1 //===--- TokenAnnotator.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 implements a token annotator, i.e. creates
12 /// \c AnnotatedTokens out of \c FormatTokens with required extra information.
14 //===----------------------------------------------------------------------===//
16 #include "TokenAnnotator.h"
17 #include "clang/Basic/SourceManager.h"
18 #include "llvm/ADT/SmallPtrSet.h"
19 #include "llvm/Support/Debug.h"
21 #define DEBUG_TYPE "format-token-annotator"
28 /// \brief A parser that gathers additional information about tokens.
30 /// The \c TokenAnnotator tries to match parenthesis and square brakets and
31 /// store a parenthesis levels. It also tries to resolve matching "<" and ">"
32 /// into template parameter lists.
33 class AnnotatingParser {
35 AnnotatingParser(const FormatStyle &Style, AnnotatedLine &Line,
36 const AdditionalKeywords &Keywords)
37 : Style(Style), Line(Line), CurrentToken(Line.First), AutoFound(false),
39 Contexts.push_back(Context(tok::unknown, 1, /*IsExpression=*/false));
40 resetTokenMetadata(CurrentToken);
45 if (!CurrentToken || !CurrentToken->Previous)
47 if (NonTemplateLess.count(CurrentToken->Previous))
50 const FormatToken& Previous = *CurrentToken->Previous;
51 if (Previous.Previous) {
52 if (Previous.Previous->Tok.isLiteral())
54 if (Previous.Previous->is(tok::r_paren) && Contexts.size() > 1 &&
55 (!Previous.Previous->MatchingParen ||
56 !Previous.Previous->MatchingParen->is(TT_OverloadedOperatorLParen)))
60 FormatToken *Left = CurrentToken->Previous;
61 Left->ParentBracket = Contexts.back().ContextKind;
62 ScopedContextCreator ContextCreator(*this, tok::less, 12);
64 // If this angle is in the context of an expression, we need to be more
65 // hesitant to detect it as opening template parameters.
66 bool InExprContext = Contexts.back().IsExpression;
68 Contexts.back().IsExpression = false;
69 // If there's a template keyword before the opening angle bracket, this is a
70 // template parameter, not an argument.
71 Contexts.back().InTemplateArgument =
72 Left->Previous && Left->Previous->Tok.isNot(tok::kw_template);
74 if (Style.Language == FormatStyle::LK_Java &&
75 CurrentToken->is(tok::question))
78 while (CurrentToken) {
79 if (CurrentToken->is(tok::greater)) {
80 Left->MatchingParen = CurrentToken;
81 CurrentToken->MatchingParen = Left;
82 CurrentToken->Type = TT_TemplateCloser;
86 if (CurrentToken->is(tok::question) &&
87 Style.Language == FormatStyle::LK_Java) {
91 if (CurrentToken->isOneOf(tok::r_paren, tok::r_square, tok::r_brace) ||
92 (CurrentToken->isOneOf(tok::colon, tok::question) && InExprContext))
94 // If a && or || is found and interpreted as a binary operator, this set
95 // of angles is likely part of something like "a < b && c > d". If the
96 // angles are inside an expression, the ||/&& might also be a binary
97 // operator that was misinterpreted because we are parsing template
99 // FIXME: This is getting out of hand, write a decent parser.
100 if (CurrentToken->Previous->isOneOf(tok::pipepipe, tok::ampamp) &&
101 CurrentToken->Previous->is(TT_BinaryOperator) &&
102 Contexts[Contexts.size() - 2].IsExpression &&
103 !Line.startsWith(tok::kw_template))
105 updateParameterCount(Left, CurrentToken);
112 bool parseParens(bool LookForDecls = false) {
115 FormatToken *Left = CurrentToken->Previous;
116 Left->ParentBracket = Contexts.back().ContextKind;
117 ScopedContextCreator ContextCreator(*this, tok::l_paren, 1);
119 // FIXME: This is a bit of a hack. Do better.
120 Contexts.back().ColonIsForRangeExpr =
121 Contexts.size() == 2 && Contexts[0].ColonIsForRangeExpr;
123 bool StartsObjCMethodExpr = false;
124 if (CurrentToken->is(tok::caret)) {
125 // (^ can start a block type.
126 Left->Type = TT_ObjCBlockLParen;
127 } else if (FormatToken *MaybeSel = Left->Previous) {
128 // @selector( starts a selector.
129 if (MaybeSel->isObjCAtKeyword(tok::objc_selector) && MaybeSel->Previous &&
130 MaybeSel->Previous->is(tok::at)) {
131 StartsObjCMethodExpr = true;
135 if (Left->is(TT_OverloadedOperatorLParen)) {
136 Contexts.back().IsExpression = false;
137 } else if (Style.Language == FormatStyle::LK_JavaScript &&
138 Line.startsWith(Keywords.kw_type, tok::identifier)) {
140 Contexts.back().IsExpression = false;
141 } else if (Left->Previous &&
142 (Left->Previous->isOneOf(tok::kw_static_assert, tok::kw_decltype,
143 tok::kw_if, tok::kw_while, tok::l_paren,
145 Left->Previous->is(TT_BinaryOperator))) {
146 // static_assert, if and while usually contain expressions.
147 Contexts.back().IsExpression = true;
148 } else if (Style.Language == FormatStyle::LK_JavaScript && Left->Previous &&
149 (Left->Previous->is(Keywords.kw_function) ||
150 (Left->Previous->endsSequence(tok::identifier,
151 Keywords.kw_function)))) {
152 // function(...) or function f(...)
153 Contexts.back().IsExpression = false;
154 } else if (Style.Language == FormatStyle::LK_JavaScript && Left->Previous &&
155 Left->Previous->is(TT_JsTypeColon)) {
156 // let x: (SomeType);
157 Contexts.back().IsExpression = false;
158 } else if (Left->Previous && Left->Previous->is(tok::r_square) &&
159 Left->Previous->MatchingParen &&
160 Left->Previous->MatchingParen->is(TT_LambdaLSquare)) {
161 // This is a parameter list of a lambda expression.
162 Contexts.back().IsExpression = false;
163 } else if (Line.InPPDirective &&
164 (!Left->Previous || !Left->Previous->is(tok::identifier))) {
165 Contexts.back().IsExpression = true;
166 } else if (Contexts[Contexts.size() - 2].CaretFound) {
167 // This is the parameter list of an ObjC block.
168 Contexts.back().IsExpression = false;
169 } else if (Left->Previous && Left->Previous->is(tok::kw___attribute)) {
170 Left->Type = TT_AttributeParen;
171 } else if (Left->Previous && Left->Previous->is(TT_ForEachMacro)) {
172 // The first argument to a foreach macro is a declaration.
173 Contexts.back().IsForEachMacro = true;
174 Contexts.back().IsExpression = false;
175 } else if (Left->Previous && Left->Previous->MatchingParen &&
176 Left->Previous->MatchingParen->is(TT_ObjCBlockLParen)) {
177 Contexts.back().IsExpression = false;
178 } else if (!Line.MustBeDeclaration && !Line.InPPDirective) {
180 Left->Previous && Left->Previous->isOneOf(tok::kw_for, tok::kw_catch);
181 Contexts.back().IsExpression = !IsForOrCatch;
184 if (StartsObjCMethodExpr) {
185 Contexts.back().ColonIsObjCMethodExpr = true;
186 Left->Type = TT_ObjCMethodExpr;
189 bool MightBeFunctionType = !Contexts[Contexts.size() - 2].IsExpression;
190 bool ProbablyFunctionType = CurrentToken->isOneOf(tok::star, tok::amp);
191 bool HasMultipleLines = false;
192 bool HasMultipleParametersOnALine = false;
193 bool MightBeObjCForRangeLoop =
194 Left->Previous && Left->Previous->is(tok::kw_for);
195 while (CurrentToken) {
196 // LookForDecls is set when "if (" has been seen. Check for
197 // 'identifier' '*' 'identifier' followed by not '=' -- this
198 // '*' has to be a binary operator but determineStarAmpUsage() will
199 // categorize it as an unary operator, so set the right type here.
200 if (LookForDecls && CurrentToken->Next) {
201 FormatToken *Prev = CurrentToken->getPreviousNonComment();
203 FormatToken *PrevPrev = Prev->getPreviousNonComment();
204 FormatToken *Next = CurrentToken->Next;
205 if (PrevPrev && PrevPrev->is(tok::identifier) &&
206 Prev->isOneOf(tok::star, tok::amp, tok::ampamp) &&
207 CurrentToken->is(tok::identifier) && Next->isNot(tok::equal)) {
208 Prev->Type = TT_BinaryOperator;
209 LookForDecls = false;
214 if (CurrentToken->Previous->is(TT_PointerOrReference) &&
215 CurrentToken->Previous->Previous->isOneOf(tok::l_paren,
217 ProbablyFunctionType = true;
218 if (CurrentToken->is(tok::comma))
219 MightBeFunctionType = false;
220 if (CurrentToken->Previous->is(TT_BinaryOperator))
221 Contexts.back().IsExpression = true;
222 if (CurrentToken->is(tok::r_paren)) {
223 if (MightBeFunctionType && ProbablyFunctionType && CurrentToken->Next &&
224 (CurrentToken->Next->is(tok::l_paren) ||
225 (CurrentToken->Next->is(tok::l_square) && Line.MustBeDeclaration)))
226 Left->Type = TT_FunctionTypeLParen;
227 Left->MatchingParen = CurrentToken;
228 CurrentToken->MatchingParen = Left;
230 if (CurrentToken->Next && CurrentToken->Next->is(tok::l_brace) &&
231 Left->Previous && Left->Previous->is(tok::l_paren)) {
232 // Detect the case where macros are used to generate lambdas or
233 // function bodies, e.g.:
234 // auto my_lambda = MARCO((Type *type, int i) { .. body .. });
235 for (FormatToken *Tok = Left; Tok != CurrentToken; Tok = Tok->Next) {
236 if (Tok->is(TT_BinaryOperator) &&
237 Tok->isOneOf(tok::star, tok::amp, tok::ampamp))
238 Tok->Type = TT_PointerOrReference;
242 if (StartsObjCMethodExpr) {
243 CurrentToken->Type = TT_ObjCMethodExpr;
244 if (Contexts.back().FirstObjCSelectorName) {
245 Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName =
246 Contexts.back().LongestObjCSelectorName;
250 if (Left->is(TT_AttributeParen))
251 CurrentToken->Type = TT_AttributeParen;
252 if (Left->Previous && Left->Previous->is(TT_JavaAnnotation))
253 CurrentToken->Type = TT_JavaAnnotation;
254 if (Left->Previous && Left->Previous->is(TT_LeadingJavaAnnotation))
255 CurrentToken->Type = TT_LeadingJavaAnnotation;
257 if (!HasMultipleLines)
258 Left->PackingKind = PPK_Inconclusive;
259 else if (HasMultipleParametersOnALine)
260 Left->PackingKind = PPK_BinPacked;
262 Left->PackingKind = PPK_OnePerLine;
267 if (CurrentToken->isOneOf(tok::r_square, tok::r_brace))
270 if (CurrentToken->is(tok::l_brace))
271 Left->Type = TT_Unknown; // Not TT_ObjCBlockLParen
272 if (CurrentToken->is(tok::comma) && CurrentToken->Next &&
273 !CurrentToken->Next->HasUnescapedNewline &&
274 !CurrentToken->Next->isTrailingComment())
275 HasMultipleParametersOnALine = true;
276 if (CurrentToken->isOneOf(tok::kw_const, tok::kw_auto) ||
277 CurrentToken->isSimpleTypeSpecifier())
278 Contexts.back().IsExpression = false;
279 if (CurrentToken->isOneOf(tok::semi, tok::colon))
280 MightBeObjCForRangeLoop = false;
281 if (MightBeObjCForRangeLoop && CurrentToken->is(Keywords.kw_in))
282 CurrentToken->Type = TT_ObjCForIn;
283 // When we discover a 'new', we set CanBeExpression to 'false' in order to
284 // parse the type correctly. Reset that after a comma.
285 if (CurrentToken->is(tok::comma))
286 Contexts.back().CanBeExpression = true;
288 FormatToken *Tok = CurrentToken;
291 updateParameterCount(Left, Tok);
292 if (CurrentToken && CurrentToken->HasUnescapedNewline)
293 HasMultipleLines = true;
302 // A '[' could be an index subscript (after an identifier or after
303 // ')' or ']'), it could be the start of an Objective-C method
304 // expression, or it could the start of an Objective-C array literal.
305 FormatToken *Left = CurrentToken->Previous;
306 Left->ParentBracket = Contexts.back().ContextKind;
307 FormatToken *Parent = Left->getPreviousNonComment();
308 bool StartsObjCMethodExpr =
309 Style.Language == FormatStyle::LK_Cpp &&
310 Contexts.back().CanBeExpression && Left->isNot(TT_LambdaLSquare) &&
311 CurrentToken->isNot(tok::l_brace) &&
313 Parent->isOneOf(tok::colon, tok::l_square, tok::l_paren,
314 tok::kw_return, tok::kw_throw) ||
315 Parent->isUnaryOperator() ||
316 Parent->isOneOf(TT_ObjCForIn, TT_CastRParen) ||
317 getBinOpPrecedence(Parent->Tok.getKind(), true, true) > prec::Unknown);
318 bool ColonFound = false;
320 unsigned BindingIncrease = 1;
321 if (Left->is(TT_Unknown)) {
322 if (StartsObjCMethodExpr) {
323 Left->Type = TT_ObjCMethodExpr;
324 } else if (Style.Language == FormatStyle::LK_JavaScript && Parent &&
325 Contexts.back().ContextKind == tok::l_brace &&
326 Parent->isOneOf(tok::l_brace, tok::comma)) {
327 Left->Type = TT_JsComputedPropertyName;
328 } else if (Style.Language == FormatStyle::LK_Proto ||
330 Parent->isOneOf(TT_BinaryOperator, TT_TemplateCloser, tok::at,
331 tok::comma, tok::l_paren, tok::l_square,
332 tok::question, tok::colon, tok::kw_return,
333 // Should only be relevant to JavaScript:
335 Left->Type = TT_ArrayInitializerLSquare;
337 BindingIncrease = 10;
338 Left->Type = TT_ArraySubscriptLSquare;
342 ScopedContextCreator ContextCreator(*this, tok::l_square, BindingIncrease);
343 Contexts.back().IsExpression = true;
344 Contexts.back().ColonIsObjCMethodExpr = StartsObjCMethodExpr;
346 while (CurrentToken) {
347 if (CurrentToken->is(tok::r_square)) {
348 if (CurrentToken->Next && CurrentToken->Next->is(tok::l_paren) &&
349 Left->is(TT_ObjCMethodExpr)) {
350 // An ObjC method call is rarely followed by an open parenthesis.
351 // FIXME: Do we incorrectly label ":" with this?
352 StartsObjCMethodExpr = false;
353 Left->Type = TT_Unknown;
355 if (StartsObjCMethodExpr && CurrentToken->Previous != Left) {
356 CurrentToken->Type = TT_ObjCMethodExpr;
357 // determineStarAmpUsage() thinks that '*' '[' is allocating an
358 // array of pointers, but if '[' starts a selector then '*' is a
360 if (Parent && Parent->is(TT_PointerOrReference))
361 Parent->Type = TT_BinaryOperator;
363 Left->MatchingParen = CurrentToken;
364 CurrentToken->MatchingParen = Left;
365 if (Contexts.back().FirstObjCSelectorName) {
366 Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName =
367 Contexts.back().LongestObjCSelectorName;
368 if (Left->BlockParameterCount > 1)
369 Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = 0;
374 if (CurrentToken->isOneOf(tok::r_paren, tok::r_brace))
376 if (CurrentToken->is(tok::colon)) {
377 if (Left->is(TT_ArraySubscriptLSquare)) {
378 Left->Type = TT_ObjCMethodExpr;
379 StartsObjCMethodExpr = true;
380 Contexts.back().ColonIsObjCMethodExpr = true;
381 if (Parent && Parent->is(tok::r_paren))
382 Parent->Type = TT_CastRParen;
386 if (CurrentToken->is(tok::comma) && Left->is(TT_ObjCMethodExpr) &&
388 Left->Type = TT_ArrayInitializerLSquare;
389 FormatToken *Tok = CurrentToken;
392 updateParameterCount(Left, Tok);
399 FormatToken *Left = CurrentToken->Previous;
400 Left->ParentBracket = Contexts.back().ContextKind;
402 if (Contexts.back().CaretFound)
403 Left->Type = TT_ObjCBlockLBrace;
404 Contexts.back().CaretFound = false;
406 ScopedContextCreator ContextCreator(*this, tok::l_brace, 1);
407 Contexts.back().ColonIsDictLiteral = true;
408 if (Left->BlockKind == BK_BracedInit)
409 Contexts.back().IsExpression = true;
411 while (CurrentToken) {
412 if (CurrentToken->is(tok::r_brace)) {
413 Left->MatchingParen = CurrentToken;
414 CurrentToken->MatchingParen = Left;
418 if (CurrentToken->isOneOf(tok::r_paren, tok::r_square))
420 updateParameterCount(Left, CurrentToken);
421 if (CurrentToken->isOneOf(tok::colon, tok::l_brace)) {
422 FormatToken *Previous = CurrentToken->getPreviousNonComment();
423 if (((CurrentToken->is(tok::colon) &&
424 (!Contexts.back().ColonIsDictLiteral ||
425 Style.Language != FormatStyle::LK_Cpp)) ||
426 Style.Language == FormatStyle::LK_Proto) &&
427 (Previous->Tok.getIdentifierInfo() ||
428 Previous->is(tok::string_literal)))
429 Previous->Type = TT_SelectorName;
430 if (CurrentToken->is(tok::colon) ||
431 Style.Language == FormatStyle::LK_JavaScript)
432 Left->Type = TT_DictLiteral;
441 void updateParameterCount(FormatToken *Left, FormatToken *Current) {
442 if (Current->is(tok::l_brace) && Current->BlockKind == BK_Block)
443 ++Left->BlockParameterCount;
444 if (Current->is(tok::comma)) {
445 ++Left->ParameterCount;
447 Left->Role.reset(new CommaSeparatedList(Style));
448 Left->Role->CommaFound(Current);
449 } else if (Left->ParameterCount == 0 && Current->isNot(tok::comment)) {
450 Left->ParameterCount = 1;
454 bool parseConditional() {
455 while (CurrentToken) {
456 if (CurrentToken->is(tok::colon)) {
457 CurrentToken->Type = TT_ConditionalExpr;
467 bool parseTemplateDeclaration() {
468 if (CurrentToken && CurrentToken->is(tok::less)) {
469 CurrentToken->Type = TT_TemplateOpener;
474 CurrentToken->Previous->ClosesTemplateDeclaration = true;
480 bool consumeToken() {
481 FormatToken *Tok = CurrentToken;
483 switch (Tok->Tok.getKind()) {
486 if (!Tok->Previous && Line.MustBeDeclaration)
487 Tok->Type = TT_ObjCMethodSpecifier;
492 // Colons from ?: are handled in parseConditional().
493 if (Style.Language == FormatStyle::LK_JavaScript) {
494 if (Contexts.back().ColonIsForRangeExpr || // colon in for loop
495 (Contexts.size() == 1 && // switch/case labels
496 !Line.First->isOneOf(tok::kw_enum, tok::kw_case)) ||
497 Contexts.back().ContextKind == tok::l_paren || // function params
498 Contexts.back().ContextKind == tok::l_square || // array type
499 (Contexts.size() == 1 &&
500 Line.MustBeDeclaration)) { // method/property declaration
501 Tok->Type = TT_JsTypeColon;
505 if (Contexts.back().ColonIsDictLiteral ||
506 Style.Language == FormatStyle::LK_Proto) {
507 Tok->Type = TT_DictLiteral;
508 } else if (Contexts.back().ColonIsObjCMethodExpr ||
509 Line.startsWith(TT_ObjCMethodSpecifier)) {
510 Tok->Type = TT_ObjCMethodExpr;
511 Tok->Previous->Type = TT_SelectorName;
512 if (Tok->Previous->ColumnWidth >
513 Contexts.back().LongestObjCSelectorName)
514 Contexts.back().LongestObjCSelectorName = Tok->Previous->ColumnWidth;
515 if (!Contexts.back().FirstObjCSelectorName)
516 Contexts.back().FirstObjCSelectorName = Tok->Previous;
517 } else if (Contexts.back().ColonIsForRangeExpr) {
518 Tok->Type = TT_RangeBasedForLoopColon;
519 } else if (CurrentToken && CurrentToken->is(tok::numeric_constant)) {
520 Tok->Type = TT_BitFieldColon;
521 } else if (Contexts.size() == 1 &&
522 !Line.First->isOneOf(tok::kw_enum, tok::kw_case)) {
523 if (Tok->Previous->isOneOf(tok::r_paren, tok::kw_noexcept))
524 Tok->Type = TT_CtorInitializerColon;
526 Tok->Type = TT_InheritanceColon;
527 } else if (Tok->Previous->is(tok::identifier) && Tok->Next &&
528 Tok->Next->isOneOf(tok::r_paren, tok::comma)) {
529 // This handles a special macro in ObjC code where selectors including
530 // the colon are passed as macro arguments.
531 Tok->Type = TT_ObjCMethodExpr;
532 } else if (Contexts.back().ContextKind == tok::l_paren) {
533 Tok->Type = TT_InlineASMColon;
538 // | and & in declarations/type expressions represent union and
539 // intersection types, respectively.
540 if (Style.Language == FormatStyle::LK_JavaScript &&
541 !Contexts.back().IsExpression)
542 Tok->Type = TT_JsTypeOperator;
546 if (CurrentToken && CurrentToken->is(tok::l_paren)) {
548 if (!parseParens(/*LookForDecls=*/true))
553 if (Style.Language == FormatStyle::LK_JavaScript && Tok->Previous &&
554 Tok->Previous->is(tok::period))
556 Contexts.back().ColonIsForRangeExpr = true;
562 // When faced with 'operator()()', the kw_operator handler incorrectly
563 // marks the first l_paren as a OverloadedOperatorLParen. Here, we make
564 // the first two parens OverloadedOperators and the second l_paren an
565 // OverloadedOperatorLParen.
567 Tok->Previous->is(tok::r_paren) &&
568 Tok->Previous->MatchingParen &&
569 Tok->Previous->MatchingParen->is(TT_OverloadedOperatorLParen)) {
570 Tok->Previous->Type = TT_OverloadedOperator;
571 Tok->Previous->MatchingParen->Type = TT_OverloadedOperator;
572 Tok->Type = TT_OverloadedOperatorLParen;
577 if (Line.MustBeDeclaration && Contexts.size() == 1 &&
578 !Contexts.back().IsExpression && !Line.startsWith(TT_ObjCProperty) &&
580 !Tok->Previous->isOneOf(tok::kw_decltype, tok::kw___attribute,
581 TT_LeadingJavaAnnotation)))
582 Line.MightBeFunctionDecl = true;
594 Tok->Type = TT_TemplateOpener;
596 Tok->Type = TT_BinaryOperator;
597 NonTemplateLess.insert(Tok);
606 // Lines can start with '}'.
611 Tok->Type = TT_BinaryOperator;
613 case tok::kw_operator:
614 while (CurrentToken &&
615 !CurrentToken->isOneOf(tok::l_paren, tok::semi, tok::r_paren)) {
616 if (CurrentToken->isOneOf(tok::star, tok::amp))
617 CurrentToken->Type = TT_PointerOrReference;
620 CurrentToken->Previous->isOneOf(TT_BinaryOperator, tok::comma))
621 CurrentToken->Previous->Type = TT_OverloadedOperator;
624 CurrentToken->Type = TT_OverloadedOperatorLParen;
625 if (CurrentToken->Previous->is(TT_BinaryOperator))
626 CurrentToken->Previous->Type = TT_OverloadedOperator;
630 if (Style.Language == FormatStyle::LK_JavaScript && Tok->Next &&
631 Tok->Next->isOneOf(tok::semi, tok::comma, tok::colon, tok::r_paren,
633 // Question marks before semicolons, colons, etc. indicate optional
634 // types (fields, parameters), e.g.
635 // function(x?: string, y?) {...}
637 Tok->Type = TT_JsTypeOptionalQuestion;
640 // Declarations cannot be conditional expressions, this can only be part
641 // of a type declaration.
642 if (Line.MustBeDeclaration && !Contexts.back().IsExpression &&
643 Style.Language == FormatStyle::LK_JavaScript)
647 case tok::kw_template:
648 parseTemplateDeclaration();
651 if (Contexts.back().InCtorInitializer)
652 Tok->Type = TT_CtorInitializerComma;
653 else if (Contexts.back().FirstStartOfName &&
654 (Contexts.size() == 1 || Line.startsWith(tok::kw_for))) {
655 Contexts.back().FirstStartOfName->PartOfMultiVariableDeclStmt = true;
656 Line.IsMultiVariableDeclStmt = true;
658 if (Contexts.back().IsForEachMacro)
659 Contexts.back().IsExpression = true;
667 void parseIncludeDirective() {
668 if (CurrentToken && CurrentToken->is(tok::less)) {
670 while (CurrentToken) {
671 if (CurrentToken->isNot(tok::comment) || CurrentToken->Next)
672 CurrentToken->Type = TT_ImplicitStringLiteral;
678 void parseWarningOrError() {
680 // We still want to format the whitespace left of the first token of the
683 while (CurrentToken) {
684 CurrentToken->Type = TT_ImplicitStringLiteral;
690 next(); // Consume "pragma".
692 CurrentToken->isOneOf(Keywords.kw_mark, Keywords.kw_option)) {
693 bool IsMark = CurrentToken->is(Keywords.kw_mark);
694 next(); // Consume "mark".
695 next(); // Consume first token (so we fix leading whitespace).
696 while (CurrentToken) {
697 if (IsMark || CurrentToken->Previous->is(TT_BinaryOperator))
698 CurrentToken->Type = TT_ImplicitStringLiteral;
704 LineType parsePreprocessorDirective() {
705 bool IsFirstToken = CurrentToken->IsFirst;
706 LineType Type = LT_PreprocessorDirective;
711 if (Style.Language == FormatStyle::LK_JavaScript && IsFirstToken) {
712 // JavaScript files can contain shebang lines of the form:
713 // #!/usr/bin/env node
714 // Treat these like C++ #include directives.
715 while (CurrentToken) {
716 // Tokens cannot be comments here.
717 CurrentToken->Type = TT_ImplicitStringLiteral;
720 return LT_ImportStatement;
723 if (CurrentToken->Tok.is(tok::numeric_constant)) {
724 CurrentToken->SpacesRequiredBefore = 1;
727 // Hashes in the middle of a line can lead to any strange token
729 if (!CurrentToken->Tok.getIdentifierInfo())
731 switch (CurrentToken->Tok.getIdentifierInfo()->getPPKeywordID()) {
732 case tok::pp_include:
733 case tok::pp_include_next:
736 parseIncludeDirective();
737 Type = LT_ImportStatement;
740 case tok::pp_warning:
741 parseWarningOrError();
748 Contexts.back().IsExpression = true;
760 LineType parseLine() {
761 NonTemplateLess.clear();
762 if (CurrentToken->is(tok::hash))
763 return parsePreprocessorDirective();
765 // Directly allow to 'import <string-literal>' to support protocol buffer
766 // definitions (code.google.com/p/protobuf) or missing "#" (either way we
767 // should not break the line).
768 IdentifierInfo *Info = CurrentToken->Tok.getIdentifierInfo();
769 if ((Style.Language == FormatStyle::LK_Java &&
770 CurrentToken->is(Keywords.kw_package)) ||
771 (Info && Info->getPPKeywordID() == tok::pp_import &&
772 CurrentToken->Next &&
773 CurrentToken->Next->isOneOf(tok::string_literal, tok::identifier,
776 parseIncludeDirective();
777 return LT_ImportStatement;
780 // If this line starts and ends in '<' and '>', respectively, it is likely
781 // part of "#define <a/b.h>".
782 if (CurrentToken->is(tok::less) && Line.Last->is(tok::greater)) {
783 parseIncludeDirective();
784 return LT_ImportStatement;
787 // In .proto files, top-level options are very similar to import statements
788 // and should not be line-wrapped.
789 if (Style.Language == FormatStyle::LK_Proto && Line.Level == 0 &&
790 CurrentToken->is(Keywords.kw_option)) {
792 if (CurrentToken && CurrentToken->is(tok::identifier))
793 return LT_ImportStatement;
796 bool KeywordVirtualFound = false;
797 bool ImportStatement = false;
799 // import {...} from '...';
800 if (Style.Language == FormatStyle::LK_JavaScript &&
801 CurrentToken->is(Keywords.kw_import))
802 ImportStatement = true;
804 while (CurrentToken) {
805 if (CurrentToken->is(tok::kw_virtual))
806 KeywordVirtualFound = true;
807 if (Style.Language == FormatStyle::LK_JavaScript) {
808 // export {...} from '...';
809 // An export followed by "from 'some string';" is a re-export from
810 // another module identified by a URI and is treated as a
811 // LT_ImportStatement (i.e. prevent wraps on it for long URIs).
812 // Just "export {...};" or "export class ..." should not be treated as
813 // an import in this sense.
814 if (Line.First->is(tok::kw_export) &&
815 CurrentToken->is(Keywords.kw_from) && CurrentToken->Next &&
816 CurrentToken->Next->isStringLiteral())
817 ImportStatement = true;
818 if (isClosureImportStatement(*CurrentToken))
819 ImportStatement = true;
824 if (KeywordVirtualFound)
825 return LT_VirtualFunctionDecl;
827 return LT_ImportStatement;
829 if (Line.startsWith(TT_ObjCMethodSpecifier)) {
830 if (Contexts.back().FirstObjCSelectorName)
831 Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName =
832 Contexts.back().LongestObjCSelectorName;
833 return LT_ObjCMethodDecl;
840 bool isClosureImportStatement(const FormatToken &Tok) {
841 // FIXME: Closure-library specific stuff should not be hard-coded but be
843 return Tok.TokenText == "goog" && Tok.Next && Tok.Next->is(tok::period) &&
844 Tok.Next->Next && (Tok.Next->Next->TokenText == "module" ||
845 Tok.Next->Next->TokenText == "provide" ||
846 Tok.Next->Next->TokenText == "require" ||
847 Tok.Next->Next->TokenText == "setTestOnly" ||
848 Tok.Next->Next->TokenText == "forwardDeclare") &&
849 Tok.Next->Next->Next && Tok.Next->Next->Next->is(tok::l_paren);
852 void resetTokenMetadata(FormatToken *Token) {
856 // Reset token type in case we have already looked at it and then
857 // recovered from an error (e.g. failure to find the matching >).
858 if (!CurrentToken->isOneOf(TT_LambdaLSquare, TT_ForEachMacro,
859 TT_FunctionLBrace, TT_ImplicitStringLiteral,
860 TT_InlineASMBrace, TT_JsFatArrow, TT_LambdaArrow,
861 TT_RegexLiteral, TT_TemplateString))
862 CurrentToken->Type = TT_Unknown;
863 CurrentToken->Role.reset();
864 CurrentToken->MatchingParen = nullptr;
865 CurrentToken->FakeLParens.clear();
866 CurrentToken->FakeRParens = 0;
871 CurrentToken->NestingLevel = Contexts.size() - 1;
872 CurrentToken->BindingStrength = Contexts.back().BindingStrength;
873 modifyContext(*CurrentToken);
874 determineTokenType(*CurrentToken);
875 CurrentToken = CurrentToken->Next;
878 resetTokenMetadata(CurrentToken);
881 /// \brief A struct to hold information valid in a specific context, e.g.
882 /// a pair of parenthesis.
884 Context(tok::TokenKind ContextKind, unsigned BindingStrength,
886 : ContextKind(ContextKind), BindingStrength(BindingStrength),
887 IsExpression(IsExpression) {}
889 tok::TokenKind ContextKind;
890 unsigned BindingStrength;
892 unsigned LongestObjCSelectorName = 0;
893 bool ColonIsForRangeExpr = false;
894 bool ColonIsDictLiteral = false;
895 bool ColonIsObjCMethodExpr = false;
896 FormatToken *FirstObjCSelectorName = nullptr;
897 FormatToken *FirstStartOfName = nullptr;
898 bool CanBeExpression = true;
899 bool InTemplateArgument = false;
900 bool InCtorInitializer = false;
901 bool CaretFound = false;
902 bool IsForEachMacro = false;
905 /// \brief Puts a new \c Context onto the stack \c Contexts for the lifetime
906 /// of each instance.
907 struct ScopedContextCreator {
910 ScopedContextCreator(AnnotatingParser &P, tok::TokenKind ContextKind,
913 P.Contexts.push_back(Context(ContextKind,
914 P.Contexts.back().BindingStrength + Increase,
915 P.Contexts.back().IsExpression));
918 ~ScopedContextCreator() { P.Contexts.pop_back(); }
921 void modifyContext(const FormatToken &Current) {
922 if (Current.getPrecedence() == prec::Assignment &&
923 !Line.First->isOneOf(tok::kw_template, tok::kw_using, tok::kw_return) &&
924 // Type aliases use `type X = ...;` in TypeScript.
925 !(Style.Language == FormatStyle::LK_JavaScript &&
926 Line.startsWith(Keywords.kw_type, tok::identifier)) &&
927 (!Current.Previous || Current.Previous->isNot(tok::kw_operator))) {
928 Contexts.back().IsExpression = true;
929 if (!Line.startsWith(TT_UnaryOperator)) {
930 for (FormatToken *Previous = Current.Previous;
931 Previous && Previous->Previous &&
932 !Previous->Previous->isOneOf(tok::comma, tok::semi);
933 Previous = Previous->Previous) {
934 if (Previous->isOneOf(tok::r_square, tok::r_paren)) {
935 Previous = Previous->MatchingParen;
939 if (Previous->opensScope())
941 if (Previous->isOneOf(TT_BinaryOperator, TT_UnaryOperator) &&
942 Previous->isOneOf(tok::star, tok::amp, tok::ampamp) &&
943 Previous->Previous && Previous->Previous->isNot(tok::equal))
944 Previous->Type = TT_PointerOrReference;
947 } else if (Current.is(tok::lessless) &&
948 (!Current.Previous || !Current.Previous->is(tok::kw_operator))) {
949 Contexts.back().IsExpression = true;
950 } else if (Current.isOneOf(tok::kw_return, tok::kw_throw)) {
951 Contexts.back().IsExpression = true;
952 } else if (Current.is(TT_TrailingReturnArrow)) {
953 Contexts.back().IsExpression = false;
954 } else if (Current.is(TT_LambdaArrow) || Current.is(Keywords.kw_assert)) {
955 Contexts.back().IsExpression = Style.Language == FormatStyle::LK_Java;
956 } else if (Current.Previous &&
957 Current.Previous->is(TT_CtorInitializerColon)) {
958 Contexts.back().IsExpression = true;
959 Contexts.back().InCtorInitializer = true;
960 } else if (Current.isOneOf(tok::r_paren, tok::greater, tok::comma)) {
961 for (FormatToken *Previous = Current.Previous;
962 Previous && Previous->isOneOf(tok::star, tok::amp);
963 Previous = Previous->Previous)
964 Previous->Type = TT_PointerOrReference;
965 if (Line.MustBeDeclaration && !Contexts.front().InCtorInitializer)
966 Contexts.back().IsExpression = false;
967 } else if (Current.is(tok::kw_new)) {
968 Contexts.back().CanBeExpression = false;
969 } else if (Current.isOneOf(tok::semi, tok::exclaim)) {
970 // This should be the condition or increment in a for-loop.
971 Contexts.back().IsExpression = true;
975 void determineTokenType(FormatToken &Current) {
976 if (!Current.is(TT_Unknown))
977 // The token type is already known.
980 // Line.MightBeFunctionDecl can only be true after the parentheses of a
981 // function declaration have been found. In this case, 'Current' is a
982 // trailing token of this declaration and thus cannot be a name.
983 if (Current.is(Keywords.kw_instanceof)) {
984 Current.Type = TT_BinaryOperator;
985 } else if (isStartOfName(Current) &&
986 (!Line.MightBeFunctionDecl || Current.NestingLevel != 0)) {
987 Contexts.back().FirstStartOfName = &Current;
988 Current.Type = TT_StartOfName;
989 } else if (Current.isOneOf(tok::kw_auto, tok::kw___auto_type)) {
991 } else if (Current.is(tok::arrow) &&
992 Style.Language == FormatStyle::LK_Java) {
993 Current.Type = TT_LambdaArrow;
994 } else if (Current.is(tok::arrow) && AutoFound && Line.MustBeDeclaration &&
995 Current.NestingLevel == 0) {
996 Current.Type = TT_TrailingReturnArrow;
997 } else if (Current.isOneOf(tok::star, tok::amp, tok::ampamp)) {
999 determineStarAmpUsage(Current, Contexts.back().CanBeExpression &&
1000 Contexts.back().IsExpression,
1001 Contexts.back().InTemplateArgument);
1002 } else if (Current.isOneOf(tok::minus, tok::plus, tok::caret)) {
1003 Current.Type = determinePlusMinusCaretUsage(Current);
1004 if (Current.is(TT_UnaryOperator) && Current.is(tok::caret))
1005 Contexts.back().CaretFound = true;
1006 } else if (Current.isOneOf(tok::minusminus, tok::plusplus)) {
1007 Current.Type = determineIncrementUsage(Current);
1008 } else if (Current.isOneOf(tok::exclaim, tok::tilde)) {
1009 Current.Type = TT_UnaryOperator;
1010 } else if (Current.is(tok::question)) {
1011 if (Style.Language == FormatStyle::LK_JavaScript &&
1012 Line.MustBeDeclaration && !Contexts.back().IsExpression) {
1013 // In JavaScript, `interface X { foo?(): bar; }` is an optional method
1014 // on the interface, not a ternary expression.
1015 Current.Type = TT_JsTypeOptionalQuestion;
1017 Current.Type = TT_ConditionalExpr;
1019 } else if (Current.isBinaryOperator() &&
1020 (!Current.Previous || Current.Previous->isNot(tok::l_square))) {
1021 Current.Type = TT_BinaryOperator;
1022 } else if (Current.is(tok::comment)) {
1023 if (Current.TokenText.startswith("/*")) {
1024 if (Current.TokenText.endswith("*/"))
1025 Current.Type = TT_BlockComment;
1027 // The lexer has for some reason determined a comment here. But we
1028 // cannot really handle it, if it isn't properly terminated.
1029 Current.Tok.setKind(tok::unknown);
1031 Current.Type = TT_LineComment;
1033 } else if (Current.is(tok::r_paren)) {
1034 if (rParenEndsCast(Current))
1035 Current.Type = TT_CastRParen;
1036 if (Current.MatchingParen && Current.Next &&
1037 !Current.Next->isBinaryOperator() &&
1038 !Current.Next->isOneOf(tok::semi, tok::colon, tok::l_brace,
1039 tok::period, tok::arrow, tok::coloncolon))
1040 if (FormatToken *AfterParen = Current.MatchingParen->Next) {
1041 // Make sure this isn't the return type of an Obj-C block declaration
1042 if (AfterParen->Tok.isNot(tok::caret)) {
1043 if (FormatToken *BeforeParen = Current.MatchingParen->Previous)
1044 if (BeforeParen->is(tok::identifier) &&
1045 BeforeParen->TokenText == BeforeParen->TokenText.upper() &&
1046 (!BeforeParen->Previous ||
1047 BeforeParen->Previous->ClosesTemplateDeclaration))
1048 Current.Type = TT_FunctionAnnotationRParen;
1051 } else if (Current.is(tok::at) && Current.Next) {
1052 if (Current.Next->isStringLiteral()) {
1053 Current.Type = TT_ObjCStringLiteral;
1055 switch (Current.Next->Tok.getObjCKeywordID()) {
1056 case tok::objc_interface:
1057 case tok::objc_implementation:
1058 case tok::objc_protocol:
1059 Current.Type = TT_ObjCDecl;
1061 case tok::objc_property:
1062 Current.Type = TT_ObjCProperty;
1068 } else if (Current.is(tok::period)) {
1069 FormatToken *PreviousNoComment = Current.getPreviousNonComment();
1070 if (PreviousNoComment &&
1071 PreviousNoComment->isOneOf(tok::comma, tok::l_brace))
1072 Current.Type = TT_DesignatedInitializerPeriod;
1073 else if (Style.Language == FormatStyle::LK_Java && Current.Previous &&
1074 Current.Previous->isOneOf(TT_JavaAnnotation,
1075 TT_LeadingJavaAnnotation)) {
1076 Current.Type = Current.Previous->Type;
1078 } else if (Current.isOneOf(tok::identifier, tok::kw_const) &&
1080 !Current.Previous->isOneOf(tok::equal, tok::at) &&
1081 Line.MightBeFunctionDecl && Contexts.size() == 1) {
1082 // Line.MightBeFunctionDecl can only be true after the parentheses of a
1083 // function declaration have been found.
1084 Current.Type = TT_TrailingAnnotation;
1085 } else if ((Style.Language == FormatStyle::LK_Java ||
1086 Style.Language == FormatStyle::LK_JavaScript) &&
1088 if (Current.Previous->is(tok::at) &&
1089 Current.isNot(Keywords.kw_interface)) {
1090 const FormatToken &AtToken = *Current.Previous;
1091 const FormatToken *Previous = AtToken.getPreviousNonComment();
1092 if (!Previous || Previous->is(TT_LeadingJavaAnnotation))
1093 Current.Type = TT_LeadingJavaAnnotation;
1095 Current.Type = TT_JavaAnnotation;
1096 } else if (Current.Previous->is(tok::period) &&
1097 Current.Previous->isOneOf(TT_JavaAnnotation,
1098 TT_LeadingJavaAnnotation)) {
1099 Current.Type = Current.Previous->Type;
1104 /// \brief Take a guess at whether \p Tok starts a name of a function or
1105 /// variable declaration.
1107 /// This is a heuristic based on whether \p Tok is an identifier following
1108 /// something that is likely a type.
1109 bool isStartOfName(const FormatToken &Tok) {
1110 if (Tok.isNot(tok::identifier) || !Tok.Previous)
1113 if (Tok.Previous->isOneOf(TT_LeadingJavaAnnotation, Keywords.kw_instanceof))
1115 if (Style.Language == FormatStyle::LK_JavaScript &&
1116 Tok.Previous->is(Keywords.kw_in))
1119 // Skip "const" as it does not have an influence on whether this is a name.
1120 FormatToken *PreviousNotConst = Tok.Previous;
1121 while (PreviousNotConst && PreviousNotConst->is(tok::kw_const))
1122 PreviousNotConst = PreviousNotConst->Previous;
1124 if (!PreviousNotConst)
1127 bool IsPPKeyword = PreviousNotConst->is(tok::identifier) &&
1128 PreviousNotConst->Previous &&
1129 PreviousNotConst->Previous->is(tok::hash);
1131 if (PreviousNotConst->is(TT_TemplateCloser))
1132 return PreviousNotConst && PreviousNotConst->MatchingParen &&
1133 PreviousNotConst->MatchingParen->Previous &&
1134 PreviousNotConst->MatchingParen->Previous->isNot(tok::period) &&
1135 PreviousNotConst->MatchingParen->Previous->isNot(tok::kw_template);
1137 if (PreviousNotConst->is(tok::r_paren) && PreviousNotConst->MatchingParen &&
1138 PreviousNotConst->MatchingParen->Previous &&
1139 PreviousNotConst->MatchingParen->Previous->is(tok::kw_decltype))
1142 return (!IsPPKeyword &&
1143 PreviousNotConst->isOneOf(tok::identifier, tok::kw_auto)) ||
1144 PreviousNotConst->is(TT_PointerOrReference) ||
1145 PreviousNotConst->isSimpleTypeSpecifier();
1148 /// \brief Determine whether ')' is ending a cast.
1149 bool rParenEndsCast(const FormatToken &Tok) {
1150 // C-style casts are only used in C++ and Java.
1151 if (Style.Language != FormatStyle::LK_Cpp &&
1152 Style.Language != FormatStyle::LK_Java)
1155 // Empty parens aren't casts and there are no casts at the end of the line.
1156 if (Tok.Previous == Tok.MatchingParen || !Tok.Next || !Tok.MatchingParen)
1159 FormatToken *LeftOfParens = Tok.MatchingParen->getPreviousNonComment();
1161 // If there is a closing parenthesis left of the current parentheses,
1162 // look past it as these might be chained casts.
1163 if (LeftOfParens->is(tok::r_paren)) {
1164 if (!LeftOfParens->MatchingParen ||
1165 !LeftOfParens->MatchingParen->Previous)
1167 LeftOfParens = LeftOfParens->MatchingParen->Previous;
1170 // If there is an identifier (or with a few exceptions a keyword) right
1171 // before the parentheses, this is unlikely to be a cast.
1172 if (LeftOfParens->Tok.getIdentifierInfo() &&
1173 !LeftOfParens->isOneOf(Keywords.kw_in, tok::kw_return, tok::kw_case,
1177 // Certain other tokens right before the parentheses are also signals that
1178 // this cannot be a cast.
1179 if (LeftOfParens->isOneOf(tok::at, tok::r_square, TT_OverloadedOperator,
1180 TT_TemplateCloser, tok::ellipsis))
1184 if (Tok.Next->is(tok::question))
1187 // As Java has no function types, a "(" after the ")" likely means that this
1189 if (Style.Language == FormatStyle::LK_Java && Tok.Next->is(tok::l_paren))
1192 // If a (non-string) literal follows, this is likely a cast.
1193 if (Tok.Next->isNot(tok::string_literal) &&
1194 (Tok.Next->Tok.isLiteral() ||
1195 Tok.Next->isOneOf(tok::kw_sizeof, tok::kw_alignof)))
1198 // Heuristically try to determine whether the parentheses contain a type.
1199 bool ParensAreType =
1201 Tok.Previous->isOneOf(TT_PointerOrReference, TT_TemplateCloser) ||
1202 Tok.Previous->isSimpleTypeSpecifier();
1203 bool ParensCouldEndDecl =
1204 Tok.Next->isOneOf(tok::equal, tok::semi, tok::l_brace, tok::greater);
1205 if (ParensAreType && !ParensCouldEndDecl)
1208 // At this point, we heuristically assume that there are no casts at the
1209 // start of the line. We assume that we have found most cases where there
1210 // are by the logic above, e.g. "(void)x;".
1214 // If the following token is an identifier or 'this', this is a cast. All
1215 // cases where this can be something else are handled above.
1216 if (Tok.Next->isOneOf(tok::identifier, tok::kw_this))
1219 if (!Tok.Next->Next)
1222 // If the next token after the parenthesis is a unary operator, assume
1223 // that this is cast, unless there are unexpected tokens inside the
1226 Tok.Next->isUnaryOperator() || Tok.Next->isOneOf(tok::amp, tok::star);
1227 if (!NextIsUnary || Tok.Next->is(tok::plus) ||
1228 !Tok.Next->Next->isOneOf(tok::identifier, tok::numeric_constant))
1230 // Search for unexpected tokens.
1231 for (FormatToken *Prev = Tok.Previous; Prev != Tok.MatchingParen;
1232 Prev = Prev->Previous) {
1233 if (!Prev->isOneOf(tok::kw_const, tok::identifier, tok::coloncolon))
1239 /// \brief Return the type of the given token assuming it is * or &.
1240 TokenType determineStarAmpUsage(const FormatToken &Tok, bool IsExpression,
1241 bool InTemplateArgument) {
1242 if (Style.Language == FormatStyle::LK_JavaScript)
1243 return TT_BinaryOperator;
1245 const FormatToken *PrevToken = Tok.getPreviousNonComment();
1247 return TT_UnaryOperator;
1249 const FormatToken *NextToken = Tok.getNextNonComment();
1251 NextToken->isOneOf(tok::arrow, Keywords.kw_final,
1252 Keywords.kw_override) ||
1253 (NextToken->is(tok::l_brace) && !NextToken->getNextNonComment()))
1254 return TT_PointerOrReference;
1256 if (PrevToken->is(tok::coloncolon))
1257 return TT_PointerOrReference;
1259 if (PrevToken->isOneOf(tok::l_paren, tok::l_square, tok::l_brace,
1260 tok::comma, tok::semi, tok::kw_return, tok::colon,
1261 tok::equal, tok::kw_delete, tok::kw_sizeof) ||
1262 PrevToken->isOneOf(TT_BinaryOperator, TT_ConditionalExpr,
1263 TT_UnaryOperator, TT_CastRParen))
1264 return TT_UnaryOperator;
1266 if (NextToken->is(tok::l_square) && NextToken->isNot(TT_LambdaLSquare))
1267 return TT_PointerOrReference;
1268 if (NextToken->is(tok::kw_operator) && !IsExpression)
1269 return TT_PointerOrReference;
1270 if (NextToken->isOneOf(tok::comma, tok::semi))
1271 return TT_PointerOrReference;
1273 if (PrevToken->is(tok::r_paren) && PrevToken->MatchingParen &&
1274 PrevToken->MatchingParen->Previous &&
1275 PrevToken->MatchingParen->Previous->isOneOf(tok::kw_typeof,
1277 return TT_PointerOrReference;
1279 if (PrevToken->Tok.isLiteral() ||
1280 PrevToken->isOneOf(tok::r_paren, tok::r_square, tok::kw_true,
1281 tok::kw_false, tok::r_brace) ||
1282 NextToken->Tok.isLiteral() ||
1283 NextToken->isOneOf(tok::kw_true, tok::kw_false) ||
1284 NextToken->isUnaryOperator() ||
1285 // If we know we're in a template argument, there are no named
1286 // declarations. Thus, having an identifier on the right-hand side
1287 // indicates a binary operator.
1288 (InTemplateArgument && NextToken->Tok.isAnyIdentifier()))
1289 return TT_BinaryOperator;
1291 // "&&(" is quite unlikely to be two successive unary "&".
1292 if (Tok.is(tok::ampamp) && NextToken && NextToken->is(tok::l_paren))
1293 return TT_BinaryOperator;
1295 // This catches some cases where evaluation order is used as control flow:
1297 const FormatToken *NextNextToken = NextToken->getNextNonComment();
1298 if (NextNextToken && NextNextToken->is(tok::arrow))
1299 return TT_BinaryOperator;
1301 // It is very unlikely that we are going to find a pointer or reference type
1302 // definition on the RHS of an assignment.
1303 if (IsExpression && !Contexts.back().CaretFound)
1304 return TT_BinaryOperator;
1306 return TT_PointerOrReference;
1309 TokenType determinePlusMinusCaretUsage(const FormatToken &Tok) {
1310 const FormatToken *PrevToken = Tok.getPreviousNonComment();
1311 if (!PrevToken || PrevToken->is(TT_CastRParen))
1312 return TT_UnaryOperator;
1314 // Use heuristics to recognize unary operators.
1315 if (PrevToken->isOneOf(tok::equal, tok::l_paren, tok::comma, tok::l_square,
1316 tok::question, tok::colon, tok::kw_return,
1317 tok::kw_case, tok::at, tok::l_brace))
1318 return TT_UnaryOperator;
1320 // There can't be two consecutive binary operators.
1321 if (PrevToken->is(TT_BinaryOperator))
1322 return TT_UnaryOperator;
1324 // Fall back to marking the token as binary operator.
1325 return TT_BinaryOperator;
1328 /// \brief Determine whether ++/-- are pre- or post-increments/-decrements.
1329 TokenType determineIncrementUsage(const FormatToken &Tok) {
1330 const FormatToken *PrevToken = Tok.getPreviousNonComment();
1331 if (!PrevToken || PrevToken->is(TT_CastRParen))
1332 return TT_UnaryOperator;
1333 if (PrevToken->isOneOf(tok::r_paren, tok::r_square, tok::identifier))
1334 return TT_TrailingUnaryOperator;
1336 return TT_UnaryOperator;
1339 SmallVector<Context, 8> Contexts;
1341 const FormatStyle &Style;
1342 AnnotatedLine &Line;
1343 FormatToken *CurrentToken;
1345 const AdditionalKeywords &Keywords;
1347 // Set of "<" tokens that do not open a template parameter list. If parseAngle
1348 // determines that a specific token can't be a template opener, it will make
1349 // same decision irrespective of the decisions for tokens leading up to it.
1350 // Store this information to prevent this from causing exponential runtime.
1351 llvm::SmallPtrSet<FormatToken *, 16> NonTemplateLess;
1354 static const int PrecedenceUnaryOperator = prec::PointerToMember + 1;
1355 static const int PrecedenceArrowAndPeriod = prec::PointerToMember + 2;
1357 /// \brief Parses binary expressions by inserting fake parenthesis based on
1358 /// operator precedence.
1359 class ExpressionParser {
1361 ExpressionParser(const FormatStyle &Style, const AdditionalKeywords &Keywords,
1362 AnnotatedLine &Line)
1363 : Style(Style), Keywords(Keywords), Current(Line.First) {}
1365 /// \brief Parse expressions with the given operatore precedence.
1366 void parse(int Precedence = 0) {
1367 // Skip 'return' and ObjC selector colons as they are not part of a binary
1369 while (Current && (Current->is(tok::kw_return) ||
1370 (Current->is(tok::colon) &&
1371 Current->isOneOf(TT_ObjCMethodExpr, TT_DictLiteral))))
1374 if (!Current || Precedence > PrecedenceArrowAndPeriod)
1377 // Conditional expressions need to be parsed separately for proper nesting.
1378 if (Precedence == prec::Conditional) {
1379 parseConditionalExpr();
1383 // Parse unary operators, which all have a higher precedence than binary
1385 if (Precedence == PrecedenceUnaryOperator) {
1386 parseUnaryOperator();
1390 FormatToken *Start = Current;
1391 FormatToken *LatestOperator = nullptr;
1392 unsigned OperatorIndex = 0;
1395 // Consume operators with higher precedence.
1396 parse(Precedence + 1);
1398 int CurrentPrecedence = getCurrentPrecedence();
1400 if (Current && Current->is(TT_SelectorName) &&
1401 Precedence == CurrentPrecedence) {
1403 addFakeParenthesis(Start, prec::Level(Precedence));
1407 // At the end of the line or when an operator with higher precedence is
1408 // found, insert fake parenthesis and return.
1409 if (!Current || (Current->closesScope() && Current->MatchingParen) ||
1410 (CurrentPrecedence != -1 && CurrentPrecedence < Precedence) ||
1411 (CurrentPrecedence == prec::Conditional &&
1412 Precedence == prec::Assignment && Current->is(tok::colon))) {
1416 // Consume scopes: (), [], <> and {}
1417 if (Current->opensScope()) {
1418 while (Current && !Current->closesScope()) {
1425 if (CurrentPrecedence == Precedence) {
1427 LatestOperator->NextOperator = Current;
1428 LatestOperator = Current;
1429 Current->OperatorIndex = OperatorIndex;
1432 next(/*SkipPastLeadingComments=*/Precedence > 0);
1436 if (LatestOperator && (Current || Precedence > 0)) {
1437 // LatestOperator->LastOperator = true;
1438 if (Precedence == PrecedenceArrowAndPeriod) {
1439 // Call expressions don't have a binary operator precedence.
1440 addFakeParenthesis(Start, prec::Unknown);
1442 addFakeParenthesis(Start, prec::Level(Precedence));
1448 /// \brief Gets the precedence (+1) of the given token for binary operators
1449 /// and other tokens that we treat like binary operators.
1450 int getCurrentPrecedence() {
1452 const FormatToken *NextNonComment = Current->getNextNonComment();
1453 if (Current->is(TT_ConditionalExpr))
1454 return prec::Conditional;
1455 if (NextNonComment && NextNonComment->is(tok::colon) &&
1456 NextNonComment->is(TT_DictLiteral))
1458 if (Current->is(TT_LambdaArrow))
1460 if (Current->is(TT_JsFatArrow))
1461 return prec::Assignment;
1462 if (Current->isOneOf(tok::semi, TT_InlineASMColon, TT_SelectorName,
1463 TT_JsComputedPropertyName) ||
1464 (Current->is(tok::comment) && NextNonComment &&
1465 NextNonComment->is(TT_SelectorName)))
1467 if (Current->is(TT_RangeBasedForLoopColon))
1469 if ((Style.Language == FormatStyle::LK_Java ||
1470 Style.Language == FormatStyle::LK_JavaScript) &&
1471 Current->is(Keywords.kw_instanceof))
1472 return prec::Relational;
1473 if (Style.Language == FormatStyle::LK_JavaScript &&
1474 Current->is(Keywords.kw_in))
1475 return prec::Relational;
1476 if (Current->is(TT_BinaryOperator) || Current->is(tok::comma))
1477 return Current->getPrecedence();
1478 if (Current->isOneOf(tok::period, tok::arrow))
1479 return PrecedenceArrowAndPeriod;
1480 if ((Style.Language == FormatStyle::LK_Java ||
1481 Style.Language == FormatStyle::LK_JavaScript) &&
1482 Current->isOneOf(Keywords.kw_extends, Keywords.kw_implements,
1483 Keywords.kw_throws))
1489 void addFakeParenthesis(FormatToken *Start, prec::Level Precedence) {
1490 Start->FakeLParens.push_back(Precedence);
1491 if (Precedence > prec::Unknown)
1492 Start->StartsBinaryExpression = true;
1494 FormatToken *Previous = Current->Previous;
1495 while (Previous->is(tok::comment) && Previous->Previous)
1496 Previous = Previous->Previous;
1497 ++Previous->FakeRParens;
1498 if (Precedence > prec::Unknown)
1499 Previous->EndsBinaryExpression = true;
1503 /// \brief Parse unary operator expressions and surround them with fake
1504 /// parentheses if appropriate.
1505 void parseUnaryOperator() {
1506 if (!Current || Current->isNot(TT_UnaryOperator)) {
1507 parse(PrecedenceArrowAndPeriod);
1511 FormatToken *Start = Current;
1513 parseUnaryOperator();
1515 // The actual precedence doesn't matter.
1516 addFakeParenthesis(Start, prec::Unknown);
1519 void parseConditionalExpr() {
1520 while (Current && Current->isTrailingComment()) {
1523 FormatToken *Start = Current;
1524 parse(prec::LogicalOr);
1525 if (!Current || !Current->is(tok::question))
1528 parse(prec::Assignment);
1529 if (!Current || Current->isNot(TT_ConditionalExpr))
1532 parse(prec::Assignment);
1533 addFakeParenthesis(Start, prec::Conditional);
1536 void next(bool SkipPastLeadingComments = true) {
1538 Current = Current->Next;
1540 (Current->NewlinesBefore == 0 || SkipPastLeadingComments) &&
1541 Current->isTrailingComment())
1542 Current = Current->Next;
1545 const FormatStyle &Style;
1546 const AdditionalKeywords &Keywords;
1547 FormatToken *Current;
1550 } // end anonymous namespace
1552 void TokenAnnotator::setCommentLineLevels(
1553 SmallVectorImpl<AnnotatedLine *> &Lines) {
1554 const AnnotatedLine *NextNonCommentLine = nullptr;
1555 for (SmallVectorImpl<AnnotatedLine *>::reverse_iterator I = Lines.rbegin(),
1558 if (NextNonCommentLine && (*I)->First->is(tok::comment) &&
1559 (*I)->First->Next == nullptr)
1560 (*I)->Level = NextNonCommentLine->Level;
1562 NextNonCommentLine = (*I)->First->isNot(tok::r_brace) ? (*I) : nullptr;
1564 setCommentLineLevels((*I)->Children);
1568 void TokenAnnotator::annotate(AnnotatedLine &Line) {
1569 for (SmallVectorImpl<AnnotatedLine *>::iterator I = Line.Children.begin(),
1570 E = Line.Children.end();
1574 AnnotatingParser Parser(Style, Line, Keywords);
1575 Line.Type = Parser.parseLine();
1576 if (Line.Type == LT_Invalid)
1579 ExpressionParser ExprParser(Style, Keywords, Line);
1582 if (Line.startsWith(TT_ObjCMethodSpecifier))
1583 Line.Type = LT_ObjCMethodDecl;
1584 else if (Line.startsWith(TT_ObjCDecl))
1585 Line.Type = LT_ObjCDecl;
1586 else if (Line.startsWith(TT_ObjCProperty))
1587 Line.Type = LT_ObjCProperty;
1589 Line.First->SpacesRequiredBefore = 1;
1590 Line.First->CanBreakBefore = Line.First->MustBreakBefore;
1593 // This function heuristically determines whether 'Current' starts the name of a
1594 // function declaration.
1595 static bool isFunctionDeclarationName(const FormatToken &Current,
1596 const AnnotatedLine &Line) {
1597 auto skipOperatorName = [](const FormatToken* Next) -> const FormatToken* {
1598 for (; Next; Next = Next->Next) {
1599 if (Next->is(TT_OverloadedOperatorLParen))
1601 if (Next->is(TT_OverloadedOperator))
1603 if (Next->isOneOf(tok::kw_new, tok::kw_delete)) {
1604 // For 'new[]' and 'delete[]'.
1605 if (Next->Next && Next->Next->is(tok::l_square) &&
1606 Next->Next->Next && Next->Next->Next->is(tok::r_square))
1607 Next = Next->Next->Next;
1616 // Find parentheses of parameter list.
1617 const FormatToken *Next = Current.Next;
1618 if (Current.is(tok::kw_operator)) {
1619 if (Current.Previous && Current.Previous->is(tok::coloncolon))
1621 Next = skipOperatorName(Next);
1623 if (!Current.is(TT_StartOfName) || Current.NestingLevel != 0)
1625 for (; Next; Next = Next->Next) {
1626 if (Next->is(TT_TemplateOpener)) {
1627 Next = Next->MatchingParen;
1628 } else if (Next->is(tok::coloncolon)) {
1632 if (Next->is(tok::kw_operator)) {
1633 Next = skipOperatorName(Next->Next);
1636 if (!Next->is(tok::identifier))
1638 } else if (Next->is(tok::l_paren)) {
1646 // Check whether parameter list can be long to a function declaration.
1647 if (!Next || !Next->is(tok::l_paren) || !Next->MatchingParen)
1649 // If the lines ends with "{", this is likely an function definition.
1650 if (Line.Last->is(tok::l_brace))
1652 if (Next->Next == Next->MatchingParen)
1653 return true; // Empty parentheses.
1654 // If there is an &/&& after the r_paren, this is likely a function.
1655 if (Next->MatchingParen->Next &&
1656 Next->MatchingParen->Next->is(TT_PointerOrReference))
1658 for (const FormatToken *Tok = Next->Next; Tok && Tok != Next->MatchingParen;
1660 if (Tok->is(tok::kw_const) || Tok->isSimpleTypeSpecifier() ||
1661 Tok->isOneOf(TT_PointerOrReference, TT_StartOfName, tok::ellipsis))
1663 if (Tok->isOneOf(tok::l_brace, tok::string_literal, TT_ObjCMethodExpr) ||
1664 Tok->Tok.isLiteral())
1670 bool TokenAnnotator::mustBreakForReturnType(const AnnotatedLine &Line) const {
1671 assert(Line.MightBeFunctionDecl);
1673 if ((Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_TopLevel ||
1674 Style.AlwaysBreakAfterReturnType ==
1675 FormatStyle::RTBS_TopLevelDefinitions) &&
1679 switch (Style.AlwaysBreakAfterReturnType) {
1680 case FormatStyle::RTBS_None:
1682 case FormatStyle::RTBS_All:
1683 case FormatStyle::RTBS_TopLevel:
1685 case FormatStyle::RTBS_AllDefinitions:
1686 case FormatStyle::RTBS_TopLevelDefinitions:
1687 return Line.mightBeFunctionDefinition();
1693 void TokenAnnotator::calculateFormattingInformation(AnnotatedLine &Line) {
1694 for (SmallVectorImpl<AnnotatedLine *>::iterator I = Line.Children.begin(),
1695 E = Line.Children.end();
1697 calculateFormattingInformation(**I);
1700 Line.First->TotalLength =
1701 Line.First->IsMultiline ? Style.ColumnLimit : Line.First->ColumnWidth;
1702 if (!Line.First->Next)
1704 FormatToken *Current = Line.First->Next;
1705 bool InFunctionDecl = Line.MightBeFunctionDecl;
1707 if (isFunctionDeclarationName(*Current, Line))
1708 Current->Type = TT_FunctionDeclarationName;
1709 if (Current->is(TT_LineComment)) {
1710 if (Current->Previous->BlockKind == BK_BracedInit &&
1711 Current->Previous->opensScope())
1712 Current->SpacesRequiredBefore = Style.Cpp11BracedListStyle ? 0 : 1;
1714 Current->SpacesRequiredBefore = Style.SpacesBeforeTrailingComments;
1716 // If we find a trailing comment, iterate backwards to determine whether
1717 // it seems to relate to a specific parameter. If so, break before that
1718 // parameter to avoid changing the comment's meaning. E.g. don't move 'b'
1719 // to the previous line in:
1723 if (!Current->HasUnescapedNewline) {
1724 for (FormatToken *Parameter = Current->Previous; Parameter;
1725 Parameter = Parameter->Previous) {
1726 if (Parameter->isOneOf(tok::comment, tok::r_brace))
1728 if (Parameter->Previous && Parameter->Previous->is(tok::comma)) {
1729 if (!Parameter->Previous->is(TT_CtorInitializerComma) &&
1730 Parameter->HasUnescapedNewline)
1731 Parameter->MustBreakBefore = true;
1736 } else if (Current->SpacesRequiredBefore == 0 &&
1737 spaceRequiredBefore(Line, *Current)) {
1738 Current->SpacesRequiredBefore = 1;
1741 Current->MustBreakBefore =
1742 Current->MustBreakBefore || mustBreakBefore(Line, *Current);
1744 if (!Current->MustBreakBefore && InFunctionDecl &&
1745 Current->is(TT_FunctionDeclarationName))
1746 Current->MustBreakBefore = mustBreakForReturnType(Line);
1748 Current->CanBreakBefore =
1749 Current->MustBreakBefore || canBreakBefore(Line, *Current);
1750 unsigned ChildSize = 0;
1751 if (Current->Previous->Children.size() == 1) {
1752 FormatToken &LastOfChild = *Current->Previous->Children[0]->Last;
1753 ChildSize = LastOfChild.isTrailingComment() ? Style.ColumnLimit
1754 : LastOfChild.TotalLength + 1;
1756 const FormatToken *Prev = Current->Previous;
1757 if (Current->MustBreakBefore || Prev->Children.size() > 1 ||
1758 (Prev->Children.size() == 1 &&
1759 Prev->Children[0]->First->MustBreakBefore) ||
1760 Current->IsMultiline)
1761 Current->TotalLength = Prev->TotalLength + Style.ColumnLimit;
1763 Current->TotalLength = Prev->TotalLength + Current->ColumnWidth +
1764 ChildSize + Current->SpacesRequiredBefore;
1766 if (Current->is(TT_CtorInitializerColon))
1767 InFunctionDecl = false;
1769 // FIXME: Only calculate this if CanBreakBefore is true once static
1770 // initializers etc. are sorted out.
1771 // FIXME: Move magic numbers to a better place.
1772 Current->SplitPenalty = 20 * Current->BindingStrength +
1773 splitPenalty(Line, *Current, InFunctionDecl);
1775 Current = Current->Next;
1778 calculateUnbreakableTailLengths(Line);
1779 for (Current = Line.First; Current != nullptr; Current = Current->Next) {
1781 Current->Role->precomputeFormattingInfos(Current);
1784 DEBUG({ printDebugInfo(Line); });
1787 void TokenAnnotator::calculateUnbreakableTailLengths(AnnotatedLine &Line) {
1788 unsigned UnbreakableTailLength = 0;
1789 FormatToken *Current = Line.Last;
1791 Current->UnbreakableTailLength = UnbreakableTailLength;
1792 if (Current->CanBreakBefore ||
1793 Current->isOneOf(tok::comment, tok::string_literal)) {
1794 UnbreakableTailLength = 0;
1796 UnbreakableTailLength +=
1797 Current->ColumnWidth + Current->SpacesRequiredBefore;
1799 Current = Current->Previous;
1803 unsigned TokenAnnotator::splitPenalty(const AnnotatedLine &Line,
1804 const FormatToken &Tok,
1805 bool InFunctionDecl) {
1806 const FormatToken &Left = *Tok.Previous;
1807 const FormatToken &Right = Tok;
1809 if (Left.is(tok::semi))
1812 if (Style.Language == FormatStyle::LK_Java) {
1813 if (Right.isOneOf(Keywords.kw_extends, Keywords.kw_throws))
1815 if (Right.is(Keywords.kw_implements))
1817 if (Left.is(tok::comma) && Left.NestingLevel == 0)
1819 } else if (Style.Language == FormatStyle::LK_JavaScript) {
1820 if (Right.is(Keywords.kw_function) && Left.isNot(tok::comma))
1822 if (Left.is(TT_JsTypeColon))
1824 if ((Left.is(TT_TemplateString) && Left.TokenText.endswith("${")) ||
1825 (Right.is(TT_TemplateString) && Right.TokenText.startswith("}")))
1829 if (Left.is(tok::comma) || (Right.is(tok::identifier) && Right.Next &&
1830 Right.Next->is(TT_DictLiteral)))
1832 if (Right.is(tok::l_square)) {
1833 if (Style.Language == FormatStyle::LK_Proto)
1835 if (Left.is(tok::r_square))
1837 // Slightly prefer formatting local lambda definitions like functions.
1838 if (Right.is(TT_LambdaLSquare) && Left.is(tok::equal))
1840 if (!Right.isOneOf(TT_ObjCMethodExpr, TT_LambdaLSquare,
1841 TT_ArrayInitializerLSquare))
1845 if (Right.isOneOf(TT_StartOfName, TT_FunctionDeclarationName) ||
1846 Right.is(tok::kw_operator)) {
1847 if (Line.startsWith(tok::kw_for) && Right.PartOfMultiVariableDeclStmt)
1849 if (Left.is(TT_StartOfName))
1851 if (InFunctionDecl && Right.NestingLevel == 0)
1852 return Style.PenaltyReturnTypeOnItsOwnLine;
1855 if (Right.is(TT_PointerOrReference))
1857 if (Right.is(TT_LambdaArrow))
1859 if (Left.is(tok::equal) && Right.is(tok::l_brace))
1861 if (Left.is(TT_CastRParen))
1863 if (Left.is(tok::coloncolon) ||
1864 (Right.is(tok::period) && Style.Language == FormatStyle::LK_Proto))
1866 if (Left.isOneOf(tok::kw_class, tok::kw_struct))
1868 if (Left.is(tok::comment))
1871 if (Left.isOneOf(TT_RangeBasedForLoopColon, TT_InheritanceColon))
1874 if (Right.isMemberAccess()) {
1875 // Breaking before the "./->" of a chained call/member access is reasonably
1876 // cheap, as formatting those with one call per line is generally
1877 // desirable. In particular, it should be cheaper to break before the call
1878 // than it is to break inside a call's parameters, which could lead to weird
1879 // "hanging" indents. The exception is the very last "./->" to support this
1880 // frequent pattern:
1882 // aaaaaaaa.aaaaaaaa.bbbbbbb().ccccccccccccccccccccc(
1885 // which might otherwise be blown up onto many lines. Here, clang-format
1886 // won't produce "hanging" indents anyway as there is no other trailing
1889 // Also apply higher penalty is not a call as that might lead to a wrapping
1893 // .aaaaaaaaa.bbbbbbbb(cccccccc);
1894 return !Right.NextOperator || !Right.NextOperator->Previous->closesScope()
1899 if (Right.is(TT_TrailingAnnotation) &&
1900 (!Right.Next || Right.Next->isNot(tok::l_paren))) {
1901 // Moving trailing annotations to the next line is fine for ObjC method
1903 if (Line.startsWith(TT_ObjCMethodSpecifier))
1905 // Generally, breaking before a trailing annotation is bad unless it is
1906 // function-like. It seems to be especially preferable to keep standard
1907 // annotations (i.e. "const", "final" and "override") on the same line.
1908 // Use a slightly higher penalty after ")" so that annotations like
1909 // "const override" are kept together.
1910 bool is_short_annotation = Right.TokenText.size() < 10;
1911 return (Left.is(tok::r_paren) ? 100 : 120) + (is_short_annotation ? 50 : 0);
1914 // In for-loops, prefer breaking at ',' and ';'.
1915 if (Line.startsWith(tok::kw_for) && Left.is(tok::equal))
1918 // In Objective-C method expressions, prefer breaking before "param:" over
1919 // breaking after it.
1920 if (Right.is(TT_SelectorName))
1922 if (Left.is(tok::colon) && Left.is(TT_ObjCMethodExpr))
1923 return Line.MightBeFunctionDecl ? 50 : 500;
1925 if (Left.is(tok::l_paren) && InFunctionDecl &&
1926 Style.AlignAfterOpenBracket != FormatStyle::BAS_DontAlign)
1928 if (Left.is(tok::l_paren) && Left.Previous &&
1929 Left.Previous->isOneOf(tok::kw_if, tok::kw_for))
1931 if (Left.is(tok::equal) && InFunctionDecl)
1933 if (Right.is(tok::r_brace))
1935 if (Left.is(TT_TemplateOpener))
1937 if (Left.opensScope()) {
1938 if (Style.AlignAfterOpenBracket == FormatStyle::BAS_DontAlign)
1940 return Left.ParameterCount > 1 ? Style.PenaltyBreakBeforeFirstCallParameter
1943 if (Left.is(TT_JavaAnnotation))
1946 if (Right.is(tok::lessless)) {
1947 if (Left.is(tok::string_literal) &&
1948 (Right.NextOperator || Right.OperatorIndex != 1)) {
1949 StringRef Content = Left.TokenText;
1950 if (Content.startswith("\""))
1951 Content = Content.drop_front(1);
1952 if (Content.endswith("\""))
1953 Content = Content.drop_back(1);
1954 Content = Content.trim();
1955 if (Content.size() > 1 &&
1956 (Content.back() == ':' || Content.back() == '='))
1959 return 1; // Breaking at a << is really cheap.
1961 if (Left.is(TT_ConditionalExpr))
1962 return prec::Conditional;
1963 prec::Level Level = Left.getPrecedence();
1964 if (Level != prec::Unknown)
1966 Level = Right.getPrecedence();
1967 if (Level != prec::Unknown)
1973 bool TokenAnnotator::spaceRequiredBetween(const AnnotatedLine &Line,
1974 const FormatToken &Left,
1975 const FormatToken &Right) {
1976 if (Left.is(tok::kw_return) && Right.isNot(tok::semi))
1978 if (Style.ObjCSpaceAfterProperty && Line.Type == LT_ObjCProperty &&
1979 Left.Tok.getObjCKeywordID() == tok::objc_property)
1981 if (Right.is(tok::hashhash))
1982 return Left.is(tok::hash);
1983 if (Left.isOneOf(tok::hashhash, tok::hash))
1984 return Right.is(tok::hash);
1985 if (Left.is(tok::l_paren) && Right.is(tok::r_paren))
1986 return Style.SpaceInEmptyParentheses;
1987 if (Left.is(tok::l_paren) || Right.is(tok::r_paren))
1988 return (Right.is(TT_CastRParen) ||
1989 (Left.MatchingParen && Left.MatchingParen->is(TT_CastRParen)))
1990 ? Style.SpacesInCStyleCastParentheses
1991 : Style.SpacesInParentheses;
1992 if (Right.isOneOf(tok::semi, tok::comma))
1994 if (Right.is(tok::less) &&
1995 Line.Type == LT_ObjCDecl && Style.ObjCSpaceBeforeProtocolList)
1997 if (Right.is(tok::less) && Left.is(tok::kw_template))
1998 return Style.SpaceAfterTemplateKeyword;
1999 if (Left.isOneOf(tok::exclaim, tok::tilde))
2001 if (Left.is(tok::at) &&
2002 Right.isOneOf(tok::identifier, tok::string_literal, tok::char_constant,
2003 tok::numeric_constant, tok::l_paren, tok::l_brace,
2004 tok::kw_true, tok::kw_false))
2006 if (Left.is(tok::colon))
2007 return !Left.is(TT_ObjCMethodExpr);
2008 if (Left.is(tok::coloncolon))
2010 if (Left.is(tok::less) || Right.isOneOf(tok::greater, tok::less))
2012 if (Right.is(tok::ellipsis))
2013 return Left.Tok.isLiteral() || (Left.is(tok::identifier) && Left.Previous &&
2014 Left.Previous->is(tok::kw_case));
2015 if (Left.is(tok::l_square) && Right.is(tok::amp))
2017 if (Right.is(TT_PointerOrReference))
2018 return (Left.is(tok::r_paren) && Line.MightBeFunctionDecl) ||
2019 (Left.Tok.isLiteral() || (Left.is(tok::kw_const) && Left.Previous &&
2020 Left.Previous->is(tok::r_paren)) ||
2021 (!Left.isOneOf(TT_PointerOrReference, tok::l_paren) &&
2022 (Style.PointerAlignment != FormatStyle::PAS_Left ||
2023 (Line.IsMultiVariableDeclStmt && Left.NestingLevel == 0))));
2024 if (Right.is(TT_FunctionTypeLParen) && Left.isNot(tok::l_paren) &&
2025 (!Left.is(TT_PointerOrReference) ||
2026 (Style.PointerAlignment != FormatStyle::PAS_Right &&
2027 !Line.IsMultiVariableDeclStmt)))
2029 if (Left.is(TT_PointerOrReference))
2030 return Right.Tok.isLiteral() ||
2031 Right.isOneOf(TT_BlockComment, Keywords.kw_final,
2032 Keywords.kw_override) ||
2033 (Right.is(tok::l_brace) && Right.BlockKind == BK_Block) ||
2034 (!Right.isOneOf(TT_PointerOrReference, TT_ArraySubscriptLSquare,
2036 (Style.PointerAlignment != FormatStyle::PAS_Right &&
2037 !Line.IsMultiVariableDeclStmt) &&
2039 !Left.Previous->isOneOf(tok::l_paren, tok::coloncolon));
2040 if (Right.is(tok::star) && Left.is(tok::l_paren))
2042 if (Left.is(tok::l_square))
2043 return (Left.is(TT_ArrayInitializerLSquare) &&
2044 Style.SpacesInContainerLiterals && Right.isNot(tok::r_square)) ||
2045 (Left.is(TT_ArraySubscriptLSquare) && Style.SpacesInSquareBrackets &&
2046 Right.isNot(tok::r_square));
2047 if (Right.is(tok::r_square))
2048 return Right.MatchingParen &&
2049 ((Style.SpacesInContainerLiterals &&
2050 Right.MatchingParen->is(TT_ArrayInitializerLSquare)) ||
2051 (Style.SpacesInSquareBrackets &&
2052 Right.MatchingParen->is(TT_ArraySubscriptLSquare)));
2053 if (Right.is(tok::l_square) &&
2054 !Right.isOneOf(TT_ObjCMethodExpr, TT_LambdaLSquare) &&
2055 !Left.isOneOf(tok::numeric_constant, TT_DictLiteral))
2057 if (Left.is(tok::l_brace) && Right.is(tok::r_brace))
2058 return !Left.Children.empty(); // No spaces in "{}".
2059 if ((Left.is(tok::l_brace) && Left.BlockKind != BK_Block) ||
2060 (Right.is(tok::r_brace) && Right.MatchingParen &&
2061 Right.MatchingParen->BlockKind != BK_Block))
2062 return !Style.Cpp11BracedListStyle;
2063 if (Left.is(TT_BlockComment))
2064 return !Left.TokenText.endswith("=*/");
2065 if (Right.is(tok::l_paren)) {
2066 if (Left.is(tok::r_paren) && Left.is(TT_AttributeParen))
2068 return Line.Type == LT_ObjCDecl || Left.is(tok::semi) ||
2069 (Style.SpaceBeforeParens != FormatStyle::SBPO_Never &&
2070 (Left.isOneOf(tok::kw_if, tok::pp_elif, tok::kw_for, tok::kw_while,
2071 tok::kw_switch, tok::kw_case, TT_ForEachMacro,
2073 (Left.isOneOf(tok::kw_try, Keywords.kw___except, tok::kw_catch,
2074 tok::kw_new, tok::kw_delete) &&
2075 (!Left.Previous || Left.Previous->isNot(tok::period))))) ||
2076 (Style.SpaceBeforeParens == FormatStyle::SBPO_Always &&
2077 (Left.is(tok::identifier) || Left.isFunctionLikeKeyword() ||
2078 Left.is(tok::r_paren)) &&
2079 Line.Type != LT_PreprocessorDirective);
2081 if (Left.is(tok::at) && Right.Tok.getObjCKeywordID() != tok::objc_not_keyword)
2083 if (Right.is(TT_UnaryOperator))
2084 return !Left.isOneOf(tok::l_paren, tok::l_square, tok::at) &&
2085 (Left.isNot(tok::colon) || Left.isNot(TT_ObjCMethodExpr));
2086 if ((Left.isOneOf(tok::identifier, tok::greater, tok::r_square,
2088 Left.isSimpleTypeSpecifier()) &&
2089 Right.is(tok::l_brace) && Right.getNextNonComment() &&
2090 Right.BlockKind != BK_Block)
2092 if (Left.is(tok::period) || Right.is(tok::period))
2094 if (Right.is(tok::hash) && Left.is(tok::identifier) && Left.TokenText == "L")
2096 if (Left.is(TT_TemplateCloser) && Left.MatchingParen &&
2097 Left.MatchingParen->Previous &&
2098 Left.MatchingParen->Previous->is(tok::period))
2099 // A.<B>DoSomething();
2101 if (Left.is(TT_TemplateCloser) && Right.is(tok::l_square))
2106 bool TokenAnnotator::spaceRequiredBefore(const AnnotatedLine &Line,
2107 const FormatToken &Right) {
2108 const FormatToken &Left = *Right.Previous;
2109 if (Right.Tok.getIdentifierInfo() && Left.Tok.getIdentifierInfo())
2110 return true; // Never ever merge two identifiers.
2111 if (Style.Language == FormatStyle::LK_Cpp) {
2112 if (Left.is(tok::kw_operator))
2113 return Right.is(tok::coloncolon);
2114 } else if (Style.Language == FormatStyle::LK_Proto) {
2115 if (Right.is(tok::period) &&
2116 Left.isOneOf(Keywords.kw_optional, Keywords.kw_required,
2117 Keywords.kw_repeated, Keywords.kw_extend))
2119 if (Right.is(tok::l_paren) &&
2120 Left.isOneOf(Keywords.kw_returns, Keywords.kw_option))
2122 } else if (Style.Language == FormatStyle::LK_JavaScript) {
2123 if (Left.is(TT_JsFatArrow))
2125 if ((Left.is(TT_TemplateString) && Left.TokenText.endswith("${")) ||
2126 (Right.is(TT_TemplateString) && Right.TokenText.startswith("}")))
2128 if (Left.is(tok::identifier) && Right.is(TT_TemplateString))
2130 if (Right.is(tok::star) &&
2131 Left.isOneOf(Keywords.kw_function, Keywords.kw_yield))
2133 // JS methods can use some keywords as names (e.g. `delete()`).
2134 if (Right.is(tok::l_paren) && Line.MustBeDeclaration &&
2135 Left.Tok.getIdentifierInfo())
2137 if (Left.isOneOf(Keywords.kw_let, Keywords.kw_var, Keywords.kw_in,
2138 Keywords.kw_of, tok::kw_const) &&
2139 (!Left.Previous || !Left.Previous->is(tok::period)))
2141 if (Left.is(Keywords.kw_as) &&
2142 Right.isOneOf(tok::l_square, tok::l_brace, tok::l_paren))
2144 if (Left.is(tok::kw_default) && Left.Previous &&
2145 Left.Previous->is(tok::kw_export))
2147 if (Left.is(Keywords.kw_is) && Right.is(tok::l_brace))
2149 if (Right.isOneOf(TT_JsTypeColon, TT_JsTypeOptionalQuestion))
2151 if (Left.is(TT_JsTypeOperator) || Right.is(TT_JsTypeOperator))
2153 if ((Left.is(tok::l_brace) || Right.is(tok::r_brace)) &&
2154 Line.First->isOneOf(Keywords.kw_import, tok::kw_export))
2156 if (Left.is(tok::ellipsis))
2158 if (Left.is(TT_TemplateCloser) &&
2159 !Right.isOneOf(tok::equal, tok::l_brace, tok::comma, tok::l_square,
2160 Keywords.kw_implements, Keywords.kw_extends))
2161 // Type assertions ('<type>expr') are not followed by whitespace. Other
2162 // locations that should have whitespace following are identified by the
2163 // above set of follower tokens.
2165 // Postfix non-null assertion operator, as in `foo!.bar()`.
2166 if (Right.is(tok::exclaim) && (Left.isOneOf(tok::identifier, tok::r_paren,
2167 tok::r_square, tok::r_brace) ||
2168 Left.Tok.isLiteral()))
2170 if (Left.is(tok::exclaim) && Right.is(Keywords.kw_as))
2171 return true; // "x! as string"
2172 } else if (Style.Language == FormatStyle::LK_Java) {
2173 if (Left.is(tok::r_square) && Right.is(tok::l_brace))
2175 if (Left.is(Keywords.kw_synchronized) && Right.is(tok::l_paren))
2176 return Style.SpaceBeforeParens != FormatStyle::SBPO_Never;
2177 if ((Left.isOneOf(tok::kw_static, tok::kw_public, tok::kw_private,
2178 tok::kw_protected) ||
2179 Left.isOneOf(Keywords.kw_final, Keywords.kw_abstract,
2180 Keywords.kw_native)) &&
2181 Right.is(TT_TemplateOpener))
2184 if (Left.is(TT_ImplicitStringLiteral))
2185 return Right.WhitespaceRange.getBegin() != Right.WhitespaceRange.getEnd();
2186 if (Line.Type == LT_ObjCMethodDecl) {
2187 if (Left.is(TT_ObjCMethodSpecifier))
2189 if (Left.is(tok::r_paren) && Right.is(tok::identifier))
2190 // Don't space between ')' and <id>
2193 if (Line.Type == LT_ObjCProperty &&
2194 (Right.is(tok::equal) || Left.is(tok::equal)))
2197 if (Right.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow) ||
2198 Left.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow))
2200 if (Right.is(TT_OverloadedOperatorLParen))
2201 return Style.SpaceBeforeParens == FormatStyle::SBPO_Always;
2202 if (Left.is(tok::comma))
2204 if (Right.is(tok::comma))
2206 if (Right.isOneOf(TT_CtorInitializerColon, TT_ObjCBlockLParen))
2208 if (Right.is(tok::colon)) {
2209 if (Line.First->isOneOf(tok::kw_case, tok::kw_default) ||
2210 !Right.getNextNonComment() || Right.getNextNonComment()->is(tok::semi))
2212 if (Right.is(TT_ObjCMethodExpr))
2214 if (Left.is(tok::question))
2216 if (Right.is(TT_InlineASMColon) && Left.is(tok::coloncolon))
2218 if (Right.is(TT_DictLiteral))
2219 return Style.SpacesInContainerLiterals;
2222 if (Left.is(TT_UnaryOperator))
2223 return Right.is(TT_BinaryOperator);
2225 // If the next token is a binary operator or a selector name, we have
2226 // incorrectly classified the parenthesis as a cast. FIXME: Detect correctly.
2227 if (Left.is(TT_CastRParen))
2228 return Style.SpaceAfterCStyleCast ||
2229 Right.isOneOf(TT_BinaryOperator, TT_SelectorName);
2231 if (Left.is(tok::greater) && Right.is(tok::greater))
2232 return Right.is(TT_TemplateCloser) && Left.is(TT_TemplateCloser) &&
2233 (Style.Standard != FormatStyle::LS_Cpp11 || Style.SpacesInAngles);
2234 if (Right.isOneOf(tok::arrow, tok::period, tok::arrowstar, tok::periodstar) ||
2235 Left.isOneOf(tok::arrow, tok::period, tok::arrowstar, tok::periodstar))
2237 if (!Style.SpaceBeforeAssignmentOperators &&
2238 Right.getPrecedence() == prec::Assignment)
2240 if (Right.is(tok::coloncolon) && !Left.isOneOf(tok::l_brace, tok::comment))
2241 return (Left.is(TT_TemplateOpener) &&
2242 Style.Standard == FormatStyle::LS_Cpp03) ||
2243 !(Left.isOneOf(tok::identifier, tok::l_paren, tok::r_paren,
2245 Left.isOneOf(TT_TemplateCloser, TT_TemplateOpener));
2246 if ((Left.is(TT_TemplateOpener)) != (Right.is(TT_TemplateCloser)))
2247 return Style.SpacesInAngles;
2248 if ((Right.is(TT_BinaryOperator) && !Left.is(tok::l_paren)) ||
2249 (Left.isOneOf(TT_BinaryOperator, TT_ConditionalExpr) &&
2250 !Right.is(tok::r_paren)))
2252 if (Left.is(TT_TemplateCloser) && Right.is(tok::l_paren) &&
2253 Right.isNot(TT_FunctionTypeLParen))
2254 return Style.SpaceBeforeParens == FormatStyle::SBPO_Always;
2255 if (Right.is(TT_TemplateOpener) && Left.is(tok::r_paren) &&
2256 Left.MatchingParen && Left.MatchingParen->is(TT_OverloadedOperatorLParen))
2258 if (Right.is(tok::less) && Left.isNot(tok::l_paren) &&
2259 Line.startsWith(tok::hash))
2261 if (Right.is(TT_TrailingUnaryOperator))
2263 if (Left.is(TT_RegexLiteral))
2265 return spaceRequiredBetween(Line, Left, Right);
2268 // Returns 'true' if 'Tok' is a brace we'd want to break before in Allman style.
2269 static bool isAllmanBrace(const FormatToken &Tok) {
2270 return Tok.is(tok::l_brace) && Tok.BlockKind == BK_Block &&
2271 !Tok.isOneOf(TT_ObjCBlockLBrace, TT_DictLiteral);
2274 bool TokenAnnotator::mustBreakBefore(const AnnotatedLine &Line,
2275 const FormatToken &Right) {
2276 const FormatToken &Left = *Right.Previous;
2277 if (Right.NewlinesBefore > 1 && Style.MaxEmptyLinesToKeep > 0)
2280 if (Style.Language == FormatStyle::LK_JavaScript) {
2281 // FIXME: This might apply to other languages and token kinds.
2282 if (Right.is(tok::string_literal) && Left.is(tok::plus) && Left.Previous &&
2283 Left.Previous->is(tok::string_literal))
2285 if (Left.is(TT_DictLiteral) && Left.is(tok::l_brace) && Line.Level == 0 &&
2286 Left.Previous && Left.Previous->is(tok::equal) &&
2287 Line.First->isOneOf(tok::identifier, Keywords.kw_import, tok::kw_export,
2289 // kw_var/kw_let are pseudo-tokens that are tok::identifier, so match
2291 !Line.First->isOneOf(Keywords.kw_var, Keywords.kw_let))
2292 // Object literals on the top level of a file are treated as "enum-style".
2293 // Each key/value pair is put on a separate line, instead of bin-packing.
2295 if (Left.is(tok::l_brace) && Line.Level == 0 &&
2296 (Line.startsWith(tok::kw_enum) ||
2297 Line.startsWith(tok::kw_export, tok::kw_enum)))
2298 // JavaScript top-level enum key/value pairs are put on separate lines
2299 // instead of bin-packing.
2301 if (Right.is(tok::r_brace) && Left.is(tok::l_brace) &&
2302 !Left.Children.empty())
2303 // Support AllowShortFunctionsOnASingleLine for JavaScript.
2304 return Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_None ||
2305 Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_Empty ||
2306 (Left.NestingLevel == 0 && Line.Level == 0 &&
2307 Style.AllowShortFunctionsOnASingleLine ==
2308 FormatStyle::SFS_Inline);
2309 } else if (Style.Language == FormatStyle::LK_Java) {
2310 if (Right.is(tok::plus) && Left.is(tok::string_literal) && Right.Next &&
2311 Right.Next->is(tok::string_literal))
2315 // If the last token before a '}' is a comma or a trailing comment, the
2316 // intention is to insert a line break after it in order to make shuffling
2317 // around entries easier.
2318 const FormatToken *BeforeClosingBrace = nullptr;
2319 if (Left.isOneOf(tok::l_brace, TT_ArrayInitializerLSquare) &&
2320 Left.BlockKind != BK_Block && Left.MatchingParen)
2321 BeforeClosingBrace = Left.MatchingParen->Previous;
2322 else if (Right.MatchingParen &&
2323 Right.MatchingParen->isOneOf(tok::l_brace,
2324 TT_ArrayInitializerLSquare))
2325 BeforeClosingBrace = &Left;
2326 if (BeforeClosingBrace && (BeforeClosingBrace->is(tok::comma) ||
2327 BeforeClosingBrace->isTrailingComment()))
2330 if (Right.is(tok::comment))
2331 return Left.BlockKind != BK_BracedInit &&
2332 Left.isNot(TT_CtorInitializerColon) &&
2333 (Right.NewlinesBefore > 0 && Right.HasUnescapedNewline);
2334 if (Left.isTrailingComment())
2336 if (Left.isStringLiteral() &&
2337 (Right.isStringLiteral() || Right.is(TT_ObjCStringLiteral)))
2339 if (Right.Previous->IsUnterminatedLiteral)
2341 if (Right.is(tok::lessless) && Right.Next &&
2342 Right.Previous->is(tok::string_literal) &&
2343 Right.Next->is(tok::string_literal))
2345 if (Right.Previous->ClosesTemplateDeclaration &&
2346 Right.Previous->MatchingParen &&
2347 Right.Previous->MatchingParen->NestingLevel == 0 &&
2348 Style.AlwaysBreakTemplateDeclarations)
2350 if ((Right.isOneOf(TT_CtorInitializerComma, TT_CtorInitializerColon)) &&
2351 Style.BreakConstructorInitializersBeforeComma &&
2352 !Style.ConstructorInitializerAllOnOneLineOrOnePerLine)
2354 if (Right.is(tok::string_literal) && Right.TokenText.startswith("R\""))
2355 // Raw string literals are special wrt. line breaks. The author has made a
2356 // deliberate choice and might have aligned the contents of the string
2357 // literal accordingly. Thus, we try keep existing line breaks.
2358 return Right.NewlinesBefore > 0;
2359 if (Right.Previous->is(tok::l_brace) && Right.NestingLevel == 1 &&
2360 Style.Language == FormatStyle::LK_Proto)
2361 // Don't put enums onto single lines in protocol buffers.
2363 if (Right.is(TT_InlineASMBrace))
2364 return Right.HasUnescapedNewline;
2365 if (isAllmanBrace(Left) || isAllmanBrace(Right))
2366 return (Line.startsWith(tok::kw_enum) && Style.BraceWrapping.AfterEnum) ||
2367 (Line.startsWith(tok::kw_class) && Style.BraceWrapping.AfterClass) ||
2368 (Line.startsWith(tok::kw_struct) && Style.BraceWrapping.AfterStruct);
2369 if (Left.is(TT_ObjCBlockLBrace) && !Style.AllowShortBlocksOnASingleLine)
2372 if ((Style.Language == FormatStyle::LK_Java ||
2373 Style.Language == FormatStyle::LK_JavaScript) &&
2374 Left.is(TT_LeadingJavaAnnotation) &&
2375 Right.isNot(TT_LeadingJavaAnnotation) && Right.isNot(tok::l_paren) &&
2376 (Line.Last->is(tok::l_brace) || Style.BreakAfterJavaFieldAnnotations))
2382 bool TokenAnnotator::canBreakBefore(const AnnotatedLine &Line,
2383 const FormatToken &Right) {
2384 const FormatToken &Left = *Right.Previous;
2386 // Language-specific stuff.
2387 if (Style.Language == FormatStyle::LK_Java) {
2388 if (Left.isOneOf(Keywords.kw_throws, Keywords.kw_extends,
2389 Keywords.kw_implements))
2391 if (Right.isOneOf(Keywords.kw_throws, Keywords.kw_extends,
2392 Keywords.kw_implements))
2394 } else if (Style.Language == FormatStyle::LK_JavaScript) {
2395 const FormatToken *NonComment = Right.getPreviousNonComment();
2396 if (Left.isOneOf(tok::kw_return, tok::kw_continue, tok::kw_break,
2399 NonComment->isOneOf(tok::kw_return, tok::kw_continue, tok::kw_break,
2401 return false; // Otherwise a semicolon is inserted.
2402 if (Left.is(TT_JsFatArrow) && Right.is(tok::l_brace))
2404 if (Left.is(TT_JsTypeColon))
2406 if (Right.NestingLevel == 0 && Right.is(Keywords.kw_is))
2408 if (Left.is(Keywords.kw_in))
2409 return Style.BreakBeforeBinaryOperators == FormatStyle::BOS_None;
2410 if (Right.is(Keywords.kw_in))
2411 return Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None;
2412 if (Right.is(Keywords.kw_as))
2413 return false; // must not break before as in 'x as type' casts
2416 if (Left.is(tok::at))
2418 if (Left.Tok.getObjCKeywordID() == tok::objc_interface)
2420 if (Left.isOneOf(TT_JavaAnnotation, TT_LeadingJavaAnnotation))
2421 return !Right.is(tok::l_paren);
2422 if (Right.is(TT_PointerOrReference))
2423 return Line.IsMultiVariableDeclStmt ||
2424 (Style.PointerAlignment == FormatStyle::PAS_Right &&
2425 (!Right.Next || Right.Next->isNot(TT_FunctionDeclarationName)));
2426 if (Right.isOneOf(TT_StartOfName, TT_FunctionDeclarationName) ||
2427 Right.is(tok::kw_operator))
2429 if (Left.is(TT_PointerOrReference))
2431 if (Right.isTrailingComment())
2432 // We rely on MustBreakBefore being set correctly here as we should not
2433 // change the "binding" behavior of a comment.
2434 // The first comment in a braced lists is always interpreted as belonging to
2435 // the first list element. Otherwise, it should be placed outside of the
2437 return Left.BlockKind == BK_BracedInit;
2438 if (Left.is(tok::question) && Right.is(tok::colon))
2440 if (Right.is(TT_ConditionalExpr) || Right.is(tok::question))
2441 return Style.BreakBeforeTernaryOperators;
2442 if (Left.is(TT_ConditionalExpr) || Left.is(tok::question))
2443 return !Style.BreakBeforeTernaryOperators;
2444 if (Right.is(TT_InheritanceColon))
2446 if (Right.is(tok::colon) &&
2447 !Right.isOneOf(TT_CtorInitializerColon, TT_InlineASMColon))
2449 if (Left.is(tok::colon) && (Left.isOneOf(TT_DictLiteral, TT_ObjCMethodExpr)))
2451 if (Right.is(TT_SelectorName) || (Right.is(tok::identifier) && Right.Next &&
2452 Right.Next->is(TT_ObjCMethodExpr)))
2453 return Left.isNot(tok::period); // FIXME: Properly parse ObjC calls.
2454 if (Left.is(tok::r_paren) && Line.Type == LT_ObjCProperty)
2456 if (Left.ClosesTemplateDeclaration || Left.is(TT_FunctionAnnotationRParen))
2458 if (Right.isOneOf(TT_RangeBasedForLoopColon, TT_OverloadedOperatorLParen,
2459 TT_OverloadedOperator))
2461 if (Left.is(TT_RangeBasedForLoopColon))
2463 if (Right.is(TT_RangeBasedForLoopColon))
2465 if (Left.isOneOf(TT_TemplateCloser, TT_UnaryOperator) ||
2466 Left.is(tok::kw_operator))
2468 if (Left.is(tok::equal) && !Right.isOneOf(tok::kw_default, tok::kw_delete) &&
2469 Line.Type == LT_VirtualFunctionDecl && Left.NestingLevel == 0)
2471 if (Left.is(tok::l_paren) && Left.is(TT_AttributeParen))
2473 if (Left.is(tok::l_paren) && Left.Previous &&
2474 (Left.Previous->isOneOf(TT_BinaryOperator, TT_CastRParen)))
2476 if (Right.is(TT_ImplicitStringLiteral))
2479 if (Right.is(tok::r_paren) || Right.is(TT_TemplateCloser))
2481 if (Right.is(tok::r_square) && Right.MatchingParen &&
2482 Right.MatchingParen->is(TT_LambdaLSquare))
2485 // We only break before r_brace if there was a corresponding break before
2486 // the l_brace, which is tracked by BreakBeforeClosingBrace.
2487 if (Right.is(tok::r_brace))
2488 return Right.MatchingParen && Right.MatchingParen->BlockKind == BK_Block;
2490 // Allow breaking after a trailing annotation, e.g. after a method
2492 if (Left.is(TT_TrailingAnnotation))
2493 return !Right.isOneOf(tok::l_brace, tok::semi, tok::equal, tok::l_paren,
2494 tok::less, tok::coloncolon);
2496 if (Right.is(tok::kw___attribute))
2499 if (Left.is(tok::identifier) && Right.is(tok::string_literal))
2502 if (Right.is(tok::identifier) && Right.Next && Right.Next->is(TT_DictLiteral))
2505 if (Left.is(TT_CtorInitializerComma) &&
2506 Style.BreakConstructorInitializersBeforeComma)
2508 if (Right.is(TT_CtorInitializerComma) &&
2509 Style.BreakConstructorInitializersBeforeComma)
2511 if ((Left.is(tok::greater) && Right.is(tok::greater)) ||
2512 (Left.is(tok::less) && Right.is(tok::less)))
2514 if (Right.is(TT_BinaryOperator) &&
2515 Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None &&
2516 (Style.BreakBeforeBinaryOperators == FormatStyle::BOS_All ||
2517 Right.getPrecedence() != prec::Assignment))
2519 if (Left.is(TT_ArrayInitializerLSquare))
2521 if (Right.is(tok::kw_typename) && Left.isNot(tok::kw_const))
2523 if ((Left.isBinaryOperator() || Left.is(TT_BinaryOperator)) &&
2524 !Left.isOneOf(tok::arrowstar, tok::lessless) &&
2525 Style.BreakBeforeBinaryOperators != FormatStyle::BOS_All &&
2526 (Style.BreakBeforeBinaryOperators == FormatStyle::BOS_None ||
2527 Left.getPrecedence() == prec::Assignment))
2529 return Left.isOneOf(tok::comma, tok::coloncolon, tok::semi, tok::l_brace,
2530 tok::kw_class, tok::kw_struct, tok::comment) ||
2531 Right.isMemberAccess() ||
2532 Right.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow, tok::lessless,
2533 tok::colon, tok::l_square, tok::at) ||
2534 (Left.is(tok::r_paren) &&
2535 Right.isOneOf(tok::identifier, tok::kw_const)) ||
2536 (Left.is(tok::l_paren) && !Right.is(tok::r_paren));
2539 void TokenAnnotator::printDebugInfo(const AnnotatedLine &Line) {
2540 llvm::errs() << "AnnotatedTokens:\n";
2541 const FormatToken *Tok = Line.First;
2543 llvm::errs() << " M=" << Tok->MustBreakBefore
2544 << " C=" << Tok->CanBreakBefore
2545 << " T=" << getTokenTypeName(Tok->Type)
2546 << " S=" << Tok->SpacesRequiredBefore
2547 << " B=" << Tok->BlockParameterCount
2548 << " P=" << Tok->SplitPenalty << " Name=" << Tok->Tok.getName()
2549 << " L=" << Tok->TotalLength << " PPK=" << Tok->PackingKind
2551 for (unsigned i = 0, e = Tok->FakeLParens.size(); i != e; ++i)
2552 llvm::errs() << Tok->FakeLParens[i] << "/";
2553 llvm::errs() << " FakeRParens=" << Tok->FakeRParens << "\n";
2555 assert(Tok == Line.Last);
2558 llvm::errs() << "----\n";
2561 } // namespace format
2562 } // namespace clang