enh(sync): add trace log

Jenkinsfile2

@@ -168,13 +168,6 @@ pipeline {
     stages {
         stage('run test') {
             parallel {
-                stage('windows test') {
-                    agent {label " windows11 "}
-                    steps {
-                        pre_test_win()
-                        pre_test_build_win()
-                    }
-                }
                 stage('linux test') {
                     agent{label " slave3_0 || slave15 || slave16 || slave17 "}
                     options { skipDefaultCheckout() } 

include/common/tdataformat.h

@@ -372,10 +372,9 @@ static FORCE_INLINE void tdCopyColOfRowBySchema(SDataRow dst, STSchema *pDstSche
 // ----------------- Data column structure
 // SDataCol arrangement: data => bitmap => dataOffset
 typedef struct SDataCol {
-  int8_t          type;            // column type
-  uint8_t         bitmap : 1;      // 0: no bitmap if all rows are NORM, 1: has bitmap if has NULL/NORM rows
-  uint8_t         bitmapMode : 1;  // default is 0(2 bits), otherwise 1(1 bit)
-  uint8_t         reserve : 6;
+  int8_t          type;        // column type
+  uint8_t         bitmap : 1;  // 0: no bitmap if all rows are NORM, 1: has bitmap if has NULL/NORM rows
+  uint8_t         reserve : 7;
   int16_t         colId;      // column ID
   int32_t         bytes;      // column data bytes defined
   int32_t         offset;     // data offset in a SDataRow (including the header size)
@@ -387,8 +386,6 @@ typedef struct SDataCol {
   TSKEY           ts;         // only used in last NULL column
 } SDataCol;
 
-
-
 #define isAllRowsNull(pCol) ((pCol)->len == 0)
 #define isAllRowsNone(pCol) ((pCol)->len == 0)
 static FORCE_INLINE void dataColReset(SDataCol *pDataCol) { pDataCol->len = 0; }
@@ -482,7 +479,7 @@ void       tdResetDataCols(SDataCols *pCols);
 int32_t    tdInitDataCols(SDataCols *pCols, STSchema *pSchema);
 SDataCols *tdDupDataCols(SDataCols *pCols, bool keepData);
 SDataCols *tdFreeDataCols(SDataCols *pCols);
-int32_t tdMergeDataCols(SDataCols *target, SDataCols *source, int32_t rowsToMerge, int32_t *pOffset, bool forceSetNull, TDRowVerT maxVer);
+int32_t tdMergeDataCols(SDataCols *target, SDataCols *source, int32_t rowsToMerge, int32_t *pOffset, bool update, TDRowVerT maxVer);
 
 // ----------------- K-V data row structure
 /* |<-------------------------------------- len -------------------------------------------->|

include/common/trow.h

@@ -42,7 +42,7 @@ extern "C" {
  * @brief value type
  *  - for data from client input and STSRow in memory, 3 types of value none/null/norm available
  */
-#define TD_VTYPE_NORM 0x00U  // normal val: not none, not null(no need assign value)
+#define TD_VTYPE_NORM 0x00U  // normal val: not none, not null
 #define TD_VTYPE_NULL 0x01U  // null val
 #define TD_VTYPE_NONE 0x02U  // none or unknown/undefined
 #define TD_VTYPE_MAX  0x03U  //
@@ -140,8 +140,6 @@ typedef struct {
   };
   /// row total length
   uint32_t len;
-  /// row version
-  // uint64_t ver;
   /// the inline data, maybe a tuple or a k-v tuple
   char data[];
 } STSRow;
@@ -241,13 +239,13 @@ static FORCE_INLINE int32_t tdGetBitmapValType(const void *pBitmap, int16_t colI
 static FORCE_INLINE bool    tdIsBitmapValTypeNorm(const void *pBitmap, int16_t idx, int8_t bitmapMode);
 bool                        tdIsBitmapBlkNorm(const void *pBitmap, int32_t numOfBits, int8_t bitmapMode);
 int32_t tdAppendValToDataCol(SDataCol *pCol, TDRowValT valType, const void *val, int32_t numOfRows, int32_t maxPoints,
-                             int8_t bitmapMode);
+                             int8_t bitmapMode, bool isMerge);
 static FORCE_INLINE int32_t tdAppendColValToTpRow(SRowBuilder *pBuilder, TDRowValT valType, const void *val,
                                                   bool isCopyVarData, int8_t colType, int16_t colIdx, int32_t offset);
 static FORCE_INLINE int32_t tdAppendColValToKvRow(SRowBuilder *pBuilder, TDRowValT valType, const void *val,
                                                   bool isCopyVarData, int8_t colType, int16_t colIdx, int32_t offset,
                                                   col_id_t colId);
-int32_t                     tdAppendSTSRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols *pCols);
+int32_t                     tdAppendSTSRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols *pCols, bool isMerge);
 
 /**
  * @brief
@@ -718,6 +716,7 @@ static int32_t tdSRowResetBuf(SRowBuilder *pBuilder, void *pBuf) {
       terrno = TSDB_CODE_INVALID_PARA;
       return terrno;
   }
+
   return TSDB_CODE_SUCCESS;
 }
 

include/libs/index/index.h

@@ -56,7 +56,8 @@ typedef enum {
   QUERY_LESS_EQUAL,
   QUERY_GREATER_THAN,
   QUERY_GREATER_EQUAL,
-  QUERY_RANGE
+  QUERY_RANGE,
+  QUERY_MAX
 } EIndexQueryType;
 
 /*
@@ -178,8 +179,8 @@ void        indexOptsDestroy(SIndexOpts* opts);
  * @param:
  */
 
-SIndexTerm* indexTermCreate(int64_t suid, SIndexOperOnColumn operType, int8_t qType, uint8_t colType,
-                            const char* colName, int32_t nColName, const char* colVal, int32_t nColVal);
+SIndexTerm* indexTermCreate(int64_t suid, SIndexOperOnColumn operType, uint8_t colType, const char* colName,
+                            int32_t nColName, const char* colVal, int32_t nColVal);
 void        indexTermDestroy(SIndexTerm* p);
 
 /*

include/util/tdef.h

@@ -29,6 +29,8 @@ extern "C" {
 #define TSKEY_MAX         (INT64_MAX - 1)
 #define TSKEY_INITIAL_VAL TSKEY_MIN
 
+#define TD_VER_MAX UINT64_MAX  // TODO: use the real max version from query handle
+
 // Bytes for each type.
 extern const int32_t TYPE_BYTES[15];
 

source/common/src/tdatablock.c

@@ -1478,11 +1478,11 @@ void blockDebugShowData(const SArray* dataBlocks) {
  * @param uid set as parameter temporarily // TODO: remove this parameter, and the executor should set uid in
  * SDataBlock->info.uid
  * @param suid  // TODO: check with Liao whether suid response is reasonable
- * 
+ *
  * TODO: colId should be set
  */
-int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks, STSchema *pTSchema, int32_t vgId, tb_uid_t uid,
-                                        tb_uid_t suid) {
+int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks, STSchema* pTSchema, int32_t vgId,
+                                    tb_uid_t uid, tb_uid_t suid) {
   int32_t sz = taosArrayGetSize(pDataBlocks);
   int32_t bufSize = sizeof(SSubmitReq);
   for (int32_t i = 0; i < sz; ++i) {
@@ -1494,16 +1494,16 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks
   ASSERT(bufSize < 3 * 1024 * 1024);
 
   *pReq = taosMemoryCalloc(1, bufSize);
-  if(!(*pReq)) {
+  if (!(*pReq)) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return TSDB_CODE_FAILED;
   }
   void* pDataBuf = *pReq;
 
-  int32_t msgLen = sizeof(SSubmitReq);
+  int32_t     msgLen = sizeof(SSubmitReq);
   int32_t     numOfBlks = 0;
   SRowBuilder rb = {0};
-  tdSRowInit(&rb, 0); // TODO: use the latest version
+  tdSRowInit(&rb, 0);  // TODO: use the latest version
 
   for (int32_t i = 0; i < sz; ++i) {
     SSDataBlock* pDataBlock = taosArrayGet(pDataBlocks, i);
@@ -1511,8 +1511,8 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks
     int32_t      rows = pDataBlock->info.rows;
     int32_t      rowSize = pDataBlock->info.rowSize;
     int64_t      groupId = pDataBlock->info.groupId;
-    
-    if(rb.nCols != colNum) {
+
+    if (rb.nCols != colNum) {
       tdSRowSetTpInfo(&rb, colNum, pTSchema->flen);
     }
 
@@ -1525,10 +1525,10 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks
 
     msgLen += sizeof(SSubmitBlk);
     int32_t dataLen = 0;
-    for (int32_t j = 0; j < rows; ++j) {     // iterate by row
-      tdSRowResetBuf(&rb, POINTER_SHIFT(pDataBuf, msgLen)); // set row buf
+    for (int32_t j = 0; j < rows; ++j) {                     // iterate by row
+      tdSRowResetBuf(&rb, POINTER_SHIFT(pDataBuf, msgLen));  // set row buf
       printf("|");
-      bool  isStartKey = false;
+      bool isStartKey = false;
       for (int32_t k = 0; k < colNum; ++k) {  // iterate by column
         SColumnInfoData* pColInfoData = taosArrayGet(pDataBlock->pDataBlock, k);
         void*            var = POINTER_SHIFT(pColInfoData->pData, j * pColInfoData->info.bytes);
@@ -1536,7 +1536,8 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks
           case TSDB_DATA_TYPE_TIMESTAMP:
             if (!isStartKey) {
               isStartKey = true;
-              tdAppendColValToRow(&rb, PRIMARYKEY_TIMESTAMP_COL_ID, TSDB_DATA_TYPE_TIMESTAMP, TD_VTYPE_NORM, var, true, 0, 0);
+              tdAppendColValToRow(&rb, PRIMARYKEY_TIMESTAMP_COL_ID, TSDB_DATA_TYPE_TIMESTAMP, TD_VTYPE_NORM, var, true,
+                                  0, 0);
             } else {
               tdAppendColValToRow(&rb, 2, TSDB_DATA_TYPE_TIMESTAMP, TD_VTYPE_NORM, var, true, 8, k);
               break;
@@ -1629,14 +1630,14 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema) {
     blkHead->uid = htobe64(pDataBlock->info.uid);
 
     int32_t rows = pDataBlock->info.rows;
-    int32_t maxLen = TD_ROW_MAX_BYTES_FROM_SCHEMA(pTSchema);
+    /*int32_t maxLen = TD_ROW_MAX_BYTES_FROM_SCHEMA(pTSchema);*/
     /*blkHead->dataLen = htonl(rows * maxLen);*/
     blkHead->dataLen = 0;
 
     void*   blockData = POINTER_SHIFT(submitBlk, sizeof(SSubmitBlk));
     STSRow* rowData = blockData;
 
-    for (int32_t j = 0; j < pDataBlock->info.rows; j++) {
+    for (int32_t j = 0; j < rows; j++) {
       SRowBuilder rb = {0};
       tdSRowInit(&rb, pTSchema->version);
       tdSRowSetTpInfo(&rb, pTSchema->numOfCols, pTSchema->flen);

source/common/src/trow.c

@@ -24,7 +24,8 @@ const uint8_t tdVTypeByte[2][3] = {{
                                    },
                                    {
                                        // 1 bit
-                                       TD_VTYPE_NORM_BYTE_I, TD_VTYPE_NULL_BYTE_I,
+                                       TD_VTYPE_NORM_BYTE_I,  // normal
+                                       TD_VTYPE_NULL_BYTE_I,
                                        TD_VTYPE_NULL_BYTE_I,  // padding
                                    }
 
@@ -33,6 +34,24 @@ const uint8_t tdVTypeByte[2][3] = {{
 // declaration
 static uint8_t tdGetBitmapByte(uint8_t byte);
 
+// static void dataColSetNEleNull(SDataCol *pCol, int nEle);
+
+/**
+ * @brief src2 data has more priority than src1
+ *
+ * @param target
+ * @param src1
+ * @param iter1
+ * @param limit1
+ * @param src2
+ * @param iter2
+ * @param limit2
+ * @param tRows
+ * @param update
+ */
+static void tdMergeTwoDataCols(SDataCols *target, SDataCols *src1, int *iter1, int limit1, SDataCols *src2, int *iter2,
+                               int limit2, int tRows, bool update);
+
 // implementation
 /**
  * @brief Compress bitmap bytes comprised of 2-bits to counterpart of 1-bit.
@@ -229,23 +248,23 @@ static uint8_t tdGetMergedBitmapByte(uint8_t byte) {
 void tdMergeBitmap(uint8_t *srcBitmap, int32_t nBits, uint8_t *dstBitmap) {
   int32_t i = 0, j = 0;
   int32_t nBytes = TD_BITMAP_BYTES(nBits);
-  int32_t nStrictBytes = nBits / 4;
-  int32_t nPartialBits = nBits - nStrictBytes * 4;
+  int32_t nRoundBytes = nBits / 4;
+  int32_t nRemainderBits = nBits - nRoundBytes * 4;
 
-  switch (nPartialBits) {
+  switch (nRemainderBits) {
     case 0:
       // NOTHING TODO
       break;
     case 1: {
-      void *lastByte = POINTER_SHIFT(srcBitmap, nStrictBytes);
+      void *lastByte = POINTER_SHIFT(srcBitmap, nRoundBytes);
       *(uint8_t *)lastByte &= 0xC0;
     } break;
     case 2: {
-      void *lastByte = POINTER_SHIFT(srcBitmap, nStrictBytes);
+      void *lastByte = POINTER_SHIFT(srcBitmap, nRoundBytes);
       *(uint8_t *)lastByte &= 0xF0;
     } break;
     case 3: {
-      void *lastByte = POINTER_SHIFT(srcBitmap, nStrictBytes);
+      void *lastByte = POINTER_SHIFT(srcBitmap, nRoundBytes);
       *(uint8_t *)lastByte &= 0xFC;
     } break;
     default:
@@ -266,10 +285,6 @@ void tdMergeBitmap(uint8_t *srcBitmap, int32_t nBits, uint8_t *dstBitmap) {
   }
 }
 
-// static void dataColSetNEleNull(SDataCol *pCol, int nEle);
-static void tdMergeTwoDataCols(SDataCols *target, SDataCols *src1, int *iter1, int limit1, SDataCols *src2, int *iter2,
-                               int limit2, int tRows, bool forceSetNull);
-
 static FORCE_INLINE void dataColSetNullAt(SDataCol *pCol, int index, bool setBitmap, int8_t bitmapMode) {
   if (IS_VAR_DATA_TYPE(pCol->type)) {
     pCol->dataOff[index] = pCol->len;
@@ -329,7 +344,7 @@ bool tdIsBitmapBlkNorm(const void *pBitmap, int32_t numOfBits, int8_t bitmapMode
     if (*((uint8_t *)pBitmap) != vTypeByte) {
       return false;
     }
-    pBitmap = POINTER_SHIFT(pBitmap, 1);
+    pBitmap = POINTER_SHIFT(pBitmap, i);
   }
 
   int32_t nLeft = numOfBits - nBytes * (bitmapMode == 0 ? TD_VTYPE_BITS : TD_VTYPE_BITS_I);
@@ -410,11 +425,12 @@ STSRow *tdRowDup(STSRow *row) {
  * @param val
  * @param numOfRows
  * @param maxPoints
- * @param bitmapMode  default is 0(2 bits), otherwise 1(1 bit)
+ * @param bitmapMode default is 0(2 bits), otherwise 1(1 bit)
+ * @param isMerge merge to current row
  * @return int
  */
 int tdAppendValToDataCol(SDataCol *pCol, TDRowValT valType, const void *val, int numOfRows, int maxPoints,
-                         int8_t bitmapMode) {
+                         int8_t bitmapMode, bool isMerge) {
   TASSERT(pCol != NULL);
 
   // Assume that the columns not specified during insert/upsert mean None.
@@ -430,33 +446,58 @@ int tdAppendValToDataCol(SDataCol *pCol, TDRowValT valType, const void *val, int
       dataColSetNEleNone(pCol, numOfRows, bitmapMode);
     }
   }
-  if (!tdValTypeIsNorm(valType)) {
+  const void *value = val;
+  if (!tdValTypeIsNorm(valType) || !val) {
     // TODO:
     // 1. back compatibility and easy to debug with codes of 2.0 to save NULL values.
     // 2. later on, considering further optimization, don't save Null/None for VarType.
-    val = getNullValue(pCol->type);
+    value = getNullValue(pCol->type);
   }
-  if (IS_VAR_DATA_TYPE(pCol->type)) {
-    // set offset
-    pCol->dataOff[numOfRows] = pCol->len;
-    // Copy data
-    memcpy(POINTER_SHIFT(pCol->pData, pCol->len), val, varDataTLen(val));
-    // Update the length
-    pCol->len += varDataTLen(val);
-  } else {
-    ASSERT(pCol->len == TYPE_BYTES[pCol->type] * numOfRows);
-    memcpy(POINTER_SHIFT(pCol->pData, pCol->len), val, pCol->bytes);
-    pCol->len += pCol->bytes;
+  if (!isMerge) {
+    if (IS_VAR_DATA_TYPE(pCol->type)) {
+      // set offset
+      pCol->dataOff[numOfRows] = pCol->len;
+      // Copy data
+      memcpy(POINTER_SHIFT(pCol->pData, pCol->len), value, varDataTLen(value));
+      // Update the length
+      pCol->len += varDataTLen(value);
+    } else {
+      ASSERT(pCol->len == TYPE_BYTES[pCol->type] * numOfRows);
+      memcpy(POINTER_SHIFT(pCol->pData, pCol->len), value, pCol->bytes);
+      pCol->len += pCol->bytes;
+    }
+  } else if (!tdValTypeIsNone(valType)) {
+    if (IS_VAR_DATA_TYPE(pCol->type)) {
+      // keep the last offset
+      // discard the last var data
+      int32_t lastVarLen = varDataTLen(POINTER_SHIFT(pCol->pData, pCol->dataOff[numOfRows]));
+      pCol->len -= lastVarLen;
+      // Copy data
+      memcpy(POINTER_SHIFT(pCol->pData, pCol->len), value, varDataTLen(value));
+      // Update the length
+      pCol->len += varDataTLen(value);
+    } else {
+      ASSERT(pCol->len - TYPE_BYTES[pCol->type] == TYPE_BYTES[pCol->type] * numOfRows);
+      memcpy(POINTER_SHIFT(pCol->pData, pCol->len - TYPE_BYTES[pCol->type]), value, pCol->bytes);
+    }
   }
+
 #ifdef TD_SUPPORT_BITMAP
-  tdSetBitmapValType(pCol->pBitmap, numOfRows, valType, bitmapMode);
+  if (!isMerge || !tdValTypeIsNone(valType)) {
+    tdSetBitmapValType(pCol->pBitmap, numOfRows, valType, bitmapMode);
+  }
 #endif
   return 0;
 }
 
 // internal
-static int32_t tdAppendTpRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols *pCols) {
+static int32_t tdAppendTpRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols *pCols, bool isMerge) {
+#if 0
   ASSERT(pCols->numOfRows == 0 || dataColsKeyLast(pCols) < TD_ROW_KEY(pRow));
+#endif
+
+  // Multi-Version rows with the same key and different versions supported
+  ASSERT(pCols->numOfRows == 0 || dataColsKeyLast(pCols) <= TD_ROW_KEY(pRow));
 
   int   rcol = 1;
   int   dcol = 1;
@@ -464,12 +505,14 @@ static int32_t tdAppendTpRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols
 
   SDataCol *pDataCol = &(pCols->cols[0]);
   ASSERT(pDataCol->colId == PRIMARYKEY_TIMESTAMP_COL_ID);
-  tdAppendValToDataCol(pDataCol, TD_VTYPE_NORM, &pRow->ts, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode);
+  tdAppendValToDataCol(pDataCol, TD_VTYPE_NORM, &pRow->ts, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode,
+                       isMerge);
 
   while (dcol < pCols->numOfCols) {
     pDataCol = &(pCols->cols[dcol]);
     if (rcol >= schemaNCols(pSchema)) {
-      tdAppendValToDataCol(pDataCol, TD_VTYPE_NULL, NULL, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode);
+      tdAppendValToDataCol(pDataCol, TD_VTYPE_NULL, NULL, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode,
+                           isMerge);
       ++dcol;
       continue;
     }
@@ -480,22 +523,26 @@ static int32_t tdAppendTpRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols
       if (tdGetTpRowValOfCol(&sVal, pRow, pBitmap, pRowCol->type, pRowCol->offset - sizeof(TSKEY), rcol - 1) < 0) {
         return terrno;
       }
-      tdAppendValToDataCol(pDataCol, sVal.valType, sVal.val, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode);
+      tdAppendValToDataCol(pDataCol, sVal.valType, sVal.val, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode,
+                           isMerge);
       ++dcol;
       ++rcol;
     } else if (pRowCol->colId < pDataCol->colId) {
       ++rcol;
     } else {
-      tdAppendValToDataCol(pDataCol, TD_VTYPE_NULL, NULL, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode);
+      tdAppendValToDataCol(pDataCol, TD_VTYPE_NULL, NULL, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode,
+                           isMerge);
       ++dcol;
     }
   }
+#if 0  
   ++pCols->numOfRows;
+#endif
 
   return TSDB_CODE_SUCCESS;
 }
 // internal
-static int32_t tdAppendKvRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols *pCols) {
+static int32_t tdAppendKvRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols *pCols, bool isMerge) {
   ASSERT(pCols->numOfRows == 0 || dataColsKeyLast(pCols) < TD_ROW_KEY(pRow));
 
   int   rcol = 0;
@@ -506,12 +553,14 @@ static int32_t tdAppendKvRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols
 
   SDataCol *pDataCol = &(pCols->cols[0]);
   ASSERT(pDataCol->colId == PRIMARYKEY_TIMESTAMP_COL_ID);
-  tdAppendValToDataCol(pDataCol, TD_VTYPE_NORM, &pRow->ts, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode);
+  tdAppendValToDataCol(pDataCol, TD_VTYPE_NORM, &pRow->ts, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode,
+                       isMerge);
 
   while (dcol < pCols->numOfCols) {
     pDataCol = &(pCols->cols[dcol]);
     if (rcol >= tRowCols || rcol >= tSchemaCols) {
-      tdAppendValToDataCol(pDataCol, TD_VTYPE_NULL, NULL, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode);
+      tdAppendValToDataCol(pDataCol, TD_VTYPE_NULL, NULL, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode,
+                           isMerge);
       ++dcol;
       continue;
     }
@@ -527,17 +576,21 @@ static int32_t tdAppendKvRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols
       if (tdGetKvRowValOfCol(&sVal, pRow, pBitmap, pIdx->offset, colIdx) < 0) {
         return terrno;
       }
-      tdAppendValToDataCol(pDataCol, sVal.valType, sVal.val, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode);
+      tdAppendValToDataCol(pDataCol, sVal.valType, sVal.val, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode,
+                           isMerge);
       ++dcol;
       ++rcol;
     } else if (pIdx->colId < pDataCol->colId) {
       ++rcol;
     } else {
-      tdAppendValToDataCol(pDataCol, TD_VTYPE_NULL, NULL, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode);
+      tdAppendValToDataCol(pDataCol, TD_VTYPE_NULL, NULL, pCols->numOfRows, pCols->maxPoints, pCols->bitmapMode,
+                           isMerge);
       ++dcol;
     }
   }
+#if 0
   ++pCols->numOfRows;
+#endif
 
   return TSDB_CODE_SUCCESS;
 }
@@ -548,20 +601,30 @@ static int32_t tdAppendKvRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols
  * @param pRow
  * @param pSchema
  * @param pCols
- * @param forceSetNull
  */
-int32_t tdAppendSTSRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols *pCols) {
+int32_t tdAppendSTSRowToDataCol(STSRow *pRow, STSchema *pSchema, SDataCols *pCols, bool isMerge) {
   if (TD_IS_TP_ROW(pRow)) {
-    return tdAppendTpRowToDataCol(pRow, pSchema, pCols);
+    return tdAppendTpRowToDataCol(pRow, pSchema, pCols, isMerge);
   } else if (TD_IS_KV_ROW(pRow)) {
-    return tdAppendKvRowToDataCol(pRow, pSchema, pCols);
+    return tdAppendKvRowToDataCol(pRow, pSchema, pCols, isMerge);
   } else {
     ASSERT(0);
   }
   return TSDB_CODE_SUCCESS;
 }
 
-int tdMergeDataCols(SDataCols *target, SDataCols *source, int rowsToMerge, int *pOffset, bool forceSetNull,
+/**
+ * @brief source data has more priority than target
+ *
+ * @param target
+ * @param source
+ * @param rowsToMerge
+ * @param pOffset
+ * @param update
+ * @param maxVer
+ * @return int
+ */
+int tdMergeDataCols(SDataCols *target, SDataCols *source, int rowsToMerge, int *pOffset, bool update,
                     TDRowVerT maxVer) {
   ASSERT(rowsToMerge > 0 && rowsToMerge <= source->numOfRows);
   ASSERT(target->numOfCols == source->numOfCols);
@@ -576,17 +639,38 @@ int tdMergeDataCols(SDataCols *target, SDataCols *source, int rowsToMerge, int *
   if ((target->numOfRows == 0) || (dataColsKeyLast(target) < dataColsKeyAtRow(source, *pOffset))) {  // No overlap
     ASSERT(target->numOfRows + rowsToMerge <= target->maxPoints);
     // TODO: filter the maxVer
-    for (int i = 0; i < rowsToMerge; i++) {
+    TSKEY lastKey = TSKEY_INITIAL_VAL;
+    for (int i = 0; i < rowsToMerge; ++i) {
+      bool merge = false;
       for (int j = 0; j < source->numOfCols; j++) {
         if (source->cols[j].len > 0 || target->cols[j].len > 0) {
           SCellVal sVal = {0};
           if (tdGetColDataOfRow(&sVal, source->cols + j, i + (*pOffset), source->bitmapMode) < 0) {
             TASSERT(0);
           }
+
+          if (j == 0) {
+            if (lastKey == *(TSKEY *)sVal.val) {
+              if (!update) {
+                break;
+              }
+              merge = true;
+            } else if (lastKey != TSKEY_INITIAL_VAL) {
+              ++target->numOfRows;
+            }
+
+            lastKey = *(TSKEY *)sVal.val;
+          }
+          if (i == 0) {
+            (target->cols + j)->bitmap = (source->cols + j)->bitmap;
+          }
+
           tdAppendValToDataCol(target->cols + j, sVal.valType, sVal.val, target->numOfRows, target->maxPoints,
-                               target->bitmapMode);
+                               target->bitmapMode, merge);
         }
       }
+    }
+    if (lastKey != TSKEY_INITIAL_VAL) {
       ++target->numOfRows;
     }
     (*pOffset) += rowsToMerge;
@@ -596,7 +680,7 @@ int tdMergeDataCols(SDataCols *target, SDataCols *source, int rowsToMerge, int *
 
     int iter1 = 0;
     tdMergeTwoDataCols(target, pTarget, &iter1, pTarget->numOfRows, source, pOffset, source->numOfRows,
-                       pTarget->numOfRows + rowsToMerge, forceSetNull);
+                       pTarget->numOfRows + rowsToMerge, update);
   }
 
   tdFreeDataCols(pTarget);
@@ -607,67 +691,95 @@ _err:
   return -1;
 }
 
-// src2 data has more priority than src1
+static void tdAppendValToDataCols(SDataCols *target, SDataCols *src, int iter, bool isMerge) {
+  for (int i = 0; i < src->numOfCols; ++i) {
+    ASSERT(target->cols[i].type == src->cols[i].type);
+    if (src->cols[i].len > 0 || target->cols[i].len > 0) {
+      SCellVal sVal = {0};
+      if (tdGetColDataOfRow(&sVal, src->cols + i, iter, src->bitmapMode) < 0) {
+        TASSERT(0);
+      }
+      if (isMerge) {
+        if (!tdValTypeIsNone(sVal.valType)) {
+          tdAppendValToDataCol(&(target->cols[i]), sVal.valType, sVal.val, target->numOfRows, target->maxPoints,
+                               target->bitmapMode, isMerge);
+        } else {
+          // Keep the origin value for None
+        }
+      } else {
+        tdAppendValToDataCol(&(target->cols[i]), sVal.valType, sVal.val, target->numOfRows, target->maxPoints,
+                             target->bitmapMode, isMerge);
+      }
+    }
+  }
+}
+/**
+ * @brief src2 data has more priority than src1
+ *
+ * @param target
+ * @param src1
+ * @param iter1
+ * @param limit1
+ * @param src2
+ * @param iter2
+ * @param limit2
+ * @param tRows
+ * @param update
+ */
 static void tdMergeTwoDataCols(SDataCols *target, SDataCols *src1, int *iter1, int limit1, SDataCols *src2, int *iter2,
-                               int limit2, int tRows, bool forceSetNull) {
+                               int limit2, int tRows, bool update) {
   tdResetDataCols(target);
+  target->bitmapMode = src1->bitmapMode;
   ASSERT(limit1 <= src1->numOfRows && limit2 <= src2->numOfRows);
+  int32_t nRows = 0;
 
-  while (target->numOfRows < tRows) {
+  // TODO: filter the maxVer
+  // TODO: handle the delete function
+  TSKEY lastKey = TSKEY_INITIAL_VAL;
+  while (nRows < tRows) {
     if (*iter1 >= limit1 && *iter2 >= limit2) break;
 
     TSKEY key1 = (*iter1 >= limit1) ? INT64_MAX : dataColsKeyAt(src1, *iter1);
-    TKEY  tkey1 = (*iter1 >= limit1) ? TKEY_NULL : dataColsTKeyAt(src1, *iter1);
+    // TKEY  tkey1 = (*iter1 >= limit1) ? TKEY_NULL : dataColsTKeyAt(src1, *iter1);
     TSKEY key2 = (*iter2 >= limit2) ? INT64_MAX : dataColsKeyAt(src2, *iter2);
     // TKEY  tkey2 = (*iter2 >= limit2) ? TKEY_NULL : dataColsTKeyAt(src2, *iter2);
 
-    ASSERT(tkey1 == TKEY_NULL || (!TKEY_IS_DELETED(tkey1)));
-    // TODO: filter the maxVer
-    if (key1 < key2) {
-      for (int i = 0; i < src1->numOfCols; ++i) {
-        ASSERT(target->cols[i].type == src1->cols[i].type);
-        if (src1->cols[i].len > 0 || target->cols[i].len > 0) {
-          SCellVal sVal = {0};
-          if (tdGetColDataOfRow(&sVal, src1->cols + i, *iter1, src1->bitmapMode) < 0) {
-            TASSERT(0);
-          }
-          tdAppendValToDataCol(&(target->cols[i]), sVal.valType, sVal.val, target->numOfRows, target->maxPoints,
-                               target->bitmapMode);
+    // ASSERT(tkey1 == TKEY_NULL || (!TKEY_IS_DELETED(tkey1)));
+
+    if (key1 <= key2) {
+      // select key1 if not delete
+      if (update && (lastKey == key1)) {
+        tdAppendValToDataCols(target, src1, *iter1, true);
+      } else if (lastKey != key1) {
+        if (lastKey != TSKEY_INITIAL_VAL) {
+          ++target->numOfRows;
         }
+        tdAppendValToDataCols(target, src1, *iter1, false);
       }
-
-      ++target->numOfRows;
+      ++nRows;
       ++(*iter1);
-    } else if (key1 >= key2) {
-      // TODO: filter the maxVer
-      if ((key1 > key2) || ((key1 == key2) && !TKEY_IS_DELETED(key2))) {
-        for (int i = 0; i < src2->numOfCols; ++i) {
-          SCellVal sVal = {0};
-          ASSERT(target->cols[i].type == src2->cols[i].type);
-          if (tdGetColDataOfRow(&sVal, src2->cols + i, *iter2, src2->bitmapMode) < 0) {
-            TASSERT(0);
-          }
-          if (src2->cols[i].len > 0 && !tdValTypeIsNull(sVal.valType)) {
-            tdAppendValToDataCol(&(target->cols[i]), sVal.valType, sVal.val, target->numOfRows, target->maxPoints,
-                                 target->bitmapMode);
-          } else if (!forceSetNull && key1 == key2 && src1->cols[i].len > 0) {
-            if (tdGetColDataOfRow(&sVal, src1->cols + i, *iter1, src1->bitmapMode) < 0) {
-              TASSERT(0);
-            }
-            tdAppendValToDataCol(&(target->cols[i]), sVal.valType, sVal.val, target->numOfRows, target->maxPoints,
-                                 target->bitmapMode);
-          } else if (target->cols[i].len > 0) {
-            dataColSetNullAt(&target->cols[i], target->numOfRows, true, target->bitmapMode);
-          }
+      lastKey = key1;
+    } else {
+      // use key2 if not deleted
+      // TODO: handle the delete function
+      if (update && (lastKey == key2)) {
+        tdAppendValToDataCols(target, src2, *iter2, true);
+      } else if (lastKey != key2) {
+        if (lastKey != TSKEY_INITIAL_VAL) {
+          ++target->numOfRows;
         }
-        ++target->numOfRows;
+        tdAppendValToDataCols(target, src2, *iter2, false);
       }
 
+      ++nRows;
       ++(*iter2);
-      if (key1 == key2) ++(*iter1);
+      lastKey = key2;
     }
 
-    ASSERT(target->numOfRows <= target->maxPoints);
+    ASSERT(target->numOfRows <= target->maxPoints - 1);
+  }
+  if (lastKey != TSKEY_INITIAL_VAL) {
+    ++target->numOfRows;
   }
 }
 
@@ -777,7 +889,7 @@ SDataCols *tdDupDataCols(SDataCols *pDataCols, bool keepData) {
   pRet->sversion = pDataCols->sversion;
   if (keepData) pRet->numOfRows = pDataCols->numOfRows;
 
-  for (int i = 0; i < pDataCols->numOfCols; i++) {
+  for (int i = 0; i < pDataCols->numOfCols; ++i) {
     pRet->cols[i].type = pDataCols->cols[i].type;
     pRet->cols[i].bitmap = pDataCols->cols[i].bitmap;
     pRet->cols[i].colId = pDataCols->cols[i].colId;
@@ -797,8 +909,7 @@ SDataCols *tdDupDataCols(SDataCols *pDataCols, bool keepData) {
           memcpy(pRet->cols[i].dataOff, pDataCols->cols[i].dataOff, dataOffSize);
         }
         if (!TD_COL_ROWS_NORM(pRet->cols + i)) {
-          int32_t nBitmapBytes = (int32_t)TD_BITMAP_BYTES(pDataCols->numOfRows);
-          memcpy(pRet->cols[i].pBitmap, pDataCols->cols[i].pBitmap, nBitmapBytes);
+          memcpy(pRet->cols[i].pBitmap, pDataCols->cols[i].pBitmap, TD_BITMAP_BYTES(pDataCols->numOfRows));
         }
       }
     }

source/common/src/ttypes.c

@@ -557,7 +557,7 @@ static const void *nullValues[] = {
 };
 
 const void *getNullValue(int32_t type) {
-  assert(type >= TSDB_DATA_TYPE_BOOL && type <= TSDB_DATA_TYPE_UBIGINT);
+  assert(type >= TSDB_DATA_TYPE_BOOL && type <= TSDB_DATA_TYPE_UBIGINT);  // TODO: extend the types
   return nullValues[type - 1];
 }
 

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

@@ -293,14 +293,10 @@ void vmInitMsgHandle(SMgmtWrapper *pWrapper) {
   dmSetMsgHandle(pWrapper, TDMT_VND_TASK_MERGE_EXEC, vmProcessMergeMsg, DEFAULT_HANDLE);
   dmSetMsgHandle(pWrapper, TDMT_VND_TASK_WRITE_EXEC, vmProcessWriteMsg, DEFAULT_HANDLE);
   dmSetMsgHandle(pWrapper, TDMT_VND_STREAM_TRIGGER, vmProcessFetchMsg, DEFAULT_HANDLE);
-
   dmSetMsgHandle(pWrapper, TDMT_VND_ALTER_VNODE, vmProcessWriteMsg, DEFAULT_HANDLE);
   dmSetMsgHandle(pWrapper, TDMT_VND_COMPACT_VNODE, vmProcessWriteMsg, DEFAULT_HANDLE);
-
   dmSetMsgHandle(pWrapper, TDMT_DND_CREATE_VNODE, vmProcessMgmtMsg, DEFAULT_HANDLE);
   dmSetMsgHandle(pWrapper, TDMT_DND_DROP_VNODE, vmProcessMgmtMsg, DEFAULT_HANDLE);
-  // dmSetMsgHandle(pWrapper, TDMT_DND_SYNC_VNODE, vmProcessMgmtMsg, DEFAULT_HANDLE);
-  // dmSetMsgHandle(pWrapper, TDMT_DND_COMPACT_VNODE, vmProcessMgmtMsg, DEFAULT_HANDLE);
 
   // sync integration
   dmSetMsgHandle(pWrapper, TDMT_VND_SYNC_TIMEOUT, (NodeMsgFp)vmProcessSyncMsg, DEFAULT_HANDLE);

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

@@ -31,6 +31,7 @@ int32_t mndCheckDropUserAuth(SUserObj *pOperUser);
 
 int32_t mndCheckNodeAuth(SUserObj *pOperUser);
 int32_t mndCheckFuncAuth(SUserObj *pOperUser);
+int32_t mndCheckTransAuth(SUserObj *pOperUser);
 
 int32_t mndCheckCreateDbAuth(SUserObj *pOperUser);
 int32_t mndCheckAlterDropCompactDbAuth(SUserObj *pOperUser, SDbObj *pDb);

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

@@ -31,6 +31,11 @@ int32_t  mndValidateStbInfo(SMnode *pMnode, SSTableMetaVersion *pStbs, int32_t n
                             int32_t *pRspLen);
 int32_t  mndGetNumOfStbs(SMnode *pMnode, char *dbName, int32_t *pNumOfStbs);
 
+int32_t mndCheckCreateStbReq(SMCreateStbReq *pCreate);
+SDbObj *mndAcquireDbByStb(SMnode *pMnode, const char *stbName);
+int32_t mndBuildStbFromReq(SMnode *pMnode, SStbObj *pDst, SMCreateStbReq *pCreate, SDbObj *pDb);
+int32_t mndAddStbToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SStbObj *pStb);
+
 #ifdef __cplusplus
 }
 #endif

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

@@ -27,10 +27,14 @@ void     mndCleanupVgroup(SMnode *pMnode);
 SVgObj  *mndAcquireVgroup(SMnode *pMnode, int32_t vgId);
 void     mndReleaseVgroup(SMnode *pMnode, SVgObj *pVgroup);
 SSdbRaw *mndVgroupActionEncode(SVgObj *pVgroup);
-int32_t  mndAllocVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj **ppVgroups);
 SEpSet   mndGetVgroupEpset(SMnode *pMnode, const SVgObj *pVgroup);
 int32_t  mndGetVnodesNum(SMnode *pMnode, int32_t dnodeId);
 
+int32_t mndAllocVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj **ppVgroups);
+SArray *mndBuildDnodesArray(SMnode *pMnode);
+int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray);
+int32_t mndRemoveVnodeFromVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray, SVnodeGid *del1, SVnodeGid *del2);
+
 void *mndBuildCreateVnodeReq(SMnode *pMnode, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);
 void *mndBuildDropVnodeReq(SMnode *pMnode, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);
 void *mndBuildAlterVnodeReq(SMnode *pMnode, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);

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

@@ -74,10 +74,7 @@ static int32_t mndProcessAuthReq(SNodeMsg *pReq) {
 }
 
 int32_t mndCheckCreateUserAuth(SUserObj *pOperUser) {
-  if (pOperUser->superUser) {
-    return 0;
-  }
-
+  if (pOperUser->superUser) return 0;
   terrno = TSDB_CODE_MND_NO_RIGHTS;
   return -1;
 }
@@ -118,28 +115,25 @@ int32_t mndCheckAlterUserAuth(SUserObj *pOperUser, SUserObj *pUser, SDbObj *pDb,
 }
 
 int32_t mndCheckDropUserAuth(SUserObj *pOperUser) {
-  if (pOperUser->superUser) {
-    return 0;
-  }
-
+  if (pOperUser->superUser) return 0;
   terrno = TSDB_CODE_MND_NO_RIGHTS;
   return -1;
 }
 
 int32_t mndCheckNodeAuth(SUserObj *pOperUser) {
-  if (pOperUser->superUser) {
-    return 0;
-  }
-
+  if (pOperUser->superUser) return 0;
   terrno = TSDB_CODE_MND_NO_RIGHTS;
   return -1;
 }
 
 int32_t mndCheckFuncAuth(SUserObj *pOperUser) {
-  if (pOperUser->superUser) {
-    return 0;
-  }
+  if (pOperUser->superUser) return 0;
+  terrno = TSDB_CODE_MND_NO_RIGHTS;
+  return -1;
+}
 
+int32_t mndCheckTransAuth(SUserObj *pOperUser) {
+  if (pOperUser->superUser) return 0;
   terrno = TSDB_CODE_MND_NO_RIGHTS;
   return -1;
 }

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

@@ -261,6 +261,104 @@ void mndReleaseDb(SMnode *pMnode, SDbObj *pDb) {
   sdbRelease(pSdb, pDb);
 }
 
+static int32_t mndAddCreateVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid,
+                                       bool isRedo) {
+  STransAction action = {0};
+
+  SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
+  if (pDnode == NULL) return -1;
+  action.epSet = mndGetDnodeEpset(pDnode);
+  mndReleaseDnode(pMnode, pDnode);
+
+  int32_t contLen = 0;
+  void   *pReq = mndBuildCreateVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
+  if (pReq == NULL) return -1;
+
+  action.pCont = pReq;
+  action.contLen = contLen;
+  action.msgType = TDMT_DND_CREATE_VNODE;
+  action.acceptableCode = TSDB_CODE_NODE_ALREADY_DEPLOYED;
+
+  if (isRedo) {
+    if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+      taosMemoryFree(pReq);
+      return -1;
+    }
+  } else {
+    if (mndTransAppendUndoAction(pTrans, &action) != 0) {
+      taosMemoryFree(pReq);
+      return -1;
+    }
+  }
+
+  return 0;
+}
+
+static int32_t mndAddAlterVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid,
+                                      bool isRedo) {
+  STransAction action = {0};
+
+  SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
+  if (pDnode == NULL) return -1;
+  action.epSet = mndGetDnodeEpset(pDnode);
+  mndReleaseDnode(pMnode, pDnode);
+
+  int32_t contLen = 0;
+  void   *pReq = mndBuildAlterVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
+  if (pReq == NULL) return -1;
+
+  action.pCont = pReq;
+  action.contLen = contLen;
+  action.msgType = TDMT_VND_ALTER_VNODE;
+
+  if (isRedo) {
+    if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+      taosMemoryFree(pReq);
+      return -1;
+    }
+  } else {
+    if (mndTransAppendUndoAction(pTrans, &action) != 0) {
+      taosMemoryFree(pReq);
+      return -1;
+    }
+  }
+
+  return 0;
+}
+
+static int32_t mndAddDropVnodeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid,
+                                     bool isRedo) {
+  STransAction action = {0};
+
+  SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
+  if (pDnode == NULL) return -1;
+  action.epSet = mndGetDnodeEpset(pDnode);
+  mndReleaseDnode(pMnode, pDnode);
+
+  int32_t contLen = 0;
+  void   *pReq = mndBuildDropVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
+  if (pReq == NULL) return -1;
+
+  action.pCont = pReq;
+  action.contLen = contLen;
+  action.msgType = TDMT_DND_DROP_VNODE;
+  action.acceptableCode = TSDB_CODE_NODE_NOT_DEPLOYED;
+
+  if (isRedo) {
+    if (mndTransAppendRedoAction(pTrans, &action) != 0) {
+      taosMemoryFree(pReq);
+      return -1;
+    }
+  } else {
+    if (mndTransAppendUndoAction(pTrans, &action) != 0) {
+      taosMemoryFree(pReq);
+      return -1;
+    }
+  }
+
+  return 0;
+}
+
 static int32_t mndCheckDbName(const char *dbName, SUserObj *pUser) {
   char *pos = strstr(dbName, TS_PATH_DELIMITER);
   if (pos == NULL) {
@@ -278,6 +376,8 @@ static int32_t mndCheckDbName(const char *dbName, SUserObj *pUser) {
 }
 
 static int32_t mndCheckDbCfg(SMnode *pMnode, SDbCfg *pCfg) {
+  terrno = TSDB_CODE_MND_INVALID_DB_OPTION;
+
   if (pCfg->numOfVgroups < TSDB_MIN_VNODES_PER_DB || pCfg->numOfVgroups > TSDB_MAX_VNODES_PER_DB) return -1;
   if (pCfg->numOfStables < TSDB_DB_STREAM_MODE_OFF || pCfg->numOfStables > TSDB_DB_STREAM_MODE_ON) return -1;
   if (pCfg->buffer < TSDB_MIN_BUFFER_PER_VNODE || pCfg->buffer > TSDB_MAX_BUFFER_PER_VNODE) return -1;
@@ -299,10 +399,12 @@ static int32_t mndCheckDbCfg(SMnode *pMnode, SDbCfg *pCfg) {
   if (pCfg->compression < TSDB_MIN_COMP_LEVEL || pCfg->compression > TSDB_MAX_COMP_LEVEL) return -1;
   if (pCfg->replications < TSDB_MIN_DB_REPLICA || pCfg->replications > TSDB_MAX_DB_REPLICA) return -1;
   if (pCfg->replications > mndGetDnodeSize(pMnode)) return -1;
+  if (pCfg->replications != 1 && pCfg->replications != 3) return -1;
   if (pCfg->strict < TSDB_DB_STRICT_OFF || pCfg->strict > TSDB_DB_STRICT_ON) return -1;
-  if (pCfg->strict > pCfg->replications) return -1;
   if (pCfg->cacheLastRow < TSDB_MIN_DB_CACHE_LAST_ROW || pCfg->cacheLastRow > TSDB_MAX_DB_CACHE_LAST_ROW) return -1;
   if (pCfg->hashMethod != 1) return -1;
+
+  terrno = 0;
   return TSDB_CODE_SUCCESS;
 }
 
@@ -381,24 +483,8 @@ static int32_t mndSetCreateDbRedoActions(SMnode *pMnode, STrans *pTrans, SDbObj
     SVgObj *pVgroup = pVgroups + vg;
 
     for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
-      STransAction action = {0};
-      SVnodeGid   *pVgid = pVgroup->vnodeGid + vn;
-
-      SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
-      if (pDnode == NULL) return -1;
-      action.epSet = mndGetDnodeEpset(pDnode);
-      mndReleaseDnode(pMnode, pDnode);
-
-      int32_t contLen = 0;
-      void   *pReq = mndBuildCreateVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
-      if (pReq == NULL) return -1;
-
-      action.pCont = pReq;
-      action.contLen = contLen;
-      action.msgType = TDMT_DND_CREATE_VNODE;
-      action.acceptableCode = TSDB_CODE_NODE_ALREADY_DEPLOYED;
-      if (mndTransAppendRedoAction(pTrans, &action) != 0) {
-        taosMemoryFree(pReq);
+      SVnodeGid *pVgid = pVgroup->vnodeGid + vn;
+      if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, pVgroup, pVgid, true) != 0) {
         return -1;
       }
     }
@@ -412,24 +498,8 @@ static int32_t mndSetCreateDbUndoActions(SMnode *pMnode, STrans *pTrans, SDbObj
     SVgObj *pVgroup = pVgroups + vg;
 
     for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
-      STransAction action = {0};
-      SVnodeGid   *pVgid = pVgroup->vnodeGid + vn;
-
-      SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
-      if (pDnode == NULL) return -1;
-      action.epSet = mndGetDnodeEpset(pDnode);
-      mndReleaseDnode(pMnode, pDnode);
-
-      int32_t contLen = 0;
-      void   *pReq = mndBuildDropVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
-      if (pReq == NULL) return -1;
-
-      action.pCont = pReq;
-      action.contLen = contLen;
-      action.msgType = TDMT_DND_DROP_VNODE;
-      action.acceptableCode = TSDB_CODE_NODE_NOT_DEPLOYED;
-      if (mndTransAppendUndoAction(pTrans, &action) != 0) {
-        taosMemoryFree(pReq);
+      SVnodeGid *pVgid = pVgroup->vnodeGid + vn;
+      if (mndAddDropVnodeAction(pMnode, pTrans, pDb, pVgroup, pVgid, false) != 0) {
         return -1;
       }
     }
@@ -482,7 +552,6 @@ static int32_t mndCreateDb(SMnode *pMnode, SNodeMsg *pReq, SCreateDbReq *pCreate
   }
 
   if (mndCheckDbCfg(pMnode, &dbObj.cfg) != 0) {
-    terrno = TSDB_CODE_MND_INVALID_DB_OPTION;
     mError("db:%s, failed to create since %s", pCreate->db, terrstr());
     return -1;
   }
@@ -570,37 +639,37 @@ _OVER:
 static int32_t mndSetDbCfgFromAlterDbReq(SDbObj *pDb, SAlterDbReq *pAlter) {
   terrno = TSDB_CODE_MND_DB_OPTION_UNCHANGED;
 
-  if (pAlter->buffer >= 0 && pAlter->buffer != pDb->cfg.buffer) {
+  if (pAlter->buffer > 0 && pAlter->buffer != pDb->cfg.buffer) {
     pDb->cfg.buffer = pAlter->buffer;
     terrno = 0;
   }
 
-  if (pAlter->pages >= 0 && pAlter->pages != pDb->cfg.pages) {
+  if (pAlter->pages > 0 && pAlter->pages != pDb->cfg.pages) {
     pDb->cfg.pages = pAlter->pages;
     terrno = 0;
   }
 
-  if (pAlter->pageSize >= 0 && pAlter->pageSize != pDb->cfg.pageSize) {
+  if (pAlter->pageSize > 0 && pAlter->pageSize != pDb->cfg.pageSize) {
     pDb->cfg.pageSize = pAlter->pageSize;
     terrno = 0;
   }
 
-  if (pAlter->daysPerFile >= 0 && pAlter->daysPerFile != pDb->cfg.daysPerFile) {
+  if (pAlter->daysPerFile > 0 && pAlter->daysPerFile != pDb->cfg.daysPerFile) {
     pDb->cfg.daysPerFile = pAlter->daysPerFile;
     terrno = 0;
   }
 
-  if (pAlter->daysToKeep0 >= 0 && pAlter->daysToKeep0 != pDb->cfg.daysToKeep0) {
+  if (pAlter->daysToKeep0 > 0 && pAlter->daysToKeep0 != pDb->cfg.daysToKeep0) {
     pDb->cfg.daysToKeep0 = pAlter->daysToKeep0;
     terrno = 0;
   }
 
-  if (pAlter->daysToKeep1 >= 0 && pAlter->daysToKeep1 != pDb->cfg.daysToKeep1) {
+  if (pAlter->daysToKeep1 > 0 && pAlter->daysToKeep1 != pDb->cfg.daysToKeep1) {
     pDb->cfg.daysToKeep1 = pAlter->daysToKeep1;
     terrno = 0;
   }
 
-  if (pAlter->daysToKeep2 >= 0 && pAlter->daysToKeep2 != pDb->cfg.daysToKeep2) {
+  if (pAlter->daysToKeep2 > 0 && pAlter->daysToKeep2 != pDb->cfg.daysToKeep2) {
     pDb->cfg.daysToKeep2 = pAlter->daysToKeep2;
     terrno = 0;
   }
@@ -625,8 +694,9 @@ static int32_t mndSetDbCfgFromAlterDbReq(SDbObj *pDb, SAlterDbReq *pAlter) {
     terrno = 0;
   }
 
-  if (pAlter->replications >= 0 && pAlter->replications != pDb->cfg.replications) {
+  if (pAlter->replications > 0 && pAlter->replications != pDb->cfg.replications) {
     pDb->cfg.replications = pAlter->replications;
+    pDb->vgVersion++;
     terrno = 0;
   }
 
@@ -651,35 +721,57 @@ static int32_t mndSetAlterDbCommitLogs(SMnode *pMnode, STrans *pTrans, SDbObj *p
   return 0;
 }
 
-static int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup) {
-  for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
-    STransAction action = {0};
-    SVnodeGid   *pVgid = pVgroup->vnodeGid + vn;
-
-    SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
-    if (pDnode == NULL) return -1;
-    action.epSet = mndGetDnodeEpset(pDnode);
-    mndReleaseDnode(pMnode, pDnode);
-
-    int32_t contLen = 0;
-    void   *pReq = mndBuildAlterVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
-    if (pReq == NULL) return -1;
+static int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SArray *pArray) {
+  if (pVgroup->replica <= 0 || pVgroup->replica == pDb->cfg.replications) {
+    for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
+      SVnodeGid *pVgid = pVgroup->vnodeGid + vn;
+      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, pVgroup, pVgid, true) != 0) {
+        return -1;
+      }
+    }
+  } else {
+    SVgObj newVgroup = {0};
+    memcpy(&newVgroup, pVgroup, sizeof(SVgObj));
+    if (newVgroup.replica < pDb->cfg.replications) {
+      mInfo("db:%s, vgId:%d, will add 2 vnodes, vn:0 dnode:%d", pVgroup->dbName, pVgroup->vgId,
+            pVgroup->vnodeGid[0].dnodeId);
+
+      if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) {
+        mError("db:%s, failed to add vnode to vgId:%d since %s", pDb->name, newVgroup.vgId, terrstr());
+        return -1;
+      }
+      newVgroup.replica = pDb->cfg.replications;
+      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[0], true) != 0) return -1;
+      if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[1], true) != 0) return -1;
+      if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[2], true) != 0) return -1;
+    } else {
+      mInfo("db:%s, vgId:%d, will remove 2 vnodes", pVgroup->dbName, pVgroup->vgId);
 
-    action.pCont = pReq;
-    action.contLen = contLen;
-    action.msgType = TDMT_VND_ALTER_VNODE;
-    if (mndTransAppendRedoAction(pTrans, &action) != 0) {
-      taosMemoryFree(pReq);
-      return -1;
+      SVnodeGid del1 = {0};
+      SVnodeGid del2 = {0};
+      if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del1, &del2) != 0) {
+        mError("db:%s, failed to remove vnode from vgId:%d since %s", pDb->name, newVgroup.vgId, terrstr());
+        return -1;
+      }
+      newVgroup.replica = pDb->cfg.replications;
+      if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[0], true) != 0) return -1;
+      if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del1, true) != 0) return -1;
+      if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del2, true) != 0) return -1;
     }
+
+    SSdbRaw *pVgRaw = mndVgroupActionEncode(&newVgroup);
+    if (pVgRaw == NULL) return -1;
+    if (mndTransAppendCommitlog(pTrans, pVgRaw) != 0) return -1;
+    if (sdbSetRawStatus(pVgRaw, SDB_STATUS_READY) != 0) return -1;
   }
 
   return 0;
 }
 
 static int32_t mndSetAlterDbRedoActions(SMnode *pMnode, STrans *pTrans, SDbObj *pOld, SDbObj *pNew) {
-  SSdb *pSdb = pMnode->pSdb;
-  void *pIter = NULL;
+  SSdb   *pSdb = pMnode->pSdb;
+  void   *pIter = NULL;
+  SArray *pArray = mndBuildDnodesArray(pMnode);
 
   while (1) {
     SVgObj *pVgroup = NULL;
@@ -687,9 +779,10 @@ static int32_t mndSetAlterDbRedoActions(SMnode *pMnode, STrans *pTrans, SDbObj *
     if (pIter == NULL) break;
 
     if (pVgroup->dbUid == pNew->uid) {
-      if (mndBuildAlterVgroupAction(pMnode, pTrans, pNew, pVgroup) != 0) {
+      if (mndBuildAlterVgroupAction(pMnode, pTrans, pNew, pVgroup, pArray) != 0) {
         sdbCancelFetch(pSdb, pIter);
         sdbRelease(pSdb, pVgroup);
+        taosArrayDestroy(pArray);
         return -1;
       }
     }
@@ -697,6 +790,7 @@ static int32_t mndSetAlterDbRedoActions(SMnode *pMnode, STrans *pTrans, SDbObj *
     sdbRelease(pSdb, pVgroup);
   }
 
+  taosArrayDestroy(pArray);
   return 0;
 }
 
@@ -726,6 +820,7 @@ static int32_t mndProcessAlterDbReq(SNodeMsg *pReq) {
   SDbObj     *pDb = NULL;
   SUserObj   *pUser = NULL;
   SAlterDbReq alterReq = {0};
+  SDbObj      dbObj = {0};
 
   if (tDeserializeSAlterDbReq(pReq->rpcMsg.pCont, pReq->rpcMsg.contLen, &alterReq) != 0) {
     terrno = TSDB_CODE_INVALID_MSG;
@@ -749,15 +844,17 @@ static int32_t mndProcessAlterDbReq(SNodeMsg *pReq) {
     goto _OVER;
   }
 
-  SDbObj dbObj = {0};
   memcpy(&dbObj, pDb, sizeof(SDbObj));
-  dbObj.cfg.numOfRetensions = 0;
-  dbObj.cfg.pRetensions = NULL;
+  if (dbObj.cfg.pRetensions != NULL) {
+    dbObj.cfg.pRetensions = taosArrayDup(pDb->cfg.pRetensions);
+    if (dbObj.cfg.pRetensions == NULL) goto _OVER;
+  }
 
   code = mndSetDbCfgFromAlterDbReq(&dbObj, &alterReq);
-  if (code != 0) {
-    goto _OVER;
-  }
+  if (code != 0) goto _OVER;
+
+  code = mndCheckDbCfg(pMnode, &dbObj.cfg);
+  if (code != 0) goto _OVER;
 
   dbObj.cfgVersion++;
   dbObj.updateTime = taosGetTimestampMs();
@@ -771,6 +868,7 @@ _OVER:
 
   mndReleaseDb(pMnode, pDb);
   mndReleaseUser(pMnode, pUser);
+  taosArrayDestroy(dbObj.cfg.pRetensions);
 
   return code;
 }
@@ -899,24 +997,8 @@ static int32_t mndSetDropDbCommitLogs(SMnode *pMnode, STrans *pTrans, SDbObj *pD
 
 static int32_t mndBuildDropVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup) {
   for (int32_t vn = 0; vn < pVgroup->replica; ++vn) {
-    STransAction action = {0};
-    SVnodeGid   *pVgid = pVgroup->vnodeGid + vn;
-
-    SDnodeObj *pDnode = mndAcquireDnode(pMnode, pVgid->dnodeId);
-    if (pDnode == NULL) return -1;
-    action.epSet = mndGetDnodeEpset(pDnode);
-    mndReleaseDnode(pMnode, pDnode);
-
-    int32_t contLen = 0;
-    void   *pReq = mndBuildDropVnodeReq(pMnode, pDnode, pDb, pVgroup, &contLen);
-    if (pReq == NULL) return -1;
-
-    action.pCont = pReq;
-    action.contLen = contLen;
-    action.msgType = TDMT_DND_DROP_VNODE;
-    action.acceptableCode = TSDB_CODE_NODE_NOT_DEPLOYED;
-    if (mndTransAppendRedoAction(pTrans, &action) != 0) {
-      taosMemoryFree(pReq);
+    SVnodeGid *pVgid = pVgroup->vnodeGid + vn;
+    if (mndAddDropVnodeAction(pMnode, pTrans, pDb, pVgroup, pVgid, true) != 0) {
       return -1;
     }
   }
@@ -988,6 +1070,17 @@ static int32_t mndDropDb(SMnode *pMnode, SNodeMsg *pReq, SDbObj *pDb) {
   /*if (mndDropTopicByDB(pMnode, pTrans, pDb) != 0) goto _OVER;*/
   if (mndSetDropDbRedoActions(pMnode, pTrans, pDb) != 0) goto _OVER;
 
+  SUserObj *pUser = mndAcquireUser(pMnode, pDb->createUser);
+  if (pUser != NULL) {
+    pUser->authVersion++;
+    SSdbRaw *pCommitRaw = mndUserActionEncode(pUser);
+    if (pCommitRaw == NULL || mndTransAppendCommitlog(pTrans, pCommitRaw) != 0) {
+      mError("trans:%d, failed to append redo log since %s", pTrans->id, terrstr());
+      goto _OVER;
+    }
+    sdbSetRawStatus(pCommitRaw, SDB_STATUS_READY);
+  }
+
   int32_t rspLen = 0;
   void   *pRsp = NULL;
   if (mndBuildDropDbRsp(pDb, &rspLen, &pRsp, false) < 0) goto _OVER;

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

@@ -21,7 +21,7 @@
 #include "mndInfoSchema.h"
 #include "mndMnode.h"
 #include "mndShow.h"
-#include "mndStb.c"
+#include "mndStb.h"
 #include "mndStream.h"
 #include "mndTrans.h"
 #include "mndUser.h"

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

@@ -13,6 +13,7 @@
  * along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
+#define _DEFAULT_SOURCE
 #include "mndStb.h"
 #include "mndAuth.h"
 #include "mndDb.h"
@@ -342,7 +343,7 @@ void mndReleaseStb(SMnode *pMnode, SStbObj *pStb) {
   sdbRelease(pSdb, pStb);
 }
 
-static SDbObj *mndAcquireDbByStb(SMnode *pMnode, const char *stbName) {
+SDbObj *mndAcquireDbByStb(SMnode *pMnode, const char *stbName) {
   SName name = {0};
   tNameFromString(&name, stbName, T_NAME_ACCT | T_NAME_DB | T_NAME_TABLE);
 
@@ -463,7 +464,7 @@ static void *mndBuildVDropStbReq(SMnode *pMnode, SVgObj *pVgroup, SStbObj *pStb,
   return pHead;
 }
 
-static int32_t mndCheckCreateStbReq(SMCreateStbReq *pCreate) {
+int32_t mndCheckCreateStbReq(SMCreateStbReq *pCreate) {
   if (pCreate->igExists < 0 || pCreate->igExists > 1) {
     terrno = TSDB_CODE_MND_INVALID_STB_OPTION;
     return -1;
@@ -634,91 +635,96 @@ static SSchema *mndFindStbColumns(const SStbObj *pStb, const char *colName) {
   return NULL;
 }
 
-static int32_t mndCreateStb(SMnode *pMnode, SNodeMsg *pReq, SMCreateStbReq *pCreate, SDbObj *pDb) {
-  SStbObj stbObj = {0};
-  memcpy(stbObj.name, pCreate->name, TSDB_TABLE_FNAME_LEN);
-  memcpy(stbObj.db, pDb->name, TSDB_DB_FNAME_LEN);
-  stbObj.createdTime = taosGetTimestampMs();
-  stbObj.updateTime = stbObj.createdTime;
-  stbObj.uid = mndGenerateUid(pCreate->name, TSDB_TABLE_FNAME_LEN);
-  stbObj.dbUid = pDb->uid;
-  stbObj.version = 1;
-  stbObj.nextColId = 1;
-  stbObj.xFilesFactor = pCreate->xFilesFactor;
-  stbObj.delay = pCreate->delay;
-  stbObj.ttl = pCreate->ttl;
-  stbObj.numOfColumns = pCreate->numOfColumns;
-  stbObj.numOfTags = pCreate->numOfTags;
-  stbObj.commentLen = pCreate->commentLen;
-  if (stbObj.commentLen > 0) {
-    stbObj.comment = taosMemoryCalloc(stbObj.commentLen, 1);
-    if (stbObj.comment == NULL) {
+int32_t mndBuildStbFromReq(SMnode *pMnode, SStbObj *pDst, SMCreateStbReq *pCreate, SDbObj *pDb) {
+  memcpy(pDst->name, pCreate->name, TSDB_TABLE_FNAME_LEN);
+  memcpy(pDst->db, pDb->name, TSDB_DB_FNAME_LEN);
+  pDst->createdTime = taosGetTimestampMs();
+  pDst->updateTime = pDst->createdTime;
+  pDst->uid = mndGenerateUid(pCreate->name, TSDB_TABLE_FNAME_LEN);
+  pDst->dbUid = pDb->uid;
+  pDst->version = 1;
+  pDst->nextColId = 1;
+  pDst->xFilesFactor = pCreate->xFilesFactor;
+  pDst->delay = pCreate->delay;
+  pDst->ttl = pCreate->ttl;
+  pDst->numOfColumns = pCreate->numOfColumns;
+  pDst->numOfTags = pCreate->numOfTags;
+  pDst->commentLen = pCreate->commentLen;
+  if (pDst->commentLen > 0) {
+    pDst->comment = taosMemoryCalloc(pDst->commentLen, 1);
+    if (pDst->comment == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
       return -1;
     }
-    memcpy(stbObj.comment, pCreate->comment, stbObj.commentLen);
+    memcpy(pDst->comment, pCreate->comment, pDst->commentLen);
   }
 
-  stbObj.ast1Len = pCreate->ast1Len;
-  if (stbObj.ast1Len > 0) {
-    stbObj.pAst1 = taosMemoryCalloc(stbObj.ast1Len, 1);
-    if (stbObj.pAst1 == NULL) {
+  pDst->ast1Len = pCreate->ast1Len;
+  if (pDst->ast1Len > 0) {
+    pDst->pAst1 = taosMemoryCalloc(pDst->ast1Len, 1);
+    if (pDst->pAst1 == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
       return -1;
     }
-    memcpy(stbObj.pAst1, pCreate->pAst1, stbObj.ast1Len);
+    memcpy(pDst->pAst1, pCreate->pAst1, pDst->ast1Len);
   }
 
-  stbObj.ast2Len = pCreate->ast2Len;
-  if (stbObj.ast2Len > 0) {
-    stbObj.pAst2 = taosMemoryCalloc(stbObj.ast2Len, 1);
-    if (stbObj.pAst2 == NULL) {
+  pDst->ast2Len = pCreate->ast2Len;
+  if (pDst->ast2Len > 0) {
+    pDst->pAst2 = taosMemoryCalloc(pDst->ast2Len, 1);
+    if (pDst->pAst2 == NULL) {
       terrno = TSDB_CODE_OUT_OF_MEMORY;
       return -1;
     }
-    memcpy(stbObj.pAst2, pCreate->pAst2, stbObj.ast2Len);
+    memcpy(pDst->pAst2, pCreate->pAst2, pDst->ast2Len);
   }
 
-  stbObj.pColumns = taosMemoryCalloc(1, stbObj.numOfColumns * sizeof(SSchema));
-  stbObj.pTags = taosMemoryCalloc(1, stbObj.numOfTags * sizeof(SSchema));
-  if (stbObj.pColumns == NULL || stbObj.pTags == NULL) {
+  pDst->pColumns = taosMemoryCalloc(1, pDst->numOfColumns * sizeof(SSchema));
+  pDst->pTags = taosMemoryCalloc(1, pDst->numOfTags * sizeof(SSchema));
+  if (pDst->pColumns == NULL || pDst->pTags == NULL) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
   }
 
-  for (int32_t i = 0; i < stbObj.numOfColumns; ++i) {
+  for (int32_t i = 0; i < pDst->numOfColumns; ++i) {
     SField  *pField = taosArrayGet(pCreate->pColumns, i);
-    SSchema *pSchema = &stbObj.pColumns[i];
+    SSchema *pSchema = &pDst->pColumns[i];
     pSchema->type = pField->type;
     pSchema->bytes = pField->bytes;
     pSchema->flags = pField->flags;
     memcpy(pSchema->name, pField->name, TSDB_COL_NAME_LEN);
-    pSchema->colId = stbObj.nextColId;
-    stbObj.nextColId++;
+    pSchema->colId = pDst->nextColId;
+    pDst->nextColId++;
   }
 
-  for (int32_t i = 0; i < stbObj.numOfTags; ++i) {
+  for (int32_t i = 0; i < pDst->numOfTags; ++i) {
     SField  *pField = taosArrayGet(pCreate->pTags, i);
-    SSchema *pSchema = &stbObj.pTags[i];
+    SSchema *pSchema = &pDst->pTags[i];
     pSchema->type = pField->type;
     pSchema->bytes = pField->bytes;
     memcpy(pSchema->name, pField->name, TSDB_COL_NAME_LEN);
-    pSchema->colId = stbObj.nextColId;
-    stbObj.nextColId++;
+    pSchema->colId = pDst->nextColId;
+    pDst->nextColId++;
   }
+  return 0;
+}
+
+static int32_t mndCreateStb(SMnode *pMnode, SNodeMsg *pReq, SMCreateStbReq *pCreate, SDbObj *pDb) {
+  SStbObj stbObj = {0};
 
   int32_t code = -1;
+
   STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_TYPE_CREATE_STB, &pReq->rpcMsg);
   if (pTrans == NULL) goto _OVER;
 
   mDebug("trans:%d, used to create stb:%s", pTrans->id, pCreate->name);
-  mndTransSetDbInfo(pTrans, pDb);
 
-  if (mndSetCreateStbRedoLogs(pMnode, pTrans, pDb, &stbObj) != 0) goto _OVER;
-  if (mndSetCreateStbUndoLogs(pMnode, pTrans, pDb, &stbObj) != 0) goto _OVER;
-  if (mndSetCreateStbCommitLogs(pMnode, pTrans, pDb, &stbObj) != 0) goto _OVER;
-  if (mndSetCreateStbRedoActions(pMnode, pTrans, pDb, &stbObj) != 0) goto _OVER;
-  if (mndSetCreateStbUndoActions(pMnode, pTrans, pDb, &stbObj) != 0) goto _OVER;
+  if (mndBuildStbFromReq(pMnode, &stbObj, pCreate, pDb) != 0) {
+    goto _OVER;
+  }
+
+  if (mndAddStbToTrans(pMnode, pTrans, pDb, &stbObj) < 0) goto _OVER;
+
   if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;
 
   code = 0;
@@ -727,6 +733,15 @@ _OVER:
   mndTransDrop(pTrans);
   return code;
 }
+int32_t mndAddStbToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SStbObj *pStb) {
+  mndTransSetDbInfo(pTrans, pDb);
+  if (mndSetCreateStbRedoLogs(pMnode, pTrans, pDb, pStb) != 0) return -1;
+  if (mndSetCreateStbUndoLogs(pMnode, pTrans, pDb, pStb) != 0) return -1;
+  if (mndSetCreateStbCommitLogs(pMnode, pTrans, pDb, pStb) != 0) return -1;
+  if (mndSetCreateStbRedoActions(pMnode, pTrans, pDb, pStb) != 0) return -1;
+  if (mndSetCreateStbUndoActions(pMnode, pTrans, pDb, pStb) != 0) return -1;
+  return 0;
+}
 
 static int32_t mndProcessMCreateStbReq(SNodeMsg *pReq) {
   SMnode        *pMnode = pReq->pNode;

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

@@ -290,6 +290,72 @@ int32_t mndAddStreamToTrans(SMnode *pMnode, SStreamObj *pStream, const char *ast
   return 0;
 }
 
+static SStbObj *mndCreateStbForStream(SMnode *pMnode, STrans *pTrans, const SStreamObj *pStream, const char *user) {
+  SStbObj  *pStb = NULL;
+  SDbObj   *pDb = NULL;
+  SUserObj *pUser = NULL;
+
+  SMCreateStbReq createReq = {0};
+  tstrncpy(createReq.name, pStream->targetSTbName, TSDB_TABLE_FNAME_LEN);
+  createReq.numOfColumns = pStream->outputSchema.nCols;
+  createReq.numOfTags = 1;  // group id
+  createReq.pColumns = taosArrayInit(createReq.numOfColumns, sizeof(SField));
+  // build fields
+  // build tags
+
+  if (mndCheckCreateStbReq(&createReq) != 0) {
+    terrno = TSDB_CODE_INVALID_MSG;
+    goto _OVER;
+  }
+
+  pStb = mndAcquireStb(pMnode, createReq.name);
+  if (pStb != NULL) {
+    terrno = TSDB_CODE_MND_STB_ALREADY_EXIST;
+    goto _OVER;
+  }
+
+  pDb = mndAcquireDbByStb(pMnode, createReq.name);
+  if (pDb == NULL) {
+    terrno = TSDB_CODE_MND_DB_NOT_SELECTED;
+    goto _OVER;
+  }
+
+  pUser = mndAcquireUser(pMnode, user);
+  if (pUser == NULL) {
+    goto _OVER;
+  }
+
+  if (mndCheckWriteAuth(pUser, pDb) != 0) {
+    goto _OVER;
+  }
+
+  int32_t numOfStbs = -1;
+  mndGetNumOfStbs(pMnode, pDb->name, &numOfStbs);
+  if (pDb->cfg.numOfStables == 1 && numOfStbs != 0) {
+    terrno = TSDB_CODE_MND_SINGLE_STB_MODE_DB;
+    goto _OVER;
+  }
+
+  SStbObj stbObj = {0};
+
+  if (mndBuildStbFromReq(pMnode, &stbObj, &createReq, pDb) != 0) {
+    goto _OVER;
+  }
+
+  if (mndBuildStbFromReq(pMnode, pStb, &createReq, pDb) != 0) {
+    goto _OVER;
+  }
+
+  if (mndAddStbToTrans(pMnode, pTrans, pDb, &stbObj) < 0) goto _OVER;
+
+  return pStb;
+_OVER:
+  mndReleaseStb(pMnode, pStb);
+  mndReleaseDb(pMnode, pDb);
+  mndReleaseUser(pMnode, pUser);
+  return NULL;
+}
+
 static int32_t mndCreateStream(SMnode *pMnode, SNodeMsg *pReq, SCMCreateStreamReq *pCreate, SDbObj *pDb) {
   mDebug("stream:%s to create", pCreate->name);
   SStreamObj streamObj = {0};
@@ -311,13 +377,21 @@ static int32_t mndCreateStream(SMnode *pMnode, SNodeMsg *pReq, SCMCreateStreamRe
   streamObj.trigger = pCreate->triggerType;
   streamObj.waterMark = pCreate->watermark;
 
-  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_TYPE_CREATE_STREAM, &pReq->rpcMsg);
+  STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_TYPE_CREATE_STREAM, &pReq->rpcMsg);
   if (pTrans == NULL) {
     mError("stream:%s, failed to create since %s", pCreate->name, terrstr());
     return -1;
   }
   mDebug("trans:%d, used to create stream:%s", pTrans->id, pCreate->name);
 
+#if 0
+  if (mndCreateStbForStream(pMnode, pTrans, &streamObj, pReq->user) < 0) {
+    mError("trans:%d, failed to add stream since %s", pTrans->id, terrstr());
+    mndTransDrop(pTrans);
+    return -1;
+  }
+#endif
+
   if (mndAddStreamToTrans(pMnode, &streamObj, pCreate->ast, pCreate->triggerType, pCreate->watermark, pTrans) != 0) {
     mError("trans:%d, failed to add stream since %s", pTrans->id, terrstr());
     mndTransDrop(pTrans);

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

@@ -1336,8 +1336,7 @@ static int32_t mndProcessKillTransReq(SNodeMsg *pReq) {
     goto _OVER;
   }
 
-  if (!pUser->superUser) {
-    terrno = TSDB_CODE_MND_NO_RIGHTS;
+  if (mndCheckTransAuth(pUser) != 0) {
     goto _OVER;
   }
 

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

@@ -106,6 +106,7 @@ SSdbRaw *mndUserActionEncode(SUserObj *pUser) {
   SDB_SET_INT64(pRaw, dataPos, pUser->createdTime, _OVER)
   SDB_SET_INT64(pRaw, dataPos, pUser->updateTime, _OVER)
   SDB_SET_INT8(pRaw, dataPos, pUser->superUser, _OVER)
+  SDB_SET_INT32(pRaw, dataPos, pUser->authVersion, _OVER)
   SDB_SET_INT32(pRaw, dataPos, numOfReadDbs, _OVER)
   SDB_SET_INT32(pRaw, dataPos, numOfWriteDbs, _OVER)
 
@@ -161,6 +162,7 @@ static SSdbRow *mndUserActionDecode(SSdbRaw *pRaw) {
   SDB_GET_INT64(pRaw, dataPos, &pUser->createdTime, _OVER)
   SDB_GET_INT64(pRaw, dataPos, &pUser->updateTime, _OVER)
   SDB_GET_INT8(pRaw, dataPos, &pUser->superUser, _OVER)
+  SDB_GET_INT32(pRaw, dataPos, &pUser->authVersion, _OVER)
 
   int32_t numOfReadDbs = 0;
   int32_t numOfWriteDbs = 0;
@@ -588,7 +590,7 @@ _OVER:
   return code;
 }
 
-static int32_t mndSetUserAuthRsp(SMnode       *pMnode, SUserObj *pUser, SGetUserAuthRsp *pRsp) {
+static int32_t mndSetUserAuthRsp(SMnode *pMnode, SUserObj *pUser, SGetUserAuthRsp *pRsp) {
   memcpy(pRsp->user, pUser->user, TSDB_USER_LEN);
   pRsp->superAuth = pUser->superUser;
   pRsp->version = pUser->authVersion;
@@ -601,7 +603,7 @@ static int32_t mndSetUserAuthRsp(SMnode       *pMnode, SUserObj *pUser, SGetUser
     terrno = TSDB_CODE_OUT_OF_MEMORY;
     return -1;
   }
-  
+
   SSdb *pSdb = pMnode->pSdb;
   void *pIter = NULL;
   while (1) {
@@ -659,7 +661,7 @@ static int32_t mndProcessGetUserAuthReq(SNodeMsg *pReq) {
   code = 0;
 
 _OVER:
-  
+
   mndReleaseUser(pMnode, pUser);
   tFreeSGetUserAuthRsp(&authRsp);
 
@@ -711,7 +713,8 @@ static void mndCancelGetNextUser(SMnode *pMnode, void *pIter) {
   sdbCancelFetch(pSdb, pIter);
 }
 
-int32_t mndValidateUserAuthInfo(SMnode *pMnode, SUserAuthVersion *pUsers, int32_t numOfUses, void **ppRsp, int32_t *pRspLen) {
+int32_t mndValidateUserAuthInfo(SMnode *pMnode, SUserAuthVersion *pUsers, int32_t numOfUses, void **ppRsp,
+                                int32_t *pRspLen) {
   SUserAuthBatchRsp batchRsp = {0};
   batchRsp.pArray = taosArrayInit(numOfUses, sizeof(SGetUserAuthRsp));
   if (batchRsp.pArray == NULL) {
@@ -731,7 +734,7 @@ int32_t mndValidateUserAuthInfo(SMnode *pMnode, SUserAuthVersion *pUsers, int32_
       mndReleaseUser(pMnode, pUser);
       continue;
     }
-    
+
     SGetUserAuthRsp rsp = {0};
     code = mndSetUserAuthRsp(pMnode, pUser, &rsp);
     if (code) {
@@ -740,7 +743,6 @@ int32_t mndValidateUserAuthInfo(SMnode *pMnode, SUserAuthVersion *pUsers, int32_
       goto _OVER;
     }
 
-
     taosArrayPush(batchRsp.pArray, &rsp);
     mndReleaseUser(pMnode, pUser);
   }
@@ -748,7 +750,7 @@ int32_t mndValidateUserAuthInfo(SMnode *pMnode, SUserAuthVersion *pUsers, int32_
   if (taosArrayGetSize(batchRsp.pArray) <= 0) {
     *ppRsp = NULL;
     *pRspLen = 0;
-    
+
     tFreeSUserAuthBatchRsp(&batchRsp);
     return 0;
   }
@@ -772,10 +774,7 @@ _OVER:
 
   *ppRsp = NULL;
   *pRspLen = 0;
-  
+
   tFreeSUserAuthBatchRsp(&batchRsp);
   return code;
 }
-
-
-

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

@@ -370,7 +370,7 @@ static bool mndBuildDnodesArrayFp(SMnode *pMnode, void *pObj, void *p1, void *p2
   return true;
 }
 
-static SArray *mndBuildDnodesArray(SMnode *pMnode) {
+SArray *mndBuildDnodesArray(SMnode *pMnode) {
   SSdb   *pSdb = pMnode->pSdb;
   int32_t numOfDnodes = mndGetDnodeSize(pMnode);
 
@@ -421,7 +421,7 @@ static int32_t mndGetAvailableDnode(SMnode *pMnode, SVgObj *pVgroup, SArray *pAr
       pVgid->role = TAOS_SYNC_STATE_FOLLOWER;
     }
 
-    mDebug("db:%s, vgId:%d, vn:%d dnode:%d is alloced", pVgroup->dbName, pVgroup->vgId, v, pVgid->dnodeId);
+    mInfo("db:%s, vgId:%d, vn:%d dnode:%d is alloced", pVgroup->dbName, pVgroup->vgId, v, pVgid->dnodeId);
     pDnode->numOfVnodes++;
   }
 
@@ -440,12 +440,10 @@ int32_t mndAllocVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj **ppVgroups) {
   }
 
   pArray = mndBuildDnodesArray(pMnode);
-  if (pArray == NULL) {
-    goto _OVER;
-  }
+  if (pArray == NULL) goto _OVER;
 
-  mDebug("db:%s, total %d dnodes used to create %d vgroups (%d vnodes)", pDb->name, (int32_t)taosArrayGetSize(pArray),
-         pDb->cfg.numOfVgroups, pDb->cfg.numOfVgroups * pDb->cfg.replications);
+  mInfo("db:%s, total %d dnodes used to create %d vgroups (%d vnodes)", pDb->name, (int32_t)taosArrayGetSize(pArray),
+        pDb->cfg.numOfVgroups, pDb->cfg.numOfVgroups * pDb->cfg.replications);
 
   int32_t  allocedVgroups = 0;
   int32_t  maxVgId = sdbGetMaxId(pMnode->pSdb, SDB_VGROUP);
@@ -483,7 +481,7 @@ int32_t mndAllocVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj **ppVgroups) {
   *ppVgroups = pVgroups;
   code = 0;
 
-  mDebug("db:%s, %d vgroups is alloced, replica:%d", pDb->name, pDb->cfg.numOfVgroups, pDb->cfg.replications);
+  mInfo("db:%s, %d vgroups is alloced, replica:%d", pDb->name, pDb->cfg.numOfVgroups, pDb->cfg.replications);
 
 _OVER:
   if (code != 0) taosMemoryFree(pVgroups);
@@ -491,6 +489,88 @@ _OVER:
   return code;
 }
 
+int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
+  taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
+  for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
+    SDnodeObj *pDnode = taosArrayGet(pArray, i);
+    mDebug("dnode:%d, equivalent vnodes:%d", pDnode->id, pDnode->numOfVnodes);
+  }
+
+  int32_t maxPos = 1;
+  for (int32_t d = 0; d < taosArrayGetSize(pArray); ++d) {
+    SDnodeObj *pDnode = taosArrayGet(pArray, d);
+
+    bool used = false;
+    for (int32_t vn = 0; vn < maxPos; ++vn) {
+      if (pDnode->id == pVgroup->vnodeGid[vn].dnodeId) {
+        used = true;
+        break;
+      }
+    }
+    if (used) continue;
+
+    if (pDnode == NULL || pDnode->numOfVnodes > pDnode->numOfSupportVnodes) {
+      terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
+      return -1;
+    }
+
+    SVnodeGid *pVgid = &pVgroup->vnodeGid[maxPos];
+    pVgid->dnodeId = pDnode->id;
+    pVgid->role = TAOS_SYNC_STATE_FOLLOWER;
+    pDnode->numOfVnodes++;
+
+    mInfo("db:%s, vgId:%d, vn:%d dnode:%d is added", pVgroup->dbName, pVgroup->vgId, maxPos, pVgid->dnodeId);
+    maxPos++;
+    if (maxPos == 3) return 0;
+  }
+
+  terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
+  return -1;
+}
+
+int32_t mndRemoveVnodeFromVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray, SVnodeGid *del1, SVnodeGid *del2) {
+  int32_t removedNum = 0;
+
+  taosArraySort(pArray, (__compar_fn_t)mndCompareDnodeVnodes);
+  for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) {
+    SDnodeObj *pDnode = taosArrayGet(pArray, i);
+    mDebug("dnode:%d, equivalent vnodes:%d", pDnode->id, pDnode->numOfVnodes);
+  }
+
+  for (int32_t d = taosArrayGetSize(pArray) - 1; d >= 0; --d) {
+    SDnodeObj *pDnode = taosArrayGet(pArray, d);
+
+    for (int32_t vn = 0; vn < TSDB_MAX_REPLICA; ++vn) {
+      SVnodeGid *pVgid = &pVgroup->vnodeGid[vn];
+      if (pVgid->dnodeId == pDnode->id) {
+        if (removedNum == 0) *del1 = *pVgid;
+        if (removedNum == 1) *del2 = *pVgid;
+
+        mInfo("db:%s, vgId:%d, vn:%d dnode:%d is removed", pVgroup->dbName, pVgroup->vgId, vn, pVgid->dnodeId);
+        memset(pVgid, 0, sizeof(SVnodeGid));
+        removedNum++;
+        pDnode->numOfVnodes--;
+
+        if (removedNum == 2) goto _OVER;
+      }
+    }
+  }
+
+_OVER:
+  if (removedNum != 2) return -1;
+
+  for (int32_t vn = 1; vn < TSDB_MAX_REPLICA; ++vn) {
+    SVnodeGid *pVgid = &pVgroup->vnodeGid[vn];
+    if (pVgid->dnodeId != 0) {
+      memcpy(&pVgroup->vnodeGid[0], pVgid, sizeof(SVnodeGid));
+      memset(pVgid, 0, sizeof(SVnodeGid));
+    }
+  }
+
+  mInfo("db:%s, vgId:%d, dnode:%d is keeped", pVgroup->dbName, pVgroup->vgId, pVgroup->vnodeGid[0].dnodeId);
+  return 0;
+}
+
 SEpSet mndGetVgroupEpset(SMnode *pMnode, const SVgObj *pVgroup) {
   SEpSet epset = {0};
 

source/dnode/mnode/impl/test/db/db.cpp

@@ -73,14 +73,19 @@ TEST_F(MndTestDb, 02_Create_Alter_Drop_Db) {
   {
     SAlterDbReq alterdbReq = {0};
     strcpy(alterdbReq.db, "1.d1");
+
     alterdbReq.buffer = 12;
-    alterdbReq.daysToKeep0 = 300;
-    alterdbReq.daysToKeep1 = 400;
-    alterdbReq.daysToKeep2 = 500;
+    alterdbReq.pageSize = -1;
+    alterdbReq.pages = -1;
+    alterdbReq.daysPerFile = -1;
+    alterdbReq.daysToKeep0 = -1;
+    alterdbReq.daysToKeep1 = -1;
+    alterdbReq.daysToKeep2 = -1;
     alterdbReq.fsyncPeriod = 4000;
     alterdbReq.walLevel = 2;
-    alterdbReq.strict = 2;
+    alterdbReq.strict = 1;
     alterdbReq.cacheLastRow = 1;
+    alterdbReq.replications = 1;
 
     int32_t contLen = tSerializeSAlterDbReq(NULL, 0, &alterdbReq);
     void*   pReq = rpcMallocCont(contLen);

source/dnode/mnode/impl/test/dnode/mdnode.cpp

@@ -266,4 +266,96 @@ TEST_F(MndTestDnode, 05_Create_Drop_Restart_Dnode) {
   taosMsleep(1300);
   test.SendShowReq(TSDB_MGMT_TABLE_DNODE, "dnodes", "");
   EXPECT_EQ(test.GetShowRows(), 4);
+
+  // alter replica
+#if 0
+  {
+    SCreateDbReq createReq = {0};
+    strcpy(createReq.db, "1.d2");
+    createReq.numOfVgroups = 2;
+    createReq.buffer = -1;
+    createReq.pageSize = -1;
+    createReq.pages = -1;
+    createReq.daysPerFile = 1000;
+    createReq.daysToKeep0 = 3650;
+    createReq.daysToKeep1 = 3650;
+    createReq.daysToKeep2 = 3650;
+    createReq.minRows = 100;
+    createReq.maxRows = 4096;
+    createReq.fsyncPeriod = 3000;
+    createReq.walLevel = 1;
+    createReq.precision = 0;
+    createReq.compression = 2;
+    createReq.replications = 1;
+    createReq.strict = 1;
+    createReq.cacheLastRow = 0;
+    createReq.ignoreExist = 1;
+    createReq.numOfStables = 0;
+    createReq.numOfRetensions = 0;
+
+    int32_t contLen = tSerializeSCreateDbReq(NULL, 0, &createReq);
+    void*   pReq = rpcMallocCont(contLen);
+    tSerializeSCreateDbReq(pReq, contLen, &createReq);
+
+    SRpcMsg* pRsp = test.SendReq(TDMT_MND_CREATE_DB, pReq, contLen);
+    ASSERT_NE(pRsp, nullptr);
+    ASSERT_EQ(pRsp->code, 0);
+
+    test.SendShowReq(TSDB_MGMT_TABLE_DB, "user_databases", "");
+    EXPECT_EQ(test.GetShowRows(), 3);
+  }
+
+  {
+    SAlterDbReq alterdbReq = {0};
+    strcpy(alterdbReq.db, "1.d2");
+
+    alterdbReq.buffer = 12;
+    alterdbReq.pageSize = -1;
+    alterdbReq.pages = -1;
+    alterdbReq.daysPerFile = -1;
+    alterdbReq.daysToKeep0 = -1;
+    alterdbReq.daysToKeep1 = -1;
+    alterdbReq.daysToKeep2 = -1;
+    alterdbReq.fsyncPeriod = 4000;
+    alterdbReq.walLevel = 2;
+    alterdbReq.strict = 1;
+    alterdbReq.cacheLastRow = 1;
+    alterdbReq.replications = 3;
+
+    int32_t contLen = tSerializeSAlterDbReq(NULL, 0, &alterdbReq);
+    void*   pReq = rpcMallocCont(contLen);
+    tSerializeSAlterDbReq(pReq, contLen, &alterdbReq);
+
+    SRpcMsg* pRsp = test.SendReq(TDMT_MND_ALTER_DB, pReq, contLen);
+    ASSERT_NE(pRsp, nullptr);
+    ASSERT_EQ(pRsp->code, 0);
+  }
+
+  {
+    SAlterDbReq alterdbReq = {0};
+    strcpy(alterdbReq.db, "1.d2");
+
+    alterdbReq.buffer = 12;
+    alterdbReq.pageSize = -1;
+    alterdbReq.pages = -1;
+    alterdbReq.daysPerFile = -1;
+    alterdbReq.daysToKeep0 = -1;
+    alterdbReq.daysToKeep1 = -1;
+    alterdbReq.daysToKeep2 = -1;
+    alterdbReq.fsyncPeriod = 4000;
+    alterdbReq.walLevel = 2;
+    alterdbReq.strict = 1;
+    alterdbReq.cacheLastRow = 1;
+    alterdbReq.replications = 1;
+
+    int32_t contLen = tSerializeSAlterDbReq(NULL, 0, &alterdbReq);
+    void*   pReq = rpcMallocCont(contLen);
+    tSerializeSAlterDbReq(pReq, contLen, &alterdbReq);
+
+    SRpcMsg* pRsp = test.SendReq(TDMT_MND_ALTER_DB, pReq, contLen);
+    ASSERT_NE(pRsp, nullptr);
+    ASSERT_EQ(pRsp->code, 0);
+  }
+
+#endif
 }

source/dnode/vnode/inc/vnode.h

@@ -126,7 +126,7 @@ int     tqReadHandleSetTbUidList(STqReadHandle *pHandle, const SArray *tbUidList
 int     tqReadHandleAddTbUidList(STqReadHandle *pHandle, const SArray *tbUidList);
 int32_t tqReadHandleSetMsg(STqReadHandle *pHandle, SSubmitReq *pMsg, int64_t ver);
 bool    tqNextDataBlock(STqReadHandle *pHandle);
-int32_t tqRetrieveDataBlock(SArray **ppCols, STqReadHandle *pHandle, uint64_t *pGroupId, int32_t *pNumOfRows,
+int32_t tqRetrieveDataBlock(SArray **ppCols, STqReadHandle *pHandle, uint64_t *pGroupId, uint64_t* pUid, int32_t *pNumOfRows,
                             int16_t *pNumOfCols);
 
 // need to reposition

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

@@ -273,7 +273,8 @@ typedef enum {
 
 typedef struct {
   uint8_t  last : 1;
-  uint8_t  blkVer : 7;
+  uint8_t  hasDupKey : 1;  // 0: no dup TS key, 1: has dup TS key(since supporting Multi-Version)
+  uint8_t  blkVer : 6;
   uint8_t  numOfSubBlocks;
   col_id_t numOfCols;    // not including timestamp column
   uint32_t len;          // data block length

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

@@ -161,7 +161,7 @@ int32_t tqPushMsgNew(STQ* pTq, void* msg, int32_t msgLen, tmsg_t msgType, int64_
       tqReadHandleSetMsg(pReader, pReq, 0);
       while (tqNextDataBlock(pReader)) {
         SSDataBlock block = {0};
-        if (tqRetrieveDataBlock(&block.pDataBlock, pReader, &block.info.groupId, &block.info.rows,
+        if (tqRetrieveDataBlock(&block.pDataBlock, pReader, &block.info.groupId, &block.info.uid, &block.info.rows,
                                 &block.info.numOfCols) < 0) {
           ASSERT(0);
         }
@@ -540,7 +540,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) {
         tqReadHandleSetMsg(pReader, pCont, 0);
         while (tqNextDataBlock(pReader)) {
           SSDataBlock block = {0};
-          if (tqRetrieveDataBlock(&block.pDataBlock, pReader, &block.info.groupId, &block.info.rows,
+          if (tqRetrieveDataBlock(&block.pDataBlock, pReader, &block.info.groupId, &block.info.uid, &block.info.rows,
                                   &block.info.numOfCols) < 0) {
             ASSERT(0);
           }

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

@@ -84,10 +84,12 @@ bool tqNextDataBlock(STqReadHandle* pHandle) {
   return false;
 }
 
-int32_t tqRetrieveDataBlock(SArray** ppCols, STqReadHandle* pHandle, uint64_t* pGroupId, int32_t* pNumOfRows,
+int32_t tqRetrieveDataBlock(SArray** ppCols, STqReadHandle* pHandle, uint64_t* pGroupId, uint64_t* pUid, int32_t* pNumOfRows,
                             int16_t* pNumOfCols) {
   /*int32_t         sversion = pHandle->pBlock->sversion;*/
   // TODO set to real sversion
+  *pUid = 0;
+
   int32_t sversion = 0;
   if (pHandle->sver != sversion) {
     pHandle->pSchema = metaGetTbTSchema(pHandle->pVnodeMeta, pHandle->msgIter.uid, sversion);
@@ -169,7 +171,10 @@ int32_t tqRetrieveDataBlock(SArray** ppCols, STqReadHandle* pHandle, uint64_t* p
   tdSTSRowIterInit(&iter, pTschema);
   STSRow* row;
   int32_t curRow = 0;
+
   tInitSubmitBlkIter(&pHandle->msgIter, pHandle->pBlock, &pHandle->blkIter);
+  *pUid = pHandle->pBlock->uid;  // set the uid of table for submit block
+
   while ((row = tGetSubmitBlkNext(&pHandle->blkIter)) != NULL) {
     tdSTSRowIterReset(&iter, row);
     // get all wanted col of that block

source/dnode/vnode/src/tsdb/tsdbCommit.c

@@ -1330,13 +1330,15 @@ static void tsdbLoadAndMergeFromCache(SDataCols *pDataCols, int *iter, SCommitIt
                                       TSKEY maxKey, int maxRows, int8_t update) {
   TSKEY     key1 = INT64_MAX;
   TSKEY     key2 = INT64_MAX;
+  TSKEY     lastKey = TSKEY_INITIAL_VAL;
   STSchema *pSchema = NULL;
 
   ASSERT(maxRows > 0 && dataColsKeyLast(pDataCols) <= maxKey);
   tdResetDataCols(pTarget);
 
   pTarget->bitmapMode = pDataCols->bitmapMode;
-
+  // TODO: filter Multi-Version
+  // TODO: support delete function
   while (true) {
     key1 = (*iter >= pDataCols->numOfRows) ? INT64_MAX : dataColsKeyAt(pDataCols, *iter);
     STSRow *row = tsdbNextIterRow(pCommitIter->pIter);
@@ -1349,6 +1351,9 @@ static void tsdbLoadAndMergeFromCache(SDataCols *pDataCols, int *iter, SCommitIt
     if (key1 == INT64_MAX && key2 == INT64_MAX) break;
 
     if (key1 < key2) {
+      if (lastKey != TSKEY_INITIAL_VAL) {
+        ++pTarget->numOfRows;
+      }
       for (int i = 0; i < pDataCols->numOfCols; ++i) {
         // TODO: dataColAppendVal may fail
         SCellVal sVal = {0};
@@ -1356,10 +1361,10 @@ static void tsdbLoadAndMergeFromCache(SDataCols *pDataCols, int *iter, SCommitIt
           TASSERT(0);
         }
         tdAppendValToDataCol(pTarget->cols + i, sVal.valType, sVal.val, pTarget->numOfRows, pTarget->maxPoints,
-                             pTarget->bitmapMode);
+                             pTarget->bitmapMode, false);
       }
 
-      ++pTarget->numOfRows;
+      lastKey = key1;
       ++(*iter);
     } else if (key1 > key2) {
       if (pSchema == NULL || schemaVersion(pSchema) != TD_ROW_SVER(row)) {
@@ -1367,7 +1372,17 @@ static void tsdbLoadAndMergeFromCache(SDataCols *pDataCols, int *iter, SCommitIt
         ASSERT(pSchema != NULL);
       }
 
-      tdAppendSTSRowToDataCol(row, pSchema, pTarget);
+      if (key2 == lastKey) {
+        if (TD_SUPPORT_UPDATE(update)) {
+          tdAppendSTSRowToDataCol(row, pSchema, pTarget, true);
+        }
+      } else {
+        if (lastKey != TSKEY_INITIAL_VAL) {
+          ++pTarget->numOfRows;
+        }
+        tdAppendSTSRowToDataCol(row, pSchema, pTarget, false);
+        lastKey = key2;
+      }
 
       tSkipListIterNext(pCommitIter->pIter);
     } else {
@@ -1397,6 +1412,12 @@ static void tsdbLoadAndMergeFromCache(SDataCols *pDataCols, int *iter, SCommitIt
       ++(*iter);
       tSkipListIterNext(pCommitIter->pIter);
 #endif
+
+      if(lastKey != key1) {
+        lastKey = key1;
+        ++pTarget->numOfRows;
+      }
+
       // copy disk data
       for (int i = 0; i < pDataCols->numOfCols; ++i) {
         SCellVal sVal = {0};
@@ -1405,7 +1426,7 @@ static void tsdbLoadAndMergeFromCache(SDataCols *pDataCols, int *iter, SCommitIt
         }
         // TODO: tdAppendValToDataCol may fail
         tdAppendValToDataCol(pTarget->cols + i, sVal.valType, sVal.val, pTarget->numOfRows, pTarget->maxPoints,
-                             pTarget->bitmapMode);
+                             pTarget->bitmapMode, false);
       }
 
       if (TD_SUPPORT_UPDATE(update)) {
@@ -1416,26 +1437,17 @@ static void tsdbLoadAndMergeFromCache(SDataCols *pDataCols, int *iter, SCommitIt
         }
 
         // TODO: merge with Multi-Version
-        STSRow *curRow = row;
-
-        ++(*iter);
-        tSkipListIterNext(pCommitIter->pIter);
-        STSRow *nextRow = tsdbNextIterRow(pCommitIter->pIter);
-
-        if (key2 < TD_ROW_KEY(nextRow)) {
-          tdAppendSTSRowToDataCol(row, pSchema, pTarget);
-        } else {
-          tdAppendSTSRowToDataCol(row, pSchema, pTarget);
-        }
-        // TODO: merge with Multi-Version
-      } else {
-        ++pTarget->numOfRows;
-        ++(*iter);
-        tSkipListIterNext(pCommitIter->pIter);
+        tdAppendSTSRowToDataCol(row, pSchema, pTarget, true);
       }
+      ++(*iter);
+      tSkipListIterNext(pCommitIter->pIter);
     }
 
-    if (pTarget->numOfRows >= maxRows) break;
+    if (pTarget->numOfRows >= (maxRows - 1)) break;
+  }
+
+  if (lastKey != TSKEY_INITIAL_VAL) {
+    ++pTarget->numOfRows;
   }
 }
 

source/dnode/vnode/src/tsdb/tsdbMemTable.c

@@ -20,7 +20,7 @@ static void     tsdbFreeTbData(STbData *pTbData);
 static char    *tsdbGetTsTupleKey(const void *data);
 static int      tsdbTbDataComp(const void *arg1, const void *arg2);
 static char    *tsdbTbDataGetUid(const void *arg);
-static int      tsdbAppendTableRowToCols(STable *pTable, SDataCols *pCols, STSchema **ppSchema, STSRow *row);
+static int tsdbAppendTableRowToCols(STable *pTable, SDataCols *pCols, STSchema **ppSchema, STSRow *row, bool merge);
 
 int tsdbMemTableCreate(STsdb *pTsdb, STsdbMemTable **ppMemTable) {
   STsdbMemTable *pMemTable;
@@ -82,14 +82,19 @@ int tsdbLoadDataFromCache(STable *pTable, SSkipListIterator *pIter, TSKEY maxKey
                           TKEY *filterKeys, int nFilterKeys, bool keepDup, SMergeInfo *pMergeInfo) {
   ASSERT(maxRowsToRead > 0 && nFilterKeys >= 0);
   if (pIter == NULL) return 0;
-  STSchema  *pSchema = NULL;
-  TSKEY      rowKey = 0;
-  TSKEY      fKey = 0;
+  STSchema *pSchema = NULL;
+  TSKEY     rowKey = 0;
+  TSKEY     fKey = 0;
+  // only fetch lastKey from mem data as file data not used in this function actually
+  TSKEY      lastKey = TSKEY_INITIAL_VAL;
   bool       isRowDel = false;
   int        filterIter = 0;
   STSRow    *row = NULL;
   SMergeInfo mInfo;
 
+  // TODO: support Multi-Version(the rows with the same TS keys in memory can't be merged if its version refered by
+  // query handle)
+
   if (pMergeInfo == NULL) pMergeInfo = &mInfo;
 
   memset(pMergeInfo, 0, sizeof(*pMergeInfo));
@@ -111,7 +116,8 @@ int tsdbLoadDataFromCache(STable *pTable, SSkipListIterator *pIter, TSKEY maxKey
   } else {
     fKey = tdGetKey(filterKeys[filterIter]);
   }
-
+  // 1. fkey - no dup since merged up to maxVersion of each query handle by tsdbLoadBlockDataCols
+  // 2. rowKey - would dup since Multi-Version supported
   while (true) {
     if (fKey == INT64_MAX && rowKey == INT64_MAX) break;
 
@@ -125,12 +131,14 @@ int tsdbLoadDataFromCache(STable *pTable, SSkipListIterator *pIter, TSKEY maxKey
       } else {
         fKey = tdGetKey(filterKeys[filterIter]);
       }
+#if 0
     } else if (fKey > rowKey) {
       if (isRowDel) {
         pMergeInfo->rowsDeleteFailed++;
       } else {
         if (pMergeInfo->rowsInserted - pMergeInfo->rowsDeleteSucceed >= maxRowsToRead) break;
         if (pCols && pMergeInfo->nOperations >= pCols->maxPoints) break;
+
         pMergeInfo->rowsInserted++;
         pMergeInfo->nOperations++;
         pMergeInfo->keyFirst = TMIN(pMergeInfo->keyFirst, rowKey);
@@ -185,6 +193,94 @@ int tsdbLoadDataFromCache(STable *pTable, SSkipListIterator *pIter, TSKEY maxKey
         fKey = tdGetKey(filterKeys[filterIter]);
       }
     }
+#endif
+#if 1
+    } else if (fKey > rowKey) {
+      if (isRowDel) {
+        // TODO: support delete function
+        pMergeInfo->rowsDeleteFailed++;
+      } else {
+        if (pMergeInfo->rowsInserted - pMergeInfo->rowsDeleteSucceed >= maxRowsToRead) break;
+        if (pCols && pMergeInfo->nOperations >= pCols->maxPoints) break;
+
+        if (lastKey != rowKey) {
+          pMergeInfo->rowsInserted++;
+          pMergeInfo->nOperations++;
+          pMergeInfo->keyFirst = TMIN(pMergeInfo->keyFirst, rowKey);
+          pMergeInfo->keyLast = TMAX(pMergeInfo->keyLast, rowKey);
+          if (pCols) {
+            if (lastKey != TSKEY_INITIAL_VAL) {
+              ++pCols->numOfRows;
+            }
+            tsdbAppendTableRowToCols(pTable, pCols, &pSchema, row, false);
+          }
+          lastKey = rowKey;
+        } else {
+          if (keepDup) {
+            tsdbAppendTableRowToCols(pTable, pCols, &pSchema, row, true);
+          } else {
+            // discard
+          }
+        }
+      }
+
+      tSkipListIterNext(pIter);
+      row = tsdbNextIterRow(pIter);
+      if (row == NULL || TD_ROW_KEY(row) > maxKey) {
+        rowKey = INT64_MAX;
+        isRowDel = false;
+      } else {
+        rowKey = TD_ROW_KEY(row);
+        isRowDel = TD_ROW_IS_DELETED(row);
+      }
+    } else {           // fkey == rowKey
+      if (isRowDel) {  // TODO: support delete function(How to stands for delete in file? rowVersion = -1?)
+        ASSERT(!keepDup);
+        if (pCols && pMergeInfo->nOperations >= pCols->maxPoints) break;
+        pMergeInfo->rowsDeleteSucceed++;
+        pMergeInfo->nOperations++;
+        tsdbAppendTableRowToCols(pTable, pCols, &pSchema, row, false);
+      } else {
+        if (keepDup) {
+          if (pCols && pMergeInfo->nOperations >= pCols->maxPoints) break;
+          if (lastKey != rowKey) {
+            pMergeInfo->rowsUpdated++;
+            pMergeInfo->nOperations++;
+            pMergeInfo->keyFirst = TMIN(pMergeInfo->keyFirst, rowKey);
+            pMergeInfo->keyLast = TMAX(pMergeInfo->keyLast, rowKey);
+            lastKey = rowKey;
+            ++pCols->numOfRows;
+            tsdbAppendTableRowToCols(pTable, pCols, &pSchema, row, false);
+          } else {
+            tsdbAppendTableRowToCols(pTable, pCols, &pSchema, row, true);
+          }
+        } else {
+          pMergeInfo->keyFirst = TMIN(pMergeInfo->keyFirst, fKey);
+          pMergeInfo->keyLast = TMAX(pMergeInfo->keyLast, fKey);
+        }
+      }
+
+      tSkipListIterNext(pIter);
+      row = tsdbNextIterRow(pIter);
+      if (row == NULL || TD_ROW_KEY(row) > maxKey) {
+        rowKey = INT64_MAX;
+        isRowDel = false;
+      } else {
+        rowKey = TD_ROW_KEY(row);
+        isRowDel = TD_ROW_IS_DELETED(row);
+      }
+
+      filterIter++;
+      if (filterIter >= nFilterKeys) {
+        fKey = INT64_MAX;
+      } else {
+        fKey = tdGetKey(filterKeys[filterIter]);
+      }
+    }
+#endif
+  }
+  if (lastKey != TSKEY_INITIAL_VAL) {
+    ++pCols->numOfRows;
   }
 
   return 0;
@@ -254,9 +350,12 @@ static STbData *tsdbNewTbData(tb_uid_t uid) {
   pTbData->keyMin = TSKEY_MAX;
   pTbData->keyMax = TSKEY_MIN;
   pTbData->nrows = 0;
-
+#if 0
   pTbData->pData = tSkipListCreate(5, TSDB_DATA_TYPE_TIMESTAMP, sizeof(int64_t), tkeyComparFn, SL_DISCARD_DUP_KEY,
                                    tsdbGetTsTupleKey);
+#endif
+  pTbData->pData =
+      tSkipListCreate(5, TSDB_DATA_TYPE_TIMESTAMP, sizeof(int64_t), tkeyComparFn, SL_ALLOW_DUP_KEY, tsdbGetTsTupleKey);
   if (pTbData->pData == NULL) {
     taosMemoryFree(pTbData);
     return NULL;
@@ -291,7 +390,7 @@ static char *tsdbTbDataGetUid(const void *arg) {
   STbData *pTbData = (STbData *)arg;
   return (char *)(&(pTbData->uid));
 }
-static int tsdbAppendTableRowToCols(STable *pTable, SDataCols *pCols, STSchema **ppSchema, STSRow *row) {
+static int tsdbAppendTableRowToCols(STable *pTable, SDataCols *pCols, STSchema **ppSchema, STSRow *row, bool merge) {
   if (pCols) {
     if (*ppSchema == NULL || schemaVersion(*ppSchema) != TD_ROW_SVER(row)) {
       *ppSchema = tsdbGetTableSchemaImpl(pTable, false, false, TD_ROW_SVER(row));
@@ -301,7 +400,7 @@ static int tsdbAppendTableRowToCols(STable *pTable, SDataCols *pCols, STSchema *
       }
     }
 
-    tdAppendSTSRowToDataCol(row, *ppSchema, pCols);
+    tdAppendSTSRowToDataCol(row, *ppSchema, pCols, merge);
   }
 
   return 0;

source/dnode/vnode/src/tsdb/tsdbReadImpl.c

@@ -246,6 +246,12 @@ int tsdbLoadBlockInfo(SReadH *pReadh, void *pTarget) {
   return 0;
 }
 
+static FORCE_INLINE void tsdbSwapDataCols(SDataCols *pDest, SDataCols *pSrc) {
+  SDataCol *pCols = pDest->cols;
+  memcpy(pDest, pSrc, sizeof(SDataCols));
+  pSrc->cols = pCols;
+}
+
 int tsdbLoadBlockData(SReadH *pReadh, SBlock *pBlock, SBlockInfo *pBlkInfo) {
   ASSERT(pBlock->numOfSubBlocks > 0);
   STsdbCfg *pCfg = REPO_CFG(pReadh->pRepo);
@@ -266,17 +272,29 @@ int tsdbLoadBlockData(SReadH *pReadh, SBlock *pBlock, SBlockInfo *pBlkInfo) {
     if (tsdbLoadBlockDataImpl(pReadh, iBlock, pReadh->pDCols[1]) < 0) return -1;
     // TODO: use the real maxVersion to replace the UINT64_MAX to support Multi-Version
     if (tdMergeDataCols(pReadh->pDCols[0], pReadh->pDCols[1], pReadh->pDCols[1]->numOfRows, NULL,
-                        update != TD_ROW_PARTIAL_UPDATE, UINT64_MAX) < 0)
+                        TD_SUPPORT_UPDATE(update), TD_VER_MAX) < 0)
+      return -1;
+  }
+  // if ((pBlock->numOfSubBlocks == 1) && (iBlock->hasDupKey)) { // TODO: use this line
+  if (pBlock->numOfSubBlocks == 1) {
+    tdResetDataCols(pReadh->pDCols[1]);
+    pReadh->pDCols[1]->bitmapMode = pReadh->pDCols[0]->bitmapMode;
+    if (tdMergeDataCols(pReadh->pDCols[1], pReadh->pDCols[0], pReadh->pDCols[0]->numOfRows, NULL,
+                        TD_SUPPORT_UPDATE(update), TD_VER_MAX) < 0) {
       return -1;
+    }
+    tsdbSwapDataCols(pReadh->pDCols[0], pReadh->pDCols[1]);
+    ASSERT(pReadh->pDCols[0]->bitmapMode != 0);
   }
 
-  ASSERT(pReadh->pDCols[0]->numOfRows == pBlock->numOfRows);
+  ASSERT(pReadh->pDCols[0]->numOfRows <= pBlock->numOfRows);
   ASSERT(dataColsKeyFirst(pReadh->pDCols[0]) == pBlock->keyFirst);
   ASSERT(dataColsKeyLast(pReadh->pDCols[0]) == pBlock->keyLast);
 
   return 0;
 }
 
+// TODO: filter by Multi-Version
 int tsdbLoadBlockDataCols(SReadH *pReadh, SBlock *pBlock, SBlockInfo *pBlkInfo, const int16_t *colIds, int numOfColsIds,
                           bool mergeBitmap) {
   ASSERT(pBlock->numOfSubBlocks > 0);
@@ -297,9 +315,21 @@ int tsdbLoadBlockDataCols(SReadH *pReadh, SBlock *pBlock, SBlockInfo *pBlkInfo,
     if (tsdbLoadBlockDataColsImpl(pReadh, iBlock, pReadh->pDCols[1], colIds, numOfColsIds) < 0) return -1;
     // TODO: use the real maxVersion to replace the UINT64_MAX to support Multi-Version
     if (tdMergeDataCols(pReadh->pDCols[0], pReadh->pDCols[1], pReadh->pDCols[1]->numOfRows, NULL,
-                        update != TD_ROW_PARTIAL_UPDATE, UINT64_MAX) < 0)
+                        TD_SUPPORT_UPDATE(update), TD_VER_MAX) < 0)
       return -1;
   }
+  // if ((pBlock->numOfSubBlocks == 1) && (iBlock->hasDupKey)) { // TODO: use this line
+  if (pBlock->numOfSubBlocks == 1) {
+    tdResetDataCols(pReadh->pDCols[1]);
+    pReadh->pDCols[1]->bitmapMode = pReadh->pDCols[0]->bitmapMode;
+    if (tdMergeDataCols(pReadh->pDCols[1], pReadh->pDCols[0], pReadh->pDCols[0]->numOfRows, NULL,
+                        TD_SUPPORT_UPDATE(update), TD_VER_MAX) < 0) {
+      return -1;
+    }
+    tsdbSwapDataCols(pReadh->pDCols[0], pReadh->pDCols[1]);
+    ASSERT(pReadh->pDCols[0]->bitmapMode != 0);
+  }
+
 
   if (mergeBitmap && !tdDataColsIsBitmapI(pReadh->pDCols[0])) {
     for (int i = 0; i < numOfColsIds; ++i) {
@@ -312,7 +342,7 @@ int tsdbLoadBlockDataCols(SReadH *pReadh, SBlock *pBlock, SBlockInfo *pBlkInfo,
     }
   }
 
-  ASSERT(pReadh->pDCols[0]->numOfRows == pBlock->numOfRows);
+  ASSERT(pReadh->pDCols[0]->numOfRows <= pBlock->numOfRows);
   ASSERT(dataColsKeyFirst(pReadh->pDCols[0]) == pBlock->keyFirst);
   ASSERT(dataColsKeyLast(pReadh->pDCols[0]) == pBlock->keyLast);
 
@@ -623,15 +653,15 @@ static int tsdbLoadBlockDataImpl(SReadH *pReadh, SBlock *pBlock, SDataCols *pDat
       }
 
       if (dcol != 0) {
-        ccol++;
+        ++ccol;
       }
-      dcol++;
+      ++dcol;
     } else if (tcolId < pDataCol->colId) {
-      ccol++;
+      ++ccol;
     } else {
       // Set current column as NULL and forward
       dataColReset(pDataCol);
-      dcol++;
+      ++dcol;
     }
   }
 

source/dnode/vnode/src/tsdb/tsdbSma.c

@@ -1635,17 +1635,20 @@ int32_t tsdbCreateTSma(STsdb *pTsdb, char *pMsg) {
   SSmaCfg vCreateSmaReq = {0};
   if (!tDeserializeSVCreateTSmaReq(pMsg, &vCreateSmaReq)) {
     terrno = TSDB_CODE_OUT_OF_MEMORY;
-    tsdbWarn("vgId:%d TDMT_VND_CREATE_SMA received but deserialize failed since %s", REPO_ID(pTsdb), terrstr(terrno));
+    tsdbWarn("vgId:%d tsma create msg received but deserialize failed since %s", REPO_ID(pTsdb), terrstr(terrno));
     return -1;
   }
-  tsdbDebug("vgId:%d TDMT_VND_CREATE_SMA msg received for %s:%" PRIi64, REPO_ID(pTsdb), vCreateSmaReq.tSma.indexName,
-            vCreateSmaReq.tSma.indexUid);
+  
+  tsdbDebug("vgId:%d tsma create msg %s:%" PRIi64 " for table %" PRIi64 " received", REPO_ID(pTsdb),
+            vCreateSmaReq.tSma.indexName, vCreateSmaReq.tSma.indexUid, vCreateSmaReq.tSma.tableUid);
 
   // record current timezone of server side
   vCreateSmaReq.tSma.timezoneInt = tsTimezone;
 
   if (metaCreateTSma(REPO_META(pTsdb), &vCreateSmaReq) < 0) {
     // TODO: handle error
+    tsdbWarn("vgId:%d tsma %s:%" PRIi64 " create failed for table %" PRIi64 " since %s", REPO_ID(pTsdb),
+             vCreateSmaReq.tSma.indexName, vCreateSmaReq.tSma.indexUid, vCreateSmaReq.tSma.tableUid, terrstr(terrno));
     tdDestroyTSma(&vCreateSmaReq.tSma);
     return -1;
   }

source/libs/executor/src/indexoperator.c

@@ -252,8 +252,8 @@ static int32_t sifExecFunction(SFunctionNode *node, SIFCtx *ctx, SIFParam *outpu
   return TSDB_CODE_QRY_INVALID_INPUT;
 }
 static int32_t sifDoIndex(SIFParam *left, SIFParam *right, int8_t operType, SIFParam *output) {
-  SIndexTerm *tm = indexTermCreate(left->suid, DEFAULT, operType, left->colValType, left->colName,
-                                   strlen(left->colName), right->condValue, strlen(right->condValue));
+  SIndexTerm *tm = indexTermCreate(left->suid, DEFAULT, left->colValType, left->colName, strlen(left->colName),
+                                   right->condValue, strlen(right->condValue));
   if (tm == NULL) {
     return TSDB_CODE_QRY_OUT_OF_MEMORY;
   }

source/libs/executor/src/scanoperator.c

@@ -539,10 +539,12 @@ static SSDataBlock* doStreamBlockScan(SOperatorInfo* pOperator) {
 
     while (tqNextDataBlock(pInfo->readerHandle)) {
       SArray*  pCols = NULL;
-      uint64_t groupId;
-      int32_t  numOfRows;
-      int16_t  outputCol;
-      int32_t  code = tqRetrieveDataBlock(&pCols, pInfo->readerHandle, &groupId, &numOfRows, &outputCol);
+      uint64_t groupId = 0;
+      uint64_t uid = 0;
+      int32_t  numOfRows = 0;
+      int16_t  outputCol = 0;
+
+      int32_t  code = tqRetrieveDataBlock(&pCols, pInfo->readerHandle, &groupId, &uid, &numOfRows, &outputCol);
 
       if (code != TSDB_CODE_SUCCESS || numOfRows == 0) {
         pTaskInfo->code = code;
@@ -551,6 +553,7 @@ static SSDataBlock* doStreamBlockScan(SOperatorInfo* pOperator) {
 
       pInfo->pRes->info.groupId = groupId;
       pInfo->pRes->info.rows = numOfRows;
+      pInfo->pRes->info.uid = uid;
 
       int32_t numOfCols = pInfo->pRes->info.numOfCols;
       for (int32_t i = 0; i < numOfCols; ++i) {
@@ -606,10 +609,8 @@ SOperatorInfo* createStreamScanOperatorInfo(void* streamReadHandle, SSDataBlock*
   SStreamBlockScanInfo* pInfo = taosMemoryCalloc(1, sizeof(SStreamBlockScanInfo));
   SOperatorInfo*        pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
   if (pInfo == NULL || pOperator == NULL) {
-    taosMemoryFreeClear(pInfo);
-    taosMemoryFreeClear(pOperator);
     terrno = TSDB_CODE_QRY_OUT_OF_MEMORY;
-    return NULL;
+    goto _error;
   }
 
   int32_t numOfOutput = taosArrayGetSize(pColList);
@@ -626,16 +627,13 @@ SOperatorInfo* createStreamScanOperatorInfo(void* streamReadHandle, SSDataBlock*
   tqReadHandleSetColIdList((STqReadHandle*)streamReadHandle, pColIds);
   int32_t code = tqReadHandleSetTbUidList(streamReadHandle, pTableIdList);
   if (code != 0) {
-    taosMemoryFreeClear(pInfo);
-    taosMemoryFreeClear(pOperator);
-    return NULL;
+    goto _error;
   }
 
   pInfo->pBlockLists = taosArrayInit(4, POINTER_BYTES);
   if (pInfo->pBlockLists == NULL) {
-    taosMemoryFreeClear(pInfo);
-    taosMemoryFreeClear(pOperator);
-    return NULL;
+    terrno = TSDB_CODE_OUT_OF_MEMORY;
+    goto _error;
   }
 
   pInfo->readerHandle = streamReadHandle;
@@ -647,7 +645,7 @@ SOperatorInfo* createStreamScanOperatorInfo(void* streamReadHandle, SSDataBlock*
   pOperator->blocking        = false;
   pOperator->status          = OP_NOT_OPENED;
   pOperator->info            = pInfo;
-  pOperator->numOfExprs     = pResBlock->info.numOfCols;
+  pOperator->numOfExprs      = pResBlock->info.numOfCols;
   pOperator->fpSet._openFn   = operatorDummyOpenFn;
   pOperator->fpSet.getNextFn = doStreamBlockScan;
   pOperator->fpSet.closeFn   = operatorDummyCloseFn;
@@ -656,6 +654,11 @@ SOperatorInfo* createStreamScanOperatorInfo(void* streamReadHandle, SSDataBlock*
   pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doStreamBlockScan, NULL, NULL, operatorDummyCloseFn, NULL, NULL, NULL);
 
   return pOperator;
+
+  _error:
+  taosMemoryFreeClear(pInfo);
+  taosMemoryFreeClear(pOperator);
+  return NULL;
 }
 
 static void destroySysScanOperator(void* param, int32_t numOfOutput) {

source/libs/executor/src/sortoperator.c

@@ -67,6 +67,10 @@ SSDataBlock* getSortedBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, i
   ASSERT(taosArrayGetSize(pColMatchInfo) == pDataBlock->info.numOfCols);
 
   SSDataBlock* p = tsortGetSortedDataBlock(pHandle);
+  if (p == NULL) {
+    return NULL;
+  }
+
   blockDataEnsureCapacity(p, capacity);
 
   while (1) {

source/libs/executor/src/tsort.c

@@ -553,7 +553,7 @@ static int32_t createInitialSortedMultiSources(SSortHandle* pHandle) {
       }
     }
 
-    if (pHandle->pDataBlock->info.rows > 0) {
+    if (pHandle->pDataBlock != NULL && pHandle->pDataBlock->info.rows > 0) {
       size_t size = blockDataGetSize(pHandle->pDataBlock);
 
       // Perform the in-memory sort and then flush data in the buffer into disk.

source/libs/index/inc/indexCache.h

@@ -31,6 +31,7 @@ extern "C" {
 typedef struct MemTable {
   T_REF_DECLARE()
   SSkipList* mem;
+  void*      pCache;
 } MemTable;
 typedef struct IndexCache {
   T_REF_DECLARE()
@@ -47,6 +48,7 @@ typedef struct IndexCache {
 } IndexCache;
 
 #define CACHE_VERSION(cache) atomic_load_32(&cache->version)
+
 typedef struct CacheTerm {
   // key
   char*   colVal;

source/libs/index/inc/indexComm.h

@@ -24,7 +24,7 @@ extern char JSON_COLUMN[];
 extern char JSON_VALUE_DELIM;
 
 char* indexPackJsonData(SIndexTerm* itm);
-
+char* indexPackJsonDataPrefix(SIndexTerm* itm, int32_t* skip);
 #ifdef __cplusplus
 }
 #endif

source/libs/index/inc/indexInt.h

@@ -166,7 +166,9 @@ int32_t indexSerialCacheKey(ICacheKey* key, char* buf);
   } while (0)
 
 #define INDEX_TYPE_CONTAIN_EXTERN_TYPE(ty, exTy) (((ty >> 4) & (exTy)) != 0)
-#define INDEX_TYPE_GET_TYPE(ty)                  (ty & 0x0F)
+
+#define INDEX_TYPE_GET_TYPE(ty) (ty & 0x0F)
+
 #define INDEX_TYPE_ADD_EXTERN_TYPE(ty, exTy) \
   do {                                       \
     uint8_t oldTy = ty;                      \

source/libs/index/src/index.c

@@ -244,8 +244,8 @@ int indexMultiTermQueryAdd(SIndexMultiTermQuery* pQuery, SIndexTerm* term, EInde
   return 0;
 }
 
-SIndexTerm* indexTermCreate(int64_t suid, SIndexOperOnColumn oper, int8_t queryType, uint8_t colType,
-                            const char* colName, int32_t nColName, const char* colVal, int32_t nColVal) {
+SIndexTerm* indexTermCreate(int64_t suid, SIndexOperOnColumn oper, uint8_t colType, const char* colName,
+                            int32_t nColName, const char* colVal, int32_t nColVal) {
   SIndexTerm* tm = (SIndexTerm*)taosMemoryCalloc(1, (sizeof(SIndexTerm)));
   if (tm == NULL) {
     return NULL;

source/libs/index/src/indexCache.c

@@ -34,44 +34,96 @@ static char*   indexCacheTermGet(const void* pData);
 
 static MemTable* indexInternalCacheCreate(int8_t type);
 
-static int32_t cacheSearchTerm(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
-static int32_t cacheSearchPrefix(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
-static int32_t cacheSearchSuffix(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
-static int32_t cacheSearchRegex(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
-static int32_t cacheSearchLessThan(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
-static int32_t cacheSearchLessEqual(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
-static int32_t cacheSearchGreaterThan(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
-static int32_t cacheSearchGreaterEqual(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
-static int32_t cacheSearchRange(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchTerm(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchPrefix(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchSuffix(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchRegex(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchLessThan(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchLessEqual(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchGreaterThan(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchGreaterEqual(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchRange(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
 /*comm func of compare, used in (LE/LT/GE/GT compare)*/
-static int32_t cacheSearchCompareFunc(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s,
+static int32_t cacheSearchCompareFunc(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s,
                                       RangeType type);
+static int32_t cacheSearchTerm_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchPrefix_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchSuffix_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchRegex_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchLessThan_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchLessEqual_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchGreaterThan_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchGreaterEqual_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+static int32_t cacheSearchRange_JSON(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s);
+
+static int32_t cacheSearchCompareFunc_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s,
+                                           RangeType type);
 
 typedef enum { MATCH, CONTINUE, BREAK } TExeCond;
 typedef TExeCond (*_cache_range_compare)(void* a, void* b, int8_t type);
 
-static TExeCond tCompareLessThan(void* a, void* b, int8_t type) { return MATCH; }
-static TExeCond tCompareLessEqual(void* a, void* b, int8_t type) { return MATCH; }
-static TExeCond tCompareGreaterThan(void* a, void* b, int8_t type) { return MATCH; }
-static TExeCond tCompareGreaterEqual(void* a, void* b, int8_t type) { return MATCH; }
+static TExeCond tDoCommpare(__compar_fn_t func, int8_t comType, void* a, void* b) {
+  int32_t ret = func(a, b);
+  switch (comType) {
+    case QUERY_LESS_THAN: {
+      if (ret < 0) return MATCH;
+    } break;
+    case QUERY_LESS_EQUAL: {
+      if (ret <= 0) return MATCH;
+      break;
+    }
+    case QUERY_GREATER_THAN: {
+      if (ret > 0) return MATCH;
+      break;
+    }
+    case QUERY_GREATER_EQUAL: {
+      if (ret >= 0) return MATCH;
+    }
+  }
+  return CONTINUE;
+}
+static TExeCond tCompareLessThan(void* a, void* b, int8_t type) {
+  __compar_fn_t func = getComparFunc(type, 0);
+  return tDoCommpare(func, QUERY_LESS_THAN, a, b);
+}
+static TExeCond tCompareLessEqual(void* a, void* b, int8_t type) {
+  __compar_fn_t func = getComparFunc(type, 0);
+  return tDoCommpare(func, QUERY_LESS_EQUAL, a, b);
+}
+static TExeCond tCompareGreaterThan(void* a, void* b, int8_t type) {
+  __compar_fn_t func = getComparFunc(type, 0);
+  return tDoCommpare(func, QUERY_GREATER_THAN, a, b);
+}
+static TExeCond tCompareGreaterEqual(void* a, void* b, int8_t type) {
+  __compar_fn_t func = getComparFunc(type, 0);
+  return tDoCommpare(func, QUERY_GREATER_EQUAL, a, b);
+}
 
 static TExeCond (*rangeCompare[])(void* a, void* b, int8_t type) = {tCompareLessThan, tCompareLessEqual,
                                                                     tCompareGreaterThan, tCompareGreaterEqual};
 
-static int32_t (*cacheSearch[])(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) = {
-    cacheSearchTerm,      cacheSearchPrefix,      cacheSearchSuffix,       cacheSearchRegex, cacheSearchLessThan,
-    cacheSearchLessEqual, cacheSearchGreaterThan, cacheSearchGreaterEqual, cacheSearchRange};
+static int32_t (*cacheSearch[][QUERY_MAX])(void* cache, SIndexTerm* ct, SIdxTempResult* tr, STermValueType* s) = {
+    {cacheSearchTerm, cacheSearchPrefix, cacheSearchSuffix, cacheSearchRegex, cacheSearchLessThan, cacheSearchLessEqual,
+     cacheSearchGreaterThan, cacheSearchGreaterEqual, cacheSearchRange},
+    {cacheSearchTerm_JSON, cacheSearchPrefix_JSON, cacheSearchSuffix_JSON, cacheSearchRegex_JSON,
+     cacheSearchLessThan_JSON, cacheSearchLessEqual_JSON, cacheSearchGreaterThan_JSON, cacheSearchGreaterEqual_JSON,
+     cacheSearchRange_JSON}};
 
 static void doMergeWork(SSchedMsg* msg);
 static bool indexCacheIteratorNext(Iterate* itera);
 
-static int32_t cacheSearchTerm(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
+static int32_t cacheSearchTerm(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
   if (cache == NULL) {
     return 0;
   }
+  MemTable*   mem = cache;
+  IndexCache* pCache = mem->pCache;
 
-  MemTable* mem = cache;
-  char*     key = indexCacheTermGet(ct);
+  CacheTerm* pCt = taosMemoryCalloc(1, sizeof(CacheTerm));
+  pCt->colVal = term->colVal;
+  pCt->version = atomic_load_32(&pCache->version);
+
+  char* key = indexCacheTermGet(pCt);
 
   SSkipListIterator* iter = tSkipListCreateIterFromVal(mem->mem, key, TSDB_DATA_TYPE_BINARY, TSDB_ORDER_ASC);
   while (tSkipListIterNext(iter)) {
@@ -80,7 +132,7 @@ static int32_t cacheSearchTerm(void* cache, CacheTerm* ct, SIdxTempResult* tr, S
       break;
     }
     CacheTerm* c = (CacheTerm*)SL_GET_NODE_DATA(node);
-    if (0 == strcmp(c->colVal, ct->colVal)) {
+    if (0 == strcmp(c->colVal, pCt->colVal)) {
       if (c->operaType == ADD_VALUE) {
         INDEX_MERGE_ADD_DEL(tr->deled, tr->added, c->uid)
         // taosArrayPush(result, &c->uid);
@@ -92,30 +144,39 @@ static int32_t cacheSearchTerm(void* cache, CacheTerm* ct, SIdxTempResult* tr, S
       break;
     }
   }
+
+  taosMemoryFree(pCt);
   tSkipListDestroyIter(iter);
   return 0;
 }
-static int32_t cacheSearchPrefix(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
+static int32_t cacheSearchPrefix(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
   // impl later
   return 0;
 }
-static int32_t cacheSearchSuffix(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
+static int32_t cacheSearchSuffix(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
   // impl later
   return 0;
 }
-static int32_t cacheSearchRegex(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
+static int32_t cacheSearchRegex(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
   // impl later
   return 0;
 }
-static int32_t cacheSearchCompareFunc(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s,
+static int32_t cacheSearchCompareFunc(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s,
                                       RangeType type) {
   if (cache == NULL) {
     return 0;
   }
+
   _cache_range_compare cmpFn = rangeCompare[type];
 
-  MemTable* mem = cache;
-  char*     key = indexCacheTermGet(ct);
+  MemTable*   mem = cache;
+  IndexCache* pCache = mem->pCache;
+
+  CacheTerm* pCt = taosMemoryCalloc(1, sizeof(CacheTerm));
+  pCt->colVal = term->colVal;
+  pCt->version = atomic_load_32(&pCache->version);
+
+  char* key = indexCacheTermGet(pCt);
 
   SSkipListIterator* iter = tSkipListCreateIter(mem->mem);
   while (tSkipListIterNext(iter)) {
@@ -124,7 +185,7 @@ static int32_t cacheSearchCompareFunc(void* cache, CacheTerm* ct, SIdxTempResult
       break;
     }
     CacheTerm* c = (CacheTerm*)SL_GET_NODE_DATA(node);
-    TExeCond   cond = cmpFn(c->colVal, ct->colVal, ct->colType);
+    TExeCond   cond = cmpFn(c->colVal, pCt->colVal, pCt->colType);
     if (cond == MATCH) {
       if (c->operaType == ADD_VALUE) {
         INDEX_MERGE_ADD_DEL(tr->deled, tr->added, c->uid)
@@ -134,26 +195,153 @@ static int32_t cacheSearchCompareFunc(void* cache, CacheTerm* ct, SIdxTempResult
         INDEX_MERGE_ADD_DEL(tr->added, tr->deled, c->uid)
       }
     } else if (cond == CONTINUE) {
+      continue;
     } else if (cond == BREAK) {
       break;
     }
   }
+  taosMemoryFree(pCt);
   tSkipListDestroyIter(iter);
   return TSDB_CODE_SUCCESS;
 }
-static int32_t cacheSearchLessThan(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
-  return cacheSearchCompareFunc(cache, ct, tr, s, LT);
+static int32_t cacheSearchLessThan(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return cacheSearchCompareFunc(cache, term, tr, s, LT);
+}
+static int32_t cacheSearchLessEqual(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return cacheSearchCompareFunc(cache, term, tr, s, LE);
+}
+static int32_t cacheSearchGreaterThan(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return cacheSearchCompareFunc(cache, term, tr, s, GT);
+}
+static int32_t cacheSearchGreaterEqual(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return cacheSearchCompareFunc(cache, term, tr, s, GE);
+}
+
+static int32_t cacheSearchTerm_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  if (cache == NULL) {
+    return 0;
+  }
+  MemTable*   mem = cache;
+  IndexCache* pCache = mem->pCache;
+
+  CacheTerm* pCt = taosMemoryCalloc(1, sizeof(CacheTerm));
+  pCt->colVal = term->colVal;
+  pCt->version = atomic_load_32(&pCache->version);
+
+  char* exBuf = NULL;
+  if (INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
+    exBuf = indexPackJsonData(term);
+    pCt->colVal = exBuf;
+  }
+  char* key = indexCacheTermGet(pCt);
+
+  SSkipListIterator* iter = tSkipListCreateIterFromVal(mem->mem, key, TSDB_DATA_TYPE_BINARY, TSDB_ORDER_ASC);
+  while (tSkipListIterNext(iter)) {
+    SSkipListNode* node = tSkipListIterGet(iter);
+    if (node == NULL) {
+      break;
+    }
+    CacheTerm* c = (CacheTerm*)SL_GET_NODE_DATA(node);
+    if (0 == strcmp(c->colVal, pCt->colVal)) {
+      if (c->operaType == ADD_VALUE) {
+        INDEX_MERGE_ADD_DEL(tr->deled, tr->added, c->uid)
+        // taosArrayPush(result, &c->uid);
+        *s = kTypeValue;
+      } else if (c->operaType == DEL_VALUE) {
+        INDEX_MERGE_ADD_DEL(tr->added, tr->deled, c->uid)
+      }
+    } else {
+      break;
+    }
+  }
+
+  taosMemoryFree(pCt);
+  taosMemoryFree(exBuf);
+  tSkipListDestroyIter(iter);
+  return 0;
+
+  return TSDB_CODE_SUCCESS;
+}
+static int32_t cacheSearchPrefix_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return TSDB_CODE_SUCCESS;
+}
+static int32_t cacheSearchSuffix_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return TSDB_CODE_SUCCESS;
+}
+static int32_t cacheSearchRegex_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return TSDB_CODE_SUCCESS;
+}
+static int32_t cacheSearchLessThan_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return cacheSearchCompareFunc_JSON(cache, term, tr, s, LT);
 }
-static int32_t cacheSearchLessEqual(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
-  return cacheSearchCompareFunc(cache, ct, tr, s, LE);
+static int32_t cacheSearchLessEqual_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return cacheSearchCompareFunc_JSON(cache, term, tr, s, LE);
 }
-static int32_t cacheSearchGreaterThan(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
-  return cacheSearchCompareFunc(cache, ct, tr, s, GT);
+static int32_t cacheSearchGreaterThan_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return cacheSearchCompareFunc_JSON(cache, term, tr, s, GT);
 }
-static int32_t cacheSearchGreaterEqual(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
-  return cacheSearchCompareFunc(cache, ct, tr, s, GE);
+static int32_t cacheSearchGreaterEqual_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return cacheSearchCompareFunc_JSON(cache, term, tr, s, GE);
 }
-static int32_t cacheSearchRange(void* cache, CacheTerm* ct, SIdxTempResult* tr, STermValueType* s) {
+static int32_t cacheSearchRange_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
+  return TSDB_CODE_SUCCESS;
+}
+
+static int32_t cacheSearchCompareFunc_JSON(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s,
+                                           RangeType type) {
+  if (cache == NULL) {
+    return 0;
+  }
+  _cache_range_compare cmpFn = rangeCompare[type];
+
+  MemTable*   mem = cache;
+  IndexCache* pCache = mem->pCache;
+
+  CacheTerm* pCt = taosMemoryCalloc(1, sizeof(CacheTerm));
+  pCt->colVal = term->colVal;
+  pCt->version = atomic_load_32(&pCache->version);
+
+  int8_t dType = INDEX_TYPE_GET_TYPE(term->colType);
+  int    skip = 0;
+  char*  exBuf = NULL;
+
+  if (INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
+    exBuf = indexPackJsonDataPrefix(term, &skip);
+    pCt->colVal = exBuf;
+  }
+  char* key = indexCacheTermGet(pCt);
+
+  SSkipListIterator* iter = tSkipListCreateIterFromVal(mem->mem, key, TSDB_DATA_TYPE_BINARY, TSDB_ORDER_ASC);
+  while (tSkipListIterNext(iter)) {
+    SSkipListNode* node = tSkipListIterGet(iter);
+    if (node == NULL) {
+      break;
+    }
+    CacheTerm* c = (CacheTerm*)SL_GET_NODE_DATA(node);
+
+    TExeCond cond = cmpFn(c->colVal + skip, term->colVal, dType);
+    if (cond == MATCH) {
+      if (c->operaType == ADD_VALUE) {
+        INDEX_MERGE_ADD_DEL(tr->deled, tr->added, c->uid)
+        // taosArrayPush(result, &c->uid);
+        *s = kTypeValue;
+      } else if (c->operaType == DEL_VALUE) {
+        INDEX_MERGE_ADD_DEL(tr->added, tr->deled, c->uid)
+      }
+    } else if (cond == CONTINUE) {
+      continue;
+    } else if (cond == BREAK) {
+      break;
+    }
+  }
+
+  taosMemoryFree(pCt);
+  taosMemoryFree(exBuf);
+  tSkipListDestroyIter(iter);
+
+  return TSDB_CODE_SUCCESS;
+}
+static int32_t cacheSearchRange(void* cache, SIndexTerm* term, SIdxTempResult* tr, STermValueType* s) {
   // impl later
   return 0;
 }
@@ -167,6 +355,7 @@ IndexCache* indexCacheCreate(SIndex* idx, uint64_t suid, const char* colName, in
   };
 
   cache->mem = indexInternalCacheCreate(type);
+  cache->mem->pCache = cache;
   cache->colName = INDEX_TYPE_CONTAIN_EXTERN_TYPE(type, TSDB_DATA_TYPE_JSON) ? tstrdup(JSON_COLUMN) : tstrdup(colName);
   cache->type = type;
   cache->index = idx;
@@ -325,6 +514,7 @@ static void indexCacheMakeRoomForWrite(IndexCache* cache) {
       indexCacheRef(cache);
       cache->imm = cache->mem;
       cache->mem = indexInternalCacheCreate(cache->type);
+      cache->mem->pCache = cache;
       cache->occupiedMem = 0;
       // sched to merge
       // unref cache in bgwork
@@ -379,11 +569,19 @@ int indexCacheDel(void* cache, const char* fieldValue, int32_t fvlen, uint64_t u
   return 0;
 }
 
-static int indexQueryMem(MemTable* mem, CacheTerm* ct, EIndexQueryType qtype, SIdxTempResult* tr, STermValueType* s) {
+static int32_t indexQueryMem(MemTable* mem, SIndexTermQuery* query, SIdxTempResult* tr, STermValueType* s) {
   if (mem == NULL) {
     return 0;
   }
-  return cacheSearch[qtype](mem, ct, tr, s);
+
+  SIndexTerm*     term = query->term;
+  EIndexQueryType qtype = query->qType;
+
+  if (INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
+    return cacheSearch[1][qtype](mem, term, tr, s);
+  } else {
+    return cacheSearch[0][qtype](mem, term, tr, s);
+  }
 }
 int indexCacheSearch(void* cache, SIndexTermQuery* query, SIdxTempResult* result, STermValueType* s) {
   int64_t st = taosGetTimestampUs();
@@ -400,25 +598,24 @@ int indexCacheSearch(void* cache, SIndexTermQuery* query, SIdxTempResult* result
   indexMemRef(imm);
   taosThreadMutexUnlock(&pCache->mtx);
 
-  SIndexTerm*     term = query->term;
-  EIndexQueryType qtype = query->qType;
+  // SIndexTerm*     term = query->term;
+  // EIndexQueryType qtype = query->qType;
 
-  bool  hasJson = INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON);
-  char* p = term->colVal;
-  if (hasJson) {
-    p = indexPackJsonData(term);
-  }
-  CacheTerm ct = {.colVal = p, .version = atomic_load_32(&pCache->version)};
+  // bool  isJson = INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON);
+  // char* p = term->colVal;
+  // if (isJson) {
+  //  p = indexPackJsonData(term);
+  //}
+  // CacheTerm ct = {.colVal = p, .version = atomic_load_32(&pCache->version)};
 
-  int ret = indexQueryMem(mem, &ct, qtype, result, s);
+  int ret = indexQueryMem(mem, query, result, s);
   if (ret == 0 && *s != kTypeDeletion) {
     // continue search in imm
-    ret = indexQueryMem(imm, &ct, qtype, result, s);
-  }
-
-  if (hasJson) {
-    taosMemoryFreeClear(p);
+    ret = indexQueryMem(imm, query, result, s);
   }
+  // if (isJson) {
+  //  taosMemoryFreeClear(p);
+  //}
 
   indexMemUnRef(mem);
   indexMemUnRef(imm);

source/libs/index/src/indexComm.c

@@ -46,3 +46,30 @@ char* indexPackJsonData(SIndexTerm* itm) {
 
   return buf;
 }
+char* indexPackJsonDataPrefix(SIndexTerm* itm, int32_t* skip) {
+  /*
+   * |<-----colname---->|<-----dataType---->|<--------colVal---------->|
+   * |<-----string----->|<-----uint8_t----->|<----depend on dataType-->|
+   */
+  uint8_t ty = INDEX_TYPE_GET_TYPE(itm->colType);
+
+  int32_t sz = itm->nColName + itm->nColVal + sizeof(uint8_t) + sizeof(JSON_VALUE_DELIM) * 2 + 1;
+  char*   buf = (char*)taosMemoryCalloc(1, sz);
+  char*   p = buf;
+
+  memcpy(p, itm->colName, itm->nColName);
+  p += itm->nColName;
+
+  memcpy(p, &JSON_VALUE_DELIM, sizeof(JSON_VALUE_DELIM));
+  p += sizeof(JSON_VALUE_DELIM);
+
+  memcpy(p, &ty, sizeof(ty));
+  p += sizeof(ty);
+
+  memcpy(p, &JSON_VALUE_DELIM, sizeof(JSON_VALUE_DELIM));
+  p += sizeof(JSON_VALUE_DELIM);
+
+  *skip = p - buf;
+
+  return buf;
+}

source/libs/index/src/indexTfile.c

@@ -308,20 +308,20 @@ static int32_t tfSearchRegex(void* reader, SIndexTerm* tem, SIdxTempResult* tr)
 }
 
 static int32_t tfSearchCompareFunc(void* reader, SIndexTerm* tem, SIdxTempResult* tr, RangeType type) {
-  bool     hasJson = INDEX_TYPE_CONTAIN_EXTERN_TYPE(tem->colType, TSDB_DATA_TYPE_JSON);
-  int      ret = 0;
-  char*    p = tem->colVal;
-  uint64_t sz = tem->nColVal;
+  bool  hasJson = INDEX_TYPE_CONTAIN_EXTERN_TYPE(tem->colType, TSDB_DATA_TYPE_JSON);
+  int   ret = 0;
+  char* p = tem->colVal;
+  int   skip = 0;
+
   if (hasJson) {
-    p = indexPackJsonData(tem);
-    sz = strlen(p);
+    p = indexPackJsonDataPrefix(tem, &skip);
   }
   SArray* offsets = taosArrayInit(16, sizeof(uint64_t));
 
   AutomationCtx*    ctx = automCtxCreate((void*)p, AUTOMATION_ALWAYS);
   FstStreamBuilder* sb = fstSearch(((TFileReader*)reader)->fst, ctx);
 
-  FstSlice h = fstSliceCreate((uint8_t*)p, sz);
+  FstSlice h = fstSliceCreate((uint8_t*)p, skip);
   fstStreamBuilderSetRange(sb, &h, type);
   fstSliceDestroy(&h);
 
@@ -606,16 +606,16 @@ static bool tfileIteratorNext(Iterate* iiter) {
 static IterateValue* tifileIterateGetValue(Iterate* iter) { return &iter->val; }
 
 static TFileFstIter* tfileFstIteratorCreate(TFileReader* reader) {
-  TFileFstIter* tIter = taosMemoryCalloc(1, sizeof(TFileFstIter));
-  if (tIter == NULL) {
+  TFileFstIter* iter = taosMemoryCalloc(1, sizeof(TFileFstIter));
+  if (iter == NULL) {
     return NULL;
   }
 
-  tIter->ctx = automCtxCreate(NULL, AUTOMATION_ALWAYS);
-  tIter->fb = fstSearch(reader->fst, tIter->ctx);
-  tIter->st = streamBuilderIntoStream(tIter->fb);
-  tIter->rdr = reader;
-  return tIter;
+  iter->ctx = automCtxCreate(NULL, AUTOMATION_ALWAYS);
+  iter->fb = fstSearch(reader->fst, iter->ctx);
+  iter->st = streamBuilderIntoStream(iter->fb);
+  iter->rdr = reader;
+  return iter;
 }
 
 Iterate* tfileIteratorCreate(TFileReader* reader) {

source/libs/index/test/indexTests.cc

@@ -483,7 +483,7 @@ TEST_F(IndexTFileEnv, test_tfile_write) {
 
   std::string     colName("voltage");
   std::string     colVal("ab");
-  SIndexTerm*     term = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+  SIndexTerm*     term = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                      colVal.c_str(), colVal.size());
   SIndexTermQuery query = {term, QUERY_TERM};
 
@@ -557,7 +557,7 @@ TEST_F(IndexCacheEnv, cache_test) {
   std::string colName("voltage");
   {
     std::string colVal("v1");
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     coj->Put(term, colId, version++, suid++);
     indexTermDestroy(term);
@@ -565,28 +565,28 @@ TEST_F(IndexCacheEnv, cache_test) {
   }
   {
     std::string colVal("v3");
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     coj->Put(term, colId, version++, suid++);
     indexTermDestroy(term);
   }
   {
     std::string colVal("v2");
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     coj->Put(term, colId, version++, suid++);
     indexTermDestroy(term);
   }
   {
     std::string colVal("v3");
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     coj->Put(term, colId, version++, suid++);
     indexTermDestroy(term);
   }
   {
     std::string colVal("v3");
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     coj->Put(term, colId, version++, suid++);
     indexTermDestroy(term);
@@ -595,14 +595,14 @@ TEST_F(IndexCacheEnv, cache_test) {
   std::cout << "--------first----------" << std::endl;
   {
     std::string colVal("v3");
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     coj->Put(term, othColId, version++, suid++);
     indexTermDestroy(term);
   }
   {
     std::string colVal("v4");
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     coj->Put(term, othColId, version++, suid++);
     indexTermDestroy(term);
@@ -613,7 +613,7 @@ TEST_F(IndexCacheEnv, cache_test) {
     std::string colVal("v4");
     for (size_t i = 0; i < 10; i++) {
       colVal[colVal.size() - 1] = 'a' + i;
-      SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+      SIndexTerm* term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                          colVal.c_str(), colVal.size());
       coj->Put(term, colId, version++, suid++);
       indexTermDestroy(term);
@@ -623,7 +623,7 @@ TEST_F(IndexCacheEnv, cache_test) {
   // begin query
   {
     std::string     colVal("v3");
-    SIndexTerm*     term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*     term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     SIndexTermQuery query = {term, QUERY_TERM};
     SArray*         ret = (SArray*)taosArrayInit(4, sizeof(suid));
@@ -638,7 +638,7 @@ TEST_F(IndexCacheEnv, cache_test) {
   }
   {
     std::string     colVal("v2");
-    SIndexTerm*     term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*     term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     SIndexTermQuery query = {term, QUERY_TERM};
     SArray*         ret = (SArray*)taosArrayInit(4, sizeof(suid));
@@ -670,7 +670,7 @@ class IndexObj {
     return ret;
   }
   void Del(const std::string& colName, const std::string& colVal, uint64_t uid) {
-    SIndexTerm*      term = indexTermCreate(0, DEL_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*      term = indexTermCreate(0, DEL_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     SIndexMultiTerm* terms = indexMultiTermCreate();
     indexMultiTermAdd(terms, term);
@@ -679,7 +679,7 @@ class IndexObj {
   }
   int WriteMillonData(const std::string& colName, const std::string& colVal = "Hello world",
                       size_t numOfTable = 100 * 10000) {
-    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     SIndexMultiTerm* terms = indexMultiTermCreate();
     indexMultiTermAdd(terms, term);
@@ -701,7 +701,7 @@ class IndexObj {
         // opt
         tColVal[taosRand() % colValSize] = 'a' + k % 26;
       }
-      SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+      SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                          tColVal.c_str(), tColVal.size());
       SIndexMultiTerm* terms = indexMultiTermCreate();
       indexMultiTermAdd(terms, term);
@@ -737,7 +737,7 @@ class IndexObj {
 
   int SearchOne(const std::string& colName, const std::string& colVal) {
     SIndexMultiTermQuery* mq = indexMultiTermQueryCreate(MUST);
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*           term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     indexMultiTermQueryAdd(mq, term, QUERY_TERM);
 
@@ -759,7 +759,7 @@ class IndexObj {
   }
   int SearchOneTarget(const std::string& colName, const std::string& colVal, uint64_t val) {
     SIndexMultiTermQuery* mq = indexMultiTermQueryCreate(MUST);
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*           term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     indexMultiTermQueryAdd(mq, term, QUERY_TERM);
 
@@ -784,7 +784,7 @@ class IndexObj {
 
   void PutOne(const std::string& colName, const std::string& colVal) {
     SIndexMultiTerm* terms = indexMultiTermCreate();
-    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     indexMultiTermAdd(terms, term);
     Put(terms, 10);
@@ -792,7 +792,7 @@ class IndexObj {
   }
   void PutOneTarge(const std::string& colName, const std::string& colVal, uint64_t val) {
     SIndexMultiTerm* terms = indexMultiTermCreate();
-    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     indexMultiTermAdd(terms, term);
     Put(terms, val);
@@ -832,7 +832,7 @@ TEST_F(IndexEnv2, testIndexOpen) {
   {
     std::string colName("tag1"), colVal("Hello");
 
-    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     SIndexMultiTerm* terms = indexMultiTermCreate();
     indexMultiTermAdd(terms, term);
@@ -847,7 +847,7 @@ TEST_F(IndexEnv2, testIndexOpen) {
     size_t      size = 200;
     std::string colName("tag1"), colVal("hello");
 
-    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     SIndexMultiTerm* terms = indexMultiTermCreate();
     indexMultiTermAdd(terms, term);
@@ -862,7 +862,7 @@ TEST_F(IndexEnv2, testIndexOpen) {
     size_t      size = 200;
     std::string colName("tag1"), colVal("Hello");
 
-    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*      term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     SIndexMultiTerm* terms = indexMultiTermCreate();
     indexMultiTermAdd(terms, term);
@@ -877,7 +877,7 @@ TEST_F(IndexEnv2, testIndexOpen) {
   {
     std::string           colName("tag1"), colVal("Hello");
     SIndexMultiTermQuery* mq = indexMultiTermQueryCreate(MUST);
-    SIndexTerm* term = indexTermCreate(0, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*           term = indexTermCreate(0, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
     indexMultiTermQueryAdd(mq, term, QUERY_TERM);
 

source/libs/index/test/jsonUT.cc

@@ -40,7 +40,7 @@ TEST_F(JsonEnv, testWrite) {
   {
     std::string colName("test");
     std::string colVal("ab");
-    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
 
     SIndexMultiTerm* terms = indexMultiTermCreate();
@@ -53,7 +53,7 @@ TEST_F(JsonEnv, testWrite) {
   {
     std::string colName("voltage");
     std::string colVal("ab1");
-    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
 
     SIndexMultiTerm* terms = indexMultiTermCreate();
@@ -66,7 +66,7 @@ TEST_F(JsonEnv, testWrite) {
   {
     std::string colName("voltage");
     std::string colVal("123");
-    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
 
     SIndexMultiTerm* terms = indexMultiTermCreate();
@@ -81,7 +81,7 @@ TEST_F(JsonEnv, testWrite) {
     std::string colVal("ab");
 
     SIndexMultiTermQuery* mq = indexMultiTermQueryCreate(MUST);
-    SIndexTerm*           q = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*           q = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                     colVal.c_str(), colVal.size());
 
     SArray* result = taosArrayInit(1, sizeof(uint64_t));
@@ -95,7 +95,7 @@ TEST_F(JsonEnv, testWriteMillonData) {
   {
     std::string colName("test");
     std::string colVal("ab");
-    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
 
     SIndexMultiTerm* terms = indexMultiTermCreate();
@@ -110,7 +110,7 @@ TEST_F(JsonEnv, testWriteMillonData) {
     std::string colVal("abxxxxxxxxxxxx");
     for (int i = 0; i < 1000; i++) {
       colVal[i % colVal.size()] = '0' + i % 128;
-      SIndexTerm* term = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+      SIndexTerm* term = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                          colVal.c_str(), colVal.size());
 
       SIndexMultiTerm* terms = indexMultiTermCreate();
@@ -124,7 +124,7 @@ TEST_F(JsonEnv, testWriteMillonData) {
   {
     std::string colName("voltagefdadfa");
     std::string colVal("abxxxxxxxxxxxx");
-    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm* term = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                        colVal.c_str(), colVal.size());
 
     SIndexMultiTerm* terms = indexMultiTermCreate();
@@ -139,7 +139,7 @@ TEST_F(JsonEnv, testWriteMillonData) {
     std::string colVal("ab");
 
     SIndexMultiTermQuery* mq = indexMultiTermQueryCreate(MUST);
-    SIndexTerm*           q = indexTermCreate(1, ADD_VALUE, 0, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
+    SIndexTerm*           q = indexTermCreate(1, ADD_VALUE, TSDB_DATA_TYPE_BINARY, colName.c_str(), colName.size(),
                                     colVal.c_str(), colVal.size());
 
     SArray* result = taosArrayInit(1, sizeof(uint64_t));

source/libs/transport/src/transSrv.c

@@ -784,6 +784,11 @@ static void uvDestroyConn(uv_handle_t* handle) {
   tDebug("server conn %p destroy", conn);
   // uv_timer_stop(&conn->pTimer);
   transQueueDestroy(&conn->srvMsgs);
+
+  if (conn->regArg.init == 1) {
+    transFreeMsg(conn->regArg.msg.pCont);
+    conn->regArg.init = 0;
+  }
   QUEUE_REMOVE(&conn->queue);
   taosMemoryFree(conn->pTcp);
   taosMemoryFree(conn);

source/libs/wal/src/walWrite.c

@@ -198,12 +198,14 @@ int walRoll(SWal *pWal) {
   if (pWal->pWriteIdxTFile != NULL) {
     code = taosCloseFile(&pWal->pWriteIdxTFile);
     if (code != 0) {
+      terrno = TAOS_SYSTEM_ERROR(errno);
       return -1;
     }
   }
   if (pWal->pWriteLogTFile != NULL) {
     code = taosCloseFile(&pWal->pWriteLogTFile);
     if (code != 0) {
+      terrno = TAOS_SYSTEM_ERROR(errno);
       return -1;
     }
   }
@@ -263,14 +265,19 @@ int64_t walWriteWithSyncInfo(SWal *pWal, int64_t index, tmsg_t msgType, SSyncLog
   if (index == pWal->vers.lastVer + 1) {
     if (taosArrayGetSize(pWal->fileInfoSet) == 0) {
       pWal->vers.firstVer = index;
-      code = walRoll(pWal);
-      ASSERT(code == 0);
+      if (walRoll(pWal) < 0) {
+        return -1;
+      }
     } else {
       int64_t passed = walGetSeq() - pWal->lastRollSeq;
       if (pWal->cfg.rollPeriod != -1 && pWal->cfg.rollPeriod != 0 && passed > pWal->cfg.rollPeriod) {
-        walRoll(pWal);
+        if (walRoll(pWal) < 0) {
+          return -1;
+        }
       } else if (pWal->cfg.segSize != -1 && pWal->cfg.segSize != 0 && walGetLastFileSize(pWal) > pWal->cfg.segSize) {
-        walRoll(pWal);
+        if (walRoll(pWal) < 0) {
+          return -1;
+        }
       }
     }
   } else {

tests/script/tsim/query/scalarFunction.sim

@@ -102,20 +102,26 @@ print ====> $data60 $data61 $data62 $data63 $data64 $data65
 print ====> $data70 $data71 $data72 $data73 $data74 $data75
 print ====> $data80 $data81 $data82 $data83 $data84 $data85
 print ====> $data90 $data91 $data92 $data93 $data94 $data95
+print ====> rows = $rows and rowNum = $rowNum for ct1
 if $rows != $rowNum then
   return -1
 endi
+
 print ====> select c1, abs(c1), c2, abs(c2), c3, abs(c3) from stb
 sql select c1, abs(c1), c2, abs(c2), c3, abs(c3) from stb
+print ====> rows = $rows and totalRows = $totalRows for stb
 if $rows != $totalRows then
   return -1
 endi
+
 print ====> select c1, abs(c1), c2, abs(c2), c3, abs(c3) from ntb
 sql select c1, abs(c1), c2, abs(c2), c3, abs(c3) from ntb
+print ====> rows = $rows and rowNum = $rowNum for ntb
 if $rows != $rowNum then
   return -1
 endi
 
+
 print ====> log
 sql select c1, log(c1, 10), c2, log(c2, 10), c3, log(c3, 10) from ct1
 print ====> select c1, log(c1, 10), c2, log(c2, 10), c3, log(c3, 10) from ct1

taos-tools @ 2f3dfddd

-Subproject commit 59e0ebaf4905e4cb6d95a01c58b3fa507abc5a20
+Subproject commit 2f3dfddd4d9a869e706ba3cf98fb6d769404cd7c