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 FormatToken *getNextToken() override { 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 unsigned getPosition() override { return PreviousTokenSource->getPosition(); } 88 89 FormatToken *setPosition(unsigned Position) override { 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.release(); 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 FormatToken *getNextToken() override { 184 ++Position; 185 return Tokens[Position]; 186 } 187 188 unsigned getPosition() override { 189 assert(Position >= 0); 190 return Position; 191 } 192 193 FormatToken *setPosition(unsigned P) override { 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::at: 361 case tok::semi: 362 case tok::kw_if: 363 case tok::kw_while: 364 case tok::kw_for: 365 case tok::kw_switch: 366 case tok::kw_try: 367 if (!LBraceStack.empty()) 368 LBraceStack.back()->BlockKind = BK_Block; 369 break; 370 default: 371 break; 372 } 373 Tok = NextTok; 374 Position += ReadTokens; 375 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 376 // Assume other blocks for all unclosed opening braces. 377 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 378 if (LBraceStack[i]->BlockKind == BK_Unknown) 379 LBraceStack[i]->BlockKind = BK_Block; 380 } 381 382 FormatTok = Tokens->setPosition(StoredPosition); 383 } 384 385 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 386 bool MunchSemi) { 387 assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected"); 388 unsigned InitialLevel = Line->Level; 389 nextToken(); 390 391 addUnwrappedLine(); 392 393 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 394 MustBeDeclaration); 395 if (AddLevel) 396 ++Line->Level; 397 parseLevel(/*HasOpeningBrace=*/true); 398 399 if (!FormatTok->Tok.is(tok::r_brace)) { 400 Line->Level = InitialLevel; 401 StructuralError = true; 402 return; 403 } 404 405 nextToken(); // Munch the closing brace. 406 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 407 nextToken(); 408 Line->Level = InitialLevel; 409 } 410 411 void UnwrappedLineParser::parseChildBlock() { 412 FormatTok->BlockKind = BK_Block; 413 nextToken(); 414 { 415 ScopedLineState LineState(*this); 416 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 417 /*MustBeDeclaration=*/false); 418 Line->Level += 1; 419 parseLevel(/*HasOpeningBrace=*/true); 420 Line->Level -= 1; 421 } 422 nextToken(); 423 } 424 425 void UnwrappedLineParser::parsePPDirective() { 426 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 427 ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError); 428 nextToken(); 429 430 if (FormatTok->Tok.getIdentifierInfo() == NULL) { 431 parsePPUnknown(); 432 return; 433 } 434 435 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 436 case tok::pp_define: 437 parsePPDefine(); 438 return; 439 case tok::pp_if: 440 parsePPIf(/*IfDef=*/false); 441 break; 442 case tok::pp_ifdef: 443 case tok::pp_ifndef: 444 parsePPIf(/*IfDef=*/true); 445 break; 446 case tok::pp_else: 447 parsePPElse(); 448 break; 449 case tok::pp_elif: 450 parsePPElIf(); 451 break; 452 case tok::pp_endif: 453 parsePPEndIf(); 454 break; 455 default: 456 parsePPUnknown(); 457 break; 458 } 459 } 460 461 void UnwrappedLineParser::pushPPConditional() { 462 if (!PPStack.empty() && PPStack.back() == PP_Unreachable) 463 PPStack.push_back(PP_Unreachable); 464 else 465 PPStack.push_back(PP_Conditional); 466 } 467 468 void UnwrappedLineParser::parsePPIf(bool IfDef) { 469 ++PPBranchLevel; 470 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 471 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 472 PPLevelBranchIndex.push_back(0); 473 PPLevelBranchCount.push_back(0); 474 } 475 PPChainBranchIndex.push(0); 476 nextToken(); 477 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 478 StringRef(FormatTok->Tok.getLiteralData(), 479 FormatTok->Tok.getLength()) == "0") || 480 FormatTok->Tok.is(tok::kw_false); 481 if ((!IfDef && IsLiteralFalse) || PPLevelBranchIndex[PPBranchLevel] > 0) { 482 PPStack.push_back(PP_Unreachable); 483 } else { 484 pushPPConditional(); 485 } 486 parsePPUnknown(); 487 } 488 489 void UnwrappedLineParser::parsePPElse() { 490 if (!PPStack.empty()) 491 PPStack.pop_back(); 492 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 493 if (!PPChainBranchIndex.empty()) 494 ++PPChainBranchIndex.top(); 495 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 496 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()) { 497 PPStack.push_back(PP_Unreachable); 498 } else { 499 pushPPConditional(); 500 } 501 parsePPUnknown(); 502 } 503 504 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 505 506 void UnwrappedLineParser::parsePPEndIf() { 507 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 508 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 509 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 510 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 511 } 512 } 513 // Guard against #endif's without #if. 514 if (PPBranchLevel > 0) 515 --PPBranchLevel; 516 if (!PPChainBranchIndex.empty()) 517 PPChainBranchIndex.pop(); 518 if (!PPStack.empty()) 519 PPStack.pop_back(); 520 parsePPUnknown(); 521 } 522 523 void UnwrappedLineParser::parsePPDefine() { 524 nextToken(); 525 526 if (FormatTok->Tok.getKind() != tok::identifier) { 527 parsePPUnknown(); 528 return; 529 } 530 nextToken(); 531 if (FormatTok->Tok.getKind() == tok::l_paren && 532 FormatTok->WhitespaceRange.getBegin() == 533 FormatTok->WhitespaceRange.getEnd()) { 534 parseParens(); 535 } 536 addUnwrappedLine(); 537 Line->Level = 1; 538 539 // Errors during a preprocessor directive can only affect the layout of the 540 // preprocessor directive, and thus we ignore them. An alternative approach 541 // would be to use the same approach we use on the file level (no 542 // re-indentation if there was a structural error) within the macro 543 // definition. 544 parseFile(); 545 } 546 547 void UnwrappedLineParser::parsePPUnknown() { 548 do { 549 nextToken(); 550 } while (!eof()); 551 addUnwrappedLine(); 552 } 553 554 // Here we blacklist certain tokens that are not usually the first token in an 555 // unwrapped line. This is used in attempt to distinguish macro calls without 556 // trailing semicolons from other constructs split to several lines. 557 bool tokenCanStartNewLine(clang::Token Tok) { 558 // Semicolon can be a null-statement, l_square can be a start of a macro or 559 // a C++11 attribute, but this doesn't seem to be common. 560 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 561 Tok.isNot(tok::l_square) && 562 // Tokens that can only be used as binary operators and a part of 563 // overloaded operator names. 564 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 565 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 566 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 567 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 568 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 569 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 570 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 571 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 572 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 573 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 574 Tok.isNot(tok::lesslessequal) && 575 // Colon is used in labels, base class lists, initializer lists, 576 // range-based for loops, ternary operator, but should never be the 577 // first token in an unwrapped line. 578 Tok.isNot(tok::colon); 579 } 580 581 void UnwrappedLineParser::parseStructuralElement() { 582 assert(!FormatTok->Tok.is(tok::l_brace)); 583 switch (FormatTok->Tok.getKind()) { 584 case tok::at: 585 nextToken(); 586 if (FormatTok->Tok.is(tok::l_brace)) { 587 parseBracedList(); 588 break; 589 } 590 switch (FormatTok->Tok.getObjCKeywordID()) { 591 case tok::objc_public: 592 case tok::objc_protected: 593 case tok::objc_package: 594 case tok::objc_private: 595 return parseAccessSpecifier(); 596 case tok::objc_interface: 597 case tok::objc_implementation: 598 return parseObjCInterfaceOrImplementation(); 599 case tok::objc_protocol: 600 return parseObjCProtocol(); 601 case tok::objc_end: 602 return; // Handled by the caller. 603 case tok::objc_optional: 604 case tok::objc_required: 605 nextToken(); 606 addUnwrappedLine(); 607 return; 608 default: 609 break; 610 } 611 break; 612 case tok::kw_namespace: 613 parseNamespace(); 614 return; 615 case tok::kw_inline: 616 nextToken(); 617 if (FormatTok->Tok.is(tok::kw_namespace)) { 618 parseNamespace(); 619 return; 620 } 621 break; 622 case tok::kw_public: 623 case tok::kw_protected: 624 case tok::kw_private: 625 parseAccessSpecifier(); 626 return; 627 case tok::kw_if: 628 parseIfThenElse(); 629 return; 630 case tok::kw_for: 631 case tok::kw_while: 632 parseForOrWhileLoop(); 633 return; 634 case tok::kw_do: 635 parseDoWhile(); 636 return; 637 case tok::kw_switch: 638 parseSwitch(); 639 return; 640 case tok::kw_default: 641 nextToken(); 642 parseLabel(); 643 return; 644 case tok::kw_case: 645 parseCaseLabel(); 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 break; 658 case tok::identifier: 659 if (FormatTok->IsForEachMacro) { 660 parseForOrWhileLoop(); 661 return; 662 } 663 // In all other cases, parse the declaration. 664 break; 665 default: 666 break; 667 } 668 do { 669 switch (FormatTok->Tok.getKind()) { 670 case tok::at: 671 nextToken(); 672 if (FormatTok->Tok.is(tok::l_brace)) 673 parseBracedList(); 674 break; 675 case tok::kw_enum: 676 parseEnum(); 677 break; 678 case tok::kw_typedef: 679 nextToken(); 680 // FIXME: Use the IdentifierTable instead. 681 if (FormatTok->TokenText == "NS_ENUM") 682 parseEnum(); 683 break; 684 case tok::kw_struct: 685 case tok::kw_union: 686 case tok::kw_class: 687 parseRecord(); 688 // A record declaration or definition is always the start of a structural 689 // element. 690 break; 691 case tok::semi: 692 nextToken(); 693 addUnwrappedLine(); 694 return; 695 case tok::r_brace: 696 addUnwrappedLine(); 697 return; 698 case tok::l_paren: 699 parseParens(); 700 break; 701 case tok::caret: 702 nextToken(); 703 if (FormatTok->Tok.isAnyIdentifier() || 704 FormatTok->isSimpleTypeSpecifier()) 705 nextToken(); 706 if (FormatTok->is(tok::l_paren)) 707 parseParens(); 708 if (FormatTok->is(tok::l_brace)) 709 parseChildBlock(); 710 break; 711 case tok::l_brace: 712 if (!tryToParseBracedList()) { 713 // A block outside of parentheses must be the last part of a 714 // structural element. 715 // FIXME: Figure out cases where this is not true, and add projections 716 // for them (the one we know is missing are lambdas). 717 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 718 addUnwrappedLine(); 719 FormatTok->Type = TT_FunctionLBrace; 720 parseBlock(/*MustBeDeclaration=*/false); 721 addUnwrappedLine(); 722 return; 723 } 724 // Otherwise this was a braced init list, and the structural 725 // element continues. 726 break; 727 case tok::identifier: { 728 StringRef Text = FormatTok->TokenText; 729 nextToken(); 730 if (Line->Tokens.size() == 1) { 731 if (FormatTok->Tok.is(tok::colon)) { 732 parseLabel(); 733 return; 734 } 735 // Recognize function-like macro usages without trailing semicolon. 736 if (FormatTok->Tok.is(tok::l_paren)) { 737 parseParens(); 738 if (FormatTok->NewlinesBefore > 0 && 739 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 740 addUnwrappedLine(); 741 return; 742 } 743 } else if (FormatTok->HasUnescapedNewline && Text.size() >= 5 && 744 Text == Text.upper()) { 745 // Recognize free-standing macros like Q_OBJECT. 746 addUnwrappedLine(); 747 return; 748 } 749 } 750 break; 751 } 752 case tok::equal: 753 nextToken(); 754 if (FormatTok->Tok.is(tok::l_brace)) { 755 parseBracedList(); 756 } 757 break; 758 case tok::l_square: 759 parseSquare(); 760 break; 761 default: 762 nextToken(); 763 break; 764 } 765 } while (!eof()); 766 } 767 768 bool UnwrappedLineParser::tryToParseLambda() { 769 // FIXME: This is a dirty way to access the previous token. Find a better 770 // solution. 771 if (!Line->Tokens.empty() && 772 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator) || 773 Line->Tokens.back().Tok->closesScope() || 774 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 775 nextToken(); 776 return false; 777 } 778 assert(FormatTok->is(tok::l_square)); 779 FormatToken &LSquare = *FormatTok; 780 if (!tryToParseLambdaIntroducer()) 781 return false; 782 783 while (FormatTok->isNot(tok::l_brace)) { 784 if (FormatTok->isSimpleTypeSpecifier()) { 785 nextToken(); 786 continue; 787 } 788 switch (FormatTok->Tok.getKind()) { 789 case tok::l_brace: 790 break; 791 case tok::l_paren: 792 parseParens(); 793 break; 794 case tok::less: 795 case tok::greater: 796 case tok::identifier: 797 case tok::coloncolon: 798 case tok::kw_mutable: 799 nextToken(); 800 break; 801 case tok::arrow: 802 FormatTok->Type = TT_TrailingReturnArrow; 803 nextToken(); 804 break; 805 default: 806 return true; 807 } 808 } 809 LSquare.Type = TT_LambdaLSquare; 810 parseChildBlock(); 811 return true; 812 } 813 814 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 815 nextToken(); 816 if (FormatTok->is(tok::equal)) { 817 nextToken(); 818 if (FormatTok->is(tok::r_square)) { 819 nextToken(); 820 return true; 821 } 822 if (FormatTok->isNot(tok::comma)) 823 return false; 824 nextToken(); 825 } else if (FormatTok->is(tok::amp)) { 826 nextToken(); 827 if (FormatTok->is(tok::r_square)) { 828 nextToken(); 829 return true; 830 } 831 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 832 return false; 833 } 834 if (FormatTok->is(tok::comma)) 835 nextToken(); 836 } else if (FormatTok->is(tok::r_square)) { 837 nextToken(); 838 return true; 839 } 840 do { 841 if (FormatTok->is(tok::amp)) 842 nextToken(); 843 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 844 return false; 845 nextToken(); 846 if (FormatTok->is(tok::comma)) { 847 nextToken(); 848 } else if (FormatTok->is(tok::r_square)) { 849 nextToken(); 850 return true; 851 } else { 852 return false; 853 } 854 } while (!eof()); 855 return false; 856 } 857 858 bool UnwrappedLineParser::tryToParseBracedList() { 859 if (FormatTok->BlockKind == BK_Unknown) 860 calculateBraceTypes(); 861 assert(FormatTok->BlockKind != BK_Unknown); 862 if (FormatTok->BlockKind == BK_Block) 863 return false; 864 parseBracedList(); 865 return true; 866 } 867 868 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 869 bool HasError = false; 870 nextToken(); 871 872 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 873 // replace this by using parseAssigmentExpression() inside. 874 do { 875 // FIXME: When we start to support lambdas, we'll want to parse them away 876 // here, otherwise our bail-out scenarios below break. The better solution 877 // might be to just implement a more or less complete expression parser. 878 switch (FormatTok->Tok.getKind()) { 879 case tok::caret: 880 nextToken(); 881 if (FormatTok->is(tok::l_brace)) { 882 parseChildBlock(); 883 } 884 break; 885 case tok::l_square: 886 tryToParseLambda(); 887 break; 888 case tok::l_brace: 889 // Assume there are no blocks inside a braced init list apart 890 // from the ones we explicitly parse out (like lambdas). 891 FormatTok->BlockKind = BK_BracedInit; 892 parseBracedList(); 893 break; 894 case tok::r_brace: 895 nextToken(); 896 return !HasError; 897 case tok::semi: 898 HasError = true; 899 if (!ContinueOnSemicolons) 900 return !HasError; 901 nextToken(); 902 break; 903 case tok::comma: 904 nextToken(); 905 break; 906 default: 907 nextToken(); 908 break; 909 } 910 } while (!eof()); 911 return false; 912 } 913 914 void UnwrappedLineParser::parseParens() { 915 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 916 nextToken(); 917 do { 918 switch (FormatTok->Tok.getKind()) { 919 case tok::l_paren: 920 parseParens(); 921 break; 922 case tok::r_paren: 923 nextToken(); 924 return; 925 case tok::r_brace: 926 // A "}" inside parenthesis is an error if there wasn't a matching "{". 927 return; 928 case tok::l_square: 929 tryToParseLambda(); 930 break; 931 case tok::l_brace: { 932 if (!tryToParseBracedList()) { 933 parseChildBlock(); 934 } 935 break; 936 } 937 case tok::at: 938 nextToken(); 939 if (FormatTok->Tok.is(tok::l_brace)) 940 parseBracedList(); 941 break; 942 default: 943 nextToken(); 944 break; 945 } 946 } while (!eof()); 947 } 948 949 void UnwrappedLineParser::parseSquare() { 950 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 951 if (tryToParseLambda()) 952 return; 953 do { 954 switch (FormatTok->Tok.getKind()) { 955 case tok::l_paren: 956 parseParens(); 957 break; 958 case tok::r_square: 959 nextToken(); 960 return; 961 case tok::r_brace: 962 // A "}" inside parenthesis is an error if there wasn't a matching "{". 963 return; 964 case tok::l_square: 965 parseSquare(); 966 break; 967 case tok::l_brace: { 968 if (!tryToParseBracedList()) { 969 parseChildBlock(); 970 } 971 break; 972 } 973 case tok::at: 974 nextToken(); 975 if (FormatTok->Tok.is(tok::l_brace)) 976 parseBracedList(); 977 break; 978 default: 979 nextToken(); 980 break; 981 } 982 } while (!eof()); 983 } 984 985 void UnwrappedLineParser::parseIfThenElse() { 986 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 987 nextToken(); 988 if (FormatTok->Tok.is(tok::l_paren)) 989 parseParens(); 990 bool NeedsUnwrappedLine = false; 991 if (FormatTok->Tok.is(tok::l_brace)) { 992 CompoundStatementIndenter Indenter(this, Style, Line->Level); 993 parseBlock(/*MustBeDeclaration=*/false); 994 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 995 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 996 addUnwrappedLine(); 997 } else { 998 NeedsUnwrappedLine = true; 999 } 1000 } else { 1001 addUnwrappedLine(); 1002 ++Line->Level; 1003 parseStructuralElement(); 1004 --Line->Level; 1005 } 1006 if (FormatTok->Tok.is(tok::kw_else)) { 1007 nextToken(); 1008 if (FormatTok->Tok.is(tok::l_brace)) { 1009 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1010 parseBlock(/*MustBeDeclaration=*/false); 1011 addUnwrappedLine(); 1012 } else if (FormatTok->Tok.is(tok::kw_if)) { 1013 parseIfThenElse(); 1014 } else { 1015 addUnwrappedLine(); 1016 ++Line->Level; 1017 parseStructuralElement(); 1018 --Line->Level; 1019 } 1020 } else if (NeedsUnwrappedLine) { 1021 addUnwrappedLine(); 1022 } 1023 } 1024 1025 void UnwrappedLineParser::parseNamespace() { 1026 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1027 nextToken(); 1028 if (FormatTok->Tok.is(tok::identifier)) 1029 nextToken(); 1030 if (FormatTok->Tok.is(tok::l_brace)) { 1031 if (Style.BreakBeforeBraces == FormatStyle::BS_Linux || 1032 Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1033 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1034 addUnwrappedLine(); 1035 1036 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1037 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1038 DeclarationScopeStack.size() > 1); 1039 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1040 // Munch the semicolon after a namespace. This is more common than one would 1041 // think. Puttin the semicolon into its own line is very ugly. 1042 if (FormatTok->Tok.is(tok::semi)) 1043 nextToken(); 1044 addUnwrappedLine(); 1045 } 1046 // FIXME: Add error handling. 1047 } 1048 1049 void UnwrappedLineParser::parseForOrWhileLoop() { 1050 assert((FormatTok->Tok.is(tok::kw_for) || FormatTok->Tok.is(tok::kw_while) || 1051 FormatTok->IsForEachMacro) && 1052 "'for', 'while' or foreach macro expected"); 1053 nextToken(); 1054 if (FormatTok->Tok.is(tok::l_paren)) 1055 parseParens(); 1056 if (FormatTok->Tok.is(tok::l_brace)) { 1057 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1058 parseBlock(/*MustBeDeclaration=*/false); 1059 addUnwrappedLine(); 1060 } else { 1061 addUnwrappedLine(); 1062 ++Line->Level; 1063 parseStructuralElement(); 1064 --Line->Level; 1065 } 1066 } 1067 1068 void UnwrappedLineParser::parseDoWhile() { 1069 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1070 nextToken(); 1071 if (FormatTok->Tok.is(tok::l_brace)) { 1072 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1073 parseBlock(/*MustBeDeclaration=*/false); 1074 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1075 addUnwrappedLine(); 1076 } else { 1077 addUnwrappedLine(); 1078 ++Line->Level; 1079 parseStructuralElement(); 1080 --Line->Level; 1081 } 1082 1083 // FIXME: Add error handling. 1084 if (!FormatTok->Tok.is(tok::kw_while)) { 1085 addUnwrappedLine(); 1086 return; 1087 } 1088 1089 nextToken(); 1090 parseStructuralElement(); 1091 } 1092 1093 void UnwrappedLineParser::parseLabel() { 1094 nextToken(); 1095 unsigned OldLineLevel = Line->Level; 1096 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1097 --Line->Level; 1098 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1099 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1100 parseBlock(/*MustBeDeclaration=*/false); 1101 if (FormatTok->Tok.is(tok::kw_break)) { 1102 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1103 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1104 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1105 addUnwrappedLine(); 1106 } 1107 parseStructuralElement(); 1108 } 1109 addUnwrappedLine(); 1110 } else { 1111 addUnwrappedLine(); 1112 } 1113 Line->Level = OldLineLevel; 1114 } 1115 1116 void UnwrappedLineParser::parseCaseLabel() { 1117 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1118 // FIXME: fix handling of complex expressions here. 1119 do { 1120 nextToken(); 1121 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1122 parseLabel(); 1123 } 1124 1125 void UnwrappedLineParser::parseSwitch() { 1126 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1127 nextToken(); 1128 if (FormatTok->Tok.is(tok::l_paren)) 1129 parseParens(); 1130 if (FormatTok->Tok.is(tok::l_brace)) { 1131 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1132 parseBlock(/*MustBeDeclaration=*/false); 1133 addUnwrappedLine(); 1134 } else { 1135 addUnwrappedLine(); 1136 ++Line->Level; 1137 parseStructuralElement(); 1138 --Line->Level; 1139 } 1140 } 1141 1142 void UnwrappedLineParser::parseAccessSpecifier() { 1143 nextToken(); 1144 // Understand Qt's slots. 1145 if (FormatTok->is(tok::identifier) && 1146 (FormatTok->TokenText == "slots" || FormatTok->TokenText == "Q_SLOTS")) 1147 nextToken(); 1148 // Otherwise, we don't know what it is, and we'd better keep the next token. 1149 if (FormatTok->Tok.is(tok::colon)) 1150 nextToken(); 1151 addUnwrappedLine(); 1152 } 1153 1154 void UnwrappedLineParser::parseEnum() { 1155 if (FormatTok->Tok.is(tok::kw_enum)) { 1156 // Won't be 'enum' for NS_ENUMs. 1157 nextToken(); 1158 } 1159 // Eat up enum class ... 1160 if (FormatTok->Tok.is(tok::kw_class) || 1161 FormatTok->Tok.is(tok::kw_struct)) 1162 nextToken(); 1163 while (FormatTok->Tok.getIdentifierInfo() || 1164 FormatTok->isOneOf(tok::colon, tok::coloncolon)) { 1165 nextToken(); 1166 // We can have macros or attributes in between 'enum' and the enum name. 1167 if (FormatTok->Tok.is(tok::l_paren)) { 1168 parseParens(); 1169 } 1170 if (FormatTok->Tok.is(tok::identifier)) 1171 nextToken(); 1172 } 1173 if (FormatTok->Tok.is(tok::l_brace)) { 1174 FormatTok->BlockKind = BK_Block; 1175 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1176 if (HasError) { 1177 if (FormatTok->is(tok::semi)) 1178 nextToken(); 1179 addUnwrappedLine(); 1180 } 1181 } 1182 // We fall through to parsing a structural element afterwards, so that in 1183 // enum A {} n, m; 1184 // "} n, m;" will end up in one unwrapped line. 1185 } 1186 1187 void UnwrappedLineParser::parseRecord() { 1188 nextToken(); 1189 if (FormatTok->Tok.is(tok::identifier) || 1190 FormatTok->Tok.is(tok::kw___attribute) || 1191 FormatTok->Tok.is(tok::kw___declspec) || 1192 FormatTok->Tok.is(tok::kw_alignas)) { 1193 nextToken(); 1194 // We can have macros or attributes in between 'class' and the class name. 1195 if (FormatTok->Tok.is(tok::l_paren)) { 1196 parseParens(); 1197 } 1198 // The actual identifier can be a nested name specifier, and in macros 1199 // it is often token-pasted. 1200 while (FormatTok->Tok.is(tok::identifier) || 1201 FormatTok->Tok.is(tok::coloncolon) || 1202 FormatTok->Tok.is(tok::hashhash)) 1203 nextToken(); 1204 1205 // Note that parsing away template declarations here leads to incorrectly 1206 // accepting function declarations as record declarations. 1207 // In general, we cannot solve this problem. Consider: 1208 // class A<int> B() {} 1209 // which can be a function definition or a class definition when B() is a 1210 // macro. If we find enough real-world cases where this is a problem, we 1211 // can parse for the 'template' keyword in the beginning of the statement, 1212 // and thus rule out the record production in case there is no template 1213 // (this would still leave us with an ambiguity between template function 1214 // and class declarations). 1215 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1216 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1217 if (FormatTok->Tok.is(tok::semi)) 1218 return; 1219 nextToken(); 1220 } 1221 } 1222 } 1223 if (FormatTok->Tok.is(tok::l_brace)) { 1224 if (Style.BreakBeforeBraces == FormatStyle::BS_Linux || 1225 Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1226 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1227 addUnwrappedLine(); 1228 1229 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1230 /*MunchSemi=*/false); 1231 } 1232 // We fall through to parsing a structural element afterwards, so 1233 // class A {} n, m; 1234 // will end up in one unwrapped line. 1235 } 1236 1237 void UnwrappedLineParser::parseObjCProtocolList() { 1238 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1239 do 1240 nextToken(); 1241 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1242 nextToken(); // Skip '>'. 1243 } 1244 1245 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1246 do { 1247 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1248 nextToken(); 1249 addUnwrappedLine(); 1250 break; 1251 } 1252 if (FormatTok->is(tok::l_brace)) { 1253 parseBlock(/*MustBeDeclaration=*/false); 1254 // In ObjC interfaces, nothing should be following the "}". 1255 addUnwrappedLine(); 1256 } else if (FormatTok->is(tok::r_brace)) { 1257 // Ignore stray "}". parseStructuralElement doesn't consume them. 1258 nextToken(); 1259 addUnwrappedLine(); 1260 } else { 1261 parseStructuralElement(); 1262 } 1263 } while (!eof()); 1264 } 1265 1266 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1267 nextToken(); 1268 nextToken(); // interface name 1269 1270 // @interface can be followed by either a base class, or a category. 1271 if (FormatTok->Tok.is(tok::colon)) { 1272 nextToken(); 1273 nextToken(); // base class name 1274 } else if (FormatTok->Tok.is(tok::l_paren)) 1275 // Skip category, if present. 1276 parseParens(); 1277 1278 if (FormatTok->Tok.is(tok::less)) 1279 parseObjCProtocolList(); 1280 1281 // If instance variables are present, keep the '{' on the first line too. 1282 if (FormatTok->Tok.is(tok::l_brace)) 1283 parseBlock(/*MustBeDeclaration=*/true); 1284 1285 // With instance variables, this puts '}' on its own line. Without instance 1286 // variables, this ends the @interface line. 1287 addUnwrappedLine(); 1288 1289 parseObjCUntilAtEnd(); 1290 } 1291 1292 void UnwrappedLineParser::parseObjCProtocol() { 1293 nextToken(); 1294 nextToken(); // protocol name 1295 1296 if (FormatTok->Tok.is(tok::less)) 1297 parseObjCProtocolList(); 1298 1299 // Check for protocol declaration. 1300 if (FormatTok->Tok.is(tok::semi)) { 1301 nextToken(); 1302 return addUnwrappedLine(); 1303 } 1304 1305 addUnwrappedLine(); 1306 parseObjCUntilAtEnd(); 1307 } 1308 1309 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1310 StringRef Prefix = "") { 1311 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1312 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1313 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1314 E = Line.Tokens.end(); 1315 I != E; ++I) { 1316 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1317 } 1318 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1319 E = Line.Tokens.end(); 1320 I != E; ++I) { 1321 const UnwrappedLineNode &Node = *I; 1322 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1323 I = Node.Children.begin(), 1324 E = Node.Children.end(); 1325 I != E; ++I) { 1326 printDebugInfo(*I, "\nChild: "); 1327 } 1328 } 1329 llvm::dbgs() << "\n"; 1330 } 1331 1332 void UnwrappedLineParser::addUnwrappedLine() { 1333 if (Line->Tokens.empty()) 1334 return; 1335 DEBUG({ 1336 if (CurrentLines == &Lines) 1337 printDebugInfo(*Line); 1338 }); 1339 CurrentLines->push_back(*Line); 1340 Line->Tokens.clear(); 1341 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1342 for (SmallVectorImpl<UnwrappedLine>::iterator 1343 I = PreprocessorDirectives.begin(), 1344 E = PreprocessorDirectives.end(); 1345 I != E; ++I) { 1346 CurrentLines->push_back(*I); 1347 } 1348 PreprocessorDirectives.clear(); 1349 } 1350 } 1351 1352 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1353 1354 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1355 bool JustComments = Line->Tokens.empty(); 1356 for (SmallVectorImpl<FormatToken *>::const_iterator 1357 I = CommentsBeforeNextToken.begin(), 1358 E = CommentsBeforeNextToken.end(); 1359 I != E; ++I) { 1360 if ((*I)->NewlinesBefore && JustComments) { 1361 addUnwrappedLine(); 1362 } 1363 pushToken(*I); 1364 } 1365 if (NewlineBeforeNext && JustComments) { 1366 addUnwrappedLine(); 1367 } 1368 CommentsBeforeNextToken.clear(); 1369 } 1370 1371 void UnwrappedLineParser::nextToken() { 1372 if (eof()) 1373 return; 1374 flushComments(FormatTok->NewlinesBefore > 0); 1375 pushToken(FormatTok); 1376 readToken(); 1377 } 1378 1379 void UnwrappedLineParser::readToken() { 1380 bool CommentsInCurrentLine = true; 1381 do { 1382 FormatTok = Tokens->getNextToken(); 1383 assert(FormatTok); 1384 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1385 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1386 // If there is an unfinished unwrapped line, we flush the preprocessor 1387 // directives only after that unwrapped line was finished later. 1388 bool SwitchToPreprocessorLines = 1389 !Line->Tokens.empty() && CurrentLines == &Lines; 1390 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1391 // Comments stored before the preprocessor directive need to be output 1392 // before the preprocessor directive, at the same level as the 1393 // preprocessor directive, as we consider them to apply to the directive. 1394 flushComments(FormatTok->NewlinesBefore > 0); 1395 parsePPDirective(); 1396 } 1397 1398 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1399 !Line->InPPDirective) { 1400 continue; 1401 } 1402 1403 if (!FormatTok->Tok.is(tok::comment)) 1404 return; 1405 if (FormatTok->NewlinesBefore > 0 || FormatTok->IsFirst) { 1406 CommentsInCurrentLine = false; 1407 } 1408 if (CommentsInCurrentLine) { 1409 pushToken(FormatTok); 1410 } else { 1411 CommentsBeforeNextToken.push_back(FormatTok); 1412 } 1413 } while (!eof()); 1414 } 1415 1416 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1417 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1418 if (MustBreakBeforeNextToken) { 1419 Line->Tokens.back().Tok->MustBreakBefore = true; 1420 MustBreakBeforeNextToken = false; 1421 } 1422 } 1423 1424 } // end namespace format 1425 } // end namespace clang 1426