fork of https://github.com/sourcegraph/zoekt
score: boost exported go ident and downrank _test.go (#675)
Right now our symbol analyser doesn't tell us if a symbol is exported. We add
a go specific tweak here to boost those results. Ideally this could be
something that is encoded in the symbol information.
Additionally we do downrank _test.go files via the doc-order. But in the case
of symbol matches the boosting overweighs doc order signficantly. I found the
extra downraking quite useful when experimenting.
Test Plan: lots of manual testing on the keyword branch
author
Keegan Carruthers-Smith
committer
GitHub
date
(Oct 26, 2023, 1:20 PM +0200)
commit
b5a5fdc8
b5a5fdc86e830d80b2be3154bb9f896fd20c3e3e
parent
1a3dddce
1a3dddce84a769a32b83f98debe60ed02988f8d0
+42
-9
+2
-2
build/e2e_test.go
+2
-2
build/e2e_test.go
···
1041
1041
&query.Symbol{Expr: &query.Substring{Pattern: "http", Content: true}},
1042
1042
&query.Symbol{Expr: &query.Substring{Pattern: "Get", Content: true}}}},
1043
1043
wantLanguage: "Go",
1044
-
// 7000 (full base match) + 800 (Go func) + 500 (word) + 200 (atom) + 10 (file order)
1045
-
wantScore: 8510,
1044
+
// 7000 (full base match) + 800 (Go func) + 50 (Exported Go) + 500 (word) + 200 (atom) + 10 (file order)
1045
+
wantScore: 8560,
1046
1046
},
1047
1047
//
1048
1048
// C++
+40
-7
contentprovider.go
+40
-7
contentprovider.go
···
22
22
"path"
23
23
"sort"
24
24
"strings"
25
+
"unicode"
25
26
"unicode/utf8"
26
27
27
28
"golang.org/x/exp/slices"
···
510
511
score.score += s
511
512
}
512
513
514
+
data := p.data(m.FileName)
515
+
filename := p.data(true)
516
+
513
517
for i, r := range m.Ranges {
514
518
// calculate the start and end offset relative to the start of the content
515
519
relStartOffset := int(r.Start.ByteOffset - m.ContentStart.ByteOffset)
···
560
564
si = p.id.symbols.data(start + uint32(secIdx))
561
565
}
562
566
if si != nil {
563
-
addScore(fmt.Sprintf("kind:%s:%s", language, si.Kind), scoreKind(language, si.Kind))
567
+
sym := sectionSlice(data, sec)
568
+
addScore(fmt.Sprintf("kind:%s:%s", language, si.Kind), scoreSymbolKind(language, filename, sym, si.Kind))
564
569
}
565
570
}
566
571
···
593
598
score.score += s
594
599
}
595
600
601
+
data := p.data(m.FileName)
602
+
filename := p.data(true)
603
+
596
604
for _, f := range m.LineFragments {
597
605
startBoundary := f.LineOffset < len(m.Line) && (f.LineOffset == 0 || byteClass(m.Line[f.LineOffset-1]) != byteClass(m.Line[f.LineOffset]))
598
606
···
639
647
}
640
648
if si != nil {
641
649
// the LineFragment may not be on a symbol, then si will be nil.
642
-
addScore(fmt.Sprintf("kind:%s:%s", language, si.Kind), scoreKind(language, si.Kind))
650
+
sym := sectionSlice(data, sec)
651
+
addScore(fmt.Sprintf("kind:%s:%s", language, si.Kind), scoreSymbolKind(language, filename, sym, si.Kind))
643
652
}
644
653
}
645
654
···
656
665
return maxScore.score, maxScore.what
657
666
}
658
667
659
-
// scoreKind boosts a match based on the combination of language and kind. The
660
-
// language string comes from go-enry, the kind string from ctags.
661
-
func scoreKind(language string, kind string) float64 {
668
+
// sectionSlice will return data[sec.Start:sec.End] but will clip Start and
669
+
// End such that it won't be out of range.
670
+
func sectionSlice(data []byte, sec DocumentSection) []byte {
671
+
l := uint32(len(data))
672
+
if sec.Start >= l {
673
+
return nil
674
+
}
675
+
if sec.End > l {
676
+
sec.End = l
677
+
}
678
+
return data[sec.Start:sec.End]
679
+
}
680
+
681
+
// scoreSymbolKind boosts a match based on the combination of language, symbol
682
+
// and kind. The language string comes from go-enry, the symbol and kind from
683
+
// ctags.
684
+
func scoreSymbolKind(language string, filename []byte, sym []byte, kind string) float64 {
662
685
var factor float64
663
686
664
687
// Generic ranking which will be overriden by language specific ranking
···
731
754
// scip-ctags regression workaround https://github.com/sourcegraph/sourcegraph/issues/57659
732
755
// for each case a description of the fields in ctags in the comment
733
756
case "type": // interface struct talias
734
-
factor = 10
735
-
case "method", "function": // methodSpec
757
+
factor = 9
758
+
case "method", "function": // methodSpec func
736
759
factor = 8
737
760
case "variable": // var member
738
761
factor = 7
···
756
779
case "var": // variables
757
780
factor = 5
758
781
}
782
+
783
+
// Boost exported go symbols. Same implementation as token.IsExported
784
+
if ch, _ := utf8.DecodeRune(sym); unicode.IsUpper(ch) {
785
+
factor += 0.5
786
+
}
787
+
788
+
if bytes.HasSuffix(filename, []byte("_test.go")) {
789
+
factor *= 0.8
790
+
}
791
+
759
792
// Could also rank on:
760
793
//
761
794
// - anonMember struct anonymous members