Merge branch 'develop' into feature/hash

# Conflicts:
#	src/vnode/src/vnodeMain.c

.travis.yml

@@ -203,6 +203,29 @@ matrix:
           ;;
         esac
 
+  - os: linux
+    dist: trusty
+    language: c
+    git:
+      - depth: 1
+
+    addons:
+      apt:
+        packages:
+          - build-essential
+          - cmake
+    env: 
+      - DESC="trusty/gcc-4.8 build"
+
+    before_script:
+      - cd ${TRAVIS_BUILD_DIR}
+      - mkdir debug
+      - cd debug
+
+    script:
+      - cmake .. > /dev/null
+      - make > /dev/null
+
   - os: linux
     language: c
     compiler: clang

src/client/inc/tscSecondaryMerge.h

@@ -116,7 +116,7 @@ int32_t tscFlushTmpBuffer(tExtMemBuffer *pMemoryBuf, tOrderDescriptor *pDesc, tF
  * create local reducer to launch the second-stage reduce process at client site
  */
 void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrderDescriptor *pDesc,
-                           SColumnModel *finalModel, SSqlCmd *pSqlCmd, SSqlRes *pRes);
+                           SColumnModel *finalModel, SSqlObj* pSql);
 
 void tscDestroyLocalReducer(SSqlObj *pSql);
 

src/client/inc/tscUtil.h

@@ -31,13 +31,13 @@ extern "C" {
 #include "tscSecondaryMerge.h"
 #include "tsclient.h"
 
-#define UTIL_TABLE_IS_SUPERTABLE(metaInfo)  \
+#define UTIL_TABLE_IS_SUPER_TABLE(metaInfo)  \
   (((metaInfo)->pTableMeta != NULL) && ((metaInfo)->pTableMeta->tableType == TSDB_SUPER_TABLE))
 #define UTIL_TABLE_IS_CHILD_TABLE(metaInfo) \
   (((metaInfo)->pTableMeta != NULL) && ((metaInfo)->pTableMeta->tableType == TSDB_CHILD_TABLE))
   
-#define UTIL_TABLE_IS_NOMRAL_TABLE(metaInfo)\
+#define UTIL_TABLE_IS_NORMAL_TABLE(metaInfo)\
-  (!(UTIL_TABLE_IS_SUPERTABLE(metaInfo) || UTIL_TABLE_IS_CHILD_TABLE(metaInfo)))
+  (!(UTIL_TABLE_IS_SUPER_TABLE(metaInfo) || UTIL_TABLE_IS_CHILD_TABLE(metaInfo)))
 
 #define TSDB_COL_IS_TAG(f) (((f)&TSDB_COL_TAG) != 0)
 
@@ -183,7 +183,7 @@ void      tscSqlExprInfoDestroy(SArray* pExprInfo);
 
 SColumn* tscColumnClone(const SColumn* src);
 SColumn* tscColumnListInsert(SArray* pColList, SColumnIndex* colIndex);
-void tscColumnListCopy(SArray* dst, const SArray* src, int16_t tableIndex);
+SArray* tscColumnListClone(const SArray* src, int16_t tableIndex);
 void tscColumnListDestroy(SArray* pColList);
 
 SColumnFilterInfo* tscFilterInfoClone(const SColumnFilterInfo* src, int32_t numOfFilters);
@@ -265,6 +265,10 @@ void tscTryQueryNextVnode(SSqlObj *pSql, __async_cb_func_t fp);
 void tscAsyncQuerySingleRowForNextVnode(void *param, TAOS_RES *tres, int numOfRows);
 void tscTryQueryNextClause(SSqlObj* pSql, void (*queryFp)());
 
+void* malloc_throw(size_t size);
+void* calloc_throw(size_t nmemb, size_t size);
+char* strdup_throw(const char* str);
+
 #ifdef __cplusplus
 }
 #endif

src/client/inc/tsclient.h

@@ -438,6 +438,9 @@ extern int (*tscBuildMsg[TSDB_SQL_MAX])(SSqlObj *pSql, SSqlInfo *pInfo);
 
 typedef void (*__async_cb_func_t)(void *param, TAOS_RES *tres, int numOfRows);
 
+int32_t tscCompareTidTags(const void* p1, const void* p2);
+void tscBuildVgroupTableInfo(STableMetaInfo* pTableMetaInfo, SArray* tables);
+
 #ifdef __cplusplus
 }
 #endif

src/client/src/tscAsync.c

@@ -145,7 +145,7 @@ static void tscAsyncFetchRowsProxy(void *param, TAOS_RES *tres, int numOfRows) {
     return;
   }
   
-  // local reducer has handle this situation during super table non-projection query.
+  // local merge has handle this situation during super table non-projection query.
   if (pCmd->command != TSDB_SQL_RETRIEVE_LOCALMERGE) {
     pRes->numOfTotalInCurrentClause += pRes->numOfRows;
   }
@@ -222,9 +222,30 @@ void taos_fetch_rows_a(TAOS_RES *taosa, void (*fp)(void *, TAOS_RES *, int), voi
   tscResetForNextRetrieve(pRes);
   
   // handle the sub queries of join query
-  if (pCmd->command == TSDB_SQL_METRIC_JOIN_RETRIEVE) {
+  if (pCmd->command == TSDB_SQL_TABLE_JOIN_RETRIEVE) {
     tscFetchDatablockFromSubquery(pSql);
+  } else if (pRes->completed && pCmd->command == TSDB_SQL_FETCH) {
+    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
+       */
+      if (pCmd->clauseIndex < pCmd->numOfClause - 1) {
+        tscTryQueryNextClause(pSql, tscAsyncQueryRowsForNextVnode);
+        return;
+      }
+    
+      /*
+       * 1. has reach the limitation
+       * 2. no remain virtual nodes to be retrieved anymore
+       */
+      (*pSql->fetchFp)(param, pSql, 0);
+    }
+    return;
+  } else { // current query is not completed, continue retrieve from node
     if (pCmd->command != TSDB_SQL_RETRIEVE_LOCALMERGE && pCmd->command < TSDB_SQL_LOCAL) {
       pCmd->command = (pCmd->command > TSDB_SQL_MGMT) ? TSDB_SQL_RETRIEVE : TSDB_SQL_FETCH;
     }
@@ -425,6 +446,11 @@ void tscTableMetaCallBack(void *param, TAOS_RES *res, int code) {
   
     if ((pQueryInfo->type & TSDB_QUERY_TYPE_STABLE_SUBQUERY) == TSDB_QUERY_TYPE_STABLE_SUBQUERY) {
       STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
+      if (pTableMetaInfo->pTableMeta == NULL){
+        code = tscGetTableMeta(pSql, pTableMetaInfo);
+        assert(code == TSDB_CODE_SUCCESS);      
+      }     
+      
       assert((tscGetNumOfTags(pTableMetaInfo->pTableMeta) != 0) && pTableMetaInfo->vgroupIndex >= 0 && pSql->param != NULL);
 
       SRetrieveSupport *trs = (SRetrieveSupport *)pSql->param;
@@ -435,11 +461,6 @@ void tscTableMetaCallBack(void *param, TAOS_RES *res, int code) {
 
       tscTrace("%p get metricMeta during super table query successfully", pSql);      
 
-      code = tscGetTableMeta(pSql, pTableMetaInfo);
-      pRes->code = code;
-
-      if (code == TSDB_CODE_ACTION_IN_PROGRESS) return;
-
       code = tscGetSTableVgroupInfo(pSql, 0);
       pRes->code = code;
 
@@ -471,7 +492,7 @@ void tscTableMetaCallBack(void *param, TAOS_RES *res, int code) {
 
     if (code == TSDB_CODE_ACTION_IN_PROGRESS) return;
 
-    if (code == TSDB_CODE_SUCCESS && UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+    if (code == TSDB_CODE_SUCCESS && UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
       code = tscGetSTableVgroupInfo(pSql, pCmd->clauseIndex);
       pRes->code = code;
 

src/client/src/tscFunctionImpl.c

@@ -2904,7 +2904,11 @@ static void leastsquares_finalizer(SQLFunctionCtx *pCtx) {
   
   param[1][2] /= param[1][1];
   
-  sprintf(pCtx->aOutputBuf, "(%lf, %lf)", param[0][2], param[1][2]);
+  int32_t maxOutputSize = TSDB_AVG_FUNCTION_INTER_BUFFER_SIZE - VARSTR_HEADER_SIZE;
+  size_t n = snprintf(varDataVal(pCtx->aOutputBuf), maxOutputSize, "{slop:%.6lf, intercept:%.6lf}",
+      param[0][2], param[1][2]);
+  
+  varDataSetLen(pCtx->aOutputBuf, n);
   doFinalizer(pCtx);
 }
 

src/client/src/tscLocal.c

@@ -122,7 +122,7 @@ static int32_t tscSetValueToResObj(SSqlObj *pSql, int32_t rowLen) {
   int32_t numOfRows = tscGetNumOfColumns(pMeta);
   int32_t totalNumOfRows = numOfRows + tscGetNumOfTags(pMeta);
 
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     numOfRows = numOfRows + tscGetNumOfTags(pMeta);
   }
 
@@ -161,7 +161,7 @@ static int32_t tscSetValueToResObj(SSqlObj *pSql, int32_t rowLen) {
     }
   }
 
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     return 0;
   }
 

src/client/src/tscParseInsert.c

@@ -796,7 +796,7 @@ static int32_t tscCheckIfCreateTable(char **sqlstr, SSqlObj *pSql) {
       return code;
     }
 
-    if (!UTIL_TABLE_IS_SUPERTABLE(pSTableMeterMetaInfo)) {
+    if (!UTIL_TABLE_IS_SUPER_TABLE(pSTableMeterMetaInfo)) {
       return tscInvalidSQLErrMsg(pCmd->payload, "create table only from super table is allowed", sToken.z);
     }
 
@@ -1097,7 +1097,7 @@ int doParseInsertSql(SSqlObj *pSql, char *str) {
       goto _error_clean;       // TODO: should _clean or _error_clean to async flow ????
     }
 
-    if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+    if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
       code = tscInvalidSQLErrMsg(pCmd->payload, "insert data into super table is not supported", NULL);
       goto _error_clean;
     }

src/client/src/tscPrepare.c

@@ -194,7 +194,7 @@ static int normalStmtPrepare(STscStmt* stmt) {
 
 static char* normalStmtBuildSql(STscStmt* stmt) {
   SNormalStmt* normal = &stmt->normal;
-  SStringBuilder sb = {0};
+  SStringBuilder sb; memset(&sb, 0, sizeof(sb));
 
   if (taosStringBuilderSetJmp(&sb) != 0) {
     taosStringBuilderDestroy(&sb);

src/client/src/tscSQLParser.c

@@ -1297,10 +1297,6 @@ SSqlExpr* doAddProjectCol(SQueryInfo* pQueryInfo, int32_t outputIndex, int32_t c
       pSchema->bytes, functionId == TSDB_FUNC_TAGPRJ);
 }
 
-void addRequiredTagColumn(STableMetaInfo* pTableMetaInfo, SColumnIndex* index) {
-
-}
-
 static void addProjectQueryCol(SQueryInfo* pQueryInfo, int32_t startPos, SColumnIndex* pIndex, tSQLExprItem* pItem) {
   SSqlExpr* pExpr = doAddProjectCol(pQueryInfo, startPos, pIndex->columnIndex, pIndex->tableIndex);
 
@@ -1352,7 +1348,7 @@ static int32_t doAddProjectionExprAndResultFields(SQueryInfo* pQueryInfo, SColum
 
   STableComInfo tinfo = tscGetTableInfo(pTableMeta);
   
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     numOfTotalColumns = tinfo.numOfColumns + tinfo.numOfTags;
   } else {
     numOfTotalColumns = tinfo.numOfColumns;
@@ -1412,7 +1408,7 @@ int32_t addProjectionExprAndResultField(SQueryInfo* pQueryInfo, tSQLExprItem* pI
       STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, index.tableIndex);
       STableMeta*     pTableMeta = pTableMetaInfo->pTableMeta;
 
-      if (index.columnIndex >= tscGetNumOfColumns(pTableMeta) && UTIL_TABLE_IS_NOMRAL_TABLE(pTableMetaInfo)) {
+      if (index.columnIndex >= tscGetNumOfColumns(pTableMeta) && UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo)) {
         return invalidSqlErrMsg(pQueryInfo->msg, msg1);
       }
 
@@ -1866,7 +1862,7 @@ int32_t addExprAndResultField(SQueryInfo* pQueryInfo, int32_t colIndex, tSQLExpr
     
     case TK_TBID: {
       pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-      if (UTIL_TABLE_IS_NOMRAL_TABLE(pTableMetaInfo)) {
+      if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo)) {
         return invalidSqlErrMsg(pQueryInfo->msg, msg7);
       }
     
@@ -2283,7 +2279,7 @@ bool validateIpAddress(const char* ip, size_t size) {
 int32_t tscTansformSQLFuncForSTableQuery(SQueryInfo* pQueryInfo) {
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
 
-  if (pTableMetaInfo->pTableMeta == NULL || !UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (pTableMetaInfo->pTableMeta == NULL || !UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     return TSDB_CODE_INVALID_SQL;
   }
 
@@ -2322,7 +2318,7 @@ int32_t tscTansformSQLFuncForSTableQuery(SQueryInfo* pQueryInfo) {
 /* transfer the field-info back to original input format */
 void tscRestoreSQLFuncForSTableQuery(SQueryInfo* pQueryInfo) {
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-  if (!UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (!UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     return;
   }
   
@@ -2546,7 +2542,7 @@ int32_t parseGroupbyClause(SQueryInfo* pQueryInfo, tVariantList* pList, SSqlCmd*
     }
     
     if (groupTag) {
-      if (!UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+      if (!UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
         return invalidSqlErrMsg(pQueryInfo->msg, msg9);
       }
 
@@ -3258,7 +3254,7 @@ static bool validateJoinExprNode(SQueryInfo* pQueryInfo, tSQLExpr* pExpr, SColum
   }
 
   // table to table/ super table to super table are allowed
-  if (UTIL_TABLE_IS_SUPERTABLE(pLeftMeterMeta) != UTIL_TABLE_IS_SUPERTABLE(pRightMeterMeta)) {
+  if (UTIL_TABLE_IS_SUPER_TABLE(pLeftMeterMeta) != UTIL_TABLE_IS_SUPER_TABLE(pRightMeterMeta)) {
     invalidSqlErrMsg(pQueryInfo->msg, msg5);
     return false;
   }
@@ -3341,7 +3337,7 @@ static int32_t handleExprInQueryCond(SQueryInfo* pQueryInfo, tSQLExpr** pExpr, S
   } else if (index.columnIndex >= tscGetNumOfColumns(pTableMeta) ||
              index.columnIndex == TSDB_TBNAME_COLUMN_INDEX) {  // query on tags
     // check for tag query condition
-    if (UTIL_TABLE_IS_NOMRAL_TABLE(pTableMetaInfo)) {
+    if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo)) {
       return invalidSqlErrMsg(pQueryInfo->msg, msg1);
     }
 
@@ -3562,7 +3558,7 @@ static int32_t setTableCondForSTableQuery(SQueryInfo* pQueryInfo, const char* ac
     return TSDB_CODE_SUCCESS;
   }
 
-  SStringBuilder sb1 = {0};
+  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};
@@ -3701,7 +3697,7 @@ static int32_t validateJoinExpr(SQueryInfo* pQueryInfo, SCondExpr* pCondExpr) {
   }
 
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {  // for stable join, tag columns
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {  // for stable join, tag columns
                                                    // must be present for join
     if (pCondExpr->pJoinExpr == NULL) {
       return invalidSqlErrMsg(pQueryInfo->msg, msg1);
@@ -3739,7 +3735,7 @@ static void cleanQueryExpr(SCondExpr* pCondExpr) {
 
 static void doAddJoinTagsColumnsIntoTagList(SQueryInfo* pQueryInfo, SCondExpr* pCondExpr) {
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-  if (QUERY_IS_JOIN_QUERY(pQueryInfo->type) && UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (QUERY_IS_JOIN_QUERY(pQueryInfo->type) && UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     SColumnIndex index = {0};
 
     getColumnIndexByName(&pCondExpr->pJoinExpr->pLeft->colInfo, pQueryInfo, &index);
@@ -3796,6 +3792,8 @@ static int32_t getTagQueryCondExpr(SQueryInfo* pQueryInfo, SCondExpr* pCondExpr,
     
     tSQLExprDestroy(p1);
     tExprTreeDestroy(&p, NULL);
+    
+    taosArrayDestroy(colList);
   }
 
   pCondExpr->pTagCond = NULL;
@@ -3815,7 +3813,7 @@ int32_t parseWhereClause(SQueryInfo* pQueryInfo, tSQLExpr** pExpr, SSqlObj* pSql
   pQueryInfo->window.ekey = INT64_MAX;
 
   // tags query condition may be larger than 512bytes, therefore, we need to prepare enough large space
-  SStringBuilder sb = {0};
+  SStringBuilder sb; memset(&sb, 0, sizeof(sb));
   SCondExpr      condExpr = {0};
 
   if ((*pExpr)->pLeft == NULL || (*pExpr)->pRight == NULL) {
@@ -4104,7 +4102,7 @@ static void setDefaultOrderInfo(SQueryInfo* pQueryInfo) {
   }
 
   /* for super table query, set default ascending order for group output */
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     pQueryInfo->groupbyExpr.orderType = TSDB_ORDER_ASC;
   }
 }
@@ -4130,7 +4128,7 @@ int32_t parseOrderbyClause(SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql, SSchema
    *
    * for super table query, the order option must be less than 3.
    */
-  if (UTIL_TABLE_IS_NOMRAL_TABLE(pTableMetaInfo)) {
+  if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo)) {
     if (pSortorder->nExpr > 1) {
       return invalidSqlErrMsg(pQueryInfo->msg, msg0);
     }
@@ -4151,7 +4149,7 @@ int32_t parseOrderbyClause(SQueryInfo* pQueryInfo, SQuerySQL* pQuerySql, SSchema
   SSQLToken    columnName = {pVar->nLen, pVar->nType, pVar->pz};
   SColumnIndex index = {0};
 
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {  // super table query
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {  // super table query
     if (getColumnIndexByName(&columnName, pQueryInfo, &index) != TSDB_CODE_SUCCESS) {
       return invalidSqlErrMsg(pQueryInfo->msg, msg1);
     }
@@ -4304,10 +4302,10 @@ int32_t setAlterTableInfo(SSqlObj* pSql, struct SSqlInfo* pInfo) {
 
   if (pAlterSQL->type == TSDB_ALTER_TABLE_ADD_TAG_COLUMN || pAlterSQL->type == TSDB_ALTER_TABLE_DROP_TAG_COLUMN ||
       pAlterSQL->type == TSDB_ALTER_TABLE_CHANGE_TAG_COLUMN) {
-    if (UTIL_TABLE_IS_NOMRAL_TABLE(pTableMetaInfo)) {
+    if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo)) {
       return invalidSqlErrMsg(pQueryInfo->msg, msg3);
     }
-  } else if ((pAlterSQL->type == TSDB_ALTER_TABLE_UPDATE_TAG_VAL) && (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo))) {
+  } else if ((pAlterSQL->type == TSDB_ALTER_TABLE_UPDATE_TAG_VAL) && (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo))) {
     return invalidSqlErrMsg(pQueryInfo->msg, msg4);
   } else if ((pAlterSQL->type == TSDB_ALTER_TABLE_ADD_COLUMN || pAlterSQL->type == TSDB_ALTER_TABLE_DROP_COLUMN) &&
              UTIL_TABLE_IS_CHILD_TABLE(pTableMetaInfo)) {
@@ -4693,7 +4691,7 @@ int32_t parseLimitClause(SQueryInfo* pQueryInfo, int32_t clauseIndex, SQuerySQL*
   }
 
   // todo refactor
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     if (!tscQueryTags(pQueryInfo)) {  // local handle the super table tag query
       if (tscIsProjectionQueryOnSTable(pQueryInfo, 0)) {
         if (pQueryInfo->slimit.limit > 0 || pQueryInfo->slimit.offset > 0) {
@@ -5629,7 +5627,7 @@ int32_t doCheckForStream(SSqlObj* pSql, SSqlInfo* pInfo) {
     return code;
   }
 
-  bool isSTable = UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo);
+  bool isSTable = UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo);
   if (parseSelectClause(&pSql->cmd, 0, pQuerySql->pSelection, isSTable) != TSDB_CODE_SUCCESS) {
     return TSDB_CODE_INVALID_SQL;
   }
@@ -5773,7 +5771,7 @@ int32_t doCheckForQuery(SSqlObj* pSql, SQuerySQL* pQuerySql, int32_t index) {
   assert(pQueryInfo->numOfTables == pQuerySql->from->nExpr);
   bool isSTable = false;
   
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     isSTable = true;
     code = tscGetSTableVgroupInfo(pSql, index);
     if (code != TSDB_CODE_SUCCESS) {

src/client/src/tscSecondaryMerge.c

@@ -55,7 +55,7 @@ int32_t treeComparator(const void *pLeft, const void *pRight, void *param) {
   }
 }
 
-static void tscInitSqlContext(SSqlCmd *pCmd, SSqlRes *pRes, SLocalReducer *pReducer, tOrderDescriptor *pDesc) {
+static void tscInitSqlContext(SSqlCmd *pCmd, SLocalReducer *pReducer, tOrderDescriptor *pDesc) {
   /*
    * the fields and offset attributes in pCmd and pModel may be different due to
    * merge requirement. So, the final result in pRes structure is formatted in accordance with the pCmd object.
@@ -132,16 +132,15 @@ static void tscInitSqlContext(SSqlCmd *pCmd, SSqlRes *pRes, SLocalReducer *pRedu
   }
 }
 
-/*
- * todo release allocated memory process with async process
- */
 void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrderDescriptor *pDesc,
-                           SColumnModel *finalmodel, SSqlCmd *pCmd, SSqlRes *pRes) {
+                           SColumnModel *finalmodel, SSqlObj* pSql) {
-  // offset of cmd in SSqlObj structure
+  SSqlCmd* pCmd = &pSql->cmd;
-  char *pSqlObjAddr = (char *)pCmd - offsetof(SSqlObj, cmd);
+  SSqlRes* pRes = &pSql->res;
   
   if (pMemBuffer == NULL) {
-    tscError("%p pMemBuffer", pMemBuffer);
+    tscLocalReducerEnvDestroy(pMemBuffer, pDesc, finalmodel, numOfBuffer);
+  
+    tscError("%p pMemBuffer is NULL", pMemBuffer);
     pRes->code = TSDB_CODE_APP_ERROR;
     return;
   }
@@ -149,7 +148,7 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
   if (pDesc->pColumnModel == NULL) {
     tscLocalReducerEnvDestroy(pMemBuffer, pDesc, finalmodel, numOfBuffer);
 
-    tscError("%p no local buffer or intermediate result format model", pSqlObjAddr);
+    tscError("%p no local buffer or intermediate result format model", pSql);
     pRes->code = TSDB_CODE_APP_ERROR;
     return;
   }
@@ -158,7 +157,7 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
   for (int32_t i = 0; i < numOfBuffer; ++i) {
     int32_t len = pMemBuffer[i]->fileMeta.flushoutData.nLength;
     if (len == 0) {
-      tscTrace("%p no data retrieved from orderOfVnode:%d", pSqlObjAddr, i + 1);
+      tscTrace("%p no data retrieved from orderOfVnode:%d", pSql, i + 1);
       continue;
     }
 
@@ -167,13 +166,13 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
 
   if (numOfFlush == 0 || numOfBuffer == 0) {
     tscLocalReducerEnvDestroy(pMemBuffer, pDesc, finalmodel, numOfBuffer);
-    tscTrace("%p retrieved no data", pSqlObjAddr);
+    tscTrace("%p retrieved no data", pSql);
 
     return;
   }
 
   if (pDesc->pColumnModel->capacity >= pMemBuffer[0]->pageSize) {
-    tscError("%p Invalid value of buffer capacity %d and page size %d ", pSqlObjAddr, pDesc->pColumnModel->capacity,
+    tscError("%p Invalid value of buffer capacity %d and page size %d ", pSql, pDesc->pColumnModel->capacity,
              pMemBuffer[0]->pageSize);
 
     tscLocalReducerEnvDestroy(pMemBuffer, pDesc, finalmodel, numOfBuffer);
@@ -181,10 +180,11 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
     return;
   }
 
-  size_t         nReducerSize = sizeof(SLocalReducer) + sizeof(void *) * numOfFlush;
+  size_t size = sizeof(SLocalReducer) + POINTER_BYTES * numOfFlush;
-  SLocalReducer *pReducer = (SLocalReducer *)calloc(1, nReducerSize);
+  
+  SLocalReducer *pReducer = (SLocalReducer *) calloc(1, size);
   if (pReducer == NULL) {
-    tscError("%p failed to create merge structure", pSqlObjAddr);
+    tscError("%p failed to create local merge structure, out of memory", pSql);
 
     tscLocalReducerEnvDestroy(pMemBuffer, pDesc, finalmodel, numOfBuffer);
     pRes->code = TSDB_CODE_CLI_OUT_OF_MEMORY;
@@ -199,48 +199,52 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
   pReducer->numOfVnode = numOfBuffer;
 
   pReducer->pDesc = pDesc;
-  tscTrace("%p the number of merged leaves is: %d", pSqlObjAddr, pReducer->numOfBuffer);
+  tscTrace("%p the number of merged leaves is: %d", pSql, pReducer->numOfBuffer);
 
   int32_t idx = 0;
   for (int32_t i = 0; i < numOfBuffer; ++i) {
     int32_t numOfFlushoutInFile = pMemBuffer[i]->fileMeta.flushoutData.nLength;
 
     for (int32_t j = 0; j < numOfFlushoutInFile; ++j) {
-      SLocalDataSource *pDS = (SLocalDataSource *)malloc(sizeof(SLocalDataSource) + pMemBuffer[0]->pageSize);
+      SLocalDataSource *ds = (SLocalDataSource *)malloc(sizeof(SLocalDataSource) + pMemBuffer[0]->pageSize);
-      if (pDS == NULL) {
+      if (ds == NULL) {
-        tscError("%p failed to create merge structure", pSqlObjAddr);
+        tscError("%p failed to create merge structure", pSql);
         pRes->code = TSDB_CODE_CLI_OUT_OF_MEMORY;
         return;
       }
-      pReducer->pLocalDataSrc[idx] = pDS;
       
-      pDS->pMemBuffer = pMemBuffer[i];
+      pReducer->pLocalDataSrc[idx] = ds;
-      pDS->flushoutIdx = j;
-      pDS->filePage.numOfElems = 0;
-      pDS->pageId = 0;
-      pDS->rowIdx = 0;
 
-      tscTrace("%p load data from disk into memory, orderOfVnode:%d, total:%d", pSqlObjAddr, i + 1, idx + 1);
+      ds->pMemBuffer = pMemBuffer[i];
-      tExtMemBufferLoadData(pMemBuffer[i], &(pDS->filePage), j, 0);
+      ds->flushoutIdx = j;
+      ds->filePage.numOfElems = 0;
+      ds->pageId = 0;
+      ds->rowIdx = 0;
+
+      tscTrace("%p load data from disk into memory, orderOfVnode:%d, total:%d", pSql, i + 1, idx + 1);
+      tExtMemBufferLoadData(pMemBuffer[i], &(ds->filePage), j, 0);
 #ifdef _DEBUG_VIEW
-      printf("load data page into mem for build loser tree: %" PRIu64 " rows\n", pDS->filePage.numOfElems);
+      printf("load data page into mem for build loser tree: %" PRIu64 " rows\n", ds->filePage.numOfElems);
       SSrcColumnInfo colInfo[256] = {0};
       SQueryInfo *   pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
 
       tscGetSrcColumnInfo(colInfo, pQueryInfo);
 
-      tColModelDisplayEx(pDesc->pColumnModel, pDS->filePage.data, pDS->filePage.numOfElems,
+      tColModelDisplayEx(pDesc->pColumnModel, ds->filePage.data, ds->filePage.numOfElems,
                          pMemBuffer[0]->numOfElemsPerPage, colInfo);
 #endif
-      if (pDS->filePage.numOfElems == 0) {  // no data in this flush
+      
-        tscTrace("%p flush data is empty, ignore %d flush record", pSqlObjAddr, idx);
+      if (ds->filePage.numOfElems == 0) {  // no data in this flush, the index does not increase
-        tfree(pDS);
+        tscTrace("%p flush data is empty, ignore %d flush record", pSql, idx);
+        tfree(ds);
         continue;
       }
+      
       idx += 1;
     }
   }
-  assert(idx >= pReducer->numOfBuffer);
+  
+  // no data actually, no need to merge result.
   if (idx == 0) {
     return;
   }
@@ -262,9 +266,7 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
 
   // the input data format follows the old format, but output in a new format.
   // so, all the input must be parsed as old format
-  size_t size = tscSqlExprNumOfExprs(pQueryInfo);
+  pReducer->pCtx = (SQLFunctionCtx *)calloc(tscSqlExprNumOfExprs(pQueryInfo), sizeof(SQLFunctionCtx));
-  
-  pReducer->pCtx = (SQLFunctionCtx *)calloc(size, sizeof(SQLFunctionCtx));
   pReducer->rowSize = pMemBuffer[0]->nElemSize;
 
   tscRestoreSQLFuncForSTableQuery(pQueryInfo);
@@ -281,7 +283,6 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
 
   // used to keep the latest input row
   pReducer->pTempBuffer = (tFilePage *)calloc(1, pReducer->rowSize + sizeof(tFilePage));
-
   pReducer->discardData = (tFilePage *)calloc(1, pReducer->rowSize + sizeof(tFilePage));
   pReducer->discard = false;
 
@@ -309,11 +310,13 @@ void tscCreateLocalReducer(tExtMemBuffer **pMemBuffer, int32_t numOfBuffer, tOrd
     return;
   }
 
+  size = tscSqlExprNumOfExprs(pQueryInfo);
+  
   pReducer->pTempBuffer->numOfElems = 0;
   pReducer->pResInfo = calloc(size, sizeof(SResultInfo));
 
   tscCreateResPointerInfo(pRes, pQueryInfo);
-  tscInitSqlContext(pCmd, pRes, pReducer, pDesc);
+  tscInitSqlContext(pCmd, pReducer, pDesc);
 
   // we change the capacity of schema to denote that there is only one row in temp buffer
   pReducer->pDesc->pColumnModel->capacity = 1;
@@ -428,8 +431,7 @@ int32_t saveToBuffer(tExtMemBuffer *pMemoryBuf, tOrderDescriptor *pDesc, tFilePa
     tColModelAppend(pModel, pPage, data, numOfRows - remain, numOfWriteElems, numOfRows);
 
     if (pPage->numOfElems == pModel->capacity) {
-      int32_t ret = tscFlushTmpBuffer(pMemoryBuf, pDesc, pPage, orderType);
+      if (tscFlushTmpBuffer(pMemoryBuf, pDesc, pPage, orderType) != TSDB_CODE_SUCCESS) {
-      if (ret != 0) {
         return -1;
       }
     } else {

src/client/src/tscServer.c

@@ -39,7 +39,7 @@ int (*tscProcessMsgRsp[TSDB_SQL_MAX])(SSqlObj *pSql);
 void tscProcessActivityTimer(void *handle, void *tmrId);
 int tscKeepConn[TSDB_SQL_MAX] = {0};
 
-TSKEY tscGetSubscriptionProgress(void* sub, int64_t uid);
+TSKEY tscGetSubscriptionProgress(void* sub, int64_t uid, TSKEY dflt);
 void tscUpdateSubscriptionProgress(void* sub, int64_t uid, TSKEY ts);
 void tscSaveSubscriptionProgress(void* sub);
 
@@ -500,7 +500,7 @@ int tscBuildFetchMsg(SSqlObj *pSql, SSqlInfo *pInfo) {
   // todo valid the vgroupId at the client side
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
   
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     int32_t vgIndex = pTableMetaInfo->vgroupIndex;
     
     SVgroupsInfo* pVgroupInfo = pTableMetaInfo->vgroupList;
@@ -550,8 +550,7 @@ int tscBuildSubmitMsg(SSqlObj *pSql, SSqlInfo *pInfo) {
 }
 
 /*
- * for meter query, simply return the size <= 1k
+ * for table query, simply return the size <= 1k
- * for metric query, estimate size according to meter tags
  */
 static int32_t tscEstimateQueryMsgSize(SSqlCmd *pCmd, int32_t clauseIndex) {
   const static int32_t MIN_QUERY_MSG_PKT_SIZE = TSDB_MAX_BYTES_PER_ROW * 5;
@@ -562,25 +561,18 @@ static int32_t tscEstimateQueryMsgSize(SSqlCmd *pCmd, int32_t clauseIndex) {
   size_t numOfExprs = tscSqlExprNumOfExprs(pQueryInfo);
   int32_t exprSize = sizeof(SSqlFuncMsg) * numOfExprs;
   
-  //STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-
-  // table query without tags values
-  //if (!UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
   return MIN_QUERY_MSG_PKT_SIZE + minMsgSize() + sizeof(SQueryTableMsg) + srcColListSize + exprSize + 4096;
-  //}
-  
-  //int32_t size = 4096;
-  //return size;
 }
 
 static char *doSerializeTableInfo(SQueryTableMsg* pQueryMsg, SSqlObj *pSql, char *pMsg) {
   STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, pSql->cmd.clauseIndex, 0);
+  TSKEY dfltKey = htobe64(pQueryMsg->window.skey);
 
   STableMeta * pTableMeta = pTableMetaInfo->pTableMeta;
-  if (UTIL_TABLE_IS_NOMRAL_TABLE(pTableMetaInfo) || pTableMetaInfo->pVgroupTables == NULL) {
+  if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo) || pTableMetaInfo->pVgroupTables == NULL) {
     
     SCMVgroupInfo* pVgroupInfo = NULL;
-    if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+    if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
       int32_t index = pTableMetaInfo->vgroupIndex;
       assert(index >= 0);
   
@@ -596,7 +588,7 @@ static char *doSerializeTableInfo(SQueryTableMsg* pQueryMsg, SSqlObj *pSql, char
     STableIdInfo *pTableIdInfo = (STableIdInfo *)pMsg;
     pTableIdInfo->tid = htonl(pTableMeta->sid);
     pTableIdInfo->uid = htobe64(pTableMeta->uid);
-    pTableIdInfo->key = htobe64(tscGetSubscriptionProgress(pSql->pSubscription, pTableMeta->uid));
+    pTableIdInfo->key = htobe64(tscGetSubscriptionProgress(pSql->pSubscription, pTableMeta->uid, dfltKey));
 
     pQueryMsg->numOfTables = htonl(1);  // set the number of tables
 
@@ -624,7 +616,7 @@ static char *doSerializeTableInfo(SQueryTableMsg* pQueryMsg, SSqlObj *pSql, char
       STableIdInfo *pTableIdInfo = (STableIdInfo *)pMsg;
       pTableIdInfo->tid = htonl(pItem->tid);
       pTableIdInfo->uid = htobe64(pItem->uid);
-//      pTableIdInfo->key = htobe64(tscGetSubscriptionProgress(pSql->pSubscription, pTableMeta->uid));
+      pTableIdInfo->key = htobe64(tscGetSubscriptionProgress(pSql->pSubscription, pItem->uid, dfltKey));
       pMsg += sizeof(STableIdInfo);
     }
   }
@@ -1453,8 +1445,9 @@ int tscProcessRetrieveLocalMergeRsp(SSqlObj *pSql) {
   }
 
   pRes->row = 0;
+  pRes->completed = (pRes->numOfRows == 0);
 
-  uint8_t code = pRes->code;
+  int32_t code = pRes->code;
   if (pRes->code == TSDB_CODE_SUCCESS) {
     (*pSql->fp)(pSql->param, pSql, pRes->numOfRows);
   } else {
@@ -2292,7 +2285,7 @@ int tscProcessAlterTableMsgRsp(SSqlObj *pSql) {
   taosCacheRelease(tscCacheHandle, (void **)&pTableMeta, true);
 
   if (pTableMetaInfo->pTableMeta) {
-    bool isSuperTable = UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo);
+    bool isSuperTable = UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo);
 
     taosCacheRelease(tscCacheHandle, (void **)&(pTableMetaInfo->pTableMeta), true);
 //    taosCacheRelease(tscCacheHandle, (void **)&(pTableMetaInfo->pMetricMeta), true);
@@ -2354,6 +2347,7 @@ int tscProcessRetrieveRspFromNode(SSqlObj *pSql) {
     for (int i = 0; i < numOfTables; i++) {
       int64_t uid = htobe64(*(int64_t*)p);
       p += sizeof(int64_t);
+      p += sizeof(int32_t); // skip tid
       TSKEY key = htobe64(*(TSKEY*)p);
       p += sizeof(TSKEY);
       tscUpdateSubscriptionProgress(pSql->pSubscription, uid, key);

src/client/src/tscSql.c

@@ -463,11 +463,11 @@ TAOS_ROW taos_fetch_row(TAOS_RES *res) {
     return NULL;
   }
   
-  // current data are exhausted, fetch more data
+  // current data set are exhausted, fetch more data from node
-  if (pRes->row >= pRes->numOfRows && pRes->completed != true &&
+  if (pRes->row >= pRes->numOfRows && (pRes->completed != true || hasMoreVnodesToTry(pSql)) &&
       (pCmd->command == TSDB_SQL_RETRIEVE ||
        pCmd->command == TSDB_SQL_RETRIEVE_LOCALMERGE ||
-       pCmd->command == TSDB_SQL_METRIC_JOIN_RETRIEVE ||
+       pCmd->command == TSDB_SQL_TABLE_JOIN_RETRIEVE ||
        pCmd->command == TSDB_SQL_FETCH ||
        pCmd->command == TSDB_SQL_SHOW ||
        pCmd->command == TSDB_SQL_SELECT ||

src/client/src/tscStream.c

@@ -79,7 +79,7 @@ static void tscProcessStreamLaunchQuery(SSchedMsg *pMsg) {
 
   if (code == TSDB_CODE_ACTION_IN_PROGRESS) return;
 
-  if (code == 0 && UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (code == 0 && UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     code = tscGetSTableVgroupInfo(pSql, 0);
     pSql->res.code = code;
 

src/client/src/tscSub.c

@@ -42,8 +42,7 @@ typedef struct SSub {
   int                     interval;
   TAOS_SUBSCRIBE_CALLBACK fp;
   void *                  param;
-  int                     numOfTables;
+  SArray* progress;
-  SSubscriptionProgress * progress;
 } SSub;
 
 
@@ -55,92 +54,113 @@ static int tscCompareSubscriptionProgress(const void* a, const void* b) {
   return 0;
 }
 
-TSKEY tscGetSubscriptionProgress(void* sub, int64_t uid) {
+TSKEY tscGetSubscriptionProgress(void* sub, int64_t uid, TSKEY dflt) {
-  if (sub == NULL)
+  if (sub == NULL) {
-    return 0;
+    return dflt;
-
-  SSub* pSub = (SSub*)sub;
-  for (int s = 0, e = pSub->numOfTables; s < e;) {
-    int m = (s + e) / 2;
-    SSubscriptionProgress* p = pSub->progress + m;
-    if (p->uid > uid)
-      e = m;
-    else if (p->uid < uid)
-      s = m + 1;
-    else
-      return p->key;
   }
+  SSub* pSub = (SSub*)sub;
 
-  return 0;
+  SSubscriptionProgress target = {.uid = uid, .key = 0};
+  SSubscriptionProgress* p = taosArraySearch(pSub->progress, tscCompareSubscriptionProgress, &target);
+  if (p == NULL) {
+    return dflt;
+  }
+  return p->key;
 }
 
 void tscUpdateSubscriptionProgress(void* sub, int64_t uid, TSKEY ts) {
   if( sub == NULL)
     return;
-
   SSub* pSub = (SSub*)sub;
-  for (int s = 0, e = pSub->numOfTables; s < e;) {
+
-    int m = (s + e) / 2;
+  SSubscriptionProgress target = {.uid = uid, .key = ts};
-    SSubscriptionProgress* p = pSub->progress + m;
+  SSubscriptionProgress* p = taosArraySearch(pSub->progress, tscCompareSubscriptionProgress, &target);
-    if (p->uid > uid)
+  if (p != NULL) {
-      e = m;
+    p->key = ts;
-    else if (p->uid < uid)
-      s = m + 1;
-    else {
-      if (ts >= p->key) p->key = ts;
-      break;
   }
 }
+
+
+static void asyncCallback(void *param, TAOS_RES *tres, int code) {
+  assert(param != NULL);
+  SSqlObj *pSql = ((SSqlObj *)param);
+  
+  pSql->res.code = code;
+  sem_post(&pSql->rspSem);
 }
 
 
 static SSub* tscCreateSubscription(STscObj* pObj, const char* topic, const char* sql) {
-  SSub* pSub = calloc(1, sizeof(SSub));
+  SSub* pSub = NULL;
-  if (pSub == NULL) {
-    terrno = TSDB_CODE_CLI_OUT_OF_MEMORY;
-    tscError("failed to allocate memory for subscription");
-    return NULL;
-  }
 
-  SSqlObj* pSql = calloc(1, sizeof(SSqlObj));
+  TRY( 8 ) {
-  if (pSql == NULL) {
+    SSqlObj* pSql = calloc_throw(1, sizeof(SSqlObj));
-    terrno = TSDB_CODE_CLI_OUT_OF_MEMORY;
+    CLEANUP_PUSH_FREE(true, pSql);
-    tscError("failed to allocate SSqlObj for subscription");
+    SSqlCmd *pCmd = &pSql->cmd;
-    goto _pSql_failed;
+    SSqlRes *pRes = &pSql->res;
+
+    if (tsem_init(&pSql->rspSem, 0, 0) == -1) {
+      THROW(TAOS_SYSTEM_ERROR(errno));
     }
+    CLEANUP_PUSH_INT_PTR(true, tsem_destroy, &pSql->rspSem);
 
     pSql->signature = pSql;
+    pSql->param = pSql;
     pSql->pTscObj = pObj;
+    pSql->maxRetry = TSDB_MAX_REPLICA_NUM;
+    pSql->fp = asyncCallback;
 
-  char* sqlstr = (char*)malloc(strlen(sql) + 1);
+    int code = tscAllocPayload(pCmd, TSDB_DEFAULT_PAYLOAD_SIZE);
-  if (sqlstr == NULL) {
+    if (code != TSDB_CODE_SUCCESS) {
-    tscError("failed to allocate sql string for subscription");
+      THROW(code);
-    goto failed;
     }
-  strcpy(sqlstr, sql);
+    CLEANUP_PUSH_FREE(true, pCmd->payload);
-  strtolower(sqlstr, sqlstr);
-  pSql->sqlstr = sqlstr;
 
-  tsem_init(&pSql->rspSem, 0, 0);
+    pRes->qhandle = 0;
-
-  SSqlRes *pRes = &pSql->res;
     pRes->numOfRows = 1;
-  pRes->numOfTotal = 0;
 
+    pSql->sqlstr = strdup_throw(sql);
+    CLEANUP_PUSH_FREE(true, pSql->sqlstr);
+    strtolower(pSql->sqlstr, pSql->sqlstr);
+
+    code = tsParseSql(pSql, false);
+    if (code == TSDB_CODE_ACTION_IN_PROGRESS) {
+      // wait for the callback function to post the semaphore
+      sem_wait(&pSql->rspSem);
+      code = pSql->res.code;
+    }
+    if (code != TSDB_CODE_SUCCESS) {
+      tscError("failed to parse sql statement: %s, error: %s", pSub->topic, tstrerror(code));
+      THROW( code );
+    }
+
+    if (pSql->cmd.command != TSDB_SQL_SELECT) {
+      tscError("only 'select' statement is allowed in subscription: %s", pSub->topic);
+      THROW( -1 );  // TODO
+    }
+
+    pSub = calloc_throw(1, sizeof(SSub));
+    CLEANUP_PUSH_FREE(true, pSub);
     pSql->pSubscription = pSub;
     pSub->pSql = pSql;
     pSub->signature = pSub;
     strncpy(pSub->topic, topic, sizeof(pSub->topic));
     pSub->topic[sizeof(pSub->topic) - 1] = 0;
+    pSub->progress = taosArrayInit(32, sizeof(SSubscriptionProgress));
+    if (pSub->progress == NULL) {
+      THROW(TSDB_CODE_CLI_OUT_OF_MEMORY);
+    }
+
+    CLEANUP_EXECUTE();
+
+  } CATCH( code ) {
+    tscError("failed to create subscription object: %s", tstrerror(code));
+    CLEANUP_EXECUTE();
+    pSub = NULL;
+
+  } END_TRY
+
   return pSub;
-
-failed:
-  tfree(sqlstr);
-
-_pSql_failed:
-  tfree(pSql);
-  tfree(pSub);
-  return NULL;
 }
 
 
@@ -157,61 +177,69 @@ static void tscProcessSubscriptionTimer(void *handle, void *tmrId) {
 }
 
 
-int tscUpdateSubscription(STscObj* pObj, SSub* pSub) {
+static SArray* getTableList( SSqlObj* pSql ) {
-  int code = (uint8_t)tsParseSql(pSub->pSql, false);
+  const char* p = strstr( pSql->sqlstr, " from " );
+  char* sql = alloca(strlen(p) + 32);
+  sprintf(sql, "select tbid(tbname)%s", p);
+  int code = taos_query( pSql->pTscObj, sql );
   if (code != TSDB_CODE_SUCCESS) {
-    tscError("failed to parse sql statement: %s", pSub->topic);
+    tscError("failed to retrieve table id: %s", tstrerror(code));
-    return 0;
+    return NULL;
   }
 
-  SSqlCmd* pCmd = &pSub->pSql->cmd;
+  TAOS_RES* res = taos_use_result( pSql->pTscObj );
-  if (pCmd->command != TSDB_SQL_SELECT) {
+  TAOS_ROW row;
-    tscError("only 'select' statement is allowed in subscription: %s", pSub->topic);
+  SArray* result = taosArrayInit( 128, sizeof(STidTags) );
-    return 0;
+  while ((row = taos_fetch_row(res))) {
+    STidTags tags;
+    memcpy(&tags, row[0], sizeof(tags));
+    taosArrayPush(result, &tags);
   }
 
-  STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, 0, 0);
+  return result;
-  int numOfTables = 0;
-  if (!UTIL_TABLE_IS_NOMRAL_TABLE(pTableMetaInfo)) {
-//    SSuperTableMeta* pMetricMeta = pTableMetaInfo->pMetricMeta;
-//    for (int32_t i = 0; i < pMetricMeta->numOfVnodes; i++) {
-//      SVnodeSidList *pVnodeSidList = tscGetVnodeSidList(pMetricMeta, i);
-//      numOfTables += pVnodeSidList->numOfSids;
-//    }
 }
 
-  SSubscriptionProgress* progress = (SSubscriptionProgress*)calloc(numOfTables, sizeof(SSubscriptionProgress));
-  if (progress == NULL) {
-    tscError("failed to allocate memory for progress: %s", pSub->topic);
-    return 0;
-  }
 
-  if (UTIL_TABLE_IS_NOMRAL_TABLE(pTableMetaInfo)) {
+static int tscUpdateSubscription(STscObj* pObj, SSub* pSub) {
-    numOfTables = 1;
+  SSqlObj* pSql = pSub->pSql;
-    int64_t uid = pTableMetaInfo->pTableMeta->uid;
-    progress[0].uid = uid;
-    progress[0].key = tscGetSubscriptionProgress(pSub, uid);
-  } else {
-//    SSuperTableMeta* pMetricMeta = pTableMetaInfo->pMetricMeta;
-//    numOfTables = 0;
-//    for (int32_t i = 0; i < pMetricMeta->numOfVnodes; i++) {
-//      SVnodeSidList *pVnodeSidList = tscGetVnodeSidList(pMetricMeta, i);
-//      for (int32_t j = 0; j < pVnodeSidList->numOfSids; j++) {
-//        STableIdInfo *pTableMetaInfo = tscGetMeterSidInfo(pVnodeSidList, j);
-//        int64_t uid = pTableMetaInfo->uid;
-//        progress[numOfTables].uid = uid;
-//        progress[numOfTables++].key = tscGetSubscriptionProgress(pSub, uid);
-//      }
-//    }
-    qsort(progress, numOfTables, sizeof(SSubscriptionProgress), tscCompareSubscriptionProgress);
-  }
 
-  free(pSub->progress);
+  SSqlCmd* pCmd = &pSql->cmd;
-  pSub->numOfTables = numOfTables;
-  pSub->progress = progress;
 
   pSub->lastSyncTime = taosGetTimestampMs();
 
+  STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0);
+  if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo)) {
+    STableMeta * pTableMeta = pTableMetaInfo->pTableMeta;
+    SSubscriptionProgress target = {.uid = pTableMeta->uid, .key = 0};
+    SSubscriptionProgress* p = taosArraySearch(pSub->progress, tscCompareSubscriptionProgress, &target);
+    if (p == NULL) {
+      taosArrayClear(pSub->progress);
+      taosArrayPush(pSub->progress, &target);
+    }
+    return 1;
+  }
+
+  SArray* tables = getTableList(pSql);
+  size_t numOfTables = taosArrayGetSize(tables);
+
+  SArray* progress = taosArrayInit(numOfTables, sizeof(SSubscriptionProgress));
+  for( size_t i = 0; i < numOfTables; i++ ) {
+    STidTags* tt = taosArrayGet( tables, i );
+    SSubscriptionProgress p = { .uid = tt->uid };
+    p.key = tscGetSubscriptionProgress(pSub, tt->uid, INT64_MIN);
+    taosArrayPush(progress, &p);
+  }
+  taosArraySort(progress, tscCompareSubscriptionProgress);
+
+  taosArrayDestroy(pSub->progress);
+  pSub->progress = progress;
+
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
+    taosArraySort( tables, tscCompareTidTags );
+    tscBuildVgroupTableInfo( pTableMetaInfo, tables );
+  }
+  taosArrayDestroy(tables);
+
   return 1;
 }
 
@@ -246,32 +274,22 @@ static int tscLoadSubscriptionProgress(SSub* pSub) {
     return 0;
   }
 
-  if (fgets(buf, sizeof(buf), fp) == NULL || atoi(buf) < 0) {
+  SArray* progress = pSub->progress;
-    tscTrace("invalid subscription progress file: %s", pSub->topic);
+  taosArrayClear(progress);
-    fclose(fp);
+  while( 1 ) {
-    return 0;
-  }
-
-  int numOfTables = atoi(buf);
-  SSubscriptionProgress* progress = calloc(numOfTables, sizeof(SSubscriptionProgress));
-  for (int i = 0; i < numOfTables; i++) {
     if (fgets(buf, sizeof(buf), fp) == NULL) {
       fclose(fp);
-      free(progress);
       return 0;
     }
-    int64_t uid, key;
+    SSubscriptionProgress p;
-    sscanf(buf, "%" SCNd64 ":%" SCNd64, &uid, &key);
+    sscanf(buf, "%" SCNd64 ":%" SCNd64, &p.uid, &p.key);
-    progress[i].uid = uid;
+    taosArrayPush(progress, &p);
-    progress[i].key = key;
   }
 
   fclose(fp);
 
-  qsort(progress, numOfTables, sizeof(SSubscriptionProgress), tscCompareSubscriptionProgress);
+  taosArraySort(progress, tscCompareSubscriptionProgress);
-  pSub->numOfTables = numOfTables;
+  tscTrace("subscription progress loaded, %d tables: %s", taosArrayGetSize(progress), pSub->topic);
-  pSub->progress = progress;
-  tscTrace("subscription progress loaded, %d tables: %s", numOfTables, pSub->topic);
   return 1;
 }
 
@@ -292,11 +310,10 @@ void tscSaveSubscriptionProgress(void* sub) {
   }
 
   fputs(pSub->pSql->sqlstr, fp);
-  fprintf(fp, "\n%d\n", pSub->numOfTables);
+  fprintf(fp, "\n");
-  for (int i = 0; i < pSub->numOfTables; i++) {
+  for(size_t i = 0; i < taosArrayGetSize(pSub->progress); i++) {
-    int64_t uid = pSub->progress[i].uid;
+    SSubscriptionProgress* p = taosArrayGet(pSub->progress, i);
-    TSKEY key = pSub->progress[i].key;
+    fprintf(fp, "%" PRId64 ":%" PRId64 "\n", p->uid, p->key);
-    fprintf(fp, "%" PRId64 ":%" PRId64 "\n", uid, key);
   }
 
   fclose(fp);
@@ -361,35 +378,34 @@ TAOS_RES *taos_consume(TAOS_SUB *tsub) {
     tscRemoveFromSqlList(pSql);
 
     if (taosGetTimestampMs() - pSub->lastSyncTime > 10 * 60 * 1000) {
-      tscTrace("begin meter synchronization");
+      tscTrace("begin table synchronization");
-      char* sqlstr = pSql->sqlstr;
-      pSql->sqlstr = NULL;
-      taos_free_result_imp(pSql, 0);
-      pSql->sqlstr = sqlstr;
-      taosCacheEmpty(tscCacheHandle);
       if (!tscUpdateSubscription(pSub->taos, pSub)) return NULL;
-      tscTrace("meter synchronization completed");
+      tscTrace("table synchronization completed");
-    } else {
+    }
+
     SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0);
     
     uint32_t type = pQueryInfo->type;
-      taos_free_result_imp(pSql, 1);
+    tscFreeSqlResult(pSql);
     pRes->numOfRows = 1;
-      pRes->numOfTotal = 0;
     pRes->qhandle = 0;
     pSql->cmd.command = TSDB_SQL_SELECT;
     pQueryInfo->type = type;
 
     tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0)->vgroupIndex = 0;
-    }
 
+    pSql->fp = asyncCallback;
+    pSql->param = pSql;
     tscDoQuery(pSql);
-    if (pRes->code != TSDB_CODE_NOT_ACTIVE_TABLE) {
+    sem_wait(&pSql->rspSem);
-      break;
+
+    if (pRes->code != TSDB_CODE_SUCCESS) {
+      continue;
     }
     // meter was removed, make sync time zero, so that next retry will
     // do synchronization first
     pSub->lastSyncTime = 0;
+    break;
   }
 
   if (pRes->code != TSDB_CODE_SUCCESS) {
@@ -419,7 +435,7 @@ void taos_unsubscribe(TAOS_SUB *tsub, int keepProgress) {
   }
 
   tscFreeSqlObj(pSub->pSql);
-  free(pSub->progress);
+  taosArrayDestroy(pSub->progress);
   memset(pSub, 0, sizeof(*pSub));
   free(pSub);
 }

src/client/src/tscSubquery.c

@@ -330,7 +330,7 @@ int32_t tscLaunchSecondPhaseSubqueries(SSqlObj* pSql) {
     pNewQueryInfo->limit = pSupporter->limit;
   
     // fetch the join tag column
-    if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+    if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
       SSqlExpr* pExpr = tscSqlExprGet(pNewQueryInfo, 0);
       assert(pQueryInfo->tagCond.joinInfo.hasJoin);
     
@@ -463,77 +463,51 @@ static void tSIntersectionAndLaunchSecQuery(SJoinSupporter* pSupporter, SSqlObj*
   }
 }
 
-int32_t tagsOrderCompar(const void* p1, const void* p2) {
+int32_t tscCompareTidTags(const void* p1, const void* p2) {
-  STidTags* t1 = (STidTags*) p1;
+  const STidTags* t1 = (const STidTags*) p1;
-  STidTags* t2 = (STidTags*) p2;
+  const STidTags* t2 = (const STidTags*) p2;
   
   if (t1->vgId != t2->vgId) {
     return (t1->vgId > t2->vgId) ? 1 : -1;
-  } else {
+  }
   if (t1->tid != t2->tid) {
     return (t1->tid > t2->tid) ? 1 : -1;
-    } else {
+  }
   return 0;
 }
-  }
-}
 
-static void doBuildVgroupTableInfo(SArray* res, STableMetaInfo* pTableMetaInfo) {
+void tscBuildVgroupTableInfo(STableMetaInfo* pTableMetaInfo, SArray* tables) {
-  SArray* pGroup = taosArrayInit(4, sizeof(SVgroupTableInfo));
+  SArray* result = taosArrayInit( 4, sizeof(SVgroupTableInfo) );
+  SArray* vgTables = NULL;
+  STidTags* prev = NULL;
 
-  SArray* vgTableIdItem = taosArrayInit(4, sizeof(STableIdInfo));
+  size_t numOfTables = taosArrayGetSize( tables );
-  int32_t size = taosArrayGetSize(res);
+  for( size_t i = 0; i < numOfTables; i++ ) {
-  
+    STidTags* tt = taosArrayGet( tables, i );
-  STidTags* prev = taosArrayGet(res, 0);
-  int32_t prevVgId = prev->vgId;
-  
-  STableIdInfo item = {.uid = prev->uid, .tid = prev->tid, .key = INT64_MIN};
-  taosArrayPush(vgTableIdItem, &item);
-  
-  for(int32_t k = 1; k < size; ++k) {
-    STidTags* t1 = taosArrayGet(res, k);
-    if (prevVgId != t1->vgId) {
-      
-      SVgroupTableInfo info = {0};
 
+    if( prev == NULL || tt->vgId != prev->vgId ) {
       SVgroupsInfo* pvg = pTableMetaInfo->vgroupList;
+
+      SVgroupTableInfo info = {{ 0 }};
       for( int32_t m = 0; m < pvg->numOfVgroups; ++m ) {
-        if (prevVgId == pvg->vgroups[m].vgId) {
+        if( tt->vgId == pvg->vgroups[m].vgId ) {
           info.vgInfo = pvg->vgroups[m];
           break;
         }
       }
-      
       assert( info.vgInfo.numOfIps != 0 );
-      info.itemList = vgTableIdItem;
-      taosArrayPush(pGroup, &info);
 
-      vgTableIdItem = taosArrayInit(4, sizeof(STableIdInfo));
+      vgTables = taosArrayInit( 4, sizeof(STableIdInfo) );
-      STableIdInfo item1 = {.uid = t1->uid, .tid = t1->tid, .key = INT64_MIN};
+      info.itemList = vgTables;
-      taosArrayPush(vgTableIdItem, &item1);
+      taosArrayPush( result, &info );
-      prevVgId = t1->vgId;
-    } else {
-      taosArrayPush(vgTableIdItem, &item);
-    }
     }
 
-  if (taosArrayGetSize(vgTableIdItem) > 0) {
+    STableIdInfo item = { .uid = tt->uid, .tid = tt->tid, .key = INT64_MIN };
-    SVgroupTableInfo info = {0};
+    taosArrayPush( vgTables, &item );
-    SVgroupsInfo* pvg = pTableMetaInfo->vgroupList;
+    prev = tt;
-    
-    for(int32_t m = 0; m < pvg->numOfVgroups; ++m) {
-      if (prevVgId == pvg->vgroups[m].vgId) {
-        info.vgInfo = pvg->vgroups[m];
-        break;
-      }
   }
 
-    assert(info.vgInfo.numOfIps != 0);
+  pTableMetaInfo->pVgroupTables = result;
-    info.itemList = vgTableIdItem;
-    taosArrayPush(pGroup, &info);
-  }
-  
-  pTableMetaInfo->pVgroupTables = pGroup;
 }
 
 static void issueTSCompQuery(SSqlObj* pSql, SJoinSupporter* pSupporter, SSqlObj* pParent) {
@@ -627,8 +601,8 @@ static void joinRetrieveCallback(void* param, TAOS_RES* tres, int numOfRows) {
       SJoinSupporter* p1 = pParentSql->pSubs[0]->param;
       SJoinSupporter* p2 = pParentSql->pSubs[1]->param;
   
-      qsort(p1->pIdTagList, p1->num, p1->tagSize, tagsOrderCompar);
+      qsort(p1->pIdTagList, p1->num, p1->tagSize, tscCompareTidTags);
-      qsort(p2->pIdTagList, p2->num, p2->tagSize, tagsOrderCompar);
+      qsort(p2->pIdTagList, p2->num, p2->tagSize, tscCompareTidTags);
       
       STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
   
@@ -668,11 +642,11 @@ static void joinRetrieveCallback(void* param, TAOS_RES* tres, int numOfRows) {
         
         SQueryInfo* pQueryInfo1 = tscGetQueryInfoDetail(pSubCmd1, 0);
         STableMetaInfo* pTableMetaInfo1 = tscGetMetaInfo(pQueryInfo1, 0);
-        doBuildVgroupTableInfo(s1, pTableMetaInfo1);
+        tscBuildVgroupTableInfo(pTableMetaInfo1, s1);
         
         SQueryInfo* pQueryInfo2 = tscGetQueryInfoDetail(pSubCmd2, 0);
         STableMetaInfo* pTableMetaInfo2 = tscGetMetaInfo(pQueryInfo2, 0);
-        doBuildVgroupTableInfo(s2, pTableMetaInfo2);
+        tscBuildVgroupTableInfo(pTableMetaInfo2, s2);
   
         pSupporter->pState->numOfCompleted = 0;
         pSupporter->pState->code = 0;
@@ -1096,7 +1070,7 @@ int32_t tscLaunchJoinSubquery(SSqlObj *pSql, int16_t tableIndex, SJoinSupporter
     tscInitQueryInfo(pNewQueryInfo);
     STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pNewQueryInfo, 0);
     
-    if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) { // return the tableId & tag
+    if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) { // return the tableId & tag
       SSchema s = {0};
       SColumnIndex index = {0};
   
@@ -1203,7 +1177,7 @@ int32_t tscHandleMasterJoinQuery(SSqlObj* pSql) {
     }
   }
   
-  pSql->cmd.command = (pSql->numOfSubs <= 0)? TSDB_SQL_RETRIEVE_EMPTY_RESULT:TSDB_SQL_METRIC_JOIN_RETRIEVE;
+  pSql->cmd.command = (pSql->numOfSubs <= 0)? TSDB_SQL_RETRIEVE_EMPTY_RESULT:TSDB_SQL_TABLE_JOIN_RETRIEVE;
   
   return TSDB_CODE_SUCCESS;
 }
@@ -1533,8 +1507,7 @@ static void tscAllDataRetrievedFromDnode(SRetrieveSupport *trsupport, SSqlObj* p
   SQueryInfo *pPQueryInfo = tscGetQueryInfoDetail(&pPObj->cmd, 0);
   tscClearInterpInfo(pPQueryInfo);
   
-  tscCreateLocalReducer(trsupport->pExtMemBuffer, pState->numOfTotal, pDesc, trsupport->pFinalColModel,
+  tscCreateLocalReducer(trsupport->pExtMemBuffer, pState->numOfTotal, pDesc, trsupport->pFinalColModel, pPObj);
-                        &pPObj->cmd, &pPObj->res);
   tscTrace("%p build loser tree completed", pPObj);
   
   pPObj->res.precision = pSql->res.precision;
@@ -1950,7 +1923,7 @@ void **doSetResultRowData(SSqlObj *pSql, bool finalResult) {
   
   assert(pRes->row >= 0 && pRes->row <= pRes->numOfRows);
   
-  if(pCmd->command == TSDB_SQL_METRIC_JOIN_RETRIEVE) {
+  if(pCmd->command == TSDB_SQL_TABLE_JOIN_RETRIEVE) {
     if (pRes->completed) {
       tfree(pRes->tsrow);
     }

src/client/src/tscUtil.c

@@ -157,7 +157,7 @@ bool tscIsTwoStageSTableQuery(SQueryInfo* pQueryInfo, int32_t tableIndex) {
   }
   
   // for select query super table, the super table vgroup list can not be null in any cases.
-  if (pQueryInfo->command == TSDB_SQL_SELECT && UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (pQueryInfo->command == TSDB_SQL_SELECT && UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     assert(pTableMetaInfo->vgroupList != NULL);
   }
   
@@ -172,7 +172,7 @@ bool tscIsTwoStageSTableQuery(SQueryInfo* pQueryInfo, int32_t tableIndex) {
 
   if (((pQueryInfo->type & TSDB_QUERY_TYPE_STABLE_SUBQUERY) != TSDB_QUERY_TYPE_STABLE_SUBQUERY) &&
       pQueryInfo->command == TSDB_SQL_SELECT) {
-    return UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo);
+    return UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo);
   }
 
   return false;
@@ -187,7 +187,7 @@ bool tscIsProjectionQueryOnSTable(SQueryInfo* pQueryInfo, int32_t tableIndex) {
    * 4. show queries, instead of a select query
    */
   size_t numOfExprs = tscSqlExprNumOfExprs(pQueryInfo);
-  if (pTableMetaInfo == NULL || !UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo) ||
+  if (pTableMetaInfo == NULL || !UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo) ||
       pQueryInfo->command == TSDB_SQL_RETRIEVE_EMPTY_RESULT || numOfExprs == 0) {
     return false;
   }
@@ -386,14 +386,16 @@ void tscPartiallyFreeSqlObj(SSqlObj* pSql) {
 
   int32_t cmd = pCmd->command;
   if (cmd < TSDB_SQL_INSERT || cmd == TSDB_SQL_RETRIEVE_LOCALMERGE || cmd == TSDB_SQL_RETRIEVE_EMPTY_RESULT ||
-      cmd == TSDB_SQL_METRIC_JOIN_RETRIEVE) {
+      cmd == TSDB_SQL_TABLE_JOIN_RETRIEVE) {
     tscRemoveFromSqlList(pSql);
   }
   
   // pSql->sqlstr will be used by tscBuildQueryStreamDesc
+  if (pObj->signature == pObj) {
     pthread_mutex_lock(&pObj->mutex);
     tfree(pSql->sqlstr);
     pthread_mutex_unlock(&pObj->mutex);
+  }
   
   tscFreeSqlResult(pSql);
   
@@ -1209,18 +1211,18 @@ void tscColumnListCopy(SArray* dst, const SArray* src, int16_t tableIndex) {
   }
 }
 
-void tscColumnListDestroy(SArray* pColumnBaseInfo) {
+void tscColumnListDestroy(SArray* pColumnList) {
-  if (pColumnBaseInfo == NULL) {
+  if (pColumnList == NULL) {
     return;
   }
 
-  size_t num = taosArrayGetSize(pColumnBaseInfo);
+  size_t num = taosArrayGetSize(pColumnList);
   for (int32_t i = 0; i < num; ++i) {
-    SColumn* pCol = taosArrayGetP(pColumnBaseInfo, i);
+    SColumn* pCol = taosArrayGetP(pColumnList, i);
     tscColumnDestroy(pCol);
   }
 
-  taosArrayDestroy(pColumnBaseInfo);
+  taosArrayDestroy(pColumnList);
 }
 
 /*
@@ -1348,7 +1350,7 @@ bool tscValidateColumnId(STableMetaInfo* pTableMetaInfo, int32_t colId) {
     return false;
   }
 
-  if (colId == -1 && UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+  if (colId == -1 && UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     return true;
   }
 
@@ -1459,10 +1461,11 @@ bool tscShouldFreeHeatBeat(SSqlObj* pHb) {
 }
 
 /*
- * the following three kinds of SqlObj should not be freed
+ * the following four kinds of SqlObj should not be freed
  * 1. SqlObj for stream computing
  * 2. main SqlObj
  * 3. heartbeat SqlObj
+ * 4. SqlObj for subscription
  *
  * If res code is error and SqlObj does not belong to above types, it should be
  * automatically freed for async query, ignoring that connection should be kept.
@@ -1475,7 +1478,7 @@ bool tscShouldBeFreed(SSqlObj* pSql) {
   }
 
   STscObj* pTscObj = pSql->pTscObj;
-  if (pSql->pStream != NULL || pTscObj->pHb == pSql || pTscObj->pSql == pSql) {
+  if (pSql->pStream != NULL || pTscObj->pHb == pSql || pTscObj->pSql == pSql || pSql->pSubscription != NULL) {
     return false;
   }
 
@@ -1644,9 +1647,11 @@ void doRemoveTableMetaInfo(SQueryInfo* pQueryInfo, int32_t index, bool removeFro
 void clearAllTableMetaInfo(SQueryInfo* pQueryInfo, const char* address, bool removeFromCache) {
   tscTrace("%p deref the table meta in cache, numOfTables:%d", address, pQueryInfo->numOfTables);
   
-  int32_t index = pQueryInfo->numOfTables;
+  for(int32_t i = 0; i < pQueryInfo->numOfTables; ++i) {
-  while (index >= 0) {
+    STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, i);
-    doRemoveTableMetaInfo(pQueryInfo, --index, removeFromCache);
+  
+    tscClearTableMetaInfo(pTableMetaInfo, removeFromCache);
+    free(pTableMetaInfo);
   }
   
   tfree(pQueryInfo->pTableMetaInfo);
@@ -1666,8 +1671,7 @@ STableMetaInfo* tscAddTableMetaInfo(SQueryInfo* pQueryInfo, const char* name, ST
   assert(pTableMetaInfo != NULL);
 
   if (name != NULL) {
-    assert(strlen(name) <= TSDB_TABLE_ID_LEN);
+    strncpy(pTableMetaInfo->name, name, TSDB_TABLE_ID_LEN);
-    strcpy(pTableMetaInfo->name, name);
   }
 
   pTableMetaInfo->pTableMeta = pTableMeta;
@@ -1678,10 +1682,9 @@ STableMetaInfo* tscAddTableMetaInfo(SQueryInfo* pQueryInfo, const char* name, ST
     memcpy(pTableMetaInfo->vgroupList, vgroupList, size);
   }
 
-  if (pTagCols == NULL) {
   pTableMetaInfo->tagColList = taosArrayInit(4, POINTER_BYTES);
-  } else {
+  if (pTagCols != NULL) {
-    pTableMetaInfo->tagColList = taosArrayClone(pTagCols);
+    tscColumnListCopy(pTableMetaInfo->tagColList, pTagCols, -1);
   }
   
   pQueryInfo->numOfTables += 1;
@@ -1700,11 +1703,9 @@ void tscClearTableMetaInfo(STableMetaInfo* pTableMetaInfo, bool removeFromCache)
   taosCacheRelease(tscCacheHandle, (void**)&(pTableMetaInfo->pTableMeta), removeFromCache);
   tfree(pTableMetaInfo->vgroupList);
   
-  if (pTableMetaInfo->tagColList != NULL) {
+  tscColumnListDestroy(pTableMetaInfo->tagColList);
-    taosArrayDestroy(pTableMetaInfo->tagColList);
   pTableMetaInfo->tagColList = NULL;
 }
-}
 
 void tscResetForNextRetrieve(SSqlRes* pRes) {
   if (pRes == NULL) {
@@ -1847,7 +1848,8 @@ SSqlObj* createSubqueryObj(SSqlObj* pSql, int16_t tableIndex, void (*fp)(), void
 
   if (pPrevSql == NULL) {
     STableMeta* pTableMeta = taosCacheAcquireByName(tscCacheHandle, name);
-
+    // todo handle error
+    assert(pTableMeta != NULL);
     pFinalInfo = tscAddTableMetaInfo(pNewQueryInfo, name, pTableMeta, pTableMetaInfo->vgroupList, pTableMetaInfo->tagColList);
   } else {  // transfer the ownership of pTableMeta to the newly create sql object.
     STableMetaInfo* pPrevInfo = tscGetTableMetaInfoFromCmd(&pPrevSql->cmd, pPrevSql->cmd.clauseIndex, 0);
@@ -1859,8 +1861,15 @@ SSqlObj* createSubqueryObj(SSqlObj* pSql, int16_t tableIndex, void (*fp)(), void
     pFinalInfo = tscAddTableMetaInfo(pNewQueryInfo, name, pPrevTableMeta, pVgroupsInfo, pTableMetaInfo->tagColList);
   }
 
-  assert(pFinalInfo->pTableMeta != NULL && pNewQueryInfo->numOfTables == 1);
+  if (pFinalInfo->pTableMeta == NULL) {
-  if (UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo)) {
+    tscError("%p new subquery failed for get pMeterMeta is NULL from cache", pSql);
+    tscFreeSqlObj(pNew);
+    return NULL;
+  }
+  
+  assert(pNewQueryInfo->numOfTables == 1);
+  
+  if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
     assert(pFinalInfo->vgroupList != NULL);
   }
   
@@ -1986,22 +1995,28 @@ char* tscGetErrorMsgPayload(SSqlCmd* pCmd) { return pCmd->payload; }
 
 /**
  *  If current vnode query does not return results anymore (pRes->numOfRows == 0), try the next vnode if exists,
- *  in case of multi-vnode super table projection query and the result does not reach the limitation.
+ *  while multi-vnode super table projection query and the result does not reach the limitation.
  */
 bool hasMoreVnodesToTry(SSqlObj* pSql) {
-//  SSqlCmd* pCmd = &pSql->cmd;
+  SSqlCmd* pCmd = &pSql->cmd;
-//  SSqlRes* pRes = &pSql->res;
+  SSqlRes* pRes = &pSql->res;
-
+  if (pCmd->command != TSDB_SQL_FETCH) {
-//  SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
-  
-//  STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-//  if (!UTIL_TABLE_IS_SUPERTABLE(pTableMetaInfo) || (pTableMetaInfo->pMetricMeta == NULL)) {
     return false;
-//  }
+  }
 
-//  int32_t totalVnode = pTableMetaInfo->pMetricMeta->numOfVnodes;
+  SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
-//  return pRes->numOfRows == 0 && tscNonOrderedProjectionQueryOnSTable(pQueryInfo, 0) &&
+  
-//         (!tscHasReachLimitation(pQueryInfo, pRes)) && (pTableMetaInfo->vgroupIndex < totalVnode - 1);
+  STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
+  assert(pRes->completed);
+  
+  // for normal table, do not try any more if result are exhausted
+  if (!UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo) || (pTableMetaInfo->vgroupList == NULL)) {
+    return false;
+  }
+  
+  int32_t numOfVgroups = pTableMetaInfo->vgroupList->numOfVgroups;
+  return tscNonOrderedProjectionQueryOnSTable(pQueryInfo, 0) &&
+         (!tscHasReachLimitation(pQueryInfo, pRes)) && (pTableMetaInfo->vgroupIndex < numOfVgroups - 1);
 }
 
 void tscTryQueryNextVnode(SSqlObj* pSql, __async_cb_func_t fp) {
@@ -2017,12 +2032,11 @@ void tscTryQueryNextVnode(SSqlObj* pSql, __async_cb_func_t fp) {
   assert(pRes->numOfRows == 0 && tscNonOrderedProjectionQueryOnSTable(pQueryInfo, 0) && !tscHasReachLimitation(pQueryInfo, pRes));
 
   STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
-  int32_t totalVnode = 0;
-//  int32_t         totalVnode = pTableMetaInfo->pMetricMeta->numOfVnodes;
   
-  while (++pTableMetaInfo->vgroupIndex < totalVnode) {
+  int32_t totalVgroups = pTableMetaInfo->vgroupList->numOfVgroups;
+  while (++pTableMetaInfo->vgroupIndex < totalVgroups) {
     tscTrace("%p current vnode:%d exhausted, try next:%d. total vnode:%d. current numOfRes:%d", pSql,
-             pTableMetaInfo->vgroupIndex - 1, pTableMetaInfo->vgroupIndex, totalVnode, pRes->numOfTotalInCurrentClause);
+             pTableMetaInfo->vgroupIndex - 1, pTableMetaInfo->vgroupIndex, totalVgroups, pRes->numOfTotalInCurrentClause);
 
     /*
      * update the limit and offset value for the query on the next vnode,
@@ -2038,10 +2052,10 @@ void tscTryQueryNextVnode(SSqlObj* pSql, __async_cb_func_t fp) {
     }
 
     pQueryInfo->limit.offset = pRes->offset;
-
     assert((pRes->offset >= 0 && pRes->numOfRows == 0) || (pRes->offset == 0 && pRes->numOfRows >= 0));
-    tscTrace("%p new query to next vnode, vnode index:%d, limit:%" PRId64 ", offset:%" PRId64 ", glimit:%" PRId64, pSql,
+    
-             pTableMetaInfo->vgroupIndex, pQueryInfo->limit.limit, pQueryInfo->limit.offset, pQueryInfo->clauseLimit);
+    tscTrace("%p new query to next vgroup, index:%d, limit:%" PRId64 ", offset:%" PRId64 ", glimit:%" PRId64,
+        pSql, pTableMetaInfo->vgroupIndex, pQueryInfo->limit.limit, pQueryInfo->limit.offset, pQueryInfo->clauseLimit);
 
     /*
      * For project query with super table join, the numOfSub is equalled to the number of all subqueries.
@@ -2055,43 +2069,46 @@ void tscTryQueryNextVnode(SSqlObj* pSql, __async_cb_func_t fp) {
     tscResetForNextRetrieve(pRes);
 
     // in case of async query, set the callback function
-    void* fp1 = pSql->fp;
+//    void* fp1 = pSql->fp;
     pSql->fp = fp;
 
-    if (fp1 != NULL) {
+//    if (fp1 != NULL) {
-      assert(fp != NULL);
+//      assert(fp != NULL);
+//    }
+
+    int32_t ret = tscProcessSql(pSql);
+    if (ret == TSDB_CODE_SUCCESS) {
+      return;
+    } else {// todo check for failure
     }
-
-    int32_t ret = tscProcessSql(pSql);  // todo check for failure
-
     // in case of async query, return now
-    if (fp != NULL) {
+//    if (fp != NULL) {
-      return;
+//      return;
+//    }
+//
+//    if (ret != TSDB_CODE_SUCCESS) {
+//      pSql->res.code = ret;
+//      return;
+//    }
+//
+//    // retrieve data
+//    assert(pCmd->command == TSDB_SQL_SELECT);
+//    pCmd->command = TSDB_SQL_FETCH;
+//
+//    if ((ret = tscProcessSql(pSql)) != TSDB_CODE_SUCCESS) {
+//      pSql->res.code = ret;
+//      return;
+//    }
+//
+//    // if the result from current virtual node are empty, try next if exists. otherwise, return the results.
+//    if (pRes->numOfRows > 0) {
+//      break;
+//    }
   }
 
-    if (ret != TSDB_CODE_SUCCESS) {
+//  if (pRes->numOfRows == 0) {
-      pSql->res.code = ret;
+//    tscTrace("%p all vnodes exhausted, prj query completed. total res:%d", pSql, totalVnode, pRes->numOfTotal);
-      return;
+//  }
-    }
-
-    // retrieve data
-    assert(pCmd->command == TSDB_SQL_SELECT);
-    pCmd->command = TSDB_SQL_FETCH;
-
-    if ((ret = tscProcessSql(pSql)) != TSDB_CODE_SUCCESS) {
-      pSql->res.code = ret;
-      return;
-    }
-
-    // if the result from current virtual node are empty, try next if exists. otherwise, return the results.
-    if (pRes->numOfRows > 0) {
-      break;
-    }
-  }
-
-  if (pRes->numOfRows == 0) {
-    tscTrace("%p all vnodes exhausted, prj query completed. total res:%d", pSql, totalVnode, pRes->numOfTotal);
-  }
 }
 
 void tscTryQueryNextClause(SSqlObj* pSql, void (*queryFp)()) {
@@ -2155,3 +2172,26 @@ void tscGetResultColumnChr(SSqlRes* pRes, SFieldInfo* pFieldInfo, int32_t column
   }
 }
 
+void* malloc_throw(size_t size) {
+  void* p = malloc(size);
+  if (p == NULL) {
+    THROW(TSDB_CODE_CLI_OUT_OF_MEMORY);
+  }
+  return p;
+}
+
+void* calloc_throw(size_t nmemb, size_t size) {
+  void* p = calloc(nmemb, size);
+  if (p == NULL) {
+    THROW(TSDB_CODE_CLI_OUT_OF_MEMORY);
+  }
+  return p;
+}
+
+char* strdup_throw(const char* str) {
+  char* p = strdup(str);
+  if (p == NULL) {
+    THROW(TSDB_CODE_CLI_OUT_OF_MEMORY);
+  }
+  return p;
+}

src/common/inc/tname.h

@@ -23,4 +23,5 @@ void extractTableName(const char *tableId, char *name);
 
 char* extractDBName(const char *tableId, char *name);
 
+
 #endif  // TDENGINE_NAME_H

src/dnode/src/dnodeMain.c

@@ -84,7 +84,7 @@ int32_t main(int32_t argc, char *argv[]) {
   }
 
   /* Set termination handler. */
-  struct sigaction act = {0};
+  struct sigaction act = {{0}};
   act.sa_flags = SA_SIGINFO;
   act.sa_sigaction = signal_handler;
   sigaction(SIGTERM, &act, NULL);

src/dnode/src/dnodeModule.c

@@ -33,7 +33,7 @@ typedef struct {
   void    (*stopFp)();
 } SModule;
 
-static SModule  tsModule[TSDB_MOD_MAX] = {0};
+static SModule  tsModule[TSDB_MOD_MAX] = {{0}};
 static uint32_t tsModuleStatus = 0;
 
 static void dnodeSetModuleStatus(int32_t module) {

src/inc/taosdef.h

@@ -32,6 +32,9 @@ extern "C" {
 #define TSKEY int64_t
 #endif
 
+#define TSWINDOW_INITIALIZER {INT64_MIN, INT64_MAX};
+#define TSKEY_INITIAL_VAL    INT64_MIN
+
 // ----------------- For variable data types such as TSDB_DATA_TYPE_BINARY and TSDB_DATA_TYPE_NCHAR
 typedef int32_t VarDataOffsetT;
 typedef int16_t VarDataLenT;
@@ -341,8 +344,6 @@ void tsDataSwap(void *pLeft, void *pRight, int32_t type, int32_t size);
 #define TSDB_MAX_DBS           100
 #define TSDB_MAX_VGROUPS       1000
 #define TSDB_MAX_SUPER_TABLES  100
-#define TSDB_MAX_NORMAL_TABLES 1000
-#define TSDB_MAX_CHILD_TABLES  100000
 
 #define TSDB_PORT_DNODESHELL 0 
 #define TSDB_PORT_DNODEDNODE 5 

src/inc/taosmsg.h

@@ -627,7 +627,6 @@ typedef struct {
 typedef struct STableMetaMsg {
   int32_t       contLen;
   char          tableId[TSDB_TABLE_ID_LEN + 1];   // table id
-  char          stableId[TSDB_TABLE_ID_LEN + 1];  // stable name if it is created according to super table
   uint8_t       numOfTags;
   uint8_t       precision;
   uint8_t       tableType;

src/inc/tsdb.h

@@ -34,12 +34,14 @@ extern "C" {
 
 #define TSDB_INVALID_SUPER_TABLE_ID -1
 
+#define TSDB_STATUS_COMMIT_START 1
+#define TSDB_STATUS_COMMIT_OVER  2
+
 // --------- TSDB APPLICATION HANDLE DEFINITION
 typedef struct {
-  // WAL handle
   void *appH;
   void *cqH;
-  int (*walCallBack)(void *);
+  int (*notifyStatus)(void *, int status);
   int (*eventCallBack)(void *);
 } STsdbAppH;
 

src/mnode/src/mgmtTable.c

@@ -1185,8 +1185,10 @@ void mgmtDropAllSuperTables(SDbObj *pDropDb) {
 
 static int32_t mgmtSetSchemaFromSuperTable(SSchema *pSchema, SSuperTableObj *pTable) {
   int32_t numOfCols = pTable->numOfColumns + pTable->numOfTags;
+  assert(numOfCols <= TSDB_MAX_COLUMNS);
+  
   for (int32_t i = 0; i < numOfCols; ++i) {
-    strncpy(pSchema->name, pTable->schema[i].name, TSDB_TABLE_ID_LEN);
+    strncpy(pSchema->name, pTable->schema[i].name, TSDB_COL_NAME_LEN);
     pSchema->type  = pTable->schema[i].type;
     pSchema->bytes = htons(pTable->schema[i].bytes);
     pSchema->colId = htons(pTable->schema[i].colId);
@@ -1659,7 +1661,6 @@ static int32_t mgmtDoGetChildTableMeta(SQueuedMsg *pMsg, STableMetaMsg *pMeta) {
     pMeta->numOfTags    = (int8_t)pTable->superTable->numOfTags;
     pMeta->numOfColumns = htons((int16_t)pTable->superTable->numOfColumns);
     pMeta->contLen      = sizeof(STableMetaMsg) + mgmtSetSchemaFromSuperTable(pMeta->schema, pTable->superTable);
-    strncpy(pMeta->stableId, pTable->superTable->info.tableId, tListLen(pMeta->stableId));
   } else {
     pMeta->sversion     = htons(pTable->sversion);
     pMeta->numOfTags    = 0;

src/os/linux/inc/os.h

@@ -253,6 +253,17 @@ void taosBlockSIGPIPE();
 #define BUILDIN_CLZ(val) __builtin_clz(val)
 #define BUILDIN_CTZ(val) __builtin_ctz(val)
 
+#undef threadlocal
+#ifdef _ISOC11_SOURCE
+  #define threadlocal _Thread_local
+#elif defined(__APPLE__)
+  #define threadlocal
+#elif defined(__GNUC__) && !defined(threadlocal)
+  #define threadlocal __thread
+#else
+  #define threadlocal
+#endif
+
 #ifdef __cplusplus
 }
 #endif

src/os/linux/src/linuxPlatform.c

@@ -157,7 +157,7 @@ void *taosProcessAlarmSignal(void *tharg) {
   void (*callback)(int) = tharg;
 
   static timer_t         timerId;
-  struct sigevent sevent = {0};
+  struct sigevent sevent = {{0}};
 
   #ifdef _ALPINE
     sevent.sigev_notify = SIGEV_THREAD;

src/query/inc/qextbuffer.h

@@ -120,12 +120,6 @@ typedef struct tExtMemBuffer {
   EXT_BUFFER_FLUSH_MODEL flushModel;
 } tExtMemBuffer;
 
-//typedef struct tTagSchema {
-//  struct SSchema *pSchema;
-//  int32_t         numOfCols;
-//  int32_t         colOffset[];
-//} tTagSchema;
-
 /**
  *
  * @param inMemSize

src/query/inc/qsqltype.h

@@ -61,7 +61,7 @@ enum _sql_type {
   TSDB_SQL_LOCAL,  // SQL below for client local
   TSDB_SQL_DESCRIBE_TABLE,
   TSDB_SQL_RETRIEVE_LOCALMERGE,
-  TSDB_SQL_METRIC_JOIN_RETRIEVE,
+  TSDB_SQL_TABLE_JOIN_RETRIEVE,
 
   /*
    * build empty result instead of accessing dnode to fetch result

src/query/inc/queryExecutor.h

@@ -182,6 +182,7 @@ typedef struct SQInfo {
   SQueryRuntimeEnv runtimeEnv;
   int32_t          groupIndex;
   int32_t          offset;            // offset in group result set of subgroup, todo refactor
+  SArray*          arrTableIdInfo;
   
   T_REF_DECLARE()
   /*

src/query/src/queryExecutor.c

@@ -113,7 +113,6 @@ static void initCtxOutputBuf(SQueryRuntimeEnv *pRuntimeEnv);
 static void destroyTableQueryInfo(STableQueryInfo *pTableQueryInfo, int32_t numOfCols);
 static void resetCtxOutputBuf(SQueryRuntimeEnv *pRuntimeEnv);
 static bool hasMainOutput(SQuery *pQuery);
-static void createTableQueryInfo(SQInfo *pQInfo);
 static void buildTagQueryResult(SQInfo *pQInfo);
 
 static int32_t setAdditionalInfo(SQInfo *pQInfo, STableId *pTableId, STableQueryInfo *pTableQueryInfo);
@@ -507,7 +506,7 @@ static STimeWindow getActiveTimeWindow(SWindowResInfo *pWindowResInfo, int64_t t
     w.ekey = pQuery->window.ekey;
   }
 
-  assert(ts >= w.skey && ts <= w.ekey && w.skey != 0);
+  assert(ts >= w.skey && ts <= w.ekey);
 
   return w;
 }
@@ -624,7 +623,7 @@ static void doCheckQueryCompleted(SQueryRuntimeEnv *pRuntimeEnv, TSKEY lastKey,
     setQueryStatus(pQuery, QUERY_COMPLETED | QUERY_RESBUF_FULL);
   } else {  // set the current index to be the last unclosed window
     int32_t i = 0;
-    int64_t skey = 0;
+    int64_t skey = TSKEY_INITIAL_VAL;
 
     for (i = 0; i < pWindowResInfo->size; ++i) {
       SWindowResult *pResult = &pWindowResInfo->pResult[i];
@@ -642,7 +641,7 @@ static void doCheckQueryCompleted(SQueryRuntimeEnv *pRuntimeEnv, TSKEY lastKey,
     }
 
     // all windows are closed, set the last one to be the skey
-    if (skey == 0) {
+    if (skey == TSKEY_INITIAL_VAL) {
       assert(i == pWindowResInfo->size);
       pWindowResInfo->curIndex = pWindowResInfo->size - 1;
     } else {
@@ -660,7 +659,7 @@ static void doCheckQueryCompleted(SQueryRuntimeEnv *pRuntimeEnv, TSKEY lastKey,
     qTrace("QInfo:%p total window:%d, closed:%d", GET_QINFO_ADDR(pRuntimeEnv), pWindowResInfo->size, n);
   }
 
-  assert(pWindowResInfo->prevSKey != 0);
+  assert(pWindowResInfo->prevSKey != TSKEY_INITIAL_VAL);
 }
 
 static int32_t getNumOfRowsInTimeWindow(SQuery *pQuery, SDataBlockInfo *pDataBlockInfo, TSKEY *pPrimaryColumn,
@@ -2399,7 +2398,7 @@ static int64_t doScanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv) {
     // todo extract methods
     if (isIntervalQuery(pQuery) && pRuntimeEnv->windowResInfo.prevSKey == 0) {
       TSKEY           skey1, ekey1;
-      STimeWindow     w = {0};
+      STimeWindow     w = TSWINDOW_INITIALIZER;
       SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo;
 
       if (QUERY_IS_ASC_QUERY(pQuery)) {
@@ -3080,6 +3079,9 @@ void disableFuncInReverseScan(SQInfo *pQInfo) {
       int32_t functId = pQuery->pSelectExpr[j].base.functionId;
 
       SQLFunctionCtx *pCtx = &pRuntimeEnv->pCtx[j];
+      if (pCtx->resultInfo == NULL) {
+        continue; // resultInfo is NULL, means no data checked in previous scan
+      }
 
       if (((functId == TSDB_FUNC_FIRST || functId == TSDB_FUNC_FIRST_DST) && order == TSDB_ORDER_ASC) ||
           ((functId == TSDB_FUNC_LAST || functId == TSDB_FUNC_LAST_DST) && order == TSDB_ORDER_DESC)) {
@@ -3460,7 +3462,11 @@ static bool hasMainOutput(SQuery *pQuery) {
   return false;
 }
 
-STableQueryInfo *createTableQueryInfoImpl(SQueryRuntimeEnv *pRuntimeEnv, STableId tableId, STimeWindow win) {
+static STableQueryInfo *createTableQueryInfo(
+  SQueryRuntimeEnv *pRuntimeEnv,
+  STableId tableId,
+  STimeWindow win
+) {
   STableQueryInfo *pTableQueryInfo = calloc(1, sizeof(STableQueryInfo));
 
   pTableQueryInfo->win = win;
@@ -3590,7 +3596,6 @@ void setIntervalQueryRange(SQInfo *pQInfo, TSKEY key) {
   if (pTableQueryInfo->queryRangeSet) {
     pTableQueryInfo->lastKey = key;
   } else {
-//    pQuery->window.skey = key;
     pTableQueryInfo->win.skey = key;
     STimeWindow win = {.skey = key, .ekey = pQuery->window.ekey};
 
@@ -3613,18 +3618,16 @@ void setIntervalQueryRange(SQInfo *pQInfo, TSKEY key) {
     getAlignQueryTimeWindow(pQuery, win.skey, win.skey, win.ekey, &skey1, &ekey1, &w);
     pWindowResInfo->startTime = pTableQueryInfo->win.skey;  // windowSKey may be 0 in case of 1970 timestamp
 
-    if (pWindowResInfo->prevSKey == 0) {
+    if (pWindowResInfo->prevSKey == TSKEY_INITIAL_VAL) {
-      if (QUERY_IS_ASC_QUERY(pQuery)) {
+      if (!QUERY_IS_ASC_QUERY(pQuery)) {
-        pWindowResInfo->prevSKey = w.skey;
-      } else {
         assert(win.ekey == pQuery->window.skey);
-        pWindowResInfo->prevSKey = w.skey;
       }
+      
+      pWindowResInfo->prevSKey = w.skey;
     }
 
     pTableQueryInfo->queryRangeSet = 1;
     pTableQueryInfo->lastKey = pTableQueryInfo->win.skey;
-    pTableQueryInfo->win.skey = pTableQueryInfo->win.skey;
   }
 }
 
@@ -3870,6 +3873,17 @@ static void doCopyQueryResultToMsg(SQInfo *pQInfo, int32_t numOfRows, char *data
     data += bytes * numOfRows;
   }
 
+  int32_t numOfTables = (int32_t)taosArrayGetSize(pQInfo->arrTableIdInfo);
+  *(int32_t*)data = htonl(numOfTables);
+  data += sizeof(int32_t);
+  for(int32_t i = 0; i < numOfTables; i++) {
+    STableIdInfo* pSrc = taosArrayGet(pQInfo->arrTableIdInfo, i);
+    STableIdInfo* pDst = (STableIdInfo*)data;
+    pDst->uid = htobe64(pSrc->uid);
+    pDst->tid = htonl(pSrc->tid);
+    pDst->key = htobe64(pSrc->key);
+    data += sizeof(STableIdInfo);
+  }
 
   // all data returned, set query over
   if (Q_STATUS_EQUAL(pQuery->status, QUERY_COMPLETED)) {
@@ -4057,10 +4071,11 @@ static bool skipTimeInterval(SQueryRuntimeEnv *pRuntimeEnv) {
    *    pQuery->limit.offset times. Since hole exists, pQuery->intervalTime*pQuery->limit.offset value is
    *    not valid. otherwise, we only forward pQuery->limit.offset number of points
    */
-  assert(pRuntimeEnv->windowResInfo.prevSKey == 0);
+  assert(pRuntimeEnv->windowResInfo.prevSKey == TSKEY_INITIAL_VAL);
 
   TSKEY       skey1, ekey1;
-  STimeWindow     w = {0};
+  STimeWindow w = TSWINDOW_INITIALIZER;
+  
   SWindowResInfo *pWindowResInfo = &pRuntimeEnv->windowResInfo;
   STableQueryInfo *pTableQueryInfo = pQuery->current;
 
@@ -4148,14 +4163,18 @@ static bool skipTimeInterval(SQueryRuntimeEnv *pRuntimeEnv) {
   return true;
 }
 
-int32_t doInitQInfo(SQInfo *pQInfo, void *param, void *tsdb, int32_t vgId, bool isSTableQuery) {
+
+static void setupQueryHandle(void* tsdb, SQInfo* pQInfo, bool isSTableQuery) {
   SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv;
-
   SQuery *pQuery = pQInfo->runtimeEnv.pQuery;
-  int32_t code = TSDB_CODE_SUCCESS;
 
-  setScanLimitationByResultBuffer(pQuery);
+  if (onlyQueryTags(pQuery)) {
-  changeExecuteScanOrder(pQuery, false);
+    return;
+  }
+
+  if (isSTableQuery && (!isIntervalQuery(pQuery)) && (!isFixedOutputQuery(pQuery))) {
+    return;
+  }
 
   STsdbQueryCond cond = {
     .twindow = pQuery->window,
@@ -4164,18 +4183,31 @@ int32_t doInitQInfo(SQInfo *pQInfo, void *param, void *tsdb, int32_t vgId, bool
     .numOfCols = pQuery->numOfCols,
   };
 
+  if (!isSTableQuery
+    && (pQInfo->groupInfo.numOfTables == 1)
+    && (cond.order == TSDB_ORDER_ASC) 
+    && (!isIntervalQuery(pQuery))
+    && (!isGroupbyNormalCol(pQuery->pGroupbyExpr))
+    && (!isFixedOutputQuery(pQuery))
+  ) {
+    SArray* pa = taosArrayGetP(pQInfo->groupInfo.pGroupList, 0);
+    SGroupItem* pItem = taosArrayGet(pa, 0);
+    cond.twindow = pItem->info->win;
+  }
 
-  // normal query setup the queryhandle here
-  if (!onlyQueryTags(pQuery)) {
-    if (!isSTableQuery && isFirstLastRowQuery(pQuery)) {  // in case of last_row query, invoke a different API.
-      pRuntimeEnv->pQueryHandle = tsdbQueryLastRow(tsdb, &cond, &pQInfo->tableIdGroupInfo);
-    } else if (!isSTableQuery || isIntervalQuery(pQuery) || isFixedOutputQuery(pQuery)) {
   pRuntimeEnv->pQueryHandle = tsdbQueryTables(tsdb, &cond, &pQInfo->tableIdGroupInfo);
 }
 
-    // create the table query support structures
+
-    createTableQueryInfo(pQInfo);
+int32_t doInitQInfo(SQInfo *pQInfo, void *param, void *tsdb, int32_t vgId, bool isSTableQuery) {
-  }
+  SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv;
+
+  SQuery *pQuery = pQInfo->runtimeEnv.pQuery;
+  int32_t code = TSDB_CODE_SUCCESS;
+
+  setScanLimitationByResultBuffer(pQuery);
+  changeExecuteScanOrder(pQuery, false);
+  setupQueryHandle(tsdb, pQInfo, isSTableQuery);
   
   pQInfo->tsdb = tsdb;
   pQInfo->vgId = vgId;
@@ -4594,6 +4626,13 @@ static void sequentialTableProcess(SQInfo *pQInfo) {
          * to ensure that, we can reset the query range once query on a meter is completed.
          */
         pQInfo->tableIndex++;
+
+        STableIdInfo tidInfo;
+        tidInfo.uid = item->id.uid;
+        tidInfo.tid = item->id.tid;
+        tidInfo.key = pQuery->current->lastKey;
+        taosArrayPush(pQInfo->arrTableIdInfo, &tidInfo);
+
         // if the buffer is full or group by each table, we need to jump out of the loop
         if (Q_STATUS_EQUAL(pQuery->status, QUERY_RESBUF_FULL) /*||
             isGroupbyEachTable(pQuery->pGroupbyExpr, pSupporter->pSidSet)*/) {
@@ -4663,35 +4702,6 @@ static void sequentialTableProcess(SQInfo *pQInfo) {
       pQuery->limit.offset);
 }
 
-static void createTableQueryInfo(SQInfo *pQInfo) {
-  SQuery *pQuery = pQInfo->runtimeEnv.pQuery;
-
-  // todo make sure the table are added the reference count to gauranteed that all involved tables are valid
-  size_t numOfGroups = taosArrayGetSize(pQInfo->groupInfo.pGroupList);
-
-  int32_t index = 0;
-  for (int32_t i = 0; i < numOfGroups; ++i) {  // load all meter meta info
-    SArray *group = *(SArray **)taosArrayGet(pQInfo->groupInfo.pGroupList, i);
-
-    size_t s = taosArrayGetSize(group);
-    for (int32_t j = 0; j < s; ++j) {
-      SGroupItem* item = (SGroupItem *)taosArrayGet(group, j);
-
-      // STableQueryInfo has been created for each table
-      if (item->info != NULL) {
-        return;
-      }
-
-      STableQueryInfo* pInfo = createTableQueryInfoImpl(&pQInfo->runtimeEnv, item->id, pQuery->window);
-      pInfo->groupIdx = i;
-      pInfo->tableIndex = index;
-      
-      item->info = pInfo;
-      index += 1;
-    }
-  }
-}
-
 static void doSaveContext(SQInfo *pQInfo) {
   SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv;
   SQuery *          pQuery = pRuntimeEnv->pQuery;
@@ -4730,7 +4740,7 @@ static void doRestoreContext(SQInfo *pQInfo) {
   SWAP(pQuery->window.skey, pQuery->window.ekey, TSKEY);
 
   if (pRuntimeEnv->pTSBuf != NULL) {
-    pRuntimeEnv->pTSBuf->cur.order = pRuntimeEnv->pTSBuf->cur.order ^ 1;
+    SWITCH_ORDER(pRuntimeEnv->pTSBuf->cur.order);
   }
 
   switchCtxOrder(pRuntimeEnv);
@@ -4913,6 +4923,12 @@ static void tableMultiOutputProcess(SQInfo *pQInfo, STableQueryInfo* pTableInfo)
   if (Q_STATUS_EQUAL(pQuery->status, QUERY_RESBUF_FULL)) {
     qTrace("QInfo:%p query paused due to output limitation, next qrange:%" PRId64 "-%" PRId64, pQInfo,
         pQuery->current->lastKey, pQuery->window.ekey);
+  } else if (Q_STATUS_EQUAL(pQuery->status, QUERY_COMPLETED)) {
+    STableIdInfo tidInfo;
+    tidInfo.uid = pQuery->current->id.uid;
+    tidInfo.tid = pQuery->current->id.tid;
+    tidInfo.key = pQuery->current->lastKey;
+    taosArrayPush(pQInfo->arrTableIdInfo, &tidInfo);
   }
 
   if (!isTSCompQuery(pQuery)) {
@@ -5196,20 +5212,10 @@ static bool validateQuerySourceCols(SQueryTableMsg *pQueryMsg, SSqlFuncMsg** pEx
 static char *createTableIdList(SQueryTableMsg *pQueryMsg, char *pMsg, SArray **pTableIdList) {
   assert(pQueryMsg->numOfTables > 0);
 
-  *pTableIdList = taosArrayInit(pQueryMsg->numOfTables, sizeof(STableId));
+  *pTableIdList = taosArrayInit(pQueryMsg->numOfTables, sizeof(STableIdInfo));
 
+  for (int32_t j = 0; j < pQueryMsg->numOfTables; ++j) {
     STableIdInfo* pTableIdInfo = (STableIdInfo *)pMsg;
-  pTableIdInfo->tid = htonl(pTableIdInfo->tid);
-  pTableIdInfo->uid = htobe64(pTableIdInfo->uid);
-  pTableIdInfo->key = htobe64(pTableIdInfo->key);
-
-  STableId id = {.uid = pTableIdInfo->uid, .tid = pTableIdInfo->tid};
-  taosArrayPush(*pTableIdList, &id);
-
-  pMsg += sizeof(STableIdInfo);
-
-  for (int32_t j = 1; j < pQueryMsg->numOfTables; ++j) {
-    pTableIdInfo = (STableIdInfo *)pMsg;
 
     pTableIdInfo->tid = htonl(pTableIdInfo->tid);
     pTableIdInfo->uid = htobe64(pTableIdInfo->uid);
@@ -5660,7 +5666,16 @@ static void doUpdateExprColumnIndex(SQuery *pQuery) {
   }
 }
 
-static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGroupbyExpr, SExprInfo *pExprs,
+
+static int compareTableIdInfo( const void* a, const void* b ) {
+  const STableIdInfo* x = (const STableIdInfo*)a;
+  const STableIdInfo* y = (const STableIdInfo*)b;
+  if (x->uid > y->uid) return 1;
+  if (x->uid < y->uid) return -1;
+  return 0;
+}
+
+static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SArray* pTableIdList, SSqlGroupbyExpr *pGroupbyExpr, SExprInfo *pExprs,
                                STableGroupInfo *groupInfo, SColumnInfo* pTagCols) {
   SQInfo *pQInfo = (SQInfo *)calloc(1, sizeof(SQInfo));
   if (pQInfo == NULL) {
@@ -5753,6 +5768,9 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou
   pQInfo->groupInfo.pGroupList = taosArrayInit(numOfGroups, POINTER_BYTES);
   pQInfo->groupInfo.numOfTables = groupInfo->numOfTables;
   
+  int tableIndex = 0;
+  STimeWindow window = pQueryMsg->window;
+  taosArraySort( pTableIdList, compareTableIdInfo );
   for(int32_t i = 0; i < numOfGroups; ++i) {
     SArray* pa = taosArrayGetP(groupInfo->pGroupList, i);
     size_t s = taosArrayGetSize(pa);
@@ -5760,13 +5778,26 @@ static SQInfo *createQInfoImpl(SQueryTableMsg *pQueryMsg, SSqlGroupbyExpr *pGrou
     SArray* p1 = taosArrayInit(s, sizeof(SGroupItem));
     
     for(int32_t j = 0; j < s; ++j) {
-      SGroupItem item = { .id = *(STableId*) taosArrayGet(pa, j), .info = NULL, };
+      STableId id = *(STableId*) taosArrayGet(pa, j);
+      SGroupItem item = { .id = id };
+      // NOTE: compare STableIdInfo with STableId
+      // not a problem at present because we only use their 1st int64_t field
+      STableIdInfo* pTableId = taosArraySearch( pTableIdList, compareTableIdInfo, &id );
+      if (pTableId != NULL ) {
+        window.skey = pTableId->key;
+      } else {
+        window.skey = pQueryMsg->window.skey;
+      }
+      item.info = createTableQueryInfo(&pQInfo->runtimeEnv, item.id, window);
+      item.info->groupIdx = i;
+      item.info->tableIndex = tableIndex++;
       taosArrayPush(p1, &item);
     }
-    
     taosArrayPush(pQInfo->groupInfo.pGroupList, &p1);
   }
 
+  pQInfo->arrTableIdInfo = taosArrayInit(tableIndex, sizeof(STableIdInfo));
+
   pQuery->pos = -1;
 
   pQuery->window = pQueryMsg->window;
@@ -5918,6 +5949,7 @@ static void freeQInfo(SQInfo *pQInfo) {
   }
   
   taosArrayDestroy(pQInfo->tableIdGroupInfo.pGroupList);
+  taosArrayDestroy(pQInfo->arrTableIdInfo);
   
   if (pQuery->pGroupbyExpr != NULL) {
     taosArrayDestroy(pQuery->pGroupbyExpr->columnInfo);
@@ -6045,13 +6077,17 @@ int32_t qCreateQueryInfo(void *tsdb, int32_t vgId, SQueryTableMsg *pQueryMsg, qi
   if (TSDB_QUERY_HAS_TYPE(pQueryMsg->queryType, TSDB_QUERY_TYPE_MULTITABLE_QUERY|TSDB_QUERY_TYPE_TABLE_QUERY)) {
     isSTableQuery = TSDB_QUERY_HAS_TYPE(pQueryMsg->queryType, TSDB_QUERY_TYPE_MULTITABLE_QUERY);
     
-    STableId *id = taosArrayGet(pTableIdList, 0);
+    STableIdInfo *id = taosArrayGet(pTableIdList, 0);
     if ((code = tsdbGetOneTableGroup(tsdb, id->uid, &groupInfo)) != TSDB_CODE_SUCCESS) {
       goto _over;
     }
   } else if (TSDB_QUERY_HAS_TYPE(pQueryMsg->queryType, TSDB_QUERY_TYPE_STABLE_QUERY)) {
     isSTableQuery = true;
-    STableId *id = taosArrayGet(pTableIdList, 0);
+    // TODO: need a macro from TSDB to check if table is super table,
+    // also note there's possiblity that only one table in the super table
+    if (taosArrayGetSize(pTableIdList) == 1) {
+      STableIdInfo *id = taosArrayGet(pTableIdList, 0);
+      // if array size is 1 and assert super table
 
       // group by normal column, do not pass the group by condition to tsdb to group table into different group
       int32_t numOfGroupByCols = pQueryMsg->numOfGroupCols;
@@ -6066,11 +6102,23 @@ int32_t qCreateQueryInfo(void *tsdb, int32_t vgId, SQueryTableMsg *pQueryMsg, qi
         code = TSDB_CODE_SUCCESS;
         goto _over;
       }
+    } else {
+      groupInfo.numOfTables = taosArrayGetSize(pTableIdList);
+      SArray* pTableGroup = taosArrayInit(1, POINTER_BYTES);
+
+      SArray* sa = taosArrayInit(groupInfo.numOfTables, sizeof(STableId));
+      for(int32_t i = 0; i < groupInfo.numOfTables; ++i) {
+        STableIdInfo* tableId = taosArrayGet(pTableIdList, i);
+        taosArrayPush(sa, tableId);
+      }
+      taosArrayPush(pTableGroup, &sa);
+      groupInfo.pGroupList = pTableGroup;
+    }
   } else {
     assert(0);
   }
 
-  (*pQInfo) = createQInfoImpl(pQueryMsg, pGroupbyExpr, pExprs, &groupInfo, pTagColumnInfo);
+  (*pQInfo) = createQInfoImpl(pQueryMsg, pTableIdList, pGroupbyExpr, pExprs, &groupInfo, pTagColumnInfo);
   if ((*pQInfo) == NULL) {
     code = TSDB_CODE_SERV_OUT_OF_MEMORY;
     goto _over;
@@ -6168,6 +6216,8 @@ int32_t qDumpRetrieveResult(qinfo_t qinfo, SRetrieveTableRsp **pRsp, int32_t *co
 
   SQuery *pQuery = pQInfo->runtimeEnv.pQuery;
   size_t  size = getResultSize(pQInfo, &pQuery->rec.rows);
+  size += sizeof(int32_t);
+  size += sizeof(STableIdInfo) * taosArrayGetSize(pQInfo->arrTableIdInfo);
   *contLen = size + sizeof(SRetrieveTableRsp);
 
   // todo handle failed to allocate memory

src/tsdb/src/tsdbMain.c

@@ -161,7 +161,7 @@ static int tsdbRestoreInfo(STsdbRepo *pRepo) {
   SFileGroup *pFGroup = NULL;
 
   SFileGroupIter iter;
-  SRWHelper      rhelper = {0};
+  SRWHelper      rhelper = {{0}};
 
   if (tsdbInitReadHelper(&rhelper, pRepo) < 0) goto _err;
   tsdbInitFileGroupIter(pFileH, &iter, TSDB_ORDER_ASC);
@@ -284,6 +284,7 @@ int32_t tsdbCloseRepo(TsdbRepoT *repo) {
   pRepo->tsdbCache->curBlock = NULL;
   tsdbUnLockRepo(repo);
 
+  if (pRepo->appH.notifyStatus) pRepo->appH.notifyStatus(pRepo->appH.appH, TSDB_STATUS_COMMIT_START);
   tsdbCommitData((void *)repo);
 
   tsdbCloseFileH(pRepo->tsdbFileH);
@@ -330,7 +331,7 @@ int32_t tsdbConfigRepo(TsdbRepoT *repo, STsdbCfg *pCfg) {
 int32_t tsdbTriggerCommit(TsdbRepoT *repo) {
   STsdbRepo *pRepo = (STsdbRepo *)repo;
 
-  if (pRepo->appH.walCallBack) pRepo->appH.walCallBack(pRepo->appH.appH);
+  if (pRepo->appH.notifyStatus) pRepo->appH.notifyStatus(pRepo->appH.appH, TSDB_STATUS_COMMIT_START);
   
   tsdbLockRepo(repo);
   if (pRepo->commit) {
@@ -942,15 +943,16 @@ static void tsdbFreeMemTable(SMemTable *pMemTable) {
 
 // Commit to file
 static void *tsdbCommitData(void *arg) {
-  printf("Starting to commit....\n");
   STsdbRepo * pRepo = (STsdbRepo *)arg;
   STsdbMeta * pMeta = pRepo->tsdbMeta;
   STsdbCache *pCache = pRepo->tsdbCache;
   STsdbCfg *  pCfg = &(pRepo->config);
   SDataCols * pDataCols = NULL;
-  SRWHelper   whelper = {0};
+  SRWHelper   whelper = {{0}};
   if (pCache->imem == NULL) return NULL;
 
+  tsdbPrint("vgId: %d, starting to commit....", pRepo->config.tsdbId);
+
   // Create the iterator to read from cache
   SSkipListIterator **iters = tsdbCreateTableIters(pMeta, pCfg->maxTables);
   if (iters == NULL) {
@@ -974,6 +976,7 @@ static void *tsdbCommitData(void *arg) {
 
   // Do retention actions
   tsdbFitRetention(pRepo);
+  if (pRepo->appH.notifyStatus) pRepo->appH.notifyStatus(pRepo->appH.appH, TSDB_STATUS_COMMIT_OVER);
 
 _exit:
   tdFreeDataCols(pDataCols);

src/tsdb/src/tsdbRead.c

@@ -54,12 +54,12 @@ typedef struct SQueryFilePos {
 typedef struct SDataBlockLoadInfo {
   SFileGroup* fileGroup;
   int32_t     slot;
-  int32_t     sid;
+  int32_t     tid;
   SArray*     pLoadedCols;
 } SDataBlockLoadInfo;
 
 typedef struct SLoadCompBlockInfo {
-  int32_t sid; /* table sid */
+  int32_t tid; /* table tid */
   int32_t fileId;
   int32_t fileListIndex;
 } SLoadCompBlockInfo;
@@ -127,12 +127,12 @@ static void changeQueryHandleForLastrowQuery(TsdbQueryHandleT pqHandle);
 
 static void tsdbInitDataBlockLoadInfo(SDataBlockLoadInfo* pBlockLoadInfo) {
   pBlockLoadInfo->slot = -1;
-  pBlockLoadInfo->sid = -1;
+  pBlockLoadInfo->tid = -1;
   pBlockLoadInfo->fileGroup = NULL;
 }
 
 static void tsdbInitCompBlockLoadInfo(SLoadCompBlockInfo* pCompBlockLoadInfo) {
-  pCompBlockLoadInfo->sid = -1;
+  pCompBlockLoadInfo->tid = -1;
   pCompBlockLoadInfo->fileId = -1;
   pCompBlockLoadInfo->fileListIndex = -1;
 }
@@ -423,7 +423,7 @@ static bool doLoadFileDataBlock(STsdbQueryHandle* pQueryHandle, SCompBlock* pBlo
   bool    blockLoaded = false;
   SArray* sa = getDefaultLoadColumns(pQueryHandle, true);
 
-  if (pCheckInfo->pDataCols == NULL) {
+  if (pCheckInfo->pDataCols == NULL) { // todo: why not the real data?
     pCheckInfo->pDataCols = tdNewDataCols(pRepo->tsdbMeta->maxRowBytes, pRepo->tsdbMeta->maxCols, pRepo->config.maxRowsPerFileBlock);
   }
 
@@ -434,7 +434,7 @@ static bool doLoadFileDataBlock(STsdbQueryHandle* pQueryHandle, SCompBlock* pBlo
 
     pBlockLoadInfo->fileGroup = pQueryHandle->pFileGroup;
     pBlockLoadInfo->slot = pQueryHandle->cur.slot;
-    pBlockLoadInfo->sid = pCheckInfo->pTableObj->tableId.tid;
+    pBlockLoadInfo->tid = pCheckInfo->pTableObj->tableId.tid;
 
     blockLoaded = true;
   }
@@ -614,7 +614,7 @@ static void filterDataInDataBlock(STsdbQueryHandle* pQueryHandle, STableCheckInf
     SColumnInfoData* pCol = taosArrayGet(pQueryHandle->pColumns, i);
     int32_t bytes = pCol->info.bytes;
 
-    for (int32_t j = 0; j < numOfCols; ++j) {
+    for (int32_t j = 0; j < numOfCols; ++j) {  //todo opt performance
       SDataCol* src = &pQueryHandle->rhelper.pDataCols[0]->cols[j];
 
       if (pCol->info.colId == src->colId) {
@@ -1197,7 +1197,9 @@ SArray* tsdbRetrieveDataBlock(TsdbQueryHandleT* pQueryHandle, SArray* pIdList) {
     } else {
       // data block has been loaded, todo extract method
       SDataBlockLoadInfo* pBlockLoadInfo = &pHandle->dataBlockLoadInfo;
-      if (pBlockLoadInfo->slot == pHandle->cur.slot && pBlockLoadInfo->sid == pCheckInfo->pTableObj->tableId.tid) {
+      
+      if (pBlockLoadInfo->slot == pHandle->cur.slot && pBlockLoadInfo->fileGroup->fileId == pHandle->cur.fid &&
+          pBlockLoadInfo->tid == pCheckInfo->pTableObj->tableId.tid) {
         return pHandle->pColumns;
       } else {
         SCompBlock* pBlock = pBlockInfoEx->pBlock.compBlock;
@@ -1347,7 +1349,7 @@ int32_t tableGroupComparFn(const void *p1, const void *p2, const void *param) {
     int32_t type = 0;
     int32_t bytes = 0;
     
-    if (colIndex == TSDB_TBNAME_COLUMN_INDEX) {
+    if (colIndex == TSDB_TBNAME_COLUMN_INDEX) {  // todo refactor extract method , to queryExecutor to generate tags values
       f1 = (char*) pTable1->name;
       f2 = (char*) pTable2->name;
       type = TSDB_DATA_TYPE_BINARY;
@@ -1355,6 +1357,7 @@ int32_t tableGroupComparFn(const void *p1, const void *p2, const void *param) {
     } else {
       STColumn* pCol = schemaColAt(pTableGroupSupp->pTagSchema, colIndex);
       bytes = pCol->bytes;
+      type = pCol->type;
       
       f1 = tdGetRowDataOfCol(pTable1->tagVal, pCol->type, TD_DATA_ROW_HEAD_SIZE + pCol->offset);
       f2 = tdGetRowDataOfCol(pTable2->tagVal, pCol->type, TD_DATA_ROW_HEAD_SIZE + pCol->offset);
@@ -1522,7 +1525,7 @@ int32_t tsdbQuerySTableByTagCond(TsdbRepoT *tsdb, int64_t uid, const char *pTagC
   }
   
   if (pTable->type != TSDB_SUPER_TABLE) {
-    uError("%p query normal tag not allowed, uid:%, tid:%d, name:%s" PRIu64,
+    uError("%p query normal tag not allowed, uid:%" PRIu64 ", tid:%d, name:%s",
            tsdb, uid, pTable->tableId.tid, pTable->name);
     
     return TSDB_CODE_OPS_NOT_SUPPORT; //basically, this error is caused by invalid sql issued by client

src/util/inc/exception.h

@@ -73,6 +73,7 @@ void cleanupPush_void_ptr_bool  ( bool failOnly, void* func, void* arg1, bool ar
 void cleanupPush_void_ptr       ( bool failOnly, void* func, void* arg );
 void cleanupPush_int_int        ( bool failOnly, void* func, int arg );
 void cleanupPush_void           ( bool failOnly, void* func );
+void cleanupPush_int_ptr        ( bool failOnly, void* func, void* arg );
 
 int32_t cleanupGetActionCount();
 void cleanupExecuteTo( int32_t anchor, bool failed );
@@ -83,8 +84,10 @@ void cleanupExecute( SExceptionNode* node, bool failed );
 #define CLEANUP_PUSH_VOID_PTR( failOnly, func, arg )             cleanupPush_void_ptr( (failOnly), (void*)(func), (void*)(arg) )
 #define CLEANUP_PUSH_INT_INT( failOnly, func, arg )              cleanupPush_void_ptr( (failOnly), (void*)(func), (int)(arg) )
 #define CLEANUP_PUSH_VOID( failOnly, func )                      cleanupPush_void( (failOnly), (void*)(func) )
+#define CLEANUP_PUSH_INT_PTR( failOnly, func, arg )              cleanupPush_int_ptr( (failOnly), (void*)(func), (void*)(arg) )
 #define CLEANUP_PUSH_FREE( failOnly, arg )                       cleanupPush_void_ptr( (failOnly), free, (void*)(arg) )
 #define CLEANUP_PUSH_CLOSE( failOnly, arg )                      cleanupPush_int_int( (failOnly), close, (int)(arg) )
+#define CLEANUP_PUSH_FCLOSE( failOnly, arg )                     cleanupPush_int_ptr( (failOnly), fclose, (void*)(arg) )
 
 #define CLEANUP_GET_ANCHOR()          cleanupGetActionCount()
 #define CLEANUP_EXECUTE_TO( anchor, failed )  cleanupExecuteTo( (anchor), (failed) )
@@ -95,7 +98,7 @@ void cleanupExecute( SExceptionNode* node, bool failed );
 
 void exceptionPushNode( SExceptionNode* node );
 int32_t exceptionPopNode();
-void exceptionThrow( int code );
+void exceptionThrow( int32_t code );
 
 #define TRY(maxCleanupActions) do { \
     SExceptionNode exceptionNode = { 0 }; \
@@ -106,10 +109,10 @@ void exceptionThrow( int code );
     int caughtException = setjmp( exceptionNode.jb ); \
     if( caughtException == 0 )
 
-#define CATCH( code ) int code = exceptionPopNode(); \
+#define CATCH( code ) int32_t code = exceptionPopNode(); \
     if( caughtException == 1 )
 
-#define FINALLY( code ) int code = exceptionPopNode();
+#define FINALLY( code ) int32_t code = exceptionPopNode();
 
 #define END_TRY } while( 0 );
 

src/util/inc/hash.h

@@ -94,7 +94,7 @@ size_t taosHashGetSize(const SHashObj *pHashObj);
  * @param size
  * @return
  */
-int32_t taosHashPut(SHashObj *pHashObj, const char *key, size_t keyLen, void *data, size_t size);
+int32_t taosHashPut(SHashObj *pHashObj, const void *key, size_t keyLen, void *data, size_t size);
 
 /**
  * return the payload data with the specified key
@@ -104,7 +104,7 @@ int32_t taosHashPut(SHashObj *pHashObj, const char *key, size_t keyLen, void *da
  * @param keyLen
  * @return
  */
-void *taosHashGet(SHashObj *pHashObj, const char *key, size_t keyLen);
+void *taosHashGet(SHashObj *pHashObj, const void *key, size_t keyLen);
 
 /**
  * remove item with the specified key
@@ -112,7 +112,7 @@ void *taosHashGet(SHashObj *pHashObj, const char *key, size_t keyLen);
  * @param key
  * @param keyLen
  */
-void taosHashRemove(SHashObj *pHashObj, const char *key, size_t keyLen);
+void taosHashRemove(SHashObj *pHashObj, const void *key, size_t keyLen);
 
 /**
  * clean up hash table

src/util/inc/tarray.h

@@ -106,6 +106,12 @@ void taosArrayCopy(SArray* pDst, const SArray* pSrc);
  */
 SArray* taosArrayClone(const SArray* pSrc);
 
+/**
+ * clear the array (remove all element)
+ * @param pArray
+ */
+void taosArrayClear(SArray* pArray);
+
 /**
  * destroy array list
  * @param pArray

src/util/src/exception.c

@@ -1,7 +1,8 @@
+#include "os.h"
 #include "exception.h"
 
 
-static _Thread_local SExceptionNode* expList;
+static threadlocal SExceptionNode* expList;
 
 void exceptionPushNode( SExceptionNode* node ) {
     node->prev = expList;
@@ -14,7 +15,7 @@ int32_t exceptionPopNode() {
     return node->code;
 }
 
-void exceptionThrow( int code ) {
+void exceptionThrow( int32_t code ) {
     expList->code = code;
     longjmp( expList->jb, 1 );
 }
@@ -38,21 +39,27 @@ static void cleanupWrapper_void_ptr( SCleanupAction* ca ) {
 
 static void cleanupWrapper_int_int( SCleanupAction* ca ) {
     int (*func)( int ) = ca->func;
-    func( (int)(intptr_t)(ca->arg1.Int) );
+    func( ca->arg1.Int );
 }
 
-static void cleanupWrapper_void_void( SCleanupAction* ca ) {
+static void cleanupWrapper_void( SCleanupAction* ca ) {
     void (*func)() = ca->func;
     func();
 }
 
+static void cleanupWrapper_int_ptr( SCleanupAction* ca ) {
+    int (*func)( void* ) = ca->func;
+    func( ca->arg1.Ptr );
+}
+
 typedef void (*wrapper)(SCleanupAction*);
 static wrapper wrappers[] = {
     cleanupWrapper_void_ptr_ptr,
     cleanupWrapper_void_ptr_bool,
     cleanupWrapper_void_ptr,
     cleanupWrapper_int_int,
-    cleanupWrapper_void_void,
+    cleanupWrapper_void,
+    cleanupWrapper_int_ptr,
 };
 
 
@@ -107,6 +114,15 @@ void cleanupPush_void( bool failOnly, void* func ) {
     ca->func = func;
 }
 
+void cleanupPush_int_ptr( bool failOnly, void* func, void* arg ) {
+    assert( expList->numCleanupAction < expList->maxCleanupAction );
+
+    SCleanupAction *ca = expList->cleanupActions + expList->numCleanupAction++;
+    ca->wrapper = 5;
+    ca->failOnly = failOnly;
+    ca->func = func;
+    ca->arg1.Ptr = arg;
+}
 
 
 int32_t cleanupGetActionCount() {
@@ -118,10 +134,11 @@ static void doExecuteCleanup( SExceptionNode* node, int32_t anchor, bool failed
     while( node->numCleanupAction > anchor ) {
         --node->numCleanupAction;
         SCleanupAction *ca = node->cleanupActions + node->numCleanupAction;
-        if( failed || !(ca->failOnly) )
+        if( failed || !(ca->failOnly) ) {
             wrappers[ca->wrapper]( ca );
         }
     }
+}
 
 void cleanupExecuteTo( int32_t anchor, bool failed ) {
     doExecuteCleanup( expList, anchor, failed );

src/util/src/hash.c

@@ -17,7 +17,6 @@
 
 #include "hash.h"
 #include "tulog.h"
-#include "ttime.h"
 #include "tutil.h"
 
 static FORCE_INLINE void __wr_lock(void *lock) {
@@ -91,153 +90,64 @@ static FORCE_INLINE int32_t taosHashCapacity(int32_t length) {
 /**
  * inplace update node in hash table
  * @param pHashObj  hash table object
- * @param pNode     data node
+ * @param pNode     hash data node
  */
-static void doUpdateHashTable(SHashObj *pHashObj, SHashNode *pNode) {
+static void doUpdateHashTable(SHashObj *pHashObj, SHashNode *pNode);
-  if (pNode->prev1) {
-    pNode->prev1->next = pNode;
-  }
-
-  if (pNode->next) {
-    (pNode->next)->prev = pNode;
-  }
-}
 
 /**
- * get SHashNode from hashlist, nodes from trash are not included.
+ * Get SHashNode from hashlist, nodes from trash are not included.
  * @param pHashObj  Cache objection
  * @param key       key for hash
  * @param keyLen    key length
+ * @param hashVal   hash value by hash function
  * @return
  */
-static SHashNode *doGetNodeFromHashTable(SHashObj *pHashObj, const char *key, uint32_t keyLen, uint32_t *hashVal) {
+static SHashNode *doGetNodeFromHashTable(SHashObj *pHashObj, const void *key, uint32_t keyLen, uint32_t *hashVal);
-  uint32_t hash = (*pHashObj->hashFp)(key, keyLen);
-
-  int32_t     slot = HASH_INDEX(hash, pHashObj->capacity);
-  SHashEntry *pEntry = pHashObj->hashList[slot];
-
-  SHashNode *pNode = pEntry->next;
-  while (pNode) {
-    if ((pNode->keyLen == keyLen) && (memcmp(pNode->key, key, keyLen) == 0)) {
-      break;
-    }
-
-    pNode = pNode->next;
-  }
-
-  if (pNode) {
-    assert(HASH_INDEX(pNode->hashVal, pHashObj->capacity) == slot);
-  }
-
-  // return the calculated hash value, to avoid calculating it again in other functions
-  if (hashVal != NULL) {
-    *hashVal = hash;
-  }
-
-  return pNode;
-}
 
 /**
- * resize the hash list if the threshold is reached
+ * Resize the hash list if the threshold is reached
  *
  * @param pHashObj
  */
-static void taosHashTableResize(SHashObj *pHashObj) {
+static void taosHashTableResize(SHashObj *pHashObj);
-  if (pHashObj->size < pHashObj->capacity * HASH_DEFAULT_LOAD_FACTOR) {
-    return;
-  }
-
-  // double the original capacity
-  SHashNode *pNode = NULL;
-  SHashNode *pNext = NULL;
-
-  int32_t newSize = pHashObj->capacity << 1u;
-  if (newSize > HASH_MAX_CAPACITY) {
-//    uTrace("current capacity:%d, maximum capacity:%d, no resize applied due to limitation is reached",
-//           pHashObj->capacity, HASH_MAX_CAPACITY);
-    return;
-  }
-
-//  int64_t st = taosGetTimestampUs();
-
-  SHashEntry **pNewEntry = realloc(pHashObj->hashList, sizeof(SHashEntry *) * newSize);
-  if (pNewEntry == NULL) {
-//    uTrace("cache resize failed due to out of memory, capacity remain:%d", pHashObj->capacity);
-    return;
-  }
-
-  pHashObj->hashList = pNewEntry;
-  for (int32_t i = pHashObj->capacity; i < newSize; ++i) {
-    pHashObj->hashList[i] = calloc(1, sizeof(SHashEntry));
-  }
-
-  pHashObj->capacity = newSize;
-
-  for (int32_t i = 0; i < pHashObj->capacity; ++i) {
-    SHashEntry *pEntry = pHashObj->hashList[i];
-
-    pNode = pEntry->next;
-    if (pNode != NULL) {
-      assert(pNode->prev1 == pEntry && pEntry->num > 0);
-    }
-
-    while (pNode) {
-      int32_t j = HASH_INDEX(pNode->hashVal, pHashObj->capacity);
-      if (j == i) {  // this key resides in the same slot, no need to relocate it
-        pNode = pNode->next;
-      } else {
-        pNext = pNode->next;
-
-        // remove from current slot
-        assert(pNode->prev1 != NULL);
-
-        if (pNode->prev1 == pEntry) {  // first node of the overflow linked list
-          pEntry->next = pNode->next;
-        } else {
-          pNode->prev->next = pNode->next;
-        }
-
-        pEntry->num--;
-        assert(pEntry->num >= 0);
-
-        if (pNode->next != NULL) {
-          (pNode->next)->prev = pNode->prev;
-        }
-
-        // added into new slot
-        pNode->next = NULL;
-        pNode->prev1 = NULL;
-
-        SHashEntry *pNewIndexEntry = pHashObj->hashList[j];
-
-        if (pNewIndexEntry->next != NULL) {
-          assert(pNewIndexEntry->next->prev1 == pNewIndexEntry);
-
-          pNewIndexEntry->next->prev = pNode;
-        }
-
-        pNode->next = pNewIndexEntry->next;
-        pNode->prev1 = pNewIndexEntry;
-
-        pNewIndexEntry->next = pNode;
-        pNewIndexEntry->num++;
-
-        // continue
-        pNode = pNext;
-      }
-    }
-  }
-
-//  int64_t et = taosGetTimestampUs();
-//  uTrace("hash table resize completed, new capacity:%d, load factor:%f, elapsed time:%fms", pHashObj->capacity,
-//         ((double)pHashObj->size) / pHashObj->capacity, (et - st) / 1000.0);
-}
 
 /**
- * @param capacity maximum slots available for hash elements
+ * @param key      key of object for hash, usually a null-terminated string
- * @param fn       hash function
+ * @param keyLen   length of key
+ * @param pData    actually data. Requires a consecutive memory block, no pointer is allowed in pData.
+ *                 Pointer copy causes memory access error.
+ * @param dsize    size of data
+ * @return         SHashNode
+ */
+static SHashNode *doCreateHashNode(const void *key, size_t keyLen, const void *pData, size_t dsize, uint32_t hashVal);
+
+/**
+ * Update the hash node
+ *
+ * @param pNode   hash node
+ * @param key     key for generate hash value
+ * @param keyLen  key length
+ * @param pData   actual data
+ * @param dsize   size of actual data
+ * @return        hash node
+ */
+static SHashNode *doUpdateHashNode(SHashNode *pNode, const void *key, size_t keyLen, const void *pData, size_t dsize);
+
+/**
+ * insert the hash node at the front of the linked list
+ *
+ * @param pHashObj
+ * @param pNode
+ */
+static void doAddToHashTable(SHashObj *pHashObj, SHashNode *pNode);
+
+/**
+ * Get the next element in hash table for iterator
+ * @param pIter
  * @return
  */
+static SHashNode *getNextHashNode(SHashMutableIterator *pIter);
+
 SHashObj *taosHashInit(size_t capacity, _hash_fn_t fn, bool threadsafe) {
   if (capacity == 0 || fn == NULL) {
     return NULL;
@@ -285,79 +195,6 @@ SHashObj *taosHashInit(size_t capacity, _hash_fn_t fn, bool threadsafe) {
   return pHashObj;
 }
 
-/**
- * @param key      key of object for hash, usually a null-terminated string
- * @param keyLen   length of key
- * @param pData    actually data. required a consecutive memory block, no pointer is allowed
- *                 in pData. Pointer copy causes memory access error.
- * @param size     size of block
- * @return         SHashNode
- */
-static SHashNode *doCreateHashNode(const char *key, size_t keyLen, const char *pData, size_t dataSize,
-                                   uint32_t hashVal) {
-  size_t totalSize = dataSize + sizeof(SHashNode) + keyLen + 1;  // one extra byte for null
-
-  SHashNode *pNewNode = calloc(1, totalSize);
-  if (pNewNode == NULL) {
-    uError("failed to allocate memory, reason:%s", strerror(errno));
-    return NULL;
-  }
-
-  memcpy(pNewNode->data, pData, dataSize);
-
-  pNewNode->key = pNewNode->data + dataSize;
-  memcpy(pNewNode->key, key, keyLen);
-  pNewNode->keyLen = keyLen;
-
-  pNewNode->hashVal = hashVal;
-
-  return pNewNode;
-}
-
-static SHashNode *doUpdateHashNode(SHashNode *pNode, const char *key, size_t keyLen, const char *pData,
-                                   size_t dataSize) {
-  size_t size = dataSize + sizeof(SHashNode) + keyLen;
-
-  SHashNode *pNewNode = (SHashNode *)realloc(pNode, size);
-  if (pNewNode == NULL) {
-    return NULL;
-  }
-
-  memcpy(pNewNode->data, pData, dataSize);
-
-  pNewNode->key = pNewNode->data + dataSize;
-
-  assert(memcmp(pNewNode->key, key, keyLen) == 0 && keyLen == pNewNode->keyLen);
-
-  memcpy(pNewNode->key, key, keyLen);
-  return pNewNode;
-}
-
-/**
- * insert the hash node at the front of the linked list
- *
- * @param pHashObj
- * @param pNode
- */
-static void doAddToHashTable(SHashObj *pHashObj, SHashNode *pNode) {
-  assert(pNode != NULL);
-
-  int32_t     index = HASH_INDEX(pNode->hashVal, pHashObj->capacity);
-  SHashEntry *pEntry = pHashObj->hashList[index];
-
-  pNode->next = pEntry->next;
-
-  if (pEntry->next) {
-    pEntry->next->prev = pNode;
-  }
-
-  pEntry->next = pNode;
-  pNode->prev1 = pEntry;
-
-  pEntry->num++;
-  pHashObj->size++;
-}
-
 size_t taosHashGetSize(const SHashObj *pHashObj) {
   if (pHashObj == NULL) {
     return 0;
@@ -366,12 +203,7 @@ size_t taosHashGetSize(const SHashObj *pHashObj) {
   return pHashObj->size;
 }
 
-/**
+int32_t taosHashPut(SHashObj *pHashObj, const void *key, size_t keyLen, void *data, size_t size) {
- * add data node into hash table
- * @param pHashObj    hash object
- * @param pNode   hash node
- */
-int32_t taosHashPut(SHashObj *pHashObj, const char *key, size_t keyLen, void *data, size_t size) {
   __wr_lock(pHashObj->lock);
 
   uint32_t   hashVal = 0;
@@ -402,7 +234,7 @@ int32_t taosHashPut(SHashObj *pHashObj, const char *key, size_t keyLen, void *da
   return 0;
 }
 
-void *taosHashGet(SHashObj *pHashObj, const char *key, size_t keyLen) {
+void *taosHashGet(SHashObj *pHashObj, const void *key, size_t keyLen) {
   __rd_lock(pHashObj->lock);
 
   uint32_t   hashVal = 0;
@@ -419,12 +251,7 @@ void *taosHashGet(SHashObj *pHashObj, const char *key, size_t keyLen) {
   }
 }
 
-/**
+void taosHashRemove(SHashObj *pHashObj, const void *key, size_t keyLen) {
- * remove node in hash list
- * @param pHashObj
- * @param pNode
- */
-void taosHashRemove(SHashObj *pHashObj, const char *key, size_t keyLen) {
   __wr_lock(pHashObj->lock);
 
   uint32_t   val = 0;
@@ -518,23 +345,6 @@ SHashMutableIterator *taosHashCreateIter(SHashObj *pHashObj) {
   return pIter;
 }
 
-static SHashNode *getNextHashNode(SHashMutableIterator *pIter) {
-  assert(pIter != NULL);
-
-  pIter->entryIndex++;
-  while (pIter->entryIndex < pIter->pHashObj->capacity) {
-    SHashEntry *pEntry = pIter->pHashObj->hashList[pIter->entryIndex];
-    if (pEntry->next == NULL) {
-      pIter->entryIndex++;
-      continue;
-    }
-
-    return pEntry->next;
-  }
-
-  return NULL;
-}
-
 bool taosHashIterNext(SHashMutableIterator *pIter) {
   if (pIter == NULL) {
     return false;
@@ -617,3 +427,205 @@ int32_t taosHashGetMaxOverflowLinkLength(const SHashObj *pHashObj) {
 
   return num;
 }
+
+void doUpdateHashTable(SHashObj *pHashObj, SHashNode *pNode) {
+  if (pNode->prev1) {
+    pNode->prev1->next = pNode;
+  }
+  
+  if (pNode->next) {
+    (pNode->next)->prev = pNode;
+  }
+}
+
+SHashNode *doGetNodeFromHashTable(SHashObj *pHashObj, const void *key, uint32_t keyLen, uint32_t *hashVal) {
+  uint32_t hash = (*pHashObj->hashFp)(key, keyLen);
+  
+  int32_t     slot = HASH_INDEX(hash, pHashObj->capacity);
+  SHashEntry *pEntry = pHashObj->hashList[slot];
+  
+  SHashNode *pNode = pEntry->next;
+  while (pNode) {
+    if ((pNode->keyLen == keyLen) && (memcmp(pNode->key, key, keyLen) == 0)) {
+      break;
+    }
+    
+    pNode = pNode->next;
+  }
+  
+  if (pNode) {
+    assert(HASH_INDEX(pNode->hashVal, pHashObj->capacity) == slot);
+  }
+  
+  // return the calculated hash value, to avoid calculating it again in other functions
+  if (hashVal != NULL) {
+    *hashVal = hash;
+  }
+  
+  return pNode;
+}
+
+void taosHashTableResize(SHashObj *pHashObj) {
+  if (pHashObj->size < pHashObj->capacity * HASH_DEFAULT_LOAD_FACTOR) {
+    return;
+  }
+  
+  // double the original capacity
+  SHashNode *pNode = NULL;
+  SHashNode *pNext = NULL;
+  
+  int32_t newSize = pHashObj->capacity << 1u;
+  if (newSize > HASH_MAX_CAPACITY) {
+//    uTrace("current capacity:%d, maximum capacity:%d, no resize applied due to limitation is reached",
+//           pHashObj->capacity, HASH_MAX_CAPACITY);
+    return;
+  }
+
+//  int64_t st = taosGetTimestampUs();
+  
+  SHashEntry **pNewEntry = realloc(pHashObj->hashList, sizeof(SHashEntry *) * newSize);
+  if (pNewEntry == NULL) {
+//    uTrace("cache resize failed due to out of memory, capacity remain:%d", pHashObj->capacity);
+    return;
+  }
+  
+  pHashObj->hashList = pNewEntry;
+  for (int32_t i = pHashObj->capacity; i < newSize; ++i) {
+    pHashObj->hashList[i] = calloc(1, sizeof(SHashEntry));
+  }
+  
+  pHashObj->capacity = newSize;
+  
+  for (int32_t i = 0; i < pHashObj->capacity; ++i) {
+    SHashEntry *pEntry = pHashObj->hashList[i];
+    
+    pNode = pEntry->next;
+    if (pNode != NULL) {
+      assert(pNode->prev1 == pEntry && pEntry->num > 0);
+    }
+    
+    while (pNode) {
+      int32_t j = HASH_INDEX(pNode->hashVal, pHashObj->capacity);
+      if (j == i) {  // this key resides in the same slot, no need to relocate it
+        pNode = pNode->next;
+      } else {
+        pNext = pNode->next;
+        
+        // remove from current slot
+        assert(pNode->prev1 != NULL);
+        
+        if (pNode->prev1 == pEntry) {  // first node of the overflow linked list
+          pEntry->next = pNode->next;
+        } else {
+          pNode->prev->next = pNode->next;
+        }
+        
+        pEntry->num--;
+        assert(pEntry->num >= 0);
+        
+        if (pNode->next != NULL) {
+          (pNode->next)->prev = pNode->prev;
+        }
+        
+        // added into new slot
+        pNode->next = NULL;
+        pNode->prev1 = NULL;
+        
+        SHashEntry *pNewIndexEntry = pHashObj->hashList[j];
+        
+        if (pNewIndexEntry->next != NULL) {
+          assert(pNewIndexEntry->next->prev1 == pNewIndexEntry);
+          
+          pNewIndexEntry->next->prev = pNode;
+        }
+        
+        pNode->next = pNewIndexEntry->next;
+        pNode->prev1 = pNewIndexEntry;
+        
+        pNewIndexEntry->next = pNode;
+        pNewIndexEntry->num++;
+        
+        // continue
+        pNode = pNext;
+      }
+    }
+  }
+
+//  int64_t et = taosGetTimestampUs();
+//  uTrace("hash table resize completed, new capacity:%d, load factor:%f, elapsed time:%fms", pHashObj->capacity,
+//         ((double)pHashObj->size) / pHashObj->capacity, (et - st) / 1000.0);
+}
+
+SHashNode *doCreateHashNode(const void *key, size_t keyLen, const void *pData, size_t dsize, uint32_t hashVal) {
+  size_t totalSize = dsize + sizeof(SHashNode) + keyLen + 1;  // one extra byte for null
+  
+  SHashNode *pNewNode = calloc(1, totalSize);
+  if (pNewNode == NULL) {
+    uError("failed to allocate memory, reason:%s", strerror(errno));
+    return NULL;
+  }
+  
+  memcpy(pNewNode->data, pData, dsize);
+  
+  pNewNode->key = pNewNode->data + dsize;
+  memcpy(pNewNode->key, key, keyLen);
+  pNewNode->keyLen = keyLen;
+  
+  pNewNode->hashVal = hashVal;
+  
+  return pNewNode;
+}
+
+SHashNode *doUpdateHashNode(SHashNode *pNode, const void *key, size_t keyLen, const void *pData, size_t dsize) {
+  size_t size = dsize + sizeof(SHashNode) + keyLen;
+  
+  SHashNode *pNewNode = (SHashNode *)realloc(pNode, size);
+  if (pNewNode == NULL) {
+    return NULL;
+  }
+  
+  memcpy(pNewNode->data, pData, dsize);
+  
+  pNewNode->key = pNewNode->data + dsize;
+  
+  assert(memcmp(pNewNode->key, key, keyLen) == 0 && keyLen == pNewNode->keyLen);
+  
+  memcpy(pNewNode->key, key, keyLen);
+  return pNewNode;
+}
+
+void doAddToHashTable(SHashObj *pHashObj, SHashNode *pNode) {
+  assert(pNode != NULL);
+  
+  int32_t     index = HASH_INDEX(pNode->hashVal, pHashObj->capacity);
+  SHashEntry *pEntry = pHashObj->hashList[index];
+  
+  pNode->next = pEntry->next;
+  
+  if (pEntry->next) {
+    pEntry->next->prev = pNode;
+  }
+  
+  pEntry->next = pNode;
+  pNode->prev1 = pEntry;
+  
+  pEntry->num++;
+  pHashObj->size++;
+}
+
+SHashNode *getNextHashNode(SHashMutableIterator *pIter) {
+  assert(pIter != NULL);
+  
+  pIter->entryIndex++;
+  while (pIter->entryIndex < pIter->pHashObj->capacity) {
+    SHashEntry *pEntry = pIter->pHashObj->hashList[pIter->entryIndex];
+    if (pEntry->next == NULL) {
+      pIter->entryIndex++;
+      continue;
+    }
+    
+    return pEntry->next;
+  }
+  
+  return NULL;
+}

src/util/src/tarray.c

@@ -176,6 +176,11 @@ SArray* taosArrayClone(const SArray* pSrc) {
   return dst;
 }
 
+void taosArrayClear(SArray* pArray) {
+  assert( pArray != NULL );
+  pArray->size = 0;
+}
+
 void taosArrayDestroy(SArray* pArray) {
   if (pArray == NULL) {
     return;

src/util/src/tcache.c

@@ -446,6 +446,8 @@ void *taosCachePut(SCacheObj *pCacheObj, const char *key, const void *pData, siz
       
       uTrace("key:%s %p added into cache, added:%" PRIu64 ", expire:%" PRIu64 ", total:%d, size:%" PRId64 " bytes",
              key, pNode, pNode->addedTime, pNode->expiredTime, pCacheObj->totalSize, dataSize);
+    } else {
+      uError("key:%s failed to added into cache, out of memory", key);
     }
   } else {  // old data exists, update the node
     pNode = taosUpdateCacheImpl(pCacheObj, pOld, key, keyLen, pData, dataSize, duration * 1000L);

src/util/src/tconfig.c

@@ -25,7 +25,7 @@
 #include "tsystem.h"
 #include "tutil.h"
 
-SGlobalCfg tsGlobalConfig[TSDB_CFG_MAX_NUM] = {0};
+SGlobalCfg tsGlobalConfig[TSDB_CFG_MAX_NUM] = {{0}};
 int32_t    tsGlobalConfigNum = 0;
 
 static char *tsGlobalUnit[] = {

src/util/src/tdes.c

@@ -32,7 +32,7 @@ char* taosDesImp(unsigned char* key, char* src, unsigned int len, int process_mo
   unsigned int  number_of_blocks = len / 8;
   unsigned char data_block[9] = {0};
   unsigned char processed_block[9] = {0};
-  key_set       key_sets[17] = {0};
+  key_set       key_sets[17]; memset(key_sets, 0, sizeof(key_sets));
   char*         dest = calloc(len + 1, 1);
   generate_sub_keys(key, key_sets);
 

src/util/src/terror.c

@@ -27,10 +27,11 @@ typedef struct {
 } STaosError;
  
 
+#include "os.h"
 #include "taoserror.h"
 
 
-static _Thread_local int32_t tsErrno;
+static threadlocal int32_t tsErrno;
 int32_t* taosGetErrno() {
   return &tsErrno;
 }

src/vnode/inc/vnodeInt.h

@@ -38,6 +38,7 @@ typedef struct {
   int          status; 
   int8_t       role;   
   int64_t      version;
+  int64_t      savedVersion;
   void        *wqueue;
   void        *rqueue;
   void        *wal;

src/vnode/src/vnodeMain.c

@@ -37,7 +37,7 @@ static int32_t  vnodeSaveCfg(SMDCreateVnodeMsg *pVnodeCfg);
 static int32_t  vnodeReadCfg(SVnodeObj *pVnode);
 static int32_t  vnodeSaveVersion(SVnodeObj *pVnode);
 static bool     vnodeReadVersion(SVnodeObj *pVnode);
-static int      vnodeWalCallback(void *arg);
+static int      vnodeProcessTsdbStatus(void *arg, int status);
 static uint32_t vnodeGetFileInfo(void *ahandle, char *name, uint32_t *index, int32_t *size);
 static int      vnodeGetWalInfo(void *ahandle, char *name, uint32_t *index);
 static void     vnodeNotifyRole(void *ahandle, int8_t role);
@@ -205,7 +205,7 @@ int32_t vnodeOpen(int32_t vnode, char *rootDir) {
 
   STsdbAppH appH = {0};
   appH.appH = (void *)pVnode;
-  appH.walCallBack = vnodeWalCallback;
+  appH.notifyStatus = vnodeProcessTsdbStatus;
   appH.cqH = pVnode->cq;
 
   sprintf(temp, "%s/tsdb", rootDir);
@@ -383,16 +383,24 @@ static void vnodeCleanUp(SVnodeObj *pVnode) {
   walClose(pVnode->wal);
   pVnode->wal = NULL;
 
-  vnodeSaveVersion(pVnode);
   vnodeRelease(pVnode);
 }
 
 // TODO: this is a simple implement
-static int vnodeWalCallback(void *arg) {
+static int vnodeProcessTsdbStatus(void *arg, int status) {
   SVnodeObj *pVnode = arg;
+
+  if (status == TSDB_STATUS_COMMIT_START) {
+    pVnode->savedVersion = pVnode->version; 
     return walRenew(pVnode->wal);
   }
 
+  if (status == TSDB_STATUS_COMMIT_OVER)
+    return vnodeSaveVersion(pVnode);
+
+  return 0; 
+}
+
 static uint32_t vnodeGetFileInfo(void *ahandle, char *name, uint32_t *index, int32_t *size) {
   SVnodeObj *pVnode = ahandle;
   return tsdbGetFileInfo(pVnode->tsdb, name, index, size);
@@ -423,7 +431,7 @@ static void vnodeNotifyFileSynced(void *ahandle) {
   tsdbCloseRepo(pVnode->tsdb);
   STsdbAppH appH = {0};
   appH.appH = (void *)pVnode;
-  appH.walCallBack = vnodeWalCallback;
+  appH.notifyStatus = vnodeProcessTsdbStatus;
   appH.cqH = pVnode->cq;
   pVnode->tsdb = tsdbOpenRepo(rootDir, &appH);
 }
@@ -694,14 +702,14 @@ static int32_t vnodeSaveVersion(SVnodeObj *pVnode) {
   char *  content = calloc(1, maxLen + 1);
 
   len += snprintf(content + len, maxLen - len, "{\n");
-  len += snprintf(content + len, maxLen - len, "  \"version\": %" PRId64 "\n", pVnode->version);
+  len += snprintf(content + len, maxLen - len, "  \"version\": %" PRId64 "\n", pVnode->savedVersion);
   len += snprintf(content + len, maxLen - len, "}\n");
 
   fwrite(content, 1, len, fp);
   fclose(fp);
   free(content);
 
-  vPrint("vgId:%d, save vnode version:%" PRId64 " successed", pVnode->vgId, pVnode->version);
+  vPrint("vgId:%d, save vnode version:%" PRId64 " succeed", pVnode->vgId, pVnode->savedVersion);
 
   return 0;
 }
@@ -743,7 +751,7 @@ static bool vnodeReadVersion(SVnodeObj *pVnode) {
 
   ret = true;
 
-  vPrint("vgId:%d, read vnode version successed, version:%%" PRId64, pVnode->vgId, pVnode->version);
+  vPrint("vgId:%d, read vnode version succeed, version:%" PRId64, pVnode->vgId, pVnode->version);
 
 PARSE_OVER:
   free(content);

tests/examples/c/subscribe.c

@@ -56,32 +56,46 @@ void run_test(TAOS* taos) {
   taos_query(taos, "drop database if exists test;");
   
   usleep(100000);
-  taos_query(taos, "create database test tables 5;");
+  //taos_query(taos, "create database test tables 5;");
+  taos_query(taos, "create database test;");
   usleep(100000);
   taos_query(taos, "use test;");
-  usleep(100000);
-  taos_query(taos, "create table meters(ts timestamp, a int, b binary(20)) tags(loc binary(20), area int);");
 
-  taos_query(taos, "insert into t0 using meters tags('beijing', 0) values('2020-01-01 00:00:00.000', 0, 'china');");
+  usleep(100000);
-  taos_query(taos, "insert into t0 using meters tags('beijing', 0) values('2020-01-01 00:01:00.000', 0, 'china');");
+  taos_query(taos, "create table meters(ts timestamp, a int) tags(area int);");
-  taos_query(taos, "insert into t0 using meters tags('beijing', 0) values('2020-01-01 00:02:00.000', 0, 'china');");
+
-  taos_query(taos, "insert into t1 using meters tags('shanghai', 0) values('2020-01-01 00:00:00.000', 0, 'china');");
+  taos_query(taos, "create table t0 using meters tags(0);");
-  taos_query(taos, "insert into t1 using meters tags('shanghai', 0) values('2020-01-01 00:01:00.000', 0, 'china');");
+  taos_query(taos, "create table t1 using meters tags(1);");
-  taos_query(taos, "insert into t1 using meters tags('shanghai', 0) values('2020-01-01 00:02:00.000', 0, 'china');");
+  taos_query(taos, "create table t2 using meters tags(2);");
-  taos_query(taos, "insert into t1 using meters tags('shanghai', 0) values('2020-01-01 00:03:00.000', 0, 'china');");
+  taos_query(taos, "create table t3 using meters tags(3);");
-  taos_query(taos, "insert into t2 using meters tags('london', 0) values('2020-01-01 00:00:00.000', 0, 'UK');");
+  taos_query(taos, "create table t4 using meters tags(4);");
-  taos_query(taos, "insert into t2 using meters tags('london', 0) values('2020-01-01 00:01:00.000', 0, 'UK');");
+  taos_query(taos, "create table t5 using meters tags(5);");
-  taos_query(taos, "insert into t2 using meters tags('london', 0) values('2020-01-01 00:01:01.000', 0, 'UK');");
+  taos_query(taos, "create table t6 using meters tags(6);");
-  taos_query(taos, "insert into t2 using meters tags('london', 0) values('2020-01-01 00:01:02.000', 0, 'UK');");
+  taos_query(taos, "create table t7 using meters tags(7);");
-  taos_query(taos, "insert into t3 using meters tags('tianjin', 0) values('2020-01-01 00:01:02.000', 0, 'china');");
+  taos_query(taos, "create table t8 using meters tags(8);");
-  taos_query(taos, "insert into t4 using meters tags('wuhan', 0) values('2020-01-01 00:01:02.000', 0, 'china');");
+  taos_query(taos, "create table t9 using meters tags(9);");
-  taos_query(taos, "insert into t5 using meters tags('jinan', 0) values('2020-01-01 00:01:02.000', 0, 'china');");
+
-  taos_query(taos, "insert into t6 using meters tags('haikou', 0) values('2020-01-01 00:01:02.000', 0, 'china');");
+  taos_query(taos, "insert into t0 values('2020-01-01 00:00:00.000', 0);");
-  taos_query(taos, "insert into t7 using meters tags('nanjing', 0) values('2020-01-01 00:01:02.000', 0, 'china');");
+  taos_query(taos, "insert into t0 values('2020-01-01 00:01:00.000', 0);");
-  taos_query(taos, "insert into t8 using meters tags('lanzhou', 0) values('2020-01-01 00:01:02.000', 0, 'china');");
+  taos_query(taos, "insert into t0 values('2020-01-01 00:02:00.000', 0);");
-  taos_query(taos, "insert into t9 using meters tags('tokyo', 0) values('2020-01-01 00:01:02.000', 0, 'japan');");
+  taos_query(taos, "insert into t1 values('2020-01-01 00:00:00.000', 0);");
+  taos_query(taos, "insert into t1 values('2020-01-01 00:01:00.000', 0);");
+  taos_query(taos, "insert into t1 values('2020-01-01 00:02:00.000', 0);");
+  taos_query(taos, "insert into t1 values('2020-01-01 00:03:00.000', 0);");
+  taos_query(taos, "insert into t2 values('2020-01-01 00:00:00.000', 0);");
+  taos_query(taos, "insert into t2 values('2020-01-01 00:01:00.000', 0);");
+  taos_query(taos, "insert into t2 values('2020-01-01 00:01:01.000', 0);");
+  taos_query(taos, "insert into t2 values('2020-01-01 00:01:02.000', 0);");
+  taos_query(taos, "insert into t3 values('2020-01-01 00:01:02.000', 0);");
+  taos_query(taos, "insert into t4 values('2020-01-01 00:01:02.000', 0);");
+  taos_query(taos, "insert into t5 values('2020-01-01 00:01:02.000', 0);");
+  taos_query(taos, "insert into t6 values('2020-01-01 00:01:02.000', 0);");
+  taos_query(taos, "insert into t7 values('2020-01-01 00:01:02.000', 0);");
+  taos_query(taos, "insert into t8 values('2020-01-01 00:01:02.000', 0);");
+  taos_query(taos, "insert into t9 values('2020-01-01 00:01:02.000', 0);");
 
   // super tables subscription
+  usleep(1000000);
 
   TAOS_SUB* tsub = taos_subscribe(taos, 0, "test", "select * from meters;", NULL, NULL, 0);
   TAOS_RES* res = taos_consume(tsub);
@@ -90,23 +104,23 @@ void run_test(TAOS* taos) {
   res = taos_consume(tsub);
   check_row_count(__LINE__, res, 0);
 
-  taos_query(taos, "insert into t0 values('2020-01-01 00:03:00.000', 0, 'china');");
+  taos_query(taos, "insert into t0 values('2020-01-01 00:02:00.001', 0);");
-  taos_query(taos, "insert into t8 values('2020-01-01 00:01:03.000', 0, 'china');");
+  taos_query(taos, "insert into t8 values('2020-01-01 00:01:03.000', 0);");
   res = taos_consume(tsub);
   check_row_count(__LINE__, res, 2);
 
-  taos_query(taos, "insert into t2 values('2020-01-01 00:01:02.001', 0, 'UK');");
+  taos_query(taos, "insert into t2 values('2020-01-01 00:01:02.001', 0);");
-  taos_query(taos, "insert into t1 values('2020-01-01 00:03:00.001', 0, 'UK');");
+  taos_query(taos, "insert into t1 values('2020-01-01 00:03:00.001', 0);");
   res = taos_consume(tsub);
   check_row_count(__LINE__, res, 2);
 
-  taos_query(taos, "insert into t1 values('2020-01-01 00:03:00.002', 0, 'china');");
+  taos_query(taos, "insert into t1 values('2020-01-01 00:03:00.002', 0);");
   res = taos_consume(tsub);
   check_row_count(__LINE__, res, 1);
 
   // keep progress information and restart subscription
   taos_unsubscribe(tsub, 1);
-  taos_query(taos, "insert into t0 values('2020-01-01 00:04:00.000', 0, 'china');");
+  taos_query(taos, "insert into t0 values('2020-01-01 00:04:00.000', 0);");
   tsub = taos_subscribe(taos, 1, "test", "select * from meters;", NULL, NULL, 0);
   res = taos_consume(tsub);
   check_row_count(__LINE__, res, 24);
@@ -133,7 +147,7 @@ void run_test(TAOS* taos) {
   res = taos_consume(tsub);
   check_row_count(__LINE__, res, 0);
 
-  taos_query(taos, "insert into t0 values('2020-01-01 00:04:00.001', 0, 'china');");
+  taos_query(taos, "insert into t0 values('2020-01-01 00:04:00.001', 0);");
   res = taos_consume(tsub);
   check_row_count(__LINE__, res, 1);
 
@@ -197,7 +211,7 @@ int main(int argc, char *argv[]) {
   // init TAOS
   taos_init();
 
-  TAOS* taos = taos_connect(host, user, passwd, "test", 0);
+  TAOS* taos = taos_connect(host, user, passwd, "", 0);
   if (taos == NULL) {
     printf("failed to connect to db, reason:%s\n", taos_errstr(taos));
     exit(1);
@@ -209,6 +223,7 @@ int main(int argc, char *argv[]) {
     exit(0);
   }
 
+  taos_query(taos, "use test;");
   TAOS_SUB* tsub = NULL;
   if (async) {
     // create an asynchronized subscription, the callback function will be called every 1s

tests/script/general/cache/new_metrics.sim

@@ -27,7 +27,7 @@ sql create table $mt (ts timestamp, tbcol int) TAGS(tgcol bool)
 $i = 0
 while $i < 5
   $tb = $tbPrefix . $i
-  sql create table $tb using $mt tags( 0 )  
+  sql create table $tb using $mt tags( $i )  
   $x = 0
   while $x < $rowNum
     $val = $x * 60000

tests/script/general/compute/leastsquare.sim

@@ -48,41 +48,41 @@ $tb = $tbPrefix . $i
 
 sql select leastsquares(tbcol, 1, 1) from $tb
 print ===> $data00
-if $data00 != @(1.000000, 1.000000)@ then 
+if $data00 != @{slop:1.000000, intercept:1.000000}@ then
   return -1
 endi
 
 print =============== step3
 sql select leastsquares(tbcol, 1, 1) from $tb where ts < now + 4m
 print ===> $data00
-if $data00 != @(1.000000, 1.000000)@ then 
+if $data00 != @{slop:1.000000, intercept:1.000000}@ then
   return -1
 endi
 
 print =============== step4
 sql select leastsquares(tbcol, 1, 1) as b from $tb
 print ===> $data00
-if $data00 != @(1.000000, 1.000000)@ then 
+if $data00 != @{slop:1.000000, intercept:1.000000}@ then
   return -1
 endi
 
 print =============== step5
 sql select leastsquares(tbcol, 1, 1) as b from $tb interval(1m)
 print ===> $data01 
-if $data01 != @(1.000000, 1.000000)@ then 
+if $data01 != @{slop:1.000000, intercept:1.000000}@ then
   return -1
 endi
 
 sql select leastsquares(tbcol, 1, 1) as b from $tb interval(1d)
 print ===> $data01
-if $data01 != @(1.000000, 1.000000)@ then 
+if $data01 != @{slop:1.000000, intercept:1.000000}@ then
   return -1
 endi
 
 print =============== step6
 sql select leastsquares(tbcol, 1, 1) as b from $tb where ts < now + 4m interval(1m)
 print ===> $data01
-if $data01 != @(1.000000, 1.000000)@ then 
+if $data01 != @{slop:1.000000, intercept:1.000000}@ then
   return -1
 endi
 print ===> $rows

tests/script/general/import/replica1.sim

@@ -112,6 +112,7 @@ sql import into tb values(1520000050001, 50001)
 sql select * from tb;
 print $rows
 if $rows != 19 then
+  print expect 19, actual: $rows
   return -1
 endi