clang  7.0.0svn
TokenAnnotator.cpp
Go to the documentation of this file.
1 //===--- TokenAnnotator.cpp - Format C++ code -----------------------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 ///
10 /// \file
11 /// This file implements a token annotator, i.e. creates
12 /// \c AnnotatedTokens out of \c FormatTokens with required extra information.
13 ///
14 //===----------------------------------------------------------------------===//
15 
16 #include "TokenAnnotator.h"
18 #include "llvm/ADT/SmallPtrSet.h"
19 #include "llvm/Support/Debug.h"
20 
21 #define DEBUG_TYPE "format-token-annotator"
22 
23 namespace clang {
24 namespace format {
25 
26 namespace {
27 
28 /// Returns \c true if the token can be used as an identifier in
29 /// an Objective-C \c @selector, \c false otherwise.
30 ///
31 /// Because getFormattingLangOpts() always lexes source code as
32 /// Objective-C++, C++ keywords like \c new and \c delete are
33 /// lexed as tok::kw_*, not tok::identifier, even for Objective-C.
34 ///
35 /// For Objective-C and Objective-C++, both identifiers and keywords
36 /// are valid inside @selector(...) (or a macro which
37 /// invokes @selector(...)). So, we allow treat any identifier or
38 /// keyword as a potential Objective-C selector component.
39 static bool canBeObjCSelectorComponent(const FormatToken &Tok) {
40  return Tok.Tok.getIdentifierInfo() != nullptr;
41 }
42 
43 /// A parser that gathers additional information about tokens.
44 ///
45 /// The \c TokenAnnotator tries to match parenthesis and square brakets and
46 /// store a parenthesis levels. It also tries to resolve matching "<" and ">"
47 /// into template parameter lists.
48 class AnnotatingParser {
49 public:
50  AnnotatingParser(const FormatStyle &Style, AnnotatedLine &Line,
51  const AdditionalKeywords &Keywords)
52  : Style(Style), Line(Line), CurrentToken(Line.First), AutoFound(false),
53  Keywords(Keywords) {
54  Contexts.push_back(Context(tok::unknown, 1, /*IsExpression=*/false));
55  resetTokenMetadata(CurrentToken);
56  }
57 
58 private:
59  bool parseAngle() {
60  if (!CurrentToken || !CurrentToken->Previous)
61  return false;
62  if (NonTemplateLess.count(CurrentToken->Previous))
63  return false;
64 
65  const FormatToken &Previous = *CurrentToken->Previous; // The '<'.
66  if (Previous.Previous) {
67  if (Previous.Previous->Tok.isLiteral())
68  return false;
69  if (Previous.Previous->is(tok::r_paren) && Contexts.size() > 1 &&
70  (!Previous.Previous->MatchingParen ||
71  !Previous.Previous->MatchingParen->is(TT_OverloadedOperatorLParen)))
72  return false;
73  }
74 
75  FormatToken *Left = CurrentToken->Previous;
76  Left->ParentBracket = Contexts.back().ContextKind;
77  ScopedContextCreator ContextCreator(*this, tok::less, 12);
78 
79  // If this angle is in the context of an expression, we need to be more
80  // hesitant to detect it as opening template parameters.
81  bool InExprContext = Contexts.back().IsExpression;
82 
83  Contexts.back().IsExpression = false;
84  // If there's a template keyword before the opening angle bracket, this is a
85  // template parameter, not an argument.
86  Contexts.back().InTemplateArgument =
87  Left->Previous && Left->Previous->Tok.isNot(tok::kw_template);
88 
89  if (Style.Language == FormatStyle::LK_Java &&
90  CurrentToken->is(tok::question))
91  next();
92 
93  while (CurrentToken) {
94  if (CurrentToken->is(tok::greater)) {
95  Left->MatchingParen = CurrentToken;
96  CurrentToken->MatchingParen = Left;
97  // In TT_Proto, we must distignuish between:
98  // map<key, value>
99  // msg < item: data >
100  // msg: < item: data >
101  // In TT_TextProto, map<key, value> does not occur.
102  if (Style.Language == FormatStyle::LK_TextProto ||
103  (Style.Language == FormatStyle::LK_Proto && Left->Previous &&
104  Left->Previous->isOneOf(TT_SelectorName, TT_DictLiteral)))
105  CurrentToken->Type = TT_DictLiteral;
106  else
107  CurrentToken->Type = TT_TemplateCloser;
108  next();
109  return true;
110  }
111  if (CurrentToken->is(tok::question) &&
112  Style.Language == FormatStyle::LK_Java) {
113  next();
114  continue;
115  }
116  if (CurrentToken->isOneOf(tok::r_paren, tok::r_square, tok::r_brace) ||
117  (CurrentToken->isOneOf(tok::colon, tok::question) && InExprContext &&
118  Style.Language != FormatStyle::LK_Proto &&
119  Style.Language != FormatStyle::LK_TextProto))
120  return false;
121  // If a && or || is found and interpreted as a binary operator, this set
122  // of angles is likely part of something like "a < b && c > d". If the
123  // angles are inside an expression, the ||/&& might also be a binary
124  // operator that was misinterpreted because we are parsing template
125  // parameters.
126  // FIXME: This is getting out of hand, write a decent parser.
127  if (CurrentToken->Previous->isOneOf(tok::pipepipe, tok::ampamp) &&
128  CurrentToken->Previous->is(TT_BinaryOperator) &&
129  Contexts[Contexts.size() - 2].IsExpression &&
130  !Line.startsWith(tok::kw_template))
131  return false;
132  updateParameterCount(Left, CurrentToken);
133  if (Style.Language == FormatStyle::LK_Proto) {
134  if (FormatToken *Previous = CurrentToken->getPreviousNonComment()) {
135  if (CurrentToken->is(tok::colon) ||
136  (CurrentToken->isOneOf(tok::l_brace, tok::less) &&
137  Previous->isNot(tok::colon)))
138  Previous->Type = TT_SelectorName;
139  }
140  }
141  if (!consumeToken())
142  return false;
143  }
144  return false;
145  }
146 
147  bool parseParens(bool LookForDecls = false) {
148  if (!CurrentToken)
149  return false;
150  FormatToken *Left = CurrentToken->Previous;
151  Left->ParentBracket = Contexts.back().ContextKind;
152  ScopedContextCreator ContextCreator(*this, tok::l_paren, 1);
153 
154  // FIXME: This is a bit of a hack. Do better.
155  Contexts.back().ColonIsForRangeExpr =
156  Contexts.size() == 2 && Contexts[0].ColonIsForRangeExpr;
157 
158  bool StartsObjCMethodExpr = false;
159  if (FormatToken *MaybeSel = Left->Previous) {
160  // @selector( starts a selector.
161  if (MaybeSel->isObjCAtKeyword(tok::objc_selector) && MaybeSel->Previous &&
162  MaybeSel->Previous->is(tok::at)) {
163  StartsObjCMethodExpr = true;
164  }
165  }
166 
167  if (Left->is(TT_OverloadedOperatorLParen)) {
168  Contexts.back().IsExpression = false;
169  } else if (Style.Language == FormatStyle::LK_JavaScript &&
170  (Line.startsWith(Keywords.kw_type, tok::identifier) ||
171  Line.startsWith(tok::kw_export, Keywords.kw_type,
172  tok::identifier))) {
173  // type X = (...);
174  // export type X = (...);
175  Contexts.back().IsExpression = false;
176  } else if (Left->Previous &&
177  (Left->Previous->isOneOf(tok::kw_static_assert, tok::kw_decltype,
178  tok::kw_if, tok::kw_while, tok::l_paren,
179  tok::comma) ||
180  Left->Previous->endsSequence(tok::kw_constexpr, tok::kw_if) ||
181  Left->Previous->is(TT_BinaryOperator))) {
182  // static_assert, if and while usually contain expressions.
183  Contexts.back().IsExpression = true;
184  } else if (Style.Language == FormatStyle::LK_JavaScript && Left->Previous &&
185  (Left->Previous->is(Keywords.kw_function) ||
186  (Left->Previous->endsSequence(tok::identifier,
187  Keywords.kw_function)))) {
188  // function(...) or function f(...)
189  Contexts.back().IsExpression = false;
190  } else if (Style.Language == FormatStyle::LK_JavaScript && Left->Previous &&
191  Left->Previous->is(TT_JsTypeColon)) {
192  // let x: (SomeType);
193  Contexts.back().IsExpression = false;
194  } else if (Left->Previous && Left->Previous->is(tok::r_square) &&
195  Left->Previous->MatchingParen &&
196  Left->Previous->MatchingParen->is(TT_LambdaLSquare)) {
197  // This is a parameter list of a lambda expression.
198  Contexts.back().IsExpression = false;
199  } else if (Line.InPPDirective &&
200  (!Left->Previous || !Left->Previous->is(tok::identifier))) {
201  Contexts.back().IsExpression = true;
202  } else if (Contexts[Contexts.size() - 2].CaretFound) {
203  // This is the parameter list of an ObjC block.
204  Contexts.back().IsExpression = false;
205  } else if (Left->Previous && Left->Previous->is(tok::kw___attribute)) {
206  Left->Type = TT_AttributeParen;
207  } else if (Left->Previous && Left->Previous->is(TT_ForEachMacro)) {
208  // The first argument to a foreach macro is a declaration.
209  Contexts.back().IsForEachMacro = true;
210  Contexts.back().IsExpression = false;
211  } else if (Left->Previous && Left->Previous->MatchingParen &&
212  Left->Previous->MatchingParen->is(TT_ObjCBlockLParen)) {
213  Contexts.back().IsExpression = false;
214  } else if (!Line.MustBeDeclaration && !Line.InPPDirective) {
215  bool IsForOrCatch =
216  Left->Previous && Left->Previous->isOneOf(tok::kw_for, tok::kw_catch);
217  Contexts.back().IsExpression = !IsForOrCatch;
218  }
219 
220  if (StartsObjCMethodExpr) {
221  Contexts.back().ColonIsObjCMethodExpr = true;
222  Left->Type = TT_ObjCMethodExpr;
223  }
224 
225  // MightBeFunctionType and ProbablyFunctionType are used for
226  // function pointer and reference types as well as Objective-C
227  // block types:
228  //
229  // void (*FunctionPointer)(void);
230  // void (&FunctionReference)(void);
231  // void (^ObjCBlock)(void);
232  bool MightBeFunctionType = !Contexts[Contexts.size() - 2].IsExpression;
233  bool ProbablyFunctionType =
234  CurrentToken->isOneOf(tok::star, tok::amp, tok::caret);
235  bool HasMultipleLines = false;
236  bool HasMultipleParametersOnALine = false;
237  bool MightBeObjCForRangeLoop =
238  Left->Previous && Left->Previous->is(tok::kw_for);
239  FormatToken *PossibleObjCForInToken = nullptr;
240  while (CurrentToken) {
241  // LookForDecls is set when "if (" has been seen. Check for
242  // 'identifier' '*' 'identifier' followed by not '=' -- this
243  // '*' has to be a binary operator but determineStarAmpUsage() will
244  // categorize it as an unary operator, so set the right type here.
245  if (LookForDecls && CurrentToken->Next) {
246  FormatToken *Prev = CurrentToken->getPreviousNonComment();
247  if (Prev) {
248  FormatToken *PrevPrev = Prev->getPreviousNonComment();
249  FormatToken *Next = CurrentToken->Next;
250  if (PrevPrev && PrevPrev->is(tok::identifier) &&
251  Prev->isOneOf(tok::star, tok::amp, tok::ampamp) &&
252  CurrentToken->is(tok::identifier) && Next->isNot(tok::equal)) {
253  Prev->Type = TT_BinaryOperator;
254  LookForDecls = false;
255  }
256  }
257  }
258 
259  if (CurrentToken->Previous->is(TT_PointerOrReference) &&
260  CurrentToken->Previous->Previous->isOneOf(tok::l_paren,
261  tok::coloncolon))
262  ProbablyFunctionType = true;
263  if (CurrentToken->is(tok::comma))
264  MightBeFunctionType = false;
265  if (CurrentToken->Previous->is(TT_BinaryOperator))
266  Contexts.back().IsExpression = true;
267  if (CurrentToken->is(tok::r_paren)) {
268  if (MightBeFunctionType && ProbablyFunctionType && CurrentToken->Next &&
269  (CurrentToken->Next->is(tok::l_paren) ||
270  (CurrentToken->Next->is(tok::l_square) && Line.MustBeDeclaration)))
271  Left->Type = Left->Next->is(tok::caret) ? TT_ObjCBlockLParen
272  : TT_FunctionTypeLParen;
273  Left->MatchingParen = CurrentToken;
274  CurrentToken->MatchingParen = Left;
275 
276  if (CurrentToken->Next && CurrentToken->Next->is(tok::l_brace) &&
277  Left->Previous && Left->Previous->is(tok::l_paren)) {
278  // Detect the case where macros are used to generate lambdas or
279  // function bodies, e.g.:
280  // auto my_lambda = MARCO((Type *type, int i) { .. body .. });
281  for (FormatToken *Tok = Left; Tok != CurrentToken; Tok = Tok->Next) {
282  if (Tok->is(TT_BinaryOperator) &&
283  Tok->isOneOf(tok::star, tok::amp, tok::ampamp))
284  Tok->Type = TT_PointerOrReference;
285  }
286  }
287 
288  if (StartsObjCMethodExpr) {
289  CurrentToken->Type = TT_ObjCMethodExpr;
290  if (Contexts.back().FirstObjCSelectorName) {
291  Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName =
292  Contexts.back().LongestObjCSelectorName;
293  }
294  }
295 
296  if (Left->is(TT_AttributeParen))
297  CurrentToken->Type = TT_AttributeParen;
298  if (Left->Previous && Left->Previous->is(TT_JavaAnnotation))
299  CurrentToken->Type = TT_JavaAnnotation;
300  if (Left->Previous && Left->Previous->is(TT_LeadingJavaAnnotation))
301  CurrentToken->Type = TT_LeadingJavaAnnotation;
302 
303  if (!HasMultipleLines)
304  Left->PackingKind = PPK_Inconclusive;
305  else if (HasMultipleParametersOnALine)
306  Left->PackingKind = PPK_BinPacked;
307  else
308  Left->PackingKind = PPK_OnePerLine;
309 
310  next();
311  return true;
312  }
313  if (CurrentToken->isOneOf(tok::r_square, tok::r_brace))
314  return false;
315 
316  if (CurrentToken->is(tok::l_brace))
317  Left->Type = TT_Unknown; // Not TT_ObjCBlockLParen
318  if (CurrentToken->is(tok::comma) && CurrentToken->Next &&
319  !CurrentToken->Next->HasUnescapedNewline &&
320  !CurrentToken->Next->isTrailingComment())
321  HasMultipleParametersOnALine = true;
322  if ((CurrentToken->Previous->isOneOf(tok::kw_const, tok::kw_auto) ||
323  CurrentToken->Previous->isSimpleTypeSpecifier()) &&
324  !CurrentToken->is(tok::l_brace))
325  Contexts.back().IsExpression = false;
326  if (CurrentToken->isOneOf(tok::semi, tok::colon)) {
327  MightBeObjCForRangeLoop = false;
328  if (PossibleObjCForInToken) {
329  PossibleObjCForInToken->Type = TT_Unknown;
330  PossibleObjCForInToken = nullptr;
331  }
332  }
333  if (MightBeObjCForRangeLoop && CurrentToken->is(Keywords.kw_in)) {
334  PossibleObjCForInToken = CurrentToken;
335  PossibleObjCForInToken->Type = TT_ObjCForIn;
336  }
337  // When we discover a 'new', we set CanBeExpression to 'false' in order to
338  // parse the type correctly. Reset that after a comma.
339  if (CurrentToken->is(tok::comma))
340  Contexts.back().CanBeExpression = true;
341 
342  FormatToken *Tok = CurrentToken;
343  if (!consumeToken())
344  return false;
345  updateParameterCount(Left, Tok);
346  if (CurrentToken && CurrentToken->HasUnescapedNewline)
347  HasMultipleLines = true;
348  }
349  return false;
350  }
351 
352  bool isCpp11AttributeSpecifier(const FormatToken &Tok) {
353  if (!Style.isCpp() || !Tok.startsSequence(tok::l_square, tok::l_square))
354  return false;
355  const FormatToken *AttrTok = Tok.Next->Next;
356  if (!AttrTok)
357  return false;
358  // C++17 '[[using ns: foo, bar(baz, blech)]]'
359  // We assume nobody will name an ObjC variable 'using'.
360  if (AttrTok->startsSequence(tok::kw_using, tok::identifier, tok::colon))
361  return true;
362  if (AttrTok->isNot(tok::identifier))
363  return false;
364  while (AttrTok && !AttrTok->startsSequence(tok::r_square, tok::r_square)) {
365  // ObjC message send. We assume nobody will use : in a C++11 attribute
366  // specifier parameter, although this is technically valid:
367  // [[foo(:)]]
368  if (AttrTok->is(tok::colon) ||
369  AttrTok->startsSequence(tok::identifier, tok::identifier))
370  return false;
371  if (AttrTok->is(tok::ellipsis))
372  return true;
373  AttrTok = AttrTok->Next;
374  }
375  return AttrTok && AttrTok->startsSequence(tok::r_square, tok::r_square);
376  }
377 
378  bool parseSquare() {
379  if (!CurrentToken)
380  return false;
381 
382  // A '[' could be an index subscript (after an identifier or after
383  // ')' or ']'), it could be the start of an Objective-C method
384  // expression, it could the start of an Objective-C array literal,
385  // or it could be a C++ attribute specifier [[foo::bar]].
386  FormatToken *Left = CurrentToken->Previous;
387  Left->ParentBracket = Contexts.back().ContextKind;
388  FormatToken *Parent = Left->getPreviousNonComment();
389 
390  // Cases where '>' is followed by '['.
391  // In C++, this can happen either in array of templates (foo<int>[10])
392  // or when array is a nested template type (unique_ptr<type1<type2>[]>).
393  bool CppArrayTemplates =
394  Style.isCpp() && Parent && Parent->is(TT_TemplateCloser) &&
395  (Contexts.back().CanBeExpression || Contexts.back().IsExpression ||
396  Contexts.back().InTemplateArgument);
397 
398  bool IsCpp11AttributeSpecifier = isCpp11AttributeSpecifier(*Left) ||
399  Contexts.back().InCpp11AttributeSpecifier;
400 
401  bool StartsObjCMethodExpr =
402  !CppArrayTemplates && Style.isCpp() && !IsCpp11AttributeSpecifier &&
403  Contexts.back().CanBeExpression && Left->isNot(TT_LambdaLSquare) &&
404  !CurrentToken->isOneOf(tok::l_brace, tok::r_square) &&
405  (!Parent ||
406  Parent->isOneOf(tok::colon, tok::l_square, tok::l_paren,
407  tok::kw_return, tok::kw_throw) ||
408  Parent->isUnaryOperator() ||
409  // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen.
410  Parent->isOneOf(TT_ObjCForIn, TT_CastRParen) ||
411  getBinOpPrecedence(Parent->Tok.getKind(), true, true) > prec::Unknown);
412  bool ColonFound = false;
413 
414  unsigned BindingIncrease = 1;
415  if (Left->isCppStructuredBinding(Style)) {
416  Left->Type = TT_StructuredBindingLSquare;
417  } else if (Left->is(TT_Unknown)) {
418  if (StartsObjCMethodExpr) {
419  Left->Type = TT_ObjCMethodExpr;
420  } else if (IsCpp11AttributeSpecifier) {
421  Left->Type = TT_AttributeSquare;
422  } else if (Style.Language == FormatStyle::LK_JavaScript && Parent &&
423  Contexts.back().ContextKind == tok::l_brace &&
424  Parent->isOneOf(tok::l_brace, tok::comma)) {
425  Left->Type = TT_JsComputedPropertyName;
426  } else if (Style.isCpp() && Contexts.back().ContextKind == tok::l_brace &&
427  Parent && Parent->isOneOf(tok::l_brace, tok::comma)) {
428  Left->Type = TT_DesignatedInitializerLSquare;
429  } else if (CurrentToken->is(tok::r_square) && Parent &&
430  Parent->is(TT_TemplateCloser)) {
431  Left->Type = TT_ArraySubscriptLSquare;
432  } else if (Style.Language == FormatStyle::LK_Proto ||
433  Style.Language == FormatStyle::LK_TextProto) {
434  // Square braces in LK_Proto can either be message field attributes:
435  //
436  // optional Aaa aaa = 1 [
437  // (aaa) = aaa
438  // ];
439  //
440  // extensions 123 [
441  // (aaa) = aaa
442  // ];
443  //
444  // or text proto extensions (in options):
445  //
446  // option (Aaa.options) = {
447  // [type.type/type] {
448  // key: value
449  // }
450  // }
451  //
452  // or repeated fields (in options):
453  //
454  // option (Aaa.options) = {
455  // keys: [ 1, 2, 3 ]
456  // }
457  //
458  // In the first and the third case we want to spread the contents inside
459  // the square braces; in the second we want to keep them inline.
460  Left->Type = TT_ArrayInitializerLSquare;
461  if (!Left->endsSequence(tok::l_square, tok::numeric_constant,
462  tok::equal) &&
463  !Left->endsSequence(tok::l_square, tok::numeric_constant,
464  tok::identifier) &&
465  !Left->endsSequence(tok::l_square, tok::colon, TT_SelectorName)) {
466  Left->Type = TT_ProtoExtensionLSquare;
467  BindingIncrease = 10;
468  }
469  } else if (!CppArrayTemplates && Parent &&
470  Parent->isOneOf(TT_BinaryOperator, TT_TemplateCloser, tok::at,
471  tok::comma, tok::l_paren, tok::l_square,
472  tok::question, tok::colon, tok::kw_return,
473  // Should only be relevant to JavaScript:
474  tok::kw_default)) {
475  Left->Type = TT_ArrayInitializerLSquare;
476  } else {
477  BindingIncrease = 10;
478  Left->Type = TT_ArraySubscriptLSquare;
479  }
480  }
481 
482  ScopedContextCreator ContextCreator(*this, tok::l_square, BindingIncrease);
483  Contexts.back().IsExpression = true;
484  if (Style.Language == FormatStyle::LK_JavaScript && Parent &&
485  Parent->is(TT_JsTypeColon))
486  Contexts.back().IsExpression = false;
487 
488  Contexts.back().ColonIsObjCMethodExpr = StartsObjCMethodExpr;
489  Contexts.back().InCpp11AttributeSpecifier = IsCpp11AttributeSpecifier;
490 
491  while (CurrentToken) {
492  if (CurrentToken->is(tok::r_square)) {
493  if (IsCpp11AttributeSpecifier)
494  CurrentToken->Type = TT_AttributeSquare;
495  else if (CurrentToken->Next && CurrentToken->Next->is(tok::l_paren) &&
496  Left->is(TT_ObjCMethodExpr)) {
497  // An ObjC method call is rarely followed by an open parenthesis.
498  // FIXME: Do we incorrectly label ":" with this?
499  StartsObjCMethodExpr = false;
500  Left->Type = TT_Unknown;
501  }
502  if (StartsObjCMethodExpr && CurrentToken->Previous != Left) {
503  CurrentToken->Type = TT_ObjCMethodExpr;
504  // If we haven't seen a colon yet, make sure the last identifier
505  // before the r_square is tagged as a selector name component.
506  if (!ColonFound && CurrentToken->Previous &&
507  CurrentToken->Previous->is(TT_Unknown) &&
508  canBeObjCSelectorComponent(*CurrentToken->Previous))
509  CurrentToken->Previous->Type = TT_SelectorName;
510  // determineStarAmpUsage() thinks that '*' '[' is allocating an
511  // array of pointers, but if '[' starts a selector then '*' is a
512  // binary operator.
513  if (Parent && Parent->is(TT_PointerOrReference))
514  Parent->Type = TT_BinaryOperator;
515  }
516  Left->MatchingParen = CurrentToken;
517  CurrentToken->MatchingParen = Left;
518  if (Contexts.back().FirstObjCSelectorName) {
519  Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName =
520  Contexts.back().LongestObjCSelectorName;
521  Contexts.back().FirstObjCSelectorName->ObjCSelectorNameParts =
522  Left->ParameterCount;
523  if (Left->BlockParameterCount > 1)
524  Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = 0;
525  }
526  next();
527  return true;
528  }
529  if (CurrentToken->isOneOf(tok::r_paren, tok::r_brace))
530  return false;
531  if (CurrentToken->is(tok::colon)) {
532  if (IsCpp11AttributeSpecifier &&
533  CurrentToken->endsSequence(tok::colon, tok::identifier,
534  tok::kw_using)) {
535  // Remember that this is a [[using ns: foo]] C++ attribute, so we
536  // don't add a space before the colon (unlike other colons).
537  CurrentToken->Type = TT_AttributeColon;
538  } else if (Left->isOneOf(TT_ArraySubscriptLSquare,
539  TT_DesignatedInitializerLSquare)) {
540  Left->Type = TT_ObjCMethodExpr;
541  StartsObjCMethodExpr = true;
542  // ParameterCount might have been set to 1 before expression was
543  // recognized as ObjCMethodExpr (as '1 + number of commas' formula is
544  // used for other expression types). Parameter counter has to be,
545  // therefore, reset to 0.
546  Left->ParameterCount = 0;
547  Contexts.back().ColonIsObjCMethodExpr = true;
548  if (Parent && Parent->is(tok::r_paren))
549  // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen.
550  Parent->Type = TT_CastRParen;
551  }
552  ColonFound = true;
553  }
554  if (CurrentToken->is(tok::comma) && Left->is(TT_ObjCMethodExpr) &&
555  !ColonFound)
556  Left->Type = TT_ArrayInitializerLSquare;
557  FormatToken *Tok = CurrentToken;
558  if (!consumeToken())
559  return false;
560  updateParameterCount(Left, Tok);
561  }
562  return false;
563  }
564 
565  bool parseBrace() {
566  if (CurrentToken) {
567  FormatToken *Left = CurrentToken->Previous;
568  Left->ParentBracket = Contexts.back().ContextKind;
569 
570  if (Contexts.back().CaretFound)
571  Left->Type = TT_ObjCBlockLBrace;
572  Contexts.back().CaretFound = false;
573 
574  ScopedContextCreator ContextCreator(*this, tok::l_brace, 1);
575  Contexts.back().ColonIsDictLiteral = true;
576  if (Left->BlockKind == BK_BracedInit)
577  Contexts.back().IsExpression = true;
578  if (Style.Language == FormatStyle::LK_JavaScript && Left->Previous &&
579  Left->Previous->is(TT_JsTypeColon))
580  Contexts.back().IsExpression = false;
581 
582  while (CurrentToken) {
583  if (CurrentToken->is(tok::r_brace)) {
584  Left->MatchingParen = CurrentToken;
585  CurrentToken->MatchingParen = Left;
586  next();
587  return true;
588  }
589  if (CurrentToken->isOneOf(tok::r_paren, tok::r_square))
590  return false;
591  updateParameterCount(Left, CurrentToken);
592  if (CurrentToken->isOneOf(tok::colon, tok::l_brace, tok::less)) {
593  FormatToken *Previous = CurrentToken->getPreviousNonComment();
594  if (Previous->is(TT_JsTypeOptionalQuestion))
595  Previous = Previous->getPreviousNonComment();
596  if ((CurrentToken->is(tok::colon) &&
597  (!Contexts.back().ColonIsDictLiteral || !Style.isCpp())) ||
598  Style.Language == FormatStyle::LK_Proto ||
599  Style.Language == FormatStyle::LK_TextProto) {
600  Left->Type = TT_DictLiteral;
601  if (Previous->Tok.getIdentifierInfo() ||
602  Previous->is(tok::string_literal))
603  Previous->Type = TT_SelectorName;
604  }
605  if (CurrentToken->is(tok::colon) ||
606  Style.Language == FormatStyle::LK_JavaScript)
607  Left->Type = TT_DictLiteral;
608  }
609  if (CurrentToken->is(tok::comma) &&
610  Style.Language == FormatStyle::LK_JavaScript)
611  Left->Type = TT_DictLiteral;
612  if (!consumeToken())
613  return false;
614  }
615  }
616  return true;
617  }
618 
619  void updateParameterCount(FormatToken *Left, FormatToken *Current) {
620  if (Current->is(tok::l_brace) && Current->BlockKind == BK_Block)
621  ++Left->BlockParameterCount;
622  if (Left->Type == TT_ObjCMethodExpr) {
623  if (Current->is(tok::colon))
624  ++Left->ParameterCount;
625  } else if (Current->is(tok::comma)) {
626  ++Left->ParameterCount;
627  if (!Left->Role)
628  Left->Role.reset(new CommaSeparatedList(Style));
629  Left->Role->CommaFound(Current);
630  } else if (Left->ParameterCount == 0 && Current->isNot(tok::comment)) {
631  Left->ParameterCount = 1;
632  }
633  }
634 
635  bool parseConditional() {
636  while (CurrentToken) {
637  if (CurrentToken->is(tok::colon)) {
638  CurrentToken->Type = TT_ConditionalExpr;
639  next();
640  return true;
641  }
642  if (!consumeToken())
643  return false;
644  }
645  return false;
646  }
647 
648  bool parseTemplateDeclaration() {
649  if (CurrentToken && CurrentToken->is(tok::less)) {
650  CurrentToken->Type = TT_TemplateOpener;
651  next();
652  if (!parseAngle())
653  return false;
654  if (CurrentToken)
655  CurrentToken->Previous->ClosesTemplateDeclaration = true;
656  return true;
657  }
658  return false;
659  }
660 
661  bool consumeToken() {
662  FormatToken *Tok = CurrentToken;
663  next();
664  switch (Tok->Tok.getKind()) {
665  case tok::plus:
666  case tok::minus:
667  if (!Tok->Previous && Line.MustBeDeclaration)
668  Tok->Type = TT_ObjCMethodSpecifier;
669  break;
670  case tok::colon:
671  if (!Tok->Previous)
672  return false;
673  // Colons from ?: are handled in parseConditional().
674  if (Style.Language == FormatStyle::LK_JavaScript) {
675  if (Contexts.back().ColonIsForRangeExpr || // colon in for loop
676  (Contexts.size() == 1 && // switch/case labels
677  !Line.First->isOneOf(tok::kw_enum, tok::kw_case)) ||
678  Contexts.back().ContextKind == tok::l_paren || // function params
679  Contexts.back().ContextKind == tok::l_square || // array type
680  (!Contexts.back().IsExpression &&
681  Contexts.back().ContextKind == tok::l_brace) || // object type
682  (Contexts.size() == 1 &&
683  Line.MustBeDeclaration)) { // method/property declaration
684  Contexts.back().IsExpression = false;
685  Tok->Type = TT_JsTypeColon;
686  break;
687  }
688  }
689  if (Contexts.back().ColonIsDictLiteral ||
690  Style.Language == FormatStyle::LK_Proto ||
691  Style.Language == FormatStyle::LK_TextProto) {
692  Tok->Type = TT_DictLiteral;
693  if (Style.Language == FormatStyle::LK_TextProto) {
694  if (FormatToken *Previous = Tok->getPreviousNonComment())
695  Previous->Type = TT_SelectorName;
696  }
697  } else if (Contexts.back().ColonIsObjCMethodExpr ||
698  Line.startsWith(TT_ObjCMethodSpecifier)) {
699  Tok->Type = TT_ObjCMethodExpr;
700  const FormatToken *BeforePrevious = Tok->Previous->Previous;
701  if (!BeforePrevious ||
702  // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen.
703  !(BeforePrevious->is(TT_CastRParen) ||
704  (BeforePrevious->is(TT_ObjCMethodExpr) &&
705  BeforePrevious->is(tok::colon))) ||
706  BeforePrevious->is(tok::r_square) ||
707  Contexts.back().LongestObjCSelectorName == 0) {
708  Tok->Previous->Type = TT_SelectorName;
709  if (!Contexts.back().FirstObjCSelectorName)
710  Contexts.back().FirstObjCSelectorName = Tok->Previous;
711  else if (Tok->Previous->ColumnWidth >
712  Contexts.back().LongestObjCSelectorName)
713  Contexts.back().LongestObjCSelectorName =
714  Tok->Previous->ColumnWidth;
715  }
716  } else if (Contexts.back().ColonIsForRangeExpr) {
717  Tok->Type = TT_RangeBasedForLoopColon;
718  } else if (CurrentToken && CurrentToken->is(tok::numeric_constant)) {
719  Tok->Type = TT_BitFieldColon;
720  } else if (Contexts.size() == 1 &&
721  !Line.First->isOneOf(tok::kw_enum, tok::kw_case)) {
722  if (Tok->getPreviousNonComment()->isOneOf(tok::r_paren,
723  tok::kw_noexcept))
724  Tok->Type = TT_CtorInitializerColon;
725  else
726  Tok->Type = TT_InheritanceColon;
727  } else if (canBeObjCSelectorComponent(*Tok->Previous) && Tok->Next &&
728  (Tok->Next->isOneOf(tok::r_paren, tok::comma) ||
729  (canBeObjCSelectorComponent(*Tok->Next) && Tok->Next->Next &&
730  Tok->Next->Next->is(tok::colon)))) {
731  // This handles a special macro in ObjC code where selectors including
732  // the colon are passed as macro arguments.
733  Tok->Type = TT_ObjCMethodExpr;
734  } else if (Contexts.back().ContextKind == tok::l_paren) {
735  Tok->Type = TT_InlineASMColon;
736  }
737  break;
738  case tok::pipe:
739  case tok::amp:
740  // | and & in declarations/type expressions represent union and
741  // intersection types, respectively.
742  if (Style.Language == FormatStyle::LK_JavaScript &&
743  !Contexts.back().IsExpression)
744  Tok->Type = TT_JsTypeOperator;
745  break;
746  case tok::kw_if:
747  case tok::kw_while:
748  if (Tok->is(tok::kw_if) && CurrentToken &&
749  CurrentToken->is(tok::kw_constexpr))
750  next();
751  if (CurrentToken && CurrentToken->is(tok::l_paren)) {
752  next();
753  if (!parseParens(/*LookForDecls=*/true))
754  return false;
755  }
756  break;
757  case tok::kw_for:
758  if (Style.Language == FormatStyle::LK_JavaScript) {
759  // x.for and {for: ...}
760  if ((Tok->Previous && Tok->Previous->is(tok::period)) ||
761  (Tok->Next && Tok->Next->is(tok::colon)))
762  break;
763  // JS' for await ( ...
764  if (CurrentToken && CurrentToken->is(Keywords.kw_await))
765  next();
766  }
767  Contexts.back().ColonIsForRangeExpr = true;
768  next();
769  if (!parseParens())
770  return false;
771  break;
772  case tok::l_paren:
773  // When faced with 'operator()()', the kw_operator handler incorrectly
774  // marks the first l_paren as a OverloadedOperatorLParen. Here, we make
775  // the first two parens OverloadedOperators and the second l_paren an
776  // OverloadedOperatorLParen.
777  if (Tok->Previous && Tok->Previous->is(tok::r_paren) &&
778  Tok->Previous->MatchingParen &&
779  Tok->Previous->MatchingParen->is(TT_OverloadedOperatorLParen)) {
780  Tok->Previous->Type = TT_OverloadedOperator;
781  Tok->Previous->MatchingParen->Type = TT_OverloadedOperator;
782  Tok->Type = TT_OverloadedOperatorLParen;
783  }
784 
785  if (!parseParens())
786  return false;
787  if (Line.MustBeDeclaration && Contexts.size() == 1 &&
788  !Contexts.back().IsExpression && !Line.startsWith(TT_ObjCProperty) &&
789  (!Tok->Previous ||
790  !Tok->Previous->isOneOf(tok::kw_decltype, tok::kw___attribute,
791  TT_LeadingJavaAnnotation)))
792  Line.MightBeFunctionDecl = true;
793  break;
794  case tok::l_square:
795  if (!parseSquare())
796  return false;
797  break;
798  case tok::l_brace:
799  if (Style.Language == FormatStyle::LK_TextProto) {
800  FormatToken *Previous = Tok->getPreviousNonComment();
801  if (Previous && Previous->Type != TT_DictLiteral)
802  Previous->Type = TT_SelectorName;
803  }
804  if (!parseBrace())
805  return false;
806  break;
807  case tok::less:
808  if (parseAngle()) {
809  Tok->Type = TT_TemplateOpener;
810  // In TT_Proto, we must distignuish between:
811  // map<key, value>
812  // msg < item: data >
813  // msg: < item: data >
814  // In TT_TextProto, map<key, value> does not occur.
815  if (Style.Language == FormatStyle::LK_TextProto ||
816  (Style.Language == FormatStyle::LK_Proto && Tok->Previous &&
817  Tok->Previous->isOneOf(TT_SelectorName, TT_DictLiteral))) {
818  Tok->Type = TT_DictLiteral;
819  FormatToken *Previous = Tok->getPreviousNonComment();
820  if (Previous && Previous->Type != TT_DictLiteral)
821  Previous->Type = TT_SelectorName;
822  }
823  } else {
824  Tok->Type = TT_BinaryOperator;
825  NonTemplateLess.insert(Tok);
826  CurrentToken = Tok;
827  next();
828  }
829  break;
830  case tok::r_paren:
831  case tok::r_square:
832  return false;
833  case tok::r_brace:
834  // Lines can start with '}'.
835  if (Tok->Previous)
836  return false;
837  break;
838  case tok::greater:
839  if (Style.Language != FormatStyle::LK_TextProto)
840  Tok->Type = TT_BinaryOperator;
841  break;
842  case tok::kw_operator:
843  if (Style.Language == FormatStyle::LK_TextProto ||
844  Style.Language == FormatStyle::LK_Proto)
845  break;
846  while (CurrentToken &&
847  !CurrentToken->isOneOf(tok::l_paren, tok::semi, tok::r_paren)) {
848  if (CurrentToken->isOneOf(tok::star, tok::amp))
849  CurrentToken->Type = TT_PointerOrReference;
850  consumeToken();
851  if (CurrentToken &&
852  CurrentToken->Previous->isOneOf(TT_BinaryOperator, TT_UnaryOperator,
853  tok::comma))
854  CurrentToken->Previous->Type = TT_OverloadedOperator;
855  }
856  if (CurrentToken) {
857  CurrentToken->Type = TT_OverloadedOperatorLParen;
858  if (CurrentToken->Previous->is(TT_BinaryOperator))
859  CurrentToken->Previous->Type = TT_OverloadedOperator;
860  }
861  break;
862  case tok::question:
863  if (Style.Language == FormatStyle::LK_JavaScript && Tok->Next &&
864  Tok->Next->isOneOf(tok::semi, tok::comma, tok::colon, tok::r_paren,
865  tok::r_brace)) {
866  // Question marks before semicolons, colons, etc. indicate optional
867  // types (fields, parameters), e.g.
868  // function(x?: string, y?) {...}
869  // class X { y?; }
870  Tok->Type = TT_JsTypeOptionalQuestion;
871  break;
872  }
873  // Declarations cannot be conditional expressions, this can only be part
874  // of a type declaration.
875  if (Line.MustBeDeclaration && !Contexts.back().IsExpression &&
876  Style.Language == FormatStyle::LK_JavaScript)
877  break;
878  parseConditional();
879  break;
880  case tok::kw_template:
881  parseTemplateDeclaration();
882  break;
883  case tok::comma:
884  if (Contexts.back().InCtorInitializer)
885  Tok->Type = TT_CtorInitializerComma;
886  else if (Contexts.back().InInheritanceList)
887  Tok->Type = TT_InheritanceComma;
888  else if (Contexts.back().FirstStartOfName &&
889  (Contexts.size() == 1 || Line.startsWith(tok::kw_for))) {
890  Contexts.back().FirstStartOfName->PartOfMultiVariableDeclStmt = true;
891  Line.IsMultiVariableDeclStmt = true;
892  }
893  if (Contexts.back().IsForEachMacro)
894  Contexts.back().IsExpression = true;
895  break;
896  case tok::identifier:
897  if (Tok->isOneOf(Keywords.kw___has_include,
898  Keywords.kw___has_include_next)) {
899  parseHasInclude();
900  }
901  break;
902  default:
903  break;
904  }
905  return true;
906  }
907 
908  void parseIncludeDirective() {
909  if (CurrentToken && CurrentToken->is(tok::less)) {
910  next();
911  while (CurrentToken) {
912  // Mark tokens up to the trailing line comments as implicit string
913  // literals.
914  if (CurrentToken->isNot(tok::comment) &&
915  !CurrentToken->TokenText.startswith("//"))
916  CurrentToken->Type = TT_ImplicitStringLiteral;
917  next();
918  }
919  }
920  }
921 
922  void parseWarningOrError() {
923  next();
924  // We still want to format the whitespace left of the first token of the
925  // warning or error.
926  next();
927  while (CurrentToken) {
928  CurrentToken->Type = TT_ImplicitStringLiteral;
929  next();
930  }
931  }
932 
933  void parsePragma() {
934  next(); // Consume "pragma".
935  if (CurrentToken &&
936  CurrentToken->isOneOf(Keywords.kw_mark, Keywords.kw_option)) {
937  bool IsMark = CurrentToken->is(Keywords.kw_mark);
938  next(); // Consume "mark".
939  next(); // Consume first token (so we fix leading whitespace).
940  while (CurrentToken) {
941  if (IsMark || CurrentToken->Previous->is(TT_BinaryOperator))
942  CurrentToken->Type = TT_ImplicitStringLiteral;
943  next();
944  }
945  }
946  }
947 
948  void parseHasInclude() {
949  if (!CurrentToken || !CurrentToken->is(tok::l_paren))
950  return;
951  next(); // '('
952  parseIncludeDirective();
953  next(); // ')'
954  }
955 
956  LineType parsePreprocessorDirective() {
957  bool IsFirstToken = CurrentToken->IsFirst;
959  next();
960  if (!CurrentToken)
961  return Type;
962 
963  if (Style.Language == FormatStyle::LK_JavaScript && IsFirstToken) {
964  // JavaScript files can contain shebang lines of the form:
965  // #!/usr/bin/env node
966  // Treat these like C++ #include directives.
967  while (CurrentToken) {
968  // Tokens cannot be comments here.
969  CurrentToken->Type = TT_ImplicitStringLiteral;
970  next();
971  }
972  return LT_ImportStatement;
973  }
974 
975  if (CurrentToken->Tok.is(tok::numeric_constant)) {
976  CurrentToken->SpacesRequiredBefore = 1;
977  return Type;
978  }
979  // Hashes in the middle of a line can lead to any strange token
980  // sequence.
981  if (!CurrentToken->Tok.getIdentifierInfo())
982  return Type;
983  switch (CurrentToken->Tok.getIdentifierInfo()->getPPKeywordID()) {
984  case tok::pp_include:
985  case tok::pp_include_next:
986  case tok::pp_import:
987  next();
988  parseIncludeDirective();
989  Type = LT_ImportStatement;
990  break;
991  case tok::pp_error:
992  case tok::pp_warning:
993  parseWarningOrError();
994  break;
995  case tok::pp_pragma:
996  parsePragma();
997  break;
998  case tok::pp_if:
999  case tok::pp_elif:
1000  Contexts.back().IsExpression = true;
1001  parseLine();
1002  break;
1003  default:
1004  break;
1005  }
1006  while (CurrentToken) {
1007  FormatToken *Tok = CurrentToken;
1008  next();
1009  if (Tok->is(tok::l_paren))
1010  parseParens();
1011  else if (Tok->isOneOf(Keywords.kw___has_include,
1012  Keywords.kw___has_include_next))
1013  parseHasInclude();
1014  }
1015  return Type;
1016  }
1017 
1018 public:
1019  LineType parseLine() {
1020  NonTemplateLess.clear();
1021  if (CurrentToken->is(tok::hash))
1022  return parsePreprocessorDirective();
1023 
1024  // Directly allow to 'import <string-literal>' to support protocol buffer
1025  // definitions (github.com/google/protobuf) or missing "#" (either way we
1026  // should not break the line).
1027  IdentifierInfo *Info = CurrentToken->Tok.getIdentifierInfo();
1028  if ((Style.Language == FormatStyle::LK_Java &&
1029  CurrentToken->is(Keywords.kw_package)) ||
1030  (Info && Info->getPPKeywordID() == tok::pp_import &&
1031  CurrentToken->Next &&
1032  CurrentToken->Next->isOneOf(tok::string_literal, tok::identifier,
1033  tok::kw_static))) {
1034  next();
1035  parseIncludeDirective();
1036  return LT_ImportStatement;
1037  }
1038 
1039  // If this line starts and ends in '<' and '>', respectively, it is likely
1040  // part of "#define <a/b.h>".
1041  if (CurrentToken->is(tok::less) && Line.Last->is(tok::greater)) {
1042  parseIncludeDirective();
1043  return LT_ImportStatement;
1044  }
1045 
1046  // In .proto files, top-level options are very similar to import statements
1047  // and should not be line-wrapped.
1048  if (Style.Language == FormatStyle::LK_Proto && Line.Level == 0 &&
1049  CurrentToken->is(Keywords.kw_option)) {
1050  next();
1051  if (CurrentToken && CurrentToken->is(tok::identifier))
1052  return LT_ImportStatement;
1053  }
1054 
1055  bool KeywordVirtualFound = false;
1056  bool ImportStatement = false;
1057 
1058  // import {...} from '...';
1059  if (Style.Language == FormatStyle::LK_JavaScript &&
1060  CurrentToken->is(Keywords.kw_import))
1061  ImportStatement = true;
1062 
1063  while (CurrentToken) {
1064  if (CurrentToken->is(tok::kw_virtual))
1065  KeywordVirtualFound = true;
1066  if (Style.Language == FormatStyle::LK_JavaScript) {
1067  // export {...} from '...';
1068  // An export followed by "from 'some string';" is a re-export from
1069  // another module identified by a URI and is treated as a
1070  // LT_ImportStatement (i.e. prevent wraps on it for long URIs).
1071  // Just "export {...};" or "export class ..." should not be treated as
1072  // an import in this sense.
1073  if (Line.First->is(tok::kw_export) &&
1074  CurrentToken->is(Keywords.kw_from) && CurrentToken->Next &&
1075  CurrentToken->Next->isStringLiteral())
1076  ImportStatement = true;
1077  if (isClosureImportStatement(*CurrentToken))
1078  ImportStatement = true;
1079  }
1080  if (!consumeToken())
1081  return LT_Invalid;
1082  }
1083  if (KeywordVirtualFound)
1084  return LT_VirtualFunctionDecl;
1085  if (ImportStatement)
1086  return LT_ImportStatement;
1087 
1088  if (Line.startsWith(TT_ObjCMethodSpecifier)) {
1089  if (Contexts.back().FirstObjCSelectorName)
1090  Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName =
1091  Contexts.back().LongestObjCSelectorName;
1092  return LT_ObjCMethodDecl;
1093  }
1094 
1095  return LT_Other;
1096  }
1097 
1098 private:
1099  bool isClosureImportStatement(const FormatToken &Tok) {
1100  // FIXME: Closure-library specific stuff should not be hard-coded but be
1101  // configurable.
1102  return Tok.TokenText == "goog" && Tok.Next && Tok.Next->is(tok::period) &&
1103  Tok.Next->Next &&
1104  (Tok.Next->Next->TokenText == "module" ||
1105  Tok.Next->Next->TokenText == "provide" ||
1106  Tok.Next->Next->TokenText == "require" ||
1107  Tok.Next->Next->TokenText == "forwardDeclare") &&
1108  Tok.Next->Next->Next && Tok.Next->Next->Next->is(tok::l_paren);
1109  }
1110 
1111  void resetTokenMetadata(FormatToken *Token) {
1112  if (!Token)
1113  return;
1114 
1115  // Reset token type in case we have already looked at it and then
1116  // recovered from an error (e.g. failure to find the matching >).
1117  if (!CurrentToken->isOneOf(TT_LambdaLSquare, TT_ForEachMacro,
1118  TT_FunctionLBrace, TT_ImplicitStringLiteral,
1119  TT_InlineASMBrace, TT_JsFatArrow, TT_LambdaArrow,
1120  TT_OverloadedOperator, TT_RegexLiteral,
1121  TT_TemplateString, TT_ObjCStringLiteral))
1122  CurrentToken->Type = TT_Unknown;
1123  CurrentToken->Role.reset();
1124  CurrentToken->MatchingParen = nullptr;
1125  CurrentToken->FakeLParens.clear();
1126  CurrentToken->FakeRParens = 0;
1127  }
1128 
1129  void next() {
1130  if (CurrentToken) {
1131  CurrentToken->NestingLevel = Contexts.size() - 1;
1132  CurrentToken->BindingStrength = Contexts.back().BindingStrength;
1133  modifyContext(*CurrentToken);
1134  determineTokenType(*CurrentToken);
1135  CurrentToken = CurrentToken->Next;
1136  }
1137 
1138  resetTokenMetadata(CurrentToken);
1139  }
1140 
1141  /// A struct to hold information valid in a specific context, e.g.
1142  /// a pair of parenthesis.
1143  struct Context {
1144  Context(tok::TokenKind ContextKind, unsigned BindingStrength,
1145  bool IsExpression)
1146  : ContextKind(ContextKind), BindingStrength(BindingStrength),
1147  IsExpression(IsExpression) {}
1148 
1153  bool ColonIsForRangeExpr = false;
1154  bool ColonIsDictLiteral = false;
1156  FormatToken *FirstObjCSelectorName = nullptr;
1157  FormatToken *FirstStartOfName = nullptr;
1158  bool CanBeExpression = true;
1159  bool InTemplateArgument = false;
1160  bool InCtorInitializer = false;
1161  bool InInheritanceList = false;
1162  bool CaretFound = false;
1163  bool IsForEachMacro = false;
1165  };
1166 
1167  /// Puts a new \c Context onto the stack \c Contexts for the lifetime
1168  /// of each instance.
1169  struct ScopedContextCreator {
1170  AnnotatingParser &P;
1171 
1172  ScopedContextCreator(AnnotatingParser &P, tok::TokenKind ContextKind,
1173  unsigned Increase)
1174  : P(P) {
1175  P.Contexts.push_back(Context(ContextKind,
1176  P.Contexts.back().BindingStrength + Increase,
1177  P.Contexts.back().IsExpression));
1178  }
1179 
1180  ~ScopedContextCreator() { P.Contexts.pop_back(); }
1181  };
1182 
1183  void modifyContext(const FormatToken &Current) {
1184  if (Current.getPrecedence() == prec::Assignment &&
1185  !Line.First->isOneOf(tok::kw_template, tok::kw_using, tok::kw_return) &&
1186  // Type aliases use `type X = ...;` in TypeScript and can be exported
1187  // using `export type ...`.
1188  !(Style.Language == FormatStyle::LK_JavaScript &&
1189  (Line.startsWith(Keywords.kw_type, tok::identifier) ||
1190  Line.startsWith(tok::kw_export, Keywords.kw_type,
1191  tok::identifier))) &&
1192  (!Current.Previous || Current.Previous->isNot(tok::kw_operator))) {
1193  Contexts.back().IsExpression = true;
1194  if (!Line.startsWith(TT_UnaryOperator)) {
1195  for (FormatToken *Previous = Current.Previous;
1196  Previous && Previous->Previous &&
1197  !Previous->Previous->isOneOf(tok::comma, tok::semi);
1198  Previous = Previous->Previous) {
1199  if (Previous->isOneOf(tok::r_square, tok::r_paren)) {
1200  Previous = Previous->MatchingParen;
1201  if (!Previous)
1202  break;
1203  }
1204  if (Previous->opensScope())
1205  break;
1206  if (Previous->isOneOf(TT_BinaryOperator, TT_UnaryOperator) &&
1207  Previous->isOneOf(tok::star, tok::amp, tok::ampamp) &&
1208  Previous->Previous && Previous->Previous->isNot(tok::equal))
1209  Previous->Type = TT_PointerOrReference;
1210  }
1211  }
1212  } else if (Current.is(tok::lessless) &&
1213  (!Current.Previous || !Current.Previous->is(tok::kw_operator))) {
1214  Contexts.back().IsExpression = true;
1215  } else if (Current.isOneOf(tok::kw_return, tok::kw_throw)) {
1216  Contexts.back().IsExpression = true;
1217  } else if (Current.is(TT_TrailingReturnArrow)) {
1218  Contexts.back().IsExpression = false;
1219  } else if (Current.is(TT_LambdaArrow) || Current.is(Keywords.kw_assert)) {
1220  Contexts.back().IsExpression = Style.Language == FormatStyle::LK_Java;
1221  } else if (Current.Previous &&
1222  Current.Previous->is(TT_CtorInitializerColon)) {
1223  Contexts.back().IsExpression = true;
1224  Contexts.back().InCtorInitializer = true;
1225  } else if (Current.Previous && Current.Previous->is(TT_InheritanceColon)) {
1226  Contexts.back().InInheritanceList = true;
1227  } else if (Current.isOneOf(tok::r_paren, tok::greater, tok::comma)) {
1228  for (FormatToken *Previous = Current.Previous;
1229  Previous && Previous->isOneOf(tok::star, tok::amp);
1230  Previous = Previous->Previous)
1231  Previous->Type = TT_PointerOrReference;
1232  if (Line.MustBeDeclaration && !Contexts.front().InCtorInitializer)
1233  Contexts.back().IsExpression = false;
1234  } else if (Current.is(tok::kw_new)) {
1235  Contexts.back().CanBeExpression = false;
1236  } else if (Current.isOneOf(tok::semi, tok::exclaim)) {
1237  // This should be the condition or increment in a for-loop.
1238  Contexts.back().IsExpression = true;
1239  }
1240  }
1241 
1242  void determineTokenType(FormatToken &Current) {
1243  if (!Current.is(TT_Unknown))
1244  // The token type is already known.
1245  return;
1246 
1247  if (Style.Language == FormatStyle::LK_JavaScript) {
1248  if (Current.is(tok::exclaim)) {
1249  if (Current.Previous &&
1250  (Current.Previous->isOneOf(tok::identifier, tok::kw_namespace,
1251  tok::r_paren, tok::r_square,
1252  tok::r_brace) ||
1253  Current.Previous->Tok.isLiteral())) {
1254  Current.Type = TT_JsNonNullAssertion;
1255  return;
1256  }
1257  if (Current.Next &&
1258  Current.Next->isOneOf(TT_BinaryOperator, Keywords.kw_as)) {
1259  Current.Type = TT_JsNonNullAssertion;
1260  return;
1261  }
1262  }
1263  }
1264 
1265  // Line.MightBeFunctionDecl can only be true after the parentheses of a
1266  // function declaration have been found. In this case, 'Current' is a
1267  // trailing token of this declaration and thus cannot be a name.
1268  if (Current.is(Keywords.kw_instanceof)) {
1269  Current.Type = TT_BinaryOperator;
1270  } else if (isStartOfName(Current) &&
1271  (!Line.MightBeFunctionDecl || Current.NestingLevel != 0)) {
1272  Contexts.back().FirstStartOfName = &Current;
1273  Current.Type = TT_StartOfName;
1274  } else if (Current.is(tok::semi)) {
1275  // Reset FirstStartOfName after finding a semicolon so that a for loop
1276  // with multiple increment statements is not confused with a for loop
1277  // having multiple variable declarations.
1278  Contexts.back().FirstStartOfName = nullptr;
1279  } else if (Current.isOneOf(tok::kw_auto, tok::kw___auto_type)) {
1280  AutoFound = true;
1281  } else if (Current.is(tok::arrow) &&
1282  Style.Language == FormatStyle::LK_Java) {
1283  Current.Type = TT_LambdaArrow;
1284  } else if (Current.is(tok::arrow) && AutoFound && Line.MustBeDeclaration &&
1285  Current.NestingLevel == 0) {
1286  Current.Type = TT_TrailingReturnArrow;
1287  } else if (Current.isOneOf(tok::star, tok::amp, tok::ampamp)) {
1288  Current.Type = determineStarAmpUsage(Current,
1289  Contexts.back().CanBeExpression &&
1290  Contexts.back().IsExpression,
1291  Contexts.back().InTemplateArgument);
1292  } else if (Current.isOneOf(tok::minus, tok::plus, tok::caret)) {
1293  Current.Type = determinePlusMinusCaretUsage(Current);
1294  if (Current.is(TT_UnaryOperator) && Current.is(tok::caret))
1295  Contexts.back().CaretFound = true;
1296  } else if (Current.isOneOf(tok::minusminus, tok::plusplus)) {
1297  Current.Type = determineIncrementUsage(Current);
1298  } else if (Current.isOneOf(tok::exclaim, tok::tilde)) {
1299  Current.Type = TT_UnaryOperator;
1300  } else if (Current.is(tok::question)) {
1301  if (Style.Language == FormatStyle::LK_JavaScript &&
1302  Line.MustBeDeclaration && !Contexts.back().IsExpression) {
1303  // In JavaScript, `interface X { foo?(): bar; }` is an optional method
1304  // on the interface, not a ternary expression.
1305  Current.Type = TT_JsTypeOptionalQuestion;
1306  } else {
1307  Current.Type = TT_ConditionalExpr;
1308  }
1309  } else if (Current.isBinaryOperator() &&
1310  (!Current.Previous || Current.Previous->isNot(tok::l_square)) &&
1311  (!Current.is(tok::greater) &&
1312  Style.Language != FormatStyle::LK_TextProto)) {
1313  Current.Type = TT_BinaryOperator;
1314  } else if (Current.is(tok::comment)) {
1315  if (Current.TokenText.startswith("/*")) {
1316  if (Current.TokenText.endswith("*/"))
1317  Current.Type = TT_BlockComment;
1318  else
1319  // The lexer has for some reason determined a comment here. But we
1320  // cannot really handle it, if it isn't properly terminated.
1321  Current.Tok.setKind(tok::unknown);
1322  } else {
1323  Current.Type = TT_LineComment;
1324  }
1325  } else if (Current.is(tok::r_paren)) {
1326  if (rParenEndsCast(Current))
1327  Current.Type = TT_CastRParen;
1328  if (Current.MatchingParen && Current.Next &&
1329  !Current.Next->isBinaryOperator() &&
1330  !Current.Next->isOneOf(tok::semi, tok::colon, tok::l_brace,
1331  tok::comma, tok::period, tok::arrow,
1332  tok::coloncolon))
1333  if (FormatToken *AfterParen = Current.MatchingParen->Next) {
1334  // Make sure this isn't the return type of an Obj-C block declaration
1335  if (AfterParen->Tok.isNot(tok::caret)) {
1336  if (FormatToken *BeforeParen = Current.MatchingParen->Previous)
1337  if (BeforeParen->is(tok::identifier) &&
1338  BeforeParen->TokenText == BeforeParen->TokenText.upper() &&
1339  (!BeforeParen->Previous ||
1340  BeforeParen->Previous->ClosesTemplateDeclaration))
1341  Current.Type = TT_FunctionAnnotationRParen;
1342  }
1343  }
1344  } else if (Current.is(tok::at) && Current.Next &&
1345  Style.Language != FormatStyle::LK_JavaScript &&
1346  Style.Language != FormatStyle::LK_Java) {
1347  // In Java & JavaScript, "@..." is a decorator or annotation. In ObjC, it
1348  // marks declarations and properties that need special formatting.
1349  switch (Current.Next->Tok.getObjCKeywordID()) {
1350  case tok::objc_interface:
1351  case tok::objc_implementation:
1352  case tok::objc_protocol:
1353  Current.Type = TT_ObjCDecl;
1354  break;
1355  case tok::objc_property:
1356  Current.Type = TT_ObjCProperty;
1357  break;
1358  default:
1359  break;
1360  }
1361  } else if (Current.is(tok::period)) {
1362  FormatToken *PreviousNoComment = Current.getPreviousNonComment();
1363  if (PreviousNoComment &&
1364  PreviousNoComment->isOneOf(tok::comma, tok::l_brace))
1365  Current.Type = TT_DesignatedInitializerPeriod;
1366  else if (Style.Language == FormatStyle::LK_Java && Current.Previous &&
1367  Current.Previous->isOneOf(TT_JavaAnnotation,
1368  TT_LeadingJavaAnnotation)) {
1369  Current.Type = Current.Previous->Type;
1370  }
1371  } else if (canBeObjCSelectorComponent(Current) &&
1372  // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen.
1373  Current.Previous && Current.Previous->is(TT_CastRParen) &&
1374  Current.Previous->MatchingParen &&
1375  Current.Previous->MatchingParen->Previous &&
1376  Current.Previous->MatchingParen->Previous->is(
1377  TT_ObjCMethodSpecifier)) {
1378  // This is the first part of an Objective-C selector name. (If there's no
1379  // colon after this, this is the only place which annotates the identifier
1380  // as a selector.)
1381  Current.Type = TT_SelectorName;
1382  } else if (Current.isOneOf(tok::identifier, tok::kw_const) &&
1383  Current.Previous &&
1384  !Current.Previous->isOneOf(tok::equal, tok::at) &&
1385  Line.MightBeFunctionDecl && Contexts.size() == 1) {
1386  // Line.MightBeFunctionDecl can only be true after the parentheses of a
1387  // function declaration have been found.
1388  Current.Type = TT_TrailingAnnotation;
1389  } else if ((Style.Language == FormatStyle::LK_Java ||
1390  Style.Language == FormatStyle::LK_JavaScript) &&
1391  Current.Previous) {
1392  if (Current.Previous->is(tok::at) &&
1393  Current.isNot(Keywords.kw_interface)) {
1394  const FormatToken &AtToken = *Current.Previous;
1395  const FormatToken *Previous = AtToken.getPreviousNonComment();
1396  if (!Previous || Previous->is(TT_LeadingJavaAnnotation))
1397  Current.Type = TT_LeadingJavaAnnotation;
1398  else
1399  Current.Type = TT_JavaAnnotation;
1400  } else if (Current.Previous->is(tok::period) &&
1401  Current.Previous->isOneOf(TT_JavaAnnotation,
1402  TT_LeadingJavaAnnotation)) {
1403  Current.Type = Current.Previous->Type;
1404  }
1405  }
1406  }
1407 
1408  /// Take a guess at whether \p Tok starts a name of a function or
1409  /// variable declaration.
1410  ///
1411  /// This is a heuristic based on whether \p Tok is an identifier following
1412  /// something that is likely a type.
1413  bool isStartOfName(const FormatToken &Tok) {
1414  if (Tok.isNot(tok::identifier) || !Tok.Previous)
1415  return false;
1416 
1417  if (Tok.Previous->isOneOf(TT_LeadingJavaAnnotation, Keywords.kw_instanceof,
1418  Keywords.kw_as))
1419  return false;
1420  if (Style.Language == FormatStyle::LK_JavaScript &&
1421  Tok.Previous->is(Keywords.kw_in))
1422  return false;
1423 
1424  // Skip "const" as it does not have an influence on whether this is a name.
1425  FormatToken *PreviousNotConst = Tok.getPreviousNonComment();
1426  while (PreviousNotConst && PreviousNotConst->is(tok::kw_const))
1427  PreviousNotConst = PreviousNotConst->getPreviousNonComment();
1428 
1429  if (!PreviousNotConst)
1430  return false;
1431 
1432  bool IsPPKeyword = PreviousNotConst->is(tok::identifier) &&
1433  PreviousNotConst->Previous &&
1434  PreviousNotConst->Previous->is(tok::hash);
1435 
1436  if (PreviousNotConst->is(TT_TemplateCloser))
1437  return PreviousNotConst && PreviousNotConst->MatchingParen &&
1438  PreviousNotConst->MatchingParen->Previous &&
1439  PreviousNotConst->MatchingParen->Previous->isNot(tok::period) &&
1440  PreviousNotConst->MatchingParen->Previous->isNot(tok::kw_template);
1441 
1442  if (PreviousNotConst->is(tok::r_paren) && PreviousNotConst->MatchingParen &&
1443  PreviousNotConst->MatchingParen->Previous &&
1444  PreviousNotConst->MatchingParen->Previous->is(tok::kw_decltype))
1445  return true;
1446 
1447  return (!IsPPKeyword &&
1448  PreviousNotConst->isOneOf(tok::identifier, tok::kw_auto)) ||
1449  PreviousNotConst->is(TT_PointerOrReference) ||
1450  PreviousNotConst->isSimpleTypeSpecifier();
1451  }
1452 
1453  /// Determine whether ')' is ending a cast.
1454  bool rParenEndsCast(const FormatToken &Tok) {
1455  // C-style casts are only used in C++ and Java.
1456  if (!Style.isCpp() && Style.Language != FormatStyle::LK_Java)
1457  return false;
1458 
1459  // Empty parens aren't casts and there are no casts at the end of the line.
1460  if (Tok.Previous == Tok.MatchingParen || !Tok.Next || !Tok.MatchingParen)
1461  return false;
1462 
1463  FormatToken *LeftOfParens = Tok.MatchingParen->getPreviousNonComment();
1464  if (LeftOfParens) {
1465  // If there is a closing parenthesis left of the current parentheses,
1466  // look past it as these might be chained casts.
1467  if (LeftOfParens->is(tok::r_paren)) {
1468  if (!LeftOfParens->MatchingParen ||
1469  !LeftOfParens->MatchingParen->Previous)
1470  return false;
1471  LeftOfParens = LeftOfParens->MatchingParen->Previous;
1472  }
1473 
1474  // If there is an identifier (or with a few exceptions a keyword) right
1475  // before the parentheses, this is unlikely to be a cast.
1476  if (LeftOfParens->Tok.getIdentifierInfo() &&
1477  !LeftOfParens->isOneOf(Keywords.kw_in, tok::kw_return, tok::kw_case,
1478  tok::kw_delete))
1479  return false;
1480 
1481  // Certain other tokens right before the parentheses are also signals that
1482  // this cannot be a cast.
1483  if (LeftOfParens->isOneOf(tok::at, tok::r_square, TT_OverloadedOperator,
1484  TT_TemplateCloser, tok::ellipsis))
1485  return false;
1486  }
1487 
1488  if (Tok.Next->is(tok::question))
1489  return false;
1490 
1491  // As Java has no function types, a "(" after the ")" likely means that this
1492  // is a cast.
1493  if (Style.Language == FormatStyle::LK_Java && Tok.Next->is(tok::l_paren))
1494  return true;
1495 
1496  // If a (non-string) literal follows, this is likely a cast.
1497  if (Tok.Next->isNot(tok::string_literal) &&
1498  (Tok.Next->Tok.isLiteral() ||
1499  Tok.Next->isOneOf(tok::kw_sizeof, tok::kw_alignof)))
1500  return true;
1501 
1502  // Heuristically try to determine whether the parentheses contain a type.
1503  bool ParensAreType =
1504  !Tok.Previous ||
1505  Tok.Previous->isOneOf(TT_PointerOrReference, TT_TemplateCloser) ||
1506  Tok.Previous->isSimpleTypeSpecifier();
1507  bool ParensCouldEndDecl =
1508  Tok.Next->isOneOf(tok::equal, tok::semi, tok::l_brace, tok::greater);
1509  if (ParensAreType && !ParensCouldEndDecl)
1510  return true;
1511 
1512  // At this point, we heuristically assume that there are no casts at the
1513  // start of the line. We assume that we have found most cases where there
1514  // are by the logic above, e.g. "(void)x;".
1515  if (!LeftOfParens)
1516  return false;
1517 
1518  // Certain token types inside the parentheses mean that this can't be a
1519  // cast.
1520  for (const FormatToken *Token = Tok.MatchingParen->Next; Token != &Tok;
1521  Token = Token->Next)
1522  if (Token->is(TT_BinaryOperator))
1523  return false;
1524 
1525  // If the following token is an identifier or 'this', this is a cast. All
1526  // cases where this can be something else are handled above.
1527  if (Tok.Next->isOneOf(tok::identifier, tok::kw_this))
1528  return true;
1529 
1530  if (!Tok.Next->Next)
1531  return false;
1532 
1533  // If the next token after the parenthesis is a unary operator, assume
1534  // that this is cast, unless there are unexpected tokens inside the
1535  // parenthesis.
1536  bool NextIsUnary =
1537  Tok.Next->isUnaryOperator() || Tok.Next->isOneOf(tok::amp, tok::star);
1538  if (!NextIsUnary || Tok.Next->is(tok::plus) ||
1539  !Tok.Next->Next->isOneOf(tok::identifier, tok::numeric_constant))
1540  return false;
1541  // Search for unexpected tokens.
1542  for (FormatToken *Prev = Tok.Previous; Prev != Tok.MatchingParen;
1543  Prev = Prev->Previous) {
1544  if (!Prev->isOneOf(tok::kw_const, tok::identifier, tok::coloncolon))
1545  return false;
1546  }
1547  return true;
1548  }
1549 
1550  /// Return the type of the given token assuming it is * or &.
1551  TokenType determineStarAmpUsage(const FormatToken &Tok, bool IsExpression,
1552  bool InTemplateArgument) {
1553  if (Style.Language == FormatStyle::LK_JavaScript)
1554  return TT_BinaryOperator;
1555 
1556  const FormatToken *PrevToken = Tok.getPreviousNonComment();
1557  if (!PrevToken)
1558  return TT_UnaryOperator;
1559 
1560  const FormatToken *NextToken = Tok.getNextNonComment();
1561  if (!NextToken ||
1562  NextToken->isOneOf(tok::arrow, tok::equal, tok::kw_const) ||
1563  (NextToken->is(tok::l_brace) && !NextToken->getNextNonComment()))
1564  return TT_PointerOrReference;
1565 
1566  if (PrevToken->is(tok::coloncolon))
1567  return TT_PointerOrReference;
1568 
1569  if (PrevToken->isOneOf(tok::l_paren, tok::l_square, tok::l_brace,
1570  tok::comma, tok::semi, tok::kw_return, tok::colon,
1571  tok::equal, tok::kw_delete, tok::kw_sizeof,
1572  tok::kw_throw) ||
1573  PrevToken->isOneOf(TT_BinaryOperator, TT_ConditionalExpr,
1574  TT_UnaryOperator, TT_CastRParen))
1575  return TT_UnaryOperator;
1576 
1577  if (NextToken->is(tok::l_square) && NextToken->isNot(TT_LambdaLSquare))
1578  return TT_PointerOrReference;
1579  if (NextToken->is(tok::kw_operator) && !IsExpression)
1580  return TT_PointerOrReference;
1581  if (NextToken->isOneOf(tok::comma, tok::semi))
1582  return TT_PointerOrReference;
1583 
1584  if (PrevToken->is(tok::r_paren) && PrevToken->MatchingParen) {
1585  FormatToken *TokenBeforeMatchingParen =
1586  PrevToken->MatchingParen->getPreviousNonComment();
1587  if (TokenBeforeMatchingParen &&
1588  TokenBeforeMatchingParen->isOneOf(tok::kw_typeof, tok::kw_decltype))
1589  return TT_PointerOrReference;
1590  }
1591 
1592  if (PrevToken->Tok.isLiteral() ||
1593  PrevToken->isOneOf(tok::r_paren, tok::r_square, tok::kw_true,
1594  tok::kw_false, tok::r_brace) ||
1595  NextToken->Tok.isLiteral() ||
1596  NextToken->isOneOf(tok::kw_true, tok::kw_false) ||
1597  NextToken->isUnaryOperator() ||
1598  // If we know we're in a template argument, there are no named
1599  // declarations. Thus, having an identifier on the right-hand side
1600  // indicates a binary operator.
1601  (InTemplateArgument && NextToken->Tok.isAnyIdentifier()))
1602  return TT_BinaryOperator;
1603 
1604  // "&&(" is quite unlikely to be two successive unary "&".
1605  if (Tok.is(tok::ampamp) && NextToken && NextToken->is(tok::l_paren))
1606  return TT_BinaryOperator;
1607 
1608  // This catches some cases where evaluation order is used as control flow:
1609  // aaa && aaa->f();
1610  const FormatToken *NextNextToken = NextToken->getNextNonComment();
1611  if (NextNextToken && NextNextToken->is(tok::arrow))
1612  return TT_BinaryOperator;
1613 
1614  // It is very unlikely that we are going to find a pointer or reference type
1615  // definition on the RHS of an assignment.
1616  if (IsExpression && !Contexts.back().CaretFound)
1617  return TT_BinaryOperator;
1618 
1619  return TT_PointerOrReference;
1620  }
1621 
1622  TokenType determinePlusMinusCaretUsage(const FormatToken &Tok) {
1623  const FormatToken *PrevToken = Tok.getPreviousNonComment();
1624  if (!PrevToken)
1625  return TT_UnaryOperator;
1626 
1627  if (PrevToken->isOneOf(TT_CastRParen, TT_UnaryOperator))
1628  // This must be a sequence of leading unary operators.
1629  return TT_UnaryOperator;
1630 
1631  // Use heuristics to recognize unary operators.
1632  if (PrevToken->isOneOf(tok::equal, tok::l_paren, tok::comma, tok::l_square,
1633  tok::question, tok::colon, tok::kw_return,
1634  tok::kw_case, tok::at, tok::l_brace))
1635  return TT_UnaryOperator;
1636 
1637  // There can't be two consecutive binary operators.
1638  if (PrevToken->is(TT_BinaryOperator))
1639  return TT_UnaryOperator;
1640 
1641  // Fall back to marking the token as binary operator.
1642  return TT_BinaryOperator;
1643  }
1644 
1645  /// Determine whether ++/-- are pre- or post-increments/-decrements.
1646  TokenType determineIncrementUsage(const FormatToken &Tok) {
1647  const FormatToken *PrevToken = Tok.getPreviousNonComment();
1648  if (!PrevToken || PrevToken->is(TT_CastRParen))
1649  return TT_UnaryOperator;
1650  if (PrevToken->isOneOf(tok::r_paren, tok::r_square, tok::identifier))
1651  return TT_TrailingUnaryOperator;
1652 
1653  return TT_UnaryOperator;
1654  }
1655 
1656  SmallVector<Context, 8> Contexts;
1657 
1658  const FormatStyle &Style;
1659  AnnotatedLine &Line;
1660  FormatToken *CurrentToken;
1661  bool AutoFound;
1662  const AdditionalKeywords &Keywords;
1663 
1664  // Set of "<" tokens that do not open a template parameter list. If parseAngle
1665  // determines that a specific token can't be a template opener, it will make
1666  // same decision irrespective of the decisions for tokens leading up to it.
1667  // Store this information to prevent this from causing exponential runtime.
1668  llvm::SmallPtrSet<FormatToken *, 16> NonTemplateLess;
1669 };
1670 
1671 static const int PrecedenceUnaryOperator = prec::PointerToMember + 1;
1672 static const int PrecedenceArrowAndPeriod = prec::PointerToMember + 2;
1673 
1674 /// Parses binary expressions by inserting fake parenthesis based on
1675 /// operator precedence.
1676 class ExpressionParser {
1677 public:
1678  ExpressionParser(const FormatStyle &Style, const AdditionalKeywords &Keywords,
1679  AnnotatedLine &Line)
1680  : Style(Style), Keywords(Keywords), Current(Line.First) {}
1681 
1682  /// Parse expressions with the given operator precedence.
1683  void parse(int Precedence = 0) {
1684  // Skip 'return' and ObjC selector colons as they are not part of a binary
1685  // expression.
1686  while (Current && (Current->is(tok::kw_return) ||
1687  (Current->is(tok::colon) &&
1688  Current->isOneOf(TT_ObjCMethodExpr, TT_DictLiteral))))
1689  next();
1690 
1691  if (!Current || Precedence > PrecedenceArrowAndPeriod)
1692  return;
1693 
1694  // Conditional expressions need to be parsed separately for proper nesting.
1695  if (Precedence == prec::Conditional) {
1696  parseConditionalExpr();
1697  return;
1698  }
1699 
1700  // Parse unary operators, which all have a higher precedence than binary
1701  // operators.
1702  if (Precedence == PrecedenceUnaryOperator) {
1703  parseUnaryOperator();
1704  return;
1705  }
1706 
1707  FormatToken *Start = Current;
1708  FormatToken *LatestOperator = nullptr;
1709  unsigned OperatorIndex = 0;
1710 
1711  while (Current) {
1712  // Consume operators with higher precedence.
1713  parse(Precedence + 1);
1714 
1715  int CurrentPrecedence = getCurrentPrecedence();
1716 
1717  if (Current && Current->is(TT_SelectorName) &&
1718  Precedence == CurrentPrecedence) {
1719  if (LatestOperator)
1720  addFakeParenthesis(Start, prec::Level(Precedence));
1721  Start = Current;
1722  }
1723 
1724  // At the end of the line or when an operator with higher precedence is
1725  // found, insert fake parenthesis and return.
1726  if (!Current ||
1727  (Current->closesScope() &&
1728  (Current->MatchingParen || Current->is(TT_TemplateString))) ||
1729  (CurrentPrecedence != -1 && CurrentPrecedence < Precedence) ||
1730  (CurrentPrecedence == prec::Conditional &&
1731  Precedence == prec::Assignment && Current->is(tok::colon))) {
1732  break;
1733  }
1734 
1735  // Consume scopes: (), [], <> and {}
1736  if (Current->opensScope()) {
1737  // In fragment of a JavaScript template string can look like '}..${' and
1738  // thus close a scope and open a new one at the same time.
1739  while (Current && (!Current->closesScope() || Current->opensScope())) {
1740  next();
1741  parse();
1742  }
1743  next();
1744  } else {
1745  // Operator found.
1746  if (CurrentPrecedence == Precedence) {
1747  if (LatestOperator)
1748  LatestOperator->NextOperator = Current;
1749  LatestOperator = Current;
1750  Current->OperatorIndex = OperatorIndex;
1751  ++OperatorIndex;
1752  }
1753  next(/*SkipPastLeadingComments=*/Precedence > 0);
1754  }
1755  }
1756 
1757  if (LatestOperator && (Current || Precedence > 0)) {
1758  // LatestOperator->LastOperator = true;
1759  if (Precedence == PrecedenceArrowAndPeriod) {
1760  // Call expressions don't have a binary operator precedence.
1761  addFakeParenthesis(Start, prec::Unknown);
1762  } else {
1763  addFakeParenthesis(Start, prec::Level(Precedence));
1764  }
1765  }
1766  }
1767 
1768 private:
1769  /// Gets the precedence (+1) of the given token for binary operators
1770  /// and other tokens that we treat like binary operators.
1771  int getCurrentPrecedence() {
1772  if (Current) {
1773  const FormatToken *NextNonComment = Current->getNextNonComment();
1774  if (Current->is(TT_ConditionalExpr))
1775  return prec::Conditional;
1776  if (NextNonComment && Current->is(TT_SelectorName) &&
1777  (NextNonComment->isOneOf(TT_DictLiteral, TT_JsTypeColon) ||
1778  ((Style.Language == FormatStyle::LK_Proto ||
1779  Style.Language == FormatStyle::LK_TextProto) &&
1780  NextNonComment->is(tok::less))))
1781  return prec::Assignment;
1782  if (Current->is(TT_JsComputedPropertyName))
1783  return prec::Assignment;
1784  if (Current->is(TT_LambdaArrow))
1785  return prec::Comma;
1786  if (Current->is(TT_JsFatArrow))
1787  return prec::Assignment;
1788  if (Current->isOneOf(tok::semi, TT_InlineASMColon, TT_SelectorName) ||
1789  (Current->is(tok::comment) && NextNonComment &&
1790  NextNonComment->is(TT_SelectorName)))
1791  return 0;
1792  if (Current->is(TT_RangeBasedForLoopColon))
1793  return prec::Comma;
1794  if ((Style.Language == FormatStyle::LK_Java ||
1795  Style.Language == FormatStyle::LK_JavaScript) &&
1796  Current->is(Keywords.kw_instanceof))
1797  return prec::Relational;
1798  if (Style.Language == FormatStyle::LK_JavaScript &&
1799  Current->isOneOf(Keywords.kw_in, Keywords.kw_as))
1800  return prec::Relational;
1801  if (Current->is(TT_BinaryOperator) || Current->is(tok::comma))
1802  return Current->getPrecedence();
1803  if (Current->isOneOf(tok::period, tok::arrow))
1804  return PrecedenceArrowAndPeriod;
1805  if ((Style.Language == FormatStyle::LK_Java ||
1806  Style.Language == FormatStyle::LK_JavaScript) &&
1807  Current->isOneOf(Keywords.kw_extends, Keywords.kw_implements,
1808  Keywords.kw_throws))
1809  return 0;
1810  }
1811  return -1;
1812  }
1813 
1814  void addFakeParenthesis(FormatToken *Start, prec::Level Precedence) {
1815  Start->FakeLParens.push_back(Precedence);
1816  if (Precedence > prec::Unknown)
1817  Start->StartsBinaryExpression = true;
1818  if (Current) {
1819  FormatToken *Previous = Current->Previous;
1820  while (Previous->is(tok::comment) && Previous->Previous)
1821  Previous = Previous->Previous;
1822  ++Previous->FakeRParens;
1823  if (Precedence > prec::Unknown)
1824  Previous->EndsBinaryExpression = true;
1825  }
1826  }
1827 
1828  /// Parse unary operator expressions and surround them with fake
1829  /// parentheses if appropriate.
1830  void parseUnaryOperator() {
1832  while (Current && Current->is(TT_UnaryOperator)) {
1833  Tokens.push_back(Current);
1834  next();
1835  }
1836  parse(PrecedenceArrowAndPeriod);
1837  for (FormatToken *Token : llvm::reverse(Tokens))
1838  // The actual precedence doesn't matter.
1839  addFakeParenthesis(Token, prec::Unknown);
1840  }
1841 
1842  void parseConditionalExpr() {
1843  while (Current && Current->isTrailingComment()) {
1844  next();
1845  }
1846  FormatToken *Start = Current;
1847  parse(prec::LogicalOr);
1848  if (!Current || !Current->is(tok::question))
1849  return;
1850  next();
1851  parse(prec::Assignment);
1852  if (!Current || Current->isNot(TT_ConditionalExpr))
1853  return;
1854  next();
1855  parse(prec::Assignment);
1856  addFakeParenthesis(Start, prec::Conditional);
1857  }
1858 
1859  void next(bool SkipPastLeadingComments = true) {
1860  if (Current)
1861  Current = Current->Next;
1862  while (Current &&
1863  (Current->NewlinesBefore == 0 || SkipPastLeadingComments) &&
1864  Current->isTrailingComment())
1865  Current = Current->Next;
1866  }
1867 
1868  const FormatStyle &Style;
1869  const AdditionalKeywords &Keywords;
1870  FormatToken *Current;
1871 };
1872 
1873 } // end anonymous namespace
1874 
1877  const AnnotatedLine *NextNonCommentLine = nullptr;
1879  E = Lines.rend();
1880  I != E; ++I) {
1881  bool CommentLine = true;
1882  for (const FormatToken *Tok = (*I)->First; Tok; Tok = Tok->Next) {
1883  if (!Tok->is(tok::comment)) {
1884  CommentLine = false;
1885  break;
1886  }
1887  }
1888 
1889  // If the comment is currently aligned with the line immediately following
1890  // it, that's probably intentional and we should keep it.
1891  if (NextNonCommentLine && CommentLine &&
1892  NextNonCommentLine->First->NewlinesBefore <= 1 &&
1893  NextNonCommentLine->First->OriginalColumn ==
1894  (*I)->First->OriginalColumn) {
1895  // Align comments for preprocessor lines with the # in column 0.
1896  // Otherwise, align with the next line.
1897  (*I)->Level = (NextNonCommentLine->Type == LT_PreprocessorDirective ||
1898  NextNonCommentLine->Type == LT_ImportStatement)
1899  ? 0
1900  : NextNonCommentLine->Level;
1901  } else {
1902  NextNonCommentLine = (*I)->First->isNot(tok::r_brace) ? (*I) : nullptr;
1903  }
1904 
1905  setCommentLineLevels((*I)->Children);
1906  }
1907 }
1908 
1909 static unsigned maxNestingDepth(const AnnotatedLine &Line) {
1910  unsigned Result = 0;
1911  for (const auto *Tok = Line.First; Tok != nullptr; Tok = Tok->Next)
1912  Result = std::max(Result, Tok->NestingLevel);
1913  return Result;
1914 }
1915 
1918  E = Line.Children.end();
1919  I != E; ++I) {
1920  annotate(**I);
1921  }
1922  AnnotatingParser Parser(Style, Line, Keywords);
1923  Line.Type = Parser.parseLine();
1924 
1925  // With very deep nesting, ExpressionParser uses lots of stack and the
1926  // formatting algorithm is very slow. We're not going to do a good job here
1927  // anyway - it's probably generated code being formatted by mistake.
1928  // Just skip the whole line.
1929  if (maxNestingDepth(Line) > 50)
1930  Line.Type = LT_Invalid;
1931 
1932  if (Line.Type == LT_Invalid)
1933  return;
1934 
1935  ExpressionParser ExprParser(Style, Keywords, Line);
1936  ExprParser.parse();
1937 
1938  if (Line.startsWith(TT_ObjCMethodSpecifier))
1939  Line.Type = LT_ObjCMethodDecl;
1940  else if (Line.startsWith(TT_ObjCDecl))
1941  Line.Type = LT_ObjCDecl;
1942  else if (Line.startsWith(TT_ObjCProperty))
1943  Line.Type = LT_ObjCProperty;
1944 
1945  Line.First->SpacesRequiredBefore = 1;
1946  Line.First->CanBreakBefore = Line.First->MustBreakBefore;
1947 }
1948 
1949 // This function heuristically determines whether 'Current' starts the name of a
1950 // function declaration.
1951 static bool isFunctionDeclarationName(const FormatToken &Current,
1952  const AnnotatedLine &Line) {
1953  auto skipOperatorName = [](const FormatToken *Next) -> const FormatToken * {
1954  for (; Next; Next = Next->Next) {
1955  if (Next->is(TT_OverloadedOperatorLParen))
1956  return Next;
1957  if (Next->is(TT_OverloadedOperator))
1958  continue;
1959  if (Next->isOneOf(tok::kw_new, tok::kw_delete)) {
1960  // For 'new[]' and 'delete[]'.
1961  if (Next->Next && Next->Next->is(tok::l_square) && Next->Next->Next &&
1962  Next->Next->Next->is(tok::r_square))
1963  Next = Next->Next->Next;
1964  continue;
1965  }
1966 
1967  break;
1968  }
1969  return nullptr;
1970  };
1971 
1972  // Find parentheses of parameter list.
1973  const FormatToken *Next = Current.Next;
1974  if (Current.is(tok::kw_operator)) {
1975  if (Current.Previous && Current.Previous->is(tok::coloncolon))
1976  return false;
1977  Next = skipOperatorName(Next);
1978  } else {
1979  if (!Current.is(TT_StartOfName) || Current.NestingLevel != 0)
1980  return false;
1981  for (; Next; Next = Next->Next) {
1982  if (Next->is(TT_TemplateOpener)) {
1983  Next = Next->MatchingParen;
1984  } else if (Next->is(tok::coloncolon)) {
1985  Next = Next->Next;
1986  if (!Next)
1987  return false;
1988  if (Next->is(tok::kw_operator)) {
1989  Next = skipOperatorName(Next->Next);
1990  break;
1991  }
1992  if (!Next->is(tok::identifier))
1993  return false;
1994  } else if (Next->is(tok::l_paren)) {
1995  break;
1996  } else {
1997  return false;
1998  }
1999  }
2000  }
2001 
2002  // Check whether parameter list can belong to a function declaration.
2003  if (!Next || !Next->is(tok::l_paren) || !Next->MatchingParen)
2004  return false;
2005  // If the lines ends with "{", this is likely an function definition.
2006  if (Line.Last->is(tok::l_brace))
2007  return true;
2008  if (Next->Next == Next->MatchingParen)
2009  return true; // Empty parentheses.
2010  // If there is an &/&& after the r_paren, this is likely a function.
2011  if (Next->MatchingParen->Next &&
2012  Next->MatchingParen->Next->is(TT_PointerOrReference))
2013  return true;
2014  for (const FormatToken *Tok = Next->Next; Tok && Tok != Next->MatchingParen;
2015  Tok = Tok->Next) {
2016  if (Tok->is(tok::l_paren) && Tok->MatchingParen) {
2017  Tok = Tok->MatchingParen;
2018  continue;
2019  }
2020  if (Tok->is(tok::kw_const) || Tok->isSimpleTypeSpecifier() ||
2021  Tok->isOneOf(TT_PointerOrReference, TT_StartOfName, tok::ellipsis))
2022  return true;
2023  if (Tok->isOneOf(tok::l_brace, tok::string_literal, TT_ObjCMethodExpr) ||
2024  Tok->Tok.isLiteral())
2025  return false;
2026  }
2027  return false;
2028 }
2029 
2030 bool TokenAnnotator::mustBreakForReturnType(const AnnotatedLine &Line) const {
2031  assert(Line.MightBeFunctionDecl);
2032 
2036  Line.Level > 0)
2037  return false;
2038 
2041  return false;
2042  case FormatStyle::RTBS_All:
2044  return true;
2047  return Line.mightBeFunctionDefinition();
2048  }
2049 
2050  return false;
2051 }
2052 
2055  E = Line.Children.end();
2056  I != E; ++I) {
2057  calculateFormattingInformation(**I);
2058  }
2059 
2060  Line.First->TotalLength =
2062  : Line.FirstStartColumn + Line.First->ColumnWidth;
2063  FormatToken *Current = Line.First->Next;
2064  bool InFunctionDecl = Line.MightBeFunctionDecl;
2065  while (Current) {
2066  if (isFunctionDeclarationName(*Current, Line))
2067  Current->Type = TT_FunctionDeclarationName;
2068  if (Current->is(TT_LineComment)) {
2069  if (Current->Previous->BlockKind == BK_BracedInit &&
2070  Current->Previous->opensScope())
2071  Current->SpacesRequiredBefore = Style.Cpp11BracedListStyle ? 0 : 1;
2072  else
2074 
2075  // If we find a trailing comment, iterate backwards to determine whether
2076  // it seems to relate to a specific parameter. If so, break before that
2077  // parameter to avoid changing the comment's meaning. E.g. don't move 'b'
2078  // to the previous line in:
2079  // SomeFunction(a,
2080  // b, // comment
2081  // c);
2082  if (!Current->HasUnescapedNewline) {
2083  for (FormatToken *Parameter = Current->Previous; Parameter;
2084  Parameter = Parameter->Previous) {
2085  if (Parameter->isOneOf(tok::comment, tok::r_brace))
2086  break;
2087  if (Parameter->Previous && Parameter->Previous->is(tok::comma)) {
2088  if (!Parameter->Previous->is(TT_CtorInitializerComma) &&
2089  Parameter->HasUnescapedNewline)
2090  Parameter->MustBreakBefore = true;
2091  break;
2092  }
2093  }
2094  }
2095  } else if (Current->SpacesRequiredBefore == 0 &&
2096  spaceRequiredBefore(Line, *Current)) {
2097  Current->SpacesRequiredBefore = 1;
2098  }
2099 
2100  Current->MustBreakBefore =
2101  Current->MustBreakBefore || mustBreakBefore(Line, *Current);
2102 
2103  if (!Current->MustBreakBefore && InFunctionDecl &&
2104  Current->is(TT_FunctionDeclarationName))
2105  Current->MustBreakBefore = mustBreakForReturnType(Line);
2106 
2107  Current->CanBreakBefore =
2108  Current->MustBreakBefore || canBreakBefore(Line, *Current);
2109  unsigned ChildSize = 0;
2110  if (Current->Previous->Children.size() == 1) {
2111  FormatToken &LastOfChild = *Current->Previous->Children[0]->Last;
2112  ChildSize = LastOfChild.isTrailingComment() ? Style.ColumnLimit
2113  : LastOfChild.TotalLength + 1;
2114  }
2115  const FormatToken *Prev = Current->Previous;
2116  if (Current->MustBreakBefore || Prev->Children.size() > 1 ||
2117  (Prev->Children.size() == 1 &&
2118  Prev->Children[0]->First->MustBreakBefore) ||
2119  Current->IsMultiline)
2120  Current->TotalLength = Prev->TotalLength + Style.ColumnLimit;
2121  else
2122  Current->TotalLength = Prev->TotalLength + Current->ColumnWidth +
2123  ChildSize + Current->SpacesRequiredBefore;
2124 
2125  if (Current->is(TT_CtorInitializerColon))
2126  InFunctionDecl = false;
2127 
2128  // FIXME: Only calculate this if CanBreakBefore is true once static
2129  // initializers etc. are sorted out.
2130  // FIXME: Move magic numbers to a better place.
2131  Current->SplitPenalty = 20 * Current->BindingStrength +
2132  splitPenalty(Line, *Current, InFunctionDecl);
2133 
2134  Current = Current->Next;
2135  }
2136 
2137  calculateUnbreakableTailLengths(Line);
2138  unsigned IndentLevel = Line.Level;
2139  for (Current = Line.First; Current != nullptr; Current = Current->Next) {
2140  if (Current->Role)
2141  Current->Role->precomputeFormattingInfos(Current);
2142  if (Current->MatchingParen &&
2144  assert(IndentLevel > 0);
2145  --IndentLevel;
2146  }
2147  Current->IndentLevel = IndentLevel;
2148  if (Current->opensBlockOrBlockTypeList(Style))
2149  ++IndentLevel;
2150  }
2151 
2152  LLVM_DEBUG({ printDebugInfo(Line); });
2153 }
2154 
2155 void TokenAnnotator::calculateUnbreakableTailLengths(AnnotatedLine &Line) {
2156  unsigned UnbreakableTailLength = 0;
2157  FormatToken *Current = Line.Last;
2158  while (Current) {
2159  Current->UnbreakableTailLength = UnbreakableTailLength;
2160  if (Current->CanBreakBefore ||
2161  Current->isOneOf(tok::comment, tok::string_literal)) {
2162  UnbreakableTailLength = 0;
2163  } else {
2164  UnbreakableTailLength +=
2165  Current->ColumnWidth + Current->SpacesRequiredBefore;
2166  }
2167  Current = Current->Previous;
2168  }
2169 }
2170 
2171 unsigned TokenAnnotator::splitPenalty(const AnnotatedLine &Line,
2172  const FormatToken &Tok,
2173  bool InFunctionDecl) {
2174  const FormatToken &Left = *Tok.Previous;
2175  const FormatToken &Right = Tok;
2176 
2177  if (Left.is(tok::semi))
2178  return 0;
2179 
2181  if (Right.isOneOf(Keywords.kw_extends, Keywords.kw_throws))
2182  return 1;
2183  if (Right.is(Keywords.kw_implements))
2184  return 2;
2185  if (Left.is(tok::comma) && Left.NestingLevel == 0)
2186  return 3;
2187  } else if (Style.Language == FormatStyle::LK_JavaScript) {
2188  if (Right.is(Keywords.kw_function) && Left.isNot(tok::comma))
2189  return 100;
2190  if (Left.is(TT_JsTypeColon))
2191  return 35;
2192  if ((Left.is(TT_TemplateString) && Left.TokenText.endswith("${")) ||
2193  (Right.is(TT_TemplateString) && Right.TokenText.startswith("}")))
2194  return 100;
2195  // Prefer breaking call chains (".foo") over empty "{}", "[]" or "()".
2196  if (Left.opensScope() && Right.closesScope())
2197  return 200;
2198  }
2199 
2200  if (Right.is(tok::identifier) && Right.Next && Right.Next->is(TT_DictLiteral))
2201  return 1;
2202  if (Right.is(tok::l_square)) {
2204  return 1;
2205  if (Left.is(tok::r_square))
2206  return 200;
2207  // Slightly prefer formatting local lambda definitions like functions.
2208  if (Right.is(TT_LambdaLSquare) && Left.is(tok::equal))
2209  return 35;
2210  if (!Right.isOneOf(TT_ObjCMethodExpr, TT_LambdaLSquare,
2211  TT_ArrayInitializerLSquare,
2212  TT_DesignatedInitializerLSquare, TT_AttributeSquare))
2213  return 500;
2214  }
2215 
2216  if (Right.isOneOf(TT_StartOfName, TT_FunctionDeclarationName) ||
2217  Right.is(tok::kw_operator)) {
2218  if (Line.startsWith(tok::kw_for) && Right.PartOfMultiVariableDeclStmt)
2219  return 3;
2220  if (Left.is(TT_StartOfName))
2221  return 110;
2222  if (InFunctionDecl && Right.NestingLevel == 0)
2224  return 200;
2225  }
2226  if (Right.is(TT_PointerOrReference))
2227  return 190;
2228  if (Right.is(TT_LambdaArrow))
2229  return 110;
2230  if (Left.is(tok::equal) && Right.is(tok::l_brace))
2231  return 160;
2232  if (Left.is(TT_CastRParen))
2233  return 100;
2234  if (Left.is(tok::coloncolon) ||
2235  (Right.is(tok::period) && Style.Language == FormatStyle::LK_Proto))
2236  return 500;
2237  if (Left.isOneOf(tok::kw_class, tok::kw_struct))
2238  return 5000;
2239  if (Left.is(tok::comment))
2240  return 1000;
2241 
2242  if (Left.isOneOf(TT_RangeBasedForLoopColon, TT_InheritanceColon,
2243  TT_CtorInitializerColon))
2244  return 2;
2245 
2246  if (Right.isMemberAccess()) {
2247  // Breaking before the "./->" of a chained call/member access is reasonably
2248  // cheap, as formatting those with one call per line is generally
2249  // desirable. In particular, it should be cheaper to break before the call
2250  // than it is to break inside a call's parameters, which could lead to weird
2251  // "hanging" indents. The exception is the very last "./->" to support this
2252  // frequent pattern:
2253  //
2254  // aaaaaaaa.aaaaaaaa.bbbbbbb().ccccccccccccccccccccc(
2255  // dddddddd);
2256  //
2257  // which might otherwise be blown up onto many lines. Here, clang-format
2258  // won't produce "hanging" indents anyway as there is no other trailing
2259  // call.
2260  //
2261  // Also apply higher penalty is not a call as that might lead to a wrapping
2262  // like:
2263  //
2264  // aaaaaaa
2265  // .aaaaaaaaa.bbbbbbbb(cccccccc);
2266  return !Right.NextOperator || !Right.NextOperator->Previous->closesScope()
2267  ? 150
2268  : 35;
2269  }
2270 
2271  if (Right.is(TT_TrailingAnnotation) &&
2272  (!Right.Next || Right.Next->isNot(tok::l_paren))) {
2273  // Moving trailing annotations to the next line is fine for ObjC method
2274  // declarations.
2275  if (Line.startsWith(TT_ObjCMethodSpecifier))
2276  return 10;
2277  // Generally, breaking before a trailing annotation is bad unless it is
2278  // function-like. It seems to be especially preferable to keep standard
2279  // annotations (i.e. "const", "final" and "override") on the same line.
2280  // Use a slightly higher penalty after ")" so that annotations like
2281  // "const override" are kept together.
2282  bool is_short_annotation = Right.TokenText.size() < 10;
2283  return (Left.is(tok::r_paren) ? 100 : 120) + (is_short_annotation ? 50 : 0);
2284  }
2285 
2286  // In for-loops, prefer breaking at ',' and ';'.
2287  if (Line.startsWith(tok::kw_for) && Left.is(tok::equal))
2288  return 4;
2289 
2290  // In Objective-C method expressions, prefer breaking before "param:" over
2291  // breaking after it.
2292  if (Right.is(TT_SelectorName))
2293  return 0;
2294  if (Left.is(tok::colon) && Left.is(TT_ObjCMethodExpr))
2295  return Line.MightBeFunctionDecl ? 50 : 500;
2296 
2297  // In Objective-C type declarations, avoid breaking after the category's
2298  // open paren (we'll prefer breaking after the protocol list's opening
2299  // angle bracket, if present).
2300  if (Line.Type == LT_ObjCDecl && Left.is(tok::l_paren) && Left.Previous &&
2301  Left.Previous->isOneOf(tok::identifier, tok::greater))
2302  return 500;
2303 
2304  if (Left.is(tok::l_paren) && InFunctionDecl &&
2306  return 100;
2307  if (Left.is(tok::l_paren) && Left.Previous &&
2308  (Left.Previous->isOneOf(tok::kw_if, tok::kw_for) ||
2309  Left.Previous->endsSequence(tok::kw_constexpr, tok::kw_if)))
2310  return 1000;
2311  if (Left.is(tok::equal) && InFunctionDecl)
2312  return 110;
2313  if (Right.is(tok::r_brace))
2314  return 1;
2315  if (Left.is(TT_TemplateOpener))
2316  return 100;
2317  if (Left.opensScope()) {
2319  return 0;
2320  if (Left.is(tok::l_brace) && !Style.Cpp11BracedListStyle)
2321  return 19;
2323  : 19;
2324  }
2325  if (Left.is(TT_JavaAnnotation))
2326  return 50;
2327 
2328  if (Left.isOneOf(tok::plus, tok::comma) && Left.Previous &&
2329  Left.Previous->isLabelString() &&
2330  (Left.NextOperator || Left.OperatorIndex != 0))
2331  return 50;
2332  if (Right.is(tok::plus) && Left.isLabelString() &&
2333  (Right.NextOperator || Right.OperatorIndex != 0))
2334  return 25;
2335  if (Left.is(tok::comma))
2336  return 1;
2337  if (Right.is(tok::lessless) && Left.isLabelString() &&
2338  (Right.NextOperator || Right.OperatorIndex != 1))
2339  return 25;
2340  if (Right.is(tok::lessless)) {
2341  // Breaking at a << is really cheap.
2342  if (!Left.is(tok::r_paren) || Right.OperatorIndex > 0)
2343  // Slightly prefer to break before the first one in log-like statements.
2344  return 2;
2345  return 1;
2346  }
2347  if (Left.ClosesTemplateDeclaration)
2349  if (Left.is(TT_ConditionalExpr))
2350  return prec::Conditional;
2351  prec::Level Level = Left.getPrecedence();
2352  if (Level == prec::Unknown)
2353  Level = Right.getPrecedence();
2354  if (Level == prec::Assignment)
2356  if (Level != prec::Unknown)
2357  return Level;
2358 
2359  return 3;
2360 }
2361 
2362 bool TokenAnnotator::spaceRequiredBetween(const AnnotatedLine &Line,
2363  const FormatToken &Left,
2364  const FormatToken &Right) {
2365  if (Left.is(tok::kw_return) && Right.isNot(tok::semi))
2366  return true;
2367  if (Left.is(Keywords.kw_assert) && Style.Language == FormatStyle::LK_Java)
2368  return true;
2370  Left.Tok.getObjCKeywordID() == tok::objc_property)
2371  return true;
2372  if (Right.is(tok::hashhash))
2373  return Left.is(tok::hash);
2374  if (Left.isOneOf(tok::hashhash, tok::hash))
2375  return Right.is(tok::hash);
2376  if (Left.is(tok::l_paren) && Right.is(tok::r_paren))
2378  if (Left.is(tok::l_paren) || Right.is(tok::r_paren))
2379  return (Right.is(TT_CastRParen) ||
2380  (Left.MatchingParen && Left.MatchingParen->is(TT_CastRParen)))
2383  if (Right.isOneOf(tok::semi, tok::comma))
2384  return false;
2385  if (Right.is(tok::less) && Line.Type == LT_ObjCDecl) {
2386  bool IsLightweightGeneric =
2387  Right.MatchingParen && Right.MatchingParen->Next &&
2388  Right.MatchingParen->Next->is(tok::colon);
2389  return !IsLightweightGeneric && Style.ObjCSpaceBeforeProtocolList;
2390  }
2391  if (Right.is(tok::less) && Left.is(tok::kw_template))
2393  if (Left.isOneOf(tok::exclaim, tok::tilde))
2394  return false;
2395  if (Left.is(tok::at) &&
2396  Right.isOneOf(tok::identifier, tok::string_literal, tok::char_constant,
2397  tok::numeric_constant, tok::l_paren, tok::l_brace,
2398  tok::kw_true, tok::kw_false))
2399  return false;
2400  if (Left.is(tok::colon))
2401  return !Left.is(TT_ObjCMethodExpr);
2402  if (Left.is(tok::coloncolon))
2403  return false;
2404  if (Left.is(tok::less) || Right.isOneOf(tok::greater, tok::less)) {
2407  (Left.is(TT_DictLiteral) || Right.is(TT_DictLiteral)))) {
2408  // Format empty list as `<>`.
2409  if (Left.is(tok::less) && Right.is(tok::greater))
2410  return false;
2411  return !Style.Cpp11BracedListStyle;
2412  }
2413  return false;
2414  }
2415  if (Right.is(tok::ellipsis))
2416  return Left.Tok.isLiteral() || (Left.is(tok::identifier) && Left.Previous &&
2417  Left.Previous->is(tok::kw_case));
2418  if (Left.is(tok::l_square) && Right.is(tok::amp))
2419  return false;
2420  if (Right.is(TT_PointerOrReference)) {
2421  if (Left.is(tok::r_paren) && Line.MightBeFunctionDecl) {
2422  if (!Left.MatchingParen)
2423  return true;
2424  FormatToken *TokenBeforeMatchingParen =
2426  if (!TokenBeforeMatchingParen ||
2427  !TokenBeforeMatchingParen->isOneOf(tok::kw_typeof, tok::kw_decltype))
2428  return true;
2429  }
2430  return (Left.Tok.isLiteral() ||
2431  (!Left.isOneOf(TT_PointerOrReference, tok::l_paren) &&
2433  (Line.IsMultiVariableDeclStmt &&
2434  (Left.NestingLevel == 0 ||
2435  (Left.NestingLevel == 1 && Line.First->is(tok::kw_for)))))));
2436  }
2437  if (Right.is(TT_FunctionTypeLParen) && Left.isNot(tok::l_paren) &&
2438  (!Left.is(TT_PointerOrReference) ||
2440  !Line.IsMultiVariableDeclStmt)))
2441  return true;
2442  if (Left.is(TT_PointerOrReference))
2443  return Right.Tok.isLiteral() || Right.is(TT_BlockComment) ||
2444  (Right.isOneOf(Keywords.kw_override, Keywords.kw_final) &&
2445  !Right.is(TT_StartOfName)) ||
2446  (Right.is(tok::l_brace) && Right.BlockKind == BK_Block) ||
2447  (!Right.isOneOf(TT_PointerOrReference, TT_ArraySubscriptLSquare,
2448  tok::l_paren) &&
2450  !Line.IsMultiVariableDeclStmt) &&
2451  Left.Previous &&
2452  !Left.Previous->isOneOf(tok::l_paren, tok::coloncolon));
2453  if (Right.is(tok::star) && Left.is(tok::l_paren))
2454  return false;
2455  const auto SpaceRequiredForArrayInitializerLSquare =
2456  [](const FormatToken &LSquareTok, const FormatStyle &Style) {
2461  LSquareTok.endsSequence(tok::l_square, tok::colon,
2462  TT_SelectorName));
2463  };
2464  if (Left.is(tok::l_square))
2465  return (Left.is(TT_ArrayInitializerLSquare) && Right.isNot(tok::r_square) &&
2466  SpaceRequiredForArrayInitializerLSquare(Left, Style)) ||
2467  (Left.isOneOf(TT_ArraySubscriptLSquare,
2468  TT_StructuredBindingLSquare) &&
2469  Style.SpacesInSquareBrackets && Right.isNot(tok::r_square));
2470  if (Right.is(tok::r_square))
2471  return Right.MatchingParen &&
2472  ((Right.MatchingParen->is(TT_ArrayInitializerLSquare) &&
2473  SpaceRequiredForArrayInitializerLSquare(*Right.MatchingParen,
2474  Style)) ||
2476  Right.MatchingParen->isOneOf(TT_ArraySubscriptLSquare,
2477  TT_StructuredBindingLSquare)) ||
2478  Right.MatchingParen->is(TT_AttributeParen));
2479  if (Right.is(tok::l_square) &&
2480  !Right.isOneOf(TT_ObjCMethodExpr, TT_LambdaLSquare,
2481  TT_DesignatedInitializerLSquare,
2482  TT_StructuredBindingLSquare, TT_AttributeSquare) &&
2483  !Left.isOneOf(tok::numeric_constant, TT_DictLiteral))
2484  return false;
2485  if (Left.is(tok::l_brace) && Right.is(tok::r_brace))
2486  return !Left.Children.empty(); // No spaces in "{}".
2487  if ((Left.is(tok::l_brace) && Left.BlockKind != BK_Block) ||
2488  (Right.is(tok::r_brace) && Right.MatchingParen &&
2489  Right.MatchingParen->BlockKind != BK_Block))
2490  return !Style.Cpp11BracedListStyle;
2491  if (Left.is(TT_BlockComment))
2492  return !Left.TokenText.endswith("=*/");
2493  if (Right.is(tok::l_paren)) {
2494  if ((Left.is(tok::r_paren) && Left.is(TT_AttributeParen)) ||
2495  (Left.is(tok::r_square) && Left.is(TT_AttributeSquare)))
2496  return true;
2497  return Line.Type == LT_ObjCDecl || Left.is(tok::semi) ||
2499  (Left.isOneOf(tok::kw_if, tok::pp_elif, tok::kw_for, tok::kw_while,
2500  tok::kw_switch, tok::kw_case, TT_ForEachMacro,
2501  TT_ObjCForIn) ||
2502  Left.endsSequence(tok::kw_constexpr, tok::kw_if) ||
2503  (Left.isOneOf(tok::kw_try, Keywords.kw___except, tok::kw_catch,
2504  tok::kw_new, tok::kw_delete) &&
2505  (!Left.Previous || Left.Previous->isNot(tok::period))))) ||
2507  (Left.is(tok::identifier) || Left.isFunctionLikeKeyword() ||
2508  Left.is(tok::r_paren)) &&
2509  Line.Type != LT_PreprocessorDirective);
2510  }
2511  if (Left.is(tok::at) && Right.Tok.getObjCKeywordID() != tok::objc_not_keyword)
2512  return false;
2513  if (Right.is(TT_UnaryOperator))
2514  return !Left.isOneOf(tok::l_paren, tok::l_square, tok::at) &&
2515  (Left.isNot(tok::colon) || Left.isNot(TT_ObjCMethodExpr));
2516  if ((Left.isOneOf(tok::identifier, tok::greater, tok::r_square,
2517  tok::r_paren) ||
2518  Left.isSimpleTypeSpecifier()) &&
2519  Right.is(tok::l_brace) && Right.getNextNonComment() &&
2520  Right.BlockKind != BK_Block)
2521  return false;
2522  if (Left.is(tok::period) || Right.is(tok::period))
2523  return false;
2524  if (Right.is(tok::hash) && Left.is(tok::identifier) && Left.TokenText == "L")
2525  return false;
2526  if (Left.is(TT_TemplateCloser) && Left.MatchingParen &&
2527  Left.MatchingParen->Previous &&
2528  Left.MatchingParen->Previous->is(tok::period))
2529  // A.<B<C<...>>>DoSomething();
2530  return false;
2531  if (Left.is(TT_TemplateCloser) && Right.is(tok::l_square))
2532  return false;
2533  if (Left.is(tok::l_brace) && Left.endsSequence(TT_DictLiteral, tok::at))
2534  // Objective-C dictionary literal -> no space after opening brace.
2535  return false;
2536  if (Right.is(tok::r_brace) && Right.MatchingParen &&
2537  Right.MatchingParen->endsSequence(TT_DictLiteral, tok::at))
2538  // Objective-C dictionary literal -> no space before closing brace.
2539  return false;
2540  return true;
2541 }
2542 
2543 bool TokenAnnotator::spaceRequiredBefore(const AnnotatedLine &Line,
2544  const FormatToken &Right) {
2545  const FormatToken &Left = *Right.Previous;
2546  if (Right.Tok.getIdentifierInfo() && Left.Tok.getIdentifierInfo())
2547  return true; // Never ever merge two identifiers.
2548  if (Style.isCpp()) {
2549  if (Left.is(tok::kw_operator))
2550  return Right.is(tok::coloncolon);
2551  if (Right.is(tok::l_brace) && Right.BlockKind == BK_BracedInit &&
2553  return true;
2554  } else if (Style.Language == FormatStyle::LK_Proto ||
2556  if (Right.is(tok::period) &&
2557  Left.isOneOf(Keywords.kw_optional, Keywords.kw_required,
2558  Keywords.kw_repeated, Keywords.kw_extend))
2559  return true;
2560  if (Right.is(tok::l_paren) &&
2561  Left.isOneOf(Keywords.kw_returns, Keywords.kw_option))
2562  return true;
2563  if (Right.isOneOf(tok::l_brace, tok::less) && Left.is(TT_SelectorName))
2564  return true;
2565  // Slashes occur in text protocol extension syntax: [type/type] { ... }.
2566  if (Left.is(tok::slash) || Right.is(tok::slash))
2567  return false;
2568  if (Left.MatchingParen && Left.MatchingParen->is(TT_ProtoExtensionLSquare) &&
2569  Right.isOneOf(tok::l_brace, tok::less))
2570  return !Style.Cpp11BracedListStyle;
2571  // A percent is probably part of a formatting specification, such as %lld.
2572  if (Left.is(tok::percent))
2573  return false;
2574  // Preserve the existence of a space before a percent for cases like 0x%04x
2575  // and "%d %d"
2576  if (Left.is(tok::numeric_constant) && Right.is(tok::percent))
2577  return Right.WhitespaceRange.getEnd() != Right.WhitespaceRange.getBegin();
2578  } else if (Style.Language == FormatStyle::LK_JavaScript) {
2579  if (Left.is(TT_JsFatArrow))
2580  return true;
2581  // for await ( ...
2582  if (Right.is(tok::l_paren) && Left.is(Keywords.kw_await) && Left.Previous &&
2583  Left.Previous->is(tok::kw_for))
2584  return true;
2585  if (Left.is(Keywords.kw_async) && Right.is(tok::l_paren) &&
2586  Right.MatchingParen) {
2587  const FormatToken *Next = Right.MatchingParen->getNextNonComment();
2588  // An async arrow function, for example: `x = async () => foo();`,
2589  // as opposed to calling a function called async: `x = async();`
2590  if (Next && Next->is(TT_JsFatArrow))
2591  return true;
2592  }
2593  if ((Left.is(TT_TemplateString) && Left.TokenText.endswith("${")) ||
2594  (Right.is(TT_TemplateString) && Right.TokenText.startswith("}")))
2595  return false;
2596  // In tagged template literals ("html`bar baz`"), there is no space between
2597  // the tag identifier and the template string. getIdentifierInfo makes sure
2598  // that the identifier is not a pseudo keyword like `yield`, either.
2599  if (Left.is(tok::identifier) && Keywords.IsJavaScriptIdentifier(Left) &&
2600  Right.is(TT_TemplateString))
2601  return false;
2602  if (Right.is(tok::star) &&
2603  Left.isOneOf(Keywords.kw_function, Keywords.kw_yield))
2604  return false;
2605  if (Right.isOneOf(tok::l_brace, tok::l_square) &&
2606  Left.isOneOf(Keywords.kw_function, Keywords.kw_yield,
2607  Keywords.kw_extends, Keywords.kw_implements))
2608  return true;
2609  if (Right.is(tok::l_paren)) {
2610  // JS methods can use some keywords as names (e.g. `delete()`).
2611  if (Line.MustBeDeclaration && Left.Tok.getIdentifierInfo())
2612  return false;
2613  // Valid JS method names can include keywords, e.g. `foo.delete()` or
2614  // `bar.instanceof()`. Recognize call positions by preceding period.
2615  if (Left.Previous && Left.Previous->is(tok::period) &&
2616  Left.Tok.getIdentifierInfo())
2617  return false;
2618  // Additional unary JavaScript operators that need a space after.
2619  if (Left.isOneOf(tok::kw_throw, Keywords.kw_await, Keywords.kw_typeof,
2620  tok::kw_void))
2621  return true;
2622  }
2623  if ((Left.isOneOf(Keywords.kw_let, Keywords.kw_var, Keywords.kw_in,
2624  tok::kw_const) ||
2625  // "of" is only a keyword if it appears after another identifier
2626  // (e.g. as "const x of y" in a for loop), or after a destructuring
2627  // operation (const [x, y] of z, const {a, b} of c).
2628  (Left.is(Keywords.kw_of) && Left.Previous &&
2629  (Left.Previous->Tok.is(tok::identifier) ||
2630  Left.Previous->isOneOf(tok::r_square, tok::r_brace)))) &&
2631  (!Left.Previous || !Left.Previous->is(tok::period)))
2632  return true;
2633  if (Left.isOneOf(tok::kw_for, Keywords.kw_as) && Left.Previous &&
2634  Left.Previous->is(tok::period) && Right.is(tok::l_paren))
2635  return false;
2636  if (Left.is(Keywords.kw_as) &&
2637  Right.isOneOf(tok::l_square, tok::l_brace, tok::l_paren))
2638  return true;
2639  if (Left.is(tok::kw_default) && Left.Previous &&
2640  Left.Previous->is(tok::kw_export))
2641  return true;
2642  if (Left.is(Keywords.kw_is) && Right.is(tok::l_brace))
2643  return true;
2644  if (Right.isOneOf(TT_JsTypeColon, TT_JsTypeOptionalQuestion))
2645  return false;
2646  if (Left.is(TT_JsTypeOperator) || Right.is(TT_JsTypeOperator))
2647  return false;
2648  if ((Left.is(tok::l_brace) || Right.is(tok::r_brace)) &&
2649  Line.First->isOneOf(Keywords.kw_import, tok::kw_export))
2650  return false;
2651  if (Left.is(tok::ellipsis))
2652  return false;
2653  if (Left.is(TT_TemplateCloser) &&
2654  !Right.isOneOf(tok::equal, tok::l_brace, tok::comma, tok::l_square,
2655  Keywords.kw_implements, Keywords.kw_extends))
2656  // Type assertions ('<type>expr') are not followed by whitespace. Other
2657  // locations that should have whitespace following are identified by the
2658  // above set of follower tokens.
2659  return false;
2660  if (Right.is(TT_JsNonNullAssertion))
2661  return false;
2662  if (Left.is(TT_JsNonNullAssertion) &&
2663  Right.isOneOf(Keywords.kw_as, Keywords.kw_in))
2664  return true; // "x! as string", "x! in y"
2665  } else if (Style.Language == FormatStyle::LK_Java) {
2666  if (Left.is(tok::r_square) && Right.is(tok::l_brace))
2667  return true;
2668  if (Left.is(Keywords.kw_synchronized) && Right.is(tok::l_paren))
2670  if ((Left.isOneOf(tok::kw_static, tok::kw_public, tok::kw_private,
2671  tok::kw_protected) ||
2672  Left.isOneOf(Keywords.kw_final, Keywords.kw_abstract,
2673  Keywords.kw_native)) &&
2674  Right.is(TT_TemplateOpener))
2675  return true;
2676  }
2677  if (Left.is(TT_ImplicitStringLiteral))
2678  return Right.WhitespaceRange.getBegin() != Right.WhitespaceRange.getEnd();
2679  if (Line.Type == LT_ObjCMethodDecl) {
2680  if (Left.is(TT_ObjCMethodSpecifier))
2681  return true;
2682  if (Left.is(tok::r_paren) && canBeObjCSelectorComponent(Right))
2683  // Don't space between ')' and <id> or ')' and 'new'. 'new' is not a
2684  // keyword in Objective-C, and '+ (instancetype)new;' is a standard class
2685  // method declaration.
2686  return false;
2687  }
2688  if (Line.Type == LT_ObjCProperty &&
2689  (Right.is(tok::equal) || Left.is(tok::equal)))
2690  return false;
2691 
2692  if (Right.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow) ||
2693  Left.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow))
2694  return true;
2695  if (Right.is(TT_OverloadedOperatorLParen))
2697  if (Left.is(tok::comma))
2698  return true;
2699  if (Right.is(tok::comma))
2700  return false;
2701  if (Right.is(TT_ObjCBlockLParen))
2702  return true;
2703  if (Right.is(TT_CtorInitializerColon))
2705  if (Right.is(TT_InheritanceColon) && !Style.SpaceBeforeInheritanceColon)
2706  return false;
2707  if (Right.is(TT_RangeBasedForLoopColon) &&
2709  return false;
2710  if (Right.is(tok::colon)) {
2711  if (Line.First->isOneOf(tok::kw_case, tok::kw_default) ||
2712  !Right.getNextNonComment() || Right.getNextNonComment()->is(tok::semi))
2713  return false;
2714  if (Right.is(TT_ObjCMethodExpr))
2715  return false;
2716  if (Left.is(tok::question))
2717  return false;
2718  if (Right.is(TT_InlineASMColon) && Left.is(tok::coloncolon))
2719  return false;
2720  if (Right.is(TT_DictLiteral))
2722  if (Right.is(TT_AttributeColon))
2723  return false;
2724  return true;
2725  }
2726  if (Left.is(TT_UnaryOperator))
2727  return Right.is(TT_BinaryOperator);
2728 
2729  // If the next token is a binary operator or a selector name, we have
2730  // incorrectly classified the parenthesis as a cast. FIXME: Detect correctly.
2731  if (Left.is(TT_CastRParen))
2732  return Style.SpaceAfterCStyleCast ||
2733  Right.isOneOf(TT_BinaryOperator, TT_SelectorName);
2734 
2735  if (Left.is(tok::greater) && Right.is(tok::greater)) {
2737  (Style.Language == FormatStyle::LK_Proto && Left.is(TT_DictLiteral)))
2738  return !Style.Cpp11BracedListStyle;
2739  return Right.is(TT_TemplateCloser) && Left.is(TT_TemplateCloser) &&
2741  }
2742  if (Right.isOneOf(tok::arrow, tok::arrowstar, tok::periodstar) ||
2743  Left.isOneOf(tok::arrow, tok::period, tok::arrowstar, tok::periodstar) ||
2744  (Right.is(tok::period) && Right.isNot(TT_DesignatedInitializerPeriod)))
2745  return false;
2747  Right.getPrecedence() == prec::Assignment)
2748  return false;
2749  if (Right.is(tok::coloncolon) && Left.is(tok::identifier))
2750  // Generally don't remove existing spaces between an identifier and "::".
2751  // The identifier might actually be a macro name such as ALWAYS_INLINE. If
2752  // this turns out to be too lenient, add analysis of the identifier itself.
2753  return Right.WhitespaceRange.getBegin() != Right.WhitespaceRange.getEnd();
2754  if (Right.is(tok::coloncolon) && !Left.isOneOf(tok::l_brace, tok::comment))
2755  return (Left.is(TT_TemplateOpener) &&
2757  !(Left.isOneOf(tok::l_paren, tok::r_paren, tok::l_square,
2758  tok::kw___super, TT_TemplateCloser,
2759  TT_TemplateOpener)) ||
2760  (Left.is(tok ::l_paren) && Style.SpacesInParentheses);
2761  if ((Left.is(TT_TemplateOpener)) != (Right.is(TT_TemplateCloser)))
2762  return Style.SpacesInAngles;
2763  // Space before TT_StructuredBindingLSquare.
2764  if (Right.is(TT_StructuredBindingLSquare))
2765  return !Left.isOneOf(tok::amp, tok::ampamp) ||
2767  // Space before & or && following a TT_StructuredBindingLSquare.
2768  if (Right.Next && Right.Next->is(TT_StructuredBindingLSquare) &&
2769  Right.isOneOf(tok::amp, tok::ampamp))
2771  if ((Right.is(TT_BinaryOperator) && !Left.is(tok::l_paren)) ||
2772  (Left.isOneOf(TT_BinaryOperator, TT_ConditionalExpr) &&
2773  !Right.is(tok::r_paren)))
2774  return true;
2775  if (Left.is(TT_TemplateCloser) && Right.is(tok::l_paren) &&
2776  Right.isNot(TT_FunctionTypeLParen))
2778  if (Right.is(TT_TemplateOpener) && Left.is(tok::r_paren) &&
2779  Left.MatchingParen && Left.MatchingParen->is(TT_OverloadedOperatorLParen))
2780  return false;
2781  if (Right.is(tok::less) && Left.isNot(tok::l_paren) &&
2782  Line.startsWith(tok::hash))
2783  return true;
2784  if (Right.is(TT_TrailingUnaryOperator))
2785  return false;
2786  if (Left.is(TT_RegexLiteral))
2787  return false;
2788  return spaceRequiredBetween(Line, Left, Right);
2789 }
2790 
2791 // Returns 'true' if 'Tok' is a brace we'd want to break before in Allman style.
2792 static bool isAllmanBrace(const FormatToken &Tok) {
2793  return Tok.is(tok::l_brace) && Tok.BlockKind == BK_Block &&
2794  !Tok.isOneOf(TT_ObjCBlockLBrace, TT_DictLiteral);
2795 }
2796 
2797 bool TokenAnnotator::mustBreakBefore(const AnnotatedLine &Line,
2798  const FormatToken &Right) {
2799  const FormatToken &Left = *Right.Previous;
2800  if (Right.NewlinesBefore > 1 && Style.MaxEmptyLinesToKeep > 0)
2801  return true;
2802 
2804  // FIXME: This might apply to other languages and token kinds.
2805  if (Right.is(tok::string_literal) && Left.is(tok::plus) && Left.Previous &&
2806  Left.Previous->is(tok::string_literal))
2807  return true;
2808  if (Left.is(TT_DictLiteral) && Left.is(tok::l_brace) && Line.Level == 0 &&
2809  Left.Previous && Left.Previous->is(tok::equal) &&
2810  Line.First->isOneOf(tok::identifier, Keywords.kw_import, tok::kw_export,
2811  tok::kw_const) &&
2812  // kw_var/kw_let are pseudo-tokens that are tok::identifier, so match
2813  // above.
2814  !Line.First->isOneOf(Keywords.kw_var, Keywords.kw_let))
2815  // Object literals on the top level of a file are treated as "enum-style".
2816  // Each key/value pair is put on a separate line, instead of bin-packing.
2817  return true;
2818  if (Left.is(tok::l_brace) && Line.Level == 0 &&
2819  (Line.startsWith(tok::kw_enum) ||
2820  Line.startsWith(tok::kw_const, tok::kw_enum) ||
2821  Line.startsWith(tok::kw_export, tok::kw_enum) ||
2822  Line.startsWith(tok::kw_export, tok::kw_const, tok::kw_enum)))
2823  // JavaScript top-level enum key/value pairs are put on separate lines
2824  // instead of bin-packing.
2825  return true;
2826  if (Right.is(tok::r_brace) && Left.is(tok::l_brace) &&
2827  !Left.Children.empty())
2828  // Support AllowShortFunctionsOnASingleLine for JavaScript.
2831  (Left.NestingLevel == 0 && Line.Level == 0 &&
2834  } else if (Style.Language == FormatStyle::LK_Java) {
2835  if (Right.is(tok::plus) && Left.is(tok::string_literal) && Right.Next &&
2836  Right.Next->is(tok::string_literal))
2837  return true;
2838  } else if (Style.Language == FormatStyle::LK_Cpp ||
2842  if (Left.isStringLiteral() && Right.isStringLiteral())
2843  return true;
2844  }
2845 
2846  // If the last token before a '}', ']', or ')' is a comma or a trailing
2847  // comment, the intention is to insert a line break after it in order to make
2848  // shuffling around entries easier. Import statements, especially in
2849  // JavaScript, can be an exception to this rule.
2851  const FormatToken *BeforeClosingBrace = nullptr;
2852  if ((Left.isOneOf(tok::l_brace, TT_ArrayInitializerLSquare) ||
2854  Left.is(tok::l_paren))) &&
2855  Left.BlockKind != BK_Block && Left.MatchingParen)
2856  BeforeClosingBrace = Left.MatchingParen->Previous;
2857  else if (Right.MatchingParen &&
2858  (Right.MatchingParen->isOneOf(tok::l_brace,
2859  TT_ArrayInitializerLSquare) ||
2861  Right.MatchingParen->is(tok::l_paren))))
2862  BeforeClosingBrace = &Left;
2863  if (BeforeClosingBrace && (BeforeClosingBrace->is(tok::comma) ||
2864  BeforeClosingBrace->isTrailingComment()))
2865  return true;
2866  }
2867 
2868  if (Right.is(tok::comment))
2869  return Left.BlockKind != BK_BracedInit &&
2870  Left.isNot(TT_CtorInitializerColon) &&
2871  (Right.NewlinesBefore > 0 && Right.HasUnescapedNewline);
2872  if (Left.isTrailingComment())
2873  return true;
2874  if (Right.Previous->IsUnterminatedLiteral)
2875  return true;
2876  if (Right.is(tok::lessless) && Right.Next &&
2877  Right.Previous->is(tok::string_literal) &&
2878  Right.Next->is(tok::string_literal))
2879  return true;
2880  if (Right.Previous->ClosesTemplateDeclaration &&
2881  Right.Previous->MatchingParen &&
2882  Right.Previous->MatchingParen->NestingLevel == 0 &&
2884  return true;
2885  if (Right.is(TT_CtorInitializerComma) &&
2888  return true;
2889  if (Right.is(TT_CtorInitializerColon) &&
2892  return true;
2893  // Break only if we have multiple inheritance.
2895  Right.is(TT_InheritanceComma))
2896  return true;
2897  if (Right.is(tok::string_literal) && Right.TokenText.startswith("R\""))
2898  // Multiline raw string literals are special wrt. line breaks. The author
2899  // has made a deliberate choice and might have aligned the contents of the
2900  // string literal accordingly. Thus, we try keep existing line breaks.
2901  return Right.IsMultiline && Right.NewlinesBefore > 0;
2902  if ((Right.Previous->is(tok::l_brace) ||
2903  (Right.Previous->is(tok::less) && Right.Previous->Previous &&
2904  Right.Previous->Previous->is(tok::equal))) &&
2905  Right.NestingLevel == 1 && Style.Language == FormatStyle::LK_Proto) {
2906  // Don't put enums or option definitions onto single lines in protocol
2907  // buffers.
2908  return true;
2909  }
2910  if (Right.is(TT_InlineASMBrace))
2911  return Right.HasUnescapedNewline;
2912  if (isAllmanBrace(Left) || isAllmanBrace(Right))
2913  return (Line.startsWith(tok::kw_enum) && Style.BraceWrapping.AfterEnum) ||
2914  (Line.startsWith(tok::kw_typedef, tok::kw_enum) &&
2916  (Line.startsWith(tok::kw_class) && Style.BraceWrapping.AfterClass) ||
2917  (Line.startsWith(tok::kw_struct) && Style.BraceWrapping.AfterStruct);
2918  if (Left.is(TT_ObjCBlockLBrace) && !Style.AllowShortBlocksOnASingleLine)
2919  return true;
2920 
2923  Left.is(TT_LeadingJavaAnnotation) &&
2924  Right.isNot(TT_LeadingJavaAnnotation) && Right.isNot(tok::l_paren) &&
2925  (Line.Last->is(tok::l_brace) || Style.BreakAfterJavaFieldAnnotations))
2926  return true;
2927 
2928  if (Right.is(TT_ProtoExtensionLSquare))
2929  return true;
2930 
2931  // In text proto instances if a submessage contains at least 2 entries and at
2932  // least one of them is a submessage, like A { ... B { ... } ... },
2933  // put all of the entries of A on separate lines by forcing the selector of
2934  // the submessage B to be put on a newline.
2935  //
2936  // Example: these can stay on one line:
2937  // a { scalar_1: 1 scalar_2: 2 }
2938  // a { b { key: value } }
2939  //
2940  // and these entries need to be on a new line even if putting them all in one
2941  // line is under the column limit:
2942  // a {
2943  // scalar: 1
2944  // b { key: value }
2945  // }
2946  //
2947  // We enforce this by breaking before a submessage field that has previous
2948  // siblings, *and* breaking before a field that follows a submessage field.
2949  //
2950  // Be careful to exclude the case [proto.ext] { ... } since the `]` is
2951  // the TT_SelectorName there, but we don't want to break inside the brackets.
2952  // We ensure elsewhere that extensions are always on their own line.
2955  Right.is(TT_SelectorName) && !Right.is(tok::r_square) && Right.Next) {
2956  // Look for the scope opener after selector in cases like:
2957  // selector { ...
2958  // selector: { ...
2959  FormatToken *LBrace =
2960  Right.Next->is(tok::colon) ? Right.Next->Next : Right.Next;
2961  if (LBrace &&
2962  // The scope opener is one of {, [, <:
2963  // selector { ... }
2964  // selector [ ... ]
2965  // selector < ... >
2966  //
2967  // In case of selector { ... }, the l_brace is TT_DictLiteral.
2968  // In case of an empty selector {}, the l_brace is not TT_DictLiteral,
2969  // so we check for immediately following r_brace.
2970  ((LBrace->is(tok::l_brace) &&
2971  (LBrace->is(TT_DictLiteral) ||
2972  (LBrace->Next && LBrace->Next->is(tok::r_brace)))) ||
2973  LBrace->is(TT_ArrayInitializerLSquare) || LBrace->is(tok::less))) {
2974  // If Left.ParameterCount is 0, then this submessage entry is not the
2975  // first in its parent submessage, and we want to break before this entry.
2976  // If Left.ParameterCount is greater than 0, then its parent submessage
2977  // might contain 1 or more entries and we want to break before this entry
2978  // if it contains at least 2 entries. We deal with this case later by
2979  // detecting and breaking before the next entry in the parent submessage.
2980  if (Left.ParameterCount == 0)
2981  return true;
2982  // However, if this submessage is the first entry in its parent
2983  // submessage, Left.ParameterCount might be 1 in some cases.
2984  // We deal with this case later by detecting an entry
2985  // following a closing paren of this submessage.
2986  }
2987 
2988  // If this is an entry immediately following a submessage, it will be
2989  // preceded by a closing paren of that submessage, like in:
2990  // left---. .---right
2991  // v v
2992  // sub: { ... } key: value
2993  // If there was a comment between `}` an `key` above, then `key` would be
2994  // put on a new line anyways.
2995  if (Left.isOneOf(tok::r_brace, tok::greater, tok::r_square))
2996  return true;
2997  }
2998 
2999  return false;
3000 }
3001 
3002 bool TokenAnnotator::canBreakBefore(const AnnotatedLine &Line,
3003  const FormatToken &Right) {
3004  const FormatToken &Left = *Right.Previous;
3005 
3006  // Language-specific stuff.
3008  if (Left.isOneOf(Keywords.kw_throws, Keywords.kw_extends,
3009  Keywords.kw_implements))
3010  return false;
3011  if (Right.isOneOf(Keywords.kw_throws, Keywords.kw_extends,
3012  Keywords.kw_implements))
3013  return true;
3014  } else if (Style.Language == FormatStyle::LK_JavaScript) {
3015  const FormatToken *NonComment = Right.getPreviousNonComment();
3016  if (NonComment &&
3017  NonComment->isOneOf(
3018  tok::kw_return, Keywords.kw_yield, tok::kw_continue, tok::kw_break,
3019  tok::kw_throw, Keywords.kw_interface, Keywords.kw_type,
3020  tok::kw_static, tok::kw_public, tok::kw_private, tok::kw_protected,
3021  Keywords.kw_readonly, Keywords.kw_abstract, Keywords.kw_get,
3022  Keywords.kw_set, Keywords.kw_async, Keywords.kw_await))
3023  return false; // Otherwise automatic semicolon insertion would trigger.
3024  if (Right.NestingLevel == 0 &&
3025  (Left.Tok.getIdentifierInfo() ||
3026  Left.isOneOf(tok::r_square, tok::r_paren)) &&
3027  Right.isOneOf(tok::l_square, tok::l_paren))
3028  return false; // Otherwise automatic semicolon insertion would trigger.
3029  if (Left.is(TT_JsFatArrow) && Right.is(tok::l_brace))
3030  return false;
3031  if (Left.is(TT_JsTypeColon))
3032  return true;
3033  // Don't wrap between ":" and "!" of a strict prop init ("field!: type;").
3034  if (Left.is(tok::exclaim) && Right.is(tok::colon))
3035  return false;
3036  if (Right.is(Keywords.kw_is))
3037  return false;
3038  if (Left.is(Keywords.kw_in))
3040  if (Right.is(Keywords.kw_in))
3042  if (Right.is(Keywords.kw_as))
3043  return false; // must not break before as in 'x as type' casts
3044  if (Left.is(Keywords.kw_as))
3045  return true;
3046  if (Left.is(TT_JsNonNullAssertion))
3047  return true;
3048  if (Left.is(Keywords.kw_declare) &&
3049  Right.isOneOf(Keywords.kw_module, tok::kw_namespace,
3050  Keywords.kw_function, tok::kw_class, tok::kw_enum,
3051  Keywords.kw_interface, Keywords.kw_type, Keywords.kw_var,
3052  Keywords.kw_let, tok::kw_const))
3053  // See grammar for 'declare' statements at:
3054  // https://github.com/Microsoft/TypeScript/blob/master/doc/spec.md#A.10
3055  return false;
3056  if (Left.isOneOf(Keywords.kw_module, tok::kw_namespace) &&
3057  Right.isOneOf(tok::identifier, tok::string_literal))
3058  return false; // must not break in "module foo { ...}"
3059  if (Right.is(TT_TemplateString) && Right.closesScope())
3060  return false;
3061  if (Left.is(TT_TemplateString) && Left.opensScope())
3062  return true;
3063  }
3064 
3065  if (Left.is(tok::at))
3066  return false;
3067  if (Left.Tok.getObjCKeywordID() == tok::objc_interface)
3068  return false;
3069  if (Left.isOneOf(TT_JavaAnnotation, TT_LeadingJavaAnnotation))
3070  return !Right.is(tok::l_paren);
3071  if (Right.is(TT_PointerOrReference))
3072  return Line.IsMultiVariableDeclStmt ||
3074  (!Right.Next || Right.Next->isNot(TT_FunctionDeclarationName)));
3075  if (Right.isOneOf(TT_StartOfName, TT_FunctionDeclarationName) ||
3076  Right.is(tok::kw_operator))
3077  return true;
3078  if (Left.is(TT_PointerOrReference))
3079  return false;
3080  if (Right.isTrailingComment())
3081  // We rely on MustBreakBefore being set correctly here as we should not
3082  // change the "binding" behavior of a comment.
3083  // The first comment in a braced lists is always interpreted as belonging to
3084  // the first list element. Otherwise, it should be placed outside of the
3085  // list.
3086  return Left.BlockKind == BK_BracedInit ||
3087  (Left.is(TT_CtorInitializerColon) &&
3089  if (Left.is(tok::question) && Right.is(tok::colon))
3090  return false;
3091  if (Right.is(TT_ConditionalExpr) || Right.is(tok::question))
3093  if (Left.is(TT_ConditionalExpr) || Left.is(tok::question))
3095  if (Left.is(TT_InheritanceColon))
3097  if (Right.is(TT_InheritanceColon))
3099  if (Right.is(TT_ObjCMethodExpr) && !Right.is(tok::r_square) &&
3100  Left.isNot(TT_SelectorName))
3101  return true;
3102 
3103  if (Right.is(tok::colon) &&
3104  !Right.isOneOf(TT_CtorInitializerColon, TT_InlineASMColon))
3105  return false;
3106  if (Left.is(tok::colon) && Left.isOneOf(TT_DictLiteral, TT_ObjCMethodExpr)) {
3110  return false;
3111  // Prevent cases like:
3112  //
3113  // submessage:
3114  // { key: valueeeeeeeeeeee }
3115  //
3116  // when the snippet does not fit into one line.
3117  // Prefer:
3118  //
3119  // submessage: {
3120  // key: valueeeeeeeeeeee
3121  // }
3122  //
3123  // instead, even if it is longer by one line.
3124  //
3125  // Note that this allows allows the "{" to go over the column limit
3126  // when the column limit is just between ":" and "{", but that does
3127  // not happen too often and alternative formattings in this case are
3128  // not much better.
3129  //
3130  // The code covers the cases:
3131  //
3132  // submessage: { ... }
3133  // submessage: < ... >
3134  // repeated: [ ... ]
3135  if (((Right.is(tok::l_brace) || Right.is(tok::less)) &&
3136  Right.is(TT_DictLiteral)) ||
3137  Right.is(TT_ArrayInitializerLSquare))
3138  return false;
3139  }
3140  return true;
3141  }
3142  if (Right.is(tok::r_square) && Right.MatchingParen &&
3143  Right.MatchingParen->is(TT_ProtoExtensionLSquare))
3144  return false;
3145  if (Right.is(TT_SelectorName) || (Right.is(tok::identifier) && Right.Next &&
3146  Right.Next->is(TT_ObjCMethodExpr)))
3147  return Left.isNot(tok::period); // FIXME: Properly parse ObjC calls.
3148  if (Left.is(tok::r_paren) && Line.Type == LT_ObjCProperty)
3149  return true;
3150  if (Left.ClosesTemplateDeclaration || Left.is(TT_FunctionAnnotationRParen))
3151  return true;
3152  if (Right.isOneOf(TT_RangeBasedForLoopColon, TT_OverloadedOperatorLParen,
3153  TT_OverloadedOperator))
3154  return false;
3155  if (Left.is(TT_RangeBasedForLoopColon))
3156  return true;
3157  if (Right.is(TT_RangeBasedForLoopColon))
3158  return false;
3159  if (Left.is(TT_TemplateCloser) && Right.is(TT_TemplateOpener))
3160  return true;
3161  if (Left.isOneOf(TT_TemplateCloser, TT_UnaryOperator) ||
3162  Left.is(tok::kw_operator))
3163  return false;
3164  if (Left.is(tok::equal) && !Right.isOneOf(tok::kw_default, tok::kw_delete) &&
3165  Line.Type == LT_VirtualFunctionDecl && Left.NestingLevel == 0)
3166  return false;
3167  if (Left.is(tok::equal) && Right.is(tok::l_brace) &&
3169  return false;
3170  if (Left.is(tok::l_paren) && Left.is(TT_AttributeParen))
3171  return false;
3172  if (Left.is(tok::l_paren) && Left.Previous &&
3173  (Left.Previous->isOneOf(TT_BinaryOperator, TT_CastRParen)))
3174  return false;
3175  if (Right.is(TT_ImplicitStringLiteral))
3176  return false;
3177 
3178  if (Right.is(tok::r_paren) || Right.is(TT_TemplateCloser))
3179  return false;
3180  if (Right.is(tok::r_square) && Right.MatchingParen &&
3181  Right.MatchingParen->is(TT_LambdaLSquare))
3182  return false;
3183 
3184  // We only break before r_brace if there was a corresponding break before
3185  // the l_brace, which is tracked by BreakBeforeClosingBrace.
3186  if (Right.is(tok::r_brace))
3187  return Right.MatchingParen && Right.MatchingParen->BlockKind == BK_Block;
3188 
3189  // Allow breaking after a trailing annotation, e.g. after a method
3190  // declaration.
3191  if (Left.is(TT_TrailingAnnotation))
3192  return !Right.isOneOf(tok::l_brace, tok::semi, tok::equal, tok::l_paren,
3193  tok::less, tok::coloncolon);
3194 
3195  if (Right.is(tok::kw___attribute) ||
3196  (Right.is(tok::l_square) && Right.is(TT_AttributeSquare)))
3197  return true;
3198 
3199  if (Left.is(tok::identifier) && Right.is(tok::string_literal))
3200  return true;
3201 
3202  if (Right.is(tok::identifier) && Right.Next && Right.Next->is(TT_DictLiteral))
3203  return true;
3204 
3205  if (Left.is(TT_CtorInitializerColon))
3207  if (Right.is(TT_CtorInitializerColon))
3209  if (Left.is(TT_CtorInitializerComma) &&
3211  return false;
3212  if (Right.is(TT_CtorInitializerComma) &&
3214  return true;
3215  if (Left.is(TT_InheritanceComma) &&
3217  return false;
3218  if (Right.is(TT_InheritanceComma) &&
3220  return true;
3221  if ((Left.is(tok::greater) && Right.is(tok::greater)) ||
3222  (Left.is(tok::less) && Right.is(tok::less)))
3223  return false;
3224  if (Right.is(TT_BinaryOperator) &&
3227  Right.getPrecedence() != prec::Assignment))
3228  return true;
3229  if (Left.is(TT_ArrayInitializerLSquare))
3230  return true;
3231  if (Right.is(tok::kw_typename) && Left.isNot(tok::kw_const))
3232  return true;
3233  if ((Left.isBinaryOperator() || Left.is(TT_BinaryOperator)) &&
3234  !Left.isOneOf(tok::arrowstar, tok::lessless) &&
3237  Left.getPrecedence() == prec::Assignment))
3238  return true;
3239  if ((Left.is(TT_AttributeSquare) && Right.is(tok::l_square)) ||
3240  (Left.is(tok::r_square) && Right.is(TT_AttributeSquare)))
3241  return false;
3242  return Left.isOneOf(tok::comma, tok::coloncolon, tok::semi, tok::l_brace,
3243  tok::kw_class, tok::kw_struct, tok::comment) ||
3244  Right.isMemberAccess() ||
3245  Right.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow, tok::lessless,
3246  tok::colon, tok::l_square, tok::at) ||
3247  (Left.is(tok::r_paren) &&
3248  Right.isOneOf(tok::identifier, tok::kw_const)) ||
3249  (Left.is(tok::l_paren) && !Right.is(tok::r_paren)) ||
3250  (Left.is(TT_TemplateOpener) && !Right.is(TT_TemplateCloser));
3251 }
3252 
3253 void TokenAnnotator::printDebugInfo(const AnnotatedLine &Line) {
3254  llvm::errs() << "AnnotatedTokens(L=" << Line.Level << "):\n";
3255  const FormatToken *Tok = Line.First;
3256  while (Tok) {
3257  llvm::errs() << " M=" << Tok->MustBreakBefore
3258  << " C=" << Tok->CanBreakBefore
3259  << " T=" << getTokenTypeName(Tok->Type)
3260  << " S=" << Tok->SpacesRequiredBefore
3261  << " B=" << Tok->BlockParameterCount
3262  << " BK=" << Tok->BlockKind << " P=" << Tok->SplitPenalty
3263  << " Name=" << Tok->Tok.getName() << " L=" << Tok->TotalLength
3264  << " PPK=" << Tok->PackingKind << " FakeLParens=";
3265  for (unsigned i = 0, e = Tok->FakeLParens.size(); i != e; ++i)
3266  llvm::errs() << Tok->FakeLParens[i] << "/";
3267  llvm::errs() << " FakeRParens=" << Tok->FakeRParens;
3268  llvm::errs() << " II=" << Tok->Tok.getIdentifierInfo();
3269  llvm::errs() << " Text='" << Tok->TokenText << "'\n";
3270  if (!Tok->Next)
3271  assert(Tok == Line.Last);
3272  Tok = Tok->Next;
3273  }
3274  llvm::errs() << "----\n";
3275 }
3276 
3277 } // namespace format
3278 } // namespace clang
Always break after template declaration.
Definition: Format.h:387
Always break after the return type of top-level definitions.
Definition: Format.h:330
bool endsSequence(A K1, Ts... Tokens) const
true if this token ends a sequence with the given tokens in order, following the Previous pointers...
Definition: FormatToken.h:332
if(T->getSizeExpr()) TRY_TO(TraverseStmt(T -> getSizeExpr()))
unsigned NestingLevel
The nesting level of this token, i.e.
Definition: FormatToken.h:231
Break inheritance list after the colon and commas.
Definition: Format.h:922
bool AlwaysBreakBeforeMultilineStrings
If true, always break before multiline string literals.
Definition: Format.h:352
unsigned PenaltyBreakBeforeFirstCallParameter
The penalty for breaking a function call after call(.
Definition: Format.h:1338
Token Tok
The Token.
Definition: FormatToken.h:127
Should be used for Protocol Buffers (https://developers.google.com/protocol-buffers/).
Definition: Format.h:1191
bool is(tok::TokenKind K) const
is/isNot - Predicates to check if this token is a specific kind, as in "if (Tok.is(tok::l_brace)) {...
Definition: Token.h:95
Defines the SourceManager interface.
bool SpaceBeforeRangeBasedForLoopColon
If false, spaces will be removed before range-based for loop colon.
Definition: Format.h:1570
std::unique_ptr< TokenRole > Role
A token can have a special role that can carry extra information about the token&#39;s formatting...
Definition: FormatToken.h:207
Break constructor initializers before the colon and commas, and align the commas with the colon...
Definition: Format.h:855
unsigned OriginalColumn
The original 0-based column of this token, including expanded tabs.
Definition: FormatToken.h:218
static LLVM_ATTRIBUTE_UNUSED void printDebugInfo(const UnwrappedLine &Line, StringRef Prefix="")
bool isMemberAccess() const
Returns true if this is a "." or "->" accessing a member.
Definition: FormatToken.h:380
bool isFunctionLikeKeyword() const
Returns true if this is a keyword that can be used like a function call (e.g.
Definition: FormatToken.h:414
const FormatToken * getNextNonComment() const
Returns the next token ignoring comments.
Definition: FormatToken.h:470
Parser - This implements a parser for the C family of languages.
Definition: Parser.h:56
bool SpaceAfterTemplateKeyword
If true, a space will be inserted after the &#39;template&#39; keyword.
Definition: Format.h:1490
PointerAlignmentStyle PointerAlignment
Pointer and reference alignment style.
Definition: Format.h:1379
Align pointer to the left.
Definition: Format.h:1365
Should be used for C, C++.
Definition: Format.h:1182
bool IsMultiline
Whether the token text contains newlines (escaped or not).
Definition: FormatToken.h:156
bool AfterEnum
Wrap enum definitions.
Definition: Format.h:664
unsigned TotalLength
The total length of the unwrapped line up to and including this token.
Definition: FormatToken.h:214
bool isBinaryOperator() const
Definition: FormatToken.h:402
unsigned NewlinesBefore
The number of newlines immediately before the Token.
Definition: FormatToken.h:133
FormatToken * Next
The next token in the unwrapped line.
Definition: FormatToken.h:290
Break after operators.
Definition: Format.h:449
bool isLiteral() const
Return true if this is a "literal", like a numeric constant, string, etc.
Definition: Token.h:113
bool CanBeExpression
tok::TokenKind ContextKind
unsigned UnbreakableTailLength
The length of following tokens until the next natural split point, or the next token that can be brok...
Definition: FormatToken.h:222
bool closesScope() const
Returns whether Tok is )]} or a closing > of a template or in protos.
Definition: FormatToken.h:370
bool JavaScriptWrapImports
Whether to wrap JavaScript import/export statements.
Definition: Format.h:1161
Always break after the return type.
Definition: Format.h:290
unsigned SplitPenalty
Penalty for inserting a line break before this token.
Definition: FormatToken.h:237
bool ColonIsForRangeExpr
prec::Level getPrecedence() const
Definition: FormatToken.h:456
unsigned ParameterCount
Number of parameters, if this is "(", "[" or "<".
Definition: FormatToken.h:195
unsigned FakeRParens
Insert this many fake ) after this token for correct indentation.
Definition: FormatToken.h:257
bool SpaceInEmptyParentheses
If true, spaces may be inserted into ().
Definition: Format.h:1582
bool CanBreakBefore
true if it is allowed to break before this token.
Definition: FormatToken.h:185
Should be used for Java.
Definition: Format.h:1184
FormatToken * Previous
The previous token in the unwrapped line.
Definition: FormatToken.h:287
Always break after the return type of function definitions.
Definition: Format.h:318
Only merge functions defined inside a class.
Definition: Format.h:209
This file implements a token annotator, i.e.
bool BreakAfterJavaFieldAnnotations
Break after each annotation on a field in Java files.
Definition: Format.h:875
bool ConstructorInitializerAllOnOneLineOrOnePerLine
If the constructor initializers don&#39;t fit on a line, put each initializer on its own line...
Definition: Format.h:967
unsigned OperatorIndex
Is this is an operator (or "."/"->") in a sequence of operators with the same precedence, contains the 0-based operator index.
Definition: FormatToken.h:267
unsigned SpacesRequiredBefore
The number of spaces that should be inserted before this token.
Definition: FormatToken.h:182
bool isNot(T Kind) const
Definition: FormatToken.h:320
bool SpacesInParentheses
If true, spaces will be inserted after ( and before ).
Definition: Format.h:1630
const FormatToken & Tok
ReturnTypeBreakingStyle AlwaysBreakAfterReturnType
The function declaration return type breaking style to use.
Definition: Format.h:338
unsigned BlockParameterCount
Number of parameters that are nested blocks, if this is "(", "[" or "<".
Definition: FormatToken.h:199
bool InInheritanceList
bool SpaceBeforeAssignmentOperators
If false, spaces will be removed before assignment operators.
Definition: Format.h:1498
SpaceBeforeParensOptions SpaceBeforeParens
Defines in which cases to put a space before opening parentheses.
Definition: Format.h:1562
bool SpaceBeforeCtorInitializerColon
If false, spaces will be removed before constructor initializer colon.
Definition: Format.h:1517
unsigned PenaltyBreakAssignment
The penalty for breaking around an assignment operator.
Definition: Format.h:1335
const char * getName() const
Definition: Token.h:166
Always put a space before opening parentheses, except when it&#39;s prohibited by the syntax rules (in fu...
Definition: Format.h:1558
bool mightBeFunctionDefinition() const
true if this line looks like a function definition instead of a function declaration.
unsigned ColumnLimit
The column limit.
Definition: Format.h:885
Never merge functions into a single line.
Definition: Format.h:195
FormatToken * getPreviousNonComment() const
Returns the previous token ignoring comments.
Definition: FormatToken.h:462
AnnotatingParser & P
BracketAlignmentStyle AlignAfterOpenBracket
If true, horizontally aligns arguments after an open bracket.
Definition: Format.h:79
NodeId Parent
Definition: ASTDiff.cpp:192
bool isLabelString() const
Returns true if this is a string literal that&#39;s like a label, e.g.
Definition: FormatToken.h:434
bool isOneOf(A K1, B K2) const
Definition: FormatToken.h:313
Break after return type automatically.
Definition: Format.h:274
Only merge empty functions.
Definition: Format.h:217
Should be used for JavaScript.
Definition: Format.h:1186
const AnnotatedLine * Line
StateNode * Previous
SmallVector< AnnotatedLine *, 0 > Children
ParameterPackingKind PackingKind
If this is an opening parenthesis, how are the parameters packed?
Definition: FormatToken.h:210
bool SpacesInContainerLiterals
If true, spaces are inserted inside container literals (e.g.
Definition: Format.h:1616
BraceWrappingFlags BraceWrapping
Control of individual brace wrapping cases.
Definition: Format.h:823
bool SpacesInAngles
If true, spaces will be inserted after < and before > in template argument lists. ...
Definition: Format.h:1607
A wrapper around a Token storing information about the whitespace characters preceding it...
Definition: FormatToken.h:123
void setCommentLineLevels(SmallVectorImpl< AnnotatedLine *> &Lines)
Adapts the indent levels of comment lines to the indent of the subsequent line.
static unsigned maxNestingDepth(const AnnotatedLine &Line)
FormatToken * FirstStartOfName
SourceLocation getEnd() const
bool isTrailingComment() const
Definition: FormatToken.h:407
void annotate(AnnotatedLine &Line)
Don&#39;t align, instead use ContinuationIndentWidth, e.g.
Definition: Format.h:65
const char * getTokenTypeName(TokenType Type)
Determines the name of a token type.
Definition: FormatToken.cpp:25
static bool isAllmanBrace(const FormatToken &Tok)
bool is(tok::TokenKind Kind) const
Definition: FormatToken.h:304
bool InCpp11AttributeSpecifier
IdentifierInfo * getIdentifierInfo() const
Definition: Token.h:177
Break constructor initializers after the colon and commas.
Definition: Format.h:862
bool BreakBeforeTernaryOperators
If true, ternary operators will be placed after line breaks.
Definition: Format.h:837
SourceRange WhitespaceRange
The range of the whitespace immediately preceding the Token.
Definition: FormatToken.h:140
TokenKind
Provides a simple uniform namespace for tokens from all C languages.
Definition: TokenKinds.h:25
bool SpacesInCStyleCastParentheses
If true, spaces may be inserted into C style casts.
Definition: Format.h:1623
bool IsUnterminatedLiteral
Set to true if this token is an unterminated literal.
Definition: FormatToken.h:174
bool startsWith(Ts... Tokens) const
true if this line starts with the given tokens in order, ignoring comments.
bool SpacesInSquareBrackets
If true, spaces will be inserted after [ and before ].
Definition: Format.h:1639
StringRef TokenText
The raw text of the token.
Definition: FormatToken.h:171
SmallVector< prec::Level, 4 > FakeLParens
Stores the number of required fake parentheses and the corresponding operator precedence.
Definition: FormatToken.h:255
tok::ObjCKeywordKind getObjCKeywordID() const
Return the ObjC keyword kind.
Definition: Lexer.cpp:67
BreakTemplateDeclarationsStyle AlwaysBreakTemplateDeclarations
The template declaration breaking style to use.
Definition: Format.h:391
bool isSimpleTypeSpecifier() const
Determine whether the token is a simple-type-specifier.
Definition: FormatToken.cpp:40
The FormatStyle is used to configure the formatting to follow specific guidelines.
Definition: Format.h:48
Never put a space before opening parentheses.
Definition: Format.h:1536
unsigned PenaltyReturnTypeOnItsOwnLine
Penalty for putting the return type of a function onto its own line.
Definition: Format.h:1357
unsigned IndentLevel
The indent level of this token. Copied from the surrounding line.
Definition: FormatToken.h:234
LanguageKind Language
Language, this format style is targeted at.
Definition: Format.h:1201
bool ColonIsObjCMethodExpr
Dataflow Directional Tag Classes.
ShortFunctionStyle AllowShortFunctionsOnASingleLine
Dependent on the value, int f() { return 0; } can be put on a single line.
Definition: Format.h:241
unsigned ColumnWidth
The width of the non-whitespace parts of the token (or its first line for multi-line tokens) in colum...
Definition: FormatToken.h:149
BreakInheritanceListStyle BreakInheritanceList
The inheritance list style to use.
Definition: Format.h:926
Use C++03-compatible syntax.
Definition: Format.h:1644
Always break after the return types of top-level functions.
Definition: Format.h:303
bool IsExpression
Use features of C++11, C++14 and C++1z (e.g.
Definition: Format.h:1647
bool Cpp11BracedListStyle
If true, format braced lists as best suited for C++11 braced lists.
Definition: Format.h:1003
static bool isFunctionDeclarationName(const FormatToken &Current, const AnnotatedLine &Line)
bool SpaceBeforeCpp11BracedList
If true, a space will be inserted before a C++11 braced list used to initialize an object (after the ...
Definition: Format.h:1509
bool AfterStruct
Wrap struct definitions.
Definition: Format.h:714
Break inheritance list before the colon and commas, and align the commas with the colon...
Definition: Format.h:914
void calculateFormattingInformation(AnnotatedLine &Line)
bool SpaceAfterCStyleCast
If true, a space is inserted after C style casts.
Definition: Format.h:1483
Should be used for Objective-C, Objective-C++.
Definition: Format.h:1188
LanguageStandard Standard
Format compatible with this standard, e.g.
Definition: Format.h:1654
FormatToken * NextOperator
If this is an operator (or "."/"->") in a sequence of operators with the same precedence, points to the next operator.
Definition: FormatToken.h:271
BreakConstructorInitializersStyle BreakConstructorInitializers
The constructor initializers style to use.
Definition: Format.h:866
unsigned BindingStrength
bool ClosesTemplateDeclaration
true if this is the ">" of "template<..>".
Definition: FormatToken.h:188
FormatToken * MatchingParen
If this is a bracket, this points to the matching one.
Definition: FormatToken.h:284
SmallVector< AnnotatedLine *, 1 > Children
If this token starts a block, this contains all the unwrapped lines in it.
Definition: FormatToken.h:294
FormatToken * FirstObjCSelectorName
bool InTemplateArgument
bool ColonIsDictLiteral
The parameter type of a method or function.
BinaryOperatorStyle BreakBeforeBinaryOperators
The way to wrap binary operators.
Definition: Format.h:477
bool opensBlockOrBlockTypeList(const FormatStyle &Style) const
Returns true if this tokens starts a block-type list, i.e.
Definition: FormatToken.h:479
Break before operators.
Definition: Format.h:473
bool ObjCSpaceBeforeProtocolList
Add a space in front of an Objective-C protocol list, i.e.
Definition: Format.h:1332
unsigned SpacesBeforeTrailingComments
The number of spaces before trailing line comments (// - comments).
Definition: Format.h:1598
bool opensScope() const
Returns whether Tok is ([{ or an opening < of a template or in protos.
Definition: FormatToken.h:360
bool SpaceBeforeInheritanceColon
If false, spaces will be removed before inheritance colon.
Definition: Format.h:1524
bool AllowShortBlocksOnASingleLine
Allows contracting simple braced statements to a single line.
Definition: Format.h:176
unsigned LongestObjCSelectorName
bool MustBreakBefore
Whether there must be a line break before this token.
Definition: FormatToken.h:165
unsigned PenaltyBreakTemplateDeclaration
The penalty for breaking after template declaration.
Definition: Format.h:1350
__DEVICE__ int max(int __a, int __b)
bool AfterClass
Wrap class definitions.
Definition: Format.h:634
Should be used for Protocol Buffer messages in text format (https://developers.google.com/protocol-buffers/).
Definition: Format.h:1196
prec::Level getBinOpPrecedence(tok::TokenKind Kind, bool GreaterThanIsOperator, bool CPlusPlus11)
Return the precedence of the specified binary operator token.
Align pointer to the right.
Definition: Format.h:1370
unsigned BindingStrength
The binding strength of a token.
Definition: FormatToken.h:227
bool CaretFound
bool InCtorInitializer
bool isStringLiteral() const
Definition: FormatToken.h:336
SourceLocation getBegin() const
unsigned MaxEmptyLinesToKeep
The maximum number of consecutive empty lines to keep.
Definition: Format.h:1245
bool HasUnescapedNewline
Whether there is at least one unescaped newline before the Token.
Definition: FormatToken.h:137
BraceBlockKind BlockKind
Contains the kind of block if this token is a brace.
Definition: FormatToken.h:177
bool PartOfMultiVariableDeclStmt
Is this token part of a DeclStmt defining multiple variables?
Definition: FormatToken.h:276
bool IsForEachMacro
const FormatStyle & Style