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(nullptr) { 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() { LineLevel = OldLineLevel; } 168 169 private: 170 unsigned &LineLevel; 171 unsigned OldLineLevel; 172 }; 173 174 namespace { 175 176 class IndexedTokenSource : public FormatTokenSource { 177 public: 178 IndexedTokenSource(ArrayRef<FormatToken *> Tokens) 179 : Tokens(Tokens), Position(-1) {} 180 181 FormatToken *getNextToken() override { 182 ++Position; 183 return Tokens[Position]; 184 } 185 186 unsigned getPosition() override { 187 assert(Position >= 0); 188 return Position; 189 } 190 191 FormatToken *setPosition(unsigned P) override { 192 Position = P; 193 return Tokens[Position]; 194 } 195 196 void reset() { Position = -1; } 197 198 private: 199 ArrayRef<FormatToken *> Tokens; 200 int Position; 201 }; 202 203 } // end anonymous namespace 204 205 UnwrappedLineParser::UnwrappedLineParser(const FormatStyle &Style, 206 ArrayRef<FormatToken *> Tokens, 207 UnwrappedLineConsumer &Callback) 208 : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), 209 CurrentLines(&Lines), StructuralError(false), Style(Style), 210 Tokens(nullptr), Callback(Callback), AllTokens(Tokens), 211 PPBranchLevel(-1) {} 212 213 void UnwrappedLineParser::reset() { 214 PPBranchLevel = -1; 215 Line.reset(new UnwrappedLine); 216 CommentsBeforeNextToken.clear(); 217 FormatTok = nullptr; 218 MustBreakBeforeNextToken = false; 219 PreprocessorDirectives.clear(); 220 CurrentLines = &Lines; 221 DeclarationScopeStack.clear(); 222 StructuralError = false; 223 PPStack.clear(); 224 } 225 226 bool UnwrappedLineParser::parse() { 227 IndexedTokenSource TokenSource(AllTokens); 228 do { 229 DEBUG(llvm::dbgs() << "----\n"); 230 reset(); 231 Tokens = &TokenSource; 232 TokenSource.reset(); 233 234 readToken(); 235 parseFile(); 236 // Create line with eof token. 237 pushToken(FormatTok); 238 addUnwrappedLine(); 239 240 for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(), 241 E = Lines.end(); 242 I != E; ++I) { 243 Callback.consumeUnwrappedLine(*I); 244 } 245 Callback.finishRun(); 246 Lines.clear(); 247 while (!PPLevelBranchIndex.empty() && 248 PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) { 249 PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1); 250 PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1); 251 } 252 if (!PPLevelBranchIndex.empty()) { 253 ++PPLevelBranchIndex.back(); 254 assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size()); 255 assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back()); 256 } 257 } while (!PPLevelBranchIndex.empty()); 258 259 return StructuralError; 260 } 261 262 void UnwrappedLineParser::parseFile() { 263 ScopedDeclarationState DeclarationState( 264 *Line, DeclarationScopeStack, 265 /*MustBeDeclaration=*/ !Line->InPPDirective); 266 parseLevel(/*HasOpeningBrace=*/false); 267 // Make sure to format the remaining tokens. 268 flushComments(true); 269 addUnwrappedLine(); 270 } 271 272 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) { 273 bool SwitchLabelEncountered = false; 274 do { 275 switch (FormatTok->Tok.getKind()) { 276 case tok::comment: 277 nextToken(); 278 addUnwrappedLine(); 279 break; 280 case tok::l_brace: 281 // FIXME: Add parameter whether this can happen - if this happens, we must 282 // be in a non-declaration context. 283 parseBlock(/*MustBeDeclaration=*/false); 284 addUnwrappedLine(); 285 break; 286 case tok::r_brace: 287 if (HasOpeningBrace) 288 return; 289 StructuralError = true; 290 nextToken(); 291 addUnwrappedLine(); 292 break; 293 case tok::kw_default: 294 case tok::kw_case: 295 if (!SwitchLabelEncountered && 296 (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1))) 297 ++Line->Level; 298 SwitchLabelEncountered = true; 299 parseStructuralElement(); 300 break; 301 default: 302 parseStructuralElement(); 303 break; 304 } 305 } while (!eof()); 306 } 307 308 void UnwrappedLineParser::calculateBraceTypes() { 309 // We'll parse forward through the tokens until we hit 310 // a closing brace or eof - note that getNextToken() will 311 // parse macros, so this will magically work inside macro 312 // definitions, too. 313 unsigned StoredPosition = Tokens->getPosition(); 314 unsigned Position = StoredPosition; 315 FormatToken *Tok = FormatTok; 316 // Keep a stack of positions of lbrace tokens. We will 317 // update information about whether an lbrace starts a 318 // braced init list or a different block during the loop. 319 SmallVector<FormatToken *, 8> LBraceStack; 320 assert(Tok->Tok.is(tok::l_brace)); 321 do { 322 // Get next none-comment token. 323 FormatToken *NextTok; 324 unsigned ReadTokens = 0; 325 do { 326 NextTok = Tokens->getNextToken(); 327 ++ReadTokens; 328 } while (NextTok->is(tok::comment)); 329 330 switch (Tok->Tok.getKind()) { 331 case tok::l_brace: 332 LBraceStack.push_back(Tok); 333 break; 334 case tok::r_brace: 335 if (!LBraceStack.empty()) { 336 if (LBraceStack.back()->BlockKind == BK_Unknown) { 337 bool ProbablyBracedList = false; 338 if (Style.Language == FormatStyle::LK_Proto) { 339 ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); 340 } else { 341 // Using OriginalColumn to distinguish between ObjC methods and 342 // binary operators is a bit hacky. 343 bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) && 344 NextTok->OriginalColumn == 0; 345 346 // If there is a comma, semicolon or right paren after the closing 347 // brace, we assume this is a braced initializer list. Note that 348 // regardless how we mark inner braces here, we will overwrite the 349 // BlockKind later if we parse a braced list (where all blocks 350 // inside are by default braced lists), or when we explicitly detect 351 // blocks (for example while parsing lambdas). 352 // 353 // We exclude + and - as they can be ObjC visibility modifiers. 354 ProbablyBracedList = 355 NextTok->isOneOf(tok::comma, tok::semi, tok::period, tok::colon, 356 tok::r_paren, tok::r_square, tok::l_brace, 357 tok::l_paren) || 358 (NextTok->isBinaryOperator() && !NextIsObjCMethod); 359 } 360 if (ProbablyBracedList) { 361 Tok->BlockKind = BK_BracedInit; 362 LBraceStack.back()->BlockKind = BK_BracedInit; 363 } else { 364 Tok->BlockKind = BK_Block; 365 LBraceStack.back()->BlockKind = BK_Block; 366 } 367 } 368 LBraceStack.pop_back(); 369 } 370 break; 371 case tok::at: 372 case tok::semi: 373 case tok::kw_if: 374 case tok::kw_while: 375 case tok::kw_for: 376 case tok::kw_switch: 377 case tok::kw_try: 378 if (!LBraceStack.empty()) 379 LBraceStack.back()->BlockKind = BK_Block; 380 break; 381 default: 382 break; 383 } 384 Tok = NextTok; 385 Position += ReadTokens; 386 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 387 // Assume other blocks for all unclosed opening braces. 388 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 389 if (LBraceStack[i]->BlockKind == BK_Unknown) 390 LBraceStack[i]->BlockKind = BK_Block; 391 } 392 393 FormatTok = Tokens->setPosition(StoredPosition); 394 } 395 396 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 397 bool MunchSemi) { 398 assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected"); 399 unsigned InitialLevel = Line->Level; 400 nextToken(); 401 402 addUnwrappedLine(); 403 404 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 405 MustBeDeclaration); 406 if (AddLevel) 407 ++Line->Level; 408 parseLevel(/*HasOpeningBrace=*/true); 409 410 if (!FormatTok->Tok.is(tok::r_brace)) { 411 Line->Level = InitialLevel; 412 StructuralError = true; 413 return; 414 } 415 416 nextToken(); // Munch the closing brace. 417 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 418 nextToken(); 419 Line->Level = InitialLevel; 420 } 421 422 static bool IsGoogScope(const UnwrappedLine &Line) { 423 if (Line.Tokens.size() < 4) 424 return false; 425 auto I = Line.Tokens.begin(); 426 if (I->Tok->TokenText != "goog") 427 return false; 428 ++I; 429 if (I->Tok->isNot(tok::period)) 430 return false; 431 ++I; 432 if (I->Tok->TokenText != "scope") 433 return false; 434 ++I; 435 return I->Tok->is(tok::l_paren); 436 } 437 438 void UnwrappedLineParser::parseChildBlock() { 439 FormatTok->BlockKind = BK_Block; 440 nextToken(); 441 { 442 bool GoogScope = 443 Style.Language == FormatStyle::LK_JavaScript && IsGoogScope(*Line); 444 ScopedLineState LineState(*this); 445 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 446 /*MustBeDeclaration=*/false); 447 Line->Level += GoogScope ? 0 : 1; 448 parseLevel(/*HasOpeningBrace=*/true); 449 Line->Level -= GoogScope ? 0 : 1; 450 } 451 nextToken(); 452 } 453 454 void UnwrappedLineParser::parsePPDirective() { 455 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 456 ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError); 457 nextToken(); 458 459 if (!FormatTok->Tok.getIdentifierInfo()) { 460 parsePPUnknown(); 461 return; 462 } 463 464 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 465 case tok::pp_define: 466 parsePPDefine(); 467 return; 468 case tok::pp_if: 469 parsePPIf(/*IfDef=*/false); 470 break; 471 case tok::pp_ifdef: 472 case tok::pp_ifndef: 473 parsePPIf(/*IfDef=*/true); 474 break; 475 case tok::pp_else: 476 parsePPElse(); 477 break; 478 case tok::pp_elif: 479 parsePPElIf(); 480 break; 481 case tok::pp_endif: 482 parsePPEndIf(); 483 break; 484 default: 485 parsePPUnknown(); 486 break; 487 } 488 } 489 490 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 491 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 492 PPStack.push_back(PP_Unreachable); 493 else 494 PPStack.push_back(PP_Conditional); 495 } 496 497 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 498 ++PPBranchLevel; 499 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 500 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 501 PPLevelBranchIndex.push_back(0); 502 PPLevelBranchCount.push_back(0); 503 } 504 PPChainBranchIndex.push(0); 505 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 506 conditionalCompilationCondition(Unreachable || Skip); 507 } 508 509 void UnwrappedLineParser::conditionalCompilationAlternative() { 510 if (!PPStack.empty()) 511 PPStack.pop_back(); 512 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 513 if (!PPChainBranchIndex.empty()) 514 ++PPChainBranchIndex.top(); 515 conditionalCompilationCondition( 516 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 517 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 518 } 519 520 void UnwrappedLineParser::conditionalCompilationEnd() { 521 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 522 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 523 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 524 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 525 } 526 } 527 // Guard against #endif's without #if. 528 if (PPBranchLevel > 0) 529 --PPBranchLevel; 530 if (!PPChainBranchIndex.empty()) 531 PPChainBranchIndex.pop(); 532 if (!PPStack.empty()) 533 PPStack.pop_back(); 534 } 535 536 void UnwrappedLineParser::parsePPIf(bool IfDef) { 537 nextToken(); 538 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 539 StringRef(FormatTok->Tok.getLiteralData(), 540 FormatTok->Tok.getLength()) == "0") || 541 FormatTok->Tok.is(tok::kw_false); 542 conditionalCompilationStart(!IfDef && IsLiteralFalse); 543 parsePPUnknown(); 544 } 545 546 void UnwrappedLineParser::parsePPElse() { 547 conditionalCompilationAlternative(); 548 parsePPUnknown(); 549 } 550 551 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 552 553 void UnwrappedLineParser::parsePPEndIf() { 554 conditionalCompilationEnd(); 555 parsePPUnknown(); 556 } 557 558 void UnwrappedLineParser::parsePPDefine() { 559 nextToken(); 560 561 if (FormatTok->Tok.getKind() != tok::identifier) { 562 parsePPUnknown(); 563 return; 564 } 565 nextToken(); 566 if (FormatTok->Tok.getKind() == tok::l_paren && 567 FormatTok->WhitespaceRange.getBegin() == 568 FormatTok->WhitespaceRange.getEnd()) { 569 parseParens(); 570 } 571 addUnwrappedLine(); 572 Line->Level = 1; 573 574 // Errors during a preprocessor directive can only affect the layout of the 575 // preprocessor directive, and thus we ignore them. An alternative approach 576 // would be to use the same approach we use on the file level (no 577 // re-indentation if there was a structural error) within the macro 578 // definition. 579 parseFile(); 580 } 581 582 void UnwrappedLineParser::parsePPUnknown() { 583 do { 584 nextToken(); 585 } while (!eof()); 586 addUnwrappedLine(); 587 } 588 589 // Here we blacklist certain tokens that are not usually the first token in an 590 // unwrapped line. This is used in attempt to distinguish macro calls without 591 // trailing semicolons from other constructs split to several lines. 592 bool tokenCanStartNewLine(clang::Token Tok) { 593 // Semicolon can be a null-statement, l_square can be a start of a macro or 594 // a C++11 attribute, but this doesn't seem to be common. 595 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 596 Tok.isNot(tok::l_square) && 597 // Tokens that can only be used as binary operators and a part of 598 // overloaded operator names. 599 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 600 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 601 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 602 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 603 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 604 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 605 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 606 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 607 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 608 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 609 Tok.isNot(tok::lesslessequal) && 610 // Colon is used in labels, base class lists, initializer lists, 611 // range-based for loops, ternary operator, but should never be the 612 // first token in an unwrapped line. 613 Tok.isNot(tok::colon) && 614 // 'noexcept' is a trailing annotation. 615 Tok.isNot(tok::kw_noexcept); 616 } 617 618 void UnwrappedLineParser::parseStructuralElement() { 619 assert(!FormatTok->Tok.is(tok::l_brace)); 620 switch (FormatTok->Tok.getKind()) { 621 case tok::at: 622 nextToken(); 623 if (FormatTok->Tok.is(tok::l_brace)) { 624 parseBracedList(); 625 break; 626 } 627 switch (FormatTok->Tok.getObjCKeywordID()) { 628 case tok::objc_public: 629 case tok::objc_protected: 630 case tok::objc_package: 631 case tok::objc_private: 632 return parseAccessSpecifier(); 633 case tok::objc_interface: 634 case tok::objc_implementation: 635 return parseObjCInterfaceOrImplementation(); 636 case tok::objc_protocol: 637 return parseObjCProtocol(); 638 case tok::objc_end: 639 return; // Handled by the caller. 640 case tok::objc_optional: 641 case tok::objc_required: 642 nextToken(); 643 addUnwrappedLine(); 644 return; 645 default: 646 break; 647 } 648 break; 649 case tok::kw_namespace: 650 parseNamespace(); 651 return; 652 case tok::kw_inline: 653 nextToken(); 654 if (FormatTok->Tok.is(tok::kw_namespace)) { 655 parseNamespace(); 656 return; 657 } 658 break; 659 case tok::kw_public: 660 case tok::kw_protected: 661 case tok::kw_private: 662 parseAccessSpecifier(); 663 return; 664 case tok::kw_if: 665 parseIfThenElse(); 666 return; 667 case tok::kw_for: 668 case tok::kw_while: 669 parseForOrWhileLoop(); 670 return; 671 case tok::kw_do: 672 parseDoWhile(); 673 return; 674 case tok::kw_switch: 675 parseSwitch(); 676 return; 677 case tok::kw_default: 678 nextToken(); 679 parseLabel(); 680 return; 681 case tok::kw_case: 682 parseCaseLabel(); 683 return; 684 case tok::kw_try: 685 parseTryCatch(); 686 return; 687 case tok::kw_extern: 688 nextToken(); 689 if (FormatTok->Tok.is(tok::string_literal)) { 690 nextToken(); 691 if (FormatTok->Tok.is(tok::l_brace)) { 692 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 693 addUnwrappedLine(); 694 return; 695 } 696 } 697 break; 698 case tok::identifier: 699 if (FormatTok->IsForEachMacro) { 700 parseForOrWhileLoop(); 701 return; 702 } 703 // In all other cases, parse the declaration. 704 break; 705 default: 706 break; 707 } 708 do { 709 switch (FormatTok->Tok.getKind()) { 710 case tok::at: 711 nextToken(); 712 if (FormatTok->Tok.is(tok::l_brace)) 713 parseBracedList(); 714 break; 715 case tok::kw_enum: 716 parseEnum(); 717 break; 718 case tok::kw_typedef: 719 nextToken(); 720 // FIXME: Use the IdentifierTable instead. 721 if (FormatTok->TokenText == "NS_ENUM") 722 parseEnum(); 723 break; 724 case tok::kw_struct: 725 case tok::kw_union: 726 case tok::kw_class: 727 parseRecord(); 728 // A record declaration or definition is always the start of a structural 729 // element. 730 break; 731 case tok::semi: 732 nextToken(); 733 addUnwrappedLine(); 734 return; 735 case tok::r_brace: 736 addUnwrappedLine(); 737 return; 738 case tok::l_paren: 739 parseParens(); 740 break; 741 case tok::caret: 742 nextToken(); 743 if (FormatTok->Tok.isAnyIdentifier() || 744 FormatTok->isSimpleTypeSpecifier()) 745 nextToken(); 746 if (FormatTok->is(tok::l_paren)) 747 parseParens(); 748 if (FormatTok->is(tok::l_brace)) 749 parseChildBlock(); 750 break; 751 case tok::l_brace: 752 if (!tryToParseBracedList()) { 753 // A block outside of parentheses must be the last part of a 754 // structural element. 755 // FIXME: Figure out cases where this is not true, and add projections 756 // for them (the one we know is missing are lambdas). 757 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 758 addUnwrappedLine(); 759 FormatTok->Type = TT_FunctionLBrace; 760 parseBlock(/*MustBeDeclaration=*/false); 761 addUnwrappedLine(); 762 return; 763 } 764 // Otherwise this was a braced init list, and the structural 765 // element continues. 766 break; 767 case tok::kw_try: 768 // We arrive here when parsing function-try blocks. 769 parseTryCatch(); 770 return; 771 case tok::identifier: { 772 StringRef Text = FormatTok->TokenText; 773 if (Style.Language == FormatStyle::LK_JavaScript && Text == "function") { 774 tryToParseJSFunction(); 775 break; 776 } 777 nextToken(); 778 if (Line->Tokens.size() == 1) { 779 if (FormatTok->Tok.is(tok::colon)) { 780 parseLabel(); 781 return; 782 } 783 // Recognize function-like macro usages without trailing semicolon. 784 if (FormatTok->Tok.is(tok::l_paren)) { 785 parseParens(); 786 if (FormatTok->NewlinesBefore > 0 && 787 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 788 addUnwrappedLine(); 789 return; 790 } 791 } else if (FormatTok->HasUnescapedNewline && Text.size() >= 5 && 792 Text == Text.upper()) { 793 // Recognize free-standing macros like Q_OBJECT. 794 addUnwrappedLine(); 795 return; 796 } 797 } 798 break; 799 } 800 case tok::equal: 801 nextToken(); 802 if (FormatTok->Tok.is(tok::l_brace)) { 803 parseBracedList(); 804 } 805 break; 806 case tok::l_square: 807 parseSquare(); 808 break; 809 default: 810 nextToken(); 811 break; 812 } 813 } while (!eof()); 814 } 815 816 bool UnwrappedLineParser::tryToParseLambda() { 817 // FIXME: This is a dirty way to access the previous token. Find a better 818 // solution. 819 if (!Line->Tokens.empty() && 820 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator) || 821 Line->Tokens.back().Tok->closesScope() || 822 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 823 nextToken(); 824 return false; 825 } 826 assert(FormatTok->is(tok::l_square)); 827 FormatToken &LSquare = *FormatTok; 828 if (!tryToParseLambdaIntroducer()) 829 return false; 830 831 while (FormatTok->isNot(tok::l_brace)) { 832 if (FormatTok->isSimpleTypeSpecifier()) { 833 nextToken(); 834 continue; 835 } 836 switch (FormatTok->Tok.getKind()) { 837 case tok::l_brace: 838 break; 839 case tok::l_paren: 840 parseParens(); 841 break; 842 case tok::less: 843 case tok::greater: 844 case tok::identifier: 845 case tok::coloncolon: 846 case tok::kw_mutable: 847 nextToken(); 848 break; 849 case tok::arrow: 850 FormatTok->Type = TT_TrailingReturnArrow; 851 nextToken(); 852 break; 853 default: 854 return true; 855 } 856 } 857 LSquare.Type = TT_LambdaLSquare; 858 parseChildBlock(); 859 return true; 860 } 861 862 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 863 nextToken(); 864 if (FormatTok->is(tok::equal)) { 865 nextToken(); 866 if (FormatTok->is(tok::r_square)) { 867 nextToken(); 868 return true; 869 } 870 if (FormatTok->isNot(tok::comma)) 871 return false; 872 nextToken(); 873 } else if (FormatTok->is(tok::amp)) { 874 nextToken(); 875 if (FormatTok->is(tok::r_square)) { 876 nextToken(); 877 return true; 878 } 879 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 880 return false; 881 } 882 if (FormatTok->is(tok::comma)) 883 nextToken(); 884 } else if (FormatTok->is(tok::r_square)) { 885 nextToken(); 886 return true; 887 } 888 do { 889 if (FormatTok->is(tok::amp)) 890 nextToken(); 891 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 892 return false; 893 nextToken(); 894 if (FormatTok->is(tok::comma)) { 895 nextToken(); 896 } else if (FormatTok->is(tok::r_square)) { 897 nextToken(); 898 return true; 899 } else { 900 return false; 901 } 902 } while (!eof()); 903 return false; 904 } 905 906 void UnwrappedLineParser::tryToParseJSFunction() { 907 nextToken(); 908 if (FormatTok->isNot(tok::l_paren)) 909 return; 910 nextToken(); 911 while (FormatTok->isNot(tok::l_brace)) { 912 // Err on the side of caution in order to avoid consuming the full file in 913 // case of incomplete code. 914 if (!FormatTok->isOneOf(tok::identifier, tok::comma, tok::r_paren, 915 tok::comment)) 916 return; 917 nextToken(); 918 } 919 parseChildBlock(); 920 } 921 922 bool UnwrappedLineParser::tryToParseBracedList() { 923 if (FormatTok->BlockKind == BK_Unknown) 924 calculateBraceTypes(); 925 assert(FormatTok->BlockKind != BK_Unknown); 926 if (FormatTok->BlockKind == BK_Block) 927 return false; 928 parseBracedList(); 929 return true; 930 } 931 932 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 933 bool HasError = false; 934 nextToken(); 935 936 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 937 // replace this by using parseAssigmentExpression() inside. 938 do { 939 if (Style.Language == FormatStyle::LK_JavaScript && 940 FormatTok->TokenText == "function") { 941 tryToParseJSFunction(); 942 continue; 943 } 944 switch (FormatTok->Tok.getKind()) { 945 case tok::caret: 946 nextToken(); 947 if (FormatTok->is(tok::l_brace)) { 948 parseChildBlock(); 949 } 950 break; 951 case tok::l_square: 952 tryToParseLambda(); 953 break; 954 case tok::l_brace: 955 // Assume there are no blocks inside a braced init list apart 956 // from the ones we explicitly parse out (like lambdas). 957 FormatTok->BlockKind = BK_BracedInit; 958 parseBracedList(); 959 break; 960 case tok::r_brace: 961 nextToken(); 962 return !HasError; 963 case tok::semi: 964 HasError = true; 965 if (!ContinueOnSemicolons) 966 return !HasError; 967 nextToken(); 968 break; 969 case tok::comma: 970 nextToken(); 971 break; 972 default: 973 nextToken(); 974 break; 975 } 976 } while (!eof()); 977 return false; 978 } 979 980 void UnwrappedLineParser::parseParens() { 981 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 982 nextToken(); 983 do { 984 switch (FormatTok->Tok.getKind()) { 985 case tok::l_paren: 986 parseParens(); 987 break; 988 case tok::r_paren: 989 nextToken(); 990 return; 991 case tok::r_brace: 992 // A "}" inside parenthesis is an error if there wasn't a matching "{". 993 return; 994 case tok::l_square: 995 tryToParseLambda(); 996 break; 997 case tok::l_brace: { 998 if (!tryToParseBracedList()) { 999 parseChildBlock(); 1000 } 1001 break; 1002 } 1003 case tok::at: 1004 nextToken(); 1005 if (FormatTok->Tok.is(tok::l_brace)) 1006 parseBracedList(); 1007 break; 1008 default: 1009 nextToken(); 1010 break; 1011 } 1012 } while (!eof()); 1013 } 1014 1015 void UnwrappedLineParser::parseSquare() { 1016 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1017 if (tryToParseLambda()) 1018 return; 1019 do { 1020 switch (FormatTok->Tok.getKind()) { 1021 case tok::l_paren: 1022 parseParens(); 1023 break; 1024 case tok::r_square: 1025 nextToken(); 1026 return; 1027 case tok::r_brace: 1028 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1029 return; 1030 case tok::l_square: 1031 parseSquare(); 1032 break; 1033 case tok::l_brace: { 1034 if (!tryToParseBracedList()) { 1035 parseChildBlock(); 1036 } 1037 break; 1038 } 1039 case tok::at: 1040 nextToken(); 1041 if (FormatTok->Tok.is(tok::l_brace)) 1042 parseBracedList(); 1043 break; 1044 default: 1045 nextToken(); 1046 break; 1047 } 1048 } while (!eof()); 1049 } 1050 1051 void UnwrappedLineParser::parseIfThenElse() { 1052 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1053 nextToken(); 1054 if (FormatTok->Tok.is(tok::l_paren)) 1055 parseParens(); 1056 bool NeedsUnwrappedLine = false; 1057 if (FormatTok->Tok.is(tok::l_brace)) { 1058 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1059 parseBlock(/*MustBeDeclaration=*/false); 1060 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1061 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1062 addUnwrappedLine(); 1063 } else { 1064 NeedsUnwrappedLine = true; 1065 } 1066 } else { 1067 addUnwrappedLine(); 1068 ++Line->Level; 1069 parseStructuralElement(); 1070 --Line->Level; 1071 } 1072 if (FormatTok->Tok.is(tok::kw_else)) { 1073 nextToken(); 1074 if (FormatTok->Tok.is(tok::l_brace)) { 1075 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1076 parseBlock(/*MustBeDeclaration=*/false); 1077 addUnwrappedLine(); 1078 } else if (FormatTok->Tok.is(tok::kw_if)) { 1079 parseIfThenElse(); 1080 } else { 1081 addUnwrappedLine(); 1082 ++Line->Level; 1083 parseStructuralElement(); 1084 --Line->Level; 1085 } 1086 } else if (NeedsUnwrappedLine) { 1087 addUnwrappedLine(); 1088 } 1089 } 1090 1091 void UnwrappedLineParser::parseTryCatch() { 1092 assert(FormatTok->is(tok::kw_try) && "'try' expected"); 1093 nextToken(); 1094 bool NeedsUnwrappedLine = false; 1095 if (FormatTok->is(tok::colon)) { 1096 // We are in a function try block, what comes is an initializer list. 1097 nextToken(); 1098 while (FormatTok->is(tok::identifier)) { 1099 nextToken(); 1100 if (FormatTok->is(tok::l_paren)) 1101 parseParens(); 1102 else 1103 StructuralError = true; 1104 if (FormatTok->is(tok::comma)) 1105 nextToken(); 1106 } 1107 } 1108 if (FormatTok->is(tok::l_brace)) { 1109 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1110 parseBlock(/*MustBeDeclaration=*/false); 1111 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1112 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1113 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1114 addUnwrappedLine(); 1115 } else { 1116 NeedsUnwrappedLine = true; 1117 } 1118 } else if (!FormatTok->is(tok::kw_catch)) { 1119 // The C++ standard requires a compound-statement after a try. 1120 // If there's none, we try to assume there's a structuralElement 1121 // and try to continue. 1122 StructuralError = true; 1123 addUnwrappedLine(); 1124 ++Line->Level; 1125 parseStructuralElement(); 1126 --Line->Level; 1127 } 1128 while (FormatTok->is(tok::kw_catch) || 1129 (Style.Language == FormatStyle::LK_JavaScript && 1130 FormatTok->TokenText == "finally")) { 1131 nextToken(); 1132 while (FormatTok->isNot(tok::l_brace)) { 1133 if (FormatTok->is(tok::l_paren)) { 1134 parseParens(); 1135 continue; 1136 } 1137 if (FormatTok->isOneOf(tok::semi, tok::r_brace)) 1138 return; 1139 nextToken(); 1140 } 1141 NeedsUnwrappedLine = false; 1142 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1143 parseBlock(/*MustBeDeclaration=*/false); 1144 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1145 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1146 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1147 addUnwrappedLine(); 1148 } else { 1149 NeedsUnwrappedLine = true; 1150 } 1151 } 1152 if (NeedsUnwrappedLine) { 1153 addUnwrappedLine(); 1154 } 1155 } 1156 1157 void UnwrappedLineParser::parseNamespace() { 1158 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1159 nextToken(); 1160 if (FormatTok->Tok.is(tok::identifier)) 1161 nextToken(); 1162 if (FormatTok->Tok.is(tok::l_brace)) { 1163 if (Style.BreakBeforeBraces == FormatStyle::BS_Linux || 1164 Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1165 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1166 addUnwrappedLine(); 1167 1168 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1169 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1170 DeclarationScopeStack.size() > 1); 1171 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1172 // Munch the semicolon after a namespace. This is more common than one would 1173 // think. Puttin the semicolon into its own line is very ugly. 1174 if (FormatTok->Tok.is(tok::semi)) 1175 nextToken(); 1176 addUnwrappedLine(); 1177 } 1178 // FIXME: Add error handling. 1179 } 1180 1181 void UnwrappedLineParser::parseForOrWhileLoop() { 1182 assert((FormatTok->Tok.is(tok::kw_for) || FormatTok->Tok.is(tok::kw_while) || 1183 FormatTok->IsForEachMacro) && 1184 "'for', 'while' or foreach macro expected"); 1185 nextToken(); 1186 if (FormatTok->Tok.is(tok::l_paren)) 1187 parseParens(); 1188 if (FormatTok->Tok.is(tok::l_brace)) { 1189 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1190 parseBlock(/*MustBeDeclaration=*/false); 1191 addUnwrappedLine(); 1192 } else { 1193 addUnwrappedLine(); 1194 ++Line->Level; 1195 parseStructuralElement(); 1196 --Line->Level; 1197 } 1198 } 1199 1200 void UnwrappedLineParser::parseDoWhile() { 1201 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1202 nextToken(); 1203 if (FormatTok->Tok.is(tok::l_brace)) { 1204 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1205 parseBlock(/*MustBeDeclaration=*/false); 1206 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1207 addUnwrappedLine(); 1208 } else { 1209 addUnwrappedLine(); 1210 ++Line->Level; 1211 parseStructuralElement(); 1212 --Line->Level; 1213 } 1214 1215 // FIXME: Add error handling. 1216 if (!FormatTok->Tok.is(tok::kw_while)) { 1217 addUnwrappedLine(); 1218 return; 1219 } 1220 1221 nextToken(); 1222 parseStructuralElement(); 1223 } 1224 1225 void UnwrappedLineParser::parseLabel() { 1226 nextToken(); 1227 unsigned OldLineLevel = Line->Level; 1228 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1229 --Line->Level; 1230 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1231 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1232 parseBlock(/*MustBeDeclaration=*/false); 1233 if (FormatTok->Tok.is(tok::kw_break)) { 1234 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1235 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1236 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1237 addUnwrappedLine(); 1238 } 1239 parseStructuralElement(); 1240 } 1241 addUnwrappedLine(); 1242 } else { 1243 addUnwrappedLine(); 1244 } 1245 Line->Level = OldLineLevel; 1246 } 1247 1248 void UnwrappedLineParser::parseCaseLabel() { 1249 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1250 // FIXME: fix handling of complex expressions here. 1251 do { 1252 nextToken(); 1253 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1254 parseLabel(); 1255 } 1256 1257 void UnwrappedLineParser::parseSwitch() { 1258 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1259 nextToken(); 1260 if (FormatTok->Tok.is(tok::l_paren)) 1261 parseParens(); 1262 if (FormatTok->Tok.is(tok::l_brace)) { 1263 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1264 parseBlock(/*MustBeDeclaration=*/false); 1265 addUnwrappedLine(); 1266 } else { 1267 addUnwrappedLine(); 1268 ++Line->Level; 1269 parseStructuralElement(); 1270 --Line->Level; 1271 } 1272 } 1273 1274 void UnwrappedLineParser::parseAccessSpecifier() { 1275 nextToken(); 1276 // Understand Qt's slots. 1277 if (FormatTok->is(tok::identifier) && 1278 (FormatTok->TokenText == "slots" || FormatTok->TokenText == "Q_SLOTS")) 1279 nextToken(); 1280 // Otherwise, we don't know what it is, and we'd better keep the next token. 1281 if (FormatTok->Tok.is(tok::colon)) 1282 nextToken(); 1283 addUnwrappedLine(); 1284 } 1285 1286 void UnwrappedLineParser::parseEnum() { 1287 if (FormatTok->Tok.is(tok::kw_enum)) { 1288 // Won't be 'enum' for NS_ENUMs. 1289 nextToken(); 1290 } 1291 // Eat up enum class ... 1292 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1293 nextToken(); 1294 while (FormatTok->Tok.getIdentifierInfo() || 1295 FormatTok->isOneOf(tok::colon, tok::coloncolon)) { 1296 nextToken(); 1297 // We can have macros or attributes in between 'enum' and the enum name. 1298 if (FormatTok->Tok.is(tok::l_paren)) { 1299 parseParens(); 1300 } 1301 if (FormatTok->Tok.is(tok::identifier)) 1302 nextToken(); 1303 } 1304 if (FormatTok->Tok.is(tok::l_brace)) { 1305 FormatTok->BlockKind = BK_Block; 1306 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1307 if (HasError) { 1308 if (FormatTok->is(tok::semi)) 1309 nextToken(); 1310 addUnwrappedLine(); 1311 } 1312 } 1313 // We fall through to parsing a structural element afterwards, so that in 1314 // enum A {} n, m; 1315 // "} n, m;" will end up in one unwrapped line. 1316 } 1317 1318 void UnwrappedLineParser::parseRecord() { 1319 nextToken(); 1320 if (FormatTok->Tok.is(tok::identifier) || 1321 FormatTok->Tok.is(tok::kw___attribute) || 1322 FormatTok->Tok.is(tok::kw___declspec) || 1323 FormatTok->Tok.is(tok::kw_alignas)) { 1324 nextToken(); 1325 // We can have macros or attributes in between 'class' and the class name. 1326 if (FormatTok->Tok.is(tok::l_paren)) { 1327 parseParens(); 1328 } 1329 // The actual identifier can be a nested name specifier, and in macros 1330 // it is often token-pasted. 1331 while (FormatTok->Tok.is(tok::identifier) || 1332 FormatTok->Tok.is(tok::coloncolon) || 1333 FormatTok->Tok.is(tok::hashhash)) 1334 nextToken(); 1335 1336 // Note that parsing away template declarations here leads to incorrectly 1337 // accepting function declarations as record declarations. 1338 // In general, we cannot solve this problem. Consider: 1339 // class A<int> B() {} 1340 // which can be a function definition or a class definition when B() is a 1341 // macro. If we find enough real-world cases where this is a problem, we 1342 // can parse for the 'template' keyword in the beginning of the statement, 1343 // and thus rule out the record production in case there is no template 1344 // (this would still leave us with an ambiguity between template function 1345 // and class declarations). 1346 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1347 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1348 if (FormatTok->Tok.is(tok::semi)) 1349 return; 1350 nextToken(); 1351 } 1352 } 1353 } 1354 if (FormatTok->Tok.is(tok::l_brace)) { 1355 if (Style.BreakBeforeBraces == FormatStyle::BS_Linux || 1356 Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1357 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1358 addUnwrappedLine(); 1359 1360 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1361 /*MunchSemi=*/false); 1362 } 1363 // We fall through to parsing a structural element afterwards, so 1364 // class A {} n, m; 1365 // will end up in one unwrapped line. 1366 } 1367 1368 void UnwrappedLineParser::parseObjCProtocolList() { 1369 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1370 do 1371 nextToken(); 1372 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1373 nextToken(); // Skip '>'. 1374 } 1375 1376 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1377 do { 1378 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1379 nextToken(); 1380 addUnwrappedLine(); 1381 break; 1382 } 1383 if (FormatTok->is(tok::l_brace)) { 1384 parseBlock(/*MustBeDeclaration=*/false); 1385 // In ObjC interfaces, nothing should be following the "}". 1386 addUnwrappedLine(); 1387 } else if (FormatTok->is(tok::r_brace)) { 1388 // Ignore stray "}". parseStructuralElement doesn't consume them. 1389 nextToken(); 1390 addUnwrappedLine(); 1391 } else { 1392 parseStructuralElement(); 1393 } 1394 } while (!eof()); 1395 } 1396 1397 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1398 nextToken(); 1399 nextToken(); // interface name 1400 1401 // @interface can be followed by either a base class, or a category. 1402 if (FormatTok->Tok.is(tok::colon)) { 1403 nextToken(); 1404 nextToken(); // base class name 1405 } else if (FormatTok->Tok.is(tok::l_paren)) 1406 // Skip category, if present. 1407 parseParens(); 1408 1409 if (FormatTok->Tok.is(tok::less)) 1410 parseObjCProtocolList(); 1411 1412 if (FormatTok->Tok.is(tok::l_brace)) { 1413 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1414 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1415 addUnwrappedLine(); 1416 parseBlock(/*MustBeDeclaration=*/true); 1417 } 1418 1419 // With instance variables, this puts '}' on its own line. Without instance 1420 // variables, this ends the @interface line. 1421 addUnwrappedLine(); 1422 1423 parseObjCUntilAtEnd(); 1424 } 1425 1426 void UnwrappedLineParser::parseObjCProtocol() { 1427 nextToken(); 1428 nextToken(); // protocol name 1429 1430 if (FormatTok->Tok.is(tok::less)) 1431 parseObjCProtocolList(); 1432 1433 // Check for protocol declaration. 1434 if (FormatTok->Tok.is(tok::semi)) { 1435 nextToken(); 1436 return addUnwrappedLine(); 1437 } 1438 1439 addUnwrappedLine(); 1440 parseObjCUntilAtEnd(); 1441 } 1442 1443 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1444 StringRef Prefix = "") { 1445 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1446 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1447 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1448 E = Line.Tokens.end(); 1449 I != E; ++I) { 1450 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1451 } 1452 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1453 E = Line.Tokens.end(); 1454 I != E; ++I) { 1455 const UnwrappedLineNode &Node = *I; 1456 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1457 I = Node.Children.begin(), 1458 E = Node.Children.end(); 1459 I != E; ++I) { 1460 printDebugInfo(*I, "\nChild: "); 1461 } 1462 } 1463 llvm::dbgs() << "\n"; 1464 } 1465 1466 void UnwrappedLineParser::addUnwrappedLine() { 1467 if (Line->Tokens.empty()) 1468 return; 1469 DEBUG({ 1470 if (CurrentLines == &Lines) 1471 printDebugInfo(*Line); 1472 }); 1473 CurrentLines->push_back(*Line); 1474 Line->Tokens.clear(); 1475 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1476 for (SmallVectorImpl<UnwrappedLine>::iterator 1477 I = PreprocessorDirectives.begin(), 1478 E = PreprocessorDirectives.end(); 1479 I != E; ++I) { 1480 CurrentLines->push_back(*I); 1481 } 1482 PreprocessorDirectives.clear(); 1483 } 1484 } 1485 1486 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1487 1488 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1489 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1490 FormatTok.NewlinesBefore > 0; 1491 } 1492 1493 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1494 bool JustComments = Line->Tokens.empty(); 1495 for (SmallVectorImpl<FormatToken *>::const_iterator 1496 I = CommentsBeforeNextToken.begin(), 1497 E = CommentsBeforeNextToken.end(); 1498 I != E; ++I) { 1499 if (isOnNewLine(**I) && JustComments) { 1500 addUnwrappedLine(); 1501 } 1502 pushToken(*I); 1503 } 1504 if (NewlineBeforeNext && JustComments) { 1505 addUnwrappedLine(); 1506 } 1507 CommentsBeforeNextToken.clear(); 1508 } 1509 1510 void UnwrappedLineParser::nextToken() { 1511 if (eof()) 1512 return; 1513 flushComments(isOnNewLine(*FormatTok)); 1514 pushToken(FormatTok); 1515 readToken(); 1516 } 1517 1518 void UnwrappedLineParser::readToken() { 1519 bool CommentsInCurrentLine = true; 1520 do { 1521 FormatTok = Tokens->getNextToken(); 1522 assert(FormatTok); 1523 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1524 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1525 // If there is an unfinished unwrapped line, we flush the preprocessor 1526 // directives only after that unwrapped line was finished later. 1527 bool SwitchToPreprocessorLines = 1528 !Line->Tokens.empty() && CurrentLines == &Lines; 1529 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1530 // Comments stored before the preprocessor directive need to be output 1531 // before the preprocessor directive, at the same level as the 1532 // preprocessor directive, as we consider them to apply to the directive. 1533 flushComments(isOnNewLine(*FormatTok)); 1534 parsePPDirective(); 1535 } 1536 while (FormatTok->Type == TT_ConflictStart || 1537 FormatTok->Type == TT_ConflictEnd || 1538 FormatTok->Type == TT_ConflictAlternative) { 1539 if (FormatTok->Type == TT_ConflictStart) { 1540 conditionalCompilationStart(/*Unreachable=*/false); 1541 } else if (FormatTok->Type == TT_ConflictAlternative) { 1542 conditionalCompilationAlternative(); 1543 } else if (FormatTok->Type == TT_ConflictEnd) { 1544 conditionalCompilationEnd(); 1545 } 1546 FormatTok = Tokens->getNextToken(); 1547 FormatTok->MustBreakBefore = true; 1548 } 1549 1550 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1551 !Line->InPPDirective) { 1552 continue; 1553 } 1554 1555 if (!FormatTok->Tok.is(tok::comment)) 1556 return; 1557 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1558 CommentsInCurrentLine = false; 1559 } 1560 if (CommentsInCurrentLine) { 1561 pushToken(FormatTok); 1562 } else { 1563 CommentsBeforeNextToken.push_back(FormatTok); 1564 } 1565 } while (!eof()); 1566 } 1567 1568 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1569 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1570 if (MustBreakBeforeNextToken) { 1571 Line->Tokens.back().Tok->MustBreakBefore = true; 1572 MustBreakBeforeNextToken = false; 1573 } 1574 } 1575 1576 } // end namespace format 1577 } // end namespace clang 1578