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 #include "llvm/Support/raw_ostream.h" 20 21 #define DEBUG_TYPE "format-parser" 22 23 namespace clang { 24 namespace format { 25 26 class FormatTokenSource { 27 public: 28 virtual ~FormatTokenSource() {} 29 virtual FormatToken *getNextToken() = 0; 30 31 virtual unsigned getPosition() = 0; 32 virtual FormatToken *setPosition(unsigned Position) = 0; 33 }; 34 35 namespace { 36 37 class ScopedDeclarationState { 38 public: 39 ScopedDeclarationState(UnwrappedLine &Line, std::vector<bool> &Stack, 40 bool MustBeDeclaration) 41 : Line(Line), Stack(Stack) { 42 Line.MustBeDeclaration = MustBeDeclaration; 43 Stack.push_back(MustBeDeclaration); 44 } 45 ~ScopedDeclarationState() { 46 Stack.pop_back(); 47 if (!Stack.empty()) 48 Line.MustBeDeclaration = Stack.back(); 49 else 50 Line.MustBeDeclaration = true; 51 } 52 53 private: 54 UnwrappedLine &Line; 55 std::vector<bool> &Stack; 56 }; 57 58 class ScopedMacroState : public FormatTokenSource { 59 public: 60 ScopedMacroState(UnwrappedLine &Line, FormatTokenSource *&TokenSource, 61 FormatToken *&ResetToken, bool &StructuralError) 62 : Line(Line), TokenSource(TokenSource), ResetToken(ResetToken), 63 PreviousLineLevel(Line.Level), PreviousTokenSource(TokenSource), 64 StructuralError(StructuralError), 65 PreviousStructuralError(StructuralError), Token(nullptr) { 66 TokenSource = this; 67 Line.Level = 0; 68 Line.InPPDirective = true; 69 } 70 71 ~ScopedMacroState() override { 72 TokenSource = PreviousTokenSource; 73 ResetToken = Token; 74 Line.InPPDirective = false; 75 Line.Level = PreviousLineLevel; 76 StructuralError = PreviousStructuralError; 77 } 78 79 FormatToken *getNextToken() override { 80 // The \c UnwrappedLineParser guards against this by never calling 81 // \c getNextToken() after it has encountered the first eof token. 82 assert(!eof()); 83 Token = PreviousTokenSource->getNextToken(); 84 if (eof()) 85 return getFakeEOF(); 86 return Token; 87 } 88 89 unsigned getPosition() override { return PreviousTokenSource->getPosition(); } 90 91 FormatToken *setPosition(unsigned Position) override { 92 Token = PreviousTokenSource->setPosition(Position); 93 return Token; 94 } 95 96 private: 97 bool eof() { return Token && Token->HasUnescapedNewline; } 98 99 FormatToken *getFakeEOF() { 100 static bool EOFInitialized = false; 101 static FormatToken FormatTok; 102 if (!EOFInitialized) { 103 FormatTok.Tok.startToken(); 104 FormatTok.Tok.setKind(tok::eof); 105 EOFInitialized = true; 106 } 107 return &FormatTok; 108 } 109 110 UnwrappedLine &Line; 111 FormatTokenSource *&TokenSource; 112 FormatToken *&ResetToken; 113 unsigned PreviousLineLevel; 114 FormatTokenSource *PreviousTokenSource; 115 bool &StructuralError; 116 bool PreviousStructuralError; 117 118 FormatToken *Token; 119 }; 120 121 } // end anonymous namespace 122 123 class ScopedLineState { 124 public: 125 ScopedLineState(UnwrappedLineParser &Parser, 126 bool SwitchToPreprocessorLines = false) 127 : Parser(Parser), OriginalLines(Parser.CurrentLines) { 128 if (SwitchToPreprocessorLines) 129 Parser.CurrentLines = &Parser.PreprocessorDirectives; 130 else if (!Parser.Line->Tokens.empty()) 131 Parser.CurrentLines = &Parser.Line->Tokens.back().Children; 132 PreBlockLine = std::move(Parser.Line); 133 Parser.Line = llvm::make_unique<UnwrappedLine>(); 134 Parser.Line->Level = PreBlockLine->Level; 135 Parser.Line->InPPDirective = PreBlockLine->InPPDirective; 136 } 137 138 ~ScopedLineState() { 139 if (!Parser.Line->Tokens.empty()) { 140 Parser.addUnwrappedLine(); 141 } 142 assert(Parser.Line->Tokens.empty()); 143 Parser.Line = std::move(PreBlockLine); 144 if (Parser.CurrentLines == &Parser.PreprocessorDirectives) 145 Parser.MustBreakBeforeNextToken = true; 146 Parser.CurrentLines = OriginalLines; 147 } 148 149 private: 150 UnwrappedLineParser &Parser; 151 152 std::unique_ptr<UnwrappedLine> PreBlockLine; 153 SmallVectorImpl<UnwrappedLine> *OriginalLines; 154 }; 155 156 class CompoundStatementIndenter { 157 public: 158 CompoundStatementIndenter(UnwrappedLineParser *Parser, 159 const FormatStyle &Style, unsigned &LineLevel) 160 : LineLevel(LineLevel), OldLineLevel(LineLevel) { 161 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman) { 162 Parser->addUnwrappedLine(); 163 } else if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 164 Parser->addUnwrappedLine(); 165 ++LineLevel; 166 } 167 } 168 ~CompoundStatementIndenter() { LineLevel = OldLineLevel; } 169 170 private: 171 unsigned &LineLevel; 172 unsigned OldLineLevel; 173 }; 174 175 namespace { 176 177 class IndexedTokenSource : public FormatTokenSource { 178 public: 179 IndexedTokenSource(ArrayRef<FormatToken *> Tokens) 180 : Tokens(Tokens), Position(-1) {} 181 182 FormatToken *getNextToken() override { 183 ++Position; 184 return Tokens[Position]; 185 } 186 187 unsigned getPosition() override { 188 assert(Position >= 0); 189 return Position; 190 } 191 192 FormatToken *setPosition(unsigned P) override { 193 Position = P; 194 return Tokens[Position]; 195 } 196 197 void reset() { Position = -1; } 198 199 private: 200 ArrayRef<FormatToken *> Tokens; 201 int Position; 202 }; 203 204 } // end anonymous namespace 205 206 UnwrappedLineParser::UnwrappedLineParser(const FormatStyle &Style, 207 const AdditionalKeywords &Keywords, 208 ArrayRef<FormatToken *> Tokens, 209 UnwrappedLineConsumer &Callback) 210 : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), 211 CurrentLines(&Lines), StructuralError(false), Style(Style), 212 Keywords(Keywords), Tokens(nullptr), Callback(Callback), 213 AllTokens(Tokens), PPBranchLevel(-1) {} 214 215 void UnwrappedLineParser::reset() { 216 PPBranchLevel = -1; 217 Line.reset(new UnwrappedLine); 218 CommentsBeforeNextToken.clear(); 219 FormatTok = nullptr; 220 MustBreakBeforeNextToken = false; 221 PreprocessorDirectives.clear(); 222 CurrentLines = &Lines; 223 DeclarationScopeStack.clear(); 224 StructuralError = false; 225 PPStack.clear(); 226 } 227 228 bool UnwrappedLineParser::parse() { 229 IndexedTokenSource TokenSource(AllTokens); 230 do { 231 DEBUG(llvm::dbgs() << "----\n"); 232 reset(); 233 Tokens = &TokenSource; 234 TokenSource.reset(); 235 236 readToken(); 237 parseFile(); 238 // Create line with eof token. 239 pushToken(FormatTok); 240 addUnwrappedLine(); 241 242 for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(), 243 E = Lines.end(); 244 I != E; ++I) { 245 Callback.consumeUnwrappedLine(*I); 246 } 247 Callback.finishRun(); 248 Lines.clear(); 249 while (!PPLevelBranchIndex.empty() && 250 PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) { 251 PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1); 252 PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1); 253 } 254 if (!PPLevelBranchIndex.empty()) { 255 ++PPLevelBranchIndex.back(); 256 assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size()); 257 assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back()); 258 } 259 } while (!PPLevelBranchIndex.empty()); 260 261 return StructuralError; 262 } 263 264 void UnwrappedLineParser::parseFile() { 265 ScopedDeclarationState DeclarationState( 266 *Line, DeclarationScopeStack, 267 /*MustBeDeclaration=*/!Line->InPPDirective); 268 parseLevel(/*HasOpeningBrace=*/false); 269 // Make sure to format the remaining tokens. 270 flushComments(true); 271 addUnwrappedLine(); 272 } 273 274 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) { 275 bool SwitchLabelEncountered = false; 276 do { 277 switch (FormatTok->Tok.getKind()) { 278 case tok::comment: 279 nextToken(); 280 addUnwrappedLine(); 281 break; 282 case tok::l_brace: 283 // FIXME: Add parameter whether this can happen - if this happens, we must 284 // be in a non-declaration context. 285 parseBlock(/*MustBeDeclaration=*/false); 286 addUnwrappedLine(); 287 break; 288 case tok::r_brace: 289 if (HasOpeningBrace) 290 return; 291 StructuralError = true; 292 nextToken(); 293 addUnwrappedLine(); 294 break; 295 case tok::kw_default: 296 case tok::kw_case: 297 if (!SwitchLabelEncountered && 298 (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1))) 299 ++Line->Level; 300 SwitchLabelEncountered = true; 301 parseStructuralElement(); 302 break; 303 default: 304 parseStructuralElement(); 305 break; 306 } 307 } while (!eof()); 308 } 309 310 void UnwrappedLineParser::calculateBraceTypes() { 311 // We'll parse forward through the tokens until we hit 312 // a closing brace or eof - note that getNextToken() will 313 // parse macros, so this will magically work inside macro 314 // definitions, too. 315 unsigned StoredPosition = Tokens->getPosition(); 316 FormatToken *Tok = FormatTok; 317 // Keep a stack of positions of lbrace tokens. We will 318 // update information about whether an lbrace starts a 319 // braced init list or a different block during the loop. 320 SmallVector<FormatToken *, 8> LBraceStack; 321 assert(Tok->Tok.is(tok::l_brace)); 322 do { 323 // Get next none-comment token. 324 FormatToken *NextTok; 325 unsigned ReadTokens = 0; 326 do { 327 NextTok = Tokens->getNextToken(); 328 ++ReadTokens; 329 } while (NextTok->is(tok::comment)); 330 331 switch (Tok->Tok.getKind()) { 332 case tok::l_brace: 333 LBraceStack.push_back(Tok); 334 break; 335 case tok::r_brace: 336 if (!LBraceStack.empty()) { 337 if (LBraceStack.back()->BlockKind == BK_Unknown) { 338 bool ProbablyBracedList = false; 339 if (Style.Language == FormatStyle::LK_Proto) { 340 ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); 341 } else { 342 // Using OriginalColumn to distinguish between ObjC methods and 343 // binary operators is a bit hacky. 344 bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) && 345 NextTok->OriginalColumn == 0; 346 347 // If there is a comma, semicolon or right paren after the closing 348 // brace, we assume this is a braced initializer list. Note that 349 // regardless how we mark inner braces here, we will overwrite the 350 // BlockKind later if we parse a braced list (where all blocks 351 // inside are by default braced lists), or when we explicitly detect 352 // blocks (for example while parsing lambdas). 353 // 354 // We exclude + and - as they can be ObjC visibility modifiers. 355 ProbablyBracedList = 356 NextTok->isOneOf(tok::comma, tok::semi, tok::period, tok::colon, 357 tok::r_paren, tok::r_square, tok::l_brace, 358 tok::l_paren, tok::ellipsis) || 359 (NextTok->isBinaryOperator() && !NextIsObjCMethod); 360 } 361 if (ProbablyBracedList) { 362 Tok->BlockKind = BK_BracedInit; 363 LBraceStack.back()->BlockKind = BK_BracedInit; 364 } else { 365 Tok->BlockKind = BK_Block; 366 LBraceStack.back()->BlockKind = BK_Block; 367 } 368 } 369 LBraceStack.pop_back(); 370 } 371 break; 372 case tok::at: 373 case tok::semi: 374 case tok::kw_if: 375 case tok::kw_while: 376 case tok::kw_for: 377 case tok::kw_switch: 378 case tok::kw_try: 379 case tok::kw___try: 380 if (!LBraceStack.empty()) 381 LBraceStack.back()->BlockKind = BK_Block; 382 break; 383 default: 384 break; 385 } 386 Tok = NextTok; 387 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 388 // Assume other blocks for all unclosed opening braces. 389 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 390 if (LBraceStack[i]->BlockKind == BK_Unknown) 391 LBraceStack[i]->BlockKind = BK_Block; 392 } 393 394 FormatTok = Tokens->setPosition(StoredPosition); 395 } 396 397 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 398 bool MunchSemi) { 399 assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected"); 400 unsigned InitialLevel = Line->Level; 401 nextToken(); 402 403 addUnwrappedLine(); 404 405 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 406 MustBeDeclaration); 407 if (AddLevel) 408 ++Line->Level; 409 parseLevel(/*HasOpeningBrace=*/true); 410 411 if (!FormatTok->Tok.is(tok::r_brace)) { 412 Line->Level = InitialLevel; 413 StructuralError = true; 414 return; 415 } 416 417 nextToken(); // Munch the closing brace. 418 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 419 nextToken(); 420 Line->Level = InitialLevel; 421 } 422 423 static bool isGoogScope(const UnwrappedLine &Line) { 424 // FIXME: Closure-library specific stuff should not be hard-coded but be 425 // configurable. 426 if (Line.Tokens.size() < 4) 427 return false; 428 auto I = Line.Tokens.begin(); 429 if (I->Tok->TokenText != "goog") 430 return false; 431 ++I; 432 if (I->Tok->isNot(tok::period)) 433 return false; 434 ++I; 435 if (I->Tok->TokenText != "scope") 436 return false; 437 ++I; 438 return I->Tok->is(tok::l_paren); 439 } 440 441 static bool ShouldBreakBeforeBrace(const FormatStyle &Style, 442 const FormatToken &InitialToken) { 443 switch (Style.BreakBeforeBraces) { 444 case FormatStyle::BS_Linux: 445 return InitialToken.isOneOf(tok::kw_namespace, tok::kw_class); 446 case FormatStyle::BS_Allman: 447 case FormatStyle::BS_GNU: 448 return true; 449 default: 450 return false; 451 } 452 } 453 454 void UnwrappedLineParser::parseChildBlock() { 455 FormatTok->BlockKind = BK_Block; 456 nextToken(); 457 { 458 bool GoogScope = 459 Style.Language == FormatStyle::LK_JavaScript && isGoogScope(*Line); 460 ScopedLineState LineState(*this); 461 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 462 /*MustBeDeclaration=*/false); 463 Line->Level += GoogScope ? 0 : 1; 464 parseLevel(/*HasOpeningBrace=*/true); 465 flushComments(isOnNewLine(*FormatTok)); 466 Line->Level -= GoogScope ? 0 : 1; 467 } 468 nextToken(); 469 } 470 471 void UnwrappedLineParser::parsePPDirective() { 472 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 473 ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError); 474 nextToken(); 475 476 if (!FormatTok->Tok.getIdentifierInfo()) { 477 parsePPUnknown(); 478 return; 479 } 480 481 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 482 case tok::pp_define: 483 parsePPDefine(); 484 return; 485 case tok::pp_if: 486 parsePPIf(/*IfDef=*/false); 487 break; 488 case tok::pp_ifdef: 489 case tok::pp_ifndef: 490 parsePPIf(/*IfDef=*/true); 491 break; 492 case tok::pp_else: 493 parsePPElse(); 494 break; 495 case tok::pp_elif: 496 parsePPElIf(); 497 break; 498 case tok::pp_endif: 499 parsePPEndIf(); 500 break; 501 default: 502 parsePPUnknown(); 503 break; 504 } 505 } 506 507 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 508 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 509 PPStack.push_back(PP_Unreachable); 510 else 511 PPStack.push_back(PP_Conditional); 512 } 513 514 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 515 ++PPBranchLevel; 516 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 517 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 518 PPLevelBranchIndex.push_back(0); 519 PPLevelBranchCount.push_back(0); 520 } 521 PPChainBranchIndex.push(0); 522 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 523 conditionalCompilationCondition(Unreachable || Skip); 524 } 525 526 void UnwrappedLineParser::conditionalCompilationAlternative() { 527 if (!PPStack.empty()) 528 PPStack.pop_back(); 529 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 530 if (!PPChainBranchIndex.empty()) 531 ++PPChainBranchIndex.top(); 532 conditionalCompilationCondition( 533 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 534 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 535 } 536 537 void UnwrappedLineParser::conditionalCompilationEnd() { 538 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 539 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 540 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 541 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 542 } 543 } 544 // Guard against #endif's without #if. 545 if (PPBranchLevel > 0) 546 --PPBranchLevel; 547 if (!PPChainBranchIndex.empty()) 548 PPChainBranchIndex.pop(); 549 if (!PPStack.empty()) 550 PPStack.pop_back(); 551 } 552 553 void UnwrappedLineParser::parsePPIf(bool IfDef) { 554 nextToken(); 555 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 556 FormatTok->Tok.getLiteralData() != nullptr && 557 StringRef(FormatTok->Tok.getLiteralData(), 558 FormatTok->Tok.getLength()) == "0") || 559 FormatTok->Tok.is(tok::kw_false); 560 conditionalCompilationStart(!IfDef && IsLiteralFalse); 561 parsePPUnknown(); 562 } 563 564 void UnwrappedLineParser::parsePPElse() { 565 conditionalCompilationAlternative(); 566 parsePPUnknown(); 567 } 568 569 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 570 571 void UnwrappedLineParser::parsePPEndIf() { 572 conditionalCompilationEnd(); 573 parsePPUnknown(); 574 } 575 576 void UnwrappedLineParser::parsePPDefine() { 577 nextToken(); 578 579 if (FormatTok->Tok.getKind() != tok::identifier) { 580 parsePPUnknown(); 581 return; 582 } 583 nextToken(); 584 if (FormatTok->Tok.getKind() == tok::l_paren && 585 FormatTok->WhitespaceRange.getBegin() == 586 FormatTok->WhitespaceRange.getEnd()) { 587 parseParens(); 588 } 589 addUnwrappedLine(); 590 Line->Level = 1; 591 592 // Errors during a preprocessor directive can only affect the layout of the 593 // preprocessor directive, and thus we ignore them. An alternative approach 594 // would be to use the same approach we use on the file level (no 595 // re-indentation if there was a structural error) within the macro 596 // definition. 597 parseFile(); 598 } 599 600 void UnwrappedLineParser::parsePPUnknown() { 601 do { 602 nextToken(); 603 } while (!eof()); 604 addUnwrappedLine(); 605 } 606 607 // Here we blacklist certain tokens that are not usually the first token in an 608 // unwrapped line. This is used in attempt to distinguish macro calls without 609 // trailing semicolons from other constructs split to several lines. 610 static bool tokenCanStartNewLine(const clang::Token &Tok) { 611 // Semicolon can be a null-statement, l_square can be a start of a macro or 612 // a C++11 attribute, but this doesn't seem to be common. 613 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 614 Tok.isNot(tok::l_square) && 615 // Tokens that can only be used as binary operators and a part of 616 // overloaded operator names. 617 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 618 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 619 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 620 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 621 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 622 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 623 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 624 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 625 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 626 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 627 Tok.isNot(tok::lesslessequal) && 628 // Colon is used in labels, base class lists, initializer lists, 629 // range-based for loops, ternary operator, but should never be the 630 // first token in an unwrapped line. 631 Tok.isNot(tok::colon) && 632 // 'noexcept' is a trailing annotation. 633 Tok.isNot(tok::kw_noexcept); 634 } 635 636 void UnwrappedLineParser::parseStructuralElement() { 637 assert(!FormatTok->Tok.is(tok::l_brace)); 638 switch (FormatTok->Tok.getKind()) { 639 case tok::at: 640 nextToken(); 641 if (FormatTok->Tok.is(tok::l_brace)) { 642 parseBracedList(); 643 break; 644 } 645 switch (FormatTok->Tok.getObjCKeywordID()) { 646 case tok::objc_public: 647 case tok::objc_protected: 648 case tok::objc_package: 649 case tok::objc_private: 650 return parseAccessSpecifier(); 651 case tok::objc_interface: 652 case tok::objc_implementation: 653 return parseObjCInterfaceOrImplementation(); 654 case tok::objc_protocol: 655 return parseObjCProtocol(); 656 case tok::objc_end: 657 return; // Handled by the caller. 658 case tok::objc_optional: 659 case tok::objc_required: 660 nextToken(); 661 addUnwrappedLine(); 662 return; 663 case tok::objc_try: 664 // This branch isn't strictly necessary (the kw_try case below would 665 // do this too after the tok::at is parsed above). But be explicit. 666 parseTryCatch(); 667 return; 668 default: 669 break; 670 } 671 break; 672 case tok::kw_asm: 673 nextToken(); 674 if (FormatTok->is(tok::l_brace)) { 675 nextToken(); 676 while (FormatTok && FormatTok->isNot(tok::eof)) { 677 if (FormatTok->is(tok::r_brace)) { 678 nextToken(); 679 break; 680 } 681 FormatTok->Finalized = true; 682 nextToken(); 683 } 684 } 685 break; 686 case tok::kw_namespace: 687 parseNamespace(); 688 return; 689 case tok::kw_inline: 690 nextToken(); 691 if (FormatTok->Tok.is(tok::kw_namespace)) { 692 parseNamespace(); 693 return; 694 } 695 break; 696 case tok::kw_public: 697 case tok::kw_protected: 698 case tok::kw_private: 699 if (Style.Language == FormatStyle::LK_Java || 700 Style.Language == FormatStyle::LK_JavaScript) 701 nextToken(); 702 else 703 parseAccessSpecifier(); 704 return; 705 case tok::kw_if: 706 parseIfThenElse(); 707 return; 708 case tok::kw_for: 709 case tok::kw_while: 710 parseForOrWhileLoop(); 711 return; 712 case tok::kw_do: 713 parseDoWhile(); 714 return; 715 case tok::kw_switch: 716 parseSwitch(); 717 return; 718 case tok::kw_default: 719 nextToken(); 720 parseLabel(); 721 return; 722 case tok::kw_case: 723 parseCaseLabel(); 724 return; 725 case tok::kw_try: 726 case tok::kw___try: 727 parseTryCatch(); 728 return; 729 case tok::kw_extern: 730 nextToken(); 731 if (FormatTok->Tok.is(tok::string_literal)) { 732 nextToken(); 733 if (FormatTok->Tok.is(tok::l_brace)) { 734 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 735 addUnwrappedLine(); 736 return; 737 } 738 } 739 break; 740 case tok::kw_export: 741 if (Style.Language == FormatStyle::LK_JavaScript) { 742 parseJavaScriptEs6ImportExport(); 743 return; 744 } 745 break; 746 case tok::identifier: 747 if (FormatTok->is(TT_ForEachMacro)) { 748 parseForOrWhileLoop(); 749 return; 750 } 751 if (Style.Language == FormatStyle::LK_JavaScript && 752 FormatTok->is(Keywords.kw_import)) { 753 parseJavaScriptEs6ImportExport(); 754 return; 755 } 756 if (FormatTok->is(Keywords.kw_signals)) { 757 nextToken(); 758 if (FormatTok->is(tok::colon)) { 759 nextToken(); 760 addUnwrappedLine(); 761 } 762 return; 763 } 764 // In all other cases, parse the declaration. 765 break; 766 default: 767 break; 768 } 769 do { 770 switch (FormatTok->Tok.getKind()) { 771 case tok::at: 772 nextToken(); 773 if (FormatTok->Tok.is(tok::l_brace)) 774 parseBracedList(); 775 break; 776 case tok::kw_enum: 777 parseEnum(); 778 break; 779 case tok::kw_typedef: 780 nextToken(); 781 if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS, 782 Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS)) 783 parseEnum(); 784 break; 785 case tok::kw_struct: 786 case tok::kw_union: 787 case tok::kw_class: 788 parseRecord(); 789 // A record declaration or definition is always the start of a structural 790 // element. 791 break; 792 case tok::period: 793 nextToken(); 794 // In Java, classes have an implicit static member "class". 795 if (Style.Language == FormatStyle::LK_Java && FormatTok && 796 FormatTok->is(tok::kw_class)) 797 nextToken(); 798 break; 799 case tok::semi: 800 nextToken(); 801 addUnwrappedLine(); 802 return; 803 case tok::r_brace: 804 addUnwrappedLine(); 805 return; 806 case tok::l_paren: 807 parseParens(); 808 break; 809 case tok::caret: 810 nextToken(); 811 if (FormatTok->Tok.isAnyIdentifier() || 812 FormatTok->isSimpleTypeSpecifier()) 813 nextToken(); 814 if (FormatTok->is(tok::l_paren)) 815 parseParens(); 816 if (FormatTok->is(tok::l_brace)) 817 parseChildBlock(); 818 break; 819 case tok::l_brace: 820 if (!tryToParseBracedList()) { 821 // A block outside of parentheses must be the last part of a 822 // structural element. 823 // FIXME: Figure out cases where this is not true, and add projections 824 // for them (the one we know is missing are lambdas). 825 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 826 addUnwrappedLine(); 827 FormatTok->Type = TT_FunctionLBrace; 828 parseBlock(/*MustBeDeclaration=*/false); 829 addUnwrappedLine(); 830 return; 831 } 832 // Otherwise this was a braced init list, and the structural 833 // element continues. 834 break; 835 case tok::kw_try: 836 // We arrive here when parsing function-try blocks. 837 parseTryCatch(); 838 return; 839 case tok::identifier: { 840 StringRef Text = FormatTok->TokenText; 841 // Parse function literal unless 'function' is the first token in a line 842 // in which case this should be treated as a free-standing function. 843 if (Style.Language == FormatStyle::LK_JavaScript && Text == "function" && 844 Line->Tokens.size() > 0) { 845 tryToParseJSFunction(); 846 break; 847 } 848 nextToken(); 849 if (Line->Tokens.size() == 1 && 850 // JS doesn't have macros, and within classes colons indicate fields, 851 // not labels. 852 Style.Language != FormatStyle::LK_JavaScript) { 853 if (FormatTok->Tok.is(tok::colon) && !Line->MustBeDeclaration) { 854 parseLabel(); 855 return; 856 } 857 // Recognize function-like macro usages without trailing semicolon as 858 // well as free-standing macros like Q_OBJECT. 859 bool FunctionLike = FormatTok->is(tok::l_paren); 860 if (FunctionLike) 861 parseParens(); 862 if (FormatTok->NewlinesBefore > 0 && 863 (Text.size() >= 5 || FunctionLike) && 864 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 865 addUnwrappedLine(); 866 return; 867 } 868 } 869 break; 870 } 871 case tok::equal: 872 nextToken(); 873 if (FormatTok->Tok.is(tok::l_brace)) { 874 parseBracedList(); 875 } 876 break; 877 case tok::l_square: 878 parseSquare(); 879 break; 880 case tok::kw_new: 881 parseNew(); 882 break; 883 default: 884 nextToken(); 885 break; 886 } 887 } while (!eof()); 888 } 889 890 bool UnwrappedLineParser::tryToParseLambda() { 891 // FIXME: This is a dirty way to access the previous token. Find a better 892 // solution. 893 if (!Line->Tokens.empty() && 894 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator, 895 tok::kw_new, tok::kw_delete) || 896 Line->Tokens.back().Tok->closesScope() || 897 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 898 nextToken(); 899 return false; 900 } 901 assert(FormatTok->is(tok::l_square)); 902 FormatToken &LSquare = *FormatTok; 903 if (!tryToParseLambdaIntroducer()) 904 return false; 905 906 while (FormatTok->isNot(tok::l_brace)) { 907 if (FormatTok->isSimpleTypeSpecifier()) { 908 nextToken(); 909 continue; 910 } 911 switch (FormatTok->Tok.getKind()) { 912 case tok::l_brace: 913 break; 914 case tok::l_paren: 915 parseParens(); 916 break; 917 case tok::amp: 918 case tok::star: 919 case tok::kw_const: 920 case tok::comma: 921 case tok::less: 922 case tok::greater: 923 case tok::identifier: 924 case tok::coloncolon: 925 case tok::kw_mutable: 926 nextToken(); 927 break; 928 case tok::arrow: 929 FormatTok->Type = TT_TrailingReturnArrow; 930 nextToken(); 931 break; 932 default: 933 return true; 934 } 935 } 936 LSquare.Type = TT_LambdaLSquare; 937 parseChildBlock(); 938 return true; 939 } 940 941 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 942 nextToken(); 943 if (FormatTok->is(tok::equal)) { 944 nextToken(); 945 if (FormatTok->is(tok::r_square)) { 946 nextToken(); 947 return true; 948 } 949 if (FormatTok->isNot(tok::comma)) 950 return false; 951 nextToken(); 952 } else if (FormatTok->is(tok::amp)) { 953 nextToken(); 954 if (FormatTok->is(tok::r_square)) { 955 nextToken(); 956 return true; 957 } 958 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 959 return false; 960 } 961 if (FormatTok->is(tok::comma)) 962 nextToken(); 963 } else if (FormatTok->is(tok::r_square)) { 964 nextToken(); 965 return true; 966 } 967 do { 968 if (FormatTok->is(tok::amp)) 969 nextToken(); 970 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 971 return false; 972 nextToken(); 973 if (FormatTok->is(tok::ellipsis)) 974 nextToken(); 975 if (FormatTok->is(tok::comma)) { 976 nextToken(); 977 } else if (FormatTok->is(tok::r_square)) { 978 nextToken(); 979 return true; 980 } else { 981 return false; 982 } 983 } while (!eof()); 984 return false; 985 } 986 987 void UnwrappedLineParser::tryToParseJSFunction() { 988 nextToken(); 989 990 // Consume function name. 991 if (FormatTok->is(tok::identifier)) 992 nextToken(); 993 994 if (FormatTok->isNot(tok::l_paren)) 995 return; 996 nextToken(); 997 while (FormatTok->isNot(tok::l_brace)) { 998 // Err on the side of caution in order to avoid consuming the full file in 999 // case of incomplete code. 1000 if (!FormatTok->isOneOf(tok::identifier, tok::comma, tok::r_paren, 1001 tok::comment)) 1002 return; 1003 nextToken(); 1004 } 1005 parseChildBlock(); 1006 } 1007 1008 bool UnwrappedLineParser::tryToParseBracedList() { 1009 if (FormatTok->BlockKind == BK_Unknown) 1010 calculateBraceTypes(); 1011 assert(FormatTok->BlockKind != BK_Unknown); 1012 if (FormatTok->BlockKind == BK_Block) 1013 return false; 1014 parseBracedList(); 1015 return true; 1016 } 1017 1018 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 1019 bool HasError = false; 1020 nextToken(); 1021 1022 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 1023 // replace this by using parseAssigmentExpression() inside. 1024 do { 1025 if (Style.Language == FormatStyle::LK_JavaScript && 1026 FormatTok->is(Keywords.kw_function)) { 1027 tryToParseJSFunction(); 1028 continue; 1029 } 1030 switch (FormatTok->Tok.getKind()) { 1031 case tok::caret: 1032 nextToken(); 1033 if (FormatTok->is(tok::l_brace)) { 1034 parseChildBlock(); 1035 } 1036 break; 1037 case tok::l_square: 1038 tryToParseLambda(); 1039 break; 1040 case tok::l_brace: 1041 // Assume there are no blocks inside a braced init list apart 1042 // from the ones we explicitly parse out (like lambdas). 1043 FormatTok->BlockKind = BK_BracedInit; 1044 parseBracedList(); 1045 break; 1046 case tok::r_paren: 1047 // JavaScript can just have free standing methods and getters/setters in 1048 // object literals. Detect them by a "{" following ")". 1049 if (Style.Language == FormatStyle::LK_JavaScript) { 1050 nextToken(); 1051 if (FormatTok->is(tok::l_brace)) 1052 parseChildBlock(); 1053 break; 1054 } 1055 nextToken(); 1056 break; 1057 case tok::r_brace: 1058 nextToken(); 1059 return !HasError; 1060 case tok::semi: 1061 HasError = true; 1062 if (!ContinueOnSemicolons) 1063 return !HasError; 1064 nextToken(); 1065 break; 1066 case tok::comma: 1067 nextToken(); 1068 break; 1069 default: 1070 nextToken(); 1071 break; 1072 } 1073 } while (!eof()); 1074 return false; 1075 } 1076 1077 void UnwrappedLineParser::parseParens() { 1078 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1079 nextToken(); 1080 do { 1081 switch (FormatTok->Tok.getKind()) { 1082 case tok::l_paren: 1083 parseParens(); 1084 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) 1085 parseChildBlock(); 1086 break; 1087 case tok::r_paren: 1088 nextToken(); 1089 return; 1090 case tok::r_brace: 1091 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1092 return; 1093 case tok::l_square: 1094 tryToParseLambda(); 1095 break; 1096 case tok::l_brace: 1097 if (!tryToParseBracedList()) { 1098 parseChildBlock(); 1099 } 1100 break; 1101 case tok::at: 1102 nextToken(); 1103 if (FormatTok->Tok.is(tok::l_brace)) 1104 parseBracedList(); 1105 break; 1106 case tok::identifier: 1107 if (Style.Language == FormatStyle::LK_JavaScript && 1108 FormatTok->is(Keywords.kw_function)) 1109 tryToParseJSFunction(); 1110 else 1111 nextToken(); 1112 break; 1113 default: 1114 nextToken(); 1115 break; 1116 } 1117 } while (!eof()); 1118 } 1119 1120 void UnwrappedLineParser::parseSquare() { 1121 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1122 if (tryToParseLambda()) 1123 return; 1124 do { 1125 switch (FormatTok->Tok.getKind()) { 1126 case tok::l_paren: 1127 parseParens(); 1128 break; 1129 case tok::r_square: 1130 nextToken(); 1131 return; 1132 case tok::r_brace: 1133 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1134 return; 1135 case tok::l_square: 1136 parseSquare(); 1137 break; 1138 case tok::l_brace: { 1139 if (!tryToParseBracedList()) { 1140 parseChildBlock(); 1141 } 1142 break; 1143 } 1144 case tok::at: 1145 nextToken(); 1146 if (FormatTok->Tok.is(tok::l_brace)) 1147 parseBracedList(); 1148 break; 1149 default: 1150 nextToken(); 1151 break; 1152 } 1153 } while (!eof()); 1154 } 1155 1156 void UnwrappedLineParser::parseIfThenElse() { 1157 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1158 nextToken(); 1159 if (FormatTok->Tok.is(tok::l_paren)) 1160 parseParens(); 1161 bool NeedsUnwrappedLine = false; 1162 if (FormatTok->Tok.is(tok::l_brace)) { 1163 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1164 parseBlock(/*MustBeDeclaration=*/false); 1165 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1166 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1167 addUnwrappedLine(); 1168 } else { 1169 NeedsUnwrappedLine = true; 1170 } 1171 } else { 1172 addUnwrappedLine(); 1173 ++Line->Level; 1174 parseStructuralElement(); 1175 --Line->Level; 1176 } 1177 if (FormatTok->Tok.is(tok::kw_else)) { 1178 if (Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) 1179 addUnwrappedLine(); 1180 nextToken(); 1181 if (FormatTok->Tok.is(tok::l_brace)) { 1182 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1183 parseBlock(/*MustBeDeclaration=*/false); 1184 addUnwrappedLine(); 1185 } else if (FormatTok->Tok.is(tok::kw_if)) { 1186 parseIfThenElse(); 1187 } else { 1188 addUnwrappedLine(); 1189 ++Line->Level; 1190 parseStructuralElement(); 1191 --Line->Level; 1192 } 1193 } else if (NeedsUnwrappedLine) { 1194 addUnwrappedLine(); 1195 } 1196 } 1197 1198 void UnwrappedLineParser::parseTryCatch() { 1199 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected"); 1200 nextToken(); 1201 bool NeedsUnwrappedLine = false; 1202 if (FormatTok->is(tok::colon)) { 1203 // We are in a function try block, what comes is an initializer list. 1204 nextToken(); 1205 while (FormatTok->is(tok::identifier)) { 1206 nextToken(); 1207 if (FormatTok->is(tok::l_paren)) 1208 parseParens(); 1209 else 1210 StructuralError = true; 1211 if (FormatTok->is(tok::comma)) 1212 nextToken(); 1213 } 1214 } 1215 // Parse try with resource. 1216 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) { 1217 parseParens(); 1218 } 1219 if (FormatTok->is(tok::l_brace)) { 1220 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1221 parseBlock(/*MustBeDeclaration=*/false); 1222 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1223 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1224 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1225 addUnwrappedLine(); 1226 } else { 1227 NeedsUnwrappedLine = true; 1228 } 1229 } else if (!FormatTok->is(tok::kw_catch)) { 1230 // The C++ standard requires a compound-statement after a try. 1231 // If there's none, we try to assume there's a structuralElement 1232 // and try to continue. 1233 StructuralError = true; 1234 addUnwrappedLine(); 1235 ++Line->Level; 1236 parseStructuralElement(); 1237 --Line->Level; 1238 } 1239 while (1) { 1240 if (FormatTok->is(tok::at)) 1241 nextToken(); 1242 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, 1243 tok::kw___finally) || 1244 ((Style.Language == FormatStyle::LK_Java || 1245 Style.Language == FormatStyle::LK_JavaScript) && 1246 FormatTok->is(Keywords.kw_finally)) || 1247 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) || 1248 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally)))) 1249 break; 1250 nextToken(); 1251 while (FormatTok->isNot(tok::l_brace)) { 1252 if (FormatTok->is(tok::l_paren)) { 1253 parseParens(); 1254 continue; 1255 } 1256 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) 1257 return; 1258 nextToken(); 1259 } 1260 NeedsUnwrappedLine = false; 1261 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1262 parseBlock(/*MustBeDeclaration=*/false); 1263 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1264 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1265 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1266 addUnwrappedLine(); 1267 } else { 1268 NeedsUnwrappedLine = true; 1269 } 1270 } 1271 if (NeedsUnwrappedLine) { 1272 addUnwrappedLine(); 1273 } 1274 } 1275 1276 void UnwrappedLineParser::parseNamespace() { 1277 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1278 1279 const FormatToken &InitialToken = *FormatTok; 1280 nextToken(); 1281 if (FormatTok->Tok.is(tok::identifier)) 1282 nextToken(); 1283 if (FormatTok->Tok.is(tok::l_brace)) { 1284 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1285 addUnwrappedLine(); 1286 1287 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1288 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1289 DeclarationScopeStack.size() > 1); 1290 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1291 // Munch the semicolon after a namespace. This is more common than one would 1292 // think. Puttin the semicolon into its own line is very ugly. 1293 if (FormatTok->Tok.is(tok::semi)) 1294 nextToken(); 1295 addUnwrappedLine(); 1296 } 1297 // FIXME: Add error handling. 1298 } 1299 1300 void UnwrappedLineParser::parseNew() { 1301 assert(FormatTok->is(tok::kw_new) && "'new' expected"); 1302 nextToken(); 1303 if (Style.Language != FormatStyle::LK_Java) 1304 return; 1305 1306 // In Java, we can parse everything up to the parens, which aren't optional. 1307 do { 1308 // There should not be a ;, { or } before the new's open paren. 1309 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) 1310 return; 1311 1312 // Consume the parens. 1313 if (FormatTok->is(tok::l_paren)) { 1314 parseParens(); 1315 1316 // If there is a class body of an anonymous class, consume that as child. 1317 if (FormatTok->is(tok::l_brace)) 1318 parseChildBlock(); 1319 return; 1320 } 1321 nextToken(); 1322 } while (!eof()); 1323 } 1324 1325 void UnwrappedLineParser::parseForOrWhileLoop() { 1326 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) && 1327 "'for', 'while' or foreach macro expected"); 1328 nextToken(); 1329 if (FormatTok->Tok.is(tok::l_paren)) 1330 parseParens(); 1331 if (FormatTok->Tok.is(tok::l_brace)) { 1332 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1333 parseBlock(/*MustBeDeclaration=*/false); 1334 addUnwrappedLine(); 1335 } else { 1336 addUnwrappedLine(); 1337 ++Line->Level; 1338 parseStructuralElement(); 1339 --Line->Level; 1340 } 1341 } 1342 1343 void UnwrappedLineParser::parseDoWhile() { 1344 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1345 nextToken(); 1346 if (FormatTok->Tok.is(tok::l_brace)) { 1347 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1348 parseBlock(/*MustBeDeclaration=*/false); 1349 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1350 addUnwrappedLine(); 1351 } else { 1352 addUnwrappedLine(); 1353 ++Line->Level; 1354 parseStructuralElement(); 1355 --Line->Level; 1356 } 1357 1358 // FIXME: Add error handling. 1359 if (!FormatTok->Tok.is(tok::kw_while)) { 1360 addUnwrappedLine(); 1361 return; 1362 } 1363 1364 nextToken(); 1365 parseStructuralElement(); 1366 } 1367 1368 void UnwrappedLineParser::parseLabel() { 1369 nextToken(); 1370 unsigned OldLineLevel = Line->Level; 1371 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1372 --Line->Level; 1373 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1374 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1375 parseBlock(/*MustBeDeclaration=*/false); 1376 if (FormatTok->Tok.is(tok::kw_break)) { 1377 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1378 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1379 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1380 addUnwrappedLine(); 1381 } 1382 parseStructuralElement(); 1383 } 1384 addUnwrappedLine(); 1385 } else { 1386 addUnwrappedLine(); 1387 } 1388 Line->Level = OldLineLevel; 1389 } 1390 1391 void UnwrappedLineParser::parseCaseLabel() { 1392 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1393 // FIXME: fix handling of complex expressions here. 1394 do { 1395 nextToken(); 1396 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1397 parseLabel(); 1398 } 1399 1400 void UnwrappedLineParser::parseSwitch() { 1401 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1402 nextToken(); 1403 if (FormatTok->Tok.is(tok::l_paren)) 1404 parseParens(); 1405 if (FormatTok->Tok.is(tok::l_brace)) { 1406 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1407 parseBlock(/*MustBeDeclaration=*/false); 1408 addUnwrappedLine(); 1409 } else { 1410 addUnwrappedLine(); 1411 ++Line->Level; 1412 parseStructuralElement(); 1413 --Line->Level; 1414 } 1415 } 1416 1417 void UnwrappedLineParser::parseAccessSpecifier() { 1418 nextToken(); 1419 // Understand Qt's slots. 1420 if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots)) 1421 nextToken(); 1422 // Otherwise, we don't know what it is, and we'd better keep the next token. 1423 if (FormatTok->Tok.is(tok::colon)) 1424 nextToken(); 1425 addUnwrappedLine(); 1426 } 1427 1428 void UnwrappedLineParser::parseEnum() { 1429 // Won't be 'enum' for NS_ENUMs. 1430 if (FormatTok->Tok.is(tok::kw_enum)) 1431 nextToken(); 1432 1433 // Eat up enum class ... 1434 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1435 nextToken(); 1436 while (FormatTok->Tok.getIdentifierInfo() || 1437 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less, 1438 tok::greater, tok::comma, tok::question)) { 1439 nextToken(); 1440 // We can have macros or attributes in between 'enum' and the enum name. 1441 if (FormatTok->is(tok::l_paren)) 1442 parseParens(); 1443 if (FormatTok->is(tok::identifier)) 1444 nextToken(); 1445 } 1446 1447 // Just a declaration or something is wrong. 1448 if (FormatTok->isNot(tok::l_brace)) 1449 return; 1450 FormatTok->BlockKind = BK_Block; 1451 1452 if (Style.Language == FormatStyle::LK_Java) { 1453 // Java enums are different. 1454 parseJavaEnumBody(); 1455 return; 1456 } 1457 1458 // Parse enum body. 1459 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1460 if (HasError) { 1461 if (FormatTok->is(tok::semi)) 1462 nextToken(); 1463 addUnwrappedLine(); 1464 } 1465 1466 // We fall through to parsing a structural element afterwards, so that in 1467 // enum A {} n, m; 1468 // "} n, m;" will end up in one unwrapped line. 1469 } 1470 1471 void UnwrappedLineParser::parseJavaEnumBody() { 1472 // Determine whether the enum is simple, i.e. does not have a semicolon or 1473 // constants with class bodies. Simple enums can be formatted like braced 1474 // lists, contracted to a single line, etc. 1475 unsigned StoredPosition = Tokens->getPosition(); 1476 bool IsSimple = true; 1477 FormatToken *Tok = Tokens->getNextToken(); 1478 while (Tok) { 1479 if (Tok->is(tok::r_brace)) 1480 break; 1481 if (Tok->isOneOf(tok::l_brace, tok::semi)) { 1482 IsSimple = false; 1483 break; 1484 } 1485 // FIXME: This will also mark enums with braces in the arguments to enum 1486 // constants as "not simple". This is probably fine in practice, though. 1487 Tok = Tokens->getNextToken(); 1488 } 1489 FormatTok = Tokens->setPosition(StoredPosition); 1490 1491 if (IsSimple) { 1492 parseBracedList(); 1493 addUnwrappedLine(); 1494 return; 1495 } 1496 1497 // Parse the body of a more complex enum. 1498 // First add a line for everything up to the "{". 1499 nextToken(); 1500 addUnwrappedLine(); 1501 ++Line->Level; 1502 1503 // Parse the enum constants. 1504 while (FormatTok) { 1505 if (FormatTok->is(tok::l_brace)) { 1506 // Parse the constant's class body. 1507 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1508 /*MunchSemi=*/false); 1509 } else if (FormatTok->is(tok::l_paren)) { 1510 parseParens(); 1511 } else if (FormatTok->is(tok::comma)) { 1512 nextToken(); 1513 addUnwrappedLine(); 1514 } else if (FormatTok->is(tok::semi)) { 1515 nextToken(); 1516 addUnwrappedLine(); 1517 break; 1518 } else if (FormatTok->is(tok::r_brace)) { 1519 addUnwrappedLine(); 1520 break; 1521 } else { 1522 nextToken(); 1523 } 1524 } 1525 1526 // Parse the class body after the enum's ";" if any. 1527 parseLevel(/*HasOpeningBrace=*/true); 1528 nextToken(); 1529 --Line->Level; 1530 addUnwrappedLine(); 1531 } 1532 1533 void UnwrappedLineParser::parseRecord() { 1534 const FormatToken &InitialToken = *FormatTok; 1535 nextToken(); 1536 if (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::kw___attribute, 1537 tok::kw___declspec, tok::kw_alignas)) { 1538 nextToken(); 1539 // We can have macros or attributes in between 'class' and the class name. 1540 if (FormatTok->Tok.is(tok::l_paren)) { 1541 parseParens(); 1542 } 1543 // The actual identifier can be a nested name specifier, and in macros 1544 // it is often token-pasted. 1545 while (FormatTok->is(tok::identifier) || FormatTok->is(tok::coloncolon) || 1546 FormatTok->is(tok::hashhash) || 1547 ((Style.Language == FormatStyle::LK_Java || 1548 Style.Language == FormatStyle::LK_JavaScript) && 1549 FormatTok->isOneOf(tok::period, tok::comma))) 1550 nextToken(); 1551 1552 // Note that parsing away template declarations here leads to incorrectly 1553 // accepting function declarations as record declarations. 1554 // In general, we cannot solve this problem. Consider: 1555 // class A<int> B() {} 1556 // which can be a function definition or a class definition when B() is a 1557 // macro. If we find enough real-world cases where this is a problem, we 1558 // can parse for the 'template' keyword in the beginning of the statement, 1559 // and thus rule out the record production in case there is no template 1560 // (this would still leave us with an ambiguity between template function 1561 // and class declarations). 1562 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1563 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1564 if (FormatTok->Tok.is(tok::semi)) 1565 return; 1566 nextToken(); 1567 } 1568 } 1569 } 1570 if (FormatTok->Tok.is(tok::l_brace)) { 1571 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1572 addUnwrappedLine(); 1573 1574 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1575 /*MunchSemi=*/false); 1576 } 1577 // We fall through to parsing a structural element afterwards, so 1578 // class A {} n, m; 1579 // will end up in one unwrapped line. 1580 // This does not apply for Java. 1581 if (Style.Language == FormatStyle::LK_Java || 1582 Style.Language == FormatStyle::LK_JavaScript) 1583 addUnwrappedLine(); 1584 } 1585 1586 void UnwrappedLineParser::parseObjCProtocolList() { 1587 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1588 do 1589 nextToken(); 1590 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1591 nextToken(); // Skip '>'. 1592 } 1593 1594 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1595 do { 1596 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1597 nextToken(); 1598 addUnwrappedLine(); 1599 break; 1600 } 1601 if (FormatTok->is(tok::l_brace)) { 1602 parseBlock(/*MustBeDeclaration=*/false); 1603 // In ObjC interfaces, nothing should be following the "}". 1604 addUnwrappedLine(); 1605 } else if (FormatTok->is(tok::r_brace)) { 1606 // Ignore stray "}". parseStructuralElement doesn't consume them. 1607 nextToken(); 1608 addUnwrappedLine(); 1609 } else { 1610 parseStructuralElement(); 1611 } 1612 } while (!eof()); 1613 } 1614 1615 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1616 nextToken(); 1617 nextToken(); // interface name 1618 1619 // @interface can be followed by either a base class, or a category. 1620 if (FormatTok->Tok.is(tok::colon)) { 1621 nextToken(); 1622 nextToken(); // base class name 1623 } else if (FormatTok->Tok.is(tok::l_paren)) 1624 // Skip category, if present. 1625 parseParens(); 1626 1627 if (FormatTok->Tok.is(tok::less)) 1628 parseObjCProtocolList(); 1629 1630 if (FormatTok->Tok.is(tok::l_brace)) { 1631 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1632 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1633 addUnwrappedLine(); 1634 parseBlock(/*MustBeDeclaration=*/true); 1635 } 1636 1637 // With instance variables, this puts '}' on its own line. Without instance 1638 // variables, this ends the @interface line. 1639 addUnwrappedLine(); 1640 1641 parseObjCUntilAtEnd(); 1642 } 1643 1644 void UnwrappedLineParser::parseObjCProtocol() { 1645 nextToken(); 1646 nextToken(); // protocol name 1647 1648 if (FormatTok->Tok.is(tok::less)) 1649 parseObjCProtocolList(); 1650 1651 // Check for protocol declaration. 1652 if (FormatTok->Tok.is(tok::semi)) { 1653 nextToken(); 1654 return addUnwrappedLine(); 1655 } 1656 1657 addUnwrappedLine(); 1658 parseObjCUntilAtEnd(); 1659 } 1660 1661 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { 1662 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export)); 1663 nextToken(); 1664 1665 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, Keywords.kw_function, 1666 Keywords.kw_var)) 1667 return; // Fall through to parsing the corresponding structure. 1668 1669 if (FormatTok->is(tok::kw_default)) { 1670 nextToken(); // export default ..., fall through after eating 'default'. 1671 return; 1672 } 1673 1674 if (FormatTok->is(tok::l_brace)) { 1675 FormatTok->BlockKind = BK_Block; 1676 parseBracedList(); 1677 } 1678 1679 while (!eof() && FormatTok->isNot(tok::semi) && 1680 FormatTok->isNot(tok::l_brace)) { 1681 nextToken(); 1682 } 1683 } 1684 1685 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1686 StringRef Prefix = "") { 1687 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1688 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1689 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1690 E = Line.Tokens.end(); 1691 I != E; ++I) { 1692 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1693 } 1694 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1695 E = Line.Tokens.end(); 1696 I != E; ++I) { 1697 const UnwrappedLineNode &Node = *I; 1698 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1699 I = Node.Children.begin(), 1700 E = Node.Children.end(); 1701 I != E; ++I) { 1702 printDebugInfo(*I, "\nChild: "); 1703 } 1704 } 1705 llvm::dbgs() << "\n"; 1706 } 1707 1708 void UnwrappedLineParser::addUnwrappedLine() { 1709 if (Line->Tokens.empty()) 1710 return; 1711 DEBUG({ 1712 if (CurrentLines == &Lines) 1713 printDebugInfo(*Line); 1714 }); 1715 CurrentLines->push_back(*Line); 1716 Line->Tokens.clear(); 1717 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1718 for (SmallVectorImpl<UnwrappedLine>::iterator 1719 I = PreprocessorDirectives.begin(), 1720 E = PreprocessorDirectives.end(); 1721 I != E; ++I) { 1722 CurrentLines->push_back(*I); 1723 } 1724 PreprocessorDirectives.clear(); 1725 } 1726 } 1727 1728 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1729 1730 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1731 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1732 FormatTok.NewlinesBefore > 0; 1733 } 1734 1735 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1736 bool JustComments = Line->Tokens.empty(); 1737 for (SmallVectorImpl<FormatToken *>::const_iterator 1738 I = CommentsBeforeNextToken.begin(), 1739 E = CommentsBeforeNextToken.end(); 1740 I != E; ++I) { 1741 if (isOnNewLine(**I) && JustComments) { 1742 addUnwrappedLine(); 1743 } 1744 pushToken(*I); 1745 } 1746 if (NewlineBeforeNext && JustComments) { 1747 addUnwrappedLine(); 1748 } 1749 CommentsBeforeNextToken.clear(); 1750 } 1751 1752 void UnwrappedLineParser::nextToken() { 1753 if (eof()) 1754 return; 1755 flushComments(isOnNewLine(*FormatTok)); 1756 pushToken(FormatTok); 1757 readToken(); 1758 } 1759 1760 void UnwrappedLineParser::readToken() { 1761 bool CommentsInCurrentLine = true; 1762 do { 1763 FormatTok = Tokens->getNextToken(); 1764 assert(FormatTok); 1765 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1766 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1767 // If there is an unfinished unwrapped line, we flush the preprocessor 1768 // directives only after that unwrapped line was finished later. 1769 bool SwitchToPreprocessorLines = !Line->Tokens.empty(); 1770 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1771 // Comments stored before the preprocessor directive need to be output 1772 // before the preprocessor directive, at the same level as the 1773 // preprocessor directive, as we consider them to apply to the directive. 1774 flushComments(isOnNewLine(*FormatTok)); 1775 parsePPDirective(); 1776 } 1777 while (FormatTok->Type == TT_ConflictStart || 1778 FormatTok->Type == TT_ConflictEnd || 1779 FormatTok->Type == TT_ConflictAlternative) { 1780 if (FormatTok->Type == TT_ConflictStart) { 1781 conditionalCompilationStart(/*Unreachable=*/false); 1782 } else if (FormatTok->Type == TT_ConflictAlternative) { 1783 conditionalCompilationAlternative(); 1784 } else if (FormatTok->Type == TT_ConflictEnd) { 1785 conditionalCompilationEnd(); 1786 } 1787 FormatTok = Tokens->getNextToken(); 1788 FormatTok->MustBreakBefore = true; 1789 } 1790 1791 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1792 !Line->InPPDirective) { 1793 continue; 1794 } 1795 1796 if (!FormatTok->Tok.is(tok::comment)) 1797 return; 1798 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1799 CommentsInCurrentLine = false; 1800 } 1801 if (CommentsInCurrentLine) { 1802 pushToken(FormatTok); 1803 } else { 1804 CommentsBeforeNextToken.push_back(FormatTok); 1805 } 1806 } while (!eof()); 1807 } 1808 1809 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1810 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1811 if (MustBreakBeforeNextToken) { 1812 Line->Tokens.back().Tok->MustBreakBefore = true; 1813 MustBreakBeforeNextToken = false; 1814 } 1815 } 1816 1817 } // end namespace format 1818 } // end namespace clang 1819