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 break; 893 } 894 895 nextToken(); 896 if (FormatTok->Tok.is(tok::l_brace)) { 897 parseBracedList(); 898 } 899 break; 900 case tok::l_square: 901 parseSquare(); 902 break; 903 case tok::kw_new: 904 parseNew(); 905 break; 906 default: 907 nextToken(); 908 break; 909 } 910 } while (!eof()); 911 } 912 913 bool UnwrappedLineParser::tryToParseLambda() { 914 if (Style.Language != FormatStyle::LK_Cpp) { 915 nextToken(); 916 return false; 917 } 918 // FIXME: This is a dirty way to access the previous token. Find a better 919 // solution. 920 if (!Line->Tokens.empty() && 921 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator, 922 tok::kw_new, tok::kw_delete) || 923 Line->Tokens.back().Tok->closesScope() || 924 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 925 nextToken(); 926 return false; 927 } 928 assert(FormatTok->is(tok::l_square)); 929 FormatToken &LSquare = *FormatTok; 930 if (!tryToParseLambdaIntroducer()) 931 return false; 932 933 while (FormatTok->isNot(tok::l_brace)) { 934 if (FormatTok->isSimpleTypeSpecifier()) { 935 nextToken(); 936 continue; 937 } 938 switch (FormatTok->Tok.getKind()) { 939 case tok::l_brace: 940 break; 941 case tok::l_paren: 942 parseParens(); 943 break; 944 case tok::amp: 945 case tok::star: 946 case tok::kw_const: 947 case tok::comma: 948 case tok::less: 949 case tok::greater: 950 case tok::identifier: 951 case tok::coloncolon: 952 case tok::kw_mutable: 953 nextToken(); 954 break; 955 case tok::arrow: 956 FormatTok->Type = TT_TrailingReturnArrow; 957 nextToken(); 958 break; 959 default: 960 return true; 961 } 962 } 963 LSquare.Type = TT_LambdaLSquare; 964 parseChildBlock(); 965 return true; 966 } 967 968 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 969 nextToken(); 970 if (FormatTok->is(tok::equal)) { 971 nextToken(); 972 if (FormatTok->is(tok::r_square)) { 973 nextToken(); 974 return true; 975 } 976 if (FormatTok->isNot(tok::comma)) 977 return false; 978 nextToken(); 979 } else if (FormatTok->is(tok::amp)) { 980 nextToken(); 981 if (FormatTok->is(tok::r_square)) { 982 nextToken(); 983 return true; 984 } 985 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 986 return false; 987 } 988 if (FormatTok->is(tok::comma)) 989 nextToken(); 990 } else if (FormatTok->is(tok::r_square)) { 991 nextToken(); 992 return true; 993 } 994 do { 995 if (FormatTok->is(tok::amp)) 996 nextToken(); 997 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 998 return false; 999 nextToken(); 1000 if (FormatTok->is(tok::ellipsis)) 1001 nextToken(); 1002 if (FormatTok->is(tok::comma)) { 1003 nextToken(); 1004 } else if (FormatTok->is(tok::r_square)) { 1005 nextToken(); 1006 return true; 1007 } else { 1008 return false; 1009 } 1010 } while (!eof()); 1011 return false; 1012 } 1013 1014 void UnwrappedLineParser::tryToParseJSFunction() { 1015 nextToken(); 1016 1017 // Consume function name. 1018 if (FormatTok->is(tok::identifier)) 1019 nextToken(); 1020 1021 if (FormatTok->isNot(tok::l_paren)) 1022 return; 1023 1024 // Parse formal parameter list. 1025 parseParens(); 1026 1027 if (FormatTok->is(tok::colon)) { 1028 // Parse a type definition. 1029 nextToken(); 1030 1031 // Eat the type declaration. For braced inline object types, balance braces, 1032 // otherwise just parse until finding an l_brace for the function body. 1033 if (FormatTok->is(tok::l_brace)) 1034 tryToParseBracedList(); 1035 else 1036 while(FormatTok->isNot(tok::l_brace) && !eof()) 1037 nextToken(); 1038 } 1039 1040 parseChildBlock(); 1041 } 1042 1043 bool UnwrappedLineParser::tryToParseBracedList() { 1044 if (FormatTok->BlockKind == BK_Unknown) 1045 calculateBraceTypes(); 1046 assert(FormatTok->BlockKind != BK_Unknown); 1047 if (FormatTok->BlockKind == BK_Block) 1048 return false; 1049 parseBracedList(); 1050 return true; 1051 } 1052 1053 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 1054 bool HasError = false; 1055 nextToken(); 1056 1057 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 1058 // replace this by using parseAssigmentExpression() inside. 1059 do { 1060 if (Style.Language == FormatStyle::LK_JavaScript) { 1061 if (FormatTok->is(Keywords.kw_function)) { 1062 tryToParseJSFunction(); 1063 continue; 1064 } 1065 if (FormatTok->is(TT_JsFatArrow)) { 1066 nextToken(); 1067 // Fat arrows can be followed by simple expressions or by child blocks 1068 // in curly braces. 1069 if (FormatTok->is(tok::l_brace)){ 1070 parseChildBlock(); 1071 continue; 1072 } 1073 } 1074 } 1075 switch (FormatTok->Tok.getKind()) { 1076 case tok::caret: 1077 nextToken(); 1078 if (FormatTok->is(tok::l_brace)) { 1079 parseChildBlock(); 1080 } 1081 break; 1082 case tok::l_square: 1083 tryToParseLambda(); 1084 break; 1085 case tok::l_brace: 1086 // Assume there are no blocks inside a braced init list apart 1087 // from the ones we explicitly parse out (like lambdas). 1088 FormatTok->BlockKind = BK_BracedInit; 1089 parseBracedList(); 1090 break; 1091 case tok::r_paren: 1092 // JavaScript can just have free standing methods and getters/setters in 1093 // object literals. Detect them by a "{" following ")". 1094 if (Style.Language == FormatStyle::LK_JavaScript) { 1095 nextToken(); 1096 if (FormatTok->is(tok::l_brace)) 1097 parseChildBlock(); 1098 break; 1099 } 1100 nextToken(); 1101 break; 1102 case tok::r_brace: 1103 nextToken(); 1104 return !HasError; 1105 case tok::semi: 1106 HasError = true; 1107 if (!ContinueOnSemicolons) 1108 return !HasError; 1109 nextToken(); 1110 break; 1111 case tok::comma: 1112 nextToken(); 1113 break; 1114 default: 1115 nextToken(); 1116 break; 1117 } 1118 } while (!eof()); 1119 return false; 1120 } 1121 1122 void UnwrappedLineParser::parseParens() { 1123 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1124 nextToken(); 1125 do { 1126 switch (FormatTok->Tok.getKind()) { 1127 case tok::l_paren: 1128 parseParens(); 1129 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) 1130 parseChildBlock(); 1131 break; 1132 case tok::r_paren: 1133 nextToken(); 1134 return; 1135 case tok::r_brace: 1136 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1137 return; 1138 case tok::l_square: 1139 tryToParseLambda(); 1140 break; 1141 case tok::l_brace: 1142 if (!tryToParseBracedList()) 1143 parseChildBlock(); 1144 break; 1145 case tok::at: 1146 nextToken(); 1147 if (FormatTok->Tok.is(tok::l_brace)) 1148 parseBracedList(); 1149 break; 1150 case tok::identifier: 1151 if (Style.Language == FormatStyle::LK_JavaScript && 1152 FormatTok->is(Keywords.kw_function)) 1153 tryToParseJSFunction(); 1154 else 1155 nextToken(); 1156 break; 1157 default: 1158 nextToken(); 1159 break; 1160 } 1161 } while (!eof()); 1162 } 1163 1164 void UnwrappedLineParser::parseSquare() { 1165 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1166 if (tryToParseLambda()) 1167 return; 1168 do { 1169 switch (FormatTok->Tok.getKind()) { 1170 case tok::l_paren: 1171 parseParens(); 1172 break; 1173 case tok::r_square: 1174 nextToken(); 1175 return; 1176 case tok::r_brace: 1177 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1178 return; 1179 case tok::l_square: 1180 parseSquare(); 1181 break; 1182 case tok::l_brace: { 1183 if (!tryToParseBracedList()) 1184 parseChildBlock(); 1185 break; 1186 } 1187 case tok::at: 1188 nextToken(); 1189 if (FormatTok->Tok.is(tok::l_brace)) 1190 parseBracedList(); 1191 break; 1192 default: 1193 nextToken(); 1194 break; 1195 } 1196 } while (!eof()); 1197 } 1198 1199 void UnwrappedLineParser::parseIfThenElse() { 1200 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1201 nextToken(); 1202 if (FormatTok->Tok.is(tok::l_paren)) 1203 parseParens(); 1204 bool NeedsUnwrappedLine = false; 1205 if (FormatTok->Tok.is(tok::l_brace)) { 1206 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1207 parseBlock(/*MustBeDeclaration=*/false); 1208 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1209 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1210 addUnwrappedLine(); 1211 } else { 1212 NeedsUnwrappedLine = true; 1213 } 1214 } else { 1215 addUnwrappedLine(); 1216 ++Line->Level; 1217 parseStructuralElement(); 1218 --Line->Level; 1219 } 1220 if (FormatTok->Tok.is(tok::kw_else)) { 1221 if (Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) 1222 addUnwrappedLine(); 1223 nextToken(); 1224 if (FormatTok->Tok.is(tok::l_brace)) { 1225 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1226 parseBlock(/*MustBeDeclaration=*/false); 1227 addUnwrappedLine(); 1228 } else if (FormatTok->Tok.is(tok::kw_if)) { 1229 parseIfThenElse(); 1230 } else { 1231 addUnwrappedLine(); 1232 ++Line->Level; 1233 parseStructuralElement(); 1234 --Line->Level; 1235 } 1236 } else if (NeedsUnwrappedLine) { 1237 addUnwrappedLine(); 1238 } 1239 } 1240 1241 void UnwrappedLineParser::parseTryCatch() { 1242 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected"); 1243 nextToken(); 1244 bool NeedsUnwrappedLine = false; 1245 if (FormatTok->is(tok::colon)) { 1246 // We are in a function try block, what comes is an initializer list. 1247 nextToken(); 1248 while (FormatTok->is(tok::identifier)) { 1249 nextToken(); 1250 if (FormatTok->is(tok::l_paren)) 1251 parseParens(); 1252 if (FormatTok->is(tok::comma)) 1253 nextToken(); 1254 } 1255 } 1256 // Parse try with resource. 1257 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) { 1258 parseParens(); 1259 } 1260 if (FormatTok->is(tok::l_brace)) { 1261 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1262 parseBlock(/*MustBeDeclaration=*/false); 1263 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1264 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1265 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1266 addUnwrappedLine(); 1267 } else { 1268 NeedsUnwrappedLine = true; 1269 } 1270 } else if (!FormatTok->is(tok::kw_catch)) { 1271 // The C++ standard requires a compound-statement after a try. 1272 // If there's none, we try to assume there's a structuralElement 1273 // and try to continue. 1274 addUnwrappedLine(); 1275 ++Line->Level; 1276 parseStructuralElement(); 1277 --Line->Level; 1278 } 1279 while (1) { 1280 if (FormatTok->is(tok::at)) 1281 nextToken(); 1282 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, 1283 tok::kw___finally) || 1284 ((Style.Language == FormatStyle::LK_Java || 1285 Style.Language == FormatStyle::LK_JavaScript) && 1286 FormatTok->is(Keywords.kw_finally)) || 1287 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) || 1288 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally)))) 1289 break; 1290 nextToken(); 1291 while (FormatTok->isNot(tok::l_brace)) { 1292 if (FormatTok->is(tok::l_paren)) { 1293 parseParens(); 1294 continue; 1295 } 1296 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) 1297 return; 1298 nextToken(); 1299 } 1300 NeedsUnwrappedLine = false; 1301 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1302 parseBlock(/*MustBeDeclaration=*/false); 1303 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1304 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1305 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1306 addUnwrappedLine(); 1307 } else { 1308 NeedsUnwrappedLine = true; 1309 } 1310 } 1311 if (NeedsUnwrappedLine) { 1312 addUnwrappedLine(); 1313 } 1314 } 1315 1316 void UnwrappedLineParser::parseNamespace() { 1317 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1318 1319 const FormatToken &InitialToken = *FormatTok; 1320 nextToken(); 1321 if (FormatTok->Tok.is(tok::identifier)) 1322 nextToken(); 1323 if (FormatTok->Tok.is(tok::l_brace)) { 1324 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1325 addUnwrappedLine(); 1326 1327 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1328 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1329 DeclarationScopeStack.size() > 1); 1330 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1331 // Munch the semicolon after a namespace. This is more common than one would 1332 // think. Puttin the semicolon into its own line is very ugly. 1333 if (FormatTok->Tok.is(tok::semi)) 1334 nextToken(); 1335 addUnwrappedLine(); 1336 } 1337 // FIXME: Add error handling. 1338 } 1339 1340 void UnwrappedLineParser::parseNew() { 1341 assert(FormatTok->is(tok::kw_new) && "'new' expected"); 1342 nextToken(); 1343 if (Style.Language != FormatStyle::LK_Java) 1344 return; 1345 1346 // In Java, we can parse everything up to the parens, which aren't optional. 1347 do { 1348 // There should not be a ;, { or } before the new's open paren. 1349 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) 1350 return; 1351 1352 // Consume the parens. 1353 if (FormatTok->is(tok::l_paren)) { 1354 parseParens(); 1355 1356 // If there is a class body of an anonymous class, consume that as child. 1357 if (FormatTok->is(tok::l_brace)) 1358 parseChildBlock(); 1359 return; 1360 } 1361 nextToken(); 1362 } while (!eof()); 1363 } 1364 1365 void UnwrappedLineParser::parseForOrWhileLoop() { 1366 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) && 1367 "'for', 'while' or foreach macro expected"); 1368 nextToken(); 1369 if (FormatTok->Tok.is(tok::l_paren)) 1370 parseParens(); 1371 if (FormatTok->Tok.is(tok::l_brace)) { 1372 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1373 parseBlock(/*MustBeDeclaration=*/false); 1374 addUnwrappedLine(); 1375 } else { 1376 addUnwrappedLine(); 1377 ++Line->Level; 1378 parseStructuralElement(); 1379 --Line->Level; 1380 } 1381 } 1382 1383 void UnwrappedLineParser::parseDoWhile() { 1384 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1385 nextToken(); 1386 if (FormatTok->Tok.is(tok::l_brace)) { 1387 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1388 parseBlock(/*MustBeDeclaration=*/false); 1389 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1390 addUnwrappedLine(); 1391 } else { 1392 addUnwrappedLine(); 1393 ++Line->Level; 1394 parseStructuralElement(); 1395 --Line->Level; 1396 } 1397 1398 // FIXME: Add error handling. 1399 if (!FormatTok->Tok.is(tok::kw_while)) { 1400 addUnwrappedLine(); 1401 return; 1402 } 1403 1404 nextToken(); 1405 parseStructuralElement(); 1406 } 1407 1408 void UnwrappedLineParser::parseLabel() { 1409 nextToken(); 1410 unsigned OldLineLevel = Line->Level; 1411 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1412 --Line->Level; 1413 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1414 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1415 parseBlock(/*MustBeDeclaration=*/false); 1416 if (FormatTok->Tok.is(tok::kw_break)) { 1417 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1418 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1419 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1420 addUnwrappedLine(); 1421 } 1422 parseStructuralElement(); 1423 } 1424 addUnwrappedLine(); 1425 } else { 1426 if (FormatTok->is(tok::semi)) 1427 nextToken(); 1428 addUnwrappedLine(); 1429 } 1430 Line->Level = OldLineLevel; 1431 } 1432 1433 void UnwrappedLineParser::parseCaseLabel() { 1434 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1435 // FIXME: fix handling of complex expressions here. 1436 do { 1437 nextToken(); 1438 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1439 parseLabel(); 1440 } 1441 1442 void UnwrappedLineParser::parseSwitch() { 1443 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1444 nextToken(); 1445 if (FormatTok->Tok.is(tok::l_paren)) 1446 parseParens(); 1447 if (FormatTok->Tok.is(tok::l_brace)) { 1448 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1449 parseBlock(/*MustBeDeclaration=*/false); 1450 addUnwrappedLine(); 1451 } else { 1452 addUnwrappedLine(); 1453 ++Line->Level; 1454 parseStructuralElement(); 1455 --Line->Level; 1456 } 1457 } 1458 1459 void UnwrappedLineParser::parseAccessSpecifier() { 1460 nextToken(); 1461 // Understand Qt's slots. 1462 if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots)) 1463 nextToken(); 1464 // Otherwise, we don't know what it is, and we'd better keep the next token. 1465 if (FormatTok->Tok.is(tok::colon)) 1466 nextToken(); 1467 addUnwrappedLine(); 1468 } 1469 1470 void UnwrappedLineParser::parseEnum() { 1471 // Won't be 'enum' for NS_ENUMs. 1472 if (FormatTok->Tok.is(tok::kw_enum)) 1473 nextToken(); 1474 1475 // Eat up enum class ... 1476 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1477 nextToken(); 1478 while (FormatTok->Tok.getIdentifierInfo() || 1479 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less, 1480 tok::greater, tok::comma, tok::question)) { 1481 nextToken(); 1482 // We can have macros or attributes in between 'enum' and the enum name. 1483 if (FormatTok->is(tok::l_paren)) 1484 parseParens(); 1485 if (FormatTok->is(tok::identifier)) 1486 nextToken(); 1487 } 1488 1489 // Just a declaration or something is wrong. 1490 if (FormatTok->isNot(tok::l_brace)) 1491 return; 1492 FormatTok->BlockKind = BK_Block; 1493 1494 if (Style.Language == FormatStyle::LK_Java) { 1495 // Java enums are different. 1496 parseJavaEnumBody(); 1497 return; 1498 } 1499 1500 // Parse enum body. 1501 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1502 if (HasError) { 1503 if (FormatTok->is(tok::semi)) 1504 nextToken(); 1505 addUnwrappedLine(); 1506 } 1507 1508 // We fall through to parsing a structural element afterwards, so that in 1509 // enum A {} n, m; 1510 // "} n, m;" will end up in one unwrapped line. 1511 } 1512 1513 void UnwrappedLineParser::parseJavaEnumBody() { 1514 // Determine whether the enum is simple, i.e. does not have a semicolon or 1515 // constants with class bodies. Simple enums can be formatted like braced 1516 // lists, contracted to a single line, etc. 1517 unsigned StoredPosition = Tokens->getPosition(); 1518 bool IsSimple = true; 1519 FormatToken *Tok = Tokens->getNextToken(); 1520 while (Tok) { 1521 if (Tok->is(tok::r_brace)) 1522 break; 1523 if (Tok->isOneOf(tok::l_brace, tok::semi)) { 1524 IsSimple = false; 1525 break; 1526 } 1527 // FIXME: This will also mark enums with braces in the arguments to enum 1528 // constants as "not simple". This is probably fine in practice, though. 1529 Tok = Tokens->getNextToken(); 1530 } 1531 FormatTok = Tokens->setPosition(StoredPosition); 1532 1533 if (IsSimple) { 1534 parseBracedList(); 1535 addUnwrappedLine(); 1536 return; 1537 } 1538 1539 // Parse the body of a more complex enum. 1540 // First add a line for everything up to the "{". 1541 nextToken(); 1542 addUnwrappedLine(); 1543 ++Line->Level; 1544 1545 // Parse the enum constants. 1546 while (FormatTok) { 1547 if (FormatTok->is(tok::l_brace)) { 1548 // Parse the constant's class body. 1549 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1550 /*MunchSemi=*/false); 1551 } else if (FormatTok->is(tok::l_paren)) { 1552 parseParens(); 1553 } else if (FormatTok->is(tok::comma)) { 1554 nextToken(); 1555 addUnwrappedLine(); 1556 } else if (FormatTok->is(tok::semi)) { 1557 nextToken(); 1558 addUnwrappedLine(); 1559 break; 1560 } else if (FormatTok->is(tok::r_brace)) { 1561 addUnwrappedLine(); 1562 break; 1563 } else { 1564 nextToken(); 1565 } 1566 } 1567 1568 // Parse the class body after the enum's ";" if any. 1569 parseLevel(/*HasOpeningBrace=*/true); 1570 nextToken(); 1571 --Line->Level; 1572 addUnwrappedLine(); 1573 } 1574 1575 void UnwrappedLineParser::parseRecord() { 1576 const FormatToken &InitialToken = *FormatTok; 1577 nextToken(); 1578 1579 1580 // The actual identifier can be a nested name specifier, and in macros 1581 // it is often token-pasted. 1582 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash, 1583 tok::kw___attribute, tok::kw___declspec, 1584 tok::kw_alignas) || 1585 ((Style.Language == FormatStyle::LK_Java || 1586 Style.Language == FormatStyle::LK_JavaScript) && 1587 FormatTok->isOneOf(tok::period, tok::comma))) { 1588 bool IsNonMacroIdentifier = 1589 FormatTok->is(tok::identifier) && 1590 FormatTok->TokenText != FormatTok->TokenText.upper(); 1591 nextToken(); 1592 // We can have macros or attributes in between 'class' and the class name. 1593 if (!IsNonMacroIdentifier && FormatTok->Tok.is(tok::l_paren)) 1594 parseParens(); 1595 } 1596 1597 // Note that parsing away template declarations here leads to incorrectly 1598 // accepting function declarations as record declarations. 1599 // In general, we cannot solve this problem. Consider: 1600 // class A<int> B() {} 1601 // which can be a function definition or a class definition when B() is a 1602 // macro. If we find enough real-world cases where this is a problem, we 1603 // can parse for the 'template' keyword in the beginning of the statement, 1604 // and thus rule out the record production in case there is no template 1605 // (this would still leave us with an ambiguity between template function 1606 // and class declarations). 1607 if (FormatTok->isOneOf(tok::colon, tok::less)) { 1608 while (!eof()) { 1609 if (FormatTok->is(tok::l_brace)) { 1610 calculateBraceTypes(/*ExpectClassBody=*/true); 1611 if (!tryToParseBracedList()) 1612 break; 1613 } 1614 if (FormatTok->Tok.is(tok::semi)) 1615 return; 1616 nextToken(); 1617 } 1618 } 1619 if (FormatTok->Tok.is(tok::l_brace)) { 1620 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1621 addUnwrappedLine(); 1622 1623 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1624 /*MunchSemi=*/false); 1625 } 1626 // We fall through to parsing a structural element afterwards, so 1627 // class A {} n, m; 1628 // will end up in one unwrapped line. 1629 // This does not apply for Java and JavaScript. 1630 if (Style.Language == FormatStyle::LK_Java || 1631 Style.Language == FormatStyle::LK_JavaScript) 1632 addUnwrappedLine(); 1633 } 1634 1635 void UnwrappedLineParser::parseObjCProtocolList() { 1636 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1637 do 1638 nextToken(); 1639 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1640 nextToken(); // Skip '>'. 1641 } 1642 1643 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1644 do { 1645 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1646 nextToken(); 1647 addUnwrappedLine(); 1648 break; 1649 } 1650 if (FormatTok->is(tok::l_brace)) { 1651 parseBlock(/*MustBeDeclaration=*/false); 1652 // In ObjC interfaces, nothing should be following the "}". 1653 addUnwrappedLine(); 1654 } else if (FormatTok->is(tok::r_brace)) { 1655 // Ignore stray "}". parseStructuralElement doesn't consume them. 1656 nextToken(); 1657 addUnwrappedLine(); 1658 } else { 1659 parseStructuralElement(); 1660 } 1661 } while (!eof()); 1662 } 1663 1664 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1665 nextToken(); 1666 nextToken(); // interface name 1667 1668 // @interface can be followed by either a base class, or a category. 1669 if (FormatTok->Tok.is(tok::colon)) { 1670 nextToken(); 1671 nextToken(); // base class name 1672 } else if (FormatTok->Tok.is(tok::l_paren)) 1673 // Skip category, if present. 1674 parseParens(); 1675 1676 if (FormatTok->Tok.is(tok::less)) 1677 parseObjCProtocolList(); 1678 1679 if (FormatTok->Tok.is(tok::l_brace)) { 1680 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1681 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1682 addUnwrappedLine(); 1683 parseBlock(/*MustBeDeclaration=*/true); 1684 } 1685 1686 // With instance variables, this puts '}' on its own line. Without instance 1687 // variables, this ends the @interface line. 1688 addUnwrappedLine(); 1689 1690 parseObjCUntilAtEnd(); 1691 } 1692 1693 void UnwrappedLineParser::parseObjCProtocol() { 1694 nextToken(); 1695 nextToken(); // protocol name 1696 1697 if (FormatTok->Tok.is(tok::less)) 1698 parseObjCProtocolList(); 1699 1700 // Check for protocol declaration. 1701 if (FormatTok->Tok.is(tok::semi)) { 1702 nextToken(); 1703 return addUnwrappedLine(); 1704 } 1705 1706 addUnwrappedLine(); 1707 parseObjCUntilAtEnd(); 1708 } 1709 1710 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { 1711 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export)); 1712 nextToken(); 1713 1714 // Consume the "default" in "export default class/function". 1715 if (FormatTok->is(tok::kw_default)) 1716 nextToken(); 1717 1718 // Consume "function" and "default function", so that these get parsed as 1719 // free-standing JS functions, i.e. do not require a trailing semicolon. 1720 if (FormatTok->is(Keywords.kw_function)) { 1721 nextToken(); 1722 return; 1723 } 1724 1725 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, Keywords.kw_var)) 1726 return; // Fall through to parsing the corresponding structure. 1727 1728 if (FormatTok->is(tok::l_brace)) { 1729 FormatTok->BlockKind = BK_Block; 1730 parseBracedList(); 1731 } 1732 1733 while (!eof() && FormatTok->isNot(tok::semi) && 1734 FormatTok->isNot(tok::l_brace)) { 1735 nextToken(); 1736 } 1737 } 1738 1739 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1740 StringRef Prefix = "") { 1741 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1742 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1743 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1744 E = Line.Tokens.end(); 1745 I != E; ++I) { 1746 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1747 } 1748 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1749 E = Line.Tokens.end(); 1750 I != E; ++I) { 1751 const UnwrappedLineNode &Node = *I; 1752 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1753 I = Node.Children.begin(), 1754 E = Node.Children.end(); 1755 I != E; ++I) { 1756 printDebugInfo(*I, "\nChild: "); 1757 } 1758 } 1759 llvm::dbgs() << "\n"; 1760 } 1761 1762 void UnwrappedLineParser::addUnwrappedLine() { 1763 if (Line->Tokens.empty()) 1764 return; 1765 DEBUG({ 1766 if (CurrentLines == &Lines) 1767 printDebugInfo(*Line); 1768 }); 1769 CurrentLines->push_back(std::move(*Line)); 1770 Line->Tokens.clear(); 1771 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1772 CurrentLines->append( 1773 std::make_move_iterator(PreprocessorDirectives.begin()), 1774 std::make_move_iterator(PreprocessorDirectives.end())); 1775 PreprocessorDirectives.clear(); 1776 } 1777 } 1778 1779 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1780 1781 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1782 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1783 FormatTok.NewlinesBefore > 0; 1784 } 1785 1786 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1787 bool JustComments = Line->Tokens.empty(); 1788 for (SmallVectorImpl<FormatToken *>::const_iterator 1789 I = CommentsBeforeNextToken.begin(), 1790 E = CommentsBeforeNextToken.end(); 1791 I != E; ++I) { 1792 if (isOnNewLine(**I) && JustComments) 1793 addUnwrappedLine(); 1794 pushToken(*I); 1795 } 1796 if (NewlineBeforeNext && JustComments) 1797 addUnwrappedLine(); 1798 CommentsBeforeNextToken.clear(); 1799 } 1800 1801 void UnwrappedLineParser::nextToken() { 1802 if (eof()) 1803 return; 1804 flushComments(isOnNewLine(*FormatTok)); 1805 pushToken(FormatTok); 1806 readToken(); 1807 } 1808 1809 void UnwrappedLineParser::readToken() { 1810 bool CommentsInCurrentLine = true; 1811 do { 1812 FormatTok = Tokens->getNextToken(); 1813 assert(FormatTok); 1814 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1815 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1816 // If there is an unfinished unwrapped line, we flush the preprocessor 1817 // directives only after that unwrapped line was finished later. 1818 bool SwitchToPreprocessorLines = !Line->Tokens.empty(); 1819 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1820 // Comments stored before the preprocessor directive need to be output 1821 // before the preprocessor directive, at the same level as the 1822 // preprocessor directive, as we consider them to apply to the directive. 1823 flushComments(isOnNewLine(*FormatTok)); 1824 parsePPDirective(); 1825 } 1826 while (FormatTok->Type == TT_ConflictStart || 1827 FormatTok->Type == TT_ConflictEnd || 1828 FormatTok->Type == TT_ConflictAlternative) { 1829 if (FormatTok->Type == TT_ConflictStart) { 1830 conditionalCompilationStart(/*Unreachable=*/false); 1831 } else if (FormatTok->Type == TT_ConflictAlternative) { 1832 conditionalCompilationAlternative(); 1833 } else if (FormatTok->Type == TT_ConflictEnd) { 1834 conditionalCompilationEnd(); 1835 } 1836 FormatTok = Tokens->getNextToken(); 1837 FormatTok->MustBreakBefore = true; 1838 } 1839 1840 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1841 !Line->InPPDirective) { 1842 continue; 1843 } 1844 1845 if (!FormatTok->Tok.is(tok::comment)) 1846 return; 1847 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1848 CommentsInCurrentLine = false; 1849 } 1850 if (CommentsInCurrentLine) { 1851 pushToken(FormatTok); 1852 } else { 1853 CommentsBeforeNextToken.push_back(FormatTok); 1854 } 1855 } while (!eof()); 1856 } 1857 1858 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1859 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1860 if (MustBreakBeforeNextToken) { 1861 Line->Tokens.back().Tok->MustBreakBefore = true; 1862 MustBreakBeforeNextToken = false; 1863 } 1864 } 1865 1866 } // end namespace format 1867 } // end namespace clang 1868