clerk/internal/lsp/textdocument_semantic_tokens.go (view raw)
Oleksandr Smirnov
Oleksandr Smirnov
olexsmir@gmail.com ast: inline leaf value+span wrapper structs, 1 month ago
olexsmir@gmail.com ast: inline leaf value+span wrapper structs, 1 month ago
| 1 | package lsp |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "slices" |
| 6 | "strconv" |
| 7 | "unicode/utf8" |
| 8 | |
| 9 | "go.lsp.dev/protocol" |
| 10 | "go.lsp.dev/uri" |
| 11 | |
| 12 | "olexsmir.xyz/clerk/internal/lsp/lsputil" |
| 13 | "olexsmir.xyz/clerk/journal/ast" |
| 14 | "olexsmir.xyz/clerk/journal/lexer" |
| 15 | "olexsmir.xyz/clerk/journal/token" |
| 16 | ) |
| 17 | |
| 18 | func (s *server) SemanticTokensFull(ctx context.Context, params *protocol.SemanticTokensParams) (*protocol.SemanticTokens, error) { |
| 19 | if !s.semanticHighlightingEnabled() { |
| 20 | return &protocol.SemanticTokens{}, nil |
| 21 | } |
| 22 | return s.semanticTokensFullResult(params.TextDocument.URI), nil |
| 23 | } |
| 24 | |
| 25 | func (s *server) SemanticTokensFullDelta(ctx context.Context, params *protocol.SemanticTokensDeltaParams) (protocol.SemanticTokensDeltaResult, error) { |
| 26 | if !s.semanticHighlightingEnabled() { |
| 27 | return &protocol.SemanticTokens{}, nil |
| 28 | } |
| 29 | |
| 30 | u := params.TextDocument.URI |
| 31 | s.mu.RLock() |
| 32 | st, ok := s.openDocs[u] |
| 33 | s.mu.RUnlock() |
| 34 | if !ok || st.sem == nil || st.sem.gen == 0 || params.PreviousResultID != st.sem.resultID() { |
| 35 | return s.semanticTokensFullResult(u), nil |
| 36 | } |
| 37 | |
| 38 | data, ok := s.semanticTokensData(u) |
| 39 | if !ok { |
| 40 | return &protocol.SemanticTokens{}, nil |
| 41 | } |
| 42 | edits := semanticTokensEdits(st.sem.baseline, data) |
| 43 | if len(edits) == 0 { |
| 44 | return &protocol.SemanticTokensDelta{ResultID: new(st.sem.resultID()), Edits: []protocol.SemanticTokensEdit{}}, nil |
| 45 | } |
| 46 | |
| 47 | rid, ok := s.storeSemResult(u, data) |
| 48 | if !ok { |
| 49 | return &protocol.SemanticTokens{Data: data}, nil |
| 50 | } |
| 51 | return &protocol.SemanticTokensDelta{ResultID: &rid, Edits: edits}, nil |
| 52 | } |
| 53 | |
| 54 | func (s *server) SemanticTokensRange(ctx context.Context, params *protocol.SemanticTokensRangeParams) (*protocol.SemanticTokens, error) { |
| 55 | if !s.semanticHighlightingEnabled() { |
| 56 | return &protocol.SemanticTokens{}, nil |
| 57 | } |
| 58 | |
| 59 | tokens, ok := s.tokensForDoc(params.TextDocument.URI) |
| 60 | if !ok { |
| 61 | return &protocol.SemanticTokens{}, nil |
| 62 | } |
| 63 | start := int(params.Range.Start.Line) |
| 64 | end := int(params.Range.End.Line) |
| 65 | var filtered []semanticToken |
| 66 | for _, t := range tokens { |
| 67 | if t.line >= uint32(start) && t.line <= uint32(end) { |
| 68 | filtered = append(filtered, t) |
| 69 | } |
| 70 | } |
| 71 | return &protocol.SemanticTokens{Data: encodeSemTokens(filtered)}, nil |
| 72 | } |
| 73 | |
| 74 | func (s *server) semanticTokensFullResult(u uri.URI) *protocol.SemanticTokens { |
| 75 | data, ok := s.semanticTokensData(u) |
| 76 | if !ok { |
| 77 | return &protocol.SemanticTokens{} |
| 78 | } |
| 79 | res := &protocol.SemanticTokens{Data: data} |
| 80 | if rid, ok := s.storeSemResult(u, data); ok { |
| 81 | res.ResultID = &rid |
| 82 | } |
| 83 | return res |
| 84 | } |
| 85 | |
| 86 | func (s *server) semanticTokensData(u uri.URI) ([]uint32, bool) { |
| 87 | tokens, ok := s.tokensForDoc(u) |
| 88 | if !ok { |
| 89 | return nil, false |
| 90 | } |
| 91 | return encodeSemTokens(tokens), true |
| 92 | } |
| 93 | |
| 94 | type semCache struct { |
| 95 | entries []semEntry // tokens per entry of the current text |
| 96 | baseline []uint32 // encoded data of the last response; diff baseline for the next delta request |
| 97 | gen uint64 // increments per response; 0 before the first |
| 98 | pending *semEdit // single edit to apply incrementally, nil when none |
| 99 | } |
| 100 | |
| 101 | type semEdit struct { |
| 102 | start, oldEnd, newEnd int |
| 103 | deltaLine int |
| 104 | } |
| 105 | |
| 106 | func (c *semCache) resultID() string { return strconv.FormatUint(c.gen, 10) } |
| 107 | func (c *semCache) nextResultID() string { |
| 108 | c.gen++ |
| 109 | return c.resultID() |
| 110 | } |
| 111 | |
| 112 | func (s *server) storeSemResult(u uri.URI, data []uint32) (string, bool) { |
| 113 | s.mu.Lock() |
| 114 | defer s.mu.Unlock() |
| 115 | st, ok := s.openDocs[u] |
| 116 | if !ok { |
| 117 | return "", false |
| 118 | } |
| 119 | if st.sem == nil { |
| 120 | st.sem = &semCache{} |
| 121 | } |
| 122 | rid := st.sem.nextResultID() |
| 123 | st.sem.baseline = data |
| 124 | s.openDocs[u] = st |
| 125 | return rid, true |
| 126 | } |
| 127 | |
| 128 | func (s *server) tokensForDoc(doc uri.URI) ([]semanticToken, bool) { |
| 129 | s.mu.RLock() |
| 130 | st, ok := s.openDocs[doc] |
| 131 | s.mu.RUnlock() |
| 132 | if !ok { |
| 133 | return nil, false |
| 134 | } |
| 135 | if st.sem != nil && st.sem.pending == nil { |
| 136 | return flattenSemTokens(st.sem.entries), true |
| 137 | } |
| 138 | rj := s.loader.ResolveBytes(doc.Path(), []byte(st.text)) |
| 139 | j := rj.Occurrences[0].Ast |
| 140 | |
| 141 | var tokens []semanticToken |
| 142 | var entries []semEntry |
| 143 | good := st.sem != nil && st.sem.pending != nil |
| 144 | if good { |
| 145 | entries, good = incrementalTokens(st.text, j, st.sem.entries, *st.sem.pending) |
| 146 | } |
| 147 | if !good { |
| 148 | tokens, entries = computeSemTokens(st.text, j) |
| 149 | } else { |
| 150 | tokens = flattenSemTokens(entries) |
| 151 | } |
| 152 | s.mu.Lock() |
| 153 | defer s.mu.Unlock() |
| 154 | cur, ok := s.openDocs[doc] |
| 155 | if !ok { |
| 156 | return nil, false |
| 157 | } |
| 158 | if cur.text == st.text { // unchanged during tokenization |
| 159 | if cur.sem == nil { |
| 160 | cur.sem = &semCache{} |
| 161 | } |
| 162 | cur.sem.entries = entries |
| 163 | cur.sem.pending = nil |
| 164 | s.openDocs[doc] = cur |
| 165 | } |
| 166 | return tokens, true |
| 167 | } |
| 168 | |
| 169 | const ( |
| 170 | semDirective uint32 = iota |
| 171 | semDate |
| 172 | semAccount |
| 173 | semCommodity |
| 174 | semAmount |
| 175 | semStatus |
| 176 | semComment |
| 177 | semString |
| 178 | semOperator |
| 179 | semProperty |
| 180 | ) |
| 181 | |
| 182 | var tokenTypeStrings = []string{ |
| 183 | string(protocol.SemanticTokenTypesKeyword), // directive |
| 184 | string(protocol.SemanticTokenTypesClass), // date |
| 185 | string(protocol.SemanticTokenTypesNamespace), // account |
| 186 | string(protocol.SemanticTokenTypesType), // commodity |
| 187 | string(protocol.SemanticTokenTypesNumber), // amount |
| 188 | string(protocol.SemanticTokenTypesOperator), // status |
| 189 | string(protocol.SemanticTokenTypesComment), // comment |
| 190 | string(protocol.SemanticTokenTypesString), // string |
| 191 | string(protocol.SemanticTokenTypesOperator), // operator |
| 192 | string(protocol.SemanticTokenTypesProperty), // property |
| 193 | } |
| 194 | |
| 195 | const ( |
| 196 | modifierAbstract = 1 << 0 // virtual account |
| 197 | modifierNegative = 1 << 1 // negative amount |
| 198 | ) |
| 199 | |
| 200 | var modifierStrings = []string{ |
| 201 | "abstract", // bit 0 |
| 202 | "negative", // bit 1 |
| 203 | } |
| 204 | |
| 205 | func getSemanticTokensLegend() protocol.SemanticTokensLegend { |
| 206 | return protocol.SemanticTokensLegend{ |
| 207 | TokenTypes: tokenTypeStrings, |
| 208 | TokenModifiers: modifierStrings, |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | type semanticToken struct { |
| 213 | line, col uint32 // 0-based |
| 214 | length uint32 |
| 215 | tokenType uint32 |
| 216 | modifiers uint32 |
| 217 | } |
| 218 | |
| 219 | type semEntry struct { |
| 220 | start, end int |
| 221 | tokens []semanticToken |
| 222 | } |
| 223 | |
| 224 | func tokenizeForSemantics(content string, j *ast.Journal) []semanticToken { |
| 225 | var raw []rawSpan |
| 226 | emit := func(s token.Span, tokType, mods uint32) { |
| 227 | if s.Start.Offset >= s.End.Offset { |
| 228 | return |
| 229 | } |
| 230 | raw = append(raw, rawSpan{s, tokType, mods}) |
| 231 | } |
| 232 | for _, e := range j.Entries { |
| 233 | visitEntry(content, e, emit) |
| 234 | } |
| 235 | if len(j.Errors) > 0 { |
| 236 | // parser recovers per line; lexer fills the unparsed regions, keeping ast tokens where the parser succeeded |
| 237 | semLexerFallback(content, raw, emit) |
| 238 | } |
| 239 | return rawToSemanticTokens(content, raw) |
| 240 | } |
| 241 | |
| 242 | // rawSpan is a source span tagged with semantic token |
| 243 | type rawSpan struct { |
| 244 | span token.Span |
| 245 | tok, mods uint32 |
| 246 | } |
| 247 | |
| 248 | func computeSemTokens(content string, j *ast.Journal) ([]semanticToken, []semEntry) { |
| 249 | if len(j.Errors) > 0 { |
| 250 | return tokenizeForSemantics(content, j), nil |
| 251 | } |
| 252 | entries := buildSemTokens(content, j) |
| 253 | return flattenSemTokens(entries), entries |
| 254 | } |
| 255 | |
| 256 | func buildSemTokens(content string, j *ast.Journal) []semEntry { |
| 257 | entries := make([]semEntry, 0, len(j.Entries)) |
| 258 | for _, e := range j.Entries { |
| 259 | if _, isBlank := e.(*ast.BlankLine); isBlank { |
| 260 | continue |
| 261 | } |
| 262 | sp := entrySpan(e) |
| 263 | entries = append(entries, semEntry{start: sp.Start.Offset, end: sp.End.Offset, tokens: tokensForEntry(content, e)}) |
| 264 | } |
| 265 | return entries |
| 266 | } |
| 267 | |
| 268 | func flattenSemTokens(entries []semEntry) []semanticToken { |
| 269 | var flat []semanticToken |
| 270 | for _, en := range entries { |
| 271 | flat = append(flat, en.tokens...) |
| 272 | } |
| 273 | return flat |
| 274 | } |
| 275 | |
| 276 | func tokensForEntry(content string, e ast.Entry) []semanticToken { |
| 277 | sp := entrySpan(e) |
| 278 | var eraw []rawSpan |
| 279 | visitEntry(content, e, func(tok token.Span, tokKind, modifier uint32) { |
| 280 | if tok.Start.Offset >= tok.End.Offset { |
| 281 | return |
| 282 | } |
| 283 | eraw = append(eraw, rawSpan{tok, tokKind, modifier}) |
| 284 | }) |
| 285 | if len(eraw) == 0 { |
| 286 | return nil |
| 287 | } |
| 288 | return rawToSemanticTokensFrom(content, eraw, sp.Start.Line-1, 0, sp.Start.Offset) |
| 289 | } |
| 290 | |
| 291 | func shiftLines(toks []semanticToken, delta int) []semanticToken { |
| 292 | if delta == 0 { |
| 293 | return toks |
| 294 | } |
| 295 | out := make([]semanticToken, len(toks)) |
| 296 | for i, t := range toks { |
| 297 | t.line = uint32(int(t.line) + delta) |
| 298 | out[i] = t |
| 299 | } |
| 300 | return out |
| 301 | } |
| 302 | |
| 303 | func incrementalTokens(content string, j *ast.Journal, old []semEntry, e semEdit) ([]semEntry, bool) { |
| 304 | if len(j.Errors) > 0 { |
| 305 | return nil, false |
| 306 | } |
| 307 | var nb, na int |
| 308 | for _, en := range j.Entries { |
| 309 | if _, isBlank := en.(*ast.BlankLine); isBlank { |
| 310 | continue |
| 311 | } |
| 312 | switch sp := entrySpan(en); { |
| 313 | case sp.End.Offset <= e.start: |
| 314 | nb++ |
| 315 | case sp.Start.Offset >= e.newEnd: |
| 316 | na++ |
| 317 | } |
| 318 | } |
| 319 | var ob, oa int |
| 320 | for _, en := range old { |
| 321 | switch { |
| 322 | case en.end <= e.start: |
| 323 | ob++ |
| 324 | case en.start >= e.oldEnd: |
| 325 | oa++ |
| 326 | } |
| 327 | } |
| 328 | if nb != ob || na != oa { |
| 329 | return nil, false |
| 330 | } |
| 331 | |
| 332 | out := make([]semEntry, 0, len(old)+1) |
| 333 | iBefore, iTail := 0, len(old)-oa |
| 334 | for _, en := range j.Entries { |
| 335 | if _, isBlank := en.(*ast.BlankLine); isBlank { |
| 336 | continue |
| 337 | } |
| 338 | sp := entrySpan(en) |
| 339 | var toks []semanticToken |
| 340 | switch { |
| 341 | case sp.End.Offset <= e.start: |
| 342 | toks = old[iBefore].tokens |
| 343 | iBefore++ |
| 344 | case sp.Start.Offset >= e.newEnd: |
| 345 | toks = shiftLines(old[iTail].tokens, e.deltaLine) |
| 346 | iTail++ |
| 347 | default: |
| 348 | toks = tokensForEntry(content, en) |
| 349 | } |
| 350 | out = append(out, semEntry{start: sp.Start.Offset, end: sp.End.Offset, tokens: toks}) |
| 351 | } |
| 352 | return out, true |
| 353 | } |
| 354 | |
| 355 | func rawToSemanticTokens(content string, raw []rawSpan) []semanticToken { |
| 356 | return rawToSemanticTokensFrom(content, raw, 0, 0, 0) |
| 357 | } |
| 358 | |
| 359 | func rawToSemanticTokensFrom(content string, raw []rawSpan, line, col, cursor int) []semanticToken { |
| 360 | if len(raw) == 0 { |
| 361 | return nil |
| 362 | } |
| 363 | slices.SortFunc(raw, func(a, b rawSpan) int { return a.span.Start.Offset - b.span.Start.Offset }) |
| 364 | out := make([]semanticToken, len(raw)) |
| 365 | advance := func(end int) { |
| 366 | for cursor < end { |
| 367 | r, size := utf8.DecodeRuneInString(content[cursor:]) |
| 368 | if r == utf8.RuneError && size <= 1 { |
| 369 | break |
| 370 | } |
| 371 | if r == '\r' { |
| 372 | cursor += size |
| 373 | if cursor < len(content) && content[cursor] == '\n' { |
| 374 | cursor++ |
| 375 | } |
| 376 | line++ |
| 377 | col = 0 |
| 378 | continue |
| 379 | } |
| 380 | if r == '\n' { |
| 381 | cursor += size |
| 382 | line++ |
| 383 | col = 0 |
| 384 | continue |
| 385 | } |
| 386 | cursor += size |
| 387 | col += utf16Units(r) |
| 388 | } |
| 389 | } |
| 390 | for i, t := range raw { |
| 391 | if cursor < t.span.Start.Offset { |
| 392 | advance(t.span.Start.Offset) |
| 393 | } |
| 394 | out[i] = semanticToken{ |
| 395 | line: uint32(line), |
| 396 | col: uint32(col), |
| 397 | length: uint32(lsputil.Utf16Len(content, t.span.Start.Offset, t.span.End.Offset)), |
| 398 | tokenType: t.tok, |
| 399 | modifiers: t.mods, |
| 400 | } |
| 401 | advance(t.span.End.Offset) |
| 402 | } |
| 403 | return out |
| 404 | } |
| 405 | |
| 406 | func utf16Units(r rune) int { |
| 407 | if r >= 0x10000 && r <= 0x10FFFF { |
| 408 | return 2 |
| 409 | } |
| 410 | return 1 |
| 411 | } |
| 412 | |
| 413 | func visitEntry(content string, e ast.Entry, emit semEmitFunc) { |
| 414 | switch e := e.(type) { |
| 415 | case *ast.Transaction: |
| 416 | visitTransaction(content, e, emit) |
| 417 | case *ast.PeriodicTransaction: |
| 418 | visitPeriodicTransaction(content, e, emit) |
| 419 | case *ast.AutomatedTransaction: |
| 420 | visitAutomatedTransaction(content, e, emit) |
| 421 | case *ast.AccountDirective: |
| 422 | emit(directiveKeyword(e.Span, "account"), semDirective, 0) |
| 423 | emit(e.Account.Span, semAccount, 0) |
| 424 | for _, sd := range e.Subdirectives { |
| 425 | if sd.Kind == ast.SubdirectiveComment { |
| 426 | emitComment(sd.Comment, emit) |
| 427 | continue |
| 428 | } |
| 429 | emit(sd.NameSpan, semDirective, 0) |
| 430 | switch sd.Kind { |
| 431 | case ast.SubdirectiveAlias: |
| 432 | emit(sd.ValueSpan, semAccount, 0) |
| 433 | case ast.SubdirectiveType, ast.SubdirectiveNote: |
| 434 | emit(sd.ValueSpan, semProperty, 0) |
| 435 | } |
| 436 | emitComment(sd.Comment, emit) |
| 437 | } |
| 438 | emitComment(e.Comment, emit) |
| 439 | case *ast.CommodityDirective: |
| 440 | emit(directiveKeyword(e.Span, "commodity"), semDirective, 0) |
| 441 | if e.FormatSub != nil { |
| 442 | if e.FormatSub.KeywordSpan.End.Offset > 0 { |
| 443 | emit(e.FormatSub.KeywordSpan, semDirective, 0) |
| 444 | } |
| 445 | semEmitAmount(content, &e.FormatSub.Amount, emit) |
| 446 | emitComment(e.FormatSub.Comment, emit) |
| 447 | } else if e.CommoditySpan.Start.Offset > 0 && e.CommoditySpan.End.Offset > 0 { |
| 448 | emit(e.CommoditySpan, semCommodity, 0) |
| 449 | } |
| 450 | emitBlockComments(e.BlockComments, emit) |
| 451 | emitComment(e.Comment, emit) |
| 452 | case *ast.IncludeDirective: |
| 453 | emitDirective(content, e.Span, len("include"), semString, e.Comment, emit) |
| 454 | case *ast.PayeeDirective: |
| 455 | emitDirective(content, e.Span, len("payee"), semProperty, e.Comment, emit) |
| 456 | case *ast.TagDirective: |
| 457 | emitDirective(content, e.Span, len("tag"), semProperty, e.Comment, emit) |
| 458 | case *ast.AliasDirective: |
| 459 | emit(directiveKeyword(e.Span, "alias"), semDirective, 0) |
| 460 | emit(e.From.Span, semAccount, 0) |
| 461 | if op, ok := betweenSpan(content, e.Span.File, e.From.Span.End.Offset, e.To.Span.Start.Offset); ok { |
| 462 | emit(op, semOperator, 0) |
| 463 | } |
| 464 | emit(e.To.Span, semAccount, 0) |
| 465 | emitComment(e.Comment, emit) |
| 466 | case *ast.YearDirective: |
| 467 | kwLen := len("year") |
| 468 | if content[e.Span.Start.Offset] == 'Y' { |
| 469 | kwLen = 1 |
| 470 | } |
| 471 | emitDirective(content, e.Span, kwLen, semProperty, e.Comment, emit) |
| 472 | case *ast.DecimalMarkDirective: |
| 473 | emitDirective(content, e.Span, len("decimal-mark"), semProperty, e.Comment, emit) |
| 474 | case *ast.DefaultCommodityDirective: |
| 475 | emit(directiveKeyword(e.Span, "D"), semDirective, 0) |
| 476 | semEmitAmount(content, &e.Amount, emit) |
| 477 | emitComment(e.Comment, emit) |
| 478 | case *ast.MarketPriceDirective: |
| 479 | emit(directiveKeyword(e.Span, "P"), semDirective, 0) |
| 480 | emit(e.DateTime.Date.Span, semDate, 0) |
| 481 | if e.DateTime.Time != nil { |
| 482 | emit(e.DateTime.Time.Span, semDate, 0) |
| 483 | } |
| 484 | // commodity: text between the date (or time) and the amount |
| 485 | commStart := e.DateTime.Date.Span.End.Offset |
| 486 | if e.DateTime.Time != nil { |
| 487 | commStart = e.DateTime.Time.Span.End.Offset |
| 488 | } |
| 489 | if comm, ok := betweenSpan(content, e.Span.File, commStart, e.Amount.Span.Start.Offset); ok { |
| 490 | emit(comm, semCommodity, 0) |
| 491 | } |
| 492 | semEmitAmount(content, &e.Amount, emit) |
| 493 | emitComment(e.Comment, emit) |
| 494 | case *ast.ConversionDirective: |
| 495 | emit(directiveKeyword(e.Span, "C"), semDirective, 0) |
| 496 | semEmitAmount(content, &e.From, emit) |
| 497 | // = operator: text between the two amounts |
| 498 | if op, ok := betweenSpan(content, e.Span.File, e.From.Span.End.Offset, e.To.Span.Start.Offset); ok { |
| 499 | emit(op, semOperator, 0) |
| 500 | } |
| 501 | semEmitAmount(content, &e.To, emit) |
| 502 | emitComment(e.Comment, emit) |
| 503 | case *ast.Comment: |
| 504 | emitComment(e, emit) |
| 505 | case *ast.CommentBlockDirective: |
| 506 | emit(e.Span, semComment, 0) |
| 507 | case *ast.IgnoredDirective: |
| 508 | emitDirective(content, e.Span, len("N"), semProperty, e.Comment, emit) |
| 509 | case *ast.ApplyDirective: |
| 510 | emitDirective(content, e.Span, len("apply"), semProperty, e.Comment, emit) |
| 511 | case *ast.EndDirective: |
| 512 | emitDirective(content, e.Span, len("end"), semProperty, e.Comment, emit) |
| 513 | case *ast.BlankLine: |
| 514 | } |
| 515 | } |
| 516 | |
| 517 | func visitTransaction(content string, t *ast.Transaction, emit semEmitFunc) { |
| 518 | emit(t.Date.Span, semDate, 0) |
| 519 | if t.SecondDate != nil { |
| 520 | emit(t.SecondDate.Span, semDate, 0) |
| 521 | } |
| 522 | if t.Status != ast.StatusNone { |
| 523 | emit(t.StatusSpan, semStatus, 0) |
| 524 | } |
| 525 | if t.Code != "" { |
| 526 | emit(t.CodeSpan, semString, 0) |
| 527 | } |
| 528 | if t.Payee != "" { |
| 529 | emit(t.PayeeSpan, semProperty, 0) |
| 530 | } |
| 531 | if t.Note != "" { |
| 532 | emit(t.NoteSpan, semProperty, 0) |
| 533 | } |
| 534 | emitComment(t.Comment, emit) |
| 535 | for i := range t.HeaderComments { |
| 536 | emitComment(t.HeaderComments[i], emit) |
| 537 | } |
| 538 | for _, p := range t.Postings { |
| 539 | visitPosting(content, p, emit) |
| 540 | } |
| 541 | } |
| 542 | |
| 543 | func visitPeriodicTransaction(content string, pt *ast.PeriodicTransaction, emit semEmitFunc) { |
| 544 | // ~ operator is at the start of the period span |
| 545 | emit(offsetSpan(pt.Span.File, pt.Span.Start.Offset, pt.Span.Start.Offset+1), semOperator, 0) |
| 546 | |
| 547 | // The period span covers the whole expr, including any "from ... to ..." dates |
| 548 | if pt.Period.Span.End.Offset > pt.Period.Span.Start.Offset { |
| 549 | var dates []*ast.Date |
| 550 | if pt.Period.From != nil { |
| 551 | dates = append(dates, pt.Period.From) |
| 552 | } |
| 553 | if pt.Period.To != nil { |
| 554 | dates = append(dates, pt.Period.To) |
| 555 | } |
| 556 | pos := pt.Period.Span.Start.Offset |
| 557 | for _, d := range dates { |
| 558 | if d.Span.Start.Offset > pos { |
| 559 | emit(offsetSpan(pt.Period.Span.File, pos, d.Span.Start.Offset), semProperty, 0) |
| 560 | } |
| 561 | emit(d.Span, semDate, 0) |
| 562 | pos = d.Span.End.Offset |
| 563 | } |
| 564 | if pos < pt.Period.Span.End.Offset { |
| 565 | emit(offsetSpan(pt.Period.Span.File, pos, pt.Period.Span.End.Offset), semProperty, 0) |
| 566 | } |
| 567 | } |
| 568 | if pt.Description != "" { |
| 569 | emit(pt.DescriptionSpan, semProperty, 0) |
| 570 | } |
| 571 | emitComment(pt.Comment, emit) |
| 572 | for i := range pt.HeaderComments { |
| 573 | emitComment(pt.HeaderComments[i], emit) |
| 574 | } |
| 575 | for _, p := range pt.Postings { |
| 576 | visitPosting(content, p, emit) |
| 577 | } |
| 578 | } |
| 579 | |
| 580 | func visitAutomatedTransaction(content string, at *ast.AutomatedTransaction, emit semEmitFunc) { |
| 581 | // = operator is at the start of the expression span |
| 582 | emit(offsetSpan(at.Span.File, at.Span.Start.Offset, at.Span.Start.Offset+1), semOperator, 0) |
| 583 | |
| 584 | if at.Expr != "" { |
| 585 | emit(at.ExprSpan, semString, 0) |
| 586 | } |
| 587 | emitComment(at.Comment, emit) |
| 588 | for i := range at.HeaderComments { |
| 589 | emitComment(at.HeaderComments[i], emit) |
| 590 | } |
| 591 | for _, p := range at.Postings { |
| 592 | visitPosting(content, p, emit) |
| 593 | } |
| 594 | } |
| 595 | |
| 596 | func visitPosting(content string, p ast.Posting, emit semEmitFunc) { |
| 597 | if p.Status != ast.StatusNone { |
| 598 | emit(p.StatusSpan, semStatus, 0) |
| 599 | } |
| 600 | |
| 601 | // virtual brackets |
| 602 | if p.Type == ast.PostingVirtualUnbalanced || p.Type == ast.PostingVirtualBalanced { |
| 603 | // opening bracket |
| 604 | for off := p.Span.Start.Offset; off < p.Account.Span.Start.Offset && off < p.Span.End.Offset; off++ { |
| 605 | if content[off] == '(' || content[off] == '[' { |
| 606 | emit(offsetSpan(p.Span.File, off, off+1), semOperator, modifierAbstract) |
| 607 | break |
| 608 | } |
| 609 | } |
| 610 | // closing bracket |
| 611 | for off := p.Account.Span.End.Offset; off < p.Span.End.Offset; off++ { |
| 612 | if content[off] == ')' || content[off] == ']' { |
| 613 | emit(offsetSpan(p.Span.File, off, off+1), semOperator, modifierAbstract) |
| 614 | break |
| 615 | } |
| 616 | } |
| 617 | } |
| 618 | |
| 619 | emit(p.Account.Span, semAccount, 0) |
| 620 | |
| 621 | if p.Amount != nil { |
| 622 | semEmitAmount(content, p.Amount, emit) |
| 623 | } |
| 624 | if p.Cost != nil { |
| 625 | semEmitCost(content, p.Cost, emit) |
| 626 | } |
| 627 | if p.Balance != nil { |
| 628 | semEmitBalanceAssertion(content, p.Balance, emit) |
| 629 | } |
| 630 | emitComment(p.Comment, emit) |
| 631 | for i := range p.Comments { |
| 632 | emitComment(&p.Comments[i], emit) |
| 633 | } |
| 634 | } |
| 635 | |
| 636 | // directiveKeyword returns the span of the leading keyword on a directive line. |
| 637 | func directiveKeyword(e token.Span, kw string) token.Span { |
| 638 | return token.Span{File: e.File, Start: e.Start, End: token.Pos{Offset: e.Start.Offset + len(kw)}} |
| 639 | } |
| 640 | |
| 641 | // directiveValue returns the trimmed span of the text after the keyword end |
| 642 | // offset, up to the inline comment or the end of the line. |
| 643 | func directiveValue(content string, e token.Span, comment *ast.Comment, kwEnd int) (token.Span, bool) { |
| 644 | end := e.End.Offset |
| 645 | if comment != nil { |
| 646 | end = comment.Span.Start.Offset |
| 647 | } |
| 648 | return betweenSpan(content, e.File, kwEnd, end) |
| 649 | } |
| 650 | |
| 651 | func semEmitAmount(content string, a *ast.Amount, emit semEmitFunc) { |
| 652 | if a == nil { |
| 653 | return |
| 654 | } |
| 655 | hasCommodity := a.CommoditySpan.Start.Offset > 0 && a.CommoditySpan.End.Offset > 0 |
| 656 | if hasCommodity && a.CommodityPos == ast.CommodityBefore { |
| 657 | emit(a.CommoditySpan, semCommodity, 0) |
| 658 | semEmitQuantity(content, a, emit) |
| 659 | return |
| 660 | } |
| 661 | semEmitQuantity(content, a, emit) |
| 662 | if hasCommodity { |
| 663 | emit(a.CommoditySpan, semCommodity, 0) |
| 664 | } |
| 665 | } |
| 666 | |
| 667 | func semEmitQuantity(content string, a *ast.Amount, emit semEmitFunc) { |
| 668 | qStart, qEnd := quantitySpan(content, a) |
| 669 | if qEnd <= qStart { |
| 670 | return |
| 671 | } |
| 672 | mods := uint32(0) |
| 673 | if a.IsNegative { |
| 674 | mods |= modifierNegative |
| 675 | } |
| 676 | emit(offsetSpan(a.Span.File, qStart, qEnd), semAmount, mods) |
| 677 | } |
| 678 | |
| 679 | func quantitySpan(content string, a *ast.Amount) (int, int) { |
| 680 | start, end := a.Span.Start.Offset, a.Span.End.Offset |
| 681 | switch { |
| 682 | case a.Commodity == "": |
| 683 | // bare quantity |
| 684 | case a.CommodityPos == ast.CommodityBefore: |
| 685 | // "$50.00" or "$ 50.00": quantity follows the commodity span |
| 686 | start = a.CommoditySpan.End.Offset |
| 687 | default: // CommodityAfter |
| 688 | // "50.00 USD" or "50.00 USD": quantity precedes the commodity |
| 689 | end = a.CommoditySpan.Start.Offset |
| 690 | } |
| 691 | for start < end && (content[start] == ' ' || content[start] == '\t') { |
| 692 | start++ |
| 693 | } |
| 694 | for end > start && (content[end-1] == ' ' || content[end-1] == '\t') { |
| 695 | end-- |
| 696 | } |
| 697 | return start, end |
| 698 | } |
| 699 | |
| 700 | type semEmitFunc func(tok token.Span, tokKind, modifier uint32) |
| 701 | |
| 702 | func emitBlockComments(cs []*ast.Comment, emit semEmitFunc) { |
| 703 | for _, c := range cs { |
| 704 | emitComment(c, emit) |
| 705 | } |
| 706 | } |
| 707 | |
| 708 | func emitComment(c *ast.Comment, emit semEmitFunc) { |
| 709 | if c == nil { |
| 710 | return |
| 711 | } |
| 712 | if len(c.Tags) == 0 { |
| 713 | emit(c.Span, semComment, 0) |
| 714 | return |
| 715 | } |
| 716 | pos := c.Span.Start.Offset |
| 717 | for _, t := range c.Tags { |
| 718 | if t.Span.Start.Offset > pos { |
| 719 | emit(offsetSpan(c.Span.File, pos, t.Span.Start.Offset), semComment, 0) |
| 720 | } |
| 721 | emit(t.Span, semProperty, 0) |
| 722 | pos = t.Span.End.Offset |
| 723 | } |
| 724 | if pos < c.Span.End.Offset { |
| 725 | emit(offsetSpan(c.Span.File, pos, c.Span.End.Offset), semComment, 0) |
| 726 | } |
| 727 | } |
| 728 | |
| 729 | func emitDirective(content string, e token.Span, kwLen int, valType uint32, comment *ast.Comment, emit semEmitFunc) { |
| 730 | kwEnd := e.Start.Offset + kwLen |
| 731 | emit(token.Span{File: e.File, Start: e.Start, End: token.Pos{Offset: kwEnd}}, semDirective, 0) |
| 732 | if v, ok := directiveValue(content, e, comment, kwEnd); ok { |
| 733 | emit(v, valType, 0) |
| 734 | } |
| 735 | emitComment(comment, emit) |
| 736 | } |
| 737 | |
| 738 | func semEmitCost(content string, c *ast.Cost, emit semEmitFunc) { |
| 739 | if c.IsTotal { |
| 740 | emit(token.Span{File: c.Span.File, Start: c.Span.Start, End: token.Pos{Offset: c.Span.Start.Offset + 2}}, semOperator, 0) |
| 741 | } else { |
| 742 | emit(token.Span{File: c.Span.File, Start: c.Span.Start, End: token.Pos{Offset: c.Span.Start.Offset + 1}}, semOperator, 0) |
| 743 | } |
| 744 | semEmitAmount(content, &c.Amount, emit) |
| 745 | } |
| 746 | |
| 747 | func semEmitBalanceAssertion(content string, ba *ast.BalanceAssertion, emit semEmitFunc) { |
| 748 | // The operator is the run of '=', ':', '*' chars from the span start. |
| 749 | // The ':' of ':=' precedes the '=' token, so back up one offset. |
| 750 | opStart := ba.Span.Start.Offset |
| 751 | if ba.IsAssignment && opStart > 0 { |
| 752 | opStart-- |
| 753 | } |
| 754 | opEnd := opStart |
| 755 | for opEnd < ba.Span.End.Offset && (content[opEnd] == '=' || content[opEnd] == ':' || content[opEnd] == '*') { |
| 756 | opEnd++ |
| 757 | } |
| 758 | emit(token.Span{File: ba.Span.File, Start: token.Pos{Offset: opStart}, End: token.Pos{Offset: opEnd}}, semOperator, 0) |
| 759 | semEmitAmount(content, &ba.Amount, emit) |
| 760 | if ba.Cost != nil { |
| 761 | semEmitCost(content, ba.Cost, emit) |
| 762 | } |
| 763 | } |
| 764 | |
| 765 | func semLexerFallback(content string, base []rawSpan, emit semEmitFunc) { |
| 766 | l := lexer.New("", []byte(content)) |
| 767 | |
| 768 | var commentStart, commentEnd int // 0 = not inside a comment line |
| 769 | lineStart := true // the next significant token starts a line |
| 770 | skipLine := false // the line starts with an unclassifiable token; emit nothing |
| 771 | i := 0 // next base span to compare against |
| 772 | |
| 773 | take := func(span token.Span, tokType uint32, mods uint32) { |
| 774 | for i < len(base) && base[i].span.End.Offset <= span.Start.Offset { |
| 775 | i++ |
| 776 | } |
| 777 | if i < len(base) && base[i].span.Start.Offset < span.End.Offset { |
| 778 | return // overlaps an AST token; AST wins |
| 779 | } |
| 780 | emit(span, tokType, mods) |
| 781 | } |
| 782 | |
| 783 | for { |
| 784 | tok := l.Next() |
| 785 | if tok.Type == token.EOF { |
| 786 | if commentStart > 0 { |
| 787 | take(token.Span{Start: token.Pos{Offset: commentStart}, End: token.Pos{Offset: commentEnd}}, semComment, 0) |
| 788 | } |
| 789 | break |
| 790 | } |
| 791 | if tok.Type == token.NEWLINE { |
| 792 | if commentStart > 0 { |
| 793 | take(token.Span{Start: token.Pos{Offset: commentStart}, End: token.Pos{Offset: commentEnd}}, semComment, 0) |
| 794 | commentStart, commentEnd = 0, 0 |
| 795 | } |
| 796 | lineStart, skipLine = true, false |
| 797 | continue |
| 798 | } |
| 799 | if tok.Type == token.WHITESPACE || tok.Type == token.INDENT { |
| 800 | continue |
| 801 | } |
| 802 | if lineStart { |
| 803 | lineStart = false |
| 804 | if !isLineStartToken(tok.Type) { |
| 805 | skipLine = true |
| 806 | } |
| 807 | } |
| 808 | if skipLine { |
| 809 | continue |
| 810 | } |
| 811 | tokType := semProperty |
| 812 | if commentStart > 0 { |
| 813 | tokType = semComment |
| 814 | if tok.Span.End.Offset > commentEnd { |
| 815 | commentEnd = tok.Span.End.Offset |
| 816 | } |
| 817 | continue |
| 818 | } |
| 819 | |
| 820 | switch tok.Type { |
| 821 | case token.SEMICOLON, token.HASH, token.PERCENT, token.STAR: |
| 822 | tokType = semComment |
| 823 | commentStart = tok.Span.Start.Offset |
| 824 | commentEnd = tok.Span.End.Offset |
| 825 | continue |
| 826 | case token.STRING: |
| 827 | tokType = semString |
| 828 | case token.DATE: |
| 829 | tokType = semDate |
| 830 | case token.INT, token.DECIMAL: |
| 831 | tokType = semAmount |
| 832 | case token.COMMODITYMARK: |
| 833 | tokType = semCommodity |
| 834 | case token.BANG: |
| 835 | tokType = semStatus |
| 836 | case token.AT, token.ATAT, token.EQ, token.EQEQ, token.EQEQEQ, token.EQSTAR: |
| 837 | tokType = semOperator |
| 838 | case token.COMMENTKW, token.ACCOUNT, token.COMMODITY, token.INCLUDE, |
| 839 | token.ALIAS, token.PAYEE, token.TAG, token.APPLY, token.END, |
| 840 | token.YEAR, token.DECIMALMARK, token.D, token.P, token.N, token.C: |
| 841 | tokType = semDirective |
| 842 | } |
| 843 | take(tok.Span, tokType, 0) |
| 844 | } |
| 845 | } |
| 846 | |
| 847 | func isLineStartToken(t token.Type) bool { |
| 848 | switch t { |
| 849 | case token.DATE, token.TILDE, token.EQ, token.BANG, token.AT, |
| 850 | token.SEMICOLON, token.HASH, token.PERCENT, token.STAR, |
| 851 | token.COMMENTKW, token.ACCOUNT, token.COMMODITY, token.INCLUDE, |
| 852 | token.ALIAS, token.PAYEE, token.TAG, token.APPLY, token.END, |
| 853 | token.YEAR, token.DECIMALMARK, token.D, token.P, token.N, token.C: |
| 854 | return true |
| 855 | } |
| 856 | return false |
| 857 | } |
| 858 | |
| 859 | // semanticTokensEdits returns the single edit turning old into new, or nil when |
| 860 | // identical. Relative delta encoding keeps the common prefix and suffix unchanged. |
| 861 | func semanticTokensEdits(old, new []uint32) []protocol.SemanticTokensEdit { |
| 862 | p := 0 |
| 863 | for p < len(old) && p < len(new) && old[p] == new[p] { |
| 864 | p++ |
| 865 | } |
| 866 | s := 0 |
| 867 | for s < len(old)-p && s < len(new)-p && old[len(old)-1-s] == new[len(new)-1-s] { |
| 868 | s++ |
| 869 | } |
| 870 | delCount := len(old) - p - s |
| 871 | ins := new[p : len(new)-s] |
| 872 | if delCount == 0 && len(ins) == 0 { |
| 873 | return nil |
| 874 | } |
| 875 | return []protocol.SemanticTokensEdit{{ |
| 876 | Start: uint32(p), |
| 877 | DeleteCount: uint32(delCount), |
| 878 | Data: ins, |
| 879 | }} |
| 880 | } |
| 881 | |
| 882 | // encodeSemTokens encodes tokens into LSP delta form. Input must be sorted by |
| 883 | // line and column; [rawToSemanticTokens] produces such order. |
| 884 | func encodeSemTokens(tokens []semanticToken) []uint32 { |
| 885 | if len(tokens) == 0 { |
| 886 | return nil |
| 887 | } |
| 888 | data := make([]uint32, 0, len(tokens)*5) |
| 889 | var prevLine, prevCol uint32 |
| 890 | for _, t := range tokens { |
| 891 | var deltaLine, deltaCol uint32 |
| 892 | if t.line == prevLine { |
| 893 | deltaLine = 0 |
| 894 | deltaCol = t.col - prevCol |
| 895 | } else { |
| 896 | deltaLine = t.line - prevLine |
| 897 | deltaCol = t.col |
| 898 | } |
| 899 | data = append(data, deltaLine, deltaCol, t.length, t.tokenType, t.modifiers) |
| 900 | prevLine = t.line |
| 901 | prevCol = t.col |
| 902 | } |
| 903 | return data |
| 904 | } |
| 905 | |
| 906 | // betweenSpan returns the span of the text between two offsets, trimmed of surrounding whitespace. |
| 907 | func betweenSpan(content, file string, start, end int) (token.Span, bool) { |
| 908 | for start < end && (content[start] == ' ' || content[start] == '\t') { |
| 909 | start++ |
| 910 | } |
| 911 | for end > start && (content[end-1] == ' ' || content[end-1] == '\t' || content[end-1] == '\n' || content[end-1] == '\r') { |
| 912 | end-- |
| 913 | } |
| 914 | if end <= start { |
| 915 | return token.Span{}, false |
| 916 | } |
| 917 | return offsetSpan(file, start, end), true |
| 918 | } |
| 919 | |
| 920 | func offsetSpan(file string, start, end int) token.Span { |
| 921 | return token.Span{File: file, Start: token.Pos{Offset: start}, End: token.Pos{Offset: end}} |
| 922 | } |