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) 62 : Line(Line), TokenSource(TokenSource), ResetToken(ResetToken), 63 PreviousLineLevel(Line.Level), PreviousTokenSource(TokenSource), 64 Token(nullptr) { 65 TokenSource = this; 66 Line.Level = 0; 67 Line.InPPDirective = true; 68 } 69 70 ~ScopedMacroState() override { 71 TokenSource = PreviousTokenSource; 72 ResetToken = Token; 73 Line.InPPDirective = false; 74 Line.Level = PreviousLineLevel; 75 } 76 77 FormatToken *getNextToken() override { 78 // The \c UnwrappedLineParser guards against this by never calling 79 // \c getNextToken() after it has encountered the first eof token. 80 assert(!eof()); 81 Token = PreviousTokenSource->getNextToken(); 82 if (eof()) 83 return getFakeEOF(); 84 return Token; 85 } 86 87 unsigned getPosition() override { return PreviousTokenSource->getPosition(); } 88 89 FormatToken *setPosition(unsigned Position) override { 90 Token = PreviousTokenSource->setPosition(Position); 91 return Token; 92 } 93 94 private: 95 bool eof() { return Token && Token->HasUnescapedNewline; } 96 97 FormatToken *getFakeEOF() { 98 static bool EOFInitialized = false; 99 static FormatToken FormatTok; 100 if (!EOFInitialized) { 101 FormatTok.Tok.startToken(); 102 FormatTok.Tok.setKind(tok::eof); 103 EOFInitialized = true; 104 } 105 return &FormatTok; 106 } 107 108 UnwrappedLine &Line; 109 FormatTokenSource *&TokenSource; 110 FormatToken *&ResetToken; 111 unsigned PreviousLineLevel; 112 FormatTokenSource *PreviousTokenSource; 113 114 FormatToken *Token; 115 }; 116 117 } // end anonymous namespace 118 119 class ScopedLineState { 120 public: 121 ScopedLineState(UnwrappedLineParser &Parser, 122 bool SwitchToPreprocessorLines = false) 123 : Parser(Parser), OriginalLines(Parser.CurrentLines) { 124 if (SwitchToPreprocessorLines) 125 Parser.CurrentLines = &Parser.PreprocessorDirectives; 126 else if (!Parser.Line->Tokens.empty()) 127 Parser.CurrentLines = &Parser.Line->Tokens.back().Children; 128 PreBlockLine = std::move(Parser.Line); 129 Parser.Line = llvm::make_unique<UnwrappedLine>(); 130 Parser.Line->Level = PreBlockLine->Level; 131 Parser.Line->InPPDirective = PreBlockLine->InPPDirective; 132 } 133 134 ~ScopedLineState() { 135 if (!Parser.Line->Tokens.empty()) { 136 Parser.addUnwrappedLine(); 137 } 138 assert(Parser.Line->Tokens.empty()); 139 Parser.Line = std::move(PreBlockLine); 140 if (Parser.CurrentLines == &Parser.PreprocessorDirectives) 141 Parser.MustBreakBeforeNextToken = true; 142 Parser.CurrentLines = OriginalLines; 143 } 144 145 private: 146 UnwrappedLineParser &Parser; 147 148 std::unique_ptr<UnwrappedLine> PreBlockLine; 149 SmallVectorImpl<UnwrappedLine> *OriginalLines; 150 }; 151 152 class CompoundStatementIndenter { 153 public: 154 CompoundStatementIndenter(UnwrappedLineParser *Parser, 155 const FormatStyle &Style, unsigned &LineLevel) 156 : LineLevel(LineLevel), OldLineLevel(LineLevel) { 157 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman) { 158 Parser->addUnwrappedLine(); 159 } else if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 160 Parser->addUnwrappedLine(); 161 ++LineLevel; 162 } 163 } 164 ~CompoundStatementIndenter() { LineLevel = OldLineLevel; } 165 166 private: 167 unsigned &LineLevel; 168 unsigned OldLineLevel; 169 }; 170 171 namespace { 172 173 class IndexedTokenSource : public FormatTokenSource { 174 public: 175 IndexedTokenSource(ArrayRef<FormatToken *> Tokens) 176 : Tokens(Tokens), Position(-1) {} 177 178 FormatToken *getNextToken() override { 179 ++Position; 180 return Tokens[Position]; 181 } 182 183 unsigned getPosition() override { 184 assert(Position >= 0); 185 return Position; 186 } 187 188 FormatToken *setPosition(unsigned P) override { 189 Position = P; 190 return Tokens[Position]; 191 } 192 193 void reset() { Position = -1; } 194 195 private: 196 ArrayRef<FormatToken *> Tokens; 197 int Position; 198 }; 199 200 } // end anonymous namespace 201 202 UnwrappedLineParser::UnwrappedLineParser(const FormatStyle &Style, 203 const AdditionalKeywords &Keywords, 204 ArrayRef<FormatToken *> Tokens, 205 UnwrappedLineConsumer &Callback) 206 : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), 207 CurrentLines(&Lines), Style(Style), Keywords(Keywords), Tokens(nullptr), 208 Callback(Callback), AllTokens(Tokens), PPBranchLevel(-1) {} 209 210 void UnwrappedLineParser::reset() { 211 PPBranchLevel = -1; 212 Line.reset(new UnwrappedLine); 213 CommentsBeforeNextToken.clear(); 214 FormatTok = nullptr; 215 MustBreakBeforeNextToken = false; 216 PreprocessorDirectives.clear(); 217 CurrentLines = &Lines; 218 DeclarationScopeStack.clear(); 219 PPStack.clear(); 220 } 221 222 void UnwrappedLineParser::parse() { 223 IndexedTokenSource TokenSource(AllTokens); 224 do { 225 DEBUG(llvm::dbgs() << "----\n"); 226 reset(); 227 Tokens = &TokenSource; 228 TokenSource.reset(); 229 230 readToken(); 231 parseFile(); 232 // Create line with eof token. 233 pushToken(FormatTok); 234 addUnwrappedLine(); 235 236 for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(), 237 E = Lines.end(); 238 I != E; ++I) { 239 Callback.consumeUnwrappedLine(*I); 240 } 241 Callback.finishRun(); 242 Lines.clear(); 243 while (!PPLevelBranchIndex.empty() && 244 PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) { 245 PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1); 246 PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1); 247 } 248 if (!PPLevelBranchIndex.empty()) { 249 ++PPLevelBranchIndex.back(); 250 assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size()); 251 assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back()); 252 } 253 } while (!PPLevelBranchIndex.empty()); 254 255 } 256 257 void UnwrappedLineParser::parseFile() { 258 // The top-level context in a file always has declarations, except for pre- 259 // processor directives and JavaScript files. 260 bool MustBeDeclaration = 261 !Line->InPPDirective && Style.Language != FormatStyle::LK_JavaScript; 262 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 263 MustBeDeclaration); 264 parseLevel(/*HasOpeningBrace=*/false); 265 // Make sure to format the remaining tokens. 266 flushComments(true); 267 addUnwrappedLine(); 268 } 269 270 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) { 271 bool SwitchLabelEncountered = false; 272 do { 273 switch (FormatTok->Tok.getKind()) { 274 case tok::comment: 275 nextToken(); 276 addUnwrappedLine(); 277 break; 278 case tok::l_brace: 279 // FIXME: Add parameter whether this can happen - if this happens, we must 280 // be in a non-declaration context. 281 parseBlock(/*MustBeDeclaration=*/false); 282 addUnwrappedLine(); 283 break; 284 case tok::r_brace: 285 if (HasOpeningBrace) 286 return; 287 nextToken(); 288 addUnwrappedLine(); 289 break; 290 case tok::kw_default: 291 case tok::kw_case: 292 if (!SwitchLabelEncountered && 293 (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1))) 294 ++Line->Level; 295 SwitchLabelEncountered = true; 296 parseStructuralElement(); 297 break; 298 default: 299 parseStructuralElement(); 300 break; 301 } 302 } while (!eof()); 303 } 304 305 void UnwrappedLineParser::calculateBraceTypes() { 306 // We'll parse forward through the tokens until we hit 307 // a closing brace or eof - note that getNextToken() will 308 // parse macros, so this will magically work inside macro 309 // definitions, too. 310 unsigned StoredPosition = Tokens->getPosition(); 311 FormatToken *Tok = FormatTok; 312 // Keep a stack of positions of lbrace tokens. We will 313 // update information about whether an lbrace starts a 314 // braced init list or a different block during the loop. 315 SmallVector<FormatToken *, 8> LBraceStack; 316 assert(Tok->Tok.is(tok::l_brace)); 317 do { 318 // Get next none-comment token. 319 FormatToken *NextTok; 320 unsigned ReadTokens = 0; 321 do { 322 NextTok = Tokens->getNextToken(); 323 ++ReadTokens; 324 } while (NextTok->is(tok::comment)); 325 326 switch (Tok->Tok.getKind()) { 327 case tok::l_brace: 328 LBraceStack.push_back(Tok); 329 break; 330 case tok::r_brace: 331 if (!LBraceStack.empty()) { 332 if (LBraceStack.back()->BlockKind == BK_Unknown) { 333 bool ProbablyBracedList = false; 334 if (Style.Language == FormatStyle::LK_Proto) { 335 ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); 336 } else { 337 // Using OriginalColumn to distinguish between ObjC methods and 338 // binary operators is a bit hacky. 339 bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) && 340 NextTok->OriginalColumn == 0; 341 342 // If there is a comma, semicolon or right paren after the closing 343 // brace, we assume this is a braced initializer list. Note that 344 // regardless how we mark inner braces here, we will overwrite the 345 // BlockKind later if we parse a braced list (where all blocks 346 // inside are by default braced lists), or when we explicitly detect 347 // blocks (for example while parsing lambdas). 348 // 349 // We exclude + and - as they can be ObjC visibility modifiers. 350 ProbablyBracedList = 351 NextTok->isOneOf(tok::comma, tok::semi, tok::period, tok::colon, 352 tok::r_paren, tok::r_square, tok::l_brace, 353 tok::l_paren, tok::ellipsis) || 354 (NextTok->isBinaryOperator() && !NextIsObjCMethod); 355 } 356 if (ProbablyBracedList) { 357 Tok->BlockKind = BK_BracedInit; 358 LBraceStack.back()->BlockKind = BK_BracedInit; 359 } else { 360 Tok->BlockKind = BK_Block; 361 LBraceStack.back()->BlockKind = BK_Block; 362 } 363 } 364 LBraceStack.pop_back(); 365 } 366 break; 367 case tok::at: 368 case tok::semi: 369 case tok::kw_if: 370 case tok::kw_while: 371 case tok::kw_for: 372 case tok::kw_switch: 373 case tok::kw_try: 374 case tok::kw___try: 375 if (!LBraceStack.empty()) 376 LBraceStack.back()->BlockKind = BK_Block; 377 break; 378 default: 379 break; 380 } 381 Tok = NextTok; 382 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 383 // Assume other blocks for all unclosed opening braces. 384 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 385 if (LBraceStack[i]->BlockKind == BK_Unknown) 386 LBraceStack[i]->BlockKind = BK_Block; 387 } 388 389 FormatTok = Tokens->setPosition(StoredPosition); 390 } 391 392 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 393 bool MunchSemi) { 394 assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected"); 395 unsigned InitialLevel = Line->Level; 396 nextToken(); 397 398 addUnwrappedLine(); 399 400 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 401 MustBeDeclaration); 402 if (AddLevel) 403 ++Line->Level; 404 parseLevel(/*HasOpeningBrace=*/true); 405 406 if (!FormatTok->Tok.is(tok::r_brace)) { 407 Line->Level = InitialLevel; 408 return; 409 } 410 411 nextToken(); // Munch the closing brace. 412 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 413 nextToken(); 414 Line->Level = InitialLevel; 415 } 416 417 static bool isGoogScope(const UnwrappedLine &Line) { 418 // FIXME: Closure-library specific stuff should not be hard-coded but be 419 // configurable. 420 if (Line.Tokens.size() < 4) 421 return false; 422 auto I = Line.Tokens.begin(); 423 if (I->Tok->TokenText != "goog") 424 return false; 425 ++I; 426 if (I->Tok->isNot(tok::period)) 427 return false; 428 ++I; 429 if (I->Tok->TokenText != "scope") 430 return false; 431 ++I; 432 return I->Tok->is(tok::l_paren); 433 } 434 435 static bool ShouldBreakBeforeBrace(const FormatStyle &Style, 436 const FormatToken &InitialToken) { 437 switch (Style.BreakBeforeBraces) { 438 case FormatStyle::BS_Linux: 439 return InitialToken.isOneOf(tok::kw_namespace, tok::kw_class); 440 case FormatStyle::BS_Allman: 441 case FormatStyle::BS_GNU: 442 return true; 443 default: 444 return false; 445 } 446 } 447 448 void UnwrappedLineParser::parseChildBlock() { 449 FormatTok->BlockKind = BK_Block; 450 nextToken(); 451 { 452 bool GoogScope = 453 Style.Language == FormatStyle::LK_JavaScript && isGoogScope(*Line); 454 ScopedLineState LineState(*this); 455 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 456 /*MustBeDeclaration=*/false); 457 Line->Level += GoogScope ? 0 : 1; 458 parseLevel(/*HasOpeningBrace=*/true); 459 flushComments(isOnNewLine(*FormatTok)); 460 Line->Level -= GoogScope ? 0 : 1; 461 } 462 nextToken(); 463 } 464 465 void UnwrappedLineParser::parsePPDirective() { 466 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 467 ScopedMacroState MacroState(*Line, Tokens, FormatTok); 468 nextToken(); 469 470 if (!FormatTok->Tok.getIdentifierInfo()) { 471 parsePPUnknown(); 472 return; 473 } 474 475 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 476 case tok::pp_define: 477 parsePPDefine(); 478 return; 479 case tok::pp_if: 480 parsePPIf(/*IfDef=*/false); 481 break; 482 case tok::pp_ifdef: 483 case tok::pp_ifndef: 484 parsePPIf(/*IfDef=*/true); 485 break; 486 case tok::pp_else: 487 parsePPElse(); 488 break; 489 case tok::pp_elif: 490 parsePPElIf(); 491 break; 492 case tok::pp_endif: 493 parsePPEndIf(); 494 break; 495 default: 496 parsePPUnknown(); 497 break; 498 } 499 } 500 501 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 502 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 503 PPStack.push_back(PP_Unreachable); 504 else 505 PPStack.push_back(PP_Conditional); 506 } 507 508 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 509 ++PPBranchLevel; 510 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 511 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 512 PPLevelBranchIndex.push_back(0); 513 PPLevelBranchCount.push_back(0); 514 } 515 PPChainBranchIndex.push(0); 516 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 517 conditionalCompilationCondition(Unreachable || Skip); 518 } 519 520 void UnwrappedLineParser::conditionalCompilationAlternative() { 521 if (!PPStack.empty()) 522 PPStack.pop_back(); 523 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 524 if (!PPChainBranchIndex.empty()) 525 ++PPChainBranchIndex.top(); 526 conditionalCompilationCondition( 527 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 528 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 529 } 530 531 void UnwrappedLineParser::conditionalCompilationEnd() { 532 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 533 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 534 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 535 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 536 } 537 } 538 // Guard against #endif's without #if. 539 if (PPBranchLevel > 0) 540 --PPBranchLevel; 541 if (!PPChainBranchIndex.empty()) 542 PPChainBranchIndex.pop(); 543 if (!PPStack.empty()) 544 PPStack.pop_back(); 545 } 546 547 void UnwrappedLineParser::parsePPIf(bool IfDef) { 548 nextToken(); 549 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 550 FormatTok->Tok.getLiteralData() != nullptr && 551 StringRef(FormatTok->Tok.getLiteralData(), 552 FormatTok->Tok.getLength()) == "0") || 553 FormatTok->Tok.is(tok::kw_false); 554 conditionalCompilationStart(!IfDef && IsLiteralFalse); 555 parsePPUnknown(); 556 } 557 558 void UnwrappedLineParser::parsePPElse() { 559 conditionalCompilationAlternative(); 560 parsePPUnknown(); 561 } 562 563 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 564 565 void UnwrappedLineParser::parsePPEndIf() { 566 conditionalCompilationEnd(); 567 parsePPUnknown(); 568 } 569 570 void UnwrappedLineParser::parsePPDefine() { 571 nextToken(); 572 573 if (FormatTok->Tok.getKind() != tok::identifier) { 574 parsePPUnknown(); 575 return; 576 } 577 nextToken(); 578 if (FormatTok->Tok.getKind() == tok::l_paren && 579 FormatTok->WhitespaceRange.getBegin() == 580 FormatTok->WhitespaceRange.getEnd()) { 581 parseParens(); 582 } 583 addUnwrappedLine(); 584 Line->Level = 1; 585 586 // Errors during a preprocessor directive can only affect the layout of the 587 // preprocessor directive, and thus we ignore them. An alternative approach 588 // would be to use the same approach we use on the file level (no 589 // re-indentation if there was a structural error) within the macro 590 // definition. 591 parseFile(); 592 } 593 594 void UnwrappedLineParser::parsePPUnknown() { 595 do { 596 nextToken(); 597 } while (!eof()); 598 addUnwrappedLine(); 599 } 600 601 // Here we blacklist certain tokens that are not usually the first token in an 602 // unwrapped line. This is used in attempt to distinguish macro calls without 603 // trailing semicolons from other constructs split to several lines. 604 static bool tokenCanStartNewLine(const clang::Token &Tok) { 605 // Semicolon can be a null-statement, l_square can be a start of a macro or 606 // a C++11 attribute, but this doesn't seem to be common. 607 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 608 Tok.isNot(tok::l_square) && 609 // Tokens that can only be used as binary operators and a part of 610 // overloaded operator names. 611 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 612 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 613 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 614 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 615 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 616 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 617 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 618 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 619 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 620 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 621 Tok.isNot(tok::lesslessequal) && 622 // Colon is used in labels, base class lists, initializer lists, 623 // range-based for loops, ternary operator, but should never be the 624 // first token in an unwrapped line. 625 Tok.isNot(tok::colon) && 626 // 'noexcept' is a trailing annotation. 627 Tok.isNot(tok::kw_noexcept); 628 } 629 630 void UnwrappedLineParser::parseStructuralElement() { 631 assert(!FormatTok->Tok.is(tok::l_brace)); 632 switch (FormatTok->Tok.getKind()) { 633 case tok::at: 634 nextToken(); 635 if (FormatTok->Tok.is(tok::l_brace)) { 636 parseBracedList(); 637 break; 638 } 639 switch (FormatTok->Tok.getObjCKeywordID()) { 640 case tok::objc_public: 641 case tok::objc_protected: 642 case tok::objc_package: 643 case tok::objc_private: 644 return parseAccessSpecifier(); 645 case tok::objc_interface: 646 case tok::objc_implementation: 647 return parseObjCInterfaceOrImplementation(); 648 case tok::objc_protocol: 649 return parseObjCProtocol(); 650 case tok::objc_end: 651 return; // Handled by the caller. 652 case tok::objc_optional: 653 case tok::objc_required: 654 nextToken(); 655 addUnwrappedLine(); 656 return; 657 case tok::objc_try: 658 // This branch isn't strictly necessary (the kw_try case below would 659 // do this too after the tok::at is parsed above). But be explicit. 660 parseTryCatch(); 661 return; 662 default: 663 break; 664 } 665 break; 666 case tok::kw_asm: 667 nextToken(); 668 if (FormatTok->is(tok::l_brace)) { 669 nextToken(); 670 while (FormatTok && FormatTok->isNot(tok::eof)) { 671 if (FormatTok->is(tok::r_brace)) { 672 nextToken(); 673 break; 674 } 675 FormatTok->Finalized = true; 676 nextToken(); 677 } 678 } 679 break; 680 case tok::kw_namespace: 681 parseNamespace(); 682 return; 683 case tok::kw_inline: 684 nextToken(); 685 if (FormatTok->Tok.is(tok::kw_namespace)) { 686 parseNamespace(); 687 return; 688 } 689 break; 690 case tok::kw_public: 691 case tok::kw_protected: 692 case tok::kw_private: 693 if (Style.Language == FormatStyle::LK_Java || 694 Style.Language == FormatStyle::LK_JavaScript) 695 nextToken(); 696 else 697 parseAccessSpecifier(); 698 return; 699 case tok::kw_if: 700 parseIfThenElse(); 701 return; 702 case tok::kw_for: 703 case tok::kw_while: 704 parseForOrWhileLoop(); 705 return; 706 case tok::kw_do: 707 parseDoWhile(); 708 return; 709 case tok::kw_switch: 710 parseSwitch(); 711 return; 712 case tok::kw_default: 713 nextToken(); 714 parseLabel(); 715 return; 716 case tok::kw_case: 717 parseCaseLabel(); 718 return; 719 case tok::kw_try: 720 case tok::kw___try: 721 parseTryCatch(); 722 return; 723 case tok::kw_extern: 724 nextToken(); 725 if (FormatTok->Tok.is(tok::string_literal)) { 726 nextToken(); 727 if (FormatTok->Tok.is(tok::l_brace)) { 728 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 729 addUnwrappedLine(); 730 return; 731 } 732 } 733 break; 734 case tok::kw_export: 735 if (Style.Language == FormatStyle::LK_JavaScript) { 736 parseJavaScriptEs6ImportExport(); 737 return; 738 } 739 break; 740 case tok::identifier: 741 if (FormatTok->is(TT_ForEachMacro)) { 742 parseForOrWhileLoop(); 743 return; 744 } 745 if (Style.Language == FormatStyle::LK_JavaScript && 746 FormatTok->is(Keywords.kw_import)) { 747 parseJavaScriptEs6ImportExport(); 748 return; 749 } 750 if (FormatTok->is(Keywords.kw_signals)) { 751 nextToken(); 752 if (FormatTok->is(tok::colon)) { 753 nextToken(); 754 addUnwrappedLine(); 755 } 756 return; 757 } 758 // In all other cases, parse the declaration. 759 break; 760 default: 761 break; 762 } 763 do { 764 switch (FormatTok->Tok.getKind()) { 765 case tok::at: 766 nextToken(); 767 if (FormatTok->Tok.is(tok::l_brace)) 768 parseBracedList(); 769 break; 770 case tok::kw_enum: 771 parseEnum(); 772 break; 773 case tok::kw_typedef: 774 nextToken(); 775 if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS, 776 Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS)) 777 parseEnum(); 778 break; 779 case tok::kw_struct: 780 case tok::kw_union: 781 case tok::kw_class: 782 parseRecord(); 783 // A record declaration or definition is always the start of a structural 784 // element. 785 break; 786 case tok::period: 787 nextToken(); 788 // In Java, classes have an implicit static member "class". 789 if (Style.Language == FormatStyle::LK_Java && FormatTok && 790 FormatTok->is(tok::kw_class)) 791 nextToken(); 792 break; 793 case tok::semi: 794 nextToken(); 795 addUnwrappedLine(); 796 return; 797 case tok::r_brace: 798 addUnwrappedLine(); 799 return; 800 case tok::l_paren: 801 parseParens(); 802 break; 803 case tok::caret: 804 nextToken(); 805 if (FormatTok->Tok.isAnyIdentifier() || 806 FormatTok->isSimpleTypeSpecifier()) 807 nextToken(); 808 if (FormatTok->is(tok::l_paren)) 809 parseParens(); 810 if (FormatTok->is(tok::l_brace)) 811 parseChildBlock(); 812 break; 813 case tok::l_brace: 814 if (!tryToParseBracedList()) { 815 // A block outside of parentheses must be the last part of a 816 // structural element. 817 // FIXME: Figure out cases where this is not true, and add projections 818 // for them (the one we know is missing are lambdas). 819 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 820 addUnwrappedLine(); 821 FormatTok->Type = TT_FunctionLBrace; 822 parseBlock(/*MustBeDeclaration=*/false); 823 addUnwrappedLine(); 824 return; 825 } 826 // Otherwise this was a braced init list, and the structural 827 // element continues. 828 break; 829 case tok::kw_try: 830 // We arrive here when parsing function-try blocks. 831 parseTryCatch(); 832 return; 833 case tok::identifier: { 834 // Parse function literal unless 'function' is the first token in a line 835 // in which case this should be treated as a free-standing function. 836 if (Style.Language == FormatStyle::LK_JavaScript && 837 FormatTok->is(Keywords.kw_function) && Line->Tokens.size() > 0) { 838 tryToParseJSFunction(); 839 break; 840 } 841 if ((Style.Language == FormatStyle::LK_JavaScript || 842 Style.Language == FormatStyle::LK_Java) && 843 FormatTok->is(Keywords.kw_interface)) { 844 parseRecord(); 845 break; 846 } 847 848 StringRef Text = FormatTok->TokenText; 849 nextToken(); 850 if (Line->Tokens.size() == 1 && 851 // JS doesn't have macros, and within classes colons indicate fields, 852 // not labels. 853 Style.Language != FormatStyle::LK_JavaScript) { 854 if (FormatTok->Tok.is(tok::colon) && !Line->MustBeDeclaration) { 855 parseLabel(); 856 return; 857 } 858 // Recognize function-like macro usages without trailing semicolon as 859 // well as free-standing macros like Q_OBJECT. 860 bool FunctionLike = FormatTok->is(tok::l_paren); 861 if (FunctionLike) 862 parseParens(); 863 if (FormatTok->NewlinesBefore > 0 && 864 (Text.size() >= 5 || FunctionLike) && 865 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 866 addUnwrappedLine(); 867 return; 868 } 869 } 870 break; 871 } 872 case tok::equal: 873 nextToken(); 874 if (FormatTok->Tok.is(tok::l_brace)) { 875 parseBracedList(); 876 } 877 break; 878 case tok::l_square: 879 parseSquare(); 880 break; 881 case tok::kw_new: 882 parseNew(); 883 break; 884 default: 885 nextToken(); 886 break; 887 } 888 } while (!eof()); 889 } 890 891 bool UnwrappedLineParser::tryToParseLambda() { 892 // FIXME: This is a dirty way to access the previous token. Find a better 893 // solution. 894 if (!Line->Tokens.empty() && 895 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator, 896 tok::kw_new, tok::kw_delete) || 897 Line->Tokens.back().Tok->closesScope() || 898 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 899 nextToken(); 900 return false; 901 } 902 assert(FormatTok->is(tok::l_square)); 903 FormatToken &LSquare = *FormatTok; 904 if (!tryToParseLambdaIntroducer()) 905 return false; 906 907 while (FormatTok->isNot(tok::l_brace)) { 908 if (FormatTok->isSimpleTypeSpecifier()) { 909 nextToken(); 910 continue; 911 } 912 switch (FormatTok->Tok.getKind()) { 913 case tok::l_brace: 914 break; 915 case tok::l_paren: 916 parseParens(); 917 break; 918 case tok::amp: 919 case tok::star: 920 case tok::kw_const: 921 case tok::comma: 922 case tok::less: 923 case tok::greater: 924 case tok::identifier: 925 case tok::coloncolon: 926 case tok::kw_mutable: 927 nextToken(); 928 break; 929 case tok::arrow: 930 FormatTok->Type = TT_TrailingReturnArrow; 931 nextToken(); 932 break; 933 default: 934 return true; 935 } 936 } 937 LSquare.Type = TT_LambdaLSquare; 938 parseChildBlock(); 939 return true; 940 } 941 942 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 943 nextToken(); 944 if (FormatTok->is(tok::equal)) { 945 nextToken(); 946 if (FormatTok->is(tok::r_square)) { 947 nextToken(); 948 return true; 949 } 950 if (FormatTok->isNot(tok::comma)) 951 return false; 952 nextToken(); 953 } else if (FormatTok->is(tok::amp)) { 954 nextToken(); 955 if (FormatTok->is(tok::r_square)) { 956 nextToken(); 957 return true; 958 } 959 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 960 return false; 961 } 962 if (FormatTok->is(tok::comma)) 963 nextToken(); 964 } else if (FormatTok->is(tok::r_square)) { 965 nextToken(); 966 return true; 967 } 968 do { 969 if (FormatTok->is(tok::amp)) 970 nextToken(); 971 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 972 return false; 973 nextToken(); 974 if (FormatTok->is(tok::ellipsis)) 975 nextToken(); 976 if (FormatTok->is(tok::comma)) { 977 nextToken(); 978 } else if (FormatTok->is(tok::r_square)) { 979 nextToken(); 980 return true; 981 } else { 982 return false; 983 } 984 } while (!eof()); 985 return false; 986 } 987 988 void UnwrappedLineParser::tryToParseJSFunction() { 989 nextToken(); 990 991 // Consume function name. 992 if (FormatTok->is(tok::identifier)) 993 nextToken(); 994 995 if (FormatTok->isNot(tok::l_paren)) 996 return; 997 nextToken(); 998 while (FormatTok->isNot(tok::l_brace)) { 999 // Err on the side of caution in order to avoid consuming the full file in 1000 // case of incomplete code. 1001 if (!FormatTok->isOneOf(tok::identifier, tok::comma, tok::r_paren, 1002 tok::comment)) 1003 return; 1004 nextToken(); 1005 } 1006 parseChildBlock(); 1007 } 1008 1009 bool UnwrappedLineParser::tryToParseBracedList() { 1010 if (FormatTok->BlockKind == BK_Unknown) 1011 calculateBraceTypes(); 1012 assert(FormatTok->BlockKind != BK_Unknown); 1013 if (FormatTok->BlockKind == BK_Block) 1014 return false; 1015 parseBracedList(); 1016 return true; 1017 } 1018 1019 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 1020 bool HasError = false; 1021 nextToken(); 1022 1023 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 1024 // replace this by using parseAssigmentExpression() inside. 1025 do { 1026 if (Style.Language == FormatStyle::LK_JavaScript && 1027 FormatTok->is(Keywords.kw_function)) { 1028 tryToParseJSFunction(); 1029 continue; 1030 } 1031 switch (FormatTok->Tok.getKind()) { 1032 case tok::caret: 1033 nextToken(); 1034 if (FormatTok->is(tok::l_brace)) { 1035 parseChildBlock(); 1036 } 1037 break; 1038 case tok::l_square: 1039 tryToParseLambda(); 1040 break; 1041 case tok::l_brace: 1042 // Assume there are no blocks inside a braced init list apart 1043 // from the ones we explicitly parse out (like lambdas). 1044 FormatTok->BlockKind = BK_BracedInit; 1045 parseBracedList(); 1046 break; 1047 case tok::r_paren: 1048 // JavaScript can just have free standing methods and getters/setters in 1049 // object literals. Detect them by a "{" following ")". 1050 if (Style.Language == FormatStyle::LK_JavaScript) { 1051 nextToken(); 1052 if (FormatTok->is(tok::l_brace)) 1053 parseChildBlock(); 1054 break; 1055 } 1056 nextToken(); 1057 break; 1058 case tok::r_brace: 1059 nextToken(); 1060 return !HasError; 1061 case tok::semi: 1062 HasError = true; 1063 if (!ContinueOnSemicolons) 1064 return !HasError; 1065 nextToken(); 1066 break; 1067 case tok::comma: 1068 nextToken(); 1069 break; 1070 default: 1071 nextToken(); 1072 break; 1073 } 1074 } while (!eof()); 1075 return false; 1076 } 1077 1078 void UnwrappedLineParser::parseParens() { 1079 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1080 nextToken(); 1081 do { 1082 switch (FormatTok->Tok.getKind()) { 1083 case tok::l_paren: 1084 parseParens(); 1085 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) 1086 parseChildBlock(); 1087 break; 1088 case tok::r_paren: 1089 nextToken(); 1090 return; 1091 case tok::r_brace: 1092 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1093 return; 1094 case tok::l_square: 1095 tryToParseLambda(); 1096 break; 1097 case tok::l_brace: 1098 if (!tryToParseBracedList()) { 1099 parseChildBlock(); 1100 } 1101 break; 1102 case tok::at: 1103 nextToken(); 1104 if (FormatTok->Tok.is(tok::l_brace)) 1105 parseBracedList(); 1106 break; 1107 case tok::identifier: 1108 if (Style.Language == FormatStyle::LK_JavaScript && 1109 FormatTok->is(Keywords.kw_function)) 1110 tryToParseJSFunction(); 1111 else 1112 nextToken(); 1113 break; 1114 default: 1115 nextToken(); 1116 break; 1117 } 1118 } while (!eof()); 1119 } 1120 1121 void UnwrappedLineParser::parseSquare() { 1122 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1123 if (tryToParseLambda()) 1124 return; 1125 do { 1126 switch (FormatTok->Tok.getKind()) { 1127 case tok::l_paren: 1128 parseParens(); 1129 break; 1130 case tok::r_square: 1131 nextToken(); 1132 return; 1133 case tok::r_brace: 1134 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1135 return; 1136 case tok::l_square: 1137 parseSquare(); 1138 break; 1139 case tok::l_brace: { 1140 if (!tryToParseBracedList()) { 1141 parseChildBlock(); 1142 } 1143 break; 1144 } 1145 case tok::at: 1146 nextToken(); 1147 if (FormatTok->Tok.is(tok::l_brace)) 1148 parseBracedList(); 1149 break; 1150 default: 1151 nextToken(); 1152 break; 1153 } 1154 } while (!eof()); 1155 } 1156 1157 void UnwrappedLineParser::parseIfThenElse() { 1158 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1159 nextToken(); 1160 if (FormatTok->Tok.is(tok::l_paren)) 1161 parseParens(); 1162 bool NeedsUnwrappedLine = false; 1163 if (FormatTok->Tok.is(tok::l_brace)) { 1164 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1165 parseBlock(/*MustBeDeclaration=*/false); 1166 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1167 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1168 addUnwrappedLine(); 1169 } else { 1170 NeedsUnwrappedLine = true; 1171 } 1172 } else { 1173 addUnwrappedLine(); 1174 ++Line->Level; 1175 parseStructuralElement(); 1176 --Line->Level; 1177 } 1178 if (FormatTok->Tok.is(tok::kw_else)) { 1179 if (Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) 1180 addUnwrappedLine(); 1181 nextToken(); 1182 if (FormatTok->Tok.is(tok::l_brace)) { 1183 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1184 parseBlock(/*MustBeDeclaration=*/false); 1185 addUnwrappedLine(); 1186 } else if (FormatTok->Tok.is(tok::kw_if)) { 1187 parseIfThenElse(); 1188 } else { 1189 addUnwrappedLine(); 1190 ++Line->Level; 1191 parseStructuralElement(); 1192 --Line->Level; 1193 } 1194 } else if (NeedsUnwrappedLine) { 1195 addUnwrappedLine(); 1196 } 1197 } 1198 1199 void UnwrappedLineParser::parseTryCatch() { 1200 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected"); 1201 nextToken(); 1202 bool NeedsUnwrappedLine = false; 1203 if (FormatTok->is(tok::colon)) { 1204 // We are in a function try block, what comes is an initializer list. 1205 nextToken(); 1206 while (FormatTok->is(tok::identifier)) { 1207 nextToken(); 1208 if (FormatTok->is(tok::l_paren)) 1209 parseParens(); 1210 if (FormatTok->is(tok::comma)) 1211 nextToken(); 1212 } 1213 } 1214 // Parse try with resource. 1215 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) { 1216 parseParens(); 1217 } 1218 if (FormatTok->is(tok::l_brace)) { 1219 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1220 parseBlock(/*MustBeDeclaration=*/false); 1221 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1222 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1223 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1224 addUnwrappedLine(); 1225 } else { 1226 NeedsUnwrappedLine = true; 1227 } 1228 } else if (!FormatTok->is(tok::kw_catch)) { 1229 // The C++ standard requires a compound-statement after a try. 1230 // If there's none, we try to assume there's a structuralElement 1231 // and try to continue. 1232 addUnwrappedLine(); 1233 ++Line->Level; 1234 parseStructuralElement(); 1235 --Line->Level; 1236 } 1237 while (1) { 1238 if (FormatTok->is(tok::at)) 1239 nextToken(); 1240 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, 1241 tok::kw___finally) || 1242 ((Style.Language == FormatStyle::LK_Java || 1243 Style.Language == FormatStyle::LK_JavaScript) && 1244 FormatTok->is(Keywords.kw_finally)) || 1245 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) || 1246 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally)))) 1247 break; 1248 nextToken(); 1249 while (FormatTok->isNot(tok::l_brace)) { 1250 if (FormatTok->is(tok::l_paren)) { 1251 parseParens(); 1252 continue; 1253 } 1254 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) 1255 return; 1256 nextToken(); 1257 } 1258 NeedsUnwrappedLine = false; 1259 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1260 parseBlock(/*MustBeDeclaration=*/false); 1261 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1262 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1263 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1264 addUnwrappedLine(); 1265 } else { 1266 NeedsUnwrappedLine = true; 1267 } 1268 } 1269 if (NeedsUnwrappedLine) { 1270 addUnwrappedLine(); 1271 } 1272 } 1273 1274 void UnwrappedLineParser::parseNamespace() { 1275 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1276 1277 const FormatToken &InitialToken = *FormatTok; 1278 nextToken(); 1279 if (FormatTok->Tok.is(tok::identifier)) 1280 nextToken(); 1281 if (FormatTok->Tok.is(tok::l_brace)) { 1282 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1283 addUnwrappedLine(); 1284 1285 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1286 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1287 DeclarationScopeStack.size() > 1); 1288 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1289 // Munch the semicolon after a namespace. This is more common than one would 1290 // think. Puttin the semicolon into its own line is very ugly. 1291 if (FormatTok->Tok.is(tok::semi)) 1292 nextToken(); 1293 addUnwrappedLine(); 1294 } 1295 // FIXME: Add error handling. 1296 } 1297 1298 void UnwrappedLineParser::parseNew() { 1299 assert(FormatTok->is(tok::kw_new) && "'new' expected"); 1300 nextToken(); 1301 if (Style.Language != FormatStyle::LK_Java) 1302 return; 1303 1304 // In Java, we can parse everything up to the parens, which aren't optional. 1305 do { 1306 // There should not be a ;, { or } before the new's open paren. 1307 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) 1308 return; 1309 1310 // Consume the parens. 1311 if (FormatTok->is(tok::l_paren)) { 1312 parseParens(); 1313 1314 // If there is a class body of an anonymous class, consume that as child. 1315 if (FormatTok->is(tok::l_brace)) 1316 parseChildBlock(); 1317 return; 1318 } 1319 nextToken(); 1320 } while (!eof()); 1321 } 1322 1323 void UnwrappedLineParser::parseForOrWhileLoop() { 1324 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) && 1325 "'for', 'while' or foreach macro expected"); 1326 nextToken(); 1327 if (FormatTok->Tok.is(tok::l_paren)) 1328 parseParens(); 1329 if (FormatTok->Tok.is(tok::l_brace)) { 1330 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1331 parseBlock(/*MustBeDeclaration=*/false); 1332 addUnwrappedLine(); 1333 } else { 1334 addUnwrappedLine(); 1335 ++Line->Level; 1336 parseStructuralElement(); 1337 --Line->Level; 1338 } 1339 } 1340 1341 void UnwrappedLineParser::parseDoWhile() { 1342 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1343 nextToken(); 1344 if (FormatTok->Tok.is(tok::l_brace)) { 1345 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1346 parseBlock(/*MustBeDeclaration=*/false); 1347 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1348 addUnwrappedLine(); 1349 } else { 1350 addUnwrappedLine(); 1351 ++Line->Level; 1352 parseStructuralElement(); 1353 --Line->Level; 1354 } 1355 1356 // FIXME: Add error handling. 1357 if (!FormatTok->Tok.is(tok::kw_while)) { 1358 addUnwrappedLine(); 1359 return; 1360 } 1361 1362 nextToken(); 1363 parseStructuralElement(); 1364 } 1365 1366 void UnwrappedLineParser::parseLabel() { 1367 nextToken(); 1368 unsigned OldLineLevel = Line->Level; 1369 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1370 --Line->Level; 1371 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1372 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1373 parseBlock(/*MustBeDeclaration=*/false); 1374 if (FormatTok->Tok.is(tok::kw_break)) { 1375 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1376 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1377 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1378 addUnwrappedLine(); 1379 } 1380 parseStructuralElement(); 1381 } 1382 addUnwrappedLine(); 1383 } else { 1384 if (FormatTok->is(tok::semi)) 1385 nextToken(); 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 1537 1538 // The actual identifier can be a nested name specifier, and in macros 1539 // it is often token-pasted. 1540 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash, 1541 tok::kw___attribute, tok::kw___declspec, 1542 tok::kw_alignas) || 1543 ((Style.Language == FormatStyle::LK_Java || 1544 Style.Language == FormatStyle::LK_JavaScript) && 1545 FormatTok->isOneOf(tok::period, tok::comma))) { 1546 bool IsNonMacroIdentifier = 1547 FormatTok->is(tok::identifier) && 1548 FormatTok->TokenText != FormatTok->TokenText.upper(); 1549 nextToken(); 1550 // We can have macros or attributes in between 'class' and the class name. 1551 if (!IsNonMacroIdentifier && FormatTok->Tok.is(tok::l_paren)) 1552 parseParens(); 1553 } 1554 1555 // Note that parsing away template declarations here leads to incorrectly 1556 // accepting function declarations as record declarations. 1557 // In general, we cannot solve this problem. Consider: 1558 // class A<int> B() {} 1559 // which can be a function definition or a class definition when B() is a 1560 // macro. If we find enough real-world cases where this is a problem, we 1561 // can parse for the 'template' keyword in the beginning of the statement, 1562 // and thus rule out the record production in case there is no template 1563 // (this would still leave us with an ambiguity between template function 1564 // and class declarations). 1565 if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) { 1566 while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) { 1567 if (FormatTok->Tok.is(tok::semi)) 1568 return; 1569 nextToken(); 1570 } 1571 } 1572 if (FormatTok->Tok.is(tok::l_brace)) { 1573 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1574 addUnwrappedLine(); 1575 1576 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1577 /*MunchSemi=*/false); 1578 } 1579 // We fall through to parsing a structural element afterwards, so 1580 // class A {} n, m; 1581 // will end up in one unwrapped line. 1582 // This does not apply for Java and JavaScript. 1583 if (Style.Language == FormatStyle::LK_Java || 1584 Style.Language == FormatStyle::LK_JavaScript) 1585 addUnwrappedLine(); 1586 } 1587 1588 void UnwrappedLineParser::parseObjCProtocolList() { 1589 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1590 do 1591 nextToken(); 1592 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1593 nextToken(); // Skip '>'. 1594 } 1595 1596 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1597 do { 1598 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1599 nextToken(); 1600 addUnwrappedLine(); 1601 break; 1602 } 1603 if (FormatTok->is(tok::l_brace)) { 1604 parseBlock(/*MustBeDeclaration=*/false); 1605 // In ObjC interfaces, nothing should be following the "}". 1606 addUnwrappedLine(); 1607 } else if (FormatTok->is(tok::r_brace)) { 1608 // Ignore stray "}". parseStructuralElement doesn't consume them. 1609 nextToken(); 1610 addUnwrappedLine(); 1611 } else { 1612 parseStructuralElement(); 1613 } 1614 } while (!eof()); 1615 } 1616 1617 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1618 nextToken(); 1619 nextToken(); // interface name 1620 1621 // @interface can be followed by either a base class, or a category. 1622 if (FormatTok->Tok.is(tok::colon)) { 1623 nextToken(); 1624 nextToken(); // base class name 1625 } else if (FormatTok->Tok.is(tok::l_paren)) 1626 // Skip category, if present. 1627 parseParens(); 1628 1629 if (FormatTok->Tok.is(tok::less)) 1630 parseObjCProtocolList(); 1631 1632 if (FormatTok->Tok.is(tok::l_brace)) { 1633 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1634 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1635 addUnwrappedLine(); 1636 parseBlock(/*MustBeDeclaration=*/true); 1637 } 1638 1639 // With instance variables, this puts '}' on its own line. Without instance 1640 // variables, this ends the @interface line. 1641 addUnwrappedLine(); 1642 1643 parseObjCUntilAtEnd(); 1644 } 1645 1646 void UnwrappedLineParser::parseObjCProtocol() { 1647 nextToken(); 1648 nextToken(); // protocol name 1649 1650 if (FormatTok->Tok.is(tok::less)) 1651 parseObjCProtocolList(); 1652 1653 // Check for protocol declaration. 1654 if (FormatTok->Tok.is(tok::semi)) { 1655 nextToken(); 1656 return addUnwrappedLine(); 1657 } 1658 1659 addUnwrappedLine(); 1660 parseObjCUntilAtEnd(); 1661 } 1662 1663 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { 1664 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export)); 1665 nextToken(); 1666 1667 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, Keywords.kw_function, 1668 Keywords.kw_var)) 1669 return; // Fall through to parsing the corresponding structure. 1670 1671 if (FormatTok->is(tok::kw_default)) { 1672 nextToken(); // export default ..., fall through after eating 'default'. 1673 return; 1674 } 1675 1676 if (FormatTok->is(tok::l_brace)) { 1677 FormatTok->BlockKind = BK_Block; 1678 parseBracedList(); 1679 } 1680 1681 while (!eof() && FormatTok->isNot(tok::semi) && 1682 FormatTok->isNot(tok::l_brace)) { 1683 nextToken(); 1684 } 1685 } 1686 1687 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1688 StringRef Prefix = "") { 1689 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1690 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1691 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1692 E = Line.Tokens.end(); 1693 I != E; ++I) { 1694 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1695 } 1696 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1697 E = Line.Tokens.end(); 1698 I != E; ++I) { 1699 const UnwrappedLineNode &Node = *I; 1700 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1701 I = Node.Children.begin(), 1702 E = Node.Children.end(); 1703 I != E; ++I) { 1704 printDebugInfo(*I, "\nChild: "); 1705 } 1706 } 1707 llvm::dbgs() << "\n"; 1708 } 1709 1710 void UnwrappedLineParser::addUnwrappedLine() { 1711 if (Line->Tokens.empty()) 1712 return; 1713 DEBUG({ 1714 if (CurrentLines == &Lines) 1715 printDebugInfo(*Line); 1716 }); 1717 CurrentLines->push_back(*Line); 1718 Line->Tokens.clear(); 1719 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1720 for (SmallVectorImpl<UnwrappedLine>::iterator 1721 I = PreprocessorDirectives.begin(), 1722 E = PreprocessorDirectives.end(); 1723 I != E; ++I) { 1724 CurrentLines->push_back(*I); 1725 } 1726 PreprocessorDirectives.clear(); 1727 } 1728 } 1729 1730 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1731 1732 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1733 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1734 FormatTok.NewlinesBefore > 0; 1735 } 1736 1737 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1738 bool JustComments = Line->Tokens.empty(); 1739 for (SmallVectorImpl<FormatToken *>::const_iterator 1740 I = CommentsBeforeNextToken.begin(), 1741 E = CommentsBeforeNextToken.end(); 1742 I != E; ++I) { 1743 if (isOnNewLine(**I) && JustComments) { 1744 addUnwrappedLine(); 1745 } 1746 pushToken(*I); 1747 } 1748 if (NewlineBeforeNext && JustComments) { 1749 addUnwrappedLine(); 1750 } 1751 CommentsBeforeNextToken.clear(); 1752 } 1753 1754 void UnwrappedLineParser::nextToken() { 1755 if (eof()) 1756 return; 1757 flushComments(isOnNewLine(*FormatTok)); 1758 pushToken(FormatTok); 1759 readToken(); 1760 } 1761 1762 void UnwrappedLineParser::readToken() { 1763 bool CommentsInCurrentLine = true; 1764 do { 1765 FormatTok = Tokens->getNextToken(); 1766 assert(FormatTok); 1767 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1768 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1769 // If there is an unfinished unwrapped line, we flush the preprocessor 1770 // directives only after that unwrapped line was finished later. 1771 bool SwitchToPreprocessorLines = !Line->Tokens.empty(); 1772 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1773 // Comments stored before the preprocessor directive need to be output 1774 // before the preprocessor directive, at the same level as the 1775 // preprocessor directive, as we consider them to apply to the directive. 1776 flushComments(isOnNewLine(*FormatTok)); 1777 parsePPDirective(); 1778 } 1779 while (FormatTok->Type == TT_ConflictStart || 1780 FormatTok->Type == TT_ConflictEnd || 1781 FormatTok->Type == TT_ConflictAlternative) { 1782 if (FormatTok->Type == TT_ConflictStart) { 1783 conditionalCompilationStart(/*Unreachable=*/false); 1784 } else if (FormatTok->Type == TT_ConflictAlternative) { 1785 conditionalCompilationAlternative(); 1786 } else if (FormatTok->Type == TT_ConflictEnd) { 1787 conditionalCompilationEnd(); 1788 } 1789 FormatTok = Tokens->getNextToken(); 1790 FormatTok->MustBreakBefore = true; 1791 } 1792 1793 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1794 !Line->InPPDirective) { 1795 continue; 1796 } 1797 1798 if (!FormatTok->Tok.is(tok::comment)) 1799 return; 1800 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1801 CommentsInCurrentLine = false; 1802 } 1803 if (CommentsInCurrentLine) { 1804 pushToken(FormatTok); 1805 } else { 1806 CommentsBeforeNextToken.push_back(FormatTok); 1807 } 1808 } while (!eof()); 1809 } 1810 1811 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1812 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1813 if (MustBreakBeforeNextToken) { 1814 Line->Tokens.back().Tok->MustBreakBefore = true; 1815 MustBreakBeforeNextToken = false; 1816 } 1817 } 1818 1819 } // end namespace format 1820 } // end namespace clang 1821