Merge pull request #3000 from taosdata/feature/query

Feature/query

src/client/inc/tscUtil.h

@@ -277,6 +277,9 @@ void tscAsyncQuerySingleRowForNextVnode(void *param, TAOS_RES *tres, int numOfRo
 void tscTryQueryNextClause(SSqlObj* pSql, __async_cb_func_t fp);
 int  tscSetMgmtEpSetFromCfg(const char *first, const char *second);
 
+bool tscSetSqlOwner(SSqlObj* pSql);
+void tscClearSqlOwner(SSqlObj* pSql);
+
 void* malloc_throw(size_t size);
 void* calloc_throw(size_t nmemb, size_t size);
 char* strdup_throw(const char* str);

src/client/inc/tsclient.h

@@ -80,7 +80,8 @@ typedef struct STableMetaInfo {
    * 2. keep the vgroup index for multi-vnode insertion
    */
   int32_t vgroupIndex;
-  char    name[TSDB_TABLE_ID_LEN];        // (super) table name
+  char    name[TSDB_TABLE_FNAME_LEN];        // (super) table name
+  char    aliasName[TSDB_TABLE_NAME_LEN];    // alias name of table specified in query sql
   SArray* tagColList;                        // SArray<SColumn*>, involved tag columns
 } STableMetaInfo;
 
@@ -106,7 +107,7 @@ typedef struct SColumnIndex {
 typedef struct SFieldSupInfo {
   bool            visible;
   SExprInfo      *pArithExprInfo;
-  SSqlExpr *      pSqlExpr;
+  SSqlExpr       *pSqlExpr;
 } SFieldSupInfo;
 
 typedef struct SFieldInfo {
@@ -128,7 +129,7 @@ typedef struct SCond {
 } SCond;
 
 typedef struct SJoinNode {
-  char     tableId[TSDB_TABLE_ID_LEN];
+  char     tableId[TSDB_TABLE_FNAME_LEN];
   uint64_t uid;
   int16_t  tagColId;
 } SJoinNode;
@@ -162,7 +163,7 @@ typedef struct SParamInfo {
 } SParamInfo;
 
 typedef struct STableDataBlocks {
-  char     tableId[TSDB_TABLE_ID_LEN];
+  char     tableId[TSDB_TABLE_FNAME_LEN];
   int8_t   tsSource;     // where does the UNIX timestamp come from, server or client
   bool     ordered;      // if current rows are ordered or not
   int64_t  vgId;         // virtual group id
@@ -255,6 +256,7 @@ typedef struct SResRec {
 
 typedef struct {
   int64_t               numOfRows;                  // num of results in current retrieved
+  int64_t               numOfRowsGroup;             // num of results of current group
   int64_t               numOfTotal;                 // num of total results
   int64_t               numOfClauseTotal;           // num of total result in current subclause
   char *                pRsp;
@@ -301,6 +303,7 @@ typedef struct STscObj {
 
 typedef struct SSqlObj {
   void            *signature;
+  pthread_t        owner;        // owner of sql object, by which it is executed
   STscObj         *pTscObj;
   void            *pRpcCtx;
   void            (*fp)();
@@ -419,7 +422,6 @@ char *tscGetErrorMsgPayload(SSqlCmd *pCmd);
 int32_t tscInvalidSQLErrMsg(char *msg, const char *additionalInfo, const char *sql);
 
 int32_t tscToSQLCmd(SSqlObj *pSql, struct SSqlInfo *pInfo);
-//void    tscGetResultColumnChr(SSqlRes *pRes, SFieldInfo* pFieldInfo, int32_t column);
 
 static FORCE_INLINE void tscGetResultColumnChr(SSqlRes* pRes, SFieldInfo* pFieldInfo, int32_t columnIndex) {
   SFieldSupInfo* pInfo = (SFieldSupInfo*) TARRAY_GET_ELEM(pFieldInfo->pSupportInfo, columnIndex);

src/client/src/tscAsync.c

@@ -220,14 +220,13 @@ void taos_fetch_rows_a(TAOS_RES *taosa, void (*fp)(void *, TAOS_RES *, int), voi
   if (pCmd->command == TSDB_SQL_TABLE_JOIN_RETRIEVE) {
     tscFetchDatablockFromSubquery(pSql);
   } else if (pRes->completed) {
-    if(pCmd->command == TSDB_SQL_FETCH) {
+    if(pCmd->command == TSDB_SQL_FETCH || (pCmd->command >= TSDB_SQL_SERV_STATUS && pCmd->command <= TSDB_SQL_CURRENT_USER)) {
       if (hasMoreVnodesToTry(pSql)) {  // sequentially retrieve data from remain vnodes.
         tscTryQueryNextVnode(pSql, tscAsyncQueryRowsForNextVnode);
-        return;
       } else {
         /*
-       * all available virtual node has been checked already, now we need to check
-       * for the next subclause queries
+         * all available virtual nodes in current clause has been checked already, now try the
+         * next one in the following union subclause
          */
         if (pCmd->clauseIndex < pCmd->numOfClause - 1) {
           tscTryQueryNextClause(pSql, tscAsyncQueryRowsForNextVnode);
@@ -240,6 +239,7 @@ void taos_fetch_rows_a(TAOS_RES *taosa, void (*fp)(void *, TAOS_RES *, int), voi
          */
         (*pSql->fetchFp)(param, pSql, 0);
       }
+
       return;
     } else if (pCmd->command == TSDB_SQL_RETRIEVE || pCmd->command == TSDB_SQL_RETRIEVE_LOCALMERGE) {
       // in case of show command, return no data

src/client/src/tscFunctionImpl.c

@@ -1942,11 +1942,12 @@ static void do_top_function_add(STopBotInfo *pInfo, int32_t maxLen, void *pData,
 
 static void do_bottom_function_add(STopBotInfo *pInfo, int32_t maxLen, void *pData, int64_t ts, uint16_t type,
                                    SExtTagsInfo *pTagInfo, char *pTags, int16_t stage) {
-  tValuePair **pList = pInfo->res;
-  
   tVariant val = {0};
   tVariantCreateFromBinary(&val, pData, tDataTypeDesc[type].nSize, type);
 
+  tValuePair **pList = pInfo->res;
+  assert(pList != NULL);
+
   if (pInfo->num < maxLen) {
     if (pInfo->num == 0) {
       valuePairAssign(pList[pInfo->num], type, (const char*) &val.i64Key, ts, pTags, pTagInfo, stage);

src/client/src/tscLocal.c

@@ -293,7 +293,7 @@ static void tscProcessCurrentDB(SSqlObj *pSql) {
   char db[TSDB_DB_NAME_LEN] = {0};
   extractDBName(pSql->pTscObj->db, db);
   
-  SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0);
+  SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, pSql->cmd.clauseIndex);
   
   SSqlExpr* pExpr = taosArrayGetP(pQueryInfo->exprList, 0);
   pExpr->resType = TSDB_DATA_TYPE_BINARY;
@@ -314,7 +314,7 @@ static void tscProcessCurrentDB(SSqlObj *pSql) {
 
 static void tscProcessServerVer(SSqlObj *pSql) {
   const char* v = pSql->pTscObj->sversion;
-  SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0);
+  SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, pSql->cmd.clauseIndex);
   
   SSqlExpr* pExpr = taosArrayGetP(pQueryInfo->exprList, 0);
   pExpr->resType = TSDB_DATA_TYPE_BINARY;

src/client/src/tscLocalMerge.c

@@ -68,7 +68,7 @@ static void tscInitSqlContext(SSqlCmd *pCmd, SLocalReducer *pReducer, tOrderDesc
     SSqlExpr *      pExpr = tscSqlExprGet(pQueryInfo, i);
 
     pCtx->aOutputBuf =
-        pReducer->pResultBuf->data + tscFieldInfoGetOffset(pQueryInfo, i) * pReducer->resColModel->capacity;
+        pReducer->pResultBuf->data + pExpr->offset * pReducer->resColModel->capacity;
     pCtx->order = pQueryInfo->order.order;
     pCtx->functionId = pExpr->functionId;
 
@@ -321,6 +321,7 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
   pReducer->finalRowSize = tscGetResRowLength(pQueryInfo->exprList);
   pReducer->resColModel = finalmodel;
   pReducer->resColModel->capacity = pReducer->nResultBufSize;
+
   assert(pReducer->finalRowSize > 0);
   if (pReducer->finalRowSize > 0) {
     pReducer->resColModel->capacity /= pReducer->finalRowSize;
@@ -328,7 +329,6 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
   assert(pReducer->finalRowSize <= pReducer->rowSize);
 
   pReducer->pFinalRes = calloc(1, pReducer->rowSize * pReducer->resColModel->capacity);
-//  pReducer->pBufForInterpo = calloc(1, pReducer->nResultBufSize);
 
   if (pReducer->pTempBuffer == NULL || pReducer->discardData == NULL || pReducer->pResultBuf == NULL ||
       /*pReducer->pBufForInterpo == NULL || */pReducer->pFinalRes == NULL || pReducer->prevRowOfInput == NULL) {
@@ -856,24 +856,6 @@ void savePrevRecordAndSetupInterpoInfo(SLocalReducer *pLocalReducer, SQueryInfo
   tColModelAppend(pModel, pLocalReducer->discardData, pLocalReducer->prevRowOfInput, 0, 1, 1);
 }
 
-// todo merge with following function
-// static void reversedCopyResultToDstBuf(SQueryInfo* pQueryInfo, SSqlRes *pRes, tFilePage *pFinalDataPage) {
-//
-//  for (int32_t i = 0; i < pQueryInfo->exprList.numOfExprs; ++i) {
-//    TAOS_FIELD *pField = tscFieldInfoGetField(&pQueryInfo->fieldsInfo, i);
-//
-//    int32_t offset = tscFieldInfoGetOffset(pQueryInfo, i);
-//    char *  src = pFinalDataPage->data + (pRes->numOfRows - 1) * pField->bytes + pRes->numOfRows * offset;
-//    char *  dst = pRes->data + pRes->numOfRows * offset;
-//
-//    for (int32_t j = 0; j < pRes->numOfRows; ++j) {
-//      memcpy(dst, src, (size_t)pField->bytes);
-//      dst += pField->bytes;
-//      src -= pField->bytes;
-//    }
-//  }
-//}
-
 static void reversedCopyFromInterpolationToDstBuf(SQueryInfo *pQueryInfo, SSqlRes *pRes, tFilePage **pResPages,
                                                   SLocalReducer *pLocalReducer) {
   assert(0);
@@ -907,20 +889,10 @@ static void doFillResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool doneO
   tFilePage * pFinalDataPage = pLocalReducer->pResultBuf;
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
 
-//  if (pRes->pLocalReducer != pLocalReducer) {
-//    /*
-//     * Release the SSqlObj is called, and it is int destroying function invoked by other thread.
-//     * However, the other thread will WAIT until current process fully completes.
-//     * Since the flag of release struct is set by doLocalReduce function
-//     */
-//    assert(pRes->pLocalReducer == NULL);
-//  }
-
   // no interval query, no fill operation
   if (pQueryInfo->intervalTime == 0 || pQueryInfo->fillType == TSDB_FILL_NONE) {
     pRes->data = pLocalReducer->pFinalRes;
     pRes->numOfRows = pFinalDataPage->num;
-    pRes->numOfClauseTotal += pRes->numOfRows;
 
     if (pQueryInfo->limit.offset > 0) {
       if (pQueryInfo->limit.offset < pRes->numOfRows) {
@@ -931,22 +903,22 @@ static void doFillResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool doneO
         tColModelCompact(pLocalReducer->resColModel, pFinalDataPage, prevSize);
 
         pRes->numOfRows -= pQueryInfo->limit.offset;
-        pRes->numOfClauseTotal -= pQueryInfo->limit.offset;
         pQueryInfo->limit.offset = 0;
       } else {
         pQueryInfo->limit.offset -= pRes->numOfRows;
         pRes->numOfRows = 0;
-        pRes->numOfClauseTotal = 0;
       }
     }
 
-    if (pQueryInfo->limit.limit >= 0 && pRes->numOfClauseTotal > pQueryInfo->limit.limit) {
+    pRes->numOfRowsGroup += pRes->numOfRows;
+
+    if (pQueryInfo->limit.limit >= 0 && pRes->numOfRowsGroup > pQueryInfo->limit.limit) {
       /* impose the limitation of output rows on the final result */
-      int32_t prevSize = (int32_t)pFinalDataPage->num;
-      int32_t overflow = (int32_t)(pRes->numOfClauseTotal - pQueryInfo->limit.limit);
+      int32_t prevSize = pFinalDataPage->num;
+      int32_t overflow = pRes->numOfRowsGroup - pQueryInfo->limit.limit;
       assert(overflow < pRes->numOfRows);
 
-      pRes->numOfClauseTotal = pQueryInfo->limit.limit;
+      pRes->numOfRowsGroup = pQueryInfo->limit.limit;
       pRes->numOfRows -= overflow;
       pFinalDataPage->num -= overflow;
 
@@ -957,6 +929,8 @@ static void doFillResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool doneO
     }
 
     memcpy(pRes->data, pFinalDataPage->data, pRes->numOfRows * pLocalReducer->finalRowSize);
+
+    pRes->numOfClauseTotal += pRes->numOfRows;
     pFinalDataPage->num = 0;
     return;
   }
@@ -986,7 +960,6 @@ static void doFillResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool doneO
 
       pRes->data = pLocalReducer->pFinalRes;
       pRes->numOfRows = newRows;
-      pRes->numOfClauseTotal += newRows;
 
       pQueryInfo->limit.offset = 0;
       break;
@@ -1010,15 +983,13 @@ static void doFillResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool doneO
   }
 
   if (pRes->numOfRows > 0) {
-    if (pQueryInfo->limit.limit >= 0 && pRes->numOfClauseTotal > pQueryInfo->limit.limit) {
-      int32_t overflow = (int32_t)(pRes->numOfClauseTotal - pQueryInfo->limit.limit);
+    if (pQueryInfo->limit.limit >= 0 && pRes->numOfRows > pQueryInfo->limit.limit) {
+      int32_t overflow = pRes->numOfRows - pQueryInfo->limit.limit;
       pRes->numOfRows -= overflow;
-
-      assert(pRes->numOfRows >= 0);
-
-      pRes->numOfClauseTotal = pQueryInfo->limit.limit;
       pFinalDataPage->num -= overflow;
 
+      assert(pRes->numOfRows >= 0 && pFinalDataPage->num > 0);
+
       /* set remain data to be discarded, and reset the interpolation information */
       savePrevRecordAndSetupInterpoInfo(pLocalReducer, pQueryInfo, pFillInfo);
     }
@@ -1032,6 +1003,9 @@ static void doFillResult(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool doneO
     } else {  // todo bug??
       reversedCopyFromInterpolationToDstBuf(pQueryInfo, pRes, pResPages, pLocalReducer);
     }
+
+    pRes->numOfRowsGroup += pRes->numOfRows;
+    pRes->numOfClauseTotal += pRes->numOfRows;
   }
 
   pFinalDataPage->num = 0;
@@ -1227,7 +1201,10 @@ static bool saveGroupResultInfo(SSqlObj *pSql) {
   SSqlRes *pRes = &pSql->res;
 
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
+
+  if (pRes->numOfRowsGroup > 0) {
     pRes->numOfGroups += 1;
+  }
 
   // the output group is limited by the slimit clause
   if (reachGroupResultLimit(pQueryInfo, pRes)) {
@@ -1267,6 +1244,11 @@ bool doGenerateFinalResults(SSqlObj *pSql, SLocalReducer *pLocalReducer, bool no
     pQueryInfo->slimit.offset -= 1;
     pLocalReducer->discard = !noMoreCurrentGroupRes;
 
+    if (pLocalReducer->discard) {
+      SColumnModel *pInternModel = pLocalReducer->pDesc->pColumnModel;
+      tColModelAppend(pInternModel, pLocalReducer->discardData, pLocalReducer->pTempBuffer->data, 0, 1, 1);
+    }
+
     return false;
   }
 
@@ -1299,7 +1281,7 @@ void resetOutputBuf(SQueryInfo *pQueryInfo, SLocalReducer *pLocalReducer) {  //
 static void resetEnvForNewResultset(SSqlRes *pRes, SSqlCmd *pCmd, SLocalReducer *pLocalReducer) {
   // In handling data in other groups, we need to reset the interpolation information for a new group data
   pRes->numOfRows = 0;
-  pRes->numOfClauseTotal = 0;
+  pRes->numOfRowsGroup = 0;
 
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
 
@@ -1363,7 +1345,8 @@ static bool doHandleLastRemainData(SSqlObj *pSql) {
   if ((isAllSourcesCompleted(pLocalReducer) && !pLocalReducer->hasPrevRow) || pLocalReducer->pLocalDataSrc[0] == NULL ||
       prevGroupCompleted) {
     // if fillType == TSDB_FILL_NONE, return directly
-    if (pQueryInfo->fillType != TSDB_FILL_NONE) {
+    if (pQueryInfo->fillType != TSDB_FILL_NONE &&
+      ((pRes->numOfRowsGroup < pQueryInfo->limit.limit && pQueryInfo->limit.limit > 0) || (pQueryInfo->limit.limit < 0))) {
       int64_t etime = (pQueryInfo->window.skey < pQueryInfo->window.ekey) ? pQueryInfo->window.ekey : pQueryInfo->window.skey;
 
       assert(pFillInfo->numOfRows == 0);

src/client/src/tscParseInsert.c

@@ -989,7 +989,7 @@ static int32_t tscCheckIfCreateTable(char **sqlstr, SSqlObj *pSql) {
 }
 
 int validateTableName(char *tblName, int len, SSQLToken* psTblToken) {
-  tstrncpy(psTblToken->z, tblName, TSDB_TABLE_ID_LEN);
+  tstrncpy(psTblToken->z, tblName, TSDB_TABLE_FNAME_LEN);
 
   psTblToken->n    = len;
   psTblToken->type = TK_ID;
@@ -1038,7 +1038,7 @@ int tsParseInsertSql(SSqlObj *pSql) {
   }
 
   if (NULL == pCmd->pTableList) {
-    pCmd->pTableList = taosHashInit(128, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false);
+    pCmd->pTableList = taosHashInit(128, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), true, false);
     pCmd->pDataBlocks = taosArrayInit(4, POINTER_BYTES);
     if (NULL == pCmd->pTableList || NULL == pSql->cmd.pDataBlocks) {
       code = TSDB_CODE_TSC_OUT_OF_MEMORY;
@@ -1077,7 +1077,7 @@ int tsParseInsertSql(SSqlObj *pSql) {
     }
 
     pCmd->curSql = sToken.z;
-    char buf[TSDB_TABLE_ID_LEN];
+    char buf[TSDB_TABLE_FNAME_LEN];
     SSQLToken sTblToken;
     sTblToken.z = buf;
     // Check if the table name available or not

src/client/src/tscSQLParser.c

@@ -86,7 +86,7 @@ static int32_t parseOrderbyClause(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, SQueryS
 static int32_t tsRewriteFieldNameIfNecessary(SSqlCmd* pCmd, SQueryInfo* pQueryInfo);
 static int32_t setAlterTableInfo(SSqlObj* pSql, struct SSqlInfo* pInfo);
 static int32_t validateSqlFunctionInStreamSql(SSqlCmd* pCmd, SQueryInfo* pQueryInfo);
-static int32_t buildArithmeticExprString(tSQLExpr* pExpr, char** exprString);
+static int32_t arithmeticExprToString(tSQLExpr* pExpr, char** exprString);
 static int32_t validateFunctionsInIntervalOrGroupbyQuery(SSqlCmd* pCmd, SQueryInfo* pQueryInfo);
 static int32_t validateArithmeticSQLExpr(SSqlCmd* pCmd, tSQLExpr* pExpr, SQueryInfo* pQueryInfo, SColumnList* pList, int32_t* type);
 static int32_t validateEp(char* ep);
@@ -1087,11 +1087,11 @@ int32_t setObjFullName(char* fullName, const char* account, SSQLToken* pDB, SSQL
     *xlen = totalLen;
   }
 
-  if (totalLen < TSDB_TABLE_ID_LEN) {
+  if (totalLen < TSDB_TABLE_FNAME_LEN) {
     fullName[totalLen] = 0;
   }
 
-  return (totalLen < TSDB_TABLE_ID_LEN) ? TSDB_CODE_SUCCESS : TSDB_CODE_TSC_INVALID_SQL;
+  return (totalLen < TSDB_TABLE_FNAME_LEN) ? TSDB_CODE_SUCCESS : TSDB_CODE_TSC_INVALID_SQL;
 }
 
 static void extractColumnNameFromString(tSQLExprItem* pItem) {
@@ -1106,48 +1106,15 @@ static void extractColumnNameFromString(tSQLExprItem* pItem) {
   }
 }
 
-int32_t parseSelectClause(SSqlCmd* pCmd, int32_t clauseIndex, tSQLExprList* pSelection, bool isSTable) {
-  assert(pSelection != NULL && pCmd != NULL);
-
+static int32_t handleArithmeticExpr(SSqlCmd* pCmd, int32_t clauseIndex, int32_t exprIndex, tSQLExprItem* pItem) {
   const char* msg1 = "invalid column name, or illegal column type";
-  const char* msg2 = "functions can not be mixed up";
-  const char* msg3 = "not support query expression";
+  const char* msg2 = "invalid arithmetic expression in select clause";
+  const char* msg3 = "tag columns can not be used in arithmetic expression";
   const char* msg4 = "columns from different table mixed up in arithmetic expression";
-  const char* msg5 = "invalid function name";
 
+  // arithmetic function in select clause
   SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, clauseIndex);
 
-  if (pQueryInfo->colList == NULL) {
-    pQueryInfo->colList = taosArrayInit(4, POINTER_BYTES);
-  }
-  
-  for (int32_t i = 0; i < pSelection->nExpr; ++i) {
-    int32_t outputIndex = (int32_t)tscSqlExprNumOfExprs(pQueryInfo);
-    tSQLExprItem* pItem = &pSelection->a[i];
-
-    // project on all fields
-    if (pItem->pNode->nSQLOptr == TK_ALL || pItem->pNode->nSQLOptr == TK_ID || pItem->pNode->nSQLOptr == TK_STRING) {
-      // it is actually a function, but the function name is invalid
-      if (pItem->pNode->nSQLOptr == TK_ID && (pItem->pNode->colInfo.z == NULL && pItem->pNode->colInfo.n == 0)) {
-        return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg5);
-      }
-
-      // if the name of column is quoted, remove it and set the right information for later process
-      extractColumnNameFromString(pItem);
-      TSDB_QUERY_SET_TYPE(pQueryInfo->type, TSDB_QUERY_TYPE_PROJECTION_QUERY);
-
-      // select table_name1.field_name1, table_name2.field_name2  from table_name1, table_name2
-      if (addProjectionExprAndResultField(pCmd, pQueryInfo, pItem) != TSDB_CODE_SUCCESS) {
-        return TSDB_CODE_TSC_INVALID_SQL;
-      }
-    } else if (pItem->pNode->nSQLOptr >= TK_COUNT && pItem->pNode->nSQLOptr <= TK_TBID) {
-      // sql function in selection clause, append sql function info in pSqlCmd structure sequentially
-      if (addExprAndResultField(pCmd, pQueryInfo, outputIndex, pItem, true) != TSDB_CODE_SUCCESS) {
-        return TSDB_CODE_TSC_INVALID_SQL;
-      }
-
-    } else if (pItem->pNode->nSQLOptr >= TK_PLUS && pItem->pNode->nSQLOptr <= TK_REM) {
-      // arithmetic function in select clause
   SColumnList columnList = {0};
   int32_t     arithmeticType = NON_ARITHMEIC_EXPR;
 
@@ -1156,7 +1123,9 @@ int32_t parseSelectClause(SSqlCmd* pCmd, int32_t clauseIndex, tSQLExprList* pSel
   }
 
   int32_t tableIndex = columnList.ids[0].tableIndex;
-      char  arithmeticExprStr[1024] = {0};
+
+  // todo potential data overflow
+  char  arithmeticExprStr[1024*12];
   char* p = arithmeticExprStr;
 
   if (arithmeticType == NORMAL_ARITHMETIC) {
@@ -1169,16 +1138,16 @@ int32_t parseSelectClause(SSqlCmd* pCmd, int32_t clauseIndex, tSQLExprList* pSel
       }
     }
 
-        if (buildArithmeticExprString(pItem->pNode, &p) != TSDB_CODE_SUCCESS) {
+    if (arithmeticExprToString(pItem->pNode, &p) != TSDB_CODE_SUCCESS) {
       return TSDB_CODE_TSC_INVALID_SQL;
     }
 
     // expr string is set as the parameter of function
     SColumnIndex index = {.tableIndex = tableIndex};
-        SSqlExpr*    pExpr = tscSqlExprAppend(pQueryInfo, TSDB_FUNC_ARITHM, &index, TSDB_DATA_TYPE_DOUBLE,
-                                              sizeof(double), sizeof(double), false);
 
-        /* todo alias name should use the original sql string */
+    SSqlExpr* pExpr = tscSqlExprAppend(pQueryInfo, TSDB_FUNC_ARITHM, &index, TSDB_DATA_TYPE_DOUBLE, sizeof(double),
+                                       sizeof(double), false);
+
     char* name = (pItem->aliasName != NULL)? pItem->aliasName:arithmeticExprStr;
     tstrncpy(pExpr->aliasName, name, sizeof(pExpr->aliasName));
 
@@ -1188,7 +1157,15 @@ int32_t parseSelectClause(SSqlCmd* pCmd, int32_t clauseIndex, tSQLExprList* pSel
     int32_t ret = exprTreeFromSqlExpr(pCmd, &pNode, pItem->pNode, pQueryInfo->exprList, pQueryInfo, colList);
     if (ret != TSDB_CODE_SUCCESS) {
       tExprTreeDestroy(&pNode, NULL);
-          return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), "invalid arithmetic expression in select clause");
+      return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg2);
+    }
+
+    size_t numOfNode = taosArrayGetSize(colList);
+    for(int32_t k = 0; k < numOfNode; ++k) {
+      SColIndex* pIndex = taosArrayGet(colList, k);
+      if (pIndex->flag == 1) {
+        return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg3);
+      }
     }
 
     SBufferWriter bw = tbufInitWriter(NULL, false);
@@ -1205,17 +1182,22 @@ int32_t parseSelectClause(SSqlCmd* pCmd, int32_t clauseIndex, tSQLExprList* pSel
     char* c = tbufGetData(&bw, true);
 
     // set the serialized binary string as the parameter of arithmetic expression
-        addExprParams(pExpr, c, TSDB_DATA_TYPE_BINARY, (int32_t)len, index.tableIndex);
+    addExprParams(pExpr, c, TSDB_DATA_TYPE_BINARY, len, index.tableIndex);
 
-        insertResultField(pQueryInfo, i, &columnList, sizeof(double), TSDB_DATA_TYPE_DOUBLE, pExpr->aliasName, pExpr);
+    insertResultField(pQueryInfo, exprIndex, &columnList, sizeof(double), TSDB_DATA_TYPE_DOUBLE, pExpr->aliasName, pExpr);
 
     taosArrayDestroy(colList);
     tExprTreeDestroy(&pNode, NULL);
   } else {
+    if (arithmeticExprToString(pItem->pNode, &p) != TSDB_CODE_SUCCESS) {
+      return TSDB_CODE_TSC_INVALID_SQL;
+    }
+
     columnList.num = 0;
     columnList.ids[0] = (SColumnIndex) {0, 0};
 
-        insertResultField(pQueryInfo, i, &columnList, sizeof(double), TSDB_DATA_TYPE_DOUBLE, "dummy_column", NULL);
+    char* name = (pItem->aliasName != NULL)? pItem->aliasName:arithmeticExprStr;
+    insertResultField(pQueryInfo, exprIndex, &columnList, sizeof(double), TSDB_DATA_TYPE_DOUBLE, name, NULL);
 
     int32_t slot = tscNumOfFields(pQueryInfo) - 1;
     SFieldSupInfo* pInfo = tscFieldInfoGetSupp(&pQueryInfo->fieldsInfo, slot);
@@ -1237,6 +1219,54 @@ int32_t parseSelectClause(SSqlCmd* pCmd, int32_t clauseIndex, tSQLExprList* pSel
       pInfo->pArithExprInfo = pArithExprInfo;
     }
   }
+
+  return TSDB_CODE_SUCCESS;
+}
+
+int32_t parseSelectClause(SSqlCmd* pCmd, int32_t clauseIndex, tSQLExprList* pSelection, bool isSTable) {
+  assert(pSelection != NULL && pCmd != NULL);
+
+  const char* msg2 = "functions can not be mixed up";
+  const char* msg3 = "not support query expression";
+  const char* msg5 = "invalid function name";
+
+  SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, clauseIndex);
+
+  if (pQueryInfo->colList == NULL) {
+    pQueryInfo->colList = taosArrayInit(4, POINTER_BYTES);
+  }
+
+  for (int32_t i = 0; i < pSelection->nExpr; ++i) {
+    int32_t outputIndex = tscSqlExprNumOfExprs(pQueryInfo);
+    tSQLExprItem* pItem = &pSelection->a[i];
+
+    // project on all fields
+    if (pItem->pNode->nSQLOptr == TK_ALL || pItem->pNode->nSQLOptr == TK_ID || pItem->pNode->nSQLOptr == TK_STRING) {
+      // it is actually a function, but the function name is invalid
+      if (pItem->pNode->nSQLOptr == TK_ID && (pItem->pNode->colInfo.z == NULL && pItem->pNode->colInfo.n == 0)) {
+        return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg5);
+      }
+
+      // if the name of column is quoted, remove it and set the right information for later process
+      extractColumnNameFromString(pItem);
+      TSDB_QUERY_SET_TYPE(pQueryInfo->type, TSDB_QUERY_TYPE_PROJECTION_QUERY);
+
+      // select table_name1.field_name1, table_name2.field_name2  from table_name1, table_name2
+      if (addProjectionExprAndResultField(pCmd, pQueryInfo, pItem) != TSDB_CODE_SUCCESS) {
+        return TSDB_CODE_TSC_INVALID_SQL;
+      }
+    } else if (pItem->pNode->nSQLOptr >= TK_COUNT && pItem->pNode->nSQLOptr <= TK_TBID) {
+      // sql function in selection clause, append sql function info in pSqlCmd structure sequentially
+      if (addExprAndResultField(pCmd, pQueryInfo, outputIndex, pItem, true) != TSDB_CODE_SUCCESS) {
+        return TSDB_CODE_TSC_INVALID_SQL;
+      }
+
+    } else if (pItem->pNode->nSQLOptr >= TK_PLUS && pItem->pNode->nSQLOptr <= TK_REM) {
+      int32_t code = handleArithmeticExpr(pCmd, clauseIndex, i, pItem);
+      if (code != TSDB_CODE_SUCCESS) {
+        return code;
+      }
+
     } else {
       /*
        * not support such expression
@@ -2106,13 +2136,10 @@ int32_t getTableIndexImpl(SSQLToken* pTableToken, SQueryInfo* pQueryInfo, SColum
   }
 
   pIndex->tableIndex = COLUMN_INDEX_INITIAL_VAL;
-  char tableName[TSDB_TABLE_ID_LEN] = {0};
-
   for (int32_t i = 0; i < pQueryInfo->numOfTables; ++i) {
     STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, i);
-    extractTableName(pTableMetaInfo->name, tableName);
-
-    if (strncasecmp(tableName, pTableToken->z, pTableToken->n) == 0 && strlen(tableName) == pTableToken->n) {
+    char* name = pTableMetaInfo->aliasName;
+    if (strncasecmp(name, pTableToken->z, pTableToken->n) == 0 && strlen(name) == pTableToken->n) {
       pIndex->tableIndex = i;
       break;
     }
@@ -3081,14 +3108,14 @@ static int32_t getJoinCondInfo(SSqlCmd* pCmd, SQueryInfo* pQueryInfo, tSQLExpr*
 }
 
 // todo error handle / such as and /or mixed with +/-/*/
-int32_t buildArithmeticExprString(tSQLExpr* pExpr, char** exprString) {
+int32_t doArithmeticExprToString(tSQLExpr* pExpr, char** exprString) {
   tSQLExpr* pLeft = pExpr->pLeft;
   tSQLExpr* pRight = pExpr->pRight;
 
   *(*exprString)++ = '(';
 
   if (pLeft->nSQLOptr >= TK_PLUS && pLeft->nSQLOptr <= TK_REM) {
-    buildArithmeticExprString(pLeft, exprString);
+    doArithmeticExprToString(pLeft, exprString);
   } else {
     int32_t ret = tSQLExprNodeToString(pLeft, exprString);
     if (ret != TSDB_CODE_SUCCESS) {
@@ -3099,7 +3126,7 @@ int32_t buildArithmeticExprString(tSQLExpr* pExpr, char** exprString) {
   optrToString(pExpr, exprString);
 
   if (pRight->nSQLOptr >= TK_PLUS && pRight->nSQLOptr <= TK_REM) {
-    buildArithmeticExprString(pRight, exprString);
+    doArithmeticExprToString(pRight, exprString);
   } else {
     int32_t ret = tSQLExprNodeToString(pRight, exprString);
     if (ret != TSDB_CODE_SUCCESS) {
@@ -3112,6 +3139,19 @@ int32_t buildArithmeticExprString(tSQLExpr* pExpr, char** exprString) {
   return TSDB_CODE_SUCCESS;
 }
 
+static int32_t arithmeticExprToString(tSQLExpr* pExpr, char** str) {
+  char* start = *str;
+
+  int32_t code = doArithmeticExprToString(pExpr, str);
+  if (code == TSDB_CODE_SUCCESS) { // remove out the parenthesis
+    int32_t len = strlen(start);
+    memmove(start, start + 1, len - 2);
+    start[len - 2] = 0;
+  }
+
+  return code;
+}
+
 static int32_t validateSQLExpr(SSqlCmd* pCmd, tSQLExpr* pExpr, SQueryInfo* pQueryInfo, SColumnList* pList, int32_t* type) {
   if (pExpr->nSQLOptr == TK_ID) {
     if (*type == NON_ARITHMEIC_EXPR) {
@@ -3615,7 +3655,7 @@ static int32_t setTableCondForSTableQuery(SSqlCmd* pCmd, SQueryInfo* pQueryInfo,
   SStringBuilder sb1; memset(&sb1, 0, sizeof(sb1));
   taosStringBuilderAppendStringLen(&sb1, QUERY_COND_REL_PREFIX_IN, QUERY_COND_REL_PREFIX_IN_LEN);
 
-  char db[TSDB_TABLE_ID_LEN] = {0};
+  char db[TSDB_TABLE_FNAME_LEN] = {0};
 
   // remove the duplicated input table names
   int32_t num = 0;
@@ -3640,7 +3680,7 @@ static int32_t setTableCondForSTableQuery(SSqlCmd* pCmd, SQueryInfo* pQueryInfo,
       taosStringBuilderAppendStringLen(&sb1, TBNAME_LIST_SEP, 1);
     }
 
-    char      idBuf[TSDB_TABLE_ID_LEN] = {0};
+    char      idBuf[TSDB_TABLE_FNAME_LEN] = {0};
     int32_t   xlen = (int32_t)strlen(segments[i]);
     SSQLToken t = {.z = segments[i], .n = xlen, .type = TK_STRING};
 
@@ -4345,6 +4385,7 @@ int32_t setAlterTableInfo(SSqlObj* pSql, struct SSqlInfo* pInfo) {
   const char* msg16 = "only support one column";
   const char* msg17 = "invalid column name";
   const char* msg18 = "primary timestamp column cannot be dropped";
+  const char* msg19 = "invalid new tag name";
 
   SSqlCmd*        pCmd = &pSql->cmd;
   SAlterTableSQL* pAlterSQL = pInfo->pAlterInfo;
@@ -4446,12 +4487,12 @@ int32_t setAlterTableInfo(SSqlObj* pSql, struct SSqlInfo* pInfo) {
 
     SSQLToken srcToken = {.z = pSrcItem->pVar.pz, .n = pSrcItem->pVar.nLen, .type = TK_STRING};
     if (getColumnIndexByName(pCmd, &srcToken, pQueryInfo, &srcIndex) != TSDB_CODE_SUCCESS) {
-      return TSDB_CODE_TSC_INVALID_SQL;
+      return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg17);
     }
 
     SSQLToken destToken = {.z = pDstItem->pVar.pz, .n = pDstItem->pVar.nLen, .type = TK_STRING};
     if (getColumnIndexByName(pCmd, &destToken, pQueryInfo, &destIndex) == TSDB_CODE_SUCCESS) {
-      return TSDB_CODE_TSC_INVALID_SQL;
+      return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg19);
     }
 
     char name[TSDB_COL_NAME_LEN] = {0};
@@ -5880,7 +5921,8 @@ int32_t doCheckForQuery(SSqlObj* pSql, SQuerySQL* pQuerySql, int32_t index) {
   const char* msg7  = "illegal number of tables in from clause";
   const char* msg8  = "too many columns in selection clause";
   const char* msg9  = "TWA query requires both the start and end time";
-  const char* msg10= "too many tables in from clause";
+  const char* msg10 = "too many tables in from clause";
+  const char* msg11 = "invalid table alias name";
 
   int32_t code = TSDB_CODE_SUCCESS;
 
@@ -5912,18 +5954,18 @@ int32_t doCheckForQuery(SSqlObj* pSql, SQuerySQL* pQuerySql, int32_t index) {
     return doLocalQueryProcess(pCmd, pQueryInfo, pQuerySql);
   }
 
-  if (pQuerySql->from->nExpr > TSDB_MAX_JOIN_TABLE_NUM) {
+  if (pQuerySql->from->nExpr > TSDB_MAX_JOIN_TABLE_NUM * 2) {
     return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg7);
   }
 
   pQueryInfo->command = TSDB_SQL_SELECT;
 
-  if (pQuerySql->from->nExpr > 2) {
+  if (pQuerySql->from->nExpr > 4) {
     return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg10);
   }
 
   // set all query tables, which are maybe more than one.
-  for (int32_t i = 0; i < pQuerySql->from->nExpr; ++i) {
+  for (int32_t i = 0; i < pQuerySql->from->nExpr; ) {
     tVariant* pTableItem = &pQuerySql->from->a[i].pVar;
 
     if (pTableItem->nType != TSDB_DATA_TYPE_BINARY) {
@@ -5937,24 +5979,34 @@ int32_t doCheckForQuery(SSqlObj* pSql, SQuerySQL* pQuerySql, int32_t index) {
       return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg0);
     }
 
-    if (pQueryInfo->numOfTables <= i) {  // more than one table
+    if (pQueryInfo->numOfTables <= i/2) {  // more than one table
       tscAddEmptyMetaInfo(pQueryInfo);
     }
 
-    STableMetaInfo* pMeterInfo1 = tscGetMetaInfo(pQueryInfo, i);
+    STableMetaInfo* pTableMetaInfo1 = tscGetMetaInfo(pQueryInfo, i/2);
 
     SSQLToken t = {.type = TSDB_DATA_TYPE_BINARY, .n = pTableItem->nLen, .z = pTableItem->pz};
-    if (tscSetTableFullName(pMeterInfo1, &t, pSql) != TSDB_CODE_SUCCESS) {
+    if (tscSetTableFullName(pTableMetaInfo1, &t, pSql) != TSDB_CODE_SUCCESS) {
       return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg1);
     }
 
-    code = tscGetTableMeta(pSql, pMeterInfo1);
+    tVariant* pTableItem1 = &pQuerySql->from->a[i + 1].pVar;
+    SSQLToken aliasName = {.z = pTableItem1->pz, .n = pTableItem1->nLen, .type = TK_STRING};
+    if (tscValidateName(&aliasName) != TSDB_CODE_SUCCESS) {
+      return invalidSqlErrMsg(tscGetErrorMsgPayload(pCmd), msg11);
+    }
+
+    tstrncpy(pTableMetaInfo1->aliasName, pTableItem1->pz, sizeof(pTableMetaInfo1->aliasName));
+
+    code = tscGetTableMeta(pSql, pTableMetaInfo1);
     if (code != TSDB_CODE_SUCCESS) {
       return code;
     }
+
+    i += 2;
   }
 
-  assert(pQueryInfo->numOfTables == pQuerySql->from->nExpr);
+  assert(pQueryInfo->numOfTables == pQuerySql->from->nExpr / 2);
   bool isSTable = false;
   
   if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
@@ -6123,11 +6175,13 @@ int32_t exprTreeFromSqlExpr(SSqlCmd* pCmd, tExprNode **pExpr, const tSQLExpr* pS
         return ret;
       }
 
+      STableMeta* pTableMeta = tscGetMetaInfo(pQueryInfo, 0)->pTableMeta;
+      int32_t numOfColumns = tscGetNumOfColumns(pTableMeta);
+
       *pExpr = calloc(1, sizeof(tExprNode));
       (*pExpr)->nodeType = TSQL_NODE_COL;
       (*pExpr)->pSchema = calloc(1, sizeof(SSchema));
 
-      STableMeta* pTableMeta = tscGetMetaInfo(pQueryInfo, 0)->pTableMeta;
       SSchema* pSchema = tscGetTableColumnSchema(pTableMeta, index.columnIndex);
       *(*pExpr)->pSchema = *pSchema;
   
@@ -6136,6 +6190,7 @@ int32_t exprTreeFromSqlExpr(SSqlCmd* pCmd, tExprNode **pExpr, const tSQLExpr* pS
         tstrncpy(colIndex.name, pSchema->name, sizeof(colIndex.name));
         colIndex.colId = pSchema->colId;
         colIndex.colIndex = index.columnIndex;
+        colIndex.flag = (index.columnIndex >= numOfColumns)? 1:0;
 
         taosArrayPush(pCols, &colIndex);
       }

src/client/src/tscServer.c

@@ -45,19 +45,27 @@ void tscSaveSubscriptionProgress(void* sub);
 static int32_t minMsgSize() { return tsRpcHeadSize + 100; }
 
 static void tscSetDnodeEpSet(SSqlObj* pSql, SCMVgroupInfo* pVgroupInfo) {
+  assert(pSql != NULL && pVgroupInfo != NULL && pVgroupInfo->numOfEps > 0);
+
   SRpcEpSet* pEpSet = &pSql->epSet;
   pEpSet->inUse = 0;
-  if (pVgroupInfo == NULL) {
-    pEpSet->numOfEps = 0;
-    return;
-  }
+
+  // apply the FQDN string length check here
+  bool hasFqdn = false;
 
   pEpSet->numOfEps = pVgroupInfo->numOfEps;
   for(int32_t i = 0; i < pVgroupInfo->numOfEps; ++i) {
     strcpy(pEpSet->fqdn[i], pVgroupInfo->epAddr[i].fqdn);
     pEpSet->port[i] = pVgroupInfo->epAddr[i].port;
+
+    if (!hasFqdn) {
+      hasFqdn = (strlen(pEpSet->fqdn[i]) > 0);
     }
+  }
+
+  assert(hasFqdn);
 }
+
 static void tscDumpMgmtEpSet(SRpcEpSet *epSet) {
   taosCorBeginRead(&tscMgmtEpSet.version);
   *epSet = tscMgmtEpSet.epSet;
@@ -127,21 +135,6 @@ void tscPrintMgmtEp() {
   }
 }
 
-/*
- * For each management node, try twice at least in case of poor network situation.
- * If the client start to connect to a non-management node from the client, and the first retry may fail due to
- * the poor network quality. And then, the second retry get the response with redirection command.
- * The retry will not be executed since only *two* retry is allowed in case of single management node in the cluster.
- * Therefore, we need to multiply the retry times by factor of 2 to fix this problem.
- */
-UNUSED_FUNC
-static int32_t tscGetMgmtConnMaxRetryTimes() {
-  int32_t factor = 2;
-  SRpcEpSet dump;
-  tscDumpMgmtEpSet(&dump);
-  return dump.numOfEps * factor;
-}
-
 void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) {
   STscObj *pObj = (STscObj *)param;
   if (pObj == NULL) return;
@@ -424,19 +417,16 @@ int doProcessSql(SSqlObj *pSql) {
 }
 
 int tscProcessSql(SSqlObj *pSql) {
-  char *   name = NULL;
+  char    *name = NULL;
   SSqlCmd *pCmd = &pSql->cmd;
   
-  SQueryInfo *    pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
+  SQueryInfo     *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
   STableMetaInfo *pTableMetaInfo = NULL;
   uint32_t        type = 0;
 
   if (pQueryInfo != NULL) {
     pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-    if (pTableMetaInfo != NULL) {
-      name = pTableMetaInfo->name;
-    }
-
+    name = (pTableMetaInfo != NULL)? pTableMetaInfo->name:NULL;
     type = pQueryInfo->type;
 
     // while numOfTables equals to 0, it must be Heartbeat
@@ -450,7 +440,6 @@ int tscProcessSql(SSqlObj *pSql) {
       return pSql->res.code;
     }
   } else if (pCmd->command < TSDB_SQL_LOCAL) {
-
     //pSql->epSet = tscMgmtEpSet;
   } else {  // local handler
     return (*tscProcessMsgRsp[pCmd->command])(pSql);
@@ -597,11 +586,11 @@ static char *doSerializeTableInfo(SQueryTableMsg* pQueryMsg, SSqlObj *pSql, char
     } else {
       pVgroupInfo = &pTableMeta->vgroupInfo;
     }
-    tscSetDnodeEpSet(pSql, pVgroupInfo);
 
-    if (pVgroupInfo != NULL) {
+    assert(pVgroupInfo != NULL);
+
+    tscSetDnodeEpSet(pSql, pVgroupInfo);
     pQueryMsg->head.vgId = htonl(pVgroupInfo->vgId);
-    } 
 
     STableIdInfo *pTableIdInfo = (STableIdInfo *)pMsg;
     pTableIdInfo->tid = htonl(pTableMeta->id.tid);
@@ -1460,7 +1449,7 @@ int tscProcessRetrieveLocalMergeRsp(SSqlObj *pSql) {
   SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
 
   if (pRes->code == TSDB_CODE_SUCCESS && pRes->numOfRows > 0) {
-    tscSetResultPointer(pQueryInfo, pRes);
+    tscCreateResPointerInfo(pRes, pQueryInfo);
   }
 
   pRes->row = 0;
@@ -1562,7 +1551,7 @@ int tscBuildMultiMeterMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) {
 
   // fill head info
   SMgmtHead *pMgmt = (SMgmtHead *)(pCmd->payload + tsRpcHeadSize);
-  memset(pMgmt->db, 0, TSDB_TABLE_ID_LEN);  // server don't need the db
+  memset(pMgmt->db, 0, TSDB_TABLE_FNAME_LEN);  // server don't need the db
 
   SCMMultiTableInfoMsg *pInfoMsg = (SCMMultiTableInfoMsg *)(pCmd->payload + tsRpcHeadSize + sizeof(SMgmtHead));
   pInfoMsg->numOfTables = htonl((int32_t)pCmd->count);
@@ -1603,7 +1592,7 @@ int tscBuildMultiMeterMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) {
 ////    tagLen += strlen(pQueryInfo->tagCond.tbnameCond.cond) * TSDB_NCHAR_SIZE;
 ////  }
 ////
-////  int32_t joinCondLen = (TSDB_TABLE_ID_LEN + sizeof(int16_t)) * 2;
+////  int32_t joinCondLen = (TSDB_TABLE_FNAME_LEN + sizeof(int16_t)) * 2;
 ////  int32_t elemSize = sizeof(SSuperTableMetaElemMsg) * pQueryInfo->numOfTables;
 ////
 ////  int32_t colSize = pQueryInfo->groupbyExpr.numOfGroupCols*sizeof(SColIndex);
@@ -1884,12 +1873,11 @@ int tscProcessSTableVgroupRsp(SSqlObj *pSql) {
 
       for (int32_t k = 0; k < pVgroups->numOfEps; ++k) {
         pVgroups->epAddr[k].port = htons(pVgroups->epAddr[k].port);
-
+      }
     }
 
     pMsg += size;
   }
-  }
   
   return pSql->res.code;
 }
@@ -1966,7 +1954,7 @@ int tscProcessShowRsp(SSqlObj *pSql) {
 }
 
 int tscProcessConnectRsp(SSqlObj *pSql) {
-  char temp[TSDB_TABLE_ID_LEN * 2];
+  char temp[TSDB_TABLE_FNAME_LEN * 2];
   STscObj *pObj = pSql->pTscObj;
   SSqlRes *pRes = &pSql->res;
 
@@ -2114,21 +2102,6 @@ int tscProcessRetrieveRspFromNode(SSqlObj *pSql) {
   return 0;
 }
 
-int tscProcessRetrieveRspFromLocal(SSqlObj *pSql) {
-  SSqlRes *   pRes = &pSql->res;
-  SSqlCmd *   pCmd = &pSql->cmd;
-  SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, 0);
-
-  SRetrieveTableRsp *pRetrieve = (SRetrieveTableRsp *)pRes->pRsp;
-
-  pRes->numOfRows = htonl(pRetrieve->numOfRows);
-  pRes->data = pRetrieve->data;
-
-  tscSetResultPointer(pQueryInfo, pRes);
-  pRes->row = 0;
-  return 0;
-}
-
 void tscTableMetaCallBack(void *param, TAOS_RES *res, int code);
 
 static int32_t getTableMetaFromMgmt(SSqlObj *pSql, STableMetaInfo *pTableMetaInfo) {

src/client/src/tscSql.c

@@ -233,15 +233,14 @@ static void waitForRetrieveRsp(void *param, TAOS_RES *tres, int numOfRows) {
   sem_post(&pSql->rspSem);
 }
 
-TAOS_RES* taos_query(TAOS *taos, const char *sqlstr) {
+TAOS_RES* taos_query_c(TAOS *taos, const char *sqlstr, uint32_t sqlLen) {
   STscObj *pObj = (STscObj *)taos;
   if (pObj == NULL || pObj->signature != pObj) {
     terrno = TSDB_CODE_TSC_DISCONNECTED;
     return NULL;
   }
   
-  int32_t sqlLen = (int32_t)strlen(sqlstr);
-  if (sqlLen > tsMaxSQLStringLen) {
+  if (sqlLen > (uint32_t)tsMaxSQLStringLen) {
     tscError("sql string exceeds max length:%d", tsMaxSQLStringLen);
     terrno = TSDB_CODE_TSC_INVALID_SQL;
     return NULL;
@@ -258,35 +257,14 @@ TAOS_RES* taos_query(TAOS *taos, const char *sqlstr) {
   
   doAsyncQuery(pObj, pSql, waitForQueryRsp, taos, sqlstr, sqlLen);
 
-  // wait for the callback function to post the semaphore
   tsem_wait(&pSql->rspSem);
   return pSql; 
 }
-TAOS_RES* taos_query_c(TAOS *taos, const char *sqlstr, uint32_t sqlLen) {
-  STscObj *pObj = (STscObj *)taos;
-  if (pObj == NULL || pObj->signature != pObj) {
-    terrno = TSDB_CODE_TSC_DISCONNECTED;
-    return NULL;
-  }
-  
-  if (sqlLen > (uint32_t)tsMaxSQLStringLen) {
-    tscError("sql string exceeds max length:%d", tsMaxSQLStringLen);
-    terrno = TSDB_CODE_TSC_INVALID_SQL;
-    return NULL;
-  }
 
-  SSqlObj* pSql = calloc(1, sizeof(SSqlObj));
-  if (pSql == NULL) {
-    tscError("failed to malloc sqlObj");
-    terrno = TSDB_CODE_TSC_OUT_OF_MEMORY;
-    return NULL;
-  }
-  
-  doAsyncQuery(pObj, pSql, waitForQueryRsp, taos, sqlstr, sqlLen);
-
-  tsem_wait(&pSql->rspSem);
-  return pSql; 
+TAOS_RES* taos_query(TAOS *taos, const char *sqlstr) {
+  return taos_query_c(taos, sqlstr, strlen(sqlstr));
 }
+
 int taos_result_precision(TAOS_RES *res) {
   SSqlObj *pSql = (SSqlObj *)res;
   if (pSql == NULL || pSql->signature != pSql) return 0;
@@ -423,6 +401,9 @@ TAOS_ROW taos_fetch_row(TAOS_RES *res) {
     return NULL;
   }
 
+  // set the sql object owner
+  tscSetSqlOwner(pSql);
+
   // current data set are exhausted, fetch more data from node
   if (pRes->row >= pRes->numOfRows && (pRes->completed != true || hasMoreVnodesToTry(pSql) || hasMoreClauseToTry(pSql)) &&
       (pCmd->command == TSDB_SQL_RETRIEVE ||
@@ -441,7 +422,10 @@ TAOS_ROW taos_fetch_row(TAOS_RES *res) {
     sem_wait(&pSql->rspSem);
   }
 
-  return doSetResultRowData(pSql, true);
+  void* data = doSetResultRowData(pSql, true);
+
+  tscClearSqlOwner(pSql);
+  return data;
 }
 
 int taos_fetch_block(TAOS_RES *res, TAOS_ROW *rows) {
@@ -509,7 +493,7 @@ int taos_select_db(TAOS *taos, const char *db) {
 }
 
 // send free message to vnode to free qhandle and corresponding resources in vnode
-static bool tscFreeQhandleInVnode(SSqlObj* pSql) {
+static bool tscKillQueryInVnode(SSqlObj* pSql) {
   SSqlCmd* pCmd = &pSql->cmd;
   SSqlRes* pRes = &pSql->res;
 
@@ -557,16 +541,14 @@ void taos_free_result(TAOS_RES *res) {
   }
 
   pQueryInfo->type = TSDB_QUERY_TYPE_FREE_RESOURCE;
-  if (!tscFreeQhandleInVnode(pSql)) {
+  if (!tscKillQueryInVnode(pSql)) {
     tscFreeSqlObj(pSql);
     tscDebug("%p sqlObj is freed by app", pSql);
   }
 }
 
-// todo should not be used in async query
 int taos_errno(TAOS_RES *tres) {
   SSqlObj *pSql = (SSqlObj *) tres;
-
   if (pSql == NULL || pSql->signature != pSql) {
     return terrno;
   }
@@ -812,7 +794,7 @@ static int tscParseTblNameList(SSqlObj *pSql, const char *tblNameList, int32_t t
   }
 
   char *nextStr;
-  char  tblName[TSDB_TABLE_ID_LEN];
+  char  tblName[TSDB_TABLE_FNAME_LEN];
   int   payloadLen = 0;
   char *pMsg = pCmd->payload;
   while (1) {

src/client/src/tscSubquery.c

@@ -1026,9 +1026,11 @@ void tscSetupOutputColumnIndex(SSqlObj* pSql) {
   }
 
   SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
-  pRes->pColumnIndex = calloc(1, sizeof(SColumnIndex) * pQueryInfo->fieldsInfo.numOfOutput);
 
-  for (int32_t i = 0; i < pQueryInfo->fieldsInfo.numOfOutput; ++i) {
+  int32_t numOfExprs = tscSqlExprNumOfExprs(pQueryInfo);
+  pRes->pColumnIndex = calloc(1, sizeof(SColumnIndex) * numOfExprs);
+
+  for (int32_t i = 0; i < numOfExprs; ++i) {
     SSqlExpr* pExpr = tscSqlExprGet(pQueryInfo, i);
 
     int32_t tableIndexOfSub = -1;
@@ -1045,8 +1047,8 @@ void tscSetupOutputColumnIndex(SSqlObj* pSql) {
     SSqlCmd* pSubCmd = &pSql->pSubs[tableIndexOfSub]->cmd;
     SQueryInfo* pSubQueryInfo = tscGetQueryInfoDetail(pSubCmd, 0);
     
-    size_t numOfExprs = taosArrayGetSize(pSubQueryInfo->exprList);
-    for (int32_t k = 0; k < numOfExprs; ++k) {
+    size_t numOfSubExpr = taosArrayGetSize(pSubQueryInfo->exprList);
+    for (int32_t k = 0; k < numOfSubExpr; ++k) {
       SSqlExpr* pSubExpr = tscSqlExprGet(pSubQueryInfo, k);
       if (pExpr->functionId == pSubExpr->functionId && pExpr->colInfo.colId == pSubExpr->colInfo.colId) {
         pRes->pColumnIndex[i] = (SColumnIndex){.tableIndex = tableIndexOfSub, .columnIndex = k};
@@ -1054,6 +1056,10 @@ void tscSetupOutputColumnIndex(SSqlObj* pSql) {
       }
     }
   }
+
+  // restore the offset value for super table query in case of final result.
+  tscRestoreSQLFuncForSTableQuery(pQueryInfo);
+  tscFieldInfoUpdateOffset(pQueryInfo);
 }
 
 void tscJoinQueryCallback(void* param, TAOS_RES* tres, int code) {
@@ -1079,7 +1085,7 @@ void tscJoinQueryCallback(void* param, TAOS_RES* tres, int code) {
   if (taos_errno(pSql) != TSDB_CODE_SUCCESS) {
     assert(taos_errno(pSql) == code);
 
-    tscError("%p abort query, code:%d, global code:%d", pSql, code, pParentSql->res.code);
+    tscError("%p abort query, code:%s, global code:%s", pSql, tstrerror(code), tstrerror(pParentSql->res.code));
     pParentSql->res.code = code;
 
     quitAllSubquery(pParentSql, pSupporter);
@@ -2052,9 +2058,7 @@ void tscBuildResFromSubqueries(SSqlObj *pSql) {
   }
 
   while (1) {
-    if (pRes->row < pRes->numOfRows) {
-      assert(0);
-    }
+    assert (pRes->row >= pRes->numOfRows);
 
     doBuildResFromSubqueries(pSql);
     sem_post(&pSql->rspSem);

src/client/src/tscUtil.c

@@ -644,7 +644,7 @@ int32_t tscMergeTableDataBlocks(SSqlObj* pSql, SArray* pTableDataBlockList) {
   STableDataBlocks* pOneTableBlock = taosArrayGetP(pTableDataBlockList, 0);
   int32_t expandSize = getRowExpandSize(pOneTableBlock->pTableMeta);
 
-  void* pVnodeDataBlockHashList = taosHashInit(128, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false);
+  void* pVnodeDataBlockHashList = taosHashInit(128, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), true, false);
   SArray* pVnodeDataBlockList = taosArrayInit(8, POINTER_BYTES);
 
   size_t total = taosArrayGetSize(pTableDataBlockList);
@@ -858,12 +858,13 @@ void tscFieldInfoCopy(SFieldInfo* dst, const SFieldInfo* src) {
 }
 
 TAOS_FIELD* tscFieldInfoGetField(SFieldInfo* pFieldInfo, int32_t index) {
+  assert(index < pFieldInfo->numOfOutput);
   return TARRAY_GET_ELEM(pFieldInfo->pFields, index);
 }
 
 int16_t tscFieldInfoGetOffset(SQueryInfo* pQueryInfo, int32_t index) {
   SFieldSupInfo* pInfo = tscFieldInfoGetSupp(&pQueryInfo->fieldsInfo, index);
-  assert(pInfo != NULL);
+  assert(pInfo != NULL && pInfo->pSqlExpr != NULL);
 
   return pInfo->pSqlExpr->offset;
 }
@@ -1680,6 +1681,77 @@ SSqlObj* createSimpleSubObj(SSqlObj* pSql, void (*fp)(), void* param, int32_t cm
   return pNew;
 }
 
+// current sql function is not direct output result, so create a dummy output field
+static void doSetNewFieldInfo(SQueryInfo* pNewQueryInfo, SSqlExpr* pExpr) {
+  TAOS_FIELD f = {.type = pExpr->resType, .bytes = pExpr->resBytes};
+  tstrncpy(f.name, pExpr->aliasName, sizeof(f.name));
+
+  SFieldSupInfo* pInfo1 = tscFieldInfoAppend(&pNewQueryInfo->fieldsInfo, &f);
+
+  pInfo1->pSqlExpr = pExpr;
+  pInfo1->visible = false;
+}
+
+static void doSetSqlExprAndResultFieldInfo(SQueryInfo* pQueryInfo, SQueryInfo* pNewQueryInfo, int64_t uid) {
+  int32_t numOfOutput = tscSqlExprNumOfExprs(pNewQueryInfo);
+  if (numOfOutput == 0) {
+    return;
+  }
+
+  size_t      numOfExprs = tscSqlExprNumOfExprs(pQueryInfo);
+  SFieldInfo* pFieldInfo = &pQueryInfo->fieldsInfo;
+
+  // set the field info in pNewQueryInfo object
+  for (int32_t i = 0; i < numOfExprs; ++i) {
+    SSqlExpr* pExpr = tscSqlExprGet(pQueryInfo, i);
+
+    if (pExpr->uid == uid) {
+      if (i < pFieldInfo->numOfOutput) {
+        SFieldSupInfo* pInfo = tscFieldInfoGetSupp(pFieldInfo, i);
+
+        if (pInfo->pSqlExpr != NULL) {
+          TAOS_FIELD* p = tscFieldInfoGetField(pFieldInfo, i);
+          assert(strcmp(p->name, pExpr->aliasName) == 0);
+
+          SFieldSupInfo* pInfo1 = tscFieldInfoAppend(&pNewQueryInfo->fieldsInfo, p);
+          *pInfo1 = *pInfo;
+        } else {
+          assert(pInfo->pArithExprInfo != NULL);
+          doSetNewFieldInfo(pNewQueryInfo, pExpr);
+        }
+      } else { // it is a arithmetic column, does not have actual field for sqlExpr, so build it
+        doSetNewFieldInfo(pNewQueryInfo, pExpr);
+      }
+    }
+  }
+
+  // make sure the the sqlExpr for each fields is correct
+  numOfExprs = tscSqlExprNumOfExprs(pNewQueryInfo);
+
+  // update the pSqlExpr pointer in SFieldSupInfo according the field name
+  // make sure the pSqlExpr point to the correct SqlExpr in pNewQueryInfo, not SqlExpr in pQueryInfo
+  for (int32_t f = 0; f < pNewQueryInfo->fieldsInfo.numOfOutput; ++f) {
+    TAOS_FIELD* field = tscFieldInfoGetField(&pNewQueryInfo->fieldsInfo, f);
+
+    bool matched = false;
+    for (int32_t k1 = 0; k1 < numOfExprs; ++k1) {
+      SSqlExpr* pExpr1 = tscSqlExprGet(pNewQueryInfo, k1);
+
+      if (strcmp(field->name, pExpr1->aliasName) == 0) {  // establish link according to the result field name
+        SFieldSupInfo* pInfo = tscFieldInfoGetSupp(&pNewQueryInfo->fieldsInfo, f);
+        pInfo->pSqlExpr = pExpr1;
+
+        matched = true;
+        break;
+      }
+    }
+
+    assert(matched);
+  }
+
+  tscFieldInfoUpdateOffset(pNewQueryInfo);
+}
+
 SSqlObj* createSubqueryObj(SSqlObj* pSql, int16_t tableIndex, void (*fp)(), void* param, int32_t cmd, SSqlObj* pPrevSql) {
   SSqlCmd* pCmd = &pSql->cmd;
   SSqlObj* pNew = (SSqlObj*)calloc(1, sizeof(SSqlObj));
@@ -1773,49 +1845,7 @@ SSqlObj* createSubqueryObj(SSqlObj* pSql, int16_t tableIndex, void (*fp)(), void
   uint64_t uid = pTableMetaInfo->pTableMeta->id.uid;
   tscSqlExprCopy(pNewQueryInfo->exprList, pQueryInfo->exprList, uid, true);
 
-  int32_t numOfOutput = (int32_t)tscSqlExprNumOfExprs(pNewQueryInfo);
-
-  if (numOfOutput > 0) {  // todo refactor to extract method
-    size_t numOfExprs = tscSqlExprNumOfExprs(pQueryInfo);
-    SFieldInfo* pFieldInfo = &pQueryInfo->fieldsInfo;
-    
-    for (int32_t i = 0; i < numOfExprs; ++i) {
-      SSqlExpr* pExpr = tscSqlExprGet(pQueryInfo, i);
-      
-      if (pExpr->uid == uid) {
-        TAOS_FIELD* p = tscFieldInfoGetField(pFieldInfo, i);
-        SFieldSupInfo* pInfo = tscFieldInfoGetSupp(pFieldInfo, i);
-  
-        SFieldSupInfo* pInfo1 = tscFieldInfoAppend(&pNewQueryInfo->fieldsInfo, p);
-        *pInfo1 = *pInfo;
-      }
-    }
-
-    // make sure the the sqlExpr for each fields is correct
-    // todo handle the agg arithmetic expression
-    numOfExprs = tscSqlExprNumOfExprs(pNewQueryInfo);
-
-    for(int32_t f = 0; f < pNewQueryInfo->fieldsInfo.numOfOutput; ++f) {
-      TAOS_FIELD* field = tscFieldInfoGetField(&pNewQueryInfo->fieldsInfo, f);
-      bool matched = false;
-
-      for(int32_t k1 = 0; k1 < numOfExprs; ++k1) {
-        SSqlExpr* pExpr1 = tscSqlExprGet(pNewQueryInfo, k1);
-
-        if (strcmp(field->name, pExpr1->aliasName) == 0) {  // establish link according to the result field name
-          SFieldSupInfo* pInfo = tscFieldInfoGetSupp(&pNewQueryInfo->fieldsInfo, f);
-          pInfo->pSqlExpr = pExpr1;
-
-          matched = true;
-          break;
-        }
-      }
-
-      assert(matched);
-    }
-  
-    tscFieldInfoUpdateOffset(pNewQueryInfo);
-  }
+  doSetSqlExprAndResultFieldInfo(pQueryInfo, pNewQueryInfo, uid);
 
   pNew->fp = fp;
   pNew->fetchFp = fp;
@@ -2013,6 +2043,10 @@ bool hasMoreVnodesToTry(SSqlObj* pSql) {
   }
   
   int32_t numOfVgroups = pTableMetaInfo->vgroupList->numOfVgroups;
+  if (pTableMetaInfo->pVgroupTables != NULL) {
+    numOfVgroups = taosArrayGetSize(pTableMetaInfo->pVgroupTables);
+  }
+
   return tscNonOrderedProjectionQueryOnSTable(pQueryInfo, 0) &&
          (!tscHasReachLimitation(pQueryInfo, pRes)) && (pTableMetaInfo->vgroupIndex < numOfVgroups - 1);
 }
@@ -2200,3 +2234,21 @@ int tscSetMgmtEpSetFromCfg(const char *first, const char *second) {
 
   return 0;
 }
+
+bool tscSetSqlOwner(SSqlObj* pSql) {
+  SSqlRes* pRes = &pSql->res;
+
+  // set the sql object owner
+  uint64_t threadId = taosGetPthreadId();
+  if (atomic_val_compare_exchange_64(&pSql->owner, 0, threadId) != 0) {
+    pRes->code = TSDB_CODE_QRY_IN_EXEC;
+    return false;
+  }
+
+  return true;
+}
+
+void tscClearSqlOwner(SSqlObj* pSql) {
+  assert(pSql->owner != 0);
+  atomic_store_64(&pSql->owner, 0);
+}
\ No newline at end of file

src/dnode/src/dnodeVRead.c

@@ -202,8 +202,9 @@ static void *dnodeProcessReadQueue(void *param) {
       break;
     }
 
-    dDebug("%p, msg:%s will be processed in vread queue, qtype:%d", pReadMsg->rpcMsg.ahandle,
-           taosMsg[pReadMsg->rpcMsg.msgType], type);
+    dDebug("%p, msg:%s will be processed in vread queue, qtype:%d, msg:%p", pReadMsg->rpcMsg.ahandle,
+           taosMsg[pReadMsg->rpcMsg.msgType], type, pReadMsg);
+
     int32_t code = vnodeProcessRead(pVnode, pReadMsg);
 
     if (type == TAOS_QTYPE_RPC && code != TSDB_CODE_QRY_NOT_READY) {

src/inc/query.h

@@ -76,6 +76,9 @@ void* qGetResultRetrieveMsg(qinfo_t qinfo);
  */
 int32_t qKillQuery(qinfo_t qinfo);
 
+int32_t qQueryCompleted(qinfo_t qinfo);
+
+
 /**
  * destroy query info structure
  * @param qHandle

src/inc/taosdef.h

@@ -232,7 +232,7 @@ void tsDataSwap(void *pLeft, void *pRight, int32_t type, int32_t size);
 #define TSDB_NODE_NAME_LEN        64
 #define TSDB_TABLE_NAME_LEN       193     // it is a null-terminated string
 #define TSDB_DB_NAME_LEN          33
-#define TSDB_TABLE_ID_LEN         (TSDB_ACCT_LEN + TSDB_DB_NAME_LEN + TSDB_TABLE_NAME_LEN)
+#define TSDB_TABLE_FNAME_LEN      (TSDB_ACCT_LEN + TSDB_DB_NAME_LEN + TSDB_TABLE_NAME_LEN)
 #define TSDB_COL_NAME_LEN         65
 #define TSDB_MAX_SAVED_SQL_LEN    TSDB_MAX_COLUMNS * 64
 #define TSDB_MAX_SQL_LEN          TSDB_PAYLOAD_SIZE

src/inc/taoserror.h

@@ -219,6 +219,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_QRY_DUP_JOIN_KEY,             0, 0x0705, "Duplicated
 TAOS_DEFINE_ERROR(TSDB_CODE_QRY_EXCEED_TAGS_LIMIT,        0, 0x0706, "Tag conditon too many")
 TAOS_DEFINE_ERROR(TSDB_CODE_QRY_NOT_READY,                0, 0x0707, "Query not ready")
 TAOS_DEFINE_ERROR(TSDB_CODE_QRY_HAS_RSP,                  0, 0x0708, "Query should response")
+TAOS_DEFINE_ERROR(TSDB_CODE_QRY_IN_EXEC,                  0, 0x0709, "Multiple retrieval of this query")
 
 // grant
 TAOS_DEFINE_ERROR(TSDB_CODE_GRANT_EXPIRED,                0, 0x0800, "License expired")

src/inc/taosmsg.h

@@ -246,13 +246,13 @@ typedef struct {
   uint64_t uid;
   uint64_t superTableUid;
   uint64_t createdTime;
-  char     tableId[TSDB_TABLE_ID_LEN];
-  char     superTableId[TSDB_TABLE_ID_LEN];
+  char     tableId[TSDB_TABLE_FNAME_LEN];
+  char     superTableId[TSDB_TABLE_FNAME_LEN];
   char     data[];
 } SMDCreateTableMsg;
 
 typedef struct {
-  char    tableId[TSDB_TABLE_ID_LEN];
+  char    tableId[TSDB_TABLE_FNAME_LEN];
   char    db[TSDB_ACCT_LEN + TSDB_DB_NAME_LEN];
   int8_t  igExists;
   int8_t  getMeta;
@@ -265,12 +265,12 @@ typedef struct {
 } SCMCreateTableMsg;
 
 typedef struct {
-  char   tableId[TSDB_TABLE_ID_LEN];
+  char   tableId[TSDB_TABLE_FNAME_LEN];
   int8_t igNotExists;
 } SCMDropTableMsg;
 
 typedef struct {
-  char    tableId[TSDB_TABLE_ID_LEN];
+  char    tableId[TSDB_TABLE_FNAME_LEN];
   char    db[TSDB_ACCT_LEN + TSDB_DB_NAME_LEN];
   int16_t type; /* operation type   */
   int16_t numOfCols; /* number of schema */
@@ -297,7 +297,7 @@ typedef struct {
 typedef struct {
   char clientVersion[TSDB_VERSION_LEN];
   char msgVersion[TSDB_VERSION_LEN];
-  char db[TSDB_TABLE_ID_LEN];
+  char db[TSDB_TABLE_FNAME_LEN];
 } SCMConnectMsg;
 
 typedef struct {
@@ -347,14 +347,14 @@ typedef struct {
   int32_t  vgId;
   int32_t  sid;
   uint64_t uid;
-  char     tableId[TSDB_TABLE_ID_LEN];
+  char     tableId[TSDB_TABLE_FNAME_LEN];
 } SMDDropTableMsg;
 
 typedef struct {
   int32_t  contLen;
   int32_t  vgId;
   uint64_t uid;
-  char    tableId[TSDB_TABLE_ID_LEN];
+  char    tableId[TSDB_TABLE_FNAME_LEN];
 } SMDDropSTableMsg;
 
 typedef struct {
@@ -527,7 +527,7 @@ typedef struct {
 } SCMCreateDbMsg, SCMAlterDbMsg;
 
 typedef struct {
-  char    db[TSDB_TABLE_ID_LEN];
+  char    db[TSDB_TABLE_FNAME_LEN];
   uint8_t ignoreNotExists;
 } SCMDropDbMsg, SCMUseDbMsg;
 
@@ -637,7 +637,7 @@ typedef struct {
 } SMDCreateVnodeMsg, SMDAlterVnodeMsg;
 
 typedef struct {
-  char    tableId[TSDB_TABLE_ID_LEN];
+  char    tableId[TSDB_TABLE_FNAME_LEN];
   int16_t createFlag;
   char    tags[];
 } SCMTableInfoMsg;
@@ -664,7 +664,7 @@ typedef struct {
 
 typedef struct STableMetaMsg {
   int32_t       contLen;
-  char          tableId[TSDB_TABLE_ID_LEN];   // table id
+  char          tableId[TSDB_TABLE_FNAME_LEN];   // table id
   uint8_t       numOfTags;
   uint8_t       precision;
   uint8_t       tableType;
@@ -685,7 +685,7 @@ typedef struct SMultiTableMeta {
 
 typedef struct {
   int32_t dataLen;
-  char name[TSDB_TABLE_ID_LEN];
+  char name[TSDB_TABLE_FNAME_LEN];
   char data[TSDB_MAX_TAGS_LEN + TD_KV_ROW_HEAD_SIZE + sizeof(SColIdx) * TSDB_MAX_TAGS];
 } STagData;
 
@@ -771,7 +771,7 @@ typedef struct {
   uint64_t uid;
   uint64_t stime;  // stream starting time
   int32_t  status;
-  char     tableId[TSDB_TABLE_ID_LEN];
+  char     tableId[TSDB_TABLE_FNAME_LEN];
 } SMDAlterStreamMsg;
 
 typedef struct {

src/inc/tsdb.h

@@ -166,10 +166,15 @@ typedef struct SDataBlockInfo {
   int32_t     tid;
 } SDataBlockInfo;
 
+typedef struct {
+  void  *pTable;
+  TSKEY  lastKey;
+} STableKeyInfo;
+
 typedef struct {
   size_t    numOfTables;
   SArray   *pGroupList;
-  SHashObj *map;         // speedup acquire the tableQueryInfo from STableId
+  SHashObj *map;         // speedup acquire the tableQueryInfo by table uid
 } STableGroupInfo;
 
 /**
@@ -177,11 +182,12 @@ typedef struct {
  *
  * @param tsdb       tsdb handle
  * @param pCond      query condition, including time window, result set order, and basic required columns for each block
- * @param tableqinfoGroupInfo  tableId list in the form of set, seperated into different groups according to group by condition
+ * @param tableInfoGroup  table object list in the form of set, grouped into different sets according to the
+ *                        group by condition
  * @param qinfo      query info handle from query processor
  * @return
  */
-TsdbQueryHandleT *tsdbQueryTables(TSDB_REPO_T *tsdb, STsdbQueryCond *pCond, STableGroupInfo *tableqinfoGroupInfo, void *qinfo);
+TsdbQueryHandleT *tsdbQueryTables(TSDB_REPO_T *tsdb, STsdbQueryCond *pCond, STableGroupInfo *tableInfoGroup, void *qinfo);
 
 /**
  * Get the last row of the given query time window for all the tables in STableGroupInfo object.
@@ -189,12 +195,11 @@ TsdbQueryHandleT *tsdbQueryTables(TSDB_REPO_T *tsdb, STsdbQueryCond *pCond, STab
  * all tables in this group.
  *
  * @param tsdb   tsdb handle
- * @param pCond       query condition, including time window, result set order, and basic required columns for each
- * block
- * @param tableqinfoGroupInfo   tableId list.
+ * @param pCond  query condition, including time window, result set order, and basic required columns for each block
+ * @param tableInfo  table list.
  * @return
  */
-TsdbQueryHandleT tsdbQueryLastRow(TSDB_REPO_T *tsdb, STsdbQueryCond *pCond, STableGroupInfo *tableqinfoGroupInfo, void *qinfo);
+TsdbQueryHandleT tsdbQueryLastRow(TSDB_REPO_T *tsdb, STsdbQueryCond *pCond, STableGroupInfo *tableInfo, void *qinfo);
 
 /**
  * get the queried table object list
@@ -260,7 +265,7 @@ SArray *tsdbRetrieveDataBlock(TsdbQueryHandleT *pQueryHandle, SArray *pColumnIdL
  * @param stableid. super table sid
  * @param pTagCond. tag query condition
  */
-int32_t tsdbQuerySTableByTagCond(TSDB_REPO_T *tsdb, uint64_t uid, const char *pTagCond, size_t len,
+int32_t tsdbQuerySTableByTagCond(TSDB_REPO_T *tsdb, uint64_t uid, TSKEY key, const char *pTagCond, size_t len,
                                  int16_t tagNameRelType, const char *tbnameCond, STableGroupInfo *pGroupList,
                                  SColIndex *pColIndex, int32_t numOfCols);
 
@@ -278,7 +283,7 @@ void tsdbDestroyTableGroup(STableGroupInfo *pGroupList);
  * @param pGroupInfo  the generated result
  * @return
  */
-int32_t tsdbGetOneTableGroup(TSDB_REPO_T *tsdb, uint64_t uid, STableGroupInfo *pGroupInfo);
+int32_t tsdbGetOneTableGroup(TSDB_REPO_T *tsdb, uint64_t uid, TSKEY startKey, STableGroupInfo *pGroupInfo);
 
 /**
  *

src/mnode/src/mnodeDb.c

@@ -193,7 +193,7 @@ void mnodeDecDbRef(SDbObj *pDb) {
 }
 
 SDbObj *mnodeGetDbByTableId(char *tableId) {
-  char db[TSDB_TABLE_ID_LEN], *pos;
+  char db[TSDB_TABLE_FNAME_LEN], *pos;
  
   // tableId format should be :  acct.db.table
   pos = strstr(tableId, TS_PATH_DELIMITER);
@@ -1046,7 +1046,7 @@ static int32_t mnodeProcessDropDbMsg(SMnodeMsg *pMsg) {
   if (pMsg->pDb == NULL) pMsg->pDb = mnodeGetDb(pDrop->db);
   if (pMsg->pDb == NULL) {
     if (pDrop->ignoreNotExists) {
-      mDebug("db:%s, db is not exist, think drop success", pDrop->db);
+      mDebug("db:%s, db is not exist, treat as success", pDrop->db);
       return TSDB_CODE_SUCCESS;
     } else {
       mError("db:%s, failed to drop, invalid db", pDrop->db);

src/mnode/src/mnodeProfile.c

@@ -73,13 +73,13 @@ int32_t mnodeInitProfile() {
 
 void mnodeCleanupProfile() {
   if (tsMnodeConnCache != NULL) {
-    mInfo("conn cache is cleanup");
     taosCacheCleanup(tsMnodeConnCache);
     tsMnodeConnCache = NULL;
   }
 }
 
 SConnObj *mnodeCreateConn(char *user, uint32_t ip, uint16_t port) {
+#if 0
   int32_t connSize = taosHashGetSize(tsMnodeConnCache->pHashTable);
   if (connSize > tsMaxShellConns) {
     mError("failed to create conn for user:%s ip:%s:%u, conns:%d larger than maxShellConns:%d, ", user, taosIpStr(ip),
@@ -87,6 +87,7 @@ SConnObj *mnodeCreateConn(char *user, uint32_t ip, uint16_t port) {
     terrno = TSDB_CODE_MND_TOO_MANY_SHELL_CONNS;
     return NULL;
   }
+#endif  
 
   int32_t connId = atomic_add_fetch_32(&tsConnIndex, 1);
   if (connId == 0) atomic_add_fetch_32(&tsConnIndex, 1);
@@ -112,7 +113,7 @@ void mnodeReleaseConn(SConnObj *pConn) {
 
 SConnObj *mnodeAccquireConn(int32_t connId, char *user, uint32_t ip, uint16_t port) {
   uint64_t expireTime = CONN_KEEP_TIME * 1000 + (uint64_t)taosGetTimestampMs();
-  SConnObj *pConn = taosCacheUpdateExpireTimeByName(tsMnodeConnCache, &connId, sizeof(int32_t), expireTime);
+  SConnObj *pConn = taosCacheAcquireByKey(tsMnodeConnCache, &connId, sizeof(int32_t));
   if (pConn == NULL) {
     mDebug("connId:%d, is already destroyed, user:%s ip:%s:%u", connId, user, taosIpStr(ip), port);
     return NULL;

src/mnode/src/mnodeSdb.c

@@ -874,7 +874,7 @@ void *sdbOpenTable(SSdbTableDesc *pDesc) {
   if (pTable->keyType == SDB_KEY_STRING || pTable->keyType == SDB_KEY_VAR_STRING) {
     hashFp = taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY);
   }
-  pTable->iHandle = taosHashInit(pTable->hashSessions, hashFp, true);
+  pTable->iHandle = taosHashInit(pTable->hashSessions, hashFp, true, true);
 
   tsSdbObj.numOfTables++;
   tsSdbObj.tableList[pTable->tableId] = pTable;

src/mnode/src/mnodeShow.c

@@ -302,7 +302,7 @@ static int32_t mnodeProcessConnectMsg(SMnodeMsg *pMsg) {
   SAcctObj *pAcct = pUser->pAcct;
 
   if (pConnectMsg->db[0]) {
-    char dbName[TSDB_TABLE_ID_LEN * 3] = {0};
+    char dbName[TSDB_TABLE_FNAME_LEN * 3] = {0};
     sprintf(dbName, "%x%s%s", pAcct->acctId, TS_PATH_DELIMITER, pConnectMsg->db);
     SDbObj *pDb = mnodeGetDb(dbName);
     if (pDb == NULL) {

src/mnode/src/mnodeTable.c

@@ -215,7 +215,7 @@ static int32_t mnodeChildTableActionEncode(SSdbOper *pOper) {
   assert(pTable != NULL && pOper->rowData != NULL);
 
   int32_t len = strlen(pTable->info.tableId);
-  if (len >= TSDB_TABLE_ID_LEN) return TSDB_CODE_MND_INVALID_TABLE_ID;
+  if (len >= TSDB_TABLE_FNAME_LEN) return TSDB_CODE_MND_INVALID_TABLE_ID;
 
   memcpy(pOper->rowData, pTable->info.tableId, len);
   memset(pOper->rowData + len, 0, 1);
@@ -246,7 +246,7 @@ static int32_t mnodeChildTableActionDecode(SSdbOper *pOper) {
   if (pTable == NULL) return TSDB_CODE_MND_OUT_OF_MEMORY;
 
   int32_t len = strlen(pOper->rowData);
-  if (len >= TSDB_TABLE_ID_LEN) {
+  if (len >= TSDB_TABLE_FNAME_LEN) {
     free(pTable);
     return TSDB_CODE_MND_INVALID_TABLE_ID;
   }
@@ -348,7 +348,7 @@ static int32_t mnodeInitChildTables() {
     .tableId      = SDB_TABLE_CTABLE,
     .tableName    = "ctables",
     .hashSessions = TSDB_DEFAULT_CTABLES_HASH_SIZE,
-    .maxRowSize   = sizeof(SChildTableObj) + sizeof(SSchema) * (TSDB_MAX_TAGS + TSDB_MAX_COLUMNS + 16) + TSDB_TABLE_ID_LEN + TSDB_CQ_SQL_SIZE,
+    .maxRowSize   = sizeof(SChildTableObj) + sizeof(SSchema) * (TSDB_MAX_TAGS + TSDB_MAX_COLUMNS + 16) + TSDB_TABLE_FNAME_LEN + TSDB_CQ_SQL_SIZE,
     .refCountPos  = (int8_t *)(&tObj.refCount) - (int8_t *)&tObj,
     .keyType      = SDB_KEY_VAR_STRING,
     .insertFp     = mnodeChildTableActionInsert,
@@ -387,7 +387,7 @@ static void mnodeAddTableIntoStable(SSuperTableObj *pStable, SChildTableObj *pCt
   atomic_add_fetch_32(&pStable->numOfTables, 1);
 
   if (pStable->vgHash == NULL) {
-    pStable->vgHash = taosHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), false);
+    pStable->vgHash = taosHashInit(64, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, false);
   }
 
   if (pStable->vgHash != NULL) {
@@ -479,7 +479,7 @@ static int32_t mnodeSuperTableActionEncode(SSdbOper *pOper) {
   assert(pOper->pObj != NULL && pOper->rowData != NULL);
 
   int32_t len = strlen(pStable->info.tableId);
-  if (len >= TSDB_TABLE_ID_LEN) len = TSDB_CODE_MND_INVALID_TABLE_ID;
+  if (len >= TSDB_TABLE_FNAME_LEN) len = TSDB_CODE_MND_INVALID_TABLE_ID;
 
   memcpy(pOper->rowData, pStable->info.tableId, len);
   memset(pOper->rowData + len, 0, 1);
@@ -503,7 +503,7 @@ static int32_t mnodeSuperTableActionDecode(SSdbOper *pOper) {
   if (pStable == NULL) return TSDB_CODE_MND_OUT_OF_MEMORY;
 
   int32_t len = strlen(pOper->rowData);
-  if (len >= TSDB_TABLE_ID_LEN){
+  if (len >= TSDB_TABLE_FNAME_LEN){
     free(pStable);
     return TSDB_CODE_MND_INVALID_TABLE_ID;
   }
@@ -539,7 +539,7 @@ static int32_t mnodeInitSuperTables() {
     .tableId      = SDB_TABLE_STABLE,
     .tableName    = "stables",
     .hashSessions = TSDB_DEFAULT_STABLES_HASH_SIZE,
-    .maxRowSize   = sizeof(SSuperTableObj) + sizeof(SSchema) * (TSDB_MAX_TAGS + TSDB_MAX_COLUMNS + 16) + TSDB_TABLE_ID_LEN,
+    .maxRowSize   = sizeof(SSuperTableObj) + sizeof(SSchema) * (TSDB_MAX_TAGS + TSDB_MAX_COLUMNS + 16) + TSDB_TABLE_FNAME_LEN,
     .refCountPos  = (int8_t *)(&tObj.refCount) - (int8_t *)&tObj,
     .keyType      = SDB_KEY_VAR_STRING,
     .insertFp     = mnodeSuperTableActionInsert,
@@ -751,7 +751,7 @@ static int32_t mnodeProcessDropTableMsg(SMnodeMsg *pMsg) {
   if (pMsg->pTable == NULL) pMsg->pTable = mnodeGetTable(pDrop->tableId);
   if (pMsg->pTable == NULL) {
     if (pDrop->igNotExists) {
-      mDebug("app:%p:%p, table:%s, table is not exist, think drop success", pMsg->rpcMsg.ahandle, pMsg, pDrop->tableId);
+      mDebug("app:%p:%p, table:%s, table is not exist, treat as success", pMsg->rpcMsg.ahandle, pMsg, pDrop->tableId);
       return TSDB_CODE_SUCCESS;
     } else {
       mError("app:%p:%p, table:%s, failed to drop table, table not exist", pMsg->rpcMsg.ahandle, pMsg, pDrop->tableId);
@@ -1464,7 +1464,7 @@ static int32_t mnodeProcessSuperTableVgroupMsg(SMnodeMsg *pMsg) {
   // reserve space
   int32_t contLen = sizeof(SCMSTableVgroupRspMsg) + 32 * sizeof(SCMVgroupInfo) + sizeof(SVgroupsInfo); 
   for (int32_t i = 0; i < numOfTable; ++i) {
-    char *stableName = (char*)pInfo + sizeof(SCMSTableVgroupMsg) + (TSDB_TABLE_ID_LEN) * i;
+    char *stableName = (char*)pInfo + sizeof(SCMSTableVgroupMsg) + (TSDB_TABLE_FNAME_LEN) * i;
     SSuperTableObj *pTable = mnodeGetSuperTable(stableName);
     if (pTable != NULL && pTable->vgHash != NULL) {
       contLen += (taosHashGetSize(pTable->vgHash) * sizeof(SCMVgroupInfo) + sizeof(SVgroupsInfo));
@@ -1481,7 +1481,7 @@ static int32_t mnodeProcessSuperTableVgroupMsg(SMnodeMsg *pMsg) {
   char *msg = (char *)pRsp + sizeof(SCMSTableVgroupRspMsg);
 
   for (int32_t i = 0; i < numOfTable; ++i) {
-    char *          stableName = (char *)pInfo + sizeof(SCMSTableVgroupMsg) + (TSDB_TABLE_ID_LEN)*i;
+    char *          stableName = (char *)pInfo + sizeof(SCMSTableVgroupMsg) + (TSDB_TABLE_FNAME_LEN)*i;
     SSuperTableObj *pTable = mnodeGetSuperTable(stableName);
     if (pTable == NULL) {
       mError("app:%p:%p, stable:%s, not exist while get stable vgroup info", pMsg->rpcMsg.ahandle, pMsg, stableName);
@@ -1828,7 +1828,7 @@ static int32_t mnodeSendDropChildTableMsg(SMnodeMsg *pMsg, bool needReturn) {
     return TSDB_CODE_MND_OUT_OF_MEMORY;
   }
 
-  tstrncpy(pDrop->tableId, pTable->info.tableId, TSDB_TABLE_ID_LEN);
+  tstrncpy(pDrop->tableId, pTable->info.tableId, TSDB_TABLE_FNAME_LEN);
   pDrop->vgId    = htonl(pTable->vgId);
   pDrop->contLen = htonl(sizeof(SMDDropTableMsg));
   pDrop->sid     = htonl(pTable->sid);
@@ -2079,7 +2079,7 @@ static int32_t mnodeDoGetChildTableMeta(SMnodeMsg *pMsg, STableMetaMsg *pMeta) {
   pMeta->sid       = htonl(pTable->sid);
   pMeta->precision = pDb->cfg.precision;
   pMeta->tableType = pTable->info.type;
-  tstrncpy(pMeta->tableId, pTable->info.tableId, TSDB_TABLE_ID_LEN);
+  tstrncpy(pMeta->tableId, pTable->info.tableId, TSDB_TABLE_FNAME_LEN);
 
   if (pTable->info.type == TSDB_CHILD_TABLE) {
     pMeta->sversion     = htons(pTable->superTable->sversion);
@@ -2448,7 +2448,7 @@ static int32_t mnodeProcessMultiTableMetaMsg(SMnodeMsg *pMsg) {
   pMultiMeta->numOfTables = 0;
 
   for (int32_t t = 0; t < pInfo->numOfTables; ++t) {
-    char * tableId = (char *)(pInfo->tableIds + t * TSDB_TABLE_ID_LEN);
+    char * tableId = (char *)(pInfo->tableIds + t * TSDB_TABLE_FNAME_LEN);
     SChildTableObj *pTable = mnodeGetChildTable(tableId);
     if (pTable == NULL) continue;
 

src/plugins/http/src/httpContext.c

@@ -167,8 +167,8 @@ bool httpInitContext(HttpContext *pContext) {
   memset(pParser, 0, sizeof(HttpParser));
   pParser->pCur = pParser->pLast = pParser->buffer;
 
-  httpDebug("context:%p, fd:%d, ip:%s, thread:%s, accessTimes:%d, parsed:%d",
-          pContext, pContext->fd, pContext->ipstr, pContext->pThread->label, pContext->accessTimes, pContext->parsed);
+  httpDebug("context:%p, fd:%d, ip:%s, accessTimes:%d, parsed:%d", pContext, pContext->fd, pContext->ipstr,
+            pContext->accessTimes, pContext->parsed);
   return true;
 }
 

src/plugins/monitor/src/monitorMain.c

@@ -192,7 +192,7 @@ static void dnodeBuildMonitorSql(char *sql, int32_t cmd) {
     snprintf(sql, SQL_LENGTH,
              "create table if not exists %s.slowquery(ts timestamp, username "
              "binary(%d), created_time timestamp, time bigint, sql binary(%d))",
-             tsMonitorDbName, TSDB_TABLE_ID_LEN - 1, TSDB_SLOW_QUERY_SQL_LEN);
+             tsMonitorDbName, TSDB_TABLE_FNAME_LEN - 1, TSDB_SLOW_QUERY_SQL_LEN);
   } else if (cmd == MONITOR_CMD_CREATE_TB_LOG) {
     snprintf(sql, SQL_LENGTH,
              "create table if not exists %s.log(ts timestamp, level tinyint, "

src/query/inc/qAst.h

@@ -48,7 +48,7 @@ typedef struct tQueryInfo {
   SSchema       sch;      // schema of tags
   char*         q;
   __compar_fn_t compare;  // filter function
-  void*         param;    // STSchema
+  bool          indexed;  // indexed columns
 } tQueryInfo;
 
 typedef struct SExprTraverseSupp {

src/query/inc/qExecutor.h

@@ -52,10 +52,10 @@ typedef struct SWindowStatus {
 
 typedef struct SWindowResult {
   uint16_t      numOfRows;   // number of rows of current  time window
+  SWindowStatus status;      // this result status: closed or opened
   SPosInfo      pos;         // Position of current result in disk-based output buffer
   SResultInfo*  resultInfo;  // For each result column, there is a resultInfo
   STimeWindow   window;      // The time window that current result covers.
-  SWindowStatus status;      // this result status: closed or opened
 } SWindowResult;
 
 /**
@@ -122,6 +122,7 @@ typedef struct SQueryCostInfo {
   uint32_t discardBlocks;
   uint64_t elapsedTime;
   uint64_t computTime;
+  uint64_t internalSupSize;
 } SQueryCostInfo;
 
 typedef struct SQuery {
@@ -184,10 +185,8 @@ enum {
 
 typedef struct SQInfo {
   void*            signature;
-  int32_t          pointsInterpo;
   int32_t          code;   // error code to returned to client
-//  sem_t            dataReady;
-
+  pthread_t        owner; // if it is in execution
   void*            tsdb;
   int32_t          vgId;
   STableGroupInfo  tableGroupInfo;       // table id list < only includes the STable list>

src/query/inc/qExtbuffer.h

@@ -29,7 +29,7 @@ extern "C" {
 
 #define MAX_TMPFILE_PATH_LENGTH        PATH_MAX
 #define INITIAL_ALLOCATION_BUFFER_SIZE 64
-#define DEFAULT_PAGE_SIZE              (4096L)  // 16k larger than the SHistoInfo
+#define DEFAULT_PAGE_SIZE              (1024L)  // 16k larger than the SHistoInfo
 
 typedef enum EXT_BUFFER_FLUSH_MODEL {
   /*

src/query/inc/sql.y

@@ -26,7 +26,12 @@
 #include <string.h>
 #include <assert.h>
 #include <stdbool.h>
+#include "qSqlparser.h"
+#include "tcmdtype.h"
+#include "tstoken.h"
+#include "ttokendef.h"
 #include "tutil.h"
+#include "tvariant.h"
 }
 
 %syntax_error {
@@ -254,7 +259,7 @@ alter_db_optr(Y) ::= alter_db_optr(Z) keep(X).        { Y = Z; Y.keep = X; }
 alter_db_optr(Y) ::= alter_db_optr(Z) blocks(X).      { Y = Z; Y.numOfBlocks = strtol(X.z, NULL, 10); }
 alter_db_optr(Y) ::= alter_db_optr(Z) comp(X).        { Y = Z; Y.compressionLevel = strtol(X.z, NULL, 10); }
 alter_db_optr(Y) ::= alter_db_optr(Z) wal(X).         { Y = Z; Y.walLevel = strtol(X.z, NULL, 10); }
-alter_db_optr(Y) ::= alter_db_optr(Z) fsync(X).       { Y = Z; Y.fsyncPeriod = strtod(X.z, NULL, 10); }
+alter_db_optr(Y) ::= alter_db_optr(Z) fsync(X).       { Y = Z; Y.fsyncPeriod = strtol(X.z, NULL, 10); }
 
 %type typename {TAOS_FIELD}
 typename(A) ::= ids(X). { 
@@ -422,8 +427,35 @@ as(X) ::= .             { X.n = 0;  }
 from(A) ::= FROM tablelist(X).                 {A = X;}
 
 %type tablelist {tVariantList*}
-tablelist(A) ::= ids(X) cpxName(Y).                     { toTSDBType(X.type); X.n += Y.n; A = tVariantListAppendToken(NULL, &X, -1);}
-tablelist(A) ::= tablelist(Y) COMMA ids(X) cpxName(Z).  { toTSDBType(X.type); X.n += Z.n; A = tVariantListAppendToken(Y, &X, -1);   }
+tablelist(A) ::= ids(X) cpxName(Y).                     {
+  toTSDBType(X.type);
+  X.n += Y.n;
+  A = tVariantListAppendToken(NULL, &X, -1);
+  A = tVariantListAppendToken(A, &X, -1);  // table alias name
+}
+
+tablelist(A) ::= ids(X) cpxName(Y) ids(Z).             {
+   toTSDBType(X.type);
+   toTSDBType(Z.type);
+   X.n += Y.n;
+   A = tVariantListAppendToken(NULL, &X, -1);
+   A = tVariantListAppendToken(A, &Z, -1);
+}
+
+tablelist(A) ::= tablelist(Y) COMMA ids(X) cpxName(Z).  {
+  toTSDBType(X.type);
+  X.n += Z.n;
+  A = tVariantListAppendToken(Y, &X, -1);
+  A = tVariantListAppendToken(A, &X, -1);
+}
+
+tablelist(A) ::= tablelist(Y) COMMA ids(X) cpxName(Z) ids(F). {
+   toTSDBType(X.type);
+   toTSDBType(F.type);
+   X.n += Z.n;
+   A = tVariantListAppendToken(Y, &X, -1);
+   A = tVariantListAppendToken(A, &F, -1);
+}
 
 // The value of interval should be the form of "number+[a,s,m,h,d,n,y]" or "now"
 %type tmvar {SSQLToken}

src/query/src/qAst.c

@@ -428,7 +428,8 @@ static void tQueryIndexColumn(SSkipList* pSkipList, tQueryInfo* pQueryInfo, SArr
           break;
         }
 
-        taosArrayPush(result, SL_GET_NODE_DATA(pNode));
+        STableKeyInfo info = {.pTable = *(void**)SL_GET_NODE_DATA(pNode), .lastKey = TSKEY_INITIAL_VAL};
+        taosArrayPush(result, &info);
       }
     } else if (optr == TSDB_RELATION_GREATER || optr == TSDB_RELATION_GREATER_EQUAL) { // greater equal
       bool comp = true;
@@ -445,7 +446,8 @@ static void tQueryIndexColumn(SSkipList* pSkipList, tQueryInfo* pQueryInfo, SArr
         if (ret == 0 && optr == TSDB_RELATION_GREATER) {
           continue;
         } else {
-          taosArrayPush(result, SL_GET_NODE_DATA(pNode));
+          STableKeyInfo info = {.pTable = *(void**)SL_GET_NODE_DATA(pNode), .lastKey = TSKEY_INITIAL_VAL};
+          taosArrayPush(result, &info);
           comp = false;
         }
       }
@@ -459,7 +461,8 @@ static void tQueryIndexColumn(SSkipList* pSkipList, tQueryInfo* pQueryInfo, SArr
           continue;
         }
 
-        taosArrayPush(result, SL_GET_NODE_DATA(pNode));
+        STableKeyInfo info = {.pTable = *(void**)SL_GET_NODE_DATA(pNode), .lastKey = TSKEY_INITIAL_VAL};
+        taosArrayPush(result, &info);
       }
       
       tSkipListDestroyIter(iter);
@@ -473,7 +476,8 @@ static void tQueryIndexColumn(SSkipList* pSkipList, tQueryInfo* pQueryInfo, SArr
           continue;
         }
 
-        taosArrayPush(result, SL_GET_NODE_DATA(pNode));
+        STableKeyInfo info = {.pTable = *(void**)SL_GET_NODE_DATA(pNode), .lastKey = TSKEY_INITIAL_VAL};
+        taosArrayPush(result, &info);
       }
   
     } else {
@@ -496,12 +500,14 @@ static void tQueryIndexColumn(SSkipList* pSkipList, tQueryInfo* pQueryInfo, SArr
         if (ret == 0 && optr == TSDB_RELATION_LESS) {
           continue;
         } else {
-          taosArrayPush(result, SL_GET_NODE_DATA(pNode));
+          STableKeyInfo info = {.pTable = *(void**)SL_GET_NODE_DATA(pNode), .lastKey = TSKEY_INITIAL_VAL};
+          taosArrayPush(result, &info);
           comp = false;  // no need to compare anymore
         }
       }
     }
   }
+
   free(cond.start); 
   free(cond.end);
   tSkipListDestroyIter(iter);
@@ -689,7 +695,8 @@ static void tQueryIndexlessColumn(SSkipList* pSkipList, tQueryInfo* pQueryInfo,
     }
 
     if (addToResult) {
-      taosArrayPush(res, pData);
+      STableKeyInfo info = {.pTable = *(void**)pData, .lastKey = TSKEY_INITIAL_VAL};
+      taosArrayPush(res, &info);
     }
   }
 
@@ -716,7 +723,7 @@ void tExprTreeTraverse(tExprNode *pExpr, SSkipList *pSkipList, SArray *result, S
     }
 
     tQueryInfo *pQueryInfo = pExpr->_node.info;
-    if (pQueryInfo->sch.colId == PRIMARYKEY_TIMESTAMP_COL_INDEX && pQueryInfo->optr != TSDB_RELATION_LIKE) {
+    if (pQueryInfo->indexed && pQueryInfo->optr != TSDB_RELATION_LIKE) {
       tQueryIndexColumn(pSkipList, pQueryInfo, result);
     } else {
       tQueryIndexlessColumn(pSkipList, pQueryInfo, result, param->nodeFilterFn);

src/query/src/qExtbuffer.c

@@ -12,15 +12,14 @@
  * You should have received a copy of the GNU Affero General Public License
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
-#include "qExtbuffer.h"
 #include "os.h"
+#include "qExtbuffer.h"
 #include "queryLog.h"
 #include "taos.h"
 #include "taosdef.h"
 #include "taosmsg.h"
 #include "tsqlfunction.h"
 #include "tulog.h"
-#include "tutil.h"
 
 #define COLMODEL_GET_VAL(data, schema, allrow, rowId, colId) \
   (data + (schema)->pFields[colId].offset * (allrow) + (rowId) * (schema)->pFields[colId].field.bytes)

src/query/src/qResultbuf.c

@@ -6,7 +6,8 @@
 #include "queryLog.h"
 #include "taoserror.h"
 
-#define GET_DATA_PAYLOAD(_p) ((tFilePage*)(((char*)(_p)->pData) + POINTER_BYTES))
+#define GET_DATA_PAYLOAD(_p) ((_p)->pData + POINTER_BYTES)
+#define NO_IN_MEM_AVAILABLE_PAGES(_b) (listNEles((_b)->lruList) >= (_b)->inMemPages)
 
 int32_t createDiskbasedResultBuffer(SDiskbasedResultBuf** pResultBuf, int32_t rowSize, int32_t pagesize,
                                     int32_t inMemBufSize, const void* handle) {
@@ -34,9 +35,9 @@ int32_t createDiskbasedResultBuffer(SDiskbasedResultBuf** pResultBuf, int32_t ro
   pResBuf->lruList = tdListNew(POINTER_BYTES);
 
   // init id hash table
-  pResBuf->groupSet  = taosHashInit(10, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), false);
+  pResBuf->groupSet  = taosHashInit(10, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, false);
   pResBuf->assistBuf = malloc(pResBuf->pageSize + 2); // EXTRA BYTES
-  pResBuf->all = taosHashInit(10, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), false);
+  pResBuf->all = taosHashInit(10, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, false);
 
   char path[PATH_MAX] = {0};
   taosGetTmpfilePath("qbuf", path);
@@ -186,8 +187,6 @@ static char* loadPageFromDisk(SDiskbasedResultBuf* pResultBuf, SPageInfo* pg) {
   return (char*)GET_DATA_PAYLOAD(pg);
 }
 
-#define NO_AVAILABLE_PAGES(_b) ((_b)->numOfPages >= (_b)->inMemPages)
-
 static SIDList addNewGroup(SDiskbasedResultBuf* pResultBuf, int32_t groupId) {
   assert(taosHashGet(pResultBuf->groupSet, (const char*) &groupId, sizeof(int32_t)) == NULL);
 
@@ -211,11 +210,12 @@ static SPageInfo* registerPage(SDiskbasedResultBuf* pResultBuf, int32_t groupId,
   pResultBuf->numOfPages += 1;
 
   SPageInfo* ppi = malloc(sizeof(SPageInfo));//{ .info = PAGE_INFO_INITIALIZER, .pageId = pageId, .pn = NULL};
-  ppi->info     = PAGE_INFO_INITIALIZER;
+
   ppi->pageId = pageId;
   ppi->pData  = NULL;
-  ppi->pn       = NULL;
+  ppi->info   = PAGE_INFO_INITIALIZER;
   ppi->used   = true;
+  ppi->pn     = NULL;
 
   return *(SPageInfo**) taosArrayPush(list, &ppi);
 }
@@ -246,7 +246,9 @@ static char* evicOneDataPage(SDiskbasedResultBuf* pResultBuf) {
   // all pages are referenced by user, try to allocate new space
   if (pn == NULL) {
     int32_t prev = pResultBuf->inMemPages;
-    pResultBuf->inMemPages = (int32_t)(pResultBuf->inMemPages * 1.5f);
+
+    // increase by 50% of previous mem pages
+    pResultBuf->inMemPages = pResultBuf->inMemPages * 1.5f;
 
     qWarn("%p in memory buf page not sufficient, expand from %d to %d, page size:%d", pResultBuf, prev,
           pResultBuf->inMemPages, pResultBuf->pageSize);
@@ -281,7 +283,7 @@ tFilePage* getNewDataBuf(SDiskbasedResultBuf* pResultBuf, int32_t groupId, int32
   pResultBuf->statis.getPages += 1;
 
   char* availablePage = NULL;
-  if (NO_AVAILABLE_PAGES(pResultBuf)) {
+  if (NO_IN_MEM_AVAILABLE_PAGES(pResultBuf)) {
     availablePage = evicOneDataPage(pResultBuf);
   }
 
@@ -340,7 +342,7 @@ tFilePage* getResBufPage(SDiskbasedResultBuf* pResultBuf, int32_t id) {
     assert((*pi)->pData == NULL && (*pi)->pn == NULL && (*pi)->info.length >= 0 && (*pi)->info.offset >= 0);
 
     char* availablePage = NULL;
-    if (NO_AVAILABLE_PAGES(pResultBuf)) {
+    if (NO_IN_MEM_AVAILABLE_PAGES(pResultBuf)) {
       availablePage = evicOneDataPage(pResultBuf);
     }
 
@@ -353,6 +355,8 @@ tFilePage* getResBufPage(SDiskbasedResultBuf* pResultBuf, int32_t id) {
     ((void**)((*pi)->pData))[0] = (*pi);
 
     lruListPushFront(pResultBuf->lruList, *pi);
+    (*pi)->used = true;
+
     loadPageFromDisk(pResultBuf, *pi);
     return GET_DATA_PAYLOAD(*pi);
   }
@@ -396,12 +400,13 @@ void destroyResultBuf(SDiskbasedResultBuf* pResultBuf) {
   }
 
   if (pResultBuf->file != NULL) {
-    qDebug("QInfo:%p disk-based output buffer closed, total:%" PRId64 " bytes, file size:%"PRId64" bytes",
-        pResultBuf->handle, pResultBuf->totalBufSize, pResultBuf->fileSize);
+    qDebug("QInfo:%p res output buffer closed, total:%" PRId64 " bytes, inmem size:%dbytes, file size:%"PRId64" bytes",
+        pResultBuf->handle, pResultBuf->totalBufSize, listNEles(pResultBuf->lruList) * pResultBuf->pageSize,
+        pResultBuf->fileSize);
 
     fclose(pResultBuf->file);
   } else {
-    qDebug("QInfo:%p disk-based output buffer closed, total:%" PRId64 " bytes, no file created", pResultBuf->handle,
+    qDebug("QInfo:%p res output buffer closed, total:%" PRId64 " bytes, no file created", pResultBuf->handle,
            pResultBuf->totalBufSize);
   }
 

src/query/src/qTokenizer.c

@@ -257,7 +257,7 @@ static void* KeywordHashTable = NULL;
 static void doInitKeywordsTable(void) {
   int numOfEntries = tListLen(keywordTable);
   
-  KeywordHashTable = taosHashInit(numOfEntries, MurmurHash3_32, false);
+  KeywordHashTable = taosHashInit(numOfEntries, MurmurHash3_32, true, false);
   for (int32_t i = 0; i < numOfEntries; i++) {
     keywordTable[i].len = (uint8_t)strlen(keywordTable[i].name);
     void* ptr = &keywordTable[i];

src/query/src/qTsbuf.c

@@ -72,6 +72,7 @@ STSBuf* tsBufCreateFromFile(const char* path, bool autoDelete) {
 
   // invalid file
   if (header.magic != TS_COMP_FILE_MAGIC) {
+    tsBufDestroy(pTSBuf);
     return NULL;
   }
   

src/query/src/qUtil.c

@@ -37,7 +37,7 @@ int32_t initWindowResInfo(SWindowResInfo *pWindowResInfo, SQueryRuntimeEnv *pRun
   
   pWindowResInfo->type = type;
   _hash_fn_t fn = taosGetDefaultHashFunction(type);
-  pWindowResInfo->hashList = taosHashInit(threshold, fn, false);
+  pWindowResInfo->hashList = taosHashInit(threshold, fn, true, false);
   if (pWindowResInfo->hashList == NULL) {
     return TSDB_CODE_QRY_OUT_OF_MEMORY;
   }
@@ -46,12 +46,17 @@ int32_t initWindowResInfo(SWindowResInfo *pWindowResInfo, SQueryRuntimeEnv *pRun
   pWindowResInfo->size     = 0;
   pWindowResInfo->prevSKey = TSKEY_INITIAL_VAL;
   
+  pRuntimeEnv->summary.internalSupSize += sizeof(SWindowResult) * threshold;
+
   // use the pointer arraylist
   pWindowResInfo->pResult = calloc(threshold, sizeof(SWindowResult));
   if (pWindowResInfo->pResult == NULL) {
     return TSDB_CODE_QRY_OUT_OF_MEMORY;
   }
 
+  pRuntimeEnv->summary.internalSupSize += sizeof(SWindowResult) * threshold;
+  pRuntimeEnv->summary.internalSupSize += (pRuntimeEnv->pQuery->numOfOutput * sizeof(SResultInfo) + pRuntimeEnv->interBufSize) * pWindowResInfo->capacity;
+
   for (int32_t i = 0; i < pWindowResInfo->capacity; ++i) {
     int32_t code = createQueryResultInfo(pRuntimeEnv->pQuery, &pWindowResInfo->pResult[i], pRuntimeEnv->stableQuery, pRuntimeEnv->interBufSize);
     if (code != TSDB_CODE_SUCCESS) {
@@ -104,7 +109,7 @@ void resetTimeWindowInfo(SQueryRuntimeEnv *pRuntimeEnv, SWindowResInfo *pWindowR
   pWindowResInfo->size = 0;
   
   _hash_fn_t fn = taosGetDefaultHashFunction(pWindowResInfo->type);
-  pWindowResInfo->hashList = taosHashInit(pWindowResInfo->capacity, fn, false);
+  pWindowResInfo->hashList = taosHashInit(pWindowResInfo->capacity, fn, true, false);
   
   pWindowResInfo->startTime = TSKEY_INITIAL_VAL;
   pWindowResInfo->prevSKey = TSKEY_INITIAL_VAL;

src/query/tests/resultBufferTest.cpp

@@ -130,7 +130,6 @@ void recyclePageTest() {
   ASSERT_TRUE(t4 == pBufPage4);
   ASSERT_TRUE(pageId == 4);
   releaseResBufPage(pResultBuf, t4);
-  releaseResBufPage(pResultBuf, t4);
 
   tFilePage* pBufPage5 = getNewDataBuf(pResultBuf, groupId, &pageId);
   tFilePage* t5 = getResBufPage(pResultBuf, pageId);

src/query/tests/tsBufTest.cpp

@@ -47,6 +47,8 @@ void simpleTest() {
   EXPECT_EQ(pTSBuf->block.numOfElem, num);
 
   tsBufDestroy(pTSBuf);
+
+  free(list);
 }
 
 // one large list of ts, the ts list need to be split into several small blocks
@@ -71,6 +73,7 @@ void largeTSTest() {
   EXPECT_EQ(pTSBuf->block.numOfElem, num);
 
   tsBufDestroy(pTSBuf);
+  free(list);
 }
 
 void multiTagsTest() {
@@ -208,6 +211,8 @@ void loadDataTest() {
 
   int64_t e = taosGetTimestampUs();
   printf("end:%" PRIu64 ", elapsed:%" PRIu64 ", total obj:%d\n", e, e - s, x);
+  tsBufDestroy(pTSBuf);
+  tsBufDestroy(pNewBuf);
 }
 
 void randomIncTsTest() {}
@@ -338,6 +343,8 @@ void TSTraverse() {
       }
     }
   }
+
+  tsBufDestroy(pTSBuf);
 }
 
 void performanceTest() {}
@@ -352,9 +359,12 @@ void invalidFileTest() {
 
   STSBuf* pNewBuf = tsBufCreateFromFile("/tmp/test", true);
   EXPECT_TRUE(pNewBuf == NULL);
+  tsBufDestroy(pNewBuf);
 
   pNewBuf = tsBufCreateFromFile("/tmp/911", true);
   EXPECT_TRUE(pNewBuf == NULL);
+
+  tsBufDestroy(pNewBuf);
 }
 
 void mergeDiffVnodeBufferTest() {

src/rpc/src/rpcMain.c

@@ -260,7 +260,7 @@ void *rpcOpen(const SRpcInit *pInit) {
   }
 
   if (pRpc->connType == TAOS_CONN_SERVER) {
-    pRpc->hash = taosHashInit(pRpc->sessions, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true);
+    pRpc->hash = taosHashInit(pRpc->sessions, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, true);
     if (pRpc->hash == NULL) {
       tError("%s failed to init string hash", pRpc->label);
       rpcClose(pRpc);

src/sync/src/syncMain.c

@@ -96,7 +96,7 @@ static void syncModuleInitFunc() {
     return;
   }
     
-  vgIdHash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true); 
+  vgIdHash = taosHashInit(TSDB_MIN_VNODES, taosGetDefaultHashFunction(TSDB_DATA_TYPE_INT), true, true);
   if (vgIdHash == NULL) {
     taosTmrCleanUp(syncTmrCtrl);
     taosCloseTcpThreadPool(tsTcpPool);

src/tsdb/src/tsdbMeta.c

@@ -443,7 +443,7 @@ STsdbMeta *tsdbNewMeta(STsdbCfg *pCfg) {
     goto _err;
   }
 
-  pMeta->uidMap = taosHashInit((size_t)(TSDB_INIT_NTABLES * 1.1), taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false);
+  pMeta->uidMap = taosHashInit((size_t)(TSDB_INIT_NTABLES * 1.1), taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), true, false);
   if (pMeta->uidMap == NULL) {
     terrno = TSDB_CODE_TDB_OUT_OF_MEMORY;
     goto _err;

src/tsdb/src/tsdbRead.c

@@ -208,6 +208,7 @@ TsdbQueryHandleT* tsdbQueryTables(TSDB_REPO_T* tsdb, STsdbQueryCond* pCond, STab
   if (pQueryHandle->statis == NULL) {
     goto out_of_memory;
   }
+
   pQueryHandle->pColumns = taosArrayInit(numOfCols, sizeof(SColumnInfoData));  // todo: use list instead of array?
   if (pQueryHandle->pColumns == NULL) {
     goto out_of_memory;
@@ -229,8 +230,9 @@ TsdbQueryHandleT* tsdbQueryTables(TSDB_REPO_T* tsdb, STsdbQueryCond* pCond, STab
   if (pQueryHandle->pTableCheckInfo == NULL) {
     goto out_of_memory;
   }
+
   STsdbMeta* pMeta = tsdbGetMeta(tsdb);
-  assert(pMeta != NULL);
+  assert(pMeta != NULL && sizeOfGroup >= 1 && pCond != NULL && pCond->numOfCols > 0);
 
   for (int32_t i = 0; i < sizeOfGroup; ++i) {
     SArray* group = *(SArray**) taosArrayGet(groupList->pGroupList, i);
@@ -239,18 +241,26 @@ TsdbQueryHandleT* tsdbQueryTables(TSDB_REPO_T* tsdb, STsdbQueryCond* pCond, STab
     assert(gsize > 0);
 
     for (int32_t j = 0; j < gsize; ++j) {
-      STable* pTable = (STable*) taosArrayGetP(group, j);
+      STableKeyInfo* pKeyInfo = (STableKeyInfo*) taosArrayGet(group, j);
 
       STableCheckInfo info = {
-          .lastKey = pQueryHandle->window.skey,
-          //.tableId = pTable->tableId,
-          .pTableObj = pTable,
+          .lastKey = pKeyInfo->lastKey,
+          .tableId = ((STable*)(pKeyInfo->pTable))->tableId,
+          .pTableObj = pKeyInfo->pTable,
       };
-      info.tableId = pTable->tableId;
 
       assert(info.pTableObj != NULL && (info.pTableObj->type == TSDB_NORMAL_TABLE ||
       info.pTableObj->type == TSDB_CHILD_TABLE || info.pTableObj->type == TSDB_STREAM_TABLE));
 
+      info.tableId.tid = info.pTableObj->tableId.tid;
+      info.tableId.uid = info.pTableObj->tableId.uid;
+
+      if (ASCENDING_TRAVERSE(pQueryHandle->order)) {
+        assert(info.lastKey >= pQueryHandle->window.skey);
+      } else {
+        assert(info.lastKey <= pQueryHandle->window.skey);
+      }
+
       taosArrayPush(pQueryHandle->pTableCheckInfo, &info);
       tsdbDebug("%p check table uid:%"PRId64", tid:%d from lastKey:%"PRId64" %p", pQueryHandle, info.tableId.uid,
           info.tableId.tid, info.lastKey, qinfo);
@@ -326,17 +336,23 @@ static bool initTableMemIterator(STsdbQueryHandle* pHandle, STableCheckInfo* pCh
 
   assert(pCheckInfo->iter == NULL && pCheckInfo->iiter == NULL);
 
-  // TODO: add uid check
-  if (pHandle->mem && pCheckInfo->tableId.tid < pHandle->mem->maxTables &&
-      pHandle->mem->tData[pCheckInfo->tableId.tid] != NULL) {
-    pCheckInfo->iter = tSkipListCreateIterFromVal(pHandle->mem->tData[pCheckInfo->tableId.tid]->pData,
-                                                  (const char*)&pCheckInfo->lastKey, TSDB_DATA_TYPE_TIMESTAMP, order);
+  STableData* pMem = NULL;
+  STableData* pIMem = NULL;
+
+  if (pHandle->mem && pCheckInfo->tableId.tid < pHandle->mem->maxTables) {
+    pMem = pHandle->mem->tData[pCheckInfo->tableId.tid];
+    if (pMem != NULL && pMem->uid == pCheckInfo->tableId.uid) { // check uid
+      pCheckInfo->iter =
+          tSkipListCreateIterFromVal(pMem->pData, (const char*)&pCheckInfo->lastKey, TSDB_DATA_TYPE_TIMESTAMP, order);
+    }
   }
 
-  if (pHandle->imem && pCheckInfo->tableId.tid < pHandle->imem->maxTables &&
-      pHandle->imem->tData[pCheckInfo->tableId.tid] != NULL) {
-    pCheckInfo->iiter = tSkipListCreateIterFromVal(pHandle->imem->tData[pCheckInfo->tableId.tid]->pData,
-                                                   (const char*)&pCheckInfo->lastKey, TSDB_DATA_TYPE_TIMESTAMP, order);
+  if (pHandle->imem && pCheckInfo->tableId.tid < pHandle->imem->maxTables) {
+    pIMem = pHandle->imem->tData[pCheckInfo->tableId.tid];
+    if (pIMem != NULL && pIMem->uid == pCheckInfo->tableId.uid) { // check uid
+      pCheckInfo->iiter =
+          tSkipListCreateIterFromVal(pIMem->pData, (const char*)&pCheckInfo->lastKey, TSDB_DATA_TYPE_TIMESTAMP, order);
+    }
   }
 
   // both iterators are NULL, no data in buffer right now
@@ -356,8 +372,17 @@ static bool initTableMemIterator(STsdbQueryHandle* pHandle, STableCheckInfo* pCh
 
     SDataRow row = *(SDataRow *)SL_GET_NODE_DATA(node);
     TSKEY key = dataRowKey(row);  // first timestamp in buffer
-    tsdbDebug("%p uid:%" PRId64", tid:%d check data in mem from skey:%" PRId64 ", order:%d, %p", pHandle,
-           pCheckInfo->tableId.uid, pCheckInfo->tableId.tid, key, order, pHandle->qinfo);
+    tsdbDebug("%p uid:%" PRId64 ", tid:%d check data in mem from skey:%" PRId64 ", order:%d, ts range in buf:%" PRId64
+              "-%" PRId64 ", lastKey:%" PRId64 ", %p",
+              pHandle, pCheckInfo->tableId.uid, pCheckInfo->tableId.tid, key, order, pMem->keyFirst, pMem->keyLast,
+              pCheckInfo->lastKey, pHandle->qinfo);
+
+    if (ASCENDING_TRAVERSE(order)) {
+      assert(pCheckInfo->lastKey <= key);
+    } else {
+      assert(pCheckInfo->lastKey >= key);
+    }
+
   } else {
     tsdbDebug("%p uid:%"PRId64", tid:%d no data in mem, %p", pHandle, pCheckInfo->tableId.uid, pCheckInfo->tableId.tid,
         pHandle->qinfo);
@@ -369,8 +394,16 @@ static bool initTableMemIterator(STsdbQueryHandle* pHandle, STableCheckInfo* pCh
 
     SDataRow row = *(SDataRow *)SL_GET_NODE_DATA(node);
     TSKEY key = dataRowKey(row);  // first timestamp in buffer
-    tsdbDebug("%p uid:%" PRId64", tid:%d check data in imem from skey:%" PRId64 ", order:%d, %p", pHandle,
-           pCheckInfo->tableId.uid, pCheckInfo->tableId.tid, key, order, pHandle->qinfo);
+    tsdbDebug("%p uid:%" PRId64 ", tid:%d check data in mem from skey:%" PRId64 ", order:%d, ts range in buf:%" PRId64
+              "-%" PRId64 ", lastKey:%" PRId64 ", %p",
+              pHandle, pCheckInfo->tableId.uid, pCheckInfo->tableId.tid, key, order, pIMem->keyFirst, pIMem->keyLast,
+              pCheckInfo->lastKey, pHandle->qinfo);
+
+    if (ASCENDING_TRAVERSE(order)) {
+      assert(pCheckInfo->lastKey <= key);
+    } else {
+      assert(pCheckInfo->lastKey >= key);
+    }
   } else {
     tsdbDebug("%p uid:%"PRId64", tid:%d no data in imem, %p", pHandle, pCheckInfo->tableId.uid, pCheckInfo->tableId.tid,
         pHandle->qinfo);
@@ -2051,7 +2084,9 @@ static int32_t getAllTableList(STable* pSuperTable, SArray* list) {
     SSkipListNode* pNode = tSkipListIterGet(iter);
 
     STable** pTable = (STable**) SL_GET_NODE_DATA((SSkipListNode*) pNode);
-    taosArrayPush(list, pTable);
+
+    STableKeyInfo info = {.pTable = *pTable, .lastKey = TSKEY_INITIAL_VAL};
+    taosArrayPush(list, &info);
   }
 
   tSkipListDestroyIter(iter);
@@ -2089,7 +2124,7 @@ void filterPrepare(void* expr, void* param) {
   pInfo->sch      = *pSchema;
   pInfo->optr     = pExpr->_node.optr;
   pInfo->compare  = getComparFunc(pSchema->type, pInfo->optr);
-  pInfo->param    = pTSSchema;
+  pInfo->indexed  = pTSSchema->columns->colId == pInfo->sch.colId;
 
   if (pInfo->optr == TSDB_RELATION_IN) {
     pInfo->q = (char*) pCond->arr;
@@ -2107,8 +2142,8 @@ typedef struct STableGroupSupporter {
 
 int32_t tableGroupComparFn(const void *p1, const void *p2, const void *param) {
   STableGroupSupporter* pTableGroupSupp = (STableGroupSupporter*) param;
-  STable* pTable1 = *(STable**) p1;
-  STable* pTable2 = *(STable**) p2;
+  STable* pTable1 = ((STableKeyInfo*) p1)->pTable;
+  STable* pTable2 = ((STableKeyInfo*) p2)->pTable;
 
   for (int32_t i = 0; i < pTableGroupSupp->numOfCols; ++i) {
     SColIndex* pColIndex = &pTableGroupSupp->pCols[i];
@@ -2158,12 +2193,14 @@ int32_t tableGroupComparFn(const void *p1, const void *p2, const void *param) {
   return 0;
 }
 
-void createTableGroupImpl(SArray* pGroups, SArray* pTableList, size_t numOfTables, STableGroupSupporter* pSupp,
+void createTableGroupImpl(SArray* pGroups, SArray* pTableList, size_t numOfTables, TSKEY skey, STableGroupSupporter* pSupp,
     __ext_compar_fn_t compareFn) {
   STable* pTable = taosArrayGetP(pTableList, 0);
 
-  SArray* g = taosArrayInit(16, POINTER_BYTES);
-  taosArrayPush(g, &pTable);
+  SArray* g = taosArrayInit(16, sizeof(STableKeyInfo));
+
+  STableKeyInfo info = {.pTable = pTable, .lastKey = skey};
+  taosArrayPush(g, &info);
   tsdbRefTable(pTable);
 
   for (int32_t i = 1; i < numOfTables; ++i) {
@@ -2177,18 +2214,21 @@ void createTableGroupImpl(SArray* pGroups, SArray* pTableList, size_t numOfTable
     assert((*p)->type == TSDB_CHILD_TABLE);
 
     if (ret == 0) {
-      taosArrayPush(g, p);
+      STableKeyInfo info1 = {.pTable = *p, .lastKey = skey};
+      taosArrayPush(g, &info1);
     } else {
       taosArrayPush(pGroups, &g);  // current group is ended, start a new group
-      g = taosArrayInit(16, POINTER_BYTES);
-      taosArrayPush(g, p);
+      g = taosArrayInit(16, sizeof(STableKeyInfo));
+
+      STableKeyInfo info1 = {.pTable = *p, .lastKey = skey};
+      taosArrayPush(g, &info1);
     }
   }
 
   taosArrayPush(pGroups, &g);
 }
 
-SArray* createTableGroup(SArray* pTableList, STSchema* pTagSchema, SColIndex* pCols, int32_t numOfOrderCols) {
+SArray* createTableGroup(SArray* pTableList, STSchema* pTagSchema, SColIndex* pCols, int32_t numOfOrderCols, TSKEY skey) {
   assert(pTableList != NULL);
   SArray* pTableGroup = taosArrayInit(1, POINTER_BYTES);
 
@@ -2199,13 +2239,16 @@ SArray* createTableGroup(SArray* pTableList, STSchema* pTagSchema, SColIndex* pC
   }
 
   if (numOfOrderCols == 0 || size == 1) { // no group by tags clause or only one table
-    SArray* sa = taosArrayInit(size, POINTER_BYTES);
+    SArray* sa = taosArrayInit(size, sizeof(STableKeyInfo));
+
     for(int32_t i = 0; i < size; ++i) {
-      STable** pTable = taosArrayGet(pTableList, i);
-      assert((*pTable)->type == TSDB_CHILD_TABLE);
+      STableKeyInfo *pKeyInfo = taosArrayGet(pTableList, i);
+      assert(((STable*)pKeyInfo->pTable)->type == TSDB_CHILD_TABLE);
 
-      tsdbRefTable(*pTable);
-      taosArrayPush(sa, pTable);
+      tsdbRefTable(pKeyInfo->pTable);
+
+      STableKeyInfo info = {.pTable = pKeyInfo->pTable, .lastKey = skey};
+      taosArrayPush(sa, &info);
     }
 
     taosArrayPush(pTableGroup, &sa);
@@ -2216,8 +2259,8 @@ SArray* createTableGroup(SArray* pTableList, STSchema* pTagSchema, SColIndex* pC
     pSupp->pTagSchema = pTagSchema;
     pSupp->pCols = pCols;
 
-    taosqsort(pTableList->pData, size, POINTER_BYTES, pSupp, tableGroupComparFn);
-    createTableGroupImpl(pTableGroup, pTableList, size, pSupp, tableGroupComparFn);
+    taosqsort(pTableList->pData, size, sizeof(STableKeyInfo), pSupp, tableGroupComparFn);
+    createTableGroupImpl(pTableGroup, pTableList, size, skey, pSupp, tableGroupComparFn);
     taosTFree(pSupp);
   }
 
@@ -2290,7 +2333,7 @@ static int32_t doQueryTableList(STable* pSTable, SArray* pRes, tExprNode* pExpr)
   return TSDB_CODE_SUCCESS;
 }
 
-int32_t tsdbQuerySTableByTagCond(TSDB_REPO_T* tsdb, uint64_t uid, const char* pTagCond, size_t len,
+int32_t tsdbQuerySTableByTagCond(TSDB_REPO_T* tsdb, uint64_t uid, TSKEY skey, const char* pTagCond, size_t len,
                                  int16_t tagNameRelType, const char* tbnameCond, STableGroupInfo* pGroupInfo,
                                  SColIndex* pColIndex, int32_t numOfCols) {
   if (tsdbRLockRepoMeta(tsdb) < 0) goto _error;
@@ -2314,7 +2357,7 @@ int32_t tsdbQuerySTableByTagCond(TSDB_REPO_T* tsdb, uint64_t uid, const char* pT
   }
 
   //NOTE: not add ref count for super table
-  SArray* res = taosArrayInit(8, POINTER_BYTES);
+  SArray* res = taosArrayInit(8, sizeof(STableKeyInfo));
   STSchema* pTagSchema = tsdbGetTableTagSchema(pTable);
 
   // no tags and tbname condition, all child tables of this stable are involved
@@ -2326,7 +2369,7 @@ int32_t tsdbQuerySTableByTagCond(TSDB_REPO_T* tsdb, uint64_t uid, const char* pT
     }
 
     pGroupInfo->numOfTables = taosArrayGetSize(res);
-    pGroupInfo->pGroupList  = createTableGroup(res, pTagSchema, pColIndex, numOfCols);
+    pGroupInfo->pGroupList  = createTableGroup(res, pTagSchema, pColIndex, numOfCols, skey);
 
     tsdbDebug("%p no table name/tag condition, all tables belong to one group, numOfTables:%zu", tsdb, pGroupInfo->numOfTables);
     taosArrayDestroy(res);
@@ -2369,7 +2412,7 @@ int32_t tsdbQuerySTableByTagCond(TSDB_REPO_T* tsdb, uint64_t uid, const char* pT
 
   doQueryTableList(pTable, res, expr);
   pGroupInfo->numOfTables = taosArrayGetSize(res);
-  pGroupInfo->pGroupList  = createTableGroup(res, pTagSchema, pColIndex, numOfCols);
+  pGroupInfo->pGroupList  = createTableGroup(res, pTagSchema, pColIndex, numOfCols, skey);
 
   tsdbDebug("%p stable tid:%d, uid:%"PRIu64" query, numOfTables:%zu, belong to %zu groups", tsdb, pTable->tableId.tid,
       pTable->tableId.uid, pGroupInfo->numOfTables, taosArrayGetSize(pGroupInfo->pGroupList));
@@ -2383,7 +2426,7 @@ int32_t tsdbQuerySTableByTagCond(TSDB_REPO_T* tsdb, uint64_t uid, const char* pT
   return terrno;
 }
 
-int32_t tsdbGetOneTableGroup(TSDB_REPO_T* tsdb, uint64_t uid, STableGroupInfo* pGroupInfo) {
+int32_t tsdbGetOneTableGroup(TSDB_REPO_T* tsdb, uint64_t uid, TSKEY startKey, STableGroupInfo* pGroupInfo) {
   if (tsdbRLockRepoMeta(tsdb) < 0) goto _error;
 
   STable* pTable = tsdbGetTableByUid(tsdbGetMeta(tsdb), uid);
@@ -2400,9 +2443,11 @@ int32_t tsdbGetOneTableGroup(TSDB_REPO_T* tsdb, uint64_t uid, STableGroupInfo* p
   pGroupInfo->numOfTables = 1;
   pGroupInfo->pGroupList = taosArrayInit(1, POINTER_BYTES);
 
-  SArray* group = taosArrayInit(1, POINTER_BYTES);
+  SArray* group = taosArrayInit(1, sizeof(STableKeyInfo));
+
+  STableKeyInfo info = {.pTable = pTable, .lastKey = startKey};
+  taosArrayPush(group, &info);
 
-  taosArrayPush(group, &pTable);
   taosArrayPush(pGroupInfo->pGroupList, &group);
 
   return TSDB_CODE_SUCCESS;
@@ -2419,7 +2464,7 @@ int32_t tsdbGetTableGroupFromIdList(TSDB_REPO_T* tsdb, SArray* pTableIdList, STa
   assert(pTableIdList != NULL);
   size_t size = taosArrayGetSize(pTableIdList);
   pGroupInfo->pGroupList = taosArrayInit(1, POINTER_BYTES);
-  SArray* group = taosArrayInit(1, POINTER_BYTES);
+  SArray* group = taosArrayInit(1, sizeof(STableKeyInfo));
 
   int32_t i = 0;
   for(; i < size; ++i) {
@@ -2437,7 +2482,9 @@ int32_t tsdbGetTableGroupFromIdList(TSDB_REPO_T* tsdb, SArray* pTableIdList, STa
     }
 
     tsdbRefTable(pTable);
-    taosArrayPush(group, &pTable);
+
+    STableKeyInfo info = {.pTable = pTable, .lastKey = id->key};
+    taosArrayPush(group, &info);
   }
 
   if (tsdbUnlockRepoMeta(tsdb) < 0) {

src/util/inc/hash.h

@@ -20,7 +20,9 @@
 extern "C" {
 #endif
 
+#include "tarray.h"
 #include "hashfunc.h"
+#include "tlockfree.h"
 
 #define HASH_MAX_CAPACITY (1024 * 1024 * 16)
 #define HASH_DEFAULT_LOAD_FACTOR (0.75)
@@ -30,37 +32,44 @@ typedef void (*_hash_free_fn_t)(void *param);
 
 typedef struct SHashNode {
   char             *key;
-//  union {
-    struct SHashNode * prev;
-//    struct SHashEntry *prev1;
-//  };
-//
+//  struct SHashNode *prev;
   struct SHashNode *next;
   uint32_t          hashVal;  // the hash value of key, if hashVal == HASH_VALUE_IN_TRASH, this node is moved to trash
   uint32_t          keyLen;   // length of the key
-  char              data[];
+  char             *data;
 } SHashNode;
 
+typedef enum SHashLockTypeE {
+  HASH_NO_LOCK     = 0,
+  HASH_ENTRY_LOCK  = 1,
+} SHashLockTypeE;
+
+typedef struct SHashEntry {
+  int32_t    num;      // number of elements in current entry
+  SRWLatch   latch;    // entry latch
+  SHashNode *next;
+} SHashEntry;
+
 typedef struct SHashObj {
-  SHashNode     **hashList;
+  SHashEntry    **hashList;
   size_t          capacity;     // number of slots
   size_t          size;         // number of elements in hash table
   _hash_fn_t      hashFp;       // hash function
   _hash_free_fn_t freeFp;       // hash node free callback function
 
-#if defined(LINUX)
-  pthread_rwlock_t *lock;
-#else
-  pthread_mutex_t *lock;
-#endif
+  SRWLatch        lock;         // read-write spin lock
+  SHashLockTypeE  type;         // lock type
+  bool            enableUpdate; // enable update
+  SArray         *pMemBlock;    // memory block allocated for SHashEntry
 } SHashObj;
 
 typedef struct SHashMutableIterator {
-  SHashObj * pHashObj;
+  SHashObj  *pHashObj;
   int32_t    entryIndex;
   SHashNode *pCur;
   SHashNode *pNext;           // current node can be deleted for mutable iterator, so keep the next one before return current
-  int32_t    num;    // already check number of elements in hash table
+  size_t     numOfChecked;    // already check number of elements in hash table
+  size_t     numOfEntries;    // number of entries while the iterator is created
 } SHashMutableIterator;
 
 /**
@@ -71,7 +80,7 @@ typedef struct SHashMutableIterator {
  * @param threadsafe  thread safe or not
  * @return
  */
-SHashObj *taosHashInit(size_t capacity, _hash_fn_t fn, bool threadsafe);
+SHashObj *taosHashInit(size_t capacity, _hash_fn_t fn, bool update, SHashLockTypeE type);
 
 /**
  * return the size of hash table
@@ -101,13 +110,19 @@ int32_t taosHashPut(SHashObj *pHashObj, const void *key, size_t keyLen, void *da
  */
 void *taosHashGet(SHashObj *pHashObj, const void *key, size_t keyLen);
 
+void *taosHashGetCB(SHashObj *pHashObj, const void *key, size_t keyLen, void(*fp)(void*));
+
 /**
  * remove item with the specified key
  * @param pHashObj
  * @param key
  * @param keyLen
  */
-void taosHashRemove(SHashObj *pHashObj, const void *key, size_t keyLen);
+int32_t taosHashRemove(SHashObj *pHashObj, const void *key, size_t keyLen);
+
+int32_t taosHashRemoveWithData(SHashObj *pHashObj, const void *key, size_t keyLen, void* data, size_t dsize);
+
+int32_t taosHashCondTraverse(SHashObj *pHashObj, bool (*fp)(void *, void *), void *param);
 
 /**
  * clean up hash table
@@ -115,13 +130,6 @@ void taosHashRemove(SHashObj *pHashObj, const void *key, size_t keyLen);
  */
 void taosHashCleanup(SHashObj *pHashObj);
 
-/**
- * Set the free callback function
- * This function if set will be invoked right before freeing each hash node
- * @param pHashObj
- */
-void taosHashSetFreecb(SHashObj *pHashObj, _hash_free_fn_t freeFp);
-
 /**
  *
  * @param pHashObj

src/util/inc/tcache.h

@@ -121,7 +121,7 @@ void *taosCacheAcquireByKey(SCacheObj *pCacheObj, const void *key, size_t keyLen
  * @param expireTime    new expire time of data
  * @return
  */ 
-void* taosCacheUpdateExpireTimeByName(SCacheObj *pCacheObj, void *key, size_t keyLen, uint64_t expireTime);
+//void* taosCacheUpdateExpireTimeByName(SCacheObj *pCacheObj, void *key, size_t keyLen, uint64_t expireTime);
 
 /**
  * Add one reference count for the exist data, and assign this data for a new owner.

src/util/inc/tstoken.h

@@ -24,8 +24,6 @@ extern "C" {
 #include "tutil.h"
 #include "ttokendef.h"
 
-
-
 #define TSQL_TBNAME   "TBNAME"
 #define TSQL_TBNAME_L "tbname"