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(bool ExpectClassBody) { 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 Tok->BlockKind = BK_Unknown; 329 LBraceStack.push_back(Tok); 330 break; 331 case tok::r_brace: 332 if (!LBraceStack.empty()) { 333 if (LBraceStack.back()->BlockKind == BK_Unknown) { 334 bool ProbablyBracedList = false; 335 if (Style.Language == FormatStyle::LK_Proto) { 336 ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); 337 } else { 338 // Using OriginalColumn to distinguish between ObjC methods and 339 // binary operators is a bit hacky. 340 bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) && 341 NextTok->OriginalColumn == 0; 342 343 // If there is a comma, semicolon or right paren after the closing 344 // brace, we assume this is a braced initializer list. Note that 345 // regardless how we mark inner braces here, we will overwrite the 346 // BlockKind later if we parse a braced list (where all blocks 347 // inside are by default braced lists), or when we explicitly detect 348 // blocks (for example while parsing lambdas). 349 // 350 // We exclude + and - as they can be ObjC visibility modifiers. 351 ProbablyBracedList = 352 NextTok->isOneOf(tok::comma, tok::period, tok::colon, 353 tok::r_paren, tok::r_square, tok::l_brace, 354 tok::l_paren, tok::ellipsis) || 355 (NextTok->is(tok::semi) && 356 (!ExpectClassBody || LBraceStack.size() != 1)) || 357 (NextTok->isBinaryOperator() && !NextIsObjCMethod); 358 } 359 if (ProbablyBracedList) { 360 Tok->BlockKind = BK_BracedInit; 361 LBraceStack.back()->BlockKind = BK_BracedInit; 362 } else { 363 Tok->BlockKind = BK_Block; 364 LBraceStack.back()->BlockKind = BK_Block; 365 } 366 } 367 LBraceStack.pop_back(); 368 } 369 break; 370 case tok::at: 371 case tok::semi: 372 case tok::kw_if: 373 case tok::kw_while: 374 case tok::kw_for: 375 case tok::kw_switch: 376 case tok::kw_try: 377 case tok::kw___try: 378 if (!LBraceStack.empty()) 379 LBraceStack.back()->BlockKind = BK_Block; 380 break; 381 default: 382 break; 383 } 384 Tok = NextTok; 385 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 386 // Assume other blocks for all unclosed opening braces. 387 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 388 if (LBraceStack[i]->BlockKind == BK_Unknown) 389 LBraceStack[i]->BlockKind = BK_Block; 390 } 391 392 FormatTok = Tokens->setPosition(StoredPosition); 393 } 394 395 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 396 bool MunchSemi) { 397 assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected"); 398 unsigned InitialLevel = Line->Level; 399 nextToken(); 400 401 addUnwrappedLine(); 402 403 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 404 MustBeDeclaration); 405 if (AddLevel) 406 ++Line->Level; 407 parseLevel(/*HasOpeningBrace=*/true); 408 409 if (!FormatTok->Tok.is(tok::r_brace)) { 410 Line->Level = InitialLevel; 411 return; 412 } 413 414 nextToken(); // Munch the closing brace. 415 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 416 nextToken(); 417 Line->Level = InitialLevel; 418 } 419 420 static bool isGoogScope(const UnwrappedLine &Line) { 421 // FIXME: Closure-library specific stuff should not be hard-coded but be 422 // configurable. 423 if (Line.Tokens.size() < 4) 424 return false; 425 auto I = Line.Tokens.begin(); 426 if (I->Tok->TokenText != "goog") 427 return false; 428 ++I; 429 if (I->Tok->isNot(tok::period)) 430 return false; 431 ++I; 432 if (I->Tok->TokenText != "scope") 433 return false; 434 ++I; 435 return I->Tok->is(tok::l_paren); 436 } 437 438 static bool ShouldBreakBeforeBrace(const FormatStyle &Style, 439 const FormatToken &InitialToken) { 440 switch (Style.BreakBeforeBraces) { 441 case FormatStyle::BS_Linux: 442 return InitialToken.isOneOf(tok::kw_namespace, tok::kw_class); 443 case FormatStyle::BS_Allman: 444 case FormatStyle::BS_GNU: 445 return true; 446 default: 447 return false; 448 } 449 } 450 451 void UnwrappedLineParser::parseChildBlock() { 452 FormatTok->BlockKind = BK_Block; 453 nextToken(); 454 { 455 bool GoogScope = 456 Style.Language == FormatStyle::LK_JavaScript && isGoogScope(*Line); 457 ScopedLineState LineState(*this); 458 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 459 /*MustBeDeclaration=*/false); 460 Line->Level += GoogScope ? 0 : 1; 461 parseLevel(/*HasOpeningBrace=*/true); 462 flushComments(isOnNewLine(*FormatTok)); 463 Line->Level -= GoogScope ? 0 : 1; 464 } 465 nextToken(); 466 } 467 468 void UnwrappedLineParser::parsePPDirective() { 469 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 470 ScopedMacroState MacroState(*Line, Tokens, FormatTok); 471 nextToken(); 472 473 if (!FormatTok->Tok.getIdentifierInfo()) { 474 parsePPUnknown(); 475 return; 476 } 477 478 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 479 case tok::pp_define: 480 parsePPDefine(); 481 return; 482 case tok::pp_if: 483 parsePPIf(/*IfDef=*/false); 484 break; 485 case tok::pp_ifdef: 486 case tok::pp_ifndef: 487 parsePPIf(/*IfDef=*/true); 488 break; 489 case tok::pp_else: 490 parsePPElse(); 491 break; 492 case tok::pp_elif: 493 parsePPElIf(); 494 break; 495 case tok::pp_endif: 496 parsePPEndIf(); 497 break; 498 default: 499 parsePPUnknown(); 500 break; 501 } 502 } 503 504 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 505 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 506 PPStack.push_back(PP_Unreachable); 507 else 508 PPStack.push_back(PP_Conditional); 509 } 510 511 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 512 ++PPBranchLevel; 513 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 514 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 515 PPLevelBranchIndex.push_back(0); 516 PPLevelBranchCount.push_back(0); 517 } 518 PPChainBranchIndex.push(0); 519 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 520 conditionalCompilationCondition(Unreachable || Skip); 521 } 522 523 void UnwrappedLineParser::conditionalCompilationAlternative() { 524 if (!PPStack.empty()) 525 PPStack.pop_back(); 526 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 527 if (!PPChainBranchIndex.empty()) 528 ++PPChainBranchIndex.top(); 529 conditionalCompilationCondition( 530 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 531 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 532 } 533 534 void UnwrappedLineParser::conditionalCompilationEnd() { 535 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 536 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 537 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 538 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 539 } 540 } 541 // Guard against #endif's without #if. 542 if (PPBranchLevel > 0) 543 --PPBranchLevel; 544 if (!PPChainBranchIndex.empty()) 545 PPChainBranchIndex.pop(); 546 if (!PPStack.empty()) 547 PPStack.pop_back(); 548 } 549 550 void UnwrappedLineParser::parsePPIf(bool IfDef) { 551 nextToken(); 552 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 553 FormatTok->Tok.getLiteralData() != nullptr && 554 StringRef(FormatTok->Tok.getLiteralData(), 555 FormatTok->Tok.getLength()) == "0") || 556 FormatTok->Tok.is(tok::kw_false); 557 conditionalCompilationStart(!IfDef && IsLiteralFalse); 558 parsePPUnknown(); 559 } 560 561 void UnwrappedLineParser::parsePPElse() { 562 conditionalCompilationAlternative(); 563 parsePPUnknown(); 564 } 565 566 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 567 568 void UnwrappedLineParser::parsePPEndIf() { 569 conditionalCompilationEnd(); 570 parsePPUnknown(); 571 } 572 573 void UnwrappedLineParser::parsePPDefine() { 574 nextToken(); 575 576 if (FormatTok->Tok.getKind() != tok::identifier) { 577 parsePPUnknown(); 578 return; 579 } 580 nextToken(); 581 if (FormatTok->Tok.getKind() == tok::l_paren && 582 FormatTok->WhitespaceRange.getBegin() == 583 FormatTok->WhitespaceRange.getEnd()) { 584 parseParens(); 585 } 586 addUnwrappedLine(); 587 Line->Level = 1; 588 589 // Errors during a preprocessor directive can only affect the layout of the 590 // preprocessor directive, and thus we ignore them. An alternative approach 591 // would be to use the same approach we use on the file level (no 592 // re-indentation if there was a structural error) within the macro 593 // definition. 594 parseFile(); 595 } 596 597 void UnwrappedLineParser::parsePPUnknown() { 598 do { 599 nextToken(); 600 } while (!eof()); 601 addUnwrappedLine(); 602 } 603 604 // Here we blacklist certain tokens that are not usually the first token in an 605 // unwrapped line. This is used in attempt to distinguish macro calls without 606 // trailing semicolons from other constructs split to several lines. 607 static bool tokenCanStartNewLine(const clang::Token &Tok) { 608 // Semicolon can be a null-statement, l_square can be a start of a macro or 609 // a C++11 attribute, but this doesn't seem to be common. 610 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 611 Tok.isNot(tok::l_square) && 612 // Tokens that can only be used as binary operators and a part of 613 // overloaded operator names. 614 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 615 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 616 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 617 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 618 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 619 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 620 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 621 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 622 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 623 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 624 Tok.isNot(tok::lesslessequal) && 625 // Colon is used in labels, base class lists, initializer lists, 626 // range-based for loops, ternary operator, but should never be the 627 // first token in an unwrapped line. 628 Tok.isNot(tok::colon) && 629 // 'noexcept' is a trailing annotation. 630 Tok.isNot(tok::kw_noexcept); 631 } 632 633 void UnwrappedLineParser::parseStructuralElement() { 634 assert(!FormatTok->Tok.is(tok::l_brace)); 635 switch (FormatTok->Tok.getKind()) { 636 case tok::at: 637 nextToken(); 638 if (FormatTok->Tok.is(tok::l_brace)) { 639 parseBracedList(); 640 break; 641 } 642 switch (FormatTok->Tok.getObjCKeywordID()) { 643 case tok::objc_public: 644 case tok::objc_protected: 645 case tok::objc_package: 646 case tok::objc_private: 647 return parseAccessSpecifier(); 648 case tok::objc_interface: 649 case tok::objc_implementation: 650 return parseObjCInterfaceOrImplementation(); 651 case tok::objc_protocol: 652 return parseObjCProtocol(); 653 case tok::objc_end: 654 return; // Handled by the caller. 655 case tok::objc_optional: 656 case tok::objc_required: 657 nextToken(); 658 addUnwrappedLine(); 659 return; 660 case tok::objc_try: 661 // This branch isn't strictly necessary (the kw_try case below would 662 // do this too after the tok::at is parsed above). But be explicit. 663 parseTryCatch(); 664 return; 665 default: 666 break; 667 } 668 break; 669 case tok::kw_asm: 670 nextToken(); 671 if (FormatTok->is(tok::l_brace)) { 672 FormatTok->Type = TT_InlineASMBrace; 673 nextToken(); 674 while (FormatTok && FormatTok->isNot(tok::eof)) { 675 if (FormatTok->is(tok::r_brace)) { 676 FormatTok->Type = TT_InlineASMBrace; 677 nextToken(); 678 addUnwrappedLine(); 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 // Parse function literal unless 'function' is the first token in a line 841 // in which case this should be treated as a free-standing function. 842 if (Style.Language == FormatStyle::LK_JavaScript && 843 FormatTok->is(Keywords.kw_function) && Line->Tokens.size() > 0) { 844 tryToParseJSFunction(); 845 break; 846 } 847 if ((Style.Language == FormatStyle::LK_JavaScript || 848 Style.Language == FormatStyle::LK_Java) && 849 FormatTok->is(Keywords.kw_interface)) { 850 parseRecord(); 851 break; 852 } 853 854 StringRef Text = FormatTok->TokenText; 855 nextToken(); 856 if (Line->Tokens.size() == 1 && 857 // JS doesn't have macros, and within classes colons indicate fields, 858 // not labels. 859 Style.Language != FormatStyle::LK_JavaScript) { 860 if (FormatTok->Tok.is(tok::colon) && !Line->MustBeDeclaration) { 861 parseLabel(); 862 return; 863 } 864 // Recognize function-like macro usages without trailing semicolon as 865 // well as free-standing macros like Q_OBJECT. 866 bool FunctionLike = FormatTok->is(tok::l_paren); 867 if (FunctionLike) 868 parseParens(); 869 870 bool FollowedByNewline = 871 CommentsBeforeNextToken.empty() 872 ? FormatTok->NewlinesBefore > 0 873 : CommentsBeforeNextToken.front()->NewlinesBefore > 0; 874 875 if (FollowedByNewline && 876 (Text.size() >= 5 || FunctionLike) && 877 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 878 addUnwrappedLine(); 879 return; 880 } 881 } 882 break; 883 } 884 case tok::equal: 885 // Fat arrows (=>) have tok::TokenKind tok::equal but TokenType 886 // TT_JsFatArrow. The always start an expression or a child block if 887 // followed by a curly. 888 if (FormatTok->is(TT_JsFatArrow)) { 889 nextToken(); 890 if (FormatTok->is(tok::l_brace)) { 891 parseChildBlock(); 892 } 893 break; 894 } 895 896 nextToken(); 897 if (FormatTok->Tok.is(tok::l_brace)) { 898 parseBracedList(); 899 } 900 break; 901 case tok::l_square: 902 parseSquare(); 903 break; 904 case tok::kw_new: 905 parseNew(); 906 break; 907 default: 908 nextToken(); 909 break; 910 } 911 } while (!eof()); 912 } 913 914 bool UnwrappedLineParser::tryToParseLambda() { 915 // FIXME: This is a dirty way to access the previous token. Find a better 916 // solution. 917 if (!Line->Tokens.empty() && 918 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator, 919 tok::kw_new, tok::kw_delete) || 920 Line->Tokens.back().Tok->closesScope() || 921 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 922 nextToken(); 923 return false; 924 } 925 assert(FormatTok->is(tok::l_square)); 926 FormatToken &LSquare = *FormatTok; 927 if (!tryToParseLambdaIntroducer()) 928 return false; 929 930 while (FormatTok->isNot(tok::l_brace)) { 931 if (FormatTok->isSimpleTypeSpecifier()) { 932 nextToken(); 933 continue; 934 } 935 switch (FormatTok->Tok.getKind()) { 936 case tok::l_brace: 937 break; 938 case tok::l_paren: 939 parseParens(); 940 break; 941 case tok::amp: 942 case tok::star: 943 case tok::kw_const: 944 case tok::comma: 945 case tok::less: 946 case tok::greater: 947 case tok::identifier: 948 case tok::coloncolon: 949 case tok::kw_mutable: 950 nextToken(); 951 break; 952 case tok::arrow: 953 FormatTok->Type = TT_TrailingReturnArrow; 954 nextToken(); 955 break; 956 default: 957 return true; 958 } 959 } 960 LSquare.Type = TT_LambdaLSquare; 961 parseChildBlock(); 962 return true; 963 } 964 965 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 966 nextToken(); 967 if (FormatTok->is(tok::equal)) { 968 nextToken(); 969 if (FormatTok->is(tok::r_square)) { 970 nextToken(); 971 return true; 972 } 973 if (FormatTok->isNot(tok::comma)) 974 return false; 975 nextToken(); 976 } else if (FormatTok->is(tok::amp)) { 977 nextToken(); 978 if (FormatTok->is(tok::r_square)) { 979 nextToken(); 980 return true; 981 } 982 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 983 return false; 984 } 985 if (FormatTok->is(tok::comma)) 986 nextToken(); 987 } else if (FormatTok->is(tok::r_square)) { 988 nextToken(); 989 return true; 990 } 991 do { 992 if (FormatTok->is(tok::amp)) 993 nextToken(); 994 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 995 return false; 996 nextToken(); 997 if (FormatTok->is(tok::ellipsis)) 998 nextToken(); 999 if (FormatTok->is(tok::comma)) { 1000 nextToken(); 1001 } else if (FormatTok->is(tok::r_square)) { 1002 nextToken(); 1003 return true; 1004 } else { 1005 return false; 1006 } 1007 } while (!eof()); 1008 return false; 1009 } 1010 1011 void UnwrappedLineParser::tryToParseJSFunction() { 1012 nextToken(); 1013 1014 // Consume function name. 1015 if (FormatTok->is(tok::identifier)) 1016 nextToken(); 1017 1018 if (FormatTok->isNot(tok::l_paren)) 1019 return; 1020 1021 // Parse formal parameter list. 1022 parseBalanced(tok::l_paren, tok::r_paren); 1023 1024 if (FormatTok->is(tok::colon)) { 1025 // Parse a type definition. 1026 nextToken(); 1027 1028 // Eat the type declaration. For braced inline object types, balance braces, 1029 // otherwise just parse until finding an l_brace for the function body. 1030 if (FormatTok->is(tok::l_brace)) { 1031 parseBalanced(tok::l_brace, tok::r_brace); 1032 } else { 1033 while(FormatTok->isNot(tok::l_brace) && !eof()) { 1034 nextToken(); 1035 } 1036 } 1037 } 1038 1039 parseChildBlock(); 1040 } 1041 1042 void UnwrappedLineParser::parseBalanced(tok::TokenKind OpenKind, 1043 tok::TokenKind CloseKind) { 1044 assert(FormatTok->is(OpenKind)); 1045 nextToken(); 1046 int Depth = 1; 1047 while (Depth > 0 && !eof()) { 1048 // Parse the formal parameter list. 1049 if (FormatTok->is(OpenKind)) { 1050 ++Depth; 1051 } else if (FormatTok->is(CloseKind)) { 1052 --Depth; 1053 } 1054 nextToken(); 1055 } 1056 } 1057 1058 bool UnwrappedLineParser::tryToParseBracedList() { 1059 if (FormatTok->BlockKind == BK_Unknown) 1060 calculateBraceTypes(); 1061 assert(FormatTok->BlockKind != BK_Unknown); 1062 if (FormatTok->BlockKind == BK_Block) 1063 return false; 1064 parseBracedList(); 1065 return true; 1066 } 1067 1068 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 1069 bool HasError = false; 1070 nextToken(); 1071 1072 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 1073 // replace this by using parseAssigmentExpression() inside. 1074 do { 1075 if (Style.Language == FormatStyle::LK_JavaScript) { 1076 if (FormatTok->is(Keywords.kw_function)) { 1077 tryToParseJSFunction(); 1078 continue; 1079 } else if (FormatTok->is(TT_JsFatArrow)) { 1080 nextToken(); 1081 // Fat arrows can be followed by simple expressions or by child blocks 1082 // in curly braces. 1083 if (FormatTok->is(tok::l_brace)){ 1084 parseChildBlock(); 1085 continue; 1086 } 1087 } 1088 } 1089 switch (FormatTok->Tok.getKind()) { 1090 case tok::caret: 1091 nextToken(); 1092 if (FormatTok->is(tok::l_brace)) { 1093 parseChildBlock(); 1094 } 1095 break; 1096 case tok::l_square: 1097 tryToParseLambda(); 1098 break; 1099 case tok::l_brace: 1100 // Assume there are no blocks inside a braced init list apart 1101 // from the ones we explicitly parse out (like lambdas). 1102 FormatTok->BlockKind = BK_BracedInit; 1103 parseBracedList(); 1104 break; 1105 case tok::r_paren: 1106 // JavaScript can just have free standing methods and getters/setters in 1107 // object literals. Detect them by a "{" following ")". 1108 if (Style.Language == FormatStyle::LK_JavaScript) { 1109 nextToken(); 1110 if (FormatTok->is(tok::l_brace)) 1111 parseChildBlock(); 1112 break; 1113 } 1114 nextToken(); 1115 break; 1116 case tok::r_brace: 1117 nextToken(); 1118 return !HasError; 1119 case tok::semi: 1120 HasError = true; 1121 if (!ContinueOnSemicolons) 1122 return !HasError; 1123 nextToken(); 1124 break; 1125 case tok::comma: 1126 nextToken(); 1127 break; 1128 default: 1129 nextToken(); 1130 break; 1131 } 1132 } while (!eof()); 1133 return false; 1134 } 1135 1136 void UnwrappedLineParser::parseParens() { 1137 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1138 nextToken(); 1139 do { 1140 switch (FormatTok->Tok.getKind()) { 1141 case tok::l_paren: 1142 parseParens(); 1143 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) 1144 parseChildBlock(); 1145 break; 1146 case tok::r_paren: 1147 nextToken(); 1148 return; 1149 case tok::r_brace: 1150 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1151 return; 1152 case tok::l_square: 1153 tryToParseLambda(); 1154 break; 1155 case tok::l_brace: 1156 if (!tryToParseBracedList()) 1157 parseChildBlock(); 1158 break; 1159 case tok::at: 1160 nextToken(); 1161 if (FormatTok->Tok.is(tok::l_brace)) 1162 parseBracedList(); 1163 break; 1164 case tok::identifier: 1165 if (Style.Language == FormatStyle::LK_JavaScript && 1166 FormatTok->is(Keywords.kw_function)) 1167 tryToParseJSFunction(); 1168 else 1169 nextToken(); 1170 break; 1171 default: 1172 nextToken(); 1173 break; 1174 } 1175 } while (!eof()); 1176 } 1177 1178 void UnwrappedLineParser::parseSquare() { 1179 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1180 if (tryToParseLambda()) 1181 return; 1182 do { 1183 switch (FormatTok->Tok.getKind()) { 1184 case tok::l_paren: 1185 parseParens(); 1186 break; 1187 case tok::r_square: 1188 nextToken(); 1189 return; 1190 case tok::r_brace: 1191 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1192 return; 1193 case tok::l_square: 1194 parseSquare(); 1195 break; 1196 case tok::l_brace: { 1197 if (!tryToParseBracedList()) 1198 parseChildBlock(); 1199 break; 1200 } 1201 case tok::at: 1202 nextToken(); 1203 if (FormatTok->Tok.is(tok::l_brace)) 1204 parseBracedList(); 1205 break; 1206 default: 1207 nextToken(); 1208 break; 1209 } 1210 } while (!eof()); 1211 } 1212 1213 void UnwrappedLineParser::parseIfThenElse() { 1214 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1215 nextToken(); 1216 if (FormatTok->Tok.is(tok::l_paren)) 1217 parseParens(); 1218 bool NeedsUnwrappedLine = false; 1219 if (FormatTok->Tok.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 addUnwrappedLine(); 1225 } else { 1226 NeedsUnwrappedLine = true; 1227 } 1228 } else { 1229 addUnwrappedLine(); 1230 ++Line->Level; 1231 parseStructuralElement(); 1232 --Line->Level; 1233 } 1234 if (FormatTok->Tok.is(tok::kw_else)) { 1235 if (Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) 1236 addUnwrappedLine(); 1237 nextToken(); 1238 if (FormatTok->Tok.is(tok::l_brace)) { 1239 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1240 parseBlock(/*MustBeDeclaration=*/false); 1241 addUnwrappedLine(); 1242 } else if (FormatTok->Tok.is(tok::kw_if)) { 1243 parseIfThenElse(); 1244 } else { 1245 addUnwrappedLine(); 1246 ++Line->Level; 1247 parseStructuralElement(); 1248 --Line->Level; 1249 } 1250 } else if (NeedsUnwrappedLine) { 1251 addUnwrappedLine(); 1252 } 1253 } 1254 1255 void UnwrappedLineParser::parseTryCatch() { 1256 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected"); 1257 nextToken(); 1258 bool NeedsUnwrappedLine = false; 1259 if (FormatTok->is(tok::colon)) { 1260 // We are in a function try block, what comes is an initializer list. 1261 nextToken(); 1262 while (FormatTok->is(tok::identifier)) { 1263 nextToken(); 1264 if (FormatTok->is(tok::l_paren)) 1265 parseParens(); 1266 if (FormatTok->is(tok::comma)) 1267 nextToken(); 1268 } 1269 } 1270 // Parse try with resource. 1271 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) { 1272 parseParens(); 1273 } 1274 if (FormatTok->is(tok::l_brace)) { 1275 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1276 parseBlock(/*MustBeDeclaration=*/false); 1277 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1278 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1279 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1280 addUnwrappedLine(); 1281 } else { 1282 NeedsUnwrappedLine = true; 1283 } 1284 } else if (!FormatTok->is(tok::kw_catch)) { 1285 // The C++ standard requires a compound-statement after a try. 1286 // If there's none, we try to assume there's a structuralElement 1287 // and try to continue. 1288 addUnwrappedLine(); 1289 ++Line->Level; 1290 parseStructuralElement(); 1291 --Line->Level; 1292 } 1293 while (1) { 1294 if (FormatTok->is(tok::at)) 1295 nextToken(); 1296 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, 1297 tok::kw___finally) || 1298 ((Style.Language == FormatStyle::LK_Java || 1299 Style.Language == FormatStyle::LK_JavaScript) && 1300 FormatTok->is(Keywords.kw_finally)) || 1301 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) || 1302 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally)))) 1303 break; 1304 nextToken(); 1305 while (FormatTok->isNot(tok::l_brace)) { 1306 if (FormatTok->is(tok::l_paren)) { 1307 parseParens(); 1308 continue; 1309 } 1310 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) 1311 return; 1312 nextToken(); 1313 } 1314 NeedsUnwrappedLine = false; 1315 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1316 parseBlock(/*MustBeDeclaration=*/false); 1317 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1318 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1319 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1320 addUnwrappedLine(); 1321 } else { 1322 NeedsUnwrappedLine = true; 1323 } 1324 } 1325 if (NeedsUnwrappedLine) { 1326 addUnwrappedLine(); 1327 } 1328 } 1329 1330 void UnwrappedLineParser::parseNamespace() { 1331 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1332 1333 const FormatToken &InitialToken = *FormatTok; 1334 nextToken(); 1335 if (FormatTok->Tok.is(tok::identifier)) 1336 nextToken(); 1337 if (FormatTok->Tok.is(tok::l_brace)) { 1338 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1339 addUnwrappedLine(); 1340 1341 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1342 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1343 DeclarationScopeStack.size() > 1); 1344 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1345 // Munch the semicolon after a namespace. This is more common than one would 1346 // think. Puttin the semicolon into its own line is very ugly. 1347 if (FormatTok->Tok.is(tok::semi)) 1348 nextToken(); 1349 addUnwrappedLine(); 1350 } 1351 // FIXME: Add error handling. 1352 } 1353 1354 void UnwrappedLineParser::parseNew() { 1355 assert(FormatTok->is(tok::kw_new) && "'new' expected"); 1356 nextToken(); 1357 if (Style.Language != FormatStyle::LK_Java) 1358 return; 1359 1360 // In Java, we can parse everything up to the parens, which aren't optional. 1361 do { 1362 // There should not be a ;, { or } before the new's open paren. 1363 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) 1364 return; 1365 1366 // Consume the parens. 1367 if (FormatTok->is(tok::l_paren)) { 1368 parseParens(); 1369 1370 // If there is a class body of an anonymous class, consume that as child. 1371 if (FormatTok->is(tok::l_brace)) 1372 parseChildBlock(); 1373 return; 1374 } 1375 nextToken(); 1376 } while (!eof()); 1377 } 1378 1379 void UnwrappedLineParser::parseForOrWhileLoop() { 1380 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) && 1381 "'for', 'while' or foreach macro expected"); 1382 nextToken(); 1383 if (FormatTok->Tok.is(tok::l_paren)) 1384 parseParens(); 1385 if (FormatTok->Tok.is(tok::l_brace)) { 1386 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1387 parseBlock(/*MustBeDeclaration=*/false); 1388 addUnwrappedLine(); 1389 } else { 1390 addUnwrappedLine(); 1391 ++Line->Level; 1392 parseStructuralElement(); 1393 --Line->Level; 1394 } 1395 } 1396 1397 void UnwrappedLineParser::parseDoWhile() { 1398 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1399 nextToken(); 1400 if (FormatTok->Tok.is(tok::l_brace)) { 1401 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1402 parseBlock(/*MustBeDeclaration=*/false); 1403 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1404 addUnwrappedLine(); 1405 } else { 1406 addUnwrappedLine(); 1407 ++Line->Level; 1408 parseStructuralElement(); 1409 --Line->Level; 1410 } 1411 1412 // FIXME: Add error handling. 1413 if (!FormatTok->Tok.is(tok::kw_while)) { 1414 addUnwrappedLine(); 1415 return; 1416 } 1417 1418 nextToken(); 1419 parseStructuralElement(); 1420 } 1421 1422 void UnwrappedLineParser::parseLabel() { 1423 nextToken(); 1424 unsigned OldLineLevel = Line->Level; 1425 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1426 --Line->Level; 1427 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1428 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1429 parseBlock(/*MustBeDeclaration=*/false); 1430 if (FormatTok->Tok.is(tok::kw_break)) { 1431 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1432 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1433 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1434 addUnwrappedLine(); 1435 } 1436 parseStructuralElement(); 1437 } 1438 addUnwrappedLine(); 1439 } else { 1440 if (FormatTok->is(tok::semi)) 1441 nextToken(); 1442 addUnwrappedLine(); 1443 } 1444 Line->Level = OldLineLevel; 1445 } 1446 1447 void UnwrappedLineParser::parseCaseLabel() { 1448 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1449 // FIXME: fix handling of complex expressions here. 1450 do { 1451 nextToken(); 1452 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1453 parseLabel(); 1454 } 1455 1456 void UnwrappedLineParser::parseSwitch() { 1457 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1458 nextToken(); 1459 if (FormatTok->Tok.is(tok::l_paren)) 1460 parseParens(); 1461 if (FormatTok->Tok.is(tok::l_brace)) { 1462 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1463 parseBlock(/*MustBeDeclaration=*/false); 1464 addUnwrappedLine(); 1465 } else { 1466 addUnwrappedLine(); 1467 ++Line->Level; 1468 parseStructuralElement(); 1469 --Line->Level; 1470 } 1471 } 1472 1473 void UnwrappedLineParser::parseAccessSpecifier() { 1474 nextToken(); 1475 // Understand Qt's slots. 1476 if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots)) 1477 nextToken(); 1478 // Otherwise, we don't know what it is, and we'd better keep the next token. 1479 if (FormatTok->Tok.is(tok::colon)) 1480 nextToken(); 1481 addUnwrappedLine(); 1482 } 1483 1484 void UnwrappedLineParser::parseEnum() { 1485 // Won't be 'enum' for NS_ENUMs. 1486 if (FormatTok->Tok.is(tok::kw_enum)) 1487 nextToken(); 1488 1489 // Eat up enum class ... 1490 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1491 nextToken(); 1492 while (FormatTok->Tok.getIdentifierInfo() || 1493 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less, 1494 tok::greater, tok::comma, tok::question)) { 1495 nextToken(); 1496 // We can have macros or attributes in between 'enum' and the enum name. 1497 if (FormatTok->is(tok::l_paren)) 1498 parseParens(); 1499 if (FormatTok->is(tok::identifier)) 1500 nextToken(); 1501 } 1502 1503 // Just a declaration or something is wrong. 1504 if (FormatTok->isNot(tok::l_brace)) 1505 return; 1506 FormatTok->BlockKind = BK_Block; 1507 1508 if (Style.Language == FormatStyle::LK_Java) { 1509 // Java enums are different. 1510 parseJavaEnumBody(); 1511 return; 1512 } 1513 1514 // Parse enum body. 1515 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1516 if (HasError) { 1517 if (FormatTok->is(tok::semi)) 1518 nextToken(); 1519 addUnwrappedLine(); 1520 } 1521 1522 // We fall through to parsing a structural element afterwards, so that in 1523 // enum A {} n, m; 1524 // "} n, m;" will end up in one unwrapped line. 1525 } 1526 1527 void UnwrappedLineParser::parseJavaEnumBody() { 1528 // Determine whether the enum is simple, i.e. does not have a semicolon or 1529 // constants with class bodies. Simple enums can be formatted like braced 1530 // lists, contracted to a single line, etc. 1531 unsigned StoredPosition = Tokens->getPosition(); 1532 bool IsSimple = true; 1533 FormatToken *Tok = Tokens->getNextToken(); 1534 while (Tok) { 1535 if (Tok->is(tok::r_brace)) 1536 break; 1537 if (Tok->isOneOf(tok::l_brace, tok::semi)) { 1538 IsSimple = false; 1539 break; 1540 } 1541 // FIXME: This will also mark enums with braces in the arguments to enum 1542 // constants as "not simple". This is probably fine in practice, though. 1543 Tok = Tokens->getNextToken(); 1544 } 1545 FormatTok = Tokens->setPosition(StoredPosition); 1546 1547 if (IsSimple) { 1548 parseBracedList(); 1549 addUnwrappedLine(); 1550 return; 1551 } 1552 1553 // Parse the body of a more complex enum. 1554 // First add a line for everything up to the "{". 1555 nextToken(); 1556 addUnwrappedLine(); 1557 ++Line->Level; 1558 1559 // Parse the enum constants. 1560 while (FormatTok) { 1561 if (FormatTok->is(tok::l_brace)) { 1562 // Parse the constant's class body. 1563 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1564 /*MunchSemi=*/false); 1565 } else if (FormatTok->is(tok::l_paren)) { 1566 parseParens(); 1567 } else if (FormatTok->is(tok::comma)) { 1568 nextToken(); 1569 addUnwrappedLine(); 1570 } else if (FormatTok->is(tok::semi)) { 1571 nextToken(); 1572 addUnwrappedLine(); 1573 break; 1574 } else if (FormatTok->is(tok::r_brace)) { 1575 addUnwrappedLine(); 1576 break; 1577 } else { 1578 nextToken(); 1579 } 1580 } 1581 1582 // Parse the class body after the enum's ";" if any. 1583 parseLevel(/*HasOpeningBrace=*/true); 1584 nextToken(); 1585 --Line->Level; 1586 addUnwrappedLine(); 1587 } 1588 1589 void UnwrappedLineParser::parseRecord() { 1590 const FormatToken &InitialToken = *FormatTok; 1591 nextToken(); 1592 1593 1594 // The actual identifier can be a nested name specifier, and in macros 1595 // it is often token-pasted. 1596 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash, 1597 tok::kw___attribute, tok::kw___declspec, 1598 tok::kw_alignas) || 1599 ((Style.Language == FormatStyle::LK_Java || 1600 Style.Language == FormatStyle::LK_JavaScript) && 1601 FormatTok->isOneOf(tok::period, tok::comma))) { 1602 bool IsNonMacroIdentifier = 1603 FormatTok->is(tok::identifier) && 1604 FormatTok->TokenText != FormatTok->TokenText.upper(); 1605 nextToken(); 1606 // We can have macros or attributes in between 'class' and the class name. 1607 if (!IsNonMacroIdentifier && FormatTok->Tok.is(tok::l_paren)) 1608 parseParens(); 1609 } 1610 1611 // Note that parsing away template declarations here leads to incorrectly 1612 // accepting function declarations as record declarations. 1613 // In general, we cannot solve this problem. Consider: 1614 // class A<int> B() {} 1615 // which can be a function definition or a class definition when B() is a 1616 // macro. If we find enough real-world cases where this is a problem, we 1617 // can parse for the 'template' keyword in the beginning of the statement, 1618 // and thus rule out the record production in case there is no template 1619 // (this would still leave us with an ambiguity between template function 1620 // and class declarations). 1621 if (FormatTok->isOneOf(tok::colon, tok::less)) { 1622 while (!eof()) { 1623 if (FormatTok->is(tok::l_brace)) { 1624 calculateBraceTypes(/*ExpectClassBody=*/true); 1625 if (!tryToParseBracedList()) 1626 break; 1627 } 1628 if (FormatTok->Tok.is(tok::semi)) 1629 return; 1630 nextToken(); 1631 } 1632 } 1633 if (FormatTok->Tok.is(tok::l_brace)) { 1634 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1635 addUnwrappedLine(); 1636 1637 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1638 /*MunchSemi=*/false); 1639 } 1640 // We fall through to parsing a structural element afterwards, so 1641 // class A {} n, m; 1642 // will end up in one unwrapped line. 1643 // This does not apply for Java and JavaScript. 1644 if (Style.Language == FormatStyle::LK_Java || 1645 Style.Language == FormatStyle::LK_JavaScript) 1646 addUnwrappedLine(); 1647 } 1648 1649 void UnwrappedLineParser::parseObjCProtocolList() { 1650 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1651 do 1652 nextToken(); 1653 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1654 nextToken(); // Skip '>'. 1655 } 1656 1657 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1658 do { 1659 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1660 nextToken(); 1661 addUnwrappedLine(); 1662 break; 1663 } 1664 if (FormatTok->is(tok::l_brace)) { 1665 parseBlock(/*MustBeDeclaration=*/false); 1666 // In ObjC interfaces, nothing should be following the "}". 1667 addUnwrappedLine(); 1668 } else if (FormatTok->is(tok::r_brace)) { 1669 // Ignore stray "}". parseStructuralElement doesn't consume them. 1670 nextToken(); 1671 addUnwrappedLine(); 1672 } else { 1673 parseStructuralElement(); 1674 } 1675 } while (!eof()); 1676 } 1677 1678 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1679 nextToken(); 1680 nextToken(); // interface name 1681 1682 // @interface can be followed by either a base class, or a category. 1683 if (FormatTok->Tok.is(tok::colon)) { 1684 nextToken(); 1685 nextToken(); // base class name 1686 } else if (FormatTok->Tok.is(tok::l_paren)) 1687 // Skip category, if present. 1688 parseParens(); 1689 1690 if (FormatTok->Tok.is(tok::less)) 1691 parseObjCProtocolList(); 1692 1693 if (FormatTok->Tok.is(tok::l_brace)) { 1694 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1695 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1696 addUnwrappedLine(); 1697 parseBlock(/*MustBeDeclaration=*/true); 1698 } 1699 1700 // With instance variables, this puts '}' on its own line. Without instance 1701 // variables, this ends the @interface line. 1702 addUnwrappedLine(); 1703 1704 parseObjCUntilAtEnd(); 1705 } 1706 1707 void UnwrappedLineParser::parseObjCProtocol() { 1708 nextToken(); 1709 nextToken(); // protocol name 1710 1711 if (FormatTok->Tok.is(tok::less)) 1712 parseObjCProtocolList(); 1713 1714 // Check for protocol declaration. 1715 if (FormatTok->Tok.is(tok::semi)) { 1716 nextToken(); 1717 return addUnwrappedLine(); 1718 } 1719 1720 addUnwrappedLine(); 1721 parseObjCUntilAtEnd(); 1722 } 1723 1724 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { 1725 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export)); 1726 nextToken(); 1727 1728 // Consume the "default" in "export default class/function". 1729 if (FormatTok->is(tok::kw_default)) 1730 nextToken(); 1731 1732 // Consume "function" and "default function", so that these get parsed as 1733 // free-standing JS functions, i.e. do not require a trailing semicolon. 1734 if (FormatTok->is(Keywords.kw_function)) { 1735 nextToken(); 1736 return; 1737 } 1738 1739 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, Keywords.kw_var)) 1740 return; // Fall through to parsing the corresponding structure. 1741 1742 if (FormatTok->is(tok::l_brace)) { 1743 FormatTok->BlockKind = BK_Block; 1744 parseBracedList(); 1745 } 1746 1747 while (!eof() && FormatTok->isNot(tok::semi) && 1748 FormatTok->isNot(tok::l_brace)) { 1749 nextToken(); 1750 } 1751 } 1752 1753 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1754 StringRef Prefix = "") { 1755 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1756 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1757 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1758 E = Line.Tokens.end(); 1759 I != E; ++I) { 1760 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1761 } 1762 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1763 E = Line.Tokens.end(); 1764 I != E; ++I) { 1765 const UnwrappedLineNode &Node = *I; 1766 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1767 I = Node.Children.begin(), 1768 E = Node.Children.end(); 1769 I != E; ++I) { 1770 printDebugInfo(*I, "\nChild: "); 1771 } 1772 } 1773 llvm::dbgs() << "\n"; 1774 } 1775 1776 void UnwrappedLineParser::addUnwrappedLine() { 1777 if (Line->Tokens.empty()) 1778 return; 1779 DEBUG({ 1780 if (CurrentLines == &Lines) 1781 printDebugInfo(*Line); 1782 }); 1783 CurrentLines->push_back(*Line); 1784 Line->Tokens.clear(); 1785 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1786 for (SmallVectorImpl<UnwrappedLine>::iterator 1787 I = PreprocessorDirectives.begin(), 1788 E = PreprocessorDirectives.end(); 1789 I != E; ++I) { 1790 CurrentLines->push_back(*I); 1791 } 1792 PreprocessorDirectives.clear(); 1793 } 1794 } 1795 1796 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1797 1798 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1799 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1800 FormatTok.NewlinesBefore > 0; 1801 } 1802 1803 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1804 bool JustComments = Line->Tokens.empty(); 1805 for (SmallVectorImpl<FormatToken *>::const_iterator 1806 I = CommentsBeforeNextToken.begin(), 1807 E = CommentsBeforeNextToken.end(); 1808 I != E; ++I) { 1809 if (isOnNewLine(**I) && JustComments) 1810 addUnwrappedLine(); 1811 pushToken(*I); 1812 } 1813 if (NewlineBeforeNext && JustComments) 1814 addUnwrappedLine(); 1815 CommentsBeforeNextToken.clear(); 1816 } 1817 1818 void UnwrappedLineParser::nextToken() { 1819 if (eof()) 1820 return; 1821 flushComments(isOnNewLine(*FormatTok)); 1822 pushToken(FormatTok); 1823 readToken(); 1824 } 1825 1826 void UnwrappedLineParser::readToken() { 1827 bool CommentsInCurrentLine = true; 1828 do { 1829 FormatTok = Tokens->getNextToken(); 1830 assert(FormatTok); 1831 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1832 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1833 // If there is an unfinished unwrapped line, we flush the preprocessor 1834 // directives only after that unwrapped line was finished later. 1835 bool SwitchToPreprocessorLines = !Line->Tokens.empty(); 1836 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1837 // Comments stored before the preprocessor directive need to be output 1838 // before the preprocessor directive, at the same level as the 1839 // preprocessor directive, as we consider them to apply to the directive. 1840 flushComments(isOnNewLine(*FormatTok)); 1841 parsePPDirective(); 1842 } 1843 while (FormatTok->Type == TT_ConflictStart || 1844 FormatTok->Type == TT_ConflictEnd || 1845 FormatTok->Type == TT_ConflictAlternative) { 1846 if (FormatTok->Type == TT_ConflictStart) { 1847 conditionalCompilationStart(/*Unreachable=*/false); 1848 } else if (FormatTok->Type == TT_ConflictAlternative) { 1849 conditionalCompilationAlternative(); 1850 } else if (FormatTok->Type == TT_ConflictEnd) { 1851 conditionalCompilationEnd(); 1852 } 1853 FormatTok = Tokens->getNextToken(); 1854 FormatTok->MustBreakBefore = true; 1855 } 1856 1857 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1858 !Line->InPPDirective) { 1859 continue; 1860 } 1861 1862 if (!FormatTok->Tok.is(tok::comment)) 1863 return; 1864 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1865 CommentsInCurrentLine = false; 1866 } 1867 if (CommentsInCurrentLine) { 1868 pushToken(FormatTok); 1869 } else { 1870 CommentsBeforeNextToken.push_back(FormatTok); 1871 } 1872 } while (!eof()); 1873 } 1874 1875 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1876 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1877 if (MustBreakBeforeNextToken) { 1878 Line->Tokens.back().Tok->MustBreakBefore = true; 1879 MustBreakBeforeNextToken = false; 1880 } 1881 } 1882 1883 } // end namespace format 1884 } // end namespace clang 1885