all repos

clerk @ cf9c1d930d43e1c0015820114a18abaf71ae60f3

missing tooling for ledger/hledger

clerk/internal/lsp/textdocument_semantic_tokens.go (view raw)

Oleksandr Smirnov Oleksandr Smirnov
olexsmir@gmail.com
remove duplication of token.Pos.File in token.Span, 1 month ago
1
package lsp
2
3
import (
4
	"context"
5
	"slices"
6
	"unicode/utf8"
7
8
	"go.lsp.dev/protocol"
9
	"go.lsp.dev/uri"
10
11
	"olexsmir.xyz/clerk/internal/lsp/lsputil"
12
	"olexsmir.xyz/clerk/journal/ast"
13
	"olexsmir.xyz/clerk/journal/lexer"
14
	"olexsmir.xyz/clerk/journal/token"
15
)
16
17
func (s *server) SemanticTokensFull(ctx context.Context, params *protocol.SemanticTokensParams) (*protocol.SemanticTokens, error) {
18
	if !s.semanticHighlightingEnabled() {
19
		return &protocol.SemanticTokens{}, nil
20
	}
21
	return s.semanticTokensFullResult(params.TextDocument.URI), nil
22
}
23
24
func (s *server) SemanticTokensFullDelta(ctx context.Context, params *protocol.SemanticTokensDeltaParams) (protocol.SemanticTokensDeltaResult, error) {
25
	if !s.semanticHighlightingEnabled() {
26
		return &protocol.SemanticTokens{}, nil
27
	}
28
29
	u := params.TextDocument.URI
30
	s.mu.RLock()
31
	st, ok := s.openDocs[u]
32
	s.mu.RUnlock()
33
	if !ok || st.semGen == 0 || params.PreviousResultID != st.resultID() {
34
		return s.semanticTokensFullResult(u), nil
35
	}
36
37
	data, ok := s.semanticTokensData(u)
38
	if !ok {
39
		return &protocol.SemanticTokens{}, nil
40
	}
41
	edits := semanticTokensEdits(st.semBaseline, data)
42
	if len(edits) == 0 {
43
		return &protocol.SemanticTokensDelta{ResultID: new(st.resultID()), Edits: []protocol.SemanticTokensEdit{}}, nil
44
	}
45
46
	rid, ok := s.storeSemResult(u, data)
47
	if !ok {
48
		return &protocol.SemanticTokens{Data: data}, nil
49
	}
50
	return &protocol.SemanticTokensDelta{ResultID: &rid, Edits: edits}, nil
51
}
52
53
func (s *server) SemanticTokensRange(ctx context.Context, params *protocol.SemanticTokensRangeParams) (*protocol.SemanticTokens, error) {
54
	if !s.semanticHighlightingEnabled() {
55
		return &protocol.SemanticTokens{}, nil
56
	}
57
58
	tokens, ok := s.tokensForDoc(params.TextDocument.URI)
59
	if !ok {
60
		return &protocol.SemanticTokens{}, nil
61
	}
62
	start := int(params.Range.Start.Line)
63
	end := int(params.Range.End.Line)
64
	var filtered []semanticToken
65
	for _, t := range tokens {
66
		if t.line >= uint32(start) && t.line <= uint32(end) {
67
			filtered = append(filtered, t)
68
		}
69
	}
70
	return &protocol.SemanticTokens{Data: encodeSemTokens(filtered)}, nil
71
}
72
73
func (s *server) semanticTokensFullResult(u uri.URI) *protocol.SemanticTokens {
74
	data, ok := s.semanticTokensData(u)
75
	if !ok {
76
		return &protocol.SemanticTokens{}
77
	}
78
	res := &protocol.SemanticTokens{Data: data}
79
	if rid, ok := s.storeSemResult(u, data); ok {
80
		res.ResultID = &rid
81
	}
82
	return res
83
}
84
85
func (s *server) semanticTokensData(u uri.URI) ([]uint32, bool) {
86
	tokens, ok := s.tokensForDoc(u)
87
	if !ok {
88
		return nil, false
89
	}
90
	return encodeSemTokens(tokens), true
91
}
92
93
func (s *server) storeSemResult(u uri.URI, data []uint32) (string, bool) {
94
	s.mu.Lock()
95
	defer s.mu.Unlock()
96
	st, ok := s.openDocs[u]
97
	if !ok {
98
		return "", false
99
	}
100
	rid := st.nextResultID()
101
	st.semBaseline = data
102
	s.openDocs[u] = st
103
	return rid, true
104
}
105
106
func (s *server) tokensForDoc(doc uri.URI) ([]semanticToken, bool) {
107
	s.mu.RLock()
108
	st, ok := s.openDocs[doc]
109
	s.mu.RUnlock()
110
	if !ok {
111
		return nil, false
112
	}
113
	if st.semTokens != nil {
114
		return st.semTokens, true
115
	}
116
	// Tokenize outside the lock: a full tokenization of a large journal is
117
	// milliseconds, during which didChange/didOpen would otherwise stall.
118
	tokens := tokenizeForSemantics(st.text, parseJournalStr(st.text))
119
	s.mu.Lock()
120
	defer s.mu.Unlock()
121
	cur, ok := s.openDocs[doc]
122
	if !ok {
123
		return nil, false
124
	}
125
	if cur.text == st.text { // unchanged during tokenization
126
		cur.semTokens = tokens
127
		s.openDocs[doc] = cur
128
	}
129
	return tokens, true
130
}
131
132
const (
133
	semDirective uint32 = iota
134
	semDate
135
	semAccount
136
	semCommodity
137
	semAmount
138
	semStatus
139
	semComment
140
	semString
141
	semOperator
142
	semProperty
143
)
144
145
var tokenTypeStrings = []string{
146
	string(protocol.SemanticTokenTypesKeyword),   // directive
147
	string(protocol.SemanticTokenTypesClass),     // date
148
	string(protocol.SemanticTokenTypesNamespace), // account
149
	string(protocol.SemanticTokenTypesType),      // commodity
150
	string(protocol.SemanticTokenTypesNumber),    // amount
151
	string(protocol.SemanticTokenTypesOperator),  // status
152
	string(protocol.SemanticTokenTypesComment),   // comment
153
	string(protocol.SemanticTokenTypesString),    // string
154
	string(protocol.SemanticTokenTypesOperator),  // operator
155
	string(protocol.SemanticTokenTypesProperty),  // property
156
}
157
158
const (
159
	modifierAbstract = 1 << 0 // virtual account
160
	modifierNegative = 1 << 1 // negative amount
161
)
162
163
var modifierStrings = []string{
164
	"abstract", // bit 0
165
	"negative", // bit 1
166
}
167
168
func getSemanticTokensLegend() protocol.SemanticTokensLegend {
169
	return protocol.SemanticTokensLegend{
170
		TokenTypes:     tokenTypeStrings,
171
		TokenModifiers: modifierStrings,
172
	}
173
}
174
175
type semanticToken struct {
176
	line, col uint32 // 0-based
177
	length    uint32
178
	tokenType uint32
179
	modifiers uint32
180
}
181
182
func tokenizeForSemantics(content string, j *ast.Journal) []semanticToken {
183
	var raw []rawSpan
184
	emit := func(s token.Span, tokType, mods uint32) {
185
		if s.Start.Offset >= s.End.Offset {
186
			return
187
		}
188
		raw = append(raw, rawSpan{s, tokType, mods})
189
	}
190
	for _, e := range j.Entries {
191
		visitEntry(content, e, emit)
192
	}
193
	if len(j.Errors) > 0 {
194
		// parser recovers per line; lexer fills the unparsed regions, keeping ast tokens where the parser succeeded
195
		semLexerFallback(content, raw, emit)
196
	}
197
	return rawToSemanticTokens(content, raw)
198
}
199
200
// rawSpan is a source span tagged with semantic token
201
type rawSpan struct {
202
	span      token.Span
203
	tok, mods uint32
204
}
205
206
func rawToSemanticTokens(content string, raw []rawSpan) []semanticToken {
207
	if len(raw) == 0 {
208
		return nil
209
	}
210
	slices.SortFunc(raw, func(a, b rawSpan) int { return a.span.Start.Offset - b.span.Start.Offset })
211
	out := make([]semanticToken, len(raw))
212
	line, col, cursor := 0, 0, 0
213
	advance := func(end int) {
214
		for cursor < end {
215
			r, size := utf8.DecodeRuneInString(content[cursor:])
216
			if r == utf8.RuneError && size <= 1 {
217
				break
218
			}
219
			if r == '\r' {
220
				cursor += size
221
				if cursor < len(content) && content[cursor] == '\n' {
222
					cursor++
223
				}
224
				line++
225
				col = 0
226
				continue
227
			}
228
			if r == '\n' {
229
				cursor += size
230
				line++
231
				col = 0
232
				continue
233
			}
234
			cursor += size
235
			col += utf16Units(r)
236
		}
237
	}
238
	for i, t := range raw {
239
		if cursor < t.span.Start.Offset {
240
			advance(t.span.Start.Offset)
241
		}
242
		out[i] = semanticToken{
243
			line:      uint32(line),
244
			col:       uint32(col),
245
			length:    uint32(lsputil.Utf16Len(content, t.span.Start.Offset, t.span.End.Offset)),
246
			tokenType: t.tok,
247
			modifiers: t.mods,
248
		}
249
		advance(t.span.End.Offset)
250
	}
251
	return out
252
}
253
254
func utf16Units(r rune) int {
255
	if r >= 0x10000 && r <= 0x10FFFF {
256
		return 2
257
	}
258
	return 1
259
}
260
261
func visitEntry(content string, e ast.Entry, emit semEmitFunc) {
262
	switch e := e.(type) {
263
	case *ast.Transaction:
264
		visitTransaction(content, e, emit)
265
	case *ast.PeriodicTransaction:
266
		visitPeriodicTransaction(content, e, emit)
267
	case *ast.AutomatedTransaction:
268
		visitAutomatedTransaction(content, e, emit)
269
	case *ast.AccountDirective:
270
		emit(directiveKeyword(e.Span, "account"), semDirective, 0)
271
		emit(e.Account.Span, semAccount, 0)
272
		for _, sd := range e.Subdirectives {
273
			if sd.Kind == ast.SubdirectiveComment {
274
				emitComment(sd.Comment, emit)
275
				continue
276
			}
277
			emit(sd.NameSpan, semDirective, 0)
278
			switch sd.Kind {
279
			case ast.SubdirectiveAlias:
280
				emit(sd.ValueSpan, semAccount, 0)
281
			case ast.SubdirectiveType, ast.SubdirectiveNote:
282
				emit(sd.ValueSpan, semProperty, 0)
283
			}
284
			emitComment(sd.Comment, emit)
285
		}
286
		emitComment(e.Comment, emit)
287
	case *ast.CommodityDirective:
288
		emit(directiveKeyword(e.Span, "commodity"), semDirective, 0)
289
		if e.FormatSub != nil {
290
			if e.FormatSub.KeywordSpan.End.Offset > 0 {
291
				emit(e.FormatSub.KeywordSpan, semDirective, 0)
292
			}
293
			semEmitAmount(content, &e.FormatSub.Amount, emit)
294
			emitComment(e.FormatSub.Comment, emit)
295
		} else if e.CommoditySpan.Start.Offset > 0 && e.CommoditySpan.End.Offset > 0 {
296
			emit(e.CommoditySpan, semCommodity, 0)
297
		}
298
		emitBlockComments(e.BlockComments, emit)
299
		emitComment(e.Comment, emit)
300
	case *ast.IncludeDirective:
301
		emitDirective(content, e.Span, len("include"), semString, e.Comment, emit)
302
	case *ast.PayeeDirective:
303
		emitDirective(content, e.Span, len("payee"), semProperty, e.Comment, emit)
304
	case *ast.TagDirective:
305
		emitDirective(content, e.Span, len("tag"), semProperty, e.Comment, emit)
306
	case *ast.AliasDirective:
307
		emit(directiveKeyword(e.Span, "alias"), semDirective, 0)
308
		emit(e.From.Span, semAccount, 0)
309
		if op, ok := betweenSpan(content, e.Span.File, e.From.Span.End.Offset, e.To.Span.Start.Offset); ok {
310
			emit(op, semOperator, 0)
311
		}
312
		emit(e.To.Span, semAccount, 0)
313
		emitComment(e.Comment, emit)
314
	case *ast.YearDirective:
315
		kwLen := len("year")
316
		if content[e.Span.Start.Offset] == 'Y' {
317
			kwLen = 1
318
		}
319
		emitDirective(content, e.Span, kwLen, semProperty, e.Comment, emit)
320
	case *ast.DecimalMarkDirective:
321
		emitDirective(content, e.Span, len("decimal-mark"), semProperty, e.Comment, emit)
322
	case *ast.DefaultCommodityDirective:
323
		emit(directiveKeyword(e.Span, "D"), semDirective, 0)
324
		semEmitAmount(content, &e.Amount, emit)
325
		emitComment(e.Comment, emit)
326
	case *ast.MarketPriceDirective:
327
		emit(directiveKeyword(e.Span, "P"), semDirective, 0)
328
		emit(e.DateTime.Date.Span, semDate, 0)
329
		if e.DateTime.Time != nil {
330
			emit(e.DateTime.Time.Span, semDate, 0)
331
		}
332
		// commodity: text between the date (or time) and the amount
333
		commStart := e.DateTime.Date.Span.End.Offset
334
		if e.DateTime.Time != nil {
335
			commStart = e.DateTime.Time.Span.End.Offset
336
		}
337
		if comm, ok := betweenSpan(content, e.Span.File, commStart, e.Amount.Span.Start.Offset); ok {
338
			emit(comm, semCommodity, 0)
339
		}
340
		semEmitAmount(content, &e.Amount, emit)
341
		emitComment(e.Comment, emit)
342
	case *ast.ConversionDirective:
343
		emit(directiveKeyword(e.Span, "C"), semDirective, 0)
344
		semEmitAmount(content, &e.From, emit)
345
		// = operator: text between the two amounts
346
		if op, ok := betweenSpan(content, e.Span.File, e.From.Span.End.Offset, e.To.Span.Start.Offset); ok {
347
			emit(op, semOperator, 0)
348
		}
349
		semEmitAmount(content, &e.To, emit)
350
		emitComment(e.Comment, emit)
351
	case *ast.Comment:
352
		emitComment(e, emit)
353
	case *ast.CommentBlockDirective:
354
		emit(e.Span, semComment, 0)
355
	case *ast.IgnoredDirective:
356
		emitDirective(content, e.Span, len("N"), semProperty, e.Comment, emit)
357
	case *ast.ApplyDirective:
358
		emitDirective(content, e.Span, len("apply"), semProperty, e.Comment, emit)
359
	case *ast.EndDirective:
360
		emitDirective(content, e.Span, len("end"), semProperty, e.Comment, emit)
361
	case *ast.BlankLine:
362
	}
363
}
364
365
func visitTransaction(content string, t *ast.Transaction, emit semEmitFunc) {
366
	emit(t.Date.Span, semDate, 0)
367
	if t.SecondDate != nil {
368
		emit(t.SecondDate.Span, semDate, 0)
369
	}
370
	if t.Status.Value != ast.StatusNone {
371
		emit(t.Status.Span, semStatus, 0)
372
	}
373
	if t.Code != nil {
374
		emit(t.Code.Span, semString, 0)
375
	}
376
	if t.Payee != nil {
377
		emit(t.Payee.Span, semProperty, 0)
378
	}
379
	if t.Note != nil {
380
		emit(t.Note.Span, semProperty, 0)
381
	}
382
	emitComment(t.Comment, emit)
383
	for i := range t.HeaderComments {
384
		emitComment(t.HeaderComments[i], emit)
385
	}
386
	for _, p := range t.Postings {
387
		visitPosting(content, p, emit)
388
	}
389
}
390
391
func visitPeriodicTransaction(content string, pt *ast.PeriodicTransaction, emit semEmitFunc) {
392
	// ~ operator is at the start of the period span
393
	emit(offsetSpan(pt.Span.File, pt.Span.Start.Offset, pt.Span.Start.Offset+1), semOperator, 0)
394
395
	// The period span covers the whole expr, including any "from ... to ..." dates
396
	if pt.Period.Span.End.Offset > pt.Period.Span.Start.Offset {
397
		var dates []*ast.Date
398
		if pt.Period.From != nil {
399
			dates = append(dates, pt.Period.From)
400
		}
401
		if pt.Period.To != nil {
402
			dates = append(dates, pt.Period.To)
403
		}
404
		pos := pt.Period.Span.Start.Offset
405
		for _, d := range dates {
406
			if d.Span.Start.Offset > pos {
407
				emit(offsetSpan(pt.Period.Span.File, pos, d.Span.Start.Offset), semProperty, 0)
408
			}
409
			emit(d.Span, semDate, 0)
410
			pos = d.Span.End.Offset
411
		}
412
		if pos < pt.Period.Span.End.Offset {
413
			emit(offsetSpan(pt.Period.Span.File, pos, pt.Period.Span.End.Offset), semProperty, 0)
414
		}
415
	}
416
	if pt.Description != nil {
417
		emit(pt.Description.Span, semProperty, 0)
418
	}
419
	emitComment(pt.Comment, emit)
420
	for i := range pt.HeaderComments {
421
		emitComment(pt.HeaderComments[i], emit)
422
	}
423
	for _, p := range pt.Postings {
424
		visitPosting(content, p, emit)
425
	}
426
}
427
428
func visitAutomatedTransaction(content string, at *ast.AutomatedTransaction, emit semEmitFunc) {
429
	// = operator is at the start of the expression span
430
	emit(offsetSpan(at.Span.File, at.Span.Start.Offset, at.Span.Start.Offset+1), semOperator, 0)
431
432
	if at.Expr.Value != "" {
433
		emit(at.Expr.Span, semString, 0)
434
	}
435
	emitComment(at.Comment, emit)
436
	for i := range at.HeaderComments {
437
		emitComment(at.HeaderComments[i], emit)
438
	}
439
	for _, p := range at.Postings {
440
		visitPosting(content, p, emit)
441
	}
442
}
443
444
func visitPosting(content string, p *ast.Posting, emit semEmitFunc) {
445
	if p.Status.Value != ast.StatusNone {
446
		emit(p.Status.Span, semStatus, 0)
447
	}
448
449
	// virtual brackets
450
	if p.Type == ast.PostingVirtualUnbalanced || p.Type == ast.PostingVirtualBalanced {
451
		// opening bracket
452
		for off := p.Span.Start.Offset; off < p.Account.Span.Start.Offset && off < p.Span.End.Offset; off++ {
453
			if content[off] == '(' || content[off] == '[' {
454
				emit(offsetSpan(p.Span.File, off, off+1), semOperator, modifierAbstract)
455
				break
456
			}
457
		}
458
		// closing bracket
459
		for off := p.Account.Span.End.Offset; off < p.Span.End.Offset; off++ {
460
			if content[off] == ')' || content[off] == ']' {
461
				emit(offsetSpan(p.Span.File, off, off+1), semOperator, modifierAbstract)
462
				break
463
			}
464
		}
465
	}
466
467
	emit(p.Account.Span, semAccount, 0)
468
469
	if p.Amount != nil {
470
		semEmitAmount(content, p.Amount, emit)
471
	}
472
	if p.Cost != nil {
473
		semEmitCost(content, p.Cost, emit)
474
	}
475
	if p.Balance != nil {
476
		semEmitBalanceAssertion(content, p.Balance, emit)
477
	}
478
	emitComment(p.Comment, emit)
479
	for i := range p.Comments {
480
		emitComment(&p.Comments[i], emit)
481
	}
482
}
483
484
// directiveKeyword returns the span of the leading keyword on a directive line.
485
func directiveKeyword(e token.Span, kw string) token.Span {
486
	return token.Span{File: e.File, Start: e.Start, End: token.Pos{Offset: e.Start.Offset + len(kw)}}
487
}
488
489
// directiveValue returns the trimmed span of the text after the keyword end
490
// offset, up to the inline comment or the end of the line.
491
func directiveValue(content string, e token.Span, comment *ast.Comment, kwEnd int) (token.Span, bool) {
492
	end := e.End.Offset
493
	if comment != nil {
494
		end = comment.Span.Start.Offset
495
	}
496
	return betweenSpan(content, e.File, kwEnd, end)
497
}
498
499
func semEmitAmount(content string, a *ast.Amount, emit semEmitFunc) {
500
	if a == nil {
501
		return
502
	}
503
	hasCommodity := a.CommoditySpan.Start.Offset > 0 && a.CommoditySpan.End.Offset > 0
504
	if hasCommodity && a.CommodityPos == ast.CommodityBefore {
505
		emit(a.CommoditySpan, semCommodity, 0)
506
		semEmitQuantity(content, a, emit)
507
		return
508
	}
509
	semEmitQuantity(content, a, emit)
510
	if hasCommodity {
511
		emit(a.CommoditySpan, semCommodity, 0)
512
	}
513
}
514
515
func semEmitQuantity(content string, a *ast.Amount, emit semEmitFunc) {
516
	qStart, qEnd := quantitySpan(content, a)
517
	if qEnd <= qStart {
518
		return
519
	}
520
	mods := uint32(0)
521
	if a.IsNegative {
522
		mods |= modifierNegative
523
	}
524
	emit(offsetSpan(a.Span.File, qStart, qEnd), semAmount, mods)
525
}
526
527
func quantitySpan(content string, a *ast.Amount) (int, int) {
528
	start, end := a.Span.Start.Offset, a.Span.End.Offset
529
	switch {
530
	case a.Commodity == "":
531
		// bare quantity
532
	case a.CommodityPos == ast.CommodityBefore:
533
		// "$50.00" or "$   50.00": quantity follows the commodity span
534
		start = a.CommoditySpan.End.Offset
535
	default: // CommodityAfter
536
		// "50.00 USD" or "50.00     USD": quantity precedes the commodity
537
		end = a.CommoditySpan.Start.Offset
538
	}
539
	for start < end && (content[start] == ' ' || content[start] == '\t') {
540
		start++
541
	}
542
	for end > start && (content[end-1] == ' ' || content[end-1] == '\t') {
543
		end--
544
	}
545
	return start, end
546
}
547
548
type semEmitFunc func(tok token.Span, tokKind, modifier uint32)
549
550
func emitBlockComments(cs []*ast.Comment, emit semEmitFunc) {
551
	for _, c := range cs {
552
		emitComment(c, emit)
553
	}
554
}
555
556
func emitComment(c *ast.Comment, emit semEmitFunc) {
557
	if c == nil {
558
		return
559
	}
560
	if len(c.Tags) == 0 {
561
		emit(c.Span, semComment, 0)
562
		return
563
	}
564
	pos := c.Span.Start.Offset
565
	for _, t := range c.Tags {
566
		if t.Span.Start.Offset > pos {
567
			emit(offsetSpan(c.Span.File, pos, t.Span.Start.Offset), semComment, 0)
568
		}
569
		emit(t.Span, semProperty, 0)
570
		pos = t.Span.End.Offset
571
	}
572
	if pos < c.Span.End.Offset {
573
		emit(offsetSpan(c.Span.File, pos, c.Span.End.Offset), semComment, 0)
574
	}
575
}
576
577
func emitDirective(content string, e token.Span, kwLen int, valType uint32, comment *ast.Comment, emit semEmitFunc) {
578
	kwEnd := e.Start.Offset + kwLen
579
	emit(token.Span{File: e.File, Start: e.Start, End: token.Pos{Offset: kwEnd}}, semDirective, 0)
580
	if v, ok := directiveValue(content, e, comment, kwEnd); ok {
581
		emit(v, valType, 0)
582
	}
583
	emitComment(comment, emit)
584
}
585
586
func semEmitCost(content string, c *ast.Cost, emit semEmitFunc) {
587
	if c.IsTotal {
588
		emit(token.Span{File: c.Span.File, Start: c.Span.Start, End: token.Pos{Offset: c.Span.Start.Offset + 2}}, semOperator, 0)
589
	} else {
590
		emit(token.Span{File: c.Span.File, Start: c.Span.Start, End: token.Pos{Offset: c.Span.Start.Offset + 1}}, semOperator, 0)
591
	}
592
	semEmitAmount(content, &c.Amount, emit)
593
}
594
595
func semEmitBalanceAssertion(content string, ba *ast.BalanceAssertion, emit semEmitFunc) {
596
	// The operator is the run of '=', ':', '*' chars from the span start.
597
	// The ':' of ':=' precedes the '=' token, so back up one offset.
598
	opStart := ba.Span.Start.Offset
599
	if ba.IsAssignment && opStart > 0 {
600
		opStart--
601
	}
602
	opEnd := opStart
603
	for opEnd < ba.Span.End.Offset && (content[opEnd] == '=' || content[opEnd] == ':' || content[opEnd] == '*') {
604
		opEnd++
605
	}
606
	emit(token.Span{File: ba.Span.File, Start: token.Pos{Offset: opStart}, End: token.Pos{Offset: opEnd}}, semOperator, 0)
607
	semEmitAmount(content, &ba.Amount, emit)
608
	if ba.Cost != nil {
609
		semEmitCost(content, ba.Cost, emit)
610
	}
611
}
612
613
func semLexerFallback(content string, base []rawSpan, emit semEmitFunc) {
614
	l := lexer.New("", []byte(content))
615
616
	var commentStart, commentEnd int // 0 = not inside a comment line
617
	lineStart := true                // the next significant token starts a line
618
	skipLine := false                // the line starts with an unclassifiable token; emit nothing
619
	i := 0                           // next base span to compare against
620
621
	take := func(span token.Span, tokType uint32, mods uint32) {
622
		for i < len(base) && base[i].span.End.Offset <= span.Start.Offset {
623
			i++
624
		}
625
		if i < len(base) && base[i].span.Start.Offset < span.End.Offset {
626
			return // overlaps an AST token; AST wins
627
		}
628
		emit(span, tokType, mods)
629
	}
630
631
	for {
632
		tok := l.Next()
633
		if tok.Type == token.EOF {
634
			if commentStart > 0 {
635
				take(token.Span{Start: token.Pos{Offset: commentStart}, End: token.Pos{Offset: commentEnd}}, semComment, 0)
636
			}
637
			break
638
		}
639
		if tok.Type == token.NEWLINE {
640
			if commentStart > 0 {
641
				take(token.Span{Start: token.Pos{Offset: commentStart}, End: token.Pos{Offset: commentEnd}}, semComment, 0)
642
				commentStart, commentEnd = 0, 0
643
			}
644
			lineStart, skipLine = true, false
645
			continue
646
		}
647
		if tok.Type == token.WHITESPACE || tok.Type == token.INDENT {
648
			continue
649
		}
650
		if lineStart {
651
			lineStart = false
652
			if !isLineStartToken(tok.Type) {
653
				skipLine = true
654
			}
655
		}
656
		if skipLine {
657
			continue
658
		}
659
		tokType := semProperty
660
		if commentStart > 0 {
661
			tokType = semComment
662
			if tok.Span.End.Offset > commentEnd {
663
				commentEnd = tok.Span.End.Offset
664
			}
665
			continue
666
		}
667
668
		switch tok.Type {
669
		case token.SEMICOLON, token.HASH, token.PERCENT, token.STAR:
670
			tokType = semComment
671
			commentStart = tok.Span.Start.Offset
672
			commentEnd = tok.Span.End.Offset
673
			continue
674
		case token.STRING:
675
			tokType = semString
676
		case token.DATE:
677
			tokType = semDate
678
		case token.INT, token.DECIMAL:
679
			tokType = semAmount
680
		case token.COMMODITYMARK:
681
			tokType = semCommodity
682
		case token.BANG:
683
			tokType = semStatus
684
		case token.AT, token.ATAT, token.EQ, token.EQEQ, token.EQEQEQ, token.EQSTAR:
685
			tokType = semOperator
686
		case token.COMMENTKW, token.ACCOUNT, token.COMMODITY, token.INCLUDE,
687
			token.ALIAS, token.PAYEE, token.TAG, token.APPLY, token.END,
688
			token.YEAR, token.DECIMALMARK, token.D, token.P, token.N, token.C:
689
			tokType = semDirective
690
		}
691
		take(tok.Span, tokType, 0)
692
	}
693
}
694
695
func isLineStartToken(t token.Type) bool {
696
	switch t {
697
	case token.DATE, token.TILDE, token.EQ, token.BANG, token.AT,
698
		token.SEMICOLON, token.HASH, token.PERCENT, token.STAR,
699
		token.COMMENTKW, token.ACCOUNT, token.COMMODITY, token.INCLUDE,
700
		token.ALIAS, token.PAYEE, token.TAG, token.APPLY, token.END,
701
		token.YEAR, token.DECIMALMARK, token.D, token.P, token.N, token.C:
702
		return true
703
	}
704
	return false
705
}
706
707
// semanticTokensEdits returns the single edit turning old into new, or nil when
708
// identical. Relative delta encoding keeps the common prefix and suffix unchanged.
709
func semanticTokensEdits(old, new []uint32) []protocol.SemanticTokensEdit {
710
	p := 0
711
	for p < len(old) && p < len(new) && old[p] == new[p] {
712
		p++
713
	}
714
	s := 0
715
	for s < len(old)-p && s < len(new)-p && old[len(old)-1-s] == new[len(new)-1-s] {
716
		s++
717
	}
718
	delCount := len(old) - p - s
719
	ins := new[p : len(new)-s]
720
	if delCount == 0 && len(ins) == 0 {
721
		return nil
722
	}
723
	return []protocol.SemanticTokensEdit{{
724
		Start:       uint32(p),
725
		DeleteCount: uint32(delCount),
726
		Data:        ins,
727
	}}
728
}
729
730
// encodeSemTokens encodes tokens into LSP delta form. Input must be sorted by
731
// line and column; [rawToSemanticTokens] produces such order.
732
func encodeSemTokens(tokens []semanticToken) []uint32 {
733
	if len(tokens) == 0 {
734
		return nil
735
	}
736
	data := make([]uint32, 0, len(tokens)*5)
737
	var prevLine, prevCol uint32
738
	for _, t := range tokens {
739
		var deltaLine, deltaCol uint32
740
		if t.line == prevLine {
741
			deltaLine = 0
742
			deltaCol = t.col - prevCol
743
		} else {
744
			deltaLine = t.line - prevLine
745
			deltaCol = t.col
746
		}
747
		data = append(data, deltaLine, deltaCol, t.length, t.tokenType, t.modifiers)
748
		prevLine = t.line
749
		prevCol = t.col
750
	}
751
	return data
752
}
753
754
// betweenSpan returns the span of the text between two offsets, trimmed of surrounding whitespace.
755
func betweenSpan(content, file string, start, end int) (token.Span, bool) {
756
	for start < end && (content[start] == ' ' || content[start] == '\t') {
757
		start++
758
	}
759
	for end > start && (content[end-1] == ' ' || content[end-1] == '\t' || content[end-1] == '\n' || content[end-1] == '\r') {
760
		end--
761
	}
762
	if end <= start {
763
		return token.Span{}, false
764
	}
765
	return offsetSpan(file, start, end), true
766
}
767
768
func offsetSpan(file string, start, end int) token.Span {
769
	return token.Span{File: file, Start: token.Pos{Offset: start}, End: token.Pos{Offset: end}}
770
}