clerk/internal/lsp/textdocument_completion.go (view raw)
| 1 | package lsp |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "fmt" |
| 6 | "math" |
| 7 | "sort" |
| 8 | "strings" |
| 9 | "time" |
| 10 | |
| 11 | "go.lsp.dev/protocol" |
| 12 | |
| 13 | "olexsmir.xyz/clerk/internal/analyzer" |
| 14 | "olexsmir.xyz/clerk/internal/lsp/fuzzy" |
| 15 | "olexsmir.xyz/clerk/internal/lsp/lsputil" |
| 16 | "olexsmir.xyz/clerk/journal/ast" |
| 17 | "olexsmir.xyz/clerk/journal/lexer" |
| 18 | "olexsmir.xyz/clerk/journal/token" |
| 19 | ) |
| 20 | |
| 21 | func (s *server) Completion(ctx context.Context, params *protocol.CompletionParams) (protocol.CompletionResult, error) { |
| 22 | state, ok := s.getDocState(params.TextDocument.URI) |
| 23 | if !ok { |
| 24 | return &protocol.CompletionList{}, nil |
| 25 | } |
| 26 | cursor := lsputil.Offset(state.text, int(params.Position.Line), int(params.Position.Character)) |
| 27 | if cursor > len(state.text) { |
| 28 | return &protocol.CompletionList{}, nil |
| 29 | } |
| 30 | detectedCtx, start := detectCompletionCtx(state.text, cursor) |
| 31 | if detectedCtx == cmplNone { |
| 32 | return &protocol.CompletionList{}, nil |
| 33 | } |
| 34 | a := s.analysis() |
| 35 | if a == nil { |
| 36 | return &protocol.CompletionList{}, nil |
| 37 | } |
| 38 | return &protocol.CompletionList{ |
| 39 | IsIncomplete: true, |
| 40 | Items: cmplItems(a, detectedCtx, state.text, start, cursor), |
| 41 | }, nil |
| 42 | } |
| 43 | |
| 44 | const maxCompletionItems = 50 |
| 45 | |
| 46 | type cmplCtx int |
| 47 | |
| 48 | const ( |
| 49 | cmplNone cmplCtx = iota |
| 50 | cmplAccount |
| 51 | cmplPayee |
| 52 | cmplCommodity |
| 53 | cmplTagName |
| 54 | cmplTagValue |
| 55 | cmplDirective |
| 56 | ) |
| 57 | |
| 58 | var directiveKeywords = []string{ |
| 59 | "account", "include", "commodity", "payee", "decimal-mark", "alias", |
| 60 | "apply", "end", "tag", "year", "D", "P", "N", "C", "Y", |
| 61 | } |
| 62 | |
| 63 | func detectCompletionCtx(content string, cursor int) (cmplCtx, int) { |
| 64 | toks := lexLine(content, cursor) |
| 65 | lineStart, _ := lineBounds(content, cursor) |
| 66 | |
| 67 | if m := commentMarker(toks, cursor); m != -1 { |
| 68 | return cmplTagContext(content, toks[m].Span.End.Offset, cursor) |
| 69 | } |
| 70 | if len(toks) == 0 { |
| 71 | return cmplDirective, lineStart |
| 72 | } |
| 73 | |
| 74 | switch toks[0].Type { |
| 75 | case token.INDENT: |
| 76 | return cmplPostingCtx(content, cursor, toks) |
| 77 | case token.DATE: |
| 78 | return cmplHeaderCtx(content, cursor, toks) |
| 79 | case token.ACCOUNT, token.COMMODITY, token.PAYEE, token.TAG: |
| 80 | return cmplDirectiveContext(cursor, lineStart, toks) |
| 81 | case token.TEXT: |
| 82 | return cmplDirective, lineStart // half-typed keyword or unparseable line |
| 83 | } |
| 84 | return cmplNone, cursor |
| 85 | } |
| 86 | |
| 87 | func cmplPostingCtx(content string, cursor int, toks []token.Token) (cmplCtx, int) { |
| 88 | if inDirectiveBody(content, cursor) { |
| 89 | return cmplNone, cursor |
| 90 | } |
| 91 | fieldStart := toks[0].Span.End.Offset |
| 92 | i := 1 |
| 93 | for i < len(toks) { |
| 94 | switch toks[i].Type { |
| 95 | case token.STAR, token.BANG, token.LPAREN, token.LBRACKET, token.WHITESPACE: |
| 96 | fieldStart = toks[i].Span.End.Offset |
| 97 | i++ |
| 98 | default: |
| 99 | goto run |
| 100 | } |
| 101 | } |
| 102 | run: |
| 103 | // account run: consecutive account-name segments and colons |
| 104 | fieldEnd := fieldStart |
| 105 | for ; i < len(toks); i++ { |
| 106 | if toks[i].Type != token.TEXT && toks[i].Type != token.COLON { |
| 107 | break |
| 108 | } |
| 109 | fieldEnd = toks[i].Span.End.Offset |
| 110 | } |
| 111 | if cursor <= fieldEnd && cursor >= fieldStart { |
| 112 | return cmplAccount, fieldStart |
| 113 | } |
| 114 | if cursor > fieldEnd { |
| 115 | if t := tokenUnder(toks, cursor); t != nil && (t.Type == token.COMMODITYMARK || t.Type == token.STRING) { |
| 116 | start := t.Span.Start.Offset |
| 117 | if t.Type == token.STRING { |
| 118 | start++ // skip opening quote |
| 119 | } |
| 120 | return cmplCommodity, start |
| 121 | } |
| 122 | if strings.TrimSpace(content[fieldEnd:cursor]) == "" { |
| 123 | return cmplCommodity, cursor |
| 124 | } |
| 125 | } |
| 126 | return cmplNone, cursor |
| 127 | } |
| 128 | |
| 129 | func cmplHeaderCtx(content string, cursor int, toks []token.Token) (cmplCtx, int) { |
| 130 | // skip date, status, code, and whitespace - where the payee beginds |
| 131 | fieldStart := toks[0].Span.End.Offset |
| 132 | fieldEnd := fieldStart |
| 133 | seen := false |
| 134 | for i := 1; i < len(toks); i++ { |
| 135 | t := toks[i] |
| 136 | switch t.Type { |
| 137 | case token.WHITESPACE, token.STAR, token.BANG, token.DATE, token.TIME, |
| 138 | token.EQ, token.EQEQ, token.EQEQEQ: |
| 139 | fieldStart = t.Span.End.Offset |
| 140 | fieldEnd = t.Span.End.Offset |
| 141 | case token.TEXT: |
| 142 | lit := content[t.Span.Start.Offset:t.Span.End.Offset] |
| 143 | if !seen && len(lit) >= 2 && lit[0] == '(' && lit[len(lit)-1] == ')' { |
| 144 | fieldStart = t.Span.End.Offset // parenthesized code |
| 145 | fieldEnd = t.Span.End.Offset |
| 146 | continue |
| 147 | } |
| 148 | if !seen { |
| 149 | fieldStart = t.Span.Start.Offset |
| 150 | seen = true |
| 151 | } |
| 152 | fieldEnd = t.Span.End.Offset |
| 153 | if p := strings.IndexByte(lit, '|'); p >= 0 { |
| 154 | fieldEnd = t.Span.Start.Offset + p // "payee|note" keeps the pipe in the token |
| 155 | return payeeAt(cursor, fieldStart, fieldEnd) |
| 156 | } |
| 157 | case token.STRING: |
| 158 | if !seen { |
| 159 | fieldStart = t.Span.Start.Offset + 1 // skip opening quote |
| 160 | seen = true |
| 161 | } |
| 162 | fieldEnd = t.Span.End.Offset |
| 163 | default: |
| 164 | // PIPE, SEMICOLON, ... |
| 165 | return payeeAt(cursor, fieldStart, fieldEnd) |
| 166 | } |
| 167 | } |
| 168 | if seen { |
| 169 | return payeeAt(cursor, fieldStart, fieldEnd) |
| 170 | } |
| 171 | // no payee yet: the payee field is the whitespace after the header meta |
| 172 | if cursor >= fieldStart && strings.TrimSpace(content[fieldStart:cursor]) == "" { |
| 173 | return cmplPayee, cursor |
| 174 | } |
| 175 | return cmplNone, cursor |
| 176 | } |
| 177 | |
| 178 | func payeeAt(cursor, start, end int) (cmplCtx, int) { |
| 179 | if cursor >= start && cursor <= end { |
| 180 | return cmplPayee, start |
| 181 | } |
| 182 | return cmplNone, cursor |
| 183 | } |
| 184 | |
| 185 | // cmplDirectiveContext classifies a directive line. keyword completion before the keyword ends, symbol completion in the value field after |
| 186 | func cmplDirectiveContext(cursor, lineStart int, toks []token.Token) (cmplCtx, int) { |
| 187 | kwEnd := toks[0].Span.End.Offset |
| 188 | if cursor <= kwEnd { |
| 189 | return cmplDirective, lineStart |
| 190 | } |
| 191 | start := kwEnd |
| 192 | for i := 1; i < len(toks); i++ { |
| 193 | if toks[i].Type == token.WHITESPACE || toks[i].Span.End.Offset <= kwEnd { |
| 194 | continue |
| 195 | } |
| 196 | start = toks[i].Span.Start.Offset |
| 197 | if toks[i].Type == token.STRING { |
| 198 | start++ // skip opening quote |
| 199 | } |
| 200 | break |
| 201 | } |
| 202 | if start > cursor { |
| 203 | start = cursor |
| 204 | } |
| 205 | switch toks[0].Type { |
| 206 | case token.ACCOUNT: |
| 207 | return cmplAccount, start |
| 208 | case token.COMMODITY: |
| 209 | return cmplCommodity, start |
| 210 | case token.PAYEE: |
| 211 | return cmplPayee, start |
| 212 | case token.TAG: |
| 213 | return cmplTagName, start |
| 214 | } |
| 215 | return cmplNone, cursor |
| 216 | } |
| 217 | |
| 218 | // commentStart completes tag names before ':' of the current tag and tag values after it |
| 219 | func cmplTagContext(content string, commentStart, cursor int) (cmplCtx, int) { |
| 220 | prefix := content[commentStart:cursor] |
| 221 | segStart := commentStart |
| 222 | seg := prefix |
| 223 | if comma := strings.LastIndexByte(prefix, ','); comma >= 0 { |
| 224 | segStart = commentStart + comma + 1 |
| 225 | seg = prefix[comma+1:] |
| 226 | } |
| 227 | if colon := strings.IndexByte(seg, ':'); colon >= 0 { |
| 228 | start := segStart + colon + 1 |
| 229 | for start < cursor && (content[start] == ' ' || content[start] == '\t') { |
| 230 | start++ |
| 231 | } |
| 232 | return cmplTagValue, start |
| 233 | } |
| 234 | keyStart := commentStart + lastSeparator(prefix) + 1 |
| 235 | return cmplTagName, keyStart |
| 236 | } |
| 237 | |
| 238 | // tagKeyAt returns the key of tag whose value region starts at start |
| 239 | func tagKeyAt(content string, start int) (string, bool) { |
| 240 | lineStart, _ := lineBounds(content, start) |
| 241 | segStart := lineStart |
| 242 | for i := start - 1; i >= lineStart; i-- { |
| 243 | switch content[i] { |
| 244 | case ',', ';', '#', '%': |
| 245 | segStart = i + 1 |
| 246 | i = lineStart - 1 // stop at the separator closest to start |
| 247 | } |
| 248 | } |
| 249 | colon := strings.IndexByte(content[segStart:start], ':') |
| 250 | if colon < 0 { |
| 251 | return "", false |
| 252 | } |
| 253 | colon += segStart |
| 254 | keyStart := lastSeparator(content[segStart:colon]) + 1 |
| 255 | key := content[segStart+keyStart : colon] |
| 256 | if key == "" { |
| 257 | return "", false |
| 258 | } |
| 259 | return key, true |
| 260 | } |
| 261 | |
| 262 | // commentMarker returns index of the first comment marker token at or before the cursor, or -1. |
| 263 | func commentMarker(toks []token.Token, cursor int) int { |
| 264 | for i, t := range toks { |
| 265 | switch t.Type { |
| 266 | case token.SEMICOLON, token.HASH, token.PERCENT: |
| 267 | if t.Span.Start.Offset <= cursor { |
| 268 | return i |
| 269 | } |
| 270 | case token.STAR: |
| 271 | if i == 0 && t.Span.Start.Offset <= cursor { |
| 272 | return i |
| 273 | } |
| 274 | } |
| 275 | } |
| 276 | return -1 |
| 277 | } |
| 278 | |
| 279 | func inDirectiveBody(content string, cursor int) bool { |
| 280 | lineStart, _ := lineBounds(content, cursor) |
| 281 | if toks := lexLine(content, lineStart); len(toks) == 0 || toks[0].Type != token.INDENT { |
| 282 | return false |
| 283 | } |
| 284 | for lineStart > 0 { |
| 285 | lineStart, _ = lineBounds(content, lineStart-1) |
| 286 | toks := lexLine(content, lineStart) |
| 287 | if len(toks) == 0 || toks[0].Type != token.INDENT { |
| 288 | return len(toks) > 0 && (toks[0].Type == token.ACCOUNT || toks[0].Type == token.COMMODITY) |
| 289 | } |
| 290 | } |
| 291 | return false |
| 292 | } |
| 293 | |
| 294 | type cmplCand struct { |
| 295 | label string |
| 296 | score float64 |
| 297 | count int |
| 298 | lastUsed ast.Date |
| 299 | } |
| 300 | |
| 301 | // cmplItems ranks candidates for the content against typed pattern |
| 302 | func cmplItems(a *analyzer.Analysis, ctx cmplCtx, content string, start, cursor int) []protocol.CompletionItem { |
| 303 | pattern := content[start:cursor] |
| 304 | |
| 305 | var kind protocol.CompletionItemKind |
| 306 | var cands []cmplCand |
| 307 | switch ctx { |
| 308 | case cmplAccount: |
| 309 | kind = protocol.CompletionItemKindClass |
| 310 | for name, info := range a.Accounts { |
| 311 | cands = append(cands, cmplCand{label: name, count: info.UsedCount, lastUsed: info.LastUsed}) |
| 312 | } |
| 313 | case cmplPayee: |
| 314 | kind = protocol.CompletionItemKindVariable |
| 315 | for name, info := range a.Payees { |
| 316 | cands = append(cands, cmplCand{label: name, count: info.UsedCount, lastUsed: info.LastUsed}) |
| 317 | } |
| 318 | case cmplCommodity: |
| 319 | kind = protocol.CompletionItemKindValue |
| 320 | for name, info := range a.Commodities { |
| 321 | cands = append(cands, cmplCand{label: name, count: info.UsedCount, lastUsed: info.LastUsed}) |
| 322 | } |
| 323 | case cmplTagName: |
| 324 | kind = protocol.CompletionItemKindProperty |
| 325 | for name, info := range a.Tags { |
| 326 | cands = append(cands, cmplCand{label: name, count: info.UsedCount, lastUsed: info.LastUsed}) |
| 327 | } |
| 328 | case cmplTagValue: |
| 329 | kind = protocol.CompletionItemKindProperty |
| 330 | if key, ok := tagKeyAt(content, start); ok { |
| 331 | if info, ok := a.Tags[key]; ok { |
| 332 | for _, v := range info.Values { |
| 333 | cands = append(cands, cmplCand{label: v}) |
| 334 | } |
| 335 | } |
| 336 | } |
| 337 | case cmplDirective: |
| 338 | kind = protocol.CompletionItemKindKeyword |
| 339 | for _, name := range directiveKeywords { |
| 340 | cands = append(cands, cmplCand{label: name}) |
| 341 | } |
| 342 | default: |
| 343 | return nil |
| 344 | } |
| 345 | |
| 346 | var newest ast.Date |
| 347 | if n := len(a.Dates); n > 0 { |
| 348 | newest = a.Dates[n-1] |
| 349 | } |
| 350 | ranked := cands[:0] |
| 351 | for i := range cands { |
| 352 | sc := fuzzy.Score(pattern, cands[i].label) |
| 353 | if sc != 0 { |
| 354 | sc *= 1 + math.Log1p(float64(cands[i].count)) |
| 355 | if cands[i].count > 0 && cands[i].lastUsed.Year != 0 && newest.Year != 0 { |
| 356 | days := daysBetween(cands[i].lastUsed, newest) |
| 357 | sc *= 1 + 0.5*max(0, 1-float64(days)/365) |
| 358 | } |
| 359 | } |
| 360 | cands[i].score = sc |
| 361 | if sc != 0 { |
| 362 | ranked = append(ranked, cands[i]) |
| 363 | } |
| 364 | } |
| 365 | sort.Slice(ranked, func(i, j int) bool { |
| 366 | if ranked[i].score != ranked[j].score { |
| 367 | return ranked[i].score > ranked[j].score |
| 368 | } |
| 369 | if ranked[i].lastUsed != ranked[j].lastUsed { |
| 370 | return ranked[i].lastUsed.Compare(ranked[j].lastUsed) > 0 |
| 371 | } |
| 372 | return ranked[i].label < ranked[j].label |
| 373 | }) |
| 374 | if len(ranked) > maxCompletionItems { |
| 375 | ranked = ranked[:maxCompletionItems] |
| 376 | } |
| 377 | |
| 378 | replace := protocol.Range{ |
| 379 | Start: lsputil.Position(content, start), |
| 380 | End: lsputil.Position(content, cursor), |
| 381 | } |
| 382 | items := make([]protocol.CompletionItem, len(ranked)) |
| 383 | for i, r := range ranked { |
| 384 | it := protocol.CompletionItem{ |
| 385 | Label: r.label, |
| 386 | Kind: kind, |
| 387 | SortText: protocol.NewOptional(fmt.Sprintf("%04d", i)), |
| 388 | FilterText: protocol.NewOptional(r.label), |
| 389 | TextEdit: &protocol.TextEdit{ |
| 390 | Range: replace, |
| 391 | NewText: r.label, |
| 392 | }, |
| 393 | } |
| 394 | if r.count > 0 { |
| 395 | it.Detail = protocol.NewOptional(fmt.Sprintf("%d uses", r.count)) |
| 396 | } |
| 397 | items[i] = it |
| 398 | } |
| 399 | return items |
| 400 | } |
| 401 | |
| 402 | func daysBetween(a, b ast.Date) int { |
| 403 | return int(time.Date(b.Year, time.Month(b.Month), b.Day, 0, 0, 0, 0, time.UTC). |
| 404 | Sub(time.Date(a.Year, time.Month(a.Month), a.Day, 0, 0, 0, 0, time.UTC)). |
| 405 | Hours() / 24) |
| 406 | } |
| 407 | |
| 408 | // lineBounds returns the byte offsets of the line containing cursor |
| 409 | func lineBounds(content string, cursor int) (start, end int) { |
| 410 | start = cursor |
| 411 | for start > 0 && content[start-1] != '\n' && content[start-1] != '\r' { |
| 412 | start-- |
| 413 | } |
| 414 | end = start |
| 415 | for end < len(content) && content[end] != '\n' && content[end] != '\r' { |
| 416 | end++ |
| 417 | } |
| 418 | return start, end |
| 419 | } |
| 420 | |
| 421 | func lastSeparator(s string) int { |
| 422 | for i := len(s) - 1; i >= 0; i-- { |
| 423 | switch s[i] { |
| 424 | case ' ', '\t', ',': |
| 425 | return i |
| 426 | } |
| 427 | } |
| 428 | return -1 |
| 429 | } |
| 430 | |
| 431 | func lexLine(content string, cursor int) []token.Token { |
| 432 | lineStart, lineEnd := lineBounds(content, cursor) |
| 433 | l := lexer.New("", []byte(content[lineStart:lineEnd])) |
| 434 | var out []token.Token |
| 435 | for { |
| 436 | t := l.Next() |
| 437 | if t.Type == token.EOF || t.Type == token.NEWLINE { |
| 438 | break |
| 439 | } |
| 440 | t.Span.Start.Offset += lineStart |
| 441 | t.Span.End.Offset += lineStart |
| 442 | out = append(out, t) |
| 443 | } |
| 444 | return out |
| 445 | } |
| 446 | |
| 447 | func tokenUnder(toks []token.Token, cursor int) *token.Token { |
| 448 | for i := range toks { |
| 449 | t := &toks[i] |
| 450 | if t.Span.Start.Offset <= cursor && cursor <= t.Span.End.Offset { |
| 451 | return t |
| 452 | } |
| 453 | } |
| 454 | return nil |
| 455 | } |