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), OriginalLines(Parser.CurrentLines) { 126 if (SwitchToPreprocessorLines) 127 Parser.CurrentLines = &Parser.PreprocessorDirectives; 128 else if (!Parser.Line->Tokens.empty()) 129 Parser.CurrentLines = &Parser.Line->Tokens.back().Children; 130 PreBlockLine = std::move(Parser.Line); 131 Parser.Line = llvm::make_unique<UnwrappedLine>(); 132 Parser.Line->Level = PreBlockLine->Level; 133 Parser.Line->InPPDirective = PreBlockLine->InPPDirective; 134 } 135 136 ~ScopedLineState() { 137 if (!Parser.Line->Tokens.empty()) { 138 Parser.addUnwrappedLine(); 139 } 140 assert(Parser.Line->Tokens.empty()); 141 Parser.Line = std::move(PreBlockLine); 142 if (Parser.CurrentLines == &Parser.PreprocessorDirectives) 143 Parser.MustBreakBeforeNextToken = true; 144 Parser.CurrentLines = OriginalLines; 145 } 146 147 private: 148 UnwrappedLineParser &Parser; 149 150 std::unique_ptr<UnwrappedLine> PreBlockLine; 151 SmallVectorImpl<UnwrappedLine> *OriginalLines; 152 }; 153 154 class CompoundStatementIndenter { 155 public: 156 CompoundStatementIndenter(UnwrappedLineParser *Parser, 157 const FormatStyle &Style, unsigned &LineLevel) 158 : LineLevel(LineLevel), OldLineLevel(LineLevel) { 159 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman) { 160 Parser->addUnwrappedLine(); 161 } else if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 162 Parser->addUnwrappedLine(); 163 ++LineLevel; 164 } 165 } 166 ~CompoundStatementIndenter() { LineLevel = OldLineLevel; } 167 168 private: 169 unsigned &LineLevel; 170 unsigned OldLineLevel; 171 }; 172 173 namespace { 174 175 class IndexedTokenSource : public FormatTokenSource { 176 public: 177 IndexedTokenSource(ArrayRef<FormatToken *> Tokens) 178 : Tokens(Tokens), Position(-1) {} 179 180 FormatToken *getNextToken() override { 181 ++Position; 182 return Tokens[Position]; 183 } 184 185 unsigned getPosition() override { 186 assert(Position >= 0); 187 return Position; 188 } 189 190 FormatToken *setPosition(unsigned P) override { 191 Position = P; 192 return Tokens[Position]; 193 } 194 195 void reset() { Position = -1; } 196 197 private: 198 ArrayRef<FormatToken *> Tokens; 199 int Position; 200 }; 201 202 } // end anonymous namespace 203 204 UnwrappedLineParser::UnwrappedLineParser(const FormatStyle &Style, 205 const AdditionalKeywords &Keywords, 206 ArrayRef<FormatToken *> Tokens, 207 UnwrappedLineConsumer &Callback) 208 : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), 209 CurrentLines(&Lines), StructuralError(false), Style(Style), 210 Keywords(Keywords), Tokens(nullptr), Callback(Callback), 211 AllTokens(Tokens), 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, tok::ellipsis) || 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 static bool ShouldBreakBeforeBrace(const FormatStyle &Style, 439 const FormatToken &InitialToken) { 440 switch (Style.BreakBeforeBraces) { 441 case FormatStyle::BS_Linux: 442 return InitialToken.isOneOf(tok::kw_namespace, tok::kw_class); 443 case FormatStyle::BS_Allman: 444 case FormatStyle::BS_GNU: 445 return true; 446 default: 447 return false; 448 } 449 } 450 451 void UnwrappedLineParser::parseChildBlock() { 452 FormatTok->BlockKind = BK_Block; 453 nextToken(); 454 { 455 bool GoogScope = 456 Style.Language == FormatStyle::LK_JavaScript && IsGoogScope(*Line); 457 ScopedLineState LineState(*this); 458 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 459 /*MustBeDeclaration=*/false); 460 Line->Level += GoogScope ? 0 : 1; 461 parseLevel(/*HasOpeningBrace=*/true); 462 Line->Level -= GoogScope ? 0 : 1; 463 } 464 nextToken(); 465 } 466 467 void UnwrappedLineParser::parsePPDirective() { 468 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 469 ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError); 470 nextToken(); 471 472 if (!FormatTok->Tok.getIdentifierInfo()) { 473 parsePPUnknown(); 474 return; 475 } 476 477 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 478 case tok::pp_define: 479 parsePPDefine(); 480 return; 481 case tok::pp_if: 482 parsePPIf(/*IfDef=*/false); 483 break; 484 case tok::pp_ifdef: 485 case tok::pp_ifndef: 486 parsePPIf(/*IfDef=*/true); 487 break; 488 case tok::pp_else: 489 parsePPElse(); 490 break; 491 case tok::pp_elif: 492 parsePPElIf(); 493 break; 494 case tok::pp_endif: 495 parsePPEndIf(); 496 break; 497 default: 498 parsePPUnknown(); 499 break; 500 } 501 } 502 503 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 504 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 505 PPStack.push_back(PP_Unreachable); 506 else 507 PPStack.push_back(PP_Conditional); 508 } 509 510 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 511 ++PPBranchLevel; 512 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 513 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 514 PPLevelBranchIndex.push_back(0); 515 PPLevelBranchCount.push_back(0); 516 } 517 PPChainBranchIndex.push(0); 518 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 519 conditionalCompilationCondition(Unreachable || Skip); 520 } 521 522 void UnwrappedLineParser::conditionalCompilationAlternative() { 523 if (!PPStack.empty()) 524 PPStack.pop_back(); 525 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 526 if (!PPChainBranchIndex.empty()) 527 ++PPChainBranchIndex.top(); 528 conditionalCompilationCondition( 529 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 530 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 531 } 532 533 void UnwrappedLineParser::conditionalCompilationEnd() { 534 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 535 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 536 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 537 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 538 } 539 } 540 // Guard against #endif's without #if. 541 if (PPBranchLevel > 0) 542 --PPBranchLevel; 543 if (!PPChainBranchIndex.empty()) 544 PPChainBranchIndex.pop(); 545 if (!PPStack.empty()) 546 PPStack.pop_back(); 547 } 548 549 void UnwrappedLineParser::parsePPIf(bool IfDef) { 550 nextToken(); 551 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 552 StringRef(FormatTok->Tok.getLiteralData(), 553 FormatTok->Tok.getLength()) == "0") || 554 FormatTok->Tok.is(tok::kw_false); 555 conditionalCompilationStart(!IfDef && IsLiteralFalse); 556 parsePPUnknown(); 557 } 558 559 void UnwrappedLineParser::parsePPElse() { 560 conditionalCompilationAlternative(); 561 parsePPUnknown(); 562 } 563 564 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 565 566 void UnwrappedLineParser::parsePPEndIf() { 567 conditionalCompilationEnd(); 568 parsePPUnknown(); 569 } 570 571 void UnwrappedLineParser::parsePPDefine() { 572 nextToken(); 573 574 if (FormatTok->Tok.getKind() != tok::identifier) { 575 parsePPUnknown(); 576 return; 577 } 578 nextToken(); 579 if (FormatTok->Tok.getKind() == tok::l_paren && 580 FormatTok->WhitespaceRange.getBegin() == 581 FormatTok->WhitespaceRange.getEnd()) { 582 parseParens(); 583 } 584 addUnwrappedLine(); 585 Line->Level = 1; 586 587 // Errors during a preprocessor directive can only affect the layout of the 588 // preprocessor directive, and thus we ignore them. An alternative approach 589 // would be to use the same approach we use on the file level (no 590 // re-indentation if there was a structural error) within the macro 591 // definition. 592 parseFile(); 593 } 594 595 void UnwrappedLineParser::parsePPUnknown() { 596 do { 597 nextToken(); 598 } while (!eof()); 599 addUnwrappedLine(); 600 } 601 602 // Here we blacklist certain tokens that are not usually the first token in an 603 // unwrapped line. This is used in attempt to distinguish macro calls without 604 // trailing semicolons from other constructs split to several lines. 605 bool tokenCanStartNewLine(clang::Token Tok) { 606 // Semicolon can be a null-statement, l_square can be a start of a macro or 607 // a C++11 attribute, but this doesn't seem to be common. 608 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 609 Tok.isNot(tok::l_square) && 610 // Tokens that can only be used as binary operators and a part of 611 // overloaded operator names. 612 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 613 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 614 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 615 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 616 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 617 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 618 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 619 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 620 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 621 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 622 Tok.isNot(tok::lesslessequal) && 623 // Colon is used in labels, base class lists, initializer lists, 624 // range-based for loops, ternary operator, but should never be the 625 // first token in an unwrapped line. 626 Tok.isNot(tok::colon) && 627 // 'noexcept' is a trailing annotation. 628 Tok.isNot(tok::kw_noexcept); 629 } 630 631 void UnwrappedLineParser::parseStructuralElement() { 632 assert(!FormatTok->Tok.is(tok::l_brace)); 633 switch (FormatTok->Tok.getKind()) { 634 case tok::at: 635 nextToken(); 636 if (FormatTok->Tok.is(tok::l_brace)) { 637 parseBracedList(); 638 break; 639 } 640 switch (FormatTok->Tok.getObjCKeywordID()) { 641 case tok::objc_public: 642 case tok::objc_protected: 643 case tok::objc_package: 644 case tok::objc_private: 645 return parseAccessSpecifier(); 646 case tok::objc_interface: 647 case tok::objc_implementation: 648 return parseObjCInterfaceOrImplementation(); 649 case tok::objc_protocol: 650 return parseObjCProtocol(); 651 case tok::objc_end: 652 return; // Handled by the caller. 653 case tok::objc_optional: 654 case tok::objc_required: 655 nextToken(); 656 addUnwrappedLine(); 657 return; 658 default: 659 break; 660 } 661 break; 662 case tok::kw_asm: 663 FormatTok->Finalized = true; 664 nextToken(); 665 if (FormatTok->is(tok::l_brace)) { 666 while (FormatTok && FormatTok->isNot(tok::eof)) { 667 FormatTok->Finalized = true; 668 if (FormatTok->is(tok::r_brace)) { 669 nextToken(); 670 break; 671 } 672 nextToken(); 673 } 674 } 675 break; 676 case tok::kw_namespace: 677 parseNamespace(); 678 return; 679 case tok::kw_inline: 680 nextToken(); 681 if (FormatTok->Tok.is(tok::kw_namespace)) { 682 parseNamespace(); 683 return; 684 } 685 break; 686 case tok::kw_public: 687 case tok::kw_protected: 688 case tok::kw_private: 689 if (Style.Language == FormatStyle::LK_Java) 690 nextToken(); 691 else 692 parseAccessSpecifier(); 693 return; 694 case tok::kw_if: 695 parseIfThenElse(); 696 return; 697 case tok::kw_for: 698 case tok::kw_while: 699 parseForOrWhileLoop(); 700 return; 701 case tok::kw_do: 702 parseDoWhile(); 703 return; 704 case tok::kw_switch: 705 parseSwitch(); 706 return; 707 case tok::kw_default: 708 nextToken(); 709 parseLabel(); 710 return; 711 case tok::kw_case: 712 parseCaseLabel(); 713 return; 714 case tok::kw_try: 715 parseTryCatch(); 716 return; 717 case tok::kw_extern: 718 nextToken(); 719 if (FormatTok->Tok.is(tok::string_literal)) { 720 nextToken(); 721 if (FormatTok->Tok.is(tok::l_brace)) { 722 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 723 addUnwrappedLine(); 724 return; 725 } 726 } 727 break; 728 case tok::identifier: 729 if (FormatTok->IsForEachMacro) { 730 parseForOrWhileLoop(); 731 return; 732 } 733 // In all other cases, parse the declaration. 734 break; 735 default: 736 break; 737 } 738 do { 739 switch (FormatTok->Tok.getKind()) { 740 case tok::at: 741 nextToken(); 742 if (FormatTok->Tok.is(tok::l_brace)) 743 parseBracedList(); 744 break; 745 case tok::kw_enum: 746 parseEnum(); 747 break; 748 case tok::kw_typedef: 749 nextToken(); 750 if (FormatTok->is(Keywords.kw_NS_ENUM)) 751 parseEnum(); 752 break; 753 case tok::kw_struct: 754 case tok::kw_union: 755 case tok::kw_class: 756 parseRecord(); 757 // A record declaration or definition is always the start of a structural 758 // element. 759 break; 760 case tok::semi: 761 nextToken(); 762 addUnwrappedLine(); 763 return; 764 case tok::r_brace: 765 addUnwrappedLine(); 766 return; 767 case tok::l_paren: 768 parseParens(); 769 break; 770 case tok::caret: 771 nextToken(); 772 if (FormatTok->Tok.isAnyIdentifier() || 773 FormatTok->isSimpleTypeSpecifier()) 774 nextToken(); 775 if (FormatTok->is(tok::l_paren)) 776 parseParens(); 777 if (FormatTok->is(tok::l_brace)) 778 parseChildBlock(); 779 break; 780 case tok::l_brace: 781 if (!tryToParseBracedList()) { 782 // A block outside of parentheses must be the last part of a 783 // structural element. 784 // FIXME: Figure out cases where this is not true, and add projections 785 // for them (the one we know is missing are lambdas). 786 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 787 addUnwrappedLine(); 788 FormatTok->Type = TT_FunctionLBrace; 789 parseBlock(/*MustBeDeclaration=*/false); 790 addUnwrappedLine(); 791 return; 792 } 793 // Otherwise this was a braced init list, and the structural 794 // element continues. 795 break; 796 case tok::kw_try: 797 // We arrive here when parsing function-try blocks. 798 parseTryCatch(); 799 return; 800 case tok::identifier: { 801 StringRef Text = FormatTok->TokenText; 802 // Parse function literal unless 'function' is the first token in a line 803 // in which case this should be treated as a free-standing function. 804 if (Style.Language == FormatStyle::LK_JavaScript && Text == "function" && 805 Line->Tokens.size() > 0) { 806 tryToParseJSFunction(); 807 break; 808 } 809 nextToken(); 810 if (Line->Tokens.size() == 1) { 811 if (FormatTok->Tok.is(tok::colon)) { 812 parseLabel(); 813 return; 814 } 815 // Recognize function-like macro usages without trailing semicolon as 816 // well as free-standing macrose like Q_OBJECT. 817 bool FunctionLike = FormatTok->is(tok::l_paren); 818 if (FunctionLike) 819 parseParens(); 820 if (FormatTok->NewlinesBefore > 0 && 821 (Text.size() >= 5 || FunctionLike) && 822 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 823 addUnwrappedLine(); 824 return; 825 } 826 } 827 break; 828 } 829 case tok::equal: 830 nextToken(); 831 if (FormatTok->Tok.is(tok::l_brace)) { 832 parseBracedList(); 833 } 834 break; 835 case tok::l_square: 836 parseSquare(); 837 break; 838 default: 839 nextToken(); 840 break; 841 } 842 } while (!eof()); 843 } 844 845 bool UnwrappedLineParser::tryToParseLambda() { 846 // FIXME: This is a dirty way to access the previous token. Find a better 847 // solution. 848 if (!Line->Tokens.empty() && 849 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator, 850 tok::kw_new, tok::kw_delete) || 851 Line->Tokens.back().Tok->closesScope() || 852 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 853 nextToken(); 854 return false; 855 } 856 assert(FormatTok->is(tok::l_square)); 857 FormatToken &LSquare = *FormatTok; 858 if (!tryToParseLambdaIntroducer()) 859 return false; 860 861 while (FormatTok->isNot(tok::l_brace)) { 862 if (FormatTok->isSimpleTypeSpecifier()) { 863 nextToken(); 864 continue; 865 } 866 switch (FormatTok->Tok.getKind()) { 867 case tok::l_brace: 868 break; 869 case tok::l_paren: 870 parseParens(); 871 break; 872 case tok::less: 873 case tok::greater: 874 case tok::identifier: 875 case tok::coloncolon: 876 case tok::kw_mutable: 877 nextToken(); 878 break; 879 case tok::arrow: 880 FormatTok->Type = TT_TrailingReturnArrow; 881 nextToken(); 882 break; 883 default: 884 return true; 885 } 886 } 887 LSquare.Type = TT_LambdaLSquare; 888 parseChildBlock(); 889 return true; 890 } 891 892 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 893 nextToken(); 894 if (FormatTok->is(tok::equal)) { 895 nextToken(); 896 if (FormatTok->is(tok::r_square)) { 897 nextToken(); 898 return true; 899 } 900 if (FormatTok->isNot(tok::comma)) 901 return false; 902 nextToken(); 903 } else if (FormatTok->is(tok::amp)) { 904 nextToken(); 905 if (FormatTok->is(tok::r_square)) { 906 nextToken(); 907 return true; 908 } 909 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 910 return false; 911 } 912 if (FormatTok->is(tok::comma)) 913 nextToken(); 914 } else if (FormatTok->is(tok::r_square)) { 915 nextToken(); 916 return true; 917 } 918 do { 919 if (FormatTok->is(tok::amp)) 920 nextToken(); 921 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 922 return false; 923 nextToken(); 924 if (FormatTok->is(tok::ellipsis)) 925 nextToken(); 926 if (FormatTok->is(tok::comma)) { 927 nextToken(); 928 } else if (FormatTok->is(tok::r_square)) { 929 nextToken(); 930 return true; 931 } else { 932 return false; 933 } 934 } while (!eof()); 935 return false; 936 } 937 938 void UnwrappedLineParser::tryToParseJSFunction() { 939 nextToken(); 940 941 // Consume function name. 942 if (FormatTok->is(tok::identifier)) 943 nextToken(); 944 945 if (FormatTok->isNot(tok::l_paren)) 946 return; 947 nextToken(); 948 while (FormatTok->isNot(tok::l_brace)) { 949 // Err on the side of caution in order to avoid consuming the full file in 950 // case of incomplete code. 951 if (!FormatTok->isOneOf(tok::identifier, tok::comma, tok::r_paren, 952 tok::comment)) 953 return; 954 nextToken(); 955 } 956 parseChildBlock(); 957 } 958 959 bool UnwrappedLineParser::tryToParseBracedList() { 960 if (FormatTok->BlockKind == BK_Unknown) 961 calculateBraceTypes(); 962 assert(FormatTok->BlockKind != BK_Unknown); 963 if (FormatTok->BlockKind == BK_Block) 964 return false; 965 parseBracedList(); 966 return true; 967 } 968 969 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 970 bool HasError = false; 971 nextToken(); 972 973 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 974 // replace this by using parseAssigmentExpression() inside. 975 do { 976 if (Style.Language == FormatStyle::LK_JavaScript && 977 FormatTok->is(Keywords.kw_function)) { 978 tryToParseJSFunction(); 979 continue; 980 } 981 switch (FormatTok->Tok.getKind()) { 982 case tok::caret: 983 nextToken(); 984 if (FormatTok->is(tok::l_brace)) { 985 parseChildBlock(); 986 } 987 break; 988 case tok::l_square: 989 tryToParseLambda(); 990 break; 991 case tok::l_brace: 992 // Assume there are no blocks inside a braced init list apart 993 // from the ones we explicitly parse out (like lambdas). 994 FormatTok->BlockKind = BK_BracedInit; 995 parseBracedList(); 996 break; 997 case tok::r_brace: 998 nextToken(); 999 return !HasError; 1000 case tok::semi: 1001 HasError = true; 1002 if (!ContinueOnSemicolons) 1003 return !HasError; 1004 nextToken(); 1005 break; 1006 case tok::comma: 1007 nextToken(); 1008 break; 1009 default: 1010 nextToken(); 1011 break; 1012 } 1013 } while (!eof()); 1014 return false; 1015 } 1016 1017 void UnwrappedLineParser::parseParens() { 1018 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1019 nextToken(); 1020 do { 1021 switch (FormatTok->Tok.getKind()) { 1022 case tok::l_paren: 1023 parseParens(); 1024 break; 1025 case tok::r_paren: 1026 nextToken(); 1027 return; 1028 case tok::r_brace: 1029 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1030 return; 1031 case tok::l_square: 1032 tryToParseLambda(); 1033 break; 1034 case tok::l_brace: { 1035 if (!tryToParseBracedList()) { 1036 parseChildBlock(); 1037 } 1038 break; 1039 } 1040 case tok::at: 1041 nextToken(); 1042 if (FormatTok->Tok.is(tok::l_brace)) 1043 parseBracedList(); 1044 break; 1045 case tok::identifier: 1046 if (Style.Language == FormatStyle::LK_JavaScript && 1047 FormatTok->is(Keywords.kw_function)) 1048 tryToParseJSFunction(); 1049 else 1050 nextToken(); 1051 break; 1052 default: 1053 nextToken(); 1054 break; 1055 } 1056 } while (!eof()); 1057 } 1058 1059 void UnwrappedLineParser::parseSquare() { 1060 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1061 if (tryToParseLambda()) 1062 return; 1063 do { 1064 switch (FormatTok->Tok.getKind()) { 1065 case tok::l_paren: 1066 parseParens(); 1067 break; 1068 case tok::r_square: 1069 nextToken(); 1070 return; 1071 case tok::r_brace: 1072 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1073 return; 1074 case tok::l_square: 1075 parseSquare(); 1076 break; 1077 case tok::l_brace: { 1078 if (!tryToParseBracedList()) { 1079 parseChildBlock(); 1080 } 1081 break; 1082 } 1083 case tok::at: 1084 nextToken(); 1085 if (FormatTok->Tok.is(tok::l_brace)) 1086 parseBracedList(); 1087 break; 1088 default: 1089 nextToken(); 1090 break; 1091 } 1092 } while (!eof()); 1093 } 1094 1095 void UnwrappedLineParser::parseIfThenElse() { 1096 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1097 nextToken(); 1098 if (FormatTok->Tok.is(tok::l_paren)) 1099 parseParens(); 1100 bool NeedsUnwrappedLine = false; 1101 if (FormatTok->Tok.is(tok::l_brace)) { 1102 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1103 parseBlock(/*MustBeDeclaration=*/false); 1104 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1105 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1106 addUnwrappedLine(); 1107 } else { 1108 NeedsUnwrappedLine = true; 1109 } 1110 } else { 1111 addUnwrappedLine(); 1112 ++Line->Level; 1113 parseStructuralElement(); 1114 --Line->Level; 1115 } 1116 if (FormatTok->Tok.is(tok::kw_else)) { 1117 if (Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) 1118 addUnwrappedLine(); 1119 nextToken(); 1120 if (FormatTok->Tok.is(tok::l_brace)) { 1121 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1122 parseBlock(/*MustBeDeclaration=*/false); 1123 addUnwrappedLine(); 1124 } else if (FormatTok->Tok.is(tok::kw_if)) { 1125 parseIfThenElse(); 1126 } else { 1127 addUnwrappedLine(); 1128 ++Line->Level; 1129 parseStructuralElement(); 1130 --Line->Level; 1131 } 1132 } else if (NeedsUnwrappedLine) { 1133 addUnwrappedLine(); 1134 } 1135 } 1136 1137 void UnwrappedLineParser::parseTryCatch() { 1138 assert(FormatTok->is(tok::kw_try) && "'try' expected"); 1139 nextToken(); 1140 bool NeedsUnwrappedLine = false; 1141 if (FormatTok->is(tok::colon)) { 1142 // We are in a function try block, what comes is an initializer list. 1143 nextToken(); 1144 while (FormatTok->is(tok::identifier)) { 1145 nextToken(); 1146 if (FormatTok->is(tok::l_paren)) 1147 parseParens(); 1148 else 1149 StructuralError = true; 1150 if (FormatTok->is(tok::comma)) 1151 nextToken(); 1152 } 1153 } 1154 if (FormatTok->is(tok::l_brace)) { 1155 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1156 parseBlock(/*MustBeDeclaration=*/false); 1157 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1158 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1159 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1160 addUnwrappedLine(); 1161 } else { 1162 NeedsUnwrappedLine = true; 1163 } 1164 } else if (!FormatTok->is(tok::kw_catch)) { 1165 // The C++ standard requires a compound-statement after a try. 1166 // If there's none, we try to assume there's a structuralElement 1167 // and try to continue. 1168 StructuralError = true; 1169 addUnwrappedLine(); 1170 ++Line->Level; 1171 parseStructuralElement(); 1172 --Line->Level; 1173 } 1174 while (FormatTok->is(tok::kw_catch) || 1175 ((Style.Language == FormatStyle::LK_Java || 1176 Style.Language == FormatStyle::LK_JavaScript) && 1177 FormatTok->is(Keywords.kw_finally))) { 1178 nextToken(); 1179 while (FormatTok->isNot(tok::l_brace)) { 1180 if (FormatTok->is(tok::l_paren)) { 1181 parseParens(); 1182 continue; 1183 } 1184 if (FormatTok->isOneOf(tok::semi, tok::r_brace)) 1185 return; 1186 nextToken(); 1187 } 1188 NeedsUnwrappedLine = false; 1189 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1190 parseBlock(/*MustBeDeclaration=*/false); 1191 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1192 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1193 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1194 addUnwrappedLine(); 1195 } else { 1196 NeedsUnwrappedLine = true; 1197 } 1198 } 1199 if (NeedsUnwrappedLine) { 1200 addUnwrappedLine(); 1201 } 1202 } 1203 1204 void UnwrappedLineParser::parseNamespace() { 1205 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1206 1207 const FormatToken &InitialToken = *FormatTok; 1208 nextToken(); 1209 if (FormatTok->Tok.is(tok::identifier)) 1210 nextToken(); 1211 if (FormatTok->Tok.is(tok::l_brace)) { 1212 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1213 addUnwrappedLine(); 1214 1215 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1216 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1217 DeclarationScopeStack.size() > 1); 1218 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1219 // Munch the semicolon after a namespace. This is more common than one would 1220 // think. Puttin the semicolon into its own line is very ugly. 1221 if (FormatTok->Tok.is(tok::semi)) 1222 nextToken(); 1223 addUnwrappedLine(); 1224 } 1225 // FIXME: Add error handling. 1226 } 1227 1228 void UnwrappedLineParser::parseForOrWhileLoop() { 1229 assert((FormatTok->Tok.is(tok::kw_for) || FormatTok->Tok.is(tok::kw_while) || 1230 FormatTok->IsForEachMacro) && 1231 "'for', 'while' or foreach macro expected"); 1232 nextToken(); 1233 if (FormatTok->Tok.is(tok::l_paren)) 1234 parseParens(); 1235 if (FormatTok->Tok.is(tok::l_brace)) { 1236 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1237 parseBlock(/*MustBeDeclaration=*/false); 1238 addUnwrappedLine(); 1239 } else { 1240 addUnwrappedLine(); 1241 ++Line->Level; 1242 parseStructuralElement(); 1243 --Line->Level; 1244 } 1245 } 1246 1247 void UnwrappedLineParser::parseDoWhile() { 1248 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1249 nextToken(); 1250 if (FormatTok->Tok.is(tok::l_brace)) { 1251 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1252 parseBlock(/*MustBeDeclaration=*/false); 1253 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1254 addUnwrappedLine(); 1255 } else { 1256 addUnwrappedLine(); 1257 ++Line->Level; 1258 parseStructuralElement(); 1259 --Line->Level; 1260 } 1261 1262 // FIXME: Add error handling. 1263 if (!FormatTok->Tok.is(tok::kw_while)) { 1264 addUnwrappedLine(); 1265 return; 1266 } 1267 1268 nextToken(); 1269 parseStructuralElement(); 1270 } 1271 1272 void UnwrappedLineParser::parseLabel() { 1273 nextToken(); 1274 unsigned OldLineLevel = Line->Level; 1275 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1276 --Line->Level; 1277 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1278 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1279 parseBlock(/*MustBeDeclaration=*/false); 1280 if (FormatTok->Tok.is(tok::kw_break)) { 1281 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1282 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1283 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1284 addUnwrappedLine(); 1285 } 1286 parseStructuralElement(); 1287 } 1288 addUnwrappedLine(); 1289 } else { 1290 addUnwrappedLine(); 1291 } 1292 Line->Level = OldLineLevel; 1293 } 1294 1295 void UnwrappedLineParser::parseCaseLabel() { 1296 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1297 // FIXME: fix handling of complex expressions here. 1298 do { 1299 nextToken(); 1300 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1301 parseLabel(); 1302 } 1303 1304 void UnwrappedLineParser::parseSwitch() { 1305 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1306 nextToken(); 1307 if (FormatTok->Tok.is(tok::l_paren)) 1308 parseParens(); 1309 if (FormatTok->Tok.is(tok::l_brace)) { 1310 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1311 parseBlock(/*MustBeDeclaration=*/false); 1312 addUnwrappedLine(); 1313 } else { 1314 addUnwrappedLine(); 1315 ++Line->Level; 1316 parseStructuralElement(); 1317 --Line->Level; 1318 } 1319 } 1320 1321 void UnwrappedLineParser::parseAccessSpecifier() { 1322 nextToken(); 1323 // Understand Qt's slots. 1324 if (FormatTok->is(tok::identifier) && 1325 (FormatTok->TokenText == "slots" || FormatTok->TokenText == "Q_SLOTS")) 1326 nextToken(); 1327 // Otherwise, we don't know what it is, and we'd better keep the next token. 1328 if (FormatTok->Tok.is(tok::colon)) 1329 nextToken(); 1330 addUnwrappedLine(); 1331 } 1332 1333 void UnwrappedLineParser::parseEnum() { 1334 if (FormatTok->Tok.is(tok::kw_enum)) { 1335 // Won't be 'enum' for NS_ENUMs. 1336 nextToken(); 1337 } 1338 // Eat up enum class ... 1339 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1340 nextToken(); 1341 while (FormatTok->Tok.getIdentifierInfo() || 1342 FormatTok->isOneOf(tok::colon, tok::coloncolon)) { 1343 nextToken(); 1344 // We can have macros or attributes in between 'enum' and the enum name. 1345 if (FormatTok->Tok.is(tok::l_paren)) { 1346 parseParens(); 1347 } 1348 if (FormatTok->Tok.is(tok::identifier)) 1349 nextToken(); 1350 } 1351 if (FormatTok->Tok.is(tok::l_brace)) { 1352 FormatTok->BlockKind = BK_Block; 1353 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1354 if (HasError) { 1355 if (FormatTok->is(tok::semi)) 1356 nextToken(); 1357 addUnwrappedLine(); 1358 } 1359 } 1360 // We fall through to parsing a structural element afterwards, so that in 1361 // enum A {} n, m; 1362 // "} n, m;" will end up in one unwrapped line. 1363 // This does not apply for Java. 1364 if (Style.Language == FormatStyle::LK_Java) 1365 addUnwrappedLine(); 1366 } 1367 1368 void UnwrappedLineParser::parseRecord() { 1369 const FormatToken &InitialToken = *FormatTok; 1370 nextToken(); 1371 if (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::kw___attribute, 1372 tok::kw___declspec, tok::kw_alignas)) { 1373 nextToken(); 1374 // We can have macros or attributes in between 'class' and the class name. 1375 if (FormatTok->Tok.is(tok::l_paren)) { 1376 parseParens(); 1377 } 1378 // The actual identifier can be a nested name specifier, and in macros 1379 // it is often token-pasted. 1380 while (FormatTok->is(tok::identifier) || FormatTok->is(tok::coloncolon) || 1381 FormatTok->is(tok::hashhash) || 1382 (Style.Language == FormatStyle::LK_Java && 1383 FormatTok->isOneOf(tok::period, tok::comma))) 1384 nextToken(); 1385 1386 // Note that parsing away template declarations here leads to incorrectly 1387 // accepting function declarations as record declarations. 1388 // In general, we cannot solve this problem. Consider: 1389 // class A<int> B() {} 1390 // which can be a function definition or a class definition when B() is a 1391 // macro. If we find enough real-world cases where this is a problem, we 1392 // can parse for the 'template' keyword in the beginning of the statement, 1393 // and thus rule out the record production in case there is no template 1394 // (this would still leave us with an ambiguity between template function 1395 // and class declarations). 1396 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1397 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1398 if (FormatTok->Tok.is(tok::semi)) 1399 return; 1400 nextToken(); 1401 } 1402 } 1403 } 1404 if (FormatTok->Tok.is(tok::l_brace)) { 1405 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1406 addUnwrappedLine(); 1407 1408 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1409 /*MunchSemi=*/false); 1410 } 1411 // We fall through to parsing a structural element afterwards, so 1412 // class A {} n, m; 1413 // will end up in one unwrapped line. 1414 // This does not apply for Java. 1415 if (Style.Language == FormatStyle::LK_Java) 1416 addUnwrappedLine(); 1417 } 1418 1419 void UnwrappedLineParser::parseObjCProtocolList() { 1420 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1421 do 1422 nextToken(); 1423 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1424 nextToken(); // Skip '>'. 1425 } 1426 1427 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1428 do { 1429 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1430 nextToken(); 1431 addUnwrappedLine(); 1432 break; 1433 } 1434 if (FormatTok->is(tok::l_brace)) { 1435 parseBlock(/*MustBeDeclaration=*/false); 1436 // In ObjC interfaces, nothing should be following the "}". 1437 addUnwrappedLine(); 1438 } else if (FormatTok->is(tok::r_brace)) { 1439 // Ignore stray "}". parseStructuralElement doesn't consume them. 1440 nextToken(); 1441 addUnwrappedLine(); 1442 } else { 1443 parseStructuralElement(); 1444 } 1445 } while (!eof()); 1446 } 1447 1448 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1449 nextToken(); 1450 nextToken(); // interface name 1451 1452 // @interface can be followed by either a base class, or a category. 1453 if (FormatTok->Tok.is(tok::colon)) { 1454 nextToken(); 1455 nextToken(); // base class name 1456 } else if (FormatTok->Tok.is(tok::l_paren)) 1457 // Skip category, if present. 1458 parseParens(); 1459 1460 if (FormatTok->Tok.is(tok::less)) 1461 parseObjCProtocolList(); 1462 1463 if (FormatTok->Tok.is(tok::l_brace)) { 1464 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1465 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1466 addUnwrappedLine(); 1467 parseBlock(/*MustBeDeclaration=*/true); 1468 } 1469 1470 // With instance variables, this puts '}' on its own line. Without instance 1471 // variables, this ends the @interface line. 1472 addUnwrappedLine(); 1473 1474 parseObjCUntilAtEnd(); 1475 } 1476 1477 void UnwrappedLineParser::parseObjCProtocol() { 1478 nextToken(); 1479 nextToken(); // protocol name 1480 1481 if (FormatTok->Tok.is(tok::less)) 1482 parseObjCProtocolList(); 1483 1484 // Check for protocol declaration. 1485 if (FormatTok->Tok.is(tok::semi)) { 1486 nextToken(); 1487 return addUnwrappedLine(); 1488 } 1489 1490 addUnwrappedLine(); 1491 parseObjCUntilAtEnd(); 1492 } 1493 1494 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1495 StringRef Prefix = "") { 1496 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1497 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1498 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1499 E = Line.Tokens.end(); 1500 I != E; ++I) { 1501 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1502 } 1503 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1504 E = Line.Tokens.end(); 1505 I != E; ++I) { 1506 const UnwrappedLineNode &Node = *I; 1507 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1508 I = Node.Children.begin(), 1509 E = Node.Children.end(); 1510 I != E; ++I) { 1511 printDebugInfo(*I, "\nChild: "); 1512 } 1513 } 1514 llvm::dbgs() << "\n"; 1515 } 1516 1517 void UnwrappedLineParser::addUnwrappedLine() { 1518 if (Line->Tokens.empty()) 1519 return; 1520 DEBUG({ 1521 if (CurrentLines == &Lines) 1522 printDebugInfo(*Line); 1523 }); 1524 CurrentLines->push_back(*Line); 1525 Line->Tokens.clear(); 1526 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1527 for (SmallVectorImpl<UnwrappedLine>::iterator 1528 I = PreprocessorDirectives.begin(), 1529 E = PreprocessorDirectives.end(); 1530 I != E; ++I) { 1531 CurrentLines->push_back(*I); 1532 } 1533 PreprocessorDirectives.clear(); 1534 } 1535 } 1536 1537 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1538 1539 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1540 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1541 FormatTok.NewlinesBefore > 0; 1542 } 1543 1544 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1545 bool JustComments = Line->Tokens.empty(); 1546 for (SmallVectorImpl<FormatToken *>::const_iterator 1547 I = CommentsBeforeNextToken.begin(), 1548 E = CommentsBeforeNextToken.end(); 1549 I != E; ++I) { 1550 if (isOnNewLine(**I) && JustComments) { 1551 addUnwrappedLine(); 1552 } 1553 pushToken(*I); 1554 } 1555 if (NewlineBeforeNext && JustComments) { 1556 addUnwrappedLine(); 1557 } 1558 CommentsBeforeNextToken.clear(); 1559 } 1560 1561 void UnwrappedLineParser::nextToken() { 1562 if (eof()) 1563 return; 1564 flushComments(isOnNewLine(*FormatTok)); 1565 pushToken(FormatTok); 1566 readToken(); 1567 } 1568 1569 void UnwrappedLineParser::readToken() { 1570 bool CommentsInCurrentLine = true; 1571 do { 1572 FormatTok = Tokens->getNextToken(); 1573 assert(FormatTok); 1574 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1575 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1576 // If there is an unfinished unwrapped line, we flush the preprocessor 1577 // directives only after that unwrapped line was finished later. 1578 bool SwitchToPreprocessorLines = 1579 !Line->Tokens.empty() && CurrentLines == &Lines; 1580 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1581 // Comments stored before the preprocessor directive need to be output 1582 // before the preprocessor directive, at the same level as the 1583 // preprocessor directive, as we consider them to apply to the directive. 1584 flushComments(isOnNewLine(*FormatTok)); 1585 parsePPDirective(); 1586 } 1587 while (FormatTok->Type == TT_ConflictStart || 1588 FormatTok->Type == TT_ConflictEnd || 1589 FormatTok->Type == TT_ConflictAlternative) { 1590 if (FormatTok->Type == TT_ConflictStart) { 1591 conditionalCompilationStart(/*Unreachable=*/false); 1592 } else if (FormatTok->Type == TT_ConflictAlternative) { 1593 conditionalCompilationAlternative(); 1594 } else if (FormatTok->Type == TT_ConflictEnd) { 1595 conditionalCompilationEnd(); 1596 } 1597 FormatTok = Tokens->getNextToken(); 1598 FormatTok->MustBreakBefore = true; 1599 } 1600 1601 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1602 !Line->InPPDirective) { 1603 continue; 1604 } 1605 1606 if (!FormatTok->Tok.is(tok::comment)) 1607 return; 1608 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1609 CommentsInCurrentLine = false; 1610 } 1611 if (CommentsInCurrentLine) { 1612 pushToken(FormatTok); 1613 } else { 1614 CommentsBeforeNextToken.push_back(FormatTok); 1615 } 1616 } while (!eof()); 1617 } 1618 1619 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1620 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1621 if (MustBreakBeforeNextToken) { 1622 Line->Tokens.back().Tok->MustBreakBefore = true; 1623 MustBreakBeforeNextToken = false; 1624 } 1625 } 1626 1627 } // end namespace format 1628 } // end namespace clang 1629