merge from 3.0

include/client/taos.h

@@ -124,8 +124,25 @@ typedef struct TAOS_MULTI_BIND {
   int       num;
 } TAOS_MULTI_BIND;
 
+typedef enum {
+  SET_CONF_RET_SUCC = 0,
+  SET_CONF_RET_ERR_PART = -1,
+  SET_CONF_RET_ERR_INNER = -2,
+  SET_CONF_RET_ERR_JSON_INVALID = -3,
+  SET_CONF_RET_ERR_JSON_PARSE = -4,
+  SET_CONF_RET_ERR_ONLY_ONCE = -5,
+  SET_CONF_RET_ERR_TOO_LONG = -6
+} SET_CONF_RET_CODE;
+
+#define RET_MSG_LENGTH 1024
+typedef struct setConfRet {
+  SET_CONF_RET_CODE retCode;
+  char   retMsg[RET_MSG_LENGTH];
+} setConfRet;
+
 DLL_EXPORT void  taos_cleanup(void);
 DLL_EXPORT int   taos_options(TSDB_OPTION option, const void *arg, ...);
+DLL_EXPORT setConfRet   taos_set_config(const char *config);
 DLL_EXPORT TAOS *taos_connect(const char *ip, const char *user, const char *pass, const char *db, uint16_t port);
 DLL_EXPORT TAOS *taos_connect_l(const char *ip, int ipLen, const char *user, int userLen, const char *pass, int passLen,
                                 const char *db, int dbLen, uint16_t port);
@@ -168,10 +185,13 @@ DLL_EXPORT int         taos_select_db(TAOS *taos, const char *db);
 DLL_EXPORT int         taos_print_row(char *str, TAOS_ROW row, TAOS_FIELD *fields, int num_fields);
 DLL_EXPORT void        taos_stop_query(TAOS_RES *res);
 DLL_EXPORT bool        taos_is_null(TAOS_RES *res, int32_t row, int32_t col);
+DLL_EXPORT bool        taos_is_update_query(TAOS_RES *res);
 DLL_EXPORT int         taos_fetch_block(TAOS_RES *res, TAOS_ROW *rows);
 DLL_EXPORT int         taos_validate_sql(TAOS *taos, const char *sql);
+DLL_EXPORT void taos_reset_current_db(TAOS *taos);
 
 DLL_EXPORT int *taos_fetch_lengths(TAOS_RES *res);
+DLL_EXPORT TAOS_ROW *taos_result_block(TAOS_RES *res);
 
 DLL_EXPORT const char *taos_get_server_info(TAOS *taos);
 DLL_EXPORT const char *taos_get_client_info();

include/common/tdatablock.h

@@ -133,7 +133,8 @@ static FORCE_INLINE int32_t colDataAppendInt32(SColumnInfoData* pColumnInfoData,
 }
 
 static FORCE_INLINE int32_t colDataAppendInt64(SColumnInfoData* pColumnInfoData, uint32_t currentRow, int64_t* v) {
-  ASSERT(pColumnInfoData->info.type == TSDB_DATA_TYPE_BIGINT || pColumnInfoData->info.type == TSDB_DATA_TYPE_UBIGINT);
+  int32_t type = pColumnInfoData->info.type;
+  ASSERT(type == TSDB_DATA_TYPE_BIGINT || type == TSDB_DATA_TYPE_UBIGINT || type == TSDB_DATA_TYPE_TIMESTAMP);
   char* p = pColumnInfoData->pData + pColumnInfoData->info.bytes * currentRow;
   *(int64_t*)p = *(int64_t*)v;
 }
@@ -175,17 +176,16 @@ size_t blockDataGetRowSize(SSDataBlock* pBlock);
 double blockDataGetSerialRowSize(const SSDataBlock* pBlock);
 size_t blockDataGetSerialMetaSize(const SSDataBlock* pBlock);
 
-SSchema* blockDataExtractSchema(const SSDataBlock* pBlock, int32_t* numOfCols);
-
 int32_t blockDataSort(SSDataBlock* pDataBlock, SArray* pOrderInfo);
 int32_t blockDataSort_rv(SSDataBlock* pDataBlock, SArray* pOrderInfo, bool nullFirst);
 
-int32_t      blockDataEnsureColumnCapacity(SColumnInfoData* pColumn, uint32_t numOfRows);
+int32_t blockDataEnsureColumnCapacity(SColumnInfoData* pColumn, uint32_t numOfRows);
-int32_t      blockDataEnsureCapacity(SSDataBlock* pDataBlock, uint32_t numOfRows);
+int32_t blockDataEnsureCapacity(SSDataBlock* pDataBlock, uint32_t numOfRows);
-void         blockDataCleanup(SSDataBlock* pDataBlock);
+void    blockDataCleanup(SSDataBlock* pDataBlock);
+size_t  blockDataGetCapacityInRow(const SSDataBlock* pBlock, size_t pageSize);
+void*   blockDataDestroy(SSDataBlock* pBlock);
+
 SSDataBlock* createOneDataBlock(const SSDataBlock* pDataBlock);
-size_t       blockDataGetCapacityInRow(const SSDataBlock* pBlock, size_t pageSize);
-void*        blockDataDestroy(SSDataBlock* pBlock);
 
 void blockDebugShowData(const SArray* dataBlocks);
 

include/common/tglobal.h

@@ -51,7 +51,7 @@ extern int32_t tsCompatibleModel;
 extern bool    tsEnableSlaveQuery;
 extern bool    tsPrintAuth;
 extern int64_t tsTickPerDay[3];
-extern int32_t tsMultiProcess;
+extern bool    tsMultiProcess;
 
 // monitor
 extern bool     tsEnableMonitor;

include/common/tmsg.h

@@ -469,8 +469,7 @@ typedef struct {
   int32_t tz;  // query client timezone
   char    intervalUnit;
   char    slidingUnit;
-  char
+  char    offsetUnit;  // TODO Remove it, the offset is the number of precision tickle, and it must be a immutable duration.
-      offsetUnit;  // TODO Remove it, the offset is the number of precision tickle, and it must be a immutable duration.
   int8_t  precision;
   int64_t interval;
   int64_t sliding;
@@ -2017,7 +2016,6 @@ typedef struct {
   int8_t   slidingUnit;   // MACRO: TIME_UNIT_XXX
   int8_t   timezoneInt;   // sma data expired if timezone changes.
   char     indexName[TSDB_INDEX_NAME_LEN];
-  char     timezone[TD_TIMEZONE_LEN];
   int32_t  exprLen;
   int32_t  tagsFilterLen;
   int64_t  indexUid;
@@ -2055,32 +2053,6 @@ void*   tDeserializeSVCreateTSmaReq(void* buf, SVCreateTSmaReq* pReq);
 int32_t tSerializeSVDropTSmaReq(void** buf, SVDropTSmaReq* pReq);
 void*   tDeserializeSVDropTSmaReq(void* buf, SVDropTSmaReq* pReq);
 
-typedef struct {
-  col_id_t colId;
-  uint16_t blockSize;  // sma data block size
-  char     data[];
-} STSmaColData;
-
-typedef struct {
-  tb_uid_t tableUid;  // super/child/normal table uid
-  int32_t  dataLen;   // not including head
-  char     data[];
-} STSmaTbData;
-
-typedef struct {
-  int64_t indexUid;
-  TSKEY   skey;  // startKey of one interval/sliding window
-  int64_t interval;
-  int32_t dataLen;  // not including head
-  int8_t  intervalUnit;
-  char    data[];
-} STSmaDataWrapper;  // sma data for a interval/sliding window
-
-// interval/sliding => window
-
-// => window->table->colId
-// => 当一个window下所有的表均计算完成时，流计算告知tsdb清除window的过期标记
-
 // RSma: Rollup SMA
 typedef struct {
   int64_t  interval;

include/common/trow.h

@@ -937,9 +937,8 @@ static FORCE_INLINE bool tdSTSRowIterNext(STSRowIter *pIter, col_id_t colId, col
     STColumn *pCol = NULL;
     STSchema *pSchema = pIter->pSchema;
     while (pIter->colIdx <= pSchema->numOfCols) {
-      pCol = &pSchema->columns[pIter->colIdx];
+      pCol = &pSchema->columns[pIter->colIdx];  // 1st column of schema is primary TS key
       if (colId == pCol->colId) {
-        ++pIter->colIdx;
         break;
       } else if (colId < pCol->colId) {
         ++pIter->colIdx;
@@ -948,7 +947,8 @@ static FORCE_INLINE bool tdSTSRowIterNext(STSRowIter *pIter, col_id_t colId, col
         return false;
       }
     }
-    return tdGetTpRowDataOfCol(pIter, pCol->type, pCol->offset - sizeof(TSKEY), pVal);
+    tdGetTpRowDataOfCol(pIter, pCol->type, pCol->offset - sizeof(TSKEY), pVal);
+    ++pIter->colIdx;
   } else if (TD_IS_KV_ROW(pIter->pRow)) {
     return tdGetKvRowValOfColEx(pIter, colId, colType, &pIter->kvIdx, pVal);
   } else {

include/common/ttokendef.h

@@ -138,55 +138,59 @@
 #define TK_INTERVAL                       120
 #define TK_TOPIC                          121
 #define TK_AS                             122
-#define TK_NK_BOOL                        123
+#define TK_EXPLAIN                        123
-#define TK_NULL                           124
+#define TK_ANALYZE                        124
-#define TK_NK_VARIABLE                    125
+#define TK_VERBOSE                        125
-#define TK_NK_UNDERLINE                   126
+#define TK_NK_BOOL                        126
-#define TK_ROWTS                          127
+#define TK_RATIO                          127
-#define TK_TBNAME                         128
+#define TK_NULL                           128
-#define TK_QSTARTTS                       129
+#define TK_NK_VARIABLE                    129
-#define TK_QENDTS                         130
+#define TK_NK_UNDERLINE                   130
-#define TK_WSTARTTS                       131
+#define TK_ROWTS                          131
-#define TK_WENDTS                         132
+#define TK_TBNAME                         132
-#define TK_WDURATION                      133
+#define TK_QSTARTTS                       133
-#define TK_BETWEEN                        134
+#define TK_QENDTS                         134
-#define TK_IS                             135
+#define TK_WSTARTTS                       135
-#define TK_NK_LT                          136
+#define TK_WENDTS                         136
-#define TK_NK_GT                          137
+#define TK_WDURATION                      137
-#define TK_NK_LE                          138
+#define TK_BETWEEN                        138
-#define TK_NK_GE                          139
+#define TK_IS                             139
-#define TK_NK_NE                          140
+#define TK_NK_LT                          140
-#define TK_MATCH                          141
+#define TK_NK_GT                          141
-#define TK_NMATCH                         142
+#define TK_NK_LE                          142
-#define TK_IN                             143
+#define TK_NK_GE                          143
-#define TK_JOIN                           144
+#define TK_NK_NE                          144
-#define TK_INNER                          145
+#define TK_MATCH                          145
-#define TK_SELECT                         146
+#define TK_NMATCH                         146
-#define TK_DISTINCT                       147
+#define TK_IN                             147
-#define TK_WHERE                          148
+#define TK_JOIN                           148
-#define TK_PARTITION                      149
+#define TK_INNER                          149
-#define TK_BY                             150
+#define TK_SELECT                         150
-#define TK_SESSION                        151
+#define TK_DISTINCT                       151
-#define TK_STATE_WINDOW                   152
+#define TK_WHERE                          152
-#define TK_SLIDING                        153
+#define TK_PARTITION                      153
-#define TK_FILL                           154
+#define TK_BY                             154
-#define TK_VALUE                          155
+#define TK_SESSION                        155
-#define TK_NONE                           156
+#define TK_STATE_WINDOW                   156
-#define TK_PREV                           157
+#define TK_SLIDING                        157
-#define TK_LINEAR                         158
+#define TK_FILL                           158
-#define TK_NEXT                           159
+#define TK_VALUE                          159
-#define TK_GROUP                          160
+#define TK_NONE                           160
-#define TK_HAVING                         161
+#define TK_PREV                           161
-#define TK_ORDER                          162
+#define TK_LINEAR                         162
-#define TK_SLIMIT                         163
+#define TK_NEXT                           163
-#define TK_SOFFSET                        164
+#define TK_GROUP                          164
-#define TK_LIMIT                          165
+#define TK_HAVING                         165
-#define TK_OFFSET                         166
+#define TK_ORDER                          166
-#define TK_ASC                            167
+#define TK_SLIMIT                         167
-#define TK_DESC                           168
+#define TK_SOFFSET                        168
-#define TK_NULLS                          169
+#define TK_LIMIT                          169
-#define TK_FIRST                          170
+#define TK_OFFSET                         170
-#define TK_LAST                           171
+#define TK_ASC                            171
+#define TK_DESC                           172
+#define TK_NULLS                          173
+#define TK_FIRST                          174
+#define TK_LAST                           175
 
 #define TK_NK_SPACE                       300
 #define TK_NK_COMMENT                     301

include/dnode/mgmt/dnode.h

@@ -46,11 +46,12 @@ typedef struct {
   char      localFqdn[TSDB_FQDN_LEN];
   char      firstEp[TSDB_EP_LEN];
   char      secondEp[TSDB_EP_LEN];
-  SDiskCfg *pDisks;
+  SDiskCfg *disks;
   int32_t   numOfDisks;
+  int8_t    ntype;
 } SDnodeOpt;
 
-typedef enum { DND_EVENT_START, DND_EVENT_STOP = 1, DND_EVENT_RELOAD } EDndEvent;
+typedef enum { DND_EVENT_START, DND_EVENT_STOP = 1, DND_EVENT_CHILD } EDndEvent;
 
 /**
  * @brief Initialize and start the dnode.

include/libs/executor/executor.h

@@ -37,6 +37,11 @@ typedef struct SReadHandle {
 #define STREAM_DATA_TYPE_SUBMIT_BLOCK 0x1
 #define STREAM_DATA_TYPE_SSDATA_BLOCK 0x2
 
+typedef enum {
+  OPTR_EXEC_MODEL_BATCH  = 0x1,
+  OPTR_EXEC_MODEL_STREAM = 0x2,
+} EOPTR_EXEC_MODEL;
+
 /**
  * Create the exec task for streaming mode
  * @param pMsg
@@ -84,7 +89,7 @@ int32_t qUpdateQualifiedTableId(qTaskInfo_t tinfo, SArray* tableIdList, bool isA
  * @return
  */
 int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, struct SSubplan* pPlan,
-                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle);
+                        qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, EOPTR_EXEC_MODEL model);
 
 /**
  * The main task execution function, including query on both table and multiple tables,

include/libs/function/function.h

@@ -41,6 +41,7 @@ typedef void (*FExecFinalize)(struct SqlFunctionCtx *pCtx);
 typedef int32_t (*FScalarExecProcess)(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 
 typedef struct SScalarFuncExecFuncs {
+  FExecGetEnv getEnv;
   FScalarExecProcess process;
 } SScalarFuncExecFuncs;
 
@@ -241,7 +242,6 @@ typedef struct tExprNode {
   };
 } tExprNode;
 
-void exprTreeToBinary(SBufferWriter* bw, tExprNode* pExprTree);
 void tExprTreeDestroy(tExprNode *pNode, void (*fp)(void *));
 
 typedef struct SAggFunctionInfo {
@@ -267,28 +267,6 @@ struct SScalarParam {
   int32_t            numOfRows;
 };
 
-typedef struct SMultiFunctionsDesc {
-  bool stableQuery;
-  bool groupbyColumn;
-  bool agg;
-  bool arithmeticOnAgg;
-  bool projectionQuery;
-  bool hasFilter;
-  bool onlyTagQuery;
-  bool orderProjectQuery;
-  bool globalMerge;
-  bool multigroupResult;
-  bool blockDistribution;
-  bool stateWindow;
-  bool timewindow;
-  bool sessionWindow;
-  bool topbotQuery;
-  bool interpQuery;
-  bool distinct;
-  bool join;
-  bool continueQuery;
-} SMultiFunctionsDesc;
-
 int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionId, int32_t param, SResultDataInfo* pInfo, int16_t extLength,
                           bool isSuperTable);
 
@@ -296,8 +274,6 @@ bool qIsValidUdf(SArray* pUdfInfo, const char* name, int32_t len, int32_t* funct
 
 tExprNode* exprTreeFromBinary(const void* data, size_t size);
 
-void extractFunctionDesc(SArray* pFunctionIdList, SMultiFunctionsDesc* pDesc);
-
 tExprNode* exprdup(tExprNode* pTree);
 
 void resetResultRowEntryResult(SqlFunctionCtx* pCtx, int32_t num);

include/libs/nodes/nodes.h

@@ -70,6 +70,7 @@ typedef enum ENodeType {
   QUERY_NODE_DATABASE_OPTIONS,
   QUERY_NODE_TABLE_OPTIONS,
   QUERY_NODE_INDEX_OPTIONS,
+  QUERY_NODE_EXPLAIN_OPTIONS,
 
   // Statement nodes are used in parser and planner module.
   QUERY_NODE_SET_OPERATOR,
@@ -99,6 +100,7 @@ typedef enum ENodeType {
   QUERY_NODE_CREATE_TOPIC_STMT,
   QUERY_NODE_DROP_TOPIC_STMT,
   QUERY_NODE_ALTER_LOCAL_STMT,
+  QUERY_NODE_EXPLAIN_STMT,
   QUERY_NODE_SHOW_DATABASES_STMT,
   QUERY_NODE_SHOW_TABLES_STMT,
   QUERY_NODE_SHOW_STABLES_STMT,

include/libs/nodes/plannodes.h

@@ -49,6 +49,7 @@ typedef struct SScanLogicNode {
   STimeWindow scanRange;
   SName tableName;
   bool showRewrite;
+  double ratio;
 } SScanLogicNode;
 
 typedef struct SJoinLogicNode {
@@ -197,7 +198,7 @@ typedef struct STableScanPhysiNode {
   SScanPhysiNode scan;
   uint8_t scanFlag;         // denotes reversed scan of data or not
   STimeWindow scanRange;
-  SNode* pScanConditions;
+  double ratio;
 } STableScanPhysiNode;
 
 typedef STableScanPhysiNode STableSeqScanPhysiNode;
@@ -252,6 +253,7 @@ typedef struct SIntervalPhysiNode {
   int64_t    sliding;
   int8_t     intervalUnit;
   int8_t     slidingUnit;
+  uint8_t precision;
   SFillNode* pFill;
 } SIntervalPhysiNode;
 
@@ -297,18 +299,23 @@ typedef struct SSubplan {
   SDataSinkNode* pDataSink;    // data of the subplan flow into the datasink
 } SSubplan;
 
-typedef enum EQueryMode {
+typedef enum EExplainMode {
-  QUERY_MODE_NORMAL = 1,
+  EXPLAIN_MODE_DISABLE = 1,
-  QUERY_MODE_EXPLAIN,
+  EXPLAIN_MODE_STATIC,
-  QUERY_MODE_EXPLAIN_AN
+  EXPLAIN_MODE_ANALYZE
-} EQueryMode;
+} EExplainMode;
+
+typedef struct SExplainInfo {
+  EExplainMode mode;
+  bool verbose;
+} SExplainInfo;
 
 typedef struct SQueryPlan {
   ENodeType type;
   uint64_t queryId;
   int32_t numOfSubplans;
-
   SNodeList* pSubplans; // Element is SNodeListNode. The execution level of subplan, starting from 0.
+  SExplainInfo explainInfo;
 } SQueryPlan;
 
 #ifdef __cplusplus

include/libs/nodes/querynodes.h

@@ -131,6 +131,7 @@ typedef struct SRealTableNode {
   struct STableMeta* pMeta;
   SVgroupsInfo* pVgroupList;
   char useDbName[TSDB_DB_NAME_LEN];
+  double ratio;
 } SRealTableNode;
 
 typedef struct STempTableNode {
@@ -282,6 +283,19 @@ typedef struct SVnodeModifOpStmt {
   const char* sql;                 // current sql statement position
 } SVnodeModifOpStmt;
 
+typedef struct SExplainOptions {
+  ENodeType type;
+  bool verbose;
+  double ratio;
+} SExplainOptions;
+
+typedef struct SExplainStmt {
+  ENodeType type;
+  bool analyze;
+  SExplainOptions* pOptions;
+  SNode* pQuery;
+} SExplainStmt;
+
 void nodesWalkSelectStmt(SSelectStmt* pSelect, ESqlClause clause, FNodeWalker walker, void* pContext);
 void nodesRewriteSelectStmt(SSelectStmt* pSelect, ESqlClause clause, FNodeRewriter rewriter, void* pContext);
 

include/libs/scalar/scalar.h

@@ -58,6 +58,14 @@ int32_t ceilFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutp
 int32_t floorFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 int32_t roundFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 
+bool getTimePseudoFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
+
+int32_t winStartTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t winEndTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t winDurFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t qStartTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t qEndTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+
 #ifdef __cplusplus
 }
 #endif

include/os/os.h

@@ -73,7 +73,6 @@ extern "C" {
 #include <wchar.h>
 #include <wctype.h>
 
-
 #include "osAtomic.h"
 #include "osDef.h"
 #include "osDir.h"
@@ -83,10 +82,12 @@ extern "C" {
 #include "osLz4.h"
 #include "osMath.h"
 #include "osMemory.h"
+#include "osProc.h"
 #include "osRand.h"
 #include "osThread.h"
 #include "osSemaphore.h"
 #include "osSignal.h"
+#include "osShm.h"
 #include "osSleep.h"
 #include "osSocket.h"
 #include "osString.h"

include/os/osProc.h

@@ -0,0 +1,33 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_OS_PROC_H_
+#define _TD_OS_PROC_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// start a copy of itself
+int32_t taosNewProc(const char *args);
+
+// the length of the new name must be less than the original name to take effect
+void taosSetProcName(char **argv, const char *name);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_OS_PROC_H_*/

include/os/osShm.h

@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_OS_SHM_H_
+#define _TD_OS_SHM_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+  int32_t id;
+  int32_t size;
+  void*   ptr;
+} SShm;
+
+int32_t taosCreateShm(SShm *pShm, int32_t shmsize) ;
+void    taosDropShm(SShm *pShm);
+int32_t taosAttachShm(SShm *pShm);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*_TD_OS_SHM_H_*/

include/os/osSignal.h

@@ -49,6 +49,8 @@ void taosSetSignal(int32_t signum, FSignalHandler sigfp);
 void taosIgnSignal(int32_t signum);
 void taosDflSignal(int32_t signum);
 
+void taosKillChildOnSelfStopped();
+
 #ifdef __cplusplus
 }
 #endif

include/os/osThread.h

@@ -32,6 +32,9 @@ typedef pthread_once_t TdThreadOnce;
 typedef pthread_rwlockattr_t TdThreadRwlockAttr;
 typedef pthread_cond_t TdThreadCond;
 typedef pthread_condattr_t TdThreadCondAttr;
+typedef pthread_key_t TdThreadKey;
+typedef pthread_barrier_t TdThreadBarrier;
+typedef pthread_barrierattr_t TdThreadBarrierAttr;
 
 #define taosThreadCleanupPush pthread_cleanup_push
 #define taosThreadCleanupPop pthread_cleanup_pop
@@ -49,68 +52,180 @@ typedef pthread_condattr_t TdThreadCondAttr;
     #define pthread_rwlockattr_t PTHREAD_RWLOCKATTR_T_TYPE_TAOS_FORBID
     #define pthread_cond_t PTHREAD_COND_T_TYPE_TAOS_FORBID
     #define pthread_condattr_t PTHREAD_CONDATTR_T_TYPE_TAOS_FORBID
-    #define pthread_spin_init PTHREAD_SPIN_INIT_FUNC_TAOS_FORBID
+    #define pthread_key_t PTHREAD_KEY_T_TYPE_TAOS_FORBID
-    #define pthread_mutex_init PTHREAD_MUTEX_INIT_FUNC_TAOS_FORBID
+    #define pthread_barrier_t PTHREAD_BARRIER_T_TYPE_TAOS_FORBID
-    #define pthread_spin_destroy PTHREAD_SPIN_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_barrierattr_t PTHREAD_BARRIERATTR_T_TYPE_TAOS_FORBID
-    #define pthread_mutex_destroy PTHREAD_MUTEX_DESTROY_FUNC_TAOS_FORBID
-    #define pthread_spin_lock PTHREAD_SPIN_LOCK_FUNC_TAOS_FORBID
-    #define pthread_mutex_lock PTHREAD_MUTEX_LOCK_FUNC_TAOS_FORBID
-    #define pthread_spin_unlock PTHREAD_SPIN_UNLOCK_FUNC_TAOS_FORBID
-    #define pthread_mutex_unlock PTHREAD_MUTEX_UNLOCK_FUNC_TAOS_FORBID
-    #define pthread_rwlock_rdlock PTHREAD_RWLOCK_RDLOCK_FUNC_TAOS_FORBID
-    #define pthread_rwlock_wrlock PTHREAD_RWLOCK_WRLOCK_FUNC_TAOS_FORBID
-    #define pthread_rwlock_unlock PTHREAD_RWLOCK_UNLOCK_FUNC_TAOS_FORBID
-    #define pthread_testcancel PTHREAD_TESTCANCEL_FUNC_TAOS_FORBID
-    #define pthread_attr_init PTHREAD_ATTR_INIT_FUNC_TAOS_FORBID
     #define pthread_create PTHREAD_CREATE_FUNC_TAOS_FORBID
-    #define pthread_once PTHREAD_ONCE_FUNC_TAOS_FORBID
-    #define pthread_attr_setdetachstate PTHREAD_ATTR_SETDETACHSTATE_FUNC_TAOS_FORBID
     #define pthread_attr_destroy PTHREAD_ATTR_DESTROY_FUNC_TAOS_FORBID
-    #define pthread_join PTHREAD_JOIN_FUNC_TAOS_FORBID
+    #define pthread_attr_getdetachstate PTHREAD_ATTR_GETDETACHSTATE_FUNC_TAOS_FORBID
-    #define pthread_rwlock_init PTHREAD_RWLOCK_INIT_FUNC_TAOS_FORBID
+    #define pthread_attr_getinheritsched PTHREAD_ATTR_GETINHERITSCHED_FUNC_TAOS_FORBID
-    #define pthread_rwlock_destroy PTHREAD_RWLOCK_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_attr_getschedparam PTHREAD_ATTR_GETSCHEDPARAM_FUNC_TAOS_FORBID
-    #define pthread_cond_signal PTHREAD_COND_SIGNAL_FUNC_TAOS_FORBID
+    #define pthread_attr_getschedpolicy PTHREAD_ATTR_GETSCHEDPOLICY_FUNC_TAOS_FORBID
-    #define pthread_cond_init PTHREAD_COND_INIT_FUNC_TAOS_FORBID
+    #define pthread_attr_getscope PTHREAD_ATTR_GETSCOPE_FUNC_TAOS_FORBID
-    #define pthread_cond_broadcast PTHREAD_COND_BROADCAST_FUNC_TAOS_FORBID
+    #define pthread_attr_getstacksize PTHREAD_ATTR_GETSTACKSIZE_FUNC_TAOS_FORBID
-    #define pthread_cond_destroy PTHREAD_COND_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_attr_init PTHREAD_ATTR_INIT_FUNC_TAOS_FORBID
-    #define pthread_cond_wait PTHREAD_COND_WAIT_FUNC_TAOS_FORBID
+    #define pthread_attr_setdetachstate PTHREAD_ATTR_SETDETACHSTATE_FUNC_TAOS_FORBID
-    #define pthread_self PTHREAD_SELF_FUNC_TAOS_FORBID
+    #define pthread_attr_setinheritsched PTHREAD_ATTR_SETINHERITSCHED_FUNC_TAOS_FORBID
-    #define pthread_equal PTHREAD_EQUAL_FUNC_TAOS_FORBID
+    #define pthread_attr_setschedparam PTHREAD_ATTR_SETSCHEDPARAM_FUNC_TAOS_FORBID
-    #define pthread_sigmask PTHREAD_SIGMASK_FUNC_TAOS_FORBID
+    #define pthread_attr_setschedpolicy PTHREAD_ATTR_SETSCHEDPOLICY_FUNC_TAOS_FORBID
+    #define pthread_attr_setscope PTHREAD_ATTR_SETSCOPE_FUNC_TAOS_FORBID
+    #define pthread_attr_setstacksize PTHREAD_ATTR_SETSTACKSIZE_FUNC_TAOS_FORBID
+    #define pthread_barrier_destroy PTHREAD_BARRIER_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_barrier_init PTHREAD_BARRIER_INIT_FUNC_TAOS_FORBID
+    #define pthread_barrier_wait PTHREAD_BARRIER_WAIT_FUNC_TAOS_FORBID
+    #define pthread_barrierattr_destroy PTHREAD_BARRIERATTR_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_barrierattr_getpshared PTHREAD_BARRIERATTR_GETPSHARED_FUNC_TAOS_FORBID
+    #define pthread_barrierattr_init PTHREAD_BARRIERATTR_INIT_FUNC_TAOS_FORBID
+    #define pthread_barrierattr_setpshared PTHREAD_BARRIERATTR_SETPSHARED_FUNC_TAOS_FORBID
     #define pthread_cancel PTHREAD_CANCEL_FUNC_TAOS_FORBID
+    #define pthread_cond_destroy PTHREAD_COND_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_cond_init PTHREAD_COND_INIT_FUNC_TAOS_FORBID
+    #define pthread_cond_signal PTHREAD_COND_SIGNAL_FUNC_TAOS_FORBID
+    #define pthread_cond_broadcast PTHREAD_COND_BROADCAST_FUNC_TAOS_FORBID
+    #define pthread_cond_wait PTHREAD_COND_WAIT_FUNC_TAOS_FORBID
+    #define pthread_cond_timedwait PTHREAD_COND_TIMEDWAIT_FUNC_TAOS_FORBID
+    #define pthread_condattr_destroy PTHREAD_CONDATTR_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_condattr_getpshared PTHREAD_CONDATTR_GETPSHARED_FUNC_TAOS_FORBID
+    #define pthread_condattr_init PTHREAD_CONDATTR_INIT_FUNC_TAOS_FORBID
+    #define pthread_condattr_setpshared PTHREAD_CONDATTR_SETPSHARED_FUNC_TAOS_FORBID
+    #define pthread_detach PTHREAD_DETACH_FUNC_TAOS_FORBID
+    #define pthread_equal PTHREAD_EQUAL_FUNC_TAOS_FORBID
+    #define pthread_exit PTHREAD_EXIT_FUNC_TAOS_FORBID
+    #define pthread_getschedparam PTHREAD_GETSCHEDPARAM_FUNC_TAOS_FORBID
+    #define pthread_getspecific PTHREAD_GETSPECIFIC_FUNC_TAOS_FORBID
+    #define pthread_join PTHREAD_JOIN_FUNC_TAOS_FORBID
+    #define pthread_key_create PTHREAD_KEY_CREATE_FUNC_TAOS_FORBID
+    #define pthread_key_delete PTHREAD_KEY_DELETE_FUNC_TAOS_FORBID
     #define pthread_kill PTHREAD_KILL_FUNC_TAOS_FORBID
+    #define pthread_mutex_consistent PTHREAD_MUTEX_CONSISTENT_FUNC_TAOS_FORBID
+    #define pthread_mutex_destroy PTHREAD_MUTEX_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_mutex_init PTHREAD_MUTEX_INIT_FUNC_TAOS_FORBID
+    #define pthread_mutex_lock PTHREAD_MUTEX_LOCK_FUNC_TAOS_FORBID
+    #define pthread_mutex_timedlock PTHREAD_MUTEX_TIMEDLOCK_FUNC_TAOS_FORBID
+    #define pthread_mutex_trylock PTHREAD_MUTEX_TRYLOCK_FUNC_TAOS_FORBID
+    #define pthread_mutex_unlock PTHREAD_MUTEX_UNLOCK_FUNC_TAOS_FORBID
+    #define pthread_mutexattr_destroy PTHREAD_MUTEXATTR_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_mutexattr_getpshared PTHREAD_MUTEXATTR_GETPSHARED_FUNC_TAOS_FORBID
+    #define pthread_mutexattr_getrobust PTHREAD_MUTEXATTR_GETROBUST_FUNC_TAOS_FORBID
+    #define pthread_mutexattr_gettype PTHREAD_MUTEXATTR_GETTYPE_FUNC_TAOS_FORBID
+    #define pthread_mutexattr_init PTHREAD_MUTEXATTR_INIT_FUNC_TAOS_FORBID
+    #define pthread_mutexattr_setpshared PTHREAD_MUTEXATTR_SETPSHARED_FUNC_TAOS_FORBID
+    #define pthread_mutexattr_setrobust PTHREAD_MUTEXATTR_SETROBUST_FUNC_TAOS_FORBID
+    #define pthread_mutexattr_settype PTHREAD_MUTEXATTR_SETTYPE_FUNC_TAOS_FORBID
+    #define pthread_once PTHREAD_ONCE_FUNC_TAOS_FORBID
+    #define pthread_rwlock_destroy PTHREAD_RWLOCK_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_rwlock_init PTHREAD_RWLOCK_INIT_FUNC_TAOS_FORBID
+    #define pthread_rwlock_rdlock PTHREAD_RWLOCK_RDLOCK_FUNC_TAOS_FORBID
+    #define pthread_rwlock_timedrdlock PTHREAD_RWLOCK_TIMEDRDLOCK_FUNC_TAOS_FORBID
+    #define pthread_rwlock_timedwrlock PTHREAD_RWLOCK_TIMEDWRLOCK_FUNC_TAOS_FORBID
+    #define pthread_rwlock_tryrdlock PTHREAD_RWLOCK_TRYRDLOCK_FUNC_TAOS_FORBID
+    #define pthread_rwlock_trywrlock PTHREAD_RWLOCK_TRYWRLOCK_FUNC_TAOS_FORBID
+    #define pthread_rwlock_unlock PTHREAD_RWLOCK_UNLOCK_FUNC_TAOS_FORBID
+    #define pthread_rwlock_wrlock PTHREAD_RWLOCK_WRLOCK_FUNC_TAOS_FORBID
+    #define pthread_rwlockattr_destroy PTHREAD_RWLOCKATTR_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_rwlockattr_getpshared PTHREAD_RWLOCKATTR_GETPSHARED_FUNC_TAOS_FORBID
+    #define pthread_rwlockattr_init PTHREAD_RWLOCKATTR_INIT_FUNC_TAOS_FORBID
+    #define pthread_rwlockattr_setpshared PTHREAD_RWLOCKATTR_SETPSHARED_FUNC_TAOS_FORBID
+    #define pthread_self PTHREAD_SELF_FUNC_TAOS_FORBID
+    #define pthread_setcancelstate PTHREAD_SETCANCELSTATE_FUNC_TAOS_FORBID
+    #define pthread_setcanceltype PTHREAD_SETCANCELTYPE_FUNC_TAOS_FORBID
+    #define pthread_setschedparam PTHREAD_SETSCHEDPARAM_FUNC_TAOS_FORBID
+    #define pthread_setspecific PTHREAD_SETSPECIFIC_FUNC_TAOS_FORBID
+    #define pthread_spin_destroy PTHREAD_SPIN_DESTROY_FUNC_TAOS_FORBID
+    #define pthread_spin_init PTHREAD_SPIN_INIT_FUNC_TAOS_FORBID
+    #define pthread_spin_lock PTHREAD_SPIN_LOCK_FUNC_TAOS_FORBID
+    #define pthread_spin_trylock PTHREAD_SPIN_TRYLOCK_FUNC_TAOS_FORBID
+    #define pthread_spin_unlock PTHREAD_SPIN_UNLOCK_FUNC_TAOS_FORBID
+    #define pthread_testcancel PTHREAD_TESTCANCEL_FUNC_TAOS_FORBID
+    #define pthread_sigmask PTHREAD_SIGMASK_FUNC_TAOS_FORBID
+    #define sigwait SIGWAIT_FUNC_TAOS_FORBID
 #endif
 
-int32_t taosThreadSpinInit(TdThreadSpinlock *lock, int pshared);
+int32_t taosThreadCreate(TdThread * tid, const TdThreadAttr * attr, void *(*start)(void *), void *arg);
-int32_t taosThreadMutexInit(TdThreadMutex *mutex, const TdThreadMutexAttr *attr);
+int32_t taosThreadAttrDestroy(TdThreadAttr * attr);
-int32_t taosThreadSpinDestroy(TdThreadSpinlock *lock);
+int32_t taosThreadAttrGetDetachState(const TdThreadAttr * attr, int32_t *detachstate);
-int32_t taosThreadMutexDestroy(TdThreadMutex * mutex);
+int32_t taosThreadAttrGetInheritSched(const TdThreadAttr * attr, int32_t *inheritsched);
-int32_t taosThreadSpinLock(TdThreadSpinlock *lock);
+int32_t taosThreadAttrGetSchedParam(const TdThreadAttr * attr, struct sched_param *param);
-int32_t taosThreadMutexLock(TdThreadMutex *mutex);
+int32_t taosThreadAttrGetSchedPolicy(const TdThreadAttr * attr, int32_t *policy);
-int32_t taosThreadRwlockRdlock(TdThreadRwlock *rwlock);
+int32_t taosThreadAttrGetScope(const TdThreadAttr * attr, int32_t *contentionscope);
-int32_t taosThreadSpinUnlock(TdThreadSpinlock *lock);
+int32_t taosThreadAttrGetStackSize(const TdThreadAttr * attr, size_t * stacksize);
-int32_t taosThreadMutexUnlock(TdThreadMutex *mutex);
+int32_t taosThreadAttrInit(TdThreadAttr * attr);
-int32_t taosThreadRwlockWrlock(TdThreadRwlock *rwlock);
+int32_t taosThreadAttrSetDetachState(TdThreadAttr * attr, int32_t detachstate);
-int32_t taosThreadRwlockUnlock(TdThreadRwlock *rwlock);
+int32_t taosThreadAttrSetInheritSched(TdThreadAttr * attr, int32_t inheritsched);
-void taosThreadTestCancel(void);
+int32_t taosThreadAttrSetSchedParam(TdThreadAttr * attr, const struct sched_param *param);
-int32_t taosThreadAttrInit(TdThreadAttr *attr);
+int32_t taosThreadAttrSetSchedPolicy(TdThreadAttr * attr, int32_t policy);
-int32_t taosThreadCreate(TdThread *tid, const TdThreadAttr *attr, void*(*start)(void*), void *arg);
+int32_t taosThreadAttrSetScope(TdThreadAttr * attr, int32_t contentionscope);
-int32_t taosThreadOnce(TdThreadOnce *onceControl, void(*initRoutine)(void));
+int32_t taosThreadAttrSetStackSize(TdThreadAttr * attr, size_t stacksize);
-int32_t taosThreadAttrSetDetachState(TdThreadAttr *attr, int32_t detachState);
+int32_t taosThreadBarrierDestroy(TdThreadBarrier * barrier);
-int32_t taosThreadAttrDestroy(TdThreadAttr *attr);
+int32_t taosThreadBarrierInit(TdThreadBarrier * barrier, const TdThreadBarrierAttr * attr, uint32_t count);
-int32_t taosThreadJoin(TdThread thread, void **pValue);
+int32_t taosThreadBarrierWait(TdThreadBarrier * barrier);
-int32_t taosThreadRwlockInit(TdThreadRwlock *rwlock, const TdThreadRwlockAttr *attr);
+int32_t taosThreadBarrierAttrDestroy(TdThreadBarrierAttr * attr);
-int32_t taosThreadRwlockDestroy(TdThreadRwlock *rwlock);
+int32_t taosThreadBarrierAttrGetPshared(const TdThreadBarrierAttr * attr, int32_t *pshared);
-int32_t taosThreadCondSignal(TdThreadCond *cond);
+int32_t taosThreadBarrierAttrInit(TdThreadBarrierAttr * attr);
-int32_t taosThreadCondInit(TdThreadCond *cond, const TdThreadCondAttr *attr);
+int32_t taosThreadBarrierAttrSetPshared(TdThreadBarrierAttr * attr, int32_t pshared);
-int32_t taosThreadCondBroadcast(TdThreadCond *cond);
-int32_t taosThreadCondDestroy(TdThreadCond *cond);
-int32_t taosThreadCondWait(TdThreadCond *cond, TdThreadMutex *mutex);
-TdThread taosThreadSelf(void);
-int32_t taosThreadEqual(TdThread t1, TdThread t2);
-int32_t taosThreadSigmask(int how, sigset_t const *set, sigset_t *oset);
 int32_t taosThreadCancel(TdThread thread);
-int32_t taosThreadKill(TdThread thread, int sig);
+int32_t taosThreadCondDestroy(TdThreadCond * cond);
+int32_t taosThreadCondInit(TdThreadCond * cond, const TdThreadCondAttr * attr);
+int32_t taosThreadCondSignal(TdThreadCond * cond);
+int32_t taosThreadCondBroadcast(TdThreadCond * cond);
+int32_t taosThreadCondWait(TdThreadCond * cond, TdThreadMutex * mutex);
+int32_t taosThreadCondTimedWait(TdThreadCond * cond, TdThreadMutex * mutex, const struct timespec *abstime);
+int32_t taosThreadCondAttrDestroy(TdThreadCondAttr * attr);
+int32_t taosThreadCondAttrGetPshared(const TdThreadCondAttr * attr, int32_t *pshared);
+int32_t taosThreadCondAttrInit(TdThreadCondAttr * attr);
+int32_t taosThreadCondAttrSetPshared(TdThreadCondAttr * attr, int32_t pshared);
+int32_t taosThreadDetach(TdThread thread);
+int32_t taosThreadEqual(TdThread t1, TdThread t2);
+void taosThreadExit(void *valuePtr);
+int32_t taosThreadGetSchedParam(TdThread thread, int32_t *policy, struct sched_param *param);
+void *taosThreadGetSpecific(TdThreadKey key);
+int32_t taosThreadJoin(TdThread thread, void **valuePtr);
+int32_t taosThreadKeyCreate(TdThreadKey * key, void(*destructor)(void *));
+int32_t taosThreadKeyDelete(TdThreadKey key);
+int32_t taosThreadKill(TdThread thread, int32_t sig);
+int32_t taosThreadMutexConsistent(TdThreadMutex* mutex);
+int32_t taosThreadMutexDestroy(TdThreadMutex * mutex);
+int32_t taosThreadMutexInit(TdThreadMutex * mutex, const TdThreadMutexAttr * attr);
+int32_t taosThreadMutexLock(TdThreadMutex * mutex);
+int32_t taosThreadMutexTimedLock(TdThreadMutex * mutex, const struct timespec *abstime);
+int32_t taosThreadMutexTryLock(TdThreadMutex * mutex);
+int32_t taosThreadMutexUnlock(TdThreadMutex * mutex);
+int32_t taosThreadMutexAttrDestroy(TdThreadMutexAttr * attr);
+int32_t taosThreadMutexAttrGetPshared(const TdThreadMutexAttr * attr, int32_t *pshared);
+int32_t taosThreadMutexAttrGetRobust(const TdThreadMutexAttr * attr, int32_t * robust);
+int32_t taosThreadMutexAttrGetType(const TdThreadMutexAttr * attr, int32_t *kind);
+int32_t taosThreadMutexAttrInit(TdThreadMutexAttr * attr);
+int32_t taosThreadMutexAttrSetPshared(TdThreadMutexAttr * attr, int32_t pshared);
+int32_t taosThreadMutexAttrSetRobust(TdThreadMutexAttr * attr, int32_t robust);
+int32_t taosThreadMutexAttrSetType(TdThreadMutexAttr * attr, int32_t kind);
+int32_t taosThreadOnce(TdThreadOnce * onceControl, void(*initRoutine)(void));
+int32_t taosThreadRwlockDestroy(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockInit(TdThreadRwlock * rwlock, const TdThreadRwlockAttr * attr);
+int32_t taosThreadRwlockRdlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockTimedRdlock(TdThreadRwlock * rwlock, const struct timespec *abstime);
+int32_t taosThreadRwlockTimedWrlock(TdThreadRwlock * rwlock, const struct timespec *abstime);
+int32_t taosThreadRwlockTryRdlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockTryWrlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockUnlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockWrlock(TdThreadRwlock * rwlock);
+int32_t taosThreadRwlockAttrDestroy(TdThreadRwlockAttr * attr);
+int32_t taosThreadRwlockAttrGetPshared(const TdThreadRwlockAttr * attr, int32_t *pshared);
+int32_t taosThreadRwlockAttrInit(TdThreadRwlockAttr * attr);
+int32_t taosThreadRwlockAttrSetPshared(TdThreadRwlockAttr * attr, int32_t pshared);
+TdThread taosThreadSelf(void);
+int32_t taosThreadSetCancelState(int32_t state, int32_t *oldstate);
+int32_t taosThreadSetCancelType(int32_t type, int32_t *oldtype);
+int32_t taosThreadSetSchedParam(TdThread thread, int32_t policy, const struct sched_param *param);
+int32_t taosThreadSetSpecific(TdThreadKey key, const void *value);
+int32_t taosThreadSpinDestroy(TdThreadSpinlock * lock);
+int32_t taosThreadSpinInit(TdThreadSpinlock * lock, int32_t pshared);
+int32_t taosThreadSpinLock(TdThreadSpinlock * lock);
+int32_t taosThreadSpinTrylock(TdThreadSpinlock * lock);
+int32_t taosThreadSpinUnlock(TdThreadSpinlock * lock);
+void taosThreadTestCancel(void);
+int32_t taosThreadSigMask(int32_t how, sigset_t const *set, sigset_t * oset);
+int32_t taosThreadSigWait(const sigset_t * set, int32_t *sig);
+
 #ifdef __cplusplus
 }
 #endif

include/os/osTime.h

@@ -20,6 +20,8 @@
 extern "C" {
 #endif
 
+#include <time.h>
+
 // If the error is in a third-party library, place this header file under the third-party library header file.
 // When you want to use this feature, you should find or add the same function in the following section.
 #ifndef ALLOW_FORBID_FUNC

include/util/taoserror.h

@@ -78,6 +78,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_CFG_NOT_FOUND                 TAOS_DEF_ERROR_CODE(0, 0x010C)
 #define TSDB_CODE_INVALID_CFG                   TAOS_DEF_ERROR_CODE(0, 0x010D)
 #define TSDB_CODE_OUT_OF_SHM_MEM                TAOS_DEF_ERROR_CODE(0, 0x010E)
+#define TSDB_CODE_INVALID_SHM_ID                TAOS_DEF_ERROR_CODE(0, 0x010F)
 #define TSDB_CODE_REF_NO_MEMORY                 TAOS_DEF_ERROR_CODE(0, 0x0110)
 #define TSDB_CODE_REF_FULL                      TAOS_DEF_ERROR_CODE(0, 0x0111)
 #define TSDB_CODE_REF_ID_REMOVED                TAOS_DEF_ERROR_CODE(0, 0x0112)

include/util/tdef.h

@@ -381,6 +381,14 @@ typedef enum ELogicConditionType {
 #define TSDB_MAX_DB_DELAY     10
 #define TSDB_DEFAULT_DB_DELAY 2
 
+#define TSDB_DEFAULT_EXPLAIN_VERBOSE false
+
+#define TSDB_MIN_EXPLAIN_RATIO       0
+#define TSDB_MAX_EXPLAIN_RATIO       1
+#define TSDB_DEFAULT_EXPLAIN_RATIO   0.001
+
+#define TSDB_EXPLAIN_RESULT_ROW_SIZE 1024
+
 #define TSDB_MAX_JOIN_TABLE_NUM 10
 #define TSDB_MAX_UNION_CLAUSE   5
 

include/util/tprocess.h

@@ -32,29 +32,25 @@ typedef void *(*ProcConsumeFp)(void *pParent, void *pHead, int16_t headLen, void
                                ProcFuncType ftype);
 
 typedef struct {
-  int32_t       childQueueSize;
   ProcConsumeFp childConsumeFp;
   ProcMallocFp  childMallocHeadFp;
   ProcFreeFp    childFreeHeadFp;
   ProcMallocFp  childMallocBodyFp;
   ProcFreeFp    childFreeBodyFp;
-  int32_t       parentQueueSize;
   ProcConsumeFp parentConsumeFp;
-  ProcMallocFp  parentdMallocHeadFp;
+  ProcMallocFp  parentMallocHeadFp;
   ProcFreeFp    parentFreeHeadFp;
   ProcMallocFp  parentMallocBodyFp;
   ProcFreeFp    parentFreeBodyFp;
-  bool          testFlag;
+  SShm          shm;
   void         *pParent;
   const char   *name;
+  bool          isChild;
 } SProcCfg;
 
 SProcObj *taosProcInit(const SProcCfg *pCfg);
 void      taosProcCleanup(SProcObj *pProc);
 int32_t   taosProcRun(SProcObj *pProc);
-void      taosProcStop(SProcObj *pProc);
-bool      taosProcIsChild(SProcObj *pProc);
-int32_t   taosProcChildId(SProcObj *pProc);
 int32_t   taosProcPutToChildQ(SProcObj *pProc, const void *pHead, int16_t headLen, const void *pBody, int32_t bodyLen,
                               ProcFuncType ftype);
 int32_t   taosProcPutToParentQ(SProcObj *pProc, const void *pHead, int16_t headLen, const void *pBody, int32_t bodyLen,

source/client/src/clientEnv.c

@@ -200,6 +200,9 @@ static void doDestroyRequest(void *p) {
     taosArrayDestroy(pRequest->body.showInfo.pArray);
   }
 
+  taosArrayDestroy(pRequest->tableList);
+  taosArrayDestroy(pRequest->dbList);
+
   deregisterRequest(pRequest);
   taosMemoryFreeClear(pRequest);
 }

source/client/src/clientMain.c

@@ -71,6 +71,12 @@ void taos_cleanup(void) {
   tscInfo("all local resources released");
 }
 
+setConfRet   taos_set_config(const char *config) {
+  // TODO
+  setConfRet ret = {SET_CONF_RET_SUCC, {0}};
+  return ret;
+}
+
 TAOS *taos_connect(const char *ip, const char *user, const char *pass, const char *db, uint16_t port) {
   tscDebug("try to connect to %s:%u, user:%s db:%s", ip, port, user, db);
   if (user == NULL) {
@@ -257,6 +263,11 @@ int *taos_fetch_lengths(TAOS_RES *res) {
   return ((SRequestObj *)res)->body.resInfo.length;
 }
 
+TAOS_ROW *taos_result_block(TAOS_RES *res) {
+  // TODO
+  return NULL;
+}
+
 // todo intergrate with tDataTypes
 const char *taos_data_type(int type) {
   switch (type) {
@@ -353,6 +364,11 @@ bool taos_is_null(TAOS_RES *res, int32_t row, int32_t col) {
   return colDataIsNull_f(pCol->nullbitmap, row);
 }
 
+bool taos_is_update_query(TAOS_RES *res) {
+  // TODO
+  return true;
+}
+
 int taos_fetch_block(TAOS_RES *res, TAOS_ROW *rows) {
   if (res == NULL) {
     return 0;
@@ -376,6 +392,11 @@ int taos_fetch_block(TAOS_RES *res, TAOS_ROW *rows) {
 
 int taos_validate_sql(TAOS *taos, const char *sql) { return true; }
 
+void taos_reset_current_db(TAOS *taos) {
+  // TODO
+  return;
+}
+
 const char *taos_get_server_info(TAOS *taos) {
   if (taos == NULL) {
     return NULL;

source/client/src/tmq.c

@@ -241,6 +241,10 @@ void tmq_list_destroy(tmq_list_t* list) {
   taosArrayDestroyEx(container, (void (*)(void*))taosMemoryFree);
 }
 
+static int32_t tmqMakeTopicVgKey(char* dst, const char* topicName, int32_t vg) {
+  return sprintf(dst, "%s:%d", topicName, vg);
+}
+
 void tmqClearUnhandleMsg(tmq_t* tmq) {
   tmq_message_t* msg = NULL;
   while (1) {
@@ -827,7 +831,7 @@ int32_t tmqPollCb(void* param, const SDataBuf* pMsg, int32_t code) {
   SMqClientVg*    pVg = pParam->pVg;
   tmq_t*          tmq = pParam->tmq;
   if (code != 0) {
-    printf("msg discard, code:%x\n", code);
+    tscWarn("msg discard, code:%x", code);
     goto WRITE_QUEUE_FAIL;
   }
 
@@ -835,12 +839,12 @@ int32_t tmqPollCb(void* param, const SDataBuf* pMsg, int32_t code) {
   int32_t tmqEpoch = atomic_load_32(&tmq->epoch);
   if (msgEpoch < tmqEpoch) {
     tsem_post(&tmq->rspSem);
-    printf("discard rsp epoch %d, current epoch %d\n", msgEpoch, tmqEpoch);
+    tscWarn("discard rsp epoch %d, current epoch %d", msgEpoch, tmqEpoch);
     return 0;
   }
 
   if (msgEpoch != tmqEpoch) {
-    printf("mismatch rsp epoch %d, current epoch %d\n", msgEpoch, tmqEpoch);
+    tscWarn("mismatch rsp epoch %d, current epoch %d", msgEpoch, tmqEpoch);
   } else {
     atomic_sub_fetch_32(&tmq->waitingRequest, 1);
   }
@@ -899,19 +903,54 @@ WRITE_QUEUE_FAIL:
 bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, SMqCMGetSubEpRsp* pRsp) {
   /*printf("call update ep %d\n", epoch);*/
   bool    set = false;
-  int32_t sz = taosArrayGetSize(pRsp->topics);
+  int32_t topicNumGet = taosArrayGetSize(pRsp->topics);
-  SArray* newTopics = taosArrayInit(sz, sizeof(SMqClientTopic));
+  char    vgKey[TSDB_TOPIC_FNAME_LEN + 22];
-  for (int32_t i = 0; i < sz; i++) {
+  SArray* newTopics = taosArrayInit(topicNumGet, sizeof(SMqClientTopic));
+  if (newTopics == NULL) {
+    return false;
+  }
+  SHashObj* pHash = taosHashInit(64, MurmurHash3_32, false, HASH_NO_LOCK);
+  if (pHash == NULL) {
+    taosArrayDestroy(newTopics);
+    return false;
+  }
+
+  // find topic, build hash
+  for (int32_t i = 0; i < topicNumGet; i++) {
     SMqClientTopic topic = {0};
     SMqSubTopicEp* pTopicEp = taosArrayGet(pRsp->topics, i);
+    taosHashClear(pHash);
     topic.topicName = strdup(pTopicEp->topic);
-    int32_t vgSz = taosArrayGetSize(pTopicEp->vgs);
+
-    topic.vgs = taosArrayInit(vgSz, sizeof(SMqClientVg));
+    int32_t topicNumCur = taosArrayGetSize(tmq->clientTopics);
-    for (int32_t j = 0; j < vgSz; j++) {
+    for (int32_t j = 0; j < topicNumCur; j++) {
+      // find old topic
+      SMqClientTopic* pTopicCur = taosArrayGet(tmq->clientTopics, j);
+      if (pTopicCur->vgs && strcmp(pTopicCur->topicName, pTopicEp->topic) == 0) {
+        int32_t vgNumCur = taosArrayGetSize(pTopicCur->vgs);
+        if (vgNumCur == 0) break;
+        for (int32_t k = 0; k < vgNumCur; k++) {
+          SMqClientVg* pVgCur = taosArrayGet(pTopicCur->vgs, k);
+          sprintf(vgKey, "%s:%d", topic.topicName, pVgCur->vgId);
+          taosHashPut(pHash, vgKey, strlen(vgKey), &pVgCur->currentOffset, sizeof(int64_t));
+        }
+        break;
+      }
+    }
+
+    int32_t vgNumGet = taosArrayGetSize(pTopicEp->vgs);
+    topic.vgs = taosArrayInit(vgNumGet, sizeof(SMqClientVg));
+    for (int32_t j = 0; j < vgNumGet; j++) {
       SMqSubVgEp* pVgEp = taosArrayGet(pTopicEp->vgs, j);
+      sprintf(vgKey, "%s:%d", topic.topicName, pVgEp->vgId);
+      int64_t* pOffset = taosHashGet(pHash, vgKey, strlen(vgKey));
+      int64_t  offset = pVgEp->offset;
+      if (pOffset != NULL) {
+        offset = *pOffset;
+      }
       SMqClientVg clientVg = {
           .pollCnt = 0,
-          .currentOffset = pVgEp->offset,
+          .currentOffset = offset,
           .vgId = pVgEp->vgId,
           .epSet = pVgEp->epSet,
           .vgStatus = TMQ_VG_STATUS__IDLE,
@@ -922,6 +961,7 @@ bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, SMqCMGetSubEpRsp* pRsp) {
     taosArrayPush(newTopics, &topic);
   }
   if (tmq->clientTopics) taosArrayDestroy(tmq->clientTopics);
+  taosHashCleanup(pHash);
   tmq->clientTopics = newTopics;
   atomic_store_32(&tmq->epoch, epoch);
   return set;
@@ -931,7 +971,7 @@ int32_t tmqAskEpCb(void* param, const SDataBuf* pMsg, int32_t code) {
   SMqAskEpCbParam* pParam = (SMqAskEpCbParam*)param;
   tmq_t*           tmq = pParam->tmq;
   if (code != 0) {
-    printf("get topic endpoint error, not ready, wait:%d\n", pParam->sync);
+    tscError("get topic endpoint error, not ready, wait:%d\n", pParam->sync);
     goto END;
   }
 
@@ -1302,6 +1342,7 @@ tmq_message_t* tmq_consumer_poll(tmq_t* tmq, int64_t blocking_time) {
 
   while (1) {
     /*printf("cycle\n");*/
+    tmqAskEp(tmq, false);
     tmqPollImpl(tmq, blocking_time);
 
     tsem_wait(&tmq->rspSem);

source/common/src/tdatablock.c

@@ -331,7 +331,6 @@ int32_t blockDataUpdateTsWindow(SSDataBlock* pDataBlock) {
     return 0;
   }
 
-  ASSERT(pColInfoData->nullbitmap == NULL);
   pDataBlock->info.window.skey = *(TSKEY*)colDataGetData(pColInfoData, 0);
   pDataBlock->info.window.ekey = *(TSKEY*)colDataGetData(pColInfoData, (pDataBlock->info.rows - 1));
   return 0;
@@ -609,22 +608,6 @@ size_t blockDataGetSerialMetaSize(const SSDataBlock* pBlock) {
   return sizeof(int32_t) + pBlock->info.numOfCols * sizeof(int32_t);
 }
 
-SSchema* blockDataExtractSchema(const SSDataBlock* pBlock, int32_t* numOfCols) {
-  SSchema* pSchema = taosMemoryCalloc(pBlock->info.numOfCols, sizeof(SSchema));
-  for (int32_t i = 0; i < pBlock->info.numOfCols; ++i) {
-    SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, i);
-    pSchema[i].bytes = pColInfoData->info.bytes;
-    pSchema[i].type = pColInfoData->info.type;
-    pSchema[i].colId = pColInfoData->info.colId;
-  }
-
-  if (numOfCols != NULL) {
-    *numOfCols = pBlock->info.numOfCols;
-  }
-
-  return pSchema;
-}
-
 double blockDataGetSerialRowSize(const SSDataBlock* pBlock) {
   ASSERT(pBlock != NULL);
   double rowSize = 0;

source/common/src/tglobal.c

@@ -45,7 +45,7 @@ float   tsRatioOfQueryCores = 1.0f;
 int32_t tsMaxBinaryDisplayWidth = 30;
 bool    tsEnableSlaveQuery = 1;
 bool    tsPrintAuth = 0;
-int32_t tsMultiProcess = 0;
+bool    tsMultiProcess = 0;
 
 // monitor
 bool     tsEnableMonitor = 1;
@@ -347,7 +347,7 @@ static int32_t taosAddServerCfg(SConfig *pCfg) {
   if (cfgAddBool(pCfg, "printAuth", tsPrintAuth, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "slaveQuery", tsEnableSlaveQuery, 0) != 0) return -1;
   if (cfgAddBool(pCfg, "deadLockKillQuery", tsDeadLockKillQuery, 0) != 0) return -1;
-  if (cfgAddInt32(pCfg, "multiProcess", tsMultiProcess, 0, 2, 0) != 0) return -1;
+  if (cfgAddBool(pCfg, "multiProcess", tsMultiProcess, 0) != 0) return -1;
 
   if (cfgAddBool(pCfg, "monitor", tsEnableMonitor, 0) != 0) return -1;
   if (cfgAddInt32(pCfg, "monitorInterval", tsMonitorInterval, 1, 360000, 0) != 0) return -1;
@@ -466,7 +466,7 @@ static int32_t taosSetServerCfg(SConfig *pCfg) {
   tsPrintAuth = cfgGetItem(pCfg, "printAuth")->bval;
   tsEnableSlaveQuery = cfgGetItem(pCfg, "slaveQuery")->bval;
   tsDeadLockKillQuery = cfgGetItem(pCfg, "deadLockKillQuery")->bval;
-  tsMultiProcess = cfgGetItem(pCfg, "multiProcess")->i32;
+  tsMultiProcess = cfgGetItem(pCfg, "multiProcess")->bval;
 
   tsEnableMonitor = cfgGetItem(pCfg, "monitor")->bval;
   tsMonitorInterval = cfgGetItem(pCfg, "monitorInterval")->i32;

source/dnode/mgmt/dm/src/dmFile.c

@@ -130,7 +130,7 @@ int32_t dmReadFile(SDnodeMgmt *pMgmt) {
   }
 
   code = 0;
-  dInfo("succcessed to read file %s", file);
+  dDebug("succcessed to read file %s", file);
   dmPrintDnodes(pMgmt);
 
 PRASE_DNODE_OVER:

source/dnode/mgmt/dm/src/dmInt.c

@@ -112,6 +112,16 @@ int32_t dmInit(SMgmtWrapper *pWrapper) {
     return -1;
   }
 
+  if (dndInitServer(pDnode) != 0) {
+    dError("failed to init trans server since %s", terrstr());
+    return -1;
+  }
+
+  if (dndInitClient(pDnode) != 0) {
+    dError("failed to init trans client since %s", terrstr());
+    return -1;
+  }
+
   pWrapper->pMgmt = pMgmt;
   dInfo("dnode-mgmt is initialized");
   return 0;
@@ -122,6 +132,7 @@ void dmCleanup(SMgmtWrapper *pWrapper) {
   if (pMgmt == NULL) return;
 
   dInfo("dnode-mgmt start to clean up");
+  SDnode *pDnode = pMgmt->pDnode;
   dmStopWorker(pMgmt);
 
   taosWLockLatch(&pMgmt->latch);
@@ -140,6 +151,9 @@ void dmCleanup(SMgmtWrapper *pWrapper) {
 
   taosMemoryFree(pMgmt);
   pWrapper->pMgmt = NULL;
+  dndCleanupServer(pDnode);
+  dndCleanupClient(pDnode);
+
   dInfo("dnode-mgmt is cleaned up");
 }
 

source/dnode/mgmt/main/exe/dndMain.c

@@ -29,7 +29,7 @@ static struct {
   ENodeType ntype;
 } global = {0};
 
-static void dndSigintHandle(int signum, void *info, void *ctx) {
+static void dndStopDnode(int signum, void *info, void *ctx) {
   dInfo("signal:%d is received", signum);
   SDnode *pDnode = atomic_val_compare_exchange_ptr(&global.pDnode, 0, global.pDnode);
   if (pDnode != NULL) {
@@ -37,12 +37,27 @@ static void dndSigintHandle(int signum, void *info, void *ctx) {
   }
 }
 
+static void dndHandleChild(int signum, void *info, void *ctx) {
+  dInfo("signal:%d is received", signum);
+  dndHandleEvent(global.pDnode, DND_EVENT_CHILD);
+}
+
 static void dndSetSignalHandle() {
-  taosSetSignal(SIGTERM, dndSigintHandle);
+  taosSetSignal(SIGTERM, dndStopDnode);
-  taosSetSignal(SIGHUP, dndSigintHandle);
+  taosSetSignal(SIGHUP, dndStopDnode);
-  taosSetSignal(SIGINT, dndSigintHandle);
+  taosSetSignal(SIGINT, dndStopDnode);
-  taosSetSignal(SIGABRT, dndSigintHandle);
+  taosSetSignal(SIGABRT, dndStopDnode);
-  taosSetSignal(SIGBREAK, dndSigintHandle);
+  taosSetSignal(SIGBREAK, dndStopDnode);
+
+  if (!tsMultiProcess) {
+    // Set the single process signal
+  } else if (global.ntype == DNODE) {
+    // When the child process exits, the parent process receives a signal
+    taosSetSignal(SIGCHLD, dndHandleChild);
+  } else {
+    // When the parent process exits, the child process will receive the SIGKILL signal
+    taosKillChildOnSelfStopped();
+  }
 }
 
 static int32_t dndParseArgs(int32_t argc, char const *argv[]) {
@@ -66,6 +81,10 @@ static int32_t dndParseArgs(int32_t argc, char const *argv[]) {
       global.generateGrant = true;
     } else if (strcmp(argv[i], "-n") == 0) {
       global.ntype = atoi(argv[++i]);
+      if (global.ntype <= DNODE || global.ntype > NODE_MAX) {
+        printf("'-n' range is [1-5], default is 0\n");
+        return -1;
+      }
     } else if (strcmp(argv[i], "-C") == 0) {
       global.dumpConfig = true;
     } else if (strcmp(argv[i], "-V") == 0) {
@@ -109,8 +128,9 @@ static SDnodeOpt dndGetOpt() {
   option.serverPort = tsServerPort;
   tstrncpy(option.localFqdn, tsLocalFqdn, sizeof(option.localFqdn));
   snprintf(option.localEp, sizeof(option.localEp), "%s:%u", option.localFqdn, option.serverPort);
-  option.pDisks = tsDiskCfg;
+  option.disks = tsDiskCfg;
   option.numOfDisks = tsDiskCfgNum;
+  option.ntype = global.ntype;
   return option;
 }
 
@@ -121,10 +141,9 @@ static int32_t dndInitLog() {
 }
 
 static void dndSetProcName(char **argv) {
-  if (global.ntype != 0) {
+  if (global.ntype != DNODE) {
     const char *name = dndNodeProcStr(global.ntype);
-    prctl(PR_SET_NAME, name);
+    taosSetProcName(argv, name);
-    strcpy(argv[0], name);
   }
 }
 

source/dnode/mgmt/main/inc/dnd.h

@@ -95,12 +95,14 @@ typedef struct SMgmtWrapper {
   bool        deployed;
   bool        required;
   EProcType   procType;
+  int32_t     procId;
   SProcObj   *pProc;
+  SShm        shm;
   void       *pMgmt;
   SDnode     *pDnode;
-  NodeMsgFp   msgFps[TDMT_MAX];
-  int32_t     msgVgIds[TDMT_MAX];  // Handle the case where the same message type is distributed to qnode or vnode
   SMgmtFp     fp;
+  int8_t      msgVgIds[TDMT_MAX];  // Handle the case where the same message type is distributed to qnode or vnode
+  NodeMsgFp   msgFps[TDMT_MAX];
 } SMgmtWrapper;
 
 typedef struct {
@@ -119,14 +121,15 @@ typedef struct SDnode {
   char        *firstEp;
   char        *secondEp;
   char        *dataDir;
-  SDiskCfg    *pDisks;
+  SDiskCfg    *disks;
   int32_t      numOfDisks;
   uint16_t     serverPort;
   bool         dropped;
+  ENodeType    ntype;
   EDndStatus   status;
   EDndEvent    event;
   SStartupReq  startup;
-  TdFilePtr    pLockFile;
+  TdFilePtr    lockfile;
   STransMgmt   trans;
   SMgmtWrapper wrappers[NODE_MAX];
 } SDnode;
@@ -135,17 +138,21 @@ const char *dndNodeLogStr(ENodeType ntype);
 const char *dndNodeProcStr(ENodeType ntype);
 EDndStatus  dndGetStatus(SDnode *pDnode);
 void        dndSetStatus(SDnode *pDnode, EDndStatus stat);
-void        dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp, int32_t vgId);
+void        dndSetMsgHandle(SMgmtWrapper *pWrapper, tmsg_t msgType, NodeMsgFp nodeMsgFp, int8_t vgId);
 void        dndReportStartup(SDnode *pDnode, const char *pName, const char *pDesc);
 void        dndSendMonitorReport(SDnode *pDnode);
 
+int32_t dndInitServer(SDnode *pDnode);
+void    dndCleanupServer(SDnode *pDnode);
+int32_t dndInitClient(SDnode *pDnode);
+void    dndCleanupClient(SDnode *pDnode);
+int32_t dndProcessNodeMsg(SDnode *pDnode, SNodeMsg *pMsg);
 int32_t dndSendReqToMnode(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
 int32_t dndSendReqToDnode(SMgmtWrapper *pWrapper, const SEpSet *pEpSet, SRpcMsg *pMsg);
 void    dndSendRsp(SMgmtWrapper *pWrapper, const SRpcMsg *pRsp);
 void    dndRegisterBrokenLinkArg(SMgmtWrapper *pWrapper, SRpcMsg *pMsg);
 SMsgCb  dndCreateMsgcb(SMgmtWrapper *pWrapper);
 
-int32_t dndProcessNodeMsg(SDnode *pDnode, SNodeMsg *pMsg);
 int32_t dndReadFile(SMgmtWrapper *pWrapper, bool *pDeployed);
 int32_t dndWriteFile(SMgmtWrapper *pWrapper, bool deployed);
 

source/dnode/mgmt/main/inc/dndInt.h

@@ -34,7 +34,6 @@ int32_t     dndInit();
 void        dndCleanup();
 const char *dndStatStr(EDndStatus stat);
 void        dndGetStartup(SDnode *pDnode, SStartupReq *pStartup);
-TdFilePtr   dndCheckRunning(const char *dataDir);
 void        dndProcessStartupReq(SDnode *pDnode, SRpcMsg *pMsg);
 
 // dndMsg.c
@@ -51,13 +50,14 @@ void    dndClose(SDnode *pDnode);
 void    dndHandleEvent(SDnode *pDnode, EDndEvent event);
 
 // dndTransport.c
-int32_t dndInitServer(SDnode *pDnode);
-void    dndCleanupServer(SDnode *pDnode);
-int32_t dndInitClient(SDnode *pDnode);
-void    dndCleanupClient(SDnode *pDnode);
 int32_t dndInitMsgHandle(SDnode *pDnode);
 void    dndSendRpcRsp(SMgmtWrapper *pWrapper, const SRpcMsg *pRsp);
 
+// dndFile.c
+TdFilePtr dndCheckRunning(const char *dataDir);
+int32_t   dndReadShmFile(SDnode *pDnode);
+int32_t   dndWriteShmFile(SDnode *pDnode);
+
 #ifdef __cplusplus
 }
 #endif

source/dnode/mgmt/main/src/dndExec.c

@@ -16,15 +16,6 @@
 #define _DEFAULT_SOURCE
 #include "dndInt.h"
 
-static void dndResetLog(SMgmtWrapper *pMgmt) {
-  char logname[24] = {0};
-  snprintf(logname, sizeof(logname), "%slog", pMgmt->name);
-
-  dInfo("node:%s, reset log to %s in child process", pMgmt->name, logname);
-  taosCloseLog();
-  taosInitLog(logname, 1);
-}
-
 static bool dndRequireNode(SMgmtWrapper *pWrapper) {
   bool required = false;
   int32_t code =(*pWrapper->fp.requiredFp)(pWrapper, &required);
@@ -37,14 +28,18 @@ static bool dndRequireNode(SMgmtWrapper *pWrapper) {
 }
 
 int32_t dndOpenNode(SMgmtWrapper *pWrapper) {
-  int32_t code = (*pWrapper->fp.openFp)(pWrapper);
+  if (taosMkDir(pWrapper->path) != 0) {
-  if (code != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("node:%s, failed to open since %s", pWrapper->name, terrstr());
+    dError("node:%s, failed to create dir:%s since %s", pWrapper->name, pWrapper->path, terrstr());
     return -1;
-  } else {
-    dDebug("node:%s, has been opened", pWrapper->name);
   }
 
+  if ((*pWrapper->fp.openFp)(pWrapper) != 0) {
+    dError("node:%s, failed to open since %s", pWrapper->name, terrstr());
+    return -1;
+  }
+
+  dDebug("node:%s, has been opened", pWrapper->name);
   pWrapper->deployed = true;
   return 0;
 }
@@ -71,23 +66,13 @@ void dndCloseNode(SMgmtWrapper *pWrapper) {
 }
 
 static int32_t dndRunInSingleProcess(SDnode *pDnode) {
-  dInfo("dnode run in single process mode");
+  dInfo("dnode start to run in single process");
 
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
+  for (ENodeType n = DNODE; n < NODE_MAX; ++n) {
     SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
     pWrapper->required = dndRequireNode(pWrapper);
     if (!pWrapper->required) continue;
-    SMsgCb msgCb = dndCreateMsgcb(pWrapper);
-    tmsgSetDefaultMsgCb(&msgCb);
 
-    if (taosMkDir(pWrapper->path) != 0) {
-      terrno = TAOS_SYSTEM_ERROR(errno);
-      dError("failed to create dir:%s since %s", pWrapper->path, terrstr());
-      return -1;
-    }
-
-    dInfo("node:%s, will start in single process", pWrapper->name);
-    pWrapper->procType = PROC_SINGLE;
     if (dndOpenNode(pWrapper) != 0) {
       dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
       return -1;
@@ -106,18 +91,10 @@ static int32_t dndRunInSingleProcess(SDnode *pDnode) {
     }
   }
 
+  dInfo("dnode running in single process");
   return 0;
 }
 
-static void dndClearNodesExecpt(SDnode *pDnode, ENodeType except) {
-  // dndCleanupServer(pDnode);
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
-    if (except == n) continue;
-    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
-    pWrapper->required = false;
-  }
-}
-
 static void dndConsumeChildQueue(SMgmtWrapper *pWrapper, SNodeMsg *pMsg, int16_t msgLen, void *pCont, int32_t contLen,
                                  ProcFuncType ftype) {
   SRpcMsg *pRpc = &pMsg->rpcMsg;
@@ -163,115 +140,159 @@ static void dndConsumeParentQueue(SMgmtWrapper *pWrapper, SRpcMsg *pMsg, int16_t
   taosMemoryFree(pMsg);
 }
 
-static int32_t dndRunInMultiProcess(SDnode *pDnode) {
+static int32_t dndRunInParentProcess(SDnode *pDnode) {
-  dInfo("dnode run in multi process mode");
+  dInfo("dnode start to run in parent process");
+  SMgmtWrapper *pDWrapper = &pDnode->wrappers[DNODE];
+  if (dndOpenNode(pDWrapper) != 0) {
+    dError("node:%s, failed to start since %s", pDWrapper->name, terrstr());
+    return -1;
+  }
 
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
+  for (ENodeType n = DNODE + 1; n < NODE_MAX; ++n) {
     SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
     pWrapper->required = dndRequireNode(pWrapper);
     if (!pWrapper->required) continue;
 
-    SMsgCb msgCb = dndCreateMsgcb(pWrapper);
+    int64_t shmsize = 1024 * 1024 * 2;  // size will be a configuration item
-    tmsgSetDefaultMsgCb(&msgCb);
+    if (taosCreateShm(&pWrapper->shm, shmsize) != 0) {
-
+      terrno = TAOS_SYSTEM_ERROR(terrno);
-    if (taosMkDir(pWrapper->path) != 0) {
+      dError("node:%s, failed to create shm size:%" PRId64 " since %s", pWrapper->name, shmsize, terrstr());
-      terrno = TAOS_SYSTEM_ERROR(errno);
-      dError("failed to create dir:%s since %s", pWrapper->path, terrstr());
       return -1;
     }
 
-    if (n == DNODE) {
+    SProcCfg cfg = {.childConsumeFp = (ProcConsumeFp)dndConsumeChildQueue,
-      dInfo("node:%s, will start in parent process", pWrapper->name);
+                    .childMallocHeadFp = (ProcMallocFp)taosAllocateQitem,
-      pWrapper->procType = PROC_SINGLE;
+                    .childFreeHeadFp = (ProcFreeFp)taosFreeQitem,
-      if (dndOpenNode(pWrapper) != 0) {
+                    .childMallocBodyFp = (ProcMallocFp)rpcMallocCont,
-        dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
+                    .childFreeBodyFp = (ProcFreeFp)rpcFreeCont,
-        return -1;
+                    .parentConsumeFp = (ProcConsumeFp)dndConsumeParentQueue,
-      }
+                    .parentMallocHeadFp = (ProcMallocFp)taosMemoryMalloc,
-      continue;
+                    .parentFreeHeadFp = (ProcFreeFp)taosMemoryFree,
-    }
+                    .parentMallocBodyFp = (ProcMallocFp)rpcMallocCont,
+                    .parentFreeBodyFp = (ProcFreeFp)rpcFreeCont,
+                    .shm = pWrapper->shm,
+                    .pParent = pWrapper,
+                    .isChild = false,
+                    .name = pWrapper->name};
 
-    SProcCfg  cfg = {.childQueueSize = 1024 * 1024 * 2,  // size will be a configuration item
+    pWrapper->procType = PROC_PARENT;
-                     .childConsumeFp = (ProcConsumeFp)dndConsumeChildQueue,
+    pWrapper->pProc = taosProcInit(&cfg);
-                     .childMallocHeadFp = (ProcMallocFp)taosAllocateQitem,
+    if (pWrapper->pProc == NULL) {
-                     .childFreeHeadFp = (ProcFreeFp)taosFreeQitem,
+      dError("node:%s, failed to create proc since %s", pWrapper->name, terrstr());
-                     .childMallocBodyFp = (ProcMallocFp)rpcMallocCont,
-                     .childFreeBodyFp = (ProcFreeFp)rpcFreeCont,
-                     .parentQueueSize = 1024 * 1024 * 2,  // size will be a configuration item
-                     .parentConsumeFp = (ProcConsumeFp)dndConsumeParentQueue,
-                     .parentdMallocHeadFp = (ProcMallocFp)taosMemoryMalloc,
-                     .parentFreeHeadFp = (ProcFreeFp)taosMemoryFree,
-                     .parentMallocBodyFp = (ProcMallocFp)rpcMallocCont,
-                     .parentFreeBodyFp = (ProcFreeFp)rpcFreeCont,
-                     .pParent = pWrapper,
-                     .name = pWrapper->name};
-    SProcObj *pProc = taosProcInit(&cfg);
-    if (pProc == NULL) {
-      dError("node:%s, failed to fork since %s", pWrapper->name, terrstr());
       return -1;
     }
+  }
 
-    pWrapper->pProc = pProc;
+  if (dndWriteShmFile(pDnode) != 0) {
+    dError("failed to write runtime file since %s", terrstr());
+    return -1;
+  }
 
-    if (taosProcIsChild(pProc)) {
+  for (ENodeType n = DNODE + 1; n < NODE_MAX; ++n) {
-      dInfo("node:%s, will start in child process", pWrapper->name);
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
-      pWrapper->procType = PROC_CHILD;
+    if (!pWrapper->required) continue;
-      dndResetLog(pWrapper);
 
-      dInfo("node:%s, clean up resources inherited from parent", pWrapper->name);
+    if (pDnode->ntype == NODE_MAX) {
-      dndClearNodesExecpt(pDnode, n);
+      dInfo("node:%s, should be started manually", pWrapper->name);
-
-      dInfo("node:%s, will be initialized in child process", pWrapper->name);
-      if (dndOpenNode(pWrapper) != 0) {
-        dInfo("node:%s, failed to init in child process since %s", pWrapper->name, terrstr());
-        return -1;
-      }
-
-      if (taosProcRun(pProc) != 0) {
-        dError("node:%s, failed to run proc since %s", pWrapper->name, terrstr());
-        return -1;
-      }
-      break;
     } else {
-      dInfo("node:%s, will not start in parent process, child pid:%d", pWrapper->name, taosProcChildId(pProc));
+      char    args[PATH_MAX];
-      pWrapper->procType = PROC_PARENT;
+      int32_t pid = taosNewProc(args);
-      if (taosProcRun(pProc) != 0) {
+      if (pid <= 0) {
-        dError("node:%s, failed to run proc since %s", pWrapper->name, terrstr());
+        terrno = TAOS_SYSTEM_ERROR(errno);
+        dError("node:%s, failed to exec in new process since %s", pWrapper->name, terrstr());
         return -1;
       }
+      pWrapper->procId = pid;
+      dInfo("node:%s, run in new process, pid:%d", pWrapper->name, pid);
+    }
+
+    if (taosProcRun(pWrapper->pProc) != 0) {
+      dError("node:%s, failed to run proc since %s", pWrapper->name, terrstr());
+      return -1;
     }
   }
 
   dndSetStatus(pDnode, DND_STAT_RUNNING);
 
-  for (ENodeType n = 0; n < NODE_MAX; ++n) {
+  if ((*pDWrapper->fp.startFp)(pDWrapper) != 0) {
-    SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
+    dError("node:%s, failed to start since %s", pDWrapper->name, terrstr());
-    if (!pWrapper->required) continue;
+    return -1;
-    if (pWrapper->fp.startFp == NULL) continue;
+  }
-    if (pWrapper->procType == PROC_PARENT && n != DNODE) continue;
+
-    if (pWrapper->procType == PROC_CHILD && n == DNODE) continue;
+  dInfo("dnode running in parent process");
+  return 0;
+}
+
+static int32_t dndRunInChildProcess(SDnode *pDnode) {
+  dInfo("dnode start to run in child process");
+  SMgmtWrapper *pWrapper = &pDnode->wrappers[pDnode->ntype];
+
+  SMsgCb msgCb = dndCreateMsgcb(pWrapper);
+  tmsgSetDefaultMsgCb(&msgCb);
+  pWrapper->procType = PROC_CHILD;
+
+  if (dndOpenNode(pWrapper) != 0) {
+    dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
+    return -1;
+  }
+
+  SProcCfg cfg = {.childConsumeFp = (ProcConsumeFp)dndConsumeChildQueue,
+                  .childMallocHeadFp = (ProcMallocFp)taosAllocateQitem,
+                  .childFreeHeadFp = (ProcFreeFp)taosFreeQitem,
+                  .childMallocBodyFp = (ProcMallocFp)rpcMallocCont,
+                  .childFreeBodyFp = (ProcFreeFp)rpcFreeCont,
+                  .parentConsumeFp = (ProcConsumeFp)dndConsumeParentQueue,
+                  .parentMallocHeadFp = (ProcMallocFp)taosMemoryMalloc,
+                  .parentFreeHeadFp = (ProcFreeFp)taosMemoryFree,
+                  .parentMallocBodyFp = (ProcMallocFp)rpcMallocCont,
+                  .parentFreeBodyFp = (ProcFreeFp)rpcFreeCont,
+                  .shm = pWrapper->shm,
+                  .pParent = pWrapper,
+                  .isChild = true,
+                  .name = pWrapper->name};
+
+  pWrapper->pProc = taosProcInit(&cfg);
+  if (pWrapper->pProc == NULL) {
+    dError("node:%s, failed to create proc since %s", pWrapper->name, terrstr());
+    return -1;
+  }
+
+  if (pWrapper->fp.startFp != NULL) {
     if ((*pWrapper->fp.startFp)(pWrapper) != 0) {
       dError("node:%s, failed to start since %s", pWrapper->name, terrstr());
       return -1;
     }
   }
 
+  if (taosProcRun(pWrapper->pProc) != 0) {
+    dError("node:%s, failed to run proc since %s", pWrapper->name, terrstr());
+    return -1;
+  }
+
+  dInfo("dnode running in child process");
   return 0;
 }
 
-int32_t dndRun(SDnode *pDnode) {
+int32_t dndRun(SDnode * pDnode) {
-  if (tsMultiProcess == 0) {
+  if (!tsMultiProcess) {
     if (dndRunInSingleProcess(pDnode) != 0) {
-      dError("failed to run dnode in single process mode since %s", terrstr());
+      dError("failed to run dnode since %s", terrstr());
+      return -1;
+    }
+  } else if (pDnode->ntype == DNODE || pDnode->ntype == NODE_MAX) {
+    if (dndRunInParentProcess(pDnode) != 0) {
+      dError("failed to run dnode in parent process since %s", terrstr());
       return -1;
     }
   } else {
-    if (dndRunInMultiProcess(pDnode) != 0) {
+    if (dndRunInChildProcess(pDnode) != 0) {
-      dError("failed to run dnode in multi process mode since %s", terrstr());
+      dError("failed to run dnode in child process since %s", terrstr());
       return -1;
     }
   }
 
   dndReportStartup(pDnode, "TDengine", "initialized successfully");
+  dInfo("TDengine initialized successfully");
 
   while (1) {
     if (pDnode->event == DND_EVENT_STOP) {

source/dnode/mgmt/main/src/dndFile.c

@@ -19,13 +19,12 @@
 #define MAXLEN 1024
 
 int32_t dndReadFile(SMgmtWrapper *pWrapper, bool *pDeployed) {
-  int32_t       code = TSDB_CODE_NODE_PARSE_FILE_ERROR;
+  int32_t   code = TSDB_CODE_NODE_PARSE_FILE_ERROR;
-  int32_t       len = 0;
+  int64_t   len = 0;
-  const int32_t maxLen = MAXLEN;
+  char      content[MAXLEN + 1] = {0};
-  char          content[MAXLEN + 1] = {0};
+  cJSON    *root = NULL;
-  cJSON        *root = NULL;
+  char      file[PATH_MAX];
-  char          file[PATH_MAX];
+  TdFilePtr pFile = NULL;
-  TdFilePtr     pFile = NULL;
 
   snprintf(file, sizeof(file), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
   pFile = taosOpenFile(file, TD_FILE_READ);
@@ -35,13 +34,12 @@ int32_t dndReadFile(SMgmtWrapper *pWrapper, bool *pDeployed) {
     goto _OVER;
   }
 
-  len = (int32_t)taosReadFile(pFile, content, maxLen);
+  len = taosReadFile(pFile, content, MAXLEN);
   if (len <= 0) {
     dError("failed to read %s since content is null", file);
     goto _OVER;
   }
 
-  content[len] = 0;
   root = cJSON_Parse(content);
   if (root == NULL) {
     dError("failed to read %s since invalid json format", file);
@@ -55,8 +53,8 @@ int32_t dndReadFile(SMgmtWrapper *pWrapper, bool *pDeployed) {
   }
   *pDeployed = deployed->valueint != 0;
 
-  code = 0;
   dDebug("succcessed to read file %s, deployed:%d", file, *pDeployed);
+  code = 0;
 
 _OVER:
   if (root != NULL) cJSON_Delete(root);
@@ -67,31 +65,41 @@ _OVER:
 }
 
 int32_t dndWriteFile(SMgmtWrapper *pWrapper, bool deployed) {
-  char file[PATH_MAX] = {0};
+  int32_t   code = -1;
-  snprintf(file, sizeof(file), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
+  int32_t   len = 0;
+  char      content[MAXLEN + 1] = {0};
+  char      file[PATH_MAX] = {0};
+  char      realfile[PATH_MAX] = {0};
+  TdFilePtr pFile = NULL;
 
-  TdFilePtr pFile = taosOpenFile(file, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  snprintf(file, sizeof(file), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
+  snprintf(realfile, sizeof(realfile), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
+
+  pFile = taosOpenFile(file, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
   if (pFile == NULL) {
     terrno = TAOS_SYSTEM_ERROR(errno);
     dError("failed to write %s since %s", file, terrstr());
-    return -1;
+    goto _OVER;
   }
 
-  int32_t       len = 0;
+  len += snprintf(content + len, MAXLEN - len, "{\n");
-  const int32_t maxLen = MAXLEN;
+  len += snprintf(content + len, MAXLEN - len, "  \"deployed\": %d\n", deployed);
-  char          content[MAXLEN + 1] = {0};
+  len += snprintf(content + len, MAXLEN - len, "}\n");
 
-  len += snprintf(content + len, maxLen - len, "{\n");
+  if (taosWriteFile(pFile, content, len) != len) {
-  len += snprintf(content + len, maxLen - len, "  \"deployed\": %d\n", deployed);
+    terrno = TAOS_SYSTEM_ERROR(errno);
-  len += snprintf(content + len, maxLen - len, "}\n");
+    dError("failed to write file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  if (taosFsyncFile(pFile) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to fsync file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
 
-  taosWriteFile(pFile, content, len);
-  taosFsyncFile(pFile);
   taosCloseFile(&pFile);
 
-  char realfile[PATH_MAX] = {0};
-  snprintf(realfile, sizeof(realfile), "%s%s%s.json", pWrapper->path, TD_DIRSEP, pWrapper->name);
-
   if (taosRenameFile(file, realfile) != 0) {
     terrno = TAOS_SYSTEM_ERROR(errno);
     dError("failed to rename %s since %s", file, terrstr());
@@ -99,5 +107,163 @@ int32_t dndWriteFile(SMgmtWrapper *pWrapper, bool deployed) {
   }
 
   dInfo("successed to write %s, deployed:%d", realfile, deployed);
-  return 0;
+  code = 0;
+
+_OVER:
+  if (pFile != NULL) {
+    taosCloseFile(&pFile);
+  }
+
+  return code;
+}
+
+TdFilePtr dndCheckRunning(const char *dataDir) {
+  char filepath[PATH_MAX] = {0};
+  snprintf(filepath, sizeof(filepath), "%s%s.running", dataDir, TD_DIRSEP);
+
+  TdFilePtr pFile = taosOpenFile(filepath, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  if (pFile == NULL) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to lock file:%s since %s", filepath, terrstr());
+    return NULL;
+  }
+
+  int32_t ret = taosLockFile(pFile);
+  if (ret != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to lock file:%s since %s", filepath, terrstr());
+    taosCloseFile(&pFile);
+    return NULL;
+  }
+
+  dDebug("file:%s is locked", filepath);
+  return pFile;
+}
+
+int32_t dndReadShmFile(SDnode *pDnode) {
+  int32_t   code = -1;
+  char      itemName[24] = {0};
+  char      content[MAXLEN + 1] = {0};
+  char      file[PATH_MAX] = {0};
+  cJSON    *root = NULL;
+  TdFilePtr pFile = NULL;
+
+  snprintf(file, sizeof(file), "%s%s.shmfile", pDnode->dataDir, TD_DIRSEP);
+  pFile = taosOpenFile(file, TD_FILE_READ);
+  if (pFile == NULL) {
+    dDebug("file %s not exist", file);
+    code = 0;
+    goto _OVER;
+  }
+
+  if (taosReadFile(pFile, content, MAXLEN) > 0) {
+    root = cJSON_Parse(content);
+    if (root == NULL) {
+      terrno = TSDB_CODE_NODE_PARSE_FILE_ERROR;
+      dError("failed to read %s since invalid json format", file);
+      goto _OVER;
+    }
+
+    for (ENodeType ntype = DNODE + 1; ntype < NODE_MAX; ++ntype) {
+      snprintf(itemName, sizeof(itemName), "%s_shmid", dndNodeProcStr(ntype));
+      cJSON *shmid = cJSON_GetObjectItem(root, itemName);
+      if (shmid && shmid->type == cJSON_Number) {
+        pDnode->wrappers[ntype].shm.id = shmid->valueint;
+      }
+
+      snprintf(itemName, sizeof(itemName), "%s_shmsize", dndNodeProcStr(ntype));
+      cJSON *shmsize = cJSON_GetObjectItem(root, itemName);
+      if (shmsize && shmsize->type == cJSON_Number) {
+        pDnode->wrappers[ntype].shm.size = shmsize->valueint;
+      }
+    }
+  }
+
+  if (!tsMultiProcess || pDnode->ntype == DNODE || pDnode->ntype == DNODE) {
+    for (ENodeType ntype = DNODE; ntype < NODE_MAX; ++ntype) {
+      SMgmtWrapper *pWrapper = &pDnode->wrappers[ntype];
+      if (pWrapper->shm.id >= 0) {
+        dDebug("shmid:%d, is closed, size:%d", pWrapper->shm.id, pWrapper->shm.size);
+        taosDropShm(&pWrapper->shm);
+      }
+    }
+  } else {
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[pDnode->ntype];
+    if (taosAttachShm(&pWrapper->shm) != 0) {
+      terrno = TAOS_SYSTEM_ERROR(errno);
+      dError("shmid:%d, failed to attach shm since %s", pWrapper->shm.id, terrstr());
+      goto _OVER;
+    }
+    dDebug("shmid:%d, is attached, size:%d", pWrapper->shm.id, pWrapper->shm.size);
+  }
+
+  dDebug("successed to open %s", file);
+  code = 0;
+
+_OVER:
+  if (root != NULL) cJSON_Delete(root);
+  if (pFile != NULL) taosCloseFile(&pFile);
+
+  return code;
+}
+
+int32_t dndWriteShmFile(SDnode *pDnode) {
+  int32_t   code = -1;
+  int32_t   len = 0;
+  char      content[MAXLEN + 1] = {0};
+  char      file[PATH_MAX] = {0};
+  char      realfile[PATH_MAX] = {0};
+  TdFilePtr pFile = NULL;
+
+  snprintf(file, sizeof(file), "%s%s.shmfile.bak", pDnode->dataDir, TD_DIRSEP);
+  snprintf(realfile, sizeof(realfile), "%s%s.shmfile", pDnode->dataDir, TD_DIRSEP);
+
+  pFile = taosOpenFile(file, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
+  if (pFile == NULL) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to open file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  len += snprintf(content + len, MAXLEN - len, "{\n");
+  for (ENodeType ntype = DNODE + 1; ntype < NODE_MAX; ++ntype) {
+    SMgmtWrapper *pWrapper = &pDnode->wrappers[ntype];
+    len += snprintf(content + len, MAXLEN - len, "  \"%s_shmid\":%d,\n", dndNodeProcStr(ntype), pWrapper->shm.id);
+    if (ntype == NODE_MAX - 1) {
+      len += snprintf(content + len, MAXLEN - len, "  \"%s_shmsize\":%d\n", dndNodeProcStr(ntype), pWrapper->shm.size);
+    } else {
+      len += snprintf(content + len, MAXLEN - len, "  \"%s_shmsize\":%d,\n", dndNodeProcStr(ntype), pWrapper->shm.size);
+    }
+  }
+  len += snprintf(content + len, MAXLEN - len, "}\n");
+
+  if (taosWriteFile(pFile, content, len) != len) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to write file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  if (taosFsyncFile(pFile) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to fsync file:%s since %s", file, terrstr());
+    goto _OVER;
+  }
+
+  taosCloseFile(&pFile);
+
+  if (taosRenameFile(file, realfile) != 0) {
+    terrno = TAOS_SYSTEM_ERROR(errno);
+    dError("failed to rename %s to %s since %s", file, realfile, terrstr());
+    return -1;
+  }
+
+  dInfo("successed to write %s", realfile);
+  code = 0;
+
+_OVER:
+  if (pFile != NULL) {
+    taosCloseFile(&pFile);
+  }
+
+  return code;
 }

source/dnode/mgmt/main/src/dndInt.c

@@ -20,7 +20,7 @@
 static int8_t once = DND_ENV_INIT;
 
 int32_t dndInit() {
-  dInfo("start to init dnode env");
+  dDebug("start to init dnode env");
   if (atomic_val_compare_exchange_8(&once, DND_ENV_INIT, DND_ENV_READY) != DND_ENV_INIT) {
     terrno = TSDB_CODE_REPEAT_INIT;
     dError("failed to init dnode env since %s", terrstr());
@@ -31,12 +31,6 @@ int32_t dndInit() {
   taosBlockSIGPIPE();
   taosResolveCRC();
 
-  if (rpcInit() != 0) {
-    dError("failed to init rpc since %s", terrstr());
-    dndCleanup();
-    return -1;
-  }
-
   SMonCfg monCfg = {0};
   monCfg.maxLogs = tsMonitorMaxLogs;
   monCfg.port = tsMonitorPort;
@@ -44,7 +38,6 @@ int32_t dndInit() {
   monCfg.comp = tsMonitorComp;
   if (monInit(&monCfg) != 0) {
     dError("failed to init monitor since %s", terrstr());
-    dndCleanup();
     return -1;
   }
 
@@ -53,20 +46,19 @@ int32_t dndInit() {
 }
 
 void dndCleanup() {
-  dInfo("start to cleanup dnode env");
+  dDebug("start to cleanup dnode env");
   if (atomic_val_compare_exchange_8(&once, DND_ENV_READY, DND_ENV_CLEANUP) != DND_ENV_READY) {
     dError("dnode env is already cleaned up");
     return;
   }
 
   monCleanup();
-  rpcCleanup();
   walCleanUp();
   taosStopCacheRefreshWorker();
   dInfo("dnode env is cleaned up");
 }
 
-void dndSetMsgHandle(SMgmtWrapper *pWrapper, int32_t msgType, NodeMsgFp nodeMsgFp, int32_t vgId) {
+void dndSetMsgHandle(SMgmtWrapper *pWrapper, tmsg_t msgType, NodeMsgFp nodeMsgFp, int8_t vgId) {
   pWrapper->msgFps[TMSG_INDEX(msgType)] = nodeMsgFp;
   pWrapper->msgVgIds[TMSG_INDEX(msgType)] = vgId;
 }
@@ -92,29 +84,6 @@ void dndGetStartup(SDnode *pDnode, SStartupReq *pStartup) {
   pStartup->finished = (dndGetStatus(pDnode) == DND_STAT_RUNNING);
 }
 
-TdFilePtr dndCheckRunning(const char *dataDir) {
-  char filepath[PATH_MAX] = {0};
-  snprintf(filepath, sizeof(filepath), "%s/.running", dataDir);
-
-  TdFilePtr pFile = taosOpenFile(filepath, TD_FILE_CTEATE | TD_FILE_WRITE | TD_FILE_TRUNC);
-  if (pFile == NULL) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to lock file:%s since %s", filepath, terrstr());
-    return NULL;
-  }
-
-  int32_t ret = taosLockFile(pFile);
-  if (ret != 0) {
-    terrno = TAOS_SYSTEM_ERROR(errno);
-    dError("failed to lock file:%s since %s", filepath, terrstr());
-    taosCloseFile(&pFile);
-    return NULL;
-  }
-
-  dDebug("file:%s is locked", filepath);
-  return pFile;
-}
-
 void dndProcessStartupReq(SDnode *pDnode, SRpcMsg *pReq) {
   dDebug("startup req is received");
   SStartupReq *pStartup = rpcMallocCont(sizeof(SStartupReq));

source/dnode/mgmt/main/src/dndObj.c

@@ -16,7 +16,7 @@
 #define _DEFAULT_SOURCE
 #include "dndInt.h"
 
-static int32_t dndInitMemory(SDnode *pDnode, const SDnodeOpt *pOption) {
+static int32_t dndInitVars(SDnode *pDnode, const SDnodeOpt *pOption) {
   pDnode->numOfSupportVnodes = pOption->numOfSupportVnodes;
   pDnode->serverPort = pOption->serverPort;
   pDnode->dataDir = strdup(pOption->dataDir);
@@ -24,8 +24,9 @@ static int32_t dndInitMemory(SDnode *pDnode, const SDnodeOpt *pOption) {
   pDnode->localFqdn = strdup(pOption->localFqdn);
   pDnode->firstEp = strdup(pOption->firstEp);
   pDnode->secondEp = strdup(pOption->secondEp);
-  pDnode->pDisks = pOption->pDisks;
+  pDnode->disks = pOption->disks;
   pDnode->numOfDisks = pOption->numOfDisks;
+  pDnode->ntype = pOption->ntype;
   pDnode->rebootTime = taosGetTimestampMs();
 
   if (pDnode->dataDir == NULL || pDnode->localEp == NULL || pDnode->localFqdn == NULL || pDnode->firstEp == NULL ||
@@ -33,18 +34,26 @@ static int32_t dndInitMemory(SDnode *pDnode, const SDnodeOpt *pOption) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
   }
+
+  if (!tsMultiProcess || pDnode->ntype == DNODE || pDnode->ntype == NODE_MAX) {
+    pDnode->lockfile = dndCheckRunning(pDnode->dataDir);
+    if (pDnode->lockfile == NULL) {
+      return -1;
+    }
+  }
+
   return 0;
 }
 
-static void dndClearMemory(SDnode *pDnode) {
+static void dndClearVars(SDnode *pDnode) {
   for (ENodeType n = 0; n < NODE_MAX; ++n) {
     SMgmtWrapper *pMgmt = &pDnode->wrappers[n];
     taosMemoryFreeClear(pMgmt->path);
   }
-  if (pDnode->pLockFile != NULL) {
+  if (pDnode->lockfile != NULL) {
-    taosUnLockFile(pDnode->pLockFile);
+    taosUnLockFile(pDnode->lockfile);
-    taosCloseFile(&pDnode->pLockFile);
+    taosCloseFile(&pDnode->lockfile);
-    pDnode->pLockFile = NULL;
+    pDnode->lockfile = NULL;
   }
   taosMemoryFreeClear(pDnode->localEp);
   taosMemoryFreeClear(pDnode->localFqdn);
@@ -56,7 +65,7 @@ static void dndClearMemory(SDnode *pDnode) {
 }
 
 SDnode *dndCreate(const SDnodeOpt *pOption) {
-  dInfo("start to create dnode object");
+  dDebug("start to create dnode object");
   int32_t code = -1;
   char    path[PATH_MAX] = {0};
   SDnode *pDnode = NULL;
@@ -67,26 +76,12 @@ SDnode *dndCreate(const SDnodeOpt *pOption) {
     goto _OVER;
   }
 
-  if (dndInitMemory(pDnode, pOption) != 0) {
+  if (dndInitVars(pDnode, pOption) != 0) {
+    dError("failed to init variables since %s", terrstr());
     goto _OVER;
   }
 
   dndSetStatus(pDnode, DND_STAT_INIT);
-  pDnode->pLockFile = dndCheckRunning(pDnode->dataDir);
-  if (pDnode->pLockFile == NULL) {
-    goto _OVER;
-  }
-
-  if (dndInitServer(pDnode) != 0) {
-    dError("failed to init trans server since %s", terrstr());
-    goto _OVER;
-  }
-
-  if (dndInitClient(pDnode) != 0) {
-    dError("failed to init trans client since %s", terrstr());
-    goto _OVER;
-  }
-
   dmGetMgmtFp(&pDnode->wrappers[DNODE]);
   mmGetMgmtFp(&pDnode->wrappers[MNODE]);
   vmGetMgmtFp(&pDnode->wrappers[VNODES]);
@@ -94,14 +89,11 @@ SDnode *dndCreate(const SDnodeOpt *pOption) {
   smGetMgmtFp(&pDnode->wrappers[SNODE]);
   bmGetMgmtFp(&pDnode->wrappers[BNODE]);
 
-  if (dndInitMsgHandle(pDnode) != 0) {
-    goto _OVER;
-  }
-
   for (ENodeType n = 0; n < NODE_MAX; ++n) {
     SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
     snprintf(path, sizeof(path), "%s%s%s", pDnode->dataDir, TD_DIRSEP, pWrapper->name);
     pWrapper->path = strdup(path);
+    pWrapper->shm.id = -1;
     pWrapper->pDnode = pDnode;
     if (pWrapper->path == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
@@ -112,15 +104,27 @@ SDnode *dndCreate(const SDnodeOpt *pOption) {
     taosInitRWLatch(&pWrapper->latch);
   }
 
+  if (dndInitMsgHandle(pDnode) != 0) {
+    dError("failed to msg handles since %s", terrstr());
+    goto _OVER;
+  }
+
+  if (dndReadShmFile(pDnode) != 0) {
+    dError("failed to read shm file since %s", terrstr());
+    goto _OVER;
+  }
+
+  SMsgCb msgCb = dndCreateMsgcb(&pDnode->wrappers[0]);
+  tmsgSetDefaultMsgCb(&msgCb);
+
+  dInfo("dnode object is created, data:%p", pDnode);
   code = 0;
 
 _OVER:
   if (code != 0 && pDnode) {
-    dndClearMemory(pDnode);
+    dndClearVars(pDnode);
     pDnode = NULL;
     dError("failed to create dnode object since %s", terrstr());
-  } else {
-    dInfo("dnode object is created, data:%p", pDnode);
   }
 
   return pDnode;
@@ -137,15 +141,12 @@ void dndClose(SDnode *pDnode) {
   dInfo("start to close dnode, data:%p", pDnode);
   dndSetStatus(pDnode, DND_STAT_STOPPED);
 
-  dndCleanupServer(pDnode);
-  dndCleanupClient(pDnode);
-
   for (ENodeType n = 0; n < NODE_MAX; ++n) {
     SMgmtWrapper *pWrapper = &pDnode->wrappers[n];
     dndCloseNode(pWrapper);
   }
 
-  dndClearMemory(pDnode);
+  dndClearVars(pDnode);
   dInfo("dnode object is closed, data:%p", pDnode);
 }
 

source/dnode/mgmt/vm/src/vmInt.c

@@ -285,7 +285,7 @@ static int32_t vmInit(SMgmtWrapper *pWrapper) {
   tstrncpy(dCfg.dir, pDnode->dataDir, TSDB_FILENAME_LEN);
   dCfg.level = 0;
   dCfg.primary = 1;
-  SDiskCfg *pDisks = pDnode->pDisks;
+  SDiskCfg *pDisks = pDnode->disks;
   int32_t   numOfDisks = pDnode->numOfDisks;
   if (numOfDisks <= 0 || pDisks == NULL) {
     pDisks = &dCfg;

source/dnode/mgmt/vm/src/vmWorker.c

@@ -76,7 +76,7 @@ static void vmProcessFetchQueue(SQueueInfo *pInfo, SNodeMsg *pMsg) {
   SVnodeObj *pVnode = pInfo->ahandle;
 
   dTrace("msg:%p, will be processed in vnode-fetch queue", pMsg);
-  int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg);
+  int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg, pInfo);
   if (code != 0) {
     vmSendRsp(pVnode->pWrapper, pMsg, code);
     dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
@@ -168,7 +168,7 @@ static void vmProcessMergeQueue(SQueueInfo *pInfo, STaosQall *qall, int32_t numO
     taosGetQitem(qall, (void **)&pMsg);
 
     dTrace("msg:%p, will be processed in vnode-merge queue", pMsg);
-    int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg);
+    int32_t code = vnodeProcessFetchMsg(pVnode->pImpl, &pMsg->rpcMsg, pInfo);
     if (code != 0) {
       vmSendRsp(pVnode->pWrapper, pMsg, code);
       dTrace("msg:%p, is freed, result:0x%04x:%s", pMsg, code & 0XFFFF, tstrerror(code));
@@ -414,8 +414,7 @@ int32_t vmStartWorker(SVnodesMgmt *pMgmt) {
   pWPool->max = maxMergeThreads;
   if (tWWorkerInit(pWPool) != 0) return -1;
 
-  SSingleWorkerCfg cfg = {
+  SSingleWorkerCfg cfg = {.min = 1, .max = 1, .name = "vnode-mgmt", .fp = (FItem)vmProcessMgmtQueue, .param = pMgmt};
-      .min = 1, .max = 1, .name = "vnode-mgmt", .fp = (FItem)vmProcessMgmtQueue, .param = pMgmt};
   if (tSingleWorkerInit(&pMgmt->mgmtWorker, &cfg) != 0) {
     dError("failed to start vnode-mgmt worker since %s", terrstr());
     return -1;

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

@@ -26,7 +26,7 @@
 #include "parser.h"
 #include "tname.h"
 
-#define MND_TOPIC_VER_NUMBER   1
+#define MND_TOPIC_VER_NUMBER 1
 #define MND_TOPIC_RESERVE_SIZE 64
 
 static int32_t mndTopicActionInsert(SSdb *pSdb, SMqTopicObj *pTopic);

source/dnode/vnode/CMakeLists.txt

@@ -46,14 +46,11 @@ target_link_libraries(
     PUBLIC stream
     PUBLIC qworker
     PUBLIC sync
+    # TODO: get rid of BDB
+    PUBLIC bdb
+    PUBLIC tdb
 )
 
-if(${META_DB_IMPL} STREQUAL "BDB")
-    target_link_libraries(vnode PUBLIC bdb)
-elseif(${META_DB_IMPL} STREQUAL "TDB")
-    target_link_libraries(vnode PUBLIC tdb)
-endif()
-
 if(${BUILD_TEST})
     add_subdirectory(test)
 endif(${BUILD_TEST})

source/dnode/vnode/inc/tsdb.h

@@ -100,10 +100,11 @@ int32_t tsdbUpdateSmaWindow(STsdb *pTsdb, const char *msg);
  * @brief Insert tSma(Time-range-wise SMA) data from stream computing engine
  *
  * @param pTsdb
+ * @param indexUid
  * @param msg
  * @return int32_t
  */
-int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg);
+int32_t tsdbInsertTSmaData(STsdb *pTsdb, int64_t indexUid, const char *msg);
 
 /**
  * @brief Drop tSma data and local cache.
@@ -130,16 +131,11 @@ int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg);
  * @param pTsdb 
  * @param pData 
  * @param indexUid 
- * @param interval 
- * @param intervalUnit 
- * @param tableUid 
- * @param colId 
  * @param querySKey 
  * @param nMaxResult 
  * @return int32_t 
  */
-int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval, int8_t intervalUnit,
+int32_t tsdbGetTSmaData(STsdb *pTsdb, char *pData, int64_t indexUid, TSKEY querySKey, int32_t nMaxResult);
-                        tb_uid_t tableUid, col_id_t colId, TSKEY querySKey, int32_t nMaxResult);
 
 // STsdbCfg
 int  tsdbOptionsInit(STsdbCfg *);

source/dnode/vnode/inc/vnode.h

@@ -17,8 +17,9 @@
 #define _TD_VNODE_H_
 
 #include "os.h"
-#include "trpc.h"
 #include "tmsgcb.h"
+#include "tqueue.h"
+#include "trpc.h"
 
 #include "meta.h"
 #include "tarray.h"
@@ -166,7 +167,7 @@ int vnodeProcessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg);
  * @param pMsg The request message
  * @return int 0 for success, -1 for failure
  */
-int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg);
+int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo);
 
 /* ------------------------ SVnodeCfg ------------------------ */
 /**
@@ -185,7 +186,6 @@ void vnodeOptionsClear(SVnodeCfg *pOptions);
 
 int vnodeValidateTableHash(SVnodeCfg *pVnodeOptions, char *tableFName);
 
-
 /* ------------------------ FOR COMPILE ------------------------ */
 
 int32_t vnodeAlter(SVnode *pVnode, const SVnodeCfg *pCfg);

source/dnode/vnode/src/inc/tsdbSma.h

@@ -44,15 +44,14 @@ int32_t tsdbRemoveTSmaData(STsdb *pTsdb, STSma *param, STimeWindow *pWin);
 #endif
 
 // internal func
-static FORCE_INLINE int32_t tsdbEncodeTSmaKey(tb_uid_t tableUid, col_id_t colId, TSKEY tsKey, void **pData) {
+static FORCE_INLINE int32_t tsdbEncodeTSmaKey(int64_t groupId, TSKEY tsKey, void **pData) {
   int32_t len = 0;
-  len += taosEncodeFixedI64(pData, tableUid);
-  len += taosEncodeFixedU16(pData, colId);
   len += taosEncodeFixedI64(pData, tsKey);
+  len += taosEncodeFixedI64(pData, groupId);
   return len;
 }
 
-static FORCE_INLINE int tsdbRLockSma(SSmaEnv *pEnv) {
+static FORCE_INLINE int32_t tsdbRLockSma(SSmaEnv *pEnv) {
   int code = taosThreadRwlockRdlock(&(pEnv->lock));
   if (code != 0) {
     terrno = TAOS_SYSTEM_ERROR(code);
@@ -61,7 +60,7 @@ static FORCE_INLINE int tsdbRLockSma(SSmaEnv *pEnv) {
   return 0;
 }
 
-static FORCE_INLINE int tsdbWLockSma(SSmaEnv *pEnv) {
+static FORCE_INLINE int32_t tsdbWLockSma(SSmaEnv *pEnv) {
   int code = taosThreadRwlockWrlock(&(pEnv->lock));
   if (code != 0) {
     terrno = TAOS_SYSTEM_ERROR(code);
@@ -70,7 +69,7 @@ static FORCE_INLINE int tsdbWLockSma(SSmaEnv *pEnv) {
   return 0;
 }
 
-static FORCE_INLINE int tsdbUnLockSma(SSmaEnv *pEnv) {
+static FORCE_INLINE int32_t tsdbUnLockSma(SSmaEnv *pEnv) {
   int code = taosThreadRwlockUnlock(&(pEnv->lock));
   if (code != 0) {
     terrno = TAOS_SYSTEM_ERROR(code);

source/dnode/vnode/src/inc/vnd.h

@@ -197,9 +197,9 @@ int tqCommit(STQ*);
 int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg);
 int32_t tqProcessSetConnReq(STQ* pTq, char* msg);
 int32_t tqProcessRebReq(STQ* pTq, char* msg);
-int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen);
+int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen, int32_t workerId);
 int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen);
-int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen);
+int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen, int32_t workerId);
 
 // sma
 void smaHandleRes(void* pVnode, int64_t smaId, const SArray* data);

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

@@ -15,131 +15,637 @@
 
 #include "metaDef.h"
 
-#include "tdb.h"
+#include "tdbInt.h"
 
 struct SMetaDB {
   TENV *pEnv;
-  TDB * pTbDB;
+  TDB  *pTbDB;
-  TDB * pSchemaDB;
+  TDB  *pSchemaDB;
-  TDB * pNameIdx;
+  TDB  *pNameIdx;
-  TDB * pStbIdx;
+  TDB  *pStbIdx;
-  TDB * pNtbIdx;
+  TDB  *pNtbIdx;
-  TDB * pCtbIdx;
+  TDB  *pCtbIdx;
-  // tag index hash table
-  // suid+colid --> TDB *
-  struct {
-  } tagIdxHt;
 };
 
-#define A(op, flag)                 \
+typedef struct __attribute__((__packed__)) {
-  do {                              \
+  tb_uid_t uid;
-    if ((ret = op) != 0) goto flag; \
+  int32_t  sver;
-  } while (0)
+} SSchemaDbKey;
+
+typedef struct {
+  char    *name;
+  tb_uid_t uid;
+} SNameIdxKey;
+
+typedef struct {
+  tb_uid_t suid;
+  tb_uid_t uid;
+} SCtbIdxKey;
+
+static int   metaEncodeTbInfo(void **buf, STbCfg *pTbCfg);
+static void *metaDecodeTbInfo(void *buf, STbCfg *pTbCfg);
+static int   metaEncodeSchema(void **buf, SSchemaWrapper *pSW);
+static void *metaDecodeSchema(void *buf, SSchemaWrapper *pSW);
+
+static inline int metaUidCmpr(const void *arg1, int len1, const void *arg2, int len2) {
+  tb_uid_t uid1, uid2;
+
+  ASSERT(len1 == sizeof(tb_uid_t));
+  ASSERT(len2 == sizeof(tb_uid_t));
+
+  uid1 = ((tb_uid_t *)arg1)[0];
+  uid2 = ((tb_uid_t *)arg2)[0];
+
+  if (uid1 < uid2) {
+    return -1;
+  }
+  if (uid1 == uid2) {
+    return 0;
+  } else {
+    return 1;
+  }
+}
+
+static inline int metaSchemaKeyCmpr(const void *arg1, int len1, const void *arg2, int len2) {
+  int           c;
+  SSchemaDbKey *pKey1 = (SSchemaDbKey *)arg1;
+  SSchemaDbKey *pKey2 = (SSchemaDbKey *)arg2;
+
+  c = metaUidCmpr(arg1, sizeof(tb_uid_t), arg2, sizeof(tb_uid_t));
+  if (c) return c;
+
+  if (pKey1->sver > pKey2->sver) {
+    return 1;
+  } else if (pKey1->sver == pKey2->sver) {
+    return 0;
+  } else {
+    return -1;
+  }
+}
+
+static inline int metaNameIdxCmpr(const void *arg1, int len1, const void *arg2, int len2) {
+  return strcmp((char *)arg1, (char *)arg2);
+}
+
+static inline int metaCtbIdxCmpr(const void *arg1, int len1, const void *arg2, int len2) {
+  int         c;
+  SCtbIdxKey *pKey1 = (SCtbIdxKey *)arg1;
+  SCtbIdxKey *pKey2 = (SCtbIdxKey *)arg2;
+
+  c = metaUidCmpr(arg1, sizeof(tb_uid_t), arg2, sizeof(tb_uid_t));
+  if (c) return c;
+
+  return metaUidCmpr(&pKey1->uid, sizeof(tb_uid_t), &pKey2->uid, sizeof(tb_uid_t));
+}
 
 int metaOpenDB(SMeta *pMeta) {
-  SMetaDB *pDb;
+  SMetaDB *pMetaDb;
-  TENV *   pEnv;
-  TDB *    pTbDB;
-  TDB *    pSchemaDB;
-  TDB *    pNameIdx;
-  TDB *    pStbIdx;
-  TDB *    pNtbIdx;
-  TDB *    pCtbIdx;
   int      ret;
 
-  pDb = (SMetaDB *)taosMemoryCalloc(1, sizeof(*pDb));
+  // allocate DB handle
-  if (pDb == NULL) {
+  pMetaDb = taosMemoryCalloc(1, sizeof(*pMetaDb));
+  if (pMetaDb == NULL) {
+    // TODO
+    ASSERT(0);
     return -1;
   }
 
-  // Create and open the ENV
+  // open the ENV
-  A((tdbEnvCreate(&pEnv)), _err);
+  ret = tdbEnvOpen(pMeta->path, 4096, 256, &(pMetaDb->pEnv));
-#if 0
+  if (ret < 0) {
-  // Set options of the environment
+    // TODO
-  A(tdbEnvSetPageSize(pEnv, 8192), _err);
+    ASSERT(0);
-  A(tdbEnvSetCacheSize(pEnv, 16 * 1024 * 1024), _err);
+    return -1;
-#endif
+  }
-  A((tdbEnvOpen(&pEnv)), _err);
 
-  // Create and open each DB
+  // open table DB
-  A(tdbCreate(&pTbDB), _err);
+  ret = tdbDbOpen("table.db", sizeof(tb_uid_t), TDB_VARIANT_LEN, metaUidCmpr, pMetaDb->pEnv, &(pMetaDb->pTbDB));
-  A(tdbOpen(&pTbDB, "table.db", NULL, pEnv), _err);
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
 
-  A(tdbCreate(&pSchemaDB), _err);
+  // open schema DB
-  A(tdbOpen(&pSchemaDB, "schema.db", NULL, pEnv), _err);
+  ret = tdbDbOpen("schema.db", sizeof(SSchemaDbKey), TDB_VARIANT_LEN, metaSchemaKeyCmpr, pMetaDb->pEnv,
+                  &(pMetaDb->pSchemaDB));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
 
-  A(tdbCreate(&pNameIdx), _err);
+  ret = tdbDbOpen("name.idx", TDB_VARIANT_LEN, 0, metaNameIdxCmpr, pMetaDb->pEnv, &(pMetaDb->pNameIdx));
-  A(tdbOpen(&pNameIdx, "name.db", NULL, pEnv), _err);
+  if (ret < 0) {
-  // tdbAssociate();
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
 
-  pDb->pEnv = pEnv;
+  ret = tdbDbOpen("stb.idx", sizeof(tb_uid_t), 0, metaUidCmpr, pMetaDb->pEnv, &(pMetaDb->pStbIdx));
-  pDb->pTbDB = pTbDB;
+  if (ret < 0) {
-  pDb->pSchemaDB = pSchemaDB;
+    // TODO
-  pMeta->pDB = pDb;
+    ASSERT(0);
+    return -1;
+  }
+
+  ret = tdbDbOpen("ntb.idx", sizeof(tb_uid_t), 0, metaUidCmpr, pMetaDb->pEnv, &(pMetaDb->pNtbIdx));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
+
+  ret = tdbDbOpen("ctb.idx", sizeof(SCtbIdxKey), 0, metaCtbIdxCmpr, pMetaDb->pEnv, &(pMetaDb->pCtbIdx));
+  if (ret < 0) {
+    // TODO
+    ASSERT(0);
+    return -1;
+  }
+
+  pMeta->pDB = pMetaDb;
   return 0;
-
-_err:
-  return -1;
 }
 
 void metaCloseDB(SMeta *pMeta) {
-  // TODO
+  if (pMeta->pDB) {
+    tdbDbClose(pMeta->pDB->pCtbIdx);
+    tdbDbClose(pMeta->pDB->pNtbIdx);
+    tdbDbClose(pMeta->pDB->pStbIdx);
+    tdbDbClose(pMeta->pDB->pNameIdx);
+    tdbDbClose(pMeta->pDB->pSchemaDB);
+    tdbDbClose(pMeta->pDB->pTbDB);
+    taosMemoryFree(pMeta->pDB);
+  }
 }
 
 int metaSaveTableToDB(SMeta *pMeta, STbCfg *pTbCfg) {
-  // TODO
+  tb_uid_t       uid;
+  SMetaDB       *pMetaDb;
+  void          *pKey;
+  void          *pVal;
+  int            kLen;
+  int            vLen;
+  int            ret;
+  char           buf[512];
+  void          *pBuf;
+  SCtbIdxKey     ctbIdxKey;
+  SSchemaDbKey   schemaDbKey;
+  SSchemaWrapper schemaWrapper;
+
+  pMetaDb = pMeta->pDB;
+
+  // TODO: make this operation pre-process
+  if (pTbCfg->type == META_SUPER_TABLE) {
+    uid = pTbCfg->stbCfg.suid;
+  } else {
+    uid = metaGenerateUid(pMeta);
+  }
+
+  // save to table.db
+  pKey = &uid;
+  kLen = sizeof(uid);
+  pVal = pBuf = buf;
+  metaEncodeTbInfo(&pBuf, pTbCfg);
+  vLen = POINTER_DISTANCE(pBuf, buf);
+  ret = tdbDbInsert(pMetaDb->pTbDB, pKey, kLen, pVal, vLen);
+  if (ret < 0) {
+    return -1;
+  }
+
+  // save to schema.db for META_SUPER_TABLE and META_NORMAL_TABLE
+  if (pTbCfg->type != META_CHILD_TABLE) {
+    schemaDbKey.uid = uid;
+    schemaDbKey.sver = 0;  // TODO
+    pKey = &schemaDbKey;
+    kLen = sizeof(schemaDbKey);
+
+    if (pTbCfg->type == META_SUPER_TABLE) {
+      schemaWrapper.nCols = pTbCfg->stbCfg.nCols;
+      schemaWrapper.pSchema = pTbCfg->stbCfg.pSchema;
+    } else {
+      schemaWrapper.nCols = pTbCfg->ntbCfg.nCols;
+      schemaWrapper.pSchema = pTbCfg->ntbCfg.pSchema;
+    }
+    pVal = pBuf = buf;
+    metaEncodeSchema(&pBuf, &schemaWrapper);
+    vLen = POINTER_DISTANCE(pBuf, buf);
+    ret = tdbDbInsert(pMetaDb->pSchemaDB, pKey, kLen, pVal, vLen);
+    if (ret < 0) {
+      return -1;
+    }
+  }
+
+  // update name.idx
+  int nameLen = strlen(pTbCfg->name);
+  memcpy(buf, pTbCfg->name, nameLen + 1);
+  ((tb_uid_t *)(buf + nameLen + 1))[0] = uid;
+  pKey = buf;
+  kLen = nameLen + 1 + sizeof(uid);
+  pVal = NULL;
+  vLen = 0;
+  ret = tdbDbInsert(pMetaDb->pNameIdx, pKey, kLen, pVal, vLen);
+  if (ret < 0) {
+    return -1;
+  }
+
+  // update other index
+  if (pTbCfg->type == META_SUPER_TABLE) {
+    pKey = &uid;
+    kLen = sizeof(uid);
+    pVal = NULL;
+    vLen = 0;
+    ret = tdbDbInsert(pMetaDb->pStbIdx, pKey, kLen, pVal, vLen);
+    if (ret < 0) {
+      return -1;
+    }
+  } else if (pTbCfg->type == META_CHILD_TABLE) {
+    ctbIdxKey.suid = pTbCfg->ctbCfg.suid;
+    ctbIdxKey.uid = uid;
+    pKey = &ctbIdxKey;
+    kLen = sizeof(ctbIdxKey);
+    pVal = NULL;
+    vLen = 0;
+    ret = tdbDbInsert(pMetaDb->pCtbIdx, pKey, kLen, pVal, vLen);
+    if (ret < 0) {
+      return -1;
+    }
+  } else if (pTbCfg->type == META_NORMAL_TABLE) {
+    pKey = &uid;
+    kLen = sizeof(uid);
+    pVal = NULL;
+    vLen = 0;
+    ret = tdbDbInsert(pMetaDb->pNtbIdx, pKey, kLen, pVal, vLen);
+    if (ret < 0) {
+      return -1;
+    }
+  }
+
   return 0;
 }
 
 int metaRemoveTableFromDb(SMeta *pMeta, tb_uid_t uid) {
   // TODO
+  ASSERT(0);
   return 0;
 }
 
 STbCfg *metaGetTbInfoByUid(SMeta *pMeta, tb_uid_t uid) {
-  // TODO
+  int      ret;
-  return NULL;
+  SMetaDB *pMetaDb = pMeta->pDB;
+  void    *pKey;
+  void    *pVal;
+  int      kLen;
+  int      vLen;
+  STbCfg  *pTbCfg;
+
+  // Fetch
+  pKey = &uid;
+  kLen = sizeof(uid);
+  pVal = NULL;
+  ret = tdbDbGet(pMetaDb->pTbDB, pKey, kLen, &pVal, &vLen);
+  if (ret < 0) {
+    return NULL;
+  }
+
+  // Decode
+  pTbCfg = taosMemoryMalloc(sizeof(*pTbCfg));
+  metaDecodeTbInfo(pVal, pTbCfg);
+
+  TDB_FREE(pVal);
+
+  return pTbCfg;
 }
 
 STbCfg *metaGetTbInfoByName(SMeta *pMeta, char *tbname, tb_uid_t *uid) {
-  // TODO
+  void *pKey;
-  return NULL;
+  void *pVal;
+  void *ppKey;
+  int   pkLen;
+  int   kLen;
+  int   vLen;
+  int   ret;
+
+  pKey = tbname;
+  kLen = strlen(tbname) + 1;
+  pVal = NULL;
+  ppKey = NULL;
+  ret = tdbDbPGet(pMeta->pDB->pNameIdx, pKey, kLen, &ppKey, &pkLen, &pVal, &vLen);
+  if (ret < 0) {
+    return NULL;
+  }
+
+  ASSERT(pkLen == kLen + sizeof(uid));
+
+  *uid = *(tb_uid_t *)POINTER_SHIFT(ppKey, kLen);
+  TDB_FREE(ppKey);
+  TDB_FREE(pVal);
+
+  return metaGetTbInfoByUid(pMeta, *uid);
 }
 
 SSchemaWrapper *metaGetTableSchema(SMeta *pMeta, tb_uid_t uid, int32_t sver, bool isinline) {
-  // TODO
+  void           *pKey;
-  return NULL;
+  void           *pVal;
+  int             kLen;
+  int             vLen;
+  int             ret;
+  SSchemaDbKey    schemaDbKey;
+  SSchemaWrapper *pSchemaWrapper;
+  void           *pBuf;
+
+  // fetch
+  schemaDbKey.uid = uid;
+  schemaDbKey.sver = sver;
+  pKey = &schemaDbKey;
+  kLen = sizeof(schemaDbKey);
+  pVal = NULL;
+  ret = tdbDbGet(pMeta->pDB->pSchemaDB, pKey, kLen, &pVal, &vLen);
+  if (ret < 0) {
+    return NULL;
+  }
+
+  // decode
+  pBuf = pVal;
+  pSchemaWrapper = taosMemoryMalloc(sizeof(*pSchemaWrapper));
+  metaDecodeSchema(pBuf, pSchemaWrapper);
+
+  TDB_FREE(pVal);
+
+  return pSchemaWrapper;
 }
 
 STSchema *metaGetTbTSchema(SMeta *pMeta, tb_uid_t uid, int32_t sver) {
-  // TODO
+  tb_uid_t        quid;
-  return NULL;
+  SSchemaWrapper *pSW;
+  STSchemaBuilder sb;
+  SSchema        *pSchema;
+  STSchema       *pTSchema;
+  STbCfg         *pTbCfg;
+
+  pTbCfg = metaGetTbInfoByUid(pMeta, uid);
+  if (pTbCfg->type == META_CHILD_TABLE) {
+    quid = pTbCfg->ctbCfg.suid;
+  } else {
+    quid = uid;
+  }
+
+  pSW = metaGetTableSchema(pMeta, quid, sver, true);
+  if (pSW == NULL) {
+    return NULL;
+  }
+
+  tdInitTSchemaBuilder(&sb, 0);
+  for (int i = 0; i < pSW->nCols; i++) {
+    pSchema = pSW->pSchema + i;
+    tdAddColToSchema(&sb, pSchema->type, pSchema->colId, pSchema->bytes);
+  }
+  pTSchema = tdGetSchemaFromBuilder(&sb);
+  tdDestroyTSchemaBuilder(&sb);
+
+  return pTSchema;
 }
 
+struct SMTbCursor {
+  TDBC *pDbc;
+};
+
 SMTbCursor *metaOpenTbCursor(SMeta *pMeta) {
-  // TODO
+  SMTbCursor *pTbCur = NULL;
-  return NULL;
+  SMetaDB    *pDB = pMeta->pDB;
+
+  pTbCur = (SMTbCursor *)taosMemoryCalloc(1, sizeof(*pTbCur));
+  if (pTbCur == NULL) {
+    return NULL;
+  }
+
+  tdbDbcOpen(pDB->pTbDB, &pTbCur->pDbc);
+
+  return pTbCur;
 }
 
 void metaCloseTbCursor(SMTbCursor *pTbCur) {
-  // TODO
+  if (pTbCur) {
+    if (pTbCur->pDbc) {
+      tdbDbcClose(pTbCur->pDbc);
+    }
+    taosMemoryFree(pTbCur);
+  }
 }
 
 char *metaTbCursorNext(SMTbCursor *pTbCur) {
-  // TODO
+  void  *pKey = NULL;
+  void  *pVal = NULL;
+  int    kLen;
+  int    vLen;
+  int    ret;
+  void  *pBuf;
+  STbCfg tbCfg;
+
+  for (;;) {
+    ret = tdbDbNext(pTbCur->pDbc, &pKey, &kLen, &pVal, &vLen);
+    if (ret < 0) break;
+    pBuf = pVal;
+    metaDecodeTbInfo(pBuf, &tbCfg);
+    if (tbCfg.type == META_SUPER_TABLE) {
+      taosMemoryFree(tbCfg.name);
+      taosMemoryFree(tbCfg.stbCfg.pTagSchema);
+      continue;
+      ;
+    } else if (tbCfg.type == META_CHILD_TABLE) {
+      kvRowFree(tbCfg.ctbCfg.pTag);
+    }
+
+    return tbCfg.name;
+  }
+
   return NULL;
 }
 
+struct SMCtbCursor {
+  TDBC    *pCur;
+  tb_uid_t suid;
+  void    *pKey;
+  void    *pVal;
+  int      kLen;
+  int      vLen;
+};
+
 SMCtbCursor *metaOpenCtbCursor(SMeta *pMeta, tb_uid_t uid) {
-  // TODO
+  SMCtbCursor *pCtbCur = NULL;
-  return NULL;
+  SMetaDB     *pDB = pMeta->pDB;
+  int          ret;
+
+  pCtbCur = (SMCtbCursor *)taosMemoryCalloc(1, sizeof(*pCtbCur));
+  if (pCtbCur == NULL) {
+    return NULL;
+  }
+
+  pCtbCur->suid = uid;
+  ret = tdbDbcOpen(pDB->pCtbIdx, &pCtbCur->pCur);
+  if (ret < 0) {
+    taosMemoryFree(pCtbCur);
+    return NULL;
+  }
+
+  // TODO: move the cursor to the suid there
+
+  return pCtbCur;
 }
 
 void metaCloseCtbCurosr(SMCtbCursor *pCtbCur) {
-  // TODO
+  if (pCtbCur) {
+    if (pCtbCur->pCur) {
+      tdbDbcClose(pCtbCur->pCur);
+
+      TDB_FREE(pCtbCur->pKey);
+      TDB_FREE(pCtbCur->pVal);
+    }
+
+    taosMemoryFree(pCtbCur);
+  }
 }
 
 tb_uid_t metaCtbCursorNext(SMCtbCursor *pCtbCur) {
+  int         ret;
+  SCtbIdxKey *pCtbIdxKey;
+
+  ret = tdbDbNext(pCtbCur->pCur, &pCtbCur->pKey, &pCtbCur->kLen, &pCtbCur->pVal, &pCtbCur->vLen);
+  if (ret < 0) {
+    return 0;
+  }
+
+  pCtbIdxKey = pCtbCur->pVal;
+
+  return pCtbIdxKey->uid;
+}
+
+int metaGetTbNum(SMeta *pMeta) {
   // TODO
+  // ASSERT(0);
   return 0;
 }
+
+STSmaWrapper *metaGetSmaInfoByTable(SMeta *pMeta, tb_uid_t uid) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+int metaRemoveSmaFromDb(SMeta *pMeta, int64_t indexUid) {
+  // TODO
+  ASSERT(0);
+  return 0;
+}
+
+int metaSaveSmaToDB(SMeta *pMeta, STSma *pSmaCfg) {
+  // TODO
+  ASSERT(0);
+  return 0;
+}
+
+STSma *metaGetSmaInfoByIndex(SMeta *pMeta, int64_t indexUid) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+const char *metaSmaCursorNext(SMSmaCursor *pCur) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+void metaCloseSmaCurosr(SMSmaCursor *pCur) {
+  // TODO
+  ASSERT(0);
+}
+
+SArray *metaGetSmaTbUids(SMeta *pMeta, bool isDup) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+SMSmaCursor *metaOpenSmaCursor(SMeta *pMeta, tb_uid_t uid) {
+  // TODO
+  ASSERT(0);
+  return NULL;
+}
+
+static int metaEncodeSchema(void **buf, SSchemaWrapper *pSW) {
+  int      tlen = 0;
+  SSchema *pSchema;
+
+  tlen += taosEncodeFixedU32(buf, pSW->nCols);
+  for (int i = 0; i < pSW->nCols; i++) {
+    pSchema = pSW->pSchema + i;
+    tlen += taosEncodeFixedI8(buf, pSchema->type);
+    tlen += taosEncodeFixedI16(buf, pSchema->colId);
+    tlen += taosEncodeFixedI32(buf, pSchema->bytes);
+    tlen += taosEncodeString(buf, pSchema->name);
+  }
+
+  return tlen;
+}
+
+static void *metaDecodeSchema(void *buf, SSchemaWrapper *pSW) {
+  SSchema *pSchema;
+
+  buf = taosDecodeFixedU32(buf, &pSW->nCols);
+  pSW->pSchema = (SSchema *)taosMemoryMalloc(sizeof(SSchema) * pSW->nCols);
+  for (int i = 0; i < pSW->nCols; i++) {
+    pSchema = pSW->pSchema + i;
+    buf = taosDecodeFixedI8(buf, &pSchema->type);
+    buf = taosDecodeFixedI16(buf, &pSchema->colId);
+    buf = taosDecodeFixedI32(buf, &pSchema->bytes);
+    buf = taosDecodeStringTo(buf, pSchema->name);
+  }
+
+  return buf;
+}
+
+static int metaEncodeTbInfo(void **buf, STbCfg *pTbCfg) {
+  int tsize = 0;
+
+  tsize += taosEncodeString(buf, pTbCfg->name);
+  tsize += taosEncodeFixedU32(buf, pTbCfg->ttl);
+  tsize += taosEncodeFixedU32(buf, pTbCfg->keep);
+  tsize += taosEncodeFixedU8(buf, pTbCfg->info);
+
+  if (pTbCfg->type == META_SUPER_TABLE) {
+    SSchemaWrapper sw = {.nCols = pTbCfg->stbCfg.nTagCols, .pSchema = pTbCfg->stbCfg.pTagSchema};
+    tsize += metaEncodeSchema(buf, &sw);
+  } else if (pTbCfg->type == META_CHILD_TABLE) {
+    tsize += taosEncodeFixedU64(buf, pTbCfg->ctbCfg.suid);
+    tsize += tdEncodeKVRow(buf, pTbCfg->ctbCfg.pTag);
+  } else if (pTbCfg->type == META_NORMAL_TABLE) {
+    // TODO
+  } else {
+    ASSERT(0);
+  }
+
+  return tsize;
+}
+
+static void *metaDecodeTbInfo(void *buf, STbCfg *pTbCfg) {
+  buf = taosDecodeString(buf, &(pTbCfg->name));
+  buf = taosDecodeFixedU32(buf, &(pTbCfg->ttl));
+  buf = taosDecodeFixedU32(buf, &(pTbCfg->keep));
+  buf = taosDecodeFixedU8(buf, &(pTbCfg->info));
+
+  if (pTbCfg->type == META_SUPER_TABLE) {
+    SSchemaWrapper sw;
+    buf = metaDecodeSchema(buf, &sw);
+    pTbCfg->stbCfg.nTagCols = sw.nCols;
+    pTbCfg->stbCfg.pTagSchema = sw.pSchema;
+  } else if (pTbCfg->type == META_CHILD_TABLE) {
+    buf = taosDecodeFixedU64(buf, &(pTbCfg->ctbCfg.suid));
+    buf = tdDecodeKVRow(buf, &(pTbCfg->ctbCfg.pTag));
+  } else if (pTbCfg->type == META_NORMAL_TABLE) {
+    // TODO
+  } else {
+    ASSERT(0);
+  }
+  return buf;
+}

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

@@ -356,7 +356,6 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
   void*   buf = rpcMallocCont(tlen);
   if (buf == NULL) {
     pMsg->code = -1;
-    ASSERT(0);
     return -1;
   }
   ((SMqRspHead*)buf)->mqMsgType = TMQ_MSG_TYPE__POLL_RSP;
@@ -490,7 +489,7 @@ int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen) {
   return 0;
 }
 
-int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen) {
+int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen, int32_t workerId) {
   void* pIter = NULL;
 
   while (1) {
@@ -498,14 +497,14 @@ int32_t tqProcessStreamTrigger(STQ* pTq, void* data, int32_t dataLen) {
     if (pIter == NULL) break;
     SStreamTask* pTask = (SStreamTask*)pIter;
 
-    if (streamExecTask(pTask, &pTq->pVnode->msgCb, data, STREAM_DATA_TYPE_SUBMIT_BLOCK, 0) < 0) {
+    if (streamExecTask(pTask, &pTq->pVnode->msgCb, data, STREAM_DATA_TYPE_SUBMIT_BLOCK, workerId) < 0) {
       // TODO
     }
   }
   return 0;
 }
 
-int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen) {
+int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen, int32_t workerId) {
   SStreamTaskExecReq req;
   tDecodeSStreamTaskExecReq(msg, &req);
 
@@ -515,7 +514,7 @@ int32_t tqProcessTaskExec(STQ* pTq, char* msg, int32_t msgLen) {
   SStreamTask* pTask = taosHashGet(pTq->pStreamTasks, &taskId, sizeof(int32_t));
   ASSERT(pTask);
 
-  if (streamExecTask(pTask, &pTq->pVnode->msgCb, req.data, STREAM_DATA_TYPE_SSDATA_BLOCK, 0) < 0) {
+  if (streamExecTask(pTask, &pTq->pVnode->msgCb, req.data, STREAM_DATA_TYPE_SSDATA_BLOCK, workerId) < 0) {
     // TODO
   }
   return 0;

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

@@ -167,8 +167,10 @@ SArray* tqRetrieveDataBlock(STqReadHandle* pHandle) {
       if (!tdSTSRowIterNext(&iter, pColData->info.colId, pColData->info.type, &sVal)) {
         break;
       }
-      // TODO handle null
+      if (colDataAppend(pColData, curRow, sVal.val, sVal.valType == TD_VTYPE_NULL) < 0) {
-      colDataAppend(pColData, curRow, sVal.val, sVal.valType == TD_VTYPE_NULL);
+        taosArrayDestroyEx(pArray, (void (*)(void*))tDeleteSSDataBlock);
+        return NULL;
+      }
     }
     curRow++;
   }

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

@@ -24,7 +24,7 @@ static const char *TSDB_SMA_DNAME[] = {
 #define SMA_STORAGE_TSDB_DAYS   30
 #define SMA_STORAGE_TSDB_TIMES  10
 #define SMA_STORAGE_SPLIT_HOURS 24
-#define SMA_KEY_LEN             18  // tableUid_colId_TSKEY 8+2+8
+#define SMA_KEY_LEN             16  // TSKEY+groupId 8+8
 #define SMA_DROP_EXPIRED_TIME   10  // default is 10 seconds
 
 #define SMA_STATE_HASH_SLOT      4
@@ -38,10 +38,10 @@ typedef enum {
 } ESmaStorageLevel;
 
 typedef struct {
-  STsdb       *pTsdb;
+  STsdb        *pTsdb;
-  SDBFile      dFile;
+  SDBFile       dFile;
-  SSDataBlock *pData;     // sma data
+  const SArray *pDataBlocks;  // sma data
-  int32_t      interval;  // interval with the precision of DB
+  int32_t       interval;     // interval with the precision of DB
 } STSmaWriteH;
 
 typedef struct {
@@ -94,26 +94,24 @@ static int32_t  tsdbUnRefSmaStat(STsdb *pTsdb, SSmaStat *pStat);
 
 // read data
 // TODO: This is the basic params, and should wrap the params to a queryHandle.
-static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval,
+static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, char *pData, int64_t indexUid, TSKEY querySKey, int32_t nMaxResult);
-                                   int8_t intervalUnit, tb_uid_t tableUid, col_id_t colId, TSKEY querySKey,
-                                   int32_t nMaxResult);
 
 // insert data
-static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, SSDataBlock *pData, int64_t interval,
+static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, const SArray *pDataBlocks, int64_t interval,
                                   int8_t intervalUnit);
 static void    tsdbDestroyTSmaWriteH(STSmaWriteH *pSmaH);
 static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, int64_t interval, int8_t intervalUnit);
 static int32_t tsdbGetSmaStorageLevel(int64_t interval, int8_t intervalUnit);
-static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *pData);
+static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, int32_t fid);
 static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t keyLen, void *pData, uint32_t dataLen);
-static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision);
+static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision, bool adjusted);
 static int32_t tsdbGetTSmaDays(STsdb *pTsdb, int64_t interval, int32_t storageLevel);
-static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int64_t indexUid, int32_t fid);
+static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, int64_t indexUid, int32_t fid);
 static int32_t tsdbInitTSmaFile(STSmaReadH *pSmaH, int64_t indexUid, TSKEY skey);
 static bool    tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, TSKEY *queryKey);
 static void    tsdbGetSmaDir(int32_t vgId, ETsdbSmaType smaType, char dirName[]);
-static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg);
+static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, int64_t indexUid, const char *msg);
-static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg);
+static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, const char *msg);
 
 // mgmt interface
 static int32_t tsdbDropTSmaDataImpl(STsdb *pTsdb, int64_t indexUid);
@@ -387,7 +385,6 @@ static int32_t tsdbCheckAndInitSmaEnv(STsdb *pTsdb, int8_t smaType) {
   return TSDB_CODE_SUCCESS;
 };
 
-
 static int32_t tsdbSetExpiredWindow(STsdb *pTsdb, SHashObj *pItemsHash, int64_t indexUid, int64_t winSKey) {
   SSmaStatItem *pItem = taosHashGet(pItemsHash, &indexUid, sizeof(indexUid));
   if (pItem == NULL) {
@@ -480,18 +477,15 @@ int32_t tsdbUpdateExpiredWindowImpl(STsdb *pTsdb, const char *msg) {
   TSKEY expiredWindows[SMA_TEST_EXPIRED_WINDOW_SIZE];
 #endif
 
-
   // Firstly, assume that tSma can only be created on super table/normal table.
   // getActiveTimeWindow
 
-
+  SSmaEnv  *pEnv = REPO_SMA_ENV(pTsdb, TSDB_SMA_TYPE_TIME_RANGE);
-  SSmaEnv *pEnv = REPO_SMA_ENV(pTsdb, TSDB_SMA_TYPE_TIME_RANGE);
   SSmaStat *pStat = SMA_ENV_STAT(pEnv);
   SHashObj *pItemsHash = SMA_ENV_STAT_ITEMS(pEnv);
 
   TASSERT(pEnv != NULL && pStat != NULL && pItemsHash != NULL);
 
-
   // basic procedure
   // TODO: optimization
   tsdbRefSmaStat(pTsdb, pStat);
@@ -523,11 +517,11 @@ int32_t tsdbUpdateExpiredWindowImpl(STsdb *pTsdb, const char *msg) {
         tdFreeTSmaWrapper(pSW);
         break;
       }
-      if(pSW == NULL) {
+      if (pSW == NULL) {
-        if((pSW =metaGetSmaInfoByTable(REPO_META(pTsdb), pBlock->suid)) == NULL) {
+        if ((pSW = metaGetSmaInfoByTable(REPO_META(pTsdb), pBlock->suid)) == NULL) {
           break;
         }
-        if((pSW->number) <= 0 || (pSW->tSma == NULL)) {
+        if ((pSW->number) <= 0 || (pSW->tSma == NULL)) {
           tdFreeTSmaWrapper(pSW);
           break;
         }
@@ -657,9 +651,8 @@ static int32_t tsdbGetSmaStorageLevel(int64_t interval, int8_t intervalUnit) {
  */
 static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t keyLen, void *pData, uint32_t dataLen) {
   SDBFile *pDBFile = &pSmaH->dFile;
-  tsdbDebug("vgId:%d insert sma data blocks into %s: smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64 ", dataLen %d",
+  printf("\nvgId:%d insert sma data blocks into %s: smaKey %" PRIx64 "-%" PRIx64 ", dataLen %" PRIu32 "\n",
-            REPO_ID(pSmaH->pTsdb), pDBFile->path, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
+         REPO_ID(pSmaH->pTsdb), pDBFile->path, *(int64_t *)smaKey, *(int64_t *)POINTER_SHIFT(smaKey, 8), dataLen);
-            *(int64_t *)POINTER_SHIFT(smaKey, 10), dataLen);
 
   // TODO: insert sma data blocks into B+Tree(TDB)
   if (tsdbSaveSmaToDB(pDBFile, smaKey, keyLen, pData, dataLen) != 0) {
@@ -683,9 +676,15 @@ static int32_t tsdbInsertTSmaBlocks(STSmaWriteH *pSmaH, void *smaKey, uint32_t k
  * @param interval
  * @param intervalUnit
  * @param precision
+ * @param adjusted Interval already adjusted according to DB precision
  * @return int64_t
  */
-static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision) {
+static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit, int8_t precision, bool adjusted) {
+  
+  if (adjusted) {
+    return interval;
+  }
+
   switch (intervalUnit) {
     case TIME_UNIT_YEAR:  // approximate value
       interval *= 365 * 86400 * 1e3;
@@ -753,59 +752,12 @@ static int64_t tsdbGetIntervalByPrecision(int64_t interval, uint8_t intervalUnit
   return interval;
 }
 
-/**
+static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, const SArray *pDataBlocks, int64_t interval,
- * @brief Split the TSma data blocks into expected size and insert into B+Tree.
+                                  int8_t intervalUnit) {
- *
- * @param pSmaH
- * @param pData
- * @param nOffset The nOffset of blocks since fid changes.
- * @param nBlocks The nBlocks with the same fid since nOffset.
- * @return int32_t
- */
-static int32_t tsdbInsertTSmaDataSection(STSmaWriteH *pSmaH, STSmaDataWrapper *pData) {
-  STsdb *pTsdb = pSmaH->pTsdb;
-
-  tsdbDebug("tsdbInsertTSmaDataSection: index %" PRIi64 ", skey %" PRIi64, pData->indexUid, pData->skey);
-
-  // TODO: check the data integrity
-
-  int32_t len = 0;
-  while (true) {
-    if (len >= pData->dataLen) {
-      break;
-    }
-    assert(pData->dataLen > 0);
-    STSmaTbData *pTbData = (STSmaTbData *)POINTER_SHIFT(pData->data, len);
-
-    int32_t tbLen = 0;
-    while (true) {
-      if (tbLen >= pTbData->dataLen) {
-        break;
-      }
-      assert(pTbData->dataLen > 0);
-      STSmaColData *pColData = (STSmaColData *)POINTER_SHIFT(pTbData->data, tbLen);
-      char          smaKey[SMA_KEY_LEN] = {0};
-      void         *pSmaKey = &smaKey;
-#if 0
-      printf("tsdbInsertTSmaDataSection: index %" PRIi64 ", skey %" PRIi64 " table[%" PRIi64 "]col[%" PRIu16 "]\n",
-             pData->indexUid, pData->skey, pTbData->tableUid, pColData->colId);
-#endif
-      tsdbEncodeTSmaKey(pTbData->tableUid, pColData->colId, pData->skey, (void **)&pSmaKey);
-      if (tsdbInsertTSmaBlocks(pSmaH, smaKey, SMA_KEY_LEN, pColData->data, pColData->blockSize) < 0) {
-        tsdbWarn("vgId:%d insert tSma blocks failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
-      }
-      tbLen += (sizeof(STSmaColData) + pColData->blockSize);
-    }
-    len += (sizeof(STSmaTbData) + pTbData->dataLen);
-  }
-
-  return TSDB_CODE_SUCCESS;
-}
-
-static int32_t tsdbInitTSmaWriteH(STSmaWriteH *pSmaH, STsdb *pTsdb, SSDataBlock *pData, int64_t interval, int8_t intervalUnit) {
   pSmaH->pTsdb = pTsdb;
-  pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision);
+  pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision, true);
-  pSmaH->pData = pData;
+  pSmaH->pDataBlocks = pDataBlocks;
+  pSmaH->dFile.fid = TSDB_IVLD_FID;
   return TSDB_CODE_SUCCESS;
 }
 
@@ -815,7 +767,7 @@ static void tsdbDestroyTSmaWriteH(STSmaWriteH *pSmaH) {
   }
 }
 
-static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int64_t indexUid, int32_t fid) {
+static int32_t tsdbSetTSmaDataFile(STSmaWriteH *pSmaH, int64_t indexUid, int32_t fid) {
   STsdb *pTsdb = pSmaH->pTsdb;
   ASSERT(pSmaH->dFile.path == NULL && pSmaH->dFile.pDB == NULL);
 
@@ -859,11 +811,10 @@ static int32_t tsdbGetTSmaDays(STsdb *pTsdb, int64_t interval, int32_t storageLe
  * @param msg
  * @return int32_t
  */
-static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
+static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, int64_t indexUid, const char *msg) {
-  STsdbCfg    *pCfg = REPO_CFG(pTsdb);
+  STsdbCfg     *pCfg = REPO_CFG(pTsdb);
-  SSDataBlock *pData = (SSDataBlock *)msg;
+  const SArray *pDataBlocks = (const SArray *)msg;
-  SSmaEnv     *pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
+  SSmaEnv      *pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
-  int64_t      indexUid = SMA_TEST_INDEX_UID;
 
   if (pEnv == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
@@ -871,15 +822,15 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
     return terrno;
   }
 
-  if (pData == NULL) {
+  if (pDataBlocks == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
-    tsdbWarn("vgId:%d insert tSma data failed since pData is NULL", REPO_ID(pTsdb));
+    tsdbWarn("vgId:%d insert tSma data failed since pDataBlocks is NULL", REPO_ID(pTsdb));
     return terrno;
   }
 
-  if (taosArrayGetSize(pData->pDataBlock) <= 0) {
+  if (taosArrayGetSize(pDataBlocks) <= 0) {
     terrno = TSDB_CODE_INVALID_PARA;
-    tsdbWarn("vgId:%d insert tSma data failed since pDataBlock is empty", REPO_ID(pTsdb));
+    tsdbWarn("vgId:%d insert tSma data failed since pDataBlocks is empty", REPO_ID(pTsdb));
     return TSDB_CODE_FAILED;
   }
 
@@ -899,10 +850,9 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
   }
 
   STSma *pSma = pItem->pSma;
-
   STSmaWriteH tSmaH = {0};
 
-  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData, pSma->interval, pSma->intervalUnit) != 0) {
+  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pDataBlocks, pSma->interval, pSma->intervalUnit) != 0) {
     return TSDB_CODE_FAILED;
   }
 
@@ -921,33 +871,134 @@ static int32_t tsdbInsertTSmaDataImpl(STsdb *pTsdb, char *msg) {
   int32_t storageLevel = tsdbGetSmaStorageLevel(pSma->interval, pSma->intervalUnit);
   int32_t daysPerFile = tsdbGetTSmaDays(pTsdb, tSmaH.interval, storageLevel);
 
+  // key: skey + groupId
+  char    smaKey[SMA_KEY_LEN] = {0};
+  char    dataBuf[512] = {0};
+  void   *pDataBuf = &dataBuf;
+  int32_t sz = taosArrayGetSize(pDataBlocks);
+  for (int32_t i = 0; i < sz; ++i) {
+    SSDataBlock *pDataBlock = *(SSDataBlock **)taosArrayGet(pDataBlocks, i);
+    int32_t      colNum = pDataBlock->info.numOfCols;
+    int32_t      rows = pDataBlock->info.rows;
+    int32_t      rowSize = pDataBlock->info.rowSize;
+    int64_t      groupId = pDataBlock->info.groupId;
+    for (int32_t j = 0; j < rows; ++j) {
+      printf("|");
+      TSKEY   skey = TSKEY_INITIAL_VAL;  // the start key of TS window by interval
+      void   *pSmaKey = &smaKey;
+      int32_t tlen = 0;
+      for (int32_t k = 0; k < colNum; ++k) {
+        SColumnInfoData *pColInfoData = *(SColumnInfoData **)taosArrayGet(pDataBlock->pDataBlock, k);
+        void            *var = POINTER_SHIFT(pColInfoData->pData, j * pColInfoData->info.bytes);
+        switch (pColInfoData->info.type) {
+          case TSDB_DATA_TYPE_TIMESTAMP:
+            skey = *(TSKEY *)var;
+            printf("==> skey = %" PRIi64 " groupId = %" PRId64 "|", skey, groupId);
+            tsdbEncodeTSmaKey(groupId, skey, &pSmaKey);
+            break;
+          case TSDB_DATA_TYPE_BOOL:
+          case TSDB_DATA_TYPE_UTINYINT:
+            printf(" %15d |", *(uint8_t *)var);
+            tlen += taosEncodeFixedU8(&pDataBuf, *(uint8_t *)var);
+            break;
+          case TSDB_DATA_TYPE_TINYINT:
+            printf(" %15d |", *(int8_t *)var);
+            tlen += taosEncodeFixedI8(&pDataBuf, *(int8_t *)var);
+            break;
+          case TSDB_DATA_TYPE_SMALLINT:
+            printf(" %15d |", *(int16_t *)var);
+            tlen += taosEncodeFixedI16(&pDataBuf, *(int16_t *)var);
+            break;
+          case TSDB_DATA_TYPE_USMALLINT:
+            printf(" %15d |", *(uint16_t *)var);
+            tlen += taosEncodeFixedU16(&pDataBuf, *(uint16_t *)var);
+            break;
+          case TSDB_DATA_TYPE_INT:
+            printf(" %15d |", *(int32_t *)var);
+            tlen += taosEncodeFixedI32(&pDataBuf, *(int32_t *)var);
+            break;
+          case TSDB_DATA_TYPE_FLOAT:
+          case TSDB_DATA_TYPE_UINT:
+            printf(" %15u |", *(uint32_t *)var);
+            tlen += taosEncodeFixedU32(&pDataBuf, *(uint32_t *)var);
+            break;
+          case TSDB_DATA_TYPE_BIGINT:
+            printf(" %15ld |", *(int64_t *)var);
+            tlen += taosEncodeFixedI64(&pDataBuf, *(int64_t *)var);
+            break;
+          case TSDB_DATA_TYPE_DOUBLE:
+          case TSDB_DATA_TYPE_UBIGINT:
+            printf(" %15lu |", *(uint64_t *)var);
+            tlen += taosEncodeFixedU64(&pDataBuf, *(uint64_t *)var);
+            break;
+          case TSDB_DATA_TYPE_NCHAR: {
+            char tmpChar[100] = {0};
+            strncpy(tmpChar, varDataVal(var), varDataLen(var));
+            printf(" %s |", tmpChar);
+            tlen += taosEncodeBinary(&pDataBuf, varDataVal(var), varDataLen(var));
+            break;
+          }
+          case TSDB_DATA_TYPE_VARCHAR: { // TSDB_DATA_TYPE_BINARY
+                      char tmpChar[100] = {0};
+            strncpy(tmpChar, varDataVal(var), varDataLen(var));
+            printf(" %s |", tmpChar);
+            tlen += taosEncodeBinary(&pDataBuf, varDataVal(var), varDataLen(var));
+            break;
+          }
+          case TSDB_DATA_TYPE_VARBINARY:
+            // TODO: add binary/varbinary
+            TASSERT(0);
+          default:
+            printf("the column type %" PRIi16 " is undefined\n", pColInfoData->info.type);
+            TASSERT(0);
+            break;
+        }
+      }
+      if ((tlen > 0) && (skey != TSKEY_INITIAL_VAL)) {
+        int32_t fid = (int32_t)(TSDB_KEY_FID(skey, daysPerFile, pCfg->precision));
 
-#if 0  
+        // Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index
-  int32_t fid = (int32_t)(TSDB_KEY_FID(pData->skey, daysPerFile, pCfg->precision));
+        // file
+        //         - Set and open the DFile or the B+Tree file
+        // TODO: tsdbStartTSmaCommit();
+        if (fid != tSmaH.dFile.fid) {
+          if (tSmaH.dFile.fid != TSDB_IVLD_FID) {
+            tsdbCloseDBF(&tSmaH.dFile);
+          }
+          tsdbSetTSmaDataFile(&tSmaH, indexUid, fid);
+          if (tsdbOpenDBF(pTsdb->pTSmaEnv->dbEnv, &tSmaH.dFile) != 0) {
+            tsdbWarn("vgId:%d open DB file %s failed since %s", REPO_ID(pTsdb),
+                     tSmaH.dFile.path ? tSmaH.dFile.path : "path is NULL", tstrerror(terrno));
+            tsdbDestroyTSmaWriteH(&tSmaH);
+            tsdbUnRefSmaStat(pTsdb, pStat);
+            return TSDB_CODE_FAILED;
+          }
+        }
 
-  // Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index file
+        if (tsdbInsertTSmaBlocks(&tSmaH, &smaKey, SMA_KEY_LEN, pDataBuf, tlen) != 0) {
-  //         - Set and open the DFile or the B+Tree file
+          tsdbWarn("vgId:%d insert tSma data blocks failed for index %" PRIi64 ", skey %" PRIi64 ", groupId %" PRIi64
-  // TODO: tsdbStartTSmaCommit();
+                   " since %s",
-  tsdbSetTSmaDataFile(&tSmaH, pData, indexUid, fid);
+                   REPO_ID(pTsdb), indexUid, skey, groupId, tstrerror(terrno));
-  if (tsdbOpenDBF(pTsdb->pTSmaEnv->dbEnv, &tSmaH.dFile) != 0) {
+          tsdbDestroyTSmaWriteH(&tSmaH);
-    tsdbWarn("vgId:%d open DB file %s failed since %s", REPO_ID(pTsdb),
+          tsdbUnRefSmaStat(pTsdb, pStat);
-             tSmaH.dFile.path ? tSmaH.dFile.path : "path is NULL", tstrerror(terrno));
+          return TSDB_CODE_FAILED;
-    tsdbDestroyTSmaWriteH(&tSmaH);
+        } else {
-    tsdbUnRefSmaStat(pTsdb, pStat);
+          tsdbWarn("vgId:%d insert tSma data blocks success for index %" PRIi64 ", skey %" PRIi64 ", groupId %" PRIi64,
-    return TSDB_CODE_FAILED;
+                   REPO_ID(pTsdb), indexUid, skey, groupId);
+        }
+        // TODO:tsdbEndTSmaCommit();
+
+        // Step 3: reset the SSmaStat
+        tsdbResetExpiredWindow(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv), indexUid, skey);
+      } else {
+        tsdbWarn("vgId:%d invalid data skey:%" PRIi64 ", tlen %" PRIi32 " during insert tSma data for %" PRIi64,
+                 REPO_ID(pTsdb), skey, tlen, indexUid);
+      }
+
+      printf("\n");
+    }
   }
 
-  if (tsdbInsertTSmaDataSection(&tSmaH, pData) != 0) {
-    tsdbWarn("vgId:%d insert tSma data section failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
-    tsdbDestroyTSmaWriteH(&tSmaH);
-    tsdbUnRefSmaStat(pTsdb, pStat);
-    return TSDB_CODE_FAILED;
-  }
-  // TODO:tsdbEndTSmaCommit();
-
-  // Step 3: reset the SSmaStat
-  tsdbResetExpiredWindow(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv), pData->indexUid, pData->skey);
-#endif
   tsdbDestroyTSmaWriteH(&tSmaH);
   tsdbUnRefSmaStat(pTsdb, pStat);
   return TSDB_CODE_SUCCESS;
@@ -1002,7 +1053,7 @@ static int32_t tsdbDropTSmaDataImpl(STsdb *pTsdb, int64_t indexUid) {
   // TODO:
 }
 
-static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData, int32_t fid) {
+static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, int32_t fid) {
   STsdb *pTsdb = pSmaH->pTsdb;
 
   char tSmaFile[TSDB_FILENAME_LEN] = {0};
@@ -1012,11 +1063,11 @@ static int32_t tsdbSetRSmaDataFile(STSmaWriteH *pSmaH, STSmaDataWrapper *pData,
   return TSDB_CODE_SUCCESS;
 }
 
-static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
+static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, const char *msg) {
-  STsdbCfg    *pCfg = REPO_CFG(pTsdb);
+  STsdbCfg     *pCfg = REPO_CFG(pTsdb);
-  SSDataBlock *pData = (SSDataBlock *)msg;
+  const SArray *pDataBlocks = (const SArray *)msg;
-  SSmaEnv     *pEnv = atomic_load_ptr(&pTsdb->pRSmaEnv);
+  SSmaEnv      *pEnv = atomic_load_ptr(&pTsdb->pRSmaEnv);
-  int64_t      indexUid = SMA_TEST_INDEX_UID;
+  int64_t       indexUid = SMA_TEST_INDEX_UID;
 
   if (pEnv == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
@@ -1030,15 +1081,15 @@ static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
     return terrno;
   }
 
-  if (pData == NULL) {
+  if (pDataBlocks == NULL) {
     terrno = TSDB_CODE_INVALID_PTR;
-    tsdbWarn("vgId:%d insert rSma data failed since pData is NULL", REPO_ID(pTsdb));
+    tsdbWarn("vgId:%d insert rSma data failed since pDataBlocks is NULL", REPO_ID(pTsdb));
     return terrno;
   }
 
-  if (taosArrayGetSize(pData->pDataBlock) <= 0) {
+  if (taosArrayGetSize(pDataBlocks) <= 0) {
     terrno = TSDB_CODE_INVALID_PARA;
-    tsdbWarn("vgId:%d insert rSma data failed since pDataBlock is empty", REPO_ID(pTsdb));
+    tsdbWarn("vgId:%d insert rSma data failed since pDataBlocks is empty", REPO_ID(pTsdb));
     return TSDB_CODE_FAILED;
   }
 
@@ -1061,12 +1112,12 @@ static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
 
   STSmaWriteH tSmaH = {0};
 
-  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pData, pSma->interval, pSma->intervalUnit) != 0) {
+  if (tsdbInitTSmaWriteH(&tSmaH, pTsdb, pDataBlocks, pSma->interval, pSma->intervalUnit) != 0) {
     return TSDB_CODE_FAILED;
   }
 
-  char    rPath[TSDB_FILENAME_LEN] = {0};
+  char rPath[TSDB_FILENAME_LEN] = {0};
-  char    aPath[TSDB_FILENAME_LEN] = {0};
+  char aPath[TSDB_FILENAME_LEN] = {0};
   snprintf(rPath, TSDB_FILENAME_LEN, "%s%s%" PRIi64, SMA_ENV_PATH(pEnv), TD_DIRSEP, indexUid);
   tfsAbsoluteName(REPO_TFS(pTsdb), SMA_ENV_DID(pEnv), rPath, aPath);
   if (!taosCheckExistFile(aPath)) {
@@ -1078,7 +1129,7 @@ static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
   // Step 1: Judge the storage level and days
   int32_t storageLevel = tsdbGetSmaStorageLevel(pSma->interval, pSma->intervalUnit);
   int32_t daysPerFile = tsdbGetTSmaDays(pTsdb, tSmaH.interval, storageLevel);
-  #if 0
+#if 0
   int32_t fid = (int32_t)(TSDB_KEY_FID(pData->skey, daysPerFile, pCfg->precision));
 
   // Step 2: Set the DFile for storage of SMA index, and iterate/split the TSma data and store to B+Tree index file
@@ -1119,7 +1170,7 @@ static int32_t tsdbInsertRSmaDataImpl(STsdb *pTsdb, char *msg) {
  */
 static int32_t tsdbInitTSmaReadH(STSmaReadH *pSmaH, STsdb *pTsdb, int64_t interval, int8_t intervalUnit) {
   pSmaH->pTsdb = pTsdb;
-  pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision);
+  pSmaH->interval = tsdbGetIntervalByPrecision(interval, intervalUnit, REPO_CFG(pTsdb)->precision, true);
   pSmaH->storageLevel = tsdbGetSmaStorageLevel(interval, intervalUnit);
   pSmaH->days = tsdbGetTSmaDays(pTsdb, pSmaH->interval, pSmaH->storageLevel);
 }
@@ -1185,17 +1236,11 @@ static bool tsdbSetAndOpenTSmaFile(STSmaReadH *pReadH, TSKEY *queryKey) {
  * @param pTsdb Return the data between queryWin and fill the pData.
  * @param pData
  * @param indexUid
- * @param interval
- * @param intervalUnit
- * @param tableUid
- * @param colId
  * @param pQuerySKey
  * @param nMaxResult The query invoker should control the nMaxResult need to return to avoid OOM.
  * @return int32_t
  */
-static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval,
+static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, char *pData, int64_t indexUid, TSKEY querySKey, int32_t nMaxResult) {
-                                   int8_t intervalUnit, tb_uid_t tableUid, col_id_t colId, TSKEY querySKey,
-                                   int32_t nMaxResult) {
   SSmaEnv *pEnv = atomic_load_ptr(&pTsdb->pTSmaEnv);
 
   if (!pEnv) {
@@ -1243,35 +1288,38 @@ static int32_t tsdbGetTSmaDataImpl(STsdb *pTsdb, STSmaDataWrapper *pData, int64_
     tsdbDebug("vgId:%d skey %" PRIi64 " of window not in expired window for index %" PRIi64, REPO_ID(pTsdb), querySKey,
               indexUid);
   }
-  tsdbUnRefSmaStat(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv));
+
+  STSma *pTSma = pItem->pSma;
+
+  
 
 #endif
 
   STSmaReadH tReadH = {0};
-  tsdbInitTSmaReadH(&tReadH, pTsdb, interval, intervalUnit);
+  tsdbInitTSmaReadH(&tReadH, pTsdb, pTSma->interval, pTSma->intervalUnit);
   tsdbCloseDBF(&tReadH.dFile);
   
+  tsdbUnRefSmaStat(pTsdb, SMA_ENV_STAT(pTsdb->pTSmaEnv));
+
   tsdbInitTSmaFile(&tReadH, indexUid, querySKey);
   if (tsdbOpenDBF(SMA_ENV_ENV(pTsdb->pTSmaEnv), &tReadH.dFile) != 0) {
     tsdbWarn("vgId:%d open DBF %s failed since %s", REPO_ID(pTsdb), tReadH.dFile.path, tstrerror(terrno));
     return TSDB_CODE_FAILED;
   }
 
-  char  smaKey[SMA_KEY_LEN] = {0};
+  char    smaKey[SMA_KEY_LEN] = {0};
-  void *pSmaKey = &smaKey;
+  void   *pSmaKey = &smaKey;
-  tsdbEncodeTSmaKey(tableUid, colId, querySKey, (void **)&pSmaKey);
+  int64_t queryGroupId = 1;
+  tsdbEncodeTSmaKey(queryGroupId, querySKey, (void **)&pSmaKey);
 
-  tsdbDebug("vgId:%d get sma data from %s: smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64 ", keyLen %d", REPO_ID(pTsdb),
+  tsdbDebug("vgId:%d get sma data from %s: smaKey %" PRIx64 "-%" PRIx64 ", keyLen %d", REPO_ID(pTsdb),
-            tReadH.dFile.path, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
+            tReadH.dFile.path, *(int64_t *)smaKey, *(int64_t *)POINTER_SHIFT(smaKey, 8), SMA_KEY_LEN);
-            *(int64_t *)POINTER_SHIFT(smaKey, 10), SMA_KEY_LEN);
 
   void    *result = NULL;
   uint32_t valueSize = 0;
   if ((result = tsdbGetSmaDataByKey(&tReadH.dFile, smaKey, SMA_KEY_LEN, &valueSize)) == NULL) {
-    tsdbWarn("vgId:%d get sma data failed from smaIndex %" PRIi64 ", smaKey %" PRIx64 "-%" PRIu16 "-%" PRIx64
+    tsdbWarn("vgId:%d get sma data failed from smaIndex %" PRIi64 ", smaKey %" PRIx64 "-%" PRIx64 " since %s",
-             " since %s",
+             REPO_ID(pTsdb), indexUid, *(int64_t *)smaKey, *(int64_t *)POINTER_SHIFT(smaKey, 8), tstrerror(terrno));
-             REPO_ID(pTsdb), indexUid, *(tb_uid_t *)smaKey, *(uint16_t *)POINTER_SHIFT(smaKey, 8),
-             *(int64_t *)POINTER_SHIFT(smaKey, 10), tstrerror(terrno));
     tsdbCloseDBF(&tReadH.dFile);
     return TSDB_CODE_FAILED;
   }
@@ -1347,11 +1395,10 @@ int32_t tsdbRemoveTSmaData(STsdb *pTsdb, void *smaIndex, STimeWindow *pWin) {
 }
 #endif
 
-
 // TODO: Who is responsible for resource allocate and release?
-int32_t tsdbInsertTSmaData(STsdb *pTsdb, char *msg) {
+int32_t tsdbInsertTSmaData(STsdb *pTsdb, int64_t indexUid, const char *msg) {
   int32_t code = TSDB_CODE_SUCCESS;
-  if ((code = tsdbInsertTSmaDataImpl(pTsdb, msg)) < 0) {
+  if ((code = tsdbInsertTSmaDataImpl(pTsdb, indexUid, msg)) < 0) {
     tsdbWarn("vgId:%d insert tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
   }
   return code;
@@ -1373,18 +1420,14 @@ int32_t tsdbInsertRSmaData(STsdb *pTsdb, char *msg) {
   return code;
 }
 
-
+int32_t tsdbGetTSmaData(STsdb *pTsdb, char*pData, int64_t indexUid, TSKEY querySKey, int32_t nMaxResult) {
-int32_t tsdbGetTSmaData(STsdb *pTsdb, STSmaDataWrapper *pData, int64_t indexUid, int64_t interval, int8_t intervalUnit,
-                        tb_uid_t tableUid, col_id_t colId, TSKEY querySKey, int32_t nMaxResult) {
   int32_t code = TSDB_CODE_SUCCESS;
-  if ((code = tsdbGetTSmaDataImpl(pTsdb, pData, indexUid, interval, intervalUnit, tableUid, colId, querySKey,
+  if ((code = tsdbGetTSmaDataImpl(pTsdb, pData, indexUid, querySKey, nMaxResult)) < 0) {
-                                  nMaxResult)) < 0) {
     tsdbWarn("vgId:%d get tSma data failed since %s", REPO_ID(pTsdb), tstrerror(terrno));
   }
   return code;
 }
 
-
 int32_t tsdbDropTSmaData(STsdb *pTsdb, int64_t indexUid) {
   int32_t code = TSDB_CODE_SUCCESS;
   if ((code = tsdbDropTSmaDataImpl(pTsdb, indexUid)) < 0) {

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

@@ -41,7 +41,7 @@ int vnodeProcessQueryMsg(SVnode *pVnode, SRpcMsg *pMsg) {
   }
 }
 
-int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg) {
+int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo) {
   vTrace("message in fetch queue is processing");
   char   *msgstr = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
   int32_t msgLen = pMsg->contLen - sizeof(SMsgHead);
@@ -69,9 +69,9 @@ int vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg) {
       return tqProcessPollReq(pVnode->pTq, pMsg);
     case TDMT_VND_TASK_PIPE_EXEC:
     case TDMT_VND_TASK_MERGE_EXEC:
-      return tqProcessTaskExec(pVnode->pTq, msgstr, msgLen);
+      return tqProcessTaskExec(pVnode->pTq, msgstr, msgLen, pInfo->workerId);
     case TDMT_VND_STREAM_TRIGGER:
-      return tqProcessStreamTrigger(pVnode->pTq, pMsg->pCont, pMsg->contLen);
+      return tqProcessStreamTrigger(pVnode->pTq, pMsg->pCont, pMsg->contLen, pInfo->workerId);
     case TDMT_VND_QUERY_HEARTBEAT:
       return qWorkerProcessHbMsg(pVnode, pVnode->pQuery, pMsg);
     default:

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

@@ -17,7 +17,9 @@
 
 void smaHandleRes(void *pVnode, int64_t smaId, const SArray *data) {
   // TODO
+
   blockDebugShowData(data);
+  tsdbInsertTSmaData(((SVnode *)pVnode)->pTsdb, smaId, (const char *)data);
 }
 
 void vnodeProcessWMsgs(SVnode *pVnode, SArray *pMsgs) {
@@ -184,8 +186,8 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
       }
     } break;
     case TDMT_VND_TASK_WRITE_EXEC: {
-      if (tqProcessTaskExec(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)),
+      if (tqProcessTaskExec(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), pMsg->contLen - sizeof(SMsgHead),
-                            pMsg->contLen - sizeof(SMsgHead)) < 0) {
+                            0) < 0) {
       }
     } break;
     case TDMT_VND_CREATE_SMA: {  // timeRangeSMA
@@ -194,14 +196,15 @@ int vnodeApplyWMsg(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
       SSmaCfg vCreateSmaReq = {0};
       if (tDeserializeSVCreateTSmaReq(POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), &vCreateSmaReq) == NULL) {
         terrno = TSDB_CODE_OUT_OF_MEMORY;
-        vWarn("vgId%d: TDMT_VND_CREATE_SMA received but deserialize failed since %s", pVnode->config.vgId, terrstr(terrno));
+        vWarn("vgId%d: TDMT_VND_CREATE_SMA received but deserialize failed since %s", pVnode->config.vgId,
+              terrstr(terrno));
         return -1;
       }
       vWarn("vgId%d: TDMT_VND_CREATE_SMA received for %s:%" PRIi64, pVnode->config.vgId, vCreateSmaReq.tSma.indexName,
             vCreateSmaReq.tSma.indexUid);
 
       // record current timezone of server side
-      tstrncpy(vCreateSmaReq.tSma.timezone, tsTimezoneStr, TD_TIMEZONE_LEN);
+      vCreateSmaReq.tSma.timezoneInt = tsTimezone;
 
       if (metaCreateTSma(pVnode->pMeta, &vCreateSmaReq) < 0) {
         // TODO: handle error

source/dnode/vnode/test/tsdbSmaTest.cpp

@@ -15,6 +15,7 @@
 
 #include <gtest/gtest.h>
 #include <tsdbDef.h>
+
 #include <taoserror.h>
 #include <tglobal.h>
 #include <iostream>
@@ -58,20 +59,21 @@ TEST(testCase, unionEncodeDecodeTest) {
 
   void   *buf = taosMemoryMalloc(1024);
   void   *pBuf = buf;
+  void   *qBuf = buf;
   int32_t tlen = 0;
-  tlen += taosEncodeFixedU8(&buf, sut.info);
+  tlen += taosEncodeFixedU8(&pBuf, sut.info);
-  tlen += taosEncodeFixedI16(&buf, sut.nBSmaCols);
+  tlen += taosEncodeFixedI16(&pBuf, sut.nBSmaCols);
   for (col_id_t i = 0; i < sut.nBSmaCols; ++i) {
-    tlen += taosEncodeFixedI16(&buf, sut.pBSmaCols[i]);
+    tlen += taosEncodeFixedI16(&pBuf, sut.pBSmaCols[i]);
   }
 
   SUnionTest dut = {0};
-  pBuf = taosDecodeFixedU8(pBuf, &dut.info);
+  qBuf = taosDecodeFixedU8(qBuf, &dut.info);
-  pBuf = taosDecodeFixedI16(pBuf, &dut.nBSmaCols);
+  qBuf = taosDecodeFixedI16(qBuf, &dut.nBSmaCols);
   if (dut.nBSmaCols > 0) {
     dut.pBSmaCols = (col_id_t *)taosMemoryMalloc(dut.nBSmaCols * sizeof(col_id_t));
     for (col_id_t i = 0; i < dut.nBSmaCols; ++i) {
-      pBuf = taosDecodeFixedI16(pBuf, dut.pBSmaCols + i);
+      qBuf = taosDecodeFixedI16(qBuf, dut.pBSmaCols + i);
     }
   } else {
     dut.pBSmaCols = NULL;
@@ -80,13 +82,17 @@ TEST(testCase, unionEncodeDecodeTest) {
   printf("sut.rollup=%" PRIu8 ", type=%" PRIu8 ", info=%" PRIu8 "\n", sut.rollup, sut.type, sut.info);
   printf("dut.rollup=%" PRIu8 ", type=%" PRIu8 ", info=%" PRIu8 "\n", dut.rollup, dut.type, dut.info);
 
-  ASSERT_EQ(sut.rollup, dut.rollup);
+  EXPECT_EQ(sut.rollup, dut.rollup);
-  ASSERT_EQ(sut.type, dut.type);
+  EXPECT_EQ(sut.type, dut.type);
-  ASSERT_EQ(sut.nBSmaCols, dut.nBSmaCols);
+  EXPECT_EQ(sut.nBSmaCols, dut.nBSmaCols);
   for (col_id_t i = 0; i < sut.nBSmaCols; ++i) {
-    ASSERT_EQ(*(col_id_t *)(sut.pBSmaCols + i), sut.pBSmaCols[i]);
+    EXPECT_EQ(*(col_id_t *)(sut.pBSmaCols + i), sut.pBSmaCols[i]);
-    ASSERT_EQ(*(col_id_t *)(sut.pBSmaCols + i), dut.pBSmaCols[i]);
+    EXPECT_EQ(*(col_id_t *)(sut.pBSmaCols + i), dut.pBSmaCols[i]);
   }
+
+  taosMemoryFreeClear(buf);
+  taosMemoryFreeClear(dut.pBSmaCols);
+  taosMemoryFreeClear(sut.pBSmaCols);
 }
 #if 1
 TEST(testCase, tSma_Meta_Encode_Decode_Test) {
@@ -106,37 +112,37 @@ TEST(testCase, tSma_Meta_Encode_Decode_Test) {
   uint32_t     bufLen = tEncodeTSmaWrapper(NULL, &tSmaWrapper);
 
   void *buf = taosMemoryCalloc(1, bufLen);
-  ASSERT_NE(buf, nullptr);
+  EXPECT_NE(buf, nullptr);
 
   STSmaWrapper *pSW = (STSmaWrapper *)buf;
   uint32_t      len = tEncodeTSmaWrapper(&buf, &tSmaWrapper);
 
-  ASSERT_EQ(len, bufLen);
+  EXPECT_EQ(len, bufLen);
 
   // decode
   STSmaWrapper dstTSmaWrapper = {0};
   void        *result = tDecodeTSmaWrapper(pSW, &dstTSmaWrapper);
-  ASSERT_NE(result, nullptr);
+  EXPECT_NE(result, nullptr);
 
-  ASSERT_EQ(tSmaWrapper.number, dstTSmaWrapper.number);
+  EXPECT_EQ(tSmaWrapper.number, dstTSmaWrapper.number);
 
   for (int i = 0; i < tSmaWrapper.number; ++i) {
     STSma *pSma = tSmaWrapper.tSma + i;
     STSma *qSma = dstTSmaWrapper.tSma + i;
 
-    ASSERT_EQ(pSma->version, qSma->version);
+    EXPECT_EQ(pSma->version, qSma->version);
-    ASSERT_EQ(pSma->intervalUnit, qSma->intervalUnit);
+    EXPECT_EQ(pSma->intervalUnit, qSma->intervalUnit);
-    ASSERT_EQ(pSma->slidingUnit, qSma->slidingUnit);
+    EXPECT_EQ(pSma->slidingUnit, qSma->slidingUnit);
-    ASSERT_STRCASEEQ(pSma->indexName, qSma->indexName);
+    EXPECT_STRCASEEQ(pSma->indexName, qSma->indexName);
-    ASSERT_EQ(pSma->timezoneInt, qSma->timezoneInt);
+    EXPECT_EQ(pSma->timezoneInt, qSma->timezoneInt);
-    ASSERT_EQ(pSma->indexUid, qSma->indexUid);
+    EXPECT_EQ(pSma->indexUid, qSma->indexUid);
-    ASSERT_EQ(pSma->tableUid, qSma->tableUid);
+    EXPECT_EQ(pSma->tableUid, qSma->tableUid);
-    ASSERT_EQ(pSma->interval, qSma->interval);
+    EXPECT_EQ(pSma->interval, qSma->interval);
-    ASSERT_EQ(pSma->sliding, qSma->sliding);
+    EXPECT_EQ(pSma->sliding, qSma->sliding);
-    ASSERT_EQ(pSma->exprLen, qSma->exprLen);
+    EXPECT_EQ(pSma->exprLen, qSma->exprLen);
-    ASSERT_STRCASEEQ(pSma->expr, qSma->expr);
+    EXPECT_STRCASEEQ(pSma->expr, qSma->expr);
-    ASSERT_EQ(pSma->tagsFilterLen, qSma->tagsFilterLen);
+    EXPECT_EQ(pSma->tagsFilterLen, qSma->tagsFilterLen);
-    ASSERT_STRCASEEQ(pSma->tagsFilter, qSma->tagsFilter);
+    EXPECT_STRCASEEQ(pSma->tagsFilter, qSma->tagsFilter);
   }
 
   // resource release
@@ -172,12 +178,12 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
 
   tSma.exprLen = strlen(expr);
   tSma.expr = (char *)taosMemoryCalloc(1, tSma.exprLen + 1);
-  ASSERT_NE(tSma.expr, nullptr);
+  EXPECT_NE(tSma.expr, nullptr);
   tstrncpy(tSma.expr, expr, tSma.exprLen + 1);
 
   tSma.tagsFilterLen = strlen(tagsFilter);
   tSma.tagsFilter = (char *)taosMemoryCalloc(tSma.tagsFilterLen + 1, 1);
-  ASSERT_NE(tSma.tagsFilter, nullptr);
+  EXPECT_NE(tSma.tagsFilter, nullptr);
   tstrncpy(tSma.tagsFilter, tagsFilter, tSma.tagsFilterLen + 1);
 
   SMeta          *pMeta = NULL;
@@ -189,7 +195,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   pMeta = metaOpen(smaTestDir, pMetaCfg, NULL);
   assert(pMeta != NULL);
   // save index 1
-  ASSERT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
+  EXPECT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
 
   pSmaCfg->indexUid = indexUid2;
   tstrncpy(pSmaCfg->indexName, smaIndexName2, TSDB_INDEX_NAME_LEN);
@@ -200,7 +206,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   pSmaCfg->sliding = 5;
 
   // save index 2
-  ASSERT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
+  EXPECT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
 
   // get value by indexName
   STSma *qSmaCfg = NULL;
@@ -210,8 +216,8 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   printf("timezone1 = %" PRIi8 "\n", qSmaCfg->timezoneInt);
   printf("expr1 = %s\n", qSmaCfg->expr != NULL ? qSmaCfg->expr : "");
   printf("tagsFilter1 = %s\n", qSmaCfg->tagsFilter != NULL ? qSmaCfg->tagsFilter : "");
-  ASSERT_STRCASEEQ(qSmaCfg->indexName, smaIndexName1);
+  EXPECT_STRCASEEQ(qSmaCfg->indexName, smaIndexName1);
-  ASSERT_EQ(qSmaCfg->tableUid, tSma.tableUid);
+  EXPECT_EQ(qSmaCfg->tableUid, tSma.tableUid);
   tdDestroyTSma(qSmaCfg);
   taosMemoryFreeClear(qSmaCfg);
 
@@ -221,8 +227,8 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
   printf("timezone2 = %" PRIi8 "\n", qSmaCfg->timezoneInt);
   printf("expr2 = %s\n", qSmaCfg->expr != NULL ? qSmaCfg->expr : "");
   printf("tagsFilter2 = %s\n", qSmaCfg->tagsFilter != NULL ? qSmaCfg->tagsFilter : "");
-  ASSERT_STRCASEEQ(qSmaCfg->indexName, smaIndexName2);
+  EXPECT_STRCASEEQ(qSmaCfg->indexName, smaIndexName2);
-  ASSERT_EQ(qSmaCfg->interval, tSma.interval);
+  EXPECT_EQ(qSmaCfg->interval, tSma.interval);
   tdDestroyTSma(qSmaCfg);
   taosMemoryFreeClear(qSmaCfg);
 
@@ -238,25 +244,25 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
     printf("indexName = %s\n", indexName);
     ++indexCnt;
   }
-  ASSERT_EQ(indexCnt, nCntTSma);
+  EXPECT_EQ(indexCnt, nCntTSma);
   metaCloseSmaCurosr(pSmaCur);
 
   // get wrapper by table uid
   STSmaWrapper *pSW = metaGetSmaInfoByTable(pMeta, tbUid);
   assert(pSW != NULL);
-  ASSERT_EQ(pSW->number, nCntTSma);
+  EXPECT_EQ(pSW->number, nCntTSma);
-  ASSERT_STRCASEEQ(pSW->tSma->indexName, smaIndexName1);
+  EXPECT_STRCASEEQ(pSW->tSma->indexName, smaIndexName1);
-  ASSERT_EQ(pSW->tSma->timezoneInt, timezone);
+  EXPECT_EQ(pSW->tSma->timezoneInt, timezone);
-  ASSERT_STRCASEEQ(pSW->tSma->expr, expr);
+  EXPECT_STRCASEEQ(pSW->tSma->expr, expr);
-  ASSERT_STRCASEEQ(pSW->tSma->tagsFilter, tagsFilter);
+  EXPECT_STRCASEEQ(pSW->tSma->tagsFilter, tagsFilter);
-  ASSERT_EQ(pSW->tSma->indexUid, indexUid1);
+  EXPECT_EQ(pSW->tSma->indexUid, indexUid1);
-  ASSERT_EQ(pSW->tSma->tableUid, tbUid);
+  EXPECT_EQ(pSW->tSma->tableUid, tbUid);
-  ASSERT_STRCASEEQ((pSW->tSma + 1)->indexName, smaIndexName2);
+  EXPECT_STRCASEEQ((pSW->tSma + 1)->indexName, smaIndexName2);
-  ASSERT_EQ((pSW->tSma + 1)->timezoneInt, timezone);
+  EXPECT_EQ((pSW->tSma + 1)->timezoneInt, timezone);
-  ASSERT_STRCASEEQ((pSW->tSma + 1)->expr, expr);
+  EXPECT_STRCASEEQ((pSW->tSma + 1)->expr, expr);
-  ASSERT_STRCASEEQ((pSW->tSma + 1)->tagsFilter, tagsFilter);
+  EXPECT_STRCASEEQ((pSW->tSma + 1)->tagsFilter, tagsFilter);
-  ASSERT_EQ((pSW->tSma + 1)->indexUid, indexUid2);
+  EXPECT_EQ((pSW->tSma + 1)->indexUid, indexUid2);
-  ASSERT_EQ((pSW->tSma + 1)->tableUid, tbUid);
+  EXPECT_EQ((pSW->tSma + 1)->tableUid, tbUid);
 
   tdDestroyTSmaWrapper(pSW);
   taosMemoryFreeClear(pSW);
@@ -268,7 +274,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
     printf("metaGetSmaTbUids: uid[%" PRIu32 "] = %" PRIi64 "\n", i, *(tb_uid_t *)taosArrayGet(pUids, i));
     // printf("metaGetSmaTbUids: index[%" PRIu32 "] = %s", i, (char *)taosArrayGet(pUids, i));
   }
-  ASSERT_EQ(taosArrayGetSize(pUids), 1);
+  EXPECT_EQ(taosArrayGetSize(pUids), 1);
   taosArrayDestroy(pUids);
 
   // resource release
@@ -280,7 +286,7 @@ TEST(testCase, tSma_metaDB_Put_Get_Del_Test) {
 }
 #endif
 
-#if 0
+#if 1
 TEST(testCase, tSma_Data_Insert_Query_Test) {
   // step 1: prepare meta
   const char    *smaIndexName1 = "sma_index_test_1";
@@ -299,9 +305,9 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   // encode
   STSma tSma = {0};
   tSma.version = 0;
-  tSma.intervalUnit = TIME_UNIT_DAY;
+  tSma.intervalUnit = TIME_UNIT_MINUTE;
   tSma.interval = 1;
-  tSma.slidingUnit = TIME_UNIT_HOUR;
+  tSma.slidingUnit = TIME_UNIT_MINUTE;
   tSma.sliding = 1;  // sliding = interval when it's convert window
   tSma.indexUid = indexUid1;
   tstrncpy(tSma.indexName, smaIndexName1, TSDB_INDEX_NAME_LEN);
@@ -310,12 +316,12 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
 
   tSma.exprLen = strlen(expr);
   tSma.expr = (char *)taosMemoryCalloc(1, tSma.exprLen + 1);
-  ASSERT_NE(tSma.expr, nullptr);
+  EXPECT_NE(tSma.expr, nullptr);
   tstrncpy(tSma.expr, expr, tSma.exprLen + 1);
 
   tSma.tagsFilterLen = strlen(tagsFilter);
   tSma.tagsFilter = (char *)taosMemoryCalloc(1, tSma.tagsFilterLen + 1);
-  ASSERT_NE(tSma.tagsFilter, nullptr);
+  EXPECT_NE(tSma.tagsFilter, nullptr);
   tstrncpy(tSma.tagsFilter, tagsFilter, tSma.tagsFilterLen + 1);
 
   SMeta          *pMeta = NULL;
@@ -327,11 +333,11 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   pMeta = metaOpen(smaTestDir, pMetaCfg, NULL);
   assert(pMeta != NULL);
   // save index 1
-  ASSERT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
+  EXPECT_EQ(metaSaveSmaToDB(pMeta, pSmaCfg), 0);
 
   // step 2: insert data
-  STsdb            *pTsdb = (STsdb *)taosMemoryCalloc(1, sizeof(STsdb));
+  STsdb    *pTsdb = (STsdb *)taosMemoryCalloc(1, sizeof(STsdb));
-  STsdbCfg         *pCfg = &pTsdb->config;
+  STsdbCfg *pCfg = &pTsdb->config;
 
   pTsdb->pMeta = pMeta;
   pTsdb->vgId = 2;
@@ -364,7 +370,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   strncpy(pDisks.dir, "/var/lib/taos", TSDB_FILENAME_LEN);
   int32_t numOfDisks = 1;
   pTsdb->pTfs = tfsOpen(&pDisks, numOfDisks);
-  ASSERT_NE(pTsdb->pTfs, nullptr);
+  EXPECT_NE(pTsdb->pTfs, nullptr);
 
   // generate SSubmitReq msg and update expired window
   int16_t  schemaVer = 0;
@@ -374,7 +380,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   uint32_t msgLen = sizeof(SSubmitReq) + mockBlkNum * sizeof(SSubmitBlk) + mockBlkNum * mockRowNum * mockRowLen;
 
   SSubmitReq *pMsg = (SSubmitReq *)taosMemoryCalloc(1, msgLen);
-  ASSERT_NE(pMsg, nullptr);
+  EXPECT_NE(pMsg, nullptr);
   pMsg->version = htobe64(schemaVer);
   pMsg->numOfBlocks = htonl(mockBlkNum);
   pMsg->length = htonl(msgLen);
@@ -400,20 +406,99 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
     }
   }
 
-  ASSERT_EQ(tdScanAndConvertSubmitMsg(pMsg), TSDB_CODE_SUCCESS);
+  EXPECT_EQ(tdScanAndConvertSubmitMsg(pMsg), TSDB_CODE_SUCCESS);
 
-  ASSERT_EQ(tsdbUpdateSmaWindow(pTsdb, (const char *)pMsg), 0);
+  EXPECT_EQ(tsdbUpdateSmaWindow(pTsdb, (const char *)pMsg), 0);
 
   // init
-  int32_t allocCnt = 0;
+  const int32_t  tSmaGroupSize = 4;
-  int32_t allocStep = 16384;
+  const int32_t  tSmaNumOfTags = 2;
-  int32_t buffer = 1024;
+  const int64_t  tSmaGroupId = 12345670;
-  void   *buf = NULL;
+  const col_id_t tSmaNumOfCols = 9;  // binary/nchar/varbinary/varchar are only used for tags for group by conditions.
-  ASSERT_EQ(tsdbMakeRoom(&buf, allocStep), 0);
+  const int32_t  tSmaNumOfRows = 2;
-  int32_t  bufSize = taosTSizeof(buf);
+
-  int32_t  numOfTables = 10;
+  SArray *pDataBlocks = taosArrayInit(tSmaGroupSize, sizeof(SSDataBlock *));
-  col_id_t numOfCols = 4096;
+  EXPECT_NE(pDataBlocks, nullptr);
-  ASSERT_GT(numOfCols, 0);
+  int32_t tSmaTypeArray[tSmaNumOfCols] = {TSDB_DATA_TYPE_TIMESTAMP, TSDB_DATA_TYPE_BOOL,     TSDB_DATA_TYPE_INT,
+                                          TSDB_DATA_TYPE_UBIGINT,   TSDB_DATA_TYPE_SMALLINT, TSDB_DATA_TYPE_FLOAT,
+                                          TSDB_DATA_TYPE_DOUBLE,    TSDB_DATA_TYPE_VARCHAR,  TSDB_DATA_TYPE_NCHAR};
+  // last 2 columns for group by tags
+  // int32_t tSmaTypeArray[tSmaNumOfCols] = {TSDB_DATA_TYPE_TIMESTAMP, TSDB_DATA_TYPE_BOOL};
+  const char *tSmaGroupbyTags[tSmaGroupSize * tSmaNumOfTags] = {"BeiJing",  "HaiDian", "BeiJing",   "ChaoYang",
+                                                                "ShangHai", "PuDong",  "ShangHai", "MinHang"};
+  TSKEY       tSmaSKeyMs = (int64_t)1648535332 * 1000;
+  int64_t     tSmaIntervalMs = tSma.interval * 60 * 1000;
+  int64_t     tSmaInitVal = 0;
+
+  for (int32_t g = 0; g < tSmaGroupSize; ++g) {
+    SSDataBlock *pDataBlock = (SSDataBlock *)taosMemoryCalloc(1, sizeof(SSDataBlock));
+    EXPECT_NE(pDataBlock, nullptr);
+    pDataBlock->pBlockAgg = NULL;
+    pDataBlock->info.numOfCols = tSmaNumOfCols;
+    pDataBlock->info.rows = tSmaNumOfRows;
+    pDataBlock->info.groupId = tSmaGroupId + g;
+
+    pDataBlock->pDataBlock = taosArrayInit(tSmaNumOfCols, sizeof(SColumnInfoData *));
+    EXPECT_NE(pDataBlock->pDataBlock, nullptr);
+    for (int32_t c = 0; c < tSmaNumOfCols; ++c) {
+      
+      SColumnInfoData *pColInfoData = (SColumnInfoData *)taosMemoryCalloc(1, sizeof(SColumnInfoData));
+      EXPECT_NE(pColInfoData, nullptr);
+
+      pColInfoData->info.type = tSmaTypeArray[c];
+      if (IS_VAR_DATA_TYPE(pColInfoData->info.type)) {
+        pColInfoData->info.bytes = 100;  // update accordingly
+      } else {
+        pColInfoData->info.bytes = TYPE_BYTES[pColInfoData->info.type];
+      }
+      pColInfoData->pData = (char *)taosMemoryCalloc(1, tSmaNumOfRows * pColInfoData->info.bytes);
+
+      for (int32_t r = 0; r < tSmaNumOfRows; ++r) {
+        void *pCellData = pColInfoData->pData + r * pColInfoData->info.bytes;
+        switch (pColInfoData->info.type) {
+          case TSDB_DATA_TYPE_TIMESTAMP:
+            *(TSKEY *)pCellData = tSmaSKeyMs + tSmaIntervalMs * r;
+            break;
+          case TSDB_DATA_TYPE_BOOL:
+            *(bool *)pCellData = (bool)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_INT:
+            *(int *)pCellData = (int)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_UBIGINT:
+            *(uint64_t *)pCellData = (uint64_t)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_SMALLINT:
+            *(int16_t *)pCellData = (int16_t)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_FLOAT:
+            *(float *)pCellData = (float)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_DOUBLE:
+            *(double *)pCellData = (double)tSmaInitVal++;
+            break;
+          case TSDB_DATA_TYPE_VARCHAR:  // city
+            varDataSetLen(pCellData, strlen(tSmaGroupbyTags[g * 2]));
+            memcpy(varDataVal(pCellData), tSmaGroupbyTags[g * 2], varDataLen(pCellData));
+            break;
+          case TSDB_DATA_TYPE_NCHAR:  // district
+            varDataSetLen(pCellData, strlen(tSmaGroupbyTags[g * 2 + 1]));
+            memcpy(varDataVal(pCellData), tSmaGroupbyTags[g * 2 + 1], varDataLen(pCellData));
+            break;
+          default:
+            EXPECT_EQ(0, 1);  // add definition
+            break;
+        }
+      }
+      // push SColumnInfoData
+      taosArrayPush(pDataBlock->pDataBlock, &pColInfoData);
+    }
+    // push SSDataBlock
+    taosArrayPush(pDataBlocks, &pDataBlock);
+  }
+
+  // execute
+  EXPECT_EQ(tsdbInsertTSmaData(pTsdb, tSma.indexUid, (const char *)pDataBlocks), TSDB_CODE_SUCCESS);
 
 #if 0
   STSmaDataWrapper *pSmaData = NULL;
@@ -432,7 +517,7 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
     int32_t tableDataLen = sizeof(STSmaTbData);
     for (col_id_t c = 0; c < numOfCols; ++c) {
       if (bufSize - len - tableDataLen < buffer) {
-        ASSERT_EQ(tsdbMakeRoom(&buf, bufSize + allocStep), 0);
+        EXPECT_EQ(tsdbMakeRoom(&buf, bufSize + allocStep), 0);
         pSmaData = (STSmaDataWrapper *)buf;
         pTbData = (STSmaTbData *)POINTER_SHIFT(pSmaData, len);
         bufSize = taosTSizeof(buf);
@@ -459,31 +544,36 @@ TEST(testCase, tSma_Data_Insert_Query_Test) {
   }
   pSmaData->dataLen = (len - sizeof(STSmaDataWrapper));
 
-  ASSERT_GE(bufSize, pSmaData->dataLen);
+  EXPECT_GE(bufSize, pSmaData->dataLen);
   // execute
-  ASSERT_EQ(tsdbInsertTSmaData(pTsdb, (char *)pSmaData), TSDB_CODE_SUCCESS);
+  EXPECT_EQ(tsdbInsertTSmaData(pTsdb, (char *)pSmaData), TSDB_CODE_SUCCESS);
 #endif
 
-  SSDataBlock *pSmaData = (SSDataBlock *)taosMemoryCalloc(1, sizeof(SSDataBlock));
-  
-
-
   // step 3: query
   uint32_t checkDataCnt = 0;
-  for (int32_t t = 0; t < numOfTables; ++t) {
+  EXPECT_EQ(tsdbGetTSmaData(pTsdb, NULL, indexUid1, skey1, 1), TSDB_CODE_SUCCESS);
-    for (col_id_t c = 0; c < numOfCols; ++c) {
+  ++checkDataCnt;
-      ASSERT_EQ(tsdbGetTSmaData(pTsdb, NULL, indexUid1, interval1, intervalUnit1, tbUid + t,
-                                c + PRIMARYKEY_TIMESTAMP_COL_ID, skey1, 1),
-                TSDB_CODE_SUCCESS);
-      ++checkDataCnt;
-    }
-  }
 
   printf("%s:%d The sma data check count for insert and query is %" PRIu32 "\n", __FILE__, __LINE__, checkDataCnt);
 
   // release data
   taosMemoryFreeClear(pMsg);
-  taosTZfree(buf);
+
+  for (int32_t i = 0; i < taosArrayGetSize(pDataBlocks); ++i) {
+    SSDataBlock *pDataBlock = *(SSDataBlock **)taosArrayGet(pDataBlocks, i);
+    int32_t      numOfOutput = taosArrayGetSize(pDataBlock->pDataBlock);
+    for (int32_t j = 0; j < numOfOutput; ++j) {
+      SColumnInfoData *pColInfoData = *(SColumnInfoData **)taosArrayGet(pDataBlock->pDataBlock, j);
+      colDataDestroy(pColInfoData);
+      taosMemoryFreeClear(pColInfoData);
+    }
+
+    taosArrayDestroy(pDataBlock->pDataBlock);
+    taosMemoryFreeClear(pDataBlock->pBlockAgg);
+    taosMemoryFreeClear(pDataBlock);
+  }
+  taosArrayDestroy(pDataBlocks);
+
   // release meta
   tdDestroyTSma(&tSma);
   tfsClose(pTsdb->pTfs);

source/libs/executor/inc/executorimpl.h

@@ -161,20 +161,8 @@ typedef struct STaskCostInfo {
 typedef struct SOperatorCostInfo {
   uint64_t openCost;
   uint64_t execCost;
-//  uint64_t totalRows;
-//  uint64_t totalBytes;
 } SOperatorCostInfo;
 
-typedef struct {
-  int64_t vgroupLimit;
-  int64_t ts;
-} SOrderedPrjQueryInfo;
-
-typedef struct {
-  char*   tags;
-  SArray* pResult;  // SArray<SStddevInterResult>
-} SInterResult;
-
 // The basic query information extracted from the SQueryInfo tree to support the
 // execution of query in a data node.
 typedef struct STaskAttr {
@@ -230,7 +218,6 @@ typedef struct STaskAttr {
   SColumnInfo*         tagColList;
   int32_t              numOfFilterCols;
   int64_t*             fillVal;
-  SOrderedPrjQueryInfo prjInfo;  // limit value for each vgroup, only available in global order projection query.
 
   SSingleColumnFilterInfo* pFilterInfo;
   //  SFilterInfo     *pFilters;
@@ -245,8 +232,9 @@ struct SOperatorInfo;
 
 typedef void (*__optr_encode_fn_t)(struct SOperatorInfo* pOperator, char **result, int32_t *length);
 typedef bool (*__optr_decode_fn_t)(struct SOperatorInfo* pOperator, char *result, int32_t length);
-typedef int32_t (*__optr_open_fn_t)(struct SOperatorInfo* param);
+
-typedef SSDataBlock* (*__optr_fn_t)(struct SOperatorInfo* param, bool* newgroup);
+typedef int32_t (*__optr_open_fn_t)(struct SOperatorInfo* pOptr);
+typedef SSDataBlock* (*__optr_fn_t)(struct SOperatorInfo* pOptr, bool* newgroup);
 typedef void (*__optr_close_fn_t)(void* param, int32_t num);
 
 typedef struct STaskIdInfo {
@@ -267,7 +255,8 @@ typedef struct SExecTaskInfo {
   uint64_t        totalRows;            // total number of rows
   STableGroupInfo tableqinfoGroupInfo;  // this is a group array list, including SArray<STableQueryInfo*> structure
   char*           sql;                  // query sql string
-  jmp_buf         env;                  //
+  jmp_buf         env;                  // jump to this position when error happens.
+  EOPTR_EXEC_MODEL execModel;            // operator execution model [batch model|stream model]
   struct SOperatorInfo* pRoot;
 } SExecTaskInfo;
 
@@ -330,11 +319,12 @@ typedef struct SOperatorInfo {
   SResultInfo            resultInfo;
   struct SOperatorInfo** pDownstream;      // downstram pointer list
   int32_t                numOfDownstream;  // number of downstream. The value is always ONE expect for join operator
-  __optr_fn_t            getNextFn;
-  __optr_fn_t            cleanupFn;
-  __optr_close_fn_t      closeFn;
   __optr_open_fn_t       _openFn;          // DO NOT invoke this function directly
-  __optr_encode_fn_t     encodeResultRow;           //
+  __optr_fn_t            getNextFn;
+  __optr_fn_t            getStreamResFn;   // execute the aggregate in the stream model.
+  __optr_fn_t            cleanupFn;        // call this function to release the allocated resources ASAP
+  __optr_close_fn_t      closeFn;
+  __optr_encode_fn_t     encodeResultRow;
   __optr_decode_fn_t     decodeResultRow;
 } SOperatorInfo;
 
@@ -363,18 +353,18 @@ typedef struct SQInfo {
   STaskCostInfo   summary;
 } SQInfo;
 
-enum {
+typedef enum {
-  DATA_NOT_READY = 0x1,
+  EX_SOURCE_DATA_NOT_READY = 0x1,
-  DATA_READY     = 0x2,
+  EX_SOURCE_DATA_READY     = 0x2,
-  DATA_EXHAUSTED = 0x3,
+  EX_SOURCE_DATA_EXHAUSTED = 0x3,
-};
+} EX_SOURCE_STATUS;
 
 typedef struct SSourceDataInfo {
   struct SExchangeInfo *pEx;
-  int32_t            index;
+  int32_t               index;
-  SRetrieveTableRsp *pRsp;
+  SRetrieveTableRsp    *pRsp;
-  uint64_t           totalRows;
+  uint64_t              totalRows;
-  int32_t            status;
+  EX_SOURCE_STATUS      status;
 } SSourceDataInfo;
 
 typedef struct SLoadRemoteDataInfo {
@@ -383,12 +373,6 @@ typedef struct SLoadRemoteDataInfo {
   uint64_t           totalElapsed;  // total elapsed time
 } SLoadRemoteDataInfo;
 
-enum {
-  EX_SOURCE_DATA_NOT_READY = 0x1,
-  EX_SOURCE_DATA_READY     = 0x2,
-  EX_SOURCE_DATA_EXHAUSTED = 0x3,
-};
-
 typedef struct SExchangeInfo {
   SArray*            pSources;
   SArray*            pSourceDataInfo;
@@ -484,16 +468,18 @@ typedef struct SAggSupporter {
 } SAggSupporter;
 
 typedef struct STableIntervalOperatorInfo {
-  SOptrBasicInfo     binfo;
+  SOptrBasicInfo     binfo;                // basic info
-  SGroupResInfo      groupResInfo;
+  SGroupResInfo      groupResInfo;         // multiple results build supporter
-  SInterval          interval;
+  SInterval          interval;             // interval info
-  STimeWindow        win;
+  STimeWindow        win;                  // query time range
-  int32_t            precision;
+  bool               timeWindowInterpo;    // interpolation needed or not
-  bool               timeWindowInterpo;
+  char             **pRow;                 // previous row/tuple of already processed datablock
-  char             **pRow;
+  SAggSupporter      aggSup;               // aggregate supporter
-  SAggSupporter      aggSup;
+  STableQueryInfo   *pCurrent;             // current tableQueryInfo struct
-  STableQueryInfo   *pCurrent;
+  int32_t            order;                // current SSDataBlock scan order
-  int32_t            order;
+  EOPTR_EXEC_MODEL    execModel;            // operator execution model [batch model|stream model]
+  SArray            *pUpdatedWindow;       // updated time window due to the input data block from the downstream operator.
+  SColumnInfoData    timeWindowData;       // query time window info for scalar function execution.
 } STableIntervalOperatorInfo;
 
 typedef struct SAggOperatorInfo {
@@ -695,12 +681,6 @@ SOperatorInfo* createSLimitOperatorInfo(STaskRuntimeEnv* pRuntimeEnv, SOperatorI
 SOperatorInfo* createJoinOperatorInfo(SOperatorInfo** pdownstream, int32_t numOfDownstream, SSchema* pSchema,
                                       int32_t numOfOutput);
 
-void doSetFilterColumnInfo(SSingleColumnFilterInfo* pFilterInfo, int32_t numOfFilterCols, SSDataBlock* pBlock);
-bool doFilterDataBlock(SSingleColumnFilterInfo* pFilterInfo, int32_t numOfFilterCols, int32_t numOfRows, int8_t* p);
-void doCompactSDataBlock(SSDataBlock* pBlock, int32_t numOfRows, int8_t* p);
-
-SSDataBlock* createOutputBuf(SExprInfo* pExpr, int32_t numOfOutput, int32_t numOfRows);
-
 void* doDestroyFilterInfo(SSingleColumnFilterInfo* pFilterInfo, int32_t numOfFilterCols);
 
 void setInputDataBlock(SOperatorInfo* pOperator, SqlFunctionCtx* pCtx, SSDataBlock* pBlock, int32_t order);
@@ -734,7 +714,7 @@ int32_t getMaximumIdleDurationSec();
 
 void    doInvokeUdf(struct SUdfInfo* pUdfInfo, SqlFunctionCtx* pCtx, int32_t idx, int32_t type);
 void    setTaskStatus(SExecTaskInfo* pTaskInfo, int8_t status);
-int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId);
+int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId, EOPTR_EXEC_MODEL model);
 
 #ifdef __cplusplus
 }

source/libs/executor/src/executor.c

@@ -113,7 +113,7 @@ qTaskInfo_t qCreateStreamExecTaskInfo(void* msg, void* streamReadHandle) {
   }
 
   qTaskInfo_t pTaskInfo = NULL;
-  code = qCreateExecTask(streamReadHandle, 0, 0, plan, &pTaskInfo, NULL);
+  code = qCreateExecTask(streamReadHandle, 0, 0, plan, &pTaskInfo, NULL, OPTR_EXEC_MODEL_STREAM);
   if (code != TSDB_CODE_SUCCESS) {
     // TODO: destroy SSubplan & pTaskInfo
     terrno = code;

source/libs/executor/src/executorMain.c

@@ -51,11 +51,12 @@ static void freeqinfoFn(void *qhandle) {
   qDestroyTask(*handle);
 }
 
-int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, SSubplan* pSubplan, qTaskInfo_t* pTaskInfo, DataSinkHandle* handle) {
+int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, SSubplan* pSubplan,
+    qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, EOPTR_EXEC_MODEL model) {
   assert(readHandle != NULL && pSubplan != NULL);
   SExecTaskInfo** pTask = (SExecTaskInfo**)pTaskInfo;
 
-  int32_t code = createExecTaskInfoImpl(pSubplan, pTask, readHandle, taskId);
+  int32_t code = createExecTaskInfoImpl(pSubplan, pTask, readHandle, taskId, model);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }

source/libs/executor/src/executorimpl.c

@@ -1014,8 +1014,35 @@ static int32_t getNumOfRowsInTimeWindow(SDataBlockInfo* pDataBlockInfo, TSKEY* p
   return num;
 }
 
-static void doApplyFunctions(SqlFunctionCtx* pCtx, STimeWindow* pWin, int32_t offset, int32_t forwardStep, TSKEY* tsCol,
+//  query_range_start, query_range_end, window_duration, window_start, window_end
+static void initExecTimeWindowInfo(SColumnInfoData* pColData, STimeWindow* pQueryWindow) {
+  pColData->info.type = TSDB_DATA_TYPE_TIMESTAMP;
+  pColData->info.bytes = sizeof(int64_t);
+
+  blockDataEnsureColumnCapacity(pColData, 5);
+  colDataAppendInt64(pColData, 0, &pQueryWindow->skey);
+  colDataAppendInt64(pColData, 1, &pQueryWindow->ekey);
+
+  int64_t interval = 0;
+  colDataAppendInt64(pColData, 2, &interval);  // this value may be variable in case of 'n' and 'y'.
+  colDataAppendInt64(pColData, 3, &pQueryWindow->skey);
+  colDataAppendInt64(pColData, 4, &pQueryWindow->ekey);
+}
+
+static void updateTimeWindowInfo(SColumnInfoData* pColData, STimeWindow* pWin) {
+  int64_t* ts = (int64_t*)pColData->pData;
+
+  int64_t duration = pWin->ekey - pWin->skey + 1;
+  ts[2] = duration;       // set the duration
+  ts[3] = pWin->skey;     // window start key
+  ts[4] = pWin->ekey + 1; // window end key
+}
+
+static void doApplyFunctions(SqlFunctionCtx* pCtx, STimeWindow* pWin, SColumnInfoData* pTimeWindowData, int32_t offset, int32_t forwardStep, TSKEY* tsCol,
                              int32_t numOfTotal, int32_t numOfOutput, int32_t order) {
+  SScalarParam intervalParam = {.numOfRows = 5, .columnData = pTimeWindowData};  //TODO move out of this function
+  updateTimeWindowInfo(pTimeWindowData, pWin);
+
   for (int32_t k = 0; k < numOfOutput; ++k) {
     pCtx[k].startTs = pWin->skey;
 
@@ -1038,6 +1065,21 @@ static void doApplyFunctions(SqlFunctionCtx* pCtx, STimeWindow* pWin, int32_t of
       pCtx[k].isAggSet = false;
     }
 
+    if (fmIsWindowPseudoColumnFunc(pCtx[k].functionId)) {
+      SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(&pCtx[k]);
+      char* p = GET_ROWCELL_INTERBUF(pEntryInfo);
+
+      SScalarParam out = {.columnData = NULL};
+      out.columnData = taosMemoryCalloc(1, sizeof(SColumnInfoData));
+      out.columnData->info.type = TSDB_DATA_TYPE_BIGINT;
+      out.columnData->info.bytes = tDataTypes[TSDB_DATA_TYPE_BIGINT].bytes;
+      out.columnData->pData = p;
+      pCtx[k].sfp.process(&intervalParam, 1, &out);
+      pEntryInfo->numOfRes = 1;
+      pEntryInfo->hasResult = ',';
+      continue;
+    }
+
     if (functionNeedToExecute(&pCtx[k])) {
       pCtx[k].fpSet.process(&pCtx[k]);
     }
@@ -1054,7 +1096,7 @@ static int32_t getNextQualifiedWindow(SInterval* pInterval, STimeWindow* pNext,
   int32_t order = pInfo->order;
   bool    ascQuery = (order == TSDB_ORDER_ASC);
 
-  int32_t precision = pInfo->precision;
+  int32_t precision = pInterval->precision;
   getNextTimeWindow(pInterval, precision, order, pNext);
 
   // next time window is not in current block
@@ -1489,15 +1531,19 @@ static void doWindowBorderInterpolation(SOperatorInfo* pOperatorInfo, SSDataBloc
   }
 }
 
-static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResultRowInfo, SSDataBlock* pSDataBlock,
+static SArray* hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResultRowInfo, SSDataBlock* pSDataBlock, int32_t tableGroupId) {
-                            int32_t tableGroupId) {
   STableIntervalOperatorInfo* pInfo = (STableIntervalOperatorInfo*)pOperatorInfo->info;
 
   SExecTaskInfo* pTaskInfo = pOperatorInfo->pTaskInfo;
   int32_t        numOfOutput = pOperatorInfo->numOfOutput;
 
+  SArray* pUpdated = NULL;
+  if (pInfo->execModel == OPTR_EXEC_MODEL_STREAM) {
+    pUpdated = taosArrayInit(4, sizeof(SResultRowPosition));
+  }
+
   int32_t step = 1;
-  bool    ascQuery = true;
+  bool    ascScan = true;
 
   int32_t prevIndex = pResultRowInfo->curPos;
 
@@ -1509,10 +1555,10 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
            tsCols[pSDataBlock->info.rows - 1] == pSDataBlock->info.window.ekey);
   }
 
-  int32_t startPos = ascQuery ? 0 : (pSDataBlock->info.rows - 1);
+  int32_t startPos = ascScan? 0 : (pSDataBlock->info.rows - 1);
-  TSKEY   ts = getStartTsKey(&pSDataBlock->info.window, tsCols, pSDataBlock->info.rows, ascQuery);
+  TSKEY   ts = getStartTsKey(&pSDataBlock->info.window, tsCols, pSDataBlock->info.rows, ascScan);
 
-  STimeWindow win = getActiveTimeWindow(pResultRowInfo, ts, &pInfo->interval, pInfo->precision, &pInfo->win);
+  STimeWindow win = getActiveTimeWindow(pResultRowInfo, ts, &pInfo->interval, pInfo->interval.precision, &pInfo->win);
   bool        masterScan = true;
 
   SResultRow* pResult = NULL;
@@ -1523,6 +1569,11 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
     longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
   }
 
+  if (pInfo->execModel == OPTR_EXEC_MODEL_STREAM) {
+    SResultRowPosition pos = {.pageId = pResult->pageId, .offset = pResult->offset};
+    taosArrayPush(pUpdated, &pos);
+  }
+
   int32_t forwardStep = 0;
   TSKEY   ekey = win.ekey;
   forwardStep =
@@ -1534,8 +1585,7 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
     for (int32_t j = prevIndex; j < curIndex; ++j) {  // previous time window may be all closed already.
       SResultRow* pRes = getResultRow(pResultRowInfo, j);
       if (pRes->closed) {
-        assert(resultRowInterpolated(pRes, RESULT_ROW_START_INTERP) &&
+        assert(resultRowInterpolated(pRes, RESULT_ROW_START_INTERP) && resultRowInterpolated(pRes, RESULT_ROW_END_INTERP));
-               resultRowInterpolated(pRes, RESULT_ROW_END_INTERP));
         continue;
       }
 
@@ -1548,14 +1598,13 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
       }
 
       assert(!resultRowInterpolated(pResult, RESULT_ROW_END_INTERP));
-
       doTimeWindowInterpolation(pOperatorInfo, &pInfo->binfo, pSDataBlock->pDataBlock, *(TSKEY*)pInfo->pRow[0], -1,
                                 tsCols[startPos], startPos, w.ekey, RESULT_ROW_END_INTERP);
 
       setResultRowInterpo(pResult, RESULT_ROW_END_INTERP);
       setNotInterpoWindowKey(pInfo->binfo.pCtx, pOperatorInfo->numOfOutput, RESULT_ROW_START_INTERP);
 
-      doApplyFunctions(pInfo->binfo.pCtx, &w, startPos, 0, tsCols, pSDataBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
+      doApplyFunctions(pInfo->binfo.pCtx, &w, &pInfo->timeWindowData, startPos, 0, tsCols, pSDataBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
     }
 
     // restore current time window
@@ -1570,8 +1619,7 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
   // window start key interpolation
   doWindowBorderInterpolation(pOperatorInfo, pSDataBlock, pInfo->binfo.pCtx, pResult, &win, startPos, forwardStep,
                               pInfo->order, false);
-  doApplyFunctions(pInfo->binfo.pCtx, &win, startPos, forwardStep, tsCols, pSDataBlock->info.rows, numOfOutput,
+  doApplyFunctions(pInfo->binfo.pCtx, &win, &pInfo->timeWindowData, startPos, forwardStep, tsCols, pSDataBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
-                   TSDB_ORDER_ASC);
 
   STimeWindow nextWin = win;
   while (1) {
@@ -1589,6 +1637,11 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
       longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
     }
 
+    if (pInfo->execModel == OPTR_EXEC_MODEL_STREAM) {
+      SResultRowPosition pos = {.pageId = pResult->pageId, .offset = pResult->offset};
+      taosArrayPush(pUpdated, &pos);
+    }
+
     ekey = nextWin.ekey;  // reviseWindowEkey(pQueryAttr, &nextWin);
     forwardStep =
         getNumOfRowsInTimeWindow(&pSDataBlock->info, tsCols, startPos, ekey, binarySearchForKey, NULL, TSDB_ORDER_ASC);
@@ -1596,15 +1649,15 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
     // window start(end) key interpolation
     doWindowBorderInterpolation(pOperatorInfo, pSDataBlock, pInfo->binfo.pCtx, pResult, &nextWin, startPos, forwardStep,
                                 pInfo->order, false);
-    doApplyFunctions(pInfo->binfo.pCtx, &nextWin, startPos, forwardStep, tsCols, pSDataBlock->info.rows, numOfOutput,
+    doApplyFunctions(pInfo->binfo.pCtx, &nextWin, &pInfo->timeWindowData, startPos, forwardStep, tsCols, pSDataBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
-                     TSDB_ORDER_ASC);
   }
 
   if (pInfo->timeWindowInterpo) {
-    int32_t rowIndex = ascQuery ? (pSDataBlock->info.rows - 1) : 0;
+    int32_t rowIndex = ascScan ? (pSDataBlock->info.rows - 1) : 0;
     saveDataBlockLastRow(pInfo->pRow, pSDataBlock->pDataBlock, rowIndex, pSDataBlock->info.numOfCols);
   }
 
+  return pUpdated;
   //  updateResultRowInfoActiveIndex(pResultRowInfo, &pInfo->win, pRuntimeEnv->current->lastKey, true, false);
 }
 
@@ -1841,7 +1894,7 @@ static void doHashGroupbyAgg(SOperatorInfo* pOperator, SSDataBlock* pBlock) {
     }
 
     int32_t rowIndex = j - num;
-    doApplyFunctions(pCtx, &w, rowIndex, num, NULL, pBlock->info.rows, pOperator->numOfOutput, TSDB_ORDER_ASC);
+    doApplyFunctions(pCtx, &w, NULL, rowIndex, num, NULL, pBlock->info.rows, pOperator->numOfOutput, TSDB_ORDER_ASC);
 
     // assign the group keys or user input constant values if required
     doAssignGroupKeys(pCtx, pOperator->numOfOutput, pBlock->info.rows, rowIndex);
@@ -1859,7 +1912,7 @@ static void doHashGroupbyAgg(SOperatorInfo* pOperator, SSDataBlock* pBlock) {
     }
 
     int32_t rowIndex = pBlock->info.rows - num;
-    doApplyFunctions(pCtx, &w, rowIndex, num, NULL, pBlock->info.rows, pOperator->numOfOutput, TSDB_ORDER_ASC);
+    doApplyFunctions(pCtx, &w, NULL, rowIndex, num, NULL, pBlock->info.rows, pOperator->numOfOutput, TSDB_ORDER_ASC);
     doAssignGroupKeys(pCtx, pOperator->numOfOutput, pBlock->info.rows, rowIndex);
   }
 }
@@ -1910,8 +1963,7 @@ static void doSessionWindowAggImpl(SOperatorInfo* pOperator, SSessionAggOperator
       }
 
       // pInfo->numOfRows data belong to the current session window
-      doApplyFunctions(pInfo->binfo.pCtx, &window, pInfo->start, pInfo->numOfRows, NULL, pBlock->info.rows, numOfOutput,
+      doApplyFunctions(pInfo->binfo.pCtx, &window, NULL, pInfo->start, pInfo->numOfRows, NULL, pBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
-                       TSDB_ORDER_ASC);
 
       pInfo->curWindow.skey = tsList[j];
       pInfo->curWindow.ekey = tsList[j];
@@ -1931,8 +1983,7 @@ static void doSessionWindowAggImpl(SOperatorInfo* pOperator, SSessionAggOperator
     longjmp(pTaskInfo->env, TSDB_CODE_QRY_APP_ERROR);
   }
 
-  doApplyFunctions(pInfo->binfo.pCtx, &window, pInfo->start, pInfo->numOfRows, NULL, pBlock->info.rows, numOfOutput,
+  doApplyFunctions(pInfo->binfo.pCtx, &window, NULL, pInfo->start, pInfo->numOfRows, NULL, pBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
-                   TSDB_ORDER_ASC);
 }
 
 static void setResultRowKey(SResultRow* pResultRow, char* pData, int16_t type) {
@@ -1999,11 +2050,7 @@ static bool functionNeedToExecute(SqlFunctionCtx* pCtx) {
     return false;
   }
 
-  if (functionId == FUNCTION_TS) {
+  if (isRowEntryCompleted(pResInfo)) {
-    return true;
-  }
-
-  if (isRowEntryCompleted(pResInfo) || functionId == FUNCTION_TAG_DUMMY || functionId == FUNCTION_TS_DUMMY) {
     return false;
   }
 
@@ -2118,6 +2165,9 @@ static SqlFunctionCtx* createSqlFunctionCtx_rv(SExprInfo* pExprInfo, int32_t num
         pCtx->fpSet.getEnv(pExpr->pExpr->_function.pFunctNode, &env);
       } else {
         fmGetScalarFuncExecFuncs(pCtx->functionId, &pCtx->sfp);
+        if (pCtx->sfp.getEnv != NULL) {
+          pCtx->sfp.getEnv(pExpr->pExpr->_function.pFunctNode, &env);
+        }
       }
       pCtx->resDataInfo.interBufSize = env.calcMemSize;
     } else if (pExpr->pExpr->nodeType == QUERY_NODE_COLUMN) {
@@ -3590,6 +3640,42 @@ void finalizeMultiTupleQueryResult(SqlFunctionCtx* pCtx, int32_t numOfOutput, SD
   }
 }
 
+void finalizeUpdatedResult(SqlFunctionCtx* pCtx, int32_t numOfOutput, SDiskbasedBuf* pBuf, SArray* pUpdateList,
+                           int32_t* rowCellInfoOffset) {
+  size_t num = taosArrayGetSize(pUpdateList);
+
+  for (int32_t i = 0; i < num; ++i) {
+    SResultRowPosition* pPos = taosArrayGet(pUpdateList, i);
+
+    SFilePage*  bufPage = getBufPage(pBuf, pPos->pageId);
+    SResultRow* pRow = (SResultRow*)((char*)bufPage + pPos->offset);
+
+    for (int32_t j = 0; j < numOfOutput; ++j) {
+      pCtx[j].resultInfo = getResultCell(pRow, j, rowCellInfoOffset);
+
+      struct SResultRowEntryInfo* pResInfo = pCtx[j].resultInfo;
+      if (isRowEntryCompleted(pResInfo) && isRowEntryInitialized(pResInfo)) {
+        continue;
+      }
+
+      if (pCtx[j].fpSet.process) {  // TODO set the dummy function.
+        pCtx[j].fpSet.finalize(&pCtx[j]);
+      }
+
+      if (pRow->numOfRows < pResInfo->numOfRes) {
+        pRow->numOfRows = pResInfo->numOfRes;
+      }
+    }
+
+    releaseBufPage(pBuf, bufPage);
+    /*
+     * set the number of output results for group by normal columns, the number of output rows usually is 1 except
+     * the top and bottom query
+     */
+    //    buf->numOfRows = (uint16_t)getNumOfResult(pCtx, numOfOutput);
+  }
+}
+
 static bool hasMainOutput(STaskAttr* pQueryAttr) {
   for (int32_t i = 0; i < pQueryAttr->numOfOutput; ++i) {
     int32_t functionId = getExprFunctionId(&pQueryAttr->pExpr1[i]);
@@ -3680,7 +3766,6 @@ void setResultRowOutputBufInitCtx(STaskRuntimeEnv* pRuntimeEnv, SResultRow* pRes
 
 void setResultRowOutputBufInitCtx_rv(SDiskbasedBuf* pBuf, SResultRow* pResult, SqlFunctionCtx* pCtx,
                                      int32_t numOfOutput, int32_t* rowCellInfoOffset) {
-  // Note: pResult->pos[i]->num == 0, there is only fixed number of results for each group
   for (int32_t i = 0; i < numOfOutput; ++i) {
     pCtx[i].resultInfo = getResultCell(pResult, i, rowCellInfoOffset);
 
@@ -3688,6 +3773,11 @@ void setResultRowOutputBufInitCtx_rv(SDiskbasedBuf* pBuf, SResultRow* pResult, S
     if (isRowEntryCompleted(pResInfo) && isRowEntryInitialized(pResInfo)) {
       continue;
     }
+
+    if (fmIsWindowPseudoColumnFunc(pCtx[i].functionId)) {
+      continue;
+    }
+
     //    int32_t functionId = pCtx[i].functionId;
     //    if (functionId < 0) {
     //      continue;
@@ -4032,8 +4122,7 @@ static void toSDatablock(SGroupResInfo* pGroupResInfo, SDiskbasedBuf* pBuf, SSDa
     return;
   }
 
-  int32_t orderType =
+  int32_t orderType = TSDB_ORDER_ASC;
-      TSDB_ORDER_ASC;  //(pQueryAttr->pGroupbyExpr != NULL) ? pQueryAttr->pGroupbyExpr->orderType : TSDB_ORDER_ASC;
   doCopyToSDataBlock(pBuf, pGroupResInfo, orderType, pBlock, rowCapacity, rowCellOffset);
 
   // add condition (pBlock->info.rows >= 1) just to runtime happy
@@ -5074,12 +5163,12 @@ static SSDataBlock* concurrentlyLoadRemoteDataImpl(SOperatorInfo* pOperator, SEx
     for (int32_t i = 0; i < totalSources; ++i) {
       SSourceDataInfo* pDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, i);
 
-      if (pDataInfo->status == DATA_EXHAUSTED) {
+      if (pDataInfo->status == EX_SOURCE_DATA_EXHAUSTED) {
         completed += 1;
         continue;
       }
 
-      if (pDataInfo->status != DATA_READY) {
+      if (pDataInfo->status != EX_SOURCE_DATA_READY) {
         continue;
       }
 
@@ -5093,7 +5182,7 @@ static SSDataBlock* concurrentlyLoadRemoteDataImpl(SOperatorInfo* pOperator, SEx
                " try next",
                GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, i + 1, pDataInfo->totalRows,
                pExchangeInfo->loadInfo.totalRows);
-        pDataInfo->status = DATA_EXHAUSTED;
+        pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
         completed += 1;
         continue;
       }
@@ -5111,16 +5200,15 @@ static SSDataBlock* concurrentlyLoadRemoteDataImpl(SOperatorInfo* pOperator, SEx
                ", totalRows:%" PRIu64 ", totalBytes:%" PRIu64 " try next %d/%" PRIzu,
                GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pRes->info.rows, pDataInfo->totalRows,
                pLoadInfo->totalRows, pLoadInfo->totalSize, i + 1, totalSources);
-        pDataInfo->status = DATA_EXHAUSTED;
+        pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
       } else {
-        qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " numOfRows:%d, totalRows:%" PRIu64
+        qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " numOfRows:%d, totalRows:%" PRIu64 ", totalBytes:%" PRIu64,
-               ", totalBytes:%" PRIu64,
                GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pRes->info.rows, pLoadInfo->totalRows,
                pLoadInfo->totalSize);
       }
 
-      if (pDataInfo->status != DATA_EXHAUSTED) {
+      if (pDataInfo->status != EX_SOURCE_DATA_EXHAUSTED) {
-        pDataInfo->status = DATA_NOT_READY;
+        pDataInfo->status = EX_SOURCE_DATA_NOT_READY;
         code = doSendFetchDataRequest(pExchangeInfo, pTaskInfo, i);
         if (code != TSDB_CODE_SUCCESS) {
           goto _error;
@@ -5223,7 +5311,7 @@ static SSDataBlock* seqLoadRemoteData(SOperatorInfo* pOperator) {
              GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pExchangeInfo->current + 1,
              pDataInfo->totalRows, pLoadInfo->totalRows);
 
-      pDataInfo->status = DATA_EXHAUSTED;
+      pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
       pExchangeInfo->current += 1;
       continue;
     }
@@ -5240,7 +5328,7 @@ static SSDataBlock* seqLoadRemoteData(SOperatorInfo* pOperator) {
              GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pRes->info.rows, pDataInfo->totalRows,
              pLoadInfo->totalRows, pLoadInfo->totalSize, pExchangeInfo->current + 1, totalSources);
 
-      pDataInfo->status = DATA_EXHAUSTED;
+      pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
       pExchangeInfo->current += 1;
     } else {
       qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " numOfRows:%d, totalRows:%" PRIu64
@@ -6845,37 +6933,6 @@ static SSDataBlock* doLimit(SOperatorInfo* pOperator, bool* newgroup) {
   return pBlock;
 }
 
-static SSDataBlock* doFilter(void* param, bool* newgroup) {
-  SOperatorInfo* pOperator = (SOperatorInfo*)param;
-  if (pOperator->status == OP_EXEC_DONE) {
-    return NULL;
-  }
-
-  SFilterOperatorInfo* pCondInfo = pOperator->info;
-  STaskRuntimeEnv*     pRuntimeEnv = pOperator->pRuntimeEnv;
-
-  while (1) {
-    publishOperatorProfEvent(pOperator->pDownstream[0], QUERY_PROF_BEFORE_OPERATOR_EXEC);
-    SSDataBlock* pBlock = pOperator->pDownstream[0]->getNextFn(pOperator->pDownstream[0], newgroup);
-    publishOperatorProfEvent(pOperator->pDownstream[0], QUERY_PROF_AFTER_OPERATOR_EXEC);
-
-    if (pBlock == NULL) {
-      break;
-    }
-
-    doSetFilterColumnInfo(pCondInfo->pFilterInfo, pCondInfo->numOfFilterCols, pBlock);
-    assert(pRuntimeEnv->pTsBuf == NULL);
-    filterRowsInDataBlock(pRuntimeEnv, pCondInfo->pFilterInfo, pCondInfo->numOfFilterCols, pBlock, true);
-
-    if (pBlock->info.rows > 0) {
-      return pBlock;
-    }
-  }
-
-  doSetOperatorCompleted(pOperator);
-  return NULL;
-}
-
 static int32_t doOpenIntervalAgg(SOperatorInfo* pOperator) {
   if (OPTR_IS_OPENED(pOperator)) {
     return TSDB_CODE_SUCCESS;
@@ -6884,8 +6941,8 @@ static int32_t doOpenIntervalAgg(SOperatorInfo* pOperator) {
   STableIntervalOperatorInfo* pInfo = pOperator->info;
 
   int32_t order = TSDB_ORDER_ASC;
-  //  STimeWindow win = pQueryAttr->window;
+  //  STimeWindow win = {0};
-  bool           newgroup = false;
+  bool newgroup = false;
   SOperatorInfo* downstream = pOperator->pDownstream[0];
 
   while (1) {
@@ -6898,7 +6955,6 @@ static int32_t doOpenIntervalAgg(SOperatorInfo* pOperator) {
     }
 
     //    setTagValue(pOperator, pRuntimeEnv->current->pTable, pInfo->pCtx, pOperator->numOfOutput);
-
     // the pDataBlock are always the same one, no need to call this again
     setInputDataBlock(pOperator, pInfo->binfo.pCtx, pBlock, order);
     hashIntervalAgg(pOperator, &pInfo->binfo.resultRowInfo, pBlock, 0);
@@ -6931,6 +6987,10 @@ static SSDataBlock* doBuildIntervalResult(SOperatorInfo* pOperator, bool* newgro
     return NULL;
   }
 
+  if (pInfo->execModel == OPTR_EXEC_MODEL_STREAM) {
+    return pOperator->getStreamResFn(pOperator, newgroup);
+  }
+
   pTaskInfo->code = pOperator->_openFn(pOperator);
   if (pTaskInfo->code != TSDB_CODE_SUCCESS) {
     return NULL;
@@ -6947,7 +7007,60 @@ static SSDataBlock* doBuildIntervalResult(SOperatorInfo* pOperator, bool* newgro
   return pInfo->binfo.pRes->info.rows == 0 ? NULL : pInfo->binfo.pRes;
 }
 
-static SSDataBlock* doAllIntervalAgg(SOperatorInfo* pOperator, bool* newgroup) {
+static SSDataBlock* doStreamIntervalAgg(SOperatorInfo *pOperator, bool* newgroup) {
+  STableIntervalOperatorInfo* pInfo = pOperator->info;
+  int32_t order = TSDB_ORDER_ASC;
+
+  if (pOperator->status == OP_EXEC_DONE) {
+    return NULL;
+  }
+
+  if (pOperator->status == OP_RES_TO_RETURN) {
+    toSDatablock(&pInfo->groupResInfo, pInfo->aggSup.pResultBuf, pInfo->binfo.pRes, pInfo->binfo.capacity,
+                 pInfo->binfo.rowCellInfoOffset);
+    if (pInfo->binfo.pRes->info.rows == 0 || !hasRemainDataInCurrentGroup(&pInfo->groupResInfo)) {
+      pOperator->status = OP_EXEC_DONE;
+    }
+    return pInfo->binfo.pRes;
+  }
+
+  //  STimeWindow win = {0};
+  *newgroup = false;
+  SOperatorInfo* downstream = pOperator->pDownstream[0];
+
+  SArray* pUpdated = NULL;
+
+  while (1) {
+    publishOperatorProfEvent(downstream, QUERY_PROF_BEFORE_OPERATOR_EXEC);
+    SSDataBlock* pBlock = downstream->getNextFn(downstream, newgroup);
+    publishOperatorProfEvent(downstream, QUERY_PROF_AFTER_OPERATOR_EXEC);
+
+    if (pBlock == NULL) {
+      break;
+    }
+
+    // The timewindows that overlaps the timestamps of the input pBlock need to be recalculated and return to the caller.
+    // Note that all the time window are not close till now.
+
+    //    setTagValue(pOperator, pRuntimeEnv->current->pTable, pInfo->pCtx, pOperator->numOfOutput);
+    // the pDataBlock are always the same one, no need to call this again
+    setInputDataBlock(pOperator, pInfo->binfo.pCtx, pBlock, order);
+    pUpdated = hashIntervalAgg(pOperator, &pInfo->binfo.resultRowInfo, pBlock, 0);
+  }
+
+  finalizeUpdatedResult(pInfo->binfo.pCtx, pOperator->numOfOutput, pInfo->aggSup.pResultBuf, pUpdated, pInfo->binfo.rowCellInfoOffset);
+
+  blockDataEnsureCapacity(pInfo->binfo.pRes, pInfo->binfo.capacity);
+  toSDatablock(&pInfo->groupResInfo, pInfo->aggSup.pResultBuf, pInfo->binfo.pRes, pInfo->binfo.capacity,
+               pInfo->binfo.rowCellInfoOffset);
+
+  ASSERT(pInfo->binfo.pRes->info.rows > 0);
+  pOperator->status = OP_RES_TO_RETURN;
+
+  return pInfo->binfo.pRes->info.rows == 0 ? NULL : pInfo->binfo.pRes;
+}
+
+static SSDataBlock* doAllIntervalAgg(SOperatorInfo *pOperator, bool* newgroup) {
   if (pOperator->status == OP_EXEC_DONE) {
     return NULL;
   }
@@ -7581,8 +7694,7 @@ SOperatorInfo* createAggregateOperatorInfo(SOperatorInfo* downstream, SExprInfo*
 
   //(int32_t)(getRowNumForMultioutput(pQueryAttr, pQueryAttr->topBotQuery, pQueryAttr->stableQuery));
   int32_t numOfRows = 1;
-  int32_t code =
+  int32_t code = initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfCols, numOfRows, pResultBlock, pTaskInfo->id.str);
-      initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfCols, numOfRows, pResultBlock, pTaskInfo->id.str);
   pInfo->pTableQueryInfo = initTableQueryInfo(pTableGroupInfo);
   if (code != TSDB_CODE_SUCCESS || pInfo->pTableQueryInfo == NULL) {
     goto _error;
@@ -7590,18 +7702,18 @@ SOperatorInfo* createAggregateOperatorInfo(SOperatorInfo* downstream, SExprInfo*
 
   setFunctionResultOutput(&pInfo->binfo, &pInfo->aggSup, MAIN_SCAN, pTaskInfo);
 
-  pOperator->name = "TableAggregate";
+  pOperator->name         = "TableAggregate";
   pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_AGG;
   pOperator->blockingOptr = true;
-  pOperator->status = OP_NOT_OPENED;
+  pOperator->status       = OP_NOT_OPENED;
-  pOperator->info = pInfo;
+  pOperator->info         = pInfo;
-  pOperator->pExpr = pExprInfo;
+  pOperator->pExpr        = pExprInfo;
-  pOperator->numOfOutput = numOfCols;
+  pOperator->numOfOutput  = numOfCols;
 
-  pOperator->pTaskInfo = pTaskInfo;
+  pOperator->pTaskInfo    = pTaskInfo;
-  pOperator->_openFn = doOpenAggregateOptr;
+  pOperator->_openFn      = doOpenAggregateOptr;
-  pOperator->getNextFn = getAggregateResult;
+  pOperator->getNextFn    = getAggregateResult;
-  pOperator->closeFn = destroyAggOperatorInfo;
+  pOperator->closeFn      = destroyAggOperatorInfo;
   pOperator->encodeResultRow = aggEncodeResultRow;
   pOperator->decodeResultRow = aggDecodeResultRow;
 
@@ -7778,16 +7890,16 @@ SOperatorInfo* createProjectOperatorInfo(SOperatorInfo* downstream, SExprInfo* p
   //  initResultRowInfo(&pBInfo->resultRowInfo, 8);
   //  setFunctionResultOutput(pBInfo, MAIN_SCAN);
 
-  pOperator->name = "ProjectOperator";
+  pOperator->name         = "ProjectOperator";
   pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_PROJECT;
   pOperator->blockingOptr = false;
-  pOperator->status = OP_NOT_OPENED;
+  pOperator->status       = OP_NOT_OPENED;
-  pOperator->info = pInfo;
+  pOperator->info         = pInfo;
-  pOperator->pExpr = pExprInfo;
+  pOperator->pExpr        = pExprInfo;
-  pOperator->numOfOutput = num;
+  pOperator->numOfOutput  = num;
-  pOperator->_openFn = operatorDummyOpenFn;
+  pOperator->_openFn      = operatorDummyOpenFn;
-  pOperator->getNextFn = doProjectOperation;
+  pOperator->getNextFn    = doProjectOperation;
-  pOperator->closeFn = destroyProjectOperatorInfo;
+  pOperator->closeFn      = destroyProjectOperatorInfo;
 
   pOperator->pTaskInfo = pTaskInfo;
   int32_t code = appendDownstream(pOperator, &downstream, 1);
@@ -7802,39 +7914,6 @@ _error:
   return NULL;
 }
 
-SColumnInfo* extractColumnFilterInfo(SExprInfo* pExpr, int32_t numOfOutput, int32_t* numOfFilterCols) {
-#if 0
-  SColumnInfo* pCols = taosMemoryCalloc(numOfOutput, sizeof(SColumnInfo));
-
-  int32_t numOfFilter = 0;
-  for(int32_t i = 0; i < numOfOutput; ++i) {
-    if (pExpr[i].base.flist.numOfFilters > 0) {
-      numOfFilter += 1;
-    }
-
-    pCols[i].type  = pExpr[i].base.resSchema.type;
-    pCols[i].bytes = pExpr[i].base.resSchema.bytes;
-    pCols[i].colId = pExpr[i].base.resSchema.colId;
-
-    pCols[i].flist.numOfFilters = pExpr[i].base.flist.numOfFilters;
-    if (pCols[i].flist.numOfFilters != 0) { 
-      pCols[i].flist.filterInfo   = taosMemoryCalloc(pCols[i].flist.numOfFilters, sizeof(SColumnFilterInfo));
-      memcpy(pCols[i].flist.filterInfo, pExpr[i].base.flist.filterInfo, pCols[i].flist.numOfFilters * sizeof(SColumnFilterInfo));
-    } else {
-      // avoid runtime error
-      pCols[i].flist.filterInfo   = NULL; 
-    }
-  }
-
-  assert(numOfFilter > 0);
-
-  *numOfFilterCols = numOfFilter;
-  return pCols;
-#endif
-
-  return 0;
-}
-
 SOperatorInfo* createLimitOperatorInfo(SOperatorInfo* downstream, SLimit* pLimit, SExecTaskInfo* pTaskInfo) {
   SLimitOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SLimitOperatorInfo));
   SOperatorInfo*      pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
@@ -7845,17 +7924,18 @@ SOperatorInfo* createLimitOperatorInfo(SOperatorInfo* downstream, SLimit* pLimit
   pInfo->limit = *pLimit;
   pInfo->currentOffset = pLimit->offset;
 
-  pOperator->name = "LimitOperator";
+  pOperator->name         = "LimitOperator";
   //  pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_LIMIT;
   pOperator->blockingOptr = false;
-  pOperator->status = OP_NOT_OPENED;
+  pOperator->status       = OP_NOT_OPENED;
-  pOperator->_openFn = operatorDummyOpenFn;
+  pOperator->_openFn      = operatorDummyOpenFn;
-  pOperator->getNextFn = doLimit;
+  pOperator->getNextFn    = doLimit;
-  pOperator->info = pInfo;
+  pOperator->info         = pInfo;
-  pOperator->pTaskInfo = pTaskInfo;
+  pOperator->pTaskInfo    = pTaskInfo;
-  int32_t code = appendDownstream(pOperator, &downstream, 1);
 
+  int32_t code = appendDownstream(pOperator, &downstream, 1);
   return pOperator;
+
 _error:
   taosMemoryFreeClear(pInfo);
   taosMemoryFreeClear(pOperator);
@@ -7872,17 +7952,18 @@ SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo*
     goto _error;
   }
 
-  pInfo->order = TSDB_ORDER_ASC;
+  pInfo->order     = TSDB_ORDER_ASC;
-  pInfo->precision = TSDB_TIME_PRECISION_MILLI;
+  pInfo->interval  = *pInterval;
-  pInfo->win = pTaskInfo->window;
+  pInfo->execModel = pTaskInfo->execModel;
-  pInfo->interval = *pInterval;
 
-  pInfo->win.skey = INT64_MIN;
+  pInfo->win       = pTaskInfo->window;
-  pInfo->win.ekey = INT64_MAX;
+  pInfo->win.skey  = 0;
+  pInfo->win.ekey  = INT64_MAX;
 
   int32_t numOfRows = 4096;
-  int32_t code =
+  int32_t code = initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfCols, numOfRows, pResBlock, pTaskInfo->id.str);
-      initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfCols, numOfRows, pResBlock, pTaskInfo->id.str);
+  initExecTimeWindowInfo(&pInfo->timeWindowData, &pInfo->win);
+
   //  pInfo->pTableQueryInfo = initTableQueryInfo(pTableGroupInfo);
   if (code != TSDB_CODE_SUCCESS /* || pInfo->pTableQueryInfo == NULL*/) {
     goto _error;
@@ -7890,7 +7971,7 @@ SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo*
 
   initResultRowInfo(&pInfo->binfo.resultRowInfo, (int32_t)1);
 
-  pOperator->name = "TimeIntervalAggOperator";
+  pOperator->name         = "TimeIntervalAggOperator";
   pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_INTERVAL;
   pOperator->blockingOptr = true;
   pOperator->status       = OP_NOT_OPENED;
@@ -7900,6 +7981,7 @@ SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo*
   pOperator->info         = pInfo;
   pOperator->_openFn      = doOpenIntervalAgg;
   pOperator->getNextFn    = doBuildIntervalResult;
+  pOperator->getStreamResFn= doStreamIntervalAgg;
   pOperator->closeFn      = destroyIntervalOperatorInfo;
   pOperator->encodeResultRow = aggEncodeResultRow;
   pOperator->decodeResultRow = aggDecodeResultRow;
@@ -8626,6 +8708,23 @@ static SResSchema createResSchema(int32_t type, int32_t bytes, int32_t slotId, i
   return s;
 }
 
+static SColumn* createColumn(int32_t blockId, int32_t slotId, SDataType* pType) {
+  SColumn* pCol = taosMemoryCalloc(1, sizeof(SColumn));
+  if (pCol == NULL) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    return NULL;
+  }
+
+  pCol->slotId      = slotId;
+  pCol->bytes       = pType->bytes;
+  pCol->type        = pType->type;
+  pCol->scale       = pType->scale;
+  pCol->precision   = pType->precision;
+  pCol->dataBlockId = blockId;
+
+  return pCol;
+}
+
 SExprInfo* createExprInfo(SNodeList* pNodeList, SNodeList* pGroupKeys, int32_t* numOfExprs) {
   int32_t numOfFuncs = LIST_LENGTH(pNodeList);
   int32_t numOfGroupKeys = 0;
@@ -8657,18 +8756,11 @@ SExprInfo* createExprInfo(SNodeList* pNodeList, SNodeList* pGroupKeys, int32_t*
 
       pExp->base.pParam = taosMemoryCalloc(1, sizeof(SFunctParam));
       pExp->base.numOfParams = 1;
-      pExp->base.pParam[0].pCol = taosMemoryCalloc(1, sizeof(SColumn));
-      pExp->base.pParam[0].type = FUNC_PARAM_TYPE_COLUMN;
 
       SDataType* pType = &pColNode->node.resType;
       pExp->base.resSchema = createResSchema(pType->type, pType->bytes, pTargetNode->slotId, pType->scale, pType->precision, pColNode->colName);
-
+      pExp->base.pParam[0].pCol = createColumn(pColNode->dataBlockId, pColNode->slotId, pType);
-      SColumn* pCol = pExp->base.pParam[0].pCol;
+      pExp->base.pParam[0].type = FUNC_PARAM_TYPE_COLUMN;
-      pCol->slotId    = pColNode->slotId;   // TODO refactor
-      pCol->bytes     = pType->bytes;
-      pCol->type      = pType->type;
-      pCol->scale     = pType->scale;
-      pCol->precision = pType->precision;
     } else if (nodeType(pTargetNode->pExpr) == QUERY_NODE_FUNCTION) {
       pExp->pExpr->nodeType = QUERY_NODE_FUNCTION;
       SFunctionNode* pFuncNode = (SFunctionNode*)pTargetNode->pExpr;
@@ -8679,8 +8771,7 @@ SExprInfo* createExprInfo(SNodeList* pNodeList, SNodeList* pGroupKeys, int32_t*
 
       pExp->pExpr->_function.functionId = pFuncNode->funcId;
       pExp->pExpr->_function.pFunctNode = pFuncNode;
-      strncpy(pExp->pExpr->_function.functionName, pFuncNode->functionName,
+      strncpy(pExp->pExpr->_function.functionName, pFuncNode->functionName, tListLen(pExp->pExpr->_function.functionName));
-              tListLen(pExp->pExpr->_function.functionName));
 
       // TODO: value parameter needs to be handled
       int32_t numOfParam = LIST_LENGTH(pFuncNode->pParameterList);
@@ -8691,21 +8782,12 @@ SExprInfo* createExprInfo(SNodeList* pNodeList, SNodeList* pGroupKeys, int32_t*
       for (int32_t j = 0; j < numOfParam; ++j) {
         SNode*       p1 = nodesListGetNode(pFuncNode->pParameterList, j);
         if (p1->type == QUERY_NODE_COLUMN) {
-          SColumnNode* pcn = (SColumnNode*)p1;  // TODO refactor
+          SColumnNode* pcn = (SColumnNode*) p1;
 
           pExp->base.pParam[j].type = FUNC_PARAM_TYPE_COLUMN;
-          pExp->base.pParam[j].pCol = taosMemoryCalloc(1, sizeof(SColumn));
+          pExp->base.pParam[j].pCol = createColumn(pcn->dataBlockId, pcn->slotId, &pcn->node.resType);
-          SColumn* pCol = pExp->base.pParam[j].pCol;
-
-          pCol->slotId      = pcn->slotId;
-          pCol->bytes       = pcn->node.resType.bytes;
-          pCol->type        = pcn->node.resType.type;
-          pCol->scale       = pcn->node.resType.scale;
-          pCol->precision   = pcn->node.resType.precision;
-          pCol->dataBlockId = pcn->dataBlockId;
         } else if (p1->type == QUERY_NODE_VALUE) {
           SValueNode* pvn = (SValueNode*)p1;
-
           pExp->base.pParam[j].type = FUNC_PARAM_TYPE_VALUE;
         }
       }
@@ -8715,21 +8797,14 @@ SExprInfo* createExprInfo(SNodeList* pNodeList, SNodeList* pGroupKeys, int32_t*
 
       pExp->base.pParam = taosMemoryCalloc(1, sizeof(SFunctParam));
       pExp->base.numOfParams = 1;
-      pExp->base.pParam[0].pCol = taosMemoryCalloc(1, sizeof(SColumn));
-      pExp->base.pParam[0].type = FUNC_PARAM_TYPE_COLUMN;
 
       SDataType* pType = &pNode->node.resType;
-      pExp->base.resSchema = createResSchema(pType->type, pType->bytes, pTargetNode->slotId, pType->scale,
+      pExp->base.resSchema = createResSchema(pType->type, pType->bytes, pTargetNode->slotId, pType->scale, pType->precision, pNode->node.aliasName);
-                                             pType->precision, pNode->node.aliasName);
 
       pExp->pExpr->_optrRoot.pRootNode = pTargetNode->pExpr;
 
-      SColumn* pCol   = pExp->base.pParam[0].pCol;
+      pExp->base.pParam[0].type = FUNC_PARAM_TYPE_COLUMN;
-      pCol->slotId    = pTargetNode->slotId;   // TODO refactor
+      pExp->base.pParam[0].pCol = createColumn(pTargetNode->dataBlockId, pTargetNode->slotId, pType);
-      pCol->bytes     = pType->bytes;
-      pCol->type      = pType->type;
-      pCol->scale     = pType->scale;
-      pCol->precision = pType->precision;
     } else {
       ASSERT(0);
     }
@@ -8738,12 +8813,13 @@ SExprInfo* createExprInfo(SNodeList* pNodeList, SNodeList* pGroupKeys, int32_t*
   return pExprs;
 }
 
-static SExecTaskInfo* createExecTaskInfo(uint64_t queryId, uint64_t taskId) {
+static SExecTaskInfo* createExecTaskInfo(uint64_t queryId, uint64_t taskId, EOPTR_EXEC_MODEL model) {
   SExecTaskInfo* pTaskInfo = taosMemoryCalloc(1, sizeof(SExecTaskInfo));
   setTaskStatus(pTaskInfo, TASK_NOT_COMPLETED);
 
   pTaskInfo->cost.created = taosGetTimestampMs();
   pTaskInfo->id.queryId = queryId;
+  pTaskInfo->execModel  = model;
 
   char* p = taosMemoryCalloc(1, 128);
   snprintf(p, 128, "TID:0x%" PRIx64 " QID:0x%" PRIx64, taskId, queryId);
@@ -8763,17 +8839,15 @@ static SArray* extractColumnInfo(SNodeList* pNodeList);
 static SArray* extractColMatchInfo(SNodeList* pNodeList, SDataBlockDescNode* pOutputNodeList, int32_t* numOfOutputCols);
 static SArray* createSortInfo(SNodeList* pNodeList);
 
-SOperatorInfo* doCreateOperatorTreeNode(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo, SReadHandle* pHandle,
+SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo, SReadHandle* pHandle,
                                         uint64_t queryId, uint64_t taskId, STableGroupInfo* pTableGroupInfo) {
   if (pPhyNode->pChildren == NULL || LIST_LENGTH(pPhyNode->pChildren) == 0) {
     if (QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN == nodeType(pPhyNode)) {
       SScanPhysiNode* pScanPhyNode = (SScanPhysiNode*)pPhyNode;
 
       int32_t     numOfCols = 0;
-      tsdbReaderT pDataReader =
+      tsdbReaderT pDataReader = doCreateDataReader((STableScanPhysiNode*)pPhyNode, pHandle, pTableGroupInfo, (uint64_t)queryId, taskId);
-          doCreateDataReader((STableScanPhysiNode*)pPhyNode, pHandle, pTableGroupInfo, (uint64_t)queryId, taskId);
+      SArray* pColList = extractColMatchInfo(pScanPhyNode->pScanCols, pScanPhyNode->node.pOutputDataBlockDesc, &numOfCols);
-      SArray* pColList =
-          extractColMatchInfo(pScanPhyNode->pScanCols, pScanPhyNode->node.pOutputDataBlockDesc, &numOfCols);
 
       return createTableScanOperatorInfo(pDataReader, pScanPhyNode->order, numOfCols, pScanPhyNode->count,
                                          pScanPhyNode->reverse, pColList, pTaskInfo);
@@ -8816,7 +8890,7 @@ SOperatorInfo* doCreateOperatorTreeNode(SPhysiNode* pPhyNode, SExecTaskInfo* pTa
     assert(size == 1);
 
     SPhysiNode*    pChildNode = (SPhysiNode*)nodesListGetNode(pPhyNode->pChildren, 0);
-    SOperatorInfo* op = doCreateOperatorTreeNode(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
+    SOperatorInfo* op = createOperatorTree(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
 
     int32_t      num = 0;
     SExprInfo*   pExprInfo = createExprInfo(((SProjectPhysiNode*)pPhyNode)->pProjections, NULL, &num);
@@ -8828,7 +8902,7 @@ SOperatorInfo* doCreateOperatorTreeNode(SPhysiNode* pPhyNode, SExecTaskInfo* pTa
 
     for (int32_t i = 0; i < size; ++i) {
       SPhysiNode*    pChildNode = (SPhysiNode*)nodesListGetNode(pPhyNode->pChildren, i);
-      SOperatorInfo* op = doCreateOperatorTreeNode(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
+      SOperatorInfo* op = createOperatorTree(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
 
       int32_t num = 0;
 
@@ -8849,7 +8923,7 @@ SOperatorInfo* doCreateOperatorTreeNode(SPhysiNode* pPhyNode, SExecTaskInfo* pTa
 
     for (int32_t i = 0; i < size; ++i) {
       SPhysiNode*    pChildNode = (SPhysiNode*)nodesListGetNode(pPhyNode->pChildren, i);
-      SOperatorInfo* op = doCreateOperatorTreeNode(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
+      SOperatorInfo* op = createOperatorTree(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
 
       SIntervalPhysiNode* pIntervalPhyNode = (SIntervalPhysiNode*)pPhyNode;
 
@@ -8858,11 +8932,12 @@ SOperatorInfo* doCreateOperatorTreeNode(SPhysiNode* pPhyNode, SExecTaskInfo* pTa
       SExprInfo*   pExprInfo = createExprInfo(pIntervalPhyNode->window.pFuncs, NULL, &num);
       SSDataBlock* pResBlock = createOutputBuf_rv1(pPhyNode->pOutputDataBlockDesc);
 
-      SInterval interval = {.interval = pIntervalPhyNode->interval,
+      SInterval interval = {.interval     = pIntervalPhyNode->interval,
-                            .sliding = pIntervalPhyNode->sliding,
+                            .sliding      = pIntervalPhyNode->sliding,
                             .intervalUnit = pIntervalPhyNode->intervalUnit,
-                            .slidingUnit = pIntervalPhyNode->slidingUnit,
+                            .slidingUnit  = pIntervalPhyNode->slidingUnit,
-                            .offset = pIntervalPhyNode->offset};
+                            .offset       = pIntervalPhyNode->offset,
+                            .precision    = TSDB_TIME_PRECISION_MILLI};
       return createIntervalOperatorInfo(op, pExprInfo, num, pResBlock, &interval, pTableGroupInfo, pTaskInfo);
     }
   } else if (QUERY_NODE_PHYSICAL_PLAN_SORT == nodeType(pPhyNode)) {
@@ -8870,7 +8945,7 @@ SOperatorInfo* doCreateOperatorTreeNode(SPhysiNode* pPhyNode, SExecTaskInfo* pTa
     assert(size == 1);
 
     SPhysiNode*    pChildNode = (SPhysiNode*)nodesListGetNode(pPhyNode->pChildren, 0);
-    SOperatorInfo* op = doCreateOperatorTreeNode(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
+    SOperatorInfo* op = createOperatorTree(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
 
     SSortPhysiNode* pSortPhyNode = (SSortPhysiNode*)pPhyNode;
 
@@ -8882,7 +8957,7 @@ SOperatorInfo* doCreateOperatorTreeNode(SPhysiNode* pPhyNode, SExecTaskInfo* pTa
     assert(size == 1);
 
     SPhysiNode*    pChildNode = (SPhysiNode*)nodesListGetNode(pPhyNode->pChildren, 0);
-    SOperatorInfo* op = doCreateOperatorTreeNode(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
+    SOperatorInfo* op = createOperatorTree(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
 
     SSessionWinodwPhysiNode* pSessionNode = (SSessionWinodwPhysiNode*)pPhyNode;
 
@@ -8898,7 +8973,7 @@ SOperatorInfo* doCreateOperatorTreeNode(SPhysiNode* pPhyNode, SExecTaskInfo* pTa
 
      for (int32_t i = 0; i < size; ++i) {
        SPhysiNode*      pChildNode = taosArrayGetP(pPhyNode->pChildren, i);
-       SOperatorInfo* op = doCreateOperatorTreeNode(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
+       SOperatorInfo* op = createOperatorTree(pChildNode, pTaskInfo, pHandle, queryId, taskId, pTableGroupInfo);
        return createMultiTableAggOperatorInfo(op, pPhyNode->pTargets, pTaskInfo, pTableGroupInfo);
      }
    }*/
@@ -9043,11 +9118,11 @@ SArray* extractColMatchInfo(SNodeList* pNodeList, SDataBlockDescNode* pOutputNod
   for (int32_t i = 0; i < num; ++i) {
     SSlotDescNode* pNode = (SSlotDescNode*)nodesListGetNode(pOutputNodeList->pSlots, i);
     SColMatchInfo* info = taosArrayGet(pList, pNode->slotId);
-    //    if (pNode->output) {
+    if (pNode->output) {
-    (*numOfOutputCols) += 1;
+      (*numOfOutputCols) += 1;
-    //    } else {
+    } else {
-    //      info->output = false;
+      info->output = false;
-    //    }
+    }
   }
 
   return pList;
@@ -9105,18 +9180,18 @@ _error:
   return NULL;
 }
 
-int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId) {
+int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId, EOPTR_EXEC_MODEL model) {
   uint64_t queryId = pPlan->id.queryId;
 
   int32_t code = TSDB_CODE_SUCCESS;
-  *pTaskInfo = createExecTaskInfo(queryId, taskId);
+  *pTaskInfo = createExecTaskInfo(queryId, taskId, model);
   if (*pTaskInfo == NULL) {
     code = TSDB_CODE_QRY_OUT_OF_MEMORY;
     goto _complete;
   }
 
   STableGroupInfo group = {0};
-  (*pTaskInfo)->pRoot = doCreateOperatorTreeNode(pPlan->pNode, *pTaskInfo, pHandle, queryId, taskId, &group);
+  (*pTaskInfo)->pRoot = createOperatorTree(pPlan->pNode, *pTaskInfo, pHandle, queryId, taskId, &group);
   if (NULL == (*pTaskInfo)->pRoot) {
     code = terrno;
     goto _complete;

source/libs/executor/test/executorTests.cpp

@@ -944,7 +944,7 @@ TEST(testCase, build_executor_tree_Test) {
   int32_t code = qStringToSubplan(msg, &plan);
   ASSERT_EQ(code, 0);
 
-  code = qCreateExecTask(&handle, 2, 1, plan, (void**) &pTaskInfo, &sinkHandle);
+  code = qCreateExecTask(&handle, 2, 1, plan, (void**) &pTaskInfo, &sinkHandle, OPTR_EXEC_MODEL_BATCH);
   ASSERT_EQ(code, 0);
 }
 

source/libs/function/src/builtins.c

@@ -315,31 +315,31 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
   {
     .name = "_qstartts",
     .type = FUNCTION_TYPE_QSTARTTS,
-    .classification = FUNC_MGT_PSEUDO_COLUMN_FUNC,
+    .classification = FUNC_MGT_PSEUDO_COLUMN_FUNC | FUNC_MGT_WINDOW_PC_FUNC,
     .checkFunc    = stubCheckAndGetResultType,
-    .getEnvFunc   = NULL,
+    .getEnvFunc   = getTimePseudoFuncEnv,
     .initFunc     = NULL,
-    .sprocessFunc = NULL,
+    .sprocessFunc = qStartTsFunction,
     .finalizeFunc = NULL
   },
   {
     .name = "_qendts",
     .type = FUNCTION_TYPE_QENDTS,
-    .classification = FUNC_MGT_PSEUDO_COLUMN_FUNC,
+    .classification = FUNC_MGT_PSEUDO_COLUMN_FUNC | FUNC_MGT_WINDOW_PC_FUNC,
     .checkFunc    = stubCheckAndGetResultType,
-    .getEnvFunc   = NULL,
+    .getEnvFunc   = getTimePseudoFuncEnv,
     .initFunc     = NULL,
-    .sprocessFunc = NULL,
+    .sprocessFunc = qEndTsFunction,
     .finalizeFunc = NULL
   },
   {
     .name = "_wstartts",
-    .type = FUNCTION_TYPE_QSTARTTS,
+    .type = FUNCTION_TYPE_WSTARTTS,
     .classification = FUNC_MGT_PSEUDO_COLUMN_FUNC | FUNC_MGT_WINDOW_PC_FUNC,
     .checkFunc    = stubCheckAndGetResultType,
-    .getEnvFunc   = NULL,
+    .getEnvFunc   = getTimePseudoFuncEnv,
     .initFunc     = NULL,
-    .sprocessFunc = NULL,
+    .sprocessFunc = winStartTsFunction,
     .finalizeFunc = NULL
   },
   {
@@ -347,9 +347,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
     .type = FUNCTION_TYPE_QENDTS,
     .classification = FUNC_MGT_PSEUDO_COLUMN_FUNC | FUNC_MGT_WINDOW_PC_FUNC,
     .checkFunc    = stubCheckAndGetResultType,
-    .getEnvFunc   = NULL,
+    .getEnvFunc   = getTimePseudoFuncEnv,
     .initFunc     = NULL,
-    .sprocessFunc = NULL,
+    .sprocessFunc = winEndTsFunction,
     .finalizeFunc = NULL
   },
   {
@@ -357,9 +357,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
     .type = FUNCTION_TYPE_WDURATION,
     .classification = FUNC_MGT_PSEUDO_COLUMN_FUNC | FUNC_MGT_WINDOW_PC_FUNC,
     .checkFunc    = stubCheckAndGetResultType,
-    .getEnvFunc   = NULL,
+    .getEnvFunc   = getTimePseudoFuncEnv,
     .initFunc     = NULL,
-    .sprocessFunc = NULL,
+    .sprocessFunc = winDurFunction,
     .finalizeFunc = NULL
   }
 };
@@ -368,6 +368,7 @@ const int32_t funcMgtBuiltinsNum = (sizeof(funcMgtBuiltins) / sizeof(SBuiltinFun
 
 int32_t stubCheckAndGetResultType(SFunctionNode* pFunc) {
   switch(pFunc->funcType) {
+    case FUNCTION_TYPE_WDURATION:
     case FUNCTION_TYPE_COUNT:
       pFunc->node.resType = (SDataType){.bytes = sizeof(int64_t), .type = TSDB_DATA_TYPE_BIGINT};
       break;
@@ -400,14 +401,18 @@ int32_t stubCheckAndGetResultType(SFunctionNode* pFunc) {
     }
     case FUNCTION_TYPE_CONCAT:
     case FUNCTION_TYPE_ROWTS:
-    case FUNCTION_TYPE_TBNAME:
+    case FUNCTION_TYPE_TBNAME: {
-    case FUNCTION_TYPE_QSTARTTS:
-    case FUNCTION_TYPE_QENDTS:
-    case FUNCTION_TYPE_WSTARTTS:
-    case FUNCTION_TYPE_WENDTS:
-    case FUNCTION_TYPE_WDURATION:
       // todo
       break;
+    }
+
+    case FUNCTION_TYPE_QENDTS:
+    case FUNCTION_TYPE_QSTARTTS:
+    case FUNCTION_TYPE_WENDTS:
+    case FUNCTION_TYPE_WSTARTTS: {
+      pFunc->node.resType = (SDataType){.bytes = sizeof(int64_t), .type = TSDB_DATA_TYPE_TIMESTAMP};
+      break;
+    }
 
     case FUNCTION_TYPE_ABS:
     case FUNCTION_TYPE_CEIL:

source/libs/function/src/functionMgt.c

@@ -92,6 +92,7 @@ int32_t fmGetScalarFuncExecFuncs(int32_t funcId, SScalarFuncExecFuncs* pFpSet) {
     return TSDB_CODE_FAILED;
   }
   pFpSet->process = funcMgtBuiltins[funcId].sprocessFunc;
+  pFpSet->getEnv  = funcMgtBuiltins[funcId].getEnvFunc;
   return TSDB_CODE_SUCCESS;
 }
 

source/libs/function/src/texpr.c

@@ -116,42 +116,6 @@ bool exprTreeApplyFilter(tExprNode *pExpr, const void *pItem, SExprTraverseSupp
   return param->nodeFilterFn(pItem, pExpr->_node.info);
 }
 
-
-
-static void exprTreeToBinaryImpl(SBufferWriter* bw, tExprNode* expr) {
-  tbufWriteUint8(bw, expr->nodeType);
-  
-  if (expr->nodeType == TEXPR_VALUE_NODE) {
-    SVariant* pVal = expr->pVal;
-    
-    tbufWriteUint32(bw, pVal->nType);
-    if (pVal->nType == TSDB_DATA_TYPE_BINARY) {
-      tbufWriteInt32(bw, pVal->nLen);
-      tbufWrite(bw, pVal->pz, pVal->nLen);
-    } else {
-      tbufWriteInt64(bw, pVal->i);
-    }
-    
-  } else if (expr->nodeType == TEXPR_COL_NODE) {
-    SSchema* pSchema = expr->pSchema;
-    tbufWriteInt16(bw, pSchema->colId);
-    tbufWriteInt16(bw, pSchema->bytes);
-    tbufWriteUint8(bw, pSchema->type);
-    tbufWriteString(bw, pSchema->name);
-    
-  } else if (expr->nodeType == TEXPR_BINARYEXPR_NODE) {
-    tbufWriteUint8(bw, expr->_node.optr);
-    exprTreeToBinaryImpl(bw, expr->_node.pLeft);
-    exprTreeToBinaryImpl(bw, expr->_node.pRight);
-  }
-}
-
-void exprTreeToBinary(SBufferWriter* bw, tExprNode* expr) {
-  if (expr != NULL) {
-    exprTreeToBinaryImpl(bw, expr);
-  }
-}
-
 // TODO: these three functions should be made global
 static void* exception_calloc(size_t nmemb, size_t size) {
   void* p = taosMemoryCalloc(nmemb, size);
@@ -230,97 +194,6 @@ tExprNode* exprTreeFromBinary(const void* data, size_t size) {
   return exprTreeFromBinaryImpl(&br);
 }
 
-tExprNode* exprTreeFromTableName(const char* tbnameCond) {
-  if (!tbnameCond) {
-    return NULL;
-  }
-
-  int32_t anchor = CLEANUP_GET_ANCHOR();
-
-  tExprNode* expr = exception_calloc(1, sizeof(tExprNode));
-  CLEANUP_PUSH_VOID_PTR_PTR(true, tExprTreeDestroy, expr, NULL);
-
-  expr->nodeType = TEXPR_BINARYEXPR_NODE;
-
-  tExprNode* left = exception_calloc(1, sizeof(tExprNode));
-  expr->_node.pLeft = left;
-
-  left->nodeType = TEXPR_COL_NODE;
-  SSchema* pSchema = exception_calloc(1, sizeof(SSchema));
-  left->pSchema = pSchema;
-
-//  *pSchema = NULL;//*tGetTbnameColumnSchema();
-
-  tExprNode* right = exception_calloc(1, sizeof(tExprNode));
-  expr->_node.pRight = right;
-
-  if (strncmp(tbnameCond, QUERY_COND_REL_PREFIX_LIKE, QUERY_COND_REL_PREFIX_LIKE_LEN) == 0) {
-    right->nodeType = TEXPR_VALUE_NODE;
-    expr->_node.optr = OP_TYPE_LIKE;
-    SVariant* pVal = exception_calloc(1, sizeof(SVariant));
-    right->pVal = pVal;
-    size_t len = strlen(tbnameCond + QUERY_COND_REL_PREFIX_LIKE_LEN) + 1;
-    pVal->pz = exception_malloc(len);
-    memcpy(pVal->pz, tbnameCond + QUERY_COND_REL_PREFIX_LIKE_LEN, len);
-    pVal->nType = TSDB_DATA_TYPE_BINARY;
-    pVal->nLen = (int32_t)len;
-
-  } else if (strncmp(tbnameCond, QUERY_COND_REL_PREFIX_MATCH, QUERY_COND_REL_PREFIX_MATCH_LEN) == 0) {
-    right->nodeType = TEXPR_VALUE_NODE;
-    expr->_node.optr = OP_TYPE_MATCH;
-    SVariant* pVal = exception_calloc(1, sizeof(SVariant));
-    right->pVal = pVal;
-    size_t len = strlen(tbnameCond + QUERY_COND_REL_PREFIX_MATCH_LEN) + 1;
-    pVal->pz = exception_malloc(len);
-    memcpy(pVal->pz, tbnameCond + QUERY_COND_REL_PREFIX_MATCH_LEN, len);
-    pVal->nType = TSDB_DATA_TYPE_BINARY;
-    pVal->nLen = (int32_t)len;
-  } else if (strncmp(tbnameCond, QUERY_COND_REL_PREFIX_NMATCH, QUERY_COND_REL_PREFIX_NMATCH_LEN) == 0) {
-    right->nodeType = TEXPR_VALUE_NODE;
-    expr->_node.optr = OP_TYPE_NMATCH;
-    SVariant* pVal = exception_calloc(1, sizeof(SVariant));
-    right->pVal = pVal;
-    size_t len = strlen(tbnameCond + QUERY_COND_REL_PREFIX_NMATCH_LEN) + 1;
-    pVal->pz = exception_malloc(len);
-    memcpy(pVal->pz, tbnameCond + QUERY_COND_REL_PREFIX_NMATCH_LEN, len);
-    pVal->nType = TSDB_DATA_TYPE_BINARY;
-    pVal->nLen = (int32_t)len;
-  } else if (strncmp(tbnameCond, QUERY_COND_REL_PREFIX_IN, QUERY_COND_REL_PREFIX_IN_LEN) == 0) {
-    right->nodeType = TEXPR_VALUE_NODE;
-    expr->_node.optr = OP_TYPE_IN;
-    SVariant* pVal = exception_calloc(1, sizeof(SVariant));
-    right->pVal = pVal;
-    pVal->nType = TSDB_DATA_TYPE_POINTER_ARRAY;
-    pVal->arr = taosArrayInit(2, POINTER_BYTES);
-
-    const char* cond = tbnameCond + QUERY_COND_REL_PREFIX_IN_LEN;
-    for (const char *e = cond; *e != 0; e++) {
-      if (*e == TS_PATH_DELIMITER[0]) {
-        cond = e + 1;
-      } else if (*e == ',') {
-        size_t len = e - cond;
-        char* p = exception_malloc(len + VARSTR_HEADER_SIZE);
-        STR_WITH_SIZE_TO_VARSTR(p, cond, (VarDataLenT)len);
-        cond += len;
-        taosArrayPush(pVal->arr, &p);
-      }
-    }
-
-    if (*cond != 0) {
-      size_t len = strlen(cond) + VARSTR_HEADER_SIZE;
-      
-      char* p = exception_malloc(len);
-      STR_WITH_SIZE_TO_VARSTR(p, cond, (VarDataLenT)(len - VARSTR_HEADER_SIZE));
-      taosArrayPush(pVal->arr, &p);
-    }
-
-    taosArraySortString(pVal->arr, taosArrayCompareString);
-  }
-
-  CLEANUP_EXECUTE_TO(anchor, false);
-  return expr;
-}
-
 void buildFilterSetFromBinary(void **q, const char *buf, int32_t len) {
   SBufferReader br = tbufInitReader(buf, len, false); 
   uint32_t type  = tbufReadUint32(&br);     

source/libs/function/src/tunaryoperator.c

@@ -1,13 +0,0 @@
-#include "tunaryoperator.h"
-
-
-
-
-// TODO dynamic define these functions
-//_unary_scalar_fn_t getUnaryScalarOperatorFn(int32_t operator) {
-//  assert(0);
-//}
-
-//bool isStringOperatorFn(int32_t op) {
-//  return op == FUNCTION_LENGTH;
-//}

source/libs/index/inc/indexFstDfa.h

@@ -0,0 +1,74 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __INDEX_FST_DFA_H__
+#define __INDEX_FST_DFA_H__
+
+#include "indexFstRegex.h"
+#include "indexFstSparse.h"
+#include "tarray.h"
+#include "thash.h"
+
+#ifdef __cplusplus
+
+extern "C" {
+#endif
+
+typedef struct FstDfa FstDfa;
+
+typedef struct {
+  SArray * insts;
+  uint32_t next[256];
+  bool     isMatch;
+} State;
+
+/*
+ * dfa builder related func
+ **/
+typedef struct FstDfaBuilder {
+  FstDfa *  dfa;
+  SHashObj *cache;
+} FstDfaBuilder;
+
+FstDfaBuilder *dfaBuilderCreate(SArray *insts);
+
+void dfaBuilderDestroy(FstDfaBuilder *builder);
+
+FstDfa *dfaBuilderBuild(FstDfaBuilder *builder);
+
+bool dfaBuilderRunState(FstDfaBuilder *builder, FstSparseSet *cur, FstSparseSet *next, uint32_t state, uint8_t bytes,
+                        uint32_t *result);
+
+bool dfaBuilderCachedState(FstDfaBuilder *builder, FstSparseSet *set, uint32_t *result);
+
+/*
+ * dfa related func
+ **/
+typedef struct FstDfa {
+  SArray *insts;
+  SArray *states;
+} FstDfa;
+
+FstDfa *dfaCreate(SArray *insts, SArray *states);
+bool    dfaIsMatch(FstDfa *dfa, uint32_t si);
+bool    dfaAccept(FstDfa *dfa, uint32_t si, uint8_t byte, uint32_t *result);
+void    dfaAdd(FstDfa *dfa, FstSparseSet *set, uint32_t ip);
+bool    dfaRun(FstDfa *dfa, FstSparseSet *from, FstSparseSet *to, uint8_t byte);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

source/libs/index/inc/indexFstRegex.h

@@ -0,0 +1,74 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _TD_INDEX_FST_REGEX_H_
+#define _TD_INDEX_FST_REGEX_H_
+
+//#include "indexFstDfa.h"
+#include "taos.h"
+#include "tarray.h"
+#include "tchecksum.h"
+#include "thash.h"
+#include "tlog.h"
+#include "tutil.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef enum { MATCH, JUMP, SPLIT, RANGE } InstType;
+
+typedef struct MatchValue {
+} MatchValue;
+typedef struct JumpValue {
+  uint32_t step;
+} JumpValue;
+
+typedef struct SplitValue {
+  uint32_t len1;
+  uint32_t len2;
+} SplitValue;
+
+typedef struct RangeValue {
+  uint8_t start;
+  uint8_t end;
+} RangeValue;
+
+typedef struct {
+  InstType ty;
+  union {
+    MatchValue mv;
+    JumpValue  jv;
+    SplitValue sv;
+    RangeValue rv;
+  };
+} Inst;
+
+typedef struct {
+  char *orig;
+  void *dfa;
+} FstRegex;
+
+FstRegex *regexCreate(const char *str);
+
+void regexSetup(FstRegex *regex, uint32_t size, const char *str);
+
+// uint32_t regexStart()
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

source/libs/index/inc/indexFstSparse.h

@@ -13,8 +13,8 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef _TD_INDEX_SPARSE_H_
+#ifndef _TD_INDEX_FST_SPARSE_H_
-#define _TD_INDEX_SPARSE_H_
+#define _TD_INDEX_FST_SPARSE_H_
 
 #include "tarray.h"
 

source/libs/index/src/indexFstDfa.c

@@ -0,0 +1,218 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "indexFstDfa.h"
+#include "thash.h"
+
+const static uint32_t STATE_LIMIT = 1000;
+
+static int dfaInstsEqual(const void *a, const void *b, size_t size) {
+  SArray *ar = (SArray *)a;
+  SArray *br = (SArray *)b;
+  size_t  al = ar != NULL ? taosArrayGetSize(ar) : 0;
+  size_t  bl = br != NULL ? taosArrayGetSize(br) : 0;
+  if (al != bl) {
+    return -1;
+  }
+  for (int i = 0; i < al; i++) {
+    uint32_t v1 = *(uint32_t *)taosArrayGet(ar, i);
+    uint32_t v2 = *(uint32_t *)taosArrayGet(br, i);
+    if (v1 != v2) {
+      return -1;
+    }
+  }
+  return 0;
+}
+FstDfaBuilder *dfaBuilderCreate(SArray *insts) {
+  FstDfaBuilder *builder = taosMemoryCalloc(1, sizeof(FstDfaBuilder));
+  if (builder == NULL) {
+    return NULL;
+  }
+
+  SArray *states = taosArrayInit(4, sizeof(State));
+
+  builder->dfa = dfaCreate(insts, states);
+  builder->cache = taosHashInit(
+      4, taosGetDefaultHashFunction(POINTER_BYTES == sizeof(int64_t) ? TSDB_DATA_TYPE_BIGINT : TSDB_DATA_TYPE_INT),
+      false, HASH_NO_LOCK);
+  taosHashSetEqualFp(builder->cache, dfaInstsEqual);
+  return builder;
+}
+void dfaBuilderDestroy(FstDfaBuilder *builder) {
+  if (builder == NULL) {
+    return;
+  }
+  void *pIter = builder->cache != NULL ? taosHashIterate(builder->cache, NULL) : NULL;
+  while (pIter) {
+    SArray **key = pIter;
+    taosArrayDestroy(*key);
+    pIter = taosHashIterate(builder->cache, pIter);
+  }
+  taosHashCleanup(builder->cache);
+}
+
+FstDfa *dfaBuilderBuild(FstDfaBuilder *builder) {
+  uint32_t      sz = taosArrayGetSize(builder->dfa->insts);
+  FstSparseSet *cur = sparSetCreate(sz);
+  FstSparseSet *nxt = sparSetCreate(sz);
+
+  dfaAdd(builder->dfa, cur, 0);
+
+  SArray * states = taosArrayInit(0, sizeof(uint32_t));
+  uint32_t result;
+  if (dfaBuilderCachedState(builder, cur, &result)) {
+    taosArrayPush(states, &result);
+  }
+  SHashObj *seen = taosHashInit(12, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), false, HASH_NO_LOCK);
+  while (taosArrayGetSize(states) != 0) {
+    result = *(uint32_t *)taosArrayPop(states);
+    for (int i = 0; i < 256; i++) {
+      uint32_t ns, dummpy = 0;
+      if (dfaBuilderRunState(builder, cur, nxt, result, i, &ns)) {
+        if (taosHashGet(seen, &ns, sizeof(ns)) == NULL) {
+          taosHashPut(seen, &ns, sizeof(ns), &dummpy, sizeof(dummpy));
+          taosArrayPush(states, &ns);
+        }
+      }
+      if (taosArrayGetSize(builder->dfa->states) > STATE_LIMIT) {
+        // Too many state;
+        //
+      }
+    }
+  }
+  taosArrayDestroy(states);
+  taosHashCleanup(seen);
+  return builder->dfa;
+}
+
+bool dfaBuilderRunState(FstDfaBuilder *builder, FstSparseSet *cur, FstSparseSet *next, uint32_t state, uint8_t byte,
+                        uint32_t *result) {
+  sparSetClear(cur);
+  State *t = taosArrayGet(builder->dfa->states, state);
+  for (int i = 0; i < taosArrayGetSize(t->insts); i++) {
+    uint32_t ip = *(int32_t *)taosArrayGet(t->insts, i);
+    sparSetAdd(cur, ip);
+  }
+  dfaRun(builder->dfa, cur, next, byte);
+
+  t = taosArrayGet(builder->dfa->states, state);
+
+  uint32_t nxtState;
+  if (dfaBuilderCachedState(builder, next, &nxtState)) {
+    t->next[byte] = nxtState;
+    *result = nxtState;
+    return true;
+  }
+  return false;
+}
+
+bool dfaBuilderCachedState(FstDfaBuilder *builder, FstSparseSet *set, uint32_t *result) {
+  SArray *tinsts = taosArrayInit(4, sizeof(uint32_t));
+  bool    isMatch = false;
+
+  for (int i = 0; i < sparSetLen(set); i++) {
+    uint32_t ip = sparSetGet(set, i);
+
+    Inst *inst = taosArrayGet(builder->dfa->insts, ip);
+    if (inst->ty == JUMP || inst->ty == SPLIT) {
+      continue;
+    } else if (inst->ty == RANGE) {
+      taosArrayPush(tinsts, &ip);
+    } else if (inst->ty == MATCH) {
+      isMatch = true;
+      taosArrayPush(tinsts, &ip);
+    }
+  }
+  if (taosArrayGetSize(tinsts) == 0) {
+    return false;
+  }
+  uint32_t *v = taosHashGet(builder->cache, &tinsts, sizeof(POINTER_BYTES));
+  if (v != NULL) {
+    *result = *v;
+    taosArrayDestroy(tinsts);
+  } else {
+    State st;
+    st.insts = tinsts;
+    st.isMatch = isMatch;
+    taosArrayPush(builder->dfa->states, &st);
+    int32_t sz = taosArrayGetSize(builder->dfa->states) - 1;
+    taosHashPut(builder->cache, &tinsts, sizeof(POINTER_BYTES), &sz, sizeof(sz));
+    *result = sz;
+  }
+  return true;
+}
+
+FstDfa *dfaCreate(SArray *insts, SArray *states) {
+  FstDfa *dfa = taosMemoryCalloc(1, sizeof(FstDfa));
+  if (dfa == NULL) {
+    return NULL;
+  }
+
+  dfa->insts = insts;
+  dfa->states = states;
+  return dfa;
+}
+bool dfaIsMatch(FstDfa *dfa, uint32_t si) {
+  if (dfa->states == NULL || si < taosArrayGetSize(dfa->states)) {
+    return false;
+  }
+  State *st = taosArrayGet(dfa->states, si);
+  return st != NULL ? st->isMatch : false;
+}
+bool dfaAccept(FstDfa *dfa, uint32_t si, uint8_t byte, uint32_t *result) {
+  if (dfa->states == NULL || si < taosArrayGetSize(dfa->states)) {
+    return false;
+  }
+  State *st = taosArrayGet(dfa->states, si);
+  *result = st->next[byte];
+  return true;
+}
+void dfaAdd(FstDfa *dfa, FstSparseSet *set, uint32_t ip) {
+  if (sparSetContains(set, ip)) {
+    return;
+  }
+  sparSetAdd(set, ip);
+  Inst *inst = taosArrayGet(dfa->insts, ip);
+  if (inst->ty == MATCH || inst->ty == RANGE) {
+    // do nothing
+  } else if (inst->ty == JUMP) {
+    dfaAdd(dfa, set, inst->jv.step);
+  } else if (inst->ty == SPLIT) {
+    dfaAdd(dfa, set, inst->sv.len1);
+    dfaAdd(dfa, set, inst->sv.len2);
+  }
+
+  return;
+}
+bool dfaRun(FstDfa *dfa, FstSparseSet *from, FstSparseSet *to, uint8_t byte) {
+  bool isMatch = false;
+  sparSetClear(to);
+  for (int i = 0; i < sparSetLen(from); i++) {
+    uint32_t ip = sparSetGet(from, i);
+
+    Inst *inst = taosArrayGet(dfa->insts, ip);
+    if (inst->ty == JUMP || inst->ty == SPLIT) {
+      continue;
+    } else if (inst->ty == MATCH) {
+      isMatch = true;
+    } else if (inst->ty == RANGE) {
+      if (inst->rv.start <= byte && byte <= inst->rv.end) {
+        dfaAdd(dfa, to, ip + 1);
+      }
+    }
+  }
+
+  return isMatch;
+}

source/libs/index/src/indexFstRegex.c

@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "indexFstRegex.h"
+#include "indexFstSparse.h"
+
+FstRegex *regexCreate(const char *str) {
+  FstRegex *regex = taosMemoryCalloc(1, sizeof(FstRegex));
+  if (regex == NULL) {
+    return NULL;
+  }
+  int32_t sz = (int32_t)strlen(str);
+  char *  orig = taosMemoryCalloc(1, sz);
+  memcpy(orig, str, sz);
+
+  regex->orig = orig;
+}
+
+void regexSetup(FstRegex *regex, uint32_t size, const char *str) {
+  // return
+  // return;
+}

source/libs/index/src/indexSparse.c

@@ -13,7 +13,7 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "indexSparse.h"
+#include "indexFstSparse.h"
 
 FstSparseSet *sparSetCreate(int32_t sz) {
   FstSparseSet *ss = taosMemoryCalloc(1, sizeof(FstSparseSet));

source/libs/nodes/src/nodesCodeFuncs.c

@@ -1057,6 +1057,7 @@ static const char* jkIntervalPhysiPlanIntervalUnit = "intervalUnit";
 static const char* jkIntervalPhysiPlanSlidingUnit  = "slidingUnit";
 static const char* jkIntervalPhysiPlanFill = "Fill";
 static const char* jkIntervalPhysiPlanTsPk = "TsPk";
+static const char* jkIntervalPhysiPlanPrecision = "Precision";
 
 static int32_t physiIntervalNodeToJson(const void* pObj, SJson* pJson) {
   const SIntervalPhysiNode* pNode = (const SIntervalPhysiNode*)pObj;
@@ -1083,6 +1084,9 @@ static int32_t physiIntervalNodeToJson(const void* pObj, SJson* pJson) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonAddObject(pJson, jkIntervalPhysiPlanTsPk, nodeToJson, pNode->pTspk);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddIntegerToObject(pJson, jkIntervalPhysiPlanPrecision, pNode->precision);
+  }
 
   return code;
 }
@@ -1112,6 +1116,9 @@ static int32_t jsonToPhysiIntervalNode(const SJson* pJson, void* pObj) {
   if (TSDB_CODE_SUCCESS == code) {
     code = jsonToNodeObject(pJson, jkIntervalPhysiPlanTsPk, (SNode**)&pNode->pTspk);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonGetUTinyIntValue(pJson, jkIntervalPhysiPlanPrecision, &pNode->precision);
+  }
 
   return code;
 }

source/libs/nodes/src/nodesUtilFuncs.c

@@ -82,6 +82,8 @@ SNodeptr nodesMakeNode(ENodeType type) {
       return makeNode(type, sizeof(STableOptions));
     case QUERY_NODE_INDEX_OPTIONS:
       return makeNode(type, sizeof(SIndexOptions));
+    case QUERY_NODE_EXPLAIN_OPTIONS:
+      return makeNode(type, sizeof(SExplainOptions));
     case QUERY_NODE_SET_OPERATOR:
       return makeNode(type, sizeof(SSetOperator));
     case QUERY_NODE_SELECT_STMT:
@@ -132,6 +134,8 @@ SNodeptr nodesMakeNode(ENodeType type) {
       return makeNode(type, sizeof(SCreateTopicStmt));
     case QUERY_NODE_DROP_TOPIC_STMT:
       return makeNode(type, sizeof(SDropTopicStmt));
+    case QUERY_NODE_EXPLAIN_STMT:
+      return makeNode(type, sizeof(SExplainStmt));
     case QUERY_NODE_SHOW_DATABASES_STMT:
     case QUERY_NODE_SHOW_TABLES_STMT:
     case QUERY_NODE_SHOW_STABLES_STMT:
@@ -204,49 +208,159 @@ SNodeptr nodesMakeNode(ENodeType type) {
   return NULL;
 }
 
-static EDealRes destroyNode(SNode** pNode, void* pContext) {
+static void destroyVgDataBlockArray(SArray* pArray) {
-  switch (nodeType(*pNode)) {
+  size_t size = taosArrayGetSize(pArray);
-    case QUERY_NODE_VALUE: {
+  for (size_t i = 0; i < size; ++i) {
-      SValueNode* pValue = (SValueNode*)*pNode;
+    SVgDataBlocks* pVg = taosArrayGetP(pArray, i);
+    taosMemoryFreeClear(pVg->pData);
+    taosMemoryFreeClear(pVg);
+  }
+  taosArrayDestroy(pArray);
+}
 
+static void destroyLogicNode(SLogicNode* pNode) {
+  nodesDestroyList(pNode->pChildren);
+  nodesDestroyNode(pNode->pConditions);
+  nodesDestroyList(pNode->pTargets);
+}
+
+static void destroyPhysiNode(SPhysiNode* pNode) {
+  nodesDestroyList(pNode->pChildren);
+  nodesDestroyNode(pNode->pConditions);
+  nodesDestroyNode(pNode->pOutputDataBlockDesc);
+}
+
+static void destroyWinodwPhysiNode(SWinodwPhysiNode* pNode) {
+  destroyPhysiNode((SPhysiNode*)pNode);
+  nodesDestroyList(pNode->pExprs);
+  nodesDestroyList(pNode->pFuncs);
+}
+
+static void destroyScanPhysiNode(SScanPhysiNode* pNode) {
+  destroyPhysiNode((SPhysiNode*)pNode);
+  nodesDestroyList(pNode->pScanCols);
+}
+
+static void destroyDataSinkNode(SDataSinkNode* pNode) {
+  nodesDestroyNode(pNode->pInputDataBlockDesc);
+}
+
+void nodesDestroyNode(SNodeptr pNode) {
+  if (NULL == pNode) {
+    return;
+  }
+
+  switch (nodeType(pNode)) {
+    case QUERY_NODE_COLUMN: // pProjectRef is weak reference, no need to release
+      break;
+    case QUERY_NODE_VALUE: {
+      SValueNode* pValue = (SValueNode*)pNode;
       taosMemoryFreeClear(pValue->literal);
       if (IS_VAR_DATA_TYPE(pValue->node.resType.type)) {
         taosMemoryFreeClear(pValue->datum.p);
       }
-      
+      break;
+    }
+    case QUERY_NODE_OPERATOR: {
+      SOperatorNode* pOp = (SOperatorNode*)pNode;
+      nodesDestroyNode(pOp->pLeft);
+      nodesDestroyNode(pOp->pRight);
       break;
     }
     case QUERY_NODE_LOGIC_CONDITION:
-      nodesClearList(((SLogicConditionNode*)(*pNode))->pParameterList);
+      nodesDestroyList(((SLogicConditionNode*)pNode)->pParameterList);
       break;
     case QUERY_NODE_FUNCTION:
-      nodesClearList(((SFunctionNode*)(*pNode))->pParameterList);
+      nodesDestroyList(((SFunctionNode*)pNode)->pParameterList);
       break;
     case QUERY_NODE_REAL_TABLE: {
-      SRealTableNode* pReal = (SRealTableNode*)*pNode;
+      SRealTableNode* pReal = (SRealTableNode*)pNode;
       taosMemoryFreeClear(pReal->pMeta);
       taosMemoryFreeClear(pReal->pVgroupList);
       break;
     }
     case QUERY_NODE_TEMP_TABLE:
-      nodesDestroyNode(((STempTableNode*)(*pNode))->pSubquery);
+      nodesDestroyNode(((STempTableNode*)pNode)->pSubquery);
       break;
+    case QUERY_NODE_JOIN_TABLE: {
+      SJoinTableNode* pJoin = (SJoinTableNode*)pNode;
+      nodesDestroyNode(pJoin->pLeft);
+      nodesDestroyNode(pJoin->pRight);
+      nodesDestroyNode(pJoin->pOnCond);
+      break;
+    }
     case QUERY_NODE_GROUPING_SET:
-      nodesClearList(((SGroupingSetNode*)(*pNode))->pParameterList);
+      nodesDestroyList(((SGroupingSetNode*)pNode)->pParameterList);
       break;
+    case QUERY_NODE_ORDER_BY_EXPR:
+      nodesDestroyNode(((SOrderByExprNode*)pNode)->pExpr);
+      break;
+    case QUERY_NODE_LIMIT: // no pointer field
+      break;
+    case QUERY_NODE_STATE_WINDOW:
+      nodesDestroyNode(((SStateWindowNode*)pNode)->pCol);
+      break;
+    case QUERY_NODE_SESSION_WINDOW: {
+      SSessionWindowNode* pSession = (SSessionWindowNode*)pNode;
+      nodesDestroyNode(pSession->pCol);
+      nodesDestroyNode(pSession->pGap);
+      break;
+    }
+    case QUERY_NODE_INTERVAL_WINDOW: {
+      SIntervalWindowNode* pJoin = (SIntervalWindowNode*)pNode;
+      nodesDestroyNode(pJoin->pCol);
+      nodesDestroyNode(pJoin->pInterval);
+      nodesDestroyNode(pJoin->pOffset);
+      nodesDestroyNode(pJoin->pSliding);
+      nodesDestroyNode(pJoin->pFill);
+      break;
+    }
     case QUERY_NODE_NODE_LIST:
-      nodesClearList(((SNodeListNode*)(*pNode))->pNodeList);
+      nodesDestroyList(((SNodeListNode*)pNode)->pNodeList);
       break;
+    case QUERY_NODE_FILL:
+      nodesDestroyNode(((SFillNode*)pNode)->pValues);
+      break;
+    case QUERY_NODE_RAW_EXPR:
+      nodesDestroyNode(((SRawExprNode*)pNode)->pNode);
+      break;
+    case QUERY_NODE_TARGET:
+      nodesDestroyNode(((STargetNode*)pNode)->pExpr);
+      break;
+    case QUERY_NODE_DATABLOCK_DESC:
+      nodesDestroyList(((SDataBlockDescNode*)pNode)->pSlots);
+      break;
+    case QUERY_NODE_SLOT_DESC:         // no pointer field
+    case QUERY_NODE_COLUMN_DEF:        // no pointer field
+    case QUERY_NODE_DOWNSTREAM_SOURCE: // no pointer field
+      break;
+    case QUERY_NODE_DATABASE_OPTIONS:
+      nodesDestroyList(((SDatabaseOptions*)pNode)->pRetentions);
+      break;
+    case QUERY_NODE_TABLE_OPTIONS: {
+      STableOptions* pStmt = (STableOptions*)pNode;
+      nodesDestroyList(pStmt->pSma);
+      nodesDestroyList(pStmt->pFuncs);
+      break;
+    }
     case QUERY_NODE_INDEX_OPTIONS: {
-      SIndexOptions* pStmt = (SIndexOptions*)*pNode;
+      SIndexOptions* pStmt = (SIndexOptions*)pNode;
       nodesDestroyList(pStmt->pFuncs);
       nodesDestroyNode(pStmt->pInterval);
       nodesDestroyNode(pStmt->pOffset);
       nodesDestroyNode(pStmt->pSliding);
       break;
     }
+    case QUERY_NODE_SET_OPERATOR: {
+      SSetOperator* pStmt = (SSetOperator*)pNode;
+      nodesDestroyNode(pStmt->pLeft);
+      nodesDestroyNode(pStmt->pRight);
+      nodesDestroyList(pStmt->pOrderByList);
+      nodesDestroyNode(pStmt->pLimit);
+      break;
+    }
     case QUERY_NODE_SELECT_STMT: {
-      SSelectStmt* pStmt = (SSelectStmt*)*pNode;
+      SSelectStmt* pStmt = (SSelectStmt*)pNode;
       nodesDestroyList(pStmt->pProjectionList);
       nodesDestroyNode(pStmt->pFromTable);
       nodesDestroyNode(pStmt->pWhere);
@@ -259,50 +373,255 @@ static EDealRes destroyNode(SNode** pNode, void* pContext) {
       nodesDestroyNode(pStmt->pSlimit);
       break;
     }
-    case QUERY_NODE_VNODE_MODIF_STMT: {
+    case QUERY_NODE_VNODE_MODIF_STMT:
-      SVnodeModifOpStmt* pStmt = (SVnodeModifOpStmt*)*pNode;
+      destroyVgDataBlockArray(((SVnodeModifOpStmt*)pNode)->pDataBlocks);
-      size_t size = taosArrayGetSize(pStmt->pDataBlocks);
+      break;
-      for (size_t i = 0; i < size; ++i) {
+    case QUERY_NODE_CREATE_DATABASE_STMT:
-        SVgDataBlocks* pVg = taosArrayGetP(pStmt->pDataBlocks, i);
+      nodesDestroyNode(((SCreateDatabaseStmt*)pNode)->pOptions);
-        taosMemoryFreeClear(pVg->pData);
+      break;
-        taosMemoryFreeClear(pVg);
+    case QUERY_NODE_DROP_DATABASE_STMT: // no pointer field
-      }
+      break;
-      taosArrayDestroy(pStmt->pDataBlocks);
+    case QUERY_NODE_ALTER_DATABASE_STMT:
+      nodesDestroyNode(((SAlterDatabaseStmt*)pNode)->pOptions);
       break;
-    }
     case QUERY_NODE_CREATE_TABLE_STMT: {
-      SCreateTableStmt* pStmt = (SCreateTableStmt*)*pNode;
+      SCreateTableStmt* pStmt = (SCreateTableStmt*)pNode;
       nodesDestroyList(pStmt->pCols);
       nodesDestroyList(pStmt->pTags);
+      nodesDestroyNode(pStmt->pOptions);
       break;
     }
     case QUERY_NODE_CREATE_SUBTABLE_CLAUSE: {
-      SCreateSubTableClause* pStmt = (SCreateSubTableClause*)*pNode;
+      SCreateSubTableClause* pStmt = (SCreateSubTableClause*)pNode;
       nodesDestroyList(pStmt->pSpecificTags);
       nodesDestroyList(pStmt->pValsOfTags);
       break;
     }
     case QUERY_NODE_CREATE_MULTI_TABLE_STMT:
-      nodesDestroyList(((SCreateMultiTableStmt*)(*pNode))->pSubTables);
+      nodesDestroyList(((SCreateMultiTableStmt*)pNode)->pSubTables);
+      break;
+    case QUERY_NODE_DROP_TABLE_CLAUSE: // no pointer field
+      break;
+    case QUERY_NODE_DROP_TABLE_STMT:
+      nodesDestroyNode(((SDropTableStmt*)pNode)->pTables);
+      break;
+    case QUERY_NODE_DROP_SUPER_TABLE_STMT: // no pointer field
+      break;
+    case QUERY_NODE_ALTER_TABLE_STMT: {
+      SAlterTableStmt* pStmt = (SAlterTableStmt*)pNode;
+      nodesDestroyNode(pStmt->pOptions);
+      nodesDestroyNode(pStmt->pVal);
+      break;
+    }
+    case QUERY_NODE_CREATE_USER_STMT:  // no pointer field
+    case QUERY_NODE_ALTER_USER_STMT:   // no pointer field
+    case QUERY_NODE_DROP_USER_STMT:    // no pointer field
+    case QUERY_NODE_USE_DATABASE_STMT: // no pointer field
+    case QUERY_NODE_CREATE_DNODE_STMT: // no pointer field
+    case QUERY_NODE_DROP_DNODE_STMT:   // no pointer field
+    case QUERY_NODE_ALTER_DNODE_STMT:  // no pointer field
       break;
     case QUERY_NODE_CREATE_INDEX_STMT: {
-      SCreateIndexStmt* pStmt = (SCreateIndexStmt*)*pNode;
+      SCreateIndexStmt* pStmt = (SCreateIndexStmt*)pNode;
       nodesDestroyNode(pStmt->pOptions);
       nodesDestroyList(pStmt->pCols);
       break;
     }
+    case QUERY_NODE_DROP_INDEX_STMT:   // no pointer field
+    case QUERY_NODE_CREATE_QNODE_STMT: // no pointer field
+    case QUERY_NODE_DROP_QNODE_STMT:   // no pointer field
+      break;
+    case QUERY_NODE_CREATE_TOPIC_STMT:
+      nodesDestroyNode(((SCreateTopicStmt*)pNode)->pQuery);
+      break;
+    case QUERY_NODE_DROP_TOPIC_STMT:  // no pointer field
+    case QUERY_NODE_ALTER_LOCAL_STMT: // no pointer field
+      break;
+    case QUERY_NODE_SHOW_DATABASES_STMT:
+    case QUERY_NODE_SHOW_TABLES_STMT:
+    case QUERY_NODE_SHOW_STABLES_STMT:
+    case QUERY_NODE_SHOW_USERS_STMT:
+    case QUERY_NODE_SHOW_DNODES_STMT:
+    case QUERY_NODE_SHOW_VGROUPS_STMT:
+    case QUERY_NODE_SHOW_MNODES_STMT:
+    case QUERY_NODE_SHOW_MODULES_STMT:
+    case QUERY_NODE_SHOW_QNODES_STMT:
+    case QUERY_NODE_SHOW_FUNCTIONS_STMT:
+    case QUERY_NODE_SHOW_INDEXES_STMT:
+    case QUERY_NODE_SHOW_STREAMS_STMT: {
+      SShowStmt* pStmt = (SShowStmt*)pNode;
+      nodesDestroyNode(pStmt->pDbName);
+      nodesDestroyNode(pStmt->pTbNamePattern);
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN_SCAN: {
+      SScanLogicNode* pLogicNode = (SScanLogicNode*)pNode;
+      destroyLogicNode((SLogicNode*)pLogicNode);
+      nodesDestroyList(pLogicNode->pScanCols);
+      taosMemoryFreeClear(pLogicNode->pMeta);
+      taosMemoryFreeClear(pLogicNode->pVgroupList);
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN_JOIN: {
+      SJoinLogicNode* pLogicNode = (SJoinLogicNode*)pNode;
+      destroyLogicNode((SLogicNode*)pLogicNode);
+      nodesDestroyNode(pLogicNode->pOnConditions);
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN_AGG: {
+      SAggLogicNode* pLogicNode = (SAggLogicNode*)pNode;
+      destroyLogicNode((SLogicNode*)pLogicNode);
+      nodesDestroyList(pLogicNode->pAggFuncs);
+      nodesDestroyList(pLogicNode->pGroupKeys);
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN_PROJECT: {
+      SProjectLogicNode* pLogicNode = (SProjectLogicNode*)pNode;
+      destroyLogicNode((SLogicNode*)pLogicNode);
+      nodesDestroyList(pLogicNode->pProjections);
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN_VNODE_MODIF: {
+      SVnodeModifLogicNode* pLogicNode = (SVnodeModifLogicNode*)pNode;
+      destroyLogicNode((SLogicNode*)pLogicNode);
+      destroyVgDataBlockArray(pLogicNode->pDataBlocks);
+      // pVgDataBlocks is weak reference
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN_EXCHANGE:
+      destroyLogicNode((SLogicNode*)pNode);
+      break;
+    case QUERY_NODE_LOGIC_PLAN_WINDOW: {
+      SWindowLogicNode* pLogicNode = (SWindowLogicNode*)pNode;
+      destroyLogicNode((SLogicNode*)pLogicNode);
+      nodesDestroyList(pLogicNode->pFuncs);
+      nodesDestroyNode(pLogicNode->pFill);
+      nodesDestroyNode(pLogicNode->pTspk);
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN_SORT: {
+      SSortLogicNode* pLogicNode = (SSortLogicNode*)pNode;
+      destroyLogicNode((SLogicNode*)pLogicNode);
+      nodesDestroyList(pLogicNode->pSortKeys);
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN_PARTITION: {
+      SPartitionLogicNode* pLogicNode = (SPartitionLogicNode*)pNode;
+      destroyLogicNode((SLogicNode*)pLogicNode);
+      nodesDestroyList(pLogicNode->pPartitionKeys);
+      break;
+    }
+    case QUERY_NODE_LOGIC_SUBPLAN: {
+      SLogicSubplan* pSubplan = (SLogicSubplan*)pNode;
+      nodesDestroyList(pSubplan->pChildren);
+      nodesDestroyNode(pSubplan->pNode);
+      nodesClearList(pSubplan->pParents);
+      taosMemoryFreeClear(pSubplan->pVgroupList);
+      break;
+    }
+    case QUERY_NODE_LOGIC_PLAN:
+      nodesDestroyList(((SQueryLogicPlan*)pNode)->pTopSubplans);
+      break;
+    case QUERY_NODE_PHYSICAL_PLAN_TAG_SCAN:
+      destroyScanPhysiNode((SScanPhysiNode*)pNode);
+      break;
+    case QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN:
+      destroyScanPhysiNode((SScanPhysiNode*)pNode);
+      break;
+    case QUERY_NODE_PHYSICAL_PLAN_TABLE_SEQ_SCAN:
+      destroyScanPhysiNode((SScanPhysiNode*)pNode);
+      break;
+    case QUERY_NODE_PHYSICAL_PLAN_STREAM_SCAN:
+      destroyScanPhysiNode((SScanPhysiNode*)pNode);
+      break;
+    case QUERY_NODE_PHYSICAL_PLAN_SYSTABLE_SCAN:
+      destroyScanPhysiNode((SScanPhysiNode*)pNode);
+      break;
+    case QUERY_NODE_PHYSICAL_PLAN_PROJECT: {
+      SProjectPhysiNode* pPhyNode = (SProjectPhysiNode*)pNode;
+      destroyPhysiNode((SPhysiNode*)pPhyNode);
+      nodesDestroyList(pPhyNode->pProjections);
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_JOIN: {
+      SJoinPhysiNode* pPhyNode = (SJoinPhysiNode*)pNode;
+      destroyPhysiNode((SPhysiNode*)pPhyNode);
+      nodesDestroyNode(pPhyNode->pOnConditions);
+      nodesDestroyList(pPhyNode->pTargets);
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_AGG: {
+      SAggPhysiNode* pPhyNode = (SAggPhysiNode*)pNode;
+      destroyPhysiNode((SPhysiNode*)pPhyNode);
+      nodesDestroyList(pPhyNode->pExprs);
+      nodesDestroyList(pPhyNode->pAggFuncs);
+      nodesDestroyList(pPhyNode->pGroupKeys);
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_EXCHANGE: {
+      SExchangePhysiNode* pPhyNode = (SExchangePhysiNode*)pNode;
+      destroyPhysiNode((SPhysiNode*)pPhyNode);
+      nodesDestroyList(pPhyNode->pSrcEndPoints);
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_SORT: {
+      SSortPhysiNode* pPhyNode = (SSortPhysiNode*)pNode;
+      destroyPhysiNode((SPhysiNode*)pPhyNode);
+      nodesDestroyNode(pPhyNode->pExprs);
+      nodesDestroyNode(pPhyNode->pSortKeys);
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_INTERVAL: {
+      SIntervalPhysiNode* pPhyNode = (SIntervalPhysiNode*)pNode;
+      destroyWinodwPhysiNode((SWinodwPhysiNode*)pPhyNode);
+      nodesDestroyNode(pPhyNode->pFill);
+      nodesDestroyNode(pPhyNode->pTspk);
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN_SESSION_WINDOW:
+      destroyWinodwPhysiNode((SWinodwPhysiNode*)pNode);
+      break;
+    case QUERY_NODE_PHYSICAL_PLAN_DISPATCH:
+      destroyDataSinkNode((SDataSinkNode*)pNode);
+      break;
+    case QUERY_NODE_PHYSICAL_PLAN_INSERT: {
+      SDataInserterNode* pSink = (SDataInserterNode*)pNode;
+      destroyDataSinkNode((SDataSinkNode*)pSink);
+      taosMemoryFreeClear(pSink->pData);
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_SUBPLAN: {
+      SSubplan* pSubplan = (SSubplan*)pNode;
+      nodesDestroyList(pSubplan->pChildren);
+      nodesDestroyNode(pSubplan->pNode);
+      nodesDestroyNode(pSubplan->pDataSink);
+      nodesClearList(pSubplan->pParents);
+      break;
+    }
+    case QUERY_NODE_PHYSICAL_PLAN: {
+      SQueryPlan* pPlan = (SQueryPlan*)pNode;
+      if (NULL != pPlan->pSubplans) {
+        // only need to destroy the top-level subplans, because they will recurse to all the subplans below
+        bool first = true;
+        SNode* pElement = NULL;
+        FOREACH(pElement, pPlan->pSubplans) {
+          if (first) {
+            first = false;
+            nodesDestroyNode(pElement);
+          } else {
+            nodesClearList(((SNodeListNode*)pElement)->pNodeList);
+            taosMemoryFreeClear(pElement);
+          }
+        }
+        nodesClearList(pPlan->pSubplans);
+      }
+      break;
+    }
     default:
       break;
   }
-  taosMemoryFreeClear(*pNode);
+  taosMemoryFreeClear(pNode);
-  return DEAL_RES_CONTINUE;
+  return;
-}
-
-void nodesDestroyNode(SNodeptr pNode) {
-  if (NULL == pNode) {
-    return;
-  }
-  nodesRewriteNodePostOrder((SNode**)&pNode, destroyNode, NULL);
 }
 
 SNodeList* nodesMakeList() {

source/libs/parser/inc/parAst.h

@@ -159,6 +159,10 @@ SNode* createDropQnodeStmt(SAstCreateContext* pCxt, const SToken* pDnodeId);
 SNode* createCreateTopicStmt(SAstCreateContext* pCxt, bool ignoreExists, const SToken* pTopicName, SNode* pQuery, const SToken* pSubscribeDbName);
 SNode* createDropTopicStmt(SAstCreateContext* pCxt, bool ignoreNotExists, const SToken* pTopicName);
 SNode* createAlterLocalStmt(SAstCreateContext* pCxt, const SToken* pConfig, const SToken* pValue);
+SNode* createDefaultExplainOptions(SAstCreateContext* pCxt);
+SNode* setExplainVerbose(SAstCreateContext* pCxt, SNode* pOptions, const SToken* pVal);
+SNode* setExplainRatio(SAstCreateContext* pCxt, SNode* pOptions, const SToken* pVal);
+SNode* createExplainStmt(SAstCreateContext* pCxt, bool analyze, SNode* pOptions, SNode* pQuery);
 
 #ifdef __cplusplus
 }

source/libs/parser/inc/sql.y

@@ -339,6 +339,18 @@ cmd ::= CREATE TOPIC not_exists_opt(A) topic_name(B) AS query_expression(C).
 cmd ::= CREATE TOPIC not_exists_opt(A) topic_name(B) AS db_name(C).               { pCxt->pRootNode = createCreateTopicStmt(pCxt, A, &B, NULL, &C); }
 cmd ::= DROP TOPIC exists_opt(A) topic_name(B).                                   { pCxt->pRootNode = createDropTopicStmt(pCxt, A, &B); }
 
+/************************************************ select **************************************************************/
+cmd ::= EXPLAIN analyze_opt(A) explain_options(B) query_expression(C).            { pCxt->pRootNode = createExplainStmt(pCxt, A, B, C); }
+
+%type analyze_opt                                                                 { bool }
+%destructor analyze_opt                                                           { }
+analyze_opt(A) ::= .                                                              { A = false; }
+analyze_opt(A) ::= ANALYZE.                                                       { A = true; }
+
+explain_options(A) ::= .                                                          { A = createDefaultExplainOptions(pCxt); }
+explain_options(A) ::= explain_options(B) VERBOSE NK_BOOL(C).                     { A = setExplainVerbose(pCxt, B, &C); }
+explain_options(A) ::= explain_options(B) RATIO NK_FLOAT(C).                      { A = setExplainRatio(pCxt, B, &C); }
+
 /************************************************ select **************************************************************/
 cmd ::= query_expression(A).                                                      { pCxt->pRootNode = A; }
 

source/libs/parser/src/parAstCreater.c

@@ -1316,3 +1316,30 @@ SNode* createAlterLocalStmt(SAstCreateContext* pCxt, const SToken* pConfig, cons
   }
   return (SNode*)pStmt;
 }
+
+SNode* createDefaultExplainOptions(SAstCreateContext* pCxt) {
+  SExplainOptions* pOptions = nodesMakeNode(QUERY_NODE_EXPLAIN_OPTIONS);
+  CHECK_OUT_OF_MEM(pOptions);
+  pOptions->verbose = TSDB_DEFAULT_EXPLAIN_VERBOSE;
+  pOptions->ratio = TSDB_DEFAULT_EXPLAIN_RATIO;
+  return (SNode*)pOptions;
+}
+
+SNode* setExplainVerbose(SAstCreateContext* pCxt, SNode* pOptions, const SToken* pVal) {
+  ((SExplainOptions*)pOptions)->verbose = (0 == strncasecmp(pVal->z, "true", pVal->n));
+  return pOptions;
+}
+
+SNode* setExplainRatio(SAstCreateContext* pCxt, SNode* pOptions, const SToken* pVal) {
+  ((SExplainOptions*)pOptions)->ratio = strtod(pVal->z, NULL);
+  return pOptions;
+}
+
+SNode* createExplainStmt(SAstCreateContext* pCxt, bool analyze, SNode* pOptions, SNode* pQuery) {
+  SExplainStmt* pStmt = nodesMakeNode(QUERY_NODE_EXPLAIN_STMT);
+  CHECK_OUT_OF_MEM(pStmt);
+  pStmt->analyze = analyze;
+  pStmt->pOptions = (SExplainOptions*)pOptions;
+  pStmt->pQuery = pQuery;
+  return (SNode*)pStmt;
+}

source/libs/parser/src/parTokenizer.c

@@ -31,6 +31,7 @@ static SKeyword keywordTable[] = {
     {"ACCOUNT",       TK_ACCOUNT},
     {"ALL",           TK_ALL},
     {"ALTER",         TK_ALTER},
+    {"ANALYZE",       TK_ANALYZE},
     {"AND",           TK_AND},
     {"AS",            TK_AS},
     {"ASC",           TK_ASC},
@@ -56,6 +57,7 @@ static SKeyword keywordTable[] = {
     {"DOUBLE",        TK_DOUBLE},
     {"DROP",          TK_DROP},
     {"EXISTS",        TK_EXISTS},
+    {"EXPLAIN",       TK_EXPLAIN},
     {"FILE_FACTOR",   TK_FILE_FACTOR},
     {"FILL",          TK_FILL},
     {"FLOAT",         TK_FLOAT},
@@ -110,6 +112,7 @@ static SKeyword keywordTable[] = {
     {"QNODES",        TK_QNODES},
     {"QSTARTTS",      TK_QSTARTTS},
     {"QUORUM",        TK_QUORUM},
+    {"RATIO",         TK_RATIO},
     {"REPLICA",       TK_REPLICA},
     {"RETENTIONS",    TK_RETENTIONS},
     {"ROLLUP",        TK_ROLLUP},
@@ -144,6 +147,7 @@ static SKeyword keywordTable[] = {
     {"USING",         TK_USING},
     {"VALUES",        TK_VALUES},
     {"VARCHAR",       TK_VARCHAR},
+    {"VERBOSE",       TK_VERBOSE},
     {"VGROUPS",       TK_VGROUPS},
     {"WAL",           TK_WAL},
     {"WDURATION",     TK_WDURATION},
@@ -224,7 +228,6 @@ static SKeyword keywordTable[] = {
     // {"DETACH",       TK_DETACH},
     // {"EACH",         TK_EACH},
     // {"END",          TK_END},
-    // {"EXPLAIN",      TK_EXPLAIN},
     // {"FAIL",         TK_FAIL},
     // {"FOR",          TK_FOR},
     // {"IGNORE",       TK_IGNORE},

source/libs/parser/src/parTranslater.c

@@ -32,6 +32,7 @@ typedef struct STranslateContext {
   SCmdMsgInfo* pCmdMsg;
   SHashObj* pDbs;
   SHashObj* pTables;
+  SExplainOptions* pExplainOpt;
 } STranslateContext;
 
 typedef struct SFullDatabaseName {
@@ -228,6 +229,9 @@ static void setColumnInfoBySchema(const SRealTableNode* pTable, const SSchema* p
   pCol->colType = isTag ? COLUMN_TYPE_TAG : COLUMN_TYPE_COLUMN;
   pCol->node.resType.type = pColSchema->type;
   pCol->node.resType.bytes = pColSchema->bytes;
+  if (TSDB_DATA_TYPE_TIMESTAMP == pCol->node.resType.type) {
+    pCol->node.resType.precision = pTable->pMeta->tableInfo.precision;
+  }
 }
 
 static void setColumnInfoByExpr(const STableNode* pTable, SExprNode* pExpr, SColumnNode* pCol) {
@@ -670,6 +674,7 @@ static int32_t translateTable(STranslateContext* pCxt, SNode* pTable) {
   switch (nodeType(pTable)) {
     case QUERY_NODE_REAL_TABLE: {
       SRealTableNode* pRealTable = (SRealTableNode*)pTable;
+      pRealTable->ratio = (NULL != pCxt->pExplainOpt ? pCxt->pExplainOpt->ratio : 1.0);
       SName name;
       code = getTableMetaImpl(pCxt,
           toName(pCxt->pParseCxt->acctId, pRealTable->table.dbName, pRealTable->table.tableName, &name), &(pRealTable->pMeta));
@@ -677,10 +682,9 @@ static int32_t translateTable(STranslateContext* pCxt, SNode* pTable) {
         return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_TABLE_NOT_EXIST, pRealTable->table.tableName);
       }
       code = setTableVgroupList(pCxt, &name, pRealTable);
-      if (TSDB_CODE_SUCCESS != code) {
+      if (TSDB_CODE_SUCCESS == code) {
-        return code;
+        code = addNamespace(pCxt, pRealTable);
       }
-      code = addNamespace(pCxt, pRealTable);
       break;
     }
     case QUERY_NODE_TEMP_TABLE: {
@@ -1801,6 +1805,13 @@ static int32_t translateAlterLocal(STranslateContext* pCxt, SAlterLocalStmt* pSt
   return TSDB_CODE_SUCCESS;
 }
 
+static int32_t translateExplain(STranslateContext* pCxt, SExplainStmt* pStmt) {
+  if (pStmt->analyze) {
+    pCxt->pExplainOpt = pStmt->pOptions;
+  }
+  return translateQuery(pCxt, pStmt->pQuery);
+}
+
 static int32_t translateQuery(STranslateContext* pCxt, SNode* pNode) {
   int32_t code = TSDB_CODE_SUCCESS;
   switch (nodeType(pNode)) {
@@ -1882,6 +1893,9 @@ static int32_t translateQuery(STranslateContext* pCxt, SNode* pNode) {
     case QUERY_NODE_ALTER_LOCAL_STMT:
       code = translateAlterLocal(pCxt, (SAlterLocalStmt*)pNode);
       break;
+    case QUERY_NODE_EXPLAIN_STMT:
+      code = translateExplain(pCxt, (SExplainStmt*)pNode);
+      break;
     default:
       break;
   }
@@ -1900,7 +1914,11 @@ static int32_t translateSubquery(STranslateContext* pCxt, SNode* pNode) {
 }
 
 int32_t qExtractResultSchema(const SNode* pRoot, int32_t* numOfCols, SSchema** pSchema) {
-  if (NULL != pRoot && QUERY_NODE_SELECT_STMT == nodeType(pRoot)) {
+  if (NULL == pRoot) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  if (QUERY_NODE_SELECT_STMT == nodeType(pRoot)) {
     SSelectStmt* pSelect = (SSelectStmt*) pRoot;
     *numOfCols = LIST_LENGTH(pSelect->pProjectionList);
     *pSchema = taosMemoryCalloc((*numOfCols), sizeof(SSchema));
@@ -1918,6 +1936,14 @@ int32_t qExtractResultSchema(const SNode* pRoot, int32_t* numOfCols, SSchema** p
       strcpy((*pSchema)[index].name, pExpr->aliasName);
       index +=1;
     }
+  } else if (QUERY_NODE_EXPLAIN_STMT == nodeType(pRoot)) {
+    *numOfCols = 1;
+    *pSchema = taosMemoryCalloc((*numOfCols), sizeof(SSchema));
+    if (NULL == (*pSchema)) {
+      return TSDB_CODE_OUT_OF_MEMORY;
+    }
+    (*pSchema)[0].type = TSDB_DATA_TYPE_BINARY;
+    (*pSchema)[0].bytes = TSDB_EXPLAIN_RESULT_ROW_SIZE;
   }
 
   return TSDB_CODE_SUCCESS;
@@ -2502,6 +2528,7 @@ static int32_t rewriteQuery(STranslateContext* pCxt, SQuery* pQuery) {
 static int32_t setQuery(STranslateContext* pCxt, SQuery* pQuery) {
   switch (nodeType(pQuery->pRoot)) {
     case QUERY_NODE_SELECT_STMT:
+    case QUERY_NODE_EXPLAIN_STMT:
       pQuery->haveResultSet = true;
       pQuery->directRpc = false;
       pQuery->msgType = TDMT_VND_QUERY;

source/libs/parser/src/parser.c

@@ -58,5 +58,7 @@ void qDestroyQuery(SQuery* pQueryNode) {
     taosMemoryFreeClear(pQueryNode->pCmdMsg->pMsg);
     taosMemoryFreeClear(pQueryNode->pCmdMsg);
   }
+  taosArrayDestroy(pQueryNode->pDbList);
+  taosArrayDestroy(pQueryNode->pTableList);
   taosMemoryFreeClear(pQueryNode);
 }

source/libs/parser/test/parserAstTest.cpp

@@ -647,3 +647,16 @@ TEST_F(ParserTest, dropTopic) {
   bind("drop topic if exists tp1");
   ASSERT_TRUE(run());
 }
+
+TEST_F(ParserTest, explain) {
+  setDatabase("root", "test");
+
+  bind("explain SELECT * FROM t1");
+  ASSERT_TRUE(run());
+
+  bind("explain analyze SELECT * FROM t1");
+  ASSERT_TRUE(run());
+
+  bind("explain analyze verbose true ratio 0.01 SELECT * FROM t1");
+  ASSERT_TRUE(run());
+}

source/libs/planner/src/planLogicCreater.c

@@ -197,6 +197,7 @@ static int32_t createScanLogicNode(SLogicPlanContext* pCxt, SSelectStmt* pSelect
   strcpy(pScan->tableName.dbname, pRealTable->table.dbName);
   strcpy(pScan->tableName.tname, pRealTable->table.tableName);
   pScan->showRewrite = pCxt->pPlanCxt->showRewrite;
+  pScan->ratio = pRealTable->ratio;
 
   // set columns to scan
   SNodeList* pCols = NULL;
@@ -692,7 +693,7 @@ static int32_t createVnodeModifLogicNode(SLogicPlanContext* pCxt, SVnodeModifOpS
   if (NULL == pModif) {
     return TSDB_CODE_OUT_OF_MEMORY;
   }
-  pModif->pDataBlocks = pStmt->pDataBlocks;
+  TSWAP(pModif->pDataBlocks, pStmt->pDataBlocks, SArray*);
   pModif->msgType = getMsgType(pStmt->sqlNodeType);
   *pLogicNode = (SLogicNode*)pModif;
   return TSDB_CODE_SUCCESS;
@@ -704,6 +705,8 @@ static int32_t createQueryLogicNode(SLogicPlanContext* pCxt, SNode* pStmt, SLogi
       return createSelectLogicNode(pCxt, (SSelectStmt*)pStmt, pLogicNode);
     case QUERY_NODE_VNODE_MODIF_STMT:
       return createVnodeModifLogicNode(pCxt, (SVnodeModifOpStmt*)pStmt, pLogicNode);
+    case QUERY_NODE_EXPLAIN_STMT:
+      return createQueryLogicNode(pCxt, ((SExplainStmt*)pStmt)->pQuery, pLogicNode);
     default:
       break;
   }

source/libs/planner/src/planPhysiCreater.c

@@ -398,6 +398,7 @@ static int32_t createTableScanPhysiNode(SPhysiPlanContext* pCxt, SSubplan* pSubp
 
   pTableScan->scanFlag = pScanLogicNode->scanFlag;
   pTableScan->scanRange = pScanLogicNode->scanRange;
+  pTableScan->ratio = pScanLogicNode->ratio;
   vgroupInfoToNodeAddr(pScanLogicNode->pVgroupList->vgroups, &pSubplan->execNode);
   taosArrayPush(pCxt->pExecNodeList, &pSubplan->execNode);
   pSubplan->execNodeStat.tableNum = pScanLogicNode->pVgroupList->vgroups[0].numOfTable;
@@ -780,6 +781,7 @@ static int32_t createIntervalPhysiNode(SPhysiPlanContext* pCxt, SNodeList* pChil
   pInterval->sliding = pWindowLogicNode->sliding;
   pInterval->intervalUnit = pWindowLogicNode->intervalUnit;
   pInterval->slidingUnit = pWindowLogicNode->slidingUnit;
+  pInterval->precision = ((SColumnNode*)pWindowLogicNode->pTspk)->node.resType.precision;
 
   pInterval->pFill = nodesCloneNode(pWindowLogicNode->pFill);
   if (NULL != pWindowLogicNode->pFill && NULL == pInterval->pFill) {
@@ -1080,6 +1082,30 @@ static int32_t doCreatePhysiPlan(SPhysiPlanContext* pCxt, SQueryLogicPlan* pLogi
   return code;
 }
 
+static void destoryLocationHash(void* p) {
+  SHashObj* pHash = *(SHashObj**)p;
+  SSlotIndex* pIndex = taosHashIterate(pHash, NULL);
+  while (NULL != pIndex) {
+    taosArrayDestroy(pIndex->pSlotIdsInfo);
+    pIndex = taosHashIterate(pHash, pIndex);
+  }
+  taosHashCleanup(pHash);
+}
+
+static void destoryPhysiPlanContext(SPhysiPlanContext* pCxt) {
+  taosArrayDestroyEx(pCxt->pLocationHelper, destoryLocationHash);
+}
+
+static void setExplainInfo(SPlanContext* pCxt, SQueryPlan* pPlan) {
+  if (QUERY_NODE_EXPLAIN_STMT == nodeType(pCxt->pAstRoot)) {
+    SExplainStmt* pStmt = (SExplainStmt*)pCxt->pAstRoot;
+    pPlan->explainInfo.mode = pStmt->analyze ? EXPLAIN_MODE_ANALYZE : EXPLAIN_MODE_STATIC;
+    pPlan->explainInfo.verbose = pStmt->pOptions->verbose;
+  } else {
+    pPlan->explainInfo.mode = EXPLAIN_MODE_DISABLE;
+  }
+}
+
 int32_t createPhysiPlan(SPlanContext* pCxt, SQueryLogicPlan* pLogicPlan, SQueryPlan** pPlan, SArray* pExecNodeList) {
   SPhysiPlanContext cxt = {
     .pPlanCxt = pCxt,
@@ -1091,5 +1117,12 @@ int32_t createPhysiPlan(SPlanContext* pCxt, SQueryLogicPlan* pLogicPlan, SQueryP
   if (NULL == cxt.pLocationHelper) {
     return TSDB_CODE_OUT_OF_MEMORY;
   }
-  return doCreatePhysiPlan(&cxt, pLogicPlan, pPlan);
+
+  int32_t code = doCreatePhysiPlan(&cxt, pLogicPlan, pPlan);
+  if (TSDB_CODE_SUCCESS == code) {
+    setExplainInfo(pCxt, *pPlan);
+  }
+
+  destoryPhysiPlanContext(&cxt);
+  return code;
 }

source/libs/planner/src/planScaleOut.c

@@ -167,6 +167,8 @@ static int32_t doScaleOut(SScaleOutContext* pCxt, SLogicSubplan* pSubplan, int32
 
   if (TSDB_CODE_SUCCESS != code) {
     nodesDestroyList(pCurrentGroup);
+  } else {
+    nodesClearList(pCurrentGroup);
   }
 
   return code;

source/libs/planner/src/planSpliter.c

@@ -140,6 +140,7 @@ static int32_t stsSplit(SSplitContext* pCxt) {
     code = stsCreateExchangeNode(pCxt, pInfo->pSubplan, pInfo->pScan);
   }
   ++(pCxt->groupId);
+  taosMemoryFreeClear(pCxt->pInfo);
   return code;
 }
 

source/libs/planner/test/plannerTest.cpp

@@ -248,6 +248,11 @@ TEST_F(PlannerTest, showTables) {
 
   bind("show tables");
   ASSERT_TRUE(run());
+
+  setDatabase("root", "information_schema");
+
+  bind("show tables");
+  ASSERT_TRUE(run());
 }
 
 TEST_F(PlannerTest, showStables) {
@@ -277,3 +282,16 @@ TEST_F(PlannerTest, createSmaIndex) {
   bind("create sma index index1 on t1 function(max(c1), min(c3 + 10), sum(c4)) INTERVAL(10s)");
   ASSERT_TRUE(run());
 }
+
+TEST_F(PlannerTest, explain) {
+  setDatabase("root", "test");
+
+  bind("explain SELECT * FROM t1");
+  ASSERT_TRUE(run());
+
+  bind("explain analyze SELECT * FROM t1");
+  ASSERT_TRUE(run());
+
+  bind("explain analyze verbose true ratio 0.01 SELECT * FROM t1");
+  ASSERT_TRUE(run());
+}

source/libs/qworker/src/qworker.c

@@ -959,7 +959,7 @@ int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, int8_t taskType) {
     QW_ERR_JRET(code);
   }
   
-  code = qCreateExecTask(qwMsg->node, mgmt->nodeId, tId, plan, &pTaskInfo, &sinkHandle);
+  code = qCreateExecTask(qwMsg->node, mgmt->nodeId, tId, plan, &pTaskInfo, &sinkHandle, OPTR_EXEC_MODEL_BATCH);
   if (code) {
     QW_TASK_ELOG("qCreateExecTask failed, code:%x - %s", code, tstrerror(code));
     QW_ERR_JRET(code);

source/libs/scalar/src/scalar.c

@@ -288,7 +288,7 @@ _return:
   SCL_RET(code);
 }
 
-int32_t sclExecFuncion(SFunctionNode *node, SScalarCtx *ctx, SScalarParam *output) {
+int32_t sclExecFunction(SFunctionNode *node, SScalarCtx *ctx, SScalarParam *output) {
   if (NULL == node->pParameterList || node->pParameterList->length <= 0) {
     sclError("invalid function parameter list, list:%p, paramNum:%d", node->pParameterList, node->pParameterList ? node->pParameterList->length : 0);
     SCL_ERR_RET(TSDB_CODE_QRY_INVALID_INPUT);
@@ -420,7 +420,7 @@ EDealRes sclRewriteFunction(SNode** pNode, SScalarCtx *ctx) {
   SFunctionNode *node = (SFunctionNode *)*pNode;
   SScalarParam output = {0};
   
-  ctx->code = sclExecFuncion(node, ctx, &output);
+  ctx->code = sclExecFunction(node, ctx, &output);
   if (ctx->code) {
     return DEAL_RES_ERROR;
   }
@@ -547,7 +547,7 @@ EDealRes sclWalkFunction(SNode* pNode, SScalarCtx *ctx) {
   SFunctionNode *node = (SFunctionNode *)pNode;
   SScalarParam output = {0};
   
-  ctx->code = sclExecFuncion(node, ctx, &output);
+  ctx->code = sclExecFunction(node, ctx, &output);
   if (ctx->code) {
     return DEAL_RES_ERROR;
   }
@@ -667,7 +667,7 @@ int32_t scalarCalculateConstants(SNode *pNode, SNode **pRes) {
 
   int32_t code = 0;
   SScalarCtx ctx = {0};
-  ctx.pRes = taosHashInit(SCL_DEFAULT_OP_NUM, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
+  ctx.pRes = taosHashInit(SCL_DEFAULT_OP_NUM, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_NO_LOCK);
   if (NULL == ctx.pRes) {
     sclError("taosHashInit failed, num:%d", SCL_DEFAULT_OP_NUM);
     SCL_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
@@ -689,7 +689,7 @@ int32_t scalarCalculate(SNode *pNode, SArray *pBlockList, SScalarParam *pDst) {
 
   int32_t code = 0;
   SScalarCtx ctx = {.code = 0, .pBlockList = pBlockList};
-  
+  // TODO: OPT performance
   ctx.pRes = taosHashInit(SCL_DEFAULT_OP_NUM, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
   if (NULL == ctx.pRes) {
     sclError("taosHashInit failed, num:%d", SCL_DEFAULT_OP_NUM);
@@ -716,6 +716,3 @@ _return:
   sclFreeRes(ctx.pRes);
   return code;
 }
-
-
-

source/libs/scalar/src/sclfunc.c

@@ -377,3 +377,34 @@ static void reverseCopy(char* dest, const char* src, int16_t type, int32_t numOf
   }
 }
 
+bool getTimePseudoFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) {
+  pEnv->calcMemSize = sizeof(int64_t);
+  return true;
+}
+
+int32_t qStartTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
+  ASSERT(inputNum == 1);
+  colDataAppendInt64(pOutput->columnData, pOutput->numOfRows, (int64_t *)colDataGetData(pInput->columnData, 0));
+}
+
+int32_t qEndTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
+  ASSERT(inputNum == 1);
+  colDataAppendInt64(pOutput->columnData, pOutput->numOfRows, (int64_t *)colDataGetData(pInput->columnData, 1));
+}
+
+int32_t winDurFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
+  ASSERT(inputNum == 1);
+  colDataAppendInt64(pOutput->columnData, pOutput->numOfRows, (int64_t *)colDataGetData(pInput->columnData, 2));
+}
+
+int32_t winStartTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
+  ASSERT(inputNum == 1);
+  colDataAppendInt64(pOutput->columnData, pOutput->numOfRows, (int64_t*) colDataGetData(pInput->columnData, 3));
+  return TSDB_CODE_SUCCESS;
+}
+
+int32_t winEndTsFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
+  ASSERT(inputNum == 1);
+  colDataAppendInt64(pOutput->columnData, pOutput->numOfRows, (int64_t*) colDataGetData(pInput->columnData, 4));
+  return TSDB_CODE_SUCCESS;
+}

source/libs/tdb/CMakeLists.txt

@@ -9,9 +9,8 @@ target_sources(tdb
     "src/db/tdbDb.c"
     "src/db/tdbEnv.c"
     "src/db/tdbTxn.c"
+    "src/db/tdbPage.c"
     "src/db/tdbOs.c"
-    "src/page/tdbPage.c"
-    "src/page/tdbPageL.c"
 )
 
 target_include_directories(

source/libs/tdb/src/db/tdbBtree.c

@@ -18,19 +18,12 @@
 #define TDB_BTREE_ROOT 0x1
 #define TDB_BTREE_LEAF 0x2
 
-#define TDB_BTREE_PAGE_IS_ROOT(flags) TDB_FLAG_HAS(flags, TDB_BTREE_ROOT)
-#define TDB_BTREE_PAGE_IS_LEAF(flags) TDB_FLAG_HAS(flags, TDB_BTREE_LEAF)
-#define TDB_BTREE_ASSERT_FLAG(flags)                                                 \
-  ASSERT(TDB_FLAG_IS(flags, TDB_BTREE_ROOT) || TDB_FLAG_IS(flags, TDB_BTREE_LEAF) || \
-         TDB_FLAG_IS(flags, TDB_BTREE_ROOT | TDB_BTREE_LEAF) || TDB_FLAG_IS(flags, 0))
-
 struct SBTree {
   SPgno          root;
   int            keyLen;
   int            valLen;
   SPager        *pPager;
   FKeyComparator kcmpr;
-  u8             fanout;
   int            pageSize;
   int            maxLocal;
   int            minLocal;
@@ -41,8 +34,13 @@ struct SBTree {
 
 #define TDB_BTREE_PAGE_COMMON_HDR u8 flags;
 
-#define TDB_BTREE_PAGE_GET_FLAGS(PAGE)        (PAGE)->pData[0]
+#define TDB_BTREE_PAGE_GET_FLAGS(PAGE) (PAGE)->pData[0]
 #define TDB_BTREE_PAGE_SET_FLAGS(PAGE, flags) ((PAGE)->pData[0] = (flags))
+#define TDB_BTREE_PAGE_IS_ROOT(PAGE) (TDB_BTREE_PAGE_GET_FLAGS(PAGE) & TDB_BTREE_ROOT)
+#define TDB_BTREE_PAGE_IS_LEAF(PAGE) (TDB_BTREE_PAGE_GET_FLAGS(PAGE) & TDB_BTREE_LEAF)
+#define TDB_BTREE_ASSERT_FLAG(flags)                                                 \
+  ASSERT(TDB_FLAG_IS(flags, TDB_BTREE_ROOT) || TDB_FLAG_IS(flags, TDB_BTREE_LEAF) || \
+         TDB_FLAG_IS(flags, TDB_BTREE_ROOT | TDB_BTREE_LEAF) || TDB_FLAG_IS(flags, 0))
 
 typedef struct __attribute__((__packed__)) {
   TDB_BTREE_PAGE_COMMON_HDR
@@ -59,15 +57,15 @@ typedef struct {
 } SBtreeInitPageArg;
 
 typedef struct {
-  int   kLen;
+  int       kLen;
-  u8   *pKey;
+  const u8 *pKey;
-  int   vLen;
+  int       vLen;
-  u8   *pVal;
+  const u8 *pVal;
-  SPgno pgno;
+  SPgno     pgno;
-  u8   *pTmpSpace;
+  u8       *pBuf;
 } SCellDecoder;
 
-static int tdbBtCursorMoveTo(SBTC *pBtc, const void *pKey, int kLen, int *pCRst);
+static int tdbBtcMoveTo(SBTC *pBtc, const void *pKey, int kLen, int *pCRst);
 static int tdbDefaultKeyCmprFn(const void *pKey1, int keyLen1, const void *pKey2, int keyLen2);
 static int tdbBtreeOpenImpl(SBTree *pBt);
 static int tdbBtreeZeroPage(SPage *pPage, void *arg);
@@ -78,13 +76,15 @@ static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pD
 static int tdbBtreeBalance(SBTC *pBtc);
 static int tdbBtreeCellSize(const SPage *pPage, SCell *pCell);
 static int tdbBtcMoveToNext(SBTC *pBtc);
-static int tdbBtcMoveDownward(SBTC *pBtc, SPgno pgno);
+static int tdbBtcMoveDownward(SBTC *pBtc);
 static int tdbBtcMoveUpward(SBTC *pBtc);
 
 int tdbBtreeOpen(int keyLen, int valLen, SPager *pPager, FKeyComparator kcmpr, SBTree **ppBt) {
   SBTree *pBt;
   int     ret;
 
+  ASSERT(keyLen != 0);
+
   *ppBt = NULL;
 
   pBt = (SBTree *)tdbOsCalloc(1, sizeof(*pBt));
@@ -93,28 +93,21 @@ int tdbBtreeOpen(int keyLen, int valLen, SPager *pPager, FKeyComparator kcmpr, S
   }
 
   // pBt->keyLen
-  pBt->keyLen = keyLen;
+  pBt->keyLen = keyLen < 0 ? TDB_VARIANT_LEN : keyLen;
   // pBt->valLen
-  pBt->valLen = valLen;
+  pBt->valLen = valLen < 0 ? TDB_VARIANT_LEN : valLen;
   // pBt->pPager
   pBt->pPager = pPager;
   // pBt->kcmpr
   pBt->kcmpr = kcmpr ? kcmpr : tdbDefaultKeyCmprFn;
-  // pBt->fanout
-  if (keyLen == TDB_VARIANT_LEN) {
-    pBt->fanout = TDB_DEFAULT_FANOUT;
-  } else {
-    ASSERT(0);
-    // TODO: pBt->fanout = 0;
-  }
   // pBt->pageSize
   pBt->pageSize = tdbPagerGetPageSize(pPager);
   // pBt->maxLocal
-  pBt->maxLocal = (pBt->pageSize - 14) / pBt->fanout;
+  pBt->maxLocal = tdbPageCapacity(pBt->pageSize, sizeof(SIntHdr)) / 4;
   // pBt->minLocal: Should not be allowed smaller than 15, which is [nPayload][nKey][nData]
-  pBt->minLocal = (pBt->pageSize - 14) / pBt->fanout / 2;
+  pBt->minLocal = pBt->maxLocal / 2;
   // pBt->maxLeaf
-  pBt->maxLeaf = pBt->pageSize - 14;
+  pBt->maxLeaf = tdbPageCapacity(pBt->pageSize, sizeof(SLeafHdr));
   // pBt->minLeaf
   pBt->minLeaf = pBt->minLocal;
 
@@ -143,7 +136,7 @@ int tdbBtCursorInsert(SBTC *pBtc, const void *pKey, int kLen, const void *pVal,
   int     cret;
   SBTree *pBt;
 
-  ret = tdbBtCursorMoveTo(pBtc, pKey, kLen, &cret);
+  ret = tdbBtcMoveTo(pBtc, pKey, kLen, &cret);
   if (ret < 0) {
     // TODO: handle error
     return -1;
@@ -206,7 +199,7 @@ int tdbBtreeGet(SBTree *pBt, const void *pKey, int kLen, void **ppVal, int *vLen
 
   tdbBtcOpen(&btc, pBt);
 
-  tdbBtCursorMoveTo(&btc, pKey, kLen, &cret);
+  tdbBtcMoveTo(&btc, pKey, kLen, &cret);
 
   if (cret) {
     return cret;
@@ -226,103 +219,40 @@ int tdbBtreeGet(SBTree *pBt, const void *pKey, int kLen, void **ppVal, int *vLen
   return 0;
 }
 
-static int tdbBtCursorMoveTo(SBTC *pBtc, const void *pKey, int kLen, int *pCRst) {
+int tdbBtreePGet(SBTree *pBt, const void *pKey, int kLen, void **ppKey, int *pkLen, void **ppVal, int *vLen) {
-  int     ret;
+  SBTC         btc;
-  SBTree *pBt;
+  SCell       *pCell;
-  SPager *pPager;
+  int          cret;
+  void        *pTKey;
+  void        *pTVal;
+  SCellDecoder cd;
 
-  pBt = pBtc->pBt;
+  tdbBtcOpen(&btc, pBt);
-  pPager = pBt->pPager;
 
-  if (pBtc->iPage < 0) {
+  tdbBtcMoveTo(&btc, pKey, kLen, &cret);
-    ASSERT(pBtc->iPage == -1);
+  if (cret) {
-    ASSERT(pBtc->idx == -1);
+    return cret;
-
-    // Move from the root
-    ret = tdbPagerFetchPage(pPager, pBt->root, &(pBtc->pPage), tdbBtreeInitPage, pBt);
-    if (ret < 0) {
-      ASSERT(0);
-      return -1;
-    }
-
-    pBtc->iPage = 0;
-
-    if (TDB_PAGE_TOTAL_CELLS(pBtc->pPage) == 0) {
-      // Current page is empty
-      // ASSERT(TDB_FLAG_IS(TDB_PAGE_FLAGS(pBtc->pPage), TDB_BTREE_ROOT | TDB_BTREE_LEAF));
-      return 0;
-    }
-
-    for (;;) {
-      int          lidx, ridx, midx, c, nCells;
-      SCell       *pCell;
-      SPage       *pPage;
-      SCellDecoder cd = {0};
-
-      pPage = pBtc->pPage;
-      nCells = TDB_PAGE_TOTAL_CELLS(pPage);
-      lidx = 0;
-      ridx = nCells - 1;
-
-      ASSERT(nCells > 0);
-
-      for (;;) {
-        if (lidx > ridx) break;
-
-        midx = (lidx + ridx) >> 1;
-
-        pCell = tdbPageGetCell(pPage, midx);
-        ret = tdbBtreeDecodeCell(pPage, pCell, &cd);
-        if (ret < 0) {
-          // TODO: handle error
-          ASSERT(0);
-          return -1;
-        }
-
-        // Compare the key values
-        c = pBt->kcmpr(pKey, kLen, cd.pKey, cd.kLen);
-        if (c < 0) {
-          /* input-key < cell-key */
-          ridx = midx - 1;
-        } else if (c > 0) {
-          /* input-key > cell-key */
-          lidx = midx + 1;
-        } else {
-          /* input-key == cell-key */
-          break;
-        }
-      }
-
-      // Move downward or break
-      u8 flags = TDB_BTREE_PAGE_GET_FLAGS(pPage);
-      u8 leaf = TDB_BTREE_PAGE_IS_LEAF(flags);
-      if (leaf) {
-        pBtc->idx = midx;
-        *pCRst = c;
-        break;
-      } else {
-        if (c <= 0) {
-          pBtc->idx = midx;
-          tdbBtcMoveDownward(pBtc, cd.pgno);
-        } else {
-          pBtc->idx = midx + 1;
-          if (midx == nCells - 1) {
-            /* Move to right-most child */
-            tdbBtcMoveDownward(pBtc, ((SIntHdr *)pBtc->pPage->pData)->pgno);
-          } else {
-            pCell = tdbPageGetCell(pPage, pBtc->idx);
-            tdbBtreeDecodeCell(pPage, pCell, &cd);
-            tdbBtcMoveDownward(pBtc, cd.pgno);
-          }
-        }
-      }
-    }
-
-  } else {
-    // TODO: Move the cursor from a some position instead of a clear state
-    ASSERT(0);
   }
 
+  pCell = tdbPageGetCell(btc.pPage, btc.idx);
+  tdbBtreeDecodeCell(btc.pPage, pCell, &cd);
+
+  pTKey = TDB_REALLOC(*ppKey, cd.kLen);
+  pTVal = TDB_REALLOC(*ppVal, cd.vLen);
+
+  if (pTKey == NULL || pTVal == NULL) {
+    TDB_FREE(pTKey);
+    TDB_FREE(pTVal);
+  }
+
+  *ppKey = pTKey;
+  *ppVal = pTVal;
+  *pkLen = cd.kLen;
+  *vLen = cd.vLen;
+
+  memcpy(*ppKey, cd.pKey, cd.kLen);
+  memcpy(*ppVal, cd.pVal, cd.vLen);
+
   return 0;
 }
 
@@ -385,7 +315,7 @@ static int tdbBtreeInitPage(SPage *pPage, void *arg) {
 
   pBt = (SBTree *)arg;
   flags = TDB_BTREE_PAGE_GET_FLAGS(pPage);
-  isLeaf = TDB_BTREE_PAGE_IS_LEAF(flags);
+  isLeaf = TDB_BTREE_PAGE_IS_LEAF(pPage);
 
   ASSERT(flags == TDB_BTREE_PAGE_GET_FLAGS(pPage));
 
@@ -411,15 +341,15 @@ static int tdbBtreeInitPage(SPage *pPage, void *arg) {
 static int tdbBtreeZeroPage(SPage *pPage, void *arg) {
   u8      flags;
   SBTree *pBt;
-  u8      isLeaf;
+  u8      leaf;
 
   flags = ((SBtreeInitPageArg *)arg)->flags;
   pBt = ((SBtreeInitPageArg *)arg)->pBt;
-  isLeaf = TDB_BTREE_PAGE_IS_LEAF(flags);
+  leaf = flags & TDB_BTREE_LEAF;
 
-  tdbPageZero(pPage, isLeaf ? sizeof(SLeafHdr) : sizeof(SIntHdr), tdbBtreeCellSize);
+  tdbPageZero(pPage, leaf ? sizeof(SLeafHdr) : sizeof(SIntHdr), tdbBtreeCellSize);
 
-  if (isLeaf) {
+  if (leaf) {
     SLeafHdr *pLeafHdr = (SLeafHdr *)(pPage->pData);
     pLeafHdr->flags = flags;
 
@@ -464,7 +394,7 @@ static int tdbBtreeBalanceDeeper(SBTree *pBt, SPage *pRoot, SPage **ppChild) {
 
   pPager = pRoot->pPager;
   flags = TDB_BTREE_PAGE_GET_FLAGS(pRoot);
-  leaf = TDB_BTREE_PAGE_IS_LEAF(flags);
+  leaf = TDB_BTREE_PAGE_IS_LEAF(pRoot);
 
   // Allocate a new child page
   zArg.flags = TDB_FLAG_REMOVE(flags, TDB_BTREE_ROOT);
@@ -530,7 +460,7 @@ static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx) {
 
       SPgno pgno;
       if (sIdx + i == nCells) {
-        ASSERT(!TDB_BTREE_PAGE_IS_LEAF(TDB_BTREE_PAGE_GET_FLAGS(pParent)));
+        ASSERT(!TDB_BTREE_PAGE_IS_LEAF(pParent));
         pgno = ((SIntHdr *)(pParent->pData))->pgno;
       } else {
         pCell = tdbPageGetCell(pParent, sIdx + i);
@@ -544,7 +474,7 @@ static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx) {
       }
     }
     // copy the parent key out if child pages are not leaf page
-    childNotLeaf = !TDB_BTREE_PAGE_IS_LEAF(TDB_BTREE_PAGE_GET_FLAGS(pOlds[0]));
+    childNotLeaf = !TDB_BTREE_PAGE_IS_LEAF(pOlds[0]);
     if (childNotLeaf) {
       for (int i = 0; i < nOlds; i++) {
         if (sIdx + i < TDB_PAGE_TOTAL_CELLS(pParent)) {
@@ -818,9 +748,8 @@ static int tdbBtreeBalance(SBTC *pBtc) {
   for (;;) {
     iPage = pBtc->iPage;
     pPage = pBtc->pPage;
-    flags = TDB_BTREE_PAGE_GET_FLAGS(pPage);
+    leaf = TDB_BTREE_PAGE_IS_LEAF(pPage);
-    leaf = TDB_BTREE_PAGE_IS_LEAF(flags);
+    root = TDB_BTREE_PAGE_IS_ROOT(pPage);
-    root = TDB_BTREE_PAGE_IS_ROOT(flags);
 
     // when the page is not overflow and not too empty, the balance work
     // is finished. Just break out the balance loop.
@@ -861,23 +790,17 @@ static int tdbBtreeBalance(SBTC *pBtc) {
 }
 #endif
 
-#ifndef TDB_BTREE_CELL  // =========================================================
+// TDB_BTREE_CELL =====================
-static int tdbBtreeEncodePayload(SPage *pPage, u8 *pPayload, const void *pKey, int kLen, const void *pVal, int vLen,
+static int tdbBtreeEncodePayload(SPage *pPage, SCell *pCell, int nHeader, const void *pKey, int kLen, const void *pVal,
-                                 int *szPayload) {
+                                 int vLen, int *szPayload) {
   int nPayload;
 
-  ASSERT(pKey != NULL);
-
-  if (pVal == NULL) {
-    vLen = 0;
-  }
-
   nPayload = kLen + vLen;
-  if (nPayload <= pPage->maxLocal) {
+  if (nPayload + nHeader <= pPage->maxLocal) {
-    // General case without overflow
+    // no overflow page is needed
-    memcpy(pPayload, pKey, kLen);
+    memcpy(pCell + nHeader, pKey, kLen);
     if (pVal) {
-      memcpy(pPayload + kLen, pVal, vLen);
+      memcpy(pCell + nHeader + kLen, pVal, vLen);
     }
 
     *szPayload = nPayload;
@@ -892,10 +815,8 @@ static int tdbBtreeEncodePayload(SPage *pPage, u8 *pPayload, const void *pKey, i
   return 0;
 }
 
-// TODO: allow vLen = 0
 static int tdbBtreeEncodeCell(SPage *pPage, const void *pKey, int kLen, const void *pVal, int vLen, SCell *pCell,
                               int *szCell) {
-  u8  flags;
   u8  leaf;
   int nHeader;
   int nPayload;
@@ -903,11 +824,11 @@ static int tdbBtreeEncodeCell(SPage *pPage, const void *pKey, int kLen, const vo
 
   ASSERT(pPage->kLen == TDB_VARIANT_LEN || pPage->kLen == kLen);
   ASSERT(pPage->vLen == TDB_VARIANT_LEN || pPage->vLen == vLen);
+  ASSERT(pKey != NULL && kLen > 0);
 
   nPayload = 0;
   nHeader = 0;
-  flags = TDB_BTREE_PAGE_GET_FLAGS(pPage);
+  leaf = TDB_BTREE_PAGE_IS_LEAF(pPage);
-  leaf = TDB_BTREE_PAGE_IS_LEAF(flags);
 
   // 1. Encode Header part
   /* Encode SPgno if interior page */
@@ -929,38 +850,42 @@ static int tdbBtreeEncodeCell(SPage *pPage, const void *pKey, int kLen, const vo
   }
 
   // 2. Encode payload part
-  if (leaf) {
+  if ((!leaf) || pPage->vLen == 0) {
-    ret = tdbBtreeEncodePayload(pPage, pCell + nHeader, pKey, kLen, pVal, vLen, &nPayload);
+    pVal = NULL;
-  } else {
+    vLen = 0;
-    ret = tdbBtreeEncodePayload(pPage, pCell + nHeader, pKey, kLen, NULL, 0, &nPayload);
   }
+
+  ret = tdbBtreeEncodePayload(pPage, pCell, nHeader, pKey, kLen, pVal, vLen, &nPayload);
   if (ret < 0) {
-    // TODO: handle error
+    // TODO
-    return -1;
+    ASSERT(0);
+    return 0;
   }
 
   *szCell = nHeader + nPayload;
   return 0;
 }
 
-static int tdbBtreeDecodePayload(SPage *pPage, const u8 *pPayload, SCellDecoder *pDecoder) {
+static int tdbBtreeDecodePayload(SPage *pPage, const SCell *pCell, int nHeader, SCellDecoder *pDecoder) {
   int nPayload;
 
-  ASSERT(pDecoder->pKey == NULL);
-
   if (pDecoder->pVal) {
-    nPayload = pDecoder->kLen + pDecoder->vLen;
+    ASSERT(!TDB_BTREE_PAGE_IS_LEAF(pPage));
-  } else {
     nPayload = pDecoder->kLen;
+  } else {
+    nPayload = pDecoder->kLen + pDecoder->vLen;
   }
 
-  if (nPayload <= pPage->maxLocal) {
+  if (nHeader + nPayload <= pPage->maxLocal) {
-    // General case without overflow
+    // no over flow case
-    pDecoder->pKey = (void *)pPayload;
+    pDecoder->pKey = pCell + nHeader;
-    if (!pDecoder->pVal) {
+    if (pDecoder->pVal == NULL && pDecoder->vLen > 0) {
-      pDecoder->pVal = (void *)(pPayload + pDecoder->kLen);
+      pDecoder->pVal = pCell + nHeader + pDecoder->kLen;
     }
-  } else {
+    return 0;
+  }
+
+  {
     // TODO: handle overflow case
     ASSERT(0);
   }
@@ -969,14 +894,12 @@ static int tdbBtreeDecodePayload(SPage *pPage, const u8 *pPayload, SCellDecoder
 }
 
 static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pDecoder) {
-  u8  flags;
   u8  leaf;
   int nHeader;
   int ret;
 
   nHeader = 0;
-  flags = TDB_BTREE_PAGE_GET_FLAGS(pPage);
+  leaf = TDB_BTREE_PAGE_IS_LEAF(pPage);
-  leaf = TDB_BTREE_PAGE_IS_LEAF(flags);
 
   // Clear the state of decoder
   pDecoder->kLen = -1;
@@ -1001,13 +924,14 @@ static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pD
   }
 
   if (pPage->vLen == TDB_VARIANT_LEN) {
+    ASSERT(leaf);
     nHeader += tdbGetVarInt(pCell + nHeader, &(pDecoder->vLen));
   } else {
     pDecoder->vLen = pPage->vLen;
   }
 
   // 2. Decode payload part
-  ret = tdbBtreeDecodePayload(pPage, pCell + nHeader, pDecoder);
+  ret = tdbBtreeDecodePayload(pPage, pCell, nHeader, pDecoder);
   if (ret < 0) {
     return -1;
   }
@@ -1016,16 +940,14 @@ static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pD
 }
 
 static int tdbBtreeCellSize(const SPage *pPage, SCell *pCell) {
-  u8  flags;
+  u8  leaf;
-  u8  isLeaf;
   int szCell;
   int kLen = 0, vLen = 0;
 
-  flags = TDB_BTREE_PAGE_GET_FLAGS(pPage);
+  leaf = TDB_BTREE_PAGE_IS_LEAF(pPage);
-  isLeaf = TDB_BTREE_PAGE_IS_LEAF(flags);
   szCell = 0;
 
-  if (!isLeaf) {
+  if (!leaf) {
     szCell += sizeof(SPgno);
   }
 
@@ -1035,21 +957,28 @@ static int tdbBtreeCellSize(const SPage *pPage, SCell *pCell) {
     kLen = pPage->kLen;
   }
 
-  if (isLeaf) {
+  if (pPage->vLen == TDB_VARIANT_LEN) {
-    if (pPage->vLen == TDB_VARIANT_LEN) {
+    ASSERT(leaf);
-      szCell += tdbGetVarInt(pCell + szCell, &vLen);
+    szCell += tdbGetVarInt(pCell + szCell, &vLen);
-    } else {
+  } else if (leaf) {
-      vLen = pPage->vLen;
+    vLen = pPage->vLen;
-    }
   }
 
   szCell = szCell + kLen + vLen;
 
-  return szCell;
+  if (szCell <= pPage->maxLocal) {
+    return szCell;
+  }
+
+  {
+    // TODO
+    ASSERT(0);
+    return 0;
+  }
 }
+// TDB_BTREE_CELL
 
-#endif
+// TDB_BTREE_CURSOR =====================
-
 int tdbBtcOpen(SBTC *pBtc, SBTree *pBt) {
   pBtc->pBt = pBt;
   pBtc->iPage = -1;
@@ -1063,7 +992,6 @@ int tdbBtcMoveToFirst(SBTC *pBtc) {
   int     ret;
   SBTree *pBt;
   SPager *pPager;
-  u8      flags;
   SCell  *pCell;
   SPgno   pgno;
 
@@ -1078,23 +1006,43 @@ int tdbBtcMoveToFirst(SBTC *pBtc) {
       return -1;
     }
 
+    ASSERT(TDB_BTREE_PAGE_IS_ROOT(pBtc->pPage));
+
     pBtc->iPage = 0;
-    pBtc->idx = 0;
+    if (TDB_PAGE_TOTAL_CELLS(pBtc->pPage) > 0) {
+      pBtc->idx = 0;
+    } else {
+      // no any data, point to an invalid position
+      ASSERT(TDB_BTREE_PAGE_IS_LEAF(pBtc->pPage));
+      pBtc->idx = -1;
+      return 0;
+    }
   } else {
     // move from a position
-    ASSERT(0);
+    int iPage = 0;
+
+    for (; iPage < pBtc->iPage; iPage++) {
+      ASSERT(pBtc->idxStack[iPage] >= 0);
+      if (pBtc->idxStack[iPage]) break;
+    }
+
+    // move upward
+    for (;;) {
+      if (pBtc->iPage == 0) {
+        pBtc->idx = 0;
+        break;
+      }
+
+      if (pBtc->iPage < iPage) break;
+      tdbBtcMoveUpward(pBtc);
+    }
   }
 
   // move downward
   for (;;) {
-    flags = TDB_BTREE_PAGE_GET_FLAGS(pBtc->pPage);
+    if (TDB_BTREE_PAGE_IS_LEAF(pBtc->pPage)) break;
 
-    if (TDB_BTREE_PAGE_IS_LEAF(flags)) break;
+    ret = tdbBtcMoveDownward(pBtc);
-
-    pCell = tdbPageGetCell(pBtc->pPage, 0);
-    pgno = *(SPgno *)pCell;
-
-    ret = tdbBtcMoveDownward(pBtc, pgno);
     if (ret < 0) {
       ASSERT(0);
       return -1;
@@ -1110,7 +1058,6 @@ int tdbBtcMoveToLast(SBTC *pBtc) {
   int     ret;
   SBTree *pBt;
   SPager *pPager;
-  u8      flags;
   SPgno   pgno;
 
   pBt = pBtc->pBt;
@@ -1132,18 +1079,15 @@ int tdbBtcMoveToLast(SBTC *pBtc) {
 
   // move downward
   for (;;) {
-    flags = TDB_BTREE_PAGE_GET_FLAGS(pBtc->pPage);
+    if (TDB_BTREE_PAGE_IS_LEAF(pBtc->pPage)) {
-
-    if (TDB_BTREE_PAGE_IS_LEAF(flags)) {
       // TODO: handle empty case
       ASSERT(TDB_PAGE_TOTAL_CELLS(pBtc->pPage) > 0);
       pBtc->idx = TDB_PAGE_TOTAL_CELLS(pBtc->pPage) - 1;
       break;
     } else {
       pBtc->idx = TDB_PAGE_TOTAL_CELLS(pBtc->pPage);
-      pgno = ((SIntHdr *)pBtc->pPage->pData)->pgno;
 
-      ret = tdbBtcMoveDownward(pBtc, pgno);
+      ret = tdbBtcMoveDownward(pBtc);
       if (ret < 0) {
         ASSERT(0);
         return -1;
@@ -1154,11 +1098,6 @@ int tdbBtcMoveToLast(SBTC *pBtc) {
   return 0;
 }
 
-int tdbBtcMoveTo(SBTC *pBtc, const void *pKey, int kLen) {
-  // TODO
-  return 0;
-}
-
 int tdbBtreeNext(SBTC *pBtc, void **ppKey, int *kLen, void **ppVal, int *vLen) {
   SCell       *pCell;
   SCellDecoder cd;
@@ -1201,11 +1140,9 @@ int tdbBtreeNext(SBTC *pBtc, void **ppKey, int *kLen, void **ppVal, int *vLen) {
 
 static int tdbBtcMoveToNext(SBTC *pBtc) {
   int    nCells;
-  SPgno  pgno;
   SCell *pCell;
-  u8     flags;
 
-  ASSERT(TDB_BTREE_PAGE_IS_LEAF(TDB_BTREE_PAGE_GET_FLAGS(pBtc->pPage)));
+  ASSERT(TDB_BTREE_PAGE_IS_LEAF(pBtc->pPage));
 
   if (pBtc->idx < 0) return -1;
 
@@ -1238,18 +1175,11 @@ static int tdbBtcMoveToNext(SBTC *pBtc) {
   // Move downward
   for (;;) {
     nCells = TDB_PAGE_TOTAL_CELLS(pBtc->pPage);
-    if (pBtc->idx < nCells) {
-      pCell = tdbPageGetCell(pBtc->pPage, pBtc->idx);
-      pgno = *(SPgno *)pCell;
-    } else {
-      pgno = ((SIntHdr *)pBtc->pPage->pData)->pgno;
-    }
 
-    tdbBtcMoveDownward(pBtc, pgno);
+    tdbBtcMoveDownward(pBtc);
     pBtc->idx = 0;
 
-    flags = TDB_BTREE_PAGE_GET_FLAGS(pBtc->pPage);
+    if (TDB_BTREE_PAGE_IS_LEAF(pBtc->pPage)) {
-    if (TDB_BTREE_PAGE_IS_LEAF(flags)) {
       break;
     }
   }
@@ -1257,13 +1187,20 @@ static int tdbBtcMoveToNext(SBTC *pBtc) {
   return 0;
 }
 
-int tdbBtcClose(SBTC *pBtc) {
+static int tdbBtcMoveDownward(SBTC *pBtc) {
-  // TODO
+  int    ret;
-  return 0;
+  SPgno  pgno;
-}
+  SCell *pCell;
 
-static int tdbBtcMoveDownward(SBTC *pBtc, SPgno pgno) {
+  ASSERT(pBtc->idx >= 0);
-  int ret;
+  ASSERT(!TDB_BTREE_PAGE_IS_LEAF(pBtc->pPage));
+
+  if (pBtc->idx < TDB_PAGE_TOTAL_CELLS(pBtc->pPage)) {
+    pCell = tdbPageGetCell(pBtc->pPage, pBtc->idx);
+    pgno = ((SPgno *)pCell)[0];
+  } else {
+    pgno = ((SIntHdr *)pBtc->pPage->pData)->pgno;
+  }
 
   pBtc->pgStack[pBtc->iPage] = pBtc->pPage;
   pBtc->idxStack[pBtc->iPage] = pBtc->idx;
@@ -1274,6 +1211,7 @@ static int tdbBtcMoveDownward(SBTC *pBtc, SPgno pgno) {
   ret = tdbPagerFetchPage(pBtc->pBt->pPager, pgno, &pBtc->pPage, tdbBtreeInitPage, pBtc->pBt);
   if (ret < 0) {
     ASSERT(0);
+    return -1;
   }
 
   return 0;
@@ -1282,7 +1220,7 @@ static int tdbBtcMoveDownward(SBTC *pBtc, SPgno pgno) {
 static int tdbBtcMoveUpward(SBTC *pBtc) {
   if (pBtc->iPage == 0) return -1;
 
-  // tdbPagerReturnPage(pBtc->pBt->pPager, pBtc->pPage);
+  tdbPagerReturnPage(pBtc->pBt->pPager, pBtc->pPage);
 
   pBtc->iPage--;
   pBtc->pPage = pBtc->pgStack[pBtc->iPage];
@@ -1291,6 +1229,117 @@ static int tdbBtcMoveUpward(SBTC *pBtc) {
   return 0;
 }
 
+static int tdbBtcMoveTo(SBTC *pBtc, const void *pKey, int kLen, int *pCRst) {
+  int     ret;
+  SBTree *pBt;
+  SPager *pPager;
+
+  pBt = pBtc->pBt;
+  pPager = pBt->pPager;
+
+  if (pBtc->iPage < 0) {
+    ASSERT(pBtc->iPage == -1);
+    ASSERT(pBtc->idx == -1);
+
+    // Move from the root
+    ret = tdbPagerFetchPage(pPager, pBt->root, &(pBtc->pPage), tdbBtreeInitPage, pBt);
+    if (ret < 0) {
+      ASSERT(0);
+      return -1;
+    }
+
+    pBtc->iPage = 0;
+
+    if (TDB_PAGE_TOTAL_CELLS(pBtc->pPage) == 0) {
+      // Current page is empty
+      // ASSERT(TDB_FLAG_IS(TDB_PAGE_FLAGS(pBtc->pPage), TDB_BTREE_ROOT | TDB_BTREE_LEAF));
+      return 0;
+    }
+
+    for (;;) {
+      int          lidx, ridx, midx, c, nCells;
+      SCell       *pCell;
+      SPage       *pPage;
+      SCellDecoder cd = {0};
+
+      pPage = pBtc->pPage;
+      nCells = TDB_PAGE_TOTAL_CELLS(pPage);
+      lidx = 0;
+      ridx = nCells - 1;
+
+      ASSERT(nCells > 0);
+
+      for (;;) {
+        if (lidx > ridx) break;
+
+        midx = (lidx + ridx) >> 1;
+
+        pCell = tdbPageGetCell(pPage, midx);
+        ret = tdbBtreeDecodeCell(pPage, pCell, &cd);
+        if (ret < 0) {
+          // TODO: handle error
+          ASSERT(0);
+          return -1;
+        }
+
+        // Compare the key values
+        c = pBt->kcmpr(pKey, kLen, cd.pKey, cd.kLen);
+        if (c < 0) {
+          /* input-key < cell-key */
+          ridx = midx - 1;
+        } else if (c > 0) {
+          /* input-key > cell-key */
+          lidx = midx + 1;
+        } else {
+          /* input-key == cell-key */
+          break;
+        }
+      }
+
+      // Move downward or break
+      u8 leaf = TDB_BTREE_PAGE_IS_LEAF(pPage);
+      if (leaf) {
+        pBtc->idx = midx;
+        *pCRst = c;
+        break;
+      } else {
+        if (c <= 0) {
+          pBtc->idx = midx;
+        } else {
+          pBtc->idx = midx + 1;
+        }
+        tdbBtcMoveDownward(pBtc);
+      }
+    }
+
+  } else {
+    // TODO: Move the cursor from a some position instead of a clear state
+    ASSERT(0);
+  }
+
+  return 0;
+}
+
+int tdbBtcClose(SBTC *pBtc) {
+  if (pBtc->iPage < 0) return 0;
+
+  for (;;) {
+    ASSERT(pBtc->pPage);
+
+    tdbPagerReturnPage(pBtc->pBt->pPager, pBtc->pPage);
+
+    pBtc->iPage--;
+    if (pBtc->iPage < 0) break;
+
+    pBtc->pPage = pBtc->pgStack[pBtc->iPage];
+    pBtc->idx = pBtc->idxStack[pBtc->iPage];
+  }
+
+  return 0;
+}
+// TDB_BTREE_CURSOR
+
+// TDB_BTREE_DEBUG =====================
 #ifndef NODEBUG
 typedef struct {
   SPgno pgno;
@@ -1304,17 +1353,14 @@ typedef struct {
 SBtPageInfo btPageInfos[20];
 
 void tdbBtPageInfo(SPage *pPage, int idx) {
-  u8           flags;
   SBtPageInfo *pBtPageInfo;
 
   pBtPageInfo = btPageInfos + idx;
 
   pBtPageInfo->pgno = TDB_PAGE_PGNO(pPage);
 
-  flags = TDB_BTREE_PAGE_GET_FLAGS(pPage);
+  pBtPageInfo->root = TDB_BTREE_PAGE_IS_ROOT(pPage);
-
+  pBtPageInfo->leaf = TDB_BTREE_PAGE_IS_LEAF(pPage);
-  pBtPageInfo->root = TDB_BTREE_PAGE_IS_ROOT(flags);
-  pBtPageInfo->leaf = TDB_BTREE_PAGE_IS_LEAF(flags);
 
   pBtPageInfo->rChild = 0;
   if (!pBtPageInfo->leaf) {
@@ -1325,3 +1371,4 @@ void tdbBtPageInfo(SPage *pPage, int idx) {
   pBtPageInfo->nOvfl = pPage->nOverflow;
 }
 #endif
+// TDB_BTREE_DEBUG

source/libs/tdb/src/db/tdbDb.c

@@ -96,6 +96,10 @@ int tdbDbGet(TDB *pDb, const void *pKey, int kLen, void **ppVal, int *vLen) {
   return tdbBtreeGet(pDb->pBt, pKey, kLen, ppVal, vLen);
 }
 
+int tdbDbPGet(TDB *pDb, const void *pKey, int kLen, void **ppKey, int *pkLen, void **ppVal, int *vLen) {
+  return tdbBtreePGet(pDb->pBt, pKey, kLen, ppKey, pkLen, ppVal, vLen);
+}
+
 int tdbDbcOpen(TDB *pDb, TDBC **ppDbc) {
   int   ret;
   TDBC *pDbc = NULL;
@@ -131,3 +135,9 @@ int tdbDbcClose(TDBC *pDbc) {
 
   return 0;
 }
+
+int tdbDbcInsert(TDBC *pDbc, const void *pKey, int keyLen, const void *pVal, int valLen) {
+  // TODO
+  ASSERT(0);
+  return 0;
+}

source/libs/tdb/src/db/tdbPCache.c

@@ -122,7 +122,7 @@ static void tdbPCacheClearLock(SPCache *pCache) { tdbMutexDestroy(&(pCache->mute
 
 static void tdbPCacheLock(SPCache *pCache) { tdbMutexLock(&(pCache->mutex)); }
 
-static void tdbPCacheUnlock(SPCache *pCache) { tdbMutexDestroy(&(pCache->mutex)); }
+static void tdbPCacheUnlock(SPCache *pCache) { tdbMutexUnlock(&(pCache->mutex)); }
 
 static bool tdbPCacheLocked(SPCache *pCache) {
   assert(0);

source/libs/tdb/src/db/tdbPage.c

@@ -242,6 +242,18 @@ void tdbPageCopy(SPage *pFromPage, SPage *pToPage) {
   pToPage->nOverflow = pFromPage->nOverflow;
 }
 
+int tdbPageCapacity(int pageSize, int amHdrSize) {
+  int szPageHdr;
+
+  if (pageSize < 65536) {
+    szPageHdr = pageMethods.szPageHdr;
+  } else {
+    szPageHdr = pageLargeMethods.szPageHdr;
+  }
+
+  return pageSize - szPageHdr - amHdrSize;
+}
+
 static int tdbPageAllocate(SPage *pPage, int szCell, SCell **ppCell) {
   SCell *pFreeCell;
   u8    *pOffset;
@@ -504,3 +516,80 @@ SPageMethods pageMethods = {
     getPageFreeCellInfo,  // getFreeCellInfo
     setPageFreeCellInfo   // setFreeCellInfo
 };
+
+typedef struct __attribute__((__packed__)) {
+  u8 cellNum[3];
+  u8 cellBody[3];
+  u8 cellFree[3];
+  u8 nFree[3];
+} SPageHdrL;
+
+typedef struct __attribute__((__packed__)) {
+  u8 szCell[3];
+  u8 nxOffset[3];
+} SFreeCellL;
+
+// cellNum
+static inline int  getLPageCellNum(SPage *pPage) { return TDB_GET_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellNum); }
+static inline void setLPageCellNum(SPage *pPage, int cellNum) {
+  TDB_PUT_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellNum, cellNum);
+}
+
+// cellBody
+static inline int  getLPageCellBody(SPage *pPage) { return TDB_GET_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellBody); }
+static inline void setLPageCellBody(SPage *pPage, int cellBody) {
+  TDB_PUT_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellBody, cellBody);
+}
+
+// cellFree
+static inline int  getLPageCellFree(SPage *pPage) { return TDB_GET_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellFree); }
+static inline void setLPageCellFree(SPage *pPage, int cellFree) {
+  TDB_PUT_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellFree, cellFree);
+}
+
+// nFree
+static inline int  getLPageNFree(SPage *pPage) { return TDB_GET_U24(((SPageHdrL *)(pPage->pPageHdr))[0].nFree); }
+static inline void setLPageNFree(SPage *pPage, int nFree) {
+  TDB_PUT_U24(((SPageHdrL *)(pPage->pPageHdr))[0].nFree, nFree);
+}
+
+// cell offset
+static inline int getLPageCellOffset(SPage *pPage, int idx) {
+  ASSERT(idx >= 0 && idx < getPageCellNum(pPage));
+  return TDB_GET_U24(pPage->pCellIdx + 3 * idx);
+}
+
+static inline void setLPageCellOffset(SPage *pPage, int idx, int offset) {
+  TDB_PUT_U24(pPage->pCellIdx + 3 * idx, offset);
+}
+
+// free cell info
+static inline void getLPageFreeCellInfo(SCell *pCell, int *szCell, int *nxOffset) {
+  SFreeCellL *pFreeCell = (SFreeCellL *)pCell;
+  *szCell = TDB_GET_U24(pFreeCell->szCell);
+  *nxOffset = TDB_GET_U24(pFreeCell->nxOffset);
+}
+
+static inline void setLPageFreeCellInfo(SCell *pCell, int szCell, int nxOffset) {
+  SFreeCellL *pFreeCell = (SFreeCellL *)pCell;
+  TDB_PUT_U24(pFreeCell->szCell, szCell);
+  TDB_PUT_U24(pFreeCell->nxOffset, nxOffset);
+}
+
+SPageMethods pageLargeMethods = {
+    3,                     // szOffset
+    sizeof(SPageHdrL),     // szPageHdr
+    sizeof(SFreeCellL),    // szFreeCell
+    getLPageCellNum,       // getCellNum
+    setLPageCellNum,       // setCellNum
+    getLPageCellBody,      // getCellBody
+    setLPageCellBody,      // setCellBody
+    getLPageCellFree,      // getCellFree
+    setLPageCellFree,      // setCellFree
+    getLPageNFree,         // getFreeBytes
+    setLPageNFree,         // setFreeBytes
+    getLPageCellOffset,    // getCellOffset
+    setLPageCellOffset,    // setCellOffset
+    getLPageFreeCellInfo,  // getFreeCellInfo
+    setLPageFreeCellInfo   // setFreeCellInfo
+};

source/libs/tdb/src/inc/tdbBtree.h

@@ -35,7 +35,6 @@ struct SBTC {
   int     idx;
   int     idxStack[BTREE_MAX_DEPTH + 1];
   SPage  *pgStack[BTREE_MAX_DEPTH + 1];
-  void   *pBuf;
 };
 
 // SBTree
@@ -43,12 +42,12 @@ int tdbBtreeOpen(int keyLen, int valLen, SPager *pFile, FKeyComparator kcmpr, SB
 int tdbBtreeClose(SBTree *pBt);
 int tdbBtCursorInsert(SBTC *pCur, const void *pKey, int kLen, const void *pVal, int vLen);
 int tdbBtreeGet(SBTree *pBt, const void *pKey, int kLen, void **ppVal, int *vLen);
+int tdbBtreePGet(SBTree *pBt, const void *pKey, int kLen, void **ppKey, int *pkLen, void **ppVal, int *vLen);
 
 // SBTC
 int tdbBtcOpen(SBTC *pCur, SBTree *pBt);
 int tdbBtcMoveToFirst(SBTC *pBtc);
 int tdbBtcMoveToLast(SBTC *pBtc);
-int tdbBtcMoveTo(SBTC *pBtc, const void *pKey, int kLen);
 int tdbBtreeNext(SBTC *pBtc, void **ppKey, int *kLen, void **ppVal, int *vLen);
 int tdbBtcClose(SBTC *pBtc);
 

source/libs/tdb/src/inc/tdbDb.h

@@ -29,11 +29,13 @@ int tdbDbClose(TDB *pDb);
 int tdbDbDrop(TDB *pDb);
 int tdbDbInsert(TDB *pDb, const void *pKey, int keyLen, const void *pVal, int valLen);
 int tdbDbGet(TDB *pDb, const void *pKey, int kLen, void **ppVal, int *vLen);
+int tdbDbPGet(TDB *pDb, const void *pKey, int kLen, void **ppKey, int *pkLen, void **ppVal, int *vLen);
 
 // TDBC
 int tdbDbcOpen(TDB *pDb, TDBC **ppDbc);
 int tdbDbNext(TDBC *pDbc, void **ppKey, int *kLen, void **ppVal, int *vLen);
 int tdbDbcClose(TDBC *pDbc);
+int tdbDbcInsert(TDBC *pDbc, const void *pKey, int keyLen, const void *pVal, int valLen);
 
 #ifdef __cplusplus
 }

source/libs/tdb/src/inc/tdbOs.h

@@ -64,7 +64,7 @@ typedef TdThreadSpinlock tdb_spinlock_t;
 #define tdbSpinlockDestroy taosThreadSpinDestroy
 #define tdbSpinlockLock    taosThreadSpinLock
 #define tdbSpinlockUnlock  taosThreadSpinUnlock
-#define tdbSpinlockTrylock pthread_spin_trylock
+#define tdbSpinlockTrylock taosThreadSpinTrylock
 
 /* mutex lock */
 typedef TdThreadMutex tdb_mutex_t;

source/libs/tdb/src/inc/tdbPage.h

@@ -111,6 +111,7 @@ void tdbPageInit(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell
 int  tdbPageInsertCell(SPage *pPage, int idx, SCell *pCell, int szCell, u8 asOvfl);
 int  tdbPageDropCell(SPage *pPage, int idx);
 void tdbPageCopy(SPage *pFromPage, SPage *pToPage);
+int  tdbPageCapacity(int pageSize, int amHdrSize);
 
 static inline SCell *tdbPageGetCell(SPage *pPage, int idx) {
   SCell *pCell;

source/libs/tdb/src/inc/tdbTxn.h

@@ -20,7 +20,7 @@
 extern "C" {
 #endif
 
-typedef struct STxn STXN;
+typedef struct STxn TXN;
 
 struct STxn {
   u64 txnId;

source/libs/tdb/src/page/tdbPageL.c

@@ -1,93 +0,0 @@
-/*
- * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
- *
- * This program is free software: you can use, redistribute, and/or modify
- * it under the terms of the GNU Affero General Public License, version 3
- * or later ("AGPL"), as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.
- *
- * You should have received a copy of the GNU Affero General Public License
- * along with this program. If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "tdbInt.h"
-
-typedef struct __attribute__((__packed__)) {
-  u8 cellNum[3];
-  u8 cellBody[3];
-  u8 cellFree[3];
-  u8 nFree[3];
-} SPageHdrL;
-
-typedef struct __attribute__((__packed__)) {
-  u8 szCell[3];
-  u8 nxOffset[3];
-} SFreeCellL;
-
-// cellNum
-static inline int  getPageCellNum(SPage *pPage) { return TDB_GET_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellNum); }
-static inline void setPageCellNum(SPage *pPage, int cellNum) {
-  TDB_PUT_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellNum, cellNum);
-}
-
-// cellBody
-static inline int  getPageCellBody(SPage *pPage) { return TDB_GET_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellBody); }
-static inline void setPageCellBody(SPage *pPage, int cellBody) {
-  TDB_PUT_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellBody, cellBody);
-}
-
-// cellFree
-static inline int  getPageCellFree(SPage *pPage) { return TDB_GET_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellFree); }
-static inline void setPageCellFree(SPage *pPage, int cellFree) {
-  TDB_PUT_U24(((SPageHdrL *)(pPage->pPageHdr))[0].cellFree, cellFree);
-}
-
-// nFree
-static inline int  getPageNFree(SPage *pPage) { return TDB_GET_U24(((SPageHdrL *)(pPage->pPageHdr))[0].nFree); }
-static inline void setPageNFree(SPage *pPage, int nFree) {
-  TDB_PUT_U24(((SPageHdrL *)(pPage->pPageHdr))[0].nFree, nFree);
-}
-
-// cell offset
-static inline int getPageCellOffset(SPage *pPage, int idx) {
-  ASSERT(idx >= 0 && idx < getPageCellNum(pPage));
-  return TDB_GET_U24(pPage->pCellIdx + 3 * idx);
-}
-
-static inline void setPageCellOffset(SPage *pPage, int idx, int offset) {
-  TDB_PUT_U24(pPage->pCellIdx + 3 * idx, offset);
-}
-
-// free cell info
-static inline void getPageFreeCellInfo(SCell *pCell, int *szCell, int *nxOffset) {
-  SFreeCellL *pFreeCell = (SFreeCellL *)pCell;
-  *szCell = TDB_GET_U24(pFreeCell->szCell);
-  *nxOffset = TDB_GET_U24(pFreeCell->nxOffset);
-}
-
-static inline void setPageFreeCellInfo(SCell *pCell, int szCell, int nxOffset) {
-  SFreeCellL *pFreeCell = (SFreeCellL *)pCell;
-  TDB_PUT_U24(pFreeCell->szCell, szCell);
-  TDB_PUT_U24(pFreeCell->nxOffset, nxOffset);
-}
-
-SPageMethods pageLargeMethods = {
-    3,                    // szOffset
-    sizeof(SPageHdrL),    // szPageHdr
-    sizeof(SFreeCellL),   // szFreeCell
-    getPageCellNum,       // getCellNum
-    setPageCellNum,       // setCellNum
-    getPageCellBody,      // getCellBody
-    setPageCellBody,      // setCellBody
-    getPageCellFree,      // getCellFree
-    setPageCellFree,      // setCellFree
-    getPageNFree,         // getFreeBytes
-    setPageNFree,         // setFreeBytes
-    getPageCellOffset,    // getCellOffset
-    setPageCellOffset,    // setCellOffset
-    getPageFreeCellInfo,  // getFreeCellInfo
-    setPageFreeCellInfo   // setFreeCellInfo
-};

source/os/src/osMemory.c

@@ -17,6 +17,8 @@
 #include <malloc.h>
 #include "os.h"
 
+#ifdef USE_TD_MEMORY
+
 #define TD_MEMORY_SYMBOL ('T'<<24|'A'<<16|'O'<<8|'S')
 
 #define TD_MEMORY_STACK_TRACE_DEPTH 10
@@ -70,6 +72,8 @@ int32_t taosBackTrace(void **buffer, int32_t size) {
 //   return backtrace_symbols(buffer, *size);
 // }
 
+#endif
+
 void *taosMemoryMalloc(int32_t size) {
 #ifdef USE_TD_MEMORY
   void *tmp = malloc(size + sizeof(TdMemoryInfo));
@@ -126,9 +130,9 @@ void *taosMemoryRealloc(void *ptr, int32_t size) {
 }
 
 void taosMemoryFree(const void *ptr) {
-#ifdef USE_TD_MEMORY
   if (ptr == NULL) return;
 
+#ifdef USE_TD_MEMORY
   TdMemoryInfoPtr pTdMemoryInfo = (TdMemoryInfoPtr)((char*)ptr - sizeof(TdMemoryInfo));
   if(pTdMemoryInfo->symbol == TD_MEMORY_SYMBOL) {
     pTdMemoryInfo->memorySize = 0;
@@ -143,9 +147,9 @@ void taosMemoryFree(const void *ptr) {
 }
 
 int32_t taosMemorySize(void *ptr) {
-#ifdef USE_TD_MEMORY
   if (ptr == NULL) return 0;
 
+#ifdef USE_TD_MEMORY
   TdMemoryInfoPtr pTdMemoryInfo = (TdMemoryInfoPtr)((char*)ptr - sizeof(TdMemoryInfo));
   assert(pTdMemoryInfo->symbol == TD_MEMORY_SYMBOL);
 

source/os/src/osProc.c

@@ -13,4 +13,15 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "consumer.h"
+#define ALLOW_FORBID_FUNC
+#define _DEFAULT_SOURCE
+#include "os.h"
+
+int32_t taosNewProc(const char *args) {
+  return 0;
+}
+
+void taosSetProcName(char **argv, const char *name) {
+  prctl(PR_SET_NAME, name);
+  strcpy(argv[0], name);
+}

source/os/src/osShm.c

@@ -0,0 +1,59 @@
+/*
+ * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#define ALLOW_FORBID_FUNC
+#define _DEFAULT_SOURCE
+#include "os.h"
+
+int32_t taosCreateShm(SShm* pShm, int32_t shmsize) {
+  pShm->id = -1;
+
+  int32_t shmid = shmget(0X95279527, shmsize, IPC_CREAT | 0600);
+  if (shmid < 0) {
+    return -1;
+  }
+
+  void* shmptr = shmat(shmid, NULL, 0);
+  if (shmptr == NULL) {
+    return -1;
+  }
+
+  pShm->id = shmid;
+  pShm->size = shmsize;
+  pShm->ptr = shmptr;
+  return 0;
+}
+
+void taosDropShm(SShm* pShm) {
+  if (pShm->id >= 0) {
+    if (pShm->ptr != NULL) {
+      shmdt(pShm->ptr);
+    }
+    shmctl(pShm->id, IPC_RMID, NULL);
+  }
+  pShm->id = -1;
+  pShm->size = 0;
+  pShm->ptr = NULL;
+}
+
+int32_t taosAttachShm(SShm* pShm) {
+  errno = 0;
+
+  void* ptr = shmat(pShm->id, NULL, 0);
+  if (errno == 0) {
+    pShm->ptr = ptr;
+  }
+  return errno;
+}

source/os/src/osSignal.c

@@ -71,4 +71,6 @@ void taosIgnSignal(int32_t signum) { signal(signum, SIG_IGN); }
 
 void taosDflSignal(int32_t signum) { signal(signum, SIG_DFL); }
 
+void taosKillChildOnSelfStopped() { prctl(PR_SET_PDEATHSIG, SIGKILL); }
+
 #endif

source/os/src/osSocket.c

@@ -758,7 +758,7 @@ void taosBlockSIGPIPE() {
   sigset_t signal_mask;
   sigemptyset(&signal_mask);
   sigaddset(&signal_mask, SIGPIPE);
-  int32_t rc = taosThreadSigmask(SIG_BLOCK, &signal_mask, NULL);
+  int32_t rc = taosThreadSigMask(SIG_BLOCK, &signal_mask, NULL);
   if (rc != 0) {
     // printf("failed to block SIGPIPE");
   }
@@ -876,7 +876,7 @@ void taosSetMaskSIGPIPE() {
   sigset_t signal_mask;
   sigemptyset(&signal_mask);
   sigaddset(&signal_mask, SIGPIPE);
-  int32_t rc = taosThreadSigmask(SIG_SETMASK, &signal_mask, NULL);
+  int32_t rc = taosThreadSigMask(SIG_SETMASK, &signal_mask, NULL);
   if (rc != 0) {
     // printf("failed to setmask SIGPIPE");
   }