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.BraceWrapping.AfterControlStatement) 158 Parser->addUnwrappedLine(); 159 if (Style.BraceWrapping.IndentBraces) 160 ++LineLevel; 161 } 162 ~CompoundStatementIndenter() { LineLevel = OldLineLevel; } 163 164 private: 165 unsigned &LineLevel; 166 unsigned OldLineLevel; 167 }; 168 169 namespace { 170 171 class IndexedTokenSource : public FormatTokenSource { 172 public: 173 IndexedTokenSource(ArrayRef<FormatToken *> Tokens) 174 : Tokens(Tokens), Position(-1) {} 175 176 FormatToken *getNextToken() override { 177 ++Position; 178 return Tokens[Position]; 179 } 180 181 unsigned getPosition() override { 182 assert(Position >= 0); 183 return Position; 184 } 185 186 FormatToken *setPosition(unsigned P) override { 187 Position = P; 188 return Tokens[Position]; 189 } 190 191 void reset() { Position = -1; } 192 193 private: 194 ArrayRef<FormatToken *> Tokens; 195 int Position; 196 }; 197 198 } // end anonymous namespace 199 200 UnwrappedLineParser::UnwrappedLineParser(const FormatStyle &Style, 201 const AdditionalKeywords &Keywords, 202 ArrayRef<FormatToken *> Tokens, 203 UnwrappedLineConsumer &Callback) 204 : Line(new UnwrappedLine), MustBreakBeforeNextToken(false), 205 CurrentLines(&Lines), Style(Style), Keywords(Keywords), Tokens(nullptr), 206 Callback(Callback), AllTokens(Tokens), PPBranchLevel(-1) {} 207 208 void UnwrappedLineParser::reset() { 209 PPBranchLevel = -1; 210 Line.reset(new UnwrappedLine); 211 CommentsBeforeNextToken.clear(); 212 FormatTok = nullptr; 213 MustBreakBeforeNextToken = false; 214 PreprocessorDirectives.clear(); 215 CurrentLines = &Lines; 216 DeclarationScopeStack.clear(); 217 PPStack.clear(); 218 } 219 220 void UnwrappedLineParser::parse() { 221 IndexedTokenSource TokenSource(AllTokens); 222 do { 223 DEBUG(llvm::dbgs() << "----\n"); 224 reset(); 225 Tokens = &TokenSource; 226 TokenSource.reset(); 227 228 readToken(); 229 parseFile(); 230 // Create line with eof token. 231 pushToken(FormatTok); 232 addUnwrappedLine(); 233 234 for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(), 235 E = Lines.end(); 236 I != E; ++I) { 237 Callback.consumeUnwrappedLine(*I); 238 } 239 Callback.finishRun(); 240 Lines.clear(); 241 while (!PPLevelBranchIndex.empty() && 242 PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) { 243 PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1); 244 PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1); 245 } 246 if (!PPLevelBranchIndex.empty()) { 247 ++PPLevelBranchIndex.back(); 248 assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size()); 249 assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back()); 250 } 251 } while (!PPLevelBranchIndex.empty()); 252 } 253 254 void UnwrappedLineParser::parseFile() { 255 // The top-level context in a file always has declarations, except for pre- 256 // processor directives and JavaScript files. 257 bool MustBeDeclaration = 258 !Line->InPPDirective && Style.Language != FormatStyle::LK_JavaScript; 259 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 260 MustBeDeclaration); 261 parseLevel(/*HasOpeningBrace=*/false); 262 // Make sure to format the remaining tokens. 263 flushComments(true); 264 addUnwrappedLine(); 265 } 266 267 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) { 268 bool SwitchLabelEncountered = false; 269 do { 270 tok::TokenKind kind = FormatTok->Tok.getKind(); 271 if (FormatTok->Type == TT_MacroBlockBegin) { 272 kind = tok::l_brace; 273 } else if (FormatTok->Type == TT_MacroBlockEnd) { 274 kind = tok::r_brace; 275 } 276 277 switch (kind) { 278 case tok::comment: 279 nextToken(); 280 addUnwrappedLine(); 281 break; 282 case tok::l_brace: 283 // FIXME: Add parameter whether this can happen - if this happens, we must 284 // be in a non-declaration context. 285 if (!FormatTok->is(TT_MacroBlockBegin) && tryToParseBracedList()) 286 continue; 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 const FormatToken *PrevTok = getPreviousToken(); 319 // Keep a stack of positions of lbrace tokens. We will 320 // update information about whether an lbrace starts a 321 // braced init list or a different block during the loop. 322 SmallVector<FormatToken *, 8> LBraceStack; 323 assert(Tok->Tok.is(tok::l_brace)); 324 do { 325 // Get next non-comment token. 326 FormatToken *NextTok; 327 unsigned ReadTokens = 0; 328 do { 329 NextTok = Tokens->getNextToken(); 330 ++ReadTokens; 331 } while (NextTok->is(tok::comment)); 332 333 switch (Tok->Tok.getKind()) { 334 case tok::l_brace: 335 if (Style.Language == FormatStyle::LK_JavaScript && PrevTok && 336 PrevTok->is(tok::colon)) 337 // In TypeScript's TypeMemberLists, there can be semicolons between the 338 // individual members. 339 Tok->BlockKind = BK_BracedInit; 340 else 341 Tok->BlockKind = BK_Unknown; 342 LBraceStack.push_back(Tok); 343 break; 344 case tok::r_brace: 345 if (LBraceStack.empty()) 346 break; 347 if (LBraceStack.back()->BlockKind == BK_Unknown) { 348 bool ProbablyBracedList = false; 349 if (Style.Language == FormatStyle::LK_Proto) { 350 ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square); 351 } else { 352 // Using OriginalColumn to distinguish between ObjC methods and 353 // binary operators is a bit hacky. 354 bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) && 355 NextTok->OriginalColumn == 0; 356 357 // If there is a comma, semicolon or right paren after the closing 358 // brace, we assume this is a braced initializer list. Note that 359 // regardless how we mark inner braces here, we will overwrite the 360 // BlockKind later if we parse a braced list (where all blocks 361 // inside are by default braced lists), or when we explicitly detect 362 // blocks (for example while parsing lambdas). 363 // 364 // We exclude + and - as they can be ObjC visibility modifiers. 365 ProbablyBracedList = 366 (Style.Language == FormatStyle::LK_JavaScript && 367 NextTok->isOneOf(Keywords.kw_of, Keywords.kw_in)) || 368 NextTok->isOneOf(tok::comma, tok::period, tok::colon, 369 tok::r_paren, tok::r_square, tok::l_brace, 370 tok::l_square, tok::l_paren, tok::ellipsis) || 371 (NextTok->is(tok::semi) && 372 (!ExpectClassBody || LBraceStack.size() != 1)) || 373 (NextTok->isBinaryOperator() && !NextIsObjCMethod); 374 } 375 if (ProbablyBracedList) { 376 Tok->BlockKind = BK_BracedInit; 377 LBraceStack.back()->BlockKind = BK_BracedInit; 378 } else { 379 Tok->BlockKind = BK_Block; 380 LBraceStack.back()->BlockKind = BK_Block; 381 } 382 } 383 LBraceStack.pop_back(); 384 break; 385 case tok::at: 386 case tok::semi: 387 case tok::kw_if: 388 case tok::kw_while: 389 case tok::kw_for: 390 case tok::kw_switch: 391 case tok::kw_try: 392 case tok::kw___try: 393 if (!LBraceStack.empty() && LBraceStack.back()->BlockKind == BK_Unknown) 394 LBraceStack.back()->BlockKind = BK_Block; 395 break; 396 default: 397 break; 398 } 399 PrevTok = Tok; 400 Tok = NextTok; 401 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty()); 402 403 // Assume other blocks for all unclosed opening braces. 404 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) { 405 if (LBraceStack[i]->BlockKind == BK_Unknown) 406 LBraceStack[i]->BlockKind = BK_Block; 407 } 408 409 FormatTok = Tokens->setPosition(StoredPosition); 410 } 411 412 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel, 413 bool MunchSemi) { 414 assert(FormatTok->isOneOf(tok::l_brace, TT_MacroBlockBegin) && 415 "'{' or macro block token expected"); 416 const bool MacroBlock = FormatTok->is(TT_MacroBlockBegin); 417 FormatTok->BlockKind = BK_Block; 418 419 unsigned InitialLevel = Line->Level; 420 nextToken(); 421 422 if (MacroBlock && FormatTok->is(tok::l_paren)) 423 parseParens(); 424 425 addUnwrappedLine(); 426 427 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 428 MustBeDeclaration); 429 if (AddLevel) 430 ++Line->Level; 431 parseLevel(/*HasOpeningBrace=*/true); 432 433 if (eof()) 434 return; 435 436 if (MacroBlock ? !FormatTok->is(TT_MacroBlockEnd) 437 : !FormatTok->is(tok::r_brace)) { 438 Line->Level = InitialLevel; 439 FormatTok->BlockKind = BK_Block; 440 return; 441 } 442 443 nextToken(); // Munch the closing brace. 444 445 if (MacroBlock && FormatTok->is(tok::l_paren)) 446 parseParens(); 447 448 if (MunchSemi && FormatTok->Tok.is(tok::semi)) 449 nextToken(); 450 Line->Level = InitialLevel; 451 } 452 453 static bool isGoogScope(const UnwrappedLine &Line) { 454 // FIXME: Closure-library specific stuff should not be hard-coded but be 455 // configurable. 456 if (Line.Tokens.size() < 4) 457 return false; 458 auto I = Line.Tokens.begin(); 459 if (I->Tok->TokenText != "goog") 460 return false; 461 ++I; 462 if (I->Tok->isNot(tok::period)) 463 return false; 464 ++I; 465 if (I->Tok->TokenText != "scope") 466 return false; 467 ++I; 468 return I->Tok->is(tok::l_paren); 469 } 470 471 static bool ShouldBreakBeforeBrace(const FormatStyle &Style, 472 const FormatToken &InitialToken) { 473 if (InitialToken.is(tok::kw_namespace)) 474 return Style.BraceWrapping.AfterNamespace; 475 if (InitialToken.is(tok::kw_class)) 476 return Style.BraceWrapping.AfterClass; 477 if (InitialToken.is(tok::kw_union)) 478 return Style.BraceWrapping.AfterUnion; 479 if (InitialToken.is(tok::kw_struct)) 480 return Style.BraceWrapping.AfterStruct; 481 return false; 482 } 483 484 void UnwrappedLineParser::parseChildBlock() { 485 FormatTok->BlockKind = BK_Block; 486 nextToken(); 487 { 488 bool GoogScope = 489 Style.Language == FormatStyle::LK_JavaScript && isGoogScope(*Line); 490 ScopedLineState LineState(*this); 491 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack, 492 /*MustBeDeclaration=*/false); 493 Line->Level += GoogScope ? 0 : 1; 494 parseLevel(/*HasOpeningBrace=*/true); 495 flushComments(isOnNewLine(*FormatTok)); 496 Line->Level -= GoogScope ? 0 : 1; 497 } 498 nextToken(); 499 } 500 501 void UnwrappedLineParser::parsePPDirective() { 502 assert(FormatTok->Tok.is(tok::hash) && "'#' expected"); 503 ScopedMacroState MacroState(*Line, Tokens, FormatTok); 504 nextToken(); 505 506 if (!FormatTok->Tok.getIdentifierInfo()) { 507 parsePPUnknown(); 508 return; 509 } 510 511 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) { 512 case tok::pp_define: 513 parsePPDefine(); 514 return; 515 case tok::pp_if: 516 parsePPIf(/*IfDef=*/false); 517 break; 518 case tok::pp_ifdef: 519 case tok::pp_ifndef: 520 parsePPIf(/*IfDef=*/true); 521 break; 522 case tok::pp_else: 523 parsePPElse(); 524 break; 525 case tok::pp_elif: 526 parsePPElIf(); 527 break; 528 case tok::pp_endif: 529 parsePPEndIf(); 530 break; 531 default: 532 parsePPUnknown(); 533 break; 534 } 535 } 536 537 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) { 538 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable)) 539 PPStack.push_back(PP_Unreachable); 540 else 541 PPStack.push_back(PP_Conditional); 542 } 543 544 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) { 545 ++PPBranchLevel; 546 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size()); 547 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) { 548 PPLevelBranchIndex.push_back(0); 549 PPLevelBranchCount.push_back(0); 550 } 551 PPChainBranchIndex.push(0); 552 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0; 553 conditionalCompilationCondition(Unreachable || Skip); 554 } 555 556 void UnwrappedLineParser::conditionalCompilationAlternative() { 557 if (!PPStack.empty()) 558 PPStack.pop_back(); 559 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 560 if (!PPChainBranchIndex.empty()) 561 ++PPChainBranchIndex.top(); 562 conditionalCompilationCondition( 563 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() && 564 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top()); 565 } 566 567 void UnwrappedLineParser::conditionalCompilationEnd() { 568 assert(PPBranchLevel < (int)PPLevelBranchIndex.size()); 569 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) { 570 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) { 571 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1; 572 } 573 } 574 // Guard against #endif's without #if. 575 if (PPBranchLevel > 0) 576 --PPBranchLevel; 577 if (!PPChainBranchIndex.empty()) 578 PPChainBranchIndex.pop(); 579 if (!PPStack.empty()) 580 PPStack.pop_back(); 581 } 582 583 void UnwrappedLineParser::parsePPIf(bool IfDef) { 584 nextToken(); 585 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() && 586 FormatTok->Tok.getLiteralData() != nullptr && 587 StringRef(FormatTok->Tok.getLiteralData(), 588 FormatTok->Tok.getLength()) == "0") || 589 FormatTok->Tok.is(tok::kw_false); 590 conditionalCompilationStart(!IfDef && IsLiteralFalse); 591 parsePPUnknown(); 592 } 593 594 void UnwrappedLineParser::parsePPElse() { 595 conditionalCompilationAlternative(); 596 parsePPUnknown(); 597 } 598 599 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); } 600 601 void UnwrappedLineParser::parsePPEndIf() { 602 conditionalCompilationEnd(); 603 parsePPUnknown(); 604 } 605 606 void UnwrappedLineParser::parsePPDefine() { 607 nextToken(); 608 609 if (FormatTok->Tok.getKind() != tok::identifier) { 610 parsePPUnknown(); 611 return; 612 } 613 nextToken(); 614 if (FormatTok->Tok.getKind() == tok::l_paren && 615 FormatTok->WhitespaceRange.getBegin() == 616 FormatTok->WhitespaceRange.getEnd()) { 617 parseParens(); 618 } 619 addUnwrappedLine(); 620 Line->Level = 1; 621 622 // Errors during a preprocessor directive can only affect the layout of the 623 // preprocessor directive, and thus we ignore them. An alternative approach 624 // would be to use the same approach we use on the file level (no 625 // re-indentation if there was a structural error) within the macro 626 // definition. 627 parseFile(); 628 } 629 630 void UnwrappedLineParser::parsePPUnknown() { 631 do { 632 nextToken(); 633 } while (!eof()); 634 addUnwrappedLine(); 635 } 636 637 // Here we blacklist certain tokens that are not usually the first token in an 638 // unwrapped line. This is used in attempt to distinguish macro calls without 639 // trailing semicolons from other constructs split to several lines. 640 static bool tokenCanStartNewLine(const clang::Token &Tok) { 641 // Semicolon can be a null-statement, l_square can be a start of a macro or 642 // a C++11 attribute, but this doesn't seem to be common. 643 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) && 644 Tok.isNot(tok::l_square) && 645 // Tokens that can only be used as binary operators and a part of 646 // overloaded operator names. 647 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) && 648 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) && 649 Tok.isNot(tok::less) && Tok.isNot(tok::greater) && 650 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) && 651 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) && 652 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) && 653 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) && 654 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) && 655 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) && 656 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) && 657 Tok.isNot(tok::lesslessequal) && 658 // Colon is used in labels, base class lists, initializer lists, 659 // range-based for loops, ternary operator, but should never be the 660 // first token in an unwrapped line. 661 Tok.isNot(tok::colon) && 662 // 'noexcept' is a trailing annotation. 663 Tok.isNot(tok::kw_noexcept); 664 } 665 666 static bool mustBeJSIdentOrValue(const AdditionalKeywords &Keywords, 667 const FormatToken *FormatTok) { 668 if (FormatTok->Tok.isLiteral()) 669 return true; 670 // FIXME: This returns true for C/C++ keywords like 'struct'. 671 return FormatTok->is(tok::identifier) && 672 (FormatTok->Tok.getIdentifierInfo() == nullptr || 673 !FormatTok->isOneOf(Keywords.kw_in, Keywords.kw_of, 674 Keywords.kw_finally, Keywords.kw_function, 675 Keywords.kw_import, Keywords.kw_is, 676 Keywords.kw_let, Keywords.kw_var, 677 Keywords.kw_abstract, Keywords.kw_extends, 678 Keywords.kw_implements, Keywords.kw_instanceof, 679 Keywords.kw_interface, Keywords.kw_throws)); 680 } 681 682 // isJSDeclOrStmt returns true if |FormatTok| starts a declaration or statement 683 // when encountered after a value (see mustBeJSIdentOrValue). 684 static bool isJSDeclOrStmt(const AdditionalKeywords &Keywords, 685 const FormatToken *FormatTok) { 686 return FormatTok->isOneOf( 687 tok::kw_return, 688 // conditionals 689 tok::kw_if, tok::kw_else, 690 // loops 691 tok::kw_for, tok::kw_while, tok::kw_do, tok::kw_continue, tok::kw_break, 692 // switch/case 693 tok::kw_switch, tok::kw_case, 694 // exceptions 695 tok::kw_throw, tok::kw_try, tok::kw_catch, Keywords.kw_finally, 696 // declaration 697 tok::kw_const, tok::kw_class, Keywords.kw_var, Keywords.kw_let, 698 Keywords.kw_function); 699 } 700 701 // readTokenWithJavaScriptASI reads the next token and terminates the current 702 // line if JavaScript Automatic Semicolon Insertion must 703 // happen between the current token and the next token. 704 // 705 // This method is conservative - it cannot cover all edge cases of JavaScript, 706 // but only aims to correctly handle certain well known cases. It *must not* 707 // return true in speculative cases. 708 void UnwrappedLineParser::readTokenWithJavaScriptASI() { 709 FormatToken *Previous = FormatTok; 710 readToken(); 711 FormatToken *Next = FormatTok; 712 713 bool IsOnSameLine = 714 CommentsBeforeNextToken.empty() 715 ? Next->NewlinesBefore == 0 716 : CommentsBeforeNextToken.front()->NewlinesBefore == 0; 717 if (IsOnSameLine) 718 return; 719 720 bool PreviousMustBeValue = mustBeJSIdentOrValue(Keywords, Previous); 721 if (PreviousMustBeValue && Line && Line->Tokens.size() > 1) { 722 // If the token before the previous one is an '@', the previous token is an 723 // annotation and can precede another identifier/value. 724 const FormatToken *PrePrevious = std::prev(Line->Tokens.end(), 2)->Tok; 725 if (PrePrevious->is(tok::at)) 726 return; 727 } 728 if (Next->is(tok::exclaim) && PreviousMustBeValue) 729 addUnwrappedLine(); 730 bool NextMustBeValue = mustBeJSIdentOrValue(Keywords, Next); 731 if (NextMustBeValue && (PreviousMustBeValue || 732 Previous->isOneOf(tok::r_square, tok::r_paren, 733 tok::plusplus, tok::minusminus))) 734 addUnwrappedLine(); 735 if (PreviousMustBeValue && isJSDeclOrStmt(Keywords, Next)) 736 addUnwrappedLine(); 737 } 738 739 void UnwrappedLineParser::parseStructuralElement() { 740 assert(!FormatTok->is(tok::l_brace)); 741 if (Style.Language == FormatStyle::LK_TableGen && 742 FormatTok->is(tok::pp_include)) { 743 nextToken(); 744 if (FormatTok->is(tok::string_literal)) 745 nextToken(); 746 addUnwrappedLine(); 747 return; 748 } 749 switch (FormatTok->Tok.getKind()) { 750 case tok::at: 751 nextToken(); 752 if (FormatTok->Tok.is(tok::l_brace)) { 753 parseBracedList(); 754 break; 755 } 756 switch (FormatTok->Tok.getObjCKeywordID()) { 757 case tok::objc_public: 758 case tok::objc_protected: 759 case tok::objc_package: 760 case tok::objc_private: 761 return parseAccessSpecifier(); 762 case tok::objc_interface: 763 case tok::objc_implementation: 764 return parseObjCInterfaceOrImplementation(); 765 case tok::objc_protocol: 766 return parseObjCProtocol(); 767 case tok::objc_end: 768 return; // Handled by the caller. 769 case tok::objc_optional: 770 case tok::objc_required: 771 nextToken(); 772 addUnwrappedLine(); 773 return; 774 case tok::objc_autoreleasepool: 775 nextToken(); 776 if (FormatTok->Tok.is(tok::l_brace)) { 777 if (Style.BraceWrapping.AfterObjCDeclaration) 778 addUnwrappedLine(); 779 parseBlock(/*MustBeDeclaration=*/false); 780 } 781 addUnwrappedLine(); 782 return; 783 case tok::objc_try: 784 // This branch isn't strictly necessary (the kw_try case below would 785 // do this too after the tok::at is parsed above). But be explicit. 786 parseTryCatch(); 787 return; 788 default: 789 break; 790 } 791 break; 792 case tok::kw_asm: 793 nextToken(); 794 if (FormatTok->is(tok::l_brace)) { 795 FormatTok->Type = TT_InlineASMBrace; 796 nextToken(); 797 while (FormatTok && FormatTok->isNot(tok::eof)) { 798 if (FormatTok->is(tok::r_brace)) { 799 FormatTok->Type = TT_InlineASMBrace; 800 nextToken(); 801 addUnwrappedLine(); 802 break; 803 } 804 FormatTok->Finalized = true; 805 nextToken(); 806 } 807 } 808 break; 809 case tok::kw_namespace: 810 parseNamespace(); 811 return; 812 case tok::kw_inline: 813 nextToken(); 814 if (FormatTok->Tok.is(tok::kw_namespace)) { 815 parseNamespace(); 816 return; 817 } 818 break; 819 case tok::kw_public: 820 case tok::kw_protected: 821 case tok::kw_private: 822 if (Style.Language == FormatStyle::LK_Java || 823 Style.Language == FormatStyle::LK_JavaScript) 824 nextToken(); 825 else 826 parseAccessSpecifier(); 827 return; 828 case tok::kw_if: 829 parseIfThenElse(); 830 return; 831 case tok::kw_for: 832 case tok::kw_while: 833 parseForOrWhileLoop(); 834 return; 835 case tok::kw_do: 836 parseDoWhile(); 837 return; 838 case tok::kw_switch: 839 parseSwitch(); 840 return; 841 case tok::kw_default: 842 nextToken(); 843 parseLabel(); 844 return; 845 case tok::kw_case: 846 parseCaseLabel(); 847 return; 848 case tok::kw_try: 849 case tok::kw___try: 850 parseTryCatch(); 851 return; 852 case tok::kw_extern: 853 nextToken(); 854 if (FormatTok->Tok.is(tok::string_literal)) { 855 nextToken(); 856 if (FormatTok->Tok.is(tok::l_brace)) { 857 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 858 addUnwrappedLine(); 859 return; 860 } 861 } 862 break; 863 case tok::kw_export: 864 if (Style.Language == FormatStyle::LK_JavaScript) { 865 parseJavaScriptEs6ImportExport(); 866 return; 867 } 868 break; 869 case tok::identifier: 870 if (FormatTok->is(TT_ForEachMacro)) { 871 parseForOrWhileLoop(); 872 return; 873 } 874 if (FormatTok->is(TT_MacroBlockBegin)) { 875 parseBlock(/*MustBeDeclaration=*/false, /*AddLevel=*/true, 876 /*MunchSemi=*/false); 877 return; 878 } 879 if (Style.Language == FormatStyle::LK_JavaScript && 880 FormatTok->is(Keywords.kw_import)) { 881 parseJavaScriptEs6ImportExport(); 882 return; 883 } 884 if (FormatTok->isOneOf(Keywords.kw_signals, Keywords.kw_qsignals, 885 Keywords.kw_slots, Keywords.kw_qslots)) { 886 nextToken(); 887 if (FormatTok->is(tok::colon)) { 888 nextToken(); 889 addUnwrappedLine(); 890 } 891 return; 892 } 893 // In all other cases, parse the declaration. 894 break; 895 default: 896 break; 897 } 898 do { 899 switch (FormatTok->Tok.getKind()) { 900 case tok::at: 901 nextToken(); 902 if (FormatTok->Tok.is(tok::l_brace)) 903 parseBracedList(); 904 break; 905 case tok::kw_enum: 906 // parseEnum falls through and does not yet add an unwrapped line as an 907 // enum definition can start a structural element. 908 if (!parseEnum()) 909 break; 910 // This only applies for C++. 911 if (Style.Language != FormatStyle::LK_Cpp) { 912 addUnwrappedLine(); 913 return; 914 } 915 break; 916 case tok::kw_typedef: 917 nextToken(); 918 if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS, 919 Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS)) 920 parseEnum(); 921 break; 922 case tok::kw_struct: 923 case tok::kw_union: 924 case tok::kw_class: 925 // parseRecord falls through and does not yet add an unwrapped line as a 926 // record declaration or definition can start a structural element. 927 parseRecord(); 928 // This does not apply for Java and JavaScript. 929 if (Style.Language == FormatStyle::LK_Java || 930 Style.Language == FormatStyle::LK_JavaScript) { 931 if (FormatTok->is(tok::semi)) 932 nextToken(); 933 addUnwrappedLine(); 934 return; 935 } 936 break; 937 case tok::period: 938 nextToken(); 939 // In Java, classes have an implicit static member "class". 940 if (Style.Language == FormatStyle::LK_Java && FormatTok && 941 FormatTok->is(tok::kw_class)) 942 nextToken(); 943 if (Style.Language == FormatStyle::LK_JavaScript && FormatTok && 944 FormatTok->Tok.getIdentifierInfo()) 945 // JavaScript only has pseudo keywords, all keywords are allowed to 946 // appear in "IdentifierName" positions. See http://es5.github.io/#x7.6 947 nextToken(); 948 break; 949 case tok::semi: 950 nextToken(); 951 addUnwrappedLine(); 952 return; 953 case tok::r_brace: 954 addUnwrappedLine(); 955 return; 956 case tok::l_paren: 957 parseParens(); 958 break; 959 case tok::kw_operator: 960 nextToken(); 961 if (FormatTok->isBinaryOperator()) 962 nextToken(); 963 break; 964 case tok::caret: 965 nextToken(); 966 if (FormatTok->Tok.isAnyIdentifier() || 967 FormatTok->isSimpleTypeSpecifier()) 968 nextToken(); 969 if (FormatTok->is(tok::l_paren)) 970 parseParens(); 971 if (FormatTok->is(tok::l_brace)) 972 parseChildBlock(); 973 break; 974 case tok::l_brace: 975 if (!tryToParseBracedList()) { 976 // A block outside of parentheses must be the last part of a 977 // structural element. 978 // FIXME: Figure out cases where this is not true, and add projections 979 // for them (the one we know is missing are lambdas). 980 if (Style.BraceWrapping.AfterFunction) 981 addUnwrappedLine(); 982 FormatTok->Type = TT_FunctionLBrace; 983 parseBlock(/*MustBeDeclaration=*/false); 984 addUnwrappedLine(); 985 return; 986 } 987 // Otherwise this was a braced init list, and the structural 988 // element continues. 989 break; 990 case tok::kw_try: 991 // We arrive here when parsing function-try blocks. 992 parseTryCatch(); 993 return; 994 case tok::identifier: { 995 if (FormatTok->is(TT_MacroBlockEnd)) { 996 addUnwrappedLine(); 997 return; 998 } 999 1000 // Parse function literal unless 'function' is the first token in a line 1001 // in which case this should be treated as a free-standing function. 1002 if (Style.Language == FormatStyle::LK_JavaScript && 1003 FormatTok->is(Keywords.kw_function) && Line->Tokens.size() > 0) { 1004 tryToParseJSFunction(); 1005 break; 1006 } 1007 if ((Style.Language == FormatStyle::LK_JavaScript || 1008 Style.Language == FormatStyle::LK_Java) && 1009 FormatTok->is(Keywords.kw_interface)) { 1010 parseRecord(); 1011 addUnwrappedLine(); 1012 return; 1013 } 1014 1015 // See if the following token should start a new unwrapped line. 1016 StringRef Text = FormatTok->TokenText; 1017 nextToken(); 1018 if (Line->Tokens.size() == 1 && 1019 // JS doesn't have macros, and within classes colons indicate fields, 1020 // not labels. 1021 Style.Language != FormatStyle::LK_JavaScript) { 1022 if (FormatTok->Tok.is(tok::colon) && !Line->MustBeDeclaration) { 1023 Line->Tokens.begin()->Tok->MustBreakBefore = true; 1024 parseLabel(); 1025 return; 1026 } 1027 // Recognize function-like macro usages without trailing semicolon as 1028 // well as free-standing macros like Q_OBJECT. 1029 bool FunctionLike = FormatTok->is(tok::l_paren); 1030 if (FunctionLike) 1031 parseParens(); 1032 1033 bool FollowedByNewline = 1034 CommentsBeforeNextToken.empty() 1035 ? FormatTok->NewlinesBefore > 0 1036 : CommentsBeforeNextToken.front()->NewlinesBefore > 0; 1037 1038 if (FollowedByNewline && (Text.size() >= 5 || FunctionLike) && 1039 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 1040 addUnwrappedLine(); 1041 return; 1042 } 1043 } 1044 break; 1045 } 1046 case tok::equal: 1047 // Fat arrows (=>) have tok::TokenKind tok::equal but TokenType 1048 // TT_JsFatArrow. The always start an expression or a child block if 1049 // followed by a curly. 1050 if (FormatTok->is(TT_JsFatArrow)) { 1051 nextToken(); 1052 if (FormatTok->is(tok::l_brace)) 1053 parseChildBlock(); 1054 break; 1055 } 1056 1057 nextToken(); 1058 if (FormatTok->Tok.is(tok::l_brace)) { 1059 parseBracedList(); 1060 } 1061 break; 1062 case tok::l_square: 1063 parseSquare(); 1064 break; 1065 case tok::kw_new: 1066 parseNew(); 1067 break; 1068 default: 1069 nextToken(); 1070 break; 1071 } 1072 } while (!eof()); 1073 } 1074 1075 bool UnwrappedLineParser::tryToParseLambda() { 1076 if (Style.Language != FormatStyle::LK_Cpp) { 1077 nextToken(); 1078 return false; 1079 } 1080 const FormatToken* Previous = getPreviousToken(); 1081 if (Previous && 1082 (Previous->isOneOf(tok::identifier, tok::kw_operator, tok::kw_new, 1083 tok::kw_delete) || 1084 Previous->closesScope() || Previous->isSimpleTypeSpecifier())) { 1085 nextToken(); 1086 return false; 1087 } 1088 assert(FormatTok->is(tok::l_square)); 1089 FormatToken &LSquare = *FormatTok; 1090 if (!tryToParseLambdaIntroducer()) 1091 return false; 1092 1093 while (FormatTok->isNot(tok::l_brace)) { 1094 if (FormatTok->isSimpleTypeSpecifier()) { 1095 nextToken(); 1096 continue; 1097 } 1098 switch (FormatTok->Tok.getKind()) { 1099 case tok::l_brace: 1100 break; 1101 case tok::l_paren: 1102 parseParens(); 1103 break; 1104 case tok::amp: 1105 case tok::star: 1106 case tok::kw_const: 1107 case tok::comma: 1108 case tok::less: 1109 case tok::greater: 1110 case tok::identifier: 1111 case tok::numeric_constant: 1112 case tok::coloncolon: 1113 case tok::kw_mutable: 1114 nextToken(); 1115 break; 1116 case tok::arrow: 1117 FormatTok->Type = TT_LambdaArrow; 1118 nextToken(); 1119 break; 1120 default: 1121 return true; 1122 } 1123 } 1124 LSquare.Type = TT_LambdaLSquare; 1125 parseChildBlock(); 1126 return true; 1127 } 1128 1129 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 1130 nextToken(); 1131 if (FormatTok->is(tok::equal)) { 1132 nextToken(); 1133 if (FormatTok->is(tok::r_square)) { 1134 nextToken(); 1135 return true; 1136 } 1137 if (FormatTok->isNot(tok::comma)) 1138 return false; 1139 nextToken(); 1140 } else if (FormatTok->is(tok::amp)) { 1141 nextToken(); 1142 if (FormatTok->is(tok::r_square)) { 1143 nextToken(); 1144 return true; 1145 } 1146 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 1147 return false; 1148 } 1149 if (FormatTok->is(tok::comma)) 1150 nextToken(); 1151 } else if (FormatTok->is(tok::r_square)) { 1152 nextToken(); 1153 return true; 1154 } 1155 do { 1156 if (FormatTok->is(tok::amp)) 1157 nextToken(); 1158 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 1159 return false; 1160 nextToken(); 1161 if (FormatTok->is(tok::ellipsis)) 1162 nextToken(); 1163 if (FormatTok->is(tok::comma)) { 1164 nextToken(); 1165 } else if (FormatTok->is(tok::r_square)) { 1166 nextToken(); 1167 return true; 1168 } else { 1169 return false; 1170 } 1171 } while (!eof()); 1172 return false; 1173 } 1174 1175 void UnwrappedLineParser::tryToParseJSFunction() { 1176 nextToken(); 1177 1178 // Consume function name. 1179 if (FormatTok->is(tok::identifier)) 1180 nextToken(); 1181 1182 if (FormatTok->isNot(tok::l_paren)) 1183 return; 1184 1185 // Parse formal parameter list. 1186 parseParens(); 1187 1188 if (FormatTok->is(tok::colon)) { 1189 // Parse a type definition. 1190 nextToken(); 1191 1192 // Eat the type declaration. For braced inline object types, balance braces, 1193 // otherwise just parse until finding an l_brace for the function body. 1194 if (FormatTok->is(tok::l_brace)) 1195 tryToParseBracedList(); 1196 else 1197 while (FormatTok->isNot(tok::l_brace) && !eof()) 1198 nextToken(); 1199 } 1200 1201 parseChildBlock(); 1202 } 1203 1204 bool UnwrappedLineParser::tryToParseBracedList() { 1205 if (FormatTok->BlockKind == BK_Unknown) 1206 calculateBraceTypes(); 1207 assert(FormatTok->BlockKind != BK_Unknown); 1208 if (FormatTok->BlockKind == BK_Block) 1209 return false; 1210 parseBracedList(); 1211 return true; 1212 } 1213 1214 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 1215 bool HasError = false; 1216 nextToken(); 1217 1218 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 1219 // replace this by using parseAssigmentExpression() inside. 1220 do { 1221 if (Style.Language == FormatStyle::LK_JavaScript) { 1222 if (FormatTok->is(Keywords.kw_function)) { 1223 tryToParseJSFunction(); 1224 continue; 1225 } 1226 if (FormatTok->is(TT_JsFatArrow)) { 1227 nextToken(); 1228 // Fat arrows can be followed by simple expressions or by child blocks 1229 // in curly braces. 1230 if (FormatTok->is(tok::l_brace)) { 1231 parseChildBlock(); 1232 continue; 1233 } 1234 } 1235 } 1236 switch (FormatTok->Tok.getKind()) { 1237 case tok::caret: 1238 nextToken(); 1239 if (FormatTok->is(tok::l_brace)) { 1240 parseChildBlock(); 1241 } 1242 break; 1243 case tok::l_square: 1244 tryToParseLambda(); 1245 break; 1246 case tok::l_brace: 1247 // Assume there are no blocks inside a braced init list apart 1248 // from the ones we explicitly parse out (like lambdas). 1249 FormatTok->BlockKind = BK_BracedInit; 1250 parseBracedList(); 1251 break; 1252 case tok::l_paren: 1253 parseParens(); 1254 // JavaScript can just have free standing methods and getters/setters in 1255 // object literals. Detect them by a "{" following ")". 1256 if (Style.Language == FormatStyle::LK_JavaScript) { 1257 if (FormatTok->is(tok::l_brace)) 1258 parseChildBlock(); 1259 break; 1260 } 1261 break; 1262 case tok::r_brace: 1263 nextToken(); 1264 return !HasError; 1265 case tok::semi: 1266 // JavaScript (or more precisely TypeScript) can have semicolons in braced 1267 // lists (in so-called TypeMemberLists). Thus, the semicolon cannot be 1268 // used for error recovery if we have otherwise determined that this is 1269 // a braced list. 1270 if (Style.Language == FormatStyle::LK_JavaScript) { 1271 nextToken(); 1272 break; 1273 } 1274 HasError = true; 1275 if (!ContinueOnSemicolons) 1276 return !HasError; 1277 nextToken(); 1278 break; 1279 case tok::comma: 1280 nextToken(); 1281 break; 1282 default: 1283 nextToken(); 1284 break; 1285 } 1286 } while (!eof()); 1287 return false; 1288 } 1289 1290 void UnwrappedLineParser::parseParens() { 1291 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1292 nextToken(); 1293 do { 1294 switch (FormatTok->Tok.getKind()) { 1295 case tok::l_paren: 1296 parseParens(); 1297 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) 1298 parseChildBlock(); 1299 break; 1300 case tok::r_paren: 1301 nextToken(); 1302 return; 1303 case tok::r_brace: 1304 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1305 return; 1306 case tok::l_square: 1307 tryToParseLambda(); 1308 break; 1309 case tok::l_brace: 1310 if (!tryToParseBracedList()) 1311 parseChildBlock(); 1312 break; 1313 case tok::at: 1314 nextToken(); 1315 if (FormatTok->Tok.is(tok::l_brace)) 1316 parseBracedList(); 1317 break; 1318 case tok::identifier: 1319 if (Style.Language == FormatStyle::LK_JavaScript && 1320 FormatTok->is(Keywords.kw_function)) 1321 tryToParseJSFunction(); 1322 else 1323 nextToken(); 1324 break; 1325 default: 1326 nextToken(); 1327 break; 1328 } 1329 } while (!eof()); 1330 } 1331 1332 void UnwrappedLineParser::parseSquare() { 1333 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1334 if (tryToParseLambda()) 1335 return; 1336 do { 1337 switch (FormatTok->Tok.getKind()) { 1338 case tok::l_paren: 1339 parseParens(); 1340 break; 1341 case tok::r_square: 1342 nextToken(); 1343 return; 1344 case tok::r_brace: 1345 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1346 return; 1347 case tok::l_square: 1348 parseSquare(); 1349 break; 1350 case tok::l_brace: { 1351 if (!tryToParseBracedList()) 1352 parseChildBlock(); 1353 break; 1354 } 1355 case tok::at: 1356 nextToken(); 1357 if (FormatTok->Tok.is(tok::l_brace)) 1358 parseBracedList(); 1359 break; 1360 default: 1361 nextToken(); 1362 break; 1363 } 1364 } while (!eof()); 1365 } 1366 1367 void UnwrappedLineParser::parseIfThenElse() { 1368 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1369 nextToken(); 1370 if (FormatTok->Tok.is(tok::l_paren)) 1371 parseParens(); 1372 bool NeedsUnwrappedLine = false; 1373 if (FormatTok->Tok.is(tok::l_brace)) { 1374 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1375 parseBlock(/*MustBeDeclaration=*/false); 1376 if (Style.BraceWrapping.BeforeElse) 1377 addUnwrappedLine(); 1378 else 1379 NeedsUnwrappedLine = true; 1380 } else { 1381 addUnwrappedLine(); 1382 ++Line->Level; 1383 parseStructuralElement(); 1384 --Line->Level; 1385 } 1386 if (FormatTok->Tok.is(tok::kw_else)) { 1387 nextToken(); 1388 if (FormatTok->Tok.is(tok::l_brace)) { 1389 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1390 parseBlock(/*MustBeDeclaration=*/false); 1391 addUnwrappedLine(); 1392 } else if (FormatTok->Tok.is(tok::kw_if)) { 1393 parseIfThenElse(); 1394 } else { 1395 addUnwrappedLine(); 1396 ++Line->Level; 1397 parseStructuralElement(); 1398 --Line->Level; 1399 } 1400 } else if (NeedsUnwrappedLine) { 1401 addUnwrappedLine(); 1402 } 1403 } 1404 1405 void UnwrappedLineParser::parseTryCatch() { 1406 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected"); 1407 nextToken(); 1408 bool NeedsUnwrappedLine = false; 1409 if (FormatTok->is(tok::colon)) { 1410 // We are in a function try block, what comes is an initializer list. 1411 nextToken(); 1412 while (FormatTok->is(tok::identifier)) { 1413 nextToken(); 1414 if (FormatTok->is(tok::l_paren)) 1415 parseParens(); 1416 if (FormatTok->is(tok::comma)) 1417 nextToken(); 1418 } 1419 } 1420 // Parse try with resource. 1421 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) { 1422 parseParens(); 1423 } 1424 if (FormatTok->is(tok::l_brace)) { 1425 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1426 parseBlock(/*MustBeDeclaration=*/false); 1427 if (Style.BraceWrapping.BeforeCatch) { 1428 addUnwrappedLine(); 1429 } else { 1430 NeedsUnwrappedLine = true; 1431 } 1432 } else if (!FormatTok->is(tok::kw_catch)) { 1433 // The C++ standard requires a compound-statement after a try. 1434 // If there's none, we try to assume there's a structuralElement 1435 // and try to continue. 1436 addUnwrappedLine(); 1437 ++Line->Level; 1438 parseStructuralElement(); 1439 --Line->Level; 1440 } 1441 while (1) { 1442 if (FormatTok->is(tok::at)) 1443 nextToken(); 1444 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, 1445 tok::kw___finally) || 1446 ((Style.Language == FormatStyle::LK_Java || 1447 Style.Language == FormatStyle::LK_JavaScript) && 1448 FormatTok->is(Keywords.kw_finally)) || 1449 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) || 1450 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally)))) 1451 break; 1452 nextToken(); 1453 while (FormatTok->isNot(tok::l_brace)) { 1454 if (FormatTok->is(tok::l_paren)) { 1455 parseParens(); 1456 continue; 1457 } 1458 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) 1459 return; 1460 nextToken(); 1461 } 1462 NeedsUnwrappedLine = false; 1463 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1464 parseBlock(/*MustBeDeclaration=*/false); 1465 if (Style.BraceWrapping.BeforeCatch) 1466 addUnwrappedLine(); 1467 else 1468 NeedsUnwrappedLine = true; 1469 } 1470 if (NeedsUnwrappedLine) 1471 addUnwrappedLine(); 1472 } 1473 1474 void UnwrappedLineParser::parseNamespace() { 1475 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1476 1477 const FormatToken &InitialToken = *FormatTok; 1478 nextToken(); 1479 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon)) 1480 nextToken(); 1481 if (FormatTok->Tok.is(tok::l_brace)) { 1482 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1483 addUnwrappedLine(); 1484 1485 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1486 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1487 DeclarationScopeStack.size() > 1); 1488 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1489 // Munch the semicolon after a namespace. This is more common than one would 1490 // think. Puttin the semicolon into its own line is very ugly. 1491 if (FormatTok->Tok.is(tok::semi)) 1492 nextToken(); 1493 addUnwrappedLine(); 1494 } 1495 // FIXME: Add error handling. 1496 } 1497 1498 void UnwrappedLineParser::parseNew() { 1499 assert(FormatTok->is(tok::kw_new) && "'new' expected"); 1500 nextToken(); 1501 if (Style.Language != FormatStyle::LK_Java) 1502 return; 1503 1504 // In Java, we can parse everything up to the parens, which aren't optional. 1505 do { 1506 // There should not be a ;, { or } before the new's open paren. 1507 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) 1508 return; 1509 1510 // Consume the parens. 1511 if (FormatTok->is(tok::l_paren)) { 1512 parseParens(); 1513 1514 // If there is a class body of an anonymous class, consume that as child. 1515 if (FormatTok->is(tok::l_brace)) 1516 parseChildBlock(); 1517 return; 1518 } 1519 nextToken(); 1520 } while (!eof()); 1521 } 1522 1523 void UnwrappedLineParser::parseForOrWhileLoop() { 1524 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) && 1525 "'for', 'while' or foreach macro expected"); 1526 nextToken(); 1527 if (FormatTok->Tok.is(tok::l_paren)) 1528 parseParens(); 1529 if (FormatTok->Tok.is(tok::l_brace)) { 1530 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1531 parseBlock(/*MustBeDeclaration=*/false); 1532 addUnwrappedLine(); 1533 } else { 1534 addUnwrappedLine(); 1535 ++Line->Level; 1536 parseStructuralElement(); 1537 --Line->Level; 1538 } 1539 } 1540 1541 void UnwrappedLineParser::parseDoWhile() { 1542 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1543 nextToken(); 1544 if (FormatTok->Tok.is(tok::l_brace)) { 1545 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1546 parseBlock(/*MustBeDeclaration=*/false); 1547 if (Style.BraceWrapping.IndentBraces) 1548 addUnwrappedLine(); 1549 } else { 1550 addUnwrappedLine(); 1551 ++Line->Level; 1552 parseStructuralElement(); 1553 --Line->Level; 1554 } 1555 1556 // FIXME: Add error handling. 1557 if (!FormatTok->Tok.is(tok::kw_while)) { 1558 addUnwrappedLine(); 1559 return; 1560 } 1561 1562 nextToken(); 1563 parseStructuralElement(); 1564 } 1565 1566 void UnwrappedLineParser::parseLabel() { 1567 nextToken(); 1568 unsigned OldLineLevel = Line->Level; 1569 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1570 --Line->Level; 1571 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1572 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1573 parseBlock(/*MustBeDeclaration=*/false); 1574 if (FormatTok->Tok.is(tok::kw_break)) { 1575 if (Style.BraceWrapping.AfterControlStatement) 1576 addUnwrappedLine(); 1577 parseStructuralElement(); 1578 } 1579 addUnwrappedLine(); 1580 } else { 1581 if (FormatTok->is(tok::semi)) 1582 nextToken(); 1583 addUnwrappedLine(); 1584 } 1585 Line->Level = OldLineLevel; 1586 if (FormatTok->isNot(tok::l_brace)) { 1587 parseStructuralElement(); 1588 addUnwrappedLine(); 1589 } 1590 } 1591 1592 void UnwrappedLineParser::parseCaseLabel() { 1593 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1594 // FIXME: fix handling of complex expressions here. 1595 do { 1596 nextToken(); 1597 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1598 parseLabel(); 1599 } 1600 1601 void UnwrappedLineParser::parseSwitch() { 1602 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1603 nextToken(); 1604 if (FormatTok->Tok.is(tok::l_paren)) 1605 parseParens(); 1606 if (FormatTok->Tok.is(tok::l_brace)) { 1607 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1608 parseBlock(/*MustBeDeclaration=*/false); 1609 addUnwrappedLine(); 1610 } else { 1611 addUnwrappedLine(); 1612 ++Line->Level; 1613 parseStructuralElement(); 1614 --Line->Level; 1615 } 1616 } 1617 1618 void UnwrappedLineParser::parseAccessSpecifier() { 1619 nextToken(); 1620 // Understand Qt's slots. 1621 if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots)) 1622 nextToken(); 1623 // Otherwise, we don't know what it is, and we'd better keep the next token. 1624 if (FormatTok->Tok.is(tok::colon)) 1625 nextToken(); 1626 addUnwrappedLine(); 1627 } 1628 1629 bool UnwrappedLineParser::parseEnum() { 1630 // Won't be 'enum' for NS_ENUMs. 1631 if (FormatTok->Tok.is(tok::kw_enum)) 1632 nextToken(); 1633 1634 // In TypeScript, "enum" can also be used as property name, e.g. in interface 1635 // declarations. An "enum" keyword followed by a colon would be a syntax 1636 // error and thus assume it is just an identifier. 1637 if (Style.Language == FormatStyle::LK_JavaScript && 1638 FormatTok->isOneOf(tok::colon, tok::question)) 1639 return false; 1640 1641 // Eat up enum class ... 1642 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1643 nextToken(); 1644 1645 while (FormatTok->Tok.getIdentifierInfo() || 1646 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less, 1647 tok::greater, tok::comma, tok::question)) { 1648 nextToken(); 1649 // We can have macros or attributes in between 'enum' and the enum name. 1650 if (FormatTok->is(tok::l_paren)) 1651 parseParens(); 1652 if (FormatTok->is(tok::identifier)) { 1653 nextToken(); 1654 // If there are two identifiers in a row, this is likely an elaborate 1655 // return type. In Java, this can be "implements", etc. 1656 if (Style.Language == FormatStyle::LK_Cpp && 1657 FormatTok->is(tok::identifier)) 1658 return false; 1659 } 1660 } 1661 1662 // Just a declaration or something is wrong. 1663 if (FormatTok->isNot(tok::l_brace)) 1664 return true; 1665 FormatTok->BlockKind = BK_Block; 1666 1667 if (Style.Language == FormatStyle::LK_Java) { 1668 // Java enums are different. 1669 parseJavaEnumBody(); 1670 return true; 1671 } 1672 if (Style.Language == FormatStyle::LK_Proto) { 1673 parseBlock(/*MustBeDeclaration=*/true); 1674 return true; 1675 } 1676 1677 // Parse enum body. 1678 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1679 if (HasError) { 1680 if (FormatTok->is(tok::semi)) 1681 nextToken(); 1682 addUnwrappedLine(); 1683 } 1684 return true; 1685 1686 // There is no addUnwrappedLine() here so that we fall through to parsing a 1687 // structural element afterwards. Thus, in "enum A {} n, m;", 1688 // "} n, m;" will end up in one unwrapped line. 1689 } 1690 1691 void UnwrappedLineParser::parseJavaEnumBody() { 1692 // Determine whether the enum is simple, i.e. does not have a semicolon or 1693 // constants with class bodies. Simple enums can be formatted like braced 1694 // lists, contracted to a single line, etc. 1695 unsigned StoredPosition = Tokens->getPosition(); 1696 bool IsSimple = true; 1697 FormatToken *Tok = Tokens->getNextToken(); 1698 while (Tok) { 1699 if (Tok->is(tok::r_brace)) 1700 break; 1701 if (Tok->isOneOf(tok::l_brace, tok::semi)) { 1702 IsSimple = false; 1703 break; 1704 } 1705 // FIXME: This will also mark enums with braces in the arguments to enum 1706 // constants as "not simple". This is probably fine in practice, though. 1707 Tok = Tokens->getNextToken(); 1708 } 1709 FormatTok = Tokens->setPosition(StoredPosition); 1710 1711 if (IsSimple) { 1712 parseBracedList(); 1713 addUnwrappedLine(); 1714 return; 1715 } 1716 1717 // Parse the body of a more complex enum. 1718 // First add a line for everything up to the "{". 1719 nextToken(); 1720 addUnwrappedLine(); 1721 ++Line->Level; 1722 1723 // Parse the enum constants. 1724 while (FormatTok) { 1725 if (FormatTok->is(tok::l_brace)) { 1726 // Parse the constant's class body. 1727 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1728 /*MunchSemi=*/false); 1729 } else if (FormatTok->is(tok::l_paren)) { 1730 parseParens(); 1731 } else if (FormatTok->is(tok::comma)) { 1732 nextToken(); 1733 addUnwrappedLine(); 1734 } else if (FormatTok->is(tok::semi)) { 1735 nextToken(); 1736 addUnwrappedLine(); 1737 break; 1738 } else if (FormatTok->is(tok::r_brace)) { 1739 addUnwrappedLine(); 1740 break; 1741 } else { 1742 nextToken(); 1743 } 1744 } 1745 1746 // Parse the class body after the enum's ";" if any. 1747 parseLevel(/*HasOpeningBrace=*/true); 1748 nextToken(); 1749 --Line->Level; 1750 addUnwrappedLine(); 1751 } 1752 1753 void UnwrappedLineParser::parseRecord() { 1754 const FormatToken &InitialToken = *FormatTok; 1755 nextToken(); 1756 1757 // The actual identifier can be a nested name specifier, and in macros 1758 // it is often token-pasted. 1759 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash, 1760 tok::kw___attribute, tok::kw___declspec, 1761 tok::kw_alignas) || 1762 ((Style.Language == FormatStyle::LK_Java || 1763 Style.Language == FormatStyle::LK_JavaScript) && 1764 FormatTok->isOneOf(tok::period, tok::comma))) { 1765 bool IsNonMacroIdentifier = 1766 FormatTok->is(tok::identifier) && 1767 FormatTok->TokenText != FormatTok->TokenText.upper(); 1768 nextToken(); 1769 // We can have macros or attributes in between 'class' and the class name. 1770 if (!IsNonMacroIdentifier && FormatTok->Tok.is(tok::l_paren)) 1771 parseParens(); 1772 } 1773 1774 // Note that parsing away template declarations here leads to incorrectly 1775 // accepting function declarations as record declarations. 1776 // In general, we cannot solve this problem. Consider: 1777 // class A<int> B() {} 1778 // which can be a function definition or a class definition when B() is a 1779 // macro. If we find enough real-world cases where this is a problem, we 1780 // can parse for the 'template' keyword in the beginning of the statement, 1781 // and thus rule out the record production in case there is no template 1782 // (this would still leave us with an ambiguity between template function 1783 // and class declarations). 1784 if (FormatTok->isOneOf(tok::colon, tok::less)) { 1785 while (!eof()) { 1786 if (FormatTok->is(tok::l_brace)) { 1787 calculateBraceTypes(/*ExpectClassBody=*/true); 1788 if (!tryToParseBracedList()) 1789 break; 1790 } 1791 if (FormatTok->Tok.is(tok::semi)) 1792 return; 1793 nextToken(); 1794 } 1795 } 1796 if (FormatTok->Tok.is(tok::l_brace)) { 1797 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1798 addUnwrappedLine(); 1799 1800 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1801 /*MunchSemi=*/false); 1802 } 1803 // There is no addUnwrappedLine() here so that we fall through to parsing a 1804 // structural element afterwards. Thus, in "class A {} n, m;", 1805 // "} n, m;" will end up in one unwrapped line. 1806 } 1807 1808 void UnwrappedLineParser::parseObjCProtocolList() { 1809 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1810 do 1811 nextToken(); 1812 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1813 nextToken(); // Skip '>'. 1814 } 1815 1816 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1817 do { 1818 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1819 nextToken(); 1820 addUnwrappedLine(); 1821 break; 1822 } 1823 if (FormatTok->is(tok::l_brace)) { 1824 parseBlock(/*MustBeDeclaration=*/false); 1825 // In ObjC interfaces, nothing should be following the "}". 1826 addUnwrappedLine(); 1827 } else if (FormatTok->is(tok::r_brace)) { 1828 // Ignore stray "}". parseStructuralElement doesn't consume them. 1829 nextToken(); 1830 addUnwrappedLine(); 1831 } else { 1832 parseStructuralElement(); 1833 } 1834 } while (!eof()); 1835 } 1836 1837 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1838 nextToken(); 1839 nextToken(); // interface name 1840 1841 // @interface can be followed by either a base class, or a category. 1842 if (FormatTok->Tok.is(tok::colon)) { 1843 nextToken(); 1844 nextToken(); // base class name 1845 } else if (FormatTok->Tok.is(tok::l_paren)) 1846 // Skip category, if present. 1847 parseParens(); 1848 1849 if (FormatTok->Tok.is(tok::less)) 1850 parseObjCProtocolList(); 1851 1852 if (FormatTok->Tok.is(tok::l_brace)) { 1853 if (Style.BraceWrapping.AfterObjCDeclaration) 1854 addUnwrappedLine(); 1855 parseBlock(/*MustBeDeclaration=*/true); 1856 } 1857 1858 // With instance variables, this puts '}' on its own line. Without instance 1859 // variables, this ends the @interface line. 1860 addUnwrappedLine(); 1861 1862 parseObjCUntilAtEnd(); 1863 } 1864 1865 void UnwrappedLineParser::parseObjCProtocol() { 1866 nextToken(); 1867 nextToken(); // protocol name 1868 1869 if (FormatTok->Tok.is(tok::less)) 1870 parseObjCProtocolList(); 1871 1872 // Check for protocol declaration. 1873 if (FormatTok->Tok.is(tok::semi)) { 1874 nextToken(); 1875 return addUnwrappedLine(); 1876 } 1877 1878 addUnwrappedLine(); 1879 parseObjCUntilAtEnd(); 1880 } 1881 1882 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { 1883 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export)); 1884 nextToken(); 1885 1886 // Consume the "default" in "export default class/function". 1887 if (FormatTok->is(tok::kw_default)) 1888 nextToken(); 1889 1890 // Consume "function" and "default function", so that these get parsed as 1891 // free-standing JS functions, i.e. do not require a trailing semicolon. 1892 if (FormatTok->is(Keywords.kw_function)) { 1893 nextToken(); 1894 return; 1895 } 1896 1897 // Consume the "abstract" in "export abstract class". 1898 if (FormatTok->is(Keywords.kw_abstract)) 1899 nextToken(); 1900 1901 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, tok::kw_enum, 1902 Keywords.kw_interface, Keywords.kw_let, 1903 Keywords.kw_var)) 1904 return; // Fall through to parsing the corresponding structure. 1905 1906 while (!eof() && FormatTok->isNot(tok::semi)) { 1907 if (FormatTok->is(tok::l_brace)) { 1908 FormatTok->BlockKind = BK_Block; 1909 parseBracedList(); 1910 } else { 1911 nextToken(); 1912 } 1913 } 1914 } 1915 1916 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1917 StringRef Prefix = "") { 1918 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1919 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1920 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1921 E = Line.Tokens.end(); 1922 I != E; ++I) { 1923 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1924 } 1925 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1926 E = Line.Tokens.end(); 1927 I != E; ++I) { 1928 const UnwrappedLineNode &Node = *I; 1929 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1930 I = Node.Children.begin(), 1931 E = Node.Children.end(); 1932 I != E; ++I) { 1933 printDebugInfo(*I, "\nChild: "); 1934 } 1935 } 1936 llvm::dbgs() << "\n"; 1937 } 1938 1939 void UnwrappedLineParser::addUnwrappedLine() { 1940 if (Line->Tokens.empty()) 1941 return; 1942 DEBUG({ 1943 if (CurrentLines == &Lines) 1944 printDebugInfo(*Line); 1945 }); 1946 CurrentLines->push_back(std::move(*Line)); 1947 Line->Tokens.clear(); 1948 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1949 CurrentLines->append( 1950 std::make_move_iterator(PreprocessorDirectives.begin()), 1951 std::make_move_iterator(PreprocessorDirectives.end())); 1952 PreprocessorDirectives.clear(); 1953 } 1954 } 1955 1956 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1957 1958 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1959 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1960 FormatTok.NewlinesBefore > 0; 1961 } 1962 1963 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1964 bool JustComments = Line->Tokens.empty(); 1965 for (SmallVectorImpl<FormatToken *>::const_iterator 1966 I = CommentsBeforeNextToken.begin(), 1967 E = CommentsBeforeNextToken.end(); 1968 I != E; ++I) { 1969 if (isOnNewLine(**I) && JustComments) 1970 addUnwrappedLine(); 1971 pushToken(*I); 1972 } 1973 if (NewlineBeforeNext && JustComments) 1974 addUnwrappedLine(); 1975 CommentsBeforeNextToken.clear(); 1976 } 1977 1978 void UnwrappedLineParser::nextToken() { 1979 if (eof()) 1980 return; 1981 flushComments(isOnNewLine(*FormatTok)); 1982 pushToken(FormatTok); 1983 if (Style.Language != FormatStyle::LK_JavaScript) 1984 readToken(); 1985 else 1986 readTokenWithJavaScriptASI(); 1987 } 1988 1989 const FormatToken *UnwrappedLineParser::getPreviousToken() { 1990 // FIXME: This is a dirty way to access the previous token. Find a better 1991 // solution. 1992 if (!Line || Line->Tokens.empty()) 1993 return nullptr; 1994 return Line->Tokens.back().Tok; 1995 } 1996 1997 void UnwrappedLineParser::readToken() { 1998 bool CommentsInCurrentLine = true; 1999 do { 2000 FormatTok = Tokens->getNextToken(); 2001 assert(FormatTok); 2002 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 2003 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 2004 // If there is an unfinished unwrapped line, we flush the preprocessor 2005 // directives only after that unwrapped line was finished later. 2006 bool SwitchToPreprocessorLines = !Line->Tokens.empty(); 2007 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 2008 // Comments stored before the preprocessor directive need to be output 2009 // before the preprocessor directive, at the same level as the 2010 // preprocessor directive, as we consider them to apply to the directive. 2011 flushComments(isOnNewLine(*FormatTok)); 2012 parsePPDirective(); 2013 } 2014 while (FormatTok->Type == TT_ConflictStart || 2015 FormatTok->Type == TT_ConflictEnd || 2016 FormatTok->Type == TT_ConflictAlternative) { 2017 if (FormatTok->Type == TT_ConflictStart) { 2018 conditionalCompilationStart(/*Unreachable=*/false); 2019 } else if (FormatTok->Type == TT_ConflictAlternative) { 2020 conditionalCompilationAlternative(); 2021 } else if (FormatTok->Type == TT_ConflictEnd) { 2022 conditionalCompilationEnd(); 2023 } 2024 FormatTok = Tokens->getNextToken(); 2025 FormatTok->MustBreakBefore = true; 2026 } 2027 2028 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 2029 !Line->InPPDirective) { 2030 continue; 2031 } 2032 2033 if (!FormatTok->Tok.is(tok::comment)) 2034 return; 2035 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 2036 CommentsInCurrentLine = false; 2037 } 2038 if (CommentsInCurrentLine) { 2039 pushToken(FormatTok); 2040 } else { 2041 CommentsBeforeNextToken.push_back(FormatTok); 2042 } 2043 } while (!eof()); 2044 } 2045 2046 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 2047 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 2048 if (MustBreakBeforeNextToken) { 2049 Line->Tokens.back().Tok->MustBreakBefore = true; 2050 MustBreakBeforeNextToken = false; 2051 } 2052 } 2053 2054 } // end namespace format 2055 } // end namespace clang 2056