func()

in eval.go [76:274]

• 165 lines of code
• 34 McCabe index (conditional complexity)

func (d *indexData) Search(ctx context.Context, q query.Q, opts *SearchOptions) (sr *SearchResult, err error) {
	copyOpts := *opts
	opts = &copyOpts
	opts.SetDefaults()
	importantMatchCount := 0

	var res SearchResult
	if len(d.fileNameIndex) == 0 {
		return &res, nil
	}

	select {
	case <-ctx.Done():
		res.Stats.ShardsSkipped++
		return &res, nil
	default:
	}

	tr := trace.New("indexData.Search", d.file.Name())
	tr.LazyPrintf("opts: %+v", opts)
	defer func() {
		if sr != nil {
			tr.LazyPrintf("num files: %d", len(sr.Files))
			tr.LazyPrintf("stats: %+v", sr.Stats)
		}
		if err != nil {
			tr.LazyPrintf("error: %v", err)
			tr.SetError()
		}
		tr.Finish()
	}()

	q = d.simplify(q)
	tr.LazyLog(q, true)
	if c, ok := q.(*query.Const); ok && !c.Value {
		return &res, nil
	}

	if opts.EstimateDocCount {
		res.Stats.ShardFilesConsidered = len(d.fileBranchMasks)
		return &res, nil
	}

	q = query.Map(q, query.ExpandFileContent)

	mt, err := d.newMatchTree(q)
	if err != nil {
		return nil, err
	}

	totalAtomCount := 0
	visitMatchTree(mt, func(t matchTree) {
		totalAtomCount++
	})

	cp := &contentProvider{
		id:    d,
		stats: &res.Stats,
	}

	docCount := uint32(len(d.fileBranchMasks))
	lastDoc := int(-1)

nextFileMatch:
	for {
		canceled := false
		select {
		case <-ctx.Done():
			canceled = true
		default:
		}

		nextDoc := mt.nextDoc()
		if int(nextDoc) <= lastDoc {
			nextDoc = uint32(lastDoc + 1)
		}
		if nextDoc >= docCount {
			break
		}
		lastDoc = int(nextDoc)

		if canceled || (res.Stats.MatchCount >= opts.ShardMaxMatchCount && opts.ShardMaxMatchCount > 0) ||
			(opts.ShardMaxImportantMatch > 0 && importantMatchCount >= opts.ShardMaxImportantMatch) {
			res.Stats.FilesSkipped += d.repoListEntry.Stats.Documents - lastDoc
			break
		}

		res.Stats.FilesConsidered++
		mt.prepare(nextDoc)

		cp.setDocument(nextDoc)

		known := make(map[matchTree]bool)
		for cost := costMin; cost <= costMax; cost++ {
			v, ok := mt.matches(cp, cost, known)
			if ok && !v {
				continue nextFileMatch
			}

			if cost == costMax && !ok {
				log.Panicf("did not decide. Repo %s, doc %d, known %v",
					d.repoMetaData.Name, nextDoc, known)
			}
		}

		fileMatch := FileMatch{
			Repository: d.repoMetaData.Name,
			FileName:   string(d.fileName(nextDoc)),
			Checksum:   d.getChecksum(nextDoc),
			Language:   d.languageMap[d.languages[nextDoc]],
		}

		if s := d.subRepos[nextDoc]; s > 0 {
			if s >= uint32(len(d.subRepoPaths)) {
				log.Panicf("corrupt index: subrepo %d beyond %v", s, d.subRepoPaths)
			}
			path := d.subRepoPaths[s]
			fileMatch.SubRepositoryPath = path
			sr := d.repoMetaData.SubRepoMap[path]
			fileMatch.SubRepositoryName = sr.Name
			if idx := d.branchIndex(nextDoc); idx >= 0 {
				fileMatch.Version = sr.Branches[idx].Version
			}
		} else {
			idx := d.branchIndex(nextDoc)
			if idx >= 0 {
				fileMatch.Version = d.repoMetaData.Branches[idx].Version
			}
		}

		atomMatchCount := 0
		visitMatches(mt, known, func(mt matchTree) {
			atomMatchCount++
		})
		finalCands := gatherMatches(mt, known)

		if len(finalCands) == 0 {
			nm := d.fileName(nextDoc)
			finalCands = append(finalCands,
				&candidateMatch{
					caseSensitive: false,
					fileName:      true,
					substrBytes:   nm,
					substrLowered: nm,
					file:          nextDoc,
					runeOffset:    0,
					byteOffset:    0,
					byteMatchSz:   uint32(len(nm)),
				})
		}
		fileMatch.LineMatches = cp.fillMatches(finalCands)

		maxFileScore := 0.0
		for i := range fileMatch.LineMatches {
			if maxFileScore < fileMatch.LineMatches[i].Score {
				maxFileScore = fileMatch.LineMatches[i].Score
			}

			// Order by ordering in file.
			fileMatch.LineMatches[i].Score += scoreLineOrderFactor * (1.0 - (float64(i) / float64(len(fileMatch.LineMatches))))
		}

		// Maintain ordering of input files. This
		// strictly dominates the in-file ordering of
		// the matches.
		fileMatch.addScore("fragment", maxFileScore)
		fileMatch.addScore("atom", float64(atomMatchCount)/float64(totalAtomCount)*scoreFactorAtomMatch)

		// Prefer earlier docs.
		fileMatch.addScore("doc-order", scoreFileOrderFactor*(1.0-float64(nextDoc)/float64(len(d.boundaries))))
		fileMatch.addScore("shard-order", scoreShardRankFactor*float64(d.repoMetaData.Rank)/maxUInt16)

		if fileMatch.Score > scoreImportantThreshold {
			importantMatchCount++
		}
		fileMatch.Branches = d.gatherBranches(nextDoc, mt, known)
		sortMatchesByScore(fileMatch.LineMatches)
		if opts.Whole {
			fileMatch.Content = cp.data(false)
		}

		res.Files = append(res.Files, fileMatch)
		res.Stats.MatchCount += len(fileMatch.LineMatches)
		res.Stats.FileCount++
	}
	SortFilesByScore(res.Files)

	addRepo(&res, &d.repoMetaData)
	for _, v := range d.repoMetaData.SubRepoMap {
		addRepo(&res, v)
	}

	visitMatchTree(mt, func(mt matchTree) {
		if atom, ok := mt.(interface{ updateStats(*Stats) }); ok {
			atom.updateStats(&res.Stats)
		}
	})
	return &res, nil
}