1 //===--- TokenAnnotator.cpp - Format C++ code -----------------------------===//
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //===----------------------------------------------------------------------===//
10 /// This file implements a token annotator, i.e. creates
11 /// \c AnnotatedTokens out of \c FormatTokens with required extra information.
13 //===----------------------------------------------------------------------===//
15 #include "TokenAnnotator.h"
16 #include "FormatToken.h"
17 #include "clang/Basic/SourceManager.h"
18 #include "clang/Basic/TokenKinds.h"
19 #include "llvm/ADT/SmallPtrSet.h"
20 #include "llvm/Support/Debug.h"
22 #define DEBUG_TYPE "format-token-annotator"
29 /// Returns \c true if the line starts with a token that can start a statement
30 /// with an initializer.
31 static bool startsWithInitStatement(const AnnotatedLine
&Line
) {
32 return Line
.startsWith(tok::kw_for
) || Line
.startsWith(tok::kw_if
) ||
33 Line
.startsWith(tok::kw_switch
);
36 /// Returns \c true if the token can be used as an identifier in
37 /// an Objective-C \c \@selector, \c false otherwise.
39 /// Because getFormattingLangOpts() always lexes source code as
40 /// Objective-C++, C++ keywords like \c new and \c delete are
41 /// lexed as tok::kw_*, not tok::identifier, even for Objective-C.
43 /// For Objective-C and Objective-C++, both identifiers and keywords
44 /// are valid inside @selector(...) (or a macro which
45 /// invokes @selector(...)). So, we allow treat any identifier or
46 /// keyword as a potential Objective-C selector component.
47 static bool canBeObjCSelectorComponent(const FormatToken
&Tok
) {
48 return Tok
.Tok
.getIdentifierInfo();
51 /// With `Left` being '(', check if we're at either `[...](` or
52 /// `[...]<...>(`, where the [ opens a lambda capture list.
53 static bool isLambdaParameterList(const FormatToken
*Left
) {
54 // Skip <...> if present.
55 if (Left
->Previous
&& Left
->Previous
->is(tok::greater
) &&
56 Left
->Previous
->MatchingParen
&&
57 Left
->Previous
->MatchingParen
->is(TT_TemplateOpener
)) {
58 Left
= Left
->Previous
->MatchingParen
;
62 return Left
->Previous
&& Left
->Previous
->is(tok::r_square
) &&
63 Left
->Previous
->MatchingParen
&&
64 Left
->Previous
->MatchingParen
->is(TT_LambdaLSquare
);
67 /// Returns \c true if the token is followed by a boolean condition, \c false
69 static bool isKeywordWithCondition(const FormatToken
&Tok
) {
70 return Tok
.isOneOf(tok::kw_if
, tok::kw_for
, tok::kw_while
, tok::kw_switch
,
71 tok::kw_constexpr
, tok::kw_catch
);
74 /// Returns \c true if the token starts a C++ attribute, \c false otherwise.
75 static bool isCppAttribute(bool IsCpp
, const FormatToken
&Tok
) {
76 if (!IsCpp
|| !Tok
.startsSequence(tok::l_square
, tok::l_square
))
78 // The first square bracket is part of an ObjC array literal
79 if (Tok
.Previous
&& Tok
.Previous
->is(tok::at
))
81 const FormatToken
*AttrTok
= Tok
.Next
->Next
;
84 // C++17 '[[using ns: foo, bar(baz, blech)]]'
85 // We assume nobody will name an ObjC variable 'using'.
86 if (AttrTok
->startsSequence(tok::kw_using
, tok::identifier
, tok::colon
))
88 if (AttrTok
->isNot(tok::identifier
))
90 while (AttrTok
&& !AttrTok
->startsSequence(tok::r_square
, tok::r_square
)) {
91 // ObjC message send. We assume nobody will use : in a C++11 attribute
92 // specifier parameter, although this is technically valid:
94 if (AttrTok
->is(tok::colon
) ||
95 AttrTok
->startsSequence(tok::identifier
, tok::identifier
) ||
96 AttrTok
->startsSequence(tok::r_paren
, tok::identifier
)) {
99 if (AttrTok
->is(tok::ellipsis
))
101 AttrTok
= AttrTok
->Next
;
103 return AttrTok
&& AttrTok
->startsSequence(tok::r_square
, tok::r_square
);
106 /// A parser that gathers additional information about tokens.
108 /// The \c TokenAnnotator tries to match parenthesis and square brakets and
109 /// store a parenthesis levels. It also tries to resolve matching "<" and ">"
110 /// into template parameter lists.
111 class AnnotatingParser
{
113 AnnotatingParser(const FormatStyle
&Style
, AnnotatedLine
&Line
,
114 const AdditionalKeywords
&Keywords
,
115 SmallVector
<ScopeType
> &Scopes
)
116 : Style(Style
), Line(Line
), CurrentToken(Line
.First
), AutoFound(false),
117 Keywords(Keywords
), Scopes(Scopes
) {
118 Contexts
.push_back(Context(tok::unknown
, 1, /*IsExpression=*/false));
119 resetTokenMetadata();
123 ScopeType
getScopeType(const FormatToken
&Token
) const {
124 switch (Token
.getType()) {
125 case TT_FunctionLBrace
:
126 case TT_LambdaLBrace
:
129 case TT_StructLBrace
:
138 if (!CurrentToken
|| !CurrentToken
->Previous
)
140 if (NonTemplateLess
.count(CurrentToken
->Previous
) > 0)
143 const FormatToken
&Previous
= *CurrentToken
->Previous
; // The '<'.
144 if (Previous
.Previous
) {
145 if (Previous
.Previous
->Tok
.isLiteral())
147 if (Previous
.Previous
->is(tok::r_brace
))
149 if (Previous
.Previous
->is(tok::r_paren
) && Contexts
.size() > 1 &&
150 (!Previous
.Previous
->MatchingParen
||
151 Previous
.Previous
->MatchingParen
->isNot(
152 TT_OverloadedOperatorLParen
))) {
157 FormatToken
*Left
= CurrentToken
->Previous
;
158 Left
->ParentBracket
= Contexts
.back().ContextKind
;
159 ScopedContextCreator
ContextCreator(*this, tok::less
, 12);
161 // If this angle is in the context of an expression, we need to be more
162 // hesitant to detect it as opening template parameters.
163 bool InExprContext
= Contexts
.back().IsExpression
;
165 Contexts
.back().IsExpression
= false;
166 // If there's a template keyword before the opening angle bracket, this is a
167 // template parameter, not an argument.
168 if (Left
->Previous
&& Left
->Previous
->isNot(tok::kw_template
))
169 Contexts
.back().ContextType
= Context::TemplateArgument
;
171 if (Style
.Language
== FormatStyle::LK_Java
&&
172 CurrentToken
->is(tok::question
)) {
176 while (CurrentToken
) {
177 if (CurrentToken
->is(tok::greater
)) {
178 // Try to do a better job at looking for ">>" within the condition of
179 // a statement. Conservatively insert spaces between consecutive ">"
180 // tokens to prevent splitting right bitshift operators and potentially
181 // altering program semantics. This check is overly conservative and
182 // will prevent spaces from being inserted in select nested template
183 // parameter cases, but should not alter program semantics.
184 if (CurrentToken
->Next
&& CurrentToken
->Next
->is(tok::greater
) &&
185 Left
->ParentBracket
!= tok::less
&&
186 CurrentToken
->getStartOfNonWhitespace() ==
187 CurrentToken
->Next
->getStartOfNonWhitespace().getLocWithOffset(
191 Left
->MatchingParen
= CurrentToken
;
192 CurrentToken
->MatchingParen
= Left
;
193 // In TT_Proto, we must distignuish between:
195 // msg < item: data >
196 // msg: < item: data >
197 // In TT_TextProto, map<key, value> does not occur.
198 if (Style
.Language
== FormatStyle::LK_TextProto
||
199 (Style
.Language
== FormatStyle::LK_Proto
&& Left
->Previous
&&
200 Left
->Previous
->isOneOf(TT_SelectorName
, TT_DictLiteral
))) {
201 CurrentToken
->setType(TT_DictLiteral
);
203 CurrentToken
->setType(TT_TemplateCloser
);
204 CurrentToken
->Tok
.setLength(1);
206 if (CurrentToken
->Next
&& CurrentToken
->Next
->Tok
.isLiteral())
211 if (CurrentToken
->is(tok::question
) &&
212 Style
.Language
== FormatStyle::LK_Java
) {
216 if (CurrentToken
->isOneOf(tok::r_paren
, tok::r_square
, tok::r_brace
) ||
217 (CurrentToken
->isOneOf(tok::colon
, tok::question
) && InExprContext
&&
218 !Style
.isCSharp() && Style
.Language
!= FormatStyle::LK_Proto
&&
219 Style
.Language
!= FormatStyle::LK_TextProto
)) {
222 // If a && or || is found and interpreted as a binary operator, this set
223 // of angles is likely part of something like "a < b && c > d". If the
224 // angles are inside an expression, the ||/&& might also be a binary
225 // operator that was misinterpreted because we are parsing template
227 // FIXME: This is getting out of hand, write a decent parser.
228 if (CurrentToken
->Previous
->isOneOf(tok::pipepipe
, tok::ampamp
) &&
229 CurrentToken
->Previous
->is(TT_BinaryOperator
) &&
230 Contexts
[Contexts
.size() - 2].IsExpression
&&
231 !Line
.startsWith(tok::kw_template
)) {
234 updateParameterCount(Left
, CurrentToken
);
235 if (Style
.Language
== FormatStyle::LK_Proto
) {
236 if (FormatToken
*Previous
= CurrentToken
->getPreviousNonComment()) {
237 if (CurrentToken
->is(tok::colon
) ||
238 (CurrentToken
->isOneOf(tok::l_brace
, tok::less
) &&
239 Previous
->isNot(tok::colon
))) {
240 Previous
->setType(TT_SelectorName
);
250 bool parseUntouchableParens() {
251 while (CurrentToken
) {
252 CurrentToken
->Finalized
= true;
253 switch (CurrentToken
->Tok
.getKind()) {
256 if (!parseUntouchableParens())
271 bool parseParens(bool LookForDecls
= false) {
274 assert(CurrentToken
->Previous
&& "Unknown previous token");
275 FormatToken
&OpeningParen
= *CurrentToken
->Previous
;
276 assert(OpeningParen
.is(tok::l_paren
));
277 FormatToken
*PrevNonComment
= OpeningParen
.getPreviousNonComment();
278 OpeningParen
.ParentBracket
= Contexts
.back().ContextKind
;
279 ScopedContextCreator
ContextCreator(*this, tok::l_paren
, 1);
281 // FIXME: This is a bit of a hack. Do better.
282 Contexts
.back().ColonIsForRangeExpr
=
283 Contexts
.size() == 2 && Contexts
[0].ColonIsForRangeExpr
;
285 if (OpeningParen
.Previous
&&
286 OpeningParen
.Previous
->is(TT_UntouchableMacroFunc
)) {
287 OpeningParen
.Finalized
= true;
288 return parseUntouchableParens();
291 bool StartsObjCMethodExpr
= false;
292 if (!Style
.isVerilog()) {
293 if (FormatToken
*MaybeSel
= OpeningParen
.Previous
) {
294 // @selector( starts a selector.
295 if (MaybeSel
->isObjCAtKeyword(tok::objc_selector
) &&
296 MaybeSel
->Previous
&& MaybeSel
->Previous
->is(tok::at
)) {
297 StartsObjCMethodExpr
= true;
302 if (OpeningParen
.is(TT_OverloadedOperatorLParen
)) {
303 // Find the previous kw_operator token.
304 FormatToken
*Prev
= &OpeningParen
;
305 while (Prev
->isNot(tok::kw_operator
)) {
306 Prev
= Prev
->Previous
;
307 assert(Prev
&& "Expect a kw_operator prior to the OperatorLParen!");
310 // If faced with "a.operator*(argument)" or "a->operator*(argument)",
311 // i.e. the operator is called as a member function,
312 // then the argument must be an expression.
313 bool OperatorCalledAsMemberFunction
=
314 Prev
->Previous
&& Prev
->Previous
->isOneOf(tok::period
, tok::arrow
);
315 Contexts
.back().IsExpression
= OperatorCalledAsMemberFunction
;
316 } else if (OpeningParen
.is(TT_VerilogInstancePortLParen
)) {
317 Contexts
.back().IsExpression
= true;
318 Contexts
.back().ContextType
= Context::VerilogInstancePortList
;
319 } else if (Style
.isJavaScript() &&
320 (Line
.startsWith(Keywords
.kw_type
, tok::identifier
) ||
321 Line
.startsWith(tok::kw_export
, Keywords
.kw_type
,
324 // export type X = (...);
325 Contexts
.back().IsExpression
= false;
326 } else if (OpeningParen
.Previous
&&
327 (OpeningParen
.Previous
->isOneOf(
328 tok::kw_static_assert
, tok::kw_noexcept
, tok::kw_explicit
,
329 tok::kw_while
, tok::l_paren
, tok::comma
,
330 TT_BinaryOperator
) ||
331 OpeningParen
.Previous
->isIf())) {
332 // static_assert, if and while usually contain expressions.
333 Contexts
.back().IsExpression
= true;
334 } else if (Style
.isJavaScript() && OpeningParen
.Previous
&&
335 (OpeningParen
.Previous
->is(Keywords
.kw_function
) ||
336 (OpeningParen
.Previous
->endsSequence(tok::identifier
,
337 Keywords
.kw_function
)))) {
338 // function(...) or function f(...)
339 Contexts
.back().IsExpression
= false;
340 } else if (Style
.isJavaScript() && OpeningParen
.Previous
&&
341 OpeningParen
.Previous
->is(TT_JsTypeColon
)) {
342 // let x: (SomeType);
343 Contexts
.back().IsExpression
= false;
344 } else if (isLambdaParameterList(&OpeningParen
)) {
345 // This is a parameter list of a lambda expression.
346 Contexts
.back().IsExpression
= false;
347 } else if (OpeningParen
.is(TT_RequiresExpressionLParen
)) {
348 Contexts
.back().IsExpression
= false;
349 } else if (OpeningParen
.Previous
&&
350 OpeningParen
.Previous
->is(tok::kw__Generic
)) {
351 Contexts
.back().ContextType
= Context::C11GenericSelection
;
352 Contexts
.back().IsExpression
= true;
353 } else if (Line
.InPPDirective
&&
354 (!OpeningParen
.Previous
||
355 OpeningParen
.Previous
->isNot(tok::identifier
))) {
356 Contexts
.back().IsExpression
= true;
357 } else if (Contexts
[Contexts
.size() - 2].CaretFound
) {
358 // This is the parameter list of an ObjC block.
359 Contexts
.back().IsExpression
= false;
360 } else if (OpeningParen
.Previous
&&
361 OpeningParen
.Previous
->is(TT_ForEachMacro
)) {
362 // The first argument to a foreach macro is a declaration.
363 Contexts
.back().ContextType
= Context::ForEachMacro
;
364 Contexts
.back().IsExpression
= false;
365 } else if (OpeningParen
.Previous
&& OpeningParen
.Previous
->MatchingParen
&&
366 OpeningParen
.Previous
->MatchingParen
->isOneOf(
367 TT_ObjCBlockLParen
, TT_FunctionTypeLParen
)) {
368 Contexts
.back().IsExpression
= false;
369 } else if (!Line
.MustBeDeclaration
&& !Line
.InPPDirective
) {
371 OpeningParen
.Previous
&&
372 OpeningParen
.Previous
->isOneOf(tok::kw_for
, tok::kw_catch
);
373 Contexts
.back().IsExpression
= !IsForOrCatch
;
376 // Infer the role of the l_paren based on the previous token if we haven't
378 if (PrevNonComment
&& OpeningParen
.is(TT_Unknown
)) {
379 if (PrevNonComment
->is(tok::kw___attribute
)) {
380 OpeningParen
.setType(TT_AttributeParen
);
381 } else if (PrevNonComment
->isOneOf(TT_TypenameMacro
, tok::kw_decltype
,
383 #define TRANSFORM_TYPE_TRAIT_DEF(_, Trait) tok::kw___##Trait,
384 #include "clang/Basic/TransformTypeTraits.def"
386 OpeningParen
.setType(TT_TypeDeclarationParen
);
387 // decltype() and typeof() usually contain expressions.
388 if (PrevNonComment
->isOneOf(tok::kw_decltype
, tok::kw_typeof
))
389 Contexts
.back().IsExpression
= true;
393 if (StartsObjCMethodExpr
) {
394 Contexts
.back().ColonIsObjCMethodExpr
= true;
395 OpeningParen
.setType(TT_ObjCMethodExpr
);
398 // MightBeFunctionType and ProbablyFunctionType are used for
399 // function pointer and reference types as well as Objective-C
402 // void (*FunctionPointer)(void);
403 // void (&FunctionReference)(void);
404 // void (&&FunctionReference)(void);
405 // void (^ObjCBlock)(void);
406 bool MightBeFunctionType
= !Contexts
[Contexts
.size() - 2].IsExpression
;
407 bool ProbablyFunctionType
=
408 CurrentToken
->isOneOf(tok::star
, tok::amp
, tok::ampamp
, tok::caret
);
409 bool HasMultipleLines
= false;
410 bool HasMultipleParametersOnALine
= false;
411 bool MightBeObjCForRangeLoop
=
412 OpeningParen
.Previous
&& OpeningParen
.Previous
->is(tok::kw_for
);
413 FormatToken
*PossibleObjCForInToken
= nullptr;
414 while (CurrentToken
) {
415 // LookForDecls is set when "if (" has been seen. Check for
416 // 'identifier' '*' 'identifier' followed by not '=' -- this
417 // '*' has to be a binary operator but determineStarAmpUsage() will
418 // categorize it as an unary operator, so set the right type here.
419 if (LookForDecls
&& CurrentToken
->Next
) {
420 FormatToken
*Prev
= CurrentToken
->getPreviousNonComment();
422 FormatToken
*PrevPrev
= Prev
->getPreviousNonComment();
423 FormatToken
*Next
= CurrentToken
->Next
;
424 if (PrevPrev
&& PrevPrev
->is(tok::identifier
) &&
425 PrevPrev
->isNot(TT_TypeName
) &&
426 Prev
->isOneOf(tok::star
, tok::amp
, tok::ampamp
) &&
427 CurrentToken
->is(tok::identifier
) && Next
->isNot(tok::equal
)) {
428 Prev
->setType(TT_BinaryOperator
);
429 LookForDecls
= false;
434 if (CurrentToken
->Previous
->is(TT_PointerOrReference
) &&
435 CurrentToken
->Previous
->Previous
->isOneOf(tok::l_paren
,
437 ProbablyFunctionType
= true;
439 if (CurrentToken
->is(tok::comma
))
440 MightBeFunctionType
= false;
441 if (CurrentToken
->Previous
->is(TT_BinaryOperator
))
442 Contexts
.back().IsExpression
= true;
443 if (CurrentToken
->is(tok::r_paren
)) {
444 if (OpeningParen
.isNot(TT_CppCastLParen
) && MightBeFunctionType
&&
445 ProbablyFunctionType
&& CurrentToken
->Next
&&
446 (CurrentToken
->Next
->is(tok::l_paren
) ||
447 (CurrentToken
->Next
->is(tok::l_square
) &&
448 Line
.MustBeDeclaration
))) {
449 OpeningParen
.setType(OpeningParen
.Next
->is(tok::caret
)
451 : TT_FunctionTypeLParen
);
453 OpeningParen
.MatchingParen
= CurrentToken
;
454 CurrentToken
->MatchingParen
= &OpeningParen
;
456 if (CurrentToken
->Next
&& CurrentToken
->Next
->is(tok::l_brace
) &&
457 OpeningParen
.Previous
&& OpeningParen
.Previous
->is(tok::l_paren
)) {
458 // Detect the case where macros are used to generate lambdas or
459 // function bodies, e.g.:
460 // auto my_lambda = MACRO((Type *type, int i) { .. body .. });
461 for (FormatToken
*Tok
= &OpeningParen
; Tok
!= CurrentToken
;
463 if (Tok
->is(TT_BinaryOperator
) &&
464 Tok
->isOneOf(tok::star
, tok::amp
, tok::ampamp
)) {
465 Tok
->setType(TT_PointerOrReference
);
470 if (StartsObjCMethodExpr
) {
471 CurrentToken
->setType(TT_ObjCMethodExpr
);
472 if (Contexts
.back().FirstObjCSelectorName
) {
473 Contexts
.back().FirstObjCSelectorName
->LongestObjCSelectorName
=
474 Contexts
.back().LongestObjCSelectorName
;
478 if (OpeningParen
.is(TT_AttributeParen
))
479 CurrentToken
->setType(TT_AttributeParen
);
480 if (OpeningParen
.is(TT_TypeDeclarationParen
))
481 CurrentToken
->setType(TT_TypeDeclarationParen
);
482 if (OpeningParen
.Previous
&&
483 OpeningParen
.Previous
->is(TT_JavaAnnotation
)) {
484 CurrentToken
->setType(TT_JavaAnnotation
);
486 if (OpeningParen
.Previous
&&
487 OpeningParen
.Previous
->is(TT_LeadingJavaAnnotation
)) {
488 CurrentToken
->setType(TT_LeadingJavaAnnotation
);
490 if (OpeningParen
.Previous
&&
491 OpeningParen
.Previous
->is(TT_AttributeSquare
)) {
492 CurrentToken
->setType(TT_AttributeSquare
);
495 if (!HasMultipleLines
)
496 OpeningParen
.setPackingKind(PPK_Inconclusive
);
497 else if (HasMultipleParametersOnALine
)
498 OpeningParen
.setPackingKind(PPK_BinPacked
);
500 OpeningParen
.setPackingKind(PPK_OnePerLine
);
505 if (CurrentToken
->isOneOf(tok::r_square
, tok::r_brace
))
508 if (CurrentToken
->is(tok::l_brace
) && OpeningParen
.is(TT_ObjCBlockLParen
))
509 OpeningParen
.setType(TT_Unknown
);
510 if (CurrentToken
->is(tok::comma
) && CurrentToken
->Next
&&
511 !CurrentToken
->Next
->HasUnescapedNewline
&&
512 !CurrentToken
->Next
->isTrailingComment()) {
513 HasMultipleParametersOnALine
= true;
515 bool ProbablyFunctionTypeLParen
=
516 (CurrentToken
->is(tok::l_paren
) && CurrentToken
->Next
&&
517 CurrentToken
->Next
->isOneOf(tok::star
, tok::amp
, tok::caret
));
518 if ((CurrentToken
->Previous
->isOneOf(tok::kw_const
, tok::kw_auto
) ||
519 CurrentToken
->Previous
->isSimpleTypeSpecifier()) &&
520 !(CurrentToken
->is(tok::l_brace
) ||
521 (CurrentToken
->is(tok::l_paren
) && !ProbablyFunctionTypeLParen
))) {
522 Contexts
.back().IsExpression
= false;
524 if (CurrentToken
->isOneOf(tok::semi
, tok::colon
)) {
525 MightBeObjCForRangeLoop
= false;
526 if (PossibleObjCForInToken
) {
527 PossibleObjCForInToken
->setType(TT_Unknown
);
528 PossibleObjCForInToken
= nullptr;
531 if (MightBeObjCForRangeLoop
&& CurrentToken
->is(Keywords
.kw_in
)) {
532 PossibleObjCForInToken
= CurrentToken
;
533 PossibleObjCForInToken
->setType(TT_ObjCForIn
);
535 // When we discover a 'new', we set CanBeExpression to 'false' in order to
536 // parse the type correctly. Reset that after a comma.
537 if (CurrentToken
->is(tok::comma
))
538 Contexts
.back().CanBeExpression
= true;
540 FormatToken
*Tok
= CurrentToken
;
543 updateParameterCount(&OpeningParen
, Tok
);
544 if (CurrentToken
&& CurrentToken
->HasUnescapedNewline
)
545 HasMultipleLines
= true;
550 bool isCSharpAttributeSpecifier(const FormatToken
&Tok
) {
551 if (!Style
.isCSharp())
554 // `identifier[i]` is not an attribute.
555 if (Tok
.Previous
&& Tok
.Previous
->is(tok::identifier
))
558 // Chains of [] in `identifier[i][j][k]` are not attributes.
559 if (Tok
.Previous
&& Tok
.Previous
->is(tok::r_square
)) {
560 auto *MatchingParen
= Tok
.Previous
->MatchingParen
;
561 if (!MatchingParen
|| MatchingParen
->is(TT_ArraySubscriptLSquare
))
565 const FormatToken
*AttrTok
= Tok
.Next
;
569 // Just an empty declaration e.g. string [].
570 if (AttrTok
->is(tok::r_square
))
573 // Move along the tokens inbetween the '[' and ']' e.g. [STAThread].
574 while (AttrTok
&& AttrTok
->isNot(tok::r_square
))
575 AttrTok
= AttrTok
->Next
;
580 // Allow an attribute to be the only content of a file.
581 AttrTok
= AttrTok
->Next
;
585 // Limit this to being an access modifier that follows.
586 if (AttrTok
->isOneOf(tok::kw_public
, tok::kw_private
, tok::kw_protected
,
587 tok::comment
, tok::kw_class
, tok::kw_static
,
588 tok::l_square
, Keywords
.kw_internal
)) {
592 // incase its a [XXX] retval func(....
594 AttrTok
->Next
->startsSequence(tok::identifier
, tok::l_paren
)) {
605 // A '[' could be an index subscript (after an identifier or after
606 // ')' or ']'), it could be the start of an Objective-C method
607 // expression, it could the start of an Objective-C array literal,
608 // or it could be a C++ attribute specifier [[foo::bar]].
609 FormatToken
*Left
= CurrentToken
->Previous
;
610 Left
->ParentBracket
= Contexts
.back().ContextKind
;
611 FormatToken
*Parent
= Left
->getPreviousNonComment();
613 // Cases where '>' is followed by '['.
614 // In C++, this can happen either in array of templates (foo<int>[10])
615 // or when array is a nested template type (unique_ptr<type1<type2>[]>).
616 bool CppArrayTemplates
=
617 Style
.isCpp() && Parent
&& Parent
->is(TT_TemplateCloser
) &&
618 (Contexts
.back().CanBeExpression
|| Contexts
.back().IsExpression
||
619 Contexts
.back().ContextType
== Context::TemplateArgument
);
621 const bool IsInnerSquare
= Contexts
.back().InCpp11AttributeSpecifier
;
622 const bool IsCpp11AttributeSpecifier
=
623 isCppAttribute(Style
.isCpp(), *Left
) || IsInnerSquare
;
625 // Treat C# Attributes [STAThread] much like C++ attributes [[...]].
626 bool IsCSharpAttributeSpecifier
=
627 isCSharpAttributeSpecifier(*Left
) ||
628 Contexts
.back().InCSharpAttributeSpecifier
;
630 bool InsideInlineASM
= Line
.startsWith(tok::kw_asm
);
631 bool IsCppStructuredBinding
= Left
->isCppStructuredBinding(Style
);
632 bool StartsObjCMethodExpr
=
633 !IsCppStructuredBinding
&& !InsideInlineASM
&& !CppArrayTemplates
&&
634 Style
.isCpp() && !IsCpp11AttributeSpecifier
&&
635 !IsCSharpAttributeSpecifier
&& Contexts
.back().CanBeExpression
&&
636 Left
->isNot(TT_LambdaLSquare
) &&
637 !CurrentToken
->isOneOf(tok::l_brace
, tok::r_square
) &&
639 Parent
->isOneOf(tok::colon
, tok::l_square
, tok::l_paren
,
640 tok::kw_return
, tok::kw_throw
) ||
641 Parent
->isUnaryOperator() ||
642 // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen.
643 Parent
->isOneOf(TT_ObjCForIn
, TT_CastRParen
) ||
644 (getBinOpPrecedence(Parent
->Tok
.getKind(), true, true) >
646 bool ColonFound
= false;
648 unsigned BindingIncrease
= 1;
649 if (IsCppStructuredBinding
) {
650 Left
->setType(TT_StructuredBindingLSquare
);
651 } else if (Left
->is(TT_Unknown
)) {
652 if (StartsObjCMethodExpr
) {
653 Left
->setType(TT_ObjCMethodExpr
);
654 } else if (InsideInlineASM
) {
655 Left
->setType(TT_InlineASMSymbolicNameLSquare
);
656 } else if (IsCpp11AttributeSpecifier
) {
657 Left
->setType(TT_AttributeSquare
);
658 if (!IsInnerSquare
&& Left
->Previous
)
659 Left
->Previous
->EndsCppAttributeGroup
= false;
660 } else if (Style
.isJavaScript() && Parent
&&
661 Contexts
.back().ContextKind
== tok::l_brace
&&
662 Parent
->isOneOf(tok::l_brace
, tok::comma
)) {
663 Left
->setType(TT_JsComputedPropertyName
);
664 } else if (Style
.isCpp() && Contexts
.back().ContextKind
== tok::l_brace
&&
665 Parent
&& Parent
->isOneOf(tok::l_brace
, tok::comma
)) {
666 Left
->setType(TT_DesignatedInitializerLSquare
);
667 } else if (IsCSharpAttributeSpecifier
) {
668 Left
->setType(TT_AttributeSquare
);
669 } else if (CurrentToken
->is(tok::r_square
) && Parent
&&
670 Parent
->is(TT_TemplateCloser
)) {
671 Left
->setType(TT_ArraySubscriptLSquare
);
672 } else if (Style
.Language
== FormatStyle::LK_Proto
||
673 Style
.Language
== FormatStyle::LK_TextProto
) {
674 // Square braces in LK_Proto can either be message field attributes:
676 // optional Aaa aaa = 1 [
684 // or text proto extensions (in options):
686 // option (Aaa.options) = {
687 // [type.type/type] {
692 // or repeated fields (in options):
694 // option (Aaa.options) = {
698 // In the first and the third case we want to spread the contents inside
699 // the square braces; in the second we want to keep them inline.
700 Left
->setType(TT_ArrayInitializerLSquare
);
701 if (!Left
->endsSequence(tok::l_square
, tok::numeric_constant
,
703 !Left
->endsSequence(tok::l_square
, tok::numeric_constant
,
705 !Left
->endsSequence(tok::l_square
, tok::colon
, TT_SelectorName
)) {
706 Left
->setType(TT_ProtoExtensionLSquare
);
707 BindingIncrease
= 10;
709 } else if (!CppArrayTemplates
&& Parent
&&
710 Parent
->isOneOf(TT_BinaryOperator
, TT_TemplateCloser
, tok::at
,
711 tok::comma
, tok::l_paren
, tok::l_square
,
712 tok::question
, tok::colon
, tok::kw_return
,
713 // Should only be relevant to JavaScript:
715 Left
->setType(TT_ArrayInitializerLSquare
);
717 BindingIncrease
= 10;
718 Left
->setType(TT_ArraySubscriptLSquare
);
722 ScopedContextCreator
ContextCreator(*this, tok::l_square
, BindingIncrease
);
723 Contexts
.back().IsExpression
= true;
724 if (Style
.isJavaScript() && Parent
&& Parent
->is(TT_JsTypeColon
))
725 Contexts
.back().IsExpression
= false;
727 Contexts
.back().ColonIsObjCMethodExpr
= StartsObjCMethodExpr
;
728 Contexts
.back().InCpp11AttributeSpecifier
= IsCpp11AttributeSpecifier
;
729 Contexts
.back().InCSharpAttributeSpecifier
= IsCSharpAttributeSpecifier
;
731 while (CurrentToken
) {
732 if (CurrentToken
->is(tok::r_square
)) {
733 if (IsCpp11AttributeSpecifier
) {
734 CurrentToken
->setType(TT_AttributeSquare
);
736 CurrentToken
->EndsCppAttributeGroup
= true;
738 if (IsCSharpAttributeSpecifier
) {
739 CurrentToken
->setType(TT_AttributeSquare
);
740 } else if (((CurrentToken
->Next
&&
741 CurrentToken
->Next
->is(tok::l_paren
)) ||
742 (CurrentToken
->Previous
&&
743 CurrentToken
->Previous
->Previous
== Left
)) &&
744 Left
->is(TT_ObjCMethodExpr
)) {
745 // An ObjC method call is rarely followed by an open parenthesis. It
746 // also can't be composed of just one token, unless it's a macro that
747 // will be expanded to more tokens.
748 // FIXME: Do we incorrectly label ":" with this?
749 StartsObjCMethodExpr
= false;
750 Left
->setType(TT_Unknown
);
752 if (StartsObjCMethodExpr
&& CurrentToken
->Previous
!= Left
) {
753 CurrentToken
->setType(TT_ObjCMethodExpr
);
754 // If we haven't seen a colon yet, make sure the last identifier
755 // before the r_square is tagged as a selector name component.
756 if (!ColonFound
&& CurrentToken
->Previous
&&
757 CurrentToken
->Previous
->is(TT_Unknown
) &&
758 canBeObjCSelectorComponent(*CurrentToken
->Previous
)) {
759 CurrentToken
->Previous
->setType(TT_SelectorName
);
761 // determineStarAmpUsage() thinks that '*' '[' is allocating an
762 // array of pointers, but if '[' starts a selector then '*' is a
764 if (Parent
&& Parent
->is(TT_PointerOrReference
))
765 Parent
->overwriteFixedType(TT_BinaryOperator
);
767 // An arrow after an ObjC method expression is not a lambda arrow.
768 if (CurrentToken
->getType() == TT_ObjCMethodExpr
&&
769 CurrentToken
->Next
&& CurrentToken
->Next
->is(TT_LambdaArrow
)) {
770 CurrentToken
->Next
->overwriteFixedType(TT_Unknown
);
772 Left
->MatchingParen
= CurrentToken
;
773 CurrentToken
->MatchingParen
= Left
;
774 // FirstObjCSelectorName is set when a colon is found. This does
775 // not work, however, when the method has no parameters.
776 // Here, we set FirstObjCSelectorName when the end of the method call is
777 // reached, in case it was not set already.
778 if (!Contexts
.back().FirstObjCSelectorName
) {
779 FormatToken
*Previous
= CurrentToken
->getPreviousNonComment();
780 if (Previous
&& Previous
->is(TT_SelectorName
)) {
781 Previous
->ObjCSelectorNameParts
= 1;
782 Contexts
.back().FirstObjCSelectorName
= Previous
;
785 Left
->ParameterCount
=
786 Contexts
.back().FirstObjCSelectorName
->ObjCSelectorNameParts
;
788 if (Contexts
.back().FirstObjCSelectorName
) {
789 Contexts
.back().FirstObjCSelectorName
->LongestObjCSelectorName
=
790 Contexts
.back().LongestObjCSelectorName
;
791 if (Left
->BlockParameterCount
> 1)
792 Contexts
.back().FirstObjCSelectorName
->LongestObjCSelectorName
= 0;
797 if (CurrentToken
->isOneOf(tok::r_paren
, tok::r_brace
))
799 if (CurrentToken
->is(tok::colon
)) {
800 if (IsCpp11AttributeSpecifier
&&
801 CurrentToken
->endsSequence(tok::colon
, tok::identifier
,
803 // Remember that this is a [[using ns: foo]] C++ attribute, so we
804 // don't add a space before the colon (unlike other colons).
805 CurrentToken
->setType(TT_AttributeColon
);
806 } else if (!Style
.isVerilog() && !Line
.InPragmaDirective
&&
807 Left
->isOneOf(TT_ArraySubscriptLSquare
,
808 TT_DesignatedInitializerLSquare
)) {
809 Left
->setType(TT_ObjCMethodExpr
);
810 StartsObjCMethodExpr
= true;
811 Contexts
.back().ColonIsObjCMethodExpr
= true;
812 if (Parent
&& Parent
->is(tok::r_paren
)) {
813 // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen.
814 Parent
->setType(TT_CastRParen
);
819 if (CurrentToken
->is(tok::comma
) && Left
->is(TT_ObjCMethodExpr
) &&
821 Left
->setType(TT_ArrayInitializerLSquare
);
823 FormatToken
*Tok
= CurrentToken
;
826 updateParameterCount(Left
, Tok
);
831 bool couldBeInStructArrayInitializer() const {
832 if (Contexts
.size() < 2)
834 // We want to back up no more then 2 context levels i.e.
836 const auto End
= std::next(Contexts
.rbegin(), 2);
837 auto Last
= Contexts
.rbegin();
839 for (; Last
!= End
; ++Last
)
840 if (Last
->ContextKind
== tok::l_brace
)
842 return Depth
== 2 && Last
->ContextKind
!= tok::l_brace
;
849 assert(CurrentToken
->Previous
);
850 FormatToken
&OpeningBrace
= *CurrentToken
->Previous
;
851 assert(OpeningBrace
.is(tok::l_brace
));
852 OpeningBrace
.ParentBracket
= Contexts
.back().ContextKind
;
854 if (Contexts
.back().CaretFound
)
855 OpeningBrace
.overwriteFixedType(TT_ObjCBlockLBrace
);
856 Contexts
.back().CaretFound
= false;
858 ScopedContextCreator
ContextCreator(*this, tok::l_brace
, 1);
859 Contexts
.back().ColonIsDictLiteral
= true;
860 if (OpeningBrace
.is(BK_BracedInit
))
861 Contexts
.back().IsExpression
= true;
862 if (Style
.isJavaScript() && OpeningBrace
.Previous
&&
863 OpeningBrace
.Previous
->is(TT_JsTypeColon
)) {
864 Contexts
.back().IsExpression
= false;
866 if (Style
.isVerilog() &&
867 (!OpeningBrace
.getPreviousNonComment() ||
868 OpeningBrace
.getPreviousNonComment()->isNot(Keywords
.kw_apostrophe
))) {
869 Contexts
.back().VerilogMayBeConcatenation
= true;
872 unsigned CommaCount
= 0;
873 while (CurrentToken
) {
874 if (CurrentToken
->is(tok::r_brace
)) {
875 assert(!Scopes
.empty());
876 assert(Scopes
.back() == getScopeType(OpeningBrace
));
878 assert(OpeningBrace
.Optional
== CurrentToken
->Optional
);
879 OpeningBrace
.MatchingParen
= CurrentToken
;
880 CurrentToken
->MatchingParen
= &OpeningBrace
;
881 if (Style
.AlignArrayOfStructures
!= FormatStyle::AIAS_None
) {
882 if (OpeningBrace
.ParentBracket
== tok::l_brace
&&
883 couldBeInStructArrayInitializer() && CommaCount
> 0) {
884 Contexts
.back().ContextType
= Context::StructArrayInitializer
;
890 if (CurrentToken
->isOneOf(tok::r_paren
, tok::r_square
))
892 updateParameterCount(&OpeningBrace
, CurrentToken
);
893 if (CurrentToken
->isOneOf(tok::colon
, tok::l_brace
, tok::less
)) {
894 FormatToken
*Previous
= CurrentToken
->getPreviousNonComment();
895 if (Previous
->is(TT_JsTypeOptionalQuestion
))
896 Previous
= Previous
->getPreviousNonComment();
897 if ((CurrentToken
->is(tok::colon
) &&
898 (!Contexts
.back().ColonIsDictLiteral
|| !Style
.isCpp())) ||
899 Style
.Language
== FormatStyle::LK_Proto
||
900 Style
.Language
== FormatStyle::LK_TextProto
) {
901 OpeningBrace
.setType(TT_DictLiteral
);
902 if (Previous
->Tok
.getIdentifierInfo() ||
903 Previous
->is(tok::string_literal
)) {
904 Previous
->setType(TT_SelectorName
);
907 if (CurrentToken
->is(tok::colon
) && OpeningBrace
.is(TT_Unknown
))
908 OpeningBrace
.setType(TT_DictLiteral
);
909 else if (Style
.isJavaScript())
910 OpeningBrace
.overwriteFixedType(TT_DictLiteral
);
912 if (CurrentToken
->is(tok::comma
)) {
913 if (Style
.isJavaScript())
914 OpeningBrace
.overwriteFixedType(TT_DictLiteral
);
923 void updateParameterCount(FormatToken
*Left
, FormatToken
*Current
) {
924 // For ObjC methods, the number of parameters is calculated differently as
925 // method declarations have a different structure (the parameters are not
926 // inside a bracket scope).
927 if (Current
->is(tok::l_brace
) && Current
->is(BK_Block
))
928 ++Left
->BlockParameterCount
;
929 if (Current
->is(tok::comma
)) {
930 ++Left
->ParameterCount
;
932 Left
->Role
.reset(new CommaSeparatedList(Style
));
933 Left
->Role
->CommaFound(Current
);
934 } else if (Left
->ParameterCount
== 0 && Current
->isNot(tok::comment
)) {
935 Left
->ParameterCount
= 1;
939 bool parseConditional() {
940 while (CurrentToken
) {
941 if (CurrentToken
->is(tok::colon
)) {
942 CurrentToken
->setType(TT_ConditionalExpr
);
952 bool parseTemplateDeclaration() {
953 if (CurrentToken
&& CurrentToken
->is(tok::less
)) {
954 CurrentToken
->setType(TT_TemplateOpener
);
959 CurrentToken
->Previous
->ClosesTemplateDeclaration
= true;
965 bool consumeToken() {
966 FormatToken
*Tok
= CurrentToken
;
968 // In Verilog primitives' state tables, `:`, `?`, and `-` aren't normal
970 if (Tok
->is(TT_VerilogTableItem
))
972 switch (Tok
->Tok
.getKind()) {
975 if (!Tok
->Previous
&& Line
.MustBeDeclaration
)
976 Tok
->setType(TT_ObjCMethodSpecifier
);
981 // Goto labels and case labels are already identified in
982 // UnwrappedLineParser.
983 if (Tok
->isTypeFinalized())
985 // Colons from ?: are handled in parseConditional().
986 if (Style
.isJavaScript()) {
987 if (Contexts
.back().ColonIsForRangeExpr
|| // colon in for loop
988 (Contexts
.size() == 1 && // switch/case labels
989 !Line
.First
->isOneOf(tok::kw_enum
, tok::kw_case
)) ||
990 Contexts
.back().ContextKind
== tok::l_paren
|| // function params
991 Contexts
.back().ContextKind
== tok::l_square
|| // array type
992 (!Contexts
.back().IsExpression
&&
993 Contexts
.back().ContextKind
== tok::l_brace
) || // object type
994 (Contexts
.size() == 1 &&
995 Line
.MustBeDeclaration
)) { // method/property declaration
996 Contexts
.back().IsExpression
= false;
997 Tok
->setType(TT_JsTypeColon
);
1000 } else if (Style
.isCSharp()) {
1001 if (Contexts
.back().InCSharpAttributeSpecifier
) {
1002 Tok
->setType(TT_AttributeColon
);
1005 if (Contexts
.back().ContextKind
== tok::l_paren
) {
1006 Tok
->setType(TT_CSharpNamedArgumentColon
);
1009 } else if (Style
.isVerilog() && Tok
->isNot(TT_BinaryOperator
)) {
1010 // The distribution weight operators are labeled
1011 // TT_BinaryOperator by the lexer.
1012 if (Keywords
.isVerilogEnd(*Tok
->Previous
) ||
1013 Keywords
.isVerilogBegin(*Tok
->Previous
)) {
1014 Tok
->setType(TT_VerilogBlockLabelColon
);
1015 } else if (Contexts
.back().ContextKind
== tok::l_square
) {
1016 Tok
->setType(TT_BitFieldColon
);
1017 } else if (Contexts
.back().ColonIsDictLiteral
) {
1018 Tok
->setType(TT_DictLiteral
);
1019 } else if (Contexts
.size() == 1) {
1020 // In Verilog a case label doesn't have the case keyword. We
1021 // assume a colon following an expression is a case label.
1022 // Colons from ?: are annotated in parseConditional().
1023 Tok
->setType(TT_CaseLabelColon
);
1024 if (Line
.Level
> 1 || (!Line
.InPPDirective
&& Line
.Level
> 0))
1029 if (Line
.First
->isOneOf(Keywords
.kw_module
, Keywords
.kw_import
) ||
1030 Line
.First
->startsSequence(tok::kw_export
, Keywords
.kw_module
) ||
1031 Line
.First
->startsSequence(tok::kw_export
, Keywords
.kw_import
)) {
1032 Tok
->setType(TT_ModulePartitionColon
);
1033 } else if (Contexts
.back().ColonIsDictLiteral
||
1034 Style
.Language
== FormatStyle::LK_Proto
||
1035 Style
.Language
== FormatStyle::LK_TextProto
) {
1036 Tok
->setType(TT_DictLiteral
);
1037 if (Style
.Language
== FormatStyle::LK_TextProto
) {
1038 if (FormatToken
*Previous
= Tok
->getPreviousNonComment())
1039 Previous
->setType(TT_SelectorName
);
1041 } else if (Contexts
.back().ColonIsObjCMethodExpr
||
1042 Line
.startsWith(TT_ObjCMethodSpecifier
)) {
1043 Tok
->setType(TT_ObjCMethodExpr
);
1044 const FormatToken
*BeforePrevious
= Tok
->Previous
->Previous
;
1045 // Ensure we tag all identifiers in method declarations as
1047 bool UnknownIdentifierInMethodDeclaration
=
1048 Line
.startsWith(TT_ObjCMethodSpecifier
) &&
1049 Tok
->Previous
->is(tok::identifier
) && Tok
->Previous
->is(TT_Unknown
);
1050 if (!BeforePrevious
||
1051 // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen.
1052 !(BeforePrevious
->is(TT_CastRParen
) ||
1053 (BeforePrevious
->is(TT_ObjCMethodExpr
) &&
1054 BeforePrevious
->is(tok::colon
))) ||
1055 BeforePrevious
->is(tok::r_square
) ||
1056 Contexts
.back().LongestObjCSelectorName
== 0 ||
1057 UnknownIdentifierInMethodDeclaration
) {
1058 Tok
->Previous
->setType(TT_SelectorName
);
1059 if (!Contexts
.back().FirstObjCSelectorName
) {
1060 Contexts
.back().FirstObjCSelectorName
= Tok
->Previous
;
1061 } else if (Tok
->Previous
->ColumnWidth
>
1062 Contexts
.back().LongestObjCSelectorName
) {
1063 Contexts
.back().LongestObjCSelectorName
=
1064 Tok
->Previous
->ColumnWidth
;
1066 Tok
->Previous
->ParameterIndex
=
1067 Contexts
.back().FirstObjCSelectorName
->ObjCSelectorNameParts
;
1068 ++Contexts
.back().FirstObjCSelectorName
->ObjCSelectorNameParts
;
1070 } else if (Contexts
.back().ColonIsForRangeExpr
) {
1071 Tok
->setType(TT_RangeBasedForLoopColon
);
1072 } else if (Contexts
.back().ContextType
== Context::C11GenericSelection
) {
1073 Tok
->setType(TT_GenericSelectionColon
);
1074 } else if (CurrentToken
&& CurrentToken
->is(tok::numeric_constant
)) {
1075 Tok
->setType(TT_BitFieldColon
);
1076 } else if (Contexts
.size() == 1 &&
1077 !Line
.First
->isOneOf(tok::kw_enum
, tok::kw_case
,
1079 FormatToken
*Prev
= Tok
->getPreviousNonComment();
1082 if (Prev
->isOneOf(tok::r_paren
, tok::kw_noexcept
) ||
1083 Prev
->ClosesRequiresClause
) {
1084 Tok
->setType(TT_CtorInitializerColon
);
1085 } else if (Prev
->is(tok::kw_try
)) {
1086 // Member initializer list within function try block.
1087 FormatToken
*PrevPrev
= Prev
->getPreviousNonComment();
1090 if (PrevPrev
&& PrevPrev
->isOneOf(tok::r_paren
, tok::kw_noexcept
))
1091 Tok
->setType(TT_CtorInitializerColon
);
1093 Tok
->setType(TT_InheritanceColon
);
1095 } else if (canBeObjCSelectorComponent(*Tok
->Previous
) && Tok
->Next
&&
1096 (Tok
->Next
->isOneOf(tok::r_paren
, tok::comma
) ||
1097 (canBeObjCSelectorComponent(*Tok
->Next
) && Tok
->Next
->Next
&&
1098 Tok
->Next
->Next
->is(tok::colon
)))) {
1099 // This handles a special macro in ObjC code where selectors including
1100 // the colon are passed as macro arguments.
1101 Tok
->setType(TT_ObjCMethodExpr
);
1102 } else if (Contexts
.back().ContextKind
== tok::l_paren
&&
1103 !Line
.InPragmaDirective
) {
1104 Tok
->setType(TT_InlineASMColon
);
1109 // | and & in declarations/type expressions represent union and
1110 // intersection types, respectively.
1111 if (Style
.isJavaScript() && !Contexts
.back().IsExpression
)
1112 Tok
->setType(TT_JsTypeOperator
);
1116 CurrentToken
->isOneOf(tok::kw_constexpr
, tok::identifier
)) {
1121 if (CurrentToken
&& CurrentToken
->is(tok::l_paren
)) {
1123 if (!parseParens(/*LookForDecls=*/true))
1128 if (Style
.isJavaScript()) {
1129 // x.for and {for: ...}
1130 if ((Tok
->Previous
&& Tok
->Previous
->is(tok::period
)) ||
1131 (Tok
->Next
&& Tok
->Next
->is(tok::colon
))) {
1134 // JS' for await ( ...
1135 if (CurrentToken
&& CurrentToken
->is(Keywords
.kw_await
))
1138 if (Style
.isCpp() && CurrentToken
&& CurrentToken
->is(tok::kw_co_await
))
1140 Contexts
.back().ColonIsForRangeExpr
= true;
1141 if (!CurrentToken
|| CurrentToken
->isNot(tok::l_paren
))
1148 // When faced with 'operator()()', the kw_operator handler incorrectly
1149 // marks the first l_paren as a OverloadedOperatorLParen. Here, we make
1150 // the first two parens OverloadedOperators and the second l_paren an
1151 // OverloadedOperatorLParen.
1152 if (Tok
->Previous
&& Tok
->Previous
->is(tok::r_paren
) &&
1153 Tok
->Previous
->MatchingParen
&&
1154 Tok
->Previous
->MatchingParen
->is(TT_OverloadedOperatorLParen
)) {
1155 Tok
->Previous
->setType(TT_OverloadedOperator
);
1156 Tok
->Previous
->MatchingParen
->setType(TT_OverloadedOperator
);
1157 Tok
->setType(TT_OverloadedOperatorLParen
);
1160 if (Style
.isVerilog()) {
1161 // Identify the parameter list and port list in a module instantiation.
1162 // This is still needed when we already have
1163 // UnwrappedLineParser::parseVerilogHierarchyHeader because that
1164 // function is only responsible for the definition, not the
1166 auto IsInstancePort
= [&]() {
1167 const FormatToken
*Prev
= Tok
->getPreviousNonComment();
1168 const FormatToken
*PrevPrev
;
1169 // In the following example all 4 left parentheses will be treated as
1170 // 'TT_VerilogInstancePortLParen'.
1172 // module_x instance_1(port_1); // Case A.
1173 // module_x #(parameter_1) // Case B.
1174 // instance_2(port_1), // Case C.
1175 // instance_3(port_1); // Case D.
1176 if (!Prev
|| !(PrevPrev
= Prev
->getPreviousNonComment()))
1179 if (Keywords
.isVerilogIdentifier(*Prev
) &&
1180 Keywords
.isVerilogIdentifier(*PrevPrev
)) {
1184 if (Prev
->is(Keywords
.kw_verilogHash
) &&
1185 Keywords
.isVerilogIdentifier(*PrevPrev
)) {
1189 if (Keywords
.isVerilogIdentifier(*Prev
) && PrevPrev
->is(tok::r_paren
))
1192 if (Keywords
.isVerilogIdentifier(*Prev
) && PrevPrev
->is(tok::comma
)) {
1193 const FormatToken
*PrevParen
= PrevPrev
->getPreviousNonComment();
1194 if (PrevParen
->is(tok::r_paren
) && PrevParen
->MatchingParen
&&
1195 PrevParen
->MatchingParen
->is(TT_VerilogInstancePortLParen
)) {
1202 if (IsInstancePort())
1203 Tok
->setFinalizedType(TT_VerilogInstancePortLParen
);
1208 if (Line
.MustBeDeclaration
&& Contexts
.size() == 1 &&
1209 !Contexts
.back().IsExpression
&& !Line
.startsWith(TT_ObjCProperty
) &&
1210 !Tok
->isOneOf(TT_TypeDeclarationParen
, TT_RequiresExpressionLParen
) &&
1212 !Tok
->Previous
->isOneOf(tok::kw___attribute
, TT_RequiresClause
,
1213 TT_LeadingJavaAnnotation
))) {
1214 Line
.MightBeFunctionDecl
= true;
1222 if (Style
.Language
== FormatStyle::LK_TextProto
) {
1223 FormatToken
*Previous
= Tok
->getPreviousNonComment();
1224 if (Previous
&& Previous
->getType() != TT_DictLiteral
)
1225 Previous
->setType(TT_SelectorName
);
1227 Scopes
.push_back(getScopeType(*Tok
));
1233 Tok
->setType(TT_TemplateOpener
);
1234 // In TT_Proto, we must distignuish between:
1236 // msg < item: data >
1237 // msg: < item: data >
1238 // In TT_TextProto, map<key, value> does not occur.
1239 if (Style
.Language
== FormatStyle::LK_TextProto
||
1240 (Style
.Language
== FormatStyle::LK_Proto
&& Tok
->Previous
&&
1241 Tok
->Previous
->isOneOf(TT_SelectorName
, TT_DictLiteral
))) {
1242 Tok
->setType(TT_DictLiteral
);
1243 FormatToken
*Previous
= Tok
->getPreviousNonComment();
1244 if (Previous
&& Previous
->getType() != TT_DictLiteral
)
1245 Previous
->setType(TT_SelectorName
);
1248 Tok
->setType(TT_BinaryOperator
);
1249 NonTemplateLess
.insert(Tok
);
1258 // Don't pop scope when encountering unbalanced r_brace.
1259 if (!Scopes
.empty())
1261 // Lines can start with '}'.
1266 if (Style
.Language
!= FormatStyle::LK_TextProto
)
1267 Tok
->setType(TT_BinaryOperator
);
1268 if (Tok
->Previous
&& Tok
->Previous
->is(TT_TemplateCloser
))
1269 Tok
->SpacesRequiredBefore
= 1;
1271 case tok::kw_operator
:
1272 if (Style
.Language
== FormatStyle::LK_TextProto
||
1273 Style
.Language
== FormatStyle::LK_Proto
) {
1276 while (CurrentToken
&&
1277 !CurrentToken
->isOneOf(tok::l_paren
, tok::semi
, tok::r_paren
)) {
1278 if (CurrentToken
->isOneOf(tok::star
, tok::amp
))
1279 CurrentToken
->setType(TT_PointerOrReference
);
1280 auto Next
= CurrentToken
->getNextNonComment();
1283 if (Next
->is(tok::less
))
1289 auto Previous
= CurrentToken
->getPreviousNonComment();
1291 if (CurrentToken
->is(tok::comma
) && Previous
->isNot(tok::kw_operator
))
1293 if (Previous
->isOneOf(TT_BinaryOperator
, TT_UnaryOperator
, tok::comma
,
1294 tok::star
, tok::arrow
, tok::amp
, tok::ampamp
) ||
1295 // User defined literal.
1296 Previous
->TokenText
.startswith("\"\"")) {
1297 Previous
->setType(TT_OverloadedOperator
);
1298 if (CurrentToken
->isOneOf(tok::less
, tok::greater
))
1302 if (CurrentToken
&& CurrentToken
->is(tok::l_paren
))
1303 CurrentToken
->setType(TT_OverloadedOperatorLParen
);
1304 if (CurrentToken
&& CurrentToken
->Previous
->is(TT_BinaryOperator
))
1305 CurrentToken
->Previous
->setType(TT_OverloadedOperator
);
1308 if (Style
.isJavaScript() && Tok
->Next
&&
1309 Tok
->Next
->isOneOf(tok::semi
, tok::comma
, tok::colon
, tok::r_paren
,
1310 tok::r_brace
, tok::r_square
)) {
1311 // Question marks before semicolons, colons, etc. indicate optional
1312 // types (fields, parameters), e.g.
1313 // function(x?: string, y?) {...}
1315 Tok
->setType(TT_JsTypeOptionalQuestion
);
1318 // Declarations cannot be conditional expressions, this can only be part
1319 // of a type declaration.
1320 if (Line
.MustBeDeclaration
&& !Contexts
.back().IsExpression
&&
1321 Style
.isJavaScript()) {
1324 if (Style
.isCSharp()) {
1325 // `Type?)`, `Type?>`, `Type? name;` and `Type? name =` can only be
1328 // `Type?)`, `Type?>`, `Type? name;`
1330 (Tok
->Next
->startsSequence(tok::question
, tok::r_paren
) ||
1331 Tok
->Next
->startsSequence(tok::question
, tok::greater
) ||
1332 Tok
->Next
->startsSequence(tok::question
, tok::identifier
,
1334 Tok
->setType(TT_CSharpNullable
);
1339 if (Tok
->Next
&& Tok
->Next
->is(tok::identifier
) && Tok
->Next
->Next
&&
1340 Tok
->Next
->Next
->is(tok::equal
)) {
1341 Tok
->setType(TT_CSharpNullable
);
1345 // Line.MustBeDeclaration will be true for `Type? name;`.
1347 // cond ? "A" : "B";
1349 // cond ? cond2 ? "A" : "B" : "C";
1350 if (!Contexts
.back().IsExpression
&& Line
.MustBeDeclaration
&&
1352 !Tok
->Next
->isOneOf(tok::identifier
, tok::string_literal
) ||
1354 !Tok
->Next
->Next
->isOneOf(tok::colon
, tok::question
))) {
1355 Tok
->setType(TT_CSharpNullable
);
1361 case tok::kw_template
:
1362 parseTemplateDeclaration();
1365 switch (Contexts
.back().ContextType
) {
1366 case Context::CtorInitializer
:
1367 Tok
->setType(TT_CtorInitializerComma
);
1369 case Context::InheritanceList
:
1370 Tok
->setType(TT_InheritanceComma
);
1372 case Context::VerilogInstancePortList
:
1373 Tok
->setFinalizedType(TT_VerilogInstancePortComma
);
1376 if (Style
.isVerilog() && Contexts
.size() == 1 &&
1377 Line
.startsWith(Keywords
.kw_assign
)) {
1378 Tok
->setFinalizedType(TT_VerilogAssignComma
);
1379 } else if (Contexts
.back().FirstStartOfName
&&
1380 (Contexts
.size() == 1 || startsWithInitStatement(Line
))) {
1381 Contexts
.back().FirstStartOfName
->PartOfMultiVariableDeclStmt
= true;
1382 Line
.IsMultiVariableDeclStmt
= true;
1386 if (Contexts
.back().ContextType
== Context::ForEachMacro
)
1387 Contexts
.back().IsExpression
= true;
1389 case tok::kw_default
:
1390 // Unindent case labels.
1391 if (Style
.isVerilog() && Keywords
.isVerilogEndOfLabel(*Tok
) &&
1392 (Line
.Level
> 1 || (!Line
.InPPDirective
&& Line
.Level
> 0))) {
1396 case tok::identifier
:
1397 if (Tok
->isOneOf(Keywords
.kw___has_include
,
1398 Keywords
.kw___has_include_next
)) {
1401 if (Style
.isCSharp() && Tok
->is(Keywords
.kw_where
) && Tok
->Next
&&
1402 Tok
->Next
->isNot(tok::l_paren
)) {
1403 Tok
->setType(TT_CSharpGenericTypeConstraint
);
1404 parseCSharpGenericTypeConstraint();
1405 if (!Tok
->getPreviousNonComment())
1406 Line
.IsContinuation
= true;
1410 if (Tok
->isNot(TT_LambdaArrow
) && Tok
->Previous
&&
1411 Tok
->Previous
->is(tok::kw_noexcept
)) {
1412 Tok
->setType(TT_TrailingReturnArrow
);
1416 if (Style
.InsertNewlineAtEOF
&& Tok
->NewlinesBefore
== 0)
1417 Tok
->NewlinesBefore
= 1;
1425 void parseCSharpGenericTypeConstraint() {
1426 int OpenAngleBracketsCount
= 0;
1427 while (CurrentToken
) {
1428 if (CurrentToken
->is(tok::less
)) {
1429 // parseAngle is too greedy and will consume the whole line.
1430 CurrentToken
->setType(TT_TemplateOpener
);
1431 ++OpenAngleBracketsCount
;
1433 } else if (CurrentToken
->is(tok::greater
)) {
1434 CurrentToken
->setType(TT_TemplateCloser
);
1435 --OpenAngleBracketsCount
;
1437 } else if (CurrentToken
->is(tok::comma
) && OpenAngleBracketsCount
== 0) {
1438 // We allow line breaks after GenericTypeConstraintComma's
1439 // so do not flag commas in Generics as GenericTypeConstraintComma's.
1440 CurrentToken
->setType(TT_CSharpGenericTypeConstraintComma
);
1442 } else if (CurrentToken
->is(Keywords
.kw_where
)) {
1443 CurrentToken
->setType(TT_CSharpGenericTypeConstraint
);
1445 } else if (CurrentToken
->is(tok::colon
)) {
1446 CurrentToken
->setType(TT_CSharpGenericTypeConstraintColon
);
1454 void parseIncludeDirective() {
1455 if (CurrentToken
&& CurrentToken
->is(tok::less
)) {
1457 while (CurrentToken
) {
1458 // Mark tokens up to the trailing line comments as implicit string
1460 if (CurrentToken
->isNot(tok::comment
) &&
1461 !CurrentToken
->TokenText
.startswith("//")) {
1462 CurrentToken
->setType(TT_ImplicitStringLiteral
);
1469 void parseWarningOrError() {
1471 // We still want to format the whitespace left of the first token of the
1472 // warning or error.
1474 while (CurrentToken
) {
1475 CurrentToken
->setType(TT_ImplicitStringLiteral
);
1480 void parsePragma() {
1481 next(); // Consume "pragma".
1483 CurrentToken
->isOneOf(Keywords
.kw_mark
, Keywords
.kw_option
,
1484 Keywords
.kw_region
)) {
1485 bool IsMarkOrRegion
=
1486 CurrentToken
->isOneOf(Keywords
.kw_mark
, Keywords
.kw_region
);
1488 next(); // Consume first token (so we fix leading whitespace).
1489 while (CurrentToken
) {
1490 if (IsMarkOrRegion
|| CurrentToken
->Previous
->is(TT_BinaryOperator
))
1491 CurrentToken
->setType(TT_ImplicitStringLiteral
);
1497 void parseHasInclude() {
1498 if (!CurrentToken
|| CurrentToken
->isNot(tok::l_paren
))
1501 parseIncludeDirective();
1505 LineType
parsePreprocessorDirective() {
1506 bool IsFirstToken
= CurrentToken
->IsFirst
;
1507 LineType Type
= LT_PreprocessorDirective
;
1512 if (Style
.isJavaScript() && IsFirstToken
) {
1513 // JavaScript files can contain shebang lines of the form:
1514 // #!/usr/bin/env node
1515 // Treat these like C++ #include directives.
1516 while (CurrentToken
) {
1517 // Tokens cannot be comments here.
1518 CurrentToken
->setType(TT_ImplicitStringLiteral
);
1521 return LT_ImportStatement
;
1524 if (CurrentToken
->is(tok::numeric_constant
)) {
1525 CurrentToken
->SpacesRequiredBefore
= 1;
1528 // Hashes in the middle of a line can lead to any strange token
1530 if (!CurrentToken
->Tok
.getIdentifierInfo())
1532 // In Verilog macro expansions start with a backtick just like preprocessor
1533 // directives. Thus we stop if the word is not a preprocessor directive.
1534 if (Style
.isVerilog() && !Keywords
.isVerilogPPDirective(*CurrentToken
))
1536 switch (CurrentToken
->Tok
.getIdentifierInfo()->getPPKeywordID()) {
1537 case tok::pp_include
:
1538 case tok::pp_include_next
:
1539 case tok::pp_import
:
1541 parseIncludeDirective();
1542 Type
= LT_ImportStatement
;
1545 case tok::pp_warning
:
1546 parseWarningOrError();
1548 case tok::pp_pragma
:
1553 Contexts
.back().IsExpression
= true;
1560 while (CurrentToken
) {
1561 FormatToken
*Tok
= CurrentToken
;
1563 if (Tok
->is(tok::l_paren
)) {
1565 } else if (Tok
->isOneOf(Keywords
.kw___has_include
,
1566 Keywords
.kw___has_include_next
)) {
1574 LineType
parseLine() {
1577 NonTemplateLess
.clear();
1578 if (!Line
.InMacroBody
&& CurrentToken
->is(tok::hash
)) {
1579 // We were not yet allowed to use C++17 optional when this was being
1580 // written. So we used LT_Invalid to mark that the line is not a
1581 // preprocessor directive.
1582 auto Type
= parsePreprocessorDirective();
1583 if (Type
!= LT_Invalid
)
1587 // Directly allow to 'import <string-literal>' to support protocol buffer
1588 // definitions (github.com/google/protobuf) or missing "#" (either way we
1589 // should not break the line).
1590 IdentifierInfo
*Info
= CurrentToken
->Tok
.getIdentifierInfo();
1591 if ((Style
.Language
== FormatStyle::LK_Java
&&
1592 CurrentToken
->is(Keywords
.kw_package
)) ||
1593 (!Style
.isVerilog() && Info
&&
1594 Info
->getPPKeywordID() == tok::pp_import
&& CurrentToken
->Next
&&
1595 CurrentToken
->Next
->isOneOf(tok::string_literal
, tok::identifier
,
1598 parseIncludeDirective();
1599 return LT_ImportStatement
;
1602 // If this line starts and ends in '<' and '>', respectively, it is likely
1603 // part of "#define <a/b.h>".
1604 if (CurrentToken
->is(tok::less
) && Line
.Last
->is(tok::greater
)) {
1605 parseIncludeDirective();
1606 return LT_ImportStatement
;
1609 // In .proto files, top-level options and package statements are very
1610 // similar to import statements and should not be line-wrapped.
1611 if (Style
.Language
== FormatStyle::LK_Proto
&& Line
.Level
== 0 &&
1612 CurrentToken
->isOneOf(Keywords
.kw_option
, Keywords
.kw_package
)) {
1614 if (CurrentToken
&& CurrentToken
->is(tok::identifier
)) {
1615 while (CurrentToken
)
1617 return LT_ImportStatement
;
1621 bool KeywordVirtualFound
= false;
1622 bool ImportStatement
= false;
1624 // import {...} from '...';
1625 if (Style
.isJavaScript() && CurrentToken
->is(Keywords
.kw_import
))
1626 ImportStatement
= true;
1628 while (CurrentToken
) {
1629 if (CurrentToken
->is(tok::kw_virtual
))
1630 KeywordVirtualFound
= true;
1631 if (Style
.isJavaScript()) {
1632 // export {...} from '...';
1633 // An export followed by "from 'some string';" is a re-export from
1634 // another module identified by a URI and is treated as a
1635 // LT_ImportStatement (i.e. prevent wraps on it for long URIs).
1636 // Just "export {...};" or "export class ..." should not be treated as
1637 // an import in this sense.
1638 if (Line
.First
->is(tok::kw_export
) &&
1639 CurrentToken
->is(Keywords
.kw_from
) && CurrentToken
->Next
&&
1640 CurrentToken
->Next
->isStringLiteral()) {
1641 ImportStatement
= true;
1643 if (isClosureImportStatement(*CurrentToken
))
1644 ImportStatement
= true;
1646 if (!consumeToken())
1649 if (KeywordVirtualFound
)
1650 return LT_VirtualFunctionDecl
;
1651 if (ImportStatement
)
1652 return LT_ImportStatement
;
1654 if (Line
.startsWith(TT_ObjCMethodSpecifier
)) {
1655 if (Contexts
.back().FirstObjCSelectorName
) {
1656 Contexts
.back().FirstObjCSelectorName
->LongestObjCSelectorName
=
1657 Contexts
.back().LongestObjCSelectorName
;
1659 return LT_ObjCMethodDecl
;
1662 for (const auto &ctx
: Contexts
)
1663 if (ctx
.ContextType
== Context::StructArrayInitializer
)
1664 return LT_ArrayOfStructInitializer
;
1670 bool isClosureImportStatement(const FormatToken
&Tok
) {
1671 // FIXME: Closure-library specific stuff should not be hard-coded but be
1673 return Tok
.TokenText
== "goog" && Tok
.Next
&& Tok
.Next
->is(tok::period
) &&
1675 (Tok
.Next
->Next
->TokenText
== "module" ||
1676 Tok
.Next
->Next
->TokenText
== "provide" ||
1677 Tok
.Next
->Next
->TokenText
== "require" ||
1678 Tok
.Next
->Next
->TokenText
== "requireType" ||
1679 Tok
.Next
->Next
->TokenText
== "forwardDeclare") &&
1680 Tok
.Next
->Next
->Next
&& Tok
.Next
->Next
->Next
->is(tok::l_paren
);
1683 void resetTokenMetadata() {
1687 // Reset token type in case we have already looked at it and then
1688 // recovered from an error (e.g. failure to find the matching >).
1689 if (!CurrentToken
->isTypeFinalized() &&
1690 !CurrentToken
->isOneOf(
1691 TT_LambdaLSquare
, TT_LambdaLBrace
, TT_AttributeMacro
, TT_IfMacro
,
1692 TT_ForEachMacro
, TT_TypenameMacro
, TT_FunctionLBrace
,
1693 TT_ImplicitStringLiteral
, TT_InlineASMBrace
, TT_FatArrow
,
1694 TT_LambdaArrow
, TT_NamespaceMacro
, TT_OverloadedOperator
,
1695 TT_RegexLiteral
, TT_TemplateString
, TT_ObjCStringLiteral
,
1696 TT_UntouchableMacroFunc
, TT_StatementAttributeLikeMacro
,
1697 TT_FunctionLikeOrFreestandingMacro
, TT_ClassLBrace
, TT_EnumLBrace
,
1698 TT_RecordLBrace
, TT_StructLBrace
, TT_UnionLBrace
, TT_RequiresClause
,
1699 TT_RequiresClauseInARequiresExpression
, TT_RequiresExpression
,
1700 TT_RequiresExpressionLParen
, TT_RequiresExpressionLBrace
,
1701 TT_CompoundRequirementLBrace
, TT_BracedListLBrace
)) {
1702 CurrentToken
->setType(TT_Unknown
);
1704 CurrentToken
->Role
.reset();
1705 CurrentToken
->MatchingParen
= nullptr;
1706 CurrentToken
->FakeLParens
.clear();
1707 CurrentToken
->FakeRParens
= 0;
1714 CurrentToken
->NestingLevel
= Contexts
.size() - 1;
1715 CurrentToken
->BindingStrength
= Contexts
.back().BindingStrength
;
1716 modifyContext(*CurrentToken
);
1717 determineTokenType(*CurrentToken
);
1718 CurrentToken
= CurrentToken
->Next
;
1720 resetTokenMetadata();
1723 /// A struct to hold information valid in a specific context, e.g.
1724 /// a pair of parenthesis.
1726 Context(tok::TokenKind ContextKind
, unsigned BindingStrength
,
1728 : ContextKind(ContextKind
), BindingStrength(BindingStrength
),
1729 IsExpression(IsExpression
) {}
1731 tok::TokenKind ContextKind
;
1732 unsigned BindingStrength
;
1734 unsigned LongestObjCSelectorName
= 0;
1735 bool ColonIsForRangeExpr
= false;
1736 bool ColonIsDictLiteral
= false;
1737 bool ColonIsObjCMethodExpr
= false;
1738 FormatToken
*FirstObjCSelectorName
= nullptr;
1739 FormatToken
*FirstStartOfName
= nullptr;
1740 bool CanBeExpression
= true;
1741 bool CaretFound
= false;
1742 bool InCpp11AttributeSpecifier
= false;
1743 bool InCSharpAttributeSpecifier
= false;
1744 bool VerilogAssignmentFound
= false;
1745 // Whether the braces may mean concatenation instead of structure or array
1747 bool VerilogMayBeConcatenation
= false;
1750 // Like the part after `:` in a constructor.
1751 // Context(...) : IsExpression(IsExpression)
1753 // Like in the parentheses in a foreach.
1755 // Like the inheritance list in a class declaration.
1756 // class Input : public IO
1758 // Like in the braced list.
1760 StructArrayInitializer
,
1761 // Like in `static_cast<int>`.
1763 // C11 _Generic selection.
1764 C11GenericSelection
,
1765 // Like in the outer parentheses in `ffnand ff1(.q());`.
1766 VerilogInstancePortList
,
1767 } ContextType
= Unknown
;
1770 /// Puts a new \c Context onto the stack \c Contexts for the lifetime
1771 /// of each instance.
1772 struct ScopedContextCreator
{
1773 AnnotatingParser
&P
;
1775 ScopedContextCreator(AnnotatingParser
&P
, tok::TokenKind ContextKind
,
1778 P
.Contexts
.push_back(Context(ContextKind
,
1779 P
.Contexts
.back().BindingStrength
+ Increase
,
1780 P
.Contexts
.back().IsExpression
));
1783 ~ScopedContextCreator() {
1784 if (P
.Style
.AlignArrayOfStructures
!= FormatStyle::AIAS_None
) {
1785 if (P
.Contexts
.back().ContextType
== Context::StructArrayInitializer
) {
1786 P
.Contexts
.pop_back();
1787 P
.Contexts
.back().ContextType
= Context::StructArrayInitializer
;
1791 P
.Contexts
.pop_back();
1795 void modifyContext(const FormatToken
&Current
) {
1796 auto AssignmentStartsExpression
= [&]() {
1797 if (Current
.getPrecedence() != prec::Assignment
)
1800 if (Line
.First
->isOneOf(tok::kw_using
, tok::kw_return
))
1802 if (Line
.First
->is(tok::kw_template
)) {
1803 assert(Current
.Previous
);
1804 if (Current
.Previous
->is(tok::kw_operator
)) {
1805 // `template ... operator=` cannot be an expression.
1809 // `template` keyword can start a variable template.
1810 const FormatToken
*Tok
= Line
.First
->getNextNonComment();
1811 assert(Tok
); // Current token is on the same line.
1812 if (Tok
->isNot(TT_TemplateOpener
)) {
1813 // Explicit template instantiations do not have `<>`.
1817 // This is the default value of a template parameter, determine if it's
1818 // type or non-type.
1819 if (Contexts
.back().ContextKind
== tok::less
) {
1820 assert(Current
.Previous
->Previous
);
1821 return !Current
.Previous
->Previous
->isOneOf(tok::kw_typename
,
1825 Tok
= Tok
->MatchingParen
;
1828 Tok
= Tok
->getNextNonComment();
1832 if (Tok
->isOneOf(tok::kw_class
, tok::kw_enum
, tok::kw_struct
,
1840 // Type aliases use `type X = ...;` in TypeScript and can be exported
1841 // using `export type ...`.
1842 if (Style
.isJavaScript() &&
1843 (Line
.startsWith(Keywords
.kw_type
, tok::identifier
) ||
1844 Line
.startsWith(tok::kw_export
, Keywords
.kw_type
,
1845 tok::identifier
))) {
1849 return !Current
.Previous
|| Current
.Previous
->isNot(tok::kw_operator
);
1852 if (AssignmentStartsExpression()) {
1853 Contexts
.back().IsExpression
= true;
1854 if (!Line
.startsWith(TT_UnaryOperator
)) {
1855 for (FormatToken
*Previous
= Current
.Previous
;
1856 Previous
&& Previous
->Previous
&&
1857 !Previous
->Previous
->isOneOf(tok::comma
, tok::semi
);
1858 Previous
= Previous
->Previous
) {
1859 if (Previous
->isOneOf(tok::r_square
, tok::r_paren
, tok::greater
)) {
1860 Previous
= Previous
->MatchingParen
;
1864 if (Previous
->opensScope())
1866 if (Previous
->isOneOf(TT_BinaryOperator
, TT_UnaryOperator
) &&
1867 Previous
->isOneOf(tok::star
, tok::amp
, tok::ampamp
) &&
1868 Previous
->Previous
&& Previous
->Previous
->isNot(tok::equal
)) {
1869 Previous
->setType(TT_PointerOrReference
);
1873 } else if (Current
.is(tok::lessless
) &&
1874 (!Current
.Previous
||
1875 Current
.Previous
->isNot(tok::kw_operator
))) {
1876 Contexts
.back().IsExpression
= true;
1877 } else if (Current
.isOneOf(tok::kw_return
, tok::kw_throw
)) {
1878 Contexts
.back().IsExpression
= true;
1879 } else if (Current
.is(TT_TrailingReturnArrow
)) {
1880 Contexts
.back().IsExpression
= false;
1881 } else if (Current
.is(TT_LambdaArrow
) || Current
.is(Keywords
.kw_assert
)) {
1882 Contexts
.back().IsExpression
= Style
.Language
== FormatStyle::LK_Java
;
1883 } else if (Current
.Previous
&&
1884 Current
.Previous
->is(TT_CtorInitializerColon
)) {
1885 Contexts
.back().IsExpression
= true;
1886 Contexts
.back().ContextType
= Context::CtorInitializer
;
1887 } else if (Current
.Previous
&& Current
.Previous
->is(TT_InheritanceColon
)) {
1888 Contexts
.back().ContextType
= Context::InheritanceList
;
1889 } else if (Current
.isOneOf(tok::r_paren
, tok::greater
, tok::comma
)) {
1890 for (FormatToken
*Previous
= Current
.Previous
;
1891 Previous
&& Previous
->isOneOf(tok::star
, tok::amp
);
1892 Previous
= Previous
->Previous
) {
1893 Previous
->setType(TT_PointerOrReference
);
1895 if (Line
.MustBeDeclaration
&&
1896 Contexts
.front().ContextType
!= Context::CtorInitializer
) {
1897 Contexts
.back().IsExpression
= false;
1899 } else if (Current
.is(tok::kw_new
)) {
1900 Contexts
.back().CanBeExpression
= false;
1901 } else if (Current
.is(tok::semi
) ||
1902 (Current
.is(tok::exclaim
) && Current
.Previous
&&
1903 Current
.Previous
->isNot(tok::kw_operator
))) {
1904 // This should be the condition or increment in a for-loop.
1905 // But not operator !() (can't use TT_OverloadedOperator here as its not
1906 // been annotated yet).
1907 Contexts
.back().IsExpression
= true;
1911 static FormatToken
*untilMatchingParen(FormatToken
*Current
) {
1912 // Used when `MatchingParen` is not yet established.
1915 if (Current
->is(tok::l_paren
))
1917 if (Current
->is(tok::r_paren
))
1921 Current
= Current
->Next
;
1926 static bool isDeductionGuide(FormatToken
&Current
) {
1927 // Look for a deduction guide template<T> A(...) -> A<...>;
1928 if (Current
.Previous
&& Current
.Previous
->is(tok::r_paren
) &&
1929 Current
.startsSequence(tok::arrow
, tok::identifier
, tok::less
)) {
1930 // Find the TemplateCloser.
1931 FormatToken
*TemplateCloser
= Current
.Next
->Next
;
1932 int NestingLevel
= 0;
1933 while (TemplateCloser
) {
1934 // Skip over an expressions in parens A<(3 < 2)>;
1935 if (TemplateCloser
->is(tok::l_paren
)) {
1936 // No Matching Paren yet so skip to matching paren
1937 TemplateCloser
= untilMatchingParen(TemplateCloser
);
1938 if (!TemplateCloser
)
1941 if (TemplateCloser
->is(tok::less
))
1943 if (TemplateCloser
->is(tok::greater
))
1945 if (NestingLevel
< 1)
1947 TemplateCloser
= TemplateCloser
->Next
;
1949 // Assuming we have found the end of the template ensure its followed
1950 // with a semi-colon.
1951 if (TemplateCloser
&& TemplateCloser
->Next
&&
1952 TemplateCloser
->Next
->is(tok::semi
) &&
1953 Current
.Previous
->MatchingParen
) {
1954 // Determine if the identifier `A` prior to the A<..>; is the same as
1955 // prior to the A(..)
1956 FormatToken
*LeadingIdentifier
=
1957 Current
.Previous
->MatchingParen
->Previous
;
1959 return LeadingIdentifier
&&
1960 LeadingIdentifier
->TokenText
== Current
.Next
->TokenText
;
1966 void determineTokenType(FormatToken
&Current
) {
1967 if (Current
.isNot(TT_Unknown
)) {
1968 // The token type is already known.
1972 if ((Style
.isJavaScript() || Style
.isCSharp()) &&
1973 Current
.is(tok::exclaim
)) {
1974 if (Current
.Previous
) {
1976 Style
.isJavaScript()
1977 ? Keywords
.IsJavaScriptIdentifier(
1978 *Current
.Previous
, /* AcceptIdentifierName= */ true)
1979 : Current
.Previous
->is(tok::identifier
);
1981 Current
.Previous
->isOneOf(
1982 tok::kw_default
, tok::kw_namespace
, tok::r_paren
, tok::r_square
,
1983 tok::r_brace
, tok::kw_false
, tok::kw_true
, Keywords
.kw_type
,
1984 Keywords
.kw_get
, Keywords
.kw_init
, Keywords
.kw_set
) ||
1985 Current
.Previous
->Tok
.isLiteral()) {
1986 Current
.setType(TT_NonNullAssertion
);
1991 Current
.Next
->isOneOf(TT_BinaryOperator
, Keywords
.kw_as
)) {
1992 Current
.setType(TT_NonNullAssertion
);
1997 // Line.MightBeFunctionDecl can only be true after the parentheses of a
1998 // function declaration have been found. In this case, 'Current' is a
1999 // trailing token of this declaration and thus cannot be a name.
2000 if (Current
.is(Keywords
.kw_instanceof
)) {
2001 Current
.setType(TT_BinaryOperator
);
2002 } else if (isStartOfName(Current
) &&
2003 (!Line
.MightBeFunctionDecl
|| Current
.NestingLevel
!= 0)) {
2004 Contexts
.back().FirstStartOfName
= &Current
;
2005 Current
.setType(TT_StartOfName
);
2006 } else if (Current
.is(tok::semi
)) {
2007 // Reset FirstStartOfName after finding a semicolon so that a for loop
2008 // with multiple increment statements is not confused with a for loop
2009 // having multiple variable declarations.
2010 Contexts
.back().FirstStartOfName
= nullptr;
2011 } else if (Current
.isOneOf(tok::kw_auto
, tok::kw___auto_type
)) {
2013 } else if (Current
.is(tok::arrow
) &&
2014 Style
.Language
== FormatStyle::LK_Java
) {
2015 Current
.setType(TT_LambdaArrow
);
2016 } else if (Current
.is(tok::arrow
) && AutoFound
&&
2017 (Line
.MightBeFunctionDecl
|| Line
.InPPDirective
) &&
2018 Current
.NestingLevel
== 0 &&
2019 !Current
.Previous
->isOneOf(tok::kw_operator
, tok::identifier
)) {
2020 // not auto operator->() -> xxx;
2021 Current
.setType(TT_TrailingReturnArrow
);
2022 } else if (Current
.is(tok::arrow
) && Current
.Previous
&&
2023 Current
.Previous
->is(tok::r_brace
)) {
2024 // Concept implicit conversion constraint needs to be treated like
2025 // a trailing return type ... } -> <type>.
2026 Current
.setType(TT_TrailingReturnArrow
);
2027 } else if (isDeductionGuide(Current
)) {
2028 // Deduction guides trailing arrow " A(...) -> A<T>;".
2029 Current
.setType(TT_TrailingReturnArrow
);
2030 } else if (Current
.isOneOf(tok::star
, tok::amp
, tok::ampamp
)) {
2031 Current
.setType(determineStarAmpUsage(
2033 Contexts
.back().CanBeExpression
&& Contexts
.back().IsExpression
,
2034 Contexts
.back().ContextType
== Context::TemplateArgument
));
2035 } else if (Current
.isOneOf(tok::minus
, tok::plus
, tok::caret
) ||
2036 (Style
.isVerilog() && Current
.is(tok::pipe
))) {
2037 Current
.setType(determinePlusMinusCaretUsage(Current
));
2038 if (Current
.is(TT_UnaryOperator
) && Current
.is(tok::caret
))
2039 Contexts
.back().CaretFound
= true;
2040 } else if (Current
.isOneOf(tok::minusminus
, tok::plusplus
)) {
2041 Current
.setType(determineIncrementUsage(Current
));
2042 } else if (Current
.isOneOf(tok::exclaim
, tok::tilde
)) {
2043 Current
.setType(TT_UnaryOperator
);
2044 } else if (Current
.is(tok::question
)) {
2045 if (Style
.isJavaScript() && Line
.MustBeDeclaration
&&
2046 !Contexts
.back().IsExpression
) {
2047 // In JavaScript, `interface X { foo?(): bar; }` is an optional method
2048 // on the interface, not a ternary expression.
2049 Current
.setType(TT_JsTypeOptionalQuestion
);
2051 Current
.setType(TT_ConditionalExpr
);
2053 } else if (Current
.isBinaryOperator() &&
2054 (!Current
.Previous
|| Current
.Previous
->isNot(tok::l_square
)) &&
2055 (Current
.isNot(tok::greater
) &&
2056 Style
.Language
!= FormatStyle::LK_TextProto
)) {
2057 if (Style
.isVerilog()) {
2058 if (Current
.is(tok::lessequal
) && Contexts
.size() == 1 &&
2059 !Contexts
.back().VerilogAssignmentFound
) {
2060 // In Verilog `<=` is assignment if in its own statement. It is a
2061 // statement instead of an expression, that is it can not be chained.
2062 Current
.ForcedPrecedence
= prec::Assignment
;
2063 Current
.setFinalizedType(TT_BinaryOperator
);
2065 if (Current
.getPrecedence() == prec::Assignment
)
2066 Contexts
.back().VerilogAssignmentFound
= true;
2068 Current
.setType(TT_BinaryOperator
);
2069 } else if (Current
.is(tok::comment
)) {
2070 if (Current
.TokenText
.startswith("/*")) {
2071 if (Current
.TokenText
.endswith("*/")) {
2072 Current
.setType(TT_BlockComment
);
2074 // The lexer has for some reason determined a comment here. But we
2075 // cannot really handle it, if it isn't properly terminated.
2076 Current
.Tok
.setKind(tok::unknown
);
2079 Current
.setType(TT_LineComment
);
2081 } else if (Current
.is(tok::string_literal
)) {
2082 if (Style
.isVerilog() && Contexts
.back().VerilogMayBeConcatenation
&&
2083 Current
.getPreviousNonComment() &&
2084 Current
.getPreviousNonComment()->isOneOf(tok::comma
, tok::l_brace
) &&
2085 Current
.getNextNonComment() &&
2086 Current
.getNextNonComment()->isOneOf(tok::comma
, tok::r_brace
)) {
2087 Current
.setType(TT_StringInConcatenation
);
2089 } else if (Current
.is(tok::l_paren
)) {
2090 if (lParenStartsCppCast(Current
))
2091 Current
.setType(TT_CppCastLParen
);
2092 } else if (Current
.is(tok::r_paren
)) {
2093 if (rParenEndsCast(Current
))
2094 Current
.setType(TT_CastRParen
);
2095 if (Current
.MatchingParen
&& Current
.Next
&&
2096 !Current
.Next
->isBinaryOperator() &&
2097 !Current
.Next
->isOneOf(tok::semi
, tok::colon
, tok::l_brace
,
2098 tok::comma
, tok::period
, tok::arrow
,
2099 tok::coloncolon
, tok::kw_noexcept
)) {
2100 if (FormatToken
*AfterParen
= Current
.MatchingParen
->Next
) {
2101 // Make sure this isn't the return type of an Obj-C block declaration
2102 if (AfterParen
->isNot(tok::caret
)) {
2103 if (FormatToken
*BeforeParen
= Current
.MatchingParen
->Previous
) {
2104 if (BeforeParen
->is(tok::identifier
) &&
2105 BeforeParen
->isNot(TT_TypenameMacro
) &&
2106 BeforeParen
->TokenText
== BeforeParen
->TokenText
.upper() &&
2107 (!BeforeParen
->Previous
||
2108 BeforeParen
->Previous
->ClosesTemplateDeclaration
)) {
2109 Current
.setType(TT_FunctionAnnotationRParen
);
2115 } else if (Current
.is(tok::at
) && Current
.Next
&& !Style
.isJavaScript() &&
2116 Style
.Language
!= FormatStyle::LK_Java
) {
2117 // In Java & JavaScript, "@..." is a decorator or annotation. In ObjC, it
2118 // marks declarations and properties that need special formatting.
2119 switch (Current
.Next
->Tok
.getObjCKeywordID()) {
2120 case tok::objc_interface
:
2121 case tok::objc_implementation
:
2122 case tok::objc_protocol
:
2123 Current
.setType(TT_ObjCDecl
);
2125 case tok::objc_property
:
2126 Current
.setType(TT_ObjCProperty
);
2131 } else if (Current
.is(tok::period
)) {
2132 FormatToken
*PreviousNoComment
= Current
.getPreviousNonComment();
2133 if (PreviousNoComment
&&
2134 PreviousNoComment
->isOneOf(tok::comma
, tok::l_brace
)) {
2135 Current
.setType(TT_DesignatedInitializerPeriod
);
2136 } else if (Style
.Language
== FormatStyle::LK_Java
&& Current
.Previous
&&
2137 Current
.Previous
->isOneOf(TT_JavaAnnotation
,
2138 TT_LeadingJavaAnnotation
)) {
2139 Current
.setType(Current
.Previous
->getType());
2141 } else if (canBeObjCSelectorComponent(Current
) &&
2142 // FIXME(bug 36976): ObjC return types shouldn't use
2144 Current
.Previous
&& Current
.Previous
->is(TT_CastRParen
) &&
2145 Current
.Previous
->MatchingParen
&&
2146 Current
.Previous
->MatchingParen
->Previous
&&
2147 Current
.Previous
->MatchingParen
->Previous
->is(
2148 TT_ObjCMethodSpecifier
)) {
2149 // This is the first part of an Objective-C selector name. (If there's no
2150 // colon after this, this is the only place which annotates the identifier
2152 Current
.setType(TT_SelectorName
);
2153 } else if (Current
.isOneOf(tok::identifier
, tok::kw_const
, tok::kw_noexcept
,
2154 tok::kw_requires
) &&
2156 !Current
.Previous
->isOneOf(tok::equal
, tok::at
,
2157 TT_CtorInitializerComma
,
2158 TT_CtorInitializerColon
) &&
2159 Line
.MightBeFunctionDecl
&& Contexts
.size() == 1) {
2160 // Line.MightBeFunctionDecl can only be true after the parentheses of a
2161 // function declaration have been found.
2162 Current
.setType(TT_TrailingAnnotation
);
2163 } else if ((Style
.Language
== FormatStyle::LK_Java
||
2164 Style
.isJavaScript()) &&
2166 if (Current
.Previous
->is(tok::at
) &&
2167 Current
.isNot(Keywords
.kw_interface
)) {
2168 const FormatToken
&AtToken
= *Current
.Previous
;
2169 const FormatToken
*Previous
= AtToken
.getPreviousNonComment();
2170 if (!Previous
|| Previous
->is(TT_LeadingJavaAnnotation
))
2171 Current
.setType(TT_LeadingJavaAnnotation
);
2173 Current
.setType(TT_JavaAnnotation
);
2174 } else if (Current
.Previous
->is(tok::period
) &&
2175 Current
.Previous
->isOneOf(TT_JavaAnnotation
,
2176 TT_LeadingJavaAnnotation
)) {
2177 Current
.setType(Current
.Previous
->getType());
2182 /// Take a guess at whether \p Tok starts a name of a function or
2183 /// variable declaration.
2185 /// This is a heuristic based on whether \p Tok is an identifier following
2186 /// something that is likely a type.
2187 bool isStartOfName(const FormatToken
&Tok
) {
2188 // Handled in ExpressionParser for Verilog.
2189 if (Style
.isVerilog())
2192 if (Tok
.isNot(tok::identifier
) || !Tok
.Previous
)
2195 if (Tok
.Previous
->isOneOf(TT_LeadingJavaAnnotation
, Keywords
.kw_instanceof
,
2199 if (Style
.isJavaScript() && Tok
.Previous
->is(Keywords
.kw_in
))
2202 // Skip "const" as it does not have an influence on whether this is a name.
2203 FormatToken
*PreviousNotConst
= Tok
.getPreviousNonComment();
2205 // For javascript const can be like "let" or "var"
2206 if (!Style
.isJavaScript())
2207 while (PreviousNotConst
&& PreviousNotConst
->is(tok::kw_const
))
2208 PreviousNotConst
= PreviousNotConst
->getPreviousNonComment();
2210 if (!PreviousNotConst
)
2213 if (PreviousNotConst
->ClosesRequiresClause
)
2216 bool IsPPKeyword
= PreviousNotConst
->is(tok::identifier
) &&
2217 PreviousNotConst
->Previous
&&
2218 PreviousNotConst
->Previous
->is(tok::hash
);
2220 if (PreviousNotConst
->is(TT_TemplateCloser
)) {
2221 return PreviousNotConst
&& PreviousNotConst
->MatchingParen
&&
2222 PreviousNotConst
->MatchingParen
->Previous
&&
2223 PreviousNotConst
->MatchingParen
->Previous
->isNot(tok::period
) &&
2224 PreviousNotConst
->MatchingParen
->Previous
->isNot(tok::kw_template
);
2227 if (PreviousNotConst
->is(tok::r_paren
) &&
2228 PreviousNotConst
->is(TT_TypeDeclarationParen
)) {
2232 // If is a preprocess keyword like #define.
2237 if (PreviousNotConst
->isOneOf(tok::identifier
, tok::kw_auto
))
2241 if (PreviousNotConst
->is(TT_PointerOrReference
))
2245 if (PreviousNotConst
->isSimpleTypeSpecifier())
2249 if (Style
.Language
== FormatStyle::LK_Java
&&
2250 PreviousNotConst
->is(tok::r_square
)) {
2254 // const a = in JavaScript.
2255 return Style
.isJavaScript() && PreviousNotConst
->is(tok::kw_const
);
2258 /// Determine whether '(' is starting a C++ cast.
2259 bool lParenStartsCppCast(const FormatToken
&Tok
) {
2260 // C-style casts are only used in C++.
2264 FormatToken
*LeftOfParens
= Tok
.getPreviousNonComment();
2265 if (LeftOfParens
&& LeftOfParens
->is(TT_TemplateCloser
) &&
2266 LeftOfParens
->MatchingParen
) {
2267 auto *Prev
= LeftOfParens
->MatchingParen
->getPreviousNonComment();
2269 Prev
->isOneOf(tok::kw_const_cast
, tok::kw_dynamic_cast
,
2270 tok::kw_reinterpret_cast
, tok::kw_static_cast
)) {
2271 // FIXME: Maybe we should handle identifiers ending with "_cast",
2279 /// Determine whether ')' is ending a cast.
2280 bool rParenEndsCast(const FormatToken
&Tok
) {
2281 // C-style casts are only used in C++, C# and Java.
2282 if (!Style
.isCSharp() && !Style
.isCpp() &&
2283 Style
.Language
!= FormatStyle::LK_Java
) {
2287 // Empty parens aren't casts and there are no casts at the end of the line.
2288 if (Tok
.Previous
== Tok
.MatchingParen
|| !Tok
.Next
|| !Tok
.MatchingParen
)
2291 if (Tok
.MatchingParen
->is(TT_OverloadedOperatorLParen
))
2294 FormatToken
*LeftOfParens
= Tok
.MatchingParen
->getPreviousNonComment();
2296 // If there is a closing parenthesis left of the current
2297 // parentheses, look past it as these might be chained casts.
2298 if (LeftOfParens
->is(tok::r_paren
) &&
2299 LeftOfParens
->isNot(TT_CastRParen
)) {
2300 if (!LeftOfParens
->MatchingParen
||
2301 !LeftOfParens
->MatchingParen
->Previous
) {
2304 LeftOfParens
= LeftOfParens
->MatchingParen
->Previous
;
2307 if (LeftOfParens
->is(tok::r_square
)) {
2308 // delete[] (void *)ptr;
2309 auto MayBeArrayDelete
= [](FormatToken
*Tok
) -> FormatToken
* {
2310 if (Tok
->isNot(tok::r_square
))
2313 Tok
= Tok
->getPreviousNonComment();
2314 if (!Tok
|| Tok
->isNot(tok::l_square
))
2317 Tok
= Tok
->getPreviousNonComment();
2318 if (!Tok
|| Tok
->isNot(tok::kw_delete
))
2322 if (FormatToken
*MaybeDelete
= MayBeArrayDelete(LeftOfParens
))
2323 LeftOfParens
= MaybeDelete
;
2326 // The Condition directly below this one will see the operator arguments
2327 // as a (void *foo) cast.
2328 // void operator delete(void *foo) ATTRIB;
2329 if (LeftOfParens
->Tok
.getIdentifierInfo() && LeftOfParens
->Previous
&&
2330 LeftOfParens
->Previous
->is(tok::kw_operator
)) {
2334 // If there is an identifier (or with a few exceptions a keyword) right
2335 // before the parentheses, this is unlikely to be a cast.
2336 if (LeftOfParens
->Tok
.getIdentifierInfo() &&
2337 !LeftOfParens
->isOneOf(Keywords
.kw_in
, tok::kw_return
, tok::kw_case
,
2338 tok::kw_delete
, tok::kw_throw
)) {
2342 // Certain other tokens right before the parentheses are also signals that
2343 // this cannot be a cast.
2344 if (LeftOfParens
->isOneOf(tok::at
, tok::r_square
, TT_OverloadedOperator
,
2345 TT_TemplateCloser
, tok::ellipsis
)) {
2350 if (Tok
.Next
->is(tok::question
))
2353 // `foreach((A a, B b) in someList)` should not be seen as a cast.
2354 if (Tok
.Next
->is(Keywords
.kw_in
) && Style
.isCSharp())
2357 // Functions which end with decorations like volatile, noexcept are unlikely
2359 if (Tok
.Next
->isOneOf(tok::kw_noexcept
, tok::kw_volatile
, tok::kw_const
,
2360 tok::kw_requires
, tok::kw_throw
, tok::arrow
,
2361 Keywords
.kw_override
, Keywords
.kw_final
) ||
2362 isCppAttribute(Style
.isCpp(), *Tok
.Next
)) {
2366 // As Java has no function types, a "(" after the ")" likely means that this
2368 if (Style
.Language
== FormatStyle::LK_Java
&& Tok
.Next
->is(tok::l_paren
))
2371 // If a (non-string) literal follows, this is likely a cast.
2372 if (Tok
.Next
->isNot(tok::string_literal
) &&
2373 (Tok
.Next
->Tok
.isLiteral() ||
2374 Tok
.Next
->isOneOf(tok::kw_sizeof
, tok::kw_alignof
))) {
2378 // Heuristically try to determine whether the parentheses contain a type.
2379 auto IsQualifiedPointerOrReference
= [](FormatToken
*T
) {
2380 // This is used to handle cases such as x = (foo *const)&y;
2381 assert(!T
->isSimpleTypeSpecifier() && "Should have already been checked");
2382 // Strip trailing qualifiers such as const or volatile when checking
2383 // whether the parens could be a cast to a pointer/reference type.
2385 if (T
->is(TT_AttributeParen
)) {
2386 // Handle `x = (foo *__attribute__((foo)))&v;`:
2387 if (T
->MatchingParen
&& T
->MatchingParen
->Previous
&&
2388 T
->MatchingParen
->Previous
->is(tok::kw___attribute
)) {
2389 T
= T
->MatchingParen
->Previous
->Previous
;
2392 } else if (T
->is(TT_AttributeSquare
)) {
2393 // Handle `x = (foo *[[clang::foo]])&v;`:
2394 if (T
->MatchingParen
&& T
->MatchingParen
->Previous
) {
2395 T
= T
->MatchingParen
->Previous
;
2398 } else if (T
->canBePointerOrReferenceQualifier()) {
2404 return T
&& T
->is(TT_PointerOrReference
);
2406 bool ParensAreType
=
2408 Tok
.Previous
->isOneOf(TT_TemplateCloser
, TT_TypeDeclarationParen
) ||
2409 Tok
.Previous
->isSimpleTypeSpecifier() ||
2410 IsQualifiedPointerOrReference(Tok
.Previous
);
2411 bool ParensCouldEndDecl
=
2412 Tok
.Next
->isOneOf(tok::equal
, tok::semi
, tok::l_brace
, tok::greater
);
2413 if (ParensAreType
&& !ParensCouldEndDecl
)
2416 // At this point, we heuristically assume that there are no casts at the
2417 // start of the line. We assume that we have found most cases where there
2418 // are by the logic above, e.g. "(void)x;".
2422 // Certain token types inside the parentheses mean that this can't be a
2424 for (const FormatToken
*Token
= Tok
.MatchingParen
->Next
; Token
!= &Tok
;
2425 Token
= Token
->Next
) {
2426 if (Token
->is(TT_BinaryOperator
))
2430 // If the following token is an identifier or 'this', this is a cast. All
2431 // cases where this can be something else are handled above.
2432 if (Tok
.Next
->isOneOf(tok::identifier
, tok::kw_this
))
2435 // Look for a cast `( x ) (`.
2436 if (Tok
.Next
->is(tok::l_paren
) && Tok
.Previous
&& Tok
.Previous
->Previous
) {
2437 if (Tok
.Previous
->is(tok::identifier
) &&
2438 Tok
.Previous
->Previous
->is(tok::l_paren
)) {
2443 if (!Tok
.Next
->Next
)
2446 // If the next token after the parenthesis is a unary operator, assume
2447 // that this is cast, unless there are unexpected tokens inside the
2449 const bool NextIsAmpOrStar
= Tok
.Next
->isOneOf(tok::amp
, tok::star
);
2450 if (!(Tok
.Next
->isUnaryOperator() || NextIsAmpOrStar
) ||
2451 Tok
.Next
->is(tok::plus
) ||
2452 !Tok
.Next
->Next
->isOneOf(tok::identifier
, tok::numeric_constant
)) {
2455 if (NextIsAmpOrStar
&&
2456 (Tok
.Next
->Next
->is(tok::numeric_constant
) || Line
.InPPDirective
)) {
2459 // Search for unexpected tokens.
2460 for (FormatToken
*Prev
= Tok
.Previous
; Prev
!= Tok
.MatchingParen
;
2461 Prev
= Prev
->Previous
) {
2462 if (!Prev
->isOneOf(tok::kw_const
, tok::identifier
, tok::coloncolon
))
2468 /// Returns true if the token is used as a unary operator.
2469 bool determineUnaryOperatorByUsage(const FormatToken
&Tok
) {
2470 const FormatToken
*PrevToken
= Tok
.getPreviousNonComment();
2474 // These keywords are deliberately not included here because they may
2475 // precede only one of unary star/amp and plus/minus but not both. They are
2476 // either included in determineStarAmpUsage or determinePlusMinusCaretUsage.
2478 // @ - It may be followed by a unary `-` in Objective-C literals. We don't
2479 // know how they can be followed by a star or amp.
2480 if (PrevToken
->isOneOf(
2481 TT_ConditionalExpr
, tok::l_paren
, tok::comma
, tok::colon
, tok::semi
,
2482 tok::equal
, tok::question
, tok::l_square
, tok::l_brace
,
2483 tok::kw_case
, tok::kw_co_await
, tok::kw_co_return
, tok::kw_co_yield
,
2484 tok::kw_delete
, tok::kw_return
, tok::kw_throw
)) {
2488 // We put sizeof here instead of only in determineStarAmpUsage. In the cases
2489 // where the unary `+` operator is overloaded, it is reasonable to write
2490 // things like `sizeof +x`. Like commit 446d6ec996c6c3.
2491 if (PrevToken
->is(tok::kw_sizeof
))
2494 // A sequence of leading unary operators.
2495 if (PrevToken
->isOneOf(TT_CastRParen
, TT_UnaryOperator
))
2498 // There can't be two consecutive binary operators.
2499 if (PrevToken
->is(TT_BinaryOperator
))
2505 /// Return the type of the given token assuming it is * or &.
2506 TokenType
determineStarAmpUsage(const FormatToken
&Tok
, bool IsExpression
,
2507 bool InTemplateArgument
) {
2508 if (Style
.isJavaScript())
2509 return TT_BinaryOperator
;
2511 // && in C# must be a binary operator.
2512 if (Style
.isCSharp() && Tok
.is(tok::ampamp
))
2513 return TT_BinaryOperator
;
2515 if (Style
.isVerilog()) {
2516 // In Verilog, `*` can only be a binary operator. `&` can be either unary
2517 // or binary. `*` also includes `*>` in module path declarations in
2518 // specify blocks because merged tokens take the type of the first one by
2520 if (Tok
.is(tok::star
))
2521 return TT_BinaryOperator
;
2522 return determineUnaryOperatorByUsage(Tok
) ? TT_UnaryOperator
2523 : TT_BinaryOperator
;
2526 const FormatToken
*PrevToken
= Tok
.getPreviousNonComment();
2528 return TT_UnaryOperator
;
2529 if (PrevToken
->is(TT_TypeName
))
2530 return TT_PointerOrReference
;
2532 const FormatToken
*NextToken
= Tok
.getNextNonComment();
2534 if (InTemplateArgument
&& NextToken
&& NextToken
->is(tok::kw_noexcept
))
2535 return TT_BinaryOperator
;
2538 NextToken
->isOneOf(tok::arrow
, tok::equal
, tok::comma
, tok::r_paren
,
2539 TT_RequiresClause
) ||
2540 (NextToken
->is(tok::kw_noexcept
) && !IsExpression
) ||
2541 NextToken
->canBePointerOrReferenceQualifier() ||
2542 (NextToken
->is(tok::l_brace
) && !NextToken
->getNextNonComment())) {
2543 return TT_PointerOrReference
;
2546 if (PrevToken
->is(tok::coloncolon
))
2547 return TT_PointerOrReference
;
2549 if (PrevToken
->is(tok::r_paren
) && PrevToken
->is(TT_TypeDeclarationParen
))
2550 return TT_PointerOrReference
;
2552 if (determineUnaryOperatorByUsage(Tok
))
2553 return TT_UnaryOperator
;
2555 if (NextToken
->is(tok::l_square
) && NextToken
->isNot(TT_LambdaLSquare
))
2556 return TT_PointerOrReference
;
2557 if (NextToken
->is(tok::kw_operator
) && !IsExpression
)
2558 return TT_PointerOrReference
;
2559 if (NextToken
->isOneOf(tok::comma
, tok::semi
))
2560 return TT_PointerOrReference
;
2562 // After right braces, star tokens are likely to be pointers to struct,
2565 // This by itself is not sufficient to distinguish from multiplication
2566 // following a brace-initialized expression, as in:
2567 // int i = int{42} * 2;
2568 // In the struct case, the part of the struct declaration until the `{` and
2569 // the `}` are put on separate unwrapped lines; in the brace-initialized
2570 // case, the matching `{` is on the same unwrapped line, so check for the
2571 // presence of the matching brace to distinguish between those.
2572 if (PrevToken
->is(tok::r_brace
) && Tok
.is(tok::star
) &&
2573 !PrevToken
->MatchingParen
) {
2574 return TT_PointerOrReference
;
2577 if (PrevToken
->endsSequence(tok::r_square
, tok::l_square
, tok::kw_delete
))
2578 return TT_UnaryOperator
;
2580 if (PrevToken
->Tok
.isLiteral() ||
2581 PrevToken
->isOneOf(tok::r_paren
, tok::r_square
, tok::kw_true
,
2582 tok::kw_false
, tok::r_brace
)) {
2583 return TT_BinaryOperator
;
2586 const FormatToken
*NextNonParen
= NextToken
;
2587 while (NextNonParen
&& NextNonParen
->is(tok::l_paren
))
2588 NextNonParen
= NextNonParen
->getNextNonComment();
2589 if (NextNonParen
&& (NextNonParen
->Tok
.isLiteral() ||
2590 NextNonParen
->isOneOf(tok::kw_true
, tok::kw_false
) ||
2591 NextNonParen
->isUnaryOperator())) {
2592 return TT_BinaryOperator
;
2595 // If we know we're in a template argument, there are no named declarations.
2596 // Thus, having an identifier on the right-hand side indicates a binary
2598 if (InTemplateArgument
&& NextToken
->Tok
.isAnyIdentifier())
2599 return TT_BinaryOperator
;
2601 // "&&(" is quite unlikely to be two successive unary "&".
2602 if (Tok
.is(tok::ampamp
) && NextToken
->is(tok::l_paren
))
2603 return TT_BinaryOperator
;
2605 // This catches some cases where evaluation order is used as control flow:
2607 if (NextToken
->Tok
.isAnyIdentifier()) {
2608 const FormatToken
*NextNextToken
= NextToken
->getNextNonComment();
2609 if (NextNextToken
&& NextNextToken
->is(tok::arrow
))
2610 return TT_BinaryOperator
;
2613 // It is very unlikely that we are going to find a pointer or reference type
2614 // definition on the RHS of an assignment.
2615 if (IsExpression
&& !Contexts
.back().CaretFound
)
2616 return TT_BinaryOperator
;
2618 // Opeartors at class scope are likely pointer or reference members.
2619 if (!Scopes
.empty() && Scopes
.back() == ST_Class
)
2620 return TT_PointerOrReference
;
2622 // Tokens that indicate member access or chained operator& use.
2623 auto IsChainedOperatorAmpOrMember
= [](const FormatToken
*token
) {
2624 return !token
|| token
->isOneOf(tok::amp
, tok::period
, tok::arrow
,
2625 tok::arrowstar
, tok::periodstar
);
2628 // It's more likely that & represents operator& than an uninitialized
2630 if (Tok
.is(tok::amp
) && PrevToken
&& PrevToken
->Tok
.isAnyIdentifier() &&
2631 IsChainedOperatorAmpOrMember(PrevToken
->getPreviousNonComment()) &&
2632 NextToken
&& NextToken
->Tok
.isAnyIdentifier()) {
2633 if (auto NextNext
= NextToken
->getNextNonComment();
2635 (IsChainedOperatorAmpOrMember(NextNext
) || NextNext
->is(tok::semi
))) {
2636 return TT_BinaryOperator
;
2640 return TT_PointerOrReference
;
2643 TokenType
determinePlusMinusCaretUsage(const FormatToken
&Tok
) {
2644 if (determineUnaryOperatorByUsage(Tok
))
2645 return TT_UnaryOperator
;
2647 const FormatToken
*PrevToken
= Tok
.getPreviousNonComment();
2649 return TT_UnaryOperator
;
2651 if (PrevToken
->is(tok::at
))
2652 return TT_UnaryOperator
;
2654 // Fall back to marking the token as binary operator.
2655 return TT_BinaryOperator
;
2658 /// Determine whether ++/-- are pre- or post-increments/-decrements.
2659 TokenType
determineIncrementUsage(const FormatToken
&Tok
) {
2660 const FormatToken
*PrevToken
= Tok
.getPreviousNonComment();
2661 if (!PrevToken
|| PrevToken
->is(TT_CastRParen
))
2662 return TT_UnaryOperator
;
2663 if (PrevToken
->isOneOf(tok::r_paren
, tok::r_square
, tok::identifier
))
2664 return TT_TrailingUnaryOperator
;
2666 return TT_UnaryOperator
;
2669 SmallVector
<Context
, 8> Contexts
;
2671 const FormatStyle
&Style
;
2672 AnnotatedLine
&Line
;
2673 FormatToken
*CurrentToken
;
2675 const AdditionalKeywords
&Keywords
;
2677 SmallVector
<ScopeType
> &Scopes
;
2679 // Set of "<" tokens that do not open a template parameter list. If parseAngle
2680 // determines that a specific token can't be a template opener, it will make
2681 // same decision irrespective of the decisions for tokens leading up to it.
2682 // Store this information to prevent this from causing exponential runtime.
2683 llvm::SmallPtrSet
<FormatToken
*, 16> NonTemplateLess
;
2686 static const int PrecedenceUnaryOperator
= prec::PointerToMember
+ 1;
2687 static const int PrecedenceArrowAndPeriod
= prec::PointerToMember
+ 2;
2689 /// Parses binary expressions by inserting fake parenthesis based on
2690 /// operator precedence.
2691 class ExpressionParser
{
2693 ExpressionParser(const FormatStyle
&Style
, const AdditionalKeywords
&Keywords
,
2694 AnnotatedLine
&Line
)
2695 : Style(Style
), Keywords(Keywords
), Line(Line
), Current(Line
.First
) {}
2697 /// Parse expressions with the given operator precedence.
2698 void parse(int Precedence
= 0) {
2699 // Skip 'return' and ObjC selector colons as they are not part of a binary
2701 while (Current
&& (Current
->is(tok::kw_return
) ||
2702 (Current
->is(tok::colon
) &&
2703 Current
->isOneOf(TT_ObjCMethodExpr
, TT_DictLiteral
)))) {
2707 if (!Current
|| Precedence
> PrecedenceArrowAndPeriod
)
2710 // Conditional expressions need to be parsed separately for proper nesting.
2711 if (Precedence
== prec::Conditional
) {
2712 parseConditionalExpr();
2716 // Parse unary operators, which all have a higher precedence than binary
2718 if (Precedence
== PrecedenceUnaryOperator
) {
2719 parseUnaryOperator();
2723 FormatToken
*Start
= Current
;
2724 FormatToken
*LatestOperator
= nullptr;
2725 unsigned OperatorIndex
= 0;
2726 // The first name of the current type in a port list.
2727 FormatToken
*VerilogFirstOfType
= nullptr;
2730 // In Verilog ports in a module header that don't have a type take the
2731 // type of the previous one. For example,
2732 // module a(output b,
2735 // In this case there need to be fake parentheses around b and c.
2736 if (Style
.isVerilog() && Precedence
== prec::Comma
) {
2737 VerilogFirstOfType
=
2738 verilogGroupDecl(VerilogFirstOfType
, LatestOperator
);
2741 // Consume operators with higher precedence.
2742 parse(Precedence
+ 1);
2744 // Do not assign fake parenthesis to tokens that are part of an
2745 // unexpanded macro call. The line within the macro call contains
2746 // the parenthesis and commas, and we will not find operators within
2748 if (Current
&& Current
->MacroParent
)
2751 int CurrentPrecedence
= getCurrentPrecedence();
2753 if (Precedence
== CurrentPrecedence
&& Current
&&
2754 Current
->is(TT_SelectorName
)) {
2756 addFakeParenthesis(Start
, prec::Level(Precedence
));
2760 if ((Style
.isCSharp() || Style
.isJavaScript() ||
2761 Style
.Language
== FormatStyle::LK_Java
) &&
2762 Precedence
== prec::Additive
&& Current
) {
2763 // A string can be broken without parentheses around it when it is
2764 // already in a sequence of strings joined by `+` signs.
2765 FormatToken
*Prev
= Current
->getPreviousNonComment();
2766 if (Prev
&& Prev
->is(tok::string_literal
) &&
2767 (Prev
== Start
|| Prev
->endsSequence(tok::string_literal
, tok::plus
,
2768 TT_StringInConcatenation
))) {
2769 Prev
->setType(TT_StringInConcatenation
);
2773 // At the end of the line or when an operator with lower precedence is
2774 // found, insert fake parenthesis and return.
2776 (Current
->closesScope() &&
2777 (Current
->MatchingParen
|| Current
->is(TT_TemplateString
))) ||
2778 (CurrentPrecedence
!= -1 && CurrentPrecedence
< Precedence
) ||
2779 (CurrentPrecedence
== prec::Conditional
&&
2780 Precedence
== prec::Assignment
&& Current
->is(tok::colon
))) {
2784 // Consume scopes: (), [], <> and {}
2785 // In addition to that we handle require clauses as scope, so that the
2786 // constraints in that are correctly indented.
2787 if (Current
->opensScope() ||
2788 Current
->isOneOf(TT_RequiresClause
,
2789 TT_RequiresClauseInARequiresExpression
)) {
2790 // In fragment of a JavaScript template string can look like '}..${' and
2791 // thus close a scope and open a new one at the same time.
2792 while (Current
&& (!Current
->closesScope() || Current
->opensScope())) {
2799 if (CurrentPrecedence
== Precedence
) {
2801 LatestOperator
->NextOperator
= Current
;
2802 LatestOperator
= Current
;
2803 Current
->OperatorIndex
= OperatorIndex
;
2806 next(/*SkipPastLeadingComments=*/Precedence
> 0);
2810 // Group variables of the same type.
2811 if (Style
.isVerilog() && Precedence
== prec::Comma
&& VerilogFirstOfType
)
2812 addFakeParenthesis(VerilogFirstOfType
, prec::Comma
);
2814 if (LatestOperator
&& (Current
|| Precedence
> 0)) {
2815 // The requires clauses do not neccessarily end in a semicolon or a brace,
2816 // but just go over to struct/class or a function declaration, we need to
2817 // intervene so that the fake right paren is inserted correctly.
2820 Start
->Previous
->isOneOf(TT_RequiresClause
,
2821 TT_RequiresClauseInARequiresExpression
))
2823 auto Ret
= Current
? Current
: Line
.Last
;
2824 while (!Ret
->ClosesRequiresClause
&& Ret
->Previous
)
2825 Ret
= Ret
->Previous
;
2830 if (Precedence
== PrecedenceArrowAndPeriod
) {
2831 // Call expressions don't have a binary operator precedence.
2832 addFakeParenthesis(Start
, prec::Unknown
, End
);
2834 addFakeParenthesis(Start
, prec::Level(Precedence
), End
);
2840 /// Gets the precedence (+1) of the given token for binary operators
2841 /// and other tokens that we treat like binary operators.
2842 int getCurrentPrecedence() {
2844 const FormatToken
*NextNonComment
= Current
->getNextNonComment();
2845 if (Current
->is(TT_ConditionalExpr
))
2846 return prec::Conditional
;
2847 if (NextNonComment
&& Current
->is(TT_SelectorName
) &&
2848 (NextNonComment
->isOneOf(TT_DictLiteral
, TT_JsTypeColon
) ||
2849 ((Style
.Language
== FormatStyle::LK_Proto
||
2850 Style
.Language
== FormatStyle::LK_TextProto
) &&
2851 NextNonComment
->is(tok::less
)))) {
2852 return prec::Assignment
;
2854 if (Current
->is(TT_JsComputedPropertyName
))
2855 return prec::Assignment
;
2856 if (Current
->is(TT_LambdaArrow
))
2858 if (Current
->is(TT_FatArrow
))
2859 return prec::Assignment
;
2860 if (Current
->isOneOf(tok::semi
, TT_InlineASMColon
, TT_SelectorName
) ||
2861 (Current
->is(tok::comment
) && NextNonComment
&&
2862 NextNonComment
->is(TT_SelectorName
))) {
2865 if (Current
->is(TT_RangeBasedForLoopColon
))
2867 if ((Style
.Language
== FormatStyle::LK_Java
|| Style
.isJavaScript()) &&
2868 Current
->is(Keywords
.kw_instanceof
)) {
2869 return prec::Relational
;
2871 if (Style
.isJavaScript() &&
2872 Current
->isOneOf(Keywords
.kw_in
, Keywords
.kw_as
)) {
2873 return prec::Relational
;
2875 if (Current
->is(TT_BinaryOperator
) || Current
->is(tok::comma
))
2876 return Current
->getPrecedence();
2877 if (Current
->isOneOf(tok::period
, tok::arrow
) &&
2878 Current
->isNot(TT_TrailingReturnArrow
)) {
2879 return PrecedenceArrowAndPeriod
;
2881 if ((Style
.Language
== FormatStyle::LK_Java
|| Style
.isJavaScript()) &&
2882 Current
->isOneOf(Keywords
.kw_extends
, Keywords
.kw_implements
,
2883 Keywords
.kw_throws
)) {
2886 // In Verilog case labels are not on separate lines straight out of
2887 // UnwrappedLineParser. The colon is not part of an expression.
2888 if (Style
.isVerilog() && Current
->is(tok::colon
))
2894 void addFakeParenthesis(FormatToken
*Start
, prec::Level Precedence
,
2895 FormatToken
*End
= nullptr) {
2896 Start
->FakeLParens
.push_back(Precedence
);
2897 if (Precedence
> prec::Unknown
)
2898 Start
->StartsBinaryExpression
= true;
2899 if (!End
&& Current
)
2900 End
= Current
->getPreviousNonComment();
2903 if (Precedence
> prec::Unknown
)
2904 End
->EndsBinaryExpression
= true;
2908 /// Parse unary operator expressions and surround them with fake
2909 /// parentheses if appropriate.
2910 void parseUnaryOperator() {
2911 llvm::SmallVector
<FormatToken
*, 2> Tokens
;
2912 while (Current
&& Current
->is(TT_UnaryOperator
)) {
2913 Tokens
.push_back(Current
);
2916 parse(PrecedenceArrowAndPeriod
);
2917 for (FormatToken
*Token
: llvm::reverse(Tokens
)) {
2918 // The actual precedence doesn't matter.
2919 addFakeParenthesis(Token
, prec::Unknown
);
2923 void parseConditionalExpr() {
2924 while (Current
&& Current
->isTrailingComment())
2926 FormatToken
*Start
= Current
;
2927 parse(prec::LogicalOr
);
2928 if (!Current
|| Current
->isNot(tok::question
))
2931 parse(prec::Assignment
);
2932 if (!Current
|| Current
->isNot(TT_ConditionalExpr
))
2935 parse(prec::Assignment
);
2936 addFakeParenthesis(Start
, prec::Conditional
);
2939 void next(bool SkipPastLeadingComments
= true) {
2941 Current
= Current
->Next
;
2943 (Current
->NewlinesBefore
== 0 || SkipPastLeadingComments
) &&
2944 Current
->isTrailingComment()) {
2945 Current
= Current
->Next
;
2949 // Add fake parenthesis around declarations of the same type for example in a
2950 // module prototype. Return the first port / variable of the current type.
2951 FormatToken
*verilogGroupDecl(FormatToken
*FirstOfType
,
2952 FormatToken
*PreviousComma
) {
2956 FormatToken
*Start
= Current
;
2959 while (Start
->startsSequence(tok::l_paren
, tok::star
)) {
2960 if (!(Start
= Start
->MatchingParen
) ||
2961 !(Start
= Start
->getNextNonComment())) {
2966 FormatToken
*Tok
= Start
;
2968 if (Tok
->is(Keywords
.kw_assign
))
2969 Tok
= Tok
->getNextNonComment();
2971 // Skip any type qualifiers to find the first identifier. It may be either a
2972 // new type name or a variable name. There can be several type qualifiers
2973 // preceding a variable name, and we can not tell them apart by looking at
2974 // the word alone since a macro can be defined as either a type qualifier or
2975 // a variable name. Thus we use the last word before the dimensions instead
2976 // of the first word as the candidate for the variable or type name.
2977 FormatToken
*First
= nullptr;
2979 FormatToken
*Next
= Tok
->getNextNonComment();
2981 if (Tok
->is(tok::hash
)) {
2982 // Start of a macro expansion.
2986 Tok
= Tok
->getNextNonComment();
2987 } else if (Tok
->is(tok::hashhash
)) {
2988 // Concatenation. Skip.
2991 Tok
= Tok
->getNextNonComment();
2992 } else if (Keywords
.isVerilogQualifier(*Tok
) ||
2993 Keywords
.isVerilogIdentifier(*Tok
)) {
2996 // The name may have dots like `interface_foo.modport_foo`.
2997 while (Tok
&& Tok
->isOneOf(tok::period
, tok::coloncolon
) &&
2998 (Tok
= Tok
->getNextNonComment())) {
2999 if (Keywords
.isVerilogIdentifier(*Tok
))
3000 Tok
= Tok
->getNextNonComment();
3004 } else if (Tok
->is(tok::l_paren
)) {
3005 // Make sure the parenthesized list is a drive strength. Otherwise the
3006 // statement may be a module instantiation in which case we have already
3007 // found the instance name.
3009 Keywords
.kw_highz0
, Keywords
.kw_highz1
, Keywords
.kw_large
,
3010 Keywords
.kw_medium
, Keywords
.kw_pull0
, Keywords
.kw_pull1
,
3011 Keywords
.kw_small
, Keywords
.kw_strong0
, Keywords
.kw_strong1
,
3012 Keywords
.kw_supply0
, Keywords
.kw_supply1
, Keywords
.kw_weak0
,
3013 Keywords
.kw_weak1
)) {
3014 Tok
->setType(TT_VerilogStrength
);
3015 Tok
= Tok
->MatchingParen
;
3017 Tok
->setType(TT_VerilogStrength
);
3018 Tok
= Tok
->getNextNonComment();
3023 } else if (Tok
->is(tok::hash
)) {
3024 if (Next
->is(tok::l_paren
))
3025 Next
= Next
->MatchingParen
;
3027 Tok
= Next
->getNextNonComment();
3033 // Find the second identifier. If it exists it will be the name.
3034 FormatToken
*Second
= nullptr;
3036 while (Tok
&& Tok
->is(tok::l_square
) && (Tok
= Tok
->MatchingParen
))
3037 Tok
= Tok
->getNextNonComment();
3038 if (Tok
&& (Tok
->is(tok::hash
) || Keywords
.isVerilogIdentifier(*Tok
)))
3041 // If the second identifier doesn't exist and there are qualifiers, the type
3043 FormatToken
*TypedName
= nullptr;
3046 if (First
&& First
->is(TT_Unknown
))
3047 First
->setType(TT_VerilogDimensionedTypeName
);
3048 } else if (First
!= Start
) {
3049 // If 'First' is null, then this isn't a declaration, 'TypedName' gets set
3050 // to null as intended.
3055 // This is a declaration with a new type.
3056 if (TypedName
->is(TT_Unknown
))
3057 TypedName
->setType(TT_StartOfName
);
3058 // Group variables of the previous type.
3059 if (FirstOfType
&& PreviousComma
) {
3060 PreviousComma
->setType(TT_VerilogTypeComma
);
3061 addFakeParenthesis(FirstOfType
, prec::Comma
, PreviousComma
->Previous
);
3064 FirstOfType
= TypedName
;
3066 // Don't let higher precedence handle the qualifiers. For example if we
3069 // We skip `parameter` here. This way the fake parentheses for the
3070 // assignment will be around `x = 0`.
3071 while (Current
&& Current
!= FirstOfType
) {
3072 if (Current
->opensScope()) {
3083 const FormatStyle
&Style
;
3084 const AdditionalKeywords
&Keywords
;
3085 const AnnotatedLine
&Line
;
3086 FormatToken
*Current
;
3089 } // end anonymous namespace
3091 void TokenAnnotator::setCommentLineLevels(
3092 SmallVectorImpl
<AnnotatedLine
*> &Lines
) const {
3093 const AnnotatedLine
*NextNonCommentLine
= nullptr;
3094 for (AnnotatedLine
*Line
: llvm::reverse(Lines
)) {
3095 assert(Line
->First
);
3097 // If the comment is currently aligned with the line immediately following
3098 // it, that's probably intentional and we should keep it.
3099 if (NextNonCommentLine
&& !NextNonCommentLine
->First
->Finalized
&&
3100 Line
->isComment() && NextNonCommentLine
->First
->NewlinesBefore
<= 1 &&
3101 NextNonCommentLine
->First
->OriginalColumn
==
3102 Line
->First
->OriginalColumn
) {
3103 const bool PPDirectiveOrImportStmt
=
3104 NextNonCommentLine
->Type
== LT_PreprocessorDirective
||
3105 NextNonCommentLine
->Type
== LT_ImportStatement
;
3106 if (PPDirectiveOrImportStmt
)
3107 Line
->Type
= LT_CommentAbovePPDirective
;
3108 // Align comments for preprocessor lines with the # in column 0 if
3109 // preprocessor lines are not indented. Otherwise, align with the next
3111 Line
->Level
= Style
.IndentPPDirectives
!= FormatStyle::PPDIS_BeforeHash
&&
3112 PPDirectiveOrImportStmt
3114 : NextNonCommentLine
->Level
;
3116 NextNonCommentLine
= Line
->First
->isNot(tok::r_brace
) ? Line
: nullptr;
3119 setCommentLineLevels(Line
->Children
);
3123 static unsigned maxNestingDepth(const AnnotatedLine
&Line
) {
3124 unsigned Result
= 0;
3125 for (const auto *Tok
= Line
.First
; Tok
; Tok
= Tok
->Next
)
3126 Result
= std::max(Result
, Tok
->NestingLevel
);
3130 // Returns the name of a function with no return type, e.g. a constructor or
3132 static FormatToken
*getFunctionName(const AnnotatedLine
&Line
) {
3133 for (FormatToken
*Tok
= Line
.getFirstNonComment(), *Name
= nullptr; Tok
;
3134 Tok
= Tok
->getNextNonComment()) {
3135 // Skip C++11 attributes both before and after the function name.
3136 if (Tok
->is(tok::l_square
) && Tok
->is(TT_AttributeSquare
)) {
3137 Tok
= Tok
->MatchingParen
;
3143 // Make sure the name is followed by a pair of parentheses.
3145 return Tok
->is(tok::l_paren
) && Tok
->MatchingParen
? Name
: nullptr;
3147 // Skip keywords that may precede the constructor/destructor name.
3148 if (Tok
->isOneOf(tok::kw_friend
, tok::kw_inline
, tok::kw_virtual
,
3149 tok::kw_constexpr
, tok::kw_consteval
, tok::kw_explicit
)) {
3153 // A qualified name may start from the global namespace.
3154 if (Tok
->is(tok::coloncolon
)) {
3160 // Skip to the unqualified part of the name.
3161 while (Tok
->startsSequence(tok::identifier
, tok::coloncolon
)) {
3163 Tok
= Tok
->Next
->Next
;
3168 // Skip the `~` if a destructor name.
3169 if (Tok
->is(tok::tilde
)) {
3175 // Make sure the name is not already annotated, e.g. as NamespaceMacro.
3176 if (Tok
->isNot(tok::identifier
) || Tok
->isNot(TT_Unknown
))
3185 // Checks if Tok is a constructor/destructor name qualified by its class name.
3186 static bool isCtorOrDtorName(const FormatToken
*Tok
) {
3187 assert(Tok
&& Tok
->is(tok::identifier
));
3188 const auto *Prev
= Tok
->Previous
;
3190 if (Prev
&& Prev
->is(tok::tilde
))
3191 Prev
= Prev
->Previous
;
3193 if (!Prev
|| !Prev
->endsSequence(tok::coloncolon
, tok::identifier
))
3196 assert(Prev
->Previous
);
3197 return Prev
->Previous
->TokenText
== Tok
->TokenText
;
3200 void TokenAnnotator::annotate(AnnotatedLine
&Line
) {
3201 for (auto &Child
: Line
.Children
)
3204 AnnotatingParser
Parser(Style
, Line
, Keywords
, Scopes
);
3205 Line
.Type
= Parser
.parseLine();
3207 // With very deep nesting, ExpressionParser uses lots of stack and the
3208 // formatting algorithm is very slow. We're not going to do a good job here
3209 // anyway - it's probably generated code being formatted by mistake.
3210 // Just skip the whole line.
3211 if (maxNestingDepth(Line
) > 50)
3212 Line
.Type
= LT_Invalid
;
3214 if (Line
.Type
== LT_Invalid
)
3217 ExpressionParser
ExprParser(Style
, Keywords
, Line
);
3220 if (Style
.isCpp()) {
3221 auto *Tok
= getFunctionName(Line
);
3222 if (Tok
&& ((!Scopes
.empty() && Scopes
.back() == ST_Class
) ||
3223 Line
.endsWith(TT_FunctionLBrace
) || isCtorOrDtorName(Tok
))) {
3224 Tok
->setFinalizedType(TT_FunctionDeclarationName
);
3228 if (Line
.startsWith(TT_ObjCMethodSpecifier
))
3229 Line
.Type
= LT_ObjCMethodDecl
;
3230 else if (Line
.startsWith(TT_ObjCDecl
))
3231 Line
.Type
= LT_ObjCDecl
;
3232 else if (Line
.startsWith(TT_ObjCProperty
))
3233 Line
.Type
= LT_ObjCProperty
;
3235 Line
.First
->SpacesRequiredBefore
= 1;
3236 Line
.First
->CanBreakBefore
= Line
.First
->MustBreakBefore
;
3239 // This function heuristically determines whether 'Current' starts the name of a
3240 // function declaration.
3241 static bool isFunctionDeclarationName(bool IsCpp
, const FormatToken
&Current
,
3242 const AnnotatedLine
&Line
) {
3243 assert(Current
.Previous
);
3245 if (Current
.is(TT_FunctionDeclarationName
))
3248 if (!Current
.Tok
.getIdentifierInfo())
3251 auto skipOperatorName
= [](const FormatToken
*Next
) -> const FormatToken
* {
3252 for (; Next
; Next
= Next
->Next
) {
3253 if (Next
->is(TT_OverloadedOperatorLParen
))
3255 if (Next
->is(TT_OverloadedOperator
))
3257 if (Next
->isOneOf(tok::kw_new
, tok::kw_delete
)) {
3258 // For 'new[]' and 'delete[]'.
3260 Next
->Next
->startsSequence(tok::l_square
, tok::r_square
)) {
3261 Next
= Next
->Next
->Next
;
3265 if (Next
->startsSequence(tok::l_square
, tok::r_square
)) {
3266 // For operator[]().
3270 if ((Next
->isSimpleTypeSpecifier() || Next
->is(tok::identifier
)) &&
3271 Next
->Next
&& Next
->Next
->isOneOf(tok::star
, tok::amp
, tok::ampamp
)) {
3272 // For operator void*(), operator char*(), operator Foo*().
3276 if (Next
->is(TT_TemplateOpener
) && Next
->MatchingParen
) {
3277 Next
= Next
->MatchingParen
;
3286 // Find parentheses of parameter list.
3287 const FormatToken
*Next
= Current
.Next
;
3288 if (Current
.is(tok::kw_operator
)) {
3289 const auto *Previous
= Current
.Previous
;
3290 if (Previous
->Tok
.getIdentifierInfo() &&
3291 !Previous
->isOneOf(tok::kw_return
, tok::kw_co_return
)) {
3294 if (!Previous
->isOneOf(tok::star
, tok::amp
, tok::ampamp
, TT_TemplateCloser
))
3296 Next
= skipOperatorName(Next
);
3298 if (Current
.isNot(TT_StartOfName
) || Current
.NestingLevel
!= 0)
3300 for (; Next
; Next
= Next
->Next
) {
3301 if (Next
->is(TT_TemplateOpener
) && Next
->MatchingParen
) {
3302 Next
= Next
->MatchingParen
;
3303 } else if (Next
->is(tok::coloncolon
)) {
3307 if (Next
->is(tok::kw_operator
)) {
3308 Next
= skipOperatorName(Next
->Next
);
3311 if (Next
->isNot(tok::identifier
))
3313 } else if (isCppAttribute(IsCpp
, *Next
)) {
3314 Next
= Next
->MatchingParen
;
3317 } else if (Next
->is(tok::l_paren
)) {
3325 // Check whether parameter list can belong to a function declaration.
3326 if (!Next
|| Next
->isNot(tok::l_paren
) || !Next
->MatchingParen
)
3328 // If the lines ends with "{", this is likely a function definition.
3329 if (Line
.Last
->is(tok::l_brace
))
3331 if (Next
->Next
== Next
->MatchingParen
)
3332 return true; // Empty parentheses.
3333 // If there is an &/&& after the r_paren, this is likely a function.
3334 if (Next
->MatchingParen
->Next
&&
3335 Next
->MatchingParen
->Next
->is(TT_PointerOrReference
)) {
3339 // Check for K&R C function definitions (and C++ function definitions with
3340 // unnamed parameters), e.g.:
3345 // bool g(size_t = 0, bool b = false)
3349 if (IsCpp
&& Next
->Next
&& Next
->Next
->is(tok::identifier
) &&
3350 !Line
.endsWith(tok::semi
)) {
3354 for (const FormatToken
*Tok
= Next
->Next
; Tok
&& Tok
!= Next
->MatchingParen
;
3356 if (Tok
->is(TT_TypeDeclarationParen
))
3358 if (Tok
->isOneOf(tok::l_paren
, TT_TemplateOpener
) && Tok
->MatchingParen
) {
3359 Tok
= Tok
->MatchingParen
;
3362 if (Tok
->is(tok::kw_const
) || Tok
->isSimpleTypeSpecifier() ||
3363 Tok
->isOneOf(TT_PointerOrReference
, TT_StartOfName
, tok::ellipsis
)) {
3366 if (Tok
->isOneOf(tok::l_brace
, tok::string_literal
, TT_ObjCMethodExpr
) ||
3367 Tok
->Tok
.isLiteral()) {
3374 bool TokenAnnotator::mustBreakForReturnType(const AnnotatedLine
&Line
) const {
3375 assert(Line
.MightBeFunctionDecl
);
3377 if ((Style
.AlwaysBreakAfterReturnType
== FormatStyle::RTBS_TopLevel
||
3378 Style
.AlwaysBreakAfterReturnType
==
3379 FormatStyle::RTBS_TopLevelDefinitions
) &&
3384 switch (Style
.AlwaysBreakAfterReturnType
) {
3385 case FormatStyle::RTBS_None
:
3387 case FormatStyle::RTBS_All
:
3388 case FormatStyle::RTBS_TopLevel
:
3390 case FormatStyle::RTBS_AllDefinitions
:
3391 case FormatStyle::RTBS_TopLevelDefinitions
:
3392 return Line
.mightBeFunctionDefinition();
3398 static bool mustBreakAfterAttributes(const FormatToken
&Tok
,
3399 const FormatStyle
&Style
) {
3400 switch (Style
.BreakAfterAttributes
) {
3401 case FormatStyle::ABS_Always
:
3403 case FormatStyle::ABS_Leave
:
3404 return Tok
.NewlinesBefore
> 0;
3410 void TokenAnnotator::calculateFormattingInformation(AnnotatedLine
&Line
) const {
3411 for (AnnotatedLine
*ChildLine
: Line
.Children
)
3412 calculateFormattingInformation(*ChildLine
);
3414 Line
.First
->TotalLength
=
3415 Line
.First
->IsMultiline
? Style
.ColumnLimit
3416 : Line
.FirstStartColumn
+ Line
.First
->ColumnWidth
;
3417 FormatToken
*Current
= Line
.First
->Next
;
3418 bool InFunctionDecl
= Line
.MightBeFunctionDecl
;
3419 bool AlignArrayOfStructures
=
3420 (Style
.AlignArrayOfStructures
!= FormatStyle::AIAS_None
&&
3421 Line
.Type
== LT_ArrayOfStructInitializer
);
3422 if (AlignArrayOfStructures
)
3423 calculateArrayInitializerColumnList(Line
);
3425 bool LineIsFunctionDeclaration
= false;
3426 for (FormatToken
*Tok
= Current
, *AfterLastAttribute
= nullptr; Tok
;
3428 if (Tok
->Previous
->EndsCppAttributeGroup
)
3429 AfterLastAttribute
= Tok
;
3430 if (isFunctionDeclarationName(Style
.isCpp(), *Tok
, Line
)) {
3431 LineIsFunctionDeclaration
= true;
3432 Tok
->setFinalizedType(TT_FunctionDeclarationName
);
3433 if (AfterLastAttribute
&&
3434 mustBreakAfterAttributes(*AfterLastAttribute
, Style
)) {
3435 AfterLastAttribute
->MustBreakBefore
= true;
3436 Line
.ReturnTypeWrapped
= true;
3442 if (Style
.isCpp() && !LineIsFunctionDeclaration
) {
3443 // Annotate */&/&& in `operator` function calls as binary operators.
3444 for (const auto *Tok
= Line
.First
; Tok
; Tok
= Tok
->Next
) {
3445 if (Tok
->isNot(tok::kw_operator
))
3449 } while (Tok
&& Tok
->isNot(TT_OverloadedOperatorLParen
));
3452 const auto *LeftParen
= Tok
;
3453 for (Tok
= Tok
->Next
; Tok
&& Tok
!= LeftParen
->MatchingParen
;
3455 if (Tok
->isNot(tok::identifier
))
3457 auto *Next
= Tok
->Next
;
3458 const bool NextIsBinaryOperator
=
3459 Next
&& Next
->isOneOf(tok::star
, tok::amp
, tok::ampamp
) &&
3460 Next
->Next
&& Next
->Next
->is(tok::identifier
);
3461 if (!NextIsBinaryOperator
)
3463 Next
->setType(TT_BinaryOperator
);
3470 const FormatToken
*Prev
= Current
->Previous
;
3471 if (Current
->is(TT_LineComment
)) {
3472 if (Prev
->is(BK_BracedInit
) && Prev
->opensScope()) {
3473 Current
->SpacesRequiredBefore
=
3474 (Style
.Cpp11BracedListStyle
&& !Style
.SpacesInParensOptions
.Other
)
3477 } else if (Prev
->is(TT_VerilogMultiLineListLParen
)) {
3478 Current
->SpacesRequiredBefore
= 0;
3480 Current
->SpacesRequiredBefore
= Style
.SpacesBeforeTrailingComments
;
3483 // If we find a trailing comment, iterate backwards to determine whether
3484 // it seems to relate to a specific parameter. If so, break before that
3485 // parameter to avoid changing the comment's meaning. E.g. don't move 'b'
3486 // to the previous line in:
3490 if (!Current
->HasUnescapedNewline
) {
3491 for (FormatToken
*Parameter
= Current
->Previous
; Parameter
;
3492 Parameter
= Parameter
->Previous
) {
3493 if (Parameter
->isOneOf(tok::comment
, tok::r_brace
))
3495 if (Parameter
->Previous
&& Parameter
->Previous
->is(tok::comma
)) {
3496 if (Parameter
->Previous
->isNot(TT_CtorInitializerComma
) &&
3497 Parameter
->HasUnescapedNewline
) {
3498 Parameter
->MustBreakBefore
= true;
3504 } else if (Current
->SpacesRequiredBefore
== 0 &&
3505 spaceRequiredBefore(Line
, *Current
)) {
3506 Current
->SpacesRequiredBefore
= 1;
3509 const auto &Children
= Prev
->Children
;
3510 if (!Children
.empty() && Children
.back()->Last
->is(TT_LineComment
)) {
3511 Current
->MustBreakBefore
= true;
3513 Current
->MustBreakBefore
=
3514 Current
->MustBreakBefore
|| mustBreakBefore(Line
, *Current
);
3515 if (!Current
->MustBreakBefore
&& InFunctionDecl
&&
3516 Current
->is(TT_FunctionDeclarationName
)) {
3517 Current
->MustBreakBefore
= mustBreakForReturnType(Line
);
3521 Current
->CanBreakBefore
=
3522 Current
->MustBreakBefore
|| canBreakBefore(Line
, *Current
);
3523 unsigned ChildSize
= 0;
3524 if (Prev
->Children
.size() == 1) {
3525 FormatToken
&LastOfChild
= *Prev
->Children
[0]->Last
;
3526 ChildSize
= LastOfChild
.isTrailingComment() ? Style
.ColumnLimit
3527 : LastOfChild
.TotalLength
+ 1;
3529 if (Current
->MustBreakBefore
|| Prev
->Children
.size() > 1 ||
3530 (Prev
->Children
.size() == 1 &&
3531 Prev
->Children
[0]->First
->MustBreakBefore
) ||
3532 Current
->IsMultiline
) {
3533 Current
->TotalLength
= Prev
->TotalLength
+ Style
.ColumnLimit
;
3535 Current
->TotalLength
= Prev
->TotalLength
+ Current
->ColumnWidth
+
3536 ChildSize
+ Current
->SpacesRequiredBefore
;
3539 if (Current
->is(TT_CtorInitializerColon
))
3540 InFunctionDecl
= false;
3542 // FIXME: Only calculate this if CanBreakBefore is true once static
3543 // initializers etc. are sorted out.
3544 // FIXME: Move magic numbers to a better place.
3546 // Reduce penalty for aligning ObjC method arguments using the colon
3547 // alignment as this is the canonical way (still prefer fitting everything
3548 // into one line if possible). Trying to fit a whole expression into one
3549 // line should not force other line breaks (e.g. when ObjC method
3550 // expression is a part of other expression).
3551 Current
->SplitPenalty
= splitPenalty(Line
, *Current
, InFunctionDecl
);
3552 if (Style
.Language
== FormatStyle::LK_ObjC
&&
3553 Current
->is(TT_SelectorName
) && Current
->ParameterIndex
> 0) {
3554 if (Current
->ParameterIndex
== 1)
3555 Current
->SplitPenalty
+= 5 * Current
->BindingStrength
;
3557 Current
->SplitPenalty
+= 20 * Current
->BindingStrength
;
3560 Current
= Current
->Next
;
3563 calculateUnbreakableTailLengths(Line
);
3564 unsigned IndentLevel
= Line
.Level
;
3565 for (Current
= Line
.First
; Current
; Current
= Current
->Next
) {
3567 Current
->Role
->precomputeFormattingInfos(Current
);
3568 if (Current
->MatchingParen
&&
3569 Current
->MatchingParen
->opensBlockOrBlockTypeList(Style
) &&
3573 Current
->IndentLevel
= IndentLevel
;
3574 if (Current
->opensBlockOrBlockTypeList(Style
))
3578 LLVM_DEBUG({ printDebugInfo(Line
); });
3581 void TokenAnnotator::calculateUnbreakableTailLengths(
3582 AnnotatedLine
&Line
) const {
3583 unsigned UnbreakableTailLength
= 0;
3584 FormatToken
*Current
= Line
.Last
;
3586 Current
->UnbreakableTailLength
= UnbreakableTailLength
;
3587 if (Current
->CanBreakBefore
||
3588 Current
->isOneOf(tok::comment
, tok::string_literal
)) {
3589 UnbreakableTailLength
= 0;
3591 UnbreakableTailLength
+=
3592 Current
->ColumnWidth
+ Current
->SpacesRequiredBefore
;
3594 Current
= Current
->Previous
;
3598 void TokenAnnotator::calculateArrayInitializerColumnList(
3599 AnnotatedLine
&Line
) const {
3600 if (Line
.First
== Line
.Last
)
3602 auto *CurrentToken
= Line
.First
;
3603 CurrentToken
->ArrayInitializerLineStart
= true;
3605 while (CurrentToken
&& CurrentToken
!= Line
.Last
) {
3606 if (CurrentToken
->is(tok::l_brace
)) {
3607 CurrentToken
->IsArrayInitializer
= true;
3608 if (CurrentToken
->Next
)
3609 CurrentToken
->Next
->MustBreakBefore
= true;
3611 calculateInitializerColumnList(Line
, CurrentToken
->Next
, Depth
+ 1);
3613 CurrentToken
= CurrentToken
->Next
;
3618 FormatToken
*TokenAnnotator::calculateInitializerColumnList(
3619 AnnotatedLine
&Line
, FormatToken
*CurrentToken
, unsigned Depth
) const {
3620 while (CurrentToken
&& CurrentToken
!= Line
.Last
) {
3621 if (CurrentToken
->is(tok::l_brace
))
3623 else if (CurrentToken
->is(tok::r_brace
))
3625 if (Depth
== 2 && CurrentToken
->isOneOf(tok::l_brace
, tok::comma
)) {
3626 CurrentToken
= CurrentToken
->Next
;
3629 CurrentToken
->StartsColumn
= true;
3630 CurrentToken
= CurrentToken
->Previous
;
3632 CurrentToken
= CurrentToken
->Next
;
3634 return CurrentToken
;
3637 unsigned TokenAnnotator::splitPenalty(const AnnotatedLine
&Line
,
3638 const FormatToken
&Tok
,
3639 bool InFunctionDecl
) const {
3640 const FormatToken
&Left
= *Tok
.Previous
;
3641 const FormatToken
&Right
= Tok
;
3643 if (Left
.is(tok::semi
))
3646 // Language specific handling.
3647 if (Style
.Language
== FormatStyle::LK_Java
) {
3648 if (Right
.isOneOf(Keywords
.kw_extends
, Keywords
.kw_throws
))
3650 if (Right
.is(Keywords
.kw_implements
))
3652 if (Left
.is(tok::comma
) && Left
.NestingLevel
== 0)
3654 } else if (Style
.isJavaScript()) {
3655 if (Right
.is(Keywords
.kw_function
) && Left
.isNot(tok::comma
))
3657 if (Left
.is(TT_JsTypeColon
))
3659 if ((Left
.is(TT_TemplateString
) && Left
.TokenText
.endswith("${")) ||
3660 (Right
.is(TT_TemplateString
) && Right
.TokenText
.startswith("}"))) {
3663 // Prefer breaking call chains (".foo") over empty "{}", "[]" or "()".
3664 if (Left
.opensScope() && Right
.closesScope())
3666 } else if (Style
.isProto()) {
3667 if (Right
.is(tok::l_square
))
3669 if (Right
.is(tok::period
))
3673 if (Right
.is(tok::identifier
) && Right
.Next
&& Right
.Next
->is(TT_DictLiteral
))
3675 if (Right
.is(tok::l_square
)) {
3676 if (Left
.is(tok::r_square
))
3678 // Slightly prefer formatting local lambda definitions like functions.
3679 if (Right
.is(TT_LambdaLSquare
) && Left
.is(tok::equal
))
3681 if (!Right
.isOneOf(TT_ObjCMethodExpr
, TT_LambdaLSquare
,
3682 TT_ArrayInitializerLSquare
,
3683 TT_DesignatedInitializerLSquare
, TT_AttributeSquare
)) {
3688 if (Left
.is(tok::coloncolon
))
3690 if (Right
.isOneOf(TT_StartOfName
, TT_FunctionDeclarationName
) ||
3691 Right
.is(tok::kw_operator
)) {
3692 if (Line
.startsWith(tok::kw_for
) && Right
.PartOfMultiVariableDeclStmt
)
3694 if (Left
.is(TT_StartOfName
))
3696 if (InFunctionDecl
&& Right
.NestingLevel
== 0)
3697 return Style
.PenaltyReturnTypeOnItsOwnLine
;
3700 if (Right
.is(TT_PointerOrReference
))
3702 if (Right
.is(TT_LambdaArrow
))
3704 if (Left
.is(tok::equal
) && Right
.is(tok::l_brace
))
3706 if (Left
.is(TT_CastRParen
))
3708 if (Left
.isOneOf(tok::kw_class
, tok::kw_struct
, tok::kw_union
))
3710 if (Left
.is(tok::comment
))
3713 if (Left
.isOneOf(TT_RangeBasedForLoopColon
, TT_InheritanceColon
,
3714 TT_CtorInitializerColon
)) {
3718 if (Right
.isMemberAccess()) {
3719 // Breaking before the "./->" of a chained call/member access is reasonably
3720 // cheap, as formatting those with one call per line is generally
3721 // desirable. In particular, it should be cheaper to break before the call
3722 // than it is to break inside a call's parameters, which could lead to weird
3723 // "hanging" indents. The exception is the very last "./->" to support this
3724 // frequent pattern:
3726 // aaaaaaaa.aaaaaaaa.bbbbbbb().ccccccccccccccccccccc(
3729 // which might otherwise be blown up onto many lines. Here, clang-format
3730 // won't produce "hanging" indents anyway as there is no other trailing
3733 // Also apply higher penalty is not a call as that might lead to a wrapping
3737 // .aaaaaaaaa.bbbbbbbb(cccccccc);
3738 return !Right
.NextOperator
|| !Right
.NextOperator
->Previous
->closesScope()
3743 if (Right
.is(TT_TrailingAnnotation
) &&
3744 (!Right
.Next
|| Right
.Next
->isNot(tok::l_paren
))) {
3745 // Moving trailing annotations to the next line is fine for ObjC method
3747 if (Line
.startsWith(TT_ObjCMethodSpecifier
))
3749 // Generally, breaking before a trailing annotation is bad unless it is
3750 // function-like. It seems to be especially preferable to keep standard
3751 // annotations (i.e. "const", "final" and "override") on the same line.
3752 // Use a slightly higher penalty after ")" so that annotations like
3753 // "const override" are kept together.
3754 bool is_short_annotation
= Right
.TokenText
.size() < 10;
3755 return (Left
.is(tok::r_paren
) ? 100 : 120) + (is_short_annotation
? 50 : 0);
3758 // In for-loops, prefer breaking at ',' and ';'.
3759 if (Line
.startsWith(tok::kw_for
) && Left
.is(tok::equal
))
3762 // In Objective-C method expressions, prefer breaking before "param:" over
3763 // breaking after it.
3764 if (Right
.is(TT_SelectorName
))
3766 if (Left
.is(tok::colon
) && Left
.is(TT_ObjCMethodExpr
))
3767 return Line
.MightBeFunctionDecl
? 50 : 500;
3769 // In Objective-C type declarations, avoid breaking after the category's
3770 // open paren (we'll prefer breaking after the protocol list's opening
3771 // angle bracket, if present).
3772 if (Line
.Type
== LT_ObjCDecl
&& Left
.is(tok::l_paren
) && Left
.Previous
&&
3773 Left
.Previous
->isOneOf(tok::identifier
, tok::greater
)) {
3777 if (Left
.is(tok::l_paren
) && Style
.PenaltyBreakOpenParenthesis
!= 0)
3778 return Style
.PenaltyBreakOpenParenthesis
;
3779 if (Left
.is(tok::l_paren
) && InFunctionDecl
&&
3780 Style
.AlignAfterOpenBracket
!= FormatStyle::BAS_DontAlign
) {
3783 if (Left
.is(tok::l_paren
) && Left
.Previous
&&
3784 (Left
.Previous
->isOneOf(tok::kw_for
, tok::kw__Generic
) ||
3785 Left
.Previous
->isIf())) {
3788 if (Left
.is(tok::equal
) && InFunctionDecl
)
3790 if (Right
.is(tok::r_brace
))
3792 if (Left
.is(TT_TemplateOpener
))
3794 if (Left
.opensScope()) {
3795 // If we aren't aligning after opening parens/braces we can always break
3796 // here unless the style does not want us to place all arguments on the
3798 if (Style
.AlignAfterOpenBracket
== FormatStyle::BAS_DontAlign
&&
3799 (Left
.ParameterCount
<= 1 || Style
.AllowAllArgumentsOnNextLine
)) {
3802 if (Left
.is(tok::l_brace
) && !Style
.Cpp11BracedListStyle
)
3804 return Left
.ParameterCount
> 1 ? Style
.PenaltyBreakBeforeFirstCallParameter
3807 if (Left
.is(TT_JavaAnnotation
))
3810 if (Left
.is(TT_UnaryOperator
))
3812 if (Left
.isOneOf(tok::plus
, tok::comma
) && Left
.Previous
&&
3813 Left
.Previous
->isLabelString() &&
3814 (Left
.NextOperator
|| Left
.OperatorIndex
!= 0)) {
3817 if (Right
.is(tok::plus
) && Left
.isLabelString() &&
3818 (Right
.NextOperator
|| Right
.OperatorIndex
!= 0)) {
3821 if (Left
.is(tok::comma
))
3823 if (Right
.is(tok::lessless
) && Left
.isLabelString() &&
3824 (Right
.NextOperator
|| Right
.OperatorIndex
!= 1)) {
3827 if (Right
.is(tok::lessless
)) {
3828 // Breaking at a << is really cheap.
3829 if (Left
.isNot(tok::r_paren
) || Right
.OperatorIndex
> 0) {
3830 // Slightly prefer to break before the first one in log-like statements.
3835 if (Left
.ClosesTemplateDeclaration
)
3836 return Style
.PenaltyBreakTemplateDeclaration
;
3837 if (Left
.ClosesRequiresClause
)
3839 if (Left
.is(TT_ConditionalExpr
))
3840 return prec::Conditional
;
3841 prec::Level Level
= Left
.getPrecedence();
3842 if (Level
== prec::Unknown
)
3843 Level
= Right
.getPrecedence();
3844 if (Level
== prec::Assignment
)
3845 return Style
.PenaltyBreakAssignment
;
3846 if (Level
!= prec::Unknown
)
3852 bool TokenAnnotator::spaceRequiredBeforeParens(const FormatToken
&Right
) const {
3853 if (Style
.SpaceBeforeParens
== FormatStyle::SBPO_Always
)
3855 if (Right
.is(TT_OverloadedOperatorLParen
) &&
3856 Style
.SpaceBeforeParensOptions
.AfterOverloadedOperator
) {
3859 if (Style
.SpaceBeforeParensOptions
.BeforeNonEmptyParentheses
&&
3860 Right
.ParameterCount
> 0) {
3866 bool TokenAnnotator::spaceRequiredBetween(const AnnotatedLine
&Line
,
3867 const FormatToken
&Left
,
3868 const FormatToken
&Right
) const {
3869 if (Left
.is(tok::kw_return
) &&
3870 !Right
.isOneOf(tok::semi
, tok::r_paren
, tok::hashhash
)) {
3873 if (Left
.is(tok::kw_throw
) && Right
.is(tok::l_paren
) && Right
.MatchingParen
&&
3874 Right
.MatchingParen
->is(TT_CastRParen
)) {
3877 if (Left
.is(Keywords
.kw_assert
) && Style
.Language
== FormatStyle::LK_Java
)
3879 if (Style
.ObjCSpaceAfterProperty
&& Line
.Type
== LT_ObjCProperty
&&
3880 Left
.Tok
.getObjCKeywordID() == tok::objc_property
) {
3883 if (Right
.is(tok::hashhash
))
3884 return Left
.is(tok::hash
);
3885 if (Left
.isOneOf(tok::hashhash
, tok::hash
))
3886 return Right
.is(tok::hash
);
3887 if ((Left
.is(tok::l_paren
) && Right
.is(tok::r_paren
)) ||
3888 (Left
.is(tok::l_brace
) && Left
.isNot(BK_Block
) &&
3889 Right
.is(tok::r_brace
) && Right
.isNot(BK_Block
))) {
3890 return Style
.SpacesInParensOptions
.InEmptyParentheses
;
3892 if (Style
.SpacesInParensOptions
.InConditionalStatements
) {
3893 const FormatToken
*LeftParen
= nullptr;
3894 if (Left
.is(tok::l_paren
))
3896 else if (Right
.is(tok::r_paren
) && Right
.MatchingParen
)
3897 LeftParen
= Right
.MatchingParen
;
3899 if (LeftParen
->is(TT_ConditionLParen
))
3901 if (LeftParen
->Previous
&& isKeywordWithCondition(*LeftParen
->Previous
))
3906 // trailing return type 'auto': []() -> auto {}, auto foo() -> auto {}
3907 if (Left
.is(tok::kw_auto
) && Right
.isOneOf(TT_LambdaLBrace
, TT_FunctionLBrace
,
3908 // function return type 'auto'
3909 TT_FunctionTypeLParen
)) {
3914 if (Left
.is(tok::kw_auto
) && Right
.isOneOf(tok::l_paren
, tok::l_brace
))
3917 // operator co_await(x)
3918 if (Right
.is(tok::l_paren
) && Left
.is(tok::kw_co_await
) && Left
.Previous
&&
3919 Left
.Previous
->is(tok::kw_operator
)) {
3922 // co_await (x), co_yield (x), co_return (x)
3923 if (Left
.isOneOf(tok::kw_co_await
, tok::kw_co_yield
, tok::kw_co_return
) &&
3924 !Right
.isOneOf(tok::semi
, tok::r_paren
)) {
3928 if (Left
.is(tok::l_paren
) || Right
.is(tok::r_paren
)) {
3929 return (Right
.is(TT_CastRParen
) ||
3930 (Left
.MatchingParen
&& Left
.MatchingParen
->is(TT_CastRParen
)))
3931 ? Style
.SpacesInParensOptions
.InCStyleCasts
3932 : Style
.SpacesInParensOptions
.Other
;
3934 if (Right
.isOneOf(tok::semi
, tok::comma
))
3936 if (Right
.is(tok::less
) && Line
.Type
== LT_ObjCDecl
) {
3937 bool IsLightweightGeneric
= Right
.MatchingParen
&&
3938 Right
.MatchingParen
->Next
&&
3939 Right
.MatchingParen
->Next
->is(tok::colon
);
3940 return !IsLightweightGeneric
&& Style
.ObjCSpaceBeforeProtocolList
;
3942 if (Right
.is(tok::less
) && Left
.is(tok::kw_template
))
3943 return Style
.SpaceAfterTemplateKeyword
;
3944 if (Left
.isOneOf(tok::exclaim
, tok::tilde
))
3946 if (Left
.is(tok::at
) &&
3947 Right
.isOneOf(tok::identifier
, tok::string_literal
, tok::char_constant
,
3948 tok::numeric_constant
, tok::l_paren
, tok::l_brace
,
3949 tok::kw_true
, tok::kw_false
)) {
3952 if (Left
.is(tok::colon
))
3953 return Left
.isNot(TT_ObjCMethodExpr
);
3954 if (Left
.is(tok::coloncolon
))
3956 if (Left
.is(tok::less
) || Right
.isOneOf(tok::greater
, tok::less
)) {
3957 if (Style
.Language
== FormatStyle::LK_TextProto
||
3958 (Style
.Language
== FormatStyle::LK_Proto
&&
3959 (Left
.is(TT_DictLiteral
) || Right
.is(TT_DictLiteral
)))) {
3960 // Format empty list as `<>`.
3961 if (Left
.is(tok::less
) && Right
.is(tok::greater
))
3963 return !Style
.Cpp11BracedListStyle
;
3965 // Don't attempt to format operator<(), as it is handled later.
3966 if (Right
.isNot(TT_OverloadedOperatorLParen
))
3969 if (Right
.is(tok::ellipsis
)) {
3970 return Left
.Tok
.isLiteral() || (Left
.is(tok::identifier
) && Left
.Previous
&&
3971 Left
.Previous
->is(tok::kw_case
));
3973 if (Left
.is(tok::l_square
) && Right
.is(tok::amp
))
3974 return Style
.SpacesInSquareBrackets
;
3975 if (Right
.is(TT_PointerOrReference
)) {
3976 if (Left
.is(tok::r_paren
) && Line
.MightBeFunctionDecl
) {
3977 if (!Left
.MatchingParen
)
3979 FormatToken
*TokenBeforeMatchingParen
=
3980 Left
.MatchingParen
->getPreviousNonComment();
3981 if (!TokenBeforeMatchingParen
|| Left
.isNot(TT_TypeDeclarationParen
))
3984 // Add a space if the previous token is a pointer qualifier or the closing
3985 // parenthesis of __attribute__(()) expression and the style requires spaces
3986 // after pointer qualifiers.
3987 if ((Style
.SpaceAroundPointerQualifiers
== FormatStyle::SAPQ_After
||
3988 Style
.SpaceAroundPointerQualifiers
== FormatStyle::SAPQ_Both
) &&
3989 (Left
.is(TT_AttributeParen
) ||
3990 Left
.canBePointerOrReferenceQualifier())) {
3993 if (Left
.Tok
.isLiteral())
3995 // for (auto a = 0, b = 0; const auto & c : {1, 2, 3})
3996 if (Left
.isTypeOrIdentifier() && Right
.Next
&& Right
.Next
->Next
&&
3997 Right
.Next
->Next
->is(TT_RangeBasedForLoopColon
)) {
3998 return getTokenPointerOrReferenceAlignment(Right
) !=
3999 FormatStyle::PAS_Left
;
4001 return !Left
.isOneOf(TT_PointerOrReference
, tok::l_paren
) &&
4002 (getTokenPointerOrReferenceAlignment(Right
) !=
4003 FormatStyle::PAS_Left
||
4004 (Line
.IsMultiVariableDeclStmt
&&
4005 (Left
.NestingLevel
== 0 ||
4006 (Left
.NestingLevel
== 1 && startsWithInitStatement(Line
)))));
4008 if (Right
.is(TT_FunctionTypeLParen
) && Left
.isNot(tok::l_paren
) &&
4009 (Left
.isNot(TT_PointerOrReference
) ||
4010 (getTokenPointerOrReferenceAlignment(Left
) != FormatStyle::PAS_Right
&&
4011 !Line
.IsMultiVariableDeclStmt
))) {
4014 if (Left
.is(TT_PointerOrReference
)) {
4015 // Add a space if the next token is a pointer qualifier and the style
4016 // requires spaces before pointer qualifiers.
4017 if ((Style
.SpaceAroundPointerQualifiers
== FormatStyle::SAPQ_Before
||
4018 Style
.SpaceAroundPointerQualifiers
== FormatStyle::SAPQ_Both
) &&
4019 Right
.canBePointerOrReferenceQualifier()) {
4023 if (Right
.Tok
.isLiteral())
4026 if (Right
.is(TT_BlockComment
))
4028 // foo() -> const Bar * override/final
4029 // S::foo() & noexcept/requires
4030 if (Right
.isOneOf(Keywords
.kw_override
, Keywords
.kw_final
, tok::kw_noexcept
,
4031 TT_RequiresClause
) &&
4032 Right
.isNot(TT_StartOfName
)) {
4036 if (Right
.is(tok::l_brace
) && Right
.is(BK_Block
))
4038 // for (auto a = 0, b = 0; const auto& c : {1, 2, 3})
4039 if (Left
.Previous
&& Left
.Previous
->isTypeOrIdentifier() && Right
.Next
&&
4040 Right
.Next
->is(TT_RangeBasedForLoopColon
)) {
4041 return getTokenPointerOrReferenceAlignment(Left
) !=
4042 FormatStyle::PAS_Right
;
4044 if (Right
.isOneOf(TT_PointerOrReference
, TT_ArraySubscriptLSquare
,
4048 if (getTokenPointerOrReferenceAlignment(Left
) == FormatStyle::PAS_Right
)
4050 // FIXME: Setting IsMultiVariableDeclStmt for the whole line is error-prone,
4051 // because it does not take into account nested scopes like lambdas.
4052 // In multi-variable declaration statements, attach */& to the variable
4053 // independently of the style. However, avoid doing it if we are in a nested
4054 // scope, e.g. lambda. We still need to special-case statements with
4056 if (Line
.IsMultiVariableDeclStmt
&&
4057 (Left
.NestingLevel
== Line
.First
->NestingLevel
||
4058 ((Left
.NestingLevel
== Line
.First
->NestingLevel
+ 1) &&
4059 startsWithInitStatement(Line
)))) {
4062 return Left
.Previous
&& !Left
.Previous
->isOneOf(
4063 tok::l_paren
, tok::coloncolon
, tok::l_square
);
4065 // Ensure right pointer alignment with ellipsis e.g. int *...P
4066 if (Left
.is(tok::ellipsis
) && Left
.Previous
&&
4067 Left
.Previous
->isOneOf(tok::star
, tok::amp
, tok::ampamp
)) {
4068 return Style
.PointerAlignment
!= FormatStyle::PAS_Right
;
4071 if (Right
.is(tok::star
) && Left
.is(tok::l_paren
))
4073 if (Left
.is(tok::star
) && Right
.isOneOf(tok::star
, tok::amp
, tok::ampamp
))
4075 if (Right
.isOneOf(tok::star
, tok::amp
, tok::ampamp
)) {
4076 const FormatToken
*Previous
= &Left
;
4077 while (Previous
&& Previous
->isNot(tok::kw_operator
)) {
4078 if (Previous
->is(tok::identifier
) || Previous
->isSimpleTypeSpecifier()) {
4079 Previous
= Previous
->getPreviousNonComment();
4082 if (Previous
->is(TT_TemplateCloser
) && Previous
->MatchingParen
) {
4083 Previous
= Previous
->MatchingParen
->getPreviousNonComment();
4086 if (Previous
->is(tok::coloncolon
)) {
4087 Previous
= Previous
->getPreviousNonComment();
4092 // Space between the type and the * in:
4095 // operator void const*()
4096 // operator void volatile*()
4097 // operator /*comment*/ const char*()
4098 // operator volatile /*comment*/ char*()
4101 // operator std::Foo*()
4102 // operator C<T>::D<U>*()
4103 // dependent on PointerAlignment style.
4105 if (Previous
->endsSequence(tok::kw_operator
))
4106 return Style
.PointerAlignment
!= FormatStyle::PAS_Left
;
4107 if (Previous
->is(tok::kw_const
) || Previous
->is(tok::kw_volatile
)) {
4108 return (Style
.PointerAlignment
!= FormatStyle::PAS_Left
) ||
4109 (Style
.SpaceAroundPointerQualifiers
==
4110 FormatStyle::SAPQ_After
) ||
4111 (Style
.SpaceAroundPointerQualifiers
== FormatStyle::SAPQ_Both
);
4115 if (Style
.isCSharp() && Left
.is(Keywords
.kw_is
) && Right
.is(tok::l_square
))
4117 const auto SpaceRequiredForArrayInitializerLSquare
=
4118 [](const FormatToken
&LSquareTok
, const FormatStyle
&Style
) {
4119 return Style
.SpacesInContainerLiterals
||
4120 ((Style
.Language
== FormatStyle::LK_Proto
||
4121 Style
.Language
== FormatStyle::LK_TextProto
) &&
4122 !Style
.Cpp11BracedListStyle
&&
4123 LSquareTok
.endsSequence(tok::l_square
, tok::colon
,
4126 if (Left
.is(tok::l_square
)) {
4127 return (Left
.is(TT_ArrayInitializerLSquare
) && Right
.isNot(tok::r_square
) &&
4128 SpaceRequiredForArrayInitializerLSquare(Left
, Style
)) ||
4129 (Left
.isOneOf(TT_ArraySubscriptLSquare
, TT_StructuredBindingLSquare
,
4130 TT_LambdaLSquare
) &&
4131 Style
.SpacesInSquareBrackets
&& Right
.isNot(tok::r_square
));
4133 if (Right
.is(tok::r_square
)) {
4134 return Right
.MatchingParen
&&
4135 ((Right
.MatchingParen
->is(TT_ArrayInitializerLSquare
) &&
4136 SpaceRequiredForArrayInitializerLSquare(*Right
.MatchingParen
,
4138 (Style
.SpacesInSquareBrackets
&&
4139 Right
.MatchingParen
->isOneOf(TT_ArraySubscriptLSquare
,
4140 TT_StructuredBindingLSquare
,
4141 TT_LambdaLSquare
)) ||
4142 Right
.MatchingParen
->is(TT_AttributeParen
));
4144 if (Right
.is(tok::l_square
) &&
4145 !Right
.isOneOf(TT_ObjCMethodExpr
, TT_LambdaLSquare
,
4146 TT_DesignatedInitializerLSquare
,
4147 TT_StructuredBindingLSquare
, TT_AttributeSquare
) &&
4148 !Left
.isOneOf(tok::numeric_constant
, TT_DictLiteral
) &&
4149 !(Left
.isNot(tok::r_square
) && Style
.SpaceBeforeSquareBrackets
&&
4150 Right
.is(TT_ArraySubscriptLSquare
))) {
4153 if (Left
.is(tok::l_brace
) && Right
.is(tok::r_brace
))
4154 return !Left
.Children
.empty(); // No spaces in "{}".
4155 if ((Left
.is(tok::l_brace
) && Left
.isNot(BK_Block
)) ||
4156 (Right
.is(tok::r_brace
) && Right
.MatchingParen
&&
4157 Right
.MatchingParen
->isNot(BK_Block
))) {
4158 return Style
.Cpp11BracedListStyle
? Style
.SpacesInParensOptions
.Other
4161 if (Left
.is(TT_BlockComment
)) {
4162 // No whitespace in x(/*foo=*/1), except for JavaScript.
4163 return Style
.isJavaScript() || !Left
.TokenText
.endswith("=*/");
4166 // Space between template and attribute.
4167 // e.g. template <typename T> [[nodiscard]] ...
4168 if (Left
.is(TT_TemplateCloser
) && Right
.is(TT_AttributeSquare
))
4170 // Space before parentheses common for all languages
4171 if (Right
.is(tok::l_paren
)) {
4172 if (Left
.is(TT_TemplateCloser
) && Right
.isNot(TT_FunctionTypeLParen
))
4173 return spaceRequiredBeforeParens(Right
);
4174 if (Left
.isOneOf(TT_RequiresClause
,
4175 TT_RequiresClauseInARequiresExpression
)) {
4176 return Style
.SpaceBeforeParensOptions
.AfterRequiresInClause
||
4177 spaceRequiredBeforeParens(Right
);
4179 if (Left
.is(TT_RequiresExpression
)) {
4180 return Style
.SpaceBeforeParensOptions
.AfterRequiresInExpression
||
4181 spaceRequiredBeforeParens(Right
);
4183 if ((Left
.is(tok::r_paren
) && Left
.is(TT_AttributeParen
)) ||
4184 (Left
.is(tok::r_square
) && Left
.is(TT_AttributeSquare
))) {
4187 if (Left
.is(TT_ForEachMacro
)) {
4188 return Style
.SpaceBeforeParensOptions
.AfterForeachMacros
||
4189 spaceRequiredBeforeParens(Right
);
4191 if (Left
.is(TT_IfMacro
)) {
4192 return Style
.SpaceBeforeParensOptions
.AfterIfMacros
||
4193 spaceRequiredBeforeParens(Right
);
4195 if (Line
.Type
== LT_ObjCDecl
)
4197 if (Left
.is(tok::semi
))
4199 if (Left
.isOneOf(tok::pp_elif
, tok::kw_for
, tok::kw_while
, tok::kw_switch
,
4200 tok::kw_case
, TT_ForEachMacro
, TT_ObjCForIn
) ||
4201 Left
.isIf(Line
.Type
!= LT_PreprocessorDirective
) ||
4202 Right
.is(TT_ConditionLParen
)) {
4203 return Style
.SpaceBeforeParensOptions
.AfterControlStatements
||
4204 spaceRequiredBeforeParens(Right
);
4207 // TODO add Operator overloading specific Options to
4208 // SpaceBeforeParensOptions
4209 if (Right
.is(TT_OverloadedOperatorLParen
))
4210 return spaceRequiredBeforeParens(Right
);
4211 // Function declaration or definition
4212 if (Line
.MightBeFunctionDecl
&& (Left
.is(TT_FunctionDeclarationName
))) {
4213 if (Line
.mightBeFunctionDefinition()) {
4214 return Style
.SpaceBeforeParensOptions
.AfterFunctionDefinitionName
||
4215 spaceRequiredBeforeParens(Right
);
4217 return Style
.SpaceBeforeParensOptions
.AfterFunctionDeclarationName
||
4218 spaceRequiredBeforeParens(Right
);
4222 if (Line
.Type
!= LT_PreprocessorDirective
&& Left
.is(tok::r_square
) &&
4223 Left
.MatchingParen
&& Left
.MatchingParen
->is(TT_LambdaLSquare
)) {
4224 return Style
.SpaceBeforeParensOptions
.AfterFunctionDefinitionName
||
4225 spaceRequiredBeforeParens(Right
);
4227 if (!Left
.Previous
|| Left
.Previous
->isNot(tok::period
)) {
4228 if (Left
.isOneOf(tok::kw_try
, Keywords
.kw___except
, tok::kw_catch
)) {
4229 return Style
.SpaceBeforeParensOptions
.AfterControlStatements
||
4230 spaceRequiredBeforeParens(Right
);
4232 if (Left
.isOneOf(tok::kw_new
, tok::kw_delete
)) {
4233 return ((!Line
.MightBeFunctionDecl
|| !Left
.Previous
) &&
4234 Style
.SpaceBeforeParens
!= FormatStyle::SBPO_Never
) ||
4235 spaceRequiredBeforeParens(Right
);
4238 if (Left
.is(tok::r_square
) && Left
.MatchingParen
&&
4239 Left
.MatchingParen
->Previous
&&
4240 Left
.MatchingParen
->Previous
->is(tok::kw_delete
)) {
4241 return (Style
.SpaceBeforeParens
!= FormatStyle::SBPO_Never
) ||
4242 spaceRequiredBeforeParens(Right
);
4245 // Handle builtins like identifiers.
4246 if (Line
.Type
!= LT_PreprocessorDirective
&&
4247 (Left
.Tok
.getIdentifierInfo() || Left
.is(tok::r_paren
))) {
4248 return spaceRequiredBeforeParens(Right
);
4252 if (Left
.is(tok::at
) && Right
.Tok
.getObjCKeywordID() != tok::objc_not_keyword
)
4254 if (Right
.is(TT_UnaryOperator
)) {
4255 return !Left
.isOneOf(tok::l_paren
, tok::l_square
, tok::at
) &&
4256 (Left
.isNot(tok::colon
) || Left
.isNot(TT_ObjCMethodExpr
));
4258 // No space between the variable name and the initializer list.
4260 // Verilog doesn't have such syntax, but it has word operators that are C++
4261 // identifiers like `a inside {b, c}`. So the rule is not applicable.
4262 if (!Style
.isVerilog() &&
4263 (Left
.isOneOf(tok::identifier
, tok::greater
, tok::r_square
,
4265 Left
.isSimpleTypeSpecifier()) &&
4266 Right
.is(tok::l_brace
) && Right
.getNextNonComment() &&
4267 Right
.isNot(BK_Block
)) {
4270 if (Left
.is(tok::period
) || Right
.is(tok::period
))
4272 // u#str, U#str, L#str, u8#str
4273 // uR#str, UR#str, LR#str, u8R#str
4274 if (Right
.is(tok::hash
) && Left
.is(tok::identifier
) &&
4275 (Left
.TokenText
== "L" || Left
.TokenText
== "u" ||
4276 Left
.TokenText
== "U" || Left
.TokenText
== "u8" ||
4277 Left
.TokenText
== "LR" || Left
.TokenText
== "uR" ||
4278 Left
.TokenText
== "UR" || Left
.TokenText
== "u8R")) {
4281 if (Left
.is(TT_TemplateCloser
) && Left
.MatchingParen
&&
4282 Left
.MatchingParen
->Previous
&&
4283 (Left
.MatchingParen
->Previous
->is(tok::period
) ||
4284 Left
.MatchingParen
->Previous
->is(tok::coloncolon
))) {
4285 // Java call to generic function with explicit type:
4286 // A.<B<C<...>>>DoSomething();
4287 // A::<B<C<...>>>DoSomething(); // With a Java 8 method reference.
4290 if (Left
.is(TT_TemplateCloser
) && Right
.is(tok::l_square
))
4292 if (Left
.is(tok::l_brace
) && Left
.endsSequence(TT_DictLiteral
, tok::at
)) {
4293 // Objective-C dictionary literal -> no space after opening brace.
4296 if (Right
.is(tok::r_brace
) && Right
.MatchingParen
&&
4297 Right
.MatchingParen
->endsSequence(TT_DictLiteral
, tok::at
)) {
4298 // Objective-C dictionary literal -> no space before closing brace.
4301 if (Right
.getType() == TT_TrailingAnnotation
&&
4302 Right
.isOneOf(tok::amp
, tok::ampamp
) &&
4303 Left
.isOneOf(tok::kw_const
, tok::kw_volatile
) &&
4304 (!Right
.Next
|| Right
.Next
->is(tok::semi
))) {
4305 // Match const and volatile ref-qualifiers without any additional
4306 // qualifiers such as
4307 // void Fn() const &;
4308 return getTokenReferenceAlignment(Right
) != FormatStyle::PAS_Left
;
4314 bool TokenAnnotator::spaceRequiredBefore(const AnnotatedLine
&Line
,
4315 const FormatToken
&Right
) const {
4316 const FormatToken
&Left
= *Right
.Previous
;
4318 // If the token is finalized don't touch it (as it could be in a
4319 // clang-format-off section).
4321 return Right
.hasWhitespaceBefore();
4323 // Never ever merge two words.
4324 if (Keywords
.isWordLike(Right
) && Keywords
.isWordLike(Left
))
4327 // Leave a space between * and /* to avoid C4138 `comment end` found outside
4329 if (Left
.is(tok::star
) && Right
.is(tok::comment
))
4332 if (Style
.isCpp()) {
4333 if (Left
.is(TT_OverloadedOperator
) &&
4334 Right
.isOneOf(TT_TemplateOpener
, TT_TemplateCloser
)) {
4337 // Space between UDL and dot: auto b = 4s .count();
4338 if (Right
.is(tok::period
) && Left
.is(tok::numeric_constant
))
4340 // Space between import <iostream>.
4342 if (Left
.is(Keywords
.kw_import
) && Right
.isOneOf(tok::less
, tok::ellipsis
))
4344 // Space between `module :` and `import :`.
4345 if (Left
.isOneOf(Keywords
.kw_module
, Keywords
.kw_import
) &&
4346 Right
.is(TT_ModulePartitionColon
)) {
4349 // No space between import foo:bar but keep a space between import :bar;
4350 if (Left
.is(tok::identifier
) && Right
.is(TT_ModulePartitionColon
))
4352 // No space between :bar;
4353 if (Left
.is(TT_ModulePartitionColon
) &&
4354 Right
.isOneOf(tok::identifier
, tok::kw_private
)) {
4357 if (Left
.is(tok::ellipsis
) && Right
.is(tok::identifier
) &&
4358 Line
.First
->is(Keywords
.kw_import
)) {
4361 // Space in __attribute__((attr)) ::type.
4362 if (Left
.is(TT_AttributeParen
) && Right
.is(tok::coloncolon
))
4365 if (Left
.is(tok::kw_operator
))
4366 return Right
.is(tok::coloncolon
);
4367 if (Right
.is(tok::l_brace
) && Right
.is(BK_BracedInit
) &&
4368 !Left
.opensScope() && Style
.SpaceBeforeCpp11BracedList
) {
4371 if (Left
.is(tok::less
) && Left
.is(TT_OverloadedOperator
) &&
4372 Right
.is(TT_TemplateOpener
)) {
4375 } else if (Style
.Language
== FormatStyle::LK_Proto
||
4376 Style
.Language
== FormatStyle::LK_TextProto
) {
4377 if (Right
.is(tok::period
) &&
4378 Left
.isOneOf(Keywords
.kw_optional
, Keywords
.kw_required
,
4379 Keywords
.kw_repeated
, Keywords
.kw_extend
)) {
4382 if (Right
.is(tok::l_paren
) &&
4383 Left
.isOneOf(Keywords
.kw_returns
, Keywords
.kw_option
)) {
4386 if (Right
.isOneOf(tok::l_brace
, tok::less
) && Left
.is(TT_SelectorName
))
4388 // Slashes occur in text protocol extension syntax: [type/type] { ... }.
4389 if (Left
.is(tok::slash
) || Right
.is(tok::slash
))
4391 if (Left
.MatchingParen
&&
4392 Left
.MatchingParen
->is(TT_ProtoExtensionLSquare
) &&
4393 Right
.isOneOf(tok::l_brace
, tok::less
)) {
4394 return !Style
.Cpp11BracedListStyle
;
4396 // A percent is probably part of a formatting specification, such as %lld.
4397 if (Left
.is(tok::percent
))
4399 // Preserve the existence of a space before a percent for cases like 0x%04x
4401 if (Left
.is(tok::numeric_constant
) && Right
.is(tok::percent
))
4402 return Right
.hasWhitespaceBefore();
4403 } else if (Style
.isJson()) {
4404 if (Right
.is(tok::colon
) && Left
.is(tok::string_literal
))
4405 return Style
.SpaceBeforeJsonColon
;
4406 } else if (Style
.isCSharp()) {
4407 // Require spaces around '{' and before '}' unless they appear in
4408 // interpolated strings. Interpolated strings are merged into a single token
4409 // so cannot have spaces inserted by this function.
4411 // No space between 'this' and '['
4412 if (Left
.is(tok::kw_this
) && Right
.is(tok::l_square
))
4415 // No space between 'new' and '('
4416 if (Left
.is(tok::kw_new
) && Right
.is(tok::l_paren
))
4419 // Space before { (including space within '{ {').
4420 if (Right
.is(tok::l_brace
))
4423 // Spaces inside braces.
4424 if (Left
.is(tok::l_brace
) && Right
.isNot(tok::r_brace
))
4427 if (Left
.isNot(tok::l_brace
) && Right
.is(tok::r_brace
))
4430 // Spaces around '=>'.
4431 if (Left
.is(TT_FatArrow
) || Right
.is(TT_FatArrow
))
4434 // No spaces around attribute target colons
4435 if (Left
.is(TT_AttributeColon
) || Right
.is(TT_AttributeColon
))
4438 // space between type and variable e.g. Dictionary<string,string> foo;
4439 if (Left
.is(TT_TemplateCloser
) && Right
.is(TT_StartOfName
))
4442 // spaces inside square brackets.
4443 if (Left
.is(tok::l_square
) || Right
.is(tok::r_square
))
4444 return Style
.SpacesInSquareBrackets
;
4446 // No space before ? in nullable types.
4447 if (Right
.is(TT_CSharpNullable
))
4450 // No space before null forgiving '!'.
4451 if (Right
.is(TT_NonNullAssertion
))
4454 // No space between consecutive commas '[,,]'.
4455 if (Left
.is(tok::comma
) && Right
.is(tok::comma
))
4458 // space after var in `var (key, value)`
4459 if (Left
.is(Keywords
.kw_var
) && Right
.is(tok::l_paren
))
4462 // space between keywords and paren e.g. "using ("
4463 if (Right
.is(tok::l_paren
)) {
4464 if (Left
.isOneOf(tok::kw_using
, Keywords
.kw_async
, Keywords
.kw_when
,
4465 Keywords
.kw_lock
)) {
4466 return Style
.SpaceBeforeParensOptions
.AfterControlStatements
||
4467 spaceRequiredBeforeParens(Right
);
4471 // space between method modifier and opening parenthesis of a tuple return
4473 if (Left
.isOneOf(tok::kw_public
, tok::kw_private
, tok::kw_protected
,
4474 tok::kw_virtual
, tok::kw_extern
, tok::kw_static
,
4475 Keywords
.kw_internal
, Keywords
.kw_abstract
,
4476 Keywords
.kw_sealed
, Keywords
.kw_override
,
4477 Keywords
.kw_async
, Keywords
.kw_unsafe
) &&
4478 Right
.is(tok::l_paren
)) {
4481 } else if (Style
.isJavaScript()) {
4482 if (Left
.is(TT_FatArrow
))
4485 if (Right
.is(tok::l_paren
) && Left
.is(Keywords
.kw_await
) && Left
.Previous
&&
4486 Left
.Previous
->is(tok::kw_for
)) {
4489 if (Left
.is(Keywords
.kw_async
) && Right
.is(tok::l_paren
) &&
4490 Right
.MatchingParen
) {
4491 const FormatToken
*Next
= Right
.MatchingParen
->getNextNonComment();
4492 // An async arrow function, for example: `x = async () => foo();`,
4493 // as opposed to calling a function called async: `x = async();`
4494 if (Next
&& Next
->is(TT_FatArrow
))
4497 if ((Left
.is(TT_TemplateString
) && Left
.TokenText
.endswith("${")) ||
4498 (Right
.is(TT_TemplateString
) && Right
.TokenText
.startswith("}"))) {
4501 // In tagged template literals ("html`bar baz`"), there is no space between
4502 // the tag identifier and the template string.
4503 if (Keywords
.IsJavaScriptIdentifier(Left
,
4504 /* AcceptIdentifierName= */ false) &&
4505 Right
.is(TT_TemplateString
)) {
4508 if (Right
.is(tok::star
) &&
4509 Left
.isOneOf(Keywords
.kw_function
, Keywords
.kw_yield
)) {
4512 if (Right
.isOneOf(tok::l_brace
, tok::l_square
) &&
4513 Left
.isOneOf(Keywords
.kw_function
, Keywords
.kw_yield
,
4514 Keywords
.kw_extends
, Keywords
.kw_implements
)) {
4517 if (Right
.is(tok::l_paren
)) {
4518 // JS methods can use some keywords as names (e.g. `delete()`).
4519 if (Line
.MustBeDeclaration
&& Left
.Tok
.getIdentifierInfo())
4521 // Valid JS method names can include keywords, e.g. `foo.delete()` or
4522 // `bar.instanceof()`. Recognize call positions by preceding period.
4523 if (Left
.Previous
&& Left
.Previous
->is(tok::period
) &&
4524 Left
.Tok
.getIdentifierInfo()) {
4527 // Additional unary JavaScript operators that need a space after.
4528 if (Left
.isOneOf(tok::kw_throw
, Keywords
.kw_await
, Keywords
.kw_typeof
,
4533 // `foo as const;` casts into a const type.
4534 if (Left
.endsSequence(tok::kw_const
, Keywords
.kw_as
))
4536 if ((Left
.isOneOf(Keywords
.kw_let
, Keywords
.kw_var
, Keywords
.kw_in
,
4538 // "of" is only a keyword if it appears after another identifier
4539 // (e.g. as "const x of y" in a for loop), or after a destructuring
4540 // operation (const [x, y] of z, const {a, b} of c).
4541 (Left
.is(Keywords
.kw_of
) && Left
.Previous
&&
4542 (Left
.Previous
->is(tok::identifier
) ||
4543 Left
.Previous
->isOneOf(tok::r_square
, tok::r_brace
)))) &&
4544 (!Left
.Previous
|| Left
.Previous
->isNot(tok::period
))) {
4547 if (Left
.isOneOf(tok::kw_for
, Keywords
.kw_as
) && Left
.Previous
&&
4548 Left
.Previous
->is(tok::period
) && Right
.is(tok::l_paren
)) {
4551 if (Left
.is(Keywords
.kw_as
) &&
4552 Right
.isOneOf(tok::l_square
, tok::l_brace
, tok::l_paren
)) {
4555 if (Left
.is(tok::kw_default
) && Left
.Previous
&&
4556 Left
.Previous
->is(tok::kw_export
)) {
4559 if (Left
.is(Keywords
.kw_is
) && Right
.is(tok::l_brace
))
4561 if (Right
.isOneOf(TT_JsTypeColon
, TT_JsTypeOptionalQuestion
))
4563 if (Left
.is(TT_JsTypeOperator
) || Right
.is(TT_JsTypeOperator
))
4565 if ((Left
.is(tok::l_brace
) || Right
.is(tok::r_brace
)) &&
4566 Line
.First
->isOneOf(Keywords
.kw_import
, tok::kw_export
)) {
4569 if (Left
.is(tok::ellipsis
))
4571 if (Left
.is(TT_TemplateCloser
) &&
4572 !Right
.isOneOf(tok::equal
, tok::l_brace
, tok::comma
, tok::l_square
,
4573 Keywords
.kw_implements
, Keywords
.kw_extends
)) {
4574 // Type assertions ('<type>expr') are not followed by whitespace. Other
4575 // locations that should have whitespace following are identified by the
4576 // above set of follower tokens.
4579 if (Right
.is(TT_NonNullAssertion
))
4581 if (Left
.is(TT_NonNullAssertion
) &&
4582 Right
.isOneOf(Keywords
.kw_as
, Keywords
.kw_in
)) {
4583 return true; // "x! as string", "x! in y"
4585 } else if (Style
.Language
== FormatStyle::LK_Java
) {
4586 if (Left
.is(tok::r_square
) && Right
.is(tok::l_brace
))
4588 if (Left
.is(Keywords
.kw_synchronized
) && Right
.is(tok::l_paren
)) {
4589 return Style
.SpaceBeforeParensOptions
.AfterControlStatements
||
4590 spaceRequiredBeforeParens(Right
);
4592 if ((Left
.isOneOf(tok::kw_static
, tok::kw_public
, tok::kw_private
,
4593 tok::kw_protected
) ||
4594 Left
.isOneOf(Keywords
.kw_final
, Keywords
.kw_abstract
,
4595 Keywords
.kw_native
)) &&
4596 Right
.is(TT_TemplateOpener
)) {
4599 } else if (Style
.isVerilog()) {
4600 // An escaped identifier ends with whitespace.
4601 if (Style
.isVerilog() && Left
.is(tok::identifier
) &&
4602 Left
.TokenText
[0] == '\\') {
4605 // Add space between things in a primitive's state table unless in a
4606 // transition like `(0?)`.
4607 if ((Left
.is(TT_VerilogTableItem
) &&
4608 !Right
.isOneOf(tok::r_paren
, tok::semi
)) ||
4609 (Right
.is(TT_VerilogTableItem
) && Left
.isNot(tok::l_paren
))) {
4610 const FormatToken
*Next
= Right
.getNextNonComment();
4611 return !(Next
&& Next
->is(tok::r_paren
));
4613 // Don't add space within a delay like `#0`.
4614 if (Left
.isNot(TT_BinaryOperator
) &&
4615 Left
.isOneOf(Keywords
.kw_verilogHash
, Keywords
.kw_verilogHashHash
)) {
4618 // Add space after a delay.
4619 if (Right
.isNot(tok::semi
) &&
4620 (Left
.endsSequence(tok::numeric_constant
, Keywords
.kw_verilogHash
) ||
4621 Left
.endsSequence(tok::numeric_constant
,
4622 Keywords
.kw_verilogHashHash
) ||
4623 (Left
.is(tok::r_paren
) && Left
.MatchingParen
&&
4624 Left
.MatchingParen
->endsSequence(tok::l_paren
, tok::at
)))) {
4627 // Don't add embedded spaces in a number literal like `16'h1?ax` or an array
4628 // literal like `'{}`.
4629 if (Left
.is(Keywords
.kw_apostrophe
) ||
4630 (Left
.is(TT_VerilogNumberBase
) && Right
.is(tok::numeric_constant
))) {
4633 // Don't add spaces between two at signs. Like in a coverage event.
4634 // Don't add spaces between at and a sensitivity list like
4635 // `@(posedge clk)`.
4636 if (Left
.is(tok::at
) && Right
.isOneOf(tok::l_paren
, tok::star
, tok::at
))
4638 // Add space between the type name and dimension like `logic [1:0]`.
4639 if (Right
.is(tok::l_square
) &&
4640 Left
.isOneOf(TT_VerilogDimensionedTypeName
, Keywords
.kw_function
)) {
4643 // Don't add spaces between a casting type and the quote or repetition count
4645 if ((Right
.is(Keywords
.kw_apostrophe
) ||
4646 (Right
.is(BK_BracedInit
) && Right
.is(tok::l_brace
))) &&
4647 !(Left
.isOneOf(Keywords
.kw_assign
, Keywords
.kw_unique
) ||
4648 Keywords
.isVerilogWordOperator(Left
)) &&
4649 (Left
.isOneOf(tok::r_square
, tok::r_paren
, tok::r_brace
,
4650 tok::numeric_constant
) ||
4651 Keywords
.isWordLike(Left
))) {
4654 // Don't add spaces in imports like `import foo::*;`.
4655 if ((Right
.is(tok::star
) && Left
.is(tok::coloncolon
)) ||
4656 (Left
.is(tok::star
) && Right
.is(tok::semi
))) {
4659 // Add space in attribute like `(* ASYNC_REG = "TRUE" *)`.
4660 if (Left
.endsSequence(tok::star
, tok::l_paren
) && Right
.is(tok::identifier
))
4662 // Add space before drive strength like in `wire (strong1, pull0)`.
4663 if (Right
.is(tok::l_paren
) && Right
.is(TT_VerilogStrength
))
4665 // Don't add space in a streaming concatenation like `{>>{j}}`.
4666 if ((Left
.is(tok::l_brace
) &&
4667 Right
.isOneOf(tok::lessless
, tok::greatergreater
)) ||
4668 (Left
.endsSequence(tok::lessless
, tok::l_brace
) ||
4669 Left
.endsSequence(tok::greatergreater
, tok::l_brace
))) {
4673 if (Left
.is(TT_ImplicitStringLiteral
))
4674 return Right
.hasWhitespaceBefore();
4675 if (Line
.Type
== LT_ObjCMethodDecl
) {
4676 if (Left
.is(TT_ObjCMethodSpecifier
))
4678 if (Left
.is(tok::r_paren
) && canBeObjCSelectorComponent(Right
)) {
4679 // Don't space between ')' and <id> or ')' and 'new'. 'new' is not a
4680 // keyword in Objective-C, and '+ (instancetype)new;' is a standard class
4681 // method declaration.
4685 if (Line
.Type
== LT_ObjCProperty
&&
4686 (Right
.is(tok::equal
) || Left
.is(tok::equal
))) {
4690 if (Right
.isOneOf(TT_TrailingReturnArrow
, TT_LambdaArrow
) ||
4691 Left
.isOneOf(TT_TrailingReturnArrow
, TT_LambdaArrow
)) {
4694 if (Left
.is(tok::comma
) && Right
.isNot(TT_OverloadedOperatorLParen
) &&
4695 // In an unexpanded macro call we only find the parentheses and commas
4696 // in a line; the commas and closing parenthesis do not require a space.
4697 (Left
.Children
.empty() || !Left
.MacroParent
)) {
4700 if (Right
.is(tok::comma
))
4702 if (Right
.is(TT_ObjCBlockLParen
))
4704 if (Right
.is(TT_CtorInitializerColon
))
4705 return Style
.SpaceBeforeCtorInitializerColon
;
4706 if (Right
.is(TT_InheritanceColon
) && !Style
.SpaceBeforeInheritanceColon
)
4708 if (Right
.is(TT_RangeBasedForLoopColon
) &&
4709 !Style
.SpaceBeforeRangeBasedForLoopColon
) {
4712 if (Left
.is(TT_BitFieldColon
)) {
4713 return Style
.BitFieldColonSpacing
== FormatStyle::BFCS_Both
||
4714 Style
.BitFieldColonSpacing
== FormatStyle::BFCS_After
;
4716 if (Right
.is(tok::colon
)) {
4717 if (Right
.is(TT_CaseLabelColon
))
4718 return Style
.SpaceBeforeCaseColon
;
4719 if (Right
.is(TT_GotoLabelColon
))
4721 // `private:` and `public:`.
4722 if (!Right
.getNextNonComment())
4724 if (Right
.is(TT_ObjCMethodExpr
))
4726 if (Left
.is(tok::question
))
4728 if (Right
.is(TT_InlineASMColon
) && Left
.is(tok::coloncolon
))
4730 if (Right
.is(TT_DictLiteral
))
4731 return Style
.SpacesInContainerLiterals
;
4732 if (Right
.is(TT_AttributeColon
))
4734 if (Right
.is(TT_CSharpNamedArgumentColon
))
4736 if (Right
.is(TT_GenericSelectionColon
))
4738 if (Right
.is(TT_BitFieldColon
)) {
4739 return Style
.BitFieldColonSpacing
== FormatStyle::BFCS_Both
||
4740 Style
.BitFieldColonSpacing
== FormatStyle::BFCS_Before
;
4744 // Do not merge "- -" into "--".
4745 if ((Left
.isOneOf(tok::minus
, tok::minusminus
) &&
4746 Right
.isOneOf(tok::minus
, tok::minusminus
)) ||
4747 (Left
.isOneOf(tok::plus
, tok::plusplus
) &&
4748 Right
.isOneOf(tok::plus
, tok::plusplus
))) {
4751 if (Left
.is(TT_UnaryOperator
)) {
4752 if (Right
.isNot(tok::l_paren
)) {
4753 // The alternative operators for ~ and ! are "compl" and "not".
4754 // If they are used instead, we do not want to combine them with
4755 // the token to the right, unless that is a left paren.
4756 if (Left
.is(tok::exclaim
) && Left
.TokenText
== "not")
4758 if (Left
.is(tok::tilde
) && Left
.TokenText
== "compl")
4760 // Lambda captures allow for a lone &, so "&]" needs to be properly
4762 if (Left
.is(tok::amp
) && Right
.is(tok::r_square
))
4763 return Style
.SpacesInSquareBrackets
;
4765 return (Style
.SpaceAfterLogicalNot
&& Left
.is(tok::exclaim
)) ||
4766 Right
.is(TT_BinaryOperator
);
4769 // If the next token is a binary operator or a selector name, we have
4770 // incorrectly classified the parenthesis as a cast. FIXME: Detect correctly.
4771 if (Left
.is(TT_CastRParen
)) {
4772 return Style
.SpaceAfterCStyleCast
||
4773 Right
.isOneOf(TT_BinaryOperator
, TT_SelectorName
);
4776 auto ShouldAddSpacesInAngles
= [this, &Right
]() {
4777 if (this->Style
.SpacesInAngles
== FormatStyle::SIAS_Always
)
4779 if (this->Style
.SpacesInAngles
== FormatStyle::SIAS_Leave
)
4780 return Right
.hasWhitespaceBefore();
4784 if (Left
.is(tok::greater
) && Right
.is(tok::greater
)) {
4785 if (Style
.Language
== FormatStyle::LK_TextProto
||
4786 (Style
.Language
== FormatStyle::LK_Proto
&& Left
.is(TT_DictLiteral
))) {
4787 return !Style
.Cpp11BracedListStyle
;
4789 return Right
.is(TT_TemplateCloser
) && Left
.is(TT_TemplateCloser
) &&
4790 ((Style
.Standard
< FormatStyle::LS_Cpp11
) ||
4791 ShouldAddSpacesInAngles());
4793 if (Right
.isOneOf(tok::arrow
, tok::arrowstar
, tok::periodstar
) ||
4794 Left
.isOneOf(tok::arrow
, tok::period
, tok::arrowstar
, tok::periodstar
) ||
4795 (Right
.is(tok::period
) && Right
.isNot(TT_DesignatedInitializerPeriod
))) {
4798 if (!Style
.SpaceBeforeAssignmentOperators
&& Left
.isNot(TT_TemplateCloser
) &&
4799 Right
.getPrecedence() == prec::Assignment
) {
4802 if (Style
.Language
== FormatStyle::LK_Java
&& Right
.is(tok::coloncolon
) &&
4803 (Left
.is(tok::identifier
) || Left
.is(tok::kw_this
))) {
4806 if (Right
.is(tok::coloncolon
) && Left
.is(tok::identifier
)) {
4807 // Generally don't remove existing spaces between an identifier and "::".
4808 // The identifier might actually be a macro name such as ALWAYS_INLINE. If
4809 // this turns out to be too lenient, add analysis of the identifier itself.
4810 return Right
.hasWhitespaceBefore();
4812 if (Right
.is(tok::coloncolon
) &&
4813 !Left
.isOneOf(tok::l_brace
, tok::comment
, tok::l_paren
)) {
4814 // Put a space between < and :: in vector< ::std::string >
4815 return (Left
.is(TT_TemplateOpener
) &&
4816 ((Style
.Standard
< FormatStyle::LS_Cpp11
) ||
4817 ShouldAddSpacesInAngles())) ||
4818 !(Left
.isOneOf(tok::l_paren
, tok::r_paren
, tok::l_square
,
4819 tok::kw___super
, TT_TemplateOpener
,
4820 TT_TemplateCloser
)) ||
4821 (Left
.is(tok::l_paren
) && Style
.SpacesInParensOptions
.Other
);
4823 if ((Left
.is(TT_TemplateOpener
)) != (Right
.is(TT_TemplateCloser
)))
4824 return ShouldAddSpacesInAngles();
4825 // Space before TT_StructuredBindingLSquare.
4826 if (Right
.is(TT_StructuredBindingLSquare
)) {
4827 return !Left
.isOneOf(tok::amp
, tok::ampamp
) ||
4828 getTokenReferenceAlignment(Left
) != FormatStyle::PAS_Right
;
4830 // Space before & or && following a TT_StructuredBindingLSquare.
4831 if (Right
.Next
&& Right
.Next
->is(TT_StructuredBindingLSquare
) &&
4832 Right
.isOneOf(tok::amp
, tok::ampamp
)) {
4833 return getTokenReferenceAlignment(Right
) != FormatStyle::PAS_Left
;
4835 if ((Right
.is(TT_BinaryOperator
) && Left
.isNot(tok::l_paren
)) ||
4836 (Left
.isOneOf(TT_BinaryOperator
, TT_ConditionalExpr
) &&
4837 Right
.isNot(tok::r_paren
))) {
4840 if (Right
.is(TT_TemplateOpener
) && Left
.is(tok::r_paren
) &&
4841 Left
.MatchingParen
&&
4842 Left
.MatchingParen
->is(TT_OverloadedOperatorLParen
)) {
4845 if (Right
.is(tok::less
) && Left
.isNot(tok::l_paren
) &&
4846 Line
.Type
== LT_ImportStatement
) {
4849 if (Right
.is(TT_TrailingUnaryOperator
))
4851 if (Left
.is(TT_RegexLiteral
))
4853 return spaceRequiredBetween(Line
, Left
, Right
);
4856 // Returns 'true' if 'Tok' is a brace we'd want to break before in Allman style.
4857 static bool isAllmanBrace(const FormatToken
&Tok
) {
4858 return Tok
.is(tok::l_brace
) && Tok
.is(BK_Block
) &&
4859 !Tok
.isOneOf(TT_ObjCBlockLBrace
, TT_LambdaLBrace
, TT_DictLiteral
);
4862 // Returns 'true' if 'Tok' is a function argument.
4863 static bool IsFunctionArgument(const FormatToken
&Tok
) {
4864 return Tok
.MatchingParen
&& Tok
.MatchingParen
->Next
&&
4865 Tok
.MatchingParen
->Next
->isOneOf(tok::comma
, tok::r_paren
);
4869 isItAnEmptyLambdaAllowed(const FormatToken
&Tok
,
4870 FormatStyle::ShortLambdaStyle ShortLambdaOption
) {
4871 return Tok
.Children
.empty() && ShortLambdaOption
!= FormatStyle::SLS_None
;
4874 static bool isAllmanLambdaBrace(const FormatToken
&Tok
) {
4875 return Tok
.is(tok::l_brace
) && Tok
.is(BK_Block
) &&
4876 !Tok
.isOneOf(TT_ObjCBlockLBrace
, TT_DictLiteral
);
4879 bool TokenAnnotator::mustBreakBefore(const AnnotatedLine
&Line
,
4880 const FormatToken
&Right
) const {
4881 const FormatToken
&Left
= *Right
.Previous
;
4882 if (Right
.NewlinesBefore
> 1 && Style
.MaxEmptyLinesToKeep
> 0)
4885 if (Style
.isCSharp()) {
4886 if (Left
.is(TT_FatArrow
) && Right
.is(tok::l_brace
) &&
4887 Style
.BraceWrapping
.AfterFunction
) {
4890 if (Right
.is(TT_CSharpNamedArgumentColon
) ||
4891 Left
.is(TT_CSharpNamedArgumentColon
)) {
4894 if (Right
.is(TT_CSharpGenericTypeConstraint
))
4896 if (Right
.Next
&& Right
.Next
->is(TT_FatArrow
) &&
4897 (Right
.is(tok::numeric_constant
) ||
4898 (Right
.is(tok::identifier
) && Right
.TokenText
== "_"))) {
4902 // Break after C# [...] and before public/protected/private/internal.
4903 if (Left
.is(TT_AttributeSquare
) && Left
.is(tok::r_square
) &&
4904 (Right
.isAccessSpecifier(/*ColonRequired=*/false) ||
4905 Right
.is(Keywords
.kw_internal
))) {
4908 // Break between ] and [ but only when there are really 2 attributes.
4909 if (Left
.is(TT_AttributeSquare
) && Right
.is(TT_AttributeSquare
) &&
4910 Left
.is(tok::r_square
) && Right
.is(tok::l_square
)) {
4914 } else if (Style
.isJavaScript()) {
4915 // FIXME: This might apply to other languages and token kinds.
4916 if (Right
.is(tok::string_literal
) && Left
.is(tok::plus
) && Left
.Previous
&&
4917 Left
.Previous
->is(tok::string_literal
)) {
4920 if (Left
.is(TT_DictLiteral
) && Left
.is(tok::l_brace
) && Line
.Level
== 0 &&
4921 Left
.Previous
&& Left
.Previous
->is(tok::equal
) &&
4922 Line
.First
->isOneOf(tok::identifier
, Keywords
.kw_import
, tok::kw_export
,
4924 // kw_var/kw_let are pseudo-tokens that are tok::identifier, so match
4926 !Line
.First
->isOneOf(Keywords
.kw_var
, Keywords
.kw_let
)) {
4927 // Object literals on the top level of a file are treated as "enum-style".
4928 // Each key/value pair is put on a separate line, instead of bin-packing.
4931 if (Left
.is(tok::l_brace
) && Line
.Level
== 0 &&
4932 (Line
.startsWith(tok::kw_enum
) ||
4933 Line
.startsWith(tok::kw_const
, tok::kw_enum
) ||
4934 Line
.startsWith(tok::kw_export
, tok::kw_enum
) ||
4935 Line
.startsWith(tok::kw_export
, tok::kw_const
, tok::kw_enum
))) {
4936 // JavaScript top-level enum key/value pairs are put on separate lines
4937 // instead of bin-packing.
4940 if (Right
.is(tok::r_brace
) && Left
.is(tok::l_brace
) && Left
.Previous
&&
4941 Left
.Previous
->is(TT_FatArrow
)) {
4942 // JS arrow function (=> {...}).
4943 switch (Style
.AllowShortLambdasOnASingleLine
) {
4944 case FormatStyle::SLS_All
:
4946 case FormatStyle::SLS_None
:
4948 case FormatStyle::SLS_Empty
:
4949 return !Left
.Children
.empty();
4950 case FormatStyle::SLS_Inline
:
4951 // allow one-lining inline (e.g. in function call args) and empty arrow
4953 return (Left
.NestingLevel
== 0 && Line
.Level
== 0) &&
4954 !Left
.Children
.empty();
4956 llvm_unreachable("Unknown FormatStyle::ShortLambdaStyle enum");
4959 if (Right
.is(tok::r_brace
) && Left
.is(tok::l_brace
) &&
4960 !Left
.Children
.empty()) {
4961 // Support AllowShortFunctionsOnASingleLine for JavaScript.
4962 return Style
.AllowShortFunctionsOnASingleLine
== FormatStyle::SFS_None
||
4963 Style
.AllowShortFunctionsOnASingleLine
== FormatStyle::SFS_Empty
||
4964 (Left
.NestingLevel
== 0 && Line
.Level
== 0 &&
4965 Style
.AllowShortFunctionsOnASingleLine
&
4966 FormatStyle::SFS_InlineOnly
);
4968 } else if (Style
.Language
== FormatStyle::LK_Java
) {
4969 if (Right
.is(tok::plus
) && Left
.is(tok::string_literal
) && Right
.Next
&&
4970 Right
.Next
->is(tok::string_literal
)) {
4973 } else if (Style
.isVerilog()) {
4974 // Break between assignments.
4975 if (Left
.is(TT_VerilogAssignComma
))
4977 // Break between ports of different types.
4978 if (Left
.is(TT_VerilogTypeComma
))
4980 // Break between ports in a module instantiation and after the parameter
4982 if (Style
.VerilogBreakBetweenInstancePorts
&&
4983 (Left
.is(TT_VerilogInstancePortComma
) ||
4984 (Left
.is(tok::r_paren
) && Keywords
.isVerilogIdentifier(Right
) &&
4985 Left
.MatchingParen
&&
4986 Left
.MatchingParen
->is(TT_VerilogInstancePortLParen
)))) {
4989 // Break after labels. In Verilog labels don't have the 'case' keyword, so
4990 // it is hard to identify them in UnwrappedLineParser.
4991 if (!Keywords
.isVerilogBegin(Right
) && Keywords
.isVerilogEndOfLabel(Left
))
4993 } else if (Style
.Language
== FormatStyle::LK_Cpp
||
4994 Style
.Language
== FormatStyle::LK_ObjC
||
4995 Style
.Language
== FormatStyle::LK_Proto
||
4996 Style
.Language
== FormatStyle::LK_TableGen
||
4997 Style
.Language
== FormatStyle::LK_TextProto
) {
4998 if (Left
.isStringLiteral() && Right
.isStringLiteral())
5002 // Basic JSON newline processing.
5003 if (Style
.isJson()) {
5004 // Always break after a JSON record opener.
5007 if (Left
.is(TT_DictLiteral
) && Left
.is(tok::l_brace
))
5009 // Always break after a JSON array opener based on BreakArrays.
5010 if ((Left
.is(TT_ArrayInitializerLSquare
) && Left
.is(tok::l_square
) &&
5011 Right
.isNot(tok::r_square
)) ||
5012 Left
.is(tok::comma
)) {
5013 if (Right
.is(tok::l_brace
))
5015 // scan to the right if an we see an object or an array inside
5017 for (const auto *Tok
= &Right
; Tok
; Tok
= Tok
->Next
) {
5018 if (Tok
->isOneOf(tok::l_brace
, tok::l_square
))
5020 if (Tok
->isOneOf(tok::r_brace
, tok::r_square
))
5023 return Style
.BreakArrays
;
5027 if (Line
.startsWith(tok::kw_asm
) && Right
.is(TT_InlineASMColon
) &&
5028 Style
.BreakBeforeInlineASMColon
== FormatStyle::BBIAS_Always
) {
5032 // If the last token before a '}', ']', or ')' is a comma or a trailing
5033 // comment, the intention is to insert a line break after it in order to make
5034 // shuffling around entries easier. Import statements, especially in
5035 // JavaScript, can be an exception to this rule.
5036 if (Style
.JavaScriptWrapImports
|| Line
.Type
!= LT_ImportStatement
) {
5037 const FormatToken
*BeforeClosingBrace
= nullptr;
5038 if ((Left
.isOneOf(tok::l_brace
, TT_ArrayInitializerLSquare
) ||
5039 (Style
.isJavaScript() && Left
.is(tok::l_paren
))) &&
5040 Left
.isNot(BK_Block
) && Left
.MatchingParen
) {
5041 BeforeClosingBrace
= Left
.MatchingParen
->Previous
;
5042 } else if (Right
.MatchingParen
&&
5043 (Right
.MatchingParen
->isOneOf(tok::l_brace
,
5044 TT_ArrayInitializerLSquare
) ||
5045 (Style
.isJavaScript() &&
5046 Right
.MatchingParen
->is(tok::l_paren
)))) {
5047 BeforeClosingBrace
= &Left
;
5049 if (BeforeClosingBrace
&& (BeforeClosingBrace
->is(tok::comma
) ||
5050 BeforeClosingBrace
->isTrailingComment())) {
5055 if (Right
.is(tok::comment
)) {
5056 return Left
.isNot(BK_BracedInit
) && Left
.isNot(TT_CtorInitializerColon
) &&
5057 (Right
.NewlinesBefore
> 0 && Right
.HasUnescapedNewline
);
5059 if (Left
.isTrailingComment())
5061 if (Left
.IsUnterminatedLiteral
)
5063 if (Right
.is(tok::lessless
) && Right
.Next
&& Left
.is(tok::string_literal
) &&
5064 Right
.Next
->is(tok::string_literal
)) {
5067 if (Right
.is(TT_RequiresClause
)) {
5068 switch (Style
.RequiresClausePosition
) {
5069 case FormatStyle::RCPS_OwnLine
:
5070 case FormatStyle::RCPS_WithFollowing
:
5076 // Can break after template<> declaration
5077 if (Left
.ClosesTemplateDeclaration
&& Left
.MatchingParen
&&
5078 Left
.MatchingParen
->NestingLevel
== 0) {
5079 // Put concepts on the next line e.g.
5080 // template<typename T>
5082 if (Right
.is(tok::kw_concept
))
5083 return Style
.BreakBeforeConceptDeclarations
== FormatStyle::BBCDS_Always
;
5084 return Style
.AlwaysBreakTemplateDeclarations
== FormatStyle::BTDS_Yes
;
5086 if (Left
.ClosesRequiresClause
&& Right
.isNot(tok::semi
)) {
5087 switch (Style
.RequiresClausePosition
) {
5088 case FormatStyle::RCPS_OwnLine
:
5089 case FormatStyle::RCPS_WithPreceding
:
5095 if (Style
.PackConstructorInitializers
== FormatStyle::PCIS_Never
) {
5096 if (Style
.BreakConstructorInitializers
== FormatStyle::BCIS_BeforeColon
&&
5097 (Left
.is(TT_CtorInitializerComma
) ||
5098 Right
.is(TT_CtorInitializerColon
))) {
5102 if (Style
.BreakConstructorInitializers
== FormatStyle::BCIS_AfterColon
&&
5103 Left
.isOneOf(TT_CtorInitializerColon
, TT_CtorInitializerComma
)) {
5107 if (Style
.PackConstructorInitializers
< FormatStyle::PCIS_CurrentLine
&&
5108 Style
.BreakConstructorInitializers
== FormatStyle::BCIS_BeforeComma
&&
5109 Right
.isOneOf(TT_CtorInitializerComma
, TT_CtorInitializerColon
)) {
5112 if (Style
.PackConstructorInitializers
== FormatStyle::PCIS_NextLineOnly
) {
5113 if ((Style
.BreakConstructorInitializers
== FormatStyle::BCIS_BeforeColon
||
5114 Style
.BreakConstructorInitializers
== FormatStyle::BCIS_BeforeComma
) &&
5115 Right
.is(TT_CtorInitializerColon
)) {
5119 if (Style
.BreakConstructorInitializers
== FormatStyle::BCIS_AfterColon
&&
5120 Left
.is(TT_CtorInitializerColon
)) {
5124 // Break only if we have multiple inheritance.
5125 if (Style
.BreakInheritanceList
== FormatStyle::BILS_BeforeComma
&&
5126 Right
.is(TT_InheritanceComma
)) {
5129 if (Style
.BreakInheritanceList
== FormatStyle::BILS_AfterComma
&&
5130 Left
.is(TT_InheritanceComma
)) {
5133 if (Right
.is(tok::string_literal
) && Right
.TokenText
.startswith("R\"")) {
5134 // Multiline raw string literals are special wrt. line breaks. The author
5135 // has made a deliberate choice and might have aligned the contents of the
5136 // string literal accordingly. Thus, we try keep existing line breaks.
5137 return Right
.IsMultiline
&& Right
.NewlinesBefore
> 0;
5139 if ((Left
.is(tok::l_brace
) || (Left
.is(tok::less
) && Left
.Previous
&&
5140 Left
.Previous
->is(tok::equal
))) &&
5141 Right
.NestingLevel
== 1 && Style
.Language
== FormatStyle::LK_Proto
) {
5142 // Don't put enums or option definitions onto single lines in protocol
5146 if (Right
.is(TT_InlineASMBrace
))
5147 return Right
.HasUnescapedNewline
;
5149 if (isAllmanBrace(Left
) || isAllmanBrace(Right
)) {
5150 auto *FirstNonComment
= Line
.getFirstNonComment();
5151 bool AccessSpecifier
=
5153 FirstNonComment
->isOneOf(Keywords
.kw_internal
, tok::kw_public
,
5154 tok::kw_private
, tok::kw_protected
);
5156 if (Style
.BraceWrapping
.AfterEnum
) {
5157 if (Line
.startsWith(tok::kw_enum
) ||
5158 Line
.startsWith(tok::kw_typedef
, tok::kw_enum
)) {
5161 // Ensure BraceWrapping for `public enum A {`.
5162 if (AccessSpecifier
&& FirstNonComment
->Next
&&
5163 FirstNonComment
->Next
->is(tok::kw_enum
)) {
5168 // Ensure BraceWrapping for `public interface A {`.
5169 if (Style
.BraceWrapping
.AfterClass
&&
5170 ((AccessSpecifier
&& FirstNonComment
->Next
&&
5171 FirstNonComment
->Next
->is(Keywords
.kw_interface
)) ||
5172 Line
.startsWith(Keywords
.kw_interface
))) {
5176 // Don't attempt to interpret struct return types as structs.
5177 if (Right
.isNot(TT_FunctionLBrace
)) {
5178 return (Line
.startsWith(tok::kw_class
) &&
5179 Style
.BraceWrapping
.AfterClass
) ||
5180 (Line
.startsWith(tok::kw_struct
) &&
5181 Style
.BraceWrapping
.AfterStruct
);
5185 if (Left
.is(TT_ObjCBlockLBrace
) &&
5186 Style
.AllowShortBlocksOnASingleLine
== FormatStyle::SBS_Never
) {
5190 // Ensure wrapping after __attribute__((XX)) and @interface etc.
5191 if (Left
.is(TT_AttributeParen
) && Right
.is(TT_ObjCDecl
))
5194 if (Left
.is(TT_LambdaLBrace
)) {
5195 if (IsFunctionArgument(Left
) &&
5196 Style
.AllowShortLambdasOnASingleLine
== FormatStyle::SLS_Inline
) {
5200 if (Style
.AllowShortLambdasOnASingleLine
== FormatStyle::SLS_None
||
5201 Style
.AllowShortLambdasOnASingleLine
== FormatStyle::SLS_Inline
||
5202 (!Left
.Children
.empty() &&
5203 Style
.AllowShortLambdasOnASingleLine
== FormatStyle::SLS_Empty
)) {
5208 if (Style
.BraceWrapping
.BeforeLambdaBody
&& Right
.is(TT_LambdaLBrace
) &&
5209 Left
.isOneOf(tok::star
, tok::amp
, tok::ampamp
, TT_TemplateCloser
)) {
5213 // Put multiple Java annotation on a new line.
5214 if ((Style
.Language
== FormatStyle::LK_Java
|| Style
.isJavaScript()) &&
5215 Left
.is(TT_LeadingJavaAnnotation
) &&
5216 Right
.isNot(TT_LeadingJavaAnnotation
) && Right
.isNot(tok::l_paren
) &&
5217 (Line
.Last
->is(tok::l_brace
) || Style
.BreakAfterJavaFieldAnnotations
)) {
5221 if (Right
.is(TT_ProtoExtensionLSquare
))
5224 // In text proto instances if a submessage contains at least 2 entries and at
5225 // least one of them is a submessage, like A { ... B { ... } ... },
5226 // put all of the entries of A on separate lines by forcing the selector of
5227 // the submessage B to be put on a newline.
5229 // Example: these can stay on one line:
5230 // a { scalar_1: 1 scalar_2: 2 }
5231 // a { b { key: value } }
5233 // and these entries need to be on a new line even if putting them all in one
5234 // line is under the column limit:
5240 // We enforce this by breaking before a submessage field that has previous
5241 // siblings, *and* breaking before a field that follows a submessage field.
5243 // Be careful to exclude the case [proto.ext] { ... } since the `]` is
5244 // the TT_SelectorName there, but we don't want to break inside the brackets.
5246 // Another edge case is @submessage { key: value }, which is a common
5247 // substitution placeholder. In this case we want to keep `@` and `submessage`
5250 // We ensure elsewhere that extensions are always on their own line.
5251 if ((Style
.Language
== FormatStyle::LK_Proto
||
5252 Style
.Language
== FormatStyle::LK_TextProto
) &&
5253 Right
.is(TT_SelectorName
) && Right
.isNot(tok::r_square
) && Right
.Next
) {
5254 // Keep `@submessage` together in:
5255 // @submessage { key: value }
5256 if (Left
.is(tok::at
))
5258 // Look for the scope opener after selector in cases like:
5261 // selector: @base { ...
5262 FormatToken
*LBrace
= Right
.Next
;
5263 if (LBrace
&& LBrace
->is(tok::colon
)) {
5264 LBrace
= LBrace
->Next
;
5265 if (LBrace
&& LBrace
->is(tok::at
)) {
5266 LBrace
= LBrace
->Next
;
5268 LBrace
= LBrace
->Next
;
5272 // The scope opener is one of {, [, <:
5277 // In case of selector { ... }, the l_brace is TT_DictLiteral.
5278 // In case of an empty selector {}, the l_brace is not TT_DictLiteral,
5279 // so we check for immediately following r_brace.
5280 ((LBrace
->is(tok::l_brace
) &&
5281 (LBrace
->is(TT_DictLiteral
) ||
5282 (LBrace
->Next
&& LBrace
->Next
->is(tok::r_brace
)))) ||
5283 LBrace
->is(TT_ArrayInitializerLSquare
) || LBrace
->is(tok::less
))) {
5284 // If Left.ParameterCount is 0, then this submessage entry is not the
5285 // first in its parent submessage, and we want to break before this entry.
5286 // If Left.ParameterCount is greater than 0, then its parent submessage
5287 // might contain 1 or more entries and we want to break before this entry
5288 // if it contains at least 2 entries. We deal with this case later by
5289 // detecting and breaking before the next entry in the parent submessage.
5290 if (Left
.ParameterCount
== 0)
5292 // However, if this submessage is the first entry in its parent
5293 // submessage, Left.ParameterCount might be 1 in some cases.
5294 // We deal with this case later by detecting an entry
5295 // following a closing paren of this submessage.
5298 // If this is an entry immediately following a submessage, it will be
5299 // preceded by a closing paren of that submessage, like in:
5300 // left---. .---right
5302 // sub: { ... } key: value
5303 // If there was a comment between `}` an `key` above, then `key` would be
5304 // put on a new line anyways.
5305 if (Left
.isOneOf(tok::r_brace
, tok::greater
, tok::r_square
))
5312 bool TokenAnnotator::canBreakBefore(const AnnotatedLine
&Line
,
5313 const FormatToken
&Right
) const {
5314 const FormatToken
&Left
= *Right
.Previous
;
5315 // Language-specific stuff.
5316 if (Style
.isCSharp()) {
5317 if (Left
.isOneOf(TT_CSharpNamedArgumentColon
, TT_AttributeColon
) ||
5318 Right
.isOneOf(TT_CSharpNamedArgumentColon
, TT_AttributeColon
)) {
5321 // Only break after commas for generic type constraints.
5322 if (Line
.First
->is(TT_CSharpGenericTypeConstraint
))
5323 return Left
.is(TT_CSharpGenericTypeConstraintComma
);
5324 // Keep nullable operators attached to their identifiers.
5325 if (Right
.is(TT_CSharpNullable
))
5327 } else if (Style
.Language
== FormatStyle::LK_Java
) {
5328 if (Left
.isOneOf(Keywords
.kw_throws
, Keywords
.kw_extends
,
5329 Keywords
.kw_implements
)) {
5332 if (Right
.isOneOf(Keywords
.kw_throws
, Keywords
.kw_extends
,
5333 Keywords
.kw_implements
)) {
5336 } else if (Style
.isJavaScript()) {
5337 const FormatToken
*NonComment
= Right
.getPreviousNonComment();
5339 NonComment
->isOneOf(
5340 tok::kw_return
, Keywords
.kw_yield
, tok::kw_continue
, tok::kw_break
,
5341 tok::kw_throw
, Keywords
.kw_interface
, Keywords
.kw_type
,
5342 tok::kw_static
, tok::kw_public
, tok::kw_private
, tok::kw_protected
,
5343 Keywords
.kw_readonly
, Keywords
.kw_override
, Keywords
.kw_abstract
,
5344 Keywords
.kw_get
, Keywords
.kw_set
, Keywords
.kw_async
,
5345 Keywords
.kw_await
)) {
5346 return false; // Otherwise automatic semicolon insertion would trigger.
5348 if (Right
.NestingLevel
== 0 &&
5349 (Left
.Tok
.getIdentifierInfo() ||
5350 Left
.isOneOf(tok::r_square
, tok::r_paren
)) &&
5351 Right
.isOneOf(tok::l_square
, tok::l_paren
)) {
5352 return false; // Otherwise automatic semicolon insertion would trigger.
5354 if (NonComment
&& NonComment
->is(tok::identifier
) &&
5355 NonComment
->TokenText
== "asserts") {
5358 if (Left
.is(TT_FatArrow
) && Right
.is(tok::l_brace
))
5360 if (Left
.is(TT_JsTypeColon
))
5362 // Don't wrap between ":" and "!" of a strict prop init ("field!: type;").
5363 if (Left
.is(tok::exclaim
) && Right
.is(tok::colon
))
5365 // Look for is type annotations like:
5366 // function f(): a is B { ... }
5367 // Do not break before is in these cases.
5368 if (Right
.is(Keywords
.kw_is
)) {
5369 const FormatToken
*Next
= Right
.getNextNonComment();
5370 // If `is` is followed by a colon, it's likely that it's a dict key, so
5371 // ignore it for this check.
5372 // For example this is common in Polymer:
5377 if (!Next
|| Next
->isNot(tok::colon
))
5380 if (Left
.is(Keywords
.kw_in
))
5381 return Style
.BreakBeforeBinaryOperators
== FormatStyle::BOS_None
;
5382 if (Right
.is(Keywords
.kw_in
))
5383 return Style
.BreakBeforeBinaryOperators
!= FormatStyle::BOS_None
;
5384 if (Right
.is(Keywords
.kw_as
))
5385 return false; // must not break before as in 'x as type' casts
5386 if (Right
.isOneOf(Keywords
.kw_extends
, Keywords
.kw_infer
)) {
5387 // extends and infer can appear as keywords in conditional types:
5388 // https://www.typescriptlang.org/docs/handbook/release-notes/typescript-2-8.html#conditional-types
5389 // do not break before them, as the expressions are subject to ASI.
5392 if (Left
.is(Keywords
.kw_as
))
5394 if (Left
.is(TT_NonNullAssertion
))
5396 if (Left
.is(Keywords
.kw_declare
) &&
5397 Right
.isOneOf(Keywords
.kw_module
, tok::kw_namespace
,
5398 Keywords
.kw_function
, tok::kw_class
, tok::kw_enum
,
5399 Keywords
.kw_interface
, Keywords
.kw_type
, Keywords
.kw_var
,
5400 Keywords
.kw_let
, tok::kw_const
)) {
5401 // See grammar for 'declare' statements at:
5402 // https://github.com/Microsoft/TypeScript/blob/main/doc/spec-ARCHIVED.md#A.10
5405 if (Left
.isOneOf(Keywords
.kw_module
, tok::kw_namespace
) &&
5406 Right
.isOneOf(tok::identifier
, tok::string_literal
)) {
5407 return false; // must not break in "module foo { ...}"
5409 if (Right
.is(TT_TemplateString
) && Right
.closesScope())
5411 // Don't split tagged template literal so there is a break between the tag
5412 // identifier and template string.
5413 if (Left
.is(tok::identifier
) && Right
.is(TT_TemplateString
))
5415 if (Left
.is(TT_TemplateString
) && Left
.opensScope())
5419 if (Left
.is(tok::at
))
5421 if (Left
.Tok
.getObjCKeywordID() == tok::objc_interface
)
5423 if (Left
.isOneOf(TT_JavaAnnotation
, TT_LeadingJavaAnnotation
))
5424 return Right
.isNot(tok::l_paren
);
5425 if (Right
.is(TT_PointerOrReference
)) {
5426 return Line
.IsMultiVariableDeclStmt
||
5427 (getTokenPointerOrReferenceAlignment(Right
) ==
5428 FormatStyle::PAS_Right
&&
5429 (!Right
.Next
|| Right
.Next
->isNot(TT_FunctionDeclarationName
)));
5431 if (Right
.isOneOf(TT_StartOfName
, TT_FunctionDeclarationName
) ||
5432 Right
.is(tok::kw_operator
)) {
5435 if (Left
.is(TT_PointerOrReference
))
5437 if (Right
.isTrailingComment()) {
5438 // We rely on MustBreakBefore being set correctly here as we should not
5439 // change the "binding" behavior of a comment.
5440 // The first comment in a braced lists is always interpreted as belonging to
5441 // the first list element. Otherwise, it should be placed outside of the
5443 return Left
.is(BK_BracedInit
) ||
5444 (Left
.is(TT_CtorInitializerColon
) && Right
.NewlinesBefore
> 0 &&
5445 Style
.BreakConstructorInitializers
== FormatStyle::BCIS_AfterColon
);
5447 if (Left
.is(tok::question
) && Right
.is(tok::colon
))
5449 if (Right
.is(TT_ConditionalExpr
) || Right
.is(tok::question
))
5450 return Style
.BreakBeforeTernaryOperators
;
5451 if (Left
.is(TT_ConditionalExpr
) || Left
.is(tok::question
))
5452 return !Style
.BreakBeforeTernaryOperators
;
5453 if (Left
.is(TT_InheritanceColon
))
5454 return Style
.BreakInheritanceList
== FormatStyle::BILS_AfterColon
;
5455 if (Right
.is(TT_InheritanceColon
))
5456 return Style
.BreakInheritanceList
!= FormatStyle::BILS_AfterColon
;
5457 if (Right
.is(TT_ObjCMethodExpr
) && Right
.isNot(tok::r_square
) &&
5458 Left
.isNot(TT_SelectorName
)) {
5462 if (Right
.is(tok::colon
) &&
5463 !Right
.isOneOf(TT_CtorInitializerColon
, TT_InlineASMColon
)) {
5466 if (Left
.is(tok::colon
) && Left
.isOneOf(TT_DictLiteral
, TT_ObjCMethodExpr
)) {
5467 if (Style
.Language
== FormatStyle::LK_Proto
||
5468 Style
.Language
== FormatStyle::LK_TextProto
) {
5469 if (!Style
.AlwaysBreakBeforeMultilineStrings
&& Right
.isStringLiteral())
5471 // Prevent cases like:
5474 // { key: valueeeeeeeeeeee }
5476 // when the snippet does not fit into one line.
5480 // key: valueeeeeeeeeeee
5483 // instead, even if it is longer by one line.
5485 // Note that this allows the "{" to go over the column limit
5486 // when the column limit is just between ":" and "{", but that does
5487 // not happen too often and alternative formattings in this case are
5490 // The code covers the cases:
5492 // submessage: { ... }
5493 // submessage: < ... >
5494 // repeated: [ ... ]
5495 if (((Right
.is(tok::l_brace
) || Right
.is(tok::less
)) &&
5496 Right
.is(TT_DictLiteral
)) ||
5497 Right
.is(TT_ArrayInitializerLSquare
)) {
5503 if (Right
.is(tok::r_square
) && Right
.MatchingParen
&&
5504 Right
.MatchingParen
->is(TT_ProtoExtensionLSquare
)) {
5507 if (Right
.is(TT_SelectorName
) || (Right
.is(tok::identifier
) && Right
.Next
&&
5508 Right
.Next
->is(TT_ObjCMethodExpr
))) {
5509 return Left
.isNot(tok::period
); // FIXME: Properly parse ObjC calls.
5511 if (Left
.is(tok::r_paren
) && Line
.Type
== LT_ObjCProperty
)
5513 if (Right
.is(tok::kw_concept
))
5514 return Style
.BreakBeforeConceptDeclarations
!= FormatStyle::BBCDS_Never
;
5515 if (Right
.is(TT_RequiresClause
))
5517 if (Left
.ClosesTemplateDeclaration
|| Left
.is(TT_FunctionAnnotationRParen
))
5519 if (Left
.ClosesRequiresClause
)
5521 if (Right
.isOneOf(TT_RangeBasedForLoopColon
, TT_OverloadedOperatorLParen
,
5522 TT_OverloadedOperator
)) {
5525 if (Left
.is(TT_RangeBasedForLoopColon
))
5527 if (Right
.is(TT_RangeBasedForLoopColon
))
5529 if (Left
.is(TT_TemplateCloser
) && Right
.is(TT_TemplateOpener
))
5531 if ((Left
.is(tok::greater
) && Right
.is(tok::greater
)) ||
5532 (Left
.is(tok::less
) && Right
.is(tok::less
))) {
5535 if (Right
.is(TT_BinaryOperator
) &&
5536 Style
.BreakBeforeBinaryOperators
!= FormatStyle::BOS_None
&&
5537 (Style
.BreakBeforeBinaryOperators
== FormatStyle::BOS_All
||
5538 Right
.getPrecedence() != prec::Assignment
)) {
5541 if (Left
.isOneOf(TT_TemplateCloser
, TT_UnaryOperator
) ||
5542 Left
.is(tok::kw_operator
)) {
5545 if (Left
.is(tok::equal
) && !Right
.isOneOf(tok::kw_default
, tok::kw_delete
) &&
5546 Line
.Type
== LT_VirtualFunctionDecl
&& Left
.NestingLevel
== 0) {
5549 if (Left
.is(tok::equal
) && Right
.is(tok::l_brace
) &&
5550 !Style
.Cpp11BracedListStyle
) {
5553 if (Left
.is(tok::l_paren
) &&
5554 Left
.isOneOf(TT_AttributeParen
, TT_TypeDeclarationParen
)) {
5557 if (Left
.is(tok::l_paren
) && Left
.Previous
&&
5558 (Left
.Previous
->isOneOf(TT_BinaryOperator
, TT_CastRParen
))) {
5561 if (Right
.is(TT_ImplicitStringLiteral
))
5564 if (Right
.is(TT_TemplateCloser
))
5566 if (Right
.is(tok::r_square
) && Right
.MatchingParen
&&
5567 Right
.MatchingParen
->is(TT_LambdaLSquare
)) {
5571 // We only break before r_brace if there was a corresponding break before
5572 // the l_brace, which is tracked by BreakBeforeClosingBrace.
5573 if (Right
.is(tok::r_brace
)) {
5574 return Right
.MatchingParen
&& (Right
.MatchingParen
->is(BK_Block
) ||
5575 (Right
.isBlockIndentedInitRBrace(Style
)));
5578 // We only break before r_paren if we're in a block indented context.
5579 if (Right
.is(tok::r_paren
)) {
5580 if (Style
.AlignAfterOpenBracket
!= FormatStyle::BAS_BlockIndent
||
5581 !Right
.MatchingParen
) {
5584 auto Next
= Right
.Next
;
5585 if (Next
&& Next
->is(tok::r_paren
))
5587 if (Next
&& Next
->is(tok::l_paren
))
5589 const FormatToken
*Previous
= Right
.MatchingParen
->Previous
;
5590 return !(Previous
&& (Previous
->is(tok::kw_for
) || Previous
->isIf()));
5593 // Allow breaking after a trailing annotation, e.g. after a method
5595 if (Left
.is(TT_TrailingAnnotation
)) {
5596 return !Right
.isOneOf(tok::l_brace
, tok::semi
, tok::equal
, tok::l_paren
,
5597 tok::less
, tok::coloncolon
);
5600 if (Right
.is(tok::kw___attribute
) ||
5601 (Right
.is(tok::l_square
) && Right
.is(TT_AttributeSquare
))) {
5602 return Left
.isNot(TT_AttributeSquare
);
5605 if (Left
.is(tok::identifier
) && Right
.is(tok::string_literal
))
5608 if (Right
.is(tok::identifier
) && Right
.Next
&& Right
.Next
->is(TT_DictLiteral
))
5611 if (Left
.is(TT_CtorInitializerColon
)) {
5612 return Style
.BreakConstructorInitializers
== FormatStyle::BCIS_AfterColon
&&
5613 (!Right
.isTrailingComment() || Right
.NewlinesBefore
> 0);
5615 if (Right
.is(TT_CtorInitializerColon
))
5616 return Style
.BreakConstructorInitializers
!= FormatStyle::BCIS_AfterColon
;
5617 if (Left
.is(TT_CtorInitializerComma
) &&
5618 Style
.BreakConstructorInitializers
== FormatStyle::BCIS_BeforeComma
) {
5621 if (Right
.is(TT_CtorInitializerComma
) &&
5622 Style
.BreakConstructorInitializers
== FormatStyle::BCIS_BeforeComma
) {
5625 if (Left
.is(TT_InheritanceComma
) &&
5626 Style
.BreakInheritanceList
== FormatStyle::BILS_BeforeComma
) {
5629 if (Right
.is(TT_InheritanceComma
) &&
5630 Style
.BreakInheritanceList
== FormatStyle::BILS_BeforeComma
) {
5633 if (Left
.is(TT_ArrayInitializerLSquare
))
5635 if (Right
.is(tok::kw_typename
) && Left
.isNot(tok::kw_const
))
5637 if ((Left
.isBinaryOperator() || Left
.is(TT_BinaryOperator
)) &&
5638 !Left
.isOneOf(tok::arrowstar
, tok::lessless
) &&
5639 Style
.BreakBeforeBinaryOperators
!= FormatStyle::BOS_All
&&
5640 (Style
.BreakBeforeBinaryOperators
== FormatStyle::BOS_None
||
5641 Left
.getPrecedence() == prec::Assignment
)) {
5644 if ((Left
.is(TT_AttributeSquare
) && Right
.is(tok::l_square
)) ||
5645 (Left
.is(tok::r_square
) && Right
.is(TT_AttributeSquare
))) {
5649 auto ShortLambdaOption
= Style
.AllowShortLambdasOnASingleLine
;
5650 if (Style
.BraceWrapping
.BeforeLambdaBody
&& Right
.is(TT_LambdaLBrace
)) {
5651 if (isAllmanLambdaBrace(Left
))
5652 return !isItAnEmptyLambdaAllowed(Left
, ShortLambdaOption
);
5653 if (isAllmanLambdaBrace(Right
))
5654 return !isItAnEmptyLambdaAllowed(Right
, ShortLambdaOption
);
5657 if (Right
.is(tok::kw_noexcept
) && Right
.is(TT_TrailingAnnotation
)) {
5658 switch (Style
.AllowBreakBeforeNoexceptSpecifier
) {
5659 case FormatStyle::BBNSS_Never
:
5661 case FormatStyle::BBNSS_Always
:
5663 case FormatStyle::BBNSS_OnlyWithParen
:
5664 return Right
.Next
&& Right
.Next
->is(tok::l_paren
);
5668 return Left
.isOneOf(tok::comma
, tok::coloncolon
, tok::semi
, tok::l_brace
,
5669 tok::kw_class
, tok::kw_struct
, tok::comment
) ||
5670 Right
.isMemberAccess() ||
5671 Right
.isOneOf(TT_TrailingReturnArrow
, TT_LambdaArrow
, tok::lessless
,
5672 tok::colon
, tok::l_square
, tok::at
) ||
5673 (Left
.is(tok::r_paren
) &&
5674 Right
.isOneOf(tok::identifier
, tok::kw_const
)) ||
5675 (Left
.is(tok::l_paren
) && Right
.isNot(tok::r_paren
)) ||
5676 (Left
.is(TT_TemplateOpener
) && Right
.isNot(TT_TemplateCloser
));
5679 void TokenAnnotator::printDebugInfo(const AnnotatedLine
&Line
) const {
5680 llvm::errs() << "AnnotatedTokens(L=" << Line
.Level
<< ", P=" << Line
.PPLevel
5681 << ", T=" << Line
.Type
<< ", C=" << Line
.IsContinuation
5683 const FormatToken
*Tok
= Line
.First
;
5685 llvm::errs() << " M=" << Tok
->MustBreakBefore
5686 << " C=" << Tok
->CanBreakBefore
5687 << " T=" << getTokenTypeName(Tok
->getType())
5688 << " S=" << Tok
->SpacesRequiredBefore
5689 << " F=" << Tok
->Finalized
<< " B=" << Tok
->BlockParameterCount
5690 << " BK=" << Tok
->getBlockKind() << " P=" << Tok
->SplitPenalty
5691 << " Name=" << Tok
->Tok
.getName() << " L=" << Tok
->TotalLength
5692 << " PPK=" << Tok
->getPackingKind() << " FakeLParens=";
5693 for (prec::Level LParen
: Tok
->FakeLParens
)
5694 llvm::errs() << LParen
<< "/";
5695 llvm::errs() << " FakeRParens=" << Tok
->FakeRParens
;
5696 llvm::errs() << " II=" << Tok
->Tok
.getIdentifierInfo();
5697 llvm::errs() << " Text='" << Tok
->TokenText
<< "'\n";
5699 assert(Tok
== Line
.Last
);
5702 llvm::errs() << "----\n";
5705 FormatStyle::PointerAlignmentStyle
5706 TokenAnnotator::getTokenReferenceAlignment(const FormatToken
&Reference
) const {
5707 assert(Reference
.isOneOf(tok::amp
, tok::ampamp
));
5708 switch (Style
.ReferenceAlignment
) {
5709 case FormatStyle::RAS_Pointer
:
5710 return Style
.PointerAlignment
;
5711 case FormatStyle::RAS_Left
:
5712 return FormatStyle::PAS_Left
;
5713 case FormatStyle::RAS_Right
:
5714 return FormatStyle::PAS_Right
;
5715 case FormatStyle::RAS_Middle
:
5716 return FormatStyle::PAS_Middle
;
5718 assert(0); //"Unhandled value of ReferenceAlignment"
5719 return Style
.PointerAlignment
;
5722 FormatStyle::PointerAlignmentStyle
5723 TokenAnnotator::getTokenPointerOrReferenceAlignment(
5724 const FormatToken
&PointerOrReference
) const {
5725 if (PointerOrReference
.isOneOf(tok::amp
, tok::ampamp
)) {
5726 switch (Style
.ReferenceAlignment
) {
5727 case FormatStyle::RAS_Pointer
:
5728 return Style
.PointerAlignment
;
5729 case FormatStyle::RAS_Left
:
5730 return FormatStyle::PAS_Left
;
5731 case FormatStyle::RAS_Right
:
5732 return FormatStyle::PAS_Right
;
5733 case FormatStyle::RAS_Middle
:
5734 return FormatStyle::PAS_Middle
;
5737 assert(PointerOrReference
.is(tok::star
));
5738 return Style
.PointerAlignment
;
5741 } // namespace format
5742 } // namespace clang