• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

taosdata / TDengine / #3608

12 Feb 2025 05:57AM UTC coverage: 63.066% (+1.4%) from 61.715%
#3608

push

travis-ci

web-flow
Merge pull request #29746 from taosdata/merge/mainto3.02

merge: from main to 3.0 branch

140199 of 286257 branches covered (48.98%)

Branch coverage included in aggregate %.

89 of 161 new or added lines in 18 files covered. (55.28%)

3211 existing lines in 190 files now uncovered.

218998 of 283298 relevant lines covered (77.3%)

5949310.66 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

69.96
/source/libs/executor/src/sortoperator.c
1
/*
2
 * Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
3
 *
4
 * This program is free software: you can use, redistribute, and/or modify
5
 * it under the terms of the GNU Affero General Public License, version 3
6
 * or later ("AGPL"), as published by the Free Software Foundation.
7
 *
8
 * This program is distributed in the hope that it will be useful, but WITHOUT
9
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10
 * FITNESS FOR A PARTICULAR PURPOSE.
11
 *
12
 * You should have received a copy of the GNU Affero General Public License
13
 * along with this program. If not, see <http://www.gnu.org/licenses/>.
14
 */
15

16
#include "executorInt.h"
17
#include "filter.h"
18
#include "operator.h"
19
#include "querytask.h"
20
#include "tdatablock.h"
21

22
typedef struct SSortOpGroupIdCalc {
23
  STupleHandle* pSavedTuple;
24
  SArray*       pSortColsArr;
25
  char*         keyBuf;
26
  int32_t       lastKeysLen; // default to be 0
27
  uint64_t      lastGroupId;
28
  bool          excludePKCol;
29
} SSortOpGroupIdCalc;
30

31
typedef struct SSortOperatorInfo {
32
  SOptrBasicInfo      binfo;
33
  uint32_t            sortBufSize;  // max buffer size for in-memory sort
34
  SArray*             pSortInfo;
35
  SSortHandle*        pSortHandle;
36
  SColMatchInfo       matchInfo;
37
  int32_t             bufPageSize;
38
  int64_t             startTs;      // sort start time
39
  uint64_t            sortElapsed;  // sort elapsed time, time to flush to disk not included.
40
  SLimitInfo          limitInfo;
41
  uint64_t            maxTupleLength;
42
  int64_t             maxRows;
43
  SSortOpGroupIdCalc* pGroupIdCalc;
44
} SSortOperatorInfo;
45

46
static int32_t doSort(SOperatorInfo* pOperator, SSDataBlock** pResBlock);
47
static int32_t doOpenSortOperator(SOperatorInfo* pOperator);
48
static int32_t getExplainExecInfo(SOperatorInfo* pOptr, void** pOptrExplain, uint32_t* len);
49
static int32_t doGroupSort(SOperatorInfo* pOperator, SSDataBlock** pResBlock);
50

51
static void destroySortOperatorInfo(void* param);
52
static void calcSortOperMaxTupleLength(SSortOperatorInfo* pSortOperInfo, SNodeList* pSortKeys);
53

54
static void destroySortOpGroupIdCalc(SSortOpGroupIdCalc* pCalc);
55

56
// todo add limit/offset impl
57
int32_t createSortOperatorInfo(SOperatorInfo* downstream, SSortPhysiNode* pSortNode, SExecTaskInfo* pTaskInfo, SOperatorInfo** pOptrInfo) {
186,744✔
58
  QRY_PARAM_CHECK(pOptrInfo);
186,744!
59

60
  int32_t code = 0;
186,744✔
61
  int32_t lino = 0;
186,744✔
62

63
  SSortOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SSortOperatorInfo));
186,744!
64
  SOperatorInfo*     pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
186,825!
65
  if (pInfo == NULL || pOperator == NULL) {
186,831!
66
    code = terrno;
×
67
    goto _error;
×
68
  }
69

70
  pOperator->pTaskInfo = pTaskInfo;
186,835✔
71
  SDataBlockDescNode* pDescNode = pSortNode->node.pOutputDataBlockDesc;
186,835✔
72

73
  int32_t numOfCols = 0;
186,835✔
74
  code = createExprInfo(pSortNode->pExprs, NULL, &pOperator->exprSupp.pExprInfo, &numOfCols);
186,835✔
75
  QUERY_CHECK_CODE(code, lino, _error);
186,828!
76

77
  pOperator->exprSupp.numOfExprs = numOfCols;
186,828✔
78
  int32_t numOfOutputCols = 0;
186,828✔
79
  code =
80
      extractColMatchInfo(pSortNode->pTargets, pDescNode, &numOfOutputCols, COL_MATCH_FROM_SLOT_ID, &pInfo->matchInfo);
186,828✔
81
  if (code != TSDB_CODE_SUCCESS) {
186,813!
82
    goto _error;
×
83
  }
84
  
85
  calcSortOperMaxTupleLength(pInfo, pSortNode->pSortKeys);
186,813✔
86
  pInfo->maxRows = -1;
186,822✔
87
  if (pSortNode->node.pLimit && ((SLimitNode*)pSortNode->node.pLimit)->limit) {
186,822✔
88
    SLimitNode* pLimit = (SLimitNode*)pSortNode->node.pLimit;
27,763✔
89
    if (pLimit->limit->datum.i > 0) {
27,763✔
90
      pInfo->maxRows = pLimit->limit->datum.i + (pLimit->offset ? pLimit->offset->datum.i : 0);
27,540✔
91
    }
92
  }
93

94
  pOperator->exprSupp.pCtx =
186,810✔
95
      createSqlFunctionCtx(pOperator->exprSupp.pExprInfo, numOfCols, &pOperator->exprSupp.rowEntryInfoOffset, &pTaskInfo->storageAPI.functionStore);
186,822✔
96
  QUERY_CHECK_NULL(pOperator->exprSupp.pCtx, code, lino, _error, terrno);
186,810!
97
  initResultSizeInfo(&pOperator->resultInfo, 1024);
186,810✔
98
  code = filterInitFromNode((SNode*)pSortNode->node.pConditions, &pOperator->exprSupp.pFilterInfo, 0);
186,784✔
99
  if (code != TSDB_CODE_SUCCESS) {
186,801!
100
    goto _error;
×
101
  }
102

103
  pInfo->binfo.pRes = createDataBlockFromDescNode(pDescNode);
186,801✔
104
  QUERY_CHECK_NULL(pInfo->binfo.pRes, code, lino, _error, terrno);
186,844!
105

106
  pInfo->pSortInfo = createSortInfo(pSortNode->pSortKeys);
186,844✔
107
  TSDB_CHECK_NULL(pInfo->pSortInfo, code, lino, _error, terrno);
186,844✔
108

109
  if (pSortNode->calcGroupId) {
186,759✔
110
    int32_t keyLen;
111
    SSortOpGroupIdCalc* pGroupIdCalc = pInfo->pGroupIdCalc = taosMemoryCalloc(1, sizeof(SSortOpGroupIdCalc));
944!
112
    if (!pGroupIdCalc) {
944!
113
      code = terrno;
×
114
      goto _error;
×
115
    }
116
    SNodeList* pSortColsNodeArr = makeColsNodeArrFromSortKeys(pSortNode->pSortKeys);
944✔
117
    if (!pSortColsNodeArr) code = terrno;
944!
118
    if (TSDB_CODE_SUCCESS == code) {
944!
119
      pGroupIdCalc->pSortColsArr = makeColumnArrayFromList(pSortColsNodeArr);
944✔
120
      if (!pGroupIdCalc->pSortColsArr) code = terrno;
944!
121
      nodesClearList(pSortColsNodeArr);
944✔
122
    }
123
    if (TSDB_CODE_SUCCESS == code) {
944!
124
      // PK ts col should always at last, see partColOptCreateSort
125
      if (pSortNode->excludePkCol) taosArrayPop(pGroupIdCalc->pSortColsArr);
944✔
126
      code = extractKeysLen(pGroupIdCalc->pSortColsArr, &keyLen);
944✔
127
      QUERY_CHECK_CODE(code, lino, _error);
944!
128
    }
129
    if (TSDB_CODE_SUCCESS == code) {
944✔
130
      pGroupIdCalc->lastKeysLen = 0;
943✔
131
      pGroupIdCalc->keyBuf = taosMemoryCalloc(1, keyLen);
943!
132
      if (!pGroupIdCalc->keyBuf) {
943!
133
        code = terrno;
×
134
      }
135
    }
136
  }
137
  if (code != TSDB_CODE_SUCCESS) goto _error;
186,759!
138

139
  pInfo->binfo.inputTsOrder = pSortNode->node.inputTsOrder;
186,759✔
140
  pInfo->binfo.outputTsOrder = pSortNode->node.outputTsOrder;
186,759✔
141
  initLimitInfo(pSortNode->node.pLimit, pSortNode->node.pSlimit, &pInfo->limitInfo);
186,759✔
142

143
  setOperatorInfo(pOperator, "SortOperator", QUERY_NODE_PHYSICAL_PLAN_SORT, true, OP_NOT_OPENED, pInfo, pTaskInfo);
186,834✔
144

145

146
  // lazy evaluation for the following parameter since the input datablock is not known till now.
147
  //  pInfo->bufPageSize  = rowSize < 1024 ? 1024 * 2 : rowSize * 2;
148
  //  there are headers, so pageSize = rowSize + header pInfo->sortBufSize  = pInfo->bufPageSize * 16;
149
  // TODO dynamic set the available sort buffer
150

151
  pOperator->fpSet =
152
      createOperatorFpSet(doOpenSortOperator, doSort, NULL, destroySortOperatorInfo, optrDefaultBufFn, getExplainExecInfo, optrDefaultGetNextExtFn, NULL);
186,823✔
153

154
  code = appendDownstream(pOperator, &downstream, 1);
186,797✔
155
  if (code != TSDB_CODE_SUCCESS) {
186,835!
156
    goto _error;
×
157
  }
158

159
  *pOptrInfo = pOperator;
186,835✔
160
  return TSDB_CODE_SUCCESS;
186,835✔
161

162
_error:
×
163
  if (pInfo != NULL) {
×
164
    destroySortOperatorInfo(pInfo);
×
165
  }
166
  destroyOperatorAndDownstreams(pOperator, &downstream, 1);
×
167
  pTaskInfo->code = code;
×
168
  return code;
×
169
}
170

171
int32_t appendOneRowToDataBlock(SSDataBlock* pBlock, STupleHandle* pTupleHandle) {
134,415,649✔
172
  int32_t code = 0;
134,415,649✔
173
  for (int32_t i = 0; i < taosArrayGetSize(pBlock->pDataBlock); ++i) {
683,171,666✔
174
    SColumnInfoData* pColInfo = taosArrayGet(pBlock->pDataBlock, i);
547,577,585✔
175
    if (pColInfo == NULL) {
545,735,196!
176
      return terrno;
×
177
    }
178

179
    bool isNull = tsortIsNullVal(pTupleHandle, i);
545,735,196✔
180
    if (isNull) {
542,582,495✔
181
      colDataSetNULL(pColInfo, pBlock->info.rows);
26,266,509✔
182
    } else {
183
      char* pData = NULL;
516,315,986✔
184
      tsortGetValue(pTupleHandle, i, (void**) &pData);
516,315,986✔
185

186
      if (pData != NULL) {
518,034,185✔
187
        code = colDataSetVal(pColInfo, pBlock->info.rows, pData, false);
517,716,927✔
188
        if (code) {
522,172,250!
189
          return code;
×
190
        }
191
      }
192
    }
193
  }
194

195
  pBlock->info.dataLoad = 1;
127,748,353✔
196

197
  SDataBlockInfo info = {0};
127,748,353✔
198
  tsortGetBlockInfo(pTupleHandle, &info);
127,748,353✔
199

200
  pBlock->info.scanFlag = info.scanFlag;
135,278,604✔
201
  pBlock->info.rows += 1;
135,278,604✔
202
  return code;
135,278,604✔
203
}
204

205
/**
206
 * @brief get next tuple with group id attached, here assume that all tuples are sorted by group keys
207
 * @param [in, out] pBlock the output block, the group id will be saved in it
208
 * @retval NULL if next group tuple arrived and this new group tuple will be saved in pInfo.pSavedTuple
209
 */
210
static int32_t nextTupleWithGroupId(SSortHandle* pHandle, SSortOperatorInfo* pInfo, SSDataBlock* pBlock,
21,863,539✔
211
                                    STupleHandle** pTupleHandle) {
212
  QRY_PARAM_CHECK(pTupleHandle);
21,863,539!
213

214
  int32_t       code = 0;
21,863,539✔
215
  STupleHandle* retTuple = pInfo->pGroupIdCalc->pSavedTuple;
21,863,539✔
216
  if (!retTuple) {
21,863,539✔
217
    code = tsortNextTuple(pHandle, &retTuple);
21,390,206✔
218
    if (code) {
21,385,154✔
219
      qError("failed to get next tuple, code:%s", tstrerror(code));
14,709!
220
      return code;
×
221
    }
222
  }
223

224
  if (retTuple) {
21,843,778!
225
    int32_t newGroup;
226
    if (pInfo->pGroupIdCalc->pSavedTuple) {
21,854,950✔
227
      newGroup = true;
479,755✔
228
      pInfo->pGroupIdCalc->pSavedTuple = NULL;
479,755✔
229
    } else {
230
      newGroup = tsortCompAndBuildKeys(pInfo->pGroupIdCalc->pSortColsArr, pInfo->pGroupIdCalc->keyBuf,
21,375,195✔
231
                                       &pInfo->pGroupIdCalc->lastKeysLen, retTuple);
21,375,195✔
232
    }
233

234
    bool emptyBlock = (pBlock->info.rows == 0);
21,804,850✔
235
    if (newGroup) {
21,804,850✔
236
      if (!emptyBlock) {
958,002✔
237
        // new group arrived, and we have already copied some tuples for cur group, save the new group tuple, return
238
        // NULL. Note that the keyBuf and lastKeysLen has been updated to new value
239
        pInfo->pGroupIdCalc->pSavedTuple = retTuple;
481,635✔
240
        retTuple = NULL;
481,635✔
241
      } else {
242
        // new group with empty block
243
        pInfo->pGroupIdCalc->lastGroupId = pBlock->info.id.groupId =
483,701✔
244
            calcGroupId(pInfo->pGroupIdCalc->keyBuf, pInfo->pGroupIdCalc->lastKeysLen);
476,367✔
245
      }
246
    } else {
247
      if (emptyBlock) {
20,846,848✔
248
        // new block but not new group, assign last group id to it
249
        pBlock->info.id.groupId = pInfo->pGroupIdCalc->lastGroupId;
6,787✔
250
      } else {
251
        // not new group and not empty block and ret NOT NULL, just return the tuple
252
      }
253
    }
254
  }
255

256
  *pTupleHandle = retTuple;
21,801,012✔
257
  return code;
21,801,012✔
258
}
259

260
static int32_t getSortedBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, int32_t capacity, SArray* pColMatchInfo,
848,574✔
261
                                SSortOperatorInfo* pInfo, SSDataBlock** pResBlock) {
262
  QRY_PARAM_CHECK(pResBlock);
848,574!
263
  blockDataCleanup(pDataBlock);
848,574✔
264

265
  int32_t       lino = 0;
843,550✔
266
  int32_t       code = 0;
843,550✔
267
  STupleHandle* pTupleHandle = NULL;
843,550✔
268
  SSDataBlock*  p = NULL;
843,550✔
269

270
  code = tsortGetSortedDataBlock(pHandle, &p);
843,550✔
271
  if (p == NULL || (code != 0)) {
845,594!
272
    return code;
33,682✔
273
  }
274

275
  code = blockDataEnsureCapacity(p, capacity);
811,912✔
276
  QUERY_CHECK_CODE(code, lino, _error);
813,289!
277

278
  while (1) {
279
    if (pInfo->pGroupIdCalc) {
49,536,786✔
280
      code = nextTupleWithGroupId(pHandle, pInfo, p, &pTupleHandle);
21,867,461✔
281
    } else {
282
      code = tsortNextTuple(pHandle, &pTupleHandle);
27,669,325✔
283
    }
284

285
    TSDB_CHECK_CODE(code, lino, _error);
49,467,074!
286
    if (pTupleHandle == NULL) {
49,467,074✔
287
      break;
783,669✔
288
    }
289

290
    code = appendOneRowToDataBlock(p, pTupleHandle);
48,683,405✔
291
    QUERY_CHECK_CODE(code, lino, _error);
48,751,979!
292

293
    if (p->info.rows >= capacity) {
48,751,979✔
294
      break;
28,482✔
295
    }
296
  }
297

298
  QUERY_CHECK_CODE(code, lino, _error);
812,151!
299

300
  if (p->info.rows > 0) {
812,151✔
301
    code = blockDataEnsureCapacity(pDataBlock, capacity);
665,462✔
302
    QUERY_CHECK_CODE(code, lino, _error);
664,436!
303

304
    // todo extract function to handle this
305
    int32_t numOfCols = taosArrayGetSize(pColMatchInfo);
664,436✔
306
    for (int32_t i = 0; i < numOfCols; ++i) {
2,801,228✔
307
      SColMatchItem* pmInfo = taosArrayGet(pColMatchInfo, i);
2,144,219✔
308
      QUERY_CHECK_NULL(pmInfo, code, lino, _error, terrno);
2,140,256!
309

310
      SColumnInfoData* pSrc = taosArrayGet(p->pDataBlock, pmInfo->srcSlotId);
2,140,256✔
311
      QUERY_CHECK_NULL(pSrc, code, lino, _error, terrno);
2,135,519!
312

313
      SColumnInfoData* pDst = taosArrayGet(pDataBlock->pDataBlock, pmInfo->dstSlotId);
2,135,519✔
314
      QUERY_CHECK_NULL(pDst, code, lino, _error, terrno);
2,131,092!
315

316
      code = colDataAssign(pDst, pSrc, p->info.rows, &pDataBlock->info);
2,131,092✔
317
      QUERY_CHECK_CODE(code, lino, _error);
2,137,743!
318
    }
319

320
    pDataBlock->info.dataLoad = 1;
657,009✔
321
    pDataBlock->info.rows = p->info.rows;
657,009✔
322
    pDataBlock->info.scanFlag = p->info.scanFlag;
657,009✔
323
    pDataBlock->info.id.groupId = p->info.id.groupId;
657,009✔
324
  }
325

326
  blockDataDestroy(p);
803,698✔
327
  *pResBlock = (pDataBlock->info.rows > 0) ? pDataBlock : NULL;
814,996✔
328
  return code;
814,996✔
329

330
  _error:
×
331
  qError("%s failed at line %d since %s", __func__, lino, tstrerror(code));
×
332

333
  blockDataDestroy(p);
×
334
  return code;
×
335
}
336

337
int32_t loadNextDataBlock(void* param, SSDataBlock** ppBlock) {
570,321✔
338
  SOperatorInfo* pOperator = (SOperatorInfo*)param;
570,321✔
339
  int32_t        code = pOperator->fpSet.getNextFn(pOperator, ppBlock);
570,321✔
340
  if (code) {
570,216!
341
    qError("failed to get next data block from upstream, %s code:%s", __func__, tstrerror(code));
×
342
  } else {
343
    code = blockDataCheck(*ppBlock);
570,216✔
344
    if (code) {
570,280!
345
      qError("failed to check block data, %s code:%s", __func__, tstrerror(code));
×
346
    }
347
  }
348
  return code;
570,277✔
349
}
350

351
// todo refactor: merged with fetch fp
352
void applyScalarFunction(SSDataBlock* pBlock, void* param) {
383,731✔
353
  SOperatorInfo*     pOperator = param;
383,731✔
354
  SSortOperatorInfo* pSort = pOperator->info;
383,731✔
355
  if (pOperator->exprSupp.pExprInfo != NULL && pOperator->exprSupp.numOfExprs > 0) {
383,731!
356
    int32_t code = projectApplyFunctions(pOperator->exprSupp.pExprInfo, pBlock, pBlock, pOperator->exprSupp.pCtx,
539✔
357
                                         pOperator->exprSupp.numOfExprs, NULL);
358
    if (code != TSDB_CODE_SUCCESS) {
539!
359
      T_LONG_JMP(pOperator->pTaskInfo->env, code);
×
360
    }
361
  }
362
}
383,731✔
363

364
int32_t doOpenSortOperator(SOperatorInfo* pOperator) {
844,577✔
365
  SSortOperatorInfo* pInfo = pOperator->info;
844,577✔
366
  SExecTaskInfo*     pTaskInfo = pOperator->pTaskInfo;
844,577✔
367
  int32_t            code = TSDB_CODE_SUCCESS;
844,577✔
368
  int32_t            lino = 0;
844,577✔
369
  SSortSource* pSource =NULL;
844,577✔
370

371
  if (OPTR_IS_OPENED(pOperator)) {
844,577✔
372
    return code;
658,066✔
373
  }
374

375
  pInfo->startTs = taosGetTimestampUs();
186,816✔
376
  //  pInfo->binfo.pRes is not equalled to the input datablock.
377
  pInfo->pSortHandle = NULL;
186,816✔
378
  code =
379
      tsortCreateSortHandle(pInfo->pSortInfo, SORT_SINGLESOURCE_SORT, -1, -1, NULL, pTaskInfo->id.str, pInfo->maxRows,
186,816✔
380
                            pInfo->maxTupleLength, tsPQSortMemThreshold * 1024 * 1024, &pInfo->pSortHandle);
186,816✔
381
  QUERY_CHECK_CODE(code, lino, _end);
186,809!
382

383
  tsortSetFetchRawDataFp(pInfo->pSortHandle, loadNextDataBlock, applyScalarFunction, pOperator);
186,809✔
384

385
  pSource = taosMemoryCalloc(1, sizeof(SSortSource));
186,779!
386
  QUERY_CHECK_NULL(pSource, code, lino, _end, terrno);
186,869!
387

388
  pSource->param = pOperator->pDownstream[0];
186,869✔
389
  pSource->onlyRef = true;
186,869✔
390

391
  code = tsortAddSource(pInfo->pSortHandle, pSource);
186,869✔
392
  QUERY_CHECK_CODE(code, lino, _end);
186,762!
393
  pSource = NULL;
186,762✔
394

395
  code = tsortOpen(pInfo->pSortHandle);
186,762✔
396
  QUERY_CHECK_CODE(code, lino, _end);
186,859!
397
  pOperator->cost.openCost = (taosGetTimestampUs() - pInfo->startTs) / 1000.0;
186,873✔
398
  pOperator->status = OP_RES_TO_RETURN;
186,873✔
399
  OPTR_SET_OPENED(pOperator);
186,873✔
400

401
_end:
186,873✔
402
  if (pSource) {
186,873!
403
    taosMemoryFree(pSource);
×
404
  }
405
  if (code != TSDB_CODE_SUCCESS) {
186,867!
406
    qError("%s failed at line %d since %s", __func__, lino, tstrerror(code));
×
407
    pTaskInfo->code = code;
×
408
    T_LONG_JMP(pTaskInfo->env, code);
×
409
  }
410
  return code;
186,867✔
411
}
412

413
int32_t doSort(SOperatorInfo* pOperator, SSDataBlock** pResBlock) {
844,832✔
414
  QRY_PARAM_CHECK(pResBlock);
844,832!
415
  int32_t code = TSDB_CODE_SUCCESS;
844,832✔
416
  int32_t lino = 0;
844,832✔
417
  if (pOperator->status == OP_EXEC_DONE) {
844,832!
UNCOV
418
    return code;
×
419
  }
420

421
  SExecTaskInfo*     pTaskInfo = pOperator->pTaskInfo;
844,832✔
422
  SSortOperatorInfo* pInfo = pOperator->info;
844,832✔
423

424
  code = pOperator->fpSet._openFn(pOperator);
844,832✔
425
  QUERY_CHECK_CODE(code, lino, _end);
844,645!
426

427
  // multi-group case not handle here
428
  SSDataBlock* pBlock = NULL;
844,645✔
429
  while (1) {
729✔
430
    if (tsortIsClosed(pInfo->pSortHandle)) {
845,374!
431
      code = TSDB_CODE_TSC_QUERY_CANCELLED;
×
432
      QUERY_CHECK_CODE(code, lino, _end);
×
433
    }
434

435
    code = getSortedBlockData(pInfo->pSortHandle, pInfo->binfo.pRes, pOperator->resultInfo.capacity,
849,206✔
436
                                pInfo->matchInfo.pList, pInfo, &pBlock);
437
    QUERY_CHECK_CODE(code, lino, _end);
847,895!
438
    if (pBlock == NULL) {
847,895✔
439
      setOperatorCompleted(pOperator);
182,629✔
440
      return code;
182,625✔
441
    }
442

443
    code = doFilter(pBlock, pOperator->exprSupp.pFilterInfo, &pInfo->matchInfo);
665,266✔
444
    QUERY_CHECK_CODE(code, lino, _end);
664,866!
445

446
    if (blockDataGetNumOfRows(pBlock) == 0) {
664,866!
447
      continue;
×
448
    }
449

450
    // there are bugs?
451
    bool limitReached = applyLimitOffset(&pInfo->limitInfo, pBlock, pTaskInfo);
664,428✔
452
    if (limitReached) {
664,037✔
453
      resetLimitInfoForNextGroup(&pInfo->limitInfo);
14,318✔
454
    }
455

456
    pOperator->resultInfo.totalRows += pBlock->info.rows;
664,139✔
457
    if (pBlock->info.rows > 0) {
664,139✔
458
      break;
663,410✔
459
    }
460
  }
461

462
  *pResBlock = blockDataGetNumOfRows(pBlock) > 0 ? pBlock : NULL;
663,410!
463
_end:
663,363✔
464
  if (code != TSDB_CODE_SUCCESS) {
663,363!
465
    qError("%s failed at line %d since %s", __func__, lino, tstrerror(code));
×
466
    pTaskInfo->code = code;
×
467
    T_LONG_JMP(pTaskInfo->env, code);
×
468
  }
469
  return code;
663,363✔
470
}
471

472
void destroySortOperatorInfo(void* param) {
186,872✔
473
  SSortOperatorInfo* pInfo = (SSortOperatorInfo*)param;
186,872✔
474
  blockDataDestroy(pInfo->binfo.pRes);
186,872✔
475
  pInfo->binfo.pRes = NULL;
186,882✔
476

477
  tsortDestroySortHandle(pInfo->pSortHandle);
186,882✔
478
  taosArrayDestroy(pInfo->pSortInfo);
186,881✔
479
  taosArrayDestroy(pInfo->matchInfo.pList);
186,881✔
480
  destroySortOpGroupIdCalc(pInfo->pGroupIdCalc);
186,884✔
481
  taosMemoryFreeClear(param);
186,873!
482
}
186,879✔
483

484
int32_t getExplainExecInfo(SOperatorInfo* pOptr, void** pOptrExplain, uint32_t* len) {
203✔
485
  SSortExecInfo* pInfo = taosMemoryCalloc(1, sizeof(SSortExecInfo));
203!
486
  if (pInfo == NULL) {
203!
487
    return terrno;
×
488
  }
489

490
  SSortOperatorInfo* pOperatorInfo = (SSortOperatorInfo*)pOptr->info;
203✔
491

492
  *pInfo = tsortGetSortExecInfo(pOperatorInfo->pSortHandle);
203✔
493
  *pOptrExplain = pInfo;
203✔
494
  *len = sizeof(SSortExecInfo);
203✔
495
  return TSDB_CODE_SUCCESS;
203✔
496
}
497

498
static void calcSortOperMaxTupleLength(SSortOperatorInfo* pSortOperInfo, SNodeList* pSortKeys) {
186,738✔
499
  SColMatchInfo* pColItem = &pSortOperInfo->matchInfo;
186,738✔
500
  size_t         size = taosArrayGetSize(pColItem->pList);
186,738✔
501
  for (size_t i = 0; i < size; ++i) {
715,126✔
502
    SColMatchItem* pInfo = taosArrayGet(pColItem->pList, i);
528,317✔
503
    if (pInfo == NULL) {
528,340!
504
      continue;
×
505
    }
506

507
    pSortOperInfo->maxTupleLength += pInfo->dataType.bytes;
528,340✔
508
  }
509

510
  size = LIST_LENGTH(pSortKeys);
186,809✔
511
  for (size_t i = 0; i < size; ++i) {
437,706✔
512
    SOrderByExprNode* pOrderExprNode = (SOrderByExprNode*)nodesListGetNode(pSortKeys, i);
250,874✔
513
    pSortOperInfo->maxTupleLength += ((SColumnNode*)pOrderExprNode->pExpr)->node.resType.bytes;
250,897✔
514
  }
515
}
186,832✔
516

517
static void destroySortOpGroupIdCalc(SSortOpGroupIdCalc* pCalc) {
186,878✔
518
  if (pCalc) {
186,878✔
519
    taosArrayDestroy(pCalc->pSortColsArr);
944✔
520
    taosMemoryFree(pCalc->keyBuf);
943!
521
    taosMemoryFree(pCalc);
943!
522
  }
523
}
186,877✔
524

525
//=====================================================================================
526
// Group Sort Operator
527
typedef enum EChildOperatorStatus { CHILD_OP_NEW_GROUP, CHILD_OP_SAME_GROUP, CHILD_OP_FINISHED } EChildOperatorStatus;
528

529
typedef struct SGroupSortOperatorInfo {
530
  SOptrBasicInfo       binfo;
531
  SArray*              pSortInfo;
532
  SColMatchInfo        matchInfo;
533
  int64_t              startTs;
534
  uint64_t             sortElapsed;
535
  bool                 hasGroupId;
536
  uint64_t             currGroupId;
537
  SSDataBlock*         prefetchedSortInput;
538
  SSortHandle*         pCurrSortHandle;
539
  EChildOperatorStatus childOpStatus;
540
  SSortExecInfo        sortExecInfo;
541
} SGroupSortOperatorInfo;
542

543
int32_t getGroupSortedBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, int32_t capacity, SArray* pColMatchInfo,
360✔
544
                                SGroupSortOperatorInfo* pInfo, SSDataBlock** pResBlock) {
545
  QRY_PARAM_CHECK(pResBlock);
360!
546

547
  blockDataCleanup(pDataBlock);
360✔
548
  int32_t code = blockDataEnsureCapacity(pDataBlock, capacity);
360✔
549
  if (code) {
360!
550
    return code;
×
551
  }
552

553
  SSDataBlock* p = NULL;
360✔
554
  code = tsortGetSortedDataBlock(pHandle, &p);
360✔
555
  if (p == NULL || (code != 0)) {
360!
556
    return code;
×
557
  }
558

559
  code = blockDataEnsureCapacity(p, capacity);
360✔
560
  if (code) {
360!
561
    return code;
×
562
  }
563

564
  while (1) {
18,000✔
565
    STupleHandle* pTupleHandle = NULL;
18,360✔
566
    code = tsortNextTuple(pHandle, &pTupleHandle);
18,360✔
567
    if (pTupleHandle == NULL || code != 0) {
18,360!
568
      break;
569
    }
570

571
    code = appendOneRowToDataBlock(p, pTupleHandle);
18,000✔
572
    if (code) {
18,000!
573
      break;
×
574
    }
575

576
    if (p->info.rows >= capacity) {
18,000!
577
      break;
×
578
    }
579
  }
580

581
  if (p->info.rows > 0) {
360✔
582
    int32_t numOfCols = taosArrayGetSize(pColMatchInfo);
180✔
583
    for (int32_t i = 0; i < numOfCols; ++i) {
540✔
584
      SColMatchItem* pmInfo = taosArrayGet(pColMatchInfo, i);
360✔
585
      if (pmInfo == NULL) {
360!
586
        return terrno;
×
587
      }
588

589
      SColumnInfoData* pSrc = taosArrayGet(p->pDataBlock, pmInfo->srcSlotId);
360✔
590
      if (pSrc == NULL) {
360!
591
        return terrno;
×
592
      }
593

594
      SColumnInfoData* pDst = taosArrayGet(pDataBlock->pDataBlock, pmInfo->dstSlotId);
360✔
595
      if (pDst == NULL) {
360!
596
        return terrno;
×
597
      }
598

599
      code = colDataAssign(pDst, pSrc, p->info.rows, &pDataBlock->info);
360✔
600
      if (code) {
360!
601
        return code;
×
602
      }
603
    }
604

605
    pDataBlock->info.rows = p->info.rows;
180✔
606
    pDataBlock->info.capacity = p->info.rows;
180✔
607
    pDataBlock->info.scanFlag = p->info.scanFlag;
180✔
608
  }
609

610
  blockDataDestroy(p);
360✔
611
  *pResBlock = (pDataBlock->info.rows > 0) ? pDataBlock : NULL;
360✔
612
  return code;
360✔
613
}
614

615
typedef struct SGroupSortSourceParam {
616
  SOperatorInfo*          childOpInfo;
617
  SGroupSortOperatorInfo* grpSortOpInfo;
618
} SGroupSortSourceParam;
619

620
int32_t fetchNextGroupSortDataBlock(void* param, SSDataBlock** ppBlock) {
450✔
621
  int32_t                 code = 0;
450✔
622
  int32_t                 lino = 0;
450✔
623
  SGroupSortSourceParam*  source = param;
450✔
624
  SGroupSortOperatorInfo* grpSortOpInfo = source->grpSortOpInfo;
450✔
625
  SSDataBlock*            block = NULL;
450✔
626

627
  QRY_PARAM_CHECK(ppBlock);
450!
628

629
  if (grpSortOpInfo->prefetchedSortInput) {
450✔
630
    block = grpSortOpInfo->prefetchedSortInput;
180✔
631
    grpSortOpInfo->prefetchedSortInput = NULL;
180✔
632
    *ppBlock = block;
180✔
633
  } else {
634
    SOperatorInfo* childOp = source->childOpInfo;
270✔
635
    code = childOp->fpSet.getNextFn(childOp, &block);
270✔
636
    QUERY_CHECK_CODE(code, lino, _end);
270!
637

638
    if (block != NULL) {
270✔
639
      code = blockDataCheck(block);
90✔
640
      QUERY_CHECK_CODE(code, lino, _end);
90!
641
      if (block->info.id.groupId == grpSortOpInfo->currGroupId) {
90!
642
        grpSortOpInfo->childOpStatus = CHILD_OP_SAME_GROUP;
90✔
643
        *ppBlock = block;
90✔
644
      } else {
645
        grpSortOpInfo->childOpStatus = CHILD_OP_NEW_GROUP;
×
646
        grpSortOpInfo->prefetchedSortInput = block;
×
647
      }
648
    } else {
649
      grpSortOpInfo->childOpStatus = CHILD_OP_FINISHED;
180✔
650
    }
651
  }
652

653
  return code;
450✔
654
_end:
×
655
  if (code != 0) {
×
656
    qError("%s failed at line %d since %s", __func__, lino, tstrerror(code));
×
657
  }
658
  return code;
×
659
}
660

661
int32_t beginSortGroup(SOperatorInfo* pOperator) {
180✔
662
  SGroupSortOperatorInfo* pInfo = pOperator->info;
180✔
663
  SExecTaskInfo*          pTaskInfo = pOperator->pTaskInfo;
180✔
664

665
  //  pInfo->binfo.pRes is not equalled to the input datablock.
666
  pInfo->pCurrSortHandle = NULL;
180✔
667

668
  int32_t code = tsortCreateSortHandle(pInfo->pSortInfo, SORT_SINGLESOURCE_SORT, -1, -1, NULL, pTaskInfo->id.str, 0, 0,
180✔
669
                                       0, &pInfo->pCurrSortHandle);
670
  if (code) {
180!
671
    return code;
×
672
  }
673

674
  tsortSetFetchRawDataFp(pInfo->pCurrSortHandle, fetchNextGroupSortDataBlock, applyScalarFunction, pOperator);
180✔
675

676
  SSortSource*           ps = taosMemoryCalloc(1, sizeof(SSortSource));
180!
677
  SGroupSortSourceParam* param = taosMemoryCalloc(1, sizeof(SGroupSortSourceParam));
180!
678
  if (ps == NULL || param == NULL) {
180!
679
    taosMemoryFree(ps);
×
680
    taosMemoryFree(param);
×
681
    return terrno;
×
682
  }
683

684
  param->childOpInfo = pOperator->pDownstream[0];
180✔
685
  param->grpSortOpInfo = pInfo;
180✔
686

687
  ps->param = param;
180✔
688
  ps->onlyRef = false;
180✔
689
  code = tsortAddSource(pInfo->pCurrSortHandle, ps);
180✔
690
  if (code != 0) {
180!
691
    return code;
×
692
  }
693

694
  code = tsortOpen(pInfo->pCurrSortHandle);
180✔
695
  return code;
180✔
696
}
697

698
int32_t finishSortGroup(SOperatorInfo* pOperator) {
180✔
699
  SGroupSortOperatorInfo* pInfo = pOperator->info;
180✔
700

701
  SSortExecInfo sortExecInfo = tsortGetSortExecInfo(pInfo->pCurrSortHandle);
180✔
702

703
  pInfo->sortExecInfo.sortMethod = sortExecInfo.sortMethod;
180✔
704
  pInfo->sortExecInfo.sortBuffer = sortExecInfo.sortBuffer;
180✔
705
  pInfo->sortExecInfo.loops += sortExecInfo.loops;
180✔
706
  pInfo->sortExecInfo.readBytes += sortExecInfo.readBytes;
180✔
707
  pInfo->sortExecInfo.writeBytes += sortExecInfo.writeBytes;
180✔
708

709
  tsortDestroySortHandle(pInfo->pCurrSortHandle);
180✔
710
  pInfo->pCurrSortHandle = NULL;
180✔
711

712
  return TSDB_CODE_SUCCESS;
180✔
713
}
714

715
int32_t doGroupSort(SOperatorInfo* pOperator, SSDataBlock** pResBlock) {
385✔
716
  QRY_PARAM_CHECK(pResBlock);
385!
717
  SExecTaskInfo*          pTaskInfo = pOperator->pTaskInfo;
385✔
718
  SGroupSortOperatorInfo* pInfo = pOperator->info;
385✔
719
  int32_t                 code = TSDB_CODE_SUCCESS;
385✔
720
  int32_t                 lino = 0;
385✔
721

722
  if (pOperator->status == OP_EXEC_DONE) {
385!
723
    return code;
×
724
  }
725

726
  code = pOperator->fpSet._openFn(pOperator);
385✔
727
  QUERY_CHECK_CODE(code, lino, _end);
385!
728

729
  if (!pInfo->hasGroupId) {
385✔
730
    pInfo->hasGroupId = true;
205✔
731

732
    pInfo->prefetchedSortInput = getNextBlockFromDownstream(pOperator, 0);
205✔
733
    if (pInfo->prefetchedSortInput == NULL) {
205✔
734
      setOperatorCompleted(pOperator);
25✔
735
      return code;
25✔
736
    }
737

738
    pInfo->currGroupId = pInfo->prefetchedSortInput->info.id.groupId;
180✔
739
    pInfo->childOpStatus = CHILD_OP_NEW_GROUP;
180✔
740
    code = beginSortGroup(pOperator);
180✔
741
    QUERY_CHECK_CODE(code, lino, _end);
180!
742
  }
743

744
  SSDataBlock* pBlock = NULL;
360✔
745
  while (pInfo->pCurrSortHandle != NULL) {
360!
746
    if (tsortIsClosed(pInfo->pCurrSortHandle)) {
360!
747
      code = TSDB_CODE_TSC_QUERY_CANCELLED;
×
748
      QUERY_CHECK_CODE(code, lino, _end);
×
749
    }
750

751
    // beginSortGroup would fetch all child blocks of pInfo->currGroupId;
752
    if (pInfo->childOpStatus == CHILD_OP_SAME_GROUP) {
360!
753
      code = TSDB_CODE_QRY_EXECUTOR_INTERNAL_ERROR;
×
754
      QUERY_CHECK_CODE(code, lino, _end);
×
755
    }
756

757
    code = getGroupSortedBlockData(pInfo->pCurrSortHandle, pInfo->binfo.pRes, pOperator->resultInfo.capacity,
360✔
758
                                     pInfo->matchInfo.pList, pInfo, &pBlock);
759
    QUERY_CHECK_CODE(code, lino, _end);
360!
760
    if (pBlock != NULL) {
360✔
761
      pBlock->info.id.groupId = pInfo->currGroupId;
180✔
762
      pOperator->resultInfo.totalRows += pBlock->info.rows;
180✔
763
      *pResBlock = pBlock;
180✔
764
      return code;
180✔
765
    } else {
766
      if (pInfo->childOpStatus == CHILD_OP_NEW_GROUP) {
180!
767
        (void) finishSortGroup(pOperator);
×
768
        pInfo->currGroupId = pInfo->prefetchedSortInput->info.id.groupId;
×
769
        code = beginSortGroup(pOperator);
×
770
        QUERY_CHECK_CODE(code, lino, _end);
×
771
      } else if (pInfo->childOpStatus == CHILD_OP_FINISHED) {
180!
772
        (void) finishSortGroup(pOperator);
180✔
773
        setOperatorCompleted(pOperator);
180✔
774
        return code;
180✔
775
      }
776
    }
777
  }
778

779
_end:
×
780
  if (code != TSDB_CODE_SUCCESS) {
×
781
    qError("%s failed at line %d since %s", __func__, lino, tstrerror(code));
×
782
    pTaskInfo->code = code;
×
783
    T_LONG_JMP(pTaskInfo->env, code);
×
784
  }
785
  return code;
×
786
}
787

788
int32_t getGroupSortExplainExecInfo(SOperatorInfo* pOptr, void** pOptrExplain, uint32_t* len) {
×
789
  SGroupSortOperatorInfo* pInfo = (SGroupSortOperatorInfo*)pOptr->info;
×
790
  *pOptrExplain = &pInfo->sortExecInfo;
×
791
  *len = sizeof(SSortExecInfo);
×
792
  return TSDB_CODE_SUCCESS;
×
793
}
794

795
void destroyGroupSortOperatorInfo(void* param) {
205✔
796
  SGroupSortOperatorInfo* pInfo = (SGroupSortOperatorInfo*)param;
205✔
797
  blockDataDestroy(pInfo->binfo.pRes);
205✔
798
  pInfo->binfo.pRes = NULL;
205✔
799

800
  taosArrayDestroy(pInfo->pSortInfo);
205✔
801
  taosArrayDestroy(pInfo->matchInfo.pList);
205✔
802

803
  tsortDestroySortHandle(pInfo->pCurrSortHandle);
205✔
804
  pInfo->pCurrSortHandle = NULL;
205✔
805

806
  taosMemoryFreeClear(param);
205!
807
}
205✔
808

809
int32_t createGroupSortOperatorInfo(SOperatorInfo* downstream, SGroupSortPhysiNode* pSortPhyNode,
205✔
810
                                    SExecTaskInfo* pTaskInfo, SOperatorInfo** pOptrInfo) {
811
  QRY_PARAM_CHECK(pOptrInfo);
205!
812
  int32_t code = 0;
205✔
813
  int32_t lino = 0;
205✔
814

815
  SGroupSortOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SGroupSortOperatorInfo));
205!
816
  SOperatorInfo*          pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
205!
817
  if (pInfo == NULL || pOperator == NULL) {
205!
818
    code = terrno;
×
819
    goto _error;
×
820
  }
821

822
  SExprSupp*          pSup = &pOperator->exprSupp;
205✔
823
  SDataBlockDescNode* pDescNode = pSortPhyNode->node.pOutputDataBlockDesc;
205✔
824

825
  int32_t    numOfCols = 0;
205✔
826
  SExprInfo* pExprInfo = NULL;
205✔
827
  code = createExprInfo(pSortPhyNode->pExprs, NULL, &pExprInfo, &numOfCols);
205✔
828
  QUERY_CHECK_CODE(code, lino, _error);
205!
829

830
  pSup->pExprInfo = pExprInfo;
205✔
831
  pSup->numOfExprs = numOfCols;
205✔
832

833
  initResultSizeInfo(&pOperator->resultInfo, 1024);
205✔
834
  pOperator->exprSupp.pCtx = createSqlFunctionCtx(pExprInfo, numOfCols, &pOperator->exprSupp.rowEntryInfoOffset,
205✔
835
                                                  &pTaskInfo->storageAPI.functionStore);
836
  QUERY_CHECK_NULL(pOperator->exprSupp.pCtx, code, lino, _error, terrno);
205!
837

838
  pInfo->binfo.pRes = createDataBlockFromDescNode(pDescNode);
205✔
839
  QUERY_CHECK_NULL(pInfo->binfo.pRes, code, lino, _error, terrno);
205!
840

841
  code = blockDataEnsureCapacity(pInfo->binfo.pRes, pOperator->resultInfo.capacity);
205✔
842
  TSDB_CHECK_CODE(code, lino, _error);
205!
843

844
  pInfo->binfo.inputTsOrder = pSortPhyNode->node.inputTsOrder;
205✔
845
  pInfo->binfo.outputTsOrder = pSortPhyNode->node.outputTsOrder;
205✔
846

847
  int32_t numOfOutputCols = 0;
205✔
848
  code = extractColMatchInfo(pSortPhyNode->pTargets, pDescNode, &numOfOutputCols, COL_MATCH_FROM_SLOT_ID,
205✔
849
                             &pInfo->matchInfo);
850
  TSDB_CHECK_CODE(code, lino, _error);
205!
851

852
  pInfo->pSortInfo = createSortInfo(pSortPhyNode->pSortKeys);
205✔
853
  setOperatorInfo(pOperator, "GroupSortOperator", QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT, false, OP_NOT_OPENED, pInfo,
205✔
854
                  pTaskInfo);
855
  pOperator->fpSet = createOperatorFpSet(optrDummyOpenFn, doGroupSort, NULL, destroyGroupSortOperatorInfo,
205✔
856
                                         optrDefaultBufFn, getGroupSortExplainExecInfo, optrDefaultGetNextExtFn, NULL);
857

858
  code = appendDownstream(pOperator, &downstream, 1);
205✔
859
  if (code != TSDB_CODE_SUCCESS) {
205!
860
    goto _error;
×
861
  }
862

863
  *pOptrInfo = pOperator;
205✔
864
  return TSDB_CODE_SUCCESS;
205✔
865

866
_error:
×
867
  pTaskInfo->code = code;
×
868
  if (pInfo != NULL) {
×
869
    destroyGroupSortOperatorInfo(pInfo);
×
870
  }
871
  destroyOperatorAndDownstreams(pOperator, &downstream, 1);
×
872
  return code;
×
873
}
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc