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 ArrayRef<FormatToken *> Tokens, 206 UnwrappedLineConsumer &Callback) 207 : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), 208 CurrentLines(&Lines), StructuralError(false), Style(Style), 209 Tokens(nullptr), Callback(Callback), AllTokens(Tokens), 210 PPBranchLevel(-1) {} 211 212 void UnwrappedLineParser::reset() { 213 PPBranchLevel = -1; 214 Line.reset(new UnwrappedLine); 215 CommentsBeforeNextToken.clear(); 216 FormatTok = nullptr; 217 MustBreakBeforeNextToken = false; 218 PreprocessorDirectives.clear(); 219 CurrentLines = &Lines; 220 DeclarationScopeStack.clear(); 221 StructuralError = false; 222 PPStack.clear(); 223 } 224 225 bool UnwrappedLineParser::parse() { 226 IndexedTokenSource TokenSource(AllTokens); 227 do { 228 DEBUG(llvm::dbgs() << "----\n"); 229 reset(); 230 Tokens = &TokenSource; 231 TokenSource.reset(); 232 233 readToken(); 234 parseFile(); 235 // Create line with eof token. 236 pushToken(FormatTok); 237 addUnwrappedLine(); 238 239 for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(), 240 E = Lines.end(); 241 I != E; ++I) { 242 Callback.consumeUnwrappedLine(*I); 243 } 244 Callback.finishRun(); 245 Lines.clear(); 246 while (!PPLevelBranchIndex.empty() && 247 PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) { 248 PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1); 249 PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1); 250 } 251 if (!PPLevelBranchIndex.empty()) { 252 ++PPLevelBranchIndex.back(); 253 assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size()); 254 assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back()); 255 } 256 } while (!PPLevelBranchIndex.empty()); 257 258 return StructuralError; 259 } 260 261 void UnwrappedLineParser::parseFile() { 262 ScopedDeclarationState DeclarationState( 263 *Line, DeclarationScopeStack, 264 /*MustBeDeclaration=*/ !Line->InPPDirective); 265 parseLevel(/*HasOpeningBrace=*/false); 266 // Make sure to format the remaining tokens. 267 flushComments(true); 268 addUnwrappedLine(); 269 } 270 271 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) { 272 bool SwitchLabelEncountered = false; 273 do { 274 switch (FormatTok->Tok.getKind()) { 275 case tok::comment: 276 nextToken(); 277 addUnwrappedLine(); 278 break; 279 case tok::l_brace: 280 // FIXME: Add parameter whether this can happen - if this happens, we must 281 // be in a non-declaration context. 282 parseBlock(/*MustBeDeclaration=*/false); 283 addUnwrappedLine(); 284 break; 285 case tok::r_brace: 286 if (HasOpeningBrace) 287 return; 288 StructuralError = true; 289 nextToken(); 290 addUnwrappedLine(); 291 break; 292 case tok::kw_default: 293 case tok::kw_case: 294 if (!SwitchLabelEncountered && 295 (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1))) 296 ++Line->Level; 297 SwitchLabelEncountered = true; 298 parseStructuralElement(); 299 break; 300 default: 301 parseStructuralElement(); 302 break; 303 } 304 } while (!eof()); 305 } 306 307 void UnwrappedLineParser::calculateBraceTypes() { 308 // We'll parse forward through the tokens until we hit 309 // a closing brace or eof - note that getNextToken() will 310 // parse macros, so this will magically work inside macro 311 // definitions, too. 312 unsigned StoredPosition = Tokens->getPosition(); 313 unsigned Position = StoredPosition; 314 FormatToken *Tok = FormatTok; 315 // Keep a stack of positions of lbrace tokens. We will 316 // update information about whether an lbrace starts a 317 // braced init list or a different block during the loop. 318 SmallVector<FormatToken *, 8> LBraceStack; 319 assert(Tok->Tok.is(tok::l_brace)); 320 do { 321 // Get next none-comment token. 322 FormatToken *NextTok; 323 unsigned ReadTokens = 0; 324 do { 325 NextTok = Tokens->getNextToken(); 326 ++ReadTokens; 327 } while (NextTok->is(tok::comment)); 328 329 switch (Tok->Tok.getKind()) { 330 case tok::l_brace: 331 LBraceStack.push_back(Tok); 332 break; 333 case tok::r_brace: 334 if (!LBraceStack.empty()) { 335 if (LBraceStack.back()->BlockKind == BK_Unknown) { 336 bool ProbablyBracedList = false; 337 if (Style.Language == FormatStyle::LK_Proto) { 338 ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); 339 } else { 340 // Using OriginalColumn to distinguish between ObjC methods and 341 // binary operators is a bit hacky. 342 bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) && 343 NextTok->OriginalColumn == 0; 344 345 // If there is a comma, semicolon or right paren after the closing 346 // brace, we assume this is a braced initializer list. Note that 347 // regardless how we mark inner braces here, we will overwrite the 348 // BlockKind later if we parse a braced list (where all blocks 349 // inside are by default braced lists), or when we explicitly detect 350 // blocks (for example while parsing lambdas). 351 // 352 // We exclude + and - as they can be ObjC visibility modifiers. 353 ProbablyBracedList = 354 NextTok->isOneOf(tok::comma, tok::semi, tok::period, tok::colon, 355 tok::r_paren, tok::r_square, tok::l_brace, 356 tok::l_paren, tok::ellipsis) || 357 (NextTok->isBinaryOperator() && !NextIsObjCMethod); 358 } 359 if (ProbablyBracedList) { 360 Tok->BlockKind = BK_BracedInit; 361 LBraceStack.back()->BlockKind = BK_BracedInit; 362 } else { 363 Tok->BlockKind = BK_Block; 364 LBraceStack.back()->BlockKind = BK_Block; 365 } 366 } 367 LBraceStack.pop_back(); 368 } 369 break; 370 case tok::at: 371 case tok::semi: 372 case tok::kw_if: 373 case tok::kw_while: 374 case tok::kw_for: 375 case tok::kw_switch: 376 case tok::kw_try: 377 if (!LBraceStack.empty()) 378 LBraceStack.back()->BlockKind = BK_Block; 379 break; 380 default: 381 break; 382 } 383 Tok = NextTok; 384 Position += ReadTokens; 385 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 386 // Assume other blocks for all unclosed opening braces. 387 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 388 if (LBraceStack[i]->BlockKind == BK_Unknown) 389 LBraceStack[i]->BlockKind = BK_Block; 390 } 391 392 FormatTok = Tokens->setPosition(StoredPosition); 393 } 394 395 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 396 bool MunchSemi) { 397 assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected"); 398 unsigned InitialLevel = Line->Level; 399 nextToken(); 400 401 addUnwrappedLine(); 402 403 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 404 MustBeDeclaration); 405 if (AddLevel) 406 ++Line->Level; 407 parseLevel(/*HasOpeningBrace=*/true); 408 409 if (!FormatTok->Tok.is(tok::r_brace)) { 410 Line->Level = InitialLevel; 411 StructuralError = true; 412 return; 413 } 414 415 nextToken(); // Munch the closing brace. 416 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 417 nextToken(); 418 Line->Level = InitialLevel; 419 } 420 421 static bool IsGoogScope(const UnwrappedLine &Line) { 422 if (Line.Tokens.size() < 4) 423 return false; 424 auto I = Line.Tokens.begin(); 425 if (I->Tok->TokenText != "goog") 426 return false; 427 ++I; 428 if (I->Tok->isNot(tok::period)) 429 return false; 430 ++I; 431 if (I->Tok->TokenText != "scope") 432 return false; 433 ++I; 434 return I->Tok->is(tok::l_paren); 435 } 436 437 static bool ShouldBreakBeforeBrace(const FormatStyle &Style, 438 const FormatToken &InitialToken) { 439 switch (Style.BreakBeforeBraces) { 440 case FormatStyle::BS_Linux: 441 return InitialToken.isOneOf(tok::kw_namespace, tok::kw_class); 442 case FormatStyle::BS_Allman: 443 case FormatStyle::BS_GNU: 444 return true; 445 default: 446 return false; 447 } 448 } 449 450 void UnwrappedLineParser::parseChildBlock() { 451 FormatTok->BlockKind = BK_Block; 452 nextToken(); 453 { 454 bool GoogScope = 455 Style.Language == FormatStyle::LK_JavaScript && IsGoogScope(*Line); 456 ScopedLineState LineState(*this); 457 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 458 /*MustBeDeclaration=*/false); 459 Line->Level += GoogScope ? 0 : 1; 460 parseLevel(/*HasOpeningBrace=*/true); 461 Line->Level -= GoogScope ? 0 : 1; 462 } 463 nextToken(); 464 } 465 466 void UnwrappedLineParser::parsePPDirective() { 467 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 468 ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError); 469 nextToken(); 470 471 if (!FormatTok->Tok.getIdentifierInfo()) { 472 parsePPUnknown(); 473 return; 474 } 475 476 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 477 case tok::pp_define: 478 parsePPDefine(); 479 return; 480 case tok::pp_if: 481 parsePPIf(/*IfDef=*/false); 482 break; 483 case tok::pp_ifdef: 484 case tok::pp_ifndef: 485 parsePPIf(/*IfDef=*/true); 486 break; 487 case tok::pp_else: 488 parsePPElse(); 489 break; 490 case tok::pp_elif: 491 parsePPElIf(); 492 break; 493 case tok::pp_endif: 494 parsePPEndIf(); 495 break; 496 default: 497 parsePPUnknown(); 498 break; 499 } 500 } 501 502 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 503 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 504 PPStack.push_back(PP_Unreachable); 505 else 506 PPStack.push_back(PP_Conditional); 507 } 508 509 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 510 ++PPBranchLevel; 511 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 512 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 513 PPLevelBranchIndex.push_back(0); 514 PPLevelBranchCount.push_back(0); 515 } 516 PPChainBranchIndex.push(0); 517 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 518 conditionalCompilationCondition(Unreachable || Skip); 519 } 520 521 void UnwrappedLineParser::conditionalCompilationAlternative() { 522 if (!PPStack.empty()) 523 PPStack.pop_back(); 524 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 525 if (!PPChainBranchIndex.empty()) 526 ++PPChainBranchIndex.top(); 527 conditionalCompilationCondition( 528 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 529 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 530 } 531 532 void UnwrappedLineParser::conditionalCompilationEnd() { 533 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 534 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 535 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 536 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 537 } 538 } 539 // Guard against #endif's without #if. 540 if (PPBranchLevel > 0) 541 --PPBranchLevel; 542 if (!PPChainBranchIndex.empty()) 543 PPChainBranchIndex.pop(); 544 if (!PPStack.empty()) 545 PPStack.pop_back(); 546 } 547 548 void UnwrappedLineParser::parsePPIf(bool IfDef) { 549 nextToken(); 550 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 551 StringRef(FormatTok->Tok.getLiteralData(), 552 FormatTok->Tok.getLength()) == "0") || 553 FormatTok->Tok.is(tok::kw_false); 554 conditionalCompilationStart(!IfDef && IsLiteralFalse); 555 parsePPUnknown(); 556 } 557 558 void UnwrappedLineParser::parsePPElse() { 559 conditionalCompilationAlternative(); 560 parsePPUnknown(); 561 } 562 563 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 564 565 void UnwrappedLineParser::parsePPEndIf() { 566 conditionalCompilationEnd(); 567 parsePPUnknown(); 568 } 569 570 void UnwrappedLineParser::parsePPDefine() { 571 nextToken(); 572 573 if (FormatTok->Tok.getKind() != tok::identifier) { 574 parsePPUnknown(); 575 return; 576 } 577 nextToken(); 578 if (FormatTok->Tok.getKind() == tok::l_paren && 579 FormatTok->WhitespaceRange.getBegin() == 580 FormatTok->WhitespaceRange.getEnd()) { 581 parseParens(); 582 } 583 addUnwrappedLine(); 584 Line->Level = 1; 585 586 // Errors during a preprocessor directive can only affect the layout of the 587 // preprocessor directive, and thus we ignore them. An alternative approach 588 // would be to use the same approach we use on the file level (no 589 // re-indentation if there was a structural error) within the macro 590 // definition. 591 parseFile(); 592 } 593 594 void UnwrappedLineParser::parsePPUnknown() { 595 do { 596 nextToken(); 597 } while (!eof()); 598 addUnwrappedLine(); 599 } 600 601 // Here we blacklist certain tokens that are not usually the first token in an 602 // unwrapped line. This is used in attempt to distinguish macro calls without 603 // trailing semicolons from other constructs split to several lines. 604 bool tokenCanStartNewLine(clang::Token Tok) { 605 // Semicolon can be a null-statement, l_square can be a start of a macro or 606 // a C++11 attribute, but this doesn't seem to be common. 607 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 608 Tok.isNot(tok::l_square) && 609 // Tokens that can only be used as binary operators and a part of 610 // overloaded operator names. 611 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 612 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 613 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 614 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 615 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 616 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 617 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 618 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 619 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 620 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 621 Tok.isNot(tok::lesslessequal) && 622 // Colon is used in labels, base class lists, initializer lists, 623 // range-based for loops, ternary operator, but should never be the 624 // first token in an unwrapped line. 625 Tok.isNot(tok::colon) && 626 // 'noexcept' is a trailing annotation. 627 Tok.isNot(tok::kw_noexcept); 628 } 629 630 void UnwrappedLineParser::parseStructuralElement() { 631 assert(!FormatTok->Tok.is(tok::l_brace)); 632 switch (FormatTok->Tok.getKind()) { 633 case tok::at: 634 nextToken(); 635 if (FormatTok->Tok.is(tok::l_brace)) { 636 parseBracedList(); 637 break; 638 } 639 switch (FormatTok->Tok.getObjCKeywordID()) { 640 case tok::objc_public: 641 case tok::objc_protected: 642 case tok::objc_package: 643 case tok::objc_private: 644 return parseAccessSpecifier(); 645 case tok::objc_interface: 646 case tok::objc_implementation: 647 return parseObjCInterfaceOrImplementation(); 648 case tok::objc_protocol: 649 return parseObjCProtocol(); 650 case tok::objc_end: 651 return; // Handled by the caller. 652 case tok::objc_optional: 653 case tok::objc_required: 654 nextToken(); 655 addUnwrappedLine(); 656 return; 657 default: 658 break; 659 } 660 break; 661 case tok::kw_asm: 662 FormatTok->Finalized = true; 663 nextToken(); 664 if (FormatTok->is(tok::l_brace)) { 665 while (FormatTok && FormatTok->isNot(tok::eof)) { 666 FormatTok->Finalized = true; 667 if (FormatTok->is(tok::r_brace)) { 668 nextToken(); 669 break; 670 } 671 nextToken(); 672 } 673 } 674 break; 675 case tok::kw_namespace: 676 parseNamespace(); 677 return; 678 case tok::kw_inline: 679 nextToken(); 680 if (FormatTok->Tok.is(tok::kw_namespace)) { 681 parseNamespace(); 682 return; 683 } 684 break; 685 case tok::kw_public: 686 case tok::kw_protected: 687 case tok::kw_private: 688 parseAccessSpecifier(); 689 return; 690 case tok::kw_if: 691 parseIfThenElse(); 692 return; 693 case tok::kw_for: 694 case tok::kw_while: 695 parseForOrWhileLoop(); 696 return; 697 case tok::kw_do: 698 parseDoWhile(); 699 return; 700 case tok::kw_switch: 701 parseSwitch(); 702 return; 703 case tok::kw_default: 704 nextToken(); 705 parseLabel(); 706 return; 707 case tok::kw_case: 708 parseCaseLabel(); 709 return; 710 case tok::kw_try: 711 parseTryCatch(); 712 return; 713 case tok::kw_extern: 714 nextToken(); 715 if (FormatTok->Tok.is(tok::string_literal)) { 716 nextToken(); 717 if (FormatTok->Tok.is(tok::l_brace)) { 718 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 719 addUnwrappedLine(); 720 return; 721 } 722 } 723 break; 724 case tok::identifier: 725 if (FormatTok->IsForEachMacro) { 726 parseForOrWhileLoop(); 727 return; 728 } 729 // In all other cases, parse the declaration. 730 break; 731 default: 732 break; 733 } 734 do { 735 switch (FormatTok->Tok.getKind()) { 736 case tok::at: 737 nextToken(); 738 if (FormatTok->Tok.is(tok::l_brace)) 739 parseBracedList(); 740 break; 741 case tok::kw_enum: 742 parseEnum(); 743 break; 744 case tok::kw_typedef: 745 nextToken(); 746 // FIXME: Use the IdentifierTable instead. 747 if (FormatTok->TokenText == "NS_ENUM") 748 parseEnum(); 749 break; 750 case tok::kw_struct: 751 case tok::kw_union: 752 case tok::kw_class: 753 parseRecord(); 754 // A record declaration or definition is always the start of a structural 755 // element. 756 break; 757 case tok::semi: 758 nextToken(); 759 addUnwrappedLine(); 760 return; 761 case tok::r_brace: 762 addUnwrappedLine(); 763 return; 764 case tok::l_paren: 765 parseParens(); 766 break; 767 case tok::caret: 768 nextToken(); 769 if (FormatTok->Tok.isAnyIdentifier() || 770 FormatTok->isSimpleTypeSpecifier()) 771 nextToken(); 772 if (FormatTok->is(tok::l_paren)) 773 parseParens(); 774 if (FormatTok->is(tok::l_brace)) 775 parseChildBlock(); 776 break; 777 case tok::l_brace: 778 if (!tryToParseBracedList()) { 779 // A block outside of parentheses must be the last part of a 780 // structural element. 781 // FIXME: Figure out cases where this is not true, and add projections 782 // for them (the one we know is missing are lambdas). 783 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 784 addUnwrappedLine(); 785 FormatTok->Type = TT_FunctionLBrace; 786 parseBlock(/*MustBeDeclaration=*/false); 787 addUnwrappedLine(); 788 return; 789 } 790 // Otherwise this was a braced init list, and the structural 791 // element continues. 792 break; 793 case tok::kw_try: 794 // We arrive here when parsing function-try blocks. 795 parseTryCatch(); 796 return; 797 case tok::identifier: { 798 StringRef Text = FormatTok->TokenText; 799 // Parse function literal unless 'function' is the first token in a line 800 // in which case this should be treated as a free-standing function. 801 if (Style.Language == FormatStyle::LK_JavaScript && Text == "function" && 802 Line->Tokens.size() > 0) { 803 tryToParseJSFunction(); 804 break; 805 } 806 nextToken(); 807 if (Line->Tokens.size() == 1) { 808 if (FormatTok->Tok.is(tok::colon)) { 809 parseLabel(); 810 return; 811 } 812 // Recognize function-like macro usages without trailing semicolon. 813 if (FormatTok->Tok.is(tok::l_paren)) { 814 parseParens(); 815 if (FormatTok->NewlinesBefore > 0 && 816 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 817 addUnwrappedLine(); 818 return; 819 } 820 } else if (FormatTok->HasUnescapedNewline && Text.size() >= 5 && 821 Text == Text.upper()) { 822 // Recognize free-standing macros like Q_OBJECT. 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 Line->Tokens.back().Tok->closesScope() || 851 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 852 nextToken(); 853 return false; 854 } 855 assert(FormatTok->is(tok::l_square)); 856 FormatToken &LSquare = *FormatTok; 857 if (!tryToParseLambdaIntroducer()) 858 return false; 859 860 while (FormatTok->isNot(tok::l_brace)) { 861 if (FormatTok->isSimpleTypeSpecifier()) { 862 nextToken(); 863 continue; 864 } 865 switch (FormatTok->Tok.getKind()) { 866 case tok::l_brace: 867 break; 868 case tok::l_paren: 869 parseParens(); 870 break; 871 case tok::less: 872 case tok::greater: 873 case tok::identifier: 874 case tok::coloncolon: 875 case tok::kw_mutable: 876 nextToken(); 877 break; 878 case tok::arrow: 879 FormatTok->Type = TT_TrailingReturnArrow; 880 nextToken(); 881 break; 882 default: 883 return true; 884 } 885 } 886 LSquare.Type = TT_LambdaLSquare; 887 parseChildBlock(); 888 return true; 889 } 890 891 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 892 nextToken(); 893 if (FormatTok->is(tok::equal)) { 894 nextToken(); 895 if (FormatTok->is(tok::r_square)) { 896 nextToken(); 897 return true; 898 } 899 if (FormatTok->isNot(tok::comma)) 900 return false; 901 nextToken(); 902 } else if (FormatTok->is(tok::amp)) { 903 nextToken(); 904 if (FormatTok->is(tok::r_square)) { 905 nextToken(); 906 return true; 907 } 908 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 909 return false; 910 } 911 if (FormatTok->is(tok::comma)) 912 nextToken(); 913 } else if (FormatTok->is(tok::r_square)) { 914 nextToken(); 915 return true; 916 } 917 do { 918 if (FormatTok->is(tok::amp)) 919 nextToken(); 920 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 921 return false; 922 nextToken(); 923 if (FormatTok->is(tok::ellipsis)) 924 nextToken(); 925 if (FormatTok->is(tok::comma)) { 926 nextToken(); 927 } else if (FormatTok->is(tok::r_square)) { 928 nextToken(); 929 return true; 930 } else { 931 return false; 932 } 933 } while (!eof()); 934 return false; 935 } 936 937 void UnwrappedLineParser::tryToParseJSFunction() { 938 nextToken(); 939 940 // Consume function name. 941 if (FormatTok->is(tok::identifier)) 942 nextToken(); 943 944 if (FormatTok->isNot(tok::l_paren)) 945 return; 946 nextToken(); 947 while (FormatTok->isNot(tok::l_brace)) { 948 // Err on the side of caution in order to avoid consuming the full file in 949 // case of incomplete code. 950 if (!FormatTok->isOneOf(tok::identifier, tok::comma, tok::r_paren, 951 tok::comment)) 952 return; 953 nextToken(); 954 } 955 parseChildBlock(); 956 } 957 958 bool UnwrappedLineParser::tryToParseBracedList() { 959 if (FormatTok->BlockKind == BK_Unknown) 960 calculateBraceTypes(); 961 assert(FormatTok->BlockKind != BK_Unknown); 962 if (FormatTok->BlockKind == BK_Block) 963 return false; 964 parseBracedList(); 965 return true; 966 } 967 968 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 969 bool HasError = false; 970 nextToken(); 971 972 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 973 // replace this by using parseAssigmentExpression() inside. 974 do { 975 if (Style.Language == FormatStyle::LK_JavaScript && 976 FormatTok->TokenText == "function") { 977 tryToParseJSFunction(); 978 continue; 979 } 980 switch (FormatTok->Tok.getKind()) { 981 case tok::caret: 982 nextToken(); 983 if (FormatTok->is(tok::l_brace)) { 984 parseChildBlock(); 985 } 986 break; 987 case tok::l_square: 988 tryToParseLambda(); 989 break; 990 case tok::l_brace: 991 // Assume there are no blocks inside a braced init list apart 992 // from the ones we explicitly parse out (like lambdas). 993 FormatTok->BlockKind = BK_BracedInit; 994 parseBracedList(); 995 break; 996 case tok::r_brace: 997 nextToken(); 998 return !HasError; 999 case tok::semi: 1000 HasError = true; 1001 if (!ContinueOnSemicolons) 1002 return !HasError; 1003 nextToken(); 1004 break; 1005 case tok::comma: 1006 nextToken(); 1007 break; 1008 default: 1009 nextToken(); 1010 break; 1011 } 1012 } while (!eof()); 1013 return false; 1014 } 1015 1016 void UnwrappedLineParser::parseParens() { 1017 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1018 nextToken(); 1019 do { 1020 switch (FormatTok->Tok.getKind()) { 1021 case tok::l_paren: 1022 parseParens(); 1023 break; 1024 case tok::r_paren: 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 tryToParseLambda(); 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::parseSquare() { 1052 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1053 if (tryToParseLambda()) 1054 return; 1055 do { 1056 switch (FormatTok->Tok.getKind()) { 1057 case tok::l_paren: 1058 parseParens(); 1059 break; 1060 case tok::r_square: 1061 nextToken(); 1062 return; 1063 case tok::r_brace: 1064 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1065 return; 1066 case tok::l_square: 1067 parseSquare(); 1068 break; 1069 case tok::l_brace: { 1070 if (!tryToParseBracedList()) { 1071 parseChildBlock(); 1072 } 1073 break; 1074 } 1075 case tok::at: 1076 nextToken(); 1077 if (FormatTok->Tok.is(tok::l_brace)) 1078 parseBracedList(); 1079 break; 1080 default: 1081 nextToken(); 1082 break; 1083 } 1084 } while (!eof()); 1085 } 1086 1087 void UnwrappedLineParser::parseIfThenElse() { 1088 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1089 nextToken(); 1090 if (FormatTok->Tok.is(tok::l_paren)) 1091 parseParens(); 1092 bool NeedsUnwrappedLine = false; 1093 if (FormatTok->Tok.is(tok::l_brace)) { 1094 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1095 parseBlock(/*MustBeDeclaration=*/false); 1096 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1097 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1098 addUnwrappedLine(); 1099 } else { 1100 NeedsUnwrappedLine = true; 1101 } 1102 } else { 1103 addUnwrappedLine(); 1104 ++Line->Level; 1105 parseStructuralElement(); 1106 --Line->Level; 1107 } 1108 if (FormatTok->Tok.is(tok::kw_else)) { 1109 if (Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) 1110 addUnwrappedLine(); 1111 nextToken(); 1112 if (FormatTok->Tok.is(tok::l_brace)) { 1113 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1114 parseBlock(/*MustBeDeclaration=*/false); 1115 addUnwrappedLine(); 1116 } else if (FormatTok->Tok.is(tok::kw_if)) { 1117 parseIfThenElse(); 1118 } else { 1119 addUnwrappedLine(); 1120 ++Line->Level; 1121 parseStructuralElement(); 1122 --Line->Level; 1123 } 1124 } else if (NeedsUnwrappedLine) { 1125 addUnwrappedLine(); 1126 } 1127 } 1128 1129 void UnwrappedLineParser::parseTryCatch() { 1130 assert(FormatTok->is(tok::kw_try) && "'try' expected"); 1131 nextToken(); 1132 bool NeedsUnwrappedLine = false; 1133 if (FormatTok->is(tok::colon)) { 1134 // We are in a function try block, what comes is an initializer list. 1135 nextToken(); 1136 while (FormatTok->is(tok::identifier)) { 1137 nextToken(); 1138 if (FormatTok->is(tok::l_paren)) 1139 parseParens(); 1140 else 1141 StructuralError = true; 1142 if (FormatTok->is(tok::comma)) 1143 nextToken(); 1144 } 1145 } 1146 if (FormatTok->is(tok::l_brace)) { 1147 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1148 parseBlock(/*MustBeDeclaration=*/false); 1149 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1150 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1151 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1152 addUnwrappedLine(); 1153 } else { 1154 NeedsUnwrappedLine = true; 1155 } 1156 } else if (!FormatTok->is(tok::kw_catch)) { 1157 // The C++ standard requires a compound-statement after a try. 1158 // If there's none, we try to assume there's a structuralElement 1159 // and try to continue. 1160 StructuralError = true; 1161 addUnwrappedLine(); 1162 ++Line->Level; 1163 parseStructuralElement(); 1164 --Line->Level; 1165 } 1166 while (FormatTok->is(tok::kw_catch) || 1167 (Style.Language == FormatStyle::LK_JavaScript && 1168 FormatTok->TokenText == "finally")) { 1169 nextToken(); 1170 while (FormatTok->isNot(tok::l_brace)) { 1171 if (FormatTok->is(tok::l_paren)) { 1172 parseParens(); 1173 continue; 1174 } 1175 if (FormatTok->isOneOf(tok::semi, tok::r_brace)) 1176 return; 1177 nextToken(); 1178 } 1179 NeedsUnwrappedLine = false; 1180 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1181 parseBlock(/*MustBeDeclaration=*/false); 1182 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1183 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1184 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1185 addUnwrappedLine(); 1186 } else { 1187 NeedsUnwrappedLine = true; 1188 } 1189 } 1190 if (NeedsUnwrappedLine) { 1191 addUnwrappedLine(); 1192 } 1193 } 1194 1195 void UnwrappedLineParser::parseNamespace() { 1196 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1197 1198 const FormatToken &InitialToken = *FormatTok; 1199 nextToken(); 1200 if (FormatTok->Tok.is(tok::identifier)) 1201 nextToken(); 1202 if (FormatTok->Tok.is(tok::l_brace)) { 1203 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1204 addUnwrappedLine(); 1205 1206 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1207 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1208 DeclarationScopeStack.size() > 1); 1209 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1210 // Munch the semicolon after a namespace. This is more common than one would 1211 // think. Puttin the semicolon into its own line is very ugly. 1212 if (FormatTok->Tok.is(tok::semi)) 1213 nextToken(); 1214 addUnwrappedLine(); 1215 } 1216 // FIXME: Add error handling. 1217 } 1218 1219 void UnwrappedLineParser::parseForOrWhileLoop() { 1220 assert((FormatTok->Tok.is(tok::kw_for) || FormatTok->Tok.is(tok::kw_while) || 1221 FormatTok->IsForEachMacro) && 1222 "'for', 'while' or foreach macro expected"); 1223 nextToken(); 1224 if (FormatTok->Tok.is(tok::l_paren)) 1225 parseParens(); 1226 if (FormatTok->Tok.is(tok::l_brace)) { 1227 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1228 parseBlock(/*MustBeDeclaration=*/false); 1229 addUnwrappedLine(); 1230 } else { 1231 addUnwrappedLine(); 1232 ++Line->Level; 1233 parseStructuralElement(); 1234 --Line->Level; 1235 } 1236 } 1237 1238 void UnwrappedLineParser::parseDoWhile() { 1239 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1240 nextToken(); 1241 if (FormatTok->Tok.is(tok::l_brace)) { 1242 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1243 parseBlock(/*MustBeDeclaration=*/false); 1244 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1245 addUnwrappedLine(); 1246 } else { 1247 addUnwrappedLine(); 1248 ++Line->Level; 1249 parseStructuralElement(); 1250 --Line->Level; 1251 } 1252 1253 // FIXME: Add error handling. 1254 if (!FormatTok->Tok.is(tok::kw_while)) { 1255 addUnwrappedLine(); 1256 return; 1257 } 1258 1259 nextToken(); 1260 parseStructuralElement(); 1261 } 1262 1263 void UnwrappedLineParser::parseLabel() { 1264 nextToken(); 1265 unsigned OldLineLevel = Line->Level; 1266 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1267 --Line->Level; 1268 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1269 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1270 parseBlock(/*MustBeDeclaration=*/false); 1271 if (FormatTok->Tok.is(tok::kw_break)) { 1272 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1273 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1274 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1275 addUnwrappedLine(); 1276 } 1277 parseStructuralElement(); 1278 } 1279 addUnwrappedLine(); 1280 } else { 1281 addUnwrappedLine(); 1282 } 1283 Line->Level = OldLineLevel; 1284 } 1285 1286 void UnwrappedLineParser::parseCaseLabel() { 1287 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1288 // FIXME: fix handling of complex expressions here. 1289 do { 1290 nextToken(); 1291 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1292 parseLabel(); 1293 } 1294 1295 void UnwrappedLineParser::parseSwitch() { 1296 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1297 nextToken(); 1298 if (FormatTok->Tok.is(tok::l_paren)) 1299 parseParens(); 1300 if (FormatTok->Tok.is(tok::l_brace)) { 1301 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1302 parseBlock(/*MustBeDeclaration=*/false); 1303 addUnwrappedLine(); 1304 } else { 1305 addUnwrappedLine(); 1306 ++Line->Level; 1307 parseStructuralElement(); 1308 --Line->Level; 1309 } 1310 } 1311 1312 void UnwrappedLineParser::parseAccessSpecifier() { 1313 nextToken(); 1314 // Understand Qt's slots. 1315 if (FormatTok->is(tok::identifier) && 1316 (FormatTok->TokenText == "slots" || FormatTok->TokenText == "Q_SLOTS")) 1317 nextToken(); 1318 // Otherwise, we don't know what it is, and we'd better keep the next token. 1319 if (FormatTok->Tok.is(tok::colon)) 1320 nextToken(); 1321 addUnwrappedLine(); 1322 } 1323 1324 void UnwrappedLineParser::parseEnum() { 1325 if (FormatTok->Tok.is(tok::kw_enum)) { 1326 // Won't be 'enum' for NS_ENUMs. 1327 nextToken(); 1328 } 1329 // Eat up enum class ... 1330 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1331 nextToken(); 1332 while (FormatTok->Tok.getIdentifierInfo() || 1333 FormatTok->isOneOf(tok::colon, tok::coloncolon)) { 1334 nextToken(); 1335 // We can have macros or attributes in between 'enum' and the enum name. 1336 if (FormatTok->Tok.is(tok::l_paren)) { 1337 parseParens(); 1338 } 1339 if (FormatTok->Tok.is(tok::identifier)) 1340 nextToken(); 1341 } 1342 if (FormatTok->Tok.is(tok::l_brace)) { 1343 FormatTok->BlockKind = BK_Block; 1344 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1345 if (HasError) { 1346 if (FormatTok->is(tok::semi)) 1347 nextToken(); 1348 addUnwrappedLine(); 1349 } 1350 } 1351 // We fall through to parsing a structural element afterwards, so that in 1352 // enum A {} n, m; 1353 // "} n, m;" will end up in one unwrapped line. 1354 } 1355 1356 void UnwrappedLineParser::parseRecord() { 1357 const FormatToken &InitialToken = *FormatTok; 1358 nextToken(); 1359 if (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::kw___attribute, 1360 tok::kw___declspec, tok::kw_alignas)) { 1361 nextToken(); 1362 // We can have macros or attributes in between 'class' and the class name. 1363 if (FormatTok->Tok.is(tok::l_paren)) { 1364 parseParens(); 1365 } 1366 // The actual identifier can be a nested name specifier, and in macros 1367 // it is often token-pasted. 1368 while (FormatTok->Tok.is(tok::identifier) || 1369 FormatTok->Tok.is(tok::coloncolon) || 1370 FormatTok->Tok.is(tok::hashhash)) 1371 nextToken(); 1372 1373 // Note that parsing away template declarations here leads to incorrectly 1374 // accepting function declarations as record declarations. 1375 // In general, we cannot solve this problem. Consider: 1376 // class A<int> B() {} 1377 // which can be a function definition or a class definition when B() is a 1378 // macro. If we find enough real-world cases where this is a problem, we 1379 // can parse for the 'template' keyword in the beginning of the statement, 1380 // and thus rule out the record production in case there is no template 1381 // (this would still leave us with an ambiguity between template function 1382 // and class declarations). 1383 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1384 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1385 if (FormatTok->Tok.is(tok::semi)) 1386 return; 1387 nextToken(); 1388 } 1389 } 1390 } 1391 if (FormatTok->Tok.is(tok::l_brace)) { 1392 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1393 addUnwrappedLine(); 1394 1395 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1396 /*MunchSemi=*/false); 1397 } 1398 // We fall through to parsing a structural element afterwards, so 1399 // class A {} n, m; 1400 // will end up in one unwrapped line. 1401 } 1402 1403 void UnwrappedLineParser::parseObjCProtocolList() { 1404 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1405 do 1406 nextToken(); 1407 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1408 nextToken(); // Skip '>'. 1409 } 1410 1411 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1412 do { 1413 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1414 nextToken(); 1415 addUnwrappedLine(); 1416 break; 1417 } 1418 if (FormatTok->is(tok::l_brace)) { 1419 parseBlock(/*MustBeDeclaration=*/false); 1420 // In ObjC interfaces, nothing should be following the "}". 1421 addUnwrappedLine(); 1422 } else if (FormatTok->is(tok::r_brace)) { 1423 // Ignore stray "}". parseStructuralElement doesn't consume them. 1424 nextToken(); 1425 addUnwrappedLine(); 1426 } else { 1427 parseStructuralElement(); 1428 } 1429 } while (!eof()); 1430 } 1431 1432 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1433 nextToken(); 1434 nextToken(); // interface name 1435 1436 // @interface can be followed by either a base class, or a category. 1437 if (FormatTok->Tok.is(tok::colon)) { 1438 nextToken(); 1439 nextToken(); // base class name 1440 } else if (FormatTok->Tok.is(tok::l_paren)) 1441 // Skip category, if present. 1442 parseParens(); 1443 1444 if (FormatTok->Tok.is(tok::less)) 1445 parseObjCProtocolList(); 1446 1447 if (FormatTok->Tok.is(tok::l_brace)) { 1448 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1449 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1450 addUnwrappedLine(); 1451 parseBlock(/*MustBeDeclaration=*/true); 1452 } 1453 1454 // With instance variables, this puts '}' on its own line. Without instance 1455 // variables, this ends the @interface line. 1456 addUnwrappedLine(); 1457 1458 parseObjCUntilAtEnd(); 1459 } 1460 1461 void UnwrappedLineParser::parseObjCProtocol() { 1462 nextToken(); 1463 nextToken(); // protocol name 1464 1465 if (FormatTok->Tok.is(tok::less)) 1466 parseObjCProtocolList(); 1467 1468 // Check for protocol declaration. 1469 if (FormatTok->Tok.is(tok::semi)) { 1470 nextToken(); 1471 return addUnwrappedLine(); 1472 } 1473 1474 addUnwrappedLine(); 1475 parseObjCUntilAtEnd(); 1476 } 1477 1478 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1479 StringRef Prefix = "") { 1480 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1481 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1482 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1483 E = Line.Tokens.end(); 1484 I != E; ++I) { 1485 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1486 } 1487 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1488 E = Line.Tokens.end(); 1489 I != E; ++I) { 1490 const UnwrappedLineNode &Node = *I; 1491 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1492 I = Node.Children.begin(), 1493 E = Node.Children.end(); 1494 I != E; ++I) { 1495 printDebugInfo(*I, "\nChild: "); 1496 } 1497 } 1498 llvm::dbgs() << "\n"; 1499 } 1500 1501 void UnwrappedLineParser::addUnwrappedLine() { 1502 if (Line->Tokens.empty()) 1503 return; 1504 DEBUG({ 1505 if (CurrentLines == &Lines) 1506 printDebugInfo(*Line); 1507 }); 1508 CurrentLines->push_back(*Line); 1509 Line->Tokens.clear(); 1510 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1511 for (SmallVectorImpl<UnwrappedLine>::iterator 1512 I = PreprocessorDirectives.begin(), 1513 E = PreprocessorDirectives.end(); 1514 I != E; ++I) { 1515 CurrentLines->push_back(*I); 1516 } 1517 PreprocessorDirectives.clear(); 1518 } 1519 } 1520 1521 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1522 1523 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1524 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1525 FormatTok.NewlinesBefore > 0; 1526 } 1527 1528 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1529 bool JustComments = Line->Tokens.empty(); 1530 for (SmallVectorImpl<FormatToken *>::const_iterator 1531 I = CommentsBeforeNextToken.begin(), 1532 E = CommentsBeforeNextToken.end(); 1533 I != E; ++I) { 1534 if (isOnNewLine(**I) && JustComments) { 1535 addUnwrappedLine(); 1536 } 1537 pushToken(*I); 1538 } 1539 if (NewlineBeforeNext && JustComments) { 1540 addUnwrappedLine(); 1541 } 1542 CommentsBeforeNextToken.clear(); 1543 } 1544 1545 void UnwrappedLineParser::nextToken() { 1546 if (eof()) 1547 return; 1548 flushComments(isOnNewLine(*FormatTok)); 1549 pushToken(FormatTok); 1550 readToken(); 1551 } 1552 1553 void UnwrappedLineParser::readToken() { 1554 bool CommentsInCurrentLine = true; 1555 do { 1556 FormatTok = Tokens->getNextToken(); 1557 assert(FormatTok); 1558 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1559 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1560 // If there is an unfinished unwrapped line, we flush the preprocessor 1561 // directives only after that unwrapped line was finished later. 1562 bool SwitchToPreprocessorLines = 1563 !Line->Tokens.empty() && CurrentLines == &Lines; 1564 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1565 // Comments stored before the preprocessor directive need to be output 1566 // before the preprocessor directive, at the same level as the 1567 // preprocessor directive, as we consider them to apply to the directive. 1568 flushComments(isOnNewLine(*FormatTok)); 1569 parsePPDirective(); 1570 } 1571 while (FormatTok->Type == TT_ConflictStart || 1572 FormatTok->Type == TT_ConflictEnd || 1573 FormatTok->Type == TT_ConflictAlternative) { 1574 if (FormatTok->Type == TT_ConflictStart) { 1575 conditionalCompilationStart(/*Unreachable=*/false); 1576 } else if (FormatTok->Type == TT_ConflictAlternative) { 1577 conditionalCompilationAlternative(); 1578 } else if (FormatTok->Type == TT_ConflictEnd) { 1579 conditionalCompilationEnd(); 1580 } 1581 FormatTok = Tokens->getNextToken(); 1582 FormatTok->MustBreakBefore = true; 1583 } 1584 1585 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1586 !Line->InPPDirective) { 1587 continue; 1588 } 1589 1590 if (!FormatTok->Tok.is(tok::comment)) 1591 return; 1592 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1593 CommentsInCurrentLine = false; 1594 } 1595 if (CommentsInCurrentLine) { 1596 pushToken(FormatTok); 1597 } else { 1598 CommentsBeforeNextToken.push_back(FormatTok); 1599 } 1600 } while (!eof()); 1601 } 1602 1603 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1604 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1605 if (MustBreakBeforeNextToken) { 1606 Line->Tokens.back().Tok->MustBreakBefore = true; 1607 MustBreakBeforeNextToken = false; 1608 } 1609 } 1610 1611 } // end namespace format 1612 } // end namespace clang 1613