summaryrefslogtreecommitdiff
path: root/src/mongo/db/pipeline/pipeline.cpp
blob: bba6bf9615f6e9f096ef9b0a721ba6edf7c4a3cb (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
/**
 * Copyright (c) 2011 10gen 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.
 */

#include "mongo/platform/basic.h"

// This file defines functions from both of these headers
#include "mongo/db/pipeline/pipeline.h"
#include "mongo/db/pipeline/pipeline_optimizations.h"

#include "mongo/db/auth/action_set.h"
#include "mongo/db/auth/privilege.h"
#include "mongo/db/catalog/document_validation.h"
#include "mongo/db/commands.h"
#include "mongo/db/jsobj.h"
#include "mongo/db/pipeline/accumulator.h"
#include "mongo/db/pipeline/document.h"
#include "mongo/db/pipeline/document_source.h"
#include "mongo/db/pipeline/expression.h"
#include "mongo/db/pipeline/expression_context.h"
#include "mongo/util/mongoutils/str.h"

namespace mongo {

    using boost::intrusive_ptr;
    using std::endl;
    using std::ostringstream;
    using std::string;
    using std::vector;

    const char Pipeline::commandName[] = "aggregate";
    const char Pipeline::pipelineName[] = "pipeline";
    const char Pipeline::explainName[] = "explain";
    const char Pipeline::fromRouterName[] = "fromRouter";
    const char Pipeline::serverPipelineName[] = "serverPipeline";
    const char Pipeline::mongosPipelineName[] = "mongosPipeline";

    Pipeline::Pipeline(const intrusive_ptr<ExpressionContext> &pTheCtx):
        explain(false),
        pCtx(pTheCtx) {
    }


    /* this structure is used to make a lookup table of operators */
    struct StageDesc {
        const char *pName;
        intrusive_ptr<DocumentSource> (*pFactory)(
            BSONElement, const intrusive_ptr<ExpressionContext> &);
    };

    /* this table must be in alphabetical order by name for bsearch() */
    static const StageDesc stageDesc[] = {
        {DocumentSourceGeoNear::geoNearName,
         DocumentSourceGeoNear::createFromBson},
        {DocumentSourceGroup::groupName,
         DocumentSourceGroup::createFromBson},
        {DocumentSourceLimit::limitName,
         DocumentSourceLimit::createFromBson},
        {DocumentSourceMatch::matchName,
         DocumentSourceMatch::createFromBson},
        {DocumentSourceMergeCursors::name,
         DocumentSourceMergeCursors::createFromBson},
        {DocumentSourceOut::outName,
         DocumentSourceOut::createFromBson},
        {DocumentSourceProject::projectName,
         DocumentSourceProject::createFromBson},
        {DocumentSourceRedact::redactName,
         DocumentSourceRedact::createFromBson},
        {DocumentSourceSkip::skipName,
         DocumentSourceSkip::createFromBson},
        {DocumentSourceSort::sortName,
         DocumentSourceSort::createFromBson},
        {DocumentSourceUnwind::unwindName,
         DocumentSourceUnwind::createFromBson},
    };
    static const size_t nStageDesc = sizeof(stageDesc) / sizeof(StageDesc);

    static int stageDescCmp(const void *pL, const void *pR) {
        return strcmp(((const StageDesc *)pL)->pName,
                      ((const StageDesc *)pR)->pName);
    }

    intrusive_ptr<Pipeline> Pipeline::parseCommand(string& errmsg,
                                                   const BSONObj& cmdObj,
                                                   const intrusive_ptr<ExpressionContext>& pCtx) {
        intrusive_ptr<Pipeline> pPipeline(new Pipeline(pCtx));
        vector<BSONElement> pipeline;

        /* gather the specification for the aggregation */
        for(BSONObj::iterator cmdIterator = cmdObj.begin();
                cmdIterator.more(); ) {
            BSONElement cmdElement(cmdIterator.next());
            const char *pFieldName = cmdElement.fieldName();

            // ignore top-level fields prefixed with $. They are for the command processor, not us.
            if (pFieldName[0] == '$') {
                continue;
            }

            // maxTimeMS is also for the command processor.
            if (pFieldName == LiteParsedQuery::cmdOptionMaxTimeMS) {
                continue;
            }

            // ignore cursor options since they are handled externally.
            if (str::equals(pFieldName, "cursor")) {
                continue;
            }

            /* look for the aggregation command */
            if (!strcmp(pFieldName, commandName)) {
                continue;
            }

            /* check for the collection name */
            if (!strcmp(pFieldName, pipelineName)) {
                pipeline = cmdElement.Array();
                continue;
            }

            /* check for explain option */
            if (!strcmp(pFieldName, explainName)) {
                pPipeline->explain = cmdElement.Bool();
                continue;
            }

            /* if the request came from the router, we're in a shard */
            if (!strcmp(pFieldName, fromRouterName)) {
                pCtx->inShard = cmdElement.Bool();
                continue;
            }

            if (str::equals(pFieldName, "allowDiskUse")) {
                uassert(16949,
                        str::stream() << "allowDiskUse must be a bool, not a "
                                      << typeName(cmdElement.type()),
                        cmdElement.type() == Bool);
                pCtx->extSortAllowed = cmdElement.Bool();
                continue;
            }

            if (pFieldName == bypassDocumentValidationCommandOption()) {
                pCtx->bypassDocumentValidation = cmdElement.trueValue();
                continue;
            }

            /* we didn't recognize a field in the command */
            ostringstream sb;
            sb << "unrecognized field '" << cmdElement.fieldName() << "'";
            errmsg = sb.str();
            return intrusive_ptr<Pipeline>();
        }

        /*
          If we get here, we've harvested the fields we expect for a pipeline.

          Set up the specified document source pipeline.
        */
        SourceContainer& sources = pPipeline->sources; // shorthand

        /* iterate over the steps in the pipeline */
        const size_t nSteps = pipeline.size();
        for(size_t iStep = 0; iStep < nSteps; ++iStep) {
            /* pull out the pipeline element as an object */
            BSONElement pipeElement(pipeline[iStep]);
            uassert(15942, str::stream() << "pipeline element " <<
                    iStep << " is not an object",
                    pipeElement.type() == Object);
            BSONObj bsonObj(pipeElement.Obj());

            // Parse a pipeline stage from 'bsonObj'.
            uassert(16435, "A pipeline stage specification object must contain exactly one field.",
                    bsonObj.nFields() == 1);
            BSONElement stageSpec = bsonObj.firstElement();
            const char* stageName = stageSpec.fieldName();

            // Create a DocumentSource pipeline stage from 'stageSpec'.
            StageDesc key;
            key.pName = stageName;
            const StageDesc* pDesc = (const StageDesc*)
                    bsearch(&key, stageDesc, nStageDesc, sizeof(StageDesc),
                            stageDescCmp);

            uassert(16436,
                    str::stream() << "Unrecognized pipeline stage name: '" << stageName << "'",
                    pDesc);
            intrusive_ptr<DocumentSource> stage = pDesc->pFactory(stageSpec, pCtx);
            verify(stage);
            sources.push_back(stage);

            // TODO find a good general way to check stages that must be first syntactically

            if (dynamic_cast<DocumentSourceOut*>(stage.get())) {
                uassert(16991, "$out can only be the final stage in the pipeline",
                        iStep == nSteps - 1);
            }
        }

        // The order in which optimizations are applied can have significant impact on the
        // efficiency of the final pipeline. Be Careful!
        Optimizations::Local::moveMatchBeforeSort(pPipeline.get());
        Optimizations::Local::moveSkipAndLimitBeforeProject(pPipeline.get());
        Optimizations::Local::moveLimitBeforeSkip(pPipeline.get());
        Optimizations::Local::coalesceAdjacent(pPipeline.get());
        Optimizations::Local::optimizeEachDocumentSource(pPipeline.get());
        Optimizations::Local::duplicateMatchBeforeInitalRedact(pPipeline.get());

        return pPipeline;
    }

    void Pipeline::Optimizations::Local::moveMatchBeforeSort(Pipeline* pipeline) {
        // TODO Keep moving matches across multiple sorts as moveLimitBeforeSkip does below.
        // TODO Check sort for limit. Not an issue currently due to order optimizations are applied,
        // but should be fixed.
        SourceContainer& sources = pipeline->sources;
        for (size_t srcn = sources.size(), srci = 1; srci < srcn; ++srci) {
            intrusive_ptr<DocumentSource> &pSource = sources[srci];
            DocumentSourceMatch* match = dynamic_cast<DocumentSourceMatch *>(pSource.get());
            if (match && !match->isTextQuery()) {
                intrusive_ptr<DocumentSource> &pPrevious = sources[srci - 1];
                if (dynamic_cast<DocumentSourceSort *>(pPrevious.get())) {
                    /* swap this item with the previous */
                    intrusive_ptr<DocumentSource> pTemp(pPrevious);
                    pPrevious = pSource;
                    pSource = pTemp;
                }
            }
        }
    }

    void Pipeline::Optimizations::Local::moveSkipAndLimitBeforeProject(Pipeline* pipeline) {
        SourceContainer& sources = pipeline->sources;
        if (sources.empty()) return;

        for (int i = sources.size() - 1; i >= 1 /* not looking at 0 */; i--) {
            // This optimization only applies when a $project comes before a $skip or $limit.
            auto project = dynamic_cast<DocumentSourceProject*>(sources[i-1].get());
            if (!project) continue;

            auto skip = dynamic_cast<DocumentSourceSkip*>(sources[i].get());
            auto limit = dynamic_cast<DocumentSourceLimit*>(sources[i].get());
            if (!(skip || limit)) continue;

            swap(sources[i], sources[i-1]);

            // Start at back again. This is needed to handle cases with more than 1 $skip or
            // $limit (S means skip, L means limit, P means project)
            //
            // These would work without second pass (assuming back to front ordering)
            // PS  -> SP
            // PL  -> LP
            // PPL -> LPP
            // PPS -> SPP
            //
            // The following cases need a second pass to handle the second skip or limit
            // PLL  -> LLP
            // PPLL -> LLPP
            // PLPL -> LLPP
            i = sources.size(); // decremented before next pass
        }
    }

    void Pipeline::Optimizations::Local::moveLimitBeforeSkip(Pipeline* pipeline) {
        SourceContainer& sources = pipeline->sources;
        if (sources.empty())
            return;

        for(int i = sources.size() - 1; i >= 1 /* not looking at 0 */; i--) {
            DocumentSourceLimit* limit =
                dynamic_cast<DocumentSourceLimit*>(sources[i].get());
            DocumentSourceSkip* skip =
                dynamic_cast<DocumentSourceSkip*>(sources[i-1].get());
            if (limit && skip) {
                // Increase limit by skip since the skipped docs now pass through the $limit
                limit->setLimit(limit->getLimit() + skip->getSkip());
                swap(sources[i], sources[i-1]);

                // Start at back again. This is needed to handle cases with more than 1 $limit
                // (S means skip, L means limit)
                //
                // These two would work without second pass (assuming back to front ordering)
                // SL   -> LS
                // SSL  -> LSS
                //
                // The following cases need a second pass to handle the second limit
                // SLL  -> LLS
                // SSLL -> LLSS
                // SLSL -> LLSS
                i = sources.size(); // decremented before next pass
            }
        }
    }

    void Pipeline::Optimizations::Local::coalesceAdjacent(Pipeline* pipeline) {
        SourceContainer& sources = pipeline->sources;
        if (sources.empty())
            return;

        // move all sources to a temporary list
        SourceContainer tempSources;
        sources.swap(tempSources);

        // move the first one to the final list
        sources.push_back(tempSources[0]);

        // run through the sources, coalescing them or keeping them
        for (size_t tempn = tempSources.size(), tempi = 1; tempi < tempn; ++tempi) {
            // If we can't coalesce the source with the last, then move it
            // to the final list, and make it the new last.  (If we succeeded,
            // then we're still on the same last, and there's no need to move
            // or do anything with the source -- the destruction of tempSources
            // will take care of the rest.)
            intrusive_ptr<DocumentSource> &pLastSource = sources.back();
            intrusive_ptr<DocumentSource> &pTemp = tempSources[tempi];
            verify(pTemp && pLastSource);
            if (!pLastSource->coalesce(pTemp))
                sources.push_back(pTemp);
        }
    }

    void Pipeline::Optimizations::Local::optimizeEachDocumentSource(Pipeline* pipeline) {
        SourceContainer& sources = pipeline->sources;
        SourceContainer newSources;
        for (SourceContainer::iterator it(sources.begin()); it != sources.end(); ++it) {
            if (auto out = (*it)->optimize()) {
                newSources.push_back(std::move(out));
            }
        }
        pipeline->sources = std::move(newSources);
    }

    void Pipeline::Optimizations::Local::duplicateMatchBeforeInitalRedact(Pipeline* pipeline) {
        SourceContainer& sources = pipeline->sources;
        if (sources.size() >= 2 && dynamic_cast<DocumentSourceRedact*>(sources[0].get())) {
            if (DocumentSourceMatch* match = dynamic_cast<DocumentSourceMatch*>(sources[1].get())) {
                const BSONObj redactSafePortion = match->redactSafePortion();
                if (!redactSafePortion.isEmpty()) {
                    sources.push_front(
                        DocumentSourceMatch::createFromBson(
                            BSON("$match" << redactSafePortion).firstElement(),
                            pipeline->pCtx));
                }
            }
        }
    }

    void Pipeline::addRequiredPrivileges(Command* commandTemplate,
                                         const string& db,
                                         BSONObj cmdObj,
                                         vector<Privilege>* out) {
        ResourcePattern inputResource(commandTemplate->parseResourcePattern(db, cmdObj));
        uassert(17138,
                mongoutils::str::stream() << "Invalid input resource, " << inputResource.toString(),
                inputResource.isExactNamespacePattern());

        out->push_back(Privilege(inputResource, ActionType::find));

        BSONObj pipeline = cmdObj.getObjectField("pipeline");
        BSONForEach(stageElem, pipeline) {
            BSONObj stage = stageElem.embeddedObjectUserCheck();
            if (str::equals(stage.firstElementFieldName(), "$out")) {
                NamespaceString outputNs(db, stage.firstElement().str());
                uassert(17139,
                        mongoutils::str::stream() << "Invalid $out target namespace, " <<
                        outputNs.ns(),
                        outputNs.isValid());

                ActionSet actions;
                actions.addAction(ActionType::remove);
                actions.addAction(ActionType::insert);
                if (shouldBypassDocumentValidationForCommand(cmdObj)) {
                    actions.addAction(ActionType::bypassDocumentValidation);
                }

                out->push_back(Privilege(ResourcePattern::forExactNamespace(outputNs), actions));
            }
        }
    }

    intrusive_ptr<Pipeline> Pipeline::splitForSharded() {
        // Create and initialize the shard spec we'll return. We start with an empty pipeline on the
        // shards and all work being done in the merger. Optimizations can move operations between
        // the pipelines to be more efficient.
        intrusive_ptr<Pipeline> shardPipeline(new Pipeline(pCtx));
        shardPipeline->explain = explain;

        // The order in which optimizations are applied can have significant impact on the
        // efficiency of the final pipeline. Be Careful!
        Optimizations::Sharded::findSplitPoint(shardPipeline.get(), this);
        Optimizations::Sharded::moveFinalUnwindFromShardsToMerger(shardPipeline.get(), this);
        Optimizations::Sharded::limitFieldsSentFromShardsToMerger(shardPipeline.get(), this);

        return shardPipeline;
    }

    void Pipeline::Optimizations::Sharded::findSplitPoint(Pipeline* shardPipe,
                                                          Pipeline* mergePipe) {
        while (!mergePipe->sources.empty()) {
            intrusive_ptr<DocumentSource> current = mergePipe->sources.front();
            mergePipe->sources.pop_front();

            // Check if this source is splittable
            SplittableDocumentSource* splittable =
                dynamic_cast<SplittableDocumentSource*>(current.get());

            if (!splittable){
                // move the source from the merger sources to the shard sources
                shardPipe->sources.push_back(current);
            }
            else {
                // split this source into Merge and Shard sources
                intrusive_ptr<DocumentSource> shardSource = splittable->getShardSource();
                intrusive_ptr<DocumentSource> mergeSource = splittable->getMergeSource();
                if (shardSource) shardPipe->sources.push_back(shardSource);
                if (mergeSource) mergePipe->sources.push_front(mergeSource);

                break;
            }
        }
    }

    void Pipeline::Optimizations::Sharded::moveFinalUnwindFromShardsToMerger(Pipeline* shardPipe,
                                                                             Pipeline* mergePipe) {
        while (!shardPipe->sources.empty()
                && dynamic_cast<DocumentSourceUnwind*>(shardPipe->sources.back().get())) {
            mergePipe->sources.push_front(shardPipe->sources.back());
            shardPipe->sources.pop_back();
        }
    }

    void Pipeline::Optimizations::Sharded::limitFieldsSentFromShardsToMerger(Pipeline* shardPipe,
                                                                             Pipeline* mergePipe) {
        DepsTracker mergeDeps = mergePipe->getDependencies(shardPipe->getInitialQuery());
        if (mergeDeps.needWholeDocument)
            return; // the merge needs all fields, so nothing we can do.

        // Empty project is "special" so if no fields are needed, we just ask for _id instead.
        if (mergeDeps.fields.empty())
            mergeDeps.fields.insert("_id");

        // Remove metadata from dependencies since it automatically flows through projection and we
        // don't want to project it in to the document.
        mergeDeps.needTextScore = false;

        // HEURISTIC: only apply optimization if none of the shard stages have an exhaustive list of
        // field dependencies. While this may not be 100% ideal in all cases, it is simple and
        // avoids the worst cases by ensuring that:
        // 1) Optimization IS applied when the shards wouldn't have known their exhaustive list of
        //    dependencies. This situation can happen when a $sort is before the first $project or
        //    $group. Without the optimization, the shards would have to reify and transmit full
        //    objects even though only a subset of fields are needed.
        // 2) Optimization IS NOT applied immediately following a $project or $group since it would
        //    add an unnecessary project (and therefore a deep-copy).
        for (size_t i = 0; i < shardPipe->sources.size(); i++) {
            DepsTracker dt; // ignored
            if (shardPipe->sources[i]->getDependencies(&dt) & DocumentSource::EXHAUSTIVE_FIELDS)
                return;
        }

        // if we get here, add the project.
        shardPipe->sources.push_back(
            DocumentSourceProject::createFromBson(
                BSON("$project" << mergeDeps.toProjection()).firstElement(),
                shardPipe->pCtx));
    }

    BSONObj Pipeline::getInitialQuery() const {
        if (sources.empty())
            return BSONObj();

        /* look for an initial $match */
        DocumentSourceMatch* match = dynamic_cast<DocumentSourceMatch*>(sources.front().get());
        if (!match)
            return BSONObj();

        return match->getQuery();
    }

    bool Pipeline::hasOutStage() const {
        if (sources.empty()) {
            return false;
        }

        // The $out stage must be the last one in the pipeline, so check if the last stage is $out.
        return dynamic_cast<DocumentSourceOut*>(sources.back().get());
    }

    Document Pipeline::serialize() const {
        MutableDocument serialized;
        // create an array out of the pipeline operations
        vector<Value> array;
        for(SourceContainer::const_iterator iter(sources.begin()),
                                            listEnd(sources.end());
                                        iter != listEnd;
                                        ++iter) {
            intrusive_ptr<DocumentSource> pSource(*iter);
            pSource->serializeToArray(array);
        }

        // add the top-level items to the command
        serialized.setField(commandName, Value(pCtx->ns.coll()));
        serialized.setField(pipelineName, Value(array));

        if (explain) {
            serialized.setField(explainName, Value(explain));
        }

        if (pCtx->extSortAllowed) {
            serialized.setField("allowDiskUse", Value(true));
        }

        if (pCtx->bypassDocumentValidation) {
            serialized.setField(bypassDocumentValidationCommandOption(), Value(true));
        }

        return serialized.freeze();
    }

    void Pipeline::stitch() {
        massert(16600, "should not have an empty pipeline",
                !sources.empty());

        /* chain together the sources we found */
        DocumentSource* prevSource = sources.front().get();
        for(SourceContainer::iterator iter(sources.begin() + 1),
                                      listEnd(sources.end());
                                    iter != listEnd;
                                    ++iter) {
            intrusive_ptr<DocumentSource> pTemp(*iter);
            pTemp->setSource(prevSource);
            prevSource = pTemp.get();
        }
    }

    void Pipeline::run(BSONObjBuilder& result) {
        // should not get here in the explain case
        verify(!explain);

        // the array in which the aggregation results reside
        // cant use subArrayStart() due to error handling
        BSONArrayBuilder resultArray;
        DocumentSource* finalSource = sources.back().get();
        while (boost::optional<Document> next = finalSource->getNext()) {
            // add the document to the result set
            BSONObjBuilder documentBuilder (resultArray.subobjStart());
            next->toBson(&documentBuilder);
            documentBuilder.doneFast();
            // object will be too large, assert. the extra 1KB is for headers
            uassert(16389,
                    str::stream() << "aggregation result exceeds maximum document size ("
                                  << BSONObjMaxUserSize / (1024 * 1024) << "MB)",
                    resultArray.len() < BSONObjMaxUserSize - 1024);
        }

        resultArray.done();
        result.appendArray("result", resultArray.arr());
    }

    vector<Value> Pipeline::writeExplainOps() const {
        vector<Value> array;
        for(SourceContainer::const_iterator it = sources.begin(); it != sources.end(); ++it) {
            (*it)->serializeToArray(array, /*explain=*/true);
        }
        return array;
    }

    void Pipeline::addInitialSource(intrusive_ptr<DocumentSource> source) {
        sources.push_front(source);
    }

    DepsTracker Pipeline::getDependencies(const BSONObj& initialQuery) const {
        DepsTracker deps;
        bool knowAllFields = false;
        bool knowAllMeta = false;
        for (size_t i=0; i < sources.size() && !(knowAllFields && knowAllMeta); i++) {
            DepsTracker localDeps;
            DocumentSource::GetDepsReturn status = sources[i]->getDependencies(&localDeps);

            if (status == DocumentSource::NOT_SUPPORTED) {
                // Assume this stage needs everything. We may still know something about our
                // dependencies if an earlier stage returned either EXHAUSTIVE_FIELDS or
                // EXHAUSTIVE_META.
                break;
            }

            if (!knowAllFields) {
                deps.fields.insert(localDeps.fields.begin(), localDeps.fields.end());
                if (localDeps.needWholeDocument)
                    deps.needWholeDocument = true;
                knowAllFields = status & DocumentSource::EXHAUSTIVE_FIELDS;
            }

            if (!knowAllMeta) {
                if (localDeps.needTextScore)
                    deps.needTextScore = true;

                knowAllMeta = status & DocumentSource::EXHAUSTIVE_META;
            }
        }

        if (!knowAllFields)
            deps.needWholeDocument = true; // don't know all fields we need

        // NOTE This code assumes that textScore can only be generated by the initial query.
        if (DocumentSourceMatch::isTextQuery(initialQuery)) {
            // If doing a text query, assume we need the score if we can't prove we don't.
            if (!knowAllMeta)
                deps.needTextScore = true;
        }
        else {
            // If we aren't doing a text query, then we don't need to ask for the textScore since we
            // know it will be missing anyway.
            deps.needTextScore = false;
        }

        return deps;
    }
} // namespace mongo