fork of https://github.com/sourcegraph/zoekt
index: support searching for symbols
The implementation only supports substrings for now. It is achieved by
intersecting the candidate matches produced by trigram search with the
document sections.
Change-Id: I4f11a3574c7ec979cbbd4d35ac2459da7e318958
author
Han-Wen Nienhuys
date
(Jan 25, 2018, 10:21 AM +0100)
commit
b1998ff4
b1998ff47724c4436ba482d29798620a0f5d869b
parent
269454bc
269454bc2e1525e4028a6e4753a7ecbd86c3ed3d
+94
+44
index_test.go
+44
index_test.go
···
1570
1570
t.Fatalf("got %v, want 2 results in f3", res.Files)
1571
1571
}
1572
1572
}
1573
+
1574
+
func TestSymbolAtom(t *testing.T) {
1575
+
content := []byte("bla\nsymblabla\nbla")
1576
+
// ----------------0123 456789012
1577
+
1578
+
b := testIndexBuilder(t, &Repository{Name: "reponame"},
1579
+
Document{
1580
+
Name: "f1",
1581
+
Content: content,
1582
+
Symbols: []DocumentSection{{4, 12}},
1583
+
},
1584
+
)
1585
+
q := &query.Symbol{&query.Substring{Pattern: "bla"}}
1586
+
res := searchForTest(t, b, q)
1587
+
if len(res.Files) != 1 && len(res.Files[0].LineMatches) != 1 {
1588
+
t.Fatalf("got %v, want 1 line in 1 file", res.Files)
1589
+
}
1590
+
m := res.Files[0].LineMatches[0].LineFragments[0]
1591
+
if m.Offset != 7 || m.MatchLength != 3 {
1592
+
t.Fatalf("got offset %d, size %d want 7 size 3", m.Offset, m.MatchLength)
1593
+
}
1594
+
}
1595
+
1596
+
func TestSymbolAtomExact(t *testing.T) {
1597
+
content := []byte("bla\nsym\nbla\nsym\nasymb")
1598
+
// ----------------0123 4567 89012
1599
+
1600
+
b := testIndexBuilder(t, &Repository{Name: "reponame"},
1601
+
Document{
1602
+
Name: "f1",
1603
+
Content: content,
1604
+
Symbols: []DocumentSection{{4, 7}},
1605
+
},
1606
+
)
1607
+
q := &query.Symbol{&query.Substring{Pattern: "sym"}}
1608
+
res := searchForTest(t, b, q)
1609
+
if len(res.Files) != 1 && len(res.Files[0].LineMatches) != 1 {
1610
+
t.Fatalf("got %v, want 1 line in 1 file", res.Files)
1611
+
}
1612
+
m := res.Files[0].LineMatches[0].LineFragments[0]
1613
+
if m.Offset != 4 {
1614
+
t.Fatalf("got offset %d, want 7", m.Offset)
1615
+
}
1616
+
}
+50
matchtree.go
+50
matchtree.go
···
532
532
return &docMatchTree{
533
533
docs: docs,
534
534
}, nil
535
+
536
+
case *query.Symbol:
537
+
mt, err := d.newSubstringMatchTree(s.Atom, stats)
538
+
if err != nil {
539
+
return nil, err
540
+
}
541
+
542
+
if _, ok := mt.(*regexpMatchTree); ok {
543
+
return nil, fmt.Errorf("regexps and short queries not implemented for symbol search")
544
+
}
545
+
subMT, ok := mt.(*substrMatchTree)
546
+
if !ok {
547
+
return nil, fmt.Errorf("found %T inside query.Symbol", mt)
548
+
}
549
+
550
+
subMT.cands = d.trimByDocSection(s.Atom, subMT.cands, d.runeDocSections)
551
+
return subMT, nil
535
552
}
536
553
log.Panicf("type %T", q)
537
554
return nil, nil
···
565
582
stats.IndexBytesLoaded += int64(result.bytesRead)
566
583
return st, nil
567
584
}
585
+
586
+
func (d *indexData) trimByDocSection(q *query.Substring, ms []*candidateMatch, secs []DocumentSection) []*candidateMatch {
587
+
trimmed := ms[:0]
588
+
589
+
patSize := utf8.RuneCount([]byte(q.Pattern))
590
+
for len(secs) > 0 && len(ms) > 0 {
591
+
var fileStart uint32
592
+
if ms[0].file > 0 {
593
+
fileStart = d.fileEndRunes[ms[0].file-1]
594
+
}
595
+
596
+
start := fileStart + ms[0].runeOffset
597
+
end := start + uint32(patSize)
598
+
if start >= secs[0].End {
599
+
secs = secs[1:]
600
+
continue
601
+
}
602
+
603
+
if start < secs[0].Start {
604
+
ms = ms[1:]
605
+
continue
606
+
}
607
+
608
+
// here we have: sec.Start <= start < sec.End
609
+
if end <= secs[0].End {
610
+
// complete match falls inside section.
611
+
trimmed = append(trimmed, ms[0])
612
+
}
613
+
614
+
ms = ms[1:]
615
+
}
616
+
return trimmed
617
+
}