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 void UnwrappedLineParser::parseFile() { 257 // The top-level context in a file always has declarations, except for pre- 258 // processor directives and JavaScript files. 259 bool MustBeDeclaration = 260 !Line->InPPDirective && Style.Language != FormatStyle::LK_JavaScript; 261 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 262 MustBeDeclaration); 263 parseLevel(/*HasOpeningBrace=*/false); 264 // Make sure to format the remaining tokens. 265 flushComments(true); 266 addUnwrappedLine(); 267 } 268 269 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) { 270 bool SwitchLabelEncountered = false; 271 do { 272 tok::TokenKind kind = FormatTok->Tok.getKind(); 273 if (FormatTok->Type == TT_MacroBlockBegin) { 274 kind = tok::l_brace; 275 } else if (FormatTok->Type == TT_MacroBlockEnd) { 276 kind = tok::r_brace; 277 } 278 279 switch (kind) { 280 case tok::comment: 281 nextToken(); 282 addUnwrappedLine(); 283 break; 284 case tok::l_brace: 285 // FIXME: Add parameter whether this can happen - if this happens, we must 286 // be in a non-declaration context. 287 parseBlock(/*MustBeDeclaration=*/false); 288 addUnwrappedLine(); 289 break; 290 case tok::r_brace: 291 if (HasOpeningBrace) 292 return; 293 nextToken(); 294 addUnwrappedLine(); 295 break; 296 case tok::kw_default: 297 case tok::kw_case: 298 if (!SwitchLabelEncountered && 299 (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1))) 300 ++Line->Level; 301 SwitchLabelEncountered = true; 302 parseStructuralElement(); 303 break; 304 default: 305 parseStructuralElement(); 306 break; 307 } 308 } while (!eof()); 309 } 310 311 void UnwrappedLineParser::calculateBraceTypes(bool ExpectClassBody) { 312 // We'll parse forward through the tokens until we hit 313 // a closing brace or eof - note that getNextToken() will 314 // parse macros, so this will magically work inside macro 315 // definitions, too. 316 unsigned StoredPosition = Tokens->getPosition(); 317 FormatToken *Tok = FormatTok; 318 // Keep a stack of positions of lbrace tokens. We will 319 // update information about whether an lbrace starts a 320 // braced init list or a different block during the loop. 321 SmallVector<FormatToken *, 8> LBraceStack; 322 assert(Tok->Tok.is(tok::l_brace)); 323 do { 324 // Get next none-comment token. 325 FormatToken *NextTok; 326 unsigned ReadTokens = 0; 327 do { 328 NextTok = Tokens->getNextToken(); 329 ++ReadTokens; 330 } while (NextTok->is(tok::comment)); 331 332 switch (Tok->Tok.getKind()) { 333 case tok::l_brace: 334 Tok->BlockKind = BK_Unknown; 335 LBraceStack.push_back(Tok); 336 break; 337 case tok::r_brace: 338 if (!LBraceStack.empty()) { 339 if (LBraceStack.back()->BlockKind == BK_Unknown) { 340 bool ProbablyBracedList = false; 341 if (Style.Language == FormatStyle::LK_Proto) { 342 ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); 343 } else { 344 // Using OriginalColumn to distinguish between ObjC methods and 345 // binary operators is a bit hacky. 346 bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) && 347 NextTok->OriginalColumn == 0; 348 349 // If there is a comma, semicolon or right paren after the closing 350 // brace, we assume this is a braced initializer list. Note that 351 // regardless how we mark inner braces here, we will overwrite the 352 // BlockKind later if we parse a braced list (where all blocks 353 // inside are by default braced lists), or when we explicitly detect 354 // blocks (for example while parsing lambdas). 355 // 356 // We exclude + and - as they can be ObjC visibility modifiers. 357 ProbablyBracedList = 358 NextTok->isOneOf(tok::comma, tok::period, tok::colon, 359 tok::r_paren, tok::r_square, tok::l_brace, 360 tok::l_paren, tok::ellipsis) || 361 (NextTok->is(tok::semi) && 362 (!ExpectClassBody || LBraceStack.size() != 1)) || 363 (NextTok->isBinaryOperator() && !NextIsObjCMethod); 364 } 365 if (ProbablyBracedList) { 366 Tok->BlockKind = BK_BracedInit; 367 LBraceStack.back()->BlockKind = BK_BracedInit; 368 } else { 369 Tok->BlockKind = BK_Block; 370 LBraceStack.back()->BlockKind = BK_Block; 371 } 372 } 373 LBraceStack.pop_back(); 374 } 375 break; 376 case tok::at: 377 case tok::semi: 378 case tok::kw_if: 379 case tok::kw_while: 380 case tok::kw_for: 381 case tok::kw_switch: 382 case tok::kw_try: 383 case tok::kw___try: 384 if (!LBraceStack.empty()) 385 LBraceStack.back()->BlockKind = BK_Block; 386 break; 387 default: 388 break; 389 } 390 Tok = NextTok; 391 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 392 // Assume other blocks for all unclosed opening braces. 393 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 394 if (LBraceStack[i]->BlockKind == BK_Unknown) 395 LBraceStack[i]->BlockKind = BK_Block; 396 } 397 398 FormatTok = Tokens->setPosition(StoredPosition); 399 } 400 401 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 402 bool MunchSemi) { 403 assert(FormatTok->isOneOf(tok::l_brace, TT_MacroBlockBegin) && 404 "'{' or macro block token expected"); 405 const bool MacroBlock = FormatTok->is(TT_MacroBlockBegin); 406 407 unsigned InitialLevel = Line->Level; 408 nextToken(); 409 410 if (MacroBlock && FormatTok->is(tok::l_paren)) 411 parseParens(); 412 413 addUnwrappedLine(); 414 415 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 416 MustBeDeclaration); 417 if (AddLevel) 418 ++Line->Level; 419 parseLevel(/*HasOpeningBrace=*/true); 420 421 if (MacroBlock ? !FormatTok->is(TT_MacroBlockEnd) 422 : !FormatTok->is(tok::r_brace)) { 423 Line->Level = InitialLevel; 424 return; 425 } 426 427 nextToken(); // Munch the closing brace. 428 429 if (MacroBlock && FormatTok->is(tok::l_paren)) 430 parseParens(); 431 432 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 433 nextToken(); 434 Line->Level = InitialLevel; 435 } 436 437 static bool isGoogScope(const UnwrappedLine &Line) { 438 // FIXME: Closure-library specific stuff should not be hard-coded but be 439 // configurable. 440 if (Line.Tokens.size() < 4) 441 return false; 442 auto I = Line.Tokens.begin(); 443 if (I->Tok->TokenText != "goog") 444 return false; 445 ++I; 446 if (I->Tok->isNot(tok::period)) 447 return false; 448 ++I; 449 if (I->Tok->TokenText != "scope") 450 return false; 451 ++I; 452 return I->Tok->is(tok::l_paren); 453 } 454 455 static bool ShouldBreakBeforeBrace(const FormatStyle &Style, 456 const FormatToken &InitialToken) { 457 switch (Style.BreakBeforeBraces) { 458 case FormatStyle::BS_Linux: 459 return InitialToken.isOneOf(tok::kw_namespace, tok::kw_class); 460 case FormatStyle::BS_Allman: 461 case FormatStyle::BS_GNU: 462 return true; 463 default: 464 return false; 465 } 466 } 467 468 void UnwrappedLineParser::parseChildBlock() { 469 FormatTok->BlockKind = BK_Block; 470 nextToken(); 471 { 472 bool GoogScope = 473 Style.Language == FormatStyle::LK_JavaScript && isGoogScope(*Line); 474 ScopedLineState LineState(*this); 475 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 476 /*MustBeDeclaration=*/false); 477 Line->Level += GoogScope ? 0 : 1; 478 parseLevel(/*HasOpeningBrace=*/true); 479 flushComments(isOnNewLine(*FormatTok)); 480 Line->Level -= GoogScope ? 0 : 1; 481 } 482 nextToken(); 483 } 484 485 void UnwrappedLineParser::parsePPDirective() { 486 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 487 ScopedMacroState MacroState(*Line, Tokens, FormatTok); 488 nextToken(); 489 490 if (!FormatTok->Tok.getIdentifierInfo()) { 491 parsePPUnknown(); 492 return; 493 } 494 495 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 496 case tok::pp_define: 497 parsePPDefine(); 498 return; 499 case tok::pp_if: 500 parsePPIf(/*IfDef=*/false); 501 break; 502 case tok::pp_ifdef: 503 case tok::pp_ifndef: 504 parsePPIf(/*IfDef=*/true); 505 break; 506 case tok::pp_else: 507 parsePPElse(); 508 break; 509 case tok::pp_elif: 510 parsePPElIf(); 511 break; 512 case tok::pp_endif: 513 parsePPEndIf(); 514 break; 515 default: 516 parsePPUnknown(); 517 break; 518 } 519 } 520 521 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 522 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 523 PPStack.push_back(PP_Unreachable); 524 else 525 PPStack.push_back(PP_Conditional); 526 } 527 528 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 529 ++PPBranchLevel; 530 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 531 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 532 PPLevelBranchIndex.push_back(0); 533 PPLevelBranchCount.push_back(0); 534 } 535 PPChainBranchIndex.push(0); 536 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 537 conditionalCompilationCondition(Unreachable || Skip); 538 } 539 540 void UnwrappedLineParser::conditionalCompilationAlternative() { 541 if (!PPStack.empty()) 542 PPStack.pop_back(); 543 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 544 if (!PPChainBranchIndex.empty()) 545 ++PPChainBranchIndex.top(); 546 conditionalCompilationCondition( 547 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 548 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 549 } 550 551 void UnwrappedLineParser::conditionalCompilationEnd() { 552 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 553 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 554 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 555 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 556 } 557 } 558 // Guard against #endif's without #if. 559 if (PPBranchLevel > 0) 560 --PPBranchLevel; 561 if (!PPChainBranchIndex.empty()) 562 PPChainBranchIndex.pop(); 563 if (!PPStack.empty()) 564 PPStack.pop_back(); 565 } 566 567 void UnwrappedLineParser::parsePPIf(bool IfDef) { 568 nextToken(); 569 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 570 FormatTok->Tok.getLiteralData() != nullptr && 571 StringRef(FormatTok->Tok.getLiteralData(), 572 FormatTok->Tok.getLength()) == "0") || 573 FormatTok->Tok.is(tok::kw_false); 574 conditionalCompilationStart(!IfDef && IsLiteralFalse); 575 parsePPUnknown(); 576 } 577 578 void UnwrappedLineParser::parsePPElse() { 579 conditionalCompilationAlternative(); 580 parsePPUnknown(); 581 } 582 583 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 584 585 void UnwrappedLineParser::parsePPEndIf() { 586 conditionalCompilationEnd(); 587 parsePPUnknown(); 588 } 589 590 void UnwrappedLineParser::parsePPDefine() { 591 nextToken(); 592 593 if (FormatTok->Tok.getKind() != tok::identifier) { 594 parsePPUnknown(); 595 return; 596 } 597 nextToken(); 598 if (FormatTok->Tok.getKind() == tok::l_paren && 599 FormatTok->WhitespaceRange.getBegin() == 600 FormatTok->WhitespaceRange.getEnd()) { 601 parseParens(); 602 } 603 addUnwrappedLine(); 604 Line->Level = 1; 605 606 // Errors during a preprocessor directive can only affect the layout of the 607 // preprocessor directive, and thus we ignore them. An alternative approach 608 // would be to use the same approach we use on the file level (no 609 // re-indentation if there was a structural error) within the macro 610 // definition. 611 parseFile(); 612 } 613 614 void UnwrappedLineParser::parsePPUnknown() { 615 do { 616 nextToken(); 617 } while (!eof()); 618 addUnwrappedLine(); 619 } 620 621 // Here we blacklist certain tokens that are not usually the first token in an 622 // unwrapped line. This is used in attempt to distinguish macro calls without 623 // trailing semicolons from other constructs split to several lines. 624 static bool tokenCanStartNewLine(const clang::Token &Tok) { 625 // Semicolon can be a null-statement, l_square can be a start of a macro or 626 // a C++11 attribute, but this doesn't seem to be common. 627 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 628 Tok.isNot(tok::l_square) && 629 // Tokens that can only be used as binary operators and a part of 630 // overloaded operator names. 631 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 632 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 633 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 634 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 635 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 636 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 637 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 638 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 639 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 640 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 641 Tok.isNot(tok::lesslessequal) && 642 // Colon is used in labels, base class lists, initializer lists, 643 // range-based for loops, ternary operator, but should never be the 644 // first token in an unwrapped line. 645 Tok.isNot(tok::colon) && 646 // 'noexcept' is a trailing annotation. 647 Tok.isNot(tok::kw_noexcept); 648 } 649 650 void UnwrappedLineParser::parseStructuralElement() { 651 assert(!FormatTok->Tok.is(tok::l_brace)); 652 switch (FormatTok->Tok.getKind()) { 653 case tok::at: 654 nextToken(); 655 if (FormatTok->Tok.is(tok::l_brace)) { 656 parseBracedList(); 657 break; 658 } 659 switch (FormatTok->Tok.getObjCKeywordID()) { 660 case tok::objc_public: 661 case tok::objc_protected: 662 case tok::objc_package: 663 case tok::objc_private: 664 return parseAccessSpecifier(); 665 case tok::objc_interface: 666 case tok::objc_implementation: 667 return parseObjCInterfaceOrImplementation(); 668 case tok::objc_protocol: 669 return parseObjCProtocol(); 670 case tok::objc_end: 671 return; // Handled by the caller. 672 case tok::objc_optional: 673 case tok::objc_required: 674 nextToken(); 675 addUnwrappedLine(); 676 return; 677 case tok::objc_autoreleasepool: 678 nextToken(); 679 if (FormatTok->Tok.is(tok::l_brace)) { 680 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 681 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 682 addUnwrappedLine(); 683 parseBlock(/*MustBeDeclaration=*/false); 684 } 685 addUnwrappedLine(); 686 return; 687 case tok::objc_try: 688 // This branch isn't strictly necessary (the kw_try case below would 689 // do this too after the tok::at is parsed above). But be explicit. 690 parseTryCatch(); 691 return; 692 default: 693 break; 694 } 695 break; 696 case tok::kw_asm: 697 nextToken(); 698 if (FormatTok->is(tok::l_brace)) { 699 FormatTok->Type = TT_InlineASMBrace; 700 nextToken(); 701 while (FormatTok && FormatTok->isNot(tok::eof)) { 702 if (FormatTok->is(tok::r_brace)) { 703 FormatTok->Type = TT_InlineASMBrace; 704 nextToken(); 705 addUnwrappedLine(); 706 break; 707 } 708 FormatTok->Finalized = true; 709 nextToken(); 710 } 711 } 712 break; 713 case tok::kw_namespace: 714 parseNamespace(); 715 return; 716 case tok::kw_inline: 717 nextToken(); 718 if (FormatTok->Tok.is(tok::kw_namespace)) { 719 parseNamespace(); 720 return; 721 } 722 break; 723 case tok::kw_public: 724 case tok::kw_protected: 725 case tok::kw_private: 726 if (Style.Language == FormatStyle::LK_Java || 727 Style.Language == FormatStyle::LK_JavaScript) 728 nextToken(); 729 else 730 parseAccessSpecifier(); 731 return; 732 case tok::kw_if: 733 parseIfThenElse(); 734 return; 735 case tok::kw_for: 736 case tok::kw_while: 737 parseForOrWhileLoop(); 738 return; 739 case tok::kw_do: 740 parseDoWhile(); 741 return; 742 case tok::kw_switch: 743 parseSwitch(); 744 return; 745 case tok::kw_default: 746 nextToken(); 747 parseLabel(); 748 return; 749 case tok::kw_case: 750 parseCaseLabel(); 751 return; 752 case tok::kw_try: 753 case tok::kw___try: 754 parseTryCatch(); 755 return; 756 case tok::kw_extern: 757 nextToken(); 758 if (FormatTok->Tok.is(tok::string_literal)) { 759 nextToken(); 760 if (FormatTok->Tok.is(tok::l_brace)) { 761 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 762 addUnwrappedLine(); 763 return; 764 } 765 } 766 break; 767 case tok::kw_export: 768 if (Style.Language == FormatStyle::LK_JavaScript) { 769 parseJavaScriptEs6ImportExport(); 770 return; 771 } 772 break; 773 case tok::identifier: 774 if (FormatTok->is(TT_ForEachMacro)) { 775 parseForOrWhileLoop(); 776 return; 777 } 778 if (FormatTok->is(TT_MacroBlockBegin)) { 779 parseBlock(/*MustBeDeclaration=*/false, /*AddLevel=*/true, 780 /*MunchSemi=*/false); 781 return; 782 } 783 if (Style.Language == FormatStyle::LK_JavaScript && 784 FormatTok->is(Keywords.kw_import)) { 785 parseJavaScriptEs6ImportExport(); 786 return; 787 } 788 if (FormatTok->is(Keywords.kw_signals)) { 789 nextToken(); 790 if (FormatTok->is(tok::colon)) { 791 nextToken(); 792 addUnwrappedLine(); 793 } 794 return; 795 } 796 // In all other cases, parse the declaration. 797 break; 798 default: 799 break; 800 } 801 do { 802 switch (FormatTok->Tok.getKind()) { 803 case tok::at: 804 nextToken(); 805 if (FormatTok->Tok.is(tok::l_brace)) 806 parseBracedList(); 807 break; 808 case tok::kw_enum: 809 // parseEnum falls through and does not yet add an unwrapped line as an 810 // enum definition can start a structural element. 811 parseEnum(); 812 // This does not apply for Java and JavaScript. 813 if (Style.Language == FormatStyle::LK_Java || 814 Style.Language == FormatStyle::LK_JavaScript) { 815 addUnwrappedLine(); 816 return; 817 } 818 break; 819 case tok::kw_typedef: 820 nextToken(); 821 if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS, 822 Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS)) 823 parseEnum(); 824 break; 825 case tok::kw_struct: 826 case tok::kw_union: 827 case tok::kw_class: 828 // parseRecord falls through and does not yet add an unwrapped line as a 829 // record declaration or definition can start a structural element. 830 parseRecord(); 831 // This does not apply for Java and JavaScript. 832 if (Style.Language == FormatStyle::LK_Java || 833 Style.Language == FormatStyle::LK_JavaScript) { 834 addUnwrappedLine(); 835 return; 836 } 837 break; 838 case tok::period: 839 nextToken(); 840 // In Java, classes have an implicit static member "class". 841 if (Style.Language == FormatStyle::LK_Java && FormatTok && 842 FormatTok->is(tok::kw_class)) 843 nextToken(); 844 break; 845 case tok::semi: 846 nextToken(); 847 addUnwrappedLine(); 848 return; 849 case tok::r_brace: 850 addUnwrappedLine(); 851 return; 852 case tok::l_paren: 853 parseParens(); 854 break; 855 case tok::caret: 856 nextToken(); 857 if (FormatTok->Tok.isAnyIdentifier() || 858 FormatTok->isSimpleTypeSpecifier()) 859 nextToken(); 860 if (FormatTok->is(tok::l_paren)) 861 parseParens(); 862 if (FormatTok->is(tok::l_brace)) 863 parseChildBlock(); 864 break; 865 case tok::l_brace: 866 if (!tryToParseBracedList()) { 867 // A block outside of parentheses must be the last part of a 868 // structural element. 869 // FIXME: Figure out cases where this is not true, and add projections 870 // for them (the one we know is missing are lambdas). 871 if (Style.BreakBeforeBraces != FormatStyle::BS_Attach) 872 addUnwrappedLine(); 873 FormatTok->Type = TT_FunctionLBrace; 874 parseBlock(/*MustBeDeclaration=*/false); 875 addUnwrappedLine(); 876 return; 877 } 878 // Otherwise this was a braced init list, and the structural 879 // element continues. 880 break; 881 case tok::kw_try: 882 // We arrive here when parsing function-try blocks. 883 parseTryCatch(); 884 return; 885 case tok::identifier: { 886 if (FormatTok->is(TT_MacroBlockEnd)) { 887 addUnwrappedLine(); 888 return; 889 } 890 891 // Parse function literal unless 'function' is the first token in a line 892 // in which case this should be treated as a free-standing function. 893 if (Style.Language == FormatStyle::LK_JavaScript && 894 FormatTok->is(Keywords.kw_function) && Line->Tokens.size() > 0) { 895 tryToParseJSFunction(); 896 break; 897 } 898 if ((Style.Language == FormatStyle::LK_JavaScript || 899 Style.Language == FormatStyle::LK_Java) && 900 FormatTok->is(Keywords.kw_interface)) { 901 parseRecord(); 902 addUnwrappedLine(); 903 return; 904 } 905 906 StringRef Text = FormatTok->TokenText; 907 nextToken(); 908 if (Line->Tokens.size() == 1 && 909 // JS doesn't have macros, and within classes colons indicate fields, 910 // not labels. 911 Style.Language != FormatStyle::LK_JavaScript) { 912 if (FormatTok->Tok.is(tok::colon) && !Line->MustBeDeclaration) { 913 parseLabel(); 914 return; 915 } 916 // Recognize function-like macro usages without trailing semicolon as 917 // well as free-standing macros like Q_OBJECT. 918 bool FunctionLike = FormatTok->is(tok::l_paren); 919 if (FunctionLike) 920 parseParens(); 921 922 bool FollowedByNewline = 923 CommentsBeforeNextToken.empty() 924 ? FormatTok->NewlinesBefore > 0 925 : CommentsBeforeNextToken.front()->NewlinesBefore > 0; 926 927 if (FollowedByNewline && (Text.size() >= 5 || FunctionLike) && 928 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 929 addUnwrappedLine(); 930 return; 931 } 932 } 933 break; 934 } 935 case tok::equal: 936 // Fat arrows (=>) have tok::TokenKind tok::equal but TokenType 937 // TT_JsFatArrow. The always start an expression or a child block if 938 // followed by a curly. 939 if (FormatTok->is(TT_JsFatArrow)) { 940 nextToken(); 941 if (FormatTok->is(tok::l_brace)) 942 parseChildBlock(); 943 break; 944 } 945 946 nextToken(); 947 if (FormatTok->Tok.is(tok::l_brace)) { 948 parseBracedList(); 949 } 950 break; 951 case tok::l_square: 952 parseSquare(); 953 break; 954 case tok::kw_new: 955 parseNew(); 956 break; 957 default: 958 nextToken(); 959 break; 960 } 961 } while (!eof()); 962 } 963 964 bool UnwrappedLineParser::tryToParseLambda() { 965 if (Style.Language != FormatStyle::LK_Cpp) { 966 nextToken(); 967 return false; 968 } 969 // FIXME: This is a dirty way to access the previous token. Find a better 970 // solution. 971 if (!Line->Tokens.empty() && 972 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator, 973 tok::kw_new, tok::kw_delete) || 974 Line->Tokens.back().Tok->closesScope() || 975 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 976 nextToken(); 977 return false; 978 } 979 assert(FormatTok->is(tok::l_square)); 980 FormatToken &LSquare = *FormatTok; 981 if (!tryToParseLambdaIntroducer()) 982 return false; 983 984 while (FormatTok->isNot(tok::l_brace)) { 985 if (FormatTok->isSimpleTypeSpecifier()) { 986 nextToken(); 987 continue; 988 } 989 switch (FormatTok->Tok.getKind()) { 990 case tok::l_brace: 991 break; 992 case tok::l_paren: 993 parseParens(); 994 break; 995 case tok::amp: 996 case tok::star: 997 case tok::kw_const: 998 case tok::comma: 999 case tok::less: 1000 case tok::greater: 1001 case tok::identifier: 1002 case tok::coloncolon: 1003 case tok::kw_mutable: 1004 nextToken(); 1005 break; 1006 case tok::arrow: 1007 FormatTok->Type = TT_LambdaArrow; 1008 nextToken(); 1009 break; 1010 default: 1011 return true; 1012 } 1013 } 1014 LSquare.Type = TT_LambdaLSquare; 1015 parseChildBlock(); 1016 return true; 1017 } 1018 1019 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 1020 nextToken(); 1021 if (FormatTok->is(tok::equal)) { 1022 nextToken(); 1023 if (FormatTok->is(tok::r_square)) { 1024 nextToken(); 1025 return true; 1026 } 1027 if (FormatTok->isNot(tok::comma)) 1028 return false; 1029 nextToken(); 1030 } else if (FormatTok->is(tok::amp)) { 1031 nextToken(); 1032 if (FormatTok->is(tok::r_square)) { 1033 nextToken(); 1034 return true; 1035 } 1036 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 1037 return false; 1038 } 1039 if (FormatTok->is(tok::comma)) 1040 nextToken(); 1041 } else if (FormatTok->is(tok::r_square)) { 1042 nextToken(); 1043 return true; 1044 } 1045 do { 1046 if (FormatTok->is(tok::amp)) 1047 nextToken(); 1048 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 1049 return false; 1050 nextToken(); 1051 if (FormatTok->is(tok::ellipsis)) 1052 nextToken(); 1053 if (FormatTok->is(tok::comma)) { 1054 nextToken(); 1055 } else if (FormatTok->is(tok::r_square)) { 1056 nextToken(); 1057 return true; 1058 } else { 1059 return false; 1060 } 1061 } while (!eof()); 1062 return false; 1063 } 1064 1065 void UnwrappedLineParser::tryToParseJSFunction() { 1066 nextToken(); 1067 1068 // Consume function name. 1069 if (FormatTok->is(tok::identifier)) 1070 nextToken(); 1071 1072 if (FormatTok->isNot(tok::l_paren)) 1073 return; 1074 1075 // Parse formal parameter list. 1076 parseParens(); 1077 1078 if (FormatTok->is(tok::colon)) { 1079 // Parse a type definition. 1080 nextToken(); 1081 1082 // Eat the type declaration. For braced inline object types, balance braces, 1083 // otherwise just parse until finding an l_brace for the function body. 1084 if (FormatTok->is(tok::l_brace)) 1085 tryToParseBracedList(); 1086 else 1087 while (FormatTok->isNot(tok::l_brace) && !eof()) 1088 nextToken(); 1089 } 1090 1091 parseChildBlock(); 1092 } 1093 1094 bool UnwrappedLineParser::tryToParseBracedList() { 1095 if (FormatTok->BlockKind == BK_Unknown) 1096 calculateBraceTypes(); 1097 assert(FormatTok->BlockKind != BK_Unknown); 1098 if (FormatTok->BlockKind == BK_Block) 1099 return false; 1100 parseBracedList(); 1101 return true; 1102 } 1103 1104 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 1105 bool HasError = false; 1106 nextToken(); 1107 1108 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 1109 // replace this by using parseAssigmentExpression() inside. 1110 do { 1111 if (Style.Language == FormatStyle::LK_JavaScript) { 1112 if (FormatTok->is(Keywords.kw_function)) { 1113 tryToParseJSFunction(); 1114 continue; 1115 } 1116 if (FormatTok->is(TT_JsFatArrow)) { 1117 nextToken(); 1118 // Fat arrows can be followed by simple expressions or by child blocks 1119 // in curly braces. 1120 if (FormatTok->is(tok::l_brace)) { 1121 parseChildBlock(); 1122 continue; 1123 } 1124 } 1125 } 1126 switch (FormatTok->Tok.getKind()) { 1127 case tok::caret: 1128 nextToken(); 1129 if (FormatTok->is(tok::l_brace)) { 1130 parseChildBlock(); 1131 } 1132 break; 1133 case tok::l_square: 1134 tryToParseLambda(); 1135 break; 1136 case tok::l_brace: 1137 // Assume there are no blocks inside a braced init list apart 1138 // from the ones we explicitly parse out (like lambdas). 1139 FormatTok->BlockKind = BK_BracedInit; 1140 parseBracedList(); 1141 break; 1142 case tok::l_paren: 1143 parseParens(); 1144 // JavaScript can just have free standing methods and getters/setters in 1145 // object literals. Detect them by a "{" following ")". 1146 if (Style.Language == FormatStyle::LK_JavaScript) { 1147 if (FormatTok->is(tok::l_brace)) 1148 parseChildBlock(); 1149 break; 1150 } 1151 break; 1152 case tok::r_brace: 1153 nextToken(); 1154 return !HasError; 1155 case tok::semi: 1156 HasError = true; 1157 if (!ContinueOnSemicolons) 1158 return !HasError; 1159 nextToken(); 1160 break; 1161 case tok::comma: 1162 nextToken(); 1163 break; 1164 default: 1165 nextToken(); 1166 break; 1167 } 1168 } while (!eof()); 1169 return false; 1170 } 1171 1172 void UnwrappedLineParser::parseParens() { 1173 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1174 nextToken(); 1175 do { 1176 switch (FormatTok->Tok.getKind()) { 1177 case tok::l_paren: 1178 parseParens(); 1179 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) 1180 parseChildBlock(); 1181 break; 1182 case tok::r_paren: 1183 nextToken(); 1184 return; 1185 case tok::r_brace: 1186 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1187 return; 1188 case tok::l_square: 1189 tryToParseLambda(); 1190 break; 1191 case tok::l_brace: 1192 if (!tryToParseBracedList()) 1193 parseChildBlock(); 1194 break; 1195 case tok::at: 1196 nextToken(); 1197 if (FormatTok->Tok.is(tok::l_brace)) 1198 parseBracedList(); 1199 break; 1200 case tok::identifier: 1201 if (Style.Language == FormatStyle::LK_JavaScript && 1202 FormatTok->is(Keywords.kw_function)) 1203 tryToParseJSFunction(); 1204 else 1205 nextToken(); 1206 break; 1207 default: 1208 nextToken(); 1209 break; 1210 } 1211 } while (!eof()); 1212 } 1213 1214 void UnwrappedLineParser::parseSquare() { 1215 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1216 if (tryToParseLambda()) 1217 return; 1218 do { 1219 switch (FormatTok->Tok.getKind()) { 1220 case tok::l_paren: 1221 parseParens(); 1222 break; 1223 case tok::r_square: 1224 nextToken(); 1225 return; 1226 case tok::r_brace: 1227 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1228 return; 1229 case tok::l_square: 1230 parseSquare(); 1231 break; 1232 case tok::l_brace: { 1233 if (!tryToParseBracedList()) 1234 parseChildBlock(); 1235 break; 1236 } 1237 case tok::at: 1238 nextToken(); 1239 if (FormatTok->Tok.is(tok::l_brace)) 1240 parseBracedList(); 1241 break; 1242 default: 1243 nextToken(); 1244 break; 1245 } 1246 } while (!eof()); 1247 } 1248 1249 void UnwrappedLineParser::parseIfThenElse() { 1250 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1251 nextToken(); 1252 if (FormatTok->Tok.is(tok::l_paren)) 1253 parseParens(); 1254 bool NeedsUnwrappedLine = false; 1255 if (FormatTok->Tok.is(tok::l_brace)) { 1256 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1257 parseBlock(/*MustBeDeclaration=*/false); 1258 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1259 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1260 addUnwrappedLine(); 1261 } else { 1262 NeedsUnwrappedLine = true; 1263 } 1264 } else { 1265 addUnwrappedLine(); 1266 ++Line->Level; 1267 parseStructuralElement(); 1268 --Line->Level; 1269 } 1270 if (FormatTok->Tok.is(tok::kw_else)) { 1271 if (Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) 1272 addUnwrappedLine(); 1273 nextToken(); 1274 if (FormatTok->Tok.is(tok::l_brace)) { 1275 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1276 parseBlock(/*MustBeDeclaration=*/false); 1277 addUnwrappedLine(); 1278 } else if (FormatTok->Tok.is(tok::kw_if)) { 1279 parseIfThenElse(); 1280 } else { 1281 addUnwrappedLine(); 1282 ++Line->Level; 1283 parseStructuralElement(); 1284 --Line->Level; 1285 } 1286 } else if (NeedsUnwrappedLine) { 1287 addUnwrappedLine(); 1288 } 1289 } 1290 1291 void UnwrappedLineParser::parseTryCatch() { 1292 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected"); 1293 nextToken(); 1294 bool NeedsUnwrappedLine = false; 1295 if (FormatTok->is(tok::colon)) { 1296 // We are in a function try block, what comes is an initializer list. 1297 nextToken(); 1298 while (FormatTok->is(tok::identifier)) { 1299 nextToken(); 1300 if (FormatTok->is(tok::l_paren)) 1301 parseParens(); 1302 if (FormatTok->is(tok::comma)) 1303 nextToken(); 1304 } 1305 } 1306 // Parse try with resource. 1307 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) { 1308 parseParens(); 1309 } 1310 if (FormatTok->is(tok::l_brace)) { 1311 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1312 parseBlock(/*MustBeDeclaration=*/false); 1313 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1314 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1315 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1316 addUnwrappedLine(); 1317 } else { 1318 NeedsUnwrappedLine = true; 1319 } 1320 } else if (!FormatTok->is(tok::kw_catch)) { 1321 // The C++ standard requires a compound-statement after a try. 1322 // If there's none, we try to assume there's a structuralElement 1323 // and try to continue. 1324 addUnwrappedLine(); 1325 ++Line->Level; 1326 parseStructuralElement(); 1327 --Line->Level; 1328 } 1329 while (1) { 1330 if (FormatTok->is(tok::at)) 1331 nextToken(); 1332 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, 1333 tok::kw___finally) || 1334 ((Style.Language == FormatStyle::LK_Java || 1335 Style.Language == FormatStyle::LK_JavaScript) && 1336 FormatTok->is(Keywords.kw_finally)) || 1337 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) || 1338 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally)))) 1339 break; 1340 nextToken(); 1341 while (FormatTok->isNot(tok::l_brace)) { 1342 if (FormatTok->is(tok::l_paren)) { 1343 parseParens(); 1344 continue; 1345 } 1346 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) 1347 return; 1348 nextToken(); 1349 } 1350 NeedsUnwrappedLine = false; 1351 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1352 parseBlock(/*MustBeDeclaration=*/false); 1353 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1354 Style.BreakBeforeBraces == FormatStyle::BS_GNU || 1355 Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) { 1356 addUnwrappedLine(); 1357 } else { 1358 NeedsUnwrappedLine = true; 1359 } 1360 } 1361 if (NeedsUnwrappedLine) { 1362 addUnwrappedLine(); 1363 } 1364 } 1365 1366 void UnwrappedLineParser::parseNamespace() { 1367 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1368 1369 const FormatToken &InitialToken = *FormatTok; 1370 nextToken(); 1371 if (FormatTok->Tok.is(tok::identifier)) 1372 nextToken(); 1373 if (FormatTok->Tok.is(tok::l_brace)) { 1374 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1375 addUnwrappedLine(); 1376 1377 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1378 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1379 DeclarationScopeStack.size() > 1); 1380 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1381 // Munch the semicolon after a namespace. This is more common than one would 1382 // think. Puttin the semicolon into its own line is very ugly. 1383 if (FormatTok->Tok.is(tok::semi)) 1384 nextToken(); 1385 addUnwrappedLine(); 1386 } 1387 // FIXME: Add error handling. 1388 } 1389 1390 void UnwrappedLineParser::parseNew() { 1391 assert(FormatTok->is(tok::kw_new) && "'new' expected"); 1392 nextToken(); 1393 if (Style.Language != FormatStyle::LK_Java) 1394 return; 1395 1396 // In Java, we can parse everything up to the parens, which aren't optional. 1397 do { 1398 // There should not be a ;, { or } before the new's open paren. 1399 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) 1400 return; 1401 1402 // Consume the parens. 1403 if (FormatTok->is(tok::l_paren)) { 1404 parseParens(); 1405 1406 // If there is a class body of an anonymous class, consume that as child. 1407 if (FormatTok->is(tok::l_brace)) 1408 parseChildBlock(); 1409 return; 1410 } 1411 nextToken(); 1412 } while (!eof()); 1413 } 1414 1415 void UnwrappedLineParser::parseForOrWhileLoop() { 1416 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) && 1417 "'for', 'while' or foreach macro expected"); 1418 nextToken(); 1419 if (FormatTok->Tok.is(tok::l_paren)) 1420 parseParens(); 1421 if (FormatTok->Tok.is(tok::l_brace)) { 1422 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1423 parseBlock(/*MustBeDeclaration=*/false); 1424 addUnwrappedLine(); 1425 } else { 1426 addUnwrappedLine(); 1427 ++Line->Level; 1428 parseStructuralElement(); 1429 --Line->Level; 1430 } 1431 } 1432 1433 void UnwrappedLineParser::parseDoWhile() { 1434 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1435 nextToken(); 1436 if (FormatTok->Tok.is(tok::l_brace)) { 1437 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1438 parseBlock(/*MustBeDeclaration=*/false); 1439 if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1440 addUnwrappedLine(); 1441 } else { 1442 addUnwrappedLine(); 1443 ++Line->Level; 1444 parseStructuralElement(); 1445 --Line->Level; 1446 } 1447 1448 // FIXME: Add error handling. 1449 if (!FormatTok->Tok.is(tok::kw_while)) { 1450 addUnwrappedLine(); 1451 return; 1452 } 1453 1454 nextToken(); 1455 parseStructuralElement(); 1456 } 1457 1458 void UnwrappedLineParser::parseLabel() { 1459 nextToken(); 1460 unsigned OldLineLevel = Line->Level; 1461 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1462 --Line->Level; 1463 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1464 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1465 parseBlock(/*MustBeDeclaration=*/false); 1466 if (FormatTok->Tok.is(tok::kw_break)) { 1467 // "break;" after "}" on its own line only for BS_Allman and BS_GNU 1468 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1469 Style.BreakBeforeBraces == FormatStyle::BS_GNU) { 1470 addUnwrappedLine(); 1471 } 1472 parseStructuralElement(); 1473 } 1474 addUnwrappedLine(); 1475 } else { 1476 if (FormatTok->is(tok::semi)) 1477 nextToken(); 1478 addUnwrappedLine(); 1479 } 1480 Line->Level = OldLineLevel; 1481 } 1482 1483 void UnwrappedLineParser::parseCaseLabel() { 1484 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1485 // FIXME: fix handling of complex expressions here. 1486 do { 1487 nextToken(); 1488 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1489 parseLabel(); 1490 } 1491 1492 void UnwrappedLineParser::parseSwitch() { 1493 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1494 nextToken(); 1495 if (FormatTok->Tok.is(tok::l_paren)) 1496 parseParens(); 1497 if (FormatTok->Tok.is(tok::l_brace)) { 1498 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1499 parseBlock(/*MustBeDeclaration=*/false); 1500 addUnwrappedLine(); 1501 } else { 1502 addUnwrappedLine(); 1503 ++Line->Level; 1504 parseStructuralElement(); 1505 --Line->Level; 1506 } 1507 } 1508 1509 void UnwrappedLineParser::parseAccessSpecifier() { 1510 nextToken(); 1511 // Understand Qt's slots. 1512 if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots)) 1513 nextToken(); 1514 // Otherwise, we don't know what it is, and we'd better keep the next token. 1515 if (FormatTok->Tok.is(tok::colon)) 1516 nextToken(); 1517 addUnwrappedLine(); 1518 } 1519 1520 void UnwrappedLineParser::parseEnum() { 1521 // Won't be 'enum' for NS_ENUMs. 1522 if (FormatTok->Tok.is(tok::kw_enum)) 1523 nextToken(); 1524 1525 // Eat up enum class ... 1526 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1527 nextToken(); 1528 1529 while (FormatTok->Tok.getIdentifierInfo() || 1530 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less, 1531 tok::greater, tok::comma, tok::question)) { 1532 nextToken(); 1533 // We can have macros or attributes in between 'enum' and the enum name. 1534 if (FormatTok->is(tok::l_paren)) 1535 parseParens(); 1536 if (FormatTok->is(tok::identifier)) { 1537 nextToken(); 1538 // If there are two identifiers in a row, this is likely an elaborate 1539 // return type. In Java, this can be "implements", etc. 1540 if (Style.Language == FormatStyle::LK_Cpp && 1541 FormatTok->is(tok::identifier)) 1542 return; 1543 } 1544 } 1545 1546 // Just a declaration or something is wrong. 1547 if (FormatTok->isNot(tok::l_brace)) 1548 return; 1549 FormatTok->BlockKind = BK_Block; 1550 1551 if (Style.Language == FormatStyle::LK_Java) { 1552 // Java enums are different. 1553 parseJavaEnumBody(); 1554 return; 1555 } 1556 1557 // Parse enum body. 1558 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1559 if (HasError) { 1560 if (FormatTok->is(tok::semi)) 1561 nextToken(); 1562 addUnwrappedLine(); 1563 } 1564 1565 // There is no addUnwrappedLine() here so that we fall through to parsing a 1566 // structural element afterwards. Thus, in "enum A {} n, m;", 1567 // "} n, m;" will end up in one unwrapped line. 1568 } 1569 1570 void UnwrappedLineParser::parseJavaEnumBody() { 1571 // Determine whether the enum is simple, i.e. does not have a semicolon or 1572 // constants with class bodies. Simple enums can be formatted like braced 1573 // lists, contracted to a single line, etc. 1574 unsigned StoredPosition = Tokens->getPosition(); 1575 bool IsSimple = true; 1576 FormatToken *Tok = Tokens->getNextToken(); 1577 while (Tok) { 1578 if (Tok->is(tok::r_brace)) 1579 break; 1580 if (Tok->isOneOf(tok::l_brace, tok::semi)) { 1581 IsSimple = false; 1582 break; 1583 } 1584 // FIXME: This will also mark enums with braces in the arguments to enum 1585 // constants as "not simple". This is probably fine in practice, though. 1586 Tok = Tokens->getNextToken(); 1587 } 1588 FormatTok = Tokens->setPosition(StoredPosition); 1589 1590 if (IsSimple) { 1591 parseBracedList(); 1592 addUnwrappedLine(); 1593 return; 1594 } 1595 1596 // Parse the body of a more complex enum. 1597 // First add a line for everything up to the "{". 1598 nextToken(); 1599 addUnwrappedLine(); 1600 ++Line->Level; 1601 1602 // Parse the enum constants. 1603 while (FormatTok) { 1604 if (FormatTok->is(tok::l_brace)) { 1605 // Parse the constant's class body. 1606 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1607 /*MunchSemi=*/false); 1608 } else if (FormatTok->is(tok::l_paren)) { 1609 parseParens(); 1610 } else if (FormatTok->is(tok::comma)) { 1611 nextToken(); 1612 addUnwrappedLine(); 1613 } else if (FormatTok->is(tok::semi)) { 1614 nextToken(); 1615 addUnwrappedLine(); 1616 break; 1617 } else if (FormatTok->is(tok::r_brace)) { 1618 addUnwrappedLine(); 1619 break; 1620 } else { 1621 nextToken(); 1622 } 1623 } 1624 1625 // Parse the class body after the enum's ";" if any. 1626 parseLevel(/*HasOpeningBrace=*/true); 1627 nextToken(); 1628 --Line->Level; 1629 addUnwrappedLine(); 1630 } 1631 1632 void UnwrappedLineParser::parseRecord() { 1633 const FormatToken &InitialToken = *FormatTok; 1634 nextToken(); 1635 1636 // The actual identifier can be a nested name specifier, and in macros 1637 // it is often token-pasted. 1638 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash, 1639 tok::kw___attribute, tok::kw___declspec, 1640 tok::kw_alignas) || 1641 ((Style.Language == FormatStyle::LK_Java || 1642 Style.Language == FormatStyle::LK_JavaScript) && 1643 FormatTok->isOneOf(tok::period, tok::comma))) { 1644 bool IsNonMacroIdentifier = 1645 FormatTok->is(tok::identifier) && 1646 FormatTok->TokenText != FormatTok->TokenText.upper(); 1647 nextToken(); 1648 // We can have macros or attributes in between 'class' and the class name. 1649 if (!IsNonMacroIdentifier && FormatTok->Tok.is(tok::l_paren)) 1650 parseParens(); 1651 } 1652 1653 // Note that parsing away template declarations here leads to incorrectly 1654 // accepting function declarations as record declarations. 1655 // In general, we cannot solve this problem. Consider: 1656 // class A<int> B() {} 1657 // which can be a function definition or a class definition when B() is a 1658 // macro. If we find enough real-world cases where this is a problem, we 1659 // can parse for the 'template' keyword in the beginning of the statement, 1660 // and thus rule out the record production in case there is no template 1661 // (this would still leave us with an ambiguity between template function 1662 // and class declarations). 1663 if (FormatTok->isOneOf(tok::colon, tok::less)) { 1664 while (!eof()) { 1665 if (FormatTok->is(tok::l_brace)) { 1666 calculateBraceTypes(/*ExpectClassBody=*/true); 1667 if (!tryToParseBracedList()) 1668 break; 1669 } 1670 if (FormatTok->Tok.is(tok::semi)) 1671 return; 1672 nextToken(); 1673 } 1674 } 1675 if (FormatTok->Tok.is(tok::l_brace)) { 1676 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1677 addUnwrappedLine(); 1678 1679 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1680 /*MunchSemi=*/false); 1681 } 1682 // There is no addUnwrappedLine() here so that we fall through to parsing a 1683 // structural element afterwards. Thus, in "class A {} n, m;", 1684 // "} n, m;" will end up in one unwrapped line. 1685 } 1686 1687 void UnwrappedLineParser::parseObjCProtocolList() { 1688 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1689 do 1690 nextToken(); 1691 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1692 nextToken(); // Skip '>'. 1693 } 1694 1695 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1696 do { 1697 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1698 nextToken(); 1699 addUnwrappedLine(); 1700 break; 1701 } 1702 if (FormatTok->is(tok::l_brace)) { 1703 parseBlock(/*MustBeDeclaration=*/false); 1704 // In ObjC interfaces, nothing should be following the "}". 1705 addUnwrappedLine(); 1706 } else if (FormatTok->is(tok::r_brace)) { 1707 // Ignore stray "}". parseStructuralElement doesn't consume them. 1708 nextToken(); 1709 addUnwrappedLine(); 1710 } else { 1711 parseStructuralElement(); 1712 } 1713 } while (!eof()); 1714 } 1715 1716 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1717 nextToken(); 1718 nextToken(); // interface name 1719 1720 // @interface can be followed by either a base class, or a category. 1721 if (FormatTok->Tok.is(tok::colon)) { 1722 nextToken(); 1723 nextToken(); // base class name 1724 } else if (FormatTok->Tok.is(tok::l_paren)) 1725 // Skip category, if present. 1726 parseParens(); 1727 1728 if (FormatTok->Tok.is(tok::less)) 1729 parseObjCProtocolList(); 1730 1731 if (FormatTok->Tok.is(tok::l_brace)) { 1732 if (Style.BreakBeforeBraces == FormatStyle::BS_Allman || 1733 Style.BreakBeforeBraces == FormatStyle::BS_GNU) 1734 addUnwrappedLine(); 1735 parseBlock(/*MustBeDeclaration=*/true); 1736 } 1737 1738 // With instance variables, this puts '}' on its own line. Without instance 1739 // variables, this ends the @interface line. 1740 addUnwrappedLine(); 1741 1742 parseObjCUntilAtEnd(); 1743 } 1744 1745 void UnwrappedLineParser::parseObjCProtocol() { 1746 nextToken(); 1747 nextToken(); // protocol name 1748 1749 if (FormatTok->Tok.is(tok::less)) 1750 parseObjCProtocolList(); 1751 1752 // Check for protocol declaration. 1753 if (FormatTok->Tok.is(tok::semi)) { 1754 nextToken(); 1755 return addUnwrappedLine(); 1756 } 1757 1758 addUnwrappedLine(); 1759 parseObjCUntilAtEnd(); 1760 } 1761 1762 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { 1763 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export)); 1764 nextToken(); 1765 1766 // Consume the "default" in "export default class/function". 1767 if (FormatTok->is(tok::kw_default)) 1768 nextToken(); 1769 1770 // Consume "function" and "default function", so that these get parsed as 1771 // free-standing JS functions, i.e. do not require a trailing semicolon. 1772 if (FormatTok->is(Keywords.kw_function)) { 1773 nextToken(); 1774 return; 1775 } 1776 1777 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, tok::kw_enum, 1778 Keywords.kw_var)) 1779 return; // Fall through to parsing the corresponding structure. 1780 1781 if (FormatTok->is(tok::l_brace)) { 1782 FormatTok->BlockKind = BK_Block; 1783 parseBracedList(); 1784 } 1785 1786 while (!eof() && FormatTok->isNot(tok::semi) && 1787 FormatTok->isNot(tok::l_brace)) { 1788 nextToken(); 1789 } 1790 } 1791 1792 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1793 StringRef Prefix = "") { 1794 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1795 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1796 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1797 E = Line.Tokens.end(); 1798 I != E; ++I) { 1799 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1800 } 1801 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1802 E = Line.Tokens.end(); 1803 I != E; ++I) { 1804 const UnwrappedLineNode &Node = *I; 1805 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1806 I = Node.Children.begin(), 1807 E = Node.Children.end(); 1808 I != E; ++I) { 1809 printDebugInfo(*I, "\nChild: "); 1810 } 1811 } 1812 llvm::dbgs() << "\n"; 1813 } 1814 1815 void UnwrappedLineParser::addUnwrappedLine() { 1816 if (Line->Tokens.empty()) 1817 return; 1818 DEBUG({ 1819 if (CurrentLines == &Lines) 1820 printDebugInfo(*Line); 1821 }); 1822 CurrentLines->push_back(std::move(*Line)); 1823 Line->Tokens.clear(); 1824 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1825 CurrentLines->append( 1826 std::make_move_iterator(PreprocessorDirectives.begin()), 1827 std::make_move_iterator(PreprocessorDirectives.end())); 1828 PreprocessorDirectives.clear(); 1829 } 1830 } 1831 1832 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1833 1834 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1835 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1836 FormatTok.NewlinesBefore > 0; 1837 } 1838 1839 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1840 bool JustComments = Line->Tokens.empty(); 1841 for (SmallVectorImpl<FormatToken *>::const_iterator 1842 I = CommentsBeforeNextToken.begin(), 1843 E = CommentsBeforeNextToken.end(); 1844 I != E; ++I) { 1845 if (isOnNewLine(**I) && JustComments) 1846 addUnwrappedLine(); 1847 pushToken(*I); 1848 } 1849 if (NewlineBeforeNext && JustComments) 1850 addUnwrappedLine(); 1851 CommentsBeforeNextToken.clear(); 1852 } 1853 1854 void UnwrappedLineParser::nextToken() { 1855 if (eof()) 1856 return; 1857 flushComments(isOnNewLine(*FormatTok)); 1858 pushToken(FormatTok); 1859 readToken(); 1860 } 1861 1862 void UnwrappedLineParser::readToken() { 1863 bool CommentsInCurrentLine = true; 1864 do { 1865 FormatTok = Tokens->getNextToken(); 1866 assert(FormatTok); 1867 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1868 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1869 // If there is an unfinished unwrapped line, we flush the preprocessor 1870 // directives only after that unwrapped line was finished later. 1871 bool SwitchToPreprocessorLines = !Line->Tokens.empty(); 1872 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1873 // Comments stored before the preprocessor directive need to be output 1874 // before the preprocessor directive, at the same level as the 1875 // preprocessor directive, as we consider them to apply to the directive. 1876 flushComments(isOnNewLine(*FormatTok)); 1877 parsePPDirective(); 1878 } 1879 while (FormatTok->Type == TT_ConflictStart || 1880 FormatTok->Type == TT_ConflictEnd || 1881 FormatTok->Type == TT_ConflictAlternative) { 1882 if (FormatTok->Type == TT_ConflictStart) { 1883 conditionalCompilationStart(/*Unreachable=*/false); 1884 } else if (FormatTok->Type == TT_ConflictAlternative) { 1885 conditionalCompilationAlternative(); 1886 } else if (FormatTok->Type == TT_ConflictEnd) { 1887 conditionalCompilationEnd(); 1888 } 1889 FormatTok = Tokens->getNextToken(); 1890 FormatTok->MustBreakBefore = true; 1891 } 1892 1893 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1894 !Line->InPPDirective) { 1895 continue; 1896 } 1897 1898 if (!FormatTok->Tok.is(tok::comment)) 1899 return; 1900 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1901 CommentsInCurrentLine = false; 1902 } 1903 if (CommentsInCurrentLine) { 1904 pushToken(FormatTok); 1905 } else { 1906 CommentsBeforeNextToken.push_back(FormatTok); 1907 } 1908 } while (!eof()); 1909 } 1910 1911 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1912 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1913 if (MustBreakBeforeNextToken) { 1914 Line->Tokens.back().Tok->MustBreakBefore = true; 1915 MustBreakBeforeNextToken = false; 1916 } 1917 } 1918 1919 } // end namespace format 1920 } // end namespace clang 1921