Merge branch '3.0' of https://github.com/taosdata/TDengine into fix/TS-4706-3.0

docs/en/12-taos-sql/06-select.md

@@ -39,7 +39,7 @@ select_expr: {
 
 from_clause: {
     table_reference [, table_reference] ...
-  | join_clause [, join_clause] ...
+  | table_reference join_clause [, join_clause] ...
 }
 
 table_reference:
@@ -52,7 +52,7 @@ table_expr: {
 }
 
 join_clause:
-    table_reference [INNER] JOIN table_reference ON condition
+    [INNER|LEFT|RIGHT|FULL] [OUTER|SEMI|ANTI|ASOF|WINDOW] JOIN table_reference [ON condition] [WINDOW_OFFSET(start_offset, end_offset)] [JLIMIT jlimit_num]
 
 window_clause: {
     SESSION(ts_col, tol_val)
@@ -408,9 +408,11 @@ SELECT AVG(CASE WHEN voltage < 200 or voltage > 250 THEN 220 ELSE voltage END) F
 
 ## JOIN
 
-TDengine supports the `INTER JOIN` based on the timestamp primary key, that is, the `JOIN` condition must contain the timestamp primary key. As long as the requirement of timestamp-based primary key is met, `INTER JOIN` can be made between normal tables, sub-tables, super tables and sub-queries at will, and there is no limit on the number of tables, primary key and other conditions must be combined with `AND` operator.
+Before the 3.3.0.0 version, TDengine only supported Inner Join queries. Since the 3.3.0.0 version, TDengine supports a wider range of JOIN types, including LEFT JOIN, RIGHT JOIN, FULL JOIN, SEMI JOIN, ANTI-SEMI JOIN in traditional databases, as well as ASOF JOIN and WINDOW JOIN in time series databases. JOIN operations are supported between subtables, normal tables, super tables, and subqueries.
 
-For standard tables:
+### Examples
+
+INNER JOIN between normal tables:
 
 ```sql
 SELECT *
@@ -418,23 +420,23 @@ FROM temp_tb_1 t1, pressure_tb_1 t2
 WHERE t1.ts = t2.ts
 ```
 
-For supertables:
+LEFT JOIN between super tables:
 
 ```sql
 SELECT *
-FROM temp_stable t1, temp_stable t2
+FROM temp_stable t1 LEFT JOIN temp_stable t2
-WHERE t1.ts = t2.ts AND t1.deviceid = t2.deviceid AND t1.status=0;
+ON t1.ts = t2.ts AND t1.deviceid = t2.deviceid AND t1.status=0;
 ```
 
-For sub-table and super table:
+LEFT ASOF JOIN between child table and super table:
 
 ```sql
 SELECT *
-FROM temp_ctable t1, temp_stable t2
+FROM temp_ctable t1 LEFT ASOF JOIN temp_stable t2
-WHERE t1.ts = t2.ts AND t1.deviceid = t2.deviceid AND t1.status=0;
+ON t1.ts = t2.ts AND t1.deviceid = t2.deviceid;
 ```
 
-Similarly, join operations can be performed on the result sets of multiple subqueries.
+For more information about JOIN operations, please refer to the page [TDengine Join] (../join).
 
 ## Nested Query
 

docs/en/12-taos-sql/07-tag-index.md

@@ -34,6 +34,13 @@ SELECT * FROM information_schema.INS_INDEXES
 
 You can also add filter conditions to limit the results.
 
+
+````sql
+SHOW INDEXES FROM tbl_name [FROM db_name];
+SHOW INDEXES FROM [db_name.]tbl_name ;
+````
+Use `show indexes` commands to show indices that have been created for the specified database or table.
+
 ## Detailed Specification
 
 1. Indexes can improve query performance significantly if they are used properly. The operators supported by tag index include  `=`, `>`, `>=`, `<`, `<=`. If you use these operators with tags, indexes can improve query performance significantly. However, for operators not in this scope, indexes don't help. More and more operators will be added in future.

docs/en/12-taos-sql/27-indexing.md

@@ -1,65 +1,134 @@
 ---
-title: Indexing
+sidebar_label: Window Pre-Aggregation
-sidebar_label: Indexing
+title: Window Pre-Aggregation
-description: This document describes the SQL statements related to indexing in TDengine.
+description: Instructions for using Window Pre-Aggregation
 ---
 
-TDengine supports SMA and tag indexing.
+To improve the performance of aggregate function queries on large datasets, you can create Time-Range Small Materialized Aggregates (TSMA) objects. These objects perform pre-computation on specified aggregate functions using fixed time windows and store the computed results. When querying, you can retrieve the pre-computed results to enhance query performance.
 
-## Create an Index
+## Creating TSMA
 
 ```sql
-CREATE INDEX index_name ON tb_name (col_name [, col_name] ...)
+-- Create TSMA based on a super table or regular table
+CREATE TSMA tsma_name ON [dbname.]table_name FUNCTION (func_name(func_param) [, ...] ) INTERVAL(time_duration);
+-- Create a large window TSMA based on a small window TSMA
+CREATE RECURSIVE TSMA tsma_name ON [db_name.]tsma_name1 INTERVAL(time_duration);
 
-CREATE SMA INDEX index_name ON tb_name index_option
+time_duration:
-
+    number unit
-index_option:
-    FUNCTION(functions) INTERVAL(interval_val [, interval_offset]) [SLIDING(sliding_val)] [WATERMARK(watermark_val)] [MAX_DELAY(max_delay_val)]
-
-functions:
-    function [, function] ...
 ```
-### tag Indexing
 
-  [tag index](../tag-index)
+To create a TSMA, you need to specify the TSMA name, table name, function list, and window size. When creating a TSMA based on an existing TSMA, using the `RECURSIVE` keyword, you don't need to specify the `FUNCTION()`. It will create a TSMA with the same function list as the existing TSMA, and the INTERVAL must be a multiple of the window of the base TSMA.
 
-### SMA Indexing
+The naming rule for TSMA is similar to the table name, with a maximum length of the table name length minus the length of the output table suffix. The table name length limit is 193, and the output table suffix is `_tsma_res_stb_`. The maximum length of the TSMA name is 178.
 
-Performs pre-aggregation on the specified column over the time window defined by the INTERVAL clause. The type is specified in functions_string. SMA indexing improves aggregate query performance for the specified time period. One supertable can only contain one SMA index.
+TSMA can only be created based on super tables and regular tables, not on subtables.
 
-- The max, min, and sum functions are supported.
+In the function list, you can only specify supported aggregate functions (see below), and the number of function parameters must be 1, even if the current function supports multiple parameters. The function parameters must be ordinary column names, not tag columns. Duplicate functions and columns in the function list will be deduplicated. When calculating TSMA, all `intermediate results of the functions` will be output to another super table, and the output super table also includes all tag columns of the original table. The maximum number of functions in the function list is the maximum number of columns in the output table (including tag columns) minus the four additional columns added for TSMA calculation, namely `_wstart`, `_wend`, `_wduration`, and a new tag column `tbname`, minus the number of tag columns in the original table. If the number of columns exceeds the limit, an error `Too many columns` will be reported.
-- WATERMARK: Enter a value between 0ms and 900000ms. The most precise unit supported is milliseconds. The default value is 5 seconds. This option can be used only on supertables.
+
-- MAX_DELAY: Enter a value between 1ms and 900000ms. The most precise unit supported is milliseconds. The default value is the value of interval provided that it does not exceed 900000ms. This option can be used only on supertables. Note: Retain the default value if possible. Configuring a small MAX_DELAY may cause results to be frequently pushed, affecting storage and query performance.
+Since the output of TSMA is a super table, the row length of the output table is subject to the maximum row length limit. The size of the `intermediate results of different functions` varies, but they are generally larger than the original data size. If the row length of the output table exceeds the maximum row length limit, an error `Row length exceeds max length` will be reported. In this case, you need to reduce the number of functions or split commonly used functions groups into multiple TSMA objects.
+
+The window size is limited to [1ms ~ 1h]. The unit of INTERVAL is the same as the INTERVAL clause in the query, such as a (milliseconds), b (nanoseconds), h (hours), m (minutes), s (seconds), u (microseconds).
+
+TSMA is a database-level object, but it is globally unique. The number of TSMA that can be created in the cluster is limited by the parameter `maxTsmaNum`, with a default value of 8 and a range of [0-12]. Note that since TSMA background calculation uses stream computing, creating a TSMA will create a stream. Therefore, the number of TSMA that can be created is also limited by the number of existing streams and the maximum number of streams that can be created.
+
+## Supported Functions
+| function |  comments |
+|---|---|
+|min||
+|max||
+|sum||
+|first||
+|last||
+|avg||
+|count| If you want to use count(*), you should create the count(ts) function|
+|spread||
+|stddev||
+|hyperloglog||
+|||
+
+## Drop TSMA
+```sql
+DROP TSMA [db_name.]tsma_name;
+```
+
+If there are other TSMA created based on the TSMA being deleted, the delete operation will report an `Invalid drop base tsma, drop recursive tsma first` error. Therefore, all Recursive TSMA must be deleted first.
+
+## TSMA Calculation
+The calculation result of TSMA is a super table in the same database as the original table, but it is not visible to users. It cannot be deleted and will be automatically deleted when `DROP TSMA` is executed. The calculation of TSMA is done through stream computing, which is a background asynchronous process. The calculation result of TSMA is not guaranteed to be real-time, but it can guarantee eventual correctness.
+
+When there is a large amount of historical data, after creating TSMA, the stream computing will first calculate the historical data. During this period, newly created TSMA will not be used. The calculation will be automatically recalculated when data updates, deletions, or expired data arrive. During the recalculation period, the TSMA query results are not guaranteed to be real-time. If you want to query real-time data, you can use the hint `/*+ skip_tsma() */` in the SQL statement or disable the `querySmaOptimize` parameter to query from the original data.
+
+## Using and Limitations of TSMA
+
+Client configuration parameter: `querySmaOptimize`, used to control whether to use TSMA during queries. Set it to `True` to use TSMA, and `False` to query from the original data.
+
+Client configuration parameter: `maxTsmaCalcDelay`, in seconds, is used to control the acceptable TSMA calculation delay for users. If the calculation progress of a TSMA is within this range from the latest time, the TSMA will be used. If it exceeds this range, it will not be used. The default value is 600 (10 minutes), with a minimum value of 600 (10 minutes) and a maximum value of 86400 (1 day).
+
+### Using TSMA Duraing Query
+
+The aggregate functions defined in TSMA can be directly used in most query scenarios. If multiple TSMA are available, the one with the larger window size is preferred. For unclosed windows, the calculation can be done using smaller window TSMA or the original data. However, there are certain scenarios where TSMA cannot be used (see below). In such cases, the entire query will be calculated using the original data.
+
+The default behavior for queries without specified window sizes is to prioritize the use of the largest window TSMA that includes all the aggregate functions used in the query. For example, `SELECT COUNT(*) FROM stable GROUP BY tbname` will use the TSMA with the largest window that includes the `count(ts)` function. Therefore, when using aggregate queries frequently, it is recommended to create TSMA objects with larger window size.
+
+When specifying the window size, which is the `INTERVAL` statement, use the largest TSMA window that is divisible by the window size of the query. In window queries, the window size of the `INTERVAL`, `OFFSET`, and `SLIDING` all affect the TSMA window size that can be used. Divisible window TSMA refers to a TSMA window size that is divisible by the `INTERVAL`, `OFFSET`, and `SLIDING` of the query statement. Therefore, when using window queries frequently, consider the window size, as well as the offset and sliding size when creating TSMA objects.
+
+Example 1. If TSMA with window size of `5m` and `10m` is created, and the query is `INTERVAL(30m)`, the TSMA with window size of `10m` will be used. If the query is `INTERVAL(30m, 10m) SLIDING(5m)`, only the TSMA with window size of `5m` can be used for the query.
+
+### Limitations of Query
+
+When the parameter `querySmaOptimize` is enabled and there is no `skip_tsma()` hint, the following query scenarios cannot use TSMA:
+
+- When the aggregate functions defined in a TSMA do not cover the function list of the current query.
+- Non-`INTERVAL` windows or the query window size (including `INTERVAL, SLIDING, OFFSET`) is not multiples of the defined window size. For example, if the defined window is 2m and the query uses a 5-minute window, but if there is a 1m window available, it can be used.
+- Query with filtering on any regular column (non-primary key time column) in the `WHERE` condition.
+- When `PARTITION` or `GROUP BY` includes any regular column or its expression
+- When other faster optimization logic can be used, such as last cache optimization, if it meets the conditions for last optimization, it will be prioritized. If last optimization is not possible, then it will be determined whether TSMA optimization can be used.
+- When the current TSMA calculation progress delay is greater than the configuration parameter `maxTsmaCalcDelay`
+
+Some examples:
 
 ```sql
-DROP DATABASE IF EXISTS d0;
+SELECT agg_func_list [, pesudo_col_list] FROM stable WHERE exprs [GROUP/PARTITION BY [tbname] [, tag_list]] [HAVING ...] [INTERVAL(time_duration, offset) SLIDING(duration)]...;
-CREATE DATABASE d0;
+
-USE d0;
+-- create
-CREATE TABLE IF NOT EXISTS st1 (ts timestamp, c1 int, c2 float, c3 double) TAGS (t1 int unsigned);
+CREATE TSMA tsma1 ON stable FUNCTION(COUNT(ts), SUM(c1), SUM(c3), MIN(c1), MIN(c3), AVG(c1)) INTERVAL(1m);
-CREATE TABLE ct1 USING st1 TAGS(1000);
+-- query
-CREATE TABLE ct2 USING st1 TAGS(2000);
+SELECT COUNT(*), SUM(c1) + SUM(c3) FROM stable; ---- use tsma1
-INSERT INTO ct1 VALUES(now+0s, 10, 2.0, 3.0);
+SELECT COUNT(*), AVG(c1) FROM stable GROUP/PARTITION BY tbname, tag1, tag2;  --- use tsma1
-INSERT INTO ct1 VALUES(now+1s, 11, 2.1, 3.1)(now+2s, 12, 2.2, 3.2)(now+3s, 13, 2.3, 3.3);
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(1h);  ---use tsma1
-CREATE SMA INDEX sma_index_name1 ON st1 FUNCTION(max(c1),max(c2),min(c1)) INTERVAL(5m,10s) SLIDING(5m) WATERMARK 5s MAX_DELAY 1m;
+SELECT COUNT(*), MIN(c1), SPREAD(c1) FROM stable INTERVAL(1h); ----- can't use, spread func not defined, although SPREAD can be calculated by MIN and MAX which are defined.
--- query from SMA Index
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(30s); ----- can't use tsma1, time_duration not fit. Normally, query_time_duration should be multple of create_duration.
-ALTER LOCAL 'querySmaOptimize' '1';
+SELECT COUNT(*), MIN(c1) FROM stable where c2 > 0; ---- can't use tsma1, can't do c2 filtering
-SELECT max(c2),min(c1) FROM st1 INTERVAL(5m,10s) SLIDING(5m);
+SELECT COUNT(*) FROM stable GROUP BY c2; ---- can't use any tsma
-SELECT _wstart,_wend,_wduration,max(c2),min(c1) FROM st1 INTERVAL(5m,10s) SLIDING(5m);
+SELECT MIN(c3), MIN(c2) FROM stable INTERVAL(1m); ---- can't use tsma1, c2 is not defined in tsma1.
--- query from raw data
+
-ALTER LOCAL 'querySmaOptimize' '0';
+-- Another tsma2 created with INTERVAL(1h) based on tsma1
+CREATE RECURSIVE TSMA tsma2 on tsma1 INTERVAL(1h);
+SELECT COUNT(*), SUM(c1) FROM stable; ---- use tsma2
+SELECT COUNT(*), AVG(c1) FROM stable GROUP/PARTITION BY tbname, tag1, tag2;  --- use tsma2
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(2h);  ---use tsma2
+SELECT COUNT(*), MIN(c1) FROM stable WHERE ts < '2023-01-01 10:10:10' INTERVAL(30m); --use tsma1
+SELECT COUNT(*), MIN(c1) + MIN(c3) FROM stable INTERVAL(30m);  ---use tsma1
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(1h) SLIDING(30m);  ---use tsma1
+SELECT COUNT(*), MIN(c1), SPREAD(c1) FROM stable INTERVAL(1h); ----- can't use tsma1 or tsma2, spread func not defined
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(30s); ----- can't use tsma1 or tsma2, time_duration not fit. Normally, query_time_duration should be multple of create_duration.
+SELECT COUNT(*), MIN(c1) FROM stable where c2 > 0; ---- can't use tsma1 or tsam2, can't do c2 filtering
 ```
 
-## Delete an Index
+### Limitations of Usage
+
+After creating a TSMA, there are certain restrictions on operations that can be performed on the original table:
+
+- You must delete all TSMAs on the table before you can delete the table itself.
+- All tag columns of the original table cannot be deleted, nor can the tag column names or sub-table tag values be modified. You must first delete the TSMA before you can delete the tag column.
+- If some columns are being used by the TSMA, these columns cannot be deleted. You must first delete the TSMA. However, adding new columns to the table is not affected. However, new columns added are not included in any TSMA, so if you want to calculate the new columns, you need to create new TSMA for them.
+
+## Show TSMA
 
 ```sql
-DROP INDEX index_name;
+SHOW [db_name.]TSMAS;
+SELECT * FROM information_schema.ins_tsma;
 ```
 
-## View Indices
+If more functions are specified during creation, and the column names are longer, the function list may be truncated when displayed (currently supports a maximum output of 256KB)
-
-````sql
-SHOW INDEXES FROM tbl_name [FROM db_name];
-SHOW INDEXES FROM [db_name.]tbl_name ;
-````
-
-Shows indices that have been created for the specified database or table.

docs/en/12-taos-sql/30-join.md

@@ -0,0 +1,298 @@
+---
+sidebar_label: JOIN
+title: JOIN
+description: JOIN Description
+---
+
+
+## Join Concept
+
+### Driving Table
+
+The table used for driving Join queries is the left table in the Left Join series and the right table in the Right Join series.
+
+### Join Conditions
+
+Join conditions refer to the conditions specified for join operation. All join queries supported by TDengine require specifying join conditions. Join conditions usually only appear in `ON` (except for Inner Join and Window Join). For Inner Join, conditions that appear in `WHERE` can also be regarded as join conditions. For Window Join join conditions are specified in `WINDOW_OFFSET` clause.
+
+ Except for ASOF Join, all join types supported by TDengine must explicitly specify join conditions. Since ASOF Join has implicit join conditions defined by default, it is not necessary to explicitly specify the join conditions (if the default conditions meet the requirements).
+
+Except for ASOF/Window Join, the join condition can include not only the primary join condition(refer below), but also any number of other join conditions. The primary join condition must have an `AND` relationship with the other join conditions, while there is no such restriction between the other join conditions. The other join conditions can include any logical operation combination of primary key columns, Tag columns, normal columns, constants, and their scalar functions or operations.
+
+
+Taking smart meters as an example, the following SQL statements all contain valid join conditions:
+
+```sql
+SELECT a.* FROM meters a LEFT JOIN meters b ON a.ts = b.ts AND a.ts > '2023-10-18 10:00:00.000';
+SELECT a.* FROM meters a LEFT JOIN meters b ON a.ts = b.ts AND (a.ts > '2023-10-18 10:00:00.000' OR a.ts < '2023-10-17 10:00:00.000');
+SELECT a.* FROM meters a LEFT JOIN meters b ON timetruncate(a.ts, 1s) = timetruncate(b.ts, 1s) AND (a.ts + 1s > '2023-10-18 10:00:00.000' OR a.groupId > 0);
+SELECT a.* FROM meters a LEFT ASOF JOIN meters b ON timetruncate(a.ts, 1s) < timetruncate(b.ts, 1s) AND a.groupId = b.groupId;
+```
+
+### Primary Join Condition
+
+As a time series database, all join queries of TDengine revolve around the primary timestamp column, so all join queries except ASOF/Window Join are required to contain equivalent join condition of the primary key column. The equivalent join condition of the primary key column that first appear in the join conditions in order will be used as the primary join condition. The primary join condition of ASOF Join can contain non-equivalent join condition, for Window Join the primary join condition is specified by `WINDOW_OFFSET` clause.
+
+Except for Window Join, TDengine supports performing `timetruncate` function operation in the primary join condition, e.g. `ON timetruncate(a.ts, 1s) = timetruncate(b.ts, 1s)`. Other functions and scalar operations to primary key column are not currently supported in the primary join condition.
+
+### Grouping Conditions
+
+ASOF/Window Join supports grouping the input data of join queries, and then performing join operations within each group. Grouping only applies to the input of join queries, and the output result will not include grouping information. Equivalent conditions that appear in `ON` in ASOF/Window Join (excluding the primary join condition of ASOF) will be used as grouping conditions.
+
+
+### Primary Key Timeline
+
+TDengine, as a time series database, requires that each table must have a primary key timestamp column, which will perform many time-related operations as the primary key timeline of the table. It is also necessary to clarify which column will be regarded as the primary key timeline for subsequent time-related operations in the results of subqueries or join operations. In subqueries, the ordered first occurrence of the primary key column (or its operation) or the pseudo-column equivalent to the primary key column (`_wstart`/`_wend`) in the query results will be regarded as the primary key timeline of the output table. The selection of the primary key timeline in the join output results follows the following rules:
+
+- The primary key column of the driving table (subquery) in the Left/Right Join series will be used as the primary key timeline for subsequent queries. In addition, in each Window Join window, because the left and right tables are ordered at the same time, the primary key column of any table can be used as the primary key timeline in the window, and the primary key column of current table is preferentially selected as the primary key timeline.
+
+- The primary key column of any table in Inner Join can be treated as the primary key timeline. When there are similar grouping conditions (equivalent conditions of TAG columns and `AND` relationship with the primary join condition), there will be no available primary key timeline.
+
+- Full Join will not result in any primary key timeline because it cannot generate any valid primary key time series, so no timeline-related operations can be performed in or after a Full Join.
+
+
+## Syntax Conventions
+Because we will introduce the Left/Right Join series simultaneously through sharing below, the introductions of Left/Right Outer, Semi, Anti-Semi, ASOF, and Window series Joins will all use a similar "left/right" approach to introduce Left/Right Join simultaneously. Here is a brief introduction to the meaning of this writing method. The words written before "/" are the words applied to Left Join, and the words written after "/" are the words applied to Right Join. 
+
+For example:
+
+The phrase "left/right table" means "left table" for Left Join and "right table" for Right Join.
+
+Similarly,
+
+The phrase "right/left table" means "right table" for Left Join and "left table" for Right Join.
+
+## Join Function
+
+### Inner Join
+
+#### Definition
+Only data from both left and right tables that meet the join conditions will be returned, which can be regarded as the intersection of data from two tables that meet the join conditions.
+
+#### Grammar
+```sql
+SELECT ... FROM table_name1 [INNER] JOIN table_name2 [ON ...] [WHERE ...] [...]
+Or
+SELECT ... FROM table_name1, table_name2 WHERE ... [...]
+```
+#### Result set
+Cartesian product set of left and right table row data that meets the join conditions.
+
+#### Scope
+Inner Join are supported between super tables, normal tables, child tables, and subqueries.
+
+#### Notes
+- For the first type syntax, the `INNER` keyword is optional. The primary join condition and other join conditions can be specified in `ON` and/or `WHERE`, and filters can also be specified in `WHERE`. At least one of `ON`/`WHERE` must be specified.
+- For the second type syntax, all primary join condition, other join conditions, and filters can be specified in `WHERE`.
+- When performing Inner Join on the super table, the Tag column equivalent conditions with the `AND` relationship of the primary join condition will be used as a similar grouping condition, so the output result cannot remain time serious ordered.
+
+#### Examples
+
+The timestamp when the voltage is greater than 220V occurs simultaneously in table d1001 and table d1002 and their respective voltage values:
+```sql
+SELECT a.ts, a.voltage, b.voltage FROM d1001 a JOIN d1002 b ON a.ts = b.ts and a.voltage > 220 and b.voltage > 220
+```
+
+
+### Left/Right Outer Join
+
+
+#### Definition
+It returns data sets that meet the join conditions for both left and right tables, as well as data sets that do not meet the join conditions in the left/right tables.
+
+#### Grammar
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT [OUTER] JOIN table_name2 ON ... [WHERE ...] [...]
+```
+
+#### Result set
+The result set of Inner Join are rows in the left/right table that do not meet the join conditions combining with null data (`NULL`) in the right/left table.
+
+#### Scope
+Left/Right Outer Join are supported between super tables, normal tables, child tables, and subqueries.
+
+#### Notes
+- the `OUTER` keyword is optional.
+
+#### Examples
+
+Timestamp and voltage values at all times in table d1001 and the timestamp when the voltage is greater than 220V occurs simultaneously in table d1001 and table d1002 and their respective voltage values:
+```sql
+SELECT a.ts, a.voltage, b.voltage FROM d1001 a LEFT JOIN d1002 b ON a.ts = b.ts and a.voltage > 220 and b.voltage > 220
+```
+
+### Left/Right Semi Join
+
+#### Definition
+It usually expresses the meaning of `IN`/`EXISTS`, which means that for any data in the left/right table, only when there is any row data in the right/left table that meets the join conditions, will the left/right table row data be returned.
+
+#### Grammar
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT SEMI JOIN table_name2 ON ... [WHERE ...] [...]
+```
+
+#### Result set
+The row data set composed of rows that meet the join conditions in the left/right table and any one row that meets the join conditions in the right/left table.
+
+#### Scope
+Left/Right Semi Join are supported between super tables, normal tables, child tables, and subqueries.
+
+#### Examples
+
+The timestamp when the voltage in table d1001 is greater than 220V and there are other meters with voltages greater than 220V at the same time:
+```sql
+SELECT a.ts FROM d1001 a LEFT SEMI JOIN meters b ON a.ts = b.ts and a.voltage > 220 and b.voltage > 220 and b.tbname != 'd1001'
+```
+
+### Left/Right Anti-Semi Join
+
+#### Definition
+Opposite meaning to the Left/Right Semi Join. It usually expresses the meaning of `NOT IN`/`NOT EXISTS`, that is, for any row data in the left/right table, only will be returned when there is no row data that meets the join conditions in the right/left table.
+
+#### Grammar
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT ANTI JOIN table_name2 ON ... [WHERE ...] [...]
+```
+
+#### Result set
+A collection of rows in the left/right table that do not meet the join conditions and null data (`NULL`) in the right/left table.
+
+#### Scope
+Left/Right Anti-Semi Join are supported between super tables, normal tables, child tables, and subqueries.
+
+#### Examples
+
+The timestamp when the voltage in table d1001 is greater than 220V and there is not any other meters with voltages greater than 220V at the same time:
+```sql
+SELECT a.ts FROM d1001 a LEFT ANTI JOIN meters b ON a.ts = b.ts and b.voltage > 220 and b.tbname != 'd1001' WHERE a.voltage > 220
+```
+
+### left/Right ASOF Join
+
+#### Definition
+Different from other traditional join's exact matching patterns, ASOF Join allows for incomplete matching in a specified matching pattern, that is, matching in the manner closest to the primary key timestamp.
+
+#### Grammar
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT ASOF JOIN table_name2 [ON ...] [JLIMIT jlimit_num] [WHERE ...] [...]
+```
+
+##### Result set
+The Cartesian product set of up to `jlimit_num` rows data or null data (`NULL`) closest to the timestamp of each row in the left/right table, ordered by primary key, that meets the join conditions in the right/left table.
+
+
+##### Scope
+Left/Right ASOF Join are supported between super tables, normal tables, child tables.
+
+#### Notes
+- Only supports ASOF Join between tables, not between subqueries.
+- The `ON` clause supports a single matching rule (primary join condition) with the primary key column or the timetruncate function operation of the primary key column (other scalar operations and functions are not supported). The supported operators and their meanings are as follows:
+
+
+  |    **Operator**   |       **Meaning for Left ASOF Join**       |
+  | :-------------: | ------------------------ |
+  | >    | Match rows in the right table whose primary key timestamp is less than and the most closed to the left table's primary key timestamp      |
+  | >=    | Match rows in the right table whose primary key timestamp is less than or equal to and the most closed to the left table's primary key timestamp  |
+  | =    | Match rows in the right table whose primary key timestamp is equal to the left table's primary key timestamp  |
+  | <    | Match rows in the right table whose the primary key timestamp is greater than and the most closed to the left table's primary key timestamp  |
+  | <=    | Match rows in the right table whose primary key timestamp is greater than or equal to and the most closed to the left table's primary key timestamp  |
+
+  For Right ASOF Join, the above operators have the opposite meaning.
+
+- If there is no `ON` clause or no primary join condition is specified in the `ON` clause, the default primary join condition operator will be “>=”， that is, (for Left ASOF Join) matching rows in the right table whose primary key timestamp is less than or equal to the left table's primary key timestamp. Multiple primary join conditions are not supported.
+- In the `ON` clause, except for the primary key column, equivalent conditions between Tag columns and ordinary columns (which do not support scalar functions and operations) can be specified for grouping calculations. Other types of conditions are not supported.
+- Only `AND` operation is supported between all `ON` conditions.
+- `JLIMIT` is used to specify the maximum number of rows for a single row match result. It's optional. The default value is 1 when not specified, which means that each row of data in the left/right table can obtain at most one row of matching results from the right/left table. The value range of `JLIMIT` is [0,1024]. All the `jlimit_num` rows data that meet the join conditions do not require the same timestamp. When there are not enough `jlimit_num` rows data that meet the conditions in the right/left table, the number of returned result rows may be less than `jlimit_num`. When there are more than `jlimit_num` rows data that meet the conditions in the right/left table and all their timestamps are the same, random `jlimit_num` rows data will be returned.
+
+#### Examples
+
+The moment that voltage in table d1001 is greater than 220V and at the same time or at the last moment the voltage in table d1002 is also greater than 220V and their respective voltage values:
+```sql
+SELECT a.ts, a.voltage, a.ts, b.voltage FROM d1001 a LEFT ASOF JOIN d1002 b ON a.ts >= b.ts where a.voltage > 220 and b.voltage > 220 
+```
+
+### Left/Right Window Join
+
+#### Definition
+Construct windows based on the primary key timestamp of each row in the left/right table and the window boundary, and then perform window join accordingly, supporting projection, scalar, and aggregation operations within the window.
+
+#### Grammar
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT WINDOW JOIN table_name2 [ON ...] WINDOW_OFFSET(start_offset, end_offset) [JLIMIT jlimit_num] [WHERE ...] [...]
+```
+
+#### Result set
+The Cartesian product of each row of data in the left/right table and null data (`NULL`) or up to `jlimit_num` rows of data in the constructed window(based on the left/right table primary key timestamp and `WINDOW_OFFSET`) in the right/left table.
+Or 
+The Cartesian product of each row of data in the left/right table and null data (`NULL`) or the aggregation result of up to `jlimit_num` rows of data in the constructed window(based on the left/right table primary key timestamp and `WINDOW_OFFSET`) in the right/left table.
+
+#### Scope
+Left/Right Window Join are supported between super tables, normal tables, child tables.
+
+#### Notes
+- Only supports Window Join between tables, not between subqueries.
+- The `ON` clause is optional. Except for the primary key column, equivalent conditions between Tag columns and ordinary columns (which do not support scalar functions and operations) can be specified in `ON` clause for grouping calculations. Other types of conditions are not supported.
+- Only `AND` operation is supported between all `ON` conditions.
+- `WINDOW_OFFSET` is used to specify the offset of the left and right boundaries of the window relative to the timestamp of the left/right table's primary key. It supports the form of built-in time units. For example: `WINDOW_OFFSET (-1a, 1a)`, for Left Window Join, it means that each window boundary is [left table primary key timestamp - 1 millisecond, left table primary key timestamp + 1 millisecond], and both the left and right boundaries are closed intervals. The time unit after the number can be `b` (nanosecond), `u` (microsecond), `a` (millisecond), `s` (second), `m` (minute), `h` (hour), `d` (day), `w` (week). Natural months (`n`) and natural years (`y`) are not supported. The minimum time unit supported is database precision. The precision of the databases where the left and right tables are located should be the same.
+- `JLIMIT` is used to specify the maximum number of matching rows in a single window. Optional. If not specified, all matching rows in each window are obtained by default. The value range of `JLIMIT` is [0,1024]. Less than `jlimit_num` rows of data will be returned when there are not enough `jlimit_num` rows of data in the right table that meet the condition. When there are more than `jlimit_num` rows of data in the right table that meet the condition, `jlimit_num` rows of data with the smallest primary key timestamp in the window will be returned.
+- No `GROUP BY`/`PARTITION BY`/Window queries could be used together with Window Join in one single SQL statement.
+- Supports scalar filtering in the `WHERE` clause, aggregation function filtering for each window in the `HAVING` clause (does not support scalar filtering), does not support `SLIMIT`, and does not support various window pseudo-columns.
+
+#### Examples
+
+The voltage value of table d1002 within 1 second before and after the moment that voltage value of table d1001 is greater than 220V:
+```sql
+SELECT a.ts, a.voltage, b.voltage FROM d1001 a LEFT WINDOW JOIN d1002 b WINDOW_OFFSET（-1s, 1s) where a.voltage > 220
+```
+
+The moment that the voltage value of table d1001 is greater than 220V and the average voltage value of table d1002 is also greater than 220V in the interval of 1 second before and after that:
+```sql
+SELECT a.ts, a.voltage, avg(b.voltage) FROM d1001 a LEFT WINDOW JOIN d1002 b WINDOW_OFFSET（-1s, 1s) where a.voltage > 220 HAVING(avg(b.voltage) > 220)
+```
+
+### Full Outer Join
+
+#### Definition
+It includes data sets that meet the join conditions for both left and right tables, as well as data sets that do not meet the join conditions in the left and right tables.
+
+#### Grammar
+SELECT ... FROM table_name1 FULL [OUTER] JOIN table_name2 ON ... [WHERE ...] [...]
+
+#### Result set
+The result set of Inner Join + rows data set composed of rows in the left table that do not meet the join conditions and null data(`NULL`) in the right table + rows data set composed of rows in the right table that do not meet the join conditions and null data(`NULL`) in the left table.
+
+#### Scope
+Full Outer Join is supported between super tables, normal tables, child tables, and subqueries.
+
+#### Notes
+- the `OUTER` keyword is optional.
+
+#### Examples
+
+All timestamps and voltage values recorded in both tables d1001 and d1002:
+```sql
+SELECT a.ts, a.voltage, b.ts, b.voltage FROM d1001 a FULL JOIN d1002 b on a.ts = b.ts
+```
+
+## Limitations
+
+### Input timeline limits
+- Currently, all types of join require input data to contain a valid primary key timeline, which can be satisfied by all table queries. Subqueries need to pay attention to whether the output data contains a valid primary key timeline.
+
+### Join conditions limits
+- Except for ASOF and Window Join, the join conditions of other types of join must include the primary join condition; 
+- Only `AND` operation is supported between the primary join condition and other join conditions.
+- The primary key column used in the primary join condition only supports `timetruncate` function operations (not other functions and scalar operations), and there are no restrictions when used as other join conditions.
+
+### Grouping conditions limits
+- Only support equivalent conditions for Tag and ordinary columns except for primary key columns.
+- Does not support scalar operations.
+- Supports multiple grouping conditions, and only supports `AND` operation between conditions.
+
+### Query result order limits
+- In scenarios where there are normal tables, subtables, and subqueries without grouping conditions or sorting, the query results will be output in the order of the primary key columns of the driving table.
+- In scenarios such as super table queries, Full Join, or with grouping conditions and without sorting, there is no fixed output order for query results.
+Therefore, in scenarios where sorting is required and the output is not in a fixed order, sorting operations need to be performed. Some functions that rely on timelines may not be able to execute without soring due to the lack of valid timeline output.
+
+### Nested join and multi-table join limits
+- Currently, except for Inner Join which supports nesting and multi-table Join, other types of join do not support nesting and multi-table join.

docs/en/14-reference/12-config/index.md

@@ -241,6 +241,16 @@ Please note the `taoskeeper` needs to be installed and running to create the `lo
 | Default Value | 0          |
 | Notes         |  When this parameter is set to 0, last(\*)/last_row(\*)/first(\*) only returns the columns of the super table; When it is 1, return the columns and tags of the super table.               |
 
+### maxTsmaCalcDelay
+
+| Attribute     | Description                                                                                                                                  |
+| --------      | -------------------------------------------------------------------------------------------------------------------------------------------- |
+| Applicable    | Client only                                                                                                                                  |
+| Meaning       | Query allowed tsma calculation delay, if the tsma calculation delay is greater than the configured value, the TSMA will not be used.         |
+| Value Range   | 600s - 86400s, 10 minutes to 1 hour                                                                                                          |
+| Default value | 600s                                                                                                                                         |
+
+
 ## Locale Parameters
 
 ### timezone
@@ -760,6 +770,15 @@ The charset that takes effect is UTF-8.
 | Value Range | 1-10000|
 | Default Value   | 20                  |
 
+### maxTsmaNum
+
+| Attribute | Description                   |
+| --------- | ----------------------------- |
+| Applicable | Server Only                  |
+| Meaning   | Max num of TSMAs              |
+| Value Range | 0-12                        |
+| Default Value | 8                         |
+
 ## 3.0 Parameters
 
 | #   |     **Parameter**      | **Applicable to 2.x ** | **Applicable to  3.0 **      | Current behavior in 3.0 |

docs/zh/12-taos-sql/06-select.md

@@ -39,7 +39,7 @@ select_expr: {
 
 from_clause: {
     table_reference [, table_reference] ...
-  | join_clause [, join_clause] ...
+  | table_reference join_clause [, join_clause] ...
 }
 
 table_reference:
@@ -52,7 +52,7 @@ table_expr: {
 }
 
 join_clause:
-    table_reference [INNER] JOIN table_reference ON condition
+    [INNER|LEFT|RIGHT|FULL] [OUTER|SEMI|ANTI|ASOF|WINDOW] JOIN table_reference [ON condition] [WINDOW_OFFSET(start_offset, end_offset)] [JLIMIT jlimit_num]
 
 window_clause: {
     SESSION(ts_col, tol_val)
@@ -410,7 +410,9 @@ SELECT AVG(CASE WHEN voltage < 200 or voltage > 250 THEN 220 ELSE voltage END) F
 
 ## JOIN 子句
 
-TDengine 支持基于时间戳主键的内连接，即 JOIN 条件必须包含时间戳主键。只要满足基于时间戳主键这个要求，普通表、子表、超级表和子查询之间可以随意的进行内连接，且对表个数没有限制，其它连接条件与主键间必须是 AND 操作。
+在 3.3.0.0 版本之前 TDengine 只支持内连接，自 3.3.0.0 版本起 TDengine 支持了更为广泛的 JOIN 类型，这其中既包括传统数据库中的 LEFT JOIN、RIGHT JOIN、FULL JOIN、SEMI JOIN、ANTI-SEMI JOIN，也包括时序库中特色的 ASOF JOIN、WINDOW JOIN。JOIN 操作支持在子表、普通表、超级表以及子查询间进行。
+
+### 示例
 
 普通表与普通表之间的 JOIN 操作：
 
@@ -420,23 +422,23 @@ FROM temp_tb_1 t1, pressure_tb_1 t2
 WHERE t1.ts = t2.ts
 ```
 
-超级表与超级表之间的 JOIN 操作：
+超级表与超级表之间的 LEFT JOIN 操作：
 
 ```sql
 SELECT *
-FROM temp_stable t1, temp_stable t2
+FROM temp_stable t1 LEFT JOIN temp_stable t2
-WHERE t1.ts = t2.ts AND t1.deviceid = t2.deviceid AND t1.status=0;
+ON t1.ts = t2.ts AND t1.deviceid = t2.deviceid AND t1.status=0;
 ```
 
-子表与超级表之间的 JOIN 操作：
+子表与超级表之间的 LEFT ASOF JOIN 操作：
 
 ```sql
 SELECT *
-FROM temp_ctable t1, temp_stable t2
+FROM temp_ctable t1 LEFT ASOF JOIN temp_stable t2
-WHERE t1.ts = t2.ts AND t1.deviceid = t2.deviceid AND t1.status=0;
+ON t1.ts = t2.ts AND t1.deviceid = t2.deviceid;
 ```
 
-类似地，也可以对多个子查询的查询结果进行 JOIN 操作。
+更多 JOIN 操作相关介绍参见页面 [TDengine 关联查询](../join)
 
 ## 嵌套查询
 

docs/zh/12-taos-sql/07-tag-index.md

@@ -34,6 +34,13 @@ SELECT * FROM information_schema.INS_INDEXES
 
 也可以为上面的查询语句加上过滤条件以缩小查询范围。
 
+或者通过 SHOW 命令查看指定表上的索引
+
+```sql
+SHOW INDEXES FROM tbl_name [FROM db_name];
+SHOW INDEXES FROM [db_name.]tbl_name;
+```
+
 ## 使用说明
 
 1. 索引使用得当能够提升数据过滤的效率，目前支持的过滤算子有 `=`, `>`, `>=`, `<`, `<=`。如果查询过滤条件中使用了这些算子，则索引能够明显提升查询效率。但如果查询过滤条件中使用的是其它算子，则索引起不到作用，查询效率没有变化。未来会逐步添加更多的算子。

docs/zh/12-taos-sql/27-indexing.md

@@ -1,66 +1,132 @@
 ---
-sidebar_label: 索引
+sidebar_label: 窗口预聚集
-title: 索引
+title: 窗口预聚集
-description: 索引功能的使用细节
+description: 窗口预聚集使用说明
 ---
 
-TDengine 从 3.0.0.0 版本开始引入了索引功能，支持 SMA 索引和 tag 索引。
+为了提高大数据量的聚合函数查询性能，通过创建窗口预聚集 (TSMA Time-Range Small Materialized Aggregates) 对象, 使用固定时间窗口对指定的聚集函数进行预计算，并将计算结果存储下来，查询时通过查询预计算结果以提高查询性能。
 
-## 创建索引
+## 创建TSMA
 
 ```sql
+-- 创建基于超级表或普通表的tsma
+CREATE TSMA tsma_name ON [dbname.]table_name FUNCTION (func_name(func_param) [, ...] ) INTERVAL(time_duration);
+-- 创建基于小窗口tsma的大窗口tsma
+CREATE RECURSIVE TSMA tsma_name ON [db_name.]tsma_name1 INTERVAL(time_duration);
 
-CREATE INDEX index_name ON tb_name index_option
+time_duration:
-
+    number unit
-CREATE SMA INDEX index_name ON tb_name index_option
-
-index_option:
-    FUNCTION(functions) INTERVAL(interval_val [, interval_offset]) [SLIDING(sliding_val)] [WATERMARK(watermark_val)] [MAX_DELAY(max_delay_val)]
-
-functions:
-    function [, function] ...
 ```
-### tag 索引
 
- [tag 索引](../tag-index)
+创建 TSMA 时需要指定 TSMA 名字, 表名字, 函数列表以及窗口大小. 当基于 TSMA 创建时 TSMA 时, 即使用 `RECURSIVE` 关键字， 不需要指定 `FUNCTION()`， 将创建与已有 TSMA 相同的函数列表的TSMA， 且 INTERVAL 必须为所基于的TSMA窗口的整数倍。
 
-### SMA 索引
+其中 TSMA 命名规则与表名字类似, 长度最大限制为表名长度限制减去输出表后缀长度, 表名长度限制为193, 输出表后缀为`_tsma_res_stb_`, TSMA 名字最大长度为178.
 
-对指定列按 INTERVAL 子句定义的时间窗口创建进行预聚合计算，预聚合计算类型由 functions_string 指定。SMA 索引能提升指定时间段的聚合查询的性能。目前，限制一个超级表只能创建一个 SMA INDEX。
+TSMA只能基于超级表和普通表创建, 不能基于子表创建.
 
-- 支持的函数包括 MAX、MIN 和 SUM。
+函数列表中只能指定支持的聚集函数(见下文), 并且函数参数必须为1个, 即使当前函数支持多个参数, 函数参数内必须为普通列名, 不能为标签列. 函数列表中完全相同的函数和列会被去重, 如同时创建两个avg(c1), 则只会计算一个输出. TSMA 计算时将会把所有`函数中间结果`都输出到另一张超级表中, 输出超级表还包含了原始表的所有tag列. 函数列表中函数个数最多支持创建表最大列个数(包括tag列)减去 TSMA 计算附加的四列, 分别为`_wstart`, `_wend`, `_wduration`, 以及一个新增tag列 `tbname`, 再减去原始表的tag列数. 若列个数超出限制, 会报`Too many columns`错误. 
-- WATERMARK: 最小单位毫秒，取值范围 [0ms, 900000ms]，默认值为 5 秒，只可用于超级表。
+
-- MAX_DELAY: 最小单位毫秒，取值范围 [1ms, 900000ms]，默认值为 interval 的值(但不能超过最大值)，只可用于超级表。注：不建议 MAX_DELAY 设置太小，否则会过于频繁的推送结果，影响存储和查询性能，如无特殊需求，取默认值即可。
+由于TSMA输出为一张超级表, 因此输出表的行长度受最大行长度限制, 不同函数的`中间结果`大小各异, 一般都大于原始数据大小, 若输出表的行长度大于最大行长度限制, 将会报`Row length exceeds max length`错误. 此时需要减少函数个数或者将常用的函数进行分组拆分到多个TSMA中.
+
+窗口大小的限制为[1ms ~ 1h]. INTERVAL 的单位与查询中INTERVAL字句相同, 如 a (毫秒), b (纳秒), h (小时), m (分钟), s (秒), u (微妙).
+
+TSMA为库内对象, 但名字全局唯一. 集群内一共可创建TSMA个数受参数`maxTsmaNum`限制, 参数默认值为8, 范围: [0-12]. 注意, 由于TSMA后台计算使用流计算, 因此每创建一条TSMA, 将会创建一条流, 因此能够创建的TSMA条数也受当前已经存在的流条数和最大可创建流条数限制.
+
+## 支持的函数列表
+| 函数|  备注 |
+|---|---|
+|min||
+|max||
+|sum||
+|first||
+|last||
+|avg||
+|count| 若想使用count(*), 则应创建count(ts)函数|
+|spread||
+|stddev||
+|hyperloglog||
+|||
+
+## 删除TSMA
+```sql
+DROP TSMA [db_name.]tsma_name;
+```
+若存在其他TSMA基于当前被删除TSMA创建, 则删除操作报`Invalid drop base tsma, drop recursive tsma first`错误. 因此需先删除 所有Recursive TSMA.
+
+## TSMA的计算
+TSMA的计算结果为与原始表相同库下的一张超级表, 此表用户不可见. 不可删除, 在`DROP TSMA`时自动删除. TSMA的计算是通过流计算完成的, 此过程为后台异步过程, TSMA的计算结果不保证实时性, 但可以保证最终正确性.
+
+当存在大量历史数据时, 创建TSMA之后, 流计算将会首先计算历史数据, 此期间新创建的TSMA不会被使用. 数据更新删除或者过期数据到来时自动重新计算影响部分数据。 在重新计算期间 TSMA 查询结果不保证实时性。若希望查询实时数据， 可以通过在 SQL 中添加 hint `/*+ skip_tsma() */` 或者关闭参数`querySmaOptimize`从原始数据查询。
+
+## TSMA的使用与限制
+
+客户端配置参数: `querySmaOptimize`, 用于控制查询时是否使用TSMA, `True`为使用, `False`为不使用即从原始数据查询.
+
+客户端配置参数：`maxTsmaCalcDelay`，单位 s，用于控制用户可以接受的 TSMA 计算延迟，若 TSMA 的计算进度与最新时间差距在此范围内， 则该 TSMA 将会被使用， 若超出该范围， 则不使用， 默认值： 600（10 分钟）， 最小值： 600（10 分钟）， 最大值： 86400（1 天）.
+
+### 查询时使用TSMA
+
+已在 TSMA 中定义的 agg 函数在大部分查询场景下都可直接使用， 若存在多个可用的 TSMA， 优先使用大窗口的 TSMA， 未闭合窗口通过查询小窗口TSMA或者原始数据计算。 同时也有某些场景不能使用 TSMA(见下文)。 不可用时整个查询将使用原始数据进行计算。 
+
+未指定窗口大小的查询语句默认优先使用包含所有查询聚合函数的最大窗口 TSMA 进行数据的计算。 如`SELECT COUNT(*) FROM stable GROUP BY tbname`将会使用包含count(ts)且窗口最大的TSMA。因此若使用聚合查询频率高时, 应当尽可能创建大窗口的TSMA.
+
+指定窗口大小时即 `INTERVAL` 语句，使用最大的可整除窗口 TSMA。 窗口查询中， `INTERVAL` 的窗口大小， `OFFSET` 以及 `SLIDING` 都影响能使用的 TSMA 窗口大小， 可整 除窗口 TSMA 即 TSMA 窗口大小可被查询语句的 `INTERVAL， OFFSET， SLIDING` 整除的窗口。因此若使用窗口查询较多时, 需要考虑经常查询的窗口大小, 以及 offset, sliding大小来创建TSMA.
+
+例 1. 如 创建 TSMA 窗口大小 `5m` 一条， `10m` 一条， 查询时 `INTERVAL(30m)`， 那么优先使用 `10m` 的 TSMA， 若查询为 `INTERVAL(30m, 10m) SLIDING(5m)`， 那么仅可使用 `5m` 的 TSMA 查询。
+
+
+### 查询限制
+
+在开启了参数`querySmaOptimize`并且无`skip_tsma()` hint时, 以下查询场景无法使用TSMA:
+
+- 某个TSMA 中定义的 agg 函数不能覆盖当前查询的函数列表时
+- 非 `INTERVAL` 的其他窗口，或者 `INTERVAL` 查询窗口大小（包括 `INTERVAL，SLIDING，OFFSET`）不是定义窗口的整数倍，如定义窗口为 2m，查询使用 5 分钟窗口，但若存在 1m 的窗口，则可以使用。
+- 查询 `WHERE` 条件中包含任意普通列(非主键时间列)的过滤。
+- `PARTITION` 或者 `GROUY BY` 包含任意普通列或其表达式时
+- 可以使用其他更快的优化逻辑时， 如last cache优化, 若符合last优化的条件, 则先走last 优化, 无法走last时, 再判断是否可以走tsma优化
+- 当前 TSMA 计算进度延迟大于配置参数 `maxTsmaCalcDelay`时
+
+下面是一些例子:
 
 ```sql
-DROP DATABASE IF EXISTS d0;
+SELECT agg_func_list [, pesudo_col_list] FROM stable WHERE exprs [GROUP/PARTITION BY [tbname] [, tag_list]] [HAVING ...] [INTERVAL(time_duration, offset) SLIDING(duration)]...;
-CREATE DATABASE d0;
+
-USE d0;
+-- 创建
-CREATE TABLE IF NOT EXISTS st1 (ts timestamp, c1 int, c2 float, c3 double) TAGS (t1 int unsigned);
+CREATE TSMA tsma1 ON stable FUNCTION(COUNT(ts), SUM(c1), SUM(c3), MIN(c1), MIN(c3), AVG(c1)) INTERVAL(1m);
-CREATE TABLE ct1 USING st1 TAGS(1000);
+-- 查询
-CREATE TABLE ct2 USING st1 TAGS(2000);
+SELECT COUNT(*), SUM(c1) + SUM(c3) FROM stable; ---- use tsma1
-INSERT INTO ct1 VALUES(now+0s, 10, 2.0, 3.0);
+SELECT COUNT(*), AVG(c1) FROM stable GROUP/PARTITION BY tbname, tag1, tag2;  --- use tsma1
-INSERT INTO ct1 VALUES(now+1s, 11, 2.1, 3.1)(now+2s, 12, 2.2, 3.2)(now+3s, 13, 2.3, 3.3);
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(1h);  ---use tsma1
-CREATE SMA INDEX sma_index_name1 ON st1 FUNCTION(max(c1),max(c2),min(c1)) INTERVAL(5m,10s) SLIDING(5m) WATERMARK 5s MAX_DELAY 1m;
+SELECT COUNT(*), MIN(c1), SPREAD(c1) FROM stable INTERVAL(1h); ----- can't use, spread func not defined, although SPREAD can be calculated by MIN and MAX which are defined.
--- 从 SMA 索引查询
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(30s); ----- can't use tsma1, time_duration not fit. Normally, query_time_duration should be multple of create_duration.
-ALTER LOCAL 'querySmaOptimize' '1';
+SELECT COUNT(*), MIN(c1) FROM stable where c2 > 0; ---- can't use tsma1, can't do c2 filtering
-SELECT max(c2),min(c1) FROM st1 INTERVAL(5m,10s) SLIDING(5m);
+SELECT COUNT(*) FROM stable GROUP BY c2; ---- can't use any tsma
-SELECT _wstart,_wend,_wduration,max(c2),min(c1) FROM st1 INTERVAL(5m,10s) SLIDING(5m);
+SELECT MIN(c3), MIN(c2) FROM stable INTERVAL(1m); ---- can't use tsma1, c2 is not defined in tsma1.
--- 从原始数据查询
+
-ALTER LOCAL 'querySmaOptimize' '0'; 
+-- Another tsma2 created with INTERVAL(1h) based on tsma1
+CREATE RECURSIVE TSMA tsma2 on tsma1 INTERVAL(1h);
+SELECT COUNT(*), SUM(c1) FROM stable; ---- use tsma2
+SELECT COUNT(*), AVG(c1) FROM stable GROUP/PARTITION BY tbname, tag1, tag2;  --- use tsma2
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(2h);  ---use tsma2
+SELECT COUNT(*), MIN(c1) FROM stable WHERE ts < '2023-01-01 10:10:10' INTERVAL(30m); --use tsma1
+SELECT COUNT(*), MIN(c1) + MIN(c3) FROM stable INTERVAL(30m);  ---use tsma1
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(1h) SLIDING(30m);  ---use tsma1
+SELECT COUNT(*), MIN(c1), SPREAD(c1) FROM stable INTERVAL(1h); ----- can't use tsma1 or tsma2, spread func not defined
+SELECT COUNT(*), MIN(c1) FROM stable INTERVAL(30s); ----- can't use tsma1 or tsma2, time_duration not fit. Normally, query_time_duration should be multple of create_duration.
+SELECT COUNT(*), MIN(c1) FROM stable where c2 > 0; ---- can't use tsma1 or tsam2, can't do c2 filtering
 ```
 
-## 删除索引
+### 使用限制
 
+创建TSMA之后, 对原始超级表的操作有以下限制:
+
+- 必须删除该表上的所有TSMA才能删除该表.
+- 原始表所有tag列不能删除, 也不能修改tag列名或子表的tag值, 必须先删除TSMA, 才能删除tag列.
+- 若某些列被TSMA使用了, 则这些列不能被删除, 必须先删除TSMA. 添加列不受影响, 但是新添加的列不在任何TSMA中, 因此若要计算新增列, 需要新创建其他的TSMA.
+
+## 查看TSMA
 ```sql
-DROP INDEX index_name;
+SHOW [db_name.]TSMAS;
+SELECT * FROM information_schema.ins_tsma;
 ```
-
+若创建时指定的较多的函数, 且列名较长, 在显示函数列表时可能会被截断(目前最大支持输出256KB).
-## 查看索引
-
-````sql
-SHOW INDEXES FROM tbl_name [FROM db_name];
-SHOW INDEXES FROM [db_name.]tbl_name;
-````
-
-显示在所指定的数据库或表上已创建的索引。

docs/zh/12-taos-sql/30-join.md

@@ -0,0 +1,290 @@
+---
+sidebar_label: 关联查询
+title: 关联查询
+description: 关联查询详细描述
+---
+
+## Join 概念
+
+### 驱动表
+
+驱动关联查询进行的表，在 Left Join 系列中左表为驱动表，在 Right Join 系列中右表为驱动表。
+
+### 连接条件
+
+连接条件是指进行表关联所指定的条件，TDengine 支持的所有关联查询都需要指定连接条件，连接条件通常（Inner Join 和 Window Join 例外）只出现在 `ON` 之后。根据语义，Inner Join 中出现在 `WHERE` 之后的条件也可以视作连接条件，而 Window Join 是通过 `WINDOW_OFFSET` 来指定连接条件。
+
+  除 ASOF Join 外，TDengine 支持的所有 Join 类型都必须显式指定连接条件，ASOF Join 因为默认定义有隐式的连接条件，所以（在默认条件可以满足需求的情况下）可以不必显式指定连接条件。
+
+除 ASOF/Window Join 外，连接条件中除了包含主连接条件外，还可以包含任意多条其他连接条件，主连接条件与其他连接条件间必须是 `AND` 关系，而其他连接条件之间则没有这个限制。其他连接条件中可以包含主键列、Tag 、普通列、常量及其标量函数或运算的任意逻辑运算组合。
+
+以智能电表为例，下面这几条 SQL 都包含合法的连接条件：
+
+```sql
+SELECT a.* FROM meters a LEFT JOIN meters b ON a.ts = b.ts AND a.ts > '2023-10-18 10:00:00.000';
+SELECT a.* FROM meters a LEFT JOIN meters b ON a.ts = b.ts AND (a.ts > '2023-10-18 10:00:00.000' OR a.ts < '2023-10-17 10:00:00.000');
+SELECT a.* FROM meters a LEFT JOIN meters b ON timetruncate(a.ts, 1s) = timetruncate(b.ts, 1s) AND (a.ts + 1s > '2023-10-18 10:00:00.000' OR a.groupId > 0);
+SELECT a.* FROM meters a LEFT ASOF JOIN meters b ON timetruncate(a.ts, 1s) < timetruncate(b.ts, 1s) AND a.groupId = b.groupId;
+```
+
+### 主连接条件
+
+作为一款时序数据库，TDengine 所有的关联查询都围绕主键时戳列进行，因此要求除 ASOF/Window Join 外的所有关联查询都必须含有主键列的等值连接条件，而按照顺序首次出现在连接条件中的主键列等值连接条件将会被作为主连接条件。ASOF Join 的主连接条件可以包含非等值的连接条件，而 Window Join 的主连接条件则是通过 `WINDOW_OFFSET` 来指定。
+
+除 Window Join 外，TDengine 支持在主连接条件中进行 `timetruncate` 函数操作，例如 `ON timetruncate(a.ts, 1s) = timetruncate(b.ts, 1s)`，除此之外，暂不支持其他函数及标量运算。
+
+### 分组条件
+
+时序数据库特色的 ASOF/Window Join 支持对关联查询的输入数据进行分组，然后每个分组内进行关联操作。分组只对关联查询的输入进行，输出结果将不包含分组信息。ASOF/Window Join 中出现在 `ON` 之后的等值条件（ASOF 的主连接条件除外）将被作为分组条件。
+
+### 主键时间线
+
+TDengine 作为时序数据库要求每个表（子表）中必须有主键时间戳列，它将作为该表的主键时间线进行很多跟时间相关的运算，而子查询的结果或者 Join 运算的结果中也需要明确哪一列将被视作主键时间线参与后续的时间相关的运算。在子查询中，查询结果中存在的有序的第一个出现的主键列（或其运算）或等同主键列的伪列（`_wstart`/`_wend`）将被视作该输出表的主键时间线。Join 输出结果中主键时间线的选择遵从以下规则：
+- Left/Right Join 系列中驱动表（子查询）的主键列将被作为后续查询的主键时间线；此外，在 Window Join 窗口内，因为左右表同时有序所以在窗口内可以把任意一个表的主键列做作主键时间线，优先选择本表的主键列作为主键时间线。
+- Inner Join 可以把任意一个表的主键列做作主键时间线，当存在类似分组条件（Tag 列的等值条件且与主连接条件 `AND` 关系）时将无法产生主键时间线。
+- Full Join 因为无法产生任何一个有效的主键时间序列，因此没有主键时间线，这也就意味着 Full Join 中无法进行时间线相关的运算。
+
+
+## 语法说明
+
+在接下来的章节中会通过共用的方式同时介绍 Left/Right Join 系列，因此后续的包括 Outer、Semi、Anti-Semi、ASOF、Window 系列介绍中都采用了类似 "left/right" 的写法来同时进行 Left/Right Join 的介绍。这里简要介绍这种写法的含义，写在 "/" 前面的表示应用于 Left Join，而写在 "/" 后面的表示应用于 Right Join。
+
+举例说明：
+
+"左/右表" 表示对 Left Join 来说，它指的是"左表"，对 Right Join 来说，它指的是“右表”；
+
+同理，
+
+"右/左表" 表示对 Left Join 来说，它指的是"右表"，对 Right Join 来说，它指的是“左表”；
+
+
+## Join 功能
+
+### Inner Join
+
+#### 定义
+内连接 - 只有左右表中同时符合连接条件的数据才会被返回，可以视为两个表符合连接条件的数据的交集。
+
+#### 语法
+```sql
+SELECT ... FROM table_name1 [INNER] JOIN table_name2 [ON ...] [WHERE ...] [...]
+或
+SELECT ... FROM table_name1, table_name2 WHERE ... [...]
+```
+#### 结果集
+符合连接条件的左右表行数据的笛卡尔积集合。
+
+#### 适用范围
+支持超级表、普通表、子表、子查询间 Inner Join。
+
+#### 说明
+- 对于第一种语法，`INNER` 关键字可选， `ON` 和/或 `WHERE` 中可以指定主连接条件和其他连接条件，`WHERE` 中还可以指定过滤条件，`ON`/`WHERE` 两者至少指定一个。
+- 对于第二种语法，可以在 `WHERE` 中指定主连接条件、其他连接条件、过滤条件。
+- 对超级表进行 Inner Join 时，与主连接条件 `AND` 关系的 Tag 列等值条件将作为类似分组条件使用，因此输出结果不能保持有序。
+
+#### 示例
+
+表 d1001 和表 d1002 中同时出现电压大于 220V 的时刻及各自的电压值：
+```sql
+SELECT a.ts, a.voltage, b.voltage FROM d1001 a JOIN d1002 b ON a.ts = b.ts and a.voltage > 220 and b.voltage > 220
+```
+
+
+### Left/Right Outer Join
+
+#### 定义
+左/右（外）连接 - 既包含左右表同时符合连接条件的数据集合，也包括左/右表中不符合连接条件的数据集合。
+
+#### 语法
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT [OUTER] JOIN table_name2 ON ... [WHERE ...] [...]
+```
+
+#### 结果集
+Inner Join 的结果集 + 左/右表中不符合连接条件的行和右/左表的空数据（`NULL`）组成的行数据集合。
+
+#### 适用范围
+支持超级表、普通表、子表、子查询间 Left/Right Join。
+
+#### 说明
+- OUTER 关键字可选。
+
+#### 示例
+
+表 d1001 所有时刻的电压值以及和表 d1002 中同时出现电压大于 220V 的时刻及各自的电压值：
+```sql
+SELECT a.ts, a.voltage, b.voltage FROM d1001 a LEFT JOIN d1002 b ON a.ts = b.ts and a.voltage > 220 and b.voltage > 220
+```
+
+### Left/Right Semi Join
+
+#### 定义
+左/右半连接 - 通常表达的是 `IN``/EXISTS` 的含义，即对左/右表任意一条数据来说，只有当右/左表中存在任一符合连接条件的数据时才返回左/右表行数据。
+
+#### 语法
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT SEMI JOIN table_name2 ON ... [WHERE ...] [...]
+```
+
+#### 结果集
+左/右表中符合连接条件的行和右/左表任一符合连接条件的行组成的行数据集合。
+
+#### 适用范围
+支持超级表、普通表、子表、子查询间 Left/Right Semi Join。
+
+#### 示例
+
+表 d1001 中出现电压大于 220V 且存在其他电表同一时刻电压也大于 220V 的时间：
+```sql
+SELECT a.ts FROM d1001 a LEFT SEMI JOIN meters b ON a.ts = b.ts and a.voltage > 220 and b.voltage > 220 and b.tbname != 'd1001'
+```
+
+### Left/Right Anti-Semi Join
+
+#### 定义
+左/右反连接 - 同左/右半连接的逻辑正好相反，通常表达的是 `NOT IN`/`NOT EXISTS` 的含义，即对左/右表任意一条数据来说，只有当右/左表中不存在任何符合连接条件的数据时才返回左/右表行数据。
+
+#### 语法
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT ANTI JOIN table_name2 ON ... [WHERE ...] [...]
+```
+
+#### 结果集
+左/右表中不符合连接条件的行和右/左表的空数据（`NULL`）组成的行数据集合。
+
+#### 适用范围
+支持超级表、普通表、子表、子查询间 Left/Right Anti-Semi Join。
+
+#### 示例
+
+表 d1001 中出现电压大于 220V 且不存在其他电表同一时刻电压也大于 220V 的时间：
+```sql
+SELECT a.ts FROM d1001 a LEFT ANTI JOIN meters b ON a.ts = b.ts and b.voltage > 220 and b.tbname != 'd1001' WHERE a.voltage > 220
+```
+
+### left/Right ASOF Join
+
+#### 定义
+左/右不完全匹配连接 - 不同于其他传统 Join 的完全匹配模式，ASOF Join 允许以指定的匹配模式进行不完全匹配，即按照主键时间戳最接近的方式进行匹配。
+
+#### 语法
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT ASOF JOIN table_name2 [ON ...] [JLIMIT jlimit_num] [WHERE ...] [...]
+```
+
+##### 结果集
+左/右表中每一行数据与右/左表中符合连接条件的按主键列排序后时间戳最接近的最多 `jlimit_num` 条数据或空数据（`NULL`）的笛卡尔积集合。
+
+##### 适用范围
+支持超级表、普通表、子表间 Left/Right ASOF Join。
+
+#### 说明
+- 只支持表间 ASOF Join，不支持子查询间 ASOF Join。
+- ON 子句中支持指定主键列或主键列的 timetruncate 函数运算（不支持其他标量运算及函数）后的单个匹配规则（主连接条件），支持的运算符及其含义如下：
+
+
+  |    **运算符**   |       **Left ASOF 时含义**       |
+  | :-------------: | ------------------------ |
+  | >    | 匹配右表中主键时间戳小于左表主键时间戳且时间戳最接近的数据行      |
+  | >=    | 匹配右表中主键时间戳小于等于左表主键时间戳且时间戳最接近的数据行  |
+  | =    | 匹配右表中主键时间戳等于左表主键时间戳的行  |
+  | <    | 匹配右表中主键时间戳大于左表主键时间戳且时间戳最接近的数据行  |
+  | <=    | 匹配右表中主键时间戳大于等于左表主键时间戳且时间戳最接近的数据行  |
+
+  对于 Right ASOF 来说，上述运算符含义正好相反。
+
+- 如果不含 `ON` 子句或 `ON` 子句中未指定主键列的匹配规则，则默认主键匹配规则运算符是 “>=”， 即（对 Left ASOF Join 来说）右表中主键时戳小于等于左表主键时戳的行数据。不支持多个主连接条件。
+- `ON` 子句中还可以指定除主键列外的 Tag、普通列（不支持标量函数及运算）之间的等值条件用于分组计算，除此之外不支持其他类型的条件。
+- 所有 ON 条件间只支持 `AND` 运算。
+- `JLIMIT` 用于指定单行匹配结果的最大行数，可选，未指定时默认值为1，即左/右表每行数据最多从右/左表中获得一行匹配结果。`JLIMIT` 取值范围为 [0, 1024]。符合匹配条件的 `jlimit_num` 条数据不要求时间戳相同，当右/左表中不存在满足条件的 `jlimit_num` 条数据时，返回的结果行数可能小于 `jlimit_num`；当右/左表中存在符合条件的多于 `jlimit_num` 条数据时，如果时间戳相同将随机返回 `jlimit_num` 条数据。
+
+#### 示例
+
+表 d1001 电压值大于 220V 且表 d1002 中同一时刻或稍早前最后时刻出现电压大于 220V 的时间及各自的电压值：
+```sql
+SELECT a.ts, a.voltage, a.ts, b.voltage FROM d1001 a LEFT ASOF JOIN d1002 b ON a.ts >= b.ts where a.voltage > 220 and b.voltage > 220 
+```
+
+### Left/Right Window Join
+
+#### 定义
+左/右窗口连接 - 根据左/右表中每一行的主键时间戳和窗口边界构造窗口并据此进行窗口连接，支持窗口内进行投影、标量和聚合操作。
+
+#### 语法
+```sql
+SELECT ... FROM table_name1 LEFT|RIGHT WINDOW JOIN table_name2 [ON ...] WINDOW_OFFSET(start_offset, end_offset) [JLIMIT jlimit_num] [WHERE ...] [...]
+```
+
+#### 结果集
+左/右表中每一行数据与右/左表中基于左/右表主键时戳列和 `WINDOW_OFFSET` 划分的窗口内的至多 `jlimit_num` 条数据或空数据（`NULL`）的笛卡尔积集合 或 
+左/右表中每一行数据与右/左表中基于左/右表主键时戳列和 `WINDOW_OFFSET` 划分的窗口内的至多 `jlimit_num` 条数据的聚合结果或空数据（`NULL`）组成的行数据集合。
+
+#### 适用范围
+支持超级表、普通表、子表间 Left/Right Window Join。
+
+#### 说明
+- 只支持表间 Window Join，不支持子查询间 Window Join；
+- `ON` 子句可选，只支持指定除主键列外的 Tag、普通列（不支持标量函数及运算）之间的等值条件用于分组计算，所有条件间只支持 `AND` 运算；
+- `WINDOW_OFFSET` 用于指定窗口的左右边界相对于左/右表主键时间戳的偏移量，支持自带时间单位的形式，例如：`WINDOW_OFFSET(-1a， 1a)`，对于 Left Window Join 来说，表示每个窗口为 [左表主键时间戳 - 1毫秒，左表主键时间戳 + 1毫秒] ，左右边界均为闭区间。数字后面的时间单位可以是 `b`（纳秒）、`u`（微秒）、`a`（毫秒）、`s`（秒）、`m`（分）、`h`（小时）、`d`（天）、`w`（周），不支持自然月（`n`）、自然年（`y`），支持的最小时间单位为数据库精度，左右表所在数据库精度需保持一致。
+- `JLIMIT` 用于指定单个窗口内的最大匹配行数，可选，未指定时默认获取每个窗口内的所有匹配行。`JLIMIT` 取值范围为 [0, 1024]，当右表中不存在满足条件的 `jlimit_num` 条数据时，返回的结果行数可能小于 `jlimit_num`；当右表中存在超过 `jlimit_num` 条满足条件的数据时，优先返回窗口内主键时间戳最小的 `jlimit_num` 条数据。
+- SQL 语句中不能含其他 `GROUP BY`/`PARTITION BY`/窗口查询；
+- 支持在 `WHERE` 子句中进行标量过滤，支持在 `HAVING` 子句中针对每个窗口进行聚合函数过滤（不支持标量过滤），不支持 `SLIMIT`，不支持各种窗口伪列；
+
+#### 示例
+
+表 d1001 电压值大于 220V 时前后1秒的区间内表 d1002 的电压值：
+```sql
+SELECT a.ts, a.voltage, b.voltage FROM d1001 a LEFT WINDOW JOIN d1002 b WINDOW_OFFSET（-1s, 1s) where a.voltage > 220
+```
+
+表 d1001 电压值大于 220V 且前后1秒的区间内表 d1002 的电压平均值也大于 220V 的时间及电压值：
+```sql
+SELECT a.ts, a.voltage, avg(b.voltage) FROM d1001 a LEFT WINDOW JOIN d1002 b WINDOW_OFFSET（-1s, 1s) where a.voltage > 220 HAVING(avg(b.voltage) > 220)
+```
+
+### Full Outer Join
+
+#### 定义
+全（外）连接 - 既包含左右表同时符合连接条件的数据集合，也包括左右表中不符合连接条件的数据集合。
+
+#### 语法
+SELECT ... FROM table_name1 FULL [OUTER] JOIN table_name2 ON ... [WHERE ...] [...]
+
+#### 结果集
+Inner Join 的结果集 + 左表中不符合连接条件的行加上右表的空数据组成的行数据集合 + 右表中不符合连接条件的行加上左表的空数据(`NULL`)组成的行数据集合。
+
+#### 适用范围
+支持超级表、普通表、子表、子查询间 Full Outer Join。
+
+#### 说明
+- OUTER 关键字可选。
+
+#### 示例
+
+表 d1001 和表 d1002 中记录的所有时刻及电压值：
+```sql
+SELECT a.ts, a.voltage, b.ts, b.voltage FROM d1001 a FULL JOIN d1002 b on a.ts = b.ts
+```
+
+## 约束和限制
+
+### 输入时间线限制
+- 目前所有 Join 都要求输入数据含有效的主键时间线，所有表查询都可以满足，子查询需要注意输出数据是否含有效的主键时间线。
+
+### 连接条件限制
+- 除 ASOF 和 Window Join 之外，其他 Join 的连接条件中必须含主键列的主连接条件； 且
+- 主连接条件与其他连接条件间只支持 `AND` 运算；
+- 作为主连接条件的主键列只支持 `timetruncate` 函数运算（不支持其他函数和标量运算），作为其他连接条件时无限制；
+
+### 分组条件限制
+- 只支持除主键列外的 Tag、普通列的等值条件；
+- 不支持标量运算；
+- 支持多个分组条件，条件间只支持 `AND` 运算；
+
+### 查询结果顺序限制
+- 普通表、子表、子查询且无分组条件无排序的场景下，查询结果会按照驱动表的主键列顺序输出；
+- 超级表查询、Full Join或有分组条件无排序的场景下，查询结果没有固定的输出顺序；
+因此，在有排序需求且输出无固定顺序的场景下，需要进行排序操作。部分依赖时间线的函数可能会因为没有有效的时间线输出而无法执行。
+
+### 嵌套 Join 与多表 Join 限制
+- 目前除 Inner Join 支持嵌套与多表 Join 外，其他类型的 JoiN 暂不支持嵌套与多表 Join。

docs/zh/14-reference/12-config/index.md

@@ -240,6 +240,16 @@ taos -C
 | 缺省值   | 0                                                                                                                                              |
 | 补充说明 | 该参数设置为 0 时，last(\*)/last_row(\*)/first(\*) 只返回超级表的普通列；为 1 时，返回超级表的普通列和标签列                                            |
 
+### maxTsmaCalcDelay
+
+| 属性     | 说明                                                                                                                                           |
+| -------- | ---------------------------------------------------------------------------------------------------------------------------------------------- |
+| 适用范围 | 仅客户端适用                                                                                                                                   |
+| 含义     | 查询时客户端可允许的tsma计算延迟, 若tsma的计算延迟大于配置值, 则该TSMA将不会被使用.                                                  |
+| 取值范围 | 600s - 86400s, 即10分钟-1小时                                                                                                                         |
+| 缺省值   | 600s                                                                                                                                              |
+
+
 ## 区域相关
 
 ### timezone
@@ -745,6 +755,15 @@ charset 的有效值是 UTF-8。
 | 取值范围 | 1-10000                     |
 | 缺省值   | 20                          |
 
+### maxTsmaNum
+
+| 属性     | 说明                        |
+| -------- | --------------------------- |
+| 适用范围 | 仅服务端适用                |
+| 含义     | 集群内可创建的TSMA个数      |
+| 取值范围 | 0-12                        |
+| 缺省值   | 8                           |
+
 ## 压缩参数
 
 ### compressMsgSize

docs/zh/21-tdinternal/01-arch.md

@@ -201,7 +201,7 @@ TDengine 采用数据驱动的方式让缓存中的数据写入硬盘进行持
 
 除此之外，TDengine 也提供了数据分级存储的功能，将不同时间段的数据存储在挂载的不同介质上的目录里，从而实现不同“热度”的数据存储在不同的存储介质上，充分利用存储，节约成本。比如，最新采集的数据需要经常访问，对硬盘的读取性能要求高，那么用户可以配置将这些数据存储在 SSD 盘上。超过一定期限的数据，查询需求量没有那么高，那么可以存储在相对便宜的 HDD 盘上。
 
-多级存储支持 3 级，每级最多可配置 16 个挂载点。
+多级存储支持 3 级，每级最多可配置 128 个挂载点。
 
 TDengine 多级存储配置方式如下（在配置文件/etc/taos/taos.cfg 中）：
 

include/common/tcol.h

@@ -89,7 +89,7 @@ bool checkColumnLevelOrSetDefault(uint8_t type, char level[TSDB_CL_COMPRESS_OPTI
 void    setColEncode(uint32_t* compress, uint8_t encode);
 void    setColCompress(uint32_t* compress, uint16_t compressType);
 void    setColLevel(uint32_t* compress, uint8_t level);
-int8_t setColCompressByOption(uint8_t type, uint8_t encode, uint16_t compressType, uint8_t level, bool check,
+int32_t setColCompressByOption(uint8_t type, uint8_t encode, uint16_t compressType, uint8_t level, bool check,
                                uint32_t* compress);
 
 int8_t validColCompressLevel(uint8_t type, uint8_t level);
@@ -97,5 +97,5 @@ int8_t validColCompress(uint8_t type, uint8_t l2);
 int8_t validColEncode(uint8_t type, uint8_t l1);
 
 uint32_t createDefaultColCmprByType(uint8_t type);
-bool     validColCmprByType(uint8_t type, uint32_t cmpr);
+int32_t  validColCmprByType(uint8_t type, uint32_t cmpr);
 #endif /*_TD_TCOL_H_*/

include/common/tcommon.h

@@ -15,7 +15,7 @@
 
 #ifndef _TD_COMMON_DEF_H_
 #define _TD_COMMON_DEF_H_
-// #include "taosdef.h"
+
 #include "tarray.h"
 #include "tmsg.h"
 #include "tvariant.h"
@@ -412,6 +412,7 @@ typedef struct STUidTagInfo {
 #define UD_TAG_COLUMN_INDEX        2
 
 int32_t taosGenCrashJsonMsg(int signum, char** pMsg, int64_t clusterId, int64_t startTime);
+int32_t dumpConfToDataBlock(SSDataBlock* pBlock, int32_t startCol);
 
 #define TSMA_RES_STB_POSTFIX "_tsma_res_stb_"
 #define MD5_OUTPUT_LEN 32

include/common/tglobal.h

@@ -260,7 +260,7 @@ int32_t taosInitCfg(const char *cfgDir, const char **envCmd, const char *envFile
                     bool tsc);
 void    taosCleanupCfg();
 
-int32_t taosCfgDynamicOptions(SConfig *pCfg, char *name, bool forServer);
+int32_t taosCfgDynamicOptions(SConfig *pCfg, const char *name, bool forServer);
 
 struct SConfig *taosGetCfg();
 

include/libs/executor/storageapi.h

@@ -35,6 +35,7 @@ extern "C" {
 #define CACHESCAN_RETRIEVE_TYPE_SINGLE 0x2
 #define CACHESCAN_RETRIEVE_LAST_ROW    0x4
 #define CACHESCAN_RETRIEVE_LAST        0x8
+#define CACHESCAN_RETRIEVE_PK          0x10
 
 #define META_READER_LOCK   0x0
 #define META_READER_NOLOCK 0x1
@@ -307,6 +308,13 @@ typedef struct SUpdateInfo {
   SScalableBf* pCloseWinSBF;
   SHashObj*    pMap;
   uint64_t     maxDataVersion;
+  int8_t       pkColType;
+  int32_t      pkColLen;
+  char*        pKeyBuff;
+  char*        pValueBuff;
+
+  int (*comparePkRowFn)(void* pValue1, void* pTs, void* pPkVal, __compar_fn_t cmpPkFn);
+  __compar_fn_t comparePkCol;
 } SUpdateInfo;
 
 typedef struct {
@@ -374,17 +382,17 @@ typedef struct SStateStore {
                                               void** ppVal, int32_t* pVLen);
   int32_t (*streamStateCountWinAdd)(SStreamState* pState, SSessionKey* pKey, void** pVal, int32_t* pVLen);
 
-  SUpdateInfo* (*updateInfoInit)(int64_t interval, int32_t precision, int64_t watermark, bool igUp);
+  SUpdateInfo* (*updateInfoInit)(int64_t interval, int32_t precision, int64_t watermark, bool igUp, int8_t pkType, int32_t pkLen);
-  TSKEY (*updateInfoFillBlockData)(SUpdateInfo* pInfo, SSDataBlock* pBlock, int32_t primaryTsCol);
+  TSKEY (*updateInfoFillBlockData)(SUpdateInfo* pInfo, SSDataBlock* pBlock, int32_t primaryTsCol, int32_t primaryKeyCol);
-  bool (*updateInfoIsUpdated)(SUpdateInfo* pInfo, uint64_t tableId, TSKEY ts);
+  bool (*updateInfoIsUpdated)(SUpdateInfo* pInfo, uint64_t tableId, TSKEY ts, void* pPkVal, int32_t len);
   bool (*updateInfoIsTableInserted)(SUpdateInfo* pInfo, int64_t tbUid);
-  bool (*isIncrementalTimeStamp)(SUpdateInfo* pInfo, uint64_t tableId, TSKEY ts);
+  bool (*isIncrementalTimeStamp)(SUpdateInfo* pInfo, uint64_t tableId, TSKEY ts, void* pPkVal, int32_t len);
 
   void (*updateInfoDestroy)(SUpdateInfo* pInfo);
   void (*windowSBfDelete)(SUpdateInfo* pInfo, uint64_t count);
   void (*windowSBfAdd)(SUpdateInfo* pInfo, uint64_t count);
 
-  SUpdateInfo* (*updateInfoInitP)(SInterval* pInterval, int64_t watermark, bool igUp);
+  SUpdateInfo* (*updateInfoInitP)(SInterval* pInterval, int64_t watermark, bool igUp, int8_t pkType, int32_t pkLen);
   void (*updateInfoAddCloseWindowSBF)(SUpdateInfo* pInfo);
   void (*updateInfoDestoryColseWinSBF)(SUpdateInfo* pInfo);
   int32_t (*updateInfoSerialize)(void* buf, int32_t bufLen, const SUpdateInfo* pInfo);

include/libs/nodes/querynodes.h

@@ -270,7 +270,6 @@ typedef struct SJoinTableNode {
   SNode*       addPrimCond;
   bool         hasSubQuery;
   bool         isLowLevelJoin;
-  SNode*       pParent;
   SNode*       pLeft;
   SNode*       pRight;
   SNode*       pOnCond;

include/libs/stream/tstream.h

@@ -832,7 +832,7 @@ bool    streamHistoryTaskSetVerRangeStep2(SStreamTask* pTask, int64_t latestVer)
 int32_t streamQueueGetNumOfItems(const SStreamQueue* pQueue);
 
 // common
-void    streamTaskPause(SStreamMeta* pMeta, SStreamTask* pTask);
+void    streamTaskPause(SStreamTask* pTask);
 void    streamTaskResume(SStreamTask* pTask);
 int32_t streamTaskStop(SStreamTask* pTask);
 int32_t streamTaskSetUpstreamInfo(SStreamTask* pTask, const SStreamTask* pUpstreamTask);

include/libs/stream/tstreamUpdate.h

@@ -25,28 +25,10 @@
 extern "C" {
 #endif
 
-typedef struct SUpdateKey {
+SUpdateInfo *updateInfoInitP(SInterval *pInterval, int64_t watermark, bool igUp, int8_t pkType, int32_t pkLen);
-  int64_t tbUid;
+SUpdateInfo *updateInfoInit(int64_t interval, int32_t precision, int64_t watermark, bool igUp, int8_t pkType, int32_t pkLen);
-  TSKEY   ts;
+TSKEY        updateInfoFillBlockData(SUpdateInfo *pInfo, SSDataBlock *pBlock, int32_t primaryTsCol, int32_t primaryKeyCol);
-} SUpdateKey;
+bool         updateInfoIsUpdated(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts, void* pPkVal, int32_t len);
-
-//typedef struct SUpdateInfo {
-//  SArray      *pTsBuckets;
-//  uint64_t     numBuckets;
-//  SArray      *pTsSBFs;
-//  uint64_t     numSBFs;
-//  int64_t      interval;
-//  int64_t      watermark;
-//  TSKEY        minTS;
-//  SScalableBf *pCloseWinSBF;
-//  SHashObj    *pMap;
-//  uint64_t     maxDataVersion;
-//} SUpdateInfo;
-
-SUpdateInfo *updateInfoInitP(SInterval *pInterval, int64_t watermark, bool igUp);
-SUpdateInfo *updateInfoInit(int64_t interval, int32_t precision, int64_t watermark, bool igUp);
-TSKEY        updateInfoFillBlockData(SUpdateInfo *pInfo, SSDataBlock *pBlock, int32_t primaryTsCol);
-bool         updateInfoIsUpdated(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts);
 bool         updateInfoIsTableInserted(SUpdateInfo *pInfo, int64_t tbUid);
 void         updateInfoDestroy(SUpdateInfo *pInfo);
 void         updateInfoAddCloseWindowSBF(SUpdateInfo *pInfo);
@@ -55,7 +37,7 @@ int32_t      updateInfoSerialize(void *buf, int32_t bufLen, const SUpdateInfo *p
 int32_t      updateInfoDeserialize(void *buf, int32_t bufLen, SUpdateInfo *pInfo);
 void         windowSBfDelete(SUpdateInfo *pInfo, uint64_t count);
 void         windowSBfAdd(SUpdateInfo *pInfo, uint64_t count);
-bool         isIncrementalTimeStamp(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts);
+bool         isIncrementalTimeStamp(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts, void* pPkVal, int32_t len);
 
 #ifdef __cplusplus
 }

include/util/talgo.h

@@ -54,6 +54,12 @@ typedef int32_t (*__ext_compar_fn_t)(const void *p1, const void *p2, const void
  */
 void taosqsort(void *src, int64_t numOfElem, int64_t size, const void *param, __ext_compar_fn_t comparFn);
 
+/**
+ * Non-recursive quick sort.
+ *
+ */
+void taosqsort_r(void *src, int64_t nelem, int64_t size, const void *arg, __ext_compar_fn_t cmp);
+
 /**
  * merge sort, with the compare function requiring additional parameters support
  *

include/util/taoserror.h

@@ -182,6 +182,8 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_TSC_STMT_CACHE_ERROR          TAOS_DEF_ERROR_CODE(0, 0X0230)
 #define TSDB_CODE_TSC_ENCODE_PARAM_ERROR        TAOS_DEF_ERROR_CODE(0, 0X0231)
 #define TSDB_CODE_TSC_ENCODE_PARAM_NULL         TAOS_DEF_ERROR_CODE(0, 0X0232)
+#define TSDB_CODE_TSC_COMPRESS_PARAM_ERROR      TAOS_DEF_ERROR_CODE(0, 0X0233)
+#define TSDB_CODE_TSC_COMPRESS_LEVEL_ERROR TAOS_DEF_ERROR_CODE(0, 0X0234)
 #define TSDB_CODE_TSC_INTERNAL_ERROR            TAOS_DEF_ERROR_CODE(0, 0X02FF)
 
 // mnode-common
@@ -283,7 +285,6 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_MND_INVALID_STB_OPTION        TAOS_DEF_ERROR_CODE(0, 0x036E)
 #define TSDB_CODE_MND_INVALID_ROW_BYTES         TAOS_DEF_ERROR_CODE(0, 0x036F)
 #define TSDB_CODE_MND_FIELD_VALUE_OVERFLOW           TAOS_DEF_ERROR_CODE(0, 0x0370)
-#define TSDB_CODE_MND_COLUMN_COMPRESS_ALREADY_EXIST TAOS_DEF_ERROR_CODE(0, 0x0371)
 
 
 // mnode-func
@@ -406,6 +407,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_MND_INVALID_TARGET_TABLE      TAOS_DEF_ERROR_CODE(0, 0x03F7)
 
 
+#define TSDB_CODE_MND_COLUMN_COMPRESS_ALREADY_EXIST TAOS_DEF_ERROR_CODE(0, 0x03F8)
 
 // dnode
 // #define TSDB_CODE_DND_MSG_NOT_PROCESSED      TAOS_DEF_ERROR_CODE(0, 0x0400) // 2.x

include/util/tconfig.h

@@ -98,34 +98,32 @@ typedef struct {
   const char *value;
 } SConfigPair;
 
-typedef struct SConfig {
+typedef struct SConfig SConfig;
-  ECfgSrcType stype;
+typedef struct SConfigIter SConfigIter;
-  SArray     *array;
-} SConfig;
 
 SConfig     *cfgInit();
 int32_t      cfgLoad(SConfig *pCfg, ECfgSrcType cfgType, const void *sourceStr);
 int32_t      cfgLoadFromArray(SConfig *pCfg, SArray *pArgs);  // SConfigPair
 void         cfgCleanup(SConfig *pCfg);
-
 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      cfgCheckRangeForDynUpdate(SConfig *pCfg, const char *name, const char *pVal, bool isServer);
+SConfigIter *cfgCreateIter(SConfig *pConf);
+SConfigItem *cfgNextIter(SConfigIter *pIter);
+void         cfgDestroyIter(SConfigIter *pIter);
 
+// clang-format off
 int32_t cfgAddBool(SConfig *pCfg, const char *name, bool defaultVal, int8_t scope, int8_t dynScope);
-int32_t cfgAddInt32(SConfig *pCfg, const char *name, int32_t defaultVal, int64_t minval, int64_t maxval, int8_t scope,
+int32_t cfgAddInt32(SConfig *pCfg, const char *name, int32_t defaultVal, int64_t minval, int64_t maxval, int8_t scope, int8_t dynScope);
-                    int8_t dynScope);
+int32_t cfgAddInt64(SConfig *pCfg, const char *name, int64_t defaultVal, int64_t minval, int64_t maxval, int8_t scope, int8_t dynScope);
-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, float minval, float maxval, int8_t scope, int8_t dynScope);
-                    int8_t dynScope);
-int32_t cfgAddFloat(SConfig *pCfg, const char *name, float defaultVal, float minval, float maxval, int8_t scope,
-                    int8_t dynScope);
 int32_t cfgAddString(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope, int8_t dynScope);
 int32_t cfgAddDir(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope, int8_t dynScope);
 int32_t cfgAddLocale(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope, int8_t dynScope);
 int32_t cfgAddCharset(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope, int8_t dynScope);
 int32_t cfgAddTimezone(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope, int8_t dynScope);
+// clang-format on
 
 const char *cfgStypeStr(ECfgSrcType type);
 const char *cfgDtypeStr(ECfgDataType type);

include/util/tdef.h

@@ -525,7 +525,7 @@ typedef enum ELogicConditionType {
 #define TSDB_ARB_DUMMY_TIME 4765104000000  // 2121-01-01 00:00:00.000, :P
 
 #define TFS_MAX_TIERS          3
-#define TFS_MAX_DISKS_PER_TIER 16
+#define TFS_MAX_DISKS_PER_TIER 128
 #define TFS_MAX_DISKS          (TFS_MAX_TIERS * TFS_MAX_DISKS_PER_TIER)
 #define TFS_MIN_LEVEL          0
 #define TFS_MAX_LEVEL          (TFS_MAX_TIERS - 1)

include/util/thash.h

@@ -180,6 +180,13 @@ void taosHashCancelIterate(SHashObj *pHashObj, void *p);
  */
 void *taosHashGetKey(void *data, size_t *keyLen);
 
+/**
+ * Get the corresponding value length for a given data in hash table
+ * @param data
+ * @return
+ */
+int32_t taosHashGetValueSize(void *data);
+
 /**
  * return the payload data with the specified key(reference number added)
  *

source/client/src/clientEnv.c

@@ -770,6 +770,32 @@ int taos_init() {
   return tscInitRes;
 }
 
+const char* getCfgName(TSDB_OPTION option) {
+  const char* name = NULL;
+
+  switch (option) {
+    case TSDB_OPTION_SHELL_ACTIVITY_TIMER:
+      name = "shellActivityTimer";
+      break;
+    case TSDB_OPTION_LOCALE:
+      name = "locale";
+      break;
+    case TSDB_OPTION_CHARSET:
+      name = "charset";
+      break;
+    case TSDB_OPTION_TIMEZONE:
+      name = "timezone";
+      break;
+    case TSDB_OPTION_USE_ADAPTER:
+      name = "useAdapter";
+      break;
+    default:
+      break;
+  }
+
+  return name;
+}
+
 int taos_options_imp(TSDB_OPTION option, const char *str) {
   if (option == TSDB_OPTION_CONFIGDIR) {
 #ifndef WINDOWS
@@ -799,39 +825,26 @@ int taos_options_imp(TSDB_OPTION option, const char *str) {
 
   SConfig     *pCfg = taosGetCfg();
   SConfigItem *pItem = NULL;
+  const char  *name = getCfgName(option);
 
-  switch (option) {
+  if (name == NULL) {
-    case TSDB_OPTION_SHELL_ACTIVITY_TIMER:
+    tscError("Invalid option %d", option);
-      pItem = cfgGetItem(pCfg, "shellActivityTimer");
+    return -1;
-      break;
-    case TSDB_OPTION_LOCALE:
-      pItem = cfgGetItem(pCfg, "locale");
-      break;
-    case TSDB_OPTION_CHARSET:
-      pItem = cfgGetItem(pCfg, "charset");
-      break;
-    case TSDB_OPTION_TIMEZONE:
-      pItem = cfgGetItem(pCfg, "timezone");
-      break;
-    case TSDB_OPTION_USE_ADAPTER:
-      pItem = cfgGetItem(pCfg, "useAdapter");
-      break;
-    default:
-      break;
   }
 
+  pItem = cfgGetItem(pCfg, name);
   if (pItem == NULL) {
     tscError("Invalid option %d", option);
     return -1;
   }
 
-  int code = cfgSetItem(pCfg, pItem->name, str, CFG_STYPE_TAOS_OPTIONS);
+  int code = cfgSetItem(pCfg, name, str, CFG_STYPE_TAOS_OPTIONS);
   if (code != 0) {
-    tscError("failed to set cfg:%s to %s since %s", pItem->name, str, terrstr());
+    tscError("failed to set cfg:%s to %s since %s", name, str, terrstr());
   } else {
-    tscInfo("set cfg:%s to %s", pItem->name, str);
+    tscInfo("set cfg:%s to %s", name, str);
     if (TSDB_OPTION_SHELL_ACTIVITY_TIMER == option || TSDB_OPTION_USE_ADAPTER == option) {
-      code = taosCfgDynamicOptions(pCfg, pItem->name, false);
+      code = taosCfgDynamicOptions(pCfg, name, false);
     }
   }
 

source/client/src/clientRawBlockWrite.c

@@ -736,8 +736,8 @@ static int32_t taosCreateStb(TAOS* taos, void* meta, int32_t metaLen) {
     SSchema*          pSchema = req.schemaRow.pSchema + i;
     SFieldWithOptions field = {.type = pSchema->type, .flags = pSchema->flags, .bytes = pSchema->bytes};
     strcpy(field.name, pSchema->name);
-    // todo get active compress param
+    SColCmpr *p = &req.colCmpr.pColCmpr[i];
-    setDefaultOptionsForField(&field);
+    field.compress = p->alg;
     taosArrayPush(pReq.pColumns, &field);
   }
   pReq.pTags = taosArrayInit(req.schemaTag.nCols, sizeof(SField));

source/client/src/clientSml.c

@@ -1290,17 +1290,24 @@ end:
   return code;
 }
 
-static void smlInsertMeta(SHashObj *metaHash, SArray *metaArray, SArray *cols) {
+static int32_t smlInsertMeta(SHashObj *metaHash, SArray *metaArray, SArray *cols, SHashObj *checkDuplicate) {
+  terrno = 0;
   for (int16_t i = 0; i < taosArrayGetSize(cols); ++i) {
     SSmlKv *kv = (SSmlKv *)taosArrayGet(cols, i);
     int     ret = taosHashPut(metaHash, kv->key, kv->keyLen, &i, SHORT_BYTES);
     if (ret == 0) {
       taosArrayPush(metaArray, kv);
+      if(taosHashGet(checkDuplicate, kv->key, kv->keyLen) != NULL) {
+        return TSDB_CODE_PAR_DUPLICATED_COLUMN;
+      }
+    }else if(terrno == TSDB_CODE_DUP_KEY){
+      return TSDB_CODE_PAR_DUPLICATED_COLUMN;
     }
   }
+  return TSDB_CODE_SUCCESS;
 }
 
-static int32_t smlUpdateMeta(SHashObj *metaHash, SArray *metaArray, SArray *cols, bool isTag, SSmlMsgBuf *msg) {
+static int32_t smlUpdateMeta(SHashObj *metaHash, SArray *metaArray, SArray *cols, bool isTag, SSmlMsgBuf *msg, SHashObj* checkDuplicate) {
   for (int i = 0; i < taosArrayGetSize(cols); ++i) {
     SSmlKv *kv = (SSmlKv *)taosArrayGet(cols, i);
 
@@ -1332,6 +1339,11 @@ static int32_t smlUpdateMeta(SHashObj *metaHash, SArray *metaArray, SArray *cols
       int     ret = taosHashPut(metaHash, kv->key, kv->keyLen, &size, SHORT_BYTES);
       if (ret == 0) {
         taosArrayPush(metaArray, kv);
+        if(taosHashGet(checkDuplicate, kv->key, kv->keyLen) != NULL) {
+          return TSDB_CODE_PAR_DUPLICATED_COLUMN;
+        }
+      }else{
+        return ret;
       }
     }
   }
@@ -1456,7 +1468,7 @@ static int32_t smlPushCols(SArray *colsArray, SArray *cols) {
     taosHashPut(kvHash, kv->key, kv->keyLen, &kv, POINTER_BYTES);
     if (terrno == TSDB_CODE_DUP_KEY) {
       taosHashCleanup(kvHash);
-      return terrno;
+      return TSDB_CODE_PAR_DUPLICATED_COLUMN;
     }
   }
 
@@ -1512,12 +1524,12 @@ static int32_t smlParseLineBottom(SSmlHandle *info) {
     if (tableMeta) {  // update meta
       uDebug("SML:0x%" PRIx64 " smlParseLineBottom update meta, format:%d, linenum:%d", info->id, info->dataFormat,
              info->lineNum);
-      ret = smlUpdateMeta((*tableMeta)->colHash, (*tableMeta)->cols, elements->colArray, false, &info->msgBuf);
+      ret = smlUpdateMeta((*tableMeta)->colHash, (*tableMeta)->cols, elements->colArray, false, &info->msgBuf, (*tableMeta)->tagHash);
       if (ret == TSDB_CODE_SUCCESS) {
-        ret = smlUpdateMeta((*tableMeta)->tagHash, (*tableMeta)->tags, tinfo->tags, true, &info->msgBuf);
+        ret = smlUpdateMeta((*tableMeta)->tagHash, (*tableMeta)->tags, tinfo->tags, true, &info->msgBuf, (*tableMeta)->colHash);
       }
       if (ret != TSDB_CODE_SUCCESS) {
-        uError("SML:0x%" PRIx64 " smlUpdateMeta failed", info->id);
+        uError("SML:0x%" PRIx64 " smlUpdateMeta failed, ret:%d", info->id, ret);
         return ret;
       }
     } else {
@@ -1527,13 +1539,19 @@ static int32_t smlParseLineBottom(SSmlHandle *info) {
       if (meta == NULL) {
         return TSDB_CODE_OUT_OF_MEMORY;
       }
-      taosHashPut(info->superTables, elements->measure, elements->measureLen, &meta, POINTER_BYTES);
+      ret = taosHashPut(info->superTables, elements->measure, elements->measureLen, &meta, POINTER_BYTES);
-      terrno = 0;
+      if (ret != TSDB_CODE_SUCCESS) {
-      smlInsertMeta(meta->tagHash, meta->tags, tinfo->tags);
+        uError("SML:0x%" PRIx64 " put measuer to hash failed", info->id);
-      if (terrno == TSDB_CODE_DUP_KEY) {
+        return ret;
-        return terrno;
+      }
+      ret = smlInsertMeta(meta->tagHash, meta->tags, tinfo->tags, NULL);
+      if (ret == TSDB_CODE_SUCCESS) {
+        ret = smlInsertMeta(meta->colHash, meta->cols, elements->colArray, meta->tagHash);
+      }
+      if (ret != TSDB_CODE_SUCCESS) {
+        uError("SML:0x%" PRIx64 " insert meta failed:%s", info->id, tstrerror(ret));
+        return ret;
       }
-      smlInsertMeta(meta->colHash, meta->cols, elements->colArray);
     }
   }
   uDebug("SML:0x%" PRIx64 " smlParseLineBottom end, format:%d, linenum:%d", info->id, info->dataFormat, info->lineNum);

source/client/src/clientTmq.c

@@ -660,13 +660,13 @@ static void asyncCommitAllOffsets(tmq_t* tmq, tmq_commit_cb* pCommitFp, void* us
 
   taosRLockLatch(&tmq->lock);
   int32_t numOfTopics = taosArrayGetSize(tmq->clientTopics);
-  tscInfo("consumer:0x%" PRIx64 " start to commit offset for %d topics", tmq->consumerId, numOfTopics);
+  tscDebug("consumer:0x%" PRIx64 " start to commit offset for %d topics", tmq->consumerId, numOfTopics);
 
   for (int32_t i = 0; i < numOfTopics; i++) {
     SMqClientTopic* pTopic = taosArrayGet(tmq->clientTopics, i);
     int32_t         numOfVgroups = taosArrayGetSize(pTopic->vgs);
 
-    tscInfo("consumer:0x%" PRIx64 " commit offset for topics:%s, numOfVgs:%d", tmq->consumerId, pTopic->topicName,
+    tscDebug("consumer:0x%" PRIx64 " commit offset for topics:%s, numOfVgs:%d", tmq->consumerId, pTopic->topicName,
             numOfVgroups);
     for (int32_t j = 0; j < numOfVgroups; j++) {
       SMqClientVg* pVg = taosArrayGet(pTopic->vgs, j);
@@ -688,19 +688,19 @@ static void asyncCommitAllOffsets(tmq_t* tmq, tmq_commit_cb* pCommitFp, void* us
           continue;
         }
 
-        tscInfo("consumer:0x%" PRIx64
+        tscDebug("consumer:0x%" PRIx64
                 " topic:%s on vgId:%d send commit msg success, send offset:%s committed:%s, ordinal:%d/%d",
                 tmq->consumerId, pTopic->topicName, pVg->vgId, offsetBuf, commitBuf, j + 1, numOfVgroups);
         tOffsetCopy(&pVg->offsetInfo.committedOffset, &pVg->offsetInfo.endOffset);
       } else {
-        tscInfo("consumer:0x%" PRIx64 " topic:%s vgId:%d, no commit, current:%" PRId64 ", ordinal:%d/%d",
+        tscDebug("consumer:0x%" PRIx64 " topic:%s vgId:%d, no commit, current:%" PRId64 ", ordinal:%d/%d",
                 tmq->consumerId, pTopic->topicName, pVg->vgId, pVg->offsetInfo.endOffset.version, j + 1, numOfVgroups);
       }
     }
   }
   taosRUnLockLatch(&tmq->lock);
 
-  tscInfo("consumer:0x%" PRIx64 " total commit:%d for %d topics", tmq->consumerId, pParamSet->waitingRspNum - 1,
+  tscDebug("consumer:0x%" PRIx64 " total commit:%d for %d topics", tmq->consumerId, pParamSet->waitingRspNum - 1,
           numOfTopics);
 
   // request is sent
@@ -815,7 +815,7 @@ void tmqSendHbReq(void* param, void* tmrId) {
       offRows->ever = pVg->offsetInfo.walVerEnd;
       char buf[TSDB_OFFSET_LEN] = {0};
       tFormatOffset(buf, TSDB_OFFSET_LEN, &offRows->offset);
-      tscInfo("consumer:0x%" PRIx64 ",report offset, group:%s vgId:%d, offset:%s/%" PRId64 ", rows:%" PRId64,
+      tscDebug("consumer:0x%" PRIx64 ",report offset, group:%s vgId:%d, offset:%s/%" PRId64 ", rows:%" PRId64,
               tmq->consumerId, tmq->groupId, offRows->vgId, buf, offRows->ever, offRows->rows);
     }
   }
@@ -1058,6 +1058,7 @@ static void tmqMgmtInit(void) {
 
 #define SET_ERROR_MSG_TMQ(MSG) \
   if (errstr != NULL) snprintf(errstr, errstrLen, MSG);
+
 tmq_t* tmq_consumer_new(tmq_conf_t* conf, char* errstr, int32_t errstrLen) {
   if (conf == NULL) {
     SET_ERROR_MSG_TMQ("configure is null")
@@ -1504,7 +1505,7 @@ static bool doUpdateLocalEp(tmq_t* tmq, int32_t epoch, const SMqAskEpRsp* pRsp)
 
   int32_t topicNumGet = taosArrayGetSize(pRsp->topics);
   if (epoch < tmq->epoch || (epoch == tmq->epoch && topicNumGet == 0)) {
-    tscInfo("consumer:0x%" PRIx64 " no update ep epoch from %d to epoch %d, incoming topics:%d", tmq->consumerId,
+    tscDebug("consumer:0x%" PRIx64 " no update ep epoch from %d to epoch %d, incoming topics:%d", tmq->consumerId,
             tmq->epoch, epoch, topicNumGet);
     if (atomic_load_8(&tmq->status) == TMQ_CONSUMER_STATUS__RECOVER) {
       atomic_store_8(&tmq->status, TMQ_CONSUMER_STATUS__READY);
@@ -1800,14 +1801,14 @@ static int32_t tmqPollImpl(tmq_t* tmq, int64_t timeout) {
     for (int j = 0; j < numOfVg; j++) {
       SMqClientVg* pVg = taosArrayGet(pTopic->vgs, j);
       if (taosGetTimestampMs() - pVg->emptyBlockReceiveTs < EMPTY_BLOCK_POLL_IDLE_DURATION) {  // less than 10ms
-        tscTrace("consumer:0x%" PRIx64 " epoch %d, vgId:%d idle for 10ms before start next poll", tmq->consumerId,
+        tscDebug("consumer:0x%" PRIx64 " epoch %d, vgId:%d idle for 10ms before start next poll", tmq->consumerId,
                  tmq->epoch, pVg->vgId);
         continue;
       }
 
       if (tmq->replayEnable &&
           taosGetTimestampMs() - pVg->blockReceiveTs < pVg->blockSleepForReplay) {  // less than 10ms
-        tscTrace("consumer:0x%" PRIx64 " epoch %d, vgId:%d idle for %" PRId64 "ms before start next poll when replay",
+        tscDebug("consumer:0x%" PRIx64 " epoch %d, vgId:%d idle for %" PRId64 "ms before start next poll when replay",
                  tmq->consumerId, tmq->epoch, pVg->vgId, pVg->blockSleepForReplay);
         continue;
       }
@@ -1815,7 +1816,7 @@ static int32_t tmqPollImpl(tmq_t* tmq, int64_t timeout) {
       int32_t vgStatus = atomic_val_compare_exchange_32(&pVg->vgStatus, TMQ_VG_STATUS__IDLE, TMQ_VG_STATUS__WAIT);
       if (vgStatus == TMQ_VG_STATUS__WAIT) {
         int32_t vgSkipCnt = atomic_add_fetch_32(&pVg->vgSkipCnt, 1);
-        tscTrace("consumer:0x%" PRIx64 " epoch %d wait poll-rsp, skip vgId:%d skip cnt %d", tmq->consumerId, tmq->epoch,
+        tscDebug("consumer:0x%" PRIx64 " epoch %d wait poll-rsp, skip vgId:%d skip cnt %d", tmq->consumerId, tmq->epoch,
                  pVg->vgId, vgSkipCnt);
         continue;
       }
@@ -1875,7 +1876,7 @@ static void* tmqHandleAllRsp(tmq_t* tmq, int64_t timeout) {
       SMqPollRspWrapper* pollRspWrapper = (SMqPollRspWrapper*)pRspWrapper;
       if (pRspWrapper->code == TSDB_CODE_TMQ_CONSUMER_MISMATCH) {
         atomic_store_8(&tmq->status, TMQ_CONSUMER_STATUS__RECOVER);
-        tscDebug("consumer:0x%" PRIx64 " wait for the re-balance, set status to be RECOVER", tmq->consumerId);
+        tscDebug("consumer:0x%" PRIx64 " wait for the rebalance, set status to be RECOVER", tmq->consumerId);
       } else if (pRspWrapper->code == TSDB_CODE_TQ_NO_COMMITTED_OFFSET) {
         terrno = pRspWrapper->code;
         tscError("consumer:0x%" PRIx64 " unexpected rsp from poll, code:%s", tmq->consumerId,
@@ -2476,7 +2477,7 @@ int32_t askEpCb(void* param, SDataBuf* pMsg, int32_t code) {
 
   SMqRspHead* head = pMsg->pData;
   int32_t     epoch = atomic_load_32(&tmq->epoch);
-  tscInfo("consumer:0x%" PRIx64 ", recv ep, msg epoch %d, current epoch %d", tmq->consumerId, head->epoch, epoch);
+  tscDebug("consumer:0x%" PRIx64 ", recv ep, msg epoch %d, current epoch %d", tmq->consumerId, head->epoch, epoch);
   if (pParam->sync) {
     SMqAskEpRsp rsp = {0};
     tDecodeSMqAskEpRsp(POINTER_SHIFT(pMsg->pData, sizeof(SMqRspHead)), &rsp);
@@ -2581,7 +2582,7 @@ void askEp(tmq_t* pTmq, void* param, bool sync, bool updateEpSet) {
   sendInfo->msgType = TDMT_MND_TMQ_ASK_EP;
 
   SEpSet epSet = getEpSet_s(&pTmq->pTscObj->pAppInfo->mgmtEp);
-  tscInfo("consumer:0x%" PRIx64 " ask ep from mnode, reqId:0x%" PRIx64, pTmq->consumerId, sendInfo->requestId);
+  tscDebug("consumer:0x%" PRIx64 " ask ep from mnode, reqId:0x%" PRIx64, pTmq->consumerId, sendInfo->requestId);
 
   int64_t transporterId = 0;
   code = asyncSendMsgToServer(pTmq->pTscObj->pAppInfo->pTransporter, &epSet, &transporterId, sendInfo);

source/client/test/clientTests.cpp

@@ -827,10 +827,14 @@ TEST(clientCase, projection_query_tables) {
   //  }
   //  taos_free_result(pRes);
 
-  TAOS_RES* pRes = taos_query(pConn, "use test");
+  TAOS_RES* pRes = taos_query(pConn, "use cache_1");
   taos_free_result(pRes);
 
-  pRes = taos_query(pConn, "create table st2 (ts timestamp, k int primary key, j varchar(1000)) tags(a int)");
+  pRes = taos_query(pConn, "select last(ts), ts from cache_1.t1");
+//  pRes = taos_query(pConn, "select last(ts), ts from cache_1.no_pk_t1");
+  if (taos_errno(pRes) != 0) {
+    printf("failed to exec query, %s\n", taos_errstr(pRes));
+  }
   taos_free_result(pRes);
 
 //  pRes = taos_query(pConn, "create stream stream_1 trigger at_once fill_history 1 ignore expired 0 into str_res1 as select _wstart as ts, count(*) from stable_1 interval(10s);");

source/common/src/tcol.c

@@ -306,22 +306,22 @@ void setColLevel(uint32_t* compress, uint8_t level) {
   return;
 }
 
-int8_t setColCompressByOption(uint8_t type, uint8_t encode, uint16_t compressType, uint8_t level, bool check,
+int32_t setColCompressByOption(uint8_t type, uint8_t encode, uint16_t compressType, uint8_t level, bool check,
                                uint32_t* compress) {
-  if (check && !validColEncode(type, encode)) return 0;
+  if (check && !validColEncode(type, encode)) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
   setColEncode(compress, encode);
 
   if (compressType == TSDB_COLVAL_COMPRESS_DISABLED) {
     setColCompress(compress, compressType);
     setColLevel(compress, TSDB_COLVAL_LEVEL_DISABLED);
   } else {
-    if (check && !validColCompress(type, compressType)) return 0;
+    if (check && !validColCompress(type, compressType)) return TSDB_CODE_TSC_COMPRESS_PARAM_ERROR;
     setColCompress(compress, compressType);
 
-    if (check && !validColCompressLevel(type, level)) return 0;
+    if (check && !validColCompressLevel(type, level)) return TSDB_CODE_TSC_COMPRESS_LEVEL_ERROR;
     setColLevel(compress, level);
   }
-  return 1;
+  return TSDB_CODE_SUCCESS;
 }
 
 bool useCompress(uint8_t tableType) { return TSDB_SUPER_TABLE == tableType || TSDB_NORMAL_TABLE == tableType; }
@@ -397,10 +397,17 @@ uint32_t createDefaultColCmprByType(uint8_t type) {
   SET_COMPRESS(encode, compress, lvl, ret);
   return ret;
 }
-bool validColCmprByType(uint8_t type, uint32_t cmpr) {
+int32_t validColCmprByType(uint8_t type, uint32_t cmpr) {
   DEFINE_VAR(cmpr);
-  if (validColEncode(type, l1) && validColCompress(type, l2) && validColCompressLevel(type, lvl)) {
+  if (!validColEncode(type, l1)) {
-    return true;
+    return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
   }
-  return false;
+  if (!validColCompress(type, l2)) {
+    return TSDB_CODE_TSC_COMPRESS_PARAM_ERROR;
+  }
+
+  if (!validColCompressLevel(type, lvl)) {
+    return TSDB_CODE_TSC_COMPRESS_LEVEL_ERROR;
+  }
+  return TSDB_CODE_SUCCESS;
 }

source/common/src/tdatablock.c

@@ -193,15 +193,15 @@ static int32_t doCopyNItems(struct SColumnInfoData* pColumnInfoData, int32_t cur
   size_t   start = 1;
   int32_t  t = 0;
   int32_t  count = log(numOfRows) / log(2);
-  uint32_t startOffset = (IS_VAR_DATA_TYPE(pColumnInfoData->info.type)) ? pColumnInfoData->varmeta.length : (currentRow * itemLen);
+  uint32_t startOffset =
+      (IS_VAR_DATA_TYPE(pColumnInfoData->info.type)) ? pColumnInfoData->varmeta.length : (currentRow * itemLen);
 
   // the first item
   memcpy(pColumnInfoData->pData + startOffset, pData, itemLen);
 
   while (t < count) {
     int32_t xlen = 1 << t;
-    memcpy(pColumnInfoData->pData + start * itemLen + startOffset, 
+    memcpy(pColumnInfoData->pData + start * itemLen + startOffset, pColumnInfoData->pData + startOffset,
-           pColumnInfoData->pData + startOffset,
            xlen * itemLen);
     t += 1;
     start += xlen;
@@ -209,8 +209,7 @@ static int32_t doCopyNItems(struct SColumnInfoData* pColumnInfoData, int32_t cur
 
   // the tail part
   if (numOfRows > start) {
-    memcpy(pColumnInfoData->pData + start * itemLen + startOffset, 
+    memcpy(pColumnInfoData->pData + start * itemLen + startOffset, pColumnInfoData->pData + startOffset,
-           pColumnInfoData->pData + startOffset,
            (numOfRows - start) * itemLen);
   }
 
@@ -491,7 +490,8 @@ int32_t colDataAssign(SColumnInfoData* pColumnInfoData, const SColumnInfoData* p
   return 0;
 }
 
-int32_t colDataAssignNRows(SColumnInfoData* pDst, int32_t dstIdx, const SColumnInfoData* pSrc, int32_t srcIdx, int32_t numOfRows) {
+int32_t colDataAssignNRows(SColumnInfoData* pDst, int32_t dstIdx, const SColumnInfoData* pSrc, int32_t srcIdx,
+                           int32_t numOfRows) {
   if (pDst->info.type != pSrc->info.type || pDst->info.bytes != pSrc->info.bytes || pSrc->reassigned) {
     return TSDB_CODE_FAILED;
   }
@@ -588,14 +588,14 @@ int32_t colDataAssignNRows(SColumnInfoData* pDst, int32_t dstIdx, const SColumnI
     }
 
     if (pSrc->pData != NULL) {
-      memcpy(pDst->pData + pDst->info.bytes * dstIdx, pSrc->pData + pSrc->info.bytes * srcIdx, pDst->info.bytes * numOfRows);
+      memcpy(pDst->pData + pDst->info.bytes * dstIdx, pSrc->pData + pSrc->info.bytes * srcIdx,
+             pDst->info.bytes * numOfRows);
     }
   }
 
   return 0;
 }
 
-
 size_t blockDataGetNumOfCols(const SSDataBlock* pBlock) { return taosArrayGetSize(pBlock->pDataBlock); }
 
 size_t blockDataGetNumOfRows(const SSDataBlock* pBlock) { return pBlock->info.rows; }
@@ -742,7 +742,6 @@ void blockDataShrinkNRows(SSDataBlock* pBlock, int32_t numOfRows) {
   pBlock->info.rows -= numOfRows;
 }
 
-
 size_t blockDataGetSize(const SSDataBlock* pBlock) {
   size_t total = 0;
   size_t numOfCols = taosArrayGetSize(pBlock->pDataBlock);
@@ -827,10 +826,8 @@ SSDataBlock* blockDataExtractBlock(SSDataBlock* pBlock, int32_t startIndex, int3
     return NULL;
   }
 
-
   blockDataEnsureCapacity(pDst, rowCount);
 
-
   /* may have disorder varchar data, TODO
     for (int32_t i = 0; i < numOfCols; ++i) {
       SColumnInfoData* pColData = taosArrayGet(pBlock->pDataBlock, i);
@@ -840,7 +837,6 @@ SSDataBlock* blockDataExtractBlock(SSDataBlock* pBlock, int32_t startIndex, int3
     }
   */
 
-
   size_t numOfCols = taosArrayGetSize(pBlock->pDataBlock);
   for (int32_t i = 0; i < numOfCols; ++i) {
     SColumnInfoData* pColData = taosArrayGet(pBlock->pDataBlock, i);
@@ -1322,7 +1318,7 @@ int32_t blockDataSort(SSDataBlock* pDataBlock, SArray* pOrderInfo) {
   }
 
   terrno = 0;
-  taosqsort(index, rows, sizeof(int32_t), &helper, dataBlockCompar);
+  taosqsort_r(index, rows, sizeof(int32_t), &helper, dataBlockCompar);
   if (terrno) return terrno;
 
   int64_t p1 = taosGetTimestampUs();
@@ -1400,7 +1396,6 @@ void blockDataReset(SSDataBlock* pDataBlock) {
   pInfo->id.groupId = 0;
 }
 
-
 /*
  * NOTE: the type of the input column may be TSDB_DATA_TYPE_NULL, which is used to denote
  * the all NULL value in this column. It is an internal representation of all NULL value column, and no visible to

source/common/src/tglobal.c

@@ -1032,7 +1032,7 @@ static void taosSetServerLogCfg(SConfig *pCfg) {
   sndDebugFlag = cfgGetItem(pCfg, "sndDebugFlag")->i32;
 }
 
-static int32_t taosSetSlowLogScope(char *pScope) {
+static int32_t taosSetSlowLogScope(const char *pScope) {
   if (NULL == pScope || 0 == strlen(pScope)) {
     tsSlowLogScope = SLOW_LOG_TYPE_ALL;
     return 0;
@@ -1505,7 +1505,7 @@ static int32_t taosCfgSetOption(OptionNameAndVar *pOptions, int32_t optionSize,
   return terrno == TSDB_CODE_SUCCESS ? 0 : -1;
 }
 
-static int32_t taosCfgDynamicOptionsForServer(SConfig *pCfg, char *name) {
+static int32_t taosCfgDynamicOptionsForServer(SConfig *pCfg, const char *name) {
   terrno = TSDB_CODE_SUCCESS;
 
   if (strcasecmp(name, "resetlog") == 0) {
@@ -1583,11 +1583,12 @@ static int32_t taosCfgDynamicOptionsForServer(SConfig *pCfg, char *name) {
   return terrno == TSDB_CODE_SUCCESS ? 0 : -1;
 }
 
-static int32_t taosCfgDynamicOptionsForClient(SConfig *pCfg, char *name) {
+// todo fix race condition caused by update of config, pItem->str may be removed
+static int32_t taosCfgDynamicOptionsForClient(SConfig *pCfg, const char *name) {
   terrno = TSDB_CODE_SUCCESS;
 
   SConfigItem *pItem = cfgGetItem(pCfg, name);
-  if (!pItem || (pItem->dynScope & CFG_DYN_CLIENT) == 0) {
+  if ((pItem == NULL) || (pItem->dynScope & CFG_DYN_CLIENT) == 0) {
     uError("failed to config:%s, not support", name);
     terrno = TSDB_CODE_INVALID_CFG;
     return -1;
@@ -1598,6 +1599,7 @@ static int32_t taosCfgDynamicOptionsForClient(SConfig *pCfg, char *name) {
   int32_t len = strlen(name);
   char    lowcaseName[CFG_NAME_MAX_LEN + 1] = {0};
   strntolower(lowcaseName, name, TMIN(CFG_NAME_MAX_LEN, len));
+
   switch (lowcaseName[0]) {
     case 'd': {
       if (strcasecmp("debugFlag", name) == 0) {
@@ -1803,10 +1805,13 @@ _out:
   return terrno == TSDB_CODE_SUCCESS ? 0 : -1;
 }
 
-int32_t taosCfgDynamicOptions(SConfig *pCfg, char *name, bool forServer) {
+int32_t taosCfgDynamicOptions(SConfig *pCfg, const char *name, bool forServer) {
-  if (forServer) return taosCfgDynamicOptionsForServer(pCfg, name);
+  if (forServer) {
+    return taosCfgDynamicOptionsForServer(pCfg, name);
+  } else {
     return taosCfgDynamicOptionsForClient(pCfg, name);
   }
+}
 
 void taosSetDebugFlag(int32_t *pFlagPtr, const char *flagName, int32_t flagVal) {
   SConfigItem *pItem = cfgGetItem(tsCfg, flagName);

source/common/src/tmisce.c

@@ -17,6 +17,8 @@
 #include "tmisce.h"
 #include "tglobal.h"
 #include "tjson.h"
+#include "tdatablock.h"
+
 int32_t taosGetFqdnPortFromEp(const char* ep, SEp* pEp) {
   pEp->port = 0;
   memset(pEp->fqdn, 0, TSDB_FQDN_LEN);
@@ -97,10 +99,10 @@ void epsetSort(SEpSet* pDst) {
       SEp* s = &pDst->eps[j + 1];
       int  cmp = strncmp(f->fqdn, s->fqdn, sizeof(f->fqdn));
       if (cmp > 0 || (cmp == 0 && f->port > s->port)) {
-        SEp ep = {0};
+        SEp ep1 = {0};
-        epAssign(&ep, f);
+        epAssign(&ep1, f);
         epAssign(f, s);
-        epAssign(s, &ep);
+        epAssign(s, &ep1);
       }
     }
   }
@@ -216,3 +218,43 @@ int32_t taosGenCrashJsonMsg(int signum, char** pMsg, int64_t clusterId, int64_t
 
   return TSDB_CODE_SUCCESS;
 }
+
+int32_t dumpConfToDataBlock(SSDataBlock* pBlock, int32_t startCol) {
+  SConfig*     pConf = taosGetCfg();
+  int32_t      numOfRows = 0;
+  int32_t      col = startCol;
+  SConfigItem* pItem = NULL;
+
+  blockDataEnsureCapacity(pBlock, cfgGetSize(pConf));
+  SConfigIter* pIter = cfgCreateIter(pConf);
+
+  while ((pItem = cfgNextIter(pIter)) != NULL) {
+    col = startCol;
+
+    // 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, col++);
+    colDataSetVal(pColInfo, numOfRows, name, false);
+
+    char    value[TSDB_CONFIG_VALUE_LEN + VARSTR_HEADER_SIZE] = {0};
+    int32_t valueLen = 0;
+    cfgDumpItemValue(pItem, &value[VARSTR_HEADER_SIZE], TSDB_CONFIG_VALUE_LEN, &valueLen);
+    varDataSetLen(value, valueLen);
+    pColInfo = taosArrayGet(pBlock->pDataBlock, col++);
+    colDataSetVal(pColInfo, numOfRows, 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, col++);
+    colDataSetVal(pColInfo, numOfRows, scope, false);
+
+    numOfRows++;
+  }
+
+  pBlock->info.rows = numOfRows;
+
+  cfgDestroyIter(pIter);
+  return TSDB_CODE_SUCCESS;
+}

source/common/src/tmsg.c

@@ -9624,6 +9624,8 @@ void tDestroySubmitTbData(SSubmitTbData *pTbData, int32_t flag) {
       taosArrayDestroy(pTbData->aRowP);
     }
   }
+
+  pTbData->aRowP = NULL;
 }
 
 void tDestroySubmitReq(SSubmitReq2 *pReq, int32_t flag) {

source/dnode/mgmt/mgmt_dnode/src/dmHandle.c

@@ -333,39 +333,10 @@ SSDataBlock *dmBuildVariablesBlock(void) {
 }
 
 int32_t dmAppendVariablesToBlock(SSDataBlock *pBlock, int32_t dnodeId) {
-  int32_t numOfCfg = taosArrayGetSize(tsCfg->array);
+  /*int32_t code = */dumpConfToDataBlock(pBlock, 1);
-  int32_t numOfRows = 0;
-  blockDataEnsureCapacity(pBlock, numOfCfg);
 
-  for (int32_t i = 0, c = 0; i < numOfCfg; ++i, c = 0) {
+  SColumnInfoData *pColInfo = taosArrayGet(pBlock->pDataBlock, 0);
-    SConfigItem *pItem = taosArrayGet(tsCfg->array, i);
+  colDataSetNItems(pColInfo, 0, (const char *)&dnodeId, pBlock->info.rows, false);
-    // GRANT_CFG_SKIP;
-
-    SColumnInfoData *pColInfo = taosArrayGet(pBlock->pDataBlock, c++);
-    colDataSetVal(pColInfo, i, (const char *)&dnodeId, false);
-
-    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);
-    pColInfo = taosArrayGet(pBlock->pDataBlock, c++);
-    colDataSetVal(pColInfo, i, name, false);
-
-    char    value[TSDB_CONFIG_VALUE_LEN + VARSTR_HEADER_SIZE] = {0};
-    int32_t valueLen = 0;
-    cfgDumpItemValue(pItem, &value[VARSTR_HEADER_SIZE], TSDB_CONFIG_VALUE_LEN, &valueLen);
-    varDataSetLen(value, valueLen);
-    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++;
-  }
-
-  pBlock->info.rows = numOfRows;
 
   return TSDB_CODE_SUCCESS;
 }

source/dnode/mnode/impl/inc/mndDef.h

@@ -166,7 +166,7 @@ typedef struct {
   int32_t       failedTimes;
   void*         rpcRsp;
   int32_t       rpcRspLen;
-  int32_t       redoActionPos;
+  int32_t       actionPos;
   SArray*       prepareActions;
   SArray*       redoActions;
   SArray*       undoActions;

source/dnode/mnode/impl/src/mndDb.c

@@ -860,11 +860,6 @@ static int32_t mndProcessCreateDbReq(SRpcMsg *pReq) {
   SUserObj    *pUser = NULL;
   SCreateDbReq createReq = {0};
 
-  if ((terrno = grantCheck(TSDB_GRANT_DB)) != 0) {
-    code = terrno;
-    goto _OVER;
-  }
-
   if (tDeserializeSCreateDbReq(pReq->pCont, pReq->contLen, &createReq) != 0) {
     terrno = TSDB_CODE_INVALID_MSG;
     goto _OVER;
@@ -903,6 +898,11 @@ static int32_t mndProcessCreateDbReq(SRpcMsg *pReq) {
     }
   }
 
+  if ((terrno = grantCheck(TSDB_GRANT_DB)) != 0) {
+    code = terrno;
+    goto _OVER;
+  }
+
   if ((code = mndCheckDbEncryptKey(pMnode, &createReq)) != 0) {
     terrno = code;
     goto _OVER;

source/dnode/mnode/impl/src/mndStb.c

@@ -1006,7 +1006,8 @@ static int32_t mndProcessTtlTimer(SRpcMsg *pReq) {
     pHead->vgId = htonl(pVgroup->vgId);
     tSerializeSVDropTtlTableReq((char *)pHead + sizeof(SMsgHead), reqLen, &ttlReq);
 
-    SRpcMsg rpcMsg = {.msgType = TDMT_VND_FETCH_TTL_EXPIRED_TBS, .pCont = pHead, .contLen = contLen, .info = pReq->info};
+    SRpcMsg rpcMsg = {
+        .msgType = TDMT_VND_FETCH_TTL_EXPIRED_TBS, .pCont = pHead, .contLen = contLen, .info = pReq->info};
     SEpSet  epSet = mndGetVgroupEpset(pMnode, pVgroup);
     int32_t code = tmsgSendReq(&epSet, &rpcMsg);
     if (code != 0) {
@@ -1752,9 +1753,10 @@ static int32_t mndUpdateSuperTableColumnCompress(SMnode *pMnode, const SStbObj *
   if (mndAllocStbSchemas(pOld, pNew) != 0) {
     return -1;
   }
-  if (!validColCmprByType(pTarget->type, p->bytes)) {
+  code = validColCmprByType(pTarget->type, p->bytes);
-    terrno = TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+  if (code != TSDB_CODE_SUCCESS) {
-    return -1;
+    terrno = code;
+    return code;
   }
 
   int8_t updated = 0;
@@ -3902,7 +3904,8 @@ typedef struct SMndDropTbsWithTsmaCtx {
   SArray   *pResTbNames;  // SArray<char*>
 } SMndDropTbsWithTsmaCtx;
 
-static int32_t mndDropTbAddTsmaResTbsForSingleVg(SMnode* pMnode, SMndDropTbsWithTsmaCtx* pCtx, SArray* pTbs, int32_t vgId);
+static int32_t mndDropTbAddTsmaResTbsForSingleVg(SMnode *pMnode, SMndDropTbsWithTsmaCtx *pCtx, SArray *pTbs,
+                                                 int32_t vgId);
 
 static void mndDestroyDropTbsWithTsmaCtx(SMndDropTbsWithTsmaCtx *p) {
   if (!p) return;
@@ -3969,8 +3972,8 @@ _end:
   return code;
 }
 
-
+static void *mndBuildVDropTbsReq(SMnode *pMnode, const SVgroupInfo *pVgInfo, const SVDropTbBatchReq *pReq,
-static void* mndBuildVDropTbsReq(SMnode* pMnode, const SVgroupInfo* pVgInfo, const SVDropTbBatchReq* pReq, int32_t *len) {
+                                 int32_t *len) {
   int32_t   contLen = 0;
   int32_t   ret = 0;
   SMsgHead *pHead = NULL;
@@ -3999,7 +4002,8 @@ static void* mndBuildVDropTbsReq(SMnode* pMnode, const SVgroupInfo* pVgInfo, con
   return pHead;
 }
 
-static int32_t mndSetDropTbsRedoActions(SMnode* pMnode, STrans* pTrans, const SVDropTbVgReqs* pVgReqs, void* pCont, int32_t contLen) {
+static int32_t mndSetDropTbsRedoActions(SMnode *pMnode, STrans *pTrans, const SVDropTbVgReqs *pVgReqs, void *pCont,
+                                        int32_t contLen) {
   STransAction action = {0};
   action.epSet = pVgReqs->info.epSet;
   action.pCont = pCont;
@@ -4055,8 +4059,7 @@ static int32_t mndProcessDropTbWithTsma(SRpcMsg* pReq) {
     terrno = mndDropTbAddTsmaResTbsForSingleVg(pMnode, pCtx, pReq->pTbs, pReq->vgInfo.vgId);
     if (terrno) goto _OVER;
   }
-  if (mndCreateDropTbsTxnPrepare(pReq, pCtx) == 0)
+  if (mndCreateDropTbsTxnPrepare(pReq, pCtx) == 0) code = 0;
-    code = 0;
 _OVER:
   tFreeSMDropTbsReq(&dropReq);
   if (pCtx) mndDestroyDropTbsWithTsmaCtx(pCtx);
@@ -4121,7 +4124,8 @@ int32_t vgHashValCmp(const void* lp, const void* rp) {
   return 0;
 }
 
-static int32_t mndDropTbAddTsmaResTbsForSingleVg(SMnode* pMnode, SMndDropTbsWithTsmaCtx* pCtx, SArray* pTbs, int32_t vgId) {
+static int32_t mndDropTbAddTsmaResTbsForSingleVg(SMnode *pMnode, SMndDropTbsWithTsmaCtx *pCtx, SArray *pTbs,
+                                                 int32_t vgId) {
   int32_t code = 0;
 
   SVgObj *pVgObj = mndAcquireVgroup(pMnode, vgId);
@@ -4129,7 +4133,10 @@ static int32_t mndDropTbAddTsmaResTbsForSingleVg(SMnode* pMnode, SMndDropTbsWith
     code = 0;
     goto _end;
   }
-  SVgroupInfo vgInfo = {.hashBegin = pVgObj->hashBegin, .hashEnd = pVgObj->hashEnd, .numOfTable = pVgObj->numOfTables, .vgId = pVgObj->vgId};
+  SVgroupInfo vgInfo = {.hashBegin = pVgObj->hashBegin,
+                        .hashEnd = pVgObj->hashEnd,
+                        .numOfTable = pVgObj->numOfTables,
+                        .vgId = pVgObj->vgId};
   vgInfo.epSet = mndGetVgroupEpset(pMnode, pVgObj);
   mndReleaseVgroup(pMnode, pVgObj);
 
@@ -4137,7 +4144,6 @@ static int32_t mndDropTbAddTsmaResTbsForSingleVg(SMnode* pMnode, SMndDropTbsWith
   for (int32_t i = 0; i < pTbs->size; ++i) {
     const SVDropTbReq *pTb = taosArrayGet(pTbs, i);
     if (taosHashGet(pCtx->pTsmaMap, &pTb->suid, sizeof(pTb->suid))) {
-
     } else {
       SMDropTbTsmaInfos infos = {0};
       infos.pTsmaInfos = taosArrayInit(2, sizeof(SMDropTbTsmaInfo));
@@ -4225,8 +4231,7 @@ static int32_t mndProcessFetchTtlExpiredTbs(SRpcMsg *pRsp) {
 
   terrno = mndDropTbAddTsmaResTbsForSingleVg(pMnode, pCtx, rsp.pExpiredTbs, rsp.vgId);
   if (terrno) goto _end;
-  if (mndCreateDropTbsTxnPrepare(pRsp, pCtx) == 0)
+  if (mndCreateDropTbsTxnPrepare(pRsp, pCtx) == 0) code = 0;
-    code = 0;
 _end:
   if (pCtx) mndDestroyDropTbsWithTsmaCtx(pCtx);
   tDecoderClear(&decoder);

source/dnode/mnode/impl/src/mndStream.c

@@ -325,7 +325,7 @@ static int32_t createSchemaByFields(const SArray* pFields, SSchemaWrapper* pWrap
   return TSDB_CODE_SUCCESS;
 }
 
-static bool hasPrimaryKey(SSchemaWrapper* pWrapper) {
+static bool hasDestPrimaryKey(SSchemaWrapper* pWrapper) {
   if (pWrapper->nCols < 2) {
     return false;
   }
@@ -442,7 +442,7 @@ static int32_t mndBuildStreamObjFromCreateReq(SMnode *pMnode, SStreamObj *pObj,
     pObj->outputSchema.pSchema = pFullSchema;
   }
 
-  bool hasKey = hasPrimaryKey(&pObj->outputSchema);
+  bool hasKey = hasDestPrimaryKey(&pObj->outputSchema);
   SPlanContext cxt = {
       .pAstRoot = pAst,
       .topicQuery = false,
@@ -699,10 +699,6 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
   int32_t     sqlLen = 0;
   terrno = TSDB_CODE_SUCCESS;
 
-  if ((terrno = grantCheck(TSDB_GRANT_STREAMS)) < 0) {
-    return terrno;
-  }
-
   SCMCreateStreamReq createReq = {0};
   if (tDeserializeSCMCreateStreamReq(pReq->pCont, pReq->contLen, &createReq) != 0) {
     terrno = TSDB_CODE_INVALID_MSG;
@@ -733,6 +729,10 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
+  if ((terrno = grantCheck(TSDB_GRANT_STREAMS)) < 0) {
+    goto _OVER;
+  }
+
   if (createReq.sql != NULL) {
     sqlLen = strlen(createReq.sql);
     sql = taosMemoryMalloc(sqlLen + 1);

source/dnode/mnode/impl/src/mndStreamTrans.c

@@ -98,7 +98,8 @@ bool mndStreamTransConflictCheck(SMnode* pMnode, int64_t streamId, const char* p
         mDebug("not conflict with checkpoint trans, name:%s, continue create trans", pTransName);
       }
     } else if ((strcmp(tInfo.name, MND_STREAM_CREATE_NAME) == 0) || (strcmp(tInfo.name, MND_STREAM_DROP_NAME) == 0) ||
-               (strcmp(tInfo.name, MND_STREAM_TASK_RESET_NAME) == 0)) {
+               (strcmp(tInfo.name, MND_STREAM_TASK_RESET_NAME) == 0) ||
+               strcmp(tInfo.name, MND_STREAM_TASK_UPDATE_NAME) == 0) {
       mWarn("conflict with other transId:%d streamUid:0x%" PRIx64 ", trans:%s", tInfo.transId, tInfo.streamId,
             tInfo.name);
       terrno = TSDB_CODE_MND_TRANS_CONFLICT;

source/dnode/mnode/impl/src/mndSubscribe.c

@@ -438,10 +438,10 @@ static void processSubOffsetRows(SMnode *pMnode, const SMqRebInputObj *pInput, S
 }
 
 static void printRebalanceLog(SMqRebOutputObj *pOutput){
-  mInfo("sub:%s mq re-balance calculation completed, re-balanced vg", pOutput->pSub->key);
+  mInfo("sub:%s mq rebalance calculation completed, re-balanced vg", pOutput->pSub->key);
   for (int32_t i = 0; i < taosArrayGetSize(pOutput->rebVgs); i++) {
     SMqRebOutputVg *pOutputRebVg = taosArrayGet(pOutput->rebVgs, i);
-    mInfo("sub:%s mq re-balance vgId:%d, moved from consumer:0x%" PRIx64 ", to consumer:0x%" PRIx64, pOutput->pSub->key,
+    mInfo("sub:%s mq rebalance vgId:%d, moved from consumer:0x%" PRIx64 ", to consumer:0x%" PRIx64, pOutput->pSub->key,
           pOutputRebVg->pVgEp->vgId, pOutputRebVg->oldConsumerId, pOutputRebVg->newConsumerId);
   }
 
@@ -451,10 +451,10 @@ static void printRebalanceLog(SMqRebOutputObj *pOutput){
     if (pIter == NULL) break;
     SMqConsumerEp *pConsumerEp = (SMqConsumerEp *)pIter;
     int32_t        sz = taosArrayGetSize(pConsumerEp->vgs);
-    mInfo("sub:%s mq re-balance final cfg: consumer:0x%" PRIx64 " has %d vg", pOutput->pSub->key, pConsumerEp->consumerId, sz);
+    mInfo("sub:%s mq rebalance final cfg: consumer:0x%" PRIx64 " has %d vg", pOutput->pSub->key, pConsumerEp->consumerId, sz);
     for (int32_t i = 0; i < sz; i++) {
       SMqVgEp *pVgEp = taosArrayGetP(pConsumerEp->vgs, i);
-      mInfo("sub:%s mq re-balance final cfg: vg %d to consumer:0x%" PRIx64, pOutput->pSub->key, pVgEp->vgId,
+      mInfo("sub:%s mq rebalance final cfg: vg %d to consumer:0x%" PRIx64, pOutput->pSub->key, pVgEp->vgId,
             pConsumerEp->consumerId);
     }
   }
@@ -762,18 +762,18 @@ static void mndCheckConsumer(SRpcMsg *pMsg, SHashObj* rebSubHash) {
 
 bool mndRebTryStart() {
   int32_t old = atomic_val_compare_exchange_32(&mqRebInExecCnt, 0, 1);
-  mInfo("rebalance counter old val:%d", old);
+  if (old > 0) mInfo("[rebalance] counter old val:%d", old)
   return old == 0;
 }
 
 void mndRebCntInc() {
   int32_t val = atomic_add_fetch_32(&mqRebInExecCnt, 1);
-  mInfo("rebalance cnt inc, value:%d", val);
+  if (val > 0) mInfo("[rebalance] cnt inc, value:%d", val)
 }
 
 void mndRebCntDec() {
   int32_t val = atomic_sub_fetch_32(&mqRebInExecCnt, 1);
-  mInfo("rebalance cnt sub, value:%d", val);
+  if (val > 0) mInfo("[rebalance] cnt sub, value:%d", val)
 }
 
 static void clearRebOutput(SMqRebOutputObj *rebOutput){
@@ -848,10 +848,10 @@ static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg) {
   int                code = 0;
   void              *pIter = NULL;
   SMnode            *pMnode = pMsg->info.node;
-  mInfo("[rebalance] start to process mq timer");
+  mDebug("[rebalance] start to process mq timer")
 
   if (!mndRebTryStart()) {
-    mInfo("[rebalance] mq rebalance already in progress, do nothing");
+    mInfo("[rebalance] mq rebalance already in progress, do nothing")
     return code;
   }
 
@@ -863,7 +863,9 @@ static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg) {
   taosHashSetFreeFp(rebSubHash, freeRebalanceItem);
 
   mndCheckConsumer(pMsg, rebSubHash);
-  mInfo("[rebalance] mq re-balance start, total required re-balanced trans:%d", taosHashGetSize(rebSubHash));
+  if (taosHashGetSize(rebSubHash) > 0) {
+    mInfo("[rebalance] mq rebalance start, total required re-balanced trans:%d", taosHashGetSize(rebSubHash))
+  }
 
   while (1) {
     pIter = taosHashIterate(rebSubHash, pIter);
@@ -887,13 +889,15 @@ static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg) {
     mndDoRebalance(pMnode, &rebInput, &rebOutput);
 
     if (mndPersistRebResult(pMnode, pMsg, &rebOutput) != 0) {
-      mError("mq re-balance persist output error, possibly vnode splitted or dropped,msg:%s", terrstr());
+      mError("mq rebalance persist output error, possibly vnode splitted or dropped,msg:%s", terrstr())
     }
 
     clearRebOutput(&rebOutput);
   }
 
-  mInfo("[rebalance] mq re-balance completed successfully, wait trans finish");
+  if (taosHashGetSize(rebSubHash) > 0) {
+    mInfo("[rebalance] mq rebalance completed successfully, wait trans finish")
+  }
 
 END:
   taosHashCancelIterate(rebSubHash, pIter);

source/dnode/mnode/impl/src/mndTopic.c

@@ -561,15 +561,6 @@ static int32_t mndProcessCreateTopicReq(SRpcMsg *pReq) {
   SMqTopicObj      *pTopic = NULL;
   SDbObj           *pDb = NULL;
   SCMCreateTopicReq createTopicReq = {0};
-  if (sdbGetSize(pMnode->pSdb, SDB_TOPIC) >= tmqMaxTopicNum){
-    terrno = TSDB_CODE_TMQ_TOPIC_OUT_OF_RANGE;
-    mError("topic num out of range");
-    return code;
-  }
-
-  if ((terrno = grantCheck(TSDB_GRANT_SUBSCRIPTION)) < 0) {
-    return code;
-  }
 
   if (tDeserializeSCMCreateTopicReq(pReq->pCont, pReq->contLen, &createTopicReq) != 0) {
     terrno = TSDB_CODE_INVALID_MSG;
@@ -609,6 +600,16 @@ static int32_t mndProcessCreateTopicReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
+  if (sdbGetSize(pMnode->pSdb, SDB_TOPIC) >= tmqMaxTopicNum){
+    terrno = TSDB_CODE_TMQ_TOPIC_OUT_OF_RANGE;
+    mError("topic num out of range");
+    goto _OVER;
+  }
+
+  if ((terrno = grantCheck(TSDB_GRANT_SUBSCRIPTION)) < 0) {
+    goto _OVER;
+  }
+
   code = mndCreateTopic(pMnode, pReq, &createTopicReq, pDb, pReq->info.conn.user);
   if (code == 0) {
     code = TSDB_CODE_ACTION_IN_PROGRESS;

source/dnode/mnode/impl/src/mndTrans.c

@@ -169,7 +169,7 @@ SSdbRaw *mndTransEncode(STrans *pTrans) {
   SDB_SET_INT64(pRaw, dataPos, pTrans->createdTime, _OVER)
   SDB_SET_BINARY(pRaw, dataPos, pTrans->dbname, TSDB_TABLE_FNAME_LEN, _OVER)
   SDB_SET_BINARY(pRaw, dataPos, pTrans->stbname, TSDB_TABLE_FNAME_LEN, _OVER)
-  SDB_SET_INT32(pRaw, dataPos, pTrans->redoActionPos, _OVER)
+  SDB_SET_INT32(pRaw, dataPos, pTrans->actionPos, _OVER)
 
   int32_t prepareActionNum = taosArrayGetSize(pTrans->prepareActions);
   int32_t redoActionNum = taosArrayGetSize(pTrans->redoActions);
@@ -317,7 +317,7 @@ SSdbRow *mndTransDecode(SSdbRaw *pRaw) {
   SDB_GET_INT64(pRaw, dataPos, &pTrans->createdTime, _OVER)
   SDB_GET_BINARY(pRaw, dataPos, pTrans->dbname, TSDB_TABLE_FNAME_LEN, _OVER)
   SDB_GET_BINARY(pRaw, dataPos, pTrans->stbname, TSDB_TABLE_FNAME_LEN, _OVER)
-  SDB_GET_INT32(pRaw, dataPos, &pTrans->redoActionPos, _OVER)
+  SDB_GET_INT32(pRaw, dataPos, &pTrans->actionPos, _OVER)
 
   if (sver > TRANS_VER1_NUMBER) {
     SDB_GET_INT32(pRaw, dataPos, &prepareActionNum, _OVER)
@@ -525,7 +525,7 @@ static int32_t mndTransActionUpdate(SSdb *pSdb, STrans *pOld, STrans *pNew) {
   mndTransUpdateActions(pOld->undoActions, pNew->undoActions);
   mndTransUpdateActions(pOld->commitActions, pNew->commitActions);
   pOld->stage = pNew->stage;
-  pOld->redoActionPos = pNew->redoActionPos;
+  pOld->actionPos = pNew->actionPos;
 
   if (pOld->stage == TRN_STAGE_COMMIT) {
     pOld->stage = TRN_STAGE_COMMIT_ACTION;
@@ -1360,22 +1360,19 @@ static int32_t mndTransExecuteCommitActions(SMnode *pMnode, STrans *pTrans, bool
   return code;
 }
 
-static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans, bool topHalf) {
+static int32_t mndTransExecuteActionsSerial(SMnode *pMnode, STrans *pTrans, SArray *pActions, bool topHalf) {
   int32_t code = 0;
-  int32_t numOfActions = taosArrayGetSize(pTrans->redoActions);
+  int32_t numOfActions = taosArrayGetSize(pActions);
   if (numOfActions == 0) return code;
 
-  taosThreadMutexLock(&pTrans->mutex);
+  if (pTrans->actionPos >= numOfActions) {
-
-  if (pTrans->redoActionPos >= numOfActions) {
-    taosThreadMutexUnlock(&pTrans->mutex);
     return code;
   }
 
-  mInfo("trans:%d, execute %d actions serial, current redoAction:%d", pTrans->id, numOfActions, pTrans->redoActionPos);
+  mInfo("trans:%d, execute %d actions serial, current redoAction:%d", pTrans->id, numOfActions, pTrans->actionPos);
 
-  for (int32_t action = pTrans->redoActionPos; action < numOfActions; ++action) {
+  for (int32_t action = pTrans->actionPos; action < numOfActions; ++action) {
-    STransAction *pAction = taosArrayGet(pTrans->redoActions, pTrans->redoActionPos);
+    STransAction *pAction = taosArrayGet(pActions, pTrans->actionPos);
 
     code = mndTransExecSingleAction(pMnode, pTrans, pAction, topHalf);
     if (code == 0) {
@@ -1409,14 +1406,14 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans,
 
     if (code == 0) {
       pTrans->code = 0;
-      pTrans->redoActionPos++;
+      pTrans->actionPos++;
       mInfo("trans:%d, %s:%d is executed and need sync to other mnodes", pTrans->id, mndTransStr(pAction->stage),
             pAction->id);
       taosThreadMutexUnlock(&pTrans->mutex);
       code = mndTransSync(pMnode, pTrans);
       taosThreadMutexLock(&pTrans->mutex);
       if (code != 0) {
-        pTrans->redoActionPos--;
+        pTrans->actionPos--;
         pTrans->code = terrno;
         mError("trans:%d, %s:%d is executed and failed to sync to other mnodes since %s", pTrans->id,
                mndTransStr(pAction->stage), pAction->id, terrstr());
@@ -1442,8 +1439,26 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans,
     }
   }
 
-  taosThreadMutexUnlock(&pTrans->mutex);
+  return code;
+}
 
+static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans, bool topHalf) {
+  int32_t code = TSDB_CODE_ACTION_IN_PROGRESS;
+  taosThreadMutexLock(&pTrans->mutex);
+  if (pTrans->stage == TRN_STAGE_REDO_ACTION) {
+    code = mndTransExecuteActionsSerial(pMnode, pTrans, pTrans->redoActions, topHalf);
+  }
+  taosThreadMutexUnlock(&pTrans->mutex);
+  return code;
+}
+
+static int32_t mndTransExecuteUndoActionsSerial(SMnode *pMnode, STrans *pTrans, bool topHalf) {
+  int32_t code = TSDB_CODE_ACTION_IN_PROGRESS;
+  taosThreadMutexLock(&pTrans->mutex);
+  if (pTrans->stage == TRN_STAGE_UNDO_ACTION) {
+    code = mndTransExecuteActionsSerial(pMnode, pTrans, pTrans->undoActions, topHalf);
+  }
+  taosThreadMutexUnlock(&pTrans->mutex);
   return code;
 }
 
@@ -1563,13 +1578,22 @@ static bool mndTransPerformCommitActionStage(SMnode *pMnode, STrans *pTrans, boo
 
 static bool mndTransPerformUndoActionStage(SMnode *pMnode, STrans *pTrans, bool topHalf) {
   bool    continueExec = true;
-  int32_t code = mndTransExecuteUndoActions(pMnode, pTrans, topHalf);
+  int32_t code = 0;
+
+  if (pTrans->exec == TRN_EXEC_SERIAL) {
+    code = mndTransExecuteUndoActionsSerial(pMnode, pTrans, topHalf);
+  } else {
+    code = mndTransExecuteUndoActions(pMnode, pTrans, topHalf);
+  }
+
+  if (mndCannotExecuteTransAction(pMnode, topHalf)) return false;
+  terrno = code;
 
   if (code == 0) {
     pTrans->stage = TRN_STAGE_PRE_FINISH;
     mInfo("trans:%d, stage from undoAction to pre-finish", pTrans->id);
     continueExec = true;
-  } else if (code == TSDB_CODE_ACTION_IN_PROGRESS) {
+  } else if (code == TSDB_CODE_ACTION_IN_PROGRESS || code == TSDB_CODE_MND_TRANS_CTX_SWITCH) {
     mInfo("trans:%d, stage keep on undoAction since %s", pTrans->id, tstrerror(code));
     continueExec = false;
   } else {

source/dnode/vnode/src/meta/metaTable.c

@@ -887,6 +887,8 @@ int metaCreateTable(SMeta *pMeta, int64_t ver, SVCreateTbReq *pReq, STableMetaRs
 
   bool sysTbl = (pReq->type == TSDB_CHILD_TABLE) && metaTbInFilterCache(pMeta, pReq->ctb.stbName, 1);
 
+  if (!sysTbl && ((terrno = grantCheck(TSDB_GRANT_TIMESERIES)) < 0)) goto _err;
+
   // build SMetaEntry
   SVnodeStats *pStats = &pMeta->pVnode->config.vndStats;
   me.version = ver;
@@ -2659,6 +2661,8 @@ int32_t metaGetColCmpr(SMeta *pMeta, tb_uid_t uid, SHashObj **ppColCmprObj) {
   SMetaEntry e = {0};
   SDecoder   dc = {0};
 
+  *ppColCmprObj = NULL;
+
   metaRLock(pMeta);
   rc = tdbTbGet(pMeta->pUidIdx, &uid, sizeof(uid), &pData, &nData);
   if (rc < 0) {

source/dnode/vnode/src/tq/tq.c

@@ -367,7 +367,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
       } while (0);
     }
 
-    // 2. check re-balance status
+    // 2. check rebalance status
     if (pHandle->consumerId != consumerId) {
       tqError("ERROR tmq poll: consumer:0x%" PRIx64
               " vgId:%d, subkey %s, mismatch for saved handle consumer:0x%" PRIx64,
@@ -485,7 +485,7 @@ int32_t tqProcessVgWalInfoReq(STQ* pTq, SRpcMsg* pMsg) {
     return -1;
   }
 
-  // 2. check re-balance status
+  // 2. check rebalance status
   if (pHandle->consumerId != consumerId) {
     tqDebug("ERROR consumer:0x%" PRIx64 " vgId:%d, subkey %s, mismatch for saved handle consumer:0x%" PRIx64,
             consumerId, vgId, req.subKey, pHandle->consumerId);
@@ -666,7 +666,7 @@ int32_t tqProcessSubscribeReq(STQ* pTq, int64_t sversion, char* msg, int32_t msg
               req.vgId, req.subKey, req.newConsumerId, req.oldConsumerId);
     }
     if (req.newConsumerId == -1) {
-      tqError("vgId:%d, tq invalid re-balance request, new consumerId %" PRId64 "", req.vgId, req.newConsumerId);
+      tqError("vgId:%d, tq invalid rebalance request, new consumerId %" PRId64 "", req.vgId, req.newConsumerId);
       goto end;
     }
     STqHandle handle = {0};

source/dnode/vnode/src/tq/tqSink.c

@@ -583,7 +583,7 @@ int32_t doConvertRows(SSubmitTbData* pTableData, const STSchema* pTSchema, SSDat
 
       if (IS_SET_NULL(pCol)) {
         if (pCol->flags & COL_IS_KEY) {
-          qError("ts:%" PRId64 " Primary key column should not be null, colId:%" PRIi16 ", colType:%" PRIi8, ts,
+          qError("ts:%" PRId64 " primary key column should not be null, colId:%" PRIi16 ", colType:%" PRIi8, ts,
                  pCol->colId, pCol->type);
           break;
         }
@@ -593,7 +593,7 @@ int32_t doConvertRows(SSubmitTbData* pTableData, const STSchema* pTSchema, SSDat
         SColumnInfoData* pColData = taosArrayGet(pDataBlock->pDataBlock, dataIndex);
         if (colDataIsNull_s(pColData, j)) {
           if (pCol->flags & COL_IS_KEY) {
-            qError("ts:%" PRId64 "Primary key column should not be null, colId:%" PRIi16 ", colType:%" PRIi8,
+            qError("ts:%" PRId64 " primary key column should not be null, colId:%" PRIi16 ", colType:%" PRIi8,
                    ts, pCol->colId, pCol->type);
             break;
           }
@@ -624,8 +624,8 @@ int32_t doConvertRows(SSubmitTbData* pTableData, const STSchema* pTSchema, SSDat
     code = tRowBuild(pVals, (STSchema*)pTSchema, &pRow);
     if (code != TSDB_CODE_SUCCESS) {
       tDestroySubmitTbData(pTableData, TSDB_MSG_FLG_ENCODE);
-      pTableData->aRowP = taosArrayDestroy(pTableData->aRowP);
       taosArrayDestroy(pVals);
+      tqError("s-task:%s build rows for submit failed, ts:%"PRId64, id, ts);
       return code;
     }
 

source/dnode/vnode/src/tqCommon/tqCommon.c

@@ -195,12 +195,15 @@ int32_t tqStreamTaskProcessUpdateReq(SStreamMeta* pMeta, SMsgCb* cb, SRpcMsg* pM
   const char*  idstr = pTask->id.idStr;
 
   if (pMeta->updateInfo.transId != req.transId) {
-    pMeta->updateInfo.transId = req.transId;
+    ASSERT(req.transId > pMeta->updateInfo.transId);
-    tqInfo("s-task:%s receive new trans to update nodeEp msg from mnode, transId:%d", idstr, req.transId);
+    tqInfo("s-task:%s vgId:%d receive new trans to update nodeEp msg from mnode, transId:%d, prev transId:%d", idstr,
+           vgId, req.transId, pMeta->updateInfo.transId);
+
     // info needs to be kept till the new trans to update the nodeEp arrived.
     taosHashClear(pMeta->updateInfo.pTasks);
+    pMeta->updateInfo.transId = req.transId;
   } else {
-    tqDebug("s-task:%s recv trans to update nodeEp from mnode, transId:%d", idstr, req.transId);
+    tqDebug("s-task:%s vgId:%d recv trans to update nodeEp from mnode, transId:%d", idstr, vgId, req.transId);
   }
 
   // duplicate update epset msg received, discard this redundant message
@@ -208,7 +211,7 @@ int32_t tqStreamTaskProcessUpdateReq(SStreamMeta* pMeta, SMsgCb* cb, SRpcMsg* pM
 
   void* pReqTask = taosHashGet(pMeta->updateInfo.pTasks, &entry, sizeof(STaskUpdateEntry));
   if (pReqTask != NULL) {
-    tqDebug("s-task:%s (vgId:%d) already update in trans:%d, discard the nodeEp update msg", idstr, vgId, req.transId);
+    tqDebug("s-task:%s (vgId:%d) already update in transId:%d, discard the nodeEp update msg", idstr, vgId, req.transId);
     rsp.code = TSDB_CODE_SUCCESS;
     streamMetaWUnLock(pMeta);
     taosArrayDestroy(req.pNodeList);
@@ -232,7 +235,7 @@ int32_t tqStreamTaskProcessUpdateReq(SStreamMeta* pMeta, SMsgCb* cb, SRpcMsg* pM
     } else {
       tqDebug("s-task:%s fill-history task update nodeEp along with stream task", (*ppHTask)->id.idStr);
       bool updateEpSet = streamTaskUpdateEpsetInfo(*ppHTask, req.pNodeList);
-      if (!updated) {
+      if (updateEpSet) {
         updated = updateEpSet;
       }
 
@@ -242,14 +245,15 @@ int32_t tqStreamTaskProcessUpdateReq(SStreamMeta* pMeta, SMsgCb* cb, SRpcMsg* pM
   }
 
   if (updated) {
-    tqDebug("s-task:%s vgId:%d save task after update epset", idstr, vgId);
+    tqDebug("s-task:%s vgId:%d save task after update epset, and stop task", idstr, vgId);
     streamMetaSaveTask(pMeta, pTask);
     if (ppHTask != NULL) {
       streamMetaSaveTask(pMeta, *ppHTask);
     }
+  } else {
+    tqDebug("s-task:%s vgId:%d not save task since not update epset actually, stop task", idstr, vgId);
   }
 
-  tqDebug("s-task:%s vgId:%d start to stop task after save task", idstr, vgId);
   streamTaskStop(pTask);
 
   // keep the already updated info
@@ -989,7 +993,7 @@ int32_t tqStreamTaskProcessTaskPauseReq(SStreamMeta* pMeta, char* pMsg) {
   }
 
   tqDebug("s-task:%s receive pause msg from mnode", pTask->id.idStr);
-  streamTaskPause(pMeta, pTask);
+  streamTaskPause(pTask);
 
   SStreamTask* pHistoryTask = NULL;
   if (HAS_RELATED_FILLHISTORY_TASK(pTask)) {
@@ -1006,7 +1010,7 @@ int32_t tqStreamTaskProcessTaskPauseReq(SStreamMeta* pMeta, char* pMsg) {
 
     tqDebug("s-task:%s fill-history task handle paused along with related stream task", pHistoryTask->id.idStr);
 
-    streamTaskPause(pMeta, pHistoryTask);
+    streamTaskPause(pHistoryTask);
     streamMetaReleaseTask(pMeta, pHistoryTask);
   }
 

source/dnode/vnode/src/tsdb/tsdbCache.c

@@ -130,7 +130,7 @@ static void tsdbClosePgCache(STsdb *pTsdb) {
 enum {
   LFLAG_LAST_ROW = 0,
   LFLAG_LAST = 1,
-  LFLAG_PRIMARY_KEY = (1 << 4),
+  LFLAG_PRIMARY_KEY = CACHESCAN_RETRIEVE_PK,
 };
 
 typedef struct {
@@ -1637,8 +1637,8 @@ int32_t tsdbCacheGetBatch(STsdb *pTsdb, tb_uid_t uid, SArray *pLastArray, SCache
       SLastCol *pLastCol = (SLastCol *)taosLRUCacheValue(pCache, h);
 
       SLastCol lastCol = *pLastCol;
-      for (int8_t i = 0; i < lastCol.rowKey.numOfPKs; i++) {
+      for (int8_t j = 0; j < lastCol.rowKey.numOfPKs; j++) {
-        reallocVarDataVal(&lastCol.rowKey.pks[i]);
+        reallocVarDataVal(&lastCol.rowKey.pks[j]);
       }
       reallocVarData(&lastCol.colVal);
       taosArrayPush(pLastArray, &lastCol);
@@ -1667,8 +1667,8 @@ int32_t tsdbCacheGetBatch(STsdb *pTsdb, tb_uid_t uid, SArray *pLastArray, SCache
         SLastCol *pLastCol = (SLastCol *)taosLRUCacheValue(pCache, h);
 
         SLastCol lastCol = *pLastCol;
-        for (int8_t i = 0; i < lastCol.rowKey.numOfPKs; i++) {
+        for (int8_t j = 0; j < lastCol.rowKey.numOfPKs; j++) {
-          reallocVarDataVal(&lastCol.rowKey.pks[i]);
+          reallocVarDataVal(&lastCol.rowKey.pks[j]);
         }
         reallocVarData(&lastCol.colVal);
         taosArraySet(pLastArray, idxKey->idx, &lastCol);

source/dnode/vnode/src/tsdb/tsdbCacheRead.c

@@ -367,11 +367,7 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32
     goto _end;
   }
 
-  int32_t pkBufLen = 0;
+  int32_t pkBufLen = (pr->rowKey.numOfPKs > 0)? pr->pkColumn.bytes:0;
-  if (pr->rowKey.numOfPKs > 0) {
-    pkBufLen = pr->pkColumn.bytes;
-  }
-
   for (int32_t j = 0; j < pr->numOfCols; ++j) {
     int32_t bytes = (slotIds[j] == -1) ? 1 : pr->pSchema->columns[slotIds[j]].bytes;
 
@@ -387,6 +383,10 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32
   }
 
   int8_t ltype = (pr->type & CACHESCAN_RETRIEVE_LAST) >> 3;
+  if (pr->rowKey.numOfPKs > 0) {
+    ltype |= CACHESCAN_RETRIEVE_PK;
+  }
+
   STableKeyInfo* pTableList = pr->pTableList;
 
   // retrieve the only one last row of all tables in the uid list.

source/dnode/vnode/src/tsdb/tsdbDataFileRW.c

@@ -48,9 +48,24 @@ static int32_t tsdbDataFileReadHeadFooter(SDataFileReader *reader) {
   if (reader->fd[ftype]) {
     int32_t encryptAlgorithm = reader->config->tsdb->pVnode->config.tsdbCfg.encryptAlgorithm;
     char* encryptKey = reader->config->tsdb->pVnode->config.tsdbCfg.encryptKey;
+#if 1
     code = tsdbReadFile(reader->fd[ftype], reader->config->files[ftype].file.size - sizeof(SHeadFooter),
                         (uint8_t *)reader->headFooter, sizeof(SHeadFooter), 0, encryptAlgorithm, encryptKey);
     TSDB_CHECK_CODE(code, lino, _exit);
+#else
+    int64_t size = reader->config->files[ftype].file.size;
+    for (; size > TSDB_FHDR_SIZE; size--) {
+      code = tsdbReadFile(reader->fd[ftype], size - sizeof(SHeadFooter), (uint8_t *)reader->headFooter,
+                          sizeof(SHeadFooter), 0, encryptAlgorithm, encryptKey);
+      if (code) continue;
+      if (reader->headFooter->brinBlkPtr->offset + reader->headFooter->brinBlkPtr->size + sizeof(SHeadFooter) == size) {
+        break;
+      }
+    }
+    if (size <= TSDB_FHDR_SIZE) {
+      TSDB_CHECK_CODE(code = TSDB_CODE_FILE_CORRUPTED, lino, _exit);
+    }
+#endif
   }
 
   reader->ctx->headFooterLoaded = true;

source/dnode/vnode/src/tsdb/tsdbFSetRW.c

@@ -46,7 +46,7 @@ static int32_t tsdbFSetWriteTableDataBegin(SFSetWriter *writer, const TABLEID *t
   code = tsdbUpdateSkmTb(writer->config->tsdb, writer->ctx->tbid, writer->skmTb);
 
   code = metaGetColCmpr(writer->config->tsdb->pVnode->pMeta, tbid->suid ? tbid->suid : tbid->uid, &writer->pColCmprObj);
-  TSDB_CHECK_CODE(code, lino, _exit);
+  // TSDB_CHECK_CODE(code, lino, _exit);
 
   writer->blockDataIdx = 0;
   for (int32_t i = 0; i < ARRAY_SIZE(writer->blockData); i++) {
@@ -301,15 +301,3 @@ _exit:
   }
   return code;
 }
-// int32_t tsdbGetCompressByUid(SFSetWriter *writer, tb_uid_t uid, struct SColCompressInfo *info) {
-//   SHashObj *p = NULL;
-//   int32_t   code = metaGetColCmpr(writer->config->tsdb->pVnode->pMeta, uid, &p);
-//   if (code < 0) {
-//     ASSERT(0);
-//     taosHashCleanup(p);
-//     p = NULL;
-//   } else {
-//   }
-//   info->pColCmpr = p;
-//   return code;
-// }

source/dnode/vnode/src/tsdb/tsdbRead2.c

@@ -1610,11 +1610,6 @@ static bool tryCopyDistinctRowFromSttBlock(TSDBROW* fRow, SSttBlockReader* pSttB
   doUnpinSttBlock(pSttBlockReader);
   if (hasVal) {
     SRowKey* pNext = getCurrentKeyInSttBlock(pSttBlockReader);
-
-    if (IS_VAR_DATA_TYPE(pSttKey->pks[0].type)) {
-      tsdbInfo("current pk:%s, next pk:%s", pSttKey->pks[0].pData, pNext->pks[0].pData);
-    }
-
     if (pkCompEx(pSttKey, pNext) != 0) {
       code = doAppendRowFromFileBlock(pReader->resBlockInfo.pResBlock, pReader, fRow->pBlockData, fRow->iRow);
       *copied = (code == TSDB_CODE_SUCCESS);

source/dnode/vnode/src/tsdb/tsdbReaderWriter.c

@@ -14,8 +14,8 @@
  */
 
 #include "cos.h"
-#include "tsdb.h"
 #include "crypt.h"
+#include "tsdb.h"
 #include "vnd.h"
 
 static int32_t tsdbOpenFileImpl(STsdbFD *pFD) {
@@ -61,6 +61,7 @@ static int32_t tsdbOpenFileImpl(STsdbFD *pFD) {
       // taosMemoryFree(pFD);
       goto _exit;
     }
+    pFD->s3File = 1;
     /*
     const char *object_name = taosDirEntryBaseName((char *)path);
     long        s3_size = 0;
@@ -86,7 +87,6 @@ static int32_t tsdbOpenFileImpl(STsdbFD *pFD) {
         goto _exit;
       }
 #else
-      pFD->s3File = 1;
       pFD->pFD = (TdFilePtr)&pFD->s3File;
       int32_t vid = 0;
       sscanf(object_name, "v%df%dver%" PRId64 ".data", &vid, &pFD->fid, &pFD->cid);
@@ -182,7 +182,7 @@ static int32_t tsdbWriteFilePage(STsdbFD *pFD, int32_t encryptAlgorithm, char* e
 
   if (pFD->pgno > 0) {
     int64_t offset = PAGE_OFFSET(pFD->pgno, pFD->szPage);
-    if (pFD->lcn > 1) {
+    if (pFD->s3File && pFD->lcn > 1) {
       SVnodeCfg *pCfg = &pFD->pTsdb->pVnode->config;
       int64_t    chunksize = (int64_t)pCfg->tsdbPageSize * pCfg->s3ChunkSize;
       int64_t    chunkoffset = chunksize * (pFD->lcn - 1);
@@ -303,8 +303,7 @@ static int32_t tsdbReadFilePage(STsdbFD *pFD, int64_t pgno, int32_t encryptAlgor
     int           NewLen;
 
     int32_t count = 0;
-    while(count < pFD->szPage)
+    while (count < pFD->szPage) {
-    {
       SCryptOpts opts = {0};
       opts.len = 128;
       opts.source = pFD->pBuf + count;
@@ -582,7 +581,7 @@ int32_t tsdbReadFile(STsdbFD *pFD, int64_t offset, uint8_t *pBuf, int64_t size,
     }
   }
 
-  if (pFD->lcn > 1 /*pFD->s3File && tsS3BlockSize < 0*/) {
+  if (pFD->s3File && pFD->lcn > 1 /* && tsS3BlockSize < 0*/) {
     return tsdbReadFileS3(pFD, offset, pBuf, size, szHint);
   } else {
     return tsdbReadFileImp(pFD, offset, pBuf, size, encryptAlgorithm, encryptKey);
@@ -598,8 +597,7 @@ int32_t tsdbReadFileToBuffer(STsdbFD *pFD, int64_t offset, int64_t size, SBuffer
 
   code = tBufferEnsureCapacity(buffer, buffer->size + size);
   if (code) return code;
-  code = tsdbReadFile(pFD, offset, (uint8_t *)tBufferGetDataEnd(buffer), size, szHint,
+  code = tsdbReadFile(pFD, offset, (uint8_t *)tBufferGetDataEnd(buffer), size, szHint, encryptAlgorithm, encryptKey);
-                      encryptAlgorithm, encryptKey);
   if (code) return code;
   buffer->size += size;
 

source/dnode/vnode/src/tsdb/tsdbRetention.c

@@ -557,6 +557,7 @@ static int32_t tsdbMigrateDataFileS3(SRTNer *rtner, const STFileObj *fobj, int64
   int32_t   lino = 0;
   STFileOp  op = {0};
   int32_t   lcn = (size - 1) / chunksize + 1;
+  TdFilePtr fdFrom = NULL, fdTo = NULL;
 
   // remove old
   op = (STFileOp){
@@ -615,7 +616,6 @@ static int32_t tsdbMigrateDataFileS3(SRTNer *rtner, const STFileObj *fobj, int64
   }
 
   // copy last chunk
-  TdFilePtr fdFrom = NULL, fdTo = NULL;
   int64_t lc_offset = (int64_t)(lcn - 1) * chunksize;
   int64_t lc_size = size - lc_offset;
 
@@ -671,7 +671,7 @@ static int32_t tsdbDoS3MigrateOnFileSet(SRTNer *rtner, STFileSet *fset) {
   int64_t chunksize = (int64_t)pCfg->tsdbPageSize * pCfg->s3ChunkSize;
   int32_t lcn = fobj->f->lcn;
 
-  if (lcn < 1 && taosCheckExistFile(fobj->fname)) {
+  if (/*lcn < 1 && */ taosCheckExistFile(fobj->fname)) {
     int32_t mtime = 0;
     int64_t size = 0;
     taosStatFile(fobj->fname, &size, &mtime, NULL);

source/dnode/vnode/src/tsdb/tsdbSttFileRW.c

@@ -65,9 +65,27 @@ int32_t tsdbSttFileReaderOpen(const char *fname, const SSttFileReaderConfig *con
 
   int32_t encryptAlgoirthm = config->tsdb->pVnode->config.tsdbCfg.encryptAlgorithm;
   char* encryptKey = config->tsdb->pVnode->config.tsdbCfg.encryptKey;
+#if 1
   code = tsdbReadFile(reader[0]->fd, offset, (uint8_t *)(reader[0]->footer), sizeof(SSttFooter), 0, encryptAlgoirthm, 
                       encryptKey);
   TSDB_CHECK_CODE(code, lino, _exit);
+#else
+  int64_t size = config->file->size;
+
+  for (; size > TSDB_FHDR_SIZE; size--) {
+    code = tsdbReadFile(reader[0]->fd, size - sizeof(SSttFooter), (uint8_t *)(reader[0]->footer), sizeof(SSttFooter), 0, encryptAlgoirthm, 
+                      encryptKey);
+    if (code) continue;
+    if ((*reader)->footer->sttBlkPtr->offset + (*reader)->footer->sttBlkPtr->size + sizeof(SSttFooter) == size ||
+        (*reader)->footer->statisBlkPtr->offset + (*reader)->footer->statisBlkPtr->size + sizeof(SSttFooter) == size ||
+        (*reader)->footer->tombBlkPtr->offset + (*reader)->footer->tombBlkPtr->size + sizeof(SSttFooter) == size) {
+      break;
+    }
+  }
+  if (size <= TSDB_FHDR_SIZE) {
+    TSDB_CHECK_CODE(code = TSDB_CODE_FILE_CORRUPTED, lino, _exit);
+  }
+#endif
 
 _exit:
   if (code) {

source/dnode/vnode/src/vnd/vnodeOpen.c

@@ -486,8 +486,6 @@ SVnode *vnodeOpen(const char *path, int32_t diskPrimary, STfs *pTfs, SMsgCb msgC
 
   if (tsEnableMonitor && pVnode->monitor.insertCounter == NULL) {
     taos_counter_t *counter = NULL;
-    counter = taos_collector_registry_get_metric(VNODE_METRIC_SQL_COUNT);
-    if(counter == NULL){
     int32_t label_count = 7;
     const char *sample_labels[] = {VNODE_METRIC_TAG_NAME_SQL_TYPE, VNODE_METRIC_TAG_NAME_CLUSTER_ID,
                                   VNODE_METRIC_TAG_NAME_DNODE_ID, VNODE_METRIC_TAG_NAME_DNODE_EP,
@@ -501,7 +499,6 @@ SVnode *vnodeOpen(const char *path, int32_t diskPrimary, STfs *pTfs, SMsgCb msgC
       counter = taos_collector_registry_get_metric(VNODE_METRIC_SQL_COUNT);
       vInfo("vgId:%d, get metric from registry:%p", TD_VID(pVnode), counter);
     }
-    }
     pVnode->monitor.insertCounter = counter;
     vInfo("vgId:%d, succeed to set metric:%p", TD_VID(pVnode), counter);
   }

source/dnode/vnode/src/vnd/vnodeSvr.c

@@ -1076,16 +1076,6 @@ static int32_t vnodeProcessCreateTbReq(SVnode *pVnode, int64_t ver, void *pReq,
     pCreateReq = req.pReqs + iReq;
     memset(&cRsp, 0, sizeof(cRsp));
 
-    if ((terrno = grantCheck(TSDB_GRANT_TIMESERIES)) < 0) {
-      rcode = -1;
-      goto _exit;
-    }
-
-    if ((terrno = grantCheck(TSDB_GRANT_TABLE)) < 0) {
-      rcode = -1;
-      goto _exit;
-    }
-
     if (tsEnableAudit && tsEnableAuditCreateTable) {
       char *str = taosMemoryCalloc(1, TSDB_TABLE_FNAME_LEN);
       if (str == NULL) {
@@ -1778,13 +1768,6 @@ static int32_t vnodeProcessSubmitReq(SVnode *pVnode, int64_t ver, void *pReq, in
 
     // create table
     if (pSubmitTbData->pCreateTbReq) {
-      // check (TODO: move check to create table)
-      code = grantCheck(TSDB_GRANT_TIMESERIES);
-      if (code) goto _exit;
-
-      code = grantCheck(TSDB_GRANT_TABLE);
-      if (code) goto _exit;
-
       // alloc if need
       if (pSubmitRsp->aCreateTbRsp == NULL &&
           (pSubmitRsp->aCreateTbRsp = taosArrayInit(TARRAY_SIZE(pSubmitReq->aSubmitTbData), sizeof(SVCreateTbRsp))) ==

source/libs/catalog/src/ctgCache.c

@@ -596,6 +596,7 @@ int32_t ctgCopyTbMeta(SCatalog *pCtg, SCtgTbMetaCtx *ctx, SCtgDBCache **pDb, SCt
   }
 
   memcpy(&(*pTableMeta)->sversion, &stbMeta->sversion, metaSize - sizeof(SCTableMeta));
+  (*pTableMeta)->schemaExt =  NULL;
 
   return TSDB_CODE_SUCCESS;
 }
@@ -2883,14 +2884,24 @@ int32_t ctgGetTbMetasFromCache(SCatalog *pCtg, SRequestConnInfo *pConn, SCtgTbMe
     SMetaRes    res = {0};
     STableMeta *pTableMeta = NULL;
     if (tbMeta->tableType != TSDB_CHILD_TABLE) {
+      int32_t schemaExtSize = 0;
       int32_t metaSize = CTG_META_SIZE(tbMeta);
-      pTableMeta = taosMemoryCalloc(1, metaSize);
+      if (tbMeta->schemaExt != NULL) {
+        schemaExtSize = tbMeta->tableInfo.numOfColumns * sizeof(SSchemaExt);
+      }
+      pTableMeta = taosMemoryCalloc(1, metaSize + schemaExtSize);
       if (NULL == pTableMeta) {
         ctgReleaseTbMetaToCache(pCtg, dbCache, pCache);
         CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
       }
 
       memcpy(pTableMeta, tbMeta, metaSize);
+      if (tbMeta->schemaExt != NULL) {
+        pTableMeta->schemaExt = (SSchemaExt *)((char *)pTableMeta + metaSize);
+        memcpy(pTableMeta->schemaExt, tbMeta->schemaExt, schemaExtSize);
+      } else {
+        pTableMeta->schemaExt = NULL;
+      }
 
       CTG_UNLOCK(CTG_READ, &pCache->metaLock);
       taosHashRelease(dbCache->tbCache, pCache);
@@ -2999,6 +3010,7 @@ int32_t ctgGetTbMetasFromCache(SCatalog *pCtg, SRequestConnInfo *pConn, SCtgTbMe
     }
 
     memcpy(&pTableMeta->sversion, &stbMeta->sversion, metaSize - sizeof(SCTableMeta));
+    pTableMeta->schemaExt = NULL;
 
     CTG_UNLOCK(CTG_READ, &pCache->metaLock);
     taosHashRelease(dbCache->tbCache, pCache);

source/libs/command/src/command.c

@@ -18,10 +18,10 @@
 #include "commandInt.h"
 #include "scheduler.h"
 #include "systable.h"
+#include "taosdef.h"
 #include "tdatablock.h"
 #include "tglobal.h"
 #include "tgrant.h"
-#include "taosdef.h"
 
 extern SConfig* tsCfg;
 
@@ -126,6 +126,7 @@ static int32_t setDescResultIntoDataBlock(bool sysInfoUser, SSDataBlock* pBlock,
     pCol7 = taosArrayGet(pBlock->pDataBlock, 6);
   }
 
+  int32_t fillTagCol = 0;
   char    buf[DESCRIBE_RESULT_FIELD_LEN] = {0};
   for (int32_t i = 0; i < numOfRows; ++i) {
     if (invisibleColumn(sysInfoUser, pMeta->tableType, pMeta->schema[i].flags)) {
@@ -140,6 +141,7 @@ static int32_t setDescResultIntoDataBlock(bool sysInfoUser, SSDataBlock* pBlock,
     if (TSDB_VIEW_TABLE != pMeta->tableType) {
       if (i >= pMeta->tableInfo.numOfColumns) {
         STR_TO_VARSTR(buf, "TAG");
+        fillTagCol = 1;
       } else if (i == 1 && pMeta->schema[i].flags & COL_IS_KEY) {
         STR_TO_VARSTR(buf, "PRIMARY KEY")
       } else {
@@ -158,15 +160,17 @@ static int32_t setDescResultIntoDataBlock(bool sysInfoUser, SSDataBlock* pBlock,
         STR_TO_VARSTR(buf, columnLevelStr(COMPRESS_L2_TYPE_LEVEL_U32(pMeta->schemaExt[i].compress)));
         colDataSetVal(pCol7, pBlock->info.rows, buf, false);
       } else {
-        STR_TO_VARSTR(buf, "");
+        STR_TO_VARSTR(buf, fillTagCol == 0 ? "" : "disabled");
         colDataSetVal(pCol5, pBlock->info.rows, buf, false);
-        STR_TO_VARSTR(buf, "");
+        STR_TO_VARSTR(buf, fillTagCol == 0 ? "" : "disabled");
         colDataSetVal(pCol6, pBlock->info.rows, buf, false);
-        STR_TO_VARSTR(buf, "");
+        STR_TO_VARSTR(buf, fillTagCol == 0 ? "" : "disabled");
         colDataSetVal(pCol7, pBlock->info.rows, buf, false);
       }
     }
 
+    fillTagCol = 0;
+
     ++(pBlock->info.rows);
   }
   if (pMeta->tableType == TSDB_SUPER_TABLE && biMode != 0) {
@@ -367,17 +371,20 @@ static void setCreateDBResultIntoDataBlock(SSDataBlock* pBlock, char* dbName, ch
   if (IS_SYS_DBNAME(dbName)) {
     len += sprintf(buf2 + VARSTR_HEADER_SIZE, "CREATE DATABASE `%s`", dbName);
   } else {
-    len += sprintf(
+    len += sprintf(buf2 + VARSTR_HEADER_SIZE,
-        buf2 + VARSTR_HEADER_SIZE,
                    "CREATE DATABASE `%s` BUFFER %d CACHESIZE %d CACHEMODEL '%s' COMP %d DURATION %dm "
-        "WAL_FSYNC_PERIOD %d MAXROWS %d MINROWS %d STT_TRIGGER %d KEEP %dm,%dm,%dm PAGES %d PAGESIZE %d PRECISION '%s' REPLICA %d "
+                   "WAL_FSYNC_PERIOD %d MAXROWS %d MINROWS %d STT_TRIGGER %d KEEP %dm,%dm,%dm PAGES %d PAGESIZE %d "
+                   "PRECISION '%s' REPLICA %d "
                    "WAL_LEVEL %d VGROUPS %d SINGLE_STABLE %d TABLE_PREFIX %d TABLE_SUFFIX %d TSDB_PAGESIZE %d "
-        "WAL_RETENTION_PERIOD %d WAL_RETENTION_SIZE %" PRId64 " KEEP_TIME_OFFSET %d ENCRYPT_ALGORITHM '%s' S3_CHUNKSIZE %d S3_KEEPLOCAL %dm S3_COMPACT %d",
+                   "WAL_RETENTION_PERIOD %d WAL_RETENTION_SIZE %" PRId64
-        dbName, pCfg->buffer, pCfg->cacheSize, cacheModelStr(pCfg->cacheLast), pCfg->compression, pCfg->daysPerFile,
+                   " KEEP_TIME_OFFSET %d ENCRYPT_ALGORITHM '%s' S3_CHUNKSIZE %d S3_KEEPLOCAL %dm S3_COMPACT %d",
-        pCfg->walFsyncPeriod, pCfg->maxRows, pCfg->minRows,  pCfg->sstTrigger, pCfg->daysToKeep0, pCfg->daysToKeep1, pCfg->daysToKeep2,
+                   dbName, pCfg->buffer, pCfg->cacheSize, cacheModelStr(pCfg->cacheLast), pCfg->compression,
-        pCfg->pages, pCfg->pageSize, prec, pCfg->replications, pCfg->walLevel, pCfg->numOfVgroups,
+                   pCfg->daysPerFile, pCfg->walFsyncPeriod, pCfg->maxRows, pCfg->minRows, pCfg->sstTrigger,
-        1 == pCfg->numOfStables, hashPrefix, pCfg->hashSuffix, pCfg->tsdbPageSize, pCfg->walRetentionPeriod, pCfg->walRetentionSize,
+                   pCfg->daysToKeep0, pCfg->daysToKeep1, pCfg->daysToKeep2, pCfg->pages, pCfg->pageSize, prec,
-        pCfg->keepTimeOffset, encryptAlgorithmStr(pCfg->encryptAlgorithm), pCfg->s3ChunkSize, pCfg->s3KeepLocal, pCfg->s3Compact);
+                   pCfg->replications, pCfg->walLevel, pCfg->numOfVgroups, 1 == pCfg->numOfStables, hashPrefix,
+                   pCfg->hashSuffix, pCfg->tsdbPageSize, pCfg->walRetentionPeriod, pCfg->walRetentionSize,
+                   pCfg->keepTimeOffset, encryptAlgorithmStr(pCfg->encryptAlgorithm), pCfg->s3ChunkSize,
+                   pCfg->s3KeepLocal, pCfg->s3Compact);
 
     if (retentions) {
       len += sprintf(buf2 + VARSTR_HEADER_SIZE + len, " RETENTIONS %s", retentions);
@@ -391,7 +398,9 @@ static void setCreateDBResultIntoDataBlock(SSDataBlock* pBlock, char* dbName, ch
   colDataSetVal(pCol2, 0, buf2, false);
 }
 
-#define CHECK_LEADER(n) (row[n] && (fields[n].type == TSDB_DATA_TYPE_VARCHAR && strncasecmp(row[n], "leader", varDataLen((char *)row[n] - VARSTR_HEADER_SIZE)) == 0))
+#define CHECK_LEADER(n)                                   \
+  (row[n] && (fields[n].type == TSDB_DATA_TYPE_VARCHAR && \
+              strncasecmp(row[n], "leader", varDataLen((char*)row[n] - VARSTR_HEADER_SIZE)) == 0))
 // on this row, if have leader return true else return false
 bool existLeaderRole(TAOS_ROW row, TAOS_FIELD* fields, int nFields) {
   // vgroup_id | db_name | tables | v1_dnode | v1_status | v2_dnode | v2_status | v3_dnode | v3_status | v4_dnode |
@@ -548,23 +557,25 @@ static int32_t buildCreateViewResultDataBlock(SSDataBlock** pOutput) {
   return code;
 }
 
-
-
 void appendColumnFields(char* buf, int32_t* len, STableCfg* pCfg) {
   for (int32_t i = 0; i < pCfg->numOfColumns; ++i) {
     SSchema* pSchema = pCfg->pSchemas + i;
     char     type[32 + 60];  // 60 byte for compress info
     sprintf(type, "%s", tDataTypes[pSchema->type].name);
-    if (TSDB_DATA_TYPE_VARCHAR == pSchema->type || TSDB_DATA_TYPE_VARBINARY == pSchema->type || TSDB_DATA_TYPE_GEOMETRY == pSchema->type) {
+    if (TSDB_DATA_TYPE_VARCHAR == pSchema->type || TSDB_DATA_TYPE_VARBINARY == pSchema->type ||
+        TSDB_DATA_TYPE_GEOMETRY == pSchema->type) {
       sprintf(type + strlen(type), "(%d)", (int32_t)(pSchema->bytes - VARSTR_HEADER_SIZE));
     } else if (TSDB_DATA_TYPE_NCHAR == pSchema->type) {
       sprintf(type + strlen(type), "(%d)", (int32_t)((pSchema->bytes - VARSTR_HEADER_SIZE) / TSDB_NCHAR_SIZE));
     }
 
     if (useCompress(pCfg->tableType)) {
-      sprintf(type + strlen(type), " ENCODE \'%s\'", columnEncodeStr(COMPRESS_L1_TYPE_U32(pCfg->pSchemaExt[i].compress)));
+      sprintf(type + strlen(type), " ENCODE \'%s\'",
-      sprintf(type + strlen(type), " COMPRESS \'%s\'", columnCompressStr(COMPRESS_L2_TYPE_U32(pCfg->pSchemaExt[i].compress)));
+              columnEncodeStr(COMPRESS_L1_TYPE_U32(pCfg->pSchemaExt[i].compress)));
-      sprintf(type + strlen(type), " LEVEL \'%s\'", columnLevelStr(COMPRESS_L2_TYPE_LEVEL_U32(pCfg->pSchemaExt[i].compress)));
+      sprintf(type + strlen(type), " COMPRESS \'%s\'",
+              columnCompressStr(COMPRESS_L2_TYPE_U32(pCfg->pSchemaExt[i].compress)));
+      sprintf(type + strlen(type), " LEVEL \'%s\'",
+              columnLevelStr(COMPRESS_L2_TYPE_LEVEL_U32(pCfg->pSchemaExt[i].compress)));
     }
     if (!(pSchema->flags & COL_IS_KEY)) {
       *len += sprintf(buf + VARSTR_HEADER_SIZE + *len, "%s`%s` %s", ((i > 0) ? ", " : ""), pSchema->name, type);
@@ -580,7 +591,8 @@ void appendTagFields(char* buf, int32_t* len, STableCfg* pCfg) {
     SSchema* pSchema = pCfg->pSchemas + pCfg->numOfColumns + i;
     char     type[32];
     sprintf(type, "%s", tDataTypes[pSchema->type].name);
-    if (TSDB_DATA_TYPE_VARCHAR == pSchema->type || TSDB_DATA_TYPE_VARBINARY == pSchema->type || TSDB_DATA_TYPE_GEOMETRY == pSchema->type) {
+    if (TSDB_DATA_TYPE_VARCHAR == pSchema->type || TSDB_DATA_TYPE_VARBINARY == pSchema->type ||
+        TSDB_DATA_TYPE_GEOMETRY == pSchema->type) {
       sprintf(type + strlen(type), "(%d)", (int32_t)(pSchema->bytes - VARSTR_HEADER_SIZE));
     } else if (TSDB_DATA_TYPE_NCHAR == pSchema->type) {
       sprintf(type + strlen(type), "(%d)", (int32_t)((pSchema->bytes - VARSTR_HEADER_SIZE) / TSDB_NCHAR_SIZE));
@@ -823,7 +835,8 @@ static int32_t setCreateViewResultIntoDataBlock(SSDataBlock* pBlock, SShowCreate
 
   SViewMeta* pMeta = pStmt->pViewMeta;
   ASSERT(pMeta);
-  snprintf(varDataVal(buf2), SHOW_CREATE_VIEW_RESULT_FIELD2_LEN - VARSTR_HEADER_SIZE, "CREATE VIEW `%s`.`%s` AS %s", pStmt->dbName, pStmt->viewName, pMeta->querySql);
+  snprintf(varDataVal(buf2), SHOW_CREATE_VIEW_RESULT_FIELD2_LEN - VARSTR_HEADER_SIZE, "CREATE VIEW `%s`.`%s` AS %s",
+           pStmt->dbName, pStmt->viewName, pMeta->querySql);
   int32_t len = strlen(varDataVal(buf2));
   varDataLen(buf2) = (len > 65535) ? 65535 : len;
   colDataSetVal(pCol2, 0, buf2, false);
@@ -833,7 +846,6 @@ static int32_t setCreateViewResultIntoDataBlock(SSDataBlock* pBlock, SShowCreate
   return TSDB_CODE_SUCCESS;
 }
 
-
 static int32_t execShowCreateTable(SShowCreateTableStmt* pStmt, SRetrieveTableRsp** pRsp) {
   SSDataBlock* pBlock = NULL;
   int32_t      code = buildCreateTbResultDataBlock(&pBlock);
@@ -955,46 +967,11 @@ static int32_t buildLocalVariablesResultDataBlock(SSDataBlock** pOutput) {
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t setLocalVariablesResultIntoDataBlock(SSDataBlock* pBlock) {
-  int32_t numOfCfg = taosArrayGetSize(tsCfg->array);
-  int32_t numOfRows = 0;
-  blockDataEnsureCapacity(pBlock, numOfCfg);
-
-  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++);
-    colDataSetVal(pColInfo, i, name, false);
-
-    char    value[TSDB_CONFIG_VALUE_LEN + VARSTR_HEADER_SIZE] = {0};
-    int32_t valueLen = 0;
-    cfgDumpItemValue(pItem, &value[VARSTR_HEADER_SIZE], TSDB_CONFIG_VALUE_LEN, &valueLen);
-    varDataSetLen(value, valueLen);
-    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++;
-  }
-
-  pBlock->info.rows = numOfRows;
-
-  return TSDB_CODE_SUCCESS;
-}
-
 static int32_t execShowLocalVariables(SRetrieveTableRsp** pRsp) {
   SSDataBlock* pBlock = NULL;
   int32_t      code = buildLocalVariablesResultDataBlock(&pBlock);
   if (TSDB_CODE_SUCCESS == code) {
-    code = setLocalVariablesResultIntoDataBlock(pBlock);
+    code = dumpConfToDataBlock(pBlock, 0);
   }
   if (TSDB_CODE_SUCCESS == code) {
     code = buildRetrieveTableRsp(pBlock, SHOW_LOCAL_VARIABLES_RESULT_COLS, pRsp);

source/libs/executor/inc/executorInt.h

@@ -500,6 +500,8 @@ typedef struct SStreamScanInfo {
   SStoreTqReader readerFn;
   SStateStore    stateStore;
   SSDataBlock*   pCheckpointRes;
+  int8_t         pkColType;
+  int32_t        pkColLen;
 } SStreamScanInfo;
 
 typedef struct {
@@ -566,8 +568,13 @@ typedef struct SOpCheckPointInfo {
   SHashObj* children;  // key:child id
 } SOpCheckPointInfo;
 
+typedef struct SSteamOpBasicInfo {
+  int32_t primaryPkIndex;
+} SSteamOpBasicInfo;
+
 typedef struct SStreamIntervalOperatorInfo {
   SOptrBasicInfo      binfo;           // basic info
+  SSteamOpBasicInfo   basic;
   SAggSupporter       aggSup;          // aggregate supporter
   SExprSupp           scalarSupp;      // supporter for perform scalar function
   SGroupResInfo       groupResInfo;    // multiple results build supporter
@@ -633,6 +640,7 @@ typedef struct SResultWindowInfo {
 
 typedef struct SStreamSessionAggOperatorInfo {
   SOptrBasicInfo      binfo;
+  SSteamOpBasicInfo   basic;
   SStreamAggSupporter streamAggSup;
   SExprSupp           scalarSupp;  // supporter for perform scalar function
   SGroupResInfo       groupResInfo;
@@ -665,6 +673,7 @@ typedef struct SStreamSessionAggOperatorInfo {
 
 typedef struct SStreamStateAggOperatorInfo {
   SOptrBasicInfo      binfo;
+  SSteamOpBasicInfo   basic;
   SStreamAggSupporter streamAggSup;
   SExprSupp           scalarSupp;  // supporter for perform scalar function
   SGroupResInfo       groupResInfo;
@@ -691,6 +700,7 @@ typedef struct SStreamStateAggOperatorInfo {
 
 typedef struct SStreamEventAggOperatorInfo {
   SOptrBasicInfo      binfo;
+  SSteamOpBasicInfo   basic;
   SStreamAggSupporter streamAggSup;
   SExprSupp           scalarSupp;  // supporter for perform scalar function
   SGroupResInfo       groupResInfo;
@@ -719,6 +729,7 @@ typedef struct SStreamEventAggOperatorInfo {
 
 typedef struct SStreamCountAggOperatorInfo {
   SOptrBasicInfo      binfo;
+  SSteamOpBasicInfo   basic;
   SStreamAggSupporter streamAggSup;
   SExprSupp           scalarSupp;  // supporter for perform scalar function
   SGroupResInfo       groupResInfo;
@@ -742,6 +753,7 @@ typedef struct SStreamCountAggOperatorInfo {
 
 typedef struct SStreamPartitionOperatorInfo {
   SOptrBasicInfo        binfo;
+  SSteamOpBasicInfo     basic;
   SPartitionBySupporter partitionSup;
   SExprSupp             scalarSup;
   SExprSupp             tbnameCalSup;
@@ -775,6 +787,7 @@ typedef struct SStreamFillSupporter {
 } SStreamFillSupporter;
 
 typedef struct SStreamFillOperatorInfo {
+  SSteamOpBasicInfo     basic;
   SStreamFillSupporter* pFillSup;
   SSDataBlock*          pRes;
   SSDataBlock*          pSrcBlock;
@@ -911,7 +924,7 @@ int32_t  initStreamAggSupporter(SStreamAggSupporter* pSup, SExprSupp* pExpSup, i
                                 SReadHandle* pHandle, STimeWindowAggSupp* pTwAggSup, const char* taskIdStr,
                                 SStorageAPI* pApi, int32_t tsIndex);
 void     initDownStream(struct SOperatorInfo* downstream, SStreamAggSupporter* pAggSup, uint16_t type, int32_t tsColIndex,
-                        STimeWindowAggSupp* pTwSup);
+                        STimeWindowAggSupp* pTwSup, struct SSteamOpBasicInfo* pBasic);
 void     getMaxTsWins(const SArray* pAllWins, SArray* pMaxWins);
 void     initGroupResInfoFromArrayList(SGroupResInfo* pGroupResInfo, SArray* pArrayList);
 void     getSessionHashKey(const SSessionKey* pKey, SSessionKey* pHashKey);
@@ -939,11 +952,12 @@ void     compactTimeWindow(SExprSupp* pSup, SStreamAggSupporter* pAggSup, STimeW
                            SSHashObj* pStUpdated, SSHashObj* pStDeleted, bool addGap);
 int32_t  releaseOutputBuf(void* pState, SRowBuffPos* pPos, SStateStore* pAPI);
 void     resetWinRange(STimeWindow* winRange);
-bool     checkExpiredData(SStateStore* pAPI, SUpdateInfo* pUpdateInfo, STimeWindowAggSupp* pTwSup, uint64_t tableId, TSKEY ts);
+bool     checkExpiredData(SStateStore* pAPI, SUpdateInfo* pUpdateInfo, STimeWindowAggSupp* pTwSup, uint64_t tableId, TSKEY ts, void* pPkVal, int32_t len);
 int64_t  getDeleteMark(SWindowPhysiNode* pWinPhyNode, int64_t interval);
 void     resetUnCloseSessionWinInfo(SSHashObj* winMap);
 void     setStreamOperatorCompleted(struct SOperatorInfo* pOperator);
 void     reloadAggSupFromDownStream(struct SOperatorInfo* downstream, SStreamAggSupporter* pAggSup);
+void     destroyFlusedPos(void* pRes);
 
 int32_t encodeSSessionKey(void** buf, SSessionKey* key);
 void*   decodeSSessionKey(void* buf, SSessionKey* key);

source/libs/executor/src/aggregateoperator.c

@@ -278,7 +278,7 @@ int32_t doAggregateImpl(SOperatorInfo* pOperator, SqlFunctionCtx* pCtx) {
   int32_t code = TSDB_CODE_SUCCESS;
   for (int32_t k = 0; k < pOperator->exprSupp.numOfExprs; ++k) {
     if (functionNeedToExecute(&pCtx[k])) {
-      // todo add a dummy funtion to avoid process check
+      // todo add a dummy function to avoid process check
       if (pCtx[k].fpSet.process == NULL) {
         continue;
       }

source/libs/executor/src/cachescanoperator.c

@@ -160,6 +160,9 @@ SOperatorInfo* createCacherowsScanOperator(SLastRowScanPhysiNode* pScanNode, SRe
   // partition by tbname
   if (oneTableForEachGroup(pTableListInfo) || (totalTables == 1)) {
     pInfo->retrieveType = CACHESCAN_RETRIEVE_TYPE_ALL | SCAN_ROW_TYPE(pScanNode->ignoreNull);
+    if (pInfo->numOfPks > 0) {
+      pInfo->retrieveType |= CACHESCAN_RETRIEVE_PK;
+    }
 
     STableKeyInfo* pList = tableListGetInfo(pTableListInfo, 0);
 
@@ -217,6 +220,8 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
   SCacheRowsScanInfo* pInfo = pOperator->info;
   SExecTaskInfo*      pTaskInfo = pOperator->pTaskInfo;
   STableListInfo*     pTableList = pInfo->pTableList;
+  SStoreCacheReader*  pReaderFn = &pInfo->readHandle.api.cacheFn;
+  SSDataBlock*        pBufRes = pInfo->pBufferedRes;
 
   uint64_t suid = tableListGetSuid(pTableList);
   int32_t  size = tableListGetSize(pTableList);
@@ -233,18 +238,18 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
       T_LONG_JMP(pTaskInfo->env, pTaskInfo->code);
     }
 
-    if (pInfo->indexOfBufferedRes >= pInfo->pBufferedRes->info.rows) {
+    if (pInfo->indexOfBufferedRes >= pBufRes->info.rows) {
-      blockDataCleanup(pInfo->pBufferedRes);
+      blockDataCleanup(pBufRes);
       taosArrayClear(pInfo->pUidList);
 
-      int32_t code = pInfo->readHandle.api.cacheFn.retrieveRows(pInfo->pLastrowReader, pInfo->pBufferedRes,
+      int32_t code =
-                                                                pInfo->pSlotIds, pInfo->pDstSlotIds, pInfo->pUidList);
+          pReaderFn->retrieveRows(pInfo->pLastrowReader, pBufRes, pInfo->pSlotIds, pInfo->pDstSlotIds, pInfo->pUidList);
       if (code != TSDB_CODE_SUCCESS) {
         T_LONG_JMP(pTaskInfo->env, code);
       }
 
       // check for tag values
-      int32_t resultRows = pInfo->pBufferedRes->info.rows;
+      int32_t resultRows = pBufRes->info.rows;
 
       // the results may be null, if last values are all null
       ASSERT(resultRows == 0 || resultRows == taosArrayGetSize(pInfo->pUidList));
@@ -253,12 +258,12 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
 
     SSDataBlock* pRes = pInfo->pRes;
 
-    if (pInfo->indexOfBufferedRes < pInfo->pBufferedRes->info.rows) {
+    if (pInfo->indexOfBufferedRes < pBufRes->info.rows) {
-      for (int32_t i = 0; i < taosArrayGetSize(pInfo->pBufferedRes->pDataBlock); ++i) {
+      for (int32_t i = 0; i < taosArrayGetSize(pBufRes->pDataBlock); ++i) {
         SColumnInfoData* pCol = taosArrayGet(pRes->pDataBlock, i);
         int32_t        slotId = pCol->info.slotId;
 
-        SColumnInfoData* pSrc = taosArrayGet(pInfo->pBufferedRes->pDataBlock, slotId);
+        SColumnInfoData* pSrc = taosArrayGet(pBufRes->pDataBlock, slotId);
         SColumnInfoData* pDst = taosArrayGet(pRes->pDataBlock, slotId);
 
         if (colDataIsNull_s(pSrc, pInfo->indexOfBufferedRes)) {
@@ -307,10 +312,10 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
       }
 
       if (NULL == pInfo->pLastrowReader) {
-        code = pInfo->readHandle.api.cacheFn.openReader(
+        code = pReaderFn->openReader(pInfo->readHandle.vnode, pInfo->retrieveType, pList, num,
-            pInfo->readHandle.vnode, pInfo->retrieveType, pList, num, taosArrayGetSize(pInfo->matchInfo.pList),
+                                     taosArrayGetSize(pInfo->matchInfo.pList), pInfo->pCidList, pInfo->pSlotIds, suid,
-            pInfo->pCidList, pInfo->pSlotIds, suid, &pInfo->pLastrowReader, pTaskInfo->id.str, pInfo->pFuncTypeList,
+                                     &pInfo->pLastrowReader, pTaskInfo->id.str, pInfo->pFuncTypeList, &pInfo->pkCol,
-            &pInfo->pkCol, pInfo->numOfPks);
+                                     pInfo->numOfPks);
 
         if (code != TSDB_CODE_SUCCESS) {
           pInfo->currentGroupIndex += 1;
@@ -318,12 +323,12 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
           continue;
         }
       } else {
-        pInfo->readHandle.api.cacheFn.reuseReader(pInfo->pLastrowReader, pList, num);
+        pReaderFn->reuseReader(pInfo->pLastrowReader, pList, num);
       }
 
       taosArrayClear(pInfo->pUidList);
 
-      code = pInfo->readHandle.api.cacheFn.retrieveRows(pInfo->pLastrowReader, pInfo->pRes, pInfo->pSlotIds, pInfo->pDstSlotIds,
+      code = pReaderFn->retrieveRows(pInfo->pLastrowReader, pInfo->pRes, pInfo->pSlotIds, pInfo->pDstSlotIds,
                                      pInfo->pUidList);
       if (code != TSDB_CODE_SUCCESS) {
         T_LONG_JMP(pTaskInfo->env, code);
@@ -357,7 +362,7 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
       }
     }
 
-    pInfo->pLastrowReader = pInfo->readHandle.api.cacheFn.closeReader(pInfo->pLastrowReader);
+    pInfo->pLastrowReader = pReaderFn->closeReader(pInfo->pLastrowReader);
     setOperatorCompleted(pOperator);
     return NULL;
   }

source/libs/executor/src/groupoperator.c

@@ -1267,7 +1267,7 @@ void initParDownStream(SOperatorInfo* downstream, SPartitionBySupporter* pParSup
   pScanInfo->pPartScalarSup = pExpr;
   pScanInfo->pPartTbnameSup = pTbnameExpr;
   if (!pScanInfo->pUpdateInfo) {
-    pScanInfo->pUpdateInfo = pAPI->stateStore.updateInfoInit(60000, TSDB_TIME_PRECISION_MILLI, 0, pScanInfo->igCheckUpdate);
+    pScanInfo->pUpdateInfo = pAPI->stateStore.updateInfoInit(60000, TSDB_TIME_PRECISION_MILLI, 0, pScanInfo->igCheckUpdate, pScanInfo->pkColType, pScanInfo->pkColLen);
   }
 }
 

source/libs/executor/src/scanoperator.c

@@ -1380,7 +1380,7 @@ bool comparePrimaryKey(SColumnInfoData* pCol, int32_t rowId, void* pVal) {
   return false;
 }
 
-bool hasPrimaryKey(SStreamScanInfo* pInfo) {
+bool hasPrimaryKeyCol(SStreamScanInfo* pInfo) {
   return pInfo->primaryKeyIndex != -1;
 }
 
@@ -1391,7 +1391,7 @@ static uint64_t getGroupIdByCol(SStreamScanInfo* pInfo, uint64_t uid, TSKEY ts,
   }
 
   int32_t rowId = 0;
-  if (hasPrimaryKey(pInfo)) {
+  if (hasPrimaryKeyCol(pInfo)) {
     SColumnInfoData* pPkCol = taosArrayGet(pPreRes->pDataBlock, pInfo->primaryKeyIndex);
     for (; rowId < pPreRes->info.rows; rowId++) {
       if (comparePrimaryKey(pPkCol, rowId, pVal)) {
@@ -1630,7 +1630,7 @@ static void getPreVersionDataBlock(uint64_t uid, TSKEY startTs, TSKEY endTs, int
 
   SColumnInfoData* pTsCol = (SColumnInfoData*)taosArrayGet(pPreRes->pDataBlock, pInfo->primaryTsIndex);
   SColumnInfoData* pPkCol = NULL;
-  if (hasPrimaryKey(pInfo)) {
+  if (hasPrimaryKeyCol(pInfo)) {
     pPkCol = (SColumnInfoData*)taosArrayGet(pPreRes->pDataBlock, pInfo->primaryKeyIndex);
   }
   for (int32_t i = 0; i < pPreRes->info.rows; i++) {
@@ -1659,7 +1659,7 @@ static int32_t generateSessionScanRange(SStreamScanInfo* pInfo, SSDataBlock* pSr
   }
   int64_t ver = pSrcBlock->info.version - 1;
 
-  if (pInfo->partitionSup.needCalc && ( startData[0] != endData[0] || (hasPrimaryKey(pInfo) && mode == STREAM_DELETE_DATA) )) {
+  if (pInfo->partitionSup.needCalc && ( startData[0] != endData[0] || (hasPrimaryKeyCol(pInfo) && mode == STREAM_DELETE_DATA) )) {
     getPreVersionDataBlock(uidCol[0], startData[0], endData[0], ver, GET_TASKID(pTaskInfo), pInfo, pSrcBlock);
     startData = (TSKEY*)pStartTsCol->pData;
     endData = (TSKEY*)pEndTsCol->pData;
@@ -1682,7 +1682,7 @@ static int32_t generateSessionScanRange(SStreamScanInfo* pInfo, SSDataBlock* pSr
     uint64_t groupId = pSrcGp[i];
     if (groupId == 0) {
       void* pVal = NULL;
-      if (hasPrimaryKey(pInfo) && pSrcPkCol) {
+      if (hasPrimaryKeyCol(pInfo) && pSrcPkCol) {
         pVal = colDataGetData(pSrcPkCol, i);
       }
       groupId = getGroupIdByData(pInfo, uidCol[i], startData[i], ver, pVal);
@@ -1736,7 +1736,7 @@ static int32_t generateCountScanRange(SStreamScanInfo* pInfo, SSDataBlock* pSrcB
   }
   int64_t ver = pSrcBlock->info.version - 1;
 
-  if (pInfo->partitionSup.needCalc && ( startData[0] != endData[0] || (hasPrimaryKey(pInfo) && mode == STREAM_DELETE_DATA) )) {
+  if (pInfo->partitionSup.needCalc && ( startData[0] != endData[0] || (hasPrimaryKeyCol(pInfo) && mode == STREAM_DELETE_DATA) )) {
     getPreVersionDataBlock(uidCol[0], startData[0], endData[0], ver, GET_TASKID(pTaskInfo), pInfo, pSrcBlock);
     startData = (TSKEY*)pStartTsCol->pData;
     endData = (TSKEY*)pEndTsCol->pData;
@@ -1759,7 +1759,7 @@ static int32_t generateCountScanRange(SStreamScanInfo* pInfo, SSDataBlock* pSrcB
     uint64_t groupId = pSrcGp[i];
     if (groupId == 0) {
       void* pVal = NULL;
-      if (hasPrimaryKey(pInfo) && pSrcPkCol) {
+      if (hasPrimaryKeyCol(pInfo) && pSrcPkCol) {
         pVal = colDataGetData(pSrcPkCol, i);
       }
       groupId = getGroupIdByData(pInfo, uidCol[i], startData[i], ver, pVal);
@@ -1800,7 +1800,7 @@ static int32_t generateIntervalScanRange(SStreamScanInfo* pInfo, SSDataBlock* pS
   TSKEY*  srcEndTsCol = (TSKEY*)pSrcEndTsCol->pData;
   int64_t ver = pSrcBlock->info.version - 1;
 
-  if (pInfo->partitionSup.needCalc && ( srcStartTsCol[0] != srcEndTsCol[0] || (hasPrimaryKey(pInfo) && mode == STREAM_DELETE_DATA) )) {
+  if (pInfo->partitionSup.needCalc && ( srcStartTsCol[0] != srcEndTsCol[0] || (hasPrimaryKeyCol(pInfo) && mode == STREAM_DELETE_DATA) )) {
     getPreVersionDataBlock(srcUidData[0], srcStartTsCol[0], srcEndTsCol[0], ver, GET_TASKID(pTaskInfo), pInfo, pSrcBlock);
     srcStartTsCol = (TSKEY*)pSrcStartTsCol->pData;
     srcEndTsCol = (TSKEY*)pSrcEndTsCol->pData;
@@ -1824,7 +1824,7 @@ static int32_t generateIntervalScanRange(SStreamScanInfo* pInfo, SSDataBlock* pS
     uint64_t groupId = srcGp[i];
     if (groupId == 0) {
       void* pVal = NULL;
-      if (hasPrimaryKey(pInfo) && pSrcPkCol) {
+      if (hasPrimaryKeyCol(pInfo) && pSrcPkCol) {
         pVal = colDataGetData(pSrcPkCol, i);
       }
       groupId = getGroupIdByData(pInfo, srcUid, srcStartTsCol[i], ver, pVal);
@@ -1915,7 +1915,7 @@ static int32_t generateDeleteResultBlockImpl(SStreamScanInfo* pInfo, SSDataBlock
     char     tbname[VARSTR_HEADER_SIZE + TSDB_TABLE_NAME_LEN] = {0};
     if (groupId == 0) {
       void* pVal = NULL;
-      if (hasPrimaryKey(pInfo) && pSrcPkCol) {
+      if (hasPrimaryKeyCol(pInfo) && pSrcPkCol) {
         pVal = colDataGetData(pSrcPkCol, i);
       }
       groupId = getGroupIdByData(pInfo, srcUid, srcStartTsCol[i], ver, pVal);
@@ -1979,9 +1979,9 @@ void appendDataToSpecialBlock(SSDataBlock* pBlock, TSKEY* pStartTs, TSKEY* pEndT
   appendOneRowToSpecialBlockImpl(pBlock, pStartTs, pEndTs, pStartTs, pEndTs, pUid, pGp, pTbName, NULL);
 }
 
-bool checkExpiredData(SStateStore* pAPI, SUpdateInfo* pUpdateInfo, STimeWindowAggSupp* pTwSup, uint64_t tableId, TSKEY ts) {
+bool checkExpiredData(SStateStore* pAPI, SUpdateInfo* pUpdateInfo, STimeWindowAggSupp* pTwSup, uint64_t tableId, TSKEY ts, void* pPkVal, int32_t len) {
   bool isExpired = false;
-  bool isInc = pAPI->isIncrementalTimeStamp(pUpdateInfo, tableId, ts);
+  bool isInc = pAPI->isIncrementalTimeStamp(pUpdateInfo, tableId, ts, pPkVal, len);
   if (!isInc) {
     isExpired = isOverdue(ts, pTwSup);
   }
@@ -1997,7 +1997,7 @@ static void checkUpdateData(SStreamScanInfo* pInfo, bool invertible, SSDataBlock
   ASSERT(pColDataInfo->info.type == TSDB_DATA_TYPE_TIMESTAMP);
   TSKEY* tsCol = (TSKEY*)pColDataInfo->pData;
   SColumnInfoData* pPkColDataInfo = NULL;
-  if (hasPrimaryKey(pInfo)) {
+  if (hasPrimaryKeyCol(pInfo)) {
     pPkColDataInfo = taosArrayGet(pBlock->pDataBlock, pInfo->primaryKeyIndex);
   }
 
@@ -2017,7 +2017,13 @@ static void checkUpdateData(SStreamScanInfo* pInfo, bool invertible, SSDataBlock
       isClosed = isCloseWindow(&win, &pInfo->twAggSup);
     }
     // must check update info first.
-    bool update = pInfo->stateStore.updateInfoIsUpdated(pInfo->pUpdateInfo, pBlock->info.id.uid, tsCol[rowId]);
+    void* pPkVal = NULL;
+    int32_t pkLen = 0;
+    if (hasPrimaryKeyCol(pInfo)) {
+      pPkVal = colDataGetData(pPkColDataInfo, rowId);
+      pkLen = colDataGetRowLength(pPkColDataInfo, rowId);
+    }
+    bool update = pInfo->stateStore.updateInfoIsUpdated(pInfo->pUpdateInfo, pBlock->info.id.uid, tsCol[rowId], pPkVal, pkLen);
     bool isDeleted = isClosed && isSignleIntervalWindow(pInfo) &&
                      isDeletedStreamWindow(&win, pBlock->info.id.groupId, pInfo->pState, &pInfo->twAggSup, &pInfo->stateStore);
     if ((update || isDeleted) && out) {
@@ -2517,7 +2523,7 @@ static SSDataBlock* doStreamScan(SOperatorInfo* pOperator) {
     if (pInfo->pRecoverRes != NULL) {
       calBlockTbName(pInfo, pInfo->pRecoverRes, 0);
       if (!pInfo->igCheckUpdate && pInfo->pUpdateInfo) {
-        TSKEY maxTs = pAPI->stateStore.updateInfoFillBlockData(pInfo->pUpdateInfo, pInfo->pRecoverRes, pInfo->primaryTsIndex);
+        TSKEY maxTs = pAPI->stateStore.updateInfoFillBlockData(pInfo->pUpdateInfo, pInfo->pRecoverRes, pInfo->primaryTsIndex, pInfo->primaryKeyIndex);
         pInfo->twAggSup.maxTs = TMAX(pInfo->twAggSup.maxTs, maxTs);
       }
       if (pInfo->pCreateTbRes->info.rows > 0) {
@@ -3202,8 +3208,10 @@ SOperatorInfo* createStreamScanOperatorInfo(SReadHandle* pHandle, STableScanPhys
   pInfo->pDeleteDataRes = createSpecialDataBlock(STREAM_DELETE_DATA);
   pInfo->updateWin = (STimeWindow){.skey = INT64_MAX, .ekey = INT64_MAX};
   pInfo->pUpdateDataRes = createSpecialDataBlock(STREAM_CLEAR);
-  if (hasPrimaryKey(pInfo)) {
+  if (hasPrimaryKeyCol(pInfo)) {
     addPrimaryKeyCol(pInfo->pUpdateDataRes, pkType.type, pkType.bytes);
+    pInfo->pkColType = pkType.type;
+    pInfo->pkColLen = pkType.bytes;
   }
   pInfo->assignBlockUid = pTableScanNode->assignBlockUid;
   pInfo->partitionSup.needCalc = false;

source/libs/executor/src/streamcountwindowoperator.c

@@ -50,12 +50,13 @@ void destroyStreamCountAggOperatorInfo(void* param) {
   destroyStreamAggSupporter(&pInfo->streamAggSup);
   cleanupExprSupp(&pInfo->scalarSupp);
   clearGroupResInfo(&pInfo->groupResInfo);
+  taosArrayDestroyP(pInfo->pUpdated, destroyFlusedPos);
+  pInfo->pUpdated = NULL;
 
   colDataDestroy(&pInfo->twAggSup.timeWindowData);
   blockDataDestroy(pInfo->pDelRes);
   tSimpleHashCleanup(pInfo->pStUpdated);
   tSimpleHashCleanup(pInfo->pStDeleted);
-  pInfo->pUpdated = taosArrayDestroy(pInfo->pUpdated);
   cleanupGroupResInfo(&pInfo->groupResInfo);
 
   taosArrayDestroy(pInfo->historyWins);
@@ -242,7 +243,7 @@ static void doStreamCountAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBl
   for (int32_t i = 0; i < rows;) {
     if (pInfo->ignoreExpiredData &&
         checkExpiredData(&pInfo->streamAggSup.stateStore, pInfo->streamAggSup.pUpdateInfo, &pInfo->twAggSup,
-                         pSDataBlock->info.id.uid, startTsCols[i])) {
+                         pSDataBlock->info.id.uid, startTsCols[i], NULL, 0)) {
       i++;
       continue;
     }
@@ -728,7 +729,7 @@ SOperatorInfo* createStreamCountAggOperatorInfo(SOperatorInfo* downstream, SPhys
   setOperatorStreamStateFn(pOperator, streamCountReleaseState, streamCountReloadState);
 
   if (downstream) {
-    initDownStream(downstream, &pInfo->streamAggSup, pOperator->operatorType, pInfo->primaryTsIndex, &pInfo->twAggSup);
+    initDownStream(downstream, &pInfo->streamAggSup, pOperator->operatorType, pInfo->primaryTsIndex, &pInfo->twAggSup, &pInfo->basic);
     code = appendDownstream(pOperator, &downstream, 1);
   }
   return pOperator;

source/libs/executor/src/streameventwindowoperator.c

@@ -46,6 +46,9 @@ void destroyStreamEventOperatorInfo(void* param) {
   cleanupBasicInfo(&pInfo->binfo);
   destroyStreamAggSupporter(&pInfo->streamAggSup);
   clearGroupResInfo(&pInfo->groupResInfo);
+  taosArrayDestroyP(pInfo->pUpdated, destroyFlusedPos);
+  pInfo->pUpdated = NULL;
+
   cleanupExprSupp(&pInfo->scalarSupp);
   if (pInfo->pChildren != NULL) {
     int32_t size = taosArrayGetSize(pInfo->pChildren);
@@ -60,7 +63,6 @@ void destroyStreamEventOperatorInfo(void* param) {
   tSimpleHashCleanup(pInfo->pSeUpdated);
   tSimpleHashCleanup(pInfo->pAllUpdated);
   tSimpleHashCleanup(pInfo->pSeDeleted);
-  pInfo->pUpdated = taosArrayDestroy(pInfo->pUpdated);
   cleanupGroupResInfo(&pInfo->groupResInfo);
 
   taosArrayDestroy(pInfo->historyWins);
@@ -310,7 +312,7 @@ static void doStreamEventAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBl
   blockDataEnsureCapacity(pAggSup->pScanBlock, rows);
   for (int32_t i = 0; i < rows; i += winRows) {
     if (pInfo->ignoreExpiredData && checkExpiredData(&pInfo->streamAggSup.stateStore, pInfo->streamAggSup.pUpdateInfo,
-                                                     &pInfo->twAggSup, pSDataBlock->info.id.uid, tsCols[i])) {
+                                                     &pInfo->twAggSup, pSDataBlock->info.id.uid, tsCols[i], NULL, 0)) {
       i++;
       continue;
     }
@@ -776,7 +778,7 @@ SOperatorInfo* createStreamEventAggOperatorInfo(SOperatorInfo* downstream, SPhys
   pOperator->fpSet = createOperatorFpSet(optrDummyOpenFn, doStreamEventAgg, NULL, destroyStreamEventOperatorInfo,
                                          optrDefaultBufFn, NULL, optrDefaultGetNextExtFn, NULL);
   setOperatorStreamStateFn(pOperator, streamEventReleaseState, streamEventReloadState);
-  initDownStream(downstream, &pInfo->streamAggSup, pOperator->operatorType, pInfo->primaryTsIndex, &pInfo->twAggSup);
+  initDownStream(downstream, &pInfo->streamAggSup, pOperator->operatorType, pInfo->primaryTsIndex, &pInfo->twAggSup, &pInfo->basic);
   code = appendDownstream(pOperator, &downstream, 1);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;

source/libs/executor/src/streamtimewindowoperator.c

@@ -388,16 +388,21 @@ static void doBuildDeleteResult(SStreamIntervalOperatorInfo* pInfo, SArray* pWin
   }
 }
 
-void clearGroupResInfo(SGroupResInfo* pGroupResInfo) {
+void destroyFlusedPos(void* pRes) {
-  if (pGroupResInfo->freeItem) {
+  SRowBuffPos* pPos = (SRowBuffPos*) pRes;
-    int32_t size = taosArrayGetSize(pGroupResInfo->pRows);
-    for (int32_t i = pGroupResInfo->index; i < size; i++) {
-      SRowBuffPos* pPos = taosArrayGetP(pGroupResInfo->pRows, i);
   if (!pPos->needFree && !pPos->pRowBuff) {
     taosMemoryFreeClear(pPos->pKey);
     taosMemoryFree(pPos);
   }
 }
+
+void clearGroupResInfo(SGroupResInfo* pGroupResInfo) {
+  if (pGroupResInfo->freeItem) {
+    int32_t size = taosArrayGetSize(pGroupResInfo->pRows);
+    for (int32_t i = pGroupResInfo->index; i < size; i++) {
+      void* pPos = taosArrayGetP(pGroupResInfo->pRows, i);
+      destroyFlusedPos(pPos);
+    }
     pGroupResInfo->freeItem = false;
   }
   pGroupResInfo->pRows = taosArrayDestroy(pGroupResInfo->pRows);
@@ -409,6 +414,8 @@ void destroyStreamFinalIntervalOperatorInfo(void* param) {
   cleanupBasicInfo(&pInfo->binfo);
   cleanupAggSup(&pInfo->aggSup);
   clearGroupResInfo(&pInfo->groupResInfo);
+  taosArrayDestroyP(pInfo->pUpdated, destroyFlusedPos);
+  pInfo->pUpdated = NULL;
 
   // it should be empty.
   void* pIte = NULL;
@@ -437,7 +444,6 @@ void destroyStreamFinalIntervalOperatorInfo(void* param) {
   cleanupExprSupp(&pInfo->scalarSupp);
   tSimpleHashCleanup(pInfo->pUpdatedMap);
   pInfo->pUpdatedMap = NULL;
-  pInfo->pUpdated = taosArrayDestroy(pInfo->pUpdated);
   tSimpleHashCleanup(pInfo->pDeletedMap);
 
   blockDataDestroy(pInfo->pCheckpointRes);
@@ -481,13 +487,14 @@ void initIntervalDownStream(SOperatorInfo* downstream, uint16_t type, SStreamInt
   pScanInfo->windowSup.pIntervalAggSup = &pInfo->aggSup;
   if (!pScanInfo->pUpdateInfo) {
     pScanInfo->pUpdateInfo =
-        pAPI->updateInfoInitP(&pInfo->interval, pInfo->twAggSup.waterMark, pScanInfo->igCheckUpdate);
+        pAPI->updateInfoInitP(&pInfo->interval, pInfo->twAggSup.waterMark, pScanInfo->igCheckUpdate, pScanInfo->pkColType, pScanInfo->pkColLen);
   }
 
   pScanInfo->interval = pInfo->interval;
   pScanInfo->twAggSup = pInfo->twAggSup;
   pScanInfo->pState = pInfo->pState;
   pInfo->pUpdateInfo = pScanInfo->pUpdateInfo;
+  pInfo->basic.primaryPkIndex = pScanInfo->primaryKeyIndex;
 }
 
 void compactFunctions(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx, int32_t numOfOutput,
@@ -820,6 +827,10 @@ static int32_t getNextQualifiedFinalWindow(SInterval* pInterval, STimeWindow* pN
   return startPos;
 }
 
+bool hasSrcPrimaryKeyCol(SSteamOpBasicInfo* pInfo) {
+  return pInfo->primaryPkIndex != -1;
+}
+
 static void doStreamIntervalAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBlock, uint64_t groupId,
                                     SSHashObj* pUpdatedMap, SSHashObj* pDeletedMap) {
   SStreamIntervalOperatorInfo* pInfo = (SStreamIntervalOperatorInfo*)pOperator->info;
@@ -839,6 +850,13 @@ static void doStreamIntervalAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDat
   SColumnInfoData* pColDataInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->primaryTsIndex);
   tsCols = (int64_t*)pColDataInfo->pData;
 
+  void* pPkVal = NULL;
+  int32_t pkLen = 0;
+  SColumnInfoData* pPkColDataInfo = NULL;
+  if (hasSrcPrimaryKeyCol(&pInfo->basic)) {
+    pPkColDataInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->primaryTsIndex);
+  }
+
   if (pSDataBlock->info.window.skey != tsCols[0] || pSDataBlock->info.window.ekey != tsCols[endRowId]) {
     qError("table uid %" PRIu64 " data block timestamp range may not be calculated! minKey %" PRId64
             ",maxKey %" PRId64,
@@ -862,9 +880,15 @@ static void doStreamIntervalAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDat
   }
   while (1) {
     bool isClosed = isCloseWindow(&nextWin, &pInfo->twAggSup);
+    if (hasSrcPrimaryKeyCol(&pInfo->basic) && !IS_FINAL_INTERVAL_OP(pOperator) && pInfo->ignoreExpiredData &&
+        pSDataBlock->info.type != STREAM_PULL_DATA) {
+      pPkVal = colDataGetData(pPkColDataInfo, startPos);
+      pkLen = colDataGetRowLength(pPkColDataInfo, startPos);
+    }
+
     if ((!IS_FINAL_INTERVAL_OP(pOperator) && pInfo->ignoreExpiredData && pSDataBlock->info.type != STREAM_PULL_DATA &&
          checkExpiredData(&pInfo->stateStore, pInfo->pUpdateInfo, &pInfo->twAggSup, pSDataBlock->info.id.uid,
-                          nextWin.ekey)) ||
+                          nextWin.ekey, pPkVal, pkLen)) ||
         !inSlidingWindow(&pInfo->interval, &nextWin, &pSDataBlock->info)) {
       startPos = getNexWindowPos(&pInfo->interval, &pSDataBlock->info, tsCols, startPos, nextWin.ekey, &nextWin);
       if (startPos < 0) {
@@ -1664,6 +1688,8 @@ void destroyStreamSessionAggOperatorInfo(void* param) {
   destroyStreamAggSupporter(&pInfo->streamAggSup);
   cleanupExprSupp(&pInfo->scalarSupp);
   clearGroupResInfo(&pInfo->groupResInfo);
+  taosArrayDestroyP(pInfo->pUpdated, destroyFlusedPos);
+  pInfo->pUpdated = NULL;
 
   if (pInfo->pChildren != NULL) {
     int32_t size = taosArrayGetSize(pInfo->pChildren);
@@ -1679,7 +1705,6 @@ void destroyStreamSessionAggOperatorInfo(void* param) {
   blockDataDestroy(pInfo->pWinBlock);
   tSimpleHashCleanup(pInfo->pStUpdated);
   tSimpleHashCleanup(pInfo->pStDeleted);
-  pInfo->pUpdated = taosArrayDestroy(pInfo->pUpdated);
   cleanupGroupResInfo(&pInfo->groupResInfo);
 
   taosArrayDestroy(pInfo->historyWins);
@@ -1715,14 +1740,14 @@ void initDummyFunction(SqlFunctionCtx* pDummy, SqlFunctionCtx* pCtx, int32_t num
 }
 
 void initDownStream(SOperatorInfo* downstream, SStreamAggSupporter* pAggSup, uint16_t type, int32_t tsColIndex,
-                    STimeWindowAggSupp* pTwSup) {
+                    STimeWindowAggSupp* pTwSup, struct SSteamOpBasicInfo* pBasic) {
   if (downstream->operatorType == QUERY_NODE_PHYSICAL_PLAN_STREAM_PARTITION) {
     SStreamPartitionOperatorInfo* pScanInfo = downstream->info;
     pScanInfo->tsColIndex = tsColIndex;
   }
 
   if (downstream->operatorType != QUERY_NODE_PHYSICAL_PLAN_STREAM_SCAN) {
-    initDownStream(downstream->pDownstream[0], pAggSup, type, tsColIndex, pTwSup);
+    initDownStream(downstream->pDownstream[0], pAggSup, type, tsColIndex, pTwSup, pBasic);
     return;
   }
   SStreamScanInfo* pScanInfo = downstream->info;
@@ -1730,10 +1755,11 @@ void initDownStream(SOperatorInfo* downstream, SStreamAggSupporter* pAggSup, uin
   pScanInfo->pState = pAggSup->pState;
   if (!pScanInfo->pUpdateInfo) {
     pScanInfo->pUpdateInfo = pAggSup->stateStore.updateInfoInit(60000, TSDB_TIME_PRECISION_MILLI, pTwSup->waterMark,
-                                                                pScanInfo->igCheckUpdate);
+                                                                pScanInfo->igCheckUpdate, pScanInfo->pkColType, pScanInfo->pkColLen);
   }
   pScanInfo->twAggSup = *pTwSup;
   pAggSup->pUpdateInfo = pScanInfo->pUpdateInfo;
+  pBasic->primaryPkIndex = pScanInfo->primaryKeyIndex;
 }
 
 static TSKEY sesionTs(void* pKey) {
@@ -2106,10 +2132,22 @@ static void doStreamSessionAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSData
   }
 
   TSKEY* endTsCols = (int64_t*)pEndTsCol->pData;
+
+  void* pPkVal = NULL;
+  int32_t pkLen = 0;
+  SColumnInfoData* pPkColDataInfo = NULL;
+  if (hasSrcPrimaryKeyCol(&pInfo->basic)) {
+    pPkColDataInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->primaryTsIndex);
+  }
+
   for (int32_t i = 0; i < rows;) {
+    if (hasSrcPrimaryKeyCol(&pInfo->basic) && !IS_FINAL_SESSION_OP(pOperator) && pInfo->ignoreExpiredData) {
+      pPkVal = colDataGetData(pPkColDataInfo, i);
+      pkLen = colDataGetRowLength(pPkColDataInfo, i);
+    }
     if (!IS_FINAL_SESSION_OP(pOperator) && pInfo->ignoreExpiredData &&
         checkExpiredData(&pInfo->streamAggSup.stateStore, pInfo->streamAggSup.pUpdateInfo, &pInfo->twAggSup,
-                         pSDataBlock->info.id.uid, endTsCols[i])) {
+                         pSDataBlock->info.id.uid, endTsCols[i], pPkVal, pkLen)) {
       i++;
       continue;
     }
@@ -3051,7 +3089,7 @@ SOperatorInfo* createStreamSessionAggOperatorInfo(SOperatorInfo* downstream, SPh
   setOperatorStreamStateFn(pOperator, streamSessionReleaseState, streamSessionReloadState);
 
   if (downstream) {
-    initDownStream(downstream, &pInfo->streamAggSup, pOperator->operatorType, pInfo->primaryTsIndex, &pInfo->twAggSup);
+    initDownStream(downstream, &pInfo->streamAggSup, pOperator->operatorType, pInfo->primaryTsIndex, &pInfo->twAggSup, &pInfo->basic);
     code = appendDownstream(pOperator, &downstream, 1);
   }
   return pOperator;
@@ -3250,6 +3288,9 @@ void destroyStreamStateOperatorInfo(void* param) {
   cleanupBasicInfo(&pInfo->binfo);
   destroyStreamAggSupporter(&pInfo->streamAggSup);
   clearGroupResInfo(&pInfo->groupResInfo);
+  taosArrayDestroyP(pInfo->pUpdated, destroyFlusedPos);
+  pInfo->pUpdated = NULL;
+
   cleanupExprSupp(&pInfo->scalarSupp);
   if (pInfo->pChildren != NULL) {
     int32_t size = taosArrayGetSize(pInfo->pChildren);
@@ -3263,7 +3304,6 @@ void destroyStreamStateOperatorInfo(void* param) {
   blockDataDestroy(pInfo->pDelRes);
   tSimpleHashCleanup(pInfo->pSeUpdated);
   tSimpleHashCleanup(pInfo->pSeDeleted);
-  pInfo->pUpdated = taosArrayDestroy(pInfo->pUpdated);
   cleanupGroupResInfo(&pInfo->groupResInfo);
 
   taosArrayDestroy(pInfo->historyWins);
@@ -3481,7 +3521,7 @@ static void doStreamStateAggImpl(SOperatorInfo* pOperator, SSDataBlock* pSDataBl
   SColumnInfoData* pKeyColInfo = taosArrayGet(pSDataBlock->pDataBlock, pInfo->stateCol.slotId);
   for (int32_t i = 0; i < rows; i += winRows) {
     if (pInfo->ignoreExpiredData && checkExpiredData(&pInfo->streamAggSup.stateStore, pInfo->streamAggSup.pUpdateInfo,
-                                                     &pInfo->twAggSup, pSDataBlock->info.id.uid, tsCols[i]) ||
+                                                     &pInfo->twAggSup, pSDataBlock->info.id.uid, tsCols[i], NULL, 0) ||
         colDataIsNull_s(pKeyColInfo, i)) {
       i++;
       continue;
@@ -3948,7 +3988,7 @@ SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream, SPhys
   pOperator->fpSet = createOperatorFpSet(optrDummyOpenFn, doStreamStateAgg, NULL, destroyStreamStateOperatorInfo,
                                          optrDefaultBufFn, NULL, optrDefaultGetNextExtFn, NULL);
   setOperatorStreamStateFn(pOperator, streamStateReleaseState, streamStateReloadState);
-  initDownStream(downstream, &pInfo->streamAggSup, pOperator->operatorType, pInfo->primaryTsIndex, &pInfo->twAggSup);
+  initDownStream(downstream, &pInfo->streamAggSup, pOperator->operatorType, pInfo->primaryTsIndex, &pInfo->twAggSup, &pInfo->basic);
   code = appendDownstream(pOperator, &downstream, 1);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;

source/libs/executor/src/sysscanoperator.c

@@ -720,7 +720,6 @@ static SSDataBlock* sysTableScanUserTags(SOperatorInfo* pOperator) {
     pAPI->metaFn.resumeTableMetaCursor(pInfo->pCur, 0, 0);
   }
 
-  bool blockFull = false;
   while ((ret = pAPI->metaFn.cursorNext(pInfo->pCur, TSDB_SUPER_TABLE)) == 0) {
     if (pInfo->pCur->mr.me.type != TSDB_CHILD_TABLE) {
       continue;
@@ -743,25 +742,19 @@ static SSDataBlock* sysTableScanUserTags(SOperatorInfo* pOperator) {
     }
 
     if ((smrSuperTable.me.stbEntry.schemaTag.nCols + numOfRows) > pOperator->resultInfo.capacity) {
-      blockFull = true;
-    } else {
-      sysTableUserTagsFillOneTableTags(pInfo, &smrSuperTable, &pInfo->pCur->mr, dbname, tableName, &numOfRows,
-                                       dataBlock);
-    }
-
-    pAPI->metaReaderFn.clearReader(&smrSuperTable);
-
-    if (blockFull || numOfRows >= pOperator->resultInfo.capacity) {
       relocateAndFilterSysTagsScanResult(pInfo, numOfRows, dataBlock, pOperator->exprSupp.pFilterInfo);
       numOfRows = 0;
 
       if (pInfo->pRes->info.rows > 0) {
         pAPI->metaFn.pauseTableMetaCursor(pInfo->pCur);
+        pAPI->metaReaderFn.clearReader(&smrSuperTable);
         break;
       }
-
+    } else {
-      blockFull = false;
+      sysTableUserTagsFillOneTableTags(pInfo, &smrSuperTable, &pInfo->pCur->mr, dbname, tableName, &numOfRows,
+                                       dataBlock);
     }
+    pAPI->metaReaderFn.clearReader(&smrSuperTable);
   }
 
   if (numOfRows > 0) {

source/libs/function/src/builtinsimpl.c

@@ -2885,6 +2885,7 @@ static int32_t firstLastFunctionMergeImpl(SqlFunctionCtx* pCtx, bool isFirstQuer
     } else {
       pInputInfo->pkData = NULL;
     }
+
     int32_t code = firstLastTransferInfo(pCtx, pInputInfo, pInfo, isFirstQuery, i);
     if (code != TSDB_CODE_SUCCESS) {
       return code;

source/libs/nodes/src/nodesCloneFuncs.c

@@ -300,6 +300,9 @@ static int32_t joinTableNodeCopy(const SJoinTableNode* pSrc, SJoinTableNode* pDs
   COPY_BASE_OBJECT_FIELD(table, tableNodeCopy);
   COPY_SCALAR_FIELD(joinType);
   COPY_SCALAR_FIELD(subType);
+  CLONE_NODE_FIELD(pWindowOffset);
+  CLONE_NODE_FIELD(pJLimit);
+  CLONE_NODE_FIELD(addPrimCond);
   COPY_SCALAR_FIELD(hasSubQuery);
   COPY_SCALAR_FIELD(isLowLevelJoin);
   CLONE_NODE_FIELD(pLeft);

source/libs/nodes/src/nodesUtilFuncs.c

@@ -2343,7 +2343,7 @@ static EDealRes collectFuncs(SNode* pNode, void* pContext) {
         return DEAL_RES_CONTINUE;
       }
     }
-    SExprNode* pExpr = (SExprNode*)pNode;
+
     bool bFound = false;
     SNode* pn = NULL;
     FOREACH(pn, pCxt->pFuncs) {

source/libs/parser/src/parAstCreater.c

@@ -745,7 +745,6 @@ SNode* createTimeOffsetValueNode(SAstCreateContext* pCxt, const SToken* pLiteral
   return (SNode*)val;
 }
 
-
 SNode* createDefaultDatabaseCondValue(SAstCreateContext* pCxt) {
   CHECK_PARSER_STATUS(pCxt);
   if (NULL == pCxt->pQueryCxt->db) {
@@ -965,7 +964,8 @@ SNode* createTempTableNode(SAstCreateContext* pCxt, SNode* pSubquery, const STok
   return (SNode*)tempTable;
 }
 
-SNode* createJoinTableNode(SAstCreateContext* pCxt, EJoinType type, EJoinSubType stype, SNode* pLeft, SNode* pRight, SNode* pJoinCond) {
+SNode* createJoinTableNode(SAstCreateContext* pCxt, EJoinType type, EJoinSubType stype, SNode* pLeft, SNode* pRight,
+                           SNode* pJoinCond) {
   CHECK_PARSER_STATUS(pCxt);
   SJoinTableNode* joinTable = (SJoinTableNode*)nodesMakeNode(QUERY_NODE_JOIN_TABLE);
   CHECK_OUT_OF_MEM(joinTable);
@@ -1264,7 +1264,6 @@ SNode* addFillClause(SAstCreateContext* pCxt, SNode* pStmt, SNode* pFill) {
   return pStmt;
 }
 
-
 SNode* addJLimitClause(SAstCreateContext* pCxt, SNode* pJoin, SNode* pJLimit) {
   CHECK_PARSER_STATUS(pCxt);
   if (NULL == pJLimit) {
@@ -1276,7 +1275,6 @@ SNode* addJLimitClause(SAstCreateContext* pCxt, SNode* pJoin, SNode* pJLimit) {
   return pJoin;
 }
 
-
 SNode* addWindowOffsetClause(SAstCreateContext* pCxt, SNode* pJoin, SNode* pWinOffset) {
   CHECK_PARSER_STATUS(pCxt);
   if (NULL == pWinOffset) {
@@ -1288,7 +1286,6 @@ SNode* addWindowOffsetClause(SAstCreateContext* pCxt, SNode* pJoin, SNode* pWinO
   return pJoin;
 }
 
-
 SNode* createSelectStmt(SAstCreateContext* pCxt, bool isDistinct, SNodeList* pProjectionList, SNode* pTable,
                         SNodeList* pHint) {
   CHECK_PARSER_STATUS(pCxt);
@@ -1744,14 +1741,14 @@ SNode* setColumnOptions(SAstCreateContext* pCxt, SNode* pOptions, EColumnOptionT
       memset(((SColumnOptions*)pOptions)->compress, 0, TSDB_CL_COMPRESS_OPTION_LEN);
       COPY_STRING_FORM_STR_TOKEN(((SColumnOptions*)pOptions)->compress, (SToken*)pVal);
       if (0 == strlen(((SColumnOptions*)pOptions)->compress)) {
-        pCxt->errCode = TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+        pCxt->errCode = TSDB_CODE_TSC_COMPRESS_PARAM_ERROR;
       }
       break;
     case COLUMN_OPTION_LEVEL:
       memset(((SColumnOptions*)pOptions)->compressLevel, 0, TSDB_CL_COMPRESS_OPTION_LEN);
       COPY_STRING_FORM_STR_TOKEN(((SColumnOptions*)pOptions)->compressLevel, (SToken*)pVal);
       if (0 == strlen(((SColumnOptions*)pOptions)->compressLevel)) {
-        pCxt->errCode = TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+        pCxt->errCode = TSDB_CODE_TSC_COMPRESS_LEVEL_ERROR;
       }
       break;
     case COLUMN_OPTION_PRIMARYKEY:
@@ -1895,8 +1892,8 @@ SNode* createAlterTableAddModifyCol(SAstCreateContext* pCxt, SNode* pRealTable,
   return createAlterTableStmtFinalize(pRealTable, pStmt);
 }
 
-SNode* createAlterTableAddModifyColOptions(SAstCreateContext* pCxt, SNode* pRealTable, int8_t alterType, SToken* pColName,
+SNode* createAlterTableAddModifyColOptions(SAstCreateContext* pCxt, SNode* pRealTable, int8_t alterType,
-                                    SNode* pOptions) {
+                                           SToken* pColName, SNode* pOptions) {
   CHECK_PARSER_STATUS(pCxt);
   if (!checkColumnName(pCxt, pColName)) {
     return NULL;

source/libs/parser/src/parTranslater.c

@@ -4394,7 +4394,6 @@ int32_t translateTable(STranslateContext* pCxt, SNode** pTable, SNode* pJoinPare
     }
     case QUERY_NODE_JOIN_TABLE: {
       SJoinTableNode* pJoinTable = (SJoinTableNode*)*pTable;
-      pJoinTable->pParent = pJoinParent;
       code = translateJoinTable(pCxt, pJoinTable);
       if (TSDB_CODE_SUCCESS == code) {
         code = translateTable(pCxt, &pJoinTable->pLeft, (SNode*)pJoinTable);
@@ -7225,9 +7224,9 @@ static int32_t checkColumnOptions(SNodeList* pList) {
     if (!checkColumnEncodeOrSetDefault(pCol->dataType.type, ((SColumnOptions*)pCol->pOptions)->encode))
       return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
     if (!checkColumnCompressOrSetDefault(pCol->dataType.type, ((SColumnOptions*)pCol->pOptions)->compress))
-      return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+      return TSDB_CODE_TSC_COMPRESS_PARAM_ERROR;
     if (!checkColumnLevelOrSetDefault(pCol->dataType.type, ((SColumnOptions*)pCol->pOptions)->compressLevel))
-      return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+      return TSDB_CODE_TSC_COMPRESS_LEVEL_ERROR;
   }
   return TSDB_CODE_SUCCESS;
 }
@@ -8118,13 +8117,15 @@ static int32_t buildAlterSuperTableReq(STranslateContext* pCxt, SAlterTableStmt*
       TAOS_FIELD field = {0};
       strcpy(field.name, pStmt->colName);
       if (!checkColumnEncode(pStmt->pColOptions->encode)) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
-      if (!checkColumnCompress(pStmt->pColOptions->compress)) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+      if (!checkColumnCompress(pStmt->pColOptions->compress)) return TSDB_CODE_TSC_COMPRESS_PARAM_ERROR;
-      if (!checkColumnLevel(pStmt->pColOptions->compressLevel)) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+      if (!checkColumnLevel(pStmt->pColOptions->compressLevel)) return TSDB_CODE_TSC_COMPRESS_LEVEL_ERROR;
-      int8_t valid =
+      int32_t code =
           setColCompressByOption(pStmt->dataType.type, columnEncodeVal(pStmt->pColOptions->encode),
                                  columnCompressVal(pStmt->pColOptions->compress),
                                  columnLevelVal(pStmt->pColOptions->compressLevel), false, (uint32_t*)&field.bytes);
-      if (!valid) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+      if (code != TSDB_CODE_SUCCESS) {
+        return code;
+      }
       taosArrayPush(pAlterReq->pFields, &field);
       break;
     }
@@ -10726,7 +10727,8 @@ static int32_t deduplicateTsmaFuncs(SNodeList* pFuncs) {
   return TSDB_CODE_SUCCESS;
 }
 
-static int32_t buildTSMAAstStreamSubTable(SCreateTSMAStmt* pStmt, SMCreateSmaReq* pReq, const SNode* pTbname, SNode** pSubTable) {
+static int32_t buildTSMAAstStreamSubTable(SCreateTSMAStmt* pStmt, SMCreateSmaReq* pReq, const SNode* pTbname,
+                                          SNode** pSubTable) {
   int32_t        code = 0;
   SFunctionNode* pMd5Func = (SFunctionNode*)nodesMakeNode(QUERY_NODE_FUNCTION);
   SFunctionNode* pConcatFunc = (SFunctionNode*)nodesMakeNode(QUERY_NODE_FUNCTION);
@@ -10768,8 +10770,8 @@ _end:
   return code;
 }
 
-static int32_t buildTSMAAst(STranslateContext* pCxt, SCreateTSMAStmt* pStmt, SMCreateSmaReq* pReq,
+static int32_t buildTSMAAst(STranslateContext* pCxt, SCreateTSMAStmt* pStmt, SMCreateSmaReq* pReq, const char* tbName,
-                            const char* tbName, int32_t numOfTags, const SSchema* pTags) {
+                            int32_t numOfTags, const SSchema* pTags) {
   int32_t        code = TSDB_CODE_SUCCESS;
   SSampleAstInfo info = {0};
   info.createSmaIndex = true;
@@ -10813,16 +10815,17 @@ static int32_t buildTSMAAst(STranslateContext* pCxt, SCreateTSMAStmt* pStmt, SMC
         if (!pTagCol) code = TSDB_CODE_OUT_OF_MEMORY;
       }
       if (code == TSDB_CODE_SUCCESS) {
-        code = buildTSMAAstStreamSubTable(pStmt, pReq, pStmt->pOptions->recursiveTsma ? pTagCol : (SNode*)pTbnameFunc, (SNode**)&pSubTable);
+        code = buildTSMAAstStreamSubTable(pStmt, pReq, pStmt->pOptions->recursiveTsma ? pTagCol : (SNode*)pTbnameFunc,
+                                          (SNode**)&pSubTable);
         info.pSubTable = (SNode*)pSubTable;
       }
       if (code == TSDB_CODE_SUCCESS)
-        code = nodesListMakeStrictAppend(&info.pTags, pStmt->pOptions->recursiveTsma ? pTagCol : nodesCloneNode((SNode*)pTbnameFunc));
+        code = nodesListMakeStrictAppend(
+            &info.pTags, pStmt->pOptions->recursiveTsma ? pTagCol : nodesCloneNode((SNode*)pTbnameFunc));
     }
   }
 
-  if (code == TSDB_CODE_SUCCESS && !pStmt->pOptions->recursiveTsma)
+  if (code == TSDB_CODE_SUCCESS && !pStmt->pOptions->recursiveTsma) code = fmCreateStateFuncs(info.pFuncs);
-    code = fmCreateStateFuncs(info.pFuncs);
 
   if (code == TSDB_CODE_SUCCESS) {
     int32_t pProjectionTotalLen = 0;
@@ -10914,7 +10917,8 @@ static int32_t rewriteTSMAFuncs(STranslateContext* pCxt, SCreateTSMAStmt* pStmt,
   return code;
 }
 
-static int32_t buildCreateTSMAReq(STranslateContext* pCxt, SCreateTSMAStmt* pStmt, SMCreateSmaReq* pReq, SName* useTbName) {
+static int32_t buildCreateTSMAReq(STranslateContext* pCxt, SCreateTSMAStmt* pStmt, SMCreateSmaReq* pReq,
+                                  SName* useTbName) {
   SName name;
   tNameExtractFullName(toName(pCxt->pParseCxt->acctId, pStmt->dbName, pStmt->tsmaName, &name), pReq->name);
   memset(&name, 0, sizeof(SName));
@@ -11063,7 +11067,8 @@ int32_t translatePostCreateTSMA(SParseContext* pParseCxt, SQuery* pQuery, SSData
 
   if (TSDB_CODE_SUCCESS == code) {
     if (interval.interval > 0) {
-      pStmt->pReq->lastTs = taosTimeAdd(taosTimeTruncate(lastTs, &interval), interval.interval, interval.intervalUnit, interval.precision);
+      pStmt->pReq->lastTs = taosTimeAdd(taosTimeTruncate(lastTs, &interval), interval.interval, interval.intervalUnit,
+                                        interval.precision);
     } else {
       pStmt->pReq->lastTs = lastTs + 1;  // start key of the next time window
     }
@@ -12033,13 +12038,13 @@ static int32_t buildNormalTableBatchReq(int32_t acctId, const SCreateTableStmt*
     toSchema(pColDef, index + 1, pScheam);
     if (pColDef->pOptions) {
       req.colCmpr.pColCmpr[index].id = index + 1;
-      int8_t valid = setColCompressByOption(
+      int32_t code = setColCompressByOption(
           pScheam->type, columnEncodeVal(((SColumnOptions*)pColDef->pOptions)->encode),
           columnCompressVal(((SColumnOptions*)pColDef->pOptions)->compress),
           columnLevelVal(((SColumnOptions*)pColDef->pOptions)->compressLevel), true, &req.colCmpr.pColCmpr[index].alg);
-      if (!valid) {
+      if (code != TSDB_CODE_SUCCESS) {
         tdDestroySVCreateTbReq(&req);
-        return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+        return code;
       }
     }
     ++index;
@@ -12499,7 +12504,6 @@ static int32_t buildDropTableVgroupHashmap(STranslateContext* pCxt, SDropTableCl
     goto over;
   }
 
-
   SVgroupInfo info = {0};
   if (TSDB_CODE_SUCCESS == code) {
     code = getTableHashVgroup(pCxt, pClause->dbName, pClause->tableName, &info);
@@ -12879,14 +12883,12 @@ static int buildAlterTableColumnCompress(STranslateContext* pCxt, SAlterTableStm
   }
 
   if (!checkColumnEncode(pStmt->pColOptions->encode)) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
-  if (!checkColumnCompress(pStmt->pColOptions->compress)) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+  if (!checkColumnCompress(pStmt->pColOptions->compress)) return TSDB_CODE_TSC_COMPRESS_PARAM_ERROR;
-  if (!checkColumnLevel(pStmt->pColOptions->compressLevel)) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+  if (!checkColumnLevel(pStmt->pColOptions->compressLevel)) return TSDB_CODE_TSC_COMPRESS_LEVEL_ERROR;
-  int8_t valid = setColCompressByOption(pSchema->type, columnEncodeVal(pStmt->pColOptions->encode),
+  int8_t code = setColCompressByOption(pSchema->type, columnEncodeVal(pStmt->pColOptions->encode),
                                        columnCompressVal(pStmt->pColOptions->compress),
                                        columnLevelVal(pStmt->pColOptions->compressLevel), true, &pReq->compress);
-  if (!valid) return TSDB_CODE_TSC_ENCODE_PARAM_ERROR;
+  return code;
-
-  return TSDB_CODE_SUCCESS;
 }
 
 static int32_t buildAlterTbReq(STranslateContext* pCxt, SAlterTableStmt* pStmt, STableMeta* pTableMeta,

source/libs/parser/src/parUtil.c

@@ -315,19 +315,15 @@ STableMeta* tableMetaDup(const STableMeta* pTableMeta) {
   size_t schemaExtSize = hasSchemaExt ? pTableMeta->tableInfo.numOfColumns * sizeof(SSchemaExt) : 0;
 
   size_t      size = sizeof(STableMeta) + numOfFields * sizeof(SSchema);
-  int32_t     cpSize = sizeof(STableMeta) - sizeof(void*);
   STableMeta* p = taosMemoryMalloc(size + schemaExtSize);
-
   if (NULL == p) return NULL;
 
-  memcpy(p, pTableMeta, cpSize);
+  memcpy(p, pTableMeta, schemaExtSize+size);
   if (hasSchemaExt) {
     p->schemaExt = (SSchemaExt*)(((char*)p) + size);
-    memcpy(p->schemaExt, pTableMeta->schemaExt, schemaExtSize);
   } else {
     p->schemaExt = NULL;
   }
-  memcpy(p->schema, pTableMeta->schema, numOfFields * sizeof(SSchema));
   return p;
 }
 

source/libs/planner/src/planLogicCreater.c

@@ -19,6 +19,9 @@
 #include "tglobal.h"
 #include "parser.h"
 
+// primary key column always the second column if exists
+#define PRIMARY_COLUMN_SLOT 1
+
 typedef struct SLogicPlanContext {
   SPlanContext* pPlanCxt;
   SLogicNode*   pCurrRoot;
@@ -304,7 +307,7 @@ static SNode* createFirstCol(SRealTableNode* pTable, const SSchema* pSchema) {
   return (SNode*)pCol;
 }
 
-static int32_t addPrimaryKeyCol(SRealTableNode* pTable, SNodeList** pCols) {
+static int32_t addPrimaryTsCol(SRealTableNode* pTable, SNodeList** pCols) {
   bool   found = false;
   SNode* pCol = NULL;
   FOREACH(pCol, *pCols) {
@@ -327,10 +330,10 @@ static int32_t addSystableFirstCol(SRealTableNode* pTable, SNodeList** pCols) {
   return nodesListMakeStrictAppend(pCols, createFirstCol(pTable, pTable->pMeta->schema));
 }
 
-static int32_t addPkCol(SRealTableNode* pTable, SNodeList** pCols) {
+static int32_t addPrimaryKeyCol(SRealTableNode* pTable, SNodeList** pCols) {
   bool   found = false;
   SNode* pCol = NULL;
-  SSchema* pSchema = &pTable->pMeta->schema[1];
+  SSchema* pSchema = &pTable->pMeta->schema[PRIMARY_COLUMN_SLOT];
   FOREACH(pCol, *pCols) {
     if (pSchema->colId == ((SColumnNode*)pCol)->colId) {
       found = true;
@@ -348,9 +351,9 @@ static int32_t addDefaultScanCol(SRealTableNode* pTable, SNodeList** pCols) {
   if (TSDB_SYSTEM_TABLE == pTable->pMeta->tableType) {
     return addSystableFirstCol(pTable, pCols);
   }
-  int32_t code = addPrimaryKeyCol(pTable, pCols);
+  int32_t code = addPrimaryTsCol(pTable, pCols);
   if (code == TSDB_CODE_SUCCESS && hasPkInTable(pTable->pMeta)) {
-    code = addPkCol(pTable, pCols);
+    code = addPrimaryKeyCol(pTable, pCols);
   }
   return code;
 }
@@ -1802,7 +1805,7 @@ static int32_t createDeleteScanLogicNode(SLogicPlanContext* pCxt, SDeleteStmt* p
 
   STableMeta* pMeta = ((SRealTableNode*)pDelete->pFromTable)->pMeta;
   if (TSDB_CODE_SUCCESS == code && hasPkInTable(pMeta)) {
-    code = addPkCol((SRealTableNode*)pDelete->pFromTable, &pScan->pScanCols);
+    code = addPrimaryKeyCol((SRealTableNode*)pDelete->pFromTable, &pScan->pScanCols);
   }
 
   if (TSDB_CODE_SUCCESS == code && NULL != pDelete->pTagCond) {

source/libs/planner/src/planOptimizer.c

@@ -3966,21 +3966,25 @@ static int32_t lastRowScanOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogic
   if (NULL != cxt.pLastCols) {
     cxt.doAgg = false;
     cxt.funcType = FUNCTION_TYPE_CACHE_LAST;
+
     lastRowScanOptSetLastTargets(pScan->pScanCols, cxt.pLastCols, pLastRowCols, true, cxt.pkBytes);
     nodesWalkExprs(pScan->pScanPseudoCols, lastRowScanOptSetColDataType, &cxt);
+
     lastRowScanOptSetLastTargets(pScan->node.pTargets, cxt.pLastCols, pLastRowCols, false, cxt.pkBytes);
     lastRowScanOptRemoveUslessTargets(pScan->node.pTargets, cxt.pLastCols, cxt.pOtherCols, pLastRowCols);
-    if (pPKTsCol && pScan->node.pTargets->length == 1) {
+    if (pPKTsCol && ((pScan->node.pTargets->length == 1) || (pScan->node.pTargets->length == 2 && cxt.pkBytes > 0))) {
       // when select last(ts),ts from ..., we add another ts to targets
       sprintf(pPKTsCol->colName, "#sel_val.%p", pPKTsCol);
       nodesListAppend(pScan->node.pTargets, nodesCloneNode((SNode*)pPKTsCol));
     }
+
     if (pNonPKCol && cxt.pLastCols->length == 1 &&
         nodesEqualNode((SNode*)pNonPKCol, nodesListGetNode(cxt.pLastCols, 0))) {
       // when select last(c1), c1 from ..., we add c1 to targets
       sprintf(pNonPKCol->colName, "#sel_val.%p", pNonPKCol);
       nodesListAppend(pScan->node.pTargets, nodesCloneNode((SNode*)pNonPKCol));
     }
+
     nodesClearList(cxt.pLastCols);
   }
   nodesClearList(cxt.pOtherCols);

source/libs/scheduler/src/schRemote.c

@@ -475,6 +475,8 @@ int32_t schHandleDropCallback(void *param, SDataBuf *pMsg, int32_t code) {
   SSchTaskCallbackParam *pParam = (SSchTaskCallbackParam *)param;
   qDebug("QID:0x%" PRIx64 ",TID:0x%" PRIx64 " drop task rsp received, code:0x%x", pParam->queryId, pParam->taskId,
          code);
+  // called if drop task rsp received code
+  rpcReleaseHandle(pMsg->handle, TAOS_CONN_CLIENT);
   if (pMsg) {
     taosMemoryFree(pMsg->pData);
     taosMemoryFree(pMsg->pEpSet);
@@ -486,7 +488,6 @@ int32_t schHandleNotifyCallback(void *param, SDataBuf *pMsg, int32_t code) {
   SSchTaskCallbackParam *pParam = (SSchTaskCallbackParam *)param;
   qDebug("QID:0x%" PRIx64 ",TID:0x%" PRIx64 " task notify rsp received, code:0x%x", pParam->queryId, pParam->taskId,
          code);
-  rpcReleaseHandle(pMsg->handle, TAOS_CONN_CLIENT);
   if (pMsg) {
     taosMemoryFree(pMsg->pData);
     taosMemoryFree(pMsg->pEpSet);

source/libs/stream/src/streamCheckpoint.c

@@ -94,6 +94,24 @@ int32_t tDecodeStreamCheckpointReadyMsg(SDecoder* pDecoder, SStreamCheckpointRea
   return 0;
 }
 
+int32_t tEncodeStreamTaskCheckpointReq(SEncoder* pEncoder, const SStreamTaskCheckpointReq* 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->nodeId) < 0) return -1;
+  tEndEncode(pEncoder);
+  return 0;
+}
+
+int32_t tDecodeStreamTaskCheckpointReq(SDecoder* pDecoder, SStreamTaskCheckpointReq* 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->nodeId) < 0) return -1;
+  tEndDecode(pDecoder);
+  return 0;
+}
+
 static int32_t streamAlignCheckpoint(SStreamTask* pTask) {
   int32_t num = taosArrayGetSize(pTask->upstreamInfo.pList);
   int64_t old = atomic_val_compare_exchange_32(&pTask->chkInfo.downstreamAlignNum, 0, num);
@@ -190,6 +208,7 @@ int32_t streamProcessCheckpointBlock(SStreamTask* pTask, SStreamDataBlock* pBloc
     code = streamTaskHandleEvent(pTask->status.pSM, TASK_EVENT_GEN_CHECKPOINT);
     if (code != TSDB_CODE_SUCCESS) {
       stError("s-task:%s handle checkpoint-trigger block failed, code:%s", id, tstrerror(code));
+      streamFreeQitem((SStreamQueueItem*)pBlock);
       return code;
     }
   }

source/libs/stream/src/streamExec.c

@@ -21,6 +21,7 @@
 #define STREAM_RESULT_DUMP_SIZE_THRESHOLD (1048576 * 1)   // 1MiB result data
 #define STREAM_SCAN_HISTORY_TIMESLICE     1000            // 1000 ms
 #define MIN_INVOKE_INTERVAL               50              // 50ms
+#define FILL_HISTORY_TASK_EXEC_INTERVAL   5000            // 5 sec
 
 static int32_t streamTransferStateDoPrepare(SStreamTask* pTask);
 
@@ -46,6 +47,7 @@ static int32_t doOutputResultBlockImpl(SStreamTask* pTask, SStreamDataBlock* pBl
     ASSERT(type == TASK_OUTPUT__FIXED_DISPATCH || type == TASK_OUTPUT__SHUFFLE_DISPATCH);
     code = streamTaskPutDataIntoOutputQ(pTask, pBlock);
     if (code != TSDB_CODE_SUCCESS) {
+      destroyStreamDataBlock(pBlock);
       return code;
     }
 
@@ -76,7 +78,6 @@ static int32_t doDumpResult(SStreamTask* pTask, SStreamQueueItem* pItem, SArray*
 
   int32_t code = doOutputResultBlockImpl(pTask, pStreamBlocks);
   if (code != TSDB_CODE_SUCCESS) {  // back pressure and record position
-    destroyStreamDataBlock(pStreamBlocks);
     return code;
   }
 
@@ -244,6 +245,10 @@ static void streamScanHistoryDataImpl(SStreamTask* pTask, SArray* pRes, int32_t*
   }
 }
 
+static SScanhistoryDataInfo buildScanhistoryExecRet(EScanHistoryCode code, int32_t idleTime) {
+  return (SScanhistoryDataInfo){code, idleTime};
+}
+
 SScanhistoryDataInfo streamScanHistoryData(SStreamTask* pTask, int64_t st) {
   ASSERT(pTask->info.taskLevel == TASK_LEVEL__SOURCE);
 
@@ -260,7 +265,7 @@ SScanhistoryDataInfo streamScanHistoryData(SStreamTask* pTask, int64_t st) {
     if (streamTaskShouldPause(pTask)) {
       stDebug("s-task:%s paused from the scan-history task", id);
       // quit from step1, not continue to handle the step2
-      return (SScanhistoryDataInfo){TASK_SCANHISTORY_QUIT, 0};
+      return buildScanhistoryExecRet(TASK_SCANHISTORY_QUIT, 0);
     }
 
     SArray* pRes = taosArrayInit(0, sizeof(SSDataBlock));
@@ -275,7 +280,7 @@ SScanhistoryDataInfo streamScanHistoryData(SStreamTask* pTask, int64_t st) {
 
     if(streamTaskShouldStop(pTask)) {
       taosArrayDestroyEx(pRes, (FDelete)blockDataFreeRes);
-      return (SScanhistoryDataInfo){TASK_SCANHISTORY_QUIT, 0};
+      return buildScanhistoryExecRet(TASK_SCANHISTORY_QUIT, 0);
     }
 
     // dispatch the generated results
@@ -285,38 +290,21 @@ SScanhistoryDataInfo streamScanHistoryData(SStreamTask* pTask, int64_t st) {
 
     // downstream task input queue is full, try in 5sec
     if (pTask->inputq.status == TASK_INPUT_STATUS__BLOCKED && (pTask->info.fillHistory == 1)) {
-      return (SScanhistoryDataInfo){TASK_SCANHISTORY_REXEC, 5000};
+      return buildScanhistoryExecRet(TASK_SCANHISTORY_REXEC, FILL_HISTORY_TASK_EXEC_INTERVAL);
     }
 
     if (finished) {
-      return (SScanhistoryDataInfo){TASK_SCANHISTORY_CONT, 0};
+      return buildScanhistoryExecRet(TASK_SCANHISTORY_CONT, 0);
     }
 
     if (el >= STREAM_SCAN_HISTORY_TIMESLICE && (pTask->info.fillHistory == 1)) {
       stDebug("s-task:%s fill-history:%d time slice exhausted, elapsed time:%.2fs, retry in 100ms", id,
               pTask->info.fillHistory, el / 1000.0);
-      return (SScanhistoryDataInfo){TASK_SCANHISTORY_REXEC, 100};
+      return buildScanhistoryExecRet(TASK_SCANHISTORY_REXEC, 100);
     }
   }
 }
 
-// wait for the stream task to be idle
-static void waitForTaskIdle(SStreamTask* pTask, SStreamTask* pStreamTask) {
-  const char* id = pTask->id.idStr;
-
-  int64_t st = taosGetTimestampMs();
-  while (!streamTaskIsIdle(pStreamTask)) {
-    stDebug("s-task:%s level:%d wait for stream task:%s to be idle, check again in 100ms", id, pTask->info.taskLevel,
-            pStreamTask->id.idStr);
-    taosMsleep(100);
-  }
-
-  double el = (taosGetTimestampMs() - st) / 1000.0;
-  if (el > 0) {
-    stDebug("s-task:%s wait for stream task:%s for %.2fs to be idle", id, pStreamTask->id.idStr, el);
-  }
-}
-
 int32_t streamTransferStateDoPrepare(SStreamTask* pTask) {
   SStreamMeta* pMeta = pTask->pMeta;
   const char*  id = pTask->id.idStr;

source/libs/stream/src/streamStart.c

@@ -29,19 +29,15 @@ typedef struct SLaunchHTaskInfo {
   STaskId      hTaskId;
 } SLaunchHTaskInfo;
 
-typedef struct STaskRecheckInfo {
-  SStreamTask*        pTask;
-  SStreamTaskCheckReq req;
-  void*               checkTimer;
-} STaskRecheckInfo;
-
 static int32_t           streamSetParamForScanHistory(SStreamTask* pTask);
 static void              streamTaskSetRangeStreamCalc(SStreamTask* pTask);
 static int32_t           initScanHistoryReq(SStreamTask* pTask, SStreamScanHistoryReq* pReq, int8_t igUntreated);
-static SLaunchHTaskInfo* createHTaskLaunchInfo(SStreamMeta* pMeta, int64_t streamId, int32_t taskId, int64_t hStreamId,
+static SLaunchHTaskInfo* createHTaskLaunchInfo(SStreamMeta* pMeta, STaskId* pTaskId, int64_t hStreamId, int32_t hTaskId);
-                                               int32_t hTaskId);
 static void              tryLaunchHistoryTask(void* param, void* tmrId);
 static void              doProcessDownstreamReadyRsp(SStreamTask* pTask);
+static void doExecScanhistoryInFuture(void* param, void* tmrId);
+static int32_t doStartScanHistoryTask(SStreamTask* pTask);
+static int32_t streamTaskStartScanHistory(SStreamTask* pTask);
 
 int32_t streamTaskSetReady(SStreamTask* pTask) {
   int32_t           numOfDowns = streamTaskGetNumOfDownstream(pTask);
@@ -83,7 +79,7 @@ int32_t streamStartScanHistoryAsync(SStreamTask* pTask, int8_t igUntreated) {
   return 0;
 }
 
-static void doExecScanhistoryInFuture(void* param, void* tmrId) {
+void doExecScanhistoryInFuture(void* param, void* tmrId) {
   SStreamTask* pTask = param;
   pTask->schedHistoryInfo.numOfTicks -= 1;
 
@@ -139,7 +135,7 @@ int32_t streamExecScanHistoryInFuture(SStreamTask* pTask, int32_t idleDuration)
   return TSDB_CODE_SUCCESS;
 }
 
-static int32_t doStartScanHistoryTask(SStreamTask* pTask) {
+int32_t doStartScanHistoryTask(SStreamTask* pTask) {
   SVersionRange* pRange = &pTask->dataRange.range;
   if (pTask->info.fillHistory) {
     streamSetParamForScanHistory(pTask);
@@ -195,7 +191,7 @@ void streamTaskCheckDownstream(SStreamTask* pTask) {
     streamTaskAddReqInfo(&pTask->taskCheckInfo, req.reqId, req.downstreamTaskId, pTask->id.idStr);
 
     stDebug("s-task:%s (vgId:%d) stage:%" PRId64 " check single downstream task:0x%x(vgId:%d) ver:%" PRId64 "-%" PRId64
-            " window:%" PRId64 "-%" PRId64 " req:0x%" PRIx64,
+            " window:%" PRId64 "-%" PRId64 " reqId:0x%" PRIx64,
             pTask->id.idStr, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId,
             pRange->range.minVer, pRange->range.maxVer, pWindow->skey, pWindow->ekey, req.reqId);
 
@@ -218,8 +214,9 @@ void streamTaskCheckDownstream(SStreamTask* pTask) {
 
       streamTaskAddReqInfo(&pTask->taskCheckInfo, req.reqId, req.downstreamTaskId, pTask->id.idStr);
 
-      stDebug("s-task:%s (vgId:%d) stage:%" PRId64 " check downstream task:0x%x (vgId:%d) (shuffle), idx:%d",
+      stDebug("s-task:%s (vgId:%d) stage:%" PRId64
-              pTask->id.idStr, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, i);
+              " check downstream task:0x%x (vgId:%d) (shuffle), idx:%d, reqId:0x%" PRIx64,
+              pTask->id.idStr, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, i, req.reqId);
       streamSendCheckMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet);
     }
   } else {  // for sink task, set it ready directly.
@@ -395,7 +392,10 @@ int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRs
   }
 
   if (pRsp->status == TASK_DOWNSTREAM_READY) {
-    streamTaskUpdateCheckInfo(pInfo, pRsp->downstreamTaskId, pRsp->status, now, pRsp->reqId, &left, id);
+    int32_t code = streamTaskUpdateCheckInfo(pInfo, pRsp->downstreamTaskId, pRsp->status, now, pRsp->reqId, &left, id);
+    if (code != TSDB_CODE_SUCCESS) {
+      return TSDB_CODE_SUCCESS;
+    }
 
     if (left == 0) {
       doProcessDownstreamReadyRsp(pTask);  // all downstream tasks are ready, set the complete check downstream flag
@@ -405,7 +405,11 @@ int32_t streamProcessCheckRsp(SStreamTask* pTask, const SStreamTaskCheckRsp* pRs
               pRsp->upstreamNodeId, pRsp->downstreamTaskId, pRsp->downstreamNodeId, pRsp->status, total, left);
     }
   } else {  // not ready, wait for 100ms and retry
-    streamTaskUpdateCheckInfo(pInfo, pRsp->downstreamTaskId, pRsp->status, now, pRsp->reqId, &left, id);
+    int32_t code = streamTaskUpdateCheckInfo(pInfo, pRsp->downstreamTaskId, pRsp->status, now, pRsp->reqId, &left, id);
+    if (code != TSDB_CODE_SUCCESS) {
+      return TSDB_CODE_SUCCESS;  // return success in any cases.
+    }
+
     if (pRsp->status == TASK_UPSTREAM_NEW_STAGE || pRsp->status == TASK_DOWNSTREAM_NOT_LEADER) {
       if (pRsp->status == TASK_UPSTREAM_NEW_STAGE) {
         stError("s-task:%s vgId:%d self vnode-transfer/leader-change/restart detected, old stage:%" PRId64
@@ -655,16 +659,15 @@ static void tryLaunchHistoryTask(void* param, void* tmrId) {
   taosMemoryFree(pInfo);
 }
 
-SLaunchHTaskInfo* createHTaskLaunchInfo(SStreamMeta* pMeta, int64_t streamId, int32_t taskId, int64_t hStreamId,
+SLaunchHTaskInfo* createHTaskLaunchInfo(SStreamMeta* pMeta, STaskId* pTaskId, int64_t hStreamId, int32_t hTaskId) {
-                                        int32_t hTaskId) {
   SLaunchHTaskInfo* pInfo = taosMemoryCalloc(1, sizeof(SLaunchHTaskInfo));
   if (pInfo == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return NULL;
   }
 
-  pInfo->id.streamId = streamId;
+  pInfo->id.streamId = pTaskId->streamId;
-  pInfo->id.taskId = taskId;
+  pInfo->id.taskId = pTaskId->taskId;
 
   pInfo->hTaskId.streamId = hStreamId;
   pInfo->hTaskId.taskId = hTaskId;
@@ -683,7 +686,8 @@ static int32_t launchNotBuiltFillHistoryTask(SStreamTask* pTask) {
 
   stWarn("s-task:%s vgId:%d failed to launch history task:0x%x, since not built yet", idStr, pMeta->vgId, hTaskId);
 
-  SLaunchHTaskInfo* pInfo = createHTaskLaunchInfo(pMeta, pTask->id.streamId, pTask->id.taskId, hStreamId, hTaskId);
+  STaskId id = streamTaskGetTaskId(pTask);
+  SLaunchHTaskInfo* pInfo = createHTaskLaunchInfo(pMeta, &id, hStreamId, hTaskId);
   if (pInfo == NULL) {
     stError("s-task:%s failed to launch related fill-history task, since Out Of Memory", idStr);
     streamMetaAddTaskLaunchResult(pMeta, hStreamId, hTaskId, pExecInfo->checkTs, pExecInfo->readyTs, false);
@@ -852,24 +856,6 @@ int32_t tDecodeStreamTaskCheckRsp(SDecoder* pDecoder, SStreamTaskCheckRsp* pRsp)
   return 0;
 }
 
-int32_t tEncodeStreamTaskCheckpointReq(SEncoder* pEncoder, const SStreamTaskCheckpointReq* 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->nodeId) < 0) return -1;
-  tEndEncode(pEncoder);
-  return 0;
-}
-
-int32_t tDecodeStreamTaskCheckpointReq(SDecoder* pDecoder, SStreamTaskCheckpointReq* 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->nodeId) < 0) return -1;
-  tEndDecode(pDecoder);
-  return 0;
-}
-
 void streamTaskSetRangeStreamCalc(SStreamTask* pTask) {
   SDataRange* pRange = &pTask->dataRange;
 

source/libs/stream/src/streamTask.c

@@ -36,15 +36,20 @@ static int32_t addToTaskset(SArray* pArray, SStreamTask* pTask) {
 
 static int32_t doUpdateTaskEpset(SStreamTask* pTask, int32_t nodeId, SEpSet* pEpSet, bool* pUpdated) {
   char buf[512] = {0};
-
   if (pTask->info.nodeId == nodeId) {  // execution task should be moved away
-    if (!(*pUpdated)) {
+    bool isEqual = isEpsetEqual(&pTask->info.epSet, pEpSet);
-      *pUpdated = isEpsetEqual(&pTask->info.epSet, pEpSet);
+    epsetToStr(pEpSet, buf, tListLen(buf));
-    }
+
+    if (!isEqual) {
+      (*pUpdated) = true;
+      char tmp[512] = {0};
+      epsetToStr(&pTask->info.epSet, tmp, tListLen(tmp));
 
       epsetAssign(&pTask->info.epSet, pEpSet);
-    epsetToStr(pEpSet, buf, tListLen(buf));
+      stDebug("s-task:0x%x (vgId:%d) self node epset is updated %s, old:%s", pTask->id.taskId, nodeId, buf, tmp);
-    stDebug("s-task:0x%x (vgId:%d) self node epset is updated %s", pTask->id.taskId, nodeId, buf);
+    } else {
+      stDebug("s-task:0x%x (vgId:%d) not updated task epset, since epset identical, %s", pTask->id.taskId, nodeId, buf);
+    }
   }
 
   // check for the dispatch info and the upstream task info
@@ -620,13 +625,21 @@ void streamTaskUpdateUpstreamInfo(SStreamTask* pTask, int32_t nodeId, const SEpS
   for (int32_t i = 0; i < numOfUpstream; ++i) {
     SStreamChildEpInfo* pInfo = taosArrayGetP(pTask->upstreamInfo.pList, i);
     if (pInfo->nodeId == nodeId) {
-      if (!(*pUpdated)) {
+      bool equal = isEpsetEqual(&pInfo->epSet, pEpSet);
-        *pUpdated = isEpsetEqual(&pInfo->epSet, pEpSet);
+      if (!equal) {
-      }
+        *pUpdated = true;
+
+        char tmp[512] = {0};
+        epsetToStr(&pInfo->epSet, tmp, tListLen(tmp));
 
         epsetAssign(&pInfo->epSet, pEpSet);
-      stDebug("s-task:0x%x update the upstreamInfo taskId:0x%x(nodeId:%d) newEpset:%s", pTask->id.taskId, pInfo->taskId,
+        stDebug("s-task:0x%x update the upstreamInfo taskId:0x%x(nodeId:%d) newEpset:%s old:%s", pTask->id.taskId,
-              nodeId, buf);
+                pInfo->taskId, nodeId, buf, tmp);
+      } else {
+        stDebug("s-task:0x%x not update upstreamInfo, since identical, task:0x%x(nodeId:%d) epset:%s", pTask->id.taskId,
+                pInfo->taskId, nodeId, buf);
+      }
+
       break;
     }
   }
@@ -653,7 +666,6 @@ void streamTaskSetFixedDownstreamInfo(SStreamTask* pTask, const SStreamTask* pDo
 void streamTaskUpdateDownstreamInfo(SStreamTask* pTask, int32_t nodeId, const SEpSet* pEpSet, bool *pUpdated) {
   char buf[512] = {0};
   epsetToStr(pEpSet, buf, tListLen(buf));
-  *pUpdated = false;
 
   int32_t id = pTask->id.taskId;
   int8_t  type = pTask->outputInfo.type;
@@ -661,29 +673,43 @@ void streamTaskUpdateDownstreamInfo(SStreamTask* pTask, int32_t nodeId, const SE
   if (type == TASK_OUTPUT__SHUFFLE_DISPATCH) {
     SArray* pVgs = pTask->outputInfo.shuffleDispatcher.dbInfo.pVgroupInfos;
 
-    int32_t numOfVgroups = taosArrayGetSize(pVgs);
+    for (int32_t i = 0; i < taosArrayGetSize(pVgs); i++) {
-    for (int32_t i = 0; i < numOfVgroups; i++) {
       SVgroupInfo* pVgInfo = taosArrayGet(pVgs, i);
 
       if (pVgInfo->vgId == nodeId) {
-        if (!(*pUpdated)) {
+        bool isEqual = isEpsetEqual(&pVgInfo->epSet, pEpSet);
-          (*pUpdated) = isEpsetEqual(&pVgInfo->epSet, pEpSet);
+        if (!isEqual) {
-        }
+          *pUpdated = true;
+          char tmp[512] = {0};
+          epsetToStr(&pVgInfo->epSet, tmp, tListLen(tmp));
 
           epsetAssign(&pVgInfo->epSet, pEpSet);
-        stDebug("s-task:0x%x update dispatch info, task:0x%x(nodeId:%d) newEpset:%s", id, pVgInfo->taskId, nodeId, buf);
+          stDebug("s-task:0x%x update dispatch info, task:0x%x(nodeId:%d) newEpset:%s old:%s", id, pVgInfo->taskId,
+                  nodeId, buf, tmp);
+        } else {
+          stDebug("s-task:0x%x not update dispatch info, since identical, task:0x%x(nodeId:%d) epset:%s", id,
+                  pVgInfo->taskId, nodeId, buf);
+        }
         break;
       }
     }
   } else if (type == TASK_OUTPUT__FIXED_DISPATCH) {
     STaskDispatcherFixed* pDispatcher = &pTask->outputInfo.fixedDispatcher;
     if (pDispatcher->nodeId == nodeId) {
-      if (!(*pUpdated)) {
+      bool equal = isEpsetEqual(&pDispatcher->epSet, pEpSet);
-        *pUpdated = isEpsetEqual(&pDispatcher->epSet, pEpSet);
+      if (!equal) {
-      }
+        *pUpdated = true;
+
+        char tmp[512] = {0};
+        epsetToStr(&pDispatcher->epSet, tmp, tListLen(tmp));
 
         epsetAssign(&pDispatcher->epSet, pEpSet);
-      stDebug("s-task:0x%x update dispatch info, task:0x%x(nodeId:%d) newEpset:%s", id, pDispatcher->taskId, nodeId, buf);
+        stDebug("s-task:0x%x update dispatch info, task:0x%x(nodeId:%d) newEpset:%s old:%s", id, pDispatcher->taskId,
+                nodeId, buf, tmp);
+      } else {
+        stDebug("s-task:0x%x not update dispatch info, since identical, task:0x%x(nodeId:%d) epset:%s", id,
+                pDispatcher->taskId, nodeId, buf);
+      }
     }
   }
 }
@@ -912,7 +938,7 @@ static int32_t taskPauseCallback(SStreamTask* pTask, void* param) {
   return TSDB_CODE_SUCCESS;
 }
 
-void streamTaskPause(SStreamMeta* pMeta, SStreamTask* pTask) {
+void streamTaskPause(SStreamTask* pTask) {
   streamTaskHandleEventAsync(pTask->status.pSM, TASK_EVENT_PAUSE, taskPauseCallback, NULL);
 }
 
@@ -983,20 +1009,28 @@ static int32_t streamTaskInitTaskCheckInfo(STaskCheckInfo* pInfo, STaskOutputInf
   return TSDB_CODE_SUCCESS;
 }
 
+static SDownstreamStatusInfo* findCheckRspStatus(STaskCheckInfo* pInfo, int32_t taskId) {
+  for (int32_t j = 0; j < taosArrayGetSize(pInfo->pList); ++j) {
+    SDownstreamStatusInfo* p = taosArrayGet(pInfo->pList, j);
+    if (p->taskId == taskId) {
+      return p;
+    }
+  }
+
+  return NULL;
+}
+
 int32_t streamTaskAddReqInfo(STaskCheckInfo* pInfo, int64_t reqId, int32_t taskId, const char* id) {
   SDownstreamStatusInfo info = {.taskId = taskId, .status = -1, .reqId = reqId, .rspTs = 0};
 
   taosThreadMutexLock(&pInfo->checkInfoLock);
 
-  for(int32_t i = 0; i < taosArrayGetSize(pInfo->pList); ++i) {
+  SDownstreamStatusInfo* p = findCheckRspStatus(pInfo, taskId);
-    SDownstreamStatusInfo* p = taosArrayGet(pInfo->pList, i);
+  if (p != NULL) {
-    if (p->taskId == taskId) {
     stDebug("s-task:%s check info to task:0x%x already sent", id, taskId);
-
     taosThreadMutexUnlock(&pInfo->checkInfoLock);
     return TSDB_CODE_SUCCESS;
   }
-  }
 
   taosArrayPush(pInfo->pList, &info);
 
@@ -1008,12 +1042,18 @@ int32_t streamTaskUpdateCheckInfo(STaskCheckInfo* pInfo, int32_t taskId, int32_t
                                   int32_t* pNotReady, const char* id) {
   taosThreadMutexLock(&pInfo->checkInfoLock);
 
-  for(int32_t i = 0; i < taosArrayGetSize(pInfo->pList); ++i) {
+  SDownstreamStatusInfo* p = findCheckRspStatus(pInfo, taskId);
-    SDownstreamStatusInfo* p = taosArrayGet(pInfo->pList, i);
+  if (p != NULL) {
-    if (p->taskId == taskId) {
-      ASSERT(reqId == p->reqId);
 
-      // count down one, since it is ready now
+    if (reqId != p->reqId) {
+      stError("s-task:%s reqId:%" PRIx64 " expected:%" PRIx64
+              " expired check-rsp recv from downstream task:0x%x, discarded",
+              id, reqId, p->reqId, taskId);
+      taosThreadMutexUnlock(&pInfo->checkInfoLock);
+      return TSDB_CODE_FAILED;
+    }
+
+    // subtract one not-ready-task, since it is ready now
     if ((p->status != TASK_DOWNSTREAM_READY) && (status == TASK_DOWNSTREAM_READY)) {
       *pNotReady = atomic_sub_fetch_32(&pInfo->notReadyTasks, 1);
     } else {
@@ -1026,10 +1066,10 @@ int32_t streamTaskUpdateCheckInfo(STaskCheckInfo* pInfo, int32_t taskId, int32_t
     taosThreadMutexUnlock(&pInfo->checkInfoLock);
     return TSDB_CODE_SUCCESS;
   }
-  }
 
   taosThreadMutexUnlock(&pInfo->checkInfoLock);
-  stError("s-task:%s unexpected check rsp msg, downstream task:0x%x, reqId:%"PRIx64, id, taskId, reqId);
+  stError("s-task:%s unexpected check rsp msg, invalid downstream task:0x%x, reqId:%" PRIx64 " discarded", id, taskId,
+          reqId);
   return TSDB_CODE_FAILED;
 }
 
@@ -1077,7 +1117,7 @@ static void doSendCheckMsg(SStreamTask* pTask, SDownstreamStatusInfo* p) {
     req.reqId = p->reqId;
     req.downstreamNodeId = pOutputInfo->fixedDispatcher.nodeId;
     req.downstreamTaskId = pOutputInfo->fixedDispatcher.taskId;
-    stDebug("s-task:%s (vgId:%d) stage:%" PRId64 " re-send check downstream task:0x%x(vgId:%d) req:0x%" PRIx64,
+    stDebug("s-task:%s (vgId:%d) stage:%" PRId64 " re-send check downstream task:0x%x(vgId:%d) reqId:0x%" PRIx64,
             pTask->id.idStr, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, req.reqId);
 
     streamSendCheckMsg(pTask, &req, pOutputInfo->fixedDispatcher.nodeId, &pOutputInfo->fixedDispatcher.epSet);
@@ -1093,8 +1133,10 @@ static void doSendCheckMsg(SStreamTask* pTask, SDownstreamStatusInfo* p) {
         req.downstreamNodeId = pVgInfo->vgId;
         req.downstreamTaskId = pVgInfo->taskId;
 
-        stDebug("s-task:%s (vgId:%d) stage:%" PRId64 " re-send check downstream task:0x%x(vgId:%d) (shuffle), idx:%d",
+        stDebug("s-task:%s (vgId:%d) stage:%" PRId64
-                pTask->id.idStr, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, i);
+                " re-send check downstream task:0x%x(vgId:%d) (shuffle), idx:%d reqId:0x%" PRIx64,
+                pTask->id.idStr, pTask->info.nodeId, req.stage, req.downstreamTaskId, req.downstreamNodeId, i,
+                p->reqId);
         streamSendCheckMsg(pTask, &req, pVgInfo->vgId, &pVgInfo->epSet);
         break;
       }
@@ -1104,6 +1146,31 @@ static void doSendCheckMsg(SStreamTask* pTask, SDownstreamStatusInfo* p) {
   }
 }
 
+static void getCheckRspStatus(STaskCheckInfo* pInfo, int64_t el, int32_t* numOfReady, int32_t* numOfFault,
+                              int32_t* numOfNotRsp, SArray* pTimeoutList, SArray* pNotReadyList, const char* id) {
+  for (int32_t i = 0; i < taosArrayGetSize(pInfo->pList); ++i) {
+    SDownstreamStatusInfo* p = taosArrayGet(pInfo->pList, i);
+    if (p->status == TASK_DOWNSTREAM_READY) {
+      (*numOfReady) += 1;
+    } else if (p->status == TASK_UPSTREAM_NEW_STAGE || p->status == TASK_DOWNSTREAM_NOT_LEADER) {
+      stDebug("s-task:%s recv status:NEW_STAGE/NOT_LEADER from downstream, task:0x%x, quit from check downstream", id,
+              p->taskId);
+      (*numOfFault) += 1;
+    } else {                // TASK_DOWNSTREAM_NOT_READY
+      if (p->rspTs == 0) {  // not response yet
+        ASSERT(p->status == -1);
+        if (el >= CHECK_NOT_RSP_DURATION) {  // not receive info for 10 sec.
+          taosArrayPush(pTimeoutList, &p->taskId);
+        } else {                // el < CHECK_NOT_RSP_DURATION
+          (*numOfNotRsp) += 1;  // do nothing and continue waiting for their rsp
+        }
+      } else {
+        taosArrayPush(pNotReadyList, &p->taskId);
+      }
+    }
+  }
+}
+
 static void rspMonitorFn(void* param, void* tmrId) {
   SStreamTask*      pTask = param;
   SStreamTaskState* pStat = streamTaskGetStatus(pTask);
@@ -1119,7 +1186,7 @@ static void rspMonitorFn(void* param, void* tmrId) {
   int32_t           numOfNotReady = 0;
   int32_t           numOfTimeout = 0;
 
-  stDebug("s-task:%s start to do check downstream rsp check", id);
+  stDebug("s-task:%s start to do check-downstream-rsp check in tmr", id);
 
   if (state == TASK_STATUS__STOP) {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
@@ -1130,17 +1197,20 @@ static void rspMonitorFn(void* param, void* tmrId) {
     taosThreadMutexUnlock(&pInfo->checkInfoLock);
 
     streamMetaAddTaskLaunchResult(pTask->pMeta, pTask->id.streamId, pTask->id.taskId, pInfo->startTs, now, false);
+    if (HAS_RELATED_FILLHISTORY_TASK(pTask)) {
+      STaskId* pHId = &pTask->hTaskInfo.id;
+      streamMetaAddTaskLaunchResult(pTask->pMeta, pHId->streamId, pHId->taskId, pInfo->startTs, now, false);
+    }
     return;
   }
 
-  if (state == TASK_STATUS__DROPPING || state == TASK_STATUS__READY) {
+  if (state == TASK_STATUS__DROPPING || state == TASK_STATUS__READY || state == TASK_STATUS__PAUSE) {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug("s-task:%s status:%s vgId:%d quit from monitor check-rsp tmr, ref:%d", id, pStat->name, vgId, ref);
 
     taosThreadMutexLock(&pInfo->checkInfoLock);
     streamTaskCompleteCheckRsp(pInfo, id);
     taosThreadMutexUnlock(&pInfo->checkInfoLock);
-
     return;
   }
 
@@ -1159,27 +1229,7 @@ static void rspMonitorFn(void* param, void* tmrId) {
   SArray* pTimeoutList = taosArrayInit(4, sizeof(int64_t));
 
   if (pStat->state == TASK_STATUS__UNINIT) {
-    for (int32_t i = 0; i < taosArrayGetSize(pInfo->pList); ++i) {
+    getCheckRspStatus(pInfo, el, &numOfReady, &numOfFault, &numOfNotRsp, pTimeoutList, pNotReadyList, id);
-      SDownstreamStatusInfo* p = taosArrayGet(pInfo->pList, i);
-      if (p->status == TASK_DOWNSTREAM_READY) {
-        numOfReady += 1;
-      } else if (p->status == TASK_UPSTREAM_NEW_STAGE || p->status == TASK_DOWNSTREAM_NOT_LEADER) {
-        stDebug("s-task:%s recv status:NEW_STAGE/NOT_LEADER from downstream, task:0x%x, quit from check downstream", id,
-                p->taskId);
-        numOfFault += 1;
-      } else {                // TASK_DOWNSTREAM_NOT_READY
-        if (p->rspTs == 0) {  // not response yet
-          ASSERT(p->status == -1);
-          if (el >= CHECK_NOT_RSP_DURATION) {  // not receive info for 10 sec.
-            taosArrayPush(pTimeoutList, &p->taskId);
-          } else {  // el < CHECK_NOT_RSP_DURATION
-            numOfNotRsp += 1;  // do nothing and continue waiting for their rsp
-          }
-        } else {
-          taosArrayPush(pNotReadyList, &p->taskId);
-        }
-      }
-    }
   } else {  // unexpected status
     stError("s-task:%s unexpected task status:%s during waiting for check rsp", id, pStat->name);
   }
@@ -1189,18 +1239,18 @@ static void rspMonitorFn(void* param, void* tmrId) {
 
   // fault tasks detected, not try anymore
   ASSERT((numOfReady + numOfFault + numOfNotReady + numOfTimeout + numOfNotRsp) == taosArrayGetSize(pInfo->pList));
-  if ((numOfNotRsp == 0) && (numOfFault > 0)) {
+  if (numOfFault > 0) {
     int32_t ref = atomic_sub_fetch_32(&pTask->status.timerActive, 1);
     stDebug(
         "s-task:%s status:%s vgId:%d all rsp. quit from monitor rsp tmr, since vnode-transfer/leader-change/restart "
         "detected, notRsp:%d, notReady:%d, fault:%d, timeout:%d, ready:%d ref:%d",
         id, pStat->name, vgId, numOfNotRsp, numOfNotReady, numOfFault, numOfTimeout, numOfReady, ref);
 
+    streamTaskCompleteCheckRsp(pInfo, id);
     taosThreadMutexUnlock(&pInfo->checkInfoLock);
+
     taosArrayDestroy(pNotReadyList);
     taosArrayDestroy(pTimeoutList);
-
-    streamTaskCompleteCheckRsp(pInfo, id);
     return;
   }
 
@@ -1221,6 +1271,9 @@ static void rspMonitorFn(void* param, void* tmrId) {
       STaskId* pHId = &pTask->hTaskInfo.id;
       streamMetaAddTaskLaunchResult(pTask->pMeta, pHId->streamId, pHId->taskId, pInfo->startTs, now, false);
     }
+
+    taosArrayDestroy(pNotReadyList);
+    taosArrayDestroy(pTimeoutList);
     return;
   }
 
@@ -1231,15 +1284,13 @@ static void rspMonitorFn(void* param, void* tmrId) {
     for (int32_t i = 0; i < numOfNotReady; ++i) {
       int32_t taskId = *(int32_t*)taosArrayGet(pNotReadyList, i);
 
-      for (int32_t j = 0; j < taosArrayGetSize(pInfo->pList); ++j) {
+      SDownstreamStatusInfo* p = findCheckRspStatus(pInfo, taskId);
-        SDownstreamStatusInfo* p = taosArrayGet(pInfo->pList, j);
+      if (p != NULL) {
-        if (p->taskId == taskId) {
         p->rspTs = 0;
         p->status = -1;
         doSendCheckMsg(pTask, p);
       }
     }
-    }
 
     stDebug("s-task:%s %d downstream task(s) not ready, send check msg again", id, numOfNotReady);
   }
@@ -1251,13 +1302,10 @@ static void rspMonitorFn(void* param, void* tmrId) {
     for (int32_t i = 0; i < numOfTimeout; ++i) {
       int32_t taskId = *(int32_t*)taosArrayGet(pTimeoutList, i);
 
-      for (int32_t j = 0; j < taosArrayGetSize(pInfo->pList); ++j) {
+      SDownstreamStatusInfo* p = findCheckRspStatus(pInfo, taskId);
-        SDownstreamStatusInfo* p = taosArrayGet(pInfo->pList, j);
+      if (p != NULL) {
-        if (p->taskId == taskId) {
         ASSERT(p->status == -1 && p->rspTs == 0);
         doSendCheckMsg(pTask, p);
-          break;
-        }
       }
     }
 
@@ -1293,7 +1341,7 @@ int32_t streamTaskStartMonitorCheckRsp(SStreamTask* pTask) {
   if (pInfo->checkRspTmr == NULL) {
     pInfo->checkRspTmr = taosTmrStart(rspMonitorFn, CHECK_RSP_INTERVAL, pTask, streamTimer);
   } else {
-    taosTmrReset(rspMonitorFn, CHECK_RSP_INTERVAL, pTask, streamTimer, pInfo->checkRspTmr);
+    taosTmrReset(rspMonitorFn, CHECK_RSP_INTERVAL, pTask, streamTimer, &pInfo->checkRspTmr);
   }
 
   taosThreadMutexUnlock(&pInfo->checkInfoLock);

source/libs/stream/src/streamUpdate.c

@@ -13,6 +13,8 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include "tcompare.h"
+#include "tdatablock.h"
 #include "tencode.h"
 #include "tstreamUpdate.h"
 #include "ttime.h"
@@ -31,6 +33,39 @@
 
 static int64_t adjustExpEntries(int64_t entries) { return TMIN(DEFAULT_EXPECTED_ENTRIES, entries); }
 
+int compareKeyTs(void* pTs1, void* pTs2, void* pPkVal, __compar_fn_t cmpPkFn) {
+  return compareInt64Val(pTs1, pTs2);;
+}
+
+int compareKeyTsAndPk(void* pValue1, void* pTs, void* pPkVal, __compar_fn_t cmpPkFn) {
+  int res = compareInt64Val(pValue1, pTs);
+  if (res != 0) {
+    return res;
+  } else {
+    void* pk1 = (char*)pValue1 + sizeof(TSKEY);
+    return cmpPkFn(pk1, pPkVal);
+  }
+}
+
+int32_t getKeyBuff(TSKEY ts, int64_t tbUid, void* pVal, int32_t len, char* buff) {
+  *(TSKEY*)buff = ts;
+  memcpy(buff+ sizeof(TSKEY), &tbUid, sizeof(int64_t));
+  if (len == 0) {
+    return sizeof(TSKEY) + sizeof(int64_t);
+  }
+  memcpy(buff, pVal, len);
+  return sizeof(TSKEY) + sizeof(int64_t) + len;
+}
+
+int32_t getValueBuff(TSKEY ts, char* pVal, int32_t len, char* buff) {
+  *(TSKEY*)buff = ts;
+  if (len == 0) {
+    return sizeof(TSKEY);
+  }
+  memcpy(buff + sizeof(TSKEY), pVal, len);
+  return sizeof(TSKEY) + len;
+}
+
 void windowSBfAdd(SUpdateInfo *pInfo, uint64_t count) {
   if (pInfo->numSBFs < count) {
     count = pInfo->numSBFs;
@@ -89,11 +124,11 @@ static int64_t adjustWatermark(int64_t adjInterval, int64_t originInt, int64_t w
   return watermark;
 }
 
-SUpdateInfo *updateInfoInitP(SInterval *pInterval, int64_t watermark, bool igUp) {
+SUpdateInfo *updateInfoInitP(SInterval *pInterval, int64_t watermark, bool igUp, int8_t pkType, int32_t pkLen) {
-  return updateInfoInit(pInterval->interval, pInterval->precision, watermark, igUp);
+  return updateInfoInit(pInterval->interval, pInterval->precision, watermark, igUp, pkType, pkLen);
 }
 
-SUpdateInfo *updateInfoInit(int64_t interval, int32_t precision, int64_t watermark, bool igUp) {
+SUpdateInfo *updateInfoInit(int64_t interval, int32_t precision, int64_t watermark, bool igUp, int8_t pkType, int32_t pkLen) {
   SUpdateInfo *pInfo = taosMemoryCalloc(1, sizeof(SUpdateInfo));
   if (pInfo == NULL) {
     return NULL;
@@ -133,6 +168,17 @@ SUpdateInfo *updateInfoInit(int64_t interval, int32_t precision, int64_t waterma
   _hash_fn_t hashFn = taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT);
   pInfo->pMap = taosHashInit(DEFAULT_MAP_CAPACITY, hashFn, true, HASH_NO_LOCK);
   pInfo->maxDataVersion = 0;
+  pInfo->pkColLen = pkLen;
+  pInfo->pkColType = pkType;
+  pInfo->pKeyBuff = taosMemoryCalloc(1, sizeof(TSKEY) + sizeof(int64_t) + pkLen);
+  pInfo->pValueBuff = taosMemoryCalloc(1, sizeof(TSKEY) + pkLen);
+  if (pkLen != 0) {
+    pInfo->comparePkRowFn = compareKeyTsAndPk;
+    pInfo->comparePkCol = getKeyComparFunc(pkType, TSDB_ORDER_ASC);;
+  } else {
+    pInfo->comparePkRowFn = compareKeyTs;
+    pInfo->comparePkCol = NULL;
+  }
   return pInfo;
 }
 
@@ -168,47 +214,60 @@ bool updateInfoIsTableInserted(SUpdateInfo *pInfo, int64_t tbUid) {
   return false;
 }
 
-TSKEY updateInfoFillBlockData(SUpdateInfo *pInfo, SSDataBlock *pBlock, int32_t primaryTsCol) {
+TSKEY updateInfoFillBlockData(SUpdateInfo *pInfo, SSDataBlock *pBlock, int32_t primaryTsCol, int32_t primaryKeyCol) {
   if (pBlock == NULL || pBlock->info.rows == 0) return INT64_MIN;
   TSKEY   maxTs = INT64_MIN;
+  void*   pPkVal = NULL;
+  void*   pMaxPkVal = NULL;
+  int32_t maxLen = 0;
+  int32_t len = 0;
   int64_t tbUid = pBlock->info.id.uid;
 
   SColumnInfoData *pColDataInfo = taosArrayGet(pBlock->pDataBlock, primaryTsCol);
+  SColumnInfoData *pPkDataInfo = NULL;
+  if (primaryKeyCol >= 0) {
+    pPkDataInfo = taosArrayGet(pBlock->pDataBlock, primaryKeyCol);
+  }
 
   for (int32_t i = 0; i < pBlock->info.rows; i++) {
     TSKEY ts = ((TSKEY *)pColDataInfo->pData)[i];
-    maxTs = TMAX(maxTs, ts);
+    if (maxTs < ts) {
+      maxTs = ts;
+      if (primaryKeyCol >= 0) {
+        pMaxPkVal = colDataGetData(pPkDataInfo, i);
+        maxLen = colDataGetRowLength(pPkDataInfo, i);
+      }
+    }
     SScalableBf *pSBf = getSBf(pInfo, ts);
     if (pSBf) {
-      SUpdateKey updateKey = {
+      if (primaryKeyCol >= 0) {
-          .tbUid = tbUid,
+        pPkVal = colDataGetData(pPkDataInfo, i);
-          .ts = ts,
+        len = colDataGetRowLength(pPkDataInfo, i);
-      };
+      }
-      tScalableBfPut(pSBf, &updateKey, sizeof(SUpdateKey));
+      int32_t buffLen = getKeyBuff(ts, tbUid, pPkVal, len, pInfo->pKeyBuff);
+      tScalableBfPut(pSBf, pInfo->pKeyBuff, buffLen);
     }
   }
-  TSKEY *pMaxTs = taosHashGet(pInfo->pMap, &tbUid, sizeof(int64_t));
+  void *pMaxTs = taosHashGet(pInfo->pMap, &tbUid, sizeof(int64_t));
-  if (pMaxTs == NULL || *pMaxTs > maxTs) {
+  if (pMaxTs == NULL || pInfo->comparePkRowFn(pMaxTs, &maxTs, pMaxPkVal, pInfo->comparePkCol) == -1) {
-    taosHashPut(pInfo->pMap, &tbUid, sizeof(int64_t), &maxTs, sizeof(TSKEY));
+    int32_t valueLen = getValueBuff(maxTs, pMaxPkVal, maxLen, pInfo->pValueBuff);
+    taosHashPut(pInfo->pMap, &tbUid, sizeof(int64_t), pInfo->pValueBuff, valueLen);
   }
   return maxTs;
 }
 
-bool updateInfoIsUpdated(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts) {
+bool updateInfoIsUpdated(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts, void* pPkVal, int32_t len) {
   int32_t res = TSDB_CODE_FAILED;
+  int32_t buffLen = 0;
 
-  SUpdateKey updateKey = {
+  buffLen = getKeyBuff(ts, tableId, pPkVal, len, pInfo->pKeyBuff);
-      .tbUid = tableId,
+  void*   *pMapMaxTs = taosHashGet(pInfo->pMap, &tableId, sizeof(uint64_t));
-      .ts = ts,
-  };
-
-  TSKEY   *pMapMaxTs = taosHashGet(pInfo->pMap, &tableId, sizeof(uint64_t));
   uint64_t index = ((uint64_t)tableId) % pInfo->numBuckets;
   TSKEY    maxTs = *(TSKEY *)taosArrayGet(pInfo->pTsBuckets, index);
   if (ts < maxTs - pInfo->watermark) {
     // this window has been closed.
     if (pInfo->pCloseWinSBF) {
-      res = tScalableBfPut(pInfo->pCloseWinSBF, &updateKey, sizeof(SUpdateKey));
+      res = tScalableBfPut(pInfo->pCloseWinSBF, pInfo->pKeyBuff, buffLen);
       if (res == TSDB_CODE_SUCCESS) {
         return false;
       } else {
@@ -221,18 +280,19 @@ bool updateInfoIsUpdated(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts) {
   SScalableBf *pSBf = getSBf(pInfo, ts);
 
   int32_t size = taosHashGetSize(pInfo->pMap);
-  if ((!pMapMaxTs && size < DEFAULT_MAP_SIZE) || (pMapMaxTs && *pMapMaxTs < ts)) {
+  if ((!pMapMaxTs && size < DEFAULT_MAP_SIZE) || (pMapMaxTs && pInfo->comparePkRowFn(pMapMaxTs, &ts, pPkVal, pInfo->comparePkCol) == -1 )) {
-    taosHashPut(pInfo->pMap, &tableId, sizeof(uint64_t), &ts, sizeof(TSKEY));
+    int32_t valueLen = getValueBuff(ts, pPkVal, len, pInfo->pValueBuff);
+    taosHashPut(pInfo->pMap, &tableId, sizeof(uint64_t), pInfo->pValueBuff, valueLen);
     // pSBf may be a null pointer
     if (pSBf) {
-      res = tScalableBfPutNoCheck(pSBf, &updateKey, sizeof(SUpdateKey));
+      res = tScalableBfPutNoCheck(pSBf, pInfo->pKeyBuff, buffLen);
     }
     return false;
   }
 
   // pSBf may be a null pointer
   if (pSBf) {
-    res = tScalableBfPut(pSBf, &updateKey, sizeof(SUpdateKey));
+    res = tScalableBfPut(pSBf, pInfo->pKeyBuff, buffLen);
   }
 
   if (!pMapMaxTs && maxTs < ts) {
@@ -262,6 +322,8 @@ void updateInfoDestroy(SUpdateInfo *pInfo) {
   }
 
   taosArrayDestroy(pInfo->pTsSBFs);
+  taosMemoryFreeClear(pInfo->pKeyBuff);
+  taosMemoryFreeClear(pInfo->pValueBuff);
   taosHashCleanup(pInfo->pMap);
   updateInfoDestoryColseWinSBF(pInfo);
   taosMemoryFree(pInfo);
@@ -322,11 +384,15 @@ int32_t updateInfoSerialize(void *buf, int32_t bufLen, const SUpdateInfo *pInfo)
   while ((pIte = taosHashIterate(pInfo->pMap, pIte)) != NULL) {
     void *key = taosHashGetKey(pIte, &keyLen);
     if (tEncodeU64(&encoder, *(uint64_t *)key) < 0) return -1;
-    if (tEncodeI64(&encoder, *(TSKEY *)pIte) < 0) return -1;
+    int32_t valueSize = taosHashGetValueSize(pIte);
+    if (tEncodeBinary(&encoder, (const uint8_t *)pIte, valueSize) < 0) return -1;
   }
 
   if (tEncodeU64(&encoder, pInfo->maxDataVersion) < 0) return -1;
 
+  if (tEncodeI32(&encoder, pInfo->pkColLen) < 0) return -1;
+  if (tEncodeI8(&encoder, pInfo->pkColType) < 0) return -1;
+
   tEndEncode(&encoder);
 
   int32_t tlen = encoder.pos;
@@ -371,28 +437,43 @@ int32_t updateInfoDeserialize(void *buf, int32_t bufLen, SUpdateInfo *pInfo) {
   _hash_fn_t hashFn = taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT);
   pInfo->pMap = taosHashInit(mapSize, hashFn, true, HASH_NO_LOCK);
   uint64_t uid = 0;
-  ts = INT64_MIN;
+  void* pVal = NULL;
+  int32_t valSize = 0;
   for (int32_t i = 0; i < mapSize; i++) {
     if (tDecodeU64(&decoder, &uid) < 0) return -1;
-    if (tDecodeI64(&decoder, &ts) < 0) return -1;
+    if (tDecodeBinary(&decoder, (uint8_t**)&pVal, &valSize) < 0) return -1;
-    taosHashPut(pInfo->pMap, &uid, sizeof(uint64_t), &ts, sizeof(TSKEY));
+    taosHashPut(pInfo->pMap, &uid, sizeof(uint64_t), pVal, valSize);
   }
   ASSERT(mapSize == taosHashGetSize(pInfo->pMap));
   if (tDecodeU64(&decoder, &pInfo->maxDataVersion) < 0) return -1;
 
+  if (tDecodeI32(&decoder, &pInfo->pkColLen) < 0) return -1;
+  if (tDecodeI8(&decoder, &pInfo->pkColType) < 0) return -1;
+
+  pInfo->pKeyBuff = taosMemoryCalloc(1, sizeof(TSKEY) + sizeof(int64_t) + pInfo->pkColLen);
+  pInfo->pValueBuff = taosMemoryCalloc(1, sizeof(TSKEY) + pInfo->pkColLen);
+  if (pInfo->pkColLen != 0) {
+    pInfo->comparePkRowFn = compareKeyTsAndPk;
+    pInfo->comparePkCol = getKeyComparFunc(pInfo->pkColType, TSDB_ORDER_ASC);;
+  } else {
+    pInfo->comparePkRowFn = compareKeyTs;
+    pInfo->comparePkCol = NULL;
+  }
+
   tEndDecode(&decoder);
 
   tDecoderClear(&decoder);
   return 0;
 }
 
-bool isIncrementalTimeStamp(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts) {
+bool isIncrementalTimeStamp(SUpdateInfo *pInfo, uint64_t tableId, TSKEY ts, void* pPkVal, int32_t len) {
   TSKEY *pMapMaxTs = taosHashGet(pInfo->pMap, &tableId, sizeof(uint64_t));
   bool   res = true;
-  if (pMapMaxTs && ts < *pMapMaxTs) {
+  if (pMapMaxTs && pInfo->comparePkRowFn(pMapMaxTs, &ts, pPkVal, pInfo->comparePkCol) == 1) {
     res = false;
   } else {
-    taosHashPut(pInfo->pMap, &tableId, sizeof(uint64_t), &ts, sizeof(TSKEY));
+    int32_t valueLen = getValueBuff(ts, pPkVal, len, pInfo->pValueBuff);
+    taosHashPut(pInfo->pMap, &tableId, sizeof(uint64_t), pInfo->pValueBuff, valueLen);
   }
   return res;
 }

source/util/src/talgo.c

@@ -153,6 +153,130 @@ void taosqsort(void *src, int64_t numOfElem, int64_t size, const void *param, __
   taosMemoryFreeClear(buf);
 }
 
+#define DOSWAP(a, b, size)        \
+  do {                            \
+    size_t __size = (size);       \
+    char  *__a = (a), *__b = (b); \
+    do {                          \
+      char __tmp = *__a;          \
+      *__a++ = *__b;              \
+      *__b++ = __tmp;             \
+    } while (--__size > 0);       \
+  } while (0)
+
+typedef struct {
+  char *lo;
+  char *hi;
+} stack_node;
+
+#define STACK_SIZE      (CHAR_BIT * sizeof(size_t))
+#define PUSH(low, high) ((void)((top->lo = (low)), (top->hi = (high)), ++top))
+#define POP(low, high)  ((void)(--top, (low = top->lo), (high = top->hi)))
+#define STACK_NOT_EMPTY (stack < top)
+
+void taosqsort_r(void *src, int64_t nelem, int64_t size, const void *arg, __ext_compar_fn_t cmp) {
+  const int32_t MAX_THRESH = 6;
+  char         *base_ptr = (char *)src;
+
+  const size_t max_thresh = MAX_THRESH * size;
+
+  if (nelem == 0) return;
+
+  if (nelem > MAX_THRESH) {
+    char       *lo = base_ptr;
+    char       *hi = &lo[size * (nelem - 1)];
+    stack_node  stack[STACK_SIZE];
+    stack_node *top = stack;
+
+    PUSH(NULL, NULL);
+
+    while (STACK_NOT_EMPTY) {
+      char *left_ptr;
+      char *right_ptr;
+
+      char *mid = lo + size * ((hi - lo) / size >> 1);
+
+      if ((*cmp)((void *)mid, (void *)lo, arg) < 0) DOSWAP(mid, lo, size);
+      if ((*cmp)((void *)hi, (void *)mid, arg) < 0)
+        DOSWAP(mid, hi, size);
+      else
+        goto jump_over;
+      if ((*cmp)((void *)mid, (void *)lo, arg) < 0) DOSWAP(mid, lo, size);
+    jump_over:;
+
+      left_ptr = lo + size;
+      right_ptr = hi - size;
+      do {
+        while ((*cmp)((void *)left_ptr, (void *)mid, arg) < 0) left_ptr += size;
+
+        while ((*cmp)((void *)mid, (void *)right_ptr, arg) < 0) right_ptr -= size;
+
+        if (left_ptr < right_ptr) {
+          DOSWAP(left_ptr, right_ptr, size);
+          if (mid == left_ptr)
+            mid = right_ptr;
+          else if (mid == right_ptr)
+            mid = left_ptr;
+          left_ptr += size;
+          right_ptr -= size;
+        } else if (left_ptr == right_ptr) {
+          left_ptr += size;
+          right_ptr -= size;
+          break;
+        }
+      } while (left_ptr <= right_ptr);
+
+      if ((size_t)(right_ptr - lo) <= max_thresh) {
+        if ((size_t)(hi - left_ptr) <= max_thresh)
+          POP(lo, hi);
+        else
+          lo = left_ptr;
+      } else if ((size_t)(hi - left_ptr) <= max_thresh)
+        hi = right_ptr;
+      else if ((right_ptr - lo) > (hi - left_ptr)) {
+        PUSH(lo, right_ptr);
+        lo = left_ptr;
+      } else {
+        PUSH(left_ptr, hi);
+        hi = right_ptr;
+      }
+    }
+  }
+#define min(x, y) ((x) < (y) ? (x) : (y))
+
+  {
+    char *const end_ptr = &base_ptr[size * (nelem - 1)];
+    char       *tmp_ptr = base_ptr;
+    char       *thresh = min(end_ptr, base_ptr + max_thresh);
+    char       *run_ptr;
+
+    for (run_ptr = tmp_ptr + size; run_ptr <= thresh; run_ptr += size)
+      if ((*cmp)((void *)run_ptr, (void *)tmp_ptr, arg) < 0) tmp_ptr = run_ptr;
+
+    if (tmp_ptr != base_ptr) DOSWAP(tmp_ptr, base_ptr, size);
+
+    run_ptr = base_ptr + size;
+    while ((run_ptr += size) <= end_ptr) {
+      tmp_ptr = run_ptr - size;
+      while ((*cmp)((void *)run_ptr, (void *)tmp_ptr, arg) < 0) tmp_ptr -= size;
+
+      tmp_ptr += size;
+      if (tmp_ptr != run_ptr) {
+        char *trav;
+
+        trav = run_ptr + size;
+        while (--trav >= run_ptr) {
+          char  c = *trav;
+          char *hi, *lo;
+
+          for (hi = lo = trav; (lo -= size) >= tmp_ptr; hi = lo) *hi = *lo;
+          *hi = c;
+        }
+      }
+    }
+  }
+}
+
 void *taosbsearch(const void *key, const void *base, int32_t nmemb, int32_t size, __compar_fn_t compar, int32_t flags) {
   uint8_t *p;
   int32_t  lidx;
@@ -351,7 +475,6 @@ int32_t msortHelper(const void *p1, const void *p2, const void *param) {
   return comparFn(p1, p2);
 }
 
-
 int32_t taosMergeSort(void *src, int64_t numOfElem, int64_t size, __compar_fn_t comparFn) {
   void *param = comparFn;
   return taosMergeSortHelper(src, numOfElem, size, param, msortHelper);

source/util/src/tconfig.c

@@ -27,12 +27,17 @@
 #define CFG_NAME_PRINT_LEN 24
 #define CFG_SRC_PRINT_LEN  12
 
+struct SConfig {
+  ECfgSrcType   stype;
+  SArray       *array;
+  TdThreadMutex lock;
+};
+
 int32_t cfgLoadFromCfgFile(SConfig *pConfig, const char *filepath);
-int32_t cfgLoadFromEnvFile(SConfig *pConfig, const char *filepath);
+int32_t cfgLoadFromEnvFile(SConfig *pConfig, const char *envFile);
 int32_t cfgLoadFromEnvVar(SConfig *pConfig);
 int32_t cfgLoadFromEnvCmd(SConfig *pConfig, const char **envCmd);
 int32_t cfgLoadFromApollUrl(SConfig *pConfig, const char *url);
-int32_t cfgSetItem(SConfig *pConfig, const char *name, const char *value, ECfgSrcType stype);
 
 extern char **environ;
 
@@ -50,6 +55,7 @@ SConfig *cfgInit() {
     return NULL;
   }
 
+  taosThreadMutexInit(&pCfg->lock, NULL);
   return pCfg;
 }
 
@@ -87,9 +93,9 @@ static void cfgFreeItem(SConfigItem *pItem) {
       pItem->dtype == CFG_DTYPE_CHARSET || pItem->dtype == CFG_DTYPE_TIMEZONE) {
     taosMemoryFreeClear(pItem->str);
   }
+
   if (pItem->array) {
-    taosArrayDestroy(pItem->array);
+    pItem->array = taosArrayDestroy(pItem->array);
-    pItem->array = NULL;
   }
 }
 
@@ -102,37 +108,18 @@ void cfgCleanup(SConfig *pCfg) {
       taosMemoryFreeClear(pItem->name);
     }
     taosArrayDestroy(pCfg->array);
+    taosThreadMutexDestroy(&pCfg->lock);
     taosMemoryFree(pCfg);
   }
 }
 
 int32_t cfgGetSize(SConfig *pCfg) { return taosArrayGetSize(pCfg->array); }
 
-static int32_t cfgCheckAndSetTimezone(SConfigItem *pItem, const char *timezone) {
+static int32_t cfgCheckAndSetConf(SConfigItem *pItem, const char *conf) {
   cfgFreeItem(pItem);
-  pItem->str = taosStrdup(timezone);
+  ASSERT(pItem->str == NULL);
-  if (pItem->str == NULL) {
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    return -1;
-  }
 
-  return 0;
+  pItem->str = taosStrdup(conf);
-}
-
-static int32_t cfgCheckAndSetCharset(SConfigItem *pItem, const char *charset) {
-  cfgFreeItem(pItem);
-  pItem->str = taosStrdup(charset);
-  if (pItem->str == NULL) {
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    return -1;
-  }
-
-  return 0;
-}
-
-static int32_t cfgCheckAndSetLocale(SConfigItem *pItem, const char *locale) {
-  cfgFreeItem(pItem);
-  pItem->str = taosStrdup(locale);
   if (pItem->str == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
@@ -229,7 +216,7 @@ static int32_t cfgSetString(SConfigItem *pItem, const char *value, ECfgSrcType s
     return -1;
   }
 
-  taosMemoryFree(pItem->str);
+  taosMemoryFreeClear(pItem->str);
   pItem->str = tmp;
   pItem->stype = stype;
   return 0;
@@ -246,20 +233,8 @@ static int32_t cfgSetDir(SConfigItem *pItem, const char *value, ECfgSrcType styp
   return 0;
 }
 
-static int32_t cfgSetLocale(SConfigItem *pItem, const char *value, ECfgSrcType stype) {
+static int32_t doSetConf(SConfigItem *pItem, const char *value, ECfgSrcType stype) {
-  if (cfgCheckAndSetLocale(pItem, value) != 0) {
+  if (cfgCheckAndSetConf(pItem, value) != 0) {
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    uError("cfg:%s, type:%s src:%s value:%s failed to dup since %s", pItem->name, cfgDtypeStr(pItem->dtype),
-           cfgStypeStr(stype), value, terrstr());
-    return -1;
-  }
-
-  pItem->stype = stype;
-  return 0;
-}
-
-static int32_t cfgSetCharset(SConfigItem *pItem, const char *value, ECfgSrcType stype) {
-  if (cfgCheckAndSetCharset(pItem, value) != 0) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     uError("cfg:%s, type:%s src:%s value:%s failed to dup since %s", pItem->name, cfgDtypeStr(pItem->dtype),
            cfgStypeStr(stype), value, terrstr());
@@ -271,18 +246,13 @@ static int32_t cfgSetCharset(SConfigItem *pItem, const char *value, ECfgSrcType
 }
 
 static int32_t cfgSetTimezone(SConfigItem *pItem, const char *value, ECfgSrcType stype) {
-  if (cfgCheckAndSetTimezone(pItem, value) != 0) {
+  int32_t code = doSetConf(pItem, value, stype);
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
+  if (code != TSDB_CODE_SUCCESS) {
-    uError("cfg:%s, type:%s src:%s value:%s failed to dup since %s", pItem->name, cfgDtypeStr(pItem->dtype),
+    return code;
-           cfgStypeStr(stype), value, terrstr());
-    return -1;
   }
-  pItem->stype = stype;
 
-  // apply new timezone
   osSetTimezone(value);
-
+  return code;
-  return 0;
 }
 
 static int32_t cfgSetTfsItem(SConfig *pCfg, const char *name, const char *value, const char *level, const char *primary,
@@ -342,40 +312,61 @@ static int32_t cfgUpdateDebugFlagItem(SConfig *pCfg, const char *name, bool rese
 
 int32_t cfgSetItem(SConfig *pCfg, const char *name, const char *value, ECfgSrcType stype) {
   // GRANT_CFG_SET;
+  int32_t      code = 0;
   SConfigItem *pItem = cfgGetItem(pCfg, name);
   if (pItem == NULL) {
     terrno = TSDB_CODE_CFG_NOT_FOUND;
     return -1;
   }
 
+  taosThreadMutexLock(&pCfg->lock);
+
   switch (pItem->dtype) {
-    case CFG_DTYPE_BOOL:
+    case CFG_DTYPE_BOOL: {
-      return cfgSetBool(pItem, value, stype);
+      code = cfgSetBool(pItem, value, stype);
-    case CFG_DTYPE_INT32:
+      break;
-      return cfgSetInt32(pItem, value, stype);
+    }
-    case CFG_DTYPE_INT64:
+    case CFG_DTYPE_INT32: {
-      return cfgSetInt64(pItem, value, stype);
+      code = cfgSetInt32(pItem, value, stype);
+      break;
+    }
+    case CFG_DTYPE_INT64: {
+      code = cfgSetInt64(pItem, value, stype);
+      break;
+    }
     case CFG_DTYPE_FLOAT:
-    case CFG_DTYPE_DOUBLE:
+    case CFG_DTYPE_DOUBLE: {
-      return cfgSetFloat(pItem, value, stype);
+      code = cfgSetFloat(pItem, value, stype);
-    case CFG_DTYPE_STRING:
+      break;
-      return cfgSetString(pItem, value, stype);
+    }
-    case CFG_DTYPE_DIR:
+    case CFG_DTYPE_STRING: {
-      return cfgSetDir(pItem, value, stype);
+      code = cfgSetString(pItem, value, stype);
-    case CFG_DTYPE_TIMEZONE:
+      break;
-      return cfgSetTimezone(pItem, value, stype);
+    }
-    case CFG_DTYPE_CHARSET:
+    case CFG_DTYPE_DIR: {
-      return cfgSetCharset(pItem, value, stype);
+      code = cfgSetDir(pItem, value, stype);
-    case CFG_DTYPE_LOCALE:
+      break;
-      return cfgSetLocale(pItem, value, stype);
+    }
+    case CFG_DTYPE_TIMEZONE: {
+      code = cfgSetTimezone(pItem, value, stype);
+      break;
+    }
+    case CFG_DTYPE_CHARSET: {
+      code = doSetConf(pItem, value, stype);
+      break;
+    }
+    case CFG_DTYPE_LOCALE: {
+      code = doSetConf(pItem, value, stype);
+      break;
+    }
     case CFG_DTYPE_NONE:
     default:
+      terrno = TSDB_CODE_INVALID_CFG;
       break;
   }
 
-_err_out:
+  taosThreadMutexUnlock(&pCfg->lock);
-  terrno = TSDB_CODE_INVALID_CFG;
+  return code;
-  return -1;
 }
 
 SConfigItem *cfgGetItem(SConfig *pCfg, const char *name) {
@@ -388,16 +379,16 @@ SConfigItem *cfgGetItem(SConfig *pCfg, const char *name) {
     }
   }
 
-  // uError("name:%s, cfg not found", name);
   terrno = TSDB_CODE_CFG_NOT_FOUND;
   return NULL;
 }
 
 int32_t cfgCheckRangeForDynUpdate(SConfig *pCfg, const char *name, const char *pVal, bool isServer) {
   ECfgDynType  dynType = isServer ? CFG_DYN_SERVER : CFG_DYN_CLIENT;
+
   SConfigItem *pItem = cfgGetItem(pCfg, name);
   if (!pItem || (pItem->dynScope & dynType) == 0) {
-    uError("failed to config:%s, not support", name);
+    uError("failed to config:%s, not support update this config", name);
     terrno = TSDB_CODE_INVALID_CFG;
     return -1;
   }
@@ -459,7 +450,7 @@ static int32_t cfgAddItem(SConfig *pCfg, SConfigItem *pItem, const char *name) {
     return -1;
   }
 
-  int size = pCfg->array->size;
+  int32_t size = taosArrayGetSize(pCfg->array);
   for (int32_t i = 0; i < size; ++i) {
     SConfigItem *existItem = taosArrayGet(pCfg->array, i);
     if (existItem != NULL && strcmp(existItem->name, pItem->name) == 0) {
@@ -559,7 +550,7 @@ int32_t cfgAddDir(SConfig *pCfg, const char *name, const char *defaultVal, int8_
 
 int32_t cfgAddLocale(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope, int8_t dynScope) {
   SConfigItem item = {.dtype = CFG_DTYPE_LOCALE, .scope = scope, .dynScope = dynScope};
-  if (cfgCheckAndSetLocale(&item, defaultVal) != 0) {
+  if (cfgCheckAndSetConf(&item, defaultVal) != 0) {
     return -1;
   }
 
@@ -568,7 +559,7 @@ int32_t cfgAddLocale(SConfig *pCfg, const char *name, const char *defaultVal, in
 
 int32_t cfgAddCharset(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope, int8_t dynScope) {
   SConfigItem item = {.dtype = CFG_DTYPE_CHARSET, .scope = scope, .dynScope = dynScope};
-  if (cfgCheckAndSetCharset(&item, defaultVal) != 0) {
+  if (cfgCheckAndSetConf(&item, defaultVal) != 0) {
     return -1;
   }
 
@@ -577,7 +568,7 @@ int32_t cfgAddCharset(SConfig *pCfg, const char *name, const char *defaultVal, i
 
 int32_t cfgAddTimezone(SConfig *pCfg, const char *name, const char *defaultVal, int8_t scope, int8_t dynScope) {
   SConfigItem item = {.dtype = CFG_DTYPE_TIMEZONE, .scope = scope, .dynScope = dynScope};
-  if (cfgCheckAndSetTimezone(&item, defaultVal) != 0) {
+  if (cfgCheckAndSetConf(&item, defaultVal) != 0) {
     return -1;
   }
 
@@ -1356,3 +1347,35 @@ int32_t cfgGetApollUrl(const char **envCmd, const char *envFile, char *apolloUrl
   uInfo("fail get apollo url from cmd env file");
   return -1;
 }
+
+struct SConfigIter {
+  int32_t  index;
+  SConfig *pConf;
+};
+
+SConfigIter *cfgCreateIter(SConfig *pConf) {
+  SConfigIter* pIter = taosMemoryCalloc(1, sizeof(SConfigIter));
+  if (pIter == NULL) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    return NULL;
+  }
+
+  pIter->pConf = pConf;
+  return pIter;
+}
+
+SConfigItem *cfgNextIter(SConfigIter* pIter) {
+  if (pIter->index < cfgGetSize(pIter->pConf)) {
+    return taosArrayGet(pIter->pConf->array, pIter->index++);
+  }
+
+  return NULL;
+}
+
+void cfgDestroyIter(SConfigIter *pIter) {
+  if (pIter == NULL) {
+    return;
+  }
+
+  taosMemoryFree(pIter);
+}

source/util/src/terror.c

@@ -151,8 +151,12 @@ TAOS_DEFINE_ERROR(TSDB_CODE_TSC_QUERY_KILLED,             "Query killed")
 TAOS_DEFINE_ERROR(TSDB_CODE_TSC_NO_EXEC_NODE,             "No available execution node in current query policy configuration")
 TAOS_DEFINE_ERROR(TSDB_CODE_TSC_NOT_STABLE_ERROR,         "Table is not a super table")
 TAOS_DEFINE_ERROR(TSDB_CODE_TSC_STMT_CACHE_ERROR,         "Stmt cache error")
-TAOS_DEFINE_ERROR(TSDB_CODE_TSC_ENCODE_PARAM_ERROR,       "Invalid compress param")
+TAOS_DEFINE_ERROR(TSDB_CODE_TSC_ENCODE_PARAM_ERROR,       "Invalid encode param")
 TAOS_DEFINE_ERROR(TSDB_CODE_TSC_ENCODE_PARAM_NULL,        "Not found compress param")
+TAOS_DEFINE_ERROR(TSDB_CODE_TSC_COMPRESS_PARAM_ERROR,     "Invalid compress param")
+TAOS_DEFINE_ERROR(TSDB_CODE_TSC_COMPRESS_LEVEL_ERROR,     "Invalid compress level param")
+
+
 TAOS_DEFINE_ERROR(TSDB_CODE_TSC_INTERNAL_ERROR,           "Internal error")
 
 // mnode-common
@@ -221,7 +225,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_MND_COLUMN_NOT_EXIST,         "Column does not exist
 TAOS_DEFINE_ERROR(TSDB_CODE_MND_INVALID_STB_OPTION,       "Invalid stable options")
 TAOS_DEFINE_ERROR(TSDB_CODE_MND_INVALID_ROW_BYTES,        "Invalid row bytes")
 TAOS_DEFINE_ERROR(TSDB_CODE_MND_FIELD_VALUE_OVERFLOW,          "out of range and overflow")
-TAOS_DEFINE_ERROR(TSDB_CODE_MND_COLUMN_COMPRESS_ALREADY_EXIST,        "Column compress already exist")  
+TAOS_DEFINE_ERROR(TSDB_CODE_MND_COLUMN_COMPRESS_ALREADY_EXIST, "Same with old param")  
 
 
 // mnode-func
@@ -397,7 +401,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_VND_ALREADY_IS_VOTER,         "Vnode already is a vo
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_DIR_ALREADY_EXIST,        "Vnode directory already exist")
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_META_DATA_UNSAFE_DELETE,  "Single replica vnode data will lost permanently after this operation, if you make sure this, please use drop dnode <id> unsafe to execute")
 TAOS_DEFINE_ERROR(TSDB_CODE_VND_ARB_NOT_SYNCED,           "Vgroup peer is not synced")
-TAOS_DEFINE_ERROR(TSDB_CODE_VND_COLUMN_COMPRESS_ALREADY_EXIST,"Column compress already exist")
+TAOS_DEFINE_ERROR(TSDB_CODE_VND_COLUMN_COMPRESS_ALREADY_EXIST,"Same with old param")
 
 
 // tsdb

source/util/src/thash.c

@@ -719,6 +719,11 @@ void *taosHashGetKey(void *data, size_t *keyLen) {
   return GET_HASH_NODE_KEY(node);
 }
 
+int32_t taosHashGetValueSize(void *data) {
+  SHashNode *node = GET_HASH_PNODE(data);
+  return node->dataLen;
+}
+
 // release the pNode, return next pNode, and lock the current entry
 static void *taosHashReleaseNode(SHashObj *pHashObj, void *p, int *slot) {
   SHashNode *pOld = (SHashNode *)GET_HASH_PNODE(p);

source/util/test/cfgTest.cpp

@@ -63,9 +63,11 @@ TEST_F(CfgTest, 02_Basic) {
 
   EXPECT_EQ(cfgGetSize(pConfig), 6);
 
-  int32_t size = taosArrayGetSize(pConfig->array);
+  int32_t size = cfgGetSize(pConfig);
-  for (int32_t i = 0; i < size; ++i) {
+
-    SConfigItem *pItem = (SConfigItem *)taosArrayGet(pConfig->array, i);
+  SConfigItem* pItem = NULL;
+  SConfigIter* pIter = cfgCreateIter(pConfig);
+  while((pItem == cfgNextIter(pIter)) != NULL) {
     switch (pItem->dtype) {
       case CFG_DTYPE_BOOL:
         printf("index:%d, cfg:%s value:%d\n", size, pItem->name, pItem->bval);
@@ -90,9 +92,12 @@ TEST_F(CfgTest, 02_Basic) {
         break;
     }
   }
+
+  cfgDestroyIter(pIter);
+
   EXPECT_EQ(cfgGetSize(pConfig), 6);
 
-  SConfigItem *pItem = cfgGetItem(pConfig, "test_bool");
+  pItem = cfgGetItem(pConfig, "test_bool");
   EXPECT_EQ(pItem->stype, CFG_STYPE_DEFAULT);
   EXPECT_EQ(pItem->dtype, CFG_DTYPE_BOOL);
   EXPECT_STREQ(pItem->name, "test_bool");

tests/army/community/cluster/snapshot.py

@@ -30,7 +30,7 @@ from frame.srvCtl import *
 
 class TDTestCase(TBase):
     updatecfgDict = {
-        "countAlwaysReturnValue" : "0",
+        "countAlwaysReturnValue" : "1",
         "lossyColumns"           : "float,double",
         "fPrecision"             : "0.000000001",
         "dPrecision"             : "0.00000000000000001",
@@ -106,7 +106,7 @@ class TDTestCase(TBase):
         # check count always return value
         sql = f"select count(*) from {self.db}.ta"
         tdSql.query(sql)
-        tdSql.checkRows(0) # countAlwaysReturnValue is false
+        tdSql.checkRows(1) # countAlwaysReturnValue is false
 
     # run
     def run(self):

tests/army/community/cmdline/fullopt.py

@@ -49,7 +49,7 @@ class TDTestCase(TBase):
 
     def checkQueryOK(self, rets):
         if rets[-2][:9] != "Query OK,":
-            tdLog.exit(f"check taos -s return unecpect: {rets}")
+            tdLog.exit(f"check taos -s return unexpect: {rets}")
 
     def doTaos(self):
         tdLog.info(f"check taos command options...")

tests/army/community/storage/compressBasic.py

@@ -118,18 +118,10 @@ class TDTestCase(TBase):
         sql = f"describe {self.db}.{self.stb}"
         tdSql.query(sql)
 
-        '''
         # see AutoGen.types 
         defEncodes = [ "delta-i","delta-i","simple8b","simple8b","simple8b","simple8b","simple8b","simple8b",
                        "simple8b","simple8b","delta-d","delta-d","bit-packing",
-                       "disabled","disabled","disabled","disabled","disabled"]
+                       "disabled","disabled","disabled","disabled"]        
-        '''
-        
-        # pass-ci have error
-        defEncodes = [ "delta-i","delta-i","simple8b","simple8b","simple8b","simple8b","simple8b","simple8b",
-                       "simple8b","simple8b","delta-d","delta-d","bit-packing",
-                       "disabled","disabled","disabled","disabled","simple8b"]
-        
 
         count = tdSql.getRows()
         for i in range(count):

tests/army/enterprise/s3/s3Basic.json

@@ -32,7 +32,7 @@
                 {
                     "name": "stb",
                     "child_table_exists": "no",
-                    "childtable_count": 10,
+                    "childtable_count": 6,
                     "insert_rows": 2000000,
                     "childtable_prefix": "d",
                     "insert_mode": "taosc",

tests/army/enterprise/s3/s3Basic.py

@@ -38,6 +38,10 @@ s3EndPoint     http://192.168.1.52:9000
 s3AccessKey    'zOgllR6bSnw2Ah3mCNel:cdO7oXAu3Cqdb1rUdevFgJMi0LtRwCXdWKQx4bhX'
 s3BucketName   ci-bucket
 s3UploadDelaySec 60
+
+for test:
+"s3AccessKey" : "fGPPyYjzytw05nw44ViA:vK1VcwxgSOykicx6hk8fL1x15uEtyDSFU3w4hTaZ"
+"s3BucketName": "test-bucket"
 '''
 
 
@@ -63,7 +67,7 @@ class TDTestCase(TBase):
 
         tdSql.execute(f"use {self.db}")
         # come from s3_basic.json
-        self.childtable_count = 10
+        self.childtable_count = 6
         self.insert_rows = 2000000
         self.timestamp_step = 1000
 
@@ -85,7 +89,7 @@ class TDTestCase(TBase):
             fileName = cols[8]
             #print(f" filesize={fileSize} fileName={fileName}  line={line}")
             if fileSize > maxFileSize:
-                tdLog.info(f"error, {fileSize} over max size({maxFileSize})\n")
+                tdLog.info(f"error, {fileSize} over max size({maxFileSize}) {fileName}\n")
                 overCnt += 1
             else:
                 tdLog.info(f"{fileName}({fileSize}) check size passed.")
@@ -99,7 +103,7 @@ class TDTestCase(TBase):
         loop = 0
         rets = []
         overCnt = 0
-        while loop < 180:
+        while loop < 100:
             time.sleep(3)
 
             # check upload to s3
@@ -335,7 +339,7 @@ class TDTestCase(TBase):
             self.snapshotAgg()
             self.doAction()
             self.checkAggCorrect()
-            self.checkInsertCorrect(difCnt=self.childtable_count*999999)
+            self.checkInsertCorrect(difCnt=self.childtable_count*1499999)
             self.checkDelete()
             self.doAction()
 

tests/army/enterprise/s3/s3Basic1.json

@@ -32,7 +32,7 @@
                 {
                     "name": "stb",
                     "child_table_exists": "yes",
-                    "childtable_count": 10,
+                    "childtable_count": 6,
                     "insert_rows": 1000000,
                     "childtable_prefix": "d",
                     "insert_mode": "taosc",

tests/army/frame/caseBase.py

@@ -140,7 +140,7 @@ class TBase:
 
         # check step
         sql = f"select count(*) from (select diff(ts) as dif from {self.stb} partition by tbname order by ts desc) where dif != {self.timestamp_step}"
-        #tdSql.checkAgg(sql, difCnt)
+        tdSql.checkAgg(sql, difCnt)
 
     # save agg result
     def snapshotAgg(self):        

tests/parallel_test/cases.task

@@ -14,6 +14,7 @@
 ,,y,army,./pytest.sh python3 ./test.py -f enterprise/s3/s3Basic.py -N 3
 ,,y,army,./pytest.sh python3 ./test.py -f community/cluster/snapshot.py -N 3 -L 3 -D 2
 ,,y,army,./pytest.sh python3 ./test.py -f community/query/function/test_func_elapsed.py
+,,y,army,./pytest.sh python3 ./test.py -f community/query/test_join.py
 ,,y,army,./pytest.sh python3 ./test.py -f community/query/fill/fill_desc.py -N 3 -L 3 -D 2
 ,,y,army,./pytest.sh python3 ./test.py -f community/cluster/incSnapshot.py -N 3
 ,,y,army,./pytest.sh python3 ./test.py -f community/query/query_basic.py -N 3
@@ -128,8 +129,8 @@
 ,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/subscribeDb0.py -N 3 -n 3
 ,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/ins_topics_test.py
 ,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/tmqMaxTopic.py
-#,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/tmqParamsTest.py
+,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/tmqParamsTest.py
-#,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/tmqParamsTest.py -R
+,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/tmqParamsTest.py -R
 ,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/tmqClientConsLog.py
 ,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/tmqMaxGroupIds.py
 ,,y,system-test,./pytest.sh python3 ./test.py -f 7-tmq/tmqConsumeDiscontinuousData.py

tests/script/tsim/join/join_scalar1.sim

@@ -1030,7 +1030,7 @@ endi
 sql_error select a.ts from sta a left window join sta b on a.t1 = b.t1 window_offset(-1s, 1s) where a.col1 < 3 or a.col1 > 4 having(count(a.ts) > 1) order by b.col1;
 sql_error select a.ts from sta a left window join sta b on a.t1 = b.t1 window_offset(-1s, 1s) where a.col1 < 3 or a.col1 > 4 having(count(a.ts) > 1) order by b.tbname;
 sql select count(b.col1) from sta a left window join sta b on a.t1 = b.t1 window_offset(-1s, 1s) where (a.col1 < 3 or a.col1 > 4) and b.col1 is not null;
-sql select count(b.col1) from sta a left window join sta b on a.t1 = b.t1 window_offset(-1s, 1s) where a.col1 < 3 or a.col1 > 4 order by a.t1;
+sql select count(b.col1),a.t1,a.ts from sta a left window join sta b on a.t1 = b.t1 window_offset(-1s, 1s) where a.col1 < 3 or a.col1 > 4 order by 2,3;
 if $rows != 5 then
   return -1
 endi

tests/script/tsim/join/join_scalar2.sim

@@ -725,7 +725,7 @@ endi
 sql_error select a.ts from sta b right window join sta a on a.t1 = b.t1 window_offset(-1s, 1s) where a.col1 < 3 or a.col1 > 4 having(count(a.ts) > 1) order by b.col1;
 sql_error select a.ts from sta b right window join sta a on a.t1 = b.t1 window_offset(-1s, 1s) where a.col1 < 3 or a.col1 > 4 having(count(a.ts) > 1) order by b.tbname;
 sql select count(b.col1) from sta b right window join sta a on a.t1 = b.t1 window_offset(-1s, 1s) where (a.col1 < 3 or a.col1 > 4) and b.col1 is not null;
-sql select count(b.col1) from sta b right window join sta a on a.t1 = b.t1 window_offset(-1s, 1s) where a.col1 < 3 or a.col1 > 4 order by a.t1;
+sql select count(b.col1),a.t1,a.ts from sta b right window join sta a on a.t1 = b.t1 window_offset(-1s, 1s) where a.col1 < 3 or a.col1 > 4 order by 2,3;
 if $rows != 5 then
   return -1
 endi

tests/system-test/0-others/multilevel.py

@@ -182,10 +182,10 @@ class TDTestCase:
             else:
                 checkFiles("%s/*/*" % i, 0)
     
-    def more_than_16_disks(self):
+    def more_than_128_disks(self):
-        tdLog.info("============== more_than_16_disks test ===============")
+        tdLog.info("============== more_than_128_disks test ===============")
         cfg={}
-        for i in range(17):
+        for i in range(129):
             if i == 0 :
                 datadir = '/mnt/data%d 0 1' % (i+1)
             else:
@@ -272,7 +272,7 @@ class TDTestCase:
         self.dir_permission_denied()
         self.file_distribution_same_level()
         self.three_level_basic()
-        self.more_than_16_disks()
+        self.more_than_128_disks()
         self.trim_database()
         self.missing_middle_level()
         

tests/system-test/2-query/pk_error.py

@@ -88,7 +88,7 @@ class TDTestCase:
         tdSql.checkData(1, 1, 8)
         tdSql.checkData(1, 2, 8)
 
-        tdSql.query('select ts,diff(f) from d1.st partition by t order by t;')
+        tdSql.query('select ts,diff(f),t from d1.st partition by t order by 3,1;')
         tdSql.checkRows(4)
         tdSql.checkData(0, 0, datetime.datetime(2021, 4, 19, 0, 0, 1))
         tdSql.checkData(0, 1, 0)
@@ -104,7 +104,7 @@ class TDTestCase:
         tdSql.checkData(0, 0, 4.0)
         tdSql.checkData(1, 0, 5.0)
 
-        tdSql.query('select ts,derivative(f, 1s, 0) from d1.st partition by t order by t;')
+        tdSql.query('select ts,derivative(f, 1s, 0),t from d1.st partition by t order by 3,1;')
         tdSql.checkRows(4)
         tdSql.checkData(0, 0, datetime.datetime(2021, 4, 19, 0, 0, 1))
         tdSql.checkData(0, 1, 0.0)

tests/system-test/win-test-file

@@ -61,8 +61,8 @@ python3 ./test.py -f 7-tmq/subscribeStb3.py
 python3 ./test.py -f 7-tmq/subscribeDb0.py -N 3 -n 3
 python3 ./test.py -f 7-tmq/ins_topics_test.py
 python3 ./test.py -f 7-tmq/tmqMaxTopic.py
-#python3 ./test.py -f 7-tmq/tmqParamsTest.py
+python3 ./test.py -f 7-tmq/tmqParamsTest.py
-#python3 ./test.py -f 7-tmq/tmqParamsTest.py -R
+python3 ./test.py -f 7-tmq/tmqParamsTest.py -R
 python3 ./test.py -f 7-tmq/tmqClientConsLog.py
 python3 ./test.py -f 7-tmq/tmqMaxGroupIds.py
 python3 ./test.py -f 7-tmq/tmqConsumeDiscontinuousData.py