1 //===--- UnwrappedLineParser.cpp - Format C++ code ------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
11 /// \brief This file contains the implementation of the UnwrappedLineParser,
12 /// which turns a stream of tokens into UnwrappedLines.
14 //===----------------------------------------------------------------------===//
16 #include "UnwrappedLineParser.h"
17 #include "llvm/ADT/STLExtras.h"
18 #include "llvm/Support/Debug.h"
19 #include "llvm/Support/raw_ostream.h"
21 #define DEBUG_TYPE "format-parser"
26 class FormatTokenSource {
28 virtual ~FormatTokenSource() {}
29 virtual FormatToken *getNextToken() = 0;
31 virtual unsigned getPosition() = 0;
32 virtual FormatToken *setPosition(unsigned Position) = 0;
37 class ScopedDeclarationState {
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);
45 ~ScopedDeclarationState() {
48 Line.MustBeDeclaration = Stack.back();
50 Line.MustBeDeclaration = true;
55 std::vector<bool> &Stack;
58 class ScopedMacroState : public FormatTokenSource {
60 ScopedMacroState(UnwrappedLine &Line, FormatTokenSource *&TokenSource,
61 FormatToken *&ResetToken)
62 : Line(Line), TokenSource(TokenSource), ResetToken(ResetToken),
63 PreviousLineLevel(Line.Level), PreviousTokenSource(TokenSource),
67 Line.InPPDirective = true;
70 ~ScopedMacroState() override {
71 TokenSource = PreviousTokenSource;
73 Line.InPPDirective = false;
74 Line.Level = PreviousLineLevel;
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.
81 Token = PreviousTokenSource->getNextToken();
87 unsigned getPosition() override { return PreviousTokenSource->getPosition(); }
89 FormatToken *setPosition(unsigned Position) override {
90 Token = PreviousTokenSource->setPosition(Position);
95 bool eof() { return Token && Token->HasUnescapedNewline; }
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;
109 FormatTokenSource *&TokenSource;
110 FormatToken *&ResetToken;
111 unsigned PreviousLineLevel;
112 FormatTokenSource *PreviousTokenSource;
117 } // end anonymous namespace
119 class ScopedLineState {
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;
135 if (!Parser.Line->Tokens.empty()) {
136 Parser.addUnwrappedLine();
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;
146 UnwrappedLineParser &Parser;
148 std::unique_ptr<UnwrappedLine> PreBlockLine;
149 SmallVectorImpl<UnwrappedLine> *OriginalLines;
152 class CompoundStatementIndenter {
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)
162 ~CompoundStatementIndenter() { LineLevel = OldLineLevel; }
166 unsigned OldLineLevel;
171 class IndexedTokenSource : public FormatTokenSource {
173 IndexedTokenSource(ArrayRef<FormatToken *> Tokens)
174 : Tokens(Tokens), Position(-1) {}
176 FormatToken *getNextToken() override {
178 return Tokens[Position];
181 unsigned getPosition() override {
182 assert(Position >= 0);
186 FormatToken *setPosition(unsigned P) override {
188 return Tokens[Position];
191 void reset() { Position = -1; }
194 ArrayRef<FormatToken *> Tokens;
198 } // end anonymous namespace
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) {}
208 void UnwrappedLineParser::reset() {
210 Line.reset(new UnwrappedLine);
211 CommentsBeforeNextToken.clear();
213 MustBreakBeforeNextToken = false;
214 PreprocessorDirectives.clear();
215 CurrentLines = &Lines;
216 DeclarationScopeStack.clear();
220 void UnwrappedLineParser::parse() {
221 IndexedTokenSource TokenSource(AllTokens);
223 DEBUG(llvm::dbgs() << "----\n");
225 Tokens = &TokenSource;
230 // Create line with eof token.
231 pushToken(FormatTok);
234 for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(),
237 Callback.consumeUnwrappedLine(*I);
239 Callback.finishRun();
241 while (!PPLevelBranchIndex.empty() &&
242 PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) {
243 PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1);
244 PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1);
246 if (!PPLevelBranchIndex.empty()) {
247 ++PPLevelBranchIndex.back();
248 assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size());
249 assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back());
251 } while (!PPLevelBranchIndex.empty());
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,
261 parseLevel(/*HasOpeningBrace=*/false);
262 // Make sure to format the remaining tokens.
267 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) {
268 bool SwitchLabelEncountered = false;
270 tok::TokenKind kind = FormatTok->Tok.getKind();
271 if (FormatTok->Type == TT_MacroBlockBegin) {
273 } else if (FormatTok->Type == TT_MacroBlockEnd) {
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())
287 parseBlock(/*MustBeDeclaration=*/false);
296 case tok::kw_default:
298 if (!SwitchLabelEncountered &&
299 (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1)))
301 SwitchLabelEncountered = true;
302 parseStructuralElement();
305 parseStructuralElement();
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
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));
325 // Get next non-comment token.
326 FormatToken *NextTok;
327 unsigned ReadTokens = 0;
329 NextTok = Tokens->getNextToken();
331 } while (NextTok->is(tok::comment));
333 switch (Tok->Tok.getKind()) {
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;
341 Tok->BlockKind = BK_Unknown;
342 LBraceStack.push_back(Tok);
345 if (LBraceStack.empty())
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);
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;
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).
364 (Style.Language == FormatStyle::LK_JavaScript &&
365 NextTok->isOneOf(Keywords.kw_of, Keywords.kw_in,
367 NextTok->isOneOf(tok::comma, tok::period, tok::colon,
368 tok::r_paren, tok::r_square, tok::l_brace,
369 tok::l_square, tok::l_paren, tok::ellipsis) ||
370 (NextTok->is(tok::identifier) &&
371 !PrevTok->isOneOf(tok::semi, tok::r_brace, tok::l_brace)) ||
372 (NextTok->is(tok::semi) &&
373 (!ExpectClassBody || LBraceStack.size() != 1)) ||
374 (NextTok->isBinaryOperator() && !NextIsObjCMethod);
376 if (ProbablyBracedList) {
377 Tok->BlockKind = BK_BracedInit;
378 LBraceStack.back()->BlockKind = BK_BracedInit;
380 Tok->BlockKind = BK_Block;
381 LBraceStack.back()->BlockKind = BK_Block;
384 LBraceStack.pop_back();
394 if (!LBraceStack.empty() && LBraceStack.back()->BlockKind == BK_Unknown)
395 LBraceStack.back()->BlockKind = BK_Block;
402 } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty());
404 // Assume other blocks for all unclosed opening braces.
405 for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) {
406 if (LBraceStack[i]->BlockKind == BK_Unknown)
407 LBraceStack[i]->BlockKind = BK_Block;
410 FormatTok = Tokens->setPosition(StoredPosition);
413 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel,
415 assert(FormatTok->isOneOf(tok::l_brace, TT_MacroBlockBegin) &&
416 "'{' or macro block token expected");
417 const bool MacroBlock = FormatTok->is(TT_MacroBlockBegin);
418 FormatTok->BlockKind = BK_Block;
420 unsigned InitialLevel = Line->Level;
423 if (MacroBlock && FormatTok->is(tok::l_paren))
428 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack,
432 parseLevel(/*HasOpeningBrace=*/true);
437 if (MacroBlock ? !FormatTok->is(TT_MacroBlockEnd)
438 : !FormatTok->is(tok::r_brace)) {
439 Line->Level = InitialLevel;
440 FormatTok->BlockKind = BK_Block;
444 nextToken(); // Munch the closing brace.
446 if (MacroBlock && FormatTok->is(tok::l_paren))
449 if (MunchSemi && FormatTok->Tok.is(tok::semi))
451 Line->Level = InitialLevel;
454 static bool isGoogScope(const UnwrappedLine &Line) {
455 // FIXME: Closure-library specific stuff should not be hard-coded but be
457 if (Line.Tokens.size() < 4)
459 auto I = Line.Tokens.begin();
460 if (I->Tok->TokenText != "goog")
463 if (I->Tok->isNot(tok::period))
466 if (I->Tok->TokenText != "scope")
469 return I->Tok->is(tok::l_paren);
472 static bool ShouldBreakBeforeBrace(const FormatStyle &Style,
473 const FormatToken &InitialToken) {
474 if (InitialToken.is(tok::kw_namespace))
475 return Style.BraceWrapping.AfterNamespace;
476 if (InitialToken.is(tok::kw_class))
477 return Style.BraceWrapping.AfterClass;
478 if (InitialToken.is(tok::kw_union))
479 return Style.BraceWrapping.AfterUnion;
480 if (InitialToken.is(tok::kw_struct))
481 return Style.BraceWrapping.AfterStruct;
485 void UnwrappedLineParser::parseChildBlock() {
486 FormatTok->BlockKind = BK_Block;
490 Style.Language == FormatStyle::LK_JavaScript && isGoogScope(*Line);
491 ScopedLineState LineState(*this);
492 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack,
493 /*MustBeDeclaration=*/false);
494 Line->Level += GoogScope ? 0 : 1;
495 parseLevel(/*HasOpeningBrace=*/true);
496 flushComments(isOnNewLine(*FormatTok));
497 Line->Level -= GoogScope ? 0 : 1;
502 void UnwrappedLineParser::parsePPDirective() {
503 assert(FormatTok->Tok.is(tok::hash) && "'#' expected");
504 ScopedMacroState MacroState(*Line, Tokens, FormatTok);
507 if (!FormatTok->Tok.getIdentifierInfo()) {
512 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) {
517 parsePPIf(/*IfDef=*/false);
521 parsePPIf(/*IfDef=*/true);
538 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) {
539 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable))
540 PPStack.push_back(PP_Unreachable);
542 PPStack.push_back(PP_Conditional);
545 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) {
547 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size());
548 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) {
549 PPLevelBranchIndex.push_back(0);
550 PPLevelBranchCount.push_back(0);
552 PPChainBranchIndex.push(0);
553 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0;
554 conditionalCompilationCondition(Unreachable || Skip);
557 void UnwrappedLineParser::conditionalCompilationAlternative() {
558 if (!PPStack.empty())
560 assert(PPBranchLevel < (int)PPLevelBranchIndex.size());
561 if (!PPChainBranchIndex.empty())
562 ++PPChainBranchIndex.top();
563 conditionalCompilationCondition(
564 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() &&
565 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top());
568 void UnwrappedLineParser::conditionalCompilationEnd() {
569 assert(PPBranchLevel < (int)PPLevelBranchIndex.size());
570 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) {
571 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) {
572 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1;
575 // Guard against #endif's without #if.
576 if (PPBranchLevel > 0)
578 if (!PPChainBranchIndex.empty())
579 PPChainBranchIndex.pop();
580 if (!PPStack.empty())
584 void UnwrappedLineParser::parsePPIf(bool IfDef) {
586 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() &&
587 FormatTok->Tok.getLiteralData() != nullptr &&
588 StringRef(FormatTok->Tok.getLiteralData(),
589 FormatTok->Tok.getLength()) == "0") ||
590 FormatTok->Tok.is(tok::kw_false);
591 conditionalCompilationStart(!IfDef && IsLiteralFalse);
595 void UnwrappedLineParser::parsePPElse() {
596 conditionalCompilationAlternative();
600 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); }
602 void UnwrappedLineParser::parsePPEndIf() {
603 conditionalCompilationEnd();
607 void UnwrappedLineParser::parsePPDefine() {
610 if (FormatTok->Tok.getKind() != tok::identifier) {
615 if (FormatTok->Tok.getKind() == tok::l_paren &&
616 FormatTok->WhitespaceRange.getBegin() ==
617 FormatTok->WhitespaceRange.getEnd()) {
623 // Errors during a preprocessor directive can only affect the layout of the
624 // preprocessor directive, and thus we ignore them. An alternative approach
625 // would be to use the same approach we use on the file level (no
626 // re-indentation if there was a structural error) within the macro
631 void UnwrappedLineParser::parsePPUnknown() {
638 // Here we blacklist certain tokens that are not usually the first token in an
639 // unwrapped line. This is used in attempt to distinguish macro calls without
640 // trailing semicolons from other constructs split to several lines.
641 static bool tokenCanStartNewLine(const clang::Token &Tok) {
642 // Semicolon can be a null-statement, l_square can be a start of a macro or
643 // a C++11 attribute, but this doesn't seem to be common.
644 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) &&
645 Tok.isNot(tok::l_square) &&
646 // Tokens that can only be used as binary operators and a part of
647 // overloaded operator names.
648 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) &&
649 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) &&
650 Tok.isNot(tok::less) && Tok.isNot(tok::greater) &&
651 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) &&
652 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) &&
653 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) &&
654 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) &&
655 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) &&
656 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) &&
657 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) &&
658 Tok.isNot(tok::lesslessequal) &&
659 // Colon is used in labels, base class lists, initializer lists,
660 // range-based for loops, ternary operator, but should never be the
661 // first token in an unwrapped line.
662 Tok.isNot(tok::colon) &&
663 // 'noexcept' is a trailing annotation.
664 Tok.isNot(tok::kw_noexcept);
667 static bool mustBeJSIdent(const AdditionalKeywords &Keywords,
668 const FormatToken *FormatTok) {
669 // FIXME: This returns true for C/C++ keywords like 'struct'.
670 return FormatTok->is(tok::identifier) &&
671 (FormatTok->Tok.getIdentifierInfo() == nullptr ||
673 Keywords.kw_in, Keywords.kw_of, Keywords.kw_as, Keywords.kw_async,
674 Keywords.kw_await, Keywords.kw_yield, Keywords.kw_finally,
675 Keywords.kw_function, Keywords.kw_import, Keywords.kw_is,
676 Keywords.kw_let, Keywords.kw_var, tok::kw_const,
677 Keywords.kw_abstract, Keywords.kw_extends, Keywords.kw_implements,
678 Keywords.kw_instanceof, Keywords.kw_interface,
679 Keywords.kw_throws));
682 static bool mustBeJSIdentOrValue(const AdditionalKeywords &Keywords,
683 const FormatToken *FormatTok) {
684 return FormatTok->Tok.isLiteral() ||
685 FormatTok->isOneOf(tok::kw_true, tok::kw_false) ||
686 mustBeJSIdent(Keywords, FormatTok);
689 // isJSDeclOrStmt returns true if |FormatTok| starts a declaration or statement
690 // when encountered after a value (see mustBeJSIdentOrValue).
691 static bool isJSDeclOrStmt(const AdditionalKeywords &Keywords,
692 const FormatToken *FormatTok) {
693 return FormatTok->isOneOf(
694 tok::kw_return, Keywords.kw_yield,
696 tok::kw_if, tok::kw_else,
698 tok::kw_for, tok::kw_while, tok::kw_do, tok::kw_continue, tok::kw_break,
700 tok::kw_switch, tok::kw_case,
702 tok::kw_throw, tok::kw_try, tok::kw_catch, Keywords.kw_finally,
704 tok::kw_const, tok::kw_class, Keywords.kw_var, Keywords.kw_let,
705 Keywords.kw_async, Keywords.kw_function,
707 Keywords.kw_import, tok::kw_export);
710 // readTokenWithJavaScriptASI reads the next token and terminates the current
711 // line if JavaScript Automatic Semicolon Insertion must
712 // happen between the current token and the next token.
714 // This method is conservative - it cannot cover all edge cases of JavaScript,
715 // but only aims to correctly handle certain well known cases. It *must not*
716 // return true in speculative cases.
717 void UnwrappedLineParser::readTokenWithJavaScriptASI() {
718 FormatToken *Previous = FormatTok;
720 FormatToken *Next = FormatTok;
723 CommentsBeforeNextToken.empty()
724 ? Next->NewlinesBefore == 0
725 : CommentsBeforeNextToken.front()->NewlinesBefore == 0;
729 bool PreviousMustBeValue = mustBeJSIdentOrValue(Keywords, Previous);
730 bool PreviousStartsTemplateExpr =
731 Previous->is(TT_TemplateString) && Previous->TokenText.endswith("${");
732 if (PreviousMustBeValue && Line && Line->Tokens.size() > 1) {
733 // If the token before the previous one is an '@', the previous token is an
734 // annotation and can precede another identifier/value.
735 const FormatToken *PrePrevious = std::prev(Line->Tokens.end(), 2)->Tok;
736 if (PrePrevious->is(tok::at))
739 if (Next->is(tok::exclaim) && PreviousMustBeValue)
740 return addUnwrappedLine();
741 bool NextMustBeValue = mustBeJSIdentOrValue(Keywords, Next);
742 bool NextEndsTemplateExpr =
743 Next->is(TT_TemplateString) && Next->TokenText.startswith("}");
744 if (NextMustBeValue && !NextEndsTemplateExpr && !PreviousStartsTemplateExpr &&
745 (PreviousMustBeValue ||
746 Previous->isOneOf(tok::r_square, tok::r_paren, tok::plusplus,
748 return addUnwrappedLine();
749 if ((PreviousMustBeValue || Previous->is(tok::r_brace)) &&
750 isJSDeclOrStmt(Keywords, Next))
751 return addUnwrappedLine();
754 void UnwrappedLineParser::parseStructuralElement() {
755 assert(!FormatTok->is(tok::l_brace));
756 if (Style.Language == FormatStyle::LK_TableGen &&
757 FormatTok->is(tok::pp_include)) {
759 if (FormatTok->is(tok::string_literal))
764 switch (FormatTok->Tok.getKind()) {
767 if (FormatTok->Tok.is(tok::l_brace)) {
771 switch (FormatTok->Tok.getObjCKeywordID()) {
772 case tok::objc_public:
773 case tok::objc_protected:
774 case tok::objc_package:
775 case tok::objc_private:
776 return parseAccessSpecifier();
777 case tok::objc_interface:
778 case tok::objc_implementation:
779 return parseObjCInterfaceOrImplementation();
780 case tok::objc_protocol:
781 return parseObjCProtocol();
783 return; // Handled by the caller.
784 case tok::objc_optional:
785 case tok::objc_required:
789 case tok::objc_autoreleasepool:
791 if (FormatTok->Tok.is(tok::l_brace)) {
792 if (Style.BraceWrapping.AfterObjCDeclaration)
794 parseBlock(/*MustBeDeclaration=*/false);
799 // This branch isn't strictly necessary (the kw_try case below would
800 // do this too after the tok::at is parsed above). But be explicit.
809 if (FormatTok->is(tok::l_brace)) {
810 FormatTok->Type = TT_InlineASMBrace;
812 while (FormatTok && FormatTok->isNot(tok::eof)) {
813 if (FormatTok->is(tok::r_brace)) {
814 FormatTok->Type = TT_InlineASMBrace;
819 FormatTok->Finalized = true;
824 case tok::kw_namespace:
829 if (FormatTok->Tok.is(tok::kw_namespace)) {
835 case tok::kw_protected:
836 case tok::kw_private:
837 if (Style.Language == FormatStyle::LK_Java ||
838 Style.Language == FormatStyle::LK_JavaScript)
841 parseAccessSpecifier();
848 parseForOrWhileLoop();
856 case tok::kw_default:
869 if (FormatTok->Tok.is(tok::string_literal)) {
871 if (FormatTok->Tok.is(tok::l_brace)) {
872 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false);
879 if (Style.Language == FormatStyle::LK_JavaScript) {
880 parseJavaScriptEs6ImportExport();
884 case tok::identifier:
885 if (FormatTok->is(TT_ForEachMacro)) {
886 parseForOrWhileLoop();
889 if (FormatTok->is(TT_MacroBlockBegin)) {
890 parseBlock(/*MustBeDeclaration=*/false, /*AddLevel=*/true,
891 /*MunchSemi=*/false);
894 if (FormatTok->is(Keywords.kw_import)) {
895 if (Style.Language == FormatStyle::LK_JavaScript) {
896 parseJavaScriptEs6ImportExport();
899 if (Style.Language == FormatStyle::LK_Proto) {
901 if (FormatTok->is(tok::kw_public))
903 if (!FormatTok->is(tok::string_literal))
906 if (FormatTok->is(tok::semi))
912 if (FormatTok->isOneOf(Keywords.kw_signals, Keywords.kw_qsignals,
913 Keywords.kw_slots, Keywords.kw_qslots)) {
915 if (FormatTok->is(tok::colon)) {
921 // In all other cases, parse the declaration.
927 const FormatToken *Previous = getPreviousToken();
928 switch (FormatTok->Tok.getKind()) {
931 if (FormatTok->Tok.is(tok::l_brace))
935 // Ignore if this is part of "template <enum ...".
936 if (Previous && Previous->is(tok::less)) {
941 // parseEnum falls through and does not yet add an unwrapped line as an
942 // enum definition can start a structural element.
945 // This only applies for C++.
946 if (Style.Language != FormatStyle::LK_Cpp) {
951 case tok::kw_typedef:
953 if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS,
954 Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS))
960 // parseRecord falls through and does not yet add an unwrapped line as a
961 // record declaration or definition can start a structural element.
963 // This does not apply for Java and JavaScript.
964 if (Style.Language == FormatStyle::LK_Java ||
965 Style.Language == FormatStyle::LK_JavaScript) {
966 if (FormatTok->is(tok::semi))
974 // In Java, classes have an implicit static member "class".
975 if (Style.Language == FormatStyle::LK_Java && FormatTok &&
976 FormatTok->is(tok::kw_class))
978 if (Style.Language == FormatStyle::LK_JavaScript && FormatTok &&
979 FormatTok->Tok.getIdentifierInfo())
980 // JavaScript only has pseudo keywords, all keywords are allowed to
981 // appear in "IdentifierName" positions. See http://es5.github.io/#x7.6
994 case tok::kw_operator:
996 if (FormatTok->isBinaryOperator())
1001 if (FormatTok->Tok.isAnyIdentifier() ||
1002 FormatTok->isSimpleTypeSpecifier())
1004 if (FormatTok->is(tok::l_paren))
1006 if (FormatTok->is(tok::l_brace))
1010 if (!tryToParseBracedList()) {
1011 // A block outside of parentheses must be the last part of a
1012 // structural element.
1013 // FIXME: Figure out cases where this is not true, and add projections
1014 // for them (the one we know is missing are lambdas).
1015 if (Style.BraceWrapping.AfterFunction)
1017 FormatTok->Type = TT_FunctionLBrace;
1018 parseBlock(/*MustBeDeclaration=*/false);
1022 // Otherwise this was a braced init list, and the structural
1023 // element continues.
1026 // We arrive here when parsing function-try blocks.
1029 case tok::identifier: {
1030 if (FormatTok->is(TT_MacroBlockEnd)) {
1035 // Parse function literal unless 'function' is the first token in a line
1036 // in which case this should be treated as a free-standing function.
1037 if (Style.Language == FormatStyle::LK_JavaScript &&
1038 (FormatTok->is(Keywords.kw_function) ||
1039 FormatTok->startsSequence(Keywords.kw_async,
1040 Keywords.kw_function)) &&
1041 Line->Tokens.size() > 0) {
1042 tryToParseJSFunction();
1045 if ((Style.Language == FormatStyle::LK_JavaScript ||
1046 Style.Language == FormatStyle::LK_Java) &&
1047 FormatTok->is(Keywords.kw_interface)) {
1048 if (Style.Language == FormatStyle::LK_JavaScript) {
1049 // In JavaScript/TypeScript, "interface" can be used as a standalone
1050 // identifier, e.g. in `var interface = 1;`. If "interface" is
1051 // followed by another identifier, it is very like to be an actual
1052 // interface declaration.
1053 unsigned StoredPosition = Tokens->getPosition();
1054 FormatToken *Next = Tokens->getNextToken();
1055 FormatTok = Tokens->setPosition(StoredPosition);
1056 if (Next && !mustBeJSIdent(Keywords, Next)) {
1066 // See if the following token should start a new unwrapped line.
1067 StringRef Text = FormatTok->TokenText;
1069 if (Line->Tokens.size() == 1 &&
1070 // JS doesn't have macros, and within classes colons indicate fields,
1072 Style.Language != FormatStyle::LK_JavaScript) {
1073 if (FormatTok->Tok.is(tok::colon) && !Line->MustBeDeclaration) {
1074 Line->Tokens.begin()->Tok->MustBreakBefore = true;
1078 // Recognize function-like macro usages without trailing semicolon as
1079 // well as free-standing macros like Q_OBJECT.
1080 bool FunctionLike = FormatTok->is(tok::l_paren);
1084 bool FollowedByNewline =
1085 CommentsBeforeNextToken.empty()
1086 ? FormatTok->NewlinesBefore > 0
1087 : CommentsBeforeNextToken.front()->NewlinesBefore > 0;
1089 if (FollowedByNewline && (Text.size() >= 5 || FunctionLike) &&
1090 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) {
1098 // Fat arrows (=>) have tok::TokenKind tok::equal but TokenType
1099 // TT_JsFatArrow. The always start an expression or a child block if
1100 // followed by a curly.
1101 if (FormatTok->is(TT_JsFatArrow)) {
1103 if (FormatTok->is(tok::l_brace))
1109 if (FormatTok->Tok.is(tok::l_brace)) {
1126 bool UnwrappedLineParser::tryToParseLambda() {
1127 if (Style.Language != FormatStyle::LK_Cpp) {
1131 const FormatToken* Previous = getPreviousToken();
1133 (Previous->isOneOf(tok::identifier, tok::kw_operator, tok::kw_new,
1135 Previous->closesScope() || Previous->isSimpleTypeSpecifier())) {
1139 assert(FormatTok->is(tok::l_square));
1140 FormatToken &LSquare = *FormatTok;
1141 if (!tryToParseLambdaIntroducer())
1144 while (FormatTok->isNot(tok::l_brace)) {
1145 if (FormatTok->isSimpleTypeSpecifier()) {
1149 switch (FormatTok->Tok.getKind()) {
1161 case tok::identifier:
1162 case tok::numeric_constant:
1163 case tok::coloncolon:
1164 case tok::kw_mutable:
1168 FormatTok->Type = TT_LambdaArrow;
1175 LSquare.Type = TT_LambdaLSquare;
1180 bool UnwrappedLineParser::tryToParseLambdaIntroducer() {
1182 if (FormatTok->is(tok::equal)) {
1184 if (FormatTok->is(tok::r_square)) {
1188 if (FormatTok->isNot(tok::comma))
1191 } else if (FormatTok->is(tok::amp)) {
1193 if (FormatTok->is(tok::r_square)) {
1197 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) {
1200 if (FormatTok->is(tok::comma))
1202 } else if (FormatTok->is(tok::r_square)) {
1207 if (FormatTok->is(tok::amp))
1209 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this))
1212 if (FormatTok->is(tok::ellipsis))
1214 if (FormatTok->is(tok::comma)) {
1216 } else if (FormatTok->is(tok::r_square)) {
1226 void UnwrappedLineParser::tryToParseJSFunction() {
1227 assert(FormatTok->is(Keywords.kw_function) ||
1228 FormatTok->startsSequence(Keywords.kw_async, Keywords.kw_function));
1229 if (FormatTok->is(Keywords.kw_async))
1231 // Consume "function".
1234 // Consume * (generator function). Treat it like C++'s overloaded operators.
1235 if (FormatTok->is(tok::star)) {
1236 FormatTok->Type = TT_OverloadedOperator;
1240 // Consume function name.
1241 if (FormatTok->is(tok::identifier))
1244 if (FormatTok->isNot(tok::l_paren))
1247 // Parse formal parameter list.
1250 if (FormatTok->is(tok::colon)) {
1251 // Parse a type definition.
1254 // Eat the type declaration. For braced inline object types, balance braces,
1255 // otherwise just parse until finding an l_brace for the function body.
1256 if (FormatTok->is(tok::l_brace))
1257 tryToParseBracedList();
1259 while (!FormatTok->isOneOf(tok::l_brace, tok::semi) && !eof())
1263 if (FormatTok->is(tok::semi))
1269 bool UnwrappedLineParser::tryToParseBracedList() {
1270 if (FormatTok->BlockKind == BK_Unknown)
1271 calculateBraceTypes();
1272 assert(FormatTok->BlockKind != BK_Unknown);
1273 if (FormatTok->BlockKind == BK_Block)
1279 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) {
1280 bool HasError = false;
1283 // FIXME: Once we have an expression parser in the UnwrappedLineParser,
1284 // replace this by using parseAssigmentExpression() inside.
1286 if (Style.Language == FormatStyle::LK_JavaScript) {
1287 if (FormatTok->is(Keywords.kw_function) ||
1288 FormatTok->startsSequence(Keywords.kw_async, Keywords.kw_function)) {
1289 tryToParseJSFunction();
1292 if (FormatTok->is(TT_JsFatArrow)) {
1294 // Fat arrows can be followed by simple expressions or by child blocks
1296 if (FormatTok->is(tok::l_brace)) {
1302 switch (FormatTok->Tok.getKind()) {
1305 if (FormatTok->is(tok::l_brace)) {
1313 // Assume there are no blocks inside a braced init list apart
1314 // from the ones we explicitly parse out (like lambdas).
1315 FormatTok->BlockKind = BK_BracedInit;
1320 // JavaScript can just have free standing methods and getters/setters in
1321 // object literals. Detect them by a "{" following ")".
1322 if (Style.Language == FormatStyle::LK_JavaScript) {
1323 if (FormatTok->is(tok::l_brace))
1332 // JavaScript (or more precisely TypeScript) can have semicolons in braced
1333 // lists (in so-called TypeMemberLists). Thus, the semicolon cannot be
1334 // used for error recovery if we have otherwise determined that this is
1336 if (Style.Language == FormatStyle::LK_JavaScript) {
1341 if (!ContinueOnSemicolons)
1356 void UnwrappedLineParser::parseParens() {
1357 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected.");
1360 switch (FormatTok->Tok.getKind()) {
1363 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace))
1370 // A "}" inside parenthesis is an error if there wasn't a matching "{".
1376 if (!tryToParseBracedList())
1381 if (FormatTok->Tok.is(tok::l_brace))
1384 case tok::identifier:
1385 if (Style.Language == FormatStyle::LK_JavaScript &&
1386 (FormatTok->is(Keywords.kw_function) ||
1387 FormatTok->startsSequence(Keywords.kw_async, Keywords.kw_function)))
1388 tryToParseJSFunction();
1399 void UnwrappedLineParser::parseSquare() {
1400 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected.");
1401 if (tryToParseLambda())
1404 switch (FormatTok->Tok.getKind()) {
1412 // A "}" inside parenthesis is an error if there wasn't a matching "{".
1417 case tok::l_brace: {
1418 if (!tryToParseBracedList())
1424 if (FormatTok->Tok.is(tok::l_brace))
1434 void UnwrappedLineParser::parseIfThenElse() {
1435 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected");
1437 if (FormatTok->Tok.is(tok::l_paren))
1439 bool NeedsUnwrappedLine = false;
1440 if (FormatTok->Tok.is(tok::l_brace)) {
1441 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1442 parseBlock(/*MustBeDeclaration=*/false);
1443 if (Style.BraceWrapping.BeforeElse)
1446 NeedsUnwrappedLine = true;
1450 parseStructuralElement();
1453 if (FormatTok->Tok.is(tok::kw_else)) {
1455 if (FormatTok->Tok.is(tok::l_brace)) {
1456 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1457 parseBlock(/*MustBeDeclaration=*/false);
1459 } else if (FormatTok->Tok.is(tok::kw_if)) {
1464 parseStructuralElement();
1465 if (FormatTok->is(tok::eof))
1469 } else if (NeedsUnwrappedLine) {
1474 void UnwrappedLineParser::parseTryCatch() {
1475 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected");
1477 bool NeedsUnwrappedLine = false;
1478 if (FormatTok->is(tok::colon)) {
1479 // We are in a function try block, what comes is an initializer list.
1481 while (FormatTok->is(tok::identifier)) {
1483 if (FormatTok->is(tok::l_paren))
1485 if (FormatTok->is(tok::comma))
1489 // Parse try with resource.
1490 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) {
1493 if (FormatTok->is(tok::l_brace)) {
1494 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1495 parseBlock(/*MustBeDeclaration=*/false);
1496 if (Style.BraceWrapping.BeforeCatch) {
1499 NeedsUnwrappedLine = true;
1501 } else if (!FormatTok->is(tok::kw_catch)) {
1502 // The C++ standard requires a compound-statement after a try.
1503 // If there's none, we try to assume there's a structuralElement
1504 // and try to continue.
1507 parseStructuralElement();
1511 if (FormatTok->is(tok::at))
1513 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except,
1514 tok::kw___finally) ||
1515 ((Style.Language == FormatStyle::LK_Java ||
1516 Style.Language == FormatStyle::LK_JavaScript) &&
1517 FormatTok->is(Keywords.kw_finally)) ||
1518 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) ||
1519 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally))))
1522 while (FormatTok->isNot(tok::l_brace)) {
1523 if (FormatTok->is(tok::l_paren)) {
1527 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof))
1531 NeedsUnwrappedLine = false;
1532 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1533 parseBlock(/*MustBeDeclaration=*/false);
1534 if (Style.BraceWrapping.BeforeCatch)
1537 NeedsUnwrappedLine = true;
1539 if (NeedsUnwrappedLine)
1543 void UnwrappedLineParser::parseNamespace() {
1544 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected");
1546 const FormatToken &InitialToken = *FormatTok;
1548 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon))
1550 if (FormatTok->Tok.is(tok::l_brace)) {
1551 if (ShouldBreakBeforeBrace(Style, InitialToken))
1554 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All ||
1555 (Style.NamespaceIndentation == FormatStyle::NI_Inner &&
1556 DeclarationScopeStack.size() > 1);
1557 parseBlock(/*MustBeDeclaration=*/true, AddLevel);
1558 // Munch the semicolon after a namespace. This is more common than one would
1559 // think. Puttin the semicolon into its own line is very ugly.
1560 if (FormatTok->Tok.is(tok::semi))
1564 // FIXME: Add error handling.
1567 void UnwrappedLineParser::parseNew() {
1568 assert(FormatTok->is(tok::kw_new) && "'new' expected");
1570 if (Style.Language != FormatStyle::LK_Java)
1573 // In Java, we can parse everything up to the parens, which aren't optional.
1575 // There should not be a ;, { or } before the new's open paren.
1576 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace))
1579 // Consume the parens.
1580 if (FormatTok->is(tok::l_paren)) {
1583 // If there is a class body of an anonymous class, consume that as child.
1584 if (FormatTok->is(tok::l_brace))
1592 void UnwrappedLineParser::parseForOrWhileLoop() {
1593 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) &&
1594 "'for', 'while' or foreach macro expected");
1596 if (FormatTok->Tok.is(tok::l_paren))
1598 if (FormatTok->Tok.is(tok::l_brace)) {
1599 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1600 parseBlock(/*MustBeDeclaration=*/false);
1605 parseStructuralElement();
1610 void UnwrappedLineParser::parseDoWhile() {
1611 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected");
1613 if (FormatTok->Tok.is(tok::l_brace)) {
1614 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1615 parseBlock(/*MustBeDeclaration=*/false);
1616 if (Style.BraceWrapping.IndentBraces)
1621 parseStructuralElement();
1625 // FIXME: Add error handling.
1626 if (!FormatTok->Tok.is(tok::kw_while)) {
1632 parseStructuralElement();
1635 void UnwrappedLineParser::parseLabel() {
1637 unsigned OldLineLevel = Line->Level;
1638 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0))
1640 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) {
1641 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1642 parseBlock(/*MustBeDeclaration=*/false);
1643 if (FormatTok->Tok.is(tok::kw_break)) {
1644 if (Style.BraceWrapping.AfterControlStatement)
1646 parseStructuralElement();
1650 if (FormatTok->is(tok::semi))
1654 Line->Level = OldLineLevel;
1655 if (FormatTok->isNot(tok::l_brace)) {
1656 parseStructuralElement();
1661 void UnwrappedLineParser::parseCaseLabel() {
1662 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected");
1663 // FIXME: fix handling of complex expressions here.
1666 } while (!eof() && !FormatTok->Tok.is(tok::colon));
1670 void UnwrappedLineParser::parseSwitch() {
1671 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected");
1673 if (FormatTok->Tok.is(tok::l_paren))
1675 if (FormatTok->Tok.is(tok::l_brace)) {
1676 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1677 parseBlock(/*MustBeDeclaration=*/false);
1682 parseStructuralElement();
1687 void UnwrappedLineParser::parseAccessSpecifier() {
1689 // Understand Qt's slots.
1690 if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots))
1692 // Otherwise, we don't know what it is, and we'd better keep the next token.
1693 if (FormatTok->Tok.is(tok::colon))
1698 bool UnwrappedLineParser::parseEnum() {
1699 // Won't be 'enum' for NS_ENUMs.
1700 if (FormatTok->Tok.is(tok::kw_enum))
1703 // In TypeScript, "enum" can also be used as property name, e.g. in interface
1704 // declarations. An "enum" keyword followed by a colon would be a syntax
1705 // error and thus assume it is just an identifier.
1706 if (Style.Language == FormatStyle::LK_JavaScript &&
1707 FormatTok->isOneOf(tok::colon, tok::question))
1710 // Eat up enum class ...
1711 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct))
1714 while (FormatTok->Tok.getIdentifierInfo() ||
1715 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less,
1716 tok::greater, tok::comma, tok::question)) {
1718 // We can have macros or attributes in between 'enum' and the enum name.
1719 if (FormatTok->is(tok::l_paren))
1721 if (FormatTok->is(tok::identifier)) {
1723 // If there are two identifiers in a row, this is likely an elaborate
1724 // return type. In Java, this can be "implements", etc.
1725 if (Style.Language == FormatStyle::LK_Cpp &&
1726 FormatTok->is(tok::identifier))
1731 // Just a declaration or something is wrong.
1732 if (FormatTok->isNot(tok::l_brace))
1734 FormatTok->BlockKind = BK_Block;
1736 if (Style.Language == FormatStyle::LK_Java) {
1737 // Java enums are different.
1738 parseJavaEnumBody();
1741 if (Style.Language == FormatStyle::LK_Proto) {
1742 parseBlock(/*MustBeDeclaration=*/true);
1747 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true);
1749 if (FormatTok->is(tok::semi))
1755 // There is no addUnwrappedLine() here so that we fall through to parsing a
1756 // structural element afterwards. Thus, in "enum A {} n, m;",
1757 // "} n, m;" will end up in one unwrapped line.
1760 void UnwrappedLineParser::parseJavaEnumBody() {
1761 // Determine whether the enum is simple, i.e. does not have a semicolon or
1762 // constants with class bodies. Simple enums can be formatted like braced
1763 // lists, contracted to a single line, etc.
1764 unsigned StoredPosition = Tokens->getPosition();
1765 bool IsSimple = true;
1766 FormatToken *Tok = Tokens->getNextToken();
1768 if (Tok->is(tok::r_brace))
1770 if (Tok->isOneOf(tok::l_brace, tok::semi)) {
1774 // FIXME: This will also mark enums with braces in the arguments to enum
1775 // constants as "not simple". This is probably fine in practice, though.
1776 Tok = Tokens->getNextToken();
1778 FormatTok = Tokens->setPosition(StoredPosition);
1786 // Parse the body of a more complex enum.
1787 // First add a line for everything up to the "{".
1792 // Parse the enum constants.
1794 if (FormatTok->is(tok::l_brace)) {
1795 // Parse the constant's class body.
1796 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true,
1797 /*MunchSemi=*/false);
1798 } else if (FormatTok->is(tok::l_paren)) {
1800 } else if (FormatTok->is(tok::comma)) {
1803 } else if (FormatTok->is(tok::semi)) {
1807 } else if (FormatTok->is(tok::r_brace)) {
1815 // Parse the class body after the enum's ";" if any.
1816 parseLevel(/*HasOpeningBrace=*/true);
1822 void UnwrappedLineParser::parseRecord() {
1823 const FormatToken &InitialToken = *FormatTok;
1826 // The actual identifier can be a nested name specifier, and in macros
1827 // it is often token-pasted.
1828 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash,
1829 tok::kw___attribute, tok::kw___declspec,
1831 ((Style.Language == FormatStyle::LK_Java ||
1832 Style.Language == FormatStyle::LK_JavaScript) &&
1833 FormatTok->isOneOf(tok::period, tok::comma))) {
1834 bool IsNonMacroIdentifier =
1835 FormatTok->is(tok::identifier) &&
1836 FormatTok->TokenText != FormatTok->TokenText.upper();
1838 // We can have macros or attributes in between 'class' and the class name.
1839 if (!IsNonMacroIdentifier && FormatTok->Tok.is(tok::l_paren))
1843 // Note that parsing away template declarations here leads to incorrectly
1844 // accepting function declarations as record declarations.
1845 // In general, we cannot solve this problem. Consider:
1846 // class A<int> B() {}
1847 // which can be a function definition or a class definition when B() is a
1848 // macro. If we find enough real-world cases where this is a problem, we
1849 // can parse for the 'template' keyword in the beginning of the statement,
1850 // and thus rule out the record production in case there is no template
1851 // (this would still leave us with an ambiguity between template function
1852 // and class declarations).
1853 if (FormatTok->isOneOf(tok::colon, tok::less)) {
1855 if (FormatTok->is(tok::l_brace)) {
1856 calculateBraceTypes(/*ExpectClassBody=*/true);
1857 if (!tryToParseBracedList())
1860 if (FormatTok->Tok.is(tok::semi))
1865 if (FormatTok->Tok.is(tok::l_brace)) {
1866 if (ShouldBreakBeforeBrace(Style, InitialToken))
1869 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true,
1870 /*MunchSemi=*/false);
1872 // There is no addUnwrappedLine() here so that we fall through to parsing a
1873 // structural element afterwards. Thus, in "class A {} n, m;",
1874 // "} n, m;" will end up in one unwrapped line.
1877 void UnwrappedLineParser::parseObjCProtocolList() {
1878 assert(FormatTok->Tok.is(tok::less) && "'<' expected.");
1881 while (!eof() && FormatTok->Tok.isNot(tok::greater));
1882 nextToken(); // Skip '>'.
1885 void UnwrappedLineParser::parseObjCUntilAtEnd() {
1887 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) {
1892 if (FormatTok->is(tok::l_brace)) {
1893 parseBlock(/*MustBeDeclaration=*/false);
1894 // In ObjC interfaces, nothing should be following the "}".
1896 } else if (FormatTok->is(tok::r_brace)) {
1897 // Ignore stray "}". parseStructuralElement doesn't consume them.
1901 parseStructuralElement();
1906 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() {
1908 nextToken(); // interface name
1910 // @interface can be followed by either a base class, or a category.
1911 if (FormatTok->Tok.is(tok::colon)) {
1913 nextToken(); // base class name
1914 } else if (FormatTok->Tok.is(tok::l_paren))
1915 // Skip category, if present.
1918 if (FormatTok->Tok.is(tok::less))
1919 parseObjCProtocolList();
1921 if (FormatTok->Tok.is(tok::l_brace)) {
1922 if (Style.BraceWrapping.AfterObjCDeclaration)
1924 parseBlock(/*MustBeDeclaration=*/true);
1927 // With instance variables, this puts '}' on its own line. Without instance
1928 // variables, this ends the @interface line.
1931 parseObjCUntilAtEnd();
1934 void UnwrappedLineParser::parseObjCProtocol() {
1936 nextToken(); // protocol name
1938 if (FormatTok->Tok.is(tok::less))
1939 parseObjCProtocolList();
1941 // Check for protocol declaration.
1942 if (FormatTok->Tok.is(tok::semi)) {
1944 return addUnwrappedLine();
1948 parseObjCUntilAtEnd();
1951 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() {
1952 bool IsImport = FormatTok->is(Keywords.kw_import);
1953 assert(IsImport || FormatTok->is(tok::kw_export));
1956 // Consume the "default" in "export default class/function".
1957 if (FormatTok->is(tok::kw_default))
1960 // Consume "async function", "function" and "default function", so that these
1961 // get parsed as free-standing JS functions, i.e. do not require a trailing
1963 if (FormatTok->is(Keywords.kw_async))
1965 if (FormatTok->is(Keywords.kw_function)) {
1970 // For imports, `export *`, `export {...}`, consume the rest of the line up
1971 // to the terminating `;`. For everything else, just return and continue
1972 // parsing the structural element, i.e. the declaration or expression for
1973 // `export default`.
1974 if (!IsImport && !FormatTok->isOneOf(tok::l_brace, tok::star) &&
1975 !FormatTok->isStringLiteral())
1979 if (FormatTok->is(tok::semi))
1981 if (Line->Tokens.size() == 0) {
1982 // Common issue: Automatic Semicolon Insertion wrapped the line, so the
1983 // import statement should terminate.
1986 if (FormatTok->is(tok::l_brace)) {
1987 FormatTok->BlockKind = BK_Block;
1995 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line,
1996 StringRef Prefix = "") {
1997 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")"
1998 << (Line.InPPDirective ? " MACRO" : "") << ": ";
1999 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(),
2000 E = Line.Tokens.end();
2002 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] ";
2004 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(),
2005 E = Line.Tokens.end();
2007 const UnwrappedLineNode &Node = *I;
2008 for (SmallVectorImpl<UnwrappedLine>::const_iterator
2009 I = Node.Children.begin(),
2010 E = Node.Children.end();
2012 printDebugInfo(*I, "\nChild: ");
2015 llvm::dbgs() << "\n";
2018 void UnwrappedLineParser::addUnwrappedLine() {
2019 if (Line->Tokens.empty())
2022 if (CurrentLines == &Lines)
2023 printDebugInfo(*Line);
2025 CurrentLines->push_back(std::move(*Line));
2026 Line->Tokens.clear();
2027 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) {
2028 CurrentLines->append(
2029 std::make_move_iterator(PreprocessorDirectives.begin()),
2030 std::make_move_iterator(PreprocessorDirectives.end()));
2031 PreprocessorDirectives.clear();
2035 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); }
2037 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) {
2038 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) &&
2039 FormatTok.NewlinesBefore > 0;
2042 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) {
2043 bool JustComments = Line->Tokens.empty();
2044 for (SmallVectorImpl<FormatToken *>::const_iterator
2045 I = CommentsBeforeNextToken.begin(),
2046 E = CommentsBeforeNextToken.end();
2048 if (isOnNewLine(**I) && JustComments)
2052 if (NewlineBeforeNext && JustComments)
2054 CommentsBeforeNextToken.clear();
2057 void UnwrappedLineParser::nextToken() {
2060 flushComments(isOnNewLine(*FormatTok));
2061 pushToken(FormatTok);
2062 if (Style.Language != FormatStyle::LK_JavaScript)
2065 readTokenWithJavaScriptASI();
2068 const FormatToken *UnwrappedLineParser::getPreviousToken() {
2069 // FIXME: This is a dirty way to access the previous token. Find a better
2071 if (!Line || Line->Tokens.empty())
2073 return Line->Tokens.back().Tok;
2076 void UnwrappedLineParser::readToken() {
2077 bool CommentsInCurrentLine = true;
2079 FormatTok = Tokens->getNextToken();
2081 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) &&
2082 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) {
2083 // If there is an unfinished unwrapped line, we flush the preprocessor
2084 // directives only after that unwrapped line was finished later.
2085 bool SwitchToPreprocessorLines = !Line->Tokens.empty();
2086 ScopedLineState BlockState(*this, SwitchToPreprocessorLines);
2087 // Comments stored before the preprocessor directive need to be output
2088 // before the preprocessor directive, at the same level as the
2089 // preprocessor directive, as we consider them to apply to the directive.
2090 flushComments(isOnNewLine(*FormatTok));
2093 while (FormatTok->Type == TT_ConflictStart ||
2094 FormatTok->Type == TT_ConflictEnd ||
2095 FormatTok->Type == TT_ConflictAlternative) {
2096 if (FormatTok->Type == TT_ConflictStart) {
2097 conditionalCompilationStart(/*Unreachable=*/false);
2098 } else if (FormatTok->Type == TT_ConflictAlternative) {
2099 conditionalCompilationAlternative();
2100 } else if (FormatTok->Type == TT_ConflictEnd) {
2101 conditionalCompilationEnd();
2103 FormatTok = Tokens->getNextToken();
2104 FormatTok->MustBreakBefore = true;
2107 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) &&
2108 !Line->InPPDirective) {
2112 if (!FormatTok->Tok.is(tok::comment))
2114 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) {
2115 CommentsInCurrentLine = false;
2117 if (CommentsInCurrentLine) {
2118 pushToken(FormatTok);
2120 CommentsBeforeNextToken.push_back(FormatTok);
2125 void UnwrappedLineParser::pushToken(FormatToken *Tok) {
2126 Line->Tokens.push_back(UnwrappedLineNode(Tok));
2127 if (MustBreakBeforeNextToken) {
2128 Line->Tokens.back().Tok->MustBreakBefore = true;
2129 MustBreakBeforeNextToken = false;
2133 } // end namespace format
2134 } // end namespace clang