Merge branch '3.0' of https://github.com/taosdata/TDengine into feature/vnode_refact1

include/common/tdatablock.h

@@ -54,12 +54,34 @@ SEpSet getEpSet_s(SCorEpSet* pEpSet);
     BMCharPos(bm_, r_) |= (1u << (7u - BitPos(r_))); \
   } while (0)
 
+#define colDataIsNull_var(pColumnInfoData, row) (pColumnInfoData->varmeta.offset[row] == -1)
+#define colDataSetNull_var(pColumnInfoData, row)  (pColumnInfoData->varmeta.offset[row] = -1)
+
+#define BitmapLen(_n) (((_n) + ((1 << NBIT) - 1)) >> NBIT)
+
+#define colDataGetVarData(p1_, r_) ((p1_)->pData + (p1_)->varmeta.offset[(r_)])
+
+#define colDataGetNumData(p1_, r_) ((p1_)->pData + ((r_) * (p1_)->info.bytes))
+// SColumnInfoData, rowNumber
+#define colDataGetData(p1_, r_)                                        \
+  ((IS_VAR_DATA_TYPE((p1_)->info.type)) ?  colDataGetVarData(p1_, r_)  \
+                                        :  colDataGetNumData(p1_, r_))
+
 static FORCE_INLINE bool colDataIsNull_s(const SColumnInfoData* pColumnInfoData, uint32_t row) {
+  if (pColumnInfoData->info.type == TSDB_DATA_TYPE_JSON){
+    if(colDataIsNull_var(pColumnInfoData, row)){
+      return true;
+    }
+    char *data = colDataGetVarData(pColumnInfoData, row);
+    return (*data == TSDB_DATA_TYPE_NULL);
+  } 
+  
   if (!pColumnInfoData->hasNull) {
     return false;
   }
-  if (IS_VAR_DATA_TYPE(pColumnInfoData->info.type)) {
-    return pColumnInfoData->varmeta.offset[row] == -1;
+
+  if (pColumnInfoData->info.type== TSDB_DATA_TYPE_VARCHAR || pColumnInfoData->info.type == TSDB_DATA_TYPE_NCHAR) {
+    return colDataIsNull_var(pColumnInfoData, row);
   } else {
     if (pColumnInfoData->nullbitmap == NULL) {
       return false;
@@ -86,7 +108,7 @@ static FORCE_INLINE bool colDataIsNull(const SColumnInfoData* pColumnInfoData, u
   }
 
   if (IS_VAR_DATA_TYPE(pColumnInfoData->info.type)) {
-    return pColumnInfoData->varmeta.offset[row] == -1;
+    return colDataIsNull_var(pColumnInfoData, row);
   } else {
     if (pColumnInfoData->nullbitmap == NULL) {
       return false;
@@ -96,17 +118,10 @@ static FORCE_INLINE bool colDataIsNull(const SColumnInfoData* pColumnInfoData, u
   }
 }
 
-#define BitmapLen(_n) (((_n) + ((1 << NBIT) - 1)) >> NBIT)
-
-// SColumnInfoData, rowNumber
-#define colDataGetData(p1_, r_)                                                        \
-  ((IS_VAR_DATA_TYPE((p1_)->info.type)) ? ((p1_)->pData + (p1_)->varmeta.offset[(r_)]) \
-                                        : ((p1_)->pData + ((r_) * (p1_)->info.bytes)))
-
 static FORCE_INLINE void colDataAppendNULL(SColumnInfoData* pColumnInfoData, uint32_t currentRow) {
   // There is a placehold for each NULL value of binary or nchar type.
   if (IS_VAR_DATA_TYPE(pColumnInfoData->info.type)) {
-    pColumnInfoData->varmeta.offset[currentRow] = -1;  // it is a null value of VAR type.
+    colDataSetNull_var(pColumnInfoData, currentRow);  // it is a null value of VAR type.
   } else {
     colDataSetNull_f(pColumnInfoData->nullbitmap, currentRow);
   }
@@ -117,7 +132,7 @@ static FORCE_INLINE void colDataAppendNULL(SColumnInfoData* pColumnInfoData, uin
 static FORCE_INLINE void colDataAppendNNULL(SColumnInfoData* pColumnInfoData, uint32_t start, size_t nRows) {
   if (IS_VAR_DATA_TYPE(pColumnInfoData->info.type)) {
     for (int32_t i = start; i < start + nRows; ++i) {
-      pColumnInfoData->varmeta.offset[i] = -1;  // it is a null value of VAR type.
+      colDataSetNull_var(pColumnInfoData,i);  // it is a null value of VAR type.
     }
   } else {
     for (int32_t i = start; i < start + nRows; ++i) {
@@ -265,3 +280,4 @@ static FORCE_INLINE void blockCompressEncode(const SSDataBlock* pBlock, char* da
 #endif
 
 #endif /*_TD_COMMON_EP_H_*/
+

include/common/tdataformat.h

@@ -502,7 +502,7 @@ typedef struct {
 
 #define TD_KV_ROW_HEAD_SIZE (sizeof(uint16_t) + sizeof(int16_t))
 
-#define kvRowLen(r)            (*(TDRowLenT *)(r))
+#define kvRowLen(r)            (*(uint16_t *)(r))
 #define kvRowNCols(r)          (*(int16_t *)POINTER_SHIFT(r, sizeof(uint16_t)))
 #define kvRowSetLen(r, len)    kvRowLen(r) = (len)
 #define kvRowSetNCols(r, n)    kvRowNCols(r) = (n)
@@ -608,7 +608,7 @@ void    tdDestroyKVRowBuilder(SKVRowBuilder *pBuilder);
 void    tdResetKVRowBuilder(SKVRowBuilder *pBuilder);
 SKVRow  tdGetKVRowFromBuilder(SKVRowBuilder *pBuilder);
 
-static FORCE_INLINE int32_t tdAddColToKVRow(SKVRowBuilder *pBuilder, col_id_t colId, int8_t type, const void *value) {
+static FORCE_INLINE int32_t tdAddColToKVRow(SKVRowBuilder *pBuilder, col_id_t colId, const void *value, int32_t tlen) {
   if (pBuilder->nCols >= pBuilder->tCols) {
     pBuilder->tCols *= 2;
     SColIdx *pColIdx = (SColIdx *)taosMemoryRealloc((void *)(pBuilder->pColIdx), sizeof(SColIdx) * pBuilder->tCols);
@@ -621,7 +621,6 @@ static FORCE_INLINE int32_t tdAddColToKVRow(SKVRowBuilder *pBuilder, col_id_t co
 
   pBuilder->nCols++;
 
-  int32_t tlen = IS_VAR_DATA_TYPE(type) ? varDataTLen(value) : TYPE_BYTES[type];
   if (tlen > pBuilder->alloc - pBuilder->size) {
     while (tlen > pBuilder->alloc - pBuilder->size) {
       pBuilder->alloc *= 2;

include/common/ttypes.h

@@ -49,7 +49,7 @@ typedef struct {
 #define varDataCopy(dst, v)    memcpy((dst), (void *)(v), varDataTLen(v))
 #define varDataLenByData(v)    (*(VarDataLenT *)(((char *)(v)) - VARSTR_HEADER_SIZE))
 #define varDataSetLen(v, _len) (((VarDataLenT *)(v))[0] = (VarDataLenT)(_len))
-#define IS_VAR_DATA_TYPE(t)    (((t) == TSDB_DATA_TYPE_VARCHAR) || ((t) == TSDB_DATA_TYPE_NCHAR))
+#define IS_VAR_DATA_TYPE(t)    (((t) == TSDB_DATA_TYPE_VARCHAR) || ((t) == TSDB_DATA_TYPE_NCHAR) || ((t) == TSDB_DATA_TYPE_JSON))
 
 #define varDataNetLen(v)  (htons(((VarDataLenT *)(v))[0]))
 #define varDataNetTLen(v) (sizeof(VarDataLenT) + varDataNetLen(v))
@@ -252,7 +252,7 @@ typedef struct tDataTypeDescriptor {
                      int64_t *max, int64_t *sum, int16_t *minindex, int16_t *maxindex, int16_t *numofnull);
 } tDataTypeDescriptor;
 
-extern tDataTypeDescriptor tDataTypes[15];
+extern tDataTypeDescriptor tDataTypes[TSDB_DATA_TYPE_MAX];
 
 bool isValidDataType(int32_t type);
 

include/libs/scalar/scalar.h

@@ -76,6 +76,7 @@ int32_t castFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutp
 /* Time related functions */
 int32_t toISO8601Function(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 int32_t toUnixtimestampFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
+int32_t toJsonFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 int32_t timeTruncateFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 int32_t timeDiffFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);
 int32_t nowFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput);

include/util/taoserror.h

@@ -609,6 +609,7 @@ int32_t* taosGetErrno();
 #define TSDB_CODE_PAR_INTER_SLIDING_UNIT        TAOS_DEF_ERROR_CODE(0, 0x2630)
 #define TSDB_CODE_PAR_INTER_SLIDING_TOO_BIG     TAOS_DEF_ERROR_CODE(0, 0x2631)
 #define TSDB_CODE_PAR_INTER_SLIDING_TOO_SMALL   TAOS_DEF_ERROR_CODE(0, 0x2632)
+#define TSDB_CODE_PAR_ONLY_ONE_JSON_TAG         TAOS_DEF_ERROR_CODE(0, 0x2633)
 
 //planner
 #define TSDB_CODE_PLAN_INTERNAL_ERROR           TAOS_DEF_ERROR_CODE(0, 0x2700)

include/util/tcompare.h

@@ -105,6 +105,8 @@ int32_t compareStrPatternNotMatch(const void *pLeft, const void *pRight);
 int32_t compareWStrPatternMatch(const void *pLeft, const void *pRight);
 int32_t compareWStrPatternNotMatch(const void *pLeft, const void *pRight);
 
+int32_t compareJsonContainsKey(const void *pLeft, const void *pRight);
+
 __compar_fn_t getComparFunc(int32_t type, int32_t optr);
 __compar_fn_t getKeyComparFunc(int32_t keyType, int32_t order);
 int32_t       doCompare(const char *a, const char *b, int32_t type, size_t size);

include/util/tdef.h

@@ -273,6 +273,8 @@ typedef enum ELogicConditionType {
 #define TSDB_MAX_TAGS           128
 #define TSDB_MAX_TAG_CONDITIONS 1024
 
+#define TSDB_MAX_JSON_TAG_LEN   16384
+
 #define TSDB_AUTH_LEN          16
 #define TSDB_PASSWORD_LEN      32
 #define TSDB_USET_PASSWORD_LEN 129

include/util/tjson.h

@@ -76,6 +76,7 @@ char* tjsonToString(const SJson* pJson);
 char* tjsonToUnformattedString(const SJson* pJson);
 
 SJson* tjsonParse(const char* pStr);
+bool   tjsonValidateJson(const char* pJson);
 
 #ifdef __cplusplus
 }

include/util/tutil.h

@@ -26,8 +26,6 @@ extern "C" {
 #endif
 
 int32_t strdequote(char *src);
-int32_t strndequote(char *dst, const char *z, int32_t len);
-int32_t strRmquote(char *z, int32_t len);
 size_t  strtrim(char *src);
 char   *strnchr(const char *haystack, char needle, int32_t len, bool skipquote);
 char  **strsplit(char *src, const char *delim, int32_t *num);

source/client/src/clientImpl.c

@@ -725,6 +725,76 @@ static int32_t doConvertUCS4(SReqResultInfo* pResultInfo, int32_t numOfRows, int
       pResultInfo->pCol[i].pData = pResultInfo->convertBuf[i];
       pResultInfo->row[i] = pResultInfo->pCol[i].pData;
     }
+
+    if (type == TSDB_DATA_TYPE_JSON) {
+      char* p = taosMemoryRealloc(pResultInfo->convertBuf[i], colLength[i]);
+      if (p == NULL) {
+        return TSDB_CODE_OUT_OF_MEMORY;
+      }
+
+      pResultInfo->convertBuf[i] = p;
+      int32_t len = 0;
+      SResultColumn* pCol = &pResultInfo->pCol[i];
+      for (int32_t j = 0; j < numOfRows; ++j) {
+        if (pCol->offset[j] != -1) {
+          char* pStart = pCol->offset[j] + pCol->pData;
+
+          int32_t jsonInnerType = *pStart;
+          char *jsonInnerData = pStart + CHAR_BYTES;
+          char dst[TSDB_MAX_JSON_TAG_LEN] = {0};
+          if(jsonInnerType == TSDB_DATA_TYPE_NULL){
+            sprintf(varDataVal(dst), "%s", TSDB_DATA_NULL_STR_L);
+            varDataSetLen(dst, strlen(varDataVal(dst)));
+          }else if(jsonInnerType == TSDB_DATA_TYPE_JSON){
+            int32_t length = taosUcs4ToMbs((TdUcs4 *)varDataVal(jsonInnerData), varDataLen(jsonInnerData), varDataVal(dst));
+
+            if (length <= 0) {
+              tscError("charset:%s to %s. val:%s convert failed.", DEFAULT_UNICODE_ENCODEC, tsCharset, varDataVal(jsonInnerData));
+              length = 0;
+            }
+            varDataSetLen(dst, length);
+          }else if (jsonInnerType == TSDB_DATA_TYPE_NCHAR) {   // value -> "value"
+            *(char*)varDataVal(dst) = '\"';
+            int32_t length = taosUcs4ToMbs((TdUcs4 *)varDataVal(jsonInnerData), varDataLen(jsonInnerData), varDataVal(dst) + CHAR_BYTES);
+            if (length <= 0) {
+              tscError("charset:%s to %s. val:%s convert failed.", DEFAULT_UNICODE_ENCODEC, tsCharset, varDataVal(jsonInnerData));
+              length = 0;
+            }
+            varDataSetLen(dst, length + CHAR_BYTES*2);
+            *(char*)(varDataVal(dst), length + CHAR_BYTES) = '\"';
+          }else if(jsonInnerType == TSDB_DATA_TYPE_DOUBLE){
+            double jsonVd = *(double*)(jsonInnerData);
+            sprintf(varDataVal(dst), "%.9lf", jsonVd);
+            varDataSetLen(dst, strlen(varDataVal(dst)));
+          }else if(jsonInnerType == TSDB_DATA_TYPE_BIGINT){
+            int64_t jsonVd = *(int64_t*)(jsonInnerData);
+            sprintf(varDataVal(dst), "%" PRId64, jsonVd);
+            varDataSetLen(dst, strlen(varDataVal(dst)));
+          }else if(jsonInnerType == TSDB_DATA_TYPE_BOOL){
+            sprintf(varDataVal(dst), "%s", (*((char *)jsonInnerData) == 1) ? "true" : "false");
+            varDataSetLen(dst, strlen(varDataVal(dst)));
+          }else {
+            ASSERT(0);
+          }
+
+          if(len + varDataTLen(dst) > colLength[i]){
+            p = taosMemoryRealloc(pResultInfo->convertBuf[i], len + varDataTLen(dst));
+            if (p == NULL) {
+              return TSDB_CODE_OUT_OF_MEMORY;
+            }
+
+            pResultInfo->convertBuf[i] = p;
+          }
+          p = pResultInfo->convertBuf[i] + len;
+          memcpy(p, dst, varDataTLen(dst));
+          pCol->offset[j] = len;
+          len += varDataTLen(dst);
+        }
+      }
+
+      pResultInfo->pCol[i].pData = pResultInfo->convertBuf[i];
+      pResultInfo->row[i] = pResultInfo->pCol[i].pData;
+    }
   }
 
   return TSDB_CODE_SUCCESS;

source/common/src/tdatablock.c

@@ -113,14 +113,27 @@ int32_t colDataAppend(SColumnInfoData* pColumnInfoData, uint32_t currentRow, con
 
   int32_t type = pColumnInfoData->info.type;
   if (IS_VAR_DATA_TYPE(type)) {
+    int32_t dataLen = varDataTLen(pData);
+    if(type == TSDB_DATA_TYPE_JSON) {
+      if(*pData == TSDB_DATA_TYPE_NULL) {
+        dataLen = 0;
+      }else if(*pData == TSDB_DATA_TYPE_NCHAR) {
+        dataLen = varDataTLen(pData+CHAR_BYTES);
+      }else if(*pData == TSDB_DATA_TYPE_BIGINT || *pData == TSDB_DATA_TYPE_DOUBLE) {
+        dataLen = LONG_BYTES;
+      }else if(*pData == TSDB_DATA_TYPE_BOOL) {
+        dataLen = CHAR_BYTES;
+      }
+      dataLen += CHAR_BYTES;
+    }
     SVarColAttr* pAttr = &pColumnInfoData->varmeta;
-    if (pAttr->allocLen < pAttr->length + varDataTLen(pData)) {
+    if (pAttr->allocLen < pAttr->length + dataLen) {
       uint32_t newSize = pAttr->allocLen;
       if (newSize == 0) {
         newSize = 8;
       }
 
-      while (newSize < pAttr->length + varDataTLen(pData)) {
+      while (newSize < pAttr->length + dataLen) {
         newSize = newSize * 1.5;
       }
 
@@ -136,8 +149,8 @@ int32_t colDataAppend(SColumnInfoData* pColumnInfoData, uint32_t currentRow, con
     uint32_t len = pColumnInfoData->varmeta.length;
     pColumnInfoData->varmeta.offset[currentRow] = len;
 
-    memcpy(pColumnInfoData->pData + len, pData, varDataTLen(pData));
-    pColumnInfoData->varmeta.length += varDataTLen(pData);
+    memcpy(pColumnInfoData->pData + len, pData, dataLen);
+    pColumnInfoData->varmeta.length += dataLen;
   } else {
     memcpy(pColumnInfoData->pData + pColumnInfoData->info.bytes * currentRow, pData, pColumnInfoData->info.bytes);
   }

source/common/src/ttypes.c

@@ -379,7 +379,7 @@ static void getStatics_nchr(int8_t bitmapMode, const void *pBitmap, const void *
   *maxIndex = 0;
 }
 
-tDataTypeDescriptor tDataTypes[15] = {
+tDataTypeDescriptor tDataTypes[TSDB_DATA_TYPE_MAX] = {
     {TSDB_DATA_TYPE_NULL, 6, 1, "NOTYPE", 0, 0, NULL, NULL, NULL},
     {TSDB_DATA_TYPE_BOOL, 4, CHAR_BYTES, "BOOL", false, true, tsCompressBool, tsDecompressBool, getStatics_bool},
     {TSDB_DATA_TYPE_TINYINT, 7, CHAR_BYTES, "TINYINT", INT8_MIN, INT8_MAX, tsCompressTinyint, tsDecompressTinyint,
@@ -402,6 +402,7 @@ tDataTypeDescriptor tDataTypes[15] = {
     {TSDB_DATA_TYPE_UINT, 12, INT_BYTES, "INT UNSIGNED", 0, UINT32_MAX, tsCompressInt, tsDecompressInt, getStatics_u32},
     {TSDB_DATA_TYPE_UBIGINT, 15, LONG_BYTES, "BIGINT UNSIGNED", 0, UINT64_MAX, tsCompressBigint, tsDecompressBigint,
      getStatics_u64},
+    {TSDB_DATA_TYPE_JSON, 4, TSDB_MAX_JSON_TAG_LEN, "JSON", 0, 0, tsCompressString, tsDecompressString, getStatics_nchr},
 };
 
 char tTokenTypeSwitcher[13] = {

source/common/src/tvariant.c

@@ -118,7 +118,7 @@ void taosVariantCreate(SVariant *pVar, const char *z, int32_t n, int32_t type) {
     }
     case TSDB_DATA_TYPE_BINARY: {
       pVar->pz = strndup(z, n);
-      pVar->nLen = strRmquote(pVar->pz, n);
+      //pVar->nLen = strRmquote(pVar->pz, n);
       break;
     }
     case TSDB_DATA_TYPE_TIMESTAMP: {

source/libs/function/src/builtins.c

@@ -939,7 +939,7 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
     .translateFunc = translateToJson,
     .getEnvFunc   = NULL,
     .initFunc     = NULL,
-    .sprocessFunc = NULL,
+    .sprocessFunc = toJsonFunction,
     .finalizeFunc = NULL
   }
 };

source/libs/nodes/src/nodesUtilFuncs.c

@@ -1009,6 +1009,7 @@ bool nodesIsComparisonOp(const SOperatorNode* pOp) {
     case OP_TYPE_NOT_LIKE:
     case OP_TYPE_MATCH:
     case OP_TYPE_NMATCH:
+    case OP_TYPE_JSON_CONTAINS:
     case OP_TYPE_IS_NULL:
     case OP_TYPE_IS_NOT_NULL:
     case OP_TYPE_IS_TRUE:
@@ -1027,7 +1028,6 @@ bool nodesIsComparisonOp(const SOperatorNode* pOp) {
 bool nodesIsJsonOp(const SOperatorNode* pOp) {
   switch (pOp->opType) {
     case OP_TYPE_JSON_GET_VALUE:
-    case OP_TYPE_JSON_CONTAINS:
       return true;
     default:
       break;

source/libs/parser/inc/parUtil.h

@@ -47,6 +47,7 @@ SSchema *getTableTagSchema(const STableMeta* pTableMeta);
 int32_t  getNumOfColumns(const STableMeta* pTableMeta);
 int32_t  getNumOfTags(const STableMeta* pTableMeta);
 STableComInfo getTableInfo(const STableMeta* pTableMeta);
+int parseJsontoTagData(const char* json, SKVRowBuilder* kvRowBuilder, SMsgBuf* errMsg, int16_t startColId);
 
 int32_t trimString(const char* src, int32_t len, char* dst, int32_t dlen);
 

source/libs/parser/src/parAstCreater.c

@@ -48,6 +48,7 @@ void initAstCreateContext(SParseContext* pParseCxt, SAstCreateContext* pCxt) {
 }
 
 static void trimEscape(SToken* pName) {
+  // todo need to deal with `ioo``ii` -> ioo`ii
   if (NULL != pName && pName->n > 1 && '`' == pName->z[0]) {
     pName->z += 1;
     pName->n -= 2;

source/libs/parser/src/parInsert.c

@@ -397,35 +397,15 @@ static FORCE_INLINE int32_t checkAndTrimValue(SToken* pToken, uint32_t type, cha
     return buildSyntaxErrMsg(pMsgBuf, "invalid data or symbol", pToken->z);
   }
 
-  if (IS_NUMERIC_TYPE(type) && pToken->n == 0) {
-    return buildSyntaxErrMsg(pMsgBuf, "invalid numeric data", pToken->z);
-  }
-
   // Remove quotation marks
   if (TK_NK_STRING == pToken->type) {
     if (pToken->n >= TSDB_MAX_BYTES_PER_ROW) {
       return buildSyntaxErrMsg(pMsgBuf, "too long string", pToken->z);
     }
 
-    // delete escape character: \\, \', \"
-    char    delim = pToken->z[0];
-    int32_t cnt = 0;
-    int32_t j = 0;
-    for (uint32_t k = 1; k < pToken->n - 1; ++k) {
-      if (pToken->z[k] == '\\' || (pToken->z[k] == delim && pToken->z[k + 1] == delim)) {
-        tmpTokenBuf[j] = pToken->z[k + 1];
-        cnt++;
-        j++;
-        k++;
-        continue;
-      }
-      tmpTokenBuf[j] = pToken->z[k];
-      j++;
-    }
-
-    tmpTokenBuf[j] = 0;
+    int32_t len = trimString(pToken->z, pToken->n, tmpTokenBuf, TSDB_MAX_BYTES_PER_ROW);
     pToken->z = tmpTokenBuf;
-    pToken->n -= 2 + cnt;
+    pToken->n = len;
   }
 
   return TSDB_CODE_SUCCESS;
@@ -603,6 +583,13 @@ static int32_t parseValueToken(char** end, SToken* pToken, SSchema* pSchema, int
       return func(pMsgBuf, pToken->z, pToken->n, param);
     }
 
+    case TSDB_DATA_TYPE_JSON: {
+      if(pToken->n > (TSDB_MAX_JSON_TAG_LEN - VARSTR_HEADER_SIZE) / TSDB_NCHAR_SIZE){
+        return buildSyntaxErrMsg(pMsgBuf, "json string too long than 4095", pToken->z);
+      }
+      return func(pMsgBuf, pToken->z, pToken->n, param);
+    }
+
     case TSDB_DATA_TYPE_TIMESTAMP: {
       int64_t tmpVal;
       if (parseTime(end, pToken, timePrec, &tmpVal, pMsgBuf) != TSDB_CODE_SUCCESS) {
@@ -723,8 +710,10 @@ static int32_t parseBoundColumns(SInsertParseContext* pCxt, SParsedDataColInfo*
     qsort(pColIdx, pColList->numOfBound, sizeof(SBoundIdxInfo), boundIdxCompar);
   }
 
-  memset(&pColList->boundColumns[pColList->numOfBound], 0,
-         sizeof(col_id_t) * (pColList->numOfCols - pColList->numOfBound));
+  if(pColList->numOfCols > pColList->numOfBound){
+    memset(&pColList->boundColumns[pColList->numOfBound], 0,
+           sizeof(col_id_t) * (pColList->numOfCols - pColList->numOfBound));
+  }
 
   return TSDB_CODE_SUCCESS;
 }
@@ -741,9 +730,18 @@ static int32_t KvRowAppend(SMsgBuf* pMsgBuf, const void* value, int32_t len, voi
   int8_t  type = pa->schema->type;
   int16_t colId = pa->schema->colId;
 
+  if(TSDB_DATA_TYPE_JSON == type){
+    return parseJsontoTagData(value, pa->builder, pMsgBuf, colId);
+  }
+
+  if (value == NULL) {  // it is a null data
+    // tdAppendColValToRow(rb, pa->schema->colId, pa->schema->type, TD_VTYPE_NULL, value, false, pa->toffset, pa->colIdx);
+    return TSDB_CODE_SUCCESS;
+  }
+
   if (TSDB_DATA_TYPE_BINARY == type) {
     STR_WITH_SIZE_TO_VARSTR(pa->buf, value, len);
-    tdAddColToKVRow(pa->builder, colId, type, pa->buf);
+    tdAddColToKVRow(pa->builder, colId, pa->buf, varDataTLen(pa->buf));
   } else if (TSDB_DATA_TYPE_NCHAR == type) {
     // if the converted output len is over than pColumnModel->bytes, return error: 'Argument list too long'
     int32_t output = 0;
@@ -751,13 +749,12 @@ static int32_t KvRowAppend(SMsgBuf* pMsgBuf, const void* value, int32_t len, voi
       char buf[512] = {0};
       snprintf(buf, tListLen(buf), "%s", strerror(errno));
       return buildSyntaxErrMsg(pMsgBuf, buf, value);
-      ;
     }
 
     varDataSetLen(pa->buf, output);
-    tdAddColToKVRow(pa->builder, colId, type, pa->buf);
+    tdAddColToKVRow(pa->builder, colId, pa->buf, varDataTLen(pa->buf));
   } else {
-    tdAddColToKVRow(pa->builder, colId, type, value);
+    tdAddColToKVRow(pa->builder, colId, value, TYPE_BYTES[type]);
   }
 
   return TSDB_CODE_SUCCESS;

source/libs/parser/src/parTranslater.c

@@ -543,14 +543,18 @@ static EDealRes translateOperator(STranslateContext* pCxt, SOperatorNode* pOp) {
       pOp->node.resType.bytes = tDataTypes[TSDB_DATA_TYPE_DOUBLE].bytes;
     }
   } else if (nodesIsComparisonOp(pOp)) {
-    if (TSDB_DATA_TYPE_JSON == ldt.type || TSDB_DATA_TYPE_BLOB == ldt.type || TSDB_DATA_TYPE_JSON == rdt.type ||
+    if (TSDB_DATA_TYPE_BLOB == ldt.type || TSDB_DATA_TYPE_JSON == rdt.type ||
         TSDB_DATA_TYPE_BLOB == rdt.type) {
       return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, ((SExprNode*)(pOp->pRight))->aliasName);
     }
     pOp->node.resType.type = TSDB_DATA_TYPE_BOOL;
     pOp->node.resType.bytes = tDataTypes[TSDB_DATA_TYPE_BOOL].bytes;
-  } else {
-    // todo json operator
+  } else if (nodesIsJsonOp(pOp)){
+    if (TSDB_DATA_TYPE_JSON != ldt.type || TSDB_DATA_TYPE_BINARY != rdt.type) {
+      return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, ((SExprNode*)(pOp->pRight))->aliasName);
+    }
+    pOp->node.resType.type = TSDB_DATA_TYPE_JSON;
+    pOp->node.resType.bytes = tDataTypes[TSDB_DATA_TYPE_JSON].bytes;
   }
   return DEAL_RES_CONTINUE;
 }
@@ -1813,6 +1817,17 @@ static int32_t checkTableSmaOption(STranslateContext* pCxt, SCreateTableStmt* pS
   return TSDB_CODE_SUCCESS;
 }
 
+static int32_t checkTableTags(STranslateContext* pCxt, SCreateTableStmt* pStmt) {
+  SNode* pNode;
+  FOREACH(pNode, pStmt->pTags) {
+    SColumnDefNode* pCol = (SColumnDefNode*)pNode;
+    if(pCol->dataType.type == TSDB_DATA_TYPE_JSON && LIST_LENGTH(pStmt->pTags) > 1){
+      return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_ONLY_ONE_JSON_TAG);
+    }
+  }
+  return TSDB_CODE_SUCCESS;
+}
+
 static int32_t checkTableRollupOption(STranslateContext* pCxt, SNodeList* pFuncs) {
   if (NULL == pFuncs) {
     return TSDB_CODE_SUCCESS;
@@ -1848,6 +1863,9 @@ static int32_t checkCreateTable(STranslateContext* pCxt, SCreateTableStmt* pStmt
   if (TSDB_CODE_SUCCESS == code) {
     code = checkRangeOption(pCxt, "delay", pStmt->pOptions->pDelay, TSDB_MIN_DB_DELAY, TSDB_MAX_DB_DELAY);
   }
+  if (TSDB_CODE_SUCCESS == code) {
+    code = checkTableTags(pCxt, pStmt);
+  }
   return code;
 }
 
@@ -3249,17 +3267,25 @@ static void addCreateTbReqIntoVgroup(int32_t acctId, SHashObj* pVgroupHashmap, c
 
 static int32_t addValToKVRow(STranslateContext* pCxt, SValueNode* pVal, const SSchema* pSchema,
                              SKVRowBuilder* pBuilder) {
+  if(pSchema->type == TSDB_DATA_TYPE_JSON){
+    if(pVal->literal && strlen(pVal->literal) > (TSDB_MAX_JSON_TAG_LEN - VARSTR_HEADER_SIZE) / TSDB_NCHAR_SIZE){
+      return buildSyntaxErrMsg(&pCxt->msgBuf, "json string too long than 4095", pVal->literal);
+    }
+
+    return parseJsontoTagData(pVal->literal, pBuilder, &pCxt->msgBuf, pSchema->colId);
+  }
+
   if (DEAL_RES_ERROR == translateValue(pCxt, pVal)) {
     return pCxt->errCode;
   }
-  SVariant var;
-  valueNodeToVariant(pVal, &var);
-  char    tagVal[TSDB_MAX_TAGS_LEN] = {0};
-  int32_t code = taosVariantDump(&var, tagVal, pSchema->type, true);
-  if (TSDB_CODE_SUCCESS == code) {
-    tdAddColToKVRow(pBuilder, pSchema->colId, pSchema->type, tagVal);
+
+  if(pVal->node.resType.type == TSDB_DATA_TYPE_NULL){
+    // todo
+  }else{
+    tdAddColToKVRow(pBuilder, pSchema->colId, &(pVal->datum.p), IS_VAR_DATA_TYPE(pSchema->type) ? varDataTLen(pVal->datum.p) : TYPE_BYTES[pSchema->type]);
   }
-  return code;
+
+  return TSDB_CODE_SUCCESS;
 }
 
 static int32_t buildKVRowForBindTags(STranslateContext* pCxt, SCreateSubTableClause* pStmt, STableMeta* pSuperTableMeta,

source/libs/parser/src/parUtil.c

@@ -14,6 +14,7 @@
  */
 
 #include "parUtil.h"
+#include "cJSON.h"
 
 static char* getSyntaxErrFormat(int32_t errCode) {
   switch (errCode) {
@@ -115,6 +116,8 @@ static char* getSyntaxErrFormat(int32_t errCode) {
       return "sliding value no larger than the interval value";
     case TSDB_CODE_PAR_INTER_SLIDING_TOO_SMALL:
       return "sliding value can not less than 1% of interval value";
+    case TSDB_CODE_PAR_ONLY_ONE_JSON_TAG:
+      return "Only one tag if there is a json tag";
     case TSDB_CODE_OUT_OF_MEMORY:
       return "Out of memory";
     default:
@@ -215,24 +218,178 @@ STableComInfo getTableInfo(const STableMeta* pTableMeta) {
 }
 
 int32_t trimString(const char* src, int32_t len, char* dst, int32_t dlen) {
-  // delete escape character: \\, \', \"
+  if (len <=0 || dlen <= 0) return 0;
+
   char delim = src[0];
-  int32_t cnt = 0;
   int32_t j = 0;
   for (uint32_t k = 1; k < len - 1; ++k) {
     if (j >= dlen) {
-      break;
+      dst[j - 1] = '\0';
+      return j;
     }
-    if (src[k] == '\\' || (src[k] == delim && src[k + 1] == delim)) {
+    if (src[k] == delim && src[k + 1] == delim) {   // deal with "", ''
       dst[j] = src[k + 1];
-      cnt++;
       j++;
       k++;
       continue;
     }
+
+    if (src[k] == '\\') {   // deal with escape character
+      if(src[k+1] == 'n'){
+        dst[j] = '\n';
+      }else if(src[k+1] == 'r'){
+        dst[j] = '\r';
+      }else if(src[k+1] == 't'){
+        dst[j] = '\t';
+      }else if(src[k+1] == '\\'){
+        dst[j] = '\\';
+      }else if(src[k+1] == '\''){
+        dst[j] = '\'';
+      }else if(src[k+1] == '"'){
+        dst[j] = '"';
+      }else if(src[k+1] == '%' || src[k+1] == '_'){
+        dst[j++] = src[k];
+        dst[j] = src[k+1];
+      }else{
+        dst[j] = src[k+1];
+      }
+      j++;
+      k++;
+      continue;
+    }
+
     dst[j] = src[k];
     j++;
   }
   dst[j] = '\0';
   return j;
 }
+
+static bool isValidateTag(char *input) {
+  if (!input) return false;
+  for (size_t i = 0; i < strlen(input); ++i) {
+    if (isprint(input[i]) == 0) return false;
+  }
+  return true;
+}
+
+int parseJsontoTagData(const char* json, SKVRowBuilder* kvRowBuilder, SMsgBuf* pMsgBuf, int16_t startColId){
+  // set json NULL data
+  uint8_t jsonNULL = TSDB_DATA_TYPE_NULL;
+  int jsonIndex = startColId + 1;
+  if (!json || strcasecmp(json, TSDB_DATA_NULL_STR_L) == 0){
+    tdAddColToKVRow(kvRowBuilder, jsonIndex, &jsonNULL, CHAR_BYTES);
+    return TSDB_CODE_SUCCESS;
+  }
+
+  // set json real data
+  cJSON *root = cJSON_Parse(json);
+  if (root == NULL){
+    return buildSyntaxErrMsg(pMsgBuf, "json parse error", json);
+  }
+
+  int size = cJSON_GetArraySize(root);
+  if(!cJSON_IsObject(root)){
+    return buildSyntaxErrMsg(pMsgBuf, "json error invalide value", json);
+  }
+
+  int retCode = 0;
+  char *tagKV = NULL;
+  SHashObj* keyHash = taosHashInit(8, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, false);
+  for(int i = 0; i < size; i++) {
+    cJSON* item = cJSON_GetArrayItem(root, i);
+    if (!item) {
+      qError("json inner error:%d", i);
+      retCode = buildSyntaxErrMsg(pMsgBuf, "json inner error", json);
+      goto end;
+    }
+
+    char *jsonKey = item->string;
+    if(!isValidateTag(jsonKey)){
+      retCode = buildSyntaxErrMsg(pMsgBuf, "json key not validate", jsonKey);
+      goto end;
+    }
+//    if(strlen(jsonKey) > TSDB_MAX_JSON_KEY_LEN){
+//      tscError("json key too long error");
+//      retCode =  tscSQLSyntaxErrMsg(errMsg, "json key too long, more than 256", NULL);
+//      goto end;
+//    }
+    size_t keyLen = strlen(jsonKey);
+    if(keyLen == 0 || taosHashGet(keyHash, jsonKey, keyLen) != NULL){
+      continue;
+    }
+    // key: keyLen + VARSTR_HEADER_SIZE, value type: CHAR_BYTES, value reserved: LONG_BYTES
+    tagKV = taosMemoryCalloc(keyLen + VARSTR_HEADER_SIZE + CHAR_BYTES + LONG_BYTES, 1);
+    if(!tagKV) {
+      retCode = TSDB_CODE_TSC_OUT_OF_MEMORY;
+      goto end;
+    }
+    strncpy(varDataVal(tagKV), jsonKey, keyLen);
+    varDataSetLen(tagKV, keyLen);
+    if(taosHashGetSize(keyHash) == 0){
+      uint8_t jsonNotNULL = TSDB_DATA_TYPE_JSON;
+      tdAddColToKVRow(kvRowBuilder, jsonIndex++, &jsonNotNULL, CHAR_BYTES);   // add json type
+    }
+    taosHashPut(keyHash, jsonKey, keyLen, &keyLen, CHAR_BYTES);  // add key to hash to remove dumplicate, value is useless
+
+    if(item->type == cJSON_String){     // add json value  format: type|data
+      char *jsonValue = item->valuestring;
+      int32_t valLen = (int32_t)strlen(jsonValue);
+      int32_t totalLen = keyLen + VARSTR_HEADER_SIZE + valLen * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE + CHAR_BYTES;
+      char *tmp = taosMemoryRealloc(tagKV, totalLen);
+      if(!tmp) {
+        retCode = TSDB_CODE_TSC_OUT_OF_MEMORY;
+        goto end;
+      }
+      tagKV = tmp;
+      char* valueType = POINTER_SHIFT(tagKV, keyLen + VARSTR_HEADER_SIZE);
+      char* valueData = POINTER_SHIFT(tagKV, keyLen + VARSTR_HEADER_SIZE + CHAR_BYTES);
+      *valueType = TSDB_DATA_TYPE_NCHAR;
+      if (valLen > 0 && !taosMbsToUcs4(jsonValue, valLen, (TdUcs4*)varDataVal(valueData),
+                                                  (int32_t)(valLen * TSDB_NCHAR_SIZE), &valLen)) {
+        qError("charset:%s to %s. val:%s, errno:%s, convert failed.", DEFAULT_UNICODE_ENCODEC, tsCharset, jsonValue, strerror(errno));
+        retCode = buildSyntaxErrMsg(pMsgBuf, "charset convert json error", jsonValue);
+        goto end;
+      }
+
+      varDataSetLen(valueData, valLen);
+      tdAddColToKVRow(kvRowBuilder, jsonIndex++, tagKV, totalLen);
+    }else if(item->type == cJSON_Number){
+      if(!isfinite(item->valuedouble)){
+        qError("json value is invalidate");
+        retCode =  buildSyntaxErrMsg(pMsgBuf, "json value number is illegal", json);
+        goto end;
+      }
+      char* valueType = POINTER_SHIFT(tagKV, keyLen + VARSTR_HEADER_SIZE);
+      char* valueData = POINTER_SHIFT(tagKV, keyLen + VARSTR_HEADER_SIZE + CHAR_BYTES);
+      *valueType = (item->valuedouble - (int64_t)(item->valuedouble) == 0) ? TSDB_DATA_TYPE_BIGINT : TSDB_DATA_TYPE_DOUBLE;
+      if(*valueType== TSDB_DATA_TYPE_DOUBLE) *((double *)valueData) = item->valuedouble;
+      else if(*valueType == TSDB_DATA_TYPE_BIGINT) *((int64_t *)valueData) = item->valueint;
+      tdAddColToKVRow(kvRowBuilder, jsonIndex++, tagKV, keyLen + VARSTR_HEADER_SIZE + CHAR_BYTES +LONG_BYTES);
+    }else if(item->type == cJSON_True || item->type == cJSON_False){
+      char* valueType = POINTER_SHIFT(tagKV, keyLen + VARSTR_HEADER_SIZE);
+      char* valueData = POINTER_SHIFT(tagKV, keyLen + VARSTR_HEADER_SIZE + CHAR_BYTES);
+      *valueType = TSDB_DATA_TYPE_BOOL;
+      *valueData = (char)(item->valueint);
+      tdAddColToKVRow(kvRowBuilder, jsonIndex++, tagKV, keyLen + VARSTR_HEADER_SIZE + CHAR_BYTES + CHAR_BYTES);
+    }else if(item->type == cJSON_NULL){
+      char* valueType = POINTER_SHIFT(tagKV, keyLen + VARSTR_HEADER_SIZE);
+      *valueType = TSDB_DATA_TYPE_NULL;
+      tdAddColToKVRow(kvRowBuilder, jsonIndex++, tagKV, keyLen + VARSTR_HEADER_SIZE + CHAR_BYTES);
+    }
+    else{
+      retCode = buildSyntaxErrMsg(pMsgBuf, "invalidate json value", json);
+      goto end;
+    }
+  }
+
+  if(taosHashGetSize(keyHash) == 0){  // set json NULL true
+    tdAddColToKVRow(kvRowBuilder, jsonIndex, &jsonNULL, CHAR_BYTES);
+  }
+
+end:
+  taosMemoryFree(tagKV);
+  taosHashCleanup(keyHash);
+  cJSON_Delete(root);
+  return retCode;
+}
\ No newline at end of file

source/libs/scalar/inc/filterInt.h

@@ -308,7 +308,6 @@ struct SFilterInfo {
 #define FILTER_GET_COL_FIELD_DATA(fi, ri) (colDataGetData(((SColumnInfoData *)(fi)->data), (ri)))
 #define FILTER_GET_VAL_FIELD_TYPE(fi) (((SValueNode *)((fi)->desc))->node.resType.type)
 #define FILTER_GET_VAL_FIELD_DATA(fi) ((char *)(fi)->data)
-#define FILTER_GET_JSON_VAL_FIELD_DATA(fi) ((char *)(fi)->desc)
 #define FILTER_GET_TYPE(fl) ((fl) & FLD_TYPE_MAX)
 
 #define FILTER_GROUP_UNIT(i, g, uid) ((i)->units + (g)->unitIdxs[uid])
@@ -321,7 +320,6 @@ struct SFilterInfo {
 #define FILTER_UNIT_COL_SIZE(i, u) FILTER_GET_COL_FIELD_SIZE(FILTER_UNIT_LEFT_FIELD(i, u))
 #define FILTER_UNIT_COL_ID(i, u) FILTER_GET_COL_FIELD_ID(FILTER_UNIT_LEFT_FIELD(i, u))
 #define FILTER_UNIT_VAL_DATA(i, u) FILTER_GET_VAL_FIELD_DATA(FILTER_UNIT_RIGHT_FIELD(i, u))
-#define FILTER_UNIT_JSON_VAL_DATA(i, u) FILTER_GET_JSON_VAL_FIELD_DATA(FILTER_UNIT_RIGHT_FIELD(i, u))
 #define FILTER_UNIT_COL_IDX(u) ((u)->left.idx)
 #define FILTER_UNIT_OPTR(u) ((u)->compare.optr)
 #define FILTER_UNIT_COMP_FUNC(u) ((u)->compare.func)

source/libs/scalar/inc/sclvector.h

@@ -56,6 +56,8 @@ static FORCE_INLINE double getVectorDoubleValue_BOOL(void *src, int32_t index) {
   return (double)*((bool *)src + index);
 }
 
+double getVectorDoubleValue_JSON(void *src, int32_t index);
+
 static FORCE_INLINE _getDoubleValue_fn_t getVectorDoubleValueFn(int32_t srcType) {
   _getDoubleValue_fn_t p = NULL;
   if (srcType == TSDB_DATA_TYPE_TINYINT) {
@@ -80,6 +82,8 @@ static FORCE_INLINE _getDoubleValue_fn_t getVectorDoubleValueFn(int32_t srcType)
     p = getVectorDoubleValue_DOUBLE;
   } else if (srcType == TSDB_DATA_TYPE_TIMESTAMP) {
     p = getVectorDoubleValue_BIGINT;
+  } else if (srcType == TSDB_DATA_TYPE_JSON) {
+    p = getVectorDoubleValue_JSON;
   } else if (srcType == TSDB_DATA_TYPE_BOOL) {
     p = getVectorDoubleValue_BOOL;
   } else {

source/libs/scalar/src/filter.c

@@ -57,28 +57,24 @@ OptrStr gOptrStr[] = {
   {OP_TYPE_IS_NOT_UNKNOWN,                 "not unknown"},
 
   // json operator
-  {OP_TYPE_JSON_GET_VALUE,                 "json get"},
+  {OP_TYPE_JSON_GET_VALUE,                 "->"},
   {OP_TYPE_JSON_CONTAINS,                  "json contains"}
 };
 
 bool filterRangeCompGi (const void *minv, const void *maxv, const void *minr, const void *maxr, __compar_fn_t cfunc) {
   int32_t result = cfunc(maxv, minr);
-  //if (result == TSDB_DATA_JSON_CAN_NOT_COMPARE) return false;
   return result >= 0;
 }
 bool filterRangeCompGe (const void *minv, const void *maxv, const void *minr, const void *maxr, __compar_fn_t cfunc) {
   int32_t result = cfunc(maxv, minr);
-  //if (result == TSDB_DATA_JSON_CAN_NOT_COMPARE) return false;
   return result > 0;
 }
 bool filterRangeCompLi (const void *minv, const void *maxv, const void *minr, const void *maxr, __compar_fn_t cfunc) {
   int32_t result = cfunc(minv, maxr);
-  //if (result == TSDB_DATA_JSON_CAN_NOT_COMPARE) return false;
   return result <= 0;
 }
 bool filterRangeCompLe (const void *minv, const void *maxv, const void *minr, const void *maxr, __compar_fn_t cfunc) {
   int32_t result = cfunc(minv, maxr);
-  //if (result == TSDB_DATA_JSON_CAN_NOT_COMPARE) return false;
   return result < 0;
 }
 bool filterRangeCompii (const void *minv, const void *maxv, const void *minr, const void *maxr, __compar_fn_t cfunc) {
@@ -170,7 +166,7 @@ __compar_fn_t gDataCompare[] = {compareInt32Val, compareInt8Val, compareInt16Val
   compareLenPrefixedWStr, compareUint8Val, compareUint16Val, compareUint32Val, compareUint64Val,
   setChkInBytes1, setChkInBytes2, setChkInBytes4, setChkInBytes8, compareStrRegexCompMatch, 
   compareStrRegexCompNMatch, setChkNotInBytes1, setChkNotInBytes2, setChkNotInBytes4, setChkNotInBytes8,
-  compareChkNotInString, compareStrPatternNotMatch, compareWStrPatternNotMatch
+  compareChkNotInString, compareStrPatternNotMatch, compareWStrPatternNotMatch, compareJsonContainsKey
 };
 
 int8_t filterGetCompFuncIdx(int32_t type, int32_t optr) {
@@ -221,7 +217,12 @@ int8_t filterGetCompFuncIdx(int32_t type, int32_t optr) {
         assert(0);
     }
   }
-  
+
+  if (optr == OP_TYPE_JSON_CONTAINS && type == TSDB_DATA_TYPE_JSON) {
+    return 28;
+  }
+
+
   switch (type) {
     case TSDB_DATA_TYPE_BOOL:
     case TSDB_DATA_TYPE_TINYINT:   comparFn = 1;   break;
@@ -1049,6 +1050,8 @@ int32_t fltAddGroupUnitFromNode(SFilterInfo *info, SNode* tree, SArray *group) {
 
       cell = cell->pNext;
     }
+    colDataDestroy(out.columnData);
+    taosMemoryFree(out.columnData);
   } else {
     filterAddFieldFromNode(info, node->pRight, &right);
     
@@ -1778,9 +1781,9 @@ int32_t fltInitValFieldData(SFilterInfo *info) {
       bytes = (len + 1) * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE;
 
       fi->data = taosMemoryCalloc(1, bytes);
-    } else if (type != TSDB_DATA_TYPE_JSON){
+    } else{
       if (dType->type == TSDB_DATA_TYPE_VALUE_ARRAY) {  //TIME RANGE
-/*      
+/*
         fi->data = taosMemoryCalloc(dType->bytes, tDataTypes[type].bytes);
         for (int32_t a = 0; a < dType->bytes; ++a) {
           int64_t *v = taosArrayGet(var->arr, a);
@@ -1791,31 +1794,29 @@ int32_t fltInitValFieldData(SFilterInfo *info) {
       } else {
         fi->data = taosMemoryCalloc(1, sizeof(int64_t));
       }
-    } else{     // type == TSDB_DATA_TYPE_JSON
-      // fi->data = null;  use fi->desc as data, because json value is variable, so use tVariant (fi->desc)
     }
 
-    if(type != TSDB_DATA_TYPE_JSON) {
-      if (dType->type == type) {
-        assignVal(fi->data, nodesGetValueFromNode(var), dType->bytes, type);
+    if (dType->type == type) {
+      assignVal(fi->data, nodesGetValueFromNode(var), dType->bytes, type);
+    } else {
+      SScalarParam out = {.columnData = taosMemoryCalloc(1, sizeof(SColumnInfoData))};
+      out.columnData->info.type = type;
+      if (IS_VAR_DATA_TYPE(type)) {
+        out.columnData->info.bytes = bytes;
       } else {
-        SScalarParam out = {.columnData = taosMemoryCalloc(1, sizeof(SColumnInfoData))};
-        out.columnData->info.type = type;
-        if (IS_VAR_DATA_TYPE(type)) {
-          out.columnData->info.bytes = bytes;
-        } else {
-          out.columnData->info.bytes = tDataTypes[type].bytes;
-        }
-
-        // todo refactor the convert
-        int32_t code = doConvertDataType(var, &out);
-        if (code != TSDB_CODE_SUCCESS) {
-          qError("convert value to type[%d] failed", type);
-          return TSDB_CODE_TSC_INVALID_OPERATION;
-        }
+        out.columnData->info.bytes = tDataTypes[type].bytes;
+      }
 
-        memcpy(fi->data, out.columnData->pData, out.columnData->info.bytes);
+      // todo refactor the convert
+      int32_t code = doConvertDataType(var, &out);
+      if (code != TSDB_CODE_SUCCESS) {
+        qError("convert value to type[%d] failed", type);
+        return TSDB_CODE_TSC_INVALID_OPERATION;
       }
+
+      memcpy(fi->data, out.columnData->pData, out.columnData->info.bytes);
+      colDataDestroy(out.columnData);
+      taosMemoryFree(out.columnData);
     }
 
     // match/nmatch for nchar type need convert from ucs4 to mbs
@@ -2556,11 +2557,7 @@ int32_t filterGenerateComInfo(SFilterInfo *info) {
     info->cunits[i].colId = FILTER_UNIT_COL_ID(info, unit);
     
     if (unit->right.type == FLD_TYPE_VALUE) {
-      if(FILTER_UNIT_DATA_TYPE(unit) == TSDB_DATA_TYPE_JSON){   // json value is tVariant
-        info->cunits[i].valData = FILTER_UNIT_JSON_VAL_DATA(info, unit);
-      }else{
-        info->cunits[i].valData = FILTER_UNIT_VAL_DATA(info, unit);
-      }
+      info->cunits[i].valData = FILTER_UNIT_VAL_DATA(info, unit);
     } else {
       info->cunits[i].valData = NULL;
     }
@@ -2886,18 +2883,8 @@ static FORCE_INLINE bool filterExecuteImplIsNull(void *pinfo, int32_t numOfRows,
   for (int32_t i = 0; i < numOfRows; ++i) {
     uint32_t uidx = info->groups[0].unitIdxs[0];
     void *colData = colDataGetData((SColumnInfoData *)info->cunits[uidx].colData, i);
-    if(info->cunits[uidx].dataType == TSDB_DATA_TYPE_JSON){
-      if (!colData){  // for json->'key' is null
-        (*p)[i] = 1;
-      }else if( *(char*)colData == TSDB_DATA_TYPE_JSON){  // for json is null
-        colData = POINTER_SHIFT(colData, CHAR_BYTES);
-        (*p)[i] = colDataIsNull((SColumnInfoData *)info->cunits[uidx].colData, 0, i, NULL);
-      }else{
-        (*p)[i] = 0;
-      }
-    }else{
-      (*p)[i] = ((colData == NULL) || colDataIsNull((SColumnInfoData *)info->cunits[uidx].colData, 0, i, NULL));
-    }
+    (*p)[i] = ((colData == NULL) || colDataIsNull((SColumnInfoData *)info->cunits[uidx].colData, 0, i, NULL));
+
     if ((*p)[i] == 0) {
       all = false;
     }    
@@ -2921,19 +2908,7 @@ static FORCE_INLINE bool filterExecuteImplNotNull(void *pinfo, int32_t numOfRows
     uint32_t uidx = info->groups[0].unitIdxs[0];
     void *colData = colDataGetData((SColumnInfoData *)info->cunits[uidx].colData, i);
 
-    if(info->cunits[uidx].dataType == TSDB_DATA_TYPE_JSON){
-      if (!colData) {   // for json->'key' is not null
-        (*p)[i] = 0;
-      }else if( *(char*)colData == TSDB_DATA_TYPE_JSON){   // for json is not null
-        colData = POINTER_SHIFT(colData, CHAR_BYTES);
-        (*p)[i] = !colDataIsNull((SColumnInfoData *)info->cunits[uidx].colData, 0, i, NULL);
-      }else{    // for json->'key' is not null
-        (*p)[i] = 1;
-      }
-    }else {
-      (*p)[i] = ((colData != NULL) && !colDataIsNull((SColumnInfoData *)info->cunits[uidx].colData, 0, i, NULL));
-    }
-
+    (*p)[i] = ((colData != NULL) && !colDataIsNull((SColumnInfoData *)info->cunits[uidx].colData, 0, i, NULL));
     if ((*p)[i] == 0) {
       all = false;
     }
@@ -3566,6 +3541,11 @@ EDealRes fltReviseRewriter(SNode** pNode, void* pContext) {
         return DEAL_RES_CONTINUE;
       }
 
+      if (OP_TYPE_JSON_CONTAINS == node->opType) {
+        stat->scalarMode = true;
+        return DEAL_RES_CONTINUE;
+      }
+
       if (QUERY_NODE_COLUMN != nodeType(node->pLeft)) {
         SNode *t = node->pLeft;
         node->pLeft = node->pRight;

source/libs/scalar/src/scalar.c

@@ -132,6 +132,7 @@ void sclFreeRes(SHashObj *res) {
 void sclFreeParam(SScalarParam *param) {
   if (param->columnData != NULL) {
     colDataDestroy(param->columnData);
+    taosMemoryFree(param->columnData);
   }
 
   if (param->pHashFilter != NULL) {
@@ -605,37 +606,48 @@ EDealRes sclWalkOperator(SNode* pNode, SScalarCtx *ctx) {
 
 EDealRes sclWalkTarget(SNode* pNode, SScalarCtx *ctx) {
   STargetNode *target = (STargetNode *)pNode;
-  
+
   if (target->dataBlockId >= taosArrayGetSize(ctx->pBlockList)) {
     sclError("target tupleId is too big, tupleId:%d, dataBlockNum:%d", target->dataBlockId, (int32_t)taosArrayGetSize(ctx->pBlockList));
     ctx->code = TSDB_CODE_QRY_INVALID_INPUT;
     return DEAL_RES_ERROR;
   }
 
-  SSDataBlock *block = *(SSDataBlock **)taosArrayGet(ctx->pBlockList, target->dataBlockId);
+  int32_t index = -1;
+  for(int32_t i = 0; i < taosArrayGetSize(ctx->pBlockList); ++i) {
+    SSDataBlock* pb = taosArrayGetP(ctx->pBlockList, i);
+    if (pb->info.blockId == target->dataBlockId) {
+      index = i;
+      break;
+    }
+  }
+
+  if (index == -1) {
+    sclError("column tupleId is too big, tupleId:%d, dataBlockNum:%d", target->dataBlockId, (int32_t)taosArrayGetSize(ctx->pBlockList));
+    ctx->code = TSDB_CODE_QRY_INVALID_INPUT;
+    return DEAL_RES_ERROR;
+  }
+
+  SSDataBlock *block = *(SSDataBlock **)taosArrayGet(ctx->pBlockList, index);
+
   if (target->slotId >= taosArrayGetSize(block->pDataBlock)) {
     sclError("target slot not exist, dataBlockId:%d, slotId:%d, dataBlockNum:%d", target->dataBlockId, target->slotId, (int32_t)taosArrayGetSize(block->pDataBlock));
     ctx->code = TSDB_CODE_QRY_INVALID_INPUT;
     return DEAL_RES_ERROR;
   }
 
+  // if block->pDataBlock is not enough, there are problems if target->slotId bigger than the size of block->pDataBlock,
   SColumnInfoData *col = taosArrayGet(block->pDataBlock, target->slotId);
-  
+
   SScalarParam *res = (SScalarParam *)taosHashGet(ctx->pRes, (void *)&target->pExpr, POINTER_BYTES);
   if (NULL == res) {
     sclError("no valid res in hash, node:%p, type:%d", target->pExpr, nodeType(target->pExpr));
     ctx->code = TSDB_CODE_QRY_APP_ERROR;
     return DEAL_RES_ERROR;
   }
-  
-  for (int32_t i = 0; i < res->numOfRows; ++i) {
-    if (colDataIsNull(res->columnData, res->numOfRows, i, NULL)) {
-      colDataAppend(col, i, NULL, true);
-    } else {
-      char *p = colDataGetData(res->columnData, i);
-      colDataAppend(col, i, p, false);
-    }
-  }
+
+  colDataAssign(col, res->columnData, res->numOfRows);
+  block->info.rows = res->numOfRows;
 
   sclFreeParam(res);
   taosHashRemove(ctx->pRes, (void *)&target->pExpr, POINTER_BYTES);

source/libs/scalar/src/sclfunc.c

@@ -4,6 +4,7 @@
 #include "ttime.h"
 #include "sclInt.h"
 #include "sclvector.h"
+#include "tjson.h"
 
 typedef float (*_float_fn)(float);
 typedef double (*_double_fn)(double);
@@ -872,6 +873,59 @@ int32_t toUnixtimestampFunction(SScalarParam *pInput, int32_t inputNum, SScalarP
   return TSDB_CODE_SUCCESS;
 }
 
+int32_t toJsonFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
+  int32_t type = GET_PARAM_TYPE(pInput);
+  if (type != TSDB_DATA_TYPE_BINARY && type != TSDB_DATA_TYPE_NCHAR) {
+    return TSDB_CODE_FAILED;
+  }
+
+  if (inputNum != 1) {
+    return TSDB_CODE_FAILED;
+  }
+
+  char *input = pInput[0].columnData->pData + pInput[0].columnData->varmeta.offset[0];
+  char *tmp = taosMemoryCalloc(pInput[0].columnData->info.bytes + 1, 1);
+  for (int32_t i = 0; i < pInput[0].numOfRows; ++i) {
+    if (colDataIsNull_s(pInput[0].columnData, i)) {
+      colDataAppendNULL(pOutput->columnData, i);
+      continue;
+    }
+
+    if(type == TSDB_DATA_TYPE_NCHAR){
+      if (varDataTLen(input) > TSDB_MAX_JSON_TAG_LEN){
+        colDataAppendNULL(pOutput->columnData, i);
+        continue;
+      }
+      int32_t len  = taosUcs4ToMbs((TdUcs4 *)varDataVal(input), varDataLen(input), tmp);
+      if (len < 0) {
+        colDataAppendNULL(pOutput->columnData, i);
+        continue;
+      }
+      tmp[len] = 0;
+    }else{
+      if (varDataLen(input) > (TSDB_MAX_JSON_TAG_LEN - VARSTR_HEADER_SIZE) / TSDB_NCHAR_SIZE){
+        colDataAppendNULL(pOutput->columnData, i);
+        continue;
+      }
+      memcpy(tmp, varDataVal(input), varDataLen(input));
+      tmp[varDataTLen(input)] = 0;
+    }
+
+    if(!tjsonValidateJson(tmp)){
+      colDataAppendNULL(pOutput->columnData, i);
+      continue;
+    }
+
+    colDataAppend(pOutput->columnData, i, input, false);
+    input += varDataTLen(input);
+  }
+  taosMemoryFree(tmp);
+
+  pOutput->numOfRows = pInput->numOfRows;
+
+  return TSDB_CODE_SUCCESS;
+}
+
 int32_t timeTruncateFunction(SScalarParam *pInput, int32_t inputNum, SScalarParam *pOutput) {
   int32_t type = GET_PARAM_TYPE(&pInput[0]);
   int32_t timePrec = GET_PARAM_PRECISON(&pInput[0]);

source/libs/scalar/test/scalar/CMakeLists.txt

@@ -8,11 +8,12 @@ AUX_SOURCE_DIRECTORY(${CMAKE_CURRENT_SOURCE_DIR} SOURCE_LIST)
 ADD_EXECUTABLE(scalarTest ${SOURCE_LIST})
 TARGET_LINK_LIBRARIES(
         scalarTest
-        PUBLIC os util common gtest qcom function nodes scalar
+        PUBLIC os util common gtest qcom function nodes scalar parser
 )
 
 TARGET_INCLUDE_DIRECTORIES(
         scalarTest
         PUBLIC "${TD_SOURCE_DIR}/include/libs/scalar/"
+        PUBLIC "${TD_SOURCE_DIR}/source/libs/parser/inc"
         PRIVATE "${TD_SOURCE_DIR}/source/libs/scalar/inc"
 )

source/libs/scalar/test/scalar/scalarTests.cpp

@@ -38,6 +38,7 @@
 #include "scalar.h"
 #include "nodes.h"
 #include "tlog.h"
+#include "parUtil.h"
 
 #define _DEBUG_PRINT_ 0
 
@@ -92,6 +93,7 @@ void scltAppendReservedSlot(SArray *pBlockList, int16_t *dataBlockId, int16_t *s
     blockDataEnsureCapacity(res, rows);
 
     *dataBlockId = taosArrayGetSize(pBlockList) - 1;
+    res->info.blockId = *dataBlockId;
     *slotId = 0;
   } else {
     SSDataBlock *res = *(SSDataBlock **)taosArrayGetLast(pBlockList);
@@ -909,6 +911,264 @@ TEST(constantTest, greater_and_lower) {
   nodesDestroyNode(res);
 }
 
+void makeJsonArrow(SSDataBlock **src, SNode **opNode, void *json, char *key){
+  char keyVar[32] = {0};
+  memcpy(varDataVal(keyVar), key, strlen(key));
+  varDataLen(keyVar) = strlen(key);
+
+  SNode *pLeft = NULL, *pRight = NULL;
+  scltMakeValueNode(&pRight, TSDB_DATA_TYPE_BINARY, keyVar);
+  scltMakeColumnNode(&pLeft, src, TSDB_DATA_TYPE_JSON, varDataLen(json), 1, json);
+  scltMakeOpNode(opNode, OP_TYPE_JSON_GET_VALUE, TSDB_DATA_TYPE_JSON, pLeft, pRight);
+}
+
+void makeOperator(SNode **opNode, SArray *blockList, EOperatorType opType, int32_t rightType, void *rightData){
+  int32_t resType = TSDB_DATA_TYPE_NULL;
+  if(opType == OP_TYPE_ADD || opType == OP_TYPE_SUB || opType == OP_TYPE_MULTI ||
+      opType == OP_TYPE_DIV || opType == OP_TYPE_MOD || opType == OP_TYPE_MINUS){
+    resType = TSDB_DATA_TYPE_DOUBLE;
+  }else if(opType == OP_TYPE_BIT_AND || opType == OP_TYPE_BIT_OR){
+    resType = TSDB_DATA_TYPE_BIGINT;
+  }else if(opType == OP_TYPE_GREATER_THAN || opType == OP_TYPE_GREATER_EQUAL ||
+             opType == OP_TYPE_LOWER_THAN || opType == OP_TYPE_LOWER_EQUAL ||
+             opType == OP_TYPE_EQUAL || opType == OP_TYPE_NOT_EQUAL ||
+             opType == OP_TYPE_IS_NULL || opType == OP_TYPE_IS_NOT_NULL || opType == OP_TYPE_IS_TRUE ||
+             opType == OP_TYPE_LIKE || opType == OP_TYPE_NOT_LIKE || opType == OP_TYPE_MATCH ||
+             opType == OP_TYPE_NMATCH){
+    resType = TSDB_DATA_TYPE_BOOL;
+  }
+
+  SNode *right = NULL;
+  scltMakeValueNode(&right, rightType, rightData);
+  scltMakeOpNode(opNode, opType, resType, *opNode, right);
+
+  SColumnInfo colInfo = createColumnInfo(1, resType, tDataTypes[resType].bytes);
+  int16_t dataBlockId = 0, slotId = 0;
+  scltAppendReservedSlot(blockList, &dataBlockId, &slotId, true, 1, &colInfo);
+  scltMakeTargetNode(opNode, dataBlockId, slotId, *opNode);
+}
+
+void makeCalculate(void *json, void *key, int32_t rightType, void *rightData, double exceptValue, EOperatorType opType){
+  SArray *blockList = taosArrayInit(2, POINTER_BYTES);
+  SSDataBlock *src = NULL;
+  SNode *opNode = NULL;
+
+  makeJsonArrow(&src, &opNode, json, (char*)key);
+  taosArrayPush(blockList, &src);
+
+  makeOperator(&opNode, blockList, opType, rightType, rightData);
+
+  int32_t code = scalarCalculate(opNode, blockList, NULL);
+  ASSERT_EQ(code, 0);
+
+  SSDataBlock *res = *(SSDataBlock **)taosArrayGetLast(blockList);
+  ASSERT_EQ(res->info.rows, 1);
+  SColumnInfoData *column = (SColumnInfoData *)taosArrayGetLast(res->pDataBlock);
+
+  if(colDataIsNull_f(column->nullbitmap, 0)){
+    ASSERT_EQ(DBL_MAX, exceptValue);
+    printf("result:NULL\n");
+
+  }else if(opType == OP_TYPE_ADD || opType == OP_TYPE_SUB || opType == OP_TYPE_MULTI || opType == OP_TYPE_DIV ||
+             opType == OP_TYPE_MOD || opType == OP_TYPE_MINUS){
+    double tmp = *((double *)colDataGetData(column, 0));
+    ASSERT_TRUE(tmp == exceptValue);
+    printf("result:%lf\n", tmp);
+  }else if(opType == OP_TYPE_BIT_AND || opType == OP_TYPE_BIT_OR){
+    ASSERT_EQ(*((int64_t *)colDataGetData(column, 0)), exceptValue);
+    printf("result:%ld\n", *((int64_t *)colDataGetData(column, 0)));
+  }else if(opType == OP_TYPE_GREATER_THAN || opType == OP_TYPE_GREATER_EQUAL || opType == OP_TYPE_LOWER_THAN ||
+             opType == OP_TYPE_LOWER_EQUAL || opType == OP_TYPE_EQUAL || opType == OP_TYPE_NOT_EQUAL ||
+             opType == OP_TYPE_IS_NULL || opType == OP_TYPE_IS_NOT_NULL || opType == OP_TYPE_IS_TRUE ||
+             opType == OP_TYPE_LIKE || opType == OP_TYPE_NOT_LIKE || opType == OP_TYPE_MATCH || opType == OP_TYPE_NMATCH){
+    ASSERT_EQ(*((bool *)colDataGetData(column, 0)), exceptValue);
+    printf("result:%d\n", *((bool *)colDataGetData(column, 0)));
+  }
+
+  taosArrayDestroyEx(blockList, scltFreeDataBlock);
+  nodesDestroyNode(opNode);
+}
+
+TEST(columnTest, json_column_arith_op) {
+  scltInitLogFile();
+  char *rightv= "{\"k1\":4,\"k2\":\"hello\",\"k3\":null,\"k4\":true,\"k5\":5.44}";
+
+  SKVRowBuilder kvRowBuilder;
+  tdInitKVRowBuilder(&kvRowBuilder);
+  parseJsontoTagData(rightv, &kvRowBuilder, NULL, 0);
+  SKVRow row = tdGetKVRowFromBuilder(&kvRowBuilder);
+
+  const int32_t len = 8;
+  EOperatorType op[len] = {OP_TYPE_ADD, OP_TYPE_SUB, OP_TYPE_MULTI, OP_TYPE_DIV,
+                         OP_TYPE_MOD, OP_TYPE_MINUS, OP_TYPE_BIT_AND, OP_TYPE_BIT_OR};
+  int32_t input[len] = {1, 8, 2, 2, 3, 0, -4, 9};
+
+  printf("--------------------json int---------------------\n");
+  char *key = "k1";
+  double eRes[len] = {5.0, -4, 8.0, 2.0, 1.0, -4, 4&-4, 4|9};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes[i], op[i]);
+  }
+
+  printf("--------------------json string---------------------\n");
+
+  key = "k2";
+  double eRes1[len] = {1.0, -8, 0, 0, 0, 0, 0, 9};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes1[i], op[i]);
+  }
+
+  printf("---------------------json null--------------------\n");
+
+  key = "k3";
+  double eRes2[len] = {DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes2[i], op[i]);
+  }
+
+  printf("---------------------json bool--------------------\n");
+
+  key = "k4";
+  double eRes3[len] = {2.0, -7, 2, 0.5, 1, -1, 1&-4, 1|9};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes3[i], op[i]);
+  }
+
+  printf("----------------------json double-------------------\n");
+
+  key = "k5";
+  double eRes4[len] = {6.44, -2.56, 10.88, 2.72, 2.44, -5.44, 5&-4, 5|9};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes4[i], op[i]);
+  }
+
+  printf("---------------------json not exist--------------------\n");
+
+  key = "k10";
+  double eRes5[len] = {DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes5[i], op[i]);
+  }
+}
+
+void *prepareNchar(char* rightData){
+  int32_t len = 0;
+  int32_t inputLen = strlen(rightData);
+
+  char* t = (char*)taosMemoryCalloc(1,(inputLen + 1) * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE);
+  taosMbsToUcs4(rightData, inputLen, (TdUcs4*) varDataVal(t), inputLen * TSDB_NCHAR_SIZE, &len);
+  varDataSetLen(t, len);
+  return t;
+}
+
+TEST(columnTest, json_column_logic_op) {
+  scltInitLogFile();
+  char *rightv= "{\"k1\":4,\"k2\":\"hello\",\"k3\":null,\"k4\":true,\"k5\":5.44,\"k6\":\"6.6hello\"}";
+
+  SKVRowBuilder kvRowBuilder;
+  tdInitKVRowBuilder(&kvRowBuilder);
+  parseJsontoTagData(rightv, &kvRowBuilder, NULL, 0);
+  SKVRow row = tdGetKVRowFromBuilder(&kvRowBuilder);
+
+  const int32_t len = 9;
+  const int32_t len1 = 4;
+  EOperatorType op[len+len1] = {OP_TYPE_GREATER_THAN, OP_TYPE_GREATER_EQUAL, OP_TYPE_LOWER_THAN, OP_TYPE_LOWER_EQUAL, OP_TYPE_EQUAL, OP_TYPE_NOT_EQUAL,
+                          OP_TYPE_IS_NULL, OP_TYPE_IS_NOT_NULL, OP_TYPE_IS_TRUE, OP_TYPE_LIKE, OP_TYPE_NOT_LIKE, OP_TYPE_MATCH, OP_TYPE_NMATCH};
+
+  int32_t input[len] = {1, 8, 2, 2, 3, 0, 0, 0, 0};
+  char *inputNchar[len1] = {"hell_", "hel%", "hell", "llll"};
+
+  printf("--------------------json int---------------------\n");
+  char *key = "k1";
+  bool eRes[len+len1] = {true, false, false, false, false, true, false, true, true, false, false, false, false};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes[i], op[i]);
+  }
+  for(int i = len; i < len + len1; i++){
+    void* rightData = prepareNchar(inputNchar[i-len]);
+    makeCalculate(row, key, TSDB_DATA_TYPE_NCHAR, rightData, eRes[i], op[i]);
+    taosMemoryFree(rightData);
+  }
+
+  printf("--------------------json string---------------------\n");
+
+  key = "k2";
+  bool eRes1[len+len1] = {false, false, true, true, false, false, false, true, false, true, false, true, true};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes1[i], op[i]);
+  }
+  for(int i = len; i < len + len1; i++){
+    void* rightData = prepareNchar(inputNchar[i-len]);
+    makeCalculate(row, key, TSDB_DATA_TYPE_NCHAR, rightData, eRes1[i], op[i]);
+    taosMemoryFree(rightData);
+  }
+
+  printf("--------------------json null---------------------\n");
+
+  key = "k3";
+  double eRes2[len+len1] = {DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, true, false, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes2[i], op[i]);
+  }
+  for(int i = len; i < len + len1; i++){
+    void* rightData = prepareNchar(inputNchar[i-len]);
+    makeCalculate(row, key, TSDB_DATA_TYPE_NCHAR, rightData, eRes2[i], op[i]);
+    taosMemoryFree(rightData);
+  }
+
+  printf("--------------------json bool---------------------\n");
+
+  key = "k4";
+  bool eRes3[len+len1] = {false, false, true, true, false, true, false, true, true, false, false, false, false};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes3[i], op[i]);
+  }
+  for(int i = len; i < len + len1; i++){
+    void* rightData = prepareNchar(inputNchar[i-len]);
+    makeCalculate(row, key, TSDB_DATA_TYPE_NCHAR, rightData, eRes3[i], op[i]);
+    taosMemoryFree(rightData);
+  }
+
+  printf("--------------------json double---------------------\n");
+
+  key = "k5";
+  bool eRes4[len+len1] = {true, false, false, false, false, true, false, true, true, false, false, false, false};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes4[i], op[i]);
+  }
+  for(int i = len; i < len + len1; i++){
+    void* rightData = prepareNchar(inputNchar[i-len]);
+    makeCalculate(row, key, TSDB_DATA_TYPE_NCHAR, rightData, eRes4[i], op[i]);
+    taosMemoryFree(rightData);
+  }
+
+  printf("--------------------json double---------------------\n");
+
+  key = "k6";
+  bool eRes5[len+len1] = {true, false, false, false, false, true, false, true, true, false, false, false, true};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes5[i], op[i]);
+  }
+  for(int i = len; i < len + len1; i++){
+    void* rightData = prepareNchar(inputNchar[i-len]);
+    makeCalculate(row, key, TSDB_DATA_TYPE_NCHAR, rightData, eRes5[i], op[i]);
+    taosMemoryFree(rightData);
+  }
+
+  printf("---------------------json not exist--------------------\n");
+
+  key = "k10";
+  double eRes10[len+len1] = {DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, true, false, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX, DBL_MAX};
+  for(int i = 0; i < len; i++){
+    makeCalculate(row, key, TSDB_DATA_TYPE_INT, &input[i], eRes10[i], op[i]);
+  }
+  for(int i = len; i < len + len1; i++){
+    void* rightData = prepareNchar(inputNchar[i-len]);
+    makeCalculate(row, key, TSDB_DATA_TYPE_NCHAR, rightData, eRes10[i], op[i]);
+    taosMemoryFree(rightData);
+  }
+}
+
 TEST(columnTest, smallint_value_add_int_column) {
   scltInitLogFile();
   
@@ -928,7 +1188,7 @@ TEST(columnTest, smallint_value_add_int_column) {
   int16_t dataBlockId = 0, slotId = 0;
   scltAppendReservedSlot(blockList, &dataBlockId, &slotId, true, rowNum, &colInfo);
   scltMakeTargetNode(&opNode, dataBlockId, slotId, opNode);
-  
+
   int32_t code = scalarCalculate(opNode, blockList, NULL);
   ASSERT_EQ(code, 0);
 
@@ -3097,3 +3357,4 @@ int main(int argc, char** argv) {
 }
 
 #pragma GCC diagnostic pop
+

source/util/src/tcompare.c

@@ -222,6 +222,11 @@ int32_t compareLenPrefixedWStrDesc(const void *pLeft, const void *pRight) {
   return compareLenPrefixedWStr(pRight, pLeft);
 }
 
+int32_t compareJsonContainsKey(const void* pLeft, const void* pRight) {
+  if(pLeft) return 0;
+  return 1;
+}
+
 /*
  * Compare two strings
  *    TSDB_MATCH:            Match

source/util/src/tjson.c

@@ -276,3 +276,37 @@ int32_t tjsonToArray(const SJson* pJson, const char* pName, FToObject func, void
 }
 
 SJson* tjsonParse(const char* pStr) { return cJSON_Parse(pStr); }
+
+bool tjsonValidateJson(const char *jIn) {
+  if (!jIn){
+    return false;
+  }
+
+  // set json real data
+  cJSON *root = cJSON_Parse(jIn);
+  if (root == NULL){
+    return false;
+  }
+
+  if(!cJSON_IsObject(root)){
+    return false;
+  }
+  int size = cJSON_GetArraySize(root);
+  for(int i = 0; i < size; i++) {
+    cJSON* item = cJSON_GetArrayItem(root, i);
+    if (!item) {
+      return false;
+    }
+
+    char* jsonKey = item->string;
+    if (!jsonKey) return false;
+    for (size_t j = 0; j < strlen(jsonKey); ++i) {
+      if (isprint(jsonKey[i]) == 0) return false;
+    }
+
+    if (item->type == cJSON_Object || item->type == cJSON_Array) {
+      return false;
+    }
+  }
+  return true;
+}
\ No newline at end of file

source/util/src/tutil.c

@@ -47,65 +47,6 @@ int32_t strdequote(char *z) {
   return j + 1;  // only one quote, do nothing
 }
 
-int32_t strRmquote(char *z, int32_t len) {
-  // delete escape character: \\, \', \"
-  char delim = z[0];
-  if (delim != '\'' && delim != '\"') {
-    return len;
-  }
-
-  int32_t cnt = 0;
-  int32_t j = 0;
-  for (uint32_t k = 1; k < len - 1; ++k) {
-    if (z[k] == '\\' || (z[k] == delim && z[k + 1] == delim)) {
-      if (z[k] == '\\' && z[k + 1] == '_') {
-        // match '_' self
-      } else {
-        z[j] = z[k + 1];
-        cnt++;
-        j++;
-        k++;
-        continue;
-      }
-    }
-
-    z[j] = z[k];
-    j++;
-  }
-
-  z[j] = 0;
-
-  return len - 2 - cnt;
-}
-
-int32_t strndequote(char *dst, const char *z, int32_t len) {
-  assert(dst != NULL);
-  if (z == NULL || len == 0) {
-    return 0;
-  }
-
-  int32_t quote = z[0];
-  int32_t i = 1, j = 0;
-
-  while (z[i] != 0) {
-    if (z[i] == quote) {
-      if (z[i + 1] == quote) {
-        dst[j++] = (char)quote;
-        i++;
-      } else {
-        dst[j++] = 0;
-        return (j - 1);
-      }
-    } else {
-      dst[j++] = z[i];
-    }
-
-    i++;
-  }
-
-  return j + 1;  // only one quote, do nothing
-}
-
 size_t strtrim(char *z) {
   int32_t i = 0;
   int32_t j = 0;

tests/script/jenkins/basic.txt

@@ -93,4 +93,7 @@
 # --- sma
 ./test.sh -f tsim/sma/tsmaCreateInsertData.sim
 
+# --- valgrind
+./test.sh -f tsim/valgrind/checkError.sim -v
+
 #======================b1-end===============

tests/script/sh/checkValgrind.sh

+#!/bin/bash
+
+set +e
+#set -x
+
+NODE_NAME=
+
+while getopts "n:" arg
+do
+  case $arg in
+    n)
+      NODE_NAME=$OPTARG
+      ;;
+    ?)
+      echo "unkown argument"
+      ;;
+  esac
+done
+
+SCRIPT_DIR=`dirname $0`
+cd $SCRIPT_DIR/../
+SCRIPT_DIR=`pwd`
+
+IN_TDINTERNAL="community"
+if [[ "$SCRIPT_DIR" == *"$IN_TDINTERNAL"* ]]; then
+  cd ../../..
+else
+  cd ../../
+fi
+
+TAOS_DIR=`pwd`
+LOG_DIR=$TAOS_DIR/sim/$NODE_NAME/log
+#CFG_DIR=$TAOS_DIR/sim/$NODE_NAME/cfg
+
+#echo ---- $LOG_DIR
+
+errors=`grep "ERROR SUMMARY:" ${LOG_DIR}/valgrind-taosd-*.log | cut -d ' ' -f 2,3,4,5 | tr -d "\n"`
+echo $errors

tests/script/tsim/valgrind/checkError.sim

+system sh/stop_dnodes.sh
+system sh/deploy.sh -n dnode1 -i 1
+#system sh/deploy.sh -n dnode2 -i 2
+#system sh/deploy.sh -n dnode3 -i 3
+#system sh/deploy.sh -n dnode4 -i 4
+#system sh/cfg.sh -n dnode1 -c supportVnodes -v 0
+system sh/exec.sh -n dnode1 -s start
+#system sh/exec.sh -n dnode2 -s start
+#system sh/exec.sh -n dnode3 -s start
+#system sh/exec.sh -n dnode4 -s start
+
+sleep 2000
+
+#$loop_cnt = 0
+#check_dnode_ready:
+#	$loop_cnt = $loop_cnt + 1
+#	sleep 200
+#	if $loop_cnt == 10 then
+#	  print ====> dnode not ready!
+#		return -1
+#	endi
+#sql show dnodes
+#print ===> $rows $data[0][0] $data[0][1] $data[0][2] $data[0][3] $data[0][4] $data[0][5] $data[0][6]
+#print ===> $rows $data[1][0] $data[1][1] $data[1][2] $data[1][3] $data[1][4] $data[1][5] $data[1][6]
+#print ===> $rows $data[2][0] $data[2][1] $data[2][2] $data[2][3] $data[2][4] $data[2][5] $data[2][6]
+#print ===> $rows $data[3][0] $data[3][1] $data[3][2] $data[3][3] $data[3][4] $data[3][5] $data[3][6]
+#if $data[0][0] != 1 then
+#  return -1
+#endi
+#if $data[0][4] != ready then
+#  goto check_dnode_ready
+#endi
+#
+##sql connect
+#sql create dnode $hostname port 7200
+#sql create dnode $hostname port 7300
+#sql create dnode $hostname port 7400
+#
+#$loop_cnt = 0
+#check_dnode_ready_1:
+#$loop_cnt = $loop_cnt + 1
+#sleep 200
+#if $loop_cnt == 10 then
+#  print ====> dnodes not ready!
+#	return -1
+#endi
+#sql show dnodes
+#print ===> $rows $data[0][0] $data[0][1] $data[0][2] $data[0][3] $data[0][4] $data[0][5] $data[0][6]
+#print ===> $rows $data[1][0] $data[1][1] $data[1][2] $data[1][3] $data[1][4] $data[1][5] $data[1][6]
+#print ===> $rows $data[2][0] $data[2][1] $data[2][2] $data[2][3] $data[2][4] $data[2][5] $data[2][6]
+#print ===> $rows $data[3][0] $data[3][1] $data[3][2] $data[3][3] $data[3][4] $data[3][5] $data[3][6]
+#if $data[0][4] != ready then
+#  goto check_dnode_ready_1
+#endi
+#if $data[1][4] != ready then
+#  goto check_dnode_ready_1
+#endi
+#if $data[2][4] != ready then
+#  goto check_dnode_ready_1
+#endi
+#if $data[3][4] != ready then
+#  goto check_dnode_ready_1
+#endi
+
+#=========== please add any actions above =================
+
+print ====> stop all dondes to output valgrind log file
+system sh/exec.sh -n dnode1 -s stop -x SIGINT
+
+print ====> start to check if there are ERRORS in vagrind log file for each dnode
+# -n : dnode[x] be check
+system_content sh/checkValgrind.sh -n dnode1 
+print cmd return result----> [ $system_content ]
+if $system_content == @ERROR SUMMARY: 0 errors@ then
+  return 0
+endi 
+
+$null=
+if $system_content == $null then
+  return 0
+endi 
+
+return -1
\ No newline at end of file

tests/tsim/src/simExe.c

@@ -441,6 +441,12 @@ void simStoreSystemContentResult(SScript *script, char *filename) {
   // if ((fd = fopen(filename, "r")) != NULL) {
   if ((pFile = taosOpenFile(filename, TD_FILE_READ)) != NULL) {
     taosReadFile(pFile, script->system_ret_content, MAX_SYSTEM_RESULT_LEN - 1);
+    int32_t len = strlen(script->system_ret_content);
+    for (int32_t i = 0; i < len; ++i) {
+      if (script->system_ret_content[i] == '\n' || script->system_ret_content[i] == '\r') {
+        script->system_ret_content[i] = 0;
+      }
+    }
     taosCloseFile(&pFile);
     char rmCmd[MAX_FILE_NAME_LEN] = {0};
     sprintf(rmCmd, "rm -f %s", filename);

tools/shell/src/shellEngine.c

@@ -224,63 +224,27 @@ int32_t shellRunCommand(TAOS *con, char *command) {
     }
   }
 
-  bool esc = false;
-  char quote = 0, *cmd = command, *p = command;
+  char quote = 0, *cmd = command;
   for (char c = *command++; c != 0; c = *command++) {
-    if (esc) {
-      switch (c) {
-        case 'n':
-          c = '\n';
-          break;
-        case 'r':
-          c = '\r';
-          break;
-        case 't':
-          c = '\t';
-          break;
-        case 'G':
-          *p++ = '\\';
-          break;
-        case '\'':
-        case '"':
-          if (quote) {
-            *p++ = '\\';
-          }
-          break;
-      }
-      *p++ = c;
-      esc = false;
+    if (c == '\\' && (*command == '\'' || *command == '"' || *command == '`')) {
+      command ++;
       continue;
     }
 
-    if (c == '\\') {
-      if (quote != 0 && (*command == '_' || *command == '\\')) {
-        // DO nothing
-      } else {
-        esc = true;
-        continue;
-      }
-    }
-
     if (quote == c) {
       quote = 0;
-    } else if (quote == 0 && (c == '\'' || c == '"')) {
+    } else if (quote == 0 && (c == '\'' || c == '"' || c == '`')) {
       quote = c;
-    }
-
-    *p++ = c;
-    if (c == ';' && quote == 0) {
-      c = *p;
-      *p = 0;
+    } else if (c == ';' && quote == 0) {
+      c = *command;
+      *command = 0;
       if (shellRunSingleCommand(con, cmd) < 0) {
         return -1;
       }
-      *p = c;
-      p = cmd;
+      *command = c;
+      cmd = command;
     }
   }
-
-  *p = 0;
   return shellRunSingleCommand(con, cmd);
 }
 
@@ -574,19 +538,23 @@ static void shellPrintNChar(const char *str, int length, int width) {
   while (pos < length) {
     TdWchar wc;
     int     bytes = taosMbToWchar(&wc, str + pos, MB_CUR_MAX);
-    if (bytes == 0) {
+    if (bytes <= 0) {
       break;
     }
-    pos += bytes;
-    if (pos > length) {
+    if (pos + bytes > length) {
       break;
     }
-
+    int w = 0;
 #ifdef WINDOWS
-    int w = bytes;
+    w = bytes;
 #else
-    int w = taosWcharWidth(wc);
+    if(*(str + pos) == '\t' || *(str + pos) == '\n' || *(str + pos) == '\r'){
+      w = bytes;
+    }else{
+      w = taosWcharWidth(wc);
+    }
 #endif
+    pos += bytes;
     if (w <= 0) {
       continue;
     }
@@ -680,6 +648,7 @@ static void printField(const char *val, TAOS_FIELD *field, int width, int32_t le
       break;
     case TSDB_DATA_TYPE_BINARY:
     case TSDB_DATA_TYPE_NCHAR:
+    case TSDB_DATA_TYPE_JSON:
       shellPrintNChar(val, length, width);
       break;
     case TSDB_DATA_TYPE_TIMESTAMP:
@@ -792,7 +761,8 @@ static int calcColWidth(TAOS_FIELD *field, int precision) {
         return TMAX(field->bytes, width);
       }
 
-    case TSDB_DATA_TYPE_NCHAR: {
+    case TSDB_DATA_TYPE_NCHAR:
+    case TSDB_DATA_TYPE_JSON:{
       int16_t bytes = field->bytes * TSDB_NCHAR_SIZE;
       if (bytes > tsMaxBinaryDisplayWidth) {
         return TMAX(tsMaxBinaryDisplayWidth, width);