• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //===--- UnwrappedLineParser.cpp - Format C++ code ------------------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 ///
10 /// \file
11 /// \brief This file contains the implementation of the UnwrappedLineParser,
12 /// which turns a stream of tokens into UnwrappedLines.
13 ///
14 //===----------------------------------------------------------------------===//
15 
16 #include "UnwrappedLineParser.h"
17 #include "llvm/Support/Debug.h"
18 
19 #define DEBUG_TYPE "format-parser"
20 
21 namespace clang {
22 namespace format {
23 
24 class FormatTokenSource {
25 public:
~FormatTokenSource()26   virtual ~FormatTokenSource() {}
27   virtual FormatToken *getNextToken() = 0;
28 
29   virtual unsigned getPosition() = 0;
30   virtual FormatToken *setPosition(unsigned Position) = 0;
31 };
32 
33 namespace {
34 
35 class ScopedDeclarationState {
36 public:
ScopedDeclarationState(UnwrappedLine & Line,std::vector<bool> & Stack,bool MustBeDeclaration)37   ScopedDeclarationState(UnwrappedLine &Line, std::vector<bool> &Stack,
38                          bool MustBeDeclaration)
39       : Line(Line), Stack(Stack) {
40     Line.MustBeDeclaration = MustBeDeclaration;
41     Stack.push_back(MustBeDeclaration);
42   }
~ScopedDeclarationState()43   ~ScopedDeclarationState() {
44     Stack.pop_back();
45     if (!Stack.empty())
46       Line.MustBeDeclaration = Stack.back();
47     else
48       Line.MustBeDeclaration = true;
49   }
50 
51 private:
52   UnwrappedLine &Line;
53   std::vector<bool> &Stack;
54 };
55 
56 class ScopedMacroState : public FormatTokenSource {
57 public:
ScopedMacroState(UnwrappedLine & Line,FormatTokenSource * & TokenSource,FormatToken * & ResetToken,bool & StructuralError)58   ScopedMacroState(UnwrappedLine &Line, FormatTokenSource *&TokenSource,
59                    FormatToken *&ResetToken, bool &StructuralError)
60       : Line(Line), TokenSource(TokenSource), ResetToken(ResetToken),
61         PreviousLineLevel(Line.Level), PreviousTokenSource(TokenSource),
62         StructuralError(StructuralError),
63         PreviousStructuralError(StructuralError), Token(nullptr) {
64     TokenSource = this;
65     Line.Level = 0;
66     Line.InPPDirective = true;
67   }
68 
~ScopedMacroState()69   ~ScopedMacroState() {
70     TokenSource = PreviousTokenSource;
71     ResetToken = Token;
72     Line.InPPDirective = false;
73     Line.Level = PreviousLineLevel;
74     StructuralError = PreviousStructuralError;
75   }
76 
getNextToken()77   FormatToken *getNextToken() override {
78     // The \c UnwrappedLineParser guards against this by never calling
79     // \c getNextToken() after it has encountered the first eof token.
80     assert(!eof());
81     Token = PreviousTokenSource->getNextToken();
82     if (eof())
83       return getFakeEOF();
84     return Token;
85   }
86 
getPosition()87   unsigned getPosition() override { return PreviousTokenSource->getPosition(); }
88 
setPosition(unsigned Position)89   FormatToken *setPosition(unsigned Position) override {
90     Token = PreviousTokenSource->setPosition(Position);
91     return Token;
92   }
93 
94 private:
eof()95   bool eof() { return Token && Token->HasUnescapedNewline; }
96 
getFakeEOF()97   FormatToken *getFakeEOF() {
98     static bool EOFInitialized = false;
99     static FormatToken FormatTok;
100     if (!EOFInitialized) {
101       FormatTok.Tok.startToken();
102       FormatTok.Tok.setKind(tok::eof);
103       EOFInitialized = true;
104     }
105     return &FormatTok;
106   }
107 
108   UnwrappedLine &Line;
109   FormatTokenSource *&TokenSource;
110   FormatToken *&ResetToken;
111   unsigned PreviousLineLevel;
112   FormatTokenSource *PreviousTokenSource;
113   bool &StructuralError;
114   bool PreviousStructuralError;
115 
116   FormatToken *Token;
117 };
118 
119 } // end anonymous namespace
120 
121 class ScopedLineState {
122 public:
ScopedLineState(UnwrappedLineParser & Parser,bool SwitchToPreprocessorLines=false)123   ScopedLineState(UnwrappedLineParser &Parser,
124                   bool SwitchToPreprocessorLines = false)
125       : Parser(Parser) {
126     OriginalLines = Parser.CurrentLines;
127     if (SwitchToPreprocessorLines)
128       Parser.CurrentLines = &Parser.PreprocessorDirectives;
129     else if (!Parser.Line->Tokens.empty())
130       Parser.CurrentLines = &Parser.Line->Tokens.back().Children;
131     PreBlockLine = Parser.Line.release();
132     Parser.Line.reset(new UnwrappedLine());
133     Parser.Line->Level = PreBlockLine->Level;
134     Parser.Line->InPPDirective = PreBlockLine->InPPDirective;
135   }
136 
~ScopedLineState()137   ~ScopedLineState() {
138     if (!Parser.Line->Tokens.empty()) {
139       Parser.addUnwrappedLine();
140     }
141     assert(Parser.Line->Tokens.empty());
142     Parser.Line.reset(PreBlockLine);
143     if (Parser.CurrentLines == &Parser.PreprocessorDirectives)
144       Parser.MustBreakBeforeNextToken = true;
145     Parser.CurrentLines = OriginalLines;
146   }
147 
148 private:
149   UnwrappedLineParser &Parser;
150 
151   UnwrappedLine *PreBlockLine;
152   SmallVectorImpl<UnwrappedLine> *OriginalLines;
153 };
154 
155 class CompoundStatementIndenter {
156 public:
CompoundStatementIndenter(UnwrappedLineParser * Parser,const FormatStyle & Style,unsigned & LineLevel)157   CompoundStatementIndenter(UnwrappedLineParser *Parser,
158                             const FormatStyle &Style, unsigned &LineLevel)
159       : LineLevel(LineLevel), OldLineLevel(LineLevel) {
160     if (Style.BreakBeforeBraces == FormatStyle::BS_Allman) {
161       Parser->addUnwrappedLine();
162     } else if (Style.BreakBeforeBraces == FormatStyle::BS_GNU) {
163       Parser->addUnwrappedLine();
164       ++LineLevel;
165     }
166   }
~CompoundStatementIndenter()167   ~CompoundStatementIndenter() { LineLevel = OldLineLevel; }
168 
169 private:
170   unsigned &LineLevel;
171   unsigned OldLineLevel;
172 };
173 
174 namespace {
175 
176 class IndexedTokenSource : public FormatTokenSource {
177 public:
IndexedTokenSource(ArrayRef<FormatToken * > Tokens)178   IndexedTokenSource(ArrayRef<FormatToken *> Tokens)
179       : Tokens(Tokens), Position(-1) {}
180 
getNextToken()181   FormatToken *getNextToken() override {
182     ++Position;
183     return Tokens[Position];
184   }
185 
getPosition()186   unsigned getPosition() override {
187     assert(Position >= 0);
188     return Position;
189   }
190 
setPosition(unsigned P)191   FormatToken *setPosition(unsigned P) override {
192     Position = P;
193     return Tokens[Position];
194   }
195 
reset()196   void reset() { Position = -1; }
197 
198 private:
199   ArrayRef<FormatToken *> Tokens;
200   int Position;
201 };
202 
203 } // end anonymous namespace
204 
UnwrappedLineParser(const FormatStyle & Style,ArrayRef<FormatToken * > Tokens,UnwrappedLineConsumer & Callback)205 UnwrappedLineParser::UnwrappedLineParser(const FormatStyle &Style,
206                                          ArrayRef<FormatToken *> Tokens,
207                                          UnwrappedLineConsumer &Callback)
208     : Line(new UnwrappedLine), MustBreakBeforeNextToken(false),
209       CurrentLines(&Lines), StructuralError(false), Style(Style),
210       Tokens(nullptr), Callback(Callback), AllTokens(Tokens),
211       PPBranchLevel(-1) {}
212 
reset()213 void UnwrappedLineParser::reset() {
214   PPBranchLevel = -1;
215   Line.reset(new UnwrappedLine);
216   CommentsBeforeNextToken.clear();
217   FormatTok = nullptr;
218   MustBreakBeforeNextToken = false;
219   PreprocessorDirectives.clear();
220   CurrentLines = &Lines;
221   DeclarationScopeStack.clear();
222   StructuralError = false;
223   PPStack.clear();
224 }
225 
parse()226 bool UnwrappedLineParser::parse() {
227   IndexedTokenSource TokenSource(AllTokens);
228   do {
229     DEBUG(llvm::dbgs() << "----\n");
230     reset();
231     Tokens = &TokenSource;
232     TokenSource.reset();
233 
234     readToken();
235     parseFile();
236     // Create line with eof token.
237     pushToken(FormatTok);
238     addUnwrappedLine();
239 
240     for (SmallVectorImpl<UnwrappedLine>::iterator I = Lines.begin(),
241                                                   E = Lines.end();
242          I != E; ++I) {
243       Callback.consumeUnwrappedLine(*I);
244     }
245     Callback.finishRun();
246     Lines.clear();
247     while (!PPLevelBranchIndex.empty() &&
248            PPLevelBranchIndex.back() + 1 >= PPLevelBranchCount.back()) {
249       PPLevelBranchIndex.resize(PPLevelBranchIndex.size() - 1);
250       PPLevelBranchCount.resize(PPLevelBranchCount.size() - 1);
251     }
252     if (!PPLevelBranchIndex.empty()) {
253       ++PPLevelBranchIndex.back();
254       assert(PPLevelBranchIndex.size() == PPLevelBranchCount.size());
255       assert(PPLevelBranchIndex.back() <= PPLevelBranchCount.back());
256     }
257   } while (!PPLevelBranchIndex.empty());
258 
259   return StructuralError;
260 }
261 
parseFile()262 void UnwrappedLineParser::parseFile() {
263   ScopedDeclarationState DeclarationState(
264       *Line, DeclarationScopeStack,
265       /*MustBeDeclaration=*/ !Line->InPPDirective);
266   parseLevel(/*HasOpeningBrace=*/false);
267   // Make sure to format the remaining tokens.
268   flushComments(true);
269   addUnwrappedLine();
270 }
271 
parseLevel(bool HasOpeningBrace)272 void UnwrappedLineParser::parseLevel(bool HasOpeningBrace) {
273   bool SwitchLabelEncountered = false;
274   do {
275     switch (FormatTok->Tok.getKind()) {
276     case tok::comment:
277       nextToken();
278       addUnwrappedLine();
279       break;
280     case tok::l_brace:
281       // FIXME: Add parameter whether this can happen - if this happens, we must
282       // be in a non-declaration context.
283       parseBlock(/*MustBeDeclaration=*/false);
284       addUnwrappedLine();
285       break;
286     case tok::r_brace:
287       if (HasOpeningBrace)
288         return;
289       StructuralError = true;
290       nextToken();
291       addUnwrappedLine();
292       break;
293     case tok::kw_default:
294     case tok::kw_case:
295       if (!SwitchLabelEncountered &&
296           (Style.IndentCaseLabels || (Line->InPPDirective && Line->Level == 1)))
297         ++Line->Level;
298       SwitchLabelEncountered = true;
299       parseStructuralElement();
300       break;
301     default:
302       parseStructuralElement();
303       break;
304     }
305   } while (!eof());
306 }
307 
calculateBraceTypes()308 void UnwrappedLineParser::calculateBraceTypes() {
309   // We'll parse forward through the tokens until we hit
310   // a closing brace or eof - note that getNextToken() will
311   // parse macros, so this will magically work inside macro
312   // definitions, too.
313   unsigned StoredPosition = Tokens->getPosition();
314   unsigned Position = StoredPosition;
315   FormatToken *Tok = FormatTok;
316   // Keep a stack of positions of lbrace tokens. We will
317   // update information about whether an lbrace starts a
318   // braced init list or a different block during the loop.
319   SmallVector<FormatToken *, 8> LBraceStack;
320   assert(Tok->Tok.is(tok::l_brace));
321   do {
322     // Get next none-comment token.
323     FormatToken *NextTok;
324     unsigned ReadTokens = 0;
325     do {
326       NextTok = Tokens->getNextToken();
327       ++ReadTokens;
328     } while (NextTok->is(tok::comment));
329 
330     switch (Tok->Tok.getKind()) {
331     case tok::l_brace:
332       LBraceStack.push_back(Tok);
333       break;
334     case tok::r_brace:
335       if (!LBraceStack.empty()) {
336         if (LBraceStack.back()->BlockKind == BK_Unknown) {
337           bool ProbablyBracedList = false;
338           if (Style.Language == FormatStyle::LK_Proto) {
339             ProbablyBracedList = NextTok->isOneOf(tok::comma, tok::r_square);
340           } else {
341             // Using OriginalColumn to distinguish between ObjC methods and
342             // binary operators is a bit hacky.
343             bool NextIsObjCMethod = NextTok->isOneOf(tok::plus, tok::minus) &&
344                                     NextTok->OriginalColumn == 0;
345 
346             // If there is a comma, semicolon or right paren after the closing
347             // brace, we assume this is a braced initializer list.  Note that
348             // regardless how we mark inner braces here, we will overwrite the
349             // BlockKind later if we parse a braced list (where all blocks
350             // inside are by default braced lists), or when we explicitly detect
351             // blocks (for example while parsing lambdas).
352             //
353             // We exclude + and - as they can be ObjC visibility modifiers.
354             ProbablyBracedList =
355                 NextTok->isOneOf(tok::comma, tok::semi, tok::period, tok::colon,
356                                  tok::r_paren, tok::r_square, tok::l_brace,
357                                  tok::l_paren) ||
358                 (NextTok->isBinaryOperator() && !NextIsObjCMethod);
359           }
360           if (ProbablyBracedList) {
361             Tok->BlockKind = BK_BracedInit;
362             LBraceStack.back()->BlockKind = BK_BracedInit;
363           } else {
364             Tok->BlockKind = BK_Block;
365             LBraceStack.back()->BlockKind = BK_Block;
366           }
367         }
368         LBraceStack.pop_back();
369       }
370       break;
371     case tok::at:
372     case tok::semi:
373     case tok::kw_if:
374     case tok::kw_while:
375     case tok::kw_for:
376     case tok::kw_switch:
377     case tok::kw_try:
378       if (!LBraceStack.empty())
379         LBraceStack.back()->BlockKind = BK_Block;
380       break;
381     default:
382       break;
383     }
384     Tok = NextTok;
385     Position += ReadTokens;
386   } while (Tok->Tok.isNot(tok::eof) && !LBraceStack.empty());
387   // Assume other blocks for all unclosed opening braces.
388   for (unsigned i = 0, e = LBraceStack.size(); i != e; ++i) {
389     if (LBraceStack[i]->BlockKind == BK_Unknown)
390       LBraceStack[i]->BlockKind = BK_Block;
391   }
392 
393   FormatTok = Tokens->setPosition(StoredPosition);
394 }
395 
parseBlock(bool MustBeDeclaration,bool AddLevel,bool MunchSemi)396 void UnwrappedLineParser::parseBlock(bool MustBeDeclaration, bool AddLevel,
397                                      bool MunchSemi) {
398   assert(FormatTok->Tok.is(tok::l_brace) && "'{' expected");
399   unsigned InitialLevel = Line->Level;
400   nextToken();
401 
402   addUnwrappedLine();
403 
404   ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack,
405                                           MustBeDeclaration);
406   if (AddLevel)
407     ++Line->Level;
408   parseLevel(/*HasOpeningBrace=*/true);
409 
410   if (!FormatTok->Tok.is(tok::r_brace)) {
411     Line->Level = InitialLevel;
412     StructuralError = true;
413     return;
414   }
415 
416   nextToken(); // Munch the closing brace.
417   if (MunchSemi && FormatTok->Tok.is(tok::semi))
418     nextToken();
419   Line->Level = InitialLevel;
420 }
421 
IsGoogScope(const UnwrappedLine & Line)422 static bool IsGoogScope(const UnwrappedLine &Line) {
423   if (Line.Tokens.size() < 4)
424     return false;
425   auto I = Line.Tokens.begin();
426   if (I->Tok->TokenText != "goog")
427     return false;
428   ++I;
429   if (I->Tok->isNot(tok::period))
430     return false;
431   ++I;
432   if (I->Tok->TokenText != "scope")
433     return false;
434   ++I;
435   return I->Tok->is(tok::l_paren);
436 }
437 
parseChildBlock()438 void UnwrappedLineParser::parseChildBlock() {
439   FormatTok->BlockKind = BK_Block;
440   nextToken();
441   {
442     bool GoogScope =
443         Style.Language == FormatStyle::LK_JavaScript && IsGoogScope(*Line);
444     ScopedLineState LineState(*this);
445     ScopedDeclarationState DeclarationState(*Line, DeclarationScopeStack,
446                                             /*MustBeDeclaration=*/false);
447     Line->Level += GoogScope ? 0 : 1;
448     parseLevel(/*HasOpeningBrace=*/true);
449     Line->Level -= GoogScope ? 0 : 1;
450   }
451   nextToken();
452 }
453 
parsePPDirective()454 void UnwrappedLineParser::parsePPDirective() {
455   assert(FormatTok->Tok.is(tok::hash) && "'#' expected");
456   ScopedMacroState MacroState(*Line, Tokens, FormatTok, StructuralError);
457   nextToken();
458 
459   if (!FormatTok->Tok.getIdentifierInfo()) {
460     parsePPUnknown();
461     return;
462   }
463 
464   switch (FormatTok->Tok.getIdentifierInfo()->getPPKeywordID()) {
465   case tok::pp_define:
466     parsePPDefine();
467     return;
468   case tok::pp_if:
469     parsePPIf(/*IfDef=*/false);
470     break;
471   case tok::pp_ifdef:
472   case tok::pp_ifndef:
473     parsePPIf(/*IfDef=*/true);
474     break;
475   case tok::pp_else:
476     parsePPElse();
477     break;
478   case tok::pp_elif:
479     parsePPElIf();
480     break;
481   case tok::pp_endif:
482     parsePPEndIf();
483     break;
484   default:
485     parsePPUnknown();
486     break;
487   }
488 }
489 
conditionalCompilationCondition(bool Unreachable)490 void UnwrappedLineParser::conditionalCompilationCondition(bool Unreachable) {
491   if (Unreachable || (!PPStack.empty() && PPStack.back() == PP_Unreachable))
492     PPStack.push_back(PP_Unreachable);
493   else
494     PPStack.push_back(PP_Conditional);
495 }
496 
conditionalCompilationStart(bool Unreachable)497 void UnwrappedLineParser::conditionalCompilationStart(bool Unreachable) {
498   ++PPBranchLevel;
499   assert(PPBranchLevel >= 0 && PPBranchLevel <= (int)PPLevelBranchIndex.size());
500   if (PPBranchLevel == (int)PPLevelBranchIndex.size()) {
501     PPLevelBranchIndex.push_back(0);
502     PPLevelBranchCount.push_back(0);
503   }
504   PPChainBranchIndex.push(0);
505   bool Skip = PPLevelBranchIndex[PPBranchLevel] > 0;
506   conditionalCompilationCondition(Unreachable || Skip);
507 }
508 
conditionalCompilationAlternative()509 void UnwrappedLineParser::conditionalCompilationAlternative() {
510   if (!PPStack.empty())
511     PPStack.pop_back();
512   assert(PPBranchLevel < (int)PPLevelBranchIndex.size());
513   if (!PPChainBranchIndex.empty())
514     ++PPChainBranchIndex.top();
515   conditionalCompilationCondition(
516       PPBranchLevel >= 0 && !PPChainBranchIndex.empty() &&
517       PPLevelBranchIndex[PPBranchLevel] != PPChainBranchIndex.top());
518 }
519 
conditionalCompilationEnd()520 void UnwrappedLineParser::conditionalCompilationEnd() {
521   assert(PPBranchLevel < (int)PPLevelBranchIndex.size());
522   if (PPBranchLevel >= 0 && !PPChainBranchIndex.empty()) {
523     if (PPChainBranchIndex.top() + 1 > PPLevelBranchCount[PPBranchLevel]) {
524       PPLevelBranchCount[PPBranchLevel] = PPChainBranchIndex.top() + 1;
525     }
526   }
527   // Guard against #endif's without #if.
528   if (PPBranchLevel > 0)
529     --PPBranchLevel;
530   if (!PPChainBranchIndex.empty())
531     PPChainBranchIndex.pop();
532   if (!PPStack.empty())
533     PPStack.pop_back();
534 }
535 
parsePPIf(bool IfDef)536 void UnwrappedLineParser::parsePPIf(bool IfDef) {
537   nextToken();
538   bool IsLiteralFalse = (FormatTok->Tok.isLiteral() &&
539                          StringRef(FormatTok->Tok.getLiteralData(),
540                                    FormatTok->Tok.getLength()) == "0") ||
541                         FormatTok->Tok.is(tok::kw_false);
542   conditionalCompilationStart(!IfDef && IsLiteralFalse);
543   parsePPUnknown();
544 }
545 
parsePPElse()546 void UnwrappedLineParser::parsePPElse() {
547   conditionalCompilationAlternative();
548   parsePPUnknown();
549 }
550 
parsePPElIf()551 void UnwrappedLineParser::parsePPElIf() { parsePPElse(); }
552 
parsePPEndIf()553 void UnwrappedLineParser::parsePPEndIf() {
554   conditionalCompilationEnd();
555   parsePPUnknown();
556 }
557 
parsePPDefine()558 void UnwrappedLineParser::parsePPDefine() {
559   nextToken();
560 
561   if (FormatTok->Tok.getKind() != tok::identifier) {
562     parsePPUnknown();
563     return;
564   }
565   nextToken();
566   if (FormatTok->Tok.getKind() == tok::l_paren &&
567       FormatTok->WhitespaceRange.getBegin() ==
568           FormatTok->WhitespaceRange.getEnd()) {
569     parseParens();
570   }
571   addUnwrappedLine();
572   Line->Level = 1;
573 
574   // Errors during a preprocessor directive can only affect the layout of the
575   // preprocessor directive, and thus we ignore them. An alternative approach
576   // would be to use the same approach we use on the file level (no
577   // re-indentation if there was a structural error) within the macro
578   // definition.
579   parseFile();
580 }
581 
parsePPUnknown()582 void UnwrappedLineParser::parsePPUnknown() {
583   do {
584     nextToken();
585   } while (!eof());
586   addUnwrappedLine();
587 }
588 
589 // Here we blacklist certain tokens that are not usually the first token in an
590 // unwrapped line. This is used in attempt to distinguish macro calls without
591 // trailing semicolons from other constructs split to several lines.
tokenCanStartNewLine(clang::Token Tok)592 bool tokenCanStartNewLine(clang::Token Tok) {
593   // Semicolon can be a null-statement, l_square can be a start of a macro or
594   // a C++11 attribute, but this doesn't seem to be common.
595   return Tok.isNot(tok::semi) && Tok.isNot(tok::l_brace) &&
596          Tok.isNot(tok::l_square) &&
597          // Tokens that can only be used as binary operators and a part of
598          // overloaded operator names.
599          Tok.isNot(tok::period) && Tok.isNot(tok::periodstar) &&
600          Tok.isNot(tok::arrow) && Tok.isNot(tok::arrowstar) &&
601          Tok.isNot(tok::less) && Tok.isNot(tok::greater) &&
602          Tok.isNot(tok::slash) && Tok.isNot(tok::percent) &&
603          Tok.isNot(tok::lessless) && Tok.isNot(tok::greatergreater) &&
604          Tok.isNot(tok::equal) && Tok.isNot(tok::plusequal) &&
605          Tok.isNot(tok::minusequal) && Tok.isNot(tok::starequal) &&
606          Tok.isNot(tok::slashequal) && Tok.isNot(tok::percentequal) &&
607          Tok.isNot(tok::ampequal) && Tok.isNot(tok::pipeequal) &&
608          Tok.isNot(tok::caretequal) && Tok.isNot(tok::greatergreaterequal) &&
609          Tok.isNot(tok::lesslessequal) &&
610          // Colon is used in labels, base class lists, initializer lists,
611          // range-based for loops, ternary operator, but should never be the
612          // first token in an unwrapped line.
613          Tok.isNot(tok::colon) &&
614          // 'noexcept' is a trailing annotation.
615          Tok.isNot(tok::kw_noexcept);
616 }
617 
parseStructuralElement()618 void UnwrappedLineParser::parseStructuralElement() {
619   assert(!FormatTok->Tok.is(tok::l_brace));
620   switch (FormatTok->Tok.getKind()) {
621   case tok::at:
622     nextToken();
623     if (FormatTok->Tok.is(tok::l_brace)) {
624       parseBracedList();
625       break;
626     }
627     switch (FormatTok->Tok.getObjCKeywordID()) {
628     case tok::objc_public:
629     case tok::objc_protected:
630     case tok::objc_package:
631     case tok::objc_private:
632       return parseAccessSpecifier();
633     case tok::objc_interface:
634     case tok::objc_implementation:
635       return parseObjCInterfaceOrImplementation();
636     case tok::objc_protocol:
637       return parseObjCProtocol();
638     case tok::objc_end:
639       return; // Handled by the caller.
640     case tok::objc_optional:
641     case tok::objc_required:
642       nextToken();
643       addUnwrappedLine();
644       return;
645     default:
646       break;
647     }
648     break;
649   case tok::kw_namespace:
650     parseNamespace();
651     return;
652   case tok::kw_inline:
653     nextToken();
654     if (FormatTok->Tok.is(tok::kw_namespace)) {
655       parseNamespace();
656       return;
657     }
658     break;
659   case tok::kw_public:
660   case tok::kw_protected:
661   case tok::kw_private:
662     parseAccessSpecifier();
663     return;
664   case tok::kw_if:
665     parseIfThenElse();
666     return;
667   case tok::kw_for:
668   case tok::kw_while:
669     parseForOrWhileLoop();
670     return;
671   case tok::kw_do:
672     parseDoWhile();
673     return;
674   case tok::kw_switch:
675     parseSwitch();
676     return;
677   case tok::kw_default:
678     nextToken();
679     parseLabel();
680     return;
681   case tok::kw_case:
682     parseCaseLabel();
683     return;
684   case tok::kw_try:
685     parseTryCatch();
686     return;
687   case tok::kw_extern:
688     nextToken();
689     if (FormatTok->Tok.is(tok::string_literal)) {
690       nextToken();
691       if (FormatTok->Tok.is(tok::l_brace)) {
692         parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/false);
693         addUnwrappedLine();
694         return;
695       }
696     }
697     break;
698   case tok::identifier:
699     if (FormatTok->IsForEachMacro) {
700       parseForOrWhileLoop();
701       return;
702     }
703     // In all other cases, parse the declaration.
704     break;
705   default:
706     break;
707   }
708   do {
709     switch (FormatTok->Tok.getKind()) {
710     case tok::at:
711       nextToken();
712       if (FormatTok->Tok.is(tok::l_brace))
713         parseBracedList();
714       break;
715     case tok::kw_enum:
716       parseEnum();
717       break;
718     case tok::kw_typedef:
719       nextToken();
720       // FIXME: Use the IdentifierTable instead.
721       if (FormatTok->TokenText == "NS_ENUM")
722         parseEnum();
723       break;
724     case tok::kw_struct:
725     case tok::kw_union:
726     case tok::kw_class:
727       parseRecord();
728       // A record declaration or definition is always the start of a structural
729       // element.
730       break;
731     case tok::semi:
732       nextToken();
733       addUnwrappedLine();
734       return;
735     case tok::r_brace:
736       addUnwrappedLine();
737       return;
738     case tok::l_paren:
739       parseParens();
740       break;
741     case tok::caret:
742       nextToken();
743       if (FormatTok->Tok.isAnyIdentifier() ||
744           FormatTok->isSimpleTypeSpecifier())
745         nextToken();
746       if (FormatTok->is(tok::l_paren))
747         parseParens();
748       if (FormatTok->is(tok::l_brace))
749         parseChildBlock();
750       break;
751     case tok::l_brace:
752       if (!tryToParseBracedList()) {
753         // A block outside of parentheses must be the last part of a
754         // structural element.
755         // FIXME: Figure out cases where this is not true, and add projections
756         // for them (the one we know is missing are lambdas).
757         if (Style.BreakBeforeBraces != FormatStyle::BS_Attach)
758           addUnwrappedLine();
759         FormatTok->Type = TT_FunctionLBrace;
760         parseBlock(/*MustBeDeclaration=*/false);
761         addUnwrappedLine();
762         return;
763       }
764       // Otherwise this was a braced init list, and the structural
765       // element continues.
766       break;
767     case tok::kw_try:
768       // We arrive here when parsing function-try blocks.
769       parseTryCatch();
770       return;
771     case tok::identifier: {
772       StringRef Text = FormatTok->TokenText;
773       // Parse function literal unless 'function' is the first token in a line
774       // in which case this should be treated as a free-standing function.
775       if (Style.Language == FormatStyle::LK_JavaScript && Text == "function" &&
776           Line->Tokens.size() > 0) {
777         tryToParseJSFunction();
778         break;
779       }
780       nextToken();
781       if (Line->Tokens.size() == 1) {
782         if (FormatTok->Tok.is(tok::colon)) {
783           parseLabel();
784           return;
785         }
786         // Recognize function-like macro usages without trailing semicolon.
787         if (FormatTok->Tok.is(tok::l_paren)) {
788           parseParens();
789           if (FormatTok->NewlinesBefore > 0 &&
790               tokenCanStartNewLine(FormatTok->Tok) && Text == Text.upper()) {
791             addUnwrappedLine();
792             return;
793           }
794         } else if (FormatTok->HasUnescapedNewline && Text.size() >= 5 &&
795                    Text == Text.upper()) {
796           // Recognize free-standing macros like Q_OBJECT.
797           addUnwrappedLine();
798           return;
799         }
800       }
801       break;
802     }
803     case tok::equal:
804       nextToken();
805       if (FormatTok->Tok.is(tok::l_brace)) {
806         parseBracedList();
807       }
808       break;
809     case tok::l_square:
810       parseSquare();
811       break;
812     default:
813       nextToken();
814       break;
815     }
816   } while (!eof());
817 }
818 
tryToParseLambda()819 bool UnwrappedLineParser::tryToParseLambda() {
820   // FIXME: This is a dirty way to access the previous token. Find a better
821   // solution.
822   if (!Line->Tokens.empty() &&
823       (Line->Tokens.back().Tok->isOneOf(tok::identifier, tok::kw_operator) ||
824        Line->Tokens.back().Tok->closesScope() ||
825        Line->Tokens.back().Tok->isSimpleTypeSpecifier())) {
826     nextToken();
827     return false;
828   }
829   assert(FormatTok->is(tok::l_square));
830   FormatToken &LSquare = *FormatTok;
831   if (!tryToParseLambdaIntroducer())
832     return false;
833 
834   while (FormatTok->isNot(tok::l_brace)) {
835     if (FormatTok->isSimpleTypeSpecifier()) {
836       nextToken();
837       continue;
838     }
839     switch (FormatTok->Tok.getKind()) {
840     case tok::l_brace:
841       break;
842     case tok::l_paren:
843       parseParens();
844       break;
845     case tok::less:
846     case tok::greater:
847     case tok::identifier:
848     case tok::coloncolon:
849     case tok::kw_mutable:
850       nextToken();
851       break;
852     case tok::arrow:
853       FormatTok->Type = TT_TrailingReturnArrow;
854       nextToken();
855       break;
856     default:
857       return true;
858     }
859   }
860   LSquare.Type = TT_LambdaLSquare;
861   parseChildBlock();
862   return true;
863 }
864 
tryToParseLambdaIntroducer()865 bool UnwrappedLineParser::tryToParseLambdaIntroducer() {
866   nextToken();
867   if (FormatTok->is(tok::equal)) {
868     nextToken();
869     if (FormatTok->is(tok::r_square)) {
870       nextToken();
871       return true;
872     }
873     if (FormatTok->isNot(tok::comma))
874       return false;
875     nextToken();
876   } else if (FormatTok->is(tok::amp)) {
877     nextToken();
878     if (FormatTok->is(tok::r_square)) {
879       nextToken();
880       return true;
881     }
882     if (!FormatTok->isOneOf(tok::comma, tok::identifier)) {
883       return false;
884     }
885     if (FormatTok->is(tok::comma))
886       nextToken();
887   } else if (FormatTok->is(tok::r_square)) {
888     nextToken();
889     return true;
890   }
891   do {
892     if (FormatTok->is(tok::amp))
893       nextToken();
894     if (!FormatTok->isOneOf(tok::identifier, tok::kw_this))
895       return false;
896     nextToken();
897     if (FormatTok->is(tok::ellipsis))
898       nextToken();
899     if (FormatTok->is(tok::comma)) {
900       nextToken();
901     } else if (FormatTok->is(tok::r_square)) {
902       nextToken();
903       return true;
904     } else {
905       return false;
906     }
907   } while (!eof());
908   return false;
909 }
910 
tryToParseJSFunction()911 void UnwrappedLineParser::tryToParseJSFunction() {
912   nextToken();
913 
914   // Consume function name.
915   if (FormatTok->is(tok::identifier))
916       nextToken();
917 
918   if (FormatTok->isNot(tok::l_paren))
919     return;
920   nextToken();
921   while (FormatTok->isNot(tok::l_brace)) {
922     // Err on the side of caution in order to avoid consuming the full file in
923     // case of incomplete code.
924     if (!FormatTok->isOneOf(tok::identifier, tok::comma, tok::r_paren,
925                             tok::comment))
926       return;
927     nextToken();
928   }
929   parseChildBlock();
930 }
931 
tryToParseBracedList()932 bool UnwrappedLineParser::tryToParseBracedList() {
933   if (FormatTok->BlockKind == BK_Unknown)
934     calculateBraceTypes();
935   assert(FormatTok->BlockKind != BK_Unknown);
936   if (FormatTok->BlockKind == BK_Block)
937     return false;
938   parseBracedList();
939   return true;
940 }
941 
parseBracedList(bool ContinueOnSemicolons)942 bool UnwrappedLineParser::parseBracedList(bool ContinueOnSemicolons) {
943   bool HasError = false;
944   nextToken();
945 
946   // FIXME: Once we have an expression parser in the UnwrappedLineParser,
947   // replace this by using parseAssigmentExpression() inside.
948   do {
949     if (Style.Language == FormatStyle::LK_JavaScript &&
950         FormatTok->TokenText == "function") {
951       tryToParseJSFunction();
952       continue;
953     }
954     switch (FormatTok->Tok.getKind()) {
955     case tok::caret:
956       nextToken();
957       if (FormatTok->is(tok::l_brace)) {
958         parseChildBlock();
959       }
960       break;
961     case tok::l_square:
962       tryToParseLambda();
963       break;
964     case tok::l_brace:
965       // Assume there are no blocks inside a braced init list apart
966       // from the ones we explicitly parse out (like lambdas).
967       FormatTok->BlockKind = BK_BracedInit;
968       parseBracedList();
969       break;
970     case tok::r_brace:
971       nextToken();
972       return !HasError;
973     case tok::semi:
974       HasError = true;
975       if (!ContinueOnSemicolons)
976         return !HasError;
977       nextToken();
978       break;
979     case tok::comma:
980       nextToken();
981       break;
982     default:
983       nextToken();
984       break;
985     }
986   } while (!eof());
987   return false;
988 }
989 
parseParens()990 void UnwrappedLineParser::parseParens() {
991   assert(FormatTok->Tok.is(tok::l_paren) && "'(' expected.");
992   nextToken();
993   do {
994     switch (FormatTok->Tok.getKind()) {
995     case tok::l_paren:
996       parseParens();
997       break;
998     case tok::r_paren:
999       nextToken();
1000       return;
1001     case tok::r_brace:
1002       // A "}" inside parenthesis is an error if there wasn't a matching "{".
1003       return;
1004     case tok::l_square:
1005       tryToParseLambda();
1006       break;
1007     case tok::l_brace: {
1008       if (!tryToParseBracedList()) {
1009         parseChildBlock();
1010       }
1011       break;
1012     }
1013     case tok::at:
1014       nextToken();
1015       if (FormatTok->Tok.is(tok::l_brace))
1016         parseBracedList();
1017       break;
1018     default:
1019       nextToken();
1020       break;
1021     }
1022   } while (!eof());
1023 }
1024 
parseSquare()1025 void UnwrappedLineParser::parseSquare() {
1026   assert(FormatTok->Tok.is(tok::l_square) && "'[' expected.");
1027   if (tryToParseLambda())
1028     return;
1029   do {
1030     switch (FormatTok->Tok.getKind()) {
1031     case tok::l_paren:
1032       parseParens();
1033       break;
1034     case tok::r_square:
1035       nextToken();
1036       return;
1037     case tok::r_brace:
1038       // A "}" inside parenthesis is an error if there wasn't a matching "{".
1039       return;
1040     case tok::l_square:
1041       parseSquare();
1042       break;
1043     case tok::l_brace: {
1044       if (!tryToParseBracedList()) {
1045         parseChildBlock();
1046       }
1047       break;
1048     }
1049     case tok::at:
1050       nextToken();
1051       if (FormatTok->Tok.is(tok::l_brace))
1052         parseBracedList();
1053       break;
1054     default:
1055       nextToken();
1056       break;
1057     }
1058   } while (!eof());
1059 }
1060 
parseIfThenElse()1061 void UnwrappedLineParser::parseIfThenElse() {
1062   assert(FormatTok->Tok.is(tok::kw_if) && "'if' expected");
1063   nextToken();
1064   if (FormatTok->Tok.is(tok::l_paren))
1065     parseParens();
1066   bool NeedsUnwrappedLine = false;
1067   if (FormatTok->Tok.is(tok::l_brace)) {
1068     CompoundStatementIndenter Indenter(this, Style, Line->Level);
1069     parseBlock(/*MustBeDeclaration=*/false);
1070     if (Style.BreakBeforeBraces == FormatStyle::BS_Allman ||
1071         Style.BreakBeforeBraces == FormatStyle::BS_GNU) {
1072       addUnwrappedLine();
1073     } else {
1074       NeedsUnwrappedLine = true;
1075     }
1076   } else {
1077     addUnwrappedLine();
1078     ++Line->Level;
1079     parseStructuralElement();
1080     --Line->Level;
1081   }
1082   if (FormatTok->Tok.is(tok::kw_else)) {
1083     nextToken();
1084     if (FormatTok->Tok.is(tok::l_brace)) {
1085       CompoundStatementIndenter Indenter(this, Style, Line->Level);
1086       parseBlock(/*MustBeDeclaration=*/false);
1087       addUnwrappedLine();
1088     } else if (FormatTok->Tok.is(tok::kw_if)) {
1089       parseIfThenElse();
1090     } else {
1091       addUnwrappedLine();
1092       ++Line->Level;
1093       parseStructuralElement();
1094       --Line->Level;
1095     }
1096   } else if (NeedsUnwrappedLine) {
1097     addUnwrappedLine();
1098   }
1099 }
1100 
parseTryCatch()1101 void UnwrappedLineParser::parseTryCatch() {
1102   assert(FormatTok->is(tok::kw_try) && "'try' expected");
1103   nextToken();
1104   bool NeedsUnwrappedLine = false;
1105   if (FormatTok->is(tok::colon)) {
1106     // We are in a function try block, what comes is an initializer list.
1107     nextToken();
1108     while (FormatTok->is(tok::identifier)) {
1109       nextToken();
1110       if (FormatTok->is(tok::l_paren))
1111         parseParens();
1112       else
1113         StructuralError = true;
1114       if (FormatTok->is(tok::comma))
1115         nextToken();
1116     }
1117   }
1118   if (FormatTok->is(tok::l_brace)) {
1119     CompoundStatementIndenter Indenter(this, Style, Line->Level);
1120     parseBlock(/*MustBeDeclaration=*/false);
1121     if (Style.BreakBeforeBraces == FormatStyle::BS_Allman ||
1122         Style.BreakBeforeBraces == FormatStyle::BS_GNU ||
1123         Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) {
1124       addUnwrappedLine();
1125     } else {
1126       NeedsUnwrappedLine = true;
1127     }
1128   } else if (!FormatTok->is(tok::kw_catch)) {
1129     // The C++ standard requires a compound-statement after a try.
1130     // If there's none, we try to assume there's a structuralElement
1131     // and try to continue.
1132     StructuralError = true;
1133     addUnwrappedLine();
1134     ++Line->Level;
1135     parseStructuralElement();
1136     --Line->Level;
1137   }
1138   while (FormatTok->is(tok::kw_catch) ||
1139          (Style.Language == FormatStyle::LK_JavaScript &&
1140           FormatTok->TokenText == "finally")) {
1141     nextToken();
1142     while (FormatTok->isNot(tok::l_brace)) {
1143       if (FormatTok->is(tok::l_paren)) {
1144         parseParens();
1145         continue;
1146       }
1147       if (FormatTok->isOneOf(tok::semi, tok::r_brace))
1148         return;
1149       nextToken();
1150     }
1151     NeedsUnwrappedLine = false;
1152     CompoundStatementIndenter Indenter(this, Style, Line->Level);
1153     parseBlock(/*MustBeDeclaration=*/false);
1154     if (Style.BreakBeforeBraces == FormatStyle::BS_Allman ||
1155         Style.BreakBeforeBraces == FormatStyle::BS_GNU ||
1156         Style.BreakBeforeBraces == FormatStyle::BS_Stroustrup) {
1157       addUnwrappedLine();
1158     } else {
1159       NeedsUnwrappedLine = true;
1160     }
1161   }
1162   if (NeedsUnwrappedLine) {
1163     addUnwrappedLine();
1164   }
1165 }
1166 
parseNamespace()1167 void UnwrappedLineParser::parseNamespace() {
1168   assert(FormatTok->Tok.is(tok::kw_namespace) && "'namespace' expected");
1169   nextToken();
1170   if (FormatTok->Tok.is(tok::identifier))
1171     nextToken();
1172   if (FormatTok->Tok.is(tok::l_brace)) {
1173     if (Style.BreakBeforeBraces == FormatStyle::BS_Linux ||
1174         Style.BreakBeforeBraces == FormatStyle::BS_Allman ||
1175         Style.BreakBeforeBraces == FormatStyle::BS_GNU)
1176       addUnwrappedLine();
1177 
1178     bool AddLevel = Style.NamespaceIndentation == FormatStyle::NI_All ||
1179                     (Style.NamespaceIndentation == FormatStyle::NI_Inner &&
1180                      DeclarationScopeStack.size() > 1);
1181     parseBlock(/*MustBeDeclaration=*/true, AddLevel);
1182     // Munch the semicolon after a namespace. This is more common than one would
1183     // think. Puttin the semicolon into its own line is very ugly.
1184     if (FormatTok->Tok.is(tok::semi))
1185       nextToken();
1186     addUnwrappedLine();
1187   }
1188   // FIXME: Add error handling.
1189 }
1190 
parseForOrWhileLoop()1191 void UnwrappedLineParser::parseForOrWhileLoop() {
1192   assert((FormatTok->Tok.is(tok::kw_for) || FormatTok->Tok.is(tok::kw_while) ||
1193           FormatTok->IsForEachMacro) &&
1194          "'for', 'while' or foreach macro expected");
1195   nextToken();
1196   if (FormatTok->Tok.is(tok::l_paren))
1197     parseParens();
1198   if (FormatTok->Tok.is(tok::l_brace)) {
1199     CompoundStatementIndenter Indenter(this, Style, Line->Level);
1200     parseBlock(/*MustBeDeclaration=*/false);
1201     addUnwrappedLine();
1202   } else {
1203     addUnwrappedLine();
1204     ++Line->Level;
1205     parseStructuralElement();
1206     --Line->Level;
1207   }
1208 }
1209 
parseDoWhile()1210 void UnwrappedLineParser::parseDoWhile() {
1211   assert(FormatTok->Tok.is(tok::kw_do) && "'do' expected");
1212   nextToken();
1213   if (FormatTok->Tok.is(tok::l_brace)) {
1214     CompoundStatementIndenter Indenter(this, Style, Line->Level);
1215     parseBlock(/*MustBeDeclaration=*/false);
1216     if (Style.BreakBeforeBraces == FormatStyle::BS_GNU)
1217       addUnwrappedLine();
1218   } else {
1219     addUnwrappedLine();
1220     ++Line->Level;
1221     parseStructuralElement();
1222     --Line->Level;
1223   }
1224 
1225   // FIXME: Add error handling.
1226   if (!FormatTok->Tok.is(tok::kw_while)) {
1227     addUnwrappedLine();
1228     return;
1229   }
1230 
1231   nextToken();
1232   parseStructuralElement();
1233 }
1234 
parseLabel()1235 void UnwrappedLineParser::parseLabel() {
1236   nextToken();
1237   unsigned OldLineLevel = Line->Level;
1238   if (Line->Level > 1 || (!Line->InPPDirective && Line->Level > 0))
1239     --Line->Level;
1240   if (CommentsBeforeNextToken.empty() && FormatTok->Tok.is(tok::l_brace)) {
1241     CompoundStatementIndenter Indenter(this, Style, Line->Level);
1242     parseBlock(/*MustBeDeclaration=*/false);
1243     if (FormatTok->Tok.is(tok::kw_break)) {
1244       // "break;" after "}" on its own line only for BS_Allman and BS_GNU
1245       if (Style.BreakBeforeBraces == FormatStyle::BS_Allman ||
1246           Style.BreakBeforeBraces == FormatStyle::BS_GNU) {
1247         addUnwrappedLine();
1248       }
1249       parseStructuralElement();
1250     }
1251     addUnwrappedLine();
1252   } else {
1253     addUnwrappedLine();
1254   }
1255   Line->Level = OldLineLevel;
1256 }
1257 
parseCaseLabel()1258 void UnwrappedLineParser::parseCaseLabel() {
1259   assert(FormatTok->Tok.is(tok::kw_case) && "'case' expected");
1260   // FIXME: fix handling of complex expressions here.
1261   do {
1262     nextToken();
1263   } while (!eof() && !FormatTok->Tok.is(tok::colon));
1264   parseLabel();
1265 }
1266 
parseSwitch()1267 void UnwrappedLineParser::parseSwitch() {
1268   assert(FormatTok->Tok.is(tok::kw_switch) && "'switch' expected");
1269   nextToken();
1270   if (FormatTok->Tok.is(tok::l_paren))
1271     parseParens();
1272   if (FormatTok->Tok.is(tok::l_brace)) {
1273     CompoundStatementIndenter Indenter(this, Style, Line->Level);
1274     parseBlock(/*MustBeDeclaration=*/false);
1275     addUnwrappedLine();
1276   } else {
1277     addUnwrappedLine();
1278     ++Line->Level;
1279     parseStructuralElement();
1280     --Line->Level;
1281   }
1282 }
1283 
parseAccessSpecifier()1284 void UnwrappedLineParser::parseAccessSpecifier() {
1285   nextToken();
1286   // Understand Qt's slots.
1287   if (FormatTok->is(tok::identifier) &&
1288       (FormatTok->TokenText == "slots" || FormatTok->TokenText == "Q_SLOTS"))
1289     nextToken();
1290   // Otherwise, we don't know what it is, and we'd better keep the next token.
1291   if (FormatTok->Tok.is(tok::colon))
1292     nextToken();
1293   addUnwrappedLine();
1294 }
1295 
parseEnum()1296 void UnwrappedLineParser::parseEnum() {
1297   if (FormatTok->Tok.is(tok::kw_enum)) {
1298     // Won't be 'enum' for NS_ENUMs.
1299     nextToken();
1300   }
1301   // Eat up enum class ...
1302   if (FormatTok->Tok.is(tok::kw_class) || FormatTok->Tok.is(tok::kw_struct))
1303     nextToken();
1304   while (FormatTok->Tok.getIdentifierInfo() ||
1305          FormatTok->isOneOf(tok::colon, tok::coloncolon)) {
1306     nextToken();
1307     // We can have macros or attributes in between 'enum' and the enum name.
1308     if (FormatTok->Tok.is(tok::l_paren)) {
1309       parseParens();
1310     }
1311     if (FormatTok->Tok.is(tok::identifier))
1312       nextToken();
1313   }
1314   if (FormatTok->Tok.is(tok::l_brace)) {
1315     FormatTok->BlockKind = BK_Block;
1316     bool HasError = !parseBracedList(/*ContinueOnSemicolons=*/true);
1317     if (HasError) {
1318       if (FormatTok->is(tok::semi))
1319         nextToken();
1320       addUnwrappedLine();
1321     }
1322   }
1323   // We fall through to parsing a structural element afterwards, so that in
1324   // enum A {} n, m;
1325   // "} n, m;" will end up in one unwrapped line.
1326 }
1327 
parseRecord()1328 void UnwrappedLineParser::parseRecord() {
1329   nextToken();
1330   if (FormatTok->Tok.is(tok::identifier) ||
1331       FormatTok->Tok.is(tok::kw___attribute) ||
1332       FormatTok->Tok.is(tok::kw___declspec) ||
1333       FormatTok->Tok.is(tok::kw_alignas)) {
1334     nextToken();
1335     // We can have macros or attributes in between 'class' and the class name.
1336     if (FormatTok->Tok.is(tok::l_paren)) {
1337       parseParens();
1338     }
1339     // The actual identifier can be a nested name specifier, and in macros
1340     // it is often token-pasted.
1341     while (FormatTok->Tok.is(tok::identifier) ||
1342            FormatTok->Tok.is(tok::coloncolon) ||
1343            FormatTok->Tok.is(tok::hashhash))
1344       nextToken();
1345 
1346     // Note that parsing away template declarations here leads to incorrectly
1347     // accepting function declarations as record declarations.
1348     // In general, we cannot solve this problem. Consider:
1349     // class A<int> B() {}
1350     // which can be a function definition or a class definition when B() is a
1351     // macro. If we find enough real-world cases where this is a problem, we
1352     // can parse for the 'template' keyword in the beginning of the statement,
1353     // and thus rule out the record production in case there is no template
1354     // (this would still leave us with an ambiguity between template function
1355     // and class declarations).
1356     if (FormatTok->Tok.is(tok::colon) || FormatTok->Tok.is(tok::less)) {
1357       while (!eof() && FormatTok->Tok.isNot(tok::l_brace)) {
1358         if (FormatTok->Tok.is(tok::semi))
1359           return;
1360         nextToken();
1361       }
1362     }
1363   }
1364   if (FormatTok->Tok.is(tok::l_brace)) {
1365     if (Style.BreakBeforeBraces == FormatStyle::BS_Linux ||
1366         Style.BreakBeforeBraces == FormatStyle::BS_Allman ||
1367         Style.BreakBeforeBraces == FormatStyle::BS_GNU)
1368       addUnwrappedLine();
1369 
1370     parseBlock(/*MustBeDeclaration=*/true, /*AddLevel=*/true,
1371                /*MunchSemi=*/false);
1372   }
1373   // We fall through to parsing a structural element afterwards, so
1374   // class A {} n, m;
1375   // will end up in one unwrapped line.
1376 }
1377 
parseObjCProtocolList()1378 void UnwrappedLineParser::parseObjCProtocolList() {
1379   assert(FormatTok->Tok.is(tok::less) && "'<' expected.");
1380   do
1381     nextToken();
1382   while (!eof() && FormatTok->Tok.isNot(tok::greater));
1383   nextToken(); // Skip '>'.
1384 }
1385 
parseObjCUntilAtEnd()1386 void UnwrappedLineParser::parseObjCUntilAtEnd() {
1387   do {
1388     if (FormatTok->Tok.isObjCAtKeyword(tok::objc_end)) {
1389       nextToken();
1390       addUnwrappedLine();
1391       break;
1392     }
1393     if (FormatTok->is(tok::l_brace)) {
1394       parseBlock(/*MustBeDeclaration=*/false);
1395       // In ObjC interfaces, nothing should be following the "}".
1396       addUnwrappedLine();
1397     } else if (FormatTok->is(tok::r_brace)) {
1398       // Ignore stray "}". parseStructuralElement doesn't consume them.
1399       nextToken();
1400       addUnwrappedLine();
1401     } else {
1402       parseStructuralElement();
1403     }
1404   } while (!eof());
1405 }
1406 
parseObjCInterfaceOrImplementation()1407 void UnwrappedLineParser::parseObjCInterfaceOrImplementation() {
1408   nextToken();
1409   nextToken(); // interface name
1410 
1411   // @interface can be followed by either a base class, or a category.
1412   if (FormatTok->Tok.is(tok::colon)) {
1413     nextToken();
1414     nextToken(); // base class name
1415   } else if (FormatTok->Tok.is(tok::l_paren))
1416     // Skip category, if present.
1417     parseParens();
1418 
1419   if (FormatTok->Tok.is(tok::less))
1420     parseObjCProtocolList();
1421 
1422   if (FormatTok->Tok.is(tok::l_brace)) {
1423     if (Style.BreakBeforeBraces == FormatStyle::BS_Allman ||
1424         Style.BreakBeforeBraces == FormatStyle::BS_GNU)
1425       addUnwrappedLine();
1426     parseBlock(/*MustBeDeclaration=*/true);
1427   }
1428 
1429   // With instance variables, this puts '}' on its own line.  Without instance
1430   // variables, this ends the @interface line.
1431   addUnwrappedLine();
1432 
1433   parseObjCUntilAtEnd();
1434 }
1435 
parseObjCProtocol()1436 void UnwrappedLineParser::parseObjCProtocol() {
1437   nextToken();
1438   nextToken(); // protocol name
1439 
1440   if (FormatTok->Tok.is(tok::less))
1441     parseObjCProtocolList();
1442 
1443   // Check for protocol declaration.
1444   if (FormatTok->Tok.is(tok::semi)) {
1445     nextToken();
1446     return addUnwrappedLine();
1447   }
1448 
1449   addUnwrappedLine();
1450   parseObjCUntilAtEnd();
1451 }
1452 
printDebugInfo(const UnwrappedLine & Line,StringRef Prefix="")1453 LLVM_ATTRIBUTE_UNUSED static void printDebugInfo(const UnwrappedLine &Line,
1454                                                  StringRef Prefix = "") {
1455   llvm::dbgs() << Prefix << "Line(" << Line.Level << ")"
1456                << (Line.InPPDirective ? " MACRO" : "") << ": ";
1457   for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(),
1458                                                     E = Line.Tokens.end();
1459        I != E; ++I) {
1460     llvm::dbgs() << I->Tok->Tok.getName() << "[" << I->Tok->Type << "] ";
1461   }
1462   for (std::list<UnwrappedLineNode>::const_iterator I = Line.Tokens.begin(),
1463                                                     E = Line.Tokens.end();
1464        I != E; ++I) {
1465     const UnwrappedLineNode &Node = *I;
1466     for (SmallVectorImpl<UnwrappedLine>::const_iterator
1467              I = Node.Children.begin(),
1468              E = Node.Children.end();
1469          I != E; ++I) {
1470       printDebugInfo(*I, "\nChild: ");
1471     }
1472   }
1473   llvm::dbgs() << "\n";
1474 }
1475 
addUnwrappedLine()1476 void UnwrappedLineParser::addUnwrappedLine() {
1477   if (Line->Tokens.empty())
1478     return;
1479   DEBUG({
1480     if (CurrentLines == &Lines)
1481       printDebugInfo(*Line);
1482   });
1483   CurrentLines->push_back(*Line);
1484   Line->Tokens.clear();
1485   if (CurrentLines == &Lines && !PreprocessorDirectives.empty()) {
1486     for (SmallVectorImpl<UnwrappedLine>::iterator
1487              I = PreprocessorDirectives.begin(),
1488              E = PreprocessorDirectives.end();
1489          I != E; ++I) {
1490       CurrentLines->push_back(*I);
1491     }
1492     PreprocessorDirectives.clear();
1493   }
1494 }
1495 
eof() const1496 bool UnwrappedLineParser::eof() const { return FormatTok->Tok.is(tok::eof); }
1497 
isOnNewLine(const FormatToken & FormatTok)1498 bool UnwrappedLineParser::isOnNewLine(const FormatToken &FormatTok) {
1499   return (Line->InPPDirective || FormatTok.HasUnescapedNewline) &&
1500          FormatTok.NewlinesBefore > 0;
1501 }
1502 
flushComments(bool NewlineBeforeNext)1503 void UnwrappedLineParser::flushComments(bool NewlineBeforeNext) {
1504   bool JustComments = Line->Tokens.empty();
1505   for (SmallVectorImpl<FormatToken *>::const_iterator
1506            I = CommentsBeforeNextToken.begin(),
1507            E = CommentsBeforeNextToken.end();
1508        I != E; ++I) {
1509     if (isOnNewLine(**I) && JustComments) {
1510       addUnwrappedLine();
1511     }
1512     pushToken(*I);
1513   }
1514   if (NewlineBeforeNext && JustComments) {
1515     addUnwrappedLine();
1516   }
1517   CommentsBeforeNextToken.clear();
1518 }
1519 
nextToken()1520 void UnwrappedLineParser::nextToken() {
1521   if (eof())
1522     return;
1523   flushComments(isOnNewLine(*FormatTok));
1524   pushToken(FormatTok);
1525   readToken();
1526 }
1527 
readToken()1528 void UnwrappedLineParser::readToken() {
1529   bool CommentsInCurrentLine = true;
1530   do {
1531     FormatTok = Tokens->getNextToken();
1532     assert(FormatTok);
1533     while (!Line->InPPDirective && FormatTok->Tok.is(tok::hash) &&
1534            (FormatTok->HasUnescapedNewline || FormatTok->IsFirst)) {
1535       // If there is an unfinished unwrapped line, we flush the preprocessor
1536       // directives only after that unwrapped line was finished later.
1537       bool SwitchToPreprocessorLines =
1538           !Line->Tokens.empty() && CurrentLines == &Lines;
1539       ScopedLineState BlockState(*this, SwitchToPreprocessorLines);
1540       // Comments stored before the preprocessor directive need to be output
1541       // before the preprocessor directive, at the same level as the
1542       // preprocessor directive, as we consider them to apply to the directive.
1543       flushComments(isOnNewLine(*FormatTok));
1544       parsePPDirective();
1545     }
1546     while (FormatTok->Type == TT_ConflictStart ||
1547            FormatTok->Type == TT_ConflictEnd ||
1548            FormatTok->Type == TT_ConflictAlternative) {
1549       if (FormatTok->Type == TT_ConflictStart) {
1550         conditionalCompilationStart(/*Unreachable=*/false);
1551       } else if (FormatTok->Type == TT_ConflictAlternative) {
1552         conditionalCompilationAlternative();
1553       } else if (FormatTok->Type == TT_ConflictEnd) {
1554         conditionalCompilationEnd();
1555       }
1556       FormatTok = Tokens->getNextToken();
1557       FormatTok->MustBreakBefore = true;
1558     }
1559 
1560     if (!PPStack.empty() && (PPStack.back() == PP_Unreachable) &&
1561         !Line->InPPDirective) {
1562       continue;
1563     }
1564 
1565     if (!FormatTok->Tok.is(tok::comment))
1566       return;
1567     if (isOnNewLine(*FormatTok) || FormatTok->IsFirst) {
1568       CommentsInCurrentLine = false;
1569     }
1570     if (CommentsInCurrentLine) {
1571       pushToken(FormatTok);
1572     } else {
1573       CommentsBeforeNextToken.push_back(FormatTok);
1574     }
1575   } while (!eof());
1576 }
1577 
pushToken(FormatToken * Tok)1578 void UnwrappedLineParser::pushToken(FormatToken *Tok) {
1579   Line->Tokens.push_back(UnwrappedLineNode(Tok));
1580   if (MustBreakBeforeNextToken) {
1581     Line->Tokens.back().Tok->MustBreakBefore = true;
1582     MustBreakBeforeNextToken = false;
1583   }
1584 }
1585 
1586 } // end namespace format
1587 } // end namespace clang
1588