diff --git a/src/client/inc/tscUtil.h b/src/client/inc/tscUtil.h
index d8b9bfab26603793b0f0266877647daf41010493..de69e23b9bb7671a6865d9360bf414865fee87c5 100644
--- a/src/client/inc/tscUtil.h
+++ b/src/client/inc/tscUtil.h
@@ -154,7 +154,7 @@ bool tscMultiRoundQuery(SQueryInfo* pQueryInfo, int32_t tableIndex);
 bool tscQueryBlockInfo(SQueryInfo* pQueryInfo);
 
 SExprInfo* tscAddFuncInSelectClause(SQueryInfo* pQueryInfo, int32_t outputColIndex, int16_t functionId,
-                                       SColumnIndex* pIndex, SSchema* pColSchema, int16_t colType);
+                                       SColumnIndex* pIndex, SSchema* pColSchema, int16_t colType, int16_t colId);
 
 int32_t tscSetTableFullName(SName* pName, SStrToken* pzTableName, SSqlObj* pSql);
 void    tscClearInterpInfo(SQueryInfo* pQueryInfo);
@@ -290,7 +290,7 @@ void registerSqlObj(SSqlObj* pSql);
 SSqlObj* createSubqueryObj(SSqlObj* pSql, int16_t tableIndex, __async_cb_func_t fp, void* param, int32_t cmd, SSqlObj* pPrevSql);
 void     addGroupInfoForSubquery(SSqlObj* pParentObj, SSqlObj* pSql, int32_t subClauseIndex, int32_t tableIndex);
 
-void doAddGroupColumnForSubquery(SQueryInfo* pQueryInfo, int32_t tagIndex);
+void doAddGroupColumnForSubquery(SQueryInfo* pQueryInfo, int32_t tagIndex, SSqlCmd* pCmd);
 
 int16_t tscGetJoinTagColIdByUid(STagCond* pTagCond, uint64_t uid);
 int16_t tscGetTagColIndexById(STableMeta* pTableMeta, int16_t colId);
diff --git a/src/client/inc/tsclient.h b/src/client/inc/tsclient.h
index 13ec02b5d14e5b35e83905fe7b08fd7a6b699c0a..87da9c8b002c018f6a6e281d0aae4e7f48cb1c90 100644
--- a/src/client/inc/tsclient.h
+++ b/src/client/inc/tsclient.h
@@ -203,7 +203,7 @@ typedef struct SQueryInfo {
   SInterval        interval;      // tumble time window
   SSessionWindow   sessionWindow; // session time window
 
-  SGroupbyExpr  groupbyExpr;   // groupby tags info
+  SGroupbyExpr     groupbyExpr;   // groupby tags info
   SArray *         colList;       // SArray<SColumn*>
   SFieldInfo       fieldsInfo;
   SArray *         exprList;      // SArray<SExprInfo*>
@@ -251,6 +251,7 @@ typedef struct {
   SVgroupsInfo *pVgroupInfo;
 } STableMetaVgroupInfo;
 
+// TODO extract sql parser supporter
 typedef struct {
   int     command;
   uint8_t msgType;
@@ -293,6 +294,7 @@ typedef struct {
 
   SHashObj    *pTableBlockHashList;     // data block for each table
   SArray      *pDataBlocks;             // SArray<STableDataBlocks*>. Merged submit block for each vgroup
+  int32_t      resColumnId;
 } SSqlCmd;
 
 typedef struct SResRec {
@@ -514,7 +516,7 @@ extern int   tscNumOfObj;     // number of existed sqlObj in current process.
 extern int (*tscBuildMsg[TSDB_SQL_MAX])(SSqlObj *pSql, SSqlInfo *pInfo);
 
 void tscBuildVgroupTableInfo(SSqlObj* pSql, STableMetaInfo* pTableMetaInfo, SArray* tables);
-int16_t getNewResColId(SQueryInfo* pQueryInfo);
+int16_t getNewResColId(SSqlCmd* pCmd);
 
 #ifdef __cplusplus
 }
diff --git a/src/client/src/tscAsync.c b/src/client/src/tscAsync.c
index 28eafa164cea1b65de568f06232f87fa2fa98f33..1456a91099f054095c4eaa5863775732d5d3cbd0 100644
--- a/src/client/src/tscAsync.c
+++ b/src/client/src/tscAsync.c
@@ -59,6 +59,7 @@ void doAsyncQuery(STscObj* pObj, SSqlObj* pSql, __async_cb_func_t fp, void* para
 
   tscDebugL("0x%"PRIx64" SQL: %s", pSql->self, pSql->sqlstr);
   pCmd->curSql = pSql->sqlstr;
+  pCmd->resColumnId = TSDB_RES_COL_ID;
 
   int32_t code = tsParseSql(pSql, true);
   if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) return;
diff --git a/src/client/src/tscLocalMerge.c b/src/client/src/tscLocalMerge.c
index 2dbc09f08dc2a93c7aeacea35691a73c8b15d12e..e1bcf3a7922bba763f2e27a4983664fcb138d963 100644
--- a/src/client/src/tscLocalMerge.c
+++ b/src/client/src/tscLocalMerge.c
@@ -871,7 +871,7 @@ SSDataBlock* doGlobalAggregate(void* param, bool* newgroup) {
   }
 
   SMultiwayMergeInfo *pAggInfo = pOperator->info;
-  SOperatorInfo      *upstream = pOperator->upstream;
+  SOperatorInfo      *upstream = pOperator->upstream[0];
 
   *newgroup = false;
   bool handleData = false;
@@ -974,7 +974,7 @@ static SSDataBlock* skipGroupBlock(SOperatorInfo* pOperator, bool* newgroup) {
 
   SSDataBlock* pBlock = NULL;
   if (pInfo->currentGroupOffset == 0) {
-    pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+    pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
     if (pBlock == NULL) {
       setQueryStatus(pOperator->pRuntimeEnv, QUERY_COMPLETED);
       pOperator->status = OP_EXEC_DONE;
@@ -982,7 +982,7 @@ static SSDataBlock* skipGroupBlock(SOperatorInfo* pOperator, bool* newgroup) {
 
     if (*newgroup == false && pInfo->limit.limit > 0 && pInfo->rowsTotal >= pInfo->limit.limit) {
       while ((*newgroup) == false) {  // ignore the remain blocks
-        pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+        pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
         if (pBlock == NULL) {
           setQueryStatus(pOperator->pRuntimeEnv, QUERY_COMPLETED);
           pOperator->status = OP_EXEC_DONE;
@@ -994,7 +994,7 @@ static SSDataBlock* skipGroupBlock(SOperatorInfo* pOperator, bool* newgroup) {
     return pBlock;
   }
 
-    pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+    pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
     if (pBlock == NULL) {
       setQueryStatus(pOperator->pRuntimeEnv, QUERY_COMPLETED);
       pOperator->status = OP_EXEC_DONE;
@@ -1008,7 +1008,7 @@ static SSDataBlock* skipGroupBlock(SOperatorInfo* pOperator, bool* newgroup) {
       }
 
       while ((*newgroup) == false) {
-        pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+        pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
         if (pBlock == NULL) {
           setQueryStatus(pOperator->pRuntimeEnv, QUERY_COMPLETED);
           pOperator->status = OP_EXEC_DONE;
diff --git a/src/client/src/tscSQLParser.c b/src/client/src/tscSQLParser.c
index 676dc51cb00e9b0fc574a207d68f10fe8d557a6b..c179ed1ec61b009c59438b6facbcda2a1c4f62cc 100644
--- a/src/client/src/tscSQLParser.c
+++ b/src/client/src/tscSQLParser.c
@@ -58,7 +58,7 @@ typedef struct SConvertFunc {
   int32_t execFuncId;
 } SConvertFunc;
 
-static SExprInfo* doAddProjectCol(SQueryInfo* pQueryInfo, int32_t colIndex, int32_t tableIndex);
+static SExprInfo* doAddProjectCol(SQueryInfo* pQueryInfo, int32_t colIndex, int32_t tableIndex, int32_t colId);
 
 static int32_t setShowInfo(SSqlObj* pSql, SSqlInfo* pInfo);
 static char*   getAccountId(SSqlObj* pSql);
@@ -137,8 +137,8 @@ static bool    isTimeWindowQuery(SQueryInfo* pQueryInfo) {
   return pQueryInfo->interval.interval > 0 || pQueryInfo->sessionWindow.gap > 0;
 }
 
-int16_t getNewResColId(SQueryInfo* pQueryInfo) {
-  return pQueryInfo->resColumnId--;
+int16_t getNewResColId(SSqlCmd* pCmd) {
+  return pCmd->resColumnId--;
 }
 
 static uint8_t convertOptr(SStrToken *pToken) {
@@ -730,7 +730,7 @@ static bool isTopBottomQuery(SQueryInfo* pQueryInfo) {
 }
 
 // need to add timestamp column in result set, if it is a time window query
-static int32_t addPrimaryTsColumnForTimeWindowQuery(SQueryInfo* pQueryInfo) {
+static int32_t addPrimaryTsColumnForTimeWindowQuery(SQueryInfo* pQueryInfo, SSqlCmd* pCmd) {
   uint64_t uid = tscExprGet(pQueryInfo, 0)->base.uid;
 
   int32_t  tableIndex = COLUMN_INDEX_INITIAL_VAL;
@@ -750,7 +750,7 @@ static int32_t addPrimaryTsColumnForTimeWindowQuery(SQueryInfo* pQueryInfo) {
   tstrncpy(s.name, aAggs[TSDB_FUNC_TS].name, sizeof(s.name));
 
   SColumnIndex index = {tableIndex, PRIMARYKEY_TIMESTAMP_COL_INDEX};
-  tscAddFuncInSelectClause(pQueryInfo, 0, TSDB_FUNC_TS, &index, &s, TSDB_COL_NORMAL);
+  tscAddFuncInSelectClause(pQueryInfo, 0, TSDB_FUNC_TS, &index, &s, TSDB_COL_NORMAL, getNewResColId(pCmd));
   return TSDB_CODE_SUCCESS;
 }
 
@@ -792,7 +792,7 @@ static int32_t checkInvalidExprForTimeWindow(SSqlCmd* pCmd, SQueryInfo* pQueryIn
     return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg1);
   }
 
-  return addPrimaryTsColumnForTimeWindowQuery(pQueryInfo);
+  return addPrimaryTsColumnForTimeWindowQuery(pQueryInfo, pCmd);
 }
 
 int32_t validateIntervalNode(SSqlObj* pSql, SQueryInfo* pQueryInfo, SSqlNode* pSqlNode) {
@@ -1442,7 +1442,7 @@ static int32_t handleArithmeticExpr(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32
     SColumnIndex index = {.tableIndex = tableIndex};
 
     SExprInfo* pExpr = tscExprAppend(pQueryInfo, TSDB_FUNC_ARITHM, &index, TSDB_DATA_TYPE_DOUBLE, sizeof(double),
-                                       getNewResColId(pQueryInfo), sizeof(double), false);
+                                       getNewResColId(pCmd), sizeof(double), false);
 
     char* name = (pItem->aliasName != NULL)? pItem->aliasName:pItem->pNode->token.z;
     size_t len = MIN(sizeof(pExpr->base.aliasName), pItem->pNode->token.n + 1);
@@ -1516,7 +1516,7 @@ static int32_t handleArithmeticExpr(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32
 
     pExprInfo->base.functionId = TSDB_FUNC_ARITHM;
     pExprInfo->base.numOfParams = 1;
-    pExprInfo->base.resColId = getNewResColId(pQueryInfo);
+    pExprInfo->base.resColId = getNewResColId(pCmd);
     strncpy(pExprInfo->base.aliasName, aliasName, tListLen(pExprInfo->base.aliasName));
     strncpy(pExprInfo->base.token, rawName, tListLen(pExprInfo->base.token));
 
@@ -1549,8 +1549,8 @@ static int32_t handleArithmeticExpr(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32
   return TSDB_CODE_SUCCESS;
 }
 
-static void addProjectQueryCol(SQueryInfo* pQueryInfo, int32_t startPos, SColumnIndex* pIndex, tSqlExprItem* pItem) {
-  SExprInfo* pExpr = doAddProjectCol(pQueryInfo, pIndex->columnIndex, pIndex->tableIndex);
+static void addProjectQueryCol(SQueryInfo* pQueryInfo, int32_t startPos, SColumnIndex* pIndex, tSqlExprItem* pItem, int32_t colId) {
+  SExprInfo* pExpr = doAddProjectCol(pQueryInfo, pIndex->columnIndex, pIndex->tableIndex, colId);
 
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, pIndex->tableIndex);
   STableMeta*     pTableMeta = pTableMetaInfo->pTableMeta;
@@ -1572,7 +1572,7 @@ static void addProjectQueryCol(SQueryInfo* pQueryInfo, int32_t startPos, SColumn
   insertResultField(pQueryInfo, startPos, &ids, pExpr->base.resBytes, (int8_t)pExpr->base.resType, pExpr->base.aliasName, pExpr);
 }
 
-static void addPrimaryTsColIntoResult(SQueryInfo* pQueryInfo) {
+static void addPrimaryTsColIntoResult(SQueryInfo* pQueryInfo, SSqlCmd* pCmd) {
   // primary timestamp column has been added already
   size_t size = tscNumOfExprs(pQueryInfo);
   for (int32_t i = 0; i < size; ++i) {
@@ -1590,7 +1590,7 @@ static void addPrimaryTsColIntoResult(SQueryInfo* pQueryInfo) {
   // add the timestamp column into the output columns
   SColumnIndex index = {0};  // primary timestamp column info
   int32_t numOfCols = (int32_t)tscNumOfExprs(pQueryInfo);
-  tscAddFuncInSelectClause(pQueryInfo, numOfCols, TSDB_FUNC_PRJ, &index, pSchema, TSDB_COL_NORMAL);
+  tscAddFuncInSelectClause(pQueryInfo, numOfCols, TSDB_FUNC_PRJ, &index, pSchema, TSDB_COL_NORMAL, getNewResColId(pCmd));
 
   SInternalField* pSupInfo = tscFieldInfoGetInternalField(&pQueryInfo->fieldsInfo, numOfCols);
   pSupInfo->visible = false;
@@ -1681,7 +1681,7 @@ int32_t validateSelectNodeList(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SArray* pS
   // there is only one user-defined column in the final result field, add the timestamp column.
   size_t numOfSrcCols = taosArrayGetSize(pQueryInfo->colList);
   if (numOfSrcCols <= 0 && !tscQueryTags(pQueryInfo) && !tscQueryBlockInfo(pQueryInfo)) {
-    addPrimaryTsColIntoResult(pQueryInfo);
+    addPrimaryTsColIntoResult(pQueryInfo, pCmd);
   }
 
   if (!functionCompatibleCheck(pQueryInfo, joinQuery, timeWindowQuery)) {
@@ -1714,7 +1714,7 @@ int32_t insertResultField(SQueryInfo* pQueryInfo, int32_t outputIndex, SColumnLi
   return TSDB_CODE_SUCCESS;
 }
 
-SExprInfo* doAddProjectCol(SQueryInfo* pQueryInfo, int32_t colIndex, int32_t tableIndex) {
+SExprInfo* doAddProjectCol(SQueryInfo* pQueryInfo, int32_t colIndex, int32_t tableIndex, int32_t colId) {
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, tableIndex);
   STableMeta*     pTableMeta = pTableMetaInfo->pTableMeta;
   int32_t numOfCols = tscGetNumOfColumns(pTableMeta);
@@ -1731,15 +1731,12 @@ SExprInfo* doAddProjectCol(SQueryInfo* pQueryInfo, int32_t colIndex, int32_t tab
     index.columnIndex = colIndex;
   }
 
-  int16_t colId = getNewResColId(pQueryInfo);
   return tscExprAppend(pQueryInfo, functionId, &index, pSchema->type, pSchema->bytes, colId, pSchema->bytes,
                           (functionId == TSDB_FUNC_TAGPRJ));
 }
 
 SExprInfo* tscAddFuncInSelectClause(SQueryInfo* pQueryInfo, int32_t outputColIndex, int16_t functionId,
-                                  SColumnIndex* pIndex, SSchema* pColSchema, int16_t flag) {
-  int16_t colId = getNewResColId(pQueryInfo);
-
+                                  SColumnIndex* pIndex, SSchema* pColSchema, int16_t flag, int16_t colId) {
   SExprInfo* pExpr = tscExprInsert(pQueryInfo, outputColIndex, functionId, pIndex, pColSchema->type,
                                      pColSchema->bytes, colId, pColSchema->bytes, TSDB_COL_IS_TAG(flag));
   tstrncpy(pExpr->base.aliasName, pColSchema->name, sizeof(pExpr->base.aliasName));
@@ -1762,7 +1759,7 @@ SExprInfo* tscAddFuncInSelectClause(SQueryInfo* pQueryInfo, int32_t outputColInd
   return pExpr;
 }
 
-static int32_t doAddProjectionExprAndResultFields(SQueryInfo* pQueryInfo, SColumnIndex* pIndex, int32_t startPos) {
+static int32_t doAddProjectionExprAndResultFields(SQueryInfo* pQueryInfo, SColumnIndex* pIndex, int32_t startPos, SSqlCmd* pCmd) {
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, pIndex->tableIndex);
 
   int32_t     numOfTotalColumns = 0;
@@ -1778,7 +1775,7 @@ static int32_t doAddProjectionExprAndResultFields(SQueryInfo* pQueryInfo, SColum
   }
 
   for (int32_t j = 0; j < numOfTotalColumns; ++j) {
-    SExprInfo* pExpr = doAddProjectCol(pQueryInfo, j, pIndex->tableIndex);
+    SExprInfo* pExpr = doAddProjectCol(pQueryInfo, j, pIndex->tableIndex, getNewResColId(pCmd));
     tstrncpy(pExpr->base.aliasName, pSchema[j].name, sizeof(pExpr->base.aliasName));
 
     pIndex->columnIndex = j;
@@ -1811,11 +1808,11 @@ int32_t addProjectionExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, t
     if (index.tableIndex == COLUMN_INDEX_INITIAL_VAL) {  // all table columns are required.
       for (int32_t i = 0; i < pQueryInfo->numOfTables; ++i) {
         index.tableIndex = i;
-        int32_t inc = doAddProjectionExprAndResultFields(pQueryInfo, &index, startPos);
+        int32_t inc = doAddProjectionExprAndResultFields(pQueryInfo, &index, startPos, pCmd);
         startPos += inc;
       }
     } else {
-      doAddProjectionExprAndResultFields(pQueryInfo, &index, startPos);
+      doAddProjectionExprAndResultFields(pQueryInfo, &index, startPos, pCmd);
     }
 
     // add the primary timestamp column even though it is not required by user
@@ -1832,7 +1829,7 @@ int32_t addProjectionExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, t
 
     SSchema colSchema = tGetUserSpecifiedColumnSchema(&pItem->pNode->value, &pItem->pNode->token, pItem->aliasName);
     SExprInfo* pExpr =
-        tscAddFuncInSelectClause(pQueryInfo, startPos, TSDB_FUNC_PRJ, &index, &colSchema, TSDB_COL_UDC);
+        tscAddFuncInSelectClause(pQueryInfo, startPos, TSDB_FUNC_PRJ, &index, &colSchema, TSDB_COL_UDC, getNewResColId(pCmd));
 
     // NOTE: the first parameter is reserved for the tag column id during join query process.
     pExpr->base.numOfParams = 2;
@@ -1846,10 +1843,10 @@ int32_t addProjectionExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, t
 
     if (index.columnIndex == TSDB_TBNAME_COLUMN_INDEX) {
       SSchema* colSchema = tGetTbnameColumnSchema();
-      tscAddFuncInSelectClause(pQueryInfo, startPos, TSDB_FUNC_TAGPRJ, &index, colSchema, TSDB_COL_TAG);
+      tscAddFuncInSelectClause(pQueryInfo, startPos, TSDB_FUNC_TAGPRJ, &index, colSchema, TSDB_COL_TAG, getNewResColId(pCmd));
     } else if (index.columnIndex == TSDB_BLOCK_DIST_COLUMN_INDEX) {
       SSchema colSchema = tGetBlockDistColumnSchema();
-      tscAddFuncInSelectClause(pQueryInfo, startPos, TSDB_FUNC_PRJ, &index, &colSchema, TSDB_COL_TAG);
+      tscAddFuncInSelectClause(pQueryInfo, startPos, TSDB_FUNC_PRJ, &index, &colSchema, TSDB_COL_TAG, getNewResColId(pCmd));
     } else {
       STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, index.tableIndex);
       STableMeta*     pTableMeta = pTableMetaInfo->pTableMeta;
@@ -1858,7 +1855,7 @@ int32_t addProjectionExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, t
         return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg1);
       }
 
-      addProjectQueryCol(pQueryInfo, startPos, &index, pItem);
+      addProjectQueryCol(pQueryInfo, startPos, &index, pItem, getNewResColId(pCmd));
       pQueryInfo->type |= TSDB_QUERY_TYPE_PROJECTION_QUERY;
     }
 
@@ -1896,7 +1893,7 @@ static int32_t setExprInfoForFunctions(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SS
     bytes = pSchema->bytes;
   }
   
-  SExprInfo* pExpr = tscExprAppend(pQueryInfo, functionID, pColIndex, type, bytes, getNewResColId(pQueryInfo), bytes, false);
+  SExprInfo* pExpr = tscExprAppend(pQueryInfo, functionID, pColIndex, type, bytes, getNewResColId(pCmd), bytes, false);
   tstrncpy(pExpr->base.aliasName, name, tListLen(pExpr->base.aliasName));
 
   if (cvtFunc.originFuncId == TSDB_FUNC_LAST_ROW && cvtFunc.originFuncId != functionID) {
@@ -2008,7 +2005,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
 
           index = (SColumnIndex){0, PRIMARYKEY_TIMESTAMP_COL_INDEX};
           int32_t size = tDataTypes[TSDB_DATA_TYPE_BIGINT].bytes;
-          pExpr = tscExprAppend(pQueryInfo, functionId, &index, TSDB_DATA_TYPE_BIGINT, size, getNewResColId(pQueryInfo), size, false);
+          pExpr = tscExprAppend(pQueryInfo, functionId, &index, TSDB_DATA_TYPE_BIGINT, size, getNewResColId(pCmd), size, false);
         } else {
           // count the number of table created according to the super table
           if (getColumnIndexByName(pCmd, pToken, pQueryInfo, &index) != TSDB_CODE_SUCCESS) {
@@ -2025,12 +2022,12 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
           }
 
           int32_t size = tDataTypes[TSDB_DATA_TYPE_BIGINT].bytes;
-          pExpr = tscExprAppend(pQueryInfo, functionId, &index, TSDB_DATA_TYPE_BIGINT, size, getNewResColId(pQueryInfo), size, isTag);
+          pExpr = tscExprAppend(pQueryInfo, functionId, &index, TSDB_DATA_TYPE_BIGINT, size, getNewResColId(pCmd), size, isTag);
         }
       } else {  // count(*) is equalled to count(primary_timestamp_key)
         index = (SColumnIndex){0, PRIMARYKEY_TIMESTAMP_COL_INDEX};
         int32_t size = tDataTypes[TSDB_DATA_TYPE_BIGINT].bytes;
-        pExpr = tscExprAppend(pQueryInfo, functionId, &index, TSDB_DATA_TYPE_BIGINT, size, getNewResColId(pQueryInfo), size, false);
+        pExpr = tscExprAppend(pQueryInfo, functionId, &index, TSDB_DATA_TYPE_BIGINT, size, getNewResColId(pCmd), size, false);
       }
 
       pTableMetaInfo = tscGetMetaInfo(pQueryInfo, index.tableIndex);
@@ -2116,7 +2113,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
         colIndex += 1;
         SColumnIndex indexTS = {.tableIndex = index.tableIndex, .columnIndex = 0};
         SExprInfo* pExpr = tscExprAppend(pQueryInfo, TSDB_FUNC_TS_DUMMY, &indexTS, TSDB_DATA_TYPE_TIMESTAMP, TSDB_KEYSIZE,
-                                           getNewResColId(pQueryInfo), TSDB_KEYSIZE, false);
+                                           getNewResColId(pCmd), TSDB_KEYSIZE, false);
 
         SColumnList ids = createColumnList(1, 0, 0);
         insertResultField(pQueryInfo, 0, &ids, TSDB_KEYSIZE, TSDB_DATA_TYPE_TIMESTAMP, aAggs[TSDB_FUNC_TS_DUMMY].name, pExpr);
@@ -2127,7 +2124,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
         return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg6);
       }
 
-      SExprInfo* pExpr = tscExprAppend(pQueryInfo, functionId, &index, resultType, resultSize, getNewResColId(pQueryInfo), resultSize, false);
+      SExprInfo* pExpr = tscExprAppend(pQueryInfo, functionId, &index, resultType, resultSize, getNewResColId(pCmd), resultSize, false);
 
       if (functionId == TSDB_FUNC_LEASTSQR) {
         /* set the leastsquares parameters */
@@ -2341,7 +2338,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
         tscInsertPrimaryTsSourceColumn(pQueryInfo, pTableMetaInfo->pTableMeta->id.uid);
         colIndex += 1;  // the first column is ts
 
-        pExpr = tscExprAppend(pQueryInfo, functionId, &index, resultType, resultSize, getNewResColId(pQueryInfo), resultSize, false);
+        pExpr = tscExprAppend(pQueryInfo, functionId, &index, resultType, resultSize, getNewResColId(pCmd), resultSize, false);
         tscExprAddParams(&pExpr->base, val, TSDB_DATA_TYPE_DOUBLE, sizeof(double));
       } else {
         tVariantDump(pVariant, val, TSDB_DATA_TYPE_BIGINT, true);
@@ -2354,7 +2351,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
         // todo REFACTOR
         // set the first column ts for top/bottom query
         SColumnIndex index1 = {index.tableIndex, PRIMARYKEY_TIMESTAMP_COL_INDEX};
-        pExpr = tscExprAppend(pQueryInfo, TSDB_FUNC_TS, &index1, TSDB_DATA_TYPE_TIMESTAMP, TSDB_KEYSIZE, getNewResColId(pQueryInfo),
+        pExpr = tscExprAppend(pQueryInfo, TSDB_FUNC_TS, &index1, TSDB_DATA_TYPE_TIMESTAMP, TSDB_KEYSIZE, getNewResColId(pCmd),
                                  TSDB_KEYSIZE, false);
         tstrncpy(pExpr->base.aliasName, aAggs[TSDB_FUNC_TS].name, sizeof(pExpr->base.aliasName));
 
@@ -2365,7 +2362,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
 
         colIndex += 1;  // the first column is ts
 
-        pExpr = tscExprAppend(pQueryInfo, functionId, &index, resultType, resultSize, getNewResColId(pQueryInfo), resultSize, false);
+        pExpr = tscExprAppend(pQueryInfo, functionId, &index, resultType, resultSize, getNewResColId(pCmd), resultSize, false);
         tscExprAddParams(&pExpr->base, val, TSDB_DATA_TYPE_BIGINT, sizeof(int64_t));
       }
   
@@ -2451,7 +2448,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
       s.bytes = bytes;
 
       TSDB_QUERY_SET_TYPE(pQueryInfo->type, TSDB_QUERY_TYPE_TAG_FILTER_QUERY);
-      tscAddFuncInSelectClause(pQueryInfo, 0, TSDB_FUNC_TID_TAG, &index, &s, TSDB_COL_TAG);
+      tscAddFuncInSelectClause(pQueryInfo, 0, TSDB_FUNC_TID_TAG, &index, &s, TSDB_COL_TAG, getNewResColId(pCmd));
       
       return TSDB_CODE_SUCCESS;
     }
@@ -2472,7 +2469,7 @@ int32_t addExprAndResultField(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, int32_t col
 
       s.bytes = bytes;
       s.type = (uint8_t)resType;
-      SExprInfo* pExpr = tscAddFuncInSelectClause(pQueryInfo, 0, TSDB_FUNC_BLKINFO, &index, &s, TSDB_COL_TAG);
+      SExprInfo* pExpr = tscAddFuncInSelectClause(pQueryInfo, 0, TSDB_FUNC_BLKINFO, &index, &s, TSDB_COL_TAG, getNewResColId(pCmd));
       pExpr->base.numOfParams = 1;
       pExpr->base.param[0].i64 = pTableMetaInfo->pTableMeta->tableInfo.rowSize;
       pExpr->base.param[0].nType = TSDB_DATA_TYPE_BIGINT;
@@ -4966,7 +4963,7 @@ int32_t validateOrderbyNode(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SSqlNode* pSq
 
         // orderby ts query on super table
         if (tscOrderedProjectionQueryOnSTable(pQueryInfo, 0)) {
-          addPrimaryTsColIntoResult(pQueryInfo);
+          addPrimaryTsColIntoResult(pQueryInfo, pCmd);
         }
       }
     }
@@ -5772,7 +5769,7 @@ void addGroupInfoForSubquery(SSqlObj* pParentObj, SSqlObj* pSql, int32_t subClau
       int16_t type = pTagSchema->type;
       int16_t bytes = pTagSchema->bytes;
 
-      pExpr = tscExprAppend(pQueryInfo, TSDB_FUNC_TAG, &index, type, bytes, getNewResColId(pQueryInfo), bytes, true);
+      pExpr = tscExprAppend(pQueryInfo, TSDB_FUNC_TAG, &index, type, bytes, getNewResColId(&pSql->cmd), bytes, true);
       pExpr->base.colInfo.flag = TSDB_COL_TAG;
 
       // NOTE: tag column does not add to source column list
@@ -5797,7 +5794,7 @@ static void doLimitOutputNormalColOfGroupby(SExprInfo* pExpr) {
   pExpr->base.numOfParams = 1;
 }
 
-void doAddGroupColumnForSubquery(SQueryInfo* pQueryInfo, int32_t tagIndex) {
+void doAddGroupColumnForSubquery(SQueryInfo* pQueryInfo, int32_t tagIndex, SSqlCmd* pCmd) {
   SColIndex* pColIndex = taosArrayGet(pQueryInfo->groupbyExpr.columnInfo, tagIndex);
   size_t size = tscNumOfExprs(pQueryInfo);
 
@@ -5807,7 +5804,7 @@ void doAddGroupColumnForSubquery(SQueryInfo* pQueryInfo, int32_t tagIndex) {
   SColumnIndex colIndex = {.tableIndex = 0, .columnIndex = pColIndex->colIndex};
 
   SExprInfo* pExprInfo = tscAddFuncInSelectClause(pQueryInfo, (int32_t)size, TSDB_FUNC_PRJ, &colIndex, pSchema,
-      TSDB_COL_NORMAL);
+      TSDB_COL_NORMAL, getNewResColId(pCmd));
 
   strncpy(pExprInfo->base.token, pExprInfo->base.colInfo.name, tListLen(pExprInfo->base.token));
 
@@ -6076,7 +6073,7 @@ static int32_t doAddGroupbyColumnsOnDemand(SSqlCmd* pCmd, SQueryInfo* pQueryInfo
     if (TSDB_COL_IS_TAG(pColIndex->flag)) {
       SColumnIndex index = {.tableIndex = pQueryInfo->groupbyExpr.tableIndex, .columnIndex = colIndex};
       SExprInfo*   pExpr = tscExprAppend(pQueryInfo, TSDB_FUNC_TAG, &index, s->type, s->bytes,
-                                          getNewResColId(pQueryInfo), s->bytes, true);
+                                          getNewResColId(pCmd), s->bytes, true);
 
       memset(pExpr->base.aliasName, 0, sizeof(pExpr->base.aliasName));
       tstrncpy(pExpr->base.aliasName, s->name, sizeof(pExpr->base.aliasName));
@@ -6104,7 +6101,7 @@ static int32_t doAddGroupbyColumnsOnDemand(SSqlCmd* pCmd, SQueryInfo* pQueryInfo
 
       //if the group by column does not required by user, add an invisible column into the final result set.
       if (!hasGroupColumn) {
-        doAddGroupColumnForSubquery(pQueryInfo, i);
+        doAddGroupColumnForSubquery(pQueryInfo, i, pCmd);
       }
     }
   }
@@ -6275,7 +6272,7 @@ int32_t doLocalQueryProcess(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SSqlNode* pSq
   
   SColumnIndex ind = {0};
   SExprInfo* pExpr1 = tscExprAppend(pQueryInfo, TSDB_FUNC_TAG_DUMMY, &ind, TSDB_DATA_TYPE_INT,
-                                      tDataTypes[TSDB_DATA_TYPE_INT].bytes, getNewResColId(pQueryInfo), tDataTypes[TSDB_DATA_TYPE_INT].bytes, false);
+                                      tDataTypes[TSDB_DATA_TYPE_INT].bytes, getNewResColId(pCmd), tDataTypes[TSDB_DATA_TYPE_INT].bytes, false);
 
   tSqlExprItem* item = taosArrayGet(pExprList, 0);
   const char* name = (item->aliasName != NULL)? item->aliasName:functionsInfo[index].name;
@@ -7345,8 +7342,10 @@ static int32_t doValidateSubquery(SSqlNode* pSqlNode, int32_t index, SSqlObj* pS
 
   // all columns are added into the table column list
   STableMeta* pMeta = pTableMetaInfo1->pTableMeta;
+  int32_t startOffset = taosArrayGetSize(pQueryInfo->colList);
+
   for(int32_t i = 0; i < pMeta->tableInfo.numOfColumns; ++i) {
-    tscColumnListInsert(pQueryInfo->colList, i, pMeta->id.uid, &pMeta->schema[i]);
+    tscColumnListInsert(pQueryInfo->colList, i + startOffset, pMeta->id.uid, &pMeta->schema[i]);
   }
 
   return TSDB_CODE_SUCCESS;
diff --git a/src/client/src/tscServer.c b/src/client/src/tscServer.c
index ae836299e1d2c4c47c66ccfecfd04491d352a422..08085d7216298a5a922a4ce5890300cc82ef64fa 100644
--- a/src/client/src/tscServer.c
+++ b/src/client/src/tscServer.c
@@ -1644,8 +1644,6 @@ int tscProcessRetrieveLocalMergeRsp(SSqlObj *pSql) {
   qTableQuery(pQueryInfo->pQInfo, &localQueryId);
   convertQueryResult(pRes, pQueryInfo);
 
-//  handleDownstreamOperator(pRes, pQueryInfo);
-
   code = pRes->code;
   if (pRes->code == TSDB_CODE_SUCCESS) {
     (*pSql->fp)(pSql->param, pSql, pRes->numOfRows);
@@ -2189,7 +2187,7 @@ int tscProcessShowRsp(SSqlObj *pSql) {
     SInternalField* pInfo = tscFieldInfoAppend(pFieldInfo, &f);
     
     pInfo->pExpr = tscExprAppend(pQueryInfo, TSDB_FUNC_TS_DUMMY, &index,
-                     pTableSchema[i].type, pTableSchema[i].bytes, getNewResColId(pQueryInfo), pTableSchema[i].bytes, false);
+                     pTableSchema[i].type, pTableSchema[i].bytes, getNewResColId(pCmd), pTableSchema[i].bytes, false);
   }
   
   pCmd->numOfCols = pQueryInfo->fieldsInfo.numOfOutput;
diff --git a/src/client/src/tscSubquery.c b/src/client/src/tscSubquery.c
index c4688ca1b0187161247ba78e51125e333db3bee3..90bc5c985f57aa74a7cbfe00f48a14e1dfa64310 100644
--- a/src/client/src/tscSubquery.c
+++ b/src/client/src/tscSubquery.c
@@ -614,7 +614,7 @@ static int32_t tscLaunchRealSubqueries(SSqlObj* pSql) {
 
       int16_t functionId = tscIsProjectionQuery(pQueryInfo)? TSDB_FUNC_PRJ : TSDB_FUNC_TS;
 
-      tscAddFuncInSelectClause(pQueryInfo, 0, functionId, &index, s, TSDB_COL_NORMAL);
+      tscAddFuncInSelectClause(pQueryInfo, 0, functionId, &index, s, TSDB_COL_NORMAL, getNewResColId(&pNew->cmd));
       tscPrintSelNodeList(pNew, 0);
       tscFieldInfoUpdateOffset(pQueryInfo);
 
@@ -817,7 +817,7 @@ static void issueTsCompQuery(SSqlObj* pSql, SJoinSupporter* pSupporter, SSqlObj*
   SSchema colSchema = {.type = TSDB_DATA_TYPE_BINARY, .bytes = 1};
   
   SColumnIndex index = {0, PRIMARYKEY_TIMESTAMP_COL_INDEX};
-  tscAddFuncInSelectClause(pQueryInfo, 0, TSDB_FUNC_TS_COMP, &index, &colSchema, TSDB_COL_NORMAL);
+  tscAddFuncInSelectClause(pQueryInfo, 0, TSDB_FUNC_TS_COMP, &index, &colSchema, TSDB_COL_NORMAL, getNewResColId(pCmd));
   
   // set the tags value for ts_comp function
   if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
@@ -1906,7 +1906,7 @@ int32_t tscCreateJoinSubquery(SSqlObj *pSql, int16_t tableIndex, SJoinSupporter
 
       // set get tags query type
       TSDB_QUERY_SET_TYPE(pNewQueryInfo->type, TSDB_QUERY_TYPE_TAG_FILTER_QUERY);
-      tscAddFuncInSelectClause(pNewQueryInfo, 0, TSDB_FUNC_TID_TAG, &colIndex, &s1, TSDB_COL_TAG);
+      tscAddFuncInSelectClause(pNewQueryInfo, 0, TSDB_FUNC_TID_TAG, &colIndex, &s1, TSDB_COL_TAG, getNewResColId(pCmd));
       size_t numOfCols = taosArrayGetSize(pNewQueryInfo->colList);
   
       tscDebug(
@@ -1917,7 +1917,7 @@ int32_t tscCreateJoinSubquery(SSqlObj *pSql, int16_t tableIndex, SJoinSupporter
     } else {
       SSchema      colSchema = {.type = TSDB_DATA_TYPE_BINARY, .bytes = 1};
       SColumnIndex colIndex = {0, PRIMARYKEY_TIMESTAMP_COL_INDEX};
-      tscAddFuncInSelectClause(pNewQueryInfo, 0, TSDB_FUNC_TS_COMP, &colIndex, &colSchema, TSDB_COL_NORMAL);
+      tscAddFuncInSelectClause(pNewQueryInfo, 0, TSDB_FUNC_TS_COMP, &colIndex, &colSchema, TSDB_COL_NORMAL, getNewResColId(pCmd));
 
       // set the tags value for ts_comp function
       SExprInfo *pExpr = tscExprGet(pNewQueryInfo, 0);
@@ -2332,7 +2332,7 @@ int32_t tscHandleFirstRoundStableQuery(SSqlObj *pSql) {
       SColumnIndex colIndex = {.tableIndex = 0, .columnIndex = PRIMARYKEY_TIMESTAMP_COL_INDEX};
       SSchema* schema = tscGetColumnSchemaById(pTableMetaInfo1->pTableMeta, pExpr->base.colInfo.colId);
 
-      SExprInfo* p = tscAddFuncInSelectClause(pNewQueryInfo, index++, TSDB_FUNC_TS, &colIndex, schema, TSDB_COL_NORMAL);
+      SExprInfo* p = tscAddFuncInSelectClause(pNewQueryInfo, index++, TSDB_FUNC_TS, &colIndex, schema, TSDB_COL_NORMAL, getNewResColId(pCmd));
       p->base.resColId = pExpr->base.resColId;  // update the result column id
     } else if (pExpr->base.functionId == TSDB_FUNC_STDDEV_DST) {
       taosArrayPush(pSup->pColsInfo, &pExpr->base.resColId);
@@ -2341,7 +2341,7 @@ int32_t tscHandleFirstRoundStableQuery(SSqlObj *pSql) {
       SSchema schema = {.type = TSDB_DATA_TYPE_DOUBLE, .bytes = sizeof(double)};
       tstrncpy(schema.name, pExpr->base.aliasName, tListLen(schema.name));
 
-      SExprInfo* p = tscAddFuncInSelectClause(pNewQueryInfo, index++, TSDB_FUNC_AVG, &colIndex, &schema, TSDB_COL_NORMAL);
+      SExprInfo* p = tscAddFuncInSelectClause(pNewQueryInfo, index++, TSDB_FUNC_AVG, &colIndex, &schema, TSDB_COL_NORMAL, getNewResColId(pCmd));
       p->base.resColId = pExpr->base.resColId;  // update the result column id
     } else if (pExpr->base.functionId == TSDB_FUNC_TAG) {
       pSup->tagLen += pExpr->base.resBytes;
@@ -2354,7 +2354,7 @@ int32_t tscHandleFirstRoundStableQuery(SSqlObj *pSql) {
         schema = tGetTbnameColumnSchema();
       }
 
-      SExprInfo* p = tscAddFuncInSelectClause(pNewQueryInfo, index++, TSDB_FUNC_TAG, &colIndex, schema, TSDB_COL_TAG);
+      SExprInfo* p = tscAddFuncInSelectClause(pNewQueryInfo, index++, TSDB_FUNC_TAG, &colIndex, schema, TSDB_COL_TAG, getNewResColId(pCmd));
       p->base.resColId = pExpr->base.resColId;
     } else if (pExpr->base.functionId == TSDB_FUNC_PRJ) {
       int32_t num = (int32_t) taosArrayGetSize(pNewQueryInfo->groupbyExpr.columnInfo);
@@ -2368,7 +2368,7 @@ int32_t tscHandleFirstRoundStableQuery(SSqlObj *pSql) {
           SSchema* schema = tscGetColumnSchemaById(pTableMetaInfo1->pTableMeta, pExpr->base.colInfo.colId);
 
           //doLimitOutputNormalColOfGroupby
-          SExprInfo* p = tscAddFuncInSelectClause(pNewQueryInfo, index++, TSDB_FUNC_PRJ, &colIndex, schema, TSDB_COL_NORMAL);
+          SExprInfo* p = tscAddFuncInSelectClause(pNewQueryInfo, index++, TSDB_FUNC_PRJ, &colIndex, schema, TSDB_COL_NORMAL, getNewResColId(pCmd));
           p->base.numOfParams = 1;
           p->base.param[0].i64 = 1;
           p->base.param[0].nType = TSDB_DATA_TYPE_INT;
diff --git a/src/client/src/tscUtil.c b/src/client/src/tscUtil.c
index c12aa4447d04b23fb7db412a8e2555dc3ca3ce6e..a35d44bba7befc0de719c1fb4e31cf7951523695 100644
--- a/src/client/src/tscUtil.c
+++ b/src/client/src/tscUtil.c
@@ -14,10 +14,11 @@
  */
 
 #include "tscUtil.h"
+#include "tsched.h"
 #include "hash.h"
 #include "os.h"
-#include "texpr.h"
 #include "taosmsg.h"
+#include "texpr.h"
 #include "tkey.h"
 #include "tmd5.h"
 #include "tscLocalMerge.h"
@@ -634,18 +635,12 @@ void tscSetResRawPtrRv(SSqlRes* pRes, SQueryInfo* pQueryInfo, SSDataBlock* pBloc
   }
 }
 
-static SColumnInfo* extractColumnInfoFromResult(STableMeta* pTableMeta, SArray* pTableCols) {
+static SColumnInfo* extractColumnInfoFromResult(SArray* pTableCols) {
   int32_t numOfCols = (int32_t) taosArrayGetSize(pTableCols);
   SColumnInfo* pColInfo = calloc(numOfCols, sizeof(SColumnInfo));
-
-  SSchema *pSchema = pTableMeta->schema;
   for(int32_t i = 0; i < numOfCols; ++i) {
     SColumn* pCol = taosArrayGetP(pTableCols, i);
-    int32_t index = pCol->columnIndex;
-
-    pColInfo[i].type  = pSchema[index].type;
-    pColInfo[i].bytes = pSchema[index].bytes;
-    pColInfo[i].colId = pSchema[index].colId;
+    pColInfo[i] = pCol->info;//[index].type;
   }
 
   return pColInfo;
@@ -656,6 +651,10 @@ typedef struct SDummyInputInfo {
   SSqlRes        *pRes;  // refactor: remove it
 } SDummyInputInfo;
 
+typedef struct SJoinOperatorInfo {
+  int32_t a;
+} SJoinOperatorInfo;
+
 SSDataBlock* doGetDataBlock(void* param, bool* newgroup) {
   SOperatorInfo *pOperator = (SOperatorInfo*) param;
 
@@ -729,6 +728,39 @@ SOperatorInfo* createDummyInputOperator(char* pResult, SSchema* pSchema, int32_t
   return pOptr;
 }
 
+SOperatorInfo* createJoinOperator(SOperatorInfo** pUpstream, int32_t numOfUpstream, SExprInfo* pExprInfo, int32_t numOfOutput) {
+  SJoinInfo* pInfo = calloc(1, sizeof(SJoinInfo));
+/*
+  pInfo->pRes  = (SSqlRes*) pResult;
+  pInfo->block = calloc(numOfCols, sizeof(SSDataBlock));
+  pInfo->block->info.numOfCols = numOfCols;
+
+  pInfo->block->pDataBlock = taosArrayInit(numOfCols, sizeof(SColumnInfoData));
+  for(int32_t i = 0; i < numOfCols; ++i) {
+    SColumnInfoData colData = {{0}};
+    colData.info.bytes = pSchema[i].bytes;
+    colData.info.type  = pSchema[i].type;
+    colData.info.colId = pSchema[i].colId;
+
+    taosArrayPush(pInfo->block->pDataBlock, &colData);
+  }
+*/
+  SOperatorInfo* pOptr = calloc(1, sizeof(SOperatorInfo));
+  pOptr->name          = "JoinOperator";
+  pOptr->operatorType  = OP_Join;
+  pOptr->numOfOutput   = numOfOutput;
+  pOptr->blockingOptr  = false;
+  pOptr->info          = pInfo;
+  pOptr->exec          = doGetDataBlock;
+  pOptr->cleanup       = destroyDummyInputOperator;
+
+  for(int32_t i = 0; i < numOfUpstream; ++i) {
+    appendUpstream(pOptr, pUpstream[0]);
+  }
+
+  return pOptr;
+}
+
 void convertQueryResult(SSqlRes* pRes, SQueryInfo* pQueryInfo) {
   // set the correct result
   SSDataBlock* p = pQueryInfo->pQInfo->runtimeEnv.outputBuf;
@@ -743,40 +775,51 @@ void convertQueryResult(SSqlRes* pRes, SQueryInfo* pQueryInfo) {
   pRes->completed = (pRes->numOfRows == 0);
 }
 
-void handleDownstreamOperator(SSqlRes* pRes, SQueryInfo* pQueryInfo, SSqlRes* pOutput) {
-  if (pQueryInfo->pDownstream != NULL) {
-    // handle the following query process
-    SQueryInfo *px = pQueryInfo->pDownstream;
-    if (px->pQInfo == NULL) {
-      SColumnInfo* pColumnInfo = extractColumnInfoFromResult(px->pTableMetaInfo[0]->pTableMeta, px->colList);
-      int32_t numOfOutput = (int32_t) tscNumOfExprs(px);
+void handleDownstreamOperator(SSqlRes* pRes, SQueryInfo* px, SSqlRes* pOutput) {
+  // handle the following query process
+  if (px->pQInfo == NULL) {
+    SColumnInfo* pColumnInfo = extractColumnInfoFromResult(px->colList);
+
+    int32_t numOfOutput = (int32_t)tscNumOfExprs(px);
+    int32_t numOfCols = (int32_t)taosArrayGetSize(px->colList);
+
+    SQueriedTableInfo info = {
+        .colList = pColumnInfo,
+        .numOfCols = numOfCols,
+    };
+
+    SSchema* pSchema = tscGetTableSchema(px->pTableMetaInfo[0]->pTableMeta);
 
-      int32_t numOfCols = (int32_t) taosArrayGetSize(px->colList);
-      SQueriedTableInfo info = {.colList = pColumnInfo, .numOfCols = numOfCols,};
-      SSchema* pSchema = tscGetTableSchema(px->pTableMetaInfo[0]->pTableMeta);
+    STableGroupInfo tableGroupInfo = {
+        .numOfTables = 1,
+        .pGroupList = taosArrayInit(1, POINTER_BYTES),
+    };
 
-      STableGroupInfo tableGroupInfo = {.numOfTables = 1, .pGroupList = taosArrayInit(1, POINTER_BYTES),};
-      tableGroupInfo.map = taosHashInit(1, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_NO_LOCK);
+    tableGroupInfo.map = taosHashInit(1, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, HASH_NO_LOCK);
 
-      STableKeyInfo tableKeyInfo = {.pTable = NULL, .lastKey = INT64_MIN};
+    STableKeyInfo tableKeyInfo = {.pTable = NULL, .lastKey = INT64_MIN};
 
-      SArray* group = taosArrayInit(1, sizeof(STableKeyInfo));
-      taosArrayPush(group, &tableKeyInfo);
+    SArray* group = taosArrayInit(1, sizeof(STableKeyInfo));
+    taosArrayPush(group, &tableKeyInfo);
 
-      taosArrayPush(tableGroupInfo.pGroupList, &group);
+    taosArrayPush(tableGroupInfo.pGroupList, &group);
 
-      SOperatorInfo* pSourceOptr = createDummyInputOperator((char*)pRes, pSchema, numOfCols);
+    // if it is a join query, create join operator here
+    SOperatorInfo* pSourceOperator = createDummyInputOperator((char*)pRes, pSchema, numOfCols);
+    if (px->numOfTables > 1) {
 
-      SExprInfo *exprInfo = NULL;
-      /*int32_t code = */createQueryFunc(&info, numOfOutput, &exprInfo, px->exprList->pData, NULL, px->type, NULL);
-      px->pQInfo = createQueryInfoFromQueryNode(px, exprInfo, &tableGroupInfo, pSourceOptr, NULL, NULL, MASTER_SCAN);
-      tfree(pColumnInfo);
+      pSourceOperator = createJoinOperator(&pSourceOperator, 1, NULL, pSourceOperator->numOfOutput);
     }
 
-    uint64_t qId = 0;
-    qTableQuery(px->pQInfo, &qId);
-    convertQueryResult(pOutput, px);
+    SExprInfo* exprInfo = NULL;
+    /*int32_t code = */ createQueryFunc(&info, numOfOutput, &exprInfo, px->exprList->pData, NULL, px->type, NULL);
+    px->pQInfo = createQueryInfoFromQueryNode(px, exprInfo, &tableGroupInfo, pSourceOperator, NULL, NULL, MASTER_SCAN);
+    tfree(pColumnInfo);
   }
+
+  uint64_t qId = 0;
+  qTableQuery(px->pQInfo, &qId);
+  convertQueryResult(pOutput, px);
 }
 
 static void tscDestroyResPointerInfo(SSqlRes* pRes) {
@@ -2340,7 +2383,6 @@ void tscInitQueryInfo(SQueryInfo* pQueryInfo) {
   pQueryInfo->exprList       = taosArrayInit(4, POINTER_BYTES);
   pQueryInfo->colList        = taosArrayInit(4, POINTER_BYTES);
   pQueryInfo->udColumnId     = TSDB_UD_COLUMN_INDEX;
-  pQueryInfo->resColumnId    = TSDB_RES_COL_ID;
   pQueryInfo->limit.limit    = -1;
   pQueryInfo->limit.offset   = 0;
 
@@ -2949,6 +2991,40 @@ void doExecuteQuery(SSqlObj* pSql, SQueryInfo* pQueryInfo) {
   }
 }
 
+static void doRetrieveSubqueryData(SSchedMsg *pMsg) {
+  SSqlObj* pSql = (SSqlObj*)taosAcquireRef(tscObjRef, (int64_t)pMsg->ahandle);
+  if (pSql == NULL || pSql->signature != pSql) {
+    tscDebug("%p SqlObj is freed, not add into queue async res", pMsg->ahandle);
+    return;
+  }
+
+  int32_t numOfRows = 0;
+  for(int32_t i = 0; i < pSql->subState.numOfSub; ++i) {
+    SSqlObj*    pSub = pSql->pSubs[i];
+    /*TAOS_ROW row = */taos_fetch_row(pSub);
+//    SQueryInfo* pQueryInfo = tscGetQueryInfo(&pSub->cmd);
+//    int32_t rows = taos_fetch_block(pSub, &row);
+    if (numOfRows == 0) {
+      numOfRows = pSub->res.numOfRows;
+    }
+  }
+
+  if (numOfRows > 0) {
+    SQueryInfo *pQueryInfo = tscGetQueryInfo(&pSql->cmd);
+    SSqlObj    *pSub = pSql->pSubs[0];
+    handleDownstreamOperator(&pSub->res, pQueryInfo, &pSql->res);
+  }
+
+//  int32_t code = pSql->res.code;
+  pSql->res.qId = -1;
+  if (pSql->res.code == TSDB_CODE_SUCCESS) {
+    (*pSql->fp)(pSql->param, pSql, pSql->res.numOfRows);
+  } else {
+    tscAsyncResultOnError(pSql);
+  }
+}
+
+// NOTE: the blocking query can not be executed in the rpc message handler thread
 static void tscSubqueryRetrieveCallback(void* param, TAOS_RES* tres, int code) {
   // handle the pDownStream process
   SRetrieveSupport* ps = param;
@@ -2960,27 +3036,37 @@ static void tscSubqueryRetrieveCallback(void* param, TAOS_RES* tres, int code) {
     return;
   }
 
+  pParentSql->cmd.active = pParentSql->cmd.pQueryInfo;
+
+  SSchedMsg schedMsg = {0};
+  schedMsg.fp = doRetrieveSubqueryData;
+  schedMsg.ahandle = (void *)pParentSql->self;
+  schedMsg.thandle = (void *)1;
+  schedMsg.msg = 0;
+  taosScheduleTask(tscQhandle, &schedMsg);
+
   // merge all subquery result
-  SSqlCmd* pCmd = &pSql->cmd;
-  SSqlRes* pRes = &pSql->res;
+//  SSqlCmd* pCmd = &pSql->cmd;
+//  SSqlRes* pRes = &pSql->res;
 
-  SQueryInfo* pQueryInfo = tscGetQueryInfo(pCmd);
-  /*TAOS_ROW* pRow = */taos_fetch_row(pSql);
-  if (pSql->res.numOfRows > 0) {
-    handleDownstreamOperator(pRes, pQueryInfo, &pParentSql->res);
-  }
+  // add it to the message queue
 
-  code = pParentSql->res.code;
-  pParentSql->res.qId = -1;
-  if (pParentSql->res.code == TSDB_CODE_SUCCESS) {
-    (*pParentSql->fp)(pParentSql->param, pParentSql, pParentSql->res.numOfRows);
-  } else {
-    tscAsyncResultOnError(pParentSql);
-  }
+//  SQueryInfo* pQueryInfo = tscGetQueryInfo(pCmd);
+//  /*TAOS_ROW* pRow = */taos_fetch_row(pSql);
+//  if (pSql->res.numOfRows > 0) {
+//    handleDownstreamOperator(pRes, pQueryInfo, &pParentSql->res);
+//  }
+//
+//  code = pParentSql->res.code;
+//  pParentSql->res.qId = -1;
+//  if (pParentSql->res.code == TSDB_CODE_SUCCESS) {
+//    (*pParentSql->fp)(pParentSql->param, pParentSql, pParentSql->res.numOfRows);
+//  } else {
+//    tscAsyncResultOnError(pParentSql);
+//  }
 }
 
 static void tscSubqueryCompleteCallback(void* param, TAOS_RES* tres, int code) {
-  printf("123\n");
   taos_fetch_rows_a(tres, tscSubqueryRetrieveCallback, param);
 }
 
@@ -3000,43 +3086,44 @@ void executeQuery(SSqlObj* pSql, SQueryInfo* pQueryInfo) {
     pSql->pSubs = calloc(pSql->subState.numOfSub, POINTER_BYTES);
     pSql->subState.states = calloc(pSql->subState.numOfSub, sizeof(int8_t));
 
-    SQueryInfo* pSub = taosArrayGetP(pQueryInfo->pUpstream, 0);
+    for(int32_t i = 0; i < pSql->subState.numOfSub; ++i) {
+      SQueryInfo* pSub = taosArrayGetP(pQueryInfo->pUpstream, i);
 
-    pSql->cmd.active = pSub;
-    pSql->cmd.command = TSDB_SQL_SELECT;
+      pSql->cmd.active = pSub;
+      pSql->cmd.command = TSDB_SQL_SELECT;
 
-    SSqlObj* pNew = (SSqlObj*)calloc(1, sizeof(SSqlObj));
-    if (pNew == NULL) {
-      terrno = TSDB_CODE_TSC_OUT_OF_MEMORY;
-//      return NULL;
-    }
-
-    pNew->pTscObj   = pSql->pTscObj;
-    pNew->signature = pNew;
-    pNew->sqlstr    = strdup(pSql->sqlstr); // todo refactor
-    pNew->fp        = tscSubqueryCompleteCallback;
+      SSqlObj* pNew = (SSqlObj*)calloc(1, sizeof(SSqlObj));
+      if (pNew == NULL) {
+        terrno = TSDB_CODE_TSC_OUT_OF_MEMORY;
+        //      return NULL;
+      }
 
-    SRetrieveSupport* ps = calloc(1, sizeof(SRetrieveSupport));// todo use object id
-    ps->pParentSql    = pSql;
-    ps->subqueryIndex = 0;
+      pNew->pTscObj = pSql->pTscObj;
+      pNew->signature = pNew;
+      pNew->sqlstr = strdup(pSql->sqlstr);  // todo refactor
+      pNew->fp = tscSubqueryCompleteCallback;
 
-    pNew->param    = ps;
-    pSql->pSubs[0] = pNew;
-    registerSqlObj(pNew);
+      SRetrieveSupport* ps = calloc(1, sizeof(SRetrieveSupport));  // todo use object id
+      ps->pParentSql = pSql;
+      ps->subqueryIndex = i;
 
-    SSqlCmd* pCmd = &pNew->cmd;
-    pCmd->command = TSDB_SQL_SELECT;
-    pCmd->parseFinished = 1;
+      pNew->param = ps;
+      pSql->pSubs[i] = pNew;
+      registerSqlObj(pNew);
 
-    if (tscAddQueryInfo(pCmd) != TSDB_CODE_SUCCESS) {
-    }
+      SSqlCmd* pCmd = &pNew->cmd;
+      pCmd->command = TSDB_SQL_SELECT;
+      pCmd->parseFinished = 1;
 
-    SQueryInfo* pNewQueryInfo = tscGetQueryInfo(pCmd);
-    tscQueryInfoCopy(pNewQueryInfo, pSub);
+      if (tscAddQueryInfo(pCmd) != TSDB_CODE_SUCCESS) {
+      }
 
-    // create sub query to handle the sub query.
-    executeQuery(pNew, pSub);
+      SQueryInfo* pNewQueryInfo = tscGetQueryInfo(pCmd);
+      tscQueryInfoCopy(pNewQueryInfo, pSub);
 
+      // create sub query to handle the sub query.
+      executeQuery(pNew, pSub);
+    }
     // merge sub query result and generate final results
     return;
   }
diff --git a/src/query/inc/qExecutor.h b/src/query/inc/qExecutor.h
index 3e0f1f5e8f9b5f9d9994c50c0d68d1b5e1e749ef..8e22b4e20814d0d1303183378404f53a1e51c204 100644
--- a/src/query/inc/qExecutor.h
+++ b/src/query/inc/qExecutor.h
@@ -302,6 +302,7 @@ enum OPERATOR_TYPE_E {
   OP_GlobalAggregate   = 18,   // global merge for the multi-way data sources.
   OP_Filter            = 19,
   OP_Distinct          = 20,
+  OP_Join              = 21,
 };
 
 typedef struct SOperatorInfo {
@@ -314,7 +315,8 @@ typedef struct SOperatorInfo {
   SExprInfo            *pExpr;
   SQueryRuntimeEnv     *pRuntimeEnv;
 
-  struct SOperatorInfo *upstream;
+  struct SOperatorInfo **upstream;     // upstream pointer list
+  int32_t               numOfUpstream; // number of upstream. The value is always ONE expect for join operator
   __operator_fn_t       exec;
   __optr_cleanup_fn_t   cleanup;
 } SOperatorInfo;
@@ -494,6 +496,8 @@ typedef struct SMultiwayMergeInfo {
   bool                 groupMix;
 } SMultiwayMergeInfo;
 
+void appendUpstream(SOperatorInfo* p, SOperatorInfo* pUpstream);
+
 SOperatorInfo* createDataBlocksOptScanInfo(void* pTsdbQueryHandle, SQueryRuntimeEnv* pRuntimeEnv, int32_t repeatTime, int32_t reverseTime);
 SOperatorInfo* createTableScanOperator(void* pTsdbQueryHandle, SQueryRuntimeEnv* pRuntimeEnv, int32_t repeatTime);
 SOperatorInfo* createTableSeqScanOperator(void* pTsdbQueryHandle, SQueryRuntimeEnv* pRuntimeEnv);
@@ -517,6 +521,8 @@ SOperatorInfo* createSLimitOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperator
 SOperatorInfo* createFilterOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperatorInfo* upstream, SExprInfo* pExpr,
                                         int32_t numOfOutput, SColumnInfo* pCols, int32_t numOfFilter);
 
+SOperatorInfo* createJoinOperator(SOperatorInfo** pUpstream, int32_t numOfUpstream, SExprInfo* pExprInfo, int32_t numOfOutput);
+
 SSDataBlock* doGlobalAggregate(void* param, bool* newgroup);
 SSDataBlock* doMultiwayMergeSort(void* param, bool* newgroup);
 SSDataBlock* doSLimit(void* param, bool* newgroup);
diff --git a/src/query/src/qExecutor.c b/src/query/src/qExecutor.c
index da857533ace6ed84d97600b36dc65f794b51ae46..5035e9cbd7f6c272c707249f74c7f6cbdd50fbcd 100644
--- a/src/query/src/qExecutor.c
+++ b/src/query/src/qExecutor.c
@@ -1710,38 +1710,38 @@ static int32_t setupQueryRuntimeEnv(SQueryRuntimeEnv *pRuntimeEnv, int32_t numOf
       case OP_MultiTableTimeInterval: {
         pRuntimeEnv->proot =
             createMultiTableTimeIntervalOperatorInfo(pRuntimeEnv, pRuntimeEnv->proot, pQueryAttr->pExpr1, pQueryAttr->numOfOutput);
-        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream->info, pRuntimeEnv->proot);
+        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream[0]->info, pRuntimeEnv->proot);
         break;
       }
       case OP_TimeWindow: {
         pRuntimeEnv->proot =
             createTimeIntervalOperatorInfo(pRuntimeEnv, pRuntimeEnv->proot, pQueryAttr->pExpr1, pQueryAttr->numOfOutput);
-        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream->info, pRuntimeEnv->proot);
+        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream[0]->info, pRuntimeEnv->proot);
         break;
       }
       case OP_Groupby: {
         pRuntimeEnv->proot =
             createGroupbyOperatorInfo(pRuntimeEnv, pRuntimeEnv->proot, pQueryAttr->pExpr1, pQueryAttr->numOfOutput);
-        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream->info, pRuntimeEnv->proot);
+        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream[0]->info, pRuntimeEnv->proot);
         break;
       }
       case OP_SessionWindow: {
         pRuntimeEnv->proot =
             createSWindowOperatorInfo(pRuntimeEnv, pRuntimeEnv->proot, pQueryAttr->pExpr1, pQueryAttr->numOfOutput);
-        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream->info, pRuntimeEnv->proot);
+        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream[0]->info, pRuntimeEnv->proot);
         break;
       }
       case OP_MultiTableAggregate: {
         pRuntimeEnv->proot =
             createMultiTableAggOperatorInfo(pRuntimeEnv, pRuntimeEnv->proot, pQueryAttr->pExpr1, pQueryAttr->numOfOutput);
-        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream->info, pRuntimeEnv->proot);
+        setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream[0]->info, pRuntimeEnv->proot);
         break;
       }
       case OP_Aggregate: {
         pRuntimeEnv->proot =
             createAggregateOperatorInfo(pRuntimeEnv, pRuntimeEnv->proot, pQueryAttr->pExpr1, pQueryAttr->numOfOutput);
-        if (pRuntimeEnv->proot->upstream->operatorType != OP_DummyInput) {
-          setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream->info, pRuntimeEnv->proot);
+        if (pRuntimeEnv->proot->upstream[0]->operatorType != OP_DummyInput) {
+          setTableScanFilterOperatorInfo(pRuntimeEnv->proot->upstream[0]->info, pRuntimeEnv->proot);
         }
         break;
       }
@@ -3927,6 +3927,15 @@ void queryCostStatis(SQInfo *pQInfo) {
 //  return true;
 //}
 
+void appendUpstream(SOperatorInfo* p, SOperatorInfo* pUpstream) {
+  if (p->upstream == NULL) {
+    assert(p->numOfOutput == 0);
+  }
+
+  p->upstream = realloc(p->upstream, POINTER_BYTES * (p->numOfOutput + 1));
+  p->upstream[p->numOfOutput++] = pUpstream;
+}
+
 static void doDestroyTableQueryInfo(STableGroupInfo* pTableqinfoGroupInfo);
 
 static int32_t setupQueryHandle(void* tsdb, SQueryRuntimeEnv* pRuntimeEnv, int64_t qId, bool isSTableQuery) {
@@ -4604,13 +4613,14 @@ SOperatorInfo* createGlobalAggregateOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv,
   pOperator->blockingOptr = true;
   pOperator->status       = OP_IN_EXECUTING;
   pOperator->info         = pInfo;
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->pRuntimeEnv  = pRuntimeEnv;
 
   pOperator->exec         = doGlobalAggregate;
   pOperator->cleanup      = destroyGlobalAggOperatorInfo;
+  appendUpstream(pOperator, upstream);
+
   return pOperator;
 }
 
@@ -4675,7 +4685,7 @@ static SSDataBlock* doAggregate(void* param, bool* newgroup) {
   SQueryAttr* pQueryAttr = pRuntimeEnv->pQueryAttr;
   int32_t order = pQueryAttr->order.order;
 
-  SOperatorInfo* upstream = pOperator->upstream;
+  SOperatorInfo* upstream = pOperator->upstream[0];
 
   while(1) {
     SSDataBlock* pBlock = upstream->exec(upstream, newgroup);
@@ -4730,7 +4740,7 @@ static SSDataBlock* doSTableAggregate(void* param, bool* newgroup) {
   SQueryAttr* pQueryAttr = pRuntimeEnv->pQueryAttr;
   int32_t order = pQueryAttr->order.order;
 
-  SOperatorInfo* upstream = pOperator->upstream;
+  SOperatorInfo* upstream = pOperator->upstream[0];
 
   while(1) {
     SSDataBlock* pBlock = upstream->exec(upstream, newgroup);
@@ -4814,7 +4824,7 @@ static SSDataBlock* doArithmeticOperation(void* param, bool* newgroup) {
     bool prevVal = *newgroup;
 
     // The upstream exec may change the value of the newgroup, so use a local variable instead.
-    SSDataBlock* pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+    SSDataBlock* pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
     if (pBlock == NULL) {
       assert(*newgroup == false);
 
@@ -4868,7 +4878,7 @@ static SSDataBlock* doLimit(void* param, bool* newgroup) {
 
   SSDataBlock* pBlock = NULL;
   while (1) {
-    pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+    pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
     if (pBlock == NULL) {
       setQueryStatus(pOperator->pRuntimeEnv, QUERY_COMPLETED);
       pOperator->status = OP_EXEC_DONE;
@@ -4939,7 +4949,7 @@ static SSDataBlock* doFilter(void* param, bool* newgroup) {
   SQueryRuntimeEnv* pRuntimeEnv = pOperator->pRuntimeEnv;
 
   while (1) {
-    SSDataBlock *pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+    SSDataBlock *pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
     if (pBlock == NULL) {
       break;
     }
@@ -4981,7 +4991,7 @@ static SSDataBlock* doIntervalAgg(void* param, bool* newgroup) {
   int32_t order = pQueryAttr->order.order;
   STimeWindow win = pQueryAttr->window;
 
-  SOperatorInfo* upstream = pOperator->upstream;
+  SOperatorInfo* upstream = pOperator->upstream[0];
 
   while(1) {
     SSDataBlock* pBlock = upstream->exec(upstream, newgroup);
@@ -5034,7 +5044,7 @@ static SSDataBlock* doSTableIntervalAgg(void* param, bool* newgroup) {
   SQueryAttr* pQueryAttr = pRuntimeEnv->pQueryAttr;
   int32_t order = pQueryAttr->order.order;
 
-  SOperatorInfo* upstream = pOperator->upstream;
+  SOperatorInfo* upstream = pOperator->upstream[0];
 
   while(1) {
     SSDataBlock* pBlock = upstream->exec(upstream, newgroup);
@@ -5089,7 +5099,7 @@ static SSDataBlock* doSessionWindowAgg(void* param, bool* newgroup) {
   int32_t order = pQueryAttr->order.order;
   STimeWindow win = pQueryAttr->window;
 
-  SOperatorInfo* upstream = pOperator->upstream;
+  SOperatorInfo* upstream = pOperator->upstream[0];
 
   while(1) {
     SSDataBlock* pBlock = upstream->exec(upstream, newgroup);
@@ -5140,7 +5150,7 @@ static SSDataBlock* hashGroupbyAggregate(void* param, bool* newgroup) {
     return pInfo->binfo.pRes;
   }
 
-  SOperatorInfo* upstream = pOperator->upstream;
+  SOperatorInfo* upstream = pOperator->upstream[0];
 
   while(1) {
     SSDataBlock* pBlock = upstream->exec(upstream, newgroup);
@@ -5209,7 +5219,7 @@ static SSDataBlock* doFill(void* param, bool* newgroup) {
   }
 
   while(1) {
-    SSDataBlock* pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+    SSDataBlock* pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
     if (*newgroup) {
       assert(pBlock != NULL);
     }
@@ -5285,7 +5295,7 @@ static void destroyOperatorInfo(SOperatorInfo* pOperator) {
     pOperator->cleanup(pOperator->info, pOperator->numOfOutput);
   }
 
-  destroyOperatorInfo(pOperator->upstream);
+  destroyOperatorInfo(pOperator->upstream[0]);
   tfree(pOperator->info);
   tfree(pOperator);
 }
@@ -5310,13 +5320,14 @@ SOperatorInfo* createAggregateOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOpera
   pOperator->blockingOptr = true;
   pOperator->status       = OP_IN_EXECUTING;
   pOperator->info         = pInfo;
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->pRuntimeEnv  = pRuntimeEnv;
 
   pOperator->exec         = doAggregate;
   pOperator->cleanup      = destroyBasicOperatorInfo;
+  appendUpstream(pOperator, upstream);
+
   return pOperator;
 }
 
@@ -5383,13 +5394,13 @@ SOperatorInfo* createMultiTableAggOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SO
   pOperator->blockingOptr = true;
   pOperator->status       = OP_IN_EXECUTING;
   pOperator->info         = pInfo;
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->pRuntimeEnv  = pRuntimeEnv;
 
   pOperator->exec         = doSTableAggregate;
   pOperator->cleanup      = destroyBasicOperatorInfo;
+  appendUpstream(pOperator, upstream);
 
   return pOperator;
 }
@@ -5413,13 +5424,13 @@ SOperatorInfo* createArithOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperatorI
   pOperator->blockingOptr = false;
   pOperator->status       = OP_IN_EXECUTING;
   pOperator->info         = pInfo;
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->pRuntimeEnv  = pRuntimeEnv;
 
   pOperator->exec         = doArithmeticOperation;
   pOperator->cleanup      = destroyArithOperatorInfo;
+  appendUpstream(pOperator, upstream);
 
   return pOperator;
 }
@@ -5464,11 +5475,11 @@ SOperatorInfo* createFilterOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperator
   pOperator->status       = OP_IN_EXECUTING;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->pExpr        = pExpr;
-  pOperator->upstream     = upstream;
   pOperator->exec         = doFilter;
   pOperator->info         = pInfo;
   pOperator->pRuntimeEnv  = pRuntimeEnv;
   pOperator->cleanup      = destroyConditionOperatorInfo;
+  appendUpstream(pOperator, upstream);
 
   return pOperator;
 }
@@ -5483,10 +5494,10 @@ SOperatorInfo* createLimitOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperatorI
   pOperator->operatorType = OP_Limit;
   pOperator->blockingOptr = false;
   pOperator->status       = OP_IN_EXECUTING;
-  pOperator->upstream     = upstream;
   pOperator->exec         = doLimit;
   pOperator->info         = pInfo;
   pOperator->pRuntimeEnv  = pRuntimeEnv;
+  appendUpstream(pOperator, upstream);
 
   return pOperator;
 }
@@ -5504,7 +5515,6 @@ SOperatorInfo* createTimeIntervalOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOp
   pOperator->operatorType = OP_TimeWindow;
   pOperator->blockingOptr = true;
   pOperator->status       = OP_IN_EXECUTING;
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->info         = pInfo;
@@ -5512,6 +5522,7 @@ SOperatorInfo* createTimeIntervalOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOp
   pOperator->exec         = doIntervalAgg;
   pOperator->cleanup      = destroyBasicOperatorInfo;
 
+  appendUpstream(pOperator, upstream);
   return pOperator;
 }
 
@@ -5529,7 +5540,6 @@ SOperatorInfo* createSWindowOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperato
   pOperator->operatorType = OP_SessionWindow;
   pOperator->blockingOptr = true;
   pOperator->status       = OP_IN_EXECUTING;
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->info         = pInfo;
@@ -5537,6 +5547,7 @@ SOperatorInfo* createSWindowOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperato
   pOperator->exec         = doSessionWindowAgg;
   pOperator->cleanup      = destroyBasicOperatorInfo;
 
+  appendUpstream(pOperator, upstream);
   return pOperator;
 }
 
@@ -5552,7 +5563,6 @@ SOperatorInfo* createMultiTableTimeIntervalOperatorInfo(SQueryRuntimeEnv* pRunti
   pOperator->operatorType = OP_MultiTableTimeInterval;
   pOperator->blockingOptr = true;
   pOperator->status       = OP_IN_EXECUTING;
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->info         = pInfo;
@@ -5561,6 +5571,7 @@ SOperatorInfo* createMultiTableTimeIntervalOperatorInfo(SQueryRuntimeEnv* pRunti
   pOperator->exec         = doSTableIntervalAgg;
   pOperator->cleanup      = destroyBasicOperatorInfo;
 
+  appendUpstream(pOperator, upstream);
   return pOperator;
 }
 
@@ -5577,7 +5588,6 @@ SOperatorInfo* createGroupbyOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperato
   pOperator->blockingOptr = true;
   pOperator->status       = OP_IN_EXECUTING;
   pOperator->operatorType = OP_Groupby;
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->info         = pInfo;
@@ -5585,6 +5595,7 @@ SOperatorInfo* createGroupbyOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperato
   pOperator->exec         = hashGroupbyAggregate;
   pOperator->cleanup      = destroyGroupbyOperatorInfo;
 
+  appendUpstream(pOperator, upstream);
   return pOperator;
 }
 
@@ -5614,8 +5625,6 @@ SOperatorInfo* createFillOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperatorIn
   pOperator->blockingOptr = false;
   pOperator->status       = OP_IN_EXECUTING;
   pOperator->operatorType = OP_Fill;
-
-  pOperator->upstream     = upstream;
   pOperator->pExpr        = pExpr;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->info         = pInfo;
@@ -5624,6 +5633,7 @@ SOperatorInfo* createFillOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperatorIn
   pOperator->exec         = doFill;
   pOperator->cleanup      = destroySFillOperatorInfo;
 
+  appendUpstream(pOperator, upstream);
   return pOperator;
 }
 
@@ -5662,11 +5672,12 @@ SOperatorInfo* createSLimitOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperator
   pOperator->operatorType = OP_SLimit;
   pOperator->blockingOptr = false;
   pOperator->status       = OP_IN_EXECUTING;
-  pOperator->upstream     = upstream;
   pOperator->exec         = doSLimit;
   pOperator->info         = pInfo;
   pOperator->pRuntimeEnv  = pRuntimeEnv;
   pOperator->cleanup      = destroySlimitOperatorInfo;
+
+  appendUpstream(pOperator, upstream);
   return pOperator;
 }
 
@@ -5833,7 +5844,7 @@ static SSDataBlock* hashDistinct(void* param, bool* newgroup) {
   pRes->info.rows = 0;
   SSDataBlock* pBlock = NULL;
   while(1) {
-    pBlock = pOperator->upstream->exec(pOperator->upstream, newgroup);
+    pBlock = pOperator->upstream[0]->exec(pOperator->upstream, newgroup);
     if (pBlock == NULL) {
       setQueryStatus(pOperator->pRuntimeEnv, QUERY_COMPLETED);
       pOperator->status = OP_EXEC_DONE;
@@ -5894,12 +5905,13 @@ SOperatorInfo* createDistinctOperatorInfo(SQueryRuntimeEnv* pRuntimeEnv, SOperat
   pOperator->blockingOptr = false;
   pOperator->status       = OP_IN_EXECUTING;
   pOperator->operatorType = OP_Distinct;
-  pOperator->upstream     = upstream;
   pOperator->numOfOutput  = numOfOutput;
   pOperator->info         = pInfo;
   pOperator->pRuntimeEnv  = pRuntimeEnv;
   pOperator->exec         = hashDistinct;
   pOperator->cleanup      = destroyDistinctOperatorInfo;
+
+  appendUpstream(pOperator, upstream);
   return pOperator;
 }