fork of https://github.com/sourcegraph/zoekt
debug: add debug score for line matches (#351)
We enrich the score of a line match with more detailed information of how it was
calculated. This is similar to what we already do on the level of file matches.
author
Stefan Hengl
committer
GitHub
date
(May 24, 2022, 11:08 AM +0200)
commit
f2bb82f7
f2bb82f75b5fb748ac68d1408a503f1b11541d28
parent
a3e0a5fd
a3e0a5fd11355b80d7da83c1a9557e42342f9142
+48
-23
+3
-1
api.go
+3
-1
api.go
+36
-17
contentprovider.go
+36
-17
contentprovider.go
···
16
16
17
17
import (
18
18
"bytes"
19
+
"fmt"
19
20
"log"
20
21
"sort"
22
+
"strings"
21
23
"unicode/utf8"
22
24
)
23
25
···
130
132
return byteOff
131
133
}
132
134
133
-
func (p *contentProvider) fillMatches(ms []*candidateMatch, numContextLines int, language string) []LineMatch {
135
+
func (p *contentProvider) fillMatches(ms []*candidateMatch, numContextLines int, language string, debug bool) []LineMatch {
134
136
var result []LineMatch
135
137
if ms[0].fileName {
136
138
// There is only "line" in a filename.
···
155
157
156
158
sects := p.docSections()
157
159
for i, m := range result {
158
-
result[i].Score = p.matchScore(sects, &m, language)
160
+
result[i].Score, result[i].DebugScore = p.matchScore(sects, &m, language, debug)
159
161
}
160
162
161
163
return result
···
296
298
return 0, false
297
299
}
298
300
299
-
func (p *contentProvider) matchScore(secs []DocumentSection, m *LineMatch, language string) float64 {
300
-
var maxScore float64
301
+
func (p *contentProvider) matchScore(secs []DocumentSection, m *LineMatch, language string, debug bool) (float64, string) {
302
+
type debugScore struct {
303
+
score float64
304
+
what string
305
+
}
306
+
307
+
score := &debugScore{}
308
+
maxScore := &debugScore{}
309
+
310
+
addScore := func(what string, s float64) {
311
+
if debug {
312
+
score.what += fmt.Sprintf("%s:%f, ", what, s)
313
+
}
314
+
score.score += s
315
+
}
316
+
301
317
for _, f := range m.LineFragments {
302
318
startBoundary := f.LineOffset < len(m.Line) && (f.LineOffset == 0 || byteClass(m.Line[f.LineOffset-1]) != byteClass(m.Line[f.LineOffset]))
303
319
304
320
end := int(f.LineOffset) + f.MatchLength
305
321
endBoundary := end > 0 && (end == len(m.Line) || byteClass(m.Line[end-1]) != byteClass(m.Line[end]))
306
322
307
-
score := 0.0
323
+
score.score = 0
324
+
score.what = ""
325
+
308
326
if startBoundary && endBoundary {
309
-
score = scoreWordMatch
327
+
addScore("WordMatch", scoreWordMatch)
310
328
} else if startBoundary || endBoundary {
311
-
score = scorePartialWordMatch
329
+
addScore("PartialWordMatch", scorePartialWordMatch)
312
330
}
313
331
314
332
if m.FileName {
···
316
334
startMatch := sep+1 == f.LineOffset
317
335
endMatch := len(m.Line) == f.LineOffset+f.MatchLength
318
336
if startMatch && endMatch {
319
-
score += scoreBase
337
+
addScore("Base", scoreBase)
320
338
} else if startMatch || endMatch {
321
-
score += (scoreBase + scorePartialBase) / 2
339
+
addScore("EdgeBase", (scoreBase+scorePartialBase)/2)
322
340
} else if sep < f.LineOffset {
323
-
score += scorePartialBase
341
+
addScore("InnerBase", scorePartialBase)
324
342
}
325
343
} else if secIdx, ok := findSection(secs, f.Offset, uint32(f.MatchLength)); ok {
326
344
sec := secs[secIdx]
327
345
startMatch := sec.Start == f.Offset
328
346
endMatch := sec.End == f.Offset+uint32(f.MatchLength)
329
347
if startMatch && endMatch {
330
-
score += scoreSymbol
348
+
addScore("Symbol", scoreSymbol)
331
349
} else if startMatch || endMatch {
332
-
score += (scoreSymbol + scorePartialSymbol) / 2
350
+
addScore("EdgeSymbol", (scoreSymbol+scorePartialSymbol)/2)
333
351
} else {
334
-
score += scorePartialSymbol
352
+
addScore("InnerSymbol", scorePartialSymbol)
335
353
}
336
354
337
355
si := f.SymbolInfo
···
342
360
}
343
361
if si != nil {
344
362
// the LineFragment may not be on a symbol, then si will be nil.
345
-
score += scoreKind(language, si.Kind)
363
+
addScore(fmt.Sprintf("kind:%s:%s", language, si.Kind), scoreKind(language, si.Kind))
346
364
}
347
365
}
348
366
349
-
if score > maxScore {
350
-
maxScore = score
367
+
if score.score > maxScore.score {
368
+
maxScore.score = score.score
369
+
maxScore.what = score.what
351
370
}
352
371
}
353
-
return maxScore
372
+
return maxScore.score, strings.TrimRight(maxScore.what, ", ")
354
373
}
355
374
356
375
// scoreKind boosts a match based on the combination of language and kind. The
+3
-2
eval.go
+3
-2
eval.go
···
23
23
"strings"
24
24
25
25
enry_data "github.com/go-enry/go-enry/v2/data"
26
+
"github.com/grafana/regexp"
27
+
26
28
"github.com/google/zoekt/query"
27
-
"github.com/grafana/regexp"
28
29
)
29
30
30
31
const maxUInt16 = 0xffff
···
337
338
byteMatchSz: uint32(len(nm)),
338
339
})
339
340
}
340
-
fileMatch.LineMatches = cp.fillMatches(finalCands, opts.NumContextLines, fileMatch.Language)
341
+
fileMatch.LineMatches = cp.fillMatches(finalCands, opts.NumContextLines, fileMatch.Language, opts.DebugScore)
341
342
342
343
maxFileScore := 0.0
343
344
for i := range fileMatch.LineMatches {
+2
-1
web/api.go
+2
-1
web/api.go
+3
-1
web/snippets.go
+3
-1
web/snippets.go
+1
-1
web/templates.go
+1
-1
web/templates.go
···
241
241
{{range .Matches}}
242
242
<tr>
243
243
<td style="background-color: rgba(238, 238, 255, 0.6);">
244
-
<pre class="inline-pre"><span class="noselect">{{if .URL}}<a href="{{.URL}}">{{end}}<u>{{.LineNum}}</u>{{if .URL}}</a>{{end}}: </span>{{range .Fragments}}{{LimitPre 100 .Pre}}<b>{{.Match}}</b>{{LimitPost 100 .Post}}{{end}} {{if $showScoreDebug}}<i>(score:{{.Score}})</i>{{end}}</pre>
244
+
<pre class="inline-pre"><span class="noselect">{{if .URL}}<a href="{{.URL}}">{{end}}<u>{{.LineNum}}</u>{{if .URL}}</a>{{end}}: </span>{{range .Fragments}}{{LimitPre 100 .Pre}}<b>{{.Match}}</b>{{LimitPost 100 .Post}}{{end}} {{if $showScoreDebug}}<i>(score:{{.Score}} <-- {{.ScoreDebug}})</i>{{end}}</pre>
245
245
</td>
246
246
</tr>
247
247
{{end}}