projectoperator.c 21.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
/*
 * 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/>.
 */

#include "executorimpl.h"
#include "functionMgt.h"

static SSDataBlock* doGenerateSourceData(SOperatorInfo* pOperator);
static SSDataBlock* doProjectOperation(SOperatorInfo* pOperator);
static SSDataBlock* doApplyIndefinitFunction(SOperatorInfo* pOperator);
static SArray*      setRowTsColumnOutputInfo(SqlFunctionCtx* pCtx, int32_t numOfCols);
static void setFunctionResultOutput(SOperatorInfo* pOperator, SOptrBasicInfo* pInfo, SAggSupporter* pSup, int32_t stage,
                                    int32_t numOfExprs);

26
static void destroyProjectOperatorInfo(void* param) {
27 28 29 30 31 32 33 34 35 36 37 38 39
  if (NULL == param) {
    return;
  }

  SProjectOperatorInfo* pInfo = (SProjectOperatorInfo*)param;
  cleanupBasicInfo(&pInfo->binfo);
  cleanupAggSup(&pInfo->aggSup);
  taosArrayDestroy(pInfo->pPseudoColInfo);

  blockDataDestroy(pInfo->pFinalRes);
  taosMemoryFreeClear(param);
}

40
static void destroyIndefinitOperatorInfo(void* param) {
41
  SIndefOperatorInfo* pInfo = (SIndefOperatorInfo*)param;
42 43 44
  if (pInfo == NULL) {
    return;
  }
45

46
  cleanupBasicInfo(&pInfo->binfo);
47 48 49 50 51 52 53 54 55
  taosArrayDestroy(pInfo->pPseudoColInfo);
  cleanupAggSup(&pInfo->aggSup);
  cleanupExprSupp(&pInfo->scalarSup);

  taosMemoryFreeClear(param);
}

SOperatorInfo* createProjectOperatorInfo(SOperatorInfo* downstream, SProjectPhysiNode* pProjPhyNode,
                                         SExecTaskInfo* pTaskInfo) {
56
  int32_t code = TSDB_CODE_SUCCESS;
57 58 59
  SProjectOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SProjectOperatorInfo));
  SOperatorInfo*        pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
  if (pInfo == NULL || pOperator == NULL) {
60
    code = TSDB_CODE_OUT_OF_MEMORY;
61 62 63
    goto _error;
  }

64 65
  pOperator->pTaskInfo = pTaskInfo;

66 67 68 69 70 71 72
  int32_t    numOfCols = 0;
  SExprInfo* pExprInfo = createExprInfo(pProjPhyNode->pProjections, NULL, &numOfCols);

  SSDataBlock* pResBlock = createResDataBlock(pProjPhyNode->node.pOutputDataBlockDesc);
  initLimitInfo(pProjPhyNode->node.pLimit, pProjPhyNode->node.pSlimit, &pInfo->limitInfo);

  pInfo->binfo.pRes = pResBlock;
73
  pInfo->pFinalRes = createOneDataBlock(pResBlock, false);
74
  pInfo->pFilterNode = pProjPhyNode->node.pConditions;
J
jiajingbin 已提交
75

H
Haojun Liao 已提交
76
  if (pTaskInfo->execModel == OPTR_EXEC_MODEL_STREAM) {
77
    pInfo->mergeDataBlocks = false;
78 79
  } else {
    pInfo->mergeDataBlocks = pProjPhyNode->mergeDataBlock;
H
Haojun Liao 已提交
80 81
  }

82 83 84 85 86 87 88 89
  int32_t numOfRows = 4096;
  size_t  keyBufSize = sizeof(int64_t) + sizeof(int64_t) + POINTER_BYTES;

  // Make sure the size of SSDataBlock will never exceed the size of 2MB.
  int32_t TWOMB = 2 * 1024 * 1024;
  if (numOfRows * pResBlock->info.rowSize > TWOMB) {
    numOfRows = TWOMB / pResBlock->info.rowSize;
  }
90

91
  initResultSizeInfo(&pOperator->resultInfo, numOfRows);
92 93 94 95
  code = initAggInfo(&pOperator->exprSupp, &pInfo->aggSup, pExprInfo, numOfCols, keyBufSize, pTaskInfo->id.str);
  if (code != TSDB_CODE_SUCCESS) {
    goto _error;
  }
96 97 98 99 100

  initBasicInfo(&pInfo->binfo, pResBlock);
  setFunctionResultOutput(pOperator, &pInfo->binfo, &pInfo->aggSup, MAIN_SCAN, numOfCols);

  pInfo->pPseudoColInfo = setRowTsColumnOutputInfo(pOperator->exprSupp.pCtx, numOfCols);
101
  pOperator->name = "ProjectOperator";
102
  pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_PROJECT;
103 104 105
  pOperator->blocking = false;
  pOperator->status = OP_NOT_OPENED;
  pOperator->info = pInfo;
106 107 108 109

  pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doProjectOperation, NULL, NULL,
                                         destroyProjectOperatorInfo, NULL, NULL, NULL);

110
  code = appendDownstream(pOperator, &downstream, 1);
111 112 113 114 115 116
  if (code != TSDB_CODE_SUCCESS) {
    goto _error;
  }

  return pOperator;

117
_error:
118
  destroyProjectOperatorInfo(pInfo);
119 120
  taosMemoryFree(pOperator);
  pTaskInfo->code = code;
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150
  return NULL;
}

static int32_t discardGroupDataBlock(SSDataBlock* pBlock, SLimitInfo* pLimitInfo) {
  if (pLimitInfo->remainGroupOffset > 0) {
    // it is the first group
    if (pLimitInfo->currentGroupId == 0 || pLimitInfo->currentGroupId == pBlock->info.groupId) {
      pLimitInfo->currentGroupId = pBlock->info.groupId;
      return PROJECT_RETRIEVE_CONTINUE;
    } else if (pLimitInfo->currentGroupId != pBlock->info.groupId) {
      // now it is the data from a new group
      pLimitInfo->remainGroupOffset -= 1;
      pLimitInfo->currentGroupId = pBlock->info.groupId;

      // ignore data block in current group
      if (pLimitInfo->remainGroupOffset > 0) {
        return PROJECT_RETRIEVE_CONTINUE;
      }
    }

    // set current group id of the project operator
    pLimitInfo->currentGroupId = pBlock->info.groupId;
  }

  return PROJECT_RETRIEVE_DONE;
}

static int32_t setInfoForNewGroup(SSDataBlock* pBlock, SLimitInfo* pLimitInfo, SOperatorInfo* pOperator) {
  // remainGroupOffset == 0
  // here check for a new group data, we need to handle the data of the previous group.
H
Haojun Liao 已提交
151
  ASSERT(pLimitInfo->remainGroupOffset == 0 || pLimitInfo->remainGroupOffset == -1);
152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168

  if (pLimitInfo->currentGroupId != 0 && pLimitInfo->currentGroupId != pBlock->info.groupId) {
    pLimitInfo->numOfOutputGroups += 1;
    if ((pLimitInfo->slimit.limit > 0) && (pLimitInfo->slimit.limit <= pLimitInfo->numOfOutputGroups)) {
      doSetOperatorCompleted(pOperator);
      return PROJECT_RETRIEVE_DONE;
    }

    // reset the value for a new group data
    // existing rows that belongs to previous group.
    pLimitInfo->numOfOutputRows = 0;
    pLimitInfo->remainOffset = pLimitInfo->limit.offset;
  }

  return PROJECT_RETRIEVE_DONE;
}

169 170
static int32_t doIngroupLimitOffset(SLimitInfo* pLimitInfo, uint64_t groupId, SSDataBlock* pBlock,
                                    SOperatorInfo* pOperator) {
171 172 173 174 175 176 177 178 179 180 181 182 183
  // set current group id
  pLimitInfo->currentGroupId = groupId;

  if (pLimitInfo->remainOffset >= pBlock->info.rows) {
    pLimitInfo->remainOffset -= pBlock->info.rows;
    blockDataCleanup(pBlock);
    return PROJECT_RETRIEVE_CONTINUE;
  } else if (pLimitInfo->remainOffset < pBlock->info.rows && pLimitInfo->remainOffset > 0) {
    blockDataTrimFirstNRows(pBlock, pLimitInfo->remainOffset);
    pLimitInfo->remainOffset = 0;
  }

  // check for the limitation in each group
184
  if (pLimitInfo->limit.limit >= 0 && pLimitInfo->numOfOutputRows + pBlock->info.rows >= pLimitInfo->limit.limit) {
185 186
    int32_t keepRows = (int32_t)(pLimitInfo->limit.limit - pLimitInfo->numOfOutputRows);
    blockDataKeepFirstNRows(pBlock, keepRows);
187
    //TODO: optimize it later when partition by + limit
188 189
    if ((pLimitInfo->slimit.limit == -1 && pLimitInfo->currentGroupId == 0) ||
        (pLimitInfo->slimit.limit > 0 && pLimitInfo->slimit.limit <= pLimitInfo->numOfOutputGroups)) {
190 191 192 193 194 195 196 197
      doSetOperatorCompleted(pOperator);
    }
  }

  pLimitInfo->numOfOutputRows += pBlock->info.rows;
  return PROJECT_RETRIEVE_DONE;
}

J
jiajingbin 已提交
198 199 200 201 202 203 204
void printDataBlock1(SSDataBlock* pBlock, const char* flag) {
  if (!pBlock || pBlock->info.rows == 0) {
    qDebug("===stream===printDataBlock: Block is Null or Empty");
    return;
  }
  char* pBuf = NULL;
  qDebug("%s", dumpBlockData(pBlock, flag, &pBuf));
L
Liu Jicong 已提交
205
  taosMemoryFreeClear(pBuf);
J
jiajingbin 已提交
206 207
}

208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236
SSDataBlock* doProjectOperation(SOperatorInfo* pOperator) {
  SProjectOperatorInfo* pProjectInfo = pOperator->info;
  SOptrBasicInfo*       pInfo = &pProjectInfo->binfo;

  SExprSupp*   pSup = &pOperator->exprSupp;
  SSDataBlock* pRes = pInfo->pRes;
  SSDataBlock* pFinalRes = pProjectInfo->pFinalRes;

  blockDataCleanup(pFinalRes);

  SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
  if (pOperator->status == OP_EXEC_DONE) {
    if (pTaskInfo->execModel == OPTR_EXEC_MODEL_QUEUE) {
      pOperator->status = OP_OPENED;
      return NULL;
    }

    return NULL;
  }

  int64_t st = 0;
  int32_t order = 0;
  int32_t scanFlag = 0;

  if (pOperator->cost.openCost == 0) {
    st = taosGetTimestampUs();
  }

  SOperatorInfo* downstream = pOperator->pDownstream[0];
237
  SLimitInfo*    pLimitInfo = &pProjectInfo->limitInfo;
238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254

  if (downstream == NULL) {
    return doGenerateSourceData(pOperator);
  }

  while (1) {
    while (1) {
      blockDataCleanup(pRes);

      // The downstream exec may change the value of the newgroup, so use a local variable instead.
      SSDataBlock* pBlock = downstream->fpSet.getNextFn(downstream);
      if (pBlock == NULL) {
        doSetOperatorCompleted(pOperator);
        break;
      }

      // for stream interval
5
54liuyao 已提交
255 256
      if (pBlock->info.type == STREAM_RETRIEVE || pBlock->info.type == STREAM_DELETE_RESULT ||
          pBlock->info.type == STREAM_DELETE_DATA) {
L
Liu Jicong 已提交
257
        // printDataBlock1(pBlock, "project1");
258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273
        return pBlock;
      }

      int32_t status = discardGroupDataBlock(pBlock, pLimitInfo);
      if (status == PROJECT_RETRIEVE_CONTINUE) {
        continue;
      }

      setInfoForNewGroup(pBlock, pLimitInfo, pOperator);
      if (pOperator->status == OP_EXEC_DONE) {
        break;
      }

      // the pDataBlock are always the same one, no need to call this again
      int32_t code = getTableScanInfo(downstream, &order, &scanFlag);
      if (code != TSDB_CODE_SUCCESS) {
274
        T_LONG_JMP(pTaskInfo->env, code);
275 276 277 278 279 280 281 282
      }

      setInputDataBlock(pOperator, pSup->pCtx, pBlock, order, scanFlag, false);
      blockDataEnsureCapacity(pInfo->pRes, pInfo->pRes->info.rows + pBlock->info.rows);

      code = projectApplyFunctions(pSup->pExprInfo, pInfo->pRes, pBlock, pSup->pCtx, pSup->numOfExprs,
                                   pProjectInfo->pPseudoColInfo);
      if (code != TSDB_CODE_SUCCESS) {
283
        T_LONG_JMP(pTaskInfo->env, code);
284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307
      }

      status = doIngroupLimitOffset(pLimitInfo, pBlock->info.groupId, pInfo->pRes, pOperator);
      if (status == PROJECT_RETRIEVE_CONTINUE) {
        continue;
      }

      break;
    }

    if (pProjectInfo->mergeDataBlocks) {
      if (pRes->info.rows > 0) {
        pFinalRes->info.groupId = pRes->info.groupId;
        pFinalRes->info.version = pRes->info.version;

        // continue merge data, ignore the group id
        blockDataMerge(pFinalRes, pRes);
        if (pFinalRes->info.rows + pRes->info.rows <= pOperator->resultInfo.threshold) {
          continue;
        }
      }

      // do apply filter
      doFilter(pProjectInfo->pFilterNode, pFinalRes, NULL);
308 309 310

      // when apply the limit/offset for each group, pRes->info.rows may be 0, due to limit constraint.
      if (pFinalRes->info.rows > 0 || (pOperator->status == OP_EXEC_DONE)) {
311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
        break;
      }
    } else {
      // do apply filter
      if (pRes->info.rows > 0) {
        doFilter(pProjectInfo->pFilterNode, pRes, NULL);
        if (pRes->info.rows == 0) {
          continue;
        }
      }

      // no results generated
      break;
    }
  }

  SSDataBlock* p = pProjectInfo->mergeDataBlocks ? pFinalRes : pRes;
  pOperator->resultInfo.totalRows += p->info.rows;

  if (pOperator->cost.openCost == 0) {
    pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
  }
333

L
Liu Jicong 已提交
334
  // printDataBlock1(p, "project");
335 336 337 338 339 340 341 342 343 344 345
  return (p->info.rows > 0) ? p : NULL;
}

SOperatorInfo* createIndefinitOutputOperatorInfo(SOperatorInfo* downstream, SPhysiNode* pNode,
                                                 SExecTaskInfo* pTaskInfo) {
  SIndefOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SIndefOperatorInfo));
  SOperatorInfo*      pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
  if (pInfo == NULL || pOperator == NULL) {
    goto _error;
  }

346 347
  pOperator->pTaskInfo = pTaskInfo;

348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376
  SExprSupp* pSup = &pOperator->exprSupp;

  SIndefRowsFuncPhysiNode* pPhyNode = (SIndefRowsFuncPhysiNode*)pNode;

  int32_t    numOfExpr = 0;
  SExprInfo* pExprInfo = createExprInfo(pPhyNode->pFuncs, NULL, &numOfExpr);

  if (pPhyNode->pExprs != NULL) {
    int32_t    num = 0;
    SExprInfo* pSExpr = createExprInfo(pPhyNode->pExprs, NULL, &num);
    int32_t    code = initExprSupp(&pInfo->scalarSup, pSExpr, num);
    if (code != TSDB_CODE_SUCCESS) {
      goto _error;
    }
  }

  SSDataBlock* pResBlock = createResDataBlock(pPhyNode->node.pOutputDataBlockDesc);

  int32_t numOfRows = 4096;
  size_t  keyBufSize = sizeof(int64_t) + sizeof(int64_t) + POINTER_BYTES;

  // Make sure the size of SSDataBlock will never exceed the size of 2MB.
  int32_t TWOMB = 2 * 1024 * 1024;
  if (numOfRows * pResBlock->info.rowSize > TWOMB) {
    numOfRows = TWOMB / pResBlock->info.rowSize;
  }

  initResultSizeInfo(&pOperator->resultInfo, numOfRows);

377 378 379 380
  int32_t code = initAggInfo(pSup, &pInfo->aggSup, pExprInfo, numOfExpr, keyBufSize, pTaskInfo->id.str);
  if (code != TSDB_CODE_SUCCESS) {
    goto _error;
  }
381

382
  initBasicInfo(&pInfo->binfo, pResBlock);
383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
  setFunctionResultOutput(pOperator, &pInfo->binfo, &pInfo->aggSup, MAIN_SCAN, numOfExpr);

  pInfo->binfo.pRes = pResBlock;
  pInfo->pCondition = pPhyNode->node.pConditions;
  pInfo->pPseudoColInfo = setRowTsColumnOutputInfo(pSup->pCtx, numOfExpr);

  pOperator->name = "IndefinitOperator";
  pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC;
  pOperator->blocking = false;
  pOperator->status = OP_NOT_OPENED;
  pOperator->info = pInfo;

  pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doApplyIndefinitFunction, NULL, NULL,
                                         destroyIndefinitOperatorInfo, NULL, NULL, NULL);

398
  code = appendDownstream(pOperator, &downstream, 1);
399 400 401 402 403 404
  if (code != TSDB_CODE_SUCCESS) {
    goto _error;
  }

  return pOperator;

405
_error:
406
  destroyIndefinitOperatorInfo(pInfo);
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
  taosMemoryFree(pOperator);
  pTaskInfo->code = TSDB_CODE_OUT_OF_MEMORY;
  return NULL;
}

static void doHandleDataBlock(SOperatorInfo* pOperator, SSDataBlock* pBlock, SOperatorInfo* downstream,
                              SExecTaskInfo* pTaskInfo) {
  int32_t order = 0;
  int32_t scanFlag = 0;

  SIndefOperatorInfo* pIndefInfo = pOperator->info;
  SOptrBasicInfo*     pInfo = &pIndefInfo->binfo;
  SExprSupp*          pSup = &pOperator->exprSupp;

  // the pDataBlock are always the same one, no need to call this again
  int32_t code = getTableScanInfo(downstream, &order, &scanFlag);
  if (code != TSDB_CODE_SUCCESS) {
424
    T_LONG_JMP(pTaskInfo->env, code);
425 426 427 428 429 430 431 432
  }

  // there is an scalar expression that needs to be calculated before apply the group aggregation.
  SExprSupp* pScalarSup = &pIndefInfo->scalarSup;
  if (pScalarSup->pExprInfo != NULL) {
    code = projectApplyFunctions(pScalarSup->pExprInfo, pBlock, pBlock, pScalarSup->pCtx, pScalarSup->numOfExprs,
                                 pIndefInfo->pPseudoColInfo);
    if (code != TSDB_CODE_SUCCESS) {
433
      T_LONG_JMP(pTaskInfo->env, code);
434 435 436 437 438 439 440 441 442
    }
  }

  setInputDataBlock(pOperator, pSup->pCtx, pBlock, order, scanFlag, false);
  blockDataEnsureCapacity(pInfo->pRes, pInfo->pRes->info.rows + pBlock->info.rows);

  code = projectApplyFunctions(pSup->pExprInfo, pInfo->pRes, pBlock, pSup->pCtx, pSup->numOfExprs,
                               pIndefInfo->pPseudoColInfo);
  if (code != TSDB_CODE_SUCCESS) {
443
    T_LONG_JMP(pTaskInfo->env, code);
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612
  }
}

SSDataBlock* doApplyIndefinitFunction(SOperatorInfo* pOperator) {
  SIndefOperatorInfo* pIndefInfo = pOperator->info;
  SOptrBasicInfo*     pInfo = &pIndefInfo->binfo;
  SExprSupp*          pSup = &pOperator->exprSupp;

  SSDataBlock* pRes = pInfo->pRes;
  blockDataCleanup(pRes);

  SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
  if (pOperator->status == OP_EXEC_DONE) {
    return NULL;
  }

  int64_t st = 0;

  if (pOperator->cost.openCost == 0) {
    st = taosGetTimestampUs();
  }

  SOperatorInfo* downstream = pOperator->pDownstream[0];

  while (1) {
    // here we need to handle the existsed group results
    if (pIndefInfo->pNextGroupRes != NULL) {  // todo extract method
      for (int32_t k = 0; k < pSup->numOfExprs; ++k) {
        SqlFunctionCtx* pCtx = &pSup->pCtx[k];

        SResultRowEntryInfo* pResInfo = GET_RES_INFO(pCtx);
        pResInfo->initialized = false;
        pCtx->pOutput = NULL;
      }

      doHandleDataBlock(pOperator, pIndefInfo->pNextGroupRes, downstream, pTaskInfo);
      pIndefInfo->pNextGroupRes = NULL;
    }

    if (pInfo->pRes->info.rows < pOperator->resultInfo.threshold) {
      while (1) {
        // The downstream exec may change the value of the newgroup, so use a local variable instead.
        SSDataBlock* pBlock = downstream->fpSet.getNextFn(downstream);
        if (pBlock == NULL) {
          doSetOperatorCompleted(pOperator);
          break;
        }

        if (pIndefInfo->groupId == 0 && pBlock->info.groupId != 0) {
          pIndefInfo->groupId = pBlock->info.groupId;  // this is the initial group result
        } else {
          if (pIndefInfo->groupId != pBlock->info.groupId) {  // reset output buffer and computing status
            pIndefInfo->groupId = pBlock->info.groupId;
            pIndefInfo->pNextGroupRes = pBlock;
            break;
          }
        }

        doHandleDataBlock(pOperator, pBlock, downstream, pTaskInfo);
        if (pInfo->pRes->info.rows >= pOperator->resultInfo.threshold) {
          break;
        }
      }
    }

    doFilter(pIndefInfo->pCondition, pInfo->pRes, NULL);
    size_t rows = pInfo->pRes->info.rows;
    if (rows > 0 || pOperator->status == OP_EXEC_DONE) {
      break;
    } else {
      blockDataCleanup(pInfo->pRes);
    }
  }

  size_t rows = pInfo->pRes->info.rows;
  pOperator->resultInfo.totalRows += rows;

  if (pOperator->cost.openCost == 0) {
    pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
  }

  return (rows > 0) ? pInfo->pRes : NULL;
}

void initCtxOutputBuffer(SqlFunctionCtx* pCtx, int32_t size) {
  for (int32_t j = 0; j < size; ++j) {
    struct SResultRowEntryInfo* pResInfo = GET_RES_INFO(&pCtx[j]);
    if (isRowEntryInitialized(pResInfo) || fmIsPseudoColumnFunc(pCtx[j].functionId) || pCtx[j].functionId == -1 ||
        fmIsScalarFunc(pCtx[j].functionId)) {
      continue;
    }

    pCtx[j].fpSet.init(&pCtx[j], pCtx[j].resultInfo);
  }
}

/*
 * The start of each column SResultRowEntryInfo is denote by RowCellInfoOffset.
 * Note that in case of top/bottom query, the whole multiple rows of result is treated as only one row of results.
 * +------------+-----------------result column 1------------+------------------result column 2-----------+
 * | SResultRow | SResultRowEntryInfo | intermediate buffer1 | SResultRowEntryInfo | intermediate buffer 2|
 * +------------+--------------------------------------------+--------------------------------------------+
 *           offset[0]                                  offset[1]                                   offset[2]
 */
// TODO refactor: some function move away
void setFunctionResultOutput(SOperatorInfo* pOperator, SOptrBasicInfo* pInfo, SAggSupporter* pSup, int32_t stage,
                             int32_t numOfExprs) {
  SExecTaskInfo*  pTaskInfo = pOperator->pTaskInfo;
  SqlFunctionCtx* pCtx = pOperator->exprSupp.pCtx;
  int32_t*        rowEntryInfoOffset = pOperator->exprSupp.rowEntryInfoOffset;

  SResultRowInfo* pResultRowInfo = &pInfo->resultRowInfo;
  initResultRowInfo(pResultRowInfo);

  int64_t     tid = 0;
  int64_t     groupId = 0;
  SResultRow* pRow = doSetResultOutBufByKey(pSup->pResultBuf, pResultRowInfo, (char*)&tid, sizeof(tid), true, groupId,
                                            pTaskInfo, false, pSup);

  for (int32_t i = 0; i < numOfExprs; ++i) {
    struct SResultRowEntryInfo* pEntry = getResultEntryInfo(pRow, i, rowEntryInfoOffset);
    cleanupResultRowEntry(pEntry);

    pCtx[i].resultInfo = pEntry;
    pCtx[i].scanFlag = stage;
  }

  initCtxOutputBuffer(pCtx, numOfExprs);
}

SArray* setRowTsColumnOutputInfo(SqlFunctionCtx* pCtx, int32_t numOfCols) {
  SArray* pList = taosArrayInit(4, sizeof(int32_t));
  for (int32_t i = 0; i < numOfCols; ++i) {
    if (fmIsPseudoColumnFunc(pCtx[i].functionId)) {
      taosArrayPush(pList, &i);
    }
  }

  return pList;
}

SSDataBlock* doGenerateSourceData(SOperatorInfo* pOperator) {
  SProjectOperatorInfo* pProjectInfo = pOperator->info;

  SExprSupp*   pSup = &pOperator->exprSupp;
  SSDataBlock* pRes = pProjectInfo->binfo.pRes;

  blockDataEnsureCapacity(pRes, pOperator->resultInfo.capacity);
  SExprInfo* pExpr = pSup->pExprInfo;

  int64_t st = taosGetTimestampUs();

  for (int32_t k = 0; k < pSup->numOfExprs; ++k) {
    int32_t outputSlotId = pExpr[k].base.resSchema.slotId;

    ASSERT(pExpr[k].pExpr->nodeType == QUERY_NODE_VALUE);
    SColumnInfoData* pColInfoData = taosArrayGet(pRes->pDataBlock, outputSlotId);

    int32_t type = pExpr[k].base.pParam[0].param.nType;
    if (TSDB_DATA_TYPE_NULL == type) {
      colDataAppendNNULL(pColInfoData, 0, 1);
    } else {
      colDataAppend(pColInfoData, 0, taosVariantGet(&pExpr[k].base.pParam[0].param, type), false);
    }
  }

  pRes->info.rows = 1;
  doFilter(pProjectInfo->pFilterNode, pRes, NULL);

613
  /*int32_t status = */ doIngroupLimitOffset(&pProjectInfo->limitInfo, 0, pRes, pOperator);
614 615 616 617 618 619 620 621 622

  pOperator->resultInfo.totalRows += pRes->info.rows;

  doSetOperatorCompleted(pOperator);
  if (pOperator->cost.openCost == 0) {
    pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
  }

  return (pRes->info.rows > 0) ? pRes : NULL;
L
Liu Jicong 已提交
623
}