1// Copyright 2018 The Go Authors. All rights reserved. 2// Use of this source code is governed by a BSD-style 3// license that can be found in the LICENSE file. 4 5package modfile 6 7import ( 8 "bytes" 9 "errors" 10 "fmt" 11 "os" 12 "strconv" 13 "strings" 14 "unicode" 15 "unicode/utf8" 16) 17 18// A Position describes an arbitrary source position in a file, including the 19// file, line, column, and byte offset. 20type Position struct { 21 Line int // line in input (starting at 1) 22 LineRune int // rune in line (starting at 1) 23 Byte int // byte in input (starting at 0) 24} 25 26// add returns the position at the end of s, assuming it starts at p. 27func (p Position) add(s string) Position { 28 p.Byte += len(s) 29 if n := strings.Count(s, "\n"); n > 0 { 30 p.Line += n 31 s = s[strings.LastIndex(s, "\n")+1:] 32 p.LineRune = 1 33 } 34 p.LineRune += utf8.RuneCountInString(s) 35 return p 36} 37 38// An Expr represents an input element. 39type Expr interface { 40 // Span returns the start and end position of the expression, 41 // excluding leading or trailing comments. 42 Span() (start, end Position) 43 44 // Comment returns the comments attached to the expression. 45 // This method would normally be named 'Comments' but that 46 // would interfere with embedding a type of the same name. 47 Comment() *Comments 48} 49 50// A Comment represents a single // comment. 51type Comment struct { 52 Start Position 53 Token string // without trailing newline 54 Suffix bool // an end of line (not whole line) comment 55} 56 57// Comments collects the comments associated with an expression. 58type Comments struct { 59 Before []Comment // whole-line comments before this expression 60 Suffix []Comment // end-of-line comments after this expression 61 62 // For top-level expressions only, After lists whole-line 63 // comments following the expression. 64 After []Comment 65} 66 67// Comment returns the receiver. This isn't useful by itself, but 68// a Comments struct is embedded into all the expression 69// implementation types, and this gives each of those a Comment 70// method to satisfy the Expr interface. 71func (c *Comments) Comment() *Comments { 72 return c 73} 74 75// A FileSyntax represents an entire go.mod file. 76type FileSyntax struct { 77 Name string // file path 78 Comments 79 Stmt []Expr 80} 81 82func (x *FileSyntax) Span() (start, end Position) { 83 if len(x.Stmt) == 0 { 84 return 85 } 86 start, _ = x.Stmt[0].Span() 87 _, end = x.Stmt[len(x.Stmt)-1].Span() 88 return start, end 89} 90 91// addLine adds a line containing the given tokens to the file. 92// 93// If the first token of the hint matches the first token of the 94// line, the new line is added at the end of the block containing hint, 95// extracting hint into a new block if it is not yet in one. 96// 97// If the hint is non-nil buts its first token does not match, 98// the new line is added after the block containing hint 99// (or hint itself, if not in a block). 100// 101// If no hint is provided, addLine appends the line to the end of 102// the last block with a matching first token, 103// or to the end of the file if no such block exists. 104func (x *FileSyntax) addLine(hint Expr, tokens ...string) *Line { 105 if hint == nil { 106 // If no hint given, add to the last statement of the given type. 107 Loop: 108 for i := len(x.Stmt) - 1; i >= 0; i-- { 109 stmt := x.Stmt[i] 110 switch stmt := stmt.(type) { 111 case *Line: 112 if stmt.Token != nil && stmt.Token[0] == tokens[0] { 113 hint = stmt 114 break Loop 115 } 116 case *LineBlock: 117 if stmt.Token[0] == tokens[0] { 118 hint = stmt 119 break Loop 120 } 121 } 122 } 123 } 124 125 newLineAfter := func(i int) *Line { 126 new := &Line{Token: tokens} 127 if i == len(x.Stmt) { 128 x.Stmt = append(x.Stmt, new) 129 } else { 130 x.Stmt = append(x.Stmt, nil) 131 copy(x.Stmt[i+2:], x.Stmt[i+1:]) 132 x.Stmt[i+1] = new 133 } 134 return new 135 } 136 137 if hint != nil { 138 for i, stmt := range x.Stmt { 139 switch stmt := stmt.(type) { 140 case *Line: 141 if stmt == hint { 142 if stmt.Token == nil || stmt.Token[0] != tokens[0] { 143 return newLineAfter(i) 144 } 145 146 // Convert line to line block. 147 stmt.InBlock = true 148 block := &LineBlock{Token: stmt.Token[:1], Line: []*Line{stmt}} 149 stmt.Token = stmt.Token[1:] 150 x.Stmt[i] = block 151 new := &Line{Token: tokens[1:], InBlock: true} 152 block.Line = append(block.Line, new) 153 return new 154 } 155 156 case *LineBlock: 157 if stmt == hint { 158 if stmt.Token[0] != tokens[0] { 159 return newLineAfter(i) 160 } 161 162 new := &Line{Token: tokens[1:], InBlock: true} 163 stmt.Line = append(stmt.Line, new) 164 return new 165 } 166 167 for j, line := range stmt.Line { 168 if line == hint { 169 if stmt.Token[0] != tokens[0] { 170 return newLineAfter(i) 171 } 172 173 // Add new line after hint within the block. 174 stmt.Line = append(stmt.Line, nil) 175 copy(stmt.Line[j+2:], stmt.Line[j+1:]) 176 new := &Line{Token: tokens[1:], InBlock: true} 177 stmt.Line[j+1] = new 178 return new 179 } 180 } 181 } 182 } 183 } 184 185 new := &Line{Token: tokens} 186 x.Stmt = append(x.Stmt, new) 187 return new 188} 189 190func (x *FileSyntax) updateLine(line *Line, tokens ...string) { 191 if line.InBlock { 192 tokens = tokens[1:] 193 } 194 line.Token = tokens 195} 196 197func (x *FileSyntax) removeLine(line *Line) { 198 line.Token = nil 199} 200 201// Cleanup cleans up the file syntax x after any edit operations. 202// To avoid quadratic behavior, removeLine marks the line as dead 203// by setting line.Token = nil but does not remove it from the slice 204// in which it appears. After edits have all been indicated, 205// calling Cleanup cleans out the dead lines. 206func (x *FileSyntax) Cleanup() { 207 w := 0 208 for _, stmt := range x.Stmt { 209 switch stmt := stmt.(type) { 210 case *Line: 211 if stmt.Token == nil { 212 continue 213 } 214 case *LineBlock: 215 ww := 0 216 for _, line := range stmt.Line { 217 if line.Token != nil { 218 stmt.Line[ww] = line 219 ww++ 220 } 221 } 222 if ww == 0 { 223 continue 224 } 225 if ww == 1 { 226 // Collapse block into single line. 227 line := &Line{ 228 Comments: Comments{ 229 Before: commentsAdd(stmt.Before, stmt.Line[0].Before), 230 Suffix: commentsAdd(stmt.Line[0].Suffix, stmt.Suffix), 231 After: commentsAdd(stmt.Line[0].After, stmt.After), 232 }, 233 Token: stringsAdd(stmt.Token, stmt.Line[0].Token), 234 } 235 x.Stmt[w] = line 236 w++ 237 continue 238 } 239 stmt.Line = stmt.Line[:ww] 240 } 241 x.Stmt[w] = stmt 242 w++ 243 } 244 x.Stmt = x.Stmt[:w] 245} 246 247func commentsAdd(x, y []Comment) []Comment { 248 return append(x[:len(x):len(x)], y...) 249} 250 251func stringsAdd(x, y []string) []string { 252 return append(x[:len(x):len(x)], y...) 253} 254 255// A CommentBlock represents a top-level block of comments separate 256// from any rule. 257type CommentBlock struct { 258 Comments 259 Start Position 260} 261 262func (x *CommentBlock) Span() (start, end Position) { 263 return x.Start, x.Start 264} 265 266// A Line is a single line of tokens. 267type Line struct { 268 Comments 269 Start Position 270 Token []string 271 InBlock bool 272 End Position 273} 274 275func (x *Line) Span() (start, end Position) { 276 return x.Start, x.End 277} 278 279// A LineBlock is a factored block of lines, like 280// 281// require ( 282// "x" 283// "y" 284// ) 285// 286type LineBlock struct { 287 Comments 288 Start Position 289 LParen LParen 290 Token []string 291 Line []*Line 292 RParen RParen 293} 294 295func (x *LineBlock) Span() (start, end Position) { 296 return x.Start, x.RParen.Pos.add(")") 297} 298 299// An LParen represents the beginning of a parenthesized line block. 300// It is a place to store suffix comments. 301type LParen struct { 302 Comments 303 Pos Position 304} 305 306func (x *LParen) Span() (start, end Position) { 307 return x.Pos, x.Pos.add(")") 308} 309 310// An RParen represents the end of a parenthesized line block. 311// It is a place to store whole-line (before) comments. 312type RParen struct { 313 Comments 314 Pos Position 315} 316 317func (x *RParen) Span() (start, end Position) { 318 return x.Pos, x.Pos.add(")") 319} 320 321// An input represents a single input file being parsed. 322type input struct { 323 // Lexing state. 324 filename string // name of input file, for errors 325 complete []byte // entire input 326 remaining []byte // remaining input 327 tokenStart []byte // token being scanned to end of input 328 token token // next token to be returned by lex, peek 329 pos Position // current input position 330 comments []Comment // accumulated comments 331 332 // Parser state. 333 file *FileSyntax // returned top-level syntax tree 334 parseErrors ErrorList // errors encountered during parsing 335 336 // Comment assignment state. 337 pre []Expr // all expressions, in preorder traversal 338 post []Expr // all expressions, in postorder traversal 339} 340 341func newInput(filename string, data []byte) *input { 342 return &input{ 343 filename: filename, 344 complete: data, 345 remaining: data, 346 pos: Position{Line: 1, LineRune: 1, Byte: 0}, 347 } 348} 349 350// parse parses the input file. 351func parse(file string, data []byte) (f *FileSyntax, err error) { 352 // The parser panics for both routine errors like syntax errors 353 // and for programmer bugs like array index errors. 354 // Turn both into error returns. Catching bug panics is 355 // especially important when processing many files. 356 in := newInput(file, data) 357 defer func() { 358 if e := recover(); e != nil && e != &in.parseErrors { 359 in.parseErrors = append(in.parseErrors, Error{ 360 Filename: in.filename, 361 Pos: in.pos, 362 Err: fmt.Errorf("internal error: %v", e), 363 }) 364 } 365 if err == nil && len(in.parseErrors) > 0 { 366 err = in.parseErrors 367 } 368 }() 369 370 // Prime the lexer by reading in the first token. It will be available 371 // in the next peek() or lex() call. 372 in.readToken() 373 374 // Invoke the parser. 375 in.parseFile() 376 if len(in.parseErrors) > 0 { 377 return nil, in.parseErrors 378 } 379 in.file.Name = in.filename 380 381 // Assign comments to nearby syntax. 382 in.assignComments() 383 384 return in.file, nil 385} 386 387// Error is called to report an error. 388// Error does not return: it panics. 389func (in *input) Error(s string) { 390 in.parseErrors = append(in.parseErrors, Error{ 391 Filename: in.filename, 392 Pos: in.pos, 393 Err: errors.New(s), 394 }) 395 panic(&in.parseErrors) 396} 397 398// eof reports whether the input has reached end of file. 399func (in *input) eof() bool { 400 return len(in.remaining) == 0 401} 402 403// peekRune returns the next rune in the input without consuming it. 404func (in *input) peekRune() int { 405 if len(in.remaining) == 0 { 406 return 0 407 } 408 r, _ := utf8.DecodeRune(in.remaining) 409 return int(r) 410} 411 412// peekPrefix reports whether the remaining input begins with the given prefix. 413func (in *input) peekPrefix(prefix string) bool { 414 // This is like bytes.HasPrefix(in.remaining, []byte(prefix)) 415 // but without the allocation of the []byte copy of prefix. 416 for i := 0; i < len(prefix); i++ { 417 if i >= len(in.remaining) || in.remaining[i] != prefix[i] { 418 return false 419 } 420 } 421 return true 422} 423 424// readRune consumes and returns the next rune in the input. 425func (in *input) readRune() int { 426 if len(in.remaining) == 0 { 427 in.Error("internal lexer error: readRune at EOF") 428 } 429 r, size := utf8.DecodeRune(in.remaining) 430 in.remaining = in.remaining[size:] 431 if r == '\n' { 432 in.pos.Line++ 433 in.pos.LineRune = 1 434 } else { 435 in.pos.LineRune++ 436 } 437 in.pos.Byte += size 438 return int(r) 439} 440 441type token struct { 442 kind tokenKind 443 pos Position 444 endPos Position 445 text string 446} 447 448type tokenKind int 449 450const ( 451 _EOF tokenKind = -(iota + 1) 452 _EOLCOMMENT 453 _IDENT 454 _STRING 455 _COMMENT 456 457 // newlines and punctuation tokens are allowed as ASCII codes. 458) 459 460func (k tokenKind) isComment() bool { 461 return k == _COMMENT || k == _EOLCOMMENT 462} 463 464// isEOL returns whether a token terminates a line. 465func (k tokenKind) isEOL() bool { 466 return k == _EOF || k == _EOLCOMMENT || k == '\n' 467} 468 469// startToken marks the beginning of the next input token. 470// It must be followed by a call to endToken, once the token's text has 471// been consumed using readRune. 472func (in *input) startToken() { 473 in.tokenStart = in.remaining 474 in.token.text = "" 475 in.token.pos = in.pos 476} 477 478// endToken marks the end of an input token. 479// It records the actual token string in tok.text. 480// A single trailing newline (LF or CRLF) will be removed from comment tokens. 481func (in *input) endToken(kind tokenKind) { 482 in.token.kind = kind 483 text := string(in.tokenStart[:len(in.tokenStart)-len(in.remaining)]) 484 if kind.isComment() { 485 if strings.HasSuffix(text, "\r\n") { 486 text = text[:len(text)-2] 487 } else { 488 text = strings.TrimSuffix(text, "\n") 489 } 490 } 491 in.token.text = text 492 in.token.endPos = in.pos 493} 494 495// peek returns the kind of the the next token returned by lex. 496func (in *input) peek() tokenKind { 497 return in.token.kind 498} 499 500// lex is called from the parser to obtain the next input token. 501func (in *input) lex() token { 502 tok := in.token 503 in.readToken() 504 return tok 505} 506 507// readToken lexes the next token from the text and stores it in in.token. 508func (in *input) readToken() { 509 // Skip past spaces, stopping at non-space or EOF. 510 for !in.eof() { 511 c := in.peekRune() 512 if c == ' ' || c == '\t' || c == '\r' { 513 in.readRune() 514 continue 515 } 516 517 // Comment runs to end of line. 518 if in.peekPrefix("//") { 519 in.startToken() 520 521 // Is this comment the only thing on its line? 522 // Find the last \n before this // and see if it's all 523 // spaces from there to here. 524 i := bytes.LastIndex(in.complete[:in.pos.Byte], []byte("\n")) 525 suffix := len(bytes.TrimSpace(in.complete[i+1:in.pos.Byte])) > 0 526 in.readRune() 527 in.readRune() 528 529 // Consume comment. 530 for len(in.remaining) > 0 && in.readRune() != '\n' { 531 } 532 533 // If we are at top level (not in a statement), hand the comment to 534 // the parser as a _COMMENT token. The grammar is written 535 // to handle top-level comments itself. 536 if !suffix { 537 in.endToken(_COMMENT) 538 return 539 } 540 541 // Otherwise, save comment for later attachment to syntax tree. 542 in.endToken(_EOLCOMMENT) 543 in.comments = append(in.comments, Comment{in.token.pos, in.token.text, suffix}) 544 return 545 } 546 547 if in.peekPrefix("/*") { 548 in.Error("mod files must use // comments (not /* */ comments)") 549 } 550 551 // Found non-space non-comment. 552 break 553 } 554 555 // Found the beginning of the next token. 556 in.startToken() 557 558 // End of file. 559 if in.eof() { 560 in.endToken(_EOF) 561 return 562 } 563 564 // Punctuation tokens. 565 switch c := in.peekRune(); c { 566 case '\n', '(', ')', '[', ']', '{', '}', ',': 567 in.readRune() 568 in.endToken(tokenKind(c)) 569 return 570 571 case '"', '`': // quoted string 572 quote := c 573 in.readRune() 574 for { 575 if in.eof() { 576 in.pos = in.token.pos 577 in.Error("unexpected EOF in string") 578 } 579 if in.peekRune() == '\n' { 580 in.Error("unexpected newline in string") 581 } 582 c := in.readRune() 583 if c == quote { 584 break 585 } 586 if c == '\\' && quote != '`' { 587 if in.eof() { 588 in.pos = in.token.pos 589 in.Error("unexpected EOF in string") 590 } 591 in.readRune() 592 } 593 } 594 in.endToken(_STRING) 595 return 596 } 597 598 // Checked all punctuation. Must be identifier token. 599 if c := in.peekRune(); !isIdent(c) { 600 in.Error(fmt.Sprintf("unexpected input character %#q", c)) 601 } 602 603 // Scan over identifier. 604 for isIdent(in.peekRune()) { 605 if in.peekPrefix("//") { 606 break 607 } 608 if in.peekPrefix("/*") { 609 in.Error("mod files must use // comments (not /* */ comments)") 610 } 611 in.readRune() 612 } 613 in.endToken(_IDENT) 614} 615 616// isIdent reports whether c is an identifier rune. 617// We treat most printable runes as identifier runes, except for a handful of 618// ASCII punctuation characters. 619func isIdent(c int) bool { 620 switch r := rune(c); r { 621 case ' ', '(', ')', '[', ']', '{', '}', ',': 622 return false 623 default: 624 return !unicode.IsSpace(r) && unicode.IsPrint(r) 625 } 626} 627 628// Comment assignment. 629// We build two lists of all subexpressions, preorder and postorder. 630// The preorder list is ordered by start location, with outer expressions first. 631// The postorder list is ordered by end location, with outer expressions last. 632// We use the preorder list to assign each whole-line comment to the syntax 633// immediately following it, and we use the postorder list to assign each 634// end-of-line comment to the syntax immediately preceding it. 635 636// order walks the expression adding it and its subexpressions to the 637// preorder and postorder lists. 638func (in *input) order(x Expr) { 639 if x != nil { 640 in.pre = append(in.pre, x) 641 } 642 switch x := x.(type) { 643 default: 644 panic(fmt.Errorf("order: unexpected type %T", x)) 645 case nil: 646 // nothing 647 case *LParen, *RParen: 648 // nothing 649 case *CommentBlock: 650 // nothing 651 case *Line: 652 // nothing 653 case *FileSyntax: 654 for _, stmt := range x.Stmt { 655 in.order(stmt) 656 } 657 case *LineBlock: 658 in.order(&x.LParen) 659 for _, l := range x.Line { 660 in.order(l) 661 } 662 in.order(&x.RParen) 663 } 664 if x != nil { 665 in.post = append(in.post, x) 666 } 667} 668 669// assignComments attaches comments to nearby syntax. 670func (in *input) assignComments() { 671 const debug = false 672 673 // Generate preorder and postorder lists. 674 in.order(in.file) 675 676 // Split into whole-line comments and suffix comments. 677 var line, suffix []Comment 678 for _, com := range in.comments { 679 if com.Suffix { 680 suffix = append(suffix, com) 681 } else { 682 line = append(line, com) 683 } 684 } 685 686 if debug { 687 for _, c := range line { 688 fmt.Fprintf(os.Stderr, "LINE %q :%d:%d #%d\n", c.Token, c.Start.Line, c.Start.LineRune, c.Start.Byte) 689 } 690 } 691 692 // Assign line comments to syntax immediately following. 693 for _, x := range in.pre { 694 start, _ := x.Span() 695 if debug { 696 fmt.Fprintf(os.Stderr, "pre %T :%d:%d #%d\n", x, start.Line, start.LineRune, start.Byte) 697 } 698 xcom := x.Comment() 699 for len(line) > 0 && start.Byte >= line[0].Start.Byte { 700 if debug { 701 fmt.Fprintf(os.Stderr, "ASSIGN LINE %q #%d\n", line[0].Token, line[0].Start.Byte) 702 } 703 xcom.Before = append(xcom.Before, line[0]) 704 line = line[1:] 705 } 706 } 707 708 // Remaining line comments go at end of file. 709 in.file.After = append(in.file.After, line...) 710 711 if debug { 712 for _, c := range suffix { 713 fmt.Fprintf(os.Stderr, "SUFFIX %q :%d:%d #%d\n", c.Token, c.Start.Line, c.Start.LineRune, c.Start.Byte) 714 } 715 } 716 717 // Assign suffix comments to syntax immediately before. 718 for i := len(in.post) - 1; i >= 0; i-- { 719 x := in.post[i] 720 721 start, end := x.Span() 722 if debug { 723 fmt.Fprintf(os.Stderr, "post %T :%d:%d #%d :%d:%d #%d\n", x, start.Line, start.LineRune, start.Byte, end.Line, end.LineRune, end.Byte) 724 } 725 726 // Do not assign suffix comments to end of line block or whole file. 727 // Instead assign them to the last element inside. 728 switch x.(type) { 729 case *FileSyntax: 730 continue 731 } 732 733 // Do not assign suffix comments to something that starts 734 // on an earlier line, so that in 735 // 736 // x ( y 737 // z ) // comment 738 // 739 // we assign the comment to z and not to x ( ... ). 740 if start.Line != end.Line { 741 continue 742 } 743 xcom := x.Comment() 744 for len(suffix) > 0 && end.Byte <= suffix[len(suffix)-1].Start.Byte { 745 if debug { 746 fmt.Fprintf(os.Stderr, "ASSIGN SUFFIX %q #%d\n", suffix[len(suffix)-1].Token, suffix[len(suffix)-1].Start.Byte) 747 } 748 xcom.Suffix = append(xcom.Suffix, suffix[len(suffix)-1]) 749 suffix = suffix[:len(suffix)-1] 750 } 751 } 752 753 // We assigned suffix comments in reverse. 754 // If multiple suffix comments were appended to the same 755 // expression node, they are now in reverse. Fix that. 756 for _, x := range in.post { 757 reverseComments(x.Comment().Suffix) 758 } 759 760 // Remaining suffix comments go at beginning of file. 761 in.file.Before = append(in.file.Before, suffix...) 762} 763 764// reverseComments reverses the []Comment list. 765func reverseComments(list []Comment) { 766 for i, j := 0, len(list)-1; i < j; i, j = i+1, j-1 { 767 list[i], list[j] = list[j], list[i] 768 } 769} 770 771func (in *input) parseFile() { 772 in.file = new(FileSyntax) 773 var cb *CommentBlock 774 for { 775 switch in.peek() { 776 case '\n': 777 in.lex() 778 if cb != nil { 779 in.file.Stmt = append(in.file.Stmt, cb) 780 cb = nil 781 } 782 case _COMMENT: 783 tok := in.lex() 784 if cb == nil { 785 cb = &CommentBlock{Start: tok.pos} 786 } 787 com := cb.Comment() 788 com.Before = append(com.Before, Comment{Start: tok.pos, Token: tok.text}) 789 case _EOF: 790 if cb != nil { 791 in.file.Stmt = append(in.file.Stmt, cb) 792 } 793 return 794 default: 795 in.parseStmt() 796 if cb != nil { 797 in.file.Stmt[len(in.file.Stmt)-1].Comment().Before = cb.Before 798 cb = nil 799 } 800 } 801 } 802} 803 804func (in *input) parseStmt() { 805 tok := in.lex() 806 start := tok.pos 807 end := tok.endPos 808 tokens := []string{tok.text} 809 for { 810 tok := in.lex() 811 switch { 812 case tok.kind.isEOL(): 813 in.file.Stmt = append(in.file.Stmt, &Line{ 814 Start: start, 815 Token: tokens, 816 End: end, 817 }) 818 return 819 820 case tok.kind == '(': 821 if next := in.peek(); next.isEOL() { 822 // Start of block: no more tokens on this line. 823 in.file.Stmt = append(in.file.Stmt, in.parseLineBlock(start, tokens, tok)) 824 return 825 } else if next == ')' { 826 rparen := in.lex() 827 if in.peek().isEOL() { 828 // Empty block. 829 in.lex() 830 in.file.Stmt = append(in.file.Stmt, &LineBlock{ 831 Start: start, 832 Token: tokens, 833 LParen: LParen{Pos: tok.pos}, 834 RParen: RParen{Pos: rparen.pos}, 835 }) 836 return 837 } 838 // '( )' in the middle of the line, not a block. 839 tokens = append(tokens, tok.text, rparen.text) 840 } else { 841 // '(' in the middle of the line, not a block. 842 tokens = append(tokens, tok.text) 843 } 844 845 default: 846 tokens = append(tokens, tok.text) 847 end = tok.endPos 848 } 849 } 850} 851 852func (in *input) parseLineBlock(start Position, token []string, lparen token) *LineBlock { 853 x := &LineBlock{ 854 Start: start, 855 Token: token, 856 LParen: LParen{Pos: lparen.pos}, 857 } 858 var comments []Comment 859 for { 860 switch in.peek() { 861 case _EOLCOMMENT: 862 // Suffix comment, will be attached later by assignComments. 863 in.lex() 864 case '\n': 865 // Blank line. Add an empty comment to preserve it. 866 in.lex() 867 if len(comments) == 0 && len(x.Line) > 0 || len(comments) > 0 && comments[len(comments)-1].Token != "" { 868 comments = append(comments, Comment{}) 869 } 870 case _COMMENT: 871 tok := in.lex() 872 comments = append(comments, Comment{Start: tok.pos, Token: tok.text}) 873 case _EOF: 874 in.Error(fmt.Sprintf("syntax error (unterminated block started at %s:%d:%d)", in.filename, x.Start.Line, x.Start.LineRune)) 875 case ')': 876 rparen := in.lex() 877 x.RParen.Before = comments 878 x.RParen.Pos = rparen.pos 879 if !in.peek().isEOL() { 880 in.Error("syntax error (expected newline after closing paren)") 881 } 882 in.lex() 883 return x 884 default: 885 l := in.parseLine() 886 x.Line = append(x.Line, l) 887 l.Comment().Before = comments 888 comments = nil 889 } 890 } 891} 892 893func (in *input) parseLine() *Line { 894 tok := in.lex() 895 if tok.kind.isEOL() { 896 in.Error("internal parse error: parseLine at end of line") 897 } 898 start := tok.pos 899 end := tok.endPos 900 tokens := []string{tok.text} 901 for { 902 tok := in.lex() 903 if tok.kind.isEOL() { 904 return &Line{ 905 Start: start, 906 Token: tokens, 907 End: end, 908 InBlock: true, 909 } 910 } 911 tokens = append(tokens, tok.text) 912 end = tok.endPos 913 } 914} 915 916var ( 917 slashSlash = []byte("//") 918 moduleStr = []byte("module") 919) 920 921// ModulePath returns the module path from the gomod file text. 922// If it cannot find a module path, it returns an empty string. 923// It is tolerant of unrelated problems in the go.mod file. 924func ModulePath(mod []byte) string { 925 for len(mod) > 0 { 926 line := mod 927 mod = nil 928 if i := bytes.IndexByte(line, '\n'); i >= 0 { 929 line, mod = line[:i], line[i+1:] 930 } 931 if i := bytes.Index(line, slashSlash); i >= 0 { 932 line = line[:i] 933 } 934 line = bytes.TrimSpace(line) 935 if !bytes.HasPrefix(line, moduleStr) { 936 continue 937 } 938 line = line[len(moduleStr):] 939 n := len(line) 940 line = bytes.TrimSpace(line) 941 if len(line) == n || len(line) == 0 { 942 continue 943 } 944 945 if line[0] == '"' || line[0] == '`' { 946 p, err := strconv.Unquote(string(line)) 947 if err != nil { 948 return "" // malformed quoted string or multiline module path 949 } 950 return p 951 } 952 953 return string(line) 954 } 955 return "" // missing module path 956} 957