Merge branch '3.0' of github.com:taosdata/TDengine into szhou/feature/project-compaction

include/common/tdatablock.h

@@ -236,7 +236,7 @@ void        blockCompressEncode(const SSDataBlock* pBlock, char* data, int32_t*
                                 int8_t needCompress);
 const char* blockCompressDecode(SSDataBlock* pBlock, int32_t numOfCols, int32_t numOfRows, const char* pData);
 
-void blockDebugShowData(const SArray* dataBlocks, const char* flag);
+void blockDebugShowDataBlocks(const SArray* dataBlocks, const char* flag);
 // for debug
 char* dumpBlockData(SSDataBlock* pDataBlock, const char* flag, char** dumpBuf);
 

include/libs/nodes/plannodes.h

@@ -24,6 +24,8 @@ extern "C" {
 #include "querynodes.h"
 #include "tname.h"
 
+#define SLOT_NAME_LEN TSDB_TABLE_NAME_LEN + TSDB_COL_NAME_LEN
+
 typedef struct SLogicNode {
   ENodeType          type;
   SNodeList*         pTargets;  // SColumnNode
@@ -74,8 +76,8 @@ typedef struct SScanLogicNode {
   int16_t       tsColId;
   double        filesFactor;
   SArray*       pSmaIndexes;
-  SNodeList*    pPartTags;
-  bool          partSort;
+  SNodeList*    pGroupTags;
+  bool          groupSort;
 } SScanLogicNode;
 
 typedef struct SJoinLogicNode {
@@ -100,6 +102,7 @@ typedef struct SProjectLogicNode {
 typedef struct SIndefRowsFuncLogicNode {
   SLogicNode node;
   SNodeList* pFuncs;
+  bool       isTailFunc;
 } SIndefRowsFuncLogicNode;
 
 typedef struct SInterpFuncLogicNode {
@@ -138,6 +141,7 @@ typedef struct SMergeLogicNode {
   SNodeList* pInputs;
   int32_t    numOfChannels;
   int32_t    srcGroupId;
+  bool       groupSort;
 } SMergeLogicNode;
 
 typedef enum EWindowType { WINDOW_TYPE_INTERVAL = 1, WINDOW_TYPE_SESSION, WINDOW_TYPE_STATE } EWindowType;
@@ -184,6 +188,7 @@ typedef struct SFillLogicNode {
 typedef struct SSortLogicNode {
   SLogicNode node;
   SNodeList* pSortKeys;
+  bool       groupSort;
 } SSortLogicNode;
 
 typedef struct SPartitionLogicNode {
@@ -230,6 +235,7 @@ typedef struct SSlotDescNode {
   bool      reserve;
   bool      output;
   bool      tag;
+  char      name[SLOT_NAME_LEN];
 } SSlotDescNode;
 
 typedef struct SDataBlockDescNode {
@@ -279,7 +285,8 @@ typedef struct STableScanPhysiNode {
   double         ratio;
   int32_t        dataRequired;
   SNodeList*     pDynamicScanFuncs;
-  SNodeList*     pPartitionTags;
+  SNodeList*     pGroupTags;
+  bool           groupSort;
   int64_t        interval;
   int64_t        offset;
   int64_t        sliding;
@@ -353,6 +360,7 @@ typedef struct SMergePhysiNode {
   SNodeList* pTargets;
   int32_t    numOfChannels;
   int32_t    srcGroupId;
+  bool       groupSort;
 } SMergePhysiNode;
 
 typedef struct SWinodwPhysiNode {

include/libs/nodes/querynodes.h

@@ -259,6 +259,7 @@ typedef struct SSelectStmt {
   bool        hasTailFunc;
   bool        hasInterpFunc;
   bool        hasLastRowFunc;
+  bool        groupSort;
 } SSelectStmt;
 
 typedef enum ESetOperatorType { SET_OP_TYPE_UNION_ALL = 1, SET_OP_TYPE_UNION } ESetOperatorType;

source/common/src/tdatablock.c

@@ -1605,7 +1605,7 @@ static char* formatTimestamp(char* buf, int64_t val, int precision) {
   return buf;
 }
 
-void blockDebugShowData(const SArray* dataBlocks, const char* flag) {
+void blockDebugShowDataBlocks(const SArray* dataBlocks, const char* flag) {
   char    pBuf[128] = {0};
   int32_t sz = taosArrayGetSize(dataBlocks);
   for (int32_t i = 0; i < sz; i++) {
@@ -1613,7 +1613,7 @@ void blockDebugShowData(const SArray* dataBlocks, const char* flag) {
     size_t       numOfCols = taosArrayGetSize(pDataBlock->pDataBlock);
 
     int32_t rows = pDataBlock->info.rows;
-    printf("%s |block type %d |child id %d|\n", flag, (int32_t)pDataBlock->info.type, pDataBlock->info.childId);
+    printf("%s |block type %d |child id %d|group id %zX\n", flag, (int32_t)pDataBlock->info.type, pDataBlock->info.childId, pDataBlock->info.groupId);
     for (int32_t j = 0; j < rows; j++) {
       printf("%s |", flag);
       for (int32_t k = 0; k < numOfCols; k++) {

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

@@ -39,8 +39,10 @@ typedef struct {
   int32_t   id;
   int32_t   errCode;
   int32_t   acceptableCode;
-  ETrnStage stage;
+  int32_t   retryCode;
   ETrnAct   actionType;
+  ETrnStage stage;
+  int8_t    reserved;
   int8_t    rawWritten;
   int8_t    msgSent;
   int8_t    msgReceived;

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

@@ -234,7 +234,7 @@ static int32_t mndProcessRetrieveSysTableReq(SRpcMsg *pReq) {
   if (retrieveReq.user[0] != 0) {
     memcpy(pReq->info.conn.user, retrieveReq.user, TSDB_USER_LEN);
   } else {
-    memcpy(pReq->info.conn.user, TSDB_DEFAULT_USER, TSDB_USER_LEN);
+    memcpy(pReq->info.conn.user, TSDB_DEFAULT_USER, strlen(TSDB_DEFAULT_USER) + 1);
   }
   if (mndCheckShowPrivilege(pMnode, pReq->info.conn.user, pShow->type, retrieveReq.db) != 0) {
     return -1;

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

@@ -15,9 +15,9 @@
 
 #define _DEFAULT_SOURCE
 #include "mndTrans.h"
-#include "mndPrivilege.h"
 #include "mndConsumer.h"
 #include "mndDb.h"
+#include "mndPrivilege.h"
 #include "mndShow.h"
 #include "mndSync.h"
 #include "mndUser.h"
@@ -55,7 +55,7 @@ static bool    mndTransPerfromFinishedStage(SMnode *pMnode, STrans *pTrans);
 static bool    mndCannotExecuteTransAction(SMnode *pMnode) { return !pMnode->deploy && !mndIsMaster(pMnode); }
 
 static void    mndTransSendRpcRsp(SMnode *pMnode, STrans *pTrans);
-static int32_t mndProcessTransReq(SRpcMsg *pReq);
+static int32_t mndProcessTransTimer(SRpcMsg *pReq);
 static int32_t mndProcessTtl(SRpcMsg *pReq);
 static int32_t mndProcessKillTransReq(SRpcMsg *pReq);
 
@@ -73,7 +73,7 @@ int32_t mndInitTrans(SMnode *pMnode) {
       .deleteFp = (SdbDeleteFp)mndTransActionDelete,
   };
 
-  mndSetMsgHandle(pMnode, TDMT_MND_TRANS_TIMER, mndProcessTransReq);
+  mndSetMsgHandle(pMnode, TDMT_MND_TRANS_TIMER, mndProcessTransTimer);
   mndSetMsgHandle(pMnode, TDMT_MND_KILL_TRANS, mndProcessKillTransReq);
 
   mndAddShowRetrieveHandle(pMnode, TSDB_MGMT_TABLE_TRANS, mndRetrieveTrans);
@@ -139,8 +139,10 @@ static SSdbRaw *mndTransActionEncode(STrans *pTrans) {
     SDB_SET_INT32(pRaw, dataPos, pAction->id, _OVER)
     SDB_SET_INT32(pRaw, dataPos, pAction->errCode, _OVER)
     SDB_SET_INT32(pRaw, dataPos, pAction->acceptableCode, _OVER)
+    SDB_SET_INT32(pRaw, dataPos, pAction->retryCode, _OVER)
     SDB_SET_INT8(pRaw, dataPos, pAction->actionType, _OVER)
     SDB_SET_INT8(pRaw, dataPos, pAction->stage, _OVER)
+    SDB_SET_INT8(pRaw, dataPos, pAction->reserved, _OVER)
     if (pAction->actionType == TRANS_ACTION_RAW) {
       int32_t len = sdbGetRawTotalSize(pAction->pRaw);
       SDB_SET_INT8(pRaw, dataPos, pAction->rawWritten, _OVER)
@@ -163,8 +165,10 @@ static SSdbRaw *mndTransActionEncode(STrans *pTrans) {
     SDB_SET_INT32(pRaw, dataPos, pAction->id, _OVER)
     SDB_SET_INT32(pRaw, dataPos, pAction->errCode, _OVER)
     SDB_SET_INT32(pRaw, dataPos, pAction->acceptableCode, _OVER)
+    SDB_SET_INT32(pRaw, dataPos, pAction->retryCode, _OVER)
     SDB_SET_INT8(pRaw, dataPos, pAction->actionType, _OVER)
     SDB_SET_INT8(pRaw, dataPos, pAction->stage, _OVER)
+    SDB_SET_INT8(pRaw, dataPos, pAction->reserved, _OVER)
     if (pAction->actionType == TRANS_ACTION_RAW) {
       int32_t len = sdbGetRawTotalSize(pAction->pRaw);
       SDB_SET_INT8(pRaw, dataPos, pAction->rawWritten, _OVER)
@@ -187,8 +191,10 @@ static SSdbRaw *mndTransActionEncode(STrans *pTrans) {
     SDB_SET_INT32(pRaw, dataPos, pAction->id, _OVER)
     SDB_SET_INT32(pRaw, dataPos, pAction->errCode, _OVER)
     SDB_SET_INT32(pRaw, dataPos, pAction->acceptableCode, _OVER)
+    SDB_SET_INT32(pRaw, dataPos, pAction->retryCode, _OVER)
     SDB_SET_INT8(pRaw, dataPos, pAction->actionType, _OVER)
     SDB_SET_INT8(pRaw, dataPos, pAction->stage, _OVER)
+    SDB_SET_INT8(pRaw, dataPos, pAction->reserved, _OVER)
     if (pAction->actionType == TRANS_ACTION_RAW) {
       int32_t len = sdbGetRawTotalSize(pAction->pRaw);
       SDB_SET_INT8(pRaw, dataPos, pAction->rawWritten, _OVER)
@@ -291,10 +297,12 @@ static SSdbRow *mndTransActionDecode(SSdbRaw *pRaw) {
     SDB_GET_INT32(pRaw, dataPos, &action.id, _OVER)
     SDB_GET_INT32(pRaw, dataPos, &action.errCode, _OVER)
     SDB_GET_INT32(pRaw, dataPos, &action.acceptableCode, _OVER)
+    SDB_GET_INT32(pRaw, dataPos, &action.retryCode, _OVER)
     SDB_GET_INT8(pRaw, dataPos, &actionType, _OVER)
     action.actionType = actionType;
     SDB_GET_INT8(pRaw, dataPos, &stage, _OVER)
     action.stage = stage;
+    SDB_GET_INT8(pRaw, dataPos, &action.reserved, _OVER)
     if (action.actionType == TRANS_ACTION_RAW) {
       SDB_GET_INT8(pRaw, dataPos, &action.rawWritten, _OVER)
       SDB_GET_INT32(pRaw, dataPos, &dataLen, _OVER)
@@ -324,10 +332,12 @@ static SSdbRow *mndTransActionDecode(SSdbRaw *pRaw) {
     SDB_GET_INT32(pRaw, dataPos, &action.id, _OVER)
     SDB_GET_INT32(pRaw, dataPos, &action.errCode, _OVER)
     SDB_GET_INT32(pRaw, dataPos, &action.acceptableCode, _OVER)
+    SDB_GET_INT32(pRaw, dataPos, &action.retryCode, _OVER)
     SDB_GET_INT8(pRaw, dataPos, &actionType, _OVER)
     action.actionType = actionType;
     SDB_GET_INT8(pRaw, dataPos, &stage, _OVER)
     action.stage = stage;
+    SDB_GET_INT8(pRaw, dataPos, &action.reserved, _OVER)
     if (action.actionType == TRANS_ACTION_RAW) {
       SDB_GET_INT8(pRaw, dataPos, &action.rawWritten, _OVER)
       SDB_GET_INT32(pRaw, dataPos, &dataLen, _OVER)
@@ -357,10 +367,12 @@ static SSdbRow *mndTransActionDecode(SSdbRaw *pRaw) {
     SDB_GET_INT32(pRaw, dataPos, &action.id, _OVER)
     SDB_GET_INT32(pRaw, dataPos, &action.errCode, _OVER)
     SDB_GET_INT32(pRaw, dataPos, &action.acceptableCode, _OVER)
+    SDB_GET_INT32(pRaw, dataPos, &action.retryCode, _OVER)
     SDB_GET_INT8(pRaw, dataPos, &actionType, _OVER)
     action.actionType = actionType;
     SDB_GET_INT8(pRaw, dataPos, &stage, _OVER)
     action.stage = stage;
+    SDB_GET_INT8(pRaw, dataPos, &action.reserved, _OVER)
     if (action.actionType) {
       SDB_GET_INT8(pRaw, dataPos, &action.rawWritten, _OVER)
       SDB_GET_INT32(pRaw, dataPos, &dataLen, _OVER)
@@ -463,15 +475,25 @@ static int32_t mndTransActionInsert(SSdb *pSdb, STrans *pTrans) {
     if (fp) {
       (*fp)(pSdb->pMnode, pTrans->param, pTrans->paramLen);
     }
+    pTrans->startFunc = 0;
   }
 
   return 0;
 }
 
 static void mndTransDropData(STrans *pTrans) {
+  if (pTrans->redoActions != NULL) {
     mndTransDropActions(pTrans->redoActions);
+    pTrans->redoActions = NULL;
+  }
+  if (pTrans->undoActions != NULL) {
     mndTransDropActions(pTrans->undoActions);
+    pTrans->undoActions = NULL;
+  }
+  if (pTrans->commitActions != NULL) {
     mndTransDropActions(pTrans->commitActions);
+    pTrans->commitActions = NULL;
+  }
   if (pTrans->rpcRsp != NULL) {
     taosMemoryFree(pTrans->rpcRsp);
     pTrans->rpcRsp = NULL;
@@ -492,6 +514,7 @@ static int32_t mndTransActionDelete(SSdb *pSdb, STrans *pTrans, bool callFunc) {
     if (fp) {
       (*fp)(pSdb->pMnode, pTrans->param, pTrans->paramLen);
     }
+    pTrans->stopFunc = 0;
   }
 
   mndTransDropData(pTrans);
@@ -805,7 +828,7 @@ static void mndTransSendRpcRsp(SMnode *pMnode, STrans *pTrans) {
       sendRsp = true;
     }
   } else {
-    if (pTrans->stage == TRN_STAGE_REDO_ACTION && pTrans->failedTimes > 3) {
+    if (pTrans->stage == TRN_STAGE_REDO_ACTION && pTrans->failedTimes > 6) {
       if (code == 0) code = TSDB_CODE_MND_TRANS_UNKNOW_ERROR;
       sendRsp = true;
     }
@@ -875,8 +898,8 @@ int32_t mndTransProcessRsp(SRpcMsg *pRsp) {
     pAction->errCode = pRsp->code;
   }
 
-  mDebug("trans:%d, %s:%d response is received, code:0x%x, accept:0x%x", transId, mndTransStr(pAction->stage), action,
-         pRsp->code, pAction->acceptableCode);
+  mDebug("trans:%d, %s:%d response is received, code:0x%x, accept:0x%x retry:0x%x", transId,
+         mndTransStr(pAction->stage), action, pRsp->code, pAction->acceptableCode, pAction->retryCode);
   mndTransExecute(pMnode, pTrans);
 
 _OVER:
@@ -884,6 +907,21 @@ _OVER:
   return 0;
 }
 
+static void mndTransResetAction(SMnode *pMnode, STrans *pTrans, STransAction *pAction) {
+  pAction->rawWritten = 0;
+  pAction->msgSent = 0;
+  pAction->msgReceived = 0;
+  if (pAction->errCode == TSDB_CODE_RPC_REDIRECT || pAction->errCode == TSDB_CODE_SYN_NEW_CONFIG_ERROR ||
+      pAction->errCode == TSDB_CODE_SYN_INTERNAL_ERROR || pAction->errCode == TSDB_CODE_SYN_NOT_LEADER) {
+    pAction->epSet.inUse = (pAction->epSet.inUse + 1) % pAction->epSet.numOfEps;
+    mDebug("trans:%d, %s:%d execute status is reset and set epset inuse:%d", pTrans->id, mndTransStr(pAction->stage),
+           pAction->id, pAction->epSet.inUse);
+  } else {
+    mDebug("trans:%d, %s:%d execute status is reset", pTrans->id, mndTransStr(pAction->stage), pAction->id);
+  }
+  pAction->errCode = 0;
+}
+
 static void mndTransResetActions(SMnode *pMnode, STrans *pTrans, SArray *pArray) {
   int32_t numOfActions = taosArrayGetSize(pArray);
 
@@ -894,18 +932,7 @@ static void mndTransResetActions(SMnode *pMnode, STrans *pTrans, SArray *pArray)
       continue;
     if (pAction->rawWritten && (pAction->errCode == 0 || pAction->errCode == pAction->acceptableCode)) continue;
 
-    pAction->rawWritten = 0;
-    pAction->msgSent = 0;
-    pAction->msgReceived = 0;
-    if (pAction->errCode == TSDB_CODE_RPC_REDIRECT || pAction->errCode == TSDB_CODE_SYN_NEW_CONFIG_ERROR ||
-        pAction->errCode == TSDB_CODE_SYN_INTERNAL_ERROR || pAction->errCode == TSDB_CODE_SYN_NOT_LEADER) {
-      pAction->epSet.inUse = (pAction->epSet.inUse + 1) % pAction->epSet.numOfEps;
-      mDebug("trans:%d, %s:%d execute status is reset and set epset inuse:%d", pTrans->id, mndTransStr(pAction->stage),
-             action, pAction->epSet.inUse);
-    } else {
-      mDebug("trans:%d, %s:%d execute status is reset", pTrans->id, mndTransStr(pAction->stage), action);
-    }
-    pAction->errCode = 0;
+    mndTransResetAction(pMnode, pTrans, pAction);
   }
 }
 
@@ -1112,9 +1139,9 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans)
         if (pAction->msgReceived) {
           if (pAction->errCode != 0 && pAction->errCode != pAction->acceptableCode) {
             code = pAction->errCode;
-            pAction->msgSent = 0;
-            pAction->msgReceived = 0;
-            mDebug("trans:%d, %s:%d execute status is reset", pTrans->id, mndTransStr(pAction->stage), action);
+            mndTransResetAction(pMnode, pTrans, pAction);
+          } else {
+            mDebug("trans:%d, %s:%d execute successfully", pTrans->id, mndTransStr(pAction->stage), action);
           }
         } else {
           code = TSDB_CODE_ACTION_IN_PROGRESS;
@@ -1123,6 +1150,8 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans)
       if (pAction->rawWritten) {
         if (pAction->errCode != 0 && pAction->errCode != pAction->acceptableCode) {
           code = pAction->errCode;
+        } else {
+          mDebug("trans:%d, %s:%d write successfully", pTrans->id, mndTransStr(pAction->stage), action);
         }
       }
     }
@@ -1156,9 +1185,15 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans)
     } else if (code == TSDB_CODE_ACTION_IN_PROGRESS) {
       mDebug("trans:%d, %s:%d is in progress and wait it finish", pTrans->id, mndTransStr(pAction->stage), pAction->id);
       break;
+    } else if (code == pAction->retryCode) {
+      mDebug("trans:%d, %s:%d receive code:0x%x and retry", pTrans->id, mndTransStr(pAction->stage), pAction->id, code);
+      taosMsleep(300);
+      continue;
     } else {
       terrno = code;
       pTrans->code = code;
+      mDebug("trans:%d, %s:%d receive code:0x%x and wait another schedule, failedTimes:%d", pTrans->id,
+             mndTransStr(pAction->stage), pAction->id, code, pTrans->failedTimes);
       break;
     }
   }
@@ -1343,7 +1378,7 @@ void mndTransExecute(SMnode *pMnode, STrans *pTrans) {
   mndTransSendRpcRsp(pMnode, pTrans);
 }
 
-static int32_t mndProcessTransReq(SRpcMsg *pReq) {
+static int32_t mndProcessTransTimer(SRpcMsg *pReq) {
   mndTransPullup(pReq->info.node);
   return 0;
 }

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

@@ -15,10 +15,10 @@
 
 #define _DEFAULT_SOURCE
 #include "mndVgroup.h"
-#include "mndPrivilege.h"
 #include "mndDb.h"
 #include "mndDnode.h"
 #include "mndMnode.h"
+#include "mndPrivilege.h"
 #include "mndShow.h"
 #include "mndTrans.h"
 #include "mndUser.h"
@@ -896,6 +896,8 @@ int32_t mndAddAlterVnodeConfirmAction(SMnode *pMnode, STrans *pTrans, SDbObj *pD
   action.pCont = pHead;
   action.contLen = contLen;
   action.msgType = TDMT_VND_ALTER_CONFIRM;
+  // incorrect redirect result will cause this erro
+  action.retryCode = TSDB_CODE_VND_INVALID_VGROUP_ID;
 
   if (mndTransAppendRedoAction(pTrans, &action) != 0) {
     taosMemoryFree(pHead);
@@ -942,6 +944,8 @@ static int32_t mndAddSetVnodeStandByAction(SMnode *pMnode, STrans *pTrans, SDbOb
   action.contLen = contLen;
   action.msgType = TDMT_SYNC_SET_VNODE_STANDBY;
   action.acceptableCode = TSDB_CODE_NODE_NOT_DEPLOYED;
+  // Keep retrying until the target vnode is not the leader
+  action.retryCode = TSDB_CODE_SYN_IS_LEADER;
 
   if (isRedo) {
     if (mndTransAppendRedoAction(pTrans, &action) != 0) {
@@ -1229,7 +1233,8 @@ static int32_t mndProcessRedistributeVgroupMsg(SRpcMsg *pReq) {
     }
 
     if (req.dnodeId1 == pVgroup->vnodeGid[0].dnodeId) {
-      terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
+      // terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
+      code = 0;
       goto _OVER;
     }
 
@@ -1351,7 +1356,8 @@ static int32_t mndProcessRedistributeVgroupMsg(SRpcMsg *pReq) {
     }
 
     if (pNew1 == NULL && pOld1 == NULL && pNew2 == NULL && pOld2 == NULL && pNew3 == NULL && pOld3 == NULL) {
-      terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
+      // terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
+      code = 0;
       goto _OVER;
     }
 

source/dnode/vnode/src/sma/smaRollup.c

@@ -487,7 +487,7 @@ static int32_t tdFetchAndSubmitRSmaResult(SRSmaInfoItem *pItem, int8_t blkType)
 #if 1
     char flag[10] = {0};
     snprintf(flag, 10, "level %" PRIi8, pItem->level);
-    blockDebugShowData(pResult, flag);
+    blockDebugShowDataBlocks(pResult, flag);
 #endif
     STsdb      *sinkTsdb = (pItem->level == TSDB_RETENTION_L1 ? pSma->pRSmaTsdb1 : pSma->pRSmaTsdb2);
     SSubmitReq *pReq = NULL;

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

@@ -299,7 +299,7 @@ int32_t vnodeProcessWriteMsg(SVnode *pVnode, int64_t version, SRpcMsg *pMsg, SRp
 void smaHandleRes(void *pVnode, int64_t smaId, const SArray *data) {
   // TODO
 
-  blockDebugShowData(data, __func__);
+  blockDebugShowDataBlocks(data, __func__);
   tdProcessTSmaInsert(((SVnode *)pVnode)->pSma, smaId, (const char *)data);
 }
 

source/libs/executor/inc/tsort.h

@@ -136,6 +136,13 @@ bool tsortIsNullVal(STupleHandle* pVHandle, int32_t colId);
  */
 void* tsortGetValue(STupleHandle* pVHandle, int32_t colId);
 
+/**
+ *
+ * @param pVHandle
+ * @return
+ */
+uint64_t tsortGetGroupId(STupleHandle* pVHandle);
+
 /**
  *
  * @param pSortHandle

source/libs/executor/src/scanoperator.c

@@ -2051,7 +2051,7 @@ int32_t createScanTableListInfo(STableScanPhysiNode* pTableScanNode, SReadHandle
     qDebug("no table qualified for query, TID:0x%" PRIx64 ", QID:0x%" PRIx64, taskId, queryId);
     return TSDB_CODE_SUCCESS;
   }
-  code = generateGroupIdMap(pTableListInfo, pHandle, pTableScanNode->pPartitionTags);
+  code = generateGroupIdMap(pTableListInfo, pHandle, pTableScanNode->pGroupTags);
   if (code != TSDB_CODE_SUCCESS) {
     return code;
   }
@@ -2455,7 +2455,7 @@ SOperatorInfo* createTableMergeScanOperatorInfo(STableScanPhysiNode* pTableScanN
   if (pInfo == NULL || pOperator == NULL) {
     goto _error;
   }
-  if (pTableScanNode->pPartitionTags) {
+  if (pTableScanNode->pGroupTags) {
     taosArraySort(pTableListInfo->pTableList, compareTableKeyInfoByGid);
   }
 

source/libs/executor/src/sortoperator.c

@@ -142,7 +142,8 @@ SSDataBlock* getSortedBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, i
 
 SSDataBlock* loadNextDataBlock(void* param) {
   SOperatorInfo* pOperator = (SOperatorInfo*)param;
-  return pOperator->fpSet.getNextFn(pOperator);
+  SSDataBlock *pBlock = pOperator->fpSet.getNextFn(pOperator);
+  return pBlock;
 }
 
 // todo refactor: merged with fetch fp
@@ -505,7 +506,9 @@ typedef struct SMultiwaySortMergeOperatorInfo {
 
   SSDataBlock* pInputBlock;
   int64_t      startTs;  // sort start time
+  bool  hasGroupId;
   uint64_t groupId;
+  STupleHandle *prefetchedTuple;
 } SMultiwaySortMergeOperatorInfo;
 
 int32_t doOpenMultiwaySortMergeOperator(SOperatorInfo* pOperator) {
@@ -560,12 +563,30 @@ SSDataBlock* getMultiwaySortedBlockData(SSortHandle* pHandle, SSDataBlock* pData
   blockDataEnsureCapacity(p, capacity);
 
   while (1) {
-    STupleHandle* pTupleHandle = tsortNextTuple(pHandle);
+    STupleHandle* pTupleHandle = NULL;
+    if (pInfo->prefetchedTuple == NULL) {
+      pTupleHandle = tsortNextTuple(pHandle);
+    } else {
+      pTupleHandle = pInfo->prefetchedTuple;
+      pInfo->groupId = tsortGetGroupId(pTupleHandle);
+      pInfo->prefetchedTuple = NULL;
+    }
+
     if (pTupleHandle == NULL) {
       break;
     }
 
+    uint64_t tupleGroupId = tsortGetGroupId(pTupleHandle);
+    if (!pInfo->hasGroupId) {
+      pInfo->groupId = tupleGroupId;
+      pInfo->hasGroupId = true;
       appendOneRowToDataBlock(p, pTupleHandle);
+    } else if (pInfo->groupId == tupleGroupId) {
+      appendOneRowToDataBlock(p, pTupleHandle);
+    } else {
+      pInfo->prefetchedTuple = pTupleHandle;
+      break;
+    }
     if (p->info.rows >= capacity) {
       break;
     }
@@ -608,7 +629,6 @@ SSDataBlock* doMultiwaySortMerge(SOperatorInfo* pOperator) {
 
   SSDataBlock* pBlock = getMultiwaySortedBlockData(pInfo->pSortHandle, pInfo->binfo.pRes,
                                                    pOperator->resultInfo.capacity, pInfo->pColMatchInfo, pOperator);
-
   if (pBlock != NULL) {
     pOperator->resultInfo.totalRows += pBlock->info.rows;
   } else {

source/libs/executor/src/tsort.c

@@ -752,6 +752,10 @@ void* tsortGetValue(STupleHandle* pVHandle, int32_t colIndex) {
   }
 }
 
+uint64_t tsortGetGroupId(STupleHandle* pVHandle) {
+  return pVHandle->pBlock->info.groupId;
+}
+
 SSortExecInfo tsortGetSortExecInfo(SSortHandle* pHandle) {
   SSortExecInfo info = {0};
 

source/libs/nodes/src/nodesCloneFuncs.c

@@ -351,7 +351,7 @@ static int32_t logicScanCopy(const SScanLogicNode* pSrc, SScanLogicNode* pDst) {
   COPY_SCALAR_FIELD(watermark);
   COPY_SCALAR_FIELD(tsColId);
   COPY_SCALAR_FIELD(filesFactor);
-  CLONE_NODE_LIST_FIELD(pPartTags);
+  CLONE_NODE_LIST_FIELD(pGroupTags);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -401,6 +401,7 @@ static int32_t logicMergeCopy(const SMergeLogicNode* pSrc, SMergeLogicNode* pDst
   CLONE_NODE_LIST_FIELD(pInputs);
   COPY_SCALAR_FIELD(numOfChannels);
   COPY_SCALAR_FIELD(srcGroupId);
+  COPY_SCALAR_FIELD(groupSort);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -436,6 +437,7 @@ static int32_t logicFillCopy(const SFillLogicNode* pSrc, SFillLogicNode* pDst) {
 static int32_t logicSortCopy(const SSortLogicNode* pSrc, SSortLogicNode* pDst) {
   COPY_BASE_OBJECT_FIELD(node, logicNodeCopy);
   CLONE_NODE_LIST_FIELD(pSortKeys);
+  COPY_SCALAR_FIELD(groupSort);
   return TSDB_CODE_SUCCESS;
 }
 
@@ -500,7 +502,7 @@ static int32_t physiTableScanCopy(const STableScanPhysiNode* pSrc, STableScanPhy
   COPY_SCALAR_FIELD(ratio);
   COPY_SCALAR_FIELD(dataRequired);
   CLONE_NODE_LIST_FIELD(pDynamicScanFuncs);
-  CLONE_NODE_LIST_FIELD(pPartitionTags);
+  CLONE_NODE_LIST_FIELD(pGroupTags);
   COPY_SCALAR_FIELD(interval);
   COPY_SCALAR_FIELD(offset);
   COPY_SCALAR_FIELD(sliding);

source/libs/nodes/src/nodesCodeFuncs.c

@@ -234,6 +234,8 @@ const char* nodesNodeName(ENodeType type) {
       return "PhysiMerge";
     case QUERY_NODE_PHYSICAL_PLAN_SORT:
       return "PhysiSort";
+    case QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT:
+      return "PhysiGroupSort";
     case QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL:
       return "PhysiHashInterval";
     case QUERY_NODE_PHYSICAL_PLAN_MERGE_ALIGNED_INTERVAL:
@@ -539,7 +541,7 @@ static const char* jkScanLogicPlanScanPseudoCols = "ScanPseudoCols";
 static const char* jkScanLogicPlanTableId = "TableId";
 static const char* jkScanLogicPlanTableType = "TableType";
 static const char* jkScanLogicPlanTagCond = "TagCond";
-static const char* jkScanLogicPlanPartTags = "PartTags";
+static const char* jkScanLogicPlanGroupTags = "GroupTags";
 
 static int32_t logicScanNodeToJson(const void* pObj, SJson* pJson) {
   const SScanLogicNode* pNode = (const SScanLogicNode*)pObj;
@@ -561,7 +563,7 @@ static int32_t logicScanNodeToJson(const void* pObj, SJson* pJson) {
     code = tjsonAddObject(pJson, jkScanLogicPlanTagCond, nodeToJson, pNode->pTagCond);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = nodeListToJson(pJson, jkScanLogicPlanPartTags, pNode->pPartTags);
+    code = nodeListToJson(pJson, jkScanLogicPlanGroupTags, pNode->pGroupTags);
   }
 
   return code;
@@ -588,7 +590,7 @@ static int32_t jsonToLogicScanNode(const SJson* pJson, void* pObj) {
     code = jsonToNodeObject(pJson, jkScanLogicPlanTagCond, &pNode->pTagCond);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = jsonToNodeList(pJson, jkScanLogicPlanPartTags, &pNode->pPartTags);
+    code = jsonToNodeList(pJson, jkScanLogicPlanGroupTags, &pNode->pGroupTags);
   }
 
   return code;
@@ -1430,7 +1432,8 @@ static const char* jkTableScanPhysiPlanTriggerType = "triggerType";
 static const char* jkTableScanPhysiPlanWatermark = "watermark";
 static const char* jkTableScanPhysiPlanTsColId = "tsColId";
 static const char* jkTableScanPhysiPlanFilesFactor = "FilesFactor";
-static const char* jkTableScanPhysiPlanPartitionTags = "PartitionTags";
+static const char* jkTableScanPhysiPlanGroupTags = "GroupTags";
+static const char* jkTableScanPhysiPlanGroupSort = "GroupSort";
 
 static int32_t physiTableScanNodeToJson(const void* pObj, SJson* pJson) {
   const STableScanPhysiNode* pNode = (const STableScanPhysiNode*)pObj;
@@ -1485,7 +1488,10 @@ static int32_t physiTableScanNodeToJson(const void* pObj, SJson* pJson) {
     code = tjsonAddDoubleToObject(pJson, jkTableScanPhysiPlanFilesFactor, pNode->filesFactor);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = nodeListToJson(pJson, jkTableScanPhysiPlanPartitionTags, pNode->pPartitionTags);
+    code = nodeListToJson(pJson, jkTableScanPhysiPlanGroupTags, pNode->pGroupTags);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddBoolToObject(pJson, jkTableScanPhysiPlanGroupSort, pNode->groupSort);
   }
 
   return code;
@@ -1544,7 +1550,10 @@ static int32_t jsonToPhysiTableScanNode(const SJson* pJson, void* pObj) {
     code = tjsonGetDoubleValue(pJson, jkTableScanPhysiPlanFilesFactor, &pNode->filesFactor);
   }
   if (TSDB_CODE_SUCCESS == code) {
-    code = jsonToNodeList(pJson, jkTableScanPhysiPlanPartitionTags, &pNode->pPartitionTags);
+    code = jsonToNodeList(pJson, jkTableScanPhysiPlanGroupTags, &pNode->pGroupTags);
+  }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonGetBoolValue(pJson, jkTableScanPhysiPlanGroupSort, &pNode->groupSort);
   }
 
   return code;
@@ -1725,6 +1734,7 @@ static const char* jkMergePhysiPlanMergeKeys = "MergeKeys";
 static const char* jkMergePhysiPlanTargets = "Targets";
 static const char* jkMergePhysiPlanNumOfChannels = "NumOfChannels";
 static const char* jkMergePhysiPlanSrcGroupId = "SrcGroupId";
+static const char* jkMergePhysiPlanGroupSort = "GroupSort";
 
 static int32_t physiMergeNodeToJson(const void* pObj, SJson* pJson) {
   const SMergePhysiNode* pNode = (const SMergePhysiNode*)pObj;
@@ -1742,6 +1752,9 @@ static int32_t physiMergeNodeToJson(const void* pObj, SJson* pJson) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonAddIntegerToObject(pJson, jkMergePhysiPlanSrcGroupId, pNode->srcGroupId);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddBoolToObject(pJson, jkMergePhysiPlanGroupSort, pNode->groupSort);
+  }
 
   return code;
 }
@@ -1762,6 +1775,9 @@ static int32_t jsonToPhysiMergeNode(const SJson* pJson, void* pObj) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonGetIntValue(pJson, jkMergePhysiPlanSrcGroupId, &pNode->srcGroupId);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonGetBoolValue(pJson, jkMergePhysiPlanGroupSort, &pNode->groupSort);
+  }
 
   return code;
 }
@@ -3369,6 +3385,7 @@ static const char* jkSlotDescSlotId = "SlotId";
 static const char* jkSlotDescDataType = "DataType";
 static const char* jkSlotDescReserve = "Reserve";
 static const char* jkSlotDescOutput = "Output";
+static const char* jkSlotDescName = "Name";
 
 static int32_t slotDescNodeToJson(const void* pObj, SJson* pJson) {
   const SSlotDescNode* pNode = (const SSlotDescNode*)pObj;
@@ -3383,6 +3400,9 @@ static int32_t slotDescNodeToJson(const void* pObj, SJson* pJson) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonAddBoolToObject(pJson, jkSlotDescOutput, pNode->output);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonAddStringToObject(pJson, jkSlotDescName, pNode->name);
+  }
 
   return code;
 }
@@ -3400,6 +3420,9 @@ static int32_t jsonToSlotDescNode(const SJson* pJson, void* pObj) {
   if (TSDB_CODE_SUCCESS == code) {
     code = tjsonGetBoolValue(pJson, jkSlotDescOutput, &pNode->output);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = tjsonGetStringValue(pJson, jkSlotDescName, pNode->name);
+  }
 
   return code;
 }
@@ -4137,6 +4160,7 @@ static int32_t specificNodeToJson(const void* pObj, SJson* pJson) {
     case QUERY_NODE_PHYSICAL_PLAN_MERGE:
       return physiMergeNodeToJson(pObj, pJson);
     case QUERY_NODE_PHYSICAL_PLAN_SORT:
+    case QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT:
       return physiSortNodeToJson(pObj, pJson);
     case QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL:
     case QUERY_NODE_PHYSICAL_PLAN_MERGE_ALIGNED_INTERVAL:
@@ -4280,6 +4304,7 @@ static int32_t jsonToSpecificNode(const SJson* pJson, void* pObj) {
     case QUERY_NODE_PHYSICAL_PLAN_MERGE:
       return jsonToPhysiMergeNode(pJson, pObj);
     case QUERY_NODE_PHYSICAL_PLAN_SORT:
+    case QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT:
       return jsonToPhysiSortNode(pJson, pObj);
     case QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL:
     case QUERY_NODE_PHYSICAL_PLAN_MERGE_ALIGNED_INTERVAL:

source/libs/nodes/src/nodesTraverseFuncs.c

@@ -500,7 +500,8 @@ static EDealRes dispatchPhysiPlan(SNode* pNode, ETraversalOrder order, FNodeWalk
       }
       break;
     }
-    case QUERY_NODE_PHYSICAL_PLAN_SORT: {
+    case QUERY_NODE_PHYSICAL_PLAN_SORT:
+    case QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT: {
       SSortPhysiNode* pSort = (SSortPhysiNode*)pNode;
       res = walkPhysiNode((SPhysiNode*)pNode, order, walker, pContext);
       if (DEAL_RES_ERROR != res && DEAL_RES_END != res) {

source/libs/nodes/src/nodesUtilFuncs.c

@@ -288,6 +288,8 @@ SNode* nodesMakeNode(ENodeType type) {
       return makeNode(type, sizeof(SMergePhysiNode));
     case QUERY_NODE_PHYSICAL_PLAN_SORT:
       return makeNode(type, sizeof(SSortPhysiNode));
+    case QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT:
+      return makeNode(type, sizeof(SGroupSortPhysiNode));
     case QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL:
       return makeNode(type, sizeof(SIntervalPhysiNode));
     case QUERY_NODE_PHYSICAL_PLAN_MERGE_ALIGNED_INTERVAL:
@@ -709,7 +711,7 @@ void nodesDestroyNode(SNode* pNode) {
       nodesDestroyNode(pLogicNode->pTagCond);
       nodesDestroyNode(pLogicNode->pTagIndexCond);
       taosArrayDestroy(pLogicNode->pSmaIndexes);
-      nodesDestroyList(pLogicNode->pPartTags);
+      nodesDestroyList(pLogicNode->pGroupTags);
       break;
     }
     case QUERY_NODE_LOGIC_PLAN_JOIN: {
@@ -813,7 +815,7 @@ void nodesDestroyNode(SNode* pNode) {
       STableScanPhysiNode* pPhyNode = (STableScanPhysiNode*)pNode;
       destroyScanPhysiNode((SScanPhysiNode*)pNode);
       nodesDestroyList(pPhyNode->pDynamicScanFuncs);
-      nodesDestroyList(pPhyNode->pPartitionTags);
+      nodesDestroyList(pPhyNode->pGroupTags);
       break;
     }
     case QUERY_NODE_PHYSICAL_PLAN_PROJECT: {
@@ -850,7 +852,8 @@ void nodesDestroyNode(SNode* pNode) {
       nodesDestroyList(pPhyNode->pTargets);
       break;
     }
-    case QUERY_NODE_PHYSICAL_PLAN_SORT: {
+    case QUERY_NODE_PHYSICAL_PLAN_SORT:
+    case QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT: {
       SSortPhysiNode* pPhyNode = (SSortPhysiNode*)pNode;
       destroyPhysiNode((SPhysiNode*)pPhyNode);
       nodesDestroyList(pPhyNode->pExprs);

source/libs/parser/src/parTranslater.c

@@ -2456,90 +2456,6 @@ static int32_t rewriteUniqueStmt(STranslateContext* pCxt, SSelectStmt* pSelect)
   return cxt.pTranslateCxt->errCode;
 }
 
-typedef struct SRwriteTailCxt {
-  STranslateContext* pTranslateCxt;
-  int64_t            limit;
-  int64_t            offset;
-} SRwriteTailCxt;
-
-static EDealRes rewriteTailFunc(SNode** pNode, void* pContext) {
-  SRwriteTailCxt* pCxt = pContext;
-  if (QUERY_NODE_FUNCTION == nodeType(*pNode)) {
-    SFunctionNode* pFunc = (SFunctionNode*)*pNode;
-    if (FUNCTION_TYPE_TAIL == pFunc->funcType) {
-      pCxt->limit = ((SValueNode*)nodesListGetNode(pFunc->pParameterList, 1))->datum.i;
-      if (3 == LIST_LENGTH(pFunc->pParameterList)) {
-        pCxt->offset = ((SValueNode*)nodesListGetNode(pFunc->pParameterList, 2))->datum.i;
-      }
-      SNode* pExpr = nodesListGetNode(pFunc->pParameterList, 0);
-      strcpy(((SExprNode*)pExpr)->aliasName, ((SExprNode*)*pNode)->aliasName);
-      NODES_CLEAR_LIST(pFunc->pParameterList);
-      nodesDestroyNode(*pNode);
-      *pNode = pExpr;
-      return DEAL_RES_IGNORE_CHILD;
-    }
-  }
-  return DEAL_RES_CONTINUE;
-}
-
-static int32_t createLimieNode(SRwriteTailCxt* pCxt, SLimitNode** pOutput) {
-  *pOutput = (SLimitNode*)nodesMakeNode(QUERY_NODE_LIMIT);
-  if (NULL == *pOutput) {
-    return TSDB_CODE_OUT_OF_MEMORY;
-  }
-  (*pOutput)->limit = pCxt->limit;
-  (*pOutput)->offset = pCxt->offset;
-  return TSDB_CODE_SUCCESS;
-}
-
-static SNode* createOrderByExpr(STranslateContext* pCxt) {
-  SOrderByExprNode* pOrder = (SOrderByExprNode*)nodesMakeNode(QUERY_NODE_ORDER_BY_EXPR);
-  if (NULL == pOrder) {
-    return NULL;
-  }
-  pCxt->errCode = createPrimaryKeyCol(pCxt, &pOrder->pExpr);
-  if (TSDB_CODE_SUCCESS != pCxt->errCode) {
-    nodesDestroyNode((SNode*)pOrder);
-    return NULL;
-  }
-  pOrder->order = ORDER_DESC;
-  pOrder->nullOrder = NULL_ORDER_FIRST;
-  return (SNode*)pOrder;
-}
-
-/* case 1:
- * in:  select tail(expr, k, f) from t where_clause
- * out: select expr from t where_clause order by _rowts desc limit k offset f
- *
- * case 2:
- * in:  select tail(expr, k, f) from t where_clause partition_by_clause
- * out: select expr from t where_clause partition_by_clause sort by _rowts desc limit k offset f
- *
- * case 3:
- * in:  select tail(expr, k, f) from t where_clause order_by_clause limit_clause
- * out: select expr from (
- *        select expr from t where_clause order by _rowts desc limit k offset f
- *      ) order_by_clause limit_clause
- *
- * case 4:
- * in:  select tail(expr, k, f) from t where_clause partition_by_clause limit_clause
- * out:
- */
-static int32_t rewriteTailStmt(STranslateContext* pCxt, SSelectStmt* pSelect) {
-  if (!pSelect->hasTailFunc) {
-    return TSDB_CODE_SUCCESS;
-  }
-
-  SRwriteTailCxt cxt = {.pTranslateCxt = pCxt, .limit = -1, .offset = -1};
-  nodesRewriteExprs(pSelect->pProjectionList, rewriteTailFunc, &cxt);
-  int32_t code = nodesListMakeStrictAppend(&pSelect->pOrderByList, createOrderByExpr(pCxt));
-  if (TSDB_CODE_SUCCESS == code) {
-    code = createLimieNode(&cxt, &pSelect->pLimit);
-  }
-  pSelect->hasIndefiniteRowsFunc = false;
-  return code;
-}
-
 typedef struct SReplaceOrderByAliasCxt {
   STranslateContext* pTranslateCxt;
   SNodeList*         pProjectionList;
@@ -2616,9 +2532,6 @@ static int32_t translateSelectFrom(STranslateContext* pCxt, SSelectStmt* pSelect
   if (TSDB_CODE_SUCCESS == code) {
     code = rewriteUniqueStmt(pCxt, pSelect);
   }
-  // if (TSDB_CODE_SUCCESS == code) {
-  //   code = rewriteTailStmt(pCxt, pSelect);
-  // }
   if (TSDB_CODE_SUCCESS == code) {
     code = rewriteTimelineFunc(pCxt, pSelect);
   }

source/libs/planner/src/planLogicCreater.c

@@ -507,6 +507,8 @@ static int32_t createIndefRowsFuncLogicNode(SLogicPlanContext* pCxt, SSelectStmt
     return TSDB_CODE_OUT_OF_MEMORY;
   }
 
+  pIdfRowsFunc->isTailFunc = pSelect->hasTailFunc;
+
   // indefinite rows functions and _select_values functions
   int32_t code = nodesCollectFuncs(pSelect, SQL_CLAUSE_SELECT, fmIsVectorFunc, &pIdfRowsFunc->pFuncs);
   if (TSDB_CODE_SUCCESS == code) {
@@ -733,6 +735,8 @@ static int32_t createSortLogicNode(SLogicPlanContext* pCxt, SSelectStmt* pSelect
     return TSDB_CODE_OUT_OF_MEMORY;
   }
 
+  pSort->groupSort = pSelect->groupSort;
+
   int32_t code = nodesCollectColumns(pSelect, SQL_CLAUSE_ORDER_BY, NULL, COLLECT_COL_TYPE_ALL, &pSort->node.pTargets);
   if (TSDB_CODE_SUCCESS == code && NULL == pSort->node.pTargets) {
     code = nodesListMakeStrictAppend(&pSort->node.pTargets,

source/libs/planner/src/planPhysiCreater.c

@@ -41,8 +41,12 @@ typedef struct SPhysiPlanContext {
 static int32_t getSlotKey(SNode* pNode, const char* pStmtName, char* pKey) {
   if (QUERY_NODE_COLUMN == nodeType(pNode)) {
     SColumnNode* pCol = (SColumnNode*)pNode;
-    if (NULL != pStmtName && '\0' != pStmtName[0]) {
+    if (NULL != pStmtName) {
+      if ('\0' != pStmtName[0]) {
         return sprintf(pKey, "%s.%s", pStmtName, pCol->node.aliasName);
+      } else {
+        return sprintf(pKey, "%s", pCol->node.aliasName);
+      }
     }
     if ('\0' == pCol->tableAlias[0]) {
       return sprintf(pKey, "%s", pCol->colName);
@@ -56,11 +60,13 @@ static int32_t getSlotKey(SNode* pNode, const char* pStmtName, char* pKey) {
   return sprintf(pKey, "%s", ((SExprNode*)pNode)->aliasName);
 }
 
-static SNode* createSlotDesc(SPhysiPlanContext* pCxt, const SNode* pNode, int16_t slotId, bool output, bool reserve) {
+static SNode* createSlotDesc(SPhysiPlanContext* pCxt, const char* pName, const SNode* pNode, int16_t slotId,
+                             bool output, bool reserve) {
   SSlotDescNode* pSlot = (SSlotDescNode*)nodesMakeNode(QUERY_NODE_SLOT_DESC);
   if (NULL == pSlot) {
     return NULL;
   }
+  strcpy(pSlot->name, pName);
   pSlot->slotId = slotId;
   pSlot->dataType = ((SExprNode*)pNode)->resType;
   pSlot->reserve = reserve;
@@ -99,10 +105,8 @@ static int32_t putSlotToHashImpl(int16_t dataBlockId, int16_t slotId, const char
   return taosHashPut(pHash, pName, len, &index, sizeof(SSlotIndex));
 }
 
-static int32_t putSlotToHash(int16_t dataBlockId, int16_t slotId, SNode* pNode, SHashObj* pHash) {
-  char    name[TSDB_TABLE_NAME_LEN + TSDB_COL_NAME_LEN];
-  int32_t len = getSlotKey(pNode, NULL, name);
-  return putSlotToHashImpl(dataBlockId, slotId, name, len, pHash);
+static int32_t putSlotToHash(const char* pName, int16_t dataBlockId, int16_t slotId, SNode* pNode, SHashObj* pHash) {
+  return putSlotToHashImpl(dataBlockId, slotId, pName, strlen(pName), pHash);
 }
 
 static int32_t createDataBlockDescHash(SPhysiPlanContext* pCxt, int32_t capacity, int16_t dataBlockId,
@@ -131,9 +135,11 @@ static int32_t buildDataBlockSlots(SPhysiPlanContext* pCxt, SNodeList* pList, SD
   int16_t slotId = 0;
   SNode*  pNode = NULL;
   FOREACH(pNode, pList) {
-    code = nodesListStrictAppend(pDataBlockDesc->pSlots, createSlotDesc(pCxt, pNode, slotId, true, false));
+    char name[TSDB_TABLE_NAME_LEN + TSDB_COL_NAME_LEN];
+    getSlotKey(pNode, NULL, name);
+    code = nodesListStrictAppend(pDataBlockDesc->pSlots, createSlotDesc(pCxt, name, pNode, slotId, true, false));
     if (TSDB_CODE_SUCCESS == code) {
-      code = putSlotToHash(pDataBlockDesc->dataBlockId, slotId, pNode, pHash);
+      code = putSlotToHash(name, pDataBlockDesc->dataBlockId, slotId, pNode, pHash);
     }
     if (TSDB_CODE_SUCCESS == code) {
       pDataBlockDesc->totalRowSize += ((SExprNode*)pNode)->resType.bytes;
@@ -196,7 +202,8 @@ static int32_t addDataBlockSlotsImpl(SPhysiPlanContext* pCxt, SNodeList* pList,
     int32_t     len = getSlotKey(pExpr, pStmtName, name);
     SSlotIndex* pIndex = taosHashGet(pHash, name, len);
     if (NULL == pIndex) {
-      code = nodesListStrictAppend(pDataBlockDesc->pSlots, createSlotDesc(pCxt, pExpr, nextSlotId, output, reserve));
+      code =
+          nodesListStrictAppend(pDataBlockDesc->pSlots, createSlotDesc(pCxt, name, pExpr, nextSlotId, output, reserve));
       if (TSDB_CODE_SUCCESS == code) {
         code = putSlotToHashImpl(pDataBlockDesc->dataBlockId, nextSlotId, name, len, pHash);
       }
@@ -513,12 +520,13 @@ static int32_t createTableScanPhysiNode(SPhysiPlanContext* pCxt, SSubplan* pSubp
   tNameGetFullDbName(&pScanLogicNode->tableName, pSubplan->dbFName);
   pTableScan->dataRequired = pScanLogicNode->dataRequired;
   pTableScan->pDynamicScanFuncs = nodesCloneList(pScanLogicNode->pDynamicScanFuncs);
-  pTableScan->pPartitionTags = nodesCloneList(pScanLogicNode->pPartTags);
+  pTableScan->pGroupTags = nodesCloneList(pScanLogicNode->pGroupTags);
   if ((NULL != pScanLogicNode->pDynamicScanFuncs && NULL == pTableScan->pDynamicScanFuncs) ||
-      (NULL != pScanLogicNode->pPartTags && NULL == pTableScan->pPartitionTags)) {
+      (NULL != pScanLogicNode->pGroupTags && NULL == pTableScan->pGroupTags)) {
     nodesDestroyNode((SNode*)pTableScan);
     return TSDB_CODE_OUT_OF_MEMORY;
   }
+  pTableScan->groupSort = pScanLogicNode->groupSort;
   pTableScan->interval = pScanLogicNode->interval;
   pTableScan->offset = pScanLogicNode->offset;
   pTableScan->sliding = pScanLogicNode->sliding;
@@ -1170,8 +1178,9 @@ static int32_t createWindowPhysiNode(SPhysiPlanContext* pCxt, SNodeList* pChildr
 
 static int32_t createSortPhysiNode(SPhysiPlanContext* pCxt, SNodeList* pChildren, SSortLogicNode* pSortLogicNode,
                                    SPhysiNode** pPhyNode) {
-  SSortPhysiNode* pSort =
-      (SSortPhysiNode*)makePhysiNode(pCxt, (SLogicNode*)pSortLogicNode, QUERY_NODE_PHYSICAL_PLAN_SORT);
+  SSortPhysiNode* pSort = (SSortPhysiNode*)makePhysiNode(
+      pCxt, (SLogicNode*)pSortLogicNode,
+      pSortLogicNode->groupSort ? QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT : QUERY_NODE_PHYSICAL_PLAN_SORT);
   if (NULL == pSort) {
     return TSDB_CODE_OUT_OF_MEMORY;
   }
@@ -1185,7 +1194,7 @@ static int32_t createSortPhysiNode(SPhysiPlanContext* pCxt, SNodeList* pChildren
   if (TSDB_CODE_SUCCESS == code && NULL != pPrecalcExprs) {
     code = setListSlotId(pCxt, pChildTupe->dataBlockId, -1, pPrecalcExprs, &pSort->pExprs);
     if (TSDB_CODE_SUCCESS == code) {
-      code = addDataBlockSlots(pCxt, pSort->pExprs, pChildTupe);
+      code = pushdownDataBlockSlots(pCxt, pSort->pExprs, pChildTupe);
     }
   }
 
@@ -1322,6 +1331,7 @@ static int32_t createMergePhysiNode(SPhysiPlanContext* pCxt, SMergeLogicNode* pM
 
   pMerge->numOfChannels = pMergeLogicNode->numOfChannels;
   pMerge->srcGroupId = pMergeLogicNode->srcGroupId;
+  pMerge->groupSort = pMergeLogicNode->groupSort;
 
   int32_t code = addDataBlockSlots(pCxt, pMergeLogicNode->pInputs, pMerge->node.pOutputDataBlockDesc);
 

source/libs/planner/src/planSpliter.c

@@ -362,7 +362,7 @@ static int32_t stbSplGetNumOfVgroups(SLogicNode* pNode) {
 }
 
 static int32_t stbSplCreateMergeNode(SSplitContext* pCxt, SLogicSubplan* pSubplan, SLogicNode* pSplitNode,
-                                     SNodeList* pMergeKeys, SLogicNode* pPartChild) {
+                                     SNodeList* pMergeKeys, SLogicNode* pPartChild, bool groupSort) {
   SMergeLogicNode* pMerge = (SMergeLogicNode*)nodesMakeNode(QUERY_NODE_LOGIC_PLAN_MERGE);
   if (NULL == pMerge) {
     return TSDB_CODE_OUT_OF_MEMORY;
@@ -371,6 +371,7 @@ static int32_t stbSplCreateMergeNode(SSplitContext* pCxt, SLogicSubplan* pSubpla
   pMerge->srcGroupId = pCxt->groupId;
   pMerge->node.precision = pPartChild->precision;
   pMerge->pMergeKeys = pMergeKeys;
+  pMerge->groupSort = groupSort;
 
   int32_t code = TSDB_CODE_SUCCESS;
   pMerge->pInputs = nodesCloneList(pPartChild->pTargets);
@@ -430,7 +431,7 @@ static int32_t stbSplSplitIntervalForBatch(SSplitContext* pCxt, SStableSplitInfo
     SNodeList* pMergeKeys = NULL;
     code = stbSplCreateMergeKeysByPrimaryKey(((SWindowLogicNode*)pInfo->pSplitNode)->pTspk, &pMergeKeys);
     if (TSDB_CODE_SUCCESS == code) {
-      code = stbSplCreateMergeNode(pCxt, NULL, pInfo->pSplitNode, pMergeKeys, pPartWindow);
+      code = stbSplCreateMergeNode(pCxt, NULL, pInfo->pSplitNode, pMergeKeys, pPartWindow, false);
     }
     if (TSDB_CODE_SUCCESS != code) {
       nodesDestroyList(pMergeKeys);
@@ -497,12 +498,16 @@ static int32_t stbSplSplitSessionForStream(SSplitContext* pCxt, SStableSplitInfo
   return code;
 }
 
-static void splSetTableScanType(SLogicNode* pNode, EScanType scanType) {
+static void stbSplSetTableMergeScan(SLogicNode* pNode) {
   if (QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pNode)) {
-    ((SScanLogicNode*)pNode)->scanType = scanType;
+    SScanLogicNode* pScan = (SScanLogicNode*)pNode;
+    pScan->scanType = SCAN_TYPE_TABLE_MERGE;
+    if (NULL != pScan->pGroupTags) {
+      pScan->groupSort = true;
+    }
   } else {
     if (1 == LIST_LENGTH(pNode->pChildren)) {
-      splSetTableScanType((SLogicNode*)nodesListGetNode(pNode->pChildren, 0), scanType);
+      stbSplSetTableMergeScan((SLogicNode*)nodesListGetNode(pNode->pChildren, 0));
     }
   }
 }
@@ -515,7 +520,7 @@ static int32_t stbSplSplitSessionOrStateForBatch(SSplitContext* pCxt, SStableSpl
   int32_t    code = stbSplCreateMergeKeysByPrimaryKey(((SWindowLogicNode*)pWindow)->pTspk, &pMergeKeys);
 
   if (TSDB_CODE_SUCCESS == code) {
-    code = stbSplCreateMergeNode(pCxt, pInfo->pSubplan, pChild, pMergeKeys, (SLogicNode*)pChild);
+    code = stbSplCreateMergeNode(pCxt, pInfo->pSubplan, pChild, pMergeKeys, (SLogicNode*)pChild, true);
   }
 
   if (TSDB_CODE_SUCCESS == code) {
@@ -524,13 +529,10 @@ static int32_t stbSplSplitSessionOrStateForBatch(SSplitContext* pCxt, SStableSpl
   }
 
   if (TSDB_CODE_SUCCESS == code) {
-    splSetTableScanType(pChild, SCAN_TYPE_TABLE_MERGE);
-    ++(pCxt->groupId);
-  }
-
-  if (TSDB_CODE_SUCCESS == code) {
+    stbSplSetTableMergeScan(pChild);
     pInfo->pSubplan->subplanType = SUBPLAN_TYPE_MERGE;
     SPLIT_FLAG_SET_MASK(pInfo->pSubplan->splitFlag, SPLIT_FLAG_STABLE_SPLIT);
+    ++(pCxt->groupId);
   } else {
     nodesDestroyList(pMergeKeys);
   }
@@ -560,7 +562,7 @@ static int32_t stbSplSplitState(SSplitContext* pCxt, SStableSplitInfo* pInfo) {
 
 static SNodeList* stbSplGetPartKeys(SLogicNode* pNode) {
   if (QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pNode)) {
-    return ((SScanLogicNode*)pNode)->pPartTags;
+    return ((SScanLogicNode*)pNode)->pGroupTags;
   } else if (QUERY_NODE_LOGIC_PLAN_PARTITION == nodeType(pNode)) {
     return ((SPartitionLogicNode*)pNode)->pPartitionKeys;
   } else {
@@ -775,6 +777,7 @@ static int32_t stbSplCreatePartSortNode(SSortLogicNode* pSort, SLogicNode** pOut
     pPartSort->node.pChildren = pChildren;
     splSetParent((SLogicNode*)pPartSort);
     pPartSort->pSortKeys = pSortKeys;
+    pPartSort->groupSort = pSort->groupSort;
     code = stbSplCreateMergeKeys(pPartSort->pSortKeys, pPartSort->node.pTargets, &pMergeKeys);
   }
 
@@ -789,12 +792,29 @@ static int32_t stbSplCreatePartSortNode(SSortLogicNode* pSort, SLogicNode** pOut
   return code;
 }
 
+static void stbSplSetScanPartSort(SLogicNode* pNode) {
+  if (QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pNode)) {
+    SScanLogicNode* pScan = (SScanLogicNode*)pNode;
+    if (NULL != pScan->pGroupTags) {
+      pScan->groupSort = true;
+    }
+  } else {
+    if (1 == LIST_LENGTH(pNode->pChildren)) {
+      stbSplSetScanPartSort((SLogicNode*)nodesListGetNode(pNode->pChildren, 0));
+    }
+  }
+}
+
 static int32_t stbSplSplitSortNode(SSplitContext* pCxt, SStableSplitInfo* pInfo) {
   SLogicNode* pPartSort = NULL;
   SNodeList*  pMergeKeys = NULL;
+  bool        groupSort = ((SSortLogicNode*)pInfo->pSplitNode)->groupSort;
   int32_t     code = stbSplCreatePartSortNode((SSortLogicNode*)pInfo->pSplitNode, &pPartSort, &pMergeKeys);
   if (TSDB_CODE_SUCCESS == code) {
-    code = stbSplCreateMergeNode(pCxt, pInfo->pSubplan, pInfo->pSplitNode, pMergeKeys, pPartSort);
+    code = stbSplCreateMergeNode(pCxt, pInfo->pSubplan, pInfo->pSplitNode, pMergeKeys, pPartSort, groupSort);
+  }
+  if (TSDB_CODE_SUCCESS == code && groupSort) {
+    stbSplSetScanPartSort(pPartSort);
   }
   if (TSDB_CODE_SUCCESS == code) {
     code = nodesListMakeStrictAppend(&pInfo->pSubplan->pChildren,
@@ -829,7 +849,7 @@ static int32_t stbSplSplitScanNodeForJoin(SSplitContext* pCxt, SLogicSubplan* pS
   SNodeList* pMergeKeys = NULL;
   int32_t    code = stbSplCreateMergeKeysByPrimaryKey(stbSplFindPrimaryKeyFromScan(pScan), &pMergeKeys);
   if (TSDB_CODE_SUCCESS == code) {
-    code = stbSplCreateMergeNode(pCxt, pSubplan, (SLogicNode*)pScan, pMergeKeys, (SLogicNode*)pScan);
+    code = stbSplCreateMergeNode(pCxt, pSubplan, (SLogicNode*)pScan, pMergeKeys, (SLogicNode*)pScan, false);
   }
   if (TSDB_CODE_SUCCESS == code) {
     code = nodesListMakeStrictAppend(&pSubplan->pChildren,

source/libs/planner/test/planBasicTest.cpp

@@ -67,6 +67,14 @@ TEST_F(PlanBasicTest, tailFunc) {
   run("SELECT TAIL(c1, 10) FROM t1");
 
   run("SELECT TAIL(c2 + 10, 10, 80) FROM t1 WHERE c1 > 10");
+
+  run("SELECT TAIL(c2 + 10, 10, 80) FROM t1 WHERE c1 > 10 PARTITION BY c1");
+
+  run("SELECT TAIL(c2 + 10, 10, 80) FROM t1 WHERE c1 > 10 ORDER BY 1");
+
+  run("SELECT TAIL(c2 + 10, 10, 80) FROM t1 WHERE c1 > 10 LIMIT 5");
+
+  run("SELECT TAIL(c2 + 10, 10, 80) FROM t1 WHERE c1 > 10 PARTITION BY c1 LIMIT 5");
 }
 
 TEST_F(PlanBasicTest, interpFunc) {

source/libs/planner/test/planTestMain.cpp

@@ -76,6 +76,7 @@ static void parseArg(int argc, char* argv[]) {
   static struct option long_options[] = {
     {"dump", optional_argument, NULL, 'd'},
     {"skipSql", required_argument, NULL, 's'},
+    {"limitSql", required_argument, NULL, 'i'},
     {"log", required_argument, NULL, 'l'},
     {0, 0, 0, 0}
   };
@@ -88,6 +89,9 @@ static void parseArg(int argc, char* argv[]) {
       case 's':
         setSkipSqlNum(optarg);
         break;
+      case 'i':
+        setLimitSqlNum(optarg);
+        break;
       case 'l':
         setLogLevel(optarg);
         break;

source/libs/planner/test/planTestUtil.cpp

@@ -51,6 +51,7 @@ enum DumpModule {
 
 DumpModule g_dumpModule = DUMP_MODULE_NOTHING;
 int32_t    g_skipSql = 0;
+int32_t    g_limitSql = 0;
 int32_t    g_logLevel = 131;
 
 void setDumpModule(const char* pModule) {
@@ -76,28 +77,33 @@ void setDumpModule(const char* pModule) {
 }
 
 void setSkipSqlNum(const char* pNum) { g_skipSql = stoi(pNum); }
-
+void setLimitSqlNum(const char* pNum) { g_limitSql = stoi(pNum); }
 void setLogLevel(const char* pLogLevel) { g_logLevel = stoi(pLogLevel); }
 
 int32_t getLogLevel() { return g_logLevel; }
 
 class PlannerTestBaseImpl {
  public:
-  PlannerTestBaseImpl() : sqlNo_(0) {}
+  PlannerTestBaseImpl() : sqlNo_(0), sqlNum_(0) {}
 
   void useDb(const string& user, const string& db) {
     caseEnv_.acctId_ = 0;
     caseEnv_.user_ = user;
     caseEnv_.db_ = db;
-    caseEnv_.nsql_ = g_skipSql;
+    caseEnv_.numOfSkipSql_ = g_skipSql;
+    caseEnv_.numOfLimitSql_ = g_limitSql;
   }
 
   void run(const string& sql) {
     ++sqlNo_;
-    if (caseEnv_.nsql_ > 0) {
-      --(caseEnv_.nsql_);
+    if (caseEnv_.numOfSkipSql_ > 0) {
+      --(caseEnv_.numOfSkipSql_);
       return;
     }
+    if (caseEnv_.numOfLimitSql_ > 0 && caseEnv_.numOfLimitSql_ == sqlNum_) {
+      return;
+    }
+    ++sqlNum_;
 
     reset();
     try {
@@ -134,7 +140,7 @@ class PlannerTestBaseImpl {
   }
 
   void prepare(const string& sql) {
-    if (caseEnv_.nsql_ > 0) {
+    if (caseEnv_.numOfSkipSql_ > 0) {
       return;
     }
 
@@ -148,7 +154,7 @@ class PlannerTestBaseImpl {
   }
 
   void bindParams(TAOS_MULTI_BIND* pParams, int32_t colIdx) {
-    if (caseEnv_.nsql_ > 0) {
+    if (caseEnv_.numOfSkipSql_ > 0) {
       return;
     }
 
@@ -161,8 +167,8 @@ class PlannerTestBaseImpl {
   }
 
   void exec() {
-    if (caseEnv_.nsql_ > 0) {
-      --(caseEnv_.nsql_);
+    if (caseEnv_.numOfSkipSql_ > 0) {
+      --(caseEnv_.numOfSkipSql_);
       return;
     }
 
@@ -197,9 +203,10 @@ class PlannerTestBaseImpl {
     int32_t acctId_;
     string  user_;
     string  db_;
-    int32_t nsql_;
+    int32_t numOfSkipSql_;
+    int32_t numOfLimitSql_;
 
-    caseEnv() : nsql_(0) {}
+    caseEnv() : numOfSkipSql_(0) {}
   };
 
   struct stmtEnv {
@@ -401,6 +408,7 @@ class PlannerTestBaseImpl {
   stmtEnv stmtEnv_;
   stmtRes res_;
   int32_t sqlNo_;
+  int32_t sqlNum_;
 };
 
 PlannerTestBase::PlannerTestBase() : impl_(new PlannerTestBaseImpl()) {}

source/libs/planner/test/planTestUtil.h

@@ -43,6 +43,7 @@ class PlannerTestBase : public testing::Test {
 
 extern void    setDumpModule(const char* pModule);
 extern void    setSkipSqlNum(const char* pNum);
+extern void    setLimitSqlNum(const char* pNum);
 extern void    setLogLevel(const char* pLogLevel);
 extern int32_t getLogLevel();
 

source/libs/sync/inc/syncRaftEntry.h

@@ -56,6 +56,31 @@ void syncEntryPrint2(char* s, const SSyncRaftEntry* pObj);
 void syncEntryLog(const SSyncRaftEntry* pObj);
 void syncEntryLog2(char* s, const SSyncRaftEntry* pObj);
 
+//-----------------------------------
+typedef struct SRaftEntryCache {
+  SHashObj*     pEntryHash;
+  int32_t       maxCount;
+  int32_t       currentCount;
+  TdThreadMutex mutex;
+  SSyncNode*    pSyncNode;
+} SRaftEntryCache;
+
+SRaftEntryCache* raftCacheCreate(SSyncNode* pSyncNode, int32_t maxCount);
+void             raftCacheDestroy(SRaftEntryCache* pCache);
+int32_t          raftCachePutEntry(struct SRaftEntryCache* pCache, SSyncRaftEntry* pEntry);
+int32_t          raftCacheGetEntry(struct SRaftEntryCache* pCache, SyncIndex index, SSyncRaftEntry** ppEntry);
+int32_t          raftCacheGetEntryP(struct SRaftEntryCache* pCache, SyncIndex index, SSyncRaftEntry** ppEntry);
+int32_t          raftCacheDelEntry(struct SRaftEntryCache* pCache, SyncIndex index);
+int32_t          raftCacheGetAndDel(struct SRaftEntryCache* pCache, SyncIndex index, SSyncRaftEntry** ppEntry);
+int32_t          raftCacheClear(struct SRaftEntryCache* pCache);
+
+cJSON* raftCache2Json(SRaftEntryCache* pObj);
+char*  raftCache2Str(SRaftEntryCache* pObj);
+void   raftCachePrint(SRaftEntryCache* pObj);
+void   raftCachePrint2(char* s, SRaftEntryCache* pObj);
+void   raftCacheLog(SRaftEntryCache* pObj);
+void   raftCacheLog2(char* s, SRaftEntryCache* pObj);
+
 #ifdef __cplusplus
 }
 #endif

source/libs/sync/src/syncMain.c

@@ -1355,11 +1355,16 @@ void syncNodeEventLog(const SSyncNode* pSyncNode, char* str) {
   int32_t userStrLen = strlen(str);
 
   SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
-  if (pSyncNode->pFsm->FpGetSnapshotInfo != NULL) {
+  if (pSyncNode->pFsm != NULL && pSyncNode->pFsm->FpGetSnapshotInfo != NULL) {
     pSyncNode->pFsm->FpGetSnapshotInfo(pSyncNode->pFsm, &snapshot);
   }
-  SyncIndex logLastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
-  SyncIndex logBeginIndex = pSyncNode->pLogStore->syncLogBeginIndex(pSyncNode->pLogStore);
+
+  SyncIndex logLastIndex = SYNC_INDEX_INVALID;
+  SyncIndex logBeginIndex = SYNC_INDEX_INVALID;
+  if (pSyncNode->pLogStore != NULL) {
+    logLastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
+    logBeginIndex = pSyncNode->pLogStore->syncLogBeginIndex(pSyncNode->pLogStore);
+  }
 
   char* pCfgStr = syncCfg2SimpleStr(&(pSyncNode->pRaftCfg->cfg));
   char* printStr = "";

source/libs/sync/src/syncRaftEntry.c

@@ -180,3 +180,247 @@ void syncEntryLog2(char* s, const SSyncRaftEntry* pObj) {
   sTrace("syncEntryLog2 | len:%zu | %s | %s", strlen(serialized), s, serialized);
   taosMemoryFree(serialized);
 }
+
+//-----------------------------------
+SRaftEntryCache* raftCacheCreate(SSyncNode* pSyncNode, int32_t maxCount) {
+  SRaftEntryCache* pCache = taosMemoryMalloc(sizeof(SRaftEntryCache));
+  if (pCache == NULL) {
+    sError("vgId:%d raft cache create error", pSyncNode->vgId);
+    return NULL;
+  }
+
+  pCache->pEntryHash =
+      taosHashInit(sizeof(SyncIndex), taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_NO_LOCK);
+  if (pCache->pEntryHash == NULL) {
+    sError("vgId:%d raft cache create hash error", pSyncNode->vgId);
+    return NULL;
+  }
+
+  taosThreadMutexInit(&(pCache->mutex), NULL);
+  pCache->maxCount = maxCount;
+  pCache->currentCount = 0;
+  pCache->pSyncNode = pSyncNode;
+
+  return pCache;
+}
+
+void raftCacheDestroy(SRaftEntryCache* pCache) {
+  if (pCache != NULL) {
+    taosThreadMutexLock(&(pCache->mutex));
+    taosHashCleanup(pCache->pEntryHash);
+    taosThreadMutexUnlock(&(pCache->mutex));
+    taosThreadMutexDestroy(&(pCache->mutex));
+    taosMemoryFree(pCache);
+  }
+}
+
+// success, return 1
+// max count, return 0
+// error, return -1
+int32_t raftCachePutEntry(struct SRaftEntryCache* pCache, SSyncRaftEntry* pEntry) {
+  taosThreadMutexLock(&(pCache->mutex));
+
+  if (pCache->currentCount >= pCache->maxCount) {
+    taosThreadMutexUnlock(&(pCache->mutex));
+    return 0;
+  }
+
+  taosHashPut(pCache->pEntryHash, &(pEntry->index), sizeof(pEntry->index), pEntry, pEntry->bytes);
+  ++(pCache->currentCount);
+
+  do {
+    char eventLog[128];
+    snprintf(eventLog, sizeof(eventLog), "raft cache add, type:%s,%d, type2:%s,%d, index:%ld, bytes:%d",
+             TMSG_INFO(pEntry->msgType), pEntry->msgType, TMSG_INFO(pEntry->originalRpcType), pEntry->originalRpcType,
+             pEntry->index, pEntry->bytes);
+    syncNodeEventLog(pCache->pSyncNode, eventLog);
+  } while (0);
+
+  taosThreadMutexUnlock(&(pCache->mutex));
+  return 1;
+}
+
+// success, return 0
+// error, return -1
+// not exist, return -1, terrno = TSDB_CODE_WAL_LOG_NOT_EXIST
+int32_t raftCacheGetEntry(struct SRaftEntryCache* pCache, SyncIndex index, SSyncRaftEntry** ppEntry) {
+  if (ppEntry == NULL) {
+    return -1;
+  }
+  *ppEntry = NULL;
+
+  taosThreadMutexLock(&(pCache->mutex));
+  void* pTmp = taosHashGet(pCache->pEntryHash, &index, sizeof(index));
+  if (pTmp != NULL) {
+    SSyncRaftEntry* pEntry = pTmp;
+    *ppEntry = taosMemoryMalloc(pEntry->bytes);
+    memcpy(*ppEntry, pTmp, pEntry->bytes);
+
+    do {
+      char eventLog[128];
+      snprintf(eventLog, sizeof(eventLog), "raft cache get, type:%s,%d, type2:%s,%d, index:%ld",
+               TMSG_INFO((*ppEntry)->msgType), (*ppEntry)->msgType, TMSG_INFO((*ppEntry)->originalRpcType),
+               (*ppEntry)->originalRpcType, (*ppEntry)->index);
+      syncNodeEventLog(pCache->pSyncNode, eventLog);
+    } while (0);
+
+    taosThreadMutexUnlock(&(pCache->mutex));
+    return 0;
+  }
+
+  taosThreadMutexUnlock(&(pCache->mutex));
+  terrno = TSDB_CODE_WAL_LOG_NOT_EXIST;
+  return -1;
+}
+
+// success, return 0
+// error, return -1
+// not exist, return -1, terrno = TSDB_CODE_WAL_LOG_NOT_EXIST
+int32_t raftCacheGetEntryP(struct SRaftEntryCache* pCache, SyncIndex index, SSyncRaftEntry** ppEntry) {
+  if (ppEntry == NULL) {
+    return -1;
+  }
+  *ppEntry = NULL;
+
+  taosThreadMutexLock(&(pCache->mutex));
+  void* pTmp = taosHashGet(pCache->pEntryHash, &index, sizeof(index));
+  if (pTmp != NULL) {
+    SSyncRaftEntry* pEntry = pTmp;
+    *ppEntry = pEntry;
+
+    do {
+      char eventLog[128];
+      snprintf(eventLog, sizeof(eventLog), "raft cache get, type:%s,%d, type2:%s,%d, index:%ld",
+               TMSG_INFO((*ppEntry)->msgType), (*ppEntry)->msgType, TMSG_INFO((*ppEntry)->originalRpcType),
+               (*ppEntry)->originalRpcType, (*ppEntry)->index);
+      syncNodeEventLog(pCache->pSyncNode, eventLog);
+    } while (0);
+
+    taosThreadMutexUnlock(&(pCache->mutex));
+    return 0;
+  }
+
+  taosThreadMutexUnlock(&(pCache->mutex));
+  terrno = TSDB_CODE_WAL_LOG_NOT_EXIST;
+  return -1;
+}
+
+int32_t raftCacheDelEntry(struct SRaftEntryCache* pCache, SyncIndex index) {
+  taosThreadMutexLock(&(pCache->mutex));
+  taosHashRemove(pCache->pEntryHash, &index, sizeof(index));
+  --(pCache->currentCount);
+  taosThreadMutexUnlock(&(pCache->mutex));
+  return 0;
+}
+
+int32_t raftCacheGetAndDel(struct SRaftEntryCache* pCache, SyncIndex index, SSyncRaftEntry** ppEntry) {
+  if (ppEntry == NULL) {
+    return -1;
+  }
+  *ppEntry = NULL;
+
+  taosThreadMutexLock(&(pCache->mutex));
+  void* pTmp = taosHashGet(pCache->pEntryHash, &index, sizeof(index));
+  if (pTmp != NULL) {
+    SSyncRaftEntry* pEntry = pTmp;
+    *ppEntry = taosMemoryMalloc(pEntry->bytes);
+    memcpy(*ppEntry, pTmp, pEntry->bytes);
+
+    do {
+      char eventLog[128];
+      snprintf(eventLog, sizeof(eventLog), "raft cache get-and-del, type:%s,%d, type2:%s,%d, index:%ld",
+               TMSG_INFO((*ppEntry)->msgType), (*ppEntry)->msgType, TMSG_INFO((*ppEntry)->originalRpcType),
+               (*ppEntry)->originalRpcType, (*ppEntry)->index);
+      syncNodeEventLog(pCache->pSyncNode, eventLog);
+    } while (0);
+
+    taosHashRemove(pCache->pEntryHash, &index, sizeof(index));
+    --(pCache->currentCount);
+
+    taosThreadMutexUnlock(&(pCache->mutex));
+    return 0;
+  }
+
+  taosThreadMutexUnlock(&(pCache->mutex));
+  terrno = TSDB_CODE_WAL_LOG_NOT_EXIST;
+  return -1;
+}
+
+int32_t raftCacheClear(struct SRaftEntryCache* pCache) {
+  taosThreadMutexLock(&(pCache->mutex));
+  taosHashClear(pCache->pEntryHash);
+  pCache->currentCount = 0;
+  taosThreadMutexUnlock(&(pCache->mutex));
+  return 0;
+}
+
+//-----------------------------------
+cJSON* raftCache2Json(SRaftEntryCache* pCache) {
+  char   u64buf[128] = {0};
+  cJSON* pRoot = cJSON_CreateObject();
+
+  if (pCache != NULL) {
+    taosThreadMutexLock(&(pCache->mutex));
+
+    snprintf(u64buf, sizeof(u64buf), "%p", pCache->pSyncNode);
+    cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
+    cJSON_AddNumberToObject(pRoot, "currentCount", pCache->currentCount);
+    cJSON_AddNumberToObject(pRoot, "maxCount", pCache->maxCount);
+    cJSON* pEntries = cJSON_CreateArray();
+    cJSON_AddItemToObject(pRoot, "entries", pEntries);
+
+    SSyncRaftEntry* pIter = (SSyncRaftEntry*)taosHashIterate(pCache->pEntryHash, NULL);
+    if (pIter != NULL) {
+      SSyncRaftEntry* pEntry = (SSyncRaftEntry*)pIter;
+      cJSON_AddItemToArray(pEntries, syncEntry2Json(pEntry));
+    }
+    while (pIter) {
+      pIter = taosHashIterate(pCache->pEntryHash, pIter);
+      if (pIter != NULL) {
+        SSyncRaftEntry* pEntry = (SSyncRaftEntry*)pIter;
+        cJSON_AddItemToArray(pEntries, syncEntry2Json(pEntry));
+      }
+    }
+
+    taosThreadMutexUnlock(&(pCache->mutex));
+  }
+
+  cJSON* pJson = cJSON_CreateObject();
+  cJSON_AddItemToObject(pJson, "SRaftEntryCache", pRoot);
+  return pJson;
+}
+
+char* raftCache2Str(SRaftEntryCache* pCache) {
+  cJSON* pJson = raftCache2Json(pCache);
+  char*  serialized = cJSON_Print(pJson);
+  cJSON_Delete(pJson);
+  return serialized;
+}
+
+void raftCachePrint(SRaftEntryCache* pCache) {
+  char* serialized = raftCache2Str(pCache);
+  printf("raftCachePrint | len:%lu | %s \n", strlen(serialized), serialized);
+  fflush(NULL);
+  taosMemoryFree(serialized);
+}
+
+void raftCachePrint2(char* s, SRaftEntryCache* pCache) {
+  char* serialized = raftCache2Str(pCache);
+  printf("raftCachePrint2 | len:%lu | %s | %s \n", strlen(serialized), s, serialized);
+  fflush(NULL);
+  taosMemoryFree(serialized);
+}
+
+void raftCacheLog(SRaftEntryCache* pCache) {
+  char* serialized = raftCache2Str(pCache);
+  sTrace("raftCacheLog | len:%lu | %s", strlen(serialized), serialized);
+  taosMemoryFree(serialized);
+}
+
+void raftCacheLog2(char* s, SRaftEntryCache* pCache) {
+  if (gRaftDetailLog) {
+    char* serialized = raftCache2Str(pCache);
+    sTraceLong("raftCacheLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
+    taosMemoryFree(serialized);
+  }
+}

source/libs/sync/src/syncRespMgr.c

@@ -32,12 +32,14 @@ SSyncRespMgr *syncRespMgrCreate(void *data, int64_t ttl) {
 }
 
 void syncRespMgrDestroy(SSyncRespMgr *pObj) {
+  if (pObj != NULL) {
     taosThreadMutexLock(&(pObj->mutex));
     taosHashCleanup(pObj->pRespHash);
     taosThreadMutexUnlock(&(pObj->mutex));
     taosThreadMutexDestroy(&(pObj->mutex));
     taosMemoryFree(pObj);
   }
+}
 
 int64_t syncRespMgrAdd(SSyncRespMgr *pObj, SRespStub *pStub) {
   taosThreadMutexLock(&(pObj->mutex));

source/libs/sync/test/CMakeLists.txt

@@ -17,6 +17,7 @@ add_executable(syncVotesRespondTest "")
 add_executable(syncIndexMgrTest "")
 add_executable(syncLogStoreTest "")
 add_executable(syncEntryTest "")
+add_executable(syncEntryCacheTest "")
 add_executable(syncRequestVoteTest "")
 add_executable(syncRequestVoteReplyTest "")
 add_executable(syncAppendEntriesTest "")
@@ -129,6 +130,10 @@ target_sources(syncEntryTest
     PRIVATE
     "syncEntryTest.cpp"
 )
+target_sources(syncEntryCacheTest
+    PRIVATE
+    "syncEntryCacheTest.cpp"
+)
 target_sources(syncRequestVoteTest
     PRIVATE
     "syncRequestVoteTest.cpp"
@@ -362,6 +367,11 @@ target_include_directories(syncEntryTest
     "${TD_SOURCE_DIR}/include/libs/sync"
     "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
 )
+target_include_directories(syncEntryCacheTest
+    PUBLIC
+    "${TD_SOURCE_DIR}/include/libs/sync"
+    "${CMAKE_CURRENT_SOURCE_DIR}/../inc"
+)
 target_include_directories(syncRequestVoteTest
     PUBLIC
     "${TD_SOURCE_DIR}/include/libs/sync"
@@ -610,6 +620,10 @@ target_link_libraries(syncEntryTest
     sync
     gtest_main
 )
+target_link_libraries(syncEntryCacheTest
+    sync
+    gtest_main
+)
 target_link_libraries(syncRequestVoteTest
     sync
     gtest_main

source/libs/sync/test/syncEntryCacheTest.cpp

@@ -0,0 +1,162 @@
+#include <stdio.h>
+#include "syncEnv.h"
+#include "syncIO.h"
+#include "syncInt.h"
+#include "syncRaftLog.h"
+#include "syncRaftStore.h"
+#include "syncUtil.h"
+
+void logTest() {
+  sTrace("--- sync log test: trace");
+  sDebug("--- sync log test: debug");
+  sInfo("--- sync log test: info");
+  sWarn("--- sync log test: warn");
+  sError("--- sync log test: error");
+  sFatal("--- sync log test: fatal");
+}
+
+SSyncRaftEntry* createEntry(int i) {
+  int32_t         dataLen = 20;
+  SSyncRaftEntry* pEntry = syncEntryBuild(dataLen);
+  assert(pEntry != NULL);
+  pEntry->msgType = 88;
+  pEntry->originalRpcType = 99;
+  pEntry->seqNum = 3;
+  pEntry->isWeak = true;
+  pEntry->term = 100 + i;
+  pEntry->index = i;
+  snprintf(pEntry->data, dataLen, "value%d", i);
+
+  return pEntry;
+}
+
+SSyncNode* createFakeNode() {
+  SSyncNode* pSyncNode = (SSyncNode*)taosMemoryMalloc(sizeof(SSyncNode));
+  ASSERT(pSyncNode != NULL);
+  memset(pSyncNode, 0, sizeof(SSyncNode));
+
+  return pSyncNode;
+}
+
+SRaftEntryCache* createCache(int maxCount) {
+  SSyncNode* pSyncNode = createFakeNode();
+  ASSERT(pSyncNode != NULL);
+
+  SRaftEntryCache* pCache = raftCacheCreate(pSyncNode, maxCount);
+  ASSERT(pCache != NULL);
+
+  return pCache;
+}
+
+void test1() {
+  int32_t          code = 0;
+  SRaftEntryCache* pCache = createCache(5);
+  for (int i = 0; i < 5; ++i) {
+    SSyncRaftEntry* pEntry = createEntry(i);
+    code = raftCachePutEntry(pCache, pEntry);
+    ASSERT(code == 1);
+    syncEntryDestory(pEntry);
+  }
+  raftCacheLog2((char*)"==test1 write 5 entries==", pCache);
+
+  SyncIndex index;
+  index = 1;
+  code = raftCacheDelEntry(pCache, index);
+  ASSERT(code == 0);
+  index = 3;
+  code = raftCacheDelEntry(pCache, index);
+  ASSERT(code == 0);
+  raftCacheLog2((char*)"==test1 delete 1,3==", pCache);
+
+  code = raftCacheClear(pCache);
+  ASSERT(code == 0);
+  raftCacheLog2((char*)"==clear all==", pCache);
+}
+
+void test2() {
+  int32_t          code = 0;
+  SRaftEntryCache* pCache = createCache(5);
+  for (int i = 0; i < 5; ++i) {
+    SSyncRaftEntry* pEntry = createEntry(i);
+    code = raftCachePutEntry(pCache, pEntry);
+    ASSERT(code == 1);
+    syncEntryDestory(pEntry);
+  }
+  raftCacheLog2((char*)"==test2 write 5 entries==", pCache);
+
+  SyncIndex index;
+  index = 1;
+  SSyncRaftEntry* pEntry;
+  code = raftCacheGetEntry(pCache, index, &pEntry);
+  ASSERT(code == 0);
+  syncEntryDestory(pEntry);
+  syncEntryLog2((char*)"==test2 get entry 1==", pEntry);
+
+  index = 2;
+  code = raftCacheGetEntryP(pCache, index, &pEntry);
+  ASSERT(code == 0);
+  syncEntryLog2((char*)"==test2 get entry pointer 2==", pEntry);
+
+  // not found
+  index = 8;
+  code = raftCacheGetEntry(pCache, index, &pEntry);
+  ASSERT(code == -1 && terrno == TSDB_CODE_WAL_LOG_NOT_EXIST);
+  sTrace("==test2 get entry 8 not found==");
+
+  // not found
+  index = 9;
+  code = raftCacheGetEntryP(pCache, index, &pEntry);
+  ASSERT(code == -1 && terrno == TSDB_CODE_WAL_LOG_NOT_EXIST);
+  sTrace("==test2 get entry pointer 9 not found==");
+}
+
+void test3() {
+  int32_t          code = 0;
+  SRaftEntryCache* pCache = createCache(5);
+  for (int i = 0; i < 5; ++i) {
+    SSyncRaftEntry* pEntry = createEntry(i);
+    code = raftCachePutEntry(pCache, pEntry);
+    ASSERT(code == 1);
+    syncEntryDestory(pEntry);
+  }
+  for (int i = 6; i < 10; ++i) {
+    SSyncRaftEntry* pEntry = createEntry(i);
+    code = raftCachePutEntry(pCache, pEntry);
+    ASSERT(code == 0);
+    syncEntryDestory(pEntry);
+  }
+  raftCacheLog2((char*)"==test3 write 10 entries, max count is 5==", pCache);
+}
+
+void test4() {
+  int32_t          code = 0;
+  SRaftEntryCache* pCache = createCache(5);
+  for (int i = 0; i < 5; ++i) {
+    SSyncRaftEntry* pEntry = createEntry(i);
+    code = raftCachePutEntry(pCache, pEntry);
+    ASSERT(code == 1);
+    syncEntryDestory(pEntry);
+  }
+  raftCacheLog2((char*)"==test4 write 5 entries==", pCache);
+
+  SyncIndex index;
+  index = 3;
+  SSyncRaftEntry* pEntry;
+  code = raftCacheGetAndDel(pCache, index, &pEntry);
+  ASSERT(code == 0);
+  syncEntryLog2((char*)"==test4 get-and-del entry 3==", pEntry);
+  raftCacheLog2((char*)"==test4 after get-and-del entry 3==", pCache);
+}
+
+int main(int argc, char** argv) {
+  gRaftDetailLog = true;
+  tsAsyncLog = 0;
+  sDebugFlag = DEBUG_TRACE + DEBUG_SCREEN + DEBUG_FILE + DEBUG_DEBUG;
+
+  test1();
+  test2();
+  test3();
+  test4();
+
+  return 0;
+}

tests/pytest/util/constant.py

@@ -1,5 +1,76 @@
 # -*- coding: utf-8 -*-
 
+# basic type
+TAOS_DATA_TYPE  = [
+    "INT", "BIGINT", "SMALLINT", "TINYINT", "INT UNSIGNED", "BIGINT UNSIGNED", "SMALLINT UNSIGNED", "TINYINT UNSIGNED",
+    "FLOAT", "DOUBLE",
+    "BOOL",
+    "BINARY", "NCHAR", "VARCHAR",
+    "TIMESTAMP",
+    # "MEDIUMBLOB", "BLOB", # add in 3.x
+    # "DECIMAL", "NUMERIC", # add in 3.x
+    "JSON", # only for tag
+]
+
+TAOS_NUM_TYPE   = [
+    "INT", "BIGINT", "SMALLINT", "TINYINT", "INT UNSIGNED", "BIGINT UNSIGNED", "SMALLINT UNSIGNED", "TINYINT UNSIGNED", "FLOAT", "DOUBLE",
+    # "DECIMAL", "NUMERIC", # add in 3.x
+]
+TAOS_CHAR_TYPE  = [
+    "BINARY", "NCHAR", "VARCHAR",
+]
+TAOS_BOOL_TYPE  = ["BOOL",]
+TAOS_TS_TYPE    = ["TIMESTAMP",]
+TAOS_BIN_TYPE   = [
+    "MEDIUMBLOB", "BLOB", # add in 3.x
+]
+
+TAOS_TIME_INIT = ["b", "u", "a", "s", "m", "h", "d", "w", "n", "y"]
+TAOS_PRECISION = ["ms", "us", "ns"]
+PRECISION_DEFAULT = "ms"
+PRECISION = PRECISION_DEFAULT
+
+TAOS_KEYWORDS = [
+    "ABORT",		"CREATE",		"IGNORE",		"NULL",			"STAR",
+    "ACCOUNT",		"CTIME",		"IMMEDIATE",	"OF",			"STATE",
+    "ACCOUNTS",		"DATABASE",		"IMPORT",		"OFFSET",		"STATEMENT",
+    "ADD",			"DATABASES",	"IN",			"OR",			"STATE_WINDOW",
+    "AFTER",		"DAYS",			"INITIALLY",	"ORDER",		"STORAGE",
+    "ALL",			"DBS",			"INSERT",		"PARTITIONS",	"STREAM",
+    "ALTER",		"DEFERRED",		"INSTEAD",		"PASS",			"STREAMS",
+    "AND",			"DELIMITERS",	"INT",			"PLUS",			"STRING",
+    "AS",			"DESC",			"INTEGER",		"PPS",			"SYNCDB",
+    "ASC",			"DESCRIBE",		"INTERVAL",		"PRECISION",	"TABLE",
+    "ATTACH",		"DETACH",		"INTO",			"PREV",			"TABLES",
+    "BEFORE",		"DISTINCT", 	"IS",			"PRIVILEGE",	"TAG",
+    "BEGIN",		"DIVIDE",		"ISNULL",		"QTIME",		"TAGS",
+    "BETWEEN",		"DNODE",		"JOIN",			"QUERIES",		"TBNAME",
+    "BIGINT",		"DNODES",		"KEEP",			"QUERY",		"TIMES",
+    "BINARY",		"DOT",			"KEY",			"QUORUM",		"TIMESTAMP",
+    "BITAND",		"DOUBLE",		"KILL",			"RAISE",		"TINYINT",
+    "BITNOT",		"DROP",			"LE",			"REM",			"TOPIC",
+    "BITOR",		"EACH",			"LIKE",			"REPLACE",		"TOPICS",
+    "BLOCKS",		"END",			"LIMIT",		"REPLICA",		"TRIGGER",
+    "BOOL",			"EQ",			"LINEAR",		"RESET",		"TSERIES",
+    "BY",			"EXISTS",		"LOCAL",		"RESTRICT",		"UMINUS",
+    "CACHE",		"EXPLAIN",		"LP",			"ROW",			"UNION",
+    "CACHELAST",	"FAIL",			"LSHIFT",		"RP",			"UNSIGNED",
+    "CASCADE",		"FILE",			"LT",			"RSHIFT",		"UPDATE",
+    "CHANGE",		"FILL",			"MATCH",		"SCORES",		"UPLUS",
+    "CLUSTER",		"FLOAT",		"MAXROWS",		"SELECT",		"USE",
+    "COLON",		"FOR",			"MINROWS",		"SEMI",			"USER",
+    "COLUMN",		"FROM",			"MINUS",		"SESSION",		"USERS",
+    "COMMA",		"FSYNC",		"MNODES",		"SET",			"USING",
+    "COMP",			"GE",			"MODIFY",		"SHOW",			"VALUES",
+    "COMPACT",		"GLOB",			"MODULES",		"SLASH",		"VARIABLE",
+    "CONCAT",		"GRANTS",		"NCHAR",		"SLIDING",		"VARIABLES",
+    "CONFLICT",		"GROUP",		"NE",			"SLIMIT",		"VGROUPS",
+    "CONNECTION",	"GT",			"NONE",			"SMALLINT",		"VIEW",
+    "CONNECTIONS",	"HAVING",		"NOT",			"SOFFSET",		"VNODES",
+    "CONNS",		"ID",			"NOTNULL",		"STABLE",		"WAL",
+    "COPY",			"IF",			"NOW",			"STABLES",		"WHERE",
+]
+
 # basic data type boundary
 TINYINT_MAX             = 127
 TINYINT_MIN             = -128
@@ -11,7 +82,7 @@ SMALLINT_MAX            = 32767
 SMALLINT_MIN            = -32768
 
 SMALLINT_UN_MAX         = 65535
-MALLINT_UN_MIN          = 0
+SMALLINT_UN_MIN         = 0
 
 INT_MAX                 = 2147483647
 INT_MIN                 = -2147483648
@@ -33,8 +104,8 @@ DOUBLE_MIN              = -1.7E+308
 
 # schema boundary
 BINARY_LENGTH_MAX       = 16374
-NCAHR_LENGTH_MAX_       = 4093
-DBNAME_LENGTH_MAX_      = 64
+NCAHR_LENGTH_MAX       = 4093
+DBNAME_LENGTH_MAX      = 64
 
 STBNAME_LENGTH_MAX      = 192
 STBNAME_LENGTH_MIN      = 1
@@ -67,3 +138,31 @@ MNODE_SHM_SIZE_DEFAULT  = 6292480
 VNODE_SHM_SIZE_MAX      = 2147483647
 VNODE_SHM_SIZE_MIN      = 6292480
 VNODE_SHM_SIZE_DEFAULT  = 31458304
+
+# time_init
+TIME_MS = 1
+TIME_US = TIME_MS/1000
+TIME_NS = TIME_US/1000
+
+TIME_S = 1000 * TIME_MS
+TIME_M = 60 * TIME_S
+TIME_H = 60 * TIME_M
+TIME_D = 24 * TIME_H
+TIME_W = 7 * TIME_D
+TIME_N = 30 * TIME_D
+TIME_Y = 365 * TIME_D
+
+
+# session parameters
+INTERVAL_MIN = 1 * TIME_MS if PRECISION == PRECISION_DEFAULT else 1 * TIME_US
+
+
+# streams and related agg-function
+SMA_INDEX_FUNCTIONS = ["MIN", "MAX"]
+ROLLUP_FUNCTIONS = ["AVG", "SUM", "MIN", "MAX", "LAST", "FIRST"]
+SMA_WATMARK_MAXDELAY_INIT = ['a', "s", "m"]
+WATERMARK_MAX = 900000
+WATERMARK_MIN = 0
+
+MAX_DELAY_MAX = 900000
+MAX_DELAY_MIN = 1

tests/pytest/util/sql.py

@@ -21,6 +21,7 @@ import psutil
 import shutil
 import pandas as pd
 from util.log import *
+from util.constant import *
 
 def _parse_datetime(timestr):
     try:
@@ -117,8 +118,7 @@ class TDSql:
             col_name_list = []
             col_type_list = []
             self.cursor.execute(sql)
-            self.queryCols = self.cursor.description
-            for query_col in self.queryCols:
+            for query_col in self.cursor.description:
                 col_name_list.append(query_col[0])
                 col_type_list.append(query_col[1])
         except Exception as e:
@@ -301,6 +301,41 @@ class TDSql:
             args = (caller.filename, caller.lineno, self.sql, elm, expect_elm)
             tdLog.exit("%s(%d) failed: sql:%s, elm:%s == expect_elm:%s" % args)
 
+    def get_times(self, time_str, precision="ms"):
+        caller = inspect.getframeinfo(inspect.stack()[1][0])
+        if time_str[-1] not in TAOS_TIME_INIT:
+            tdLog.exit(f"{caller.filename}({caller.lineno}) failed: {time_str} not a standard taos time init")
+        if precision not in TAOS_PRECISION:
+            tdLog.exit(f"{caller.filename}({caller.lineno}) failed: {precision} not a standard taos time precision")
+
+        if time_str[-1] == TAOS_TIME_INIT[0]:
+            times =  int(time_str[:-1]) * TIME_NS
+        if time_str[-1] == TAOS_TIME_INIT[1]:
+            times =  int(time_str[:-1]) * TIME_US
+        if time_str[-1] == TAOS_TIME_INIT[2]:
+            times =  int(time_str[:-1]) * TIME_MS
+        if time_str[-1] == TAOS_TIME_INIT[3]:
+            times =  int(time_str[:-1]) * TIME_S
+        if time_str[-1] == TAOS_TIME_INIT[4]:
+            times =  int(time_str[:-1]) * TIME_M
+        if time_str[-1] == TAOS_TIME_INIT[5]:
+            times =  int(time_str[:-1]) * TIME_H
+        if time_str[-1] == TAOS_TIME_INIT[6]:
+            times =  int(time_str[:-1]) * TIME_D
+        if time_str[-1] == TAOS_TIME_INIT[7]:
+            times =  int(time_str[:-1]) * TIME_W
+        if time_str[-1] == TAOS_TIME_INIT[8]:
+            times =  int(time_str[:-1]) * TIME_N
+        if time_str[-1] == TAOS_TIME_INIT[9]:
+            times =  int(time_str[:-1]) * TIME_Y
+
+        if precision == "ms":
+            return int(times)
+        elif precision == "us":
+            return int(times*1000)
+        elif precision == "ns":
+            return int(times*1000*1000)
+
     def taosdStatus(self, state):
         tdLog.sleep(5)
         pstate = 0

tests/script/jenkins/basic.txt

@@ -29,7 +29,7 @@
 #./test.sh -f tsim/dnode/drop_dnode_has_vnode_replica3.sim
 #./test.sh -f tsim/dnode/drop_dnode_has_multi_vnode_replica1.sim
 #./test.sh -f tsim/dnode/drop_dnode_has_multi_vnode_replica3.sim
-#./test.sh -f tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
+./test.sh -f tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
 ./test.sh -f tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim
 
 # ---- insert

tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_follower.sim

@@ -65,7 +65,7 @@ sql_error redistribute vgroup 3 dnode 6 dnode 3 dnode 4
 # vgroup not exist
 sql_error redistribute vgroup 3 dnode 5 dnode 3 dnode 4 
 # un changed
-sql_error redistribute vgroup 2 dnode 2 dnode 3 dnode 4
+# sql_error redistribute vgroup 2 dnode 2 dnode 3 dnode 4
 # no enought vnodes
 sql_error redistribute vgroup 2 dnode 1 dnode 3 dnode 4 
 # offline vnodes
@@ -176,8 +176,6 @@ if $rows != 1 then
   return -1
 endi
 
-return
-
 print =============== step32: 
 print redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 
 sql redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5 

tests/system-test/1-insert/create_retentions.py

@@ -21,9 +21,9 @@ SINT_UN_COL = "c_sint_un"
 BINT_UN_COL = "c_bint_un"
 INT_UN_COL  = "c_int_un"
 
-BINARY_COL  = "c8"
-NCHAR_COL   = "c9"
-TS_COL      = "c10"
+BINARY_COL  = "c_binary"
+NCHAR_COL   = "c_nchar"
+TS_COL      = "c_ts"
 
 NUM_COL     = [ INT_COL, BINT_COL, SINT_COL, TINT_COL, FLOAT_COL, DOUBLE_COL, ]
 CHAR_COL    = [ BINARY_COL, NCHAR_COL, ]
@@ -51,12 +51,28 @@ class DataSet:
     binary_data     : List[str]     = None
     nchar_data      : List[str]     = None
 
+    def __post_init__(self):
+        self.ts_data        = []
+        self.int_data       = []
+        self.bint_data      = []
+        self.sint_data      = []
+        self.tint_data      = []
+        self.int_un_data    = []
+        self.bint_un_data   = []
+        self.sint_un_data   = []
+        self.tint_un_data   = []
+        self.float_data     = []
+        self.double_data    = []
+        self.bool_data      = []
+        self.binary_data    = []
+        self.nchar_data     = []
+
 
 class TDTestCase:
 
     def init(self, conn, logSql):
         tdLog.debug(f"start to excute {__file__}")
-        tdSql.init(conn.cursor(), True)
+        tdSql.init(conn.cursor(), False)
 
     @property
     def create_databases_sql_err(self):
@@ -87,28 +103,28 @@ class TDTestCase:
     @property
     def create_stable_sql_err(self):
         return [
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(ceil) watermark 1s maxdelay 1m",
+            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(ceil) watermark 1s max_delay 1m",
             f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(count) watermark  1min",
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) maxdelay -1s",
+            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) max_delay -1s",
             f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) watermark -1m",
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) watermark 1m ",
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) maxdelay 1m ",
+            # f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) watermark 1m ",
+            # f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) max_delay 1m ",
             f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int, {BINARY_COL} binary(16)) tags (tag1 int) rollup(avg) watermark 1s",
-            f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int, {BINARY_COL} nchar(16)) tags (tag1 int) rollup(avg) maxdelay 1m",
-            # f"create table ntb_1 ({PRIMARY_COL} timestamp, {INT_COL} int, {BINARY_COL} nchar(16)) rollup(avg) watermark 1s maxdelay 1s",
+            f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int, {BINARY_COL} nchar(16)) tags (tag1 int) rollup(avg) max_delay 1m",
+            # f"create table ntb_1 ({PRIMARY_COL} timestamp, {INT_COL} int, {BINARY_COL} nchar(16)) rollup(avg) watermark 1s max_delay 1s",
             # f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int, {BINARY_COL} nchar(16)) tags (tag1 int) " ,
             # f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) " ,
             # f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int) " ,
             # f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int, {BINARY_COL} nchar(16)) " ,
 
-            # watermark, maxdelay: [0, 900000], [ms, s, m, ?]
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) maxdelay 1u",
+            # watermark, max_delay: [0, 900000], [ms, s, m, ?]
+            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) max_delay 1u",
             f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) watermark 1b",
             f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) watermark 900001ms",
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) maxdelay 16m",
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) maxdelay 901s",
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) maxdelay 1h",
-            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) maxdelay 0.2h",
+            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) max_delay 16m",
+            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) max_delay 901s",
+            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) max_delay 1h",
+            f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) max_delay 0.2h",
             f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) watermark 0.002d",
 
         ]
@@ -117,11 +133,11 @@ class TDTestCase:
     def create_stable_sql_current(self):
         return [
             f"create stable stb1 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(avg)",
-            f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) watermark 5s maxdelay 1m",
-            f"create stable stb3 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(max) watermark 5s maxdelay 1m",
-            f"create stable stb4 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(sum) watermark 5s maxdelay 1m",
-            # f"create stable stb5 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(last) watermark 5s maxdelay 1m",
-            # f"create stable stb6 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(first) watermark 5s maxdelay 1m",
+            f"create stable stb2 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(min) watermark 5s max_delay 1m",
+            f"create stable stb3 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(max) watermark 5s max_delay 1m",
+            f"create stable stb4 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(sum) watermark 5s max_delay 1m",
+            # f"create stable stb5 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(last) watermark 5s max_delay 1m",
+            # f"create stable stb6 ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) rollup(first) watermark 5s max_delay 1m",
         ]
 
     def test_create_stb(self):
@@ -135,7 +151,7 @@ class TDTestCase:
         tdSql.checkRows(len(self.create_stable_sql_current))
 
         # tdSql.execute("use db")  # because db is a noraml database, not a rollup database, should not be able to create a rollup database
-        # tdSql.error(f"create stable nor_db_rollup_stb ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) file_factor 5.0")
+        # tdSql.error(f"create stable nor_db_rollup_stb ({PRIMARY_COL} timestamp, {INT_COL} int) tags (tag1 int) watermark 5s max_delay 1m")
 
 
     def test_create_databases(self):
@@ -177,21 +193,6 @@ class TDTestCase:
 
     def __data_set(self, rows):
         data_set = DataSet()
-        # neg_data_set = DataSet()
-        data_set.ts_data = []
-        data_set.int_data = []
-        data_set.bint_data = []
-        data_set.sint_data = []
-        data_set.tint_data = []
-        data_set.int_un_data = []
-        data_set.bint_un_data = []
-        data_set.sint_un_data = []
-        data_set.tint_un_data = []
-        data_set.float_data = []
-        data_set.double_data = []
-        data_set.bool_data = []
-        data_set.binary_data = []
-        data_set.nchar_data = []
 
         for i in range(rows):
             data_set.ts_data.append(NOW + 1 * (rows - i))
@@ -226,6 +227,7 @@ class TDTestCase:
         return data_set
 
     def __insert_data(self):
+        tdLog.printNoPrefix("==========step: start inser data into tables now.....")
         data = self.__data_set(rows=self.rows)
 
         # now_time = int(datetime.datetime.timestamp(datetime.datetime.now()) * 1000)
@@ -264,10 +266,10 @@ class TDTestCase:
 
     def run(self):
         self.rows = 10
-
+        tdSql.prepare()
 
         tdLog.printNoPrefix("==========step0:all check")
-        # self.all_test()
+        self.all_test()
 
         tdLog.printNoPrefix("==========step1:create table in normal database")
         tdSql.prepare()

tests/system-test/1-insert/time_range_wise.py

@@ -17,25 +17,34 @@ TINT_COL = "c_tint"
 FLOAT_COL = "c_float"
 DOUBLE_COL = "c_double"
 BOOL_COL = "c_bool"
-TINT_UN_COL = "c_tint_un"
-SINT_UN_COL = "c_sint_un"
-BINT_UN_COL = "c_bint_un"
-INT_UN_COL = "c_int_un"
+TINT_UN_COL = "c_utint"
+SINT_UN_COL = "c_usint"
+BINT_UN_COL = "c_ubint"
+INT_UN_COL = "c_uint"
 BINARY_COL = "c_binary"
 NCHAR_COL = "c_nchar"
 TS_COL = "c_ts"
 
-
-
 NUM_COL = [INT_COL, BINT_COL, SINT_COL, TINT_COL, FLOAT_COL, DOUBLE_COL, ]
 CHAR_COL = [BINARY_COL, NCHAR_COL, ]
 BOOLEAN_COL = [BOOL_COL, ]
 TS_TYPE_COL = [TS_COL, ]
 
+INT_TAG = "t_int"
+
+ALL_COL = [PRIMARY_COL, INT_COL, BINT_COL, SINT_COL, TINT_COL, FLOAT_COL, DOUBLE_COL, BINARY_COL, NCHAR_COL, BOOL_COL, TS_COL]
+TAG_COL = [INT_TAG]
+
 # insert data args：
 TIME_STEP = 10000
 NOW = int(datetime.datetime.timestamp(datetime.datetime.now()) * 1000)
 
+# init db/table
+DBNAME  = "db"
+STBNAME = "stb1"
+CTBNAME = "ct1"
+NTBNAME = "nt1"
+
 
 @dataclass
 class DataSet:
@@ -78,14 +87,20 @@ class SMAschema:
     index_flag          : str           = "SMA INDEX"
     operator            : str           = "ON"
     tbname              : str           = None
-    watermark   : str           = None
-    maxdelay    : str           = None
+    watermark           : str           = "5s"
+    max_delay           : str           = "6m"
     func                : Tuple[str]    = None
-    interval    : Tuple[str]    = None
-    sliding     : str           = None
+    interval            : Tuple[str]    = ("6m", "10s")
+    sliding             : str           = "6m"
     other               : Any           = None
     drop                : str           = "DROP"
     drop_flag           : str           = "INDEX"
+    querySmaOptimize    : int           = 1
+    show                : str           = "SHOW"
+    show_msg            : str           = "INDEXES"
+    show_oper           : str           = "FROM"
+    dbname              : str           = None
+    rollup_db           : bool          = False
 
     def __post_init__(self):
         if isinstance(self.other, dict):
@@ -111,8 +126,8 @@ class SMAschema:
                     self.watermark = v
                     del self.other[k]
 
-                if k.lower() == "maxdelay" and isinstance(v, str) and not self.maxdelay:
-                    self.maxdelay = v
+                if k.lower() == "max_delay" and isinstance(v, str) and not self.max_delay:
+                    self.max_delay = v
                     del self.other[k]
 
                 if k.lower() == "functions" and isinstance(v, tuple) and not self.func:
@@ -131,12 +146,36 @@ class SMAschema:
                     self.drop_flag = v
                     del self.other[k]
 
+                if k.lower() == "show_msg" and isinstance(v, str) and not self.show_msg:
+                    self.show_msg = v
+                    del self.other[k]
+
+                if k.lower() == "dbname" and isinstance(v, str) and not self.dbname:
+                    self.dbname = v
+                    del self.other[k]
+
+                if k.lower() == "show_oper" and isinstance(v, str) and not self.show_oper:
+                    self.show_oper = v
+                    del self.other[k]
+
+                if k.lower() == "rollup_db" and isinstance(v, bool) and not self.rollup_db:
+                    self.rollup_db = v
+                    del self.other[k]
+
+
+
+# from ...pytest.util.sql import *
+# from ...pytest.util.constant import *
 
 class TDTestCase:
+    updatecfgDict = {"querySmaOptimize": 1}
 
     def init(self, conn, logSql):
         tdLog.debug(f"start to excute {__file__}")
         tdSql.init(conn.cursor(), False)
+        self.precision = "ms"
+        self.sma_count = 0
+        self.sma_created_index = []
 
     """
         create sma index :
@@ -155,13 +194,17 @@ class TDTestCase:
         if sma.func:
             sql += f" function({', '.join(sma.func)})"
         if sma.interval:
-            sql += f" interval({', '.join(sma.interval)})"
+            interval, offset = self.__get_interval_offset(sma.interval)
+            if offset:
+                sql += f" interval({interval}, {offset})"
+            else:
+                sql += f" interval({interval})"
         if sma.sliding:
             sql += f" sliding({sma.sliding})"
         if sma.watermark:
             sql += f" watermark {sma.watermark}"
-        if sma.maxdelay:
-            sql += f" maxdelay {sma.maxdelay}"
+        if sma.max_delay:
+            sql += f" max_delay {sma.max_delay}"
         if isinstance(sma.other, dict):
             for k,v in sma.other.items():
                 if isinstance(v,tuple) or isinstance(v, list):
@@ -171,53 +214,305 @@ class TDTestCase:
         if isinstance(sma.other, tuple) or isinstance(sma.other, list):
             sql += " ".join(sma.other)
         if isinstance(sma.other, int) or isinstance(sma.other, float) or isinstance(sma.other, str):
-            sql += sma.other
+            sql += f" {sma.other}"
 
         return sql
 
-    def sma_create_check(self, sma:SMAschema):
+    def __get_sma_func_col(self, func):
+        cols = []
+        if isinstance(func, str):
+            cols.append( func.split("(")[-1].split(")")[0] )
+        elif isinstance(func, tuple) or isinstance(func, list):
+            for func_col in func:
+                cols.append(func_col.split("(")[-1].split(")")[0])
+        else:
+            cols = []
+        return cols
+
+    def __check_sma_func(self, func:tuple):
+        if not isinstance(func, str) and not isinstance(func, tuple) and not isinstance(func, list):
+            return False
+        if isinstance(func, str) :
+            if "(" not in func or ")" not in func:
+                return False
+            if func.split("(")[0].upper() not in SMA_INDEX_FUNCTIONS:
+                return False
+            if func.split("(")[1].split(")")[0] not in ALL_COL and func.split("(")[1].split(")")[0] not in TAG_COL :
+                return False
+        if isinstance(func, tuple) or isinstance(func, list):
+            for arg in func:
+                if not isinstance(arg, str):
+                    return False
+                if "(" not in arg or ")" not in arg:
+                    return False
+                if arg.split("(")[0].upper() not in SMA_INDEX_FUNCTIONS:
+                    return False
+                if arg.split("(")[1].split(")")[0] not in ALL_COL and arg.split("(")[1].split(")")[0] not in TAG_COL :
+                    return False
+        return True
+
+    def __check_sma_watermark(self, arg):
+        if not arg:
+            return False
+        if not isinstance(arg, str):
+            return False
+        if arg[-1] not in SMA_WATMARK_MAXDELAY_INIT:
+            return False
+        if len(arg) == 1:
+            return False
+        if not arg[:-1].isdecimal():
+            return False
+        if tdSql.get_times(arg) > WATERMARK_MAX:
+            return False
+        if tdSql.get_times(arg) < WATERMARK_MIN:
+            return False
+
+        return True
+
+    def __check_sma_max_delay(self, arg):
+        if not self.__check_sma_watermark(arg):
+            return False
+        if tdSql.get_times(arg) < MAX_DELAY_MIN:
+            return False
+
+        return True
+
+    def __check_sma_sliding(self, arg):
+        if not isinstance(arg, str):
+            return False
+        if arg[-1] not in TAOS_TIME_INIT:
+            return False
+        if len(arg) == 1:
+            return False
+        if not arg[:-1].isdecimal():
+            return False
+
+        return True
+
+    def __get_interval_offset(self, args):
+        if isinstance(args, str):
+            interval, offset = args, None
+        elif isinstance(args,tuple) or isinstance(args, list):
+            if len(args) == 1:
+                interval, offset = args[0], None
+            elif len(args) == 2:
+                interval, offset = args
+            else:
+                interval, offset = False, False
+        else:
+            interval, offset = False, False
+
+        return interval, offset
+
+    def __check_sma_interval(self, args):
+        if not isinstance(args, tuple) and not isinstance(args,str):
+            return False
+        interval, offset =  self.__get_interval_offset(args)
+        if not interval:
+            return False
+        if not self.__check_sma_sliding(interval):
+            return False
+        if tdSql.get_times(interval) < INTERVAL_MIN:
+            return False
+        if offset:
+            if not self.__check_sma_sliding(offset):
+                return False
+            if tdSql.get_times(interval) <= tdSql.get_times(offset) :
+                return False
+
+        return True
+
+    def __sma_create_check(self, sma:SMAschema):
+        if  self.updatecfgDict["querySmaOptimize"] == 0:
+            return False
+        # # TODO: if database is a rollup-db, can not create sma index
+        # tdSql.query("select database()")
+        # if sma.rollup_db :
+        #     return False
         tdSql.query("show stables")
+        if not sma.tbname:
+            return False
         stb_in_list = False
         for row in tdSql.queryResult:
             if sma.tbname == row[0]:
                 stb_in_list = True
-                break
         if not stb_in_list:
-            tdSql.error(self.__create_sma_index(sma))
-        if not sma.creation:
-            tdSql.error(self.__create_sma_index(sma))
-        if not sma.index_flag:
-            tdSql.error(self.__create_sma_index(sma))
-        if not sma.index_name:
-            tdSql.error(self.__create_sma_index(sma))
-        if not sma.operator:
-            tdSql.error(self.__create_sma_index(sma))
-        if not sma.tbname:
-            tdSql.error(self.__create_sma_index(sma))
-        if not sma.func:
-            tdSql.error(self.__create_sma_index(sma))
-        if not sma.interval:
-            tdSql.error(self.__create_sma_index(sma))
-        if not sma.sliding:
-            tdSql.error(self.__create_sma_index(sma))
+            return False
+        if not sma.creation or not isinstance(sma.creation, str) or sma.creation.upper() != "CREATE":
+            return False
+        if not sma.index_flag or not isinstance(sma.index_flag, str) or  sma.index_flag.upper() != "SMA INDEX" :
+            return False
+        if not sma.index_name or not isinstance(sma.index_name, str) or sma.index_name.upper() in TAOS_KEYWORDS:
+            return False
+        if not sma.operator or not isinstance(sma.operator, str) or sma.operator.upper() != "ON":
+            return False
+
+        if not sma.func or not self.__check_sma_func(sma.func):
+            return False
+        tdSql.query(f"desc {sma.tbname}")
+        _col_list = []
+        for col_row in  tdSql.queryResult:
+            _col_list.append(col_row[0])
+        _sma_func_cols = self.__get_sma_func_col(sma.func)
+        for  _sma_func_col in _sma_func_cols:
+            if _sma_func_col not in _col_list:
+                return False
+
+        if sma.sliding and not self.__check_sma_sliding(sma.sliding):
+            return False
+        interval, _ =  self.__get_interval_offset(sma.interval)
+        if not sma.interval or not self.__check_sma_interval(sma.interval) :
+            return False
+        if sma.sliding and tdSql.get_times(interval) < tdSql.get_times(sma.sliding):
+            return False
+        if sma.watermark and not self.__check_sma_watermark(sma.watermark):
+            return False
+        if sma.max_delay and not self.__check_sma_max_delay(sma.max_delay):
+            return False
         if sma.other:
+            return False
+
+        return True
+
+    def sma_create_check(self, sma:SMAschema):
+        if self.__sma_create_check(sma):
+            tdSql.query(self.__create_sma_index(sma))
+            self.sma_count += 1
+            self.sma_created_index.append(sma.index_name)
+            tdSql.query("show streams")
+            tdSql.checkRows(self.sma_count)
+
+        else:
             tdSql.error(self.__create_sma_index(sma))
 
+    def __drop_sma_index(self, sma:SMAschema):
+        sql = f"{sma.drop} {sma.drop_flag} {sma.index_name}"
+        return sql
+
+    def __sma_drop_check(self, sma:SMAschema):
+        if not sma.drop:
+            return False
+        if not sma.drop_flag:
+            return False
+        if not sma.index_name:
+            return False
+
+        return True
+
+    def sma_drop_check(self, sma:SMAschema):
+        if self.__sma_drop_check(sma):
+            tdSql.query(self.__drop_sma_index(sma))
+            print(self.__drop_sma_index(sma))
+            self.sma_count -= 1
+            self.sma_created_index = list(filter(lambda x: x != sma.index_name, self.sma_created_index))
+            tdSql.query("show streams")
+            tdSql.checkRows(self.sma_count)
+
+        else:
+            tdSql.error(self.__drop_sma_index(sma))
+
+    def __show_sma_index(self, sma:SMAschema):
+        sql = f"{sma.show} {sma.show_msg} {sma.show_oper} {sma.tbname}"
+        return sql
+
+    def __sma_show_check(self, sma:SMAschema):
+        if not sma.show:
+            return False
+        if not sma.show_msg:
+            return False
+        if not sma.show_oper:
+            return False
+        if not sma.tbname:
+            return False
+
+        return True
+
+    def sma_show_check(self, sma:SMAschema):
+        if self.__sma_show_check(sma):
+            tdSql.query(self.__show_sma_index(sma))
+            tdSql.checkRows(self.sma_count)
+        else:
+            tdSql.error(self.__show_sma_index(sma))
+
+    @property
+    def __create_sma_sql(self):
+        err_sqls = []
+        cur_sqls = []
+        # err_set
+        # # case 1: required fields check
+        err_sqls.append( SMAschema(creation="", tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(index_name="",tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(index_flag="",tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(operator="",tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(tbname="", func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(func=("",),tbname=STBNAME ) )
+        err_sqls.append( SMAschema(interval=(""),tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+
+        # # case 2: err fields
+        err_sqls.append( SMAschema(creation="show",tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(creation="alter",tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(creation="select",tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+
+        err_sqls.append( SMAschema(index_flag="SMA INDEXES", tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(index_flag="SMA INDEX ,", tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        err_sqls.append( SMAschema(index_name="tbname", tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+
+
+        # current_set
+
+        cur_sqls.append( SMAschema(max_delay="",tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        cur_sqls.append( SMAschema(watermark="",index_name="sma_index_2",tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+        cur_sqls.append( SMAschema(sliding="",index_name='sma_index_3',tbname=STBNAME, func=(f"min({INT_COL})",f"max({INT_COL})") ) )
+
+
+        return err_sqls, cur_sqls
+
+    def test_create_sma(self):
+        err_sqls , cur_sqls = self.__create_sma_sql
+        for err_sql in err_sqls:
+            self.sma_create_check(err_sql)
+        for cur_sql in cur_sqls:
+            self.sma_create_check(cur_sql)
+
+    @property
+    def __drop_sma_sql(self):
+        err_sqls = []
+        cur_sqls = []
+        # err_set
+        ## case 1: required fields check
+        err_sqls.append( SMAschema(drop="") )
+        err_sqls.append( SMAschema(drop_flag="") )
+        err_sqls.append( SMAschema(index_name="") )
+
+        for index in self.sma_created_index:
+            cur_sqls.append(SMAschema(index_name=index))
+
+        return err_sqls, cur_sqls
+
+    def test_drop_sma(self):
+        err_sqls , cur_sqls = self.__drop_sma_sql
+        for err_sql in err_sqls:
+            self.sma_drop_check(err_sql)
+        # for cur_sql in cur_sqls:
+        #     self.sma_drop_check(cur_sql)
 
     def all_test(self):
+        self.test_create_sma()
+        self.test_drop_sma()
+
         pass
 
     def __create_tb(self):
         tdLog.printNoPrefix("==========step: create table")
-        create_stb_sql = f'''create table stb1(
+        create_stb_sql = f'''create table {STBNAME}(
                 ts timestamp, {INT_COL} int, {BINT_COL} bigint, {SINT_COL} smallint, {TINT_COL} tinyint,
                 {FLOAT_COL} float, {DOUBLE_COL} double, {BOOL_COL} bool,
                 {BINARY_COL} binary(16), {NCHAR_COL} nchar(32), {TS_COL} timestamp,
                 {TINT_UN_COL} tinyint unsigned, {SINT_UN_COL} smallint unsigned,
                 {INT_UN_COL} int unsigned, {BINT_UN_COL} bigint unsigned
-            ) tags (tag1 int)
+            ) tags ({INT_TAG} int)
             '''
-        create_ntb_sql = f'''create table t1(
+        create_ntb_sql = f'''create table {NTBNAME}(
                 ts timestamp, {INT_COL} int, {BINT_COL} bigint, {SINT_COL} smallint, {TINT_COL} tinyint,
                 {FLOAT_COL} float, {DOUBLE_COL} double, {BOOL_COL} bool,
                 {BINARY_COL} binary(16), {NCHAR_COL} nchar(32), {TS_COL} timestamp,
@@ -253,6 +548,7 @@ class TDTestCase:
         return data_set
 
     def __insert_data(self):
+        tdLog.printNoPrefix("==========step: start inser data into tables now.....")
         data = self.__data_set(rows=self.rows)
 
         # now_time = int(datetime.datetime.timestamp(datetime.datetime.now()) * 1000)
@@ -278,7 +574,7 @@ class TDTestCase:
             tdSql.execute(
                 f"insert into ct4 values ( {NOW - i * int(TIME_STEP * 0.8) }, {row_data} )")
             tdSql.execute(
-                f"insert into t1 values ( {NOW - i * int(TIME_STEP * 1.2)}, {row_data} )")
+                f"insert into {NTBNAME} values ( {NOW - i * int(TIME_STEP * 1.2)}, {row_data} )")
 
         tdSql.execute(
             f"insert into ct2 values ( {NOW + int(TIME_STEP * 0.6)}, {null_data} )")
@@ -295,28 +591,31 @@ class TDTestCase:
             f"insert into ct4 values ( {NOW - self.rows * int(TIME_STEP * 0.39)}, {null_data} )")
 
         tdSql.execute(
-            f"insert into t1 values ( {NOW + int(TIME_STEP * 1.2)}, {null_data} )")
+            f"insert into {NTBNAME} values ( {NOW + int(TIME_STEP * 1.2)}, {null_data} )")
         tdSql.execute(
-            f"insert into t1 values ( {NOW - (self.rows + 1) * int(TIME_STEP * 1.2)}, {null_data} )")
+            f"insert into {NTBNAME} values ( {NOW - (self.rows + 1) * int(TIME_STEP * 1.2)}, {null_data} )")
         tdSql.execute(
-            f"insert into t1 values ( {NOW - self.rows * int(TIME_STEP * 0.59)}, {null_data} )")
+            f"insert into {NTBNAME} values ( {NOW - self.rows * int(TIME_STEP * 0.59)}, {null_data} )")
 
     def run(self):
-        sma1 = SMAschema(func=("min(c1)","max(c2)"))
-        sql1 = self.__create_sma_index(sma1)
-        print("================")
-        print(sql1)
-        # a = DataSet()
-        # return
         self.rows = 10
 
         tdLog.printNoPrefix("==========step0:all check")
-        # self.all_test()
 
         tdLog.printNoPrefix("==========step1:create table in normal database")
         tdSql.prepare()
         self.__create_tb()
-        self.__insert_data()
+        # self.__insert_data()
+        self.all_test()
+
+        # drop databases, create same name db、stb and sma index
+        # tdSql.prepare()
+        # self.__create_tb()
+        # self.__insert_data()
+        # self.all_test()
+
+
+
         return
 
         tdLog.printNoPrefix("==========step2:create table in rollup database")

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

@@ -28,7 +28,7 @@ class TDTestCase:
 
     def init(self, conn, logSql):
         tdLog.debug(f"start to excute {__file__}")
-        tdSql.init(conn.cursor(), True)
+        tdSql.init(conn.cursor(), False)
 
     def __query_condition(self,tbname):
         query_condition = []

tests/system-test/2-query/join2.py

@@ -28,7 +28,7 @@ class TDTestCase:
 
     def init(self, conn, logSql):
         tdLog.debug(f"start to excute {__file__}")
-        tdSql.init(conn.cursor(), True)
+        tdSql.init(conn.cursor(), False)
 
     def __query_condition(self,tbname):
         query_condition = []

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

@@ -31,7 +31,7 @@ class TDTestCase:
 
     def init(self, conn, logSql):
         tdLog.debug(f"start to excute {__file__}")
-        tdSql.init(conn.cursor())
+        tdSql.init(conn.cursor(),False)
 
     def __substr_condition(self):  # sourcery skip: extract-method
         substr_condition = []

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

@@ -188,8 +188,8 @@ class TDTestCase:
 
     def check_tail_table(self , tbname , col_name , tail_rows , offset):
         tail_sql = f"select tail({col_name} , {tail_rows} , {offset}) from {tbname}"
-        equal_sql = f"select {col_name} from (select ts , {col_name} from {tbname} order by ts desc limit {tail_rows} offset {offset}) order by ts"
-        #equal_sql = f"select {col_name} from {tbname} order by ts desc limit {tail_rows} offset {offset}"
+        #equal_sql = f"select {col_name} from (select ts , {col_name} from {tbname} order by ts desc limit {tail_rows} offset {offset}) order by ts"
+        equal_sql = f"select {col_name} from {tbname} order by ts desc limit {tail_rows} offset {offset}"
         tdSql.query(tail_sql)
         tail_result = tdSql.queryResult
 
@@ -404,7 +404,7 @@ class TDTestCase:
                 f"insert into sub1_bound values ( now()+1s, 2147483648, 9223372036854775808, 32768, 128, 3.40E+38, 1.7e+308, True, 'binary_tb1', 'nchar_tb1', now() )"
             )
         
-        tdSql.query("select tail(c2,2) from sub1_bound")
+        tdSql.query("select tail(c2,2) from sub1_bound order by 1 desc")
         tdSql.checkRows(2)
         tdSql.checkData(0,0,9223372036854775803)