diff --git a/docs/en/12-taos-sql/01-data-type.md b/docs/en/12-taos-sql/01-data-type.md index 13007d5bb1..b9d51bcfcd 100644 --- a/docs/en/12-taos-sql/01-data-type.md +++ b/docs/en/12-taos-sql/01-data-type.md @@ -42,10 +42,20 @@ In TDengine, the data types below can be used when specifying a column or tag. | 14 | NCHAR | User Defined | Multi-byte string that can include multi byte characters like Chinese characters. Each character of NCHAR type consumes 4 bytes storage. The string value should be quoted with single quotes. Literal single quote inside the string must be preceded with backslash, like `\'`. The length must be specified when defining a column or tag of NCHAR type, for example nchar(10) means it can store at most 10 characters of nchar type and will consume fixed storage of 40 bytes. An error will be reported if the string value exceeds the length defined. | | 15 | JSON | | JSON type can only be used on tags. A tag of json type is excluded with any other tags of any other type. | | 16 | VARCHAR | User-defined | Alias of BINARY | +| 16 | GEOMETRY | User-defined | Geometry | :::note +- Each row of the table cannot be longer than 48KB (64KB since version 3.0.5.0) (note that each BINARY/NCHAR/GEOMETRY column takes up an additional 2 bytes of storage space). - Only ASCII visible characters are suggested to be used in a column or tag of BINARY type. Multi-byte characters must be stored in NCHAR type. - The length of BINARY can be up to 16,374(data column is 65,517 and tag column is 16,382 since version 3.0.5.0) bytes. The string value must be quoted with single quotes. You must specify a length in bytes for a BINARY value, for example binary(20) for up to twenty single-byte characters. If the data exceeds the specified length, an error will occur. The literal single quote inside the string must be preceded with back slash like `\'` +- The maximum length of the GEOMETRY data column is 65,517 bytes, and the maximum length of the tag column is 16,382 bytes. Supports POINT, LINESTRING, and POLYGON subtypes of 2D. The following table describes the length calculation method: + + | # | **Syntax** | **MinLen** | **MaxLen** | **Growth of each point** | + |---|--------------------------------------|------------|------------|--------------------------| + | 1 | POINT(1.0 1.0) | 21 | 21 | NA | + | 2 | LINESTRING(1.0 1.0, 2.0 2.0) | 9+2*16 | 9+4094*16 | +16 | + | 3 | POLYGON((1.0 1.0, 2.0 2.0, 1.0 1.0)) | 13+3*16 | 13+4094*16 | +16 | + - Numeric values in SQL statements will be determined as integer or float type according to whether there is decimal point or whether scientific notation is used, so attention must be paid to avoid overflow. For example, 9999999999999999999 will be considered as overflow because it exceeds the upper limit of long integer, but 9999999999999999999.0 will be considered as a legal float number. ::: diff --git a/docs/en/12-taos-sql/03-table.md b/docs/en/12-taos-sql/03-table.md index 7f39fb5867..10c44848c9 100644 --- a/docs/en/12-taos-sql/03-table.md +++ b/docs/en/12-taos-sql/03-table.md @@ -9,27 +9,27 @@ You create standard tables and subtables with the `CREATE TABLE` statement. ```sql CREATE TABLE [IF NOT EXISTS] [db_name.]tb_name (create_definition [, create_definition] ...) [table_options] - + CREATE TABLE create_subtable_clause - + CREATE TABLE [IF NOT EXISTS] [db_name.]tb_name (create_definition [, create_definition] ...) [TAGS (create_definition [, create_definition] ...)] [table_options] - + create_subtable_clause: { create_subtable_clause [create_subtable_clause] ... | [IF NOT EXISTS] [db_name.]tb_name USING [db_name.]stb_name [(tag_name [, tag_name] ...)] TAGS (tag_value [, tag_value] ...) } - + create_definition: col_name column_definition - + column_definition: type_name [comment 'string_value'] - + table_options: table_option ... - + table_option: { COMMENT 'string_value' | WATERMARK duration[,duration] @@ -45,9 +45,9 @@ table_option: { 1. The first column of a table MUST be of type TIMESTAMP. It is automatically set as the primary key. 2. The maximum length of the table name is 192 bytes. -3. The maximum length of each row is 48k(64k since version 3.0.5.0) bytes, please note that the extra 2 bytes used by each BINARY/NCHAR column are also counted. +3. The maximum length of each row is 48k(64k since version 3.0.5.0) bytes, please note that the extra 2 bytes used by each BINARY/NCHAR/GEOMETRY column are also counted. 4. The name of the subtable can only consist of characters from the English alphabet, digits and underscore. Table names can't start with a digit. Table names are case insensitive. -5. The maximum length in bytes must be specified when using BINARY or NCHAR types. +5. The maximum length in bytes must be specified when using BINARY/NCHAR/GEOMETRY types. 6. Escape character "\`" can be used to avoid the conflict between table names and reserved keywords, above rules will be bypassed when using escape character on table names, but the upper limit for the name length is still valid. The table names specified using escape character are case sensitive. For example \`aBc\` and \`abc\` are different table names but `abc` and `aBc` are same table names because they are both converted to `abc` internally. Only ASCII visible characters can be used with escape character. @@ -58,7 +58,7 @@ table_option: { 3. MAX_DELAY: specifies the maximum latency for pushing computation results. The default value is 15 minutes or the value of the INTERVAL parameter, whichever is smaller. Enter a value between 0 and 15 minutes in milliseconds, seconds, or minutes. You can enter multiple values separated by commas (,). Note: Retain the default value if possible. Configuring a small MAX_DELAY may cause results to be frequently pushed, affecting storage and query performance. This parameter applies only to supertables and takes effect only when the RETENTIONS parameter has been specified for the database. 4. ROLLUP: specifies aggregate functions to roll up. Rolling up a function provides downsampled results based on multiple axes. This parameter applies only to supertables and takes effect only when the RETENTIONS parameter has been specified for the database. You can specify only one function to roll up. The rollup takes effect on all columns except TS. Enter one of the following values: avg, sum, min, max, last, or first. 5. SMA: specifies functions on which to enable small materialized aggregates (SMA). SMA is user-defined precomputation of aggregates based on data blocks. Enter one of the following values: max, min, or sum This parameter can be used with supertables and standard tables. -6. TTL: specifies the time to live (TTL) for the table. If TTL is specified when creatinga table, after the time period for which the table has been existing is over TTL, TDengine will automatically delete the table. Please be noted that the system may not delete the table at the exact moment that the TTL expires but guarantee there is such a system and finally the table will be deleted. The unit of TTL is in days. The default value is 0, i.e. never expire. +6. TTL: specifies the time to live (TTL) for the table. If TTL is specified when creatinga table, after the time period for which the table has been existing is over TTL, TDengine will automatically delete the table. Please be noted that the system may not delete the table at the exact moment that the TTL expires but guarantee there is such a system and finally the table will be deleted. The unit of TTL is in days. The default value is 0, i.e. never expire. ## Create Subtables @@ -88,7 +88,7 @@ You can create multiple subtables in a single SQL statement provided that all su ```sql ALTER TABLE [db_name.]tb_name alter_table_clause - + alter_table_clause: { alter_table_options | ADD COLUMN col_name column_type @@ -96,10 +96,10 @@ alter_table_clause: { | MODIFY COLUMN col_name column_type | RENAME COLUMN old_col_name new_col_name } - + alter_table_options: alter_table_option ... - + alter_table_option: { TTL value | COMMENT 'string_value' @@ -142,15 +142,15 @@ ALTER TABLE tb_name RENAME COLUMN old_col_name new_col_name ```sql ALTER TABLE [db_name.]tb_name alter_table_clause - + alter_table_clause: { alter_table_options | SET TAG tag_name = new_tag_value } - + alter_table_options: alter_table_option ... - + alter_table_option: { TTL value | COMMENT 'string_value' diff --git a/docs/en/12-taos-sql/16-operators.md b/docs/en/12-taos-sql/16-operators.md index 32ad4e7075..9328d1688a 100644 --- a/docs/en/12-taos-sql/16-operators.md +++ b/docs/en/12-taos-sql/16-operators.md @@ -39,7 +39,7 @@ TDengine supports the `UNION` and `UNION ALL` operations. UNION ALL collects all | 3 | \>, < | All types except BLOB, MEDIUMBLOB, and JSON | Greater than and less than | | 4 | \>=, <= | All types except BLOB, MEDIUMBLOB, and JSON | Greater than or equal to and less than or equal to | | 5 | IS [NOT] NULL | All types | Indicates whether the value is null | -| 6 | [NOT] BETWEEN AND | All types except BLOB, MEDIUMBLOB, and JSON | Closed interval comparison | +| 6 | [NOT] BETWEEN AND | All types except BLOB, MEDIUMBLOB, JSON and GEOMETRY | Closed interval comparison | | 7 | IN | All types except BLOB, MEDIUMBLOB, and JSON; the primary key (timestamp) is also not supported | Equal to any value in the list | | 8 | LIKE | BINARY, NCHAR, and VARCHAR | Wildcard match | | 9 | MATCH, NMATCH | BINARY, NCHAR, and VARCHAR | Regular expression match | diff --git a/docs/en/12-taos-sql/29-changes.md b/docs/en/12-taos-sql/29-changes.md index d668aa8345..bbb52db4d9 100644 --- a/docs/en/12-taos-sql/29-changes.md +++ b/docs/en/12-taos-sql/29-changes.md @@ -18,6 +18,7 @@ description: This document describes how TDengine SQL has changed in version 3.0 | 8 | Mixed operations | Enhanced | Mixing scalar and vector operations in queries has been enhanced and is supported in all SELECT clauses. | 9 | Tag operations | Added | Tag columns can be used in queries and clauses like data columns. | 10 | Timeline clauses and time functions in supertables | Enhanced | When PARTITION BY is not used, data in supertables is merged into a single timeline. +| 11 | GEOMETRY | Added | Geometry ## SQL Syntax diff --git a/docs/en/14-reference/12-config/index.md b/docs/en/14-reference/12-config/index.md index bf6a45735d..e4bb646965 100755 --- a/docs/en/14-reference/12-config/index.md +++ b/docs/en/14-reference/12-config/index.md @@ -166,7 +166,7 @@ Please note the `taoskeeper` needs to be installed and running to create the `lo | Attribute | Description | | ------------- | ---------------------------------------------------------------------------- | -| Applicable | Server Only | +| Applicable | Server and Client | | Meaning | Switch for allowing TDengine to collect and report service usage information | | Value Range | 0: Not allowed; 1: Allowed | | Default Value | 1 | @@ -174,7 +174,7 @@ Please note the `taoskeeper` needs to be installed and running to create the `lo | Attribute | Description | | ------------- | ---------------------------------------------------------------------------- | -| Applicable | Server Only | +| Applicable | Server and Client | | Meaning | Switch for allowing TDengine to collect and report crash related information | | Value Range | 0,1 0: Not allowed; 1: allowed | | Default Value | 1 | diff --git a/docs/zh/12-taos-sql/01-data-type.md b/docs/zh/12-taos-sql/01-data-type.md index 4a4c1d6ec6..1df07e7e7f 100644 --- a/docs/zh/12-taos-sql/01-data-type.md +++ b/docs/zh/12-taos-sql/01-data-type.md @@ -42,12 +42,21 @@ CREATE DATABASE db_name PRECISION 'ns'; | 14 | NCHAR | 自定义 | 记录包含多字节字符在内的字符串，如中文字符。每个 NCHAR 字符占用 4 字节的存储空间。字符串两端使用单引号引用，字符串内的单引号需用转义字符 `\'`。NCHAR 使用时须指定字符串大小，类型为 NCHAR(10) 的列表示此列的字符串最多存储 10 个 NCHAR 字符。如果用户字符串长度超出声明长度，将会报错。 | | 15 | JSON | | JSON 数据类型， 只有 Tag 可以是 JSON 格式 | | 16 | VARCHAR | 自定义 | BINARY 类型的别名 | +| 17 | GEOMETRY | 自定义 | 几何类型 | :::note -- 表的每行长度不能超过 48KB（从 3.0.5.0 版本开始为 64KB）（注意：每个 BINARY/NCHAR 类型的列还会额外占用 2 个字节的存储位置）。 +- 表的每行长度不能超过 48KB（从 3.0.5.0 版本开始为 64KB）（注意：每个 BINARY/NCHAR/GEOMETRY 类型的列还会额外占用 2 个字节的存储位置）。 - 虽然 BINARY 类型在底层存储上支持字节型的二进制字符，但不同编程语言对二进制数据的处理方式并不保证一致，因此建议在 BINARY 类型中只存储 ASCII 可见字符，而避免存储不可见字符。多字节的数据，例如中文字符，则需要使用 NCHAR 类型进行保存。如果强行使用 BINARY 类型保存中文字符，虽然有时也能正常读写，但并不带有字符集信息，很容易出现数据乱码甚至数据损坏等情况。 - BINARY 类型理论上最长可以有 16,374（从 3.0.5.0 版本开始，数据列为 65,517，标签列为 16,382） 字节。BINARY 仅支持字符串输入，字符串两端需使用单引号引用。使用时须指定大小，如 BINARY(20) 定义了最长为 20 个单字节字符的字符串，每个字符占 1 字节的存储空间，总共固定占用 20 字节的空间，此时如果用户字符串超出 20 字节将会报错。对于字符串内的单引号，可以用转义字符反斜线加单引号来表示，即 `\'`。 +- GEOMETRY 类型数据列为最大长度为 65,517 字节，标签列最大长度为 16,382 字节。支持 2D 的 POINT、LINESTRING 和 POLYGON 子类型数据。长度计算方式如下表所示： + + | # | **语法** | **最小长度** | **最大长度** | **每组坐标长度增长** | + |---|--------------------------------------|----------|------------|--------------| + | 1 | POINT(1.0 1.0) | 21 | 21 | 无 | + | 2 | LINESTRING(1.0 1.0, 2.0 2.0) | 9+2*16 | 9+4094*16 | +16 | + | 3 | POLYGON((1.0 1.0, 2.0 2.0, 1.0 1.0)) | 13+3*16 | 13+4094*16 | +16 | + - SQL 语句中的数值类型将依据是否存在小数点，或使用科学计数法表示，来判断数值类型是否为整型或者浮点型，因此在使用时要注意相应类型越界的情况。例如，9999999999999999999 会认为超过长整型的上边界而溢出，而 9999999999999999999.0 会被认为是有效的浮点数。 ::: diff --git a/docs/zh/12-taos-sql/03-table.md b/docs/zh/12-taos-sql/03-table.md index 2e66ac4002..f13a4ac94a 100644 --- a/docs/zh/12-taos-sql/03-table.md +++ b/docs/zh/12-taos-sql/03-table.md @@ -43,9 +43,9 @@ table_option: { 1. 表的第一个字段必须是 TIMESTAMP，并且系统自动将其设为主键； 2. 表名最大长度为 192； -3. 表的每行长度不能超过 48KB（从 3.0.5.0 版本开始为 64KB）;（注意：每个 BINARY/NCHAR 类型的列还会额外占用 2 个字节的存储位置） +3. 表的每行长度不能超过 48KB（从 3.0.5.0 版本开始为 64KB）;（注意：每个 BINARY/NCHAR/GEOMETRY 类型的列还会额外占用 2 个字节的存储位置） 4. 子表名只能由字母、数字和下划线组成，且不能以数字开头，不区分大小写 -5. 使用数据类型 binary 或 nchar，需指定其最长的字节数，如 binary(20)，表示 20 字节； +5. 使用数据类型 BINARY/NCHAR/GEOMETRY，需指定其最长的字节数，如 BINARY(20)，表示 20 字节； 6. 为了兼容支持更多形式的表名，TDengine 引入新的转义符 "\`"，可以让表名与关键词不冲突，同时不受限于上述表名称合法性约束检查。但是同样具有长度限制要求。使用转义字符以后，不再对转义字符中的内容进行大小写统一。 例如：\`aBc\` 和 \`abc\` 是不同的表名，但是 abc 和 aBc 是相同的表名。 需要注意的是转义字符中的内容必须是可打印字符。 diff --git a/docs/zh/12-taos-sql/16-operators.md b/docs/zh/12-taos-sql/16-operators.md index 48e9991799..0636121edd 100644 --- a/docs/zh/12-taos-sql/16-operators.md +++ b/docs/zh/12-taos-sql/16-operators.md @@ -39,7 +39,7 @@ TDengine 支持 `UNION ALL` 和 `UNION` 操作符。UNION ALL 将查询返回的 | 3 | \>, < | 除 BLOB、MEDIUMBLOB 和 JSON 外的所有类型 | 大于，小于 | | 4 | \>=, <= | 除 BLOB、MEDIUMBLOB 和 JSON 外的所有类型 | 大于等于，小于等于 | | 5 | IS [NOT] NULL | 所有类型 | 是否为空值 | -| 6 | [NOT] BETWEEN AND | 除 BOOL、BLOB、MEDIUMBLOB 和 JSON 外的所有类型 | 闭区间比较 | +| 6 | [NOT] BETWEEN AND | 除 BOOL、BLOB、MEDIUMBLOB、JSON 和 GEOMETRY 外的所有类型 | 闭区间比较 | | 7 | IN | 除 BLOB、MEDIUMBLOB 和 JSON 外的所有类型，且不可以为表的时间戳主键列 | 与列表内的任意值相等 | | 8 | LIKE | BINARY、NCHAR 和 VARCHAR | 通配符匹配 | | 9 | MATCH, NMATCH | BINARY、NCHAR 和 VARCHAR | 正则表达式匹配 | diff --git a/docs/zh/12-taos-sql/29-changes.md b/docs/zh/12-taos-sql/29-changes.md index 4177fa547e..2a1e5f092c 100644 --- a/docs/zh/12-taos-sql/29-changes.md +++ b/docs/zh/12-taos-sql/29-changes.md @@ -18,6 +18,7 @@ description: "TDengine 3.0 版本的语法变更说明" | 8 | 混合运算 | 增强 | 查询中的混合运算（标量运算和矢量运算混合）全面增强，SELECT的各个子句均全面支持符合语法语义的混合运算。 | 9 | 标签运算 | 新增 |在查询中，标签列可以像普通列一样参与各种运算，用于各种子句。 | 10 | 时间线子句和时间函数用于超级表查询 | 增强 |没有PARTITION BY时，超级表的数据会被合并成一条时间线。 +| 11 | GEOMETRY | 新增 | 几何类型。 ## SQL 语句变更 diff --git a/docs/zh/14-reference/12-config/index.md b/docs/zh/14-reference/12-config/index.md index 0e269e59dc..719a0e28c7 100755 --- a/docs/zh/14-reference/12-config/index.md +++ b/docs/zh/14-reference/12-config/index.md @@ -184,7 +184,7 @@ taos -C | 属性 | 说明 | | -------- | ------------------------ | -| 适用范围 | 仅服务端适用 | +| 适用范围 | 客户端和服务端都适用 | | 含义 | 是否上传 telemetry | | 取值范围 | 0,1 0: 不上传；1：上传 | | 缺省值 | 1 | @@ -193,7 +193,7 @@ taos -C | 属性 | 说明 | | -------- | ------------------------ | -| 适用范围 | 仅服务端适用 | +| 适用范围 | 客户端和服务端都适用 | | 含义 | 是否上传 crash 信息 | | 取值范围 | 0,1 0: 不上传；1：上传 | | 缺省值 | 1 | diff --git a/include/common/tmsg.h b/include/common/tmsg.h index b9365172d5..9e100be42a 100644 --- a/include/common/tmsg.h +++ b/include/common/tmsg.h @@ -1495,6 +1495,7 @@ int32_t tDeserializeSShowVariablesReq(void* buf, int32_t bufLen, SShowVariablesR typedef struct { char name[TSDB_CONFIG_OPTION_LEN + 1]; char value[TSDB_CONFIG_VALUE_LEN + 1]; + char scope[TSDB_CONFIG_SCOPE_LEN + 1]; } SVariablesInfo; typedef struct { diff --git a/include/libs/nodes/cmdnodes.h b/include/libs/nodes/cmdnodes.h index bd0b70c310..e45a2c492d 100644 --- a/include/libs/nodes/cmdnodes.h +++ b/include/libs/nodes/cmdnodes.h @@ -36,9 +36,10 @@ extern "C" { #define SHOW_CREATE_TB_RESULT_FIELD1_LEN (TSDB_TABLE_NAME_LEN + VARSTR_HEADER_SIZE) #define SHOW_CREATE_TB_RESULT_FIELD2_LEN (TSDB_MAX_ALLOWED_SQL_LEN * 3) -#define SHOW_LOCAL_VARIABLES_RESULT_COLS 2 +#define SHOW_LOCAL_VARIABLES_RESULT_COLS 3 #define SHOW_LOCAL_VARIABLES_RESULT_FIELD1_LEN (TSDB_CONFIG_OPTION_LEN + VARSTR_HEADER_SIZE) #define SHOW_LOCAL_VARIABLES_RESULT_FIELD2_LEN (TSDB_CONFIG_VALUE_LEN + VARSTR_HEADER_SIZE) +#define SHOW_LOCAL_VARIABLES_RESULT_FIELD3_LEN (TSDB_CONFIG_SCOPE_LEN + VARSTR_HEADER_SIZE) #define SHOW_ALIVE_RESULT_COLS 1 diff --git a/include/libs/nodes/plannodes.h b/include/libs/nodes/plannodes.h index 453c5d4914..821defc797 100644 --- a/include/libs/nodes/plannodes.h +++ b/include/libs/nodes/plannodes.h @@ -55,6 +55,7 @@ typedef struct SLogicNode { EGroupAction groupAction; EOrder inputTsOrder; EOrder outputTsOrder; + bool forceCreateNonBlockingOptr; // true if the operator can use non-blocking(pipeline) mode } SLogicNode; typedef enum EScanType { @@ -105,6 +106,7 @@ typedef struct SScanLogicNode { bool hasNormalCols; // neither tag column nor primary key tag column bool sortPrimaryKey; bool igLastNull; + bool groupOrderScan; } SScanLogicNode; typedef struct SJoinLogicNode { @@ -316,6 +318,7 @@ typedef struct SPhysiNode { struct SPhysiNode* pParent; SNode* pLimit; SNode* pSlimit; + bool forceCreateNonBlockingOptr; } SPhysiNode; typedef struct SScanPhysiNode { @@ -326,6 +329,7 @@ typedef struct SScanPhysiNode { uint64_t suid; int8_t tableType; SName tableName; + bool groupOrderScan; } SScanPhysiNode; typedef SScanPhysiNode STagScanPhysiNode; diff --git a/include/libs/stream/tstream.h b/include/libs/stream/tstream.h index 354f202e55..33b2646f0b 100644 --- a/include/libs/stream/tstream.h +++ b/include/libs/stream/tstream.h @@ -45,8 +45,8 @@ enum { TASK_STATUS__FAIL, TASK_STATUS__STOP, TASK_STATUS__SCAN_HISTORY, // stream task scan history data by using tsdbread in the stream scanner - TASK_STATUS__HALT, // stream task will handle all data in the input queue, and then paused - TASK_STATUS__PAUSE, + TASK_STATUS__HALT, // pause, but not be manipulated by user command + TASK_STATUS__PAUSE, // pause TASK_STATUS__CK, // stream task is in checkpoint status, no data are allowed to put into inputQ anymore TASK_STATUS__CK_READY, }; @@ -262,12 +262,13 @@ typedef struct SCheckpointInfo { } SCheckpointInfo; typedef struct SStreamStatus { - int8_t taskStatus; - int8_t downstreamReady; // downstream tasks are all ready now, if this flag is set - int8_t schedStatus; - int8_t keepTaskStatus; - bool transferState; - int8_t timerActive; // timer is active + int8_t taskStatus; + int8_t downstreamReady; // downstream tasks are all ready now, if this flag is set + int8_t schedStatus; + int8_t keepTaskStatus; + bool transferState; + int8_t timerActive; // timer is active + int8_t pauseAllowed; // allowed task status to be set to be paused } SStreamStatus; typedef struct SHistDataRange { @@ -292,15 +293,15 @@ typedef struct SDispatchMsgInfo { } SDispatchMsgInfo; typedef struct { - int8_t outputType; - int8_t outputStatus; - SStreamQueue* outputQueue; -} SSTaskOutputInfo; + int8_t type; + int8_t status; + SStreamQueue* queue; +} STaskOutputInfo; struct SStreamTask { SStreamId id; SSTaskBasicInfo info; - int8_t outputType; + STaskOutputInfo outputInfo; SDispatchMsgInfo msgInfo; SStreamStatus status; SCheckpointInfo chkInfo; @@ -308,6 +309,10 @@ struct SStreamTask { SHistDataRange dataRange; SStreamId historyTaskId; SStreamId streamTaskId; + SArray* pUpstreamEpInfoList; // SArray, // children info + int32_t nextCheckId; + SArray* checkpointInfo; // SArray + int64_t initTs; SArray* pUpstreamInfoList; // SArray, // children info SArray* pReadyMsgList; // SArray TdThreadMutex lock; // secure the operation of set task status and puting data into inputQ @@ -322,9 +327,7 @@ struct SStreamTask { }; int8_t inputStatus; - int8_t outputStatus; SStreamQueue* inputQueue; - SStreamQueue* outputQueue; // trigger int8_t triggerStatus; @@ -333,6 +336,7 @@ struct SStreamTask { void* launchTaskTimer; SMsgCb* pMsgCb; // msg handle SStreamState* pState; // state backend + SArray* pRspMsgList; // the followings attributes don't be serialized int32_t notReadyTasks; @@ -465,7 +469,9 @@ typedef struct { typedef struct { int64_t streamId; - int32_t taskId; + int32_t upstreamTaskId; + int32_t downstreamTaskId; + int32_t upstreamNodeId; int32_t childId; } SStreamScanHistoryFinishReq, SStreamTransferReq; @@ -510,6 +516,17 @@ typedef struct { int32_t tEncodeStreamCheckpointReadyMsg(SEncoder* pEncoder, const SStreamCheckpointReadyMsg* pRsp); int32_t tDecodeStreamCheckpointReadyMsg(SDecoder* pDecoder, SStreamCheckpointReadyMsg* pRsp); +typedef struct { + int64_t streamId; + int32_t upstreamTaskId; + int32_t upstreamNodeId; + int32_t downstreamId; + int32_t downstreamNode; +} SStreamCompleteHistoryMsg; + +int32_t tEncodeCompleteHistoryDataMsg(SEncoder* pEncoder, const SStreamCompleteHistoryMsg* pReq); +int32_t tDecodeCompleteHistoryDataMsg(SDecoder* pDecoder, SStreamCompleteHistoryMsg* pReq); + typedef struct { int64_t streamId; int32_t downstreamTaskId; @@ -543,7 +560,7 @@ int32_t streamProcessRunReq(SStreamTask* pTask); int32_t streamProcessDispatchMsg(SStreamTask* pTask, SStreamDispatchReq* pReq, SRpcMsg* pMsg, bool exec); int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, int32_t code); -int32_t streamProcessRetrieveReq(SStreamTask* pTask, SStreamRetrieveReq* pReq, SRpcMsg* pMsg); +int32_t streamProcessRetrieveReq(SStreamTask* pTask, SStreamRetrieveReq* pReq, SRpcMsg* pMsg); void streamTaskOpenAllUpstreamInput(SStreamTask* pTask); void streamTaskCloseUpstreamInput(SStreamTask* pTask, int32_t taskId); SStreamChildEpInfo* streamTaskGetUpstreamTaskEpInfo(SStreamTask* pTask, int32_t taskId); @@ -556,17 +573,20 @@ bool streamTaskShouldStop(const SStreamStatus* pStatus); bool streamTaskShouldPause(const SStreamStatus* pStatus); bool streamTaskIsIdle(const SStreamTask* pTask); +SStreamChildEpInfo * streamTaskGetUpstreamTaskEpInfo(SStreamTask* pTask, int32_t taskId); int32_t streamScanExec(SStreamTask* pTask, int32_t batchSz); char* createStreamTaskIdStr(int64_t streamId, int32_t taskId); // recover and fill history -void streamPrepareNdoCheckDownstream(SStreamTask* pTask); -int32_t streamTaskCheckDownstreamTasks(SStreamTask* pTask); +void streamTaskCheckDownstreamTasks(SStreamTask* pTask); +int32_t streamTaskDoCheckDownstreamTasks(SStreamTask* pTask); int32_t streamTaskLaunchScanHistory(SStreamTask* pTask); int32_t streamTaskCheckStatus(SStreamTask* pTask); +int32_t streamSendCheckRsp(const SStreamMeta* pMeta, const SStreamTaskCheckReq* pReq, SStreamTaskCheckRsp* pRsp, + SRpcHandleInfo* pRpcInfo, int32_t taskId); int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRsp); -int32_t streamCheckHistoryTaskDownstream(SStreamTask* pTask); +int32_t streamLaunchFillHistoryTask(SStreamTask* pTask); int32_t streamTaskScanHistoryDataComplete(SStreamTask* pTask); int32_t streamStartRecoverTask(SStreamTask* pTask, int8_t igUntreated); void streamHistoryTaskSetVerRangeStep2(SStreamTask* pTask); @@ -581,6 +601,12 @@ int32_t streamSetParamForScanHistory(SStreamTask* pTask); int32_t streamRestoreParam(SStreamTask* pTask); int32_t streamSetStatusNormal(SStreamTask* pTask); const char* streamGetTaskStatusStr(int32_t status); +void streamTaskPause(SStreamTask* pTask); +void streamTaskResume(SStreamTask* pTask); +void streamTaskHalt(SStreamTask* pTask); +void streamTaskResumeFromHalt(SStreamTask* pTask); +void streamTaskDisablePause(SStreamTask* pTask); +void streamTaskEnablePause(SStreamTask* pTask); // source level int32_t streamSetParamForStreamScannerStep1(SStreamTask* pTask, SVersionRange* pVerRange, STimeWindow* pWindow); @@ -592,8 +618,9 @@ int32_t streamDispatchScanHistoryFinishMsg(SStreamTask* pTask); int32_t streamDispatchTransferStateMsg(SStreamTask* pTask); // agg level -int32_t streamAggScanHistoryPrepare(SStreamTask* pTask); -int32_t streamProcessScanHistoryFinishReq(SStreamTask* pTask, int32_t childId); +int32_t streamTaskScanHistoryPrepare(SStreamTask* pTask); +int32_t streamProcessScanHistoryFinishReq(SStreamTask* pTask, SStreamScanHistoryFinishReq *pReq, SRpcHandleInfo* pRpcInfo); +int32_t streamProcessScanHistoryFinishRsp(SStreamTask* pTask); // stream task meta void streamMetaInit(); diff --git a/include/util/tarray.h b/include/util/tarray.h index a93c695370..f56c9e3a17 100644 --- a/include/util/tarray.h +++ b/include/util/tarray.h @@ -22,7 +22,7 @@ extern "C" { #endif -#define TARRAY_MIN_SIZE 8 +#define TARRAY_MIN_SIZE 4 #define TARRAY_GET_ELEM(array, index) ((void*)((char*)((array)->pData) + (index) * (array)->elemSize)) #define TARRAY_ELEM_IDX(array, ele) (POINTER_DISTANCE(ele, (array)->pData) / (array)->elemSize) @@ -138,7 +138,7 @@ size_t taosArrayGetSize(const SArray* pArray); * @param index * @param pData */ -void* taosArrayInsert(SArray* pArray, size_t index, void* pData); +void* taosArrayInsert(SArray* pArray, size_t index, const void* pData); /** * set data in array @@ -204,9 +204,9 @@ void taosArrayClearEx(SArray* pArray, void (*fp)(void*)); void* taosArrayDestroy(SArray* pArray); -void taosArrayDestroyP(SArray* pArray, FDelete fp); +void taosArrayDestroyP(SArray* pArray, FDelete fp); -void taosArrayDestroyEx(SArray* pArray, FDelete fp); +void taosArrayDestroyEx(SArray* pArray, FDelete fp); void taosArraySwap(SArray* a, SArray* b); diff --git a/include/util/tarray2.h b/include/util/tarray2.h new file mode 100644 index 0000000000..cd49e64789 --- /dev/null +++ b/include/util/tarray2.h @@ -0,0 +1,172 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "talgo.h" + +#ifndef _TD_UTIL_TARRAY2_H_ +#define _TD_UTIL_TARRAY2_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +// a: a +// e: element +// ep: element pointer +// cmp: compare function +// idx: index +// cb: callback function + +#define TARRAY2(TYPE) \ + struct { \ + int32_t size; \ + int32_t capacity; \ + TYPE *data; \ + } + +typedef void (*TArray2Cb)(void *); + +#define TARRAY2_SIZE(a) ((a)->size) +#define TARRAY2_CAPACITY(a) ((a)->capacity) +#define TARRAY2_DATA(a) ((a)->data) +#define TARRAY2_GET(a, i) ((a)->data[i]) +#define TARRAY2_GET_PTR(a, i) ((a)->data + i) +#define TARRAY2_FIRST(a) ((a)->data[0]) +#define TARRAY2_LAST(a) ((a)->data[(a)->size - 1]) +#define TARRAY2_DATA_LEN(a) ((a)->size * sizeof(((a)->data[0]))) + +static FORCE_INLINE int32_t tarray2_make_room(void *arr, int32_t expSize, int32_t eleSize) { + TARRAY2(void) *a = arr; + + int32_t capacity = (a->capacity > 0) ? (a->capacity << 1) : 32; + while (capacity < expSize) { + capacity <<= 1; + } + void *p = taosMemoryRealloc(a->data, capacity * eleSize); + if (p == NULL) return TSDB_CODE_OUT_OF_MEMORY; + a->capacity = capacity; + a->data = p; + return 0; +} + +static FORCE_INLINE int32_t tarray2InsertBatch(void *arr, int32_t idx, const void *elePtr, int32_t numEle, + int32_t eleSize) { + TARRAY2(uint8_t) *a = arr; + + int32_t ret = 0; + if (a->size + numEle > a->capacity) { + ret = tarray2_make_room(a, a->size + numEle, eleSize); + } + if (ret == 0) { + if (idx < a->size) { + memmove(a->data + (idx + numEle) * eleSize, a->data + idx * eleSize, (a->size - idx) * eleSize); + } + memcpy(a->data + idx * eleSize, elePtr, numEle * eleSize); + a->size += numEle; + } + return ret; +} + +static FORCE_INLINE void *tarray2Search(void *arr, const void *elePtr, int32_t eleSize, __compar_fn_t compar, + int32_t flag) { + TARRAY2(void) *a = arr; + return taosbsearch(elePtr, a->data, a->size, eleSize, compar, flag); +} + +static FORCE_INLINE int32_t tarray2SearchIdx(void *arr, const void *elePtr, int32_t eleSize, __compar_fn_t compar, + int32_t flag) { + TARRAY2(void) *a = arr; + void *p = taosbsearch(elePtr, a->data, a->size, eleSize, compar, flag); + if (p == NULL) { + return -1; + } else { + return (int32_t)(((uint8_t *)p - (uint8_t *)a->data) / eleSize); + } +} + +static FORCE_INLINE int32_t tarray2SortInsert(void *arr, const void *elePtr, int32_t eleSize, __compar_fn_t compar) { + TARRAY2(void) *a = arr; + int32_t idx = tarray2SearchIdx(arr, elePtr, eleSize, compar, TD_GT); + return tarray2InsertBatch(arr, idx < 0 ? a->size : idx, elePtr, 1, eleSize); +} + +#define TARRAY2_INIT_EX(a, size_, capacity_, data_) \ + do { \ + (a)->size = (size_); \ + (a)->capacity = (capacity_); \ + (a)->data = (data_); \ + } while (0) + +#define TARRAY2_INIT(a) TARRAY2_INIT_EX(a, 0, 0, NULL) + +#define TARRAY2_CLEAR(a, cb) \ + do { \ + if ((cb) && (a)->size > 0) { \ + TArray2Cb cb_ = (TArray2Cb)(cb); \ + for (int32_t i = 0; i < (a)->size; ++i) { \ + cb_((a)->data + i); \ + } \ + } \ + (a)->size = 0; \ + } while (0) + +#define TARRAY2_DESTROY(a, cb) \ + do { \ + TARRAY2_CLEAR(a, cb); \ + if ((a)->data) { \ + taosMemoryFree((a)->data); \ + (a)->data = NULL; \ + } \ + (a)->capacity = 0; \ + } while (0) + +#define TARRAY2_INSERT_PTR(a, idx, ep) tarray2InsertBatch(a, idx, ep, 1, sizeof((a)->data[0])) +#define TARRAY2_APPEND_PTR(a, ep) tarray2InsertBatch(a, (a)->size, ep, 1, sizeof((a)->data[0])) +#define TARRAY2_APPEND_BATCH(a, ep, n) tarray2InsertBatch(a, (a)->size, ep, n, sizeof((a)->data[0])) +#define TARRAY2_APPEND(a, e) TARRAY2_APPEND_PTR(a, &(e)) + +// return (TYPE *) +#define TARRAY2_SEARCH(a, ep, cmp, flag) tarray2Search(a, ep, sizeof(((a)->data[0])), (__compar_fn_t)cmp, flag) + +#define TARRAY2_SEARCH_IDX(a, ep, cmp, flag) tarray2SearchIdx(a, ep, sizeof(((a)->data[0])), (__compar_fn_t)cmp, flag) + +#define TARRAY2_SORT_INSERT(a, e, cmp) tarray2SortInsert(a, &(e), sizeof(((a)->data[0])), (__compar_fn_t)cmp) +#define TARRAY2_SORT_INSERT_P(a, ep, cmp) tarray2SortInsert(a, ep, sizeof(((a)->data[0])), (__compar_fn_t)cmp) + +#define TARRAY2_REMOVE(a, idx, cb) \ + do { \ + if ((idx) < (a)->size) { \ + if (cb) { \ + TArray2Cb cb_ = (TArray2Cb)(cb); \ + cb_((a)->data + (idx)); \ + } \ + if ((idx) < (a)->size - 1) { \ + memmove((a)->data + (idx), (a)->data + (idx) + 1, sizeof((*(a)->data)) * ((a)->size - (idx)-1)); \ + } \ + (a)->size--; \ + } \ + } while (0) + +#define TARRAY2_FOREACH(a, e) for (int32_t __i = 0; __i < (a)->size && ((e) = (a)->data[__i], 1); __i++) +#define TARRAY2_FOREACH_REVERSE(a, e) for (int32_t __i = (a)->size - 1; __i >= 0 && ((e) = (a)->data[__i], 1); __i--) +#define TARRAY2_FOREACH_PTR(a, ep) for (int32_t __i = 0; __i < (a)->size && ((ep) = &(a)->data[__i], 1); __i++) +#define TARRAY2_FOREACH_PTR_REVERSE(a, ep) \ + for (int32_t __i = (a)->size - 1; __i >= 0 && ((ep) = &(a)->data[__i], 1); __i--) + +#ifdef __cplusplus +} +#endif + +#endif /*_TD_UTIL_TARRAY2_H_*/ diff --git a/include/util/tconfig.h b/include/util/tconfig.h index ca3c5eb0eb..ed8eedb7e5 100644 --- a/include/util/tconfig.h +++ b/include/util/tconfig.h @@ -50,11 +50,17 @@ typedef enum { CFG_DTYPE_TIMEZONE } ECfgDataType; +typedef enum { + CFG_SCOPE_SERVER, + CFG_SCOPE_CLIENT, + CFG_SCOPE_BOTH +} ECfgScopeType; + typedef struct SConfigItem { ECfgSrcType stype; ECfgDataType dtype; - bool tsc; - char *name; + int8_t scope; + char *name; union { bool bval; float fval; @@ -92,20 +98,21 @@ int32_t cfgGetSize(SConfig *pCfg); SConfigItem *cfgGetItem(SConfig *pCfg, const char *name); int32_t cfgSetItem(SConfig *pCfg, const char *name, const char *value, ECfgSrcType stype); -int32_t cfgAddBool(SConfig *pCfg, const char *name, bool defaultVal, bool tsc); -int32_t cfgAddInt32(SConfig *pCfg, const char *name, int32_t defaultVal, int64_t minval, int64_t maxval, bool tsc); -int32_t cfgAddInt64(SConfig *pCfg, const char *name, int64_t defaultVal, int64_t minval, int64_t maxval, bool tsc); -int32_t cfgAddFloat(SConfig *pCfg, const char *name, float defaultVal, double minval, double maxval, bool tsc); -int32_t cfgAddString(SConfig *pCfg, const char *name, const char *defaultVal, bool tsc); -int32_t cfgAddDir(SConfig *pCfg, const char *name, const char *defaultVal, bool tsc); -int32_t cfgAddLocale(SConfig *pCfg, const char *name, const char *defaultVal); -int32_t cfgAddCharset(SConfig *pCfg, const char *name, const char *defaultVal); -int32_t cfgAddTimezone(SConfig *pCfg, const char *name, const char *defaultVal); +int32_t cfgAddBool(SConfig *pCfg, const char *name, bool defaultVal, int8_t scope); +int32_t cfgAddInt32(SConfig *pCfg, const char *name, int32_t defaultVal, int64_t minval, int64_t maxval, int8_t scope); +int32_t cfgAddInt64(SConfig *pCfg, const char *name, int64_t defaultVal, int64_t minval, int64_t maxval, int8_t scope); +int32_t cfgAddFloat(SConfig *pCfg, const char *name, float defaultVal, double minval, double maxval, int8_t scope); +int32_t cfgAddString(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope); +int32_t cfgAddDir(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope); +int32_t cfgAddLocale(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope); +int32_t cfgAddCharset(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope); +int32_t cfgAddTimezone(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope); const char *cfgStypeStr(ECfgSrcType type); const char *cfgDtypeStr(ECfgDataType type); void cfgDumpItemValue(SConfigItem *pItem, char *buf, int32_t bufSize, int32_t *pLen); +void cfgDumpItemScope(SConfigItem *pItem, char *buf, int32_t bufSize, int32_t *pLen); void cfgDumpCfg(SConfig *pCfg, bool tsc, bool dump); diff --git a/include/util/tdef.h b/include/util/tdef.h index 69b012ecea..8b27e039fc 100644 --- a/include/util/tdef.h +++ b/include/util/tdef.h @@ -191,16 +191,16 @@ typedef enum ELogicConditionType { #define TSDB_MAX_COLUMNS 4096 #define TSDB_MIN_COLUMNS 2 // PRIMARY COLUMN(timestamp) + other columns -#define TSDB_NODE_NAME_LEN 64 -#define TSDB_TABLE_NAME_LEN 193 // it is a null-terminated string -#define TSDB_TOPIC_NAME_LEN 193 // it is a null-terminated string -#define TSDB_CGROUP_LEN 193 // it is a null-terminated string -#define TSDB_OFFSET_LEN 64 // it is a null-terminated string -#define TSDB_USER_CGROUP_LEN (TSDB_USER_LEN + TSDB_CGROUP_LEN) // it is a null-terminated string -#define TSDB_STREAM_NAME_LEN 193 // it is a null-terminated string -#define TSDB_DB_NAME_LEN 65 -#define TSDB_DB_FNAME_LEN (TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN + TSDB_NAME_DELIMITER_LEN) -#define TSDB_PRIVILEDGE_CONDITION_LEN 200 +#define TSDB_NODE_NAME_LEN 64 +#define TSDB_TABLE_NAME_LEN 193 // it is a null-terminated string +#define TSDB_TOPIC_NAME_LEN 193 // it is a null-terminated string +#define TSDB_CGROUP_LEN 193 // it is a null-terminated string +#define TSDB_OFFSET_LEN 64 // it is a null-terminated string +#define TSDB_USER_CGROUP_LEN (TSDB_USER_LEN + TSDB_CGROUP_LEN) // it is a null-terminated string +#define TSDB_STREAM_NAME_LEN 193 // it is a null-terminated string +#define TSDB_DB_NAME_LEN 65 +#define TSDB_DB_FNAME_LEN (TSDB_ACCT_ID_LEN + TSDB_DB_NAME_LEN + TSDB_NAME_DELIMITER_LEN) +#define TSDB_PRIVILEDGE_CONDITION_LEN 200 #define TSDB_FUNC_NAME_LEN 65 #define TSDB_FUNC_COMMENT_LEN 1024 * 1024 @@ -249,15 +249,15 @@ typedef enum ELogicConditionType { #define TSDB_LABEL_LEN 8 #define TSDB_JOB_STATUS_LEN 32 -#define TSDB_CLUSTER_ID_LEN 40 -#define TSDB_FQDN_LEN 128 -#define TSDB_EP_LEN (TSDB_FQDN_LEN + 6) -#define TSDB_IPv4ADDR_LEN 16 -#define TSDB_FILENAME_LEN 128 -#define TSDB_SHOW_SQL_LEN 2048 +#define TSDB_CLUSTER_ID_LEN 40 +#define TSDB_FQDN_LEN 128 +#define TSDB_EP_LEN (TSDB_FQDN_LEN + 6) +#define TSDB_IPv4ADDR_LEN 16 +#define TSDB_FILENAME_LEN 128 +#define TSDB_SHOW_SQL_LEN 2048 #define TSDB_SHOW_SCHEMA_JSON_LEN TSDB_MAX_COLUMNS * 256 -#define TSDB_SLOW_QUERY_SQL_LEN 512 -#define TSDB_SHOW_SUBQUERY_LEN 1000 +#define TSDB_SLOW_QUERY_SQL_LEN 512 +#define TSDB_SHOW_SUBQUERY_LEN 1000 #define TSDB_TRANS_STAGE_LEN 12 #define TSDB_TRANS_TYPE_LEN 16 @@ -370,7 +370,7 @@ typedef enum ELogicConditionType { #define TSDB_DEFAULT_DB_SCHEMALESS TSDB_DB_SCHEMALESS_OFF #define TSDB_MIN_STT_TRIGGER 1 #define TSDB_MAX_STT_TRIGGER 16 -#define TSDB_DEFAULT_SST_TRIGGER 1 +#define TSDB_DEFAULT_SST_TRIGGER 2 #define TSDB_MIN_HASH_PREFIX (2 - TSDB_TABLE_NAME_LEN) #define TSDB_MAX_HASH_PREFIX (TSDB_TABLE_NAME_LEN - 2) #define TSDB_DEFAULT_HASH_PREFIX 0 @@ -410,10 +410,10 @@ typedef enum ELogicConditionType { #define TSDB_EXPLAIN_RESULT_ROW_SIZE (16 * 1024) #define TSDB_EXPLAIN_RESULT_COLUMN_NAME "QUERY_PLAN" -#define TSDB_MAX_FIELD_LEN 65519 // 16384:65519 -#define TSDB_MAX_BINARY_LEN TSDB_MAX_FIELD_LEN // 16384-8:65519 -#define TSDB_MAX_NCHAR_LEN TSDB_MAX_FIELD_LEN // 16384-8:65519 -#define TSDB_MAX_GEOMETRY_LEN TSDB_MAX_FIELD_LEN // 16384-8:65519 +#define TSDB_MAX_FIELD_LEN 65519 // 16384:65519 +#define TSDB_MAX_BINARY_LEN TSDB_MAX_FIELD_LEN // 16384-8:65519 +#define TSDB_MAX_NCHAR_LEN TSDB_MAX_FIELD_LEN // 16384-8:65519 +#define TSDB_MAX_GEOMETRY_LEN TSDB_MAX_FIELD_LEN // 16384-8:65519 #define PRIMARYKEY_TIMESTAMP_COL_ID 1 #define COL_REACH_END(colId, maxColId) ((colId) > (maxColId)) @@ -492,6 +492,7 @@ enum { #define TSDB_CONFIG_OPTION_LEN 32 #define TSDB_CONFIG_VALUE_LEN 64 +#define TSDB_CONFIG_SCOPE_LEN 8 #define TSDB_CONFIG_NUMBER 8 #define QUERY_ID_SIZE 20 diff --git a/include/util/tlist.h b/include/util/tlist.h index c684e90a33..0924c133b9 100644 --- a/include/util/tlist.h +++ b/include/util/tlist.h @@ -241,6 +241,54 @@ void tdListNodeGetData(SList *list, SListNode *node, void *target); void tdListInitIter(SList *list, SListIter *pIter, TD_LIST_DIRECTION_T direction); SListNode *tdListNext(SListIter *pIter); +// macros ==================================================================================== + +// q: for queue +// n: for node +// m: for member + +#define LISTD(TYPE) \ + struct { \ + TYPE *next, *prev; \ + } + +#define LISTD_NEXT(n, m) ((n)->m.next) +#define LISTD_PREV(n, m) ((n)->m.prev) +#define LISTD_INIT(q, m) (LISTD_NEXT(q, m) = LISTD_PREV(q, m) = (q)) +#define LISTD_HEAD(q, m) (LISTD_NEXT(q, m)) +#define LISTD_TAIL(q, m) (LISTD_PREV(q, m)) +#define LISTD_PREV_NEXT(n, m) (LISTD_NEXT(LISTD_PREV(n, m), m)) +#define LISTD_NEXT_PREV(n, m) (LISTD_PREV(LISTD_NEXT(n, m), m)) + +#define LISTD_INSERT_HEAD(q, n, m) \ + do { \ + LISTD_NEXT(n, m) = LISTD_NEXT(q, m); \ + LISTD_PREV(n, m) = (q); \ + LISTD_NEXT_PREV(n, m) = (n); \ + LISTD_NEXT(q, m) = (n); \ + } while (0) + +#define LISTD_INSERT_TAIL(q, n, m) \ + do { \ + LISTD_NEXT(n, m) = (q); \ + LISTD_PREV(n, m) = LISTD_PREV(q, m); \ + LISTD_PREV_NEXT(n, m) = (n); \ + LISTD_PREV(q, m) = (n); \ + } while (0) + +#define LISTD_REMOVE(n, m) \ + do { \ + LISTD_PREV_NEXT(n, m) = LISTD_NEXT(n, m); \ + LISTD_NEXT_PREV(n, m) = LISTD_PREV(n, m); \ + } while (0) + +#define LISTD_FOREACH(q, n, m) for ((n) = LISTD_HEAD(q, m); (n) != (q); (n) = LISTD_NEXT(n, m)) +#define LISTD_FOREACH_REVERSE(q, n, m) for ((n) = LISTD_TAIL(q, m); (n) != (q); (n) = LISTD_PREV(n, m)) +#define LISTD_FOREACH_SAFE(q, n, t, m) \ + for ((n) = LISTD_HEAD(q, m), (t) = LISTD_NEXT(n, m); (n) != (q); (n) = (t), (t) = LISTD_NEXT(n, m)) +#define LISTD_FOREACH_REVERSE_SAFE(q, n, t, m) \ + for ((n) = LISTD_TAIL(q, m), (t) = LISTD_PREV(n, m); (n) != (q); (n) = (t), (t) = LISTD_PREV(n, m)) + #ifdef __cplusplus } #endif diff --git a/include/util/trbtree.h b/include/util/trbtree.h index e226419440..8353a91f0a 100644 --- a/include/util/trbtree.h +++ b/include/util/trbtree.h @@ -39,7 +39,7 @@ void tRBTreeDrop(SRBTree *pTree, SRBTreeNode *z); SRBTreeNode *tRBTreeDropByKey(SRBTree *pTree, void *pKey); SRBTreeNode *tRBTreeDropMin(SRBTree *pTree); SRBTreeNode *tRBTreeDropMax(SRBTree *pTree); -SRBTreeNode *tRBTreeGet(SRBTree *pTree, const SRBTreeNode *pKeyNode); +SRBTreeNode *tRBTreeGet(const SRBTree *pTree, const SRBTreeNode *pKeyNode); // SRBTreeIter ============================================= #define tRBTreeIterCreate(tree, ascend) \ @@ -67,9 +67,9 @@ struct SRBTree { }; struct SRBTreeIter { - int8_t asc; - SRBTree *pTree; - SRBTreeNode *pNode; + int8_t asc; + const SRBTree *pTree; + SRBTreeNode *pNode; }; #ifdef __cplusplus diff --git a/include/util/tutil.h b/include/util/tutil.h index 7a59aa170a..a2cfa4cfe5 100644 --- a/include/util/tutil.h +++ b/include/util/tutil.h @@ -29,7 +29,7 @@ extern "C" { int32_t strdequote(char *src); size_t strtrim(char *src); char *strnchr(const char *haystack, char needle, int32_t len, bool skipquote); -TdUcs4* wcsnchr(const TdUcs4* haystack, TdUcs4 needle, size_t len); +TdUcs4 *wcsnchr(const TdUcs4 *haystack, TdUcs4 needle, size_t len); char **strsplit(char *src, const char *delim, int32_t *num); char *strtolower(char *dst, const char *src); @@ -37,11 +37,11 @@ char *strntolower(char *dst, const char *src, int32_t n); char *strntolower_s(char *dst, const char *src, int32_t n); int64_t strnatoi(char *num, int32_t len); -size_t tstrncspn(const char *str, size_t ssize, const char *reject, size_t rsize); -size_t twcsncspn(const TdUcs4 *wcs, size_t size, const TdUcs4 *reject, size_t rsize); +size_t tstrncspn(const char *str, size_t ssize, const char *reject, size_t rsize); +size_t twcsncspn(const TdUcs4 *wcs, size_t size, const TdUcs4 *reject, size_t rsize); -char *strbetween(char *string, char *begin, char *end); -char *paGetToken(char *src, char **token, int32_t *tokenLen); +char *strbetween(char *string, char *begin, char *end); +char *paGetToken(char *src, char **token, int32_t *tokenLen); int32_t taosByteArrayToHexStr(char bytes[], int32_t len, char hexstr[]); int32_t taosHexStrToByteArray(char hexstr[], char bytes[]); @@ -81,12 +81,13 @@ static FORCE_INLINE void taosEncryptPass_c(uint8_t *inBuf, size_t len, char *tar static FORCE_INLINE int32_t taosGetTbHashVal(const char *tbname, int32_t tblen, int32_t method, int32_t prefix, int32_t suffix) { - if ((prefix == 0 && suffix == 0) || (tblen <= (prefix + suffix)) || (tblen <= -1 * (prefix + suffix)) || prefix * suffix < 0) { + if ((prefix == 0 && suffix == 0) || (tblen <= (prefix + suffix)) || (tblen <= -1 * (prefix + suffix)) || + prefix * suffix < 0) { return MurmurHash3_32(tbname, tblen); } else if (prefix > 0 || suffix > 0) { return MurmurHash3_32(tbname + prefix, tblen - prefix - suffix); } else { - char tbName[TSDB_TABLE_FNAME_LEN]; + char tbName[TSDB_TABLE_FNAME_LEN]; int32_t offset = 0; if (prefix < 0) { offset = -1 * prefix; @@ -94,20 +95,33 @@ static FORCE_INLINE int32_t taosGetTbHashVal(const char *tbname, int32_t tblen, } if (suffix < 0) { strncpy(tbName + offset, tbname + tblen + suffix, -1 * suffix); - offset += -1 *suffix; + offset += -1 * suffix; } return MurmurHash3_32(tbName, offset); } } #define TSDB_CHECK_CODE(CODE, LINO, LABEL) \ - if (CODE) { \ - LINO = __LINE__; \ - goto LABEL; \ + do { \ + if ((CODE)) { \ + LINO = __LINE__; \ + goto LABEL; \ + } \ + } while (0) + +#define TSDB_CHECK_NULL(ptr, CODE, LINO, LABEL, ERRNO) \ + if ((ptr) == NULL) { \ + (CODE) = (ERRNO); \ + (LINO) = __LINE__; \ + goto LABEL; \ } +#define ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0])) + #define VND_CHECK_CODE(CODE, LINO, LABEL) TSDB_CHECK_CODE(CODE, LINO, LABEL) +#define TCONTAINER_OF(ptr, type, member) ((type *)((char *)(ptr)-offsetof(type, member))) + #ifdef __cplusplus } #endif diff --git a/source/client/inc/clientInt.h b/source/client/inc/clientInt.h index 736582dff2..aa7caaaba3 100644 --- a/source/client/inc/clientInt.h +++ b/source/client/inc/clientInt.h @@ -46,9 +46,10 @@ enum { RES_TYPE__TMQ_METADATA, }; -#define SHOW_VARIABLES_RESULT_COLS 2 +#define SHOW_VARIABLES_RESULT_COLS 3 #define SHOW_VARIABLES_RESULT_FIELD1_LEN (TSDB_CONFIG_OPTION_LEN + VARSTR_HEADER_SIZE) #define SHOW_VARIABLES_RESULT_FIELD2_LEN (TSDB_CONFIG_VALUE_LEN + VARSTR_HEADER_SIZE) +#define SHOW_VARIABLES_RESULT_FIELD3_LEN (TSDB_CONFIG_SCOPE_LEN + VARSTR_HEADER_SIZE) #define TD_RES_QUERY(res) (*(int8_t*)res == RES_TYPE__QUERY) #define TD_RES_TMQ(res) (*(int8_t*)res == RES_TYPE__TMQ) diff --git a/source/client/src/clientMsgHandler.c b/source/client/src/clientMsgHandler.c index 9ab618cf3a..f1263d0341 100644 --- a/source/client/src/clientMsgHandler.c +++ b/source/client/src/clientMsgHandler.c @@ -428,13 +428,16 @@ static int32_t buildShowVariablesBlock(SArray* pVars, SSDataBlock** block) { SColumnInfoData infoData = {0}; infoData.info.type = TSDB_DATA_TYPE_VARCHAR; infoData.info.bytes = SHOW_VARIABLES_RESULT_FIELD1_LEN; - taosArrayPush(pBlock->pDataBlock, &infoData); infoData.info.type = TSDB_DATA_TYPE_VARCHAR; infoData.info.bytes = SHOW_VARIABLES_RESULT_FIELD2_LEN; taosArrayPush(pBlock->pDataBlock, &infoData); + infoData.info.type = TSDB_DATA_TYPE_VARCHAR; + infoData.info.bytes = SHOW_VARIABLES_RESULT_FIELD3_LEN; + taosArrayPush(pBlock->pDataBlock, &infoData); + int32_t numOfCfg = taosArrayGetSize(pVars); blockDataEnsureCapacity(pBlock, numOfCfg); @@ -450,6 +453,11 @@ static int32_t buildShowVariablesBlock(SArray* pVars, SSDataBlock** block) { STR_WITH_MAXSIZE_TO_VARSTR(value, pInfo->value, TSDB_CONFIG_VALUE_LEN + VARSTR_HEADER_SIZE); pColInfo = taosArrayGet(pBlock->pDataBlock, c++); colDataSetVal(pColInfo, i, value, false); + + char scope[TSDB_CONFIG_SCOPE_LEN + VARSTR_HEADER_SIZE] = {0}; + STR_WITH_MAXSIZE_TO_VARSTR(scope, pInfo->scope, TSDB_CONFIG_SCOPE_LEN + VARSTR_HEADER_SIZE); + pColInfo = taosArrayGet(pBlock->pDataBlock, c++); + colDataSetVal(pColInfo, i, scope, false); } pBlock->info.rows = numOfCfg; diff --git a/source/common/src/systable.c b/source/common/src/systable.c index 6fdc74f692..8fd235b9b3 100644 --- a/source/common/src/systable.c +++ b/source/common/src/systable.c @@ -271,6 +271,7 @@ static const SSysDbTableSchema variablesSchema[] = { {.name = "dnode_id", .bytes = 4, .type = TSDB_DATA_TYPE_INT}, {.name = "name", .bytes = TSDB_CONFIG_OPTION_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true}, {.name = "value", .bytes = TSDB_CONFIG_VALUE_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true}, + {.name = "scope", .bytes = TSDB_CONFIG_SCOPE_LEN + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR, .sysInfo = true}, }; static const SSysDbTableSchema topicSchema[] = { diff --git a/source/common/src/tdatablock.c b/source/common/src/tdatablock.c index b2f03fa7ba..887a110831 100644 --- a/source/common/src/tdatablock.c +++ b/source/common/src/tdatablock.c @@ -632,7 +632,10 @@ int32_t blockDataToBuf(char* buf, const SSDataBlock* pBlock) { pStart += colSize; } } else { - memcpy(pStart, pCol->pData, dataSize); + if (dataSize != 0) { + // ubsan reports error if pCol->pData==NULL && dataSize==0 + memcpy(pStart, pCol->pData, dataSize); + } pStart += dataSize; } } @@ -684,8 +687,10 @@ int32_t blockDataFromBuf(SSDataBlock* pBlock, const char* buf) { return TSDB_CODE_FAILED; } } - - memcpy(pCol->pData, pStart, colLength); + if (colLength != 0) { + // ubsan reports error if colLength==0 && pCol->pData == 0 + memcpy(pCol->pData, pStart, colLength); + } pStart += colLength; } diff --git a/source/common/src/tdataformat.c b/source/common/src/tdataformat.c index 0b12177754..7c6939635a 100644 --- a/source/common/src/tdataformat.c +++ b/source/common/src/tdataformat.c @@ -2245,15 +2245,18 @@ static int32_t tColDataUpdateValue72(SColData *pColData, uint8_t *pData, uint32_ } return 0; } +static FORCE_INLINE int32_t tColDataUpdateNothing(SColData *pColData, uint8_t *pData, uint32_t nData, bool forward) { + return 0; +} static int32_t (*tColDataUpdateValueImpl[8][3])(SColData *pColData, uint8_t *pData, uint32_t nData, bool forward) = { - {NULL, NULL, NULL}, // 0 - {tColDataUpdateValue10, NULL, tColDataUpdateValue12}, // HAS_NONE - {tColDataUpdateValue20, NULL, NULL}, // HAS_NULL - {tColDataUpdateValue30, NULL, tColDataUpdateValue32}, // HAS_NULL|HAS_NONE - {tColDataUpdateValue40, NULL, tColDataUpdateValue42}, // HAS_VALUE - {tColDataUpdateValue50, NULL, tColDataUpdateValue52}, // HAS_VALUE|HAS_NONE - {tColDataUpdateValue60, NULL, tColDataUpdateValue62}, // HAS_VALUE|HAS_NULL - {tColDataUpdateValue70, NULL, tColDataUpdateValue72}, // HAS_VALUE|HAS_NULL|HAS_NONE + {NULL, NULL, NULL}, // 0 + {tColDataUpdateValue10, tColDataUpdateNothing, tColDataUpdateValue12}, // HAS_NONE + {tColDataUpdateValue20, tColDataUpdateNothing, tColDataUpdateNothing}, // HAS_NULL + {tColDataUpdateValue30, tColDataUpdateNothing, tColDataUpdateValue32}, // HAS_NULL|HAS_NONE + {tColDataUpdateValue40, tColDataUpdateNothing, tColDataUpdateValue42}, // HAS_VALUE + {tColDataUpdateValue50, tColDataUpdateNothing, tColDataUpdateValue52}, // HAS_VALUE|HAS_NONE + {tColDataUpdateValue60, tColDataUpdateNothing, tColDataUpdateValue62}, // HAS_VALUE|HAS_NULL + {tColDataUpdateValue70, tColDataUpdateNothing, tColDataUpdateValue72}, // HAS_VALUE|HAS_NULL|HAS_NONE // VALUE NONE NULL }; diff --git a/source/common/src/tglobal.c b/source/common/src/tglobal.c index a4abc898a5..073558c73e 100644 --- a/source/common/src/tglobal.c +++ b/source/common/src/tglobal.c @@ -225,6 +225,7 @@ bool tsDisableStream = false; int64_t tsStreamBufferSize = 128 * 1024 * 1024; bool tsFilterScalarMode = false; int32_t tsKeepTimeOffset = 0; // latency of data migration +int32_t tsCheckpointInterval = 20; #ifndef _STORAGE int32_t taosSetTfsCfg(SConfig *pCfg) { @@ -298,38 +299,38 @@ static int32_t taosLoadCfg(SConfig *pCfg, const char **envCmd, const char *input } int32_t taosAddClientLogCfg(SConfig *pCfg) { - if (cfgAddDir(pCfg, "configDir", configDir, 1) != 0) return -1; - if (cfgAddDir(pCfg, "scriptDir", configDir, 1) != 0) return -1; - if (cfgAddDir(pCfg, "logDir", tsLogDir, 1) != 0) return -1; - if (cfgAddFloat(pCfg, "minimalLogDirGB", 1.0f, 0.001f, 10000000, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "numOfLogLines", tsNumOfLogLines, 1000, 2000000000, 1) != 0) return -1; - if (cfgAddBool(pCfg, "asyncLog", tsAsyncLog, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "logKeepDays", 0, -365000, 365000, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "debugFlag", 0, 0, 255, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "simDebugFlag", 143, 0, 255, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "tmrDebugFlag", tmrDebugFlag, 0, 255, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "uDebugFlag", uDebugFlag, 0, 255, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "rpcDebugFlag", rpcDebugFlag, 0, 255, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "jniDebugFlag", jniDebugFlag, 0, 255, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "qDebugFlag", qDebugFlag, 0, 255, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "cDebugFlag", cDebugFlag, 0, 255, 1) != 0) return -1; + if (cfgAddDir(pCfg, "configDir", configDir, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddDir(pCfg, "scriptDir", configDir, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddDir(pCfg, "logDir", tsLogDir, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddFloat(pCfg, "minimalLogDirGB", 1.0f, 0.001f, 10000000, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfLogLines", tsNumOfLogLines, 1000, 2000000000, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "asyncLog", tsAsyncLog, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "logKeepDays", 0, -365000, 365000, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "debugFlag", 0, 0, 255, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "simDebugFlag", 143, 0, 255, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "tmrDebugFlag", tmrDebugFlag, 0, 255, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "uDebugFlag", uDebugFlag, 0, 255, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "rpcDebugFlag", rpcDebugFlag, 0, 255, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "jniDebugFlag", jniDebugFlag, 0, 255, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "qDebugFlag", qDebugFlag, 0, 255, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "cDebugFlag", cDebugFlag, 0, 255, CFG_SCOPE_CLIENT) != 0) return -1; return 0; } static int32_t taosAddServerLogCfg(SConfig *pCfg) { - if (cfgAddInt32(pCfg, "dDebugFlag", dDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "vDebugFlag", vDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "mDebugFlag", mDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "wDebugFlag", wDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "sDebugFlag", sDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "tsdbDebugFlag", tsdbDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "tqDebugFlag", tqDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "fsDebugFlag", fsDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "udfDebugFlag", udfDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "smaDebugFlag", smaDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "idxDebugFlag", idxDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "tdbDebugFlag", tdbDebugFlag, 0, 255, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "metaDebugFlag", metaDebugFlag, 0, 255, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "dDebugFlag", dDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "vDebugFlag", vDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "mDebugFlag", mDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "wDebugFlag", wDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "sDebugFlag", sDebugFlag, 0, 255, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "tsdbDebugFlag", tsdbDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "tqDebugFlag", tqDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "fsDebugFlag", fsDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "udfDebugFlag", udfDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "smaDebugFlag", smaDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "idxDebugFlag", idxDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "tdbDebugFlag", tdbDebugFlag, 0, 255, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "metaDebugFlag", metaDebugFlag, 0, 255, 0) != CFG_SCOPE_SERVER) return -1; return 0; } @@ -340,53 +341,52 @@ static int32_t taosAddClientCfg(SConfig *pCfg) { strcpy(defaultFqdn, "localhost"); } - if (cfgAddString(pCfg, "firstEp", "", 1) != 0) return -1; - if (cfgAddString(pCfg, "secondEp", "", 1) != 0) return -1; - if (cfgAddString(pCfg, "fqdn", defaultFqdn, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "serverPort", defaultServerPort, 1, 65056, 1) != 0) return -1; - if (cfgAddDir(pCfg, "tempDir", tsTempDir, 1) != 0) return -1; - if (cfgAddFloat(pCfg, "minimalTmpDirGB", 1.0f, 0.001f, 10000000, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "shellActivityTimer", tsShellActivityTimer, 1, 120, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "compressMsgSize", tsCompressMsgSize, -1, 100000000, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "compressColData", tsCompressColData, -1, 100000000, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "queryPolicy", tsQueryPolicy, 1, 4, 1) != 0) return -1; - if (cfgAddBool(pCfg, "enableQueryHb", tsEnableQueryHb, false) != 0) return -1; - if (cfgAddBool(pCfg, "enableScience", tsEnableScience, false) != 0) return -1; - if (cfgAddInt32(pCfg, "querySmaOptimize", tsQuerySmaOptimize, 0, 1, 1) != 0) return -1; - if (cfgAddBool(pCfg, "queryPlannerTrace", tsQueryPlannerTrace, true) != 0) return -1; - if (cfgAddInt32(pCfg, "queryNodeChunkSize", tsQueryNodeChunkSize, 1024, 128 * 1024, true) != 0) return -1; - if (cfgAddBool(pCfg, "queryUseNodeAllocator", tsQueryUseNodeAllocator, true) != 0) return -1; - if (cfgAddBool(pCfg, "keepColumnName", tsKeepColumnName, true) != 0) return -1; - if (cfgAddString(pCfg, "smlChildTableName", "", 1) != 0) return -1; - if (cfgAddString(pCfg, "smlTagName", tsSmlTagName, 1) != 0) return -1; - // if (cfgAddBool(pCfg, "smlDataFormat", tsSmlDataFormat, 1) != 0) return -1; - // if (cfgAddInt32(pCfg, "smlBatchSize", tsSmlBatchSize, 1, INT32_MAX, true) != 0) return -1; - if (cfgAddInt32(pCfg, "maxInsertBatchRows", tsMaxInsertBatchRows, 1, INT32_MAX, true) != 0) return -1; - if (cfgAddInt32(pCfg, "maxRetryWaitTime", tsMaxRetryWaitTime, 0, 86400000, 0) != 0) return -1; - if (cfgAddBool(pCfg, "useAdapter", tsUseAdapter, true) != 0) return -1; - if (cfgAddBool(pCfg, "crashReporting", tsEnableCrashReport, true) != 0) return -1; - if (cfgAddInt64(pCfg, "queryMaxConcurrentTables", tsQueryMaxConcurrentTables, INT64_MIN, INT64_MAX, 1) != 0) - return -1; - if (cfgAddInt32(pCfg, "metaCacheMaxSize", tsMetaCacheMaxSize, -1, INT32_MAX, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "slowLogThreshold", tsSlowLogThreshold, 0, INT32_MAX, true) != 0) return -1; - if (cfgAddString(pCfg, "slowLogScope", "", true) != 0) return -1; + if (cfgAddString(pCfg, "firstEp", "", CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "secondEp", "", CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "fqdn", defaultFqdn, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "serverPort", defaultServerPort, 1, 65056, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddDir(pCfg, "tempDir", tsTempDir, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddFloat(pCfg, "minimalTmpDirGB", 1.0f, 0.001f, 10000000, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "shellActivityTimer", tsShellActivityTimer, 1, 120, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "compressMsgSize", tsCompressMsgSize, -1, 100000000, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "compressColData", tsCompressColData, -1, 100000000, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "queryPolicy", tsQueryPolicy, 1, 4, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddBool(pCfg, "enableQueryHb", tsEnableQueryHb, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddBool(pCfg, "enableScience", tsEnableScience, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "querySmaOptimize", tsQuerySmaOptimize, 0, 1, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddBool(pCfg, "queryPlannerTrace", tsQueryPlannerTrace, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "queryNodeChunkSize", tsQueryNodeChunkSize, 1024, 128 * 1024, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddBool(pCfg, "queryUseNodeAllocator", tsQueryUseNodeAllocator, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddBool(pCfg, "keepColumnName", tsKeepColumnName, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddString(pCfg, "smlChildTableName", "", CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddString(pCfg, "smlTagName", tsSmlTagName, CFG_SCOPE_CLIENT) != 0) return -1; + // if (cfgAddBool(pCfg, "smlDataFormat", tsSmlDataFormat, CFG_SCOPE_CLIENT) != 0) return -1; + // if (cfgAddInt32(pCfg, "smlBatchSize", tsSmlBatchSize, 1, INT32_MAX, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "maxInsertBatchRows", tsMaxInsertBatchRows, 1, INT32_MAX, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "maxRetryWaitTime", tsMaxRetryWaitTime, 0, 86400000, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "useAdapter", tsUseAdapter, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddBool(pCfg, "crashReporting", tsEnableCrashReport, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt64(pCfg, "queryMaxConcurrentTables", tsQueryMaxConcurrentTables, INT64_MIN, INT64_MAX, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "metaCacheMaxSize", tsMetaCacheMaxSize, -1, INT32_MAX, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "slowLogThreshold", tsSlowLogThreshold, 0, INT32_MAX, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddString(pCfg, "slowLogScope", "", CFG_SCOPE_CLIENT) != 0) return -1; tsNumOfRpcThreads = tsNumOfCores / 2; tsNumOfRpcThreads = TRANGE(tsNumOfRpcThreads, 2, TSDB_MAX_RPC_THREADS); - if (cfgAddInt32(pCfg, "numOfRpcThreads", tsNumOfRpcThreads, 1, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfRpcThreads", tsNumOfRpcThreads, 1, 1024, CFG_SCOPE_BOTH) != 0) return -1; tsNumOfRpcSessions = TRANGE(tsNumOfRpcSessions, 100, 100000); - if (cfgAddInt32(pCfg, "numOfRpcSessions", tsNumOfRpcSessions, 1, 100000, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfRpcSessions", tsNumOfRpcSessions, 1, 100000, CFG_SCOPE_BOTH) != 0) return -1; tsTimeToGetAvailableConn = TRANGE(tsTimeToGetAvailableConn, 20, 10000000); - if (cfgAddInt32(pCfg, "timeToGetAvailableConn", tsTimeToGetAvailableConn, 20, 1000000, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "timeToGetAvailableConn", tsTimeToGetAvailableConn, 20, 1000000, CFG_SCOPE_BOTH) != 0) return -1; tsNumOfTaskQueueThreads = tsNumOfCores / 2; tsNumOfTaskQueueThreads = TMAX(tsNumOfTaskQueueThreads, 4); if (tsNumOfTaskQueueThreads >= 10) { tsNumOfTaskQueueThreads = 10; } - if (cfgAddInt32(pCfg, "numOfTaskQueueThreads", tsNumOfTaskQueueThreads, 4, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfTaskQueueThreads", tsNumOfTaskQueueThreads, 4, 1024, CFG_SCOPE_CLIENT) != 0) return -1; return 0; } @@ -394,92 +394,92 @@ static int32_t taosAddClientCfg(SConfig *pCfg) { static int32_t taosAddSystemCfg(SConfig *pCfg) { SysNameInfo info = taosGetSysNameInfo(); - if (cfgAddTimezone(pCfg, "timezone", tsTimezoneStr) != 0) return -1; - if (cfgAddLocale(pCfg, "locale", tsLocale) != 0) return -1; - if (cfgAddCharset(pCfg, "charset", tsCharset) != 0) return -1; - if (cfgAddBool(pCfg, "assert", 1, 1) != 0) return -1; - if (cfgAddBool(pCfg, "enableCoreFile", 1, 1) != 0) return -1; - if (cfgAddFloat(pCfg, "numOfCores", tsNumOfCores, 1, 100000, 1) != 0) return -1; + if (cfgAddTimezone(pCfg, "timezone", tsTimezoneStr, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddLocale(pCfg, "locale", tsLocale, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddCharset(pCfg, "charset", tsCharset, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "assert", 1, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "enableCoreFile", 1, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddFloat(pCfg, "numOfCores", tsNumOfCores, 1, 100000, CFG_SCOPE_BOTH) != 0) return -1; - if (cfgAddBool(pCfg, "SSE42", tsSSE42Enable, 0) != 0) return -1; - if (cfgAddBool(pCfg, "AVX", tsAVXEnable, 0) != 0) return -1; - if (cfgAddBool(pCfg, "AVX2", tsAVX2Enable, 0) != 0) return -1; - if (cfgAddBool(pCfg, "FMA", tsFMAEnable, 0) != 0) return -1; - if (cfgAddBool(pCfg, "SIMD-builtins", tsSIMDBuiltins, 0) != 0) return -1; - if (cfgAddBool(pCfg, "tagFilterCache", tsTagFilterCache, 0) != 0) return -1; + if (cfgAddBool(pCfg, "SSE42", tsSSE42Enable, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "AVX", tsAVXEnable, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "AVX2", tsAVX2Enable, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "FMA", tsFMAEnable, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "SIMD-builtins", tsSIMDBuiltins, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "tagFilterCache", tsTagFilterCache, CFG_SCOPE_BOTH) != 0) return -1; - if (cfgAddInt64(pCfg, "openMax", tsOpenMax, 0, INT64_MAX, 1) != 0) return -1; + if (cfgAddInt64(pCfg, "openMax", tsOpenMax, 0, INT64_MAX, CFG_SCOPE_BOTH) != 0) return -1; #if !defined(_ALPINE) - if (cfgAddInt64(pCfg, "streamMax", tsStreamMax, 0, INT64_MAX, 1) != 0) return -1; + if (cfgAddInt64(pCfg, "streamMax", tsStreamMax, 0, INT64_MAX, CFG_SCOPE_BOTH) != 0) return -1; #endif - if (cfgAddInt32(pCfg, "pageSizeKB", tsPageSizeKB, 0, INT64_MAX, 1) != 0) return -1; - if (cfgAddInt64(pCfg, "totalMemoryKB", tsTotalMemoryKB, 0, INT64_MAX, 1) != 0) return -1; - if (cfgAddString(pCfg, "os sysname", info.sysname, 1) != 0) return -1; - if (cfgAddString(pCfg, "os nodename", info.nodename, 1) != 0) return -1; - if (cfgAddString(pCfg, "os release", info.release, 1) != 0) return -1; - if (cfgAddString(pCfg, "os version", info.version, 1) != 0) return -1; - if (cfgAddString(pCfg, "os machine", info.machine, 1) != 0) return -1; + if (cfgAddInt32(pCfg, "pageSizeKB", tsPageSizeKB, 0, INT64_MAX, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt64(pCfg, "totalMemoryKB", tsTotalMemoryKB, 0, INT64_MAX, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "os sysname", info.sysname, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "os nodename", info.nodename, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "os release", info.release, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "os version", info.version, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "os machine", info.machine, CFG_SCOPE_BOTH) != 0) return -1; - if (cfgAddString(pCfg, "version", version, 1) != 0) return -1; - if (cfgAddString(pCfg, "compatible_version", compatible_version, 1) != 0) return -1; - if (cfgAddString(pCfg, "gitinfo", gitinfo, 1) != 0) return -1; - if (cfgAddString(pCfg, "buildinfo", buildinfo, 1) != 0) return -1; + if (cfgAddString(pCfg, "version", version, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "compatible_version", compatible_version, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "gitinfo", gitinfo, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "buildinfo", buildinfo, CFG_SCOPE_BOTH) != 0) return -1; return 0; } static int32_t taosAddServerCfg(SConfig *pCfg) { - if (cfgAddDir(pCfg, "dataDir", tsDataDir, 0) != 0) return -1; - if (cfgAddFloat(pCfg, "minimalDataDirGB", 2.0f, 0.001f, 10000000, 0) != 0) return -1; + if (cfgAddDir(pCfg, "dataDir", tsDataDir, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddFloat(pCfg, "minimalDataDirGB", 2.0f, 0.001f, 10000000, CFG_SCOPE_SERVER) != 0) return -1; tsNumOfSupportVnodes = tsNumOfCores * 2; tsNumOfSupportVnodes = TMAX(tsNumOfSupportVnodes, 2); - if (cfgAddInt32(pCfg, "supportVnodes", tsNumOfSupportVnodes, 0, 4096, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "supportVnodes", tsNumOfSupportVnodes, 0, 4096, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddInt32(pCfg, "maxShellConns", tsMaxShellConns, 10, 50000000, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "statusInterval", tsStatusInterval, 1, 30, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "minSlidingTime", tsMinSlidingTime, 1, 1000000, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "minIntervalTime", tsMinIntervalTime, 1, 1000000, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "maxNumOfDistinctRes", tsMaxNumOfDistinctResults, 10 * 10000, 10000 * 10000, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "countAlwaysReturnValue", tsCountAlwaysReturnValue, 0, 1, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "queryBufferSize", tsQueryBufferSize, -1, 500000000000, 0) != 0) return -1; - if (cfgAddBool(pCfg, "printAuth", tsPrintAuth, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "queryRspPolicy", tsQueryRspPolicy, 0, 1, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "maxShellConns", tsMaxShellConns, 10, 50000000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "statusInterval", tsStatusInterval, 1, 30, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "minSlidingTime", tsMinSlidingTime, 1, 1000000, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "minIntervalTime", tsMinIntervalTime, 1, 1000000, CFG_SCOPE_CLIENT) != 0) return -1; + if (cfgAddInt32(pCfg, "maxNumOfDistinctRes", tsMaxNumOfDistinctResults, 10 * 10000, 10000 * 10000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "countAlwaysReturnValue", tsCountAlwaysReturnValue, 0, 1, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "queryBufferSize", tsQueryBufferSize, -1, 500000000000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddBool(pCfg, "printAuth", tsPrintAuth, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "queryRspPolicy", tsQueryRspPolicy, 0, 1, CFG_SCOPE_SERVER) != 0) return -1; tsNumOfRpcThreads = tsNumOfCores / 2; tsNumOfRpcThreads = TRANGE(tsNumOfRpcThreads, 2, TSDB_MAX_RPC_THREADS); - if (cfgAddInt32(pCfg, "numOfRpcThreads", tsNumOfRpcThreads, 1, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfRpcThreads", tsNumOfRpcThreads, 1, 1024, CFG_SCOPE_BOTH) != 0) return -1; tsNumOfRpcSessions = TRANGE(tsNumOfRpcSessions, 100, 10000); - if (cfgAddInt32(pCfg, "numOfRpcSessions", tsNumOfRpcSessions, 1, 100000, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfRpcSessions", tsNumOfRpcSessions, 1, 100000, CFG_SCOPE_BOTH) != 0) return -1; tsTimeToGetAvailableConn = TRANGE(tsTimeToGetAvailableConn, 20, 1000000); - if (cfgAddInt32(pCfg, "timeToGetAvailableConn", tsNumOfRpcSessions, 20, 1000000, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "timeToGetAvailableConn", tsNumOfRpcSessions, 20, 1000000, CFG_SCOPE_BOTH) != 0) return -1; tsNumOfCommitThreads = tsNumOfCores / 2; tsNumOfCommitThreads = TRANGE(tsNumOfCommitThreads, 2, 4); - if (cfgAddInt32(pCfg, "numOfCommitThreads", tsNumOfCommitThreads, 1, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfCommitThreads", tsNumOfCommitThreads, 1, 1024, CFG_SCOPE_SERVER) != 0) return -1; tsNumOfMnodeReadThreads = tsNumOfCores / 8; tsNumOfMnodeReadThreads = TRANGE(tsNumOfMnodeReadThreads, 1, 4); - if (cfgAddInt32(pCfg, "numOfMnodeReadThreads", tsNumOfMnodeReadThreads, 1, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfMnodeReadThreads", tsNumOfMnodeReadThreads, 1, 1024, CFG_SCOPE_SERVER) != 0) return -1; tsNumOfVnodeQueryThreads = tsNumOfCores * 2; tsNumOfVnodeQueryThreads = TMAX(tsNumOfVnodeQueryThreads, 4); - if (cfgAddInt32(pCfg, "numOfVnodeQueryThreads", tsNumOfVnodeQueryThreads, 4, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfVnodeQueryThreads", tsNumOfVnodeQueryThreads, 4, 1024, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddFloat(pCfg, "ratioOfVnodeStreamThreads", tsRatioOfVnodeStreamThreads, 0.01, 100, 0) != 0) return -1; + if (cfgAddFloat(pCfg, "ratioOfVnodeStreamThreads", tsRatioOfVnodeStreamThreads, 0.01, 100, CFG_SCOPE_SERVER) != 0) return -1; tsNumOfVnodeFetchThreads = tsNumOfCores / 4; tsNumOfVnodeFetchThreads = TMAX(tsNumOfVnodeFetchThreads, 4); - if (cfgAddInt32(pCfg, "numOfVnodeFetchThreads", tsNumOfVnodeFetchThreads, 4, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfVnodeFetchThreads", tsNumOfVnodeFetchThreads, 4, 1024, CFG_SCOPE_SERVER) != 0) return -1; tsNumOfVnodeRsmaThreads = tsNumOfCores; tsNumOfVnodeRsmaThreads = TMAX(tsNumOfVnodeRsmaThreads, 4); - if (cfgAddInt32(pCfg, "numOfVnodeRsmaThreads", tsNumOfVnodeRsmaThreads, 1, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfVnodeRsmaThreads", tsNumOfVnodeRsmaThreads, 1, 1024, CFG_SCOPE_SERVER) != 0) return -1; tsNumOfQnodeQueryThreads = tsNumOfCores * 2; tsNumOfQnodeQueryThreads = TMAX(tsNumOfQnodeQueryThreads, 4); - if (cfgAddInt32(pCfg, "numOfQnodeQueryThreads", tsNumOfQnodeQueryThreads, 4, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfQnodeQueryThreads", tsNumOfQnodeQueryThreads, 4, 1024, CFG_SCOPE_SERVER) != 0) return -1; // tsNumOfQnodeFetchThreads = tsNumOfCores / 2; // tsNumOfQnodeFetchThreads = TMAX(tsNumOfQnodeFetchThreads, 4); @@ -487,66 +487,67 @@ static int32_t taosAddServerCfg(SConfig *pCfg) { tsNumOfSnodeStreamThreads = tsNumOfCores / 4; tsNumOfSnodeStreamThreads = TRANGE(tsNumOfSnodeStreamThreads, 2, 4); - if (cfgAddInt32(pCfg, "numOfSnodeSharedThreads", tsNumOfSnodeStreamThreads, 2, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfSnodeSharedThreads", tsNumOfSnodeStreamThreads, 2, 1024, CFG_SCOPE_SERVER) != 0) return -1; tsNumOfSnodeWriteThreads = tsNumOfCores / 4; tsNumOfSnodeWriteThreads = TRANGE(tsNumOfSnodeWriteThreads, 2, 4); - if (cfgAddInt32(pCfg, "numOfSnodeUniqueThreads", tsNumOfSnodeWriteThreads, 2, 1024, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "numOfSnodeUniqueThreads", tsNumOfSnodeWriteThreads, 2, 1024, CFG_SCOPE_SERVER) != 0) return -1; tsRpcQueueMemoryAllowed = tsTotalMemoryKB * 1024 * 0.1; tsRpcQueueMemoryAllowed = TRANGE(tsRpcQueueMemoryAllowed, TSDB_MAX_MSG_SIZE * 10LL, TSDB_MAX_MSG_SIZE * 10000LL); - if (cfgAddInt64(pCfg, "rpcQueueMemoryAllowed", tsRpcQueueMemoryAllowed, TSDB_MAX_MSG_SIZE * 10L, INT64_MAX, 0) != 0) + if (cfgAddInt64(pCfg, "rpcQueueMemoryAllowed", tsRpcQueueMemoryAllowed, TSDB_MAX_MSG_SIZE * 10L, INT64_MAX, CFG_SCOPE_BOTH) != 0) return -1; - if (cfgAddInt32(pCfg, "syncElectInterval", tsElectInterval, 10, 1000 * 60 * 24 * 2, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "syncHeartbeatInterval", tsHeartbeatInterval, 10, 1000 * 60 * 24 * 2, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "syncHeartbeatTimeout", tsHeartbeatTimeout, 10, 1000 * 60 * 24 * 2, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "syncElectInterval", tsElectInterval, 10, 1000 * 60 * 24 * 2, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "syncHeartbeatInterval", tsHeartbeatInterval, 10, 1000 * 60 * 24 * 2, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "syncHeartbeatTimeout", tsHeartbeatTimeout, 10, 1000 * 60 * 24 * 2, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddInt64(pCfg, "vndCommitMaxInterval", tsVndCommitMaxIntervalMs, 1000, 1000 * 60 * 60, 0) != 0) return -1; + if (cfgAddInt64(pCfg, "vndCommitMaxInterval", tsVndCommitMaxIntervalMs, 1000, 1000 * 60 * 60, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddInt64(pCfg, "mndSdbWriteDelta", tsMndSdbWriteDelta, 20, 10000, 0) != 0) return -1; - if (cfgAddInt64(pCfg, "mndLogRetention", tsMndLogRetention, 500, 10000, 0) != 0) return -1; - if (cfgAddBool(pCfg, "skipGrant", tsMndSkipGrant, 0) != 0) return -1; + if (cfgAddInt64(pCfg, "mndSdbWriteDelta", tsMndSdbWriteDelta, 20, 10000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt64(pCfg, "mndLogRetention", tsMndLogRetention, 500, 10000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddBool(pCfg, "skipGrant", tsMndSkipGrant, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddBool(pCfg, "monitor", tsEnableMonitor, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "monitorInterval", tsMonitorInterval, 1, 200000, 0) != 0) return -1; - if (cfgAddString(pCfg, "monitorFqdn", tsMonitorFqdn, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "monitorPort", tsMonitorPort, 1, 65056, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "monitorMaxLogs", tsMonitorMaxLogs, 1, 1000000, 0) != 0) return -1; - if (cfgAddBool(pCfg, "monitorComp", tsMonitorComp, 0) != 0) return -1; + if (cfgAddBool(pCfg, "monitor", tsEnableMonitor, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "monitorInterval", tsMonitorInterval, 1, 200000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddString(pCfg, "monitorFqdn", tsMonitorFqdn, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "monitorPort", tsMonitorPort, 1, 65056, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "monitorMaxLogs", tsMonitorMaxLogs, 1, 1000000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddBool(pCfg, "monitorComp", tsMonitorComp, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddBool(pCfg, "crashReporting", tsEnableCrashReport, 0) != 0) return -1; - if (cfgAddBool(pCfg, "telemetryReporting", tsEnableTelem, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "telemetryInterval", tsTelemInterval, 1, 200000, 0) != 0) return -1; - if (cfgAddString(pCfg, "telemetryServer", tsTelemServer, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "telemetryPort", tsTelemPort, 1, 65056, 0) != 0) return -1; + if (cfgAddBool(pCfg, "crashReporting", tsEnableCrashReport, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddBool(pCfg, "telemetryReporting", tsEnableTelem, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "telemetryInterval", tsTelemInterval, 1, 200000, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddString(pCfg, "telemetryServer", tsTelemServer, CFG_SCOPE_BOTH) != 0) return -1; + if (cfgAddInt32(pCfg, "telemetryPort", tsTelemPort, 1, 65056, CFG_SCOPE_BOTH) != 0) return -1; - if (cfgAddInt32(pCfg, "tmqMaxTopicNum", tmqMaxTopicNum, 1, 10000, 1) != 0) return -1; + if (cfgAddInt32(pCfg, "tmqMaxTopicNum", tmqMaxTopicNum, 1, 10000, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddInt32(pCfg, "transPullupInterval", tsTransPullupInterval, 1, 10000, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "mqRebalanceInterval", tsMqRebalanceInterval, 1, 10000, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "ttlUnit", tsTtlUnit, 1, 86400 * 365, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "ttlPushInterval", tsTtlPushInterval, 1, 100000, 1) != 0) return -1; - if (cfgAddBool(pCfg, "ttlChangeOnWrite", tsTtlChangeOnWrite, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "uptimeInterval", tsUptimeInterval, 1, 100000, 1) != 0) return -1; - if (cfgAddInt32(pCfg, "queryRsmaTolerance", tsQueryRsmaTolerance, 0, 900000, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "transPullupInterval", tsTransPullupInterval, 1, 10000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "mqRebalanceInterval", tsMqRebalanceInterval, 1, 10000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "ttlUnit", tsTtlUnit, 1, 86400 * 365, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "ttlPushInterval", tsTtlPushInterval, 1, 100000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddBool(pCfg, "ttlChangeOnWrite", tsTtlChangeOnWrite, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "uptimeInterval", tsUptimeInterval, 1, 100000, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "queryRsmaTolerance", tsQueryRsmaTolerance, 0, 900000, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddInt64(pCfg, "walFsyncDataSizeLimit", tsWalFsyncDataSizeLimit, 100 * 1024 * 1024, INT64_MAX, 0) != 0) + if (cfgAddInt64(pCfg, "walFsyncDataSizeLimit", tsWalFsyncDataSizeLimit, 100 * 1024 * 1024, INT64_MAX, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddBool(pCfg, "udf", tsStartUdfd, 0) != 0) return -1; - if (cfgAddString(pCfg, "udfdResFuncs", tsUdfdResFuncs, 0) != 0) return -1; - if (cfgAddString(pCfg, "udfdLdLibPath", tsUdfdLdLibPath, 0) != 0) return -1; + if (cfgAddBool(pCfg, "udf", tsStartUdfd, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddString(pCfg, "udfdResFuncs", tsUdfdResFuncs, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddString(pCfg, "udfdLdLibPath", tsUdfdLdLibPath, CFG_SCOPE_SERVER) != 0) return -1; if (cfgAddBool(pCfg, "disableStream", tsDisableStream, 0) != 0) return -1; if (cfgAddInt64(pCfg, "streamBufferSize", tsStreamBufferSize, 0, INT64_MAX, 0) != 0) return -1; + if (cfgAddInt64(pCfg, "checkpointInterval", tsCheckpointInterval, 0, INT64_MAX, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddInt32(pCfg, "cacheLazyLoadThreshold", tsCacheLazyLoadThreshold, 0, 100000, 0) != 0) return -1; + if (cfgAddInt32(pCfg, "cacheLazyLoadThreshold", tsCacheLazyLoadThreshold, 0, 100000, CFG_SCOPE_SERVER) != 0) return -1; - if (cfgAddBool(pCfg, "filterScalarMode", tsFilterScalarMode, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "keepTimeOffset", tsKeepTimeOffset, 0, 23, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "maxStreamBackendCache", tsMaxStreamBackendCache, 16, 1024, 0) != 0) return -1; - if (cfgAddInt32(pCfg, "pqSortMemThreshold", tsPQSortMemThreshold, 1, 10240, 0) != 0) return -1; + if (cfgAddBool(pCfg, "filterScalarMode", tsFilterScalarMode, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "keepTimeOffset", tsKeepTimeOffset, 0, 23, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "maxStreamBackendCache", tsMaxStreamBackendCache, 16, 1024, CFG_SCOPE_SERVER) != 0) return -1; + if (cfgAddInt32(pCfg, "pqSortMemThreshold", tsPQSortMemThreshold, 1, 10240, CFG_SCOPE_SERVER) != 0) return -1; GRANT_CFG_ADD; return 0; diff --git a/source/common/src/tmsg.c b/source/common/src/tmsg.c index a25ee04a1b..a29c8556f2 100644 --- a/source/common/src/tmsg.c +++ b/source/common/src/tmsg.c @@ -3484,12 +3484,14 @@ int32_t tDeserializeSShowVariablesReq(void *buf, int32_t bufLen, SShowVariablesR int32_t tEncodeSVariablesInfo(SEncoder *pEncoder, SVariablesInfo *pInfo) { if (tEncodeCStr(pEncoder, pInfo->name) < 0) return -1; if (tEncodeCStr(pEncoder, pInfo->value) < 0) return -1; + if (tEncodeCStr(pEncoder, pInfo->scope) < 0) return -1; return 0; } int32_t tDecodeSVariablesInfo(SDecoder *pDecoder, SVariablesInfo *pInfo) { if (tDecodeCStrTo(pDecoder, pInfo->name) < 0) return -1; if (tDecodeCStrTo(pDecoder, pInfo->value) < 0) return -1; + if (tDecodeCStrTo(pDecoder, pInfo->scope) < 0) return -1; return 0; } diff --git a/source/dnode/mgmt/mgmt_dnode/src/dmHandle.c b/source/dnode/mgmt/mgmt_dnode/src/dmHandle.c index 1998269896..45e17e995f 100644 --- a/source/dnode/mgmt/mgmt_dnode/src/dmHandle.c +++ b/source/dnode/mgmt/mgmt_dnode/src/dmHandle.c @@ -265,6 +265,12 @@ int32_t dmAppendVariablesToBlock(SSDataBlock *pBlock, int32_t dnodeId) { pColInfo = taosArrayGet(pBlock->pDataBlock, c++); colDataSetVal(pColInfo, i, value, false); + char scope[TSDB_CONFIG_SCOPE_LEN + VARSTR_HEADER_SIZE] = {0}; + cfgDumpItemScope(pItem, &scope[VARSTR_HEADER_SIZE], TSDB_CONFIG_SCOPE_LEN, &valueLen); + varDataSetLen(scope, valueLen); + pColInfo = taosArrayGet(pBlock->pDataBlock, c++); + colDataSetVal(pColInfo, i, scope, false); + numOfRows++; } diff --git a/source/dnode/mgmt/mgmt_snode/src/smHandle.c b/source/dnode/mgmt/mgmt_snode/src/smHandle.c index 311cbdc43b..f0cfd6d86a 100644 --- a/source/dnode/mgmt/mgmt_snode/src/smHandle.c +++ b/source/dnode/mgmt/mgmt_snode/src/smHandle.c @@ -79,6 +79,7 @@ SArray *smGetMsgHandles() { if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_CHECK, smPutNodeMsgToStreamQueue, 1) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_CHECK_RSP, smPutNodeMsgToStreamQueue, 1) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_STREAM_SCAN_HISTORY_FINISH, smPutNodeMsgToStreamQueue, 1) == NULL) goto _OVER; + if (dmSetMgmtHandle(pArray, TDMT_STREAM_SCAN_HISTORY_FINISH_RSP, smPutNodeMsgToStreamQueue, 1) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_VND_STREAM_CHECK_POINT_SOURCE, smPutNodeMsgToMgmtQueue, 1) == NULL) goto _OVER; code = 0; diff --git a/source/dnode/mgmt/mgmt_vnode/src/vmHandle.c b/source/dnode/mgmt/mgmt_vnode/src/vmHandle.c index 55e4bd3a2a..6dc38db273 100644 --- a/source/dnode/mgmt/mgmt_vnode/src/vmHandle.c +++ b/source/dnode/mgmt/mgmt_vnode/src/vmHandle.c @@ -737,6 +737,7 @@ SArray *vmGetMsgHandles() { if (dmSetMgmtHandle(pArray, TDMT_STREAM_RETRIEVE, vmPutMsgToStreamQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_STREAM_RETRIEVE_RSP, vmPutMsgToStreamQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_STREAM_SCAN_HISTORY_FINISH, vmPutMsgToStreamQueue, 0) == NULL) goto _OVER; + if (dmSetMgmtHandle(pArray, TDMT_STREAM_SCAN_HISTORY_FINISH_RSP, vmPutMsgToStreamQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_STREAM_TRANSFER_STATE, vmPutMsgToStreamQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_CHECK, vmPutMsgToStreamQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_CHECK_RSP, vmPutMsgToStreamQueue, 0) == NULL) goto _OVER; diff --git a/source/dnode/mnode/impl/inc/mndStb.h b/source/dnode/mnode/impl/inc/mndStb.h index 99af413539..db960d790f 100644 --- a/source/dnode/mnode/impl/inc/mndStb.h +++ b/source/dnode/mnode/impl/inc/mndStb.h @@ -39,6 +39,7 @@ int32_t mndBuildSMCreateStbRsp(SMnode *pMnode, char *dbFName, char *stbFName, vo void mndExtractDbNameFromStbFullName(const char *stbFullName, char *dst); void mndExtractShortDbNameFromStbFullName(const char *stbFullName, char *dst); +void mndExtractShortDbNameFromDbFullName(const char *stbFullName, char *dst); void mndExtractTbNameFromStbFullName(const char *stbFullName, char *dst, int32_t dstSize); const char *mndGetStbStr(const char *src); diff --git a/source/dnode/mnode/impl/src/mndDnode.c b/source/dnode/mnode/impl/src/mndDnode.c index 100513b932..81b591e8b5 100644 --- a/source/dnode/mnode/impl/src/mndDnode.c +++ b/source/dnode/mnode/impl/src/mndDnode.c @@ -785,18 +785,22 @@ static int32_t mndProcessShowVariablesReq(SRpcMsg *pReq) { strcpy(info.name, "statusInterval"); snprintf(info.value, TSDB_CONFIG_VALUE_LEN, "%d", tsStatusInterval); + strcpy(info.scope, "server"); taosArrayPush(rsp.variables, &info); strcpy(info.name, "timezone"); snprintf(info.value, TSDB_CONFIG_VALUE_LEN, "%s", tsTimezoneStr); + strcpy(info.scope, "both"); taosArrayPush(rsp.variables, &info); strcpy(info.name, "locale"); snprintf(info.value, TSDB_CONFIG_VALUE_LEN, "%s", tsLocale); + strcpy(info.scope, "both"); taosArrayPush(rsp.variables, &info); strcpy(info.name, "charset"); snprintf(info.value, TSDB_CONFIG_VALUE_LEN, "%s", tsCharset); + strcpy(info.scope, "both"); taosArrayPush(rsp.variables, &info); int32_t rspLen = tSerializeSShowVariablesRsp(NULL, 0, &rsp); diff --git a/source/dnode/mnode/impl/src/mndScheduler.c b/source/dnode/mnode/impl/src/mndScheduler.c index 1f9f686f75..7a522d9d8f 100644 --- a/source/dnode/mnode/impl/src/mndScheduler.c +++ b/source/dnode/mnode/impl/src/mndScheduler.c @@ -88,10 +88,10 @@ END: int32_t mndSetSinkTaskInfo(SStreamObj* pStream, SStreamTask* pTask) { if (pStream->smaId != 0) { - pTask->outputType = TASK_OUTPUT__SMA; + pTask->outputInfo.type = TASK_OUTPUT__SMA; pTask->smaSink.smaId = pStream->smaId; } else { - pTask->outputType = TASK_OUTPUT__TABLE; + pTask->outputInfo.type = TASK_OUTPUT__TABLE; pTask->tbSink.stbUid = pStream->targetStbUid; memcpy(pTask->tbSink.stbFullName, pStream->targetSTbName, TSDB_TABLE_FNAME_LEN); pTask->tbSink.pSchemaWrapper = tCloneSSchemaWrapper(&pStream->outputSchema); @@ -111,7 +111,7 @@ int32_t mndAddDispatcherForInternalTask(SMnode* pMnode, SStreamObj* pStream, SAr SDbObj* pDb = mndAcquireDb(pMnode, pStream->targetDb); if (pDb != NULL && pDb->cfg.numOfVgroups > 1) { isShuffle = true; - pTask->outputType = TASK_OUTPUT__SHUFFLE_DISPATCH; + pTask->outputInfo.type = TASK_OUTPUT__SHUFFLE_DISPATCH; pTask->msgInfo.msgType = TDMT_STREAM_TASK_DISPATCH; if (mndExtractDbInfo(pMnode, pDb, &pTask->shuffleDispatcher.dbInfo, NULL) < 0) { return -1; @@ -297,7 +297,7 @@ void setFixedDownstreamEpInfo(SStreamTask* pDstTask, const SStreamTask* pTask) { pDispatcher->nodeId = pTask->info.nodeId; pDispatcher->epSet = pTask->info.epSet; - pDstTask->outputType = TASK_OUTPUT__FIXED_DISPATCH; + pDstTask->outputInfo.type = TASK_OUTPUT__FIXED_DISPATCH; pDstTask->msgInfo.msgType = TDMT_STREAM_TASK_DISPATCH; } diff --git a/source/dnode/mnode/impl/src/mndStb.c b/source/dnode/mnode/impl/src/mndStb.c index fc56ae2ead..4b6bdb6d30 100644 --- a/source/dnode/mnode/impl/src/mndStb.c +++ b/source/dnode/mnode/impl/src/mndStb.c @@ -2500,12 +2500,14 @@ static int32_t mndProcessTableCfgReq(SRpcMsg *pReq) { goto _OVER; } - if (0 == strcmp(cfgReq.dbFName, TSDB_INFORMATION_SCHEMA_DB)) { + char dbName[TSDB_DB_NAME_LEN] = {0}; + mndExtractShortDbNameFromDbFullName(cfgReq.dbFName, dbName); + if (0 == strcmp(dbName, TSDB_INFORMATION_SCHEMA_DB)) { mInfo("information_schema table:%s.%s, start to retrieve cfg", cfgReq.dbFName, cfgReq.tbName); if (mndBuildInsTableCfg(pMnode, cfgReq.dbFName, cfgReq.tbName, &cfgRsp) != 0) { goto _OVER; } - } else if (0 == strcmp(cfgReq.dbFName, TSDB_PERFORMANCE_SCHEMA_DB)) { + } else if (0 == strcmp(dbName, TSDB_PERFORMANCE_SCHEMA_DB)) { mInfo("performance_schema table:%s.%s, start to retrieve cfg", cfgReq.dbFName, cfgReq.tbName); if (mndBuildPerfsTableCfg(pMnode, cfgReq.dbFName, cfgReq.tbName, &cfgRsp) != 0) { goto _OVER; @@ -2674,6 +2676,13 @@ void mndExtractShortDbNameFromStbFullName(const char *stbFullName, char *dst) { tNameGetDbName(&name, dst); } +void mndExtractShortDbNameFromDbFullName(const char *stbFullName, char *dst) { + SName name = {0}; + tNameFromString(&name, stbFullName, T_NAME_ACCT | T_NAME_DB); + + tNameGetDbName(&name, dst); +} + void mndExtractTbNameFromStbFullName(const char *stbFullName, char *dst, int32_t dstSize) { int32_t pos = -1; int32_t num = 0; diff --git a/source/dnode/snode/src/snode.c b/source/dnode/snode/src/snode.c index 1c70dfca0d..2ef419dee3 100644 --- a/source/dnode/snode/src/snode.c +++ b/source/dnode/snode/src/snode.c @@ -66,14 +66,15 @@ int32_t sndExpandTask(SSnode *pSnode, SStreamTask *pTask, int64_t ver) { pTask->status.schedStatus = TASK_SCHED_STATUS__INACTIVE; pTask->inputQueue = streamQueueOpen(512 << 10); - pTask->outputQueue = streamQueueOpen(512 << 10); + pTask->outputInfo.queue = streamQueueOpen(512 << 10); - if (pTask->inputQueue == NULL || pTask->outputQueue == NULL) { + if (pTask->inputQueue == NULL || pTask->outputInfo.queue == NULL) { return -1; } + pTask->initTs = taosGetTimestampMs(); pTask->inputStatus = TASK_INPUT_STATUS__NORMAL; - pTask->outputStatus = TASK_OUTPUT_STATUS__NORMAL; + pTask->outputInfo.status = TASK_OUTPUT_STATUS__NORMAL; pTask->pMsgCb = &pSnode->msgCb; pTask->pMeta = pSnode->pMeta; taosThreadMutexInit(&pTask->lock, NULL); @@ -90,6 +91,7 @@ int32_t sndExpandTask(SSnode *pSnode, SStreamTask *pTask, int64_t ver) { pTask->exec.pExecutor = qCreateStreamExecTaskInfo(pTask->exec.qmsg, &handle, 0); ASSERT(pTask->exec.pExecutor); + taosThreadMutexInit(&pTask->lock, NULL); streamTaskOpenAllUpstreamInput(pTask); streamSetupScheduleTrigger(pTask); @@ -167,11 +169,10 @@ int32_t sndProcessTaskDeployReq(SSnode *pSnode, char *msg, int32_t msgLen) { int32_t numOfTasks = streamMetaGetNumOfTasks(pSnode->pMeta); taosWUnLockLatch(&pSnode->pMeta->lock); - - streamPrepareNdoCheckDownstream(pTask); qDebug("snode:%d s-task:%s is deployed on snode and add into meta, status:%s, numOfTasks:%d", SNODE_HANDLE, pTask->id.idStr, - streamGetTaskStatusStr(pTask->status.taskStatus), numOfTasks); + streamGetTaskStatusStr(pTask->status.taskStatus), numOfTasks); + streamTaskCheckDownstreamTasks(pTask); return 0; } @@ -275,7 +276,7 @@ int32_t sndProcessWriteMsg(SSnode *pSnode, SRpcMsg *pMsg, SRpcMsg *pRsp) { return 0; } -int32_t sndProcessTaskRecoverFinishReq(SSnode *pSnode, SRpcMsg *pMsg) { +int32_t sndProcessStreamTaskScanHistoryFinishReq(SSnode *pSnode, SRpcMsg *pMsg) { char *msg = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)); int32_t msgLen = pMsg->contLen - sizeof(SMsgHead); @@ -288,12 +289,12 @@ int32_t sndProcessTaskRecoverFinishReq(SSnode *pSnode, SRpcMsg *pMsg) { tDecoderClear(&decoder); // find task - SStreamTask *pTask = streamMetaAcquireTask(pSnode->pMeta, req.taskId); + SStreamTask *pTask = streamMetaAcquireTask(pSnode->pMeta, req.downstreamTaskId); if (pTask == NULL) { return -1; } // do process request - if (streamProcessScanHistoryFinishReq(pTask, req.childId) < 0) { + if (streamProcessScanHistoryFinishReq(pTask, &req, &pMsg->info) < 0) { streamMetaReleaseTask(pSnode->pMeta, pTask); return -1; } @@ -416,7 +417,7 @@ int32_t sndProcessStreamMsg(SSnode *pSnode, SRpcMsg *pMsg) { case TDMT_STREAM_RETRIEVE_RSP: return sndProcessTaskRetrieveRsp(pSnode, pMsg); case TDMT_STREAM_SCAN_HISTORY_FINISH: - return sndProcessTaskRecoverFinishReq(pSnode, pMsg); + return sndProcessStreamTaskScanHistoryFinishReq(pSnode, pMsg); case TDMT_STREAM_SCAN_HISTORY_FINISH_RSP: return sndProcessTaskRecoverFinishRsp(pSnode, pMsg); case TDMT_STREAM_TASK_CHECK: diff --git a/source/dnode/vnode/CMakeLists.txt b/source/dnode/vnode/CMakeLists.txt index a904f5f87f..c7b81f0dad 100644 --- a/source/dnode/vnode/CMakeLists.txt +++ b/source/dnode/vnode/CMakeLists.txt @@ -1,21 +1,18 @@ # vnode add_library(vnode STATIC "") -target_sources( - vnode - PRIVATE - - # vnode - "src/vnd/vnodeOpen.c" - "src/vnd/vnodeBufPool.c" - "src/vnd/vnodeCfg.c" - "src/vnd/vnodeCommit.c" - "src/vnd/vnodeQuery.c" - "src/vnd/vnodeModule.c" - "src/vnd/vnodeSvr.c" - "src/vnd/vnodeSync.c" - "src/vnd/vnodeSnapshot.c" - "src/vnd/vnodeRetention.c" - "src/vnd/vnodeInitApi.c" +set( + VNODE_SOURCE_FILES + "src/vnd/vnodeOpen.c" + "src/vnd/vnodeBufPool.c" + "src/vnd/vnodeCfg.c" + "src/vnd/vnodeCommit.c" + "src/vnd/vnodeQuery.c" + "src/vnd/vnodeModule.c" + "src/vnd/vnodeSvr.c" + "src/vnd/vnodeSync.c" + "src/vnd/vnodeSnapshot.c" + "src/vnd/vnodeRetention.c" + "src/vnd/vnodeInitApi.c" # meta "src/meta/metaOpen.c" @@ -38,23 +35,23 @@ target_sources( "src/sma/smaSnapshot.c" "src/sma/smaTimeRange.c" - # tsdb - "src/tsdb/tsdbCommit.c" - "src/tsdb/tsdbFile.c" - "src/tsdb/tsdbFS.c" - "src/tsdb/tsdbOpen.c" - "src/tsdb/tsdbMemTable.c" - "src/tsdb/tsdbRead.c" - "src/tsdb/tsdbCache.c" - "src/tsdb/tsdbWrite.c" - "src/tsdb/tsdbReaderWriter.c" - "src/tsdb/tsdbUtil.c" - "src/tsdb/tsdbSnapshot.c" - "src/tsdb/tsdbCacheRead.c" - "src/tsdb/tsdbRetention.c" - "src/tsdb/tsdbDiskData.c" - "src/tsdb/tsdbMergeTree.c" - "src/tsdb/tsdbDataIter.c" + # # tsdb + # "src/tsdb/tsdbCommit.c" + # "src/tsdb/tsdbFile.c" + # "src/tsdb/tsdbFS.c" + # "src/tsdb/tsdbOpen.c" + # "src/tsdb/tsdbMemTable.c" + # "src/tsdb/tsdbRead.c" + # "src/tsdb/tsdbCache.c" + # "src/tsdb/tsdbWrite.c" + # "src/tsdb/tsdbReaderWriter.c" + # "src/tsdb/tsdbUtil.c" + # "src/tsdb/tsdbSnapshot.c" + # "src/tsdb/tsdbCacheRead.c" + # "src/tsdb/tsdbRetention.c" + # "src/tsdb/tsdbDiskData.c" + # "src/tsdb/tsdbMergeTree.c" + # "src/tsdb/tsdbDataIter.c" # tq "src/tq/tq.c" @@ -74,6 +71,19 @@ target_sources( ) +aux_source_directory("src/tsdb/" TSDB_SOURCE_FILES) +list( + APPEND + VNODE_SOURCE_FILES + ${TSDB_SOURCE_FILES} +) + +target_sources( + vnode + PRIVATE + ${VNODE_SOURCE_FILES} +) + IF (TD_VNODE_PLUGINS) target_sources( vnode diff --git a/source/dnode/vnode/inc/vnode.h b/source/dnode/vnode/inc/vnode.h index 38216e1414..f60cc2f406 100644 --- a/source/dnode/vnode/inc/vnode.h +++ b/source/dnode/vnode/inc/vnode.h @@ -168,6 +168,27 @@ uint64_t tsdbGetReaderMaxVersion(STsdbReader *pReader); void tsdbReaderSetCloseFlag(STsdbReader *pReader); int64_t tsdbGetLastTimestamp(SVnode *pVnode, void *pTableList, int32_t numOfTables, const char *pIdStr); +//====================================================================================================================== +int32_t tsdbReaderOpen2(void *pVnode, SQueryTableDataCond *pCond, void *pTableList, int32_t numOfTables, + SSDataBlock *pResBlock, void **ppReader, const char *idstr, bool countOnly, + SHashObj **pIgnoreTables); +int32_t tsdbSetTableList2(STsdbReader *pReader, const void *pTableList, int32_t num); +void tsdbReaderSetId2(STsdbReader *pReader, const char *idstr); +void tsdbReaderClose2(STsdbReader *pReader); +int32_t tsdbNextDataBlock2(STsdbReader *pReader, bool *hasNext); +int32_t tsdbRetrieveDatablockSMA2(STsdbReader *pReader, SSDataBlock *pDataBlock, bool *allHave, bool *hasNullSMA); +void tsdbReleaseDataBlock2(STsdbReader *pReader); +SSDataBlock *tsdbRetrieveDataBlock2(STsdbReader *pTsdbReadHandle, SArray *pColumnIdList); +int32_t tsdbReaderReset2(STsdbReader *pReader, SQueryTableDataCond *pCond); +int32_t tsdbGetFileBlocksDistInfo2(STsdbReader *pReader, STableBlockDistInfo *pTableBlockInfo); +int64_t tsdbGetNumOfRowsInMemTable2(STsdbReader *pHandle); +void *tsdbGetIdx2(SMeta *pMeta); +void *tsdbGetIvtIdx2(SMeta *pMeta); +uint64_t tsdbGetReaderMaxVersion2(STsdbReader *pReader); +void tsdbReaderSetCloseFlag2(STsdbReader *pReader); +int64_t tsdbGetLastTimestamp2(SVnode *pVnode, void *pTableList, int32_t numOfTables, const char *pIdStr); +//====================================================================================================================== + int32_t tsdbReuseCacherowsReader(void *pReader, void *pTableIdList, int32_t numOfTables); int32_t tsdbCacherowsReaderOpen(void *pVnode, int32_t type, void *pTableIdList, int32_t numOfTables, int32_t numOfCols, SArray *pCidList, int32_t *pSlotIds, uint64_t suid, void **pReader, const char *idstr); diff --git a/source/dnode/vnode/src/inc/tsdb.h b/source/dnode/vnode/src/inc/tsdb.h index 71af169752..ad72c5924c 100644 --- a/source/dnode/vnode/src/inc/tsdb.h +++ b/source/dnode/vnode/src/inc/tsdb.h @@ -16,6 +16,9 @@ #ifndef _TD_VNODE_TSDB_H_ #define _TD_VNODE_TSDB_H_ +// #include "../tsdb/tsdbFile2.h" +// #include "../tsdb/tsdbMerge.h" +// #include "../tsdb/tsdbSttFileRW.h" #include "tsimplehash.h" #include "vnodeInt.h" @@ -75,9 +78,8 @@ typedef struct STsdbFilterInfo STsdbFilterInfo; #define TSDBROW_ROW_FMT ((int8_t)0x0) #define TSDBROW_COL_FMT ((int8_t)0x1) -#define TSDB_FILE_DLMT ((uint32_t)0xF00AFA0F) -#define TSDB_MAX_SUBBLOCKS 8 -#define TSDB_FHDR_SIZE 512 +#define TSDB_FILE_DLMT ((uint32_t)0xF00AFA0F) +#define TSDB_FHDR_SIZE 512 #define VERSION_MIN 0 #define VERSION_MAX INT64_MAX @@ -165,6 +167,7 @@ void tBlockDataDestroy(SBlockData *pBlockData); int32_t tBlockDataInit(SBlockData *pBlockData, TABLEID *pId, STSchema *pTSchema, int16_t *aCid, int32_t nCid); void tBlockDataReset(SBlockData *pBlockData); int32_t tBlockDataAppendRow(SBlockData *pBlockData, TSDBROW *pRow, STSchema *pTSchema, int64_t uid); +int32_t tBlockDataUpdateRow(SBlockData *pBlockData, TSDBROW *pRow, STSchema *pTSchema); int32_t tBlockDataTryUpsertRow(SBlockData *pBlockData, TSDBROW *pRow, int64_t uid); int32_t tBlockDataUpsertRow(SBlockData *pBlockData, TSDBROW *pRow, STSchema *pTSchema, int64_t uid); void tBlockDataClear(SBlockData *pBlockData); @@ -198,7 +201,7 @@ int32_t tMapDataToArray(SMapData *pMapData, int32_t itemSize, int32_t (*tGetItem // other int32_t tsdbKeyFid(TSKEY key, int32_t minutes, int8_t precision); void tsdbFidKeyRange(int32_t fid, int32_t minutes, int8_t precision, TSKEY *minKey, TSKEY *maxKey); -int32_t tsdbFidLevel(int32_t fid, STsdbKeepCfg *pKeepCfg, int64_t now); +int32_t tsdbFidLevel(int32_t fid, STsdbKeepCfg *pKeepCfg, int64_t nowSec); int32_t tsdbBuildDeleteSkyline(SArray *aDelData, int32_t sidx, int32_t eidx, SArray *aSkyline); int32_t tPutColumnDataAgg(uint8_t *p, SColumnDataAgg *pColAgg); int32_t tGetColumnDataAgg(uint8_t *p, SColumnDataAgg *pColAgg); @@ -302,8 +305,11 @@ int32_t tsdbReadDelIdx(SDelFReader *pReader, SArray *aDelIdx); // tsdbRead.c ============================================================================================== int32_t tsdbTakeReadSnap(STsdbReader *pReader, _query_reseek_func_t reseek, STsdbReadSnap **ppSnap); void tsdbUntakeReadSnap(STsdbReader *pReader, STsdbReadSnap *pSnap, bool proactive); + +int32_t tsdbTakeReadSnap2(STsdbReader *pReader, _query_reseek_func_t reseek, STsdbReadSnap **ppSnap); +void tsdbUntakeReadSnap2(STsdbReader *pReader, STsdbReadSnap *pSnap, bool proactive); // tsdbMerge.c ============================================================================================== -int32_t tsdbMerge(STsdb *pTsdb); +int32_t tsdbMerge(void *arg); // tsdbDiskData ============================================================================================== int32_t tDiskDataBuilderCreate(SDiskDataBuilder **ppBuilder); @@ -362,19 +368,20 @@ typedef struct { } SCacheFlushState; struct STsdb { - char *path; - SVnode *pVnode; - STsdbKeepCfg keepCfg; - TdThreadRwlock rwLock; - SMemTable *mem; - SMemTable *imem; - STsdbFS fs; - SLRUCache *lruCache; - SCacheFlushState flushState; - TdThreadMutex lruMutex; - SLRUCache *biCache; - TdThreadMutex biMutex; - SRocksCache rCache; + char *path; + SVnode *pVnode; + STsdbKeepCfg keepCfg; + TdThreadRwlock rwLock; + SMemTable *mem; + SMemTable *imem; + STsdbFS fs; // old + SLRUCache *lruCache; + SCacheFlushState flushState; + TdThreadMutex lruMutex; + SLRUCache *biCache; + TdThreadMutex biMutex; + struct STFileSystem *pFS; // new + SRocksCache rCache; }; struct TSDBKEY { @@ -410,6 +417,7 @@ struct STbData { SDelData *pTail; SMemSkipList sl; STbData *next; + SRBTreeNode rbtn[1]; }; struct SMemTable { @@ -423,11 +431,10 @@ struct SMemTable { TSKEY maxKey; int64_t nRow; int64_t nDel; - struct { - int32_t nTbData; - int32_t nBucket; - STbData **aBucket; - }; + int32_t nTbData; + int32_t nBucket; + STbData **aBucket; + SRBTree tbDataTree[1]; }; struct TSDBROW { @@ -500,7 +507,7 @@ struct SDataBlk { int32_t nRow; int8_t hasDup; int8_t nSubBlock; - SBlockInfo aSubBlock[TSDB_MAX_SUBBLOCKS]; + SBlockInfo aSubBlock[1]; SSmaInfo smaInfo; }; @@ -652,12 +659,19 @@ struct SDelFWriter { uint8_t *aBuf[1]; }; +#include "tarray2.h" +//#include "tsdbFS2.h" +// struct STFileSet; +typedef struct STFileSet STFileSet; +typedef TARRAY2(STFileSet *) TFileSetArray; + struct STsdbReadSnap { - SMemTable *pMem; - SQueryNode *pNode; - SMemTable *pIMem; - SQueryNode *pINode; - STsdbFS fs; + SMemTable *pMem; + SQueryNode *pNode; + SMemTable *pIMem; + SQueryNode *pINode; + TFileSetArray *pfSetArray; + STsdbFS fs; }; struct SDataFWriter { @@ -696,6 +710,7 @@ typedef struct { typedef struct SSttBlockLoadInfo { SBlockData blockData[2]; + void *pSttStatisBlkArray; SArray *aSttBlk; int32_t blockIndex[2]; // to denote the loaded block in the corresponding position. int32_t currentLoadBlockIndex; @@ -704,10 +719,9 @@ typedef struct SSttBlockLoadInfo { STSchema *pSchema; int16_t *colIds; int32_t numOfCols; - bool checkRemainingRow; + bool checkRemainingRow; // todo: no assign value? bool isLast; bool sttBlockLoaded; - int32_t numOfStt; // keep the last access position, this position may be used to reduce the binary times for // starting last block data for a new table @@ -766,60 +780,106 @@ struct SDiskDataBuilder { }; typedef struct SLDataIter { - SRBTreeNode node; - SSttBlk *pSttBlk; - SDataFReader *pReader; - int32_t iStt; - int8_t backward; - int32_t iSttBlk; - int32_t iRow; - SRowInfo rInfo; - uint64_t uid; - STimeWindow timeWindow; - SVersionRange verRange; - SSttBlockLoadInfo *pBlockLoadInfo; - bool ignoreEarlierTs; + SRBTreeNode node; + SSttBlk *pSttBlk; + int32_t iStt; // for debug purpose + int8_t backward; + int32_t iSttBlk; + int32_t iRow; + SRowInfo rInfo; + uint64_t uid; + STimeWindow timeWindow; + SVersionRange verRange; + SSttBlockLoadInfo *pBlockLoadInfo; + bool ignoreEarlierTs; + struct SSttFileReader *pReader; } SLDataIter; #define tMergeTreeGetRow(_t) (&((_t)->pIter->rInfo.row)) int32_t tMergeTreeOpen(SMergeTree *pMTree, int8_t backward, SDataFReader *pFReader, uint64_t suid, uint64_t uid, STimeWindow *pTimeWindow, SVersionRange *pVerRange, SSttBlockLoadInfo *pBlockLoadInfo, bool destroyLoadInfo, const char *idStr, bool strictTimeRange, SLDataIter *pLDataIter); -void tMergeTreeAddIter(SMergeTree *pMTree, SLDataIter *pIter); -bool tMergeTreeNext(SMergeTree *pMTree); -bool tMergeTreeIgnoreEarlierTs(SMergeTree *pMTree); -void tMergeTreeClose(SMergeTree *pMTree); + +struct SSttFileReader; +typedef int32_t (*_load_tomb_fn)(STsdbReader *pReader, struct SSttFileReader *pSttFileReader, + SSttBlockLoadInfo *pLoadInfo); + +typedef struct { + int8_t backward; + STsdb *pTsdb; + uint64_t suid; + uint64_t uid; + STimeWindow timewindow; + SVersionRange verRange; + bool strictTimeRange; + SArray *pSttFileBlockIterArray; + void *pCurrentFileset; + STSchema *pSchema; + int16_t *pCols; + int32_t numOfCols; + _load_tomb_fn loadTombFn; + void *pReader; + void *idstr; +} SMergeTreeConf; +int32_t tMergeTreeOpen2(SMergeTree *pMTree, SMergeTreeConf *pConf); + +void tMergeTreeAddIter(SMergeTree *pMTree, SLDataIter *pIter); +bool tMergeTreeNext(SMergeTree *pMTree); +bool tMergeTreeIgnoreEarlierTs(SMergeTree *pMTree); +void tMergeTreeClose(SMergeTree *pMTree); SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols, int32_t numOfStt); +SSttBlockLoadInfo *tCreateOneLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols); void resetLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo); void getLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo, int64_t *blocks, double *el); void *destroyLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo); +void *destroySttBlockReader(SArray *pLDataIterArray, int64_t *blocks, double *el); // tsdbCache ============================================================================================== +typedef enum { + READ_MODE_COUNT_ONLY = 0x1, + READ_MODE_ALL, +} EReadMode; + +typedef struct STsdbReaderInfo { + uint64_t suid; + STSchema *pSchema; + EReadMode readMode; + uint64_t rowsNum; + STimeWindow window; + SVersionRange verRange; + int16_t order; +} STsdbReaderInfo; + +typedef struct { + SArray *pTombData; +} STableLoadInfo; + +struct SDataFileReader; + typedef struct SCacheRowsReader { - STsdb *pTsdb; - SVersionRange verRange; - TdThreadMutex readerMutex; - SVnode *pVnode; - STSchema *pSchema; - STSchema *pCurrSchema; - uint64_t uid; - uint64_t suid; - char **transferBuf; // todo remove it soon - int32_t numOfCols; - SArray *pCidList; - int32_t *pSlotIds; - int32_t type; - int32_t tableIndex; // currently returned result tables - STableKeyInfo *pTableList; // table id list - int32_t numOfTables; - SSttBlockLoadInfo *pLoadInfo; - SLDataIter *pDataIter; - STsdbReadSnap *pReadSnap; - SDataFReader *pDataFReader; - SDataFReader *pDataFReaderLast; - const char *idstr; - int64_t lastTs; + STsdb *pTsdb; + STsdbReaderInfo info; + TdThreadMutex readerMutex; + SVnode *pVnode; + STSchema *pSchema; + STSchema *pCurrSchema; + uint64_t uid; + char **transferBuf; // todo remove it soon + int32_t numOfCols; + SArray *pCidList; + int32_t *pSlotIds; + int32_t type; + int32_t tableIndex; // currently returned result tables + STableKeyInfo *pTableList; // table id list + int32_t numOfTables; + uint64_t *uidList; + SSHashObj *pTableMap; + SArray *pLDataIterArray; + struct SDataFileReader *pFileReader; + STsdbReadSnap *pReadSnap; + char *idstr; + int64_t lastTs; } SCacheRowsReader; typedef struct { diff --git a/source/dnode/vnode/src/inc/vnd.h b/source/dnode/vnode/src/inc/vnd.h index 5e7d522fb9..85ef384ea9 100644 --- a/source/dnode/vnode/src/inc/vnd.h +++ b/source/dnode/vnode/src/inc/vnd.h @@ -49,7 +49,8 @@ int32_t vnodeEncodeConfig(const void* pObj, SJson* pJson); int32_t vnodeDecodeConfig(const SJson* pJson, void* pObj); // vnodeModule.c -int32_t vnodeScheduleTask(int32_t (*execute)(void*), void* arg); +int vnodeScheduleTask(int (*execute)(void*), void* arg); +int vnodeScheduleTaskEx(int tpid, int (*execute)(void*), void* arg); // vnodeBufPool.c typedef struct SVBufPoolNode SVBufPoolNode; diff --git a/source/dnode/vnode/src/inc/vnodeInt.h b/source/dnode/vnode/src/inc/vnodeInt.h index 2f37048d5c..709787b7ca 100644 --- a/source/dnode/vnode/src/inc/vnodeInt.h +++ b/source/dnode/vnode/src/inc/vnodeInt.h @@ -181,8 +181,8 @@ SArray* metaGetSmaTbUids(SMeta* pMeta); void* metaGetIdx(SMeta* pMeta); void* metaGetIvtIdx(SMeta* pMeta); -int64_t metaGetTbNum(SMeta *pMeta); -void metaReaderDoInit(SMetaReader *pReader, SMeta *pMeta, int32_t flags); +int64_t metaGetTbNum(SMeta* pMeta); +void metaReaderDoInit(SMetaReader* pReader, SMeta* pMeta, int32_t flags); int32_t metaCreateTSma(SMeta* pMeta, int64_t version, SSmaCfg* pCfg); int32_t metaDropTSma(SMeta* pMeta, int64_t indexUid); @@ -199,12 +199,12 @@ int32_t metaGetInfo(SMeta* pMeta, int64_t uid, SMetaInfo* pInfo, SMetaReader* pR int tsdbOpen(SVnode* pVnode, STsdb** ppTsdb, const char* dir, STsdbKeepCfg* pKeepCfg, int8_t rollback); int tsdbClose(STsdb** pTsdb); int32_t tsdbBegin(STsdb* pTsdb); -int32_t tsdbPrepareCommit(STsdb* pTsdb); -int32_t tsdbCommit(STsdb* pTsdb, SCommitInfo* pInfo); +// int32_t tsdbPrepareCommit(STsdb* pTsdb); +// int32_t tsdbCommit(STsdb* pTsdb, SCommitInfo* pInfo); int32_t tsdbCacheCommit(STsdb* pTsdb); int32_t tsdbCompact(STsdb* pTsdb, SCompactInfo* pInfo); -int32_t tsdbFinishCommit(STsdb* pTsdb); -int32_t tsdbRollbackCommit(STsdb* pTsdb); +// int32_t tsdbFinishCommit(STsdb* pTsdb); +// int32_t tsdbRollbackCommit(STsdb* pTsdb); int tsdbScanAndConvertSubmitMsg(STsdb* pTsdb, SSubmitReq2* pMsg); int tsdbInsertData(STsdb* pTsdb, int64_t version, SSubmitReq2* pMsg, SSubmitRsp2* pRsp); int32_t tsdbInsertTableData(STsdb* pTsdb, int64_t version, SSubmitTbData* pSubmitTbData, int32_t* affectedRows); @@ -253,8 +253,8 @@ int32_t tqProcessTaskRetrieveReq(STQ* pTq, SRpcMsg* pMsg); int32_t tqProcessTaskRetrieveRsp(STQ* pTq, SRpcMsg* pMsg); int32_t tqProcessTaskScanHistory(STQ* pTq, SRpcMsg* pMsg); int32_t tqProcessTaskTransferStateReq(STQ* pTq, SRpcMsg* pMsg); -int32_t tqProcessStreamTaskScanHistoryFinishReq(STQ* pTq, SRpcMsg* pMsg); -int32_t tqProcessTaskRecoverFinishRsp(STQ* pTq, SRpcMsg* pMsg); +int32_t tqProcessTaskScanHistoryFinishReq(STQ* pTq, SRpcMsg* pMsg); +int32_t tqProcessTaskScanHistoryFinishRsp(STQ* pTq, SRpcMsg* pMsg); int32_t tqCheckLogInWal(STQ* pTq, int64_t version); // sma diff --git a/source/dnode/vnode/src/meta/metaTtl.c b/source/dnode/vnode/src/meta/metaTtl.c index 045a759fad..45f697258c 100644 --- a/source/dnode/vnode/src/meta/metaTtl.c +++ b/source/dnode/vnode/src/meta/metaTtl.c @@ -379,7 +379,7 @@ _out: int ttlMgrFlush(STtlManger *pTtlMgr, TXN *pTxn) { ttlMgrWLock(pTtlMgr); - metaInfo("%s, ttl mgr flush start. dirty uids:%d", pTtlMgr->logPrefix, taosHashGetSize(pTtlMgr->pDirtyUids)); + metaDebug("%s, ttl mgr flush start. dirty uids:%d", pTtlMgr->logPrefix, taosHashGetSize(pTtlMgr->pDirtyUids)); int ret = -1; @@ -433,7 +433,7 @@ int ttlMgrFlush(STtlManger *pTtlMgr, TXN *pTxn) { _out: ttlMgrULock(pTtlMgr); - metaInfo("%s, ttl mgr flush end.", pTtlMgr->logPrefix); + metaDebug("%s, ttl mgr flush end.", pTtlMgr->logPrefix); return ret; } diff --git a/source/dnode/vnode/src/sma/smaCommit.c b/source/dnode/vnode/src/sma/smaCommit.c index d1c4314091..c26157f4b7 100644 --- a/source/dnode/vnode/src/sma/smaCommit.c +++ b/source/dnode/vnode/src/sma/smaCommit.c @@ -103,15 +103,16 @@ _exit: return code; } -int32_t smaFinishCommit(SSma *pSma) { +extern int32_t tsdbCommitCommit(STsdb *tsdb); +int32_t smaFinishCommit(SSma *pSma) { int32_t code = 0; int32_t lino = 0; SVnode *pVnode = pSma->pVnode; - if (VND_RSMA1(pVnode) && (code = tsdbFinishCommit(VND_RSMA1(pVnode))) < 0) { + if (VND_RSMA1(pVnode) && (code = tsdbCommitCommit(VND_RSMA1(pVnode))) < 0) { TSDB_CHECK_CODE(code, lino, _exit); } - if (VND_RSMA2(pVnode) && (code = tsdbFinishCommit(VND_RSMA2(pVnode))) < 0) { + if (VND_RSMA2(pVnode) && (code = tsdbCommitCommit(VND_RSMA2(pVnode))) < 0) { TSDB_CHECK_CODE(code, lino, _exit); } _exit: @@ -130,6 +131,7 @@ _exit: * @param isCommit * @return int32_t */ +extern int32_t tsdbPreCommit(STsdb *tsdb); static int32_t tdProcessRSmaAsyncPreCommitImpl(SSma *pSma, bool isCommit) { int32_t code = 0; int32_t lino = 0; @@ -186,11 +188,11 @@ static int32_t tdProcessRSmaAsyncPreCommitImpl(SSma *pSma, bool isCommit) { // all rsma results are written completely STsdb *pTsdb = NULL; if ((pTsdb = VND_RSMA1(pSma->pVnode))) { - code = tsdbPrepareCommit(pTsdb); + code = tsdbPreCommit(pTsdb); TSDB_CHECK_CODE(code, lino, _exit); } if ((pTsdb = VND_RSMA2(pSma->pVnode))) { - code = tsdbPrepareCommit(pTsdb); + code = tsdbPreCommit(pTsdb); TSDB_CHECK_CODE(code, lino, _exit); } @@ -207,6 +209,7 @@ _exit: * @param pSma * @return int32_t */ +extern int32_t tsdbCommitBegin(STsdb *tsdb, SCommitInfo *info); static int32_t tdProcessRSmaAsyncCommitImpl(SSma *pSma, SCommitInfo *pInfo) { int32_t code = 0; int32_t lino = 0; @@ -217,10 +220,10 @@ static int32_t tdProcessRSmaAsyncCommitImpl(SSma *pSma, SCommitInfo *pInfo) { goto _exit; } - code = tsdbCommit(VND_RSMA1(pVnode), pInfo); + code = tsdbCommitBegin(VND_RSMA1(pVnode), pInfo); TSDB_CHECK_CODE(code, lino, _exit); - code = tsdbCommit(VND_RSMA2(pVnode), pInfo); + code = tsdbCommitBegin(VND_RSMA2(pVnode), pInfo); TSDB_CHECK_CODE(code, lino, _exit); _exit: diff --git a/source/dnode/vnode/src/sma/smaRollup.c b/source/dnode/vnode/src/sma/smaRollup.c index 490bcd1238..9fd4938448 100644 --- a/source/dnode/vnode/src/sma/smaRollup.c +++ b/source/dnode/vnode/src/sma/smaRollup.c @@ -264,7 +264,7 @@ static int32_t tdSetRSmaInfoItemParams(SSma *pSma, SRSmaParam *param, SRSmaStat return TSDB_CODE_FAILED; } - SReadHandle handle = { .vnode = pVnode, .initTqReader = 1, .pStateBackend = pStreamState }; + SReadHandle handle = {.vnode = pVnode, .initTqReader = 1, .pStateBackend = pStreamState}; initStorageAPI(&handle.api); pRSmaInfo->taskInfo[idx] = qCreateStreamExecTaskInfo(param->qmsg[idx], &handle, TD_VID(pVnode)); @@ -572,8 +572,8 @@ int32_t smaDoRetention(SSma *pSma, int64_t now) { for (int32_t i = 0; i < TSDB_RETENTION_L2; ++i) { if (pSma->pRSmaTsdb[i]) { - code = tsdbDoRetention(pSma->pRSmaTsdb[i], now); - if (code) goto _end; + // code = tsdbDoRetention(pSma->pRSmaTsdb[i], now); + // if (code) goto _end; } } @@ -612,7 +612,6 @@ static int32_t tdRSmaExecAndSubmitResult(SSma *pSma, qTaskInfo_t taskInfo, SRSma blockDebugShowDataBlocks(pResList, flag); #endif for (int32_t i = 0; i < taosArrayGetSize(pResList); ++i) { - output = taosArrayGetP(pResList, i); smaDebug("vgId:%d, result block, uid:%" PRIu64 ", groupid:%" PRIu64 ", rows:%" PRIi64, SMA_VID(pSma), output->info.id.uid, output->info.id.groupId, output->info.rows); @@ -1114,8 +1113,8 @@ static void tdRSmaFetchTrigger(void *param, void *tmrId) { } if (!(pStat = (SRSmaStat *)tdAcquireSmaRef(smaMgmt.rsetId, pRSmaRef->refId))) { - smaWarn("rsma fetch task not start since rsma stat already destroyed, rsetId:%d refId:%" PRIi64 ")", - smaMgmt.rsetId, pRSmaRef->refId); // pRSmaRef freed in taosHashRemove + smaWarn("rsma fetch task not start since rsma stat already destroyed, rsetId:%d refId:%" PRIi64 ")", smaMgmt.rsetId, + pRSmaRef->refId); // pRSmaRef freed in taosHashRemove taosHashRemove(smaMgmt.refHash, ¶m, POINTER_BYTES); return; } diff --git a/source/dnode/vnode/src/tq/tq.c b/source/dnode/vnode/src/tq/tq.c index dcd44209eb..6537e8ea3c 100644 --- a/source/dnode/vnode/src/tq/tq.c +++ b/source/dnode/vnode/src/tq/tq.c @@ -815,17 +815,20 @@ int32_t tqExpandTask(STQ* pTq, SStreamTask* pTask, int64_t ver) { pTask->refCnt = 1; pTask->status.schedStatus = TASK_SCHED_STATUS__INACTIVE; pTask->inputQueue = streamQueueOpen(512 << 10); - pTask->outputQueue = streamQueueOpen(512 << 10); + pTask->outputInfo.queue = streamQueueOpen(512 << 10); - if (pTask->inputQueue == NULL || pTask->outputQueue == NULL) { + if (pTask->inputQueue == NULL || pTask->outputInfo.queue == NULL) { + tqError("s-task:%s failed to prepare the input/output queue, initialize task failed", pTask->id.idStr); return -1; } + pTask->initTs = taosGetTimestampMs(); pTask->inputStatus = TASK_INPUT_STATUS__NORMAL; - pTask->outputStatus = TASK_OUTPUT_STATUS__NORMAL; + pTask->outputInfo.status = TASK_OUTPUT_STATUS__NORMAL; pTask->pMsgCb = &pTq->pVnode->msgCb; pTask->pMeta = pTq->pStreamMeta; + // backup the initial status, and set it to be TASK_STATUS__INIT pTask->chkInfo.currentVer = ver; pTask->dataRange.range.maxVer = ver; pTask->dataRange.range.minVer = ver; @@ -890,15 +893,14 @@ int32_t tqExpandTask(STQ* pTq, SStreamTask* pTask, int64_t ver) { if (pTask->exec.pExecutor == NULL) { return -1; } - qSetTaskId(pTask->exec.pExecutor, pTask->id.taskId, pTask->id.streamId); } // sink - if (pTask->outputType == TASK_OUTPUT__SMA) { + if (pTask->outputInfo.type == TASK_OUTPUT__SMA) { pTask->smaSink.vnode = pTq->pVnode; pTask->smaSink.smaSink = smaHandleRes; - } else if (pTask->outputType == TASK_OUTPUT__TABLE) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__TABLE) { pTask->tbSink.vnode = pTq->pVnode; pTask->tbSink.tbSinkFunc = tqSinkToTablePipeline; @@ -924,6 +926,13 @@ int32_t tqExpandTask(STQ* pTq, SStreamTask* pTask, int64_t ver) { pTask->exec.pWalReader = walOpenReader(pTq->pVnode->pWal, &cond); } + // reset the task status from unfinished transaction + if (pTask->status.taskStatus == TASK_STATUS__PAUSE) { + tqWarn("s-task:%s reset task status to be normal, kept in meta status: Paused", pTask->id.idStr); + pTask->status.taskStatus = TASK_STATUS__NORMAL; + } + + taosThreadMutexInit(&pTask->lock, NULL); streamTaskOpenAllUpstreamInput(pTask); streamSetupScheduleTrigger(pTask); SCheckpointInfo* pChkInfo = &pTask->chkInfo; @@ -984,29 +993,7 @@ int32_t tqProcessStreamTaskCheckReq(STQ* pTq, SRpcMsg* pMsg) { taskId, rsp.reqId, rsp.upstreamTaskId, rsp.upstreamNodeId, rsp.status); } - // todo extract method - SEncoder encoder; - int32_t code; - int32_t len; - - tEncodeSize(tEncodeStreamTaskCheckRsp, &rsp, len, code); - if (code < 0) { - tqError("vgId:%d failed to encode task check rsp, task:0x%x", pTq->pStreamMeta->vgId, taskId); - return -1; - } - - void* buf = rpcMallocCont(sizeof(SMsgHead) + len); - ((SMsgHead*)buf)->vgId = htonl(req.upstreamNodeId); - - void* abuf = POINTER_SHIFT(buf, sizeof(SMsgHead)); - tEncoderInit(&encoder, (uint8_t*)abuf, len); - tEncodeStreamTaskCheckRsp(&encoder, &rsp); - tEncoderClear(&encoder); - - SRpcMsg rspMsg = {.code = 0, .pCont = buf, .contLen = sizeof(SMsgHead) + len, .info = pMsg->info}; - - tmsgSendRsp(&rspMsg); - return 0; + return streamSendCheckRsp(pTq->pStreamMeta, &req, &rsp, &pMsg->info, taskId); } int32_t tqProcessStreamTaskCheckRsp(STQ* pTq, int64_t sversion, SRpcMsg* pMsg) { @@ -1072,9 +1059,11 @@ int32_t tqProcessTaskDeployReq(STQ* pTq, int64_t sversion, char* msg, int32_t ms SStreamMeta* pStreamMeta = pTq->pStreamMeta; // 2.save task, use the newest commit version as the initial start version of stream task. + int32_t taskId = 0; taosWLockLatch(&pStreamMeta->lock); code = streamMetaAddDeployedTask(pStreamMeta, sversion, pTask); + taskId = pTask->id.taskId; int32_t numOfTasks = streamMetaGetNumOfTasks(pStreamMeta); if (code < 0) { tqError("vgId:%d failed to add s-task:%s, total:%d", vgId, pTask->id.idStr, numOfTasks); @@ -1083,13 +1072,16 @@ int32_t tqProcessTaskDeployReq(STQ* pTq, int64_t sversion, char* msg, int32_t ms } taosWUnLockLatch(&pStreamMeta->lock); - - // 3. It's an fill history task, do nothing. wait for the main task to start it - streamPrepareNdoCheckDownstream(pTask); - tqDebug("vgId:%d s-task:%s is deployed and add into meta, status:%s, numOfTasks:%d", vgId, pTask->id.idStr, streamGetTaskStatusStr(pTask->status.taskStatus), numOfTasks); + // 3. It's an fill history task, do nothing. wait for the main task to start it + SStreamTask* p = streamMetaAcquireTask(pStreamMeta, taskId); + if (p != NULL) { + streamTaskCheckDownstreamTasks(pTask); + } + + streamMetaReleaseTask(pStreamMeta, p); return 0; } @@ -1108,31 +1100,51 @@ int32_t tqProcessTaskScanHistory(STQ* pTq, SRpcMsg* pMsg) { } // do recovery step 1 - const char* pId = pTask->id.idStr; - tqDebug("s-task:%s start history data scan stage(step 1), status:%s", pId, - streamGetTaskStatusStr(pTask->status.taskStatus)); + const char* id = pTask->id.idStr; + const char* pStatus = streamGetTaskStatusStr(pTask->status.taskStatus); + tqDebug("s-task:%s start history data scan stage(step 1), status:%s", id, pStatus); int64_t st = taosGetTimestampMs(); - int8_t schedStatus = atomic_val_compare_exchange_8(&pTask->status.schedStatus, TASK_SCHED_STATUS__INACTIVE, - TASK_SCHED_STATUS__WAITING); - if (schedStatus != TASK_SCHED_STATUS__INACTIVE) { - ASSERT(0); - return 0; + + // we have to continue retrying to successfully execute the scan history task. + while (1) { + int8_t schedStatus = atomic_val_compare_exchange_8(&pTask->status.schedStatus, TASK_SCHED_STATUS__INACTIVE, + TASK_SCHED_STATUS__WAITING); + if (schedStatus == TASK_SCHED_STATUS__INACTIVE) { + break; + } + + tqError("s-task:%s failed to start scan history in current time window, unexpected sched-status:%d, retry in 100ms", + id, schedStatus); + taosMsleep(100); + } + + ASSERT(pTask->status.pauseAllowed == false); + + if (pTask->info.fillHistory == 1) { + streamTaskEnablePause(pTask); } if (!streamTaskRecoverScanStep1Finished(pTask)) { streamSourceScanHistoryData(pTask); } + // disable the pause when handling the step2 scan of tsdb data. + // the whole next procedure cann't be stopped. + // todo fix it: the following procedure should be executed completed and then shutdown when trying to close vnode. + if (pTask->info.fillHistory == 1) { + streamTaskDisablePause(pTask); + } + if (atomic_load_8(&pTask->status.taskStatus) == TASK_STATUS__DROPPING || streamTaskShouldPause(&pTask->status)) { - tqDebug("s-task:%s is dropped or paused, abort recover in step1", pId); + tqDebug("s-task:%s is dropped or paused, abort recover in step1", id); atomic_store_8(&pTask->status.schedStatus, TASK_SCHED_STATUS__INACTIVE); streamMetaReleaseTask(pMeta, pTask); return 0; } double el = (taosGetTimestampMs() - st) / 1000.0; - tqDebug("s-task:%s history data scan stage(step 1) ended, elapsed time:%.2fs", pId, el); + tqDebug("s-task:%s history data scan stage(step 1) ended, elapsed time:%.2fs", id, el); if (pTask->info.fillHistory) { SVersionRange* pRange = NULL; @@ -1142,11 +1154,11 @@ int32_t tqProcessTaskScanHistory(STQ* pTq, SRpcMsg* pMsg) { // 1. stop the related stream task, get the current scan wal version of stream task, ver. pStreamTask = streamMetaAcquireTask(pMeta, pTask->streamTaskId.taskId); if (pStreamTask == NULL) { - qError("failed to find s-task:0x%x, it may have been destroyed, drop fill history task:%s", + qError("failed to find s-task:0x%x, it may have been destroyed, drop fill-history task:%s", pTask->streamTaskId.taskId, pTask->id.idStr); pTask->status.taskStatus = TASK_STATUS__DROPPING; - tqDebug("s-task:%s scan-history-task set status to be dropping", pId); + tqDebug("s-task:%s scan-history-task set status to be dropping", id); streamMetaSaveTask(pMeta, pTask); streamMetaReleaseTask(pMeta, pTask); @@ -1155,28 +1167,30 @@ int32_t tqProcessTaskScanHistory(STQ* pTq, SRpcMsg* pMsg) { ASSERT(pStreamTask->info.taskLevel == TASK_LEVEL__SOURCE); + // stream task in TASK_STATUS__SCAN_HISTORY can not be paused. // wait for the stream task get ready for scan history data - while (((pStreamTask->status.downstreamReady == 0) && (pStreamTask->status.taskStatus != TASK_STATUS__STOP)) || - pStreamTask->status.taskStatus == TASK_STATUS__SCAN_HISTORY) { + while (pStreamTask->status.taskStatus == TASK_STATUS__SCAN_HISTORY) { tqDebug( "s-task:%s level:%d related stream task:%s(status:%s) not ready for halt, wait for it and recheck in 100ms", - pId, pTask->info.taskLevel, pStreamTask->id.idStr, streamGetTaskStatusStr(pStreamTask->status.taskStatus)); + id, pTask->info.taskLevel, pStreamTask->id.idStr, streamGetTaskStatusStr(pStreamTask->status.taskStatus)); taosMsleep(100); } + streamTaskHalt(pTask); + // now we can stop the stream task execution + // todo upgrade the statu to be HALT from PAUSE or NORMAL pStreamTask->status.taskStatus = TASK_STATUS__HALT; - tqDebug("s-task:%s level:%d status is set to halt by history scan task:%s", pStreamTask->id.idStr, - pStreamTask->info.taskLevel, pId); + tqDebug("s-task:%s level:%d status is set to halt by fill-history task:%s", pStreamTask->id.idStr, + pStreamTask->info.taskLevel, id); // if it's an source task, extract the last version in wal. streamHistoryTaskSetVerRangeStep2(pTask); } if (!streamTaskRecoverScanStep1Finished(pTask)) { - tqDebug("s-task:%s level:%d verRange:%" PRId64 " - %" PRId64 - " do secondary scan-history-data after halt the related stream task:%s", - pId, pTask->info.taskLevel, pRange->minVer, pRange->maxVer, pId); + tqDebug("s-task:%s level:%d verRange:%" PRId64 " - %" PRId64 " do secondary scan-history-data after halt the related stream task:%s", + id, pTask->info.taskLevel, pRange->minVer, pRange->maxVer, id); ASSERT(pTask->status.schedStatus == TASK_SCHED_STATUS__WAITING); st = taosGetTimestampMs(); @@ -1187,7 +1201,7 @@ int32_t tqProcessTaskScanHistory(STQ* pTq, SRpcMsg* pMsg) { streamSourceScanHistoryData(pTask); if (atomic_load_8(&pTask->status.taskStatus) == TASK_STATUS__DROPPING || streamTaskShouldPause(&pTask->status)) { - tqDebug("s-task:%s is dropped or paused, abort recover in step1", pId); + tqDebug("s-task:%s is dropped or paused, abort recover in step1", id); streamMetaReleaseTask(pMeta, pTask); return 0; } @@ -1196,7 +1210,7 @@ int32_t tqProcessTaskScanHistory(STQ* pTq, SRpcMsg* pMsg) { } el = (taosGetTimestampMs() - st) / 1000.0; - tqDebug("s-task:%s history data scan stage(step 2) ended, elapsed time:%.2fs", pId, el); + tqDebug("s-task:%s history data scan stage(step 2) ended, elapsed time:%.2fs", id, el); // 3. notify downstream tasks to transfer executor state after handle all history blocks. if (!pTask->status.transferState) { @@ -1212,20 +1226,8 @@ int32_t tqProcessTaskScanHistory(STQ* pTq, SRpcMsg* pMsg) { // 5. resume the related stream task. streamTryExec(pTask); - pTask->status.taskStatus = TASK_STATUS__DROPPING; - tqDebug("s-task:%s scan-history-task set status to be dropping", pId); - - streamMetaSaveTask(pMeta, pTask); - streamMetaSaveTask(pMeta, pStreamTask); - streamMetaReleaseTask(pMeta, pTask); streamMetaReleaseTask(pMeta, pStreamTask); - - taosWLockLatch(&pMeta->lock); - if (streamMetaCommit(pTask->pMeta) < 0) { - // persist to disk - } - taosWUnLockLatch(&pMeta->lock); } else { // todo update the chkInfo version for current task. // this task has an associated history stream task, so we need to scan wal from the end version of @@ -1234,24 +1236,23 @@ int32_t tqProcessTaskScanHistory(STQ* pTq, SRpcMsg* pMsg) { if (pTask->historyTaskId.taskId == 0) { *pWindow = (STimeWindow){INT64_MIN, INT64_MAX}; - tqDebug("s-task:%s no related scan-history-data task, reset the time window:%" PRId64 " - %" PRId64, pId, - pWindow->skey, pWindow->ekey); + tqDebug( + "s-task:%s scan history in stream time window completed, no related fill history task, reset the time " + "window:%" PRId64 " - %" PRId64, + id, pWindow->skey, pWindow->ekey); } else { tqDebug( - "s-task:%s history data in current time window scan completed, now start to handle data from WAL, start " + "s-task:%s scan history in stream time window completed, now start to handle data from WAL, start " "ver:%" PRId64 ", window:%" PRId64 " - %" PRId64, - pId, pTask->chkInfo.currentVer, pWindow->skey, pWindow->ekey); + id, pTask->chkInfo.currentVer, pWindow->skey, pWindow->ekey); } // notify the downstream agg tasks that upstream tasks are ready to processing the WAL data, update the code = streamTaskScanHistoryDataComplete(pTask); streamMetaReleaseTask(pMeta, pTask); - // let's start the stream task by extracting data from wal - if (pTask->info.taskLevel == TASK_LEVEL__SOURCE) { - tqStartStreamTasks(pTq); - } - + // when all source task complete to scan history data in stream time window, they are allowed to handle stream data + // at the same time. return code; } @@ -1270,35 +1271,24 @@ int32_t tqProcessTaskTransferStateReq(STQ* pTq, SRpcMsg* pMsg) { int32_t code = tDecodeStreamScanHistoryFinishReq(&decoder, &req); tDecoderClear(&decoder); - SStreamTask* pTask = streamMetaAcquireTask(pTq->pStreamMeta, req.taskId); + tqDebug("vgId:%d start to process transfer state msg, from s-task:0x%x", pTq->pStreamMeta->vgId, req.downstreamTaskId); + + SStreamTask* pTask = streamMetaAcquireTask(pTq->pStreamMeta, req.downstreamTaskId); if (pTask == NULL) { - tqError("failed to find task:0x%x, it may have been dropped already", req.taskId); + tqError("failed to find task:0x%x, it may have been dropped already. process transfer state failed", req.downstreamTaskId); return -1; } int32_t remain = streamAlignTransferState(pTask); if (remain > 0) { - tqDebug("s-task:%s receive transfer state msg, remain:%d", pTask->id.idStr, remain); + tqDebug("s-task:%s receive upstream transfer state msg, remain:%d", pTask->id.idStr, remain); return 0; } // transfer the ownership of executor state - tqDebug("s-task:%s all upstream tasks end transfer msg", pTask->id.idStr); + tqDebug("s-task:%s all upstream tasks send transfer msg, open transfer state flag", pTask->id.idStr); + ASSERT(pTask->streamTaskId.taskId != 0 && pTask->info.fillHistory == 1); - // related stream task load the state from the state storage backend - SStreamTask* pStreamTask = streamMetaAcquireTask(pTq->pStreamMeta, pTask->streamTaskId.taskId); - if (pStreamTask == NULL) { - streamMetaReleaseTask(pTq->pStreamMeta, pTask); - tqError("failed to find related stream task:0x%x, it may have been dropped already", req.taskId); - return -1; - } - - // when all upstream tasks have notified the this task to start transfer state, then we start the transfer procedure. - streamTaskReleaseState(pTask); - streamTaskReloadState(pStreamTask); - streamMetaReleaseTask(pTq->pStreamMeta, pStreamTask); - - ASSERT(pTask->streamTaskId.taskId != 0); pTask->status.transferState = true; streamSchedExec(pTask); @@ -1306,7 +1296,7 @@ int32_t tqProcessTaskTransferStateReq(STQ* pTq, SRpcMsg* pMsg) { return 0; } -int32_t tqProcessStreamTaskScanHistoryFinishReq(STQ* pTq, SRpcMsg* pMsg) { +int32_t tqProcessTaskScanHistoryFinishReq(STQ* pTq, SRpcMsg* pMsg) { char* msg = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)); int32_t msgLen = pMsg->contLen - sizeof(SMsgHead); @@ -1318,20 +1308,50 @@ int32_t tqProcessStreamTaskScanHistoryFinishReq(STQ* pTq, SRpcMsg* pMsg) { tDecodeStreamScanHistoryFinishReq(&decoder, &req); tDecoderClear(&decoder); - // find task - SStreamTask* pTask = streamMetaAcquireTask(pTq->pStreamMeta, req.taskId); + SStreamTask* pTask = streamMetaAcquireTask(pTq->pStreamMeta, req.downstreamTaskId); if (pTask == NULL) { - tqError("failed to find task:0x%x, it may be destroyed, vgId:%d", req.taskId, pTq->pStreamMeta->vgId); + tqError("vgId:%d process scan history finish msg, failed to find task:0x%x, it may be destroyed", + pTq->pStreamMeta->vgId, req.downstreamTaskId); return -1; } - int32_t code = streamProcessScanHistoryFinishReq(pTask, req.childId); + tqDebug("s-task:%s receive scan-history finish msg from task:0x%x", pTask->id.idStr, req.upstreamTaskId); + + int32_t code = streamProcessScanHistoryFinishReq(pTask, &req, &pMsg->info); streamMetaReleaseTask(pTq->pStreamMeta, pTask); return code; } -int32_t tqProcessTaskRecoverFinishRsp(STQ* pTq, SRpcMsg* pMsg) { - // +int32_t tqProcessTaskScanHistoryFinishRsp(STQ* pTq, SRpcMsg* pMsg) { + char* msg = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)); + int32_t msgLen = pMsg->contLen - sizeof(SMsgHead); + + // deserialize + SStreamCompleteHistoryMsg req = {0}; + + SDecoder decoder; + tDecoderInit(&decoder, (uint8_t*)msg, msgLen); + tDecodeCompleteHistoryDataMsg(&decoder, &req); + tDecoderClear(&decoder); + + SStreamTask* pTask = streamMetaAcquireTask(pTq->pStreamMeta, req.upstreamTaskId); + if (pTask == NULL) { + tqError("vgId:%d process scan history finish rsp, failed to find task:0x%x, it may be destroyed", + pTq->pStreamMeta->vgId, req.upstreamTaskId); + return -1; + } + + tqDebug("s-task:%s scan-history finish rsp received from downstream task:0x%x", pTask->id.idStr, req.downstreamId); + + int32_t remain = atomic_sub_fetch_32(&pTask->notReadyTasks, 1); + if (remain > 0) { + tqDebug("s-task:%s remain:%d not send finish rsp", pTask->id.idStr, remain); + } else { + tqDebug("s-task:%s all downstream tasks rsp scan-history completed msg", pTask->id.idStr); + streamProcessScanHistoryFinishRsp(pTask); + } + + streamMetaReleaseTask(pTq->pStreamMeta, pTask); return 0; } @@ -1352,26 +1372,23 @@ int32_t tqProcessTaskRunReq(STQ* pTq, SRpcMsg* pMsg) { } SStreamTask* pTask = streamMetaAcquireTask(pTq->pStreamMeta, taskId); - if (pTask == NULL) { - tqError("vgId:%d failed to found s-task, taskId:0x%x", vgId, taskId); - return -1; - } + if (pTask != NULL) { + // even in halt status, the data in inputQ must be processed + int8_t status = pTask->status.taskStatus; + if (status == TASK_STATUS__NORMAL || status == TASK_STATUS__SCAN_HISTORY || status == TASK_STATUS__CK) { + tqDebug("vgId:%d s-task:%s start to process block from inputQ, status:%s, checkpointVer:%" PRId64, vgId, + pTask->id.idStr, streamGetTaskStatusStr(status), pTask->chkInfo.checkpointVer); + streamProcessRunReq(pTask); + } else { + atomic_store_8(&pTask->status.schedStatus, TASK_SCHED_STATUS__INACTIVE); + tqDebug("vgId:%d s-task:%s ignore run req since not in ready state, status:%s, sched-status:%d", vgId, + pTask->id.idStr, streamGetTaskStatusStr(status), pTask->status.schedStatus); + } - // even in halt status, the data in inputQ must be processed - int8_t status = pTask->status.taskStatus; - if (status == TASK_STATUS__NORMAL || status == TASK_STATUS__HALT || status == TASK_STATUS__CK) { - tqDebug("vgId:%d s-task:%s start to process block from inputQ, status:%s, checkpointVer:%" PRId64, vgId, - pTask->id.idStr, streamGetTaskStatusStr(status), pTask->chkInfo.checkpointVer); - streamProcessRunReq(pTask); - } else { - atomic_store_8(&pTask->status.schedStatus, TASK_SCHED_STATUS__INACTIVE); - tqDebug("vgId:%d s-task:%s ignore run req since not in ready state, status:%s, sched-status:%d", vgId, - pTask->id.idStr, streamGetTaskStatusStr(status), pTask->status.schedStatus); + streamMetaReleaseTask(pTq->pStreamMeta, pTask); + tqStartStreamTasks(pTq); } - streamMetaReleaseTask(pTq->pStreamMeta, pTask); - tqStartStreamTasks(pTq); - return 0; } @@ -1425,32 +1442,45 @@ int32_t tqProcessTaskDropReq(STQ* pTq, int64_t sversion, char* msg, int32_t msgL return 0; } -int32_t tqProcessTaskPauseImpl(SStreamMeta* pStreamMeta, SStreamTask* pTask) { - if (pTask) { - if (!streamTaskShouldPause(&pTask->status)) { - tqDebug("vgId:%d s-task:%s set pause flag", pStreamMeta->vgId, pTask->id.idStr); - atomic_store_8(&pTask->status.keepTaskStatus, pTask->status.taskStatus); - atomic_store_8(&pTask->status.taskStatus, TASK_STATUS__PAUSE); - } - streamMetaReleaseTask(pStreamMeta, pTask); - } else { - return -1; - } - return 0; -} - int32_t tqProcessTaskPauseReq(STQ* pTq, int64_t sversion, char* msg, int32_t msgLen) { SVPauseStreamTaskReq* pReq = (SVPauseStreamTaskReq*)msg; - SStreamTask* pTask = streamMetaAcquireTask(pTq->pStreamMeta, pReq->taskId); - int32_t code = tqProcessTaskPauseImpl(pTq->pStreamMeta, pTask); - if (code != 0) { - return code; + + SStreamMeta* pMeta = pTq->pStreamMeta; + SStreamTask* pTask = streamMetaAcquireTask(pMeta, pReq->taskId); + if (pTask == NULL) { + tqError("vgId:%d failed to acquire task:0x%x, it may have been dropped already", pMeta->vgId, + pReq->taskId); + + // since task is in [STOP|DROPPING] state, it is safe to assume the pause is active + return TSDB_CODE_SUCCESS; } - SStreamTask* pHistoryTask = streamMetaAcquireTask(pTq->pStreamMeta, pTask->historyTaskId.taskId); - if (pHistoryTask) { - code = tqProcessTaskPauseImpl(pTq->pStreamMeta, pHistoryTask); + + tqDebug("s-task:%s receive pause msg from mnode", pTask->id.idStr); + streamTaskPause(pTask); + + SStreamTask* pHistoryTask = NULL; + if (pTask->historyTaskId.taskId != 0) { + pHistoryTask = streamMetaAcquireTask(pMeta, pTask->historyTaskId.taskId); + if (pHistoryTask == NULL) { + tqError("vgId:%d failed to acquire fill-history task:0x%x, it may have been dropped already. Pause success", + pMeta->vgId, pTask->historyTaskId.taskId); + + streamMetaReleaseTask(pMeta, pTask); + + // since task is in [STOP|DROPPING] state, it is safe to assume the pause is active + return TSDB_CODE_SUCCESS; + } + + tqDebug("s-task:%s fill-history task handle paused along with related stream task", pHistoryTask->id.idStr); + streamTaskPause(pHistoryTask); } - return code; + + streamMetaReleaseTask(pMeta, pTask); + if (pHistoryTask != NULL) { + streamMetaReleaseTask(pMeta, pHistoryTask); + } + + return TSDB_CODE_SUCCESS; } int32_t tqProcessTaskResumeImpl(STQ* pTq, SStreamTask* pTask, int64_t sversion, int8_t igUntreated) { @@ -1459,11 +1489,14 @@ int32_t tqProcessTaskResumeImpl(STQ* pTq, SStreamTask* pTask, int64_t sversion, return -1; } - if (streamTaskShouldPause(&pTask->status)) { - atomic_store_8(&pTask->status.taskStatus, pTask->status.keepTaskStatus); + // todo: handle the case: resume from halt to pause/ from halt to normal/ from pause to normal + streamTaskResume(pTask); + int32_t level = pTask->info.taskLevel; + int8_t status = pTask->status.taskStatus; + if (status == TASK_STATUS__NORMAL || status == TASK_STATUS__SCAN_HISTORY) { // no lock needs to secure the access of the version - if (igUntreated && pTask->info.taskLevel == TASK_LEVEL__SOURCE && !pTask->info.fillHistory) { + if (igUntreated && level == TASK_LEVEL__SOURCE && !pTask->info.fillHistory) { // discard all the data when the stream task is suspended. walReaderSetSkipToVersion(pTask->exec.pWalReader, sversion); tqDebug("vgId:%d s-task:%s resume to exec, prev paused version:%" PRId64 ", start from vnode ver:%" PRId64 @@ -1474,9 +1507,9 @@ int32_t tqProcessTaskResumeImpl(STQ* pTq, SStreamTask* pTask, int64_t sversion, vgId, pTask->id.idStr, pTask->chkInfo.currentVer, sversion, pTask->status.schedStatus); } - if (pTask->info.fillHistory && pTask->info.taskLevel == TASK_LEVEL__SOURCE) { + if (level == TASK_LEVEL__SOURCE && pTask->info.fillHistory) { streamStartRecoverTask(pTask, igUntreated); - } else if (pTask->info.taskLevel == TASK_LEVEL__SOURCE && taosQueueItemSize(pTask->inputQueue->queue) == 0) { + } else if (level == TASK_LEVEL__SOURCE && (taosQueueItemSize(pTask->inputQueue->queue) == 0)) { tqStartStreamTasks(pTq); } else { streamSchedExec(pTask); @@ -1552,9 +1585,8 @@ int32_t vnodeEnqueueStreamMsg(SVnode* pVnode, SRpcMsg* pMsg) { tDecoderClear(&decoder); int32_t taskId = req.taskId; - SStreamTask* pTask = streamMetaAcquireTask(pTq->pStreamMeta, taskId); - if (pTask) { + if (pTask != NULL) { SRpcMsg rsp = {.info = pMsg->info, .code = 0}; streamProcessDispatchMsg(pTask, &req, &rsp, false); streamMetaReleaseTask(pTq->pStreamMeta, pTask); @@ -1562,18 +1594,22 @@ int32_t vnodeEnqueueStreamMsg(SVnode* pVnode, SRpcMsg* pMsg) { taosFreeQitem(pMsg); return 0; } else { + tDeleteStreamDispatchReq(&req); } code = TSDB_CODE_STREAM_TASK_NOT_EXIST; FAIL: - if (pMsg->info.handle == NULL) return -1; + if (pMsg->info.handle == NULL) { + tqError("s-task:0x%x vgId:%d msg handle is null, abort enqueue dispatch msg", pTq->pStreamMeta->vgId, taskId); + return -1; + } SMsgHead* pRspHead = rpcMallocCont(sizeof(SMsgHead) + sizeof(SStreamDispatchRsp)); if (pRspHead == NULL) { SRpcMsg rsp = {.code = TSDB_CODE_OUT_OF_MEMORY, .info = pMsg->info}; - tqDebug("send dispatch error rsp, code: %x", code); + tqError("s-task:0x%x send dispatch error rsp, code:%s", taskId, tstrerror(code)); tmsgSendRsp(&rsp); rpcFreeCont(pMsg->pCont); taosFreeQitem(pMsg); @@ -1589,9 +1625,10 @@ FAIL: pRsp->downstreamTaskId = htonl(req.taskId); pRsp->inputStatus = TASK_OUTPUT_STATUS__NORMAL; - SRpcMsg rsp = { - .code = code, .info = pMsg->info, .contLen = sizeof(SMsgHead) + sizeof(SStreamDispatchRsp), .pCont = pRspHead}; - tqDebug("send dispatch error rsp, code: %x", code); + int32_t len = sizeof(SMsgHead) + sizeof(SStreamDispatchRsp); + SRpcMsg rsp = { .code = code, .info = pMsg->info, .contLen = len, .pCont = pRspHead}; + tqError("s-task:0x%x send dispatch error rsp, code:%s", taskId, tstrerror(code)); + tmsgSendRsp(&rsp); rpcFreeCont(pMsg->pCont); taosFreeQitem(pMsg); diff --git a/source/dnode/vnode/src/tq/tqRestore.c b/source/dnode/vnode/src/tq/tqRestore.c index 70277a8356..a0680b0184 100644 --- a/source/dnode/vnode/src/tq/tqRestore.c +++ b/source/dnode/vnode/src/tq/tqRestore.c @@ -81,7 +81,13 @@ int32_t tqStreamTasksStatusCheck(STQ* pTq) { continue; } - streamTaskCheckDownstreamTasks(pTask); + if (pTask->info.fillHistory == 1) { + tqDebug("s-task:%s fill-history task, wait for related stream task:0x%x to launch it", pTask->id.idStr, + pTask->streamTaskId.taskId); + continue; + } + + streamTaskDoCheckDownstreamTasks(pTask); streamMetaReleaseTask(pMeta, pTask); } @@ -234,12 +240,14 @@ int32_t createStreamTaskRunReq(SStreamMeta* pStreamMeta, bool* pScanIdle) { continue; } - if (pTask->info.taskLevel != TASK_LEVEL__SOURCE) { + int32_t status = pTask->status.taskStatus; + + // non-source or fill-history tasks don't need to response the WAL scan action. + if (pTask->info.taskLevel != TASK_LEVEL__SOURCE || pTask->info.fillHistory == 1) { streamMetaReleaseTask(pStreamMeta, pTask); continue; } - int32_t status = pTask->status.taskStatus; const char* pStatus = streamGetTaskStatusStr(status); if (status != TASK_STATUS__NORMAL) { tqDebug("s-task:%s not ready for new submit block from wal, status:%s", pTask->id.idStr, pStatus); diff --git a/source/dnode/vnode/src/tsdb/tsdbCache.c b/source/dnode/vnode/src/tsdb/tsdbCache.c index d0986b25f0..db07223106 100644 --- a/source/dnode/vnode/src/tsdb/tsdbCache.c +++ b/source/dnode/vnode/src/tsdb/tsdbCache.c @@ -13,6 +13,8 @@ * along with this program. If not, see . */ #include "tsdb.h" +#include "tsdbDataFileRW.h" +#include "tsdbReadUtil.h" #include "vnd.h" #define ROCKS_BATCH_SIZE (4096) @@ -1689,188 +1691,188 @@ _err: } return code; } -/* -static int32_t getTableDelIdx(SDelFReader *pDelFReader, tb_uid_t suid, tb_uid_t uid, SDelIdx *pDelIdx) { - int32_t code = 0; - SArray *pDelIdxArray = NULL; - // SMapData delIdxMap; - pDelIdxArray = taosArrayInit(32, sizeof(SDelIdx)); - SDelIdx idx = {.suid = suid, .uid = uid}; +static int32_t loadTombFromBlk(const TTombBlkArray *pTombBlkArray, SCacheRowsReader *pReader, void *pFileReader, + bool isFile) { + int32_t code = 0; + uint64_t *uidList = pReader->uidList; + int32_t numOfTables = pReader->numOfTables; + int64_t suid = pReader->info.suid; - // tMapDataReset(&delIdxMap); - code = tsdbReadDelIdx(pDelFReader, pDelIdxArray); - if (code) goto _err; - - // code = tMapDataSearch(&delIdxMap, &idx, tGetDelIdx, tCmprDelIdx, pDelIdx); - SDelIdx *pIdx = taosArraySearch(pDelIdxArray, &idx, tCmprDelIdx, TD_EQ); - - *pDelIdx = *pIdx; - -_err: - if (pDelIdxArray) { - taosArrayDestroy(pDelIdxArray); - } - return code; -} -*/ -typedef enum { - SFSLASTNEXTROW_FS, - SFSLASTNEXTROW_FILESET, - SFSLASTNEXTROW_BLOCKDATA, - SFSLASTNEXTROW_BLOCKROW -} SFSLASTNEXTROWSTATES; - -typedef struct { - SFSLASTNEXTROWSTATES state; // [input] - STsdb *pTsdb; // [input] - STSchema *pTSchema; // [input] - tb_uid_t suid; - tb_uid_t uid; - int32_t nFileSet; - int32_t iFileSet; - SArray *aDFileSet; - SDataFReader **pDataFReader; - TSDBROW row; - - bool checkRemainingRow; - SMergeTree mergeTree; - SMergeTree *pMergeTree; - SSttBlockLoadInfo *pLoadInfo; - SLDataIter *pDataIter; - int64_t lastTs; -} SFSLastNextRowIter; - -static int32_t getNextRowFromFSLast(void *iter, TSDBROW **ppRow, bool *pIgnoreEarlierTs, bool isLast, int16_t *aCols, - int nCols) { - SFSLastNextRowIter *state = (SFSLastNextRowIter *)iter; - int32_t code = 0; - bool checkRemainingRow = true; - - switch (state->state) { - case SFSLASTNEXTROW_FS: - state->nFileSet = taosArrayGetSize(state->aDFileSet); - state->iFileSet = state->nFileSet; - - case SFSLASTNEXTROW_FILESET: { - SDFileSet *pFileSet = NULL; - _next_fileset: - if (state->pMergeTree != NULL) { - tMergeTreeClose(state->pMergeTree); - state->pMergeTree = NULL; - } - - if (--state->iFileSet >= 0) { - pFileSet = (SDFileSet *)taosArrayGet(state->aDFileSet, state->iFileSet); - } else { - *ppRow = NULL; - return code; - } - - if (*state->pDataFReader == NULL || (*state->pDataFReader)->pSet->fid != pFileSet->fid) { - if (*state->pDataFReader != NULL) { - tsdbDataFReaderClose(state->pDataFReader); - - resetLastBlockLoadInfo(state->pLoadInfo); - } - - code = tsdbDataFReaderOpen(state->pDataFReader, state->pTsdb, pFileSet); - if (code) goto _err; - } - - int nTmpCols = nCols; - bool hasTs = false; - if (aCols[0] == PRIMARYKEY_TIMESTAMP_COL_ID) { - --nTmpCols; - hasTs = true; - } - for (int i = 0; i < state->pLoadInfo->numOfStt; ++i) { - state->pLoadInfo[i].colIds = hasTs ? aCols + 1 : aCols; - state->pLoadInfo[i].numOfCols = nTmpCols; - state->pLoadInfo[i].isLast = isLast; - } - tMergeTreeOpen(&state->mergeTree, 1, *state->pDataFReader, state->suid, state->uid, - &(STimeWindow){.skey = state->lastTs, .ekey = TSKEY_MAX}, - &(SVersionRange){.minVer = 0, .maxVer = UINT64_MAX}, state->pLoadInfo, false, NULL, true, - state->pDataIter); - state->pMergeTree = &state->mergeTree; - state->state = SFSLASTNEXTROW_BLOCKROW; + for (int i = 0, j = 0; i < pTombBlkArray->size && j < numOfTables; ++i) { + STombBlk *pTombBlk = &pTombBlkArray->data[i]; + if (pTombBlk->maxTbid.suid < suid || (pTombBlk->maxTbid.suid == suid && pTombBlk->maxTbid.uid < uidList[0])) { + continue; } - case SFSLASTNEXTROW_BLOCKROW: { - if (nCols != state->pLoadInfo->numOfCols) { - for (int i = 0; i < state->pLoadInfo->numOfStt; ++i) { - state->pLoadInfo[i].numOfCols = nCols; - state->pLoadInfo[i].checkRemainingRow = state->checkRemainingRow; - } - } - bool hasVal = tMergeTreeNext(&state->mergeTree); - if (!hasVal) { - if (tMergeTreeIgnoreEarlierTs(&state->mergeTree)) { - *pIgnoreEarlierTs = true; - *ppRow = NULL; - return code; - } - state->state = SFSLASTNEXTROW_FILESET; - goto _next_fileset; - } - state->row = *tMergeTreeGetRow(&state->mergeTree); - *ppRow = &state->row; + if (pTombBlk->minTbid.suid > suid || + (pTombBlk->minTbid.suid == suid && pTombBlk->minTbid.uid > uidList[numOfTables - 1])) { + break; + } - if (TSDBROW_TS(&state->row) <= state->lastTs) { - *pIgnoreEarlierTs = true; - *ppRow = NULL; - return code; - } - - *pIgnoreEarlierTs = false; - /* - if (!hasVal) { - state->state = SFSLASTNEXTROW_FILESET; - } - */ - if (!state->checkRemainingRow) { - state->checkRemainingRow = true; - } + STombBlock block = {0}; + code = isFile ? tsdbDataFileReadTombBlock(pFileReader, &pTombBlkArray->data[i], &block) + : tsdbSttFileReadTombBlock(pFileReader, &pTombBlkArray->data[i], &block); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + uint64_t uid = uidList[j]; + STableLoadInfo *pInfo = *(STableLoadInfo **)tSimpleHashGet(pReader->pTableMap, &uid, sizeof(uid)); + if (pInfo->pTombData == NULL) { + pInfo->pTombData = taosArrayInit(4, sizeof(SDelData)); + } + + STombRecord record = {0}; + bool finished = false; + for (int32_t k = 0; k < TARRAY2_SIZE(block.suid); ++k) { + code = tTombBlockGet(&block, k, &record); + if (code != TSDB_CODE_SUCCESS) { + finished = true; + break; + } + + if (record.suid < suid) { + continue; + } + if (record.suid > suid) { + finished = true; + break; + } + + bool newTable = false; + if (uid < record.uid) { + while (j < numOfTables && uidList[j] < record.uid) { + ++j; + newTable = true; + } + + if (j >= numOfTables) { + finished = true; + break; + } + + uid = uidList[j]; + } + + if (record.uid < uid) { + continue; + } + + if (newTable) { + pInfo = *(STableLoadInfo **)tSimpleHashGet(pReader->pTableMap, &uid, sizeof(uid)); + if (pInfo->pTombData == NULL) { + pInfo->pTombData = taosArrayInit(4, sizeof(SDelData)); + } + } + + if (record.version <= pReader->info.verRange.maxVer) { + SDelData delData = {.version = record.version, .sKey = record.skey, .eKey = record.ekey}; + taosArrayPush(pInfo->pTombData, &delData); + } + } + + tTombBlockDestroy(&block); + + if (finished) { return code; } - default: - ASSERT(0); - break; } -_err: - /*if (state->pDataFReader) { - tsdbDataFReaderClose(&state->pDataFReader); - state->pDataFReader = NULL; - }*/ - if (state->pMergeTree != NULL) { - tMergeTreeClose(state->pMergeTree); - state->pMergeTree = NULL; + return TSDB_CODE_SUCCESS; +} + +static int32_t loadDataTomb(SCacheRowsReader *pReader, SDataFileReader *pFileReader) { + int32_t code = 0; + + const TTombBlkArray *pBlkArray = NULL; + code = tsdbDataFileReadTombBlk(pFileReader, &pBlkArray); + if (code != TSDB_CODE_SUCCESS) { + return code; } - *ppRow = NULL; + return loadTombFromBlk(pBlkArray, pReader, pFileReader, true); +} + +static int32_t loadSttTomb(STsdbReader *pTsdbReader, SSttFileReader *pSttFileReader, SSttBlockLoadInfo *pLoadInfo) { + int32_t code = 0; + + SCacheRowsReader *pReader = (SCacheRowsReader *)pTsdbReader; + + const TTombBlkArray *pBlkArray = NULL; + code = tsdbSttFileReadTombBlk(pSttFileReader, &pBlkArray); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + return loadTombFromBlk(pBlkArray, pReader, pSttFileReader, false); +} + +typedef struct { + SMergeTree mergeTree; + SMergeTree *pMergeTree; +} SFSLastIter; + +static int32_t lastIterOpen(SFSLastIter *iter, STFileSet *pFileSet, STsdb *pTsdb, STSchema *pTSchema, tb_uid_t suid, + tb_uid_t uid, SCacheRowsReader *pr, int64_t lastTs, int16_t *aCols, int nCols) { + int32_t code = 0; + + int64_t loadBlocks = 0; + double elapse = 0; + pr->pLDataIterArray = destroySttBlockReader(pr->pLDataIterArray, &loadBlocks, &elapse); + pr->pLDataIterArray = taosArrayInit(4, POINTER_BYTES); + + SMergeTreeConf conf = { + .uid = uid, + .suid = suid, + .pTsdb = pTsdb, + .timewindow = (STimeWindow){.skey = lastTs, .ekey = TSKEY_MAX}, + .verRange = (SVersionRange){.minVer = 0, .maxVer = UINT64_MAX}, + .strictTimeRange = false, + .pSchema = pTSchema, + .pCurrentFileset = pFileSet, + .backward = 1, + .pSttFileBlockIterArray = pr->pLDataIterArray, + .pCols = aCols, + .numOfCols = nCols, + .loadTombFn = loadSttTomb, + .pReader = pr, + .idstr = pr->idstr, + }; + + code = tMergeTreeOpen2(&iter->mergeTree, &conf); + if (code != TSDB_CODE_SUCCESS) { + return -1; + } + + iter->pMergeTree = &iter->mergeTree; return code; } -int32_t clearNextRowFromFSLast(void *iter) { - SFSLastNextRowIter *state = (SFSLastNextRowIter *)iter; - int32_t code = 0; +static int32_t lastIterClose(SFSLastIter **iter) { + int32_t code = 0; - if (!state) { + if ((*iter)->pMergeTree) { + tMergeTreeClose((*iter)->pMergeTree); + (*iter)->pMergeTree = NULL; + } + + *iter = NULL; + + return code; +} + +static int32_t lastIterNext(SFSLastIter *iter, TSDBROW **ppRow) { + int32_t code = 0; + + bool hasVal = tMergeTreeNext(iter->pMergeTree); + if (!hasVal) { + *ppRow = NULL; return code; } - /* - if (state->pDataFReader) { - tsdbDataFReaderClose(&state->pDataFReader); - state->pDataFReader = NULL; - } - */ - if (state->pMergeTree != NULL) { - tMergeTreeClose(state->pMergeTree); - state->pMergeTree = NULL; - } + + *ppRow = tMergeTreeGetRow(iter->pMergeTree); return code; } @@ -1878,309 +1880,334 @@ int32_t clearNextRowFromFSLast(void *iter) { typedef enum SFSNEXTROWSTATES { SFSNEXTROW_FS, SFSNEXTROW_FILESET, + SFSNEXTROW_INDEXLIST, + SFSNEXTROW_BRINBLOCK, + SFSNEXTROW_BRINRECORD, SFSNEXTROW_BLOCKDATA, - SFSNEXTROW_BLOCKROW + SFSNEXTROW_BLOCKROW, + SFSNEXTROW_NEXTSTTROW } SFSNEXTROWSTATES; +struct CacheNextRowIter; + typedef struct SFSNextRowIter { - SFSNEXTROWSTATES state; // [input] - STsdb *pTsdb; // [input] - SBlockIdx *pBlockIdxExp; // [input] - STSchema *pTSchema; // [input] - tb_uid_t suid; - tb_uid_t uid; - int32_t nFileSet; - int32_t iFileSet; - SArray *aDFileSet; - SDataFReader **pDataFReader; - SArray *aBlockIdx; - LRUHandle *aBlockIdxHandle; - SBlockIdx *pBlockIdx; - SMapData blockMap; - int32_t nBlock; - int32_t iBlock; - SDataBlk block; - SBlockData blockData; - SBlockData *pBlockData; - int32_t nRow; - int32_t iRow; - TSDBROW row; - SSttBlockLoadInfo *pLoadInfo; - int64_t lastTs; + SFSNEXTROWSTATES state; // [input] + STsdb *pTsdb; // [input] + SBlockIdx *pBlockIdxExp; // [input] + STSchema *pTSchema; // [input] + tb_uid_t suid; + tb_uid_t uid; + int32_t nFileSet; + int32_t iFileSet; + STFileSet *pFileSet; + TFileSetArray *aDFileSet; + SArray *pIndexList; + int32_t iBrinIndex; + SBrinBlock brinBlock; + int32_t iBrinRecord; + SBrinRecord brinRecord; + SBlockData blockData; + SBlockData *pBlockData; + int32_t nRow; + int32_t iRow; + TSDBROW row; + int64_t lastTs; + SFSLastIter lastIter; + SFSLastIter *pLastIter; + int8_t lastEmpty; + TSDBROW *pLastRow; + SRow *pTSRow; + SRowMerger rowMerger; + SCacheRowsReader *pr; + struct CacheNextRowIter *pRowIter; } SFSNextRowIter; +static void clearLastFileSet(SFSNextRowIter *state); + static int32_t getNextRowFromFS(void *iter, TSDBROW **ppRow, bool *pIgnoreEarlierTs, bool isLast, int16_t *aCols, int nCols) { SFSNextRowIter *state = (SFSNextRowIter *)iter; int32_t code = 0; - bool checkRemainingRow = true; - switch (state->state) { - case SFSNEXTROW_FS: - // state->aDFileSet = state->pTsdb->pFS->cState->aDFileSet; - state->nFileSet = taosArrayGetSize(state->aDFileSet); - state->iFileSet = state->nFileSet; + if (SFSNEXTROW_FS == state->state) { + state->nFileSet = TARRAY2_SIZE(state->aDFileSet); + state->iFileSet = state->nFileSet; - state->pBlockData = NULL; + state->state = SFSNEXTROW_FILESET; + } - case SFSNEXTROW_FILESET: { - SDFileSet *pFileSet = NULL; - _next_fileset: - if (--state->iFileSet >= 0) { - pFileSet = (SDFileSet *)taosArrayGet(state->aDFileSet, state->iFileSet); - } else { - // tBlockDataDestroy(&state->blockData, 1); - if (state->pBlockData) { - tBlockDataDestroy(state->pBlockData); - state->pBlockData = NULL; - } + if (SFSNEXTROW_FILESET == state->state) { + _next_fileset: + if (--state->iFileSet < 0) { + clearLastFileSet(state); - *ppRow = NULL; - return code; + *ppRow = NULL; + return code; + } else { + state->pFileSet = TARRAY2_GET(state->aDFileSet, state->iFileSet); + } + + STFileObj **pFileObj = state->pFileSet->farr; + if (pFileObj[0] != NULL || pFileObj[3] != NULL) { + SDataFileReaderConfig conf = {.tsdb = state->pTsdb, .szPage = state->pTsdb->pVnode->config.szPage}; + const char *filesName[4] = {0}; + if (pFileObj[0] != NULL) { + conf.files[0].file = *pFileObj[0]->f; + conf.files[0].exist = true; + filesName[0] = pFileObj[0]->fname; + + conf.files[1].file = *pFileObj[1]->f; + conf.files[1].exist = true; + filesName[1] = pFileObj[1]->fname; + + conf.files[2].file = *pFileObj[2]->f; + conf.files[2].exist = true; + filesName[2] = pFileObj[2]->fname; } - if (*state->pDataFReader == NULL || (*state->pDataFReader)->pSet->fid != pFileSet->fid) { - if (*state->pDataFReader != NULL) { - tsdbDataFReaderClose(state->pDataFReader); - - // resetLastBlockLoadInfo(state->pLoadInfo); - } - - code = tsdbDataFReaderOpen(state->pDataFReader, state->pTsdb, pFileSet); - if (code) goto _err; + if (pFileObj[3] != NULL) { + conf.files[3].exist = true; + conf.files[3].file = *pFileObj[3]->f; + filesName[3] = pFileObj[3]->fname; } - // tMapDataReset(&state->blockIdxMap); - /* - if (!state->aBlockIdx) { - state->aBlockIdx = taosArrayInit(0, sizeof(SBlockIdx)); - } else { - taosArrayClear(state->aBlockIdx); - } - code = tsdbReadBlockIdx(*state->pDataFReader, state->aBlockIdx); - if (code) goto _err; - */ - int32_t code = tsdbCacheGetBlockIdx(state->pTsdb->biCache, *state->pDataFReader, &state->aBlockIdxHandle); - if (code != TSDB_CODE_SUCCESS || state->aBlockIdxHandle == NULL) { + code = tsdbDataFileReaderOpen(filesName, &conf, &state->pr->pFileReader); + if (code != TSDB_CODE_SUCCESS) { goto _err; } - state->aBlockIdx = (SArray *)taosLRUCacheValue(state->pTsdb->biCache, state->aBlockIdxHandle); - /* if (state->pBlockIdx) { */ - /* } */ - /* code = tMapDataSearch(&state->blockIdxMap, state->pBlockIdxExp, tGetBlockIdx, tCmprBlockIdx, - * &state->blockIdx); - */ - state->pBlockIdx = taosArraySearch(state->aBlockIdx, state->pBlockIdxExp, tCmprBlockIdx, TD_EQ); - if (!state->pBlockIdx) { - tsdbBICacheRelease(state->pTsdb->biCache, state->aBlockIdxHandle); + loadDataTomb(state->pr, state->pr->pFileReader); - state->aBlockIdxHandle = NULL; - state->aBlockIdx = NULL; - /* - tsdbDataFReaderClose(state->pDataFReader); - *state->pDataFReader = NULL; - resetLastBlockLoadInfo(state->pLoadInfo);*/ + if (!state->pIndexList) { + state->pIndexList = taosArrayInit(1, sizeof(SBrinBlk)); + } else { + taosArrayClear(state->pIndexList); + } + const TBrinBlkArray *pBlkArray = NULL; + + int32_t code = tsdbDataFileReadBrinBlk(state->pr->pFileReader, &pBlkArray); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + + for (int i = TARRAY2_SIZE(pBlkArray) - 1; i >= 0; --i) { + SBrinBlk *pBrinBlk = &pBlkArray->data[i]; + if (state->suid >= pBrinBlk->minTbid.suid && state->suid <= pBrinBlk->maxTbid.suid) { + if (state->uid >= pBrinBlk->minTbid.uid && state->uid <= pBrinBlk->maxTbid.uid) { + taosArrayPush(state->pIndexList, pBrinBlk); + } + } else if (state->suid > pBrinBlk->maxTbid.suid || + (state->suid == pBrinBlk->maxTbid.suid && state->uid > pBrinBlk->maxTbid.uid)) { + break; + } + } + + int indexSize = TARRAY_SIZE(state->pIndexList); + if (indexSize <= 0) { + clearLastFileSet(state); + state->state = SFSNEXTROW_FILESET; goto _next_fileset; } - tMapDataReset(&state->blockMap); - /* - if (state->blockMap.pData != NULL) { - tMapDataClear(&state->blockMap); + state->state = SFSNEXTROW_INDEXLIST; + state->iBrinIndex = indexSize; + } + + code = lastIterOpen(&state->lastIter, state->pFileSet, state->pTsdb, state->pTSchema, state->suid, state->uid, + state->pr, state->lastTs, aCols, nCols); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + + code = lastIterNext(&state->lastIter, &state->pLastRow); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + + if (!state->pLastRow) { + state->lastEmpty = 1; + + if (SFSNEXTROW_INDEXLIST != state->state) { + clearLastFileSet(state); + goto _next_fileset; } - */ - code = tsdbReadDataBlk(*state->pDataFReader, state->pBlockIdx, &state->blockMap); - if (code) goto _err; + } else { + state->lastEmpty = 0; - state->nBlock = state->blockMap.nItem; - state->iBlock = state->nBlock - 1; + if (SFSNEXTROW_INDEXLIST != state->state) { + state->state = SFSNEXTROW_NEXTSTTROW; - if (!state->pBlockData) { - state->pBlockData = &state->blockData; - - code = tBlockDataCreate(&state->blockData); - if (code) goto _err; + *ppRow = state->pLastRow; + state->pLastRow = NULL; + return code; } } - case SFSNEXTROW_BLOCKDATA: - _next_datablock: - if (state->iBlock >= 0) { - SDataBlk block = {0}; - bool skipBlock = true; - int inputColIndex = 0; - tDataBlkReset(&block); - tBlockDataReset(state->pBlockData); + state->pLastIter = &state->lastIter; + } - tMapDataGetItemByIdx(&state->blockMap, state->iBlock, &block, tGetDataBlk); - if (block.maxKey.ts <= state->lastTs) { - *pIgnoreEarlierTs = true; + if (SFSNEXTROW_NEXTSTTROW == state->state) { + code = lastIterNext(&state->lastIter, &state->pLastRow); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } - tBlockDataDestroy(state->pBlockData); - state->pBlockData = NULL; + if (!state->pLastRow) { + lastIterClose(&state->pLastIter); - *ppRow = NULL; - return code; - } - *pIgnoreEarlierTs = false; - tBlockDataReset(state->pBlockData); - TABLEID tid = {.suid = state->suid, .uid = state->uid}; - int nTmpCols = nCols; - bool hasTs = false; - if (aCols[0] == PRIMARYKEY_TIMESTAMP_COL_ID) { - --nTmpCols; - skipBlock = false; - hasTs = true; - } - code = tBlockDataInit(state->pBlockData, &tid, state->pTSchema, hasTs ? aCols + 1 : aCols, nTmpCols); - if (code) goto _err; + clearLastFileSet(state); + state->state = SFSNEXTROW_FILESET; + goto _next_fileset; + } else { + *ppRow = state->pLastRow; + state->pLastRow = NULL; + return code; + } + } - code = tsdbReadDataBlock(*state->pDataFReader, &block, state->pBlockData); - if (code) goto _err; + if (SFSNEXTROW_INDEXLIST == state->state) { + SBrinBlk *pBrinBlk = NULL; + _next_brinindex: + if (--state->iBrinIndex < 0) { // no index left, goto next fileset + clearLastFileSet(state); + goto _next_fileset; + } else { + pBrinBlk = taosArrayGet(state->pIndexList, state->iBrinIndex); + } - for (int colIndex = 0; colIndex < state->pBlockData->nColData; ++colIndex) { - SColData *pColData = &state->pBlockData->aColData[colIndex]; + code = tsdbDataFileReadBrinBlock(state->pr->pFileReader, pBrinBlk, &state->brinBlock); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } - if (isLast && (pColData->flag & HAS_VALUE)) { - skipBlock = false; - break; - } /*else if (pColData->flag & (HAS_VALUE | HAS_NULL)) { - skipBlock = false; - break; - }*/ - } + state->iBrinRecord = BRIN_BLOCK_SIZE(&state->brinBlock) - 1; + state->state = SFSNEXTROW_BRINBLOCK; + } - if (!isLast) { - skipBlock = false; - } + if (SFSNEXTROW_BRINBLOCK == state->state) { + _next_brinrecord: + if (state->iBrinRecord < 0) { // empty brin block, goto _next_brinindex + tBrinBlockClear(&state->brinBlock); + goto _next_brinindex; + } + code = tBrinBlockGet(&state->brinBlock, state->iBrinRecord, &state->brinRecord); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } - if (skipBlock) { - if (--state->iBlock < 0) { - tsdbDataFReaderClose(state->pDataFReader); - *state->pDataFReader = NULL; - // resetLastBlockLoadInfo(state->pLoadInfo); + SBrinRecord *pRecord = &state->brinRecord; + if (pRecord->uid != state->uid) { + // TODO: goto next brin block early + --state->iBrinRecord; + goto _next_brinrecord; + } - if (state->aBlockIdx) { - // taosArrayDestroy(state->aBlockIdx); - tsdbBICacheRelease(state->pTsdb->biCache, state->aBlockIdxHandle); + state->state = SFSNEXTROW_BRINRECORD; + } - state->aBlockIdxHandle = NULL; - state->aBlockIdx = NULL; - } + if (SFSNEXTROW_BRINRECORD == state->state) { + SBrinRecord *pRecord = &state->brinRecord; - state->state = SFSNEXTROW_FILESET; - goto _next_fileset; - } else { - goto _next_datablock; - } - } + if (!state->pBlockData) { + state->pBlockData = &state->blockData; + code = tBlockDataCreate(&state->blockData); + if (code) goto _err; + } else { + tBlockDataReset(state->pBlockData); + } - state->nRow = state->blockData.nRow; - state->iRow = state->nRow - 1; + if (aCols[0] == PRIMARYKEY_TIMESTAMP_COL_ID) { + --nCols; + ++aCols; + } + code = tsdbDataFileReadBlockDataByColumn(state->pr->pFileReader, pRecord, state->pBlockData, state->pTSchema, aCols, + nCols); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } - state->state = SFSNEXTROW_BLOCKROW; - checkRemainingRow = false; + state->nRow = state->blockData.nRow; + state->iRow = state->nRow - 1; + + state->state = SFSNEXTROW_BLOCKROW; + } + + if (SFSNEXTROW_BLOCKROW == state->state) { + if (state->iRow < 0) { + --state->iBrinRecord; + goto _next_brinrecord; + } + + state->row = tsdbRowFromBlockData(state->pBlockData, state->iRow); + if (!state->pLastIter) { + *ppRow = &state->row; + --state->iRow; + return code; + } + + if (!state->pLastRow) { + // get next row from fslast and process with fs row, --state->Row if select fs row + code = lastIterNext(&state->lastIter, &state->pLastRow); + if (code != TSDB_CODE_SUCCESS) { + goto _err; } - case SFSNEXTROW_BLOCKROW: { - if (checkRemainingRow) { - bool skipBlock = true; - int inputColIndex = 0; - if (aCols[0] == PRIMARYKEY_TIMESTAMP_COL_ID) { - ++inputColIndex; - } - for (int colIndex = 0; colIndex < state->pBlockData->nColData; ++colIndex) { - SColData *pColData = &state->pBlockData->aColData[colIndex]; - int16_t cid = pColData->cid; + } - if (inputColIndex < nCols && cid == aCols[inputColIndex]) { - if (isLast && (pColData->flag & HAS_VALUE)) { - skipBlock = false; - break; - } /*else if (pColData->flag & (HAS_VALUE | HAS_NULL)) { - skipBlock = false; - break; - }*/ + if (!state->pLastRow) { + lastIterClose(&state->pLastIter); - ++inputColIndex; - } - } + *ppRow = &state->row; + --state->iRow; + return code; + } - if (!isLast) { - skipBlock = false; - } + // process state->pLastRow & state->row + TSKEY rowTs = TSDBROW_TS(&state->row); + TSKEY lastRowTs = TSDBROW_TS(state->pLastRow); + if (lastRowTs > rowTs) { + *ppRow = state->pLastRow; + state->pLastRow = NULL; + return code; + } else if (lastRowTs < rowTs) { + *ppRow = &state->row; + --state->iRow; + return code; + } else { + // TODO: merge rows and *ppRow = mergedRow + SRowMerger *pMerger = &state->rowMerger; + tsdbRowMergerInit(pMerger, state->pTSchema); - if (skipBlock) { - if (--state->iBlock < 0) { - tsdbDataFReaderClose(state->pDataFReader); - *state->pDataFReader = NULL; - // resetLastBlockLoadInfo(state->pLoadInfo); - - if (state->aBlockIdx) { - // taosArrayDestroy(state->aBlockIdx); - tsdbBICacheRelease(state->pTsdb->biCache, state->aBlockIdxHandle); - - state->aBlockIdxHandle = NULL; - state->aBlockIdx = NULL; - } - - state->state = SFSNEXTROW_FILESET; - goto _next_fileset; - } else { - goto _next_datablock; - } - } + code = tsdbRowMergerAdd(pMerger, &state->row, state->pTSchema); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + code = tsdbRowMergerAdd(pMerger, state->pLastRow, state->pTSchema); + if (code != TSDB_CODE_SUCCESS) { + goto _err; } - if (state->iRow >= 0) { - state->row = tsdbRowFromBlockData(state->pBlockData, state->iRow); - *ppRow = &state->row; - - if (--state->iRow < 0) { - state->state = SFSNEXTROW_BLOCKDATA; - if (--state->iBlock < 0) { - tsdbDataFReaderClose(state->pDataFReader); - *state->pDataFReader = NULL; - // resetLastBlockLoadInfo(state->pLoadInfo); - - if (state->aBlockIdx) { - // taosArrayDestroy(state->aBlockIdx); - tsdbBICacheRelease(state->pTsdb->biCache, state->aBlockIdxHandle); - - state->aBlockIdxHandle = NULL; - state->aBlockIdx = NULL; - } - - state->state = SFSNEXTROW_FILESET; - } - } + if (state->pTSRow) { + taosMemoryFree(state->pTSRow); + state->pTSRow = NULL; } + code = tsdbRowMergerGetRow(pMerger, &state->pTSRow); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + + state->row = tsdbRowFromTSRow(TSDBROW_VERSION(&state->row), state->pTSRow); + *ppRow = &state->row; + --state->iRow; + + tsdbRowMergerClear(pMerger); + return code; } - default: - ASSERT(0); - break; } _err: - /* - if (*state->pDataFReader) { - tsdbDataFReaderClose(state->pDataFReader); - *state->pDataFReader = NULL; - resetLastBlockLoadInfo(state->pLoadInfo); - }*/ - if (state->aBlockIdx) { - // taosArrayDestroy(state->aBlockIdx); - tsdbBICacheRelease(state->pTsdb->biCache, state->aBlockIdxHandle); - - state->aBlockIdxHandle = NULL; - state->aBlockIdx = NULL; - } - if (state->pBlockData) { - tBlockDataDestroy(state->pBlockData); - state->pBlockData = NULL; - } + clearLastFileSet(state); *ppRow = NULL; @@ -2194,27 +2221,8 @@ int32_t clearNextRowFromFS(void *iter) { if (!state) { return code; } - /* - if (state->pDataFReader) { - tsdbDataFReaderClose(&state->pDataFReader); - state->pDataFReader = NULL; - }*/ - if (state->aBlockIdx) { - // taosArrayDestroy(state->aBlockIdx); - tsdbBICacheRelease(state->pTsdb->biCache, state->aBlockIdxHandle); - state->aBlockIdxHandle = NULL; - state->aBlockIdx = NULL; - } - if (state->pBlockData) { - // tBlockDataDestroy(&state->blockData, 1); - tBlockDataDestroy(state->pBlockData); - state->pBlockData = NULL; - } - - if (state->blockMap.pData != NULL) { - tMapDataClear(&state->blockMap); - } + clearLastFileSet(state); return code; } @@ -2229,8 +2237,6 @@ typedef struct SMemNextRowIter { STbData *pMem; // [input] STbDataIter iter; // mem buffer skip list iterator int64_t lastTs; - // bool iterOpened; - // TSDBROW *curRow; } SMemNextRowIter; static int32_t getNextRowFromMem(void *iter, TSDBROW **ppRow, bool *pIgnoreEarlierTs, bool isLast, int16_t *aCols, @@ -2281,45 +2287,6 @@ _err: return code; } -/* static int32_t tsRowFromTsdbRow(STSchema *pTSchema, TSDBROW *pRow, STSRow **ppRow) { */ -/* int32_t code = 0; */ - -/* SColVal *pColVal = &(SColVal){0}; */ - -/* if (pRow->type == 0) { */ -/* *ppRow = tdRowDup(pRow->pTSRow); */ -/* } else { */ -/* SArray *pArray = taosArrayInit(pTSchema->numOfCols, sizeof(SColVal)); */ -/* if (pArray == NULL) { */ -/* code = TSDB_CODE_OUT_OF_MEMORY; */ -/* goto _exit; */ -/* } */ - -/* TSDBKEY key = TSDBROW_KEY(pRow); */ -/* STColumn *pTColumn = &pTSchema->columns[0]; */ -/* *pColVal = COL_VAL_VALUE(pTColumn->colId, pTColumn->type, (SValue){.ts = key.ts}); */ - -/* if (taosArrayPush(pArray, pColVal) == NULL) { */ -/* code = TSDB_CODE_OUT_OF_MEMORY; */ -/* goto _exit; */ -/* } */ - -/* for (int16_t iCol = 1; iCol < pTSchema->numOfCols; iCol++) { */ -/* tsdbRowGetColVal(pRow, pTSchema, iCol, pColVal); */ -/* if (taosArrayPush(pArray, pColVal) == NULL) { */ -/* code = TSDB_CODE_OUT_OF_MEMORY; */ -/* goto _exit; */ -/* } */ -/* } */ - -/* code = tdSTSRowNew(pArray, pTSchema, ppRow); */ -/* if (code) goto _exit; */ -/* } */ - -/* _exit: */ -/* return code; */ -/* } */ - static bool tsdbKeyDeleted(TSDBKEY *key, SArray *pSkyline, int64_t *iSkyline) { bool deleted = false; while (*iSkyline > 0) { @@ -2365,24 +2332,56 @@ typedef struct { _next_row_clear_fn_t nextRowClearFn; } TsdbNextRowState; -typedef struct { - SArray *pSkyline; - int64_t iSkyline; - - SBlockIdx idx; - SMemNextRowIter memState; - SMemNextRowIter imemState; - SFSLastNextRowIter fsLastState; - SFSNextRowIter fsState; - TSDBROW memRow, imemRow, fsLastRow, fsRow; - - TsdbNextRowState input[4]; - STsdb *pTsdb; +typedef struct CacheNextRowIter { + SArray *pMemDelData; + SArray *pSkyline; + int64_t iSkyline; + SBlockIdx idx; + SMemNextRowIter memState; + SMemNextRowIter imemState; + SFSNextRowIter fsState; + TSDBROW memRow, imemRow, fsLastRow, fsRow; + TsdbNextRowState input[3]; + SCacheRowsReader *pr; + STsdb *pTsdb; } CacheNextRowIter; +static void clearLastFileSet(SFSNextRowIter *state) { + if (state->pLastIter) { + lastIterClose(&state->pLastIter); + } + + if (state->pBlockData) { + tBlockDataDestroy(state->pBlockData); + state->pBlockData = NULL; + } + + if (state->pr->pFileReader) { + tsdbDataFileReaderClose(&state->pr->pFileReader); + state->pr->pFileReader = NULL; + } + + if (state->pTSRow) { + taosMemoryFree(state->pTSRow); + state->pTSRow = NULL; + } + + if (state->pRowIter->pSkyline) { + taosArrayDestroy(state->pRowIter->pSkyline); + state->pRowIter->pSkyline = NULL; + + void *pe = NULL; + int32_t iter = 0; + while ((pe = tSimpleHashIterate(state->pr->pTableMap, pe, &iter)) != NULL) { + STableLoadInfo *pInfo = *(STableLoadInfo **)pe; + pInfo->pTombData = taosArrayDestroy(pInfo->pTombData); + } + } +} + static int32_t nextRowIterOpen(CacheNextRowIter *pIter, tb_uid_t uid, STsdb *pTsdb, STSchema *pTSchema, tb_uid_t suid, - SSttBlockLoadInfo *pLoadInfo, SLDataIter *pLDataIter, STsdbReadSnap *pReadSnap, - SDataFReader **pDataFReader, SDataFReader **pDataFReaderLast, int64_t lastTs) { + SArray *pLDataIterArray, STsdbReadSnap *pReadSnap, int64_t lastTs, + SCacheRowsReader *pr) { int code = 0; STbData *pMem = NULL; @@ -2397,71 +2396,26 @@ static int32_t nextRowIterOpen(CacheNextRowIter *pIter, tb_uid_t uid, STsdb *pTs pIter->pTsdb = pTsdb; - pIter->pSkyline = taosArrayInit(32, sizeof(TSDBKEY)); + pIter->pMemDelData = NULL; - SDelFile *pDelFile = pReadSnap->fs.pDelFile; - if (pDelFile) { - SDelFReader *pDelFReader; - - code = tsdbDelFReaderOpen(&pDelFReader, pDelFile, pTsdb); - if (code) goto _err; - - SArray *pDelIdxArray = taosArrayInit(32, sizeof(SDelIdx)); - - code = tsdbReadDelIdx(pDelFReader, pDelIdxArray); - if (code) { - taosArrayDestroy(pDelIdxArray); - tsdbDelFReaderClose(&pDelFReader); - goto _err; - } - - SDelIdx *delIdx = taosArraySearch(pDelIdxArray, &(SDelIdx){.suid = suid, .uid = uid}, tCmprDelIdx, TD_EQ); - - code = getTableDelSkyline(pMem, pIMem, pDelFReader, delIdx, pIter->pSkyline); - if (code) { - taosArrayDestroy(pDelIdxArray); - tsdbDelFReaderClose(&pDelFReader); - goto _err; - } - - taosArrayDestroy(pDelIdxArray); - tsdbDelFReaderClose(&pDelFReader); - } else { - code = getTableDelSkyline(pMem, pIMem, NULL, NULL, pIter->pSkyline); - if (code) goto _err; - } - - pIter->iSkyline = taosArrayGetSize(pIter->pSkyline) - 1; + loadMemTombData(&pIter->pMemDelData, pMem, pIMem, pr->info.verRange.maxVer); pIter->idx = (SBlockIdx){.suid = suid, .uid = uid}; - pIter->fsLastState.state = (SFSLASTNEXTROWSTATES)SFSNEXTROW_FS; - pIter->fsLastState.pTsdb = pTsdb; - pIter->fsLastState.aDFileSet = pReadSnap->fs.aDFileSet; - pIter->fsLastState.pTSchema = pTSchema; - pIter->fsLastState.suid = suid; - pIter->fsLastState.uid = uid; - pIter->fsLastState.pLoadInfo = pLoadInfo; - pIter->fsLastState.pDataFReader = pDataFReaderLast; - pIter->fsLastState.lastTs = lastTs; - pIter->fsLastState.pDataIter = pLDataIter; - + pIter->fsState.pRowIter = pIter; pIter->fsState.state = SFSNEXTROW_FS; pIter->fsState.pTsdb = pTsdb; - pIter->fsState.aDFileSet = pReadSnap->fs.aDFileSet; + pIter->fsState.aDFileSet = pReadSnap->pfSetArray; pIter->fsState.pBlockIdxExp = &pIter->idx; pIter->fsState.pTSchema = pTSchema; pIter->fsState.suid = suid; pIter->fsState.uid = uid; - pIter->fsState.pLoadInfo = pLoadInfo; - pIter->fsState.pDataFReader = pDataFReader; pIter->fsState.lastTs = lastTs; + pIter->fsState.pr = pr; pIter->input[0] = (TsdbNextRowState){&pIter->memRow, true, false, false, &pIter->memState, getNextRowFromMem, NULL}; pIter->input[1] = (TsdbNextRowState){&pIter->imemRow, true, false, false, &pIter->imemState, getNextRowFromMem, NULL}; - pIter->input[2] = (TsdbNextRowState){ - &pIter->fsLastRow, false, true, false, &pIter->fsLastState, getNextRowFromFSLast, clearNextRowFromFSLast}; - pIter->input[3] = + pIter->input[2] = (TsdbNextRowState){&pIter->fsRow, false, true, false, &pIter->fsState, getNextRowFromFS, clearNextRowFromFS}; if (pMem) { @@ -2480,7 +2434,7 @@ static int32_t nextRowIterOpen(CacheNextRowIter *pIter, tb_uid_t uid, STsdb *pTs pIter->input[1].next = true; } - return code; + pIter->pr = pr; _err: return code; } @@ -2488,7 +2442,7 @@ _err: static int32_t nextRowIterClose(CacheNextRowIter *pIter) { int code = 0; - for (int i = 0; i < 4; ++i) { + for (int i = 0; i < 3; ++i) { if (pIter->input[i].nextRowClearFn) { pIter->input[i].nextRowClearFn(pIter->input[i].iter); } @@ -2498,6 +2452,10 @@ static int32_t nextRowIterClose(CacheNextRowIter *pIter) { taosArrayDestroy(pIter->pSkyline); } + if (pIter->pMemDelData) { + taosArrayDestroy(pIter->pMemDelData); + } + _err: return code; } @@ -2507,7 +2465,7 @@ static int32_t nextRowIterGet(CacheNextRowIter *pIter, TSDBROW **ppRow, bool *pI int16_t *aCols, int nCols) { int code = 0; for (;;) { - for (int i = 0; i < 4; ++i) { + for (int i = 0; i < 3; ++i) { if (pIter->input[i].next && !pIter->input[i].stop) { code = pIter->input[i].nextRowFn(pIter->input[i].iter, &pIter->input[i].pRow, &pIter->input[i].ignoreEarlierTs, isLast, aCols, nCols); @@ -2520,10 +2478,10 @@ static int32_t nextRowIterGet(CacheNextRowIter *pIter, TSDBROW **ppRow, bool *pI } } - if (pIter->input[0].stop && pIter->input[1].stop && pIter->input[2].stop && pIter->input[3].stop) { + if (pIter->input[0].stop && pIter->input[1].stop && pIter->input[2].stop) { *ppRow = NULL; - *pIgnoreEarlierTs = (pIter->input[0].ignoreEarlierTs || pIter->input[1].ignoreEarlierTs || - pIter->input[2].ignoreEarlierTs || pIter->input[3].ignoreEarlierTs); + *pIgnoreEarlierTs = + (pIter->input[0].ignoreEarlierTs || pIter->input[1].ignoreEarlierTs || pIter->input[2].ignoreEarlierTs); return code; } @@ -2533,7 +2491,7 @@ static int32_t nextRowIterGet(CacheNextRowIter *pIter, TSDBROW **ppRow, bool *pI int nMax = 0; TSKEY maxKey = TSKEY_MIN; - for (int i = 0; i < 4; ++i) { + for (int i = 0; i < 3; ++i) { if (!pIter->input[i].stop && pIter->input[i].pRow != NULL) { TSDBKEY key = TSDBROW_KEY(pIter->input[i].pRow); @@ -2559,6 +2517,21 @@ static int32_t nextRowIterGet(CacheNextRowIter *pIter, TSDBROW **ppRow, bool *pI for (int i = 0; i < nMax; ++i) { TSDBKEY maxKey1 = TSDBROW_KEY(max[i]); + if (!pIter->pSkyline) { + pIter->pSkyline = taosArrayInit(32, sizeof(TSDBKEY)); + + uint64_t uid = pIter->idx.uid; + STableLoadInfo *pInfo = *(STableLoadInfo **)tSimpleHashGet(pIter->pr->pTableMap, &uid, sizeof(uid)); + SArray *pTombData = pInfo->pTombData; + if (pTombData) { + taosArrayAddAll(pTombData, pIter->pMemDelData); + + code = tsdbBuildDeleteSkyline(pTombData, 0, (int32_t)(TARRAY_SIZE(pTombData) - 1), pIter->pSkyline); + } + + pIter->iSkyline = taosArrayGetSize(pIter->pSkyline) - 1; + } + bool deleted = tsdbKeyDeleted(&maxKey1, pIter->pSkyline, &pIter->iSkyline); if (!deleted) { iMerge[nMerge] = iMax[i]; @@ -2629,322 +2602,7 @@ static int32_t updateTSchema(int32_t sversion, SCacheRowsReader *pReader, uint64 } taosMemoryFreeClear(pReader->pCurrSchema); - return metaGetTbTSchemaEx(pReader->pTsdb->pVnode->pMeta, pReader->suid, uid, sversion, &pReader->pCurrSchema); -} - -static int32_t mergeLastRow(tb_uid_t uid, STsdb *pTsdb, bool *dup, SArray **ppColArray, SCacheRowsReader *pr) { - STSchema *pTSchema = pr->pSchema; // metaGetTbTSchema(pTsdb->pVnode->pMeta, uid, -1, 1); - int16_t nLastCol = pTSchema->numOfCols; - int16_t iCol = 0; - int16_t noneCol = 0; - bool setNoneCol = false; - bool hasRow = false; - bool ignoreEarlierTs = false; - SArray *pColArray = NULL; - SColVal *pColVal = &(SColVal){0}; - - int32_t code = initLastColArray(pTSchema, &pColArray); - if (TSDB_CODE_SUCCESS != code) { - return code; - } - - TSKEY lastRowTs = TSKEY_MAX; - - CacheNextRowIter iter = {0}; - nextRowIterOpen(&iter, uid, pTsdb, pTSchema, pr->suid, pr->pLoadInfo, pr->pDataIter, pr->pReadSnap, &pr->pDataFReader, - &pr->pDataFReaderLast, pr->lastTs); - - do { - TSDBROW *pRow = NULL; - nextRowIterGet(&iter, &pRow, &ignoreEarlierTs, false, NULL, 0); - - if (!pRow) { - break; - } - - hasRow = true; - - int32_t sversion = TSDBROW_SVERSION(pRow); - if (sversion != -1) { - code = updateTSchema(sversion, pr, uid); - if (TSDB_CODE_SUCCESS != code) { - goto _err; - } - pTSchema = pr->pCurrSchema; - } - int16_t nCol = pTSchema->numOfCols; - - TSKEY rowTs = TSDBROW_TS(pRow); - - if (lastRowTs == TSKEY_MAX) { - lastRowTs = rowTs; - STColumn *pTColumn = &pTSchema->columns[0]; - - *pColVal = COL_VAL_VALUE(pTColumn->colId, pTColumn->type, (SValue){.val = lastRowTs}); - taosArraySet(pColArray, 0, &(SLastCol){.ts = lastRowTs, .colVal = *pColVal}); - - for (iCol = 1; iCol < nCol; ++iCol) { - if (iCol >= nLastCol) { - break; - } - SLastCol *pCol = taosArrayGet(pColArray, iCol); - if (pCol->colVal.cid != pTSchema->columns[iCol].colId) { - continue; - } - tsdbRowGetColVal(pRow, pTSchema, iCol, pColVal); - - *pCol = (SLastCol){.ts = lastRowTs, .colVal = *pColVal}; - if (IS_VAR_DATA_TYPE(pColVal->type) && pColVal->value.nData > 0) { - pCol->colVal.value.pData = taosMemoryMalloc(pCol->colVal.value.nData); - if (pCol->colVal.value.pData == NULL) { - terrno = TSDB_CODE_OUT_OF_MEMORY; - code = TSDB_CODE_OUT_OF_MEMORY; - goto _err; - } - memcpy(pCol->colVal.value.pData, pColVal->value.pData, pColVal->value.nData); - } - - if (COL_VAL_IS_NONE(pColVal) && !setNoneCol) { - noneCol = iCol; - setNoneCol = true; - } - } - if (!setNoneCol) { - // done, goto return pColArray - break; - } else { - continue; - } - } - - if ((rowTs < lastRowTs)) { - // done, goto return pColArray - break; - } - - // merge into pColArray - setNoneCol = false; - for (iCol = noneCol; iCol < nCol; ++iCol) { - // high version's column value - SColVal *tColVal = (SColVal *)taosArrayGet(pColArray, iCol); - - tsdbRowGetColVal(pRow, pTSchema, iCol, pColVal); - if (COL_VAL_IS_NONE(tColVal) && !COL_VAL_IS_NONE(pColVal)) { - SLastCol lastCol = {.ts = rowTs, .colVal = *pColVal}; - if (IS_VAR_DATA_TYPE(pColVal->type) && pColVal->value.nData > 0) { - SLastCol *pLastCol = (SLastCol *)taosArrayGet(pColArray, iCol); - taosMemoryFree(pLastCol->colVal.value.pData); - - lastCol.colVal.value.pData = taosMemoryMalloc(lastCol.colVal.value.nData); - if (lastCol.colVal.value.pData == NULL) { - terrno = TSDB_CODE_OUT_OF_MEMORY; - code = TSDB_CODE_OUT_OF_MEMORY; - goto _err; - } - memcpy(lastCol.colVal.value.pData, pColVal->value.pData, pColVal->value.nData); - } - - taosArraySet(pColArray, iCol, &lastCol); - } else if (COL_VAL_IS_NONE(tColVal) && COL_VAL_IS_NONE(pColVal) && !setNoneCol) { - noneCol = iCol; - setNoneCol = true; - } - } - } while (setNoneCol); - - // build the result ts row here - *dup = false; - // if (taosArrayGetSize(pColArray) != nCol) { - //*ppColArray = NULL; - // taosArrayDestroy(pColArray); - //} else { - if (!hasRow) { - if (ignoreEarlierTs) { - taosArrayDestroy(pColArray); - pColArray = NULL; - } else { - taosArrayClear(pColArray); - } - } - *ppColArray = pColArray; - //} - - nextRowIterClose(&iter); - // taosMemoryFreeClear(pTSchema); - return code; - -_err: - nextRowIterClose(&iter); - taosArrayDestroy(pColArray); - // taosMemoryFreeClear(pTSchema); - return code; -} - -static int32_t mergeLast(tb_uid_t uid, STsdb *pTsdb, SArray **ppLastArray, SCacheRowsReader *pr) { - STSchema *pTSchema = pr->pSchema; // metaGetTbTSchema(pTsdb->pVnode->pMeta, uid, -1, 1); - int16_t nLastCol = pTSchema->numOfCols; - int16_t noneCol = 0; - bool setNoneCol = false; - bool hasRow = false; - bool ignoreEarlierTs = false; - SArray *pColArray = NULL; - SColVal *pColVal = &(SColVal){0}; - int16_t nCols = nLastCol; - - int32_t code = initLastColArray(pTSchema, &pColArray); - if (TSDB_CODE_SUCCESS != code) { - return code; - } - SArray *aColArray = taosArrayInit(nCols, sizeof(int16_t)); - if (NULL == aColArray) { - taosArrayDestroy(pColArray); - - return TSDB_CODE_OUT_OF_MEMORY; - } - for (int i = 1; i < pTSchema->numOfCols; ++i) { - taosArrayPush(aColArray, &pTSchema->columns[i].colId); - } - - TSKEY lastRowTs = TSKEY_MAX; - - CacheNextRowIter iter = {0}; - nextRowIterOpen(&iter, uid, pTsdb, pTSchema, pr->suid, pr->pLoadInfo, pr->pDataIter, pr->pReadSnap, &pr->pDataFReader, - &pr->pDataFReaderLast, pr->lastTs); - - do { - TSDBROW *pRow = NULL; - nextRowIterGet(&iter, &pRow, &ignoreEarlierTs, true, TARRAY_DATA(aColArray), TARRAY_SIZE(aColArray)); - - if (!pRow) { - break; - } - - hasRow = true; - - int32_t sversion = TSDBROW_SVERSION(pRow); - if (sversion != -1) { - code = updateTSchema(sversion, pr, uid); - if (TSDB_CODE_SUCCESS != code) { - goto _err; - } - pTSchema = pr->pCurrSchema; - } - int16_t nCol = pTSchema->numOfCols; - - TSKEY rowTs = TSDBROW_TS(pRow); - - if (lastRowTs == TSKEY_MAX) { - lastRowTs = rowTs; - STColumn *pTColumn = &pTSchema->columns[0]; - - *pColVal = COL_VAL_VALUE(pTColumn->colId, pTColumn->type, (SValue){.val = lastRowTs}); - taosArraySet(pColArray, 0, &(SLastCol){.ts = lastRowTs, .colVal = *pColVal}); - - for (int16_t iCol = 1; iCol < nCol; ++iCol) { - if (iCol >= nLastCol) { - break; - } - SLastCol *pCol = taosArrayGet(pColArray, iCol); - if (pCol->colVal.cid != pTSchema->columns[iCol].colId) { - continue; - } - tsdbRowGetColVal(pRow, pTSchema, iCol, pColVal); - - *pCol = (SLastCol){.ts = lastRowTs, .colVal = *pColVal}; - if (IS_VAR_DATA_TYPE(pColVal->type) && pColVal->value.nData > 0) { - pCol->colVal.value.pData = taosMemoryMalloc(pCol->colVal.value.nData); - if (pCol->colVal.value.pData == NULL) { - terrno = TSDB_CODE_OUT_OF_MEMORY; - code = TSDB_CODE_OUT_OF_MEMORY; - goto _err; - } - memcpy(pCol->colVal.value.pData, pColVal->value.pData, pColVal->value.nData); - } - - if (!COL_VAL_IS_VALUE(pColVal)) { - if (!setNoneCol) { - noneCol = iCol; - setNoneCol = true; - } - } else { - int32_t aColIndex = taosArraySearchIdx(aColArray, &pColVal->cid, compareInt16Val, TD_EQ); - taosArrayRemove(aColArray, aColIndex); - } - } - if (!setNoneCol) { - // done, goto return pColArray - break; - } else { - continue; - } - } - - // merge into pColArray - setNoneCol = false; - for (int16_t iCol = noneCol; iCol < nCol; ++iCol) { - if (iCol >= nLastCol) { - break; - } - // high version's column value - SLastCol *lastColVal = (SLastCol *)taosArrayGet(pColArray, iCol); - if (lastColVal->colVal.cid != pTSchema->columns[iCol].colId) { - continue; - } - SColVal *tColVal = &lastColVal->colVal; - - tsdbRowGetColVal(pRow, pTSchema, iCol, pColVal); - if (!COL_VAL_IS_VALUE(tColVal) && COL_VAL_IS_VALUE(pColVal)) { - SLastCol lastCol = {.ts = rowTs, .colVal = *pColVal}; - if (IS_VAR_DATA_TYPE(pColVal->type) && pColVal->value.nData > 0) { - SLastCol *pLastCol = (SLastCol *)taosArrayGet(pColArray, iCol); - taosMemoryFree(pLastCol->colVal.value.pData); - - lastCol.colVal.value.pData = taosMemoryMalloc(lastCol.colVal.value.nData); - if (lastCol.colVal.value.pData == NULL) { - terrno = TSDB_CODE_OUT_OF_MEMORY; - code = TSDB_CODE_OUT_OF_MEMORY; - goto _err; - } - memcpy(lastCol.colVal.value.pData, pColVal->value.pData, pColVal->value.nData); - } - - taosArraySet(pColArray, iCol, &lastCol); - int32_t aColIndex = taosArraySearchIdx(aColArray, &lastCol.colVal.cid, compareInt16Val, TD_EQ); - taosArrayRemove(aColArray, aColIndex); - } else if (!COL_VAL_IS_VALUE(tColVal) && !COL_VAL_IS_VALUE(pColVal) && !setNoneCol) { - noneCol = iCol; - setNoneCol = true; - } - } - } while (setNoneCol); - - // if (taosArrayGetSize(pColArray) <= 0) { - //*ppLastArray = NULL; - // taosArrayDestroy(pColArray); - //} else { - if (!hasRow) { - if (ignoreEarlierTs) { - taosArrayDestroy(pColArray); - pColArray = NULL; - } else { - taosArrayClear(pColArray); - } - } - *ppLastArray = pColArray; - //} - - nextRowIterClose(&iter); - taosArrayDestroy(aColArray); - // taosMemoryFreeClear(pTSchema); - return code; - -_err: - nextRowIterClose(&iter); - // taosMemoryFreeClear(pTSchema); - *ppLastArray = NULL; - taosArrayDestroy(pColArray); - taosArrayDestroy(aColArray); - return code; + return metaGetTbTSchemaEx(pReader->pTsdb->pVnode->pMeta, pReader->info.suid, uid, sversion, &pReader->pCurrSchema); } static int32_t mergeLastCid(tb_uid_t uid, STsdb *pTsdb, SArray **ppLastArray, SCacheRowsReader *pr, int16_t *aCols, @@ -2976,8 +2634,7 @@ static int32_t mergeLastCid(tb_uid_t uid, STsdb *pTsdb, SArray **ppLastArray, SC TSKEY lastRowTs = TSKEY_MAX; CacheNextRowIter iter = {0}; - nextRowIterOpen(&iter, uid, pTsdb, pTSchema, pr->suid, pr->pLoadInfo, pr->pDataIter, pr->pReadSnap, &pr->pDataFReader, - &pr->pDataFReaderLast, pr->lastTs); + nextRowIterOpen(&iter, uid, pTsdb, pTSchema, pr->info.suid, pr->pLDataIterArray, pr->pReadSnap, pr->lastTs, pr); do { TSDBROW *pRow = NULL; @@ -3146,8 +2803,7 @@ static int32_t mergeLastRowCid(tb_uid_t uid, STsdb *pTsdb, SArray **ppLastArray, TSKEY lastRowTs = TSKEY_MAX; CacheNextRowIter iter = {0}; - nextRowIterOpen(&iter, uid, pTsdb, pTSchema, pr->suid, pr->pLoadInfo, pr->pDataIter, pr->pReadSnap, &pr->pDataFReader, - &pr->pDataFReaderLast, pr->lastTs); + nextRowIterOpen(&iter, uid, pTsdb, pTSchema, pr->info.suid, pr->pLDataIterArray, pr->pReadSnap, pr->lastTs, pr); do { TSDBROW *pRow = NULL; @@ -3236,92 +2892,6 @@ _err: return code; } -int32_t tsdbCacheGetLastrowH(SLRUCache *pCache, tb_uid_t uid, SCacheRowsReader *pr, LRUHandle **handle) { - int32_t code = 0; - char key[32] = {0}; - int keyLen = 0; - - // getTableCacheKeyS(uid, "lr", key, &keyLen); - getTableCacheKey(uid, 0, key, &keyLen); - LRUHandle *h = taosLRUCacheLookup(pCache, key, keyLen); - if (!h) { - STsdb *pTsdb = pr->pVnode->pTsdb; - taosThreadMutexLock(&pTsdb->lruMutex); - - h = taosLRUCacheLookup(pCache, key, keyLen); - if (!h) { - SArray *pArray = NULL; - bool dup = false; // which is always false for now - code = mergeLastRow(uid, pTsdb, &dup, &pArray, pr); - // if table's empty or error or ignore ignore earlier ts, set handle NULL and return - if (code < 0 || pArray == NULL) { - if (!dup && pArray) { - taosArrayDestroy(pArray); - } - - taosThreadMutexUnlock(&pTsdb->lruMutex); - - *handle = NULL; - - return 0; - } - - size_t charge = pArray->capacity * pArray->elemSize + sizeof(*pArray); - _taos_lru_deleter_t deleter = deleteTableCacheLast; - LRUStatus status = - taosLRUCacheInsert(pCache, key, keyLen, pArray, charge, deleter, &h, TAOS_LRU_PRIORITY_LOW, NULL); - if (status != TAOS_LRU_STATUS_OK) { - code = -1; - } - } - taosThreadMutexUnlock(&pTsdb->lruMutex); - } - - *handle = h; - - return code; -} - -int32_t tsdbCacheGetLastH(SLRUCache *pCache, tb_uid_t uid, SCacheRowsReader *pr, LRUHandle **handle) { - int32_t code = 0; - char key[32] = {0}; - int keyLen = 0; - - // getTableCacheKeyS(uid, "l", key, &keyLen); - getTableCacheKey(uid, 1, key, &keyLen); - LRUHandle *h = taosLRUCacheLookup(pCache, key, keyLen); - if (!h) { - STsdb *pTsdb = pr->pVnode->pTsdb; - taosThreadMutexLock(&pTsdb->lruMutex); - - h = taosLRUCacheLookup(pCache, key, keyLen); - if (!h) { - SArray *pLastArray = NULL; - code = mergeLast(uid, pTsdb, &pLastArray, pr); - // if table's empty or error or ignore ignore earlier ts, set handle NULL and return - if (code < 0 || pLastArray == NULL) { - taosThreadMutexUnlock(&pTsdb->lruMutex); - - *handle = NULL; - return 0; - } - - size_t charge = pLastArray->capacity * pLastArray->elemSize + sizeof(*pLastArray); - _taos_lru_deleter_t deleter = deleteTableCacheLast; - LRUStatus status = - taosLRUCacheInsert(pCache, key, keyLen, pLastArray, charge, deleter, &h, TAOS_LRU_PRIORITY_LOW, NULL); - if (status != TAOS_LRU_STATUS_OK) { - code = -1; - } - } - taosThreadMutexUnlock(&pTsdb->lruMutex); - } - - *handle = h; - - return code; -} - int32_t tsdbCacheRelease(SLRUCache *pCache, LRUHandle *h) { int32_t code = 0; diff --git a/source/dnode/vnode/src/tsdb/tsdbCacheRead.c b/source/dnode/vnode/src/tsdb/tsdbCacheRead.c index 6138b1f7b4..f17041e98b 100644 --- a/source/dnode/vnode/src/tsdb/tsdbCacheRead.c +++ b/source/dnode/vnode/src/tsdb/tsdbCacheRead.c @@ -17,6 +17,7 @@ #include "tarray.h" #include "tcommon.h" #include "tsdb.h" +#include "tsdbDataFileRW.h" #define HASTYPE(_type, _t) (((_type) & (_t)) == (_t)) @@ -124,11 +125,29 @@ int32_t tsdbReuseCacherowsReader(void* reader, void* pTableIdList, int32_t numOf pReader->numOfTables = numOfTables; pReader->lastTs = INT64_MIN; - resetLastBlockLoadInfo(pReader->pLoadInfo); + int64_t blocks; + double elapse; + pReader->pLDataIterArray = destroySttBlockReader(pReader->pLDataIterArray, &blocks, &elapse); + pReader->pLDataIterArray = taosArrayInit(4, POINTER_BYTES); return TSDB_CODE_SUCCESS; } +static int32_t uidComparFunc(const void* p1, const void* p2) { + uint64_t pu1 = *(uint64_t*)p1; + uint64_t pu2 = *(uint64_t*)p2; + if (pu1 == pu2) { + return 0; + } else { + return (pu1 < pu2) ? -1 : 1; + } +} + +static void freeTableInfoFunc(void* param) { + void** p = (void**)param; + taosMemoryFreeClear(*p); +} + int32_t tsdbCacherowsReaderOpen(void* pVnode, int32_t type, void* pTableIdList, int32_t numOfTables, int32_t numOfCols, SArray* pCidList, int32_t* pSlotIds, uint64_t suid, void** pReader, const char* idstr) { *pReader = NULL; @@ -140,11 +159,11 @@ int32_t tsdbCacherowsReaderOpen(void* pVnode, int32_t type, void* pTableIdList, p->type = type; p->pVnode = pVnode; p->pTsdb = p->pVnode->pTsdb; - p->verRange = (SVersionRange){.minVer = 0, .maxVer = UINT64_MAX}; + p->info.verRange = (SVersionRange){.minVer = 0, .maxVer = UINT64_MAX}; + p->info.suid = suid; p->numOfCols = numOfCols; p->pCidList = pCidList; p->pSlotIds = pSlotIds; - p->suid = suid; if (numOfTables == 0) { *pReader = p; @@ -154,6 +173,27 @@ int32_t tsdbCacherowsReaderOpen(void* pVnode, int32_t type, void* pTableIdList, p->pTableList = pTableIdList; p->numOfTables = numOfTables; + p->pTableMap = tSimpleHashInit(numOfTables, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT)); + if (p->pTableMap == NULL) { + tsdbCacherowsReaderClose(p); + return TSDB_CODE_OUT_OF_MEMORY; + } + p->uidList = taosMemoryMalloc(numOfTables * sizeof(uint64_t)); + if (p->uidList == NULL) { + tsdbCacherowsReaderClose(p); + return TSDB_CODE_OUT_OF_MEMORY; + } + for (int32_t i = 0; i < numOfTables; ++i) { + uint64_t uid = p->pTableList[i].uid; + p->uidList[i] = uid; + STableLoadInfo* pInfo = taosMemoryCalloc(1, sizeof(STableLoadInfo)); + tSimpleHashPut(p->pTableMap, &uid, sizeof(uint64_t), &pInfo, POINTER_BYTES); + } + + tSimpleHashSetFreeFp(p->pTableMap, freeTableInfoFunc); + + taosSort(p->uidList, numOfTables, sizeof(uint64_t), uidComparFunc); + int32_t code = setTableSchema(p, suid, idstr); if (code != TSDB_CODE_SUCCESS) { tsdbCacherowsReaderClose(p); @@ -178,14 +218,8 @@ int32_t tsdbCacherowsReaderOpen(void* pVnode, int32_t type, void* pTableIdList, SVnodeCfg* pCfg = &((SVnode*)pVnode)->config; int32_t numOfStt = pCfg->sttTrigger; - p->pLoadInfo = tCreateLastBlockLoadInfo(p->pSchema, NULL, 0, numOfStt); - if (p->pLoadInfo == NULL) { - tsdbCacherowsReaderClose(p); - return TSDB_CODE_OUT_OF_MEMORY; - } - - p->pDataIter = taosMemoryCalloc(pCfg->sttTrigger, sizeof(SLDataIter)); - if (p->pDataIter == NULL) { + p->pLDataIterArray = taosArrayInit(4, POINTER_BYTES); + if (p->pLDataIterArray == NULL) { tsdbCacherowsReaderClose(p); return TSDB_CODE_OUT_OF_MEMORY; } @@ -214,14 +248,34 @@ void* tsdbCacherowsReaderClose(void* pReader) { taosMemoryFree(p->pSchema); } - taosMemoryFree(p->pDataIter); taosMemoryFree(p->pCurrSchema); - destroyLastBlockLoadInfo(p->pLoadInfo); + int64_t loadBlocks = 0; + double elapse = 0; + destroySttBlockReader(p->pLDataIterArray, &loadBlocks, &elapse); + + if (p->pFileReader) { + tsdbDataFileReaderClose(&p->pFileReader); + p->pFileReader = NULL; + } taosMemoryFree((void*)p->idstr); taosThreadMutexDestroy(&p->readerMutex); + if (p->pTableMap) { + void* pe = NULL; + int32_t iter = 0; + while ((pe = tSimpleHashIterate(p->pTableMap, pe, &iter)) != NULL) { + STableLoadInfo* pInfo = *(STableLoadInfo**)pe; + pInfo->pTombData = taosArrayDestroy(pInfo->pTombData); + } + + tSimpleHashCleanup(p->pTableMap); + } + if (p->uidList) { + taosMemoryFree(p->uidList); + } + taosMemoryFree(pReader); return NULL; } @@ -298,12 +352,10 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32 } taosThreadMutexLock(&pr->readerMutex); - code = tsdbTakeReadSnap((STsdbReader*)pr, tsdbCacheQueryReseek, &pr->pReadSnap); + code = tsdbTakeReadSnap2((STsdbReader*)pr, tsdbCacheQueryReseek, &pr->pReadSnap); if (code != TSDB_CODE_SUCCESS) { goto _end; } - pr->pDataFReader = NULL; - pr->pDataFReaderLast = NULL; int8_t ltype = (pr->type & CACHESCAN_RETRIEVE_LAST) >> 3; @@ -424,11 +476,13 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32 } _end: - tsdbDataFReaderClose(&pr->pDataFReaderLast); - tsdbDataFReaderClose(&pr->pDataFReader); + tsdbUntakeReadSnap2((STsdbReader*)pr, pr->pReadSnap, true); + + int64_t loadBlocks = 0; + double elapse = 0; + pr->pLDataIterArray = destroySttBlockReader(pr->pLDataIterArray, &loadBlocks, &elapse); + pr->pLDataIterArray = taosArrayInit(4, POINTER_BYTES); - resetLastBlockLoadInfo(pr->pLoadInfo); - tsdbUntakeReadSnap((STsdbReader*)pr, pr->pReadSnap, true); taosThreadMutexUnlock(&pr->readerMutex); if (pRes != NULL) { diff --git a/source/dnode/vnode/src/tsdb/tsdbCommit2.c b/source/dnode/vnode/src/tsdb/tsdbCommit2.c new file mode 100644 index 0000000000..8a9dbf37b9 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbCommit2.c @@ -0,0 +1,605 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbCommit2.h" + +// extern dependencies +typedef struct { + STsdb *tsdb; + TFileSetArray *fsetArr; + TFileOpArray fopArray[1]; + + // SSkmInfo skmTb[1]; + // SSkmInfo skmRow[1]; + + int32_t minutes; + int8_t precision; + int32_t minRow; + int32_t maxRow; + int8_t cmprAlg; + int32_t sttTrigger; + int32_t szPage; + int64_t compactVersion; + + struct { + int64_t cid; + int64_t now; + TSKEY nextKey; + int32_t fid; + int32_t expLevel; + SDiskID did; + TSKEY minKey; + TSKEY maxKey; + STFileSet *fset; + TABLEID tbid[1]; + bool hasTSData; + } ctx[1]; + + // reader + SSttFileReader *sttReader; + + // iter + TTsdbIterArray dataIterArray[1]; + SIterMerger *dataIterMerger; + TTsdbIterArray tombIterArray[1]; + SIterMerger *tombIterMerger; + + // writer + SFSetWriter *writer; +} SCommitter2; + +static int32_t tsdbCommitOpenWriter(SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + + SFSetWriterConfig config = { + .tsdb = committer->tsdb, + .toSttOnly = true, + .compactVersion = committer->compactVersion, + .minRow = committer->minRow, + .maxRow = committer->maxRow, + .szPage = committer->szPage, + .cmprAlg = committer->cmprAlg, + .fid = committer->ctx->fid, + .cid = committer->ctx->cid, + .did = committer->ctx->did, + .level = 0, + }; + + if (committer->sttTrigger == 1) { + config.toSttOnly = false; + + if (committer->ctx->fset) { + for (int32_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ftype++) { + if (committer->ctx->fset->farr[ftype] != NULL) { + config.files[ftype].exist = true; + config.files[ftype].file = committer->ctx->fset->farr[ftype]->f[0]; + } + } + } + } + + code = tsdbFSetWriterOpen(&config, &committer->writer); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(committer->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbCommitCloseWriter(SCommitter2 *committer) { + return tsdbFSetWriterClose(&committer->writer, 0, committer->fopArray); +} + +static int32_t tsdbCommitTSData(SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + int64_t numOfRow = 0; + SMetaInfo info; + + committer->ctx->hasTSData = false; + + committer->ctx->tbid->suid = 0; + committer->ctx->tbid->uid = 0; + for (SRowInfo *row; (row = tsdbIterMergerGetData(committer->dataIterMerger)) != NULL;) { + if (row->uid != committer->ctx->tbid->uid) { + committer->ctx->tbid->suid = row->suid; + committer->ctx->tbid->uid = row->uid; + + if (metaGetInfo(committer->tsdb->pVnode->pMeta, row->uid, &info, NULL) != 0) { + code = tsdbIterMergerSkipTableData(committer->dataIterMerger, committer->ctx->tbid); + TSDB_CHECK_CODE(code, lino, _exit); + continue; + } + } + + int64_t ts = TSDBROW_TS(&row->row); + if (ts > committer->ctx->maxKey) { + committer->ctx->nextKey = TMIN(committer->ctx->nextKey, ts); + code = tsdbIterMergerSkipTableData(committer->dataIterMerger, committer->ctx->tbid); + TSDB_CHECK_CODE(code, lino, _exit); + continue; + } + + committer->ctx->hasTSData = true; + numOfRow++; + + code = tsdbFSetWriteRow(committer->writer, row); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbIterMergerNext(committer->dataIterMerger); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(committer->tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d fid:%d commit %" PRId64 " rows", TD_VID(committer->tsdb->pVnode), committer->ctx->fid, numOfRow); + } + return code; +} + +static int32_t tsdbCommitTombData(SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + int64_t numRecord = 0; + SMetaInfo info; + + if (committer->ctx->fset == NULL && !committer->ctx->hasTSData) { + return 0; + } + + committer->ctx->tbid->suid = 0; + committer->ctx->tbid->uid = 0; + for (STombRecord *record; (record = tsdbIterMergerGetTombRecord(committer->tombIterMerger));) { + if (record->uid != committer->ctx->tbid->uid) { + committer->ctx->tbid->suid = record->suid; + committer->ctx->tbid->uid = record->uid; + + if (metaGetInfo(committer->tsdb->pVnode->pMeta, record->uid, &info, NULL) != 0) { + code = tsdbIterMergerSkipTableData(committer->dataIterMerger, committer->ctx->tbid); + TSDB_CHECK_CODE(code, lino, _exit); + continue; + } + } + + if (record->ekey < committer->ctx->minKey) { + goto _next; + } else if (record->skey > committer->ctx->maxKey) { + committer->ctx->maxKey = TMIN(record->skey, committer->ctx->maxKey); + goto _next; + } + + if (record->ekey > committer->ctx->maxKey) { + committer->ctx->maxKey = committer->ctx->maxKey + 1; + } + + record->skey = TMAX(record->skey, committer->ctx->minKey); + record->ekey = TMIN(record->ekey, committer->ctx->maxKey); + + numRecord++; + code = tsdbFSetWriteTombRecord(committer->writer, record); + TSDB_CHECK_CODE(code, lino, _exit); + + _next: + code = tsdbIterMergerNext(committer->tombIterMerger); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(committer->tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d fid:%d commit %" PRId64 " tomb records", TD_VID(committer->tsdb->pVnode), committer->ctx->fid, + numRecord); + } + return code; +} + +static int32_t tsdbCommitOpenReader(SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + + ASSERT(committer->sttReader == NULL); + + if (committer->ctx->fset == NULL // + || committer->sttTrigger > 1 // + || TARRAY2_SIZE(committer->ctx->fset->lvlArr) == 0 // + ) { + return 0; + } + + ASSERT(TARRAY2_SIZE(committer->ctx->fset->lvlArr) == 1); + + SSttLvl *lvl = TARRAY2_FIRST(committer->ctx->fset->lvlArr); + + ASSERT(lvl->level == 0); + + if (TARRAY2_SIZE(lvl->fobjArr) == 0) { + return 0; + } + + ASSERT(TARRAY2_SIZE(lvl->fobjArr) == 1); + + STFileObj *fobj = TARRAY2_FIRST(lvl->fobjArr); + + SSttFileReaderConfig config = { + .tsdb = committer->tsdb, + .szPage = committer->szPage, + .file = fobj->f[0], + }; + code = tsdbSttFileReaderOpen(fobj->fname, &config, &committer->sttReader); + TSDB_CHECK_CODE(code, lino, _exit); + + STFileOp op = { + .optype = TSDB_FOP_REMOVE, + .fid = fobj->f->fid, + .of = fobj->f[0], + }; + + code = TARRAY2_APPEND(committer->fopArray, op); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(committer->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbCommitCloseReader(SCommitter2 *committer) { return tsdbSttFileReaderClose(&committer->sttReader); } + +static int32_t tsdbCommitOpenIter(SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + + ASSERT(TARRAY2_SIZE(committer->dataIterArray) == 0); + ASSERT(committer->dataIterMerger == NULL); + ASSERT(TARRAY2_SIZE(committer->tombIterArray) == 0); + ASSERT(committer->tombIterMerger == NULL); + + STsdbIter *iter; + STsdbIterConfig config = {0}; + + // mem data iter + config.type = TSDB_ITER_TYPE_MEMT; + config.memt = committer->tsdb->imem; + config.from->ts = committer->ctx->minKey; + config.from->version = VERSION_MIN; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(committer->dataIterArray, iter); + TSDB_CHECK_CODE(code, lino, _exit); + + // mem tomb iter + config.type = TSDB_ITER_TYPE_MEMT_TOMB; + config.memt = committer->tsdb->imem; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(committer->tombIterArray, iter); + TSDB_CHECK_CODE(code, lino, _exit); + + // STT + if (committer->sttReader) { + // data iter + config.type = TSDB_ITER_TYPE_STT; + config.sttReader = committer->sttReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(committer->dataIterArray, iter); + TSDB_CHECK_CODE(code, lino, _exit); + + // tomb iter + config.type = TSDB_ITER_TYPE_STT_TOMB; + config.sttReader = committer->sttReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(committer->tombIterArray, iter); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // open merger + code = tsdbIterMergerOpen(committer->dataIterArray, &committer->dataIterMerger, false); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbIterMergerOpen(committer->tombIterArray, &committer->tombIterMerger, true); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(committer->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbCommitCloseIter(SCommitter2 *committer) { + tsdbIterMergerClose(&committer->tombIterMerger); + tsdbIterMergerClose(&committer->dataIterMerger); + TARRAY2_CLEAR(committer->tombIterArray, tsdbIterClose); + TARRAY2_CLEAR(committer->dataIterArray, tsdbIterClose); + return 0; +} + +static int32_t tsdbCommitFileSetBegin(SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + STsdb *tsdb = committer->tsdb; + + committer->ctx->fid = tsdbKeyFid(committer->ctx->nextKey, committer->minutes, committer->precision); + committer->ctx->expLevel = tsdbFidLevel(committer->ctx->fid, &tsdb->keepCfg, committer->ctx->now); + tsdbFidKeyRange(committer->ctx->fid, committer->minutes, committer->precision, &committer->ctx->minKey, + &committer->ctx->maxKey); + code = tfsAllocDisk(committer->tsdb->pVnode->pTfs, committer->ctx->expLevel, &committer->ctx->did); + TSDB_CHECK_CODE(code, lino, _exit); + tfsMkdirRecurAt(committer->tsdb->pVnode->pTfs, committer->tsdb->path, committer->ctx->did); + STFileSet fset = {.fid = committer->ctx->fid}; + committer->ctx->fset = &fset; + STFileSet **fsetPtr = TARRAY2_SEARCH(committer->fsetArr, &committer->ctx->fset, tsdbTFileSetCmprFn, TD_EQ); + committer->ctx->fset = (fsetPtr == NULL) ? NULL : *fsetPtr; + committer->ctx->tbid->suid = 0; + committer->ctx->tbid->uid = 0; + + ASSERT(TARRAY2_SIZE(committer->dataIterArray) == 0); + ASSERT(committer->dataIterMerger == NULL); + ASSERT(committer->writer == NULL); + + code = tsdbCommitOpenReader(committer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbCommitOpenIter(committer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbCommitOpenWriter(committer); + TSDB_CHECK_CODE(code, lino, _exit); + + // reset nextKey + committer->ctx->nextKey = TSKEY_MAX; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d %s done, fid:%d minKey:%" PRId64 " maxKey:%" PRId64 " expLevel:%d", TD_VID(tsdb->pVnode), + __func__, committer->ctx->fid, committer->ctx->minKey, committer->ctx->maxKey, committer->ctx->expLevel); + } + return 0; +} + +static int32_t tsdbCommitFileSetEnd(SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + + code = tsdbCommitCloseWriter(committer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbCommitCloseIter(committer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbCommitCloseReader(committer); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(committer->tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d %s done, fid:%d", TD_VID(committer->tsdb->pVnode), __func__, committer->ctx->fid); + } + return code; +} + +static int32_t tsdbCommitFileSet(SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + + // fset commit start + code = tsdbCommitFileSetBegin(committer); + TSDB_CHECK_CODE(code, lino, _exit); + + // commit fset + code = tsdbCommitTSData(committer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbCommitTombData(committer); + TSDB_CHECK_CODE(code, lino, _exit); + + // fset commit end + code = tsdbCommitFileSetEnd(committer); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(committer->tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d %s done, fid:%d", TD_VID(committer->tsdb->pVnode), __func__, committer->ctx->fid); + } + return code; +} + +static int32_t tsdbOpenCommitter(STsdb *tsdb, SCommitInfo *info, SCommitter2 *committer) { + int32_t code = 0; + int32_t lino = 0; + + memset(committer, 0, sizeof(committer[0])); + + committer->tsdb = tsdb; + code = tsdbFSCreateCopySnapshot(tsdb->pFS, &committer->fsetArr); + TSDB_CHECK_CODE(code, lino, _exit); + committer->minutes = tsdb->keepCfg.days; + committer->precision = tsdb->keepCfg.precision; + committer->minRow = info->info.config.tsdbCfg.minRows; + committer->maxRow = info->info.config.tsdbCfg.maxRows; + committer->cmprAlg = info->info.config.tsdbCfg.compression; + committer->sttTrigger = info->info.config.sttTrigger; + committer->szPage = info->info.config.tsdbPageSize; + committer->compactVersion = INT64_MAX; + committer->ctx->cid = tsdbFSAllocEid(tsdb->pFS); + committer->ctx->now = taosGetTimestampSec(); + + committer->ctx->nextKey = tsdb->imem->minKey; + if (tsdb->imem->nDel > 0) { + SRBTreeIter iter[1] = {tRBTreeIterCreate(tsdb->imem->tbDataTree, 1)}; + + for (SRBTreeNode *node = tRBTreeIterNext(iter); node; node = tRBTreeIterNext(iter)) { + STbData *tbData = TCONTAINER_OF(node, STbData, rbtn); + + for (SDelData *delData = tbData->pHead; delData; delData = delData->pNext) { + if (delData->sKey < committer->ctx->nextKey) { + committer->ctx->nextKey = delData->sKey; + } + } + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d %s done", TD_VID(tsdb->pVnode), __func__); + } + return code; +} + +static int32_t tsdbCloseCommitter(SCommitter2 *committer, int32_t eno) { + int32_t code = 0; + int32_t lino = 0; + + if (eno == 0) { + code = tsdbFSEditBegin(committer->tsdb->pFS, committer->fopArray, TSDB_FEDIT_COMMIT); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + // TODO + ASSERT(0); + } + + ASSERT(committer->writer == NULL); + ASSERT(committer->dataIterMerger == NULL); + ASSERT(committer->tombIterMerger == NULL); + TARRAY2_DESTROY(committer->dataIterArray, NULL); + TARRAY2_DESTROY(committer->tombIterArray, NULL); + TARRAY2_DESTROY(committer->fopArray, NULL); + tsdbFSDestroyCopySnapshot(&committer->fsetArr); + +_exit: + if (code) { + tsdbError("vgId:%d %s failed at line %d since %s, eid:%" PRId64, TD_VID(committer->tsdb->pVnode), __func__, lino, + tstrerror(code), committer->ctx->cid); + } else { + tsdbDebug("vgId:%d %s done, eid:%" PRId64, TD_VID(committer->tsdb->pVnode), __func__, committer->ctx->cid); + } + return code; +} + +int32_t tsdbPreCommit(STsdb *tsdb) { + taosThreadRwlockWrlock(&tsdb->rwLock); + ASSERT(tsdb->imem == NULL); + tsdb->imem = tsdb->mem; + tsdb->mem = NULL; + taosThreadRwlockUnlock(&tsdb->rwLock); + return 0; +} + +int32_t tsdbCommitBegin(STsdb *tsdb, SCommitInfo *info) { + if (!tsdb) return 0; + + int32_t code = 0; + int32_t lino = 0; + + SMemTable *imem = tsdb->imem; + int64_t nRow = imem->nRow; + int64_t nDel = imem->nDel; + + if (nRow == 0 && nDel == 0) { + taosThreadRwlockWrlock(&tsdb->rwLock); + tsdb->imem = NULL; + taosThreadRwlockUnlock(&tsdb->rwLock); + tsdbUnrefMemTable(imem, NULL, true); + } else { + SCommitter2 committer[1]; + + code = tsdbOpenCommitter(tsdb, info, committer); + TSDB_CHECK_CODE(code, lino, _exit); + + while (committer->ctx->nextKey != TSKEY_MAX) { + code = tsdbCommitFileSet(committer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbCloseCommitter(committer, code); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } else { + tsdbInfo("vgId:%d %s done, nRow:%" PRId64 " nDel:%" PRId64, TD_VID(tsdb->pVnode), __func__, nRow, nDel); + } + return code; +} + +int32_t tsdbCommitCommit(STsdb *tsdb) { + int32_t code = 0; + int32_t lino = 0; + + if (tsdb->imem == NULL) goto _exit; + + SMemTable *pMemTable = tsdb->imem; + taosThreadRwlockWrlock(&tsdb->rwLock); + code = tsdbFSEditCommit(tsdb->pFS); + if (code) { + taosThreadRwlockUnlock(&tsdb->rwLock); + TSDB_CHECK_CODE(code, lino, _exit); + } + tsdb->imem = NULL; + taosThreadRwlockUnlock(&tsdb->rwLock); + tsdbUnrefMemTable(pMemTable, NULL, true); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } else { + tsdbInfo("vgId:%d %s done", TD_VID(tsdb->pVnode), __func__); + } + return code; +} + +int32_t tsdbCommitAbort(STsdb *pTsdb) { + int32_t code = 0; + int32_t lino = 0; + + if (pTsdb->imem == NULL) goto _exit; + + code = tsdbFSEditAbort(pTsdb->pFS); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + tsdbError("vgId:%d, %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); + } else { + tsdbInfo("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); + } + return code; +} \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbCommit2.h b/source/dnode/vnode/src/tsdb/tsdbCommit2.h new file mode 100644 index 0000000000..41f72f345b --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbCommit2.h @@ -0,0 +1,33 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbDataFileRW.h" +#include "tsdbFS2.h" +#include "tsdbFSetRW.h" +#include "tsdbIter.h" +#include "tsdbSttFileRW.h" + +#ifndef _TSDB_COMMIT_H_ +#define _TSDB_COMMIT_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_COMMIT_H_*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbDataFileRW.c b/source/dnode/vnode/src/tsdb/tsdbDataFileRW.c new file mode 100644 index 0000000000..9d07bcb446 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbDataFileRW.c @@ -0,0 +1,1686 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbDataFileRW.h" + +extern int32_t tsdbFileWriteTombBlock(STsdbFD *fd, STombBlock *tombBlock, int8_t cmprAlg, int64_t *fileSize, + TTombBlkArray *tombBlkArray, uint8_t **bufArr); +extern int32_t tsdbFileWriteTombBlk(STsdbFD *fd, const TTombBlkArray *tombBlkArray, SFDataPtr *ptr, int64_t *fileSize); + +// SDataFileReader ============================================= +struct SDataFileReader { + SDataFileReaderConfig config[1]; + + uint8_t *bufArr[5]; + + struct { + bool headFooterLoaded; + bool tombFooterLoaded; + bool brinBlkLoaded; + bool tombBlkLoaded; + } ctx[1]; + + STsdbFD *fd[TSDB_FTYPE_MAX]; + + SHeadFooter headFooter[1]; + STombFooter tombFooter[1]; + TBrinBlkArray brinBlkArray[1]; + TTombBlkArray tombBlkArray[1]; +}; + +static int32_t tsdbDataFileReadHeadFooter(SDataFileReader *reader) { + if (reader->ctx->headFooterLoaded) return 0; + + int32_t code = 0; + int32_t lino = 0; + + int32_t ftype = TSDB_FTYPE_HEAD; + if (reader->fd[ftype]) { + code = tsdbReadFile(reader->fd[ftype], reader->config->files[ftype].file.size - sizeof(SHeadFooter), + (uint8_t *)reader->headFooter, sizeof(SHeadFooter)); + TSDB_CHECK_CODE(code, lino, _exit); + } + + reader->ctx->headFooterLoaded = true; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileReadTombFooter(SDataFileReader *reader) { + if (reader->ctx->tombFooterLoaded) return 0; + + int32_t code = 0; + int32_t lino = 0; + + int32_t ftype = TSDB_FTYPE_TOMB; + if (reader->fd[ftype]) { + code = tsdbReadFile(reader->fd[ftype], reader->config->files[ftype].file.size - sizeof(STombFooter), + (uint8_t *)reader->tombFooter, sizeof(STombFooter)); + TSDB_CHECK_CODE(code, lino, _exit); + } + reader->ctx->tombFooterLoaded = true; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileReaderOpen(const char *fname[], const SDataFileReaderConfig *config, SDataFileReader **reader) { + int32_t code = 0; + int32_t lino = 0; + + reader[0] = taosMemoryCalloc(1, sizeof(**reader)); + if (reader[0] == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + TSDB_CHECK_CODE(code, lino, _exit); + } + + reader[0]->config[0] = config[0]; + if (reader[0]->config->bufArr == NULL) { + reader[0]->config->bufArr = reader[0]->bufArr; + } + + if (fname) { + for (int32_t i = 0; i < TSDB_FTYPE_MAX; ++i) { + if (fname[i]) { + code = tsdbOpenFile(fname[i], config->szPage, TD_FILE_READ, &reader[0]->fd[i]); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + } else { + for (int32_t i = 0; i < TSDB_FTYPE_MAX; ++i) { + if (config->files[i].exist) { + char fname1[TSDB_FILENAME_LEN]; + tsdbTFileName(config->tsdb, &config->files[i].file, fname1); + code = tsdbOpenFile(fname1, config->szPage, TD_FILE_READ, &reader[0]->fd[i]); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileReaderClose(SDataFileReader **reader) { + if (reader[0] == NULL) return 0; + + TARRAY2_DESTROY(reader[0]->tombBlkArray, NULL); + TARRAY2_DESTROY(reader[0]->brinBlkArray, NULL); + +#if 0 + TARRAY2_DESTROY(reader[0]->dataBlkArray, NULL); + TARRAY2_DESTROY(reader[0]->blockIdxArray, NULL); +#endif + + for (int32_t i = 0; i < TSDB_FTYPE_MAX; ++i) { + if (reader[0]->fd[i]) { + tsdbCloseFile(&reader[0]->fd[i]); + } + } + + for (int32_t i = 0; i < ARRAY_SIZE(reader[0]->bufArr); ++i) { + tFree(reader[0]->bufArr[i]); + } + + taosMemoryFree(reader[0]); + reader[0] = NULL; + return 0; +} + +int32_t tsdbDataFileReadBrinBlk(SDataFileReader *reader, const TBrinBlkArray **brinBlkArray) { + int32_t code = 0; + int32_t lino = 0; + + if (!reader->ctx->brinBlkLoaded) { + code = tsdbDataFileReadHeadFooter(reader); + TSDB_CHECK_CODE(code, lino, _exit); + + if (reader->headFooter->brinBlkPtr->size > 0) { + void *data = taosMemoryMalloc(reader->headFooter->brinBlkPtr->size); + if (data == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbReadFile(reader->fd[TSDB_FTYPE_HEAD], reader->headFooter->brinBlkPtr->offset, data, + reader->headFooter->brinBlkPtr->size); + if (code) { + taosMemoryFree(data); + TSDB_CHECK_CODE(code, lino, _exit); + } + + int32_t size = reader->headFooter->brinBlkPtr->size / sizeof(SBrinBlk); + TARRAY2_INIT_EX(reader->brinBlkArray, size, size, data); + } else { + TARRAY2_INIT(reader->brinBlkArray); + } + + reader->ctx->brinBlkLoaded = true; + } + brinBlkArray[0] = reader->brinBlkArray; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileReadBrinBlock(SDataFileReader *reader, const SBrinBlk *brinBlk, SBrinBlock *brinBlock) { + int32_t code = 0; + int32_t lino = 0; + + code = tRealloc(&reader->config->bufArr[0], brinBlk->dp->size); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd[TSDB_FTYPE_HEAD], brinBlk->dp->offset, reader->config->bufArr[0], brinBlk->dp->size); + TSDB_CHECK_CODE(code, lino, _exit); + + int32_t size = 0; + tBrinBlockClear(brinBlock); + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr1); i++) { + code = tsdbDecmprData(reader->config->bufArr[0] + size, brinBlk->size[i], TSDB_DATA_TYPE_BIGINT, brinBlk->cmprAlg, + &reader->config->bufArr[1], brinBlk->numRec * sizeof(int64_t), &reader->config->bufArr[2]); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND_BATCH(&brinBlock->dataArr1[i], reader->config->bufArr[1], brinBlk->numRec); + TSDB_CHECK_CODE(code, lino, _exit); + + size += brinBlk->size[i]; + } + + for (int32_t i = 0, j = ARRAY_SIZE(brinBlock->dataArr1); i < ARRAY_SIZE(brinBlock->dataArr2); i++, j++) { + code = tsdbDecmprData(reader->config->bufArr[0] + size, brinBlk->size[j], TSDB_DATA_TYPE_INT, brinBlk->cmprAlg, + &reader->config->bufArr[1], brinBlk->numRec * sizeof(int32_t), &reader->config->bufArr[2]); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND_BATCH(&brinBlock->dataArr2[i], reader->config->bufArr[1], brinBlk->numRec); + TSDB_CHECK_CODE(code, lino, _exit); + + size += brinBlk->size[j]; + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileReadBlockData(SDataFileReader *reader, const SBrinRecord *record, SBlockData *bData) { + int32_t code = 0; + int32_t lino = 0; + + code = tRealloc(&reader->config->bufArr[0], record->blockSize); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd[TSDB_FTYPE_DATA], record->blockOffset, reader->config->bufArr[0], record->blockSize); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tDecmprBlockData(reader->config->bufArr[0], record->blockSize, bData, &reader->config->bufArr[1]); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileReadBlockDataByColumn(SDataFileReader *reader, const SBrinRecord *record, SBlockData *bData, + STSchema *pTSchema, int16_t cids[], int32_t ncid) { + int32_t code = 0; + int32_t lino = 0; + + code = tBlockDataInit(bData, (TABLEID *)record, pTSchema, cids, ncid); + TSDB_CHECK_CODE(code, lino, _exit); + + // uid + version + tskey + code = tRealloc(&reader->config->bufArr[0], record->blockKeySize); + TSDB_CHECK_CODE(code, lino, _exit); + + code = + tsdbReadFile(reader->fd[TSDB_FTYPE_DATA], record->blockOffset, reader->config->bufArr[0], record->blockKeySize); + TSDB_CHECK_CODE(code, lino, _exit); + + // hdr + SDiskDataHdr hdr[1]; + int32_t size = 0; + + size += tGetDiskDataHdr(reader->config->bufArr[0] + size, hdr); + + ASSERT(hdr->delimiter == TSDB_FILE_DLMT); + ASSERT(record->uid == hdr->uid); + + bData->nRow = hdr->nRow; + + // uid + ASSERT(hdr->uid); + + // version + code = tsdbDecmprData(reader->config->bufArr[0] + size, hdr->szVer, TSDB_DATA_TYPE_BIGINT, hdr->cmprAlg, + (uint8_t **)&bData->aVersion, sizeof(int64_t) * hdr->nRow, &reader->config->bufArr[1]); + TSDB_CHECK_CODE(code, lino, _exit); + size += hdr->szVer; + + // ts + code = tsdbDecmprData(reader->config->bufArr[0] + size, hdr->szKey, TSDB_DATA_TYPE_TIMESTAMP, hdr->cmprAlg, + (uint8_t **)&bData->aTSKEY, sizeof(TSKEY) * hdr->nRow, &reader->config->bufArr[1]); + TSDB_CHECK_CODE(code, lino, _exit); + size += hdr->szKey; + + ASSERT(size == record->blockKeySize); + + // other columns + if (bData->nColData > 0) { + if (hdr->szBlkCol > 0) { + code = tRealloc(&reader->config->bufArr[0], hdr->szBlkCol); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd[TSDB_FTYPE_DATA], record->blockOffset + record->blockKeySize, + reader->config->bufArr[0], hdr->szBlkCol); + TSDB_CHECK_CODE(code, lino, _exit); + } + + SBlockCol bc[1] = {{.cid = 0}}; + SBlockCol *blockCol = bc; + + size = 0; + for (int32_t i = 0; i < bData->nColData; i++) { + SColData *colData = tBlockDataGetColDataByIdx(bData, i); + + while (blockCol && blockCol->cid < colData->cid) { + if (size < hdr->szBlkCol) { + size += tGetBlockCol(reader->config->bufArr[0] + size, blockCol); + } else { + ASSERT(size == hdr->szBlkCol); + blockCol = NULL; + } + } + + if (blockCol == NULL || blockCol->cid > colData->cid) { + for (int32_t iRow = 0; iRow < hdr->nRow; iRow++) { + code = tColDataAppendValue(colData, &COL_VAL_NONE(colData->cid, colData->type)); + TSDB_CHECK_CODE(code, lino, _exit); + } + } else { + ASSERT(blockCol->type == colData->type); + ASSERT(blockCol->flag && blockCol->flag != HAS_NONE); + + if (blockCol->flag == HAS_NULL) { + for (int32_t iRow = 0; iRow < hdr->nRow; iRow++) { + code = tColDataAppendValue(colData, &COL_VAL_NULL(blockCol->cid, blockCol->type)); + TSDB_CHECK_CODE(code, lino, _exit); + } + } else { + int32_t size1 = blockCol->szBitmap + blockCol->szOffset + blockCol->szValue; + + code = tRealloc(&reader->config->bufArr[1], size1); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd[TSDB_FTYPE_DATA], + record->blockOffset + record->blockKeySize + hdr->szBlkCol + blockCol->offset, + reader->config->bufArr[1], size1); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDecmprColData(reader->config->bufArr[1], blockCol, hdr->cmprAlg, hdr->nRow, colData, + &reader->config->bufArr[2]); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileReadBlockSma(SDataFileReader *reader, const SBrinRecord *record, + TColumnDataAggArray *columnDataAggArray) { + int32_t code = 0; + int32_t lino = 0; + + TARRAY2_CLEAR(columnDataAggArray, NULL); + if (record->smaSize > 0) { + code = tRealloc(&reader->config->bufArr[0], record->smaSize); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd[TSDB_FTYPE_SMA], record->smaOffset, reader->config->bufArr[0], record->smaSize); + TSDB_CHECK_CODE(code, lino, _exit); + + // decode sma data + int32_t size = 0; + while (size < record->smaSize) { + SColumnDataAgg sma[1]; + + size += tGetColumnDataAgg(reader->config->bufArr[0] + size, sma); + + code = TARRAY2_APPEND_PTR(columnDataAggArray, sma); + TSDB_CHECK_CODE(code, lino, _exit); + } + ASSERT(size == record->smaSize); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileReadTombBlk(SDataFileReader *reader, const TTombBlkArray **tombBlkArray) { + int32_t code = 0; + int32_t lino = 0; + + if (!reader->ctx->tombBlkLoaded) { + code = tsdbDataFileReadTombFooter(reader); + TSDB_CHECK_CODE(code, lino, _exit); + + if (reader->tombFooter->tombBlkPtr->size > 0) { + void *data = taosMemoryMalloc(reader->tombFooter->tombBlkPtr->size); + if (data == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbReadFile(reader->fd[TSDB_FTYPE_TOMB], reader->tombFooter->tombBlkPtr->offset, data, + reader->tombFooter->tombBlkPtr->size); + if (code) { + taosMemoryFree(data); + TSDB_CHECK_CODE(code, lino, _exit); + } + + int32_t size = reader->tombFooter->tombBlkPtr->size / sizeof(STombBlk); + TARRAY2_INIT_EX(reader->tombBlkArray, size, size, data); + } else { + TARRAY2_INIT(reader->tombBlkArray); + } + + reader->ctx->tombBlkLoaded = true; + } + tombBlkArray[0] = reader->tombBlkArray; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileReadTombBlock(SDataFileReader *reader, const STombBlk *tombBlk, STombBlock *tData) { + int32_t code = 0; + int32_t lino = 0; + + code = tRealloc(&reader->config->bufArr[0], tombBlk->dp->size); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd[TSDB_FTYPE_TOMB], tombBlk->dp->offset, reader->config->bufArr[0], tombBlk->dp->size); + TSDB_CHECK_CODE(code, lino, _exit); + + int32_t size = 0; + tTombBlockClear(tData); + for (int32_t i = 0; i < ARRAY_SIZE(tData->dataArr); ++i) { + code = tsdbDecmprData(reader->config->bufArr[0] + size, tombBlk->size[i], TSDB_DATA_TYPE_BIGINT, tombBlk->cmprAlg, + &reader->config->bufArr[1], sizeof(int64_t) * tombBlk->numRec, &reader->config->bufArr[2]); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND_BATCH(&tData->dataArr[i], reader->config->bufArr[1], tombBlk->numRec); + TSDB_CHECK_CODE(code, lino, _exit); + + size += tombBlk->size[i]; + } + ASSERT(size == tombBlk->dp->size); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +// SDataFileWriter ============================================= +struct SDataFileWriter { + SDataFileWriterConfig config[1]; + + SSkmInfo skmTb[1]; + SSkmInfo skmRow[1]; + uint8_t *bufArr[5]; + + struct { + bool opened; + SDataFileReader *reader; + + // for ts data + TABLEID tbid[1]; + bool tbHasOldData; + + const TBrinBlkArray *brinBlkArray; + int32_t brinBlkArrayIdx; + SBrinBlock brinBlock[1]; + int32_t brinBlockIdx; + SBlockData blockData[1]; + int32_t blockDataIdx; + // for tomb data + bool hasOldTomb; + const TTombBlkArray *tombBlkArray; + int32_t tombBlkArrayIdx; + STombBlock tombBlock[1]; + int32_t tombBlockIdx; + } ctx[1]; + + STFile files[TSDB_FTYPE_MAX]; + STsdbFD *fd[TSDB_FTYPE_MAX]; + + SHeadFooter headFooter[1]; + STombFooter tombFooter[1]; + + TBrinBlkArray brinBlkArray[1]; + SBrinBlock brinBlock[1]; + SBlockData blockData[1]; + + TTombBlkArray tombBlkArray[1]; + STombBlock tombBlock[1]; +}; + +static int32_t tsdbDataFileWriterCloseAbort(SDataFileWriter *writer) { + ASSERT(0); + return 0; +} + +static int32_t tsdbDataFileWriterDoClose(SDataFileWriter *writer) { + if (writer->ctx->reader) { + tsdbDataFileReaderClose(&writer->ctx->reader); + } + + tTombBlockDestroy(writer->tombBlock); + TARRAY2_DESTROY(writer->tombBlkArray, NULL); + tBlockDataDestroy(writer->blockData); + tBrinBlockDestroy(writer->brinBlock); + TARRAY2_DESTROY(writer->brinBlkArray, NULL); + + tTombBlockDestroy(writer->ctx->tombBlock); + tBlockDataDestroy(writer->ctx->blockData); + tBrinBlockDestroy(writer->ctx->brinBlock); + + for (int32_t i = 0; i < ARRAY_SIZE(writer->bufArr); ++i) { + tFree(writer->bufArr[i]); + } + + tDestroyTSchema(writer->skmRow->pTSchema); + tDestroyTSchema(writer->skmTb->pTSchema); + return 0; +} + +static int32_t tsdbDataFileWriterDoOpenReader(SDataFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + + for (int32_t i = 0; i < TSDB_FTYPE_MAX; ++i) { + if (writer->config->files[i].exist) { + SDataFileReaderConfig config[1] = {{ + .tsdb = writer->config->tsdb, + .szPage = writer->config->szPage, + .bufArr = writer->config->bufArr, + }}; + + for (int32_t i = 0; i < TSDB_FTYPE_MAX; ++i) { + config->files[i].exist = writer->config->files[i].exist; + if (config->files[i].exist) { + config->files[i].file = writer->config->files[i].file; + } + } + + code = tsdbDataFileReaderOpen(NULL, config, &writer->ctx->reader); + TSDB_CHECK_CODE(code, lino, _exit); + break; + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileWriterDoOpen(SDataFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + int32_t ftype; + + if (!writer->config->skmTb) writer->config->skmTb = writer->skmTb; + if (!writer->config->skmRow) writer->config->skmRow = writer->skmRow; + if (!writer->config->bufArr) writer->config->bufArr = writer->bufArr; + + // open reader + code = tsdbDataFileWriterDoOpenReader(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + // .head + ftype = TSDB_FTYPE_HEAD; + writer->files[ftype] = (STFile){ + .type = ftype, + .did = writer->config->did, + .fid = writer->config->fid, + .cid = writer->config->cid, + .size = 0, + }; + + // .data + ftype = TSDB_FTYPE_DATA; + if (writer->config->files[ftype].exist) { + writer->files[ftype] = writer->config->files[ftype].file; + } else { + writer->files[ftype] = (STFile){ + .type = ftype, + .did = writer->config->did, + .fid = writer->config->fid, + .cid = writer->config->cid, + .size = 0, + }; + } + + // .sma + ftype = TSDB_FTYPE_SMA; + if (writer->config->files[ftype].exist) { + writer->files[ftype] = writer->config->files[ftype].file; + } else { + writer->files[ftype] = (STFile){ + .type = ftype, + .did = writer->config->did, + .fid = writer->config->fid, + .cid = writer->config->cid, + .size = 0, + }; + } + + // .tomb + ftype = TSDB_FTYPE_TOMB; + writer->files[ftype] = (STFile){ + .type = ftype, + .did = writer->config->did, + .fid = writer->config->fid, + .cid = writer->config->cid, + .size = 0, + }; + + writer->ctx->opened = true; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFileWriteBrinBlock(STsdbFD *fd, SBrinBlock *brinBlock, int8_t cmprAlg, int64_t *fileSize, + TBrinBlkArray *brinBlkArray, uint8_t **bufArr) { + if (BRIN_BLOCK_SIZE(brinBlock) == 0) return 0; + + int32_t code; + + // get SBrinBlk + SBrinBlk brinBlk[1] = { + { + .dp[0] = + { + .offset = *fileSize, + .size = 0, + }, + .minTbid = + { + .suid = TARRAY2_FIRST(brinBlock->suid), + .uid = TARRAY2_FIRST(brinBlock->uid), + }, + .maxTbid = + { + .suid = TARRAY2_LAST(brinBlock->suid), + .uid = TARRAY2_LAST(brinBlock->uid), + }, + .minVer = TARRAY2_FIRST(brinBlock->minVer), + .maxVer = TARRAY2_FIRST(brinBlock->minVer), + .numRec = BRIN_BLOCK_SIZE(brinBlock), + .cmprAlg = cmprAlg, + }, + }; + + for (int32_t i = 1; i < BRIN_BLOCK_SIZE(brinBlock); i++) { + if (brinBlk->minVer > TARRAY2_GET(brinBlock->minVer, i)) { + brinBlk->minVer = TARRAY2_GET(brinBlock->minVer, i); + } + if (brinBlk->maxVer < TARRAY2_GET(brinBlock->maxVer, i)) { + brinBlk->maxVer = TARRAY2_GET(brinBlock->maxVer, i); + } + } + + // write to file + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr1); i++) { + code = tsdbCmprData((uint8_t *)TARRAY2_DATA(brinBlock->dataArr1 + i), TARRAY2_DATA_LEN(brinBlock->dataArr1 + i), + TSDB_DATA_TYPE_BIGINT, brinBlk->cmprAlg, &bufArr[0], 0, &brinBlk->size[i], &bufArr[1]); + if (code) return code; + + code = tsdbWriteFile(fd, *fileSize, bufArr[0], brinBlk->size[i]); + if (code) return code; + + brinBlk->dp->size += brinBlk->size[i]; + *fileSize += brinBlk->size[i]; + } + + for (int32_t i = 0, j = ARRAY_SIZE(brinBlock->dataArr1); i < ARRAY_SIZE(brinBlock->dataArr2); i++, j++) { + code = tsdbCmprData((uint8_t *)TARRAY2_DATA(brinBlock->dataArr2 + i), TARRAY2_DATA_LEN(brinBlock->dataArr2 + i), + TSDB_DATA_TYPE_INT, brinBlk->cmprAlg, &bufArr[0], 0, &brinBlk->size[j], &bufArr[1]); + if (code) return code; + + code = tsdbWriteFile(fd, *fileSize, bufArr[0], brinBlk->size[j]); + if (code) return code; + + brinBlk->dp->size += brinBlk->size[j]; + *fileSize += brinBlk->size[j]; + } + +#if 0 + SBrinRecord record; + for (int32_t i = 0; i < BRIN_BLOCK_SIZE(brinBlock); i++) { + tBrinBlockGet(brinBlock, i, &record); + tsdbInfo("write brin block, block num:%04d, idx:%04d suid:%ld, uid:%ld, offset:%ld, numRow:%d, count:%d", + TARRAY2_SIZE(brinBlkArray), i, record.suid, record.uid, record.blockOffset, record.numRow, record.count); + } +#endif + + // append to brinBlkArray + code = TARRAY2_APPEND_PTR(brinBlkArray, brinBlk); + if (code) return code; + + tBrinBlockClear(brinBlock); + + return 0; +} + +static int32_t tsdbDataFileWriteBrinBlock(SDataFileWriter *writer) { + if (BRIN_BLOCK_SIZE(writer->brinBlock) == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileWriteBrinBlock(writer->fd[TSDB_FTYPE_HEAD], writer->brinBlock, writer->config->cmprAlg, + &writer->files[TSDB_FTYPE_HEAD].size, writer->brinBlkArray, writer->config->bufArr); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileWriteBrinRecord(SDataFileWriter *writer, const SBrinRecord *record) { + int32_t code = 0; + int32_t lino = 0; + + code = tBrinBlockPut(writer->brinBlock, record); + TSDB_CHECK_CODE(code, lino, _exit); + + if (BRIN_BLOCK_SIZE(writer->brinBlock) >= writer->config->maxRow) { + code = tsdbDataFileWriteBrinBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileDoWriteBlockData(SDataFileWriter *writer, SBlockData *bData) { + if (bData->nRow == 0) return 0; + + ASSERT(bData->uid); + + int32_t code = 0; + int32_t lino = 0; + + SBrinRecord record[1] = {{ + .suid = bData->suid, + .uid = bData->uid, + .firstKey = bData->aTSKEY[0], + .firstKeyVer = bData->aVersion[0], + .lastKey = bData->aTSKEY[bData->nRow - 1], + .lastKeyVer = bData->aVersion[bData->nRow - 1], + .minVer = bData->aVersion[0], + .maxVer = bData->aVersion[0], + .blockOffset = writer->files[TSDB_FTYPE_DATA].size, + .smaOffset = writer->files[TSDB_FTYPE_SMA].size, + .blockSize = 0, + .blockKeySize = 0, + .smaSize = 0, + .numRow = bData->nRow, + .count = 1, + }}; + + for (int32_t i = 1; i < bData->nRow; ++i) { + if (bData->aTSKEY[i] != bData->aTSKEY[i - 1]) { + record->count++; + } + if (bData->aVersion[i] < record->minVer) { + record->minVer = bData->aVersion[i]; + } + if (bData->aVersion[i] > record->maxVer) { + record->maxVer = bData->aVersion[i]; + } + } + + // to .data file + int32_t sizeArr[5] = {0}; + + code = tCmprBlockData(bData, writer->config->cmprAlg, NULL, NULL, writer->config->bufArr, sizeArr); + TSDB_CHECK_CODE(code, lino, _exit); + + record->blockKeySize = sizeArr[3] + sizeArr[2]; + record->blockSize = sizeArr[0] + sizeArr[1] + record->blockKeySize; + + for (int32_t i = 3; i >= 0; --i) { + if (sizeArr[i]) { + code = tsdbWriteFile(writer->fd[TSDB_FTYPE_DATA], writer->files[TSDB_FTYPE_DATA].size, writer->config->bufArr[i], + sizeArr[i]); + TSDB_CHECK_CODE(code, lino, _exit); + writer->files[TSDB_FTYPE_DATA].size += sizeArr[i]; + } + } + + // to .sma file + for (int32_t i = 0; i < bData->nColData; ++i) { + SColData *colData = bData->aColData + i; + if ((!colData->smaOn) || ((colData->flag & HAS_VALUE) == 0)) continue; + + SColumnDataAgg sma[1] = {{.colId = colData->cid}}; + tColDataCalcSMA[colData->type](colData, &sma->sum, &sma->max, &sma->min, &sma->numOfNull); + + int32_t size = tPutColumnDataAgg(NULL, sma); + + code = tRealloc(&writer->config->bufArr[0], record->smaSize + size); + TSDB_CHECK_CODE(code, lino, _exit); + + tPutColumnDataAgg(writer->config->bufArr[0] + record->smaSize, sma); + record->smaSize += size; + } + + if (record->smaSize > 0) { + code = tsdbWriteFile(writer->fd[TSDB_FTYPE_SMA], record->smaOffset, writer->config->bufArr[0], record->smaSize); + TSDB_CHECK_CODE(code, lino, _exit); + writer->files[TSDB_FTYPE_SMA].size += record->smaSize; + } + + // append SBrinRecord + code = tsdbDataFileWriteBrinRecord(writer, record); + TSDB_CHECK_CODE(code, lino, _exit); + + tBlockDataClear(bData); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileWriteDataBlk(SDataFileWriter *writer, const TDataBlkArray *dataBlkArray) { + if (TARRAY2_SIZE(dataBlkArray) == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + int32_t ftype = TSDB_FTYPE_HEAD; + SBlockIdx blockIdx[1] = {{ + .suid = writer->ctx->tbid->suid, + .uid = writer->ctx->tbid->uid, + .offset = writer->files[ftype].size, + .size = TARRAY2_DATA_LEN(dataBlkArray), + }}; + + code = + tsdbWriteFile(writer->fd[ftype], blockIdx->offset, (const uint8_t *)TARRAY2_DATA(dataBlkArray), blockIdx->size); + TSDB_CHECK_CODE(code, lino, _exit); + writer->files[ftype].size += blockIdx->size; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileDoWriteTSRow(SDataFileWriter *writer, TSDBROW *row) { + int32_t code = 0; + int32_t lino = 0; + + // update/append + if (row->type == TSDBROW_ROW_FMT) { + code = tsdbUpdateSkmRow(writer->config->tsdb, writer->ctx->tbid, TSDBROW_SVERSION(row), writer->config->skmRow); + TSDB_CHECK_CODE(code, lino, _exit); + } + + TSDBKEY key[1]; + if (row->type == TSDBROW_ROW_FMT) { + key->ts = row->pTSRow->ts; + key->version = row->version; + } else { + key->ts = row->pBlockData->aTSKEY[row->iRow]; + key->version = row->pBlockData->aVersion[row->iRow]; + } + if (key->version <= writer->config->compactVersion // + && writer->blockData->nRow > 0 // + && writer->blockData->aTSKEY[writer->blockData->nRow - 1] == key->ts // + ) { + code = tBlockDataUpdateRow(writer->blockData, row, writer->config->skmRow->pTSchema); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + if (writer->blockData->nRow >= writer->config->maxRow) { + code = tsdbDataFileDoWriteBlockData(writer, writer->blockData); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tBlockDataAppendRow(writer->blockData, row, writer->config->skmRow->pTSchema, writer->ctx->tbid->uid); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileDoWriteTableOldData(SDataFileWriter *writer, const TSDBKEY *key) { + if (writer->ctx->tbHasOldData == false) return 0; + + int32_t code = 0; + int32_t lino = 0; + + for (;;) { + for (;;) { + // SBlockData + for (; writer->ctx->blockDataIdx < writer->ctx->blockData->nRow; writer->ctx->blockDataIdx++) { + if (key->ts < writer->ctx->blockData->aTSKEY[writer->ctx->blockDataIdx] // + || (key->ts == writer->ctx->blockData->aTSKEY[writer->ctx->blockDataIdx] && + key->version < writer->ctx->blockData->aVersion[writer->ctx->blockDataIdx])) { + goto _exit; + } else { + TSDBROW row = tsdbRowFromBlockData(writer->ctx->blockData, writer->ctx->blockDataIdx); + code = tsdbDataFileDoWriteTSRow(writer, &row); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + + // SBrinBlock + if (writer->ctx->brinBlockIdx >= BRIN_BLOCK_SIZE(writer->ctx->brinBlock)) { + break; + } + + for (; writer->ctx->brinBlockIdx < BRIN_BLOCK_SIZE(writer->ctx->brinBlock); writer->ctx->brinBlockIdx++) { + if (TARRAY2_GET(writer->ctx->brinBlock->uid, writer->ctx->brinBlockIdx) != writer->ctx->tbid->uid) { + writer->ctx->tbHasOldData = false; + goto _exit; + } + + if (key->ts < TARRAY2_GET(writer->ctx->brinBlock->firstKey, writer->ctx->brinBlockIdx) // + || (key->ts == TARRAY2_GET(writer->ctx->brinBlock->firstKey, writer->ctx->brinBlockIdx) && + key->version < TARRAY2_GET(writer->ctx->brinBlock->firstKeyVer, writer->ctx->brinBlockIdx))) { + goto _exit; + } else { + SBrinRecord record[1]; + tBrinBlockGet(writer->ctx->brinBlock, writer->ctx->brinBlockIdx, record); + if (key->ts > record->lastKey || (key->ts == record->lastKey && key->version > record->maxVer)) { + if (writer->blockData->nRow > 0) { + code = tsdbDataFileDoWriteBlockData(writer, writer->blockData); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbDataFileWriteBrinRecord(writer, record); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + code = tsdbDataFileReadBlockData(writer->ctx->reader, record, writer->ctx->blockData); + TSDB_CHECK_CODE(code, lino, _exit); + + writer->ctx->blockDataIdx = 0; + writer->ctx->brinBlockIdx++; + break; + } + } + } + } + + // SBrinBlk + if (writer->ctx->brinBlkArrayIdx >= TARRAY2_SIZE(writer->ctx->brinBlkArray)) { + writer->ctx->brinBlkArray = NULL; + writer->ctx->tbHasOldData = false; + goto _exit; + } + + for (; writer->ctx->brinBlkArrayIdx < TARRAY2_SIZE(writer->ctx->brinBlkArray); writer->ctx->brinBlkArrayIdx++) { + const SBrinBlk *brinBlk = TARRAY2_GET_PTR(writer->ctx->brinBlkArray, writer->ctx->brinBlkArrayIdx); + + if (brinBlk->minTbid.uid != writer->ctx->tbid->uid) { + writer->ctx->tbHasOldData = false; + goto _exit; + } + + code = tsdbDataFileReadBrinBlock(writer->ctx->reader, brinBlk, writer->ctx->brinBlock); + TSDB_CHECK_CODE(code, lino, _exit); + + writer->ctx->brinBlockIdx = 0; + writer->ctx->brinBlkArrayIdx++; + break; + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileDoWriteTSData(SDataFileWriter *writer, TSDBROW *row) { + int32_t code = 0; + int32_t lino = 0; + + if (writer->ctx->tbHasOldData) { + TSDBKEY key[1]; + if (row->type == TSDBROW_ROW_FMT) { + key->ts = row->pTSRow->ts; + key->version = row->version; + } else { + key->ts = row->pBlockData->aTSKEY[row->iRow]; + key->version = row->pBlockData->aVersion[row->iRow]; + } + + code = tsdbDataFileDoWriteTableOldData(writer, key); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbDataFileDoWriteTSRow(writer, row); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileWriteTableDataEnd(SDataFileWriter *writer) { + if (writer->ctx->tbid->uid == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + if (writer->ctx->tbHasOldData) { + TSDBKEY key = { + .ts = TSKEY_MAX, + .version = VERSION_MAX, + }; + + code = tsdbDataFileDoWriteTableOldData(writer, &key); + TSDB_CHECK_CODE(code, lino, _exit); + + ASSERT(writer->ctx->tbHasOldData == false); + } + + code = tsdbDataFileDoWriteBlockData(writer, writer->blockData); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileWriteTableDataBegin(SDataFileWriter *writer, const TABLEID *tbid) { + int32_t code = 0; + int32_t lino = 0; + + ASSERT(writer->ctx->blockDataIdx == writer->ctx->blockData->nRow); + ASSERT(writer->blockData->nRow == 0); + + SMetaInfo info; + bool drop = false; + TABLEID tbid1[1]; + writer->ctx->tbHasOldData = false; + while (writer->ctx->brinBlkArray) { // skip data of previous table + for (; writer->ctx->brinBlockIdx < BRIN_BLOCK_SIZE(writer->ctx->brinBlock); writer->ctx->brinBlockIdx++) { + TABLEID tbid2[1] = {{ + .suid = TARRAY2_GET(writer->ctx->brinBlock->suid, writer->ctx->brinBlockIdx), + .uid = TARRAY2_GET(writer->ctx->brinBlock->uid, writer->ctx->brinBlockIdx), + }}; + + if (tbid2->uid == tbid->uid) { + writer->ctx->tbHasOldData = true; + goto _begin; + } else if (tbid2->suid > tbid->suid || (tbid2->suid == tbid->suid && tbid2->uid > tbid->uid)) { + goto _begin; + } else { + if (tbid2->uid != writer->ctx->tbid->uid) { + if (drop && tbid1->uid == tbid2->uid) { + continue; + } else if (metaGetInfo(writer->config->tsdb->pVnode->pMeta, tbid2->uid, &info, NULL) != 0) { + drop = true; + *tbid1 = *tbid2; + continue; + } else { + drop = false; + writer->ctx->tbid[0] = *tbid2; + } + } + + SBrinRecord record[1]; + tBrinBlockGet(writer->ctx->brinBlock, writer->ctx->brinBlockIdx, record); + + code = tsdbDataFileWriteBrinRecord(writer, record); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + + if (writer->ctx->brinBlkArrayIdx >= TARRAY2_SIZE(writer->ctx->brinBlkArray)) { + writer->ctx->brinBlkArray = NULL; + break; + } + + for (; writer->ctx->brinBlkArrayIdx < TARRAY2_SIZE(writer->ctx->brinBlkArray); writer->ctx->brinBlkArrayIdx++) { + const SBrinBlk *brinBlk = TARRAY2_GET_PTR(writer->ctx->brinBlkArray, writer->ctx->brinBlkArrayIdx); + + code = tsdbDataFileReadBrinBlock(writer->ctx->reader, brinBlk, writer->ctx->brinBlock); + TSDB_CHECK_CODE(code, lino, _exit); + + writer->ctx->brinBlockIdx = 0; + writer->ctx->brinBlkArrayIdx++; + break; + } + } + +_begin: + writer->ctx->tbid[0] = *tbid; + + if (tbid->uid == INT64_MAX) goto _exit; + + code = tsdbUpdateSkmTb(writer->config->tsdb, tbid, writer->config->skmTb); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tBlockDataInit(writer->blockData, writer->ctx->tbid, writer->config->skmTb->pTSchema, NULL, 0); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFileWriteHeadFooter(STsdbFD *fd, int64_t *fileSize, const SHeadFooter *footer) { + int32_t code = tsdbWriteFile(fd, *fileSize, (const uint8_t *)footer, sizeof(*footer)); + if (code) return code; + *fileSize += sizeof(*footer); + return 0; +} + +static int32_t tsdbDataFileWriteHeadFooter(SDataFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileWriteHeadFooter(writer->fd[TSDB_FTYPE_HEAD], &writer->files[TSDB_FTYPE_HEAD].size, writer->headFooter); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileDoWriteTombBlock(SDataFileWriter *writer) { + if (TOMB_BLOCK_SIZE(writer->tombBlock) == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileWriteTombBlock(writer->fd[TSDB_FTYPE_TOMB], writer->tombBlock, writer->config->cmprAlg, + &writer->files[TSDB_FTYPE_TOMB].size, writer->tombBlkArray, writer->config->bufArr); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileDoWriteTombBlk(SDataFileWriter *writer) { + ASSERT(TARRAY2_SIZE(writer->tombBlkArray) > 0); + + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileWriteTombBlk(writer->fd[TSDB_FTYPE_TOMB], writer->tombBlkArray, writer->tombFooter->tombBlkPtr, + &writer->files[TSDB_FTYPE_TOMB].size); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFileWriteTombFooter(STsdbFD *fd, const STombFooter *footer, int64_t *fileSize) { + int32_t code = tsdbWriteFile(fd, *fileSize, (const uint8_t *)footer, sizeof(*footer)); + if (code) return code; + *fileSize += sizeof(*footer); + return 0; +} + +static int32_t tsdbDataFileWriteTombFooter(SDataFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileWriteTombFooter(writer->fd[TSDB_FTYPE_TOMB], writer->tombFooter, &writer->files[TSDB_FTYPE_TOMB].size); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileDoWriteTombRecord(SDataFileWriter *writer, const STombRecord *record) { + int32_t code = 0; + int32_t lino = 0; + + while (writer->ctx->hasOldTomb) { + for (; writer->ctx->tombBlockIdx < TOMB_BLOCK_SIZE(writer->ctx->tombBlock); writer->ctx->tombBlockIdx++) { + STombRecord record1[1]; + tTombBlockGet(writer->ctx->tombBlock, writer->ctx->tombBlockIdx, record1); + + int32_t c = tTombRecordCompare(record, record1); + if (c < 0) { + break; + } else if (c > 0) { + code = tTombBlockPut(writer->tombBlock, record1); + TSDB_CHECK_CODE(code, lino, _exit); + + if (TOMB_BLOCK_SIZE(writer->tombBlock) >= writer->config->maxRow) { + code = tsdbDataFileDoWriteTombBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + } else { + ASSERT(0); + } + } + + if (writer->ctx->tombBlkArrayIdx >= TARRAY2_SIZE(writer->ctx->tombBlkArray)) { + writer->ctx->hasOldTomb = false; + break; + } + + for (; writer->ctx->tombBlkArrayIdx < TARRAY2_SIZE(writer->ctx->tombBlkArray); ++writer->ctx->tombBlkArrayIdx) { + const STombBlk *tombBlk = TARRAY2_GET_PTR(writer->ctx->tombBlkArray, writer->ctx->tombBlkArrayIdx); + + code = tsdbDataFileReadTombBlock(writer->ctx->reader, tombBlk, writer->ctx->tombBlock); + TSDB_CHECK_CODE(code, lino, _exit); + + writer->ctx->tombBlockIdx = 0; + writer->ctx->tombBlkArrayIdx++; + break; + } + } + +_write: + if (record->suid == INT64_MAX) goto _exit; + + code = tTombBlockPut(writer->tombBlock, record); + TSDB_CHECK_CODE(code, lino, _exit); + + if (TOMB_BLOCK_SIZE(writer->tombBlock) >= writer->config->maxRow) { + code = tsdbDataFileDoWriteTombBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFileWriteBrinBlk(STsdbFD *fd, TBrinBlkArray *brinBlkArray, SFDataPtr *ptr, int64_t *fileSize) { + ASSERT(TARRAY2_SIZE(brinBlkArray) > 0); + ptr->offset = *fileSize; + ptr->size = TARRAY2_DATA_LEN(brinBlkArray); + + int32_t code = tsdbWriteFile(fd, ptr->offset, (uint8_t *)TARRAY2_DATA(brinBlkArray), ptr->size); + if (code) return code; + + *fileSize += ptr->size; + return 0; +} + +static int32_t tsdbDataFileWriteBrinBlk(SDataFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileWriteBrinBlk(writer->fd[TSDB_FTYPE_HEAD], writer->brinBlkArray, writer->headFooter->brinBlkPtr, + &writer->files[TSDB_FTYPE_HEAD].size); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileWriterCloseCommit(SDataFileWriter *writer, TFileOpArray *opArr) { + int32_t code = 0; + int32_t lino = 0; + + int32_t ftype; + STFileOp op; + + if (writer->fd[TSDB_FTYPE_HEAD]) { + TABLEID tbid[1] = {{ + .suid = INT64_MAX, + .uid = INT64_MAX, + }}; + + code = tsdbDataFileWriteTableDataEnd(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileWriteTableDataBegin(writer, tbid); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileWriteBrinBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileWriteBrinBlk(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileWriteHeadFooter(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + // .head + ftype = TSDB_FTYPE_HEAD; + if (writer->config->files[ftype].exist) { + op = (STFileOp){ + .optype = TSDB_FOP_REMOVE, + .fid = writer->config->fid, + .of = writer->config->files[ftype].file, + }; + code = TARRAY2_APPEND(opArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + } + op = (STFileOp){ + .optype = TSDB_FOP_CREATE, + .fid = writer->config->fid, + .nf = writer->files[ftype], + }; + code = TARRAY2_APPEND(opArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + + // .data + ftype = TSDB_FTYPE_DATA; + if (!writer->config->files[ftype].exist) { + op = (STFileOp){ + .optype = TSDB_FOP_CREATE, + .fid = writer->config->fid, + .nf = writer->files[ftype], + }; + code = TARRAY2_APPEND(opArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + } else if (writer->config->files[ftype].file.size != writer->files[ftype].size) { + op = (STFileOp){ + .optype = TSDB_FOP_MODIFY, + .fid = writer->config->fid, + .of = writer->config->files[ftype].file, + .nf = writer->files[ftype], + }; + code = TARRAY2_APPEND(opArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // .sma + ftype = TSDB_FTYPE_SMA; + if (!writer->config->files[ftype].exist) { + op = (STFileOp){ + .optype = TSDB_FOP_CREATE, + .fid = writer->config->fid, + .nf = writer->files[ftype], + }; + code = TARRAY2_APPEND(opArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + } else if (writer->config->files[ftype].file.size != writer->files[ftype].size) { + op = (STFileOp){ + .optype = TSDB_FOP_MODIFY, + .fid = writer->config->fid, + .of = writer->config->files[ftype].file, + .nf = writer->files[ftype], + }; + code = TARRAY2_APPEND(opArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + + if (writer->fd[TSDB_FTYPE_TOMB]) { + STombRecord record[1] = {{ + .suid = INT64_MAX, + .uid = INT64_MAX, + .version = INT64_MAX, + }}; + + code = tsdbDataFileDoWriteTombRecord(writer, record); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileDoWriteTombBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileDoWriteTombBlk(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileWriteTombFooter(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + ftype = TSDB_FTYPE_TOMB; + if (writer->config->files[ftype].exist) { + op = (STFileOp){ + .optype = TSDB_FOP_REMOVE, + .fid = writer->config->fid, + .of = writer->config->files[ftype].file, + }; + code = TARRAY2_APPEND(opArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + } + op = (STFileOp){ + .optype = TSDB_FOP_CREATE, + .fid = writer->config->fid, + .nf = writer->files[ftype], + }; + code = TARRAY2_APPEND(opArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + } + + for (int32_t i = 0; i < TSDB_FTYPE_MAX; ++i) { + if (writer->fd[i]) { + code = tsdbFsyncFile(writer->fd[i]); + TSDB_CHECK_CODE(code, lino, _exit); + tsdbCloseFile(&writer->fd[i]); + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDataFileWriterOpenDataFD(SDataFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + + int32_t ftypes[] = {TSDB_FTYPE_HEAD, TSDB_FTYPE_DATA, TSDB_FTYPE_SMA}; + + for (int32_t i = 0; i < ARRAY_SIZE(ftypes); ++i) { + int32_t ftype = ftypes[i]; + + char fname[TSDB_FILENAME_LEN]; + int32_t flag = TD_FILE_READ | TD_FILE_WRITE; + + if (writer->files[ftype].size == 0) { + flag |= (TD_FILE_CREATE | TD_FILE_TRUNC); + } + + tsdbTFileName(writer->config->tsdb, &writer->files[ftype], fname); + code = tsdbOpenFile(fname, writer->config->szPage, flag, &writer->fd[ftype]); + TSDB_CHECK_CODE(code, lino, _exit); + + if (writer->files[ftype].size == 0) { + uint8_t hdr[TSDB_FHDR_SIZE] = {0}; + + code = tsdbWriteFile(writer->fd[ftype], 0, hdr, TSDB_FHDR_SIZE); + TSDB_CHECK_CODE(code, lino, _exit); + + writer->files[ftype].size += TSDB_FHDR_SIZE; + } + } + + if (writer->ctx->reader) { + code = tsdbDataFileReadBrinBlk(writer->ctx->reader, &writer->ctx->brinBlkArray); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileWriterOpen(const SDataFileWriterConfig *config, SDataFileWriter **writer) { + writer[0] = taosMemoryCalloc(1, sizeof(*writer[0])); + if (!writer[0]) return TSDB_CODE_OUT_OF_MEMORY; + + writer[0]->config[0] = config[0]; + return 0; +} + +int32_t tsdbDataFileWriterClose(SDataFileWriter **writer, bool abort, TFileOpArray *opArr) { + if (writer[0] == NULL) return 0; + + int32_t code = 0; + int32_t lino = 0; + + if (writer[0]->ctx->opened) { + if (abort) { + code = tsdbDataFileWriterCloseAbort(writer[0]); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + code = tsdbDataFileWriterCloseCommit(writer[0], opArr); + TSDB_CHECK_CODE(code, lino, _exit); + } + tsdbDataFileWriterDoClose(writer[0]); + } + taosMemoryFree(writer[0]); + writer[0] = NULL; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer[0]->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileWriteRow(SDataFileWriter *writer, SRowInfo *row) { + int32_t code = 0; + int32_t lino = 0; + + if (!writer->ctx->opened) { + code = tsdbDataFileWriterDoOpen(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (writer->fd[TSDB_FTYPE_HEAD] == NULL) { + code = tsdbDataFileWriterOpenDataFD(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (row->uid != writer->ctx->tbid->uid) { + code = tsdbDataFileWriteTableDataEnd(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileWriteTableDataBegin(writer, (TABLEID *)row); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbDataFileDoWriteTSData(writer, &row->row); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileWriteBlockData(SDataFileWriter *writer, SBlockData *bData) { + if (bData->nRow == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + ASSERT(bData->uid); + + if (!writer->ctx->opened) { + code = tsdbDataFileWriterDoOpen(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (writer->fd[TSDB_FTYPE_DATA] == NULL) { + code = tsdbDataFileWriterOpenDataFD(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (bData->uid != writer->ctx->tbid->uid) { + code = tsdbDataFileWriteTableDataEnd(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileWriteTableDataBegin(writer, (TABLEID *)bData); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (writer->ctx->tbHasOldData) { + TSDBKEY key = { + .ts = bData->aTSKEY[0], + .version = bData->aVersion[0], + }; + + code = tsdbDataFileDoWriteTableOldData(writer, &key); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (!writer->ctx->tbHasOldData // + && writer->blockData->nRow == 0 // + ) { + code = tsdbDataFileDoWriteBlockData(writer, bData); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + for (int32_t i = 0; i < bData->nRow; ++i) { + TSDBROW row[1] = {tsdbRowFromBlockData(bData, i)}; + code = tsdbDataFileDoWriteTSData(writer, row); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileFlush(SDataFileWriter *writer) { + ASSERT(writer->ctx->opened); + + if (writer->blockData->nRow == 0) return 0; + if (writer->ctx->tbHasOldData) return 0; + + return tsdbDataFileDoWriteBlockData(writer, writer->blockData); +} + +static int32_t tsdbDataFileWriterOpenTombFD(SDataFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + + char fname[TSDB_FILENAME_LEN]; + int32_t ftype = TSDB_FTYPE_TOMB; + + ASSERT(writer->files[ftype].size == 0); + + int32_t flag = (TD_FILE_READ | TD_FILE_WRITE | TD_FILE_CREATE | TD_FILE_TRUNC); + + tsdbTFileName(writer->config->tsdb, writer->files + ftype, fname); + code = tsdbOpenFile(fname, writer->config->szPage, flag, &writer->fd[ftype]); + TSDB_CHECK_CODE(code, lino, _exit); + + uint8_t hdr[TSDB_FHDR_SIZE] = {0}; + code = tsdbWriteFile(writer->fd[ftype], 0, hdr, TSDB_FHDR_SIZE); + TSDB_CHECK_CODE(code, lino, _exit); + writer->files[ftype].size += TSDB_FHDR_SIZE; + + if (writer->ctx->reader) { + code = tsdbDataFileReadTombBlk(writer->ctx->reader, &writer->ctx->tombBlkArray); + TSDB_CHECK_CODE(code, lino, _exit); + + if (TARRAY2_SIZE(writer->ctx->tombBlkArray) > 0) { + writer->ctx->hasOldTomb = true; + } + + writer->ctx->tombBlkArrayIdx = 0; + tTombBlockClear(writer->ctx->tombBlock); + writer->ctx->tombBlockIdx = 0; + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbDataFileWriteTombRecord(SDataFileWriter *writer, const STombRecord *record) { + int32_t code = 0; + int32_t lino = 0; + + if (!writer->ctx->opened) { + code = tsdbDataFileWriterDoOpen(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (writer->fd[TSDB_FTYPE_TOMB] == NULL) { + code = tsdbDataFileWriterOpenTombFD(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbDataFileDoWriteTombRecord(writer, record); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbDataFileRW.h b/source/dnode/vnode/src/tsdb/tsdbDataFileRW.h new file mode 100644 index 0000000000..827b58fb4a --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbDataFileRW.h @@ -0,0 +1,104 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbDef.h" +#include "tsdbFSet2.h" +#include "tsdbSttFileRW.h" +#include "tsdbUtil2.h" + +#ifndef _TSDB_DATA_FILE_RW_H +#define _TSDB_DATA_FILE_RW_H + +#ifdef __cplusplus +extern "C" { +#endif + +typedef TARRAY2(SBlockIdx) TBlockIdxArray; +typedef TARRAY2(SDataBlk) TDataBlkArray; +typedef TARRAY2(SColumnDataAgg) TColumnDataAggArray; + +typedef struct { + SFDataPtr brinBlkPtr[1]; + SFDataPtr rsrvd[2]; +} SHeadFooter; + +typedef struct { + SFDataPtr tombBlkPtr[1]; + SFDataPtr rsrvd[2]; +} STombFooter; + +// SDataFileReader ============================================= +typedef struct SDataFileReader SDataFileReader; +typedef struct SDataFileReaderConfig { + STsdb *tsdb; + int32_t szPage; + struct { + bool exist; + STFile file; + } files[TSDB_FTYPE_MAX]; + uint8_t **bufArr; +} SDataFileReaderConfig; + +int32_t tsdbDataFileReaderOpen(const char *fname[/* TSDB_FTYPE_MAX */], const SDataFileReaderConfig *config, + SDataFileReader **reader); +int32_t tsdbDataFileReaderClose(SDataFileReader **reader); +// .head +int32_t tsdbDataFileReadBrinBlk(SDataFileReader *reader, const TBrinBlkArray **brinBlkArray); +int32_t tsdbDataFileReadBrinBlock(SDataFileReader *reader, const SBrinBlk *brinBlk, SBrinBlock *brinBlock); +// .data +int32_t tsdbDataFileReadBlockData(SDataFileReader *reader, const SBrinRecord *record, SBlockData *bData); +int32_t tsdbDataFileReadBlockDataByColumn(SDataFileReader *reader, const SBrinRecord *record, SBlockData *bData, + STSchema *pTSchema, int16_t cids[], int32_t ncid); +// .sma +int32_t tsdbDataFileReadBlockSma(SDataFileReader *reader, const SBrinRecord *record, + TColumnDataAggArray *columnDataAggArray); +// .tomb +int32_t tsdbDataFileReadTombBlk(SDataFileReader *reader, const TTombBlkArray **tombBlkArray); +int32_t tsdbDataFileReadTombBlock(SDataFileReader *reader, const STombBlk *tombBlk, STombBlock *tData); + +// SDataFileWriter ============================================= +typedef struct SDataFileWriter SDataFileWriter; +typedef struct SDataFileWriterConfig { + STsdb *tsdb; + int8_t cmprAlg; + int32_t maxRow; + int32_t szPage; + int32_t fid; + int64_t cid; + SDiskID did; + int64_t compactVersion; + struct { + bool exist; + STFile file; + } files[TSDB_FTYPE_MAX]; + SSkmInfo *skmTb; + SSkmInfo *skmRow; + uint8_t **bufArr; +} SDataFileWriterConfig; + +int32_t tsdbDataFileWriterOpen(const SDataFileWriterConfig *config, SDataFileWriter **writer); +int32_t tsdbDataFileWriterClose(SDataFileWriter **writer, bool abort, TFileOpArray *opArr); + +int32_t tsdbDataFileWriteRow(SDataFileWriter *writer, SRowInfo *row); +int32_t tsdbDataFileWriteBlockData(SDataFileWriter *writer, SBlockData *bData); +int32_t tsdbDataFileFlush(SDataFileWriter *writer); + +int32_t tsdbDataFileWriteTombRecord(SDataFileWriter *writer, const STombRecord *record); + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_DATA_FILE_RW_H*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbDef.h b/source/dnode/vnode/src/tsdb/tsdbDef.h new file mode 100644 index 0000000000..e768f68b15 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbDef.h @@ -0,0 +1,44 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tarray2.h" +#include "tsdb.h" + +#ifndef _TD_TSDB_DEF_H_ +#define _TD_TSDB_DEF_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +#define TSDB_ERROR_LOG(vid, lino, code) \ + tsdbError("vgId:%d %s failed at line %d since %s", vid, __func__, lino, tstrerror(code)) + +typedef struct SFDataPtr { + int64_t offset; + int64_t size; +} SFDataPtr; + +extern int32_t tsdbOpenFile(const char *path, int32_t szPage, int32_t flag, STsdbFD **ppFD); +extern void tsdbCloseFile(STsdbFD **ppFD); +extern int32_t tsdbWriteFile(STsdbFD *pFD, int64_t offset, const uint8_t *pBuf, int64_t size); +extern int32_t tsdbReadFile(STsdbFD *pFD, int64_t offset, uint8_t *pBuf, int64_t size); +extern int32_t tsdbFsyncFile(STsdbFD *pFD); + +#ifdef __cplusplus +} +#endif + +#endif /*_TD_TSDB_DEF_H_*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbFS.c b/source/dnode/vnode/src/tsdb/tsdbFS.c index 41fdd05741..ec116c717e 100644 --- a/source/dnode/vnode/src/tsdb/tsdbFS.c +++ b/source/dnode/vnode/src/tsdb/tsdbFS.c @@ -181,10 +181,10 @@ static int32_t tsdbScanAndTryFixFS(STsdb *pTsdb) { TSDB_CHECK_CODE(code, lino, _exit); } - if (size != tsdbLogicToFileSize(pTsdb->fs.pDelFile->size, pTsdb->pVnode->config.tsdbPageSize)) { - code = TSDB_CODE_FILE_CORRUPTED; - TSDB_CHECK_CODE(code, lino, _exit); - } + // if (size != tsdbLogicToFileSize(pTsdb->fs.pDelFile->size, pTsdb->pVnode->config.tsdbPageSize)) { + // code = TSDB_CODE_FILE_CORRUPTED; + // TSDB_CHECK_CODE(code, lino, _exit); + // } } // SArray @@ -199,10 +199,10 @@ static int32_t tsdbScanAndTryFixFS(STsdb *pTsdb) { code = TAOS_SYSTEM_ERROR(errno); TSDB_CHECK_CODE(code, lino, _exit); } - if (size != tsdbLogicToFileSize(pSet->pHeadF->size, pTsdb->pVnode->config.tsdbPageSize)) { - code = TSDB_CODE_FILE_CORRUPTED; - TSDB_CHECK_CODE(code, lino, _exit); - } + // if (size != tsdbLogicToFileSize(pSet->pHeadF->size, pTsdb->pVnode->config.tsdbPageSize)) { + // code = TSDB_CODE_FILE_CORRUPTED; + // TSDB_CHECK_CODE(code, lino, _exit); + // } // data ========= tsdbDataFileName(pTsdb, pSet->diskId, pSet->fid, pSet->pDataF, fname); @@ -210,10 +210,10 @@ static int32_t tsdbScanAndTryFixFS(STsdb *pTsdb) { code = TAOS_SYSTEM_ERROR(errno); TSDB_CHECK_CODE(code, lino, _exit); } - if (size < tsdbLogicToFileSize(pSet->pDataF->size, pTsdb->pVnode->config.tsdbPageSize)) { - code = TSDB_CODE_FILE_CORRUPTED; - TSDB_CHECK_CODE(code, lino, _exit); - } + // if (size < tsdbLogicToFileSize(pSet->pDataF->size, pTsdb->pVnode->config.tsdbPageSize)) { + // code = TSDB_CODE_FILE_CORRUPTED; + // TSDB_CHECK_CODE(code, lino, _exit); + // } // else if (size > tsdbLogicToFileSize(pSet->pDataF->size, pTsdb->pVnode->config.tsdbPageSize)) { // code = tsdbDFileRollback(pTsdb, pSet, TSDB_DATA_FILE); // TSDB_CHECK_CODE(code, lino, _exit); @@ -225,10 +225,10 @@ static int32_t tsdbScanAndTryFixFS(STsdb *pTsdb) { code = TAOS_SYSTEM_ERROR(errno); TSDB_CHECK_CODE(code, lino, _exit); } - if (size < tsdbLogicToFileSize(pSet->pSmaF->size, pTsdb->pVnode->config.tsdbPageSize)) { - code = TSDB_CODE_FILE_CORRUPTED; - TSDB_CHECK_CODE(code, lino, _exit); - } + // if (size < tsdbLogicToFileSize(pSet->pSmaF->size, pTsdb->pVnode->config.tsdbPageSize)) { + // code = TSDB_CODE_FILE_CORRUPTED; + // TSDB_CHECK_CODE(code, lino, _exit); + // } // else if (size > tsdbLogicToFileSize(pSet->pSmaF->size, pTsdb->pVnode->config.tsdbPageSize)) { // code = tsdbDFileRollback(pTsdb, pSet, TSDB_SMA_FILE); // TSDB_CHECK_CODE(code, lino, _exit); @@ -241,10 +241,10 @@ static int32_t tsdbScanAndTryFixFS(STsdb *pTsdb) { code = TAOS_SYSTEM_ERROR(errno); TSDB_CHECK_CODE(code, lino, _exit); } - if (size != tsdbLogicToFileSize(pSet->aSttF[iStt]->size, pTsdb->pVnode->config.tsdbPageSize)) { - code = TSDB_CODE_FILE_CORRUPTED; - TSDB_CHECK_CODE(code, lino, _exit); - } + // if (size != tsdbLogicToFileSize(pSet->aSttF[iStt]->size, pTsdb->pVnode->config.tsdbPageSize)) { + // code = TSDB_CODE_FILE_CORRUPTED; + // TSDB_CHECK_CODE(code, lino, _exit); + // } } } @@ -270,7 +270,7 @@ int32_t tDFileSetCmprFn(const void *p1, const void *p2) { return 0; } -static void tsdbGetCurrentFName(STsdb *pTsdb, char *current, char *current_t) { +void tsdbGetCurrentFName(STsdb *pTsdb, char *current, char *current_t) { SVnode *pVnode = pTsdb->pVnode; int32_t offset = 0; @@ -289,7 +289,7 @@ static void tsdbGetCurrentFName(STsdb *pTsdb, char *current, char *current_t) { } } -static int32_t tsdbLoadFSFromFile(const char *fname, STsdbFS *pFS) { +static int32_t load_fs(const char *fname, STsdbFS *pFS) { int32_t code = 0; int32_t lino = 0; uint8_t *pData = NULL; @@ -666,7 +666,7 @@ static int32_t tsdbFSApplyChange(STsdb *pTsdb, STsdbFS *pFS) { taosArrayRemove(pTsdb->fs.aDFileSet, iOld); } else { code = tsdbNewFileSet(pTsdb, &fSet, pSetNew); - TSDB_CHECK_CODE(code, lino, _exit) + TSDB_CHECK_CODE(code, lino, _exit); if (taosArrayInsert(pTsdb->fs.aDFileSet, iOld, &fSet) == NULL) { code = TSDB_CODE_OUT_OF_MEMORY; @@ -682,7 +682,7 @@ static int32_t tsdbFSApplyChange(STsdb *pTsdb, STsdbFS *pFS) { taosArrayRemove(pTsdb->fs.aDFileSet, iOld); } else { code = tsdbNewFileSet(pTsdb, &fSet, pSetNew); - TSDB_CHECK_CODE(code, lino, _exit) + TSDB_CHECK_CODE(code, lino, _exit); if (taosArrayInsert(pTsdb->fs.aDFileSet, iOld, &fSet) == NULL) { code = TSDB_CODE_OUT_OF_MEMORY; @@ -723,7 +723,7 @@ int32_t tsdbFSCommit(STsdb *pTsdb) { code = tsdbFSCreate(&fs); TSDB_CHECK_CODE(code, lino, _exit); - code = tsdbLoadFSFromFile(current, &fs); + code = load_fs(current, &fs); TSDB_CHECK_CODE(code, lino, _exit); // apply file change @@ -768,7 +768,7 @@ int32_t tsdbFSOpen(STsdb *pTsdb, int8_t rollback) { tsdbGetCurrentFName(pTsdb, current, current_t); if (taosCheckExistFile(current)) { - code = tsdbLoadFSFromFile(current, &pTsdb->fs); + code = load_fs(current, &pTsdb->fs); TSDB_CHECK_CODE(code, lino, _exit); if (taosCheckExistFile(current_t)) { diff --git a/source/dnode/vnode/src/tsdb/tsdbFS2.c b/source/dnode/vnode/src/tsdb/tsdbFS2.c new file mode 100644 index 0000000000..1a0104945f --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbFS2.c @@ -0,0 +1,885 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbFS2.h" +#include "tsdbUpgrade.h" +#include "vnd.h" + +extern int vnodeScheduleTask(int (*execute)(void *), void *arg); +extern int vnodeScheduleTaskEx(int tpid, int (*execute)(void *), void *arg); + +#define TSDB_FS_EDIT_MIN TSDB_FEDIT_COMMIT +#define TSDB_FS_EDIT_MAX (TSDB_FEDIT_MERGE + 1) + +enum { + TSDB_FS_STATE_NONE = 0, + TSDB_FS_STATE_OPEN, + TSDB_FS_STATE_EDIT, + TSDB_FS_STATE_CLOSE, +}; + +static const char *gCurrentFname[] = { + [TSDB_FCURRENT] = "current.json", + [TSDB_FCURRENT_C] = "current.c.json", + [TSDB_FCURRENT_M] = "current.m.json", +}; + +static int32_t create_fs(STsdb *pTsdb, STFileSystem **fs) { + fs[0] = taosMemoryCalloc(1, sizeof(*fs[0])); + if (fs[0] == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + fs[0]->tsdb = pTsdb; + tsem_init(&fs[0]->canEdit, 0, 1); + fs[0]->state = TSDB_FS_STATE_NONE; + fs[0]->neid = 0; + TARRAY2_INIT(fs[0]->fSetArr); + TARRAY2_INIT(fs[0]->fSetArrTmp); + + // background task queue + taosThreadMutexInit(fs[0]->mutex, NULL); + fs[0]->bgTaskQueue->next = fs[0]->bgTaskQueue; + fs[0]->bgTaskQueue->prev = fs[0]->bgTaskQueue; + + return 0; +} + +static int32_t destroy_fs(STFileSystem **fs) { + if (fs[0] == NULL) return 0; + taosThreadMutexDestroy(fs[0]->mutex); + + ASSERT(fs[0]->bgTaskNum == 0); + + TARRAY2_DESTROY(fs[0]->fSetArr, NULL); + TARRAY2_DESTROY(fs[0]->fSetArrTmp, NULL); + tsem_destroy(&fs[0]->canEdit); + taosMemoryFree(fs[0]); + fs[0] = NULL; + return 0; +} + +int32_t current_fname(STsdb *pTsdb, char *fname, EFCurrentT ftype) { + int32_t offset = 0; + + vnodeGetPrimaryDir(pTsdb->path, pTsdb->pVnode->diskPrimary, pTsdb->pVnode->pTfs, fname, TSDB_FILENAME_LEN); + offset = strlen(fname); + snprintf(fname + offset, TSDB_FILENAME_LEN - offset - 1, "%s%s", TD_DIRSEP, gCurrentFname[ftype]); + + return 0; +} + +static int32_t save_json(const cJSON *json, const char *fname) { + int32_t code = 0; + + char *data = cJSON_PrintUnformatted(json); + if (data == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + TdFilePtr fp = taosOpenFile(fname, TD_FILE_WRITE | TD_FILE_CREATE | TD_FILE_TRUNC); + if (fp == NULL) { + code = TAOS_SYSTEM_ERROR(code); + goto _exit; + } + + if (taosWriteFile(fp, data, strlen(data)) < 0) { + code = TAOS_SYSTEM_ERROR(code); + goto _exit; + } + + if (taosFsyncFile(fp) < 0) { + code = TAOS_SYSTEM_ERROR(code); + goto _exit; + } + + taosCloseFile(&fp); + +_exit: + taosMemoryFree(data); + return code; +} + +static int32_t load_json(const char *fname, cJSON **json) { + int32_t code = 0; + char *data = NULL; + + TdFilePtr fp = taosOpenFile(fname, TD_FILE_READ); + if (fp == NULL) return TAOS_SYSTEM_ERROR(code); + + int64_t size; + if (taosFStatFile(fp, &size, NULL) < 0) { + code = TAOS_SYSTEM_ERROR(code); + goto _exit; + } + + data = taosMemoryMalloc(size + 1); + if (data == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + goto _exit; + } + + if (taosReadFile(fp, data, size) < 0) { + code = TAOS_SYSTEM_ERROR(code); + goto _exit; + } + data[size] = '\0'; + + json[0] = cJSON_Parse(data); + if (json[0] == NULL) { + code = TSDB_CODE_FILE_CORRUPTED; + goto _exit; + } + +_exit: + taosCloseFile(&fp); + if (data) taosMemoryFree(data); + if (code) json[0] = NULL; + return code; +} + +int32_t save_fs(const TFileSetArray *arr, const char *fname) { + int32_t code = 0; + int32_t lino = 0; + + cJSON *json = cJSON_CreateObject(); + if (!json) return TSDB_CODE_OUT_OF_MEMORY; + + // fmtv + if (cJSON_AddNumberToObject(json, "fmtv", 1) == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + TSDB_CHECK_CODE(code, lino, _exit); + } + + // fset + cJSON *ajson = cJSON_AddArrayToObject(json, "fset"); + if (!ajson) TSDB_CHECK_CODE(code = TSDB_CODE_OUT_OF_MEMORY, lino, _exit); + const STFileSet *fset; + TARRAY2_FOREACH(arr, fset) { + cJSON *item = cJSON_CreateObject(); + if (!item) TSDB_CHECK_CODE(code = TSDB_CODE_OUT_OF_MEMORY, lino, _exit); + cJSON_AddItemToArray(ajson, item); + + code = tsdbTFileSetToJson(fset, item); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = save_json(json, fname); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + tsdbError("%s failed at line %d since %s", __func__, lino, tstrerror(code)); + } + cJSON_Delete(json); + return code; +} + +static int32_t load_fs(STsdb *pTsdb, const char *fname, TFileSetArray *arr) { + int32_t code = 0; + int32_t lino = 0; + + TARRAY2_CLEAR(arr, tsdbTFileSetClear); + + // load json + cJSON *json = NULL; + code = load_json(fname, &json); + TSDB_CHECK_CODE(code, lino, _exit); + + // parse json + const cJSON *item1; + + /* fmtv */ + item1 = cJSON_GetObjectItem(json, "fmtv"); + if (cJSON_IsNumber(item1)) { + ASSERT(item1->valuedouble == 1); + } else { + TSDB_CHECK_CODE(code = TSDB_CODE_FILE_CORRUPTED, lino, _exit); + } + + /* fset */ + item1 = cJSON_GetObjectItem(json, "fset"); + if (cJSON_IsArray(item1)) { + const cJSON *item2; + cJSON_ArrayForEach(item2, item1) { + STFileSet *fset; + code = tsdbJsonToTFileSet(pTsdb, item2, &fset); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(arr, fset); + TSDB_CHECK_CODE(code, lino, _exit); + } + } else { + TSDB_CHECK_CODE(code = TSDB_CODE_FILE_CORRUPTED, lino, _exit); + } + +_exit: + if (code) { + tsdbError("%s failed at line %d since %s, fname:%s", __func__, lino, tstrerror(code), fname); + } + if (json) cJSON_Delete(json); + return code; +} + +static bool is_same_file(const STFile *f1, const STFile f2) { + if (f1->type != f2.type) return false; + if (f1->did.level != f2.did.level) return false; + if (f1->did.id != f2.did.id) return false; + if (f1->cid != f2.cid) return false; + return true; +} + +static int32_t apply_commit(STFileSystem *fs) { + int32_t code = 0; + TFileSetArray *fsetArray1 = fs->fSetArr; + TFileSetArray *fsetArray2 = fs->fSetArrTmp; + int32_t i1 = 0, i2 = 0; + + while (i1 < TARRAY2_SIZE(fsetArray1) || i2 < TARRAY2_SIZE(fsetArray2)) { + STFileSet *fset1 = i1 < TARRAY2_SIZE(fsetArray1) ? TARRAY2_GET(fsetArray1, i1) : NULL; + STFileSet *fset2 = i2 < TARRAY2_SIZE(fsetArray2) ? TARRAY2_GET(fsetArray2, i2) : NULL; + + if (fset1 && fset2) { + if (fset1->fid < fset2->fid) { + // delete fset1 + TARRAY2_REMOVE(fsetArray1, i1, tsdbTFileSetRemove); + } else if (fset1->fid > fset2->fid) { + // create new file set with fid of fset2->fid + code = tsdbTFileSetInitDup(fs->tsdb, fset2, &fset1); + if (code) return code; + code = TARRAY2_SORT_INSERT(fsetArray1, fset1, tsdbTFileSetCmprFn); + if (code) return code; + i1++; + i2++; + } else { + // edit + code = tsdbTFileSetApplyEdit(fs->tsdb, fset2, fset1); + if (code) return code; + i1++; + i2++; + } + } else if (fset1) { + // delete fset1 + TARRAY2_REMOVE(fsetArray1, i1, tsdbTFileSetRemove); + } else { + // create new file set with fid of fset2->fid + code = tsdbTFileSetInitDup(fs->tsdb, fset2, &fset1); + if (code) return code; + code = TARRAY2_SORT_INSERT(fsetArray1, fset1, tsdbTFileSetCmprFn); + if (code) return code; + i1++; + i2++; + } + } + + return 0; +} + +static int32_t commit_edit(STFileSystem *fs) { + char current[TSDB_FILENAME_LEN]; + char current_t[TSDB_FILENAME_LEN]; + + current_fname(fs->tsdb, current, TSDB_FCURRENT); + if (fs->etype == TSDB_FEDIT_COMMIT) { + current_fname(fs->tsdb, current_t, TSDB_FCURRENT_C); + } else if (fs->etype == TSDB_FEDIT_MERGE) { + current_fname(fs->tsdb, current_t, TSDB_FCURRENT_M); + } else { + ASSERT(0); + } + + int32_t code; + int32_t lino; + if ((code = taosRenameFile(current_t, current))) { + TSDB_CHECK_CODE(code = TAOS_SYSTEM_ERROR(code), lino, _exit); + } + + code = apply_commit(fs); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(fs->tsdb->pVnode), __func__, lino, tstrerror(code)); + } else { + tsdbInfo("vgId:%d %s success, etype:%d", TD_VID(fs->tsdb->pVnode), __func__, fs->etype); + } + return code; +} + +// static int32_t +static int32_t apply_abort(STFileSystem *fs) { + // TODO + return 0; +} + +static int32_t abort_edit(STFileSystem *fs) { + char fname[TSDB_FILENAME_LEN]; + + if (fs->etype == TSDB_FEDIT_COMMIT) { + current_fname(fs->tsdb, fname, TSDB_FCURRENT_C); + } else if (fs->etype == TSDB_FEDIT_MERGE) { + current_fname(fs->tsdb, fname, TSDB_FCURRENT_M); + } else { + ASSERT(0); + } + + int32_t code; + int32_t lino; + if ((code = taosRemoveFile(fname))) { + TSDB_CHECK_CODE(code = TAOS_SYSTEM_ERROR(code), lino, _exit); + } + + code = apply_abort(fs); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + tsdbError("vgId:%d %s failed since %s", TD_VID(fs->tsdb->pVnode), __func__, tstrerror(code)); + } else { + tsdbInfo("vgId:%d %s success, etype:%d", TD_VID(fs->tsdb->pVnode), __func__, fs->etype); + } + return code; +} + +static int32_t tsdbFSScanAndFix(STFileSystem *fs) { + fs->neid = 0; + + // get max commit id + const STFileSet *fset; + TARRAY2_FOREACH(fs->fSetArr, fset) { fs->neid = TMAX(fs->neid, tsdbTFileSetMaxCid(fset)); } + + // TODO + return 0; +} + +static int32_t tsdbFSDupState(STFileSystem *fs) { + int32_t code; + + const TFileSetArray *src = fs->fSetArr; + TFileSetArray *dst = fs->fSetArrTmp; + + TARRAY2_CLEAR(dst, tsdbTFileSetClear); + + const STFileSet *fset1; + TARRAY2_FOREACH(src, fset1) { + STFileSet *fset2; + code = tsdbTFileSetInitDup(fs->tsdb, fset1, &fset2); + if (code) return code; + code = TARRAY2_APPEND(dst, fset2); + if (code) return code; + } + + return 0; +} + +static int32_t open_fs(STFileSystem *fs, int8_t rollback) { + int32_t code = 0; + int32_t lino = 0; + STsdb *pTsdb = fs->tsdb; + + char fCurrent[TSDB_FILENAME_LEN]; + char cCurrent[TSDB_FILENAME_LEN]; + char mCurrent[TSDB_FILENAME_LEN]; + + current_fname(pTsdb, fCurrent, TSDB_FCURRENT); + current_fname(pTsdb, cCurrent, TSDB_FCURRENT_C); + current_fname(pTsdb, mCurrent, TSDB_FCURRENT_M); + + if (taosCheckExistFile(fCurrent)) { // current.json exists + code = load_fs(pTsdb, fCurrent, fs->fSetArr); + TSDB_CHECK_CODE(code, lino, _exit); + + if (taosCheckExistFile(cCurrent)) { + // current.c.json exists + + fs->etype = TSDB_FEDIT_COMMIT; + if (rollback) { + code = abort_edit(fs); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + code = load_fs(pTsdb, cCurrent, fs->fSetArrTmp); + TSDB_CHECK_CODE(code, lino, _exit); + + code = commit_edit(fs); + TSDB_CHECK_CODE(code, lino, _exit); + } + } else if (taosCheckExistFile(mCurrent)) { + // current.m.json exists + fs->etype = TSDB_FEDIT_MERGE; + code = abort_edit(fs); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbFSDupState(fs); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbFSScanAndFix(fs); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + code = save_fs(fs->fSetArr, fCurrent); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); + } else { + tsdbInfo("vgId:%d %s success", TD_VID(pTsdb->pVnode), __func__); + } + return 0; +} + +static int32_t close_file_system(STFileSystem *fs) { + TARRAY2_CLEAR(fs->fSetArr, tsdbTFileSetClear); + TARRAY2_CLEAR(fs->fSetArrTmp, tsdbTFileSetClear); + // TODO + return 0; +} + +static int32_t apply_edit(STFileSystem *pFS) { + int32_t code = 0; + ASSERTS(0, "TODO: Not implemented yet"); + return code; +} + +static int32_t fset_cmpr_fn(const struct STFileSet *pSet1, const struct STFileSet *pSet2) { + if (pSet1->fid < pSet2->fid) { + return -1; + } else if (pSet1->fid > pSet2->fid) { + return 1; + } + return 0; +} + +static int32_t edit_fs(STFileSystem *fs, const TFileOpArray *opArray) { + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFSDupState(fs); + if (code) return code; + + TFileSetArray *fsetArray = fs->fSetArrTmp; + STFileSet *fset = NULL; + const STFileOp *op; + TARRAY2_FOREACH_PTR(opArray, op) { + if (!fset || fset->fid != op->fid) { + STFileSet tfset = {.fid = op->fid}; + fset = &tfset; + STFileSet **fsetPtr = TARRAY2_SEARCH(fsetArray, &fset, tsdbTFileSetCmprFn, TD_EQ); + fset = (fsetPtr == NULL) ? NULL : *fsetPtr; + + if (!fset) { + code = tsdbTFileSetInit(op->fid, &fset); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_SORT_INSERT(fsetArray, fset, tsdbTFileSetCmprFn); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + + code = tsdbTFileSetEdit(fs->tsdb, fset, op); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // remove empty file set + int32_t i = 0; + while (i < TARRAY2_SIZE(fsetArray)) { + fset = TARRAY2_GET(fsetArray, i); + if (tsdbTFileSetIsEmpty(fset)) { + TARRAY2_REMOVE(fsetArray, i, tsdbTFileSetClear); + } else { + i++; + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(fs->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbOpenFS(STsdb *pTsdb, STFileSystem **fs, int8_t rollback) { + int32_t code; + int32_t lino; + + code = tsdbCheckAndUpgradeFileSystem(pTsdb, rollback); + TSDB_CHECK_CODE(code, lino, _exit); + + code = create_fs(pTsdb, fs); + TSDB_CHECK_CODE(code, lino, _exit); + + code = open_fs(fs[0], rollback); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); + destroy_fs(fs); + } else { + tsdbInfo("vgId:%d %s success", TD_VID(pTsdb->pVnode), __func__); + } + return 0; +} + +static void tsdbDoWaitBgTask(STFileSystem *fs, STFSBgTask *task) { + task->numWait++; + taosThreadCondWait(task->done, fs->mutex); + task->numWait--; + + if (task->numWait == 0) { + taosThreadCondDestroy(task->done); + taosMemoryFree(task); + } +} + +static void tsdbDoDoneBgTask(STFileSystem *fs, STFSBgTask *task) { + if (task->numWait > 0) { + taosThreadCondBroadcast(task->done); + } else { + taosThreadCondDestroy(task->done); + taosMemoryFree(task); + } +} + +int32_t tsdbCloseFS(STFileSystem **fs) { + if (fs[0] == NULL) return 0; + + taosThreadMutexLock(fs[0]->mutex); + fs[0]->stop = true; + + if (fs[0]->bgTaskRunning) { + tsdbDoWaitBgTask(fs[0], fs[0]->bgTaskRunning); + } + taosThreadMutexUnlock(fs[0]->mutex); + + close_file_system(fs[0]); + destroy_fs(fs); + return 0; +} + +int64_t tsdbFSAllocEid(STFileSystem *fs) { + taosThreadRwlockRdlock(&fs->tsdb->rwLock); + int64_t cid = ++fs->neid; + taosThreadRwlockUnlock(&fs->tsdb->rwLock); + return cid; +} + +int32_t tsdbFSEditBegin(STFileSystem *fs, const TFileOpArray *opArray, EFEditT etype) { + int32_t code = 0; + int32_t lino; + char current_t[TSDB_FILENAME_LEN]; + + switch (etype) { + case TSDB_FEDIT_COMMIT: + current_fname(fs->tsdb, current_t, TSDB_FCURRENT_C); + break; + case TSDB_FEDIT_MERGE: + current_fname(fs->tsdb, current_t, TSDB_FCURRENT_M); + break; + default: + ASSERT(0); + } + + tsem_wait(&fs->canEdit); + fs->etype = etype; + + // edit + code = edit_fs(fs, opArray); + TSDB_CHECK_CODE(code, lino, _exit); + + // save fs + code = save_fs(fs->fSetArrTmp, current_t); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + tsdbError("vgId:%d %s failed at line %d since %s, etype:%d", TD_VID(fs->tsdb->pVnode), __func__, lino, + tstrerror(code), etype); + } else { + tsdbInfo("vgId:%d %s done, etype:%d", TD_VID(fs->tsdb->pVnode), __func__, etype); + } + return code; +} + +int32_t tsdbFSEditCommit(STFileSystem *fs) { + int32_t code = 0; + int32_t lino = 0; + + // commit + code = commit_edit(fs); + TSDB_CHECK_CODE(code, lino, _exit); + + // schedule merge + if (fs->tsdb->pVnode->config.sttTrigger != 1) { + STFileSet *fset; + TARRAY2_FOREACH_REVERSE(fs->fSetArr, fset) { + if (TARRAY2_SIZE(fset->lvlArr) == 0) continue; + + SSttLvl *lvl = TARRAY2_FIRST(fset->lvlArr); + if (lvl->level != 0 || TARRAY2_SIZE(lvl->fobjArr) < fs->tsdb->pVnode->config.sttTrigger) continue; + + code = tsdbFSScheduleBgTask(fs, TSDB_BG_TASK_MERGER, tsdbMerge, fs->tsdb, NULL); + TSDB_CHECK_CODE(code, lino, _exit); + + break; + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(fs->tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d %s done, etype:%d", TD_VID(fs->tsdb->pVnode), __func__, fs->etype); + tsem_post(&fs->canEdit); + } + return code; +} + +int32_t tsdbFSEditAbort(STFileSystem *fs) { + int32_t code = abort_edit(fs); + tsem_post(&fs->canEdit); + return code; +} + +int32_t tsdbFSGetFSet(STFileSystem *fs, int32_t fid, STFileSet **fset) { + STFileSet tfset = {.fid = fid}; + STFileSet *pset = &tfset; + STFileSet **fsetPtr = TARRAY2_SEARCH(fs->fSetArr, &pset, tsdbTFileSetCmprFn, TD_EQ); + fset[0] = (fsetPtr == NULL) ? NULL : fsetPtr[0]; + return 0; +} + +int32_t tsdbFSCreateCopySnapshot(STFileSystem *fs, TFileSetArray **fsetArr) { + int32_t code = 0; + STFileSet *fset; + STFileSet *fset1; + + fsetArr[0] = taosMemoryMalloc(sizeof(TFileSetArray)); + if (fsetArr == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + TARRAY2_INIT(fsetArr[0]); + + taosThreadRwlockRdlock(&fs->tsdb->rwLock); + TARRAY2_FOREACH(fs->fSetArr, fset) { + code = tsdbTFileSetInitDup(fs->tsdb, fset, &fset1); + if (code) break; + + code = TARRAY2_APPEND(fsetArr[0], fset1); + if (code) break; + } + taosThreadRwlockUnlock(&fs->tsdb->rwLock); + + if (code) { + TARRAY2_DESTROY(fsetArr[0], tsdbTFileSetClear); + taosMemoryFree(fsetArr[0]); + fsetArr[0] = NULL; + } + return code; +} + +int32_t tsdbFSDestroyCopySnapshot(TFileSetArray **fsetArr) { + if (fsetArr[0]) { + TARRAY2_DESTROY(fsetArr[0], tsdbTFileSetClear); + taosMemoryFree(fsetArr[0]); + fsetArr[0] = NULL; + } + return 0; +} + +int32_t tsdbFSCreateRefSnapshot(STFileSystem *fs, TFileSetArray **fsetArr) { + int32_t code = 0; + STFileSet *fset, *fset1; + + fsetArr[0] = taosMemoryCalloc(1, sizeof(*fsetArr[0])); + if (fsetArr[0] == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + taosThreadRwlockRdlock(&fs->tsdb->rwLock); + TARRAY2_FOREACH(fs->fSetArr, fset) { + code = tsdbTFileSetInitRef(fs->tsdb, fset, &fset1); + if (code) break; + + code = TARRAY2_APPEND(fsetArr[0], fset1); + if (code) break; + } + taosThreadRwlockUnlock(&fs->tsdb->rwLock); + + if (code) { + TARRAY2_DESTROY(fsetArr[0], tsdbTFileSetClear); + fsetArr[0] = NULL; + } + return code; +} + +int32_t tsdbFSDestroyRefSnapshot(TFileSetArray **fsetArr) { + if (fsetArr[0]) { + TARRAY2_DESTROY(fsetArr[0], tsdbTFileSetClear); + taosMemoryFreeClear(fsetArr[0]); + fsetArr[0] = NULL; + } + return 0; +} + +const char *gFSBgTaskName[] = {NULL, "MERGE", "RETENTION", "COMPACT"}; + +static int32_t tsdbFSRunBgTask(void *arg) { + STFileSystem *fs = (STFileSystem *)arg; + + ASSERT(fs->bgTaskRunning != NULL); + + fs->bgTaskRunning->launchTime = taosGetTimestampMs(); + fs->bgTaskRunning->run(fs->bgTaskRunning->arg); + fs->bgTaskRunning->finishTime = taosGetTimestampMs(); + + tsdbDebug("vgId:%d bg task:%s task id:%" PRId64 " finished, schedule time:%" PRId64 " launch time:%" PRId64 + " finish time:%" PRId64, + TD_VID(fs->tsdb->pVnode), gFSBgTaskName[fs->bgTaskRunning->type], fs->bgTaskRunning->taskid, + fs->bgTaskRunning->scheduleTime, fs->bgTaskRunning->launchTime, fs->bgTaskRunning->finishTime); + + taosThreadMutexLock(fs->mutex); + + // free last + tsdbDoDoneBgTask(fs, fs->bgTaskRunning); + fs->bgTaskRunning = NULL; + + // schedule next + if (fs->bgTaskNum > 0) { + if (fs->stop) { + while (fs->bgTaskNum > 0) { + STFSBgTask *task = fs->bgTaskQueue->next; + task->prev->next = task->next; + task->next->prev = task->prev; + fs->bgTaskNum--; + tsdbDoDoneBgTask(fs, task); + } + } else { + // pop task from head + fs->bgTaskRunning = fs->bgTaskQueue->next; + fs->bgTaskRunning->prev->next = fs->bgTaskRunning->next; + fs->bgTaskRunning->next->prev = fs->bgTaskRunning->prev; + fs->bgTaskNum--; + vnodeScheduleTaskEx(1, tsdbFSRunBgTask, arg); + } + } + + taosThreadMutexUnlock(fs->mutex); + return 0; +} + +static int32_t tsdbFSScheduleBgTaskImpl(STFileSystem *fs, EFSBgTaskT type, int32_t (*run)(void *), void *arg, + int64_t *taskid) { + if (fs->stop) { + return 0; // TODO: use a better error code + } + + // check if same task is on + // if (fs->bgTaskRunning && fs->bgTaskRunning->type == type) { + // return 0; + // } + + for (STFSBgTask *task = fs->bgTaskQueue->next; task != fs->bgTaskQueue; task = task->next) { + if (task->type == type) { + return 0; + } + } + + // do schedule task + STFSBgTask *task = taosMemoryCalloc(1, sizeof(STFSBgTask)); + if (task == NULL) return TSDB_CODE_OUT_OF_MEMORY; + taosThreadCondInit(task->done, NULL); + + task->type = type; + task->run = run; + task->arg = arg; + task->scheduleTime = taosGetTimestampMs(); + task->taskid = ++fs->taskid; + + if (fs->bgTaskRunning == NULL && fs->bgTaskNum == 0) { + // launch task directly + fs->bgTaskRunning = task; + vnodeScheduleTaskEx(1, tsdbFSRunBgTask, fs); + } else { + // add to the queue tail + fs->bgTaskNum++; + task->next = fs->bgTaskQueue; + task->prev = fs->bgTaskQueue->prev; + task->prev->next = task; + task->next->prev = task; + } + + if (taskid) *taskid = task->taskid; + return 0; +} + +int32_t tsdbFSScheduleBgTask(STFileSystem *fs, EFSBgTaskT type, int32_t (*run)(void *), void *arg, int64_t *taskid) { + taosThreadMutexLock(fs->mutex); + int32_t code = tsdbFSScheduleBgTaskImpl(fs, type, run, arg, taskid); + taosThreadMutexUnlock(fs->mutex); + return code; +} + +int32_t tsdbFSWaitBgTask(STFileSystem *fs, int64_t taskid) { + STFSBgTask *task = NULL; + + taosThreadMutexLock(fs->mutex); + + if (fs->bgTaskRunning && fs->bgTaskRunning->taskid == taskid) { + task = fs->bgTaskRunning; + } else { + for (STFSBgTask *taskt = fs->bgTaskQueue->next; taskt != fs->bgTaskQueue; taskt = taskt->next) { + if (taskt->taskid == taskid) { + task = taskt; + break; + } + } + } + + if (task) { + tsdbDoWaitBgTask(fs, task); + } + + taosThreadMutexUnlock(fs->mutex); + return 0; +} + +int32_t tsdbFSWaitAllBgTask(STFileSystem *fs) { + taosThreadMutexLock(fs->mutex); + + while (fs->bgTaskRunning) { + taosThreadCondWait(fs->bgTaskRunning->done, fs->mutex); + } + + taosThreadMutexUnlock(fs->mutex); + return 0; +} + +static int32_t tsdbFSDoDisableBgTask(STFileSystem *fs) { + fs->stop = true; + + if (fs->bgTaskRunning) { + tsdbDoWaitBgTask(fs, fs->bgTaskRunning); + } + return 0; +} + +int32_t tsdbFSDisableBgTask(STFileSystem *fs) { + taosThreadMutexLock(fs->mutex); + int32_t code = tsdbFSDoDisableBgTask(fs); + taosThreadMutexUnlock(fs->mutex); + return code; +} + +int32_t tsdbFSEnableBgTask(STFileSystem *fs) { + taosThreadMutexLock(fs->mutex); + fs->stop = false; + taosThreadMutexUnlock(fs->mutex); + return 0; +} diff --git a/source/dnode/vnode/src/tsdb/tsdbFS2.h b/source/dnode/vnode/src/tsdb/tsdbFS2.h new file mode 100644 index 0000000000..8270581e58 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbFS2.h @@ -0,0 +1,110 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbFSet2.h" + +#ifndef _TSDB_FILE_SYSTEM_H +#define _TSDB_FILE_SYSTEM_H + +#ifdef __cplusplus +extern "C" { +#endif + +/* Exposed Handle */ +typedef struct STFileSystem STFileSystem; +typedef struct STFSBgTask STFSBgTask; +// typedef TARRAY2(STFileSet *) TFileSetArray; + +typedef enum { + TSDB_FEDIT_COMMIT = 1, // + TSDB_FEDIT_MERGE +} EFEditT; + +typedef enum { + TSDB_BG_TASK_MERGER = 1, + TSDB_BG_TASK_RETENTION, + TSDB_BG_TASK_COMPACT, +} EFSBgTaskT; + +typedef enum { + TSDB_FCURRENT = 1, + TSDB_FCURRENT_C, // for commit + TSDB_FCURRENT_M, // for merge +} EFCurrentT; + +/* Exposed APIs */ +// open/close +int32_t tsdbOpenFS(STsdb *pTsdb, STFileSystem **fs, int8_t rollback); +int32_t tsdbCloseFS(STFileSystem **fs); +// snapshot +int32_t tsdbFSCreateCopySnapshot(STFileSystem *fs, TFileSetArray **fsetArr); +int32_t tsdbFSDestroyCopySnapshot(TFileSetArray **fsetArr); +int32_t tsdbFSCreateRefSnapshot(STFileSystem *fs, TFileSetArray **fsetArr); +int32_t tsdbFSDestroyRefSnapshot(TFileSetArray **fsetArr); +// txn +int64_t tsdbFSAllocEid(STFileSystem *fs); +int32_t tsdbFSEditBegin(STFileSystem *fs, const TFileOpArray *opArray, EFEditT etype); +int32_t tsdbFSEditCommit(STFileSystem *fs); +int32_t tsdbFSEditAbort(STFileSystem *fs); +// background task +int32_t tsdbFSScheduleBgTask(STFileSystem *fs, EFSBgTaskT type, int32_t (*run)(void *), void *arg, int64_t *taskid); +int32_t tsdbFSWaitBgTask(STFileSystem *fs, int64_t taskid); +int32_t tsdbFSWaitAllBgTask(STFileSystem *fs); +int32_t tsdbFSDisableBgTask(STFileSystem *fs); +int32_t tsdbFSEnableBgTask(STFileSystem *fs); +// other +int32_t tsdbFSGetFSet(STFileSystem *fs, int32_t fid, STFileSet **fset); + +struct STFSBgTask { + EFSBgTaskT type; + int32_t (*run)(void *arg); + void *arg; + + TdThreadCond done[1]; + int32_t numWait; + + int64_t taskid; + int64_t scheduleTime; + int64_t launchTime; + int64_t finishTime; + + struct STFSBgTask *prev; + struct STFSBgTask *next; +}; + +/* Exposed Structs */ +struct STFileSystem { + STsdb *tsdb; + tsem_t canEdit; + int32_t state; + int64_t neid; + EFEditT etype; + TFileSetArray fSetArr[1]; + TFileSetArray fSetArrTmp[1]; + + // background task queue + TdThreadMutex mutex[1]; + bool stop; + int64_t taskid; + int32_t bgTaskNum; + STFSBgTask bgTaskQueue[1]; + STFSBgTask *bgTaskRunning; +}; + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_FILE_SYSTEM_H*/ diff --git a/source/dnode/vnode/src/tsdb/tsdbFSet2.c b/source/dnode/vnode/src/tsdb/tsdbFSet2.c new file mode 100644 index 0000000000..7bc9743ecb --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbFSet2.c @@ -0,0 +1,542 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbFSet2.h" + +int32_t tsdbSttLvlInit(int32_t level, SSttLvl **lvl) { + if (!(lvl[0] = taosMemoryMalloc(sizeof(SSttLvl)))) return TSDB_CODE_OUT_OF_MEMORY; + lvl[0]->level = level; + TARRAY2_INIT(lvl[0]->fobjArr); + return 0; +} + +static void tsdbSttLvlClearFObj(void *data) { tsdbTFileObjUnref(*(STFileObj **)data); } + +int32_t tsdbSttLvlClear(SSttLvl **lvl) { + if (lvl[0] != NULL) { + TARRAY2_DESTROY(lvl[0]->fobjArr, tsdbSttLvlClearFObj); + taosMemoryFree(lvl[0]); + lvl[0] = NULL; + } + return 0; +} + +static int32_t tsdbSttLvlInitEx(STsdb *pTsdb, const SSttLvl *lvl1, SSttLvl **lvl) { + int32_t code = tsdbSttLvlInit(lvl1->level, lvl); + if (code) return code; + + const STFileObj *fobj1; + TARRAY2_FOREACH(lvl1->fobjArr, fobj1) { + STFileObj *fobj; + code = tsdbTFileObjInit(pTsdb, fobj1->f, &fobj); + if (code) { + tsdbSttLvlClear(lvl); + return code; + } + + TARRAY2_APPEND(lvl[0]->fobjArr, fobj); + } + return 0; +} + +static int32_t tsdbSttLvlInitRef(STsdb *pTsdb, const SSttLvl *lvl1, SSttLvl **lvl) { + int32_t code = tsdbSttLvlInit(lvl1->level, lvl); + if (code) return code; + + STFileObj *fobj1; + TARRAY2_FOREACH(lvl1->fobjArr, fobj1) { + tsdbTFileObjRef(fobj1); + code = TARRAY2_APPEND(lvl[0]->fobjArr, fobj1); + if (code) return code; + } + return 0; +} + +static void tsdbSttLvlRemoveFObj(void *data) { tsdbTFileObjRemove(*(STFileObj **)data); } +static void tsdbSttLvlRemove(SSttLvl **lvl) { + TARRAY2_DESTROY(lvl[0]->fobjArr, tsdbSttLvlRemoveFObj); + taosMemoryFree(lvl[0]); + lvl[0] = NULL; +} + +static int32_t tsdbSttLvlApplyEdit(STsdb *pTsdb, const SSttLvl *lvl1, SSttLvl *lvl2) { + int32_t code = 0; + + ASSERT(lvl1->level == lvl2->level); + + int32_t i1 = 0, i2 = 0; + while (i1 < TARRAY2_SIZE(lvl1->fobjArr) || i2 < TARRAY2_SIZE(lvl2->fobjArr)) { + STFileObj *fobj1 = i1 < TARRAY2_SIZE(lvl1->fobjArr) ? TARRAY2_GET(lvl1->fobjArr, i1) : NULL; + STFileObj *fobj2 = i2 < TARRAY2_SIZE(lvl2->fobjArr) ? TARRAY2_GET(lvl2->fobjArr, i2) : NULL; + + if (fobj1 && fobj2) { + if (fobj1->f->cid < fobj2->f->cid) { + // create a file obj + code = tsdbTFileObjInit(pTsdb, fobj1->f, &fobj2); + if (code) return code; + code = TARRAY2_APPEND(lvl2->fobjArr, fobj2); + if (code) return code; + i1++; + i2++; + } else if (fobj1->f->cid > fobj2->f->cid) { + // remove a file obj + TARRAY2_REMOVE(lvl2->fobjArr, i2, tsdbSttLvlRemoveFObj); + } else { + if (tsdbIsSameTFile(fobj1->f, fobj2->f)) { + if (tsdbIsTFileChanged(fobj1->f, fobj2->f)) { + fobj2->f[0] = fobj1->f[0]; + } + } else { + TARRAY2_REMOVE(lvl2->fobjArr, i2, tsdbSttLvlRemoveFObj); + code = tsdbTFileObjInit(pTsdb, fobj1->f, &fobj2); + if (code) return code; + code = TARRAY2_SORT_INSERT(lvl2->fobjArr, fobj2, tsdbTFileObjCmpr); + if (code) return code; + } + i1++; + i2++; + } + } else if (fobj1) { + // create a file obj + code = tsdbTFileObjInit(pTsdb, fobj1->f, &fobj2); + if (code) return code; + code = TARRAY2_APPEND(lvl2->fobjArr, fobj2); + if (code) return code; + i1++; + i2++; + } else { + // remove a file obj + TARRAY2_REMOVE(lvl2->fobjArr, i2, tsdbSttLvlRemoveFObj); + } + } + return 0; +} + +static int32_t tsdbSttLvlCmprFn(const SSttLvl **lvl1, const SSttLvl **lvl2) { + if (lvl1[0]->level < lvl2[0]->level) return -1; + if (lvl1[0]->level > lvl2[0]->level) return 1; + return 0; +} + +static int32_t tsdbSttLvlToJson(const SSttLvl *lvl, cJSON *json) { + if (cJSON_AddNumberToObject(json, "level", lvl->level) == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + cJSON *ajson = cJSON_AddArrayToObject(json, "files"); + if (ajson == NULL) return TSDB_CODE_OUT_OF_MEMORY; + const STFileObj *fobj; + TARRAY2_FOREACH(lvl->fobjArr, fobj) { + cJSON *item = cJSON_CreateObject(); + if (item == NULL) return TSDB_CODE_OUT_OF_MEMORY; + cJSON_AddItemToArray(ajson, item); + + int32_t code = tsdbTFileToJson(fobj->f, item); + if (code) return code; + } + + return 0; +} + +static int32_t tsdbJsonToSttLvl(STsdb *pTsdb, const cJSON *json, SSttLvl **lvl) { + const cJSON *item1, *item2; + int32_t level; + + item1 = cJSON_GetObjectItem(json, "level"); + if (cJSON_IsNumber(item1)) { + level = item1->valuedouble; + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + int32_t code = tsdbSttLvlInit(level, lvl); + if (code) return code; + + item1 = cJSON_GetObjectItem(json, "files"); + if (!cJSON_IsArray(item1)) { + tsdbSttLvlClear(lvl); + return TSDB_CODE_FILE_CORRUPTED; + } + + cJSON_ArrayForEach(item2, item1) { + STFile tf; + code = tsdbJsonToTFile(item2, TSDB_FTYPE_STT, &tf); + if (code) { + tsdbSttLvlClear(lvl); + return code; + } + + STFileObj *fobj; + code = tsdbTFileObjInit(pTsdb, &tf, &fobj); + if (code) { + tsdbSttLvlClear(lvl); + return code; + } + + TARRAY2_APPEND(lvl[0]->fobjArr, fobj); + } + return 0; +} + +int32_t tsdbTFileSetToJson(const STFileSet *fset, cJSON *json) { + int32_t code = 0; + cJSON *item1, *item2; + + // fid + if (cJSON_AddNumberToObject(json, "fid", fset->fid) == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + for (int32_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (fset->farr[ftype] == NULL) continue; + + code = tsdbTFileToJson(fset->farr[ftype]->f, json); + if (code) return code; + } + + // each level + item1 = cJSON_AddArrayToObject(json, "stt lvl"); + if (item1 == NULL) return TSDB_CODE_OUT_OF_MEMORY; + const SSttLvl *lvl; + TARRAY2_FOREACH(fset->lvlArr, lvl) { + item2 = cJSON_CreateObject(); + if (!item2) return TSDB_CODE_OUT_OF_MEMORY; + cJSON_AddItemToArray(item1, item2); + + code = tsdbSttLvlToJson(lvl, item2); + if (code) return code; + } + + return 0; +} + +int32_t tsdbJsonToTFileSet(STsdb *pTsdb, const cJSON *json, STFileSet **fset) { + int32_t code; + const cJSON *item1, *item2; + int32_t fid; + STFile tf; + + // fid + item1 = cJSON_GetObjectItem(json, "fid"); + if (cJSON_IsNumber(item1)) { + fid = item1->valuedouble; + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + code = tsdbTFileSetInit(fid, fset); + if (code) return code; + + for (tsdb_ftype_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + code = tsdbJsonToTFile(json, ftype, &tf); + if (code == TSDB_CODE_NOT_FOUND) { + continue; + } else if (code) { + tsdbTFileSetClear(fset); + return code; + } else { + code = tsdbTFileObjInit(pTsdb, &tf, &(*fset)->farr[ftype]); + if (code) return code; + } + } + + // each level + item1 = cJSON_GetObjectItem(json, "stt lvl"); + if (cJSON_IsArray(item1)) { + cJSON_ArrayForEach(item2, item1) { + SSttLvl *lvl; + code = tsdbJsonToSttLvl(pTsdb, item2, &lvl); + if (code) { + tsdbTFileSetClear(fset); + return code; + } + + TARRAY2_APPEND((*fset)->lvlArr, lvl); + } + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + return 0; +} + +// NOTE: the api does not remove file, only do memory operation +int32_t tsdbTFileSetEdit(STsdb *pTsdb, STFileSet *fset, const STFileOp *op) { + int32_t code = 0; + + if (op->optype == TSDB_FOP_CREATE) { + // create a new file + STFileObj *fobj; + code = tsdbTFileObjInit(pTsdb, &op->nf, &fobj); + if (code) return code; + + if (fobj->f->type == TSDB_FTYPE_STT) { + SSttLvl *lvl = tsdbTFileSetGetSttLvl(fset, fobj->f->stt->level); + if (!lvl) { + code = tsdbSttLvlInit(fobj->f->stt->level, &lvl); + if (code) return code; + + code = TARRAY2_SORT_INSERT(fset->lvlArr, lvl, tsdbSttLvlCmprFn); + if (code) return code; + } + + code = TARRAY2_SORT_INSERT(lvl->fobjArr, fobj, tsdbTFileObjCmpr); + if (code) return code; + } else { + ASSERT(fset->farr[fobj->f->type] == NULL); + fset->farr[fobj->f->type] = fobj; + } + } else if (op->optype == TSDB_FOP_REMOVE) { + // delete a file + if (op->of.type == TSDB_FTYPE_STT) { + SSttLvl *lvl = tsdbTFileSetGetSttLvl(fset, op->of.stt->level); + ASSERT(lvl); + + STFileObj tfobj = {.f[0] = {.cid = op->of.cid}}; + STFileObj *tfobjp = &tfobj; + int32_t idx = TARRAY2_SEARCH_IDX(lvl->fobjArr, &tfobjp, tsdbTFileObjCmpr, TD_EQ); + ASSERT(idx >= 0); + TARRAY2_REMOVE(lvl->fobjArr, idx, tsdbSttLvlClearFObj); + + if (TARRAY2_SIZE(lvl->fobjArr) == 0) { + // TODO: remove the stt level if no file exists anymore + // TARRAY2_REMOVE(&fset->lvlArr, lvl - fset->lvlArr.data, tsdbSttLvlClear); + } + } else { + ASSERT(tsdbIsSameTFile(&op->of, fset->farr[op->of.type]->f)); + tsdbTFileObjUnref(fset->farr[op->of.type]); + fset->farr[op->of.type] = NULL; + } + } else { + if (op->nf.type == TSDB_FTYPE_STT) { + SSttLvl *lvl = tsdbTFileSetGetSttLvl(fset, op->of.stt->level); + ASSERT(lvl); + + STFileObj tfobj = {.f[0] = {.cid = op->of.cid}}, *tfobjp = &tfobj; + STFileObj **fobjPtr = TARRAY2_SEARCH(lvl->fobjArr, &tfobjp, tsdbTFileObjCmpr, TD_EQ); + tfobjp = (fobjPtr ? *fobjPtr : NULL); + + ASSERT(tfobjp); + + tfobjp->f[0] = op->nf; + } else { + fset->farr[op->nf.type]->f[0] = op->nf; + } + } + + return 0; +} + +int32_t tsdbTFileSetApplyEdit(STsdb *pTsdb, const STFileSet *fset1, STFileSet *fset2) { + int32_t code = 0; + + ASSERT(fset1->fid == fset2->fid); + + for (tsdb_ftype_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (!fset1->farr[ftype] && !fset2->farr[ftype]) continue; + + STFileObj *fobj1 = fset1->farr[ftype]; + STFileObj *fobj2 = fset2->farr[ftype]; + + if (fobj1 && fobj2) { + if (tsdbIsSameTFile(fobj1->f, fobj2->f)) { + if (tsdbIsTFileChanged(fobj1->f, fobj2->f)) { + fobj2->f[0] = fobj1->f[0]; + } + } else { + tsdbTFileObjRemove(fobj2); + code = tsdbTFileObjInit(pTsdb, fobj1->f, &fset2->farr[ftype]); + if (code) return code; + } + } else if (fobj1) { + // create a new file + code = tsdbTFileObjInit(pTsdb, fobj1->f, &fset2->farr[ftype]); + if (code) return code; + } else { + // remove the file + tsdbTFileObjRemove(fobj2); + fset2->farr[ftype] = NULL; + } + } + + // stt part + int32_t i1 = 0, i2 = 0; + while (i1 < TARRAY2_SIZE(fset1->lvlArr) || i2 < TARRAY2_SIZE(fset2->lvlArr)) { + SSttLvl *lvl1 = i1 < TARRAY2_SIZE(fset1->lvlArr) ? TARRAY2_GET(fset1->lvlArr, i1) : NULL; + SSttLvl *lvl2 = i2 < TARRAY2_SIZE(fset2->lvlArr) ? TARRAY2_GET(fset2->lvlArr, i2) : NULL; + + if (lvl1 && lvl2) { + if (lvl1->level < lvl2->level) { + // add a new stt level + code = tsdbSttLvlInitEx(pTsdb, lvl1, &lvl2); + if (code) return code; + code = TARRAY2_SORT_INSERT(fset2->lvlArr, lvl2, tsdbSttLvlCmprFn); + if (code) return code; + i1++; + i2++; + } else if (lvl1->level > lvl2->level) { + // remove the stt level + TARRAY2_REMOVE(fset2->lvlArr, i2, tsdbSttLvlRemove); + } else { + // apply edit on stt level + code = tsdbSttLvlApplyEdit(pTsdb, lvl1, lvl2); + if (code) return code; + i1++; + i2++; + } + } else if (lvl1) { + // add a new stt level + code = tsdbSttLvlInitEx(pTsdb, lvl1, &lvl2); + if (code) return code; + code = TARRAY2_SORT_INSERT(fset2->lvlArr, lvl2, tsdbSttLvlCmprFn); + if (code) return code; + i1++; + i2++; + } else { + // remove the stt level + TARRAY2_REMOVE(fset2->lvlArr, i2, tsdbSttLvlRemove); + } + } + + return 0; +} + +int32_t tsdbTFileSetInit(int32_t fid, STFileSet **fset) { + fset[0] = taosMemoryCalloc(1, sizeof(STFileSet)); + if (fset[0] == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + fset[0]->fid = fid; + TARRAY2_INIT(fset[0]->lvlArr); + return 0; +} + +int32_t tsdbTFileSetInitDup(STsdb *pTsdb, const STFileSet *fset1, STFileSet **fset) { + int32_t code = tsdbTFileSetInit(fset1->fid, fset); + if (code) return code; + + for (int32_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (fset1->farr[ftype] == NULL) continue; + + code = tsdbTFileObjInit(pTsdb, fset1->farr[ftype]->f, &fset[0]->farr[ftype]); + if (code) { + tsdbTFileSetClear(fset); + return code; + } + } + + const SSttLvl *lvl1; + TARRAY2_FOREACH(fset1->lvlArr, lvl1) { + SSttLvl *lvl; + code = tsdbSttLvlInitEx(pTsdb, lvl1, &lvl); + if (code) { + tsdbTFileSetClear(fset); + return code; + } + + code = TARRAY2_APPEND(fset[0]->lvlArr, lvl); + if (code) return code; + } + + return 0; +} + +int32_t tsdbTFileSetInitRef(STsdb *pTsdb, const STFileSet *fset1, STFileSet **fset) { + int32_t code = tsdbTFileSetInit(fset1->fid, fset); + if (code) return code; + + for (int32_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (fset1->farr[ftype] == NULL) continue; + + tsdbTFileObjRef(fset1->farr[ftype]); + fset[0]->farr[ftype] = fset1->farr[ftype]; + } + + const SSttLvl *lvl1; + TARRAY2_FOREACH(fset1->lvlArr, lvl1) { + SSttLvl *lvl; + code = tsdbSttLvlInitRef(pTsdb, lvl1, &lvl); + if (code) { + tsdbTFileSetClear(fset); + return code; + } + + code = TARRAY2_APPEND(fset[0]->lvlArr, lvl); + if (code) return code; + } + + return 0; +} + +int32_t tsdbTFileSetClear(STFileSet **fset) { + if (!fset[0]) return 0; + + for (tsdb_ftype_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (fset[0]->farr[ftype] == NULL) continue; + tsdbTFileObjUnref(fset[0]->farr[ftype]); + } + + TARRAY2_DESTROY(fset[0]->lvlArr, tsdbSttLvlClear); + + taosMemoryFree(fset[0]); + fset[0] = NULL; + + return 0; +} + +int32_t tsdbTFileSetRemove(STFileSet **fset) { + for (tsdb_ftype_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (fset[0]->farr[ftype] == NULL) continue; + tsdbTFileObjRemove(fset[0]->farr[ftype]); + } + + TARRAY2_DESTROY(fset[0]->lvlArr, tsdbSttLvlRemove); + taosMemoryFree(fset[0]); + fset[0] = NULL; + return 0; +} + +SSttLvl *tsdbTFileSetGetSttLvl(STFileSet *fset, int32_t level) { + SSttLvl sttLvl = {.level = level}; + SSttLvl *lvl = &sttLvl; + SSttLvl **lvlPtr = TARRAY2_SEARCH(fset->lvlArr, &lvl, tsdbSttLvlCmprFn, TD_EQ); + return lvlPtr ? lvlPtr[0] : NULL; +} + +int32_t tsdbTFileSetCmprFn(const STFileSet **fset1, const STFileSet **fset2) { + if (fset1[0]->fid < fset2[0]->fid) return -1; + if (fset1[0]->fid > fset2[0]->fid) return 1; + return 0; +} + +int64_t tsdbTFileSetMaxCid(const STFileSet *fset) { + int64_t maxCid = 0; + for (tsdb_ftype_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (fset->farr[ftype] == NULL) continue; + maxCid = TMAX(maxCid, fset->farr[ftype]->f->cid); + } + const SSttLvl *lvl; + const STFileObj *fobj; + TARRAY2_FOREACH(fset->lvlArr, lvl) { + TARRAY2_FOREACH(lvl->fobjArr, fobj) { maxCid = TMAX(maxCid, fobj->f->cid); } + } + return maxCid; +} + +bool tsdbTFileSetIsEmpty(const STFileSet *fset) { + for (tsdb_ftype_t ftype = TSDB_FTYPE_MIN; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (fset->farr[ftype] != NULL) return false; + } + return TARRAY2_SIZE(fset->lvlArr) == 0; +} \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbFSet2.h b/source/dnode/vnode/src/tsdb/tsdbFSet2.h new file mode 100644 index 0000000000..d7b3c1fc8c --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbFSet2.h @@ -0,0 +1,85 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbFile2.h" + +#ifndef _TSDB_FILE_SET2_H +#define _TSDB_FILE_SET2_H + +#ifdef __cplusplus +extern "C" { +#endif + +typedef struct STFileSet STFileSet; +typedef struct STFileOp STFileOp; +typedef struct SSttLvl SSttLvl; +typedef TARRAY2(STFileObj *) TFileObjArray; +typedef TARRAY2(SSttLvl *) TSttLvlArray; +typedef TARRAY2(STFileOp) TFileOpArray; + +typedef enum { + TSDB_FOP_NONE = 0, + TSDB_FOP_CREATE, + TSDB_FOP_REMOVE, + TSDB_FOP_MODIFY, +} tsdb_fop_t; + +#define TFILE_SET(fid_) \ + (STFileSet) { .fid = (fid_) } + +// init/clear +int32_t tsdbTFileSetInit(int32_t fid, STFileSet **fset); +int32_t tsdbTFileSetInitDup(STsdb *pTsdb, const STFileSet *fset1, STFileSet **fset); +int32_t tsdbTFileSetInitRef(STsdb *pTsdb, const STFileSet *fset1, STFileSet **fset); +int32_t tsdbTFileSetClear(STFileSet **fset); +int32_t tsdbTFileSetRemove(STFileSet **fset); +// to/from json +int32_t tsdbTFileSetToJson(const STFileSet *fset, cJSON *json); +int32_t tsdbJsonToTFileSet(STsdb *pTsdb, const cJSON *json, STFileSet **fset); +// cmpr +int32_t tsdbTFileSetCmprFn(const STFileSet **fset1, const STFileSet **fset2); +// edit +int32_t tsdbTFileSetEdit(STsdb *pTsdb, STFileSet *fset, const STFileOp *op); +int32_t tsdbTFileSetApplyEdit(STsdb *pTsdb, const STFileSet *fset1, STFileSet *fset); +// max commit id +int64_t tsdbTFileSetMaxCid(const STFileSet *fset); +// get +SSttLvl *tsdbTFileSetGetSttLvl(STFileSet *fset, int32_t level); +// is empty +bool tsdbTFileSetIsEmpty(const STFileSet *fset); + +struct STFileOp { + tsdb_fop_t optype; + int32_t fid; + STFile of; // old file state + STFile nf; // new file state +}; + +struct SSttLvl { + int32_t level; + TFileObjArray fobjArr[1]; +}; + +struct STFileSet { + int32_t fid; + STFileObj *farr[TSDB_FTYPE_MAX]; // file array + TSttLvlArray lvlArr[1]; // level array +}; + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_FILE_SET2_H*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbFSetRW.c b/source/dnode/vnode/src/tsdb/tsdbFSetRW.c new file mode 100644 index 0000000000..83ae8c2429 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbFSetRW.c @@ -0,0 +1,295 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbFSetRW.h" + +// SFSetWriter ================================================== +struct SFSetWriter { + SFSetWriterConfig config[1]; + + SSkmInfo skmTb[1]; + SSkmInfo skmRow[1]; + uint8_t *bufArr[10]; + + struct { + TABLEID tbid[1]; + } ctx[1]; + + // writer + SBlockData blockData[2]; + int32_t blockDataIdx; + SDataFileWriter *dataWriter; + SSttFileWriter *sttWriter; +}; + +static int32_t tsdbFSetWriteTableDataBegin(SFSetWriter *writer, const TABLEID *tbid) { + int32_t code = 0; + int32_t lino = 0; + + writer->ctx->tbid->suid = tbid->suid; + writer->ctx->tbid->uid = tbid->uid; + + code = tsdbUpdateSkmTb(writer->config->tsdb, writer->ctx->tbid, writer->skmTb); + TSDB_CHECK_CODE(code, lino, _exit); + + writer->blockDataIdx = 0; + for (int32_t i = 0; i < ARRAY_SIZE(writer->blockData); i++) { + code = tBlockDataInit(&writer->blockData[i], writer->ctx->tbid, writer->skmTb->pTSchema, NULL, 0); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbFSetWriteTableDataEnd(SFSetWriter *writer) { + if (writer->ctx->tbid->uid == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + int32_t cidx = writer->blockDataIdx; + int32_t pidx = ((cidx + 1) & 1); + int32_t numRow = ((writer->blockData[pidx].nRow + writer->blockData[cidx].nRow) >> 1); + + if (writer->blockData[pidx].nRow > 0 && numRow >= writer->config->minRow) { + ASSERT(writer->blockData[pidx].nRow == writer->config->maxRow); + + SRowInfo row = { + .suid = writer->ctx->tbid->suid, + .uid = writer->ctx->tbid->uid, + .row = tsdbRowFromBlockData(writer->blockData + pidx, 0), + }; + + for (int32_t i = 0; i < numRow; i++) { + row.row.iRow = i; + + code = tsdbDataFileWriteRow(writer->dataWriter, &row); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbDataFileFlush(writer->dataWriter); + TSDB_CHECK_CODE(code, lino, _exit); + + for (int32_t i = numRow; i < writer->blockData[pidx].nRow; i++) { + row.row.iRow = i; + code = tsdbDataFileWriteRow(writer->dataWriter, &row); + TSDB_CHECK_CODE(code, lino, _exit); + } + + row.row = tsdbRowFromBlockData(writer->blockData + cidx, 0); + for (int32_t i = 0; i < writer->blockData[cidx].nRow; i++) { + row.row.iRow = i; + code = tsdbDataFileWriteRow(writer->dataWriter, &row); + TSDB_CHECK_CODE(code, lino, _exit); + } + } else { + // pidx + if (writer->blockData[pidx].nRow > 0) { + code = tsdbDataFileWriteBlockData(writer->dataWriter, &writer->blockData[pidx]); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // cidx + if (writer->blockData[cidx].nRow < writer->config->minRow) { + code = tsdbSttFileWriteBlockData(writer->sttWriter, &writer->blockData[cidx]); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + code = tsdbDataFileWriteBlockData(writer->dataWriter, &writer->blockData[cidx]); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + + for (int32_t i = 0; i < ARRAY_SIZE(writer->blockData); i++) { + tBlockDataReset(&writer->blockData[i]); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFSetWriterOpen(SFSetWriterConfig *config, SFSetWriter **writer) { + int32_t code = 0; + int32_t lino = 0; + + writer[0] = taosMemoryCalloc(1, sizeof(*writer[0])); + if (writer[0] == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + writer[0]->config[0] = config[0]; + + // data writer + if (!config->toSttOnly) { + SDataFileWriterConfig dataWriterConfig = { + .tsdb = config->tsdb, + .cmprAlg = config->cmprAlg, + .maxRow = config->maxRow, + .szPage = config->szPage, + .fid = config->fid, + .cid = config->cid, + .did = config->did, + .compactVersion = config->compactVersion, + .skmTb = writer[0]->skmTb, + .skmRow = writer[0]->skmRow, + .bufArr = writer[0]->bufArr, + }; + for (int32_t ftype = 0; ftype < TSDB_FTYPE_MAX; ++ftype) { + dataWriterConfig.files[ftype].exist = config->files[ftype].exist; + dataWriterConfig.files[ftype].file = config->files[ftype].file; + } + + code = tsdbDataFileWriterOpen(&dataWriterConfig, &writer[0]->dataWriter); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // stt writer + SSttFileWriterConfig sttWriterConfig = { + .tsdb = config->tsdb, + .maxRow = config->maxRow, + .szPage = config->szPage, + .cmprAlg = config->cmprAlg, + .compactVersion = config->compactVersion, + .did = config->did, + .fid = config->fid, + .cid = config->cid, + .level = config->level, + .skmTb = writer[0]->skmTb, + .skmRow = writer[0]->skmRow, + .bufArr = writer[0]->bufArr, + }; + code = tsdbSttFileWriterOpen(&sttWriterConfig, &writer[0]->sttWriter); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFSetWriterClose(SFSetWriter **writer, bool abort, TFileOpArray *fopArr) { + if (writer[0] == NULL) return 0; + + int32_t code = 0; + int32_t lino = 0; + + STsdb *tsdb = writer[0]->config->tsdb; + + // end + if (!writer[0]->config->toSttOnly) { + code = tsdbFSetWriteTableDataEnd(writer[0]); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFileWriterClose(&writer[0]->dataWriter, abort, fopArr); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbSttFileWriterClose(&writer[0]->sttWriter, abort, fopArr); + TSDB_CHECK_CODE(code, lino, _exit); + + // free + for (int32_t i = 0; i < ARRAY_SIZE(writer[0]->blockData); i++) { + tBlockDataDestroy(&writer[0]->blockData[i]); + } + for (int32_t i = 0; i < ARRAY_SIZE(writer[0]->bufArr); i++) { + tFree(writer[0]->bufArr[i]); + } + tDestroyTSchema(writer[0]->skmRow->pTSchema); + tDestroyTSchema(writer[0]->skmTb->pTSchema); + taosMemoryFree(writer[0]); + writer[0] = NULL; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFSetWriteRow(SFSetWriter *writer, SRowInfo *row) { + int32_t code = 0; + int32_t lino = 0; + + if (writer->config->toSttOnly) { + code = tsdbSttFileWriteRow(writer->sttWriter, row); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + if (writer->ctx->tbid->uid != row->uid) { + code = tsdbFSetWriteTableDataEnd(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbFSetWriteTableDataBegin(writer, (TABLEID *)row); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (row->row.type == TSDBROW_ROW_FMT) { + code = tsdbUpdateSkmRow(writer->config->tsdb, writer->ctx->tbid, TSDBROW_SVERSION(&row->row), writer->skmRow); + TSDB_CHECK_CODE(code, lino, _exit); + } + + TSDBKEY key = TSDBROW_KEY(&row->row); + if (key.version <= writer->config->compactVersion // + && writer->blockData[writer->blockDataIdx].nRow > 0 // + && key.ts == writer->blockData[writer->blockDataIdx].aTSKEY[writer->blockData[writer->blockDataIdx].nRow - 1]) { + code = tBlockDataUpdateRow(&writer->blockData[writer->blockDataIdx], &row->row, writer->skmRow->pTSchema); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + if (writer->blockData[writer->blockDataIdx].nRow >= writer->config->maxRow) { + int32_t idx = ((writer->blockDataIdx + 1) & 1); + if (writer->blockData[idx].nRow >= writer->config->maxRow) { + code = tsdbDataFileWriteBlockData(writer->dataWriter, &writer->blockData[idx]); + TSDB_CHECK_CODE(code, lino, _exit); + + tBlockDataClear(&writer->blockData[idx]); + } + writer->blockDataIdx = idx; + } + + code = + tBlockDataAppendRow(&writer->blockData[writer->blockDataIdx], &row->row, writer->skmRow->pTSchema, row->uid); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFSetWriteTombRecord(SFSetWriter *writer, const STombRecord *tombRecord) { + int32_t code = 0; + int32_t lino = 0; + + if (writer->config->toSttOnly || tsdbSttFileWriterIsOpened(writer->sttWriter)) { + code = tsdbSttFileWriteTombRecord(writer->sttWriter, tombRecord); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + code = tsdbDataFileWriteTombRecord(writer->dataWriter, tombRecord); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbFSetRW.h b/source/dnode/vnode/src/tsdb/tsdbFSetRW.h new file mode 100644 index 0000000000..b5710407cf --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbFSetRW.h @@ -0,0 +1,55 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbDataFileRW.h" +#include "tsdbSttFileRW.h" + +#ifndef _TSDB_FSET_RW_H +#define _TSDB_FSET_RW_H + +#ifdef __cplusplus +extern "C" { +#endif + +// +typedef struct SFSetWriter SFSetWriter; +typedef struct { + STsdb *tsdb; + bool toSttOnly; + int64_t compactVersion; + int32_t minRow; + int32_t maxRow; + int32_t szPage; + int8_t cmprAlg; + int32_t fid; + int64_t cid; + SDiskID did; + int32_t level; + struct { + bool exist; + STFile file; + } files[TSDB_FTYPE_MAX]; +} SFSetWriterConfig; + +int32_t tsdbFSetWriterOpen(SFSetWriterConfig *config, SFSetWriter **writer); +int32_t tsdbFSetWriterClose(SFSetWriter **writer, bool abort, TFileOpArray *fopArr); +int32_t tsdbFSetWriteRow(SFSetWriter *writer, SRowInfo *row); +int32_t tsdbFSetWriteTombRecord(SFSetWriter *writer, const STombRecord *tombRecord); + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_FSET_RW_H*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbFile2.c b/source/dnode/vnode/src/tsdb/tsdbFile2.c new file mode 100644 index 0000000000..be021169cd --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbFile2.c @@ -0,0 +1,295 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbFile2.h" + +// to_json +static int32_t head_to_json(const STFile *file, cJSON *json); +static int32_t data_to_json(const STFile *file, cJSON *json); +static int32_t sma_to_json(const STFile *file, cJSON *json); +static int32_t tomb_to_json(const STFile *file, cJSON *json); +static int32_t stt_to_json(const STFile *file, cJSON *json); + +// from_json +static int32_t head_from_json(const cJSON *json, STFile *file); +static int32_t data_from_json(const cJSON *json, STFile *file); +static int32_t sma_from_json(const cJSON *json, STFile *file); +static int32_t tomb_from_json(const cJSON *json, STFile *file); +static int32_t stt_from_json(const cJSON *json, STFile *file); + +static const struct { + const char *suffix; + int32_t (*to_json)(const STFile *file, cJSON *json); + int32_t (*from_json)(const cJSON *json, STFile *file); +} g_tfile_info[] = { + [TSDB_FTYPE_HEAD] = {"head", head_to_json, head_from_json}, + [TSDB_FTYPE_DATA] = {"data", data_to_json, data_from_json}, + [TSDB_FTYPE_SMA] = {"sma", sma_to_json, sma_from_json}, + [TSDB_FTYPE_TOMB] = {"tomb", tomb_to_json, tomb_from_json}, + [TSDB_FTYPE_STT] = {"stt", stt_to_json, stt_from_json}, +}; + +static void remove_file(const char *fname) { + taosRemoveFile(fname); + tsdbInfo("file:%s is removed", fname); +} + +static int32_t tfile_to_json(const STFile *file, cJSON *json) { + /* did.level */ + if (cJSON_AddNumberToObject(json, "did.level", file->did.level) == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + /* did.id */ + if (cJSON_AddNumberToObject(json, "did.id", file->did.id) == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + /* fid */ + if (cJSON_AddNumberToObject(json, "fid", file->fid) == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + /* cid */ + if (cJSON_AddNumberToObject(json, "cid", file->cid) == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + /* size */ + if (cJSON_AddNumberToObject(json, "size", file->size) == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + return 0; +} + +static int32_t tfile_from_json(const cJSON *json, STFile *file) { + const cJSON *item; + + /* did.level */ + item = cJSON_GetObjectItem(json, "did.level"); + if (cJSON_IsNumber(item)) { + file->did.level = item->valuedouble; + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + /* did.id */ + item = cJSON_GetObjectItem(json, "did.id"); + if (cJSON_IsNumber(item)) { + file->did.id = item->valuedouble; + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + /* fid */ + item = cJSON_GetObjectItem(json, "fid"); + if (cJSON_IsNumber(item)) { + file->fid = item->valuedouble; + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + /* cid */ + item = cJSON_GetObjectItem(json, "cid"); + if (cJSON_IsNumber(item)) { + file->cid = item->valuedouble; + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + /* size */ + item = cJSON_GetObjectItem(json, "size"); + if (cJSON_IsNumber(item)) { + file->size = item->valuedouble; + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + return 0; +} + +static int32_t head_to_json(const STFile *file, cJSON *json) { return tfile_to_json(file, json); } +static int32_t data_to_json(const STFile *file, cJSON *json) { return tfile_to_json(file, json); } +static int32_t sma_to_json(const STFile *file, cJSON *json) { return tfile_to_json(file, json); } +static int32_t tomb_to_json(const STFile *file, cJSON *json) { return tfile_to_json(file, json); } +static int32_t stt_to_json(const STFile *file, cJSON *json) { + int32_t code = tfile_to_json(file, json); + if (code) return code; + + /* lvl */ + if (cJSON_AddNumberToObject(json, "level", file->stt->level) == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + return 0; +} + +static int32_t head_from_json(const cJSON *json, STFile *file) { return tfile_from_json(json, file); } +static int32_t data_from_json(const cJSON *json, STFile *file) { return tfile_from_json(json, file); } +static int32_t sma_from_json(const cJSON *json, STFile *file) { return tfile_from_json(json, file); } +static int32_t tomb_from_json(const cJSON *json, STFile *file) { return tfile_from_json(json, file); } +static int32_t stt_from_json(const cJSON *json, STFile *file) { + int32_t code = tfile_from_json(json, file); + if (code) return code; + + const cJSON *item; + + /* lvl */ + item = cJSON_GetObjectItem(json, "level"); + if (cJSON_IsNumber(item)) { + file->stt->level = item->valuedouble; + } else { + return TSDB_CODE_FILE_CORRUPTED; + } + + return 0; +} + +int32_t tsdbTFileToJson(const STFile *file, cJSON *json) { + if (file->type == TSDB_FTYPE_STT) { + return g_tfile_info[file->type].to_json(file, json); + } else { + cJSON *item = cJSON_AddObjectToObject(json, g_tfile_info[file->type].suffix); + if (item == NULL) return TSDB_CODE_OUT_OF_MEMORY; + return g_tfile_info[file->type].to_json(file, item); + } +} + +int32_t tsdbJsonToTFile(const cJSON *json, tsdb_ftype_t ftype, STFile *f) { + f[0] = (STFile){.type = ftype}; + + if (ftype == TSDB_FTYPE_STT) { + int32_t code = g_tfile_info[ftype].from_json(json, f); + if (code) return code; + } else { + const cJSON *item = cJSON_GetObjectItem(json, g_tfile_info[ftype].suffix); + if (cJSON_IsObject(item)) { + int32_t code = g_tfile_info[ftype].from_json(item, f); + if (code) return code; + } else { + return TSDB_CODE_NOT_FOUND; + } + } + + return 0; +} + +int32_t tsdbTFileObjInit(STsdb *pTsdb, const STFile *f, STFileObj **fobj) { + fobj[0] = taosMemoryMalloc(sizeof(*fobj[0])); + if (!fobj[0]) return TSDB_CODE_OUT_OF_MEMORY; + + taosThreadMutexInit(&fobj[0]->mutex, NULL); + fobj[0]->f[0] = f[0]; + fobj[0]->state = TSDB_FSTATE_LIVE; + fobj[0]->ref = 1; + tsdbTFileName(pTsdb, f, fobj[0]->fname); + return 0; +} + +int32_t tsdbTFileObjRef(STFileObj *fobj) { + int32_t nRef; + taosThreadMutexLock(&fobj->mutex); + ASSERT(fobj->ref > 0 && fobj->state == TSDB_FSTATE_LIVE); + nRef = ++fobj->ref; + taosThreadMutexUnlock(&fobj->mutex); + tsdbTrace("ref file %s, fobj:%p ref %d", fobj->fname, fobj, nRef); + return 0; +} + +int32_t tsdbTFileObjUnref(STFileObj *fobj) { + taosThreadMutexLock(&fobj->mutex); + int32_t nRef = --fobj->ref; + taosThreadMutexUnlock(&fobj->mutex); + ASSERT(nRef >= 0); + tsdbTrace("unref file %s, fobj:%p ref %d", fobj->fname, fobj, nRef); + if (nRef == 0) { + if (fobj->state == TSDB_FSTATE_DEAD) { + remove_file(fobj->fname); + } + taosMemoryFree(fobj); + } + + return 0; +} + +int32_t tsdbTFileObjRemove(STFileObj *fobj) { + taosThreadMutexLock(&fobj->mutex); + ASSERT(fobj->state == TSDB_FSTATE_LIVE && fobj->ref > 0); + fobj->state = TSDB_FSTATE_DEAD; + int32_t nRef = --fobj->ref; + taosThreadMutexUnlock(&fobj->mutex); + tsdbTrace("remove unref file %s, fobj:%p ref %d", fobj->fname, fobj, nRef); + if (nRef == 0) { + remove_file(fobj->fname); + taosMemoryFree(fobj); + } + return 0; +} + +int32_t tsdbTFileName(STsdb *pTsdb, const STFile *f, char fname[]) { + SVnode *pVnode = pTsdb->pVnode; + STfs *pTfs = pVnode->pTfs; + + if (pTfs) { + snprintf(fname, // + TSDB_FILENAME_LEN, // + "%s%s%s%sv%df%dver%" PRId64 ".%s", // + tfsGetDiskPath(pTfs, f->did), // + TD_DIRSEP, // + pTsdb->path, // + TD_DIRSEP, // + TD_VID(pVnode), // + f->fid, // + f->cid, // + g_tfile_info[f->type].suffix); + } else { + snprintf(fname, // + TSDB_FILENAME_LEN, // + "%s%sv%df%dver%" PRId64 ".%s", // + pTsdb->path, // + TD_DIRSEP, // + TD_VID(pVnode), // + f->fid, // + f->cid, // + g_tfile_info[f->type].suffix); + } + return 0; +} + +bool tsdbIsSameTFile(const STFile *f1, const STFile *f2) { + if (f1->type != f2->type) return false; + if (f1->did.level != f2->did.level) return false; + if (f1->did.id != f2->did.id) return false; + if (f1->fid != f2->fid) return false; + if (f1->cid != f2->cid) return false; + return true; +} + +bool tsdbIsTFileChanged(const STFile *f1, const STFile *f2) { + if (f1->size != f2->size) return true; + // if (f1->type == TSDB_FTYPE_STT && f1->stt->nseg != f2->stt->nseg) return true; + return false; +} + +int32_t tsdbTFileObjCmpr(const STFileObj **fobj1, const STFileObj **fobj2) { + if (fobj1[0]->f->cid < fobj2[0]->f->cid) { + return -1; + } else if (fobj1[0]->f->cid > fobj2[0]->f->cid) { + return 1; + } else { + return 0; + } +} \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbFile2.h b/source/dnode/vnode/src/tsdb/tsdbFile2.h new file mode 100644 index 0000000000..11d08e45e6 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbFile2.h @@ -0,0 +1,83 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbDef.h" + +#ifndef _TSDB_FILE_H +#define _TSDB_FILE_H + +#ifdef __cplusplus +extern "C" { +#endif + +typedef struct STFile STFile; +typedef struct STFileObj STFileObj; + +typedef enum { + TSDB_FTYPE_HEAD = 0, // .head + TSDB_FTYPE_DATA, // .data + TSDB_FTYPE_SMA, // .sma + TSDB_FTYPE_TOMB, // .tomb + TSDB_FTYPE_STT = TSDB_FTYPE_TOMB + 2, // .stt +} tsdb_ftype_t; + +enum { + TSDB_FSTATE_LIVE = 1, + TSDB_FSTATE_DEAD, +}; + +#define TSDB_FTYPE_MIN TSDB_FTYPE_HEAD +#define TSDB_FTYPE_MAX (TSDB_FTYPE_TOMB + 1) + +// STFile +int32_t tsdbTFileToJson(const STFile *f, cJSON *json); +int32_t tsdbJsonToTFile(const cJSON *json, tsdb_ftype_t ftype, STFile *f); +int32_t tsdbTFileName(STsdb *pTsdb, const STFile *f, char fname[]); +bool tsdbIsSameTFile(const STFile *f1, const STFile *f2); +bool tsdbIsTFileChanged(const STFile *f1, const STFile *f2); + +// STFileObj +int32_t tsdbTFileObjInit(STsdb *pTsdb, const STFile *f, STFileObj **fobj); +int32_t tsdbTFileObjRef(STFileObj *fobj); +int32_t tsdbTFileObjUnref(STFileObj *fobj); +int32_t tsdbTFileObjRemove(STFileObj *fobj); +int32_t tsdbTFileObjCmpr(const STFileObj **fobj1, const STFileObj **fobj2); + +struct STFile { + tsdb_ftype_t type; + SDiskID did; // disk id + int32_t fid; // file id + int64_t cid; // commit id + int64_t size; + union { + struct { + int32_t level; + } stt[1]; + }; +}; + +struct STFileObj { + TdThreadMutex mutex; + STFile f[1]; + int32_t state; + int32_t ref; + char fname[TSDB_FILENAME_LEN]; +}; + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_FILE_H*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbIter.c b/source/dnode/vnode/src/tsdb/tsdbIter.c new file mode 100644 index 0000000000..9780cc6be6 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbIter.c @@ -0,0 +1,780 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbIter.h" + +// STsdbIter ================ +struct STsdbIter { + EIterType type; + bool noMoreData; + bool filterByVersion; + int64_t range[2]; + union { + SRowInfo row[1]; + STombRecord record[1]; + }; + SRBTreeNode node[1]; + union { + struct { + SSttFileReader *reader; + const TSttBlkArray *sttBlkArray; + int32_t sttBlkArrayIdx; + SBlockData blockData[1]; + int32_t blockDataIdx; + } sttData[1]; + struct { + SDataFileReader *reader; + const TBrinBlkArray *brinBlkArray; + int32_t brinBlkArrayIdx; + SBrinBlock brinBlock[1]; + int32_t brinBlockIdx; + SBlockData blockData[1]; + int32_t blockDataIdx; + } dataData[1]; + struct { + SMemTable *memt; + TSDBKEY from[1]; + SRBTreeIter iter[1]; + STbData *tbData; + STbDataIter tbIter[1]; + } memtData[1]; + struct { + SSttFileReader *reader; + const TTombBlkArray *tombBlkArray; + int32_t tombBlkArrayIdx; + STombBlock tombBlock[1]; + int32_t tombBlockIdx; + } sttTomb[1]; + struct { + SDataFileReader *reader; + const TTombBlkArray *tombBlkArray; + int32_t tombBlkArrayIdx; + STombBlock tombBlock[1]; + int32_t tombBlockIdx; + } dataTomb[1]; + struct { + SMemTable *memt; + SRBTreeIter rbtIter[1]; + STbData *tbData; + SDelData *delData; + } memtTomb[1]; + }; +}; + +static int32_t tsdbSttIterNext(STsdbIter *iter, const TABLEID *tbid) { + while (!iter->noMoreData) { + for (; iter->sttData->blockDataIdx < iter->sttData->blockData->nRow; iter->sttData->blockDataIdx++) { + int64_t version = iter->sttData->blockData->aVersion[iter->sttData->blockDataIdx]; + + if (iter->filterByVersion && (version < iter->range[0] || version > iter->range[1])) { + continue; + } + + iter->row->suid = iter->sttData->blockData->suid; + iter->row->uid = iter->sttData->blockData->uid ? iter->sttData->blockData->uid + : iter->sttData->blockData->aUid[iter->sttData->blockDataIdx]; + + if (tbid && iter->row->suid == tbid->suid && iter->row->uid == tbid->uid) { + continue; + } + + iter->row->row = tsdbRowFromBlockData(iter->sttData->blockData, iter->sttData->blockDataIdx); + iter->sttData->blockDataIdx++; + goto _exit; + } + + if (iter->sttData->sttBlkArrayIdx >= TARRAY2_SIZE(iter->sttData->sttBlkArray)) { + iter->noMoreData = true; + break; + } + + for (; iter->sttData->sttBlkArrayIdx < TARRAY2_SIZE(iter->sttData->sttBlkArray); iter->sttData->sttBlkArrayIdx++) { + const SSttBlk *sttBlk = TARRAY2_GET_PTR(iter->sttData->sttBlkArray, iter->sttData->sttBlkArrayIdx); + + if (iter->filterByVersion && (sttBlk->maxVer < iter->range[0] || sttBlk->minVer > iter->range[1])) { + continue; + } + + if (tbid && tbid->suid == sttBlk->suid && tbid->uid == sttBlk->minUid && tbid->uid == sttBlk->maxUid) { + continue; + } + + int32_t code = tsdbSttFileReadBlockData(iter->sttData->reader, sttBlk, iter->sttData->blockData); + if (code) return code; + + iter->sttData->blockDataIdx = 0; + iter->sttData->sttBlkArrayIdx++; + break; + } + } + +_exit: + return 0; +} + +static int32_t tsdbDataIterNext(STsdbIter *iter, const TABLEID *tbid) { + int32_t code; + + while (!iter->noMoreData) { + for (;;) { + // SBlockData + for (; iter->dataData->blockDataIdx < iter->dataData->blockData->nRow; iter->dataData->blockDataIdx++) { + int64_t version = iter->dataData->blockData->aVersion[iter->dataData->blockDataIdx]; + if (iter->filterByVersion && (version < iter->range[0] || version > iter->range[1])) { + continue; + } + + if (tbid && tbid->suid == iter->dataData->blockData->suid && tbid->uid == iter->dataData->blockData->uid) { + iter->dataData->blockDataIdx = iter->dataData->blockData->nRow; + break; + } + + iter->row->row = tsdbRowFromBlockData(iter->dataData->blockData, iter->dataData->blockDataIdx); + iter->dataData->blockDataIdx++; + goto _exit; + } + + // SBrinBlock + if (iter->dataData->brinBlockIdx >= BRIN_BLOCK_SIZE(iter->dataData->brinBlock)) { + break; + } + + for (; iter->dataData->brinBlockIdx < BRIN_BLOCK_SIZE(iter->dataData->brinBlock); + iter->dataData->brinBlockIdx++) { + SBrinRecord record[1]; + tBrinBlockGet(iter->dataData->brinBlock, iter->dataData->brinBlockIdx, record); + + if (iter->filterByVersion && (record->maxVer < iter->range[0] || record->minVer > iter->range[1])) { + continue; + } + + if (tbid && tbid->suid == record->suid && tbid->uid == record->uid) { + continue; + } + + iter->row->suid = record->suid; + iter->row->uid = record->uid; + + code = tsdbDataFileReadBlockData(iter->dataData->reader, record, iter->dataData->blockData); + if (code) return code; + + iter->dataData->blockDataIdx = 0; + iter->dataData->brinBlockIdx++; + break; + } + } + + if (iter->dataData->brinBlkArrayIdx >= TARRAY2_SIZE(iter->dataData->brinBlkArray)) { + iter->noMoreData = true; + break; + } + + for (; iter->dataData->brinBlkArrayIdx < TARRAY2_SIZE(iter->dataData->brinBlkArray); + iter->dataData->brinBlkArrayIdx++) { + const SBrinBlk *brinBlk = TARRAY2_GET_PTR(iter->dataData->brinBlkArray, iter->dataData->brinBlkArrayIdx); + + if (iter->filterByVersion && (brinBlk->maxVer < iter->range[0] || brinBlk->minVer > iter->range[1])) { + continue; + } + + if (tbid && tbid->uid == brinBlk->minTbid.uid && tbid->uid == brinBlk->maxTbid.uid) { + continue; + } + + code = tsdbDataFileReadBrinBlock(iter->dataData->reader, brinBlk, iter->dataData->brinBlock); + if (code) return code; + + iter->dataData->brinBlockIdx = 0; + iter->dataData->brinBlkArrayIdx++; + break; + } + } + +_exit: + return 0; +} + +static int32_t tsdbMemTableIterNext(STsdbIter *iter, const TABLEID *tbid) { + SRBTreeNode *node; + + while (!iter->noMoreData) { + for (TSDBROW *row; iter->memtData->tbData && (row = tsdbTbDataIterGet(iter->memtData->tbIter));) { + if (tbid && tbid->suid == iter->memtData->tbData->suid && tbid->uid == iter->memtData->tbData->uid) { + iter->memtData->tbData = NULL; + break; + } + + if (iter->filterByVersion) { + int64_t version = TSDBROW_VERSION(row); + if (version < iter->range[0] || version > iter->range[1]) { + continue; + } + } + + iter->row->row = row[0]; + + tsdbTbDataIterNext(iter->memtData->tbIter); + goto _exit; + } + + for (;;) { + node = tRBTreeIterNext(iter->memtData->iter); + if (!node) { + iter->noMoreData = true; + goto _exit; + } + + iter->memtData->tbData = TCONTAINER_OF(node, STbData, rbtn); + if (tbid && tbid->suid == iter->memtData->tbData->suid && tbid->uid == iter->memtData->tbData->uid) { + continue; + } else { + iter->row->suid = iter->memtData->tbData->suid; + iter->row->uid = iter->memtData->tbData->uid; + tsdbTbDataIterOpen(iter->memtData->tbData, iter->memtData->from, 0, iter->memtData->tbIter); + break; + } + } + } + +_exit: + return 0; +} + +static int32_t tsdbDataTombIterNext(STsdbIter *iter, const TABLEID *tbid) { + while (!iter->noMoreData) { + for (; iter->dataTomb->tombBlockIdx < TOMB_BLOCK_SIZE(iter->dataTomb->tombBlock); iter->dataTomb->tombBlockIdx++) { + iter->record->suid = TARRAY2_GET(iter->dataTomb->tombBlock->suid, iter->dataTomb->tombBlockIdx); + iter->record->uid = TARRAY2_GET(iter->dataTomb->tombBlock->uid, iter->dataTomb->tombBlockIdx); + iter->record->version = TARRAY2_GET(iter->dataTomb->tombBlock->version, iter->dataTomb->tombBlockIdx); + + if (iter->filterByVersion && (iter->record->version < iter->range[0] || iter->record->version > iter->range[1])) { + continue; + } + + if (tbid && iter->record->suid == tbid->suid && iter->record->uid == tbid->uid) { + continue; + } + + iter->record->skey = TARRAY2_GET(iter->dataTomb->tombBlock->skey, iter->dataTomb->tombBlockIdx); + iter->record->ekey = TARRAY2_GET(iter->dataTomb->tombBlock->ekey, iter->dataTomb->tombBlockIdx); + iter->dataTomb->tombBlockIdx++; + goto _exit; + } + + if (iter->dataTomb->tombBlkArrayIdx >= TARRAY2_SIZE(iter->dataTomb->tombBlkArray)) { + iter->noMoreData = true; + goto _exit; + } + + for (; iter->dataTomb->tombBlkArrayIdx < TARRAY2_SIZE(iter->dataTomb->tombBlkArray); + iter->dataTomb->tombBlkArrayIdx++) { + const STombBlk *tombBlk = TARRAY2_GET_PTR(iter->dataTomb->tombBlkArray, iter->dataTomb->tombBlkArrayIdx); + + if (tbid && tbid->suid == tombBlk->minTbid.suid && tbid->uid == tombBlk->minTbid.uid && + tbid->suid == tombBlk->maxTbid.suid && tbid->uid == tombBlk->maxTbid.uid) { + continue; + } + + int32_t code = tsdbDataFileReadTombBlock(iter->dataTomb->reader, tombBlk, iter->dataTomb->tombBlock); + if (code) return code; + + iter->dataTomb->tombBlockIdx = 0; + iter->dataTomb->tombBlkArrayIdx++; + break; + } + } + +_exit: + return 0; +} + +static int32_t tsdbMemTombIterNext(STsdbIter *iter, const TABLEID *tbid) { + while (!iter->noMoreData) { + for (; iter->memtTomb->delData;) { + if (tbid && tbid->uid == iter->memtTomb->tbData->uid) { + iter->memtTomb->delData = NULL; + break; + } + + if (iter->filterByVersion && + (iter->memtTomb->delData->version < iter->range[0] || iter->memtTomb->delData->version > iter->range[1])) { + continue; + } + + iter->record->suid = iter->memtTomb->tbData->suid; + iter->record->uid = iter->memtTomb->tbData->uid; + iter->record->version = iter->memtTomb->delData->version; + iter->record->skey = iter->memtTomb->delData->sKey; + iter->record->ekey = iter->memtTomb->delData->eKey; + + iter->memtTomb->delData = iter->memtTomb->delData->pNext; + goto _exit; + } + + for (;;) { + SRBTreeNode *node = tRBTreeIterNext(iter->memtTomb->rbtIter); + if (node == NULL) { + iter->noMoreData = true; + goto _exit; + } + + iter->memtTomb->tbData = TCONTAINER_OF(node, STbData, rbtn); + if (tbid && tbid->uid == iter->memtTomb->tbData->uid) { + continue; + } else { + iter->memtTomb->delData = iter->memtTomb->tbData->pHead; + break; + } + } + } + +_exit: + return 0; +} + +static int32_t tsdbSttIterOpen(STsdbIter *iter) { + int32_t code; + + code = tsdbSttFileReadSttBlk(iter->sttData->reader, &iter->sttData->sttBlkArray); + if (code) return code; + + if (TARRAY2_SIZE(iter->sttData->sttBlkArray) == 0) { + iter->noMoreData = true; + return 0; + } + + iter->sttData->sttBlkArrayIdx = 0; + tBlockDataCreate(iter->sttData->blockData); + iter->sttData->blockDataIdx = 0; + + return tsdbSttIterNext(iter, NULL); +} + +static int32_t tsdbDataIterOpen(STsdbIter *iter) { + int32_t code; + + // SBrinBlk + code = tsdbDataFileReadBrinBlk(iter->dataData->reader, &iter->dataData->brinBlkArray); + if (code) return code; + + if (TARRAY2_SIZE(iter->dataData->brinBlkArray) == 0) { + iter->noMoreData = true; + return 0; + } + + iter->dataData->brinBlkArrayIdx = 0; + + // SBrinBlock + tBrinBlockInit(iter->dataData->brinBlock); + iter->dataData->brinBlockIdx = 0; + + // SBlockData + tBlockDataCreate(iter->dataData->blockData); + iter->dataData->blockDataIdx = 0; + + return tsdbDataIterNext(iter, NULL); +} + +static int32_t tsdbMemTableIterOpen(STsdbIter *iter) { + if (iter->memtData->memt->nRow == 0) { + iter->noMoreData = true; + return 0; + } + + iter->memtData->iter[0] = tRBTreeIterCreate(iter->memtData->memt->tbDataTree, 1); + return tsdbMemTableIterNext(iter, NULL); +} + +static int32_t tsdbSttIterClose(STsdbIter *iter) { + tBlockDataDestroy(iter->sttData->blockData); + return 0; +} + +static int32_t tsdbDataTombIterOpen(STsdbIter *iter) { + int32_t code; + + code = tsdbDataFileReadTombBlk(iter->dataTomb->reader, &iter->dataTomb->tombBlkArray); + if (code) return code; + + if (TARRAY2_SIZE(iter->dataTomb->tombBlkArray) == 0) { + iter->noMoreData = true; + return 0; + } + iter->dataTomb->tombBlkArrayIdx = 0; + + tTombBlockInit(iter->dataTomb->tombBlock); + iter->dataTomb->tombBlockIdx = 0; + + return tsdbDataTombIterNext(iter, NULL); +} + +static int32_t tsdbMemTombIterOpen(STsdbIter *iter) { + int32_t code; + + if (iter->memtTomb->memt->nDel == 0) { + iter->noMoreData = true; + return 0; + } + + iter->memtTomb->rbtIter[0] = tRBTreeIterCreate(iter->memtTomb->memt->tbDataTree, 1); + return tsdbMemTombIterNext(iter, NULL); +} + +static int32_t tsdbDataIterClose(STsdbIter *iter) { + tBrinBlockDestroy(iter->dataData->brinBlock); + tBlockDataDestroy(iter->dataData->blockData); + return 0; +} + +static int32_t tsdbMemTableIterClose(STsdbIter *iter) { return 0; } + +static int32_t tsdbSttTombIterNext(STsdbIter *iter, const TABLEID *tbid) { + while (!iter->noMoreData) { + for (; iter->sttTomb->tombBlockIdx < TOMB_BLOCK_SIZE(iter->sttTomb->tombBlock); iter->sttTomb->tombBlockIdx++) { + iter->record->suid = TARRAY2_GET(iter->sttTomb->tombBlock->suid, iter->sttTomb->tombBlockIdx); + iter->record->uid = TARRAY2_GET(iter->sttTomb->tombBlock->uid, iter->sttTomb->tombBlockIdx); + iter->record->version = TARRAY2_GET(iter->sttTomb->tombBlock->version, iter->sttTomb->tombBlockIdx); + + if (iter->filterByVersion && (iter->record->version < iter->range[0] || iter->record->version > iter->range[1])) { + continue; + } + + if (tbid && iter->record->suid == tbid->suid && iter->record->uid == tbid->uid) { + continue; + } + + iter->record->skey = TARRAY2_GET(iter->sttTomb->tombBlock->skey, iter->sttTomb->tombBlockIdx); + iter->record->ekey = TARRAY2_GET(iter->sttTomb->tombBlock->ekey, iter->sttTomb->tombBlockIdx); + iter->sttTomb->tombBlockIdx++; + goto _exit; + } + + if (iter->sttTomb->tombBlkArrayIdx >= TARRAY2_SIZE(iter->sttTomb->tombBlkArray)) { + iter->noMoreData = true; + goto _exit; + } + + for (; iter->sttTomb->tombBlkArrayIdx < TARRAY2_SIZE(iter->sttTomb->tombBlkArray); + iter->sttTomb->tombBlkArrayIdx++) { + const STombBlk *tombBlk = TARRAY2_GET_PTR(iter->sttTomb->tombBlkArray, iter->sttTomb->tombBlkArrayIdx); + + if (iter->filterByVersion && (tombBlk->maxVer < iter->range[0] || tombBlk->minVer > iter->range[1])) { + continue; + } + + if (tbid && tbid->suid == tombBlk->minTbid.suid && tbid->uid == tombBlk->minTbid.uid && + tbid->suid == tombBlk->maxTbid.suid && tbid->uid == tombBlk->maxTbid.uid) { + continue; + } + + int32_t code = tsdbSttFileReadTombBlock(iter->sttTomb->reader, tombBlk, iter->sttTomb->tombBlock); + if (code) return code; + + iter->sttTomb->tombBlockIdx = 0; + iter->sttTomb->tombBlkArrayIdx++; + break; + } + } + +_exit: + return 0; +} + +static int32_t tsdbSttTombIterOpen(STsdbIter *iter) { + int32_t code; + + code = tsdbSttFileReadTombBlk(iter->sttTomb->reader, &iter->sttTomb->tombBlkArray); + if (code) return code; + + if (TARRAY2_SIZE(iter->sttTomb->tombBlkArray) == 0) { + iter->noMoreData = true; + return 0; + } + + iter->sttTomb->tombBlkArrayIdx = 0; + tTombBlockInit(iter->sttTomb->tombBlock); + iter->sttTomb->tombBlockIdx = 0; + + return tsdbSttTombIterNext(iter, NULL); +} + +int32_t tsdbIterOpen(const STsdbIterConfig *config, STsdbIter **iter) { + int32_t code; + + iter[0] = taosMemoryCalloc(1, sizeof(*iter[0])); + if (iter[0] == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + iter[0]->type = config->type; + iter[0]->noMoreData = false; + iter[0]->filterByVersion = config->filterByVersion; + if (iter[0]->filterByVersion) { + iter[0]->range[0] = config->verRange[0]; + iter[0]->range[1] = config->verRange[1]; + } + + switch (config->type) { + case TSDB_ITER_TYPE_STT: + iter[0]->sttData->reader = config->sttReader; + code = tsdbSttIterOpen(iter[0]); + break; + case TSDB_ITER_TYPE_DATA: + iter[0]->dataData->reader = config->dataReader; + code = tsdbDataIterOpen(iter[0]); + break; + case TSDB_ITER_TYPE_MEMT: + iter[0]->memtData->memt = config->memt; + iter[0]->memtData->from[0] = config->from[0]; + code = tsdbMemTableIterOpen(iter[0]); + break; + case TSDB_ITER_TYPE_STT_TOMB: + iter[0]->sttTomb->reader = config->sttReader; + code = tsdbSttTombIterOpen(iter[0]); + break; + case TSDB_ITER_TYPE_DATA_TOMB: + iter[0]->dataTomb->reader = config->dataReader; + code = tsdbDataTombIterOpen(iter[0]); + break; + case TSDB_ITER_TYPE_MEMT_TOMB: + iter[0]->memtTomb->memt = config->memt; + code = tsdbMemTombIterOpen(iter[0]); + break; + default: + code = TSDB_CODE_INVALID_PARA; + ASSERTS(false, "Not implemented"); + } + + if (code) { + taosMemoryFree(iter[0]); + iter[0] = NULL; + } + return code; +} + +static int32_t tsdbSttTombIterClose(STsdbIter *iter) { + tTombBlockDestroy(iter->sttTomb->tombBlock); + return 0; +} + +static int32_t tsdbDataTombIterClose(STsdbIter *iter) { + tTombBlockDestroy(iter->dataTomb->tombBlock); + return 0; +} + +int32_t tsdbIterClose(STsdbIter **iter) { + switch (iter[0]->type) { + case TSDB_ITER_TYPE_STT: + tsdbSttIterClose(iter[0]); + break; + case TSDB_ITER_TYPE_DATA: + tsdbDataIterClose(iter[0]); + break; + case TSDB_ITER_TYPE_MEMT: + tsdbMemTableIterClose(iter[0]); + break; + case TSDB_ITER_TYPE_STT_TOMB: + tsdbSttTombIterClose(iter[0]); + break; + case TSDB_ITER_TYPE_DATA_TOMB: + tsdbDataTombIterClose(iter[0]); + break; + case TSDB_ITER_TYPE_MEMT_TOMB: + break; + default: + ASSERT(false); + } + taosMemoryFree(iter[0]); + iter[0] = NULL; + return 0; +} + +int32_t tsdbIterNext(STsdbIter *iter) { + switch (iter->type) { + case TSDB_ITER_TYPE_STT: + return tsdbSttIterNext(iter, NULL); + case TSDB_ITER_TYPE_DATA: + return tsdbDataIterNext(iter, NULL); + case TSDB_ITER_TYPE_MEMT: + return tsdbMemTableIterNext(iter, NULL); + case TSDB_ITER_TYPE_STT_TOMB: + return tsdbSttTombIterNext(iter, NULL); + case TSDB_ITER_TYPE_DATA_TOMB: + return tsdbDataTombIterNext(iter, NULL); + case TSDB_ITER_TYPE_MEMT_TOMB: + return tsdbMemTombIterNext(iter, NULL); + default: + ASSERT(false); + } + return 0; +} + +static int32_t tsdbIterSkipTableData(STsdbIter *iter, const TABLEID *tbid) { + switch (iter->type) { + case TSDB_ITER_TYPE_STT: + return tsdbSttIterNext(iter, tbid); + case TSDB_ITER_TYPE_DATA: + return tsdbDataIterNext(iter, tbid); + case TSDB_ITER_TYPE_MEMT: + return tsdbMemTableIterNext(iter, tbid); + case TSDB_ITER_TYPE_STT_TOMB: + return tsdbSttTombIterNext(iter, tbid); + case TSDB_ITER_TYPE_DATA_TOMB: + return tsdbDataTombIterNext(iter, tbid); + case TSDB_ITER_TYPE_MEMT_TOMB: + return tsdbMemTombIterNext(iter, tbid); + default: + ASSERT(false); + } + return 0; +} + +static int32_t tsdbIterCmprFn(const SRBTreeNode *n1, const SRBTreeNode *n2) { + STsdbIter *iter1 = TCONTAINER_OF(n1, STsdbIter, node); + STsdbIter *iter2 = TCONTAINER_OF(n2, STsdbIter, node); + return tRowInfoCmprFn(&iter1->row, &iter2->row); +} + +static int32_t tsdbTombIterCmprFn(const SRBTreeNode *n1, const SRBTreeNode *n2) { + STsdbIter *iter1 = TCONTAINER_OF(n1, STsdbIter, node); + STsdbIter *iter2 = TCONTAINER_OF(n2, STsdbIter, node); + + if (iter1->record->suid < iter2->record->suid) { + return -1; + } else if (iter1->record->suid > iter2->record->suid) { + return 1; + } + + if (iter1->record->uid < iter2->record->uid) { + return -1; + } else if (iter1->record->uid > iter2->record->uid) { + return 1; + } + + if (iter1->record->version < iter2->record->version) { + return -1; + } else if (iter1->record->version > iter2->record->version) { + return 1; + } + + return 0; +} + +// SIterMerger ================ +struct SIterMerger { + bool isTomb; + STsdbIter *iter; + SRBTree iterTree[1]; +}; + +int32_t tsdbIterMergerOpen(const TTsdbIterArray *iterArray, SIterMerger **merger, bool isTomb) { + STsdbIter *iter; + SRBTreeNode *node; + + merger[0] = taosMemoryCalloc(1, sizeof(*merger[0])); + if (merger[0] == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + merger[0]->isTomb = isTomb; + if (isTomb) { + tRBTreeCreate(merger[0]->iterTree, tsdbTombIterCmprFn); + } else { + tRBTreeCreate(merger[0]->iterTree, tsdbIterCmprFn); + } + TARRAY2_FOREACH(iterArray, iter) { + if (iter->noMoreData) continue; + node = tRBTreePut(merger[0]->iterTree, iter->node); + ASSERT(node); + } + + return tsdbIterMergerNext(merger[0]); +} + +int32_t tsdbIterMergerClose(SIterMerger **merger) { + if (merger[0]) { + taosMemoryFree(merger[0]); + merger[0] = NULL; + } + return 0; +} + +int32_t tsdbIterMergerNext(SIterMerger *merger) { + int32_t code; + int32_t c; + SRBTreeNode *node; + + if (merger->iter) { + code = tsdbIterNext(merger->iter); + if (code) return code; + + if (merger->iter->noMoreData) { + merger->iter = NULL; + } else if ((node = tRBTreeMin(merger->iterTree))) { + c = merger->iterTree->cmprFn(merger->iter->node, node); + ASSERT(c); + if (c > 0) { + node = tRBTreePut(merger->iterTree, merger->iter->node); + ASSERT(node); + merger->iter = NULL; + } + } + } + + if (merger->iter == NULL && (node = tRBTreeDropMin(merger->iterTree))) { + merger->iter = TCONTAINER_OF(node, STsdbIter, node); + } + + return 0; +} + +SRowInfo *tsdbIterMergerGetData(SIterMerger *merger) { + ASSERT(!merger->isTomb); + return merger->iter ? merger->iter->row : NULL; +} + +STombRecord *tsdbIterMergerGetTombRecord(SIterMerger *merger) { + ASSERT(merger->isTomb); + return merger->iter ? merger->iter->record : NULL; +} + +int32_t tsdbIterMergerSkipTableData(SIterMerger *merger, const TABLEID *tbid) { + int32_t code; + int32_t c; + SRBTreeNode *node; + + while (merger->iter && tbid->suid == merger->iter->row->suid && tbid->uid == merger->iter->row->uid) { + int32_t code = tsdbIterSkipTableData(merger->iter, tbid); + if (code) return code; + + if (merger->iter->noMoreData) { + merger->iter = NULL; + } else if ((node = tRBTreeMin(merger->iterTree))) { + c = merger->iterTree->cmprFn(merger->iter->node, node); + ASSERT(c); + if (c > 0) { + node = tRBTreePut(merger->iterTree, merger->iter->node); + ASSERT(node); + merger->iter = NULL; + } + } + + if (!merger->iter && (node = tRBTreeDropMin(merger->iterTree))) { + merger->iter = TCONTAINER_OF(node, STsdbIter, node); + } + } + + return 0; +} \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbIter.h b/source/dnode/vnode/src/tsdb/tsdbIter.h new file mode 100644 index 0000000000..367901bd84 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbIter.h @@ -0,0 +1,73 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "trbtree.h" +#include "tsdbDataFileRW.h" +#include "tsdbDef.h" +#include "tsdbSttFileRW.h" + +#ifndef _TSDB_ITER_H_ +#define _TSDB_ITER_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +typedef struct SIterMerger SIterMerger; +typedef struct STsdbIter STsdbIter; +typedef TARRAY2(STsdbIter *) TTsdbIterArray; + +typedef enum { + TSDB_ITER_TYPE_STT = 1, + TSDB_ITER_TYPE_DATA, + TSDB_ITER_TYPE_MEMT, + TSDB_ITER_TYPE_STT_TOMB, + TSDB_ITER_TYPE_DATA_TOMB, + TSDB_ITER_TYPE_MEMT_TOMB, +} EIterType; + +typedef struct { + EIterType type; + union { + SSttFileReader *sttReader; // TSDB_ITER_TYPE_STT || TSDB_ITER_TYPE_STT_TOMB + SDataFileReader *dataReader; // TSDB_ITER_TYPE_DATA || TSDB_ITER_TYPE_DATA_TOMB + struct { + SMemTable *memt; // TSDB_ITER_TYPE_MEMT_TOMB + TSDBKEY from[1]; + }; // TSDB_ITER_TYPE_MEMT + }; + bool filterByVersion; + int64_t verRange[2]; +} STsdbIterConfig; + +// STsdbIter =============== +int32_t tsdbIterOpen(const STsdbIterConfig *config, STsdbIter **iter); +int32_t tsdbIterClose(STsdbIter **iter); +int32_t tsdbIterNext(STsdbIter *iter); + +// SIterMerger =============== +int32_t tsdbIterMergerOpen(const TTsdbIterArray *iterArray, SIterMerger **merger, bool isTomb); +int32_t tsdbIterMergerClose(SIterMerger **merger); +int32_t tsdbIterMergerNext(SIterMerger *merger); +int32_t tsdbIterMergerSkipTableData(SIterMerger *merger, const TABLEID *tbid); + +SRowInfo *tsdbIterMergerGetData(SIterMerger *merger); +STombRecord *tsdbIterMergerGetTombRecord(SIterMerger *merger); + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_ITER_H_*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbMemTable.c b/source/dnode/vnode/src/tsdb/tsdbMemTable.c index 6d223e00c5..ee3abf7559 100644 --- a/source/dnode/vnode/src/tsdb/tsdbMemTable.c +++ b/source/dnode/vnode/src/tsdb/tsdbMemTable.c @@ -38,6 +38,16 @@ static int32_t tsdbInsertRowDataToTable(SMemTable *pMemTable, STbData *pTbData, static int32_t tsdbInsertColDataToTable(SMemTable *pMemTable, STbData *pTbData, int64_t version, SSubmitTbData *pSubmitTbData, int32_t *affectedRows); +static int32_t tTbDataCmprFn(const SRBTreeNode *n1, const SRBTreeNode *n2) { + STbData *tbData1 = TCONTAINER_OF(n1, STbData, rbtn); + STbData *tbData2 = TCONTAINER_OF(n2, STbData, rbtn); + if (tbData1->suid < tbData2->suid) return -1; + if (tbData1->suid > tbData2->suid) return 1; + if (tbData1->uid < tbData2->uid) return -1; + if (tbData1->uid > tbData2->uid) return 1; + return 0; +} + int32_t tsdbMemTableCreate(STsdb *pTsdb, SMemTable **ppMemTable) { int32_t code = 0; SMemTable *pMemTable = NULL; @@ -66,6 +76,7 @@ int32_t tsdbMemTableCreate(STsdb *pTsdb, SMemTable **ppMemTable) { goto _err; } vnodeBufPoolRef(pMemTable->pPool); + tRBTreeCreate(pMemTable->tbDataTree, tTbDataCmprFn); *ppMemTable = pMemTable; return code; @@ -406,6 +417,8 @@ static int32_t tsdbGetOrCreateTbData(SMemTable *pMemTable, tb_uid_t suid, tb_uid pMemTable->aBucket[idx] = pTbData; pMemTable->nTbData++; + tRBTreePut(pMemTable->tbDataTree, pTbData->rbtn); + taosWUnLockLatch(&pMemTable->latch); _exit: diff --git a/source/dnode/vnode/src/tsdb/tsdbMerge.c b/source/dnode/vnode/src/tsdb/tsdbMerge.c new file mode 100644 index 0000000000..ec0ea3c60f --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbMerge.c @@ -0,0 +1,454 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbMerge.h" + +typedef struct { + STsdb *tsdb; + TFileSetArray *fsetArr; + + int32_t sttTrigger; + int32_t maxRow; + int32_t minRow; + int32_t szPage; + int8_t cmprAlg; + int64_t compactVersion; + int64_t cid; + + // context + struct { + bool opened; + int64_t now; + STFileSet *fset; + bool toData; + int32_t level; + SSttLvl *lvl; + TABLEID tbid[1]; + } ctx[1]; + + TFileOpArray fopArr[1]; + + // reader + TSttFileReaderArray sttReaderArr[1]; + // iter + TTsdbIterArray dataIterArr[1]; + SIterMerger *dataIterMerger; + TTsdbIterArray tombIterArr[1]; + SIterMerger *tombIterMerger; + // writer + SFSetWriter *writer; +} SMerger; + +static int32_t tsdbMergerOpen(SMerger *merger) { + merger->ctx->now = taosGetTimestampSec(); + merger->maxRow = merger->tsdb->pVnode->config.tsdbCfg.maxRows; + merger->minRow = merger->tsdb->pVnode->config.tsdbCfg.minRows; + merger->szPage = merger->tsdb->pVnode->config.tsdbPageSize; + merger->cmprAlg = merger->tsdb->pVnode->config.tsdbCfg.compression; + merger->compactVersion = INT64_MAX; + merger->cid = tsdbFSAllocEid(merger->tsdb->pFS); + merger->ctx->opened = true; + return 0; +} + +static int32_t tsdbMergerClose(SMerger *merger) { + int32_t code = 0; + int32_t lino = 0; + SVnode *pVnode = merger->tsdb->pVnode; + + // edit file system + code = tsdbFSEditBegin(merger->tsdb->pFS, merger->fopArr, TSDB_FEDIT_MERGE); + TSDB_CHECK_CODE(code, lino, _exit); + + taosThreadRwlockWrlock(&merger->tsdb->rwLock); + code = tsdbFSEditCommit(merger->tsdb->pFS); + if (code) { + taosThreadRwlockUnlock(&merger->tsdb->rwLock); + TSDB_CHECK_CODE(code, lino, _exit); + } + taosThreadRwlockUnlock(&merger->tsdb->rwLock); + + ASSERT(merger->writer == NULL); + ASSERT(merger->dataIterMerger == NULL); + ASSERT(merger->tombIterMerger == NULL); + ASSERT(TARRAY2_SIZE(merger->dataIterArr) == 0); + ASSERT(TARRAY2_SIZE(merger->tombIterArr) == 0); + ASSERT(TARRAY2_SIZE(merger->sttReaderArr) == 0); + + // clear the merge + TARRAY2_DESTROY(merger->tombIterArr, NULL); + TARRAY2_DESTROY(merger->dataIterArr, NULL); + TARRAY2_DESTROY(merger->sttReaderArr, NULL); + TARRAY2_DESTROY(merger->fopArr, NULL); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(pVnode), lino, code); + } + return code; +} + +static int32_t tsdbMergeFileSetBeginOpenReader(SMerger *merger) { + int32_t code = 0; + int32_t lino = 0; + + merger->ctx->toData = true; + merger->ctx->level = 0; + + // TODO: optimize merge strategy + for (int32_t i = 0;; ++i) { + if (i >= TARRAY2_SIZE(merger->ctx->fset->lvlArr)) { + merger->ctx->lvl = NULL; + break; + } + + merger->ctx->lvl = TARRAY2_GET(merger->ctx->fset->lvlArr, i); + if (merger->ctx->lvl->level != merger->ctx->level || + TARRAY2_SIZE(merger->ctx->lvl->fobjArr) + 1 < merger->sttTrigger) { + merger->ctx->toData = false; + merger->ctx->lvl = NULL; + break; + } + + merger->ctx->level++; + + STFileObj *fobj; + int32_t numFile = 0; + TARRAY2_FOREACH(merger->ctx->lvl->fobjArr, fobj) { + if (numFile == merger->sttTrigger) { + break; + } + + STFileOp op = { + .optype = TSDB_FOP_REMOVE, + .fid = merger->ctx->fset->fid, + .of = fobj->f[0], + }; + code = TARRAY2_APPEND(merger->fopArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + + SSttFileReader *reader; + SSttFileReaderConfig config = { + .tsdb = merger->tsdb, + .szPage = merger->szPage, + .file[0] = fobj->f[0], + }; + + code = tsdbSttFileReaderOpen(fobj->fname, &config, &reader); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(merger->sttReaderArr, reader); + TSDB_CHECK_CODE(code, lino, _exit); + + numFile++; + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(merger->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbMergeFileSetBeginOpenIter(SMerger *merger) { + int32_t code = 0; + int32_t lino = 0; + int32_t vid = TD_VID(merger->tsdb->pVnode); + + SSttFileReader *sttReader; + TARRAY2_FOREACH(merger->sttReaderArr, sttReader) { + STsdbIter *iter; + STsdbIterConfig config = {0}; + + // data iter + config.type = TSDB_ITER_TYPE_STT; + config.sttReader = sttReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(merger->dataIterArr, iter); + TSDB_CHECK_CODE(code, lino, _exit); + + // tomb iter + config.type = TSDB_ITER_TYPE_STT_TOMB; + config.sttReader = sttReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(merger->tombIterArr, iter); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbIterMergerOpen(merger->dataIterArr, &merger->dataIterMerger, false); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbIterMergerOpen(merger->tombIterArr, &merger->tombIterMerger, true); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(vid, lino, code); + } + return code; +} + +static int32_t tsdbMergeFileSetBeginOpenWriter(SMerger *merger) { + int32_t code = 0; + int32_t lino = 0; + int32_t vid = TD_VID(merger->tsdb->pVnode); + + SDiskID did; + int32_t level = tsdbFidLevel(merger->ctx->fset->fid, &merger->tsdb->keepCfg, merger->ctx->now); + if (tfsAllocDisk(merger->tsdb->pVnode->pTfs, level, &did) < 0) { + code = TSDB_CODE_FS_NO_VALID_DISK; + TSDB_CHECK_CODE(code, lino, _exit); + } + tfsMkdirRecurAt(merger->tsdb->pVnode->pTfs, merger->tsdb->path, did); + SFSetWriterConfig config = { + .tsdb = merger->tsdb, + .toSttOnly = true, + .compactVersion = merger->compactVersion, + .minRow = merger->minRow, + .maxRow = merger->maxRow, + .szPage = merger->szPage, + .cmprAlg = merger->cmprAlg, + .fid = merger->ctx->fset->fid, + .cid = merger->cid, + .did = did, + .level = merger->ctx->level, + }; + + if (merger->ctx->toData) { + config.toSttOnly = false; + + for (int32_t ftype = 0; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (merger->ctx->fset->farr[ftype]) { + config.files[ftype].exist = true; + config.files[ftype].file = merger->ctx->fset->farr[ftype]->f[0]; + } else { + config.files[ftype].exist = false; + } + } + } + + code = tsdbFSetWriterOpen(&config, &merger->writer); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(vid, lino, code); + } + return code; +} + +static int32_t tsdbMergeFileSetBegin(SMerger *merger) { + int32_t code = 0; + int32_t lino = 0; + + ASSERT(TARRAY2_SIZE(merger->sttReaderArr) == 0); + ASSERT(TARRAY2_SIZE(merger->dataIterArr) == 0); + ASSERT(merger->dataIterMerger == NULL); + ASSERT(merger->writer == NULL); + + merger->ctx->tbid->suid = 0; + merger->ctx->tbid->uid = 0; + + // open reader + code = tsdbMergeFileSetBeginOpenReader(merger); + TSDB_CHECK_CODE(code, lino, _exit); + + // open iterator + code = tsdbMergeFileSetBeginOpenIter(merger); + TSDB_CHECK_CODE(code, lino, _exit); + + // open writer + code = tsdbMergeFileSetBeginOpenWriter(merger); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(merger->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbMergeFileSetEndCloseWriter(SMerger *merger) { + return tsdbFSetWriterClose(&merger->writer, 0, merger->fopArr); +} + +static int32_t tsdbMergeFileSetEndCloseIter(SMerger *merger) { + tsdbIterMergerClose(&merger->tombIterMerger); + TARRAY2_CLEAR(merger->tombIterArr, tsdbIterClose); + tsdbIterMergerClose(&merger->dataIterMerger); + TARRAY2_CLEAR(merger->dataIterArr, tsdbIterClose); + return 0; +} + +static int32_t tsdbMergeFileSetEndCloseReader(SMerger *merger) { + TARRAY2_CLEAR(merger->sttReaderArr, tsdbSttFileReaderClose); + return 0; +} + +static int32_t tsdbMergeFileSetEnd(SMerger *merger) { + int32_t code = 0; + int32_t lino = 0; + + code = tsdbMergeFileSetEndCloseWriter(merger); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbMergeFileSetEndCloseIter(merger); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbMergeFileSetEndCloseReader(merger); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(merger->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbMergeFileSet(SMerger *merger, STFileSet *fset) { + int32_t code = 0; + int32_t lino = 0; + + merger->ctx->fset = fset; + code = tsdbMergeFileSetBegin(merger); + TSDB_CHECK_CODE(code, lino, _exit); + + // data + SMetaInfo info; + SRowInfo *row; + merger->ctx->tbid->suid = 0; + merger->ctx->tbid->uid = 0; + while ((row = tsdbIterMergerGetData(merger->dataIterMerger)) != NULL) { + if (row->uid != merger->ctx->tbid->uid) { + merger->ctx->tbid->uid = row->uid; + merger->ctx->tbid->suid = row->suid; + + if (metaGetInfo(merger->tsdb->pVnode->pMeta, row->uid, &info, NULL) != 0) { + code = tsdbIterMergerSkipTableData(merger->dataIterMerger, merger->ctx->tbid); + TSDB_CHECK_CODE(code, lino, _exit); + continue; + } + } + + code = tsdbFSetWriteRow(merger->writer, row); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbIterMergerNext(merger->dataIterMerger); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // tomb + merger->ctx->tbid->suid = 0; + merger->ctx->tbid->uid = 0; + for (STombRecord *record; (record = tsdbIterMergerGetTombRecord(merger->tombIterMerger)) != NULL;) { + if (record->uid != merger->ctx->tbid->uid) { + merger->ctx->tbid->uid = record->uid; + merger->ctx->tbid->suid = record->suid; + + if (metaGetInfo(merger->tsdb->pVnode->pMeta, record->uid, &info, NULL) != 0) { + code = tsdbIterMergerSkipTableData(merger->tombIterMerger, merger->ctx->tbid); + TSDB_CHECK_CODE(code, lino, _exit); + continue; + } + } + code = tsdbFSetWriteTombRecord(merger->writer, record); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbIterMergerNext(merger->tombIterMerger); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbMergeFileSetEnd(merger); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(merger->tsdb->pVnode), __func__, lino, tstrerror(code)); + } else { + tsdbDebug("vgId:%d %s done, fid:%d", TD_VID(merger->tsdb->pVnode), __func__, fset->fid); + } + return code; +} + +static int32_t tsdbDoMerge(SMerger *merger) { + int32_t code = 0; + int32_t lino = 0; + + STFileSet *fset; + TARRAY2_FOREACH(merger->fsetArr, fset) { + if (TARRAY2_SIZE(fset->lvlArr) == 0) continue; + + SSttLvl *lvl = TARRAY2_FIRST(fset->lvlArr); + + if (lvl->level != 0 || TARRAY2_SIZE(lvl->fobjArr) < merger->sttTrigger) continue; + + if (!merger->ctx->opened) { + code = tsdbMergerOpen(merger); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbMergeFileSet(merger, fset); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (merger->ctx->opened) { + code = tsdbMergerClose(merger); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(merger->tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d %s done", TD_VID(merger->tsdb->pVnode), __func__); + } + return code; +} + +int32_t tsdbMerge(void *arg) { + int32_t code = 0; + int32_t lino = 0; + STsdb *tsdb = (STsdb *)arg; + + SMerger merger[1] = {{ + .tsdb = tsdb, + .sttTrigger = tsdb->pVnode->config.sttTrigger, + }}; + + ASSERT(merger->sttTrigger > 1); + + code = tsdbFSCreateCopySnapshot(tsdb->pFS, &merger->fsetArr); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDoMerge(merger); + TSDB_CHECK_CODE(code, lino, _exit); + + tsdbFSDestroyCopySnapshot(&merger->fsetArr); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } else if (merger->ctx->opened) { + tsdbDebug("vgId:%d %s done", TD_VID(tsdb->pVnode), __func__); + } + return code; +} diff --git a/source/dnode/vnode/src/tsdb/tsdbMerge.h b/source/dnode/vnode/src/tsdb/tsdbMerge.h new file mode 100644 index 0000000000..69d802fd27 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbMerge.h @@ -0,0 +1,40 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbDataFileRW.h" +#include "tsdbFS2.h" +#include "tsdbFSetRW.h" +#include "tsdbIter.h" +#include "tsdbSttFileRW.h" +#include "tsdbUtil2.h" + +#ifndef _TD_TSDB_MERGE_H_ +#define _TD_TSDB_MERGE_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +/* Exposed Handle */ + +/* Exposed APIs */ + +/* Exposed Structs */ + +#ifdef __cplusplus +} +#endif + +#endif /*_TD_TSDB_MERGE_H_*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbMergeTree.c b/source/dnode/vnode/src/tsdb/tsdbMergeTree.c index 79f4a17f65..b79b5c908d 100644 --- a/source/dnode/vnode/src/tsdb/tsdbMergeTree.c +++ b/source/dnode/vnode/src/tsdb/tsdbMergeTree.c @@ -14,6 +14,11 @@ */ #include "tsdb.h" +#include "tsdbFSet2.h" +#include "tsdbReadUtil.h" +#include "tsdbSttFileRW.h" + +static void tLDataIterClose2(SLDataIter *pIter); // SLDataIter ================================================= SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols, @@ -24,8 +29,6 @@ SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, return NULL; } - pLoadInfo->numOfStt = numOfSttTrigger; - for (int32_t i = 0; i < numOfSttTrigger; ++i) { pLoadInfo[i].blockIndex[0] = -1; pLoadInfo[i].blockIndex[1] = -1; @@ -50,8 +53,37 @@ SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, return pLoadInfo; } +SSttBlockLoadInfo *tCreateOneLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols) { + SSttBlockLoadInfo *pLoadInfo = taosMemoryCalloc(1, sizeof(SSttBlockLoadInfo)); + if (pLoadInfo == NULL) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + return NULL; + } + + pLoadInfo->blockIndex[0] = -1; + pLoadInfo->blockIndex[1] = -1; + pLoadInfo->currentLoadBlockIndex = 1; + + int32_t code = tBlockDataCreate(&pLoadInfo->blockData[0]); + if (code) { + terrno = code; + } + + code = tBlockDataCreate(&pLoadInfo->blockData[1]); + if (code) { + terrno = code; + } + + pLoadInfo->aSttBlk = taosArrayInit(4, sizeof(SSttBlk)); + pLoadInfo->pSchema = pSchema; + pLoadInfo->colIds = colList; + pLoadInfo->numOfCols = numOfCols; + + return pLoadInfo; +} + void resetLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo) { - for (int32_t i = 0; i < pLoadInfo->numOfStt; ++i) { + for (int32_t i = 0; i < 1; ++i) { pLoadInfo[i].currentLoadBlockIndex = 1; pLoadInfo[i].blockIndex[0] = -1; pLoadInfo[i].blockIndex[1] = -1; @@ -65,18 +97,24 @@ void resetLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo) { } void getLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo, int64_t *blocks, double *el) { - for (int32_t i = 0; i < pLoadInfo->numOfStt; ++i) { + for (int32_t i = 0; i < 1; ++i) { *el += pLoadInfo[i].elapsedTime; *blocks += pLoadInfo[i].loadBlocks; } } +static void freeTombBlock(void *param) { + STombBlock **pTombBlock = (STombBlock **)param; + tTombBlockDestroy(*pTombBlock); + taosMemoryFree(*pTombBlock); +} + void *destroyLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo) { if (pLoadInfo == NULL) { return NULL; } - for (int32_t i = 0; i < pLoadInfo->numOfStt; ++i) { + for (int32_t i = 0; i < 1; ++i) { pLoadInfo[i].currentLoadBlockIndex = 1; pLoadInfo[i].blockIndex[0] = -1; pLoadInfo[i].blockIndex[1] = -1; @@ -91,6 +129,33 @@ void *destroyLastBlockLoadInfo(SSttBlockLoadInfo *pLoadInfo) { return NULL; } +static void destroyLDataIter(SLDataIter *pIter) { + tLDataIterClose2(pIter); + destroyLastBlockLoadInfo(pIter->pBlockLoadInfo); + taosMemoryFree(pIter); +} + +void *destroySttBlockReader(SArray *pLDataIterArray, int64_t *blocks, double *el) { + if (pLDataIterArray == NULL) { + return NULL; + } + + int32_t numOfLevel = taosArrayGetSize(pLDataIterArray); + for (int32_t i = 0; i < numOfLevel; ++i) { + SArray *pList = taosArrayGetP(pLDataIterArray, i); + for (int32_t j = 0; j < taosArrayGetSize(pList); ++j) { + SLDataIter *pIter = taosArrayGetP(pList, j); + *el += pIter->pBlockLoadInfo->elapsedTime; + *blocks += pIter->pBlockLoadInfo->loadBlocks; + destroyLDataIter(pIter); + } + taosArrayDestroy(pList); + } + + taosArrayDestroy(pLDataIterArray); + return NULL; +} + static SBlockData *loadLastBlock(SLDataIter *pIter, const char *idStr) { int32_t code = 0; @@ -122,20 +187,8 @@ static SBlockData *loadLastBlock(SLDataIter *pIter, const char *idStr) { int64_t st = taosGetTimestampUs(); SBlockData *pBlock = &pInfo->blockData[pInfo->currentLoadBlockIndex]; - - TABLEID id = {0}; - if (pIter->pSttBlk->suid != 0) { - id.suid = pIter->pSttBlk->suid; - } else { - id.uid = pIter->uid; - } - - code = tBlockDataInit(pBlock, &id, pInfo->pSchema, pInfo->colIds, pInfo->numOfCols); - if (code != TSDB_CODE_SUCCESS) { - goto _exit; - } - - code = tsdbReadSttBlock(pIter->pReader, pIter->iStt, pIter->pSttBlk, pBlock); + code = tsdbSttFileReadBlockDataByColumn(pIter->pReader, pIter->pSttBlk, pBlock, pInfo->pSchema, &pInfo->colIds[1], + pInfo->numOfCols - 1); if (code != TSDB_CODE_SUCCESS) { goto _exit; } @@ -255,74 +308,153 @@ static int32_t binarySearchForStartRowIndex(uint64_t *uidList, int32_t num, uint int32_t tLDataIterOpen(struct SLDataIter *pIter, SDataFReader *pReader, int32_t iStt, int8_t backward, uint64_t suid, uint64_t uid, STimeWindow *pTimeWindow, SVersionRange *pRange, SSttBlockLoadInfo *pBlockLoadInfo, const char *idStr, bool strictTimeRange) { + return 0; +} + +static int32_t extractSttBlockInfo(SLDataIter *pIter, const TSttBlkArray *pArray, SSttBlockLoadInfo *pBlockLoadInfo, + uint64_t suid) { + if (TARRAY2_SIZE(pArray) <= 0) { + return TSDB_CODE_SUCCESS; + } + + SSttBlk *pStart = &pArray->data[0]; + SSttBlk *pEnd = &pArray->data[TARRAY2_SIZE(pArray) - 1]; + + // all identical + if (pStart->suid == pEnd->suid) { + if (pStart->suid != suid) { // no qualified stt block existed + taosArrayClear(pBlockLoadInfo->aSttBlk); + pIter->iSttBlk = -1; + return TSDB_CODE_SUCCESS; + } else { // all blocks are qualified + taosArrayClear(pBlockLoadInfo->aSttBlk); + taosArrayAddBatch(pBlockLoadInfo->aSttBlk, pArray->data, pArray->size); + } + } else { + SArray *pTmp = taosArrayInit(TARRAY2_SIZE(pArray), sizeof(SSttBlk)); + for (int32_t i = 0; i < TARRAY2_SIZE(pArray); ++i) { + SSttBlk *p = &pArray->data[i]; + if (p->suid < suid) { + continue; + } + + if (p->suid == suid) { + taosArrayPush(pTmp, p); + } else if (p->suid > suid) { + break; + } + } + + taosArrayDestroy(pBlockLoadInfo->aSttBlk); + pBlockLoadInfo->aSttBlk = pTmp; + } + + return TSDB_CODE_SUCCESS; +} + +static int32_t uidComparFn(const void* p1, const void* p2) { + const uint64_t* uid1 = p1; + const uint64_t* uid2 = p2; + return (*uid1) - (*uid2); +} + +static bool existsFromSttBlkStatis(const TStatisBlkArray *pStatisBlkArray, uint64_t suid, uint64_t uid, + SSttFileReader *pReader) { + if (TARRAY2_SIZE(pStatisBlkArray) <= 0) { + return true; + } + + int32_t i = 0; + for (i = 0; i < TARRAY2_SIZE(pStatisBlkArray); ++i) { + SStatisBlk *p = &pStatisBlkArray->data[i]; + if (p->minTbid.suid <= suid && p->maxTbid.suid >= suid) { + break; + } + } + +// for (; i < TARRAY2_SIZE(pStatisBlkArray); ++i) { +// SStatisBlk *p = &pStatisBlkArray->data[i]; +// if (p->minTbid.uid <= uid && p->maxTbid.uid >= uid) { +// break; +// } +// +// if (p->maxTbid.uid < uid) { +// break; +// } +// } + + if (i >= TARRAY2_SIZE(pStatisBlkArray)) { + return false; + } + + SStatisBlk *p = &pStatisBlkArray->data[i]; + STbStatisBlock block = {0}; + tsdbSttFileReadStatisBlock(pReader, p, &block); + + int32_t index = tarray2SearchIdx(block.uid, &uid, sizeof(int64_t), uidComparFn, TD_EQ); + tStatisBlockDestroy(&block); + + return (index != -1); +} + +int32_t tLDataIterOpen2(struct SLDataIter *pIter, SSttFileReader *pSttFileReader, int32_t iStt, int8_t backward, + uint64_t suid, uint64_t uid, STimeWindow *pTimeWindow, SVersionRange *pRange, + SSttBlockLoadInfo *pBlockLoadInfo, const char *idStr, bool strictTimeRange, + _load_tomb_fn loadTombFn, void *pReader1) { int32_t code = TSDB_CODE_SUCCESS; pIter->uid = uid; - pIter->pReader = pReader; pIter->iStt = iStt; pIter->backward = backward; pIter->verRange.minVer = pRange->minVer; pIter->verRange.maxVer = pRange->maxVer; pIter->timeWindow.skey = pTimeWindow->skey; pIter->timeWindow.ekey = pTimeWindow->ekey; - + pIter->pReader = pSttFileReader; pIter->pBlockLoadInfo = pBlockLoadInfo; if (!pBlockLoadInfo->sttBlockLoaded) { int64_t st = taosGetTimestampUs(); + + const TSttBlkArray*pSttBlkArray = NULL; pBlockLoadInfo->sttBlockLoaded = true; - code = tsdbReadSttBlk(pReader, iStt, pBlockLoadInfo->aSttBlk); - if (code) { + // load the stt block info for each stt-block + code = tsdbSttFileReadSttBlk(pIter->pReader, &pSttBlkArray); + if (code != TSDB_CODE_SUCCESS) { + tsdbError("load stt blk failed, code:%s, %s", tstrerror(code), idStr); return code; } - // only apply to the child tables, ordinary tables will not incur this filter procedure. - size_t size = taosArrayGetSize(pBlockLoadInfo->aSttBlk); - - if (size >= 1) { - SSttBlk *pStart = taosArrayGet(pBlockLoadInfo->aSttBlk, 0); - SSttBlk *pEnd = taosArrayGet(pBlockLoadInfo->aSttBlk, size - 1); - - // all identical - if (pStart->suid == pEnd->suid) { - if (pStart->suid != suid) { - // no qualified stt block existed - taosArrayClear(pBlockLoadInfo->aSttBlk); - - pIter->iSttBlk = -1; - double el = (taosGetTimestampUs() - st) / 1000.0; - tsdbDebug("load the last file info completed, elapsed time:%.2fms, %s", el, idStr); - return code; - } - } else { - SArray *pTmp = taosArrayInit(size, sizeof(SSttBlk)); - for (int32_t i = 0; i < size; ++i) { - SSttBlk *p = taosArrayGet(pBlockLoadInfo->aSttBlk, i); - uint64_t s = p->suid; - if (s < suid) { - continue; - } - - if (s == suid) { - taosArrayPush(pTmp, p); - } else if (s > suid) { - break; - } - } - - taosArrayDestroy(pBlockLoadInfo->aSttBlk); - pBlockLoadInfo->aSttBlk = pTmp; - } + code = extractSttBlockInfo(pIter, pSttBlkArray, pBlockLoadInfo, suid); + if (code != TSDB_CODE_SUCCESS) { + tsdbError("load stt block info failed, code:%s, %s", tstrerror(code), idStr); + return code; } + // load stt blocks statis for all stt-blocks, to decide if the data of queried table exists in current stt file + code = tsdbSttFileReadStatisBlk(pIter->pReader, (const TStatisBlkArray **)&pBlockLoadInfo->pSttStatisBlkArray); + if (code != TSDB_CODE_SUCCESS) { + tsdbError("failed to load stt block statistics, code:%s, %s", tstrerror(code), idStr); + return code; + } + + code = loadTombFn(pReader1, pIter->pReader, pIter->pBlockLoadInfo); + double el = (taosGetTimestampUs() - st) / 1000.0; - tsdbDebug("load the last file info completed, elapsed time:%.2fms, %s", el, idStr); + tsdbDebug("load the stt file info completed, elapsed time:%.2fms, %s", el, idStr); } - size_t size = taosArrayGetSize(pBlockLoadInfo->aSttBlk); +// bool exists = existsFromSttBlkStatis(pBlockLoadInfo->pSttStatisBlkArray, suid, uid, pIter->pReader); +// if (!exists) { +// pIter->iSttBlk = -1; +// pIter->pSttBlk = NULL; +// return TSDB_CODE_SUCCESS; +// } - // find the start block + // find the start block, actually we could load the position to avoid repeatly searching for the start position when + // the skey is updated. + size_t size = taosArrayGetSize(pBlockLoadInfo->aSttBlk); pIter->iSttBlk = binarySearchForStartBlock(pBlockLoadInfo->aSttBlk->pData, size, uid, backward); if (pIter->iSttBlk != -1) { pIter->pSttBlk = taosArrayGet(pBlockLoadInfo->aSttBlk, pIter->iSttBlk); @@ -343,7 +475,10 @@ int32_t tLDataIterOpen(struct SLDataIter *pIter, SDataFReader *pReader, int32_t return code; } -void tLDataIterClose(SLDataIter *pIter) { /*taosMemoryFree(pIter); */} +void tLDataIterClose2(SLDataIter *pIter) { + tsdbSttFileReaderClose(&pIter->pReader); + pIter->pReader = NULL; +} void tLDataIterNextBlock(SLDataIter *pIter, const char *idStr) { int32_t step = pIter->backward ? -1 : 1; @@ -395,25 +530,23 @@ void tLDataIterNextBlock(SLDataIter *pIter, const char *idStr) { if (index != -1) { pIter->iSttBlk = index; pIter->pSttBlk = (SSttBlk *)taosArrayGet(pIter->pBlockLoadInfo->aSttBlk, pIter->iSttBlk); - tsdbDebug("try next last file block:%d from %d, trigger by uid:%" PRIu64 ", file index:%d, %s", pIter->iSttBlk, - oldIndex, pIter->uid, pIter->iStt, idStr); + tsdbDebug("try next last file block:%d from stt fileIdx:%d, trigger by uid:%" PRIu64 ", file index:%d, %s", + pIter->iSttBlk, oldIndex, pIter->uid, pIter->iStt, idStr); } else { tsdbDebug("no more last block qualified, uid:%" PRIu64 ", file index:%d, %s", pIter->uid, oldIndex, idStr); } } static void findNextValidRow(SLDataIter *pIter, const char *idStr) { - int32_t step = pIter->backward ? -1 : 1; - bool hasVal = false; + int32_t step = pIter->backward ? -1 : 1; int32_t i = pIter->iRow; - SBlockData *pBlockData = loadLastBlock(pIter, idStr); + SBlockData *pData = loadLastBlock(pIter, idStr); // mostly we only need to find the start position for a given table - if ((((i == 0) && (!pIter->backward)) || (i == pBlockData->nRow - 1 && pIter->backward)) && - pBlockData->aUid != NULL) { - i = binarySearchForStartRowIndex((uint64_t *)pBlockData->aUid, pBlockData->nRow, pIter->uid, pIter->backward); + if ((((i == 0) && (!pIter->backward)) || (i == pData->nRow - 1 && pIter->backward)) && pData->aUid != NULL) { + i = binarySearchForStartRowIndex((uint64_t *)pData->aUid, pData->nRow, pIter->uid, pIter->backward); if (i == -1) { tsdbDebug("failed to find the data in pBlockData, uid:%" PRIu64 " , %s", pIter->uid, idStr); pIter->iRow = -1; @@ -421,20 +554,20 @@ static void findNextValidRow(SLDataIter *pIter, const char *idStr) { } } - for (; i < pBlockData->nRow && i >= 0; i += step) { - if (pBlockData->aUid != NULL) { + for (; i < pData->nRow && i >= 0; i += step) { + if (pData->aUid != NULL) { if (!pIter->backward) { - if (pBlockData->aUid[i] > pIter->uid) { + if (pData->aUid[i] > pIter->uid) { break; } } else { - if (pBlockData->aUid[i] < pIter->uid) { + if (pData->aUid[i] < pIter->uid) { break; } } } - int64_t ts = pBlockData->aTSKEY[i]; + int64_t ts = pData->aTSKEY[i]; if (!pIter->backward) { // asc if (ts > pIter->timeWindow.ekey) { // no more data break; @@ -449,7 +582,7 @@ static void findNextValidRow(SLDataIter *pIter, const char *idStr) { } } - int64_t ver = pBlockData->aVersion[i]; + int64_t ver = pData->aVersion[i]; if (ver < pIter->verRange.minVer) { continue; } @@ -485,7 +618,6 @@ bool tLDataIterNextRow(SLDataIter *pIter, const char *idStr) { while (1) { bool skipBlock = false; - findNextValidRow(pIter, idStr); if (pIter->pBlockLoadInfo->checkRemainingRow) { @@ -570,7 +702,7 @@ static FORCE_INLINE int32_t tLDataIterDescCmprFn(const SRBTreeNode *p1, const SR int32_t tMergeTreeOpen(SMergeTree *pMTree, int8_t backward, SDataFReader *pFReader, uint64_t suid, uint64_t uid, STimeWindow *pTimeWindow, SVersionRange *pVerRange, SSttBlockLoadInfo *pBlockLoadInfo, - bool destroyLoadInfo, const char *idStr, bool strictTimeRange, SLDataIter* pLDataIter) { + bool destroyLoadInfo, const char *idStr, bool strictTimeRange, SLDataIter *pLDataIter) { int32_t code = TSDB_CODE_SUCCESS; pMTree->backward = backward; @@ -612,6 +744,95 @@ _end: return code; } +int32_t tMergeTreeOpen2(SMergeTree *pMTree, SMergeTreeConf *pConf) { + int32_t code = TSDB_CODE_SUCCESS; + + pMTree->pIter = NULL; + pMTree->backward = pConf->backward; + pMTree->idStr = pConf->idstr; + + if (!pMTree->backward) { // asc + tRBTreeCreate(&pMTree->rbt, tLDataIterCmprFn); + } else { // desc + tRBTreeCreate(&pMTree->rbt, tLDataIterDescCmprFn); + } + + pMTree->ignoreEarlierTs = false; + + // todo handle other level of stt files, here only deal with the first level stt + int32_t size = ((STFileSet *)pConf->pCurrentFileset)->lvlArr->size; + if (size == 0) { + goto _end; + } + + // add the list/iter placeholder + while (taosArrayGetSize(pConf->pSttFileBlockIterArray) < size) { + SArray *pList = taosArrayInit(4, POINTER_BYTES); + taosArrayPush(pConf->pSttFileBlockIterArray, &pList); + } + + for (int32_t j = 0; j < size; ++j) { + SSttLvl *pSttLevel = ((STFileSet *)pConf->pCurrentFileset)->lvlArr->data[j]; + ASSERT(pSttLevel->level == j); + + SArray *pList = taosArrayGetP(pConf->pSttFileBlockIterArray, j); + int32_t numOfIter = taosArrayGetSize(pList); + + if (numOfIter < TARRAY2_SIZE(pSttLevel->fobjArr)) { + int32_t inc = TARRAY2_SIZE(pSttLevel->fobjArr) - numOfIter; + for (int32_t k = 0; k < inc; ++k) { + SLDataIter *pIter = taosMemoryCalloc(1, sizeof(SLDataIter)); + taosArrayPush(pList, &pIter); + } + } + + for (int32_t i = 0; i < TARRAY2_SIZE(pSttLevel->fobjArr); ++i) { // open all last file + SLDataIter *pIter = taosArrayGetP(pList, i); + + SSttFileReader *pSttFileReader = pIter->pReader; + SSttBlockLoadInfo *pLoadInfo = pIter->pBlockLoadInfo; + + // open stt file reader if not + if (pSttFileReader == NULL) { + SSttFileReaderConfig conf = {.tsdb = pConf->pTsdb, .szPage = pConf->pTsdb->pVnode->config.szPage}; + conf.file[0] = *pSttLevel->fobjArr->data[i]->f; + + code = tsdbSttFileReaderOpen(pSttLevel->fobjArr->data[i]->fname, &conf, &pSttFileReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + if (pLoadInfo == NULL) { + pLoadInfo = tCreateOneLastBlockLoadInfo(pConf->pSchema, pConf->pCols, pConf->numOfCols); + } + + memset(pIter, 0, sizeof(SLDataIter)); + code = tLDataIterOpen2(pIter, pSttFileReader, i, pMTree->backward, pConf->suid, pConf->uid, &pConf->timewindow, + &pConf->verRange, pLoadInfo, pMTree->idStr, pConf->strictTimeRange, pConf->loadTombFn, + pConf->pReader); + if (code != TSDB_CODE_SUCCESS) { + goto _end; + } + + bool hasVal = tLDataIterNextRow(pIter, pMTree->idStr); + if (hasVal) { + tMergeTreeAddIter(pMTree, pIter); + } else { + if (!pMTree->ignoreEarlierTs) { + pMTree->ignoreEarlierTs = pIter->ignoreEarlierTs; + } + } + } + } + + return code; + +_end: + tMergeTreeClose(pMTree); + return code; +} + void tMergeTreeAddIter(SMergeTree *pMTree, SLDataIter *pIter) { tRBTreePut(&pMTree->rbt, (SRBTreeNode *)pIter); } bool tMergeTreeIgnoreEarlierTs(SMergeTree *pMTree) { return pMTree->ignoreEarlierTs; } diff --git a/source/dnode/vnode/src/tsdb/tsdbOpen.c b/source/dnode/vnode/src/tsdb/tsdbOpen.c index 8901f64459..c684ad5184 100644 --- a/source/dnode/vnode/src/tsdb/tsdbOpen.c +++ b/source/dnode/vnode/src/tsdb/tsdbOpen.c @@ -14,6 +14,7 @@ */ #include "tsdb.h" +#include "tsdbFS2.h" int32_t tsdbSetKeepCfg(STsdb *pTsdb, STsdbCfg *pCfg) { STsdbKeepCfg *pKeepCfg = &pTsdb->keepCfg; @@ -66,7 +67,7 @@ int tsdbOpen(SVnode *pVnode, STsdb **ppTsdb, const char *dir, STsdbKeepCfg *pKee } // open tsdb - if (tsdbFSOpen(pTsdb, rollback) < 0) { + if (tsdbOpenFS(pTsdb, &pTsdb->pFS, rollback) < 0) { goto _err; } @@ -94,7 +95,7 @@ int tsdbClose(STsdb **pTsdb) { taosThreadRwlockDestroy(&(*pTsdb)->rwLock); - tsdbFSClose(*pTsdb); + tsdbCloseFS(&(*pTsdb)->pFS); tsdbCloseCache(*pTsdb); taosMemoryFreeClear(*pTsdb); } diff --git a/source/dnode/vnode/src/tsdb/tsdbRead.c b/source/dnode/vnode/src/tsdb/tsdbRead.c index cfeb1288d4..f4109d5b99 100644 --- a/source/dnode/vnode/src/tsdb/tsdbRead.c +++ b/source/dnode/vnode/src/tsdb/tsdbRead.c @@ -17,7 +17,7 @@ #include "tsdb.h" #include "tsimplehash.h" -#define ASCENDING_TRAVERSE(o) (o == TSDB_ORDER_ASC) +#define ASCENDING_TRAVERSE(o) (o == TSDB_ORDER_ASC) #define getCurrentKeyInLastBlock(_r) ((_r)->currentKey) typedef enum { @@ -30,12 +30,12 @@ typedef enum { EXTERNAL_ROWS_MAIN = 0x2, EXTERNAL_ROWS_NEXT = 0x3, } EContentData; - +/* typedef enum { READ_MODE_COUNT_ONLY = 0x1, READ_MODE_ALL, } EReadMode; - +*/ typedef struct { STbDataIter* iter; int32_t index; @@ -166,7 +166,7 @@ typedef struct SReaderStatus { SDataBlockIter blockIter; SLDataIter* pLDataIter; SRowMerger merger; - SColumnInfoData* pPrimaryTsCol; // primary time stamp output col info data + SColumnInfoData* pPrimaryTsCol; // primary time stamp output col info data } SReaderStatus; typedef struct SBlockInfoBuf { @@ -292,7 +292,7 @@ static int32_t updateBlockSMAInfo(STSchema* pSchema, SBlockLoadSuppInfo* pSupInf if (j < pSupInfo->numOfCols && PRIMARYKEY_TIMESTAMP_COL_ID == pSupInfo->colId[j]) { j += 1; } - + while (i < pSchema->numOfCols && j < pSupInfo->numOfCols) { STColumn* pTCol = &pSchema->columns[i]; if (pTCol->colId == pSupInfo->colId[j]) { @@ -410,7 +410,7 @@ static int32_t uidComparFunc(const void* p1, const void* p2) { // NOTE: speedup the whole processing by preparing the buffer for STableBlockScanInfo in batch model static SSHashObj* createDataBlockScanInfo(STsdbReader* pTsdbReader, SBlockInfoBuf* pBuf, const STableKeyInfo* idList, - STableUidList* pUidList, int32_t numOfTables) { + STableUidList* pUidList, int32_t numOfTables) { // allocate buffer in order to load data blocks from file // todo use simple hash instead, optimize the memory consumption SSHashObj* pTableMap = tSimpleHashInit(numOfTables, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT)); @@ -461,7 +461,7 @@ static SSHashObj* createDataBlockScanInfo(STsdbReader* pTsdbReader, SBlockInfoBu } static void resetAllDataBlockScanInfo(SSHashObj* pTableMap, int64_t ts, int32_t step) { - void *p = NULL; + void* p = NULL; int32_t iter = 0; while ((p = tSimpleHashIterate(pTableMap, p, &iter)) != NULL) { @@ -505,7 +505,7 @@ static void clearBlockScanInfo(STableBlockScanInfo* p) { } static void destroyAllBlockScanInfo(SSHashObj* pTableMap) { - void* p = NULL; + void* p = NULL; int32_t iter = 0; while ((p = tSimpleHashIterate(pTableMap, p, &iter)) != NULL) { @@ -743,7 +743,8 @@ void tsdbReleaseDataBlock(STsdbReader* pReader) { } } -static int32_t initResBlockInfo(SResultBlockInfo* pResBlockInfo, int64_t capacity, SSDataBlock* pResBlock, SQueryTableDataCond* pCond) { +static int32_t initResBlockInfo(SResultBlockInfo* pResBlockInfo, int64_t capacity, SSDataBlock* pResBlock, + SQueryTableDataCond* pCond) { pResBlockInfo->capacity = capacity; pResBlockInfo->pResBlock = pResBlock; terrno = 0; @@ -921,9 +922,9 @@ static void cleanupTableScanInfo(SReaderStatus* pStatus) { return; } - SSHashObj* pTableMap = pStatus->pTableMap; + SSHashObj* pTableMap = pStatus->pTableMap; STableBlockScanInfo** px = NULL; - int32_t iter = 0; + int32_t iter = 0; while (1) { px = tSimpleHashIterate(pTableMap, px, &iter); @@ -937,9 +938,10 @@ static void cleanupTableScanInfo(SReaderStatus* pStatus) { pStatus->mapDataCleaned = true; } -static int32_t doLoadFileBlock(STsdbReader* pReader, SArray* pIndexList, SBlockNumber* pBlockNum, SArray* pTableScanInfoList) { - size_t sizeInDisk = 0; - size_t numOfTables = taosArrayGetSize(pIndexList); +static int32_t doLoadFileBlock(STsdbReader* pReader, SArray* pIndexList, SBlockNumber* pBlockNum, + SArray* pTableScanInfoList) { + size_t sizeInDisk = 0; + size_t numOfTables = taosArrayGetSize(pIndexList); int64_t st = taosGetTimestampUs(); cleanupTableScanInfo(&pReader->status); @@ -1125,18 +1127,18 @@ static int32_t getEndPosInDataBlock(STsdbReader* pReader, SBlockData* pBlockData endPos = doBinarySearchKey(pBlockData->aTSKEY, pBlock->nRow, pos, key, pReader->order); } - if ((pReader->verRange.maxVer >= pBlock->minVer && pReader->verRange.maxVer < pBlock->maxVer)|| + if ((pReader->verRange.maxVer >= pBlock->minVer && pReader->verRange.maxVer < pBlock->maxVer) || (pReader->verRange.minVer <= pBlock->maxVer && pReader->verRange.minVer > pBlock->minVer)) { int32_t i = endPos; if (asc) { - for(; i >= 0; --i) { + for (; i >= 0; --i) { if (pBlockData->aVersion[i] <= pReader->verRange.maxVer) { break; } } } else { - for(; i < pBlock->nRow; ++i) { + for (; i < pBlock->nRow; ++i) { if (pBlockData->aVersion[i] >= pReader->verRange.minVer) { break; } @@ -1309,17 +1311,17 @@ static int32_t copyBlockDataToSDataBlock(STsdbReader* pReader) { ASSERT(pReader->verRange.minVer <= pBlock->maxVer && pReader->verRange.maxVer >= pBlock->minVer); // find the appropriate start position that satisfies the version requirement. - if ((pReader->verRange.maxVer >= pBlock->minVer && pReader->verRange.maxVer < pBlock->maxVer)|| + if ((pReader->verRange.maxVer >= pBlock->minVer && pReader->verRange.maxVer < pBlock->maxVer) || (pReader->verRange.minVer <= pBlock->maxVer && pReader->verRange.minVer > pBlock->minVer)) { int32_t i = pDumpInfo->rowIndex; if (asc) { - for(; i < pBlock->nRow; ++i) { + for (; i < pBlock->nRow; ++i) { if (pBlockData->aVersion[i] >= pReader->verRange.minVer) { break; } } } else { - for(; i >= 0; --i) { + for (; i >= 0; --i) { if (pBlockData->aVersion[i] <= pReader->verRange.maxVer) { break; } @@ -1562,7 +1564,8 @@ static int32_t doSetCurrentBlock(SDataBlockIter* pBlockIter, const char* idStr) return TSDB_CODE_SUCCESS; } -static int32_t initBlockIterator(STsdbReader* pReader, SDataBlockIter* pBlockIter, int32_t numOfBlocks, SArray* pTableList) { +static int32_t initBlockIterator(STsdbReader* pReader, SDataBlockIter* pBlockIter, int32_t numOfBlocks, + SArray* pTableList) { bool asc = ASCENDING_TRAVERSE(pReader->order); SBlockOrderSupporter sup = {0}; @@ -1967,13 +1970,14 @@ static bool nextRowFromLastBlocks(SLastBlockReader* pLastBlockReader, STableBloc } TSDBROW* pRow = tMergeTreeGetRow(&pLastBlockReader->mergeTree); - int64_t key = pRow->pBlockData->aTSKEY[pRow->iRow]; - int64_t ver = pRow->pBlockData->aVersion[pRow->iRow]; + int64_t key = pRow->pBlockData->aTSKEY[pRow->iRow]; + int64_t ver = pRow->pBlockData->aVersion[pRow->iRow]; pLastBlockReader->currentKey = key; pScanInfo->lastKeyInStt = key; - if (!hasBeenDropped(pScanInfo->delSkyline, &pScanInfo->lastBlockDelIndex, key, ver, pLastBlockReader->order, pVerRange)) { + if (!hasBeenDropped(pScanInfo->delSkyline, &pScanInfo->lastBlockDelIndex, key, ver, pLastBlockReader->order, + pVerRange)) { return true; } } @@ -2030,7 +2034,7 @@ static FORCE_INLINE STSchema* doGetSchemaForTSRow(int32_t sversion, STsdbReader* } STSchema* ptr = NULL; - int32_t code = metaGetTbTSchemaEx(pReader->pTsdb->pVnode->pMeta, pReader->suid, uid, sversion, &ptr); + int32_t code = metaGetTbTSchemaEx(pReader->pTsdb->pVnode->pMeta, pReader->suid, uid, sversion, &ptr); if (code != TSDB_CODE_SUCCESS) { terrno = code; return NULL; @@ -2153,7 +2157,7 @@ static int32_t doMergeBufAndFileRows(STsdbReader* pReader, STableBlockScanInfo* return terrno; } - int32_t code = tsdbRowMergerAdd(pMerger, pRow, pSchema); + int32_t code = tsdbRowMergerAdd(pMerger, pRow, pSchema); if (code != TSDB_CODE_SUCCESS) { return code; } @@ -2208,7 +2212,7 @@ static int32_t doMergeBufAndFileRows(STsdbReader* pReader, STableBlockScanInfo* static int32_t doMergeFileBlockAndLastBlock(SLastBlockReader* pLastBlockReader, STsdbReader* pReader, STableBlockScanInfo* pBlockScanInfo, SBlockData* pBlockData, bool mergeBlockData) { - SRowMerger* pMerger = &pReader->status.merger; + SRowMerger* pMerger = &pReader->status.merger; SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; int64_t tsLastBlock = getCurrentKeyInLastBlock(pLastBlockReader); @@ -2218,9 +2222,10 @@ static int32_t doMergeFileBlockAndLastBlock(SLastBlockReader* pLastBlockReader, TSDBROW* pRow = tMergeTreeGetRow(&pLastBlockReader->mergeTree); // create local variable to hold the row value - TSDBROW fRow = {.iRow = pRow->iRow, .type = TSDBROW_COL_FMT, .pBlockData = pRow->pBlockData}; + TSDBROW fRow = {.iRow = pRow->iRow, .type = TSDBROW_COL_FMT, .pBlockData = pRow->pBlockData}; - tsdbTrace("fRow ptr:%p, %d, uid:%" PRIu64 ", %s", pRow->pBlockData, pRow->iRow, pLastBlockReader->uid, pReader->idStr); + tsdbTrace("fRow ptr:%p, %d, uid:%" PRIu64 ", %s", pRow->pBlockData, pRow->iRow, pLastBlockReader->uid, + pReader->idStr); // only last block exists if ((!mergeBlockData) || (tsLastBlock != pBlockData->aTSKEY[pDumpInfo->rowIndex])) { @@ -2240,7 +2245,8 @@ static int32_t doMergeFileBlockAndLastBlock(SLastBlockReader* pLastBlockReader, TSDBROW* pRow1 = tMergeTreeGetRow(&pLastBlockReader->mergeTree); tsdbRowMergerAdd(pMerger, pRow1, NULL); - doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, tsLastBlock, pMerger, &pReader->verRange, pReader->idStr); + doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, tsLastBlock, pMerger, &pReader->verRange, + pReader->idStr); code = tsdbRowMergerGetRow(pMerger, &pTSRow); if (code != TSDB_CODE_SUCCESS) { @@ -2290,7 +2296,7 @@ static int32_t doMergeFileBlockAndLastBlock(SLastBlockReader* pLastBlockReader, static int32_t mergeFileBlockAndLastBlock(STsdbReader* pReader, SLastBlockReader* pLastBlockReader, int64_t key, STableBlockScanInfo* pBlockScanInfo, SBlockData* pBlockData) { SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; - SRowMerger* pMerger = &pReader->status.merger; + SRowMerger* pMerger = &pReader->status.merger; // merge is not initialized yet, due to the fact that the pReader->pSchema is not initialized if (pMerger->pArray == NULL) { @@ -2316,7 +2322,7 @@ static int32_t mergeFileBlockAndLastBlock(STsdbReader* pReader, SLastBlockReader if (key < ts) { // imem, mem are all empty, file blocks (data blocks and last block) exist return mergeRowsInFileBlocks(pBlockData, pBlockScanInfo, key, pReader); } else if (key == ts) { - SRow* pTSRow = NULL; + SRow* pTSRow = NULL; int32_t code = tsdbRowMergerAdd(pMerger, &fRow, pReader->pSchema); if (code != TSDB_CODE_SUCCESS) { return code; @@ -2723,7 +2729,7 @@ int32_t mergeRowsInFileBlocks(SBlockData* pBlockData, STableBlockScanInfo* pBloc } else { TSDBROW fRow = tsdbRowFromBlockData(pBlockData, pDumpInfo->rowIndex); - SRow* pTSRow = NULL; + SRow* pTSRow = NULL; code = tsdbRowMergerAdd(pMerger, &fRow, pReader->pSchema); if (code != TSDB_CODE_SUCCESS) { return code; @@ -2837,11 +2843,11 @@ static int32_t buildComposedDataBlock(STsdbReader* pReader) { SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(&pReader->status.blockIter); SLastBlockReader* pLastBlockReader = pReader->status.fileIter.pLastBlockReader; - bool asc = ASCENDING_TRAVERSE(pReader->order); - int64_t st = taosGetTimestampUs(); - int32_t step = asc ? 1 : -1; - double el = 0; - SDataBlk* pBlock = getCurrentBlock(&pReader->status.blockIter); + bool asc = ASCENDING_TRAVERSE(pReader->order); + int64_t st = taosGetTimestampUs(); + int32_t step = asc ? 1 : -1; + double el = 0; + SDataBlk* pBlock = getCurrentBlock(&pReader->status.blockIter); SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; STableBlockScanInfo* pBlockScanInfo = NULL; @@ -2874,7 +2880,8 @@ static int32_t buildComposedDataBlock(STsdbReader* pReader) { } } else { // file blocks not exist pBlockScanInfo = *pReader->status.pTableIter; - if (pReader->pIgnoreTables && taosHashGet(*pReader->pIgnoreTables, &pBlockScanInfo->uid, sizeof(pBlockScanInfo->uid))) { + if (pReader->pIgnoreTables && + taosHashGet(*pReader->pIgnoreTables, &pBlockScanInfo->uid, sizeof(pBlockScanInfo->uid))) { setBlockAllDumped(pDumpInfo, pBlock->maxKey.ts, pReader->order); return code; } @@ -3238,7 +3245,7 @@ static int32_t doLoadLastBlockSequentially(STsdbReader* pReader) { } static int32_t doBuildDataBlock(STsdbReader* pReader) { - int32_t code = TSDB_CODE_SUCCESS; + int32_t code = TSDB_CODE_SUCCESS; SReaderStatus* pStatus = &pReader->status; SDataBlockIter* pBlockIter = &pStatus->blockIter; @@ -3261,7 +3268,6 @@ static int32_t doBuildDataBlock(STsdbReader* pReader) { return terrno; } - initLastBlockReader(pLastBlockReader, pScanInfo, pReader); TSDBKEY keyInBuf = getCurrentKeyInBuf(pScanInfo, pReader); @@ -3338,7 +3344,7 @@ static int32_t doBuildDataBlock(STsdbReader* pReader) { } } - return (pReader->code != TSDB_CODE_SUCCESS)? pReader->code:code; + return (pReader->code != TSDB_CODE_SUCCESS) ? pReader->code : code; } static int32_t doSumFileBlockRows(STsdbReader* pReader, SDataFReader* pFileReader) { @@ -3493,14 +3499,15 @@ static int32_t buildBlockFromBufferSequentially(STsdbReader* pReader) { } STableBlockScanInfo** pBlockScanInfo = pStatus->pTableIter; - if (pReader->pIgnoreTables && taosHashGet(*pReader->pIgnoreTables, &(*pBlockScanInfo)->uid, sizeof((*pBlockScanInfo)->uid))) { + if (pReader->pIgnoreTables && + taosHashGet(*pReader->pIgnoreTables, &(*pBlockScanInfo)->uid, sizeof((*pBlockScanInfo)->uid))) { bool hasNexTable = moveToNextTable(pUidList, pStatus); if (!hasNexTable) { return TSDB_CODE_SUCCESS; } pBlockScanInfo = pStatus->pTableIter; } - + initMemDataIterator(*pBlockScanInfo, pReader); int64_t endKey = (ASCENDING_TRAVERSE(pReader->order)) ? INT64_MAX : INT64_MIN; @@ -3544,7 +3551,7 @@ static void initBlockDumpInfo(STsdbReader* pReader, SDataBlockIter* pBlockIter) static int32_t initForFirstBlockInFile(STsdbReader* pReader, SDataBlockIter* pBlockIter) { SBlockNumber num = {0}; - SArray* pTableList = taosArrayInit(40, POINTER_BYTES); + SArray* pTableList = taosArrayInit(40, POINTER_BYTES); int32_t code = moveToNextFile(pReader, &num, pTableList); if (code != TSDB_CODE_SUCCESS) { @@ -3589,7 +3596,7 @@ static ERetrieveType doReadDataFromLastFiles(STsdbReader* pReader) { SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; SDataBlockIter* pBlockIter = &pReader->status.blockIter; - while(1) { + while (1) { terrno = 0; code = doLoadLastBlockSequentially(pReader); @@ -3612,7 +3619,7 @@ static ERetrieveType doReadDataFromLastFiles(STsdbReader* pReader) { return TSDB_READ_RETURN; } - if (pBlockIter->numOfBlocks > 0) { // there are data blocks existed. + if (pBlockIter->numOfBlocks > 0) { // there are data blocks existed. return TSDB_READ_CONTINUE; } else { // all blocks in data file are checked, let's check the data in last files resetTableListIndex(&pReader->status); @@ -3625,7 +3632,7 @@ static int32_t buildBlockFromFiles(STsdbReader* pReader) { bool asc = ASCENDING_TRAVERSE(pReader->order); SDataBlockIter* pBlockIter = &pReader->status.blockIter; - SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; if (pBlockIter->numOfBlocks == 0) { // let's try to extract data from stt files. @@ -3737,13 +3744,14 @@ SVersionRange getQueryVerRange(SVnode* pVnode, SQueryTableDataCond* pCond, const endVer = (pCond->endVersion > pVnode->state.applied) ? pVnode->state.applied : pCond->endVersion; } - tsdbDebug("queried verRange:%"PRId64"-%"PRId64", revised query verRange:%"PRId64"-%"PRId64", %s", pCond->startVersion, - pCond->endVersion, startVer, endVer, id); + tsdbDebug("queried verRange:%" PRId64 "-%" PRId64 ", revised query verRange:%" PRId64 "-%" PRId64 ", %s", + pCond->startVersion, pCond->endVersion, startVer, endVer, id); return (SVersionRange){.minVer = startVer, .maxVer = endVer}; } -bool hasBeenDropped(const SArray* pDelList, int32_t* index, int64_t key, int64_t ver, int32_t order, SVersionRange* pVerRange) { +bool hasBeenDropped(const SArray* pDelList, int32_t* index, int64_t key, int64_t ver, int32_t order, + SVersionRange* pVerRange) { if (pDelList == NULL) { return false; } @@ -3761,8 +3769,7 @@ bool hasBeenDropped(const SArray* pDelList, int32_t* index, int64_t key, int64_t return false; } else if (key == last->ts) { TSDBKEY* prev = taosArrayGet(pDelList, num - 2); - return (prev->version >= ver && prev->version <= pVerRange->maxVer && - prev->version >= pVerRange->minVer); + return (prev->version >= ver && prev->version <= pVerRange->maxVer && prev->version >= pVerRange->minVer); } } else { TSDBKEY* pCurrent = taosArrayGet(pDelList, *index); @@ -3971,9 +3978,9 @@ int32_t doMergeRowsInFileBlocks(SBlockData* pBlockData, STableBlockScanInfo* pSc SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; SRowMerger* pMerger = &pReader->status.merger; - bool asc = ASCENDING_TRAVERSE(pReader->order); - int64_t key = pBlockData->aTSKEY[pDumpInfo->rowIndex]; - int32_t step = asc ? 1 : -1; + bool asc = ASCENDING_TRAVERSE(pReader->order); + int64_t key = pBlockData->aTSKEY[pDumpInfo->rowIndex]; + int32_t step = asc ? 1 : -1; pDumpInfo->rowIndex += step; if ((pDumpInfo->rowIndex <= pBlockData->nRow - 1 && asc) || (pDumpInfo->rowIndex >= 0 && !asc)) { @@ -4070,14 +4077,14 @@ int32_t doMergeMemTableMultiRows(TSDBROW* pRow, uint64_t uid, SIterInfo* pIter, return terrno; } - tsdbRowMergerAdd(&pReader->status.merger,pNextRow, pTSchema1); + tsdbRowMergerAdd(&pReader->status.merger, pNextRow, pTSchema1); } else { // let's merge rows in file block code = tsdbRowMergerAdd(&pReader->status.merger, ¤t, pReader->pSchema); if (code != TSDB_CODE_SUCCESS) { return code; } - tsdbRowMergerAdd(&pReader->status.merger,pNextRow, NULL); + tsdbRowMergerAdd(&pReader->status.merger, pNextRow, NULL); } code = doMergeRowsInBuf(pIter, uid, TSDBROW_TS(¤t), pDelList, pReader); @@ -4124,9 +4131,8 @@ int32_t doMergeMemIMemRows(TSDBROW* pRow, TSDBROW* piRow, STableBlockScanInfo* p return code; } - tsdbRowMergerAdd(&pReader->status.merger,pRow, pSchema); - code = - doMergeRowsInBuf(&pBlockScanInfo->iter, pBlockScanInfo->uid, k.ts, pBlockScanInfo->delSkyline, pReader); + tsdbRowMergerAdd(&pReader->status.merger, pRow, pSchema); + code = doMergeRowsInBuf(&pBlockScanInfo->iter, pBlockScanInfo->uid, k.ts, pBlockScanInfo->delSkyline, pReader); if (code != TSDB_CODE_SUCCESS) { return code; } @@ -4365,7 +4371,7 @@ int32_t tsdbSetTableList(STsdbReader* pReader, const void* pTableList, int32_t n int32_t size = tSimpleHashGetSize(pReader->status.pTableMap); STableBlockScanInfo** p = NULL; - int32_t iter = 0; + int32_t iter = 0; while ((p = tSimpleHashIterate(pReader->status.pTableMap, p, &iter)) != NULL) { clearBlockScanInfo(*p); @@ -4452,15 +4458,16 @@ static int32_t doOpenReaderImpl(STsdbReader* pReader) { } static void freeSchemaFunc(void* param) { - void **p = (void **)param; + void** p = (void**)param; taosMemoryFreeClear(*p); } // ====================================== EXPOSED APIs ====================================== int32_t tsdbReaderOpen(void* pVnode, SQueryTableDataCond* pCond, void* pTableList, int32_t numOfTables, - SSDataBlock* pResBlock, void** ppReader, const char* idstr, bool countOnly, SHashObj** pIgnoreTables) { + SSDataBlock* pResBlock, void** ppReader, const char* idstr, bool countOnly, + SHashObj** pIgnoreTables) { STimeWindow window = pCond->twindows; - SVnodeCfg* pConf = &(((SVnode*)pVnode)->config); + SVnodeCfg* pConf = &(((SVnode*)pVnode)->config); int32_t capacity = pConf->tsdbCfg.maxRows; if (pResBlock != NULL) { @@ -4729,7 +4736,7 @@ int32_t tsdbReaderSuspend(STsdbReader* pReader) { // resetDataBlockScanInfo excluding lastKey STableBlockScanInfo** p = NULL; - int32_t iter = 0; + int32_t iter = 0; while ((p = tSimpleHashIterate(pStatus->pTableMap, p, &iter)) != NULL) { STableBlockScanInfo* pInfo = *(STableBlockScanInfo**)p; @@ -4751,7 +4758,7 @@ int32_t tsdbReaderSuspend(STsdbReader* pReader) { } else { // resetDataBlockScanInfo excluding lastKey STableBlockScanInfo** p = NULL; - int32_t iter = 0; + int32_t iter = 0; while ((p = tSimpleHashIterate(pStatus->pTableMap, p, &iter)) != NULL) { STableBlockScanInfo* pInfo = *(STableBlockScanInfo**)p; @@ -4950,8 +4957,9 @@ int32_t tsdbNextDataBlock(STsdbReader* pReader, bool* hasNext) { *hasNext = false; - if (isEmptyQueryTimeWindow(&pReader->window) || pReader->step == EXTERNAL_ROWS_NEXT || pReader->code != TSDB_CODE_SUCCESS) { - return (pReader->code != TSDB_CODE_SUCCESS)? pReader->code:code; + if (isEmptyQueryTimeWindow(&pReader->window) || pReader->step == EXTERNAL_ROWS_NEXT || + pReader->code != TSDB_CODE_SUCCESS) { + return (pReader->code != TSDB_CODE_SUCCESS) ? pReader->code : code; } SReaderStatus* pStatus = &pReader->status; @@ -5087,7 +5095,7 @@ static bool doFillNullColSMA(SBlockLoadSuppInfo* pSup, int32_t numOfRows, int32_ return hasNullSMA; } -int32_t tsdbRetrieveDatablockSMA(STsdbReader* pReader, SSDataBlock* pDataBlock, bool* allHave, bool *hasNullSMA) { +int32_t tsdbRetrieveDatablockSMA(STsdbReader* pReader, SSDataBlock* pDataBlock, bool* allHave, bool* hasNullSMA) { SColumnDataAgg*** pBlockSMA = &pDataBlock->pBlockAgg; int32_t code = 0; @@ -5196,9 +5204,9 @@ STableBlockScanInfo* getTableBlockScanInfo(SSHashObj* pTableMap, uint64_t uid, c } static SSDataBlock* doRetrieveDataBlock(STsdbReader* pReader) { - SReaderStatus* pStatus = &pReader->status; - int32_t code = TSDB_CODE_SUCCESS; - SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(&pStatus->blockIter); + SReaderStatus* pStatus = &pReader->status; + int32_t code = TSDB_CODE_SUCCESS; + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(&pStatus->blockIter); if (pReader->code != TSDB_CODE_SUCCESS) { return NULL; diff --git a/source/dnode/vnode/src/tsdb/tsdbRead2.c b/source/dnode/vnode/src/tsdb/tsdbRead2.c new file mode 100644 index 0000000000..56ac52a22c --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbRead2.c @@ -0,0 +1,4945 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "osDef.h" +#include "tsdb.h" +#include "tsdbDataFileRW.h" +#include "tsdbFS2.h" +#include "tsdbMerge.h" +#include "tsdbReadUtil.h" +#include "tsdbUtil2.h" +#include "tsimplehash.h" + +#define ASCENDING_TRAVERSE(o) (o == TSDB_ORDER_ASC) +#define getCurrentKeyInLastBlock(_r) ((_r)->currentKey) + +static SFileDataBlockInfo* getCurrentBlockInfo(SDataBlockIter* pBlockIter); +static int32_t buildDataBlockFromBufImpl(STableBlockScanInfo* pBlockScanInfo, int64_t endKey, int32_t capacity, + STsdbReader* pReader); +static TSDBROW* getValidMemRow(SIterInfo* pIter, const SArray* pDelList, STsdbReader* pReader); +static int32_t doMergeRowsInFileBlocks(SBlockData* pBlockData, STableBlockScanInfo* pScanInfo, STsdbReader* pReader); +static int32_t doMergeRowsInLastBlock(SLastBlockReader* pLastBlockReader, STableBlockScanInfo* pScanInfo, int64_t ts, + SRowMerger* pMerger, SVersionRange* pVerRange, const char* id); +static int32_t doMergeRowsInBuf(SIterInfo* pIter, uint64_t uid, int64_t ts, SArray* pDelList, STsdbReader* pReader); +static int32_t doAppendRowFromTSRow(SSDataBlock* pBlock, STsdbReader* pReader, SRow* pTSRow, + STableBlockScanInfo* pScanInfo); +static int32_t doAppendRowFromFileBlock(SSDataBlock* pResBlock, STsdbReader* pReader, SBlockData* pBlockData, + int32_t rowIndex); +static void setComposedBlockFlag(STsdbReader* pReader, bool composed); +static bool hasBeenDropped(const SArray* pDelList, int32_t* index, int64_t key, int64_t ver, int32_t order, + SVersionRange* pVerRange); + +static int32_t doMergeMemTableMultiRows(TSDBROW* pRow, uint64_t uid, SIterInfo* pIter, SArray* pDelList, + TSDBROW* pResRow, STsdbReader* pReader, bool* freeTSRow); +static int32_t doMergeMemIMemRows(TSDBROW* pRow, TSDBROW* piRow, STableBlockScanInfo* pBlockScanInfo, + STsdbReader* pReader, SRow** pTSRow); +static int32_t mergeRowsInFileBlocks(SBlockData* pBlockData, STableBlockScanInfo* pBlockScanInfo, int64_t key, + STsdbReader* pReader); + +static int32_t initDelSkylineIterator(STableBlockScanInfo* pBlockScanInfo, int32_t order, SCostSummary* pCost); +static STsdb* getTsdbByRetentions(SVnode* pVnode, TSKEY winSKey, SRetention* retentions, const char* idstr, + int8_t* pLevel); +static SVersionRange getQueryVerRange(SVnode* pVnode, SQueryTableDataCond* pCond, int8_t level); +static bool hasDataInLastBlock(SLastBlockReader* pLastBlockReader); +static int32_t doBuildDataBlock(STsdbReader* pReader); +static TSDBKEY getCurrentKeyInBuf(STableBlockScanInfo* pScanInfo, STsdbReader* pReader); +static bool hasDataInFileBlock(const SBlockData* pBlockData, const SFileBlockDumpInfo* pDumpInfo); +static void initBlockDumpInfo(STsdbReader* pReader, SDataBlockIter* pBlockIter); +static int32_t getInitialDelIndex(const SArray* pDelSkyline, int32_t order); + +static bool outOfTimeWindow(int64_t ts, STimeWindow* pWindow) { return (ts > pWindow->ekey) || (ts < pWindow->skey); } + +static int32_t setColumnIdSlotList(SBlockLoadSuppInfo* pSupInfo, SColumnInfo* pCols, const int32_t* pSlotIdList, + int32_t numOfCols) { + pSupInfo->smaValid = true; + pSupInfo->numOfCols = numOfCols; + pSupInfo->colId = taosMemoryMalloc(numOfCols * (sizeof(int16_t) * 2 + POINTER_BYTES)); + if (pSupInfo->colId == NULL) { + taosMemoryFree(pSupInfo->colId); + return TSDB_CODE_OUT_OF_MEMORY; + } + + pSupInfo->slotId = (int16_t*)((char*)pSupInfo->colId + (sizeof(int16_t) * numOfCols)); + pSupInfo->buildBuf = (char**)((char*)pSupInfo->slotId + (sizeof(int16_t) * numOfCols)); + for (int32_t i = 0; i < numOfCols; ++i) { + pSupInfo->colId[i] = pCols[i].colId; + pSupInfo->slotId[i] = pSlotIdList[i]; + + if (IS_VAR_DATA_TYPE(pCols[i].type)) { + pSupInfo->buildBuf[i] = taosMemoryMalloc(pCols[i].bytes); + } else { + pSupInfo->buildBuf[i] = NULL; + } + } + + return TSDB_CODE_SUCCESS; +} + +static int32_t updateBlockSMAInfo(STSchema* pSchema, SBlockLoadSuppInfo* pSupInfo) { + int32_t i = 0, j = 0; + + while (i < pSchema->numOfCols && j < pSupInfo->numOfCols) { + STColumn* pTCol = &pSchema->columns[i]; + if (pTCol->colId == pSupInfo->colId[j]) { + if (!IS_BSMA_ON(pTCol)) { + pSupInfo->smaValid = false; + return TSDB_CODE_SUCCESS; + } + + i += 1; + j += 1; + } else if (pTCol->colId < pSupInfo->colId[j]) { // do nothing + i += 1; + } else { + return TSDB_CODE_INVALID_PARA; + } + } + + return TSDB_CODE_SUCCESS; +} + +static bool isEmptyQueryTimeWindow(STimeWindow* pWindow) { return pWindow->skey > pWindow->ekey; } + +// Update the query time window according to the data time to live(TTL) information, in order to avoid to return +// the expired data to client, even it is queried already. +static STimeWindow updateQueryTimeWindow(STsdb* pTsdb, STimeWindow* pWindow) { + STsdbKeepCfg* pCfg = &pTsdb->keepCfg; + + int64_t now = taosGetTimestamp(pCfg->precision); + int64_t earilyTs = now - (tsTickPerMin[pCfg->precision] * pCfg->keep2) + 1; // needs to add one tick + + STimeWindow win = *pWindow; + if (win.skey < earilyTs) { + win.skey = earilyTs; + } + + return win; +} + +// init file iterator +static int32_t initFilesetIterator(SFilesetIter* pIter, TFileSetArray* pFileSetArray, STsdbReader* pReader) { + size_t numOfFileset = TARRAY2_SIZE(pFileSetArray); + + pIter->index = ASCENDING_TRAVERSE(pReader->info.order) ? -1 : numOfFileset; + pIter->order = pReader->info.order; + pIter->pFilesetList = pFileSetArray; + pIter->numOfFiles = numOfFileset; + + if (pIter->pLastBlockReader == NULL) { + pIter->pLastBlockReader = taosMemoryCalloc(1, sizeof(struct SLastBlockReader)); + if (pIter->pLastBlockReader == NULL) { + int32_t code = TSDB_CODE_OUT_OF_MEMORY; + tsdbError("failed to prepare the last block iterator, since:%s %s", tstrerror(code), pReader->idStr); + return code; + } + } + + SLastBlockReader* pLReader = pIter->pLastBlockReader; + pLReader->order = pReader->info.order; + pLReader->window = pReader->info.window; + pLReader->verRange = pReader->info.verRange; + + pLReader->uid = 0; + tMergeTreeClose(&pLReader->mergeTree); + tsdbDebug("init fileset iterator, total files:%d %s", pIter->numOfFiles, pReader->idStr); + return TSDB_CODE_SUCCESS; +} + +static int32_t filesetIteratorNext(SFilesetIter* pIter, STsdbReader* pReader, bool* hasNext) { + bool asc = ASCENDING_TRAVERSE(pIter->order); + int32_t step = asc ? 1 : -1; + int32_t code = 0; + + pIter->index += step; + if ((asc && pIter->index >= pIter->numOfFiles) || ((!asc) && pIter->index < 0)) { + *hasNext = false; + return TSDB_CODE_SUCCESS; + } + + SCostSummary* pSum = &pReader->cost; + + pIter->pLastBlockReader->uid = 0; + tMergeTreeClose(&pIter->pLastBlockReader->mergeTree); + + pReader->status.pLDataIterArray = + destroySttBlockReader(pReader->status.pLDataIterArray, &pSum->lastBlockLoad, &pSum->lastBlockLoadTime); + pReader->status.pLDataIterArray = taosArrayInit(4, POINTER_BYTES); + + // check file the time range of coverage + STimeWindow win = {0}; + + while (1) { + if (pReader->pFileReader != NULL) { + tsdbDataFileReaderClose(&pReader->pFileReader); + } + + pReader->status.pCurrentFileset = pIter->pFilesetList->data[pIter->index]; + + STFileObj** pFileObj = pReader->status.pCurrentFileset->farr; + if (pFileObj[0] != NULL || pFileObj[3] != NULL) { + SDataFileReaderConfig conf = {.tsdb = pReader->pTsdb, .szPage = pReader->pTsdb->pVnode->config.szPage}; + + const char* filesName[4] = {0}; + + if (pFileObj[0] != NULL) { + conf.files[0].file = *pFileObj[0]->f; + conf.files[0].exist = true; + filesName[0] = pFileObj[0]->fname; + + conf.files[1].file = *pFileObj[1]->f; + conf.files[1].exist = true; + filesName[1] = pFileObj[1]->fname; + + conf.files[2].file = *pFileObj[2]->f; + conf.files[2].exist = true; + filesName[2] = pFileObj[2]->fname; + } + + if (pFileObj[3] != NULL) { + conf.files[3].exist = true; + conf.files[3].file = *pFileObj[3]->f; + filesName[3] = pFileObj[3]->fname; + } + + code = tsdbDataFileReaderOpen(filesName, &conf, &pReader->pFileReader); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + + pReader->cost.headFileLoad += 1; + } + + int32_t fid = pReader->status.pCurrentFileset->fid; + tsdbFidKeyRange(fid, pReader->pTsdb->keepCfg.days, pReader->pTsdb->keepCfg.precision, &win.skey, &win.ekey); + + // current file are no longer overlapped with query time window, ignore remain files + if ((asc && win.skey > pReader->info.window.ekey) || (!asc && win.ekey < pReader->info.window.skey)) { + tsdbDebug("%p remain files are not qualified for qrange:%" PRId64 "-%" PRId64 ", ignore, %s", pReader, + pReader->info.window.skey, pReader->info.window.ekey, pReader->idStr); + *hasNext = false; + return TSDB_CODE_SUCCESS; + } + + if ((asc && (win.ekey < pReader->info.window.skey)) || ((!asc) && (win.skey > pReader->info.window.ekey))) { + pIter->index += step; + if ((asc && pIter->index >= pIter->numOfFiles) || ((!asc) && pIter->index < 0)) { + *hasNext = false; + return TSDB_CODE_SUCCESS; + } + continue; + } + + tsdbDebug("%p file found fid:%d for qrange:%" PRId64 "-%" PRId64 ", %s", pReader, fid, pReader->info.window.skey, + pReader->info.window.ekey, pReader->idStr); + *hasNext = true; + return TSDB_CODE_SUCCESS; + } + +_err: + *hasNext = false; + return code; +} + +static void resetDataBlockIterator(SDataBlockIter* pIter, int32_t order) { + pIter->order = order; + pIter->index = -1; + pIter->numOfBlocks = 0; + if (pIter->blockList == NULL) { + pIter->blockList = taosArrayInit(4, sizeof(SFileDataBlockInfo)); + } else { + taosArrayClear(pIter->blockList); + } +} + +static void cleanupDataBlockIterator(SDataBlockIter* pIter) { taosArrayDestroy(pIter->blockList); } + +static void initReaderStatus(SReaderStatus* pStatus) { + pStatus->pTableIter = NULL; + pStatus->loadFromFile = true; +} + +static SSDataBlock* createResBlock(SQueryTableDataCond* pCond, int32_t capacity) { + SSDataBlock* pResBlock = createDataBlock(); + if (pResBlock == NULL) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + return NULL; + } + + for (int32_t i = 0; i < pCond->numOfCols; ++i) { + SColumnInfoData colInfo = {0}; + colInfo.info = pCond->colList[i]; + blockDataAppendColInfo(pResBlock, &colInfo); + } + + int32_t code = blockDataEnsureCapacity(pResBlock, capacity); + if (code != TSDB_CODE_SUCCESS) { + terrno = code; + taosMemoryFree(pResBlock); + return NULL; + } + return pResBlock; +} + +static int32_t tsdbInitReaderLock(STsdbReader* pReader) { + int32_t code = -1; + qTrace("tsdb/read: %p, pre-init read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + code = taosThreadMutexInit(&pReader->readerMutex, NULL); + + qTrace("tsdb/read: %p, post-init read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + return code; +} + +static int32_t tsdbUninitReaderLock(STsdbReader* pReader) { + int32_t code = -1; + qTrace("tsdb/read: %p, pre-uninit read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + code = taosThreadMutexDestroy(&pReader->readerMutex); + + qTrace("tsdb/read: %p, post-uninit read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + return code; +} + +static int32_t tsdbAcquireReader(STsdbReader* pReader) { + int32_t code = -1; + qTrace("tsdb/read: %p, pre-take read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + code = taosThreadMutexLock(&pReader->readerMutex); + + qTrace("tsdb/read: %p, post-take read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + return code; +} + +static int32_t tsdbTryAcquireReader(STsdbReader* pReader) { + int32_t code = -1; + qTrace("tsdb/read: %p, pre-trytake read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + code = taosThreadMutexTryLock(&pReader->readerMutex); + + qTrace("tsdb/read: %p, post-trytake read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + return code; +} + +static int32_t tsdbReleaseReader(STsdbReader* pReader) { + int32_t code = -1; + qTrace("tsdb/read: %p, pre-untake read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + code = taosThreadMutexUnlock(&pReader->readerMutex); + + qTrace("tsdb/read: %p, post-untake read mutex: %p, code: %d", pReader, &pReader->readerMutex, code); + + return code; +} + +void tsdbReleaseDataBlock2(STsdbReader* pReader) { + SReaderStatus* pStatus = &pReader->status; + if (!pStatus->composedDataBlock) { + tsdbReleaseReader(pReader); + } +} + +static int32_t initResBlockInfo(SResultBlockInfo* pResBlockInfo, int64_t capacity, SSDataBlock* pResBlock, + SQueryTableDataCond* pCond) { + pResBlockInfo->capacity = capacity; + pResBlockInfo->pResBlock = pResBlock; + terrno = 0; + + if (pResBlockInfo->pResBlock == NULL) { + pResBlockInfo->freeBlock = true; + pResBlockInfo->pResBlock = createResBlock(pCond, pResBlockInfo->capacity); + } else { + pResBlockInfo->freeBlock = false; + } + + return terrno; +} + +static int32_t tsdbReaderCreate(SVnode* pVnode, SQueryTableDataCond* pCond, void** ppReader, int32_t capacity, + SSDataBlock* pResBlock, const char* idstr) { + int32_t code = 0; + int8_t level = 0; + STsdbReader* pReader = (STsdbReader*)taosMemoryCalloc(1, sizeof(*pReader)); + if (pReader == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + goto _end; + } + + if (VND_IS_TSMA(pVnode)) { + tsdbDebug("vgId:%d, tsma is selected to query, %s", TD_VID(pVnode), idstr); + } + + initReaderStatus(&pReader->status); + + pReader->pTsdb = getTsdbByRetentions(pVnode, pCond->twindows.skey, pVnode->config.tsdbCfg.retentions, idstr, &level); + pReader->info.suid = pCond->suid; + pReader->info.order = pCond->order; + + pReader->idStr = (idstr != NULL) ? taosStrdup(idstr) : NULL; + pReader->info.verRange = getQueryVerRange(pVnode, pCond, level); + pReader->type = pCond->type; + pReader->info.window = updateQueryTimeWindow(pReader->pTsdb, &pCond->twindows); + pReader->blockInfoBuf.numPerBucket = 1000; // 1000 tables per bucket + + code = initResBlockInfo(&pReader->resBlockInfo, capacity, pResBlock, pCond); + if (code != TSDB_CODE_SUCCESS) { + goto _end; + } + + if (pCond->numOfCols <= 0) { + tsdbError("vgId:%d, invalid column number %d in query cond, %s", TD_VID(pVnode), pCond->numOfCols, idstr); + code = TSDB_CODE_INVALID_PARA; + goto _end; + } + + // allocate buffer in order to load data blocks from file + SBlockLoadSuppInfo* pSup = &pReader->suppInfo; + pSup->tsColAgg.colId = PRIMARYKEY_TIMESTAMP_COL_ID; + setColumnIdSlotList(pSup, pCond->colList, pCond->pSlotList, pCond->numOfCols); + + code = tBlockDataCreate(&pReader->status.fileBlockData); + if (code != TSDB_CODE_SUCCESS) { + terrno = code; + goto _end; + } + + if (pReader->suppInfo.colId[0] != PRIMARYKEY_TIMESTAMP_COL_ID) { + tsdbError("the first column isn't primary timestamp, %d, %s", pReader->suppInfo.colId[0], pReader->idStr); + code = TSDB_CODE_INVALID_PARA; + goto _end; + } + + pReader->status.pPrimaryTsCol = taosArrayGet(pReader->resBlockInfo.pResBlock->pDataBlock, pSup->slotId[0]); + int32_t type = pReader->status.pPrimaryTsCol->info.type; + if (type != TSDB_DATA_TYPE_TIMESTAMP) { + tsdbError("the first column isn't primary timestamp in result block, actual: %s, %s", tDataTypes[type].name, + pReader->idStr); + code = TSDB_CODE_INVALID_PARA; + goto _end; + } + + tsdbInitReaderLock(pReader); + + *ppReader = pReader; + return code; + +_end: + tsdbReaderClose(pReader); + *ppReader = NULL; + return code; +} + +static int32_t doLoadBlockIndex(STsdbReader* pReader, SDataFileReader* pFileReader, SArray* pIndexList) { + int64_t st = taosGetTimestampUs(); + int32_t numOfTables = tSimpleHashGetSize(pReader->status.pTableMap); + if (pFileReader == NULL) { + return TSDB_CODE_SUCCESS; + } + + const TBrinBlkArray* pBlkArray = NULL; + + int32_t code = tsdbDataFileReadBrinBlk(pFileReader, &pBlkArray); + +#if 0 + LRUHandle* handle = NULL; + + int32_t code = tsdbCacheGetBlockIdx(pFileReader->pTsdb->biCache, pFileReader, &handle); + if (code != TSDB_CODE_SUCCESS || handle == NULL) { + goto _end; + } + + + SArray* aBlockIdx = (SArray*)taosLRUCacheValue(pFileReader->pTsdb->biCache, handle); + size_t num = taosArrayGetSize(aBlockIdx); + if (num == 0) { + tsdbBICacheRelease(pFileReader->pTsdb->biCache, handle); + return TSDB_CODE_SUCCESS; + } +#endif + + // todo binary search to the start position + int64_t et1 = taosGetTimestampUs(); + + SBrinBlk* pBrinBlk = NULL; + STableUidList* pList = &pReader->status.uidList; + + int32_t i = 0; + + while (i < TARRAY2_SIZE(pBlkArray)) { + pBrinBlk = &pBlkArray->data[i]; + if (pBrinBlk->maxTbid.suid < pReader->info.suid) { + i += 1; + continue; + } + + if (pBrinBlk->minTbid.suid > pReader->info.suid) { // not include the queried table/super table, quit the loop + break; + } + + ASSERT(pBrinBlk->minTbid.suid <= pReader->info.suid && pBrinBlk->maxTbid.suid >= pReader->info.suid); + if (pBrinBlk->maxTbid.suid == pReader->info.suid && pBrinBlk->maxTbid.uid < pList->tableUidList[0]) { + i += 1; + continue; + } + + if (pBrinBlk->minTbid.suid == pReader->info.suid && pBrinBlk->minTbid.uid > pList->tableUidList[numOfTables - 1]) { + break; + } + + taosArrayPush(pIndexList, pBrinBlk); + i += 1; + } + + int64_t et2 = taosGetTimestampUs(); + tsdbDebug("load block index for %d/%d tables completed, elapsed time:%.2f ms, set BrinBlk:%.2f ms, size:%.2f Kb %s", + numOfTables, (int32_t)pBlkArray->size, (et1 - st) / 1000.0, (et2 - et1) / 1000.0, + pBlkArray->size * sizeof(SBrinBlk) / 1024.0, pReader->idStr); + + pReader->cost.headFileLoadTime += (et1 - st) / 1000.0; + +_end: + // tsdbBICacheRelease(pFileReader->pTsdb->biCache, handle); + return code; +} + +static int32_t doLoadFileBlock(STsdbReader* pReader, SArray* pIndexList, SBlockNumber* pBlockNum, + SArray* pTableScanInfoList) { + size_t sizeInDisk = 0; + int64_t st = taosGetTimestampUs(); + + // clear info for the new file + cleanupInfoFoxNextFileset(pReader->status.pTableMap); + + int32_t k = 0; + int32_t numOfTables = tSimpleHashGetSize(pReader->status.pTableMap); + int32_t step = ASCENDING_TRAVERSE(pReader->info.order) ? 1 : -1; + STimeWindow w = pReader->info.window; + SBrinRecord* pRecord = NULL; + + SBrinRecordIter iter = {0}; + initBrinRecordIter(&iter, pReader->pFileReader, pIndexList); + + while (((pRecord = getNextBrinRecord(&iter)) != NULL)) { + if (pRecord->suid > pReader->info.suid) { + break; + } + + uint64_t uid = pReader->status.uidList.tableUidList[k]; + if (pRecord->suid < pReader->info.suid) { + continue; + } + + if (uid < pRecord->uid) { // forward the table uid index + while (k < numOfTables && pReader->status.uidList.tableUidList[k] < pRecord->uid) { + k += 1; + } + + if (k >= numOfTables) { + break; + } + + uid = pReader->status.uidList.tableUidList[k]; + } + + if (pRecord->uid < uid) { + continue; + } + + ASSERT(pRecord->suid == pReader->info.suid && uid == pRecord->uid); + + STableBlockScanInfo* pScanInfo = getTableBlockScanInfo(pReader->status.pTableMap, uid, pReader->idStr); + if (ASCENDING_TRAVERSE(pReader->info.order)) { + w.skey = pScanInfo->lastKey + step; + } else { + w.ekey = pScanInfo->lastKey + step; + } + + if (isEmptyQueryTimeWindow(&w)) { + k += 1; + continue; + } + + // 1. time range check + if (pRecord->firstKey > w.ekey || pRecord->lastKey < w.skey) { + continue; + } + + // 2. version range check + if (pRecord->minVer > pReader->info.verRange.maxVer || pRecord->maxVer < pReader->info.verRange.minVer) { + continue; + } + + if (pScanInfo->pBlockList == NULL) { + pScanInfo->pBlockList = taosArrayInit(4, sizeof(SBrinRecord)); + } + + void* p1 = taosArrayPush(pScanInfo->pBlockList, pRecord); + if (p1 == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + pBlockNum->numOfBlocks += 1; + if (taosArrayGetSize(pTableScanInfoList) == 0) { + taosArrayPush(pTableScanInfoList, &pScanInfo); + } else { + STableBlockScanInfo** p = taosArrayGetLast(pTableScanInfoList); + if ((*p)->uid != uid) { + taosArrayPush(pTableScanInfoList, &pScanInfo); + } + } + } + + clearBrinBlockIter(&iter); + + pBlockNum->numOfLastFiles = pReader->status.pCurrentFileset->lvlArr->size; + int32_t total = pBlockNum->numOfLastFiles + pBlockNum->numOfBlocks; + + double el = (taosGetTimestampUs() - st) / 1000.0; + tsdbDebug( + "load block of %d tables completed, blocks:%d in %d tables, last-files:%d, block-info-size:%.2f Kb, elapsed " + "time:%.2f ms %s", + numOfTables, pBlockNum->numOfBlocks, (int32_t)taosArrayGetSize(pTableScanInfoList), pBlockNum->numOfLastFiles, + sizeInDisk / 1000.0, el, pReader->idStr); + + pReader->cost.numOfBlocks += total; + pReader->cost.headFileLoadTime += el; + + return TSDB_CODE_SUCCESS; +} + +static void setBlockAllDumped(SFileBlockDumpInfo* pDumpInfo, int64_t maxKey, int32_t order) { + int32_t step = ASCENDING_TRAVERSE(order) ? 1 : -1; + pDumpInfo->allDumped = true; + pDumpInfo->lastKey = maxKey + step; +} + +static int32_t doCopyColVal(SColumnInfoData* pColInfoData, int32_t rowIndex, int32_t colIndex, SColVal* pColVal, + SBlockLoadSuppInfo* pSup) { + if (IS_VAR_DATA_TYPE(pColVal->type)) { + if (!COL_VAL_IS_VALUE(pColVal)) { + colDataSetNULL(pColInfoData, rowIndex); + } else { + varDataSetLen(pSup->buildBuf[colIndex], pColVal->value.nData); + if (pColVal->value.nData > pColInfoData->info.bytes) { + tsdbWarn("column cid:%d actual data len %d is bigger than schema len %d", pColVal->cid, pColVal->value.nData, + pColInfoData->info.bytes); + return TSDB_CODE_TDB_INVALID_TABLE_SCHEMA_VER; + } + if (pColVal->value.nData > 0) { // pData may be null, if nData is 0 + memcpy(varDataVal(pSup->buildBuf[colIndex]), pColVal->value.pData, pColVal->value.nData); + } + + colDataSetVal(pColInfoData, rowIndex, pSup->buildBuf[colIndex], false); + } + } else { + colDataSetVal(pColInfoData, rowIndex, (const char*)&pColVal->value, !COL_VAL_IS_VALUE(pColVal)); + } + + return TSDB_CODE_SUCCESS; +} + +static SFileDataBlockInfo* getCurrentBlockInfo(SDataBlockIter* pBlockIter) { + size_t num = TARRAY_SIZE(pBlockIter->blockList); + if (num == 0) { + ASSERT(pBlockIter->numOfBlocks == num); + return NULL; + } + + SFileDataBlockInfo* pBlockInfo = taosArrayGet(pBlockIter->blockList, pBlockIter->index); + return pBlockInfo; +} + +static int doBinarySearchKey(TSKEY* keyList, int num, int pos, TSKEY key, int order) { + // start end position + int s, e; + s = pos; + + // check + ASSERT(pos >= 0 && pos < num && num > 0); + if (order == TSDB_ORDER_ASC) { + // find the first position which is smaller than the key + e = num - 1; + if (key < keyList[pos]) return -1; + while (1) { + // check can return + if (key >= keyList[e]) return e; + if (key <= keyList[s]) return s; + if (e - s <= 1) return s; + + // change start or end position + int mid = s + (e - s + 1) / 2; + if (keyList[mid] > key) + e = mid; + else if (keyList[mid] < key) + s = mid; + else + return mid; + } + } else { // DESC + // find the first position which is bigger than the key + e = 0; + if (key > keyList[pos]) return -1; + while (1) { + // check can return + if (key <= keyList[e]) return e; + if (key >= keyList[s]) return s; + if (s - e <= 1) return s; + + // change start or end position + int mid = s - (s - e + 1) / 2; + if (keyList[mid] < key) + e = mid; + else if (keyList[mid] > key) + s = mid; + else + return mid; + } + } +} + +static int32_t getEndPosInDataBlock(STsdbReader* pReader, SBlockData* pBlockData, SBrinRecord* pRecord, int32_t pos) { + // NOTE: reverse the order to find the end position in data block + int32_t endPos = -1; + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + + if (asc && pReader->info.window.ekey >= pRecord->lastKey) { + endPos = pRecord->numRow - 1; + } else if (!asc && pReader->info.window.skey <= pRecord->firstKey) { + endPos = 0; + } else { + int64_t key = asc ? pReader->info.window.ekey : pReader->info.window.skey; + endPos = doBinarySearchKey(pBlockData->aTSKEY, pRecord->numRow, pos, key, pReader->info.order); + } + + if ((pReader->info.verRange.maxVer >= pRecord->minVer && pReader->info.verRange.maxVer < pRecord->maxVer) || + (pReader->info.verRange.minVer <= pRecord->maxVer && pReader->info.verRange.minVer > pRecord->minVer)) { + int32_t i = endPos; + + if (asc) { + for (; i >= 0; --i) { + if (pBlockData->aVersion[i] <= pReader->info.verRange.maxVer) { + break; + } + } + } else { + for (; i < pRecord->numRow; ++i) { + if (pBlockData->aVersion[i] >= pReader->info.verRange.minVer) { + break; + } + } + } + + endPos = i; + } + + return endPos; +} + +static void copyPrimaryTsCol(const SBlockData* pBlockData, SFileBlockDumpInfo* pDumpInfo, SColumnInfoData* pColData, + int32_t dumpedRows, bool asc) { + if (asc) { + memcpy(pColData->pData, &pBlockData->aTSKEY[pDumpInfo->rowIndex], dumpedRows * sizeof(int64_t)); + } else { + int32_t startIndex = pDumpInfo->rowIndex - dumpedRows + 1; + memcpy(pColData->pData, &pBlockData->aTSKEY[startIndex], dumpedRows * sizeof(int64_t)); + + // todo: opt perf by extract the loop + // reverse the array list + int32_t mid = dumpedRows >> 1u; + int64_t* pts = (int64_t*)pColData->pData; + for (int32_t j = 0; j < mid; ++j) { + int64_t t = pts[j]; + pts[j] = pts[dumpedRows - j - 1]; + pts[dumpedRows - j - 1] = t; + } + } +} + +// a faster version of copy procedure. +static void copyNumericCols(const SColData* pData, SFileBlockDumpInfo* pDumpInfo, SColumnInfoData* pColData, + int32_t dumpedRows, bool asc) { + uint8_t* p = NULL; + if (asc) { + p = pData->pData + tDataTypes[pData->type].bytes * pDumpInfo->rowIndex; + } else { + int32_t startIndex = pDumpInfo->rowIndex - dumpedRows + 1; + p = pData->pData + tDataTypes[pData->type].bytes * startIndex; + } + + int32_t step = asc ? 1 : -1; + + // make sure it is aligned to 8bit, the allocated memory address is aligned to 256bit + // ASSERT((((uint64_t)pColData->pData) & (0x8 - 1)) == 0); + + // 1. copy data in a batch model + memcpy(pColData->pData, p, dumpedRows * tDataTypes[pData->type].bytes); + + // 2. reverse the array list in case of descending order scan data block + if (!asc) { + switch (pColData->info.type) { + case TSDB_DATA_TYPE_TIMESTAMP: + case TSDB_DATA_TYPE_DOUBLE: + case TSDB_DATA_TYPE_BIGINT: + case TSDB_DATA_TYPE_UBIGINT: { + int32_t mid = dumpedRows >> 1u; + int64_t* pts = (int64_t*)pColData->pData; + for (int32_t j = 0; j < mid; ++j) { + int64_t t = pts[j]; + pts[j] = pts[dumpedRows - j - 1]; + pts[dumpedRows - j - 1] = t; + } + break; + } + + case TSDB_DATA_TYPE_BOOL: + case TSDB_DATA_TYPE_TINYINT: + case TSDB_DATA_TYPE_UTINYINT: { + int32_t mid = dumpedRows >> 1u; + int8_t* pts = (int8_t*)pColData->pData; + for (int32_t j = 0; j < mid; ++j) { + int8_t t = pts[j]; + pts[j] = pts[dumpedRows - j - 1]; + pts[dumpedRows - j - 1] = t; + } + break; + } + + case TSDB_DATA_TYPE_SMALLINT: + case TSDB_DATA_TYPE_USMALLINT: { + int32_t mid = dumpedRows >> 1u; + int16_t* pts = (int16_t*)pColData->pData; + for (int32_t j = 0; j < mid; ++j) { + int64_t t = pts[j]; + pts[j] = pts[dumpedRows - j - 1]; + pts[dumpedRows - j - 1] = t; + } + break; + } + + case TSDB_DATA_TYPE_FLOAT: + case TSDB_DATA_TYPE_INT: + case TSDB_DATA_TYPE_UINT: { + int32_t mid = dumpedRows >> 1u; + int32_t* pts = (int32_t*)pColData->pData; + for (int32_t j = 0; j < mid; ++j) { + int32_t t = pts[j]; + pts[j] = pts[dumpedRows - j - 1]; + pts[dumpedRows - j - 1] = t; + } + break; + } + } + } + + // 3. if the null value exists, check items one-by-one + if (pData->flag != HAS_VALUE) { + int32_t rowIndex = 0; + + for (int32_t j = pDumpInfo->rowIndex; rowIndex < dumpedRows; j += step, rowIndex++) { + uint8_t v = tColDataGetBitValue(pData, j); + if (v == 0 || v == 1) { + colDataSetNull_f(pColData->nullbitmap, rowIndex); + pColData->hasNull = true; + } + } + } +} + +static int32_t copyBlockDataToSDataBlock(STsdbReader* pReader) { + SReaderStatus* pStatus = &pReader->status; + SDataBlockIter* pBlockIter = &pStatus->blockIter; + SBlockLoadSuppInfo* pSupInfo = &pReader->suppInfo; + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + + SBlockData* pBlockData = &pStatus->fileBlockData; + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(pBlockIter); + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + int32_t numOfOutputCols = pSupInfo->numOfCols; + int32_t code = TSDB_CODE_SUCCESS; + + SColVal cv = {0}; + int64_t st = taosGetTimestampUs(); + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + int32_t step = asc ? 1 : -1; + + SBrinRecord* pRecord = &pBlockInfo->record; + + // no data exists, return directly. + if (pBlockData->nRow == 0 || pBlockData->aTSKEY == 0) { + tsdbWarn("%p no need to copy since no data in blockData, table uid:%" PRIu64 " has been dropped, %s", pReader, + pBlockInfo->uid, pReader->idStr); + pResBlock->info.rows = 0; + return 0; + } + + // row index of dump info remain the initial position, let's find the appropriate start position. + if ((pDumpInfo->rowIndex == 0 && asc) || (pDumpInfo->rowIndex == pRecord->numRow - 1 && (!asc))) { + if (asc && pReader->info.window.skey <= pRecord->firstKey && pReader->info.verRange.minVer <= pRecord->minVer) { + // pDumpInfo->rowIndex = 0; + } else if (!asc && pReader->info.window.ekey >= pRecord->lastKey && + pReader->info.verRange.maxVer >= pRecord->maxVer) { + // pDumpInfo->rowIndex = pRecord->numRow - 1; + } else { // find the appropriate the start position in current block, and set it to be the current rowIndex + int32_t pos = asc ? pRecord->numRow - 1 : 0; + int32_t order = asc ? TSDB_ORDER_DESC : TSDB_ORDER_ASC; + int64_t key = asc ? pReader->info.window.skey : pReader->info.window.ekey; + pDumpInfo->rowIndex = doBinarySearchKey(pBlockData->aTSKEY, pRecord->numRow, pos, key, order); + + if (pDumpInfo->rowIndex < 0) { + tsdbError( + "%p failed to locate the start position in current block, global index:%d, table index:%d, brange:%" PRId64 + "-%" PRId64 ", minVer:%" PRId64 ", maxVer:%" PRId64 " %s", + pReader, pBlockIter->index, pBlockInfo->tbBlockIdx, pRecord->firstKey, pRecord->lastKey, pRecord->minVer, + pRecord->maxVer, pReader->idStr); + return TSDB_CODE_INVALID_PARA; + } + + ASSERT(pReader->info.verRange.minVer <= pRecord->maxVer && pReader->info.verRange.maxVer >= pRecord->minVer); + + // find the appropriate start position that satisfies the version requirement. + if ((pReader->info.verRange.maxVer >= pRecord->minVer && pReader->info.verRange.maxVer < pRecord->maxVer) || + (pReader->info.verRange.minVer <= pRecord->maxVer && pReader->info.verRange.minVer > pRecord->minVer)) { + int32_t i = pDumpInfo->rowIndex; + if (asc) { + for (; i < pRecord->numRow; ++i) { + if (pBlockData->aVersion[i] >= pReader->info.verRange.minVer) { + break; + } + } + } else { + for (; i >= 0; --i) { + if (pBlockData->aVersion[i] <= pReader->info.verRange.maxVer) { + break; + } + } + } + + pDumpInfo->rowIndex = i; + } + } + } + + // time window check + int32_t endIndex = getEndPosInDataBlock(pReader, pBlockData, pRecord, pDumpInfo->rowIndex); + if (endIndex == -1) { + setBlockAllDumped(pDumpInfo, pReader->info.window.ekey, pReader->info.order); + return TSDB_CODE_SUCCESS; + } + + endIndex += step; + int32_t dumpedRows = asc ? (endIndex - pDumpInfo->rowIndex) : (pDumpInfo->rowIndex - endIndex); + if (dumpedRows > pReader->resBlockInfo.capacity) { // output buffer check + dumpedRows = pReader->resBlockInfo.capacity; + } else if (dumpedRows <= 0) { // no qualified rows in current data block, abort directly. + setBlockAllDumped(pDumpInfo, pReader->info.window.ekey, pReader->info.order); + return TSDB_CODE_SUCCESS; + } + + int32_t i = 0; + int32_t rowIndex = 0; + + SColumnInfoData* pColData = taosArrayGet(pResBlock->pDataBlock, pSupInfo->slotId[i]); + if (pSupInfo->colId[i] == PRIMARYKEY_TIMESTAMP_COL_ID) { + copyPrimaryTsCol(pBlockData, pDumpInfo, pColData, dumpedRows, asc); + i += 1; + } + + int32_t colIndex = 0; + int32_t num = pBlockData->nColData; + while (i < numOfOutputCols && colIndex < num) { + rowIndex = 0; + + SColData* pData = tBlockDataGetColDataByIdx(pBlockData, colIndex); + if (pData->cid < pSupInfo->colId[i]) { + colIndex += 1; + } else if (pData->cid == pSupInfo->colId[i]) { + pColData = taosArrayGet(pResBlock->pDataBlock, pSupInfo->slotId[i]); + + if (pData->flag == HAS_NONE || pData->flag == HAS_NULL || pData->flag == (HAS_NULL | HAS_NONE)) { + colDataSetNNULL(pColData, 0, dumpedRows); + } else { + if (IS_MATHABLE_TYPE(pColData->info.type)) { + copyNumericCols(pData, pDumpInfo, pColData, dumpedRows, asc); + } else { // varchar/nchar type + for (int32_t j = pDumpInfo->rowIndex; rowIndex < dumpedRows; j += step) { + tColDataGetValue(pData, j, &cv); + code = doCopyColVal(pColData, rowIndex++, i, &cv, pSupInfo); + if (code) { + return code; + } + } + } + } + + colIndex += 1; + i += 1; + } else { // the specified column does not exist in file block, fill with null data + pColData = taosArrayGet(pResBlock->pDataBlock, pSupInfo->slotId[i]); + colDataSetNNULL(pColData, 0, dumpedRows); + i += 1; + } + } + + // fill the mis-matched columns with null value + while (i < numOfOutputCols) { + pColData = taosArrayGet(pResBlock->pDataBlock, pSupInfo->slotId[i]); + colDataSetNNULL(pColData, 0, dumpedRows); + i += 1; + } + + pResBlock->info.dataLoad = 1; + pResBlock->info.rows = dumpedRows; + pDumpInfo->rowIndex += step * dumpedRows; + + // check if current block are all handled + if (pDumpInfo->rowIndex >= 0 && pDumpInfo->rowIndex < pRecord->numRow) { + int64_t ts = pBlockData->aTSKEY[pDumpInfo->rowIndex]; + if (outOfTimeWindow(ts, + &pReader->info.window)) { // the remain data has out of query time window, ignore current block + setBlockAllDumped(pDumpInfo, ts, pReader->info.order); + } + } else { + int64_t ts = asc ? pRecord->lastKey : pRecord->firstKey; + setBlockAllDumped(pDumpInfo, ts, pReader->info.order); + } + + double elapsedTime = (taosGetTimestampUs() - st) / 1000.0; + pReader->cost.blockLoadTime += elapsedTime; + + int32_t unDumpedRows = asc ? pRecord->numRow - pDumpInfo->rowIndex : pDumpInfo->rowIndex + 1; + tsdbDebug("%p copy file block to sdatablock, global index:%d, table index:%d, brange:%" PRId64 "-%" PRId64 + ", rows:%d, remain:%d, minVer:%" PRId64 ", maxVer:%" PRId64 ", uid:%" PRIu64 " elapsed time:%.2f ms, %s", + pReader, pBlockIter->index, pBlockInfo->tbBlockIdx, pRecord->firstKey, pRecord->lastKey, dumpedRows, + unDumpedRows, pRecord->minVer, pRecord->maxVer, pBlockInfo->uid, elapsedTime, pReader->idStr); + + return TSDB_CODE_SUCCESS; +} + +static FORCE_INLINE STSchema* getTableSchemaImpl(STsdbReader* pReader, uint64_t uid) { + ASSERT(pReader->info.pSchema == NULL); + + int32_t code = metaGetTbTSchemaEx(pReader->pTsdb->pVnode->pMeta, pReader->info.suid, uid, -1, &pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS || pReader->info.pSchema == NULL) { + terrno = code; + tsdbError("failed to get table schema, uid:%" PRIu64 ", it may have been dropped, ver:-1, %s", uid, pReader->idStr); + return NULL; + } + + code = tsdbRowMergerInit(&pReader->status.merger, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + terrno = code; + tsdbError("failed to init merger, code:%s, %s", tstrerror(code), pReader->idStr); + return NULL; + } + + return pReader->info.pSchema; +} + +static int32_t doLoadFileBlockData(STsdbReader* pReader, SDataBlockIter* pBlockIter, SBlockData* pBlockData, + uint64_t uid) { + int32_t code = 0; + STSchema* pSchema = pReader->info.pSchema; + int64_t st = taosGetTimestampUs(); + + tBlockDataReset(pBlockData); + + if (pReader->info.pSchema == NULL) { + pSchema = getTableSchemaImpl(pReader, uid); + if (pSchema == NULL) { + tsdbDebug("%p table uid:%" PRIu64 " has been dropped, no data existed, %s", pReader, uid, pReader->idStr); + return code; + } + } + + SBlockLoadSuppInfo* pSup = &pReader->suppInfo; + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(pBlockIter); + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + + SBrinRecord* pRecord = &pBlockInfo->record; + code = tsdbDataFileReadBlockDataByColumn(pReader->pFileReader, pRecord, pBlockData, pSchema, &pSup->colId[1], + pSup->numOfCols - 1); + if (code != TSDB_CODE_SUCCESS) { + tsdbError("%p error occurs in loading file block, global index:%d, table index:%d, brange:%" PRId64 "-%" PRId64 + ", rows:%d, code:%s %s", + pReader, pBlockIter->index, pBlockInfo->tbBlockIdx, pBlockInfo->record.firstKey, + pBlockInfo->record.lastKey, pBlockInfo->record.numRow, tstrerror(code), pReader->idStr); + return code; + } + + double elapsedTime = (taosGetTimestampUs() - st) / 1000.0; + + tsdbDebug("%p load file block into buffer, global index:%d, index in table block list:%d, brange:%" PRId64 "-%" PRId64 + ", rows:%d, minVer:%" PRId64 ", maxVer:%" PRId64 ", elapsed time:%.2f ms, %s", + pReader, pBlockIter->index, pBlockInfo->tbBlockIdx, pRecord->firstKey, pRecord->lastKey, pRecord->numRow, + pRecord->minVer, pRecord->maxVer, elapsedTime, pReader->idStr); + + pReader->cost.blockLoadTime += elapsedTime; + pDumpInfo->allDumped = false; + + return TSDB_CODE_SUCCESS; +} + +/** + * This is an two rectangles overlap cases. + */ +static int32_t dataBlockPartiallyRequired(STimeWindow* pWindow, SVersionRange* pVerRange, SFileDataBlockInfo* pBlock) { + return (pWindow->ekey < pBlock->record.lastKey && pWindow->ekey >= pBlock->record.firstKey) || + (pWindow->skey > pBlock->record.firstKey && pWindow->skey <= pBlock->record.lastKey) || + (pVerRange->minVer > pBlock->record.minVer && pVerRange->minVer <= pBlock->record.maxVer) || + (pVerRange->maxVer < pBlock->record.maxVer && pVerRange->maxVer >= pBlock->record.minVer); +} + +static bool getNeighborBlockOfSameTable(SFileDataBlockInfo* pBlockInfo, STableBlockScanInfo* pTableBlockScanInfo, + int32_t* nextIndex, int32_t order, SBrinRecord* pRecord) { + bool asc = ASCENDING_TRAVERSE(order); + if (asc && pBlockInfo->tbBlockIdx >= taosArrayGetSize(pTableBlockScanInfo->pBlockList) - 1) { + return false; + } + + if (!asc && pBlockInfo->tbBlockIdx == 0) { + return false; + } + + int32_t step = asc ? 1 : -1; + // *nextIndex = pBlockInfo->tbBlockIdx + step; + // *pBlockIndex = *(SBlockIndex*)taosArrayGet(pTableBlockScanInfo->pBlockList, *nextIndex); + SBrinRecord* p = taosArrayGet(pTableBlockScanInfo->pBlockList, pBlockInfo->tbBlockIdx + step); + memcpy(pRecord, p, sizeof(SBrinRecord)); + + // tMapDataGetItemByIdx(&pTableBlockScanInfo->mapData, pIndex->ordinalIndex, pBlock, tGetDataBlk); + return true; +} + +static int32_t findFileBlockInfoIndex(SDataBlockIter* pBlockIter, SFileDataBlockInfo* pFBlockInfo) { + int32_t step = ASCENDING_TRAVERSE(pBlockIter->order) ? 1 : -1; + int32_t index = pBlockIter->index; + + while (index < pBlockIter->numOfBlocks && index >= 0) { + SFileDataBlockInfo* pFBlock = taosArrayGet(pBlockIter->blockList, index); + if (pFBlock->uid == pFBlockInfo->uid && pFBlock->tbBlockIdx == pFBlockInfo->tbBlockIdx) { + return index; + } + + index += step; + } + + return -1; +} + +static int32_t setFileBlockActiveInBlockIter(SDataBlockIter* pBlockIter, int32_t index, int32_t step) { + if (index < 0 || index >= pBlockIter->numOfBlocks) { + return -1; + } + + SFileDataBlockInfo fblock = *(SFileDataBlockInfo*)taosArrayGet(pBlockIter->blockList, index); + pBlockIter->index += step; + + if (index != pBlockIter->index) { + taosArrayRemove(pBlockIter->blockList, index); + taosArrayInsert(pBlockIter->blockList, pBlockIter->index, &fblock); + + SFileDataBlockInfo* pBlockInfo = taosArrayGet(pBlockIter->blockList, pBlockIter->index); + ASSERT(pBlockInfo->uid == fblock.uid && pBlockInfo->tbBlockIdx == fblock.tbBlockIdx); + } + + return TSDB_CODE_SUCCESS; +} + +// todo: this attribute could be acquired during extractin the global ordered block list. +static bool overlapWithNeighborBlock2(SFileDataBlockInfo* pBlock, SBrinRecord* pRec, int32_t order) { + // it is the last block in current file, no chance to overlap with neighbor blocks. + if (ASCENDING_TRAVERSE(order)) { + return pBlock->record.lastKey == pRec->firstKey; + } else { + return pBlock->record.firstKey == pRec->lastKey; + } +} + +static bool bufferDataInFileBlockGap(int32_t order, TSDBKEY key, SFileDataBlockInfo* pBlock) { + bool ascScan = ASCENDING_TRAVERSE(order); + + return (ascScan && (key.ts != TSKEY_INITIAL_VAL && key.ts <= pBlock->record.firstKey)) || + (!ascScan && (key.ts != TSKEY_INITIAL_VAL && key.ts >= pBlock->record.lastKey)); +} + +static bool keyOverlapFileBlock(TSDBKEY key, SFileDataBlockInfo* pBlock, SVersionRange* pVerRange) { + return (key.ts >= pBlock->record.firstKey && key.ts <= pBlock->record.lastKey) && + (pBlock->record.maxVer >= pVerRange->minVer) && (pBlock->record.minVer <= pVerRange->maxVer); +} + +static bool doCheckforDatablockOverlap(STableBlockScanInfo* pBlockScanInfo, const SBrinRecord* pRecord, + int32_t startIndex) { + size_t num = taosArrayGetSize(pBlockScanInfo->delSkyline); + + for (int32_t i = startIndex; i < num; i += 1) { + TSDBKEY* p = taosArrayGet(pBlockScanInfo->delSkyline, i); + if (p->ts >= pRecord->firstKey && p->ts <= pRecord->lastKey) { + if (p->version >= pRecord->minVer) { + return true; + } + } else if (p->ts < pRecord->firstKey) { // p->ts < pBlock->minKey.ts + if (p->version >= pRecord->minVer) { + if (i < num - 1) { + TSDBKEY* pnext = taosArrayGet(pBlockScanInfo->delSkyline, i + 1); + if (pnext->ts >= pRecord->firstKey) { + return true; + } + } else { // it must be the last point + ASSERT(p->version == 0); + } + } + } else { // (p->ts > pBlock->maxKey.ts) { + return false; + } + } + + return false; +} + +static bool overlapWithDelSkyline(STableBlockScanInfo* pBlockScanInfo, const SBrinRecord* pRecord, int32_t order) { + if (pBlockScanInfo->delSkyline == NULL || (taosArrayGetSize(pBlockScanInfo->delSkyline) == 0)) { + return false; + } + + // ts is not overlap + TSDBKEY* pFirst = taosArrayGet(pBlockScanInfo->delSkyline, 0); + TSDBKEY* pLast = taosArrayGetLast(pBlockScanInfo->delSkyline); + if (pRecord->firstKey > pLast->ts || pRecord->lastKey < pFirst->ts) { + return false; + } + + // version is not overlap + if (ASCENDING_TRAVERSE(order)) { + return doCheckforDatablockOverlap(pBlockScanInfo, pRecord, pBlockScanInfo->fileDelIndex); + } else { + int32_t index = pBlockScanInfo->fileDelIndex; + while (1) { + TSDBKEY* p = taosArrayGet(pBlockScanInfo->delSkyline, index); + if (p->ts > pRecord->firstKey && index > 0) { + index -= 1; + } else { // find the first point that is smaller than the minKey.ts of dataBlock. + if (p->ts == pRecord->firstKey && p->version < pRecord->maxVer && index > 0) { + index -= 1; + } + break; + } + } + + return doCheckforDatablockOverlap(pBlockScanInfo, pRecord, index); + } +} + +typedef struct { + bool overlapWithNeighborBlock; + bool hasDupTs; + bool overlapWithDelInfo; + bool overlapWithLastBlock; + bool overlapWithKeyInBuf; + bool partiallyRequired; + bool moreThanCapcity; +} SDataBlockToLoadInfo; + +static void getBlockToLoadInfo(SDataBlockToLoadInfo* pInfo, SFileDataBlockInfo* pBlockInfo, + STableBlockScanInfo* pScanInfo, TSDBKEY keyInBuf, SLastBlockReader* pLastBlockReader, + STsdbReader* pReader) { + int32_t neighborIndex = 0; + SBrinRecord rec = {0}; + + bool hasNeighbor = getNeighborBlockOfSameTable(pBlockInfo, pScanInfo, &neighborIndex, pReader->info.order, &rec); + + // overlap with neighbor + if (hasNeighbor) { + pInfo->overlapWithNeighborBlock = overlapWithNeighborBlock2(pBlockInfo, &rec, pReader->info.order); + } + + // has duplicated ts of different version in this block + pInfo->hasDupTs = (pBlockInfo->record.numRow > pBlockInfo->record.count); + pInfo->overlapWithDelInfo = overlapWithDelSkyline(pScanInfo, &pBlockInfo->record, pReader->info.order); + + if (hasDataInLastBlock(pLastBlockReader)) { + int64_t tsLast = getCurrentKeyInLastBlock(pLastBlockReader); + pInfo->overlapWithLastBlock = !(pBlockInfo->record.lastKey < tsLast || pBlockInfo->record.firstKey > tsLast); + } + + pInfo->moreThanCapcity = pBlockInfo->record.numRow > pReader->resBlockInfo.capacity; + pInfo->partiallyRequired = dataBlockPartiallyRequired(&pReader->info.window, &pReader->info.verRange, pBlockInfo); + pInfo->overlapWithKeyInBuf = keyOverlapFileBlock(keyInBuf, pBlockInfo, &pReader->info.verRange); +} + +// 1. the version of all rows should be less than the endVersion +// 2. current block should not overlap with next neighbor block +// 3. current timestamp should not be overlap with each other +// 4. output buffer should be large enough to hold all rows in current block +// 5. delete info should not overlap with current block data +// 6. current block should not contain the duplicated ts +static bool fileBlockShouldLoad(STsdbReader* pReader, SFileDataBlockInfo* pBlockInfo, STableBlockScanInfo* pScanInfo, + TSDBKEY keyInBuf, SLastBlockReader* pLastBlockReader) { + SDataBlockToLoadInfo info = {0}; + getBlockToLoadInfo(&info, pBlockInfo, pScanInfo, keyInBuf, pLastBlockReader, pReader); + + bool loadDataBlock = + (info.overlapWithNeighborBlock || info.hasDupTs || info.partiallyRequired || info.overlapWithKeyInBuf || + info.moreThanCapcity || info.overlapWithDelInfo || info.overlapWithLastBlock); + + // log the reason why load the datablock for profile + if (loadDataBlock) { + tsdbDebug("%p uid:%" PRIu64 + " need to load the datablock, overlapneighbor:%d, hasDup:%d, partiallyRequired:%d, " + "overlapWithKey:%d, greaterThanBuf:%d, overlapWithDel:%d, overlapWithlastBlock:%d, %s", + pReader, pBlockInfo->uid, info.overlapWithNeighborBlock, info.hasDupTs, info.partiallyRequired, + info.overlapWithKeyInBuf, info.moreThanCapcity, info.overlapWithDelInfo, info.overlapWithLastBlock, + pReader->idStr); + } + + return loadDataBlock; +} + +static bool isCleanFileDataBlock(STsdbReader* pReader, SFileDataBlockInfo* pBlockInfo, STableBlockScanInfo* pScanInfo, + TSDBKEY keyInBuf, SLastBlockReader* pLastBlockReader) { + SDataBlockToLoadInfo info = {0}; + getBlockToLoadInfo(&info, pBlockInfo, pScanInfo, keyInBuf, pLastBlockReader, pReader); + bool isCleanFileBlock = !(info.overlapWithNeighborBlock || info.hasDupTs || info.overlapWithKeyInBuf || + info.overlapWithDelInfo || info.overlapWithLastBlock); + return isCleanFileBlock; +} + +static int32_t buildDataBlockFromBuf(STsdbReader* pReader, STableBlockScanInfo* pBlockScanInfo, int64_t endKey) { + if (!(pBlockScanInfo->iiter.hasVal || pBlockScanInfo->iter.hasVal)) { + return TSDB_CODE_SUCCESS; + } + + SSDataBlock* pBlock = pReader->resBlockInfo.pResBlock; + + int64_t st = taosGetTimestampUs(); + int32_t code = buildDataBlockFromBufImpl(pBlockScanInfo, endKey, pReader->resBlockInfo.capacity, pReader); + + blockDataUpdateTsWindow(pBlock, pReader->suppInfo.slotId[0]); + pBlock->info.id.uid = pBlockScanInfo->uid; + + setComposedBlockFlag(pReader, true); + + double elapsedTime = (taosGetTimestampUs() - st) / 1000.0; + tsdbDebug("%p build data block from cache completed, elapsed time:%.2f ms, numOfRows:%" PRId64 ", brange:%" PRId64 + " - %" PRId64 ", uid:%" PRIu64 ", %s", + pReader, elapsedTime, pBlock->info.rows, pBlock->info.window.skey, pBlock->info.window.ekey, + pBlockScanInfo->uid, pReader->idStr); + + pReader->cost.buildmemBlock += elapsedTime; + return code; +} + +static bool tryCopyDistinctRowFromFileBlock(STsdbReader* pReader, SBlockData* pBlockData, int64_t key, + SFileBlockDumpInfo* pDumpInfo, bool* copied) { + // opt version + // 1. it is not a border point + // 2. the direct next point is not an duplicated timestamp + int32_t code = TSDB_CODE_SUCCESS; + + *copied = false; + bool asc = (pReader->info.order == TSDB_ORDER_ASC); + if ((pDumpInfo->rowIndex < pDumpInfo->totalRows - 1 && asc) || (pDumpInfo->rowIndex > 0 && (!asc))) { + int32_t step = pReader->info.order == TSDB_ORDER_ASC ? 1 : -1; + + int64_t nextKey = pBlockData->aTSKEY[pDumpInfo->rowIndex + step]; + if (nextKey != key) { // merge is not needed + code = doAppendRowFromFileBlock(pReader->resBlockInfo.pResBlock, pReader, pBlockData, pDumpInfo->rowIndex); + if (code) { + return code; + } + pDumpInfo->rowIndex += step; + *copied = true; + } + } + + return code; +} + +static bool nextRowFromLastBlocks(SLastBlockReader* pLastBlockReader, STableBlockScanInfo* pScanInfo, + SVersionRange* pVerRange) { + int32_t step = ASCENDING_TRAVERSE(pLastBlockReader->order) ? 1 : -1; + + while (1) { + bool hasVal = tMergeTreeNext(&pLastBlockReader->mergeTree); + if (!hasVal) { // the next value will be the accessed key in stt + pScanInfo->lastKeyInStt += step; + return false; + } + + TSDBROW* pRow = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + int64_t key = pRow->pBlockData->aTSKEY[pRow->iRow]; + int64_t ver = pRow->pBlockData->aVersion[pRow->iRow]; + + pLastBlockReader->currentKey = key; + pScanInfo->lastKeyInStt = key; + + if (!hasBeenDropped(pScanInfo->delSkyline, &pScanInfo->lastBlockDelIndex, key, ver, pLastBlockReader->order, + pVerRange)) { + return true; + } + } +} + +static bool tryCopyDistinctRowFromSttBlock(TSDBROW* fRow, SLastBlockReader* pLastBlockReader, + STableBlockScanInfo* pScanInfo, int64_t ts, STsdbReader* pReader, + bool* copied) { + int32_t code = TSDB_CODE_SUCCESS; + + *copied = false; + + bool hasVal = nextRowFromLastBlocks(pLastBlockReader, pScanInfo, &pReader->info.verRange); + if (hasVal) { + int64_t next1 = getCurrentKeyInLastBlock(pLastBlockReader); + if (next1 != ts) { + code = doAppendRowFromFileBlock(pReader->resBlockInfo.pResBlock, pReader, fRow->pBlockData, fRow->iRow); + if (code) { + return code; + } + + *copied = true; + return code; + } + } else { + code = doAppendRowFromFileBlock(pReader->resBlockInfo.pResBlock, pReader, fRow->pBlockData, fRow->iRow); + if (code) { + return code; + } + + *copied = true; + return code; + } + + return code; +} + +static FORCE_INLINE STSchema* doGetSchemaForTSRow(int32_t sversion, STsdbReader* pReader, uint64_t uid) { + // always set the newest schema version in pReader->info.pSchema + if (pReader->info.pSchema == NULL) { + STSchema* ps = getTableSchemaImpl(pReader, uid); + if (ps == NULL) { + return NULL; + } + } + + if (pReader->info.pSchema && sversion == pReader->info.pSchema->version) { + return pReader->info.pSchema; + } + + void** p = tSimpleHashGet(pReader->pSchemaMap, &sversion, sizeof(sversion)); + if (p != NULL) { + return *(STSchema**)p; + } + + STSchema* ptr = NULL; + int32_t code = metaGetTbTSchemaEx(pReader->pTsdb->pVnode->pMeta, pReader->info.suid, uid, sversion, &ptr); + if (code != TSDB_CODE_SUCCESS) { + terrno = code; + return NULL; + } else { + code = tSimpleHashPut(pReader->pSchemaMap, &sversion, sizeof(sversion), &ptr, POINTER_BYTES); + if (code != TSDB_CODE_SUCCESS) { + terrno = code; + return NULL; + } + return ptr; + } +} + +static int32_t doMergeBufAndFileRows(STsdbReader* pReader, STableBlockScanInfo* pBlockScanInfo, TSDBROW* pRow, + SIterInfo* pIter, int64_t key, SLastBlockReader* pLastBlockReader) { + SRowMerger* pMerger = &pReader->status.merger; + SRow* pTSRow = NULL; + SBlockData* pBlockData = &pReader->status.fileBlockData; + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + + int64_t tsLast = INT64_MIN; + if (hasDataInLastBlock(pLastBlockReader)) { + tsLast = getCurrentKeyInLastBlock(pLastBlockReader); + } + + TSDBKEY k = TSDBROW_KEY(pRow); + TSDBROW fRow = tsdbRowFromBlockData(pBlockData, pDumpInfo->rowIndex); + + // merge is not initialized yet, due to the fact that the pReader->info.pSchema is not initialized + if (pMerger->pArray == NULL) { + ASSERT(pReader->info.pSchema == NULL); + STSchema* ps = getTableSchemaImpl(pReader, pBlockScanInfo->uid); + if (ps == NULL) { + return terrno; + } + } + + int64_t minKey = 0; + if (pReader->info.order == TSDB_ORDER_ASC) { + minKey = INT64_MAX; // chosen the minimum value + if (minKey > tsLast && hasDataInLastBlock(pLastBlockReader)) { + minKey = tsLast; + } + + if (minKey > k.ts) { + minKey = k.ts; + } + + if (minKey > key && hasDataInFileBlock(pBlockData, pDumpInfo)) { + minKey = key; + } + } else { + minKey = INT64_MIN; + if (minKey < tsLast && hasDataInLastBlock(pLastBlockReader)) { + minKey = tsLast; + } + + if (minKey < k.ts) { + minKey = k.ts; + } + + if (minKey < key && hasDataInFileBlock(pBlockData, pDumpInfo)) { + minKey = key; + } + } + + // todo remove init + bool init = false; + + // ASC: file block ---> last block -----> imem -----> mem + // DESC: mem -----> imem -----> last block -----> file block + if (pReader->info.order == TSDB_ORDER_ASC) { + if (minKey == key) { + init = true; + int32_t code = tsdbRowMergerAdd(pMerger, &fRow, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader); + } + + if (minKey == tsLast) { + TSDBROW* fRow1 = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + if (init) { + tsdbRowMergerAdd(pMerger, fRow1, NULL); + } else { + init = true; + int32_t code = tsdbRowMergerAdd(pMerger, fRow1, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, tsLast, pMerger, &pReader->info.verRange, + pReader->idStr); + } + + if (minKey == k.ts) { + STSchema* pSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(pRow), pReader, pBlockScanInfo->uid); + if (pSchema == NULL) { + return terrno; + } + if (init) { + tsdbRowMergerAdd(pMerger, pRow, pSchema); + } else { + init = true; + int32_t code = tsdbRowMergerAdd(pMerger, pRow, pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + int32_t code = doMergeRowsInBuf(pIter, pBlockScanInfo->uid, k.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + } else { + if (minKey == k.ts) { + init = true; + STSchema* pSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(pRow), pReader, pBlockScanInfo->uid); + if (pSchema == NULL) { + return terrno; + } + + int32_t code = tsdbRowMergerAdd(pMerger, pRow, pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doMergeRowsInBuf(pIter, pBlockScanInfo->uid, k.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS || pMerger->pTSchema == NULL) { + return code; + } + } + + if (minKey == tsLast) { + TSDBROW* fRow1 = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + if (init) { + tsdbRowMergerAdd(pMerger, fRow1, NULL); + } else { + init = true; + int32_t code = tsdbRowMergerAdd(pMerger, fRow1, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, tsLast, pMerger, &pReader->info.verRange, + pReader->idStr); + } + + if (minKey == key) { + if (init) { + tsdbRowMergerAdd(pMerger, &fRow, NULL); + } else { + init = true; + int32_t code = tsdbRowMergerAdd(pMerger, &fRow, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader); + } + } + + int32_t code = tsdbRowMergerGetRow(pMerger, &pTSRow); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doAppendRowFromTSRow(pReader->resBlockInfo.pResBlock, pReader, pTSRow, pBlockScanInfo); + + taosMemoryFree(pTSRow); + tsdbRowMergerClear(pMerger); + + return code; +} + +static int32_t doMergeFileBlockAndLastBlock(SLastBlockReader* pLastBlockReader, STsdbReader* pReader, + STableBlockScanInfo* pBlockScanInfo, SBlockData* pBlockData, + bool mergeBlockData) { + SRowMerger* pMerger = &pReader->status.merger; + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + + int64_t tsLastBlock = getCurrentKeyInLastBlock(pLastBlockReader); + bool copied = false; + int32_t code = TSDB_CODE_SUCCESS; + SRow* pTSRow = NULL; + TSDBROW* pRow = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + + // create local variable to hold the row value + TSDBROW fRow = {.iRow = pRow->iRow, .type = TSDBROW_COL_FMT, .pBlockData = pRow->pBlockData}; + + tsdbTrace("fRow ptr:%p, %d, uid:%" PRIu64 ", %s", pRow->pBlockData, pRow->iRow, pLastBlockReader->uid, + pReader->idStr); + + // only last block exists + if ((!mergeBlockData) || (tsLastBlock != pBlockData->aTSKEY[pDumpInfo->rowIndex])) { + code = tryCopyDistinctRowFromSttBlock(&fRow, pLastBlockReader, pBlockScanInfo, tsLastBlock, pReader, &copied); + if (code) { + return code; + } + + if (copied) { + pBlockScanInfo->lastKey = tsLastBlock; + return TSDB_CODE_SUCCESS; + } else { + code = tsdbRowMergerAdd(pMerger, &fRow, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + TSDBROW* pRow1 = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + tsdbRowMergerAdd(pMerger, pRow1, NULL); + doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, tsLastBlock, pMerger, &pReader->info.verRange, + pReader->idStr); + + code = tsdbRowMergerGetRow(pMerger, &pTSRow); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doAppendRowFromTSRow(pReader->resBlockInfo.pResBlock, pReader, pTSRow, pBlockScanInfo); + + taosMemoryFree(pTSRow); + tsdbRowMergerClear(pMerger); + + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + } else { // not merge block data + code = tsdbRowMergerAdd(pMerger, &fRow, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, tsLastBlock, pMerger, &pReader->info.verRange, + pReader->idStr); + + // merge with block data if ts == key + if (tsLastBlock == pBlockData->aTSKEY[pDumpInfo->rowIndex]) { + doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader); + } + + code = tsdbRowMergerGetRow(pMerger, &pTSRow); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doAppendRowFromTSRow(pReader->resBlockInfo.pResBlock, pReader, pTSRow, pBlockScanInfo); + + taosMemoryFree(pTSRow); + tsdbRowMergerClear(pMerger); + + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + return TSDB_CODE_SUCCESS; +} + +static int32_t mergeFileBlockAndLastBlock(STsdbReader* pReader, SLastBlockReader* pLastBlockReader, int64_t key, + STableBlockScanInfo* pBlockScanInfo, SBlockData* pBlockData) { + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + SRowMerger* pMerger = &pReader->status.merger; + + // merge is not initialized yet, due to the fact that the pReader->info.pSchema is not initialized + if (pMerger->pArray == NULL) { + ASSERT(pReader->info.pSchema == NULL); + STSchema* ps = getTableSchemaImpl(pReader, pBlockScanInfo->uid); + if (ps == NULL) { + return terrno; + } + } + + if (hasDataInFileBlock(pBlockData, pDumpInfo)) { + // no last block available, only data block exists + if (!hasDataInLastBlock(pLastBlockReader)) { + return mergeRowsInFileBlocks(pBlockData, pBlockScanInfo, key, pReader); + } + + // row in last file block + TSDBROW fRow = tsdbRowFromBlockData(pBlockData, pDumpInfo->rowIndex); + int64_t ts = getCurrentKeyInLastBlock(pLastBlockReader); + + if (ASCENDING_TRAVERSE(pReader->info.order)) { + if (key < ts) { // imem, mem are all empty, file blocks (data blocks and last block) exist + return mergeRowsInFileBlocks(pBlockData, pBlockScanInfo, key, pReader); + } else if (key == ts) { + SRow* pTSRow = NULL; + int32_t code = tsdbRowMergerAdd(pMerger, &fRow, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader); + + TSDBROW* pRow1 = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + tsdbRowMergerAdd(pMerger, pRow1, NULL); + + doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, ts, pMerger, &pReader->info.verRange, pReader->idStr); + + code = tsdbRowMergerGetRow(pMerger, &pTSRow); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doAppendRowFromTSRow(pReader->resBlockInfo.pResBlock, pReader, pTSRow, pBlockScanInfo); + + taosMemoryFree(pTSRow); + tsdbRowMergerClear(pMerger); + return code; + } else { // key > ts + return doMergeFileBlockAndLastBlock(pLastBlockReader, pReader, pBlockScanInfo, NULL, false); + } + } else { // desc order + return doMergeFileBlockAndLastBlock(pLastBlockReader, pReader, pBlockScanInfo, pBlockData, true); + } + } else { // only last block exists + return doMergeFileBlockAndLastBlock(pLastBlockReader, pReader, pBlockScanInfo, NULL, false); + } +} + +static int32_t doMergeMultiLevelRows(STsdbReader* pReader, STableBlockScanInfo* pBlockScanInfo, SBlockData* pBlockData, + SLastBlockReader* pLastBlockReader) { + SRowMerger* pMerger = &pReader->status.merger; + SRow* pTSRow = NULL; + int32_t code = TSDB_CODE_SUCCESS; + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + SArray* pDelList = pBlockScanInfo->delSkyline; + + TSDBROW* pRow = getValidMemRow(&pBlockScanInfo->iter, pDelList, pReader); + TSDBROW* piRow = getValidMemRow(&pBlockScanInfo->iiter, pDelList, pReader); + + int64_t tsLast = INT64_MIN; + if (hasDataInLastBlock(pLastBlockReader)) { + tsLast = getCurrentKeyInLastBlock(pLastBlockReader); + } + + int64_t key = hasDataInFileBlock(pBlockData, pDumpInfo) ? pBlockData->aTSKEY[pDumpInfo->rowIndex] : INT64_MIN; + + TSDBKEY k = TSDBROW_KEY(pRow); + TSDBKEY ik = TSDBROW_KEY(piRow); + STSchema* pSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(pRow), pReader, pBlockScanInfo->uid); + if (pSchema == NULL) { + return code; + } + + STSchema* piSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(piRow), pReader, pBlockScanInfo->uid); + if (piSchema == NULL) { + return code; + } + + // merge is not initialized yet, due to the fact that the pReader->info.pSchema is not initialized + if (pMerger->pArray == NULL) { + ASSERT(pReader->info.pSchema == NULL); + STSchema* ps = getTableSchemaImpl(pReader, pBlockScanInfo->uid); + if (ps == NULL) { + return terrno; + } + } + + int64_t minKey = 0; + if (ASCENDING_TRAVERSE(pReader->info.order)) { + minKey = INT64_MAX; // let's find the minimum + if (minKey > k.ts) { + minKey = k.ts; + } + + if (minKey > ik.ts) { + minKey = ik.ts; + } + + if (minKey > key && hasDataInFileBlock(pBlockData, pDumpInfo)) { + minKey = key; + } + + if (minKey > tsLast && hasDataInLastBlock(pLastBlockReader)) { + minKey = tsLast; + } + } else { + minKey = INT64_MIN; // let find the maximum ts value + if (minKey < k.ts) { + minKey = k.ts; + } + + if (minKey < ik.ts) { + minKey = ik.ts; + } + + if (minKey < key && hasDataInFileBlock(pBlockData, pDumpInfo)) { + minKey = key; + } + + if (minKey < tsLast && hasDataInLastBlock(pLastBlockReader)) { + minKey = tsLast; + } + } + + bool init = false; + + // ASC: file block -----> last block -----> imem -----> mem + // DESC: mem -----> imem -----> last block -----> file block + if (ASCENDING_TRAVERSE(pReader->info.order)) { + if (minKey == key) { + init = true; + TSDBROW fRow = tsdbRowFromBlockData(pBlockData, pDumpInfo->rowIndex); + code = tsdbRowMergerAdd(pMerger, &fRow, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader); + } + + if (minKey == tsLast) { + TSDBROW* pRow1 = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + if (init) { + tsdbRowMergerAdd(pMerger, pRow1, NULL); + } else { + init = true; + code = tsdbRowMergerAdd(pMerger, pRow1, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, tsLast, pMerger, &pReader->info.verRange, + pReader->idStr); + } + + if (minKey == ik.ts) { + if (init) { + tsdbRowMergerAdd(pMerger, piRow, piSchema); + } else { + init = true; + code = tsdbRowMergerAdd(pMerger, piRow, piSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + code = doMergeRowsInBuf(&pBlockScanInfo->iiter, pBlockScanInfo->uid, ik.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + if (minKey == k.ts) { + if (init) { + tsdbRowMergerAdd(pMerger, pRow, pSchema); + } else { + // STSchema* pSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(pRow), pReader, pBlockScanInfo->uid); + code = tsdbRowMergerAdd(pMerger, pRow, pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + code = doMergeRowsInBuf(&pBlockScanInfo->iter, pBlockScanInfo->uid, k.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + } else { + if (minKey == k.ts) { + init = true; + code = tsdbRowMergerAdd(pMerger, pRow, pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doMergeRowsInBuf(&pBlockScanInfo->iter, pBlockScanInfo->uid, k.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + if (minKey == ik.ts) { + if (init) { + tsdbRowMergerAdd(pMerger, piRow, piSchema); + } else { + init = true; + code = tsdbRowMergerAdd(pMerger, piRow, piSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + code = doMergeRowsInBuf(&pBlockScanInfo->iiter, pBlockScanInfo->uid, ik.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + if (minKey == tsLast) { + TSDBROW* pRow1 = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + if (init) { + tsdbRowMergerAdd(pMerger, pRow1, NULL); + } else { + init = true; + code = tsdbRowMergerAdd(pMerger, pRow1, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + doMergeRowsInLastBlock(pLastBlockReader, pBlockScanInfo, tsLast, pMerger, &pReader->info.verRange, + pReader->idStr); + } + + if (minKey == key) { + TSDBROW fRow = tsdbRowFromBlockData(pBlockData, pDumpInfo->rowIndex); + if (!init) { + code = tsdbRowMergerAdd(pMerger, &fRow, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } else { + tsdbRowMergerAdd(pMerger, &fRow, NULL); + } + doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader); + } + } + + code = tsdbRowMergerGetRow(pMerger, &pTSRow); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doAppendRowFromTSRow(pReader->resBlockInfo.pResBlock, pReader, pTSRow, pBlockScanInfo); + + taosMemoryFree(pTSRow); + tsdbRowMergerClear(pMerger); + return code; +} + +int32_t doInitMemDataIter(STsdbReader* pReader, STbData** pData, STableBlockScanInfo* pBlockScanInfo, TSDBKEY* pKey, + SMemTable* pMem, SIterInfo* pIter, const char* type) { + int32_t code = TSDB_CODE_SUCCESS; + int32_t backward = (!ASCENDING_TRAVERSE(pReader->info.order)); + pIter->hasVal = false; + + if (pMem != NULL) { + *pData = tsdbGetTbDataFromMemTable(pMem, pReader->info.suid, pBlockScanInfo->uid); + + if ((*pData) != NULL) { + code = tsdbTbDataIterCreate((*pData), pKey, backward, &pIter->iter); + if (code == TSDB_CODE_SUCCESS) { + pIter->hasVal = (tsdbTbDataIterGet(pIter->iter) != NULL); + + tsdbDebug("%p uid:%" PRIu64 ", check data in %s from skey:%" PRId64 ", order:%d, ts range in buf:%" PRId64 + "-%" PRId64 " %s", + pReader, pBlockScanInfo->uid, type, pKey->ts, pReader->info.order, (*pData)->minKey, (*pData)->maxKey, + pReader->idStr); + } else { + tsdbError("%p uid:%" PRIu64 ", failed to create iterator for %s, code:%s, %s", pReader, pBlockScanInfo->uid, + type, tstrerror(code), pReader->idStr); + return code; + } + } + } else { + tsdbDebug("%p uid:%" PRIu64 ", no data in %s, %s", pReader, pBlockScanInfo->uid, type, pReader->idStr); + } + + return code; +} + +static int32_t initMemDataIterator(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader) { + if (pBlockScanInfo->iterInit) { + return TSDB_CODE_SUCCESS; + } + + STbData* d = NULL; + TSDBKEY startKey = {0}; + if (ASCENDING_TRAVERSE(pReader->info.order)) { + startKey = (TSDBKEY){.ts = pBlockScanInfo->lastKey + 1, .version = pReader->info.verRange.minVer}; + } else { + startKey = (TSDBKEY){.ts = pBlockScanInfo->lastKey - 1, .version = pReader->info.verRange.maxVer}; + } + + int32_t code = + doInitMemDataIter(pReader, &d, pBlockScanInfo, &startKey, pReader->pReadSnap->pMem, &pBlockScanInfo->iter, "mem"); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + STbData* di = NULL; + code = doInitMemDataIter(pReader, &di, pBlockScanInfo, &startKey, pReader->pReadSnap->pIMem, &pBlockScanInfo->iiter, + "imem"); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + loadMemTombData(&pBlockScanInfo->pMemDelData, d, di, pReader->info.verRange.maxVer); + + pBlockScanInfo->iterInit = true; + return TSDB_CODE_SUCCESS; +} + +static bool isValidFileBlockRow(SBlockData* pBlockData, SFileBlockDumpInfo* pDumpInfo, + STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader) { + // it is an multi-table data block + if (pBlockData->aUid != NULL) { + uint64_t uid = pBlockData->aUid[pDumpInfo->rowIndex]; + if (uid != pBlockScanInfo->uid) { // move to next row + return false; + } + } + + // check for version and time range + int64_t ver = pBlockData->aVersion[pDumpInfo->rowIndex]; + if (ver > pReader->info.verRange.maxVer || ver < pReader->info.verRange.minVer) { + return false; + } + + int64_t ts = pBlockData->aTSKEY[pDumpInfo->rowIndex]; + if (ts > pReader->info.window.ekey || ts < pReader->info.window.skey) { + return false; + } + + if (hasBeenDropped(pBlockScanInfo->delSkyline, &pBlockScanInfo->fileDelIndex, ts, ver, pReader->info.order, + &pReader->info.verRange)) { + return false; + } + + return true; +} + +static bool initLastBlockReader(SLastBlockReader* pLBlockReader, STableBlockScanInfo* pScanInfo, STsdbReader* pReader) { + // the last block reader has been initialized for this table. + if (pLBlockReader->uid == pScanInfo->uid) { + return hasDataInLastBlock(pLBlockReader); + } + + if (pLBlockReader->uid != 0) { + tMergeTreeClose(&pLBlockReader->mergeTree); + } + + pLBlockReader->uid = pScanInfo->uid; + + STimeWindow w = pLBlockReader->window; + if (ASCENDING_TRAVERSE(pLBlockReader->order)) { + w.skey = pScanInfo->lastKeyInStt; + } else { + w.ekey = pScanInfo->lastKeyInStt; + } + + int64_t st = taosGetTimestampUs(); + tsdbDebug("init last block reader, window:%" PRId64 "-%" PRId64 ", uid:%" PRIu64 ", %s", w.skey, w.ekey, + pScanInfo->uid, pReader->idStr); + + SMergeTreeConf conf = { + .uid = pScanInfo->uid, + .suid = pReader->info.suid, + .pTsdb = pReader->pTsdb, + .timewindow = w, + .verRange = pLBlockReader->verRange, + .strictTimeRange = false, + .pSchema = pReader->info.pSchema, + .pCurrentFileset = pReader->status.pCurrentFileset, + .backward = (pLBlockReader->order == TSDB_ORDER_DESC), + .pSttFileBlockIterArray = pReader->status.pLDataIterArray, + .pCols = pReader->suppInfo.colId, + .numOfCols = pReader->suppInfo.numOfCols, + .loadTombFn = loadSttTombDataForAll, + .pReader = pReader, + .idstr = pReader->idStr, + }; + + int32_t code = tMergeTreeOpen2(&pLBlockReader->mergeTree, &conf); + if (code != TSDB_CODE_SUCCESS) { + return false; + } + + initMemDataIterator(pScanInfo, pReader); + initDelSkylineIterator(pScanInfo, pReader->info.order, &pReader->cost); + + code = nextRowFromLastBlocks(pLBlockReader, pScanInfo, &pReader->info.verRange); + + int64_t el = taosGetTimestampUs() - st; + pReader->cost.initLastBlockReader += (el / 1000.0); + + tsdbDebug("init last block reader completed, elapsed time:%" PRId64 "us %s", el, pReader->idStr); + return code; +} + +static bool hasDataInLastBlock(SLastBlockReader* pLastBlockReader) { return pLastBlockReader->mergeTree.pIter != NULL; } + +bool hasDataInFileBlock(const SBlockData* pBlockData, const SFileBlockDumpInfo* pDumpInfo) { + if ((pBlockData->nRow > 0) && (pBlockData->nRow != pDumpInfo->totalRows)) { + return false; // this is an invalid result. + } + return pBlockData->nRow > 0 && (!pDumpInfo->allDumped); +} + +int32_t mergeRowsInFileBlocks(SBlockData* pBlockData, STableBlockScanInfo* pBlockScanInfo, int64_t key, + STsdbReader* pReader) { + SRowMerger* pMerger = &pReader->status.merger; + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + bool copied = false; + + int32_t code = tryCopyDistinctRowFromFileBlock(pReader, pBlockData, key, pDumpInfo, &copied); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + // merge is not initialized yet, due to the fact that the pReader->info.pSchema is not initialized + if (pMerger->pArray == NULL) { + ASSERT(pReader->info.pSchema == NULL); + STSchema* ps = getTableSchemaImpl(pReader, pBlockScanInfo->uid); + if (ps == NULL) { + return terrno; + } + } + + if (copied) { + pBlockScanInfo->lastKey = key; + return TSDB_CODE_SUCCESS; + } else { + TSDBROW fRow = tsdbRowFromBlockData(pBlockData, pDumpInfo->rowIndex); + + SRow* pTSRow = NULL; + code = tsdbRowMergerAdd(pMerger, &fRow, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + doMergeRowsInFileBlocks(pBlockData, pBlockScanInfo, pReader); + code = tsdbRowMergerGetRow(pMerger, &pTSRow); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doAppendRowFromTSRow(pReader->resBlockInfo.pResBlock, pReader, pTSRow, pBlockScanInfo); + + taosMemoryFree(pTSRow); + tsdbRowMergerClear(pMerger); + return code; + } +} + +static int32_t buildComposedDataBlockImpl(STsdbReader* pReader, STableBlockScanInfo* pBlockScanInfo, + SBlockData* pBlockData, SLastBlockReader* pLastBlockReader) { + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + + TSDBROW *pRow = NULL, *piRow = NULL; + int64_t key = (pBlockData->nRow > 0 && (!pDumpInfo->allDumped)) ? pBlockData->aTSKEY[pDumpInfo->rowIndex] : INT64_MIN; + if (pBlockScanInfo->iter.hasVal) { + pRow = getValidMemRow(&pBlockScanInfo->iter, pBlockScanInfo->delSkyline, pReader); + } + + if (pBlockScanInfo->iiter.hasVal) { + piRow = getValidMemRow(&pBlockScanInfo->iiter, pBlockScanInfo->delSkyline, pReader); + } + + // two levels of mem-table does contain the valid rows + if (pRow != NULL && piRow != NULL) { + return doMergeMultiLevelRows(pReader, pBlockScanInfo, pBlockData, pLastBlockReader); + } + + // imem + file + last block + if (pBlockScanInfo->iiter.hasVal) { + return doMergeBufAndFileRows(pReader, pBlockScanInfo, piRow, &pBlockScanInfo->iiter, key, pLastBlockReader); + } + + // mem + file + last block + if (pBlockScanInfo->iter.hasVal) { + return doMergeBufAndFileRows(pReader, pBlockScanInfo, pRow, &pBlockScanInfo->iter, key, pLastBlockReader); + } + + // files data blocks + last block + return mergeFileBlockAndLastBlock(pReader, pLastBlockReader, key, pBlockScanInfo, pBlockData); +} + +static int32_t loadNeighborIfOverlap(SFileDataBlockInfo* pBlockInfo, STableBlockScanInfo* pBlockScanInfo, + STsdbReader* pReader, bool* loadNeighbor) { + int32_t code = TSDB_CODE_SUCCESS; + int32_t step = ASCENDING_TRAVERSE(pReader->info.order) ? 1 : -1; + int32_t nextIndex = -1; + + *loadNeighbor = false; + + SBrinRecord rec = {0}; + bool hasNeighbor = getNeighborBlockOfSameTable(pBlockInfo, pBlockScanInfo, &nextIndex, pReader->info.order, &rec); + if (!hasNeighbor) { // do nothing + return code; + } + + if (overlapWithNeighborBlock2(pBlockInfo, &rec, pReader->info.order)) { // load next block + SReaderStatus* pStatus = &pReader->status; + SDataBlockIter* pBlockIter = &pStatus->blockIter; + + // 1. find the next neighbor block in the scan block list + SFileDataBlockInfo fb = {.uid = pBlockInfo->uid, .tbBlockIdx = nextIndex}; + int32_t neighborIndex = findFileBlockInfoIndex(pBlockIter, &fb); + + // 2. remove it from the scan block list + setFileBlockActiveInBlockIter(pBlockIter, neighborIndex, step); + + // 3. load the neighbor block, and set it to be the currently accessed file data block + code = doLoadFileBlockData(pReader, pBlockIter, &pStatus->fileBlockData, pBlockInfo->uid); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + // 4. check the data values + initBlockDumpInfo(pReader, pBlockIter); + *loadNeighbor = true; + } + + return code; +} + +static void updateComposedBlockInfo(STsdbReader* pReader, double el, STableBlockScanInfo* pBlockScanInfo) { + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + + pResBlock->info.id.uid = (pBlockScanInfo != NULL) ? pBlockScanInfo->uid : 0; + pResBlock->info.dataLoad = 1; + blockDataUpdateTsWindow(pResBlock, pReader->suppInfo.slotId[0]); + + setComposedBlockFlag(pReader, true); + + pReader->cost.composedBlocks += 1; + pReader->cost.buildComposedBlockTime += el; +} + +static int32_t buildComposedDataBlock(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(&pReader->status.blockIter); + SLastBlockReader* pLastBlockReader = pReader->status.fileIter.pLastBlockReader; + + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + int64_t st = taosGetTimestampUs(); + int32_t step = asc ? 1 : -1; + double el = 0; + SBrinRecord* pRecord = &pBlockInfo->record; + + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + + STableBlockScanInfo* pBlockScanInfo = NULL; + if (pBlockInfo != NULL) { + if (pReader->pIgnoreTables && taosHashGet(*pReader->pIgnoreTables, &pBlockInfo->uid, sizeof(pBlockInfo->uid))) { + setBlockAllDumped(pDumpInfo, pRecord->lastKey, pReader->info.order); + return code; + } + + pBlockScanInfo = getTableBlockScanInfo(pReader->status.pTableMap, pBlockInfo->uid, pReader->idStr); + if (pBlockScanInfo == NULL) { + goto _end; + } + + pRecord = &pBlockInfo->record; + TSDBKEY keyInBuf = getCurrentKeyInBuf(pBlockScanInfo, pReader); + + // it is a clean block, load it directly + if (isCleanFileDataBlock(pReader, pBlockInfo, pBlockScanInfo, keyInBuf, pLastBlockReader) && + (pRecord->numRow <= pReader->resBlockInfo.capacity)) { + if (asc || (!hasDataInLastBlock(pLastBlockReader))) { + code = copyBlockDataToSDataBlock(pReader); + if (code) { + goto _end; + } + + // record the last key value + pBlockScanInfo->lastKey = asc ? pRecord->lastKey : pRecord->firstKey; + goto _end; + } + } + } else { // file blocks not exist + ASSERT(0); + pBlockScanInfo = *pReader->status.pTableIter; + if (pReader->pIgnoreTables && + taosHashGet(*pReader->pIgnoreTables, &pBlockScanInfo->uid, sizeof(pBlockScanInfo->uid))) { + // setBlockAllDumped(pDumpInfo, pBlock->maxKey.ts, pReader->info.order); + return code; + } + } + + SBlockData* pBlockData = &pReader->status.fileBlockData; + + while (1) { + bool hasBlockData = false; + { + while (pBlockData->nRow > 0 && pBlockData->uid == pBlockScanInfo->uid) { + // find the first qualified row in data block + if (isValidFileBlockRow(pBlockData, pDumpInfo, pBlockScanInfo, pReader)) { + hasBlockData = true; + break; + } + + pDumpInfo->rowIndex += step; + + if (pDumpInfo->rowIndex >= pBlockData->nRow || pDumpInfo->rowIndex < 0) { + pBlockInfo = getCurrentBlockInfo(&pReader->status.blockIter); // NOTE: get the new block info + + // continue check for the next file block if the last ts in the current block + // is overlapped with the next neighbor block + bool loadNeighbor = false; + code = loadNeighborIfOverlap(pBlockInfo, pBlockScanInfo, pReader, &loadNeighbor); + if ((!loadNeighbor) || (code != 0)) { + setBlockAllDumped(pDumpInfo, pRecord->lastKey, pReader->info.order); + break; + } + } + } + } + + // no data in last block and block, no need to proceed. + if (hasBlockData == false) { + break; + } + + code = buildComposedDataBlockImpl(pReader, pBlockScanInfo, pBlockData, pLastBlockReader); + if (code) { + goto _end; + } + + // currently loaded file data block is consumed + if ((pBlockData->nRow > 0) && (pDumpInfo->rowIndex >= pBlockData->nRow || pDumpInfo->rowIndex < 0)) { + setBlockAllDumped(pDumpInfo, pRecord->lastKey, pReader->info.order); + break; + } + + if (pResBlock->info.rows >= pReader->resBlockInfo.capacity) { + break; + } + } + +_end: + el = (taosGetTimestampUs() - st) / 1000.0; + updateComposedBlockInfo(pReader, el, pBlockScanInfo); + + if (pResBlock->info.rows > 0) { + tsdbDebug("%p uid:%" PRIu64 ", composed data block created, brange:%" PRIu64 "-%" PRIu64 " rows:%" PRId64 + ", elapsed time:%.2f ms %s", + pReader, pResBlock->info.id.uid, pResBlock->info.window.skey, pResBlock->info.window.ekey, + pResBlock->info.rows, el, pReader->idStr); + } + + return code; +} + +void setComposedBlockFlag(STsdbReader* pReader, bool composed) { pReader->status.composedDataBlock = composed; } + +int32_t getInitialDelIndex(const SArray* pDelSkyline, int32_t order) { + if (pDelSkyline == NULL) { + return 0; + } + + return ASCENDING_TRAVERSE(order) ? 0 : taosArrayGetSize(pDelSkyline) - 1; +} + +int32_t initDelSkylineIterator(STableBlockScanInfo* pBlockScanInfo, int32_t order, SCostSummary* pCost) { + int32_t code = 0; + int32_t newDelDataInFile = taosArrayGetSize(pBlockScanInfo->pfileDelData); + if (newDelDataInFile == 0 && + ((pBlockScanInfo->delSkyline != NULL) || (TARRAY_SIZE(pBlockScanInfo->pMemDelData) == 0))) { + return code; + } + + int64_t st = taosGetTimestampUs(); + + if (pBlockScanInfo->delSkyline != NULL) { + taosArrayClear(pBlockScanInfo->delSkyline); + } else { + pBlockScanInfo->delSkyline = taosArrayInit(4, sizeof(TSDBKEY)); + } + + SArray* pSource = pBlockScanInfo->pfileDelData; + if (pSource == NULL) { + pSource = pBlockScanInfo->pMemDelData; + } else { + taosArrayAddAll(pSource, pBlockScanInfo->pMemDelData); + } + + code = tsdbBuildDeleteSkyline(pSource, 0, taosArrayGetSize(pSource) - 1, pBlockScanInfo->delSkyline); + + taosArrayClear(pBlockScanInfo->pfileDelData); + int32_t index = getInitialDelIndex(pBlockScanInfo->delSkyline, order); + + pBlockScanInfo->iter.index = index; + pBlockScanInfo->iiter.index = index; + pBlockScanInfo->fileDelIndex = index; + pBlockScanInfo->lastBlockDelIndex = index; + + double el = taosGetTimestampUs() - st; + pCost->createSkylineIterTime = el / 1000.0; + + return code; +} + +TSDBKEY getCurrentKeyInBuf(STableBlockScanInfo* pScanInfo, STsdbReader* pReader) { + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + TSDBKEY key = {.ts = TSKEY_INITIAL_VAL}, ikey = {.ts = TSKEY_INITIAL_VAL}; + + bool hasKey = false, hasIKey = false; + TSDBROW* pRow = getValidMemRow(&pScanInfo->iter, pScanInfo->delSkyline, pReader); + if (pRow != NULL) { + hasKey = true; + key = TSDBROW_KEY(pRow); + } + + TSDBROW* pIRow = getValidMemRow(&pScanInfo->iiter, pScanInfo->delSkyline, pReader); + if (pIRow != NULL) { + hasIKey = true; + ikey = TSDBROW_KEY(pIRow); + } + + if (hasKey) { + if (hasIKey) { // has data in mem & imem + if (asc) { + return key.ts <= ikey.ts ? key : ikey; + } else { + return key.ts <= ikey.ts ? ikey : key; + } + } else { // no data in imem + return key; + } + } else { + // no data in mem & imem, return the initial value + // only imem has data, return ikey + return ikey; + } +} + +static int32_t moveToNextFile(STsdbReader* pReader, SBlockNumber* pBlockNum, SArray* pTableList) { + SReaderStatus* pStatus = &pReader->status; + pBlockNum->numOfBlocks = 0; + pBlockNum->numOfLastFiles = 0; + + size_t numOfTables = tSimpleHashGetSize(pReader->status.pTableMap); + SArray* pIndexList = taosArrayInit(numOfTables, sizeof(SBrinBlk)); + + while (1) { + // only check here, since the iterate data in memory is very fast. + if (pReader->code != TSDB_CODE_SUCCESS) { + tsdbWarn("tsdb reader is stopped ASAP, code:%s, %s", strerror(pReader->code), pReader->idStr); + return pReader->code; + } + + bool hasNext = false; + int32_t code = filesetIteratorNext(&pStatus->fileIter, pReader, &hasNext); + if (code != TSDB_CODE_SUCCESS) { + taosArrayDestroy(pIndexList); + return code; + } + + if (!hasNext) { // no data files on disk + break; + } + + taosArrayClear(pIndexList); + code = doLoadBlockIndex(pReader, pReader->pFileReader, pIndexList); + if (code != TSDB_CODE_SUCCESS) { + taosArrayDestroy(pIndexList); + return code; + } + + if (taosArrayGetSize(pIndexList) > 0 || pReader->status.pCurrentFileset->lvlArr->size > 0) { + code = doLoadFileBlock(pReader, pIndexList, pBlockNum, pTableList); + if (code != TSDB_CODE_SUCCESS) { + taosArrayDestroy(pIndexList); + return code; + } + + if (pBlockNum->numOfBlocks + pBlockNum->numOfLastFiles > 0) { + break; + } + } + + // no blocks in current file, try next files + } + + taosArrayDestroy(pIndexList); + return loadDataFileTombDataForAll(pReader); +} + +static void resetTableListIndex(SReaderStatus* pStatus) { + STableUidList* pList = &pStatus->uidList; + + pList->currentIndex = 0; + uint64_t uid = pList->tableUidList[0]; + pStatus->pTableIter = tSimpleHashGet(pStatus->pTableMap, &uid, sizeof(uid)); +} + +static bool moveToNextTable(STableUidList* pOrderedCheckInfo, SReaderStatus* pStatus) { + pOrderedCheckInfo->currentIndex += 1; + if (pOrderedCheckInfo->currentIndex >= tSimpleHashGetSize(pStatus->pTableMap)) { + pStatus->pTableIter = NULL; + return false; + } + + uint64_t uid = pOrderedCheckInfo->tableUidList[pOrderedCheckInfo->currentIndex]; + pStatus->pTableIter = tSimpleHashGet(pStatus->pTableMap, &uid, sizeof(uid)); + return (pStatus->pTableIter != NULL); +} + +static int32_t doLoadLastBlockSequentially(STsdbReader* pReader) { + SReaderStatus* pStatus = &pReader->status; + SLastBlockReader* pLastBlockReader = pStatus->fileIter.pLastBlockReader; + STableUidList* pUidList = &pStatus->uidList; + int32_t code = TSDB_CODE_SUCCESS; + + if (tSimpleHashGetSize(pStatus->pTableMap) == 0) { + return TSDB_CODE_SUCCESS; + } + + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + + while (1) { + if (pReader->code != TSDB_CODE_SUCCESS) { + tsdbWarn("tsdb reader is stopped ASAP, code:%s, %s", strerror(pReader->code), pReader->idStr); + return pReader->code; + } + + // load the last data block of current table + STableBlockScanInfo* pScanInfo = *(STableBlockScanInfo**)pStatus->pTableIter; + if (pReader->pIgnoreTables && taosHashGet(*pReader->pIgnoreTables, &pScanInfo->uid, sizeof(pScanInfo->uid))) { + // reset the index in last block when handing a new file + // doCleanupTableScanInfo(pScanInfo); + bool hasNexTable = moveToNextTable(pUidList, pStatus); + if (!hasNexTable) { + return TSDB_CODE_SUCCESS; + } + + continue; + } + + // reset the index in last block when handing a new file + // doCleanupTableScanInfo(pScanInfo); + + bool hasDataInLastFile = initLastBlockReader(pLastBlockReader, pScanInfo, pReader); + if (!hasDataInLastFile) { + bool hasNexTable = moveToNextTable(pUidList, pStatus); + if (!hasNexTable) { + return TSDB_CODE_SUCCESS; + } + + continue; + } + + int64_t st = taosGetTimestampUs(); + while (1) { + bool hasBlockLData = hasDataInLastBlock(pLastBlockReader); + + // no data in last block and block, no need to proceed. + if (hasBlockLData == false) { + break; + } + + code = buildComposedDataBlockImpl(pReader, pScanInfo, &pReader->status.fileBlockData, pLastBlockReader); + if (code) { + return code; + } + + if (pResBlock->info.rows >= pReader->resBlockInfo.capacity) { + break; + } + } + + double el = (taosGetTimestampUs() - st) / 1000.0; + updateComposedBlockInfo(pReader, el, pScanInfo); + + if (pResBlock->info.rows > 0) { + tsdbDebug("%p uid:%" PRIu64 ", composed data block created, brange:%" PRIu64 "-%" PRIu64 " rows:%" PRId64 + ", elapsed time:%.2f ms %s", + pReader, pResBlock->info.id.uid, pResBlock->info.window.skey, pResBlock->info.window.ekey, + pResBlock->info.rows, el, pReader->idStr); + return TSDB_CODE_SUCCESS; + } + + // current table is exhausted, let's try next table + bool hasNexTable = moveToNextTable(pUidList, pStatus); + if (!hasNexTable) { + return TSDB_CODE_SUCCESS; + } + } +} + +static int32_t doBuildDataBlock(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + + SReaderStatus* pStatus = &pReader->status; + SDataBlockIter* pBlockIter = &pStatus->blockIter; + STableBlockScanInfo* pScanInfo = NULL; + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(pBlockIter); + SLastBlockReader* pLastBlockReader = pReader->status.fileIter.pLastBlockReader; + + if (pReader->pIgnoreTables && taosHashGet(*pReader->pIgnoreTables, &pBlockInfo->uid, sizeof(pBlockInfo->uid))) { + setBlockAllDumped(&pStatus->fBlockDumpInfo, pBlockInfo->record.lastKey, pReader->info.order); + return code; + } + + if (pReader->code != TSDB_CODE_SUCCESS) { + return pReader->code; + } + + pScanInfo = getTableBlockScanInfo(pReader->status.pTableMap, pBlockInfo->uid, pReader->idStr); + if (pScanInfo == NULL) { + return terrno; + } + + initLastBlockReader(pLastBlockReader, pScanInfo, pReader); + TSDBKEY keyInBuf = getCurrentKeyInBuf(pScanInfo, pReader); + + if (fileBlockShouldLoad(pReader, pBlockInfo, pScanInfo, keyInBuf, pLastBlockReader)) { + code = doLoadFileBlockData(pReader, pBlockIter, &pStatus->fileBlockData, pScanInfo->uid); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + // build composed data block + code = buildComposedDataBlock(pReader); + } else if (bufferDataInFileBlockGap(pReader->info.order, keyInBuf, pBlockInfo)) { + // data in memory that are earlier than current file block + // rows in buffer should be less than the file block in asc, greater than file block in desc + int64_t endKey = + (ASCENDING_TRAVERSE(pReader->info.order)) ? pBlockInfo->record.firstKey : pBlockInfo->record.lastKey; + code = buildDataBlockFromBuf(pReader, pScanInfo, endKey); + } else { + if (hasDataInLastBlock(pLastBlockReader) && !ASCENDING_TRAVERSE(pReader->info.order)) { + // only return the rows in last block + int64_t tsLast = getCurrentKeyInLastBlock(pLastBlockReader); + ASSERT(tsLast >= pBlockInfo->record.lastKey); + + SBlockData* pBData = &pReader->status.fileBlockData; + tBlockDataReset(pBData); + + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + tsdbDebug("load data in last block firstly, due to desc scan data, %s", pReader->idStr); + + int64_t st = taosGetTimestampUs(); + + while (1) { + bool hasBlockLData = hasDataInLastBlock(pLastBlockReader); + + // no data in last block and block, no need to proceed. + if (hasBlockLData == false) { + break; + } + + code = buildComposedDataBlockImpl(pReader, pScanInfo, &pReader->status.fileBlockData, pLastBlockReader); + if (code) { + return code; + } + + if (pResBlock->info.rows >= pReader->resBlockInfo.capacity) { + break; + } + } + + double el = (taosGetTimestampUs() - st) / 1000.0; + updateComposedBlockInfo(pReader, el, pScanInfo); + + if (pResBlock->info.rows > 0) { + tsdbDebug("%p uid:%" PRIu64 ", composed data block created, brange:%" PRIu64 "-%" PRIu64 " rows:%" PRId64 + ", elapsed time:%.2f ms %s", + pReader, pResBlock->info.id.uid, pResBlock->info.window.skey, pResBlock->info.window.ekey, + pResBlock->info.rows, el, pReader->idStr); + } + } else { // whole block is required, return it directly + SDataBlockInfo* pInfo = &pReader->resBlockInfo.pResBlock->info; + pInfo->rows = pBlockInfo->record.numRow; + pInfo->id.uid = pScanInfo->uid; + pInfo->dataLoad = 0; + pInfo->window = (STimeWindow){.skey = pBlockInfo->record.firstKey, .ekey = pBlockInfo->record.lastKey}; + setComposedBlockFlag(pReader, false); + setBlockAllDumped(&pStatus->fBlockDumpInfo, pBlockInfo->record.lastKey, pReader->info.order); + + // update the last key for the corresponding table + pScanInfo->lastKey = ASCENDING_TRAVERSE(pReader->info.order) ? pInfo->window.ekey : pInfo->window.skey; + tsdbDebug("%p uid:%" PRIu64 + " clean file block retrieved from file, global index:%d, " + "table index:%d, rows:%d, brange:%" PRId64 "-%" PRId64 ", %s", + pReader, pScanInfo->uid, pBlockIter->index, pBlockInfo->tbBlockIdx, pBlockInfo->record.numRow, + pBlockInfo->record.firstKey, pBlockInfo->record.lastKey, pReader->idStr); + } + } + + return (pReader->code != TSDB_CODE_SUCCESS) ? pReader->code : code; +} + +static int32_t doSumFileBlockRows(STsdbReader* pReader, SDataFReader* pFileReader) { + int64_t st = taosGetTimestampUs(); + LRUHandle* handle = NULL; + int32_t code = tsdbCacheGetBlockIdx(pFileReader->pTsdb->biCache, pFileReader, &handle); + if (code != TSDB_CODE_SUCCESS || handle == NULL) { + goto _end; + } + + int32_t numOfTables = tSimpleHashGetSize(pReader->status.pTableMap); + + SArray* aBlockIdx = (SArray*)taosLRUCacheValue(pFileReader->pTsdb->biCache, handle); + size_t num = taosArrayGetSize(aBlockIdx); + if (num == 0) { + tsdbBICacheRelease(pFileReader->pTsdb->biCache, handle); + return TSDB_CODE_SUCCESS; + } + + SBlockIdx* pBlockIdx = NULL; + for (int32_t i = 0; i < num; ++i) { + pBlockIdx = (SBlockIdx*)taosArrayGet(aBlockIdx, i); + if (pBlockIdx->suid != pReader->info.suid) { + continue; + } + + STableBlockScanInfo** p = tSimpleHashGet(pReader->status.pTableMap, &pBlockIdx->uid, sizeof(pBlockIdx->uid)); + if (p == NULL) { + continue; + } + + STableBlockScanInfo* pScanInfo = *p; + SDataBlk block = {0}; + // for (int32_t j = 0; j < pScanInfo->mapData.nItem; ++j) { + // tGetDataBlk(pScanInfo->mapData.pData + pScanInfo->mapData.aOffset[j], &block); + // pReader->rowsNum += block.nRow; + // } + } + +_end: + tsdbBICacheRelease(pFileReader->pTsdb->biCache, handle); + return code; +} + +static int32_t doSumSttBlockRows(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + SLastBlockReader* pLastBlockReader = pReader->status.fileIter.pLastBlockReader; + SSttBlockLoadInfo* pBlockLoadInfo = NULL; +#if 0 + for (int32_t i = 0; i < pReader->pFileReader->pSet->nSttF; ++i) { // open all last file + pBlockLoadInfo = &pLastBlockReader->pInfo[i]; + + code = tsdbReadSttBlk(pReader->pFileReader, i, pBlockLoadInfo->aSttBlk); + if (code) { + return code; + } + + size_t size = taosArrayGetSize(pBlockLoadInfo->aSttBlk); + if (size >= 1) { + SSttBlk* pStart = taosArrayGet(pBlockLoadInfo->aSttBlk, 0); + SSttBlk* pEnd = taosArrayGet(pBlockLoadInfo->aSttBlk, size - 1); + + // all identical + if (pStart->suid == pEnd->suid) { + if (pStart->suid != pReader->info.suid) { + // no qualified stt block existed + taosArrayClear(pBlockLoadInfo->aSttBlk); + continue; + } + for (int32_t j = 0; j < size; ++j) { + SSttBlk* p = taosArrayGet(pBlockLoadInfo->aSttBlk, j); + pReader->rowsNum += p->nRow; + } + } else { + for (int32_t j = 0; j < size; ++j) { + SSttBlk* p = taosArrayGet(pBlockLoadInfo->aSttBlk, j); + uint64_t s = p->suid; + if (s < pReader->info.suid) { + continue; + } + + if (s == pReader->info.suid) { + pReader->rowsNum += p->nRow; + } else if (s > pReader->info.suid) { + break; + } + } + } + } + } +#endif + + return code; +} + +static int32_t readRowsCountFromFiles(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + + while (1) { + bool hasNext = false; + code = filesetIteratorNext(&pReader->status.fileIter, pReader, &hasNext); + if (code) { + return code; + } + + if (!hasNext) { // no data files on disk + break; + } + + // code = doSumFileBlockRows(pReader, pReader->pFileReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doSumSttBlockRows(pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + pReader->status.loadFromFile = false; + + return code; +} + +static int32_t readRowsCountFromMem(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + int64_t memNum = 0, imemNum = 0; + if (pReader->pReadSnap->pMem != NULL) { + tsdbMemTableCountRows(pReader->pReadSnap->pMem, pReader->status.pTableMap, &memNum); + } + + if (pReader->pReadSnap->pIMem != NULL) { + tsdbMemTableCountRows(pReader->pReadSnap->pIMem, pReader->status.pTableMap, &imemNum); + } + + pReader->rowsNum += memNum + imemNum; + + return code; +} + +static int32_t buildBlockFromBufferSequentially(STsdbReader* pReader) { + SReaderStatus* pStatus = &pReader->status; + STableUidList* pUidList = &pStatus->uidList; + + while (1) { + if (pReader->code != TSDB_CODE_SUCCESS) { + tsdbWarn("tsdb reader is stopped ASAP, code:%s, %s", strerror(pReader->code), pReader->idStr); + return pReader->code; + } + + STableBlockScanInfo** pBlockScanInfo = pStatus->pTableIter; + if (pReader->pIgnoreTables && + taosHashGet(*pReader->pIgnoreTables, &(*pBlockScanInfo)->uid, sizeof((*pBlockScanInfo)->uid))) { + bool hasNexTable = moveToNextTable(pUidList, pStatus); + if (!hasNexTable) { + return TSDB_CODE_SUCCESS; + } + pBlockScanInfo = pStatus->pTableIter; + } + + initMemDataIterator(*pBlockScanInfo, pReader); + initDelSkylineIterator(*pBlockScanInfo, pReader->info.order, &pReader->cost); + + int64_t endKey = (ASCENDING_TRAVERSE(pReader->info.order)) ? INT64_MAX : INT64_MIN; + int32_t code = buildDataBlockFromBuf(pReader, *pBlockScanInfo, endKey); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + if (pReader->resBlockInfo.pResBlock->info.rows > 0) { + return TSDB_CODE_SUCCESS; + } + + // current table is exhausted, let's try next table + bool hasNexTable = moveToNextTable(pUidList, pStatus); + if (!hasNexTable) { + return TSDB_CODE_SUCCESS; + } + } +} + +// set the correct start position in case of the first/last file block, according to the query time window +static void initBlockDumpInfo(STsdbReader* pReader, SDataBlockIter* pBlockIter) { + int64_t lastKey = ASCENDING_TRAVERSE(pReader->info.order) ? INT64_MIN : INT64_MAX; + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(pBlockIter); + SReaderStatus* pStatus = &pReader->status; + SFileBlockDumpInfo* pDumpInfo = &pStatus->fBlockDumpInfo; + + if (pBlockInfo) { + STableBlockScanInfo* pScanInfo = tSimpleHashGet(pBlockIter->pTableMap, &pBlockInfo->uid, sizeof(pBlockInfo->uid)); + if (pScanInfo) { + lastKey = pScanInfo->lastKey; + } + + pDumpInfo->totalRows = pBlockInfo->record.numRow; + pDumpInfo->rowIndex = ASCENDING_TRAVERSE(pReader->info.order) ? 0 : pBlockInfo->record.numRow - 1; + } else { + pDumpInfo->totalRows = 0; + pDumpInfo->rowIndex = 0; + } + + pDumpInfo->allDumped = false; + pDumpInfo->lastKey = lastKey; +} + +static int32_t initForFirstBlockInFile(STsdbReader* pReader, SDataBlockIter* pBlockIter) { + SBlockNumber num = {0}; + SArray* pTableList = taosArrayInit(40, POINTER_BYTES); + + int32_t code = moveToNextFile(pReader, &num, pTableList); + if (code != TSDB_CODE_SUCCESS) { + taosArrayDestroy(pTableList); + return code; + } + + // all data files are consumed, try data in buffer + if (num.numOfBlocks + num.numOfLastFiles == 0) { + pReader->status.loadFromFile = false; + taosArrayDestroy(pTableList); + return code; + } + + // initialize the block iterator for a new fileset + if (num.numOfBlocks > 0) { + code = initBlockIterator(pReader, pBlockIter, num.numOfBlocks, pTableList); + } else { // no block data, only last block exists + tBlockDataReset(&pReader->status.fileBlockData); + resetDataBlockIterator(pBlockIter, pReader->info.order); + resetTableListIndex(&pReader->status); + } + + // set the correct start position according to the query time window + initBlockDumpInfo(pReader, pBlockIter); + taosArrayDestroy(pTableList); + return code; +} + +static bool fileBlockPartiallyRead(SFileBlockDumpInfo* pDumpInfo, bool asc) { + return (!pDumpInfo->allDumped) && + ((pDumpInfo->rowIndex > 0 && asc) || (pDumpInfo->rowIndex < (pDumpInfo->totalRows - 1) && (!asc))); +} + +typedef enum { + TSDB_READ_RETURN = 0x1, + TSDB_READ_CONTINUE = 0x2, +} ERetrieveType; + +static ERetrieveType doReadDataFromLastFiles(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + SDataBlockIter* pBlockIter = &pReader->status.blockIter; + + while (1) { + terrno = 0; + + code = doLoadLastBlockSequentially(pReader); + if (code != TSDB_CODE_SUCCESS) { + terrno = code; + return TSDB_READ_RETURN; + } + + if (pResBlock->info.rows > 0) { + return TSDB_READ_RETURN; + } + + // all data blocks are checked in this last block file, now let's try the next file + ASSERT(pReader->status.pTableIter == NULL); + code = initForFirstBlockInFile(pReader, pBlockIter); + + // error happens or all the data files are completely checked + if ((code != TSDB_CODE_SUCCESS) || (pReader->status.loadFromFile == false)) { + terrno = code; + return TSDB_READ_RETURN; + } + + if (pBlockIter->numOfBlocks > 0) { // there are data blocks existed. + return TSDB_READ_CONTINUE; + } else { // all blocks in data file are checked, let's check the data in last files + resetTableListIndex(&pReader->status); + } + } +} + +static int32_t buildBlockFromFiles(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + + SDataBlockIter* pBlockIter = &pReader->status.blockIter; + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + + if (pBlockIter->numOfBlocks == 0) { + // let's try to extract data from stt files. + ERetrieveType type = doReadDataFromLastFiles(pReader); + if (type == TSDB_READ_RETURN) { + return terrno; + } + + code = doBuildDataBlock(pReader); + if (code != TSDB_CODE_SUCCESS || pResBlock->info.rows > 0) { + return code; + } + } + + while (1) { + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + + if (fileBlockPartiallyRead(pDumpInfo, asc)) { // file data block is partially loaded + code = buildComposedDataBlock(pReader); + } else { + // current block are exhausted, try the next file block + if (pDumpInfo->allDumped) { + // try next data block in current file + bool hasNext = blockIteratorNext(&pReader->status.blockIter, pReader->idStr); + if (hasNext) { // check for the next block in the block accessed order list + initBlockDumpInfo(pReader, pBlockIter); + } else { + // all data blocks in files are checked, let's check the data in last files. + // data blocks in current file are exhausted, let's try the next file now + SBlockData* pBlockData = &pReader->status.fileBlockData; + if (pBlockData->uid != 0) { + tBlockDataClear(pBlockData); + } + + tBlockDataReset(pBlockData); + resetDataBlockIterator(pBlockIter, pReader->info.order); + resetTableListIndex(&pReader->status); + + ERetrieveType type = doReadDataFromLastFiles(pReader); + if (type == TSDB_READ_RETURN) { + return terrno; + } + } + } + + code = doBuildDataBlock(pReader); + } + + if (code != TSDB_CODE_SUCCESS || pResBlock->info.rows > 0) { + return code; + } + } +} + +static STsdb* getTsdbByRetentions(SVnode* pVnode, TSKEY winSKey, SRetention* retentions, const char* idStr, + int8_t* pLevel) { + if (VND_IS_RSMA(pVnode)) { + int8_t level = 0; + int8_t precision = pVnode->config.tsdbCfg.precision; + int64_t now = taosGetTimestamp(precision); + int64_t offset = tsQueryRsmaTolerance * ((precision == TSDB_TIME_PRECISION_MILLI) ? 1L + : (precision == TSDB_TIME_PRECISION_MICRO) ? 1000L + : 1000000L); + + for (int8_t i = 0; i < TSDB_RETENTION_MAX; ++i) { + SRetention* pRetention = retentions + level; + if (pRetention->keep <= 0) { + if (level > 0) { + --level; + } + break; + } + if ((now - pRetention->keep) <= (winSKey + offset)) { + break; + } + ++level; + } + + const char* str = (idStr != NULL) ? idStr : ""; + + if (level == TSDB_RETENTION_L0) { + *pLevel = TSDB_RETENTION_L0; + tsdbDebug("vgId:%d, rsma level %d is selected to query %s", TD_VID(pVnode), TSDB_RETENTION_L0, str); + return VND_RSMA0(pVnode); + } else if (level == TSDB_RETENTION_L1) { + *pLevel = TSDB_RETENTION_L1; + tsdbDebug("vgId:%d, rsma level %d is selected to query %s", TD_VID(pVnode), TSDB_RETENTION_L1, str); + return VND_RSMA1(pVnode); + } else { + *pLevel = TSDB_RETENTION_L2; + tsdbDebug("vgId:%d, rsma level %d is selected to query %s", TD_VID(pVnode), TSDB_RETENTION_L2, str); + return VND_RSMA2(pVnode); + } + } + + return VND_TSDB(pVnode); +} + +SVersionRange getQueryVerRange(SVnode* pVnode, SQueryTableDataCond* pCond, int8_t level) { + int64_t startVer = (pCond->startVersion == -1) ? 0 : pCond->startVersion; + + int64_t endVer = 0; + if (pCond->endVersion == -1) { + // user not specified end version, set current maximum version of vnode as the endVersion + endVer = pVnode->state.applied; + } else { + endVer = (pCond->endVersion > pVnode->state.applied) ? pVnode->state.applied : pCond->endVersion; + } + + return (SVersionRange){.minVer = startVer, .maxVer = endVer}; +} + +bool hasBeenDropped(const SArray* pDelList, int32_t* index, int64_t key, int64_t ver, int32_t order, + SVersionRange* pVerRange) { + if (pDelList == NULL || (taosArrayGetSize(pDelList) == 0)) { + return false; + } + + size_t num = taosArrayGetSize(pDelList); + bool asc = ASCENDING_TRAVERSE(order); + int32_t step = asc ? 1 : -1; + + if (asc) { + if (*index >= num - 1) { + TSDBKEY* last = taosArrayGetLast(pDelList); + ASSERT(key >= last->ts); + + if (key > last->ts) { + return false; + } else if (key == last->ts) { + TSDBKEY* prev = taosArrayGet(pDelList, num - 2); + return (prev->version >= ver && prev->version <= pVerRange->maxVer && prev->version >= pVerRange->minVer); + } + } else { + TSDBKEY* pCurrent = taosArrayGet(pDelList, *index); + TSDBKEY* pNext = taosArrayGet(pDelList, (*index) + 1); + + if (key < pCurrent->ts) { + return false; + } + + if (pCurrent->ts <= key && pNext->ts >= key && pCurrent->version >= ver && + pVerRange->maxVer >= pCurrent->version) { + return true; + } + + while (pNext->ts <= key && (*index) < num - 1) { + (*index) += 1; + + if ((*index) < num - 1) { + pCurrent = taosArrayGet(pDelList, *index); + pNext = taosArrayGet(pDelList, (*index) + 1); + + // it is not a consecutive deletion range, ignore it + if (pCurrent->version == 0 && pNext->version > 0) { + continue; + } + + if (pCurrent->ts <= key && pNext->ts >= key && pCurrent->version >= ver && + pVerRange->maxVer >= pCurrent->version) { + return true; + } + } + } + + return false; + } + } else { + if (*index <= 0) { + TSDBKEY* pFirst = taosArrayGet(pDelList, 0); + + if (key < pFirst->ts) { + return false; + } else if (key == pFirst->ts) { + return pFirst->version >= ver; + } else { + ASSERT(0); + } + } else { + TSDBKEY* pCurrent = taosArrayGet(pDelList, *index); + TSDBKEY* pPrev = taosArrayGet(pDelList, (*index) - 1); + + if (key > pCurrent->ts) { + return false; + } + + if (pPrev->ts <= key && pCurrent->ts >= key && pPrev->version >= ver) { + return true; + } + + while (pPrev->ts >= key && (*index) > 1) { + (*index) += step; + + if ((*index) >= 1) { + pCurrent = taosArrayGet(pDelList, *index); + pPrev = taosArrayGet(pDelList, (*index) - 1); + + // it is not a consecutive deletion range, ignore it + if (pCurrent->version > 0 && pPrev->version == 0) { + continue; + } + + if (pPrev->ts <= key && pCurrent->ts >= key && pPrev->version >= ver) { + return true; + } + } + } + + return false; + } + } + + return false; +} + +TSDBROW* getValidMemRow(SIterInfo* pIter, const SArray* pDelList, STsdbReader* pReader) { + if (!pIter->hasVal) { + return NULL; + } + + TSDBROW* pRow = tsdbTbDataIterGet(pIter->iter); + TSDBKEY key = TSDBROW_KEY(pRow); + + if (outOfTimeWindow(key.ts, &pReader->info.window)) { + pIter->hasVal = false; + return NULL; + } + + // it is a valid data version + if ((key.version <= pReader->info.verRange.maxVer && key.version >= pReader->info.verRange.minVer) && + (!hasBeenDropped(pDelList, &pIter->index, key.ts, key.version, pReader->info.order, &pReader->info.verRange))) { + return pRow; + } + + while (1) { + pIter->hasVal = tsdbTbDataIterNext(pIter->iter); + if (!pIter->hasVal) { + return NULL; + } + + pRow = tsdbTbDataIterGet(pIter->iter); + + key = TSDBROW_KEY(pRow); + if (outOfTimeWindow(key.ts, &pReader->info.window)) { + pIter->hasVal = false; + return NULL; + } + + if (key.version <= pReader->info.verRange.maxVer && key.version >= pReader->info.verRange.minVer && + (!hasBeenDropped(pDelList, &pIter->index, key.ts, key.version, pReader->info.order, &pReader->info.verRange))) { + return pRow; + } + } +} + +int32_t doMergeRowsInBuf(SIterInfo* pIter, uint64_t uid, int64_t ts, SArray* pDelList, STsdbReader* pReader) { + SRowMerger* pMerger = &pReader->status.merger; + + while (1) { + pIter->hasVal = tsdbTbDataIterNext(pIter->iter); + if (!pIter->hasVal) { + break; + } + + // data exists but not valid + TSDBROW* pRow = getValidMemRow(pIter, pDelList, pReader); + if (pRow == NULL) { + break; + } + + // ts is not identical, quit + TSDBKEY k = TSDBROW_KEY(pRow); + if (k.ts != ts) { + break; + } + + if (pRow->type == TSDBROW_ROW_FMT) { + STSchema* pTSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(pRow), pReader, uid); + if (pTSchema == NULL) { + return terrno; + } + + tsdbRowMergerAdd(pMerger, pRow, pTSchema); + } else { // column format + tsdbRowMergerAdd(pMerger, pRow, NULL); + } + } + + return TSDB_CODE_SUCCESS; +} + +static int32_t doMergeRowsInFileBlockImpl(SBlockData* pBlockData, int32_t rowIndex, int64_t key, SRowMerger* pMerger, + SVersionRange* pVerRange, int32_t step) { + while (rowIndex < pBlockData->nRow && rowIndex >= 0 && pBlockData->aTSKEY[rowIndex] == key) { + if (pBlockData->aVersion[rowIndex] > pVerRange->maxVer || pBlockData->aVersion[rowIndex] < pVerRange->minVer) { + rowIndex += step; + continue; + } + + TSDBROW fRow = tsdbRowFromBlockData(pBlockData, rowIndex); + tsdbRowMergerAdd(pMerger, &fRow, NULL); + rowIndex += step; + } + + return rowIndex; +} + +typedef enum { + CHECK_FILEBLOCK_CONT = 0x1, + CHECK_FILEBLOCK_QUIT = 0x2, +} CHECK_FILEBLOCK_STATE; + +static int32_t checkForNeighborFileBlock(STsdbReader* pReader, STableBlockScanInfo* pScanInfo, + SFileDataBlockInfo* pFBlock, SRowMerger* pMerger, int64_t key, + CHECK_FILEBLOCK_STATE* state) { + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + SBlockData* pBlockData = &pReader->status.fileBlockData; + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + + *state = CHECK_FILEBLOCK_QUIT; + int32_t step = ASCENDING_TRAVERSE(pReader->info.order) ? 1 : -1; + + bool loadNeighbor = true; + int32_t code = loadNeighborIfOverlap(pFBlock, pScanInfo, pReader, &loadNeighbor); + + if (loadNeighbor && (code == TSDB_CODE_SUCCESS)) { + pDumpInfo->rowIndex = + doMergeRowsInFileBlockImpl(pBlockData, pDumpInfo->rowIndex, key, pMerger, &pReader->info.verRange, step); + if ((pDumpInfo->rowIndex >= pDumpInfo->totalRows && asc) || (pDumpInfo->rowIndex < 0 && !asc)) { + *state = CHECK_FILEBLOCK_CONT; + } + } + + return code; +} + +int32_t doMergeRowsInFileBlocks(SBlockData* pBlockData, STableBlockScanInfo* pScanInfo, STsdbReader* pReader) { + SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo; + + SRowMerger* pMerger = &pReader->status.merger; + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + int64_t key = pBlockData->aTSKEY[pDumpInfo->rowIndex]; + int32_t step = asc ? 1 : -1; + + pDumpInfo->rowIndex += step; + if ((pDumpInfo->rowIndex <= pBlockData->nRow - 1 && asc) || (pDumpInfo->rowIndex >= 0 && !asc)) { + pDumpInfo->rowIndex = + doMergeRowsInFileBlockImpl(pBlockData, pDumpInfo->rowIndex, key, pMerger, &pReader->info.verRange, step); + } + + // all rows are consumed, let's try next file block + if ((pDumpInfo->rowIndex >= pBlockData->nRow && asc) || (pDumpInfo->rowIndex < 0 && !asc)) { + while (1) { + CHECK_FILEBLOCK_STATE st; + + SFileDataBlockInfo* pFileBlockInfo = getCurrentBlockInfo(&pReader->status.blockIter); + if (pFileBlockInfo == NULL) { + st = CHECK_FILEBLOCK_QUIT; + break; + } + + checkForNeighborFileBlock(pReader, pScanInfo, pFileBlockInfo, pMerger, key, &st); + if (st == CHECK_FILEBLOCK_QUIT) { + break; + } + } + } + + return TSDB_CODE_SUCCESS; +} + +int32_t doMergeRowsInLastBlock(SLastBlockReader* pLastBlockReader, STableBlockScanInfo* pScanInfo, int64_t ts, + SRowMerger* pMerger, SVersionRange* pVerRange, const char* idStr) { + while (nextRowFromLastBlocks(pLastBlockReader, pScanInfo, pVerRange)) { + int64_t next1 = getCurrentKeyInLastBlock(pLastBlockReader); + if (next1 == ts) { + TSDBROW* pRow1 = tMergeTreeGetRow(&pLastBlockReader->mergeTree); + tsdbRowMergerAdd(pMerger, pRow1, NULL); + } else { + tsdbTrace("uid:%" PRIu64 " last del index:%d, del range:%d, lastKeyInStt:%" PRId64 ", %s", pScanInfo->uid, + pScanInfo->lastBlockDelIndex, (int32_t)taosArrayGetSize(pScanInfo->delSkyline), pScanInfo->lastKeyInStt, + idStr); + break; + } + } + + return TSDB_CODE_SUCCESS; +} + +int32_t doMergeMemTableMultiRows(TSDBROW* pRow, uint64_t uid, SIterInfo* pIter, SArray* pDelList, TSDBROW* pResRow, + STsdbReader* pReader, bool* freeTSRow) { + TSDBROW* pNextRow = NULL; + TSDBROW current = *pRow; + + { // if the timestamp of the next valid row has a different ts, return current row directly + pIter->hasVal = tsdbTbDataIterNext(pIter->iter); + + if (!pIter->hasVal) { + *pResRow = *pRow; + *freeTSRow = false; + return TSDB_CODE_SUCCESS; + } else { // has next point in mem/imem + pNextRow = getValidMemRow(pIter, pDelList, pReader); + if (pNextRow == NULL) { + *pResRow = current; + *freeTSRow = false; + return TSDB_CODE_SUCCESS; + } + + if (TSDBROW_TS(¤t) != TSDBROW_TS(pNextRow)) { + *pResRow = current; + *freeTSRow = false; + return TSDB_CODE_SUCCESS; + } + } + } + + terrno = 0; + int32_t code = 0; + + // start to merge duplicated rows + if (current.type == TSDBROW_ROW_FMT) { + // get the correct schema for data in memory + STSchema* pTSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(¤t), pReader, uid); + if (pTSchema == NULL) { + return terrno; + } + + code = tsdbRowMergerAdd(&pReader->status.merger, ¤t, pTSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + STSchema* pTSchema1 = doGetSchemaForTSRow(TSDBROW_SVERSION(pNextRow), pReader, uid); + if (pTSchema1 == NULL) { + return terrno; + } + + tsdbRowMergerAdd(&pReader->status.merger, pNextRow, pTSchema1); + } else { // let's merge rows in file block + code = tsdbRowMergerAdd(&pReader->status.merger, ¤t, pReader->info.pSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + tsdbRowMergerAdd(&pReader->status.merger, pNextRow, NULL); + } + + code = doMergeRowsInBuf(pIter, uid, TSDBROW_TS(¤t), pDelList, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = tsdbRowMergerGetRow(&pReader->status.merger, &pResRow->pTSRow); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + pResRow->type = TSDBROW_ROW_FMT; + tsdbRowMergerClear(&pReader->status.merger); + *freeTSRow = true; + + return TSDB_CODE_SUCCESS; +} + +int32_t doMergeMemIMemRows(TSDBROW* pRow, TSDBROW* piRow, STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader, + SRow** pTSRow) { + SRowMerger* pMerger = &pReader->status.merger; + + TSDBKEY k = TSDBROW_KEY(pRow); + TSDBKEY ik = TSDBROW_KEY(piRow); + STSchema* pSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(pRow), pReader, pBlockScanInfo->uid); + if (pSchema == NULL) { + return terrno; + } + + STSchema* piSchema = doGetSchemaForTSRow(TSDBROW_SVERSION(piRow), pReader, pBlockScanInfo->uid); + if (piSchema == NULL) { + return terrno; + } + + if (ASCENDING_TRAVERSE(pReader->info.order)) { // ascending order imem --> mem + int32_t code = tsdbRowMergerAdd(&pReader->status.merger, piRow, piSchema); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doMergeRowsInBuf(&pBlockScanInfo->iiter, pBlockScanInfo->uid, ik.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + tsdbRowMergerAdd(&pReader->status.merger, pRow, pSchema); + code = doMergeRowsInBuf(&pBlockScanInfo->iter, pBlockScanInfo->uid, k.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + } else { + int32_t code = tsdbRowMergerAdd(&pReader->status.merger, pRow, pSchema); + if (code != TSDB_CODE_SUCCESS || pMerger->pTSchema == NULL) { + return code; + } + + code = doMergeRowsInBuf(&pBlockScanInfo->iter, pBlockScanInfo->uid, k.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + tsdbRowMergerAdd(&pReader->status.merger, piRow, piSchema); + code = doMergeRowsInBuf(&pBlockScanInfo->iiter, pBlockScanInfo->uid, ik.ts, pBlockScanInfo->delSkyline, pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + int32_t code = tsdbRowMergerGetRow(pMerger, pTSRow); + tsdbRowMergerClear(pMerger); + return code; +} + +static int32_t tsdbGetNextRowInMem(STableBlockScanInfo* pBlockScanInfo, STsdbReader* pReader, TSDBROW* pResRow, + int64_t endKey, bool* freeTSRow) { + TSDBROW* pRow = getValidMemRow(&pBlockScanInfo->iter, pBlockScanInfo->delSkyline, pReader); + TSDBROW* piRow = getValidMemRow(&pBlockScanInfo->iiter, pBlockScanInfo->delSkyline, pReader); + SArray* pDelList = pBlockScanInfo->delSkyline; + uint64_t uid = pBlockScanInfo->uid; + + // todo refactor + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + if (pBlockScanInfo->iter.hasVal) { + TSDBKEY k = TSDBROW_KEY(pRow); + if ((k.ts >= endKey && asc) || (k.ts <= endKey && !asc)) { + pRow = NULL; + } + } + + if (pBlockScanInfo->iiter.hasVal) { + TSDBKEY k = TSDBROW_KEY(piRow); + if ((k.ts >= endKey && asc) || (k.ts <= endKey && !asc)) { + piRow = NULL; + } + } + + if (pBlockScanInfo->iter.hasVal && pBlockScanInfo->iiter.hasVal && pRow != NULL && piRow != NULL) { + TSDBKEY k = TSDBROW_KEY(pRow); + TSDBKEY ik = TSDBROW_KEY(piRow); + + int32_t code = TSDB_CODE_SUCCESS; + if (ik.ts != k.ts) { + if (((ik.ts < k.ts) && asc) || ((ik.ts > k.ts) && (!asc))) { // ik.ts < k.ts + code = doMergeMemTableMultiRows(piRow, uid, &pBlockScanInfo->iiter, pDelList, pResRow, pReader, freeTSRow); + } else if (((k.ts < ik.ts) && asc) || ((k.ts > ik.ts) && (!asc))) { + code = doMergeMemTableMultiRows(pRow, uid, &pBlockScanInfo->iter, pDelList, pResRow, pReader, freeTSRow); + } + } else { // ik.ts == k.ts + *freeTSRow = true; + pResRow->type = TSDBROW_ROW_FMT; + code = doMergeMemIMemRows(pRow, piRow, pBlockScanInfo, pReader, &pResRow->pTSRow); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + + return code; + } + + if (pBlockScanInfo->iter.hasVal && pRow != NULL) { + return doMergeMemTableMultiRows(pRow, pBlockScanInfo->uid, &pBlockScanInfo->iter, pDelList, pResRow, pReader, + freeTSRow); + } + + if (pBlockScanInfo->iiter.hasVal && piRow != NULL) { + return doMergeMemTableMultiRows(piRow, uid, &pBlockScanInfo->iiter, pDelList, pResRow, pReader, freeTSRow); + } + + return TSDB_CODE_SUCCESS; +} + +int32_t doAppendRowFromTSRow(SSDataBlock* pBlock, STsdbReader* pReader, SRow* pTSRow, STableBlockScanInfo* pScanInfo) { + int32_t outputRowIndex = pBlock->info.rows; + int64_t uid = pScanInfo->uid; + int32_t code = TSDB_CODE_SUCCESS; + + SBlockLoadSuppInfo* pSupInfo = &pReader->suppInfo; + STSchema* pSchema = doGetSchemaForTSRow(pTSRow->sver, pReader, uid); + if (pSchema == NULL) { + return terrno; + } + + SColVal colVal = {0}; + int32_t i = 0, j = 0; + + if (pSupInfo->colId[i] == PRIMARYKEY_TIMESTAMP_COL_ID) { + SColumnInfoData* pColData = taosArrayGet(pBlock->pDataBlock, pSupInfo->slotId[i]); + ((int64_t*)pColData->pData)[outputRowIndex] = pTSRow->ts; + i += 1; + } + + while (i < pSupInfo->numOfCols && j < pSchema->numOfCols) { + col_id_t colId = pSupInfo->colId[i]; + + if (colId == pSchema->columns[j].colId) { + SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, pSupInfo->slotId[i]); + + tRowGet(pTSRow, pSchema, j, &colVal); + code = doCopyColVal(pColInfoData, outputRowIndex, i, &colVal, pSupInfo); + if (code) { + return code; + } + i += 1; + j += 1; + } else if (colId < pSchema->columns[j].colId) { + SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, pSupInfo->slotId[i]); + + colDataSetNULL(pColInfoData, outputRowIndex); + i += 1; + } else if (colId > pSchema->columns[j].colId) { + j += 1; + } + } + + // set null value since current column does not exist in the "pSchema" + while (i < pSupInfo->numOfCols) { + SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, pSupInfo->slotId[i]); + colDataSetNULL(pColInfoData, outputRowIndex); + i += 1; + } + + pBlock->info.dataLoad = 1; + pBlock->info.rows += 1; + pScanInfo->lastKey = pTSRow->ts; + return TSDB_CODE_SUCCESS; +} + +int32_t doAppendRowFromFileBlock(SSDataBlock* pResBlock, STsdbReader* pReader, SBlockData* pBlockData, + int32_t rowIndex) { + int32_t i = 0, j = 0; + int32_t outputRowIndex = pResBlock->info.rows; + int32_t code = TSDB_CODE_SUCCESS; + + SBlockLoadSuppInfo* pSupInfo = &pReader->suppInfo; + ((int64_t*)pReader->status.pPrimaryTsCol->pData)[outputRowIndex] = pBlockData->aTSKEY[rowIndex]; + i += 1; + + SColVal cv = {0}; + int32_t numOfInputCols = pBlockData->nColData; + int32_t numOfOutputCols = pSupInfo->numOfCols; + + while (i < numOfOutputCols && j < numOfInputCols) { + SColData* pData = tBlockDataGetColDataByIdx(pBlockData, j); + if (pData->cid < pSupInfo->colId[i]) { + j += 1; + continue; + } + + SColumnInfoData* pCol = TARRAY_GET_ELEM(pResBlock->pDataBlock, pSupInfo->slotId[i]); + if (pData->cid == pSupInfo->colId[i]) { + tColDataGetValue(pData, rowIndex, &cv); + code = doCopyColVal(pCol, outputRowIndex, i, &cv, pSupInfo); + if (code) { + return code; + } + j += 1; + } else if (pData->cid > pCol->info.colId) { + // the specified column does not exist in file block, fill with null data + colDataSetNULL(pCol, outputRowIndex); + } + + i += 1; + } + + while (i < numOfOutputCols) { + SColumnInfoData* pCol = taosArrayGet(pResBlock->pDataBlock, pSupInfo->slotId[i]); + colDataSetNULL(pCol, outputRowIndex); + i += 1; + } + + pResBlock->info.dataLoad = 1; + pResBlock->info.rows += 1; + return TSDB_CODE_SUCCESS; +} + +int32_t buildDataBlockFromBufImpl(STableBlockScanInfo* pBlockScanInfo, int64_t endKey, int32_t capacity, + STsdbReader* pReader) { + SSDataBlock* pBlock = pReader->resBlockInfo.pResBlock; + int32_t code = TSDB_CODE_SUCCESS; + + do { + // SRow* pTSRow = NULL; + TSDBROW row = {.type = -1}; + bool freeTSRow = false; + tsdbGetNextRowInMem(pBlockScanInfo, pReader, &row, endKey, &freeTSRow); + if (row.type == -1) { + break; + } + + if (row.type == TSDBROW_ROW_FMT) { + code = doAppendRowFromTSRow(pBlock, pReader, row.pTSRow, pBlockScanInfo); + + if (freeTSRow) { + taosMemoryFree(row.pTSRow); + } + + if (code) { + return code; + } + } else { + code = doAppendRowFromFileBlock(pBlock, pReader, row.pBlockData, row.iRow); + if (code) { + break; + } + } + + // no data in buffer, return immediately + if (!(pBlockScanInfo->iter.hasVal || pBlockScanInfo->iiter.hasVal)) { + break; + } + + if (pBlock->info.rows >= capacity) { + break; + } + } while (1); + + return code; +} + +// TODO refactor: with createDataBlockScanInfo +int32_t tsdbSetTableList2(STsdbReader* pReader, const void* pTableList, int32_t num) { + int32_t size = tSimpleHashGetSize(pReader->status.pTableMap); + + STableBlockScanInfo** p = NULL; + int32_t iter = 0; + + while ((p = tSimpleHashIterate(pReader->status.pTableMap, p, &iter)) != NULL) { + clearBlockScanInfo(*p); + } + + if (size < num) { + int32_t code = ensureBlockScanInfoBuf(&pReader->blockInfoBuf, num); + if (code) { + return code; + } + + char* p1 = taosMemoryRealloc(pReader->status.uidList.tableUidList, sizeof(uint64_t) * num); + if (p1 == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + pReader->status.uidList.tableUidList = (uint64_t*)p1; + } + + tSimpleHashClear(pReader->status.pTableMap); + STableUidList* pUidList = &pReader->status.uidList; + pUidList->currentIndex = 0; + + STableKeyInfo* pList = (STableKeyInfo*)pTableList; + for (int32_t i = 0; i < num; ++i) { + STableBlockScanInfo* pInfo = getPosInBlockInfoBuf(&pReader->blockInfoBuf, i); + pInfo->uid = pList[i].uid; + pUidList->tableUidList[i] = pList[i].uid; + + // todo extract method + if (ASCENDING_TRAVERSE(pReader->info.order)) { + int64_t skey = pReader->info.window.skey; + pInfo->lastKey = (skey > INT64_MIN) ? (skey - 1) : skey; + pInfo->lastKeyInStt = skey; + } else { + int64_t ekey = pReader->info.window.ekey; + pInfo->lastKey = (ekey < INT64_MAX) ? (ekey + 1) : ekey; + pInfo->lastKeyInStt = ekey; + } + + tSimpleHashPut(pReader->status.pTableMap, &pInfo->uid, sizeof(uint64_t), &pInfo, POINTER_BYTES); + } + + return TDB_CODE_SUCCESS; +} + +void* tsdbGetIdx2(SMeta* pMeta) { + if (pMeta == NULL) { + return NULL; + } + return metaGetIdx(pMeta); +} + +void* tsdbGetIvtIdx2(SMeta* pMeta) { + if (pMeta == NULL) { + return NULL; + } + return metaGetIvtIdx(pMeta); +} + +uint64_t tsdbGetReaderMaxVersion2(STsdbReader* pReader) { return pReader->info.verRange.maxVer; } + +static int32_t doOpenReaderImpl(STsdbReader* pReader) { + SReaderStatus* pStatus = &pReader->status; + SDataBlockIter* pBlockIter = &pStatus->blockIter; + + initFilesetIterator(&pStatus->fileIter, pReader->pReadSnap->pfSetArray, pReader); + resetDataBlockIterator(&pStatus->blockIter, pReader->info.order); + + int32_t code = TSDB_CODE_SUCCESS; + if (pStatus->fileIter.numOfFiles == 0) { + pStatus->loadFromFile = false; + } else if (READ_MODE_COUNT_ONLY == pReader->info.readMode) { + // DO NOTHING + } else { + code = initForFirstBlockInFile(pReader, pBlockIter); + } + + if (!pStatus->loadFromFile) { + resetTableListIndex(pStatus); + } + + return code; +} + +static void freeSchemaFunc(void* param) { + void** p = (void**)param; + taosMemoryFreeClear(*p); +} + +static void clearSharedPtr(STsdbReader* p) { + p->status.pTableMap = NULL; + p->status.uidList.tableUidList = NULL; + p->info.pSchema = NULL; + p->pReadSnap = NULL; + p->pSchemaMap = NULL; +} + +static void setSharedPtr(STsdbReader* pDst, const STsdbReader* pSrc) { + pDst->status.pTableMap = pSrc->status.pTableMap; + pDst->status.uidList = pSrc->status.uidList; + pDst->info.pSchema = pSrc->info.pSchema; + pDst->pSchemaMap = pSrc->pSchemaMap; + pDst->pReadSnap = pSrc->pReadSnap; + pDst->pReadSnap->pfSetArray = pSrc->pReadSnap->pfSetArray; + + if (pDst->info.pSchema) { + tsdbRowMergerInit(&pDst->status.merger, pDst->info.pSchema); + } +} + +// ====================================== EXPOSED APIs ====================================== +int32_t tsdbReaderOpen2(void* pVnode, SQueryTableDataCond* pCond, void* pTableList, int32_t numOfTables, + SSDataBlock* pResBlock, void** ppReader, const char* idstr, bool countOnly, + SHashObj** pIgnoreTables) { + STimeWindow window = pCond->twindows; + SVnodeCfg* pConf = &(((SVnode*)pVnode)->config); + + int32_t capacity = pConf->tsdbCfg.maxRows; + if (pResBlock != NULL) { + blockDataEnsureCapacity(pResBlock, capacity); + } + + int32_t code = tsdbReaderCreate(pVnode, pCond, ppReader, capacity, pResBlock, idstr); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + + // check for query time window + STsdbReader* pReader = *ppReader; + if (isEmptyQueryTimeWindow(&pReader->info.window) && pCond->type == TIMEWINDOW_RANGE_CONTAINED) { + tsdbDebug("%p query window not overlaps with the data set, no result returned, %s", pReader, pReader->idStr); + return TSDB_CODE_SUCCESS; + } + + if (pCond->type == TIMEWINDOW_RANGE_EXTERNAL) { + // update the SQueryTableDataCond to create inner reader + int32_t order = pCond->order; + if (order == TSDB_ORDER_ASC) { + pCond->twindows.ekey = window.skey - 1; + pCond->twindows.skey = INT64_MIN; + pCond->order = TSDB_ORDER_DESC; + } else { + pCond->twindows.skey = window.ekey + 1; + pCond->twindows.ekey = INT64_MAX; + pCond->order = TSDB_ORDER_ASC; + } + + // here we only need one more row, so the capacity is set to be ONE. + code = tsdbReaderCreate(pVnode, pCond, (void**)&((STsdbReader*)pReader)->innerReader[0], 1, pResBlock, idstr); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + + if (order == TSDB_ORDER_ASC) { + pCond->twindows.skey = window.ekey + 1; + pCond->twindows.ekey = INT64_MAX; + } else { + pCond->twindows.skey = INT64_MIN; + pCond->twindows.ekey = window.ekey - 1; + } + pCond->order = order; + + code = tsdbReaderCreate(pVnode, pCond, (void**)&((STsdbReader*)pReader)->innerReader[1], 1, pResBlock, idstr); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + } + + // NOTE: the endVersion in pCond is the data version not schema version, so pCond->endVersion is not correct here. + // no valid error code set in metaGetTbTSchema, so let's set the error code here. + // we should proceed in case of tmq processing. + if (pCond->suid != 0) { + pReader->info.pSchema = metaGetTbTSchema(pReader->pTsdb->pVnode->pMeta, pReader->info.suid, -1, 1); + if (pReader->info.pSchema == NULL) { + tsdbError("failed to get table schema, suid:%" PRIu64 ", ver:-1, %s", pReader->info.suid, pReader->idStr); + } + } else if (numOfTables > 0) { + STableKeyInfo* pKey = pTableList; + pReader->info.pSchema = metaGetTbTSchema(pReader->pTsdb->pVnode->pMeta, pKey->uid, -1, 1); + if (pReader->info.pSchema == NULL) { + tsdbError("failed to get table schema, uid:%" PRIu64 ", ver:-1, %s", pKey->uid, pReader->idStr); + } + } + + if (pReader->info.pSchema != NULL) { + tsdbRowMergerInit(&pReader->status.merger, pReader->info.pSchema); + } + + pReader->pSchemaMap = tSimpleHashInit(8, taosFastHash); + if (pReader->pSchemaMap == NULL) { + tsdbError("failed init schema hash for reader %s", pReader->idStr); + code = TSDB_CODE_OUT_OF_MEMORY; + goto _err; + } + + tSimpleHashSetFreeFp(pReader->pSchemaMap, freeSchemaFunc); + if (pReader->info.pSchema != NULL) { + code = updateBlockSMAInfo(pReader->info.pSchema, &pReader->suppInfo); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + } + + STsdbReader* p = (pReader->innerReader[0] != NULL) ? pReader->innerReader[0] : pReader; + pReader->status.pTableMap = + createDataBlockScanInfo(p, &pReader->blockInfoBuf, pTableList, &pReader->status.uidList, numOfTables); + if (pReader->status.pTableMap == NULL) { + *ppReader = NULL; + code = TSDB_CODE_OUT_OF_MEMORY; + goto _err; + } + + pReader->status.pLDataIterArray = taosArrayInit(4, POINTER_BYTES); + if (pReader->status.pLDataIterArray == NULL) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + goto _err; + } + + pReader->flag = READER_STATUS_SUSPEND; + + if (countOnly) { + pReader->info.readMode = READ_MODE_COUNT_ONLY; + } + + pReader->pIgnoreTables = pIgnoreTables; + + tsdbDebug("%p total numOfTable:%d, window:%" PRId64 " - %" PRId64 ", verRange:%" PRId64 " - %" PRId64 + " in this query %s", + pReader, numOfTables, pReader->info.window.skey, pReader->info.window.ekey, pReader->info.verRange.minVer, + pReader->info.verRange.maxVer, pReader->idStr); + + return code; + +_err: + tsdbError("failed to create data reader, code:%s %s", tstrerror(code), idstr); + tsdbReaderClose2(*ppReader); + *ppReader = NULL; // reset the pointer value. + return code; +} + +void tsdbReaderClose2(STsdbReader* pReader) { + if (pReader == NULL) { + return; + } + + tsdbAcquireReader(pReader); + + { + if (pReader->innerReader[0] != NULL || pReader->innerReader[1] != NULL) { + STsdbReader* p = pReader->innerReader[0]; + clearSharedPtr(p); + + p = pReader->innerReader[1]; + clearSharedPtr(p); + + tsdbReaderClose2(pReader->innerReader[0]); + tsdbReaderClose2(pReader->innerReader[1]); + } + } + + SBlockLoadSuppInfo* pSupInfo = &pReader->suppInfo; + TARRAY2_DESTROY(&pSupInfo->colAggArray, NULL); + for (int32_t i = 0; i < pSupInfo->numOfCols; ++i) { + if (pSupInfo->buildBuf[i] != NULL) { + taosMemoryFreeClear(pSupInfo->buildBuf[i]); + } + } + + if (pReader->resBlockInfo.freeBlock) { + pReader->resBlockInfo.pResBlock = blockDataDestroy(pReader->resBlockInfo.pResBlock); + } + + taosMemoryFree(pSupInfo->colId); + tBlockDataDestroy(&pReader->status.fileBlockData); + cleanupDataBlockIterator(&pReader->status.blockIter); + + size_t numOfTables = tSimpleHashGetSize(pReader->status.pTableMap); + if (pReader->status.pTableMap != NULL) { + destroyAllBlockScanInfo(pReader->status.pTableMap); + clearBlockScanInfoBuf(&pReader->blockInfoBuf); + pReader->status.pTableMap = NULL; + } + + if (pReader->pFileReader != NULL) { + tsdbDataFileReaderClose(&pReader->pFileReader); + } + + qTrace("tsdb/reader-close: %p, untake snapshot", pReader); + tsdbUntakeReadSnap2(pReader, pReader->pReadSnap, true); + pReader->pReadSnap = NULL; + + tsdbReleaseReader(pReader); + tsdbUninitReaderLock(pReader); + + SCostSummary* pCost = &pReader->cost; + SFilesetIter* pFilesetIter = &pReader->status.fileIter; + if (pFilesetIter->pLastBlockReader != NULL) { + SLastBlockReader* pLReader = pFilesetIter->pLastBlockReader; + tMergeTreeClose(&pLReader->mergeTree); + taosMemoryFree(pLReader); + } + + destroySttBlockReader(pReader->status.pLDataIterArray, &pCost->lastBlockLoad, &pCost->lastBlockLoadTime); + taosMemoryFreeClear(pReader->status.uidList.tableUidList); + + tsdbDebug( + "%p :io-cost summary: head-file:%" PRIu64 ", head-file time:%.2f ms, SMA:%" PRId64 + " SMA-time:%.2f ms, fileBlocks:%" PRId64 + ", fileBlocks-load-time:%.2f ms, " + "build in-memory-block-time:%.2f ms, lastBlocks:%" PRId64 ", lastBlocks-time:%.2f ms, composed-blocks:%" PRId64 + ", composed-blocks-time:%.2fms, STableBlockScanInfo size:%.2f Kb, createTime:%.2f ms,createSkylineIterTime:%.2f " + "ms, initLastBlockReader:%.2fms, %s", + pReader, pCost->headFileLoad, pCost->headFileLoadTime, pCost->smaDataLoad, pCost->smaLoadTime, pCost->numOfBlocks, + pCost->blockLoadTime, pCost->buildmemBlock, pCost->lastBlockLoad, pCost->lastBlockLoadTime, pCost->composedBlocks, + pCost->buildComposedBlockTime, numOfTables * sizeof(STableBlockScanInfo) / 1000.0, pCost->createScanInfoList, + pCost->createSkylineIterTime, pCost->initLastBlockReader, pReader->idStr); + + taosMemoryFree(pReader->idStr); + + tsdbRowMergerCleanup(&pReader->status.merger); + taosMemoryFree(pReader->info.pSchema); + + tSimpleHashCleanup(pReader->pSchemaMap); + taosMemoryFreeClear(pReader); +} + +int32_t tsdbReaderSuspend2(STsdbReader* pReader) { + int32_t code = 0; + + // save reader's base state & reset top state to be reconstructed from base state + SReaderStatus* pStatus = &pReader->status; + STableBlockScanInfo* pBlockScanInfo = NULL; + + if (pStatus->loadFromFile) { + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(&pReader->status.blockIter); + if (pBlockInfo != NULL) { + pBlockScanInfo = getTableBlockScanInfo(pStatus->pTableMap, pBlockInfo->uid, pReader->idStr); + if (pBlockScanInfo == NULL) { + goto _err; + } + } else { + pBlockScanInfo = *pStatus->pTableIter; + } + + tsdbDataFileReaderClose(&pReader->pFileReader); + + // resetDataBlockScanInfo excluding lastKey + STableBlockScanInfo** p = NULL; + int32_t iter = 0; + + while ((p = tSimpleHashIterate(pStatus->pTableMap, p, &iter)) != NULL) { + STableBlockScanInfo* pInfo = *(STableBlockScanInfo**)p; + + pInfo->iterInit = false; + pInfo->iter.hasVal = false; + pInfo->iiter.hasVal = false; + + if (pInfo->iter.iter != NULL) { + pInfo->iter.iter = tsdbTbDataIterDestroy(pInfo->iter.iter); + } + + if (pInfo->iiter.iter != NULL) { + pInfo->iiter.iter = tsdbTbDataIterDestroy(pInfo->iiter.iter); + } + + pInfo->delSkyline = taosArrayDestroy(pInfo->delSkyline); + pInfo->pfileDelData = taosArrayDestroy(pInfo->pfileDelData); + } + } else { + // resetDataBlockScanInfo excluding lastKey + STableBlockScanInfo** p = NULL; + int32_t iter = 0; + + while ((p = tSimpleHashIterate(pStatus->pTableMap, p, &iter)) != NULL) { + STableBlockScanInfo* pInfo = *(STableBlockScanInfo**)p; + + pInfo->iterInit = false; + pInfo->iter.hasVal = false; + pInfo->iiter.hasVal = false; + + if (pInfo->iter.iter != NULL) { + pInfo->iter.iter = tsdbTbDataIterDestroy(pInfo->iter.iter); + } + + if (pInfo->iiter.iter != NULL) { + pInfo->iiter.iter = tsdbTbDataIterDestroy(pInfo->iiter.iter); + } + + pInfo->delSkyline = taosArrayDestroy(pInfo->delSkyline); + } + + pBlockScanInfo = pStatus->pTableIter == NULL ? NULL : *pStatus->pTableIter; + if (pBlockScanInfo) { + // save lastKey to restore memory iterator + STimeWindow w = pReader->resBlockInfo.pResBlock->info.window; + pBlockScanInfo->lastKey = ASCENDING_TRAVERSE(pReader->info.order) ? w.ekey : w.skey; + + // reset current current table's data block scan info, + pBlockScanInfo->iterInit = false; + + pBlockScanInfo->iter.hasVal = false; + pBlockScanInfo->iiter.hasVal = false; + if (pBlockScanInfo->iter.iter != NULL) { + pBlockScanInfo->iter.iter = tsdbTbDataIterDestroy(pBlockScanInfo->iter.iter); + } + + if (pBlockScanInfo->iiter.iter != NULL) { + pBlockScanInfo->iiter.iter = tsdbTbDataIterDestroy(pBlockScanInfo->iiter.iter); + } + + pBlockScanInfo->pBlockList = taosArrayDestroy(pBlockScanInfo->pBlockList); + // TODO: keep skyline for reuse + pBlockScanInfo->delSkyline = taosArrayDestroy(pBlockScanInfo->delSkyline); + } + } + + tsdbUntakeReadSnap(pReader, pReader->pReadSnap, false); + pReader->pReadSnap = NULL; + pReader->flag = READER_STATUS_SUSPEND; + + tsdbDebug("reader: %p suspended uid %" PRIu64 " in this query %s", pReader, pBlockScanInfo ? pBlockScanInfo->uid : 0, + pReader->idStr); + return code; + +_err: + tsdbError("failed to suspend data reader, code:%s %s", tstrerror(code), pReader->idStr); + return code; +} + +static int32_t tsdbSetQueryReseek(void* pQHandle) { + int32_t code = 0; + STsdbReader* pReader = pQHandle; + + code = tsdbTryAcquireReader(pReader); + if (code == 0) { + if (pReader->flag == READER_STATUS_SUSPEND) { + tsdbReleaseReader(pReader); + return code; + } + + tsdbReaderSuspend2(pReader); + + tsdbReleaseReader(pReader); + + return code; + } else if (code == EBUSY) { + return TSDB_CODE_VND_QUERY_BUSY; + } else { + terrno = TAOS_SYSTEM_ERROR(code); + return TSDB_CODE_FAILED; + } +} + +int32_t tsdbReaderResume2(STsdbReader* pReader) { + int32_t code = 0; + + STableBlockScanInfo** pBlockScanInfo = pReader->status.pTableIter; + + // restore reader's state + // task snapshot + int32_t numOfTables = tSimpleHashGetSize(pReader->status.pTableMap); + if (numOfTables > 0) { + qTrace("tsdb/reader: %p, take snapshot", pReader); + code = tsdbTakeReadSnap2(pReader, tsdbSetQueryReseek, &pReader->pReadSnap); + if (code != TSDB_CODE_SUCCESS) { + goto _err; + } + + if (pReader->type == TIMEWINDOW_RANGE_CONTAINED) { + code = doOpenReaderImpl(pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } else { + STsdbReader* pPrevReader = pReader->innerReader[0]; + STsdbReader* pNextReader = pReader->innerReader[1]; + + // we need only one row + pPrevReader->resBlockInfo.capacity = 1; + setSharedPtr(pPrevReader, pReader); + + pNextReader->resBlockInfo.capacity = 1; + setSharedPtr(pNextReader, pReader); + + code = doOpenReaderImpl(pPrevReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + } + + pReader->flag = READER_STATUS_NORMAL; + tsdbDebug("reader: %p resumed uid %" PRIu64 ", numOfTable:%" PRId32 ", in this query %s", pReader, + pBlockScanInfo ? (*pBlockScanInfo)->uid : 0, numOfTables, pReader->idStr); + return code; + +_err: + tsdbError("failed to resume data reader, code:%s %s", tstrerror(code), pReader->idStr); + return code; +} + +static bool tsdbReadRowsCountOnly(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + SSDataBlock* pBlock = pReader->resBlockInfo.pResBlock; + + if (pReader->status.loadFromFile == false) { + return false; + } + + code = readRowsCountFromFiles(pReader); + if (code != TSDB_CODE_SUCCESS) { + return false; + } + + code = readRowsCountFromMem(pReader); + if (code != TSDB_CODE_SUCCESS) { + return false; + } + + pBlock->info.rows = pReader->rowsNum; + pBlock->info.id.uid = 0; + pBlock->info.dataLoad = 0; + + pReader->rowsNum = 0; + + return pBlock->info.rows > 0; +} + +static int32_t doTsdbNextDataBlock2(STsdbReader* pReader, bool* hasNext) { + int32_t code = TSDB_CODE_SUCCESS; + + // cleanup the data that belongs to the previous data block + SSDataBlock* pBlock = pReader->resBlockInfo.pResBlock; + blockDataCleanup(pBlock); + + *hasNext = false; + + SReaderStatus* pStatus = &pReader->status; + if (tSimpleHashGetSize(pStatus->pTableMap) == 0) { + return code; + } + + if (READ_MODE_COUNT_ONLY == pReader->info.readMode) { + return tsdbReadRowsCountOnly(pReader); + } + + if (pStatus->loadFromFile) { + code = buildBlockFromFiles(pReader); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + if (pBlock->info.rows <= 0) { + resetTableListIndex(&pReader->status); + code = buildBlockFromBufferSequentially(pReader); + } + } else { // no data in files, let's try the buffer + code = buildBlockFromBufferSequentially(pReader); + } + + *hasNext = pBlock->info.rows > 0; + + return code; +} + +int32_t tsdbNextDataBlock2(STsdbReader* pReader, bool* hasNext) { + int32_t code = TSDB_CODE_SUCCESS; + + *hasNext = false; + + if (isEmptyQueryTimeWindow(&pReader->info.window) || pReader->step == EXTERNAL_ROWS_NEXT || + pReader->code != TSDB_CODE_SUCCESS) { + return (pReader->code != TSDB_CODE_SUCCESS) ? pReader->code : code; + } + + SReaderStatus* pStatus = &pReader->status; + + code = tsdbAcquireReader(pReader); + qTrace("tsdb/read: %p, take read mutex, code: %d", pReader, code); + + if (pReader->flag == READER_STATUS_SUSPEND) { + code = tsdbReaderResume2(pReader); + if (code != TSDB_CODE_SUCCESS) { + tsdbReleaseReader(pReader); + return code; + } + } + + if (pReader->innerReader[0] != NULL && pReader->step == 0) { + code = doTsdbNextDataBlock2(pReader->innerReader[0], hasNext); + if (code) { + tsdbReleaseReader(pReader); + return code; + } + + pReader->step = EXTERNAL_ROWS_PREV; + if (*hasNext) { + pStatus = &pReader->innerReader[0]->status; + if (pStatus->composedDataBlock) { + qTrace("tsdb/read: %p, unlock read mutex", pReader); + tsdbReleaseReader(pReader); + } + + return code; + } + } + + if (pReader->step == EXTERNAL_ROWS_PREV) { + // prepare for the main scan + code = doOpenReaderImpl(pReader); + int32_t step = 1; + resetAllDataBlockScanInfo(pReader->status.pTableMap, pReader->innerReader[0]->info.window.ekey, step); + + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + pReader->step = EXTERNAL_ROWS_MAIN; + } + + code = doTsdbNextDataBlock2(pReader, hasNext); + if (code != TSDB_CODE_SUCCESS) { + tsdbReleaseReader(pReader); + return code; + } + + if (*hasNext) { + if (pStatus->composedDataBlock) { + qTrace("tsdb/read: %p, unlock read mutex", pReader); + tsdbReleaseReader(pReader); + } + + return code; + } + + if (pReader->step == EXTERNAL_ROWS_MAIN && pReader->innerReader[1] != NULL) { + // prepare for the next row scan + int32_t step = -1; + code = doOpenReaderImpl(pReader->innerReader[1]); + resetAllDataBlockScanInfo(pReader->innerReader[1]->status.pTableMap, pReader->info.window.ekey, step); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + code = doTsdbNextDataBlock2(pReader->innerReader[1], hasNext); + if (code != TSDB_CODE_SUCCESS) { + tsdbReleaseReader(pReader); + return code; + } + + pReader->step = EXTERNAL_ROWS_NEXT; + if (*hasNext) { + pStatus = &pReader->innerReader[1]->status; + if (pStatus->composedDataBlock) { + qTrace("tsdb/read: %p, unlock read mutex", pReader); + tsdbReleaseReader(pReader); + } + + return code; + } + } + + qTrace("tsdb/read: %p, unlock read mutex", pReader); + tsdbReleaseReader(pReader); + + return code; +} + +static void doFillNullColSMA(SBlockLoadSuppInfo* pSup, int32_t numOfRows, int32_t numOfCols, SColumnDataAgg* pTsAgg) { + // do fill all null column value SMA info + int32_t i = 0, j = 0; + int32_t size = (int32_t)TARRAY2_SIZE(&pSup->colAggArray); + TARRAY2_INSERT_PTR(&pSup->colAggArray, 0, pTsAgg); + size++; + + while (j < numOfCols && i < size) { + SColumnDataAgg* pAgg = &pSup->colAggArray.data[i]; + if (pAgg->colId == pSup->colId[j]) { + i += 1; + j += 1; + } else if (pAgg->colId < pSup->colId[j]) { + i += 1; + } else if (pSup->colId[j] < pAgg->colId) { + if (pSup->colId[j] != PRIMARYKEY_TIMESTAMP_COL_ID) { + SColumnDataAgg nullColAgg = {.colId = pSup->colId[j], .numOfNull = numOfRows}; + TARRAY2_INSERT_PTR(&pSup->colAggArray, i, &nullColAgg); + i += 1; + size++; + } + j += 1; + } + } + + while (j < numOfCols) { + if (pSup->colId[j] != PRIMARYKEY_TIMESTAMP_COL_ID) { + SColumnDataAgg nullColAgg = {.colId = pSup->colId[j], .numOfNull = numOfRows}; + TARRAY2_INSERT_PTR(&pSup->colAggArray, i, &nullColAgg); + i += 1; + } + j++; + } +} + +int32_t tsdbRetrieveDatablockSMA2(STsdbReader* pReader, SSDataBlock* pDataBlock, bool* allHave, bool* hasNullSMA) { + SColumnDataAgg*** pBlockSMA = &pDataBlock->pBlockAgg; + + int32_t code = 0; + *allHave = false; + *pBlockSMA = NULL; + + if (pReader->type == TIMEWINDOW_RANGE_EXTERNAL) { + return TSDB_CODE_SUCCESS; + } + + // there is no statistics data for composed block + if (pReader->status.composedDataBlock || (!pReader->suppInfo.smaValid)) { + return TSDB_CODE_SUCCESS; + } + + SFileDataBlockInfo* pFBlock = getCurrentBlockInfo(&pReader->status.blockIter); + SBlockLoadSuppInfo* pSup = &pReader->suppInfo; + + SSDataBlock* pResBlock = pReader->resBlockInfo.pResBlock; + if (pResBlock->info.id.uid != pFBlock->uid) { + return TSDB_CODE_SUCCESS; + } + + // int64_t st = taosGetTimestampUs(); + TARRAY2_CLEAR(&pSup->colAggArray, 0); + + code = tsdbDataFileReadBlockSma(pReader->pFileReader, &pFBlock->record, &pSup->colAggArray); + if (code != TSDB_CODE_SUCCESS) { + tsdbDebug("vgId:%d, failed to load block SMA for uid %" PRIu64 ", code:%s, %s", 0, pFBlock->uid, tstrerror(code), + pReader->idStr); + return code; + } + + if (pSup->colAggArray.size > 0) { + *allHave = true; + } else { + *pBlockSMA = NULL; + return TSDB_CODE_SUCCESS; + } + + // always load the first primary timestamp column data + SColumnDataAgg* pTsAgg = &pSup->tsColAgg; + + pTsAgg->numOfNull = 0; + pTsAgg->colId = PRIMARYKEY_TIMESTAMP_COL_ID; + pTsAgg->min = pResBlock->info.window.skey; + pTsAgg->max = pResBlock->info.window.ekey; + + // update the number of NULL data rows + size_t numOfCols = pSup->numOfCols; + + if (pResBlock->pBlockAgg == NULL) { + size_t num = taosArrayGetSize(pResBlock->pDataBlock); + pResBlock->pBlockAgg = taosMemoryCalloc(num, POINTER_BYTES); + } + + // do fill all null column value SMA info + doFillNullColSMA(pSup, pFBlock->record.numRow, numOfCols, pTsAgg); + + size_t size = pSup->colAggArray.size; + + int32_t i = 0, j = 0; + while (j < numOfCols && i < size) { + SColumnDataAgg* pAgg = &pSup->colAggArray.data[i]; + if (pAgg->colId == pSup->colId[j]) { + pResBlock->pBlockAgg[pSup->slotId[j]] = pAgg; + i += 1; + j += 1; + } else if (pAgg->colId < pSup->colId[j]) { + i += 1; + } else if (pSup->colId[j] < pAgg->colId) { + pResBlock->pBlockAgg[pSup->slotId[j]] = NULL; + *allHave = false; + j += 1; + } + } + + *pBlockSMA = pResBlock->pBlockAgg; + pReader->cost.smaDataLoad += 1; + + // double elapsedTime = (taosGetTimestampUs() - st) / 1000.0; + pReader->cost.smaLoadTime += 0; // elapsedTime; + + tsdbDebug("vgId:%d, succeed to load block SMA for uid %" PRIu64 ", %s", 0, pFBlock->uid, pReader->idStr); + return code; +} + +static SSDataBlock* doRetrieveDataBlock(STsdbReader* pReader) { + SReaderStatus* pStatus = &pReader->status; + int32_t code = TSDB_CODE_SUCCESS; + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(&pStatus->blockIter); + + if (pReader->code != TSDB_CODE_SUCCESS) { + return NULL; + } + + STableBlockScanInfo* pBlockScanInfo = getTableBlockScanInfo(pStatus->pTableMap, pBlockInfo->uid, pReader->idStr); + if (pBlockScanInfo == NULL) { + return NULL; + } + + code = doLoadFileBlockData(pReader, &pStatus->blockIter, &pStatus->fileBlockData, pBlockScanInfo->uid); + if (code != TSDB_CODE_SUCCESS) { + tBlockDataDestroy(&pStatus->fileBlockData); + terrno = code; + return NULL; + } + + code = copyBlockDataToSDataBlock(pReader); + if (code != TSDB_CODE_SUCCESS) { + tBlockDataDestroy(&pStatus->fileBlockData); + terrno = code; + return NULL; + } + + return pReader->resBlockInfo.pResBlock; +} + +SSDataBlock* tsdbRetrieveDataBlock2(STsdbReader* pReader, SArray* pIdList) { + STsdbReader* pTReader = pReader; + if (pReader->type == TIMEWINDOW_RANGE_EXTERNAL) { + if (pReader->step == EXTERNAL_ROWS_PREV) { + pTReader = pReader->innerReader[0]; + } else if (pReader->step == EXTERNAL_ROWS_NEXT) { + pTReader = pReader->innerReader[1]; + } + } + + SReaderStatus* pStatus = &pTReader->status; + if (pStatus->composedDataBlock) { + return pTReader->resBlockInfo.pResBlock; + } + + SSDataBlock* ret = doRetrieveDataBlock(pTReader); + + qTrace("tsdb/read-retrieve: %p, unlock read mutex", pReader); + tsdbReleaseReader(pReader); + + return ret; +} + +int32_t tsdbReaderReset2(STsdbReader* pReader, SQueryTableDataCond* pCond) { + int32_t code = TSDB_CODE_SUCCESS; + + qTrace("tsdb/reader-reset: %p, take read mutex", pReader); + tsdbAcquireReader(pReader); + + if (pReader->flag == READER_STATUS_SUSPEND) { + code = tsdbReaderResume2(pReader); + if (code != TSDB_CODE_SUCCESS) { + tsdbReleaseReader(pReader); + return code; + } + } + + if (isEmptyQueryTimeWindow(&pReader->info.window) || pReader->pReadSnap == NULL) { + tsdbDebug("tsdb reader reset return %p, %s", pReader->pReadSnap, pReader->idStr); + tsdbReleaseReader(pReader); + return TSDB_CODE_SUCCESS; + } + + SReaderStatus* pStatus = &pReader->status; + SDataBlockIter* pBlockIter = &pStatus->blockIter; + + pReader->info.order = pCond->order; + pReader->type = TIMEWINDOW_RANGE_CONTAINED; + pStatus->loadFromFile = true; + pStatus->pTableIter = NULL; + pReader->info.window = updateQueryTimeWindow(pReader->pTsdb, &pCond->twindows); + + // allocate buffer in order to load data blocks from file + memset(&pReader->suppInfo.tsColAgg, 0, sizeof(SColumnDataAgg)); + + pReader->suppInfo.tsColAgg.colId = PRIMARYKEY_TIMESTAMP_COL_ID; + tsdbDataFileReaderClose(&pReader->pFileReader); + + int32_t numOfTables = tSimpleHashGetSize(pStatus->pTableMap); + + initFilesetIterator(&pStatus->fileIter, pReader->pReadSnap->pfSetArray, pReader); + resetDataBlockIterator(pBlockIter, pReader->info.order); + resetTableListIndex(&pReader->status); + + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + int32_t step = asc ? 1 : -1; + int64_t ts = asc ? pReader->info.window.skey - 1 : pReader->info.window.ekey + 1; + resetAllDataBlockScanInfo(pStatus->pTableMap, ts, step); + + // no data in files, let's try buffer in memory + if (pStatus->fileIter.numOfFiles == 0) { + pStatus->loadFromFile = false; + resetTableListIndex(pStatus); + } else { + code = initForFirstBlockInFile(pReader, pBlockIter); + if (code != TSDB_CODE_SUCCESS) { + tsdbError("%p reset reader failed, numOfTables:%d, query range:%" PRId64 " - %" PRId64 " in query %s", pReader, + numOfTables, pReader->info.window.skey, pReader->info.window.ekey, pReader->idStr); + + tsdbReleaseReader(pReader); + return code; + } + } + + tsdbDebug("%p reset reader, suid:%" PRIu64 ", numOfTables:%d, skey:%" PRId64 ", query range:%" PRId64 " - %" PRId64 + " in query %s", + pReader, pReader->info.suid, numOfTables, pCond->twindows.skey, pReader->info.window.skey, + pReader->info.window.ekey, pReader->idStr); + + tsdbReleaseReader(pReader); + + return code; +} + +static int32_t getBucketIndex(int32_t startRow, int32_t bucketRange, int32_t numOfRows, int32_t numOfBucket) { + if (numOfRows < startRow) { + return 0; + } + int32_t bucketIndex = ((numOfRows - startRow) / bucketRange); + if (bucketIndex == numOfBucket) { + bucketIndex -= 1; + } + return bucketIndex; +} + +int32_t tsdbGetFileBlocksDistInfo2(STsdbReader* pReader, STableBlockDistInfo* pTableBlockInfo) { + int32_t code = TSDB_CODE_SUCCESS; + pTableBlockInfo->totalSize = 0; + pTableBlockInfo->totalRows = 0; + pTableBlockInfo->numOfVgroups = 1; + + const int32_t numOfBuckets = 20.0; + const int32_t defaultRows = 4096; + + // find the start data block in file + tsdbAcquireReader(pReader); + if (pReader->flag == READER_STATUS_SUSPEND) { + code = tsdbReaderResume2(pReader); + if (code != TSDB_CODE_SUCCESS) { + tsdbReleaseReader(pReader); + return code; + } + } + SReaderStatus* pStatus = &pReader->status; + + STsdbCfg* pc = &pReader->pTsdb->pVnode->config.tsdbCfg; + pTableBlockInfo->defMinRows = pc->minRows; + pTableBlockInfo->defMaxRows = pc->maxRows; + + int32_t bucketRange = ceil(((double)(pc->maxRows - pc->minRows)) / numOfBuckets); + + pTableBlockInfo->numOfFiles += 1; + + int32_t numOfTables = (int32_t)tSimpleHashGetSize(pStatus->pTableMap); + + SDataBlockIter* pBlockIter = &pStatus->blockIter; + pTableBlockInfo->numOfFiles += pStatus->fileIter.numOfFiles; + + if (pBlockIter->numOfBlocks > 0) { + pTableBlockInfo->numOfBlocks += pBlockIter->numOfBlocks; + } + + pTableBlockInfo->numOfTables = numOfTables; + bool hasNext = (pBlockIter->numOfBlocks > 0); + + while (true) { + if (hasNext) { + SFileDataBlockInfo* pBlockInfo = getCurrentBlockInfo(pBlockIter); + int32_t numOfRows = pBlockInfo->record.numRow; + + pTableBlockInfo->totalRows += numOfRows; + + if (numOfRows > pTableBlockInfo->maxRows) { + pTableBlockInfo->maxRows = numOfRows; + } + + if (numOfRows < pTableBlockInfo->minRows) { + pTableBlockInfo->minRows = numOfRows; + } + + if (numOfRows < defaultRows) { + pTableBlockInfo->numOfSmallBlocks += 1; + } + + pTableBlockInfo->totalSize += pBlockInfo->record.blockSize; + + int32_t bucketIndex = getBucketIndex(pTableBlockInfo->defMinRows, bucketRange, numOfRows, numOfBuckets); + pTableBlockInfo->blockRowsHisto[bucketIndex]++; + + hasNext = blockIteratorNext(&pStatus->blockIter, pReader->idStr); + } else { + code = initForFirstBlockInFile(pReader, pBlockIter); + if ((code != TSDB_CODE_SUCCESS) || (pStatus->loadFromFile == false)) { + break; + } + + pTableBlockInfo->numOfBlocks += pBlockIter->numOfBlocks; + hasNext = (pBlockIter->numOfBlocks > 0); + } + + // tsdbDebug("%p %d blocks found in file for %d table(s), fid:%d, %s", pReader, numOfBlocks, numOfTables, + // pReader->pFileGroup->fid, pReader->idStr); + } + tsdbReleaseReader(pReader); + return code; +} + +int64_t tsdbGetNumOfRowsInMemTable2(STsdbReader* pReader) { + int32_t code = TSDB_CODE_SUCCESS; + int64_t rows = 0; + + SReaderStatus* pStatus = &pReader->status; + tsdbAcquireReader(pReader); + if (pReader->flag == READER_STATUS_SUSPEND) { + code = tsdbReaderResume2(pReader); + if (code != TSDB_CODE_SUCCESS) { + tsdbReleaseReader(pReader); + return code; + } + } + + int32_t iter = 0; + pStatus->pTableIter = tSimpleHashIterate(pStatus->pTableMap, NULL, &iter); + + while (pStatus->pTableIter != NULL) { + STableBlockScanInfo* pBlockScanInfo = *(STableBlockScanInfo**)pStatus->pTableIter; + + STbData* d = NULL; + if (pReader->pReadSnap->pMem != NULL) { + d = tsdbGetTbDataFromMemTable(pReader->pReadSnap->pMem, pReader->info.suid, pBlockScanInfo->uid); + if (d != NULL) { + rows += tsdbGetNRowsInTbData(d); + } + } + + STbData* di = NULL; + if (pReader->pReadSnap->pIMem != NULL) { + di = tsdbGetTbDataFromMemTable(pReader->pReadSnap->pIMem, pReader->info.suid, pBlockScanInfo->uid); + if (di != NULL) { + rows += tsdbGetNRowsInTbData(di); + } + } + + // current table is exhausted, let's try the next table + pStatus->pTableIter = tSimpleHashIterate(pStatus->pTableMap, pStatus->pTableIter, &iter); + } + + tsdbReleaseReader(pReader); + + return rows; +} + +int32_t tsdbGetTableSchema2(void* pVnode, int64_t uid, STSchema** pSchema, int64_t* suid) { + SMetaReader mr = {0}; + metaReaderDoInit(&mr, ((SVnode*)pVnode)->pMeta, 0); + int32_t code = metaReaderGetTableEntryByUidCache(&mr, uid); + if (code != TSDB_CODE_SUCCESS) { + terrno = TSDB_CODE_TDB_INVALID_TABLE_ID; + metaReaderClear(&mr); + return terrno; + } + + *suid = 0; + + // only child table and ordinary table is allowed, super table is not allowed. + if (mr.me.type == TSDB_CHILD_TABLE) { + tDecoderClear(&mr.coder); + *suid = mr.me.ctbEntry.suid; + code = metaReaderGetTableEntryByUidCache(&mr, *suid); + if (code != TSDB_CODE_SUCCESS) { + terrno = TSDB_CODE_TDB_INVALID_TABLE_ID; + metaReaderClear(&mr); + return terrno; + } + } else if (mr.me.type == TSDB_NORMAL_TABLE) { // do nothing + } else { + terrno = TSDB_CODE_INVALID_PARA; + metaReaderClear(&mr); + return terrno; + } + + metaReaderClear(&mr); + + // get the newest table schema version + code = metaGetTbTSchemaEx(((SVnode*)pVnode)->pMeta, *suid, uid, -1, pSchema); + return code; +} + +int32_t tsdbTakeReadSnap2(STsdbReader* pReader, _query_reseek_func_t reseek, STsdbReadSnap** ppSnap) { + int32_t code = 0; + STsdb* pTsdb = pReader->pTsdb; + SVersionRange* pRange = &pReader->info.verRange; + + // lock + taosThreadRwlockRdlock(&pTsdb->rwLock); + + // alloc + STsdbReadSnap* pSnap = (STsdbReadSnap*)taosMemoryCalloc(1, sizeof(STsdbReadSnap)); + if (pSnap == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + goto _exit; + } + + // take snapshot + if (pTsdb->mem && (pRange->minVer <= pTsdb->mem->maxVer && pRange->maxVer >= pTsdb->mem->minVer)) { + pSnap->pMem = pTsdb->mem; + pSnap->pNode = taosMemoryMalloc(sizeof(*pSnap->pNode)); + if (pSnap->pNode == NULL) { + taosThreadRwlockUnlock(&pTsdb->rwLock); + code = TSDB_CODE_OUT_OF_MEMORY; + goto _exit; + } + pSnap->pNode->pQHandle = pReader; + pSnap->pNode->reseek = reseek; + + tsdbRefMemTable(pTsdb->mem, pSnap->pNode); + } + + if (pTsdb->imem && (pRange->minVer <= pTsdb->imem->maxVer && pRange->maxVer >= pTsdb->imem->minVer)) { + pSnap->pIMem = pTsdb->imem; + pSnap->pINode = taosMemoryMalloc(sizeof(*pSnap->pINode)); + if (pSnap->pINode == NULL) { + taosThreadRwlockUnlock(&pTsdb->rwLock); + code = TSDB_CODE_OUT_OF_MEMORY; + goto _exit; + } + pSnap->pINode->pQHandle = pReader; + pSnap->pINode->reseek = reseek; + + tsdbRefMemTable(pTsdb->imem, pSnap->pINode); + } + + // fs + code = tsdbFSCreateRefSnapshot(pTsdb->pFS, &pSnap->pfSetArray); + if (code) { + taosThreadRwlockUnlock(&pTsdb->rwLock); + goto _exit; + } + + // unlock + taosThreadRwlockUnlock(&pTsdb->rwLock); + + tsdbTrace("vgId:%d, take read snapshot", TD_VID(pTsdb->pVnode)); + +_exit: + if (code) { + *ppSnap = NULL; + if (pSnap) { + if (pSnap->pNode) taosMemoryFree(pSnap->pNode); + if (pSnap->pINode) taosMemoryFree(pSnap->pINode); + taosMemoryFree(pSnap); + } + } else { + *ppSnap = pSnap; + } + + return code; +} + +void tsdbUntakeReadSnap2(STsdbReader* pReader, STsdbReadSnap* pSnap, bool proactive) { + STsdb* pTsdb = pReader->pTsdb; + + if (pSnap) { + if (pSnap->pMem) { + tsdbUnrefMemTable(pSnap->pMem, pSnap->pNode, proactive); + } + + if (pSnap->pIMem) { + tsdbUnrefMemTable(pSnap->pIMem, pSnap->pINode, proactive); + } + + tsdbFSUnref(pTsdb, &pSnap->fs); + if (pSnap->pNode) taosMemoryFree(pSnap->pNode); + if (pSnap->pINode) taosMemoryFree(pSnap->pINode); + + tsdbFSDestroyRefSnapshot(&pSnap->pfSetArray); + + taosMemoryFree(pSnap); + } + tsdbTrace("vgId:%d, untake read snapshot", TD_VID(pTsdb->pVnode)); +} + +// if failed, do nothing +void tsdbReaderSetId2(STsdbReader* pReader, const char* idstr) { + taosMemoryFreeClear(pReader->idStr); + pReader->idStr = taosStrdup(idstr); + pReader->status.fileIter.pLastBlockReader->mergeTree.idStr = pReader->idStr; +} + +void tsdbReaderSetCloseFlag2(STsdbReader* pReader) { pReader->code = TSDB_CODE_TSC_QUERY_CANCELLED; } diff --git a/source/dnode/vnode/src/tsdb/tsdbReadUtil.c b/source/dnode/vnode/src/tsdb/tsdbReadUtil.c new file mode 100644 index 0000000000..635a74d8dd --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbReadUtil.c @@ -0,0 +1,630 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbReadUtil.h" +#include "osDef.h" +#include "tsdb.h" +#include "tsdbDataFileRW.h" +#include "tsdbFS2.h" +#include "tsdbMerge.h" +#include "tsdbUtil2.h" +#include "tsimplehash.h" + +static int32_t uidComparFunc(const void* p1, const void* p2) { + uint64_t pu1 = *(uint64_t*)p1; + uint64_t pu2 = *(uint64_t*)p2; + if (pu1 == pu2) { + return 0; + } else { + return (pu1 < pu2) ? -1 : 1; + } +} + +static int32_t initBlockScanInfoBuf(SBlockInfoBuf* pBuf, int32_t numOfTables) { + int32_t num = numOfTables / pBuf->numPerBucket; + int32_t remainder = numOfTables % pBuf->numPerBucket; + if (pBuf->pData == NULL) { + pBuf->pData = taosArrayInit(num + 1, POINTER_BYTES); + } + + for (int32_t i = 0; i < num; ++i) { + char* p = taosMemoryCalloc(pBuf->numPerBucket, sizeof(STableBlockScanInfo)); + if (p == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + taosArrayPush(pBuf->pData, &p); + } + + if (remainder > 0) { + char* p = taosMemoryCalloc(remainder, sizeof(STableBlockScanInfo)); + if (p == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + taosArrayPush(pBuf->pData, &p); + } + + pBuf->numOfTables = numOfTables; + + return TSDB_CODE_SUCCESS; +} + +int32_t ensureBlockScanInfoBuf(SBlockInfoBuf* pBuf, int32_t numOfTables) { + if (numOfTables <= pBuf->numOfTables) { + return TSDB_CODE_SUCCESS; + } + + if (pBuf->numOfTables > 0) { + STableBlockScanInfo** p = (STableBlockScanInfo**)taosArrayPop(pBuf->pData); + taosMemoryFree(*p); + pBuf->numOfTables /= pBuf->numPerBucket; + } + + int32_t num = (numOfTables - pBuf->numOfTables) / pBuf->numPerBucket; + int32_t remainder = (numOfTables - pBuf->numOfTables) % pBuf->numPerBucket; + if (pBuf->pData == NULL) { + pBuf->pData = taosArrayInit(num + 1, POINTER_BYTES); + } + + for (int32_t i = 0; i < num; ++i) { + char* p = taosMemoryCalloc(pBuf->numPerBucket, sizeof(STableBlockScanInfo)); + if (p == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + taosArrayPush(pBuf->pData, &p); + } + + if (remainder > 0) { + char* p = taosMemoryCalloc(remainder, sizeof(STableBlockScanInfo)); + if (p == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + taosArrayPush(pBuf->pData, &p); + } + + pBuf->numOfTables = numOfTables; + + return TSDB_CODE_SUCCESS; +} + +void clearBlockScanInfoBuf(SBlockInfoBuf* pBuf) { + size_t num = taosArrayGetSize(pBuf->pData); + for (int32_t i = 0; i < num; ++i) { + char** p = taosArrayGet(pBuf->pData, i); + taosMemoryFree(*p); + } + + taosArrayDestroy(pBuf->pData); +} + +void* getPosInBlockInfoBuf(SBlockInfoBuf* pBuf, int32_t index) { + int32_t bucketIndex = index / pBuf->numPerBucket; + char** pBucket = taosArrayGet(pBuf->pData, bucketIndex); + return (*pBucket) + (index % pBuf->numPerBucket) * sizeof(STableBlockScanInfo); +} + +STableBlockScanInfo* getTableBlockScanInfo(SSHashObj* pTableMap, uint64_t uid, const char* id) { + STableBlockScanInfo** p = tSimpleHashGet(pTableMap, &uid, sizeof(uid)); + if (p == NULL || *p == NULL) { + terrno = TSDB_CODE_INVALID_PARA; + int32_t size = tSimpleHashGetSize(pTableMap); + tsdbError("failed to locate the uid:%" PRIu64 " in query table uid list, total tables:%d, %s", uid, size, id); + return NULL; + } + + return *p; +} + +// NOTE: speedup the whole processing by preparing the buffer for STableBlockScanInfo in batch model +SSHashObj* createDataBlockScanInfo(STsdbReader* pTsdbReader, SBlockInfoBuf* pBuf, const STableKeyInfo* idList, + STableUidList* pUidList, int32_t numOfTables) { + // allocate buffer in order to load data blocks from file + // todo use simple hash instead, optimize the memory consumption + SSHashObj* pTableMap = tSimpleHashInit(numOfTables, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT)); + if (pTableMap == NULL) { + return NULL; + } + + int64_t st = taosGetTimestampUs(); + initBlockScanInfoBuf(pBuf, numOfTables); + + pUidList->tableUidList = taosMemoryMalloc(numOfTables * sizeof(uint64_t)); + if (pUidList->tableUidList == NULL) { + tSimpleHashCleanup(pTableMap); + return NULL; + } + + pUidList->currentIndex = 0; + + for (int32_t j = 0; j < numOfTables; ++j) { + STableBlockScanInfo* pScanInfo = getPosInBlockInfoBuf(pBuf, j); + + pScanInfo->uid = idList[j].uid; + pUidList->tableUidList[j] = idList[j].uid; + + if (ASCENDING_TRAVERSE(pTsdbReader->info.order)) { + int64_t skey = pTsdbReader->info.window.skey; + pScanInfo->lastKey = (skey > INT64_MIN) ? (skey - 1) : skey; + pScanInfo->lastKeyInStt = skey; + } else { + int64_t ekey = pTsdbReader->info.window.ekey; + pScanInfo->lastKey = (ekey < INT64_MAX) ? (ekey + 1) : ekey; + pScanInfo->lastKeyInStt = ekey; + } + + tSimpleHashPut(pTableMap, &pScanInfo->uid, sizeof(uint64_t), &pScanInfo, POINTER_BYTES); + tsdbTrace("%p check table uid:%" PRId64 " from lastKey:%" PRId64 " %s", pTsdbReader, pScanInfo->uid, + pScanInfo->lastKey, pTsdbReader->idStr); + } + + taosSort(pUidList->tableUidList, numOfTables, sizeof(uint64_t), uidComparFunc); + + pTsdbReader->cost.createScanInfoList = (taosGetTimestampUs() - st) / 1000.0; + tsdbDebug("%p create %d tables scan-info, size:%.2f Kb, elapsed time:%.2f ms, %s", pTsdbReader, numOfTables, + (sizeof(STableBlockScanInfo) * numOfTables) / 1024.0, pTsdbReader->cost.createScanInfoList, + pTsdbReader->idStr); + + return pTableMap; +} + +void resetAllDataBlockScanInfo(SSHashObj* pTableMap, int64_t ts, int32_t step) { + void* p = NULL; + int32_t iter = 0; + + while ((p = tSimpleHashIterate(pTableMap, p, &iter)) != NULL) { + STableBlockScanInfo* pInfo = *(STableBlockScanInfo**)p; + + pInfo->iterInit = false; + pInfo->iter.hasVal = false; + pInfo->iiter.hasVal = false; + + if (pInfo->iter.iter != NULL) { + pInfo->iter.iter = tsdbTbDataIterDestroy(pInfo->iter.iter); + } + + if (pInfo->iiter.iter != NULL) { + pInfo->iiter.iter = tsdbTbDataIterDestroy(pInfo->iiter.iter); + } + + pInfo->delSkyline = taosArrayDestroy(pInfo->delSkyline); + pInfo->lastKey = ts; + pInfo->lastKeyInStt = ts + step; + } +} + +void clearBlockScanInfo(STableBlockScanInfo* p) { + p->iterInit = false; + p->iter.hasVal = false; + p->iiter.hasVal = false; + + if (p->iter.iter != NULL) { + p->iter.iter = tsdbTbDataIterDestroy(p->iter.iter); + } + + if (p->iiter.iter != NULL) { + p->iiter.iter = tsdbTbDataIterDestroy(p->iiter.iter); + } + + p->delSkyline = taosArrayDestroy(p->delSkyline); + p->pBlockList = taosArrayDestroy(p->pBlockList); + p->pMemDelData = taosArrayDestroy(p->pMemDelData); + p->pfileDelData = taosArrayDestroy(p->pfileDelData); +} + +void destroyAllBlockScanInfo(SSHashObj* pTableMap) { + void* p = NULL; + int32_t iter = 0; + + while ((p = tSimpleHashIterate(pTableMap, p, &iter)) != NULL) { + clearBlockScanInfo(*(STableBlockScanInfo**)p); + } + + tSimpleHashCleanup(pTableMap); +} + +static void doCleanupInfoForNextFileset(STableBlockScanInfo* pScanInfo) { + // reset the index in last block when handing a new file + taosArrayClear(pScanInfo->pBlockList); + taosArrayClear(pScanInfo->pfileDelData); // del data from each file set +} + +void cleanupInfoFoxNextFileset(SSHashObj* pTableMap) { + STableBlockScanInfo** p = NULL; + + int32_t iter = 0; + while ((p = tSimpleHashIterate(pTableMap, p, &iter)) != NULL) { + doCleanupInfoForNextFileset(*p); + } +} + +// brin records iterator +void initBrinRecordIter(SBrinRecordIter* pIter, SDataFileReader* pReader, SArray* pList) { + memset(&pIter->block, 0, sizeof(SBrinBlock)); + memset(&pIter->record, 0, sizeof(SBrinRecord)); + pIter->blockIndex = -1; + pIter->recordIndex = -1; + + pIter->pReader = pReader; + pIter->pBrinBlockList = pList; +} + +SBrinRecord* getNextBrinRecord(SBrinRecordIter* pIter) { + if (pIter->blockIndex == -1 || (pIter->recordIndex + 1) >= TARRAY2_SIZE(pIter->block.numRow)) { + pIter->blockIndex += 1; + if (pIter->blockIndex >= taosArrayGetSize(pIter->pBrinBlockList)) { + return NULL; + } + + pIter->pCurrentBlk = taosArrayGet(pIter->pBrinBlockList, pIter->blockIndex); + + tBrinBlockClear(&pIter->block); + tsdbDataFileReadBrinBlock(pIter->pReader, pIter->pCurrentBlk, &pIter->block); + pIter->recordIndex = -1; + } + + pIter->recordIndex += 1; + tBrinBlockGet(&pIter->block, pIter->recordIndex, &pIter->record); + return &pIter->record; +} + +void clearBrinBlockIter(SBrinRecordIter* pIter) { tBrinBlockDestroy(&pIter->block); } + +// initialize the file block access order +// sort the file blocks according to the offset of each data block in the files +static void cleanupBlockOrderSupporter(SBlockOrderSupporter* pSup) { + taosMemoryFreeClear(pSup->numOfBlocksPerTable); + taosMemoryFreeClear(pSup->indexPerTable); + + for (int32_t i = 0; i < pSup->numOfTables; ++i) { + SBlockOrderWrapper* pBlockInfo = pSup->pDataBlockInfo[i]; + taosMemoryFreeClear(pBlockInfo); + } + + taosMemoryFreeClear(pSup->pDataBlockInfo); +} + +static int32_t initBlockOrderSupporter(SBlockOrderSupporter* pSup, int32_t numOfTables) { + pSup->numOfBlocksPerTable = taosMemoryCalloc(1, sizeof(int32_t) * numOfTables); + pSup->indexPerTable = taosMemoryCalloc(1, sizeof(int32_t) * numOfTables); + pSup->pDataBlockInfo = taosMemoryCalloc(1, POINTER_BYTES * numOfTables); + + if (pSup->numOfBlocksPerTable == NULL || pSup->indexPerTable == NULL || pSup->pDataBlockInfo == NULL) { + cleanupBlockOrderSupporter(pSup); + return TSDB_CODE_OUT_OF_MEMORY; + } + + return TSDB_CODE_SUCCESS; +} + +static int32_t fileDataBlockOrderCompar(const void* pLeft, const void* pRight, void* param) { + int32_t leftIndex = *(int32_t*)pLeft; + int32_t rightIndex = *(int32_t*)pRight; + + SBlockOrderSupporter* pSupporter = (SBlockOrderSupporter*)param; + + int32_t leftTableBlockIndex = pSupporter->indexPerTable[leftIndex]; + int32_t rightTableBlockIndex = pSupporter->indexPerTable[rightIndex]; + + if (leftTableBlockIndex > pSupporter->numOfBlocksPerTable[leftIndex]) { + /* left block is empty */ + return 1; + } else if (rightTableBlockIndex > pSupporter->numOfBlocksPerTable[rightIndex]) { + /* right block is empty */ + return -1; + } + + SBlockOrderWrapper* pLeftBlock = &pSupporter->pDataBlockInfo[leftIndex][leftTableBlockIndex]; + SBlockOrderWrapper* pRightBlock = &pSupporter->pDataBlockInfo[rightIndex][rightTableBlockIndex]; + + return pLeftBlock->offset > pRightBlock->offset ? 1 : -1; +} + +int32_t initBlockIterator(STsdbReader* pReader, SDataBlockIter* pBlockIter, int32_t numOfBlocks, SArray* pTableList) { + bool asc = ASCENDING_TRAVERSE(pReader->info.order); + + SBlockOrderSupporter sup = {0}; + pBlockIter->numOfBlocks = numOfBlocks; + taosArrayClear(pBlockIter->blockList); + + pBlockIter->pTableMap = pReader->status.pTableMap; + + // access data blocks according to the offset of each block in asc/desc order. + int32_t numOfTables = taosArrayGetSize(pTableList); + + int64_t st = taosGetTimestampUs(); + int32_t code = initBlockOrderSupporter(&sup, numOfTables); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + int32_t cnt = 0; + + for (int32_t i = 0; i < numOfTables; ++i) { + STableBlockScanInfo* pTableScanInfo = taosArrayGetP(pTableList, i); + // ASSERT(pTableScanInfo->pBlockList != NULL && taosArrayGetSize(pTableScanInfo->pBlockList) > 0); + + size_t num = taosArrayGetSize(pTableScanInfo->pBlockList); + sup.numOfBlocksPerTable[sup.numOfTables] = num; + + char* buf = taosMemoryMalloc(sizeof(SBlockOrderWrapper) * num); + if (buf == NULL) { + cleanupBlockOrderSupporter(&sup); + return TSDB_CODE_OUT_OF_MEMORY; + } + + sup.pDataBlockInfo[sup.numOfTables] = (SBlockOrderWrapper*)buf; + + for (int32_t k = 0; k < num; ++k) { + SBrinRecord* pRecord = taosArrayGet(pTableScanInfo->pBlockList, k); + sup.pDataBlockInfo[sup.numOfTables][k] = + (SBlockOrderWrapper){.uid = pTableScanInfo->uid, .offset = pRecord->blockOffset, .pInfo = pTableScanInfo}; + cnt++; + } + + sup.numOfTables += 1; + } + + if (numOfBlocks != cnt && sup.numOfTables != numOfTables) { + cleanupBlockOrderSupporter(&sup); + return TSDB_CODE_INVALID_PARA; + } + + // since there is only one table qualified, blocks are not sorted + if (sup.numOfTables == 1) { + for (int32_t i = 0; i < numOfBlocks; ++i) { + SFileDataBlockInfo blockInfo = {.uid = sup.pDataBlockInfo[0][i].uid, .tbBlockIdx = i}; + blockInfo.record = *(SBrinRecord*)taosArrayGet(sup.pDataBlockInfo[0][i].pInfo->pBlockList, i); + + taosArrayPush(pBlockIter->blockList, &blockInfo); + } + + int64_t et = taosGetTimestampUs(); + tsdbDebug("%p create blocks info struct completed for one table, %d blocks not sorted, elapsed time:%.2f ms %s", + pReader, numOfBlocks, (et - st) / 1000.0, pReader->idStr); + + pBlockIter->index = asc ? 0 : (numOfBlocks - 1); + cleanupBlockOrderSupporter(&sup); + return TSDB_CODE_SUCCESS; + } + + tsdbDebug("%p create data blocks info struct completed, %d blocks in %d tables %s", pReader, cnt, sup.numOfTables, + pReader->idStr); + + SMultiwayMergeTreeInfo* pTree = NULL; + + uint8_t ret = tMergeTreeCreate(&pTree, sup.numOfTables, &sup, fileDataBlockOrderCompar); + if (ret != TSDB_CODE_SUCCESS) { + cleanupBlockOrderSupporter(&sup); + return TSDB_CODE_OUT_OF_MEMORY; + } + + int32_t numOfTotal = 0; + while (numOfTotal < cnt) { + int32_t pos = tMergeTreeGetChosenIndex(pTree); + int32_t index = sup.indexPerTable[pos]++; + + SFileDataBlockInfo blockInfo = {.uid = sup.pDataBlockInfo[pos][index].uid, .tbBlockIdx = index}; + blockInfo.record = *(SBrinRecord*)taosArrayGet(sup.pDataBlockInfo[pos][index].pInfo->pBlockList, index); + + taosArrayPush(pBlockIter->blockList, &blockInfo); + + // set data block index overflow, in order to disable the offset comparator + if (sup.indexPerTable[pos] >= sup.numOfBlocksPerTable[pos]) { + sup.indexPerTable[pos] = sup.numOfBlocksPerTable[pos] + 1; + } + + numOfTotal += 1; + tMergeTreeAdjust(pTree, tMergeTreeGetAdjustIndex(pTree)); + } + + int64_t et = taosGetTimestampUs(); + tsdbDebug("%p %d data blocks access order completed, elapsed time:%.2f ms %s", pReader, numOfBlocks, + (et - st) / 1000.0, pReader->idStr); + cleanupBlockOrderSupporter(&sup); + taosMemoryFree(pTree); + + pBlockIter->index = asc ? 0 : (numOfBlocks - 1); + return TSDB_CODE_SUCCESS; +} + +bool blockIteratorNext(SDataBlockIter* pBlockIter, const char* idStr) { + bool asc = ASCENDING_TRAVERSE(pBlockIter->order); + + int32_t step = asc ? 1 : -1; + if ((pBlockIter->index >= pBlockIter->numOfBlocks - 1 && asc) || (pBlockIter->index <= 0 && (!asc))) { + return false; + } + + pBlockIter->index += step; + return true; +} + +typedef enum { + BLK_CHECK_CONTINUE = 0x1, + BLK_CHECK_QUIT = 0x2, +} ETombBlkCheckEnum; + +static int32_t doCheckTombBlock(STombBlock* pBlock, STsdbReader* pReader, int32_t numOfTables, int32_t* j, + STableBlockScanInfo** pScanInfo, ETombBlkCheckEnum* pRet) { + int32_t code = 0; + STombRecord record = {0}; + uint64_t uid = pReader->status.uidList.tableUidList[*j]; + + for (int32_t k = 0; k < TARRAY2_SIZE(pBlock->suid); ++k) { + code = tTombBlockGet(pBlock, k, &record); + if (code != TSDB_CODE_SUCCESS) { + *pRet = BLK_CHECK_QUIT; + return code; + } + + if (record.suid < pReader->info.suid) { + continue; + } + + if (record.suid > pReader->info.suid) { + *pRet = BLK_CHECK_QUIT; + return TSDB_CODE_SUCCESS; + } + + bool newTable = false; + if (uid < record.uid) { + while ((*j) < numOfTables && pReader->status.uidList.tableUidList[*j] < record.uid) { + (*j) += 1; + newTable = true; + } + + if ((*j) >= numOfTables) { + *pRet = BLK_CHECK_QUIT; + return TSDB_CODE_SUCCESS; + } + + uid = pReader->status.uidList.tableUidList[*j]; + } + + if (record.uid < uid) { + continue; + } + + ASSERT(record.suid == pReader->info.suid && uid == record.uid); + + if (newTable) { + (*pScanInfo) = getTableBlockScanInfo(pReader->status.pTableMap, uid, pReader->idStr); + if ((*pScanInfo)->pfileDelData == NULL) { + (*pScanInfo)->pfileDelData = taosArrayInit(4, sizeof(SDelData)); + } + } + + if (record.version <= pReader->info.verRange.maxVer) { + SDelData delData = {.version = record.version, .sKey = record.skey, .eKey = record.ekey}; + taosArrayPush((*pScanInfo)->pfileDelData, &delData); + } + } + + *pRet = BLK_CHECK_CONTINUE; + return TSDB_CODE_SUCCESS; +} + +// load tomb data API +static int32_t doLoadTombDataFromTombBlk(const TTombBlkArray* pTombBlkArray, STsdbReader* pReader, void* pFileReader, + bool isFile) { + int32_t code = 0; + STableUidList* pList = &pReader->status.uidList; + int32_t numOfTables = tSimpleHashGetSize(pReader->status.pTableMap); + + int32_t i = 0, j = 0; + while (i < pTombBlkArray->size && j < numOfTables) { + STombBlk* pTombBlk = &pTombBlkArray->data[i]; + if (pTombBlk->maxTbid.suid < pReader->info.suid) { + i += 1; + continue; + } + + if (pTombBlk->minTbid.suid > pReader->info.suid) { + break; + } + + ASSERT(pTombBlk->minTbid.suid <= pReader->info.suid && pTombBlk->maxTbid.suid >= pReader->info.suid); + if (pTombBlk->maxTbid.suid == pReader->info.suid && pTombBlk->maxTbid.uid < pList->tableUidList[0]) { + i += 1; + continue; + } + + if (pTombBlk->minTbid.suid == pReader->info.suid && pTombBlk->minTbid.uid > pList->tableUidList[numOfTables - 1]) { + break; + } + + STombBlock block = {0}; + code = isFile ? tsdbDataFileReadTombBlock(pFileReader, &pTombBlkArray->data[i], &block) + : tsdbSttFileReadTombBlock(pFileReader, &pTombBlkArray->data[i], &block); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + uint64_t uid = pReader->status.uidList.tableUidList[j]; + + STableBlockScanInfo* pScanInfo = getTableBlockScanInfo(pReader->status.pTableMap, uid, pReader->idStr); + if (pScanInfo->pfileDelData == NULL) { + pScanInfo->pfileDelData = taosArrayInit(4, sizeof(SDelData)); + } + + ETombBlkCheckEnum ret = 0; + code = doCheckTombBlock(&block, pReader, numOfTables, &j, &pScanInfo, &ret); + + tTombBlockDestroy(&block); + if (code != TSDB_CODE_SUCCESS || ret == BLK_CHECK_QUIT) { + return code; + } + + i += 1; + } + + return TSDB_CODE_SUCCESS; +} + +int32_t loadDataFileTombDataForAll(STsdbReader* pReader) { + if (pReader->status.pCurrentFileset == NULL || pReader->status.pCurrentFileset->farr[3] == NULL) { + return TSDB_CODE_SUCCESS; + } + + const TTombBlkArray* pBlkArray = NULL; + + int32_t code = tsdbDataFileReadTombBlk(pReader->pFileReader, &pBlkArray); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + return doLoadTombDataFromTombBlk(pBlkArray, pReader, pReader->pFileReader, true); +} + +int32_t loadSttTombDataForAll(STsdbReader* pReader, SSttFileReader* pSttFileReader, SSttBlockLoadInfo* pLoadInfo) { + const TTombBlkArray* pBlkArray = NULL; + int32_t code = tsdbSttFileReadTombBlk(pSttFileReader, &pBlkArray); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + + return doLoadTombDataFromTombBlk(pBlkArray, pReader, pSttFileReader, false); +} + +void loadMemTombData(SArray** ppMemDelData, STbData* pMemTbData, STbData* piMemTbData, int64_t ver) { + if (*ppMemDelData == NULL) { + *ppMemDelData = taosArrayInit(4, sizeof(SDelData)); + } + + SArray* pMemDelData = *ppMemDelData; + + SDelData* p = NULL; + if (pMemTbData != NULL) { + p = pMemTbData->pHead; + while (p) { + if (p->version <= ver) { + taosArrayPush(pMemDelData, p); + } + + p = p->pNext; + } + } + + if (piMemTbData != NULL) { + p = piMemTbData->pHead; + while (p) { + if (p->version <= ver) { + taosArrayPush(pMemDelData, p); + } + p = p->pNext; + } + } +} diff --git a/source/dnode/vnode/src/tsdb/tsdbReadUtil.h b/source/dnode/vnode/src/tsdb/tsdbReadUtil.h new file mode 100644 index 0000000000..5c4737440d --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbReadUtil.h @@ -0,0 +1,248 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef TDENGINE_TSDBREADUTIL_H +#define TDENGINE_TSDBREADUTIL_H + +#ifdef __cplusplus +extern "C" { +#endif + +#include "tsdbDataFileRW.h" +#include "tsdbUtil2.h" + +#define ASCENDING_TRAVERSE(o) (o == TSDB_ORDER_ASC) + +typedef enum { + READER_STATUS_SUSPEND = 0x1, + READER_STATUS_NORMAL = 0x2, +} EReaderStatus; + +typedef enum { + EXTERNAL_ROWS_PREV = 0x1, + EXTERNAL_ROWS_MAIN = 0x2, + EXTERNAL_ROWS_NEXT = 0x3, +} EContentData; + +typedef struct SBlockInfoBuf { + int32_t currentIndex; + SArray* pData; + int32_t numPerBucket; + int32_t numOfTables; +} SBlockInfoBuf; + +typedef struct { + STbDataIter* iter; + int32_t index; + bool hasVal; +} SIterInfo; + +typedef struct STableBlockScanInfo { + uint64_t uid; + TSKEY lastKey; + TSKEY lastKeyInStt; // last accessed key in stt + SArray* pBlockList; // block data index list, SArray + SArray* pMemDelData; // SArray + SArray* pfileDelData; // SArray from each file set + SIterInfo iter; // mem buffer skip list iterator + SIterInfo iiter; // imem buffer skip list iterator + SArray* delSkyline; // delete info for this table + int32_t fileDelIndex; // file block delete index + int32_t lastBlockDelIndex; // delete index for last block + bool iterInit; // whether to initialize the in-memory skip list iterator or not +} STableBlockScanInfo; + +typedef struct SResultBlockInfo { + SSDataBlock* pResBlock; + bool freeBlock; + int64_t capacity; +} SResultBlockInfo; + +typedef struct SCostSummary { + int64_t numOfBlocks; + double blockLoadTime; + double buildmemBlock; + int64_t headFileLoad; + double headFileLoadTime; + int64_t smaDataLoad; + double smaLoadTime; + int64_t lastBlockLoad; + double lastBlockLoadTime; + int64_t composedBlocks; + double buildComposedBlockTime; + double createScanInfoList; + double createSkylineIterTime; + double initLastBlockReader; +} SCostSummary; + +typedef struct STableUidList { + uint64_t* tableUidList; // access table uid list in uid ascending order list + int32_t currentIndex; // index in table uid list +} STableUidList; + +typedef struct { + int32_t numOfBlocks; + int32_t numOfLastFiles; +} SBlockNumber; + +typedef struct SBlockIndex { + int32_t ordinalIndex; + int64_t inFileOffset; + STimeWindow window; // todo replace it with overlap flag. +} SBlockIndex; + +typedef struct SBlockOrderWrapper { + int64_t uid; + int64_t offset; + STableBlockScanInfo* pInfo; +} SBlockOrderWrapper; + +typedef struct SBlockOrderSupporter { + SBlockOrderWrapper** pDataBlockInfo; + int32_t* indexPerTable; + int32_t* numOfBlocksPerTable; + int32_t numOfTables; +} SBlockOrderSupporter; + +typedef struct SBlockLoadSuppInfo { + TColumnDataAggArray colAggArray; + SColumnDataAgg tsColAgg; + int16_t* colId; + int16_t* slotId; + int32_t numOfCols; + char** buildBuf; // build string tmp buffer, todo remove it later after all string format being updated. + bool smaValid; // the sma on all queried columns are activated +} SBlockLoadSuppInfo; + +typedef struct SLastBlockReader { + STimeWindow window; + SVersionRange verRange; + int32_t order; + uint64_t uid; + SMergeTree mergeTree; + SSttBlockLoadInfo* pInfo; + int64_t currentKey; +} SLastBlockReader; + +typedef struct SFilesetIter { + int32_t numOfFiles; // number of total files + int32_t index; // current accessed index in the list + TFileSetArray* pFilesetList; // data file set list + int32_t order; + SLastBlockReader* pLastBlockReader; // last file block reader +} SFilesetIter; + +typedef struct SFileDataBlockInfo { + // index position in STableBlockScanInfo in order to check whether neighbor block overlaps with it + uint64_t uid; + int32_t tbBlockIdx; + SBrinRecord record; +} SFileDataBlockInfo; + +typedef struct SDataBlockIter { + int32_t numOfBlocks; + int32_t index; + SArray* blockList; // SArray + int32_t order; + SDataBlk block; // current SDataBlk data + SSHashObj* pTableMap; +} SDataBlockIter; + +typedef struct SFileBlockDumpInfo { + int32_t totalRows; + int32_t rowIndex; + int64_t lastKey; + bool allDumped; +} SFileBlockDumpInfo; + +typedef struct SReaderStatus { + bool loadFromFile; // check file stage + bool composedDataBlock; // the returned data block is a composed block or not + SSHashObj* pTableMap; // SHash + STableBlockScanInfo** pTableIter; // table iterator used in building in-memory buffer data blocks. + STableUidList uidList; // check tables in uid order, to avoid the repeatly load of blocks in STT. + SFileBlockDumpInfo fBlockDumpInfo; + STFileSet* pCurrentFileset; // current opened file set + SBlockData fileBlockData; + SFilesetIter fileIter; + SDataBlockIter blockIter; + SArray* pLDataIterArray; + SRowMerger merger; + SColumnInfoData* pPrimaryTsCol; // primary time stamp output col info data +} SReaderStatus; + +struct STsdbReader { + STsdb* pTsdb; + STsdbReaderInfo info; + TdThreadMutex readerMutex; + EReaderStatus flag; + int32_t code; + uint64_t rowsNum; + SResultBlockInfo resBlockInfo; + SReaderStatus status; + char* idStr; // query info handle, for debug purpose + int32_t type; // query type: 1. retrieve all data blocks, 2. retrieve direct prev|next rows + SBlockLoadSuppInfo suppInfo; + STsdbReadSnap* pReadSnap; + SCostSummary cost; + SHashObj** pIgnoreTables; + SSHashObj* pSchemaMap; // keep the retrieved schema info, to avoid the overhead by repeatly load schema + SDataFileReader* pFileReader; // the file reader + SBlockInfoBuf blockInfoBuf; + EContentData step; + STsdbReader* innerReader[2]; +}; + +typedef struct SBrinRecordIter { + SArray* pBrinBlockList; + SBrinBlk* pCurrentBlk; + int32_t blockIndex; + int32_t recordIndex; + SDataFileReader* pReader; + SBrinBlock block; + SBrinRecord record; +} SBrinRecordIter; + +STableBlockScanInfo* getTableBlockScanInfo(SSHashObj* pTableMap, uint64_t uid, const char* id); + +SSHashObj* createDataBlockScanInfo(STsdbReader* pTsdbReader, SBlockInfoBuf* pBuf, const STableKeyInfo* idList, + STableUidList* pUidList, int32_t numOfTables); +void clearBlockScanInfo(STableBlockScanInfo* p); +void destroyAllBlockScanInfo(SSHashObj* pTableMap); +void resetAllDataBlockScanInfo(SSHashObj* pTableMap, int64_t ts, int32_t step); +void cleanupInfoFoxNextFileset(SSHashObj* pTableMap); +int32_t ensureBlockScanInfoBuf(SBlockInfoBuf* pBuf, int32_t numOfTables); +void clearBlockScanInfoBuf(SBlockInfoBuf* pBuf); +void* getPosInBlockInfoBuf(SBlockInfoBuf* pBuf, int32_t index); + +// brin records iterator +void initBrinRecordIter(SBrinRecordIter* pIter, SDataFileReader* pReader, SArray* pList); +SBrinRecord* getNextBrinRecord(SBrinRecordIter* pIter); +void clearBrinBlockIter(SBrinRecordIter* pIter); + +// initialize block iterator API +int32_t initBlockIterator(STsdbReader* pReader, SDataBlockIter* pBlockIter, int32_t numOfBlocks, SArray* pTableList); +bool blockIteratorNext(SDataBlockIter* pBlockIter, const char* idStr); + +// load tomb data API (stt/mem only for one table each, tomb data from data files are load for all tables at one time) +void loadMemTombData(SArray** ppMemDelData, STbData* pMemTbData, STbData* piMemTbData, int64_t ver); +int32_t loadDataFileTombDataForAll(STsdbReader* pReader); +int32_t loadSttTombDataForAll(STsdbReader* pReader, SSttFileReader* pSttFileReader, SSttBlockLoadInfo* pLoadInfo); + +#ifdef __cplusplus +} +#endif + +#endif // TDENGINE_TSDBREADUTIL_H diff --git a/source/dnode/vnode/src/tsdb/tsdbReaderWriter.c b/source/dnode/vnode/src/tsdb/tsdbReaderWriter.c index 4b677533e7..edd44e3727 100644 --- a/source/dnode/vnode/src/tsdb/tsdbReaderWriter.c +++ b/source/dnode/vnode/src/tsdb/tsdbReaderWriter.c @@ -16,7 +16,7 @@ #include "tsdb.h" // =============== PAGE-WISE FILE =============== -static int32_t tsdbOpenFile(const char *path, int32_t szPage, int32_t flag, STsdbFD **ppFD) { +int32_t tsdbOpenFile(const char *path, int32_t szPage, int32_t flag, STsdbFD **ppFD) { int32_t code = 0; STsdbFD *pFD = NULL; @@ -68,7 +68,7 @@ _exit: return code; } -static void tsdbCloseFile(STsdbFD **ppFD) { +void tsdbCloseFile(STsdbFD **ppFD) { STsdbFD *pFD = *ppFD; if (pFD) { taosMemoryFree(pFD->pBuf); @@ -141,7 +141,7 @@ _exit: return code; } -static int32_t tsdbWriteFile(STsdbFD *pFD, int64_t offset, const uint8_t *pBuf, int64_t size) { +int32_t tsdbWriteFile(STsdbFD *pFD, int64_t offset, const uint8_t *pBuf, int64_t size) { int32_t code = 0; int64_t fOffset = LOGIC_TO_FILE_OFFSET(offset, pFD->szPage); int64_t pgno = OFFSET_PGNO(fOffset, pFD->szPage); @@ -173,7 +173,7 @@ _exit: return code; } -static int32_t tsdbReadFile(STsdbFD *pFD, int64_t offset, uint8_t *pBuf, int64_t size) { +int32_t tsdbReadFile(STsdbFD *pFD, int64_t offset, uint8_t *pBuf, int64_t size) { int32_t code = 0; int64_t n = 0; int64_t fOffset = LOGIC_TO_FILE_OFFSET(offset, pFD->szPage); @@ -202,7 +202,7 @@ _exit: return code; } -static int32_t tsdbFsyncFile(STsdbFD *pFD) { +int32_t tsdbFsyncFile(STsdbFD *pFD) { int32_t code = 0; code = tsdbWriteFilePage(pFD); @@ -1489,7 +1489,7 @@ int32_t tsdbDelFReaderClose(SDelFReader **ppReader) { } int32_t tsdbReadDelData(SDelFReader *pReader, SDelIdx *pDelIdx, SArray *aDelData) { - return tsdbReadDelDatav1(pReader, pDelIdx, aDelData, INT64_MAX); + return tsdbReadDelDatav1(pReader, pDelIdx, aDelData, INT64_MAX); } int32_t tsdbReadDelDatav1(SDelFReader *pReader, SDelIdx *pDelIdx, SArray *aDelData, int64_t maxVer) { @@ -1517,10 +1517,10 @@ int32_t tsdbReadDelDatav1(SDelFReader *pReader, SDelIdx *pDelIdx, SArray *aDelDa if (delData.version > maxVer) { continue; } - if (taosArrayPush(aDelData, &delData) == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - goto _err; - } + if (taosArrayPush(aDelData, &delData) == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + goto _err; + } } ASSERT(n == size); diff --git a/source/dnode/vnode/src/tsdb/tsdbRetention.c b/source/dnode/vnode/src/tsdb/tsdbRetention.c index 7c7e1bd0f7..ae5ac4ae36 100644 --- a/source/dnode/vnode/src/tsdb/tsdbRetention.c +++ b/source/dnode/vnode/src/tsdb/tsdbRetention.c @@ -14,101 +14,265 @@ */ #include "tsdb.h" +#include "tsdbFS2.h" -static bool tsdbShouldDoRetentionImpl(STsdb *pTsdb, int64_t now) { - for (int32_t iSet = 0; iSet < taosArrayGetSize(pTsdb->fs.aDFileSet); iSet++) { - SDFileSet *pSet = (SDFileSet *)taosArrayGet(pTsdb->fs.aDFileSet, iSet); - int32_t expLevel = tsdbFidLevel(pSet->fid, &pTsdb->keepCfg, now); - SDiskID did; +typedef struct { + STsdb *tsdb; + int32_t szPage; + int64_t now; + int64_t cid; - if (expLevel == pSet->diskId.level) continue; + TFileSetArray *fsetArr; + TFileOpArray fopArr[1]; - if (expLevel < 0) { - return true; - } else { - if (tfsAllocDisk(pTsdb->pVnode->pTfs, expLevel, &did) < 0) { - return false; - } + struct { + int32_t fsetArrIdx; + STFileSet *fset; + } ctx[1]; +} SRTNer; - if (did.level == pSet->diskId.level) continue; +static int32_t tsdbDoRemoveFileObject(SRTNer *rtner, const STFileObj *fobj) { + STFileOp op = { + .optype = TSDB_FOP_REMOVE, + .fid = fobj->f->fid, + .of = fobj->f[0], + }; - return true; - } - } - - return false; -} -bool tsdbShouldDoRetention(STsdb *pTsdb, int64_t now) { - bool should; - taosThreadRwlockRdlock(&pTsdb->rwLock); - should = tsdbShouldDoRetentionImpl(pTsdb, now); - taosThreadRwlockUnlock(&pTsdb->rwLock); - return should; + return TARRAY2_APPEND(rtner->fopArr, op); } -int32_t tsdbDoRetention(STsdb *pTsdb, int64_t now) { +static int32_t tsdbDoCopyFile(SRTNer *rtner, const STFileObj *from, const STFile *to) { int32_t code = 0; int32_t lino = 0; - STsdbFS fs = {0}; - code = tsdbFSCopy(pTsdb, &fs); + char fname[TSDB_FILENAME_LEN]; + TdFilePtr fdFrom = NULL; + TdFilePtr fdTo = NULL; + + tsdbTFileName(rtner->tsdb, to, fname); + + fdFrom = taosOpenFile(from->fname, TD_FILE_READ); + if (fdFrom == NULL) code = terrno; TSDB_CHECK_CODE(code, lino, _exit); - for (int32_t iSet = 0; iSet < taosArrayGetSize(fs.aDFileSet); iSet++) { - SDFileSet *pSet = (SDFileSet *)taosArrayGet(fs.aDFileSet, iSet); - int32_t expLevel = tsdbFidLevel(pSet->fid, &pTsdb->keepCfg, now); - SDiskID did; + fdTo = taosOpenFile(fname, TD_FILE_WRITE | TD_FILE_CREATE | TD_FILE_TRUNC); + if (fdTo == NULL) code = terrno; + TSDB_CHECK_CODE(code, lino, _exit); - if (expLevel < 0) { - taosMemoryFree(pSet->pHeadF); - taosMemoryFree(pSet->pDataF); - taosMemoryFree(pSet->pSmaF); - for (int32_t iStt = 0; iStt < pSet->nSttF; iStt++) { - taosMemoryFree(pSet->aSttF[iStt]); - } - taosArrayRemove(fs.aDFileSet, iSet); - iSet--; - } else { - if (expLevel == 0) continue; - if (tfsAllocDisk(pTsdb->pVnode->pTfs, expLevel, &did) < 0) { - code = terrno; - goto _exit; - } - - if (did.level == pSet->diskId.level) continue; - - // copy file to new disk (todo) - SDFileSet fSet = *pSet; - fSet.diskId = did; - - code = tsdbDFileSetCopy(pTsdb, pSet, &fSet); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbFSUpsertFSet(&fs, &fSet); - TSDB_CHECK_CODE(code, lino, _exit); - } + int64_t n = taosFSendFile(fdTo, fdFrom, 0, tsdbLogicToFileSize(from->f->size, rtner->szPage)); + if (n < 0) { + code = TAOS_SYSTEM_ERROR(errno); + TSDB_CHECK_CODE(code, lino, _exit); } + taosCloseFile(&fdFrom); + taosCloseFile(&fdTo); - // do change fs - code = tsdbFSPrepareCommit(pTsdb, &fs); +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(rtner->tsdb->pVnode), lino, code); + taosCloseFile(&fdFrom); + taosCloseFile(&fdTo); + } + return code; +} + +static int32_t tsdbDoMigrateFileObj(SRTNer *rtner, const STFileObj *fobj, const SDiskID *did) { + int32_t code = 0; + int32_t lino = 0; + STFileOp op = {0}; + + // remove old + op = (STFileOp){ + .optype = TSDB_FOP_REMOVE, + .fid = fobj->f->fid, + .of = fobj->f[0], + }; + + code = TARRAY2_APPEND(rtner->fopArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + + // create new + op = (STFileOp){ + .optype = TSDB_FOP_CREATE, + .fid = fobj->f->fid, + .nf = + { + .type = fobj->f->type, + .did = did[0], + .fid = fobj->f->fid, + .cid = rtner->cid, + .size = fobj->f->size, + .stt[0] = + { + .level = fobj->f->stt[0].level, + }, + }, + }; + + code = TARRAY2_APPEND(rtner->fopArr, op); + TSDB_CHECK_CODE(code, lino, _exit); + + // do copy the file + code = tsdbDoCopyFile(rtner, fobj, &op.nf); TSDB_CHECK_CODE(code, lino, _exit); _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbInfo("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); + TSDB_ERROR_LOG(TD_VID(rtner->tsdb->pVnode), lino, code); } - tsdbFSDestroy(&fs); return code; } -static int32_t tsdbCommitRetentionImpl(STsdb *pTsdb) { return tsdbFSCommit(pTsdb); } +typedef struct { + STsdb *tsdb; + int64_t now; +} SRtnArg; -int32_t tsdbCommitRetention(STsdb *pTsdb) { - taosThreadRwlockWrlock(&pTsdb->rwLock); - tsdbCommitRetentionImpl(pTsdb); - taosThreadRwlockUnlock(&pTsdb->rwLock); - tsdbInfo("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); - return 0; +static int32_t tsdbDoRetentionBegin(SRtnArg *arg, SRTNer *rtner) { + int32_t code = 0; + int32_t lino = 0; + + STsdb *tsdb = arg->tsdb; + + rtner->tsdb = tsdb; + rtner->szPage = tsdb->pVnode->config.tsdbPageSize; + rtner->now = arg->now; + rtner->cid = tsdbFSAllocEid(tsdb->pFS); + + code = tsdbFSCreateCopySnapshot(tsdb->pFS, &rtner->fsetArr); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(rtner->tsdb->pVnode), lino, code); + } else { + tsdbInfo("vid:%d, cid:%" PRId64 ", %s done", TD_VID(rtner->tsdb->pVnode), rtner->cid, __func__); + } + return code; +} + +static int32_t tsdbDoRetentionEnd(SRTNer *rtner) { + int32_t code = 0; + int32_t lino = 0; + + if (TARRAY2_SIZE(rtner->fopArr) == 0) goto _exit; + + code = tsdbFSEditBegin(rtner->tsdb->pFS, rtner->fopArr, TSDB_FEDIT_MERGE); + TSDB_CHECK_CODE(code, lino, _exit); + + taosThreadRwlockWrlock(&rtner->tsdb->rwLock); + + code = tsdbFSEditCommit(rtner->tsdb->pFS); + if (code) { + taosThreadRwlockUnlock(&rtner->tsdb->rwLock); + TSDB_CHECK_CODE(code, lino, _exit); + } + + taosThreadRwlockUnlock(&rtner->tsdb->rwLock); + + TARRAY2_DESTROY(rtner->fopArr, NULL); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(rtner->tsdb->pVnode), lino, code); + } else { + tsdbInfo("vid:%d, cid:%" PRId64 ", %s done", TD_VID(rtner->tsdb->pVnode), rtner->cid, __func__); + } + tsdbFSDestroyCopySnapshot(&rtner->fsetArr); + return code; +} + +static int32_t tsdbDoRetention2(void *arg) { + int32_t code = 0; + int32_t lino = 0; + SRTNer rtner[1] = {0}; + + code = tsdbDoRetentionBegin(arg, rtner); + TSDB_CHECK_CODE(code, lino, _exit); + + for (rtner->ctx->fsetArrIdx = 0; rtner->ctx->fsetArrIdx < TARRAY2_SIZE(rtner->fsetArr); rtner->ctx->fsetArrIdx++) { + rtner->ctx->fset = TARRAY2_GET(rtner->fsetArr, rtner->ctx->fsetArrIdx); + + STFileObj *fobj; + int32_t expLevel = tsdbFidLevel(rtner->ctx->fset->fid, &rtner->tsdb->keepCfg, rtner->now); + + if (expLevel < 0) { // remove the file set + for (int32_t ftype = 0; (ftype < TSDB_FTYPE_MAX) && (fobj = rtner->ctx->fset->farr[ftype], 1); ++ftype) { + if (fobj == NULL) continue; + + code = tsdbDoRemoveFileObject(rtner, fobj); + TSDB_CHECK_CODE(code, lino, _exit); + } + + SSttLvl *lvl; + TARRAY2_FOREACH(rtner->ctx->fset->lvlArr, lvl) { + TARRAY2_FOREACH(lvl->fobjArr, fobj) { + code = tsdbDoRemoveFileObject(rtner, fobj); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + } else if (expLevel == 0) { + continue; + } else { + SDiskID did; + + if (tfsAllocDisk(rtner->tsdb->pVnode->pTfs, expLevel, &did) < 0) { + code = terrno; + TSDB_CHECK_CODE(code, lino, _exit); + } + tfsMkdirRecurAt(rtner->tsdb->pVnode->pTfs, rtner->tsdb->path, did); + + // data + for (int32_t ftype = 0; ftype < TSDB_FTYPE_MAX && (fobj = rtner->ctx->fset->farr[ftype], 1); ++ftype) { + if (fobj == NULL) continue; + + if (fobj->f->did.level == did.level) continue; + code = tsdbDoMigrateFileObj(rtner, fobj, &did); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // stt + SSttLvl *lvl; + TARRAY2_FOREACH(rtner->ctx->fset->lvlArr, lvl) { + TARRAY2_FOREACH(lvl->fobjArr, fobj) { + if (fobj->f->did.level == did.level) continue; + + code = tsdbDoMigrateFileObj(rtner, fobj, &did); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + } + } + + code = tsdbDoRetentionEnd(rtner); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(rtner->tsdb->pVnode), lino, code); + } + taosMemoryFree(arg); + return code; +} + +int32_t tsdbAsyncRetention(STsdb *tsdb, int64_t now, int64_t *taskid) { + SRtnArg *arg = taosMemoryMalloc(sizeof(*arg)); + if (arg == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + arg->tsdb = tsdb; + arg->now = now; + + int32_t code = tsdbFSScheduleBgTask(tsdb->pFS, TSDB_BG_TASK_RETENTION, tsdbDoRetention2, arg, taskid); + if (code) taosMemoryFree(arg); + + return code; +} + +int32_t tsdbSyncRetention(STsdb *tsdb, int64_t now) { + int64_t taskid; + + int32_t code = tsdbAsyncRetention(tsdb, now, &taskid); + if (code) return code; + + return tsdbFSWaitBgTask(tsdb->pFS, taskid); } \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbSnapshot.c b/source/dnode/vnode/src/tsdb/tsdbSnapshot.c index df2aebe45b..011b9bd5a4 100644 --- a/source/dnode/vnode/src/tsdb/tsdbSnapshot.c +++ b/source/dnode/vnode/src/tsdb/tsdbSnapshot.c @@ -14,553 +14,519 @@ */ #include "tsdb.h" +#include "tsdbDataFileRW.h" +#include "tsdbFS2.h" +#include "tsdbFSetRW.h" +#include "tsdbIter.h" +#include "tsdbSttFileRW.h" extern int32_t tsdbUpdateTableSchema(SMeta* pMeta, int64_t suid, int64_t uid, SSkmInfo* pSkmInfo); -extern int32_t tsdbWriteDataBlock(SDataFWriter* pWriter, SBlockData* pBlockData, SMapData* mDataBlk, int8_t cmprAlg); -extern int32_t tsdbWriteSttBlock(SDataFWriter* pWriter, SBlockData* pBlockData, SArray* aSttBlk, int8_t cmprAlg); // STsdbSnapReader ======================================== struct STsdbSnapReader { - STsdb* pTsdb; - int64_t sver; - int64_t ever; - int8_t type; + STsdb* tsdb; + int64_t sver; + int64_t ever; + int8_t type; + uint8_t* aBuf[5]; + SSkmInfo skmTb[1]; - STsdbFS fs; - TABLEID tbid; - SSkmInfo skmTable; + TFileSetArray* fsetArr; - // timeseries data - int8_t dataDone; - int32_t fid; + // context + struct { + int32_t fsetArrIdx; + STFileSet* fset; + bool isDataDone; + bool isTombDone; + } ctx[1]; - SDataFReader* pDataFReader; - STsdbDataIter2* iterList; - STsdbDataIter2* pIter; - SRBTree rbt; - SBlockData bData; + // reader + SDataFileReader* dataReader; + TSttFileReaderArray sttReaderArr[1]; - // tombstone data - int8_t delDone; - SDelFReader* pDelFReader; - STsdbDataIter2* pTIter; - SArray* aDelData; + // iter + TTsdbIterArray dataIterArr[1]; + SIterMerger* dataIterMerger; + TTsdbIterArray tombIterArr[1]; + SIterMerger* tombIterMerger; + + // data + SBlockData blockData[1]; + STombBlock tombBlock[1]; }; -static int32_t tsdbSnapReadFileDataStart(STsdbSnapReader* pReader) { +static int32_t tsdbSnapReadFileSetOpenReader(STsdbSnapReader* reader) { int32_t code = 0; int32_t lino = 0; - SDFileSet* pSet = taosArraySearch(pReader->fs.aDFileSet, &(SDFileSet){.fid = pReader->fid}, tDFileSetCmprFn, TD_GT); - if (pSet == NULL) { - pReader->fid = INT32_MAX; - goto _exit; - } + ASSERT(reader->dataReader == NULL); + ASSERT(TARRAY2_SIZE(reader->sttReaderArr) == 0); - pReader->fid = pSet->fid; - - tRBTreeCreate(&pReader->rbt, tsdbDataIterCmprFn); - - code = tsdbDataFReaderOpen(&pReader->pDataFReader, pReader->pTsdb, pSet); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbOpenDataFileDataIter(pReader->pDataFReader, &pReader->pIter); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pReader->pIter) { - // iter to next with filter info (sver, ever) - code = tsdbDataIterNext2( - pReader->pIter, - &(STsdbFilterInfo){.flag = TSDB_FILTER_FLAG_BY_VERSION | TSDB_FILTER_FLAG_IGNORE_DROPPED_TABLE, // flag - .sver = pReader->sver, - .ever = pReader->ever}); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pReader->pIter->rowInfo.suid || pReader->pIter->rowInfo.uid) { - // add to rbtree - tRBTreePut(&pReader->rbt, &pReader->pIter->rbtn); - - // add to iterList - pReader->pIter->next = pReader->iterList; - pReader->iterList = pReader->pIter; - } else { - tsdbCloseDataIter2(pReader->pIter); + // data + SDataFileReaderConfig config = { + .tsdb = reader->tsdb, + .szPage = reader->tsdb->pVnode->config.tsdbPageSize, + .bufArr = reader->aBuf, + }; + bool hasDataFile = false; + for (int32_t ftype = 0; ftype < TSDB_FTYPE_MAX; ftype++) { + if (reader->ctx->fset->farr[ftype] != NULL) { + hasDataFile = true; + config.files[ftype].exist = true; + config.files[ftype].file = reader->ctx->fset->farr[ftype]->f[0]; } } - for (int32_t iStt = 0; iStt < pSet->nSttF; ++iStt) { - code = tsdbOpenSttFileDataIter(pReader->pDataFReader, iStt, &pReader->pIter); + if (hasDataFile) { + code = tsdbDataFileReaderOpen(NULL, &config, &reader->dataReader); TSDB_CHECK_CODE(code, lino, _exit); + } - if (pReader->pIter) { - // iter to valid row - code = tsdbDataIterNext2( - pReader->pIter, - &(STsdbFilterInfo){.flag = TSDB_FILTER_FLAG_BY_VERSION | TSDB_FILTER_FLAG_IGNORE_DROPPED_TABLE, // flag - .sver = pReader->sver, - .ever = pReader->ever}); + // stt + SSttLvl* lvl; + TARRAY2_FOREACH(reader->ctx->fset->lvlArr, lvl) { + STFileObj* fobj; + TARRAY2_FOREACH(lvl->fobjArr, fobj) { + SSttFileReader* sttReader; + SSttFileReaderConfig config = { + .tsdb = reader->tsdb, + .szPage = reader->tsdb->pVnode->config.tsdbPageSize, + .file = fobj->f[0], + .bufArr = reader->aBuf, + }; + + code = tsdbSttFileReaderOpen(fobj->fname, &config, &sttReader); TSDB_CHECK_CODE(code, lino, _exit); - if (pReader->pIter->rowInfo.suid || pReader->pIter->rowInfo.uid) { - // add to rbtree - tRBTreePut(&pReader->rbt, &pReader->pIter->rbtn); - - // add to iterList - pReader->pIter->next = pReader->iterList; - pReader->iterList = pReader->pIter; - } else { - tsdbCloseDataIter2(pReader->pIter); - } + code = TARRAY2_APPEND(reader->sttReaderArr, sttReader); + TSDB_CHECK_CODE(code, lino, _exit); } } - pReader->pIter = NULL; - _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pReader->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbInfo("vgId:%d %s done, fid:%d", TD_VID(pReader->pTsdb->pVnode), __func__, pReader->fid); + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), code, lino); } return code; } -static void tsdbSnapReadFileDataEnd(STsdbSnapReader* pReader) { - while (pReader->iterList) { - STsdbDataIter2* pIter = pReader->iterList; - pReader->iterList = pIter->next; - tsdbCloseDataIter2(pIter); - } - - tsdbDataFReaderClose(&pReader->pDataFReader); -} - -static int32_t tsdbSnapReadNextRow(STsdbSnapReader* pReader, SRowInfo** ppRowInfo) { +static int32_t tsdbSnapReadFileSetCloseReader(STsdbSnapReader* reader) { int32_t code = 0; int32_t lino = 0; - if (pReader->pIter) { - code = tsdbDataIterNext2(pReader->pIter, &(STsdbFilterInfo){.flag = TSDB_FILTER_FLAG_BY_VERSION | - TSDB_FILTER_FLAG_IGNORE_DROPPED_TABLE, // flag - .sver = pReader->sver, - .ever = pReader->ever}); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pReader->pIter->rowInfo.suid == 0 && pReader->pIter->rowInfo.uid == 0) { - pReader->pIter = NULL; - } else { - SRBTreeNode* pNode = tRBTreeMin(&pReader->rbt); - if (pNode) { - int32_t c = tsdbDataIterCmprFn(&pReader->pIter->rbtn, pNode); - if (c > 0) { - tRBTreePut(&pReader->rbt, &pReader->pIter->rbtn); - pReader->pIter = NULL; - } else if (c == 0) { - ASSERT(0); - } - } - } - } - - if (pReader->pIter == NULL) { - SRBTreeNode* pNode = tRBTreeMin(&pReader->rbt); - if (pNode) { - tRBTreeDrop(&pReader->rbt, pNode); - pReader->pIter = TSDB_RBTN_TO_DATA_ITER(pNode); - } - } - - if (ppRowInfo) { - if (pReader->pIter) { - *ppRowInfo = &pReader->pIter->rowInfo; - } else { - *ppRowInfo = NULL; - } - } + TARRAY2_CLEAR(reader->sttReaderArr, tsdbSttFileReaderClose); + tsdbDataFileReaderClose(&reader->dataReader); _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pReader->pTsdb->pVnode), __func__, lino, tstrerror(code)); + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), code, lino); } return code; } -static int32_t tsdbSnapReadGetRow(STsdbSnapReader* pReader, SRowInfo** ppRowInfo) { - if (pReader->pIter) { - *ppRowInfo = &pReader->pIter->rowInfo; - return 0; +static int32_t tsdbSnapReadFileSetOpenIter(STsdbSnapReader* reader) { + int32_t code = 0; + int32_t lino = 0; + + ASSERT(reader->dataIterMerger == NULL); + ASSERT(reader->tombIterMerger == NULL); + ASSERT(TARRAY2_SIZE(reader->dataIterArr) == 0); + ASSERT(TARRAY2_SIZE(reader->tombIterArr) == 0); + + STsdbIter* iter; + STsdbIterConfig config = { + .filterByVersion = true, + .verRange[0] = reader->sver, + .verRange[1] = reader->ever, + }; + + // data file + if (reader->dataReader) { + // data + config.type = TSDB_ITER_TYPE_DATA; + config.dataReader = reader->dataReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(reader->dataIterArr, iter); + TSDB_CHECK_CODE(code, lino, _exit); + + // tomb + config.type = TSDB_ITER_TYPE_DATA_TOMB; + config.dataReader = reader->dataReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(reader->tombIterArr, iter); + TSDB_CHECK_CODE(code, lino, _exit); } - return tsdbSnapReadNextRow(pReader, ppRowInfo); + // stt file + SSttFileReader* sttReader; + TARRAY2_FOREACH(reader->sttReaderArr, sttReader) { + // data + config.type = TSDB_ITER_TYPE_STT; + config.sttReader = sttReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(reader->dataIterArr, iter); + TSDB_CHECK_CODE(code, lino, _exit); + + // tomb + config.type = TSDB_ITER_TYPE_STT_TOMB; + config.sttReader = sttReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(reader->tombIterArr, iter); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // merger + code = tsdbIterMergerOpen(reader->dataIterArr, &reader->dataIterMerger, false); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbIterMergerOpen(reader->tombIterArr, &reader->tombIterMerger, true); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), code, lino); + } + return code; } -static int32_t tsdbSnapCmprData(STsdbSnapReader* pReader, uint8_t** ppData) { - int32_t code = 0; +static int32_t tsdbSnapReadFileSetCloseIter(STsdbSnapReader* reader) { + tsdbIterMergerClose(&reader->dataIterMerger); + tsdbIterMergerClose(&reader->tombIterMerger); + TARRAY2_CLEAR(reader->dataIterArr, tsdbIterClose); + TARRAY2_CLEAR(reader->tombIterArr, tsdbIterClose); + return 0; +} - ASSERT(pReader->bData.nRow); +static int32_t tsdbSnapReadFileSetBegin(STsdbSnapReader* reader) { + int32_t code = 0; + int32_t lino = 0; + + ASSERT(reader->ctx->fset == NULL); + + if (reader->ctx->fsetArrIdx < TARRAY2_SIZE(reader->fsetArr)) { + reader->ctx->fset = TARRAY2_GET(reader->fsetArr, reader->ctx->fsetArrIdx++); + reader->ctx->isDataDone = false; + reader->ctx->isTombDone = false; + + code = tsdbSnapReadFileSetOpenReader(reader); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSnapReadFileSetOpenIter(reader); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), code, lino); + } + return code; +} + +static int32_t tsdbSnapReadFileSetEnd(STsdbSnapReader* reader) { + tsdbSnapReadFileSetCloseIter(reader); + tsdbSnapReadFileSetCloseReader(reader); + reader->ctx->fset = NULL; + return 0; +} + +static int32_t tsdbSnapCmprData(STsdbSnapReader* reader, uint8_t** data) { + int32_t code = 0; + int32_t lino = 0; int32_t aBufN[5] = {0}; - code = tCmprBlockData(&pReader->bData, NO_COMPRESSION, NULL, NULL, pReader->aBuf, aBufN); - if (code) goto _exit; + code = tCmprBlockData(reader->blockData, NO_COMPRESSION, NULL, NULL, reader->aBuf, aBufN); + TSDB_CHECK_CODE(code, lino, _exit); int32_t size = aBufN[0] + aBufN[1] + aBufN[2] + aBufN[3]; - *ppData = taosMemoryMalloc(sizeof(SSnapDataHdr) + size); - if (*ppData == NULL) { + *data = taosMemoryMalloc(sizeof(SSnapDataHdr) + size); + if (*data == NULL) { code = TSDB_CODE_OUT_OF_MEMORY; - goto _exit; + TSDB_CHECK_CODE(code, lino, _exit); } - SSnapDataHdr* pHdr = (SSnapDataHdr*)*ppData; - pHdr->type = pReader->type; + SSnapDataHdr* pHdr = (SSnapDataHdr*)*data; + pHdr->type = reader->type; pHdr->size = size; - memcpy(pHdr->data, pReader->aBuf[3], aBufN[3]); - memcpy(pHdr->data + aBufN[3], pReader->aBuf[2], aBufN[2]); + memcpy(pHdr->data, reader->aBuf[3], aBufN[3]); + memcpy(pHdr->data + aBufN[3], reader->aBuf[2], aBufN[2]); if (aBufN[1]) { - memcpy(pHdr->data + aBufN[3] + aBufN[2], pReader->aBuf[1], aBufN[1]); + memcpy(pHdr->data + aBufN[3] + aBufN[2], reader->aBuf[1], aBufN[1]); } if (aBufN[0]) { - memcpy(pHdr->data + aBufN[3] + aBufN[2] + aBufN[1], pReader->aBuf[0], aBufN[0]); + memcpy(pHdr->data + aBufN[3] + aBufN[2] + aBufN[1], reader->aBuf[0], aBufN[0]); } _exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), lino, code); + } return code; } -static int32_t tsdbSnapReadTimeSeriesData(STsdbSnapReader* pReader, uint8_t** ppData) { - int32_t code = 0; - int32_t lino = 0; +static int32_t tsdbSnapReadTimeSeriesData(STsdbSnapReader* reader, uint8_t** data) { + int32_t code = 0; + int32_t lino = 0; + SMetaInfo info; - STsdb* pTsdb = pReader->pTsdb; + tBlockDataReset(reader->blockData); - tBlockDataReset(&pReader->bData); - - for (;;) { - // start a new file read if need - if (pReader->pDataFReader == NULL) { - code = tsdbSnapReadFileDataStart(pReader); - TSDB_CHECK_CODE(code, lino, _exit); - } - - if (pReader->pDataFReader == NULL) break; - - SRowInfo* pRowInfo; - code = tsdbSnapReadGetRow(pReader, &pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pRowInfo == NULL) { - tsdbSnapReadFileDataEnd(pReader); - continue; - } - - code = tsdbUpdateTableSchema(pTsdb->pVnode->pMeta, pRowInfo->suid, pRowInfo->uid, &pReader->skmTable); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tBlockDataInit(&pReader->bData, (TABLEID*)pRowInfo, pReader->skmTable.pTSchema, NULL, 0); - TSDB_CHECK_CODE(code, lino, _exit); - - do { - if (!TABLE_SAME_SCHEMA(pReader->bData.suid, pReader->bData.uid, pRowInfo->suid, pRowInfo->uid)) break; - - if (pReader->bData.uid && pReader->bData.uid != pRowInfo->uid) { - code = tRealloc((uint8_t**)&pReader->bData.aUid, sizeof(int64_t) * (pReader->bData.nRow + 1)); + TABLEID tbid[1] = {0}; + for (SRowInfo* row; (row = tsdbIterMergerGetData(reader->dataIterMerger));) { + // skip dropped table + if (row->uid != tbid->uid) { + tbid->suid = row->suid; + tbid->uid = row->uid; + if (metaGetInfo(reader->tsdb->pVnode->pMeta, tbid->uid, &info, NULL) != 0) { + code = tsdbIterMergerSkipTableData(reader->dataIterMerger, tbid); TSDB_CHECK_CODE(code, lino, _exit); - - for (int32_t iRow = 0; iRow < pReader->bData.nRow; ++iRow) { - pReader->bData.aUid[iRow] = pReader->bData.uid; - } - pReader->bData.uid = 0; + continue; } - - code = tBlockDataAppendRow(&pReader->bData, &pRowInfo->row, NULL, pRowInfo->uid); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbSnapReadNextRow(pReader, &pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pReader->bData.nRow >= 81920) break; - } while (pRowInfo); - - ASSERT(pReader->bData.nRow > 0); - - break; - } - - if (pReader->bData.nRow > 0) { - ASSERT(pReader->bData.suid || pReader->bData.uid); - - code = tsdbSnapCmprData(pReader, ppData); - TSDB_CHECK_CODE(code, lino, _exit); - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - return code; -} - -static int32_t tsdbSnapCmprTombData(STsdbSnapReader* pReader, uint8_t** ppData) { - int32_t code = 0; - int32_t lino = 0; - - int64_t size = sizeof(TABLEID); - for (int32_t iDelData = 0; iDelData < taosArrayGetSize(pReader->aDelData); ++iDelData) { - size += tPutDelData(NULL, taosArrayGet(pReader->aDelData, iDelData)); - } - - uint8_t* pData = (uint8_t*)taosMemoryMalloc(sizeof(SSnapDataHdr) + size); - if (pData == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - - SSnapDataHdr* pHdr = (SSnapDataHdr*)pData; - pHdr->type = SNAP_DATA_DEL; - pHdr->size = size; - - TABLEID* pId = (TABLEID*)(pData + sizeof(SSnapDataHdr)); - *pId = pReader->tbid; - - size = sizeof(SSnapDataHdr) + sizeof(TABLEID); - for (int32_t iDelData = 0; iDelData < taosArrayGetSize(pReader->aDelData); ++iDelData) { - size += tPutDelData(pData + size, taosArrayGet(pReader->aDelData, iDelData)); - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pReader->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - *ppData = pData; - return code; -} - -static void tsdbSnapReadGetTombData(STsdbSnapReader* pReader, SDelInfo** ppDelInfo) { - if (pReader->pTIter == NULL || (pReader->pTIter->delInfo.suid == 0 && pReader->pTIter->delInfo.uid == 0)) { - *ppDelInfo = NULL; - } else { - *ppDelInfo = &pReader->pTIter->delInfo; - } -} - -static int32_t tsdbSnapReadNextTombData(STsdbSnapReader* pReader, SDelInfo** ppDelInfo) { - int32_t code = 0; - int32_t lino = 0; - - code = tsdbDataIterNext2( - pReader->pTIter, &(STsdbFilterInfo){.flag = TSDB_FILTER_FLAG_BY_VERSION | TSDB_FILTER_FLAG_IGNORE_DROPPED_TABLE, - .sver = pReader->sver, - .ever = pReader->ever}); - TSDB_CHECK_CODE(code, lino, _exit); - - if (ppDelInfo) { - tsdbSnapReadGetTombData(pReader, ppDelInfo); - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pReader->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - return code; -} - -static int32_t tsdbSnapReadTombData(STsdbSnapReader* pReader, uint8_t** ppData) { - int32_t code = 0; - int32_t lino = 0; - - STsdb* pTsdb = pReader->pTsdb; - - // open tombstone data iter if need - if (pReader->pDelFReader == NULL) { - if (pReader->fs.pDelFile == NULL) goto _exit; - - // open - code = tsdbDelFReaderOpen(&pReader->pDelFReader, pReader->fs.pDelFile, pTsdb); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbOpenTombFileDataIter(pReader->pDelFReader, &pReader->pTIter); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pReader->pTIter) { - code = tsdbSnapReadNextTombData(pReader, NULL); - TSDB_CHECK_CODE(code, lino, _exit); } - } - // loop to get tombstone data - SDelInfo* pDelInfo; - tsdbSnapReadGetTombData(pReader, &pDelInfo); + if (reader->blockData->suid == 0 && reader->blockData->uid == 0) { + code = tsdbUpdateSkmTb(reader->tsdb, (TABLEID*)row, reader->skmTb); + TSDB_CHECK_CODE(code, lino, _exit); - if (pDelInfo == NULL) goto _exit; - - pReader->tbid = *(TABLEID*)pDelInfo; - - if (pReader->aDelData) { - taosArrayClear(pReader->aDelData); - } else if ((pReader->aDelData = taosArrayInit(16, sizeof(SDelData))) == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - - while (pDelInfo && pDelInfo->suid == pReader->tbid.suid && pDelInfo->uid == pReader->tbid.uid) { - if (taosArrayPush(pReader->aDelData, &pDelInfo->delData) == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; + TABLEID tbid1 = { + .suid = row->suid, + .uid = row->suid ? 0 : row->uid, + }; + code = tBlockDataInit(reader->blockData, &tbid1, reader->skmTb->pTSchema, NULL, 0); TSDB_CHECK_CODE(code, lino, _exit); } - code = tsdbSnapReadNextTombData(pReader, &pDelInfo); + if (!TABLE_SAME_SCHEMA(reader->blockData->suid, reader->blockData->uid, row->suid, row->uid)) { + break; + } + + code = tBlockDataAppendRow(reader->blockData, &row->row, NULL, row->uid); TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbIterMergerNext(reader->dataIterMerger); + TSDB_CHECK_CODE(code, lino, _exit); + + if (reader->blockData->nRow >= 81920) { + break; + } } - // encode tombstone data - if (taosArrayGetSize(pReader->aDelData) > 0) { - code = tsdbSnapCmprTombData(pReader, ppData); + if (reader->blockData->nRow > 0) { + ASSERT(reader->blockData->suid || reader->blockData->uid); + code = tsdbSnapCmprData(reader, data); TSDB_CHECK_CODE(code, lino, _exit); } _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbDebug("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), code, lino); } return code; } -int32_t tsdbSnapReaderOpen(STsdb* pTsdb, int64_t sver, int64_t ever, int8_t type, STsdbSnapReader** ppReader) { +static int32_t tsdbSnapCmprTombData(STsdbSnapReader* reader, uint8_t** data) { int32_t code = 0; int32_t lino = 0; - // alloc - STsdbSnapReader* pReader = (STsdbSnapReader*)taosMemoryCalloc(1, sizeof(*pReader)); - if (pReader == NULL) { + int64_t size = sizeof(SSnapDataHdr); + for (int32_t i = 0; i < ARRAY_SIZE(reader->tombBlock->dataArr); i++) { + size += TARRAY2_DATA_LEN(reader->tombBlock->dataArr + i); + } + + data[0] = taosMemoryMalloc(size); + if (data[0] == NULL) { code = TSDB_CODE_OUT_OF_MEMORY; TSDB_CHECK_CODE(code, lino, _exit); } - pReader->pTsdb = pTsdb; - pReader->sver = sver; - pReader->ever = ever; - pReader->type = type; - taosThreadRwlockRdlock(&pTsdb->rwLock); - code = tsdbFSRef(pTsdb, &pReader->fs); + SSnapDataHdr* hdr = (SSnapDataHdr*)data[0]; + hdr->type = SNAP_DATA_DEL; + hdr->size = size; + + uint8_t* tdata = hdr->data; + for (int32_t i = 0; i < ARRAY_SIZE(reader->tombBlock->dataArr); i++) { + memcpy(tdata, TARRAY2_DATA(reader->tombBlock->dataArr + i), TARRAY2_DATA_LEN(reader->tombBlock->dataArr + i)); + tdata += TARRAY2_DATA_LEN(reader->tombBlock->dataArr + i); + } + +_exit: if (code) { - taosThreadRwlockUnlock(&pTsdb->rwLock); + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), code, lino); + } + return code; +} + +static int32_t tsdbSnapReadTombData(STsdbSnapReader* reader, uint8_t** data) { + int32_t code = 0; + int32_t lino = 0; + SMetaInfo info; + + tTombBlockClear(reader->tombBlock); + + TABLEID tbid[1] = {0}; + for (STombRecord* record; (record = tsdbIterMergerGetTombRecord(reader->tombIterMerger)) != NULL;) { + if (record->uid != tbid->uid) { + tbid->suid = record->suid; + tbid->uid = record->uid; + if (metaGetInfo(reader->tsdb->pVnode->pMeta, tbid->uid, &info, NULL) != 0) { + code = tsdbIterMergerSkipTableData(reader->tombIterMerger, tbid); + TSDB_CHECK_CODE(code, lino, _exit); + continue; + } + } + + code = tTombBlockPut(reader->tombBlock, record); + TSDB_CHECK_CODE(code, lino, _exit); + + if (TOMB_BLOCK_SIZE(reader->tombBlock) >= 81920) { + break; + } + } + + if (TOMB_BLOCK_SIZE(reader->tombBlock) > 0) { + code = tsdbSnapCmprTombData(reader, data); TSDB_CHECK_CODE(code, lino, _exit); } - taosThreadRwlockUnlock(&pTsdb->rwLock); - // init - pReader->fid = INT32_MIN; +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), code, lino); + } + return code; +} - code = tBlockDataCreate(&pReader->bData); +int32_t tsdbSnapReaderOpen(STsdb* tsdb, int64_t sver, int64_t ever, int8_t type, STsdbSnapReader** reader) { + int32_t code = 0; + int32_t lino = 0; + + reader[0] = (STsdbSnapReader*)taosMemoryCalloc(1, sizeof(*reader[0])); + if (reader[0] == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + reader[0]->tsdb = tsdb; + reader[0]->sver = sver; + reader[0]->ever = ever; + reader[0]->type = type; + + code = tsdbFSCreateRefSnapshot(tsdb->pFS, &reader[0]->fsetArr); TSDB_CHECK_CODE(code, lino, _exit); _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s, sver:%" PRId64 " ever:%" PRId64 " type:%d", TD_VID(pTsdb->pVnode), + tsdbError("vgId:%d %s failed at line %d since %s, sver:%" PRId64 " ever:%" PRId64 " type:%d", TD_VID(tsdb->pVnode), __func__, lino, tstrerror(code), sver, ever, type); - if (pReader) { - tBlockDataDestroy(&pReader->bData); - tsdbFSUnref(pTsdb, &pReader->fs); - taosMemoryFree(pReader); - pReader = NULL; - } + tsdbFSDestroyRefSnapshot(&reader[0]->fsetArr); + taosMemoryFree(reader[0]); + reader[0] = NULL; } else { - tsdbInfo("vgId:%d %s done, sver:%" PRId64 " ever:%" PRId64 " type:%d", TD_VID(pTsdb->pVnode), __func__, sver, ever, + tsdbInfo("vgId:%d %s done, sver:%" PRId64 " ever:%" PRId64 " type:%d", TD_VID(tsdb->pVnode), __func__, sver, ever, type); } - *ppReader = pReader; return code; } -int32_t tsdbSnapReaderClose(STsdbSnapReader** ppReader) { +int32_t tsdbSnapReaderClose(STsdbSnapReader** reader) { + if (reader[0] == NULL) return 0; + int32_t code = 0; int32_t lino = 0; - STsdbSnapReader* pReader = *ppReader; - STsdb* pTsdb = pReader->pTsdb; + STsdb* tsdb = reader[0]->tsdb; - // tombstone - if (pReader->pTIter) { - tsdbCloseDataIter2(pReader->pTIter); - pReader->pTIter = NULL; - } - if (pReader->pDelFReader) { - tsdbDelFReaderClose(&pReader->pDelFReader); - } - taosArrayDestroy(pReader->aDelData); + tTombBlockDestroy(reader[0]->tombBlock); + tBlockDataDestroy(reader[0]->blockData); - // timeseries - while (pReader->iterList) { - STsdbDataIter2* pIter = pReader->iterList; - pReader->iterList = pIter->next; - tsdbCloseDataIter2(pIter); - } - if (pReader->pDataFReader) { - tsdbDataFReaderClose(&pReader->pDataFReader); - } - tBlockDataDestroy(&pReader->bData); + tsdbIterMergerClose(&reader[0]->dataIterMerger); + tsdbIterMergerClose(&reader[0]->tombIterMerger); + TARRAY2_DESTROY(reader[0]->dataIterArr, tsdbIterClose); + TARRAY2_DESTROY(reader[0]->tombIterArr, tsdbIterClose); + TARRAY2_DESTROY(reader[0]->sttReaderArr, tsdbSttFileReaderClose); + tsdbDataFileReaderClose(&reader[0]->dataReader); - // other - tDestroyTSchema(pReader->skmTable.pTSchema); - tsdbFSUnref(pReader->pTsdb, &pReader->fs); - for (int32_t iBuf = 0; iBuf < sizeof(pReader->aBuf) / sizeof(pReader->aBuf[0]); iBuf++) { - tFree(pReader->aBuf[iBuf]); + tsdbFSDestroyRefSnapshot(&reader[0]->fsetArr); + tDestroyTSchema(reader[0]->skmTb->pTSchema); + + for (int32_t i = 0; i < ARRAY_SIZE(reader[0]->aBuf); ++i) { + tFree(reader[0]->aBuf[i]); } - taosMemoryFree(pReader); + + taosMemoryFree(reader[0]); + reader[0] = NULL; _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); } else { - tsdbDebug("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); + tsdbDebug("vgId:%d %s done", TD_VID(tsdb->pVnode), __func__); } - *ppReader = NULL; return code; } -int32_t tsdbSnapRead(STsdbSnapReader* pReader, uint8_t** ppData) { +int32_t tsdbSnapRead(STsdbSnapReader* reader, uint8_t** data) { int32_t code = 0; int32_t lino = 0; - *ppData = NULL; + data[0] = NULL; - // read data file - if (!pReader->dataDone) { - code = tsdbSnapReadTimeSeriesData(pReader, ppData); - TSDB_CHECK_CODE(code, lino, _exit); - if (*ppData) { - goto _exit; - } else { - pReader->dataDone = 1; - } - } + for (;;) { + if (reader->ctx->fset == NULL) { + code = tsdbSnapReadFileSetBegin(reader); + TSDB_CHECK_CODE(code, lino, _exit); - // read del file - if (!pReader->delDone) { - code = tsdbSnapReadTombData(pReader, ppData); - TSDB_CHECK_CODE(code, lino, _exit); - if (*ppData) { - goto _exit; - } else { - pReader->delDone = 1; + if (reader->ctx->fset == NULL) { + break; + } } + + if (!reader->ctx->isDataDone) { + code = tsdbSnapReadTimeSeriesData(reader, data); + TSDB_CHECK_CODE(code, lino, _exit); + if (data[0]) { + goto _exit; + } else { + reader->ctx->isDataDone = true; + } + } + + if (!reader->ctx->isTombDone) { + code = tsdbSnapReadTombData(reader, data); + TSDB_CHECK_CODE(code, lino, _exit); + if (data[0]) { + goto _exit; + } else { + reader->ctx->isTombDone = true; + } + } + + code = tsdbSnapReadFileSetEnd(reader); + TSDB_CHECK_CODE(code, lino, _exit); } _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pReader->pTsdb->pVnode), __func__, lino, tstrerror(code)); + TSDB_ERROR_LOG(TD_VID(reader->tsdb->pVnode), code, lino); } else { - tsdbDebug("vgId:%d %s done", TD_VID(pReader->pTsdb->pVnode), __func__); + tsdbDebug("vgId:%d %s done", TD_VID(reader->tsdb->pVnode), __func__); } return code; } // STsdbSnapWriter ======================================== struct STsdbSnapWriter { - STsdb* pTsdb; + STsdb* tsdb; int64_t sver; int64_t ever; int32_t minutes; @@ -569,973 +535,595 @@ struct STsdbSnapWriter { int32_t maxRow; int8_t cmprAlg; int64_t commitID; + int32_t szPage; + int64_t compactVersion; + int64_t now; uint8_t* aBuf[5]; - STsdbFS fs; - TABLEID tbid; + TFileSetArray* fsetArr; + TFileOpArray fopArr[1]; - // time-series data - SBlockData inData; + struct { + bool fsetWriteBegin; + int32_t fid; + STFileSet* fset; + SDiskID did; + bool hasData; + bool hasTomb; - int32_t fid; - SSkmInfo skmTable; + // reader + SDataFileReader* dataReader; + TSttFileReaderArray sttReaderArr[1]; - /* reader */ - SDataFReader* pDataFReader; - STsdbDataIter2* iterList; - STsdbDataIter2* pDIter; - STsdbDataIter2* pSIter; - SRBTree rbt; // SRBTree + // iter/merger + TTsdbIterArray dataIterArr[1]; + SIterMerger* dataIterMerger; + TTsdbIterArray tombIterArr[1]; + SIterMerger* tombIterMerger; - /* writer */ - SDataFWriter* pDataFWriter; - SArray* aBlockIdx; - SMapData mDataBlk; // SMapData - SArray* aSttBlk; // SArray - SBlockData bData; - SBlockData sData; - - // tombstone data - /* reader */ - SDelFReader* pDelFReader; - STsdbDataIter2* pTIter; - - /* writer */ - SDelFWriter* pDelFWriter; - SArray* aDelIdx; - SArray* aDelData; + // writer + SFSetWriter* fsetWriter; + } ctx[1]; }; -// SNAP_DATA_TSDB -static int32_t tsdbSnapWriteTableDataStart(STsdbSnapWriter* pWriter, TABLEID* pId) { +// APIs +static int32_t tsdbSnapWriteTimeSeriesRow(STsdbSnapWriter* writer, SRowInfo* row) { int32_t code = 0; int32_t lino = 0; - if (pId) { - pWriter->tbid = *pId; - } else { - pWriter->tbid = (TABLEID){INT64_MAX, INT64_MAX}; - } - - if (pWriter->pDIter) { - STsdbDataIter2* pIter = pWriter->pDIter; - - // assert last table data end - ASSERT(pIter->dIter.iRow >= pIter->dIter.bData.nRow); - ASSERT(pIter->dIter.iDataBlk >= pIter->dIter.mDataBlk.nItem); - - for (;;) { - if (pIter->dIter.iBlockIdx >= taosArrayGetSize(pIter->dIter.aBlockIdx)) { - pWriter->pDIter = NULL; - break; - } - - SBlockIdx* pBlockIdx = (SBlockIdx*)taosArrayGet(pIter->dIter.aBlockIdx, pIter->dIter.iBlockIdx); - - int32_t c = tTABLEIDCmprFn(pBlockIdx, &pWriter->tbid); - if (c < 0) { - code = tsdbReadDataBlk(pIter->dIter.pReader, pBlockIdx, &pIter->dIter.mDataBlk); - TSDB_CHECK_CODE(code, lino, _exit); - - SBlockIdx* pNewBlockIdx = taosArrayReserve(pWriter->aBlockIdx, 1); - if (pNewBlockIdx == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - - pNewBlockIdx->suid = pBlockIdx->suid; - pNewBlockIdx->uid = pBlockIdx->uid; - - code = tsdbWriteDataBlk(pWriter->pDataFWriter, &pIter->dIter.mDataBlk, pNewBlockIdx); - TSDB_CHECK_CODE(code, lino, _exit); - - pIter->dIter.iBlockIdx++; - } else if (c == 0) { - code = tsdbReadDataBlk(pIter->dIter.pReader, pBlockIdx, &pIter->dIter.mDataBlk); - TSDB_CHECK_CODE(code, lino, _exit); - - pIter->dIter.iDataBlk = 0; - pIter->dIter.iBlockIdx++; - - break; - } else { - pIter->dIter.iDataBlk = pIter->dIter.mDataBlk.nItem; - break; - } + while (writer->ctx->hasData) { + SRowInfo* row1 = tsdbIterMergerGetData(writer->ctx->dataIterMerger); + if (row1 == NULL) { + writer->ctx->hasData = false; + break; } - } - if (pId) { - code = tsdbUpdateTableSchema(pWriter->pTsdb->pVnode->pMeta, pId->suid, pId->uid, &pWriter->skmTable); - TSDB_CHECK_CODE(code, lino, _exit); - - tMapDataReset(&pWriter->mDataBlk); - - code = tBlockDataInit(&pWriter->bData, pId, pWriter->skmTable.pTSchema, NULL, 0); - TSDB_CHECK_CODE(code, lino, _exit); - } - - if (!TABLE_SAME_SCHEMA(pWriter->tbid.suid, pWriter->tbid.uid, pWriter->sData.suid, pWriter->sData.uid)) { - if ((pWriter->sData.nRow > 0)) { - code = tsdbWriteSttBlock(pWriter->pDataFWriter, &pWriter->sData, pWriter->aSttBlk, pWriter->cmprAlg); + int32_t c = tRowInfoCmprFn(row1, row); + if (c <= 0) { + code = tsdbFSetWriteRow(writer->ctx->fsetWriter, row1); TSDB_CHECK_CODE(code, lino, _exit); - } - if (pId) { - TABLEID id = {.suid = pWriter->tbid.suid, .uid = pWriter->tbid.suid ? 0 : pWriter->tbid.uid}; - code = tBlockDataInit(&pWriter->sData, &id, pWriter->skmTable.pTSchema, NULL, 0); + code = tsdbIterMergerNext(writer->ctx->dataIterMerger); TSDB_CHECK_CODE(code, lino, _exit); - } - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbTrace("vgId:%d %s done, suid:%" PRId64 " uid:%" PRId64, TD_VID(pWriter->pTsdb->pVnode), __func__, - pWriter->tbid.suid, pWriter->tbid.uid); - } - return code; -} - -static int32_t tsdbSnapWriteTableRowImpl(STsdbSnapWriter* pWriter, TSDBROW* pRow) { - int32_t code = 0; - int32_t lino = 0; - - code = tBlockDataAppendRow(&pWriter->bData, pRow, pWriter->skmTable.pTSchema, pWriter->tbid.uid); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pWriter->bData.nRow >= pWriter->maxRow) { - code = tsdbWriteDataBlock(pWriter->pDataFWriter, &pWriter->bData, &pWriter->mDataBlk, pWriter->cmprAlg); - TSDB_CHECK_CODE(code, lino, _exit); - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - return code; -} - -static int32_t tsdbSnapWriteTableRow(STsdbSnapWriter* pWriter, TSDBROW* pRow) { - int32_t code = 0; - int32_t lino = 0; - - TSDBKEY inKey = pRow ? TSDBROW_KEY(pRow) : TSDBKEY_MAX; - - if (pWriter->pDIter == NULL || (pWriter->pDIter->dIter.iRow >= pWriter->pDIter->dIter.bData.nRow && - pWriter->pDIter->dIter.iDataBlk >= pWriter->pDIter->dIter.mDataBlk.nItem)) { - goto _write_row; - } else { - for (;;) { - while (pWriter->pDIter->dIter.iRow < pWriter->pDIter->dIter.bData.nRow) { - TSDBROW row = tsdbRowFromBlockData(&pWriter->pDIter->dIter.bData, pWriter->pDIter->dIter.iRow); - - int32_t c = tsdbKeyCmprFn(&inKey, &TSDBROW_KEY(&row)); - if (c < 0) { - goto _write_row; - } else if (c > 0) { - code = tsdbSnapWriteTableRowImpl(pWriter, &row); - TSDB_CHECK_CODE(code, lino, _exit); - - pWriter->pDIter->dIter.iRow++; - } else { - ASSERT(0); - } - } - - for (;;) { - if (pWriter->pDIter->dIter.iDataBlk >= pWriter->pDIter->dIter.mDataBlk.nItem) goto _write_row; - - // FIXME: Here can be slow, use array instead - SDataBlk dataBlk; - tMapDataGetItemByIdx(&pWriter->pDIter->dIter.mDataBlk, pWriter->pDIter->dIter.iDataBlk, &dataBlk, tGetDataBlk); - - int32_t c = tDataBlkCmprFn(&dataBlk, &(SDataBlk){.minKey = inKey, .maxKey = inKey}); - if (c > 0) { - goto _write_row; - } else if (c < 0) { - if (pWriter->bData.nRow > 0) { - code = tsdbWriteDataBlock(pWriter->pDataFWriter, &pWriter->bData, &pWriter->mDataBlk, pWriter->cmprAlg); - TSDB_CHECK_CODE(code, lino, _exit); - } - - tMapDataPutItem(&pWriter->mDataBlk, &dataBlk, tPutDataBlk); - pWriter->pDIter->dIter.iDataBlk++; - } else { - code = tsdbReadDataBlockEx(pWriter->pDataFReader, &dataBlk, &pWriter->pDIter->dIter.bData); - TSDB_CHECK_CODE(code, lino, _exit); - - pWriter->pDIter->dIter.iRow = 0; - pWriter->pDIter->dIter.iDataBlk++; - break; - } - } - } - } - -_write_row: - if (pRow) { - code = tsdbSnapWriteTableRowImpl(pWriter, pRow); - TSDB_CHECK_CODE(code, lino, _exit); - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - return code; -} - -static int32_t tsdbSnapWriteTableDataEnd(STsdbSnapWriter* pWriter) { - int32_t code = 0; - int32_t lino = 0; - - // write a NULL row to end current table data write - code = tsdbSnapWriteTableRow(pWriter, NULL); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pWriter->bData.nRow > 0) { - if (pWriter->bData.nRow < pWriter->minRow) { - ASSERT(TABLE_SAME_SCHEMA(pWriter->sData.suid, pWriter->sData.uid, pWriter->tbid.suid, pWriter->tbid.uid)); - for (int32_t iRow = 0; iRow < pWriter->bData.nRow; iRow++) { - code = - tBlockDataAppendRow(&pWriter->sData, &tsdbRowFromBlockData(&pWriter->bData, iRow), NULL, pWriter->tbid.uid); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pWriter->sData.nRow >= pWriter->maxRow) { - code = tsdbWriteSttBlock(pWriter->pDataFWriter, &pWriter->sData, pWriter->aSttBlk, pWriter->cmprAlg); - TSDB_CHECK_CODE(code, lino, _exit); - } - } - - tBlockDataClear(&pWriter->bData); } else { - code = tsdbWriteDataBlock(pWriter->pDataFWriter, &pWriter->bData, &pWriter->mDataBlk, pWriter->cmprAlg); - TSDB_CHECK_CODE(code, lino, _exit); + break; } } - if (pWriter->mDataBlk.nItem) { - SBlockIdx* pBlockIdx = taosArrayReserve(pWriter->aBlockIdx, 1); - if (pBlockIdx == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - - pBlockIdx->suid = pWriter->tbid.suid; - pBlockIdx->uid = pWriter->tbid.uid; - - code = tsdbWriteDataBlk(pWriter->pDataFWriter, &pWriter->mDataBlk, pBlockIdx); - TSDB_CHECK_CODE(code, lino, _exit); - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - return code; -} - -static int32_t tsdbSnapWriteFileDataStart(STsdbSnapWriter* pWriter, int32_t fid) { - int32_t code = 0; - int32_t lino = 0; - - ASSERT(pWriter->pDataFWriter == NULL && pWriter->fid < fid); - - STsdb* pTsdb = pWriter->pTsdb; - - pWriter->fid = fid; - pWriter->tbid = (TABLEID){0}; - SDFileSet* pSet = taosArraySearch(pWriter->fs.aDFileSet, &(SDFileSet){.fid = fid}, tDFileSetCmprFn, TD_EQ); - - // open reader - pWriter->pDataFReader = NULL; - pWriter->iterList = NULL; - pWriter->pDIter = NULL; - pWriter->pSIter = NULL; - tRBTreeCreate(&pWriter->rbt, tsdbDataIterCmprFn); - if (pSet) { - code = tsdbDataFReaderOpen(&pWriter->pDataFReader, pTsdb, pSet); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbOpenDataFileDataIter(pWriter->pDataFReader, &pWriter->pDIter); - TSDB_CHECK_CODE(code, lino, _exit); - if (pWriter->pDIter) { - pWriter->pDIter->next = pWriter->iterList; - pWriter->iterList = pWriter->pDIter; - } - - for (int32_t iStt = 0; iStt < pSet->nSttF; iStt++) { - code = tsdbOpenSttFileDataIter(pWriter->pDataFReader, iStt, &pWriter->pSIter); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pWriter->pSIter) { - code = tsdbDataIterNext2(pWriter->pSIter, NULL); - TSDB_CHECK_CODE(code, lino, _exit); - - // add to tree - tRBTreePut(&pWriter->rbt, &pWriter->pSIter->rbtn); - - // add to list - pWriter->pSIter->next = pWriter->iterList; - pWriter->iterList = pWriter->pSIter; - } - } - - pWriter->pSIter = NULL; - } - - // open writer - SDiskID diskId; - if (pSet) { - diskId = pSet->diskId; - } else { - code = tfsAllocDisk(pTsdb->pVnode->pTfs, 0 /*TODO*/, &diskId); - TSDB_CHECK_CODE(code, lino, _exit); - code = tfsMkdirRecurAt(pTsdb->pVnode->pTfs, pTsdb->path, diskId); - TSDB_CHECK_CODE(code, lino, _exit); - } - SDFileSet wSet = {.diskId = diskId, - .fid = fid, - .pHeadF = &(SHeadFile){.commitID = pWriter->commitID}, - .pDataF = (pSet) ? pSet->pDataF : &(SDataFile){.commitID = pWriter->commitID}, - .pSmaF = (pSet) ? pSet->pSmaF : &(SSmaFile){.commitID = pWriter->commitID}, - .nSttF = 1, - .aSttF = {&(SSttFile){.commitID = pWriter->commitID}}}; - code = tsdbDataFWriterOpen(&pWriter->pDataFWriter, pTsdb, &wSet); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pWriter->aBlockIdx) { - taosArrayClear(pWriter->aBlockIdx); - } else if ((pWriter->aBlockIdx = taosArrayInit(0, sizeof(SBlockIdx))) == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - - tMapDataReset(&pWriter->mDataBlk); - - if (pWriter->aSttBlk) { - taosArrayClear(pWriter->aSttBlk); - } else if ((pWriter->aSttBlk = taosArrayInit(0, sizeof(SSttBlk))) == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - - tBlockDataReset(&pWriter->bData); - tBlockDataReset(&pWriter->sData); - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s, fid:%d", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code), - fid); - } else { - tsdbDebug("vgId:%d %s done, fid:%d", TD_VID(pTsdb->pVnode), __func__, fid); - } - return code; -} - -static int32_t tsdbSnapWriteTableData(STsdbSnapWriter* pWriter, SRowInfo* pRowInfo) { - int32_t code = 0; - int32_t lino = 0; - - // switch to new table if need - if (pRowInfo == NULL || pRowInfo->uid != pWriter->tbid.uid) { - if (pWriter->tbid.uid) { - code = tsdbSnapWriteTableDataEnd(pWriter); - TSDB_CHECK_CODE(code, lino, _exit); - } - - code = tsdbSnapWriteTableDataStart(pWriter, (TABLEID*)pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); - } - - if (pRowInfo == NULL) goto _exit; - - code = tsdbSnapWriteTableRow(pWriter, &pRowInfo->row); - TSDB_CHECK_CODE(code, lino, _exit); - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - return code; -} - -static int32_t tsdbSnapWriteNextRow(STsdbSnapWriter* pWriter, SRowInfo** ppRowInfo) { - int32_t code = 0; - int32_t lino = 0; - - if (pWriter->pSIter) { - code = tsdbDataIterNext2(pWriter->pSIter, NULL); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pWriter->pSIter->rowInfo.suid == 0 && pWriter->pSIter->rowInfo.uid == 0) { - pWriter->pSIter = NULL; - } else { - SRBTreeNode* pNode = tRBTreeMin(&pWriter->rbt); - if (pNode) { - int32_t c = tsdbDataIterCmprFn(&pWriter->pSIter->rbtn, pNode); - if (c > 0) { - tRBTreePut(&pWriter->rbt, &pWriter->pSIter->rbtn); - pWriter->pSIter = NULL; - } else if (c == 0) { - ASSERT(0); - } - } - } - } - - if (pWriter->pSIter == NULL) { - SRBTreeNode* pNode = tRBTreeMin(&pWriter->rbt); - if (pNode) { - tRBTreeDrop(&pWriter->rbt, pNode); - pWriter->pSIter = TSDB_RBTN_TO_DATA_ITER(pNode); - } - } - - if (ppRowInfo) { - if (pWriter->pSIter) { - *ppRowInfo = &pWriter->pSIter->rowInfo; - } else { - *ppRowInfo = NULL; - } - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - return code; -} - -static int32_t tsdbSnapWriteGetRow(STsdbSnapWriter* pWriter, SRowInfo** ppRowInfo) { - int32_t code = 0; - int32_t lino = 0; - - if (pWriter->pSIter) { - *ppRowInfo = &pWriter->pSIter->rowInfo; + if (row->suid == INT64_MAX) { + ASSERT(writer->ctx->hasData == false); goto _exit; } - code = tsdbSnapWriteNextRow(pWriter, ppRowInfo); + code = tsdbFSetWriteRow(writer->ctx->fsetWriter, row); TSDB_CHECK_CODE(code, lino, _exit); _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); } return code; } -static int32_t tsdbSnapWriteFileDataEnd(STsdbSnapWriter* pWriter) { +static int32_t tsdbSnapWriteFileSetOpenReader(STsdbSnapWriter* writer) { int32_t code = 0; int32_t lino = 0; - ASSERT(pWriter->pDataFWriter); + ASSERT(writer->ctx->dataReader == NULL); + ASSERT(TARRAY2_SIZE(writer->ctx->sttReaderArr) == 0); - // consume remain data and end with a NULL table row - SRowInfo* pRowInfo; - code = tsdbSnapWriteGetRow(pWriter, &pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); - for (;;) { - code = tsdbSnapWriteTableData(pWriter, pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); + if (writer->ctx->fset) { + // open data reader + SDataFileReaderConfig dataFileReaderConfig = { + .tsdb = writer->tsdb, + .bufArr = writer->aBuf, + .szPage = writer->szPage, + }; - if (pRowInfo == NULL) break; + for (int32_t ftype = 0; ftype < TSDB_FTYPE_MAX; ++ftype) { + if (writer->ctx->fset->farr[ftype] == NULL) { + continue; + } - code = tsdbSnapWriteNextRow(pWriter, &pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); - } - - // do file-level updates - code = tsdbWriteSttBlk(pWriter->pDataFWriter, pWriter->aSttBlk); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbWriteBlockIdx(pWriter->pDataFWriter, pWriter->aBlockIdx); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbUpdateDFileSetHeader(pWriter->pDataFWriter); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbFSUpsertFSet(&pWriter->fs, &pWriter->pDataFWriter->wSet); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbDataFWriterClose(&pWriter->pDataFWriter, 1); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pWriter->pDataFReader) { - code = tsdbDataFReaderClose(&pWriter->pDataFReader); - TSDB_CHECK_CODE(code, lino, _exit); - } - - // clear sources - while (pWriter->iterList) { - STsdbDataIter2* pIter = pWriter->iterList; - pWriter->iterList = pIter->next; - tsdbCloseDataIter2(pIter); - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, tstrerror(code)); - } else { - tsdbDebug("vgId:%d %s is done", TD_VID(pWriter->pTsdb->pVnode), __func__); - } - return code; -} - -static int32_t tsdbSnapWriteTimeSeriesData(STsdbSnapWriter* pWriter, SSnapDataHdr* pHdr) { - int32_t code = 0; - int32_t lino = 0; - - code = tDecmprBlockData(pHdr->data, pHdr->size, &pWriter->inData, pWriter->aBuf); - TSDB_CHECK_CODE(code, lino, _exit); - - ASSERT(pWriter->inData.nRow > 0); - - // switch to new data file if need - int32_t fid = tsdbKeyFid(pWriter->inData.aTSKEY[0], pWriter->minutes, pWriter->precision); - if (pWriter->fid != fid) { - if (pWriter->pDataFWriter) { - code = tsdbSnapWriteFileDataEnd(pWriter); - TSDB_CHECK_CODE(code, lino, _exit); + dataFileReaderConfig.files[ftype].exist = true; + dataFileReaderConfig.files[ftype].file = writer->ctx->fset->farr[ftype]->f[0]; } - code = tsdbSnapWriteFileDataStart(pWriter, fid); + code = tsdbDataFileReaderOpen(NULL, &dataFileReaderConfig, &writer->ctx->dataReader); TSDB_CHECK_CODE(code, lino, _exit); - } - // loop write each row - SRowInfo* pRowInfo; - code = tsdbSnapWriteGetRow(pWriter, &pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); - for (int32_t iRow = 0; iRow < pWriter->inData.nRow; ++iRow) { - SRowInfo rInfo = {.suid = pWriter->inData.suid, - .uid = pWriter->inData.uid ? pWriter->inData.uid : pWriter->inData.aUid[iRow], - .row = tsdbRowFromBlockData(&pWriter->inData, iRow)}; + // open stt reader array + SSttLvl* lvl; + TARRAY2_FOREACH(writer->ctx->fset->lvlArr, lvl) { + STFileObj* fobj; + TARRAY2_FOREACH(lvl->fobjArr, fobj) { + SSttFileReader* reader; + SSttFileReaderConfig sttFileReaderConfig = { + .tsdb = writer->tsdb, + .szPage = writer->szPage, + .bufArr = writer->aBuf, + .file = fobj->f[0], + }; - for (;;) { - if (pRowInfo == NULL) { - code = tsdbSnapWriteTableData(pWriter, &rInfo); + code = tsdbSttFileReaderOpen(fobj->fname, &sttFileReaderConfig, &reader); TSDB_CHECK_CODE(code, lino, _exit); - break; - } else { - int32_t c = tRowInfoCmprFn(&rInfo, pRowInfo); - if (c < 0) { - code = tsdbSnapWriteTableData(pWriter, &rInfo); - TSDB_CHECK_CODE(code, lino, _exit); - break; - } else if (c > 0) { - code = tsdbSnapWriteTableData(pWriter, pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); - code = tsdbSnapWriteNextRow(pWriter, &pRowInfo); - TSDB_CHECK_CODE(code, lino, _exit); - } else { - ASSERT(0); - } + code = TARRAY2_APPEND(writer->ctx->sttReaderArr, reader); + TSDB_CHECK_CODE(code, lino, _exit); } } } _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbDebug("vgId:%d %s done, suid:%" PRId64 " uid:%" PRId64 " nRow:%d", TD_VID(pWriter->pTsdb->pVnode), __func__, - pWriter->inData.suid, pWriter->inData.uid, pWriter->inData.nRow); + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); } return code; } -// SNAP_DATA_DEL -static int32_t tsdbSnapWriteDelTableDataStart(STsdbSnapWriter* pWriter, TABLEID* pId) { - int32_t code = 0; - int32_t lino = 0; - - if (pId) { - pWriter->tbid = *pId; - } else { - pWriter->tbid = (TABLEID){.suid = INT64_MAX, .uid = INT64_MAX}; - } - - taosArrayClear(pWriter->aDelData); - - if (pWriter->pTIter) { - while (pWriter->pTIter->tIter.iDelIdx < taosArrayGetSize(pWriter->pTIter->tIter.aDelIdx)) { - SDelIdx* pDelIdx = taosArrayGet(pWriter->pTIter->tIter.aDelIdx, pWriter->pTIter->tIter.iDelIdx); - - int32_t c = tTABLEIDCmprFn(pDelIdx, &pWriter->tbid); - if (c < 0) { - code = tsdbReadDelDatav1(pWriter->pDelFReader, pDelIdx, pWriter->pTIter->tIter.aDelData, INT64_MAX); - TSDB_CHECK_CODE(code, lino, _exit); - - SDelIdx* pDelIdxNew = taosArrayReserve(pWriter->aDelIdx, 1); - if (pDelIdxNew == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - - pDelIdxNew->suid = pDelIdx->suid; - pDelIdxNew->uid = pDelIdx->uid; - - code = tsdbWriteDelData(pWriter->pDelFWriter, pWriter->pTIter->tIter.aDelData, pDelIdxNew); - TSDB_CHECK_CODE(code, lino, _exit); - - pWriter->pTIter->tIter.iDelIdx++; - } else if (c == 0) { - code = tsdbReadDelDatav1(pWriter->pDelFReader, pDelIdx, pWriter->aDelData, INT64_MAX); - TSDB_CHECK_CODE(code, lino, _exit); - - pWriter->pTIter->tIter.iDelIdx++; - break; - } else { - break; - } - } - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbTrace("vgId:%d %s done, suid:%" PRId64 " uid:%" PRId64, TD_VID(pWriter->pTsdb->pVnode), __func__, - pWriter->tbid.suid, pWriter->tbid.uid); - } - return code; +static int32_t tsdbSnapWriteFileSetCloseReader(STsdbSnapWriter* writer) { + TARRAY2_CLEAR(writer->ctx->sttReaderArr, tsdbSttFileReaderClose); + tsdbDataFileReaderClose(&writer->ctx->dataReader); + return 0; } -static int32_t tsdbSnapWriteDelTableDataEnd(STsdbSnapWriter* pWriter) { +static int32_t tsdbSnapWriteFileSetOpenIter(STsdbSnapWriter* writer) { int32_t code = 0; int32_t lino = 0; - if (taosArrayGetSize(pWriter->aDelData) > 0) { - SDelIdx* pDelIdx = taosArrayReserve(pWriter->aDelIdx, 1); - if (pDelIdx == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } + // data ieter + if (writer->ctx->dataReader) { + STsdbIter* iter; + STsdbIterConfig config = {0}; - pDelIdx->suid = pWriter->tbid.suid; - pDelIdx->uid = pWriter->tbid.uid; + // data + config.type = TSDB_ITER_TYPE_DATA; + config.dataReader = writer->ctx->dataReader; - code = tsdbWriteDelData(pWriter->pDelFWriter, pWriter->aDelData, pDelIdx); + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(writer->ctx->dataIterArr, iter); + TSDB_CHECK_CODE(code, lino, _exit); + + // tome + config.type = TSDB_ITER_TYPE_DATA_TOMB; + config.dataReader = writer->ctx->dataReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(writer->ctx->tombIterArr, iter); TSDB_CHECK_CODE(code, lino, _exit); } -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbTrace("vgId:%d %s done", TD_VID(pWriter->pTsdb->pVnode), __func__); - } - return code; -} + // stt iter + SSttFileReader* sttFileReader; + TARRAY2_FOREACH(writer->ctx->sttReaderArr, sttFileReader) { + STsdbIter* iter; + STsdbIterConfig config = {0}; -static int32_t tsdbSnapWriteDelTableData(STsdbSnapWriter* pWriter, TABLEID* pId, uint8_t* pData, int64_t size) { - int32_t code = 0; - int32_t lino = 0; + // data + config.type = TSDB_ITER_TYPE_STT; + config.sttReader = sttFileReader; - if (pId == NULL || pId->uid != pWriter->tbid.uid) { - if (pWriter->tbid.uid) { - code = tsdbSnapWriteDelTableDataEnd(pWriter); - TSDB_CHECK_CODE(code, lino, _exit); - } + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); - code = tsdbSnapWriteDelTableDataStart(pWriter, pId); + code = TARRAY2_APPEND(writer->ctx->dataIterArr, iter); + TSDB_CHECK_CODE(code, lino, _exit); + + // tomb + config.type = TSDB_ITER_TYPE_STT_TOMB; + config.sttReader = sttFileReader; + + code = tsdbIterOpen(&config, &iter); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(writer->ctx->tombIterArr, iter); TSDB_CHECK_CODE(code, lino, _exit); } - if (pId == NULL) goto _exit; - - int64_t n = 0; - while (n < size) { - SDelData delData; - n += tGetDelData(pData + n, &delData); - - if (taosArrayPush(pWriter->aDelData, &delData) == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - } - ASSERT(n == size); - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); - } - return code; -} - -static int32_t tsdbSnapWriteDelDataStart(STsdbSnapWriter* pWriter) { - int32_t code = 0; - int32_t lino = 0; - - STsdb* pTsdb = pWriter->pTsdb; - SDelFile* pDelFile = pWriter->fs.pDelFile; - - pWriter->tbid = (TABLEID){0}; - - // reader - if (pDelFile) { - code = tsdbDelFReaderOpen(&pWriter->pDelFReader, pDelFile, pTsdb); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbOpenTombFileDataIter(pWriter->pDelFReader, &pWriter->pTIter); - TSDB_CHECK_CODE(code, lino, _exit); - } - - // writer - code = tsdbDelFWriterOpen(&pWriter->pDelFWriter, &(SDelFile){.commitID = pWriter->commitID}, pTsdb); + // open merger + code = tsdbIterMergerOpen(writer->ctx->dataIterArr, &writer->ctx->dataIterMerger, false); TSDB_CHECK_CODE(code, lino, _exit); - if ((pWriter->aDelIdx = taosArrayInit(0, sizeof(SDelIdx))) == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; + code = tsdbIterMergerOpen(writer->ctx->tombIterArr, &writer->ctx->tombIterMerger, true); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbSnapWriteFileSetCloseIter(STsdbSnapWriter* writer) { + tsdbIterMergerClose(&writer->ctx->dataIterMerger); + tsdbIterMergerClose(&writer->ctx->tombIterMerger); + TARRAY2_CLEAR(writer->ctx->dataIterArr, tsdbIterClose); + TARRAY2_CLEAR(writer->ctx->tombIterArr, tsdbIterClose); + return 0; +} + +static int32_t tsdbSnapWriteFileSetOpenWriter(STsdbSnapWriter* writer) { + int32_t code = 0; + int32_t lino = 0; + + SFSetWriterConfig config = { + .tsdb = writer->tsdb, + .toSttOnly = false, + .compactVersion = writer->compactVersion, + .minRow = writer->minRow, + .maxRow = writer->maxRow, + .szPage = writer->szPage, + .cmprAlg = writer->cmprAlg, + .fid = writer->ctx->fid, + .cid = writer->commitID, + .did = writer->ctx->did, + .level = 0, + }; + + code = tsdbFSetWriterOpen(&config, &writer->ctx->fsetWriter); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbSnapWriteFileSetCloseWriter(STsdbSnapWriter* writer) { + return tsdbFSetWriterClose(&writer->ctx->fsetWriter, 0, writer->fopArr); +} + +static int32_t tsdbSnapWriteFileSetBegin(STsdbSnapWriter* writer, int32_t fid) { + int32_t code = 0; + int32_t lino = 0; + + ASSERT(writer->ctx->fsetWriteBegin == false); + + STFileSet* fset = &(STFileSet){.fid = fid}; + + writer->ctx->fid = fid; + STFileSet** fsetPtr = TARRAY2_SEARCH(writer->fsetArr, &fset, tsdbTFileSetCmprFn, TD_EQ); + writer->ctx->fset = (fsetPtr == NULL) ? NULL : *fsetPtr; + + int32_t level = tsdbFidLevel(fid, &writer->tsdb->keepCfg, taosGetTimestampSec()); + if (tfsAllocDisk(writer->tsdb->pVnode->pTfs, level, &writer->ctx->did)) { + code = TSDB_CODE_NO_AVAIL_DISK; TSDB_CHECK_CODE(code, lino, _exit); } - if ((pWriter->aDelData = taosArrayInit(0, sizeof(SDelData))) == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; + tfsMkdirRecurAt(writer->tsdb->pVnode->pTfs, writer->tsdb->path, writer->ctx->did); + + writer->ctx->hasData = true; + writer->ctx->hasTomb = true; + + code = tsdbSnapWriteFileSetOpenReader(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSnapWriteFileSetOpenIter(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSnapWriteFileSetOpenWriter(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + writer->ctx->fsetWriteBegin = true; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbSnapWriteTombRecord(STsdbSnapWriter* writer, const STombRecord* record) { + int32_t code = 0; + int32_t lino = 0; + + while (writer->ctx->hasTomb) { + STombRecord* record1 = tsdbIterMergerGetTombRecord(writer->ctx->tombIterMerger); + if (record1 == NULL) { + writer->ctx->hasTomb = false; + break; + } + + int32_t c = tTombRecordCompare(record1, record); + if (c <= 0) { + code = tsdbFSetWriteTombRecord(writer->ctx->fsetWriter, record1); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + break; + } + } + + if (record->suid == INT64_MAX) { + ASSERT(writer->ctx->hasTomb == false); + goto _exit; + } + + code = tsdbFSetWriteTombRecord(writer->ctx->fsetWriter, record); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbSnapWriteFileSetEnd(STsdbSnapWriter* writer) { + if (!writer->ctx->fsetWriteBegin) return 0; + + int32_t code = 0; + int32_t lino = 0; + + SRowInfo row = { + .suid = INT64_MAX, + .uid = INT64_MAX, + }; + + code = tsdbSnapWriteTimeSeriesRow(writer, &row); + TSDB_CHECK_CODE(code, lino, _exit); + + STombRecord record = { + .suid = INT64_MAX, + .uid = INT64_MAX, + }; + + code = tsdbSnapWriteTombRecord(writer, &record); + TSDB_CHECK_CODE(code, lino, _exit); + + // close write + code = tsdbSnapWriteFileSetCloseWriter(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSnapWriteFileSetCloseIter(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSnapWriteFileSetCloseReader(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + writer->ctx->fsetWriteBegin = false; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbSnapWriteTimeSeriesData(STsdbSnapWriter* writer, SSnapDataHdr* hdr) { + int32_t code = 0; + int32_t lino = 0; + + SBlockData blockData[1] = {0}; + + code = tDecmprBlockData(hdr->data, hdr->size - sizeof(*hdr), blockData, writer->aBuf); + TSDB_CHECK_CODE(code, lino, _exit); + + int32_t fid = tsdbKeyFid(blockData->aTSKEY[0], writer->minutes, writer->precision); + if (!writer->ctx->fsetWriteBegin || fid != writer->ctx->fid) { + code = tsdbSnapWriteFileSetEnd(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSnapWriteFileSetBegin(writer, fid); TSDB_CHECK_CODE(code, lino, _exit); } + for (int32_t i = 0; i < blockData->nRow; ++i) { + SRowInfo rowInfo = { + .suid = blockData->suid, + .uid = blockData->uid ? blockData->uid : blockData->aUid[i], + .row = tsdbRowFromBlockData(blockData, i), + }; + + code = tsdbSnapWriteTimeSeriesRow(writer, &rowInfo); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); + } else { + tsdbDebug("vgId:%d %s done, suid:%" PRId64 " uid:%" PRId64 " nRow:%d", TD_VID(writer->tsdb->pVnode), __func__, + blockData->suid, blockData->uid, blockData->nRow); + } + tBlockDataDestroy(blockData); + return code; +} + +static int32_t tsdbSnapWriteDecmprTombBlock(SSnapDataHdr* hdr, STombBlock* tombBlock) { + int32_t code = 0; + int32_t lino = 0; + + int64_t size = hdr->size - sizeof(*hdr); + ASSERT(size % TOMB_RECORD_ELEM_NUM == 0); + size = size / TOMB_RECORD_ELEM_NUM; + ASSERT(size % sizeof(int64_t) == 0); + + int64_t* data = (int64_t*)hdr->data; + for (int32_t i = 0; i < TOMB_RECORD_ELEM_NUM; ++i) { + code = TARRAY2_APPEND_BATCH(&tombBlock->dataArr[i], hdr->data + i * size, size / sizeof(int64_t)); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + return code; +} + +static int32_t tsdbSnapWriteTombData(STsdbSnapWriter* writer, SSnapDataHdr* hdr) { + int32_t code = 0; + int32_t lino = 0; + + STombRecord record; + STombBlock tombBlock[1] = {0}; + + code = tsdbSnapWriteDecmprTombBlock(hdr, tombBlock); + TSDB_CHECK_CODE(code, lino, _exit); + + tTombBlockGet(tombBlock, 0, &record); + int32_t fid = tsdbKeyFid(record.skey, writer->minutes, writer->precision); + if (!writer->ctx->fsetWriteBegin || fid != writer->ctx->fid) { + code = tsdbSnapWriteFileSetEnd(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSnapWriteFileSetBegin(writer, fid); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (writer->ctx->hasData) { + SRowInfo row = { + .suid = INT64_MAX, + .uid = INT64_MAX, + }; + + code = tsdbSnapWriteTimeSeriesRow(writer, &row); + TSDB_CHECK_CODE(code, lino, _exit); + } + + ASSERT(writer->ctx->hasData == false); + + for (int32_t i = 0; i < TOMB_BLOCK_SIZE(tombBlock); ++i) { + tTombBlockGet(tombBlock, i, &record); + + code = tsdbSnapWriteTombRecord(writer, &record); + TSDB_CHECK_CODE(code, lino, _exit); + } + + tTombBlockDestroy(tombBlock); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbSnapWriterOpen(STsdb* pTsdb, int64_t sver, int64_t ever, STsdbSnapWriter** writer) { + int32_t code = 0; + int32_t lino = 0; + + writer[0] = taosMemoryCalloc(1, sizeof(*writer[0])); + if (writer[0] == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + writer[0]->tsdb = pTsdb; + writer[0]->sver = sver; + writer[0]->ever = ever; + writer[0]->minutes = pTsdb->keepCfg.days; + writer[0]->precision = pTsdb->keepCfg.precision; + writer[0]->minRow = pTsdb->pVnode->config.tsdbCfg.minRows; + writer[0]->maxRow = pTsdb->pVnode->config.tsdbCfg.maxRows; + writer[0]->commitID = tsdbFSAllocEid(pTsdb->pFS); + writer[0]->szPage = pTsdb->pVnode->config.tsdbPageSize; + writer[0]->compactVersion = INT64_MAX; + writer[0]->now = taosGetTimestampMs(); + + code = tsdbFSCreateCopySnapshot(pTsdb->pFS, &writer[0]->fsetArr); + TSDB_CHECK_CODE(code, lino, _exit); + + tsdbFSDisableBgTask(pTsdb->pFS); + _exit: if (code) { tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbDebug("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); - } - return code; -} - -static int32_t tsdbSnapWriteDelDataEnd(STsdbSnapWriter* pWriter) { - int32_t code = 0; - int32_t lino = 0; - - STsdb* pTsdb = pWriter->pTsdb; - - // end remaining table with NULL data - code = tsdbSnapWriteDelTableData(pWriter, NULL, NULL, 0); - TSDB_CHECK_CODE(code, lino, _exit); - - // update file-level info - code = tsdbWriteDelIdx(pWriter->pDelFWriter, pWriter->aDelIdx); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbUpdateDelFileHdr(pWriter->pDelFWriter); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbFSUpsertDelFile(&pWriter->fs, &pWriter->pDelFWriter->fDel); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tsdbDelFWriterClose(&pWriter->pDelFWriter, 1); - TSDB_CHECK_CODE(code, lino, _exit); - - if (pWriter->pDelFReader) { - code = tsdbDelFReaderClose(&pWriter->pDelFReader); - TSDB_CHECK_CODE(code, lino, _exit); - } - - if (pWriter->pTIter) { - tsdbCloseDataIter2(pWriter->pTIter); - pWriter->pTIter = NULL; - } - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); - } else { - tsdbInfo("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); - } - return code; -} - -static int32_t tsdbSnapWriteDelData(STsdbSnapWriter* pWriter, SSnapDataHdr* pHdr) { - int32_t code = 0; - int32_t lino = 0; - - STsdb* pTsdb = pWriter->pTsdb; - - // start to write del data if need - if (pWriter->pDelFWriter == NULL) { - code = tsdbSnapWriteDelDataStart(pWriter); - TSDB_CHECK_CODE(code, lino, _exit); - } - - // do write del data - code = tsdbSnapWriteDelTableData(pWriter, (TABLEID*)pHdr->data, pHdr->data + sizeof(TABLEID), - pHdr->size - sizeof(TABLEID)); - TSDB_CHECK_CODE(code, lino, _exit); - -_exit: - if (code) { - tsdbError("vgId:%d %s failed since %s", TD_VID(pTsdb->pVnode), __func__, tstrerror(code)); - } else { - tsdbTrace("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); - } - return code; -} - -// APIs -int32_t tsdbSnapWriterOpen(STsdb* pTsdb, int64_t sver, int64_t ever, STsdbSnapWriter** ppWriter) { - int32_t code = 0; - int32_t lino = 0; - - // alloc - STsdbSnapWriter* pWriter = (STsdbSnapWriter*)taosMemoryCalloc(1, sizeof(*pWriter)); - if (pWriter == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - pWriter->pTsdb = pTsdb; - pWriter->sver = sver; - pWriter->ever = ever; - pWriter->minutes = pTsdb->keepCfg.days; - pWriter->precision = pTsdb->keepCfg.precision; - pWriter->minRow = pTsdb->pVnode->config.tsdbCfg.minRows; - pWriter->maxRow = pTsdb->pVnode->config.tsdbCfg.maxRows; - pWriter->cmprAlg = pTsdb->pVnode->config.tsdbCfg.compression; - pWriter->commitID = pTsdb->pVnode->state.commitID; - - code = tsdbFSCopy(pTsdb, &pWriter->fs); - TSDB_CHECK_CODE(code, lino, _exit); - - // SNAP_DATA_TSDB - code = tBlockDataCreate(&pWriter->inData); - TSDB_CHECK_CODE(code, lino, _exit); - - pWriter->fid = INT32_MIN; - - code = tBlockDataCreate(&pWriter->bData); - TSDB_CHECK_CODE(code, lino, _exit); - - code = tBlockDataCreate(&pWriter->sData); - TSDB_CHECK_CODE(code, lino, _exit); - - // SNAP_DATA_DEL - -_exit: - if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); - if (pWriter) { - tBlockDataDestroy(&pWriter->sData); - tBlockDataDestroy(&pWriter->bData); - tBlockDataDestroy(&pWriter->inData); - tsdbFSDestroy(&pWriter->fs); - taosMemoryFree(pWriter); - pWriter = NULL; - } } else { tsdbInfo("vgId:%d %s done, sver:%" PRId64 " ever:%" PRId64, TD_VID(pTsdb->pVnode), __func__, sver, ever); } - *ppWriter = pWriter; return code; } -int32_t tsdbSnapWriterPrepareClose(STsdbSnapWriter* pWriter) { +int32_t tsdbSnapWriterPrepareClose(STsdbSnapWriter* writer) { int32_t code = 0; int32_t lino = 0; - if (pWriter->pDataFWriter) { - code = tsdbSnapWriteFileDataEnd(pWriter); - TSDB_CHECK_CODE(code, lino, _exit); - } + code = tsdbSnapWriteFileSetEnd(writer); + TSDB_CHECK_CODE(code, lino, _exit); - if (pWriter->pDelFWriter) { - code = tsdbSnapWriteDelDataEnd(pWriter); - TSDB_CHECK_CODE(code, lino, _exit); - } - - code = tsdbFSPrepareCommit(pWriter->pTsdb, &pWriter->fs); + code = tsdbFSEditBegin(writer->tsdb->pFS, writer->fopArr, TSDB_FEDIT_COMMIT); TSDB_CHECK_CODE(code, lino, _exit); _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code)); + TSDB_ERROR_LOG(TD_VID(writer->tsdb->pVnode), lino, code); } else { - tsdbDebug("vgId:%d %s done", TD_VID(pWriter->pTsdb->pVnode), __func__); + tsdbDebug("vgId:%d %s done", TD_VID(writer->tsdb->pVnode), __func__); } return code; } -int32_t tsdbSnapWriterClose(STsdbSnapWriter** ppWriter, int8_t rollback) { +int32_t tsdbSnapWriterClose(STsdbSnapWriter** writer, int8_t rollback) { + if (writer[0] == NULL) return 0; + int32_t code = 0; int32_t lino = 0; - STsdbSnapWriter* pWriter = *ppWriter; - STsdb* pTsdb = pWriter->pTsdb; + STsdb* tsdb = writer[0]->tsdb; if (rollback) { - tsdbRollbackCommit(pWriter->pTsdb); + code = tsdbFSEditAbort(writer[0]->tsdb->pFS); + TSDB_CHECK_CODE(code, lino, _exit); } else { - // lock - taosThreadRwlockWrlock(&pTsdb->rwLock); + taosThreadRwlockWrlock(&writer[0]->tsdb->rwLock); - code = tsdbFSCommit(pWriter->pTsdb); + code = tsdbFSEditCommit(writer[0]->tsdb->pFS); if (code) { - taosThreadRwlockUnlock(&pTsdb->rwLock); + taosThreadRwlockUnlock(&writer[0]->tsdb->rwLock); TSDB_CHECK_CODE(code, lino, _exit); } - // unlock - taosThreadRwlockUnlock(&pTsdb->rwLock); + taosThreadRwlockUnlock(&writer[0]->tsdb->rwLock); + } + tsdbFSEnableBgTask(tsdb->pFS); + + tsdbIterMergerClose(&writer[0]->ctx->tombIterMerger); + tsdbIterMergerClose(&writer[0]->ctx->dataIterMerger); + TARRAY2_DESTROY(writer[0]->ctx->tombIterArr, tsdbIterClose); + TARRAY2_DESTROY(writer[0]->ctx->dataIterArr, tsdbIterClose); + TARRAY2_DESTROY(writer[0]->ctx->sttReaderArr, tsdbSttFileReaderClose); + tsdbDataFileReaderClose(&writer[0]->ctx->dataReader); + + TARRAY2_DESTROY(writer[0]->fopArr, NULL); + tsdbFSDestroyCopySnapshot(&writer[0]->fsetArr); + + for (int32_t i = 0; i < ARRAY_SIZE(writer[0]->aBuf); ++i) { + tFree(writer[0]->aBuf[i]); } - // SNAP_DATA_DEL - taosArrayDestroy(pWriter->aDelData); - taosArrayDestroy(pWriter->aDelIdx); - - // SNAP_DATA_TSDB - tBlockDataDestroy(&pWriter->sData); - tBlockDataDestroy(&pWriter->bData); - taosArrayDestroy(pWriter->aSttBlk); - tMapDataClear(&pWriter->mDataBlk); - taosArrayDestroy(pWriter->aBlockIdx); - tDestroyTSchema(pWriter->skmTable.pTSchema); - tBlockDataDestroy(&pWriter->inData); - - for (int32_t iBuf = 0; iBuf < sizeof(pWriter->aBuf) / sizeof(uint8_t*); iBuf++) { - tFree(pWriter->aBuf[iBuf]); - } - tsdbFSDestroy(&pWriter->fs); - taosMemoryFree(pWriter); - *ppWriter = NULL; + taosMemoryFree(writer[0]); + writer[0] = NULL; _exit: if (code) { - tsdbError("vgId:%d %s failed at line %d since %s", TD_VID(pTsdb->pVnode), __func__, lino, tstrerror(code)); + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); } else { - tsdbInfo("vgId:%d %s done", TD_VID(pTsdb->pVnode), __func__); + tsdbInfo("vgId:%d %s done", TD_VID(tsdb->pVnode), __func__); } return code; } -int32_t tsdbSnapWrite(STsdbSnapWriter* pWriter, SSnapDataHdr* pHdr) { +int32_t tsdbSnapWrite(STsdbSnapWriter* writer, SSnapDataHdr* hdr) { int32_t code = 0; int32_t lino = 0; - if (pHdr->type == SNAP_DATA_TSDB) { - code = tsdbSnapWriteTimeSeriesData(pWriter, pHdr); + if (hdr->type == SNAP_DATA_TSDB) { + code = tsdbSnapWriteTimeSeriesData(writer, hdr); TSDB_CHECK_CODE(code, lino, _exit); - goto _exit; - } else if (pWriter->pDataFWriter) { - code = tsdbSnapWriteFileDataEnd(pWriter); + } else if (hdr->type == SNAP_DATA_DEL) { + code = tsdbSnapWriteTombData(writer, hdr); TSDB_CHECK_CODE(code, lino, _exit); - } - - if (pHdr->type == SNAP_DATA_DEL) { - code = tsdbSnapWriteDelData(pWriter, pHdr); - TSDB_CHECK_CODE(code, lino, _exit); - goto _exit; + } else { + ASSERT(0); } _exit: if (code) { tsdbError("vgId:%d %s failed at line %d since %s, type:%d index:%" PRId64 " size:%" PRId64, - TD_VID(pWriter->pTsdb->pVnode), __func__, lino, tstrerror(code), pHdr->type, pHdr->index, pHdr->size); + TD_VID(writer->tsdb->pVnode), __func__, lino, tstrerror(code), hdr->type, hdr->index, hdr->size); } else { - tsdbDebug("vgId:%d %s done, type:%d index:%" PRId64 " size:%" PRId64, TD_VID(pWriter->pTsdb->pVnode), __func__, - pHdr->type, pHdr->index, pHdr->size); + tsdbDebug("vgId:%d %s done, type:%d index:%" PRId64 " size:%" PRId64, TD_VID(writer->tsdb->pVnode), __func__, + hdr->type, hdr->index, hdr->size); } return code; } diff --git a/source/dnode/vnode/src/tsdb/tsdbSttFileRW.c b/source/dnode/vnode/src/tsdb/tsdbSttFileRW.c new file mode 100644 index 0000000000..db8d14d228 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbSttFileRW.c @@ -0,0 +1,982 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbSttFileRW.h" + +// SSttFReader ============================================================ +struct SSttFileReader { + SSttFileReaderConfig config[1]; + STsdbFD *fd; + SSttFooter footer[1]; + struct { + bool sttBlkLoaded; + bool statisBlkLoaded; + bool tombBlkLoaded; + } ctx[1]; + TSttBlkArray sttBlkArray[1]; + TStatisBlkArray statisBlkArray[1]; + TTombBlkArray tombBlkArray[1]; + uint8_t *bufArr[5]; +}; + +// SSttFileReader +int32_t tsdbSttFileReaderOpen(const char *fname, const SSttFileReaderConfig *config, SSttFileReader **reader) { + int32_t code = 0; + int32_t lino = 0; + + reader[0] = taosMemoryCalloc(1, sizeof(*reader[0])); + if (reader[0] == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + reader[0]->config[0] = config[0]; + if (reader[0]->config->bufArr == NULL) { + reader[0]->config->bufArr = reader[0]->bufArr; + } + + // open file + if (fname) { + code = tsdbOpenFile(fname, config->szPage, TD_FILE_READ, &reader[0]->fd); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + char fname1[TSDB_FILENAME_LEN]; + tsdbTFileName(config->tsdb, config->file, fname1); + code = tsdbOpenFile(fname1, config->szPage, TD_FILE_READ, &reader[0]->fd); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // // open each segment reader + int64_t offset = config->file->size - sizeof(SSttFooter); + ASSERT(offset >= TSDB_FHDR_SIZE); + + code = tsdbReadFile(reader[0]->fd, offset, (uint8_t *)(reader[0]->footer), sizeof(SSttFooter)); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(config->tsdb->pVnode), lino, code); + tsdbSttFileReaderClose(reader); + } + return code; +} + +int32_t tsdbSttFileReaderClose(SSttFileReader **reader) { + if (reader[0]) { + for (int32_t i = 0; i < ARRAY_SIZE(reader[0]->bufArr); ++i) { + tFree(reader[0]->bufArr[i]); + } + tsdbCloseFile(&reader[0]->fd); + TARRAY2_DESTROY(reader[0]->tombBlkArray, NULL); + TARRAY2_DESTROY(reader[0]->statisBlkArray, NULL); + TARRAY2_DESTROY(reader[0]->sttBlkArray, NULL); + taosMemoryFree(reader[0]); + reader[0] = NULL; + } + return 0; +} + +// SSttFSegReader +int32_t tsdbSttFileReadStatisBlk(SSttFileReader *reader, const TStatisBlkArray **statisBlkArray) { + if (!reader->ctx->statisBlkLoaded) { + if (reader->footer->statisBlkPtr->size > 0) { + ASSERT(reader->footer->statisBlkPtr->size % sizeof(SStatisBlk) == 0); + + int32_t size = reader->footer->statisBlkPtr->size / sizeof(SStatisBlk); + void *data = taosMemoryMalloc(reader->footer->statisBlkPtr->size); + if (!data) return TSDB_CODE_OUT_OF_MEMORY; + + int32_t code = + tsdbReadFile(reader->fd, reader->footer->statisBlkPtr->offset, data, reader->footer->statisBlkPtr->size); + if (code) { + taosMemoryFree(data); + return code; + } + + TARRAY2_INIT_EX(reader->statisBlkArray, size, size, data); + } else { + TARRAY2_INIT(reader->statisBlkArray); + } + + reader->ctx->statisBlkLoaded = true; + } + + statisBlkArray[0] = reader->statisBlkArray; + return 0; +} + +int32_t tsdbSttFileReadTombBlk(SSttFileReader *reader, const TTombBlkArray **tombBlkArray) { + if (!reader->ctx->tombBlkLoaded) { + if (reader->footer->tombBlkPtr->size > 0) { + ASSERT(reader->footer->tombBlkPtr->size % sizeof(STombBlk) == 0); + + int32_t size = reader->footer->tombBlkPtr->size / sizeof(STombBlk); + void *data = taosMemoryMalloc(reader->footer->tombBlkPtr->size); + if (!data) return TSDB_CODE_OUT_OF_MEMORY; + + int32_t code = + tsdbReadFile(reader->fd, reader->footer->tombBlkPtr->offset, data, reader->footer->tombBlkPtr->size); + if (code) { + taosMemoryFree(data); + return code; + } + + TARRAY2_INIT_EX(reader->tombBlkArray, size, size, data); + } else { + TARRAY2_INIT(reader->tombBlkArray); + } + + reader->ctx->tombBlkLoaded = true; + } + + tombBlkArray[0] = reader->tombBlkArray; + return 0; +} + +int32_t tsdbSttFileReadSttBlk(SSttFileReader *reader, const TSttBlkArray **sttBlkArray) { + if (!reader->ctx->sttBlkLoaded) { + if (reader->footer->sttBlkPtr->size > 0) { + ASSERT(reader->footer->sttBlkPtr->size % sizeof(SSttBlk) == 0); + + int32_t size = reader->footer->sttBlkPtr->size / sizeof(SSttBlk); + void *data = taosMemoryMalloc(reader->footer->sttBlkPtr->size); + if (!data) return TSDB_CODE_OUT_OF_MEMORY; + + int32_t code = tsdbReadFile(reader->fd, reader->footer->sttBlkPtr->offset, data, reader->footer->sttBlkPtr->size); + if (code) { + taosMemoryFree(data); + return code; + } + + TARRAY2_INIT_EX(reader->sttBlkArray, size, size, data); + } else { + TARRAY2_INIT(reader->sttBlkArray); + } + + reader->ctx->sttBlkLoaded = true; + } + + sttBlkArray[0] = reader->sttBlkArray; + return 0; +} + +int32_t tsdbSttFileReadBlockData(SSttFileReader *reader, const SSttBlk *sttBlk, SBlockData *bData) { + int32_t code = 0; + int32_t lino = 0; + + code = tRealloc(&reader->config->bufArr[0], sttBlk->bInfo.szBlock); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd, sttBlk->bInfo.offset, reader->config->bufArr[0], sttBlk->bInfo.szBlock); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tDecmprBlockData(reader->config->bufArr[0], sttBlk->bInfo.szBlock, bData, &reader->config->bufArr[1]); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbSttFileReadBlockDataByColumn(SSttFileReader *reader, const SSttBlk *sttBlk, SBlockData *bData, + STSchema *pTSchema, int16_t cids[], int32_t ncid) { + int32_t code = 0; + int32_t lino = 0; + + TABLEID tbid = {.suid = sttBlk->suid}; + if (tbid.suid == 0) { + tbid.uid = sttBlk->minUid; + } else { + tbid.uid = 0; + } + + code = tBlockDataInit(bData, &tbid, pTSchema, cids, ncid); + TSDB_CHECK_CODE(code, lino, _exit); + + // uid + version + tskey + code = tRealloc(&reader->config->bufArr[0], sttBlk->bInfo.szKey); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd, sttBlk->bInfo.offset, reader->config->bufArr[0], sttBlk->bInfo.szKey); + TSDB_CHECK_CODE(code, lino, _exit); + + // hdr + SDiskDataHdr hdr[1]; + int32_t size = 0; + + size += tGetDiskDataHdr(reader->config->bufArr[0] + size, hdr); + + ASSERT(hdr->delimiter == TSDB_FILE_DLMT); + + bData->nRow = hdr->nRow; + bData->uid = hdr->uid; + + // uid + if (hdr->uid == 0) { + ASSERT(hdr->szUid); + code = tsdbDecmprData(reader->config->bufArr[0] + size, hdr->szUid, TSDB_DATA_TYPE_BIGINT, hdr->cmprAlg, + (uint8_t **)&bData->aUid, sizeof(int64_t) * hdr->nRow, &reader->config->bufArr[1]); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + ASSERT(hdr->szUid == 0); + } + size += hdr->szUid; + + // version + code = tsdbDecmprData(reader->config->bufArr[0] + size, hdr->szVer, TSDB_DATA_TYPE_BIGINT, hdr->cmprAlg, + (uint8_t **)&bData->aVersion, sizeof(int64_t) * hdr->nRow, &reader->config->bufArr[1]); + TSDB_CHECK_CODE(code, lino, _exit); + size += hdr->szVer; + + // ts + code = tsdbDecmprData(reader->config->bufArr[0] + size, hdr->szKey, TSDB_DATA_TYPE_TIMESTAMP, hdr->cmprAlg, + (uint8_t **)&bData->aTSKEY, sizeof(TSKEY) * hdr->nRow, &reader->config->bufArr[1]); + TSDB_CHECK_CODE(code, lino, _exit); + size += hdr->szKey; + + ASSERT(size == sttBlk->bInfo.szKey); + + // other columns + if (bData->nColData > 0) { + if (hdr->szBlkCol > 0) { + code = tRealloc(&reader->config->bufArr[0], hdr->szBlkCol); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd, sttBlk->bInfo.offset + sttBlk->bInfo.szKey, reader->config->bufArr[0], + hdr->szBlkCol); + TSDB_CHECK_CODE(code, lino, _exit); + } + + SBlockCol bc[1] = {{.cid = 0}}; + SBlockCol *blockCol = bc; + + size = 0; + for (int32_t i = 0; i < bData->nColData; i++) { + SColData *colData = tBlockDataGetColDataByIdx(bData, i); + + while (blockCol && blockCol->cid < colData->cid) { + if (size < hdr->szBlkCol) { + size += tGetBlockCol(reader->config->bufArr[0] + size, blockCol); + } else { + ASSERT(size == hdr->szBlkCol); + blockCol = NULL; + } + } + + if (blockCol == NULL || blockCol->cid > colData->cid) { + for (int32_t iRow = 0; iRow < hdr->nRow; iRow++) { + code = tColDataAppendValue(colData, &COL_VAL_NONE(colData->cid, colData->type)); + TSDB_CHECK_CODE(code, lino, _exit); + } + } else { + ASSERT(blockCol->type == colData->type); + ASSERT(blockCol->flag && blockCol->flag != HAS_NONE); + + if (blockCol->flag == HAS_NULL) { + for (int32_t iRow = 0; iRow < hdr->nRow; iRow++) { + code = tColDataAppendValue(colData, &COL_VAL_NULL(blockCol->cid, blockCol->type)); + TSDB_CHECK_CODE(code, lino, _exit); + } + } else { + int32_t size1 = blockCol->szBitmap + blockCol->szOffset + blockCol->szValue; + + code = tRealloc(&reader->config->bufArr[1], size1); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd, sttBlk->bInfo.offset + sttBlk->bInfo.szKey + hdr->szBlkCol + blockCol->offset, + reader->config->bufArr[1], size1); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDecmprColData(reader->config->bufArr[1], blockCol, hdr->cmprAlg, hdr->nRow, colData, + &reader->config->bufArr[2]); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbSttFileReadTombBlock(SSttFileReader *reader, const STombBlk *tombBlk, STombBlock *tombBlock) { + int32_t code = 0; + int32_t lino = 0; + + code = tRealloc(&reader->config->bufArr[0], tombBlk->dp->size); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd, tombBlk->dp->offset, reader->config->bufArr[0], tombBlk->dp->size); + if (code) TSDB_CHECK_CODE(code, lino, _exit); + + int64_t size = 0; + tTombBlockClear(tombBlock); + for (int32_t i = 0; i < ARRAY_SIZE(tombBlock->dataArr); ++i) { + code = tsdbDecmprData(reader->config->bufArr[0] + size, tombBlk->size[i], TSDB_DATA_TYPE_BIGINT, tombBlk->cmprAlg, + &reader->config->bufArr[1], sizeof(int64_t) * tombBlk->numRec, &reader->config->bufArr[2]); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND_BATCH(&tombBlock->dataArr[i], reader->config->bufArr[1], tombBlk->numRec); + TSDB_CHECK_CODE(code, lino, _exit); + + size += tombBlk->size[i]; + } + + ASSERT(size == tombBlk->dp->size); +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbSttFileReadStatisBlock(SSttFileReader *reader, const SStatisBlk *statisBlk, STbStatisBlock *statisBlock) { + int32_t code = 0; + int32_t lino = 0; + + code = tRealloc(&reader->config->bufArr[0], statisBlk->dp->size); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadFile(reader->fd, statisBlk->dp->offset, reader->config->bufArr[0], statisBlk->dp->size); + TSDB_CHECK_CODE(code, lino, _exit); + + int64_t size = 0; + tStatisBlockClear(statisBlock); + for (int32_t i = 0; i < ARRAY_SIZE(statisBlock->dataArr); ++i) { + code = + tsdbDecmprData(reader->config->bufArr[0] + size, statisBlk->size[i], TSDB_DATA_TYPE_BIGINT, statisBlk->cmprAlg, + &reader->config->bufArr[1], sizeof(int64_t) * statisBlk->numRec, &reader->config->bufArr[2]); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND_BATCH(statisBlock->dataArr + i, reader->config->bufArr[1], statisBlk->numRec); + TSDB_CHECK_CODE(code, lino, _exit); + + size += statisBlk->size[i]; + } + + ASSERT(size == statisBlk->dp->size); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(reader->config->tsdb->pVnode), lino, code); + } + return code; +} + +// SSttFWriter ============================================================ +struct SSttFileWriter { + SSttFileWriterConfig config[1]; + struct { + bool opened; + TABLEID tbid[1]; + } ctx[1]; + // file + STsdbFD *fd; + STFile file[1]; + // data + SSttFooter footer[1]; + TTombBlkArray tombBlkArray[1]; + TSttBlkArray sttBlkArray[1]; + TStatisBlkArray statisBlkArray[1]; + STombBlock tombBlock[1]; + STbStatisBlock staticBlock[1]; + SBlockData blockData[1]; + // helper data + SSkmInfo skmTb[1]; + SSkmInfo skmRow[1]; + uint8_t *bufArr[5]; +}; + +int32_t tsdbFileDoWriteBlockData(STsdbFD *fd, SBlockData *blockData, int8_t cmprAlg, int64_t *fileSize, + TSttBlkArray *sttBlkArray, uint8_t **bufArr) { + if (blockData->nRow == 0) return 0; + + int32_t code = 0; + + SSttBlk sttBlk[1] = {{ + .suid = blockData->suid, + .minUid = blockData->uid ? blockData->uid : blockData->aUid[0], + .maxUid = blockData->uid ? blockData->uid : blockData->aUid[blockData->nRow - 1], + .minKey = blockData->aTSKEY[0], + .maxKey = blockData->aTSKEY[0], + .minVer = blockData->aVersion[0], + .maxVer = blockData->aVersion[0], + .nRow = blockData->nRow, + }}; + + for (int32_t iRow = 1; iRow < blockData->nRow; iRow++) { + if (sttBlk->minKey > blockData->aTSKEY[iRow]) sttBlk->minKey = blockData->aTSKEY[iRow]; + if (sttBlk->maxKey < blockData->aTSKEY[iRow]) sttBlk->maxKey = blockData->aTSKEY[iRow]; + if (sttBlk->minVer > blockData->aVersion[iRow]) sttBlk->minVer = blockData->aVersion[iRow]; + if (sttBlk->maxVer < blockData->aVersion[iRow]) sttBlk->maxVer = blockData->aVersion[iRow]; + } + + int32_t sizeArr[5] = {0}; + code = tCmprBlockData(blockData, cmprAlg, NULL, NULL, bufArr, sizeArr); + if (code) return code; + + sttBlk->bInfo.offset = *fileSize; + sttBlk->bInfo.szKey = sizeArr[2] + sizeArr[3]; + sttBlk->bInfo.szBlock = sizeArr[0] + sizeArr[1] + sttBlk->bInfo.szKey; + + for (int32_t i = 3; i >= 0; i--) { + if (sizeArr[i]) { + code = tsdbWriteFile(fd, *fileSize, bufArr[i], sizeArr[i]); + if (code) return code; + *fileSize += sizeArr[i]; + } + } + + code = TARRAY2_APPEND_PTR(sttBlkArray, sttBlk); + if (code) return code; + + tBlockDataClear(blockData); + + return 0; +} + +static int32_t tsdbSttFileDoWriteBlockData(SSttFileWriter *writer) { + if (writer->blockData->nRow == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileDoWriteBlockData(writer->fd, writer->blockData, writer->config->cmprAlg, &writer->file->size, + writer->sttBlkArray, writer->config->bufArr); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbSttFileDoWriteStatisBlock(SSttFileWriter *writer) { + if (STATIS_BLOCK_SIZE(writer->staticBlock) == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + SStatisBlk statisBlk[1] = {{ + .dp[0] = + { + .offset = writer->file->size, + .size = 0, + }, + .minTbid = + { + .suid = TARRAY2_FIRST(writer->staticBlock->suid), + .uid = TARRAY2_FIRST(writer->staticBlock->uid), + }, + .maxTbid = + { + .suid = TARRAY2_LAST(writer->staticBlock->suid), + .uid = TARRAY2_LAST(writer->staticBlock->uid), + }, + .numRec = STATIS_BLOCK_SIZE(writer->staticBlock), + .cmprAlg = writer->config->cmprAlg, + }}; + + for (int32_t i = 0; i < STATIS_RECORD_NUM_ELEM; i++) { + code = tsdbCmprData((uint8_t *)TARRAY2_DATA(writer->staticBlock->dataArr + i), + TARRAY2_DATA_LEN(&writer->staticBlock->dataArr[i]), TSDB_DATA_TYPE_BIGINT, statisBlk->cmprAlg, + &writer->config->bufArr[0], 0, &statisBlk->size[i], &writer->config->bufArr[1]); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbWriteFile(writer->fd, writer->file->size, writer->config->bufArr[0], statisBlk->size[i]); + TSDB_CHECK_CODE(code, lino, _exit); + + statisBlk->dp->size += statisBlk->size[i]; + writer->file->size += statisBlk->size[i]; + } + + code = TARRAY2_APPEND_PTR(writer->statisBlkArray, statisBlk); + TSDB_CHECK_CODE(code, lino, _exit); + + tStatisBlockClear(writer->staticBlock); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFileWriteTombBlock(STsdbFD *fd, STombBlock *tombBlock, int8_t cmprAlg, int64_t *fileSize, + TTombBlkArray *tombBlkArray, uint8_t **bufArr) { + int32_t code; + + if (TOMB_BLOCK_SIZE(tombBlock) == 0) return 0; + + STombBlk tombBlk[1] = {{ + .dp[0] = + { + .offset = *fileSize, + .size = 0, + }, + .minTbid = + { + .suid = TARRAY2_FIRST(tombBlock->suid), + .uid = TARRAY2_FIRST(tombBlock->uid), + }, + .maxTbid = + { + .suid = TARRAY2_LAST(tombBlock->suid), + .uid = TARRAY2_LAST(tombBlock->uid), + }, + .minVer = TARRAY2_FIRST(tombBlock->version), + .maxVer = TARRAY2_FIRST(tombBlock->version), + .numRec = TOMB_BLOCK_SIZE(tombBlock), + .cmprAlg = cmprAlg, + }}; + + for (int32_t i = 1; i < TOMB_BLOCK_SIZE(tombBlock); i++) { + if (tombBlk->minVer > TARRAY2_GET(tombBlock->version, i)) { + tombBlk->minVer = TARRAY2_GET(tombBlock->version, i); + } + if (tombBlk->maxVer < TARRAY2_GET(tombBlock->version, i)) { + tombBlk->maxVer = TARRAY2_GET(tombBlock->version, i); + } + } + + for (int32_t i = 0; i < ARRAY_SIZE(tombBlock->dataArr); i++) { + code = tsdbCmprData((uint8_t *)TARRAY2_DATA(&tombBlock->dataArr[i]), TARRAY2_DATA_LEN(&tombBlock->dataArr[i]), + TSDB_DATA_TYPE_BIGINT, tombBlk->cmprAlg, &bufArr[0], 0, &tombBlk->size[i], &bufArr[1]); + if (code) return code; + + code = tsdbWriteFile(fd, *fileSize, bufArr[0], tombBlk->size[i]); + if (code) return code; + + tombBlk->dp->size += tombBlk->size[i]; + *fileSize += tombBlk->size[i]; + } + + code = TARRAY2_APPEND_PTR(tombBlkArray, tombBlk); + if (code) return code; + + tTombBlockClear(tombBlock); + return 0; +} + +static int32_t tsdbSttFileDoWriteTombBlock(SSttFileWriter *writer) { + if (TOMB_BLOCK_SIZE(writer->tombBlock) == 0) return 0; + + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileWriteTombBlock(writer->fd, writer->tombBlock, writer->config->cmprAlg, &writer->file->size, + writer->tombBlkArray, writer->config->bufArr); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFileWriteSttBlk(STsdbFD *fd, const TSttBlkArray *sttBlkArray, SFDataPtr *ptr, int64_t *fileSize) { + ptr->size = TARRAY2_DATA_LEN(sttBlkArray); + if (ptr->size > 0) { + ptr->offset = *fileSize; + + int32_t code = tsdbWriteFile(fd, *fileSize, (const uint8_t *)TARRAY2_DATA(sttBlkArray), ptr->size); + if (code) { + return code; + } + + *fileSize += ptr->size; + } + return 0; +} + +static int32_t tsdbSttFileDoWriteSttBlk(SSttFileWriter *writer) { + int32_t code = 0; + int32_t lino; + + code = tsdbFileWriteSttBlk(writer->fd, writer->sttBlkArray, writer->footer->sttBlkPtr, &writer->file->size); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbSttFileDoWriteStatisBlk(SSttFileWriter *writer) { + int32_t code = 0; + int32_t lino; + + writer->footer->statisBlkPtr->size = TARRAY2_DATA_LEN(writer->statisBlkArray); + if (writer->footer->statisBlkPtr->size) { + writer->footer->statisBlkPtr->offset = writer->file->size; + code = tsdbWriteFile(writer->fd, writer->file->size, (const uint8_t *)TARRAY2_DATA(writer->statisBlkArray), + writer->footer->statisBlkPtr->size); + TSDB_CHECK_CODE(code, lino, _exit); + writer->file->size += writer->footer->statisBlkPtr->size; + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFileWriteTombBlk(STsdbFD *fd, const TTombBlkArray *tombBlkArray, SFDataPtr *ptr, int64_t *fileSize) { + ptr->size = TARRAY2_DATA_LEN(tombBlkArray); + if (ptr->size > 0) { + ptr->offset = *fileSize; + + int32_t code = tsdbWriteFile(fd, *fileSize, (const uint8_t *)TARRAY2_DATA(tombBlkArray), ptr->size); + if (code) { + return code; + } + + *fileSize += ptr->size; + } + return 0; +} + +static int32_t tsdbSttFileDoWriteTombBlk(SSttFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + + code = tsdbFileWriteTombBlk(writer->fd, writer->tombBlkArray, writer->footer->tombBlkPtr, &writer->file->size); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbFileWriteSttFooter(STsdbFD *fd, const SSttFooter *footer, int64_t *fileSize) { + int32_t code = tsdbWriteFile(fd, *fileSize, (const uint8_t *)footer, sizeof(*footer)); + if (code) return code; + *fileSize += sizeof(*footer); + return 0; +} + +static int32_t tsdbSttFileDoWriteFooter(SSttFileWriter *writer) { + return tsdbFileWriteSttFooter(writer->fd, writer->footer, &writer->file->size); +} + +static int32_t tsdbSttFWriterDoOpen(SSttFileWriter *writer) { + int32_t code = 0; + int32_t lino = 0; + + // set + if (!writer->config->skmTb) writer->config->skmTb = writer->skmTb; + if (!writer->config->skmRow) writer->config->skmRow = writer->skmRow; + if (!writer->config->bufArr) writer->config->bufArr = writer->bufArr; + + writer->file[0] = (STFile){ + .type = TSDB_FTYPE_STT, + .did = writer->config->did, + .fid = writer->config->fid, + .cid = writer->config->cid, + .size = 0, + .stt[0] = + { + .level = writer->config->level, + }, + }; + + // open file + int32_t flag = TD_FILE_READ | TD_FILE_WRITE | TD_FILE_CREATE | TD_FILE_TRUNC; + char fname[TSDB_FILENAME_LEN]; + + tsdbTFileName(writer->config->tsdb, writer->file, fname); + code = tsdbOpenFile(fname, writer->config->szPage, flag, &writer->fd); + TSDB_CHECK_CODE(code, lino, _exit); + + uint8_t hdr[TSDB_FHDR_SIZE] = {0}; + code = tsdbWriteFile(writer->fd, 0, hdr, sizeof(hdr)); + TSDB_CHECK_CODE(code, lino, _exit); + writer->file->size += sizeof(hdr); + + writer->ctx->opened = true; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static void tsdbSttFWriterDoClose(SSttFileWriter *writer) { + ASSERT(writer->fd == NULL); + + for (int32_t i = 0; i < ARRAY_SIZE(writer->bufArr); ++i) { + tFree(writer->bufArr[i]); + } + tDestroyTSchema(writer->skmRow->pTSchema); + tDestroyTSchema(writer->skmTb->pTSchema); + tTombBlockDestroy(writer->tombBlock); + tStatisBlockDestroy(writer->staticBlock); + tBlockDataDestroy(writer->blockData); + TARRAY2_DESTROY(writer->tombBlkArray, NULL); + TARRAY2_DESTROY(writer->statisBlkArray, NULL); + TARRAY2_DESTROY(writer->sttBlkArray, NULL); +} + +static int32_t tsdbSttFileDoUpdateHeader(SSttFileWriter *writer) { + // TODO + return 0; +} + +static int32_t tsdbSttFWriterCloseCommit(SSttFileWriter *writer, TFileOpArray *opArray) { + int32_t lino; + int32_t code; + + code = tsdbSttFileDoWriteBlockData(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSttFileDoWriteStatisBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSttFileDoWriteTombBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSttFileDoWriteSttBlk(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSttFileDoWriteStatisBlk(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSttFileDoWriteTombBlk(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSttFileDoWriteFooter(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbSttFileDoUpdateHeader(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbFsyncFile(writer->fd); + TSDB_CHECK_CODE(code, lino, _exit); + + tsdbCloseFile(&writer->fd); + + ASSERT(writer->file->size > 0); + STFileOp op = (STFileOp){ + .optype = TSDB_FOP_CREATE, + .fid = writer->config->fid, + .nf = writer->file[0], + }; + + code = TARRAY2_APPEND(opArray, op); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbSttFWriterCloseAbort(SSttFileWriter *writer) { + char fname[TSDB_FILENAME_LEN]; + tsdbTFileName(writer->config->tsdb, writer->file, fname); + tsdbCloseFile(&writer->fd); + taosRemoveFile(fname); + return 0; +} + +int32_t tsdbSttFileWriterOpen(const SSttFileWriterConfig *config, SSttFileWriter **writer) { + writer[0] = taosMemoryCalloc(1, sizeof(*writer[0])); + if (writer[0] == NULL) return TSDB_CODE_OUT_OF_MEMORY; + + writer[0]->config[0] = config[0]; + writer[0]->ctx->opened = false; + return 0; +} + +int32_t tsdbSttFileWriterClose(SSttFileWriter **writer, int8_t abort, TFileOpArray *opArray) { + int32_t code = 0; + int32_t lino = 0; + + if (writer[0]->ctx->opened) { + if (abort) { + code = tsdbSttFWriterCloseAbort(writer[0]); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + code = tsdbSttFWriterCloseCommit(writer[0], opArray); + TSDB_CHECK_CODE(code, lino, _exit); + } + tsdbSttFWriterDoClose(writer[0]); + } + taosMemoryFree(writer[0]); + writer[0] = NULL; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer[0]->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbSttFileWriteRow(SSttFileWriter *writer, SRowInfo *row) { + int32_t code = 0; + int32_t lino = 0; + + if (!writer->ctx->opened) { + code = tsdbSttFWriterDoOpen(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (!TABLE_SAME_SCHEMA(row->suid, row->uid, writer->ctx->tbid->suid, writer->ctx->tbid->uid)) { + code = tsdbSttFileDoWriteBlockData(writer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbUpdateSkmTb(writer->config->tsdb, (TABLEID *)row, writer->config->skmTb); + TSDB_CHECK_CODE(code, lino, _exit); + + TABLEID id = {.suid = row->suid, .uid = row->suid ? 0 : row->uid}; + code = tBlockDataInit(writer->blockData, &id, writer->config->skmTb->pTSchema, NULL, 0); + TSDB_CHECK_CODE(code, lino, _exit); + } + + TSDBKEY key[1]; + if (row->row.type == TSDBROW_ROW_FMT) { + key->ts = row->row.pTSRow->ts; + key->version = row->row.version; + } else { + key->ts = row->row.pBlockData->aTSKEY[row->row.iRow]; + key->version = row->row.pBlockData->aVersion[row->row.iRow]; + } + + if (writer->ctx->tbid->uid != row->uid) { + writer->ctx->tbid->suid = row->suid; + writer->ctx->tbid->uid = row->uid; + + if (STATIS_BLOCK_SIZE(writer->staticBlock) >= writer->config->maxRow) { + code = tsdbSttFileDoWriteStatisBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + STbStatisRecord record = { + .suid = row->suid, + .uid = row->uid, + .firstKey = key->ts, + .lastKey = key->ts, + .count = 1, + }; + code = tStatisBlockPut(writer->staticBlock, &record); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + ASSERT(key->ts >= TARRAY2_LAST(writer->staticBlock->lastKey)); + + if (key->ts > TARRAY2_LAST(writer->staticBlock->lastKey)) { + TARRAY2_LAST(writer->staticBlock->count)++; + TARRAY2_LAST(writer->staticBlock->lastKey) = key->ts; + } + } + + if (row->row.type == TSDBROW_ROW_FMT) { + code = tsdbUpdateSkmRow(writer->config->tsdb, writer->ctx->tbid, // + TSDBROW_SVERSION(&row->row), writer->config->skmRow); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // row to col conversion + if (key->version <= writer->config->compactVersion // + && writer->blockData->nRow > 0 // + && writer->blockData->aTSKEY[writer->blockData->nRow - 1] == key->ts // + && (writer->blockData->uid // + ? writer->blockData->uid // + : writer->blockData->aUid[writer->blockData->nRow - 1]) == row->uid // + ) { + code = tBlockDataUpdateRow(writer->blockData, &row->row, writer->config->skmRow->pTSchema); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + if (writer->blockData->nRow >= writer->config->maxRow) { + code = tsdbSttFileDoWriteBlockData(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tBlockDataAppendRow(writer->blockData, &row->row, writer->config->skmRow->pTSchema, row->uid); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbSttFileWriteBlockData(SSttFileWriter *writer, SBlockData *bdata) { + int32_t code = 0; + int32_t lino = 0; + + SRowInfo row[1]; + row->suid = bdata->suid; + for (int32_t i = 0; i < bdata->nRow; i++) { + row->uid = bdata->uid ? bdata->uid : bdata->aUid[i]; + row->row = tsdbRowFromBlockData(bdata, i); + + code = tsdbSttFileWriteRow(writer, row); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +int32_t tsdbSttFileWriteTombRecord(SSttFileWriter *writer, const STombRecord *record) { + int32_t code; + int32_t lino; + + if (!writer->ctx->opened) { + code = tsdbSttFWriterDoOpen(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + if (writer->blockData->nRow > 0) { + code = tsdbSttFileDoWriteBlockData(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (STATIS_BLOCK_SIZE(writer->staticBlock) > 0) { + code = tsdbSttFileDoWriteStatisBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + + code = tTombBlockPut(writer->tombBlock, record); + TSDB_CHECK_CODE(code, lino, _exit); + + if (TOMB_BLOCK_SIZE(writer->tombBlock) >= writer->config->maxRow) { + code = tsdbSttFileDoWriteTombBlock(writer); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(writer->config->tsdb->pVnode), lino, code); + } + return code; +} + +bool tsdbSttFileWriterIsOpened(SSttFileWriter *writer) { return writer->ctx->opened; } \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbSttFileRW.h b/source/dnode/vnode/src/tsdb/tsdbSttFileRW.h new file mode 100644 index 0000000000..242b55795c --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbSttFileRW.h @@ -0,0 +1,93 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbFS2.h" +#include "tsdbUtil2.h" + +#ifndef _TSDB_STT_FILE_RW_H +#define _TSDB_STT_FILE_RW_H + +#ifdef __cplusplus +extern "C" { +#endif + +typedef TARRAY2(SSttBlk) TSttBlkArray; +typedef TARRAY2(SStatisBlk) TStatisBlkArray; + +typedef struct { + SFDataPtr sttBlkPtr[1]; + SFDataPtr statisBlkPtr[1]; + SFDataPtr tombBlkPtr[1]; + SFDataPtr rsrvd[2]; +} SSttFooter; + +// SSttFileReader ========================================== +typedef struct SSttFileReader SSttFileReader; +typedef struct SSttFileReaderConfig SSttFileReaderConfig; +typedef TARRAY2(SSttFileReader *) TSttFileReaderArray; + +// SSttFileReader +int32_t tsdbSttFileReaderOpen(const char *fname, const SSttFileReaderConfig *config, SSttFileReader **reader); +int32_t tsdbSttFileReaderClose(SSttFileReader **reader); + +// SSttSegReader +int32_t tsdbSttFileReadSttBlk(SSttFileReader *reader, const TSttBlkArray **sttBlkArray); +int32_t tsdbSttFileReadStatisBlk(SSttFileReader *reader, const TStatisBlkArray **statisBlkArray); +int32_t tsdbSttFileReadTombBlk(SSttFileReader *reader, const TTombBlkArray **delBlkArray); + +int32_t tsdbSttFileReadBlockData(SSttFileReader *reader, const SSttBlk *sttBlk, SBlockData *bData); +int32_t tsdbSttFileReadBlockDataByColumn(SSttFileReader *reader, const SSttBlk *sttBlk, SBlockData *bData, + STSchema *pTSchema, int16_t cids[], int32_t ncid); +int32_t tsdbSttFileReadStatisBlock(SSttFileReader *reader, const SStatisBlk *statisBlk, STbStatisBlock *sData); +int32_t tsdbSttFileReadTombBlock(SSttFileReader *reader, const STombBlk *delBlk, STombBlock *dData); + +struct SSttFileReaderConfig { + STsdb *tsdb; + int32_t szPage; + STFile file[1]; + uint8_t **bufArr; +}; + +// SSttFileWriter ========================================== +typedef struct SSttFileWriter SSttFileWriter; +typedef struct SSttFileWriterConfig SSttFileWriterConfig; + +int32_t tsdbSttFileWriterOpen(const SSttFileWriterConfig *config, SSttFileWriter **writer); +int32_t tsdbSttFileWriterClose(SSttFileWriter **writer, int8_t abort, TFileOpArray *opArray); +int32_t tsdbSttFileWriteRow(SSttFileWriter *writer, SRowInfo *row); +int32_t tsdbSttFileWriteBlockData(SSttFileWriter *writer, SBlockData *pBlockData); +int32_t tsdbSttFileWriteTombRecord(SSttFileWriter *writer, const STombRecord *record); +bool tsdbSttFileWriterIsOpened(SSttFileWriter *writer); + +struct SSttFileWriterConfig { + STsdb *tsdb; + int32_t maxRow; + int32_t szPage; + int8_t cmprAlg; + int64_t compactVersion; + SDiskID did; + int32_t fid; + int64_t cid; + int32_t level; + SSkmInfo *skmTb; + SSkmInfo *skmRow; + uint8_t **bufArr; +}; + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_STT_FILE_RW_H*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbUpgrade.c b/source/dnode/vnode/src/tsdb/tsdbUpgrade.c new file mode 100644 index 0000000000..59ba51c371 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbUpgrade.c @@ -0,0 +1,640 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbUpgrade.h" + +// old +extern void tsdbGetCurrentFName(STsdb *pTsdb, char *current, char *current_t); +extern int32_t tsdbReadDataBlockEx(SDataFReader *pReader, SDataBlk *pDataBlk, SBlockData *pBlockData); + +// new +extern int32_t save_fs(const TFileSetArray *arr, const char *fname); +extern int32_t current_fname(STsdb *pTsdb, char *fname, EFCurrentT ftype); +extern int32_t tsdbFileWriteBrinBlock(STsdbFD *fd, SBrinBlock *brinBlock, int8_t cmprAlg, int64_t *fileSize, + TBrinBlkArray *brinBlkArray, uint8_t **bufArr); +extern int32_t tsdbFileWriteBrinBlk(STsdbFD *fd, TBrinBlkArray *brinBlkArray, SFDataPtr *ptr, int64_t *fileSize); +extern int32_t tsdbFileWriteHeadFooter(STsdbFD *fd, int64_t *fileSize, const SHeadFooter *footer); +extern int32_t tsdbSttLvlInit(int32_t level, SSttLvl **lvl); +extern int32_t tsdbSttLvlClear(SSttLvl **lvl); +extern int32_t tsdbFileWriteSttBlk(STsdbFD *fd, const TSttBlkArray *sttBlkArray, SFDataPtr *ptr, int64_t *fileSize); +extern int32_t tsdbFileWriteSttFooter(STsdbFD *fd, const SSttFooter *footer, int64_t *fileSize); +extern int32_t tsdbFileWriteTombBlock(STsdbFD *fd, STombBlock *tombBlock, int8_t cmprAlg, int64_t *fileSize, + TTombBlkArray *tombBlkArray, uint8_t **bufArr); +extern int32_t tsdbFileWriteTombBlk(STsdbFD *fd, const TTombBlkArray *tombBlkArray, SFDataPtr *ptr, int64_t *fileSize); +extern int32_t tsdbFileWriteTombFooter(STsdbFD *fd, const STombFooter *footer, int64_t *fileSize); + +static int32_t tsdbUpgradeHead(STsdb *tsdb, SDFileSet *pDFileSet, SDataFReader *reader, STFileSet *fset) { + int32_t code = 0; + int32_t lino = 0; + + // init + struct { + // config + int32_t maxRow; + int8_t cmprAlg; + int32_t szPage; + uint8_t *bufArr[8]; + // reader + SArray *aBlockIdx; + SMapData mDataBlk[1]; + SBlockData blockData[1]; + // writer + STsdbFD *fd; + SBrinBlock brinBlock[1]; + TBrinBlkArray brinBlkArray[1]; + SHeadFooter footer[1]; + } ctx[1] = {{ + .maxRow = tsdb->pVnode->config.tsdbCfg.maxRows, + .cmprAlg = tsdb->pVnode->config.tsdbCfg.compression, + .szPage = tsdb->pVnode->config.tsdbPageSize, + }}; + + // read SBlockIdx array + if ((ctx->aBlockIdx = taosArrayInit(0, sizeof(SBlockIdx))) == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbReadBlockIdx(reader, ctx->aBlockIdx); + TSDB_CHECK_CODE(code, lino, _exit); + + if (taosArrayGetSize(ctx->aBlockIdx) > 0) { + // init/open file fd + STFile file = { + .type = TSDB_FTYPE_HEAD, + .did = pDFileSet->diskId, + .fid = fset->fid, + .cid = pDFileSet->pHeadF->commitID, + .size = pDFileSet->pHeadF->size, + }; + + code = tsdbTFileObjInit(tsdb, &file, &fset->farr[TSDB_FTYPE_HEAD]); + TSDB_CHECK_CODE(code, lino, _exit); + + // open fd + char fname[TSDB_FILENAME_LEN]; + tsdbTFileName(tsdb, &file, fname); + + code = tsdbOpenFile(fname, ctx->szPage, TD_FILE_READ | TD_FILE_WRITE, &ctx->fd); + TSDB_CHECK_CODE(code, lino, _exit); + + // convert + for (int32_t iBlockIdx = 0; iBlockIdx < taosArrayGetSize(ctx->aBlockIdx); ++iBlockIdx) { + SBlockIdx *pBlockIdx = taosArrayGet(ctx->aBlockIdx, iBlockIdx); + + code = tsdbReadDataBlk(reader, pBlockIdx, ctx->mDataBlk); + TSDB_CHECK_CODE(code, lino, _exit); + + for (int32_t iDataBlk = 0; iDataBlk < ctx->mDataBlk->nItem; ++iDataBlk) { + SDataBlk dataBlk[1]; + tMapDataGetItemByIdx(ctx->mDataBlk, iDataBlk, dataBlk, tGetDataBlk); + + SBrinRecord record = { + .suid = pBlockIdx->suid, + .uid = pBlockIdx->uid, + .firstKey = dataBlk->minKey.ts, + .firstKeyVer = dataBlk->minKey.version, + .lastKey = dataBlk->maxKey.ts, + .lastKeyVer = dataBlk->maxKey.version, + .minVer = dataBlk->minVer, + .maxVer = dataBlk->maxVer, + .blockOffset = dataBlk->aSubBlock->offset, + .smaOffset = dataBlk->smaInfo.offset, + .blockSize = dataBlk->aSubBlock->szBlock, + .blockKeySize = dataBlk->aSubBlock->szKey, + .smaSize = dataBlk->smaInfo.size, + .numRow = dataBlk->nRow, + .count = dataBlk->nRow, + }; + + if (dataBlk->hasDup) { + code = tsdbReadDataBlockEx(reader, dataBlk, ctx->blockData); + TSDB_CHECK_CODE(code, lino, _exit); + + record.count = 1; + for (int32_t i = 1; i < ctx->blockData->nRow; ++i) { + if (ctx->blockData->aTSKEY[i] != ctx->blockData->aTSKEY[i - 1]) { + record.count++; + } + } + } + + code = tBrinBlockPut(ctx->brinBlock, &record); + TSDB_CHECK_CODE(code, lino, _exit); + + if (BRIN_BLOCK_SIZE(ctx->brinBlock) >= ctx->maxRow) { + code = tsdbFileWriteBrinBlock(ctx->fd, ctx->brinBlock, ctx->cmprAlg, &fset->farr[TSDB_FTYPE_HEAD]->f->size, + ctx->brinBlkArray, ctx->bufArr); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + } + + if (BRIN_BLOCK_SIZE(ctx->brinBlock) > 0) { + code = tsdbFileWriteBrinBlock(ctx->fd, ctx->brinBlock, ctx->cmprAlg, &fset->farr[TSDB_FTYPE_HEAD]->f->size, + ctx->brinBlkArray, ctx->bufArr); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbFileWriteBrinBlk(ctx->fd, ctx->brinBlkArray, ctx->footer->brinBlkPtr, + &fset->farr[TSDB_FTYPE_HEAD]->f->size); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbFileWriteHeadFooter(ctx->fd, &fset->farr[TSDB_FTYPE_HEAD]->f->size, ctx->footer); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbFsyncFile(ctx->fd); + TSDB_CHECK_CODE(code, lino, _exit); + + tsdbCloseFile(&ctx->fd); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + TARRAY2_DESTROY(ctx->brinBlkArray, NULL); + tBrinBlockDestroy(ctx->brinBlock); + tBlockDataDestroy(ctx->blockData); + tMapDataClear(ctx->mDataBlk); + taosArrayDestroy(ctx->aBlockIdx); + for (int32_t i = 0; i < ARRAY_SIZE(ctx->bufArr); ++i) { + tFree(ctx->bufArr[i]); + } + return code; +} + +static int32_t tsdbUpgradeData(STsdb *tsdb, SDFileSet *pDFileSet, SDataFReader *reader, STFileSet *fset) { + int32_t code = 0; + int32_t lino = 0; + + if (fset->farr[TSDB_FTYPE_HEAD] == NULL) { + return 0; + } + + STFile file = { + .type = TSDB_FTYPE_DATA, + .did = pDFileSet->diskId, + .fid = fset->fid, + .cid = pDFileSet->pDataF->commitID, + .size = pDFileSet->pDataF->size, + }; + + code = tsdbTFileObjInit(tsdb, &file, &fset->farr[TSDB_FTYPE_DATA]); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbUpgradeSma(STsdb *tsdb, SDFileSet *pDFileSet, SDataFReader *reader, STFileSet *fset) { + int32_t code = 0; + int32_t lino = 0; + + if (fset->farr[TSDB_FTYPE_HEAD] == NULL) { + return 0; + } + + STFile file = { + .type = TSDB_FTYPE_SMA, + .did = pDFileSet->diskId, + .fid = fset->fid, + .cid = pDFileSet->pSmaF->commitID, + .size = pDFileSet->pSmaF->size, + }; + + code = tsdbTFileObjInit(tsdb, &file, &fset->farr[TSDB_FTYPE_SMA]); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbUpgradeSttFile(STsdb *tsdb, SDFileSet *pDFileSet, SDataFReader *reader, STFileSet *fset, + int32_t iStt, SSttLvl *lvl) { + int32_t code = 0; + int32_t lino = 0; + + SArray *aSttBlk = taosArrayInit(0, sizeof(SSttBlk)); + if (aSttBlk == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbReadSttBlk(reader, iStt, aSttBlk); + TSDB_CHECK_CODE(code, lino, _exit); + + if (taosArrayGetSize(aSttBlk) > 0) { + SSttFile *pSttF = pDFileSet->aSttF[iStt]; + STFileObj *fobj; + struct { + int32_t szPage; + // writer + STsdbFD *fd; + TSttBlkArray sttBlkArray[1]; + SSttFooter footer[1]; + } ctx[1] = {{ + .szPage = tsdb->pVnode->config.tsdbPageSize, + }}; + + STFile file = { + .type = TSDB_FTYPE_STT, + .did = pDFileSet->diskId, + .fid = fset->fid, + .cid = pSttF->commitID, + .size = pSttF->size, + }; + code = tsdbTFileObjInit(tsdb, &file, &fobj); + TSDB_CHECK_CODE(code, lino, _exit1); + + code = tsdbOpenFile(fobj->fname, ctx->szPage, TD_FILE_READ | TD_FILE_WRITE, &ctx->fd); + TSDB_CHECK_CODE(code, lino, _exit1); + + for (int32_t iSttBlk = 0; iSttBlk < taosArrayGetSize(aSttBlk); iSttBlk++) { + code = TARRAY2_APPEND_PTR(ctx->sttBlkArray, (SSttBlk *)taosArrayGet(aSttBlk, iSttBlk)); + TSDB_CHECK_CODE(code, lino, _exit1); + } + + code = tsdbFileWriteSttBlk(ctx->fd, ctx->sttBlkArray, ctx->footer->sttBlkPtr, &fobj->f->size); + TSDB_CHECK_CODE(code, lino, _exit1); + + code = tsdbFileWriteSttFooter(ctx->fd, ctx->footer, &fobj->f->size); + TSDB_CHECK_CODE(code, lino, _exit1); + + code = tsdbFsyncFile(ctx->fd); + TSDB_CHECK_CODE(code, lino, _exit1); + + tsdbCloseFile(&ctx->fd); + + code = TARRAY2_APPEND(lvl->fobjArr, fobj); + TSDB_CHECK_CODE(code, lino, _exit1); + + _exit1: + TARRAY2_DESTROY(ctx->sttBlkArray, NULL); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + taosArrayDestroy(aSttBlk); + return code; +} + +static int32_t tsdbUpgradeStt(STsdb *tsdb, SDFileSet *pDFileSet, SDataFReader *reader, STFileSet *fset) { + int32_t code = 0; + int32_t lino = 0; + + if (pDFileSet->nSttF == 0) { + return 0; + } + + SSttLvl *lvl; + code = tsdbSttLvlInit(0, &lvl); + TSDB_CHECK_CODE(code, lino, _exit); + + for (int32_t iStt = 0; iStt < pDFileSet->nSttF; ++iStt) { + code = tsdbUpgradeSttFile(tsdb, pDFileSet, reader, fset, iStt, lvl); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (TARRAY2_SIZE(lvl->fobjArr) > 0) { + code = TARRAY2_APPEND(fset->lvlArr, lvl); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + tsdbSttLvlClear(&lvl); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbUpgradeFileSet(STsdb *tsdb, SDFileSet *pDFileSet, TFileSetArray *fileSetArray) { + int32_t code = 0; + int32_t lino = 0; + + SDataFReader *reader; + STFileSet *fset; + + code = tsdbTFileSetInit(pDFileSet->fid, &fset); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDataFReaderOpen(&reader, tsdb, pDFileSet); + TSDB_CHECK_CODE(code, lino, _exit); + + // .head + code = tsdbUpgradeHead(tsdb, pDFileSet, reader, fset); + TSDB_CHECK_CODE(code, lino, _exit); + + // .data + code = tsdbUpgradeData(tsdb, pDFileSet, reader, fset); + TSDB_CHECK_CODE(code, lino, _exit); + + // .sma + code = tsdbUpgradeSma(tsdb, pDFileSet, reader, fset); + TSDB_CHECK_CODE(code, lino, _exit); + + // .stt + if (pDFileSet->nSttF > 0) { + code = tsdbUpgradeStt(tsdb, pDFileSet, reader, fset); + TSDB_CHECK_CODE(code, lino, _exit); + } + + tsdbDataFReaderClose(&reader); + + code = TARRAY2_APPEND(fileSetArray, fset); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbUpgradeOpenTombFile(STsdb *tsdb, STFileSet *fset, STsdbFD **fd, STFileObj **fobj, bool *toStt) { + int32_t code = 0; + int32_t lino = 0; + + if (TARRAY2_SIZE(fset->lvlArr) == 0) { // to .tomb file + *toStt = false; + + STFile file = { + .type = TSDB_FTYPE_TOMB, + .did = fset->farr[TSDB_FTYPE_HEAD]->f->did, + .fid = fset->fid, + .cid = 0, + .size = 0, + }; + + code = tsdbTFileObjInit(tsdb, &file, fobj); + TSDB_CHECK_CODE(code, lino, _exit); + + fset->farr[TSDB_FTYPE_TOMB] = *fobj; + } else { // to .stt file + *toStt = true; + SSttLvl *lvl = TARRAY2_GET(fset->lvlArr, 0); + + STFile file = { + .type = TSDB_FTYPE_STT, + .did = TARRAY2_GET(lvl->fobjArr, 0)->f->did, + .fid = fset->fid, + .cid = 0, + .size = 0, + }; + + code = tsdbTFileObjInit(tsdb, &file, fobj); + TSDB_CHECK_CODE(code, lino, _exit); + + code = TARRAY2_APPEND(lvl->fobjArr, fobj[0]); + TSDB_CHECK_CODE(code, lino, _exit); + } + + char fname[TSDB_FILENAME_LEN] = {0}; + code = tsdbOpenFile(fobj[0]->fname, tsdb->pVnode->config.tsdbPageSize, + TD_FILE_READ | TD_FILE_WRITE | TD_FILE_TRUNC | TD_FILE_CREATE, fd); + TSDB_CHECK_CODE(code, lino, _exit); + + uint8_t hdr[TSDB_FHDR_SIZE] = {0}; + code = tsdbWriteFile(fd[0], 0, hdr, TSDB_FHDR_SIZE); + TSDB_CHECK_CODE(code, lino, _exit); + fobj[0]->f->size += TSDB_FHDR_SIZE; + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbDumpTombDataToFSet(STsdb *tsdb, SDelFReader *reader, SArray *aDelIdx, STFileSet *fset) { + int32_t code = 0; + int32_t lino = 0; + + struct { + // context + bool toStt; + int8_t cmprAlg; + int32_t maxRow; + int64_t minKey; + int64_t maxKey; + uint8_t *bufArr[8]; + // reader + SArray *aDelData; + // writer + STsdbFD *fd; + STFileObj *fobj; + STombBlock tombBlock[1]; + TTombBlkArray tombBlkArray[1]; + STombFooter tombFooter[1]; + SSttFooter sttFooter[1]; + } ctx[1] = {{ + .maxRow = tsdb->pVnode->config.tsdbCfg.maxRows, + .cmprAlg = tsdb->pVnode->config.tsdbCfg.compression, + }}; + + tsdbFidKeyRange(fset->fid, tsdb->keepCfg.days, tsdb->keepCfg.precision, &ctx->minKey, &ctx->maxKey); + + if ((ctx->aDelData = taosArrayInit(0, sizeof(SDelData))) == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + TSDB_CHECK_CODE(code, lino, _exit); + } + + for (int32_t iDelIdx = 0; iDelIdx < taosArrayGetSize(aDelIdx); iDelIdx++) { + SDelIdx *pDelIdx = (SDelIdx *)taosArrayGet(aDelIdx, iDelIdx); + + code = tsdbReadDelData(reader, pDelIdx, ctx->aDelData); + TSDB_CHECK_CODE(code, lino, _exit); + + for (int32_t iDelData = 0; iDelData < taosArrayGetSize(ctx->aDelData); iDelData++) { + SDelData *pDelData = (SDelData *)taosArrayGet(ctx->aDelData, iDelData); + + STombRecord record = { + .suid = pDelIdx->suid, + .uid = pDelIdx->uid, + .version = pDelData->version, + .skey = pDelData->sKey, + .ekey = pDelData->eKey, + }; + + code = tTombBlockPut(ctx->tombBlock, &record); + TSDB_CHECK_CODE(code, lino, _exit); + + if (TOMB_BLOCK_SIZE(ctx->tombBlock) > ctx->maxRow) { + if (ctx->fd == NULL) { + code = tsdbUpgradeOpenTombFile(tsdb, fset, &ctx->fd, &ctx->fobj, &ctx->toStt); + TSDB_CHECK_CODE(code, lino, _exit); + } + code = tsdbFileWriteTombBlock(ctx->fd, ctx->tombBlock, ctx->cmprAlg, &ctx->fobj->f->size, ctx->tombBlkArray, + ctx->bufArr); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + } + + if (TOMB_BLOCK_SIZE(ctx->tombBlock) > 0) { + if (ctx->fd == NULL) { + code = tsdbUpgradeOpenTombFile(tsdb, fset, &ctx->fd, &ctx->fobj, &ctx->toStt); + TSDB_CHECK_CODE(code, lino, _exit); + } + code = tsdbFileWriteTombBlock(ctx->fd, ctx->tombBlock, ctx->cmprAlg, &ctx->fobj->f->size, ctx->tombBlkArray, + ctx->bufArr); + TSDB_CHECK_CODE(code, lino, _exit); + } + + if (ctx->fd != NULL) { + if (ctx->toStt) { + code = tsdbFileWriteTombBlk(ctx->fd, ctx->tombBlkArray, ctx->sttFooter->tombBlkPtr, &ctx->fobj->f->size); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbFileWriteSttFooter(ctx->fd, ctx->sttFooter, &ctx->fobj->f->size); + TSDB_CHECK_CODE(code, lino, _exit); + } else { + code = tsdbFileWriteTombBlk(ctx->fd, ctx->tombBlkArray, ctx->tombFooter->tombBlkPtr, &ctx->fobj->f->size); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbFileWriteTombFooter(ctx->fd, ctx->tombFooter, &ctx->fobj->f->size); + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbFsyncFile(ctx->fd); + TSDB_CHECK_CODE(code, lino, _exit); + + tsdbCloseFile(&ctx->fd); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + for (int32_t i = 0; i < ARRAY_SIZE(ctx->bufArr); i++) { + tFree(ctx->bufArr[i]); + } + TARRAY2_DESTROY(ctx->tombBlkArray, NULL); + tTombBlockDestroy(ctx->tombBlock); + taosArrayDestroy(ctx->aDelData); + return code; +} + +static int32_t tsdbUpgradeTombFile(STsdb *tsdb, SDelFile *pDelFile, TFileSetArray *fileSetArray) { + int32_t code = 0; + int32_t lino = 0; + + SDelFReader *reader = NULL; + SArray *aDelIdx = NULL; + + if ((aDelIdx = taosArrayInit(0, sizeof(SDelIdx))) == NULL) { + code = TSDB_CODE_OUT_OF_MEMORY; + TSDB_CHECK_CODE(code, lino, _exit); + } + + code = tsdbDelFReaderOpen(&reader, pDelFile, tsdb); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbReadDelIdx(reader, aDelIdx); + TSDB_CHECK_CODE(code, lino, _exit); + + if (taosArrayGetSize(aDelIdx) > 0) { + STFileSet *fset; + TARRAY2_FOREACH(fileSetArray, fset) { + code = tsdbDumpTombDataToFSet(tsdb, reader, aDelIdx, fset); + TSDB_CHECK_CODE(code, lino, _exit); + } + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + tsdbDelFReaderClose(&reader); + taosArrayDestroy(aDelIdx); + return code; +} + +static int32_t tsdbDoUpgradeFileSystem(STsdb *tsdb, TFileSetArray *fileSetArray) { + int32_t code = 0; + int32_t lino = 0; + + // upgrade each file set + for (int32_t i = 0; i < taosArrayGetSize(tsdb->fs.aDFileSet); i++) { + code = tsdbUpgradeFileSet(tsdb, taosArrayGet(tsdb->fs.aDFileSet, i), fileSetArray); + TSDB_CHECK_CODE(code, lino, _exit); + } + + // upgrade tomb file + if (tsdb->fs.pDelFile != NULL) { + code = tsdbUpgradeTombFile(tsdb, tsdb->fs.pDelFile, fileSetArray); + TSDB_CHECK_CODE(code, lino, _exit); + } + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + return code; +} + +static int32_t tsdbUpgradeFileSystem(STsdb *tsdb, int8_t rollback) { + int32_t code = 0; + int32_t lino = 0; + + TFileSetArray fileSetArray[1] = {0}; + + // open old file system + code = tsdbFSOpen(tsdb, rollback); + TSDB_CHECK_CODE(code, lino, _exit); + + code = tsdbDoUpgradeFileSystem(tsdb, fileSetArray); + TSDB_CHECK_CODE(code, lino, _exit); + + // close file system + code = tsdbFSClose(tsdb); + TSDB_CHECK_CODE(code, lino, _exit); + + // save new file system + char fname[TSDB_FILENAME_LEN]; + current_fname(tsdb, fname, TSDB_FCURRENT); + code = save_fs(fileSetArray, fname); + TSDB_CHECK_CODE(code, lino, _exit); + +_exit: + if (code) { + TSDB_ERROR_LOG(TD_VID(tsdb->pVnode), lino, code); + } + TARRAY2_DESTROY(fileSetArray, tsdbTFileSetClear); + return code; +} + +int32_t tsdbCheckAndUpgradeFileSystem(STsdb *tsdb, int8_t rollback) { + char fname[TSDB_FILENAME_LEN]; + + tsdbGetCurrentFName(tsdb, fname, NULL); + if (!taosCheckExistFile(fname)) return 0; + + int32_t code = tsdbUpgradeFileSystem(tsdb, rollback); + if (code) return code; + + taosRemoveFile(fname); + return 0; +} \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbUpgrade.h b/source/dnode/vnode/src/tsdb/tsdbUpgrade.h new file mode 100644 index 0000000000..f9aac94e00 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbUpgrade.h @@ -0,0 +1,35 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdb.h" +#include "tsdbDataFileRW.h" +#include "tsdbDef.h" +#include "tsdbFS2.h" +#include "tsdbUtil2.h" + +#ifndef _TSDB_UPGRADE_H_ +#define _TSDB_UPGRADE_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +int32_t tsdbCheckAndUpgradeFileSystem(STsdb *tsdb, int8_t rollback); + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_UPGRADE_H_*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbUtil2.c b/source/dnode/vnode/src/tsdb/tsdbUtil2.c new file mode 100644 index 0000000000..e938caa118 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbUtil2.c @@ -0,0 +1,191 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "tsdbUtil2.h" + +// SDelBlock ---------- +int32_t tTombBlockInit(STombBlock *tombBlock) { + for (int32_t i = 0; i < TOMB_RECORD_ELEM_NUM; ++i) { + TARRAY2_INIT(&tombBlock->dataArr[i]); + } + return 0; +} + +int32_t tTombBlockDestroy(STombBlock *tombBlock) { + for (int32_t i = 0; i < TOMB_RECORD_ELEM_NUM; ++i) { + TARRAY2_DESTROY(&tombBlock->dataArr[i], NULL); + } + return 0; +} + +int32_t tTombBlockClear(STombBlock *tombBlock) { + for (int32_t i = 0; i < TOMB_RECORD_ELEM_NUM; ++i) { + TARRAY2_CLEAR(&tombBlock->dataArr[i], NULL); + } + return 0; +} + +int32_t tTombBlockPut(STombBlock *tombBlock, const STombRecord *record) { + int32_t code; + for (int32_t i = 0; i < TOMB_RECORD_ELEM_NUM; ++i) { + code = TARRAY2_APPEND(&tombBlock->dataArr[i], record->dataArr[i]); + if (code) return code; + } + return 0; +} + +int32_t tTombBlockGet(STombBlock *tombBlock, int32_t idx, STombRecord *record) { + if (idx >= TOMB_BLOCK_SIZE(tombBlock)) return TSDB_CODE_OUT_OF_RANGE; + for (int32_t i = 0; i < TOMB_RECORD_ELEM_NUM; ++i) { + record->dataArr[i] = TARRAY2_GET(&tombBlock->dataArr[i], idx); + } + return 0; +} + +int32_t tTombRecordCompare(const STombRecord *r1, const STombRecord *r2) { + if (r1->suid < r2->suid) return -1; + if (r1->suid > r2->suid) return 1; + if (r1->uid < r2->uid) return -1; + if (r1->uid > r2->uid) return 1; + if (r1->version < r2->version) return -1; + if (r1->version > r2->version) return 1; + return 0; +} + +// STbStatisBlock ---------- +int32_t tStatisBlockInit(STbStatisBlock *statisBlock) { + for (int32_t i = 0; i < STATIS_RECORD_NUM_ELEM; ++i) { + TARRAY2_INIT(&statisBlock->dataArr[i]); + } + return 0; +} + +int32_t tStatisBlockDestroy(STbStatisBlock *statisBlock) { + for (int32_t i = 0; i < STATIS_RECORD_NUM_ELEM; ++i) { + TARRAY2_DESTROY(&statisBlock->dataArr[i], NULL); + } + return 0; +} + +int32_t tStatisBlockClear(STbStatisBlock *statisBlock) { + for (int32_t i = 0; i < STATIS_RECORD_NUM_ELEM; ++i) { + TARRAY2_CLEAR(&statisBlock->dataArr[i], NULL); + } + return 0; +} + +int32_t tStatisBlockPut(STbStatisBlock *statisBlock, const STbStatisRecord *record) { + int32_t code; + for (int32_t i = 0; i < STATIS_RECORD_NUM_ELEM; ++i) { + code = TARRAY2_APPEND(&statisBlock->dataArr[i], record->dataArr[i]); + if (code) return code; + } + return 0; +} + +int32_t tStatisBlockGet(STbStatisBlock *statisBlock, int32_t idx, STbStatisRecord *record) { + if (idx >= STATIS_BLOCK_SIZE(statisBlock)) return TSDB_CODE_OUT_OF_RANGE; + for (int32_t i = 0; i < STATIS_RECORD_NUM_ELEM; ++i) { + record->dataArr[i] = TARRAY2_GET(&statisBlock->dataArr[i], idx); + } + return 0; +} + +// SBrinRecord ---------- +int32_t tBrinBlockInit(SBrinBlock *brinBlock) { + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr1); ++i) { + TARRAY2_INIT(&brinBlock->dataArr1[i]); + } + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr2); ++i) { + TARRAY2_INIT(&brinBlock->dataArr2[i]); + } + return 0; +} + +int32_t tBrinBlockDestroy(SBrinBlock *brinBlock) { + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr1); ++i) { + TARRAY2_DESTROY(&brinBlock->dataArr1[i], NULL); + } + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr2); ++i) { + TARRAY2_DESTROY(&brinBlock->dataArr2[i], NULL); + } + return 0; +} + +int32_t tBrinBlockClear(SBrinBlock *brinBlock) { + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr1); ++i) { + TARRAY2_CLEAR(&brinBlock->dataArr1[i], NULL); + } + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr2); ++i) { + TARRAY2_CLEAR(&brinBlock->dataArr2[i], NULL); + } + return 0; +} + +int32_t tBrinBlockPut(SBrinBlock *brinBlock, const SBrinRecord *record) { + int32_t code; + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr1); ++i) { + code = TARRAY2_APPEND(&brinBlock->dataArr1[i], record->dataArr1[i]); + if (code) return code; + } + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr2); ++i) { + code = TARRAY2_APPEND(&brinBlock->dataArr2[i], record->dataArr2[i]); + if (code) return code; + } + return 0; +} + +int32_t tBrinBlockGet(SBrinBlock *brinBlock, int32_t idx, SBrinRecord *record) { + if (idx >= BRIN_BLOCK_SIZE(brinBlock)) return TSDB_CODE_OUT_OF_RANGE; + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr1); ++i) { + record->dataArr1[i] = TARRAY2_GET(&brinBlock->dataArr1[i], idx); + } + for (int32_t i = 0; i < ARRAY_SIZE(brinBlock->dataArr2); ++i) { + record->dataArr2[i] = TARRAY2_GET(&brinBlock->dataArr2[i], idx); + } + return 0; +} + +// other apis ---------- +int32_t tsdbUpdateSkmTb(STsdb *pTsdb, const TABLEID *tbid, SSkmInfo *pSkmTb) { + if (tbid->suid) { + if (pSkmTb->suid == tbid->suid) { + pSkmTb->uid = tbid->uid; + return 0; + } + } else if (pSkmTb->uid == tbid->uid) { + return 0; + } + + pSkmTb->suid = tbid->suid; + pSkmTb->uid = tbid->uid; + tDestroyTSchema(pSkmTb->pTSchema); + return metaGetTbTSchemaEx(pTsdb->pVnode->pMeta, tbid->suid, tbid->uid, -1, &pSkmTb->pTSchema); +} + +int32_t tsdbUpdateSkmRow(STsdb *pTsdb, const TABLEID *tbid, int32_t sver, SSkmInfo *pSkmRow) { + if (pSkmRow->pTSchema && pSkmRow->suid == tbid->suid) { + if (pSkmRow->suid) { + if (sver == pSkmRow->pTSchema->version) return 0; + } else if (pSkmRow->uid == tbid->uid && pSkmRow->pTSchema->version == sver) { + return 0; + } + } + + pSkmRow->suid = tbid->suid; + pSkmRow->uid = tbid->uid; + tDestroyTSchema(pSkmRow->pTSchema); + return metaGetTbTSchemaEx(pTsdb->pVnode->pMeta, tbid->suid, tbid->uid, sver, &pSkmRow->pTSchema); +} \ No newline at end of file diff --git a/source/dnode/vnode/src/tsdb/tsdbUtil2.h b/source/dnode/vnode/src/tsdb/tsdbUtil2.h new file mode 100644 index 0000000000..fa06368341 --- /dev/null +++ b/source/dnode/vnode/src/tsdb/tsdbUtil2.h @@ -0,0 +1,193 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#ifndef _TSDB_UTIL_H +#define _TSDB_UTIL_H + +#include "tsdbDef.h" + +#ifdef __cplusplus +extern "C" { +#endif + +// STombRecord ---------- +#define TOMB_RECORD_ELEM_NUM 5 +typedef union { + int64_t dataArr[TOMB_RECORD_ELEM_NUM]; + struct { + int64_t suid; + int64_t uid; + int64_t version; + int64_t skey; + int64_t ekey; + }; +} STombRecord; + +typedef union { + TARRAY2(int64_t) dataArr[TOMB_RECORD_ELEM_NUM]; + struct { + TARRAY2(int64_t) suid[1]; + TARRAY2(int64_t) uid[1]; + TARRAY2(int64_t) version[1]; + TARRAY2(int64_t) skey[1]; + TARRAY2(int64_t) ekey[1]; + }; +} STombBlock; + +typedef struct { + SFDataPtr dp[1]; + TABLEID minTbid; + TABLEID maxTbid; + int64_t minVer; + int64_t maxVer; + int32_t numRec; + int32_t size[TOMB_RECORD_ELEM_NUM]; + int8_t cmprAlg; + int8_t rsvd[7]; +} STombBlk; + +typedef TARRAY2(STombBlk) TTombBlkArray; + +#define TOMB_BLOCK_SIZE(db) TARRAY2_SIZE((db)->suid) + +int32_t tTombBlockInit(STombBlock *tombBlock); +int32_t tTombBlockDestroy(STombBlock *tombBlock); +int32_t tTombBlockClear(STombBlock *tombBlock); +int32_t tTombBlockPut(STombBlock *tombBlock, const STombRecord *record); +int32_t tTombBlockGet(STombBlock *tombBlock, int32_t idx, STombRecord *record); +int32_t tTombRecordCompare(const STombRecord *record1, const STombRecord *record2); + +// STbStatisRecord ---------- +#define STATIS_RECORD_NUM_ELEM 5 +typedef union { + int64_t dataArr[STATIS_RECORD_NUM_ELEM]; + struct { + int64_t suid; + int64_t uid; + int64_t firstKey; + int64_t lastKey; + int64_t count; + }; +} STbStatisRecord; + +typedef union { + TARRAY2(int64_t) dataArr[STATIS_RECORD_NUM_ELEM]; + struct { + TARRAY2(int64_t) suid[1]; + TARRAY2(int64_t) uid[1]; + TARRAY2(int64_t) firstKey[1]; + TARRAY2(int64_t) lastKey[1]; + TARRAY2(int64_t) count[1]; + }; +} STbStatisBlock; + +typedef struct { + SFDataPtr dp[1]; + TABLEID minTbid; + TABLEID maxTbid; + int32_t numRec; + int32_t size[STATIS_RECORD_NUM_ELEM]; + int8_t cmprAlg; + int8_t rsvd[7]; +} SStatisBlk; + +#define STATIS_BLOCK_SIZE(db) TARRAY2_SIZE((db)->suid) + +int32_t tStatisBlockInit(STbStatisBlock *statisBlock); +int32_t tStatisBlockDestroy(STbStatisBlock *statisBlock); +int32_t tStatisBlockClear(STbStatisBlock *statisBlock); +int32_t tStatisBlockPut(STbStatisBlock *statisBlock, const STbStatisRecord *record); +int32_t tStatisBlockGet(STbStatisBlock *statisBlock, int32_t idx, STbStatisRecord *record); + +// SBrinRecord ---------- +typedef union { + struct { + int64_t dataArr1[10]; + int32_t dataArr2[5]; + }; + struct { + int64_t suid; + int64_t uid; + int64_t firstKey; + int64_t firstKeyVer; + int64_t lastKey; + int64_t lastKeyVer; + int64_t minVer; + int64_t maxVer; + int64_t blockOffset; + int64_t smaOffset; + int32_t blockSize; + int32_t blockKeySize; + int32_t smaSize; + int32_t numRow; + int32_t count; + }; +} SBrinRecord; + +typedef union { + struct { + TARRAY2(int64_t) dataArr1[10]; + TARRAY2(int32_t) dataArr2[5]; + }; + struct { + TARRAY2(int64_t) suid[1]; + TARRAY2(int64_t) uid[1]; + TARRAY2(int64_t) firstKey[1]; + TARRAY2(int64_t) firstKeyVer[1]; + TARRAY2(int64_t) lastKey[1]; + TARRAY2(int64_t) lastKeyVer[1]; + TARRAY2(int64_t) minVer[1]; + TARRAY2(int64_t) maxVer[1]; + TARRAY2(int64_t) blockOffset[1]; + TARRAY2(int64_t) smaOffset[1]; + TARRAY2(int32_t) blockSize[1]; + TARRAY2(int32_t) blockKeySize[1]; + TARRAY2(int32_t) smaSize[1]; + TARRAY2(int32_t) numRow[1]; + TARRAY2(int32_t) count[1]; + }; +} SBrinBlock; + +typedef struct { + SFDataPtr dp[1]; + TABLEID minTbid; + TABLEID maxTbid; + int64_t minVer; + int64_t maxVer; + int32_t numRec; + int32_t size[15]; + int8_t cmprAlg; + int8_t rsvd[7]; +} SBrinBlk; + +typedef TARRAY2(SBrinBlk) TBrinBlkArray; + +#define BRIN_BLOCK_SIZE(db) TARRAY2_SIZE((db)->suid) + +int32_t tBrinBlockInit(SBrinBlock *brinBlock); +int32_t tBrinBlockDestroy(SBrinBlock *brinBlock); +int32_t tBrinBlockClear(SBrinBlock *brinBlock); +int32_t tBrinBlockPut(SBrinBlock *brinBlock, const SBrinRecord *record); +int32_t tBrinBlockGet(SBrinBlock *brinBlock, int32_t idx, SBrinRecord *record); + +// other apis +int32_t tsdbUpdateSkmTb(STsdb *pTsdb, const TABLEID *tbid, SSkmInfo *pSkmTb); +int32_t tsdbUpdateSkmRow(STsdb *pTsdb, const TABLEID *tbid, int32_t sver, SSkmInfo *pSkmRow); + +#ifdef __cplusplus +} +#endif + +#endif /*_TSDB_UTIL_H*/ \ No newline at end of file diff --git a/source/dnode/vnode/src/vnd/vnodeCfg.c b/source/dnode/vnode/src/vnd/vnodeCfg.c index efe82e1783..2e161d728f 100644 --- a/source/dnode/vnode/src/vnd/vnodeCfg.c +++ b/source/dnode/vnode/src/vnd/vnodeCfg.c @@ -49,7 +49,7 @@ const SVnodeCfg vnodeCfgDefault = {.vgId = -1, .hashBegin = 0, .hashEnd = 0, .hashMethod = 0, - .sttTrigger = TSDB_DEFAULT_STT_FILE, + .sttTrigger = TSDB_DEFAULT_SST_TRIGGER, .tsdbPageSize = TSDB_DEFAULT_PAGE_SIZE}; int vnodeCheckCfg(const SVnodeCfg *pCfg) { @@ -57,7 +57,7 @@ int vnodeCheckCfg(const SVnodeCfg *pCfg) { return 0; } -const char* vnodeRoleToStr(ESyncRole role) { +const char *vnodeRoleToStr(ESyncRole role) { switch (role) { case TAOS_SYNC_ROLE_VOTER: return "true"; @@ -68,11 +68,11 @@ const char* vnodeRoleToStr(ESyncRole role) { } } -const ESyncRole vnodeStrToRole(char* str) { - if(strcmp(str, "true") == 0){ +const ESyncRole vnodeStrToRole(char *str) { + if (strcmp(str, "true") == 0) { return TAOS_SYNC_ROLE_VOTER; } - if(strcmp(str, "false") == 0){ + if (strcmp(str, "false") == 0) { return TAOS_SYNC_ROLE_LEARNER; } @@ -295,10 +295,9 @@ int vnodeDecodeConfig(const SJson *pJson, void *pObj) { char role[10] = {0}; code = tjsonGetStringValue(info, "isReplica", role); if (code < 0) return -1; - if(strlen(role) != 0){ + if (strlen(role) != 0) { pNode->nodeRole = vnodeStrToRole(role); - } - else{ + } else { pNode->nodeRole = TAOS_SYNC_ROLE_VOTER; } vDebug("vgId:%d, decode config, replica:%d ep:%s:%u dnode:%d", pCfg->vgId, i, pNode->nodeFqdn, pNode->nodePort, diff --git a/source/dnode/vnode/src/vnd/vnodeCommit.c b/source/dnode/vnode/src/vnd/vnodeCommit.c index 7109f8466e..33c6f9d533 100644 --- a/source/dnode/vnode/src/vnd/vnodeCommit.c +++ b/source/dnode/vnode/src/vnd/vnodeCommit.c @@ -16,6 +16,13 @@ #include "vnd.h" #include "vnodeInt.h" +extern int32_t tsdbPreCommit(STsdb *pTsdb); +extern int32_t tsdbCommitBegin(STsdb *pTsdb, SCommitInfo *pInfo); +extern int32_t tsdbCommitCommit(STsdb *pTsdb); +extern int32_t tsdbCommitAbort(STsdb *pTsdb); + +#define VND_INFO_FNAME_TMP "vnode_tmp.json" + static int vnodeEncodeInfo(const SVnodeInfo *pInfo, char **ppData); static int vnodeCommitImpl(SCommitInfo *pInfo); @@ -296,7 +303,7 @@ static int32_t vnodePrepareCommit(SVnode *pVnode, SCommitInfo *pInfo) { TSDB_CHECK_CODE(code, lino, _exit); } - tsdbPrepareCommit(pVnode->pTsdb); + tsdbPreCommit(pVnode->pTsdb); metaPrepareAsyncCommit(pVnode->pMeta); @@ -430,8 +437,7 @@ static int vnodeCommitImpl(SCommitInfo *pInfo) { syncBeginSnapshot(pVnode->sync, pInfo->info.state.committed); - // commit each sub-system - code = tsdbCommit(pVnode->pTsdb, pInfo); + code = tsdbCommitBegin(pVnode->pTsdb, pInfo); TSDB_CHECK_CODE(code, lino, _exit); if (!TSDB_CACHE_NO(pVnode->config)) { @@ -455,7 +461,7 @@ static int vnodeCommitImpl(SCommitInfo *pInfo) { TSDB_CHECK_CODE(code, lino, _exit); } - code = tsdbFinishCommit(pVnode->pTsdb); + code = tsdbCommitCommit(pVnode->pTsdb); TSDB_CHECK_CODE(code, lino, _exit); if (VND_IS_RSMA(pVnode)) { diff --git a/source/dnode/vnode/src/vnd/vnodeInitApi.c b/source/dnode/vnode/src/vnd/vnodeInitApi.c index 28a88561af..e2f0fe8f82 100644 --- a/source/dnode/vnode/src/vnd/vnodeInitApi.c +++ b/source/dnode/vnode/src/vnd/vnodeInitApi.c @@ -42,24 +42,24 @@ void initStorageAPI(SStorageAPI* pAPI) { void initTsdbReaderAPI(TsdReader* pReader) { pReader->tsdReaderOpen = (int32_t(*)(void*, SQueryTableDataCond*, void*, int32_t, SSDataBlock*, void**, const char*, - bool, SHashObj**))tsdbReaderOpen; - pReader->tsdReaderClose = tsdbReaderClose; + bool, SHashObj**))tsdbReaderOpen2; + pReader->tsdReaderClose = tsdbReaderClose2; - pReader->tsdNextDataBlock = tsdbNextDataBlock; + pReader->tsdNextDataBlock = tsdbNextDataBlock2; - pReader->tsdReaderRetrieveDataBlock = tsdbRetrieveDataBlock; - pReader->tsdReaderReleaseDataBlock = tsdbReleaseDataBlock; + pReader->tsdReaderRetrieveDataBlock = tsdbRetrieveDataBlock2; + pReader->tsdReaderReleaseDataBlock = tsdbReleaseDataBlock2; - pReader->tsdReaderRetrieveBlockSMAInfo = tsdbRetrieveDatablockSMA; + pReader->tsdReaderRetrieveBlockSMAInfo = tsdbRetrieveDatablockSMA2; pReader->tsdReaderNotifyClosing = tsdbReaderSetCloseFlag; - pReader->tsdReaderResetStatus = tsdbReaderReset; + pReader->tsdReaderResetStatus = tsdbReaderReset2; - pReader->tsdReaderGetDataBlockDistInfo = tsdbGetFileBlocksDistInfo; - pReader->tsdReaderGetNumOfInMemRows = tsdbGetNumOfRowsInMemTable; // todo this function should be moved away + pReader->tsdReaderGetDataBlockDistInfo = tsdbGetFileBlocksDistInfo2; + pReader->tsdReaderGetNumOfInMemRows = tsdbGetNumOfRowsInMemTable2; // todo this function should be moved away - pReader->tsdSetQueryTableList = tsdbSetTableList; - pReader->tsdSetReaderTaskId = (void (*)(void*, const char*))tsdbReaderSetId; + pReader->tsdSetQueryTableList = tsdbSetTableList2; + pReader->tsdSetReaderTaskId = (void (*)(void*, const char*))tsdbReaderSetId2; } void initMetadataAPI(SStoreMeta* pMeta) { diff --git a/source/dnode/vnode/src/vnd/vnodeModule.c b/source/dnode/vnode/src/vnd/vnodeModule.c index 782ffd788d..74a8d14a86 100644 --- a/source/dnode/vnode/src/vnd/vnodeModule.c +++ b/source/dnode/vnode/src/vnd/vnodeModule.c @@ -23,26 +23,24 @@ struct SVnodeTask { void* arg; }; -struct SVnodeGlobal { - int8_t init; - int8_t stop; +typedef struct { int nthreads; TdThread* threads; TdThreadMutex mutex; TdThreadCond hasTask; SVnodeTask queue; +} SVnodeThreadPool; + +struct SVnodeGlobal { + int8_t init; + int8_t stop; + SVnodeThreadPool tp[2]; }; struct SVnodeGlobal vnodeGlobal; static void* loop(void* arg); -static tsem_t canCommit = {0}; - -static void vnodeInitCommit() { tsem_init(&canCommit, 0, 4); }; -void vnode_wait_commit() { tsem_wait(&canCommit); } -void vnode_done_commit() { tsem_wait(&canCommit); } - int vnodeInit(int nthreads) { int8_t init; int ret; @@ -51,28 +49,30 @@ int vnodeInit(int nthreads) { if (init) { return 0; } - - taosThreadMutexInit(&vnodeGlobal.mutex, NULL); - taosThreadCondInit(&vnodeGlobal.hasTask, NULL); - - taosThreadMutexLock(&vnodeGlobal.mutex); - vnodeGlobal.stop = 0; - vnodeGlobal.queue.next = &vnodeGlobal.queue; - vnodeGlobal.queue.prev = &vnodeGlobal.queue; - taosThreadMutexUnlock(&(vnodeGlobal.mutex)); + for (int32_t i = 0; i < ARRAY_SIZE(vnodeGlobal.tp); i++) { + taosThreadMutexInit(&vnodeGlobal.tp[i].mutex, NULL); + taosThreadCondInit(&vnodeGlobal.tp[i].hasTask, NULL); - vnodeGlobal.nthreads = nthreads; - vnodeGlobal.threads = taosMemoryCalloc(nthreads, sizeof(TdThread)); - if (vnodeGlobal.threads == NULL) { - terrno = TSDB_CODE_OUT_OF_MEMORY; - vError("failed to init vnode module since:%s", tstrerror(terrno)); - return -1; - } + taosThreadMutexLock(&vnodeGlobal.tp[i].mutex); - for (int i = 0; i < nthreads; i++) { - taosThreadCreate(&(vnodeGlobal.threads[i]), NULL, loop, NULL); + vnodeGlobal.tp[i].queue.next = &vnodeGlobal.tp[i].queue; + vnodeGlobal.tp[i].queue.prev = &vnodeGlobal.tp[i].queue; + + taosThreadMutexUnlock(&(vnodeGlobal.tp[i].mutex)); + + vnodeGlobal.tp[i].nthreads = nthreads; + vnodeGlobal.tp[i].threads = taosMemoryCalloc(nthreads, sizeof(TdThread)); + if (vnodeGlobal.tp[i].threads == NULL) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + vError("failed to init vnode module since:%s", tstrerror(terrno)); + return -1; + } + + for (int j = 0; j < nthreads; j++) { + taosThreadCreate(&(vnodeGlobal.tp[i].threads[j]), NULL, loop, &vnodeGlobal.tp[i]); + } } if (walInit() < 0) { @@ -92,27 +92,29 @@ void vnodeCleanup() { if (init == 0) return; // set stop - taosThreadMutexLock(&(vnodeGlobal.mutex)); vnodeGlobal.stop = 1; - taosThreadCondBroadcast(&(vnodeGlobal.hasTask)); - taosThreadMutexUnlock(&(vnodeGlobal.mutex)); + for (int32_t i = 0; i < ARRAY_SIZE(vnodeGlobal.tp); i++) { + taosThreadMutexLock(&(vnodeGlobal.tp[i].mutex)); + taosThreadCondBroadcast(&(vnodeGlobal.tp[i].hasTask)); + taosThreadMutexUnlock(&(vnodeGlobal.tp[i].mutex)); - // wait for threads - for (int i = 0; i < vnodeGlobal.nthreads; i++) { - taosThreadJoin(vnodeGlobal.threads[i], NULL); + // wait for threads + for (int j = 0; j < vnodeGlobal.tp[i].nthreads; j++) { + taosThreadJoin(vnodeGlobal.tp[i].threads[j], NULL); + } + + // clear source + taosMemoryFreeClear(vnodeGlobal.tp[i].threads); + taosThreadCondDestroy(&(vnodeGlobal.tp[i].hasTask)); + taosThreadMutexDestroy(&(vnodeGlobal.tp[i].mutex)); } - // clear source - taosMemoryFreeClear(vnodeGlobal.threads); - taosThreadCondDestroy(&(vnodeGlobal.hasTask)); - taosThreadMutexDestroy(&(vnodeGlobal.mutex)); - walCleanUp(); tqCleanUp(); smaCleanUp(); } -int vnodeScheduleTask(int (*execute)(void*), void* arg) { +int vnodeScheduleTaskEx(int tpid, int (*execute)(void*), void* arg) { SVnodeTask* pTask; ASSERT(!vnodeGlobal.stop); @@ -126,35 +128,42 @@ int vnodeScheduleTask(int (*execute)(void*), void* arg) { pTask->execute = execute; pTask->arg = arg; - taosThreadMutexLock(&(vnodeGlobal.mutex)); - pTask->next = &vnodeGlobal.queue; - pTask->prev = vnodeGlobal.queue.prev; - vnodeGlobal.queue.prev->next = pTask; - vnodeGlobal.queue.prev = pTask; - taosThreadCondSignal(&(vnodeGlobal.hasTask)); - taosThreadMutexUnlock(&(vnodeGlobal.mutex)); + taosThreadMutexLock(&(vnodeGlobal.tp[tpid].mutex)); + pTask->next = &vnodeGlobal.tp[tpid].queue; + pTask->prev = vnodeGlobal.tp[tpid].queue.prev; + vnodeGlobal.tp[tpid].queue.prev->next = pTask; + vnodeGlobal.tp[tpid].queue.prev = pTask; + taosThreadCondSignal(&(vnodeGlobal.tp[tpid].hasTask)); + taosThreadMutexUnlock(&(vnodeGlobal.tp[tpid].mutex)); return 0; } +int vnodeScheduleTask(int (*execute)(void*), void* arg) { return vnodeScheduleTaskEx(0, execute, arg); } + /* ------------------------ STATIC METHODS ------------------------ */ static void* loop(void* arg) { - SVnodeTask* pTask; - int ret; + SVnodeThreadPool* tp = (SVnodeThreadPool*)arg; + SVnodeTask* pTask; + int ret; - setThreadName("vnode-commit"); + if (tp == &vnodeGlobal.tp[0]) { + setThreadName("vnode-commit"); + } else if (tp == &vnodeGlobal.tp[1]) { + setThreadName("vnode-merge"); + } for (;;) { - taosThreadMutexLock(&(vnodeGlobal.mutex)); + taosThreadMutexLock(&(tp->mutex)); for (;;) { - pTask = vnodeGlobal.queue.next; - if (pTask == &vnodeGlobal.queue) { + pTask = tp->queue.next; + if (pTask == &tp->queue) { // no task if (vnodeGlobal.stop) { - taosThreadMutexUnlock(&(vnodeGlobal.mutex)); + taosThreadMutexUnlock(&(tp->mutex)); return NULL; } else { - taosThreadCondWait(&(vnodeGlobal.hasTask), &(vnodeGlobal.mutex)); + taosThreadCondWait(&(tp->hasTask), &(tp->mutex)); } } else { // has task @@ -164,7 +173,7 @@ static void* loop(void* arg) { } } - taosThreadMutexUnlock(&(vnodeGlobal.mutex)); + taosThreadMutexUnlock(&(tp->mutex)); pTask->execute(pTask->arg); taosMemoryFree(pTask); diff --git a/source/dnode/vnode/src/vnd/vnodeRetention.c b/source/dnode/vnode/src/vnd/vnodeRetention.c index f582d5e4a4..7af1f8e28f 100644 --- a/source/dnode/vnode/src/vnd/vnodeRetention.c +++ b/source/dnode/vnode/src/vnd/vnodeRetention.c @@ -15,111 +15,27 @@ #include "vnd.h" -typedef struct { - SVnode *pVnode; - int64_t now; - int64_t commitID; - SVnodeInfo info; -} SRetentionInfo; +extern int32_t tsdbSyncRetention(STsdb *tsdb, int64_t now); +extern int32_t tsdbAsyncRetention(STsdb *tsdb, int64_t now, int64_t *taskid); -extern bool tsdbShouldDoRetention(STsdb *pTsdb, int64_t now); -extern int32_t tsdbDoRetention(STsdb *pTsdb, int64_t now); -extern int32_t tsdbCommitRetention(STsdb *pTsdb); +int32_t vnodeDoRetention(SVnode *pVnode, int64_t now) { + int32_t code; + int32_t lino; -static int32_t vnodePrepareRentention(SVnode *pVnode, SRetentionInfo *pInfo) { - int32_t code = 0; - int32_t lino = 0; - - tsem_wait(&pVnode->canCommit); - - pInfo->commitID = ++pVnode->state.commitID; - - char dir[TSDB_FILENAME_LEN] = {0}; - vnodeGetPrimaryDir(pVnode->path, pVnode->diskPrimary, pVnode->pTfs, dir, TSDB_FILENAME_LEN); - - if (vnodeLoadInfo(dir, &pInfo->info) < 0) { - code = terrno; + if (pVnode->config.sttTrigger == 1) { + tsem_wait(&pVnode->canCommit); + code = tsdbSyncRetention(pVnode->pTsdb, now); TSDB_CHECK_CODE(code, lino, _exit); - } -_exit: - if (code) { - vError("vgId:%d %s failed at line %d since %s", TD_VID(pVnode), __func__, lino, tstrerror(code)); + // code = smaDoRetention(pVnode->pSma, now); + // TSDB_CHECK_CODE(code, lino, _exit); tsem_post(&pVnode->canCommit); } else { - vInfo("vgId:%d %s done", TD_VID(pVnode), __func__); - } - return code; -} - -static int32_t vnodeRetentionTask(void *param) { - int32_t code = 0; - int32_t lino = 0; - - SRetentionInfo *pInfo = (SRetentionInfo *)param; - SVnode *pVnode = pInfo->pVnode; - char dir[TSDB_FILENAME_LEN] = {0}; - - vnodeGetPrimaryDir(pVnode->path, pVnode->diskPrimary, pVnode->pTfs, dir, TSDB_FILENAME_LEN); - - // save info - pInfo->info.state.commitID = pInfo->commitID; - - if (vnodeSaveInfo(dir, &pInfo->info) < 0) { - code = terrno; + int64_t taskid; + code = tsdbAsyncRetention(pVnode->pTsdb, now, &taskid); TSDB_CHECK_CODE(code, lino, _exit); } - // do job - code = tsdbDoRetention(pInfo->pVnode->pTsdb, pInfo->now); - TSDB_CHECK_CODE(code, lino, _exit); - - code = smaDoRetention(pInfo->pVnode->pSma, pInfo->now); - TSDB_CHECK_CODE(code, lino, _exit); - - // commit info - vnodeCommitInfo(dir); - - // commit sub-job - tsdbCommitRetention(pVnode->pTsdb); - _exit: - if (code) { - vError("vgId:%d %s failed at line %d since %s", TD_VID(pInfo->pVnode), __func__, lino, tstrerror(code)); - } else { - vInfo("vgId:%d %s done", TD_VID(pInfo->pVnode), __func__); - } - tsem_post(&pInfo->pVnode->canCommit); - taosMemoryFree(pInfo); return code; -} - -int32_t vnodeAsyncRentention(SVnode *pVnode, int64_t now) { - int32_t code = 0; - int32_t lino = 0; - - if (!tsdbShouldDoRetention(pVnode->pTsdb, now)) return code; - - SRetentionInfo *pInfo = (SRetentionInfo *)taosMemoryCalloc(1, sizeof(*pInfo)); - if (pInfo == NULL) { - code = TSDB_CODE_OUT_OF_MEMORY; - TSDB_CHECK_CODE(code, lino, _exit); - } - - pInfo->pVnode = pVnode; - pInfo->now = now; - - code = vnodePrepareRentention(pVnode, pInfo); - TSDB_CHECK_CODE(code, lino, _exit); - - vnodeScheduleTask(vnodeRetentionTask, pInfo); - -_exit: - if (code) { - vError("vgId:%d %s failed at line %d since %s", TD_VID(pVnode), __func__, lino, tstrerror(code)); - if (pInfo) taosMemoryFree(pInfo); - } else { - vInfo("vgId:%d %s done", TD_VID(pInfo->pVnode), __func__); - } - return 0; -} +} \ No newline at end of file diff --git a/source/dnode/vnode/src/vnd/vnodeSvr.c b/source/dnode/vnode/src/vnd/vnodeSvr.c index ee18653820..ec355947e2 100644 --- a/source/dnode/vnode/src/vnd/vnodeSvr.c +++ b/source/dnode/vnode/src/vnd/vnodeSvr.c @@ -665,9 +665,9 @@ int32_t vnodeProcessStreamMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo) case TDMT_STREAM_TRANSFER_STATE: return tqProcessTaskTransferStateReq(pVnode->pTq, pMsg); case TDMT_STREAM_SCAN_HISTORY_FINISH: - return tqProcessStreamTaskScanHistoryFinishReq(pVnode->pTq, pMsg); + return tqProcessTaskScanHistoryFinishReq(pVnode->pTq, pMsg); case TDMT_STREAM_SCAN_HISTORY_FINISH_RSP: - return tqProcessTaskRecoverFinishRsp(pVnode->pTq, pMsg); + return tqProcessTaskScanHistoryFinishRsp(pVnode->pTq, pMsg); case TDMT_VND_STREAM_CHECK_POINT_SOURCE: return tqProcessStreamCheckPointSourceReq(pVnode->pTq, pMsg); case TDMT_STREAM_TASK_CHECKPOINT_READY: @@ -694,7 +694,8 @@ void vnodeUpdateMetaRsp(SVnode *pVnode, STableMetaRsp *pMetaRsp) { pMetaRsp->precision = pVnode->config.tsdbCfg.precision; } -extern int32_t vnodeAsyncRentention(SVnode *pVnode, int64_t now); +extern int32_t vnodeDoRetention(SVnode *pVnode, int64_t now); + static int32_t vnodeProcessTrimReq(SVnode *pVnode, int64_t ver, void *pReq, int32_t len, SRpcMsg *pRsp) { int32_t code = 0; SVTrimDbReq trimReq = {0}; @@ -707,10 +708,7 @@ static int32_t vnodeProcessTrimReq(SVnode *pVnode, int64_t ver, void *pReq, int3 vInfo("vgId:%d, trim vnode request will be processed, time:%d", pVnode->config.vgId, trimReq.timestamp); - // process - vnodeAsyncRentention(pVnode, trimReq.timestamp); - tsem_wait(&pVnode->canCommit); - tsem_post(&pVnode->canCommit); + code = vnodeDoRetention(pVnode, trimReq.timestamp); _exit: return code; @@ -735,7 +733,7 @@ static int32_t vnodeProcessDropTtlTbReq(SVnode *pVnode, int64_t ver, void *pReq, tqUpdateTbUidList(pVnode->pTq, tbUids, false); } - vnodeAsyncRentention(pVnode, ttlReq.timestampSec); + vnodeDoRetention(pVnode, ttlReq.timestampSec); end: taosArrayDestroy(tbUids); diff --git a/source/libs/catalog/inc/catalogInt.h b/source/libs/catalog/inc/catalogInt.h index 5746ea2340..7d47e82164 100644 --- a/source/libs/catalog/inc/catalogInt.h +++ b/source/libs/catalog/inc/catalogInt.h @@ -938,7 +938,7 @@ int32_t ctgInitJob(SCatalog* pCtg, SRequestConnInfo* pConn, SCtgJob** job, const void* param); int32_t ctgLaunchJob(SCtgJob* pJob); int32_t ctgMakeAsyncRes(SCtgJob* pJob); -int32_t ctgLaunchSubTask(SCtgTask* pTask, CTG_TASK_TYPE type, ctgSubTaskCbFp fp, void* param); +int32_t ctgLaunchSubTask(SCtgTask** ppTask, CTG_TASK_TYPE type, ctgSubTaskCbFp fp, void* param); int32_t ctgGetTbCfgCb(SCtgTask* pTask); void ctgFreeHandle(SCatalog* pCatalog); diff --git a/source/libs/catalog/src/ctgAsync.c b/source/libs/catalog/src/ctgAsync.c index 562343c9c7..fb5ecf7ad2 100644 --- a/source/libs/catalog/src/ctgAsync.c +++ b/source/libs/catalog/src/ctgAsync.c @@ -2090,25 +2090,25 @@ int32_t ctgLaunchGetTbCfgTask(SCtgTask* pTask) { } CTG_CACHE_NHIT_INC(CTG_CI_TBL_CFG, 1); - + if (pCtx->tbType <= 0) { CTG_ERR_JRET(ctgReadTbTypeFromCache(pCtg, dbFName, pCtx->pName->tname, &pCtx->tbType)); if (pCtx->tbType <= 0) { SCtgTbMetaParam param; param.pName = pCtx->pName; param.flag = 0; - CTG_ERR_JRET(ctgLaunchSubTask(pTask, CTG_TASK_GET_TB_META, ctgGetTbCfgCb, ¶m)); + CTG_ERR_JRET(ctgLaunchSubTask(&pTask, CTG_TASK_GET_TB_META, ctgGetTbCfgCb, ¶m)); return TSDB_CODE_SUCCESS; } } - if (TSDB_SUPER_TABLE == pCtx->tbType) { + if (TSDB_SUPER_TABLE == pCtx->tbType || TSDB_SYSTEM_TABLE == pCtx->tbType) { CTG_ERR_JRET(ctgGetTableCfgFromMnode(pCtg, pConn, pCtx->pName, NULL, pTask)); } else { if (NULL == pCtx->pVgInfo) { CTG_ERR_JRET(ctgGetTbHashVgroupFromCache(pCtg, pCtx->pName, &pCtx->pVgInfo)); if (NULL == pCtx->pVgInfo) { - CTG_ERR_JRET(ctgLaunchSubTask(pTask, CTG_TASK_GET_DB_VGROUP, ctgGetTbCfgCb, dbFName)); + CTG_ERR_JRET(ctgLaunchSubTask(&pTask, CTG_TASK_GET_DB_VGROUP, ctgGetTbCfgCb, dbFName)); return TSDB_CODE_SUCCESS; } } @@ -2145,7 +2145,7 @@ int32_t ctgLaunchGetTbTagTask(SCtgTask* pTask) { if (NULL == pCtx->pVgInfo) { CTG_ERR_JRET(ctgGetTbHashVgroupFromCache(pCtg, pCtx->pName, &pCtx->pVgInfo)); if (NULL == pCtx->pVgInfo) { - CTG_ERR_JRET(ctgLaunchSubTask(pTask, CTG_TASK_GET_DB_VGROUP, ctgGetTbTagCb, dbFName)); + CTG_ERR_JRET(ctgLaunchSubTask(&pTask, CTG_TASK_GET_DB_VGROUP, ctgGetTbTagCb, dbFName)); return TSDB_CODE_SUCCESS; } } @@ -2331,7 +2331,7 @@ int32_t ctgLaunchGetUserTask(SCtgTask* pTask) { SCtgTbMetaParam param; param.pName = &pCtx->user.tbName; param.flag = CTG_FLAG_SYNC_OP; - CTG_ERR_RET(ctgLaunchSubTask(pTask, CTG_TASK_GET_TB_META, ctgGetUserCb, ¶m)); + CTG_ERR_RET(ctgLaunchSubTask(&pTask, CTG_TASK_GET_TB_META, ctgGetUserCb, ¶m)); } else { CTG_ERR_RET(ctgGetUserDbAuthFromMnode(pCtg, pConn, pCtx->user.user, NULL, pTask)); } @@ -2541,19 +2541,35 @@ _return: CTG_RET(code); } -int32_t ctgLaunchSubTask(SCtgTask* pTask, CTG_TASK_TYPE type, ctgSubTaskCbFp fp, void* param) { - SCtgJob* pJob = pTask->pJob; +SCtgTask* ctgGetTask(SCtgJob* pJob, int32_t taskId) { + int32_t taskNum = taosArrayGetSize(pJob->pTasks); + + for (int32_t i = 0; i < taskNum; ++i) { + SCtgTask* pTask = taosArrayGet(pJob->pTasks, i); + if (pTask->taskId == taskId) { + return pTask; + } + } + + return NULL; +} + + +int32_t ctgLaunchSubTask(SCtgTask** ppTask, CTG_TASK_TYPE type, ctgSubTaskCbFp fp, void* param) { + SCtgJob* pJob = (*ppTask)->pJob; int32_t subTaskId = -1; bool newTask = false; + int32_t taskId = (*ppTask)->taskId; - ctgClearSubTaskRes(&pTask->subRes); - pTask->subRes.type = type; - pTask->subRes.fp = fp; + ctgClearSubTaskRes(&(*ppTask)->subRes); + (*ppTask)->subRes.type = type; + (*ppTask)->subRes.fp = fp; CTG_ERR_RET(ctgSearchExistingTask(pJob, type, param, &subTaskId)); if (subTaskId < 0) { CTG_ERR_RET(ctgInitTask(pJob, type, param, &subTaskId)); newTask = true; + *ppTask = ctgGetTask(pJob, taskId); } SCtgTask* pSub = taosArrayGet(pJob->pTasks, subTaskId); @@ -2561,10 +2577,10 @@ int32_t ctgLaunchSubTask(SCtgTask* pTask, CTG_TASK_TYPE type, ctgSubTaskCbFp fp, pSub->subTask = true; } - CTG_ERR_RET(ctgSetSubTaskCb(pSub, pTask)); + CTG_ERR_RET(ctgSetSubTaskCb(pSub, *ppTask)); if (newTask) { - SCtgMsgCtx* pMsgCtx = CTG_GET_TASK_MSGCTX(pTask, -1); + SCtgMsgCtx* pMsgCtx = CTG_GET_TASK_MSGCTX(*ppTask, -1); SCtgMsgCtx* pSubMsgCtx = CTG_GET_TASK_MSGCTX(pSub, -1); pSubMsgCtx->pBatchs = pMsgCtx->pBatchs; @@ -2584,6 +2600,7 @@ int32_t ctgLaunchJob(SCtgJob* pJob) { qDebug("QID:0x%" PRIx64 " ctg launch [%dth] task", pJob->queryId, pTask->taskId); CTG_ERR_RET((*gCtgAsyncFps[pTask->type].launchFp)(pTask)); + pTask = taosArrayGet(pJob->pTasks, i); pTask->status = CTG_TASK_LAUNCHED; } diff --git a/source/libs/catalog/src/ctgCache.c b/source/libs/catalog/src/ctgCache.c index 605f5efeb4..44de83b7ef 100644 --- a/source/libs/catalog/src/ctgCache.c +++ b/source/libs/catalog/src/ctgCache.c @@ -773,12 +773,6 @@ int32_t ctgGetCachedStbNameFromSuid(SCatalog* pCtg, char* dbFName, uint64_t suid int32_t ctgChkAuthFromCache(SCatalog *pCtg, SUserAuthInfo *pReq, bool *inCache, SCtgAuthRsp *pRes) { int32_t code = 0; - if (IS_SYS_DBNAME(pReq->tbName.dbname)) { - *inCache = true; - pRes->pRawRes->pass = true; - ctgDebug("sysdb %s, pass", pReq->tbName.dbname); - return TSDB_CODE_SUCCESS; - } SCtgUserAuth *pUser = (SCtgUserAuth *)taosHashGet(pCtg->userCache, pReq->user, strlen(pReq->user)); if (NULL == pUser) { diff --git a/source/libs/catalog/src/ctgUtil.c b/source/libs/catalog/src/ctgUtil.c index 86f6a51d9b..dab007aa47 100644 --- a/source/libs/catalog/src/ctgUtil.c +++ b/source/libs/catalog/src/ctgUtil.c @@ -1589,6 +1589,12 @@ int32_t ctgChkSetAuthRes(SCatalog* pCtg, SCtgAuthReq* req, SCtgAuthRsp* res) { return TSDB_CODE_SUCCESS; } + if (IS_SYS_DBNAME(pReq->tbName.dbname)) { + pRes->pass = true; + ctgDebug("sysdb %s, pass", pReq->tbName.dbname); + return TSDB_CODE_SUCCESS; + } + char dbFName[TSDB_DB_FNAME_LEN]; tNameGetFullDbName(&pReq->tbName, dbFName); diff --git a/source/libs/command/src/command.c b/source/libs/command/src/command.c index 54352d0a53..8ddf730d5a 100644 --- a/source/libs/command/src/command.c +++ b/source/libs/command/src/command.c @@ -802,15 +802,19 @@ static int32_t buildLocalVariablesResultDataBlock(SSDataBlock** pOutput) { pBlock->pDataBlock = taosArrayInit(SHOW_LOCAL_VARIABLES_RESULT_COLS, sizeof(SColumnInfoData)); SColumnInfoData infoData = {0}; + infoData.info.type = TSDB_DATA_TYPE_VARCHAR; infoData.info.bytes = SHOW_LOCAL_VARIABLES_RESULT_FIELD1_LEN; - taosArrayPush(pBlock->pDataBlock, &infoData); infoData.info.type = TSDB_DATA_TYPE_VARCHAR; infoData.info.bytes = SHOW_LOCAL_VARIABLES_RESULT_FIELD2_LEN; taosArrayPush(pBlock->pDataBlock, &infoData); + infoData.info.type = TSDB_DATA_TYPE_VARCHAR; + infoData.info.bytes = SHOW_LOCAL_VARIABLES_RESULT_FIELD3_LEN; + taosArrayPush(pBlock->pDataBlock, &infoData); + *pOutput = pBlock; return TSDB_CODE_SUCCESS; } @@ -823,6 +827,7 @@ int32_t setLocalVariablesResultIntoDataBlock(SSDataBlock* pBlock) { for (int32_t i = 0, c = 0; i < numOfCfg; ++i, c = 0) { SConfigItem* pItem = taosArrayGet(tsCfg->array, i); GRANT_CFG_SKIP; + char name[TSDB_CONFIG_OPTION_LEN + VARSTR_HEADER_SIZE] = {0}; STR_WITH_MAXSIZE_TO_VARSTR(name, pItem->name, TSDB_CONFIG_OPTION_LEN + VARSTR_HEADER_SIZE); SColumnInfoData* pColInfo = taosArrayGet(pBlock->pDataBlock, c++); @@ -835,6 +840,12 @@ int32_t setLocalVariablesResultIntoDataBlock(SSDataBlock* pBlock) { pColInfo = taosArrayGet(pBlock->pDataBlock, c++); colDataSetVal(pColInfo, i, value, false); + char scope[TSDB_CONFIG_SCOPE_LEN + VARSTR_HEADER_SIZE] = {0}; + cfgDumpItemScope(pItem, &scope[VARSTR_HEADER_SIZE], TSDB_CONFIG_SCOPE_LEN, &valueLen); + varDataSetLen(scope, valueLen); + pColInfo = taosArrayGet(pBlock->pDataBlock, c++); + colDataSetVal(pColInfo, i, scope, false); + numOfRows++; } diff --git a/source/libs/executor/inc/executorInt.h b/source/libs/executor/inc/executorInt.h index 9e5626a8d9..b3ce52a0f1 100644 --- a/source/libs/executor/inc/executorInt.h +++ b/source/libs/executor/inc/executorInt.h @@ -232,19 +232,20 @@ typedef struct STableMergeScanInfo { int32_t tableEndIndex; bool hasGroupId; uint64_t groupId; - SArray* queryConds; // array of queryTableDataCond STableScanBase base; int32_t bufPageSize; uint32_t sortBufSize; // max buffer size for in-memory sort SArray* pSortInfo; SSortHandle* pSortHandle; SSDataBlock* pSortInputBlock; + SSDataBlock* pReaderBlock; int64_t startTs; // sort start time SArray* sortSourceParams; SLimitInfo limitInfo; int64_t numOfRows; SScanInfo scanInfo; int32_t scanTimes; + int32_t readIdx; SSDataBlock* pResBlock; SSampleExecInfo sample; // sample execution info SSortExecInfo sortExecInfo; @@ -362,7 +363,6 @@ typedef struct SStreamScanInfo { SNode* pTagIndexCond; // recover - int32_t blockRecoverContiCnt; int32_t blockRecoverTotCnt; SSDataBlock* pRecoverRes; @@ -498,6 +498,7 @@ typedef struct SStreamSessionAggOperatorInfo { STimeWindowAggSupp twAggSup; SSDataBlock* pWinBlock; // window result SSDataBlock* pDelRes; // delete result + SSDataBlock* pUpdateRes; // update window bool returnUpdate; SSHashObj* pStDeleted; void* pDelIterator; diff --git a/source/libs/executor/inc/tsort.h b/source/libs/executor/inc/tsort.h index 627aa825c6..538a9f18f6 100644 --- a/source/libs/executor/inc/tsort.h +++ b/source/libs/executor/inc/tsort.h @@ -26,6 +26,7 @@ extern "C" { enum { SORT_MULTISOURCE_MERGE = 0x1, SORT_SINGLESOURCE_SORT = 0x2, + SORT_BLOCK_TS_MERGE = 0x3 }; typedef struct SMultiMergeSource { @@ -53,6 +54,12 @@ typedef struct SMsortComparParam { int32_t numOfSources; SArray* orderInfo; // SArray bool cmpGroupId; + + int32_t sortType; + // the following field to speed up when sortType == SORT_BLOCK_TS_MERGE + int32_t tsSlotId; + int32_t order; + __compar_fn_t cmpFn; } SMsortComparParam; typedef struct SSortHandle SSortHandle; @@ -70,8 +77,8 @@ typedef int32_t (*_sort_merge_compar_fn_t)(const void* p1, const void* p2, void* * @return */ SSortHandle* tsortCreateSortHandle(SArray* pOrderInfo, int32_t type, int32_t pageSize, int32_t numOfPages, - SSDataBlock* pBlock, const char* idstr, uint64_t maxRows, uint32_t maxTupleLength, - uint32_t sortBufSize); + SSDataBlock* pBlock, const char* idstr, uint64_t pqMaxRows, uint32_t pqMaxTupleLength, + uint32_t pqSortBufSize); void tsortSetForceUsePQSort(SSortHandle* pHandle); @@ -110,6 +117,10 @@ int32_t tsortSetFetchRawDataFp(SSortHandle* pHandle, _sort_fetch_block_fn_t fetc */ int32_t tsortSetComparFp(SSortHandle* pHandle, _sort_merge_compar_fn_t fp); +/** + * +*/ +void tsortSetMergeLimit(SSortHandle* pHandle, int64_t mergeLimit); /** * */ diff --git a/source/libs/executor/src/aggregateoperator.c b/source/libs/executor/src/aggregateoperator.c index be0ad1c239..176c4b53be 100644 --- a/source/libs/executor/src/aggregateoperator.c +++ b/source/libs/executor/src/aggregateoperator.c @@ -45,6 +45,8 @@ typedef struct SAggOperatorInfo { SGroupResInfo groupResInfo; SExprSupp scalarExprSup; bool groupKeyOptimized; + bool hasValidBlock; + SSDataBlock* pNewGroupBlock; } SAggOperatorInfo; static void destroyAggOperatorInfo(void* param); @@ -53,7 +55,6 @@ static void setExecutionContext(SOperatorInfo* pOperator, int32_t numOfOutput, u static int32_t createDataBlockForEmptyInput(SOperatorInfo* pOperator, SSDataBlock** ppBlock); static void destroyDataBlockForEmptyInput(bool blockAllocated, SSDataBlock** ppBlock); -static int32_t doOpenAggregateOptr(SOperatorInfo* pOperator); static int32_t doAggregateImpl(SOperatorInfo* pOperator, SqlFunctionCtx* pCtx); static SSDataBlock* getAggregateResult(SOperatorInfo* pOperator); @@ -111,9 +112,9 @@ SOperatorInfo* createAggregateOperatorInfo(SOperatorInfo* downstream, SAggPhysiN pInfo->binfo.inputTsOrder = pAggNode->node.inputTsOrder; pInfo->binfo.outputTsOrder = pAggNode->node.outputTsOrder; - setOperatorInfo(pOperator, "TableAggregate", QUERY_NODE_PHYSICAL_PLAN_HASH_AGG, true, OP_NOT_OPENED, pInfo, - pTaskInfo); - pOperator->fpSet = createOperatorFpSet(doOpenAggregateOptr, getAggregateResult, NULL, destroyAggOperatorInfo, + setOperatorInfo(pOperator, "TableAggregate", QUERY_NODE_PHYSICAL_PLAN_HASH_AGG, + !pAggNode->node.forceCreateNonBlockingOptr, OP_NOT_OPENED, pInfo, pTaskInfo); + pOperator->fpSet = createOperatorFpSet(optrDummyOpenFn, getAggregateResult, NULL, destroyAggOperatorInfo, optrDefaultBufFn, NULL); if (downstream->operatorType == QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN) { @@ -153,28 +154,42 @@ void destroyAggOperatorInfo(void* param) { taosMemoryFreeClear(param); } -// this is a blocking operator -int32_t doOpenAggregateOptr(SOperatorInfo* pOperator) { - if (OPTR_IS_OPENED(pOperator)) { - return TSDB_CODE_SUCCESS; - } - +/** + * @brief get blocks from downstream and fill results into groupedRes after aggragation + * @retval false if no more groups + * @retval true if there could have new groups coming + * @note if pOperator.blocking is true, scan all blocks from downstream, all groups are handled + * if false, fill results of ONE GROUP + * */ +static bool nextGroupedResult(SOperatorInfo* pOperator) { SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo; SAggOperatorInfo* pAggInfo = pOperator->info; + if (pOperator->blocking && pAggInfo->hasValidBlock) return false; + SExprSupp* pSup = &pOperator->exprSupp; SOperatorInfo* downstream = pOperator->pDownstream[0]; - int64_t st = taosGetTimestampUs(); - int32_t code = TSDB_CODE_SUCCESS; - int32_t order = pAggInfo->binfo.inputTsOrder; - bool hasValidBlock = false; + int64_t st = taosGetTimestampUs(); + int32_t code = TSDB_CODE_SUCCESS; + int32_t order = pAggInfo->binfo.inputTsOrder; + SSDataBlock* pBlock = pAggInfo->pNewGroupBlock; + if (pBlock) { + pAggInfo->pNewGroupBlock = NULL; + tSimpleHashClear(pAggInfo->aggSup.pResultRowHashTable); + setExecutionContext(pOperator, pOperator->exprSupp.numOfExprs, pBlock->info.id.groupId); + setInputDataBlock(pSup, pBlock, order, pBlock->info.scanFlag, true); + code = doAggregateImpl(pOperator, pSup->pCtx); + if (code != TSDB_CODE_SUCCESS) { + T_LONG_JMP(pTaskInfo->env, code); + } + } while (1) { bool blockAllocated = false; - SSDataBlock* pBlock = downstream->fpSet.getNextFn(downstream); + pBlock = downstream->fpSet.getNextFn(downstream); if (pBlock == NULL) { - if (!hasValidBlock) { + if (!pAggInfo->hasValidBlock) { createDataBlockForEmptyInput(pOperator, &pBlock); if (pBlock == NULL) { break; @@ -184,7 +199,7 @@ int32_t doOpenAggregateOptr(SOperatorInfo* pOperator) { break; } } - hasValidBlock = true; + pAggInfo->hasValidBlock = true; pAggInfo->binfo.pRes->info.scanFlag = pBlock->info.scanFlag; // there is an scalar expression that needs to be calculated before apply the group aggregation. @@ -196,7 +211,11 @@ int32_t doOpenAggregateOptr(SOperatorInfo* pOperator) { T_LONG_JMP(pTaskInfo->env, code); } } - + // if non-blocking mode and new group arrived, save the block and break + if (!pOperator->blocking && pAggInfo->groupId != UINT64_MAX && pBlock->info.id.groupId != pAggInfo->groupId) { + pAggInfo->pNewGroupBlock = pBlock; + break; + } // the pDataBlock are always the same one, no need to call this again setExecutionContext(pOperator, pOperator->exprSupp.numOfExprs, pBlock->info.id.groupId); setInputDataBlock(pSup, pBlock, order, pBlock->info.scanFlag, true); @@ -215,10 +234,7 @@ int32_t doOpenAggregateOptr(SOperatorInfo* pOperator) { } initGroupedResultInfo(&pAggInfo->groupResInfo, pAggInfo->aggSup.pResultRowHashTable, 0); - OPTR_SET_OPENED(pOperator); - - pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0; - return pTaskInfo->code; + return pBlock != NULL; } SSDataBlock* getAggregateResult(SOperatorInfo* pOperator) { @@ -230,26 +246,25 @@ SSDataBlock* getAggregateResult(SOperatorInfo* pOperator) { } SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo; - pTaskInfo->code = pOperator->fpSet._openFn(pOperator); - if (pTaskInfo->code != TSDB_CODE_SUCCESS) { - setOperatorCompleted(pOperator); - return NULL; - } + bool hasNewGroups = false; + do { + hasNewGroups = nextGroupedResult(pOperator); + blockDataEnsureCapacity(pInfo->pRes, pOperator->resultInfo.capacity); - blockDataEnsureCapacity(pInfo->pRes, pOperator->resultInfo.capacity); - while (1) { - doBuildResultDatablock(pOperator, pInfo, &pAggInfo->groupResInfo, pAggInfo->aggSup.pResultBuf); - doFilter(pInfo->pRes, pOperator->exprSupp.pFilterInfo, NULL); + while (1) { + doBuildResultDatablock(pOperator, pInfo, &pAggInfo->groupResInfo, pAggInfo->aggSup.pResultBuf); + doFilter(pInfo->pRes, pOperator->exprSupp.pFilterInfo, NULL); - if (!hasRemainResults(&pAggInfo->groupResInfo)) { - setOperatorCompleted(pOperator); - break; + if (!hasRemainResults(&pAggInfo->groupResInfo)) { + if (!hasNewGroups) setOperatorCompleted(pOperator); + break; + } + + if (pInfo->pRes->info.rows > 0) { + break; + } } - - if (pInfo->pRes->info.rows > 0) { - break; - } - } + } while (pInfo->pRes->info.rows == 0 && hasNewGroups); size_t rows = blockDataGetNumOfRows(pInfo->pRes); pOperator->resultInfo.totalRows += rows; diff --git a/source/libs/executor/src/executil.c b/source/libs/executor/src/executil.c index 832750e967..e1bf4e7cb0 100644 --- a/source/libs/executor/src/executil.c +++ b/source/libs/executor/src/executil.c @@ -127,6 +127,10 @@ void initGroupedResultInfo(SGroupResInfo* pGroupResInfo, SSHashObj* pHashmap, in if (pGroupResInfo->pRows != NULL) { taosArrayDestroy(pGroupResInfo->pRows); } + if (pGroupResInfo->pBuf) { + taosMemoryFree(pGroupResInfo->pBuf); + pGroupResInfo->pBuf = NULL; + } // extract the result rows information from the hash map int32_t size = tSimpleHashGetSize(pHashmap); @@ -2104,6 +2108,8 @@ int32_t buildGroupIdMapForAllTables(STableListInfo* pTableListInfo, SReadHandle* if (groupSort && groupByTbname) { taosArraySort(pTableListInfo->pTableList, orderbyGroupIdComparFn); pTableListInfo->numOfOuputGroups = numOfTables; + } else if (groupByTbname && pScanNode->groupOrderScan){ + pTableListInfo->numOfOuputGroups = numOfTables; } else { pTableListInfo->numOfOuputGroups = 1; } diff --git a/source/libs/executor/src/executor.c b/source/libs/executor/src/executor.c index 75e15876cc..4335624486 100644 --- a/source/libs/executor/src/executor.c +++ b/source/libs/executor/src/executor.c @@ -122,7 +122,7 @@ void qResetStreamInfoTimeWindow(qTaskInfo_t tinfo) { return; } - qDebug("%s set fill history start key:%"PRId64, GET_TASKID(pTaskInfo), INT64_MIN); + qDebug("%s set fill history start key:%" PRId64, GET_TASKID(pTaskInfo), INT64_MIN); pTaskInfo->streamInfo.fillHistoryWindow.skey = INT64_MIN; } diff --git a/source/libs/executor/src/scanoperator.c b/source/libs/executor/src/scanoperator.c index 3b75292556..35b37b860c 100644 --- a/source/libs/executor/src/scanoperator.c +++ b/source/libs/executor/src/scanoperator.c @@ -13,8 +13,6 @@ * along with this program. If not, see . */ -// clang-format off - #include "executorInt.h" #include "filter.h" #include "function.h" @@ -56,8 +54,7 @@ typedef struct STableMergeScanSortSourceParam { SOperatorInfo* pOperator; int32_t readerIdx; uint64_t uid; - SSDataBlock* inputBlock; - STsdbReader* dataReader; + STsdbReader* reader; } STableMergeScanSortSourceParam; typedef struct STableCountScanOperatorInfo { @@ -1841,9 +1838,7 @@ static SSDataBlock* doStreamScan(SOperatorInfo* pOperator) { if (pStreamInfo->recoverStep == STREAM_RECOVER_STEP__SCAN1 || pStreamInfo->recoverStep == STREAM_RECOVER_STEP__SCAN2) { - if (pInfo->blockRecoverContiCnt > 100) { - pInfo->blockRecoverTotCnt += pInfo->blockRecoverContiCnt; - pInfo->blockRecoverContiCnt = 0; + if (isTaskKilled(pTaskInfo)) { return NULL; } @@ -1888,7 +1883,6 @@ static SSDataBlock* doStreamScan(SOperatorInfo* pOperator) { pInfo->pRecoverRes = doTableScan(pInfo->pTableScanOp); if (pInfo->pRecoverRes != NULL) { - pInfo->blockRecoverContiCnt++; calBlockTbName(pInfo, pInfo->pRecoverRes); if (!pInfo->igCheckUpdate && pInfo->pUpdateInfo) { if (pStreamInfo->recoverStep == STREAM_RECOVER_STEP__SCAN1) { @@ -2781,32 +2775,17 @@ static SSDataBlock* getTableDataBlockImpl(void* param) { SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo; SStorageAPI* pAPI = &pTaskInfo->storageAPI; - int32_t readIdx = source->readerIdx; - SSDataBlock* pBlock = source->inputBlock; + SSDataBlock* pBlock = pInfo->pReaderBlock; int32_t code = 0; - SQueryTableDataCond* pQueryCond = taosArrayGet(pInfo->queryConds, readIdx); - int64_t st = taosGetTimestampUs(); - void* p = tableListGetInfo(pInfo->base.pTableListInfo, readIdx + pInfo->tableStartIndex); - SReadHandle* pHandle = &pInfo->base.readHandle; - if (NULL == source->dataReader) { - code = pAPI->tsdReader.tsdReaderOpen(pHandle->vnode, pQueryCond, p, 1, pBlock, (void**)&source->dataReader, GET_TASKID(pTaskInfo), false, NULL); - if (code != 0) { - T_LONG_JMP(pTaskInfo->env, code); - } - } - - pInfo->base.dataReader = source->dataReader; - STsdbReader* reader = pInfo->base.dataReader; bool hasNext = false; - qTrace("tsdb/read-table-data: %p, enter next reader", reader); + STsdbReader* reader = pInfo->base.dataReader; while (true) { code = pAPI->tsdReader.tsdNextDataBlock(reader, &hasNext); if (code != 0) { pAPI->tsdReader.tsdReaderReleaseDataBlock(reader); - pInfo->base.dataReader = NULL; T_LONG_JMP(pTaskInfo->env, code); } @@ -2816,7 +2795,6 @@ static SSDataBlock* getTableDataBlockImpl(void* param) { if (isTaskKilled(pTaskInfo)) { pAPI->tsdReader.tsdReaderReleaseDataBlock(reader); - pInfo->base.dataReader = NULL; T_LONG_JMP(pTaskInfo->env, pTaskInfo->code); } @@ -2826,12 +2804,6 @@ static SSDataBlock* getTableDataBlockImpl(void* param) { continue; } - if (pQueryCond->order == TSDB_ORDER_ASC) { - pQueryCond->twindows.skey = pBlock->info.window.ekey + 1; - } else { - pQueryCond->twindows.ekey = pBlock->info.window.skey - 1; - } - uint32_t status = 0; code = loadDataBlock(pOperator, &pInfo->base, pBlock, &status); // code = loadDataBlockFromOneTable(pOperator, pTableScanInfo, pBlock, &status); @@ -2853,16 +2825,9 @@ static SSDataBlock* getTableDataBlockImpl(void* param) { pOperator->resultInfo.totalRows += pBlock->info.rows; pInfo->base.readRecorder.elapsedTime += (taosGetTimestampUs() - st) / 1000.0; - qTrace("tsdb/read-table-data: %p, close reader", reader); - pInfo->base.dataReader = NULL; return pBlock; } - pAPI->tsdReader.tsdReaderClose(source->dataReader); - source->dataReader = NULL; - pInfo->base.dataReader = NULL; - blockDataDestroy(source->inputBlock); - source->inputBlock = NULL; return NULL; } @@ -2898,6 +2863,8 @@ int32_t dumpQueryTableCond(const SQueryTableDataCond* src, SQueryTableDataCond* int32_t startGroupTableMergeScan(SOperatorInfo* pOperator) { STableMergeScanInfo* pInfo = pOperator->info; SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo; + SReadHandle* pHandle = &pInfo->base.readHandle; + SStorageAPI* pAPI = &pTaskInfo->storageAPI; { size_t numOfTables = tableListGetSize(pInfo->base.pTableListInfo); @@ -2914,53 +2881,29 @@ int32_t startGroupTableMergeScan(SOperatorInfo* pOperator) { int32_t tableStartIdx = pInfo->tableStartIndex; int32_t tableEndIdx = pInfo->tableEndIndex; - pInfo->base.dataReader = NULL; - - // todo the total available buffer should be determined by total capacity of buffer of this task. - // the additional one is reserved for merge result - // pInfo->sortBufSize = pInfo->bufPageSize * (tableEndIdx - tableStartIdx + 1 + 1); - int32_t kWay = (TSDB_MAX_BYTES_PER_ROW * 2) / (pInfo->pResBlock->info.rowSize); - if (kWay >= 128) { - kWay = 128; - } else if (kWay <= 2) { - kWay = 2; - } else { - int i = 2; - while (i * 2 <= kWay) i = i * 2; - kWay = i; - } - - pInfo->sortBufSize = pInfo->bufPageSize * (kWay + 1); + pInfo->sortBufSize = 2048 * pInfo->bufPageSize; int32_t numOfBufPage = pInfo->sortBufSize / pInfo->bufPageSize; - pInfo->pSortHandle = tsortCreateSortHandle(pInfo->pSortInfo, SORT_MULTISOURCE_MERGE, pInfo->bufPageSize, numOfBufPage, + pInfo->pSortHandle = tsortCreateSortHandle(pInfo->pSortInfo, SORT_BLOCK_TS_MERGE, pInfo->bufPageSize, numOfBufPage, pInfo->pSortInputBlock, pTaskInfo->id.str, 0, 0, 0); - + int64_t mergeLimit = -1; + if (pInfo->limitInfo.limit.limit != -1 || pInfo->limitInfo.limit.offset != -1) { + mergeLimit = pInfo->limitInfo.limit.limit + pInfo->limitInfo.limit.offset; + } + tsortSetMergeLimit(pInfo->pSortHandle, mergeLimit); tsortSetFetchRawDataFp(pInfo->pSortHandle, getTableDataBlockImpl, NULL, NULL); // one table has one data block int32_t numOfTable = tableEndIdx - tableStartIdx + 1; - pInfo->queryConds = taosArrayInit(numOfTable, sizeof(SQueryTableDataCond)); - for (int32_t i = 0; i < numOfTable; ++i) { - STableMergeScanSortSourceParam param = {0}; - param.readerIdx = i; - param.pOperator = pOperator; - param.inputBlock = createOneDataBlock(pInfo->pResBlock, false); + STableMergeScanSortSourceParam param = {0}; + param.pOperator = pOperator; + STableKeyInfo* startKeyInfo = tableListGetInfo(pInfo->base.pTableListInfo, tableStartIdx); + pAPI->tsdReader.tsdReaderOpen(pHandle->vnode, &pInfo->base.cond, startKeyInfo, numOfTable, pInfo->pReaderBlock, (void**)&pInfo->base.dataReader, GET_TASKID(pTaskInfo), false, NULL); - taosArrayPush(pInfo->sortSourceParams, ¶m); - - SQueryTableDataCond cond; - dumpQueryTableCond(&pInfo->base.cond, &cond); - taosArrayPush(pInfo->queryConds, &cond); - } - - for (int32_t i = 0; i < numOfTable; ++i) { - SSortSource* ps = taosMemoryCalloc(1, sizeof(SSortSource)); - STableMergeScanSortSourceParam* param = taosArrayGet(pInfo->sortSourceParams, i); - ps->param = param; - ps->onlyRef = true; - tsortAddSource(pInfo->pSortHandle, ps); - } + SSortSource* ps = taosMemoryCalloc(1, sizeof(SSortSource)); + ps->param = ¶m; + ps->onlyRef = true; + tsortAddSource(pInfo->pSortHandle, ps); int32_t code = tsortOpen(pInfo->pSortHandle); @@ -2976,8 +2919,6 @@ int32_t stopGroupTableMergeScan(SOperatorInfo* pOperator) { SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo; SStorageAPI* pAPI = &pTaskInfo->storageAPI; - int32_t numOfTable = taosArrayGetSize(pInfo->queryConds); - SSortExecInfo sortExecInfo = tsortGetSortExecInfo(pInfo->pSortHandle); pInfo->sortExecInfo.sortMethod = sortExecInfo.sortMethod; pInfo->sortExecInfo.sortBuffer = sortExecInfo.sortBuffer; @@ -2985,24 +2926,14 @@ int32_t stopGroupTableMergeScan(SOperatorInfo* pOperator) { pInfo->sortExecInfo.readBytes += sortExecInfo.readBytes; pInfo->sortExecInfo.writeBytes += sortExecInfo.writeBytes; - for (int32_t i = 0; i < numOfTable; ++i) { - STableMergeScanSortSourceParam* param = taosArrayGet(pInfo->sortSourceParams, i); - blockDataDestroy(param->inputBlock); - pAPI->tsdReader.tsdReaderClose(param->dataReader); - param->dataReader = NULL; + if (pInfo->base.dataReader != NULL) { + pAPI->tsdReader.tsdReaderClose(pInfo->base.dataReader); + pInfo->base.dataReader = NULL; } - taosArrayClear(pInfo->sortSourceParams); tsortDestroySortHandle(pInfo->pSortHandle); pInfo->pSortHandle = NULL; - for (int32_t i = 0; i < taosArrayGetSize(pInfo->queryConds); i++) { - SQueryTableDataCond* cond = taosArrayGet(pInfo->queryConds, i); - taosMemoryFree(cond->colList); - } - taosArrayDestroy(pInfo->queryConds); - pInfo->queryConds = NULL; - resetLimitInfoForNextGroup(&pInfo->limitInfo); return TSDB_CODE_SUCCESS; } @@ -3015,28 +2946,32 @@ SSDataBlock* getSortedTableMergeScanBlockData(SSortHandle* pHandle, SSDataBlock* SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo; blockDataCleanup(pResBlock); - + STupleHandle* pTupleHandle = NULL; while (1) { - STupleHandle* pTupleHandle = tsortNextTuple(pHandle); - if (pTupleHandle == NULL) { - break; + while (1) { + pTupleHandle = tsortNextTuple(pHandle); + if (pTupleHandle == NULL) { + break; + } + + appendOneRowToDataBlock(pResBlock, pTupleHandle); + if (pResBlock->info.rows >= capacity) { + break; + } } - appendOneRowToDataBlock(pResBlock, pTupleHandle); - if (pResBlock->info.rows >= capacity) { - break; + if (tsortIsClosed(pHandle)) { + terrno = TSDB_CODE_TSC_QUERY_CANCELLED; + T_LONG_JMP(pOperator->pTaskInfo->env, terrno); } + + bool limitReached = applyLimitOffset(&pInfo->limitInfo, pResBlock, pTaskInfo); + qDebug("%s get sorted row block, rows:%" PRId64 ", limit:%" PRId64, GET_TASKID(pTaskInfo), pResBlock->info.rows, + pInfo->limitInfo.numOfOutputRows); + if (pTupleHandle == NULL || limitReached || pResBlock->info.rows > 0) { + break; + } } - - if (tsortIsClosed(pHandle)) { - terrno = TSDB_CODE_TSC_QUERY_CANCELLED; - T_LONG_JMP(pOperator->pTaskInfo->env, terrno); - } - - bool limitReached = applyLimitOffset(&pInfo->limitInfo, pResBlock, pTaskInfo); - qDebug("%s get sorted row block, rows:%" PRId64 ", limit:%" PRId64, GET_TASKID(pTaskInfo), pResBlock->info.rows, - pInfo->limitInfo.numOfOutputRows); - return (pResBlock->info.rows > 0) ? pResBlock : NULL; } @@ -3100,14 +3035,7 @@ void destroyTableMergeScanOperatorInfo(void* param) { STableMergeScanInfo* pTableScanInfo = (STableMergeScanInfo*)param; cleanupQueryTableDataCond(&pTableScanInfo->base.cond); - int32_t numOfTable = taosArrayGetSize(pTableScanInfo->queryConds); - - for (int32_t i = 0; i < numOfTable; i++) { - STableMergeScanSortSourceParam* p = taosArrayGet(pTableScanInfo->sortSourceParams, i); - blockDataDestroy(p->inputBlock); - pTableScanInfo->base.readerAPI.tsdReaderClose(p->dataReader); - p->dataReader = NULL; - } + int32_t numOfTable = taosArrayGetSize(pTableScanInfo->sortSourceParams); pTableScanInfo->base.readerAPI.tsdReaderClose(pTableScanInfo->base.dataReader); pTableScanInfo->base.dataReader = NULL; @@ -3116,16 +3044,11 @@ void destroyTableMergeScanOperatorInfo(void* param) { tsortDestroySortHandle(pTableScanInfo->pSortHandle); pTableScanInfo->pSortHandle = NULL; - for (int i = 0; i < taosArrayGetSize(pTableScanInfo->queryConds); i++) { - SQueryTableDataCond* pCond = taosArrayGet(pTableScanInfo->queryConds, i); - taosMemoryFree(pCond->colList); - } - - taosArrayDestroy(pTableScanInfo->queryConds); destroyTableScanBase(&pTableScanInfo->base, &pTableScanInfo->base.readerAPI); pTableScanInfo->pResBlock = blockDataDestroy(pTableScanInfo->pResBlock); pTableScanInfo->pSortInputBlock = blockDataDestroy(pTableScanInfo->pSortInputBlock); + pTableScanInfo->pReaderBlock = blockDataDestroy(pTableScanInfo->pReaderBlock); taosArrayDestroy(pTableScanInfo->pSortInfo); taosMemoryFreeClear(param); @@ -3187,6 +3110,8 @@ SOperatorInfo* createTableMergeScanOperatorInfo(STableScanPhysiNode* pTableScanN pInfo->base.scanFlag = MAIN_SCAN; pInfo->base.readHandle = *readHandle; + pInfo->readIdx = -1; + pInfo->base.limitInfo.limit.limit = -1; pInfo->base.limitInfo.slimit.limit = -1; pInfo->base.pTableListInfo = pTableListInfo; @@ -3209,6 +3134,8 @@ SOperatorInfo* createTableMergeScanOperatorInfo(STableScanPhysiNode* pTableScanN pInfo->pSortInputBlock = createOneDataBlock(pInfo->pResBlock, false); initLimitInfo(pTableScanNode->scan.node.pLimit, pTableScanNode->scan.node.pSlimit, &pInfo->limitInfo); + pInfo->pReaderBlock = createOneDataBlock(pInfo->pResBlock, false); + int32_t rowSize = pInfo->pResBlock->info.rowSize; uint32_t nCols = taosArrayGetSize(pInfo->pResBlock->pDataBlock); pInfo->bufPageSize = getProperSortPageSize(rowSize, nCols); @@ -3617,6 +3544,4 @@ static void destoryTableCountScanOperator(void* param) { taosArrayDestroy(pTableCountScanInfo->stbUidList); taosMemoryFreeClear(param); -} - -// clang-format on +} \ No newline at end of file diff --git a/source/libs/executor/src/tfill.c b/source/libs/executor/src/tfill.c index 55ef019d76..4e0dff9d4f 100644 --- a/source/libs/executor/src/tfill.c +++ b/source/libs/executor/src/tfill.c @@ -66,20 +66,25 @@ static void setNullRow(SSDataBlock* pBlock, SFillInfo* pFillInfo, int32_t rowInd } static void doSetUserSpecifiedValue(SColumnInfoData* pDst, SVariant* pVar, int32_t rowIndex, int64_t currentKey) { + bool isNull = (TSDB_DATA_TYPE_NULL == pVar->nType) ? true : false; if (pDst->info.type == TSDB_DATA_TYPE_FLOAT) { float v = 0; - GET_TYPED_DATA(v, float, pVar->nType, &pVar->i); - colDataSetVal(pDst, rowIndex, (char*)&v, false); + GET_TYPED_DATA(v, float, pVar->nType, &pVar->f); + colDataSetVal(pDst, rowIndex, (char*)&v, isNull); } else if (pDst->info.type == TSDB_DATA_TYPE_DOUBLE) { double v = 0; - GET_TYPED_DATA(v, double, pVar->nType, &pVar->i); - colDataSetVal(pDst, rowIndex, (char*)&v, false); + GET_TYPED_DATA(v, double, pVar->nType, &pVar->d); + colDataSetVal(pDst, rowIndex, (char*)&v, isNull); } else if (IS_SIGNED_NUMERIC_TYPE(pDst->info.type)) { int64_t v = 0; GET_TYPED_DATA(v, int64_t, pVar->nType, &pVar->i); - colDataSetVal(pDst, rowIndex, (char*)&v, false); + colDataSetVal(pDst, rowIndex, (char*)&v, isNull); + } else if (IS_UNSIGNED_NUMERIC_TYPE(pDst->info.type)) { + uint64_t v = 0; + GET_TYPED_DATA(v, uint64_t, pVar->nType, &pVar->u); + colDataSetVal(pDst, rowIndex, (char*)&v, isNull); } else if (pDst->info.type == TSDB_DATA_TYPE_TIMESTAMP) { - colDataSetVal(pDst, rowIndex, (const char*)¤tKey, false); + colDataSetVal(pDst, rowIndex, (const char*)¤tKey, isNull); } else { // varchar/nchar data colDataSetNULL(pDst, rowIndex); } diff --git a/source/libs/executor/src/timesliceoperator.c b/source/libs/executor/src/timesliceoperator.c index cb74392a10..b019985645 100644 --- a/source/libs/executor/src/timesliceoperator.c +++ b/source/libs/executor/src/timesliceoperator.c @@ -312,6 +312,7 @@ static bool genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp case TSDB_FILL_SET_VALUE_F: { SVariant* pVar = &pSliceInfo->pFillColInfo[fillColIndex].fillVal; + bool isNull = (TSDB_DATA_TYPE_NULL == pVar->nType) ? true : false; if (pDst->info.type == TSDB_DATA_TYPE_FLOAT) { float v = 0; if (!IS_VAR_DATA_TYPE(pVar->nType)) { @@ -319,7 +320,7 @@ static bool genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp } else { v = taosStr2Float(varDataVal(pVar->pz), NULL); } - colDataSetVal(pDst, rows, (char*)&v, false); + colDataSetVal(pDst, rows, (char*)&v, isNull); } else if (pDst->info.type == TSDB_DATA_TYPE_DOUBLE) { double v = 0; if (!IS_VAR_DATA_TYPE(pVar->nType)) { @@ -327,7 +328,7 @@ static bool genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp } else { v = taosStr2Double(varDataVal(pVar->pz), NULL); } - colDataSetVal(pDst, rows, (char*)&v, false); + colDataSetVal(pDst, rows, (char*)&v, isNull); } else if (IS_SIGNED_NUMERIC_TYPE(pDst->info.type)) { int64_t v = 0; if (!IS_VAR_DATA_TYPE(pVar->nType)) { @@ -335,7 +336,7 @@ static bool genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp } else { v = taosStr2Int64(varDataVal(pVar->pz), NULL, 10); } - colDataSetVal(pDst, rows, (char*)&v, false); + colDataSetVal(pDst, rows, (char*)&v, isNull); } else if (IS_UNSIGNED_NUMERIC_TYPE(pDst->info.type)) { uint64_t v = 0; if (!IS_VAR_DATA_TYPE(pVar->nType)) { @@ -343,7 +344,7 @@ static bool genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp } else { v = taosStr2UInt64(varDataVal(pVar->pz), NULL, 10); } - colDataSetVal(pDst, rows, (char*)&v, false); + colDataSetVal(pDst, rows, (char*)&v, isNull); } else if (IS_BOOLEAN_TYPE(pDst->info.type)) { bool v = false; if (!IS_VAR_DATA_TYPE(pVar->nType)) { @@ -351,7 +352,7 @@ static bool genInterpolationResult(STimeSliceOperatorInfo* pSliceInfo, SExprSupp } else { v = taosStr2Int8(varDataVal(pVar->pz), NULL, 10); } - colDataSetVal(pDst, rows, (char*)&v, false); + colDataSetVal(pDst, rows, (char*)&v, isNull); } ++fillColIndex; diff --git a/source/libs/executor/src/timewindowoperator.c b/source/libs/executor/src/timewindowoperator.c index 63bcaa2804..d5fd562691 100644 --- a/source/libs/executor/src/timewindowoperator.c +++ b/source/libs/executor/src/timewindowoperator.c @@ -2974,6 +2974,7 @@ void streamIntervalReloadState(SOperatorInfo* pOperator) { strlen(STREAM_INTERVAL_OP_STATE_NAME), &pBuf, &size); TSKEY ts = *(TSKEY*)pBuf; taosMemoryFree(pBuf); + pInfo->twAggSup.maxTs = TMAX(pInfo->twAggSup.maxTs, ts); pInfo->stateStore.streamStateReloadInfo(pInfo->pState, ts); } SOperatorInfo* downstream = pOperator->pDownstream[0]; @@ -3143,6 +3144,8 @@ void destroyStreamSessionAggOperatorInfo(void* param) { colDataDestroy(&pInfo->twAggSup.timeWindowData); blockDataDestroy(pInfo->pDelRes); blockDataDestroy(pInfo->pWinBlock); + blockDataDestroy(pInfo->pUpdateRes); + tSimpleHashCleanup(pInfo->pStUpdated); tSimpleHashCleanup(pInfo->pStDeleted); taosArrayDestroy(pInfo->historyWins); @@ -3240,14 +3243,6 @@ int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, SqlFunctionCtx* pCtx, pCtx[i].saveHandle.pBuf = pSup->pResultBuf; } - if (pHandle) { - pSup->winRange = pHandle->winRange; - // temporary - if (pSup->winRange.ekey <= 0) { - pSup->winRange.ekey = INT64_MAX; - } - } - pSup->pSessionAPI = pApi; return TSDB_CODE_SUCCESS; @@ -3295,11 +3290,12 @@ void setSessionOutputBuf(SStreamAggSupporter* pAggSup, TSKEY startTs, TSKEY endT if (code == TSDB_CODE_SUCCESS && !inWinRange(&pAggSup->winRange, &pCurWin->sessionWin.win)) { code = TSDB_CODE_FAILED; releaseOutputBuf(pAggSup->pState, NULL, (SResultRow*)pCurWin->pOutputBuf, &pAggSup->pSessionAPI->stateStore); - pCurWin->pOutputBuf = taosMemoryMalloc(size); + pCurWin->pOutputBuf = taosMemoryCalloc(1, size); } if (code == TSDB_CODE_SUCCESS) { pCurWin->isOutput = true; + pAggSup->stateStore.streamStateSessionDel(pAggSup->pState, &pCurWin->sessionWin); } else { pCurWin->sessionWin.win.skey = startTs; pCurWin->sessionWin.win.ekey = endTs; @@ -3432,10 +3428,10 @@ SStreamStateCur* getNextSessionWinInfo(SStreamAggSupporter* pAggSup, SSHashObj* } static void compactSessionWindow(SOperatorInfo* pOperator, SResultWindowInfo* pCurWin, SSHashObj* pStUpdated, - SSHashObj* pStDeleted) { - SExprSupp* pSup = &pOperator->exprSupp; - SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo; - SStorageAPI* pAPI = &pOperator->pTaskInfo->storageAPI; + SSHashObj* pStDeleted, bool addGap) { + SExprSupp* pSup = &pOperator->exprSupp; + SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo; + SStorageAPI* pAPI = &pOperator->pTaskInfo->storageAPI; SStreamSessionAggOperatorInfo* pInfo = pOperator->info; SResultRow* pCurResult = NULL; @@ -3456,7 +3452,7 @@ static void compactSessionWindow(SOperatorInfo* pOperator, SResultWindowInfo* pC initSessionOutputBuf(&winInfo, &pWinResult, pAggSup->pDummyCtx, numOfOutput, pSup->rowEntryInfoOffset); pCurWin->sessionWin.win.ekey = TMAX(pCurWin->sessionWin.win.ekey, winInfo.sessionWin.win.ekey); int64_t winDelta = 0; - if (IS_FINAL_OP(pInfo)) { + if (addGap) { winDelta = pAggSup->gap; } updateTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pCurWin->sessionWin.win, winDelta); @@ -3473,8 +3469,8 @@ static void compactSessionWindow(SOperatorInfo* pOperator, SResultWindowInfo* pC } int32_t saveSessionOutputBuf(SStreamAggSupporter* pAggSup, SResultWindowInfo* pWinInfo) { - saveSessionDiscBuf(pAggSup->pState, &pWinInfo->sessionWin, pWinInfo->pOutputBuf, pAggSup->resultRowSize, - &pAggSup->stateStore); + saveSessionDiscBuf(pAggSup->pState, &pWinInfo->sessionWin, pWinInfo->pOutputBuf, pAggSup->resultRowSize, &pAggSup->stateStore); + pWinInfo->pOutputBuf = NULL; return TSDB_CODE_SUCCESS; } @@ -3488,8 +3484,13 @@ static void doStreamSessionAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSData SResultRow* pResult = NULL; int32_t rows = pSDataBlock->info.rows; int32_t winRows = 0; + SStreamAggSupporter* pAggSup = &pInfo->streamAggSup; pInfo->dataVersion = TMAX(pInfo->dataVersion, pSDataBlock->info.version); + pAggSup->winRange = pTaskInfo->streamInfo.fillHistoryWindow; + if (pAggSup->winRange.ekey <= 0) { + pAggSup->winRange.ekey = INT64_MAX; + } SColumnInfoData* pStartTsCol = taosArrayGet(pSDataBlock->pDataBlock, pInfo->primaryTsIndex); TSKEY* startTsCols = (int64_t*)pStartTsCol->pData; @@ -3501,7 +3502,6 @@ static void doStreamSessionAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSData } TSKEY* endTsCols = (int64_t*)pEndTsCol->pData; - SStreamAggSupporter* pAggSup = &pInfo->streamAggSup; for (int32_t i = 0; i < rows;) { if (pInfo->ignoreExpiredData && isOverdue(endTsCols[i], &pInfo->twAggSup)) { i++; @@ -3526,7 +3526,7 @@ static void doStreamSessionAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSData if (code != TSDB_CODE_SUCCESS || pResult == NULL) { T_LONG_JMP(pTaskInfo->env, TSDB_CODE_OUT_OF_MEMORY); } - compactSessionWindow(pOperator, &winInfo, pStUpdated, pStDeleted); + compactSessionWindow(pOperator, &winInfo, pStUpdated, pStDeleted, addGap); saveSessionOutputBuf(pAggSup, &winInfo); if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE && pStUpdated) { @@ -3690,7 +3690,7 @@ static void rebuildSessionWindow(SOperatorInfo* pOperator, SArray* pWinArray, SS initSessionOutputBuf(&childWin, &pChResult, pChild->exprSupp.pCtx, numOfOutput, pChild->exprSupp.rowEntryInfoOffset); compactFunctions(pSup->pCtx, pChild->exprSupp.pCtx, numOfOutput, pTaskInfo, &pInfo->twAggSup.timeWindowData); - compactSessionWindow(pOperator, &parentWin, pStUpdated, NULL); + compactSessionWindow(pOperator, &parentWin, pStUpdated, NULL, true); saveResult(parentWin, pStUpdated); } else { break; @@ -4091,8 +4091,8 @@ void streamSessionReleaseState(SOperatorInfo* pOperator) { } void resetWinRange(STimeWindow* winRange) { - winRange->skey = INT16_MIN; - winRange->skey = INT16_MAX; + winRange->skey = INT64_MIN; + winRange->ekey = INT64_MAX; } void streamSessionReloadState(SOperatorInfo* pOperator) { @@ -4108,10 +4108,16 @@ void streamSessionReloadState(SOperatorInfo* pOperator) { int32_t num = size / sizeof(SSessionKey); SSessionKey* pSeKeyBuf = (SSessionKey*)pBuf; ASSERT(size == num * sizeof(SSessionKey)); + if (!pInfo->pStUpdated && num > 0) { + _hash_fn_t hashFn = taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY); + pInfo->pStUpdated = tSimpleHashInit(64, hashFn); + } for (int32_t i = 0; i < num; i++) { SResultWindowInfo winInfo = {0}; setSessionOutputBuf(pAggSup, pSeKeyBuf[i].win.skey, pSeKeyBuf[i].win.ekey, pSeKeyBuf[i].groupId, &winInfo); - compactSessionWindow(pOperator, &winInfo, pInfo->pStUpdated, pInfo->pStDeleted); + compactSessionWindow(pOperator, &winInfo, pInfo->pStUpdated, pInfo->pStDeleted, true); + saveSessionOutputBuf(pAggSup, &winInfo); + saveResult(winInfo, pInfo->pStUpdated); } taosMemoryFree(pBuf); @@ -4431,6 +4437,7 @@ void destroyStreamStateOperatorInfo(void* param) { colDataDestroy(&pInfo->twAggSup.timeWindowData); blockDataDestroy(pInfo->pDelRes); taosArrayDestroy(pInfo->historyWins); + tSimpleHashCleanup(pInfo->pSeUpdated); tSimpleHashCleanup(pInfo->pSeDeleted); blockDataDestroy(pInfo->pCheckpointRes); @@ -4487,6 +4494,7 @@ void setStateOutputBuf(SStreamAggSupporter* pAggSup, TSKEY ts, uint64_t groupId, if (code == TSDB_CODE_SUCCESS) { pCurWin->winInfo.isOutput = true; + pAggSup->stateStore.streamStateSessionDel(pAggSup->pState, &pCurWin->winInfo.sessionWin); } else if (pKeyData) { if (IS_VAR_DATA_TYPE(pAggSup->stateKeyType)) { varDataCopy(pCurWin->pStateKey->pData, pKeyData); @@ -4559,8 +4567,13 @@ static void doStreamStateAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBl TSKEY* tsCols = NULL; SResultRow* pResult = NULL; int32_t winRows = 0; + SStreamAggSupporter* pAggSup = &pInfo->streamAggSup; pInfo->dataVersion = TMAX(pInfo->dataVersion, pSDataBlock->info.version); + pAggSup->winRange = pTaskInfo->streamInfo.fillHistoryWindow; + if (pAggSup->winRange.ekey <= 0) { + pAggSup->winRange.ekey = INT64_MAX; + } if (pSDataBlock->pDataBlock != NULL) { SColumnInfoData* pColDataInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->primaryTsIndex); @@ -4569,7 +4582,6 @@ static void doStreamStateAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBl return; } - SStreamAggSupporter* pAggSup = &pInfo->streamAggSup; int32_t rows = pSDataBlock->info.rows; blockDataEnsureCapacity(pAggSup->pScanBlock, rows); SColumnInfoData* pKeyColInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->stateCol.slotId); @@ -4859,7 +4871,7 @@ static void compactStateWindow(SOperatorInfo* pOperator, SResultWindowInfo* pCur SResultRow* pWinResult = NULL; initSessionOutputBuf(pNextWin, &pWinResult, pAggSup->pDummyCtx, numOfOutput, pSup->rowEntryInfoOffset); - updateTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pCurWin->sessionWin.win, true); + updateTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pCurWin->sessionWin.win, 1); compactFunctions(pSup->pCtx, pAggSup->pDummyCtx, numOfOutput, pTaskInfo, &pInfo->twAggSup.timeWindowData); tSimpleHashRemove(pStUpdated, &pNextWin->sessionWin, sizeof(SSessionKey)); if (pNextWin->isOutput && pStDeleted) { @@ -4868,12 +4880,11 @@ static void compactStateWindow(SOperatorInfo* pOperator, SResultWindowInfo* pCur removeSessionResult(pStUpdated, pAggSup->pResultRows, pNextWin->sessionWin); doDeleteSessionWindow(pAggSup, &pNextWin->sessionWin); taosMemoryFree(pNextWin->pOutputBuf); - saveSessionOutputBuf(pAggSup, pCurWin); } void streamStateReloadState(SOperatorInfo* pOperator) { - SStreamSessionAggOperatorInfo* pInfo = pOperator->info; - SStreamAggSupporter* pAggSup = &pInfo->streamAggSup; + SStreamStateAggOperatorInfo* pInfo = pOperator->info; + SStreamAggSupporter* pAggSup = &pInfo->streamAggSup; resetWinRange(&pAggSup->winRange); SSessionKey seKey = {.win.skey = INT64_MIN, .win.ekey = INT64_MIN, .groupId = 0}; @@ -4884,14 +4895,19 @@ void streamStateReloadState(SOperatorInfo* pOperator) { int32_t num = size / sizeof(SSessionKey); SSessionKey* pSeKeyBuf = (SSessionKey*)pBuf; ASSERT(size == num * sizeof(SSessionKey)); + if (!pInfo->pSeUpdated && num > 0) { + _hash_fn_t hashFn = taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY); + pInfo->pSeUpdated = tSimpleHashInit(64, hashFn); + } for (int32_t i = 0; i < num; i++) { SStateWindowInfo curInfo = {0}; SStateWindowInfo nextInfo = {0}; SStateWindowInfo dummy = {0}; setStateOutputBuf(pAggSup, pSeKeyBuf[i].win.skey, pSeKeyBuf[i].groupId, NULL, &curInfo, &nextInfo); - if (compareStateKey(curInfo.pStateKey, nextInfo.pStateKey)) { - compactStateWindow(pOperator, &curInfo.winInfo, &nextInfo.winInfo, pInfo->pStUpdated, pInfo->pStDeleted); - saveResult(curInfo.winInfo, pInfo->pStUpdated); + if (compareStateKey(curInfo.pStateKey,nextInfo.pStateKey)) { + compactStateWindow(pOperator, &curInfo.winInfo, &nextInfo.winInfo, pInfo->pSeUpdated, pInfo->pSeUpdated); + saveSessionOutputBuf(pAggSup, &curInfo.winInfo); + saveResult(curInfo.winInfo, pInfo->pSeUpdated); } if (IS_VALID_SESSION_WIN(curInfo.winInfo)) { @@ -5392,7 +5408,7 @@ static void doMergeIntervalAggImpl(SOperatorInfo* pOperatorInfo, SResultRowInfo* doWindowBorderInterpolation(iaInfo, pBlock, pResult, &win, startPos, forwardRows, pExprSup); } - updateTimeWindowInfo(&iaInfo->twAggSup.timeWindowData, &win, true); + updateTimeWindowInfo(&iaInfo->twAggSup.timeWindowData, &win, 1); applyAggFunctionOnPartialTuples(pTaskInfo, pExprSup->pCtx, &iaInfo->twAggSup.timeWindowData, startPos, forwardRows, pBlock->info.rows, numOfOutput); doCloseWindow(pResultRowInfo, iaInfo, pResult); diff --git a/source/libs/executor/src/tsort.c b/source/libs/executor/src/tsort.c index d26db6536f..30e7148736 100644 --- a/source/libs/executor/src/tsort.c +++ b/source/libs/executor/src/tsort.c @@ -24,6 +24,7 @@ #include "tpagedbuf.h" #include "tsort.h" #include "tutil.h" +#include "tsimplehash.h" struct STupleHandle { SSDataBlock* pBlock; @@ -42,13 +43,15 @@ struct SSortHandle { int64_t startTs; uint64_t totalElapsed; - uint64_t maxRows; - uint32_t maxTupleLength; - uint32_t sortBufSize; + uint64_t pqMaxRows; + uint32_t pqMaxTupleLength; + uint32_t pqSortBufSize; bool forceUsePQSort; BoundedQueue* pBoundedQueue; uint32_t tmpRowIdx; + int64_t mergeLimit; + int32_t sourceId; SSDataBlock* pDataBlock; SMsortComparParam cmpParam; @@ -173,8 +176,8 @@ void destroyTuple(void* t) { * @return */ SSortHandle* tsortCreateSortHandle(SArray* pSortInfo, int32_t type, int32_t pageSize, int32_t numOfPages, - SSDataBlock* pBlock, const char* idstr, uint64_t maxRows, uint32_t maxTupleLength, - uint32_t sortBufSize) { + SSDataBlock* pBlock, const char* idstr, uint64_t pqMaxRows, uint32_t pqMaxTupleLength, + uint32_t pqSortBufSize) { SSortHandle* pSortHandle = taosMemoryCalloc(1, sizeof(SSortHandle)); pSortHandle->type = type; @@ -183,10 +186,10 @@ SSortHandle* tsortCreateSortHandle(SArray* pSortInfo, int32_t type, int32_t page pSortHandle->pSortInfo = pSortInfo; pSortHandle->loops = 0; - pSortHandle->maxTupleLength = maxTupleLength; - if (maxRows != 0) { - pSortHandle->sortBufSize = sortBufSize; - pSortHandle->maxRows = maxRows; + pSortHandle->pqMaxTupleLength = pqMaxTupleLength; + if (pqMaxRows != 0) { + pSortHandle->pqSortBufSize = pqSortBufSize; + pSortHandle->pqMaxRows = pqMaxRows; } pSortHandle->forceUsePQSort = false; @@ -194,10 +197,18 @@ SSortHandle* tsortCreateSortHandle(SArray* pSortInfo, int32_t type, int32_t page pSortHandle->pDataBlock = createOneDataBlock(pBlock, false); } + pSortHandle->mergeLimit = -1; + pSortHandle->pOrderedSource = taosArrayInit(4, POINTER_BYTES); pSortHandle->cmpParam.orderInfo = pSortInfo; pSortHandle->cmpParam.cmpGroupId = false; - + pSortHandle->cmpParam.sortType = type; + if (type == SORT_BLOCK_TS_MERGE) { + SBlockOrderInfo* pOrder = TARRAY_GET_ELEM(pSortInfo, 0); + pSortHandle->cmpParam.tsSlotId = pOrder->slotId; + pSortHandle->cmpParam.order = pOrder->order; + pSortHandle->cmpParam.cmpFn = (pOrder->order == TSDB_ORDER_ASC) ? compareInt64Val : compareInt64ValDesc; + } tsortSetComparFp(pSortHandle, msortComparFn); if (idstr != NULL) { @@ -469,11 +480,14 @@ static int32_t adjustMergeTreeForNextTuple(SSortSource* pSource, SMultiwayMergeT if (pHandle->type == SORT_SINGLESOURCE_SORT) { pSource->pageIndex++; if (pSource->pageIndex >= taosArrayGetSize(pSource->pageIdList)) { + qDebug("adjust merge tree. %d source completed %d", *numOfCompleted, pSource->pageIndex); (*numOfCompleted) += 1; pSource->src.rowIndex = -1; pSource->pageIndex = -1; pSource->src.pBlock = blockDataDestroy(pSource->src.pBlock); } else { + if (pSource->pageIndex % 512 == 0) qDebug("begin source %p page %d", pSource, pSource->pageIndex); + int32_t* pPgId = taosArrayGet(pSource->pageIdList, pSource->pageIndex); void* pPage = getBufPage(pHandle->pBuf, *pPgId); @@ -486,7 +500,6 @@ static int32_t adjustMergeTreeForNextTuple(SSortSource* pSource, SMultiwayMergeT if (code != TSDB_CODE_SUCCESS) { return code; } - releaseBufPage(pHandle->pBuf, pPage); } } else { @@ -497,6 +510,7 @@ static int32_t adjustMergeTreeForNextTuple(SSortSource* pSource, SMultiwayMergeT if (pSource->src.pBlock == NULL) { (*numOfCompleted) += 1; pSource->src.rowIndex = -1; + qDebug("adjust merge tree. %d source completed", *numOfCompleted); } } } @@ -577,53 +591,63 @@ int32_t msortComparFn(const void* pLeft, const void* pRight, void* param) { } } - for (int32_t i = 0; i < pInfo->size; ++i) { - SBlockOrderInfo* pOrder = TARRAY_GET_ELEM(pInfo, i); - SColumnInfoData* pLeftColInfoData = TARRAY_GET_ELEM(pLeftBlock->pDataBlock, pOrder->slotId); + if (pParam->sortType == SORT_BLOCK_TS_MERGE) { + SColumnInfoData* pLeftColInfoData = TARRAY_GET_ELEM(pLeftBlock->pDataBlock, pParam->tsSlotId); + SColumnInfoData* pRightColInfoData = TARRAY_GET_ELEM(pRightBlock->pDataBlock, pParam->tsSlotId); + int64_t* left1 = (int64_t*)(pLeftColInfoData->pData) + pLeftSource->src.rowIndex; + int64_t* right1 = (int64_t*)(pRightColInfoData->pData) + pRightSource->src.rowIndex; - bool leftNull = false; - if (pLeftColInfoData->hasNull) { - if (pLeftBlock->pBlockAgg == NULL) { - leftNull = colDataIsNull_s(pLeftColInfoData, pLeftSource->src.rowIndex); - } else { - leftNull = - colDataIsNull(pLeftColInfoData, pLeftBlock->info.rows, pLeftSource->src.rowIndex, pLeftBlock->pBlockAgg[i]); + int ret = pParam->cmpFn(left1, right1); + return ret; + } else { + for (int32_t i = 0; i < pInfo->size; ++i) { + SBlockOrderInfo* pOrder = TARRAY_GET_ELEM(pInfo, i); + SColumnInfoData* pLeftColInfoData = TARRAY_GET_ELEM(pLeftBlock->pDataBlock, pOrder->slotId); + SColumnInfoData* pRightColInfoData = TARRAY_GET_ELEM(pRightBlock->pDataBlock, pOrder->slotId); + + bool leftNull = false; + if (pLeftColInfoData->hasNull) { + if (pLeftBlock->pBlockAgg == NULL) { + leftNull = colDataIsNull_s(pLeftColInfoData, pLeftSource->src.rowIndex); + } else { + leftNull = colDataIsNull(pLeftColInfoData, pLeftBlock->info.rows, pLeftSource->src.rowIndex, + pLeftBlock->pBlockAgg[i]); + } } - } - SColumnInfoData* pRightColInfoData = TARRAY_GET_ELEM(pRightBlock->pDataBlock, pOrder->slotId); - bool rightNull = false; - if (pRightColInfoData->hasNull) { - if (pRightBlock->pBlockAgg == NULL) { - rightNull = colDataIsNull_s(pRightColInfoData, pRightSource->src.rowIndex); - } else { - rightNull = colDataIsNull(pRightColInfoData, pRightBlock->info.rows, pRightSource->src.rowIndex, - pRightBlock->pBlockAgg[i]); + bool rightNull = false; + if (pRightColInfoData->hasNull) { + if (pRightBlock->pBlockAgg == NULL) { + rightNull = colDataIsNull_s(pRightColInfoData, pRightSource->src.rowIndex); + } else { + rightNull = colDataIsNull(pRightColInfoData, pRightBlock->info.rows, pRightSource->src.rowIndex, + pRightBlock->pBlockAgg[i]); + } } - } - if (leftNull && rightNull) { - continue; // continue to next slot - } + if (leftNull && rightNull) { + continue; // continue to next slot + } - if (rightNull) { - return pOrder->nullFirst ? 1 : -1; - } + if (rightNull) { + return pOrder->nullFirst ? 1 : -1; + } - if (leftNull) { - return pOrder->nullFirst ? -1 : 1; - } + if (leftNull) { + return pOrder->nullFirst ? -1 : 1; + } - void* left1 = colDataGetData(pLeftColInfoData, pLeftSource->src.rowIndex); - void* right1 = colDataGetData(pRightColInfoData, pRightSource->src.rowIndex); + void* left1 = colDataGetData(pLeftColInfoData, pLeftSource->src.rowIndex); + void* right1 = colDataGetData(pRightColInfoData, pRightSource->src.rowIndex); - __compar_fn_t fn = getKeyComparFunc(pLeftColInfoData->info.type, pOrder->order); + __compar_fn_t fn = getKeyComparFunc(pLeftColInfoData->info.type, pOrder->order); - int ret = fn(left1, right1); - if (ret == 0) { - continue; - } else { - return ret; + int ret = fn(left1, right1); + if (ret == 0) { + continue; + } else { + return ret; + } } } return 0; @@ -668,6 +692,7 @@ static int32_t doInternalMergeSort(SSortHandle* pHandle) { // Only *numOfInputSources* can be loaded into buffer to perform the external sort. for (int32_t i = 0; i < sortGroup; ++i) { + qDebug("internal merge sort pass %d group %d. num input sources %d ", t, i, numOfInputSources); pHandle->sourceId += 1; int32_t end = (i + 1) * numOfInputSources - 1; @@ -690,13 +715,15 @@ static int32_t doInternalMergeSort(SSortHandle* pHandle) { return code; } + int nMergedRows = 0; + SArray* pPageIdList = taosArrayInit(4, sizeof(int32_t)); while (1) { if (tsortIsClosed(pHandle)) { code = terrno = TSDB_CODE_TSC_QUERY_CANCELLED; return code; } - + SSDataBlock* pDataBlock = getSortedBlockDataInner(pHandle, &pHandle->cmpParam, numOfRows); if (pDataBlock == NULL) { break; @@ -720,8 +747,12 @@ static int32_t doInternalMergeSort(SSortHandle* pHandle) { setBufPageDirty(pPage, true); releaseBufPage(pHandle->pBuf, pPage); + nMergedRows += pDataBlock->info.rows; blockDataCleanup(pDataBlock); + if ((pHandle->mergeLimit != -1) && (nMergedRows >= pHandle->mergeLimit)) { + break; + } } sortComparCleanup(&pHandle->cmpParam); @@ -769,39 +800,285 @@ int32_t getProperSortPageSize(size_t rowSize, uint32_t numOfCols) { return pgSize; } -static int32_t createInitialSources(SSortHandle* pHandle) { - size_t sortBufSize = pHandle->numOfPages * pHandle->pageSize; +static int32_t createPageBuf(SSortHandle* pHandle) { + if (pHandle->pBuf == NULL) { + if (!osTempSpaceAvailable()) { + terrno = TSDB_CODE_NO_DISKSPACE; + qError("create page buf failed since %s, tempDir:%s", terrstr(), tsTempDir); + return terrno; + } + + int32_t code = createDiskbasedBuf(&pHandle->pBuf, pHandle->pageSize, pHandle->numOfPages * pHandle->pageSize, + "tableBlocksBuf", tsTempDir); + dBufSetPrintInfo(pHandle->pBuf); + if (code != TSDB_CODE_SUCCESS) { + return code; + } + } + return 0; +} + +typedef struct SBlkMergeSupport { + int64_t** aTs; + int32_t* aRowIdx; + int32_t order; +} SBlkMergeSupport; + +static int32_t blockCompareTsFn(const void* pLeft, const void* pRight, void* param) { + int32_t left = *(int32_t*)pLeft; + int32_t right = *(int32_t*)pRight; + + SBlkMergeSupport* pSup = (SBlkMergeSupport*)param; + if (pSup->aRowIdx[left] == -1) { + return 1; + } else if (pSup->aRowIdx[right] == -1) { + return -1; + } + + int64_t leftTs = pSup->aTs[left][pSup->aRowIdx[left]]; + int64_t rightTs = pSup->aTs[right][pSup->aRowIdx[right]]; + + int32_t ret = leftTs>rightTs ? 1 : ((leftTs < rightTs) ? -1 : 0); + if (pSup->order == TSDB_ORDER_DESC) { + ret = -1 * ret; + } + return ret; +} + +static int32_t appendDataBlockToPageBuf(SSortHandle* pHandle, SSDataBlock* blk, SArray* aPgId) { + int32_t pageId = -1; + void* pPage = getNewBufPage(pHandle->pBuf, &pageId); + taosArrayPush(aPgId, &pageId); + + int32_t size = blockDataGetSize(blk) + sizeof(int32_t) + taosArrayGetSize(blk->pDataBlock) * sizeof(int32_t); + ASSERT(size <= getBufPageSize(pHandle->pBuf)); + + blockDataToBuf(pPage, blk); + + setBufPageDirty(pPage, true); + releaseBufPage(pHandle->pBuf, pPage); + + return 0; +} + +static int32_t getPageBufIncForRow(SSDataBlock* blk, int32_t row, int32_t rowIdxInPage) { + int sz = 0; + int numCols = taosArrayGetSize(blk->pDataBlock); + if (!blk->info.hasVarCol) { + sz += numCols * ((rowIdxInPage & 0x7) == 0 ? 1: 0); + sz += blockDataGetRowSize(blk); + } else { + for (int32_t i = 0; i < numCols; ++i) { + SColumnInfoData* pColInfoData = TARRAY_GET_ELEM(blk->pDataBlock, i); + if (IS_VAR_DATA_TYPE(pColInfoData->info.type)) { + if (pColInfoData->varmeta.offset[row] != -1) { + char* p = colDataGetData(pColInfoData, row); + sz += varDataTLen(p); + } + + sz += sizeof(pColInfoData->varmeta.offset[0]); + } else { + sz += pColInfoData->info.bytes; + + if (((rowIdxInPage) & 0x07) == 0) { + sz += 1; // bitmap + } + } + } + } + return sz; +} + +static int32_t sortBlocksToExtSource(SSortHandle* pHandle, SArray* aBlk, SBlockOrderInfo* order, SArray* aExtSrc) { + int pgHeaderSz = sizeof(int32_t) + sizeof(int32_t) * taosArrayGetSize(pHandle->pDataBlock->pDataBlock); + int32_t rowCap = blockDataGetCapacityInRow(pHandle->pDataBlock, pHandle->pageSize, pgHeaderSz); + blockDataEnsureCapacity(pHandle->pDataBlock, rowCap); + blockDataCleanup(pHandle->pDataBlock); + int32_t numBlks = taosArrayGetSize(aBlk); + + SBlkMergeSupport sup; + sup.aRowIdx = taosMemoryCalloc(numBlks, sizeof(int32_t)); + sup.aTs = taosMemoryCalloc(numBlks, sizeof(int64_t*)); + sup.order = order->order; + for (int i = 0; i < numBlks; ++i) { + SSDataBlock* blk = taosArrayGetP(aBlk, i); + SColumnInfoData* col = taosArrayGet(blk->pDataBlock, order->slotId); + sup.aTs[i] = (int64_t*)col->pData; + sup.aRowIdx[i] = 0; + } + + int32_t totalRows = 0; + for (int i = 0; i < numBlks; ++i) { + SSDataBlock* blk = taosArrayGetP(aBlk, i); + totalRows += blk->info.rows; + } + + SArray* aPgId = taosArrayInit(8, sizeof(int32_t)); + + SMultiwayMergeTreeInfo* pTree = NULL; + tMergeTreeCreate(&pTree, taosArrayGetSize(aBlk), &sup, blockCompareTsFn); + int32_t nRows = 0; + int32_t nMergedRows = 0; + bool mergeLimitReached = false; + size_t blkPgSz = pgHeaderSz; + + while (nRows < totalRows) { + int32_t minIdx = tMergeTreeGetChosenIndex(pTree); + SSDataBlock* minBlk = taosArrayGetP(aBlk, minIdx); + int32_t minRow = sup.aRowIdx[minIdx]; + int32_t bufInc = getPageBufIncForRow(minBlk, minRow, pHandle->pDataBlock->info.rows); + + if (blkPgSz <= pHandle->pageSize && blkPgSz + bufInc > pHandle->pageSize) { + appendDataBlockToPageBuf(pHandle, pHandle->pDataBlock, aPgId); + nMergedRows += pHandle->pDataBlock->info.rows; + + blockDataCleanup(pHandle->pDataBlock); + blkPgSz = pgHeaderSz; + bufInc = getPageBufIncForRow(minBlk, minRow, 0); + if ((pHandle->mergeLimit != -1) && (nMergedRows >= pHandle->mergeLimit)) { + mergeLimitReached = true; + break; + } + } + blockDataEnsureCapacity(pHandle->pDataBlock, pHandle->pDataBlock->info.rows + 1); + appendOneRowToDataBlock(pHandle->pDataBlock, minBlk, &minRow); + blkPgSz += bufInc; + + ++nRows; + + if (sup.aRowIdx[minIdx] == minBlk->info.rows - 1) { + sup.aRowIdx[minIdx] = -1; + } else { + ++sup.aRowIdx[minIdx]; + } + tMergeTreeAdjust(pTree, tMergeTreeGetAdjustIndex(pTree)); + } + if (pHandle->pDataBlock->info.rows > 0) { + if (!mergeLimitReached) { + appendDataBlockToPageBuf(pHandle, pHandle->pDataBlock, aPgId); + nMergedRows += pHandle->pDataBlock->info.rows; + } + blockDataCleanup(pHandle->pDataBlock); + } + SSDataBlock* pMemSrcBlk = createOneDataBlock(pHandle->pDataBlock, false); + doAddNewExternalMemSource(pHandle->pBuf, aExtSrc, pMemSrcBlk, &pHandle->sourceId, aPgId); + + taosMemoryFree(sup.aRowIdx); + taosMemoryFree(sup.aTs); + + tMergeTreeDestroy(&pTree); + + return 0; +} + +static int32_t createBlocksMergeSortInitialSources(SSortHandle* pHandle) { + SBlockOrderInfo* pOrder = taosArrayGet(pHandle->pSortInfo, 0); + size_t nSrc = taosArrayGetSize(pHandle->pOrderedSource); + SArray* aExtSrc = taosArrayInit(nSrc, POINTER_BYTES); + + size_t maxBufSize = pHandle->numOfPages * pHandle->pageSize; + createPageBuf(pHandle); + + SSortSource* pSrc = taosArrayGetP(pHandle->pOrderedSource, 0); + int32_t szSort = 0; + + SArray* aBlkSort = taosArrayInit(8, POINTER_BYTES); + SSHashObj* mUidBlk = tSimpleHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT)); + while (1) { + SSDataBlock* pBlk = pHandle->fetchfp(pSrc->param); + + if (pBlk != NULL) { + szSort += blockDataGetSize(pBlk); + + void* ppBlk = tSimpleHashGet(mUidBlk, &pBlk->info.id.uid, sizeof(pBlk->info.id.uid)); + if (ppBlk != NULL) { + SSDataBlock* tBlk = *(SSDataBlock**)(ppBlk); + blockDataMerge(tBlk, pBlk); + } else { + SSDataBlock* tBlk = createOneDataBlock(pBlk, true); + tSimpleHashPut(mUidBlk, &pBlk->info.id.uid, sizeof(pBlk->info.id.uid), &tBlk, POINTER_BYTES); + taosArrayPush(aBlkSort, &tBlk); + } + } + + if ((pBlk != NULL && szSort > maxBufSize) || (pBlk == NULL && szSort > 0)) { + tSimpleHashClear(mUidBlk); + + int64_t p = taosGetTimestampUs(); + sortBlocksToExtSource(pHandle, aBlkSort, pOrder, aExtSrc); + int64_t el = taosGetTimestampUs() - p; + pHandle->sortElapsed += el; + + for (int i = 0; i < taosArrayGetSize(aBlkSort); ++i) { + blockDataDestroy(taosArrayGetP(aBlkSort, i)); + } + taosArrayClear(aBlkSort); + szSort = 0; + qDebug("source %zu created", taosArrayGetSize(aExtSrc)); + } + if (pBlk == NULL) { + break; + }; + } + tSimpleHashCleanup(mUidBlk); + taosArrayDestroy(aBlkSort); + tsortClearOrderdSource(pHandle->pOrderedSource, NULL, NULL); + taosArrayAddAll(pHandle->pOrderedSource, aExtSrc); + taosArrayDestroy(aExtSrc); + + pHandle->type = SORT_SINGLESOURCE_SORT; + return 0; +} + +static int32_t createBlocksQuickSortInitialSources(SSortHandle* pHandle) { int32_t code = 0; + size_t sortBufSize = pHandle->numOfPages * pHandle->pageSize; - if (pHandle->type == SORT_SINGLESOURCE_SORT) { - SSortSource** pSource = taosArrayGet(pHandle->pOrderedSource, 0); - SSortSource* source = *pSource; - *pSource = NULL; + SSortSource** pSource = taosArrayGet(pHandle->pOrderedSource, 0); + SSortSource* source = *pSource; + *pSource = NULL; - tsortClearOrderdSource(pHandle->pOrderedSource, NULL, NULL); + tsortClearOrderdSource(pHandle->pOrderedSource, NULL, NULL); - while (1) { - SSDataBlock* pBlock = pHandle->fetchfp(source->param); - if (pBlock == NULL) { - break; + while (1) { + SSDataBlock* pBlock = pHandle->fetchfp(source->param); + if (pBlock == NULL) { + break; + } + + if (pHandle->pDataBlock == NULL) { + uint32_t numOfCols = taosArrayGetSize(pBlock->pDataBlock); + pHandle->pageSize = getProperSortPageSize(blockDataGetRowSize(pBlock), numOfCols); + + // todo, number of pages are set according to the total available sort buffer + pHandle->numOfPages = 1024; + sortBufSize = pHandle->numOfPages * pHandle->pageSize; + pHandle->pDataBlock = createOneDataBlock(pBlock, false); + } + + if (pHandle->beforeFp != NULL) { + pHandle->beforeFp(pBlock, pHandle->param); + } + + code = blockDataMerge(pHandle->pDataBlock, pBlock); + if (code != TSDB_CODE_SUCCESS) { + if (source->param && !source->onlyRef) { + taosMemoryFree(source->param); } - - if (pHandle->pDataBlock == NULL) { - uint32_t numOfCols = taosArrayGetSize(pBlock->pDataBlock); - pHandle->pageSize = getProperSortPageSize(blockDataGetRowSize(pBlock), numOfCols); - - // todo, number of pages are set according to the total available sort buffer - pHandle->numOfPages = 1024; - sortBufSize = pHandle->numOfPages * pHandle->pageSize; - pHandle->pDataBlock = createOneDataBlock(pBlock, false); + if (!source->onlyRef && source->src.pBlock) { + blockDataDestroy(source->src.pBlock); + source->src.pBlock = NULL; } + taosMemoryFree(source); + return code; + } - if (pHandle->beforeFp != NULL) { - pHandle->beforeFp(pBlock, pHandle->param); - } - - code = blockDataMerge(pHandle->pDataBlock, pBlock); - if (code != TSDB_CODE_SUCCESS) { + size_t size = blockDataGetSize(pHandle->pDataBlock); + if (size > sortBufSize) { + // Perform the in-memory sort and then flush data in the buffer into disk. + int64_t p = taosGetTimestampUs(); + code = blockDataSort(pHandle->pDataBlock, pHandle->pSortInfo); + if (code != 0) { if (source->param && !source->onlyRef) { taosMemoryFree(source->param); } @@ -809,74 +1086,67 @@ static int32_t createInitialSources(SSortHandle* pHandle) { blockDataDestroy(source->src.pBlock); source->src.pBlock = NULL; } + taosMemoryFree(source); return code; } - size_t size = blockDataGetSize(pHandle->pDataBlock); - if (size > sortBufSize) { - // Perform the in-memory sort and then flush data in the buffer into disk. - int64_t p = taosGetTimestampUs(); - code = blockDataSort(pHandle->pDataBlock, pHandle->pSortInfo); - if (code != 0) { - if (source->param && !source->onlyRef) { - taosMemoryFree(source->param); - } - if (!source->onlyRef && source->src.pBlock) { - blockDataDestroy(source->src.pBlock); - source->src.pBlock = NULL; - } - - taosMemoryFree(source); - return code; - } - - int64_t el = taosGetTimestampUs() - p; - pHandle->sortElapsed += el; - if (pHandle->maxRows > 0) blockDataKeepFirstNRows(pHandle->pDataBlock, pHandle->maxRows); - code = doAddToBuf(pHandle->pDataBlock, pHandle); - if (code != TSDB_CODE_SUCCESS) { - return code; - } - } - } - - if (source->param && !source->onlyRef) { - taosMemoryFree(source->param); - } - - taosMemoryFree(source); - - if (pHandle->pDataBlock != NULL && pHandle->pDataBlock->info.rows > 0) { - size_t size = blockDataGetSize(pHandle->pDataBlock); - - // Perform the in-memory sort and then flush data in the buffer into disk. - int64_t p = taosGetTimestampUs(); - - code = blockDataSort(pHandle->pDataBlock, pHandle->pSortInfo); - if (code != 0) { - return code; - } - - if (pHandle->maxRows > 0) blockDataKeepFirstNRows(pHandle->pDataBlock, pHandle->maxRows); int64_t el = taosGetTimestampUs() - p; pHandle->sortElapsed += el; - - // All sorted data can fit in memory, external memory sort is not needed. Return to directly - if (size <= sortBufSize && pHandle->pBuf == NULL) { - pHandle->cmpParam.numOfSources = 1; - pHandle->inMemSort = true; - - pHandle->loops = 1; - pHandle->tupleHandle.rowIndex = -1; - pHandle->tupleHandle.pBlock = pHandle->pDataBlock; - return 0; - } else { - code = doAddToBuf(pHandle->pDataBlock, pHandle); + if (pHandle->pqMaxRows > 0) blockDataKeepFirstNRows(pHandle->pDataBlock, pHandle->pqMaxRows); + code = doAddToBuf(pHandle->pDataBlock, pHandle); + if (code != TSDB_CODE_SUCCESS) { + return code; } } } + if (source->param && !source->onlyRef) { + taosMemoryFree(source->param); + } + + taosMemoryFree(source); + + if (pHandle->pDataBlock != NULL && pHandle->pDataBlock->info.rows > 0) { + size_t size = blockDataGetSize(pHandle->pDataBlock); + + // Perform the in-memory sort and then flush data in the buffer into disk. + int64_t p = taosGetTimestampUs(); + + code = blockDataSort(pHandle->pDataBlock, pHandle->pSortInfo); + if (code != 0) { + return code; + } + + if (pHandle->pqMaxRows > 0) blockDataKeepFirstNRows(pHandle->pDataBlock, pHandle->pqMaxRows); + int64_t el = taosGetTimestampUs() - p; + pHandle->sortElapsed += el; + + // All sorted data can fit in memory, external memory sort is not needed. Return to directly + if (size <= sortBufSize && pHandle->pBuf == NULL) { + pHandle->cmpParam.numOfSources = 1; + pHandle->inMemSort = true; + + pHandle->loops = 1; + pHandle->tupleHandle.rowIndex = -1; + pHandle->tupleHandle.pBlock = pHandle->pDataBlock; + return 0; + } else { + code = doAddToBuf(pHandle->pDataBlock, pHandle); + } + } + return code; +} + +static int32_t createInitialSources(SSortHandle* pHandle) { + int32_t code = 0; + + if (pHandle->type == SORT_SINGLESOURCE_SORT) { + code = createBlocksQuickSortInitialSources(pHandle); + } else if (pHandle->type == SORT_BLOCK_TS_MERGE) { + code = createBlocksMergeSortInitialSources(pHandle); + } + qDebug("%zu sources created", taosArrayGetSize(pHandle->pOrderedSource)); return code; } @@ -923,6 +1193,10 @@ void tsortSetClosed(SSortHandle* pHandle) { atomic_store_8(&pHandle->closed, 2); } +void tsortSetMergeLimit(SSortHandle* pHandle, int64_t mergeLimit) { + pHandle->mergeLimit = mergeLimit; +} + int32_t tsortSetFetchRawDataFp(SSortHandle* pHandle, _sort_fetch_block_fn_t fetchFp, void (*fp)(SSDataBlock*, void*), void* param) { pHandle->fetchfp = fetchFp; @@ -1002,8 +1276,8 @@ void tsortSetForceUsePQSort(SSortHandle* pHandle) { static bool tsortIsPQSortApplicable(SSortHandle* pHandle) { if (pHandle->type != SORT_SINGLESOURCE_SORT) return false; if (tsortIsForceUsePQSort(pHandle)) return true; - uint64_t maxRowsFitInMemory = pHandle->sortBufSize / (pHandle->maxTupleLength + sizeof(char*)); - return maxRowsFitInMemory > pHandle->maxRows; + uint64_t maxRowsFitInMemory = pHandle->pqSortBufSize / (pHandle->pqMaxTupleLength + sizeof(char*)); + return maxRowsFitInMemory > pHandle->pqMaxRows; } static bool tsortPQCompFn(void* a, void* b, void* param) { @@ -1049,7 +1323,7 @@ static int32_t tupleComparFn(const void* pLeft, const void* pRight, void* param) } static int32_t tsortOpenForPQSort(SSortHandle* pHandle) { - pHandle->pBoundedQueue = createBoundedQueue(pHandle->maxRows, tsortPQCompFn, destroyTuple, pHandle); + pHandle->pBoundedQueue = createBoundedQueue(pHandle->pqMaxRows, tsortPQCompFn, destroyTuple, pHandle); if (NULL == pHandle->pBoundedQueue) return TSDB_CODE_OUT_OF_MEMORY; tsortSetComparFp(pHandle, tupleComparFn); diff --git a/source/libs/nodes/src/nodesCloneFuncs.c b/source/libs/nodes/src/nodesCloneFuncs.c index 6e4dde4ec1..f5eacf0bd5 100644 --- a/source/libs/nodes/src/nodesCloneFuncs.c +++ b/source/libs/nodes/src/nodesCloneFuncs.c @@ -361,6 +361,7 @@ static int32_t logicNodeCopy(const SLogicNode* pSrc, SLogicNode* pDst) { COPY_SCALAR_FIELD(groupAction); COPY_SCALAR_FIELD(inputTsOrder); COPY_SCALAR_FIELD(outputTsOrder); + COPY_SCALAR_FIELD(forceCreateNonBlockingOptr); return TSDB_CODE_SUCCESS; } @@ -397,6 +398,7 @@ static int32_t logicScanCopy(const SScanLogicNode* pSrc, SScanLogicNode* pDst) { CLONE_NODE_LIST_FIELD(pTags); CLONE_NODE_FIELD(pSubtable); COPY_SCALAR_FIELD(igLastNull); + COPY_SCALAR_FIELD(groupOrderScan); return TSDB_CODE_SUCCESS; } @@ -545,6 +547,7 @@ static int32_t physiNodeCopy(const SPhysiNode* pSrc, SPhysiNode* pDst) { CLONE_NODE_LIST_FIELD(pChildren); COPY_SCALAR_FIELD(inputTsOrder); COPY_SCALAR_FIELD(outputTsOrder); + COPY_SCALAR_FIELD(forceCreateNonBlockingOptr); return TSDB_CODE_SUCCESS; } @@ -556,6 +559,7 @@ static int32_t physiScanCopy(const SScanPhysiNode* pSrc, SScanPhysiNode* pDst) { COPY_SCALAR_FIELD(suid); COPY_SCALAR_FIELD(tableType); COPY_OBJECT_FIELD(tableName, sizeof(SName)); + COPY_SCALAR_FIELD(groupOrderScan); return TSDB_CODE_SUCCESS; } diff --git a/source/libs/nodes/src/nodesCodeFuncs.c b/source/libs/nodes/src/nodesCodeFuncs.c index 81116a60b0..f25616065e 100644 --- a/source/libs/nodes/src/nodesCodeFuncs.c +++ b/source/libs/nodes/src/nodesCodeFuncs.c @@ -1559,6 +1559,7 @@ static const char* jkScanPhysiPlanTableId = "TableId"; static const char* jkScanPhysiPlanSTableId = "STableId"; static const char* jkScanPhysiPlanTableType = "TableType"; static const char* jkScanPhysiPlanTableName = "TableName"; +static const char* jkScanPhysiPlanGroupOrderScan = "GroupOrderScan"; static int32_t physiScanNodeToJson(const void* pObj, SJson* pJson) { const STagScanPhysiNode* pNode = (const STagScanPhysiNode*)pObj; @@ -1582,6 +1583,9 @@ static int32_t physiScanNodeToJson(const void* pObj, SJson* pJson) { if (TSDB_CODE_SUCCESS == code) { code = tjsonAddObject(pJson, jkScanPhysiPlanTableName, nameToJson, &pNode->tableName); } + if (TSDB_CODE_SUCCESS == code) { + code = tjsonAddBoolToObject(pJson, jkScanPhysiPlanGroupOrderScan, pNode->groupOrderScan); + } return code; } @@ -1608,6 +1612,9 @@ static int32_t jsonToPhysiScanNode(const SJson* pJson, void* pObj) { if (TSDB_CODE_SUCCESS == code) { code = tjsonToObject(pJson, jkScanPhysiPlanTableName, jsonToName, &pNode->tableName); } + if (TSDB_CODE_SUCCESS == code) { + code = tjsonGetBoolValue(pJson, jkScanPhysiPlanGroupOrderScan, &pNode->groupOrderScan); + } return code; } diff --git a/source/libs/nodes/src/nodesMsgFuncs.c b/source/libs/nodes/src/nodesMsgFuncs.c index 1ca37defa4..20e829766d 100644 --- a/source/libs/nodes/src/nodesMsgFuncs.c +++ b/source/libs/nodes/src/nodesMsgFuncs.c @@ -1853,7 +1853,8 @@ enum { PHY_NODE_CODE_LIMIT, PHY_NODE_CODE_SLIMIT, PHY_NODE_CODE_INPUT_TS_ORDER, - PHY_NODE_CODE_OUTPUT_TS_ORDER + PHY_NODE_CODE_OUTPUT_TS_ORDER, + PHY_NODE_CODE_FORCE_NONBLOCKING_OPTR }; static int32_t physiNodeToMsg(const void* pObj, STlvEncoder* pEncoder) { @@ -1878,6 +1879,9 @@ static int32_t physiNodeToMsg(const void* pObj, STlvEncoder* pEncoder) { if (TSDB_CODE_SUCCESS == code) { code = tlvEncodeEnum(pEncoder, PHY_NODE_CODE_OUTPUT_TS_ORDER, pNode->outputTsOrder); } + if (TSDB_CODE_SUCCESS == code) { + code = tlvEncodeBool(pEncoder, PHY_NODE_CODE_FORCE_NONBLOCKING_OPTR, pNode->forceCreateNonBlockingOptr); + } return code; } @@ -1910,6 +1914,8 @@ static int32_t msgToPhysiNode(STlvDecoder* pDecoder, void* pObj) { case PHY_NODE_CODE_OUTPUT_TS_ORDER: code = tlvDecodeEnum(pTlv, &pNode->outputTsOrder, sizeof(pNode->outputTsOrder)); break; + case PHY_NODE_CODE_FORCE_NONBLOCKING_OPTR: + code = tlvDecodeBool(pTlv, &pNode->forceCreateNonBlockingOptr); default: break; } @@ -1925,7 +1931,8 @@ enum { PHY_SCAN_CODE_BASE_UID, PHY_SCAN_CODE_BASE_SUID, PHY_SCAN_CODE_BASE_TABLE_TYPE, - PHY_SCAN_CODE_BASE_TABLE_NAME + PHY_SCAN_CODE_BASE_TABLE_NAME, + PHY_SCAN_CODE_BASE_GROUP_ORDER_SCAN }; static int32_t physiScanNodeToMsg(const void* pObj, STlvEncoder* pEncoder) { @@ -1950,6 +1957,9 @@ static int32_t physiScanNodeToMsg(const void* pObj, STlvEncoder* pEncoder) { if (TSDB_CODE_SUCCESS == code) { code = tlvEncodeObj(pEncoder, PHY_SCAN_CODE_BASE_TABLE_NAME, nameToMsg, &pNode->tableName); } + if (TSDB_CODE_SUCCESS == code) { + code = tlvEncodeBool(pEncoder, PHY_SCAN_CODE_BASE_GROUP_ORDER_SCAN, pNode->groupOrderScan); + } return code; } @@ -1982,6 +1992,9 @@ static int32_t msgToPhysiScanNode(STlvDecoder* pDecoder, void* pObj) { case PHY_SCAN_CODE_BASE_TABLE_NAME: code = tlvDecodeObjFromTlv(pTlv, msgToName, &pNode->tableName); break; + case PHY_SCAN_CODE_BASE_GROUP_ORDER_SCAN: + code = tlvDecodeBool(pTlv, &pNode->groupOrderScan); + break; default: break; } diff --git a/source/libs/parser/src/parTranslater.c b/source/libs/parser/src/parTranslater.c index 1427ada6da..a03dab2480 100644 --- a/source/libs/parser/src/parTranslater.c +++ b/source/libs/parser/src/parTranslater.c @@ -7719,7 +7719,7 @@ static int32_t extractShowCreateTableResultSchema(int32_t* numOfCols, SSchema** } static int32_t extractShowVariablesResultSchema(int32_t* numOfCols, SSchema** pSchema) { - *numOfCols = 2; + *numOfCols = 3; *pSchema = taosMemoryCalloc((*numOfCols), sizeof(SSchema)); if (NULL == (*pSchema)) { return TSDB_CODE_OUT_OF_MEMORY; @@ -7733,6 +7733,10 @@ static int32_t extractShowVariablesResultSchema(int32_t* numOfCols, SSchema** pS (*pSchema)[1].bytes = TSDB_CONFIG_VALUE_LEN; strcpy((*pSchema)[1].name, "value"); + (*pSchema)[2].type = TSDB_DATA_TYPE_BINARY; + (*pSchema)[2].bytes = TSDB_CONFIG_SCOPE_LEN; + strcpy((*pSchema)[2].name, "scope"); + return TSDB_CODE_SUCCESS; } diff --git a/source/libs/planner/inc/planInt.h b/source/libs/planner/inc/planInt.h index 82abc5d1a9..092fe17411 100644 --- a/source/libs/planner/inc/planInt.h +++ b/source/libs/planner/inc/planInt.h @@ -43,6 +43,9 @@ int32_t splitLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan); int32_t scaleOutLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan, SQueryLogicPlan** pLogicPlan); int32_t createPhysiPlan(SPlanContext* pCxt, SQueryLogicPlan* pLogicPlan, SQueryPlan** pPlan, SArray* pExecNodeList); +bool isPartTableAgg(SAggLogicNode* pAgg); +bool isPartTableWinodw(SWindowLogicNode* pWindow); + #ifdef __cplusplus } #endif diff --git a/source/libs/planner/src/planOptimizer.c b/source/libs/planner/src/planOptimizer.c index 05f478b116..32721d8060 100644 --- a/source/libs/planner/src/planOptimizer.c +++ b/source/libs/planner/src/planOptimizer.c @@ -363,6 +363,18 @@ static void scanPathOptSetScanOrder(EScanOrder scanOrder, SScanLogicNode* pScan) } } +static void scanPathOptSetGroupOrderScan(SScanLogicNode* pScan) { + if (pScan->tableType != TSDB_SUPER_TABLE) return; + + if (pScan->node.pParent && nodeType(pScan->node.pParent) == QUERY_NODE_LOGIC_PLAN_AGG) { + SAggLogicNode* pAgg = (SAggLogicNode*)pScan->node.pParent; + bool withSlimit = pAgg->node.pSlimit != NULL || (pAgg->node.pParent && pAgg->node.pParent->pSlimit); + if (withSlimit && isPartTableAgg(pAgg)) { + pScan->groupOrderScan = pAgg->node.forceCreateNonBlockingOptr = true; + } + } +} + static int32_t scanPathOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan) { SOsdInfo info = {.scanOrder = SCAN_ORDER_ASC}; int32_t code = scanPathOptMatch(pCxt, pLogicSubplan->pNode, &info); @@ -371,6 +383,7 @@ static int32_t scanPathOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSub if (!pCxt->pPlanCxt->streamQuery) { scanPathOptSetScanOrder(info.scanOrder, info.pScan); } + scanPathOptSetGroupOrderScan(info.pScan); } if (TSDB_CODE_SUCCESS == code && (NULL != info.pDsoFuncs || NULL != info.pSdrFuncs)) { info.pScan->dataRequired = scanPathOptGetDataRequired(info.pSdrFuncs); @@ -1675,6 +1688,7 @@ static int32_t partTagsOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSub if (TSDB_CODE_SUCCESS == code) { if (QUERY_NODE_LOGIC_PLAN_AGG == pNode->pParent->type) { SAggLogicNode* pParent = (SAggLogicNode*)(pNode->pParent); + scanPathOptSetGroupOrderScan(pScan); pParent->hasGroupKeyOptimized = true; } diff --git a/source/libs/planner/src/planPhysiCreater.c b/source/libs/planner/src/planPhysiCreater.c index b3d94a5e47..1b92dcd2e7 100644 --- a/source/libs/planner/src/planPhysiCreater.c +++ b/source/libs/planner/src/planPhysiCreater.c @@ -447,6 +447,7 @@ static int32_t createScanPhysiNodeFinalize(SPhysiPlanContext* pCxt, SSubplan* pS pScanPhysiNode->uid = pScanLogicNode->tableId; pScanPhysiNode->suid = pScanLogicNode->stableId; pScanPhysiNode->tableType = pScanLogicNode->tableType; + pScanPhysiNode->groupOrderScan = pScanLogicNode->groupOrderScan; memcpy(&pScanPhysiNode->tableName, &pScanLogicNode->tableName, sizeof(SName)); if (NULL != pScanLogicNode->pTagCond) { pSubplan->pTagCond = nodesCloneNode(pScanLogicNode->pTagCond); @@ -880,6 +881,7 @@ static int32_t createAggPhysiNode(SPhysiPlanContext* pCxt, SNodeList* pChildren, pAgg->mergeDataBlock = (GROUP_ACTION_KEEP == pAggLogicNode->node.groupAction ? false : true); pAgg->groupKeyOptimized = pAggLogicNode->hasGroupKeyOptimized; + pAgg->node.forceCreateNonBlockingOptr = pAggLogicNode->node.forceCreateNonBlockingOptr; SNodeList* pPrecalcExprs = NULL; SNodeList* pGroupKeys = NULL; diff --git a/source/libs/planner/src/planSpliter.c b/source/libs/planner/src/planSpliter.c index 246ee13fb0..84a486649e 100644 --- a/source/libs/planner/src/planSpliter.c +++ b/source/libs/planner/src/planSpliter.c @@ -306,54 +306,6 @@ static bool stbSplIsTableCountQuery(SLogicNode* pNode) { return QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pChild) && SCAN_TYPE_TABLE_COUNT == ((SScanLogicNode*)pChild)->scanType; } -static SNodeList* stbSplGetPartKeys(SLogicNode* pNode) { - if (QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pNode)) { - return ((SScanLogicNode*)pNode)->pGroupTags; - } else if (QUERY_NODE_LOGIC_PLAN_PARTITION == nodeType(pNode)) { - return ((SPartitionLogicNode*)pNode)->pPartitionKeys; - } else { - return NULL; - } -} - -static bool stbSplHasPartTbname(SNodeList* pPartKeys) { - if (NULL == pPartKeys) { - return false; - } - SNode* pPartKey = NULL; - FOREACH(pPartKey, pPartKeys) { - if (QUERY_NODE_GROUPING_SET == nodeType(pPartKey)) { - pPartKey = nodesListGetNode(((SGroupingSetNode*)pPartKey)->pParameterList, 0); - } - if ((QUERY_NODE_FUNCTION == nodeType(pPartKey) && FUNCTION_TYPE_TBNAME == ((SFunctionNode*)pPartKey)->funcType) || - (QUERY_NODE_COLUMN == nodeType(pPartKey) && COLUMN_TYPE_TBNAME == ((SColumnNode*)pPartKey)->colType)) { - return true; - } - } - return false; -} - -static bool stbSplNotSystemScan(SLogicNode* pNode) { - if (QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pNode)) { - return SCAN_TYPE_SYSTEM_TABLE != ((SScanLogicNode*)pNode)->scanType; - } else if (QUERY_NODE_LOGIC_PLAN_PARTITION == nodeType(pNode)) { - return stbSplNotSystemScan((SLogicNode*)nodesListGetNode(pNode->pChildren, 0)); - } else { - return true; - } -} - -static bool stbSplIsPartTableAgg(SAggLogicNode* pAgg) { - if (1 != LIST_LENGTH(pAgg->node.pChildren)) { - return false; - } - if (NULL != pAgg->pGroupKeys) { - return stbSplHasPartTbname(pAgg->pGroupKeys) && - stbSplNotSystemScan((SLogicNode*)nodesListGetNode(pAgg->node.pChildren, 0)); - } - return stbSplHasPartTbname(stbSplGetPartKeys((SLogicNode*)nodesListGetNode(pAgg->node.pChildren, 0))); -} - static bool stbSplNeedSplit(bool streamQuery, SLogicNode* pNode) { switch (nodeType(pNode)) { case QUERY_NODE_LOGIC_PLAN_SCAN: @@ -364,7 +316,7 @@ static bool stbSplNeedSplit(bool streamQuery, SLogicNode* pNode) { return streamQuery ? false : stbSplIsMultiTbScanChild(streamQuery, pNode); case QUERY_NODE_LOGIC_PLAN_AGG: return (!stbSplHasGatherExecFunc(((SAggLogicNode*)pNode)->pAggFuncs) || - stbSplIsPartTableAgg((SAggLogicNode*)pNode)) && + isPartTableAgg((SAggLogicNode*)pNode)) && stbSplHasMultiTbScan(streamQuery, pNode) && !stbSplIsTableCountQuery(pNode); case QUERY_NODE_LOGIC_PLAN_WINDOW: return stbSplNeedSplitWindow(streamQuery, pNode); @@ -778,10 +730,6 @@ static int32_t stbSplSplitEvent(SSplitContext* pCxt, SStableSplitInfo* pInfo) { } } -static bool stbSplIsPartTableWinodw(SWindowLogicNode* pWindow) { - return stbSplHasPartTbname(stbSplGetPartKeys((SLogicNode*)nodesListGetNode(pWindow->node.pChildren, 0))); -} - static int32_t stbSplSplitWindowForCrossTable(SSplitContext* pCxt, SStableSplitInfo* pInfo) { switch (((SWindowLogicNode*)pInfo->pSplitNode)->winType) { case WINDOW_TYPE_INTERVAL: @@ -834,7 +782,7 @@ static int32_t stbSplSplitWindowForPartTable(SSplitContext* pCxt, SStableSplitIn } static int32_t stbSplSplitWindowNode(SSplitContext* pCxt, SStableSplitInfo* pInfo) { - if (stbSplIsPartTableWinodw((SWindowLogicNode*)pInfo->pSplitNode)) { + if (isPartTableWinodw((SWindowLogicNode*)pInfo->pSplitNode)) { return stbSplSplitWindowForPartTable(pCxt, pInfo); } else { return stbSplSplitWindowForCrossTable(pCxt, pInfo); @@ -920,7 +868,7 @@ static int32_t stbSplSplitAggNodeForCrossTable(SSplitContext* pCxt, SStableSplit } static int32_t stbSplSplitAggNode(SSplitContext* pCxt, SStableSplitInfo* pInfo) { - if (stbSplIsPartTableAgg((SAggLogicNode*)pInfo->pSplitNode)) { + if (isPartTableAgg((SAggLogicNode*)pInfo->pSplitNode)) { return stbSplSplitAggNodeForPartTable(pCxt, pInfo); } return stbSplSplitAggNodeForCrossTable(pCxt, pInfo); diff --git a/source/libs/planner/src/planUtil.c b/source/libs/planner/src/planUtil.c index 29e87b34ce..88086cde1d 100644 --- a/source/libs/planner/src/planUtil.c +++ b/source/libs/planner/src/planUtil.c @@ -321,3 +321,57 @@ int32_t adjustLogicNodeDataRequirement(SLogicNode* pNode, EDataOrderLevel requir } return code; } + +static bool stbNotSystemScan(SLogicNode* pNode) { + if (QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pNode)) { + return SCAN_TYPE_SYSTEM_TABLE != ((SScanLogicNode*)pNode)->scanType; + } else if (QUERY_NODE_LOGIC_PLAN_PARTITION == nodeType(pNode)) { + return stbNotSystemScan((SLogicNode*)nodesListGetNode(pNode->pChildren, 0)); + } else { + return true; + } +} + +static bool stbHasPartTbname(SNodeList* pPartKeys) { + if (NULL == pPartKeys) { + return false; + } + SNode* pPartKey = NULL; + FOREACH(pPartKey, pPartKeys) { + if (QUERY_NODE_GROUPING_SET == nodeType(pPartKey)) { + pPartKey = nodesListGetNode(((SGroupingSetNode*)pPartKey)->pParameterList, 0); + } + if ((QUERY_NODE_FUNCTION == nodeType(pPartKey) && FUNCTION_TYPE_TBNAME == ((SFunctionNode*)pPartKey)->funcType) || + (QUERY_NODE_COLUMN == nodeType(pPartKey) && COLUMN_TYPE_TBNAME == ((SColumnNode*)pPartKey)->colType)) { + return true; + } + } + return false; +} + +static SNodeList* stbSplGetPartKeys(SLogicNode* pNode) { + if (QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pNode)) { + return ((SScanLogicNode*)pNode)->pGroupTags; + } else if (QUERY_NODE_LOGIC_PLAN_PARTITION == nodeType(pNode)) { + return ((SPartitionLogicNode*)pNode)->pPartitionKeys; + } else { + return NULL; + } +} + +bool isPartTableAgg(SAggLogicNode* pAgg) { + if (1 != LIST_LENGTH(pAgg->node.pChildren)) { + return false; + } + if (NULL != pAgg->pGroupKeys) { + return stbHasPartTbname(pAgg->pGroupKeys) && + stbNotSystemScan((SLogicNode*)nodesListGetNode(pAgg->node.pChildren, 0)); + } + return stbHasPartTbname(stbSplGetPartKeys((SLogicNode*)nodesListGetNode(pAgg->node.pChildren, 0))); +} + +bool isPartTableWinodw(SWindowLogicNode* pWindow) { + return stbHasPartTbname(stbSplGetPartKeys((SLogicNode*)nodesListGetNode(pWindow->node.pChildren, 0))); +} + + diff --git a/source/libs/stream/inc/streamInt.h b/source/libs/stream/inc/streamInt.h index 51394143ac..fdfef66d63 100644 --- a/source/libs/stream/inc/streamInt.h +++ b/source/libs/stream/inc/streamInt.h @@ -31,6 +31,12 @@ typedef struct { void* timer; } SStreamGlobalEnv; +typedef struct { + SEpSet epset; + int32_t taskId; + SRpcMsg msg; +} SStreamContinueExecInfo; + extern SStreamGlobalEnv streamEnv; extern int32_t streamBackendId; extern int32_t streamBackendCfWrapperId; @@ -60,6 +66,12 @@ int32_t streamTaskGetNumOfDownstream(const SStreamTask* pTask); int32_t extractBlocksFromInputQ(SStreamTask* pTask, SStreamQueueItem** pInput, int32_t* numOfBlocks); SStreamQueueItem* streamMergeQueueItem(SStreamQueueItem* dst, SStreamQueueItem* pElem); +int32_t streamAddEndScanHistoryMsg(SStreamTask* pTask, SRpcHandleInfo* pRpcInfo, SStreamScanHistoryFinishReq* pReq); +int32_t streamNotifyUpstreamContinue(SStreamTask* pTask); + +extern int32_t streamBackendId; +extern int32_t streamBackendCfWrapperId; + #ifdef __cplusplus } #endif diff --git a/source/libs/stream/src/stream.c b/source/libs/stream/src/stream.c index cc9bc4295e..43c085a4fb 100644 --- a/source/libs/stream/src/stream.c +++ b/source/libs/stream/src/stream.c @@ -219,15 +219,16 @@ int32_t streamTaskEnqueueRetrieve(SStreamTask* pTask, SStreamRetrieveReq* pReq, // todo add log int32_t streamTaskOutputResultBlock(SStreamTask* pTask, SStreamDataBlock* pBlock) { int32_t code = 0; - if (pTask->outputType == TASK_OUTPUT__TABLE) { + int32_t type = pTask->outputInfo.type; + if (type == TASK_OUTPUT__TABLE) { pTask->tbSink.tbSinkFunc(pTask, pTask->tbSink.vnode, 0, pBlock->blocks); destroyStreamDataBlock(pBlock); - } else if (pTask->outputType == TASK_OUTPUT__SMA) { + } else if (type == TASK_OUTPUT__SMA) { pTask->smaSink.smaSink(pTask->smaSink.vnode, pTask->smaSink.smaId, pBlock->blocks); destroyStreamDataBlock(pBlock); } else { - ASSERT(pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH || pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH); - code = taosWriteQitem(pTask->outputQueue->queue, pBlock); + ASSERT(type == TASK_OUTPUT__FIXED_DISPATCH || type == TASK_OUTPUT__SHUFFLE_DISPATCH); + code = taosWriteQitem(pTask->outputInfo.queue->queue, pBlock); if (code != 0) { // todo failed to add it into the output queue, free it. return code; } @@ -274,6 +275,7 @@ int32_t streamProcessDispatchMsg(SStreamTask* pTask, SStreamDispatchReq* pReq, S tDeleteStreamDispatchReq(pReq); streamSchedExec(pTask); + return 0; } @@ -314,6 +316,9 @@ int32_t tAppendDataToInputQueue(SStreamTask* pTask, SStreamQueueItem* pItem) { return -1; } + int32_t msgLen = px->submit.msgLen; + int64_t ver = px->submit.ver; + int32_t code = taosWriteQitem(pTask->inputQueue->queue, pItem); if (code != TSDB_CODE_SUCCESS) { streamDataSubmitDestroy(px); @@ -321,8 +326,9 @@ int32_t tAppendDataToInputQueue(SStreamTask* pTask, SStreamQueueItem* pItem) { return code; } + // use the local variable to avoid the pItem be freed by other threads, since it has been put into queue already. qDebug("s-task:%s submit enqueue msgLen:%d ver:%" PRId64 ", total in queue:%d, size:%.2fMiB", pTask->id.idStr, - px->submit.msgLen, px->submit.ver, total, size + px->submit.msgLen / 1048576.0); + msgLen, ver, total, size + msgLen/1048576.0); } else if (type == STREAM_INPUT__DATA_BLOCK || type == STREAM_INPUT__DATA_RETRIEVE || type == STREAM_INPUT__REF_DATA_BLOCK) { if ((pTask->info.taskLevel == TASK_LEVEL__SOURCE) && (tInputQueueIsFull(pTask))) { diff --git a/source/libs/stream/src/streamBackendRocksdb.c b/source/libs/stream/src/streamBackendRocksdb.c index 0c385e2362..c5f955245e 100644 --- a/source/libs/stream/src/streamBackendRocksdb.c +++ b/source/libs/stream/src/streamBackendRocksdb.c @@ -2141,7 +2141,6 @@ int32_t streamStateSessionAddIfNotExist_rocksdb(SStreamState* pState, SSessionKe if (sessionRangeKeyCmpr(&searchKey, key) == 0) { memcpy(tmp, *pVal, valSize); taosMemoryFreeClear(*pVal); - streamStateSessionDel_rocksdb(pState, key); goto _end; } taosMemoryFreeClear(*pVal); @@ -2157,7 +2156,6 @@ int32_t streamStateSessionAddIfNotExist_rocksdb(SStreamState* pState, SSessionKe if (code == 0) { if (sessionRangeKeyCmpr(&searchKey, key) == 0) { memcpy(tmp, *pVal, valSize); - streamStateSessionDel_rocksdb(pState, key); goto _end; } } @@ -2215,14 +2213,12 @@ int32_t streamStateStateAddIfNotExist_rocksdb(SStreamState* pState, SSessionKey* if (code == 0) { if (key->win.skey <= tmpKey.win.skey && tmpKey.win.ekey <= key->win.ekey) { memcpy(tmp, *pVal, valSize); - streamStateSessionDel_rocksdb(pState, key); goto _end; } void* stateKey = (char*)(*pVal) + (valSize - keyDataLen); if (fn(pKeyData, stateKey) == true) { memcpy(tmp, *pVal, valSize); - streamStateSessionDel_rocksdb(pState, key); goto _end; } @@ -2238,7 +2234,6 @@ int32_t streamStateStateAddIfNotExist_rocksdb(SStreamState* pState, SSessionKey* void* stateKey = (char*)(*pVal) + (valSize - keyDataLen); if (fn(pKeyData, stateKey) == true) { memcpy(tmp, *pVal, valSize); - streamStateSessionDel_rocksdb(pState, key); goto _end; } } diff --git a/source/libs/stream/src/streamCheckpoint.c b/source/libs/stream/src/streamCheckpoint.c index 4e7a291d8b..352561bb43 100644 --- a/source/libs/stream/src/streamCheckpoint.c +++ b/source/libs/stream/src/streamCheckpoint.c @@ -151,7 +151,7 @@ static int32_t continueDispatchCheckpointBlock(SStreamDataBlock* pBlock, SStream pBlock->srcTaskId = pTask->id.taskId; pBlock->srcVgId = pTask->pMeta->vgId; - int32_t code = taosWriteQitem(pTask->outputQueue->queue, pBlock); + int32_t code = taosWriteQitem(pTask->outputInfo.queue->queue, pBlock); if (code == 0) { streamDispatchStreamBlock(pTask); } else { @@ -187,7 +187,7 @@ int32_t streamProcessCheckpointBlock(SStreamTask* pTask, SStreamDataBlock* pBloc //todo fix race condition: set the status and append checkpoint block int32_t taskLevel = pTask->info.taskLevel; if (taskLevel == TASK_LEVEL__SOURCE) { - if (pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH || pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + if (pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH || pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { qDebug("s-task:%s set childIdx:%d, and add checkpoint block into outputQ", id, pTask->info.selfChildId); continueDispatchCheckpointBlock(pBlock, pTask); } else { // only one task exists, no need to dispatch downstream info diff --git a/source/libs/stream/src/streamDispatch.c b/source/libs/stream/src/streamDispatch.c index e3839470ff..d145ff5636 100644 --- a/source/libs/stream/src/streamDispatch.c +++ b/source/libs/stream/src/streamDispatch.c @@ -32,18 +32,18 @@ typedef struct { SEpSet epset; } SStreamChkptReadyInfo; -static void doRetryDispatchData(void* param, void* tmrId); -static int32_t streamAddBlockIntoDispatchMsg(const SSDataBlock* pBlock, SStreamDispatchReq* pReq); -static int32_t doDispatchScanHistoryFinishMsg(SStreamTask* pTask, const SStreamScanHistoryFinishReq* pReq, int32_t vgId, - SEpSet* pEpSet); +static void doRetryDispatchData(void* param, void* tmrId); static int32_t doSendDispatchMsg(SStreamTask* pTask, const SStreamDispatchReq* pReq, int32_t vgId, SEpSet* pEpSet); +static int32_t streamAddBlockIntoDispatchMsg(const SSDataBlock* pBlock, SStreamDispatchReq* pReq); static int32_t streamSearchAndAddBlock(SStreamTask* pTask, SStreamDispatchReq* pReqs, SSDataBlock* pDataBlock, int32_t vgSz, int64_t groupId); +static int32_t doDispatchScanHistoryFinishMsg(SStreamTask* pTask, const SStreamScanHistoryFinishReq* pReq, int32_t vgId, + SEpSet* pEpSet); static void initRpcMsg(SRpcMsg* pMsg, int32_t msgType, void* pCont, int32_t contLen) { - pMsg->msgType = msgType; - pMsg->pCont = pCont; - pMsg->contLen = contLen; + pMsg->msgType = msgType; + pMsg->pCont = pCont; + pMsg->contLen = contLen; } int32_t tDecodeStreamDispatchReq(SDecoder* pDecoder, SStreamDispatchReq* pReq) { @@ -250,7 +250,7 @@ static int32_t doDispatchAllBlocks(SStreamTask* pTask, const SStreamDataBlock* p int32_t numOfBlocks = taosArrayGetSize(pData->blocks); ASSERT(numOfBlocks != 0); - if (pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH) { + if (pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH) { SStreamDispatchReq req = {0}; int32_t downstreamTaskId = pTask->fixedEpDispatcher.taskId; @@ -280,7 +280,7 @@ static int32_t doDispatchAllBlocks(SStreamTask* pTask, const SStreamDataBlock* p taosArrayDestroyP(req.data, taosMemoryFree); taosArrayDestroy(req.dataLen); return code; - } else if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { int32_t rspCnt = atomic_load_32(&pTask->shuffleDispatcher.waitingRspCnt); ASSERT(rspCnt == 0); @@ -360,10 +360,77 @@ void streamRetryDispatchStreamBlock(SStreamTask* pTask, int64_t waitDuration) { taosTmrReset(doRetryDispatchData, waitDuration, pTask, streamEnv.timer, &pTask->launchTaskTimer); } -int32_t streamDispatchStreamBlock(SStreamTask* pTask) { - ASSERT((pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH || pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH)); +int32_t streamSearchAndAddBlock(SStreamTask* pTask, SStreamDispatchReq* pReqs, SSDataBlock* pDataBlock, int32_t vgSz, + int64_t groupId) { + uint32_t hashValue = 0; + SArray* vgInfo = pTask->shuffleDispatcher.dbInfo.pVgroupInfos; + if (pTask->pNameMap == NULL) { + pTask->pNameMap = tSimpleHashInit(1024, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT)); + } - int32_t numOfElems = taosQueueItemSize(pTask->outputQueue->queue); + void* pVal = tSimpleHashGet(pTask->pNameMap, &groupId, sizeof(int64_t)); + if (pVal) { + SBlockName* pBln = (SBlockName*)pVal; + hashValue = pBln->hashValue; + if (!pDataBlock->info.parTbName[0]) { + memset(pDataBlock->info.parTbName, 0, TSDB_TABLE_NAME_LEN); + memcpy(pDataBlock->info.parTbName, pBln->parTbName, strlen(pBln->parTbName)); + } + } else { + char* ctbName = taosMemoryCalloc(1, TSDB_TABLE_FNAME_LEN); + if (ctbName == NULL) { + return -1; + } + + if (pDataBlock->info.parTbName[0]) { + snprintf(ctbName, TSDB_TABLE_NAME_LEN, "%s.%s", pTask->shuffleDispatcher.dbInfo.db, pDataBlock->info.parTbName); + } else { + buildCtbNameByGroupIdImpl(pTask->shuffleDispatcher.stbFullName, groupId, pDataBlock->info.parTbName); + snprintf(ctbName, TSDB_TABLE_NAME_LEN, "%s.%s", pTask->shuffleDispatcher.dbInfo.db, pDataBlock->info.parTbName); + } + + /*uint32_t hashValue = MurmurHash3_32(ctbName, strlen(ctbName));*/ + SUseDbRsp* pDbInfo = &pTask->shuffleDispatcher.dbInfo; + hashValue = + taosGetTbHashVal(ctbName, strlen(ctbName), pDbInfo->hashMethod, pDbInfo->hashPrefix, pDbInfo->hashSuffix); + taosMemoryFree(ctbName); + SBlockName bln = {0}; + bln.hashValue = hashValue; + memcpy(bln.parTbName, pDataBlock->info.parTbName, strlen(pDataBlock->info.parTbName)); + if (tSimpleHashGetSize(pTask->pNameMap) < MAX_BLOCK_NAME_NUM) { + tSimpleHashPut(pTask->pNameMap, &groupId, sizeof(int64_t), &bln, sizeof(SBlockName)); + } + } + + bool found = false; + // TODO: optimize search + int32_t j; + for (j = 0; j < vgSz; j++) { + SVgroupInfo* pVgInfo = taosArrayGet(vgInfo, j); + ASSERT(pVgInfo->vgId > 0); + + if (hashValue >= pVgInfo->hashBegin && hashValue <= pVgInfo->hashEnd) { + if (streamAddBlockIntoDispatchMsg(pDataBlock, &pReqs[j]) < 0) { + return -1; + } + + if (pReqs[j].blockNum == 0) { + atomic_add_fetch_32(&pTask->shuffleDispatcher.waitingRspCnt, 1); + } + + pReqs[j].blockNum++; + found = true; + break; + } + } + ASSERT(found); + return 0; +} + +int32_t streamDispatchStreamBlock(SStreamTask* pTask) { + ASSERT((pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH || pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH)); + + int32_t numOfElems = taosQueueItemSize(pTask->outputInfo.queue->queue); if (numOfElems > 0) { qDebug("s-task:%s try to dispatch intermediate block to downstream, elem in outputQ:%d", pTask->id.idStr, numOfElems); @@ -371,19 +438,19 @@ int32_t streamDispatchStreamBlock(SStreamTask* pTask) { // to make sure only one dispatch is running int8_t old = - atomic_val_compare_exchange_8(&pTask->outputStatus, TASK_OUTPUT_STATUS__NORMAL, TASK_OUTPUT_STATUS__WAIT); + atomic_val_compare_exchange_8(&pTask->outputInfo.status, TASK_OUTPUT_STATUS__NORMAL, TASK_OUTPUT_STATUS__WAIT); if (old != TASK_OUTPUT_STATUS__NORMAL) { qDebug("s-task:%s wait for dispatch rsp, not dispatch now, output status:%d", pTask->id.idStr, old); return 0; } ASSERT(pTask->msgInfo.pData == NULL); - qDebug("s-task:%s start to dispatch msg, set output status:%d", pTask->id.idStr, pTask->outputStatus); + qDebug("s-task:%s start to dispatch msg, set output status:%d", pTask->id.idStr, pTask->outputInfo.status); - SStreamDataBlock* pBlock = streamQueueNextItem(pTask->outputQueue); + SStreamDataBlock* pBlock = streamQueueNextItem(pTask->outputInfo.queue); if (pBlock == NULL) { - atomic_store_8(&pTask->outputStatus, TASK_OUTPUT_STATUS__NORMAL); - qDebug("s-task:%s not dispatch since no elems in outputQ, output status:%d", pTask->id.idStr, pTask->outputStatus); + atomic_store_8(&pTask->outputInfo.status, TASK_OUTPUT_STATUS__NORMAL); + qDebug("s-task:%s not dispatch since no elems in outputQ, output status:%d", pTask->id.idStr, pTask->outputInfo.status); return 0; } @@ -399,7 +466,7 @@ int32_t streamDispatchStreamBlock(SStreamTask* pTask) { } qDebug("s-task:%s failed to dispatch msg to downstream, code:%s, output status:%d, retry cnt:%d", pTask->id.idStr, - tstrerror(terrno), pTask->outputStatus, retryCount); + tstrerror(terrno), pTask->outputInfo.status, retryCount); // todo deal with only partially success dispatch case atomic_store_32(&pTask->shuffleDispatcher.waitingRspCnt, 0); @@ -425,10 +492,10 @@ int32_t streamDispatchScanHistoryFinishMsg(SStreamTask* pTask) { SStreamScanHistoryFinishReq req = {.streamId = pTask->id.streamId, .childId = pTask->info.selfChildId}; // serialize - if (pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH) { - req.taskId = pTask->fixedEpDispatcher.taskId; + if (pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH) { + req.downstreamTaskId = pTask->fixedEpDispatcher.taskId; doDispatchScanHistoryFinishMsg(pTask, &req, pTask->fixedEpDispatcher.nodeId, &pTask->fixedEpDispatcher.epSet); - } else if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { SArray* vgInfo = pTask->shuffleDispatcher.dbInfo.pVgroupInfos; int32_t numOfVgs = taosArrayGetSize(vgInfo); @@ -436,7 +503,7 @@ int32_t streamDispatchScanHistoryFinishMsg(SStreamTask* pTask) { pTask->id.idStr, numOfVgs, streamGetTaskStatusStr(pTask->status.taskStatus)); for (int32_t i = 0; i < numOfVgs; i++) { SVgroupInfo* pVgInfo = taosArrayGet(vgInfo, i); - req.taskId = pVgInfo->taskId; + req.downstreamTaskId = pVgInfo->taskId; doDispatchScanHistoryFinishMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet); } } @@ -561,14 +628,12 @@ int32_t doDispatchScanHistoryFinishMsg(SStreamTask* pTask, const SStreamScanHist tEncoderClear(&encoder); - msg.info.noResp = 1; initRpcMsg(&msg, TDMT_STREAM_SCAN_HISTORY_FINISH, buf, tlen + sizeof(SMsgHead)); tmsgSendReq(pEpSet, &msg); const char* pStatus = streamGetTaskStatusStr(pTask->status.taskStatus); - qDebug("s-task:%s status:%s dispatch scan-history-data finish msg to taskId:0x%x (vgId:%d)", pTask->id.idStr, pStatus, - pReq->taskId, vgId); - + qDebug("s-task:%s status:%s dispatch scan-history finish msg to taskId:0x%x (vgId:%d)", pTask->id.idStr, pStatus, + pReq->downstreamTaskId, vgId); return 0; } @@ -613,76 +678,9 @@ FAIL: return code; } -int32_t streamSearchAndAddBlock(SStreamTask* pTask, SStreamDispatchReq* pReqs, SSDataBlock* pDataBlock, int32_t vgSz, - int64_t groupId) { - uint32_t hashValue = 0; - SArray* vgInfo = pTask->shuffleDispatcher.dbInfo.pVgroupInfos; - if (pTask->pNameMap == NULL) { - pTask->pNameMap = tSimpleHashInit(1024, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT)); - } - - void* pVal = tSimpleHashGet(pTask->pNameMap, &groupId, sizeof(int64_t)); - if (pVal) { - SBlockName* pBln = (SBlockName*)pVal; - hashValue = pBln->hashValue; - if (!pDataBlock->info.parTbName[0]) { - memset(pDataBlock->info.parTbName, 0, TSDB_TABLE_NAME_LEN); - memcpy(pDataBlock->info.parTbName, pBln->parTbName, strlen(pBln->parTbName)); - } - } else { - char* ctbName = taosMemoryCalloc(1, TSDB_TABLE_FNAME_LEN); - if (ctbName == NULL) { - return -1; - } - - if (pDataBlock->info.parTbName[0]) { - snprintf(ctbName, TSDB_TABLE_NAME_LEN, "%s.%s", pTask->shuffleDispatcher.dbInfo.db, pDataBlock->info.parTbName); - } else { - buildCtbNameByGroupIdImpl(pTask->shuffleDispatcher.stbFullName, groupId, pDataBlock->info.parTbName); - snprintf(ctbName, TSDB_TABLE_NAME_LEN, "%s.%s", pTask->shuffleDispatcher.dbInfo.db, pDataBlock->info.parTbName); - } - - /*uint32_t hashValue = MurmurHash3_32(ctbName, strlen(ctbName));*/ - SUseDbRsp* pDbInfo = &pTask->shuffleDispatcher.dbInfo; - hashValue = - taosGetTbHashVal(ctbName, strlen(ctbName), pDbInfo->hashMethod, pDbInfo->hashPrefix, pDbInfo->hashSuffix); - taosMemoryFree(ctbName); - SBlockName bln = {0}; - bln.hashValue = hashValue; - memcpy(bln.parTbName, pDataBlock->info.parTbName, strlen(pDataBlock->info.parTbName)); - if (tSimpleHashGetSize(pTask->pNameMap) < MAX_BLOCK_NAME_NUM) { - tSimpleHashPut(pTask->pNameMap, &groupId, sizeof(int64_t), &bln, sizeof(SBlockName)); - } - } - - bool found = false; - // TODO: optimize search - int32_t j; - for (j = 0; j < vgSz; j++) { - SVgroupInfo* pVgInfo = taosArrayGet(vgInfo, j); - ASSERT(pVgInfo->vgId > 0); - - if (hashValue >= pVgInfo->hashBegin && hashValue <= pVgInfo->hashEnd) { - if (streamAddBlockIntoDispatchMsg(pDataBlock, &pReqs[j]) < 0) { - return -1; - } - - if (pReqs[j].blockNum == 0) { - atomic_add_fetch_32(&pTask->shuffleDispatcher.waitingRspCnt, 1); - } - - pReqs[j].blockNum++; - found = true; - break; - } - } - ASSERT(found); - return 0; -} - -void doRetryDispatchData(void* param, void* tmrId) { +static void doRetryDispatchData(void* param, void* tmrId) { SStreamTask* pTask = param; - ASSERT(pTask->outputStatus == TASK_OUTPUT_STATUS__WAIT); + ASSERT(pTask->outputInfo.status == TASK_OUTPUT_STATUS__WAIT); int32_t code = doDispatchAllBlocks(pTask, pTask->msgInfo.pData); if (code != TSDB_CODE_SUCCESS) { @@ -814,7 +812,7 @@ int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, i code); // there are other dispatch message not response yet - if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { int32_t leftRsp = atomic_sub_fetch_32(&pTask->shuffleDispatcher.waitingRspCnt, 1); qDebug("s-task:%s is shuffle, left waiting rsp %d", pTask->id.idStr, leftRsp); if (leftRsp > 0) { @@ -823,9 +821,9 @@ int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, i } pTask->msgInfo.retryCount = 0; - ASSERT(pTask->outputStatus == TASK_OUTPUT_STATUS__WAIT); + ASSERT(pTask->outputInfo.status == TASK_OUTPUT_STATUS__WAIT); - qDebug("s-task:%s output status is set to:%d", pTask->id.idStr, pTask->outputStatus); + qDebug("s-task:%s output status is set to:%d", pTask->id.idStr, pTask->outputInfo.status); // the input queue of the (down stream) task that receive the output data is full, // so the TASK_INPUT_STATUS_BLOCKED is rsp @@ -851,11 +849,101 @@ int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp, i } // now ready for next data output - atomic_store_8(&pTask->outputStatus, TASK_OUTPUT_STATUS__NORMAL); + atomic_store_8(&pTask->outputInfo.status, TASK_OUTPUT_STATUS__NORMAL); // otherwise, continue dispatch the first block to down stream task in pipeline streamDispatchStreamBlock(pTask); } return 0; -} \ No newline at end of file +} + +int32_t tEncodeCompleteHistoryDataMsg(SEncoder* pEncoder, const SStreamCompleteHistoryMsg* pReq) { + if (tStartEncode(pEncoder) < 0) return -1; + if (tEncodeI64(pEncoder, pReq->streamId) < 0) return -1; + if (tEncodeI32(pEncoder, pReq->downstreamId) < 0) return -1; + if (tEncodeI32(pEncoder, pReq->downstreamNode) < 0) return -1; + if (tEncodeI32(pEncoder, pReq->upstreamTaskId) < 0) return -1; + if (tEncodeI32(pEncoder, pReq->upstreamNodeId) < 0) return -1; + tEndEncode(pEncoder); + return pEncoder->pos; +} + +int32_t tDecodeCompleteHistoryDataMsg(SDecoder* pDecoder, SStreamCompleteHistoryMsg* pRsp) { + if (tStartDecode(pDecoder) < 0) return -1; + if (tDecodeI64(pDecoder, &pRsp->streamId) < 0) return -1; + if (tDecodeI32(pDecoder, &pRsp->downstreamId) < 0) return -1; + if (tDecodeI32(pDecoder, &pRsp->downstreamNode) < 0) return -1; + if (tDecodeI32(pDecoder, &pRsp->upstreamTaskId) < 0) return -1; + if (tDecodeI32(pDecoder, &pRsp->upstreamNodeId) < 0) return -1; + tEndDecode(pDecoder); + return 0; +} + +int32_t streamAddEndScanHistoryMsg(SStreamTask* pTask, SRpcHandleInfo* pRpcInfo, SStreamScanHistoryFinishReq* pReq) { + int32_t len = 0; + int32_t code = 0; + SEncoder encoder; + + SStreamCompleteHistoryMsg msg = { + .streamId = pReq->streamId, + .upstreamTaskId = pReq->upstreamTaskId, + .upstreamNodeId = pReq->upstreamNodeId, + .downstreamId = pReq->downstreamTaskId, + .downstreamNode = pTask->pMeta->vgId, + }; + + tEncodeSize(tEncodeCompleteHistoryDataMsg, &msg, len, code); + if (code < 0) { + return code; + } + + void* pBuf = rpcMallocCont(sizeof(SMsgHead) + len); + if (pBuf == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + ((SMsgHead*)pBuf)->vgId = htonl(pReq->upstreamNodeId); + + void* abuf = POINTER_SHIFT(pBuf, sizeof(SMsgHead)); + + tEncoderInit(&encoder, (uint8_t*)abuf, len); + tEncodeCompleteHistoryDataMsg(&encoder, &msg); + tEncoderClear(&encoder); + + SStreamChildEpInfo* pInfo = streamTaskGetUpstreamTaskEpInfo(pTask, pReq->upstreamTaskId); + + SStreamContinueExecInfo info = {.taskId = pReq->upstreamTaskId, .epset = pInfo->epSet}; + initRpcMsg(&info.msg, 0, pBuf, sizeof(SMsgHead) + len); + info.msg.info = *pRpcInfo; + + taosThreadMutexLock(&pTask->lock); + if (pTask->pRspMsgList == NULL) { + pTask->pRspMsgList = taosArrayInit(4, sizeof(SStreamContinueExecInfo)); + } + taosArrayPush(pTask->pRspMsgList, &info); + taosThreadMutexUnlock(&pTask->lock); + + int32_t num = taosArrayGetSize(pTask->pRspMsgList); + qDebug("s-task:%s add scan history finish rsp msg for task:0x%x, total:%d", pTask->id.idStr, pReq->upstreamTaskId, + num); + return TSDB_CODE_SUCCESS; +} + +int32_t streamNotifyUpstreamContinue(SStreamTask* pTask) { + ASSERT(pTask->info.taskLevel == TASK_LEVEL__AGG || pTask->info.taskLevel == TASK_LEVEL__SINK); + + int32_t num = taosArrayGetSize(pTask->pRspMsgList); + for (int32_t i = 0; i < num; ++i) { + SStreamContinueExecInfo* pInfo = taosArrayGet(pTask->pRspMsgList, i); + tmsgSendRsp(&pInfo->msg); + + qDebug("s-task:%s level:%d notify upstream:0x%x to continue process data from WAL", pTask->id.idStr, pTask->info.taskLevel, + pInfo->taskId); + } + + taosArrayClear(pTask->pRspMsgList); + qDebug("s-task:%s level:%d checkpoint ready msg sent to all %d upstreams", pTask->id.idStr, pTask->info.taskLevel, + num); + return 0; +} diff --git a/source/libs/stream/src/streamExec.c b/source/libs/stream/src/streamExec.c index ce2747a5c7..dfad0b2618 100644 --- a/source/libs/stream/src/streamExec.c +++ b/source/libs/stream/src/streamExec.c @@ -16,7 +16,11 @@ #include "streamInt.h" // maximum allowed processed block batches. One block may include several submit blocks -#define MAX_STREAM_RESULT_DUMP_THRESHOLD 100 +#define MAX_STREAM_EXEC_BATCH_NUM 32 +#define MIN_STREAM_EXEC_BATCH_NUM 4 +#define MAX_STREAM_RESULT_DUMP_THRESHOLD 100 + +static int32_t updateCheckPointInfo(SStreamTask* pTask); bool streamTaskShouldStop(const SStreamStatus* pStatus) { int32_t status = atomic_load_8((int8_t*)&pStatus->taskStatus); @@ -314,41 +318,45 @@ static void waitForTaskIdle(SStreamTask* pTask, SStreamTask* pStreamTask) { double el = (taosGetTimestampMs() - st) / 1000.0; if (el > 0) { - qDebug("s-task:%s wait for stream task:%s for %.2fs to handle all data in inputQ", pTask->id.idStr, + qDebug("s-task:%s wait for stream task:%s for %.2fs to be idle", pTask->id.idStr, pStreamTask->id.idStr, el); } } static int32_t streamTransferStateToStreamTask(SStreamTask* pTask) { - SStreamTask* pStreamTask = streamMetaAcquireTask(pTask->pMeta, pTask->streamTaskId.taskId); + SStreamMeta* pMeta = pTask->pMeta; + + SStreamTask* pStreamTask = streamMetaAcquireTask(pMeta, pTask->streamTaskId.taskId); if (pStreamTask == NULL) { - qError("s-task:%s failed to find related stream task:0x%x, it may have been destroyed or closed", - pTask->id.idStr, pTask->streamTaskId.taskId); + pTask->status.transferState = false; // reset this value, to avoid transfer state again + + qError("s-task:%s failed to find related stream task:0x%x, it may have been destroyed or closed", pTask->id.idStr, + pTask->streamTaskId.taskId); return TSDB_CODE_STREAM_TASK_NOT_EXIST; } else { qDebug("s-task:%s fill-history task end, update related stream task:%s info, transfer exec state", pTask->id.idStr, pStreamTask->id.idStr); } - pTask->status.transferState = false; // reset this value, to avoid transfer state again + ASSERT(pStreamTask->historyTaskId.taskId == pTask->id.taskId && pTask->status.transferState == true); - ASSERT(pStreamTask != NULL && pStreamTask->historyTaskId.taskId == pTask->id.taskId); STimeWindow* pTimeWindow = &pStreamTask->dataRange.window; // It must be halted for a source stream task, since when the related scan-history-data task start scan the history // for the step 2. For a agg task + int8_t status = pStreamTask->status.taskStatus; if (pStreamTask->info.taskLevel == TASK_LEVEL__SOURCE) { - ASSERT(pStreamTask->status.taskStatus == TASK_STATUS__HALT); + ASSERT(status == TASK_STATUS__HALT); } else { - ASSERT(pStreamTask->status.taskStatus == TASK_STATUS__NORMAL); + ASSERT(status == TASK_STATUS__SCAN_HISTORY); pStreamTask->status.taskStatus = TASK_STATUS__HALT; - qDebug("s-task:%s status: halt by related fill history task:%s", pStreamTask->id.idStr, pTask->id.idStr); + qDebug("s-task:%s halt by related fill-history task:%s", pStreamTask->id.idStr, pTask->id.idStr); } - // wait for the stream task to be idle + // wait for the stream task to handle all in the inputQ, and to be idle waitForTaskIdle(pTask, pStreamTask); - // In case of sink tasks, no need to be halted for them. + // In case of sink tasks, no need to halt them. // In case of source tasks and agg tasks, we should HALT them, and wait for them to be idle. And then, it's safe to // start the task state transfer procedure. // When a task is idle with halt status, all data in inputQ are consumed. @@ -370,10 +378,25 @@ static int32_t streamTransferStateToStreamTask(SStreamTask* pTask) { streamTaskReleaseState(pTask); streamTaskReloadState(pStreamTask); - streamSetStatusNormal(pStreamTask); + streamTaskResumeFromHalt(pStreamTask); + + pTask->status.taskStatus = TASK_STATUS__DROPPING; + qDebug("s-task:%s fill-history task set status to be dropping, save the state into disk", pTask->id.idStr); + + // save to disk + taosWLockLatch(&pMeta->lock); + streamMetaSaveTask(pMeta, pTask); + streamMetaSaveTask(pMeta, pStreamTask); + if (streamMetaCommit(pMeta) < 0) { + // persist to disk + } + taosWUnLockLatch(&pMeta->lock); + + // pause allowed + streamTaskEnablePause(pStreamTask); streamSchedExec(pStreamTask); - streamMetaReleaseTask(pTask->pMeta, pStreamTask); + streamMetaReleaseTask(pMeta, pStreamTask); return TSDB_CODE_SUCCESS; } @@ -427,6 +450,106 @@ static void doSetStreamInputBlock(SStreamTask* pTask, const void* pInput, int64_ } } +static int32_t extractMsgFromInputQ(SStreamTask* pTask, SStreamQueueItem** pInput, int32_t* numOfBlocks, + const char* id) { + int32_t retryTimes = 0; + int32_t MAX_RETRY_TIMES = 5; + + while (1) { + if (streamTaskShouldPause(&pTask->status)) { + qDebug("s-task:%s task should pause, input blocks:%d", pTask->id.idStr, *numOfBlocks); + return TSDB_CODE_SUCCESS; + } + + SStreamQueueItem* qItem = streamQueueNextItem(pTask->inputQueue); + if (qItem == NULL) { + if (pTask->info.taskLevel == TASK_LEVEL__SOURCE && (++retryTimes) < MAX_RETRY_TIMES) { + taosMsleep(10); + qDebug("===stream===try again batchSize:%d, retry:%d", *numOfBlocks, retryTimes); + continue; + } + + qDebug("===stream===break batchSize:%d", *numOfBlocks); + return TSDB_CODE_SUCCESS; + } + + // do not merge blocks for sink node + if (pTask->info.taskLevel == TASK_LEVEL__SINK) { + *numOfBlocks = 1; + *pInput = qItem; + return TSDB_CODE_SUCCESS; + } + + if (*pInput == NULL) { + ASSERT((*numOfBlocks) == 0); + *pInput = qItem; + } else { + // todo we need to sort the data block, instead of just appending into the array list. + void* newRet = streamMergeQueueItem(*pInput, qItem); + if (newRet == NULL) { + qError("s-task:%s failed to merge blocks from inputQ, numOfBlocks:%d", id, *numOfBlocks); + streamQueueProcessFail(pTask->inputQueue); + return TSDB_CODE_SUCCESS; + } + + *pInput = newRet; + } + + *numOfBlocks += 1; + streamQueueProcessSuccess(pTask->inputQueue); + + if (*numOfBlocks >= MAX_STREAM_EXEC_BATCH_NUM) { + qDebug("s-task:%s batch size limit:%d reached, start to process blocks", id, MAX_STREAM_EXEC_BATCH_NUM); + return TSDB_CODE_SUCCESS; + } + } +} + +///** +// * todo: the batch of blocks should be tuned dynamic, according to the total elapsed time of each batch of blocks, the +// * appropriate batch of blocks should be handled in 5 to 10 sec. +// */ +//int32_t streamExecForAll(SStreamTask* pTask) { +// const char* id = pTask->id.idStr; +// +// while (1) { +// int32_t batchSize = 0; +// SStreamQueueItem* pInput = NULL; +// +// // merge multiple input data if possible in the input queue. +// qDebug("s-task:%s start to extract data block from inputQ", id); +// +// /*int32_t code = */extractMsgFromInputQ(pTask, &pInput, &batchSize, id); +// if (pInput == NULL) { +// ASSERT(batchSize == 0); +// if (pTask->info.fillHistory && pTask->status.transferState) { +// int32_t code = streamTransferStateToStreamTask(pTask); +// if (code != TSDB_CODE_SUCCESS) { // todo handle this +// return 0; +// } +// } +// +// break; +// } +// +// if (pTask->info.taskLevel == TASK_LEVEL__SINK) { +// ASSERT(pInput->type == STREAM_INPUT__DATA_BLOCK); +// qDebug("s-task:%s sink task start to sink %d blocks", id, batchSize); +// streamTaskOutputResultBlock(pTask, (SStreamDataBlock*)pInput); +// continue; +// } +// +// int64_t st = taosGetTimestampMs(); +// qDebug("s-task:%s start to process batch of blocks, num:%d", id, batchSize); +// +//// { +// // set input +// const SStreamQueueItem* pItem = pInput; +// qDebug("s-task:%s start to process batch of blocks, num:%d, type:%d", id, numOfBlocks, pItem->type); +// +// int64_t ver = pTask->chkInfo.checkpointVer; +// doSetStreamInputBlock(pTask, pInput, &ver, id); + /** * todo: the batch of blocks should be tuned dynamic, according to the total elapsed time of each batch of blocks, the * appropriate batch of blocks should be handled in 5 to 10 sec. @@ -508,21 +631,6 @@ int32_t streamExecForAll(SStreamTask* pTask) { } bool streamTaskIsIdle(const SStreamTask* pTask) { - int32_t numOfItems = taosQueueItemSize(pTask->inputQueue->queue); - if (numOfItems > 0) { - return false; - } - - numOfItems = taosQallItemSize(pTask->inputQueue->qall); - if (numOfItems > 0) { - return false; - } - - // blocked by downstream task - if (pTask->inputStatus == TASK_INPUT_STATUS__BLOCKED) { - return false; - } - return (pTask->status.schedStatus == TASK_SCHED_STATUS__INACTIVE); } diff --git a/source/libs/stream/src/streamRecover.c b/source/libs/stream/src/streamRecover.c index 0740566d15..9c747d949d 100644 --- a/source/libs/stream/src/streamRecover.c +++ b/source/libs/stream/src/streamRecover.c @@ -18,6 +18,18 @@ #include "wal.h" #include "trpc.h" +static void launchFillHistoryTask(SStreamTask* pTask); +static void streamTaskSetRangeStreamCalc(SStreamTask* pTask); + +static void streamTaskSetForReady(SStreamTask* pTask, int32_t numOfReqs) { + ASSERT(pTask->status.downstreamReady == 0); + pTask->status.downstreamReady = 1; + int64_t el = (taosGetTimestampMs() - pTask->initTs); + + qDebug("s-task:%s all %d downstream ready, init completed, elapsed time:%dms, task status:%s", + pTask->id.idStr, numOfReqs, (int32_t) el, streamGetTaskStatusStr(pTask->status.taskStatus)); +} + int32_t streamStartRecoverTask(SStreamTask* pTask, int8_t igUntreated) { SStreamScanHistoryReq req; streamBuildSourceRecover1Req(pTask, &req, igUntreated); @@ -53,10 +65,6 @@ const char* streamGetTaskStatusStr(int32_t status) { static int32_t doLaunchScanHistoryTask(SStreamTask* pTask) { SVersionRange* pRange = &pTask->dataRange.range; - - qDebug("s-task:%s vgId:%d status:%s, start scan-history-data task, verRange:%" PRId64 " - %" PRId64, pTask->id.idStr, - pTask->info.nodeId, streamGetTaskStatusStr(pTask->status.taskStatus), pRange->minVer, pRange->maxVer); - streamSetParamForScanHistory(pTask); streamSetParamForStreamScannerStep1(pTask, pRange, &pTask->dataRange.window); @@ -75,19 +83,17 @@ int32_t streamTaskLaunchScanHistory(SStreamTask* pTask) { walReaderGetCurrentVer(pTask->exec.pWalReader)); } } else if (pTask->info.taskLevel == TASK_LEVEL__AGG) { - streamSetStatusNormal(pTask); streamSetParamForScanHistory(pTask); - streamAggScanHistoryPrepare(pTask); + streamTaskScanHistoryPrepare(pTask); } else if (pTask->info.taskLevel == TASK_LEVEL__SINK) { - streamSetStatusNormal(pTask); - qDebug("s-task:%s sink task convert to normal immediately", pTask->id.idStr); + qDebug("s-task:%s sink task do nothing to handle scan-history", pTask->id.idStr); + streamTaskScanHistoryPrepare(pTask); } - return 0; } // check status -int32_t streamTaskCheckDownstreamTasks(SStreamTask* pTask) { +int32_t streamTaskDoCheckDownstreamTasks(SStreamTask* pTask) { SHistDataRange* pRange = &pTask->dataRange; STimeWindow* pWindow = &pRange->window; @@ -99,7 +105,7 @@ int32_t streamTaskCheckDownstreamTasks(SStreamTask* pTask) { }; // serialize - if (pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH) { + if (pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH) { req.reqId = tGenIdPI64(); req.downstreamNodeId = pTask->fixedEpDispatcher.nodeId; req.downstreamTaskId = pTask->fixedEpDispatcher.taskId; @@ -111,7 +117,7 @@ int32_t streamTaskCheckDownstreamTasks(SStreamTask* pTask) { pWindow->skey, pWindow->ekey, req.reqId); streamDispatchCheckMsg(pTask, &req, pTask->fixedEpDispatcher.nodeId, &pTask->fixedEpDispatcher.epSet); - } else if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { SArray* vgInfo = pTask->shuffleDispatcher.dbInfo.pVgroupInfos; int32_t numOfVgs = taosArrayGetSize(vgInfo); @@ -132,11 +138,13 @@ int32_t streamTaskCheckDownstreamTasks(SStreamTask* pTask) { streamDispatchCheckMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet); } } else { - pTask->status.downstreamReady = 1; - qDebug("s-task:%s (vgId:%d) set downstream checked since no downstream, try to launch scan-history-data, status:%s", - pTask->id.idStr, pTask->info.nodeId, streamGetTaskStatusStr(pTask->status.taskStatus)); + qDebug("s-task:%s (vgId:%d) set downstream ready, since no downstream", pTask->id.idStr, pTask->info.nodeId); + streamTaskSetForReady(pTask, 0); + streamTaskSetRangeStreamCalc(pTask); streamTaskLaunchScanHistory(pTask); + + launchFillHistoryTask(pTask); } return 0; @@ -156,9 +164,9 @@ int32_t streamRecheckDownstream(SStreamTask* pTask, const SStreamTaskCheckRsp* p qDebug("s-task:%s (vgId:%d) check downstream task:0x%x (vgId:%d) (recheck)", pTask->id.idStr, pTask->info.nodeId, req.downstreamTaskId, req.downstreamNodeId); - if (pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH) { + if (pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH) { streamDispatchCheckMsg(pTask, &req, pRsp->downstreamNodeId, &pTask->fixedEpDispatcher.epSet); - } else if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { SArray* vgInfo = pTask->shuffleDispatcher.dbInfo.pVgroupInfos; int32_t numOfVgs = taosArrayGetSize(vgInfo); @@ -174,7 +182,28 @@ int32_t streamRecheckDownstream(SStreamTask* pTask, const SStreamTaskCheckRsp* p } int32_t streamTaskCheckStatus(SStreamTask* pTask) { - return atomic_load_8(&pTask->status.taskStatus) == TASK_STATUS__NORMAL? 1:0; + return (pTask->status.downstreamReady == 1)? 1:0; +} + +static void doProcessDownstreamReadyRsp(SStreamTask* pTask, int32_t numOfReqs) { + streamTaskSetForReady(pTask, numOfReqs); + const char* id = pTask->id.idStr; + + int8_t status = pTask->status.taskStatus; + const char* str = streamGetTaskStatusStr(status); + + ASSERT(status == TASK_STATUS__SCAN_HISTORY || status == TASK_STATUS__NORMAL); + streamTaskSetRangeStreamCalc(pTask); + + if (status == TASK_STATUS__SCAN_HISTORY) { + qDebug("s-task:%s enter into scan-history data stage, status:%s", id, str); + streamTaskLaunchScanHistory(pTask); + } else { + qDebug("s-task:%s downstream tasks are ready, now ready for data from wal, status:%s", id, str); + } + + // when current stream task is ready, check the related fill history task. + launchFillHistoryTask(pTask); } int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRsp) { @@ -182,7 +211,7 @@ int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRs const char* id = pTask->id.idStr; if (pRsp->status == 1) { - if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { bool found = false; int32_t numOfReqs = taosArrayGetSize(pTask->checkReqIds); @@ -204,41 +233,20 @@ int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRs if (left == 0) { taosArrayDestroy(pTask->checkReqIds); pTask->checkReqIds = NULL; - pTask->status.downstreamReady = 1; - if (pTask->status.taskStatus == TASK_STATUS__SCAN_HISTORY) { - qDebug("s-task:%s all %d downstream tasks are ready, now enter into scan-history-data stage, status:%s", id, - numOfReqs, streamGetTaskStatusStr(pTask->status.taskStatus)); - streamTaskLaunchScanHistory(pTask); - } else { - ASSERT(pTask->status.taskStatus == TASK_STATUS__NORMAL); - qDebug("s-task:%s fixed downstream task is ready, now ready for data from wal, status:%s", id, - streamGetTaskStatusStr(pTask->status.taskStatus)); - } + doProcessDownstreamReadyRsp(pTask, numOfReqs); } else { int32_t total = taosArrayGetSize(pTask->shuffleDispatcher.dbInfo.pVgroupInfos); qDebug("s-task:%s (vgId:%d) recv check rsp from task:0x%x (vgId:%d) status:%d, total:%d not ready:%d", id, pRsp->upstreamNodeId, pRsp->downstreamTaskId, pRsp->downstreamNodeId, pRsp->status, total, left); } } else { - ASSERT(pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH); + ASSERT(pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH); if (pRsp->reqId != pTask->checkReqId) { return -1; } - // set the downstream tasks have been checked flag - ASSERT(pTask->status.downstreamReady == 0); - pTask->status.downstreamReady = 1; - - ASSERT(pTask->status.taskStatus == TASK_STATUS__SCAN_HISTORY || pTask->status.taskStatus == TASK_STATUS__NORMAL); - if (pTask->status.taskStatus == TASK_STATUS__SCAN_HISTORY) { - qDebug("s-task:%s fixed downstream task is ready, now enter into scan-history-data stage, status:%s", id, - streamGetTaskStatusStr(pTask->status.taskStatus)); - streamTaskLaunchScanHistory(pTask); - } else { - qDebug("s-task:%s fixed downstream task is ready, ready for data from inputQ, status:%s", id, - streamGetTaskStatusStr(pTask->status.taskStatus)); - } + doProcessDownstreamReadyRsp(pTask, 1); } } else { // not ready, wait for 100ms and retry qDebug("s-task:%s downstream taskId:0x%x (vgId:%d) not ready, wait for 100ms and retry", id, pRsp->downstreamTaskId, @@ -251,6 +259,32 @@ int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRs return 0; } +int32_t streamSendCheckRsp(const SStreamMeta* pMeta, const SStreamTaskCheckReq* pReq, SStreamTaskCheckRsp* pRsp, + SRpcHandleInfo *pRpcInfo, int32_t taskId) { + SEncoder encoder; + int32_t code; + int32_t len; + + tEncodeSize(tEncodeStreamTaskCheckRsp, pRsp, len, code); + if (code < 0) { + qError("vgId:%d failed to encode task check rsp, s-task:0x%x", pMeta->vgId, taskId); + return -1; + } + + void* buf = rpcMallocCont(sizeof(SMsgHead) + len); + ((SMsgHead*)buf)->vgId = htonl(pReq->upstreamNodeId); + + void* abuf = POINTER_SHIFT(buf, sizeof(SMsgHead)); + tEncoderInit(&encoder, (uint8_t*)abuf, len); + tEncodeStreamTaskCheckRsp(&encoder, pRsp); + tEncoderClear(&encoder); + + SRpcMsg rspMsg = {.code = 0, .pCont = buf, .contLen = sizeof(SMsgHead) + len, .info = *pRpcInfo}; + + tmsgSendRsp(&rspMsg); + return 0; +} + // common int32_t streamSetParamForScanHistory(SStreamTask* pTask) { qDebug("s-task:%s set operator option for scan-history-data", pTask->id.idStr); @@ -258,7 +292,7 @@ int32_t streamSetParamForScanHistory(SStreamTask* pTask) { } int32_t streamRestoreParam(SStreamTask* pTask) { - qDebug("s-task:%s restore operator param after scan-history-data", pTask->id.idStr); + qDebug("s-task:%s restore operator param after scan-history", pTask->id.idStr); return qRestoreStreamOperatorOption(pTask->exec.pExecutor); } @@ -333,7 +367,7 @@ static int32_t doDispatchTransferMsg(SStreamTask* pTask, const SStreamTransferRe tmsgSendReq(pEpSet, &msg); qDebug("s-task:%s level:%d, status:%s dispatch transfer state msg to taskId:0x%x (vgId:%d)", pTask->id.idStr, - pTask->info.taskLevel, streamGetTaskStatusStr(pTask->status.taskStatus), pReq->taskId, vgId); + pTask->info.taskLevel, streamGetTaskStatusStr(pTask->status.taskStatus), pReq->downstreamTaskId, vgId); return 0; } @@ -342,16 +376,16 @@ int32_t streamDispatchTransferStateMsg(SStreamTask* pTask) { SStreamTransferReq req = { .streamId = pTask->id.streamId, .childId = pTask->info.selfChildId }; // serialize - if (pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH) { - req.taskId = pTask->fixedEpDispatcher.taskId; + if (pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH) { + req.downstreamTaskId = pTask->fixedEpDispatcher.taskId; doDispatchTransferMsg(pTask, &req, pTask->fixedEpDispatcher.nodeId, &pTask->fixedEpDispatcher.epSet); - } else if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { SArray* vgInfo = pTask->shuffleDispatcher.dbInfo.pVgroupInfos; int32_t numOfVgs = taosArrayGetSize(vgInfo); for (int32_t i = 0; i < numOfVgs; i++) { SVgroupInfo* pVgInfo = taosArrayGet(vgInfo, i); - req.taskId = pVgInfo->taskId; + req.downstreamTaskId = pVgInfo->taskId; doDispatchTransferMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet); } } @@ -360,10 +394,11 @@ int32_t streamDispatchTransferStateMsg(SStreamTask* pTask) { } // agg -int32_t streamAggScanHistoryPrepare(SStreamTask* pTask) { - pTask->numOfWaitingUpstream = taosArrayGetSize(pTask->pUpstreamInfoList); - qDebug("s-task:%s agg task is ready and wait for %d upstream tasks complete scan-history procedure", pTask->id.idStr, - pTask->numOfWaitingUpstream); +int32_t streamTaskScanHistoryPrepare(SStreamTask* pTask) { + pTask->numOfWaitingUpstream = taosArrayGetSize(pTask->pUpstreamEpInfoList); + qDebug("s-task:%s level:%d task wait for %d upstream tasks complete scan-history procedure, status:%s", + pTask->id.idStr, pTask->info.taskLevel, pTask->numOfWaitingUpstream, + streamGetTaskStatusStr(pTask->status.taskStatus)); return 0; } @@ -379,27 +414,64 @@ int32_t streamAggUpstreamScanHistoryFinish(SStreamTask* pTask) { return 0; } -int32_t streamProcessScanHistoryFinishReq(SStreamTask* pTask, int32_t childId) { - if (pTask->info.taskLevel == TASK_LEVEL__AGG) { - int32_t left = atomic_sub_fetch_32(&pTask->numOfWaitingUpstream, 1); - ASSERT(left >= 0); +int32_t streamProcessScanHistoryFinishReq(SStreamTask* pTask, SStreamScanHistoryFinishReq* pReq, + SRpcHandleInfo* pRpcInfo) { + int32_t taskLevel = pTask->info.taskLevel; + ASSERT(taskLevel == TASK_LEVEL__AGG || taskLevel == TASK_LEVEL__SINK); - if (left == 0) { - int32_t numOfTasks = taosArrayGetSize(pTask->pUpstreamInfoList); - qDebug("s-task:%s all %d upstream tasks finish scan-history data, set param for agg task for stream data", - pTask->id.idStr, numOfTasks); + // sink node do not send end of scan history msg to its upstream, which is agg task. + streamAddEndScanHistoryMsg(pTask, pRpcInfo, pReq); + int32_t left = atomic_sub_fetch_32(&pTask->numOfWaitingUpstream, 1); + ASSERT(left >= 0); + + if (left == 0) { + int32_t numOfTasks = taosArrayGetSize(pTask->pUpstreamEpInfoList); + qDebug( + "s-task:%s all %d upstream tasks finish scan-history data, set param for agg task for stream data and send " + "rsp to all upstream tasks", + pTask->id.idStr, numOfTasks); + + if (pTask->info.taskLevel == TASK_LEVEL__AGG) { streamAggUpstreamScanHistoryFinish(pTask); - } else { - qDebug("s-task:%s receive scan-history data finish msg from upstream (index:%d), unfinished:%d", - pTask->id.idStr, childId, left); } + streamNotifyUpstreamContinue(pTask); + + // sink node does not receive the pause msg from mnode, so does not need enable it + if (pTask->info.taskLevel == TASK_LEVEL__AGG) { + streamTaskEnablePause(pTask); + } + } else { + qDebug("s-task:%s receive scan-history data finish msg from upstream:0x%x(index:%d), unfinished:%d", + pTask->id.idStr, pReq->upstreamTaskId, pReq->childId, left); } return 0; } +int32_t streamProcessScanHistoryFinishRsp(SStreamTask* pTask) { + ASSERT(pTask->status.taskStatus == TASK_STATUS__SCAN_HISTORY); + SStreamMeta* pMeta = pTask->pMeta; + + // execute in the scan history complete call back msg, ready to process data from inputQ + streamSetStatusNormal(pTask); + atomic_store_8(&pTask->status.schedStatus, TASK_SCHED_STATUS__INACTIVE); + + taosWLockLatch(&pMeta->lock); + streamMetaSaveTask(pMeta, pTask); + taosWUnLockLatch(&pMeta->lock); + + // history data scan in the stream time window finished, now let's enable the pause + streamTaskEnablePause(pTask); + + if (pTask->info.taskLevel == TASK_LEVEL__SOURCE) { + streamSchedExec(pTask); + } + + return TSDB_CODE_SUCCESS; +} + static void doCheckDownstreamStatus(SStreamTask* pTask, SStreamTask* pHTask) { pHTask->dataRange.range.minVer = 0; pHTask->dataRange.range.maxVer = pTask->chkInfo.currentVer; @@ -414,7 +486,7 @@ static void doCheckDownstreamStatus(SStreamTask* pTask, SStreamTask* pHTask) { } // check if downstream tasks have been ready - streamTaskCheckDownstreamTasks(pHTask); + streamTaskDoCheckDownstreamTasks(pHTask); } typedef struct SStreamTaskRetryInfo { @@ -480,7 +552,7 @@ static void tryLaunchHistoryTask(void* param, void* tmrId) { // todo fix the bug: 2. race condition // an fill history task needs to be started. -int32_t streamCheckHistoryTaskDownstream(SStreamTask* pTask) { +int32_t streamLaunchFillHistoryTask(SStreamTask* pTask) { SStreamMeta* pMeta = pTask->pMeta; int32_t hTaskId = pTask->historyTaskId.taskId; @@ -518,7 +590,6 @@ int32_t streamCheckHistoryTaskDownstream(SStreamTask* pTask) { } int32_t streamTaskScanHistoryDataComplete(SStreamTask* pTask) { - SStreamMeta* pMeta = pTask->pMeta; if (atomic_load_8(&pTask->status.taskStatus) == TASK_STATUS__DROPPING) { return 0; } @@ -535,16 +606,6 @@ int32_t streamTaskScanHistoryDataComplete(SStreamTask* pTask) { return -1; } - ASSERT(pTask->status.taskStatus == TASK_STATUS__SCAN_HISTORY); - - // ready to process data from inputQ - streamSetStatusNormal(pTask); - atomic_store_8(&pTask->status.schedStatus, TASK_SCHED_STATUS__INACTIVE); - - taosWLockLatch(&pMeta->lock); - streamMetaSaveTask(pMeta, pTask); - taosWUnLockLatch(&pMeta->lock); - return 0; } @@ -641,54 +702,182 @@ int32_t tDecodeStreamTaskCheckRsp(SDecoder* pDecoder, SStreamTaskCheckRsp* pRsp) int32_t tEncodeStreamScanHistoryFinishReq(SEncoder* pEncoder, const SStreamScanHistoryFinishReq* pReq) { if (tStartEncode(pEncoder) < 0) return -1; if (tEncodeI64(pEncoder, pReq->streamId) < 0) return -1; - if (tEncodeI32(pEncoder, pReq->taskId) < 0) return -1; + if (tEncodeI32(pEncoder, pReq->upstreamTaskId) < 0) return -1; + if (tEncodeI32(pEncoder, pReq->upstreamNodeId) < 0) return -1; + if (tEncodeI32(pEncoder, pReq->downstreamTaskId) < 0) return -1; if (tEncodeI32(pEncoder, pReq->childId) < 0) return -1; tEndEncode(pEncoder); return pEncoder->pos; } + int32_t tDecodeStreamScanHistoryFinishReq(SDecoder* pDecoder, SStreamScanHistoryFinishReq* pReq) { if (tStartDecode(pDecoder) < 0) return -1; if (tDecodeI64(pDecoder, &pReq->streamId) < 0) return -1; - if (tDecodeI32(pDecoder, &pReq->taskId) < 0) return -1; + if (tDecodeI32(pDecoder, &pReq->upstreamTaskId) < 0) return -1; + if (tDecodeI32(pDecoder, &pReq->upstreamNodeId) < 0) return -1; + if (tDecodeI32(pDecoder, &pReq->downstreamTaskId) < 0) return -1; if (tDecodeI32(pDecoder, &pReq->childId) < 0) return -1; tEndDecode(pDecoder); return 0; } -// todo handle race condition, this task may be destroyed -void streamPrepareNdoCheckDownstream(SStreamTask* pTask) { - if (pTask->info.fillHistory) { - qDebug("s-task:%s fill history task, wait for being launched", pTask->id.idStr); - } else { - // calculate the correct start time window, and start the handle the history data for the main task. - if (pTask->historyTaskId.taskId != 0) { - // check downstream tasks for associated scan-history-data tasks - streamCheckHistoryTaskDownstream(pTask); - - // launch current task - SHistDataRange* pRange = &pTask->dataRange; - int64_t ekey = pRange->window.ekey + 1; - int64_t ver = pRange->range.minVer; - - pRange->window.skey = ekey; - pRange->window.ekey = INT64_MAX; - pRange->range.minVer = 0; - pRange->range.maxVer = ver; - - qDebug("s-task:%s level:%d related fill-history task exists, update stream time window:%" PRId64 " - %" PRId64 - ", ver range:%" PRId64 " - %" PRId64, - pTask->id.idStr, pTask->info.taskLevel, pRange->window.skey, pRange->window.ekey, pRange->range.minVer, - pRange->range.maxVer); +void streamTaskSetRangeStreamCalc(SStreamTask* pTask) { + if (pTask->historyTaskId.taskId == 0) { + SHistDataRange* pRange = &pTask->dataRange; + if (pTask->info.fillHistory == 1) { + qDebug("s-task:%s fill-history task, time window:%" PRId64 "-%" PRId64 ", verRange:%" PRId64 + "-%" PRId64, + pTask->id.idStr, pRange->window.skey, pRange->window.ekey, pRange->range.minVer, pRange->range.maxVer); } else { - SHistDataRange* pRange = &pTask->dataRange; - qDebug("s-task:%s no related fill-history task, stream time window:%" PRId64 " - %" PRId64 - ", ver range:%" PRId64 " - %" PRId64, + qDebug("s-task:%s no related fill-history task, stream time window:%" PRId64 "-%" PRId64 ", verRange:%" PRId64 + "-%" PRId64, pTask->id.idStr, pRange->window.skey, pRange->window.ekey, pRange->range.minVer, pRange->range.maxVer); } + } else { + SHistDataRange* pRange = &pTask->dataRange; - ASSERT(pTask->status.downstreamReady == 0); + int64_t ekey = pRange->window.ekey + 1; + int64_t ver = pRange->range.minVer; - // check downstream tasks for itself - streamTaskCheckDownstreamTasks(pTask); + pRange->window.skey = ekey; + pRange->window.ekey = INT64_MAX; + pRange->range.minVer = 0; + pRange->range.maxVer = ver; + + qDebug("s-task:%s level:%d related fill-history task exists, update stream calc time window:%" PRId64 " - %" PRId64 + ", verRang:%" PRId64 " - %" PRId64, + pTask->id.idStr, pTask->info.taskLevel, pRange->window.skey, pRange->window.ekey, pRange->range.minVer, + pRange->range.maxVer); } } + +void launchFillHistoryTask(SStreamTask* pTask) { + int32_t tId = pTask->historyTaskId.taskId; + if (tId == 0) { + return; + } + + ASSERT(pTask->status.downstreamReady == 1); + qDebug("s-task:%s start to launch related fill-history task:0x%x", pTask->id.idStr, tId); + + // launch associated fill history task + streamLaunchFillHistoryTask(pTask); +} + +void streamTaskCheckDownstreamTasks(SStreamTask* pTask) { + if (pTask->info.fillHistory) { + qDebug("s-task:%s fill history task, wait for being launched", pTask->id.idStr); + return; + } + + ASSERT(pTask->status.downstreamReady == 0); + + // check downstream tasks for itself + streamTaskDoCheckDownstreamTasks(pTask); +} + +// normal -> pause, pause/stop/dropping -> pause, halt -> pause, scan-history -> pause +void streamTaskPause(SStreamTask* pTask) { + SStreamMeta* pMeta = pTask->pMeta; + + int64_t st = taosGetTimestampMs(); + + int8_t status = pTask->status.taskStatus; + if (status == TASK_STATUS__DROPPING) { + qDebug("vgId:%d s-task:%s task already dropped, do nothing", pMeta->vgId, pTask->id.idStr); + return; + } + + const char* str = streamGetTaskStatusStr(status); + if (status == TASK_STATUS__STOP || status == TASK_STATUS__PAUSE) { + qDebug("vgId:%d s-task:%s task already stopped/paused, status:%s, do nothing", pMeta->vgId, pTask->id.idStr, str); + return; + } + + while(!pTask->status.pauseAllowed || (pTask->status.taskStatus == TASK_STATUS__HALT)) { + status = pTask->status.taskStatus; + if (status == TASK_STATUS__DROPPING) { + qDebug("vgId:%d s-task:%s task already dropped, do nothing", pMeta->vgId, pTask->id.idStr); + return; + } + + if (status == TASK_STATUS__STOP || status == TASK_STATUS__PAUSE) { + qDebug("vgId:%d s-task:%s task already stopped/paused, status:%s, do nothing", pMeta->vgId, pTask->id.idStr, str); + return; + } + + qDebug("s-task:%s wait for the task can be paused, vgId:%d", pTask->id.idStr, pMeta->vgId); + taosMsleep(100); + } + + atomic_store_8(&pTask->status.keepTaskStatus, pTask->status.taskStatus); + atomic_store_8(&pTask->status.taskStatus, TASK_STATUS__PAUSE); + + int64_t el = taosGetTimestampMs() - st; + qDebug("vgId:%d s-task:%s set pause flag, prev:%s, elapsed time:%dms", pMeta->vgId, pTask->id.idStr, + streamGetTaskStatusStr(pTask->status.keepTaskStatus), (int32_t)el); +} + +void streamTaskResume(SStreamTask* pTask) { + int8_t status = pTask->status.taskStatus; + if (status == TASK_STATUS__PAUSE) { + pTask->status.taskStatus = pTask->status.keepTaskStatus; + pTask->status.keepTaskStatus = TASK_STATUS__NORMAL; + qDebug("s-task:%s resume from pause", pTask->id.idStr); + } else { + qError("s-task:%s not in pause, failed to resume, status:%s", pTask->id.idStr, streamGetTaskStatusStr(status)); + } +} + +// todo fix race condition +void streamTaskDisablePause(SStreamTask* pTask) { + // pre-condition check + const char* id = pTask->id.idStr; + while (pTask->status.taskStatus == TASK_STATUS__PAUSE) { + qDebug("s-task:%s already in pause, wait for pause being cancelled, and set pause disabled, recheck in 100ms", id); + taosMsleep(100); + } + + qDebug("s-task:%s disable task pause", id); + pTask->status.pauseAllowed = 0; +} + +void streamTaskEnablePause(SStreamTask* pTask) { + qDebug("s-task:%s enable task pause", pTask->id.idStr); + pTask->status.pauseAllowed = 1; +} + +void streamTaskHalt(SStreamTask* pTask) { + int8_t status = pTask->status.taskStatus; + if (status == TASK_STATUS__DROPPING || status == TASK_STATUS__STOP) { + return; + } + + if (status == TASK_STATUS__HALT) { + return; + } + + // upgrade to halt status + if (status == TASK_STATUS__PAUSE) { + qDebug("s-task:%s upgrade status to %s from %s", pTask->id.idStr, streamGetTaskStatusStr(TASK_STATUS__HALT), + streamGetTaskStatusStr(TASK_STATUS__PAUSE)); + } else { + qDebug("s-task:%s halt task", pTask->id.idStr); + } + + pTask->status.keepTaskStatus = status; + pTask->status.taskStatus = TASK_STATUS__HALT; +} + +void streamTaskResumeFromHalt(SStreamTask* pTask) { + const char* id = pTask->id.idStr; + int8_t status = pTask->status.taskStatus; + if (status != TASK_STATUS__HALT) { + qError("s-task:%s not in halt status, status:%s", id, streamGetTaskStatusStr(status)); + return; + } + + pTask->status.taskStatus = pTask->status.keepTaskStatus; + pTask->status.keepTaskStatus = TASK_STATUS__NORMAL; + qDebug("s-task:%s resume from halt, current status:%s", id, streamGetTaskStatusStr(pTask->status.taskStatus)); +} diff --git a/source/libs/stream/src/streamTask.c b/source/libs/stream/src/streamTask.c index 8689072f81..31149a3f22 100644 --- a/source/libs/stream/src/streamTask.c +++ b/source/libs/stream/src/streamTask.c @@ -13,6 +13,8 @@ * along with this program. If not, see . */ +#include +#include #include "executor.h" #include "tstream.h" #include "wal.h" @@ -45,7 +47,7 @@ SStreamTask* tNewStreamTask(int64_t streamId, int8_t taskLevel, int8_t fillHisto pTask->status.schedStatus = TASK_SCHED_STATUS__INACTIVE; pTask->status.taskStatus = TASK_STATUS__SCAN_HISTORY; pTask->inputStatus = TASK_INPUT_STATUS__NORMAL; - pTask->outputStatus = TASK_OUTPUT_STATUS__NORMAL; + pTask->outputInfo.status = TASK_OUTPUT_STATUS__NORMAL; addToTaskset(pTaskList, pTask); return pTask; @@ -75,7 +77,7 @@ int32_t tEncodeStreamTask(SEncoder* pEncoder, const SStreamTask* pTask) { if (tEncodeI32(pEncoder, pTask->id.taskId) < 0) return -1; if (tEncodeI32(pEncoder, pTask->info.totalLevel) < 0) return -1; if (tEncodeI8(pEncoder, pTask->info.taskLevel) < 0) return -1; - if (tEncodeI8(pEncoder, pTask->outputType) < 0) return -1; + if (tEncodeI8(pEncoder, pTask->outputInfo.type) < 0) return -1; if (tEncodeI16(pEncoder, pTask->msgInfo.msgType) < 0) return -1; if (tEncodeI8(pEncoder, pTask->status.taskStatus) < 0) return -1; @@ -110,19 +112,19 @@ int32_t tEncodeStreamTask(SEncoder* pEncoder, const SStreamTask* pTask) { if (tEncodeCStr(pEncoder, pTask->exec.qmsg) < 0) return -1; } - if (pTask->outputType == TASK_OUTPUT__TABLE) { + if (pTask->outputInfo.type == TASK_OUTPUT__TABLE) { if (tEncodeI64(pEncoder, pTask->tbSink.stbUid) < 0) return -1; if (tEncodeCStr(pEncoder, pTask->tbSink.stbFullName) < 0) return -1; if (tEncodeSSchemaWrapper(pEncoder, pTask->tbSink.pSchemaWrapper) < 0) return -1; - } else if (pTask->outputType == TASK_OUTPUT__SMA) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SMA) { if (tEncodeI64(pEncoder, pTask->smaSink.smaId) < 0) return -1; - } else if (pTask->outputType == TASK_OUTPUT__FETCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__FETCH) { if (tEncodeI8(pEncoder, pTask->fetchSink.reserved) < 0) return -1; - } else if (pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH) { if (tEncodeI32(pEncoder, pTask->fixedEpDispatcher.taskId) < 0) return -1; if (tEncodeI32(pEncoder, pTask->fixedEpDispatcher.nodeId) < 0) return -1; if (tEncodeSEpSet(pEncoder, &pTask->fixedEpDispatcher.epSet) < 0) return -1; - } else if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { if (tSerializeSUseDbRspImp(pEncoder, &pTask->shuffleDispatcher.dbInfo) < 0) return -1; if (tEncodeCStr(pEncoder, pTask->shuffleDispatcher.stbFullName) < 0) return -1; } @@ -167,7 +169,7 @@ int32_t tDecodeStreamTask(SDecoder* pDecoder, SStreamTask* pTask) { if (tDecodeI32(pDecoder, &pTask->id.taskId) < 0) return -1; if (tDecodeI32(pDecoder, &pTask->info.totalLevel) < 0) return -1; if (tDecodeI8(pDecoder, &pTask->info.taskLevel) < 0) return -1; - if (tDecodeI8(pDecoder, &pTask->outputType) < 0) return -1; + if (tDecodeI8(pDecoder, &pTask->outputInfo.type) < 0) return -1; if (tDecodeI16(pDecoder, &pTask->msgInfo.msgType) < 0) return -1; if (tDecodeI8(pDecoder, &pTask->status.taskStatus) < 0) return -1; @@ -209,21 +211,21 @@ int32_t tDecodeStreamTask(SDecoder* pDecoder, SStreamTask* pTask) { if (tDecodeCStrAlloc(pDecoder, &pTask->exec.qmsg) < 0) return -1; } - if (pTask->outputType == TASK_OUTPUT__TABLE) { + if (pTask->outputInfo.type == TASK_OUTPUT__TABLE) { if (tDecodeI64(pDecoder, &pTask->tbSink.stbUid) < 0) return -1; if (tDecodeCStrTo(pDecoder, pTask->tbSink.stbFullName) < 0) return -1; pTask->tbSink.pSchemaWrapper = taosMemoryCalloc(1, sizeof(SSchemaWrapper)); if (pTask->tbSink.pSchemaWrapper == NULL) return -1; if (tDecodeSSchemaWrapper(pDecoder, pTask->tbSink.pSchemaWrapper) < 0) return -1; - } else if (pTask->outputType == TASK_OUTPUT__SMA) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SMA) { if (tDecodeI64(pDecoder, &pTask->smaSink.smaId) < 0) return -1; - } else if (pTask->outputType == TASK_OUTPUT__FETCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__FETCH) { if (tDecodeI8(pDecoder, &pTask->fetchSink.reserved) < 0) return -1; - } else if (pTask->outputType == TASK_OUTPUT__FIXED_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__FIXED_DISPATCH) { if (tDecodeI32(pDecoder, &pTask->fixedEpDispatcher.taskId) < 0) return -1; if (tDecodeI32(pDecoder, &pTask->fixedEpDispatcher.nodeId) < 0) return -1; if (tDecodeSEpSet(pDecoder, &pTask->fixedEpDispatcher.epSet) < 0) return -1; - } else if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { if (tDeserializeSUseDbRspImp(pDecoder, &pTask->shuffleDispatcher.dbInfo) < 0) return -1; if (tDecodeCStrTo(pDecoder, pTask->shuffleDispatcher.stbFullName) < 0) return -1; } @@ -233,6 +235,11 @@ int32_t tDecodeStreamTask(SDecoder* pDecoder, SStreamTask* pTask) { return 0; } +static void freeItem(void* p) { + SStreamContinueExecInfo* pInfo = p; + rpcFreeCont(pInfo->msg.pCont); +} + void tFreeStreamTask(SStreamTask* pTask) { qDebug("free s-task:%s", pTask->id.idStr); @@ -241,8 +248,8 @@ void tFreeStreamTask(SStreamTask* pTask) { streamQueueClose(pTask->inputQueue); } - if (pTask->outputQueue) { - streamQueueClose(pTask->outputQueue); + if (pTask->outputInfo.queue) { + streamQueueClose(pTask->outputInfo.queue); } if (pTask->exec.qmsg) { @@ -258,12 +265,12 @@ void tFreeStreamTask(SStreamTask* pTask) { walCloseReader(pTask->exec.pWalReader); } - taosArrayDestroyP(pTask->pUpstreamInfoList, taosMemoryFree); - if (pTask->outputType == TASK_OUTPUT__TABLE) { + taosArrayDestroyP(pTask->pUpstreamEpInfoList, taosMemoryFree); + if (pTask->outputInfo.type == TASK_OUTPUT__TABLE) { tDeleteSchemaWrapper(pTask->tbSink.pSchemaWrapper); taosMemoryFree(pTask->tbSink.pTSchema); tSimpleHashCleanup(pTask->tbSink.pTblInfo); - } else if (pTask->outputType == TASK_OUTPUT__SHUFFLE_DISPATCH) { + } else if (pTask->outputInfo.type == TASK_OUTPUT__SHUFFLE_DISPATCH) { taosArrayDestroy(pTask->shuffleDispatcher.dbInfo.pVgroupInfos); pTask->checkReqIds = taosArrayDestroy(pTask->checkReqIds); } @@ -283,6 +290,12 @@ void tFreeStreamTask(SStreamTask* pTask) { tSimpleHashCleanup(pTask->pNameMap); } + if (pTask->pRspMsgList != NULL) { + taosArrayDestroyEx(pTask->pRspMsgList, freeItem); + pTask->pRspMsgList = NULL; + } + + taosThreadMutexDestroy(&pTask->lock); taosMemoryFree(pTask); } @@ -290,7 +303,7 @@ int32_t streamTaskGetNumOfDownstream(const SStreamTask* pTask) { if (pTask->info.taskLevel == TASK_LEVEL__SINK) { return 0; } else { - int32_t type = pTask->outputType; + int32_t type = pTask->outputInfo.type; if (type == TASK_OUTPUT__FIXED_DISPATCH || type == TASK_OUTPUT__TABLE) { return 1; } else { diff --git a/source/util/src/tarray.c b/source/util/src/tarray.c index 8906391a9a..f5e15e7436 100644 --- a/source/util/src/tarray.c +++ b/source/util/src/tarray.c @@ -191,7 +191,7 @@ void* taosArrayGet(const SArray* pArray, size_t index) { } if (index >= pArray->size) { - uError("index is out of range, current:%"PRIzu" max:%d", index, pArray->capacity); + uError("index is out of range, current:%" PRIzu " max:%d", index, pArray->capacity); return NULL; } @@ -221,7 +221,7 @@ size_t taosArrayGetSize(const SArray* pArray) { return TARRAY_SIZE(pArray); } -void* taosArrayInsert(SArray* pArray, size_t index, void* pData) { +void* taosArrayInsert(SArray* pArray, size_t index, const void* pData) { if (pArray == NULL || pData == NULL) { return NULL; } @@ -492,7 +492,7 @@ void* taosDecodeArray(const void* buf, SArray** pArray, FDecode decode, int32_t // order array void taosArraySortPWithExt(SArray* pArray, __ext_compar_fn_t fn, const void* param) { taosqsort(pArray->pData, pArray->size, pArray->elemSize, param, fn); -// taosArrayGetSize(pArray) > 8 ? taosArrayQuickSort(pArray, fn, param) : taosArrayInsertSort(pArray, fn, param); + // taosArrayGetSize(pArray) > 8 ? taosArrayQuickSort(pArray, fn, param) : taosArrayInsertSort(pArray, fn, param); } void taosArraySwap(SArray* a, SArray* b) { diff --git a/source/util/src/tconfig.c b/source/util/src/tconfig.c index 288ea6052b..3681c7a423 100644 --- a/source/util/src/tconfig.c +++ b/source/util/src/tconfig.c @@ -380,43 +380,43 @@ static int32_t cfgAddItem(SConfig *pCfg, SConfigItem *pItem, const char *name) { return 0; } -int32_t cfgAddBool(SConfig *pCfg, const char *name, bool defaultVal, bool tsc) { - SConfigItem item = {.dtype = CFG_DTYPE_BOOL, .bval = defaultVal, .tsc = tsc}; +int32_t cfgAddBool(SConfig *pCfg, const char *name, bool defaultVal, int8_t scope) { + SConfigItem item = {.dtype = CFG_DTYPE_BOOL, .bval = defaultVal, .scope = scope}; return cfgAddItem(pCfg, &item, name); } -int32_t cfgAddInt32(SConfig *pCfg, const char *name, int32_t defaultVal, int64_t minval, int64_t maxval, bool tsc) { +int32_t cfgAddInt32(SConfig *pCfg, const char *name, int32_t defaultVal, int64_t minval, int64_t maxval, int8_t scope) { if (defaultVal < minval || defaultVal > maxval) { terrno = TSDB_CODE_OUT_OF_RANGE; return -1; } - SConfigItem item = {.dtype = CFG_DTYPE_INT32, .i32 = defaultVal, .imin = minval, .imax = maxval, .tsc = tsc}; + SConfigItem item = {.dtype = CFG_DTYPE_INT32, .i32 = defaultVal, .imin = minval, .imax = maxval, .scope = scope}; return cfgAddItem(pCfg, &item, name); } -int32_t cfgAddInt64(SConfig *pCfg, const char *name, int64_t defaultVal, int64_t minval, int64_t maxval, bool tsc) { +int32_t cfgAddInt64(SConfig *pCfg, const char *name, int64_t defaultVal, int64_t minval, int64_t maxval, int8_t scope) { if (defaultVal < minval || defaultVal > maxval) { terrno = TSDB_CODE_OUT_OF_RANGE; return -1; } - SConfigItem item = {.dtype = CFG_DTYPE_INT64, .i64 = defaultVal, .imin = minval, .imax = maxval, .tsc = tsc}; + SConfigItem item = {.dtype = CFG_DTYPE_INT64, .i64 = defaultVal, .imin = minval, .imax = maxval, .scope = scope}; return cfgAddItem(pCfg, &item, name); } -int32_t cfgAddFloat(SConfig *pCfg, const char *name, float defaultVal, double minval, double maxval, bool tsc) { +int32_t cfgAddFloat(SConfig *pCfg, const char *name, float defaultVal, double minval, double maxval, int8_t scope) { if (defaultVal < minval || defaultVal > maxval) { terrno = TSDB_CODE_OUT_OF_RANGE; return -1; } - SConfigItem item = {.dtype = CFG_DTYPE_FLOAT, .fval = defaultVal, .fmin = minval, .fmax = maxval, .tsc = tsc}; + SConfigItem item = {.dtype = CFG_DTYPE_FLOAT, .fval = defaultVal, .fmin = minval, .fmax = maxval, .scope = scope}; return cfgAddItem(pCfg, &item, name); } -int32_t cfgAddString(SConfig *pCfg, const char *name, const char *defaultVal, bool tsc) { - SConfigItem item = {.dtype = CFG_DTYPE_STRING, .tsc = tsc}; +int32_t cfgAddString(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope) { + SConfigItem item = {.dtype = CFG_DTYPE_STRING, .scope = scope}; item.str = taosStrdup(defaultVal); if (item.str == NULL) { terrno = TSDB_CODE_OUT_OF_MEMORY; @@ -425,8 +425,8 @@ int32_t cfgAddString(SConfig *pCfg, const char *name, const char *defaultVal, bo return cfgAddItem(pCfg, &item, name); } -int32_t cfgAddDir(SConfig *pCfg, const char *name, const char *defaultVal, bool tsc) { - SConfigItem item = {.dtype = CFG_DTYPE_DIR, .tsc = tsc}; +int32_t cfgAddDir(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope) { + SConfigItem item = {.dtype = CFG_DTYPE_DIR, .scope = scope}; if (cfgCheckAndSetDir(&item, defaultVal) != 0) { return -1; } @@ -434,8 +434,8 @@ int32_t cfgAddDir(SConfig *pCfg, const char *name, const char *defaultVal, bool return cfgAddItem(pCfg, &item, name); } -int32_t cfgAddLocale(SConfig *pCfg, const char *name, const char *defaultVal) { - SConfigItem item = {.dtype = CFG_DTYPE_LOCALE, .tsc = 1}; +int32_t cfgAddLocale(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope) { + SConfigItem item = {.dtype = CFG_DTYPE_LOCALE, .scope = scope}; if (cfgCheckAndSetLocale(&item, defaultVal) != 0) { return -1; } @@ -443,8 +443,8 @@ int32_t cfgAddLocale(SConfig *pCfg, const char *name, const char *defaultVal) { return cfgAddItem(pCfg, &item, name); } -int32_t cfgAddCharset(SConfig *pCfg, const char *name, const char *defaultVal) { - SConfigItem item = {.dtype = CFG_DTYPE_CHARSET, .tsc = 1}; +int32_t cfgAddCharset(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope) { + SConfigItem item = {.dtype = CFG_DTYPE_CHARSET, .scope = scope}; if (cfgCheckAndSetCharset(&item, defaultVal) != 0) { return -1; } @@ -452,8 +452,8 @@ int32_t cfgAddCharset(SConfig *pCfg, const char *name, const char *defaultVal) { return cfgAddItem(pCfg, &item, name); } -int32_t cfgAddTimezone(SConfig *pCfg, const char *name, const char *defaultVal) { - SConfigItem item = {.dtype = CFG_DTYPE_TIMEZONE, .tsc = 1}; +int32_t cfgAddTimezone(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope) { + SConfigItem item = {.dtype = CFG_DTYPE_TIMEZONE, .scope = scope}; if (cfgCheckAndSetTimezone(&item, defaultVal) != 0) { return -1; } @@ -543,6 +543,27 @@ void cfgDumpItemValue(SConfigItem *pItem, char *buf, int32_t bufSize, int32_t *p *pLen = len; } +void cfgDumpItemScope(SConfigItem *pItem, char *buf, int32_t bufSize, int32_t *pLen) { + int32_t len = 0; + switch (pItem->scope) { + case CFG_SCOPE_SERVER: + len = snprintf(buf, bufSize, "server"); + break; + case CFG_SCOPE_CLIENT: + len = snprintf(buf, bufSize, "client"); + break; + case CFG_SCOPE_BOTH: + len = snprintf(buf, bufSize, "both"); + break; + } + + if (len > bufSize) { + len = bufSize; + } + + *pLen = len; +} + void cfgDumpCfg(SConfig *pCfg, bool tsc, bool dump) { if (dump) { printf(" global config"); @@ -560,7 +581,7 @@ void cfgDumpCfg(SConfig *pCfg, bool tsc, bool dump) { int32_t size = taosArrayGetSize(pCfg->array); for (int32_t i = 0; i < size; ++i) { SConfigItem *pItem = taosArrayGet(pCfg->array, i); - if (tsc && !pItem->tsc) continue; + if (tsc && pItem->scope == CFG_SCOPE_SERVER) continue; if (dump && strcmp(pItem->name, "scriptDir") == 0) continue; if (dump && strcmp(pItem->name, "simDebugFlag") == 0) continue; tstrncpy(src, cfgStypeStr(pItem->stype), CFG_SRC_PRINT_LEN); diff --git a/source/util/src/trbtree.c b/source/util/src/trbtree.c index e7386d5912..e1000f7bc1 100644 --- a/source/util/src/trbtree.c +++ b/source/util/src/trbtree.c @@ -105,7 +105,7 @@ static void tRBTreeTransplant(SRBTree *pTree, SRBTreeNode *u, SRBTreeNode *v) { v->parent = u->parent; } -static SRBTreeNode *tRBTreeSuccessor(SRBTree *pTree, SRBTreeNode *pNode) { +static SRBTreeNode *tRBTreeSuccessor(const SRBTree *pTree, SRBTreeNode *pNode) { if (pNode->right != pTree->NIL) { pNode = pNode->right; while (pNode->left != pTree->NIL) { @@ -125,7 +125,7 @@ static SRBTreeNode *tRBTreeSuccessor(SRBTree *pTree, SRBTreeNode *pNode) { return pNode; } -static SRBTreeNode *tRBTreePredecessor(SRBTree *pTree, SRBTreeNode *pNode) { +static SRBTreeNode *tRBTreePredecessor(const SRBTree *pTree, SRBTreeNode *pNode) { if (pNode->left != pTree->NIL) { pNode = pNode->left; while (pNode->right != pTree->NIL) { @@ -443,7 +443,7 @@ SRBTreeNode *tRBTreeDropMax(SRBTree *pTree) { return pNode; } -SRBTreeNode *tRBTreeGet(SRBTree *pTree, const SRBTreeNode *pKeyNode) { +SRBTreeNode *tRBTreeGet(const SRBTree *pTree, const SRBTreeNode *pKeyNode) { SRBTreeNode *pNode = pTree->root; while (pNode != pTree->NIL) { diff --git a/tests/develop-test/2-query/show_create_db.py b/tests/develop-test/2-query/show_create_db.py index d4bff819c9..07ca357f9e 100644 --- a/tests/develop-test/2-query/show_create_db.py +++ b/tests/develop-test/2-query/show_create_db.py @@ -42,7 +42,7 @@ class TDTestCase: tdSql.query('show create database scd;') tdSql.checkRows(1) tdSql.checkData(0, 0, 'scd') - tdSql.checkData(0, 1, "CREATE DATABASE `scd` BUFFER 256 CACHESIZE 1 CACHEMODEL 'none' COMP 2 DURATION 14400m WAL_FSYNC_PERIOD 3000 MAXROWS 4096 MINROWS 100 STT_TRIGGER 1 KEEP 5256000m,5256000m,5256000m PAGES 256 PAGESIZE 4 PRECISION 'ms' REPLICA 1 WAL_LEVEL 1 VGROUPS 2 SINGLE_STABLE 0 TABLE_PREFIX 0 TABLE_SUFFIX 0 TSDB_PAGESIZE 4 WAL_RETENTION_PERIOD 0 WAL_RETENTION_SIZE 0") + tdSql.checkData(0, 1, "CREATE DATABASE `scd` BUFFER 256 CACHESIZE 1 CACHEMODEL 'none' COMP 2 DURATION 14400m WAL_FSYNC_PERIOD 3000 MAXROWS 4096 MINROWS 100 STT_TRIGGER 2 KEEP 5256000m,5256000m,5256000m PAGES 256 PAGESIZE 4 PRECISION 'ms' REPLICA 1 WAL_LEVEL 1 VGROUPS 2 SINGLE_STABLE 0 TABLE_PREFIX 0 TABLE_SUFFIX 0 TSDB_PAGESIZE 4 WAL_RETENTION_PERIOD 0 WAL_RETENTION_SIZE 0") tdSql.query('show create database scd2;') tdSql.checkRows(1) @@ -60,7 +60,7 @@ class TDTestCase: tdSql.query('show create database scd;') tdSql.checkRows(1) tdSql.checkData(0, 0, 'scd') - tdSql.checkData(0, 1, "CREATE DATABASE `scd` BUFFER 256 CACHESIZE 1 CACHEMODEL 'none' COMP 2 DURATION 14400m WAL_FSYNC_PERIOD 3000 MAXROWS 4096 MINROWS 100 STT_TRIGGER 1 KEEP 5256000m,5256000m,5256000m PAGES 256 PAGESIZE 4 PRECISION 'ms' REPLICA 1 WAL_LEVEL 1 VGROUPS 2 SINGLE_STABLE 0 TABLE_PREFIX 0 TABLE_SUFFIX 0 TSDB_PAGESIZE 4 WAL_RETENTION_PERIOD 0 WAL_RETENTION_SIZE 0") + tdSql.checkData(0, 1, "CREATE DATABASE `scd` BUFFER 256 CACHESIZE 1 CACHEMODEL 'none' COMP 2 DURATION 14400m WAL_FSYNC_PERIOD 3000 MAXROWS 4096 MINROWS 100 STT_TRIGGER 2 KEEP 5256000m,5256000m,5256000m PAGES 256 PAGESIZE 4 PRECISION 'ms' REPLICA 1 WAL_LEVEL 1 VGROUPS 2 SINGLE_STABLE 0 TABLE_PREFIX 0 TABLE_SUFFIX 0 TSDB_PAGESIZE 4 WAL_RETENTION_PERIOD 0 WAL_RETENTION_SIZE 0") tdSql.query('show create database scd2;') tdSql.checkRows(1) diff --git a/tests/parallel_test/cases.task b/tests/parallel_test/cases.task index 08c23da5f0..3502475337 100644 --- a/tests/parallel_test/cases.task +++ b/tests/parallel_test/cases.task @@ -160,6 +160,7 @@ ,,n,system-test,python3 ./test.py -f 0-others/udfpy_main.py ,,n,system-test,python3 ./test.py -N 3 -f 0-others/walRetention.py ,,n,system-test,python3 ./test.py -f 0-others/splitVGroup.py -N 5 +,,n,system-test,python3 ./test.py -f 0-others/timeRangeWise.py -N 3 ,,y,system-test,./pytest.sh python3 ./test.py -f 1-insert/alter_database.py ,,y,system-test,./pytest.sh python3 ./test.py -f 1-insert/alter_replica.py -N 3 ,,y,system-test,./pytest.sh python3 ./test.py -f 1-insert/influxdb_line_taosc_insert.py diff --git a/tests/pytest/util/sql.py b/tests/pytest/util/sql.py index 2af8f721b6..2fa21b1983 100644 --- a/tests/pytest/util/sql.py +++ b/tests/pytest/util/sql.py @@ -440,8 +440,10 @@ class TDSql: time.sleep(1) continue - def execute(self, sql,queryTimes=30): + def execute(self, sql, queryTimes=30, show=False): self.sql = sql + if show: + tdLog.info(sql) i=1 while i <= queryTimes: try: diff --git a/tests/script/tsim/parser/fill.sim b/tests/script/tsim/parser/fill.sim index a66e7d6ab7..0534aa5d5b 100644 --- a/tests/script/tsim/parser/fill.sim +++ b/tests/script/tsim/parser/fill.sim @@ -1224,4 +1224,104 @@ if $data42 != NULL then return -1 endi +print ===================== TD-3625 test fill value NULL +sql use $db + +sql select _wstart,_wend,count(*) from tm0 where ts >= '2020-01-01 01:03:06.000' and ts <= '2020-01-01 01:03:10.000' interval(1s) fill(value, NULL); + +if $rows != 5 then + return -1 +endi + +if $data02 != NULL then + return -1 +endi + +if $data12 != 1 then + return -1 +endi + +if $data22 != 1 then + return -1 +endi + +if $data32 != 1 then + return -1 +endi + +if $data42 != NULL then + return -1 +endi + +sql select _wstart,_wend,count(*),sum(k),avg(k) from tm0 where ts >= '2020-01-01 01:03:06.000' and ts <= '2020-01-01 01:03:10.000' interval(1s) fill(value, 1, NULL, 1); + +if $rows != 5 then + return -1 +endi + +if $data02 != 1 then + return -1 +endi + +if $data12 != 1 then + return -1 +endi + +if $data22 != 1 then + return -1 +endi + +if $data32 != 1 then + return -1 +endi + +if $data42 != 1 then + return -1 +endi + + +if $data03 != NULL then + return -1 +endi + +if $data13 != 7 then + return -1 +endi + +if $data23 != 8 then + return -1 +endi + +if $data33 != 9 then + return -1 +endi + +if $data43 != NULL then + return -1 +endi + + +if $data04 != 1.000000000 then + return -1 +endi + +if $data14 != 7.000000000 then + return -1 +endi + +if $data24 != 8.000000000 then + return -1 +endi + +if $data34 != 9.000000000 then + return -1 +endi + +if $data44 != 1.000000000 then + return -1 +endi + + system sh/exec.sh -n dnode1 -s stop -x SIGINT + + diff --git a/tests/system-test/0-others/show.py b/tests/system-test/0-others/show.py index 50a1662ba0..9d26b3a2ae 100644 --- a/tests/system-test/0-others/show.py +++ b/tests/system-test/0-others/show.py @@ -95,6 +95,23 @@ class TDTestCase: tdSql.checkEqual(f'{db}',tdSql.queryResult[0][0]) tdSql.checkEqual(f'CREATE DATABASE `{db}`',tdSql.queryResult[0][1]) + def show_create_systb_sql(self): + for param in self.ins_param_list: + tdSql.query(f'show create table information_schema.ins_{param}') + tdSql.checkEqual(f'ins_{param}',tdSql.queryResult[0][0]) + + tdSql.execute(f'use information_schema') + tdSql.query(f'show create table ins_{param}') + tdSql.checkEqual(f'ins_{param}',tdSql.queryResult[0][0]) + + for param in self.perf_param_list: + tdSql.query(f'show create table performance_schema.perf_{param}') + tdSql.checkEqual(f'perf_{param}',tdSql.queryResult[0][0]) + + tdSql.execute(f'use performance_schema') + tdSql.query(f'show create table perf_{param}') + tdSql.checkEqual(f'perf_{param}',tdSql.queryResult[0][0]) + def show_create_sql(self): create_db_sql = self.set_create_database_sql(self.db_param) print(create_db_sql) @@ -200,6 +217,7 @@ class TDTestCase: self.perf_check() self.show_create_sql() self.show_create_sysdb_sql() + self.show_create_systb_sql() def stop(self): tdSql.close() diff --git a/tests/system-test/0-others/splitVGroup.py b/tests/system-test/0-others/splitVGroup.py index 4509961066..9fd00892e4 100644 --- a/tests/system-test/0-others/splitVGroup.py +++ b/tests/system-test/0-others/splitVGroup.py @@ -283,14 +283,14 @@ class TDTestCase: # normal table - # all rows - sql = "select * from @db_name.ta" - self.queryDouble(sql) - # count sql = "select count(*) from @db_name.ta" self.queryDouble(sql) + # all rows + sql = "select * from @db_name.ta" + self.queryDouble(sql) + # sum sql = "select sum(c1) from @db_name.ta" self.queryDouble(sql) @@ -316,7 +316,8 @@ class TDTestCase: tdSql.execute(sql) # wait end - for i in range(100): + seconds = 300 + for i in range(seconds): sql ="show transactions;" rows = tdSql.query(sql) if rows == 0: @@ -325,7 +326,7 @@ class TDTestCase: #tdLog.info(f"i={i} wait split vgroup ...") time.sleep(1) - tdLog.exit("split vgroup transaction is not finished after executing 50s") + tdLog.exit(f"split vgroup transaction is not finished after executing {seconds}s") return False # split error @@ -382,6 +383,14 @@ class TDTestCase: self.expectSplitError("topicdb") tdSql.execute("drop topic toa;") self.expectSplitOk("topicdb") + + # compact and check db2 + def compactAndCheck(self): + tdLog.info("compact db2 and check result ...") + # compact + tdSql.execute(f"compact database {self.db2};") + # check result + self.checkResult() # run def run(self): @@ -390,12 +399,15 @@ class TDTestCase: for i in range(5): # split vgroup on db2 + start = time.time() self.splitVGroup(self.db2) + end = time.time() self.vgroups2 += 1 - + # check two db query result same self.checkResult() - tdLog.info(f"split vgroup i={i} passed.") + spend = "%.3f"%(end-start) + tdLog.info(f"split vgroup i={i} passed. spend = {spend}s") # split empty db self.splitEmptyDB() @@ -403,6 +415,9 @@ class TDTestCase: # check topic and stream forib self.checkForbid() + # compact database + self.compactAndCheck() + # stop def stop(self): tdSql.close() diff --git a/tests/system-test/0-others/timeRangeWise.py b/tests/system-test/0-others/timeRangeWise.py new file mode 100644 index 0000000000..a7dc18aa82 --- /dev/null +++ b/tests/system-test/0-others/timeRangeWise.py @@ -0,0 +1,309 @@ +################################################################### +# Copyright (c) 2016 by TAOS Technologies, Inc. +# All rights reserved. +# +# This file is proprietary and confidential to TAOS Technologies. +# No part of this file may be reproduced, stored, transmitted, +# disclosed or used in any form or by any means other than as +# expressly provided by the written permission from Jianhui Tao +# +################################################################### + +# -*- coding: utf-8 -*- + +import sys +import random +import time +import copy +import string + +import taos +from util.log import * +from util.cases import * +from util.sql import * + +class TDTestCase: + + # random string + def random_string(self, count): + letters = string.ascii_letters + return ''.join(random.choice(letters) for i in range(count)) + + # get col value and total max min ... + def getColsValue(self, i, j): + # c1 value + if random.randint(1, 10) == 5: + c1 = None + else: + c1 = 1 + + # c2 value + if j % 3200 == 0: + c2 = 8764231 + elif random.randint(1, 10) == 5: + c2 = None + else: + c2 = random.randint(-87654297, 98765321) + + + value = f"({self.ts}, " + + # c1 + if c1 is None: + value += "null," + else: + self.c1Cnt += 1 + value += f"{c1}," + # c2 + if c2 is None: + value += "null," + else: + value += f"{c2}," + # total count + self.c2Cnt += 1 + # max + if self.c2Max is None: + self.c2Max = c2 + else: + if c2 > self.c2Max: + self.c2Max = c2 + # min + if self.c2Min is None: + self.c2Min = c2 + else: + if c2 < self.c2Min: + self.c2Min = c2 + # sum + if self.c2Sum is None: + self.c2Sum = c2 + else: + self.c2Sum += c2 + + # c3 same with ts + value += f"{self.ts})" + + # move next 1s interval + self.ts += 1 + + return value + + # insert data + def insertData(self): + tdLog.info("insert data ....") + sqls = "" + for i in range(self.childCnt): + # insert child table + values = "" + pre_insert = f"insert into @db_name.t{i} values " + for j in range(self.childRow): + if values == "": + values = self.getColsValue(i, j) + else: + values += "," + self.getColsValue(i, j) + + # batch insert + if j % self.batchSize == 0 and values != "": + sql = pre_insert + values + self.exeDouble(sql) + values = "" + # append last + if values != "": + sql = pre_insert + values + self.exeDouble(sql) + values = "" + + # insert finished + tdLog.info(f"insert data successfully.\n" + f" inserted child table = {self.childCnt}\n" + f" inserted child rows = {self.childRow}\n" + f" total inserted rows = {self.childCnt*self.childRow}\n") + return + + def exeDouble(self, sql): + # dbname replace + sql1 = sql.replace("@db_name", self.db1) + + if len(sql1) > 100: + tdLog.info(sql1[:100]) + else: + tdLog.info(sql1) + tdSql.execute(sql1) + + sql2 = sql.replace("@db_name", self.db2) + if len(sql2) > 100: + tdLog.info(sql2[:100]) + else: + tdLog.info(sql2) + tdSql.execute(sql2) + + + # prepareEnv + def prepareEnv(self): + # init + self.ts = 1680000000000 + self.childCnt = 2 + self.childRow = 100000 + self.batchSize = 5000 + self.vgroups1 = 4 + self.vgroups2 = 4 + self.db1 = "db1" # no sma + self.db2 = "db2" # have sma + self.smaClause = "interval(10s)" + + # total + self.c1Cnt = 0 + self.c2Cnt = 0 + self.c2Max = None + self.c2Min = None + self.c2Sum = None + + # alter local optimization to treu + sql = "alter local 'querysmaoptimize 1'" + tdSql.execute(sql, 5, True) + + # check forbid mulit-replic on create sma index + sql = f"create database db vgroups {self.vgroups1} replica 3" + tdSql.execute(sql, 5, True) + sql = f"create table db.st(ts timestamp, c1 int, c2 bigint, ts1 timestamp) tags(area int)" + tdSql.execute(sql, 5, True) + + sql = f"create sma index sma_test on db.st function(max(c1),max(c2),min(c1),min(c2)) {self.smaClause};" + tdLog.info(sql) + tdSql.error(sql) + + + # create database db + sql = f"create database @db_name vgroups {self.vgroups1} replica 1" + self.exeDouble(sql) + + # create super talbe st + sql = f"create table @db_name.st(ts timestamp, c1 int, c2 bigint, ts1 timestamp) tags(area int)" + self.exeDouble(sql) + + # create child table + for i in range(self.childCnt): + sql = f"create table @db_name.t{i} using @db_name.st tags({i}) " + self.exeDouble(sql) + + # create sma index on db2 + sql = f"use {self.db2}" + tdSql.execute(sql) + sql = f"create sma index sma_index_maxmin on {self.db2}.st function(max(c1),max(c2),min(c1),min(c2)) {self.smaClause};" + tdLog.info(sql) + tdSql.execute(sql) + + # insert data + self.insertData() + + # check data correct + def checkExpect(self, sql, expectVal): + tdSql.query(sql) + rowCnt = tdSql.getRows() + for i in range(rowCnt): + val = tdSql.getData(i,0) + if val != expectVal: + tdLog.exit(f"Not expect . query={val} expect={expectVal} i={i} sql={sql}") + return False + + tdLog.info(f"check expect ok. sql={sql} expect ={expectVal} rowCnt={rowCnt}") + return True + + # init + def init(self, conn, logSql, replicaVar=1): + seed = time.clock_gettime(time.CLOCK_REALTIME) + random.seed(seed) + self.replicaVar = int(replicaVar) + tdLog.debug(f"start to excute {__file__}") + tdSql.init(conn.cursor(), True) + + # check query result same + def queryDoubleImpl(self, sql): + # sql + sql1 = sql.replace('@db_name', self.db1) + tdLog.info(sql1) + start1 = time.time() + rows1 = tdSql.query(sql1) + spend1 = time.time() - start1 + res1 = copy.copy(tdSql.queryResult) + + sql2 = sql.replace('@db_name', self.db2) + tdLog.info(sql2) + start2 = time.time() + tdSql.query(sql2) + spend2 = time.time() - start2 + res2 = tdSql.queryResult + + rowlen1 = len(res1) + rowlen2 = len(res2) + + if rowlen1 != rowlen2: + tdLog.info(f"check error. rowlen1={rowlen1} rowlen2={rowlen2} both not equal.") + return False + + for i in range(rowlen1): + row1 = res1[i] + row2 = res2[i] + collen1 = len(row1) + collen2 = len(row2) + if collen1 != collen2: + tdLog.info(f"checkerror. collen1={collen1} collen2={collen2} both not equal.") + return False + for j in range(collen1): + if row1[j] != row2[j]: + tdLog.exit(f"col={j} col1={row1[j]} col2={row2[j]} both col not equal.") + return False + + # warning performance + multiple = spend1/spend2 + tdLog.info("spend1=%.6fs spend2=%.6fs multiple=%.1f"%(spend1, spend2, multiple)) + if spend2 > spend1 and multiple < 4: + tdLog.info(f"performace not reached: multiple(spend1/spend)={multiple} require is >=4 ") + return False + + return True + + # check query result same + def queryDouble(self, sql, tryCount=60, gap=1): + for i in range(tryCount): + if self.queryDoubleImpl(sql): + return True + # error + tdLog.info(f"queryDouble return false, try loop={i}") + time.sleep(gap) + + tdLog.exit(f"queryDouble try {tryCount} times, but all failed.") + return False + + # check result + def checkResult(self): + + # max + sql = f"select max(c1) from @db_name.st {self.smaClause}" + self.queryDouble(sql) + + # min + sql = f"select max(c2) from @db_name.st {self.smaClause}" + self.queryDouble(sql) + + # mix + sql = f"select max(c1),max(c2),min(c1),min(c2) from @db_name.st {self.smaClause}" + self.queryDouble(sql) + + + # run + def run(self): + # prepare env + self.prepareEnv() + + # check two db query result same + tdLog.info(f"check have sma(db1) and no sma(db2) performace...") + self.checkResult() + + # stop + def stop(self): + tdSql.close() + tdLog.success(f"{__file__} successfully executed") + + +tdCases.addLinux(__file__, TDTestCase()) +tdCases.addWindows(__file__, TDTestCase()) diff --git a/tests/system-test/1-insert/delete_stable.py b/tests/system-test/1-insert/delete_stable.py index 8ebe7b6692..67561c51e5 100644 --- a/tests/system-test/1-insert/delete_stable.py +++ b/tests/system-test/1-insert/delete_stable.py @@ -24,10 +24,11 @@ from util.common import * from util.sqlset import TDSetSql class TDTestCase: + updatecfgDict = {'tsdbdebugFlag': 143} def init(self, conn, logSql, replicaVar=1): self.replicaVar = int(replicaVar) tdLog.debug("start to execute %s" % __file__) - tdSql.init(conn.cursor()) + tdSql.init(conn.cursor(), True) self.dbname = 'db_test' self.ns_dbname = 'ns_test' self.us_dbname = 'us_test' diff --git a/tests/system-test/2-query/columnLenUpdated.py b/tests/system-test/2-query/columnLenUpdated.py index e43b32a716..93d9a492f9 100644 --- a/tests/system-test/2-query/columnLenUpdated.py +++ b/tests/system-test/2-query/columnLenUpdated.py @@ -202,7 +202,7 @@ class TDTestCase: if retCode != "TAOS_OK": tdLog.exit("taos -s fail") - tdSql.query("select count(*) from stb group by tg1") + tdSql.query("select count(*) from stb group by tg1 order by count(*) desc") tdSql.checkData(0, 0, 2) tdSql.checkData(1, 0, 1) diff --git a/tests/system-test/2-query/interp.py b/tests/system-test/2-query/interp.py index b6cefbe36f..986c63839b 100644 --- a/tests/system-test/2-query/interp.py +++ b/tests/system-test/2-query/interp.py @@ -147,6 +147,57 @@ class TDTestCase: tdSql.checkData(11, 0, 15) tdSql.checkData(12, 0, 1) + for col in col_list: + tdSql.query(f"select interp({col}) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, 1.0)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, 1) + tdSql.checkData(1, 0, 5) + tdSql.checkData(2, 0, 1) + tdSql.checkData(3, 0, 1) + tdSql.checkData(4, 0, 1) + tdSql.checkData(5, 0, 1) + tdSql.checkData(6, 0, 10) + tdSql.checkData(7, 0, 1) + tdSql.checkData(8, 0, 1) + tdSql.checkData(9, 0, 1) + tdSql.checkData(10, 0, 1) + tdSql.checkData(11, 0, 15) + tdSql.checkData(12, 0, 1) + + for col in col_list: + tdSql.query(f"select interp({col}) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, true)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, 1) + tdSql.checkData(1, 0, 5) + tdSql.checkData(2, 0, 1) + tdSql.checkData(3, 0, 1) + tdSql.checkData(4, 0, 1) + tdSql.checkData(5, 0, 1) + tdSql.checkData(6, 0, 10) + tdSql.checkData(7, 0, 1) + tdSql.checkData(8, 0, 1) + tdSql.checkData(9, 0, 1) + tdSql.checkData(10, 0, 1) + tdSql.checkData(11, 0, 15) + tdSql.checkData(12, 0, 1) + + for col in col_list: + tdSql.query(f"select interp({col}) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, NULL)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, None) + tdSql.checkData(1, 0, 5) + tdSql.checkData(2, 0, None) + tdSql.checkData(3, 0, None) + tdSql.checkData(4, 0, None) + tdSql.checkData(5, 0, None) + tdSql.checkData(6, 0, 10) + tdSql.checkData(7, 0, None) + tdSql.checkData(8, 0, None) + tdSql.checkData(9, 0, None) + tdSql.checkData(10, 0, None) + tdSql.checkData(11, 0, 15) + tdSql.checkData(12, 0, None) + tdSql.query(f"select interp(c4) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, 1)") tdSql.checkRows(13) tdSql.checkData(0, 0, 1.0) @@ -163,6 +214,54 @@ class TDTestCase: tdSql.checkData(11, 0, 15.0) tdSql.checkData(12, 0, 1.0) + tdSql.query(f"select interp(c4) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, 1.0)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, 1.0) + tdSql.checkData(1, 0, 5.0) + tdSql.checkData(2, 0, 1.0) + tdSql.checkData(3, 0, 1.0) + tdSql.checkData(4, 0, 1.0) + tdSql.checkData(5, 0, 1.0) + tdSql.checkData(6, 0, 10.0) + tdSql.checkData(7, 0, 1.0) + tdSql.checkData(8, 0, 1.0) + tdSql.checkData(9, 0, 1.0) + tdSql.checkData(10, 0, 1.0) + tdSql.checkData(11, 0, 15.0) + tdSql.checkData(12, 0, 1.0) + + tdSql.query(f"select interp(c4) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, true)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, 1.0) + tdSql.checkData(1, 0, 5.0) + tdSql.checkData(2, 0, 1.0) + tdSql.checkData(3, 0, 1.0) + tdSql.checkData(4, 0, 1.0) + tdSql.checkData(5, 0, 1.0) + tdSql.checkData(6, 0, 10.0) + tdSql.checkData(7, 0, 1.0) + tdSql.checkData(8, 0, 1.0) + tdSql.checkData(9, 0, 1.0) + tdSql.checkData(10, 0, 1.0) + tdSql.checkData(11, 0, 15.0) + tdSql.checkData(12, 0, 1.0) + + tdSql.query(f"select interp(c4) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, NULL)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, None) + tdSql.checkData(1, 0, 5.0) + tdSql.checkData(2, 0, None) + tdSql.checkData(3, 0, None) + tdSql.checkData(4, 0, None) + tdSql.checkData(5, 0, None) + tdSql.checkData(6, 0, 10.0) + tdSql.checkData(7, 0, None) + tdSql.checkData(8, 0, None) + tdSql.checkData(9, 0, None) + tdSql.checkData(10, 0, None) + tdSql.checkData(11, 0, 15.0) + tdSql.checkData(12, 0, None) + tdSql.query(f"select interp(c5) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, 1)") tdSql.checkRows(13) tdSql.checkData(0, 0, 1.0) @@ -179,6 +278,54 @@ class TDTestCase: tdSql.checkData(11, 0, 15.0) tdSql.checkData(12, 0, 1.0) + tdSql.query(f"select interp(c5) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, 1.0)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, 1.0) + tdSql.checkData(1, 0, 5.0) + tdSql.checkData(2, 0, 1.0) + tdSql.checkData(3, 0, 1.0) + tdSql.checkData(4, 0, 1.0) + tdSql.checkData(5, 0, 1.0) + tdSql.checkData(6, 0, 10.0) + tdSql.checkData(7, 0, 1.0) + tdSql.checkData(8, 0, 1.0) + tdSql.checkData(9, 0, 1.0) + tdSql.checkData(10, 0, 1.0) + tdSql.checkData(11, 0, 15.0) + tdSql.checkData(12, 0, 1.0) + + tdSql.query(f"select interp(c5) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, true)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, 1.0) + tdSql.checkData(1, 0, 5.0) + tdSql.checkData(2, 0, 1.0) + tdSql.checkData(3, 0, 1.0) + tdSql.checkData(4, 0, 1.0) + tdSql.checkData(5, 0, 1.0) + tdSql.checkData(6, 0, 10.0) + tdSql.checkData(7, 0, 1.0) + tdSql.checkData(8, 0, 1.0) + tdSql.checkData(9, 0, 1.0) + tdSql.checkData(10, 0, 1.0) + tdSql.checkData(11, 0, 15.0) + tdSql.checkData(12, 0, 1.0) + + tdSql.query(f"select interp(c5) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, NULL)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, None) + tdSql.checkData(1, 0, 5.0) + tdSql.checkData(2, 0, None) + tdSql.checkData(3, 0, None) + tdSql.checkData(4, 0, None) + tdSql.checkData(5, 0, None) + tdSql.checkData(6, 0, 10.0) + tdSql.checkData(7, 0, None) + tdSql.checkData(8, 0, None) + tdSql.checkData(9, 0, None) + tdSql.checkData(10, 0, None) + tdSql.checkData(11, 0, 15.0) + tdSql.checkData(12, 0, None) + tdSql.query(f"select interp(c6) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, 1)") tdSql.checkRows(13) tdSql.checkData(0, 0, True) @@ -195,6 +342,54 @@ class TDTestCase: tdSql.checkData(11, 0, True) tdSql.checkData(12, 0, True) + tdSql.query(f"select interp(c6) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, 1.0)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, True) + tdSql.checkData(1, 0, True) + tdSql.checkData(2, 0, True) + tdSql.checkData(3, 0, True) + tdSql.checkData(4, 0, True) + tdSql.checkData(5, 0, True) + tdSql.checkData(6, 0, True) + tdSql.checkData(7, 0, True) + tdSql.checkData(8, 0, True) + tdSql.checkData(9, 0, True) + tdSql.checkData(10, 0, True) + tdSql.checkData(11, 0, True) + tdSql.checkData(12, 0, True) + + tdSql.query(f"select interp(c6) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, true)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, True) + tdSql.checkData(1, 0, True) + tdSql.checkData(2, 0, True) + tdSql.checkData(3, 0, True) + tdSql.checkData(4, 0, True) + tdSql.checkData(5, 0, True) + tdSql.checkData(6, 0, True) + tdSql.checkData(7, 0, True) + tdSql.checkData(8, 0, True) + tdSql.checkData(9, 0, True) + tdSql.checkData(10, 0, True) + tdSql.checkData(11, 0, True) + tdSql.checkData(12, 0, True) + + tdSql.query(f"select interp(c6) from {dbname}.{tbname} range('2020-02-01 00:00:04', '2020-02-01 00:00:16') every(1s) fill(value, NULL)") + tdSql.checkRows(13) + tdSql.checkData(0, 0, None) + tdSql.checkData(1, 0, True) + tdSql.checkData(2, 0, None) + tdSql.checkData(3, 0, None) + tdSql.checkData(4, 0, None) + tdSql.checkData(5, 0, None) + tdSql.checkData(6, 0, True) + tdSql.checkData(7, 0, None) + tdSql.checkData(8, 0, None) + tdSql.checkData(9, 0, None) + tdSql.checkData(10, 0, None) + tdSql.checkData(11, 0, True) + tdSql.checkData(12, 0, None) + ## {} ... tdSql.query(f"select interp(c0) from {dbname}.{tbname} range('2020-02-01 00:00:01', '2020-02-01 00:00:04') every(1s) fill(value, 1)") tdSql.checkRows(4) @@ -2587,25 +2782,25 @@ class TDTestCase: tdSql.checkData(0, 0, '2020-02-02 00:00:00.000') - tdSql.checkData(0, 2, False) + tdSql.checkData(0, 2, None) tdSql.checkData(1, 2, False) - tdSql.checkData(2, 2, False) + tdSql.checkData(2, 2, None) tdSql.checkData(3, 2, True) - tdSql.checkData(4, 2, False) + tdSql.checkData(4, 2, None) tdSql.checkData(5, 2, False) - tdSql.checkData(6, 2, False) + tdSql.checkData(6, 2, None) tdSql.checkData(7, 2, True) - tdSql.checkData(8, 2, False) + tdSql.checkData(8, 2, None) tdSql.checkData(9, 2, True) - tdSql.checkData(10, 2, False) + tdSql.checkData(10, 2, None) tdSql.checkData(11, 2, False) - tdSql.checkData(12, 2, False) + tdSql.checkData(12, 2, None) tdSql.checkData(13, 2, False) - tdSql.checkData(14, 2, False) + tdSql.checkData(14, 2, None) tdSql.checkData(15, 2, None) - tdSql.checkData(16, 2, False) + tdSql.checkData(16, 2, None) tdSql.checkData(17, 2, None) - tdSql.checkData(18, 2, False) + tdSql.checkData(18, 2, None) tdSql.checkData(18, 0, '2020-02-02 00:00:18.000') diff --git a/tests/system-test/2-query/smaBasic.py b/tests/system-test/2-query/smaBasic.py index 43c379ee53..c221a70605 100644 --- a/tests/system-test/2-query/smaBasic.py +++ b/tests/system-test/2-query/smaBasic.py @@ -127,7 +127,7 @@ class TDTestCase: self.c2Sum = None # create database db - sql = f"create database db vgroups 5 replica 3" + sql = f"create database db vgroups 5 replica 3 stt_trigger 1" tdLog.info(sql) tdSql.execute(sql) sql = f"use db" diff --git a/tests/system-test/5-taos-tools/taosbenchmark/insertMix.py b/tests/system-test/5-taos-tools/taosbenchmark/insertMix.py index 60daa8cdc2..b4046b8c98 100644 --- a/tests/system-test/5-taos-tools/taosbenchmark/insertMix.py +++ b/tests/system-test/5-taos-tools/taosbenchmark/insertMix.py @@ -79,6 +79,11 @@ class TDTestCase: tdSql.query("select count(*) from (select * from meters order by ts desc)") tdSql.checkData(0, 0, allCnt) + rowCnt = tdSql.query("select tbname, count(*) from meters partition by tbname slimit 11") + if rowCnt != 10: + tdLog.exit("partition by tbname should return 10 rows of table data which is " + str(rowCnt)) + return + def run(self): binPath = self.getPath()