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 #include "UnwrappedLineParser.h" 17 #include "llvm/Support/Debug.h" 18 19 #define DEBUG_TYPE "format-parser" 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 bool ProbablyBracedList = false; 339 if (Style.Language == FormatStyle::LK_Proto) { 340 ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); 341 } else { 342 // If there is a comma, semicolon or right paren after the closing 343 // brace, we assume this is a braced initializer list. Note that 344 // regardless how we mark inner braces here, we will overwrite the 345 // BlockKind later if we parse a braced list (where all blocks 346 // inside are by default braced lists), or when we explicitly detect 347 // blocks (for example while parsing lambdas). 348 // 349 // We exclude + and - as they can be ObjC visibility modifiers. 350 ProbablyBracedList = 351 NextTok->isOneOf(tok::comma, tok::semi, tok::period, tok::colon, 352 tok::r_paren, tok::r_square, tok::l_brace) || 353 (NextTok->isBinaryOperator() && 354 !NextTok->isOneOf(tok::plus, tok::minus)); 355 } 356 if (ProbablyBracedList) { 357 Tok->BlockKind = BK_BracedInit; 358 LBraceStack.back()->BlockKind = BK_BracedInit; 359 } else { 360 Tok->BlockKind = BK_Block; 361 LBraceStack.back()->BlockKind = BK_Block; 362 } 363 } 364 LBraceStack.pop_back(); 365 } 366 break; 367 case tok::at: 368 case tok::semi: 369 case tok::kw_if: 370 case tok::kw_while: 371 case tok::kw_for: 372 case tok::kw_switch: 373 case tok::kw_try: 374 if (!LBraceStack.empty()) 375 LBraceStack.back()->BlockKind = BK_Block; 376 break; 377 default: 378 break; 379 } 380 Tok = NextTok; 381 Position += ReadTokens; 382 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 383 // Assume other blocks for all unclosed opening braces. 384 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 385 if (LBraceStack[i]->BlockKind == BK_Unknown) 386 LBraceStack[i]->BlockKind = BK_Block; 387 } 388 389 FormatTok = Tokens->setPosition(StoredPosition); 390 } 391 392 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 393 bool MunchSemi) { 394 assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected"); 395 unsigned InitialLevel = Line->Level; 396 nextToken(); 397 398 addUnwrappedLine(); 399 400 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 401 MustBeDeclaration); 402 if (AddLevel) 403 ++Line->Level; 404 parseLevel(/*HasOpeningBrace=*/true); 405 406 if (!FormatTok->Tok.is(tok::r_brace)) { 407 Line->Level = InitialLevel; 408 StructuralError = true; 409 return; 410 } 411 412 nextToken(); // Munch the closing brace. 413 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 414 nextToken(); 415 Line->Level = InitialLevel; 416 } 417 418 void UnwrappedLineParser::parseChildBlock() { 419 FormatTok->BlockKind = BK_Block; 420 nextToken(); 421 { 422 ScopedLineState LineState(*this); 423 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 424 /*MustBeDeclaration=*/false); 425 Line->Level += 1; 426 parseLevel(/*HasOpeningBrace=*/true); 427 Line->Level -= 1; 428 } 429 nextToken(); 430 } 431 432 void UnwrappedLineParser::parsePPDirective() { 433 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 434 ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError); 435 nextToken(); 436 437 if (FormatTok->Tok.getIdentifierInfo() == NULL) { 438 parsePPUnknown(); 439 return; 440 } 441 442 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 443 case tok::pp_define: 444 parsePPDefine(); 445 return; 446 case tok::pp_if: 447 parsePPIf(/*IfDef=*/false); 448 break; 449 case tok::pp_ifdef: 450 case tok::pp_ifndef: 451 parsePPIf(/*IfDef=*/true); 452 break; 453 case tok::pp_else: 454 parsePPElse(); 455 break; 456 case tok::pp_elif: 457 parsePPElIf(); 458 break; 459 case tok::pp_endif: 460 parsePPEndIf(); 461 break; 462 default: 463 parsePPUnknown(); 464 break; 465 } 466 } 467 468 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 469 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 470 PPStack.push_back(PP_Unreachable); 471 else 472 PPStack.push_back(PP_Conditional); 473 } 474 475 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 476 ++PPBranchLevel; 477 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 478 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 479 PPLevelBranchIndex.push_back(0); 480 PPLevelBranchCount.push_back(0); 481 } 482 PPChainBranchIndex.push(0); 483 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 484 conditionalCompilationCondition(Unreachable || Skip); 485 } 486 487 void UnwrappedLineParser::conditionalCompilationAlternative() { 488 if (!PPStack.empty()) 489 PPStack.pop_back(); 490 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 491 if (!PPChainBranchIndex.empty()) 492 ++PPChainBranchIndex.top(); 493 conditionalCompilationCondition( 494 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 495 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 496 } 497 498 void UnwrappedLineParser::conditionalCompilationEnd() { 499 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 500 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 501 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 502 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 503 } 504 } 505 // Guard against #endif's without #if. 506 if (PPBranchLevel > 0) 507 --PPBranchLevel; 508 if (!PPChainBranchIndex.empty()) 509 PPChainBranchIndex.pop(); 510 if (!PPStack.empty()) 511 PPStack.pop_back(); 512 } 513 514 void UnwrappedLineParser::parsePPIf(bool IfDef) { 515 nextToken(); 516 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 517 StringRef(FormatTok->Tok.getLiteralData(), 518 FormatTok->Tok.getLength()) == "0") || 519 FormatTok->Tok.is(tok::kw_false); 520 conditionalCompilationStart(!IfDef && IsLiteralFalse); 521 parsePPUnknown(); 522 } 523 524 void UnwrappedLineParser::parsePPElse() { 525 conditionalCompilationAlternative(); 526 parsePPUnknown(); 527 } 528 529 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 530 531 void UnwrappedLineParser::parsePPEndIf() { 532 conditionalCompilationEnd(); 533 parsePPUnknown(); 534 } 535 536 void UnwrappedLineParser::parsePPDefine() { 537 nextToken(); 538 539 if (FormatTok->Tok.getKind() != tok::identifier) { 540 parsePPUnknown(); 541 return; 542 } 543 nextToken(); 544 if (FormatTok->Tok.getKind() == tok::l_paren && 545 FormatTok->WhitespaceRange.getBegin() == 546 FormatTok->WhitespaceRange.getEnd()) { 547 parseParens(); 548 } 549 addUnwrappedLine(); 550 Line->Level = 1; 551 552 // Errors during a preprocessor directive can only affect the layout of the 553 // preprocessor directive, and thus we ignore them. An alternative approach 554 // would be to use the same approach we use on the file level (no 555 // re-indentation if there was a structural error) within the macro 556 // definition. 557 parseFile(); 558 } 559 560 void UnwrappedLineParser::parsePPUnknown() { 561 do { 562 nextToken(); 563 } while (!eof()); 564 addUnwrappedLine(); 565 } 566 567 // Here we blacklist certain tokens that are not usually the first token in an 568 // unwrapped line. This is used in attempt to distinguish macro calls without 569 // trailing semicolons from other constructs split to several lines. 570 bool tokenCanStartNewLine(clang::Token Tok) { 571 // Semicolon can be a null-statement, l_square can be a start of a macro or 572 // a C++11 attribute, but this doesn't seem to be common. 573 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 574 Tok.isNot(tok::l_square) && 575 // Tokens that can only be used as binary operators and a part of 576 // overloaded operator names. 577 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 578 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 579 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 580 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 581 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 582 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 583 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 584 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 585 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 586 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 587 Tok.isNot(tok::lesslessequal) && 588 // Colon is used in labels, base class lists, initializer lists, 589 // range-based for loops, ternary operator, but should never be the 590 // first token in an unwrapped line. 591 Tok.isNot(tok::colon); 592 } 593 594 void UnwrappedLineParser::parseStructuralElement() { 595 assert(!FormatTok->Tok.is(tok::l_brace)); 596 switch (FormatTok->Tok.getKind()) { 597 case tok::at: 598 nextToken(); 599 if (FormatTok->Tok.is(tok::l_brace)) { 600 parseBracedList(); 601 break; 602 } 603 switch (FormatTok->Tok.getObjCKeywordID()) { 604 case tok::objc_public: 605 case tok::objc_protected: 606 case tok::objc_package: 607 case tok::objc_private: 608 return parseAccessSpecifier(); 609 case tok::objc_interface: 610 case tok::objc_implementation: 611 return parseObjCInterfaceOrImplementation(); 612 case tok::objc_protocol: 613 return parseObjCProtocol(); 614 case tok::objc_end: 615 return; // Handled by the caller. 616 case tok::objc_optional: 617 case tok::objc_required: 618 nextToken(); 619 addUnwrappedLine(); 620 return; 621 default: 622 break; 623 } 624 break; 625 case tok::kw_namespace: 626 parseNamespace(); 627 return; 628 case tok::kw_inline: 629 nextToken(); 630 if (FormatTok->Tok.is(tok::kw_namespace)) { 631 parseNamespace(); 632 return; 633 } 634 break; 635 case tok::kw_public: 636 case tok::kw_protected: 637 case tok::kw_private: 638 parseAccessSpecifier(); 639 return; 640 case tok::kw_if: 641 parseIfThenElse(); 642 return; 643 case tok::kw_for: 644 case tok::kw_while: 645 parseForOrWhileLoop(); 646 return; 647 case tok::kw_do: 648 parseDoWhile(); 649 return; 650 case tok::kw_switch: 651 parseSwitch(); 652 return; 653 case tok::kw_default: 654 nextToken(); 655 parseLabel(); 656 return; 657 case tok::kw_case: 658 parseCaseLabel(); 659 return; 660 case tok::kw_extern: 661 nextToken(); 662 if (FormatTok->Tok.is(tok::string_literal)) { 663 nextToken(); 664 if (FormatTok->Tok.is(tok::l_brace)) { 665 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 666 addUnwrappedLine(); 667 return; 668 } 669 } 670 break; 671 case tok::identifier: 672 if (FormatTok->IsForEachMacro) { 673 parseForOrWhileLoop(); 674 return; 675 } 676 // In all other cases, parse the declaration. 677 break; 678 default: 679 break; 680 } 681 do { 682 switch (FormatTok->Tok.getKind()) { 683 case tok::at: 684 nextToken(); 685 if (FormatTok->Tok.is(tok::l_brace)) 686 parseBracedList(); 687 break; 688 case tok::kw_enum: 689 parseEnum(); 690 break; 691 case tok::kw_typedef: 692 nextToken(); 693 // FIXME: Use the IdentifierTable instead. 694 if (FormatTok->TokenText == "NS_ENUM") 695 parseEnum(); 696 break; 697 case tok::kw_struct: 698 case tok::kw_union: 699 case tok::kw_class: 700 parseRecord(); 701 // A record declaration or definition is always the start of a structural 702 // element. 703 break; 704 case tok::semi: 705 nextToken(); 706 addUnwrappedLine(); 707 return; 708 case tok::r_brace: 709 addUnwrappedLine(); 710 return; 711 case tok::l_paren: 712 parseParens(); 713 break; 714 case tok::caret: 715 nextToken(); 716 if (FormatTok->Tok.isAnyIdentifier() || 717 FormatTok->isSimpleTypeSpecifier()) 718 nextToken(); 719 if (FormatTok->is(tok::l_paren)) 720 parseParens(); 721 if (FormatTok->is(tok::l_brace)) 722 parseChildBlock(); 723 break; 724 case tok::l_brace: 725 if (!tryToParseBracedList()) { 726 // A block outside of parentheses must be the last part of a 727 // structural element. 728 // FIXME: Figure out cases where this is not true, and add projections 729 // for them (the one we know is missing are lambdas). 730 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 731 addUnwrappedLine(); 732 FormatTok->Type = TT_FunctionLBrace; 733 parseBlock(/*MustBeDeclaration=*/false); 734 addUnwrappedLine(); 735 return; 736 } 737 // Otherwise this was a braced init list, and the structural 738 // element continues. 739 break; 740 case tok::identifier: { 741 StringRef Text = FormatTok->TokenText; 742 nextToken(); 743 if (Line->Tokens.size() == 1) { 744 if (FormatTok->Tok.is(tok::colon)) { 745 parseLabel(); 746 return; 747 } 748 // Recognize function-like macro usages without trailing semicolon. 749 if (FormatTok->Tok.is(tok::l_paren)) { 750 parseParens(); 751 if (FormatTok->NewlinesBefore > 0 && 752 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 753 addUnwrappedLine(); 754 return; 755 } 756 } else if (FormatTok->HasUnescapedNewline && Text.size() >= 5 && 757 Text == Text.upper()) { 758 // Recognize free-standing macros like Q_OBJECT. 759 addUnwrappedLine(); 760 return; 761 } 762 } 763 break; 764 } 765 case tok::equal: 766 nextToken(); 767 if (FormatTok->Tok.is(tok::l_brace)) { 768 parseBracedList(); 769 } 770 break; 771 case tok::l_square: 772 parseSquare(); 773 break; 774 default: 775 nextToken(); 776 break; 777 } 778 } while (!eof()); 779 } 780 781 bool UnwrappedLineParser::tryToParseLambda() { 782 // FIXME: This is a dirty way to access the previous token. Find a better 783 // solution. 784 if (!Line->Tokens.empty() && 785 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator) || 786 Line->Tokens.back().Tok->closesScope() || 787 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 788 nextToken(); 789 return false; 790 } 791 assert(FormatTok->is(tok::l_square)); 792 FormatToken &LSquare = *FormatTok; 793 if (!tryToParseLambdaIntroducer()) 794 return false; 795 796 while (FormatTok->isNot(tok::l_brace)) { 797 if (FormatTok->isSimpleTypeSpecifier()) { 798 nextToken(); 799 continue; 800 } 801 switch (FormatTok->Tok.getKind()) { 802 case tok::l_brace: 803 break; 804 case tok::l_paren: 805 parseParens(); 806 break; 807 case tok::less: 808 case tok::greater: 809 case tok::identifier: 810 case tok::coloncolon: 811 case tok::kw_mutable: 812 nextToken(); 813 break; 814 case tok::arrow: 815 FormatTok->Type = TT_TrailingReturnArrow; 816 nextToken(); 817 break; 818 default: 819 return true; 820 } 821 } 822 LSquare.Type = TT_LambdaLSquare; 823 parseChildBlock(); 824 return true; 825 } 826 827 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 828 nextToken(); 829 if (FormatTok->is(tok::equal)) { 830 nextToken(); 831 if (FormatTok->is(tok::r_square)) { 832 nextToken(); 833 return true; 834 } 835 if (FormatTok->isNot(tok::comma)) 836 return false; 837 nextToken(); 838 } else if (FormatTok->is(tok::amp)) { 839 nextToken(); 840 if (FormatTok->is(tok::r_square)) { 841 nextToken(); 842 return true; 843 } 844 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 845 return false; 846 } 847 if (FormatTok->is(tok::comma)) 848 nextToken(); 849 } else if (FormatTok->is(tok::r_square)) { 850 nextToken(); 851 return true; 852 } 853 do { 854 if (FormatTok->is(tok::amp)) 855 nextToken(); 856 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 857 return false; 858 nextToken(); 859 if (FormatTok->is(tok::comma)) { 860 nextToken(); 861 } else if (FormatTok->is(tok::r_square)) { 862 nextToken(); 863 return true; 864 } else { 865 return false; 866 } 867 } while (!eof()); 868 return false; 869 } 870 871 bool UnwrappedLineParser::tryToParseBracedList() { 872 if (FormatTok->BlockKind == BK_Unknown) 873 calculateBraceTypes(); 874 assert(FormatTok->BlockKind != BK_Unknown); 875 if (FormatTok->BlockKind == BK_Block) 876 return false; 877 parseBracedList(); 878 return true; 879 } 880 881 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 882 bool HasError = false; 883 nextToken(); 884 885 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 886 // replace this by using parseAssigmentExpression() inside. 887 do { 888 // FIXME: When we start to support lambdas, we'll want to parse them away 889 // here, otherwise our bail-out scenarios below break. The better solution 890 // might be to just implement a more or less complete expression parser. 891 switch (FormatTok->Tok.getKind()) { 892 case tok::caret: 893 nextToken(); 894 if (FormatTok->is(tok::l_brace)) { 895 parseChildBlock(); 896 } 897 break; 898 case tok::l_square: 899 tryToParseLambda(); 900 break; 901 case tok::l_brace: 902 // Assume there are no blocks inside a braced init list apart 903 // from the ones we explicitly parse out (like lambdas). 904 FormatTok->BlockKind = BK_BracedInit; 905 parseBracedList(); 906 break; 907 case tok::r_brace: 908 nextToken(); 909 return !HasError; 910 case tok::semi: 911 HasError = true; 912 if (!ContinueOnSemicolons) 913 return !HasError; 914 nextToken(); 915 break; 916 case tok::comma: 917 nextToken(); 918 break; 919 default: 920 nextToken(); 921 break; 922 } 923 } while (!eof()); 924 return false; 925 } 926 927 void UnwrappedLineParser::parseParens() { 928 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 929 nextToken(); 930 do { 931 switch (FormatTok->Tok.getKind()) { 932 case tok::l_paren: 933 parseParens(); 934 break; 935 case tok::r_paren: 936 nextToken(); 937 return; 938 case tok::r_brace: 939 // A "}" inside parenthesis is an error if there wasn't a matching "{". 940 return; 941 case tok::l_square: 942 tryToParseLambda(); 943 break; 944 case tok::l_brace: { 945 if (!tryToParseBracedList()) { 946 parseChildBlock(); 947 } 948 break; 949 } 950 case tok::at: 951 nextToken(); 952 if (FormatTok->Tok.is(tok::l_brace)) 953 parseBracedList(); 954 break; 955 default: 956 nextToken(); 957 break; 958 } 959 } while (!eof()); 960 } 961 962 void UnwrappedLineParser::parseSquare() { 963 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 964 if (tryToParseLambda()) 965 return; 966 do { 967 switch (FormatTok->Tok.getKind()) { 968 case tok::l_paren: 969 parseParens(); 970 break; 971 case tok::r_square: 972 nextToken(); 973 return; 974 case tok::r_brace: 975 // A "}" inside parenthesis is an error if there wasn't a matching "{". 976 return; 977 case tok::l_square: 978 parseSquare(); 979 break; 980 case tok::l_brace: { 981 if (!tryToParseBracedList()) { 982 parseChildBlock(); 983 } 984 break; 985 } 986 case tok::at: 987 nextToken(); 988 if (FormatTok->Tok.is(tok::l_brace)) 989 parseBracedList(); 990 break; 991 default: 992 nextToken(); 993 break; 994 } 995 } while (!eof()); 996 } 997 998 void UnwrappedLineParser::parseIfThenElse() { 999 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1000 nextToken(); 1001 if (FormatTok->Tok.is(tok::l_paren)) 1002 parseParens(); 1003 bool NeedsUnwrappedLine = false; 1004 if (FormatTok->Tok.is(tok::l_brace)) { 1005 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1006 parseBlock(/*MustBeDeclaration=*/false); 1007 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1008 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1009 addUnwrappedLine(); 1010 } else { 1011 NeedsUnwrappedLine = true; 1012 } 1013 } else { 1014 addUnwrappedLine(); 1015 ++Line->Level; 1016 parseStructuralElement(); 1017 --Line->Level; 1018 } 1019 if (FormatTok->Tok.is(tok::kw_else)) { 1020 nextToken(); 1021 if (FormatTok->Tok.is(tok::l_brace)) { 1022 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1023 parseBlock(/*MustBeDeclaration=*/false); 1024 addUnwrappedLine(); 1025 } else if (FormatTok->Tok.is(tok::kw_if)) { 1026 parseIfThenElse(); 1027 } else { 1028 addUnwrappedLine(); 1029 ++Line->Level; 1030 parseStructuralElement(); 1031 --Line->Level; 1032 } 1033 } else if (NeedsUnwrappedLine) { 1034 addUnwrappedLine(); 1035 } 1036 } 1037 1038 void UnwrappedLineParser::parseNamespace() { 1039 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1040 nextToken(); 1041 if (FormatTok->Tok.is(tok::identifier)) 1042 nextToken(); 1043 if (FormatTok->Tok.is(tok::l_brace)) { 1044 if (Style.BreakBeforeBraces == FormatStyle::BS_Linux || 1045 Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1046 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1047 addUnwrappedLine(); 1048 1049 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1050 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1051 DeclarationScopeStack.size() > 1); 1052 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1053 // Munch the semicolon after a namespace. This is more common than one would 1054 // think. Puttin the semicolon into its own line is very ugly. 1055 if (FormatTok->Tok.is(tok::semi)) 1056 nextToken(); 1057 addUnwrappedLine(); 1058 } 1059 // FIXME: Add error handling. 1060 } 1061 1062 void UnwrappedLineParser::parseForOrWhileLoop() { 1063 assert((FormatTok->Tok.is(tok::kw_for) || FormatTok->Tok.is(tok::kw_while) || 1064 FormatTok->IsForEachMacro) && 1065 "'for', 'while' or foreach macro expected"); 1066 nextToken(); 1067 if (FormatTok->Tok.is(tok::l_paren)) 1068 parseParens(); 1069 if (FormatTok->Tok.is(tok::l_brace)) { 1070 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1071 parseBlock(/*MustBeDeclaration=*/false); 1072 addUnwrappedLine(); 1073 } else { 1074 addUnwrappedLine(); 1075 ++Line->Level; 1076 parseStructuralElement(); 1077 --Line->Level; 1078 } 1079 } 1080 1081 void UnwrappedLineParser::parseDoWhile() { 1082 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1083 nextToken(); 1084 if (FormatTok->Tok.is(tok::l_brace)) { 1085 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1086 parseBlock(/*MustBeDeclaration=*/false); 1087 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1088 addUnwrappedLine(); 1089 } else { 1090 addUnwrappedLine(); 1091 ++Line->Level; 1092 parseStructuralElement(); 1093 --Line->Level; 1094 } 1095 1096 // FIXME: Add error handling. 1097 if (!FormatTok->Tok.is(tok::kw_while)) { 1098 addUnwrappedLine(); 1099 return; 1100 } 1101 1102 nextToken(); 1103 parseStructuralElement(); 1104 } 1105 1106 void UnwrappedLineParser::parseLabel() { 1107 nextToken(); 1108 unsigned OldLineLevel = Line->Level; 1109 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1110 --Line->Level; 1111 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1112 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1113 parseBlock(/*MustBeDeclaration=*/false); 1114 if (FormatTok->Tok.is(tok::kw_break)) { 1115 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1116 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1117 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1118 addUnwrappedLine(); 1119 } 1120 parseStructuralElement(); 1121 } 1122 addUnwrappedLine(); 1123 } else { 1124 addUnwrappedLine(); 1125 } 1126 Line->Level = OldLineLevel; 1127 } 1128 1129 void UnwrappedLineParser::parseCaseLabel() { 1130 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1131 // FIXME: fix handling of complex expressions here. 1132 do { 1133 nextToken(); 1134 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1135 parseLabel(); 1136 } 1137 1138 void UnwrappedLineParser::parseSwitch() { 1139 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1140 nextToken(); 1141 if (FormatTok->Tok.is(tok::l_paren)) 1142 parseParens(); 1143 if (FormatTok->Tok.is(tok::l_brace)) { 1144 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1145 parseBlock(/*MustBeDeclaration=*/false); 1146 addUnwrappedLine(); 1147 } else { 1148 addUnwrappedLine(); 1149 ++Line->Level; 1150 parseStructuralElement(); 1151 --Line->Level; 1152 } 1153 } 1154 1155 void UnwrappedLineParser::parseAccessSpecifier() { 1156 nextToken(); 1157 // Understand Qt's slots. 1158 if (FormatTok->is(tok::identifier) && 1159 (FormatTok->TokenText == "slots" || FormatTok->TokenText == "Q_SLOTS")) 1160 nextToken(); 1161 // Otherwise, we don't know what it is, and we'd better keep the next token. 1162 if (FormatTok->Tok.is(tok::colon)) 1163 nextToken(); 1164 addUnwrappedLine(); 1165 } 1166 1167 void UnwrappedLineParser::parseEnum() { 1168 if (FormatTok->Tok.is(tok::kw_enum)) { 1169 // Won't be 'enum' for NS_ENUMs. 1170 nextToken(); 1171 } 1172 // Eat up enum class ... 1173 if (FormatTok->Tok.is(tok::kw_class) || 1174 FormatTok->Tok.is(tok::kw_struct)) 1175 nextToken(); 1176 while (FormatTok->Tok.getIdentifierInfo() || 1177 FormatTok->isOneOf(tok::colon, tok::coloncolon)) { 1178 nextToken(); 1179 // We can have macros or attributes in between 'enum' and the enum name. 1180 if (FormatTok->Tok.is(tok::l_paren)) { 1181 parseParens(); 1182 } 1183 if (FormatTok->Tok.is(tok::identifier)) 1184 nextToken(); 1185 } 1186 if (FormatTok->Tok.is(tok::l_brace)) { 1187 FormatTok->BlockKind = BK_Block; 1188 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1189 if (HasError) { 1190 if (FormatTok->is(tok::semi)) 1191 nextToken(); 1192 addUnwrappedLine(); 1193 } 1194 } 1195 // We fall through to parsing a structural element afterwards, so that in 1196 // enum A {} n, m; 1197 // "} n, m;" will end up in one unwrapped line. 1198 } 1199 1200 void UnwrappedLineParser::parseRecord() { 1201 nextToken(); 1202 if (FormatTok->Tok.is(tok::identifier) || 1203 FormatTok->Tok.is(tok::kw___attribute) || 1204 FormatTok->Tok.is(tok::kw___declspec) || 1205 FormatTok->Tok.is(tok::kw_alignas)) { 1206 nextToken(); 1207 // We can have macros or attributes in between 'class' and the class name. 1208 if (FormatTok->Tok.is(tok::l_paren)) { 1209 parseParens(); 1210 } 1211 // The actual identifier can be a nested name specifier, and in macros 1212 // it is often token-pasted. 1213 while (FormatTok->Tok.is(tok::identifier) || 1214 FormatTok->Tok.is(tok::coloncolon) || 1215 FormatTok->Tok.is(tok::hashhash)) 1216 nextToken(); 1217 1218 // Note that parsing away template declarations here leads to incorrectly 1219 // accepting function declarations as record declarations. 1220 // In general, we cannot solve this problem. Consider: 1221 // class A<int> B() {} 1222 // which can be a function definition or a class definition when B() is a 1223 // macro. If we find enough real-world cases where this is a problem, we 1224 // can parse for the 'template' keyword in the beginning of the statement, 1225 // and thus rule out the record production in case there is no template 1226 // (this would still leave us with an ambiguity between template function 1227 // and class declarations). 1228 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1229 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1230 if (FormatTok->Tok.is(tok::semi)) 1231 return; 1232 nextToken(); 1233 } 1234 } 1235 } 1236 if (FormatTok->Tok.is(tok::l_brace)) { 1237 if (Style.BreakBeforeBraces == FormatStyle::BS_Linux || 1238 Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1239 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1240 addUnwrappedLine(); 1241 1242 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1243 /*MunchSemi=*/false); 1244 } 1245 // We fall through to parsing a structural element afterwards, so 1246 // class A {} n, m; 1247 // will end up in one unwrapped line. 1248 } 1249 1250 void UnwrappedLineParser::parseObjCProtocolList() { 1251 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1252 do 1253 nextToken(); 1254 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1255 nextToken(); // Skip '>'. 1256 } 1257 1258 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1259 do { 1260 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1261 nextToken(); 1262 addUnwrappedLine(); 1263 break; 1264 } 1265 if (FormatTok->is(tok::l_brace)) { 1266 parseBlock(/*MustBeDeclaration=*/false); 1267 // In ObjC interfaces, nothing should be following the "}". 1268 addUnwrappedLine(); 1269 } else if (FormatTok->is(tok::r_brace)) { 1270 // Ignore stray "}". parseStructuralElement doesn't consume them. 1271 nextToken(); 1272 addUnwrappedLine(); 1273 } else { 1274 parseStructuralElement(); 1275 } 1276 } while (!eof()); 1277 } 1278 1279 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1280 nextToken(); 1281 nextToken(); // interface name 1282 1283 // @interface can be followed by either a base class, or a category. 1284 if (FormatTok->Tok.is(tok::colon)) { 1285 nextToken(); 1286 nextToken(); // base class name 1287 } else if (FormatTok->Tok.is(tok::l_paren)) 1288 // Skip category, if present. 1289 parseParens(); 1290 1291 if (FormatTok->Tok.is(tok::less)) 1292 parseObjCProtocolList(); 1293 1294 if (FormatTok->Tok.is(tok::l_brace)) { 1295 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1296 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1297 addUnwrappedLine(); 1298 parseBlock(/*MustBeDeclaration=*/true); 1299 } 1300 1301 // With instance variables, this puts '}' on its own line. Without instance 1302 // variables, this ends the @interface line. 1303 addUnwrappedLine(); 1304 1305 parseObjCUntilAtEnd(); 1306 } 1307 1308 void UnwrappedLineParser::parseObjCProtocol() { 1309 nextToken(); 1310 nextToken(); // protocol name 1311 1312 if (FormatTok->Tok.is(tok::less)) 1313 parseObjCProtocolList(); 1314 1315 // Check for protocol declaration. 1316 if (FormatTok->Tok.is(tok::semi)) { 1317 nextToken(); 1318 return addUnwrappedLine(); 1319 } 1320 1321 addUnwrappedLine(); 1322 parseObjCUntilAtEnd(); 1323 } 1324 1325 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1326 StringRef Prefix = "") { 1327 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1328 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1329 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1330 E = Line.Tokens.end(); 1331 I != E; ++I) { 1332 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1333 } 1334 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1335 E = Line.Tokens.end(); 1336 I != E; ++I) { 1337 const UnwrappedLineNode &Node = *I; 1338 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1339 I = Node.Children.begin(), 1340 E = Node.Children.end(); 1341 I != E; ++I) { 1342 printDebugInfo(*I, "\nChild: "); 1343 } 1344 } 1345 llvm::dbgs() << "\n"; 1346 } 1347 1348 void UnwrappedLineParser::addUnwrappedLine() { 1349 if (Line->Tokens.empty()) 1350 return; 1351 DEBUG({ 1352 if (CurrentLines == &Lines) 1353 printDebugInfo(*Line); 1354 }); 1355 CurrentLines->push_back(*Line); 1356 Line->Tokens.clear(); 1357 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1358 for (SmallVectorImpl<UnwrappedLine>::iterator 1359 I = PreprocessorDirectives.begin(), 1360 E = PreprocessorDirectives.end(); 1361 I != E; ++I) { 1362 CurrentLines->push_back(*I); 1363 } 1364 PreprocessorDirectives.clear(); 1365 } 1366 } 1367 1368 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1369 1370 bool UnwrappedLineParser::isOnNewLine(const FormatToken& FormatTok) { 1371 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1372 FormatTok.NewlinesBefore > 0; 1373 } 1374 1375 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1376 bool JustComments = Line->Tokens.empty(); 1377 for (SmallVectorImpl<FormatToken *>::const_iterator 1378 I = CommentsBeforeNextToken.begin(), 1379 E = CommentsBeforeNextToken.end(); 1380 I != E; ++I) { 1381 if (isOnNewLine(**I) && JustComments) { 1382 addUnwrappedLine(); 1383 } 1384 pushToken(*I); 1385 } 1386 if (NewlineBeforeNext && JustComments) { 1387 addUnwrappedLine(); 1388 } 1389 CommentsBeforeNextToken.clear(); 1390 } 1391 1392 void UnwrappedLineParser::nextToken() { 1393 if (eof()) 1394 return; 1395 flushComments(isOnNewLine(*FormatTok)); 1396 pushToken(FormatTok); 1397 readToken(); 1398 } 1399 1400 void UnwrappedLineParser::readToken() { 1401 bool CommentsInCurrentLine = true; 1402 do { 1403 FormatTok = Tokens->getNextToken(); 1404 assert(FormatTok); 1405 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1406 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1407 // If there is an unfinished unwrapped line, we flush the preprocessor 1408 // directives only after that unwrapped line was finished later. 1409 bool SwitchToPreprocessorLines = 1410 !Line->Tokens.empty() && CurrentLines == &Lines; 1411 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1412 // Comments stored before the preprocessor directive need to be output 1413 // before the preprocessor directive, at the same level as the 1414 // preprocessor directive, as we consider them to apply to the directive. 1415 flushComments(isOnNewLine(*FormatTok)); 1416 parsePPDirective(); 1417 } 1418 while (FormatTok->Type == TT_ConflictStart || 1419 FormatTok->Type == TT_ConflictEnd || 1420 FormatTok->Type == TT_ConflictAlternative) { 1421 if (FormatTok->Type == TT_ConflictStart) { 1422 conditionalCompilationStart(/*Unreachable=*/false); 1423 } else if (FormatTok->Type == TT_ConflictAlternative) { 1424 conditionalCompilationAlternative(); 1425 } else if(FormatTok->Type == TT_ConflictEnd) { 1426 conditionalCompilationEnd(); 1427 } 1428 FormatTok = Tokens->getNextToken(); 1429 FormatTok->MustBreakBefore = true; 1430 } 1431 1432 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1433 !Line->InPPDirective) { 1434 continue; 1435 } 1436 1437 if (!FormatTok->Tok.is(tok::comment)) 1438 return; 1439 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1440 CommentsInCurrentLine = false; 1441 } 1442 if (CommentsInCurrentLine) { 1443 pushToken(FormatTok); 1444 } else { 1445 CommentsBeforeNextToken.push_back(FormatTok); 1446 } 1447 } while (!eof()); 1448 } 1449 1450 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1451 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1452 if (MustBreakBeforeNextToken) { 1453 Line->Tokens.back().Tok->MustBreakBefore = true; 1454 MustBreakBeforeNextToken = false; 1455 } 1456 } 1457 1458 } // end namespace format 1459 } // end namespace clang 1460