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 // 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 if (InitialToken.is(tok::kw_namespace)) 458 return Style.BraceWrapping.AfterNamespace; 459 if (InitialToken.is(tok::kw_class)) 460 return Style.BraceWrapping.AfterClass; 461 if (InitialToken.is(tok::kw_union)) 462 return Style.BraceWrapping.AfterUnion; 463 if (InitialToken.is(tok::kw_struct)) 464 return Style.BraceWrapping.AfterStruct; 465 return false; 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.BraceWrapping.AfterObjCDeclaration) 681 addUnwrappedLine(); 682 parseBlock(/*MustBeDeclaration=*/false); 683 } 684 addUnwrappedLine(); 685 return; 686 case tok::objc_try: 687 // This branch isn't strictly necessary (the kw_try case below would 688 // do this too after the tok::at is parsed above). But be explicit. 689 parseTryCatch(); 690 return; 691 default: 692 break; 693 } 694 break; 695 case tok::kw_asm: 696 nextToken(); 697 if (FormatTok->is(tok::l_brace)) { 698 FormatTok->Type = TT_InlineASMBrace; 699 nextToken(); 700 while (FormatTok && FormatTok->isNot(tok::eof)) { 701 if (FormatTok->is(tok::r_brace)) { 702 FormatTok->Type = TT_InlineASMBrace; 703 nextToken(); 704 addUnwrappedLine(); 705 break; 706 } 707 FormatTok->Finalized = true; 708 nextToken(); 709 } 710 } 711 break; 712 case tok::kw_namespace: 713 parseNamespace(); 714 return; 715 case tok::kw_inline: 716 nextToken(); 717 if (FormatTok->Tok.is(tok::kw_namespace)) { 718 parseNamespace(); 719 return; 720 } 721 break; 722 case tok::kw_public: 723 case tok::kw_protected: 724 case tok::kw_private: 725 if (Style.Language == FormatStyle::LK_Java || 726 Style.Language == FormatStyle::LK_JavaScript) 727 nextToken(); 728 else 729 parseAccessSpecifier(); 730 return; 731 case tok::kw_if: 732 parseIfThenElse(); 733 return; 734 case tok::kw_for: 735 case tok::kw_while: 736 parseForOrWhileLoop(); 737 return; 738 case tok::kw_do: 739 parseDoWhile(); 740 return; 741 case tok::kw_switch: 742 parseSwitch(); 743 return; 744 case tok::kw_default: 745 nextToken(); 746 parseLabel(); 747 return; 748 case tok::kw_case: 749 parseCaseLabel(); 750 return; 751 case tok::kw_try: 752 case tok::kw___try: 753 parseTryCatch(); 754 return; 755 case tok::kw_extern: 756 nextToken(); 757 if (FormatTok->Tok.is(tok::string_literal)) { 758 nextToken(); 759 if (FormatTok->Tok.is(tok::l_brace)) { 760 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false); 761 addUnwrappedLine(); 762 return; 763 } 764 } 765 break; 766 case tok::kw_export: 767 if (Style.Language == FormatStyle::LK_JavaScript) { 768 parseJavaScriptEs6ImportExport(); 769 return; 770 } 771 break; 772 case tok::identifier: 773 if (FormatTok->is(TT_ForEachMacro)) { 774 parseForOrWhileLoop(); 775 return; 776 } 777 if (FormatTok->is(TT_MacroBlockBegin)) { 778 parseBlock(/*MustBeDeclaration=*/false, /*AddLevel=*/true, 779 /*MunchSemi=*/false); 780 return; 781 } 782 if (Style.Language == FormatStyle::LK_JavaScript && 783 FormatTok->is(Keywords.kw_import)) { 784 parseJavaScriptEs6ImportExport(); 785 return; 786 } 787 if (FormatTok->is(Keywords.kw_signals)) { 788 nextToken(); 789 if (FormatTok->is(tok::colon)) { 790 nextToken(); 791 addUnwrappedLine(); 792 } 793 return; 794 } 795 // In all other cases, parse the declaration. 796 break; 797 default: 798 break; 799 } 800 do { 801 switch (FormatTok->Tok.getKind()) { 802 case tok::at: 803 nextToken(); 804 if (FormatTok->Tok.is(tok::l_brace)) 805 parseBracedList(); 806 break; 807 case tok::kw_enum: 808 // parseEnum falls through and does not yet add an unwrapped line as an 809 // enum definition can start a structural element. 810 parseEnum(); 811 // This only applies for C++. 812 if (Style.Language != FormatStyle::LK_Cpp) { 813 addUnwrappedLine(); 814 return; 815 } 816 break; 817 case tok::kw_typedef: 818 nextToken(); 819 if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS, 820 Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS)) 821 parseEnum(); 822 break; 823 case tok::kw_struct: 824 case tok::kw_union: 825 case tok::kw_class: 826 // parseRecord falls through and does not yet add an unwrapped line as a 827 // record declaration or definition can start a structural element. 828 parseRecord(); 829 // This does not apply for Java and JavaScript. 830 if (Style.Language == FormatStyle::LK_Java || 831 Style.Language == FormatStyle::LK_JavaScript) { 832 addUnwrappedLine(); 833 return; 834 } 835 break; 836 case tok::period: 837 nextToken(); 838 // In Java, classes have an implicit static member "class". 839 if (Style.Language == FormatStyle::LK_Java && FormatTok && 840 FormatTok->is(tok::kw_class)) 841 nextToken(); 842 if (Style.Language == FormatStyle::LK_JavaScript && FormatTok && 843 FormatTok->Tok.getIdentifierInfo()) 844 // JavaScript only has pseudo keywords, all keywords are allowed to 845 // appear in "IdentifierName" positions. See http://es5.github.io/#x7.6 846 nextToken(); 847 break; 848 case tok::semi: 849 nextToken(); 850 addUnwrappedLine(); 851 return; 852 case tok::r_brace: 853 addUnwrappedLine(); 854 return; 855 case tok::l_paren: 856 parseParens(); 857 break; 858 case tok::caret: 859 nextToken(); 860 if (FormatTok->Tok.isAnyIdentifier() || 861 FormatTok->isSimpleTypeSpecifier()) 862 nextToken(); 863 if (FormatTok->is(tok::l_paren)) 864 parseParens(); 865 if (FormatTok->is(tok::l_brace)) 866 parseChildBlock(); 867 break; 868 case tok::l_brace: 869 if (!tryToParseBracedList()) { 870 // A block outside of parentheses must be the last part of a 871 // structural element. 872 // FIXME: Figure out cases where this is not true, and add projections 873 // for them (the one we know is missing are lambdas). 874 if (Style.BraceWrapping.AfterFunction) 875 addUnwrappedLine(); 876 FormatTok->Type = TT_FunctionLBrace; 877 parseBlock(/*MustBeDeclaration=*/false); 878 addUnwrappedLine(); 879 return; 880 } 881 // Otherwise this was a braced init list, and the structural 882 // element continues. 883 break; 884 case tok::kw_try: 885 // We arrive here when parsing function-try blocks. 886 parseTryCatch(); 887 return; 888 case tok::identifier: { 889 if (FormatTok->is(TT_MacroBlockEnd)) { 890 addUnwrappedLine(); 891 return; 892 } 893 894 // Parse function literal unless 'function' is the first token in a line 895 // in which case this should be treated as a free-standing function. 896 if (Style.Language == FormatStyle::LK_JavaScript && 897 FormatTok->is(Keywords.kw_function) && Line->Tokens.size() > 0) { 898 tryToParseJSFunction(); 899 break; 900 } 901 if ((Style.Language == FormatStyle::LK_JavaScript || 902 Style.Language == FormatStyle::LK_Java) && 903 FormatTok->is(Keywords.kw_interface)) { 904 parseRecord(); 905 addUnwrappedLine(); 906 return; 907 } 908 909 StringRef Text = FormatTok->TokenText; 910 nextToken(); 911 if (Line->Tokens.size() == 1 && 912 // JS doesn't have macros, and within classes colons indicate fields, 913 // not labels. 914 Style.Language != FormatStyle::LK_JavaScript) { 915 if (FormatTok->Tok.is(tok::colon) && !Line->MustBeDeclaration) { 916 parseLabel(); 917 return; 918 } 919 // Recognize function-like macro usages without trailing semicolon as 920 // well as free-standing macros like Q_OBJECT. 921 bool FunctionLike = FormatTok->is(tok::l_paren); 922 if (FunctionLike) 923 parseParens(); 924 925 bool FollowedByNewline = 926 CommentsBeforeNextToken.empty() 927 ? FormatTok->NewlinesBefore > 0 928 : CommentsBeforeNextToken.front()->NewlinesBefore > 0; 929 930 if (FollowedByNewline && (Text.size() >= 5 || FunctionLike) && 931 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) { 932 addUnwrappedLine(); 933 return; 934 } 935 } 936 break; 937 } 938 case tok::equal: 939 // Fat arrows (=>) have tok::TokenKind tok::equal but TokenType 940 // TT_JsFatArrow. The always start an expression or a child block if 941 // followed by a curly. 942 if (FormatTok->is(TT_JsFatArrow)) { 943 nextToken(); 944 if (FormatTok->is(tok::l_brace)) 945 parseChildBlock(); 946 break; 947 } 948 949 nextToken(); 950 if (FormatTok->Tok.is(tok::l_brace)) { 951 parseBracedList(); 952 } 953 break; 954 case tok::l_square: 955 parseSquare(); 956 break; 957 case tok::kw_new: 958 parseNew(); 959 break; 960 default: 961 nextToken(); 962 break; 963 } 964 } while (!eof()); 965 } 966 967 bool UnwrappedLineParser::tryToParseLambda() { 968 if (Style.Language != FormatStyle::LK_Cpp) { 969 nextToken(); 970 return false; 971 } 972 // FIXME: This is a dirty way to access the previous token. Find a better 973 // solution. 974 if (!Line->Tokens.empty() && 975 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator, 976 tok::kw_new, tok::kw_delete) || 977 Line->Tokens.back().Tok->closesScope() || 978 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) { 979 nextToken(); 980 return false; 981 } 982 assert(FormatTok->is(tok::l_square)); 983 FormatToken &LSquare = *FormatTok; 984 if (!tryToParseLambdaIntroducer()) 985 return false; 986 987 while (FormatTok->isNot(tok::l_brace)) { 988 if (FormatTok->isSimpleTypeSpecifier()) { 989 nextToken(); 990 continue; 991 } 992 switch (FormatTok->Tok.getKind()) { 993 case tok::l_brace: 994 break; 995 case tok::l_paren: 996 parseParens(); 997 break; 998 case tok::amp: 999 case tok::star: 1000 case tok::kw_const: 1001 case tok::comma: 1002 case tok::less: 1003 case tok::greater: 1004 case tok::identifier: 1005 case tok::numeric_constant: 1006 case tok::coloncolon: 1007 case tok::kw_mutable: 1008 nextToken(); 1009 break; 1010 case tok::arrow: 1011 FormatTok->Type = TT_LambdaArrow; 1012 nextToken(); 1013 break; 1014 default: 1015 return true; 1016 } 1017 } 1018 LSquare.Type = TT_LambdaLSquare; 1019 parseChildBlock(); 1020 return true; 1021 } 1022 1023 bool UnwrappedLineParser::tryToParseLambdaIntroducer() { 1024 nextToken(); 1025 if (FormatTok->is(tok::equal)) { 1026 nextToken(); 1027 if (FormatTok->is(tok::r_square)) { 1028 nextToken(); 1029 return true; 1030 } 1031 if (FormatTok->isNot(tok::comma)) 1032 return false; 1033 nextToken(); 1034 } else if (FormatTok->is(tok::amp)) { 1035 nextToken(); 1036 if (FormatTok->is(tok::r_square)) { 1037 nextToken(); 1038 return true; 1039 } 1040 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) { 1041 return false; 1042 } 1043 if (FormatTok->is(tok::comma)) 1044 nextToken(); 1045 } else if (FormatTok->is(tok::r_square)) { 1046 nextToken(); 1047 return true; 1048 } 1049 do { 1050 if (FormatTok->is(tok::amp)) 1051 nextToken(); 1052 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this)) 1053 return false; 1054 nextToken(); 1055 if (FormatTok->is(tok::ellipsis)) 1056 nextToken(); 1057 if (FormatTok->is(tok::comma)) { 1058 nextToken(); 1059 } else if (FormatTok->is(tok::r_square)) { 1060 nextToken(); 1061 return true; 1062 } else { 1063 return false; 1064 } 1065 } while (!eof()); 1066 return false; 1067 } 1068 1069 void UnwrappedLineParser::tryToParseJSFunction() { 1070 nextToken(); 1071 1072 // Consume function name. 1073 if (FormatTok->is(tok::identifier)) 1074 nextToken(); 1075 1076 if (FormatTok->isNot(tok::l_paren)) 1077 return; 1078 1079 // Parse formal parameter list. 1080 parseParens(); 1081 1082 if (FormatTok->is(tok::colon)) { 1083 // Parse a type definition. 1084 nextToken(); 1085 1086 // Eat the type declaration. For braced inline object types, balance braces, 1087 // otherwise just parse until finding an l_brace for the function body. 1088 if (FormatTok->is(tok::l_brace)) 1089 tryToParseBracedList(); 1090 else 1091 while (FormatTok->isNot(tok::l_brace) && !eof()) 1092 nextToken(); 1093 } 1094 1095 parseChildBlock(); 1096 } 1097 1098 bool UnwrappedLineParser::tryToParseBracedList() { 1099 if (FormatTok->BlockKind == BK_Unknown) 1100 calculateBraceTypes(); 1101 assert(FormatTok->BlockKind != BK_Unknown); 1102 if (FormatTok->BlockKind == BK_Block) 1103 return false; 1104 parseBracedList(); 1105 return true; 1106 } 1107 1108 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) { 1109 bool HasError = false; 1110 nextToken(); 1111 1112 // FIXME: Once we have an expression parser in the UnwrappedLineParser, 1113 // replace this by using parseAssigmentExpression() inside. 1114 do { 1115 if (Style.Language == FormatStyle::LK_JavaScript) { 1116 if (FormatTok->is(Keywords.kw_function)) { 1117 tryToParseJSFunction(); 1118 continue; 1119 } 1120 if (FormatTok->is(TT_JsFatArrow)) { 1121 nextToken(); 1122 // Fat arrows can be followed by simple expressions or by child blocks 1123 // in curly braces. 1124 if (FormatTok->is(tok::l_brace)) { 1125 parseChildBlock(); 1126 continue; 1127 } 1128 } 1129 } 1130 switch (FormatTok->Tok.getKind()) { 1131 case tok::caret: 1132 nextToken(); 1133 if (FormatTok->is(tok::l_brace)) { 1134 parseChildBlock(); 1135 } 1136 break; 1137 case tok::l_square: 1138 tryToParseLambda(); 1139 break; 1140 case tok::l_brace: 1141 // Assume there are no blocks inside a braced init list apart 1142 // from the ones we explicitly parse out (like lambdas). 1143 FormatTok->BlockKind = BK_BracedInit; 1144 parseBracedList(); 1145 break; 1146 case tok::l_paren: 1147 parseParens(); 1148 // JavaScript can just have free standing methods and getters/setters in 1149 // object literals. Detect them by a "{" following ")". 1150 if (Style.Language == FormatStyle::LK_JavaScript) { 1151 if (FormatTok->is(tok::l_brace)) 1152 parseChildBlock(); 1153 break; 1154 } 1155 break; 1156 case tok::r_brace: 1157 nextToken(); 1158 return !HasError; 1159 case tok::semi: 1160 HasError = true; 1161 if (!ContinueOnSemicolons) 1162 return !HasError; 1163 nextToken(); 1164 break; 1165 case tok::comma: 1166 nextToken(); 1167 break; 1168 default: 1169 nextToken(); 1170 break; 1171 } 1172 } while (!eof()); 1173 return false; 1174 } 1175 1176 void UnwrappedLineParser::parseParens() { 1177 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected."); 1178 nextToken(); 1179 do { 1180 switch (FormatTok->Tok.getKind()) { 1181 case tok::l_paren: 1182 parseParens(); 1183 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace)) 1184 parseChildBlock(); 1185 break; 1186 case tok::r_paren: 1187 nextToken(); 1188 return; 1189 case tok::r_brace: 1190 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1191 return; 1192 case tok::l_square: 1193 tryToParseLambda(); 1194 break; 1195 case tok::l_brace: 1196 if (!tryToParseBracedList()) 1197 parseChildBlock(); 1198 break; 1199 case tok::at: 1200 nextToken(); 1201 if (FormatTok->Tok.is(tok::l_brace)) 1202 parseBracedList(); 1203 break; 1204 case tok::identifier: 1205 if (Style.Language == FormatStyle::LK_JavaScript && 1206 FormatTok->is(Keywords.kw_function)) 1207 tryToParseJSFunction(); 1208 else 1209 nextToken(); 1210 break; 1211 default: 1212 nextToken(); 1213 break; 1214 } 1215 } while (!eof()); 1216 } 1217 1218 void UnwrappedLineParser::parseSquare() { 1219 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected."); 1220 if (tryToParseLambda()) 1221 return; 1222 do { 1223 switch (FormatTok->Tok.getKind()) { 1224 case tok::l_paren: 1225 parseParens(); 1226 break; 1227 case tok::r_square: 1228 nextToken(); 1229 return; 1230 case tok::r_brace: 1231 // A "}" inside parenthesis is an error if there wasn't a matching "{". 1232 return; 1233 case tok::l_square: 1234 parseSquare(); 1235 break; 1236 case tok::l_brace: { 1237 if (!tryToParseBracedList()) 1238 parseChildBlock(); 1239 break; 1240 } 1241 case tok::at: 1242 nextToken(); 1243 if (FormatTok->Tok.is(tok::l_brace)) 1244 parseBracedList(); 1245 break; 1246 default: 1247 nextToken(); 1248 break; 1249 } 1250 } while (!eof()); 1251 } 1252 1253 void UnwrappedLineParser::parseIfThenElse() { 1254 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected"); 1255 nextToken(); 1256 if (FormatTok->Tok.is(tok::l_paren)) 1257 parseParens(); 1258 bool NeedsUnwrappedLine = false; 1259 if (FormatTok->Tok.is(tok::l_brace)) { 1260 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1261 parseBlock(/*MustBeDeclaration=*/false); 1262 if (Style.BraceWrapping.BeforeElse) 1263 addUnwrappedLine(); 1264 else 1265 NeedsUnwrappedLine = true; 1266 } else { 1267 addUnwrappedLine(); 1268 ++Line->Level; 1269 parseStructuralElement(); 1270 --Line->Level; 1271 } 1272 if (FormatTok->Tok.is(tok::kw_else)) { 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.BraceWrapping.BeforeCatch) { 1314 addUnwrappedLine(); 1315 } else { 1316 NeedsUnwrappedLine = true; 1317 } 1318 } else if (!FormatTok->is(tok::kw_catch)) { 1319 // The C++ standard requires a compound-statement after a try. 1320 // If there's none, we try to assume there's a structuralElement 1321 // and try to continue. 1322 addUnwrappedLine(); 1323 ++Line->Level; 1324 parseStructuralElement(); 1325 --Line->Level; 1326 } 1327 while (1) { 1328 if (FormatTok->is(tok::at)) 1329 nextToken(); 1330 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except, 1331 tok::kw___finally) || 1332 ((Style.Language == FormatStyle::LK_Java || 1333 Style.Language == FormatStyle::LK_JavaScript) && 1334 FormatTok->is(Keywords.kw_finally)) || 1335 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) || 1336 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally)))) 1337 break; 1338 nextToken(); 1339 while (FormatTok->isNot(tok::l_brace)) { 1340 if (FormatTok->is(tok::l_paren)) { 1341 parseParens(); 1342 continue; 1343 } 1344 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof)) 1345 return; 1346 nextToken(); 1347 } 1348 NeedsUnwrappedLine = false; 1349 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1350 parseBlock(/*MustBeDeclaration=*/false); 1351 if (Style.BraceWrapping.BeforeCatch) 1352 addUnwrappedLine(); 1353 else 1354 NeedsUnwrappedLine = true; 1355 } 1356 if (NeedsUnwrappedLine) 1357 addUnwrappedLine(); 1358 } 1359 1360 void UnwrappedLineParser::parseNamespace() { 1361 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected"); 1362 1363 const FormatToken &InitialToken = *FormatTok; 1364 nextToken(); 1365 if (FormatTok->Tok.is(tok::identifier)) 1366 nextToken(); 1367 if (FormatTok->Tok.is(tok::l_brace)) { 1368 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1369 addUnwrappedLine(); 1370 1371 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All || 1372 (Style.NamespaceIndentation == FormatStyle::NI_Inner && 1373 DeclarationScopeStack.size() > 1); 1374 parseBlock(/*MustBeDeclaration=*/true, AddLevel); 1375 // Munch the semicolon after a namespace. This is more common than one would 1376 // think. Puttin the semicolon into its own line is very ugly. 1377 if (FormatTok->Tok.is(tok::semi)) 1378 nextToken(); 1379 addUnwrappedLine(); 1380 } 1381 // FIXME: Add error handling. 1382 } 1383 1384 void UnwrappedLineParser::parseNew() { 1385 assert(FormatTok->is(tok::kw_new) && "'new' expected"); 1386 nextToken(); 1387 if (Style.Language != FormatStyle::LK_Java) 1388 return; 1389 1390 // In Java, we can parse everything up to the parens, which aren't optional. 1391 do { 1392 // There should not be a ;, { or } before the new's open paren. 1393 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace)) 1394 return; 1395 1396 // Consume the parens. 1397 if (FormatTok->is(tok::l_paren)) { 1398 parseParens(); 1399 1400 // If there is a class body of an anonymous class, consume that as child. 1401 if (FormatTok->is(tok::l_brace)) 1402 parseChildBlock(); 1403 return; 1404 } 1405 nextToken(); 1406 } while (!eof()); 1407 } 1408 1409 void UnwrappedLineParser::parseForOrWhileLoop() { 1410 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) && 1411 "'for', 'while' or foreach macro expected"); 1412 nextToken(); 1413 if (FormatTok->Tok.is(tok::l_paren)) 1414 parseParens(); 1415 if (FormatTok->Tok.is(tok::l_brace)) { 1416 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1417 parseBlock(/*MustBeDeclaration=*/false); 1418 addUnwrappedLine(); 1419 } else { 1420 addUnwrappedLine(); 1421 ++Line->Level; 1422 parseStructuralElement(); 1423 --Line->Level; 1424 } 1425 } 1426 1427 void UnwrappedLineParser::parseDoWhile() { 1428 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected"); 1429 nextToken(); 1430 if (FormatTok->Tok.is(tok::l_brace)) { 1431 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1432 parseBlock(/*MustBeDeclaration=*/false); 1433 if (Style.BraceWrapping.IndentBraces) 1434 addUnwrappedLine(); 1435 } else { 1436 addUnwrappedLine(); 1437 ++Line->Level; 1438 parseStructuralElement(); 1439 --Line->Level; 1440 } 1441 1442 // FIXME: Add error handling. 1443 if (!FormatTok->Tok.is(tok::kw_while)) { 1444 addUnwrappedLine(); 1445 return; 1446 } 1447 1448 nextToken(); 1449 parseStructuralElement(); 1450 } 1451 1452 void UnwrappedLineParser::parseLabel() { 1453 nextToken(); 1454 unsigned OldLineLevel = Line->Level; 1455 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0)) 1456 --Line->Level; 1457 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) { 1458 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1459 parseBlock(/*MustBeDeclaration=*/false); 1460 if (FormatTok->Tok.is(tok::kw_break)) { 1461 if (Style.BraceWrapping.AfterControlStatement) 1462 addUnwrappedLine(); 1463 parseStructuralElement(); 1464 } 1465 addUnwrappedLine(); 1466 } else { 1467 if (FormatTok->is(tok::semi)) 1468 nextToken(); 1469 addUnwrappedLine(); 1470 } 1471 Line->Level = OldLineLevel; 1472 } 1473 1474 void UnwrappedLineParser::parseCaseLabel() { 1475 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected"); 1476 // FIXME: fix handling of complex expressions here. 1477 do { 1478 nextToken(); 1479 } while (!eof() && !FormatTok->Tok.is(tok::colon)); 1480 parseLabel(); 1481 } 1482 1483 void UnwrappedLineParser::parseSwitch() { 1484 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected"); 1485 nextToken(); 1486 if (FormatTok->Tok.is(tok::l_paren)) 1487 parseParens(); 1488 if (FormatTok->Tok.is(tok::l_brace)) { 1489 CompoundStatementIndenter Indenter(this, Style, Line->Level); 1490 parseBlock(/*MustBeDeclaration=*/false); 1491 addUnwrappedLine(); 1492 } else { 1493 addUnwrappedLine(); 1494 ++Line->Level; 1495 parseStructuralElement(); 1496 --Line->Level; 1497 } 1498 } 1499 1500 void UnwrappedLineParser::parseAccessSpecifier() { 1501 nextToken(); 1502 // Understand Qt's slots. 1503 if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots)) 1504 nextToken(); 1505 // Otherwise, we don't know what it is, and we'd better keep the next token. 1506 if (FormatTok->Tok.is(tok::colon)) 1507 nextToken(); 1508 addUnwrappedLine(); 1509 } 1510 1511 void UnwrappedLineParser::parseEnum() { 1512 // Won't be 'enum' for NS_ENUMs. 1513 if (FormatTok->Tok.is(tok::kw_enum)) 1514 nextToken(); 1515 1516 // Eat up enum class ... 1517 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct)) 1518 nextToken(); 1519 1520 while (FormatTok->Tok.getIdentifierInfo() || 1521 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less, 1522 tok::greater, tok::comma, tok::question)) { 1523 nextToken(); 1524 // We can have macros or attributes in between 'enum' and the enum name. 1525 if (FormatTok->is(tok::l_paren)) 1526 parseParens(); 1527 if (FormatTok->is(tok::identifier)) { 1528 nextToken(); 1529 // If there are two identifiers in a row, this is likely an elaborate 1530 // return type. In Java, this can be "implements", etc. 1531 if (Style.Language == FormatStyle::LK_Cpp && 1532 FormatTok->is(tok::identifier)) 1533 return; 1534 } 1535 } 1536 1537 // Just a declaration or something is wrong. 1538 if (FormatTok->isNot(tok::l_brace)) 1539 return; 1540 FormatTok->BlockKind = BK_Block; 1541 1542 if (Style.Language == FormatStyle::LK_Java) { 1543 // Java enums are different. 1544 parseJavaEnumBody(); 1545 return; 1546 } else if (Style.Language == FormatStyle::LK_Proto) { 1547 parseBlock(/*MustBeDeclaration=*/true); 1548 return; 1549 } 1550 1551 // Parse enum body. 1552 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true); 1553 if (HasError) { 1554 if (FormatTok->is(tok::semi)) 1555 nextToken(); 1556 addUnwrappedLine(); 1557 } 1558 1559 // There is no addUnwrappedLine() here so that we fall through to parsing a 1560 // structural element afterwards. Thus, in "enum A {} n, m;", 1561 // "} n, m;" will end up in one unwrapped line. 1562 } 1563 1564 void UnwrappedLineParser::parseJavaEnumBody() { 1565 // Determine whether the enum is simple, i.e. does not have a semicolon or 1566 // constants with class bodies. Simple enums can be formatted like braced 1567 // lists, contracted to a single line, etc. 1568 unsigned StoredPosition = Tokens->getPosition(); 1569 bool IsSimple = true; 1570 FormatToken *Tok = Tokens->getNextToken(); 1571 while (Tok) { 1572 if (Tok->is(tok::r_brace)) 1573 break; 1574 if (Tok->isOneOf(tok::l_brace, tok::semi)) { 1575 IsSimple = false; 1576 break; 1577 } 1578 // FIXME: This will also mark enums with braces in the arguments to enum 1579 // constants as "not simple". This is probably fine in practice, though. 1580 Tok = Tokens->getNextToken(); 1581 } 1582 FormatTok = Tokens->setPosition(StoredPosition); 1583 1584 if (IsSimple) { 1585 parseBracedList(); 1586 addUnwrappedLine(); 1587 return; 1588 } 1589 1590 // Parse the body of a more complex enum. 1591 // First add a line for everything up to the "{". 1592 nextToken(); 1593 addUnwrappedLine(); 1594 ++Line->Level; 1595 1596 // Parse the enum constants. 1597 while (FormatTok) { 1598 if (FormatTok->is(tok::l_brace)) { 1599 // Parse the constant's class body. 1600 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1601 /*MunchSemi=*/false); 1602 } else if (FormatTok->is(tok::l_paren)) { 1603 parseParens(); 1604 } else if (FormatTok->is(tok::comma)) { 1605 nextToken(); 1606 addUnwrappedLine(); 1607 } else if (FormatTok->is(tok::semi)) { 1608 nextToken(); 1609 addUnwrappedLine(); 1610 break; 1611 } else if (FormatTok->is(tok::r_brace)) { 1612 addUnwrappedLine(); 1613 break; 1614 } else { 1615 nextToken(); 1616 } 1617 } 1618 1619 // Parse the class body after the enum's ";" if any. 1620 parseLevel(/*HasOpeningBrace=*/true); 1621 nextToken(); 1622 --Line->Level; 1623 addUnwrappedLine(); 1624 } 1625 1626 void UnwrappedLineParser::parseRecord() { 1627 const FormatToken &InitialToken = *FormatTok; 1628 nextToken(); 1629 1630 // The actual identifier can be a nested name specifier, and in macros 1631 // it is often token-pasted. 1632 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash, 1633 tok::kw___attribute, tok::kw___declspec, 1634 tok::kw_alignas) || 1635 ((Style.Language == FormatStyle::LK_Java || 1636 Style.Language == FormatStyle::LK_JavaScript) && 1637 FormatTok->isOneOf(tok::period, tok::comma))) { 1638 bool IsNonMacroIdentifier = 1639 FormatTok->is(tok::identifier) && 1640 FormatTok->TokenText != FormatTok->TokenText.upper(); 1641 nextToken(); 1642 // We can have macros or attributes in between 'class' and the class name. 1643 if (!IsNonMacroIdentifier && FormatTok->Tok.is(tok::l_paren)) 1644 parseParens(); 1645 } 1646 1647 // Note that parsing away template declarations here leads to incorrectly 1648 // accepting function declarations as record declarations. 1649 // In general, we cannot solve this problem. Consider: 1650 // class A<int> B() {} 1651 // which can be a function definition or a class definition when B() is a 1652 // macro. If we find enough real-world cases where this is a problem, we 1653 // can parse for the 'template' keyword in the beginning of the statement, 1654 // and thus rule out the record production in case there is no template 1655 // (this would still leave us with an ambiguity between template function 1656 // and class declarations). 1657 if (FormatTok->isOneOf(tok::colon, tok::less)) { 1658 while (!eof()) { 1659 if (FormatTok->is(tok::l_brace)) { 1660 calculateBraceTypes(/*ExpectClassBody=*/true); 1661 if (!tryToParseBracedList()) 1662 break; 1663 } 1664 if (FormatTok->Tok.is(tok::semi)) 1665 return; 1666 nextToken(); 1667 } 1668 } 1669 if (FormatTok->Tok.is(tok::l_brace)) { 1670 if (ShouldBreakBeforeBrace(Style, InitialToken)) 1671 addUnwrappedLine(); 1672 1673 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true, 1674 /*MunchSemi=*/false); 1675 } 1676 // There is no addUnwrappedLine() here so that we fall through to parsing a 1677 // structural element afterwards. Thus, in "class A {} n, m;", 1678 // "} n, m;" will end up in one unwrapped line. 1679 } 1680 1681 void UnwrappedLineParser::parseObjCProtocolList() { 1682 assert(FormatTok->Tok.is(tok::less) && "'<' expected."); 1683 do 1684 nextToken(); 1685 while (!eof() && FormatTok->Tok.isNot(tok::greater)); 1686 nextToken(); // Skip '>'. 1687 } 1688 1689 void UnwrappedLineParser::parseObjCUntilAtEnd() { 1690 do { 1691 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) { 1692 nextToken(); 1693 addUnwrappedLine(); 1694 break; 1695 } 1696 if (FormatTok->is(tok::l_brace)) { 1697 parseBlock(/*MustBeDeclaration=*/false); 1698 // In ObjC interfaces, nothing should be following the "}". 1699 addUnwrappedLine(); 1700 } else if (FormatTok->is(tok::r_brace)) { 1701 // Ignore stray "}". parseStructuralElement doesn't consume them. 1702 nextToken(); 1703 addUnwrappedLine(); 1704 } else { 1705 parseStructuralElement(); 1706 } 1707 } while (!eof()); 1708 } 1709 1710 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() { 1711 nextToken(); 1712 nextToken(); // interface name 1713 1714 // @interface can be followed by either a base class, or a category. 1715 if (FormatTok->Tok.is(tok::colon)) { 1716 nextToken(); 1717 nextToken(); // base class name 1718 } else if (FormatTok->Tok.is(tok::l_paren)) 1719 // Skip category, if present. 1720 parseParens(); 1721 1722 if (FormatTok->Tok.is(tok::less)) 1723 parseObjCProtocolList(); 1724 1725 if (FormatTok->Tok.is(tok::l_brace)) { 1726 if (Style.BraceWrapping.AfterObjCDeclaration) 1727 addUnwrappedLine(); 1728 parseBlock(/*MustBeDeclaration=*/true); 1729 } 1730 1731 // With instance variables, this puts '}' on its own line. Without instance 1732 // variables, this ends the @interface line. 1733 addUnwrappedLine(); 1734 1735 parseObjCUntilAtEnd(); 1736 } 1737 1738 void UnwrappedLineParser::parseObjCProtocol() { 1739 nextToken(); 1740 nextToken(); // protocol name 1741 1742 if (FormatTok->Tok.is(tok::less)) 1743 parseObjCProtocolList(); 1744 1745 // Check for protocol declaration. 1746 if (FormatTok->Tok.is(tok::semi)) { 1747 nextToken(); 1748 return addUnwrappedLine(); 1749 } 1750 1751 addUnwrappedLine(); 1752 parseObjCUntilAtEnd(); 1753 } 1754 1755 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() { 1756 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export)); 1757 nextToken(); 1758 1759 // Consume the "default" in "export default class/function". 1760 if (FormatTok->is(tok::kw_default)) 1761 nextToken(); 1762 1763 // Consume "function" and "default function", so that these get parsed as 1764 // free-standing JS functions, i.e. do not require a trailing semicolon. 1765 if (FormatTok->is(Keywords.kw_function)) { 1766 nextToken(); 1767 return; 1768 } 1769 1770 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, tok::kw_enum, 1771 Keywords.kw_let, Keywords.kw_var)) 1772 return; // Fall through to parsing the corresponding structure. 1773 1774 if (FormatTok->is(tok::l_brace)) { 1775 FormatTok->BlockKind = BK_Block; 1776 parseBracedList(); 1777 } 1778 1779 while (!eof() && FormatTok->isNot(tok::semi) && 1780 FormatTok->isNot(tok::l_brace)) { 1781 nextToken(); 1782 } 1783 } 1784 1785 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line, 1786 StringRef Prefix = "") { 1787 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")" 1788 << (Line.InPPDirective ? " MACRO" : "") << ": "; 1789 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1790 E = Line.Tokens.end(); 1791 I != E; ++I) { 1792 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] "; 1793 } 1794 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(), 1795 E = Line.Tokens.end(); 1796 I != E; ++I) { 1797 const UnwrappedLineNode &Node = *I; 1798 for (SmallVectorImpl<UnwrappedLine>::const_iterator 1799 I = Node.Children.begin(), 1800 E = Node.Children.end(); 1801 I != E; ++I) { 1802 printDebugInfo(*I, "\nChild: "); 1803 } 1804 } 1805 llvm::dbgs() << "\n"; 1806 } 1807 1808 void UnwrappedLineParser::addUnwrappedLine() { 1809 if (Line->Tokens.empty()) 1810 return; 1811 DEBUG({ 1812 if (CurrentLines == &Lines) 1813 printDebugInfo(*Line); 1814 }); 1815 CurrentLines->push_back(std::move(*Line)); 1816 Line->Tokens.clear(); 1817 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) { 1818 CurrentLines->append( 1819 std::make_move_iterator(PreprocessorDirectives.begin()), 1820 std::make_move_iterator(PreprocessorDirectives.end())); 1821 PreprocessorDirectives.clear(); 1822 } 1823 } 1824 1825 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); } 1826 1827 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) { 1828 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) && 1829 FormatTok.NewlinesBefore > 0; 1830 } 1831 1832 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) { 1833 bool JustComments = Line->Tokens.empty(); 1834 for (SmallVectorImpl<FormatToken *>::const_iterator 1835 I = CommentsBeforeNextToken.begin(), 1836 E = CommentsBeforeNextToken.end(); 1837 I != E; ++I) { 1838 if (isOnNewLine(**I) && JustComments) 1839 addUnwrappedLine(); 1840 pushToken(*I); 1841 } 1842 if (NewlineBeforeNext && JustComments) 1843 addUnwrappedLine(); 1844 CommentsBeforeNextToken.clear(); 1845 } 1846 1847 void UnwrappedLineParser::nextToken() { 1848 if (eof()) 1849 return; 1850 flushComments(isOnNewLine(*FormatTok)); 1851 pushToken(FormatTok); 1852 readToken(); 1853 } 1854 1855 void UnwrappedLineParser::readToken() { 1856 bool CommentsInCurrentLine = true; 1857 do { 1858 FormatTok = Tokens->getNextToken(); 1859 assert(FormatTok); 1860 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) && 1861 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) { 1862 // If there is an unfinished unwrapped line, we flush the preprocessor 1863 // directives only after that unwrapped line was finished later. 1864 bool SwitchToPreprocessorLines = !Line->Tokens.empty(); 1865 ScopedLineState BlockState(*this, SwitchToPreprocessorLines); 1866 // Comments stored before the preprocessor directive need to be output 1867 // before the preprocessor directive, at the same level as the 1868 // preprocessor directive, as we consider them to apply to the directive. 1869 flushComments(isOnNewLine(*FormatTok)); 1870 parsePPDirective(); 1871 } 1872 while (FormatTok->Type == TT_ConflictStart || 1873 FormatTok->Type == TT_ConflictEnd || 1874 FormatTok->Type == TT_ConflictAlternative) { 1875 if (FormatTok->Type == TT_ConflictStart) { 1876 conditionalCompilationStart(/*Unreachable=*/false); 1877 } else if (FormatTok->Type == TT_ConflictAlternative) { 1878 conditionalCompilationAlternative(); 1879 } else if (FormatTok->Type == TT_ConflictEnd) { 1880 conditionalCompilationEnd(); 1881 } 1882 FormatTok = Tokens->getNextToken(); 1883 FormatTok->MustBreakBefore = true; 1884 } 1885 1886 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) && 1887 !Line->InPPDirective) { 1888 continue; 1889 } 1890 1891 if (!FormatTok->Tok.is(tok::comment)) 1892 return; 1893 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) { 1894 CommentsInCurrentLine = false; 1895 } 1896 if (CommentsInCurrentLine) { 1897 pushToken(FormatTok); 1898 } else { 1899 CommentsBeforeNextToken.push_back(FormatTok); 1900 } 1901 } while (!eof()); 1902 } 1903 1904 void UnwrappedLineParser::pushToken(FormatToken *Tok) { 1905 Line->Tokens.push_back(UnwrappedLineNode(Tok)); 1906 if (MustBreakBeforeNextToken) { 1907 Line->Tokens.back().Tok->MustBreakBefore = true; 1908 MustBreakBeforeNextToken = false; 1909 } 1910 } 1911 1912 } // end namespace format 1913 } // end namespace clang 1914