summaryrefslogtreecommitdiff
path: root/src/mongo/db/pipeline/pipeline_d.cpp
blob: 085f161ebaa88884dd180c2729d2e2e234a0c895 (plain)
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
26
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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
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
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
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
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
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
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
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
/**
 * Copyright (c) 2012-2014 MongoDB Inc.
 *
 * This program is free software: you can redistribute it and/or  modify
 * it under the terms of the GNU Affero General Public License, version 3,
 * 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.  See the
 * GNU Affero General Public License for more details.
 *
 * 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/>.
 *
 * As a special exception, the copyright holders give permission to link the
 * code of portions of this program with the OpenSSL library under certain
 * conditions as described in each individual source file and distribute
 * linked combinations including the program with the OpenSSL library. You
 * must comply with the GNU Affero General Public License in all respects for
 * all of the code used other than as permitted herein. If you modify file(s)
 * with this exception, you may extend this exception to your version of the
 * file(s), but you are not obligated to do so. If you do not wish to do so,
 * delete this exception statement from your version. If you delete this
 * exception statement from all source files in the program, then also delete
 * it in the license file.
 */

#define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kQuery

#include "mongo/platform/basic.h"

#include "mongo/db/pipeline/pipeline_d.h"

#include "mongo/bson/simple_bsonobj_comparator.h"
#include "mongo/db/catalog/collection.h"
#include "mongo/db/catalog/database.h"
#include "mongo/db/catalog/database_holder.h"
#include "mongo/db/catalog/index_catalog.h"
#include "mongo/db/concurrency/d_concurrency.h"
#include "mongo/db/concurrency/write_conflict_exception.h"
#include "mongo/db/db_raii.h"
#include "mongo/db/exec/fetch.h"
#include "mongo/db/exec/multi_iterator.h"
#include "mongo/db/exec/shard_filter.h"
#include "mongo/db/exec/working_set.h"
#include "mongo/db/index/index_access_method.h"
#include "mongo/db/matcher/extensions_callback_real.h"
#include "mongo/db/namespace_string.h"
#include "mongo/db/ops/write_ops_exec.h"
#include "mongo/db/ops/write_ops_gen.h"
#include "mongo/db/pipeline/document_source.h"
#include "mongo/db/pipeline/document_source_change_stream.h"
#include "mongo/db/pipeline/document_source_cursor.h"
#include "mongo/db/pipeline/document_source_geo_near.h"
#include "mongo/db/pipeline/document_source_geo_near_cursor.h"
#include "mongo/db/pipeline/document_source_group.h"
#include "mongo/db/pipeline/document_source_match.h"
#include "mongo/db/pipeline/document_source_sample.h"
#include "mongo/db/pipeline/document_source_sample_from_random_cursor.h"
#include "mongo/db/pipeline/document_source_single_document_transformation.h"
#include "mongo/db/pipeline/document_source_sort.h"
#include "mongo/db/pipeline/pipeline.h"
#include "mongo/db/query/collation/collator_interface.h"
#include "mongo/db/query/get_executor.h"
#include "mongo/db/query/plan_summary_stats.h"
#include "mongo/db/query/query_planner.h"
#include "mongo/db/s/collection_sharding_state.h"
#include "mongo/db/s/sharding_state.h"
#include "mongo/db/service_context.h"
#include "mongo/db/stats/top.h"
#include "mongo/db/storage/record_store.h"
#include "mongo/db/storage/sorted_data_interface.h"
#include "mongo/db/transaction_participant.h"
#include "mongo/rpc/metadata/client_metadata_ismaster.h"
#include "mongo/s/catalog_cache.h"
#include "mongo/s/chunk_manager.h"
#include "mongo/s/chunk_version.h"
#include "mongo/s/grid.h"
#include "mongo/s/query/document_source_merge_cursors.h"
#include "mongo/s/write_ops/cluster_write.h"
#include "mongo/stdx/memory.h"
#include "mongo/util/log.h"
#include "mongo/util/time_support.h"

namespace mongo {

using boost::intrusive_ptr;
using std::shared_ptr;
using std::string;
using std::unique_ptr;
using write_ops::Insert;

namespace {

/**
 * Returns a PlanExecutor which uses a random cursor to sample documents if successful. Returns {}
 * if the storage engine doesn't support random cursors, or if 'sampleSize' is a large enough
 * percentage of the collection.
 */
StatusWith<unique_ptr<PlanExecutor, PlanExecutor::Deleter>> createRandomCursorExecutor(
    Collection* collection, OperationContext* opCtx, long long sampleSize, long long numRecords) {
    double kMaxSampleRatioForRandCursor = 0.05;
    if (sampleSize > numRecords * kMaxSampleRatioForRandCursor || numRecords <= 100) {
        return {nullptr};
    }

    // Attempt to get a random cursor from the RecordStore.
    auto rsRandCursor = collection->getRecordStore()->getRandomCursor(opCtx);
    if (!rsRandCursor) {
        // The storage engine has no random cursor support.
        return {nullptr};
    }

    auto ws = stdx::make_unique<WorkingSet>();
    auto stage = stdx::make_unique<MultiIteratorStage>(opCtx, ws.get(), collection);
    stage->addIterator(std::move(rsRandCursor));

    {
        AutoGetCollectionForRead autoColl(opCtx, collection->ns());

        // If we're in a sharded environment, we need to filter out documents we don't own.
        if (ShardingState::get(opCtx)->needCollectionMetadata(opCtx, collection->ns().ns())) {
            auto shardFilterStage = stdx::make_unique<ShardFilterStage>(
                opCtx,
                CollectionShardingState::get(opCtx, collection->ns())->getMetadata(opCtx),
                ws.get(),
                stage.release());
            return PlanExecutor::make(opCtx,
                                      std::move(ws),
                                      std::move(shardFilterStage),
                                      collection,
                                      PlanExecutor::YIELD_AUTO);
        }
    }

    return PlanExecutor::make(
        opCtx, std::move(ws), std::move(stage), collection, PlanExecutor::YIELD_AUTO);
}

StatusWith<std::unique_ptr<PlanExecutor, PlanExecutor::Deleter>> attemptToGetExecutor(
    OperationContext* opCtx,
    Collection* collection,
    const NamespaceString& nss,
    const intrusive_ptr<ExpressionContext>& pExpCtx,
    bool oplogReplay,
    BSONObj queryObj,
    BSONObj projectionObj,
    BSONObj sortObj,
    boost::optional<std::string> groupIdForDistinctScan,
    const AggregationRequest* aggRequest,
    const size_t plannerOpts,
    const MatchExpressionParser::AllowedFeatureSet& matcherFeatures) {
    auto qr = stdx::make_unique<QueryRequest>(nss);
    qr->setTailableMode(pExpCtx->tailableMode);
    qr->setOplogReplay(oplogReplay);
    qr->setFilter(queryObj);
    qr->setProj(projectionObj);
    qr->setSort(sortObj);
    if (aggRequest) {
        qr->setExplain(static_cast<bool>(aggRequest->getExplain()));
        qr->setHint(aggRequest->getHint());
    }

    // If the pipeline has a non-null collator, set the collation option to the result of
    // serializing the collator's spec back into BSON. We do this in order to fill in all options
    // that the user omitted.
    //
    // If pipeline has a null collator (representing the "simple" collation), we simply set the
    // collation option to the original user BSON, which is either the empty object (unspecified),
    // or the specification for the "simple" collation.
    qr->setCollation(pExpCtx->getCollator() ? pExpCtx->getCollator()->getSpec().toBSON()
                                            : pExpCtx->collation);

    const ExtensionsCallbackReal extensionsCallback(pExpCtx->opCtx, &nss);

    auto cq = CanonicalQuery::canonicalize(
        opCtx, std::move(qr), pExpCtx, extensionsCallback, matcherFeatures);

    if (!cq.isOK()) {
        // Return an error instead of uasserting, since there are cases where the combination of
        // sort and projection will result in a bad query, but when we try with a different
        // combination it will be ok. e.g. a sort by {$meta: 'textScore'}, without any projection
        // will fail, but will succeed when the corresponding '$meta' projection is passed in
        // another attempt.
        return {cq.getStatus()};
    }

    if (groupIdForDistinctScan) {
        // When the pipeline includes a $group that groups by a single field
        // (groupIdForDistinctScan), we use getExecutorDistinct() to attempt to get an executor that
        // uses a DISTINCT_SCAN to scan exactly one document for each group. When that's not
        // possible, we return nullptr, and the caller is responsible for trying again without
        // passing a 'groupIdForDistinctScan' value.
        ParsedDistinct parsedDistinct(std::move(cq.getValue()), *groupIdForDistinctScan);
        auto distinctExecutor =
            getExecutorDistinct(opCtx,
                                collection,
                                plannerOpts | QueryPlannerParams::STRICT_DISTINCT_ONLY,
                                &parsedDistinct);
        if (!distinctExecutor.isOK()) {
            return distinctExecutor.getStatus().withContext(
                "Unable to use distinct scan to optimize $group stage");
        } else if (!distinctExecutor.getValue()) {
            return {ErrorCodes::OperationFailed,
                    "Unable to use distinct scan to optimize $group stage"};
        } else {
            return distinctExecutor;
        }
    }

    return getExecutorFind(opCtx, collection, nss, std::move(cq.getValue()), plannerOpts);
}

BSONObj removeSortKeyMetaProjection(BSONObj projectionObj) {
    if (!projectionObj[Document::metaFieldSortKey]) {
        return projectionObj;
    }
    return projectionObj.removeField(Document::metaFieldSortKey);
}

/**
 * Examines the indexes in 'collection' and returns the field name of a geo-indexed field suitable
 * for use in $geoNear. 2d indexes are given priority over 2dsphere indexes.
 *
 * The 'collection' is required to exist. Throws if no usable 2d or 2dsphere index could be found.
 */
StringData extractGeoNearFieldFromIndexes(OperationContext* opCtx, Collection* collection) {
    invariant(collection);

    std::vector<IndexDescriptor*> idxs;
    collection->getIndexCatalog()->findIndexByType(opCtx, IndexNames::GEO_2D, idxs);
    uassert(ErrorCodes::IndexNotFound,
            str::stream() << "There is more than one 2d index on " << collection->ns().ns()
                          << "; unsure which to use for $geoNear",
            idxs.size() <= 1U);
    if (idxs.size() == 1U) {
        for (auto&& elem : idxs.front()->keyPattern()) {
            if (elem.type() == BSONType::String && elem.valueStringData() == IndexNames::GEO_2D) {
                return elem.fieldNameStringData();
            }
        }
        MONGO_UNREACHABLE;
    }

    // If there are no 2d indexes, look for a 2dsphere index.
    idxs.clear();
    collection->getIndexCatalog()->findIndexByType(opCtx, IndexNames::GEO_2DSPHERE, idxs);
    uassert(ErrorCodes::IndexNotFound,
            "$geoNear requires a 2d or 2dsphere index, but none were found",
            !idxs.empty());
    uassert(ErrorCodes::IndexNotFound,
            str::stream() << "There is more than one 2dsphere index on " << collection->ns().ns()
                          << "; unsure which to use for $geoNear",
            idxs.size() <= 1U);

    invariant(idxs.size() == 1U);
    for (auto&& elem : idxs.front()->keyPattern()) {
        if (elem.type() == BSONType::String && elem.valueStringData() == IndexNames::GEO_2DSPHERE) {
            return elem.fieldNameStringData();
        }
    }
    MONGO_UNREACHABLE;
}
}  // namespace

void PipelineD::prepareCursorSource(Collection* collection,
                                    const NamespaceString& nss,
                                    const AggregationRequest* aggRequest,
                                    Pipeline* pipeline) {
    auto expCtx = pipeline->getContext();

    // We will be modifying the source vector as we go.
    Pipeline::SourceContainer& sources = pipeline->_sources;

    if (!sources.empty() && !sources.front()->constraints().requiresInputDocSource) {
        return;
    }

    // We are going to generate an input cursor, so we need to be holding the collection lock.
    dassert(expCtx->opCtx->lockState()->isCollectionLockedForMode(nss.ns(), MODE_IS));

    if (!sources.empty()) {
        auto sampleStage = dynamic_cast<DocumentSourceSample*>(sources.front().get());
        // Optimize an initial $sample stage if possible.
        if (collection && sampleStage) {
            const long long sampleSize = sampleStage->getSampleSize();
            const long long numRecords = collection->getRecordStore()->numRecords(expCtx->opCtx);
            auto exec = uassertStatusOK(
                createRandomCursorExecutor(collection, expCtx->opCtx, sampleSize, numRecords));
            if (exec) {
                // Replace $sample stage with $sampleFromRandomCursor stage.
                sources.pop_front();
                std::string idString = collection->ns().isOplog() ? "ts" : "_id";
                sources.emplace_front(DocumentSourceSampleFromRandomCursor::create(
                    expCtx, sampleSize, idString, numRecords));

                addCursorSource(
                    pipeline,
                    DocumentSourceCursor::create(collection, std::move(exec), expCtx),
                    pipeline->getDependencies(DepsTracker::MetadataAvailable::kNoMetadata));
                return;
            }
        }
    }

    // If the first stage is $geoNear, prepare a special DocumentSourceGeoNearCursor stage;
    // otherwise, create a generic DocumentSourceCursor.
    const auto geoNearStage =
        sources.empty() ? nullptr : dynamic_cast<DocumentSourceGeoNear*>(sources.front().get());
    if (geoNearStage) {
        prepareGeoNearCursorSource(collection, nss, aggRequest, pipeline);
    } else {
        prepareGenericCursorSource(collection, nss, aggRequest, pipeline);
    }
}

namespace {

/**
 * Look for $sort, $group at the beginning of the pipeline, potentially returning either or both.
 * Returns nullptr for any of the stages that are not found. Note that we are not looking for the
 * opposite pattern ($group, $sort). In that case, this function will return only the $group stage.
 *
 * This function will not return the $group in the case that there is an initial $sort with
 * intermediate stages that separate it from the $group (e.g.: $sort, $limit, $group). That includes
 * the case of a $sort with a non-null value for getLimitSrc(), indicating that there was previously
 * a $limit stage that was optimized away.
 */
std::pair<boost::intrusive_ptr<DocumentSourceSort>, boost::intrusive_ptr<DocumentSourceGroup>>
getSortAndGroupStagesFromPipeline(const Pipeline::SourceContainer& sources) {
    boost::intrusive_ptr<DocumentSourceSort> sortStage = nullptr;
    boost::intrusive_ptr<DocumentSourceGroup> groupStage = nullptr;

    auto sourcesIt = sources.begin();
    if (sourcesIt != sources.end()) {
        sortStage = dynamic_cast<DocumentSourceSort*>(sourcesIt->get());
        if (sortStage) {
            if (!sortStage->getLimitSrc()) {
                ++sourcesIt;
            } else {
                // This $sort stage was previously followed by a $limit stage.
                sourcesIt = sources.end();
            }
        }
    }

    if (sourcesIt != sources.end()) {
        groupStage = dynamic_cast<DocumentSourceGroup*>(sourcesIt->get());
    }

    return std::make_pair(sortStage, groupStage);
}

}  // namespace

void PipelineD::prepareGenericCursorSource(Collection* collection,
                                           const NamespaceString& nss,
                                           const AggregationRequest* aggRequest,
                                           Pipeline* pipeline) {
    Pipeline::SourceContainer& sources = pipeline->_sources;
    auto expCtx = pipeline->getContext();

    // Look for an initial match. This works whether we got an initial query or not. If not, it
    // results in a "{}" query, which will be what we want in that case.
    bool oplogReplay = false;
    const BSONObj queryObj = pipeline->getInitialQuery();
    if (!queryObj.isEmpty()) {
        auto matchStage = dynamic_cast<DocumentSourceMatch*>(sources.front().get());
        if (matchStage) {
            oplogReplay = dynamic_cast<DocumentSourceOplogMatch*>(matchStage) != nullptr;
            // If a $match query is pulled into the cursor, the $match is redundant, and can be
            // removed from the pipeline.
            sources.pop_front();
        } else {
            // A $geoNear stage, the only other stage that can produce an initial query, is also
            // a valid initial stage. However, we should be in prepareGeoNearCursorSource() instead.
            MONGO_UNREACHABLE;
        }
    }

    // Find the set of fields in the source documents depended on by this pipeline.
    DepsTracker deps = pipeline->getDependencies(DocumentSourceMatch::isTextQuery(queryObj)
                                                     ? DepsTracker::MetadataAvailable::kTextScore
                                                     : DepsTracker::MetadataAvailable::kNoMetadata);

    BSONObj projForQuery = deps.toProjection();

    boost::intrusive_ptr<DocumentSourceSort> sortStage;
    boost::intrusive_ptr<DocumentSourceGroup> groupStage;
    std::tie(sortStage, groupStage) = getSortAndGroupStagesFromPipeline(pipeline->_sources);

    BSONObj sortObj;
    if (sortStage) {
        sortObj = sortStage
                      ->sortKeyPattern(
                          DocumentSourceSort::SortKeySerialization::kForPipelineSerialization)
                      .toBson();
    }

    std::unique_ptr<GroupFromFirstDocumentTransformation> rewrittenGroupStage;
    if (groupStage) {
        rewrittenGroupStage = groupStage->rewriteGroupAsTransformOnFirstDocument();
    }

    // Create the PlanExecutor.
    auto exec = uassertStatusOK(prepareExecutor(expCtx->opCtx,
                                                collection,
                                                nss,
                                                pipeline,
                                                expCtx,
                                                oplogReplay,
                                                sortStage,
                                                std::move(rewrittenGroupStage),
                                                deps,
                                                queryObj,
                                                aggRequest,
                                                Pipeline::kAllowedMatcherFeatures,
                                                &sortObj,
                                                &projForQuery));


    if (!projForQuery.isEmpty() && !sources.empty()) {
        // Check for redundant $project in query with the same specification as the inclusion
        // projection generated by the dependency optimization.
        auto proj =
            dynamic_cast<DocumentSourceSingleDocumentTransformation*>(sources.front().get());
        if (proj && proj->isSubsetOfProjection(projForQuery)) {
            sources.pop_front();
        }
    }

    addCursorSource(pipeline,
                    DocumentSourceCursor::create(collection, std::move(exec), expCtx),
                    deps,
                    queryObj,
                    sortObj,
                    projForQuery);
}

void PipelineD::prepareGeoNearCursorSource(Collection* collection,
                                           const NamespaceString& nss,
                                           const AggregationRequest* aggRequest,
                                           Pipeline* pipeline) {
    uassert(ErrorCodes::NamespaceNotFound,
            str::stream() << "$geoNear requires a geo index to run, but " << nss.ns()
                          << " does not exist",
            collection);

    Pipeline::SourceContainer& sources = pipeline->_sources;
    auto expCtx = pipeline->getContext();
    const auto geoNearStage = dynamic_cast<DocumentSourceGeoNear*>(sources.front().get());
    invariant(geoNearStage);

    auto deps = pipeline->getDependencies(DepsTracker::kAllGeoNearDataAvailable);

    // If the user specified a "key" field, use that field to satisfy the "near" query. Otherwise,
    // look for a geo-indexed field in 'collection' that can.
    auto nearFieldName =
        (geoNearStage->getKeyField() ? geoNearStage->getKeyField()->fullPath()
                                     : extractGeoNearFieldFromIndexes(expCtx->opCtx, collection))
            .toString();

    // Create a PlanExecutor whose query is the "near" predicate on 'nearFieldName' combined with
    // the optional "query" argument in the $geoNear stage.
    BSONObj fullQuery = geoNearStage->asNearQuery(nearFieldName);
    BSONObj proj = deps.toProjection();
    BSONObj sortFromQuerySystem;
    auto exec = uassertStatusOK(prepareExecutor(expCtx->opCtx,
                                                collection,
                                                nss,
                                                pipeline,
                                                expCtx,
                                                false,   /* oplogReplay */
                                                nullptr, /* sortStage */
                                                nullptr, /* rewrittenGroupStage */
                                                deps,
                                                std::move(fullQuery),
                                                aggRequest,
                                                Pipeline::kGeoNearMatcherFeatures,
                                                &sortFromQuerySystem,
                                                &proj));

    invariant(sortFromQuerySystem.isEmpty(),
              str::stream() << "Unexpectedly got the following sort from the query system: "
                            << sortFromQuerySystem.jsonString());

    auto geoNearCursor =
        DocumentSourceGeoNearCursor::create(collection,
                                            std::move(exec),
                                            expCtx,
                                            geoNearStage->getDistanceField(),
                                            geoNearStage->getLocationField(),
                                            geoNearStage->getDistanceMultiplier().value_or(1.0));

    // Remove the initial $geoNear; it will be replaced by $geoNearCursor.
    sources.pop_front();
    addCursorSource(pipeline, std::move(geoNearCursor), std::move(deps));
}

StatusWith<std::unique_ptr<PlanExecutor, PlanExecutor::Deleter>> PipelineD::prepareExecutor(
    OperationContext* opCtx,
    Collection* collection,
    const NamespaceString& nss,
    Pipeline* pipeline,
    const intrusive_ptr<ExpressionContext>& expCtx,
    bool oplogReplay,
    const boost::intrusive_ptr<DocumentSourceSort>& sortStage,
    std::unique_ptr<GroupFromFirstDocumentTransformation> rewrittenGroupStage,
    const DepsTracker& deps,
    const BSONObj& queryObj,
    const AggregationRequest* aggRequest,
    const MatchExpressionParser::AllowedFeatureSet& matcherFeatures,
    BSONObj* sortObj,
    BSONObj* projectionObj) {
    // The query system has the potential to use an index to provide a non-blocking sort and/or to
    // use the projection to generate a covered plan. If this is possible, it is more efficient to
    // let the query system handle those parts of the pipeline. If not, it is more efficient to use
    // a $sort and/or a ParsedDeps object. Thus, we will determine whether the query system can
    // provide a non-blocking sort or a covered projection before we commit to a PlanExecutor.
    //
    // To determine if the query system can provide a non-blocking sort, we pass the
    // NO_BLOCKING_SORT planning option, meaning 'getExecutor' will not produce a PlanExecutor if
    // the query system would use a blocking sort stage.
    //
    // To determine if the query system can provide a covered projection, we pass the
    // NO_UNCOVERED_PROJECTS planning option, meaning 'getExecutor' will not produce a PlanExecutor
    // if the query system would need to fetch the document to do the projection. The following
    // logic uses the above strategies, with multiple calls to 'attemptToGetExecutor' to determine
    // the most efficient way to handle the $sort and $project stages.
    //
    // LATER - We should attempt to determine if the results from the query are returned in some
    // order so we can then apply other optimizations there are tickets for, such as SERVER-4507.
    size_t plannerOpts = QueryPlannerParams::DEFAULT | QueryPlannerParams::NO_BLOCKING_SORT;

    if (deps.hasNoRequirements()) {
        // If we don't need any fields from the input document, performing a count is faster, and
        // will output empty documents, which is okay.
        plannerOpts |= QueryPlannerParams::IS_COUNT;
    }

    if (expCtx->needsMerge && expCtx->tailableMode == TailableModeEnum::kTailableAndAwaitData) {
        plannerOpts |= QueryPlannerParams::TRACK_LATEST_OPLOG_TS;
    }

    if (rewrittenGroupStage) {
        BSONObj emptySort;

        // See if the query system can handle the $group and $sort stage using a DISTINCT_SCAN
        // (SERVER-9507). Note that passing the empty projection (as we do for some of the
        // attemptToGetExecutor() calls below) causes getExecutorDistinct() to ignore some otherwise
        // valid DISTINCT_SCAN plans, so we pass the projection and exclude the
        // NO_UNCOVERED_PROJECTIONS planner parameter.
        auto swExecutorGrouped = attemptToGetExecutor(opCtx,
                                                      collection,
                                                      nss,
                                                      expCtx,
                                                      oplogReplay,
                                                      queryObj,
                                                      *projectionObj,
                                                      sortObj ? *sortObj : emptySort,
                                                      rewrittenGroupStage->groupId(),
                                                      aggRequest,
                                                      plannerOpts,
                                                      matcherFeatures);

        if (swExecutorGrouped.isOK()) {
            // Any $limit stage before the $group stage should make the pipeline ineligible for this
            // optimization.
            invariant(!sortStage || !sortStage->getLimitSrc());

            // We remove the $sort and $group stages that begin the pipeline, because the executor
            // will handle the sort, and the groupTransform (added below) will handle the $group
            // stage.
            pipeline->popFrontWithName(DocumentSourceSort::kStageName);
            pipeline->popFrontWithName(DocumentSourceGroup::kStageName);

            boost::intrusive_ptr<DocumentSource> groupTransform(
                new DocumentSourceSingleDocumentTransformation(
                    expCtx,
                    std::move(rewrittenGroupStage),
                    "$groupByDistinctScan",
                    false /* independentOfAnyCollection */));
            pipeline->addInitialSource(groupTransform);

            return swExecutorGrouped;
        } else if (swExecutorGrouped == ErrorCodes::QueryPlanKilled) {
            return {ErrorCodes::OperationFailed,
                    str::stream() << "Failed to determine whether query system can provide a "
                                     "DISTINCT_SCAN grouping: "
                                  << swExecutorGrouped.getStatus().toString()};
        }
    }

    const BSONObj emptyProjection;
    const BSONObj metaSortProjection = BSON("$meta"
                                            << "sortKey");

    // The only way to get meta information (e.g. the text score) is to let the query system handle
    // the projection. In all other cases, unless the query system can do an index-covered
    // projection and avoid going to the raw record at all, it is faster to have ParsedDeps filter
    // the fields we need.
    if (!deps.getNeedsAnyMetadata()) {
        plannerOpts |= QueryPlannerParams::NO_UNCOVERED_PROJECTIONS;
    }

    if (sortStage) {
        // See if the query system can provide a non-blocking sort.
        auto swExecutorSort =
            attemptToGetExecutor(opCtx,
                                 collection,
                                 nss,
                                 expCtx,
                                 oplogReplay,
                                 queryObj,
                                 expCtx->needsMerge ? metaSortProjection : emptyProjection,
                                 *sortObj,
                                 boost::none, /* groupIdForDistinctScan */
                                 aggRequest,
                                 plannerOpts,
                                 matcherFeatures);

        if (swExecutorSort.isOK()) {
            // Success! Now see if the query system can also cover the projection.
            auto swExecutorSortAndProj =
                attemptToGetExecutor(opCtx,
                                     collection,
                                     nss,
                                     expCtx,
                                     oplogReplay,
                                     queryObj,
                                     *projectionObj,
                                     *sortObj,
                                     boost::none, /* groupIdForDistinctScan */
                                     aggRequest,
                                     plannerOpts,
                                     matcherFeatures);

            std::unique_ptr<PlanExecutor, PlanExecutor::Deleter> exec;
            if (swExecutorSortAndProj.isOK()) {
                // Success! We have a non-blocking sort and a covered projection.
                exec = std::move(swExecutorSortAndProj.getValue());
            } else if (swExecutorSortAndProj == ErrorCodes::QueryPlanKilled) {
                return {ErrorCodes::OperationFailed,
                        str::stream() << "Failed to determine whether query system can provide a "
                                         "covered projection in addition to a non-blocking sort: "
                                      << swExecutorSortAndProj.getStatus().toString()};
            } else {
                // The query system couldn't cover the projection.
                *projectionObj = BSONObj();
                exec = std::move(swExecutorSort.getValue());
            }

            // We know the sort is being handled by the query system, so remove the $sort stage.
            pipeline->_sources.pop_front();

            if (sortStage->getLimitSrc()) {
                // We need to reinsert the coalesced $limit after removing the $sort.
                pipeline->_sources.push_front(sortStage->getLimitSrc());
            }
            return std::move(exec);
        } else if (swExecutorSort == ErrorCodes::QueryPlanKilled) {
            return {
                ErrorCodes::OperationFailed,
                str::stream()
                    << "Failed to determine whether query system can provide a non-blocking sort: "
                    << swExecutorSort.getStatus().toString()};
        }
        // The query system can't provide a non-blocking sort.
        *sortObj = BSONObj();
    }

    // Either there was no $sort stage, or the query system could not provide a non-blocking
    // sort.
    dassert(sortObj->isEmpty());
    *projectionObj = removeSortKeyMetaProjection(*projectionObj);
    const auto metadataRequired = deps.getAllRequiredMetadataTypes();
    if (metadataRequired.size() == 1 &&
        metadataRequired.front() == DepsTracker::MetadataType::SORT_KEY) {
        // A sort key requirement would have prevented us from being able to add this parameter
        // before, but now we know the query system won't cover the sort, so we will be able to
        // compute the sort key ourselves during the $sort stage, and thus don't need a query
        // projection to do so.
        plannerOpts |= QueryPlannerParams::NO_UNCOVERED_PROJECTIONS;
    }

    // See if the query system can cover the projection.
    auto swExecutorProj = attemptToGetExecutor(opCtx,
                                               collection,
                                               nss,
                                               expCtx,
                                               oplogReplay,
                                               queryObj,
                                               *projectionObj,
                                               *sortObj,
                                               boost::none, /* groupIdForDistinctScan */
                                               aggRequest,
                                               plannerOpts,
                                               matcherFeatures);
    if (swExecutorProj.isOK()) {
        // Success! We have a covered projection.
        return std::move(swExecutorProj.getValue());
    } else if (swExecutorProj == ErrorCodes::QueryPlanKilled) {
        return {ErrorCodes::OperationFailed,
                str::stream()
                    << "Failed to determine whether query system can provide a covered projection: "
                    << swExecutorProj.getStatus().toString()};
    }

    // The query system couldn't provide a covered projection.
    *projectionObj = BSONObj();
    // If this doesn't work, nothing will.
    return attemptToGetExecutor(opCtx,
                                collection,
                                nss,
                                expCtx,
                                oplogReplay,
                                queryObj,
                                *projectionObj,
                                *sortObj,
                                boost::none, /* groupIdForDistinctScan */
                                aggRequest,
                                plannerOpts,
                                matcherFeatures);
}

void PipelineD::addCursorSource(Pipeline* pipeline,
                                boost::intrusive_ptr<DocumentSourceCursor> cursor,
                                DepsTracker deps,
                                const BSONObj& queryObj,
                                const BSONObj& sortObj,
                                const BSONObj& projectionObj) {
    cursor->setQuery(queryObj);
    cursor->setSort(sortObj);
    if (deps.hasNoRequirements()) {
        cursor->shouldProduceEmptyDocs();
    }

    if (!projectionObj.isEmpty()) {
        cursor->setProjection(projectionObj, boost::none);
    } else {
        // There may be fewer dependencies now if the sort was covered.
        if (!sortObj.isEmpty()) {
            deps = pipeline->getDependencies(DocumentSourceMatch::isTextQuery(queryObj)
                                                 ? DepsTracker::MetadataAvailable::kTextScore
                                                 : DepsTracker::MetadataAvailable::kNoMetadata);
        }

        cursor->setProjection(deps.toProjection(), deps.toParsedDeps());
    }
    pipeline->addInitialSource(std::move(cursor));
}

Timestamp PipelineD::getLatestOplogTimestamp(const Pipeline* pipeline) {
    if (auto docSourceCursor =
            dynamic_cast<DocumentSourceCursor*>(pipeline->_sources.front().get())) {
        return docSourceCursor->getLatestOplogTimestamp();
    }
    return Timestamp();
}

std::string PipelineD::getPlanSummaryStr(const Pipeline* pipeline) {
    if (auto docSourceCursor =
            dynamic_cast<DocumentSourceCursor*>(pipeline->_sources.front().get())) {
        return docSourceCursor->getPlanSummaryStr();
    }

    return "";
}

void PipelineD::getPlanSummaryStats(const Pipeline* pipeline, PlanSummaryStats* statsOut) {
    invariant(statsOut);

    if (auto docSourceCursor =
            dynamic_cast<DocumentSourceCursor*>(pipeline->_sources.front().get())) {
        *statsOut = docSourceCursor->getPlanSummaryStats();
    }

    bool hasSortStage{false};
    bool usedDisk{false};
    for (auto&& source : pipeline->_sources) {
        if (dynamic_cast<DocumentSourceSort*>(source.get()))
            hasSortStage = true;

        usedDisk = usedDisk || source->usedDisk();
        if (usedDisk && hasSortStage)
            break;
    }
    statsOut->hasSortStage = hasSortStage;
    statsOut->usedDisk = usedDisk;
}

}  // namespace mongo