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 // 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));
324 // Get next non-comment token.
325 FormatToken *NextTok;
326 unsigned ReadTokens = 0;
328 NextTok = Tokens->getNextToken();
330 } while (NextTok->is(tok::comment));
332 switch (Tok->Tok.getKind()) {
334 Tok->BlockKind = BK_Unknown;
335 LBraceStack.push_back(Tok);
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);
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;
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).
356 // We exclude + and - as they can be ObjC visibility modifiers.
358 NextTok->isOneOf(tok::comma, tok::period, tok::colon,
359 tok::r_paren, tok::r_square, tok::l_brace,
360 tok::l_square, tok::l_paren, tok::ellipsis) ||
361 (NextTok->is(tok::semi) &&
362 (!ExpectClassBody || LBraceStack.size() != 1)) ||
363 (NextTok->isBinaryOperator() && !NextIsObjCMethod);
365 if (ProbablyBracedList) {
366 Tok->BlockKind = BK_BracedInit;
367 LBraceStack.back()->BlockKind = BK_BracedInit;
369 Tok->BlockKind = BK_Block;
370 LBraceStack.back()->BlockKind = BK_Block;
373 LBraceStack.pop_back();
384 if (!LBraceStack.empty())
385 LBraceStack.back()->BlockKind = BK_Block;
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;
398 FormatTok = Tokens->setPosition(StoredPosition);
401 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel,
403 assert(FormatTok->isOneOf(tok::l_brace, TT_MacroBlockBegin) &&
404 "'{' or macro block token expected");
405 const bool MacroBlock = FormatTok->is(TT_MacroBlockBegin);
406 FormatTok->BlockKind = BK_Block;
408 unsigned InitialLevel = Line->Level;
411 if (MacroBlock && FormatTok->is(tok::l_paren))
416 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack,
420 parseLevel(/*HasOpeningBrace=*/true);
422 if (MacroBlock ? !FormatTok->is(TT_MacroBlockEnd)
423 : !FormatTok->is(tok::r_brace)) {
424 Line->Level = InitialLevel;
425 FormatTok->BlockKind = BK_Block;
429 nextToken(); // Munch the closing brace.
431 if (MacroBlock && FormatTok->is(tok::l_paren))
434 if (MunchSemi && FormatTok->Tok.is(tok::semi))
436 Line->Level = InitialLevel;
439 static bool isGoogScope(const UnwrappedLine &Line) {
440 // FIXME: Closure-library specific stuff should not be hard-coded but be
442 if (Line.Tokens.size() < 4)
444 auto I = Line.Tokens.begin();
445 if (I->Tok->TokenText != "goog")
448 if (I->Tok->isNot(tok::period))
451 if (I->Tok->TokenText != "scope")
454 return I->Tok->is(tok::l_paren);
457 static bool ShouldBreakBeforeBrace(const FormatStyle &Style,
458 const FormatToken &InitialToken) {
459 if (InitialToken.is(tok::kw_namespace))
460 return Style.BraceWrapping.AfterNamespace;
461 if (InitialToken.is(tok::kw_class))
462 return Style.BraceWrapping.AfterClass;
463 if (InitialToken.is(tok::kw_union))
464 return Style.BraceWrapping.AfterUnion;
465 if (InitialToken.is(tok::kw_struct))
466 return Style.BraceWrapping.AfterStruct;
470 void UnwrappedLineParser::parseChildBlock() {
471 FormatTok->BlockKind = BK_Block;
475 Style.Language == FormatStyle::LK_JavaScript && isGoogScope(*Line);
476 ScopedLineState LineState(*this);
477 ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack,
478 /*MustBeDeclaration=*/false);
479 Line->Level += GoogScope ? 0 : 1;
480 parseLevel(/*HasOpeningBrace=*/true);
481 flushComments(isOnNewLine(*FormatTok));
482 Line->Level -= GoogScope ? 0 : 1;
487 void UnwrappedLineParser::parsePPDirective() {
488 assert(FormatTok->Tok.is(tok::hash) && "'#' expected");
489 ScopedMacroState MacroState(*Line, Tokens, FormatTok);
492 if (!FormatTok->Tok.getIdentifierInfo()) {
497 switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) {
502 parsePPIf(/*IfDef=*/false);
506 parsePPIf(/*IfDef=*/true);
523 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) {
524 if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable))
525 PPStack.push_back(PP_Unreachable);
527 PPStack.push_back(PP_Conditional);
530 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) {
532 assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size());
533 if (PPBranchLevel == (int)PPLevelBranchIndex.size()) {
534 PPLevelBranchIndex.push_back(0);
535 PPLevelBranchCount.push_back(0);
537 PPChainBranchIndex.push(0);
538 bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0;
539 conditionalCompilationCondition(Unreachable || Skip);
542 void UnwrappedLineParser::conditionalCompilationAlternative() {
543 if (!PPStack.empty())
545 assert(PPBranchLevel < (int)PPLevelBranchIndex.size());
546 if (!PPChainBranchIndex.empty())
547 ++PPChainBranchIndex.top();
548 conditionalCompilationCondition(
549 PPBranchLevel >= 0 && !PPChainBranchIndex.empty() &&
550 PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top());
553 void UnwrappedLineParser::conditionalCompilationEnd() {
554 assert(PPBranchLevel < (int)PPLevelBranchIndex.size());
555 if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) {
556 if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) {
557 PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1;
560 // Guard against #endif's without #if.
561 if (PPBranchLevel > 0)
563 if (!PPChainBranchIndex.empty())
564 PPChainBranchIndex.pop();
565 if (!PPStack.empty())
569 void UnwrappedLineParser::parsePPIf(bool IfDef) {
571 bool IsLiteralFalse = (FormatTok->Tok.isLiteral() &&
572 FormatTok->Tok.getLiteralData() != nullptr &&
573 StringRef(FormatTok->Tok.getLiteralData(),
574 FormatTok->Tok.getLength()) == "0") ||
575 FormatTok->Tok.is(tok::kw_false);
576 conditionalCompilationStart(!IfDef && IsLiteralFalse);
580 void UnwrappedLineParser::parsePPElse() {
581 conditionalCompilationAlternative();
585 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); }
587 void UnwrappedLineParser::parsePPEndIf() {
588 conditionalCompilationEnd();
592 void UnwrappedLineParser::parsePPDefine() {
595 if (FormatTok->Tok.getKind() != tok::identifier) {
600 if (FormatTok->Tok.getKind() == tok::l_paren &&
601 FormatTok->WhitespaceRange.getBegin() ==
602 FormatTok->WhitespaceRange.getEnd()) {
608 // Errors during a preprocessor directive can only affect the layout of the
609 // preprocessor directive, and thus we ignore them. An alternative approach
610 // would be to use the same approach we use on the file level (no
611 // re-indentation if there was a structural error) within the macro
616 void UnwrappedLineParser::parsePPUnknown() {
623 // Here we blacklist certain tokens that are not usually the first token in an
624 // unwrapped line. This is used in attempt to distinguish macro calls without
625 // trailing semicolons from other constructs split to several lines.
626 static bool tokenCanStartNewLine(const clang::Token &Tok) {
627 // Semicolon can be a null-statement, l_square can be a start of a macro or
628 // a C++11 attribute, but this doesn't seem to be common.
629 return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) &&
630 Tok.isNot(tok::l_square) &&
631 // Tokens that can only be used as binary operators and a part of
632 // overloaded operator names.
633 Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) &&
634 Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) &&
635 Tok.isNot(tok::less) && Tok.isNot(tok::greater) &&
636 Tok.isNot(tok::slash) && Tok.isNot(tok::percent) &&
637 Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) &&
638 Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) &&
639 Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) &&
640 Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) &&
641 Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) &&
642 Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) &&
643 Tok.isNot(tok::lesslessequal) &&
644 // Colon is used in labels, base class lists, initializer lists,
645 // range-based for loops, ternary operator, but should never be the
646 // first token in an unwrapped line.
647 Tok.isNot(tok::colon) &&
648 // 'noexcept' is a trailing annotation.
649 Tok.isNot(tok::kw_noexcept);
652 void UnwrappedLineParser::parseStructuralElement() {
653 assert(!FormatTok->is(tok::l_brace));
654 if (Style.Language == FormatStyle::LK_TableGen &&
655 FormatTok->is(tok::pp_include)) {
657 if (FormatTok->is(tok::string_literal))
662 switch (FormatTok->Tok.getKind()) {
665 if (FormatTok->Tok.is(tok::l_brace)) {
669 switch (FormatTok->Tok.getObjCKeywordID()) {
670 case tok::objc_public:
671 case tok::objc_protected:
672 case tok::objc_package:
673 case tok::objc_private:
674 return parseAccessSpecifier();
675 case tok::objc_interface:
676 case tok::objc_implementation:
677 return parseObjCInterfaceOrImplementation();
678 case tok::objc_protocol:
679 return parseObjCProtocol();
681 return; // Handled by the caller.
682 case tok::objc_optional:
683 case tok::objc_required:
687 case tok::objc_autoreleasepool:
689 if (FormatTok->Tok.is(tok::l_brace)) {
690 if (Style.BraceWrapping.AfterObjCDeclaration)
692 parseBlock(/*MustBeDeclaration=*/false);
697 // This branch isn't strictly necessary (the kw_try case below would
698 // do this too after the tok::at is parsed above). But be explicit.
707 if (FormatTok->is(tok::l_brace)) {
708 FormatTok->Type = TT_InlineASMBrace;
710 while (FormatTok && FormatTok->isNot(tok::eof)) {
711 if (FormatTok->is(tok::r_brace)) {
712 FormatTok->Type = TT_InlineASMBrace;
717 FormatTok->Finalized = true;
722 case tok::kw_namespace:
727 if (FormatTok->Tok.is(tok::kw_namespace)) {
733 case tok::kw_protected:
734 case tok::kw_private:
735 if (Style.Language == FormatStyle::LK_Java ||
736 Style.Language == FormatStyle::LK_JavaScript)
739 parseAccessSpecifier();
746 parseForOrWhileLoop();
754 case tok::kw_default:
767 if (FormatTok->Tok.is(tok::string_literal)) {
769 if (FormatTok->Tok.is(tok::l_brace)) {
770 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false);
777 if (Style.Language == FormatStyle::LK_JavaScript) {
778 parseJavaScriptEs6ImportExport();
782 case tok::identifier:
783 if (FormatTok->is(TT_ForEachMacro)) {
784 parseForOrWhileLoop();
787 if (FormatTok->is(TT_MacroBlockBegin)) {
788 parseBlock(/*MustBeDeclaration=*/false, /*AddLevel=*/true,
789 /*MunchSemi=*/false);
792 if (Style.Language == FormatStyle::LK_JavaScript &&
793 FormatTok->is(Keywords.kw_import)) {
794 parseJavaScriptEs6ImportExport();
797 if (FormatTok->isOneOf(Keywords.kw_signals, Keywords.kw_qsignals,
798 Keywords.kw_slots, Keywords.kw_qslots)) {
800 if (FormatTok->is(tok::colon)) {
806 // In all other cases, parse the declaration.
812 switch (FormatTok->Tok.getKind()) {
815 if (FormatTok->Tok.is(tok::l_brace))
819 // parseEnum falls through and does not yet add an unwrapped line as an
820 // enum definition can start a structural element.
823 // This only applies for C++.
824 if (Style.Language != FormatStyle::LK_Cpp) {
829 case tok::kw_typedef:
831 if (FormatTok->isOneOf(Keywords.kw_NS_ENUM, Keywords.kw_NS_OPTIONS,
832 Keywords.kw_CF_ENUM, Keywords.kw_CF_OPTIONS))
838 // parseRecord falls through and does not yet add an unwrapped line as a
839 // record declaration or definition can start a structural element.
841 // This does not apply for Java and JavaScript.
842 if (Style.Language == FormatStyle::LK_Java ||
843 Style.Language == FormatStyle::LK_JavaScript) {
850 // In Java, classes have an implicit static member "class".
851 if (Style.Language == FormatStyle::LK_Java && FormatTok &&
852 FormatTok->is(tok::kw_class))
854 if (Style.Language == FormatStyle::LK_JavaScript && FormatTok &&
855 FormatTok->Tok.getIdentifierInfo())
856 // JavaScript only has pseudo keywords, all keywords are allowed to
857 // appear in "IdentifierName" positions. See http://es5.github.io/#x7.6
870 case tok::kw_operator:
872 if (FormatTok->isBinaryOperator())
877 if (FormatTok->Tok.isAnyIdentifier() ||
878 FormatTok->isSimpleTypeSpecifier())
880 if (FormatTok->is(tok::l_paren))
882 if (FormatTok->is(tok::l_brace))
886 if (!tryToParseBracedList()) {
887 // A block outside of parentheses must be the last part of a
888 // structural element.
889 // FIXME: Figure out cases where this is not true, and add projections
890 // for them (the one we know is missing are lambdas).
891 if (Style.BraceWrapping.AfterFunction)
893 FormatTok->Type = TT_FunctionLBrace;
894 parseBlock(/*MustBeDeclaration=*/false);
898 // Otherwise this was a braced init list, and the structural
899 // element continues.
902 // We arrive here when parsing function-try blocks.
905 case tok::identifier: {
906 if (FormatTok->is(TT_MacroBlockEnd)) {
911 // Parse function literal unless 'function' is the first token in a line
912 // in which case this should be treated as a free-standing function.
913 if (Style.Language == FormatStyle::LK_JavaScript &&
914 FormatTok->is(Keywords.kw_function) && Line->Tokens.size() > 0) {
915 tryToParseJSFunction();
918 if ((Style.Language == FormatStyle::LK_JavaScript ||
919 Style.Language == FormatStyle::LK_Java) &&
920 FormatTok->is(Keywords.kw_interface)) {
926 StringRef Text = FormatTok->TokenText;
928 if (Line->Tokens.size() == 1 &&
929 // JS doesn't have macros, and within classes colons indicate fields,
931 Style.Language != FormatStyle::LK_JavaScript) {
932 if (FormatTok->Tok.is(tok::colon) && !Line->MustBeDeclaration) {
936 // Recognize function-like macro usages without trailing semicolon as
937 // well as free-standing macros like Q_OBJECT.
938 bool FunctionLike = FormatTok->is(tok::l_paren);
942 bool FollowedByNewline =
943 CommentsBeforeNextToken.empty()
944 ? FormatTok->NewlinesBefore > 0
945 : CommentsBeforeNextToken.front()->NewlinesBefore > 0;
947 if (FollowedByNewline && (Text.size() >= 5 || FunctionLike) &&
948 tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) {
956 // Fat arrows (=>) have tok::TokenKind tok::equal but TokenType
957 // TT_JsFatArrow. The always start an expression or a child block if
958 // followed by a curly.
959 if (FormatTok->is(TT_JsFatArrow)) {
961 if (FormatTok->is(tok::l_brace))
967 if (FormatTok->Tok.is(tok::l_brace)) {
984 bool UnwrappedLineParser::tryToParseLambda() {
985 if (Style.Language != FormatStyle::LK_Cpp) {
989 // FIXME: This is a dirty way to access the previous token. Find a better
991 if (!Line->Tokens.empty() &&
992 (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator,
993 tok::kw_new, tok::kw_delete) ||
994 Line->Tokens.back().Tok->closesScope() ||
995 Line->Tokens.back().Tok->isSimpleTypeSpecifier())) {
999 assert(FormatTok->is(tok::l_square));
1000 FormatToken &LSquare = *FormatTok;
1001 if (!tryToParseLambdaIntroducer())
1004 while (FormatTok->isNot(tok::l_brace)) {
1005 if (FormatTok->isSimpleTypeSpecifier()) {
1009 switch (FormatTok->Tok.getKind()) {
1021 case tok::identifier:
1022 case tok::numeric_constant:
1023 case tok::coloncolon:
1024 case tok::kw_mutable:
1028 FormatTok->Type = TT_LambdaArrow;
1035 LSquare.Type = TT_LambdaLSquare;
1040 bool UnwrappedLineParser::tryToParseLambdaIntroducer() {
1042 if (FormatTok->is(tok::equal)) {
1044 if (FormatTok->is(tok::r_square)) {
1048 if (FormatTok->isNot(tok::comma))
1051 } else if (FormatTok->is(tok::amp)) {
1053 if (FormatTok->is(tok::r_square)) {
1057 if (!FormatTok->isOneOf(tok::comma, tok::identifier)) {
1060 if (FormatTok->is(tok::comma))
1062 } else if (FormatTok->is(tok::r_square)) {
1067 if (FormatTok->is(tok::amp))
1069 if (!FormatTok->isOneOf(tok::identifier, tok::kw_this))
1072 if (FormatTok->is(tok::ellipsis))
1074 if (FormatTok->is(tok::comma)) {
1076 } else if (FormatTok->is(tok::r_square)) {
1086 void UnwrappedLineParser::tryToParseJSFunction() {
1089 // Consume function name.
1090 if (FormatTok->is(tok::identifier))
1093 if (FormatTok->isNot(tok::l_paren))
1096 // Parse formal parameter list.
1099 if (FormatTok->is(tok::colon)) {
1100 // Parse a type definition.
1103 // Eat the type declaration. For braced inline object types, balance braces,
1104 // otherwise just parse until finding an l_brace for the function body.
1105 if (FormatTok->is(tok::l_brace))
1106 tryToParseBracedList();
1108 while (FormatTok->isNot(tok::l_brace) && !eof())
1115 bool UnwrappedLineParser::tryToParseBracedList() {
1116 if (FormatTok->BlockKind == BK_Unknown)
1117 calculateBraceTypes();
1118 assert(FormatTok->BlockKind != BK_Unknown);
1119 if (FormatTok->BlockKind == BK_Block)
1125 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) {
1126 bool HasError = false;
1129 // FIXME: Once we have an expression parser in the UnwrappedLineParser,
1130 // replace this by using parseAssigmentExpression() inside.
1132 if (Style.Language == FormatStyle::LK_JavaScript) {
1133 if (FormatTok->is(Keywords.kw_function)) {
1134 tryToParseJSFunction();
1137 if (FormatTok->is(TT_JsFatArrow)) {
1139 // Fat arrows can be followed by simple expressions or by child blocks
1141 if (FormatTok->is(tok::l_brace)) {
1147 switch (FormatTok->Tok.getKind()) {
1150 if (FormatTok->is(tok::l_brace)) {
1158 // Assume there are no blocks inside a braced init list apart
1159 // from the ones we explicitly parse out (like lambdas).
1160 FormatTok->BlockKind = BK_BracedInit;
1165 // JavaScript can just have free standing methods and getters/setters in
1166 // object literals. Detect them by a "{" following ")".
1167 if (Style.Language == FormatStyle::LK_JavaScript) {
1168 if (FormatTok->is(tok::l_brace))
1178 if (!ContinueOnSemicolons)
1193 void UnwrappedLineParser::parseParens() {
1194 assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected.");
1197 switch (FormatTok->Tok.getKind()) {
1200 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_brace))
1207 // A "}" inside parenthesis is an error if there wasn't a matching "{".
1213 if (!tryToParseBracedList())
1218 if (FormatTok->Tok.is(tok::l_brace))
1221 case tok::identifier:
1222 if (Style.Language == FormatStyle::LK_JavaScript &&
1223 FormatTok->is(Keywords.kw_function))
1224 tryToParseJSFunction();
1235 void UnwrappedLineParser::parseSquare() {
1236 assert(FormatTok->Tok.is(tok::l_square) && "'[' expected.");
1237 if (tryToParseLambda())
1240 switch (FormatTok->Tok.getKind()) {
1248 // A "}" inside parenthesis is an error if there wasn't a matching "{".
1253 case tok::l_brace: {
1254 if (!tryToParseBracedList())
1260 if (FormatTok->Tok.is(tok::l_brace))
1270 void UnwrappedLineParser::parseIfThenElse() {
1271 assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected");
1273 if (FormatTok->Tok.is(tok::l_paren))
1275 bool NeedsUnwrappedLine = false;
1276 if (FormatTok->Tok.is(tok::l_brace)) {
1277 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1278 parseBlock(/*MustBeDeclaration=*/false);
1279 if (Style.BraceWrapping.BeforeElse)
1282 NeedsUnwrappedLine = true;
1286 parseStructuralElement();
1289 if (FormatTok->Tok.is(tok::kw_else)) {
1291 if (FormatTok->Tok.is(tok::l_brace)) {
1292 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1293 parseBlock(/*MustBeDeclaration=*/false);
1295 } else if (FormatTok->Tok.is(tok::kw_if)) {
1300 parseStructuralElement();
1303 } else if (NeedsUnwrappedLine) {
1308 void UnwrappedLineParser::parseTryCatch() {
1309 assert(FormatTok->isOneOf(tok::kw_try, tok::kw___try) && "'try' expected");
1311 bool NeedsUnwrappedLine = false;
1312 if (FormatTok->is(tok::colon)) {
1313 // We are in a function try block, what comes is an initializer list.
1315 while (FormatTok->is(tok::identifier)) {
1317 if (FormatTok->is(tok::l_paren))
1319 if (FormatTok->is(tok::comma))
1323 // Parse try with resource.
1324 if (Style.Language == FormatStyle::LK_Java && FormatTok->is(tok::l_paren)) {
1327 if (FormatTok->is(tok::l_brace)) {
1328 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1329 parseBlock(/*MustBeDeclaration=*/false);
1330 if (Style.BraceWrapping.BeforeCatch) {
1333 NeedsUnwrappedLine = true;
1335 } else if (!FormatTok->is(tok::kw_catch)) {
1336 // The C++ standard requires a compound-statement after a try.
1337 // If there's none, we try to assume there's a structuralElement
1338 // and try to continue.
1341 parseStructuralElement();
1345 if (FormatTok->is(tok::at))
1347 if (!(FormatTok->isOneOf(tok::kw_catch, Keywords.kw___except,
1348 tok::kw___finally) ||
1349 ((Style.Language == FormatStyle::LK_Java ||
1350 Style.Language == FormatStyle::LK_JavaScript) &&
1351 FormatTok->is(Keywords.kw_finally)) ||
1352 (FormatTok->Tok.isObjCAtKeyword(tok::objc_catch) ||
1353 FormatTok->Tok.isObjCAtKeyword(tok::objc_finally))))
1356 while (FormatTok->isNot(tok::l_brace)) {
1357 if (FormatTok->is(tok::l_paren)) {
1361 if (FormatTok->isOneOf(tok::semi, tok::r_brace, tok::eof))
1365 NeedsUnwrappedLine = false;
1366 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1367 parseBlock(/*MustBeDeclaration=*/false);
1368 if (Style.BraceWrapping.BeforeCatch)
1371 NeedsUnwrappedLine = true;
1373 if (NeedsUnwrappedLine)
1377 void UnwrappedLineParser::parseNamespace() {
1378 assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected");
1380 const FormatToken &InitialToken = *FormatTok;
1382 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon))
1384 if (FormatTok->Tok.is(tok::l_brace)) {
1385 if (ShouldBreakBeforeBrace(Style, InitialToken))
1388 bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All ||
1389 (Style.NamespaceIndentation == FormatStyle::NI_Inner &&
1390 DeclarationScopeStack.size() > 1);
1391 parseBlock(/*MustBeDeclaration=*/true, AddLevel);
1392 // Munch the semicolon after a namespace. This is more common than one would
1393 // think. Puttin the semicolon into its own line is very ugly.
1394 if (FormatTok->Tok.is(tok::semi))
1398 // FIXME: Add error handling.
1401 void UnwrappedLineParser::parseNew() {
1402 assert(FormatTok->is(tok::kw_new) && "'new' expected");
1404 if (Style.Language != FormatStyle::LK_Java)
1407 // In Java, we can parse everything up to the parens, which aren't optional.
1409 // There should not be a ;, { or } before the new's open paren.
1410 if (FormatTok->isOneOf(tok::semi, tok::l_brace, tok::r_brace))
1413 // Consume the parens.
1414 if (FormatTok->is(tok::l_paren)) {
1417 // If there is a class body of an anonymous class, consume that as child.
1418 if (FormatTok->is(tok::l_brace))
1426 void UnwrappedLineParser::parseForOrWhileLoop() {
1427 assert(FormatTok->isOneOf(tok::kw_for, tok::kw_while, TT_ForEachMacro) &&
1428 "'for', 'while' or foreach macro expected");
1430 if (FormatTok->Tok.is(tok::l_paren))
1432 if (FormatTok->Tok.is(tok::l_brace)) {
1433 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1434 parseBlock(/*MustBeDeclaration=*/false);
1439 parseStructuralElement();
1444 void UnwrappedLineParser::parseDoWhile() {
1445 assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected");
1447 if (FormatTok->Tok.is(tok::l_brace)) {
1448 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1449 parseBlock(/*MustBeDeclaration=*/false);
1450 if (Style.BraceWrapping.IndentBraces)
1455 parseStructuralElement();
1459 // FIXME: Add error handling.
1460 if (!FormatTok->Tok.is(tok::kw_while)) {
1466 parseStructuralElement();
1469 void UnwrappedLineParser::parseLabel() {
1471 unsigned OldLineLevel = Line->Level;
1472 if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0))
1474 if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) {
1475 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1476 parseBlock(/*MustBeDeclaration=*/false);
1477 if (FormatTok->Tok.is(tok::kw_break)) {
1478 if (Style.BraceWrapping.AfterControlStatement)
1480 parseStructuralElement();
1484 if (FormatTok->is(tok::semi))
1488 Line->Level = OldLineLevel;
1491 void UnwrappedLineParser::parseCaseLabel() {
1492 assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected");
1493 // FIXME: fix handling of complex expressions here.
1496 } while (!eof() && !FormatTok->Tok.is(tok::colon));
1500 void UnwrappedLineParser::parseSwitch() {
1501 assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected");
1503 if (FormatTok->Tok.is(tok::l_paren))
1505 if (FormatTok->Tok.is(tok::l_brace)) {
1506 CompoundStatementIndenter Indenter(this, Style, Line->Level);
1507 parseBlock(/*MustBeDeclaration=*/false);
1512 parseStructuralElement();
1517 void UnwrappedLineParser::parseAccessSpecifier() {
1519 // Understand Qt's slots.
1520 if (FormatTok->isOneOf(Keywords.kw_slots, Keywords.kw_qslots))
1522 // Otherwise, we don't know what it is, and we'd better keep the next token.
1523 if (FormatTok->Tok.is(tok::colon))
1528 bool UnwrappedLineParser::parseEnum() {
1529 // Won't be 'enum' for NS_ENUMs.
1530 if (FormatTok->Tok.is(tok::kw_enum))
1533 // In TypeScript, "enum" can also be used as property name, e.g. in interface
1534 // declarations. An "enum" keyword followed by a colon would be a syntax
1535 // error and thus assume it is just an identifier.
1536 if (Style.Language == FormatStyle::LK_JavaScript && FormatTok->is(tok::colon))
1539 // Eat up enum class ...
1540 if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct))
1543 while (FormatTok->Tok.getIdentifierInfo() ||
1544 FormatTok->isOneOf(tok::colon, tok::coloncolon, tok::less,
1545 tok::greater, tok::comma, tok::question)) {
1547 // We can have macros or attributes in between 'enum' and the enum name.
1548 if (FormatTok->is(tok::l_paren))
1550 if (FormatTok->is(tok::identifier)) {
1552 // If there are two identifiers in a row, this is likely an elaborate
1553 // return type. In Java, this can be "implements", etc.
1554 if (Style.Language == FormatStyle::LK_Cpp &&
1555 FormatTok->is(tok::identifier))
1560 // Just a declaration or something is wrong.
1561 if (FormatTok->isNot(tok::l_brace))
1563 FormatTok->BlockKind = BK_Block;
1565 if (Style.Language == FormatStyle::LK_Java) {
1566 // Java enums are different.
1567 parseJavaEnumBody();
1570 if (Style.Language == FormatStyle::LK_Proto) {
1571 parseBlock(/*MustBeDeclaration=*/true);
1576 bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true);
1578 if (FormatTok->is(tok::semi))
1584 // There is no addUnwrappedLine() here so that we fall through to parsing a
1585 // structural element afterwards. Thus, in "enum A {} n, m;",
1586 // "} n, m;" will end up in one unwrapped line.
1589 void UnwrappedLineParser::parseJavaEnumBody() {
1590 // Determine whether the enum is simple, i.e. does not have a semicolon or
1591 // constants with class bodies. Simple enums can be formatted like braced
1592 // lists, contracted to a single line, etc.
1593 unsigned StoredPosition = Tokens->getPosition();
1594 bool IsSimple = true;
1595 FormatToken *Tok = Tokens->getNextToken();
1597 if (Tok->is(tok::r_brace))
1599 if (Tok->isOneOf(tok::l_brace, tok::semi)) {
1603 // FIXME: This will also mark enums with braces in the arguments to enum
1604 // constants as "not simple". This is probably fine in practice, though.
1605 Tok = Tokens->getNextToken();
1607 FormatTok = Tokens->setPosition(StoredPosition);
1615 // Parse the body of a more complex enum.
1616 // First add a line for everything up to the "{".
1621 // Parse the enum constants.
1623 if (FormatTok->is(tok::l_brace)) {
1624 // Parse the constant's class body.
1625 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true,
1626 /*MunchSemi=*/false);
1627 } else if (FormatTok->is(tok::l_paren)) {
1629 } else if (FormatTok->is(tok::comma)) {
1632 } else if (FormatTok->is(tok::semi)) {
1636 } else if (FormatTok->is(tok::r_brace)) {
1644 // Parse the class body after the enum's ";" if any.
1645 parseLevel(/*HasOpeningBrace=*/true);
1651 void UnwrappedLineParser::parseRecord() {
1652 const FormatToken &InitialToken = *FormatTok;
1655 // The actual identifier can be a nested name specifier, and in macros
1656 // it is often token-pasted.
1657 while (FormatTok->isOneOf(tok::identifier, tok::coloncolon, tok::hashhash,
1658 tok::kw___attribute, tok::kw___declspec,
1660 ((Style.Language == FormatStyle::LK_Java ||
1661 Style.Language == FormatStyle::LK_JavaScript) &&
1662 FormatTok->isOneOf(tok::period, tok::comma))) {
1663 bool IsNonMacroIdentifier =
1664 FormatTok->is(tok::identifier) &&
1665 FormatTok->TokenText != FormatTok->TokenText.upper();
1667 // We can have macros or attributes in between 'class' and the class name.
1668 if (!IsNonMacroIdentifier && FormatTok->Tok.is(tok::l_paren))
1672 // Note that parsing away template declarations here leads to incorrectly
1673 // accepting function declarations as record declarations.
1674 // In general, we cannot solve this problem. Consider:
1675 // class A<int> B() {}
1676 // which can be a function definition or a class definition when B() is a
1677 // macro. If we find enough real-world cases where this is a problem, we
1678 // can parse for the 'template' keyword in the beginning of the statement,
1679 // and thus rule out the record production in case there is no template
1680 // (this would still leave us with an ambiguity between template function
1681 // and class declarations).
1682 if (FormatTok->isOneOf(tok::colon, tok::less)) {
1684 if (FormatTok->is(tok::l_brace)) {
1685 calculateBraceTypes(/*ExpectClassBody=*/true);
1686 if (!tryToParseBracedList())
1689 if (FormatTok->Tok.is(tok::semi))
1694 if (FormatTok->Tok.is(tok::l_brace)) {
1695 if (ShouldBreakBeforeBrace(Style, InitialToken))
1698 parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true,
1699 /*MunchSemi=*/false);
1701 // There is no addUnwrappedLine() here so that we fall through to parsing a
1702 // structural element afterwards. Thus, in "class A {} n, m;",
1703 // "} n, m;" will end up in one unwrapped line.
1706 void UnwrappedLineParser::parseObjCProtocolList() {
1707 assert(FormatTok->Tok.is(tok::less) && "'<' expected.");
1710 while (!eof() && FormatTok->Tok.isNot(tok::greater));
1711 nextToken(); // Skip '>'.
1714 void UnwrappedLineParser::parseObjCUntilAtEnd() {
1716 if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) {
1721 if (FormatTok->is(tok::l_brace)) {
1722 parseBlock(/*MustBeDeclaration=*/false);
1723 // In ObjC interfaces, nothing should be following the "}".
1725 } else if (FormatTok->is(tok::r_brace)) {
1726 // Ignore stray "}". parseStructuralElement doesn't consume them.
1730 parseStructuralElement();
1735 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() {
1737 nextToken(); // interface name
1739 // @interface can be followed by either a base class, or a category.
1740 if (FormatTok->Tok.is(tok::colon)) {
1742 nextToken(); // base class name
1743 } else if (FormatTok->Tok.is(tok::l_paren))
1744 // Skip category, if present.
1747 if (FormatTok->Tok.is(tok::less))
1748 parseObjCProtocolList();
1750 if (FormatTok->Tok.is(tok::l_brace)) {
1751 if (Style.BraceWrapping.AfterObjCDeclaration)
1753 parseBlock(/*MustBeDeclaration=*/true);
1756 // With instance variables, this puts '}' on its own line. Without instance
1757 // variables, this ends the @interface line.
1760 parseObjCUntilAtEnd();
1763 void UnwrappedLineParser::parseObjCProtocol() {
1765 nextToken(); // protocol name
1767 if (FormatTok->Tok.is(tok::less))
1768 parseObjCProtocolList();
1770 // Check for protocol declaration.
1771 if (FormatTok->Tok.is(tok::semi)) {
1773 return addUnwrappedLine();
1777 parseObjCUntilAtEnd();
1780 void UnwrappedLineParser::parseJavaScriptEs6ImportExport() {
1781 assert(FormatTok->isOneOf(Keywords.kw_import, tok::kw_export));
1784 // Consume the "default" in "export default class/function".
1785 if (FormatTok->is(tok::kw_default))
1788 // Consume "function" and "default function", so that these get parsed as
1789 // free-standing JS functions, i.e. do not require a trailing semicolon.
1790 if (FormatTok->is(Keywords.kw_function)) {
1795 if (FormatTok->isOneOf(tok::kw_const, tok::kw_class, tok::kw_enum,
1796 Keywords.kw_let, Keywords.kw_var))
1797 return; // Fall through to parsing the corresponding structure.
1799 if (FormatTok->is(tok::l_brace)) {
1800 FormatTok->BlockKind = BK_Block;
1804 while (!eof() && FormatTok->isNot(tok::semi) &&
1805 FormatTok->isNot(tok::l_brace)) {
1810 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line,
1811 StringRef Prefix = "") {
1812 llvm::dbgs() << Prefix << "Line(" << Line.Level << ")"
1813 << (Line.InPPDirective ? " MACRO" : "") << ": ";
1814 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(),
1815 E = Line.Tokens.end();
1817 llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] ";
1819 for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(),
1820 E = Line.Tokens.end();
1822 const UnwrappedLineNode &Node = *I;
1823 for (SmallVectorImpl<UnwrappedLine>::const_iterator
1824 I = Node.Children.begin(),
1825 E = Node.Children.end();
1827 printDebugInfo(*I, "\nChild: ");
1830 llvm::dbgs() << "\n";
1833 void UnwrappedLineParser::addUnwrappedLine() {
1834 if (Line->Tokens.empty())
1837 if (CurrentLines == &Lines)
1838 printDebugInfo(*Line);
1840 CurrentLines->push_back(std::move(*Line));
1841 Line->Tokens.clear();
1842 if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) {
1843 CurrentLines->append(
1844 std::make_move_iterator(PreprocessorDirectives.begin()),
1845 std::make_move_iterator(PreprocessorDirectives.end()));
1846 PreprocessorDirectives.clear();
1850 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); }
1852 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) {
1853 return (Line->InPPDirective || FormatTok.HasUnescapedNewline) &&
1854 FormatTok.NewlinesBefore > 0;
1857 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) {
1858 bool JustComments = Line->Tokens.empty();
1859 for (SmallVectorImpl<FormatToken *>::const_iterator
1860 I = CommentsBeforeNextToken.begin(),
1861 E = CommentsBeforeNextToken.end();
1863 if (isOnNewLine(**I) && JustComments)
1867 if (NewlineBeforeNext && JustComments)
1869 CommentsBeforeNextToken.clear();
1872 void UnwrappedLineParser::nextToken() {
1875 flushComments(isOnNewLine(*FormatTok));
1876 pushToken(FormatTok);
1880 void UnwrappedLineParser::readToken() {
1881 bool CommentsInCurrentLine = true;
1883 FormatTok = Tokens->getNextToken();
1885 while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) &&
1886 (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) {
1887 // If there is an unfinished unwrapped line, we flush the preprocessor
1888 // directives only after that unwrapped line was finished later.
1889 bool SwitchToPreprocessorLines = !Line->Tokens.empty();
1890 ScopedLineState BlockState(*this, SwitchToPreprocessorLines);
1891 // Comments stored before the preprocessor directive need to be output
1892 // before the preprocessor directive, at the same level as the
1893 // preprocessor directive, as we consider them to apply to the directive.
1894 flushComments(isOnNewLine(*FormatTok));
1897 while (FormatTok->Type == TT_ConflictStart ||
1898 FormatTok->Type == TT_ConflictEnd ||
1899 FormatTok->Type == TT_ConflictAlternative) {
1900 if (FormatTok->Type == TT_ConflictStart) {
1901 conditionalCompilationStart(/*Unreachable=*/false);
1902 } else if (FormatTok->Type == TT_ConflictAlternative) {
1903 conditionalCompilationAlternative();
1904 } else if (FormatTok->Type == TT_ConflictEnd) {
1905 conditionalCompilationEnd();
1907 FormatTok = Tokens->getNextToken();
1908 FormatTok->MustBreakBefore = true;
1911 if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) &&
1912 !Line->InPPDirective) {
1916 if (!FormatTok->Tok.is(tok::comment))
1918 if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) {
1919 CommentsInCurrentLine = false;
1921 if (CommentsInCurrentLine) {
1922 pushToken(FormatTok);
1924 CommentsBeforeNextToken.push_back(FormatTok);
1929 void UnwrappedLineParser::pushToken(FormatToken *Tok) {
1930 Line->Tokens.push_back(UnwrappedLineNode(Tok));
1931 if (MustBreakBeforeNextToken) {
1932 Line->Tokens.back().Tok->MustBreakBefore = true;
1933 MustBreakBeforeNextToken = false;
1937 } // end namespace format
1938 } // end namespace clang