Merge branch '3.0' of github.com:taosdata/TDengine into 3.0

examples/c/stream_demo.c

@@ -25,7 +25,7 @@ int32_t init_env() {
     return -1;
   }
 
-  TAOS_RES* pRes = taos_query(pConn, "create database if not exists abc1 vgroups 1");
+  TAOS_RES* pRes = taos_query(pConn, "create database if not exists abc1 vgroups 2");
   if (taos_errno(pRes) != 0) {
     printf("error in create db, reason:%s\n", taos_errstr(pRes));
     return -1;
@@ -82,7 +82,7 @@ int32_t create_stream() {
   /*const char* sql = "select sum(k) from tu1 interval(10m)";*/
   /*pRes = tmq_create_stream(pConn, "stream1", "out1", sql);*/
   pRes = taos_query(
-      pConn, "create stream stream1 trigger at_once into outstb as select _wstartts, sum(k) from tu1 interval(10m)");
+      pConn, "create stream stream1 trigger at_once into outstb as select _wstartts, sum(k) from st1 interval(10m)");
   if (taos_errno(pRes) != 0) {
     printf("failed to create stream stream1, reason:%s\n", taos_errstr(pRes));
     return -1;

include/common/tcommon.h

@@ -77,7 +77,9 @@ typedef struct SDataBlockInfo {
   int16_t     numOfCols;
   int16_t     hasVarCol;
   int32_t     capacity;
-  EStreamType type;
+  // TODO: optimize and remove following
+  int32_t     childId;  // used for stream, do not serialize
+  EStreamType type;     // used for stream, do not serialize
 } SDataBlockInfo;
 
 typedef struct SSDataBlock {
@@ -105,14 +107,14 @@ typedef struct SColumnInfoData {
 } SColumnInfoData;
 
 typedef struct SQueryTableDataCond {
-  //STimeWindow  twindow;
+  // STimeWindow  twindow;
   int32_t      order;  // desc|asc order to iterate the data block
   int32_t      numOfCols;
-  SColumnInfo *colList;
+  SColumnInfo* colList;
   bool         loadExternalRows;  // load external rows or not
   int32_t      type;              // data block load type:
   int32_t      numOfTWindows;
-  STimeWindow *twindows;
+  STimeWindow* twindows;
 } SQueryTableDataCond;
 
 void*   blockDataDestroy(SSDataBlock* pBlock);
@@ -205,7 +207,7 @@ typedef struct {
   const char* key;
   int32_t     keyLen;
   uint8_t     type;
-  union{
+  union {
     const char* value;
     int64_t     i;
     uint64_t    u;

include/common/tmsg.h

@@ -933,6 +933,7 @@ typedef struct {
   int64_t numOfProcessedFetch;
   int64_t numOfProcessedDrop;
   int64_t numOfProcessedHb;
+  int64_t numOfProcessedDelete;
   int64_t cacheDataSize;
   int64_t numOfQueryInQueue;
   int64_t numOfFetchInQueue;
@@ -1126,8 +1127,8 @@ typedef struct {
   STableMetaRsp* pMeta;
 } SMAlterStbRsp;
 
-int32_t tEncodeSMAlterStbRsp(SEncoder *pEncoder, const SMAlterStbRsp *pRsp);
+int32_t tEncodeSMAlterStbRsp(SEncoder* pEncoder, const SMAlterStbRsp* pRsp);
-int32_t tDecodeSMAlterStbRsp(SDecoder *pDecoder, SMAlterStbRsp *pRsp);
+int32_t tDecodeSMAlterStbRsp(SDecoder* pDecoder, SMAlterStbRsp* pRsp);
 void    tFreeSMAlterStbRsp(SMAlterStbRsp* pRsp);
 
 int32_t tSerializeSTableMetaRsp(void* buf, int32_t bufLen, STableMetaRsp* pRsp);
@@ -2319,7 +2320,7 @@ typedef struct {
   int64_t   sliding;
   char*     expr;  // sma expression
   char*     tagsFilter;
-  SVgEpSet vgEpSet[];
+  SVgEpSet* pVgEpSet;
 } STSma;  // Time-range-wise SMA
 
 typedef STSma SVCreateTSmaReq;
@@ -2689,20 +2690,20 @@ int32_t tEncodeSVSubmitReq(SEncoder* pCoder, const SVSubmitReq* pReq);
 int32_t tDecodeSVSubmitReq(SDecoder* pCoder, SVSubmitReq* pReq);
 
 typedef struct {
-  int64_t     delUid;
+  SMsgHead header;
-  int64_t     tbUid;  // super/child/normal table
+  uint64_t sId;
-  int8_t      type;   // table type
+  uint64_t queryId;
-  int16_t     nWnds;
+  uint64_t taskId;
-  char*       tbFullName;
+  uint32_t sqlLen;
-  char*       subPlan;
+  uint32_t phyLen;
-  STimeWindow wnds[];
+  char*    sql;
+  char*    msg;
 } SVDeleteReq;
 
-int32_t tEncodeSVDeleteReq(SEncoder* pCoder, const SVDeleteReq* pReq);
+int32_t tSerializeSVDeleteReq(void *buf, int32_t bufLen, SVDeleteReq *pReq);
-int32_t tDecodeSVDeleteReq(SDecoder* pCoder, SVDeleteReq* pReq);
+int32_t tDeserializeSVDeleteReq(void *buf, int32_t bufLen, SVDeleteReq *pReq);
 
 typedef struct {
-  int32_t code;
   int64_t affectedRows;
 } SVDeleteRsp;
 

include/common/tmsgdef.h

@@ -187,6 +187,7 @@ enum {
   TD_DEF_MSG_TYPE(TDMT_VND_STREAM_TRIGGER, "vnode-stream-trigger", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_TASK_RUN, "vnode-stream-task-run", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_TASK_DISPATCH, "vnode-stream-task-dispatch", NULL, NULL)
+  TD_DEF_MSG_TYPE(TDMT_VND_TASK_DISPATCH_WRITE, "vnode-stream-task-dispatch-write", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_TASK_RECOVER, "vnode-stream-task-recover", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_CREATE_SMA, "vnode-create-sma", NULL, NULL)
   TD_DEF_MSG_TYPE(TDMT_VND_CANCEL_SMA, "vnode-cancel-sma", NULL, NULL)

include/libs/executor/dataSinkMgt.h

@@ -32,6 +32,18 @@ extern "C" {
 struct SDataSink;
 struct SSDataBlock;
 
+typedef struct SDeleterRes {
+  uint64_t uid;
+  SArray*  uidList;
+  int64_t  skey;
+  int64_t  ekey;
+  int64_t  affectedRows;
+} SDeleterRes;
+
+typedef struct SDeleterParam {
+  SArray* pUidList;
+} SDeleterParam;
+
 typedef struct SDataSinkStat {
   uint64_t cachedSize;
 } SDataSinkStat;
@@ -64,7 +76,7 @@ typedef struct SOutputData {
  * @param pHandle output
  * @return error code
  */
-int32_t dsCreateDataSinker(const SDataSinkNode* pDataSink, DataSinkHandle* pHandle);
+int32_t dsCreateDataSinker(const SDataSinkNode* pDataSink, DataSinkHandle* pHandle, void* pParam);
 
 int32_t dsDataSinkGetCacheSize(SDataSinkStat *pStat);
 

include/libs/nodes/nodes.h

@@ -195,6 +195,7 @@ typedef enum ENodeType {
   QUERY_NODE_LOGIC_PLAN_FILL,
   QUERY_NODE_LOGIC_PLAN_SORT,
   QUERY_NODE_LOGIC_PLAN_PARTITION,
+  QUERY_NODE_LOGIC_PLAN_INDEF_ROWS_FUNC,
   QUERY_NODE_LOGIC_SUBPLAN,
   QUERY_NODE_LOGIC_PLAN,
 
@@ -211,7 +212,7 @@ typedef enum ENodeType {
   QUERY_NODE_PHYSICAL_PLAN_MERGE,
   QUERY_NODE_PHYSICAL_PLAN_SORT,
   QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL,
-  QUERY_NODE_PHYSICAL_PLAN_SORT_MERGE_INTERVAL,
+  QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL,
   QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL,
   QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_INTERVAL,
   QUERY_NODE_PHYSICAL_PLAN_STREAM_SEMI_INTERVAL,
@@ -222,6 +223,7 @@ typedef enum ENodeType {
   QUERY_NODE_PHYSICAL_PLAN_STATE_WINDOW,
   QUERY_NODE_PHYSICAL_PLAN_STREAM_STATE_WINDOW,
   QUERY_NODE_PHYSICAL_PLAN_PARTITION,
+  QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC,
   QUERY_NODE_PHYSICAL_PLAN_DISPATCH,
   QUERY_NODE_PHYSICAL_PLAN_INSERT,
   QUERY_NODE_PHYSICAL_PLAN_DELETE,

include/libs/nodes/plannodes.h

@@ -42,6 +42,7 @@ typedef struct SScanLogicNode {
   SNodeList*    pScanPseudoCols;
   int8_t        tableType;
   uint64_t      tableId;
+  uint64_t      stableId;
   SVgroupsInfo* pVgroupList;
   EScanType     scanType;
   uint8_t       scanSeq[2];  // first is scan count, and second is reverse scan count
@@ -86,6 +87,11 @@ typedef struct SProjectLogicNode {
   int64_t    soffset;
 } SProjectLogicNode;
 
+typedef struct SIndefRowsFuncLogicNode {
+  SLogicNode node;
+  SNodeList* pVectorFuncs;
+} SIndefRowsFuncLogicNode;
+
 typedef enum EModifyTableType { MODIFY_TABLE_TYPE_INSERT = 1, MODIFY_TABLE_TYPE_DELETE } EModifyTableType;
 
 typedef struct SVnodeModifyLogicNode {
@@ -94,7 +100,7 @@ typedef struct SVnodeModifyLogicNode {
   int32_t          msgType;
   SArray*          pDataBlocks;
   SVgDataBlocks*   pVgDataBlocks;
-  SNode*           pModifyRows;  // SColumnNode
+  SNode*           pAffectedRows;  // SColumnNode
   uint64_t         tableId;
   int8_t           tableType;  // table type
   char             tableFName[TSDB_TABLE_FNAME_LEN];
@@ -109,6 +115,7 @@ typedef struct SExchangeLogicNode {
 typedef struct SMergeLogicNode {
   SLogicNode node;
   SNodeList* pMergeKeys;
+  SNodeList* pInputs;
   int32_t    numOfChannels;
   int32_t    srcGroupId;
 } SMergeLogicNode;
@@ -117,7 +124,7 @@ typedef enum EWindowType { WINDOW_TYPE_INTERVAL = 1, WINDOW_TYPE_SESSION, WINDOW
 
 typedef enum EIntervalAlgorithm {
   INTERVAL_ALGO_HASH = 1,
-  INTERVAL_ALGO_SORT_MERGE,
+  INTERVAL_ALGO_MERGE,
   INTERVAL_ALGO_STREAM_FINAL,
   INTERVAL_ALGO_STREAM_SEMI,
   INTERVAL_ALGO_STREAM_SINGLE,
@@ -220,6 +227,7 @@ typedef struct SScanPhysiNode {
   SNodeList* pScanCols;
   SNodeList* pScanPseudoCols;
   uint64_t   uid;  // unique id of the table
+  uint64_t   suid;
   int8_t     tableType;
   SName      tableName;
 } SScanPhysiNode;
@@ -264,6 +272,12 @@ typedef struct SProjectPhysiNode {
   int64_t    soffset;
 } SProjectPhysiNode;
 
+typedef struct SIndefRowsFuncPhysiNode {
+  SPhysiNode node;
+  SNodeList* pExprs;
+  SNodeList* pVectorFuncs;
+} SIndefRowsFuncPhysiNode;
+
 typedef struct SJoinPhysiNode {
   SPhysiNode node;
   EJoinType  joinType;
@@ -296,6 +310,7 @@ typedef struct SExchangePhysiNode {
 typedef struct SMergePhysiNode {
   SPhysiNode node;
   SNodeList* pMergeKeys;
+  SNodeList* pTargets;
   int32_t    numOfChannels;
   int32_t    srcGroupId;
 } SMergePhysiNode;
@@ -319,7 +334,7 @@ typedef struct SIntervalPhysiNode {
   int8_t           slidingUnit;
 } SIntervalPhysiNode;
 
-typedef SIntervalPhysiNode SSortMergeIntervalPhysiNode;
+typedef SIntervalPhysiNode SMergeIntervalPhysiNode;
 typedef SIntervalPhysiNode SStreamIntervalPhysiNode;
 typedef SIntervalPhysiNode SStreamFinalIntervalPhysiNode;
 typedef SIntervalPhysiNode SStreamSemiIntervalPhysiNode;
@@ -388,6 +403,7 @@ typedef struct SDataDeleterNode {
   int8_t        tableType;  // table type
   char          tableFName[TSDB_TABLE_FNAME_LEN];
   STimeWindow   deleteTimeRange;
+  SNode*        pAffectedRows;
 } SDataDeleterNode;
 
 typedef struct SSubplan {

include/libs/qcom/query.h

@@ -224,7 +224,7 @@ extern int32_t (*queryProcessMsgRsp[TDMT_MAX])(void* output, char* msg, int32_t
 #define NEED_SCHEDULER_RETRY_ERROR(_code) \
   ((_code) == TSDB_CODE_RPC_REDIRECT || (_code) == TSDB_CODE_RPC_NETWORK_UNAVAIL || (_code) == TSDB_CODE_SCH_TIMEOUT_ERROR)
 
-#define REQUEST_MAX_TRY_TIMES 1
+#define REQUEST_TOTAL_EXEC_TIMES 2
 
 #define qFatal(...)                                                                           \
   do {                                                                                        \

include/libs/qworker/qworker.h

@@ -31,7 +31,12 @@ enum {
   NODE_TYPE_MNODE,
 };
 
-
+typedef struct SDeleteRes {
+  uint64_t uid;
+  SArray*  uidList;
+  int64_t  skey;
+  int64_t  ekey;
+} SDeleteRes;
 
 typedef struct SQWorkerCfg {
   uint32_t maxSchedulerNum;
@@ -47,6 +52,7 @@ typedef struct {
   uint64_t fetchProcessed;
   uint64_t dropProcessed;
   uint64_t hbProcessed;
+  uint64_t deleteProcessed;
   
   uint64_t numOfQueryInQueue;
   uint64_t numOfFetchInQueue;
@@ -74,6 +80,8 @@ int32_t qWorkerProcessDropMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, int6
 
 int32_t qWorkerProcessHbMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, int64_t ts);
 
+int32_t qWorkerProcessDeleteMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, SRpcMsg *pRsp, SDeleteRes *pRes);
+
 void qWorkerDestroy(void **qWorkerMgmt);
 
 int32_t qWorkerGetStat(SReadHandle *handle, void *qWorkerMgmt, SQWorkerStat *pStat);

include/libs/stream/tstream.h

@@ -24,8 +24,8 @@
 extern "C" {
 #endif
 
-#ifndef _TSTREAM_H_
+#ifndef _STREAM_H_
-#define _TSTREAM_H_
+#define _STREAM_H_
 
 typedef struct SStreamTask SStreamTask;
 
@@ -39,6 +39,7 @@ enum {
   TASK_INPUT_STATUS__NORMAL = 1,
   TASK_INPUT_STATUS__BLOCKED,
   TASK_INPUT_STATUS__RECOVER,
+  TASK_INPUT_STATUS__PROCESSING,
   TASK_INPUT_STATUS__STOP,
   TASK_INPUT_STATUS__FAILED,
 };
@@ -60,6 +61,10 @@ enum {
   STREAM_INPUT__CHECKPOINT,
 };
 
+typedef struct {
+  int8_t type;
+} SStreamQueueItem;
+
 typedef struct {
   int8_t      type;
   int64_t     ver;
@@ -80,55 +85,51 @@ typedef struct {
   int8_t type;
 } SStreamCheckpoint;
 
+enum {
+  STREAM_QUEUE__SUCESS = 1,
+  STREAM_QUEUE__FAILED,
+  STREAM_QUEUE__PROCESSING,
+};
+
 typedef struct {
   STaosQueue* queue;
   STaosQall*  qall;
   void*       qItem;
-  int8_t      failed;
+  int8_t      status;
-} SStreamQ;
+} SStreamQueue;
 
-static FORCE_INLINE void* streamQCurItem(SStreamQ* queue) {
+SStreamQueue* streamQueueOpen();
-  //
+void          streamQueueClose(SStreamQueue* queue);
-  return queue->qItem;
+
+static FORCE_INLINE void streamQueueProcessSuccess(SStreamQueue* queue) {
+  ASSERT(atomic_load_8(&queue->status) == STREAM_QUEUE__PROCESSING);
+  queue->qItem = NULL;
+  atomic_store_8(&queue->status, STREAM_QUEUE__SUCESS);
 }
 
-static FORCE_INLINE void* streamQNextItem(SStreamQ* queue) {
+static FORCE_INLINE void streamQueueProcessFail(SStreamQueue* queue) {
-  int8_t failed = atomic_load_8(&queue->failed);
+  ASSERT(atomic_load_8(&queue->status) == STREAM_QUEUE__PROCESSING);
-  if (failed) {
+  atomic_store_8(&queue->status, STREAM_QUEUE__FAILED);
+}
+
+static FORCE_INLINE void* streamQueueCurItem(SStreamQueue* queue) { return queue->qItem; }
+
+static FORCE_INLINE void* streamQueueNextItem(SStreamQueue* queue) {
+  int8_t dequeueFlag = atomic_exchange_8(&queue->status, STREAM_QUEUE__PROCESSING);
+  if (dequeueFlag == STREAM_QUEUE__FAILED) {
     ASSERT(queue->qItem != NULL);
-    return streamQCurItem(queue);
+    return streamQueueCurItem(queue);
   } else {
     taosGetQitem(queue->qall, &queue->qItem);
     if (queue->qItem == NULL) {
       taosReadAllQitems(queue->queue, queue->qall);
       taosGetQitem(queue->qall, &queue->qItem);
     }
-    return streamQCurItem(queue);
+    return streamQueueCurItem(queue);
   }
 }
 
-static FORCE_INLINE void streamQSetFail(SStreamQ* queue) { atomic_store_8(&queue->failed, 1); }
+SStreamDataSubmit* streamDataSubmitNew(SSubmitReq* pReq);
-
-static FORCE_INLINE void streamQSetSuccess(SStreamQ* queue) { atomic_store_8(&queue->failed, 0); }
-
-static FORCE_INLINE SStreamDataSubmit* streamDataSubmitNew(SSubmitReq* pReq) {
-  SStreamDataSubmit* pDataSubmit = (SStreamDataSubmit*)taosAllocateQitem(sizeof(SStreamDataSubmit), DEF_QITEM);
-  if (pDataSubmit == NULL) return NULL;
-  pDataSubmit->dataRef = (int32_t*)taosMemoryMalloc(sizeof(int32_t));
-  if (pDataSubmit->dataRef == NULL) goto FAIL;
-  pDataSubmit->data = pReq;
-  *pDataSubmit->dataRef = 1;
-  pDataSubmit->type = STREAM_INPUT__DATA_SUBMIT;
-  return pDataSubmit;
-FAIL:
-  taosFreeQitem(pDataSubmit);
-  return NULL;
-}
-
-static FORCE_INLINE void streamDataSubmitRefInc(SStreamDataSubmit* pDataSubmit) {
-  //
-  atomic_add_fetch_32(pDataSubmit->dataRef, 1);
-}
 
 static FORCE_INLINE void streamDataSubmitRefDec(SStreamDataSubmit* pDataSubmit) {
   int32_t ref = atomic_sub_fetch_32(pDataSubmit->dataRef, 1);
@@ -141,9 +142,31 @@ static FORCE_INLINE void streamDataSubmitRefDec(SStreamDataSubmit* pDataSubmit)
 
 SStreamDataSubmit* streamSubmitRefClone(SStreamDataSubmit* pSubmit);
 
+#if 0
 int32_t streamDataBlockEncode(void** buf, const SStreamDataBlock* pOutput);
 void*   streamDataBlockDecode(const void* buf, SStreamDataBlock* pInput);
 
+static FORCE_INLINE int32_t streamEnqueue1(SStreamQueue* queue, SStreamQueueItem* pItem) {
+  int8_t inputStatus = atomic_load_8(&queue->enqueueStatus);
+  if (inputStatus == TASK_INPUT_STATUS__NORMAL) {
+    if (pItem->type == STREAM_INPUT__DATA_SUBMIT) {
+      SStreamDataSubmit* pSubmitClone = streamSubmitRefClone((SStreamDataSubmit*)pItem);
+      if (pSubmitClone == NULL) {
+        atomic_store_8(&queue->enqueueStatus, TASK_INPUT_STATUS__FAILED);
+        return -1;
+      }
+      taosWriteQitem(queue->queue, pSubmitClone);
+    } else if (pItem->type == STREAM_INPUT__DATA_BLOCK) {
+      taosWriteQitem(queue->queue, pItem);
+    } else if (pItem->type == STREAM_INPUT__CHECKPOINT) {
+      taosWriteQitem(queue->queue, pItem);
+    }
+    return 0;
+  }
+  return 0;
+}
+#endif
+
 typedef struct {
   int8_t parallelizable;
   char*  qmsg;
@@ -236,12 +259,16 @@ struct SStreamTask {
   int8_t  dispatchType;
   int16_t dispatchMsgType;
 
+  // node info
+  int32_t childId;
   int32_t nodeId;
   SEpSet  epSet;
 
   // exec
   STaskExec exec;
 
+  // TODO: merge sink and dispatch
+
   //  local sink
   union {
     STaskSinkTb    tbSink;
@@ -259,24 +286,54 @@ struct SStreamTask {
   int8_t inputStatus;
   int8_t outputStatus;
 
-  STaosQueue* inputQ;
+  SStreamQueue* inputQueue;
-  STaosQall*  inputQAll;
+  SStreamQueue* outputQueue;
-  STaosQueue* outputQ;
-  STaosQall*  outputQAll;
 
   // application storage
   void* ahandle;
 };
 
-SStreamTask* tNewSStreamTask(int64_t streamId);
+SStreamTask* tNewSStreamTask(int64_t streamId, int32_t childId);
 int32_t      tEncodeSStreamTask(SEncoder* pEncoder, const SStreamTask* pTask);
 int32_t      tDecodeSStreamTask(SDecoder* pDecoder, SStreamTask* pTask);
 void         tFreeSStreamTask(SStreamTask* pTask);
 
-typedef struct {
+static FORCE_INLINE int32_t streamTaskInput(SStreamTask* pTask, SStreamQueueItem* pItem) {
-  // SMsgHead     head;
+  while (1) {
-  SStreamTask* task;
+    int8_t inputStatus =
-} SStreamTaskDeployReq;
+        atomic_val_compare_exchange_8(&pTask->inputStatus, TASK_INPUT_STATUS__NORMAL, TASK_INPUT_STATUS__PROCESSING);
+    if (inputStatus == TASK_INPUT_STATUS__NORMAL) {
+      break;
+    }
+    ASSERT(0);
+  }
+
+  if (pItem->type == STREAM_INPUT__DATA_SUBMIT) {
+    SStreamDataSubmit* pSubmitClone = streamSubmitRefClone((SStreamDataSubmit*)pItem);
+    if (pSubmitClone == NULL) {
+      atomic_store_8(&pTask->inputStatus, TASK_INPUT_STATUS__FAILED);
+      return -1;
+    }
+    taosWriteQitem(pTask->inputQueue->queue, pSubmitClone);
+  } else if (pItem->type == STREAM_INPUT__DATA_BLOCK) {
+    taosWriteQitem(pTask->inputQueue->queue, pItem);
+  } else if (pItem->type == STREAM_INPUT__CHECKPOINT) {
+    taosWriteQitem(pTask->inputQueue->queue, pItem);
+  }
+
+  // TODO: back pressure
+  atomic_store_8(&pTask->inputStatus, TASK_INPUT_STATUS__NORMAL);
+  return 0;
+}
+
+static FORCE_INLINE void streamTaskInputFail(SStreamTask* pTask) {
+  atomic_store_8(&pTask->inputStatus, TASK_INPUT_STATUS__FAILED);
+}
+
+static FORCE_INLINE int32_t streamTaskOutput(SStreamTask* pTask, SStreamDataBlock* pBlock) {
+  taosWriteQitem(pTask->outputQueue->queue, pBlock);
+  return 0;
+}
 
 typedef struct {
   int32_t reserved;
@@ -289,6 +346,11 @@ typedef struct {
   SArray* data;  // SArray<SSDataBlock>
 } SStreamTaskExecReq;
 
+typedef struct {
+  // SMsgHead     head;
+  SStreamTask* task;
+} SStreamTaskDeployReq;
+
 int32_t tEncodeSStreamTaskExecReq(void** buf, const SStreamTaskExecReq* pReq);
 void*   tDecodeSStreamTaskExecReq(const void* buf, SStreamTaskExecReq* pReq);
 void    tFreeSStreamTaskExecReq(SStreamTaskExecReq* pReq);
@@ -297,13 +359,6 @@ typedef struct {
   int32_t reserved;
 } SStreamTaskExecRsp;
 
-typedef struct {
-  // SMsgHead head;
-  int64_t streamId;
-  int64_t version;
-  SArray* res;  // SArray<SSDataBlock>
-} SStreamSinkReq;
-
 typedef struct {
   SMsgHead head;
   int64_t  streamId;
@@ -315,10 +370,13 @@ typedef struct {
   int32_t taskId;
   int32_t sourceTaskId;
   int32_t sourceVg;
+  int32_t sourceChildId;
 #if 0
   int64_t sourceVer;
 #endif
-  SArray* data;  // SArray<SSDataBlock>
+  int32_t blockNum;
+  SArray* dataLen;  // SArray<int32_t>
+  SArray* data;     // SArray<SRetrieveTableRsp*>
 } SStreamDispatchReq;
 
 typedef struct {
@@ -340,9 +398,9 @@ typedef struct {
   int8_t  inputStatus;
 } SStreamTaskRecoverRsp;
 
-int32_t streamEnqueueDataSubmit(SStreamTask* pTask, SStreamDataSubmit* input);
+int32_t tDecodeStreamDispatchReq(SDecoder* pDecoder, SStreamDispatchReq* pReq);
-int32_t streamEnqueueDataBlk(SStreamTask* pTask, SStreamDataBlock* input);
+
-int32_t streamDequeueOutput(SStreamTask* pTask, void** output);
+int32_t streamTriggerByWrite(SStreamTask* pTask, int32_t vgId, SMsgCb* pMsgCb);
 
 int32_t streamTaskRun(SStreamTask* pTask);
 
@@ -356,4 +414,4 @@ int32_t streamProcessRecoverRsp(SStreamTask* pTask, SStreamTaskRecoverRsp* pRsp)
 }
 #endif
 
-#endif /* ifndef _TSTREAM_H_ */
+#endif /* ifndef _STREAM_H_ */

include/util/taoserror.h

@@ -218,6 +218,8 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_MND_VGROUP_NOT_EXIST          TAOS_DEF_ERROR_CODE(0, 0x0390)
 #define TSDB_CODE_MND_VGROUP_NOT_IN_DNODE       TAOS_DEF_ERROR_CODE(0, 0x0391)
 #define TSDB_CODE_MND_VGROUP_ALREADY_IN_DNODE   TAOS_DEF_ERROR_CODE(0, 0x0392)
+#define TSDB_CODE_MND_VGROUP_UN_CHANGED         TAOS_DEF_ERROR_CODE(0, 0x0393)
+#define TSDB_CODE_MND_HAS_OFFLINE_DNODE         TAOS_DEF_ERROR_CODE(0, 0x0394)
 
 // mnode-stable
 #define TSDB_CODE_MND_STB_ALREADY_EXIST         TAOS_DEF_ERROR_CODE(0, 0x03A0)

include/util/tcompare.h

@@ -105,8 +105,6 @@ int32_t compareStrPatternNotMatch(const void *pLeft, const void *pRight);
 int32_t compareWStrPatternMatch(const void *pLeft, const void *pRight);
 int32_t compareWStrPatternNotMatch(const void *pLeft, const void *pRight);
 
-int32_t compareJsonContainsKey(const void *pLeft, const void *pRight);
-
 __compar_fn_t getComparFunc(int32_t type, int32_t optr);
 __compar_fn_t getKeyComparFunc(int32_t keyType, int32_t order);
 int32_t       doCompare(const char *a, const char *b, int32_t type, size_t size);

source/client/inc/clientInt.h

@@ -212,6 +212,9 @@ typedef struct SRequestObj {
   SArray*              tableList;
   SQueryExecMetric     metric;
   SRequestSendRecvBody body;
+
+  uint32_t             prevCode; //previous error code: todo refactor, add update flag for catalog
+  uint32_t             retry;
 } SRequestObj;
 
 typedef struct SSyncQueryParam {
@@ -263,8 +266,8 @@ extern SAppInfo appInfo;
 extern int32_t  clientReqRefPool;
 extern int32_t  clientConnRefPool;
 
-extern int (*handleRequestRspFp[TDMT_MAX])(void*, const SDataBuf* pMsg, int32_t code);
+__async_send_cb_fn_t getMsgRspHandle(int32_t msgType);
-int           genericRspCallback(void* param, const SDataBuf* pMsg, int32_t code);
+
 SMsgSendInfo* buildMsgInfoImpl(SRequestObj* pReqObj);
 
 void*    createTscObj(const char* user, const char* auth, const char* db, int32_t connType, SAppInstInfo* pAppInfo);
@@ -274,7 +277,7 @@ int32_t  releaseTscObj(int64_t rid);
 
 uint64_t generateRequestId();
 
-void*        createRequest(STscObj* pObj, void* param, int32_t type);
+void*        createRequest(STscObj* pObj, int32_t type);
 void         destroyRequest(SRequestObj* pRequest);
 SRequestObj* acquireRequest(int64_t rid);
 int32_t      releaseRequest(int64_t rid);
@@ -290,8 +293,6 @@ void* openTransporter(const char* user, const char* auth, int32_t numOfThreads);
 bool persistConnForSpecificMsg(void* parenct, tmsg_t msgType);
 void processMsgFromServer(void* parent, SRpcMsg* pMsg, SEpSet* pEpSet);
 
-void initMsgHandleFp();
-
 TAOS* taos_connect_internal(const char* ip, const char* user, const char* pass, const char* auth, const char* db,
                             uint16_t port, int connType);
 
@@ -325,6 +326,9 @@ int32_t      scheduleQuery(SRequestObj* pRequest, SQueryPlan* pDag, SArray* pNod
 void         launchAsyncQuery(SRequestObj* pRequest, SQuery* pQuery);
 int32_t      refreshMeta(STscObj* pTscObj, SRequestObj* pRequest);
 int32_t      updateQnodeList(SAppInstInfo* pInfo, SArray* pNodeList);
+void         doAsyncQuery(SRequestObj* pRequest, bool forceUpdateMeta);
+int32_t      removeMeta(STscObj* pTscObj, SArray* tbList);// todo move to clientImpl.c and become a static function
+int32_t      handleAlterTbExecRes(void* res, struct SCatalog* pCatalog);// todo move to xxx
 
 #ifdef __cplusplus
 }

source/client/src/clientEnv.c

@@ -177,7 +177,7 @@ STscObj *acquireTscObj(int64_t rid) { return (STscObj *)taosAcquireRef(clientCon
 
 int32_t releaseTscObj(int64_t rid) { return taosReleaseRef(clientConnRefPool, rid); }
 
-void *createRequest(STscObj *pObj, void *param, int32_t type) {
+void *createRequest(STscObj *pObj, int32_t type) {
   assert(pObj != NULL);
 
   SRequestObj *pRequest = (SRequestObj *)taosMemoryCalloc(1, sizeof(SRequestObj));
@@ -191,8 +191,6 @@ void *createRequest(STscObj *pObj, void *param, int32_t type) {
   pRequest->requestId = generateRequestId();
   pRequest->metric.start = taosGetTimestampUs();
 
-  pRequest->body.param = param;
-
   pRequest->type = type;
   pRequest->pTscObj = pObj;
   pRequest->msgBuf = taosMemoryCalloc(1, ERROR_MSG_BUF_DEFAULT_SIZE);
@@ -280,7 +278,6 @@ void taos_init_imp(void) {
     return;
   }
 
-  initMsgHandleFp();
   initQueryModuleMsgHandle();
 
   rpcInit();

source/client/src/clientImpl.c

@@ -133,7 +133,7 @@ TAOS* taos_connect_internal(const char* ip, const char* user, const char* pass,
 }
 
 int32_t buildRequest(STscObj* pTscObj, const char* sql, int sqlLen, SRequestObj** pRequest) {
-  *pRequest = createRequest(pTscObj, NULL, TSDB_SQL_SELECT);
+  *pRequest = createRequest(pTscObj, TSDB_SQL_SELECT);
   if (*pRequest == NULL) {
     tscError("failed to malloc sqlObj");
     return TSDB_CODE_TSC_OUT_OF_MEMORY;
@@ -207,6 +207,7 @@ int32_t execLocalCmd(SRequestObj* pRequest, SQuery* pQuery) {
   if (TSDB_CODE_SUCCESS == code && NULL != pRsp) {
     code = setQueryResultFromRsp(&pRequest->body.resInfo, pRsp, false, false);
   }
+
   return code;
 }
 
@@ -235,6 +236,31 @@ static SAppInstInfo* getAppInfo(SRequestObj* pRequest) {
   return pRequest->pTscObj->pAppInfo;
 }
 
+void asyncExecLocalCmd(SRequestObj* pRequest, SQuery* pQuery) {
+  SRetrieveTableRsp* pRsp = NULL;
+
+  int32_t code = qExecCommand(pQuery->pRoot, &pRsp);
+  if (TSDB_CODE_SUCCESS == code && NULL != pRsp) {
+    code = setQueryResultFromRsp(&pRequest->body.resInfo, pRsp, false, false);
+  }
+
+  SReqResultInfo* pResultInfo = &pRequest->body.resInfo;
+
+  if (pRequest->code != TSDB_CODE_SUCCESS) {
+    pResultInfo->numOfRows = 0;
+    pRequest->code = code;
+    tscError("0x%" PRIx64 " fetch results failed, code:%s, reqId:0x%" PRIx64, pRequest->self, tstrerror(code),
+             pRequest->requestId);
+  } else {
+    tscDebug("0x%" PRIx64 " fetch results, numOfRows:%d total Rows:%" PRId64 ", complete:%d, reqId:0x%" PRIx64,
+             pRequest->self, pResultInfo->numOfRows, pResultInfo->totalRows, pResultInfo->completed,
+             pRequest->requestId);
+  }
+
+  pRequest->body.queryFp(pRequest->body.param, pRequest, 0);
+//  pRequest->body.fetchFp(pRequest->body.param, pRequest, pResultInfo->numOfRows);
+}
+
 int32_t asyncExecDdlQuery(SRequestObj* pRequest, SQuery* pQuery) {
   // drop table if exists not_exists_table
   if (NULL == pQuery->pCmdMsg) {
@@ -431,7 +457,7 @@ int32_t scheduleQuery(SRequestObj* pRequest, SQueryPlan* pDag, SArray* pNodeList
     return pRequest->code;
   }
 
-  if (TDMT_VND_SUBMIT == pRequest->type || TDMT_VND_CREATE_TABLE == pRequest->type) {
+  if (TDMT_VND_SUBMIT == pRequest->type || TDMT_VND_DELETE == pRequest->type || TDMT_VND_CREATE_TABLE == pRequest->type) {
     pRequest->body.resInfo.numOfRows = res.numOfRows;
 
     if (pRequest->body.queryJob != 0) {
@@ -509,19 +535,20 @@ int32_t handleAlterTbExecRes(void* res, SCatalog* pCatalog) {
   return catalogUpdateTableMeta(pCatalog, (STableMetaRsp*)res);
 }
 
-int32_t handleExecRes(SRequestObj* pRequest) {
+int32_t handleQueryExecRsp(SRequestObj* pRequest) {
   if (NULL == pRequest->body.resInfo.execRes.res) {
     return TSDB_CODE_SUCCESS;
   }
 
-  int32_t code = 0;
   SCatalog* pCatalog = NULL;
-  code = catalogGetHandle(pRequest->pTscObj->pAppInfo->clusterId, &pCatalog);
+  SAppInstInfo*   pAppInfo = getAppInfo(pRequest);
+
+  int32_t code = catalogGetHandle(pAppInfo->clusterId, &pCatalog);
   if (code) {
     return code;
   }
 
-  SEpSet epset = getEpSet_s(&pRequest->pTscObj->pAppInfo->mgmtEp);
+  SEpSet epset = getEpSet_s(&pAppInfo->mgmtEp);
   SQueryExecRes* pRes = &pRequest->body.resInfo.execRes;
 
   switch (pRes->msgType) {
@@ -539,8 +566,9 @@ int32_t handleExecRes(SRequestObj* pRequest) {
       break;
     }
     default:
-      tscError("invalid exec result for request type %d", pRequest->type);
+      tscError("0x%"PRIx64", invalid exec result for request type %d, reqId:0x%"PRIx64, pRequest->self,
-      return TSDB_CODE_APP_ERROR;
+          pRequest->type, pRequest->requestId);
+      code = TSDB_CODE_APP_ERROR;
   }
 
   return code;
@@ -548,6 +576,25 @@ int32_t handleExecRes(SRequestObj* pRequest) {
 
 void schedulerExecCb(SQueryResult* pResult, void* param, int32_t code) {
   SRequestObj* pRequest = (SRequestObj*) param;
+  pRequest->code = code;
+
+  STscObj* pTscObj = pRequest->pTscObj;
+  if (code != TSDB_CODE_SUCCESS && NEED_CLIENT_HANDLE_ERROR(code)) {
+    tscDebug("0x%"PRIx64" client retry to handle the error, code:%s, reqId:0x%"PRIx64, pRequest->self, tstrerror(code), pRequest->requestId);
+    pRequest->prevCode = code;
+    doAsyncQuery(pRequest, true);
+    return;
+  }
+
+  if (code == TSDB_CODE_SUCCESS) {
+    code = handleQueryExecRsp(pRequest);
+    ASSERT(pRequest->code ==  TSDB_CODE_SUCCESS);
+    pRequest->code = code;
+  }
+
+  if (NEED_CLIENT_RM_TBLMETA_REQ(pRequest->type)) {
+    removeMeta(pTscObj, pRequest->tableList);
+  }
 
   // return to client
   pRequest->body.queryFp(pRequest->body.param, pRequest, code);
@@ -583,7 +630,7 @@ SRequestObj* launchQueryImpl(SRequestObj* pRequest, SQuery* pQuery, bool keepQue
     qDestroyQuery(pQuery);
   }
 
-  handleExecRes(pRequest);
+  handleQueryExecRsp(pRequest);
 
   if (NULL != pRequest && TSDB_CODE_SUCCESS != code) {
     pRequest->code = terrno;
@@ -617,13 +664,12 @@ SRequestObj* launchQuery(STscObj* pTscObj, const char* sql, int sqlLen) {
 }
 
 void launchAsyncQuery(SRequestObj* pRequest, SQuery* pQuery) {
-  void* pRes = NULL;
-
   int32_t code = 0;
+
   switch (pQuery->execMode) {
     case QUERY_EXEC_MODE_LOCAL:
-      code = execLocalCmd(pRequest, pQuery);
+      asyncExecLocalCmd(pRequest, pQuery);
-      break;
+      return;
     case QUERY_EXEC_MODE_RPC:
       code = asyncExecDdlQuery(pRequest, pQuery);
       break;
@@ -649,11 +695,9 @@ void launchAsyncQuery(SRequestObj* pRequest, SQuery* pQuery) {
       if (TSDB_CODE_SUCCESS == code) {
         schedulerAsyncExecJob(pAppInfo->pTransporter, pNodeList, pRequest->body.pDag, &pRequest->body.queryJob,
                               pRequest->sqlstr, pRequest->metric.start, schedulerExecCb, pRequest);
-        //        if (NULL != pRes) {
-        //          code = validateSversion(pRequest, pRes);
-        //        }
       }
 
+      //todo not to be released here
       taosArrayDestroy(pNodeList);
       break;
     }
@@ -671,12 +715,6 @@ void launchAsyncQuery(SRequestObj* pRequest, SQuery* pQuery) {
   if (NULL != pRequest && TSDB_CODE_SUCCESS != code) {
     pRequest->code = terrno;
   }
-
-  //    if (res) {
-  //      *res = pRes;
-  //    } else {
-//  freeRequestRes(pRequest, pRes);
-//  pRes = NULL;
 }
 
 int32_t refreshMeta(STscObj* pTscObj, SRequestObj* pRequest) {
@@ -750,7 +788,7 @@ SRequestObj* execQuery(STscObj* pTscObj, const char* sql, int sqlLen) {
       pRequest->code = code;
       break;
     }
-  } while (retryNum++ < REQUEST_MAX_TRY_TIMES);
+  } while (retryNum++ < REQUEST_TOTAL_EXEC_TIMES);
 
   if (NEED_CLIENT_RM_TBLMETA_REQ(pRequest->type)) {
     removeMeta(pTscObj, pRequest->tableList);
@@ -808,7 +846,7 @@ STscObj* taosConnectImpl(const char* user, const char* auth, const char* db, __t
     return pTscObj;
   }
 
-  SRequestObj* pRequest = createRequest(pTscObj, param, TDMT_MND_CONNECT);
+  SRequestObj* pRequest = createRequest(pTscObj, TDMT_MND_CONNECT);
   if (pRequest == NULL) {
     destroyTscObj(pTscObj);
     terrno = TSDB_CODE_TSC_OUT_OF_MEMORY;
@@ -850,7 +888,7 @@ static SMsgSendInfo* buildConnectMsg(SRequestObj* pRequest) {
 
   pMsgSendInfo->requestObjRefId = pRequest->self;
   pMsgSendInfo->requestId = pRequest->requestId;
-  pMsgSendInfo->fp = handleRequestRspFp[TMSG_INDEX(pMsgSendInfo->msgType)];
+  pMsgSendInfo->fp = getMsgRspHandle(pMsgSendInfo->msgType);
   pMsgSendInfo->param = pRequest;
 
   SConnectReq connectReq = {0};
@@ -1429,7 +1467,7 @@ TSDB_SERVER_STATUS taos_check_server_status(const char* fqdn, int port, char* de
   }
 
   code = statusRsp.statusCode;
-  if (details != NULL && statusRsp.details != NULL) {
+  if (details != NULL) {
     tstrncpy(details, statusRsp.details, maxlen);
   }
 

source/client/src/clientMain.c

@@ -30,6 +30,7 @@
 #define TSC_VAR_RELEASED    0
 
 static int32_t sentinel = TSC_VAR_NOT_RELEASE;
+static int32_t createParseContext(const SRequestObj *pRequest, SParseContext** pCxt);
 
 int taos_options(TSDB_OPTION option, const void *arg, ...) {
   static int32_t lock = 0;
@@ -192,7 +193,7 @@ TAOS_RES *taos_query(TAOS *taos, const char *sql) {
   STscObj* pTscObj = (STscObj*)taos;
 
 #if SYNC_ON_TOP_OF_ASYNC
-  SSyncQueryParam* param = taosMemoryCalloc(1, sizeof(struct SSyncQueryParam));
+  SSyncQueryParam* param = taosMemoryCalloc(1, sizeof(SSyncQueryParam));
   tsem_init(&param->sem, 0, 0);
 
   taos_query_a(pTscObj, sql, syncQueryFn, param);
@@ -608,18 +609,14 @@ void retrieveMetaCallback(SMetaData* pResultMeta, void* param, int32_t code) {
   SQuery* pQuery = pWrapper->pQuery;
   SRequestObj* pRequest = pWrapper->pRequest;
 
-  if (code != TSDB_CODE_SUCCESS) {
+  if (code == TSDB_CODE_SUCCESS) {
-    goto _error;
-  }
-
     code = qAnalyseSqlSemantic(pWrapper->pCtx, &pWrapper->catalogReq, pResultMeta, pQuery);
-  if (code != TSDB_CODE_SUCCESS) {
-    goto _error;
   }
 
+  if (code == TSDB_CODE_SUCCESS) {
     if (pQuery->haveResultSet) {
-    setResSchemaInfo(&pRequest->body.resInfo, pQuery->pResSchema, (pQuery)->numOfResCols);
+      setResSchemaInfo(&pRequest->body.resInfo, pQuery->pResSchema, pQuery->numOfResCols);
-    setResPrecision(&pRequest->body.resInfo, (pQuery)->precision);
+      setResPrecision(&pRequest->body.resInfo, pQuery->precision);
     }
 
     TSWAP(pRequest->dbList, (pQuery)->pDbList);
@@ -627,17 +624,23 @@ void retrieveMetaCallback(SMetaData* pResultMeta, void* param, int32_t code) {
 
     taosMemoryFree(pWrapper);
     launchAsyncQuery(pRequest, pQuery);
+  } else {
+    if (NEED_CLIENT_HANDLE_ERROR(code)) {
+      tscDebug("0x%"PRIx64" client retry to handle the error, code:%s, reqId:0x%"PRIx64, pRequest->self, tstrerror(code), pRequest->requestId);
+      pRequest->prevCode = code;
+      doAsyncQuery(pRequest, true);
       return;
+    }
 
-  _error:
+    // return to app directly
     taosMemoryFree(pWrapper);
-  tscError("0x%" PRIx64 " error occurs, code:%s, return to user app, reqId:%" PRIx64, pRequest->self, tstrerror(code),
+    tscError("0x%" PRIx64 " error occurs, code:%s, return to user app, reqId:0x%" PRIx64, pRequest->self, tstrerror(code),
              pRequest->requestId);
     pRequest->code = code;
     pRequest->body.queryFp(pRequest->body.param, pRequest, code);
+  }
 }
 
-// todo add retry before return user's callback
 void taos_query_a(TAOS *taos, const char *sql, __taos_async_fn_t fp, void *param) {
   ASSERT(fp != NULL);
 
@@ -657,24 +660,27 @@ void taos_query_a(TAOS *taos, const char *sql, __taos_async_fn_t fp, void *param
   }
 
   SRequestObj *pRequest = NULL;
-  int32_t      retryNum = 0;
+  int32_t code = buildRequest(taos, sql, sqlLen, &pRequest);
-  int32_t      code = 0;
-
-  //  while (retryNum++ < REQUEST_MAX_TRY_TIMES) {
-  code = buildRequest(taos, sql, sqlLen, &pRequest);
   if (code != TSDB_CODE_SUCCESS) {
     terrno = code;
-    fp(param, NULL, code);
+    fp(param, NULL, terrno);
     return;
   }
 
   pRequest->body.queryFp = fp;
   pRequest->body.param   = param;
+  doAsyncQuery(pRequest, false);
+}
 
-  STscObj *pTscObj = pRequest->pTscObj;
+int32_t createParseContext(const SRequestObj *pRequest, SParseContext** pCxt) {
+  const STscObj *pTscObj = pRequest->pTscObj;
 
-  SParseContext* pCxt = taosMemoryCalloc(1, sizeof(SParseContext));
+  *pCxt = taosMemoryCalloc(1, sizeof(SParseContext));
-  *pCxt = (SParseContext){.requestId = pRequest->requestId,
+  if (*pCxt == NULL) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  **pCxt = (SParseContext){.requestId = pRequest->requestId,
                        .acctId = pTscObj->acctId,
                        .db = pRequest->pDb,
                        .topicQuery = false,
@@ -687,6 +693,22 @@ void taos_query_a(TAOS *taos, const char *sql, __taos_async_fn_t fp, void *param
                        .pUser = pTscObj->user,
                        .isSuperUser = (0 == strcmp(pTscObj->user, TSDB_DEFAULT_USER)),
                        .async = true,};
+  return TSDB_CODE_SUCCESS;
+}
+
+void doAsyncQuery(SRequestObj* pRequest, bool updateMetaForce) {
+  SParseContext* pCxt = NULL;
+  STscObj *pTscObj = pRequest->pTscObj;
+
+  if (pRequest->retry++ > REQUEST_TOTAL_EXEC_TIMES) {
+    pRequest->code = pRequest->prevCode;
+    goto _error;
+  }
+
+  int32_t code = createParseContext(pRequest, &pCxt);
+  if (code != TSDB_CODE_SUCCESS) {
+    goto _error;
+  }
 
   pCxt->mgmtEpSet = getEpSet_s(&pTscObj->pAppInfo->mgmtEp);
   code = catalogGetHandle(pTscObj->pAppInfo->clusterId, &pCxt->pCatalog);
@@ -694,14 +716,20 @@ void taos_query_a(TAOS *taos, const char *sql, __taos_async_fn_t fp, void *param
     goto _error;
   }
 
-  SQuery *    pQuery = NULL;
+  SQuery *pQuery = NULL;
-  SCatalogReq catalogReq = {0};
+
+  SCatalogReq catalogReq = {.forceUpdate = updateMetaForce};
   code = qParseSqlSyntax(pCxt, &pQuery, &catalogReq);
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }
 
   SqlParseWrapper *pWrapper = taosMemoryCalloc(1, sizeof(SqlParseWrapper));
+  if (pWrapper == NULL) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+    goto _error;
+  }
+
   pWrapper->pCtx = pCxt;
   pWrapper->pQuery = pQuery;
   pWrapper->pRequest = pRequest;
@@ -709,24 +737,15 @@ void taos_query_a(TAOS *taos, const char *sql, __taos_async_fn_t fp, void *param
 
   code = catalogAsyncGetAllMeta(pCxt->pCatalog, pCxt->pTransporter, &pCxt->mgmtEpSet, pRequest->requestId,
                                 &catalogReq, retrieveMetaCallback, pWrapper, &pRequest->body.queryJob);
-
+  if (code == TSDB_CODE_SUCCESS) {
-  if (code != TSDB_CODE_SUCCESS) {
+    return;
-    goto _error;
   }
 
-  return;
-
-  // todo handle the retry process
-
-  //    if (TSDB_CODE_SUCCESS == code || NEED_CLIENT_HANDLE_ERROR(code)) {
-  //      TSWAP(pRequest->dbList, (pQuery)->pDbList);
-  //      TSWAP(pRequest->tableList, (pQuery)->pTableList);
-  //    }
-
   _error:
+  tscError("0x%"PRIx64" error happens, code:%s, reqId:0x%"PRIx64, pRequest->self, tstrerror(code), pRequest->requestId);
   terrno = code;
   pRequest->code = code;
-  fp(param, pRequest, code);
+  pRequest->body.queryFp(pRequest->body.param, pRequest, code);
 }
 
 static void fetchCallback(void* pResult, void* param, int32_t code) {
@@ -751,13 +770,14 @@ static void fetchCallback(void* pResult, void* param, int32_t code) {
   pRequest->code = setQueryResultFromRsp(&pRequest->body.resInfo, (SRetrieveTableRsp*)pResultInfo->pData, true, false);
   if (pRequest->code != TSDB_CODE_SUCCESS) {
     pResultInfo->numOfRows = 0;
-
     pRequest->code = code;
-    pRequest->body.fetchFp(pRequest->body.param, pRequest, 0);
+    tscError("0x%" PRIx64 " fetch results failed, code:%s, reqId:0x%" PRIx64, pRequest->self, tstrerror(code),
-  }
+             pRequest->requestId);
-
+  } else {
     tscDebug("0x%" PRIx64 " fetch results, numOfRows:%d total Rows:%" PRId64 ", complete:%d, reqId:0x%" PRIx64,
-           pRequest->self, pResultInfo->numOfRows, pResultInfo->totalRows, pResultInfo->completed, pRequest->requestId);
+             pRequest->self, pResultInfo->numOfRows, pResultInfo->totalRows, pResultInfo->completed,
+             pRequest->requestId);
+  }
 
   pRequest->body.fetchFp(pRequest->body.param, pRequest, pResultInfo->numOfRows);
 }

source/client/src/clientMsgHandler.c

@@ -21,8 +21,6 @@
 #include "tdef.h"
 #include "tname.h"
 
-int32_t (*handleRequestRspFp[TDMT_MAX])(void*, const SDataBuf* pMsg, int32_t code);
-
 static void setErrno(SRequestObj* pRequest, int32_t code) {
   pRequest->code = code;
   terrno = code;
@@ -107,10 +105,7 @@ SMsgSendInfo* buildMsgInfoImpl(SRequestObj* pRequest) {
 
   assert(pRequest != NULL);
   pMsgSendInfo->msgInfo = pRequest->body.requestMsg;
-
+  pMsgSendInfo->fp = getMsgRspHandle(pRequest->type);
-  pMsgSendInfo->fp = (handleRequestRspFp[TMSG_INDEX(pRequest->type)] == NULL)
-                         ? genericRspCallback
-                         : handleRequestRspFp[TMSG_INDEX(pRequest->type)];
   return pMsgSendInfo;
 }
 
@@ -209,7 +204,7 @@ int32_t processUseDbRsp(void* param, const SDataBuf* pMsg, int32_t code) {
   return 0;
 }
 
-int32_t processCreateTableRsp(void* param, const SDataBuf* pMsg, int32_t code) {
+int32_t processCreateSTableRsp(void* param, const SDataBuf* pMsg, int32_t code) {
   assert(pMsg != NULL && param != NULL);
   SRequestObj* pRequest = param;
 
@@ -219,6 +214,7 @@ int32_t processCreateTableRsp(void* param, const SDataBuf* pMsg, int32_t code) {
   }
 
   if (pRequest->body.queryFp != NULL) {
+    removeMeta(pRequest->pTscObj, pRequest->tableList);
     pRequest->body.queryFp(pRequest->body.param, pRequest, code);
   } else {
     tsem_post(&pRequest->body.rspSem);
@@ -251,10 +247,7 @@ int32_t processAlterStbRsp(void* param, const SDataBuf* pMsg, int32_t code) {
   SRequestObj* pRequest = param;
   if (code != TSDB_CODE_SUCCESS) {
     setErrno(pRequest, code);
-    tsem_post(&pRequest->body.rspSem);
+  } else {
-    return code;
-  }
-
     SMAlterStbRsp alterRsp = {0};
     SDecoder      coder = {0};
     tDecoderInit(&coder, pMsg->pData, pMsg->len);
@@ -263,18 +256,45 @@ int32_t processAlterStbRsp(void* param, const SDataBuf* pMsg, int32_t code) {
 
     pRequest->body.resInfo.execRes.msgType = TDMT_MND_ALTER_STB;
     pRequest->body.resInfo.execRes.res = alterRsp.pMeta;
+  }
 
+  if (pRequest->body.queryFp != NULL) {
+    SQueryExecRes* pRes = &pRequest->body.resInfo.execRes;
+
+    if (code == TSDB_CODE_SUCCESS) {
+      SCatalog* pCatalog = NULL;
+      int32_t   ret = catalogGetHandle(pRequest->pTscObj->pAppInfo->clusterId, &pCatalog);
+      if (pRes->res != NULL) {
+        ret = handleAlterTbExecRes(pRes->res, pCatalog);
+      }
+
+      if (ret != TSDB_CODE_SUCCESS) {
+        code = ret;
+      }
+    }
+
+    pRequest->body.queryFp(pRequest->body.param, pRequest, code);
+  } else {
     tsem_post(&pRequest->body.rspSem);
+  }
   return code;
 }
 
-
+__async_send_cb_fn_t getMsgRspHandle(int32_t msgType) {
-// todo refactor: this arraylist is too large
+  switch (msgType) {
-void initMsgHandleFp() {
+    case TDMT_MND_CONNECT:
-  handleRequestRspFp[TMSG_INDEX(TDMT_MND_CONNECT)] = processConnectRsp;
+      return processConnectRsp;
-  handleRequestRspFp[TMSG_INDEX(TDMT_MND_CREATE_DB)] = processCreateDbRsp;
+    case TDMT_MND_CREATE_DB:
-  handleRequestRspFp[TMSG_INDEX(TDMT_MND_USE_DB)] = processUseDbRsp;
+      return processCreateDbRsp;
-  handleRequestRspFp[TMSG_INDEX(TDMT_MND_CREATE_STB)] = processCreateTableRsp;
+    case TDMT_MND_USE_DB:
-  handleRequestRspFp[TMSG_INDEX(TDMT_MND_DROP_DB)] = processDropDbRsp;
+      return processUseDbRsp;
-  handleRequestRspFp[TMSG_INDEX(TDMT_MND_ALTER_STB)] = processAlterStbRsp;
+    case TDMT_MND_CREATE_STB:
+      return processCreateSTableRsp;
+    case TDMT_MND_DROP_DB:
+      return processDropDbRsp;
+    case TDMT_MND_ALTER_STB:
+      return processAlterStbRsp;
+    default:
+      return genericRspCallback;
+  }
 }

source/client/src/clientSml.c

@@ -2364,7 +2364,7 @@ static int32_t isSchemalessDb(SSmlHandle* info){
  */
 
 TAOS_RES* taos_schemaless_insert(TAOS* taos, char* lines[], int numLines, int protocol, int precision) {
-  SRequestObj* request = (SRequestObj*)createRequest((STscObj *)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj* request = (SRequestObj*)createRequest((STscObj *)taos, TSDB_SQL_INSERT);
   if(!request){
     uError("SML:taos_schemaless_insert error request is null");
     return NULL;

source/client/test/clientTests.cpp

@@ -778,7 +778,35 @@ TEST(testCase, async_api_test) {
   TAOS* pConn = taos_connect("localhost", "root", "taosdata", NULL, 0);
   ASSERT_NE(pConn, nullptr);
 
-  taos_query_a(pConn, "drop table test.tm0", queryCallback, pConn);
+  taos_query(pConn, "use test");
+
+  TAOS_RES* pRes = taos_query(pConn, "select * from t1");
+
+  taos_query(pConn, "alter table t1 add column b int");
+  pRes = taos_query(pConn, "insert into t1 values(now, 1, 2)");
+  if (taos_errno(pRes) != 0) {
+    printf("failed, reason:%s\n", taos_errstr(pRes));
+  }
+
+//  int32_t n = 0;
+//  TAOS_ROW    pRow = NULL;
+//  TAOS_FIELD* pFields = taos_fetch_fields(pRes);
+//  int32_t     numOfFields = taos_num_fields(pRes);
+//
+//  char str[512] = {0};
+//  while ((pRow = taos_fetch_row(pRes)) != NULL) {
+//    int32_t* length = taos_fetch_lengths(pRes);
+//    for(int32_t i = 0; i < numOfFields; ++i) {
+//      printf("(%d):%d " , i, length[i]);
+//    }
+//    printf("\n");
+//
+//    int32_t code = taos_print_row(str, pRow, pFields, numOfFields);
+//    printf("%s\n", str);
+//    memset(str, 0, sizeof(str));
+//  }
+
+  taos_query_a(pConn, "alter table test.m1 comment 'abcde' ", queryCallback, pConn);
   getchar();
   taos_close(pConn);
 }

source/client/test/smlTest.cpp

@@ -486,7 +486,7 @@ TEST(testCase, smlProcess_influx_Test) {
   pRes = taos_query(taos, "use inflx_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_LINE_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -607,7 +607,7 @@ TEST(testCase, smlParseLine_error_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_LINE_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -656,7 +656,7 @@ TEST(testCase, smlProcess_telnet_Test) {
   pRes = taos_query(taos, "use telnet_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_TELNET_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -710,7 +710,7 @@ TEST(testCase, smlProcess_json1_Test) {
   pRes = taos_query(taos, "use json_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj *)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj *)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_JSON_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -779,7 +779,7 @@ TEST(testCase, smlProcess_json2_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj *)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj *)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_JSON_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -823,7 +823,7 @@ TEST(testCase, smlProcess_json3_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj *)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj *)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_JSON_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -895,7 +895,7 @@ TEST(testCase, smlProcess_json4_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_JSON_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -957,7 +957,7 @@ TEST(testCase, smlParseTelnetLine_error_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_TELNET_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -1006,7 +1006,7 @@ TEST(testCase, smlParseTelnetLine_diff_type_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_TELNET_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -1033,7 +1033,7 @@ TEST(testCase, smlParseTelnetLine_json_error_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_TELNET_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -1101,7 +1101,7 @@ TEST(testCase, smlParseTelnetLine_diff_json_type1_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_TELNET_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -1146,7 +1146,7 @@ TEST(testCase, smlParseTelnetLine_diff_json_type2_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_TELNET_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -1191,7 +1191,7 @@ TEST(testCase, sml_TD15662_Test) {
   pRes = taos_query(taos, "use db_15662");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj *)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj *)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_LINE_PROTOCOL, TSDB_SML_TIMESTAMP_MILLI_SECONDS);
@@ -1218,7 +1218,7 @@ TEST(testCase, sml_TD15735_Test) {
   pRes = taos_query(taos, "use sml_db");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_TELNET_PROTOCOL, TSDB_SML_TIMESTAMP_NANO_SECONDS);
@@ -1244,7 +1244,7 @@ TEST(testCase, sml_TD15742_Test) {
   pRes = taos_query(taos, "use TD15742");
   taos_free_result(pRes);
 
-  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, NULL, TSDB_SQL_INSERT);
+  SRequestObj *request = (SRequestObj *)createRequest((STscObj*)taos, TSDB_SQL_INSERT);
   ASSERT_NE(request, nullptr);
 
   SSmlHandle *info = smlBuildSmlInfo(taos, request, TSDB_SML_LINE_PROTOCOL, TSDB_SML_TIMESTAMP_MILLI_SECONDS);

source/common/src/tdatablock.c

@@ -1590,7 +1590,6 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks
       for (int32_t k = 0; k < colNum; ++k) {  // iterate by column
         SColumnInfoData* pColInfoData = taosArrayGet(pDataBlock->pDataBlock, k);
         STColumn*         pCol = &pTSchema->columns[k];
-        ASSERT(pCol->type == pColInfoData->info.type);
         void* var = POINTER_SHIFT(pColInfoData->pData, j * pColInfoData->info.bytes);
         switch (pColInfoData->info.type) {
           case TSDB_DATA_TYPE_TIMESTAMP:
@@ -1614,20 +1613,39 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks
           case TSDB_DATA_TYPE_VARBINARY:
           case TSDB_DATA_TYPE_DECIMAL:
           case TSDB_DATA_TYPE_BLOB:
+          case TSDB_DATA_TYPE_JSON:
           case TSDB_DATA_TYPE_MEDIUMBLOB:
             uError("the column type %" PRIi16 " is defined but not implemented yet", pColInfoData->info.type);
             TASSERT(0);
             break;
           default:
             if (pColInfoData->info.type < TSDB_DATA_TYPE_MAX && pColInfoData->info.type > TSDB_DATA_TYPE_NULL) {
-              tdAppendColValToRow(&rb, PRIMARYKEY_TIMESTAMP_COL_ID + k, pColInfoData->info.type, TD_VTYPE_NORM, var, true, offset, k);
+              char tv[8] = {0};
+              if (pColInfoData->info.type == TSDB_DATA_TYPE_FLOAT) {
+                float v = 0;
+                GET_TYPED_DATA(v, float, pColInfoData->info.type, var);
+                SET_TYPED_DATA(&tv, pCol->type, v);
+              } else if (pColInfoData->info.type == TSDB_DATA_TYPE_DOUBLE) {
+                double v = 0;
+                GET_TYPED_DATA(v, double, pColInfoData->info.type, var);
+                SET_TYPED_DATA(&tv, pCol->type, v);
+              } else if (IS_SIGNED_NUMERIC_TYPE(pColInfoData->info.type)) {
+                int64_t v = 0;
+                GET_TYPED_DATA(v, int64_t, pColInfoData->info.type, var);
+                SET_TYPED_DATA(&tv, pCol->type, v);
+              } else {
+                uint64_t v = 0;
+                GET_TYPED_DATA(v, uint64_t, pColInfoData->info.type, var);
+                SET_TYPED_DATA(&tv, pCol->type, v);
+              }
+              tdAppendColValToRow(&rb, PRIMARYKEY_TIMESTAMP_COL_ID + k, pCol->type, TD_VTYPE_NORM, tv, true, offset, k);
             } else {
               uError("the column type %" PRIi16 " is undefined\n", pColInfoData->info.type);
               TASSERT(0);
             }
             break;
         }
-        offset += TYPE_BYTES[pColInfoData->info.type];
+        offset += TYPE_BYTES[pCol->type]; // sum/avg would convert to int64_t/uint64_t/double during aggregation
       }
       dataLen += TD_ROW_LEN(rb.pBuf);
 #ifdef TD_DEBUG_PRINT_ROW

source/common/src/tmsg.c

@@ -694,7 +694,6 @@ void tFreeSMAltertbReq(SMAlterStbReq *pReq) {
   pReq->pFields = NULL;
 }
 
-
 int32_t tSerializeSEpSet(void *buf, int32_t bufLen, const SEpSet *pEpset) {
   SEncoder encoder = {0};
   tEncoderInit(&encoder, buf, bufLen);
@@ -933,6 +932,7 @@ int32_t tSerializeSStatusReq(void *buf, int32_t bufLen, SStatusReq *pReq) {
   if (tEncodeI64(&encoder, pReq->qload.numOfProcessedFetch) < 0) return -1;
   if (tEncodeI64(&encoder, pReq->qload.numOfProcessedDrop) < 0) return -1;
   if (tEncodeI64(&encoder, pReq->qload.numOfProcessedHb) < 0) return -1;
+  if (tEncodeI64(&encoder, pReq->qload.numOfProcessedDelete) < 0) return -1;
   if (tEncodeI64(&encoder, pReq->qload.cacheDataSize) < 0) return -1;
   if (tEncodeI64(&encoder, pReq->qload.numOfQueryInQueue) < 0) return -1;
   if (tEncodeI64(&encoder, pReq->qload.numOfFetchInQueue) < 0) return -1;
@@ -1002,6 +1002,7 @@ int32_t tDeserializeSStatusReq(void *buf, int32_t bufLen, SStatusReq *pReq) {
   if (tDecodeI64(&decoder, &pReq->qload.numOfProcessedFetch) < 0) return -1;
   if (tDecodeI64(&decoder, &pReq->qload.numOfProcessedDrop) < 0) return -1;
   if (tDecodeI64(&decoder, &pReq->qload.numOfProcessedHb) < 0) return -1;
+  if (tDecodeI64(&decoder, &pReq->qload.numOfProcessedDelete) < 0) return -1;
   if (tDecodeI64(&decoder, &pReq->qload.cacheDataSize) < 0) return -1;
   if (tDecodeI64(&decoder, &pReq->qload.numOfQueryInQueue) < 0) return -1;
   if (tDecodeI64(&decoder, &pReq->qload.numOfFetchInQueue) < 0) return -1;
@@ -3674,12 +3675,12 @@ int32_t tEncodeTSma(SEncoder *pCoder, const STSma *pSma) {
     if (tEncodeCStr(pCoder, pSma->tagsFilter) < 0) return -1;
   }
   for (int32_t v = 0; v < pSma->numOfVgroups; ++v) {
-    if (tEncodeI32(pCoder, pSma->vgEpSet[v].vgId) < 0) return -1;
+    if (tEncodeI32(pCoder, pSma->pVgEpSet[v].vgId) < 0) return -1;
-    if (tEncodeI8(pCoder, pSma->vgEpSet[v].epSet.inUse) < 0) return -1;
+    if (tEncodeI8(pCoder, pSma->pVgEpSet[v].epSet.inUse) < 0) return -1;
-    int8_t numOfEps = pSma->vgEpSet[v].epSet.numOfEps;
+    int8_t numOfEps = pSma->pVgEpSet[v].epSet.numOfEps;
     if (tEncodeI8(pCoder, numOfEps) < 0) return -1;
     for (int32_t n = 0; n < numOfEps; ++n) {
-      const SEp *pEp = &pSma->vgEpSet[v].epSet.eps[n];
+      const SEp *pEp = &pSma->pVgEpSet[v].epSet.eps[n];
       if (tEncodeCStr(pCoder, pEp->fqdn) < 0) return -1;
       if (tEncodeU16(pCoder, pEp->port) < 0) return -1;
     }
@@ -3712,17 +3713,27 @@ int32_t tDecodeTSma(SDecoder *pCoder, STSma *pSma) {
   } else {
     pSma->tagsFilter = NULL;
   }
+  if (pSma->numOfVgroups > 0) {
+    pSma->pVgEpSet = (SVgEpSet *)tDecoderMalloc(pCoder, pSma->numOfVgroups * sizeof(SVgEpSet));
+    if (!pSma->pVgEpSet) {
+      terrno = TSDB_CODE_OUT_OF_MEMORY;
+      return -1;
+    }
+
+    memset(pSma->pVgEpSet, 0, pSma->numOfVgroups * sizeof(SVgEpSet));
+
     for (int32_t v = 0; v < pSma->numOfVgroups; ++v) {
-    if (tDecodeI32(pCoder, &pSma->vgEpSet[v].vgId) < 0) return -1;
+      if (tDecodeI32(pCoder, &pSma->pVgEpSet[v].vgId) < 0) return -1;
-    if (tDecodeI8(pCoder, &pSma->vgEpSet[v].epSet.inUse) < 0) return -1;
+      if (tDecodeI8(pCoder, &pSma->pVgEpSet[v].epSet.inUse) < 0) return -1;
-    if (tDecodeI8(pCoder, &pSma->vgEpSet[v].epSet.numOfEps) < 0) return -1;
+      if (tDecodeI8(pCoder, &pSma->pVgEpSet[v].epSet.numOfEps) < 0) return -1;
-    int8_t numOfEps = pSma->vgEpSet[v].epSet.numOfEps;
+      int8_t numOfEps = pSma->pVgEpSet[v].epSet.numOfEps;
       for (int32_t n = 0; n < numOfEps; ++n) {
-      SEp *pEp = &pSma->vgEpSet[v].epSet.eps[n];
+        SEp *pEp = &pSma->pVgEpSet[v].epSet.eps[n];
         if (tDecodeCStrTo(pCoder, pEp->fqdn) < 0) return -1;
         if (tDecodeU16(pCoder, &pEp->port) < 0) return -1;
       }
     }
+  }
 
   return 0;
 }
@@ -3765,7 +3776,7 @@ int32_t tDecodeSVDropTSmaReq(SDecoder *pCoder, SVDropTSmaReq *pReq) {
   return 0;
 }
 
-int32_t tEncodeSVGetTSmaExpWndsReq(SEncoder* pCoder, const SVGetTsmaExpWndsReq* pReq) {
+int32_t tEncodeSVGetTSmaExpWndsReq(SEncoder *pCoder, const SVGetTsmaExpWndsReq *pReq) {
   if (tStartEncode(pCoder) < 0) return -1;
 
   if (tEncodeI64(pCoder, pReq->indexUid) < 0) return -1;
@@ -3776,7 +3787,7 @@ int32_t tEncodeSVGetTSmaExpWndsReq(SEncoder* pCoder, const SVGetTsmaExpWndsReq*
   return 0;
 }
 
-int32_t tDecodeSVGetTsmaExpWndsReq(SDecoder* pCoder, SVGetTsmaExpWndsReq* pReq) {
+int32_t tDecodeSVGetTsmaExpWndsReq(SDecoder *pCoder, SVGetTsmaExpWndsReq *pReq) {
   if (tStartDecode(pCoder) < 0) return -1;
 
   if (tDecodeI64(pCoder, &pReq->indexUid) < 0) return -1;
@@ -3787,7 +3798,7 @@ int32_t tDecodeSVGetTsmaExpWndsReq(SDecoder* pCoder, SVGetTsmaExpWndsReq* pReq)
   return 0;
 }
 
-int32_t tEncodeSVGetTSmaExpWndsRsp(SEncoder* pCoder, const SVGetTsmaExpWndsRsp* pReq) {
+int32_t tEncodeSVGetTSmaExpWndsRsp(SEncoder *pCoder, const SVGetTsmaExpWndsRsp *pReq) {
   if (tStartEncode(pCoder) < 0) return -1;
 
   if (tEncodeI64(pCoder, pReq->indexUid) < 0) return -1;
@@ -3814,56 +3825,79 @@ int32_t tDecodeSVGetTsmaExpWndsRsp(SDecoder *pCoder, SVGetTsmaExpWndsRsp *pReq)
   return 0;
 }
 
-int32_t tEncodeSVDeleteReq(SEncoder* pCoder, const SVDeleteReq* pReq) {
+int32_t tSerializeSVDeleteReq(void *buf, int32_t bufLen, SVDeleteReq *pReq) {
-  if (tStartEncode(pCoder) < 0) return -1;
+  int32_t headLen = sizeof(SMsgHead);
-
+  if (buf != NULL) {
-  if (tEncodeI64(pCoder, pReq->delUid) < 0) return -1;
+    buf = (char *)buf + headLen;
-  if (tEncodeI64(pCoder, pReq->tbUid) < 0) return -1;
+    bufLen -= headLen;
-  if (tEncodeI8(pCoder, pReq->type) < 0) return -1;
-  if (tEncodeI16v(pCoder, pReq->nWnds) < 0) return -1;
-  if (tEncodeCStr(pCoder, pReq->tbFullName) < 0) return -1;
-  if (tEncodeCStr(pCoder, pReq->subPlan) < 0) return -1;
-  for (int16_t i = 0; i < pReq->nWnds; ++i) {
-    if (tEncodeI64(pCoder, pReq->wnds[i].skey) < 0) return -1;
-    if (tEncodeI64(pCoder, pReq->wnds[i].ekey) < 0) return -1;
   }
 
-  tEndEncode(pCoder);
+  SEncoder encoder = {0};
+  tEncoderInit(&encoder, buf, bufLen);
+
+  if (tStartEncode(&encoder) < 0) return -1;
+  if (tEncodeU64(&encoder, pReq->sId) < 0) return -1;
+  if (tEncodeU64(&encoder, pReq->queryId) < 0) return -1;
+  if (tEncodeU64(&encoder, pReq->taskId) < 0) return -1;
+  if (tEncodeU32(&encoder, pReq->sqlLen) < 0) return -1;
+  if (tEncodeU32(&encoder, pReq->phyLen) < 0) return -1;
+  if (tEncodeCStr(&encoder, pReq->sql) < 0) return -1;
+  if (tEncodeCStr(&encoder, pReq->msg) < 0) return -1;
+  tEndEncode(&encoder);
+
+  int32_t tlen = encoder.pos;
+  tEncoderClear(&encoder);
+
+  if (buf != NULL) {
+    SMsgHead *pHead = (SMsgHead *)((char *)buf - headLen);
+    pHead->vgId = htonl(pReq->header.vgId);
+    pHead->contLen = htonl(tlen + headLen);
+  }
+
+  return tlen + headLen;
+}
+
+int32_t tDeserializeSVDeleteReq(void *buf, int32_t bufLen, SVDeleteReq *pReq) {
+  int32_t headLen = sizeof(SMsgHead);
+
+  SMsgHead *pHead = buf;
+  pHead->vgId = pReq->header.vgId;
+  pHead->contLen = pReq->header.contLen;
+
+  SDecoder decoder = {0};
+  tDecoderInit(&decoder, (char *)buf + headLen, bufLen - headLen);
+
+  if (tStartDecode(&decoder) < 0) return -1;
+  if (tDecodeU64(&decoder, &pReq->sId) < 0) return -1;
+  if (tDecodeU64(&decoder, &pReq->queryId) < 0) return -1;
+  if (tDecodeU64(&decoder, &pReq->taskId) < 0) return -1;
+  if (tDecodeU32(&decoder, &pReq->sqlLen) < 0) return -1;
+  if (tDecodeU32(&decoder, &pReq->phyLen) < 0) return -1;
+  pReq->sql = taosMemoryCalloc(1, pReq->sqlLen + 1);
+  if (NULL == pReq->sql) return -1;
+  pReq->msg = taosMemoryCalloc(1, pReq->phyLen + 1);
+  if (NULL == pReq->msg) return -1;
+  if (tDecodeCStrTo(&decoder, pReq->sql) < 0) return -1;
+  if (tDecodeCStrTo(&decoder, pReq->msg) < 0) return -1;
+
+  tEndDecode(&decoder);
+
+  tDecoderClear(&decoder);
   return 0;
 }
 
-int32_t tDecodeSVDeleteReq(SDecoder* pCoder, SVDeleteReq* pReq) {
+int32_t tEncodeSVDeleteRsp(SEncoder *pCoder, const SVDeleteRsp *pReq) {
-  if (tStartDecode(pCoder) < 0) return -1;
-
-  if (tDecodeI64(pCoder, &pReq->delUid) < 0) return -1;
-  if (tDecodeI64(pCoder, &pReq->tbUid) < 0) return -1;
-  if (tDecodeI8(pCoder, &pReq->type) < 0) return -1;
-  if (tDecodeI16v(pCoder, &pReq->nWnds) < 0) return -1;
-  if (tDecodeCStr(pCoder, &pReq->tbFullName) < 0) return -1;
-  if (tDecodeCStr(pCoder, &pReq->subPlan) < 0) return -1;
-  for (int16_t i = 0; i < pReq->nWnds; ++i) {
-    if (tDecodeI64(pCoder, &pReq->wnds[i].skey) < 0) return -1;
-    if (tDecodeI64(pCoder, &pReq->wnds[i].ekey) < 0) return -1;
-  }
-
-  tEndDecode(pCoder);
-  return 0;
-}
-
-int32_t tEncodeSVDeleteRsp(SEncoder* pCoder, const SVDeleteRsp* pReq) {
   if (tStartEncode(pCoder) < 0) return -1;
 
-  if (tEncodeI32(pCoder, pReq->code) < 0) return -1;
   if (tEncodeI64(pCoder, pReq->affectedRows) < 0) return -1;
 
   tEndEncode(pCoder);
   return 0;
 }
 
-int32_t tDecodeSVDeleteRsp(SDecoder* pCoder, SVDeleteRsp* pReq) {
+int32_t tDecodeSVDeleteRsp(SDecoder *pCoder, SVDeleteRsp *pReq) {
   if (tStartDecode(pCoder) < 0) return -1;
 
-  if (tDecodeI32(pCoder, &pReq->code) < 0) return -1;
   if (tDecodeI64(pCoder, &pReq->affectedRows) < 0) return -1;
 
   tEndDecode(pCoder);
@@ -4559,7 +4593,7 @@ int32_t tDeserializeSMAlterStbRsp(void *buf, int32_t bufLen, SMAlterStbRsp *pRsp
   return 0;
 }
 
-void tFreeSMAlterStbRsp(SMAlterStbRsp* pRsp) {
+void tFreeSMAlterStbRsp(SMAlterStbRsp *pRsp) {
   if (NULL == pRsp) {
     return;
   }
@@ -4569,6 +4603,3 @@ void tFreeSMAlterStbRsp(SMAlterStbRsp* pRsp) {
     taosMemoryFree(pRsp->pMeta);
   }
 }
-
-
-

source/dnode/mgmt/mgmt_vnode/src/vmHandle.c

@@ -358,6 +358,7 @@ SArray *vmGetMsgHandles() {
   if (dmSetMgmtHandle(pArray, TDMT_VND_TASK_RUN, vmPutMsgToFetchQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_TASK_DISPATCH, vmPutMsgToFetchQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_TASK_RECOVER, vmPutMsgToFetchQueue, 0) == NULL) goto _OVER;
+  if (dmSetMgmtHandle(pArray, TDMT_VND_DELETE, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
 
   if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_REPLICA, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
   if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIG, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;

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

@@ -316,10 +316,12 @@ typedef struct {
   int32_t   tagsFilterLen;
   int32_t   sqlLen;
   int32_t   astLen;
+  int32_t   numOfVgroups;
   char*     expr;
   char*     tagsFilter;
   char*     sql;
   char*     ast;
+  SVgEpSet* pVgEpSet;
 } SSmaObj;
 
 typedef struct {

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

@@ -28,7 +28,7 @@ SDnodeObj *mndAcquireDnode(SMnode *pMnode, int32_t dnodeId);
 void       mndReleaseDnode(SMnode *pMnode, SDnodeObj *pDnode);
 SEpSet     mndGetDnodeEpset(SDnodeObj *pDnode);
 int32_t    mndGetDnodeSize(SMnode *pMnode);
-bool       mndIsDnodeOnline(SMnode *pMnode, SDnodeObj *pDnode, int64_t curMs);
+bool       mndIsDnodeOnline(SDnodeObj *pDnode, int64_t curMs);
 
 #ifdef __cplusplus
 }

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

@@ -28,6 +28,7 @@ SMnodeObj *mndAcquireMnode(SMnode *pMnode, int32_t mnodeId);
 void       mndReleaseMnode(SMnode *pMnode, SMnodeObj *pObj);
 bool       mndIsMnode(SMnode *pMnode, int32_t dnodeId);
 void       mndGetMnodeEpSet(SMnode *pMnode, SEpSet *pEpSet);
+int32_t    mndSetDropMnodeInfoToTrans(SMnode *pMnode, STrans *pTrans, SMnodeObj *pObj);
 
 #ifdef __cplusplus
 }

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

@@ -30,15 +30,21 @@ SSdbRaw *mndVgroupActionEncode(SVgObj *pVgroup);
 SEpSet   mndGetVgroupEpset(SMnode *pMnode, const SVgObj *pVgroup);
 int32_t  mndGetVnodesNum(SMnode *pMnode, int32_t dnodeId);
 
-int32_t mndAllocSmaVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj *pVgroup);
+SArray *mndBuildDnodesArray(SMnode *, int32_t exceptDnodeId);
-int32_t mndAllocVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj **ppVgroups);
+int32_t mndAllocSmaVgroup(SMnode *, SDbObj *pDb, SVgObj *pVgroup);
-SArray *mndBuildDnodesArray(SMnode *pMnode);
+int32_t mndAllocVgroup(SMnode *, SDbObj *pDb, SVgObj **ppVgroups);
-int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray);
+int32_t mndAddVnodeToVgroup(SMnode *, SVgObj *pVgroup, SArray *pArray);
-int32_t mndRemoveVnodeFromVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray, SVnodeGid *pVgId);
+int32_t mndRemoveVnodeFromVgroup(SMnode *, SVgObj *pVgroup, SArray *pArray, SVnodeGid *pDelVgid);
+int32_t mndAddCreateVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid, bool standby);
+int32_t mndAddAlterVnodeConfirmAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup);
+int32_t mndAddAlterVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, tmsg_t msgType);
+int32_t mndAddDropVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid, bool isRedo);
+int32_t mndSetMoveVgroupInfoToTrans(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, int32_t vn, SArray *pArray);
+int32_t mndSetMoveVgroupsInfoToTrans(SMnode *, STrans *pTrans, int32_t dropDnodeId);
 
-void *mndBuildCreateVnodeReq(SMnode *pMnode, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen, bool standby);
+void *mndBuildCreateVnodeReq(SMnode *, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *cntlen, bool standby);
-void *mndBuildDropVnodeReq(SMnode *pMnode, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);
+void *mndBuildDropVnodeReq(SMnode *, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);
-void *mndBuildAlterVnodeReq(SMnode *pMnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);
+void *mndBuildAlterVnodeReq(SMnode *, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);
 
 #ifdef __cplusplus
 }

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

@@ -299,7 +299,7 @@ static int32_t mndProcessCreateBnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto _OVER;
   }
 
@@ -409,7 +409,7 @@ static int32_t mndProcessDropBnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto _OVER;
   }
 

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

@@ -263,111 +263,6 @@ void mndReleaseDb(SMnode *pMnode, SDbObj *pDb) {
   sdbRelease(pSdb, pDb);
 }
 
-static int32_t mndAddCreateVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid,
-                                       bool standby) {
-  STransAction action = {0};
-
-  SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
-  if (pDnode == NULL) return -1;
-  action.epSet = mndGetDnodeEpset(pDnode);
-  mndReleaseDnode(pMnode, pDnode);
-
-  int32_t contLen = 0;
-  void   *pReq = mndBuildCreateVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen, standby);
-  if (pReq == NULL) return -1;
-
-  action.pCont = pReq;
-  action.contLen = contLen;
-  action.msgType = TDMT_DND_CREATE_VNODE;
-  action.acceptableCode = TSDB_CODE_NODE_ALREADY_DEPLOYED;
-
-  if (mndTransAppendRedoAction(pTrans, &action) != 0) {
-    taosMemoryFree(pReq);
-    return -1;
-  }
-
-  return 0;
-}
-
-static int32_t mndAddAlterVnodeConfirmAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup) {
-  STransAction action = {0};
-  action.epSet = mndGetVgroupEpset(pMnode, pVgroup);
-
-  int32_t   contLen = sizeof(SMsgHead);
-  SMsgHead *pHead = taosMemoryMalloc(contLen);
-  if (pHead == NULL) {
-    terrno = TSDB_CODE_OUT_OF_MEMORY;
-    return -1;
-  }
-
-  pHead->contLen = htonl(contLen);
-  pHead->vgId = htonl(pVgroup->vgId);
-
-  action.pCont = pHead;
-  action.contLen = contLen;
-  action.msgType = TDMT_VND_ALTER_CONFIRM;
-
-  if (mndTransAppendRedoAction(pTrans, &action) != 0) {
-    taosMemoryFree(pHead);
-    return -1;
-  }
-
-  return 0;
-}
-
-static int32_t mndAddAlterVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, tmsg_t msgType) {
-  STransAction action = {0};
-  action.epSet = mndGetVgroupEpset(pMnode, pVgroup);
-
-  int32_t contLen = 0;
-  void   *pReq = mndBuildAlterVnodeReq(pMnode, pDb, pVgroup, &contLen);
-  if (pReq == NULL) return -1;
-
-  action.pCont = pReq;
-  action.contLen = contLen;
-  action.msgType = msgType;
-
-  if (mndTransAppendRedoAction(pTrans, &action) != 0) {
-    taosMemoryFree(pReq);
-    return -1;
-  }
-
-  return 0;
-}
-
-static int32_t mndAddDropVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid,
-                                     bool isRedo) {
-  STransAction action = {0};
-
-  SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
-  if (pDnode == NULL) return -1;
-  action.epSet = mndGetDnodeEpset(pDnode);
-  mndReleaseDnode(pMnode, pDnode);
-
-  int32_t contLen = 0;
-  void   *pReq = mndBuildDropVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
-  if (pReq == NULL) return -1;
-
-  action.pCont = pReq;
-  action.contLen = contLen;
-  action.msgType = TDMT_DND_DROP_VNODE;
-  action.acceptableCode = TSDB_CODE_NODE_NOT_DEPLOYED;
-
-  if (isRedo) {
-    if (mndTransAppendRedoAction(pTrans, &action) != 0) {
-      taosMemoryFree(pReq);
-      return -1;
-    }
-  } else {
-    if (mndTransAppendUndoAction(pTrans, &action) != 0) {
-      taosMemoryFree(pReq);
-      return -1;
-    }
-  }
-
-  return 0;
-}
-
 static int32_t mndCheckDbName(const char *dbName, SUserObj *pUser) {
   char *pos = strstr(dbName, TS_PATH_DELIMITER);
   if (pos == NULL) {
@@ -795,7 +690,7 @@ static int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj
 static int32_t mndSetAlterDbRedoActions(SMnode *pMnode, STrans *pTrans, SDbObj *pOld, SDbObj *pNew) {
   SSdb   *pSdb = pMnode->pSdb;
   void   *pIter = NULL;
-  SArray *pArray = mndBuildDnodesArray(pMnode);
+  SArray *pArray = mndBuildDnodesArray(pMnode, 0);
 
   while (1) {
     SVgObj *pVgroup = NULL;
@@ -1497,19 +1392,25 @@ char *buildRetension(SArray *pRetension) {
 
   int64_t v1 = getValOfDiffPrecision(p->freqUnit, p->freq);
   int64_t v2 = getValOfDiffPrecision(p->keepUnit, p->keep);
-  len += sprintf(p1 + len, "%" PRId64 "%c:%" PRId64 "%c,", v1, p->freqUnit, v2, p->keepUnit);
+  len += sprintf(p1 + len, "%" PRId64 "%c:%" PRId64 "%c", v1, p->freqUnit, v2, p->keepUnit);
 
+  if (size > 1) {
+    len += sprintf(p1 + len, ",");
     p = taosArrayGet(pRetension, 1);
 
     v1 = getValOfDiffPrecision(p->freqUnit, p->freq);
     v2 = getValOfDiffPrecision(p->keepUnit, p->keep);
-  len += sprintf(p1 + len, "%" PRId64 "%c:%" PRId64 "%c,", v1, p->freqUnit, v2, p->keepUnit);
+    len += sprintf(p1 + len, "%" PRId64 "%c:%" PRId64 "%c", v1, p->freqUnit, v2, p->keepUnit);
+  }
 
+  if (size > 2) {
+    len += sprintf(p1 + len, ",");
     p = taosArrayGet(pRetension, 2);
 
     v1 = getValOfDiffPrecision(p->freqUnit, p->freq);
     v2 = getValOfDiffPrecision(p->keepUnit, p->keep);
     len += sprintf(p1 + len, "%" PRId64 "%c:%" PRId64 "%c", v1, p->freqUnit, v2, p->keepUnit);
+  }
 
   varDataSetLen(p1, len);
   return p1;
@@ -1742,3 +1643,4 @@ static void mndCancelGetNextDb(SMnode *pMnode, void *pIter) {
   SSdb *pSdb = pMnode->pSdb;
   sdbCancelFetch(pSdb, pIter);
 }
+

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

@@ -254,7 +254,7 @@ int32_t mndGetDnodeSize(SMnode *pMnode) {
   return sdbGetSize(pSdb, SDB_DNODE);
 }
 
-bool mndIsDnodeOnline(SMnode *pMnode, SDnodeObj *pDnode, int64_t curMs) {
+bool mndIsDnodeOnline(SDnodeObj *pDnode, int64_t curMs) {
   int64_t interval = TABS(pDnode->lastAccessTime - curMs);
   if (interval > 5000 * tsStatusInterval) {
     if (pDnode->rebootTime > 0) {
@@ -393,7 +393,7 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
 
   int64_t dnodeVer = sdbGetTableVer(pMnode->pSdb, SDB_DNODE) + sdbGetTableVer(pMnode->pSdb, SDB_MNODE);
   int64_t curMs = taosGetTimestampMs();
-  bool    online = mndIsDnodeOnline(pMnode, pDnode, curMs);
+  bool    online = mndIsDnodeOnline(pDnode, curMs);
   bool    dnodeChanged = (statusReq.dnodeVer != dnodeVer);
   bool    reboot = (pDnode->rebootTime != statusReq.rebootTime);
   bool    needCheck = !online || dnodeChanged || reboot;
@@ -542,7 +542,7 @@ static int32_t mndProcessCreateDnodeReq(SRpcMsg *pReq) {
     goto CREATE_DNODE_OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto CREATE_DNODE_OVER;
   }
 
@@ -559,30 +559,36 @@ CREATE_DNODE_OVER:
   return code;
 }
 
-static int32_t mndDropDnode(SMnode *pMnode, SRpcMsg *pReq, SDnodeObj *pDnode) {
+static int32_t mndDropDnode(SMnode *pMnode, SRpcMsg *pReq, SDnodeObj *pDnode, SMnodeObj *pMObj) {
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_GLOBAL, pReq);
+  int32_t  code = -1;
-  if (pTrans == NULL) {
+  SSdbRaw *pRaw = NULL;
-    mError("dnode:%d, failed to drop since %s", pDnode->id, terrstr());
+  STrans  *pTrans = NULL;
-    return -1;
+
-  }
+  pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
+  if (pTrans == NULL) goto _OVER;
+  mndTransSetSerial(pTrans);
   mDebug("trans:%d, used to drop dnode:%d", pTrans->id, pDnode->id);
 
-  SSdbRaw *pCommitRaw = mndDnodeActionEncode(pDnode);
+  pRaw = mndDnodeActionEncode(pDnode);
-  if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) {
+  if (pRaw == NULL || mndTransAppendRedolog(pTrans, pRaw) != 0) goto _OVER;
-    mError("trans:%d, failed to append commit log since %s", pTrans->id, terrstr());
+  sdbSetRawStatus(pRaw, SDB_STATUS_DROPPING);
-    mndTransDrop(pTrans);
+  pRaw = NULL;
-    return -1;
-  }
-  sdbSetRawStatus(pCommitRaw, SDB_STATUS_DROPPED);
 
-  if (mndTransPrepare(pMnode, pTrans) != 0) {
+  pRaw = mndDnodeActionEncode(pDnode);
-    mError("trans:%d, failed to prepare since %s", pTrans->id, terrstr());
+  if (pRaw == NULL || mndTransAppendCommitlog(pTrans, pRaw) != 0) goto _OVER;
-    mndTransDrop(pTrans);
+  sdbSetRawStatus(pRaw, SDB_STATUS_DROPPED);
-    return -1;
+  pRaw = NULL;
-  }
 
+  if (mndSetDropMnodeInfoToTrans(pMnode, pTrans, pMObj) != 0) goto _OVER;
+  if (mndSetMoveVgroupsInfoToTrans(pMnode, pTrans, pDnode->id) != 0) goto _OVER;
+  if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
+
+  code = 0;
+
+_OVER:
   mndTransDrop(pTrans);
-  return 0;
+  sdbFreeRaw(pRaw);
+  return code;
 }
 
 static int32_t mndProcessDropDnodeReq(SRpcMsg *pReq) {
@@ -595,42 +601,53 @@ static int32_t mndProcessDropDnodeReq(SRpcMsg *pReq) {
 
   if (tDeserializeSCreateDropMQSBNodeReq(pReq->pCont, pReq->contLen, &dropReq) != 0) {
     terrno = TSDB_CODE_INVALID_MSG;
-    goto DROP_DNODE_OVER;
+    goto _OVER;
   }
 
   mDebug("dnode:%d, start to drop", dropReq.dnodeId);
 
   if (dropReq.dnodeId <= 0) {
     terrno = TSDB_CODE_MND_INVALID_DNODE_ID;
-    goto DROP_DNODE_OVER;
+    goto _OVER;
   }
 
   pDnode = mndAcquireDnode(pMnode, dropReq.dnodeId);
   if (pDnode == NULL) {
     terrno = TSDB_CODE_MND_DNODE_NOT_EXIST;
-    goto DROP_DNODE_OVER;
+    goto _OVER;
+  }
+
+  if (!mndIsDnodeOnline(pDnode, taosGetTimestampMs())) {
+    terrno = TSDB_CODE_NODE_OFFLINE;
+    goto _OVER;
   }
 
   pMObj = mndAcquireMnode(pMnode, dropReq.dnodeId);
   if (pMObj != NULL) {
-    terrno = TSDB_CODE_MND_MNODE_NOT_EXIST;
+    if (sdbGetSize(pMnode->pSdb, SDB_MNODE) <= 1) {
-    goto DROP_DNODE_OVER;
+      terrno = TSDB_CODE_MND_TOO_FEW_MNODES;
+      goto _OVER;
+    }
+    if (pMnode->selfDnodeId == dropReq.dnodeId) {
+      terrno = TSDB_CODE_MND_CANT_DROP_MASTER;
+      goto _OVER;
+    }
   }
 
   pUser = mndAcquireUser(pMnode, pReq->conn.user);
   if (pUser == NULL) {
     terrno = TSDB_CODE_MND_NO_USER_FROM_CONN;
-    goto DROP_DNODE_OVER;
+    goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
-    goto DROP_DNODE_OVER;
+    goto _OVER;
   }
 
-  code = mndDropDnode(pMnode, pReq, pDnode);
+  code = mndDropDnode(pMnode, pReq, pDnode, pMObj);
   if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;
 
-DROP_DNODE_OVER:
+_OVER:
   if (code != 0 && code != TSDB_CODE_ACTION_IN_PROGRESS) {
     mError("dnode:%d, failed to drop since %s", dropReq.dnodeId, terrstr());
   }
@@ -638,7 +655,6 @@ DROP_DNODE_OVER:
   mndReleaseDnode(pMnode, pDnode);
   mndReleaseUser(pMnode, pUser);
   mndReleaseMnode(pMnode, pMObj);
-
   return code;
 }
 
@@ -736,7 +752,7 @@ static int32_t mndRetrieveDnodes(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pB
   while (numOfRows < rows) {
     pShow->pIter = sdbFetch(pSdb, SDB_DNODE, pShow->pIter, (void **)&pDnode);
     if (pShow->pIter == NULL) break;
-    bool online = mndIsDnodeOnline(pMnode, pDnode, curMs);
+    bool online = mndIsDnodeOnline(pDnode, curMs);
 
     cols = 0;
 

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

@@ -529,7 +529,7 @@ int32_t mndGetMonitorInfo(SMnode *pMnode, SMonClusterInfo *pClusterInfo, SMonVgr
     SMonDnodeDesc desc = {0};
     desc.dnode_id = pObj->id;
     tstrncpy(desc.dnode_ep, pObj->ep, sizeof(desc.dnode_ep));
-    if (mndIsDnodeOnline(pMnode, pObj, ms)) {
+    if (mndIsDnodeOnline(pObj, ms)) {
       tstrncpy(desc.status, "ready", sizeof(desc.status));
     } else {
       tstrncpy(desc.status, "offline", sizeof(desc.status));

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

@@ -358,9 +358,9 @@ static int32_t mndCreateMnode(SMnode *pMnode, SRpcMsg *pReq, SDnodeObj *pDnode,
 
   STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
   if (pTrans == NULL) goto _OVER;
-
-  mDebug("trans:%d, used to create mnode:%d", pTrans->id, pCreate->dnodeId);
   mndTransSetSerial(pTrans);
+  mDebug("trans:%d, used to create mnode:%d", pTrans->id, pCreate->dnodeId);
+
   if (mndSetCreateMnodeRedoLogs(pMnode, pTrans, &mnodeObj) != 0) goto _OVER;
   if (mndSetCreateMnodeCommitLogs(pMnode, pTrans, &mnodeObj) != 0) goto _OVER;
   if (mndSetCreateMnodeRedoActions(pMnode, pTrans, pDnode, &mnodeObj) != 0) goto _OVER;
@@ -408,7 +408,7 @@ static int32_t mndProcessCreateMnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (!mndIsDnodeOnline(pMnode, pDnode, taosGetTimestampMs())) {
+  if (!mndIsDnodeOnline(pDnode, taosGetTimestampMs())) {
     terrno = TSDB_CODE_NODE_OFFLINE;
     goto _OVER;
   }
@@ -419,7 +419,7 @@ static int32_t mndProcessCreateMnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto _OVER;
   }
 
@@ -535,18 +535,25 @@ static int32_t mndSetDropMnodeRedoActions(SMnode *pMnode, STrans *pTrans, SDnode
   return 0;
 }
 
+int32_t mndSetDropMnodeInfoToTrans(SMnode *pMnode, STrans *pTrans, SMnodeObj *pObj) {
+  if (pObj == NULL) return 0;
+  if (mndSetDropMnodeRedoLogs(pMnode, pTrans, pObj) != 0) return -1;
+  if (mndSetDropMnodeCommitLogs(pMnode, pTrans, pObj) != 0) return -1;
+  if (mndSetDropMnodeRedoActions(pMnode, pTrans, pObj->pDnode, pObj) != 0) return -1;
+  if (mndTransAppendNullLog(pTrans) != 0) return -1;
+  return 0;
+}
+
 static int32_t mndDropMnode(SMnode *pMnode, SRpcMsg *pReq, SMnodeObj *pObj) {
   int32_t code = -1;
+  STrans *pTrans = NULL;
 
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
+  pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
   if (pTrans == NULL) goto _OVER;
-
-  mDebug("trans:%d, used to drop mnode:%d", pTrans->id, pObj->id);
   mndTransSetSerial(pTrans);
-  if (mndSetDropMnodeRedoLogs(pMnode, pTrans, pObj) != 0) goto _OVER;
+  mDebug("trans:%d, used to drop mnode:%d", pTrans->id, pObj->id);
-  if (mndSetDropMnodeCommitLogs(pMnode, pTrans, pObj) != 0) goto _OVER;
+
-  if (mndSetDropMnodeRedoActions(pMnode, pTrans, pObj->pDnode, pObj) != 0) goto _OVER;
+  if (mndSetDropMnodeInfoToTrans(pMnode, pTrans, pObj) != 0) goto _OVER;
-  if (mndTransAppendNullLog(pTrans) != 0) goto _OVER;
   if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
 
   code = 0;
@@ -596,7 +603,7 @@ static int32_t mndProcessDropMnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto _OVER;
   }
 
@@ -642,7 +649,7 @@ static int32_t mndRetrieveMnodes(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pB
     if (pObj->id == pMnode->selfDnodeId) {
       roles = syncStr(TAOS_SYNC_STATE_LEADER);
     }
-    if (pObj->pDnode && mndIsDnodeOnline(pMnode, pObj->pDnode, curMs)) {
+    if (pObj->pDnode && mndIsDnodeOnline(pObj->pDnode, curMs)) {
       roles = syncStr(pObj->state);
     }
     char b2[12 + VARSTR_HEADER_SIZE] = {0};

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

@@ -301,7 +301,7 @@ static int32_t mndProcessCreateQnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto _OVER;
   }
 
@@ -411,7 +411,7 @@ static int32_t mndProcessDropQnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto _OVER;
   }
 

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

@@ -190,7 +190,7 @@ int32_t mndAddShuffledSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* p
       sdbRelease(pSdb, pVgroup);
       continue;
     }
-    SStreamTask* pTask = tNewSStreamTask(pStream->uid);
+    SStreamTask* pTask = tNewSStreamTask(pStream->uid, 0);
     if (pTask == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
       return -1;
@@ -230,7 +230,7 @@ int32_t mndAddShuffledSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* p
 int32_t mndAddFixedSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
   ASSERT(pStream->fixedSinkVgId != 0);
   SArray*      tasks = taosArrayGetP(pStream->tasks, 0);
-  SStreamTask* pTask = tNewSStreamTask(pStream->uid);
+  SStreamTask* pTask = tNewSStreamTask(pStream->uid, 0);
   if (pTask == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
@@ -322,7 +322,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
           sdbRelease(pSdb, pVgroup);
           continue;
         }
-        SStreamTask* pTask = tNewSStreamTask(pStream->uid);
+        SStreamTask* pTask = tNewSStreamTask(pStream->uid, 0);
         // source part
         pTask->sourceType = TASK_SOURCE__SCAN;
         pTask->inputType = TASK_INPUT_TYPE__SUMBIT_BLOCK;
@@ -387,7 +387,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
 
       // else, assign to vnode
       ASSERT(plan->subplanType == SUBPLAN_TYPE_MERGE);
-      SStreamTask* pTask = tNewSStreamTask(pStream->uid);
+      SStreamTask* pTask = tNewSStreamTask(pStream->uid, 0);
 
       // source part, currently only support multi source
       pTask->sourceType = TASK_SOURCE__PIPE;
@@ -477,7 +477,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
         sdbRelease(pSdb, pVgroup);
         continue;
       }
-      SStreamTask* pTask = tNewSStreamTask(pStream->uid);
+      SStreamTask* pTask = tNewSStreamTask(pStream->uid, 0);
 
       // source part
       pTask->sourceType = TASK_SOURCE__MERGE;

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

@@ -36,6 +36,7 @@ static SSdbRow *mndSmaActionDecode(SSdbRaw *pRaw);
 static int32_t  mndSmaActionInsert(SSdb *pSdb, SSmaObj *pSma);
 static int32_t  mndSmaActionDelete(SSdb *pSdb, SSmaObj *pSpSmatb);
 static int32_t  mndSmaActionUpdate(SSdb *pSdb, SSmaObj *pOld, SSmaObj *pNew);
+static int32_t  mndSmaGetVgEpSet(SMnode *pMnode, SDbObj *pDb, SVgEpSet **ppVgEpSet, int32_t *numOfVgroups);
 static int32_t  mndProcessMCreateSmaReq(SRpcMsg *pReq);
 static int32_t  mndProcessMDropSmaReq(SRpcMsg *pReq);
 static int32_t  mndProcessGetSmaReq(SRpcMsg *pReq);
@@ -262,6 +263,8 @@ static void *mndBuildVCreateSmaReq(SMnode *pMnode, SVgObj *pVgroup, SSmaObj *pSm
   req.sliding = pSma->sliding;
   req.expr = pSma->expr;
   req.tagsFilter = pSma->tagsFilter;
+  req.numOfVgroups = pSma->numOfVgroups;
+  req.pVgEpSet = pSma->pVgEpSet;
   
   // get length
   int32_t ret = 0;
@@ -420,6 +423,15 @@ static int32_t mndSetCreateSmaVgroupRedoActions(SMnode *pMnode, STrans *pTrans,
   mndReleaseDnode(pMnode, pDnode);
 
   // todo add sma info here
+  SVgEpSet *pVgEpSet = NULL;
+  int32_t   numOfVgroups = 0;
+  if (mndSmaGetVgEpSet(pMnode, pDb, &pVgEpSet, &numOfVgroups) != 0) {
+    return -1;
+  }
+
+  pSma->pVgEpSet = pVgEpSet;
+  pSma->numOfVgroups = numOfVgroups;
+
   int32_t smaContLen = 0;
   void   *pSmaReq = mndBuildVCreateSmaReq(pMnode, pVgroup, pSma, &smaContLen);
   if (pSmaReq == NULL) return -1;
@@ -510,10 +522,9 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea
   int32_t code = -1;
   STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_DB, pReq);
   if (pTrans == NULL) goto _OVER;
-
-  mDebug("trans:%d, used to create sma:%s", pTrans->id, pCreate->name);
   mndTransSetDbName(pTrans, pDb->name);
   mndTransSetSerial(pTrans);
+  mDebug("trans:%d, used to create sma:%s", pTrans->id, pCreate->name);
 
   if (mndSetCreateSmaRedoLogs(pMnode, pTrans, &smaObj) != 0) goto _OVER;
   if (mndSetCreateSmaVgroupRedoLogs(pMnode, pTrans, &streamObj.fixedSinkVg) != 0) goto _OVER;
@@ -964,3 +975,52 @@ static void mndCancelGetNextSma(SMnode *pMnode, void *pIter) {
   SSdb *pSdb = pMnode->pSdb;
   sdbCancelFetch(pSdb, pIter);
 }
+
+static int32_t mndSmaGetVgEpSet(SMnode *pMnode, SDbObj *pDb, SVgEpSet **ppVgEpSet, int32_t *numOfVgroups) {
+  SSdb     *pSdb = pMnode->pSdb;
+  SVgObj   *pVgroup = NULL;
+  void     *pIter = NULL;
+  SVgEpSet *pVgEpSet = NULL;
+  int32_t   nAllocVgs = 16;
+  int32_t   nVgs = 0;
+
+  pVgEpSet = taosMemoryCalloc(nAllocVgs, sizeof(SVgEpSet));
+  if (!pVgEpSet) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    return -1;
+  }
+
+  while (1) {
+    pIter = sdbFetch(pSdb, SDB_VGROUP, pIter, (void **)&pVgroup);
+    if (pIter == NULL) break;
+    if (pVgroup->dbUid != pDb->uid) {
+      sdbRelease(pSdb, pVgroup);
+      continue;
+    }
+
+    if (nVgs >= nAllocVgs) {
+      void *p = taosMemoryRealloc(pVgEpSet, nAllocVgs * 2 * sizeof(SVgEpSet));
+      if (!p) {
+        taosMemoryFree(pVgEpSet);
+        sdbCancelFetch(pSdb, pIter);
+        sdbRelease(pSdb, pVgroup);
+        terrno = TSDB_CODE_OUT_OF_MEMORY;
+        return -1;
+      }
+      pVgEpSet = (SVgEpSet *)p;
+      nAllocVgs *= 2;
+    }
+
+    (pVgEpSet + nVgs)->vgId = pVgroup->vgId;
+    (pVgEpSet + nVgs)->epSet = mndGetVgroupEpset(pMnode, pVgroup);
+
+    ++nVgs;
+
+    sdbRelease(pSdb, pVgroup);
+  }
+
+  *ppVgEpSet = pVgEpSet;
+  *numOfVgroups = nVgs;
+
+  return 0;
+}

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

@@ -307,7 +307,7 @@ static int32_t mndProcessCreateSnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto _OVER;
   }
 
@@ -419,7 +419,7 @@ static int32_t mndProcessDropSnodeReq(SRpcMsg *pReq) {
     goto _OVER;
   }
 
-  if (mndCheckNodeAuth(pUser)) {
+  if (mndCheckNodeAuth(pUser) != 0) {
     goto _OVER;
   }
 

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

@@ -1108,7 +1108,7 @@ static int32_t mndDropSuperTableColumn(SMnode *pMnode, const SStbObj *pOld, SStb
     return -1;
   }
 
-  col_id_t colId = pOld->pTags[col].colId;
+  col_id_t colId = pOld->pColumns[col].colId;
   if (mndCheckColAndTagModifiable(pMnode, pOld->uid, colId) != 0) {
     return -1;
   }
@@ -1142,7 +1142,7 @@ static int32_t mndAlterStbColumnBytes(SMnode *pMnode, const SStbObj *pOld, SStbO
     return -1;
   }
 
-  col_id_t colId = pOld->pTags[col].colId;
+  col_id_t colId = pOld->pColumns[col].colId;
   if (mndCheckColAndTagModifiable(pMnode, pOld->uid, colId) != 0) {
     return -1;
   }

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

@@ -91,7 +91,7 @@ int32_t mndCheckColAndTagModifiable(SMnode *pMnode, int64_t suid, col_id_t colId
     }
 
     SHashObj  *pColHash = NULL;
-    SNodeList *pNodeList;
+    SNodeList *pNodeList = NULL;
     nodesCollectColumns((SSelectStmt *)pAst, SQL_CLAUSE_FROM, NULL, COLLECT_COL_TYPE_ALL, &pNodeList);
     SNode *pNode = NULL;
     FOREACH(pNode, pNodeList) {
@@ -103,6 +103,7 @@ int32_t mndCheckColAndTagModifiable(SMnode *pMnode, int64_t suid, col_id_t colId
       if (pCol->colId > 0) {
         taosHashPut(pColHash, &pCol->colId, sizeof(int16_t), NULL, 0);
       }
+      mTrace("topic:%s, colId:%d is used", pTopic->name, pCol->colId);
     }
 
     if (taosHashGet(pColHash, &colId, sizeof(int16_t)) != NULL) {

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

@@ -15,11 +15,13 @@
 
 #define _DEFAULT_SOURCE
 #include "mndVgroup.h"
+#include "mndAuth.h"
 #include "mndDb.h"
 #include "mndDnode.h"
 #include "mndMnode.h"
 #include "mndShow.h"
 #include "mndTrans.h"
+#include "mndUser.h"
 
 #define VGROUP_VER_NUMBER   1
 #define VGROUP_RESERVE_SIZE 64
@@ -34,6 +36,10 @@ static void    mndCancelGetNextVgroup(SMnode *pMnode, void *pIter);
 static int32_t mndRetrieveVnodes(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows);
 static void    mndCancelGetNextVnode(SMnode *pMnode, void *pIter);
 
+static int32_t mndProcessRedistributeVgroupMsg(SRpcMsg *pReq);
+static int32_t mndProcessSplitVgroupMsg(SRpcMsg *pReq);
+static int32_t mndProcessBalanceVgroupMsg(SRpcMsg *pReq);
+
 int32_t mndInitVgroup(SMnode *pMnode) {
   SSdbTable table = {
       .sdbType = SDB_VGROUP,
@@ -344,9 +350,14 @@ static bool mndResetDnodesArrayFp(SMnode *pMnode, void *pObj, void *p1, void *p2
 static bool mndBuildDnodesArrayFp(SMnode *pMnode, void *pObj, void *p1, void *p2, void *p3) {
   SDnodeObj *pDnode = pObj;
   SArray    *pArray = p1;
+  int32_t    exceptDnodeId = *(int32_t *)p2;
+
+  if (exceptDnodeId == pDnode->id) {
+    return true;
+  }
 
   int64_t curMs = taosGetTimestampMs();
-  bool    online = mndIsDnodeOnline(pMnode, pDnode, curMs);
+  bool    online = mndIsDnodeOnline(pDnode, curMs);
   bool    isMnode = mndIsMnode(pMnode, pDnode->id);
   pDnode->numOfVnodes = mndGetVnodesNum(pMnode, pDnode->id);
 
@@ -363,7 +374,7 @@ static bool mndBuildDnodesArrayFp(SMnode *pMnode, void *pObj, void *p1, void *p2
   return true;
 }
 
-SArray *mndBuildDnodesArray(SMnode *pMnode) {
+SArray *mndBuildDnodesArray(SMnode *pMnode, int32_t exceptDnodeId) {
   SSdb   *pSdb = pMnode->pSdb;
   int32_t numOfDnodes = mndGetDnodeSize(pMnode);
 
@@ -374,7 +385,7 @@ SArray *mndBuildDnodesArray(SMnode *pMnode) {
   }
 
   sdbTraverse(pSdb, SDB_DNODE, mndResetDnodesArrayFp, NULL, NULL, NULL);
-  sdbTraverse(pSdb, SDB_DNODE, mndBuildDnodesArrayFp, pArray, NULL, NULL);
+  sdbTraverse(pSdb, SDB_DNODE, mndBuildDnodesArrayFp, pArray, &exceptDnodeId, NULL);
   return pArray;
 }
 
@@ -422,7 +433,7 @@ static int32_t mndGetAvailableDnode(SMnode *pMnode, SVgObj *pVgroup, SArray *pAr
 }
 
 int32_t mndAllocSmaVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj *pVgroup) {
-  SArray *pArray = mndBuildDnodesArray(pMnode);
+  SArray *pArray = mndBuildDnodesArray(pMnode, 0);
   if (pArray == NULL) return -1;
 
   pVgroup->vgId = sdbGetMaxId(pMnode->pSdb, SDB_VGROUP);
@@ -451,7 +462,7 @@ int32_t mndAllocVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj **ppVgroups) {
     goto _OVER;
   }
 
-  pArray = mndBuildDnodesArray(pMnode);
+  pArray = mndBuildDnodesArray(pMnode, 0);
   if (pArray == NULL) goto _OVER;
 
   mInfo("db:%s, total %d dnodes used to create %d vgroups (%d vnodes)", pDb->name, (int32_t)taosArrayGetSize(pArray),
@@ -501,86 +512,6 @@ _OVER:
   return code;
 }
 
-int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
-  taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
-  for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
-    SDnodeObj *pDnode = taosArrayGet(pArray, i);
-    mDebug("dnode:%d, equivalent vnodes:%d", pDnode->id, pDnode->numOfVnodes);
-  }
-
-  SVnodeGid *pVgid = &pVgroup->vnodeGid[pVgroup->replica];
-  for (int32_t d = 0; d < taosArrayGetSize(pArray); ++d) {
-    SDnodeObj *pDnode = taosArrayGet(pArray, d);
-
-    bool used = false;
-    for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
-      if (pDnode->id == pVgroup->vnodeGid[vn].dnodeId) {
-        used = true;
-        break;
-      }
-    }
-    if (used) continue;
-
-    if (pDnode == NULL || pDnode->numOfVnodes > pDnode->numOfSupportVnodes) {
-      terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
-      return -1;
-    }
-
-    pVgid->dnodeId = pDnode->id;
-    pVgid->role = TAOS_SYNC_STATE_ERROR;
-    mInfo("db:%s, vgId:%d, vn:%d dnode:%d, is added", pVgroup->dbName, pVgroup->vgId, pVgroup->replica,
-          pVgid->dnodeId);
-
-    pVgroup->replica++;
-    pDnode->numOfVnodes++;
-    return 0;
-  }
-
-  terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
-  mError("db:%s, failed to add vnode to vgId:%d since %s", pVgroup->dbName, pVgroup->vgId, terrstr());
-  return -1;
-}
-
-int32_t mndRemoveVnodeFromVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray, SVnodeGid *pDelVgid) {
-  taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
-  for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
-    SDnodeObj *pDnode = taosArrayGet(pArray, i);
-    mDebug("dnode:%d, equivalent vnodes:%d", pDnode->id, pDnode->numOfVnodes);
-  }
-
-  int32_t code = -1;
-  for (int32_t d = taosArrayGetSize(pArray) - 1; d >= 0; --d) {
-    SDnodeObj *pDnode = taosArrayGet(pArray, d);
-
-    for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
-      SVnodeGid *pVgid = &pVgroup->vnodeGid[vn];
-      if (pVgid->dnodeId == pDnode->id) {
-        mInfo("db:%s, vgId:%d, vn:%d dnode:%d, is removed", pVgroup->dbName, pVgroup->vgId, vn, pVgid->dnodeId);
-        pDnode->numOfVnodes--;
-        pVgroup->replica--;
-        *pDelVgid = *pVgid;
-        *pVgid = pVgroup->vnodeGid[pVgroup->replica];
-        memset(&pVgroup->vnodeGid[pVgroup->replica], 0, sizeof(SVnodeGid));
-        code = 0;
-        goto _OVER;
-      }
-    }
-  }
-
-_OVER:
-  if (code != 0) {
-    terrno = TSDB_CODE_APP_ERROR;
-    mError("db:%s, failed to remove vnode from vgId:%d since %s", pVgroup->dbName, pVgroup->vgId, terrstr());
-    return -1;
-  }
-
-  for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
-    SVnodeGid *pVgid = &pVgroup->vnodeGid[vn];
-    mInfo("db:%s, vgId:%d, vn:%d dnode:%d is reserved", pVgroup->dbName, pVgroup->vgId, vn, pVgid->dnodeId);
-  }
-  return 0;
-}
-
 SEpSet mndGetVgroupEpset(SMnode *pMnode, const SVgObj *pVgroup) {
   SEpSet epset = {0};
 
@@ -678,7 +609,7 @@ static int32_t mndRetrieveVgroups(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *p
         bool       online = false;
         SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgroup->vnodeGid[i].dnodeId);
         if (pDnode != NULL) {
-          online = mndIsDnodeOnline(pMnode, pDnode, curMs);
+          online = mndIsDnodeOnline(pDnode, curMs);
           mndReleaseDnode(pMnode, pDnode);
         }
 
@@ -797,3 +728,597 @@ static void mndCancelGetNextVnode(SMnode *pMnode, void *pIter) {
   SSdb *pSdb = pMnode->pSdb;
   sdbCancelFetch(pSdb, pIter);
 }
+
+int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
+  taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
+  for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
+    SDnodeObj *pDnode = taosArrayGet(pArray, i);
+    mDebug("dnode:%d, equivalent vnodes:%d", pDnode->id, pDnode->numOfVnodes);
+  }
+
+  SVnodeGid *pVgid = &pVgroup->vnodeGid[pVgroup->replica];
+  for (int32_t d = 0; d < taosArrayGetSize(pArray); ++d) {
+    SDnodeObj *pDnode = taosArrayGet(pArray, d);
+
+    bool used = false;
+    for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
+      if (pDnode->id == pVgroup->vnodeGid[vn].dnodeId) {
+        used = true;
+        break;
+      }
+    }
+    if (used) continue;
+
+    if (pDnode == NULL || pDnode->numOfVnodes > pDnode->numOfSupportVnodes) {
+      terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
+      return -1;
+    }
+
+    pVgid->dnodeId = pDnode->id;
+    pVgid->role = TAOS_SYNC_STATE_ERROR;
+    mInfo("db:%s, vgId:%d, vn:%d dnode:%d, is added", pVgroup->dbName, pVgroup->vgId, pVgroup->replica, pVgid->dnodeId);
+
+    pVgroup->replica++;
+    pDnode->numOfVnodes++;
+    return 0;
+  }
+
+  terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
+  mError("db:%s, failed to add vnode to vgId:%d since %s", pVgroup->dbName, pVgroup->vgId, terrstr());
+  return -1;
+}
+
+int32_t mndRemoveVnodeFromVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray, SVnodeGid *pDelVgid) {
+  taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
+  for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
+    SDnodeObj *pDnode = taosArrayGet(pArray, i);
+    mDebug("dnode:%d, equivalent vnodes:%d", pDnode->id, pDnode->numOfVnodes);
+  }
+
+  int32_t code = -1;
+  for (int32_t d = taosArrayGetSize(pArray) - 1; d >= 0; --d) {
+    SDnodeObj *pDnode = taosArrayGet(pArray, d);
+
+    for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
+      SVnodeGid *pVgid = &pVgroup->vnodeGid[vn];
+      if (pVgid->dnodeId == pDnode->id) {
+        mInfo("db:%s, vgId:%d, vn:%d dnode:%d, is removed", pVgroup->dbName, pVgroup->vgId, vn, pVgid->dnodeId);
+        pDnode->numOfVnodes--;
+        pVgroup->replica--;
+        *pDelVgid = *pVgid;
+        *pVgid = pVgroup->vnodeGid[pVgroup->replica];
+        memset(&pVgroup->vnodeGid[pVgroup->replica], 0, sizeof(SVnodeGid));
+        code = 0;
+        goto _OVER;
+      }
+    }
+  }
+
+_OVER:
+  if (code != 0) {
+    terrno = TSDB_CODE_APP_ERROR;
+    mError("db:%s, failed to remove vnode from vgId:%d since %s", pVgroup->dbName, pVgroup->vgId, terrstr());
+    return -1;
+  }
+
+  for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
+    SVnodeGid *pVgid = &pVgroup->vnodeGid[vn];
+    mInfo("db:%s, vgId:%d, vn:%d dnode:%d is reserved", pVgroup->dbName, pVgroup->vgId, vn, pVgid->dnodeId);
+  }
+  return 0;
+}
+
+int32_t mndAddCreateVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid,
+                                bool standby) {
+  STransAction action = {0};
+
+  SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
+  if (pDnode == NULL) return -1;
+  action.epSet = mndGetDnodeEpset(pDnode);
+  mndReleaseDnode(pMnode, pDnode);
+
+  int32_t contLen = 0;
+  void   *pReq = mndBuildCreateVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen, standby);
+  if (pReq == NULL) return -1;
+
+  action.pCont = pReq;
+  action.contLen = contLen;
+  action.msgType = TDMT_DND_CREATE_VNODE;
+  action.acceptableCode = TSDB_CODE_NODE_ALREADY_DEPLOYED;
+
+  if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+    taosMemoryFree(pReq);
+    return -1;
+  }
+
+  return 0;
+}
+
+int32_t mndAddAlterVnodeConfirmAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup) {
+  STransAction action = {0};
+  action.epSet = mndGetVgroupEpset(pMnode, pVgroup);
+
+  int32_t   contLen = sizeof(SMsgHead);
+  SMsgHead *pHead = taosMemoryMalloc(contLen);
+  if (pHead == NULL) {
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    return -1;
+  }
+
+  pHead->contLen = htonl(contLen);
+  pHead->vgId = htonl(pVgroup->vgId);
+
+  action.pCont = pHead;
+  action.contLen = contLen;
+  action.msgType = TDMT_VND_ALTER_CONFIRM;
+
+  if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+    taosMemoryFree(pHead);
+    return -1;
+  }
+
+  return 0;
+}
+
+int32_t mndAddAlterVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, tmsg_t msgType) {
+  STransAction action = {0};
+  action.epSet = mndGetVgroupEpset(pMnode, pVgroup);
+
+  int32_t contLen = 0;
+  void   *pReq = mndBuildAlterVnodeReq(pMnode, pDb, pVgroup, &contLen);
+  if (pReq == NULL) return -1;
+
+  action.pCont = pReq;
+  action.contLen = contLen;
+  action.msgType = msgType;
+
+  if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+    taosMemoryFree(pReq);
+    return -1;
+  }
+
+  return 0;
+}
+
+int32_t mndAddDropVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid,
+                              bool isRedo) {
+  STransAction action = {0};
+
+  SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
+  if (pDnode == NULL) return -1;
+  action.epSet = mndGetDnodeEpset(pDnode);
+  mndReleaseDnode(pMnode, pDnode);
+
+  int32_t contLen = 0;
+  void   *pReq = mndBuildDropVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
+  if (pReq == NULL) return -1;
+
+  action.pCont = pReq;
+  action.contLen = contLen;
+  action.msgType = TDMT_DND_DROP_VNODE;
+  action.acceptableCode = TSDB_CODE_NODE_NOT_DEPLOYED;
+
+  if (isRedo) {
+    if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+      taosMemoryFree(pReq);
+      return -1;
+    }
+  } else {
+    if (mndTransAppendUndoAction(pTrans, &action) != 0) {
+      taosMemoryFree(pReq);
+      return -1;
+    }
+  }
+
+  return 0;
+}
+
+int32_t mndSetMoveVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, int32_t vnIndex,
+                                    SArray *pArray) {
+  SVgObj newVg = {0};
+  memcpy(&newVg, pVgroup, sizeof(SVgObj));
+
+  mInfo("vgId:%d, vgroup info before move, replica:%d", newVg.vgId, newVg.replica);
+  for (int32_t i = 0; i < newVg.replica; ++i) {
+    mInfo("vgId:%d, vnode:%d dnode:%d", newVg.vgId, i, newVg.vnodeGid[i].dnodeId);
+  }
+
+  mInfo("vgId:%d, will add 1 vnodes", pVgroup->vgId);
+  if (mndAddVnodeToVgroup(pMnode, &newVg, pArray) != 0) return -1;
+  if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVg, &newVg.vnodeGid[1], true) != 0) return -1;
+  if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg, TDMT_VND_ALTER_REPLICA) != 0) return -1;
+  if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVg) != 0) return -1;
+
+  mInfo("vgId:%d, will remove 1 vnodes", pVgroup->vgId);
+  newVg.replica--;
+  SVnodeGid del = newVg.vnodeGid[vnIndex];
+  newVg.vnodeGid[vnIndex] = newVg.vnodeGid[newVg.replica];
+  memset(&newVg.vnodeGid[newVg.replica], 0, sizeof(SVnodeGid));
+  if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg, TDMT_VND_ALTER_REPLICA) != 0) return -1;
+  if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVg, &del, true) != 0) return -1;
+  if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVg) != 0) return -1;
+
+  mInfo("vgId:%d, vgroup info after move, replica:%d", newVg.vgId, newVg.replica);
+  for (int32_t i = 0; i < newVg.replica; ++i) {
+    mInfo("vgId:%d, vnode:%d dnode:%d", newVg.vgId, i, newVg.vnodeGid[i].dnodeId);
+  }
+  return 0;
+}
+
+int32_t mndSetMoveVgroupsInfoToTrans(SMnode *pMnode, STrans *pTrans, int32_t delDnodeId) {
+  SArray *pArray = mndBuildDnodesArray(pMnode, delDnodeId);
+  if (pArray == NULL) return -1;
+
+  void *pIter = NULL;
+  while (1) {
+    SVgObj *pVgroup = NULL;
+    pIter = sdbFetch(pMnode->pSdb, SDB_VGROUP, pIter, (void **)&pVgroup);
+    if (pIter == NULL) break;
+
+    int32_t vnIndex = -1;
+    for (int32_t i = 0; i < pVgroup->replica; ++i) {
+      if (pVgroup->vnodeGid[i].dnodeId == delDnodeId) {
+        vnIndex = i;
+        break;
+      }
+    }
+
+    if (vnIndex != -1) {
+      mInfo("vgId:%d, vnode:%d will be removed from dnode:%d", pVgroup->vgId, vnIndex, delDnodeId);
+      SDbObj *pDb = mndAcquireDb(pMnode, pVgroup->dbName);
+      mndSetMoveVgroupInfoToTrans(pMnode, pTrans, pDb, pVgroup, vnIndex, pArray);
+      mndReleaseDb(pMnode, pDb);
+    }
+
+    sdbRelease(pMnode->pSdb, pVgroup);
+  }
+
+  taosArrayDestroy(pArray);
+  return 0;
+}
+
+static int32_t mndAddIncVgroupReplicaToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup,
+                                             int32_t newDnodeId) {
+  mDebug("vgId:%d, will add 1 vnode, replica:%d, dnode:%d", pVgroup->vgId, pVgroup->replica, newDnodeId);
+
+  SVnodeGid *pGid = &pVgroup->vnodeGid[pVgroup->replica];
+  pVgroup->replica++;
+  pGid->dnodeId = newDnodeId;
+  pGid->role = TAOS_SYNC_STATE_ERROR;
+
+  if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, pVgroup, pGid, true) != 0) return -1;
+  if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, pVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
+  if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, pVgroup) != 0) return -1;
+
+  return 0;
+}
+
+static int32_t mndAddDecVgroupReplicaFromTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup,
+                                               int32_t delDnodeId) {
+  mDebug("vgId:%d, will remove 1 vnode, replica:%d, dnode:%d", pVgroup->vgId, pVgroup->replica, delDnodeId);
+
+  SVnodeGid *pGid = NULL;
+  SVnodeGid  delGid = {0};
+  for (int32_t i = 0; i < pVgroup->replica; ++i) {
+    if (pVgroup->vnodeGid[i].dnodeId == delDnodeId) {
+      pGid = &pVgroup->vnodeGid[i];
+      break;
+    }
+  }
+
+  if (pGid == NULL) return 0;
+
+  memcpy(&delGid, pGid, sizeof(SVnodeGid));
+  memcpy(pGid, &pVgroup->vnodeGid[pVgroup->replica], sizeof(SVnodeGid));
+  memset(&pVgroup->vnodeGid[pVgroup->replica], 0, sizeof(SVnodeGid));
+  pVgroup->replica--;
+
+  if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, pVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
+  if (mndAddDropVnodeAction(pMnode, pTrans, pDb, pVgroup, &delGid, true) != 0) return -1;
+  if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, pVgroup) != 0) return -1;
+
+  return 0;
+}
+
+static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb, SVgObj *pVgroup, SDnodeObj *pNew1,
+                                     SDnodeObj *pOld1, SDnodeObj *pNew2, SDnodeObj *pOld2, SDnodeObj *pNew3,
+                                     SDnodeObj *pOld3) {
+  int32_t  code = -1;
+  SSdbRaw *pRaw = NULL;
+  STrans  *pTrans = NULL;
+
+  pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
+  if (pTrans == NULL) goto _OVER;
+  mndTransSetSerial(pTrans);
+  mDebug("trans:%d, used to drop redistribute vgId:%d", pTrans->id, pVgroup->vgId);
+
+  SVgObj newVg = {0};
+  memcpy(&newVg, pVgroup, sizeof(SVgObj));
+  mInfo("vgId:%d, vgroup info before redistribute, replica:%d", newVg.vgId, newVg.replica);
+  for (int32_t i = 0; i < newVg.replica; ++i) {
+    mInfo("vgId:%d, vnode:%d dnode:%d", newVg.vgId, i, newVg.vnodeGid[i].dnodeId);
+  }
+
+  if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew1->id) != 0) goto _OVER;
+  if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld1->id) != 0) goto _OVER;
+  if (pNew2 != NULL) {
+    if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew2->id) != 0) goto _OVER;
+    if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld2->id) != 0) goto _OVER;
+    if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew3->id) != 0) goto _OVER;
+    if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld3->id) != 0) goto _OVER;
+  }
+
+  pRaw = mndVgroupActionEncode(&newVg);
+  if (pRaw == NULL || mndTransAppendCommitlog(pTrans, pRaw) != 0) goto _OVER;
+  sdbSetRawStatus(pRaw, SDB_STATUS_READY);
+  pRaw = NULL;
+
+  mInfo("vgId:%d, vgroup info after redistribute, replica:%d", newVg.vgId, newVg.replica);
+  for (int32_t i = 0; i < newVg.replica; ++i) {
+    mInfo("vgId:%d, vnode:%d dnode:%d", newVg.vgId, i, newVg.vnodeGid[i].dnodeId);
+  }
+
+  if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
+  code = 0;
+
+_OVER:
+  mndTransDrop(pTrans);
+  sdbFreeRaw(pRaw);
+  mndReleaseDb(pMnode, pDb);
+  return code;
+}
+
+static int32_t mndProcessRedistributeVgroupMsg(SRpcMsg *pReq) {
+  SMnode        *pMnode = pReq->info.node;
+  SUserObj      *pUser = NULL;
+  SDnodeObj     *pNew1 = NULL;
+  SDnodeObj     *pNew2 = NULL;
+  SDnodeObj     *pNew3 = NULL;
+  SDnodeObj     *pOld1 = NULL;
+  SDnodeObj     *pOld2 = NULL;
+  SDnodeObj     *pOld3 = NULL;
+  SVgObj        *pVgroup = NULL;
+  SDbObj        *pDb = NULL;
+  int32_t        code = -1;
+  int64_t        curMs = taosGetTimestampMs();
+  SMDropMnodeReq redReq = {0};
+
+#if 0
+  if (tDeserializeSCreateDropMQSBNodeReq(pReq->pCont, pReq->contLen, &dropReq) != 0) {
+    terrno = TSDB_CODE_INVALID_MSG;
+    goto _OVER;
+  }
+#endif
+
+  mDebug("vgId:%d, start to redistribute", 2);
+  pUser = mndAcquireUser(pMnode, pReq->conn.user);
+  if (pUser == NULL) {
+    terrno = TSDB_CODE_MND_NO_USER_FROM_CONN;
+    goto _OVER;
+  }
+
+  if (mndCheckNodeAuth(pUser) != 0) {
+    goto _OVER;
+  }
+
+  pVgroup = mndAcquireVgroup(pMnode, 2);
+  if (pVgroup == NULL) goto _OVER;
+
+  pDb = mndAcquireDb(pMnode, pVgroup->dbName);
+  if (pDb == NULL) goto _OVER;
+
+  if (pVgroup->replica == 1) {
+    pNew1 = mndAcquireDnode(pMnode, 1);
+    pOld1 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[0].dnodeId);
+    if (pNew1 == NULL || pOld1 == NULL) goto _OVER;
+    if (!mndIsDnodeOnline(pNew1, curMs) || !mndIsDnodeOnline(pOld1, curMs)) {
+      terrno = TSDB_CODE_NODE_OFFLINE;
+      goto _OVER;
+    }
+    if (pNew1 == pOld1) {
+      terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
+      goto _OVER;
+    }
+    if (mndRedistributeVgroup(pMnode, pReq, pDb, pVgroup, pNew1, pOld1, NULL, NULL, NULL, NULL) != 0) goto _OVER;
+  }
+
+  if (pVgroup->replica == 3) {
+    pNew1 = mndAcquireDnode(pMnode, 1);
+    pNew2 = mndAcquireDnode(pMnode, 2);
+    pNew3 = mndAcquireDnode(pMnode, 3);
+    pOld1 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[0].dnodeId);
+    pOld2 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[1].dnodeId);
+    pOld3 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[2].dnodeId);
+    if (pNew1 == NULL || pOld1 == NULL || pNew2 == NULL || pOld2 == NULL || pNew3 == NULL || pOld3 == NULL) goto _OVER;
+    if (!mndIsDnodeOnline(pNew1, curMs) || !mndIsDnodeOnline(pOld1, curMs) || !mndIsDnodeOnline(pNew2, curMs) ||
+        !mndIsDnodeOnline(pOld2, curMs) || !mndIsDnodeOnline(pNew3, curMs) || !mndIsDnodeOnline(pOld3, curMs)) {
+      terrno = TSDB_CODE_NODE_OFFLINE;
+      goto _OVER;
+    }
+    bool changed = true;
+    if (pNew1 != pOld1 || pNew1 != pOld2 || pNew1 != pOld3) changed = true;
+    if (pNew2 != pOld1 || pNew2 != pOld2 || pNew2 != pOld3) changed = true;
+    if (pNew3 != pOld1 || pNew3 != pOld2 || pNew3 != pOld3) changed = true;
+    if (!changed) {
+      terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
+      goto _OVER;
+    }
+    if (mndRedistributeVgroup(pMnode, pReq, pDb, pVgroup, pNew1, pOld1, pNew2, pOld2, pNew3, pOld3) != 0) goto _OVER;
+  }
+
+  if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;
+
+_OVER:
+  if (code != 0 && code != TSDB_CODE_ACTION_IN_PROGRESS) {
+    mDebug("vgId:%d, failed to redistribute since %s", 1, terrstr());
+  }
+
+  mndReleaseDnode(pMnode, pNew1);
+  mndReleaseDnode(pMnode, pNew2);
+  mndReleaseDnode(pMnode, pNew3);
+  mndReleaseDnode(pMnode, pOld1);
+  mndReleaseDnode(pMnode, pOld2);
+  mndReleaseDnode(pMnode, pOld3);
+  mndReleaseUser(pMnode, pUser);
+  mndReleaseVgroup(pMnode, pVgroup);
+  mndReleaseDb(pMnode, pDb);
+
+  return code;
+}
+
+static int32_t mndProcessSplitVgroupMsg(SRpcMsg *pReq) { return 0; }
+
+static int32_t mndSetBalanceVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup,
+                                              SDnodeObj *pSrc, SDnodeObj *pDst) {
+  SVgObj newVg = {0};
+  memcpy(&newVg, pVgroup, sizeof(SVgObj));
+  mInfo("vgId:%d, vgroup info before balance, replica:%d", newVg.vgId, newVg.replica);
+  for (int32_t i = 0; i < newVg.replica; ++i) {
+    mInfo("vgId:%d, vnode:%d dnode:%d", newVg.vgId, i, newVg.vnodeGid[i].dnodeId);
+  }
+
+  if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pDst->id) != 0) return -1;
+  if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pSrc->id) != 0) return -1;
+
+  SSdbRaw *pRaw = mndVgroupActionEncode(&newVg);
+  if (pRaw == NULL || mndTransAppendCommitlog(pTrans, pRaw) != 0) {
+    sdbFreeRaw(pRaw);
+    return -1;
+  }
+  sdbSetRawStatus(pRaw, SDB_STATUS_READY);
+
+  mInfo("vgId:%d, vgroup info after balance, replica:%d", newVg.vgId, newVg.replica);
+  for (int32_t i = 0; i < newVg.replica; ++i) {
+    mInfo("vgId:%d, vnode:%d dnode:%d", newVg.vgId, i, newVg.vnodeGid[i].dnodeId);
+  }
+  return 0;
+}
+
+static int32_t mndBalanceVgroupBetweenDnode(SMnode *pMnode, STrans *pTrans, SDnodeObj *pSrc, SDnodeObj *pDst) {
+  void   *pIter = NULL;
+  int32_t code = -1;
+
+  while (1) {
+    SVgObj *pVgroup = NULL;
+    pIter = sdbFetch(pMnode->pSdb, SDB_VGROUP, pIter, (void **)&pVgroup);
+    if (pIter == NULL) break;
+
+    bool existInSrc = false;
+    bool existInDst = false;
+    for (int32_t i = 0; i < pVgroup->replica; ++i) {
+      SVnodeGid *pGid = &pVgroup->vnodeGid[i];
+      if (pGid->dnodeId == pSrc->id) existInSrc = true;
+      if (pGid->dnodeId == pDst->id) existInDst = true;
+    }
+
+    if (!existInSrc || existInDst) {
+      sdbRelease(pMnode->pSdb, pVgroup);
+    }
+
+    SDbObj *pDb = mndAcquireDb(pMnode, pVgroup->dbName);
+    code = mndSetBalanceVgroupInfoToTrans(pMnode, pTrans, pDb, pVgroup, pSrc, pDst);
+    mndReleaseDb(pMnode, pDb);
+    sdbRelease(pMnode->pSdb, pVgroup);
+    break;
+  }
+
+  return code;
+}
+
+static int32_t mndBalanceVgroup(SMnode *pMnode, SRpcMsg *pReq, SArray *pArray) {
+  int32_t code = -1;
+  int32_t numOfVgroups = 0;
+  STrans *pTrans = NULL;
+
+  pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
+  if (pTrans == NULL) goto _OVER;
+  mndTransSetSerial(pTrans);
+  mDebug("trans:%d, used to balance vgroup", pTrans->id);
+
+  while (1) {
+    taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
+    SDnodeObj *pSrc = taosArrayGet(pArray, 0);
+    SDnodeObj *pDst = taosArrayGet(pArray, taosArrayGetSize(pArray) - 1);
+
+    float srcScore = (float)(pSrc->numOfVnodes - 1) / pSrc->numOfSupportVnodes;
+    float dstScore = (float)(pDst->numOfVnodes + 1) / pDst->numOfSupportVnodes;
+    if (srcScore + 0.0001 < dstScore) {
+      mDebug("trans:%d, balance vgroup from dnode:%d to dnode:%d", pTrans->id, pSrc->id, pDst->id);
+      code = mndBalanceVgroupBetweenDnode(pMnode, pTrans, pSrc, pDst);
+      if (code == 0) {
+        numOfVgroups++;
+        continue;
+      } else {
+        mError("trans:%d, failed to balance vgroup from dnode:%d to dnode:%d", pTrans->id, pSrc->id, pDst->id);
+        return -1;
+      }
+    } else {
+      mDebug("trans:%d, no vgroup need to balance vgroup any more", pTrans->id);
+      break;
+    }
+  }
+
+  if (numOfVgroups <= 0) {
+    mDebug("no need to balance vgroup");
+    code = 0;
+  } else {
+    mDebug("start to balance vgroup, numOfVgroups:%d", numOfVgroups);
+    if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
+    code = TSDB_CODE_ACTION_IN_PROGRESS;
+  }
+
+_OVER:
+  mndTransDrop(pTrans);
+  return code;
+}
+
+static int32_t mndProcessBalanceVgroupMsg(SRpcMsg *pReq) {
+  SMnode   *pMnode = pReq->info.node;
+  int32_t   code = -1;
+  SUserObj *pUser = NULL;
+  SArray   *pArray = NULL;
+  void     *pIter = NULL;
+  int64_t   curMs = taosGetTimestampMs();
+
+  mDebug("start to balance vgroup");
+  pUser = mndAcquireUser(pMnode, pReq->conn.user);
+  if (pUser == NULL) {
+    terrno = TSDB_CODE_MND_NO_USER_FROM_CONN;
+    goto _OVER;
+  }
+
+  if (mndCheckNodeAuth(pUser) != 0) goto _OVER;
+
+  while (1) {
+    SDnodeObj *pDnode = NULL;
+    pIter = sdbFetch(pMnode->pSdb, SDB_DNODE, pIter, (void **)&pDnode);
+    if (pIter == NULL) break;
+    if (!mndIsDnodeOnline(pDnode, curMs)) {
+      terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
+      mError("failed to balance vgroup since %s, dnode:%d", terrstr(), pDnode->id);
+      sdbRelease(pMnode->pSdb, pDnode);
+      goto _OVER;
+    }
+
+    sdbRelease(pMnode->pSdb, pDnode);
+  }
+
+  pArray = mndBuildDnodesArray(pMnode, 0);
+  if (pArray == NULL) goto _OVER;
+
+  if (taosArrayGetSize(pArray) < 2) {
+    mDebug("no need to balance vgroup since dnode num less than 2");
+    code = 0;
+  } else {
+    code = mndBalanceVgroup(pMnode, pReq, pArray);
+  }
+
+_OVER:
+  if (code != 0 && code != TSDB_CODE_ACTION_IN_PROGRESS) {
+    mError("failed to balance vgroup since %s", terrstr());
+  }
+
+  mndReleaseUser(pMnode, pUser);
+  taosArrayDestroy(pArray);
+  return code;
+}

source/dnode/mnode/sdb/src/sdbRaw.c

@@ -42,8 +42,10 @@ SSdbRaw *sdbAllocRaw(ESdbType type, int8_t sver, int32_t dataLen) {
 }
 
 void sdbFreeRaw(SSdbRaw *pRaw) {
+  if (pRaw != NULL) {
     mTrace("raw:%p, is freed", pRaw);
     taosMemoryFree(pRaw);
+  }
 }
 
 int32_t sdbSetRawInt8(SSdbRaw *pRaw, int32_t dataPos, int8_t val) {

source/dnode/qnode/src/qnode.c

@@ -59,6 +59,7 @@ int32_t qndGetLoad(SQnode *pQnode, SQnodeLoad *pLoad) {
   pLoad->numOfProcessedFetch = stat.fetchProcessed;
   pLoad->numOfProcessedDrop = stat.dropProcessed;
   pLoad->numOfProcessedHb = stat.hbProcessed;
+  pLoad->numOfProcessedDelete = stat.deleteProcessed;
   
   return 0; 
 }

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

@@ -80,7 +80,7 @@ typedef struct {
   // exec
   int8_t       inputStatus;
   int8_t       execStatus;
-  SStreamQ inputQ;
+  SStreamQueue inputQ;
   SRWLatch     lock;
 } STqPushHandle;
 
@@ -107,7 +107,7 @@ typedef struct {
     STqExecCol execCol;
     STqExecTb  execTb;
     STqExecDb  execDb;
-  } exec;
+  };
 } STqExecHandle;
 
 typedef struct {

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

@@ -103,7 +103,7 @@ static int metaSaveJsonVarToIdx(SMeta *pMeta, const SMetaEntry *pCtbEntry, const
   tIndexJsonPut(pMeta->pTagIvtIdx, terms, tuid);
   indexMultiTermDestroy(terms);
 #endif
-  return -1;
+  return 0;
 }
 
 int metaCreateSTable(SMeta *pMeta, int64_t version, SVCreateStbReq *pReq) {

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

@@ -216,8 +216,8 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) {
   }
 
   // TODO wrap in destroy func
-  taosArrayDestroy(rsp.blockData);
   taosArrayDestroy(rsp.blockDataLen);
+  taosArrayDestroyP(rsp.blockData, (FDelete)taosMemoryFree);
 
   if (rsp.withSchema) {
     taosArrayDestroyP(rsp.blockSchema, (FDelete)tDeleteSSchemaWrapper);
@@ -266,7 +266,7 @@ int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen) {
       pHandle->execHandle.pExecReader[i] = tqInitSubmitMsgScanner(pTq->pVnode->pMeta);
     }
     if (pHandle->execHandle.subType == TOPIC_SUB_TYPE__COLUMN) {
-      pHandle->execHandle.exec.execCol.qmsg = req.qmsg;
+      pHandle->execHandle.execCol.qmsg = req.qmsg;
       req.qmsg = NULL;
       for (int32_t i = 0; i < 5; i++) {
         SReadHandle handle = {
@@ -274,15 +274,14 @@ int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen) {
             .meta = pTq->pVnode->pMeta,
             .pMsgCb = &pTq->pVnode->msgCb,
         };
-        pHandle->execHandle.exec.execCol.task[i] =
+        pHandle->execHandle.execCol.task[i] = qCreateStreamExecTaskInfo(pHandle->execHandle.execCol.qmsg, &handle);
-            qCreateStreamExecTaskInfo(pHandle->execHandle.exec.execCol.qmsg, &handle);
+        ASSERT(pHandle->execHandle.execCol.task[i]);
-        ASSERT(pHandle->execHandle.exec.execCol.task[i]);
       }
     } else if (pHandle->execHandle.subType == TOPIC_SUB_TYPE__DB) {
-      pHandle->execHandle.exec.execDb.pFilterOutTbUid =
+      pHandle->execHandle.execDb.pFilterOutTbUid =
           taosHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_NO_LOCK);
     } else if (pHandle->execHandle.subType == TOPIC_SUB_TYPE__TABLE) {
-      pHandle->execHandle.exec.execTb.suid = req.suid;
+      pHandle->execHandle.execTb.suid = req.suid;
       SArray* tbUidList = taosArrayInit(0, sizeof(int64_t));
       tsdbGetCtbIdList(pTq->pVnode->pMeta, req.suid, tbUidList);
       tqDebug("vg %d, tq try get suid: %ld", pTq->pVnode->config.vgId, req.suid);
@@ -296,17 +295,20 @@ int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen) {
       taosArrayDestroy(tbUidList);
     }
     taosHashPut(pTq->handles, req.subKey, strlen(req.subKey), pHandle, sizeof(STqHandle));
+    if (tqMetaSaveHandle(pTq, req.subKey, pHandle) < 0) {
+      // TODO
+    }
   } else {
     /*ASSERT(pExec->consumerId == req.oldConsumerId);*/
     // TODO handle qmsg and exec modification
     atomic_store_32(&pHandle->epoch, -1);
     atomic_store_64(&pHandle->consumerId, req.newConsumerId);
     atomic_add_fetch_32(&pHandle->epoch, 1);
-  }
-
     if (tqMetaSaveHandle(pTq, req.subKey, pHandle) < 0) {
       // TODO
     }
+  }
+
   return 0;
 }
 
@@ -323,16 +325,13 @@ int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen) {
   tDecoderClear(&decoder);
 
   pTask->status = TASK_STATUS__IDLE;
+
+  pTask->inputQueue = streamQueueOpen();
+  pTask->outputQueue = streamQueueOpen();
   pTask->inputStatus = TASK_INPUT_STATUS__NORMAL;
   pTask->outputStatus = TASK_OUTPUT_STATUS__NORMAL;
 
-  pTask->inputQ = taosOpenQueue();
+  if (pTask->inputQueue == NULL || pTask->outputQueue == NULL) goto FAIL;
-  pTask->outputQ = taosOpenQueue();
-  pTask->inputQAll = taosAllocateQall();
-  pTask->outputQAll = taosAllocateQall();
-
-  if (pTask->inputQ == NULL || pTask->outputQ == NULL || pTask->inputQAll == NULL || pTask->outputQAll == NULL)
-    goto FAIL;
 
   // exec
   if (pTask->execType != TASK_EXEC__NONE) {
@@ -369,10 +368,8 @@ int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen) {
 
   return 0;
 FAIL:
-  if (pTask->inputQ) taosCloseQueue(pTask->inputQ);
+  if (pTask->inputQueue) streamQueueClose(pTask->inputQueue);
-  if (pTask->outputQ) taosCloseQueue(pTask->outputQ);
+  if (pTask->outputQueue) streamQueueClose(pTask->outputQueue);
-  if (pTask->inputQAll) taosFreeQall(pTask->inputQAll);
-  if (pTask->outputQAll) taosFreeQall(pTask->outputQAll);
   if (pTask) taosMemoryFree(pTask);
   return -1;
 }
@@ -393,38 +390,16 @@ int32_t tqProcessStreamTrigger(STQ* pTq, SSubmitReq* pReq) {
     SStreamTask* pTask = (SStreamTask*)pIter;
     if (pTask->inputType != STREAM_INPUT__DATA_SUBMIT) continue;
 
-    int8_t inputStatus = atomic_load_8(&pTask->inputStatus);
+    if (!failed) {
-    if (inputStatus == TASK_INPUT_STATUS__NORMAL) {
+      if (streamTaskInput(pTask, (SStreamQueueItem*)pSubmit) < 0) {
-      if (failed) {
-        atomic_store_8(&pTask->inputStatus, TASK_INPUT_STATUS__FAILED);
         continue;
       }
 
-      SStreamDataSubmit* pSubmitClone = streamSubmitRefClone(pSubmit);
+      if (streamTriggerByWrite(pTask, pTq->pVnode->config.vgId, &pTq->pVnode->msgCb) < 0) {
-      if (pSubmitClone == NULL) {
-        atomic_store_8(&pTask->inputStatus, TASK_INPUT_STATUS__FAILED);
         continue;
       }
-      taosWriteQitem(pTask->inputQ, pSubmitClone);
-
-      int8_t execStatus = atomic_load_8(&pTask->status);
-      if (execStatus == TASK_STATUS__IDLE || execStatus == TASK_STATUS__CLOSING) {
-        SStreamTaskRunReq* pRunReq = taosMemoryMalloc(sizeof(SStreamTaskRunReq));
-        if (pRunReq == NULL) continue;
-        // TODO: do we need htonl?
-        pRunReq->head.vgId = pTq->pVnode->config.vgId;
-        pRunReq->streamId = pTask->streamId;
-        pRunReq->taskId = pTask->taskId;
-        SRpcMsg msg = {
-            .msgType = TDMT_VND_TASK_RUN,
-            .pCont = pRunReq,
-            .contLen = sizeof(SStreamTaskRunReq),
-        };
-        tmsgPutToQueue(&pTq->pVnode->msgCb, FETCH_QUEUE, &msg);
-      }
-
     } else {
-      // blocked or stopped, do nothing
+      streamTaskInputFail(pTask);
     }
   }
 
@@ -446,10 +421,20 @@ int32_t tqProcessTaskRunReq(STQ* pTq, SRpcMsg* pMsg) {
 }
 
 int32_t tqProcessTaskDispatchReq(STQ* pTq, SRpcMsg* pMsg) {
-  SStreamDispatchReq* pReq = pMsg->pCont;
+  char*              msgStr = pMsg->pCont;
-  int32_t             taskId = pReq->taskId;
+  char*              msgBody = POINTER_SHIFT(msgStr, sizeof(SMsgHead));
+  int32_t            msgLen = pMsg->contLen - sizeof(SMsgHead);
+  SStreamDispatchReq req;
+  SDecoder           decoder;
+  tDecoderInit(&decoder, msgBody, msgLen);
+  tDecodeStreamDispatchReq(&decoder, &req);
+  int32_t      taskId = req.taskId;
   SStreamTask* pTask = taosHashGet(pTq->pStreamTasks, &taskId, sizeof(int32_t));
-  streamProcessDispatchReq(pTask, &pTq->pVnode->msgCb, pReq, pMsg);
+  SRpcMsg      rsp = {
+           .info = pMsg->info,
+           .code = 0,
+  };
+  streamProcessDispatchReq(pTask, &pTq->pVnode->msgCb, &req, &rsp);
   return 0;
 }
 

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

@@ -59,7 +59,7 @@ static int32_t tqAddTbNameToRsp(const STQ* pTq, const STqExecHandle* pExec, SMqD
 
 int32_t tqDataExec(STQ* pTq, STqExecHandle* pExec, SSubmitReq* pReq, SMqDataBlkRsp* pRsp, int32_t workerId) {
   if (pExec->subType == TOPIC_SUB_TYPE__COLUMN) {
-    qTaskInfo_t task = pExec->exec.execCol.task[workerId];
+    qTaskInfo_t task = pExec->execCol.task[workerId];
     ASSERT(task);
     qSetStreamInput(task, pReq, STREAM_DATA_TYPE_SUBMIT_BLOCK, false);
     while (1) {
@@ -101,7 +101,7 @@ int32_t tqDataExec(STQ* pTq, STqExecHandle* pExec, SSubmitReq* pReq, SMqDataBlkR
     pRsp->withSchema = 1;
     STqReadHandle* pReader = pExec->pExecReader[workerId];
     tqReadHandleSetMsg(pReader, pReq, 0);
-    while (tqNextDataBlockFilterOut(pReader, pExec->exec.execDb.pFilterOutTbUid)) {
+    while (tqNextDataBlockFilterOut(pReader, pExec->execDb.pFilterOutTbUid)) {
       SSDataBlock block = {0};
       if (tqRetrieveDataBlock(&block.pDataBlock, pReader, &block.info.groupId, &block.info.uid, &block.info.rows,
                               &block.info.numOfCols) < 0) {

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

@@ -22,7 +22,7 @@ static int32_t tEncodeSTqHandle(SEncoder* pEncoder, const STqHandle* pHandle) {
   if (tEncodeI32(pEncoder, pHandle->epoch) < 0) return -1;
   if (tEncodeI8(pEncoder, pHandle->execHandle.subType) < 0) return -1;
   if (pHandle->execHandle.subType == TOPIC_SUB_TYPE__COLUMN) {
-    if (tEncodeCStr(pEncoder, pHandle->execHandle.exec.execCol.qmsg) < 0) return -1;
+    if (tEncodeCStr(pEncoder, pHandle->execHandle.execCol.qmsg) < 0) return -1;
   }
   tEndEncode(pEncoder);
   return pEncoder->pos;
@@ -35,7 +35,7 @@ static int32_t tDecodeSTqHandle(SDecoder* pDecoder, STqHandle* pHandle) {
   if (tDecodeI32(pDecoder, &pHandle->epoch) < 0) return -1;
   if (tDecodeI8(pDecoder, &pHandle->execHandle.subType) < 0) return -1;
   if (pHandle->execHandle.subType == TOPIC_SUB_TYPE__COLUMN) {
-    if (tDecodeCStrAlloc(pDecoder, &pHandle->execHandle.exec.execCol.qmsg) < 0) return -1;
+    if (tDecodeCStrAlloc(pDecoder, &pHandle->execHandle.execCol.qmsg) < 0) return -1;
   }
   tEndDecode(pDecoder);
   return 0;
@@ -88,12 +88,11 @@ int32_t tqMetaOpen(STQ* pTq) {
             .meta = pTq->pVnode->pMeta,
             .pMsgCb = &pTq->pVnode->msgCb,
         };
-        handle.execHandle.exec.execCol.task[i] =
+        handle.execHandle.execCol.task[i] = qCreateStreamExecTaskInfo(handle.execHandle.execCol.qmsg, &reader);
-            qCreateStreamExecTaskInfo(handle.execHandle.exec.execCol.qmsg, &reader);
+        ASSERT(handle.execHandle.execCol.task[i]);
-        ASSERT(handle.execHandle.exec.execCol.task[i]);
       }
     } else {
-      handle.execHandle.exec.execDb.pFilterOutTbUid =
+      handle.execHandle.execDb.pFilterOutTbUid =
           taosHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_NO_LOCK);
     }
     taosHashPut(pTq->handles, pKey, kLen, &handle, sizeof(STqHandle));

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

@@ -29,13 +29,13 @@ static int32_t tqLoopExecFromQueue(STQ* pTq, STqHandle* pHandle, SStreamDataSubm
     }
     // update processed
     atomic_store_64(&pHandle->pushHandle.processedVer, pSubmit->ver);
-    streamQSetSuccess(&pHandle->pushHandle.inputQ);
+    streamQueueProcessSuccess(&pHandle->pushHandle.inputQ);
     streamDataSubmitRefDec(pSubmit);
     if (pRsp->blockNum > 0) {
       *ppSubmit = pSubmit;
       return 0;
     } else {
-      pSubmit = streamQNextItem(&pHandle->pushHandle.inputQ);
+      pSubmit = streamQueueNextItem(&pHandle->pushHandle.inputQ);
     }
   }
   *ppSubmit = pSubmit;
@@ -52,14 +52,14 @@ int32_t tqExecFromInputQ(STQ* pTq, STqHandle* pHandle) {
     // 2. check processedVer
     // 2.1. if not missed, get msg from queue
     // 2.2. if missed, scan wal
-    pSubmit = streamQNextItem(&pHandle->pushHandle.inputQ);
+    pSubmit = streamQueueNextItem(&pHandle->pushHandle.inputQ);
     while (pHandle->pushHandle.processedVer <= pSubmit->ver) {
       // read from wal
     }
     while (pHandle->pushHandle.processedVer > pSubmit->ver + 1) {
-      streamQSetSuccess(&pHandle->pushHandle.inputQ);
+      streamQueueProcessSuccess(&pHandle->pushHandle.inputQ);
       streamDataSubmitRefDec(pSubmit);
-      pSubmit = streamQNextItem(&pHandle->pushHandle.inputQ);
+      pSubmit = streamQueueNextItem(&pHandle->pushHandle.inputQ);
       if (pSubmit == NULL) break;
     }
     // 3. exec, after each success, update processed ver

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

@@ -307,7 +307,7 @@ int32_t tqUpdateTbUidList(STQ* pTq, const SArray* tbUidList, bool isAdd) {
     STqHandle* pExec = (STqHandle*)pIter;
     if (pExec->execHandle.subType == TOPIC_SUB_TYPE__COLUMN) {
       for (int32_t i = 0; i < 5; i++) {
-        int32_t code = qUpdateQualifiedTableId(pExec->execHandle.exec.execCol.task[i], tbUidList, isAdd);
+        int32_t code = qUpdateQualifiedTableId(pExec->execHandle.execCol.task[i], tbUidList, isAdd);
         ASSERT(code == 0);
       }
     } else if (pExec->execHandle.subType == TOPIC_SUB_TYPE__DB) {
@@ -315,7 +315,7 @@ int32_t tqUpdateTbUidList(STQ* pTq, const SArray* tbUidList, bool isAdd) {
         int32_t sz = taosArrayGetSize(tbUidList);
         for (int32_t i = 0; i < sz; i++) {
           int64_t tbUid = *(int64_t*)taosArrayGet(tbUidList, i);
-          taosHashPut(pExec->execHandle.exec.execDb.pFilterOutTbUid, &tbUid, sizeof(int64_t), NULL, 0);
+          taosHashPut(pExec->execHandle.execDb.pFilterOutTbUid, &tbUid, sizeof(int64_t), NULL, 0);
         }
       }
     } else {

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

@@ -24,6 +24,7 @@ static int32_t vnodeProcessDropTbReq(SVnode *pVnode, int64_t version, void *pReq
 static int32_t vnodeProcessSubmitReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
 static int32_t vnodeProcessCreateTSmaReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
 static int32_t vnodeProcessAlterConfirmReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
+static int32_t vnodeProcessWriteMsg(SVnode *pVnode, int64_t version, SRpcMsg *pMsg, SRpcMsg *pRsp);
 
 int32_t vnodePreprocessReq(SVnode *pVnode, SRpcMsg *pMsg) {
   int32_t  code = 0;
@@ -142,6 +143,9 @@ int32_t vnodeProcessWriteReq(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
     case TDMT_VND_SUBMIT:
       if (vnodeProcessSubmitReq(pVnode, version, pMsg->pCont, pMsg->contLen, pRsp) < 0) goto _err;
       break;
+    case TDMT_VND_DELETE:
+      if (vnodeProcessWriteMsg(pVnode, version, pMsg, pRsp) < 0) goto _err;
+      break;
     /* TQ */
     case TDMT_VND_MQ_VG_CHANGE:
       if (tqProcessVgChangeReq(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)),
@@ -256,6 +260,22 @@ int32_t vnodeProcessFetchMsg(SVnode *pVnode, SRpcMsg *pMsg, SQueueInfo *pInfo) {
   }
 }
 
+int32_t vnodeProcessWriteMsg(SVnode *pVnode, int64_t version, SRpcMsg *pMsg, SRpcMsg *pRsp) {
+  vTrace("message in write queue is processing");
+  char   *msgstr = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
+  int32_t msgLen = pMsg->contLen - sizeof(SMsgHead);
+  SDeleteRes res = {0};
+  SReadHandle handle = {.meta = pVnode->pMeta, .config = &pVnode->config, .vnode = pVnode, .pMsgCb = &pVnode->msgCb};
+  
+  switch (pMsg->msgType) {
+    case TDMT_VND_DELETE:
+      return qWorkerProcessDeleteMsg(&handle, pVnode->pQuery, pMsg, pRsp, &res);
+    default:
+      vError("unknown msg type:%d in write queue", pMsg->msgType);
+      return TSDB_CODE_VND_APP_ERROR;
+  }
+}
+
 // TODO: remove the function
 void smaHandleRes(void *pVnode, int64_t smaId, const SArray *data) {
   // TODO

source/libs/catalog/inc/catalogInt.h

@@ -518,6 +518,7 @@ int32_t ctgStbVersionSearchCompare(const void* key1, const void* key2);
 int32_t ctgDbVgVersionSearchCompare(const void* key1, const void* key2);
 void ctgFreeSTableMetaOutput(STableMetaOutput* pOutput);
 int32_t ctgUpdateMsgCtx(SCtgMsgCtx* pCtx, int32_t reqType, void* out, char* target);
+char *ctgTaskTypeStr(CTG_TASK_TYPE type);
 
 
 extern SCatalogMgmt gCtgMgmt;

source/libs/catalog/src/ctgAsync.c

@@ -44,7 +44,7 @@ int32_t ctgInitGetTbMetaTask(SCtgJob *pJob, int32_t taskIdx, SName *name) {
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:0x%" PRIx64 " task %d type %d initialized, tableName:%s", pJob->queryId, taskIdx, task.type, name->tname);
+  qDebug("QID:0x%" PRIx64 " the %d task type %s initialized, tableName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), name->tname);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -67,7 +67,7 @@ int32_t ctgInitGetDbVgTask(SCtgJob *pJob, int32_t taskIdx, char *dbFName) {
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:%" PRIx64 " task %d type %d initialized, dbFName:%s", pJob->queryId, taskIdx, task.type, dbFName);
+  qDebug("QID:0x%" PRIx64 " the %d task type %s initialized, dbFName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), dbFName);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -90,7 +90,7 @@ int32_t ctgInitGetDbCfgTask(SCtgJob *pJob, int32_t taskIdx, char *dbFName) {
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:%" PRIx64 " task %d type %d initialized, dbFName:%s", pJob->queryId, taskIdx, task.type, dbFName);
+  qDebug("QID:0x%" PRIx64 " the %d task type %s initialized, dbFName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), dbFName);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -113,7 +113,7 @@ int32_t ctgInitGetDbInfoTask(SCtgJob *pJob, int32_t taskIdx, char *dbFName) {
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:%" PRIx64 " task %d type %d initialized, dbFName:%s", pJob->queryId, taskIdx, task.type, dbFName);
+  qDebug("QID:0x%" PRIx64 " the %d task type %s initialized, dbFName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), dbFName);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -143,7 +143,7 @@ int32_t ctgInitGetTbHashTask(SCtgJob *pJob, int32_t taskIdx, SName *name) {
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:0x%" PRIx64 " task %d type %d initialized, tableName:%s", pJob->queryId, taskIdx, task.type, name->tname);
+  qDebug("QID:0x%" PRIx64 " the %d task type %s initialized, tableName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), name->tname);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -158,7 +158,7 @@ int32_t ctgInitGetQnodeTask(SCtgJob *pJob, int32_t taskIdx) {
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:%" PRIx64 " task %d type %d initialized", pJob->queryId, taskIdx, task.type);
+  qDebug("QID:%" PRIx64 " the %d task type %s initialized", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type));
 
   return TSDB_CODE_SUCCESS;
 }
@@ -181,7 +181,7 @@ int32_t ctgInitGetIndexTask(SCtgJob *pJob, int32_t taskIdx, char *name) {
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:%" PRIx64 " task %d type %d initialized, indexFName:%s", pJob->queryId, taskIdx, task.type, name);
+  qDebug("QID:%" PRIx64 " the %d task type %s initialized, indexFName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), name);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -204,7 +204,7 @@ int32_t ctgInitGetUdfTask(SCtgJob *pJob, int32_t taskIdx, char *name) {
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:%" PRIx64 " task %d type %d initialized, udfName:%s", pJob->queryId, taskIdx, task.type, name);
+  qDebug("QID:%" PRIx64 " the %d task type %s initialized, udfName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), name);
 
   return TSDB_CODE_SUCCESS;
 }
@@ -227,11 +227,96 @@ int32_t ctgInitGetUserTask(SCtgJob *pJob, int32_t taskIdx, SUserAuthInfo *user)
 
   taosArrayPush(pJob->pTasks, &task);
 
-  qDebug("QID:%" PRIx64 " task %d type %d initialized, user:%s", pJob->queryId, taskIdx, task.type, user->user);
+  qDebug("QID:%" PRIx64 " the %d task type %s initialized, user:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), user->user);
 
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t ctgHandleForceUpdate(SCatalog* pCtg, SCtgJob *pJob, const SCatalogReq* pReq) {
+  int32_t dbNum = pJob->dbCfgNum + pJob->dbVgNum + pJob->dbInfoNum;
+  if (dbNum > 0) {
+    if (dbNum > pJob->dbCfgNum && dbNum > pJob->dbVgNum && dbNum > pJob->dbInfoNum) {
+      SHashObj* pDb = taosHashInit(dbNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
+      if (NULL == pDb) {
+        CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+      }
+      
+      for (int32_t i = 0; i < pJob->dbVgNum; ++i) {
+        char* dbFName = taosArrayGet(pReq->pDbVgroup, i);
+        taosHashPut(pDb, dbFName, strlen(dbFName), dbFName, TSDB_DB_FNAME_LEN);
+      }
+
+      for (int32_t i = 0; i < pJob->dbCfgNum; ++i) {
+        char* dbFName = taosArrayGet(pReq->pDbCfg, i);
+        taosHashPut(pDb, dbFName, strlen(dbFName), dbFName, TSDB_DB_FNAME_LEN);
+      }
+
+      for (int32_t i = 0; i < pJob->dbInfoNum; ++i) {
+        char* dbFName = taosArrayGet(pReq->pDbInfo, i);
+        taosHashPut(pDb, dbFName, strlen(dbFName), dbFName, TSDB_DB_FNAME_LEN);
+      }
+
+      char* dbFName = taosHashIterate(pDb, NULL);
+      while (dbFName) {
+        ctgDropDbVgroupEnqueue(pCtg, dbFName, true);
+        dbFName = taosHashIterate(pDb, dbFName);
+      }
+
+      taosHashCleanup(pDb);      
+    } else {
+      for (int32_t i = 0; i < pJob->dbVgNum; ++i) {
+        char* dbFName = taosArrayGet(pReq->pDbVgroup, i);
+        CTG_ERR_RET(ctgDropDbVgroupEnqueue(pCtg, dbFName, true));
+      }
+      
+      for (int32_t i = 0; i < pJob->dbCfgNum; ++i) {
+        char* dbFName = taosArrayGet(pReq->pDbCfg, i);
+        CTG_ERR_RET(ctgDropDbVgroupEnqueue(pCtg, dbFName, true));
+      }
+      
+      for (int32_t i = 0; i < pJob->dbInfoNum; ++i) {
+        char* dbFName = taosArrayGet(pReq->pDbInfo, i);
+        CTG_ERR_RET(ctgDropDbVgroupEnqueue(pCtg, dbFName, true));
+      }
+    }
+  }
+
+  int32_t tbNum = pJob->tbMetaNum + pJob->tbHashNum;
+  if (tbNum > 0) {
+    if (tbNum > pJob->tbMetaNum && tbNum > pJob->tbHashNum) {
+      SHashObj* pTb = taosHashInit(tbNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK);
+      for (int32_t i = 0; i < pJob->tbMetaNum; ++i) {
+        SName* name = taosArrayGet(pReq->pTableMeta, i);
+        taosHashPut(pTb, name, sizeof(SName), name, sizeof(SName));
+      }
+      
+      for (int32_t i = 0; i < pJob->tbHashNum; ++i) {
+        SName* name = taosArrayGet(pReq->pTableHash, i);
+        taosHashPut(pTb, name, sizeof(SName), name, sizeof(SName));
+      }
+
+      SName* name = taosHashIterate(pTb, NULL);
+      while (name) {
+        catalogRemoveTableMeta(pCtg, name);
+        name = taosHashIterate(pTb, name);
+      }
+
+      taosHashCleanup(pTb);
+    } else {
+      for (int32_t i = 0; i < pJob->tbMetaNum; ++i) {
+        SName* name = taosArrayGet(pReq->pTableMeta, i);
+        catalogRemoveTableMeta(pCtg, name);
+      }
+      
+      for (int32_t i = 0; i < pJob->tbHashNum; ++i) {
+        SName* name = taosArrayGet(pReq->pTableHash, i);
+        catalogRemoveTableMeta(pCtg, name);
+      }
+    }
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
 
 int32_t ctgInitJob(CTG_PARAMS, SCtgJob** job, uint64_t reqId, const SCatalogReq* pReq, catalogCallback fp, void* param, int32_t* taskNum) {
   int32_t code = 0;
@@ -283,12 +368,13 @@ int32_t ctgInitJob(CTG_PARAMS, SCtgJob** job, uint64_t reqId, const SCatalogReq*
     CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY);
   }
 
+  if (pReq->forceUpdate) {
+    CTG_ERR_JRET(ctgHandleForceUpdate(pCtg, pJob, pReq));
+  }
+
   int32_t taskIdx = 0;
   for (int32_t i = 0; i < dbVgNum; ++i) {
     char* dbFName = taosArrayGet(pReq->pDbVgroup, i);
-    if (pReq->forceUpdate) {
-      ctgDropDbVgroupEnqueue(pCtg, dbFName, true);
-    }
     CTG_ERR_JRET(ctgInitGetDbVgTask(pJob, taskIdx++, dbFName));
   }
 
@@ -304,9 +390,6 @@ int32_t ctgInitJob(CTG_PARAMS, SCtgJob** job, uint64_t reqId, const SCatalogReq*
 
   for (int32_t i = 0; i < tbMetaNum; ++i) {
     SName* name = taosArrayGet(pReq->pTableMeta, i);
-    if (pReq->forceUpdate) {
-      catalogRemoveTableMeta(pCtg, name);
-    }
     CTG_ERR_JRET(ctgInitGetTbMetaTask(pJob, taskIdx++, name));
   }
 
@@ -342,7 +425,7 @@ int32_t ctgInitJob(CTG_PARAMS, SCtgJob** job, uint64_t reqId, const SCatalogReq*
 
   taosAcquireRef(gCtgMgmt.jobPool, pJob->refId);
 
-  qDebug("QID:%" PRIx64 ", job %" PRIx64 " initialized, task num %d", pJob->queryId, pJob->refId, *taskNum);
+  qDebug("QID:0x%" PRIx64 ", jobId: 0x%" PRIx64 " initialized, task num %d, forceUpdate %d", pJob->queryId, pJob->refId, *taskNum, pReq->forceUpdate);
   return TSDB_CODE_SUCCESS;
 
 
@@ -1108,7 +1191,7 @@ int32_t ctgLaunchJob(SCtgJob *pJob) {
   for (int32_t i = 0; i < taskNum; ++i) {
     SCtgTask *pTask = taosArrayGet(pJob->pTasks, i);
 
-    qDebug("QID:%" PRIx64 " start to launch task %d", pJob->queryId, pTask->taskId); 
+    qDebug("QID:0x%" PRIx64 " start to launch task %d", pJob->queryId, pTask->taskId);
     CTG_ERR_RET((*gCtgAsyncFps[pTask->type].launchFp)(pTask));
   }
 

source/libs/catalog/src/ctgRemote.c

@@ -177,7 +177,7 @@ int32_t ctgHandleMsgCallback(void *param, const SDataBuf *pMsg, int32_t rspCode)
 
   SCtgTask *pTask = taosArrayGet(pJob->pTasks, cbParam->taskId);
 
-  qDebug("QID:%" PRIx64 " task %d start to handle rsp %s", pJob->queryId, pTask->taskId, TMSG_INFO(cbParam->reqType + 1));
+  qDebug("QID:0x%" PRIx64 " task %d start to handle rsp %s", pJob->queryId, pTask->taskId, TMSG_INFO(cbParam->reqType + 1));
   
   CTG_ERR_JRET((*gCtgAsyncFps[pTask->type].handleRspFp)(pTask, cbParam->reqType, pMsg, rspCode));
  
@@ -244,7 +244,7 @@ int32_t ctgAsyncSendMsg(CTG_PARAMS, SCtgTask* pTask, int32_t msgType, void *msg,
     CTG_ERR_JRET(code);
   }
 
-  ctgDebug("req msg sent, reqId:%" PRIx64 ", msg type:%d, %s", pTask->pJob->queryId, msgType, TMSG_INFO(msgType));
+  ctgDebug("req msg sent, reqId:0x%" PRIx64 ", msg type:%d, %s", pTask->pJob->queryId, msgType, TMSG_INFO(msgType));
   return TSDB_CODE_SUCCESS;
 
 _return:
@@ -565,7 +565,7 @@ int32_t ctgGetTbMetaFromVnode(CTG_PARAMS, const SName* pTableName, SVgroupInfo *
     }
     CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, tbFName));
     
-    CTG_RET(ctgAsyncSendMsg(CTG_PARAMS_LIST(), pTask, reqType, msg, msgLen));
+    CTG_RET(ctgAsyncSendMsg(pCtg, pTrans, &vgroupInfo->epSet, pTask, reqType, msg, msgLen));
   }
 
   SRpcMsg rpcMsg = {

source/libs/catalog/src/ctgUtil.c

@@ -19,6 +19,31 @@
 #include "catalogInt.h"
 #include "systable.h"
 
+char *ctgTaskTypeStr(CTG_TASK_TYPE type) {
+  switch (type) {
+    case CTG_TASK_GET_QNODE:
+      return "[get qnode list]";
+    case CTG_TASK_GET_DB_VGROUP:
+      return "[get db vgroup]";
+    case CTG_TASK_GET_DB_CFG:
+      return "[get db cfg]";
+    case CTG_TASK_GET_DB_INFO:
+      return "[get db info]";
+    case CTG_TASK_GET_TB_META:
+      return "[get table meta]";
+    case CTG_TASK_GET_TB_HASH:
+      return "[get table hash]";
+    case CTG_TASK_GET_INDEX:
+      return "[get index]";
+    case CTG_TASK_GET_UDF:
+      return "[get udf]";
+    case CTG_TASK_GET_USER:
+      return "[get user]";
+    default:
+      return "unknown";
+  }
+}
+
 void ctgFreeSMetaData(SMetaData* pData) {
   taosArrayDestroy(pData->pTableMeta);
   pData->pTableMeta = NULL;
@@ -477,6 +502,9 @@ int32_t ctgGetVgInfoFromHashValue(SCatalog *pCtg, SDBVgInfo *dbInfo, const SName
 
   *pVgroup = *vgInfo;
 
+  ctgDebug("Got tb %s hash vgroup, vgId %d, epNum %d, current %s port %d", tbFullName, vgInfo->vgId, vgInfo->epSet.numOfEps,
+    vgInfo->epSet.eps[vgInfo->epSet.inUse].fqdn, vgInfo->epSet.eps[vgInfo->epSet.inUse].port);
+
   CTG_RET(code);
 }
 

source/libs/executor/inc/dataSinkInt.h

@@ -49,6 +49,7 @@ typedef struct SDataSinkHandle {
 } SDataSinkHandle;
 
 int32_t createDataDispatcher(SDataSinkManager* pManager, const SDataSinkNode* pDataSink, DataSinkHandle* pHandle);
+int32_t createDataDeleter(SDataSinkManager* pManager, const SDataSinkNode* pDataSink, DataSinkHandle* pHandle, void *pParam);
 
 #ifdef __cplusplus
 }

source/libs/executor/inc/executorimpl.h

@@ -524,6 +524,17 @@ typedef struct SProjectOperatorInfo {
   int64_t            curOutput;
 } SProjectOperatorInfo;
 
+typedef struct SIndefOperatorInfo {
+  SOptrBasicInfo     binfo;
+  SAggSupporter      aggSup;
+  SArray*            pPseudoColInfo;
+
+  SExprInfo*         pScalarExpr;
+  int32_t            numOfScalarExpr;
+  SqlFunctionCtx*    pScalarCtx;
+  int32_t*           rowCellInfoOffset;
+} SIndefOperatorInfo;
+
 typedef struct SFillOperatorInfo {
   struct SFillInfo* pFillInfo;
   SSDataBlock*      pRes;
@@ -770,10 +781,13 @@ SOperatorInfo* createSysTableScanOperatorInfo(void* readHandle, SSystemTableScan
 SOperatorInfo* createAggregateOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResultBlock, SExprInfo* pScalarExprInfo,
                                            int32_t numOfScalarExpr, SExecTaskInfo* pTaskInfo);
 
+SOperatorInfo* createIndefinitOutputOperatorInfo(SOperatorInfo* downstream, SPhysiNode *pNode, SExecTaskInfo* pTaskInfo);
 SOperatorInfo* createProjectOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t num, SSDataBlock* pResBlock, SLimit* pLimit, SLimit* pSlimit, SNode* pCondition, SExecTaskInfo* pTaskInfo);
 SOperatorInfo *createSortOperatorInfo(SOperatorInfo* downstream, SSDataBlock* pResBlock, SArray* pSortInfo, SExprInfo* pExprInfo, int32_t numOfCols,
                                       SArray* pIndexMap, SExecTaskInfo* pTaskInfo);
-
+SOperatorInfo* createMultiwaySortMergeOperatorInfo(SOperatorInfo** downStreams, int32_t numStreams,
+                                                   SSDataBlock* pResBlock, SArray* pSortInfo, SArray* pColMatchColInfo,
+                                                   SExecTaskInfo* pTaskInfo);
 SOperatorInfo* createSortedMergeOperatorInfo(SOperatorInfo** downstream, int32_t numOfDownstream, SExprInfo* pExprInfo, int32_t num, SArray* pSortInfo, SArray* pGroupInfo, SExecTaskInfo* pTaskInfo);
 
 SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
@@ -855,6 +869,7 @@ int32_t decodeOperator(SOperatorInfo* ops, char* data, int32_t length);
 void    setTaskStatus(SExecTaskInfo* pTaskInfo, int8_t status);
 int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId,
                                EOPTR_EXEC_MODEL model);
+int32_t createDataSinkParam(SDataSinkNode *pNode, void **pParam, qTaskInfo_t* pTaskInfo);                               
 int32_t getOperatorExplainExecInfo(SOperatorInfo* operatorInfo, SExplainExecInfo** pRes, int32_t* capacity,
                                    int32_t* resNum);
 

source/libs/executor/src/dataDeleter.c

@@ -0,0 +1,254 @@
+/*
+ * 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 "dataSinkInt.h"
+#include "dataSinkMgt.h"
+#include "executorimpl.h"
+#include "planner.h"
+#include "tcompression.h"
+#include "tdatablock.h"
+#include "tglobal.h"
+#include "tqueue.h"
+
+extern SDataSinkStat gDataSinkStat;
+
+typedef struct SDataDeleterBuf {
+  int32_t useSize;
+  int32_t allocSize;
+  char*   pData;
+} SDataDeleterBuf;
+
+typedef struct SDataCacheEntry {
+  int32_t dataLen;
+  int32_t numOfRows;
+  int32_t numOfCols;
+  int8_t  compressed;
+  char    data[];
+} SDataCacheEntry;
+
+typedef struct SDataDeleterHandle {
+  SDataSinkHandle     sink;
+  SDataSinkManager*   pManager;
+  SDataBlockDescNode* pSchema;
+  SDataDeleterNode*   pDeleter;
+  SDeleterParam*      pParam;
+  STaosQueue*         pDataBlocks;
+  SDataDeleterBuf     nextOutput;
+  int32_t             status;
+  bool                queryEnd;
+  uint64_t            useconds;
+  uint64_t            cachedSize;
+  TdThreadMutex       mutex;
+} SDataDeleterHandle;
+
+static bool needCompress(const SSDataBlock* pData, int32_t numOfCols) {
+  if (tsCompressColData < 0 || 0 == pData->info.rows) {
+    return false;
+  }
+
+  for (int32_t col = 0; col < numOfCols; ++col) {
+    SColumnInfoData* pColRes = taosArrayGet(pData->pDataBlock, col);
+    int32_t          colSize = pColRes->info.bytes * pData->info.rows;
+    if (NEEDTO_COMPRESS_QUERY(colSize)) {
+      return true;
+    }
+  }
+
+  return false;
+}
+
+static void toDataCacheEntry(SDataDeleterHandle* pHandle, const SInputData* pInput, SDataDeleterBuf* pBuf) {
+  int32_t numOfCols = LIST_LENGTH(pHandle->pSchema->pSlots);
+
+  SDataCacheEntry* pEntry = (SDataCacheEntry*)pBuf->pData;
+  pEntry->compressed = 0;
+  pEntry->numOfRows = pInput->pData->info.rows;
+  pEntry->numOfCols = pInput->pData->info.numOfCols;
+  pEntry->dataLen = sizeof(SDeleterRes);
+
+  ASSERT(1 == pEntry->numOfRows);
+  ASSERT(1 == pEntry->numOfCols);
+
+  pBuf->useSize = sizeof(SDataCacheEntry);
+
+  SColumnInfoData* pColRes = (SColumnInfoData*)taosArrayGet(pInput->pData->pDataBlock, 0);
+
+  SDeleterRes* pRes = (SDeleterRes*)pEntry->data;
+  pRes->uid = pHandle->pDeleter->tableId;
+  pRes->uidList = pHandle->pParam->pUidList;
+  pRes->skey = pHandle->pDeleter->deleteTimeRange.skey;
+  pRes->ekey = pHandle->pDeleter->deleteTimeRange.ekey;
+  pRes->affectedRows = *(int64_t*)pColRes->pData;
+
+  pBuf->useSize += pEntry->dataLen;
+  
+  atomic_add_fetch_64(&pHandle->cachedSize, pEntry->dataLen); 
+  atomic_add_fetch_64(&gDataSinkStat.cachedSize, pEntry->dataLen); 
+}
+
+static bool allocBuf(SDataDeleterHandle* pDeleter, const SInputData* pInput, SDataDeleterBuf* pBuf) {
+  uint32_t capacity = pDeleter->pManager->cfg.maxDataBlockNumPerQuery;
+  if (taosQueueItemSize(pDeleter->pDataBlocks) > capacity) {
+    qError("SinkNode queue is full, no capacity, max:%d, current:%d, no capacity", capacity,
+           taosQueueItemSize(pDeleter->pDataBlocks));
+    return false;
+  }
+
+  pBuf->allocSize = sizeof(SDataCacheEntry) + sizeof(SDeleterRes);
+
+  pBuf->pData = taosMemoryMalloc(pBuf->allocSize);
+  if (pBuf->pData == NULL) {
+    qError("SinkNode failed to malloc memory, size:%d, code:%d", pBuf->allocSize, TAOS_SYSTEM_ERROR(errno));
+  }
+
+  return NULL != pBuf->pData;
+}
+
+static int32_t updateStatus(SDataDeleterHandle* pDeleter) {
+  taosThreadMutexLock(&pDeleter->mutex);
+  int32_t blockNums = taosQueueItemSize(pDeleter->pDataBlocks);
+  int32_t status =
+      (0 == blockNums ? DS_BUF_EMPTY
+                      : (blockNums < pDeleter->pManager->cfg.maxDataBlockNumPerQuery ? DS_BUF_LOW : DS_BUF_FULL));
+  pDeleter->status = status;
+  taosThreadMutexUnlock(&pDeleter->mutex);
+  return status;
+}
+
+static int32_t getStatus(SDataDeleterHandle* pDeleter) {
+  taosThreadMutexLock(&pDeleter->mutex);
+  int32_t status = pDeleter->status;
+  taosThreadMutexUnlock(&pDeleter->mutex);
+  return status;
+}
+
+static int32_t putDataBlock(SDataSinkHandle* pHandle, const SInputData* pInput, bool* pContinue) {
+  SDataDeleterHandle* pDeleter = (SDataDeleterHandle*)pHandle;
+  SDataDeleterBuf*    pBuf = taosAllocateQitem(sizeof(SDataDeleterBuf), DEF_QITEM);
+  if (NULL == pBuf || !allocBuf(pDeleter, pInput, pBuf)) {
+    return TSDB_CODE_QRY_OUT_OF_MEMORY;
+  }
+  toDataCacheEntry(pDeleter, pInput, pBuf);
+  taosWriteQitem(pDeleter->pDataBlocks, pBuf);
+  *pContinue = (DS_BUF_LOW == updateStatus(pDeleter) ? true : false);
+  return TSDB_CODE_SUCCESS;
+}
+
+static void endPut(struct SDataSinkHandle* pHandle, uint64_t useconds) {
+  SDataDeleterHandle* pDeleter = (SDataDeleterHandle*)pHandle;
+  taosThreadMutexLock(&pDeleter->mutex);
+  pDeleter->queryEnd = true;
+  pDeleter->useconds = useconds;
+  taosThreadMutexUnlock(&pDeleter->mutex);
+}
+
+static void getDataLength(SDataSinkHandle* pHandle, int32_t* pLen, bool* pQueryEnd) {
+  SDataDeleterHandle* pDeleter = (SDataDeleterHandle*)pHandle;
+  if (taosQueueEmpty(pDeleter->pDataBlocks)) {
+    *pQueryEnd = pDeleter->queryEnd;
+    *pLen = 0;
+    return;
+  }
+
+  SDataDeleterBuf* pBuf = NULL;
+  taosReadQitem(pDeleter->pDataBlocks, (void**)&pBuf);
+  memcpy(&pDeleter->nextOutput, pBuf, sizeof(SDataDeleterBuf));
+  taosFreeQitem(pBuf);
+  *pLen = ((SDataCacheEntry*)(pDeleter->nextOutput.pData))->dataLen;
+  *pQueryEnd = pDeleter->queryEnd;
+  qDebug("got data len %d, row num %d in sink", *pLen, ((SDataCacheEntry*)(pDeleter->nextOutput.pData))->numOfRows);
+}
+
+static int32_t getDataBlock(SDataSinkHandle* pHandle, SOutputData* pOutput) {
+  SDataDeleterHandle* pDeleter = (SDataDeleterHandle*)pHandle;
+  if (NULL == pDeleter->nextOutput.pData) {
+    assert(pDeleter->queryEnd);
+    pOutput->useconds = pDeleter->useconds;
+    pOutput->precision = pDeleter->pSchema->precision;
+    pOutput->bufStatus = DS_BUF_EMPTY;
+    pOutput->queryEnd = pDeleter->queryEnd;
+    return TSDB_CODE_SUCCESS;
+  }
+  SDataCacheEntry* pEntry = (SDataCacheEntry*)(pDeleter->nextOutput.pData);
+  memcpy(pOutput->pData, pEntry->data, pEntry->dataLen);
+  pOutput->numOfRows = pEntry->numOfRows;
+  pOutput->numOfCols = pEntry->numOfCols;
+  pOutput->compressed = pEntry->compressed;
+
+  atomic_sub_fetch_64(&pDeleter->cachedSize, pEntry->dataLen);  
+  atomic_sub_fetch_64(&gDataSinkStat.cachedSize, pEntry->dataLen); 
+
+  taosMemoryFreeClear(pDeleter->nextOutput.pData);  // todo persistent
+  pOutput->bufStatus = updateStatus(pDeleter);
+  taosThreadMutexLock(&pDeleter->mutex);
+  pOutput->queryEnd = pDeleter->queryEnd;
+  pOutput->useconds = pDeleter->useconds;
+  pOutput->precision = pDeleter->pSchema->precision;
+  taosThreadMutexUnlock(&pDeleter->mutex);
+  
+  return TSDB_CODE_SUCCESS;
+}
+
+static int32_t destroyDataSinker(SDataSinkHandle* pHandle) {
+  SDataDeleterHandle* pDeleter = (SDataDeleterHandle*)pHandle;
+  atomic_sub_fetch_64(&gDataSinkStat.cachedSize, pDeleter->cachedSize);
+  taosMemoryFreeClear(pDeleter->nextOutput.pData);
+  while (!taosQueueEmpty(pDeleter->pDataBlocks)) {
+    SDataDeleterBuf* pBuf = NULL;
+    taosReadQitem(pDeleter->pDataBlocks, (void**)&pBuf);
+    taosMemoryFreeClear(pBuf->pData);
+    taosFreeQitem(pBuf);
+  }
+  taosCloseQueue(pDeleter->pDataBlocks);
+  taosThreadMutexDestroy(&pDeleter->mutex);
+  return TSDB_CODE_SUCCESS;
+}
+
+static int32_t getCacheSize(struct SDataSinkHandle* pHandle, uint64_t* size) {
+  SDataDeleterHandle* pDispatcher = (SDataDeleterHandle*)pHandle;
+
+  *size = atomic_load_64(&pDispatcher->cachedSize);
+  return TSDB_CODE_SUCCESS;
+}
+
+int32_t createDataDeleter(SDataSinkManager* pManager, const SDataSinkNode* pDataSink, DataSinkHandle* pHandle, void *pParam) {
+  SDataDeleterHandle* deleter = taosMemoryCalloc(1, sizeof(SDataDeleterHandle));
+  if (NULL == deleter) {
+    terrno = TSDB_CODE_QRY_OUT_OF_MEMORY;
+    return TSDB_CODE_QRY_OUT_OF_MEMORY;
+  }
+
+  SDataDeleterNode* pDeleterNode = (SDataDeleterNode *)pDataSink;
+  deleter->sink.fPut = putDataBlock;
+  deleter->sink.fEndPut = endPut;
+  deleter->sink.fGetLen = getDataLength;
+  deleter->sink.fGetData = getDataBlock;
+  deleter->sink.fDestroy = destroyDataSinker;
+  deleter->sink.fGetCacheSize = getCacheSize;
+  deleter->pManager = pManager;
+  deleter->pDeleter = pDeleterNode;
+  deleter->pSchema = pDataSink->pInputDataBlockDesc;
+  deleter->pParam = pParam;
+  deleter->status = DS_BUF_EMPTY;
+  deleter->queryEnd = false;
+  deleter->pDataBlocks = taosOpenQueue();
+  taosThreadMutexInit(&deleter->mutex, NULL);
+  if (NULL == deleter->pDataBlocks) {
+    terrno = TSDB_CODE_QRY_OUT_OF_MEMORY;
+    return TSDB_CODE_QRY_OUT_OF_MEMORY;
+  }
+  *pHandle = deleter;
+  return TSDB_CODE_SUCCESS;
+}

source/libs/executor/src/dataDispatcher.c

@@ -83,7 +83,7 @@ static void toDataCacheEntry(SDataDispatchHandle* pHandle, const SInputData* pIn
   pEntry->numOfCols = pInput->pData->info.numOfCols;
   pEntry->dataLen = 0;
 
-  pBuf->useSize = sizeof(SRetrieveTableRsp);
+  pBuf->useSize = sizeof(SDataCacheEntry);
   blockCompressEncode(pInput->pData, pEntry->data, &pEntry->dataLen, numOfCols, pEntry->compressed);
 
   pBuf->useSize += pEntry->dataLen;
@@ -100,7 +100,7 @@ static bool allocBuf(SDataDispatchHandle* pDispatcher, const SInputData* pInput,
     return false;
   }
 
-  pBuf->allocSize = sizeof(SRetrieveTableRsp) + blockGetEncodeSize(pInput->pData);
+  pBuf->allocSize = sizeof(SDataCacheEntry) + blockGetEncodeSize(pInput->pData);
 
   pBuf->pData = taosMemoryMalloc(pBuf->allocSize);
   if (pBuf->pData == NULL) {
@@ -211,7 +211,7 @@ static int32_t destroyDataSinker(SDataSinkHandle* pHandle) {
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t getCacheSize(struct SDataSinkHandle* pHandle, uint64_t* size) {
+static int32_t getCacheSize(struct SDataSinkHandle* pHandle, uint64_t* size) {
   SDataDispatchHandle* pDispatcher = (SDataDispatchHandle*)pHandle;
 
   *size = atomic_load_64(&pDispatcher->cachedSize);

source/libs/executor/src/dataSinkMgt.c

@@ -34,9 +34,12 @@ int32_t dsDataSinkGetCacheSize(SDataSinkStat *pStat) {
 }
 
 
-int32_t dsCreateDataSinker(const SDataSinkNode *pDataSink, DataSinkHandle* pHandle) {
+int32_t dsCreateDataSinker(const SDataSinkNode *pDataSink, DataSinkHandle* pHandle, void* pParam) {
-  if (QUERY_NODE_PHYSICAL_PLAN_DISPATCH == nodeType(pDataSink)) {
+  switch (nodeType(pDataSink)) {
+    case QUERY_NODE_PHYSICAL_PLAN_DISPATCH:
       return createDataDispatcher(&gDataSinkManager, pDataSink, pHandle);
+    case QUERY_NODE_PHYSICAL_PLAN_DELETE:
+      return createDataDeleter(&gDataSinkManager, pDataSink, pHandle, pParam);
   }
   return TSDB_CODE_FAILED;
 }

source/libs/executor/src/executorMain.c

@@ -45,8 +45,15 @@ int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId,
   if (code != TSDB_CODE_SUCCESS) {
     goto _error;
   }
+  
   if (handle) {
-    code = dsCreateDataSinker(pSubplan->pDataSink, handle);
+    void* pSinkParam = NULL;
+    code = createDataSinkParam(pSubplan->pDataSink, &pSinkParam, pTaskInfo);
+    if (code != TSDB_CODE_SUCCESS) {
+      goto _error;
+    }
+    
+    code = dsCreateDataSinker(pSubplan->pDataSink, handle, pSinkParam);
   }
 
   _error:

source/libs/executor/src/executorimpl.c

@@ -108,8 +108,6 @@ static SColumnInfo* extractColumnFilterInfo(SExprInfo* pExpr, int32_t numOfOutpu
 
 static void releaseQueryBuf(size_t numOfTables);
 
-static int32_t getNumOfScanTimes(STaskAttr* pQueryAttr);
-
 static void destroySFillOperatorInfo(void* param, int32_t numOfOutput);
 static void destroyProjectOperatorInfo(void* param, int32_t numOfOutput);
 static void destroyTagScanOperatorInfo(void* param, int32_t numOfOutput);
@@ -2856,7 +2854,7 @@ static SSDataBlock* doLoadRemoteData(SOperatorInfo* pOperator) {
     return seqLoadRemoteData(pOperator);
   } else {
     return concurrentlyLoadRemoteDataImpl(pOperator, pExchangeInfo, pTaskInfo);
-//    return concurrentlyLoadRemoteData(pOperator);
+    //    return concurrentlyLoadRemoteData(pOperator);
   }
 }
 
@@ -3636,18 +3634,6 @@ static SSDataBlock* doProjectOperation(SOperatorInfo* pOperator) {
       break;
     }
 
-#if 0
-    // Return result of the previous group in the firstly.
-    if (false) {
-      if (pRes->info.rows > 0) {
-        pProjectInfo->existDataBlock = pBlock;
-        break;
-      } else {  // init output buffer for a new group data
-        initCtxOutputBuffer(pInfo->pCtx, pOperator->numOfExprs);
-      }
-    }
-#endif
-
     // the pDataBlock are always the same one, no need to call this again
     int32_t code = getTableScanInfo(pOperator->pDownstream[0], &order, &scanFlag);
     if (code != TSDB_CODE_SUCCESS) {
@@ -3793,6 +3779,17 @@ static SSDataBlock* doFill(SOperatorInfo* pOperator) {
   }
 }
 
+static void destroyExprInfo(SExprInfo* pExpr, int32_t numOfExprs) {
+  for (int32_t i = 0; i < numOfExprs; ++i) {
+    SExprInfo* pExprInfo = &pExpr[i];
+    if (pExprInfo->pExpr->nodeType == QUERY_NODE_COLUMN) {
+      taosMemoryFree(pExprInfo->base.pParam[0].pCol);
+    }
+    taosMemoryFree(pExprInfo->base.pParam);
+    taosMemoryFree(pExprInfo->pExpr);
+  }
+}
+
 static void destroyOperatorInfo(SOperatorInfo* pOperator) {
   if (pOperator == NULL) {
     return;
@@ -3812,14 +3809,7 @@ static void destroyOperatorInfo(SOperatorInfo* pOperator) {
   }
 
   if (pOperator->pExpr != NULL) {
-    for (int32_t i = 0; i < pOperator->numOfExprs; ++i) {
+    destroyExprInfo(pOperator->pExpr, pOperator->numOfExprs);
-      SExprInfo* pExprInfo = &pOperator->pExpr[i];
-      if (pExprInfo->pExpr->nodeType == QUERY_NODE_COLUMN) {
-        taosMemoryFree(pExprInfo->base.pParam[0].pCol);
-      }
-      taosMemoryFree(pExprInfo->base.pParam);
-      taosMemoryFree(pExprInfo->pExpr);
-    }
   }
 
   taosMemoryFreeClear(pOperator->pExpr);
@@ -4008,6 +3998,19 @@ static void destroyProjectOperatorInfo(void* param, int32_t numOfOutput) {
   taosArrayDestroy(pInfo->pPseudoColInfo);
 }
 
+static void destroyIndefinitOperatorInfo(void* param, int32_t numOfOutput) {
+  SIndefOperatorInfo* pInfo = (SIndefOperatorInfo*)param;
+  doDestroyBasicInfo(&pInfo->binfo, numOfOutput);
+
+  taosArrayDestroy(pInfo->pPseudoColInfo);
+  cleanupAggSup(&pInfo->aggSup);
+
+  destroySqlFunctionCtx(pInfo->pScalarCtx, numOfOutput);
+  destroyExprInfo(pInfo->pScalarExpr, pInfo->numOfScalarExpr);
+
+  taosMemoryFree(pInfo->rowCellInfoOffset);
+}
+
 void destroyExchangeOperatorInfo(void* param, int32_t numOfOutput) {
   SExchangeInfo* pExInfo = (SExchangeInfo*)param;
   taosArrayDestroy(pExInfo->pSources);
@@ -4085,6 +4088,136 @@ _error:
   return NULL;
 }
 
+static SSDataBlock* doApplyIndefinitFunction(SOperatorInfo* pOperator) {
+  SIndefOperatorInfo* pIndefInfo = pOperator->info;
+  SOptrBasicInfo*     pInfo = &pIndefInfo->binfo;
+
+  SSDataBlock* pRes = pInfo->pRes;
+  blockDataCleanup(pRes);
+
+  SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
+  if (pOperator->status == OP_EXEC_DONE) {
+    return NULL;
+  }
+
+  int64_t st = 0;
+  int32_t order = 0;
+  int32_t scanFlag = 0;
+
+  if (pOperator->cost.openCost == 0) {
+    st = taosGetTimestampUs();
+  }
+
+  SOperatorInfo* downstream = pOperator->pDownstream[0];
+
+  while (1) {
+    // The downstream exec may change the value of the newgroup, so use a local variable instead.
+    SSDataBlock* pBlock = downstream->fpSet.getNextFn(downstream);
+    if (pBlock == NULL) {
+      doSetOperatorCompleted(pOperator);
+      break;
+    }
+
+    // the pDataBlock are always the same one, no need to call this again
+    int32_t code = getTableScanInfo(pOperator->pDownstream[0], &order, &scanFlag);
+    if (code != TSDB_CODE_SUCCESS) {
+      longjmp(pTaskInfo->env, code);
+    }
+
+    // there is an scalar expression that needs to be calculated before apply the group aggregation.
+    if (pIndefInfo->pScalarExpr != NULL) {
+      code = projectApplyFunctions(pIndefInfo->pScalarExpr, pBlock, pBlock, pIndefInfo->pScalarCtx,
+                                   pIndefInfo->numOfScalarExpr, pIndefInfo->pPseudoColInfo);
+      if (code != TSDB_CODE_SUCCESS) {
+        longjmp(pTaskInfo->env, code);
+      }
+    }
+
+    setInputDataBlock(pOperator, pInfo->pCtx, pBlock, order, scanFlag, false);
+    blockDataEnsureCapacity(pInfo->pRes, pInfo->pRes->info.rows + pBlock->info.rows);
+
+    code = projectApplyFunctions(pOperator->pExpr, pInfo->pRes, pBlock, pInfo->pCtx, pOperator->numOfExprs,
+                                 pIndefInfo->pPseudoColInfo);
+    if (code != TSDB_CODE_SUCCESS) {
+      longjmp(pTaskInfo->env, code);
+    }
+  }
+
+  size_t rows = pInfo->pRes->info.rows;
+  pOperator->resultInfo.totalRows += rows;
+
+  if (pOperator->cost.openCost == 0) {
+    pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
+  }
+
+  return (rows > 0) ? pInfo->pRes : NULL;
+}
+
+SOperatorInfo* createIndefinitOutputOperatorInfo(SOperatorInfo* downstream, SPhysiNode* pNode,
+                                                 SExecTaskInfo* pTaskInfo) {
+  SIndefOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SIndefOperatorInfo));
+  SOperatorInfo*      pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
+  if (pInfo == NULL || pOperator == NULL) {
+    goto _error;
+  }
+
+  SIndefRowsFuncPhysiNode* pPhyNode = (SIndefRowsFuncPhysiNode*)pNode;
+
+  int32_t    numOfExpr = 0;
+  SExprInfo* pExprInfo = createExprInfo(pPhyNode->pVectorFuncs, NULL, &numOfExpr);
+
+  int32_t numOfScalarExpr = 0;
+  if (pPhyNode->pExprs != NULL) {
+    pInfo->pScalarExpr = createExprInfo(pPhyNode->pExprs, NULL, &numOfScalarExpr);
+    pInfo->pScalarCtx = createSqlFunctionCtx(pInfo->pScalarExpr, numOfScalarExpr, &pInfo->rowCellInfoOffset);
+  }
+
+  SSDataBlock* pResBlock = createResDataBlock(pPhyNode->node.pOutputDataBlockDesc);
+  ;
+
+  int32_t numOfRows = 4096;
+  size_t  keyBufSize = sizeof(int64_t) + sizeof(int64_t) + POINTER_BYTES;
+
+  // Make sure the size of SSDataBlock will never exceed the size of 2MB.
+  int32_t TWOMB = 2 * 1024 * 1024;
+  if (numOfRows * pResBlock->info.rowSize > TWOMB) {
+    numOfRows = TWOMB / pResBlock->info.rowSize;
+  }
+  initResultSizeInfo(pOperator, numOfRows);
+
+  initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfExpr, pResBlock, keyBufSize, pTaskInfo->id.str);
+  setFunctionResultOutput(&pInfo->binfo, &pInfo->aggSup, MAIN_SCAN, numOfExpr, pTaskInfo);
+
+  pInfo->binfo.pRes = pResBlock;
+  pInfo->numOfScalarExpr = numOfScalarExpr;
+  pInfo->pPseudoColInfo = setRowTsColumnOutputInfo(pInfo->binfo.pCtx, numOfExpr);
+
+  pOperator->name = "IndefinitOperator";
+  pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_PROJECT;
+  pOperator->blocking = false;
+  pOperator->status = OP_NOT_OPENED;
+  pOperator->info = pInfo;
+  pOperator->pExpr = pExprInfo;
+  pOperator->numOfExprs = numOfExpr;
+  pOperator->pTaskInfo = pTaskInfo;
+
+  pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doApplyIndefinitFunction, NULL, NULL,
+                                         destroyIndefinitOperatorInfo, NULL, NULL, NULL);
+
+  int32_t code = appendDownstream(pOperator, &downstream, 1);
+  if (code != TSDB_CODE_SUCCESS) {
+    goto _error;
+  }
+
+  return pOperator;
+
+_error:
+  taosMemoryFree(pInfo);
+  taosMemoryFree(pOperator);
+  pTaskInfo->code = TSDB_CODE_OUT_OF_MEMORY;
+  return NULL;
+}
+
 static int32_t initFillInfo(SFillOperatorInfo* pInfo, SExprInfo* pExpr, int32_t numOfCols, SNodeListNode* pValNode,
                             STimeWindow win, int32_t capacity, const char* id, SInterval* pInterval, int32_t fillType) {
   SFillColInfo* pColInfo = createFillColInfo(pExpr, numOfCols, pValNode);
@@ -4517,6 +4650,18 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
         extractColMatchInfo(pSortPhyNode->pTargets, pDescNode, &numOfOutputCols, pTaskInfo, COL_MATCH_FROM_SLOT_ID);
 
     pOptr = createSortOperatorInfo(ops[0], pResBlock, info, pExprInfo, numOfCols, pColList, pTaskInfo);
+  } else if (QUERY_NODE_PHYSICAL_PLAN_MERGE == type) {
+    SMergePhysiNode* pMergePhyNode = (SMergePhysiNode*)pPhyNode;
+
+    SDataBlockDescNode* pDescNode = pPhyNode->pOutputDataBlockDesc;
+    SSDataBlock*        pResBlock = createResDataBlock(pDescNode);
+
+    SArray* sortInfo = createSortInfo(pMergePhyNode->pMergeKeys);
+    int32_t numOfOutputCols = 0;
+    SArray* pColList =
+        extractColMatchInfo(pMergePhyNode->pTargets, pDescNode, &numOfOutputCols, pTaskInfo, COL_MATCH_FROM_SLOT_ID);
+
+    pOptr = createMultiwaySortMergeOperatorInfo(ops, size, pResBlock, sortInfo, pColList, pTaskInfo);
   } else if (QUERY_NODE_PHYSICAL_PLAN_SESSION_WINDOW == type) {
     SSessionWinodwPhysiNode* pSessionNode = (SSessionWinodwPhysiNode*)pPhyNode;
 
@@ -4577,6 +4722,8 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
     SInterval* pInterval = &((SIntervalAggOperatorInfo*)ops[0]->info)->interval;
     pOptr = createFillOperatorInfo(ops[0], pExprInfo, num, pInterval, &pFillNode->timeRange, pResBlock, pFillNode->mode,
                                    (SNodeListNode*)pFillNode->pValues, false, pTaskInfo);
+  } else if (QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC == type) {
+    pOptr = createIndefinitOutputOperatorInfo(ops[0], pPhyNode, pTaskInfo);
   } else {
     ASSERT(0);
   }
@@ -4952,6 +5099,37 @@ int32_t decodeOperator(SOperatorInfo* ops, char* result, int32_t length) {
   return TDB_CODE_SUCCESS;
 }
 
+int32_t createDataSinkParam(SDataSinkNode *pNode, void **pParam, qTaskInfo_t* pTaskInfo) {
+  SExecTaskInfo* pTask = *(SExecTaskInfo**)pTaskInfo;
+  
+  switch (pNode->type) {
+    case QUERY_NODE_PHYSICAL_PLAN_DELETE: {
+      SDeleterParam *pDeleterParam = taosMemoryCalloc(1, sizeof(SDeleterParam));
+      if (NULL == pDeleterParam) {
+        return TSDB_CODE_OUT_OF_MEMORY;
+      }
+      int32_t tbNum = taosArrayGetSize(pTask->tableqinfoList.pTableList);
+      pDeleterParam->pUidList = taosArrayInit(tbNum, sizeof(uint64_t));
+      if (NULL == pDeleterParam->pUidList) {
+        taosMemoryFree(pDeleterParam);
+        return TSDB_CODE_OUT_OF_MEMORY;
+      }
+      for (int32_t i = 0; i < tbNum; ++i) {
+        STableKeyInfo *pTable = taosArrayGet(pTask->tableqinfoList.pTableList, i);
+        taosArrayPush(pDeleterParam->pUidList, &pTable->uid);
+      }
+
+      *pParam = pDeleterParam;
+      break;
+    }
+    default:
+      break;
+  }
+
+  return TSDB_CODE_SUCCESS;
+}
+
+
 int32_t createExecTaskInfoImpl(SSubplan* pPlan, SExecTaskInfo** pTaskInfo, SReadHandle* pHandle, uint64_t taskId,
                                EOPTR_EXEC_MODEL model) {
   uint64_t queryId = pPlan->id.queryId;

source/libs/executor/src/sortoperator.c

@@ -13,8 +13,8 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include "tdatablock.h"
 #include "executorimpl.h"
+#include "tdatablock.h"
 
 static SSDataBlock* doSort(SOperatorInfo* pOperator);
 static int32_t      doOpenSortOperator(SOperatorInfo* pOperator);
@@ -22,8 +22,9 @@ static int32_t getExplainExecInfo(SOperatorInfo* pOptr, void** pOptrExplain, uin
 
 static void destroyOrderOperatorInfo(void* param, int32_t numOfOutput);
 
-SOperatorInfo* createSortOperatorInfo(SOperatorInfo* downstream, SSDataBlock* pResBlock, SArray* pSortInfo, SExprInfo* pExprInfo, int32_t numOfCols,
+SOperatorInfo* createSortOperatorInfo(SOperatorInfo* downstream, SSDataBlock* pResBlock, SArray* pSortInfo,
-                                      SArray* pColMatchColInfo, SExecTaskInfo* pTaskInfo) {
+                                      SExprInfo* pExprInfo, int32_t numOfCols, SArray* pColMatchColInfo,
+                                      SExecTaskInfo* pTaskInfo) {
   SSortOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SSortOperatorInfo));
   SOperatorInfo*     pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
   int32_t            rowSize = pResBlock->info.rowSize;
@@ -40,7 +41,7 @@ SOperatorInfo* createSortOperatorInfo(SOperatorInfo* downstream, SSDataBlock* pR
   initResultSizeInfo(pOperator, 1024);
 
   pInfo->pSortInfo = pSortInfo;
-  pInfo->pColMatchInfo= pColMatchColInfo;
+  pInfo->pColMatchInfo = pColMatchColInfo;
   pOperator->name = "SortOperator";
   pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_SORT;
   pOperator->blocking = true;
@@ -48,17 +49,17 @@ SOperatorInfo* createSortOperatorInfo(SOperatorInfo* downstream, SSDataBlock* pR
   pOperator->info = pInfo;
 
   // lazy evaluation for the following parameter since the input datablock is not known till now.
-//  pInfo->bufPageSize  = rowSize < 1024 ? 1024 * 2 : rowSize * 2;  // there are headers, so pageSize = rowSize + header
+  //  pInfo->bufPageSize  = rowSize < 1024 ? 1024 * 2 : rowSize * 2;  // there are headers, so pageSize = rowSize +
-//  pInfo->sortBufSize  = pInfo->bufPageSize * 16;  // TODO dynamic set the available sort buffer
+  //  header pInfo->sortBufSize  = pInfo->bufPageSize * 16;  // TODO dynamic set the available sort buffer
 
   pOperator->pTaskInfo = pTaskInfo;
-  pOperator->fpSet =
+  pOperator->fpSet = createOperatorFpSet(doOpenSortOperator, doSort, NULL, NULL, destroyOrderOperatorInfo, NULL, NULL,
-      createOperatorFpSet(doOpenSortOperator, doSort, NULL, NULL, destroyOrderOperatorInfo, NULL, NULL, getExplainExecInfo);
+                                         getExplainExecInfo);
 
   int32_t code = appendDownstream(pOperator, &downstream, 1);
   return pOperator;
 
-  _error:
+_error:
   pTaskInfo->code = TSDB_CODE_OUT_OF_MEMORY;
   taosMemoryFree(pInfo);
   taosMemoryFree(pOperator);
@@ -80,7 +81,8 @@ void appendOneRowToDataBlock(SSDataBlock* pBlock, STupleHandle* pTupleHandle) {
   pBlock->info.rows += 1;
 }
 
-SSDataBlock* getSortedBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, int32_t capacity, SArray* pColMatchInfo) {
+SSDataBlock* getSortedBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, int32_t capacity,
+                                SArray* pColMatchInfo) {
   blockDataCleanup(pDataBlock);
   ASSERT(taosArrayGetSize(pColMatchInfo) == pDataBlock->info.numOfCols);
 
@@ -132,7 +134,8 @@ void applyScalarFunction(SSDataBlock* pBlock, void* param) {
   SOperatorInfo*     pOperator = param;
   SSortOperatorInfo* pSort = pOperator->info;
   if (pOperator->pExpr != NULL) {
-    int32_t code = projectApplyFunctions(pOperator->pExpr, pBlock, pBlock, pSort->binfo.pCtx, pOperator->numOfExprs, NULL);
+    int32_t code =
+        projectApplyFunctions(pOperator->pExpr, pBlock, pBlock, pSort->binfo.pCtx, pOperator->numOfExprs, NULL);
     if (code != TSDB_CODE_SUCCESS) {
       longjmp(pOperator->pTaskInfo->env, code);
     }
@@ -150,8 +153,8 @@ int32_t doOpenSortOperator(SOperatorInfo* pOperator) {
   pInfo->startTs = taosGetTimestampUs();
 
   //  pInfo->binfo.pRes is not equalled to the input datablock.
-  pInfo->pSortHandle = tsortCreateSortHandle(pInfo->pSortInfo, pInfo->pColMatchInfo, SORT_SINGLESOURCE_SORT,
+  pInfo->pSortHandle = tsortCreateSortHandle(pInfo->pSortInfo, pInfo->pColMatchInfo, SORT_SINGLESOURCE_SORT, -1, -1,
-                                             -1, -1, NULL, pTaskInfo->id.str);
+                                             NULL, pTaskInfo->id.str);
 
   tsortSetFetchRawDataFp(pInfo->pSortHandle, loadNextDataBlock, applyScalarFunction, pOperator);
 
@@ -166,7 +169,7 @@ int32_t doOpenSortOperator(SOperatorInfo* pOperator) {
     longjmp(pTaskInfo->env, terrno);
   }
 
-  pOperator->cost.openCost = (taosGetTimestampUs() - pInfo->startTs)/1000.0;
+  pOperator->cost.openCost = (taosGetTimestampUs() - pInfo->startTs) / 1000.0;
   pOperator->status = OP_RES_TO_RETURN;
 
   OPTR_SET_OPENED(pOperator);
@@ -186,7 +189,8 @@ SSDataBlock* doSort(SOperatorInfo* pOperator) {
     longjmp(pTaskInfo->env, code);
   }
 
-  SSDataBlock* pBlock = getSortedBlockData(pInfo->pSortHandle, pInfo->binfo.pRes, pOperator->resultInfo.capacity, pInfo->pColMatchInfo);
+  SSDataBlock* pBlock =
+      getSortedBlockData(pInfo->pSortHandle, pInfo->binfo.pRes, pOperator->resultInfo.capacity, pInfo->pColMatchInfo);
 
   if (pBlock != NULL) {
     pOperator->resultInfo.totalRows += pBlock->info.rows;
@@ -208,10 +212,147 @@ int32_t getExplainExecInfo(SOperatorInfo* pOptr, void** pOptrExplain, uint32_t*
   ASSERT(pOptr != NULL);
   SSortExecInfo* pInfo = taosMemoryCalloc(1, sizeof(SSortExecInfo));
 
-  SSortOperatorInfo *pOperatorInfo = (SSortOperatorInfo*)pOptr->info;
+  SSortOperatorInfo* pOperatorInfo = (SSortOperatorInfo*)pOptr->info;
 
   *pInfo = tsortGetSortExecInfo(pOperatorInfo->pSortHandle);
   *pOptrExplain = pInfo;
   *len = sizeof(SSortExecInfo);
   return TSDB_CODE_SUCCESS;
 }
+
+typedef struct SMultiwaySortMergeOperatorInfo {
+  SOptrBasicInfo binfo;
+
+  int32_t  bufPageSize;
+  uint32_t sortBufSize;  // max buffer size for in-memory sort
+
+  SArray*      pSortInfo;
+  SSortHandle* pSortHandle;
+  SArray*      pColMatchInfo;  // for index map from table scan output
+
+  int64_t startTs;  // sort start time
+} SMultiwaySortMergeOperatorInfo;
+
+int32_t doOpenMultiwaySortMergeOperator(SOperatorInfo* pOperator) {
+  SMultiwaySortMergeOperatorInfo* pInfo = pOperator->info;
+  SExecTaskInfo*                  pTaskInfo = pOperator->pTaskInfo;
+
+  if (OPTR_IS_OPENED(pOperator)) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  pInfo->startTs = taosGetTimestampUs();
+
+  int32_t numOfBufPage = pInfo->sortBufSize / pInfo->bufPageSize;
+
+  pInfo->pSortHandle = tsortCreateSortHandle(pInfo->pSortInfo, pInfo->pColMatchInfo, SORT_MULTISOURCE_MERGE,
+                                             pInfo->bufPageSize, numOfBufPage, NULL, pTaskInfo->id.str);
+
+  tsortSetFetchRawDataFp(pInfo->pSortHandle, loadNextDataBlock, NULL, NULL);
+
+  for (int32_t i = 0; i < pOperator->numOfDownstream; ++i) {
+    SSortSource ps = {0};
+    ps.param = pOperator->pDownstream[i];
+    tsortAddSource(pInfo->pSortHandle, &ps);
+  }
+
+  int32_t code = tsortOpen(pInfo->pSortHandle);
+
+  if (code != TSDB_CODE_SUCCESS) {
+    longjmp(pTaskInfo->env, terrno);
+  }
+
+  pOperator->cost.openCost = (taosGetTimestampUs() - pInfo->startTs) / 1000.0;
+  pOperator->status = OP_RES_TO_RETURN;
+
+  OPTR_SET_OPENED(pOperator);
+  return TSDB_CODE_SUCCESS;
+}
+
+SSDataBlock* doMultiwaySortMerge(SOperatorInfo* pOperator) {
+  if (pOperator->status == OP_EXEC_DONE) {
+    return NULL;
+  }
+
+  SExecTaskInfo*                  pTaskInfo = pOperator->pTaskInfo;
+  SMultiwaySortMergeOperatorInfo* pInfo = pOperator->info;
+
+  int32_t code = pOperator->fpSet._openFn(pOperator);
+  if (code != TSDB_CODE_SUCCESS) {
+    longjmp(pTaskInfo->env, code);
+  }
+
+  SSDataBlock* pBlock =
+      getSortedBlockData(pInfo->pSortHandle, pInfo->binfo.pRes, pOperator->resultInfo.capacity, pInfo->pColMatchInfo);
+
+  if (pBlock != NULL) {
+    pOperator->resultInfo.totalRows += pBlock->info.rows;
+  } else {
+    doSetOperatorCompleted(pOperator);
+  }
+  return pBlock;
+}
+
+void destroyMultiwaySortMergeOperatorInfo(void* param, int32_t numOfOutput) {
+  SMultiwaySortMergeOperatorInfo * pInfo = (SMultiwaySortMergeOperatorInfo*)param;
+  pInfo->binfo.pRes = blockDataDestroy(pInfo->binfo.pRes);
+
+  taosArrayDestroy(pInfo->pSortInfo);
+  taosArrayDestroy(pInfo->pColMatchInfo);
+}
+
+int32_t getMultiwaySortMergeExplainExecInfo(SOperatorInfo* pOptr, void** pOptrExplain, uint32_t* len) {
+  ASSERT(pOptr != NULL);
+  SSortExecInfo* pInfo = taosMemoryCalloc(1, sizeof(SSortExecInfo));
+
+  SMultiwaySortMergeOperatorInfo* pOperatorInfo = (SMultiwaySortMergeOperatorInfo*)pOptr->info;
+
+  *pInfo = tsortGetSortExecInfo(pOperatorInfo->pSortHandle);
+  *pOptrExplain = pInfo;
+  *len = sizeof(SSortExecInfo);
+  return TSDB_CODE_SUCCESS;
+}
+
+SOperatorInfo* createMultiwaySortMergeOperatorInfo(SOperatorInfo** downStreams, int32_t numStreams,
+                                                   SSDataBlock* pResBlock, SArray* pSortInfo, SArray* pColMatchColInfo,
+                                                   SExecTaskInfo* pTaskInfo) {
+  SMultiwaySortMergeOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SMultiwaySortMergeOperatorInfo));
+  SOperatorInfo*                  pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
+  int32_t                         rowSize = pResBlock->info.rowSize;
+
+  if (pInfo == NULL || pOperator == NULL || rowSize > 100 * 1024 * 1024) {
+    goto _error;
+  }
+
+  pInfo->binfo.pRes = pResBlock;
+
+  initResultSizeInfo(pOperator, 1024);
+
+  pInfo->pSortInfo = pSortInfo;
+  pInfo->pColMatchInfo = pColMatchColInfo;
+  pOperator->name = "MultiwaySortMerge";
+  pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE;
+  pOperator->blocking = true;
+  pOperator->status = OP_NOT_OPENED;
+  pOperator->info = pInfo;
+
+  pInfo->bufPageSize = rowSize < 1024 ? 1024 : rowSize * 2;
+  pInfo->sortBufSize = pInfo->bufPageSize * 16;
+
+  pOperator->pTaskInfo = pTaskInfo;
+  pOperator->fpSet =
+      createOperatorFpSet(doOpenMultiwaySortMergeOperator, doMultiwaySortMerge, NULL, NULL,
+                          destroyMultiwaySortMergeOperatorInfo, NULL, NULL, getMultiwaySortMergeExplainExecInfo);
+
+  int32_t code = appendDownstream(pOperator, downStreams, numStreams);
+  if (code != TSDB_CODE_SUCCESS) {
+    goto _error;
+  }
+  return pOperator;
+
+_error:
+  pTaskInfo->code = TSDB_CODE_OUT_OF_MEMORY;
+  taosMemoryFree(pInfo);
+  taosMemoryFree(pOperator);
+  return NULL;
+}

source/libs/function/src/builtins.c

@@ -177,13 +177,6 @@ static int32_t translatePercentile(SFunctionNode* pFunc, char* pErrBuf, int32_t
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParamNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (nodeType(pParamNode0) != QUERY_NODE_COLUMN) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of PERCENTILE function can only be column");
-  }
-
   // param1
   SValueNode* pValue = (SValueNode*)nodesListGetNode(pFunc->pParameterList, 1);
 
@@ -218,13 +211,6 @@ static int32_t translateApercentileImpl(SFunctionNode* pFunc, char* pErrBuf, int
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParamNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (nodeType(pParamNode0) != QUERY_NODE_COLUMN) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of APERCENTILE function can only be column");
-  }
-
   // param1
   SNode* pParamNode1 = nodesListGetNode(pFunc->pParameterList, 1);
   if (nodeType(pParamNode1) != QUERY_NODE_VALUE) {
@@ -296,13 +282,6 @@ static int32_t translateTop(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
     return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParamNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (nodeType(pParamNode0) != QUERY_NODE_COLUMN) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of TOP/BOTTOM function can only be column");
-  }
-
   // param1
   SNode* pParamNode1 = nodesListGetNode(pFunc->pParameterList, 1);
   if (nodeType(pParamNode1) != QUERY_NODE_VALUE) {
@@ -350,13 +329,6 @@ static int32_t translateElapsed(SFunctionNode* pFunc, char* pErrBuf, int32_t len
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParaNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pParaNode0)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of ELAPSED function can only be column");
-  }
-
   uint8_t paraType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
   if (TSDB_DATA_TYPE_TIMESTAMP != paraType) {
     return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
@@ -422,13 +394,6 @@ static int32_t translateHistogram(SFunctionNode* pFunc, char* pErrBuf, int32_t l
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParaNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pParaNode0)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of HISTOGRAM function can only be column");
-  }
-
   uint8_t colType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
   if (!IS_NUMERIC_TYPE(colType)) {
     return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
@@ -461,12 +426,6 @@ static int32_t translateHLL(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  SNode* pPara = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pPara)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The input parameter of HYPERLOGLOG function can only be column");
-  }
-
   pFunc->node.resType = (SDataType){.bytes = tDataTypes[TSDB_DATA_TYPE_BIGINT].bytes, .type = TSDB_DATA_TYPE_BIGINT};
   return TSDB_CODE_SUCCESS;
 }
@@ -486,12 +445,6 @@ static int32_t translateStateCount(SFunctionNode* pFunc, char* pErrBuf, int32_t
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParaNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pParaNode0)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The input parameter of STATECOUNT function can only be column");
-  }
   uint8_t colType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
   if (!IS_NUMERIC_TYPE(colType)) {
     return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
@@ -532,12 +485,6 @@ static int32_t translateStateDuration(SFunctionNode* pFunc, char* pErrBuf, int32
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParaNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pParaNode0)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The input parameter of STATEDURATION function can only be column");
-  }
   uint8_t colType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
   if (!IS_NUMERIC_TYPE(colType)) {
     return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
@@ -585,12 +532,6 @@ static int32_t translateCsum(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  SNode* pPara = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pPara)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The input parameter of CSUM function can only be column");
-  }
-
   uint8_t colType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
   uint8_t resType;
   if (!IS_NUMERIC_TYPE(colType)) {
@@ -616,13 +557,6 @@ static int32_t translateMavg(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParaNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pParaNode0)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of MAVG function can only be column");
-  }
-
   uint8_t colType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
 
   // param1
@@ -652,13 +586,6 @@ static int32_t translateSample(SFunctionNode* pFunc, char* pErrBuf, int32_t len)
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParamNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pParamNode0)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of SAMPLE function can only be column");
-  }
-
   SExprNode* pCol = (SExprNode*)nodesListGetNode(pFunc->pParameterList, 0);
   uint8_t    colType = pCol->resType.type;
 
@@ -696,12 +623,6 @@ static int32_t translateTail(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pPara = nodesListGetNode(pFunc->pParameterList, 0);
-  if (QUERY_NODE_COLUMN != nodeType(pPara)) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of TAIL function can only be column");
-  }
   SExprNode* pCol = (SExprNode*)nodesListGetNode(pFunc->pParameterList, 0);
   uint8_t    colType = pCol->resType.type;
 
@@ -778,13 +699,6 @@ static int32_t translateDiff(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
     return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
   }
 
-  // param0
-  SNode* pParamNode0 = nodesListGetNode(pFunc->pParameterList, 0);
-  if (nodeType(pParamNode0) != QUERY_NODE_COLUMN) {
-    return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
-                           "The first parameter of DIFF function can only be column");
-  }
-
   uint8_t colType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
   if (!IS_SIGNED_NUMERIC_TYPE(colType) && !IS_FLOAT_TYPE(colType) && TSDB_DATA_TYPE_BOOL != colType) {
     return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);

source/libs/index/inc/indexInt.h

@@ -35,15 +35,15 @@ extern "C" {
 #endif
 
 // clang-format off
-#define indexFatal(...) do { if (idxDebugFlag & DEBUG_FATAL) {  taosPrintLog("INDEX FATAL ", DEBUG_FATAL, 255, __VA_ARGS__); }} while (0)
+#define indexFatal(...) do { if (idxDebugFlag & DEBUG_FATAL) {  taosPrintLog("IDX FATAL ", DEBUG_FATAL, 255, __VA_ARGS__); }} while (0)
-#define indexError(...) do { if (idxDebugFlag & DEBUG_ERROR) {  taosPrintLog("INDEX ERROR ", DEBUG_ERROR, 255, __VA_ARGS__); }} while (0)
+#define indexError(...) do { if (idxDebugFlag & DEBUG_ERROR) {  taosPrintLog("IDX ERROR ", DEBUG_ERROR, 255, __VA_ARGS__); }} while (0)
-#define indexWarn(...)  do { if (idxDebugFlag & DEBUG_WARN)  {  taosPrintLog("INDEX WARN ", DEBUG_WARN, 255, __VA_ARGS__); }} while (0)
+#define indexWarn(...)  do { if (idxDebugFlag & DEBUG_WARN)  {  taosPrintLog("IDX WARN ", DEBUG_WARN, 255, __VA_ARGS__); }} while (0)
-#define indexInfo(...)  do { if (idxDebugFlag & DEBUG_INFO)  { taosPrintLog("INDEX ", DEBUG_INFO, 255, __VA_ARGS__); } } while (0)
+#define indexInfo(...)  do { if (idxDebugFlag & DEBUG_INFO)  { taosPrintLog("IDX ", DEBUG_INFO, 255, __VA_ARGS__); } } while (0)
-#define indexDebug(...) do { if (idxDebugFlag & DEBUG_DEBUG) { taosPrintLog("INDEX ", DEBUG_DEBUG, sDebugFlag, __VA_ARGS__);} } while (0)
+#define indexDebug(...) do { if (idxDebugFlag & DEBUG_DEBUG) { taosPrintLog("IDX ", DEBUG_DEBUG, idxDebugFlag, __VA_ARGS__);} } while (0)
-#define indexTrace(...) do { if (idxDebugFlag & DEBUG_TRACE) { taosPrintLog("INDEX ", DEBUG_TRACE, sDebugFlag, __VA_ARGS__);} } while (0)
+#define indexTrace(...) do { if (idxDebugFlag & DEBUG_TRACE) { taosPrintLog("IDX", DEBUG_TRACE, idxDebugFlag, __VA_ARGS__);} } while (0)
 // clang-format on
 
-typedef enum { LT, LE, GT, GE } RangeType;
+typedef enum { LT, LE, GT, GE, CONTAINS } RangeType;
 typedef enum { kTypeValue, kTypeDeletion } STermValueType;
 
 typedef struct SIndexStat {

source/libs/index/src/indexCache.c

@@ -90,7 +90,7 @@ static int32_t cacheSearchTerm(void* cache, SIndexTerm* term, SIdxTRslt* tr, STe
       break;
     }
     CacheTerm* c = (CacheTerm*)SL_GET_NODE_DATA(node);
-    if (0 == strcmp(c->colVal, pCt->colVal)) {
+    if (0 == strcmp(c->colVal, pCt->colVal) && strlen(pCt->colVal) == strlen(c->colVal)) {
       if (c->operaType == ADD_VALUE) {
         INDEX_MERGE_ADD_DEL(tr->del, tr->add, c->uid)
         // taosArrayPush(result, &c->uid);
@@ -222,7 +222,7 @@ static int32_t cacheSearchTerm_JSON(void* cache, SIndexTerm* term, SIdxTRslt* tr
   return TSDB_CODE_SUCCESS;
 }
 static int32_t cacheSearchPrefix_JSON(void* cache, SIndexTerm* term, SIdxTRslt* tr, STermValueType* s) {
-  return TSDB_CODE_SUCCESS;
+  return cacheSearchCompareFunc_JSON(cache, term, tr, s, CONTAINS);
 }
 static int32_t cacheSearchSuffix_JSON(void* cache, SIndexTerm* term, SIdxTRslt* tr, STermValueType* s) {
   return TSDB_CODE_SUCCESS;
@@ -242,6 +242,9 @@ static int32_t cacheSearchGreaterThan_JSON(void* cache, SIndexTerm* term, SIdxTR
 static int32_t cacheSearchGreaterEqual_JSON(void* cache, SIndexTerm* term, SIdxTRslt* tr, STermValueType* s) {
   return cacheSearchCompareFunc_JSON(cache, term, tr, s, GE);
 }
+static int32_t cacheSearchContain_JSON(void* cache, SIndexTerm* term, SIdxTRslt* tr, STermValueType* s) {
+  return cacheSearchCompareFunc_JSON(cache, term, tr, s, CONTAINS);
+}
 static int32_t cacheSearchRange_JSON(void* cache, SIndexTerm* term, SIdxTRslt* tr, STermValueType* s) {
   return TSDB_CODE_SUCCESS;
 }

source/libs/index/src/indexComm.c

@@ -97,6 +97,11 @@ static TExeCond tCompareGreaterEqual(void* a, void* b, int8_t type) {
   __compar_fn_t func = indexGetCompar(type);
   return tCompare(func, QUERY_GREATER_EQUAL, a, b, type);
 }
+
+static TExeCond tCompareContains(void* a, void* b, int8_t type) {
+  __compar_fn_t func = indexGetCompar(type);
+  return tCompare(func, QUERY_TERM, a, b, type);
+}
 TExeCond tCompare(__compar_fn_t func, int8_t cmptype, void* a, void* b, int8_t dtype) {
   if (dtype == TSDB_DATA_TYPE_BINARY || dtype == TSDB_DATA_TYPE_NCHAR || dtype == TSDB_DATA_TYPE_VARBINARY) {
     return tDoCompare(func, cmptype, a, b);
@@ -185,12 +190,14 @@ TExeCond tDoCompare(__compar_fn_t func, int8_t comparType, void* a, void* b) {
     case QUERY_TERM: {
       if (ret == 0) return MATCH;
     }
+    default:
+      return BREAK;
   }
   return CONTINUE;
 }
 
-static TExeCond (*rangeCompare[])(void* a, void* b, int8_t type) = {tCompareLessThan, tCompareLessEqual,
+static TExeCond (*rangeCompare[])(void* a, void* b, int8_t type) = {
-                                                                    tCompareGreaterThan, tCompareGreaterEqual};
+    tCompareLessThan, tCompareLessEqual, tCompareGreaterThan, tCompareGreaterEqual, tCompareContains};
 
 _cache_range_compare indexGetCompare(RangeType ty) { return rangeCompare[ty]; }
 

source/libs/index/src/indexFilter.c

@@ -38,7 +38,7 @@ typedef struct SIFParam {
   col_id_t      colId;
   int64_t       suid;  // add later
   char          dbName[TSDB_DB_NAME_LEN];
-  char          colName[TSDB_COL_NAME_LEN];
+  char          colName[TSDB_COL_NAME_LEN * 2 + 4];
 
   SIndexMetaArg arg;
 } SIFParam;
@@ -64,6 +64,8 @@ static int32_t sifGetFuncFromSql(EOperatorType src, EIndexQueryType *dst) {
     *dst = QUERY_TERM;
   } else if (src == OP_TYPE_LIKE || src == OP_TYPE_MATCH || src == OP_TYPE_NMATCH) {
     *dst = QUERY_REGEX;
+  } else if (src == OP_TYPE_JSON_CONTAINS) {
+    *dst = QUERY_PREFIX;
   } else {
     return TSDB_CODE_QRY_INVALID_INPUT;
   }
@@ -171,7 +173,10 @@ static int32_t sifInitJsonParam(SNode *node, SIFParam *param, SIFCtx *ctx) {
   param->colId = l->colId;
   param->colValType = l->node.resType.type;
   memcpy(param->dbName, l->dbName, sizeof(l->dbName));
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wformat-overflow"
   sprintf(param->colName, "%s_%s", l->colName, r->literal);
+#pragma GCC diagnostic pop
   param->colValType = r->typeData;
   return 0;
   // memcpy(param->colName, l->colName, sizeof(l->colName));
@@ -183,6 +188,7 @@ static int32_t sifInitParam(SNode *node, SIFParam *param, SIFCtx *ctx) {
       SIF_ERR_RET(sifGetValueFromNode(node, &param->condValue));
       param->colId = -1;
       param->colValType = (uint8_t)(vn->node.resType.type);
+      memcpy(param->colName, vn->literal, strlen(vn->literal));
       break;
     }
     case QUERY_NODE_COLUMN: {
@@ -234,7 +240,7 @@ static int32_t sifInitOperParams(SIFParam **params, SOperatorNode *node, SIFCtx
     indexError("invalid operation node, left: %p, rigth: %p", node->pLeft, node->pRight);
     SIF_ERR_RET(TSDB_CODE_QRY_INVALID_INPUT);
   }
-  if (node->opType == OP_TYPE_JSON_GET_VALUE || node->opType == OP_TYPE_JSON_CONTAINS) {
+  if (node->opType == OP_TYPE_JSON_GET_VALUE) {
     return code;
   }
   SIFParam *paramList = taosMemoryCalloc(nParam, sizeof(SIFParam));
@@ -417,8 +423,8 @@ static int32_t sifNotMatchFunc(SIFParam *left, SIFParam *right, SIFParam *output
   return sifDoIndex(left, right, id, output);
 }
 static int32_t sifJsonContains(SIFParam *left, SIFParam *right, SIFParam *output) {
-  // return 0
+  int id = OP_TYPE_JSON_CONTAINS;
-  return 0;
+  return sifDoIndex(left, right, id, output);
 }
 static int32_t sifJsonGetValue(SIFParam *left, SIFParam *rigth, SIFParam *output) {
   // return 0
@@ -498,9 +504,11 @@ static int32_t sifExecOper(SOperatorNode *node, SIFCtx *ctx, SIFParam *output) {
   int32_t code = 0;
   int32_t nParam = sifGetOperParamNum(node->opType);
   if (nParam <= 1) {
-    SIF_ERR_RET(TSDB_CODE_QRY_INVALID_INPUT);
+    output->status = SFLT_NOT_INDEX;
+    return code;
+    // SIF_ERR_RET(TSDB_CODE_QRY_INVALID_INPUT);
   }
-  if (node->opType == OP_TYPE_JSON_GET_VALUE || node->opType == OP_TYPE_JSON_CONTAINS) {
+  if (node->opType == OP_TYPE_JSON_GET_VALUE) {
     return code;
   }
   SIFParam *params = NULL;
@@ -614,11 +622,11 @@ EDealRes sifCalcWalker(SNode *node, void *context) {
   }
 
   if (QUERY_NODE_OPERATOR == nodeType(node)) {
-    indexInfo("node type for index filter, type: %d", nodeType(node));
+    // indexInfo("node type for index filter, type: %d", nodeType(node));
     return sifWalkOper(node, ctx);
   }
 
-  indexError("invalid node type for index filter calculating, type:%d", nodeType(node));
+  // indexError("invalid node type for index filter calculating, type:%d", nodeType(node));
   ctx->code = TSDB_CODE_QRY_INVALID_INPUT;
   return DEAL_RES_ERROR;
 }

source/libs/index/src/indexTfile.c

@@ -425,8 +425,7 @@ static int32_t tfSearchTerm_JSON(void* reader, SIndexTerm* tem, SIdxTRslt* tr) {
   return TSDB_CODE_SUCCESS;
 }
 static int32_t tfSearchPrefix_JSON(void* reader, SIndexTerm* tem, SIdxTRslt* tr) {
-  // impl later
+  return tfSearchCompareFunc_JSON(reader, tem, tr, CONTAINS);
-  return TSDB_CODE_SUCCESS;
 }
 static int32_t tfSearchSuffix_JSON(void* reader, SIndexTerm* tem, SIdxTRslt* tr) {
   // impl later
@@ -466,10 +465,6 @@ static int32_t tfSearchCompareFunc_JSON(void* reader, SIndexTerm* tem, SIdxTRslt
   AutomationCtx*    ctx = automCtxCreate((void*)p, AUTOMATION_PREFIX);
   FstStreamBuilder* sb = fstSearch(((TFileReader*)reader)->fst, ctx);
 
-  // FstSlice h = fstSliceCreate((uint8_t*)p, skip);
-  // fstStreamBuilderSetRange(sb, &h, ctype);
-  // fstSliceDestroy(&h);
-
   StreamWithState*       st = streamBuilderIntoStream(sb);
   StreamWithStateResult* rt = NULL;
   while ((rt = streamWithStateNextWith(st, NULL)) != NULL) {

source/libs/monitor/src/monMsg.c

@@ -569,6 +569,7 @@ int32_t tSerializeSQnodeLoad(void *buf, int32_t bufLen, SQnodeLoad *pInfo) {
   if (tEncodeI64(&encoder, pInfo->numOfProcessedFetch) < 0) return -1;
   if (tEncodeI64(&encoder, pInfo->numOfProcessedDrop) < 0) return -1;
   if (tEncodeI64(&encoder, pInfo->numOfProcessedHb) < 0) return -1;
+  if (tEncodeI64(&encoder, pInfo->numOfProcessedDelete) < 0) return -1;
   if (tEncodeI64(&encoder, pInfo->cacheDataSize) < 0) return -1;
   if (tEncodeI64(&encoder, pInfo->numOfQueryInQueue) < 0) return -1;
   if (tEncodeI64(&encoder, pInfo->numOfFetchInQueue) < 0) return -1;
@@ -591,6 +592,7 @@ int32_t tDeserializeSQnodeLoad(void *buf, int32_t bufLen, SQnodeLoad *pInfo) {
   if (tDecodeI64(&decoder, &pInfo->numOfProcessedFetch) < 0) return -1;
   if (tDecodeI64(&decoder, &pInfo->numOfProcessedDrop) < 0) return -1;
   if (tDecodeI64(&decoder, &pInfo->numOfProcessedHb) < 0) return -1;
+  if (tDecodeI64(&decoder, &pInfo->numOfProcessedDelete) < 0) return -1;
   if (tDecodeI64(&decoder, &pInfo->cacheDataSize) < 0) return -1;
   if (tDecodeI64(&decoder, &pInfo->numOfQueryInQueue) < 0) return -1;
   if (tDecodeI64(&decoder, &pInfo->numOfFetchInQueue) < 0) return -1;

source/libs/nodes/src/nodesCloneFuncs.c

@@ -318,6 +318,7 @@ static SNode* logicScanCopy(const SScanLogicNode* pSrc, SScanLogicNode* pDst) {
   CLONE_NODE_LIST_FIELD(pScanPseudoCols);
   COPY_SCALAR_FIELD(tableType);
   COPY_SCALAR_FIELD(tableId);
+  COPY_SCALAR_FIELD(stableId);
   CLONE_OBJECT_FIELD(pVgroupList, vgroupsInfoClone);
   COPY_SCALAR_FIELD(scanType);
   COPY_OBJECT_FIELD(scanSeq[0], sizeof(uint8_t) * 2);
@@ -370,7 +371,7 @@ static SNode* logicVnodeModifCopy(const SVnodeModifyLogicNode* pSrc, SVnodeModif
   COPY_BASE_OBJECT_FIELD(node, logicNodeCopy);
   COPY_SCALAR_FIELD(modifyType);
   COPY_SCALAR_FIELD(msgType);
-  CLONE_NODE_FIELD(pModifyRows);
+  CLONE_NODE_FIELD(pAffectedRows);
   COPY_SCALAR_FIELD(tableId);
   COPY_SCALAR_FIELD(tableType);
   COPY_CHAR_ARRAY_FIELD(tableFName);
@@ -387,6 +388,7 @@ static SNode* logicExchangeCopy(const SExchangeLogicNode* pSrc, SExchangeLogicNo
 static SNode* logicMergeCopy(const SMergeLogicNode* pSrc, SMergeLogicNode* pDst) {
   COPY_BASE_OBJECT_FIELD(node, logicNodeCopy);
   CLONE_NODE_LIST_FIELD(pMergeKeys);
+  CLONE_NODE_LIST_FIELD(pInputs);
   COPY_SCALAR_FIELD(numOfChannels);
   COPY_SCALAR_FIELD(srcGroupId);
   return (SNode*)pDst;
@@ -432,6 +434,12 @@ static SNode* logicPartitionCopy(const SPartitionLogicNode* pSrc, SPartitionLogi
   return (SNode*)pDst;
 }
 
+static SNode* logicIndefRowsFuncCopy(const SIndefRowsFuncLogicNode* pSrc, SIndefRowsFuncLogicNode* pDst) {
+  COPY_BASE_OBJECT_FIELD(node, logicNodeCopy);
+  CLONE_NODE_LIST_FIELD(pVectorFuncs);
+  return (SNode*)pDst;
+}
+
 static SNode* logicSubplanCopy(const SLogicSubplan* pSrc, SLogicSubplan* pDst) {
   COPY_OBJECT_FIELD(id, sizeof(SSubplanId));
   CLONE_NODE_FIELD(pNode);
@@ -563,6 +571,8 @@ SNodeptr nodesCloneNode(const SNodeptr pNode) {
       return logicSortCopy((const SSortLogicNode*)pNode, (SSortLogicNode*)pDst);
     case QUERY_NODE_LOGIC_PLAN_PARTITION:
       return logicPartitionCopy((const SPartitionLogicNode*)pNode, (SPartitionLogicNode*)pDst);
+    case QUERY_NODE_LOGIC_PLAN_INDEF_ROWS_FUNC:
+      return logicIndefRowsFuncCopy((const SIndefRowsFuncLogicNode*)pNode, (SIndefRowsFuncLogicNode*)pDst);
     case QUERY_NODE_LOGIC_SUBPLAN:
       return logicSubplanCopy((const SLogicSubplan*)pNode, (SLogicSubplan*)pDst);
     default:

source/libs/nodes/src/nodesCodeFuncs.c

@@ -202,6 +202,8 @@ const char* nodesNodeName(ENodeType type) {
       return "LogicSort";
     case QUERY_NODE_LOGIC_PLAN_PARTITION:
       return "LogicPartition";
+    case QUERY_NODE_LOGIC_PLAN_INDEF_ROWS_FUNC:
+      return "LogicIndefRowsFunc";
     case QUERY_NODE_LOGIC_SUBPLAN:
       return "LogicSubplan";
     case QUERY_NODE_LOGIC_PLAN:
@@ -230,6 +232,8 @@ const char* nodesNodeName(ENodeType type) {
       return "PhysiSort";
     case QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL:
       return "PhysiHashInterval";
+    case QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL:
+      return "PhysiMergeInterval";
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL:
       return "PhysiStreamInterval";
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_INTERVAL:
@@ -248,6 +252,8 @@ const char* nodesNodeName(ENodeType type) {
       return "PhysiStreamStateWindow";
     case QUERY_NODE_PHYSICAL_PLAN_PARTITION:
       return "PhysiPartition";
+    case QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC:
+      return "PhysiIndefRowsFunc";
     case QUERY_NODE_PHYSICAL_PLAN_DISPATCH:
       return "PhysiDispatch";
     case QUERY_NODE_PHYSICAL_PLAN_INSERT:
@@ -611,7 +617,7 @@ static int32_t jsonToLogicProjectNode(const SJson* pJson, void* pObj) {
 
 static const char* jkVnodeModifyLogicPlanModifyType = "ModifyType";
 static const char* jkVnodeModifyLogicPlanMsgType = "MsgType";
-static const char* jkVnodeModifyLogicPlanModifyRows = "ModifyRows";
+static const char* jkVnodeModifyLogicPlanAffectedRows = "AffectedRows";
 
 static int32_t logicVnodeModifyNodeToJson(const void* pObj, SJson* pJson) {
   const SVnodeModifyLogicNode* pNode = (const SVnodeModifyLogicNode*)pObj;
@@ -624,7 +630,7 @@ static int32_t logicVnodeModifyNodeToJson(const void* pObj, SJson* pJson) {
     code = tjsonAddIntegerToObject(pJson, jkVnodeModifyLogicPlanMsgType, pNode->msgType);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = tjsonAddObject(pJson, jkVnodeModifyLogicPlanModifyRows, nodeToJson, pNode->pModifyRows);
+    code = tjsonAddObject(pJson, jkVnodeModifyLogicPlanAffectedRows, nodeToJson, pNode->pAffectedRows);
   }
 
   return code;
@@ -641,7 +647,7 @@ static int32_t jsonToLogicVnodeModifyNode(const SJson* pJson, void* pObj) {
     code = tjsonGetIntValue(pJson, jkVnodeModifyLogicPlanMsgType, &pNode->msgType);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = jsonToNodeObject(pJson, jkVnodeModifyLogicPlanModifyRows, &pNode->pModifyRows);
+    code = jsonToNodeObject(pJson, jkVnodeModifyLogicPlanAffectedRows, &pNode->pAffectedRows);
   }
 
   return code;
@@ -911,6 +917,30 @@ static int32_t jsonToLogicPartitionNode(const SJson* pJson, void* pObj) {
   return code;
 }
 
+static const char* jkIndefRowsFuncLogicPlanVectorFuncs = "VectorFuncs";
+
+static int32_t logicIndefRowsFuncNodeToJson(const void* pObj, SJson* pJson) {
+  const SIndefRowsFuncLogicNode* pNode = (const SIndefRowsFuncLogicNode*)pObj;
+
+  int32_t code = logicPlanNodeToJson(pObj, pJson);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = nodeListToJson(pJson, jkIndefRowsFuncLogicPlanVectorFuncs, pNode->pVectorFuncs);
+  }
+
+  return code;
+}
+
+static int32_t jsonToLogicIndefRowsFuncNode(const SJson* pJson, void* pObj) {
+  SIndefRowsFuncLogicNode* pNode = (SIndefRowsFuncLogicNode*)pObj;
+
+  int32_t code = jsonToLogicPlanNode(pJson, pObj);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = jsonToNodeList(pJson, jkIndefRowsFuncLogicPlanVectorFuncs, &pNode->pVectorFuncs);
+  }
+
+  return code;
+}
+
 static const char* jkSubplanIdQueryId = "QueryId";
 static const char* jkSubplanIdGroupId = "GroupId";
 static const char* jkSubplanIdSubplanId = "SubplanId";
@@ -1249,6 +1279,7 @@ static int32_t jsonToName(const SJson* pJson, void* pObj) {
 static const char* jkScanPhysiPlanScanCols = "ScanCols";
 static const char* jkScanPhysiPlanScanPseudoCols = "ScanPseudoCols";
 static const char* jkScanPhysiPlanTableId = "TableId";
+static const char* jkScanPhysiPlanSTableId = "STableId";
 static const char* jkScanPhysiPlanTableType = "TableType";
 static const char* jkScanPhysiPlanTableName = "TableName";
 
@@ -1265,6 +1296,9 @@ static int32_t physiScanNodeToJson(const void* pObj, SJson* pJson) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonAddIntegerToObject(pJson, jkScanPhysiPlanTableId, pNode->uid);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddIntegerToObject(pJson, jkScanPhysiPlanSTableId, pNode->suid);
+  }
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonAddIntegerToObject(pJson, jkScanPhysiPlanTableType, pNode->tableType);
   }
@@ -1288,6 +1322,9 @@ static int32_t jsonToPhysiScanNode(const SJson* pJson, void* pObj) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonGetUBigIntValue(pJson, jkScanPhysiPlanTableId, &pNode->uid);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonGetUBigIntValue(pJson, jkScanPhysiPlanSTableId, &pNode->suid);
+  }
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonGetTinyIntValue(pJson, jkScanPhysiPlanTableType, &pNode->tableType);
   }
@@ -1644,6 +1681,7 @@ static int32_t jsonToPhysiExchangeNode(const SJson* pJson, void* pObj) {
 }
 
 static const char* jkMergePhysiPlanMergeKeys = "MergeKeys";
+static const char* jkMergePhysiPlanTargets = "Targets";
 static const char* jkMergePhysiPlanNumOfChannels = "NumOfChannels";
 static const char* jkMergePhysiPlanSrcGroupId = "SrcGroupId";
 
@@ -1654,6 +1692,9 @@ static int32_t physiMergeNodeToJson(const void* pObj, SJson* pJson) {
   if (TSDB_CODE_SUCCESS == code) {
     code = nodeListToJson(pJson, jkMergePhysiPlanMergeKeys, pNode->pMergeKeys);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = nodeListToJson(pJson, jkMergePhysiPlanTargets, pNode->pTargets);
+  }
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonAddIntegerToObject(pJson, jkMergePhysiPlanNumOfChannels, pNode->numOfChannels);
   }
@@ -1671,6 +1712,9 @@ static int32_t jsonToPhysiMergeNode(const SJson* pJson, void* pObj) {
   if (TSDB_CODE_SUCCESS == code) {
     code = jsonToNodeList(pJson, jkMergePhysiPlanMergeKeys, &pNode->pMergeKeys);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = jsonToNodeList(pJson, jkMergePhysiPlanTargets, &pNode->pTargets);
+  }
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonGetIntValue(pJson, jkMergePhysiPlanNumOfChannels, &pNode->numOfChannels);
   }
@@ -1979,6 +2023,37 @@ static int32_t jsonToPhysiPartitionNode(const SJson* pJson, void* pObj) {
   return code;
 }
 
+static const char* jkIndefRowsFuncPhysiPlanExprs = "Exprs";
+static const char* jkIndefRowsFuncPhysiPlanVectorFuncs = "VectorFuncs";
+
+static int32_t physiIndefRowsFuncNodeToJson(const void* pObj, SJson* pJson) {
+  const SIndefRowsFuncPhysiNode* pNode = (const SIndefRowsFuncPhysiNode*)pObj;
+
+  int32_t code = physicPlanNodeToJson(pObj, pJson);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = nodeListToJson(pJson, jkIndefRowsFuncPhysiPlanExprs, pNode->pExprs);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = nodeListToJson(pJson, jkIndefRowsFuncPhysiPlanVectorFuncs, pNode->pVectorFuncs);
+  }
+
+  return code;
+}
+
+static int32_t jsonToPhysiIndefRowsFuncNode(const SJson* pJson, void* pObj) {
+  SIndefRowsFuncPhysiNode* pNode = (SIndefRowsFuncPhysiNode*)pObj;
+
+  int32_t code = jsonToPhysicPlanNode(pJson, pObj);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = jsonToNodeList(pJson, jkIndefRowsFuncPhysiPlanExprs, &pNode->pExprs);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = jsonToNodeList(pJson, jkIndefRowsFuncPhysiPlanVectorFuncs, &pNode->pVectorFuncs);
+  }
+
+  return code;
+}
+
 static const char* jkDataSinkInputDataBlockDesc = "InputDataBlockDesc";
 
 static int32_t physicDataSinkNodeToJson(const void* pObj, SJson* pJson) {
@@ -2000,6 +2075,7 @@ static const char* jkDeletePhysiPlanTableType = "TableType";
 static const char* jkDeletePhysiPlanTableFName = "TableFName";
 static const char* jkDeletePhysiPlanDeleteTimeRangeStartKey = "DeleteTimeRangeStartKey";
 static const char* jkDeletePhysiPlanDeleteTimeRangeEndKey = "DeleteTimeRangeEndKey";
+static const char* jkDeletePhysiPlanAffectedRows = "AffectedRows";
 
 static int32_t physiDeleteNodeToJson(const void* pObj, SJson* pJson) {
   const SDataDeleterNode* pNode = (const SDataDeleterNode*)pObj;
@@ -2020,6 +2096,9 @@ static int32_t physiDeleteNodeToJson(const void* pObj, SJson* pJson) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonAddIntegerToObject(pJson, jkDeletePhysiPlanDeleteTimeRangeEndKey, pNode->deleteTimeRange.ekey);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddObject(pJson, jkDeletePhysiPlanAffectedRows, nodeToJson, pNode->pAffectedRows);
+  }
 
   return code;
 }
@@ -2043,6 +2122,9 @@ static int32_t jsonToPhysiDeleteNode(const SJson* pJson, void* pObj) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonGetBigIntValue(pJson, jkDeletePhysiPlanDeleteTimeRangeEndKey, &pNode->deleteTimeRange.ekey);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = jsonToNodeObject(pJson, jkDeletePhysiPlanAffectedRows, &pNode->pAffectedRows);
+  }
 
   return code;
 }
@@ -3777,6 +3859,8 @@ static int32_t specificNodeToJson(const void* pObj, SJson* pJson) {
       return logicSortNodeToJson(pObj, pJson);
     case QUERY_NODE_LOGIC_PLAN_PARTITION:
       return logicPartitionNodeToJson(pObj, pJson);
+    case QUERY_NODE_LOGIC_PLAN_INDEF_ROWS_FUNC:
+      return logicIndefRowsFuncNodeToJson(pObj, pJson);
     case QUERY_NODE_LOGIC_SUBPLAN:
       return logicSubplanToJson(pObj, pJson);
     case QUERY_NODE_LOGIC_PLAN:
@@ -3802,6 +3886,7 @@ static int32_t specificNodeToJson(const void* pObj, SJson* pJson) {
     case QUERY_NODE_PHYSICAL_PLAN_SORT:
       return physiSortNodeToJson(pObj, pJson);
     case QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL:
+    case QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL:
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL:
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_INTERVAL:
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_SEMI_INTERVAL:
@@ -3816,6 +3901,8 @@ static int32_t specificNodeToJson(const void* pObj, SJson* pJson) {
       return physiStateWindowNodeToJson(pObj, pJson);
     case QUERY_NODE_PHYSICAL_PLAN_PARTITION:
       return physiPartitionNodeToJson(pObj, pJson);
+    case QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC:
+      return physiIndefRowsFuncNodeToJson(pObj, pJson);
     case QUERY_NODE_PHYSICAL_PLAN_DISPATCH:
       return physiDispatchNodeToJson(pObj, pJson);
     case QUERY_NODE_PHYSICAL_PLAN_INSERT:
@@ -3905,6 +3992,8 @@ static int32_t jsonToSpecificNode(const SJson* pJson, void* pObj) {
       return jsonToLogicSortNode(pJson, pObj);
     case QUERY_NODE_LOGIC_PLAN_PARTITION:
       return jsonToLogicPartitionNode(pJson, pObj);
+    case QUERY_NODE_LOGIC_PLAN_INDEF_ROWS_FUNC:
+      return jsonToLogicIndefRowsFuncNode(pJson, pObj);
     case QUERY_NODE_LOGIC_SUBPLAN:
       return jsonToLogicSubplan(pJson, pObj);
     case QUERY_NODE_LOGIC_PLAN:
@@ -3930,6 +4019,7 @@ static int32_t jsonToSpecificNode(const SJson* pJson, void* pObj) {
     case QUERY_NODE_PHYSICAL_PLAN_SORT:
       return jsonToPhysiSortNode(pJson, pObj);
     case QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL:
+    case QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL:
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL:
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_INTERVAL:
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_SEMI_INTERVAL:
@@ -3944,6 +4034,8 @@ static int32_t jsonToSpecificNode(const SJson* pJson, void* pObj) {
       return jsonToPhysiStateWindowNode(pJson, pObj);
     case QUERY_NODE_PHYSICAL_PLAN_PARTITION:
       return jsonToPhysiPartitionNode(pJson, pObj);
+    case QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC:
+      return jsonToPhysiIndefRowsFuncNode(pJson, pObj);
     case QUERY_NODE_PHYSICAL_PLAN_DISPATCH:
       return jsonToPhysiDispatchNode(pJson, pObj);
     case QUERY_NODE_PHYSICAL_PLAN_DELETE:

source/libs/nodes/src/nodesTraverseFuncs.c

@@ -350,6 +350,7 @@ void nodesWalkSelectStmt(SSelectStmt* pSelect, ESqlClause clause, FNodeWalker wa
     case SQL_CLAUSE_GROUP_BY:
       nodesWalkExpr(pSelect->pHaving, walker, pContext);
     case SQL_CLAUSE_HAVING:
+    case SQL_CLAUSE_SELECT:
     case SQL_CLAUSE_DISTINCT:
       nodesWalkExprs(pSelect->pOrderByList, walker, pContext);
     case SQL_CLAUSE_ORDER_BY:
@@ -382,6 +383,7 @@ void nodesRewriteSelectStmt(SSelectStmt* pSelect, ESqlClause clause, FNodeRewrit
     case SQL_CLAUSE_GROUP_BY:
       nodesRewriteExpr(&(pSelect->pHaving), rewriter, pContext);
     case SQL_CLAUSE_HAVING:
+    case SQL_CLAUSE_SELECT:
     case SQL_CLAUSE_DISTINCT:
       nodesRewriteExprs(pSelect->pOrderByList, rewriter, pContext);
     case SQL_CLAUSE_ORDER_BY:

source/libs/nodes/src/nodesUtilFuncs.c

@@ -232,6 +232,8 @@ SNodeptr nodesMakeNode(ENodeType type) {
       return makeNode(type, sizeof(SSortLogicNode));
     case QUERY_NODE_LOGIC_PLAN_PARTITION:
       return makeNode(type, sizeof(SPartitionLogicNode));
+    case QUERY_NODE_LOGIC_PLAN_INDEF_ROWS_FUNC:
+      return makeNode(type, sizeof(SIndefRowsFuncLogicNode));
     case QUERY_NODE_LOGIC_SUBPLAN:
       return makeNode(type, sizeof(SLogicSubplan));
     case QUERY_NODE_LOGIC_PLAN:
@@ -260,8 +262,8 @@ SNodeptr nodesMakeNode(ENodeType type) {
       return makeNode(type, sizeof(SSortPhysiNode));
     case QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL:
       return makeNode(type, sizeof(SIntervalPhysiNode));
-    case QUERY_NODE_PHYSICAL_PLAN_SORT_MERGE_INTERVAL:
+    case QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL:
-      return makeNode(type, sizeof(SSortMergeIntervalPhysiNode));
+      return makeNode(type, sizeof(SMergeIntervalPhysiNode));
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL:
       return makeNode(type, sizeof(SStreamIntervalPhysiNode));
     case QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_INTERVAL:
@@ -280,6 +282,8 @@ SNodeptr nodesMakeNode(ENodeType type) {
       return makeNode(type, sizeof(SStreamStateWinodwPhysiNode));
     case QUERY_NODE_PHYSICAL_PLAN_PARTITION:
       return makeNode(type, sizeof(SPartitionPhysiNode));
+    case QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC:
+      return makeNode(type, sizeof(SIndefRowsFuncPhysiNode));
     case QUERY_NODE_PHYSICAL_PLAN_DISPATCH:
       return makeNode(type, sizeof(SDataDispatcherNode));
     case QUERY_NODE_PHYSICAL_PLAN_INSERT:

source/libs/parser/src/parAstParser.c

@@ -258,6 +258,17 @@ static int32_t collectMetaKeyFromExplain(SCollectMetaKeyCxt* pCxt, SExplainStmt*
   return collectMetaKeyFromQuery(pCxt, pStmt->pQuery);
 }
 
+static int32_t collectMetaKeyFromDescribe(SCollectMetaKeyCxt* pCxt, SDescribeStmt* pStmt) {
+  SName name = {.type = TSDB_TABLE_NAME_T, .acctId = pCxt->pParseCxt->acctId};
+  strcpy(name.dbname, pStmt->dbName);
+  strcpy(name.tname, pStmt->tableName);
+  int32_t code = catalogRemoveTableMeta(pCxt->pParseCxt->pCatalog, &name);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = reserveTableMetaInCache(pCxt->pParseCxt->acctId, pStmt->dbName, pStmt->tableName, pCxt->pMetaCache);
+  }
+  return code;
+}
+
 static int32_t collectMetaKeyFromCreateStream(SCollectMetaKeyCxt* pCxt, SCreateStreamStmt* pStmt) {
   return collectMetaKeyFromQuery(pCxt, pStmt->pQuery);
 }
@@ -381,6 +392,8 @@ static int32_t collectMetaKeyFromQuery(SCollectMetaKeyCxt* pCxt, SNode* pStmt) {
       return collectMetaKeyFromCreateTopic(pCxt, (SCreateTopicStmt*)pStmt);
     case QUERY_NODE_EXPLAIN_STMT:
       return collectMetaKeyFromExplain(pCxt, (SExplainStmt*)pStmt);
+    case QUERY_NODE_DESCRIBE_STMT:
+      return collectMetaKeyFromDescribe(pCxt, (SDescribeStmt*)pStmt);
     case QUERY_NODE_CREATE_STREAM_STMT:
       return collectMetaKeyFromCreateStream(pCxt, (SCreateStreamStmt*)pStmt);
     case QUERY_NODE_SHOW_DNODES_STMT:

source/libs/parser/src/parInsert.c

@@ -1006,7 +1006,7 @@ static int32_t parseTagsClause(SInsertParseContext* pCxt, SSchema* pSchema, uint
     }
 
     SSchema* pTagSchema = &pSchema[pCxt->tags.boundColumns[i]];
-    char*    tmpTokenBuf = taosMemoryCalloc(1, sToken.n);  // this can be optimize with parse column
+    char*    tmpTokenBuf = taosMemoryCalloc(1, sToken.n);  //todo this can be optimize with parse column
     code = checkAndTrimValue(&sToken, tmpTokenBuf, &pCxt->msg);
     if (code != TSDB_CODE_SUCCESS) {
       taosMemoryFree(tmpTokenBuf);
@@ -1018,7 +1018,11 @@ static int32_t parseTagsClause(SInsertParseContext* pCxt, SSchema* pSchema, uint
         taosMemoryFree(tmpTokenBuf);
         goto end;
       }
+      if(isNullStr(&sToken)) {
+        code = tTagNew(pTagVals, 1, true, &pTag);
+      } else {
         code = parseJsontoTagData(sToken.z, pTagVals, &pTag, &pCxt->msg);
+      }
       taosMemoryFree(tmpTokenBuf);
       if (code != TSDB_CODE_SUCCESS) {
         goto end;

source/libs/parser/src/parTokenizer.c

@@ -64,6 +64,7 @@ static SKeyword keywordTable[] = {
     {"CONSUMER",      TK_CONSUMER},
     {"COUNT",         TK_COUNT},
     {"CREATE",        TK_CREATE},
+    {"CONTAINS",      TK_CONTAINS}, 
     {"DATABASE",      TK_DATABASE},
     {"DATABASES",     TK_DATABASES},
     {"DAYS",          TK_DAYS},

source/libs/parser/src/parTranslater.c

@@ -847,8 +847,12 @@ static EDealRes translateJsonOperator(STranslateContext* pCxt, SOperatorNode* pO
   if (TSDB_DATA_TYPE_JSON != ldt.type || TSDB_DATA_TYPE_BINARY != rdt.type) {
     return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, ((SExprNode*)(pOp->pRight))->aliasName);
   }
+  if(pOp->opType == OP_TYPE_JSON_GET_VALUE){
     pOp->node.resType.type = TSDB_DATA_TYPE_JSON;
-  pOp->node.resType.bytes = tDataTypes[TSDB_DATA_TYPE_JSON].bytes;
+  }else if(pOp->opType == OP_TYPE_JSON_CONTAINS){
+    pOp->node.resType.type = TSDB_DATA_TYPE_BOOL;
+  }
+  pOp->node.resType.bytes = tDataTypes[pOp->node.resType.type].bytes;
   return DEAL_RES_CONTINUE;
 }
 
@@ -2903,7 +2907,6 @@ static int32_t buildRollupAst(STranslateContext* pCxt, SCreateTableStmt* pStmt,
     }
   }
 
-  taosArrayDestroy(dbCfg.pRetensions);
   return code;
 }
 
@@ -5012,6 +5015,10 @@ static int32_t setQuery(STranslateContext* pCxt, SQuery* pQuery) {
       pQuery->haveResultSet = true;
       pQuery->msgType = TDMT_VND_QUERY;
       break;
+    case QUERY_NODE_DELETE_STMT:
+      pQuery->execMode = QUERY_EXEC_MODE_SCHEDULE;
+      pQuery->msgType = TDMT_VND_DELETE;
+      break;
     case QUERY_NODE_VNODE_MODIF_STMT:
       pQuery->execMode = QUERY_EXEC_MODE_SCHEDULE;
       pQuery->msgType = toMsgType(((SVnodeModifOpStmt*)pQuery->pRoot)->sqlNodeType);

source/libs/parser/test/mockCatalog.cpp

@@ -207,6 +207,13 @@ int32_t __catalogGetUdfInfo(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps
   return g_mockCatalogService->catalogGetUdfInfo(funcName, pInfo);
 }
 
+int32_t __catalogRefreshGetTableMeta(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps,
+                                     const SName* pTableName, STableMeta** pTableMeta, int32_t isSTable) {
+  return g_mockCatalogService->catalogGetTableMeta(pTableName, pTableMeta);
+}
+
+int32_t __catalogRemoveTableMeta(SCatalog* pCtg, SName* pTableName) { return 0; }
+
 void initMetaDataEnv() {
   g_mockCatalogService.reset(new MockCatalogService());
 
@@ -221,6 +228,8 @@ void initMetaDataEnv() {
   stub.set(catalogGetDBCfg, __catalogGetDBCfg);
   stub.set(catalogChkAuth, __catalogChkAuth);
   stub.set(catalogGetUdfInfo, __catalogGetUdfInfo);
+  stub.set(catalogRefreshGetTableMeta, __catalogRefreshGetTableMeta);
+  stub.set(catalogRemoveTableMeta, __catalogRemoveTableMeta);
   // {
   //   AddrAny any("libcatalog.so");
   //   std::map<std::string,void*> result;

source/libs/parser/test/parInitialDTest.cpp

@@ -21,7 +21,7 @@ namespace ParserTest {
 
 class ParserInitialDTest : public ParserDdlTest {};
 
-// DELETE FROM tb_name [WHERE condition]
+// DELETE FROM table_name [WHERE condition]
 TEST_F(ParserInitialDTest, delete) {
   useDb("root", "test");
 
@@ -40,7 +40,15 @@ TEST_F(ParserInitialDTest, deleteSemanticCheck) {
   run("DELETE FROM t1 WHERE c1 > 10", TSDB_CODE_PAR_INVALID_DELETE_WHERE, PARSER_STAGE_TRANSLATE);
 }
 
-// todo desc
+// DESC table_name
+TEST_F(ParserInitialDTest, describe) {
+  useDb("root", "test");
+
+  run("DESC t1");
+
+  run("DESCRIBE st1");
+}
+
 // todo describe
 // todo DROP account
 
@@ -51,7 +59,7 @@ TEST_F(ParserInitialDTest, dropBnode) {
 }
 
 // DROP CONSUMER GROUP [ IF EXISTS ] cgroup_name ON topic_name
-TEST_F(ParserInitialDTest, dropCGroup) {
+TEST_F(ParserInitialDTest, dropConsumerGroup) {
   useDb("root", "test");
 
   SMDropCgroupReq expect = {0};

source/libs/parser/test/parTestUtil.h

@@ -50,11 +50,13 @@ class ParserDdlTest : public ParserTestBase {
 
   virtual void checkDdl(const SQuery* pQuery, ParserStage stage) {
     ASSERT_NE(pQuery, nullptr);
-    ASSERT_EQ(pQuery->haveResultSet, false);
     ASSERT_NE(pQuery->pRoot, nullptr);
+    if (QUERY_EXEC_MODE_RPC == pQuery->execMode) {
+      ASSERT_EQ(pQuery->haveResultSet, false);
       ASSERT_EQ(pQuery->numOfResCols, 0);
       ASSERT_EQ(pQuery->pResSchema, nullptr);
       ASSERT_EQ(pQuery->precision, 0);
+    }
     if (nullptr != checkDdl_) {
       checkDdl_(pQuery, stage);
     }

source/libs/planner/src/planLogicCreater.c

@@ -219,9 +219,9 @@ static int32_t makeScanLogicNode(SLogicPlanContext* pCxt, SRealTableNode* pRealT
     return TSDB_CODE_OUT_OF_MEMORY;
   }
 
-  // TSWAP(pScan->pMeta, pRealTable->pMeta);
   TSWAP(pScan->pVgroupList, pRealTable->pVgroupList);
   pScan->tableId = pRealTable->pMeta->uid;
+  pScan->stableId = pRealTable->pMeta->suid;
   pScan->tableType = pRealTable->pMeta->tableType;
   pScan->scanSeq[0] = hasRepeatScanFuncs ? 2 : 1;
   pScan->scanSeq[1] = 0;
@@ -456,6 +456,37 @@ static int32_t createAggLogicNode(SLogicPlanContext* pCxt, SSelectStmt* pSelect,
   return code;
 }
 
+static int32_t createIndefRowsFuncLogicNode(SLogicPlanContext* pCxt, SSelectStmt* pSelect, SLogicNode** pLogicNode) {
+  // top/bottom are both an aggregate function and a indefinite rows function
+  if (!pSelect->hasIndefiniteRowsFunc || pSelect->hasAggFuncs || NULL != pSelect->pWindow) {
+    return TSDB_CODE_SUCCESS;
+  }
+
+  SIndefRowsFuncLogicNode* pIdfRowsFunc =
+      (SIndefRowsFuncLogicNode*)nodesMakeNode(QUERY_NODE_LOGIC_PLAN_INDEF_ROWS_FUNC);
+  if (NULL == pIdfRowsFunc) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  int32_t code = nodesCollectFuncs(pSelect, SQL_CLAUSE_SELECT, fmIsVectorFunc, &pIdfRowsFunc->pVectorFuncs);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = rewriteExprForSelect(pIdfRowsFunc->pVectorFuncs, pSelect, SQL_CLAUSE_SELECT);
+  }
+
+  // set the output
+  if (TSDB_CODE_SUCCESS == code) {
+    code = createColumnByRewriteExps(pIdfRowsFunc->pVectorFuncs, &pIdfRowsFunc->node.pTargets);
+  }
+
+  if (TSDB_CODE_SUCCESS == code) {
+    *pLogicNode = (SLogicNode*)pIdfRowsFunc;
+  } else {
+    nodesDestroyNode(pIdfRowsFunc);
+  }
+
+  return code;
+}
+
 static int32_t createWindowLogicNodeFinalize(SLogicPlanContext* pCxt, SSelectStmt* pSelect, SWindowLogicNode* pWindow,
                                              SLogicNode** pLogicNode) {
   int32_t code = nodesCollectFuncs(pSelect, SQL_CLAUSE_WINDOW, fmIsWindowClauseFunc, &pWindow->pFuncs);
@@ -772,6 +803,9 @@ static int32_t createSelectLogicNode(SLogicPlanContext* pCxt, SSelectStmt* pSele
   if (TSDB_CODE_SUCCESS == code) {
     code = createSelectRootLogicNode(pCxt, pSelect, createAggLogicNode, &pRoot);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = createSelectRootLogicNode(pCxt, pSelect, createIndefRowsFuncLogicNode, &pRoot);
+  }
   if (TSDB_CODE_SUCCESS == code) {
     code = createSelectRootLogicNode(pCxt, pSelect, createDistinctLogicNode, &pRoot);
   }
@@ -1057,8 +1091,8 @@ static int32_t createVnodeModifLogicNodeByDelete(SLogicPlanContext* pCxt, SDelet
   snprintf(pModify->tableFName, sizeof(pModify->tableFName), "%d.%s.%s", pCxt->pPlanCxt->acctId,
            pRealTable->table.dbName, pRealTable->table.tableName);
   pModify->deleteTimeRange = pDelete->timeRange;
-  pModify->pModifyRows = nodesCloneNode(pDelete->pCountFunc);
+  pModify->pAffectedRows = nodesCloneNode(pDelete->pCountFunc);
-  if (NULL == pModify->pModifyRows) {
+  if (NULL == pModify->pAffectedRows) {
     nodesDestroyNode(pModify);
     return TSDB_CODE_OUT_OF_MEMORY;
   }

source/libs/planner/src/planPhysiCreater.c

@@ -425,6 +425,7 @@ static int32_t createScanPhysiNodeFinalize(SPhysiPlanContext* pCxt, SSubplan* pS
 
   if (TSDB_CODE_SUCCESS == code) {
     pScanPhysiNode->uid = pScanLogicNode->tableId;
+    pScanPhysiNode->suid = pScanLogicNode->stableId;
     pScanPhysiNode->tableType = pScanLogicNode->tableType;
     memcpy(&pScanPhysiNode->tableName, &pScanLogicNode->tableName, sizeof(SName));
     if (NULL != pScanLogicNode->pTagCond) {
@@ -790,6 +791,43 @@ static int32_t createAggPhysiNode(SPhysiPlanContext* pCxt, SNodeList* pChildren,
   return code;
 }
 
+static int32_t createIndefRowsFuncPhysiNode(SPhysiPlanContext* pCxt, SNodeList* pChildren,
+                                            SIndefRowsFuncLogicNode* pFuncLogicNode, SPhysiNode** pPhyNode) {
+  SIndefRowsFuncPhysiNode* pIdfRowsFunc = (SIndefRowsFuncPhysiNode*)makePhysiNode(
+      pCxt, (SLogicNode*)pFuncLogicNode, QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC);
+  if (NULL == pIdfRowsFunc) {
+    return TSDB_CODE_OUT_OF_MEMORY;
+  }
+
+  SNodeList* pPrecalcExprs = NULL;
+  SNodeList* pVectorFuncs = NULL;
+  int32_t    code = rewritePrecalcExprs(pCxt, pFuncLogicNode->pVectorFuncs, &pPrecalcExprs, &pVectorFuncs);
+
+  SDataBlockDescNode* pChildTupe = (((SPhysiNode*)nodesListGetNode(pChildren, 0))->pOutputDataBlockDesc);
+  // push down expression to pOutputDataBlockDesc of child node
+  if (TSDB_CODE_SUCCESS == code && NULL != pPrecalcExprs) {
+    code = setListSlotId(pCxt, pChildTupe->dataBlockId, -1, pPrecalcExprs, &pIdfRowsFunc->pExprs);
+    if (TSDB_CODE_SUCCESS == code) {
+      code = pushdownDataBlockSlots(pCxt, pIdfRowsFunc->pExprs, pChildTupe);
+    }
+  }
+
+  if (TSDB_CODE_SUCCESS == code && NULL != pVectorFuncs) {
+    code = setListSlotId(pCxt, pChildTupe->dataBlockId, -1, pVectorFuncs, &pIdfRowsFunc->pVectorFuncs);
+    if (TSDB_CODE_SUCCESS == code) {
+      code = addDataBlockSlots(pCxt, pIdfRowsFunc->pVectorFuncs, pIdfRowsFunc->node.pOutputDataBlockDesc);
+    }
+  }
+
+  if (TSDB_CODE_SUCCESS == code) {
+    *pPhyNode = (SPhysiNode*)pIdfRowsFunc;
+  } else {
+    nodesDestroyNode(pIdfRowsFunc);
+  }
+
+  return code;
+}
+
 static int32_t createProjectPhysiNode(SPhysiPlanContext* pCxt, SNodeList* pChildren,
                                       SProjectLogicNode* pProjectLogicNode, SPhysiNode** pPhyNode) {
   SProjectPhysiNode* pProject =
@@ -923,8 +961,8 @@ static ENodeType getIntervalOperatorType(EIntervalAlgorithm intervalAlgo) {
   switch (intervalAlgo) {
     case INTERVAL_ALGO_HASH:
       return QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL;
-    case INTERVAL_ALGO_SORT_MERGE:
+    case INTERVAL_ALGO_MERGE:
-      return QUERY_NODE_PHYSICAL_PLAN_SORT_MERGE_INTERVAL;
+      return QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL;
     case INTERVAL_ALGO_STREAM_FINAL:
       return QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_INTERVAL;
     case INTERVAL_ALGO_STREAM_SEMI:
@@ -1155,6 +1193,8 @@ static int32_t createExchangePhysiNodeByMerge(SMergePhysiNode* pMerge) {
     nodesDestroyNode(pExchange);
     return TSDB_CODE_OUT_OF_MEMORY;
   }
+  SNode* pSlot = NULL;
+  FOREACH(pSlot, pExchange->node.pOutputDataBlockDesc->pSlots) { ((SSlotDescNode*)pSlot)->output = true; }
   return nodesListMakeStrictAppend(&pMerge->node.pChildren, pExchange);
 }
 
@@ -1168,20 +1208,30 @@ static int32_t createMergePhysiNode(SPhysiPlanContext* pCxt, SMergeLogicNode* pM
   pMerge->numOfChannels = pMergeLogicNode->numOfChannels;
   pMerge->srcGroupId = pMergeLogicNode->srcGroupId;
 
-  int32_t code = TSDB_CODE_SUCCESS;
+  int32_t code = addDataBlockSlots(pCxt, pMergeLogicNode->pInputs, pMerge->node.pOutputDataBlockDesc);
 
+  if (TSDB_CODE_SUCCESS == code) {
     for (int32_t i = 0; i < pMerge->numOfChannels; ++i) {
       code = createExchangePhysiNodeByMerge(pMerge);
       if (TSDB_CODE_SUCCESS != code) {
         break;
       }
     }
+  }
 
   if (TSDB_CODE_SUCCESS == code) {
     code = setListSlotId(pCxt, pMerge->node.pOutputDataBlockDesc->dataBlockId, -1, pMergeLogicNode->pMergeKeys,
                          &pMerge->pMergeKeys);
   }
 
+  if (TSDB_CODE_SUCCESS == code) {
+    code = setListSlotId(pCxt, pMerge->node.pOutputDataBlockDesc->dataBlockId, -1, pMergeLogicNode->node.pTargets,
+                         &pMerge->pTargets);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = addDataBlockSlots(pCxt, pMerge->pTargets, pMerge->node.pOutputDataBlockDesc);
+  }
+
   if (TSDB_CODE_SUCCESS == code) {
     *pPhyNode = (SPhysiNode*)pMerge;
   } else {
@@ -1212,6 +1262,8 @@ static int32_t doCreatePhysiNode(SPhysiPlanContext* pCxt, SLogicNode* pLogicNode
       return createPartitionPhysiNode(pCxt, pChildren, (SPartitionLogicNode*)pLogicNode, pPhyNode);
     case QUERY_NODE_LOGIC_PLAN_FILL:
       return createFillPhysiNode(pCxt, pChildren, (SFillLogicNode*)pLogicNode, pPhyNode);
+    case QUERY_NODE_LOGIC_PLAN_INDEF_ROWS_FUNC:
+      return createIndefRowsFuncPhysiNode(pCxt, pChildren, (SIndefRowsFuncLogicNode*)pLogicNode, pPhyNode);
     case QUERY_NODE_LOGIC_PLAN_MERGE:
       return createMergePhysiNode(pCxt, (SMergeLogicNode*)pLogicNode, pPhyNode);
     default:
@@ -1319,13 +1371,21 @@ static int32_t createDataDeleter(SPhysiPlanContext* pCxt, SVnodeModifyLogicNode*
   strcpy(pDeleter->tableFName, pModify->tableFName);
   pDeleter->deleteTimeRange = pModify->deleteTimeRange;
 
+  int32_t code = setNodeSlotId(pCxt, pRoot->pOutputDataBlockDesc->dataBlockId, -1, pModify->pAffectedRows,
+                               &pDeleter->pAffectedRows);
+  if (TSDB_CODE_SUCCESS == code) {
     pDeleter->sink.pInputDataBlockDesc = nodesCloneNode(pRoot->pOutputDataBlockDesc);
     if (NULL == pDeleter->sink.pInputDataBlockDesc) {
-    nodesDestroyNode(pDeleter);
+      code = TSDB_CODE_OUT_OF_MEMORY;
-    return TSDB_CODE_OUT_OF_MEMORY;
+    }
   }
 
+  if (TSDB_CODE_SUCCESS == code) {
     *pSink = (SDataSinkNode*)pDeleter;
+  } else {
+    nodesDestroyNode(pDeleter);
+  }
+
   return TSDB_CODE_SUCCESS;
 }
 
@@ -1335,6 +1395,7 @@ static int32_t buildDeleteSubplan(SPhysiPlanContext* pCxt, SVnodeModifyLogicNode
   if (TSDB_CODE_SUCCESS == code) {
     code = createDataDeleter(pCxt, pModify, pSubplan->pNode, &pSubplan->pDataSink);
   }
+  pSubplan->msgType = TDMT_VND_DELETE;
   return code;
 }
 

source/libs/planner/src/planSpliter.c

@@ -80,30 +80,36 @@ static int32_t splCreateExchangeNode(SSplitContext* pCxt, SLogicNode* pChild, SE
   return TSDB_CODE_SUCCESS;
 }
 
-static int32_t splCreateExchangeNodeForSubplan(SSplitContext* pCxt, SLogicSubplan* pSubplan, SLogicNode* pSplitNode,
+static int32_t splReplaceLogicNode(SLogicSubplan* pSubplan, SLogicNode* pOld, SLogicNode* pNew) {
-                                               ESubplanType subplanType) {
+  if (NULL == pOld->pParent) {
-  SExchangeLogicNode* pExchange = NULL;
+    pSubplan->pNode = (SLogicNode*)pNew;
-  if (TSDB_CODE_SUCCESS != splCreateExchangeNode(pCxt, pSplitNode, &pExchange)) {
-    return TSDB_CODE_OUT_OF_MEMORY;
-  }
-
-  pSubplan->subplanType = subplanType;
-
-  if (NULL == pSplitNode->pParent) {
-    pSubplan->pNode = (SLogicNode*)pExchange;
     return TSDB_CODE_SUCCESS;
   }
 
   SNode* pNode;
-  FOREACH(pNode, pSplitNode->pParent->pChildren) {
+  FOREACH(pNode, pOld->pParent->pChildren) {
-    if (nodesEqualNode(pNode, pSplitNode)) {
+    if (nodesEqualNode(pNode, pOld)) {
-      REPLACE_NODE(pExchange);
+      REPLACE_NODE(pNew);
-      pExchange->node.pParent = pSplitNode->pParent;
+      pNew->pParent = pOld->pParent;
       return TSDB_CODE_SUCCESS;
     }
   }
+  return TSDB_CODE_PLAN_INTERNAL_ERROR;
+}
+
+static int32_t splCreateExchangeNodeForSubplan(SSplitContext* pCxt, SLogicSubplan* pSubplan, SLogicNode* pSplitNode,
+                                               ESubplanType subplanType) {
+  SExchangeLogicNode* pExchange = NULL;
+  int32_t             code = splCreateExchangeNode(pCxt, pSplitNode, &pExchange);
+  if (TSDB_CODE_SUCCESS == code) {
+    code = splReplaceLogicNode(pSubplan, pSplitNode, (SLogicNode*)pExchange);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    pSubplan->subplanType = subplanType;
+  } else {
     nodesDestroyNode(pExchange);
-  return TSDB_CODE_FAILED;
+  }
+  return code;
 }
 
 static bool splMatch(SSplitContext* pCxt, SLogicSubplan* pSubplan, int32_t flag, FSplFindSplitNode func, void* pInfo) {
@@ -295,24 +301,34 @@ static int32_t stbSplCreatePartWindowNode(SWindowLogicNode* pMergeWindow, SLogic
   return code;
 }
 
-static int32_t stbSplCreateMergeNode(SSplitContext* pCxt, SLogicNode* pParent, SNodeList* pMergeKeys,
+static int32_t stbSplCreateMergeNode(SSplitContext* pCxt, SLogicSubplan* pSubplan, SLogicNode* pSplitNode,
-                                     SLogicNode* pPartChild) {
+                                     SNodeList* pMergeKeys, SLogicNode* pPartChild) {
   SMergeLogicNode* pMerge = nodesMakeNode(QUERY_NODE_LOGIC_PLAN_MERGE);
   if (NULL == pMerge) {
     return TSDB_CODE_OUT_OF_MEMORY;
   }
   pMerge->numOfChannels = ((SScanLogicNode*)nodesListGetNode(pPartChild->pChildren, 0))->pVgroupList->numOfVgroups;
   pMerge->srcGroupId = pCxt->groupId;
-  pMerge->node.pParent = pParent;
   pMerge->node.precision = pPartChild->precision;
   pMerge->pMergeKeys = pMergeKeys;
-  pMerge->node.pTargets = nodesCloneList(pPartChild->pTargets);
-  if (NULL == pMerge->node.pTargets) {
-    nodesDestroyNode(pMerge);
-    return TSDB_CODE_OUT_OF_MEMORY;
-  }
 
-  return nodesListMakeAppend(&pParent->pChildren, pMerge);
+  int32_t code = TSDB_CODE_SUCCESS;
+  pMerge->pInputs = nodesCloneList(pPartChild->pTargets);
+  pMerge->node.pTargets = nodesCloneList(pSplitNode->pTargets);
+  if (NULL == pMerge->node.pTargets || NULL == pMerge->pInputs) {
+    code = TSDB_CODE_OUT_OF_MEMORY;
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    if (NULL == pSubplan) {
+      code = nodesListMakeAppend(&pSplitNode->pChildren, pMerge);
+    } else {
+      code = splReplaceLogicNode(pSubplan, pSplitNode, (SLogicNode*)pMerge);
+    }
+  }
+  if (TSDB_CODE_SUCCESS != code) {
+    nodesDestroyNode(pMerge);
+  }
+  return code;
 }
 
 static int32_t stbSplCreateExchangeNode(SSplitContext* pCxt, SLogicNode* pParent, SLogicNode* pPartChild) {
@@ -329,8 +345,15 @@ static int32_t stbSplSplitWindowNodeForBatch(SSplitContext* pCxt, SStableSplitIn
   int32_t     code = stbSplCreatePartWindowNode((SWindowLogicNode*)pInfo->pSplitNode, &pPartWindow);
   if (TSDB_CODE_SUCCESS == code) {
     ((SWindowLogicNode*)pPartWindow)->intervalAlgo = INTERVAL_ALGO_HASH;
-    ((SWindowLogicNode*)pInfo->pSplitNode)->intervalAlgo = INTERVAL_ALGO_SORT_MERGE;
+    ((SWindowLogicNode*)pInfo->pSplitNode)->intervalAlgo = INTERVAL_ALGO_MERGE;
-    code = stbSplCreateExchangeNode(pCxt, pInfo->pSplitNode, pPartWindow);
+    SNodeList* pMergeKeys = NULL;
+    code = nodesListMakeStrictAppend(&pMergeKeys, nodesCloneNode(((SWindowLogicNode*)pInfo->pSplitNode)->pTspk));
+    if (TSDB_CODE_SUCCESS == code) {
+      code = stbSplCreateMergeNode(pCxt, NULL, pInfo->pSplitNode, pMergeKeys, pPartWindow);
+    }
+    if (TSDB_CODE_SUCCESS != code) {
+      nodesDestroyList(pMergeKeys);
+    }
   }
   if (TSDB_CODE_SUCCESS == code) {
     code = nodesListMakeStrictAppend(&pInfo->pSubplan->pChildren,
@@ -424,37 +447,99 @@ static int32_t stbSplSplitAggNode(SSplitContext* pCxt, SStableSplitInfo* pInfo)
   return code;
 }
 
-static int32_t stbSplCreatePartSortNode(SSortLogicNode* pMergeSort, SLogicNode** pOutput) {
+static SNode* stbSplCreateColumnNode(SExprNode* pExpr) {
-  SNodeList* pSortKeys = pMergeSort->pSortKeys;
+  SColumnNode* pCol = nodesMakeNode(QUERY_NODE_COLUMN);
-  pMergeSort->pSortKeys = NULL;
+  if (NULL == pCol) {
-  SNodeList* pTargets = pMergeSort->node.pTargets;
+    return NULL;
-  pMergeSort->node.pTargets = NULL;
+  }
-  SNodeList* pChildren = pMergeSort->node.pChildren;
+  if (QUERY_NODE_COLUMN == nodeType(pExpr)) {
-  pMergeSort->node.pChildren = NULL;
+    strcpy(pCol->tableAlias, ((SColumnNode*)pExpr)->tableAlias);
+  }
+  strcpy(pCol->colName, pExpr->aliasName);
+  strcpy(pCol->node.aliasName, pExpr->aliasName);
+  pCol->node.resType = pExpr->resType;
+  return (SNode*)pCol;
+}
+
+static SNode* stbSplCreateOrderByExpr(SOrderByExprNode* pSortKey, SNode* pCol) {
+  SOrderByExprNode* pOutput = nodesMakeNode(QUERY_NODE_ORDER_BY_EXPR);
+  if (NULL == pOutput) {
+    return NULL;
+  }
+  pOutput->pExpr = nodesCloneNode(pCol);
+  if (NULL == pOutput->pExpr) {
+    nodesDestroyNode(pOutput);
+    return NULL;
+  }
+  pOutput->order = pSortKey->order;
+  pOutput->nullOrder = pSortKey->nullOrder;
+  return (SNode*)pOutput;
+}
+
+static int32_t stbSplCreateMergeKeys(SNodeList* pSortKeys, SNodeList* pTargets, SNodeList** pOutput) {
+  int32_t    code = TSDB_CODE_SUCCESS;
+  SNodeList* pMergeKeys = NULL;
+  SNode*     pNode = NULL;
+  FOREACH(pNode, pSortKeys) {
+    SOrderByExprNode* pSortKey = (SOrderByExprNode*)pNode;
+    SNode*            pTarget = NULL;
+    bool              found = false;
+    FOREACH(pTarget, pTargets) {
+      if (0 == strcmp(((SExprNode*)pSortKey->pExpr)->aliasName, ((SColumnNode*)pTarget)->colName)) {
+        code = nodesListMakeStrictAppend(&pMergeKeys, stbSplCreateOrderByExpr(pSortKey, pTarget));
+        if (TSDB_CODE_SUCCESS != code) {
+          break;
+        }
+        found = true;
+      }
+    }
+    if (TSDB_CODE_SUCCESS == code && !found) {
+      SNode* pCol = stbSplCreateColumnNode((SExprNode*)pSortKey->pExpr);
+      code = nodesListMakeStrictAppend(&pMergeKeys, stbSplCreateOrderByExpr(pSortKey, pCol));
+      if (TSDB_CODE_SUCCESS == code) {
+        code = nodesListStrictAppend(pTargets, pCol);
+      } else {
+        nodesDestroyNode(pCol);
+      }
+    }
+    if (TSDB_CODE_SUCCESS != code) {
+      break;
+    }
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    *pOutput = pMergeKeys;
+  } else {
+    nodesDestroyList(pMergeKeys);
+  }
+  return code;
+}
+
+static int32_t stbSplCreatePartSortNode(SSortLogicNode* pSort, SLogicNode** pOutputPartSort,
+                                        SNodeList** pOutputMergeKeys) {
+  SNodeList* pSortKeys = pSort->pSortKeys;
+  pSort->pSortKeys = NULL;
+  SNodeList* pChildren = pSort->node.pChildren;
+  pSort->node.pChildren = NULL;
 
   int32_t         code = TSDB_CODE_SUCCESS;
-  SSortLogicNode* pPartSort = nodesCloneNode(pMergeSort);
+  SSortLogicNode* pPartSort = nodesCloneNode(pSort);
   if (NULL == pPartSort) {
     code = TSDB_CODE_OUT_OF_MEMORY;
   }
 
-  pMergeSort->node.pTargets = pTargets;
+  SNodeList* pMergeKeys = NULL;
+  if (TSDB_CODE_SUCCESS == code) {
     pPartSort->node.pChildren = pChildren;
-  if (TSDB_CODE_SUCCESS == code) {
     pPartSort->pSortKeys = pSortKeys;
-    code = createColumnByRewriteExps(pPartSort->pSortKeys, &pPartSort->node.pTargets);
+    code = stbSplCreateMergeKeys(pPartSort->pSortKeys, pPartSort->node.pTargets, &pMergeKeys);
-  }
-  if (TSDB_CODE_SUCCESS == code) {
-    pMergeSort->pSortKeys = nodesCloneList(pPartSort->node.pTargets);
-    if (NULL == pMergeSort->pSortKeys) {
-      code = TSDB_CODE_OUT_OF_MEMORY;
-    }
   }
 
   if (TSDB_CODE_SUCCESS == code) {
-    *pOutput = (SLogicNode*)pPartSort;
+    *pOutputPartSort = (SLogicNode*)pPartSort;
+    *pOutputMergeKeys = pMergeKeys;
   } else {
     nodesDestroyNode(pPartSort);
+    nodesDestroyList(pMergeKeys);
   }
 
   return code;
@@ -462,17 +547,10 @@ static int32_t stbSplCreatePartSortNode(SSortLogicNode* pMergeSort, SLogicNode**
 
 static int32_t stbSplSplitSortNode(SSplitContext* pCxt, SStableSplitInfo* pInfo) {
   SLogicNode* pPartSort = NULL;
-  int32_t     code = stbSplCreatePartSortNode((SSortLogicNode*)pInfo->pSplitNode, &pPartSort);
+  SNodeList*  pMergeKeys = NULL;
+  int32_t     code = stbSplCreatePartSortNode((SSortLogicNode*)pInfo->pSplitNode, &pPartSort, &pMergeKeys);
   if (TSDB_CODE_SUCCESS == code) {
-    SNodeList* pMergeKeys = nodesCloneList(((SSortLogicNode*)pInfo->pSplitNode)->pSortKeys);
+    code = stbSplCreateMergeNode(pCxt, pInfo->pSubplan, pInfo->pSplitNode, pMergeKeys, pPartSort);
-    if (NULL != pMergeKeys) {
-      code = stbSplCreateMergeNode(pCxt, pInfo->pSplitNode, pMergeKeys, pPartSort);
-      if (TSDB_CODE_SUCCESS != code) {
-        nodesDestroyList(pMergeKeys);
-      }
-    } else {
-      code = TSDB_CODE_OUT_OF_MEMORY;
-    }
   }
   if (TSDB_CODE_SUCCESS == code) {
     code = nodesListMakeStrictAppend(&pInfo->pSubplan->pChildren,

source/libs/planner/src/planner.c

@@ -88,7 +88,7 @@ int32_t qSetSubplanExecutionNode(SSubplan* subplan, int32_t groupId, SDownstream
 }
 
 int32_t qSubPlanToString(const SSubplan* pSubplan, char** pStr, int32_t* pLen) {
-  if (SUBPLAN_TYPE_MODIFY == pSubplan->subplanType) {
+  if (SUBPLAN_TYPE_MODIFY == pSubplan->subplanType && NULL == pSubplan->pNode) {
     SDataInserterNode* insert = (SDataInserterNode*)pSubplan->pDataSink;
     *pLen = insert->size;
     *pStr = insert->pData;

source/libs/planner/test/planIntervalTest.cpp

@@ -50,6 +50,8 @@ TEST_F(PlanIntervalTest, selectFunc) {
   run("SELECT MAX(c1), MIN(c1) FROM t1 INTERVAL(10s)");
   // select function along with the columns of select row, and with INTERVAL clause
   run("SELECT MAX(c1), c2 FROM t1 INTERVAL(10s)");
+
+  run("SELECT TOP(c1, 1) FROM t1 INTERVAL(10s) ORDER BY c1");
 }
 
 TEST_F(PlanIntervalTest, stable) {

source/libs/planner/test/planOrderByTest.cpp

@@ -46,4 +46,7 @@ TEST_F(PlanOrderByTest, stable) {
 
   // ORDER BY key is in the projection list
   run("SELECT c1 FROM st1 ORDER BY c1");
+
+  // ORDER BY key is not in the projection list
+  run("SELECT c2 FROM st1 ORDER BY c1");
 }

source/libs/planner/test/planProjectTest.cpp

@@ -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 "planTestUtil.h"
+
+using namespace std;
+
+class PlanProjectTest : public PlannerTestBase {};
+
+TEST_F(PlanProjectTest, basic) {
+  useDb("root", "test");
+
+  run("SELECT CEIL(c1) FROM t1");
+}
+
+TEST_F(PlanProjectTest, indefiniteRowsFunc) {
+  useDb("root", "test");
+
+  run("SELECT MAVG(c1, 10) FROM t1");
+
+  run("SELECT MAVG(CEIL(c1), 20) + 2 FROM t1");
+}

source/libs/planner/test/planTestUtil.cpp

@@ -81,6 +81,8 @@ int32_t getLogLevel() { return g_logLevel; }
 
 class PlannerTestBaseImpl {
  public:
+  PlannerTestBaseImpl() : sqlNo_(0) {}
+
   void useDb(const string& acctId, const string& db) {
     caseEnv_.acctId_ = acctId;
     caseEnv_.db_ = db;
@@ -88,6 +90,7 @@ class PlannerTestBaseImpl {
   }
 
   void run(const string& sql) {
+    ++sqlNo_;
     if (caseEnv_.nsql_ > 0) {
       --(caseEnv_.nsql_);
       return;
@@ -187,6 +190,8 @@ class PlannerTestBaseImpl {
     string  acctId_;
     string  db_;
     int32_t nsql_;
+
+    caseEnv() : nsql_(0) {}
   };
 
   struct stmtEnv {
@@ -194,6 +199,7 @@ class PlannerTestBaseImpl {
     array<char, 1024> msgBuf_;
     SQuery*           pQuery_;
 
+    stmtEnv() : pQuery_(nullptr) {}
     ~stmtEnv() { qDestroyQuery(pQuery_); }
   };
 
@@ -229,7 +235,7 @@ class PlannerTestBaseImpl {
       return;
     }
 
-    cout << "==========================================sql : [" << stmtEnv_.sql_ << "]" << endl;
+    cout << "========================================== " << sqlNo_ << " sql : [" << stmtEnv_.sql_ << "]" << endl;
 
     if (DUMP_MODULE_ALL == module || DUMP_MODULE_PARSER == module) {
       if (res_.prepareAst_.empty()) {
@@ -382,6 +388,7 @@ class PlannerTestBaseImpl {
   caseEnv caseEnv_;
   stmtEnv stmtEnv_;
   stmtRes res_;
+  int32_t sqlNo_;
 };
 
 PlannerTestBase::PlannerTestBase() : impl_(new PlannerTestBaseImpl()) {}

source/libs/qcom/src/querymsg.c

@@ -37,6 +37,8 @@ int32_t queryBuildUseDbOutput(SUseDbOutput *pOut, SUseDbRsp *usedbRsp) {
   pOut->dbVgroup->vgVersion = usedbRsp->vgVersion;
   pOut->dbVgroup->hashMethod = usedbRsp->hashMethod;
 
+  qDebug("Got %d vgroup for db %s", usedbRsp->vgNum, usedbRsp->db);
+
   if (usedbRsp->vgNum <= 0) {
     return TSDB_CODE_SUCCESS;
   }
@@ -50,6 +52,8 @@ int32_t queryBuildUseDbOutput(SUseDbOutput *pOut, SUseDbRsp *usedbRsp) {
   for (int32_t i = 0; i < usedbRsp->vgNum; ++i) {
     SVgroupInfo *pVgInfo = taosArrayGet(usedbRsp->pVgroupInfos, i);
     pOut->dbVgroup->numOfTable += pVgInfo->numOfTable;
+    qDebug("the %dth vgroup, id %d, epNum %d, current %s port %d", i, pVgInfo->vgId, pVgInfo->epSet.numOfEps,
+      pVgInfo->epSet.eps[pVgInfo->epSet.inUse].fqdn, pVgInfo->epSet.eps[pVgInfo->epSet.inUse].port);
     if (0 != taosHashPut(pOut->dbVgroup->vgHash, &pVgInfo->vgId, sizeof(int32_t), pVgInfo, sizeof(SVgroupInfo))) {
       return TSDB_CODE_TSC_OUT_OF_MEMORY;
     }

source/libs/qworker/inc/qwInt.h

@@ -160,6 +160,7 @@ typedef struct SQWMsgStat {
   uint64_t cancelProcessed;
   uint64_t dropProcessed;
   uint64_t hbProcessed;
+  uint64_t deleteProcessed;
 } SQWMsgStat;
 
 typedef struct SQWRTStat {
@@ -357,6 +358,7 @@ int32_t qwUpdateTimeInQueue(SQWorker *mgmt, int64_t ts, EQueueType type);
 int64_t qwGetTimeInQueue(SQWorker *mgmt, EQueueType type);
 void qwClearExpiredSch(SArray* pExpiredSch);
 int32_t qwAcquireScheduler(SQWorker *mgmt, uint64_t sId, int32_t rwType, SQWSchStatus **sch);
+void qwFreeTaskCtx(QW_FPARAMS_DEF, SQWTaskCtx *ctx);
 
 void qwDbgDumpMgmtInfo(SQWorker *mgmt);
 int32_t qwDbgValidateStatus(QW_FPARAMS_DEF, int8_t oriStatus, int8_t newStatus, bool *ignore);

source/libs/qworker/inc/qwMsg.h

@@ -30,6 +30,7 @@ int32_t qwProcessReady(QW_FPARAMS_DEF, SQWMsg *qwMsg);
 int32_t qwProcessFetch(QW_FPARAMS_DEF, SQWMsg *qwMsg);
 int32_t qwProcessDrop(QW_FPARAMS_DEF, SQWMsg *qwMsg);
 int32_t qwProcessHb(SQWorker *mgmt, SQWMsg *qwMsg, SSchedulerHbReq *req);
+int32_t qwProcessDelete(QW_FPARAMS_DEF, SQWMsg *qwMsg, SRpcMsg *pRsp, SDeleteRes *pRes);
 
 int32_t qwBuildAndSendDropRsp(SRpcHandleInfo *pConn, int32_t code);
 int32_t qwBuildAndSendCancelRsp(SRpcHandleInfo *pConn, int32_t code);

source/libs/qworker/src/qwMsg.c

@@ -300,13 +300,6 @@ int32_t qWorkerProcessQueryMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, int
     QW_ERR_RET(TSDB_CODE_QRY_INVALID_INPUT);
   }
 
-  msg->sId = msg->sId;
-  msg->queryId = msg->queryId;
-  msg->taskId = msg->taskId;
-  msg->refId = msg->refId;
-  msg->phyLen = msg->phyLen;
-  msg->sqlLen = msg->sqlLen;
-
   uint64_t sId = msg->sId;
   uint64_t qId = msg->queryId;
   uint64_t tId = msg->taskId;
@@ -523,3 +516,37 @@ int32_t qWorkerProcessHbMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, int64_
 
   return TSDB_CODE_SUCCESS;
 }
+
+
+int32_t qWorkerProcessDeleteMsg(void *node, void *qWorkerMgmt, SRpcMsg *pMsg, SRpcMsg *pRsp, SDeleteRes *pRes) {
+  if (NULL == node || NULL == qWorkerMgmt || NULL == pMsg || NULL == pRsp) {
+    QW_ERR_RET(TSDB_CODE_QRY_INVALID_INPUT);
+  }
+
+  int32_t       code = 0;
+  SVDeleteReq req = {0};
+  SQWorker *    mgmt = (SQWorker *)qWorkerMgmt;
+
+  QW_STAT_INC(mgmt->stat.msgStat.deleteProcessed, 1);
+
+  tDeserializeSVDeleteReq(pMsg->pCont, pMsg->contLen, &req);
+  
+  uint64_t sId = req.sId;
+  uint64_t qId = req.queryId;
+  uint64_t tId = req.taskId;
+  int64_t  rId = 0;
+
+  SQWMsg qwMsg = {.node = node, .msg = req.msg, .msgLen = req.phyLen, .connInfo = pMsg->info};
+  QW_SCH_TASK_DLOG("processDelete start, node:%p, handle:%p, sql:%s", node, pMsg->info.handle, req.sql);
+  taosMemoryFreeClear(req.sql);
+
+  QW_ERR_JRET(qwProcessDelete(QW_FPARAMS(), &qwMsg, pRsp, pRes));
+
+  QW_SCH_TASK_DLOG("processDelete end, node:%p", node);
+
+_return:
+
+  QW_RET(code);
+}
+
+

source/libs/qworker/src/qwUtil.c

@@ -290,8 +290,11 @@ int32_t qwKillTaskHandle(QW_FPARAMS_DEF, SQWTaskCtx *ctx) {
   QW_RET(code);
 }
 
-void qwFreeTask(QW_FPARAMS_DEF, SQWTaskCtx *ctx) {
+void qwFreeTaskCtx(QW_FPARAMS_DEF, SQWTaskCtx *ctx) {
+  if (ctx->ctrlConnInfo.handle) {
     tmsgReleaseHandle(&ctx->ctrlConnInfo, TAOS_CONN_SERVER);
+  }
+  
   ctx->ctrlConnInfo.handle = NULL;
   ctx->ctrlConnInfo.refId = -1;
 
@@ -333,7 +336,7 @@ int32_t qwDropTaskCtx(QW_FPARAMS_DEF) {
     QW_ERR_RET(TSDB_CODE_QRY_TASK_CTX_NOT_EXIST);
   }
 
-  qwFreeTask(QW_FPARAMS(), &octx);
+  qwFreeTaskCtx(QW_FPARAMS(), &octx);
 
   QW_TASK_DLOG_E("task ctx dropped");
 

source/libs/qworker/src/qworker.c

@@ -183,7 +183,7 @@ int32_t qwGenerateSchHbRsp(SQWorker *mgmt, SQWSchStatus *sch, SQWHbInfo *hbInfo)
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t qwGetResFromSink(QW_FPARAMS_DEF, SQWTaskCtx *ctx, int32_t *dataLen, void **rspMsg, SOutputData *pOutput) {
+int32_t qwGetQueryResFromSink(QW_FPARAMS_DEF, SQWTaskCtx *ctx, int32_t *dataLen, void **rspMsg, SOutputData *pOutput) {
   int32_t            len = 0;
   SRetrieveTableRsp *rsp = NULL;
   bool               queryEnd = false;
@@ -242,6 +242,53 @@ int32_t qwGetResFromSink(QW_FPARAMS_DEF, SQWTaskCtx *ctx, int32_t *dataLen, void
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t qwGetDeleteResFromSink(QW_FPARAMS_DEF, SQWTaskCtx *ctx, int32_t *dataLen, void **rspMsg, SDeleteRes *pRes) {
+  int32_t            len = 0;
+  SVDeleteRsp        rsp = {0};
+  bool               queryEnd = false;
+  int32_t            code = 0;
+  SOutputData        output = {0};
+
+  dsGetDataLength(ctx->sinkHandle, &len, &queryEnd);
+
+  if (len <= 0 || len != sizeof(SDeleterRes)) {
+    QW_TASK_ELOG("invalid length from dsGetDataLength, length:%d", len);
+    QW_ERR_RET(TSDB_CODE_QRY_INVALID_INPUT);
+  }
+
+  output.pData = taosMemoryCalloc(1, len);
+  if (NULL == output.pData) {
+    QW_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
+  }
+  
+  code = dsGetDataBlock(ctx->sinkHandle, &output);
+  if (code) {
+    QW_TASK_ELOG("dsGetDataBlock failed, code:%x - %s", code, tstrerror(code));
+    taosMemoryFree(output.pData);
+    QW_ERR_RET(code);
+  }
+
+  SDeleterRes* pDelRes = (SDeleterRes*)output.pData;
+  
+  rsp.affectedRows = pDelRes->affectedRows;
+  pRes->uid = pDelRes->uid;
+  pRes->uidList = pDelRes->uidList;
+  pRes->skey = pDelRes->skey;
+  pRes->ekey = pDelRes->ekey;
+
+  SEncoder coder = {0};  
+  tEncodeSize(tEncodeSVDeleteRsp, &rsp, len, code);
+  void *msg = rpcMallocCont(len);
+  tEncoderInit(&coder, msg, len);
+  tEncodeSVDeleteRsp(&coder, &rsp);
+  tEncoderClear(&coder);
+
+  *rspMsg = msg;
+  *dataLen = len;
+  
+  return TSDB_CODE_SUCCESS;
+}
+
 
 int32_t qwHandlePrePhaseEvents(QW_FPARAMS_DEF, int8_t phase, SQWPhaseInput *input, SQWPhaseOutput *output) {
   int32_t         code = 0;
@@ -547,7 +594,7 @@ int32_t qwProcessCQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg) {
 
     if (QW_IS_EVENT_RECEIVED(ctx, QW_EVENT_FETCH)) {
       SOutputData sOutput = {0};
-      QW_ERR_JRET(qwGetResFromSink(QW_FPARAMS(), ctx, &dataLen, &rsp, &sOutput));
+      QW_ERR_JRET(qwGetQueryResFromSink(QW_FPARAMS(), ctx, &dataLen, &rsp, &sOutput));
 
       if ((!sOutput.queryEnd) && (DS_BUF_LOW == sOutput.bufStatus || DS_BUF_EMPTY == sOutput.bufStatus)) {
         QW_TASK_DLOG("task not end and buf is %s, need to continue query", qwBufStatusStr(sOutput.bufStatus));
@@ -620,7 +667,7 @@ int32_t qwProcessFetch(QW_FPARAMS_DEF, SQWMsg *qwMsg) {
   QW_ERR_JRET(qwGetTaskCtx(QW_FPARAMS(), &ctx));
 
   SOutputData sOutput = {0};
-  QW_ERR_JRET(qwGetResFromSink(QW_FPARAMS(), ctx, &dataLen, &rsp, &sOutput));
+  QW_ERR_JRET(qwGetQueryResFromSink(QW_FPARAMS(), ctx, &dataLen, &rsp, &sOutput));
 
   if (NULL == rsp) {
     ctx->dataConnInfo = qwMsg->connInfo;
@@ -875,6 +922,47 @@ _return:
   qwRelease(refId);
 }
 
+int32_t qwProcessDelete(QW_FPARAMS_DEF, SQWMsg *qwMsg, SRpcMsg *pRsp, SDeleteRes *pRes) {
+  int32_t        code = 0;
+  SSubplan      *plan = NULL;
+  qTaskInfo_t    pTaskInfo = NULL;
+  DataSinkHandle sinkHandle = NULL;
+  SQWTaskCtx     ctx = {0};
+
+  code = qStringToSubplan(qwMsg->msg, &plan);
+  if (TSDB_CODE_SUCCESS != code) {
+    code = TSDB_CODE_INVALID_MSG;
+    QW_TASK_ELOG("task physical plan to subplan failed, code:%x - %s", code, tstrerror(code));
+    QW_ERR_JRET(code);
+  }
+
+  ctx.plan = plan;
+
+  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);
+  }
+
+  if (NULL == sinkHandle || NULL == pTaskInfo) {
+    QW_TASK_ELOG("create task result error, taskHandle:%p, sinkHandle:%p", pTaskInfo, sinkHandle);
+    QW_ERR_JRET(TSDB_CODE_QRY_APP_ERROR);
+  }
+
+  ctx.taskHandle = pTaskInfo;
+  ctx.sinkHandle = sinkHandle;
+
+  QW_ERR_JRET(qwExecTask(QW_FPARAMS(), &ctx, NULL));
+
+  QW_ERR_JRET(qwGetDeleteResFromSink(QW_FPARAMS(), &ctx, &pRsp->contLen, &pRsp->pCont, pRes));
+
+_return:
+
+  qwFreeTaskCtx(QW_FPARAMS(), &ctx);
+
+  QW_RET(TSDB_CODE_SUCCESS);
+}
+
 
 int32_t qWorkerInit(int8_t nodeType, int32_t nodeId, SQWorkerCfg *cfg, void **qWorkerMgmt, const SMsgCb *pMsgCb) {
   if (NULL == qWorkerMgmt || pMsgCb->mgmt == NULL) {
@@ -1007,6 +1095,7 @@ int32_t qWorkerGetStat(SReadHandle *handle, void *qWorkerMgmt, SQWorkerStat *pSt
   pStat->fetchProcessed = QW_STAT_GET(mgmt->stat.msgStat.fetchProcessed);
   pStat->dropProcessed = QW_STAT_GET(mgmt->stat.msgStat.dropProcessed);
   pStat->hbProcessed = QW_STAT_GET(mgmt->stat.msgStat.hbProcessed);
+  pStat->deleteProcessed = QW_STAT_GET(mgmt->stat.msgStat.deleteProcessed);
 
   pStat->numOfQueryInQueue = handle->pMsgCb->qsizeFp(handle->pMsgCb->mgmt, mgmt->nodeId, QUERY_QUEUE);
   pStat->numOfFetchInQueue = handle->pMsgCb->qsizeFp(handle->pMsgCb->mgmt, mgmt->nodeId, FETCH_QUEUE);

source/libs/scalar/src/filter.c

@@ -168,7 +168,7 @@ __compar_fn_t gDataCompare[] = {compareInt32Val, compareInt8Val, compareInt16Val
   compareLenPrefixedWStr, compareUint8Val, compareUint16Val, compareUint32Val, compareUint64Val,
   setChkInBytes1, setChkInBytes2, setChkInBytes4, setChkInBytes8, compareStrRegexCompMatch, 
   compareStrRegexCompNMatch, setChkNotInBytes1, setChkNotInBytes2, setChkNotInBytes4, setChkNotInBytes8,
-  compareChkNotInString, compareStrPatternNotMatch, compareWStrPatternNotMatch, compareJsonContainsKey
+  compareChkNotInString, compareStrPatternNotMatch, compareWStrPatternNotMatch
 };
 
 int8_t filterGetCompFuncIdx(int32_t type, int32_t optr) {

source/libs/scalar/src/sclvector.c

@@ -88,7 +88,7 @@ void convertNumberToNumber(const void *inData, void *outData, int8_t inType, int
   }
 }
 
-void convertStringToDouble(const void *inData, void *outData, int8_t inType, int8_t outType){
+void convertNcharToDouble(const void *inData, void *outData){
   char *tmp = taosMemoryMalloc(varDataTLen(inData));
   int len = taosUcs4ToMbs((TdUcs4 *)varDataVal(inData), varDataLen(inData), tmp);
   if (len < 0) {
@@ -97,13 +97,24 @@ void convertStringToDouble(const void *inData, void *outData, int8_t inType, int
 
   tmp[len] = 0;
 
-  ASSERT(outType == TSDB_DATA_TYPE_DOUBLE);
   double value = taosStr2Double(tmp, NULL);
 
   *((double *)outData) = value;
   taosMemoryFreeClear(tmp);
 }
 
+void convertBinaryToDouble(const void *inData, void *outData){
+  char *tmp = taosMemoryCalloc(1, varDataTLen(inData));
+  if(tmp == NULL){
+    *((double *)outData) = 0.;
+    return;
+  }
+  memcpy(tmp, varDataVal(inData), varDataLen(inData));
+  double ret = taosStr2Double(tmp, NULL);
+  taosMemoryFree(tmp);
+  *((double *)outData) = ret;
+}
+
 typedef int64_t (*_getBigintValue_fn_t)(void *src, int32_t index);
 
 int64_t getVectorBigintValue_TINYINT(void *src, int32_t index) {
@@ -147,7 +158,7 @@ int64_t getVectorBigintValue_JSON(void *src, int32_t index){
   if (*data == TSDB_DATA_TYPE_NULL){
     return 0;
   } else if(*data == TSDB_DATA_TYPE_NCHAR) {   // json inner type can not be BINARY
-    convertStringToDouble(data+CHAR_BYTES, &out, *data, TSDB_DATA_TYPE_DOUBLE);
+    convertNcharToDouble(data+CHAR_BYTES, &out);
   } else {
     convertNumberToNumber(data+CHAR_BYTES, &out, *data, TSDB_DATA_TYPE_DOUBLE);
   }
@@ -445,14 +456,30 @@ double getVectorDoubleValue_JSON(void *src, int32_t index){
   if (*data == TSDB_DATA_TYPE_NULL){
     return out;
   } else if(*data == TSDB_DATA_TYPE_NCHAR) {   // json inner type can not be BINARY
-    convertStringToDouble(data+CHAR_BYTES, &out, *data, TSDB_DATA_TYPE_DOUBLE);
+    convertNcharToDouble(data+CHAR_BYTES, &out);
   } else {
     convertNumberToNumber(data+CHAR_BYTES, &out, *data, TSDB_DATA_TYPE_DOUBLE);
   }
   return out;
 }
 
-bool convertJsonValue(__compar_fn_t *fp, int32_t optr, int8_t typeLeft, int8_t typeRight, char **pLeftData, char **pRightData, void *pLeftOut, void *pRightOut, bool *isNull){
+void* ncharTobinary(void *buf){            // todo need to remove , if tobinary is nchar
+  int32_t inputLen = varDataLen(buf);
+
+  void* t = taosMemoryCalloc(1, inputLen);
+  int32_t len  = taosUcs4ToMbs((TdUcs4 *)varDataVal(buf), varDataLen(buf), varDataVal(t));
+  if (len < 0) {
+    sclError("charset:%s to %s. val:%s convert ncharTobinary failed.", DEFAULT_UNICODE_ENCODEC, tsCharset,
+             (char*)varDataVal(buf));
+    taosMemoryFree(t);
+    return NULL;
+  }
+  varDataSetLen(t, len);
+  return t;
+}
+
+bool convertJsonValue(__compar_fn_t *fp, int32_t optr, int8_t typeLeft, int8_t typeRight, char **pLeftData, char **pRightData,
+                      void *pLeftOut, void *pRightOut, bool *isNull, bool *freeLeft, bool *freeRight){
   if(optr == OP_TYPE_JSON_CONTAINS) {
     return true;
   }
@@ -489,8 +516,12 @@ bool convertJsonValue(__compar_fn_t *fp, int32_t optr, int8_t typeLeft, int8_t t
 
   *fp = filterGetCompFunc(type, optr);
 
+  if(IS_NUMERIC_TYPE(type) || IS_FLOAT_TYPE(type)){
     if(typeLeft == TSDB_DATA_TYPE_NCHAR) {
-    convertStringToDouble(*pLeftData, pLeftOut, typeLeft, type);
+      convertNcharToDouble(*pLeftData, pLeftOut);
+      *pLeftData = pLeftOut;
+    } else if(typeLeft == TSDB_DATA_TYPE_BINARY) {
+      convertBinaryToDouble(*pLeftData, pLeftOut);
       *pLeftData = pLeftOut;
     } else if(typeLeft != type) {
       convertNumberToNumber(*pLeftData, pLeftOut, typeLeft, type);
@@ -498,12 +529,28 @@ bool convertJsonValue(__compar_fn_t *fp, int32_t optr, int8_t typeLeft, int8_t t
     }
 
     if(typeRight == TSDB_DATA_TYPE_NCHAR) {
-    convertStringToDouble(*pRightData, pRightOut, typeRight, type);
+      convertNcharToDouble(*pRightData, pRightOut);
+      *pRightData = pRightOut;
+    } else if(typeRight == TSDB_DATA_TYPE_BINARY) {
+      convertBinaryToDouble(*pRightData, pRightOut);
       *pRightData = pRightOut;
     } else if(typeRight != type) {
       convertNumberToNumber(*pRightData, pRightOut, typeRight, type);
       *pRightData = pRightOut;
     }
+  }else if(type == TSDB_DATA_TYPE_BINARY){
+    if(typeLeft == TSDB_DATA_TYPE_NCHAR){
+      *pLeftData = ncharTobinary(*pLeftData);
+      *freeLeft = true;
+    }
+    if(typeRight == TSDB_DATA_TYPE_NCHAR){
+      *pRightData = ncharTobinary(*pRightData);
+      *freeRight = true;
+    }
+  }else{
+    ASSERT(0);
+  }
+
   return true;
 }
 
@@ -935,44 +982,6 @@ static void doReleaseVec(SColumnInfoData* pCol, int32_t type) {
   }
 }
 
-STagVal getJsonValue(char *json, char *key, bool *isExist) {
-  STagVal val = {.pKey = key};
-  bool find = tTagGet(((const STag *)json), &val);  // json value is null and not exist is different
-  if(isExist){
-    *isExist = find;
-  }
-  return val;
-}
-
-void vectorJsonArrow(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut, int32_t _ord) {
-  SColumnInfoData *pOutputCol = pOut->columnData;
-
-  int32_t i = ((_ord) == TSDB_ORDER_ASC)? 0 : TMAX(pLeft->numOfRows, pRight->numOfRows) - 1;
-  int32_t step = ((_ord) == TSDB_ORDER_ASC)? 1 : -1;
-
-  pOut->numOfRows = TMAX(pLeft->numOfRows, pRight->numOfRows);
-
-  char *pRightData = colDataGetVarData(pRight->columnData, 0);
-  char *jsonKey = taosMemoryCalloc(1, varDataLen(pRightData) + 1);
-  memcpy(jsonKey, varDataVal(pRightData), varDataLen(pRightData));
-  for (; i >= 0 && i < pLeft->numOfRows; i += step) {
-    if (colDataIsNull_var(pLeft->columnData, i)) {
-      colDataSetNull_var(pOutputCol, i);
-      pOutputCol->hasNull = true;
-      continue;
-    }
-    char *pLeftData = colDataGetVarData(pLeft->columnData, i);
-    bool isExist = false;
-    STagVal value = getJsonValue(pLeftData, jsonKey, &isExist);
-    char *data = isExist ? tTagValToData(&value, true) : NULL;
-    colDataAppend(pOutputCol, i, data, data == NULL);
-    if(isExist && IS_VAR_DATA_TYPE(value.type) && data){
-      taosMemoryFree(data);
-    }
-  }
-  taosMemoryFree(jsonKey);
-}
-
 void vectorMathAdd(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut, int32_t _ord) {
   SColumnInfoData *pOutputCol = pOut->columnData;
 
@@ -1510,6 +1519,38 @@ void vectorBitOr(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut,
   doReleaseVec(pRightCol, rightConvert);
 }
 
+#define VEC_COM_INNER(pCol, index1, index2)    \
+  for (; i < pCol->numOfRows && i >= 0; i += step) {\
+    if (IS_HELPER_NULL(pLeft->columnData, index1) || IS_HELPER_NULL(pRight->columnData, index2)) {\
+      bool  res = false;\
+      colDataAppendInt8(pOut->columnData, i, (int8_t*)&res);\
+      continue;\
+    }\
+    char *pLeftData = colDataGetData(pLeft->columnData, index1);\
+    char *pRightData = colDataGetData(pRight->columnData, index2);\
+    int64_t leftOut = 0;\
+    int64_t rightOut = 0;\
+    bool freeLeft = false;\
+    bool freeRight = false;\
+    bool isJsonnull = false;\
+    bool result = convertJsonValue(&fp, optr, GET_PARAM_TYPE(pLeft), GET_PARAM_TYPE(pRight),\
+                                      &pLeftData, &pRightData, &leftOut, &rightOut, &isJsonnull, &freeLeft, &freeRight);\
+    if(isJsonnull){\
+      ASSERT(0);\
+    }\
+    if(!pLeftData || !pRightData){\
+      result = false;\
+    }\
+    if(!result){\
+      colDataAppendInt8(pOut->columnData, i, (int8_t*)&result);\
+    }else{\
+      bool  res = filterDoCompare(fp, optr, pLeftData, pRightData);\
+      colDataAppendInt8(pOut->columnData, i, (int8_t*)&res);\
+    }\
+    if(freeLeft) taosMemoryFreeClear(pLeftData);\
+    if(freeRight) taosMemoryFreeClear(pRightData);\
+  }
+
 void vectorCompareImpl(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut, int32_t _ord, int32_t optr) {
   int32_t       i = ((_ord) == TSDB_ORDER_ASC) ? 0 : TMAX(pLeft->numOfRows, pRight->numOfRows) - 1;
   int32_t       step = ((_ord) == TSDB_ORDER_ASC) ? 1 : -1;
@@ -1533,79 +1574,11 @@ void vectorCompareImpl(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *
   }
 
   if (pLeft->numOfRows == pRight->numOfRows) {
-    for (; i < pRight->numOfRows && i >= 0; i += step) {
+    VEC_COM_INNER(pLeft, i, i)
-      if (IS_HELPER_NULL(pLeft->columnData, i) || IS_HELPER_NULL(pRight->columnData, i)) {
-        bool  res = false;
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&res);
-        continue;  // TODO set null or ignore
-      }
-
-      char *pLeftData = colDataGetData(pLeft->columnData, i);
-      char *pRightData = colDataGetData(pRight->columnData, i);
-
-      int64_t leftOut = 0;
-      int64_t rightOut = 0;
-      bool isJsonnull = false;
-      bool result = convertJsonValue(&fp, optr, GET_PARAM_TYPE(pLeft), GET_PARAM_TYPE(pRight), &pLeftData, &pRightData, &leftOut, &rightOut, &isJsonnull);
-      if(isJsonnull){
-        ASSERT(0);
-      }
-      if(!result){
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&result);
-      }else{
-        bool  res = filterDoCompare(fp, optr, pLeftData, pRightData);
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&res);
-      }
-    }
   } else if (pRight->numOfRows == 1) {
-    ASSERT(pLeft->pHashFilter == NULL);
+    VEC_COM_INNER(pLeft, i, 0)
-    for (; i >= 0 && i < pLeft->numOfRows; i += step) {
-      if (IS_HELPER_NULL(pLeft->columnData, i) || IS_HELPER_NULL(pRight->columnData, 0)) {
-        bool  res = false;
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&res);
-        continue;
-      }
-
-      char *pLeftData = colDataGetData(pLeft->columnData, i);
-      char *pRightData = colDataGetData(pRight->columnData, 0);
-      int64_t leftOut = 0;
-      int64_t rightOut = 0;
-      bool isJsonnull = false;
-      bool result = convertJsonValue(&fp, optr, GET_PARAM_TYPE(pLeft), GET_PARAM_TYPE(pRight), &pLeftData, &pRightData, &leftOut, &rightOut, &isJsonnull);
-      if(isJsonnull){
-        ASSERT(0);
-      }
-      if(!result){
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&result);
-      }else{
-        bool  res = filterDoCompare(fp, optr, pLeftData, pRightData);
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&res);
-      }
-    }
   } else if (pLeft->numOfRows == 1) {
-    for (; i >= 0 && i < pRight->numOfRows; i += step) {
+    VEC_COM_INNER(pRight, 0, i)
-      if (IS_HELPER_NULL(pRight->columnData, i) || IS_HELPER_NULL(pLeft->columnData, 0)) {
-        bool  res = false;
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&res);
-        continue;
-      }
-
-      char *pLeftData = colDataGetData(pLeft->columnData, 0);
-      char *pRightData = colDataGetData(pLeft->columnData, i);
-      int64_t leftOut = 0;
-      int64_t rightOut = 0;
-      bool isJsonnull = false;
-      bool result = convertJsonValue(&fp, optr, GET_PARAM_TYPE(pLeft), GET_PARAM_TYPE(pRight), &pLeftData, &pRightData, &leftOut, &rightOut, &isJsonnull);
-      if(isJsonnull){
-        ASSERT(0);
-      }
-      if(!result){
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&result);
-      }else{
-        bool  res = filterDoCompare(fp, optr, pLeftData, pRightData);
-        colDataAppendInt8(pOut->columnData, i, (int8_t*)&res);
-      }
-    }
   }
 }
 
@@ -1683,10 +1656,6 @@ void vectorNotMatch(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOu
   vectorCompare(pLeft, pRight, pOut, _ord, OP_TYPE_NMATCH);
 }
 
-void vectorJsonContains(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut, int32_t _ord) {
-  vectorCompare(pLeft, pRight, pOut, _ord, OP_TYPE_JSON_CONTAINS);
-}
-
 void vectorIsNull(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut, int32_t _ord) {
   for(int32_t i = 0; i < pLeft->numOfRows; ++i) {
     int8_t v = IS_HELPER_NULL(pLeft->columnData, i) ? 1 : 0;
@@ -1707,6 +1676,69 @@ void vectorIsTrue(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut,
   vectorConvertImpl(pLeft, pOut);
 }
 
+STagVal getJsonValue(char *json, char *key, bool *isExist) {
+  STagVal val = {.pKey = key};
+  bool find = tTagGet(((const STag *)json), &val);  // json value is null and not exist is different
+  if(isExist){
+    *isExist = find;
+  }
+  return val;
+}
+
+void vectorJsonContains(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut, int32_t _ord) {
+  SColumnInfoData *pOutputCol = pOut->columnData;
+
+  int32_t i = ((_ord) == TSDB_ORDER_ASC)? 0 : TMAX(pLeft->numOfRows, pRight->numOfRows) - 1;
+  int32_t step = ((_ord) == TSDB_ORDER_ASC)? 1 : -1;
+
+  pOut->numOfRows = TMAX(pLeft->numOfRows, pRight->numOfRows);
+
+  char *pRightData = colDataGetVarData(pRight->columnData, 0);
+  char *jsonKey = taosMemoryCalloc(1, varDataLen(pRightData) + 1);
+  memcpy(jsonKey, varDataVal(pRightData), varDataLen(pRightData));
+  for (; i >= 0 && i < pLeft->numOfRows; i += step) {
+    bool isExist = false;
+
+    if (!colDataIsNull_var(pLeft->columnData, i)) {
+      char *pLeftData = colDataGetVarData(pLeft->columnData, i);
+      getJsonValue(pLeftData, jsonKey, &isExist);
+    }
+
+    colDataAppend(pOutputCol, i, (const char*)(&isExist), false);
+
+  }
+  taosMemoryFree(jsonKey);
+}
+
+void vectorJsonArrow(SScalarParam* pLeft, SScalarParam* pRight, SScalarParam *pOut, int32_t _ord) {
+  SColumnInfoData *pOutputCol = pOut->columnData;
+
+  int32_t i = ((_ord) == TSDB_ORDER_ASC)? 0 : TMAX(pLeft->numOfRows, pRight->numOfRows) - 1;
+  int32_t step = ((_ord) == TSDB_ORDER_ASC)? 1 : -1;
+
+  pOut->numOfRows = TMAX(pLeft->numOfRows, pRight->numOfRows);
+
+  char *pRightData = colDataGetVarData(pRight->columnData, 0);
+  char *jsonKey = taosMemoryCalloc(1, varDataLen(pRightData) + 1);
+  memcpy(jsonKey, varDataVal(pRightData), varDataLen(pRightData));
+  for (; i >= 0 && i < pLeft->numOfRows; i += step) {
+    if (colDataIsNull_var(pLeft->columnData, i)) {
+      colDataSetNull_var(pOutputCol, i);
+      pOutputCol->hasNull = true;
+      continue;
+    }
+    char *pLeftData = colDataGetVarData(pLeft->columnData, i);
+    bool isExist = false;
+    STagVal value = getJsonValue(pLeftData, jsonKey, &isExist);
+    char *data = isExist ? tTagValToData(&value, true) : NULL;
+    colDataAppend(pOutputCol, i, data, data == NULL);
+    if(isExist && IS_VAR_DATA_TYPE(value.type) && data){
+      taosMemoryFree(data);
+    }
+  }
+  taosMemoryFree(jsonKey);
+}
+
 _bin_scalar_fn_t getBinScalarOperatorFn(int32_t binFunctionId) {
   switch (binFunctionId) {
     case OP_TYPE_ADD:

source/libs/scheduler/src/schJob.c

@@ -108,7 +108,7 @@ int32_t schInitJob(SSchJob **pSchJob, SQueryPlan *pDag, void *pTrans, SArray *pN
 
   pJob->refId = refId;
 
-  SCH_JOB_DLOG("job refId:%" PRIx64, pJob->refId);
+  SCH_JOB_DLOG("job refId:0x%" PRIx64" created", pJob->refId);
 
   pJob->status = JOB_TASK_STATUS_NOT_START;
 
@@ -1476,7 +1476,7 @@ int32_t schExecJobImpl(void *pTrans, SArray *pNodeList, SQueryPlan *pDag, int64_
   SSchJob *pJob = NULL;
   SCH_ERR_RET(schInitJob(&pJob, pDag, pTrans, pNodeList, sql, pRes, startTs, sync));
 
-  qDebug("QID:0x%" PRIx64 " jobId:0x%"PRIx64 " started", pDag->queryId, pJob->refId);
+  qDebug("QID:0x%" PRIx64 " job refId 0x%"PRIx64 " started", pDag->queryId, pJob->refId);
   *job = pJob->refId;
 
   SCH_ERR_JRET(schLaunchJob(pJob));

source/libs/scheduler/src/schRemote.c

@@ -62,10 +62,11 @@ int32_t schValidateReceivedMsgType(SSchJob *pJob, SSchTask *pTask, int32_t msgTy
     case TDMT_VND_DROP_TABLE_RSP:
     case TDMT_VND_ALTER_TABLE_RSP:
     case TDMT_VND_SUBMIT_RSP:
+    case TDMT_VND_DELETE_RSP:
       break;
     default:
       SCH_TASK_ELOG("unknown rsp msg, type:%s, status:%s", TMSG_INFO(msgType), jobTaskStatusStr(taskStatus));
-      SCH_ERR_RET(TSDB_CODE_QRY_INVALID_INPUT);
+      SCH_ERR_RET(TSDB_CODE_INVALID_MSG);
   }
 
   if (lastMsgType != reqMsgType) {
@@ -227,6 +228,25 @@ int32_t schHandleResponseMsg(SSchJob *pJob, SSchTask *pTask, int32_t msgType, ch
 
       break;
     }
+    case TDMT_VND_DELETE_RSP: {
+      SCH_ERR_JRET(rspCode);
+
+      if (msg) {
+        SDecoder coder = {0};
+        SVDeleteRsp rsp = {0};
+        tDecoderInit(&coder, msg, msgSize);
+        tDecodeSVDeleteRsp(&coder, &rsp);
+
+        atomic_add_fetch_32(&pJob->resNumOfRows, rsp.affectedRows);
+        SCH_TASK_DLOG("delete succeed, affectedRows:%" PRId64, rsp.affectedRows);
+      }
+
+      taosMemoryFreeClear(msg);
+
+      SCH_ERR_RET(schProcessOnTaskSuccess(pJob, pTask));
+
+      break;
+    }    
     case TDMT_VND_QUERY_RSP: {
       SQueryTableRsp *rsp = (SQueryTableRsp *)msg;
 
@@ -411,6 +431,10 @@ int32_t schHandleQueryCallback(void *param, const SDataBuf *pMsg, int32_t code)
   return schHandleCallback(param, pMsg, TDMT_VND_QUERY_RSP, code);
 }
 
+int32_t schHandleDeleteCallback(void *param, const SDataBuf *pMsg, int32_t code) {
+  return schHandleCallback(param, pMsg, TDMT_VND_DELETE_RSP, code);
+}
+
 int32_t schHandleFetchCallback(void *param, const SDataBuf *pMsg, int32_t code) {
   return schHandleCallback(param, pMsg, TDMT_VND_FETCH_RSP, code);
 }
@@ -501,6 +525,9 @@ int32_t schGetCallbackFp(int32_t msgType, __async_send_cb_fn_t *fp) {
     case TDMT_VND_QUERY:
       *fp = schHandleQueryCallback;
       break;
+    case TDMT_VND_DELETE:
+      *fp = schHandleDeleteCallback;
+      break;
     case TDMT_VND_EXPLAIN:
       *fp = schHandleExplainCallback;
       break;
@@ -982,6 +1009,26 @@ int32_t schBuildAndSendMsg(SSchJob *pJob, SSchTask *pTask, SQueryNodeAddr *addr,
       break;
     }
 
+    case TDMT_VND_DELETE: {
+      SVDeleteReq req = {0};
+      req.header.vgId = addr->nodeId;
+      req.sId = schMgmt.sId;
+      req.queryId = pJob->queryId;
+      req.taskId = pTask->taskId;
+      req.phyLen = pTask->msgLen;
+      req.sqlLen = strlen(pJob->sql);
+      req.sql = (char*)pJob->sql;
+      req.msg = pTask->msg;
+      msgSize = tSerializeSVDeleteReq(NULL, 0, &req);
+      msg = taosMemoryCalloc(1, msgSize);
+      if (NULL == msg) {
+        SCH_TASK_ELOG("calloc %d failed", msgSize);
+        SCH_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY);
+      }
+
+      tSerializeSVDeleteReq(msg, msgSize, &req);
+      break;
+    }
     case TDMT_VND_QUERY: {
       SCH_ERR_RET(schMakeQueryRpcCtx(pJob, pTask, &rpcCtx));
 

source/libs/scheduler/src/schUtil.c

@@ -257,6 +257,8 @@ void schFreeRpcCtx(SRpcCtx *pCtx) {
 
   taosHashCleanup(pCtx->args);
 
+  if (pCtx->freeFunc) {
     (*pCtx->freeFunc)(pCtx->brokenVal.val);
+  }
 }
 

source/libs/stream/inc/streamInc.h

@@ -13,15 +13,22 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
+#ifndef _STREAM_INC_H_
+#define _STREAM_INC_H_
+
+#include "executor.h"
+#include "tstream.h"
+
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-#ifndef _TSTREAM_H_
+int32_t streamExec(SStreamTask* pTask, SMsgCb* pMsgCb);
-#define _TSTREAM_H_
+int32_t streamSink1(SStreamTask* pTask, SMsgCb* pMsgCb);
+int32_t streamDispatch(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDataBlock* data);
 
 #ifdef __cplusplus
 }
 #endif
 
-#endif /* ifndef _TSTREAM_H_ */
+#endif /* ifndef _STREAM_INC_H_ */

source/libs/stream/src/stream.c

@@ -0,0 +1,118 @@
+/*
+ * 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 "streamInc.h"
+
+int32_t streamTriggerByWrite(SStreamTask* pTask, int32_t vgId, SMsgCb* pMsgCb) {
+  int8_t execStatus = atomic_load_8(&pTask->status);
+  if (execStatus == TASK_STATUS__IDLE || execStatus == TASK_STATUS__CLOSING) {
+    SStreamTaskRunReq* pRunReq = rpcMallocCont(sizeof(SStreamTaskRunReq));
+    if (pRunReq == NULL) return -1;
+
+    // TODO: do we need htonl?
+    pRunReq->head.vgId = vgId;
+    pRunReq->streamId = pTask->streamId;
+    pRunReq->taskId = pTask->taskId;
+    SRpcMsg msg = {
+        .msgType = TDMT_VND_TASK_RUN,
+        .pCont = pRunReq,
+        .contLen = sizeof(SStreamTaskRunReq),
+    };
+    tmsgPutToQueue(pMsgCb, FETCH_QUEUE, &msg);
+  }
+  return 0;
+}
+
+#if 1
+int32_t streamTaskEnqueue(SStreamTask* pTask, SStreamDispatchReq* pReq, SRpcMsg* pRsp) {
+  SStreamDataBlock* pBlock = taosAllocateQitem(sizeof(SStreamDataBlock), DEF_QITEM);
+  int8_t            status;
+
+  // enqueue
+  if (pBlock != NULL) {
+    pBlock->type = STREAM_DATA_TYPE_SSDATA_BLOCK;
+    pBlock->sourceVg = pReq->sourceVg;
+    pBlock->blocks = pReq->data;
+    /*pBlock->sourceVer = pReq->sourceVer;*/
+    if (streamTaskInput(pTask, (SStreamQueueItem*)pBlock) == 0) {
+      status = TASK_INPUT_STATUS__NORMAL;
+    } else {
+      status = TASK_INPUT_STATUS__FAILED;
+    }
+  } else {
+    streamTaskInputFail(pTask);
+    status = TASK_INPUT_STATUS__FAILED;
+  }
+
+  // rsp by input status
+  void* buf = rpcMallocCont(sizeof(SMsgHead) + sizeof(SStreamDispatchRsp));
+  ((SMsgHead*)buf)->vgId = htonl(pReq->sourceVg);
+  SStreamDispatchRsp* pCont = POINTER_SHIFT(buf, sizeof(SMsgHead));
+  pCont->inputStatus = status;
+  pCont->streamId = pReq->streamId;
+  pCont->taskId = pReq->sourceTaskId;
+  pRsp->pCont = buf;
+  pRsp->contLen = sizeof(SMsgHead) + sizeof(SStreamDispatchRsp);
+  tmsgSendRsp(pRsp);
+  return status == TASK_INPUT_STATUS__NORMAL ? 0 : -1;
+}
+#endif
+
+int32_t streamProcessDispatchReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDispatchReq* pReq, SRpcMsg* pRsp) {
+  // 1. handle input
+  streamTaskEnqueue(pTask, pReq, pRsp);
+
+  // 2. try exec
+  // 2.1. idle: exec
+  // 2.2. executing: return
+  // 2.3. closing: keep trying
+  streamExec(pTask, pMsgCb);
+
+  // 3. handle output
+  // 3.1 check and set status
+  // 3.2 dispatch / sink
+  streamSink1(pTask, pMsgCb);
+
+  return 0;
+}
+
+int32_t streamProcessDispatchRsp(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDispatchRsp* pRsp) {
+  ASSERT(pRsp->inputStatus == TASK_OUTPUT_STATUS__NORMAL || pRsp->inputStatus == TASK_OUTPUT_STATUS__BLOCKED);
+  int8_t old = atomic_exchange_8(&pTask->outputStatus, pRsp->inputStatus);
+  ASSERT(old == TASK_OUTPUT_STATUS__WAIT);
+  if (pRsp->inputStatus == TASK_INPUT_STATUS__BLOCKED) {
+    // TODO: init recover timer
+    return 0;
+  }
+  // continue dispatch
+  streamSink1(pTask, pMsgCb);
+  return 0;
+}
+
+int32_t streamTaskProcessRunReq(SStreamTask* pTask, SMsgCb* pMsgCb) {
+  streamExec(pTask, pMsgCb);
+  streamSink1(pTask, pMsgCb);
+  return 0;
+}
+
+int32_t streamProcessRecoverReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamTaskRecoverReq* pReq, SRpcMsg* pMsg) {
+  //
+  return 0;
+}
+
+int32_t streamProcessRecoverRsp(SStreamTask* pTask, SStreamTaskRecoverRsp* pRsp) {
+  //
+  return 0;
+}