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/ADT/STLExtras.h" 18 #include "llvm/Support/Debug.h" 19 20 #define DEBUG_TYPE "format-parser" 21 22 namespace clang { 23 namespace format { 24 25 class FormatTokenSource { 26 public: 27 virtual ~FormatTokenSource() {} 28 virtual FormatToken *getNextToken() = 0; 29 30 virtual unsigned getPosition() = 0; 31 virtual FormatToken *setPosition(unsigned Position) = 0; 32 }; 33 34 namespace { 35 36 class ScopedDeclarationState { 37 public: 38 ScopedDeclarationState(UnwrappedLine &Line, std::vector<bool> &Stack, 39 bool MustBeDeclaration) 40 : Line(Line), Stack(Stack) { 41 Line.MustBeDeclaration = MustBeDeclaration; 42 Stack.push_back(MustBeDeclaration); 43 } 44 ~ScopedDeclarationState() { 45 Stack.pop_back(); 46 if (!Stack.empty()) 47 Line.MustBeDeclaration = Stack.back(); 48 else 49 Line.MustBeDeclaration = true; 50 } 51 52 private: 53 UnwrappedLine &Line; 54 std::vector<bool> &Stack; 55 }; 56 57 class ScopedMacroState : public FormatTokenSource { 58 public: 59 ScopedMacroState(UnwrappedLine &Line, FormatTokenSource *&TokenSource, 60 FormatToken *&ResetToken, bool &StructuralError) 61 : Line(Line), TokenSource(TokenSource), ResetToken(ResetToken), 62 PreviousLineLevel(Line.Level), PreviousTokenSource(TokenSource), 63 StructuralError(StructuralError), 64 PreviousStructuralError(StructuralError), Token(nullptr) { 65 TokenSource = this; 66 Line.Level = 0; 67 Line.InPPDirective = true; 68 } 69 70 ~ScopedMacroState() { 71 TokenSource = PreviousTokenSource; 72 ResetToken = Token; 73 Line.InPPDirective = false; 74 Line.Level = PreviousLineLevel; 75 StructuralError = PreviousStructuralError; 76 } 77 78 FormatToken *getNextToken() override { 79 // The \c UnwrappedLineParser guards against this by never calling 80 // \c getNextToken() after it has encountered the first eof token. 81 assert(!eof()); 82 Token = PreviousTokenSource->getNextToken(); 83 if (eof()) 84 return getFakeEOF(); 85 return Token; 86 } 87 88 unsigned getPosition() override { return PreviousTokenSource->getPosition(); } 89 90 FormatToken *setPosition(unsigned Position) override { 91 Token = PreviousTokenSource->setPosition(Position); 92 return Token; 93 } 94 95 private: 96 bool eof() { return Token && Token->HasUnescapedNewline; } 97 98 FormatToken *getFakeEOF() { 99 static bool EOFInitialized = false; 100 static FormatToken FormatTok; 101 if (!EOFInitialized) { 102 FormatTok.Tok.startToken(); 103 FormatTok.Tok.setKind(tok::eof); 104 EOFInitialized = true; 105 } 106 return &FormatTok; 107 } 108 109 UnwrappedLine &Line; 110 FormatTokenSource *&TokenSource; 111 FormatToken *&ResetToken; 112 unsigned PreviousLineLevel; 113 FormatTokenSource *PreviousTokenSource; 114 bool &StructuralError; 115 bool PreviousStructuralError; 116 117 FormatToken *Token; 118 }; 119 120 } // end anonymous namespace 121 122 class ScopedLineState { 123 public: 124 ScopedLineState(UnwrappedLineParser &Parser, 125 bool SwitchToPreprocessorLines = false) 126 : Parser(Parser), OriginalLines(Parser.CurrentLines) { 127 if (SwitchToPreprocessorLines) 128 Parser.CurrentLines = &Parser.PreprocessorDirectives; 129 else if (!Parser.Line->Tokens.empty()) 130 Parser.CurrentLines = &Parser.Line->Tokens.back().Children; 131 PreBlockLine = std::move(Parser.Line); 132 Parser.Line = llvm::make_unique<UnwrappedLine>(); 133 Parser.Line->Level = PreBlockLine->Level; 134 Parser.Line->InPPDirective = PreBlockLine->InPPDirective; 135 } 136 137 ~ScopedLineState() { 138 if (!Parser.Line->Tokens.empty()) { 139 Parser.addUnwrappedLine(); 140 } 141 assert(Parser.Line->Tokens.empty()); 142 Parser.Line = std::move(PreBlockLine); 143 if (Parser.CurrentLines == &Parser.PreprocessorDirectives) 144 Parser.MustBreakBeforeNextToken = true; 145 Parser.CurrentLines = OriginalLines; 146 } 147 148 private: 149 UnwrappedLineParser &Parser; 150 151 std::unique_ptr<UnwrappedLine> PreBlockLine; 152 SmallVectorImpl<UnwrappedLine> *OriginalLines; 153 }; 154 155 class CompoundStatementIndenter { 156 public: 157 CompoundStatementIndenter(UnwrappedLineParser *Parser, 158 const FormatStyle &Style, unsigned &LineLevel) 159 : LineLevel(LineLevel), OldLineLevel(LineLevel) { 160 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman) { 161 Parser->addUnwrappedLine(); 162 } else if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 163 Parser->addUnwrappedLine(); 164 ++LineLevel; 165 } 166 } 167 ~CompoundStatementIndenter() { LineLevel = OldLineLevel; } 168 169 private: 170 unsigned &LineLevel; 171 unsigned OldLineLevel; 172 }; 173 174 namespace { 175 176 class IndexedTokenSource : public FormatTokenSource { 177 public: 178 IndexedTokenSource(ArrayRef<FormatToken *> Tokens) 179 : Tokens(Tokens), Position(-1) {} 180 181 FormatToken *getNextToken() override { 182 ++Position; 183 return Tokens[Position]; 184 } 185 186 unsigned getPosition() override { 187 assert(Position >= 0); 188 return Position; 189 } 190 191 FormatToken *setPosition(unsigned P) override { 192 Position = P; 193 return Tokens[Position]; 194 } 195 196 void reset() { Position = -1; } 197 198 private: 199 ArrayRef<FormatToken *> Tokens; 200 int Position; 201 }; 202 203 } // end anonymous namespace 204 205 UnwrappedLineParser::UnwrappedLineParser(const FormatStyle &Style, 206 const AdditionalKeywords &Keywords, 207 ArrayRef<FormatToken *> Tokens, 208 UnwrappedLineConsumer &Callback) 209 : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), 210 CurrentLines(&Lines), StructuralError(false), Style(Style), 211 Keywords(Keywords), Tokens(nullptr), Callback(Callback), 212 AllTokens(Tokens), PPBranchLevel(-1) {} 213 214 void UnwrappedLineParser::reset() { 215 PPBranchLevel = -1; 216 Line.reset(new UnwrappedLine); 217 CommentsBeforeNextToken.clear(); 218 FormatTok = nullptr; 219 MustBreakBeforeNextToken = false; 220 PreprocessorDirectives.clear(); 221 CurrentLines = &Lines; 222 DeclarationScopeStack.clear(); 223 StructuralError = false; 224 PPStack.clear(); 225 } 226 227 bool UnwrappedLineParser::parse() { 228 IndexedTokenSource TokenSource(AllTokens); 229 do { 230 DEBUG(llvm::dbgs() << "----\n"); 231 reset(); 232 Tokens = &TokenSource; 233 TokenSource.reset(); 234 235 readToken(); 236 parseFile(); 237 // Create line with eof token. 238 pushToken(FormatTok); 239 addUnwrappedLine(); 240 241 for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(), 242 E = Lines.end(); 243 I != E; ++I) { 244 Callback.consumeUnwrappedLine(*I); 245 } 246 Callback.finishRun(); 247 Lines.clear(); 248 while (!PPLevelBranchIndex.empty() && 249 PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) { 250 PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1); 251 PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1); 252 } 253 if (!PPLevelBranchIndex.empty()) { 254 ++PPLevelBranchIndex.back(); 255 assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size()); 256 assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back()); 257 } 258 } while (!PPLevelBranchIndex.empty()); 259 260 return StructuralError; 261 } 262 263 void UnwrappedLineParser::parseFile() { 264 ScopedDeclarationState DeclarationState( 265 *Line, DeclarationScopeStack, 266 /*MustBeDeclaration=*/!Line->InPPDirective); 267 parseLevel(/*HasOpeningBrace=*/false); 268 // Make sure to format the remaining tokens. 269 flushComments(true); 270 addUnwrappedLine(); 271 } 272 273 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) { 274 bool SwitchLabelEncountered = false; 275 do { 276 switch (FormatTok->Tok.getKind()) { 277 case tok::comment: 278 nextToken(); 279 addUnwrappedLine(); 280 break; 281 case tok::l_brace: 282 // FIXME: Add parameter whether this can happen - if this happens, we must 283 // be in a non-declaration context. 284 parseBlock(/*MustBeDeclaration=*/false); 285 addUnwrappedLine(); 286 break; 287 case tok::r_brace: 288 if (HasOpeningBrace) 289 return; 290 StructuralError = true; 291 nextToken(); 292 addUnwrappedLine(); 293 break; 294 case tok::kw_default: 295 case tok::kw_case: 296 if (!SwitchLabelEncountered && 297 (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1))) 298 ++Line->Level; 299 SwitchLabelEncountered = true; 300 parseStructuralElement(); 301 break; 302 default: 303 parseStructuralElement(); 304 break; 305 } 306 } while (!eof()); 307 } 308 309 void UnwrappedLineParser::calculateBraceTypes() { 310 // We'll parse forward through the tokens until we hit 311 // a closing brace or eof - note that getNextToken() will 312 // parse macros, so this will magically work inside macro 313 // definitions, too. 314 unsigned StoredPosition = Tokens->getPosition(); 315 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 case tok::kw___try: 379 if (!LBraceStack.empty()) 380 LBraceStack.back()->BlockKind = BK_Block; 381 break; 382 default: 383 break; 384 } 385 Tok = NextTok; 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 // FIXME: Closure-library specific stuff should not be hard-coded but be 424 // configurable. 425 if (Line.Tokens.size() < 4) 426 return false; 427 auto I = Line.Tokens.begin(); 428 if (I->Tok->TokenText != "goog") 429 return false; 430 ++I; 431 if (I->Tok->isNot(tok::period)) 432 return false; 433 ++I; 434 if (I->Tok->TokenText != "scope") 435 return false; 436 ++I; 437 return I->Tok->is(tok::l_paren); 438 } 439 440 static bool ShouldBreakBeforeBrace(const FormatStyle &Style, 441 const FormatToken &InitialToken) { 442 switch (Style.BreakBeforeBraces) { 443 case FormatStyle::BS_Linux: 444 return InitialToken.isOneOf(tok::kw_namespace, tok::kw_class); 445 case FormatStyle::BS_Allman: 446 case FormatStyle::BS_GNU: 447 return true; 448 default: 449 return false; 450 } 451 } 452 453 void UnwrappedLineParser::parseChildBlock() { 454 FormatTok->BlockKind = BK_Block; 455 nextToken(); 456 { 457 bool GoogScope = 458 Style.Language == FormatStyle::LK_JavaScript && IsGoogScope(*Line); 459 ScopedLineState LineState(*this); 460 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 461 /*MustBeDeclaration=*/false); 462 Line->Level += GoogScope ? 0 : 1; 463 parseLevel(/*HasOpeningBrace=*/true); 464 Line->Level -= GoogScope ? 0 : 1; 465 } 466 nextToken(); 467 } 468 469 void UnwrappedLineParser::parsePPDirective() { 470 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 471 ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError); 472 nextToken(); 473 474 if (!FormatTok->Tok.getIdentifierInfo()) { 475 parsePPUnknown(); 476 return; 477 } 478 479 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 480 case tok::pp_define: 481 parsePPDefine(); 482 return; 483 case tok::pp_if: 484 parsePPIf(/*IfDef=*/false); 485 break; 486 case tok::pp_ifdef: 487 case tok::pp_ifndef: 488 parsePPIf(/*IfDef=*/true); 489 break; 490 case tok::pp_else: 491 parsePPElse(); 492 break; 493 case tok::pp_elif: 494 parsePPElIf(); 495 break; 496 case tok::pp_endif: 497 parsePPEndIf(); 498 break; 499 default: 500 parsePPUnknown(); 501 break; 502 } 503 } 504 505 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 506 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 507 PPStack.push_back(PP_Unreachable); 508 else 509 PPStack.push_back(PP_Conditional); 510 } 511 512 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 513 ++PPBranchLevel; 514 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 515 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 516 PPLevelBranchIndex.push_back(0); 517 PPLevelBranchCount.push_back(0); 518 } 519 PPChainBranchIndex.push(0); 520 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 521 conditionalCompilationCondition(Unreachable || Skip); 522 } 523 524 void UnwrappedLineParser::conditionalCompilationAlternative() { 525 if (!PPStack.empty()) 526 PPStack.pop_back(); 527 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 528 if (!PPChainBranchIndex.empty()) 529 ++PPChainBranchIndex.top(); 530 conditionalCompilationCondition( 531 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 532 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 533 } 534 535 void UnwrappedLineParser::conditionalCompilationEnd() { 536 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 537 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 538 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 539 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 540 } 541 } 542 // Guard against #endif's without #if. 543 if (PPBranchLevel > 0) 544 --PPBranchLevel; 545 if (!PPChainBranchIndex.empty()) 546 PPChainBranchIndex.pop(); 547 if (!PPStack.empty()) 548 PPStack.pop_back(); 549 } 550 551 void UnwrappedLineParser::parsePPIf(bool IfDef) { 552 nextToken(); 553 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 554 FormatTok->Tok.getLiteralData() != nullptr && 555 StringRef(FormatTok->Tok.getLiteralData(), 556 FormatTok->Tok.getLength()) == "0") || 557 FormatTok->Tok.is(tok::kw_false); 558 conditionalCompilationStart(!IfDef && IsLiteralFalse); 559 parsePPUnknown(); 560 } 561 562 void UnwrappedLineParser::parsePPElse() { 563 conditionalCompilationAlternative(); 564 parsePPUnknown(); 565 } 566 567 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 568 569 void UnwrappedLineParser::parsePPEndIf() { 570 conditionalCompilationEnd(); 571 parsePPUnknown(); 572 } 573 574 void UnwrappedLineParser::parsePPDefine() { 575 nextToken(); 576 577 if (FormatTok->Tok.getKind() != tok::identifier) { 578 parsePPUnknown(); 579 return; 580 } 581 nextToken(); 582 if (FormatTok->Tok.getKind() == tok::l_paren && 583 FormatTok->WhitespaceRange.getBegin() == 584 FormatTok->WhitespaceRange.getEnd()) { 585 parseParens(); 586 } 587 addUnwrappedLine(); 588 Line->Level = 1; 589 590 // Errors during a preprocessor directive can only affect the layout of the 591 // preprocessor directive, and thus we ignore them. An alternative approach 592 // would be to use the same approach we use on the file level (no 593 // re-indentation if there was a structural error) within the macro 594 // definition. 595 parseFile(); 596 } 597 598 void UnwrappedLineParser::parsePPUnknown() { 599 do { 600 nextToken(); 601 } while (!eof()); 602 addUnwrappedLine(); 603 } 604 605 // Here we blacklist certain tokens that are not usually the first token in an 606 // unwrapped line. This is used in attempt to distinguish macro calls without 607 // trailing semicolons from other constructs split to several lines. 608 static bool tokenCanStartNewLine(const clang::Token &Tok) { 609 // Semicolon can be a null-statement, l_square can be a start of a macro or 610 // a C++11 attribute, but this doesn't seem to be common. 611 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 612 Tok.isNot(tok::l_square) && 613 // Tokens that can only be used as binary operators and a part of 614 // overloaded operator names. 615 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 616 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 617 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 618 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 619 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 620 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 621 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 622 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 623 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 624 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 625 Tok.isNot(tok::lesslessequal) && 626 // Colon is used in labels, base class lists, initializer lists, 627 // range-based for loops, ternary operator, but should never be the 628 // first token in an unwrapped line. 629 Tok.isNot(tok::colon) && 630 // 'noexcept' is a trailing annotation. 631 Tok.isNot(tok::kw_noexcept); 632 } 633 634 void UnwrappedLineParser::parseStructuralElement() { 635 assert(!FormatTok->Tok.is(tok::l_brace)); 636 switch (FormatTok->Tok.getKind()) { 637 case tok::at: 638 nextToken(); 639 if (FormatTok->Tok.is(tok::l_brace)) { 640 parseBracedList(); 641 break; 642 } 643 switch (FormatTok->Tok.getObjCKeywordID()) { 644 case tok::objc_public: 645 case tok::objc_protected: 646 case tok::objc_package: 647 case tok::objc_private: 648 return parseAccessSpecifier(); 649 case tok::objc_interface: 650 case tok::objc_implementation: 651 return parseObjCInterfaceOrImplementation(); 652 case tok::objc_protocol: 653 return parseObjCProtocol(); 654 case tok::objc_end: 655 return; // Handled by the caller. 656 case tok::objc_optional: 657 case tok::objc_required: 658 nextToken(); 659 addUnwrappedLine(); 660 return; 661 case tok::objc_try: 662 // This branch isn't strictly necessary (the kw_try case below would 663 // do this too after the tok::at is parsed above). But be explicit. 664 parseTryCatch(); 665 return; 666 default: 667 break; 668 } 669 break; 670 case tok::kw_asm: 671 nextToken(); 672 if (FormatTok->is(tok::l_brace)) { 673 nextToken(); 674 while (FormatTok && FormatTok->isNot(tok::eof)) { 675 if (FormatTok->is(tok::r_brace)) { 676 nextToken(); 677 break; 678 } 679 FormatTok->Finalized = true; 680 nextToken(); 681 } 682 } 683 break; 684 case tok::kw_namespace: 685 parseNamespace(); 686 return; 687 case tok::kw_inline: 688 nextToken(); 689 if (FormatTok->Tok.is(tok::kw_namespace)) { 690 parseNamespace(); 691 return; 692 } 693 break; 694 case tok::kw_public: 695 case tok::kw_protected: 696 case tok::kw_private: 697 if (Style.Language == FormatStyle::LK_Java || 698 Style.Language == FormatStyle::LK_JavaScript) 699 nextToken(); 700 else 701 parseAccessSpecifier(); 702 return; 703 case tok::kw_if: 704 parseIfThenElse(); 705 return; 706 case tok::kw_for: 707 case tok::kw_while: 708 parseForOrWhileLoop(); 709 return; 710 case tok::kw_do: 711 parseDoWhile(); 712 return; 713 case tok::kw_switch: 714 parseSwitch(); 715 return; 716 case tok::kw_default: 717 nextToken(); 718 parseLabel(); 719 return; 720 case tok::kw_case: 721 parseCaseLabel(); 722 return; 723 case tok::kw_try: 724 case tok::kw___try: 725 parseTryCatch(); 726 return; 727 case tok::kw_extern: 728 nextToken(); 729 if (FormatTok->Tok.is(tok::string_literal)) { 730 nextToken(); 731 if (FormatTok->Tok.is(tok::l_brace)) { 732 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 733 addUnwrappedLine(); 734 return; 735 } 736 } 737 break; 738 case tok::kw_export: 739 if (Style.Language == FormatStyle::LK_JavaScript) { 740 parseJavaScriptEs6ImportExport(); 741 return; 742 } 743 break; 744 case tok::identifier: 745 if (FormatTok->IsForEachMacro) { 746 parseForOrWhileLoop(); 747 return; 748 } 749 if (Style.Language == FormatStyle::LK_JavaScript && 750 FormatTok->is(Keywords.kw_import)) { 751 parseJavaScriptEs6ImportExport(); 752 return; 753 } 754 // In all other cases, parse the declaration. 755 break; 756 default: 757 break; 758 } 759 do { 760 switch (FormatTok->Tok.getKind()) { 761 case tok::at: 762 nextToken(); 763 if (FormatTok->Tok.is(tok::l_brace)) 764 parseBracedList(); 765 break; 766 case tok::kw_enum: 767 parseEnum(); 768 break; 769 case tok::kw_typedef: 770 nextToken(); 771 if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS, 772 Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS)) 773 parseEnum(); 774 break; 775 case tok::kw_struct: 776 case tok::kw_union: 777 case tok::kw_class: 778 parseRecord(); 779 // A record declaration or definition is always the start of a structural 780 // element. 781 break; 782 case tok::period: 783 nextToken(); 784 // In Java, classes have an implicit static member "class". 785 if (Style.Language == FormatStyle::LK_Java && FormatTok && 786 FormatTok->is(tok::kw_class)) 787 nextToken(); 788 break; 789 case tok::semi: 790 nextToken(); 791 addUnwrappedLine(); 792 return; 793 case tok::r_brace: 794 addUnwrappedLine(); 795 return; 796 case tok::l_paren: 797 parseParens(); 798 break; 799 case tok::caret: 800 nextToken(); 801 if (FormatTok->Tok.isAnyIdentifier() || 802 FormatTok->isSimpleTypeSpecifier()) 803 nextToken(); 804 if (FormatTok->is(tok::l_paren)) 805 parseParens(); 806 if (FormatTok->is(tok::l_brace)) 807 parseChildBlock(); 808 break; 809 case tok::l_brace: 810 if (!tryToParseBracedList()) { 811 // A block outside of parentheses must be the last part of a 812 // structural element. 813 // FIXME: Figure out cases where this is not true, and add projections 814 // for them (the one we know is missing are lambdas). 815 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 816 addUnwrappedLine(); 817 FormatTok->Type = TT_FunctionLBrace; 818 parseBlock(/*MustBeDeclaration=*/false); 819 addUnwrappedLine(); 820 return; 821 } 822 // Otherwise this was a braced init list, and the structural 823 // element continues. 824 break; 825 case tok::kw_try: 826 // We arrive here when parsing function-try blocks. 827 parseTryCatch(); 828 return; 829 case tok::identifier: { 830 StringRef Text = FormatTok->TokenText; 831 // Parse function literal unless 'function' is the first token in a line 832 // in which case this should be treated as a free-standing function. 833 if (Style.Language == FormatStyle::LK_JavaScript && Text == "function" && 834 Line->Tokens.size() > 0) { 835 tryToParseJSFunction(); 836 break; 837 } 838 nextToken(); 839 if (Line->Tokens.size() == 1 && 840 // JS doesn't have macros, and within classes colons indicate fields, 841 // not labels. 842 (Style.Language != FormatStyle::LK_JavaScript || 843 !Line->MustBeDeclaration)) { 844 if (FormatTok->Tok.is(tok::colon)) { 845 parseLabel(); 846 return; 847 } 848 // Recognize function-like macro usages without trailing semicolon as 849 // well as free-standing macros like Q_OBJECT. 850 bool FunctionLike = FormatTok->is(tok::l_paren); 851 if (FunctionLike) 852 parseParens(); 853 if (FormatTok->NewlinesBefore > 0 && 854 (Text.size() >= 5 || FunctionLike) && 855 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 856 addUnwrappedLine(); 857 return; 858 } 859 } 860 break; 861 } 862 case tok::equal: 863 nextToken(); 864 if (FormatTok->Tok.is(tok::l_brace)) { 865 parseBracedList(); 866 } 867 break; 868 case tok::l_square: 869 parseSquare(); 870 break; 871 case tok::kw_new: 872 parseNew(); 873 break; 874 default: 875 nextToken(); 876 break; 877 } 878 } while (!eof()); 879 } 880 881 bool UnwrappedLineParser::tryToParseLambda() { 882 // FIXME: This is a dirty way to access the previous token. Find a better 883 // solution. 884 if (!Line->Tokens.empty() && 885 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator, 886 tok::kw_new, tok::kw_delete) || 887 Line->Tokens.back().Tok->closesScope() || 888 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 889 nextToken(); 890 return false; 891 } 892 assert(FormatTok->is(tok::l_square)); 893 FormatToken &LSquare = *FormatTok; 894 if (!tryToParseLambdaIntroducer()) 895 return false; 896 897 while (FormatTok->isNot(tok::l_brace)) { 898 if (FormatTok->isSimpleTypeSpecifier()) { 899 nextToken(); 900 continue; 901 } 902 switch (FormatTok->Tok.getKind()) { 903 case tok::l_brace: 904 break; 905 case tok::l_paren: 906 parseParens(); 907 break; 908 case tok::amp: 909 case tok::star: 910 case tok::kw_const: 911 case tok::comma: 912 case tok::less: 913 case tok::greater: 914 case tok::identifier: 915 case tok::coloncolon: 916 case tok::kw_mutable: 917 nextToken(); 918 break; 919 case tok::arrow: 920 FormatTok->Type = TT_TrailingReturnArrow; 921 nextToken(); 922 break; 923 default: 924 return true; 925 } 926 } 927 LSquare.Type = TT_LambdaLSquare; 928 parseChildBlock(); 929 return true; 930 } 931 932 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 933 nextToken(); 934 if (FormatTok->is(tok::equal)) { 935 nextToken(); 936 if (FormatTok->is(tok::r_square)) { 937 nextToken(); 938 return true; 939 } 940 if (FormatTok->isNot(tok::comma)) 941 return false; 942 nextToken(); 943 } else if (FormatTok->is(tok::amp)) { 944 nextToken(); 945 if (FormatTok->is(tok::r_square)) { 946 nextToken(); 947 return true; 948 } 949 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 950 return false; 951 } 952 if (FormatTok->is(tok::comma)) 953 nextToken(); 954 } else if (FormatTok->is(tok::r_square)) { 955 nextToken(); 956 return true; 957 } 958 do { 959 if (FormatTok->is(tok::amp)) 960 nextToken(); 961 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 962 return false; 963 nextToken(); 964 if (FormatTok->is(tok::ellipsis)) 965 nextToken(); 966 if (FormatTok->is(tok::comma)) { 967 nextToken(); 968 } else if (FormatTok->is(tok::r_square)) { 969 nextToken(); 970 return true; 971 } else { 972 return false; 973 } 974 } while (!eof()); 975 return false; 976 } 977 978 void UnwrappedLineParser::tryToParseJSFunction() { 979 nextToken(); 980 981 // Consume function name. 982 if (FormatTok->is(tok::identifier)) 983 nextToken(); 984 985 if (FormatTok->isNot(tok::l_paren)) 986 return; 987 nextToken(); 988 while (FormatTok->isNot(tok::l_brace)) { 989 // Err on the side of caution in order to avoid consuming the full file in 990 // case of incomplete code. 991 if (!FormatTok->isOneOf(tok::identifier, tok::comma, tok::r_paren, 992 tok::comment)) 993 return; 994 nextToken(); 995 } 996 parseChildBlock(); 997 } 998 999 bool UnwrappedLineParser::tryToParseBracedList() { 1000 if (FormatTok->BlockKind == BK_Unknown) 1001 calculateBraceTypes(); 1002 assert(FormatTok->BlockKind != BK_Unknown); 1003 if (FormatTok->BlockKind == BK_Block) 1004 return false; 1005 parseBracedList(); 1006 return true; 1007 } 1008 1009 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 1010 bool HasError = false; 1011 nextToken(); 1012 1013 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 1014 // replace this by using parseAssigmentExpression() inside. 1015 do { 1016 if (Style.Language == FormatStyle::LK_JavaScript && 1017 FormatTok->is(Keywords.kw_function)) { 1018 tryToParseJSFunction(); 1019 continue; 1020 } 1021 switch (FormatTok->Tok.getKind()) { 1022 case tok::caret: 1023 nextToken(); 1024 if (FormatTok->is(tok::l_brace)) { 1025 parseChildBlock(); 1026 } 1027 break; 1028 case tok::l_square: 1029 tryToParseLambda(); 1030 break; 1031 case tok::l_brace: 1032 // Assume there are no blocks inside a braced init list apart 1033 // from the ones we explicitly parse out (like lambdas). 1034 FormatTok->BlockKind = BK_BracedInit; 1035 parseBracedList(); 1036 break; 1037 case tok::r_brace: 1038 nextToken(); 1039 return !HasError; 1040 case tok::semi: 1041 HasError = true; 1042 if (!ContinueOnSemicolons) 1043 return !HasError; 1044 nextToken(); 1045 break; 1046 case tok::comma: 1047 nextToken(); 1048 break; 1049 default: 1050 nextToken(); 1051 break; 1052 } 1053 } while (!eof()); 1054 return false; 1055 } 1056 1057 void UnwrappedLineParser::parseParens() { 1058 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1059 nextToken(); 1060 do { 1061 switch (FormatTok->Tok.getKind()) { 1062 case tok::l_paren: 1063 parseParens(); 1064 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) 1065 parseChildBlock(); 1066 break; 1067 case tok::r_paren: 1068 nextToken(); 1069 return; 1070 case tok::r_brace: 1071 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1072 return; 1073 case tok::l_square: 1074 tryToParseLambda(); 1075 break; 1076 case tok::l_brace: 1077 if (!tryToParseBracedList()) { 1078 parseChildBlock(); 1079 } 1080 break; 1081 case tok::at: 1082 nextToken(); 1083 if (FormatTok->Tok.is(tok::l_brace)) 1084 parseBracedList(); 1085 break; 1086 case tok::identifier: 1087 if (Style.Language == FormatStyle::LK_JavaScript && 1088 FormatTok->is(Keywords.kw_function)) 1089 tryToParseJSFunction(); 1090 else 1091 nextToken(); 1092 break; 1093 default: 1094 nextToken(); 1095 break; 1096 } 1097 } while (!eof()); 1098 } 1099 1100 void UnwrappedLineParser::parseSquare() { 1101 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1102 if (tryToParseLambda()) 1103 return; 1104 do { 1105 switch (FormatTok->Tok.getKind()) { 1106 case tok::l_paren: 1107 parseParens(); 1108 break; 1109 case tok::r_square: 1110 nextToken(); 1111 return; 1112 case tok::r_brace: 1113 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1114 return; 1115 case tok::l_square: 1116 parseSquare(); 1117 break; 1118 case tok::l_brace: { 1119 if (!tryToParseBracedList()) { 1120 parseChildBlock(); 1121 } 1122 break; 1123 } 1124 case tok::at: 1125 nextToken(); 1126 if (FormatTok->Tok.is(tok::l_brace)) 1127 parseBracedList(); 1128 break; 1129 default: 1130 nextToken(); 1131 break; 1132 } 1133 } while (!eof()); 1134 } 1135 1136 void UnwrappedLineParser::parseIfThenElse() { 1137 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1138 nextToken(); 1139 if (FormatTok->Tok.is(tok::l_paren)) 1140 parseParens(); 1141 bool NeedsUnwrappedLine = false; 1142 if (FormatTok->Tok.is(tok::l_brace)) { 1143 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1144 parseBlock(/*MustBeDeclaration=*/false); 1145 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1146 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1147 addUnwrappedLine(); 1148 } else { 1149 NeedsUnwrappedLine = true; 1150 } 1151 } else { 1152 addUnwrappedLine(); 1153 ++Line->Level; 1154 parseStructuralElement(); 1155 --Line->Level; 1156 } 1157 if (FormatTok->Tok.is(tok::kw_else)) { 1158 if (Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) 1159 addUnwrappedLine(); 1160 nextToken(); 1161 if (FormatTok->Tok.is(tok::l_brace)) { 1162 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1163 parseBlock(/*MustBeDeclaration=*/false); 1164 addUnwrappedLine(); 1165 } else if (FormatTok->Tok.is(tok::kw_if)) { 1166 parseIfThenElse(); 1167 } else { 1168 addUnwrappedLine(); 1169 ++Line->Level; 1170 parseStructuralElement(); 1171 --Line->Level; 1172 } 1173 } else if (NeedsUnwrappedLine) { 1174 addUnwrappedLine(); 1175 } 1176 } 1177 1178 void UnwrappedLineParser::parseTryCatch() { 1179 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected"); 1180 nextToken(); 1181 bool NeedsUnwrappedLine = false; 1182 if (FormatTok->is(tok::colon)) { 1183 // We are in a function try block, what comes is an initializer list. 1184 nextToken(); 1185 while (FormatTok->is(tok::identifier)) { 1186 nextToken(); 1187 if (FormatTok->is(tok::l_paren)) 1188 parseParens(); 1189 else 1190 StructuralError = true; 1191 if (FormatTok->is(tok::comma)) 1192 nextToken(); 1193 } 1194 } 1195 // Parse try with resource. 1196 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) { 1197 parseParens(); 1198 } 1199 if (FormatTok->is(tok::l_brace)) { 1200 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1201 parseBlock(/*MustBeDeclaration=*/false); 1202 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1203 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1204 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1205 addUnwrappedLine(); 1206 } else { 1207 NeedsUnwrappedLine = true; 1208 } 1209 } else if (!FormatTok->is(tok::kw_catch)) { 1210 // The C++ standard requires a compound-statement after a try. 1211 // If there's none, we try to assume there's a structuralElement 1212 // and try to continue. 1213 StructuralError = true; 1214 addUnwrappedLine(); 1215 ++Line->Level; 1216 parseStructuralElement(); 1217 --Line->Level; 1218 } 1219 while (1) { 1220 if (FormatTok->is(tok::at)) 1221 nextToken(); 1222 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, 1223 tok::kw___finally) || 1224 ((Style.Language == FormatStyle::LK_Java || 1225 Style.Language == FormatStyle::LK_JavaScript) && 1226 FormatTok->is(Keywords.kw_finally)) || 1227 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) || 1228 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally)))) 1229 break; 1230 nextToken(); 1231 while (FormatTok->isNot(tok::l_brace)) { 1232 if (FormatTok->is(tok::l_paren)) { 1233 parseParens(); 1234 continue; 1235 } 1236 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) 1237 return; 1238 nextToken(); 1239 } 1240 NeedsUnwrappedLine = false; 1241 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1242 parseBlock(/*MustBeDeclaration=*/false); 1243 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1244 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1245 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1246 addUnwrappedLine(); 1247 } else { 1248 NeedsUnwrappedLine = true; 1249 } 1250 } 1251 if (NeedsUnwrappedLine) { 1252 addUnwrappedLine(); 1253 } 1254 } 1255 1256 void UnwrappedLineParser::parseNamespace() { 1257 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1258 1259 const FormatToken &InitialToken = *FormatTok; 1260 nextToken(); 1261 if (FormatTok->Tok.is(tok::identifier)) 1262 nextToken(); 1263 if (FormatTok->Tok.is(tok::l_brace)) { 1264 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1265 addUnwrappedLine(); 1266 1267 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1268 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1269 DeclarationScopeStack.size() > 1); 1270 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1271 // Munch the semicolon after a namespace. This is more common than one would 1272 // think. Puttin the semicolon into its own line is very ugly. 1273 if (FormatTok->Tok.is(tok::semi)) 1274 nextToken(); 1275 addUnwrappedLine(); 1276 } 1277 // FIXME: Add error handling. 1278 } 1279 1280 void UnwrappedLineParser::parseNew() { 1281 assert(FormatTok->is(tok::kw_new) && "'new' expected"); 1282 nextToken(); 1283 if (Style.Language != FormatStyle::LK_Java) 1284 return; 1285 1286 // In Java, we can parse everything up to the parens, which aren't optional. 1287 do { 1288 // There should not be a ;, { or } before the new's open paren. 1289 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) 1290 return; 1291 1292 // Consume the parens. 1293 if (FormatTok->is(tok::l_paren)) { 1294 parseParens(); 1295 1296 // If there is a class body of an anonymous class, consume that as child. 1297 if (FormatTok->is(tok::l_brace)) 1298 parseChildBlock(); 1299 return; 1300 } 1301 nextToken(); 1302 } while (!eof()); 1303 } 1304 1305 void UnwrappedLineParser::parseForOrWhileLoop() { 1306 assert((FormatTok->Tok.is(tok::kw_for) || FormatTok->Tok.is(tok::kw_while) || 1307 FormatTok->IsForEachMacro) && 1308 "'for', 'while' or foreach macro expected"); 1309 nextToken(); 1310 if (FormatTok->Tok.is(tok::l_paren)) 1311 parseParens(); 1312 if (FormatTok->Tok.is(tok::l_brace)) { 1313 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1314 parseBlock(/*MustBeDeclaration=*/false); 1315 addUnwrappedLine(); 1316 } else { 1317 addUnwrappedLine(); 1318 ++Line->Level; 1319 parseStructuralElement(); 1320 --Line->Level; 1321 } 1322 } 1323 1324 void UnwrappedLineParser::parseDoWhile() { 1325 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1326 nextToken(); 1327 if (FormatTok->Tok.is(tok::l_brace)) { 1328 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1329 parseBlock(/*MustBeDeclaration=*/false); 1330 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1331 addUnwrappedLine(); 1332 } else { 1333 addUnwrappedLine(); 1334 ++Line->Level; 1335 parseStructuralElement(); 1336 --Line->Level; 1337 } 1338 1339 // FIXME: Add error handling. 1340 if (!FormatTok->Tok.is(tok::kw_while)) { 1341 addUnwrappedLine(); 1342 return; 1343 } 1344 1345 nextToken(); 1346 parseStructuralElement(); 1347 } 1348 1349 void UnwrappedLineParser::parseLabel() { 1350 nextToken(); 1351 unsigned OldLineLevel = Line->Level; 1352 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1353 --Line->Level; 1354 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1355 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1356 parseBlock(/*MustBeDeclaration=*/false); 1357 if (FormatTok->Tok.is(tok::kw_break)) { 1358 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1359 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1360 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1361 addUnwrappedLine(); 1362 } 1363 parseStructuralElement(); 1364 } 1365 addUnwrappedLine(); 1366 } else { 1367 addUnwrappedLine(); 1368 } 1369 Line->Level = OldLineLevel; 1370 } 1371 1372 void UnwrappedLineParser::parseCaseLabel() { 1373 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1374 // FIXME: fix handling of complex expressions here. 1375 do { 1376 nextToken(); 1377 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1378 parseLabel(); 1379 } 1380 1381 void UnwrappedLineParser::parseSwitch() { 1382 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1383 nextToken(); 1384 if (FormatTok->Tok.is(tok::l_paren)) 1385 parseParens(); 1386 if (FormatTok->Tok.is(tok::l_brace)) { 1387 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1388 parseBlock(/*MustBeDeclaration=*/false); 1389 addUnwrappedLine(); 1390 } else { 1391 addUnwrappedLine(); 1392 ++Line->Level; 1393 parseStructuralElement(); 1394 --Line->Level; 1395 } 1396 } 1397 1398 void UnwrappedLineParser::parseAccessSpecifier() { 1399 nextToken(); 1400 // Understand Qt's slots. 1401 if (FormatTok->is(tok::identifier) && 1402 (FormatTok->TokenText == "slots" || FormatTok->TokenText == "Q_SLOTS")) 1403 nextToken(); 1404 // Otherwise, we don't know what it is, and we'd better keep the next token. 1405 if (FormatTok->Tok.is(tok::colon)) 1406 nextToken(); 1407 addUnwrappedLine(); 1408 } 1409 1410 void UnwrappedLineParser::parseEnum() { 1411 // Won't be 'enum' for NS_ENUMs. 1412 if (FormatTok->Tok.is(tok::kw_enum)) 1413 nextToken(); 1414 1415 // Eat up enum class ... 1416 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1417 nextToken(); 1418 while (FormatTok->Tok.getIdentifierInfo() || 1419 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less, 1420 tok::greater, tok::comma, tok::question)) { 1421 nextToken(); 1422 // We can have macros or attributes in between 'enum' and the enum name. 1423 if (FormatTok->is(tok::l_paren)) 1424 parseParens(); 1425 if (FormatTok->is(tok::identifier)) 1426 nextToken(); 1427 } 1428 1429 // Just a declaration or something is wrong. 1430 if (FormatTok->isNot(tok::l_brace)) 1431 return; 1432 FormatTok->BlockKind = BK_Block; 1433 1434 if (Style.Language == FormatStyle::LK_Java) { 1435 // Java enums are different. 1436 parseJavaEnumBody(); 1437 return; 1438 } 1439 1440 // Parse enum body. 1441 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1442 if (HasError) { 1443 if (FormatTok->is(tok::semi)) 1444 nextToken(); 1445 addUnwrappedLine(); 1446 } 1447 1448 // We fall through to parsing a structural element afterwards, so that in 1449 // enum A {} n, m; 1450 // "} n, m;" will end up in one unwrapped line. 1451 } 1452 1453 void UnwrappedLineParser::parseJavaEnumBody() { 1454 // Determine whether the enum is simple, i.e. does not have a semicolon or 1455 // constants with class bodies. Simple enums can be formatted like braced 1456 // lists, contracted to a single line, etc. 1457 unsigned StoredPosition = Tokens->getPosition(); 1458 bool IsSimple = true; 1459 FormatToken *Tok = Tokens->getNextToken(); 1460 while (Tok) { 1461 if (Tok->is(tok::r_brace)) 1462 break; 1463 if (Tok->isOneOf(tok::l_brace, tok::semi)) { 1464 IsSimple = false; 1465 break; 1466 } 1467 // FIXME: This will also mark enums with braces in the arguments to enum 1468 // constants as "not simple". This is probably fine in practice, though. 1469 Tok = Tokens->getNextToken(); 1470 } 1471 FormatTok = Tokens->setPosition(StoredPosition); 1472 1473 if (IsSimple) { 1474 parseBracedList(); 1475 addUnwrappedLine(); 1476 return; 1477 } 1478 1479 // Parse the body of a more complex enum. 1480 // First add a line for everything up to the "{". 1481 nextToken(); 1482 addUnwrappedLine(); 1483 ++Line->Level; 1484 1485 // Parse the enum constants. 1486 while (FormatTok) { 1487 if (FormatTok->is(tok::l_brace)) { 1488 // Parse the constant's class body. 1489 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1490 /*MunchSemi=*/false); 1491 } else if (FormatTok->is(tok::l_paren)) { 1492 parseParens(); 1493 } else if (FormatTok->is(tok::comma)) { 1494 nextToken(); 1495 addUnwrappedLine(); 1496 } else if (FormatTok->is(tok::semi)) { 1497 nextToken(); 1498 addUnwrappedLine(); 1499 break; 1500 } else if (FormatTok->is(tok::r_brace)) { 1501 addUnwrappedLine(); 1502 break; 1503 } else { 1504 nextToken(); 1505 } 1506 } 1507 1508 // Parse the class body after the enum's ";" if any. 1509 parseLevel(/*HasOpeningBrace=*/true); 1510 nextToken(); 1511 --Line->Level; 1512 addUnwrappedLine(); 1513 } 1514 1515 void UnwrappedLineParser::parseRecord() { 1516 const FormatToken &InitialToken = *FormatTok; 1517 nextToken(); 1518 if (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::kw___attribute, 1519 tok::kw___declspec, tok::kw_alignas)) { 1520 nextToken(); 1521 // We can have macros or attributes in between 'class' and the class name. 1522 if (FormatTok->Tok.is(tok::l_paren)) { 1523 parseParens(); 1524 } 1525 // The actual identifier can be a nested name specifier, and in macros 1526 // it is often token-pasted. 1527 while (FormatTok->is(tok::identifier) || FormatTok->is(tok::coloncolon) || 1528 FormatTok->is(tok::hashhash) || 1529 (Style.Language == FormatStyle::LK_Java && 1530 FormatTok->isOneOf(tok::period, tok::comma))) 1531 nextToken(); 1532 1533 // Note that parsing away template declarations here leads to incorrectly 1534 // accepting function declarations as record declarations. 1535 // In general, we cannot solve this problem. Consider: 1536 // class A<int> B() {} 1537 // which can be a function definition or a class definition when B() is a 1538 // macro. If we find enough real-world cases where this is a problem, we 1539 // can parse for the 'template' keyword in the beginning of the statement, 1540 // and thus rule out the record production in case there is no template 1541 // (this would still leave us with an ambiguity between template function 1542 // and class declarations). 1543 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1544 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1545 if (FormatTok->Tok.is(tok::semi)) 1546 return; 1547 nextToken(); 1548 } 1549 } 1550 } 1551 if (FormatTok->Tok.is(tok::l_brace)) { 1552 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1553 addUnwrappedLine(); 1554 1555 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1556 /*MunchSemi=*/false); 1557 } 1558 // We fall through to parsing a structural element afterwards, so 1559 // class A {} n, m; 1560 // will end up in one unwrapped line. 1561 // This does not apply for Java. 1562 if (Style.Language == FormatStyle::LK_Java || 1563 Style.Language == FormatStyle::LK_JavaScript) 1564 addUnwrappedLine(); 1565 } 1566 1567 void UnwrappedLineParser::parseObjCProtocolList() { 1568 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1569 do 1570 nextToken(); 1571 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1572 nextToken(); // Skip '>'. 1573 } 1574 1575 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1576 do { 1577 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1578 nextToken(); 1579 addUnwrappedLine(); 1580 break; 1581 } 1582 if (FormatTok->is(tok::l_brace)) { 1583 parseBlock(/*MustBeDeclaration=*/false); 1584 // In ObjC interfaces, nothing should be following the "}". 1585 addUnwrappedLine(); 1586 } else if (FormatTok->is(tok::r_brace)) { 1587 // Ignore stray "}". parseStructuralElement doesn't consume them. 1588 nextToken(); 1589 addUnwrappedLine(); 1590 } else { 1591 parseStructuralElement(); 1592 } 1593 } while (!eof()); 1594 } 1595 1596 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1597 nextToken(); 1598 nextToken(); // interface name 1599 1600 // @interface can be followed by either a base class, or a category. 1601 if (FormatTok->Tok.is(tok::colon)) { 1602 nextToken(); 1603 nextToken(); // base class name 1604 } else if (FormatTok->Tok.is(tok::l_paren)) 1605 // Skip category, if present. 1606 parseParens(); 1607 1608 if (FormatTok->Tok.is(tok::less)) 1609 parseObjCProtocolList(); 1610 1611 if (FormatTok->Tok.is(tok::l_brace)) { 1612 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1613 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1614 addUnwrappedLine(); 1615 parseBlock(/*MustBeDeclaration=*/true); 1616 } 1617 1618 // With instance variables, this puts '}' on its own line. Without instance 1619 // variables, this ends the @interface line. 1620 addUnwrappedLine(); 1621 1622 parseObjCUntilAtEnd(); 1623 } 1624 1625 void UnwrappedLineParser::parseObjCProtocol() { 1626 nextToken(); 1627 nextToken(); // protocol name 1628 1629 if (FormatTok->Tok.is(tok::less)) 1630 parseObjCProtocolList(); 1631 1632 // Check for protocol declaration. 1633 if (FormatTok->Tok.is(tok::semi)) { 1634 nextToken(); 1635 return addUnwrappedLine(); 1636 } 1637 1638 addUnwrappedLine(); 1639 parseObjCUntilAtEnd(); 1640 } 1641 1642 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { 1643 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export)); 1644 nextToken(); 1645 1646 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, Keywords.kw_function, 1647 Keywords.kw_var)) 1648 return; // Fall through to parsing the corresponding structure. 1649 1650 if (FormatTok->is(tok::kw_default)) { 1651 nextToken(); // export default ..., fall through after eating 'default'. 1652 return; 1653 } 1654 1655 if (FormatTok->is(tok::l_brace)) { 1656 FormatTok->BlockKind = BK_Block; 1657 parseBracedList(); 1658 } 1659 1660 while (!eof() && FormatTok->isNot(tok::semi) && 1661 FormatTok->isNot(tok::l_brace)) { 1662 nextToken(); 1663 } 1664 } 1665 1666 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1667 StringRef Prefix = "") { 1668 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1669 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1670 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1671 E = Line.Tokens.end(); 1672 I != E; ++I) { 1673 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1674 } 1675 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1676 E = Line.Tokens.end(); 1677 I != E; ++I) { 1678 const UnwrappedLineNode &Node = *I; 1679 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1680 I = Node.Children.begin(), 1681 E = Node.Children.end(); 1682 I != E; ++I) { 1683 printDebugInfo(*I, "\nChild: "); 1684 } 1685 } 1686 llvm::dbgs() << "\n"; 1687 } 1688 1689 void UnwrappedLineParser::addUnwrappedLine() { 1690 if (Line->Tokens.empty()) 1691 return; 1692 DEBUG({ 1693 if (CurrentLines == &Lines) 1694 printDebugInfo(*Line); 1695 }); 1696 CurrentLines->push_back(*Line); 1697 Line->Tokens.clear(); 1698 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1699 for (SmallVectorImpl<UnwrappedLine>::iterator 1700 I = PreprocessorDirectives.begin(), 1701 E = PreprocessorDirectives.end(); 1702 I != E; ++I) { 1703 CurrentLines->push_back(*I); 1704 } 1705 PreprocessorDirectives.clear(); 1706 } 1707 } 1708 1709 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1710 1711 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1712 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1713 FormatTok.NewlinesBefore > 0; 1714 } 1715 1716 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1717 bool JustComments = Line->Tokens.empty(); 1718 for (SmallVectorImpl<FormatToken *>::const_iterator 1719 I = CommentsBeforeNextToken.begin(), 1720 E = CommentsBeforeNextToken.end(); 1721 I != E; ++I) { 1722 if (isOnNewLine(**I) && JustComments) { 1723 addUnwrappedLine(); 1724 } 1725 pushToken(*I); 1726 } 1727 if (NewlineBeforeNext && JustComments) { 1728 addUnwrappedLine(); 1729 } 1730 CommentsBeforeNextToken.clear(); 1731 } 1732 1733 void UnwrappedLineParser::nextToken() { 1734 if (eof()) 1735 return; 1736 flushComments(isOnNewLine(*FormatTok)); 1737 pushToken(FormatTok); 1738 readToken(); 1739 } 1740 1741 void UnwrappedLineParser::readToken() { 1742 bool CommentsInCurrentLine = true; 1743 do { 1744 FormatTok = Tokens->getNextToken(); 1745 assert(FormatTok); 1746 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1747 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1748 // If there is an unfinished unwrapped line, we flush the preprocessor 1749 // directives only after that unwrapped line was finished later. 1750 bool SwitchToPreprocessorLines = !Line->Tokens.empty(); 1751 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1752 // Comments stored before the preprocessor directive need to be output 1753 // before the preprocessor directive, at the same level as the 1754 // preprocessor directive, as we consider them to apply to the directive. 1755 flushComments(isOnNewLine(*FormatTok)); 1756 parsePPDirective(); 1757 } 1758 while (FormatTok->Type == TT_ConflictStart || 1759 FormatTok->Type == TT_ConflictEnd || 1760 FormatTok->Type == TT_ConflictAlternative) { 1761 if (FormatTok->Type == TT_ConflictStart) { 1762 conditionalCompilationStart(/*Unreachable=*/false); 1763 } else if (FormatTok->Type == TT_ConflictAlternative) { 1764 conditionalCompilationAlternative(); 1765 } else if (FormatTok->Type == TT_ConflictEnd) { 1766 conditionalCompilationEnd(); 1767 } 1768 FormatTok = Tokens->getNextToken(); 1769 FormatTok->MustBreakBefore = true; 1770 } 1771 1772 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1773 !Line->InPPDirective) { 1774 continue; 1775 } 1776 1777 if (!FormatTok->Tok.is(tok::comment)) 1778 return; 1779 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1780 CommentsInCurrentLine = false; 1781 } 1782 if (CommentsInCurrentLine) { 1783 pushToken(FormatTok); 1784 } else { 1785 CommentsBeforeNextToken.push_back(FormatTok); 1786 } 1787 } while (!eof()); 1788 } 1789 1790 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1791 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1792 if (MustBreakBeforeNextToken) { 1793 Line->Tokens.back().Tok->MustBreakBefore = true; 1794 MustBreakBeforeNextToken = false; 1795 } 1796 } 1797 1798 } // end namespace format 1799 } // end namespace clang 1800