tdataformat.h 23.1 KB
Newer Older
H
more  
hzcheng 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/*
 * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
 *
 * This program is free software: you can use, redistribute, and/or modify
 * it under the terms of the GNU Affero General Public License, version 3
 * or later ("AGPL"), as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.
 *
 * You should have received a copy of the GNU Affero General Public License
 * along with this program. If not, see <http://www.gnu.org/licenses/>.
 */
H
hzcheng 已提交
15
#ifndef _TD_DATA_FORMAT_H_
H
more  
Hongze Cheng 已提交
16 17
#define _TD_DATA_FORMAT_H_

S
TD-4088  
Shengliang Guan 已提交
18
#include "os.h"
H
Hongze Cheng 已提交
19
#include "talgo.h"
H
Haojun Liao 已提交
20
#include "ttype.h"
H
TD-166  
hzcheng 已提交
21
#include "tutil.h"
H
hzcheng 已提交
22

H
more  
hzcheng 已提交
23 24 25
#ifdef __cplusplus
extern "C" {
#endif
H
hzcheng 已提交
26

C
Cary Xu 已提交
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51
typedef struct {
  VarDataLenT len;
  uint8_t     data;
} SBinaryNullT;

typedef struct {
  VarDataLenT len;
  uint32_t    data;
} SNCharNullT;

extern const uint8_t      BoolNull;
extern const uint8_t      TinyintNull;
extern const uint16_t     SmallintNull;
extern const uint32_t     IntNull;
extern const uint64_t     BigintNull;
extern const uint64_t     TimestampNull;
extern const uint8_t      UTinyintNull;
extern const uint16_t     USmallintNull;
extern const uint32_t     UIntNull;
extern const uint64_t     UBigintNull;
extern const uint32_t     FloatNull;
extern const uint64_t     DoubleNull;
extern const SBinaryNullT BinaryNull;
extern const SNCharNullT  NcharNull;

52 53 54 55 56
#define STR_TO_VARSTR(x, str)                     \
  do {                                            \
    VarDataLenT __len = (VarDataLenT)strlen(str); \
    *(VarDataLenT *)(x) = __len;                  \
    memcpy(varDataVal(x), (str), __len);          \
H
Hongze Cheng 已提交
57 58
  } while (0);

59 60
#define STR_WITH_MAXSIZE_TO_VARSTR(x, str, _maxs)                         \
  do {                                                                    \
H
Hui Li 已提交
61
    char *_e = stpncpy(varDataVal(x), (str), (_maxs)-VARSTR_HEADER_SIZE); \
62
    varDataSetLen(x, (_e - (x)-VARSTR_HEADER_SIZE));                      \
H
Hongze Cheng 已提交
63 64
  } while (0)

65 66 67 68
#define STR_WITH_SIZE_TO_VARSTR(x, str, _size)  \
  do {                                          \
    *(VarDataLenT *)(x) = (VarDataLenT)(_size); \
    memcpy(varDataVal(x), (str), (_size));      \
H
Hongze Cheng 已提交
69
  } while (0);
H
hjxilinx 已提交
70

H
hzcheng 已提交
71 72
// ----------------- TSDB COLUMN DEFINITION
typedef struct {
C
Cary Xu 已提交
73 74 75 76
  int8_t   type;    // Column type
  int16_t  colId;   // column ID
  uint16_t bytes;   // column bytes
  uint16_t offset;  // point offset in SDataRow/SKVRow after the header part.
H
hzcheng 已提交
77 78 79 80 81 82 83 84 85 86 87 88 89 90
} STColumn;

#define colType(col) ((col)->type)
#define colColId(col) ((col)->colId)
#define colBytes(col) ((col)->bytes)
#define colOffset(col) ((col)->offset)

#define colSetType(col, t) (colType(col) = (t))
#define colSetColId(col, id) (colColId(col) = (id))
#define colSetBytes(col, b) (colBytes(col) = (b))
#define colSetOffset(col, o) (colOffset(col) = (o))

// ----------------- TSDB SCHEMA DEFINITION
typedef struct {
H
Hongze Cheng 已提交
91
  int      version;    // version
H
hzcheng 已提交
92
  int      numOfCols;  // Number of columns appended
H
refact  
Hongze Cheng 已提交
93
  int      tlen;       // maximum length of a SDataRow without the header part (sizeof(VarDataOffsetT) + sizeof(VarDataLenT) + (bytes))
H
TD-353  
Hongze Cheng 已提交
94
  uint16_t flen;       // First part length in a SDataRow after the header part
H
refact  
Hongze Cheng 已提交
95
  uint16_t vlen;       // pure value part length, excluded the overhead (bytes only)
H
hzcheng 已提交
96 97 98 99
  STColumn columns[];
} STSchema;

#define schemaNCols(s) ((s)->numOfCols)
H
Hongze Cheng 已提交
100
#define schemaVersion(s) ((s)->version)
H
TD-166  
hzcheng 已提交
101 102
#define schemaTLen(s) ((s)->tlen)
#define schemaFLen(s) ((s)->flen)
T
Tao Liu 已提交
103
#define schemaVLen(s) ((s)->vlen)
H
hzcheng 已提交
104
#define schemaColAt(s, i) ((s)->columns + i)
S
TD-1848  
Shengliang Guan 已提交
105
#define tdFreeSchema(s) tfree((s))
H
hzcheng 已提交
106 107

STSchema *tdDupSchema(STSchema *pSchema);
H
TD-353  
Hongze Cheng 已提交
108
int       tdEncodeSchema(void **buf, STSchema *pSchema);
H
TD-353  
Hongze Cheng 已提交
109
void *    tdDecodeSchema(void *buf, STSchema **pRSchema);
H
hzcheng 已提交
110

H
Hongze Cheng 已提交
111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126
static FORCE_INLINE int comparColId(const void *key1, const void *key2) {
  if (*(int16_t *)key1 > ((STColumn *)key2)->colId) {
    return 1;
  } else if (*(int16_t *)key1 < ((STColumn *)key2)->colId) {
    return -1;
  } else {
    return 0;
  }
}

static FORCE_INLINE STColumn *tdGetColOfID(STSchema *pSchema, int16_t colId) {
  void *ptr = bsearch(&colId, (void *)pSchema->columns, schemaNCols(pSchema), sizeof(STColumn), comparColId);
  if (ptr == NULL) return NULL;
  return (STColumn *)ptr;
}

H
Hongze Cheng 已提交
127 128 129 130 131
// ----------------- SCHEMA BUILDER DEFINITION
typedef struct {
  int       tCols;
  int       nCols;
  int       tlen;
H
TD-353  
Hongze Cheng 已提交
132 133
  uint16_t  flen;
  uint16_t  vlen;
H
Hongze Cheng 已提交
134 135 136 137 138 139 140
  int       version;
  STColumn *columns;
} STSchemaBuilder;

int       tdInitTSchemaBuilder(STSchemaBuilder *pBuilder, int32_t version);
void      tdDestroyTSchemaBuilder(STSchemaBuilder *pBuilder);
void      tdResetTSchemaBuilder(STSchemaBuilder *pBuilder, int32_t version);
141
int       tdAddColToSchema(STSchemaBuilder *pBuilder, int8_t type, int16_t colId, int16_t bytes);
H
Hongze Cheng 已提交
142 143
STSchema *tdGetSchemaFromBuilder(STSchemaBuilder *pBuilder);

H
TD-1548  
Hongze Cheng 已提交
144 145 146 147 148
// ----------------- Semantic timestamp key definition
typedef uint64_t TKEY;

#define TKEY_INVALID UINT64_MAX
#define TKEY_NULL TKEY_INVALID
H
Hongze Cheng 已提交
149 150
#define TKEY_NEGATIVE_FLAG (((TKEY)1) << 63)
#define TKEY_DELETE_FLAG (((TKEY)1) << 62)
H
TD-1548  
Hongze Cheng 已提交
151 152 153 154 155 156 157 158
#define TKEY_VALUE_FILTER (~(TKEY_NEGATIVE_FLAG | TKEY_DELETE_FLAG))

#define TKEY_IS_NEGATIVE(tkey) (((tkey)&TKEY_NEGATIVE_FLAG) != 0)
#define TKEY_IS_DELETED(tkey) (((tkey)&TKEY_DELETE_FLAG) != 0)
#define tdSetTKEYDeleted(tkey) ((tkey) | TKEY_DELETE_FLAG)
#define tdGetTKEY(key) (((TKEY)ABS(key)) | (TKEY_NEGATIVE_FLAG & (TKEY)(key)))
#define tdGetKey(tkey) (((TSKEY)((tkey)&TKEY_VALUE_FILTER)) * (TKEY_IS_NEGATIVE(tkey) ? -1 : 1))

D
dapan1121 已提交
159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
#define MIN_TS_KEY ((TSKEY)0x8000000000000001)
#define MAX_TS_KEY ((TSKEY)0x3fffffffffffffff)

#define TD_TO_TKEY(key) tdGetTKEY(((key) < MIN_TS_KEY) ? MIN_TS_KEY : (((key) > MAX_TS_KEY) ? MAX_TS_KEY : key))

static FORCE_INLINE TKEY keyToTkey(TSKEY key) {
  TSKEY lkey = key;
  if (key > MAX_TS_KEY) {
    lkey = MAX_TS_KEY;
  } else if (key < MIN_TS_KEY) {
    lkey = MIN_TS_KEY;
  }

  return tdGetTKEY(lkey);
}

H
TD-1548  
Hongze Cheng 已提交
175 176 177 178 179 180 181 182 183 184 185 186
static FORCE_INLINE int tkeyComparFn(const void *tkey1, const void *tkey2) {
  TSKEY key1 = tdGetKey(*(TKEY *)tkey1);
  TSKEY key2 = tdGetKey(*(TKEY *)tkey2);

  if (key1 < key2) {
    return -1;
  } else if (key1 > key2) {
    return 1;
  } else {
    return 0;
  }
}
C
Cary Xu 已提交
187
// ----------------- Sequential Data row structure
H
more  
Hongze Cheng 已提交
188

C
Cary Xu 已提交
189
/* A sequential data row, the format is like below:
H
TD-90  
Hongze Cheng 已提交
190 191 192
 * |<--------------------+--------------------------- len ---------------------------------->|
 * |<--     Head      -->|<---------   flen -------------->|                                 |
 * +---------------------+---------------------------------+---------------------------------+
B
Bomin Zhang 已提交
193
 * | uint16_t |  int16_t |                                 |                                 |
H
TD-90  
Hongze Cheng 已提交
194 195 196
 * +----------+----------+---------------------------------+---------------------------------+
 * |   len    | sversion |           First part            |             Second part         |
 * +----------+----------+---------------------------------+---------------------------------+
197
 *
H
TD-1548  
Hongze Cheng 已提交
198
 * NOTE: timestamp in this row structure is TKEY instead of TSKEY
H
more  
Hongze Cheng 已提交
199
 */
H
hzcheng 已提交
200
typedef void *SDataRow;
C
Cary Xu 已提交
201 202 203 204 205 206 207 208 209 210 211 212
/* A memory data row, the format is like below:
 *|---------+---------------------+--------------------------- len ---------------------------------->|
 *|<- type->|<--     Head      -->|<---------   flen -------------->|                                 |
 *|---------+---------------------+---------------------------------+---------------------------------+
 *| uint8_t | uint16_t |  int16_t |                                 |                                 |
 *|---------+----------+----------+---------------------------------+---------------------------------+
 *| flag    |   len    | sversion |           First part            |             Second part         |
 *|---------+----------+----------+---------------------------------+---------------------------------+
 *
 * NOTE: timestamp in this row structure is TKEY instead of TSKEY
 */
typedef void *SMemRow;
H
hzcheng 已提交
213

B
Bomin Zhang 已提交
214
#define TD_DATA_ROW_HEAD_SIZE (sizeof(uint16_t) + sizeof(int16_t))
H
hzcheng 已提交
215

B
Bomin Zhang 已提交
216
#define dataRowLen(r) (*(uint16_t *)(r))
C
Cary Xu 已提交
217
#define dataRowVersion(r) (*(int16_t *)POINTER_SHIFT(r, sizeof(int16_t)))
H
hzcheng 已提交
218
#define dataRowTuple(r) POINTER_SHIFT(r, TD_DATA_ROW_HEAD_SIZE)
H
TD-1548  
Hongze Cheng 已提交
219 220
#define dataRowTKey(r) (*(TKEY *)(dataRowTuple(r)))
#define dataRowKey(r) tdGetKey(dataRowTKey(r))
H
hzcheng 已提交
221
#define dataRowSetLen(r, l) (dataRowLen(r) = (l))
H
TD-90  
Hongze Cheng 已提交
222
#define dataRowSetVersion(r, v) (dataRowVersion(r) = (v))
H
hzcheng 已提交
223
#define dataRowCpy(dst, r) memcpy((dst), (r), dataRowLen(r))
H
TD-166  
hzcheng 已提交
224
#define dataRowMaxBytesFromSchema(s) (schemaTLen(s) + TD_DATA_ROW_HEAD_SIZE)
H
TD-1548  
Hongze Cheng 已提交
225
#define dataRowDeleted(r) TKEY_IS_DELETED(dataRowTKey(r))
H
hzcheng 已提交
226

C
Cary Xu 已提交
227 228
SDataRow tdNewDataRowFromSchema(STSchema *pSchema);
void     tdFreeDataRow(SDataRow row);
H
TD-166  
hzcheng 已提交
229
void     tdInitDataRow(SDataRow row, STSchema *pSchema);
C
Cary Xu 已提交
230
SDataRow tdDataRowDup(SDataRow row);
C
Cary Xu 已提交
231
SMemRow tdMemRowDup(SMemRow row);
H
more  
Hongze Cheng 已提交
232

233
// offset here not include dataRow header length
C
Cary Xu 已提交
234
static FORCE_INLINE int tdAppendColVal(SDataRow row, const void *value, int8_t type, int32_t bytes, int32_t offset) {
235 236 237 238
  ASSERT(value != NULL);
  int32_t toffset = offset + TD_DATA_ROW_HEAD_SIZE;
  char *  ptr = (char *)POINTER_SHIFT(row, dataRowLen(row));

H
TD-1548  
Hongze Cheng 已提交
239 240 241 242 243 244 245 246
  if (IS_VAR_DATA_TYPE(type)) {
    *(VarDataOffsetT *)POINTER_SHIFT(row, toffset) = dataRowLen(row);
    memcpy(ptr, value, varDataTLen(value));
    dataRowLen(row) += varDataTLen(value);
  } else {
    if (offset == 0) {
      ASSERT(type == TSDB_DATA_TYPE_TIMESTAMP);
      TKEY tvalue = tdGetTKEY(*(TSKEY *)value);
H
Hongze Cheng 已提交
247
      memcpy(POINTER_SHIFT(row, toffset), (void *)(&tvalue), TYPE_BYTES[type]);
H
TD-1548  
Hongze Cheng 已提交
248
    } else {
249
      memcpy(POINTER_SHIFT(row, toffset), value, TYPE_BYTES[type]);
H
TD-1548  
Hongze Cheng 已提交
250
    }
251 252 253 254 255
  }

  return 0;
}

C
Cary Xu 已提交
256 257 258 259 260 261 262 263 264 265
// NOTE: offset here including the header size
static FORCE_INLINE void *tdGetRowDataOfCol(void *row, int8_t type, int32_t offset) {
  if (IS_VAR_DATA_TYPE(type)) {
    return POINTER_SHIFT(row, *(VarDataOffsetT *)POINTER_SHIFT(row, offset));
  } else {
    return POINTER_SHIFT(row, offset);
  }
  return NULL;
}

H
TD-34  
hzcheng 已提交
266 267
// ----------------- Data column structure
typedef struct SDataCol {
H
TD-166  
hzcheng 已提交
268 269 270
  int8_t          type;       // column type
  int16_t         colId;      // column ID
  int             bytes;      // column data bytes defined
H
TD-166  
hzcheng 已提交
271
  int             offset;     // data offset in a SDataRow (including the header size)
H
TD-166  
hzcheng 已提交
272 273 274 275
  int             spaceSize;  // Total space size for this column
  int             len;        // column data length
  VarDataOffsetT *dataOff;    // For binary and nchar data, the offset in the data column
  void *          pData;      // Actual data pointer
L
lichuang 已提交
276
  TSKEY           ts;         // only used in last NULL column
H
TD-34  
hzcheng 已提交
277 278
} SDataCol;

C
Cary Xu 已提交
279
#define isAllRowOfColNull(pCol) ((pCol)->len == 0)
H
TD-166  
hzcheng 已提交
280 281 282 283 284
static FORCE_INLINE void dataColReset(SDataCol *pDataCol) { pDataCol->len = 0; }

void dataColInit(SDataCol *pDataCol, STColumn *pCol, void **pBuf, int maxPoints);
void dataColSetOffset(SDataCol *pCol, int nEle);

H
TD-166  
hzcheng 已提交
285 286
bool isNEleNull(SDataCol *pCol, int nEle);
void dataColSetNEleNull(SDataCol *pCol, int nEle, int maxPoints);
H
TD-166  
hzcheng 已提交
287

L
liuyq-617 已提交
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319
FORCE_INLINE void dataColAppendVal(SDataCol *pCol, const void *value, int numOfRows, int maxPoints);
// value from timestamp should be TKEY here instead of TSKEY
FORCE_INLINE void dataColAppendVal(SDataCol *pCol, const void *value, int numOfRows, int maxPoints) {
  ASSERT(pCol != NULL && value != NULL);

  if (pCol->len == 0) {
    if (isNull(value, pCol->type)) {
      // all null value yet, just return
      return;
    }

    if (numOfRows > 0) {
      // Find the first not null value, fill all previous values as NULL
      dataColSetNEleNull(pCol, numOfRows, maxPoints);
    }
  }

  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;
  }
}

static FORCE_INLINE const void *tdGetNullVal(int8_t type) {
C
Cary Xu 已提交
320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
  switch (type) {
    case TSDB_DATA_TYPE_BOOL:
      return &BoolNull;
    case TSDB_DATA_TYPE_TINYINT:
      return &TinyintNull;
    case TSDB_DATA_TYPE_SMALLINT:
      return &SmallintNull;
    case TSDB_DATA_TYPE_INT:
      return &IntNull;
    case TSDB_DATA_TYPE_BIGINT:
      return &BigintNull;
    case TSDB_DATA_TYPE_FLOAT:
      return &FloatNull;
    case TSDB_DATA_TYPE_DOUBLE:
      return &DoubleNull;
    case TSDB_DATA_TYPE_BINARY:
      return &BinaryNull;
    case TSDB_DATA_TYPE_TIMESTAMP:
      return &TimestampNull;
    case TSDB_DATA_TYPE_NCHAR:
      return &NcharNull;
    case TSDB_DATA_TYPE_UTINYINT:
      return &UTinyintNull;
    case TSDB_DATA_TYPE_USMALLINT:
      return &USmallintNull;
    case TSDB_DATA_TYPE_UINT:
      return &UIntNull;
    case TSDB_DATA_TYPE_UBIGINT:
      return &UBigintNull;
    default:
      ASSERT(0);
      return NULL;
  }
}
H
TD-166  
hzcheng 已提交
354
// Get the data pointer from a column-wised data
C
Cary Xu 已提交
355 356 357 358
static FORCE_INLINE const void *tdGetColDataOfRow(SDataCol *pCol, int row) {
  if (isAllRowOfColNull(pCol)) {
    return tdGetNullVal(pCol->type);
  }
H
Hongze Cheng 已提交
359 360 361 362
  if (IS_VAR_DATA_TYPE(pCol->type)) {
    return POINTER_SHIFT(pCol->pData, pCol->dataOff[row]);
  } else {
    return POINTER_SHIFT(pCol->pData, TYPE_BYTES[pCol->type] * row);
H
TD-166  
hzcheng 已提交
363 364 365
  }
}

H
TD-166  
hzcheng 已提交
366
static FORCE_INLINE int32_t dataColGetNEleLen(SDataCol *pDataCol, int rows) {
H
TD-166  
hzcheng 已提交
367 368
  ASSERT(rows > 0);

H
Hongze Cheng 已提交
369 370 371 372
  if (IS_VAR_DATA_TYPE(pDataCol->type)) {
    return pDataCol->dataOff[rows - 1] + varDataTLen(tdGetColDataOfRow(pDataCol, rows - 1));
  } else {
    return TYPE_BYTES[pDataCol->type] * rows;
H
TD-166  
hzcheng 已提交
373 374 375
  }
}

H
TD-34  
hzcheng 已提交
376
typedef struct {
H
Hongze Cheng 已提交
377 378 379 380
  int maxRowSize;
  int maxCols;    // max number of columns
  int maxPoints;  // max number of points
  int bufSize;
H
TD-166  
hzcheng 已提交
381

H
Hongze Cheng 已提交
382 383 384 385 386
  int       numOfRows;
  int       numOfCols;  // Total number of cols
  int       sversion;   // TODO: set sversion
  void *    buf;
  SDataCol *cols;
H
TD-34  
hzcheng 已提交
387 388
} SDataCols;

H
TD-34  
hzcheng 已提交
389
#define keyCol(pCols) (&((pCols)->cols[0]))  // Key column
C
Cary Xu 已提交
390
#define dataColsTKeyAt(pCols, idx) ((TKEY *)(keyCol(pCols)->pData))[(idx)]  // the idx row of column-wised data
H
TD-1548  
Hongze Cheng 已提交
391
#define dataColsKeyAt(pCols, idx) tdGetKey(dataColsTKeyAt(pCols, idx))
392 393 394 395 396 397 398 399
static FORCE_INLINE TKEY dataColsTKeyFirst(SDataCols *pCols) {
  if (pCols->numOfRows) {
    return dataColsTKeyAt(pCols, 0);
  } else {
    return TKEY_INVALID;
  }
}

400 401 402 403 404
static FORCE_INLINE TSKEY dataColsKeyAtRow(SDataCols *pCols, int row) {
  ASSERT(row < pCols->numOfRows);
  return dataColsKeyAt(pCols, row);
}

405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427
static FORCE_INLINE TSKEY dataColsKeyFirst(SDataCols *pCols) {
  if (pCols->numOfRows) {
    return dataColsKeyAt(pCols, 0);
  } else {
    return TSDB_DATA_TIMESTAMP_NULL;
  }
}

static FORCE_INLINE TKEY dataColsTKeyLast(SDataCols *pCols) {
  if (pCols->numOfRows) {
    return dataColsTKeyAt(pCols, pCols->numOfRows - 1);
  } else {
    return TKEY_INVALID;
  }
}

static FORCE_INLINE TSKEY dataColsKeyLast(SDataCols *pCols) {
  if (pCols->numOfRows) {
    return dataColsKeyAt(pCols, pCols->numOfRows - 1);
  } else {
    return TSDB_DATA_TIMESTAMP_NULL;
  }
}
H
TD-34  
hzcheng 已提交
428

H
TD-166  
hzcheng 已提交
429
SDataCols *tdNewDataCols(int maxRowSize, int maxCols, int maxRows);
H
TD-34  
hzcheng 已提交
430
void       tdResetDataCols(SDataCols *pCols);
H
Hongze Cheng 已提交
431
int        tdInitDataCols(SDataCols *pCols, STSchema *pSchema);
H
TD-100  
hzcheng 已提交
432
SDataCols *tdDupDataCols(SDataCols *pCols, bool keepData);
H
Hongze Cheng 已提交
433
SDataCols *tdFreeDataCols(SDataCols *pCols);
C
Cary Xu 已提交
434
void       tdAppendMemRowToDataCol(SMemRow row, STSchema *pSchema, SDataCols *pCols);
H
Hongze Cheng 已提交
435
int        tdMergeDataCols(SDataCols *target, SDataCols *source, int rowsToMerge, int *pOffset);
H
more  
Hongze Cheng 已提交
436

H
Hongze Cheng 已提交
437 438 439
// ----------------- K-V data row structure
/*
 * +----------+----------+---------------------------------+---------------------------------+
C
Cary Xu 已提交
440
 * | uint16_t |  int16_t |                                 |                                 |
H
Hongze Cheng 已提交
441 442 443 444
 * +----------+----------+---------------------------------+---------------------------------+
 * |    len   |   ncols  |           cols index            |             data part           |
 * +----------+----------+---------------------------------+---------------------------------+
 */
H
Hongze Cheng 已提交
445
typedef void *SKVRow;
H
Hongze Cheng 已提交
446 447

typedef struct {
C
Cary Xu 已提交
448 449
  int16_t  colId;
  uint16_t offset;
H
Hongze Cheng 已提交
450 451
} SColIdx;

C
Cary Xu 已提交
452
#define TD_KV_ROW_HEAD_SIZE (sizeof(uint16_t) + sizeof(int16_t))
H
Hongze Cheng 已提交
453

C
Cary Xu 已提交
454
#define kvRowLen(r) (*(uint16_t *)(r))
C
Cary Xu 已提交
455
#define kvRowNCols(r) (*(int16_t *)POINTER_SHIFT(r, sizeof(uint16_t)))
H
Hongze Cheng 已提交
456 457
#define kvRowSetLen(r, len) kvRowLen(r) = (len)
#define kvRowSetNCols(r, n) kvRowNCols(r) = (n)
C
Cary Xu 已提交
458
#define kvRowColIdx(r) ((SColIdx *)POINTER_SHIFT(r, TD_KV_ROW_HEAD_SIZE))
H
Hongze Cheng 已提交
459 460 461 462
#define kvRowValues(r) POINTER_SHIFT(r, TD_KV_ROW_HEAD_SIZE + sizeof(SColIdx) * kvRowNCols(r))
#define kvRowCpy(dst, r) memcpy((dst), (r), kvRowLen(r))
#define kvRowColVal(r, colIdx) POINTER_SHIFT(kvRowValues(r), (colIdx)->offset)
#define kvRowColIdxAt(r, i) (kvRowColIdx(r) + (i))
S
TD-1848  
Shengliang Guan 已提交
463
#define kvRowFree(r) tfree(r)
H
TD-90  
Hongze Cheng 已提交
464
#define kvRowEnd(r) POINTER_SHIFT(r, kvRowLen(r))
C
Cary Xu 已提交
465 466 467 468 469
#define kvRowVersion(r) (-1)

#define kvRowTKey(r) (*(TKEY *)(kvRowValues(r)))
#define kvRowKey(r) tdGetKey(kvRowTKey(r))
#define kvRowDeleted(r) TKEY_IS_DELETED(kvRowTKey(r))
H
Hongze Cheng 已提交
470

H
Hongze Cheng 已提交
471
SKVRow tdKVRowDup(SKVRow row);
H
TD-90  
Hongze Cheng 已提交
472
int    tdSetKVRowDataOfCol(SKVRow *orow, int16_t colId, int8_t type, void *value);
H
TD-353  
Hongze Cheng 已提交
473
int    tdEncodeKVRow(void **buf, SKVRow row);
H
Hongze Cheng 已提交
474
void * tdDecodeKVRow(void *buf, SKVRow *row);
B
Bomin Zhang 已提交
475
void   tdSortKVRowByColIdx(SKVRow row);
H
Hongze Cheng 已提交
476 477 478 479 480 481 482 483 484 485 486

static FORCE_INLINE int comparTagId(const void *key1, const void *key2) {
  if (*(int16_t *)key1 > ((SColIdx *)key2)->colId) {
    return 1;
  } else if (*(int16_t *)key1 < ((SColIdx *)key2)->colId) {
    return -1;
  } else {
    return 0;
  }
}

H
Hongze Cheng 已提交
487
static FORCE_INLINE void *tdGetKVRowValOfCol(SKVRow row, int16_t colId) {
H
Hongze Cheng 已提交
488
  void *ret = taosbsearch(&colId, kvRowColIdx(row), kvRowNCols(row), sizeof(SColIdx), comparTagId, TD_EQ);
H
Hongze Cheng 已提交
489
  if (ret == NULL) return NULL;
H
Hongze Cheng 已提交
490
  return kvRowColVal(row, (SColIdx *)ret);
H
Hongze Cheng 已提交
491 492
}

L
liuyq-617 已提交
493 494 495 496
static FORCE_INLINE void *tdGetKVRowIdxOfCol(SKVRow row, int16_t colId) {
  return taosbsearch(&colId, kvRowColIdx(row), kvRowNCols(row), sizeof(SColIdx), comparTagId, TD_EQ);
}

C
Cary Xu 已提交
497 498 499 500 501 502 503 504
// offset here not include kvRow header length
static FORCE_INLINE int tdAppendKvColVal(SKVRow row, const void *value, int16_t colId, int8_t type, int32_t offset) {
  ASSERT(value != NULL);
  int32_t  toffset = offset + TD_KV_ROW_HEAD_SIZE;
  SColIdx *pColIdx = (SColIdx *)POINTER_SHIFT(row, toffset);
  char *   ptr = (char *)POINTER_SHIFT(row, kvRowLen(row));

  pColIdx->colId = colId;
C
Cary Xu 已提交
505
  pColIdx->offset = kvRowLen(row);  // offset of pColIdx including the TD_KV_ROW_HEAD_SIZE
C
Cary Xu 已提交
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523

  if (IS_VAR_DATA_TYPE(type)) {
    memcpy(ptr, value, varDataTLen(value));
    kvRowLen(row) += varDataTLen(value);
  } else {
    if (offset == 0) {
      ASSERT(type == TSDB_DATA_TYPE_TIMESTAMP);
      TKEY tvalue = tdGetTKEY(*(TSKEY *)value);
      memcpy(ptr, (void *)(&tvalue), TYPE_BYTES[type]);
    } else {
      memcpy(ptr, value, TYPE_BYTES[type]);
    }
    kvRowLen(row) += TYPE_BYTES[type];
  }

  return 0;
}

H
Hongze Cheng 已提交
524 525 526 527 528
// ----------------- K-V data row builder
typedef struct {
  int16_t  tCols;
  int16_t  nCols;
  SColIdx *pColIdx;
C
Cary Xu 已提交
529 530
  uint16_t alloc;
  uint16_t size;
H
Hongze Cheng 已提交
531
  void *   buf;
H
Hongze Cheng 已提交
532
} SKVRowBuilder;
H
Hongze Cheng 已提交
533

H
Hongze Cheng 已提交
534 535 536 537
int    tdInitKVRowBuilder(SKVRowBuilder *pBuilder);
void   tdDestroyKVRowBuilder(SKVRowBuilder *pBuilder);
void   tdResetKVRowBuilder(SKVRowBuilder *pBuilder);
SKVRow tdGetKVRowFromBuilder(SKVRowBuilder *pBuilder);
H
Hongze Cheng 已提交
538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564

static FORCE_INLINE int tdAddColToKVRow(SKVRowBuilder *pBuilder, int16_t colId, int8_t type, void *value) {
  if (pBuilder->nCols >= pBuilder->tCols) {
    pBuilder->tCols *= 2;
    pBuilder->pColIdx = (SColIdx *)realloc((void *)(pBuilder->pColIdx), sizeof(SColIdx) * pBuilder->tCols);
    if (pBuilder->pColIdx == NULL) return -1;
  }

  pBuilder->pColIdx[pBuilder->nCols].colId = colId;
  pBuilder->pColIdx[pBuilder->nCols].offset = pBuilder->size;

  pBuilder->nCols++;

  int 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;
    }
    pBuilder->buf = realloc(pBuilder->buf, pBuilder->alloc);
    if (pBuilder->buf == NULL) return -1;
  }

  memcpy(POINTER_SHIFT(pBuilder->buf, pBuilder->size), value, tlen);
  pBuilder->size += tlen;

  return 0;
}
H
Hongze Cheng 已提交
565

C
Cary Xu 已提交
566
// ----------------- Sequential Data row structure
C
Cary Xu 已提交
567 568 569 570 571 572 573 574
/*
 * |-------------------------------+--------------------------- len ---------------------------------->|
 * |<--------     Head      ------>|<---------   flen -------------->|                                 |
 * |---------+---------------------+---------------------------------+---------------------------------+
 * | uint8_t | uint16_t |  int16_t |                                 |                                 |
 * |---------+----------+----------+---------------------------------+---------------------------------+
 * |  flag   |   len    | sversion |           First part            |             Second part         |
 * +---------+----------+----------+---------------------------------+---------------------------------+
C
Cary Xu 已提交
575 576 577
 *
 * NOTE: timestamp in this row structure is TKEY instead of TSKEY
 */
C
Cary Xu 已提交
578

C
Cary Xu 已提交
579 580
// ----------------- K-V data row structure
/*
C
Cary Xu 已提交
581 582 583 584 585
 * |--------------------+----------+---------------------------------+---------------------------------+
 * | uint8_t | uint16_t |  int16_t |                                 |                                 |
 * |---------+----------+----------+---------------------------------+---------------------------------+
 * |   flag  |    len   |   ncols  |           cols index            |             data part           |
 * |---------+----------+----------+---------------------------------+---------------------------------+
C
Cary Xu 已提交
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
 */

#define TD_MEM_ROW_TYPE_SIZE sizeof(uint8_t)
#define TD_MEM_ROW_HEAD_SIZE (TD_MEM_ROW_TYPE_SIZE + sizeof(uint16_t) + sizeof(int16_t))

#define SMEM_ROW_DATA 0U  // SDataRow
#define SMEM_ROW_KV 1U    // SKVRow
#define TD_DO_NOTHING \
  do {                \
  } while (0)

#define memRowType(r) (*(uint8_t *)(r))
#define memRowBody(r) POINTER_SHIFT(r, TD_MEM_ROW_TYPE_SIZE)
#define memRowLen(r) (*(uint16_t *)POINTER_SHIFT(r, TD_MEM_ROW_TYPE_SIZE))
#define memRowTLen(r) (*(uint16_t *)POINTER_SHIFT(r, TD_MEM_ROW_TYPE_SIZE) + (uint16_t)TD_MEM_ROW_TYPE_SIZE)

#define isDataRow(r) (SMEM_ROW_DATA == memRowType(r))
#define isKvRow(r) (SMEM_ROW_KV == memRowType(r))
#define memRowVersion(r) (isDataRow(r) ? dataRowVersion(memRowBody(r)) : kvRowVersion(r))  // schema version

#define memRowTuple(r) (isDataRow(r) ? dataRowTuple(memRowBody(r)) : kvRowValues(memRowBody(r)))

#define memRowTKey(r) (isDataRow(r) ? dataRowTKey(memRowBody(r)) : kvRowTKey(memRowBody(r)))
#define memRowKey(r) (isDataRow(r) ? dataRowKey(memRowBody(r)) : kvRowKey(memRowBody(r)))

#define memRowSetType(r, t) (memRowType(r) = (t))
#define memRowSetLen(r, l) (memRowLen(r) = (l))
#define memRowSetVersion(r, v) (isDataRow(r) ? dataRowSetVersion(r, v) : TD_DO_NOTHING)
#define memRowCpy(dst, r) memcpy((dst), (r), memRowTLen(r))
#define memRowMaxBytesFromSchema(s) (schemaTLen(s) + TD_MEM_ROW_HEAD_SIZE)
#define memRowDeleted(r) TKEY_IS_DELETED(memRowTKey(r))

// NOTE: offset here including the header size
L
liuyq-617 已提交
619
static FORCE_INLINE void *tdGetKvRowDataOfCol(void *row, int32_t offset) { return POINTER_SHIFT(row, offset); }
C
Cary Xu 已提交
620
// NOTE: offset here including the header size
C
Cary Xu 已提交
621 622 623 624
static FORCE_INLINE void *tdGetMemRowDataOfCol(void *row, int8_t type, int32_t offset) {
  if (isDataRow(row)) {
    return tdGetRowDataOfCol(row, type, offset);
  } else if (isKvRow(row)) {
L
liuyq-617 已提交
625
    return tdGetKvRowDataOfCol(row, offset);
C
Cary Xu 已提交
626 627 628 629 630 631
  } else {
    ASSERT(0);
  }
  return NULL;
}

H
more  
hzcheng 已提交
632 633 634 635
#ifdef __cplusplus
}
#endif

H
hzcheng 已提交
636
#endif  // _TD_DATA_FORMAT_H_