1 //===--- UnwrappedLineParser.cpp - Format C++ code ------------------------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 /// 10 /// \file 11 /// \brief This file contains the implementation of the UnwrappedLineParser, 12 /// which turns a stream of tokens into UnwrappedLines. 13 /// 14 //===----------------------------------------------------------------------===// 15 16 #define DEBUG_TYPE "format-parser" 17 18 #include "UnwrappedLineParser.h" 19 #include "llvm/Support/Debug.h" 20 21 namespace clang { 22 namespace format { 23 24 class FormatTokenSource { 25 public: 26 virtual ~FormatTokenSource() {} 27 virtual FormatToken *getNextToken() = 0; 28 29 virtual unsigned getPosition() = 0; 30 virtual FormatToken *setPosition(unsigned Position) = 0; 31 }; 32 33 namespace { 34 35 class ScopedDeclarationState { 36 public: 37 ScopedDeclarationState(UnwrappedLine &Line, std::vector<bool> &Stack, 38 bool MustBeDeclaration) 39 : Line(Line), Stack(Stack) { 40 Line.MustBeDeclaration = MustBeDeclaration; 41 Stack.push_back(MustBeDeclaration); 42 } 43 ~ScopedDeclarationState() { 44 Stack.pop_back(); 45 if (!Stack.empty()) 46 Line.MustBeDeclaration = Stack.back(); 47 else 48 Line.MustBeDeclaration = true; 49 } 50 51 private: 52 UnwrappedLine &Line; 53 std::vector<bool> &Stack; 54 }; 55 56 class ScopedMacroState : public FormatTokenSource { 57 public: 58 ScopedMacroState(UnwrappedLine &Line, FormatTokenSource *&TokenSource, 59 FormatToken *&ResetToken, bool &StructuralError) 60 : Line(Line), TokenSource(TokenSource), ResetToken(ResetToken), 61 PreviousLineLevel(Line.Level), PreviousTokenSource(TokenSource), 62 StructuralError(StructuralError), 63 PreviousStructuralError(StructuralError), Token(NULL) { 64 TokenSource = this; 65 Line.Level = 0; 66 Line.InPPDirective = true; 67 } 68 69 ~ScopedMacroState() { 70 TokenSource = PreviousTokenSource; 71 ResetToken = Token; 72 Line.InPPDirective = false; 73 Line.Level = PreviousLineLevel; 74 StructuralError = PreviousStructuralError; 75 } 76 77 virtual FormatToken *getNextToken() { 78 // The \c UnwrappedLineParser guards against this by never calling 79 // \c getNextToken() after it has encountered the first eof token. 80 assert(!eof()); 81 Token = PreviousTokenSource->getNextToken(); 82 if (eof()) 83 return getFakeEOF(); 84 return Token; 85 } 86 87 virtual unsigned getPosition() { return PreviousTokenSource->getPosition(); } 88 89 virtual FormatToken *setPosition(unsigned Position) { 90 Token = PreviousTokenSource->setPosition(Position); 91 return Token; 92 } 93 94 private: 95 bool eof() { return Token && Token->HasUnescapedNewline; } 96 97 FormatToken *getFakeEOF() { 98 static bool EOFInitialized = false; 99 static FormatToken FormatTok; 100 if (!EOFInitialized) { 101 FormatTok.Tok.startToken(); 102 FormatTok.Tok.setKind(tok::eof); 103 EOFInitialized = true; 104 } 105 return &FormatTok; 106 } 107 108 UnwrappedLine &Line; 109 FormatTokenSource *&TokenSource; 110 FormatToken *&ResetToken; 111 unsigned PreviousLineLevel; 112 FormatTokenSource *PreviousTokenSource; 113 bool &StructuralError; 114 bool PreviousStructuralError; 115 116 FormatToken *Token; 117 }; 118 119 } // end anonymous namespace 120 121 class ScopedLineState { 122 public: 123 ScopedLineState(UnwrappedLineParser &Parser, 124 bool SwitchToPreprocessorLines = false) 125 : Parser(Parser) { 126 OriginalLines = Parser.CurrentLines; 127 if (SwitchToPreprocessorLines) 128 Parser.CurrentLines = &Parser.PreprocessorDirectives; 129 else if (!Parser.Line->Tokens.empty()) 130 Parser.CurrentLines = &Parser.Line->Tokens.back().Children; 131 PreBlockLine = Parser.Line.take(); 132 Parser.Line.reset(new UnwrappedLine()); 133 Parser.Line->Level = PreBlockLine->Level; 134 Parser.Line->InPPDirective = PreBlockLine->InPPDirective; 135 } 136 137 ~ScopedLineState() { 138 if (!Parser.Line->Tokens.empty()) { 139 Parser.addUnwrappedLine(); 140 } 141 assert(Parser.Line->Tokens.empty()); 142 Parser.Line.reset(PreBlockLine); 143 if (Parser.CurrentLines == &Parser.PreprocessorDirectives) 144 Parser.MustBreakBeforeNextToken = true; 145 Parser.CurrentLines = OriginalLines; 146 } 147 148 private: 149 UnwrappedLineParser &Parser; 150 151 UnwrappedLine *PreBlockLine; 152 SmallVectorImpl<UnwrappedLine> *OriginalLines; 153 }; 154 155 class CompoundStatementIndenter { 156 public: 157 CompoundStatementIndenter(UnwrappedLineParser *Parser, 158 const FormatStyle &Style, unsigned &LineLevel) 159 : LineLevel(LineLevel), OldLineLevel(LineLevel) { 160 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman) { 161 Parser->addUnwrappedLine(); 162 } else if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 163 Parser->addUnwrappedLine(); 164 ++LineLevel; 165 } 166 } 167 ~CompoundStatementIndenter() { 168 LineLevel = OldLineLevel; 169 } 170 171 private: 172 unsigned &LineLevel; 173 unsigned OldLineLevel; 174 }; 175 176 namespace { 177 178 class IndexedTokenSource : public FormatTokenSource { 179 public: 180 IndexedTokenSource(ArrayRef<FormatToken *> Tokens) 181 : Tokens(Tokens), Position(-1) {} 182 183 virtual FormatToken *getNextToken() { 184 ++Position; 185 return Tokens[Position]; 186 } 187 188 virtual unsigned getPosition() { 189 assert(Position >= 0); 190 return Position; 191 } 192 193 virtual FormatToken *setPosition(unsigned P) { 194 Position = P; 195 return Tokens[Position]; 196 } 197 198 void reset() { Position = -1; } 199 200 private: 201 ArrayRef<FormatToken *> Tokens; 202 int Position; 203 }; 204 205 } // end anonymous namespace 206 207 UnwrappedLineParser::UnwrappedLineParser(const FormatStyle &Style, 208 ArrayRef<FormatToken *> Tokens, 209 UnwrappedLineConsumer &Callback) 210 : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), 211 CurrentLines(&Lines), StructuralError(false), Style(Style), Tokens(NULL), 212 Callback(Callback), AllTokens(Tokens), PPBranchLevel(-1) {} 213 214 void UnwrappedLineParser::reset() { 215 PPBranchLevel = -1; 216 Line.reset(new UnwrappedLine); 217 CommentsBeforeNextToken.clear(); 218 FormatTok = NULL; 219 MustBreakBeforeNextToken = false; 220 PreprocessorDirectives.clear(); 221 CurrentLines = &Lines; 222 DeclarationScopeStack.clear(); 223 StructuralError = false; 224 PPStack.clear(); 225 } 226 227 bool UnwrappedLineParser::parse() { 228 IndexedTokenSource TokenSource(AllTokens); 229 do { 230 DEBUG(llvm::dbgs() << "----\n"); 231 reset(); 232 Tokens = &TokenSource; 233 TokenSource.reset(); 234 235 readToken(); 236 parseFile(); 237 // Create line with eof token. 238 pushToken(FormatTok); 239 addUnwrappedLine(); 240 241 for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(), 242 E = Lines.end(); 243 I != E; ++I) { 244 Callback.consumeUnwrappedLine(*I); 245 } 246 Callback.finishRun(); 247 Lines.clear(); 248 while (!PPLevelBranchIndex.empty() && 249 PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) { 250 PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1); 251 PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1); 252 } 253 if (!PPLevelBranchIndex.empty()) { 254 ++PPLevelBranchIndex.back(); 255 assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size()); 256 assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back()); 257 } 258 } while (!PPLevelBranchIndex.empty()); 259 260 return StructuralError; 261 } 262 263 void UnwrappedLineParser::parseFile() { 264 ScopedDeclarationState DeclarationState( 265 *Line, DeclarationScopeStack, 266 /*MustBeDeclaration=*/ !Line->InPPDirective); 267 parseLevel(/*HasOpeningBrace=*/false); 268 // Make sure to format the remaining tokens. 269 flushComments(true); 270 addUnwrappedLine(); 271 } 272 273 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) { 274 bool SwitchLabelEncountered = false; 275 do { 276 switch (FormatTok->Tok.getKind()) { 277 case tok::comment: 278 nextToken(); 279 addUnwrappedLine(); 280 break; 281 case tok::l_brace: 282 // FIXME: Add parameter whether this can happen - if this happens, we must 283 // be in a non-declaration context. 284 parseBlock(/*MustBeDeclaration=*/false); 285 addUnwrappedLine(); 286 break; 287 case tok::r_brace: 288 if (HasOpeningBrace) 289 return; 290 StructuralError = true; 291 nextToken(); 292 addUnwrappedLine(); 293 break; 294 case tok::kw_default: 295 case tok::kw_case: 296 if (!SwitchLabelEncountered && 297 (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1))) 298 ++Line->Level; 299 SwitchLabelEncountered = true; 300 parseStructuralElement(); 301 break; 302 default: 303 parseStructuralElement(); 304 break; 305 } 306 } while (!eof()); 307 } 308 309 void UnwrappedLineParser::calculateBraceTypes() { 310 // We'll parse forward through the tokens until we hit 311 // a closing brace or eof - note that getNextToken() will 312 // parse macros, so this will magically work inside macro 313 // definitions, too. 314 unsigned StoredPosition = Tokens->getPosition(); 315 unsigned Position = StoredPosition; 316 FormatToken *Tok = FormatTok; 317 // Keep a stack of positions of lbrace tokens. We will 318 // update information about whether an lbrace starts a 319 // braced init list or a different block during the loop. 320 SmallVector<FormatToken *, 8> LBraceStack; 321 assert(Tok->Tok.is(tok::l_brace)); 322 do { 323 // Get next none-comment token. 324 FormatToken *NextTok; 325 unsigned ReadTokens = 0; 326 do { 327 NextTok = Tokens->getNextToken(); 328 ++ReadTokens; 329 } while (NextTok->is(tok::comment)); 330 331 switch (Tok->Tok.getKind()) { 332 case tok::l_brace: 333 LBraceStack.push_back(Tok); 334 break; 335 case tok::r_brace: 336 if (!LBraceStack.empty()) { 337 if (LBraceStack.back()->BlockKind == BK_Unknown) { 338 // If there is a comma, semicolon or right paren after the closing 339 // brace, we assume this is a braced initializer list. Note that 340 // regardless how we mark inner braces here, we will overwrite the 341 // BlockKind later if we parse a braced list (where all blocks inside 342 // are by default braced lists), or when we explicitly detect blocks 343 // (for example while parsing lambdas). 344 // 345 // We exclude + and - as they can be ObjC visibility modifiers. 346 if (NextTok->isOneOf(tok::comma, tok::semi, tok::r_paren, tok::period, 347 tok::r_square, tok::l_brace, tok::colon) || 348 (NextTok->isBinaryOperator() && 349 !NextTok->isOneOf(tok::plus, tok::minus))) { 350 Tok->BlockKind = BK_BracedInit; 351 LBraceStack.back()->BlockKind = BK_BracedInit; 352 } else { 353 Tok->BlockKind = BK_Block; 354 LBraceStack.back()->BlockKind = BK_Block; 355 } 356 } 357 LBraceStack.pop_back(); 358 } 359 break; 360 case tok::semi: 361 case tok::kw_if: 362 case tok::kw_while: 363 case tok::kw_for: 364 case tok::kw_switch: 365 case tok::kw_try: 366 if (!LBraceStack.empty()) 367 LBraceStack.back()->BlockKind = BK_Block; 368 break; 369 default: 370 break; 371 } 372 Tok = NextTok; 373 Position += ReadTokens; 374 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 375 // Assume other blocks for all unclosed opening braces. 376 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 377 if (LBraceStack[i]->BlockKind == BK_Unknown) 378 LBraceStack[i]->BlockKind = BK_Block; 379 } 380 381 FormatTok = Tokens->setPosition(StoredPosition); 382 } 383 384 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 385 bool MunchSemi) { 386 assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected"); 387 unsigned InitialLevel = Line->Level; 388 nextToken(); 389 390 addUnwrappedLine(); 391 392 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 393 MustBeDeclaration); 394 if (AddLevel) 395 ++Line->Level; 396 parseLevel(/*HasOpeningBrace=*/true); 397 398 if (!FormatTok->Tok.is(tok::r_brace)) { 399 Line->Level = InitialLevel; 400 StructuralError = true; 401 return; 402 } 403 404 nextToken(); // Munch the closing brace. 405 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 406 nextToken(); 407 Line->Level = InitialLevel; 408 } 409 410 void UnwrappedLineParser::parseChildBlock() { 411 FormatTok->BlockKind = BK_Block; 412 nextToken(); 413 { 414 ScopedLineState LineState(*this); 415 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 416 /*MustBeDeclaration=*/false); 417 Line->Level += 1; 418 parseLevel(/*HasOpeningBrace=*/true); 419 Line->Level -= 1; 420 } 421 nextToken(); 422 } 423 424 void UnwrappedLineParser::parsePPDirective() { 425 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 426 ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError); 427 nextToken(); 428 429 if (FormatTok->Tok.getIdentifierInfo() == NULL) { 430 parsePPUnknown(); 431 return; 432 } 433 434 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 435 case tok::pp_define: 436 parsePPDefine(); 437 return; 438 case tok::pp_if: 439 parsePPIf(/*IfDef=*/false); 440 break; 441 case tok::pp_ifdef: 442 case tok::pp_ifndef: 443 parsePPIf(/*IfDef=*/true); 444 break; 445 case tok::pp_else: 446 parsePPElse(); 447 break; 448 case tok::pp_elif: 449 parsePPElIf(); 450 break; 451 case tok::pp_endif: 452 parsePPEndIf(); 453 break; 454 default: 455 parsePPUnknown(); 456 break; 457 } 458 } 459 460 void UnwrappedLineParser::pushPPConditional() { 461 if (!PPStack.empty() && PPStack.back() == PP_Unreachable) 462 PPStack.push_back(PP_Unreachable); 463 else 464 PPStack.push_back(PP_Conditional); 465 } 466 467 void UnwrappedLineParser::parsePPIf(bool IfDef) { 468 ++PPBranchLevel; 469 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 470 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 471 PPLevelBranchIndex.push_back(0); 472 PPLevelBranchCount.push_back(0); 473 } 474 PPChainBranchIndex.push(0); 475 nextToken(); 476 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 477 StringRef(FormatTok->Tok.getLiteralData(), 478 FormatTok->Tok.getLength()) == "0") || 479 FormatTok->Tok.is(tok::kw_false); 480 if ((!IfDef && IsLiteralFalse) || PPLevelBranchIndex[PPBranchLevel] > 0) { 481 PPStack.push_back(PP_Unreachable); 482 } else { 483 pushPPConditional(); 484 } 485 parsePPUnknown(); 486 } 487 488 void UnwrappedLineParser::parsePPElse() { 489 if (!PPStack.empty()) 490 PPStack.pop_back(); 491 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 492 if (!PPChainBranchIndex.empty()) 493 ++PPChainBranchIndex.top(); 494 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 495 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()) { 496 PPStack.push_back(PP_Unreachable); 497 } else { 498 pushPPConditional(); 499 } 500 parsePPUnknown(); 501 } 502 503 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 504 505 void UnwrappedLineParser::parsePPEndIf() { 506 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 507 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 508 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 509 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 510 } 511 } 512 --PPBranchLevel; 513 if (!PPChainBranchIndex.empty()) 514 PPChainBranchIndex.pop(); 515 if (!PPStack.empty()) 516 PPStack.pop_back(); 517 parsePPUnknown(); 518 } 519 520 void UnwrappedLineParser::parsePPDefine() { 521 nextToken(); 522 523 if (FormatTok->Tok.getKind() != tok::identifier) { 524 parsePPUnknown(); 525 return; 526 } 527 nextToken(); 528 if (FormatTok->Tok.getKind() == tok::l_paren && 529 FormatTok->WhitespaceRange.getBegin() == 530 FormatTok->WhitespaceRange.getEnd()) { 531 parseParens(); 532 } 533 addUnwrappedLine(); 534 Line->Level = 1; 535 536 // Errors during a preprocessor directive can only affect the layout of the 537 // preprocessor directive, and thus we ignore them. An alternative approach 538 // would be to use the same approach we use on the file level (no 539 // re-indentation if there was a structural error) within the macro 540 // definition. 541 parseFile(); 542 } 543 544 void UnwrappedLineParser::parsePPUnknown() { 545 do { 546 nextToken(); 547 } while (!eof()); 548 addUnwrappedLine(); 549 } 550 551 // Here we blacklist certain tokens that are not usually the first token in an 552 // unwrapped line. This is used in attempt to distinguish macro calls without 553 // trailing semicolons from other constructs split to several lines. 554 bool tokenCanStartNewLine(clang::Token Tok) { 555 // Semicolon can be a null-statement, l_square can be a start of a macro or 556 // a C++11 attribute, but this doesn't seem to be common. 557 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 558 Tok.isNot(tok::l_square) && 559 // Tokens that can only be used as binary operators and a part of 560 // overloaded operator names. 561 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 562 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 563 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 564 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 565 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 566 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 567 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 568 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 569 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 570 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 571 Tok.isNot(tok::lesslessequal) && 572 // Colon is used in labels, base class lists, initializer lists, 573 // range-based for loops, ternary operator, but should never be the 574 // first token in an unwrapped line. 575 Tok.isNot(tok::colon); 576 } 577 578 void UnwrappedLineParser::parseStructuralElement() { 579 assert(!FormatTok->Tok.is(tok::l_brace)); 580 switch (FormatTok->Tok.getKind()) { 581 case tok::at: 582 nextToken(); 583 if (FormatTok->Tok.is(tok::l_brace)) { 584 parseBracedList(); 585 break; 586 } 587 switch (FormatTok->Tok.getObjCKeywordID()) { 588 case tok::objc_public: 589 case tok::objc_protected: 590 case tok::objc_package: 591 case tok::objc_private: 592 return parseAccessSpecifier(); 593 case tok::objc_interface: 594 case tok::objc_implementation: 595 return parseObjCInterfaceOrImplementation(); 596 case tok::objc_protocol: 597 return parseObjCProtocol(); 598 case tok::objc_end: 599 return; // Handled by the caller. 600 case tok::objc_optional: 601 case tok::objc_required: 602 nextToken(); 603 addUnwrappedLine(); 604 return; 605 default: 606 break; 607 } 608 break; 609 case tok::kw_namespace: 610 parseNamespace(); 611 return; 612 case tok::kw_inline: 613 nextToken(); 614 if (FormatTok->Tok.is(tok::kw_namespace)) { 615 parseNamespace(); 616 return; 617 } 618 break; 619 case tok::kw_public: 620 case tok::kw_protected: 621 case tok::kw_private: 622 parseAccessSpecifier(); 623 return; 624 case tok::kw_if: 625 parseIfThenElse(); 626 return; 627 case tok::kw_for: 628 case tok::kw_while: 629 parseForOrWhileLoop(); 630 return; 631 case tok::kw_do: 632 parseDoWhile(); 633 return; 634 case tok::kw_switch: 635 parseSwitch(); 636 return; 637 case tok::kw_default: 638 nextToken(); 639 parseLabel(); 640 return; 641 case tok::kw_case: 642 parseCaseLabel(); 643 return; 644 case tok::kw_return: 645 parseReturn(); 646 return; 647 case tok::kw_extern: 648 nextToken(); 649 if (FormatTok->Tok.is(tok::string_literal)) { 650 nextToken(); 651 if (FormatTok->Tok.is(tok::l_brace)) { 652 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 653 addUnwrappedLine(); 654 return; 655 } 656 } 657 // In all other cases, parse the declaration. 658 break; 659 default: 660 break; 661 } 662 do { 663 switch (FormatTok->Tok.getKind()) { 664 case tok::at: 665 nextToken(); 666 if (FormatTok->Tok.is(tok::l_brace)) 667 parseBracedList(); 668 break; 669 case tok::kw_enum: 670 parseEnum(); 671 break; 672 case tok::kw_struct: 673 case tok::kw_union: 674 case tok::kw_class: 675 parseRecord(); 676 // A record declaration or definition is always the start of a structural 677 // element. 678 break; 679 case tok::semi: 680 nextToken(); 681 addUnwrappedLine(); 682 return; 683 case tok::r_brace: 684 addUnwrappedLine(); 685 return; 686 case tok::l_paren: 687 parseParens(); 688 break; 689 case tok::caret: 690 nextToken(); 691 if (FormatTok->is(tok::l_brace)) { 692 parseChildBlock(); 693 } 694 break; 695 case tok::l_brace: 696 if (!tryToParseBracedList()) { 697 // A block outside of parentheses must be the last part of a 698 // structural element. 699 // FIXME: Figure out cases where this is not true, and add projections 700 // for them (the one we know is missing are lambdas). 701 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 702 addUnwrappedLine(); 703 FormatTok->Type = TT_FunctionLBrace; 704 parseBlock(/*MustBeDeclaration=*/false); 705 addUnwrappedLine(); 706 return; 707 } 708 // Otherwise this was a braced init list, and the structural 709 // element continues. 710 break; 711 case tok::identifier: { 712 StringRef Text = FormatTok->TokenText; 713 nextToken(); 714 if (Line->Tokens.size() == 1) { 715 if (FormatTok->Tok.is(tok::colon)) { 716 parseLabel(); 717 return; 718 } 719 // Recognize function-like macro usages without trailing semicolon. 720 if (FormatTok->Tok.is(tok::l_paren)) { 721 parseParens(); 722 if (FormatTok->NewlinesBefore > 0 && 723 tokenCanStartNewLine(FormatTok->Tok)) { 724 addUnwrappedLine(); 725 return; 726 } 727 } else if (FormatTok->HasUnescapedNewline && Text.size() >= 5 && 728 Text == Text.upper()) { 729 // Recognize free-standing macros like Q_OBJECT. 730 addUnwrappedLine(); 731 return; 732 } 733 } 734 break; 735 } 736 case tok::equal: 737 nextToken(); 738 if (FormatTok->Tok.is(tok::l_brace)) { 739 parseBracedList(); 740 } 741 break; 742 case tok::l_square: 743 parseSquare(); 744 break; 745 default: 746 nextToken(); 747 break; 748 } 749 } while (!eof()); 750 } 751 752 bool UnwrappedLineParser::tryToParseLambda() { 753 // FIXME: This is a dirty way to access the previous token. Find a better 754 // solution. 755 if (!Line->Tokens.empty() && 756 Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator)) { 757 nextToken(); 758 return false; 759 } 760 assert(FormatTok->is(tok::l_square)); 761 FormatToken &LSquare = *FormatTok; 762 if (!tryToParseLambdaIntroducer()) 763 return false; 764 765 while (FormatTok && FormatTok->isNot(tok::l_brace)) { 766 if (FormatTok->isSimpleTypeSpecifier()) { 767 nextToken(); 768 continue; 769 } 770 switch (FormatTok->Tok.getKind()) { 771 case tok::l_brace: 772 break; 773 case tok::l_paren: 774 parseParens(); 775 break; 776 case tok::less: 777 case tok::greater: 778 case tok::identifier: 779 case tok::kw_mutable: 780 case tok::arrow: 781 nextToken(); 782 break; 783 default: 784 return true; 785 } 786 } 787 LSquare.Type = TT_LambdaLSquare; 788 parseChildBlock(); 789 return true; 790 } 791 792 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 793 nextToken(); 794 if (FormatTok->is(tok::equal)) { 795 nextToken(); 796 if (FormatTok->is(tok::r_square)) { 797 nextToken(); 798 return true; 799 } 800 if (FormatTok->isNot(tok::comma)) 801 return false; 802 nextToken(); 803 } else if (FormatTok->is(tok::amp)) { 804 nextToken(); 805 if (FormatTok->is(tok::r_square)) { 806 nextToken(); 807 return true; 808 } 809 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 810 return false; 811 } 812 if (FormatTok->is(tok::comma)) 813 nextToken(); 814 } else if (FormatTok->is(tok::r_square)) { 815 nextToken(); 816 return true; 817 } 818 do { 819 if (FormatTok->is(tok::amp)) 820 nextToken(); 821 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 822 return false; 823 nextToken(); 824 if (FormatTok->is(tok::comma)) { 825 nextToken(); 826 } else if (FormatTok->is(tok::r_square)) { 827 nextToken(); 828 return true; 829 } else { 830 return false; 831 } 832 } while (!eof()); 833 return false; 834 } 835 836 bool UnwrappedLineParser::tryToParseBracedList() { 837 if (FormatTok->BlockKind == BK_Unknown) 838 calculateBraceTypes(); 839 assert(FormatTok->BlockKind != BK_Unknown); 840 if (FormatTok->BlockKind == BK_Block) 841 return false; 842 parseBracedList(); 843 return true; 844 } 845 846 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 847 bool HasError = false; 848 nextToken(); 849 850 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 851 // replace this by using parseAssigmentExpression() inside. 852 do { 853 // FIXME: When we start to support lambdas, we'll want to parse them away 854 // here, otherwise our bail-out scenarios below break. The better solution 855 // might be to just implement a more or less complete expression parser. 856 switch (FormatTok->Tok.getKind()) { 857 case tok::caret: 858 nextToken(); 859 if (FormatTok->is(tok::l_brace)) { 860 parseChildBlock(); 861 } 862 break; 863 case tok::l_square: 864 tryToParseLambda(); 865 break; 866 case tok::l_brace: 867 // Assume there are no blocks inside a braced init list apart 868 // from the ones we explicitly parse out (like lambdas). 869 FormatTok->BlockKind = BK_BracedInit; 870 parseBracedList(); 871 break; 872 case tok::r_brace: 873 nextToken(); 874 return !HasError; 875 case tok::semi: 876 HasError = true; 877 if (!ContinueOnSemicolons) 878 return !HasError; 879 nextToken(); 880 break; 881 case tok::comma: 882 nextToken(); 883 break; 884 default: 885 nextToken(); 886 break; 887 } 888 } while (!eof()); 889 return false; 890 } 891 892 void UnwrappedLineParser::parseReturn() { 893 nextToken(); 894 895 do { 896 switch (FormatTok->Tok.getKind()) { 897 case tok::l_brace: 898 parseBracedList(); 899 if (FormatTok->Tok.isNot(tok::semi)) { 900 // Assume missing ';'. 901 addUnwrappedLine(); 902 return; 903 } 904 break; 905 case tok::l_paren: 906 parseParens(); 907 break; 908 case tok::r_brace: 909 // Assume missing ';'. 910 addUnwrappedLine(); 911 return; 912 case tok::semi: 913 nextToken(); 914 addUnwrappedLine(); 915 return; 916 case tok::l_square: 917 tryToParseLambda(); 918 break; 919 default: 920 nextToken(); 921 break; 922 } 923 } while (!eof()); 924 } 925 926 void UnwrappedLineParser::parseParens() { 927 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 928 nextToken(); 929 do { 930 switch (FormatTok->Tok.getKind()) { 931 case tok::l_paren: 932 parseParens(); 933 break; 934 case tok::r_paren: 935 nextToken(); 936 return; 937 case tok::r_brace: 938 // A "}" inside parenthesis is an error if there wasn't a matching "{". 939 return; 940 case tok::l_square: 941 tryToParseLambda(); 942 break; 943 case tok::l_brace: { 944 if (!tryToParseBracedList()) { 945 parseChildBlock(); 946 } 947 break; 948 } 949 case tok::at: 950 nextToken(); 951 if (FormatTok->Tok.is(tok::l_brace)) 952 parseBracedList(); 953 break; 954 default: 955 nextToken(); 956 break; 957 } 958 } while (!eof()); 959 } 960 961 void UnwrappedLineParser::parseSquare() { 962 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 963 if (tryToParseLambda()) 964 return; 965 do { 966 switch (FormatTok->Tok.getKind()) { 967 case tok::l_paren: 968 parseParens(); 969 break; 970 case tok::r_square: 971 nextToken(); 972 return; 973 case tok::r_brace: 974 // A "}" inside parenthesis is an error if there wasn't a matching "{". 975 return; 976 case tok::l_square: 977 parseSquare(); 978 break; 979 case tok::l_brace: { 980 if (!tryToParseBracedList()) { 981 parseChildBlock(); 982 } 983 break; 984 } 985 case tok::at: 986 nextToken(); 987 if (FormatTok->Tok.is(tok::l_brace)) 988 parseBracedList(); 989 break; 990 default: 991 nextToken(); 992 break; 993 } 994 } while (!eof()); 995 } 996 997 void UnwrappedLineParser::parseIfThenElse() { 998 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 999 nextToken(); 1000 if (FormatTok->Tok.is(tok::l_paren)) 1001 parseParens(); 1002 bool NeedsUnwrappedLine = false; 1003 if (FormatTok->Tok.is(tok::l_brace)) { 1004 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1005 parseBlock(/*MustBeDeclaration=*/false); 1006 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1007 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1008 addUnwrappedLine(); 1009 } else { 1010 NeedsUnwrappedLine = true; 1011 } 1012 } else { 1013 addUnwrappedLine(); 1014 ++Line->Level; 1015 parseStructuralElement(); 1016 --Line->Level; 1017 } 1018 if (FormatTok->Tok.is(tok::kw_else)) { 1019 nextToken(); 1020 if (FormatTok->Tok.is(tok::l_brace)) { 1021 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1022 parseBlock(/*MustBeDeclaration=*/false); 1023 addUnwrappedLine(); 1024 } else if (FormatTok->Tok.is(tok::kw_if)) { 1025 parseIfThenElse(); 1026 } else { 1027 addUnwrappedLine(); 1028 ++Line->Level; 1029 parseStructuralElement(); 1030 --Line->Level; 1031 } 1032 } else if (NeedsUnwrappedLine) { 1033 addUnwrappedLine(); 1034 } 1035 } 1036 1037 void UnwrappedLineParser::parseNamespace() { 1038 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1039 nextToken(); 1040 if (FormatTok->Tok.is(tok::identifier)) 1041 nextToken(); 1042 if (FormatTok->Tok.is(tok::l_brace)) { 1043 if (Style.BreakBeforeBraces == FormatStyle::BS_Linux || 1044 Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1045 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1046 addUnwrappedLine(); 1047 1048 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1049 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1050 DeclarationScopeStack.size() > 1); 1051 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1052 // Munch the semicolon after a namespace. This is more common than one would 1053 // think. Puttin the semicolon into its own line is very ugly. 1054 if (FormatTok->Tok.is(tok::semi)) 1055 nextToken(); 1056 addUnwrappedLine(); 1057 } 1058 // FIXME: Add error handling. 1059 } 1060 1061 void UnwrappedLineParser::parseForOrWhileLoop() { 1062 assert((FormatTok->Tok.is(tok::kw_for) || FormatTok->Tok.is(tok::kw_while)) && 1063 "'for' or 'while' expected"); 1064 nextToken(); 1065 if (FormatTok->Tok.is(tok::l_paren)) 1066 parseParens(); 1067 if (FormatTok->Tok.is(tok::l_brace)) { 1068 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1069 parseBlock(/*MustBeDeclaration=*/false); 1070 addUnwrappedLine(); 1071 } else { 1072 addUnwrappedLine(); 1073 ++Line->Level; 1074 parseStructuralElement(); 1075 --Line->Level; 1076 } 1077 } 1078 1079 void UnwrappedLineParser::parseDoWhile() { 1080 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1081 nextToken(); 1082 if (FormatTok->Tok.is(tok::l_brace)) { 1083 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1084 parseBlock(/*MustBeDeclaration=*/false); 1085 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1086 addUnwrappedLine(); 1087 } else { 1088 addUnwrappedLine(); 1089 ++Line->Level; 1090 parseStructuralElement(); 1091 --Line->Level; 1092 } 1093 1094 // FIXME: Add error handling. 1095 if (!FormatTok->Tok.is(tok::kw_while)) { 1096 addUnwrappedLine(); 1097 return; 1098 } 1099 1100 nextToken(); 1101 parseStructuralElement(); 1102 } 1103 1104 void UnwrappedLineParser::parseLabel() { 1105 nextToken(); 1106 unsigned OldLineLevel = Line->Level; 1107 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1108 --Line->Level; 1109 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1110 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1111 parseBlock(/*MustBeDeclaration=*/false); 1112 if (FormatTok->Tok.is(tok::kw_break)) { 1113 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1114 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1115 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1116 addUnwrappedLine(); 1117 } 1118 parseStructuralElement(); 1119 } 1120 addUnwrappedLine(); 1121 } else { 1122 addUnwrappedLine(); 1123 } 1124 Line->Level = OldLineLevel; 1125 } 1126 1127 void UnwrappedLineParser::parseCaseLabel() { 1128 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1129 // FIXME: fix handling of complex expressions here. 1130 do { 1131 nextToken(); 1132 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1133 parseLabel(); 1134 } 1135 1136 void UnwrappedLineParser::parseSwitch() { 1137 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1138 nextToken(); 1139 if (FormatTok->Tok.is(tok::l_paren)) 1140 parseParens(); 1141 if (FormatTok->Tok.is(tok::l_brace)) { 1142 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1143 parseBlock(/*MustBeDeclaration=*/false); 1144 addUnwrappedLine(); 1145 } else { 1146 addUnwrappedLine(); 1147 ++Line->Level; 1148 parseStructuralElement(); 1149 --Line->Level; 1150 } 1151 } 1152 1153 void UnwrappedLineParser::parseAccessSpecifier() { 1154 nextToken(); 1155 // Understand Qt's slots. 1156 if (FormatTok->is(tok::identifier) && 1157 (FormatTok->TokenText == "slots" || FormatTok->TokenText == "Q_SLOTS")) 1158 nextToken(); 1159 // Otherwise, we don't know what it is, and we'd better keep the next token. 1160 if (FormatTok->Tok.is(tok::colon)) 1161 nextToken(); 1162 addUnwrappedLine(); 1163 } 1164 1165 void UnwrappedLineParser::parseEnum() { 1166 nextToken(); 1167 // Eat up enum class ... 1168 if (FormatTok->Tok.is(tok::kw_class) || 1169 FormatTok->Tok.is(tok::kw_struct)) 1170 nextToken(); 1171 while (FormatTok->Tok.getIdentifierInfo() || 1172 FormatTok->isOneOf(tok::colon, tok::coloncolon)) { 1173 nextToken(); 1174 // We can have macros or attributes in between 'enum' and the enum name. 1175 if (FormatTok->Tok.is(tok::l_paren)) { 1176 parseParens(); 1177 } 1178 if (FormatTok->Tok.is(tok::identifier)) 1179 nextToken(); 1180 } 1181 if (FormatTok->Tok.is(tok::l_brace)) { 1182 FormatTok->BlockKind = BK_Block; 1183 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1184 if (HasError) { 1185 if (FormatTok->is(tok::semi)) 1186 nextToken(); 1187 addUnwrappedLine(); 1188 } 1189 } 1190 // We fall through to parsing a structural element afterwards, so that in 1191 // enum A {} n, m; 1192 // "} n, m;" will end up in one unwrapped line. 1193 } 1194 1195 void UnwrappedLineParser::parseRecord() { 1196 nextToken(); 1197 if (FormatTok->Tok.is(tok::identifier) || 1198 FormatTok->Tok.is(tok::kw___attribute) || 1199 FormatTok->Tok.is(tok::kw___declspec) || 1200 FormatTok->Tok.is(tok::kw_alignas)) { 1201 nextToken(); 1202 // We can have macros or attributes in between 'class' and the class name. 1203 if (FormatTok->Tok.is(tok::l_paren)) { 1204 parseParens(); 1205 } 1206 // The actual identifier can be a nested name specifier, and in macros 1207 // it is often token-pasted. 1208 while (FormatTok->Tok.is(tok::identifier) || 1209 FormatTok->Tok.is(tok::coloncolon) || 1210 FormatTok->Tok.is(tok::hashhash)) 1211 nextToken(); 1212 1213 // Note that parsing away template declarations here leads to incorrectly 1214 // accepting function declarations as record declarations. 1215 // In general, we cannot solve this problem. Consider: 1216 // class A<int> B() {} 1217 // which can be a function definition or a class definition when B() is a 1218 // macro. If we find enough real-world cases where this is a problem, we 1219 // can parse for the 'template' keyword in the beginning of the statement, 1220 // and thus rule out the record production in case there is no template 1221 // (this would still leave us with an ambiguity between template function 1222 // and class declarations). 1223 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1224 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1225 if (FormatTok->Tok.is(tok::semi)) 1226 return; 1227 nextToken(); 1228 } 1229 } 1230 } 1231 if (FormatTok->Tok.is(tok::l_brace)) { 1232 if (Style.BreakBeforeBraces == FormatStyle::BS_Linux || 1233 Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1234 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1235 addUnwrappedLine(); 1236 1237 parseBlock(/*MustBeDeclaration=*/true, /*Addlevel=*/true, 1238 /*MunchSemi=*/false); 1239 } 1240 // We fall through to parsing a structural element afterwards, so 1241 // class A {} n, m; 1242 // will end up in one unwrapped line. 1243 } 1244 1245 void UnwrappedLineParser::parseObjCProtocolList() { 1246 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1247 do 1248 nextToken(); 1249 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1250 nextToken(); // Skip '>'. 1251 } 1252 1253 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1254 do { 1255 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1256 nextToken(); 1257 addUnwrappedLine(); 1258 break; 1259 } 1260 if (FormatTok->is(tok::l_brace)) { 1261 parseBlock(/*MustBeDeclaration=*/false); 1262 // In ObjC interfaces, nothing should be following the "}". 1263 addUnwrappedLine(); 1264 } else if (FormatTok->is(tok::r_brace)) { 1265 // Ignore stray "}". parseStructuralElement doesn't consume them. 1266 nextToken(); 1267 addUnwrappedLine(); 1268 } else { 1269 parseStructuralElement(); 1270 } 1271 } while (!eof()); 1272 } 1273 1274 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1275 nextToken(); 1276 nextToken(); // interface name 1277 1278 // @interface can be followed by either a base class, or a category. 1279 if (FormatTok->Tok.is(tok::colon)) { 1280 nextToken(); 1281 nextToken(); // base class name 1282 } else if (FormatTok->Tok.is(tok::l_paren)) 1283 // Skip category, if present. 1284 parseParens(); 1285 1286 if (FormatTok->Tok.is(tok::less)) 1287 parseObjCProtocolList(); 1288 1289 // If instance variables are present, keep the '{' on the first line too. 1290 if (FormatTok->Tok.is(tok::l_brace)) 1291 parseBlock(/*MustBeDeclaration=*/true); 1292 1293 // With instance variables, this puts '}' on its own line. Without instance 1294 // variables, this ends the @interface line. 1295 addUnwrappedLine(); 1296 1297 parseObjCUntilAtEnd(); 1298 } 1299 1300 void UnwrappedLineParser::parseObjCProtocol() { 1301 nextToken(); 1302 nextToken(); // protocol name 1303 1304 if (FormatTok->Tok.is(tok::less)) 1305 parseObjCProtocolList(); 1306 1307 // Check for protocol declaration. 1308 if (FormatTok->Tok.is(tok::semi)) { 1309 nextToken(); 1310 return addUnwrappedLine(); 1311 } 1312 1313 addUnwrappedLine(); 1314 parseObjCUntilAtEnd(); 1315 } 1316 1317 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1318 StringRef Prefix = "") { 1319 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1320 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1321 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1322 E = Line.Tokens.end(); 1323 I != E; ++I) { 1324 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1325 } 1326 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1327 E = Line.Tokens.end(); 1328 I != E; ++I) { 1329 const UnwrappedLineNode &Node = *I; 1330 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1331 I = Node.Children.begin(), 1332 E = Node.Children.end(); 1333 I != E; ++I) { 1334 printDebugInfo(*I, "\nChild: "); 1335 } 1336 } 1337 llvm::dbgs() << "\n"; 1338 } 1339 1340 void UnwrappedLineParser::addUnwrappedLine() { 1341 if (Line->Tokens.empty()) 1342 return; 1343 DEBUG({ 1344 if (CurrentLines == &Lines) 1345 printDebugInfo(*Line); 1346 }); 1347 CurrentLines->push_back(*Line); 1348 Line->Tokens.clear(); 1349 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1350 for (SmallVectorImpl<UnwrappedLine>::iterator 1351 I = PreprocessorDirectives.begin(), 1352 E = PreprocessorDirectives.end(); 1353 I != E; ++I) { 1354 CurrentLines->push_back(*I); 1355 } 1356 PreprocessorDirectives.clear(); 1357 } 1358 } 1359 1360 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1361 1362 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1363 bool JustComments = Line->Tokens.empty(); 1364 for (SmallVectorImpl<FormatToken *>::const_iterator 1365 I = CommentsBeforeNextToken.begin(), 1366 E = CommentsBeforeNextToken.end(); 1367 I != E; ++I) { 1368 if ((*I)->NewlinesBefore && JustComments) { 1369 addUnwrappedLine(); 1370 } 1371 pushToken(*I); 1372 } 1373 if (NewlineBeforeNext && JustComments) { 1374 addUnwrappedLine(); 1375 } 1376 CommentsBeforeNextToken.clear(); 1377 } 1378 1379 void UnwrappedLineParser::nextToken() { 1380 if (eof()) 1381 return; 1382 flushComments(FormatTok->NewlinesBefore > 0); 1383 pushToken(FormatTok); 1384 readToken(); 1385 } 1386 1387 void UnwrappedLineParser::readToken() { 1388 bool CommentsInCurrentLine = true; 1389 do { 1390 FormatTok = Tokens->getNextToken(); 1391 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1392 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1393 // If there is an unfinished unwrapped line, we flush the preprocessor 1394 // directives only after that unwrapped line was finished later. 1395 bool SwitchToPreprocessorLines = 1396 !Line->Tokens.empty() && CurrentLines == &Lines; 1397 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1398 // Comments stored before the preprocessor directive need to be output 1399 // before the preprocessor directive, at the same level as the 1400 // preprocessor directive, as we consider them to apply to the directive. 1401 flushComments(FormatTok->NewlinesBefore > 0); 1402 parsePPDirective(); 1403 } 1404 1405 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1406 !Line->InPPDirective) { 1407 continue; 1408 } 1409 1410 if (!FormatTok->Tok.is(tok::comment)) 1411 return; 1412 if (FormatTok->NewlinesBefore > 0 || FormatTok->IsFirst) { 1413 CommentsInCurrentLine = false; 1414 } 1415 if (CommentsInCurrentLine) { 1416 pushToken(FormatTok); 1417 } else { 1418 CommentsBeforeNextToken.push_back(FormatTok); 1419 } 1420 } while (!eof()); 1421 } 1422 1423 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1424 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1425 if (MustBreakBeforeNextToken) { 1426 Line->Tokens.back().Tok->MustBreakBefore = true; 1427 MustBreakBeforeNextToken = false; 1428 } 1429 } 1430 1431 } // end namespace format 1432 } // end namespace clang 1433