1 | //===--- TokenAnnotator.cpp - Format C++ code -----------------------------===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | /// |
9 | /// \file |
10 | /// This file implements a token annotator, i.e. creates |
11 | /// \c AnnotatedTokens out of \c FormatTokens with required extra information. |
12 | /// |
13 | //===----------------------------------------------------------------------===// |
14 | |
15 | #include "TokenAnnotator.h" |
16 | #include "FormatToken.h" |
17 | #include "clang/Basic/SourceManager.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 | if (Left->Previous && Left->Previous->is(TT_AttributeSquare)) |
303 | CurrentToken->Type = TT_AttributeSquare; |
304 | |
305 | if (!HasMultipleLines) |
306 | Left->PackingKind = PPK_Inconclusive; |
307 | else if (HasMultipleParametersOnALine) |
308 | Left->PackingKind = PPK_BinPacked; |
309 | else |
310 | Left->PackingKind = PPK_OnePerLine; |
311 | |
312 | next(); |
313 | return true; |
314 | } |
315 | if (CurrentToken->isOneOf(tok::r_square, tok::r_brace)) |
316 | return false; |
317 | |
318 | if (CurrentToken->is(tok::l_brace)) |
319 | Left->Type = TT_Unknown; // Not TT_ObjCBlockLParen |
320 | if (CurrentToken->is(tok::comma) && CurrentToken->Next && |
321 | !CurrentToken->Next->HasUnescapedNewline && |
322 | !CurrentToken->Next->isTrailingComment()) |
323 | HasMultipleParametersOnALine = true; |
324 | if ((CurrentToken->Previous->isOneOf(tok::kw_const, tok::kw_auto) || |
325 | CurrentToken->Previous->isSimpleTypeSpecifier()) && |
326 | !CurrentToken->is(tok::l_brace)) |
327 | Contexts.back().IsExpression = false; |
328 | if (CurrentToken->isOneOf(tok::semi, tok::colon)) { |
329 | MightBeObjCForRangeLoop = false; |
330 | if (PossibleObjCForInToken) { |
331 | PossibleObjCForInToken->Type = TT_Unknown; |
332 | PossibleObjCForInToken = nullptr; |
333 | } |
334 | } |
335 | if (MightBeObjCForRangeLoop && CurrentToken->is(Keywords.kw_in)) { |
336 | PossibleObjCForInToken = CurrentToken; |
337 | PossibleObjCForInToken->Type = TT_ObjCForIn; |
338 | } |
339 | // When we discover a 'new', we set CanBeExpression to 'false' in order to |
340 | // parse the type correctly. Reset that after a comma. |
341 | if (CurrentToken->is(tok::comma)) |
342 | Contexts.back().CanBeExpression = true; |
343 | |
344 | FormatToken *Tok = CurrentToken; |
345 | if (!consumeToken()) |
346 | return false; |
347 | updateParameterCount(Left, Tok); |
348 | if (CurrentToken && CurrentToken->HasUnescapedNewline) |
349 | HasMultipleLines = true; |
350 | } |
351 | return false; |
352 | } |
353 | |
354 | bool isCSharpAttributeSpecifier(const FormatToken &Tok) { |
355 | if (!Style.isCSharp()) |
356 | return false; |
357 | |
358 | const FormatToken *AttrTok = Tok.Next; |
359 | if (!AttrTok) |
360 | return false; |
361 | |
362 | // Just an empty declaration e.g. string []. |
363 | if (AttrTok->is(tok::r_square)) |
364 | return false; |
365 | |
366 | // Move along the tokens inbetween the '[' and ']' e.g. [STAThread]. |
367 | while (AttrTok && AttrTok->isNot(tok::r_square)) { |
368 | AttrTok = AttrTok->Next; |
369 | } |
370 | |
371 | if (!AttrTok) |
372 | return false; |
373 | |
374 | // Move past the end of ']'. |
375 | AttrTok = AttrTok->Next; |
376 | if (!AttrTok) |
377 | return false; |
378 | |
379 | // Limit this to being an access modifier that follows. |
380 | if (AttrTok->isOneOf(tok::kw_public, tok::kw_private, tok::kw_protected, |
381 | tok::kw_class, tok::kw_static, tok::l_square, |
382 | Keywords.kw_internal)) { |
383 | return true; |
384 | } |
385 | return false; |
386 | } |
387 | |
388 | bool isCpp11AttributeSpecifier(const FormatToken &Tok) { |
389 | if (!Style.isCpp() || !Tok.startsSequence(tok::l_square, tok::l_square)) |
390 | return false; |
391 | const FormatToken *AttrTok = Tok.Next->Next; |
392 | if (!AttrTok) |
393 | return false; |
394 | // C++17 '[[using ns: foo, bar(baz, blech)]]' |
395 | // We assume nobody will name an ObjC variable 'using'. |
396 | if (AttrTok->startsSequence(tok::kw_using, tok::identifier, tok::colon)) |
397 | return true; |
398 | if (AttrTok->isNot(tok::identifier)) |
399 | return false; |
400 | while (AttrTok && !AttrTok->startsSequence(tok::r_square, tok::r_square)) { |
401 | // ObjC message send. We assume nobody will use : in a C++11 attribute |
402 | // specifier parameter, although this is technically valid: |
403 | // [[foo(:)]] |
404 | if (AttrTok->is(tok::colon) || |
405 | AttrTok->startsSequence(tok::identifier, tok::identifier) || |
406 | AttrTok->startsSequence(tok::r_paren, tok::identifier)) |
407 | return false; |
408 | if (AttrTok->is(tok::ellipsis)) |
409 | return true; |
410 | AttrTok = AttrTok->Next; |
411 | } |
412 | return AttrTok && AttrTok->startsSequence(tok::r_square, tok::r_square); |
413 | } |
414 | |
415 | bool parseSquare() { |
416 | if (!CurrentToken) |
417 | return false; |
418 | |
419 | // A '[' could be an index subscript (after an identifier or after |
420 | // ')' or ']'), it could be the start of an Objective-C method |
421 | // expression, it could the start of an Objective-C array literal, |
422 | // or it could be a C++ attribute specifier [[foo::bar]]. |
423 | FormatToken *Left = CurrentToken->Previous; |
424 | Left->ParentBracket = Contexts.back().ContextKind; |
425 | FormatToken *Parent = Left->getPreviousNonComment(); |
426 | |
427 | // Cases where '>' is followed by '['. |
428 | // In C++, this can happen either in array of templates (foo<int>[10]) |
429 | // or when array is a nested template type (unique_ptr<type1<type2>[]>). |
430 | bool CppArrayTemplates = |
431 | Style.isCpp() && Parent && Parent->is(TT_TemplateCloser) && |
432 | (Contexts.back().CanBeExpression || Contexts.back().IsExpression || |
433 | Contexts.back().InTemplateArgument); |
434 | |
435 | bool IsCpp11AttributeSpecifier = isCpp11AttributeSpecifier(*Left) || |
436 | Contexts.back().InCpp11AttributeSpecifier; |
437 | |
438 | // Treat C# Attributes [STAThread] much like C++ attributes [[...]]. |
439 | bool IsCSharp11AttributeSpecifier = |
440 | isCSharpAttributeSpecifier(*Left) || |
441 | Contexts.back().InCSharpAttributeSpecifier; |
442 | |
443 | bool InsideInlineASM = Line.startsWith(tok::kw_asm); |
444 | bool IsCppStructuredBinding = Left->isCppStructuredBinding(Style); |
445 | bool StartsObjCMethodExpr = |
446 | !IsCppStructuredBinding && !InsideInlineASM && !CppArrayTemplates && |
447 | Style.isCpp() && !IsCpp11AttributeSpecifier && |
448 | Contexts.back().CanBeExpression && Left->isNot(TT_LambdaLSquare) && |
449 | !CurrentToken->isOneOf(tok::l_brace, tok::r_square) && |
450 | (!Parent || |
451 | Parent->isOneOf(tok::colon, tok::l_square, tok::l_paren, |
452 | tok::kw_return, tok::kw_throw) || |
453 | Parent->isUnaryOperator() || |
454 | // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen. |
455 | Parent->isOneOf(TT_ObjCForIn, TT_CastRParen) || |
456 | (getBinOpPrecedence(Parent->Tok.getKind(), true, true) > |
457 | prec::Unknown)); |
458 | bool ColonFound = false; |
459 | |
460 | unsigned BindingIncrease = 1; |
461 | if (IsCppStructuredBinding) { |
462 | Left->Type = TT_StructuredBindingLSquare; |
463 | } else if (Left->is(TT_Unknown)) { |
464 | if (StartsObjCMethodExpr) { |
465 | Left->Type = TT_ObjCMethodExpr; |
466 | } else if (IsCpp11AttributeSpecifier) { |
467 | Left->Type = TT_AttributeSquare; |
468 | } else if (Style.Language == FormatStyle::LK_JavaScript && Parent && |
469 | Contexts.back().ContextKind == tok::l_brace && |
470 | Parent->isOneOf(tok::l_brace, tok::comma)) { |
471 | Left->Type = TT_JsComputedPropertyName; |
472 | } else if (Style.isCpp() && Contexts.back().ContextKind == tok::l_brace && |
473 | Parent && Parent->isOneOf(tok::l_brace, tok::comma)) { |
474 | Left->Type = TT_DesignatedInitializerLSquare; |
475 | } else if (CurrentToken->is(tok::r_square) && Parent && |
476 | Parent->is(TT_TemplateCloser)) { |
477 | Left->Type = TT_ArraySubscriptLSquare; |
478 | } else if (Style.Language == FormatStyle::LK_Proto || |
479 | Style.Language == FormatStyle::LK_TextProto) { |
480 | // Square braces in LK_Proto can either be message field attributes: |
481 | // |
482 | // optional Aaa aaa = 1 [ |
483 | // (aaa) = aaa |
484 | // ]; |
485 | // |
486 | // extensions 123 [ |
487 | // (aaa) = aaa |
488 | // ]; |
489 | // |
490 | // or text proto extensions (in options): |
491 | // |
492 | // option (Aaa.options) = { |
493 | // [type.type/type] { |
494 | // key: value |
495 | // } |
496 | // } |
497 | // |
498 | // or repeated fields (in options): |
499 | // |
500 | // option (Aaa.options) = { |
501 | // keys: [ 1, 2, 3 ] |
502 | // } |
503 | // |
504 | // In the first and the third case we want to spread the contents inside |
505 | // the square braces; in the second we want to keep them inline. |
506 | Left->Type = TT_ArrayInitializerLSquare; |
507 | if (!Left->endsSequence(tok::l_square, tok::numeric_constant, |
508 | tok::equal) && |
509 | !Left->endsSequence(tok::l_square, tok::numeric_constant, |
510 | tok::identifier) && |
511 | !Left->endsSequence(tok::l_square, tok::colon, TT_SelectorName)) { |
512 | Left->Type = TT_ProtoExtensionLSquare; |
513 | BindingIncrease = 10; |
514 | } |
515 | } else if (!CppArrayTemplates && Parent && |
516 | Parent->isOneOf(TT_BinaryOperator, TT_TemplateCloser, tok::at, |
517 | tok::comma, tok::l_paren, tok::l_square, |
518 | tok::question, tok::colon, tok::kw_return, |
519 | // Should only be relevant to JavaScript: |
520 | tok::kw_default)) { |
521 | Left->Type = TT_ArrayInitializerLSquare; |
522 | } else if (IsCSharp11AttributeSpecifier) { |
523 | Left->Type = TT_AttributeSquare; |
524 | } else { |
525 | BindingIncrease = 10; |
526 | Left->Type = TT_ArraySubscriptLSquare; |
527 | } |
528 | } |
529 | |
530 | ScopedContextCreator ContextCreator(*this, tok::l_square, BindingIncrease); |
531 | Contexts.back().IsExpression = true; |
532 | if (Style.Language == FormatStyle::LK_JavaScript && Parent && |
533 | Parent->is(TT_JsTypeColon)) |
534 | Contexts.back().IsExpression = false; |
535 | |
536 | Contexts.back().ColonIsObjCMethodExpr = StartsObjCMethodExpr; |
537 | Contexts.back().InCpp11AttributeSpecifier = IsCpp11AttributeSpecifier; |
538 | Contexts.back().InCSharpAttributeSpecifier = IsCSharp11AttributeSpecifier; |
539 | |
540 | while (CurrentToken) { |
541 | if (CurrentToken->is(tok::r_square)) { |
542 | if (IsCpp11AttributeSpecifier) |
543 | CurrentToken->Type = TT_AttributeSquare; |
544 | if (IsCSharp11AttributeSpecifier) |
545 | CurrentToken->Type = TT_AttributeSquare; |
546 | else if (((CurrentToken->Next && |
547 | CurrentToken->Next->is(tok::l_paren)) || |
548 | (CurrentToken->Previous && |
549 | CurrentToken->Previous->Previous == Left)) && |
550 | Left->is(TT_ObjCMethodExpr)) { |
551 | // An ObjC method call is rarely followed by an open parenthesis. It |
552 | // also can't be composed of just one token, unless it's a macro that |
553 | // will be expanded to more tokens. |
554 | // FIXME: Do we incorrectly label ":" with this? |
555 | StartsObjCMethodExpr = false; |
556 | Left->Type = TT_Unknown; |
557 | } |
558 | if (StartsObjCMethodExpr && CurrentToken->Previous != Left) { |
559 | CurrentToken->Type = TT_ObjCMethodExpr; |
560 | // If we haven't seen a colon yet, make sure the last identifier |
561 | // before the r_square is tagged as a selector name component. |
562 | if (!ColonFound && CurrentToken->Previous && |
563 | CurrentToken->Previous->is(TT_Unknown) && |
564 | canBeObjCSelectorComponent(*CurrentToken->Previous)) |
565 | CurrentToken->Previous->Type = TT_SelectorName; |
566 | // determineStarAmpUsage() thinks that '*' '[' is allocating an |
567 | // array of pointers, but if '[' starts a selector then '*' is a |
568 | // binary operator. |
569 | if (Parent && Parent->is(TT_PointerOrReference)) |
570 | Parent->Type = TT_BinaryOperator; |
571 | } |
572 | // An arrow after an ObjC method expression is not a lambda arrow. |
573 | if (CurrentToken->Type == TT_ObjCMethodExpr && CurrentToken->Next && |
574 | CurrentToken->Next->is(TT_LambdaArrow)) |
575 | CurrentToken->Next->Type = TT_Unknown; |
576 | Left->MatchingParen = CurrentToken; |
577 | CurrentToken->MatchingParen = Left; |
578 | // FirstObjCSelectorName is set when a colon is found. This does |
579 | // not work, however, when the method has no parameters. |
580 | // Here, we set FirstObjCSelectorName when the end of the method call is |
581 | // reached, in case it was not set already. |
582 | if (!Contexts.back().FirstObjCSelectorName) { |
583 | FormatToken *Previous = CurrentToken->getPreviousNonComment(); |
584 | if (Previous && Previous->is(TT_SelectorName)) { |
585 | Previous->ObjCSelectorNameParts = 1; |
586 | Contexts.back().FirstObjCSelectorName = Previous; |
587 | } |
588 | } else { |
589 | Left->ParameterCount = |
590 | Contexts.back().FirstObjCSelectorName->ObjCSelectorNameParts; |
591 | } |
592 | if (Contexts.back().FirstObjCSelectorName) { |
593 | Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = |
594 | Contexts.back().LongestObjCSelectorName; |
595 | if (Left->BlockParameterCount > 1) |
596 | Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = 0; |
597 | } |
598 | next(); |
599 | return true; |
600 | } |
601 | if (CurrentToken->isOneOf(tok::r_paren, tok::r_brace)) |
602 | return false; |
603 | if (CurrentToken->is(tok::colon)) { |
604 | if (IsCpp11AttributeSpecifier && |
605 | CurrentToken->endsSequence(tok::colon, tok::identifier, |
606 | tok::kw_using)) { |
607 | // Remember that this is a [[using ns: foo]] C++ attribute, so we |
608 | // don't add a space before the colon (unlike other colons). |
609 | CurrentToken->Type = TT_AttributeColon; |
610 | } else if (Left->isOneOf(TT_ArraySubscriptLSquare, |
611 | TT_DesignatedInitializerLSquare)) { |
612 | Left->Type = TT_ObjCMethodExpr; |
613 | StartsObjCMethodExpr = true; |
614 | Contexts.back().ColonIsObjCMethodExpr = true; |
615 | if (Parent && Parent->is(tok::r_paren)) |
616 | // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen. |
617 | Parent->Type = TT_CastRParen; |
618 | } |
619 | ColonFound = true; |
620 | } |
621 | if (CurrentToken->is(tok::comma) && Left->is(TT_ObjCMethodExpr) && |
622 | !ColonFound) |
623 | Left->Type = TT_ArrayInitializerLSquare; |
624 | FormatToken *Tok = CurrentToken; |
625 | if (!consumeToken()) |
626 | return false; |
627 | updateParameterCount(Left, Tok); |
628 | } |
629 | return false; |
630 | } |
631 | |
632 | bool parseBrace() { |
633 | if (CurrentToken) { |
634 | FormatToken *Left = CurrentToken->Previous; |
635 | Left->ParentBracket = Contexts.back().ContextKind; |
636 | |
637 | if (Contexts.back().CaretFound) |
638 | Left->Type = TT_ObjCBlockLBrace; |
639 | Contexts.back().CaretFound = false; |
640 | |
641 | ScopedContextCreator ContextCreator(*this, tok::l_brace, 1); |
642 | Contexts.back().ColonIsDictLiteral = true; |
643 | if (Left->BlockKind == BK_BracedInit) |
644 | Contexts.back().IsExpression = true; |
645 | if (Style.Language == FormatStyle::LK_JavaScript && Left->Previous && |
646 | Left->Previous->is(TT_JsTypeColon)) |
647 | Contexts.back().IsExpression = false; |
648 | |
649 | while (CurrentToken) { |
650 | if (CurrentToken->is(tok::r_brace)) { |
651 | Left->MatchingParen = CurrentToken; |
652 | CurrentToken->MatchingParen = Left; |
653 | next(); |
654 | return true; |
655 | } |
656 | if (CurrentToken->isOneOf(tok::r_paren, tok::r_square)) |
657 | return false; |
658 | updateParameterCount(Left, CurrentToken); |
659 | if (CurrentToken->isOneOf(tok::colon, tok::l_brace, tok::less)) { |
660 | FormatToken *Previous = CurrentToken->getPreviousNonComment(); |
661 | if (Previous->is(TT_JsTypeOptionalQuestion)) |
662 | Previous = Previous->getPreviousNonComment(); |
663 | if ((CurrentToken->is(tok::colon) && |
664 | (!Contexts.back().ColonIsDictLiteral || !Style.isCpp())) || |
665 | Style.Language == FormatStyle::LK_Proto || |
666 | Style.Language == FormatStyle::LK_TextProto) { |
667 | Left->Type = TT_DictLiteral; |
668 | if (Previous->Tok.getIdentifierInfo() || |
669 | Previous->is(tok::string_literal)) |
670 | Previous->Type = TT_SelectorName; |
671 | } |
672 | if (CurrentToken->is(tok::colon) || |
673 | Style.Language == FormatStyle::LK_JavaScript) |
674 | Left->Type = TT_DictLiteral; |
675 | } |
676 | if (CurrentToken->is(tok::comma) && |
677 | Style.Language == FormatStyle::LK_JavaScript) |
678 | Left->Type = TT_DictLiteral; |
679 | if (!consumeToken()) |
680 | return false; |
681 | } |
682 | } |
683 | return true; |
684 | } |
685 | |
686 | void updateParameterCount(FormatToken *Left, FormatToken *Current) { |
687 | // For ObjC methods, the number of parameters is calculated differently as |
688 | // method declarations have a different structure (the parameters are not |
689 | // inside a bracket scope). |
690 | if (Current->is(tok::l_brace) && Current->BlockKind == BK_Block) |
691 | ++Left->BlockParameterCount; |
692 | if (Current->is(tok::comma)) { |
693 | ++Left->ParameterCount; |
694 | if (!Left->Role) |
695 | Left->Role.reset(new CommaSeparatedList(Style)); |
696 | Left->Role->CommaFound(Current); |
697 | } else if (Left->ParameterCount == 0 && Current->isNot(tok::comment)) { |
698 | Left->ParameterCount = 1; |
699 | } |
700 | } |
701 | |
702 | bool parseConditional() { |
703 | while (CurrentToken) { |
704 | if (CurrentToken->is(tok::colon)) { |
705 | CurrentToken->Type = TT_ConditionalExpr; |
706 | next(); |
707 | return true; |
708 | } |
709 | if (!consumeToken()) |
710 | return false; |
711 | } |
712 | return false; |
713 | } |
714 | |
715 | bool parseTemplateDeclaration() { |
716 | if (CurrentToken && CurrentToken->is(tok::less)) { |
717 | CurrentToken->Type = TT_TemplateOpener; |
718 | next(); |
719 | if (!parseAngle()) |
720 | return false; |
721 | if (CurrentToken) |
722 | CurrentToken->Previous->ClosesTemplateDeclaration = true; |
723 | return true; |
724 | } |
725 | return false; |
726 | } |
727 | |
728 | bool consumeToken() { |
729 | FormatToken *Tok = CurrentToken; |
730 | next(); |
731 | switch (Tok->Tok.getKind()) { |
732 | case tok::plus: |
733 | case tok::minus: |
734 | if (!Tok->Previous && Line.MustBeDeclaration) |
735 | Tok->Type = TT_ObjCMethodSpecifier; |
736 | break; |
737 | case tok::colon: |
738 | if (!Tok->Previous) |
739 | return false; |
740 | // Colons from ?: are handled in parseConditional(). |
741 | if (Style.Language == FormatStyle::LK_JavaScript) { |
742 | if (Contexts.back().ColonIsForRangeExpr || // colon in for loop |
743 | (Contexts.size() == 1 && // switch/case labels |
744 | !Line.First->isOneOf(tok::kw_enum, tok::kw_case)) || |
745 | Contexts.back().ContextKind == tok::l_paren || // function params |
746 | Contexts.back().ContextKind == tok::l_square || // array type |
747 | (!Contexts.back().IsExpression && |
748 | Contexts.back().ContextKind == tok::l_brace) || // object type |
749 | (Contexts.size() == 1 && |
750 | Line.MustBeDeclaration)) { // method/property declaration |
751 | Contexts.back().IsExpression = false; |
752 | Tok->Type = TT_JsTypeColon; |
753 | break; |
754 | } |
755 | } |
756 | if (Contexts.back().ColonIsDictLiteral || |
757 | Style.Language == FormatStyle::LK_Proto || |
758 | Style.Language == FormatStyle::LK_TextProto) { |
759 | Tok->Type = TT_DictLiteral; |
760 | if (Style.Language == FormatStyle::LK_TextProto) { |
761 | if (FormatToken *Previous = Tok->getPreviousNonComment()) |
762 | Previous->Type = TT_SelectorName; |
763 | } |
764 | } else if (Contexts.back().ColonIsObjCMethodExpr || |
765 | Line.startsWith(TT_ObjCMethodSpecifier)) { |
766 | Tok->Type = TT_ObjCMethodExpr; |
767 | const FormatToken *BeforePrevious = Tok->Previous->Previous; |
768 | // Ensure we tag all identifiers in method declarations as |
769 | // TT_SelectorName. |
770 | bool UnknownIdentifierInMethodDeclaration = |
771 | Line.startsWith(TT_ObjCMethodSpecifier) && |
772 | Tok->Previous->is(tok::identifier) && Tok->Previous->is(TT_Unknown); |
773 | if (!BeforePrevious || |
774 | // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen. |
775 | !(BeforePrevious->is(TT_CastRParen) || |
776 | (BeforePrevious->is(TT_ObjCMethodExpr) && |
777 | BeforePrevious->is(tok::colon))) || |
778 | BeforePrevious->is(tok::r_square) || |
779 | Contexts.back().LongestObjCSelectorName == 0 || |
780 | UnknownIdentifierInMethodDeclaration) { |
781 | Tok->Previous->Type = TT_SelectorName; |
782 | if (!Contexts.back().FirstObjCSelectorName) |
783 | Contexts.back().FirstObjCSelectorName = Tok->Previous; |
784 | else if (Tok->Previous->ColumnWidth > |
785 | Contexts.back().LongestObjCSelectorName) |
786 | Contexts.back().LongestObjCSelectorName = |
787 | Tok->Previous->ColumnWidth; |
788 | Tok->Previous->ParameterIndex = |
789 | Contexts.back().FirstObjCSelectorName->ObjCSelectorNameParts; |
790 | ++Contexts.back().FirstObjCSelectorName->ObjCSelectorNameParts; |
791 | } |
792 | } else if (Contexts.back().ColonIsForRangeExpr) { |
793 | Tok->Type = TT_RangeBasedForLoopColon; |
794 | } else if (CurrentToken && CurrentToken->is(tok::numeric_constant)) { |
795 | Tok->Type = TT_BitFieldColon; |
796 | } else if (Contexts.size() == 1 && |
797 | !Line.First->isOneOf(tok::kw_enum, tok::kw_case)) { |
798 | if (Tok->getPreviousNonComment()->isOneOf(tok::r_paren, |
799 | tok::kw_noexcept)) |
800 | Tok->Type = TT_CtorInitializerColon; |
801 | else |
802 | Tok->Type = TT_InheritanceColon; |
803 | } else if (canBeObjCSelectorComponent(*Tok->Previous) && Tok->Next && |
804 | (Tok->Next->isOneOf(tok::r_paren, tok::comma) || |
805 | (canBeObjCSelectorComponent(*Tok->Next) && Tok->Next->Next && |
806 | Tok->Next->Next->is(tok::colon)))) { |
807 | // This handles a special macro in ObjC code where selectors including |
808 | // the colon are passed as macro arguments. |
809 | Tok->Type = TT_ObjCMethodExpr; |
810 | } else if (Contexts.back().ContextKind == tok::l_paren) { |
811 | Tok->Type = TT_InlineASMColon; |
812 | } |
813 | break; |
814 | case tok::pipe: |
815 | case tok::amp: |
816 | // | and & in declarations/type expressions represent union and |
817 | // intersection types, respectively. |
818 | if (Style.Language == FormatStyle::LK_JavaScript && |
819 | !Contexts.back().IsExpression) |
820 | Tok->Type = TT_JsTypeOperator; |
821 | break; |
822 | case tok::kw_if: |
823 | case tok::kw_while: |
824 | if (Tok->is(tok::kw_if) && CurrentToken && |
825 | CurrentToken->is(tok::kw_constexpr)) |
826 | next(); |
827 | if (CurrentToken && CurrentToken->is(tok::l_paren)) { |
828 | next(); |
829 | if (!parseParens(/*LookForDecls=*/true)) |
830 | return false; |
831 | } |
832 | break; |
833 | case tok::kw_for: |
834 | if (Style.Language == FormatStyle::LK_JavaScript) { |
835 | // x.for and {for: ...} |
836 | if ((Tok->Previous && Tok->Previous->is(tok::period)) || |
837 | (Tok->Next && Tok->Next->is(tok::colon))) |
838 | break; |
839 | // JS' for await ( ... |
840 | if (CurrentToken && CurrentToken->is(Keywords.kw_await)) |
841 | next(); |
842 | } |
843 | Contexts.back().ColonIsForRangeExpr = true; |
844 | next(); |
845 | if (!parseParens()) |
846 | return false; |
847 | break; |
848 | case tok::l_paren: |
849 | // When faced with 'operator()()', the kw_operator handler incorrectly |
850 | // marks the first l_paren as a OverloadedOperatorLParen. Here, we make |
851 | // the first two parens OverloadedOperators and the second l_paren an |
852 | // OverloadedOperatorLParen. |
853 | if (Tok->Previous && Tok->Previous->is(tok::r_paren) && |
854 | Tok->Previous->MatchingParen && |
855 | Tok->Previous->MatchingParen->is(TT_OverloadedOperatorLParen)) { |
856 | Tok->Previous->Type = TT_OverloadedOperator; |
857 | Tok->Previous->MatchingParen->Type = TT_OverloadedOperator; |
858 | Tok->Type = TT_OverloadedOperatorLParen; |
859 | } |
860 | |
861 | if (!parseParens()) |
862 | return false; |
863 | if (Line.MustBeDeclaration && Contexts.size() == 1 && |
864 | !Contexts.back().IsExpression && !Line.startsWith(TT_ObjCProperty) && |
865 | (!Tok->Previous || |
866 | !Tok->Previous->isOneOf(tok::kw_decltype, tok::kw___attribute, |
867 | TT_LeadingJavaAnnotation))) |
868 | Line.MightBeFunctionDecl = true; |
869 | break; |
870 | case tok::l_square: |
871 | if (!parseSquare()) |
872 | return false; |
873 | break; |
874 | case tok::l_brace: |
875 | if (Style.Language == FormatStyle::LK_TextProto) { |
876 | FormatToken *Previous = Tok->getPreviousNonComment(); |
877 | if (Previous && Previous->Type != TT_DictLiteral) |
878 | Previous->Type = TT_SelectorName; |
879 | } |
880 | if (!parseBrace()) |
881 | return false; |
882 | break; |
883 | case tok::less: |
884 | if (parseAngle()) { |
885 | Tok->Type = TT_TemplateOpener; |
886 | // In TT_Proto, we must distignuish between: |
887 | // map<key, value> |
888 | // msg < item: data > |
889 | // msg: < item: data > |
890 | // In TT_TextProto, map<key, value> does not occur. |
891 | if (Style.Language == FormatStyle::LK_TextProto || |
892 | (Style.Language == FormatStyle::LK_Proto && Tok->Previous && |
893 | Tok->Previous->isOneOf(TT_SelectorName, TT_DictLiteral))) { |
894 | Tok->Type = TT_DictLiteral; |
895 | FormatToken *Previous = Tok->getPreviousNonComment(); |
896 | if (Previous && Previous->Type != TT_DictLiteral) |
897 | Previous->Type = TT_SelectorName; |
898 | } |
899 | } else { |
900 | Tok->Type = TT_BinaryOperator; |
901 | NonTemplateLess.insert(Tok); |
902 | CurrentToken = Tok; |
903 | next(); |
904 | } |
905 | break; |
906 | case tok::r_paren: |
907 | case tok::r_square: |
908 | return false; |
909 | case tok::r_brace: |
910 | // Lines can start with '}'. |
911 | if (Tok->Previous) |
912 | return false; |
913 | break; |
914 | case tok::greater: |
915 | if (Style.Language != FormatStyle::LK_TextProto) |
916 | Tok->Type = TT_BinaryOperator; |
917 | break; |
918 | case tok::kw_operator: |
919 | if (Style.Language == FormatStyle::LK_TextProto || |
920 | Style.Language == FormatStyle::LK_Proto) |
921 | break; |
922 | while (CurrentToken && |
923 | !CurrentToken->isOneOf(tok::l_paren, tok::semi, tok::r_paren)) { |
924 | if (CurrentToken->isOneOf(tok::star, tok::amp)) |
925 | CurrentToken->Type = TT_PointerOrReference; |
926 | consumeToken(); |
927 | if (CurrentToken && |
928 | CurrentToken->Previous->isOneOf(TT_BinaryOperator, TT_UnaryOperator, |
929 | tok::comma)) |
930 | CurrentToken->Previous->Type = TT_OverloadedOperator; |
931 | } |
932 | if (CurrentToken) { |
933 | CurrentToken->Type = TT_OverloadedOperatorLParen; |
934 | if (CurrentToken->Previous->is(TT_BinaryOperator)) |
935 | CurrentToken->Previous->Type = TT_OverloadedOperator; |
936 | } |
937 | break; |
938 | case tok::question: |
939 | if (Style.Language == FormatStyle::LK_JavaScript && Tok->Next && |
940 | Tok->Next->isOneOf(tok::semi, tok::comma, tok::colon, tok::r_paren, |
941 | tok::r_brace)) { |
942 | // Question marks before semicolons, colons, etc. indicate optional |
943 | // types (fields, parameters), e.g. |
944 | // function(x?: string, y?) {...} |
945 | // class X { y?; } |
946 | Tok->Type = TT_JsTypeOptionalQuestion; |
947 | break; |
948 | } |
949 | // Declarations cannot be conditional expressions, this can only be part |
950 | // of a type declaration. |
951 | if (Line.MustBeDeclaration && !Contexts.back().IsExpression && |
952 | Style.Language == FormatStyle::LK_JavaScript) |
953 | break; |
954 | parseConditional(); |
955 | break; |
956 | case tok::kw_template: |
957 | parseTemplateDeclaration(); |
958 | break; |
959 | case tok::comma: |
960 | if (Contexts.back().InCtorInitializer) |
961 | Tok->Type = TT_CtorInitializerComma; |
962 | else if (Contexts.back().InInheritanceList) |
963 | Tok->Type = TT_InheritanceComma; |
964 | else if (Contexts.back().FirstStartOfName && |
965 | (Contexts.size() == 1 || Line.startsWith(tok::kw_for))) { |
966 | Contexts.back().FirstStartOfName->PartOfMultiVariableDeclStmt = true; |
967 | Line.IsMultiVariableDeclStmt = true; |
968 | } |
969 | if (Contexts.back().IsForEachMacro) |
970 | Contexts.back().IsExpression = true; |
971 | break; |
972 | case tok::identifier: |
973 | if (Tok->isOneOf(Keywords.kw___has_include, |
974 | Keywords.kw___has_include_next)) { |
975 | parseHasInclude(); |
976 | } |
977 | break; |
978 | default: |
979 | break; |
980 | } |
981 | return true; |
982 | } |
983 | |
984 | void parseIncludeDirective() { |
985 | if (CurrentToken && CurrentToken->is(tok::less)) { |
986 | next(); |
987 | while (CurrentToken) { |
988 | // Mark tokens up to the trailing line comments as implicit string |
989 | // literals. |
990 | if (CurrentToken->isNot(tok::comment) && |
991 | !CurrentToken->TokenText.startswith("//")) |
992 | CurrentToken->Type = TT_ImplicitStringLiteral; |
993 | next(); |
994 | } |
995 | } |
996 | } |
997 | |
998 | void parseWarningOrError() { |
999 | next(); |
1000 | // We still want to format the whitespace left of the first token of the |
1001 | // warning or error. |
1002 | next(); |
1003 | while (CurrentToken) { |
1004 | CurrentToken->Type = TT_ImplicitStringLiteral; |
1005 | next(); |
1006 | } |
1007 | } |
1008 | |
1009 | void parsePragma() { |
1010 | next(); // Consume "pragma". |
1011 | if (CurrentToken && |
1012 | CurrentToken->isOneOf(Keywords.kw_mark, Keywords.kw_option)) { |
1013 | bool IsMark = CurrentToken->is(Keywords.kw_mark); |
1014 | next(); // Consume "mark". |
1015 | next(); // Consume first token (so we fix leading whitespace). |
1016 | while (CurrentToken) { |
1017 | if (IsMark || CurrentToken->Previous->is(TT_BinaryOperator)) |
1018 | CurrentToken->Type = TT_ImplicitStringLiteral; |
1019 | next(); |
1020 | } |
1021 | } |
1022 | } |
1023 | |
1024 | void parseHasInclude() { |
1025 | if (!CurrentToken || !CurrentToken->is(tok::l_paren)) |
1026 | return; |
1027 | next(); // '(' |
1028 | parseIncludeDirective(); |
1029 | next(); // ')' |
1030 | } |
1031 | |
1032 | LineType parsePreprocessorDirective() { |
1033 | bool IsFirstToken = CurrentToken->IsFirst; |
1034 | LineType Type = LT_PreprocessorDirective; |
1035 | next(); |
1036 | if (!CurrentToken) |
1037 | return Type; |
1038 | |
1039 | if (Style.Language == FormatStyle::LK_JavaScript && IsFirstToken) { |
1040 | // JavaScript files can contain shebang lines of the form: |
1041 | // #!/usr/bin/env node |
1042 | // Treat these like C++ #include directives. |
1043 | while (CurrentToken) { |
1044 | // Tokens cannot be comments here. |
1045 | CurrentToken->Type = TT_ImplicitStringLiteral; |
1046 | next(); |
1047 | } |
1048 | return LT_ImportStatement; |
1049 | } |
1050 | |
1051 | if (CurrentToken->Tok.is(tok::numeric_constant)) { |
1052 | CurrentToken->SpacesRequiredBefore = 1; |
1053 | return Type; |
1054 | } |
1055 | // Hashes in the middle of a line can lead to any strange token |
1056 | // sequence. |
1057 | if (!CurrentToken->Tok.getIdentifierInfo()) |
1058 | return Type; |
1059 | switch (CurrentToken->Tok.getIdentifierInfo()->getPPKeywordID()) { |
1060 | case tok::pp_include: |
1061 | case tok::pp_include_next: |
1062 | case tok::pp_import: |
1063 | next(); |
1064 | parseIncludeDirective(); |
1065 | Type = LT_ImportStatement; |
1066 | break; |
1067 | case tok::pp_error: |
1068 | case tok::pp_warning: |
1069 | parseWarningOrError(); |
1070 | break; |
1071 | case tok::pp_pragma: |
1072 | parsePragma(); |
1073 | break; |
1074 | case tok::pp_if: |
1075 | case tok::pp_elif: |
1076 | Contexts.back().IsExpression = true; |
1077 | parseLine(); |
1078 | break; |
1079 | default: |
1080 | break; |
1081 | } |
1082 | while (CurrentToken) { |
1083 | FormatToken *Tok = CurrentToken; |
1084 | next(); |
1085 | if (Tok->is(tok::l_paren)) |
1086 | parseParens(); |
1087 | else if (Tok->isOneOf(Keywords.kw___has_include, |
1088 | Keywords.kw___has_include_next)) |
1089 | parseHasInclude(); |
1090 | } |
1091 | return Type; |
1092 | } |
1093 | |
1094 | public: |
1095 | LineType parseLine() { |
1096 | NonTemplateLess.clear(); |
1097 | if (CurrentToken->is(tok::hash)) |
1098 | return parsePreprocessorDirective(); |
1099 | |
1100 | // Directly allow to 'import <string-literal>' to support protocol buffer |
1101 | // definitions (github.com/google/protobuf) or missing "#" (either way we |
1102 | // should not break the line). |
1103 | IdentifierInfo *Info = CurrentToken->Tok.getIdentifierInfo(); |
1104 | if ((Style.Language == FormatStyle::LK_Java && |
1105 | CurrentToken->is(Keywords.kw_package)) || |
1106 | (Info && Info->getPPKeywordID() == tok::pp_import && |
1107 | CurrentToken->Next && |
1108 | CurrentToken->Next->isOneOf(tok::string_literal, tok::identifier, |
1109 | tok::kw_static))) { |
1110 | next(); |
1111 | parseIncludeDirective(); |
1112 | return LT_ImportStatement; |
1113 | } |
1114 | |
1115 | // If this line starts and ends in '<' and '>', respectively, it is likely |
1116 | // part of "#define <a/b.h>". |
1117 | if (CurrentToken->is(tok::less) && Line.Last->is(tok::greater)) { |
1118 | parseIncludeDirective(); |
1119 | return LT_ImportStatement; |
1120 | } |
1121 | |
1122 | // In .proto files, top-level options are very similar to import statements |
1123 | // and should not be line-wrapped. |
1124 | if (Style.Language == FormatStyle::LK_Proto && Line.Level == 0 && |
1125 | CurrentToken->is(Keywords.kw_option)) { |
1126 | next(); |
1127 | if (CurrentToken && CurrentToken->is(tok::identifier)) |
1128 | return LT_ImportStatement; |
1129 | } |
1130 | |
1131 | bool KeywordVirtualFound = false; |
1132 | bool ImportStatement = false; |
1133 | |
1134 | // import {...} from '...'; |
1135 | if (Style.Language == FormatStyle::LK_JavaScript && |
1136 | CurrentToken->is(Keywords.kw_import)) |
1137 | ImportStatement = true; |
1138 | |
1139 | while (CurrentToken) { |
1140 | if (CurrentToken->is(tok::kw_virtual)) |
1141 | KeywordVirtualFound = true; |
1142 | if (Style.Language == FormatStyle::LK_JavaScript) { |
1143 | // export {...} from '...'; |
1144 | // An export followed by "from 'some string';" is a re-export from |
1145 | // another module identified by a URI and is treated as a |
1146 | // LT_ImportStatement (i.e. prevent wraps on it for long URIs). |
1147 | // Just "export {...};" or "export class ..." should not be treated as |
1148 | // an import in this sense. |
1149 | if (Line.First->is(tok::kw_export) && |
1150 | CurrentToken->is(Keywords.kw_from) && CurrentToken->Next && |
1151 | CurrentToken->Next->isStringLiteral()) |
1152 | ImportStatement = true; |
1153 | if (isClosureImportStatement(*CurrentToken)) |
1154 | ImportStatement = true; |
1155 | } |
1156 | if (!consumeToken()) |
1157 | return LT_Invalid; |
1158 | } |
1159 | if (KeywordVirtualFound) |
1160 | return LT_VirtualFunctionDecl; |
1161 | if (ImportStatement) |
1162 | return LT_ImportStatement; |
1163 | |
1164 | if (Line.startsWith(TT_ObjCMethodSpecifier)) { |
1165 | if (Contexts.back().FirstObjCSelectorName) |
1166 | Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = |
1167 | Contexts.back().LongestObjCSelectorName; |
1168 | return LT_ObjCMethodDecl; |
1169 | } |
1170 | |
1171 | return LT_Other; |
1172 | } |
1173 | |
1174 | private: |
1175 | bool isClosureImportStatement(const FormatToken &Tok) { |
1176 | // FIXME: Closure-library specific stuff should not be hard-coded but be |
1177 | // configurable. |
1178 | return Tok.TokenText == "goog" && Tok.Next && Tok.Next->is(tok::period) && |
1179 | Tok.Next->Next && |
1180 | (Tok.Next->Next->TokenText == "module" || |
1181 | Tok.Next->Next->TokenText == "provide" || |
1182 | Tok.Next->Next->TokenText == "require" || |
1183 | Tok.Next->Next->TokenText == "requireType" || |
1184 | Tok.Next->Next->TokenText == "forwardDeclare") && |
1185 | Tok.Next->Next->Next && Tok.Next->Next->Next->is(tok::l_paren); |
1186 | } |
1187 | |
1188 | void resetTokenMetadata(FormatToken *Token) { |
1189 | if (!Token) |
1190 | return; |
1191 | |
1192 | // Reset token type in case we have already looked at it and then |
1193 | // recovered from an error (e.g. failure to find the matching >). |
1194 | if (!CurrentToken->isOneOf( |
1195 | TT_LambdaLSquare, TT_LambdaLBrace, TT_ForEachMacro, |
1196 | TT_FunctionLBrace, TT_ImplicitStringLiteral, TT_InlineASMBrace, |
1197 | TT_JsFatArrow, TT_LambdaArrow, TT_OverloadedOperator, |
1198 | TT_RegexLiteral, TT_TemplateString, TT_ObjCStringLiteral)) |
1199 | CurrentToken->Type = TT_Unknown; |
1200 | CurrentToken->Role.reset(); |
1201 | CurrentToken->MatchingParen = nullptr; |
1202 | CurrentToken->FakeLParens.clear(); |
1203 | CurrentToken->FakeRParens = 0; |
1204 | } |
1205 | |
1206 | void next() { |
1207 | if (CurrentToken) { |
1208 | CurrentToken->NestingLevel = Contexts.size() - 1; |
1209 | CurrentToken->BindingStrength = Contexts.back().BindingStrength; |
1210 | modifyContext(*CurrentToken); |
1211 | determineTokenType(*CurrentToken); |
1212 | CurrentToken = CurrentToken->Next; |
1213 | } |
1214 | |
1215 | resetTokenMetadata(CurrentToken); |
1216 | } |
1217 | |
1218 | /// A struct to hold information valid in a specific context, e.g. |
1219 | /// a pair of parenthesis. |
1220 | struct Context { |
1221 | Context(tok::TokenKind ContextKind, unsigned BindingStrength, |
1222 | bool IsExpression) |
1223 | : ContextKind(ContextKind), BindingStrength(BindingStrength), |
1224 | IsExpression(IsExpression) {} |
1225 | |
1226 | tok::TokenKind ContextKind; |
1227 | unsigned BindingStrength; |
1228 | bool IsExpression; |
1229 | unsigned LongestObjCSelectorName = 0; |
1230 | bool ColonIsForRangeExpr = false; |
1231 | bool ColonIsDictLiteral = false; |
1232 | bool ColonIsObjCMethodExpr = false; |
1233 | FormatToken *FirstObjCSelectorName = nullptr; |
1234 | FormatToken *FirstStartOfName = nullptr; |
1235 | bool CanBeExpression = true; |
1236 | bool InTemplateArgument = false; |
1237 | bool InCtorInitializer = false; |
1238 | bool InInheritanceList = false; |
1239 | bool CaretFound = false; |
1240 | bool IsForEachMacro = false; |
1241 | bool InCpp11AttributeSpecifier = false; |
1242 | bool InCSharpAttributeSpecifier = false; |
1243 | }; |
1244 | |
1245 | /// Puts a new \c Context onto the stack \c Contexts for the lifetime |
1246 | /// of each instance. |
1247 | struct ScopedContextCreator { |
1248 | AnnotatingParser &P; |
1249 | |
1250 | ScopedContextCreator(AnnotatingParser &P, tok::TokenKind ContextKind, |
1251 | unsigned Increase) |
1252 | : P(P) { |
1253 | P.Contexts.push_back(Context(ContextKind, |
1254 | P.Contexts.back().BindingStrength + Increase, |
1255 | P.Contexts.back().IsExpression)); |
1256 | } |
1257 | |
1258 | ~ScopedContextCreator() { P.Contexts.pop_back(); } |
1259 | }; |
1260 | |
1261 | void modifyContext(const FormatToken &Current) { |
1262 | if (Current.getPrecedence() == prec::Assignment && |
1263 | !Line.First->isOneOf(tok::kw_template, tok::kw_using, tok::kw_return) && |
1264 | // Type aliases use `type X = ...;` in TypeScript and can be exported |
1265 | // using `export type ...`. |
1266 | !(Style.Language == FormatStyle::LK_JavaScript && |
1267 | (Line.startsWith(Keywords.kw_type, tok::identifier) || |
1268 | Line.startsWith(tok::kw_export, Keywords.kw_type, |
1269 | tok::identifier))) && |
1270 | (!Current.Previous || Current.Previous->isNot(tok::kw_operator))) { |
1271 | Contexts.back().IsExpression = true; |
1272 | if (!Line.startsWith(TT_UnaryOperator)) { |
1273 | for (FormatToken *Previous = Current.Previous; |
1274 | Previous && Previous->Previous && |
1275 | !Previous->Previous->isOneOf(tok::comma, tok::semi); |
1276 | Previous = Previous->Previous) { |
1277 | if (Previous->isOneOf(tok::r_square, tok::r_paren)) { |
1278 | Previous = Previous->MatchingParen; |
1279 | if (!Previous) |
1280 | break; |
1281 | } |
1282 | if (Previous->opensScope()) |
1283 | break; |
1284 | if (Previous->isOneOf(TT_BinaryOperator, TT_UnaryOperator) && |
1285 | Previous->isOneOf(tok::star, tok::amp, tok::ampamp) && |
1286 | Previous->Previous && Previous->Previous->isNot(tok::equal)) |
1287 | Previous->Type = TT_PointerOrReference; |
1288 | } |
1289 | } |
1290 | } else if (Current.is(tok::lessless) && |
1291 | (!Current.Previous || !Current.Previous->is(tok::kw_operator))) { |
1292 | Contexts.back().IsExpression = true; |
1293 | } else if (Current.isOneOf(tok::kw_return, tok::kw_throw)) { |
1294 | Contexts.back().IsExpression = true; |
1295 | } else if (Current.is(TT_TrailingReturnArrow)) { |
1296 | Contexts.back().IsExpression = false; |
1297 | } else if (Current.is(TT_LambdaArrow) || Current.is(Keywords.kw_assert)) { |
1298 | Contexts.back().IsExpression = Style.Language == FormatStyle::LK_Java; |
1299 | } else if (Current.Previous && |
1300 | Current.Previous->is(TT_CtorInitializerColon)) { |
1301 | Contexts.back().IsExpression = true; |
1302 | Contexts.back().InCtorInitializer = true; |
1303 | } else if (Current.Previous && Current.Previous->is(TT_InheritanceColon)) { |
1304 | Contexts.back().InInheritanceList = true; |
1305 | } else if (Current.isOneOf(tok::r_paren, tok::greater, tok::comma)) { |
1306 | for (FormatToken *Previous = Current.Previous; |
1307 | Previous && Previous->isOneOf(tok::star, tok::amp); |
1308 | Previous = Previous->Previous) |
1309 | Previous->Type = TT_PointerOrReference; |
1310 | if (Line.MustBeDeclaration && !Contexts.front().InCtorInitializer) |
1311 | Contexts.back().IsExpression = false; |
1312 | } else if (Current.is(tok::kw_new)) { |
1313 | Contexts.back().CanBeExpression = false; |
1314 | } else if (Current.isOneOf(tok::semi, tok::exclaim)) { |
1315 | // This should be the condition or increment in a for-loop. |
1316 | Contexts.back().IsExpression = true; |
1317 | } |
1318 | } |
1319 | |
1320 | void determineTokenType(FormatToken &Current) { |
1321 | if (!Current.is(TT_Unknown)) |
1322 | // The token type is already known. |
1323 | return; |
1324 | |
1325 | if (Style.Language == FormatStyle::LK_JavaScript) { |
1326 | if (Current.is(tok::exclaim)) { |
1327 | if (Current.Previous && |
1328 | (Current.Previous->isOneOf(tok::identifier, tok::kw_namespace, |
1329 | tok::r_paren, tok::r_square, |
1330 | tok::r_brace) || |
1331 | Current.Previous->Tok.isLiteral())) { |
1332 | Current.Type = TT_JsNonNullAssertion; |
1333 | return; |
1334 | } |
1335 | if (Current.Next && |
1336 | Current.Next->isOneOf(TT_BinaryOperator, Keywords.kw_as)) { |
1337 | Current.Type = TT_JsNonNullAssertion; |
1338 | return; |
1339 | } |
1340 | } |
1341 | } |
1342 | |
1343 | // Line.MightBeFunctionDecl can only be true after the parentheses of a |
1344 | // function declaration have been found. In this case, 'Current' is a |
1345 | // trailing token of this declaration and thus cannot be a name. |
1346 | if (Current.is(Keywords.kw_instanceof)) { |
1347 | Current.Type = TT_BinaryOperator; |
1348 | } else if (isStartOfName(Current) && |
1349 | (!Line.MightBeFunctionDecl || Current.NestingLevel != 0)) { |
1350 | Contexts.back().FirstStartOfName = &Current; |
1351 | Current.Type = TT_StartOfName; |
1352 | } else if (Current.is(tok::semi)) { |
1353 | // Reset FirstStartOfName after finding a semicolon so that a for loop |
1354 | // with multiple increment statements is not confused with a for loop |
1355 | // having multiple variable declarations. |
1356 | Contexts.back().FirstStartOfName = nullptr; |
1357 | } else if (Current.isOneOf(tok::kw_auto, tok::kw___auto_type)) { |
1358 | AutoFound = true; |
1359 | } else if (Current.is(tok::arrow) && |
1360 | Style.Language == FormatStyle::LK_Java) { |
1361 | Current.Type = TT_LambdaArrow; |
1362 | } else if (Current.is(tok::arrow) && AutoFound && Line.MustBeDeclaration && |
1363 | Current.NestingLevel == 0) { |
1364 | Current.Type = TT_TrailingReturnArrow; |
1365 | } else if (Current.isOneOf(tok::star, tok::amp, tok::ampamp)) { |
1366 | Current.Type = determineStarAmpUsage(Current, |
1367 | Contexts.back().CanBeExpression && |
1368 | Contexts.back().IsExpression, |
1369 | Contexts.back().InTemplateArgument); |
1370 | } else if (Current.isOneOf(tok::minus, tok::plus, tok::caret)) { |
1371 | Current.Type = determinePlusMinusCaretUsage(Current); |
1372 | if (Current.is(TT_UnaryOperator) && Current.is(tok::caret)) |
1373 | Contexts.back().CaretFound = true; |
1374 | } else if (Current.isOneOf(tok::minusminus, tok::plusplus)) { |
1375 | Current.Type = determineIncrementUsage(Current); |
1376 | } else if (Current.isOneOf(tok::exclaim, tok::tilde)) { |
1377 | Current.Type = TT_UnaryOperator; |
1378 | } else if (Current.is(tok::question)) { |
1379 | if (Style.Language == FormatStyle::LK_JavaScript && |
1380 | Line.MustBeDeclaration && !Contexts.back().IsExpression) { |
1381 | // In JavaScript, `interface X { foo?(): bar; }` is an optional method |
1382 | // on the interface, not a ternary expression. |
1383 | Current.Type = TT_JsTypeOptionalQuestion; |
1384 | } else { |
1385 | Current.Type = TT_ConditionalExpr; |
1386 | } |
1387 | } else if (Current.isBinaryOperator() && |
1388 | (!Current.Previous || Current.Previous->isNot(tok::l_square)) && |
1389 | (!Current.is(tok::greater) && |
1390 | Style.Language != FormatStyle::LK_TextProto)) { |
1391 | Current.Type = TT_BinaryOperator; |
1392 | } else if (Current.is(tok::comment)) { |
1393 | if (Current.TokenText.startswith("/*")) { |
1394 | if (Current.TokenText.endswith("*/")) |
1395 | Current.Type = TT_BlockComment; |
1396 | else |
1397 | // The lexer has for some reason determined a comment here. But we |
1398 | // cannot really handle it, if it isn't properly terminated. |
1399 | Current.Tok.setKind(tok::unknown); |
1400 | } else { |
1401 | Current.Type = TT_LineComment; |
1402 | } |
1403 | } else if (Current.is(tok::r_paren)) { |
1404 | if (rParenEndsCast(Current)) |
1405 | Current.Type = TT_CastRParen; |
1406 | if (Current.MatchingParen && Current.Next && |
1407 | !Current.Next->isBinaryOperator() && |
1408 | !Current.Next->isOneOf(tok::semi, tok::colon, tok::l_brace, |
1409 | tok::comma, tok::period, tok::arrow, |
1410 | tok::coloncolon)) |
1411 | if (FormatToken *AfterParen = Current.MatchingParen->Next) { |
1412 | // Make sure this isn't the return type of an Obj-C block declaration |
1413 | if (AfterParen->Tok.isNot(tok::caret)) { |
1414 | if (FormatToken *BeforeParen = Current.MatchingParen->Previous) |
1415 | if (BeforeParen->is(tok::identifier) && |
1416 | BeforeParen->TokenText == BeforeParen->TokenText.upper() && |
1417 | (!BeforeParen->Previous || |
1418 | BeforeParen->Previous->ClosesTemplateDeclaration)) |
1419 | Current.Type = TT_FunctionAnnotationRParen; |
1420 | } |
1421 | } |
1422 | } else if (Current.is(tok::at) && Current.Next && |
1423 | Style.Language != FormatStyle::LK_JavaScript && |
1424 | Style.Language != FormatStyle::LK_Java) { |
1425 | // In Java & JavaScript, "@..." is a decorator or annotation. In ObjC, it |
1426 | // marks declarations and properties that need special formatting. |
1427 | switch (Current.Next->Tok.getObjCKeywordID()) { |
1428 | case tok::objc_interface: |
1429 | case tok::objc_implementation: |
1430 | case tok::objc_protocol: |
1431 | Current.Type = TT_ObjCDecl; |
1432 | break; |
1433 | case tok::objc_property: |
1434 | Current.Type = TT_ObjCProperty; |
1435 | break; |
1436 | default: |
1437 | break; |
1438 | } |
1439 | } else if (Current.is(tok::period)) { |
1440 | FormatToken *PreviousNoComment = Current.getPreviousNonComment(); |
1441 | if (PreviousNoComment && |
1442 | PreviousNoComment->isOneOf(tok::comma, tok::l_brace)) |
1443 | Current.Type = TT_DesignatedInitializerPeriod; |
1444 | else if (Style.Language == FormatStyle::LK_Java && Current.Previous && |
1445 | Current.Previous->isOneOf(TT_JavaAnnotation, |
1446 | TT_LeadingJavaAnnotation)) { |
1447 | Current.Type = Current.Previous->Type; |
1448 | } |
1449 | } else if (canBeObjCSelectorComponent(Current) && |
1450 | // FIXME(bug 36976): ObjC return types shouldn't use |
1451 | // TT_CastRParen. |
1452 | Current.Previous && Current.Previous->is(TT_CastRParen) && |
1453 | Current.Previous->MatchingParen && |
1454 | Current.Previous->MatchingParen->Previous && |
1455 | Current.Previous->MatchingParen->Previous->is( |
1456 | TT_ObjCMethodSpecifier)) { |
1457 | // This is the first part of an Objective-C selector name. (If there's no |
1458 | // colon after this, this is the only place which annotates the identifier |
1459 | // as a selector.) |
1460 | Current.Type = TT_SelectorName; |
1461 | } else if (Current.isOneOf(tok::identifier, tok::kw_const) && |
1462 | Current.Previous && |
1463 | !Current.Previous->isOneOf(tok::equal, tok::at) && |
1464 | Line.MightBeFunctionDecl && Contexts.size() == 1) { |
1465 | // Line.MightBeFunctionDecl can only be true after the parentheses of a |
1466 | // function declaration have been found. |
1467 | Current.Type = TT_TrailingAnnotation; |
1468 | } else if ((Style.Language == FormatStyle::LK_Java || |
1469 | Style.Language == FormatStyle::LK_JavaScript) && |
1470 | Current.Previous) { |
1471 | if (Current.Previous->is(tok::at) && |
1472 | Current.isNot(Keywords.kw_interface)) { |
1473 | const FormatToken &AtToken = *Current.Previous; |
1474 | const FormatToken *Previous = AtToken.getPreviousNonComment(); |
1475 | if (!Previous || Previous->is(TT_LeadingJavaAnnotation)) |
1476 | Current.Type = TT_LeadingJavaAnnotation; |
1477 | else |
1478 | Current.Type = TT_JavaAnnotation; |
1479 | } else if (Current.Previous->is(tok::period) && |
1480 | Current.Previous->isOneOf(TT_JavaAnnotation, |
1481 | TT_LeadingJavaAnnotation)) { |
1482 | Current.Type = Current.Previous->Type; |
1483 | } |
1484 | } |
1485 | } |
1486 | |
1487 | /// Take a guess at whether \p Tok starts a name of a function or |
1488 | /// variable declaration. |
1489 | /// |
1490 | /// This is a heuristic based on whether \p Tok is an identifier following |
1491 | /// something that is likely a type. |
1492 | bool isStartOfName(const FormatToken &Tok) { |
1493 | if (Tok.isNot(tok::identifier) || !Tok.Previous) |
1494 | return false; |
1495 | |
1496 | if (Tok.Previous->isOneOf(TT_LeadingJavaAnnotation, Keywords.kw_instanceof, |
1497 | Keywords.kw_as)) |
1498 | return false; |
1499 | if (Style.Language == FormatStyle::LK_JavaScript && |
1500 | Tok.Previous->is(Keywords.kw_in)) |
1501 | return false; |
1502 | |
1503 | // Skip "const" as it does not have an influence on whether this is a name. |
1504 | FormatToken *PreviousNotConst = Tok.getPreviousNonComment(); |
1505 | while (PreviousNotConst && PreviousNotConst->is(tok::kw_const)) |
1506 | PreviousNotConst = PreviousNotConst->getPreviousNonComment(); |
1507 | |
1508 | if (!PreviousNotConst) |
1509 | return false; |
1510 | |
1511 | bool IsPPKeyword = PreviousNotConst->is(tok::identifier) && |
1512 | PreviousNotConst->Previous && |
1513 | PreviousNotConst->Previous->is(tok::hash); |
1514 | |
1515 | if (PreviousNotConst->is(TT_TemplateCloser)) |
1516 | return PreviousNotConst && PreviousNotConst->MatchingParen && |
1517 | PreviousNotConst->MatchingParen->Previous && |
1518 | PreviousNotConst->MatchingParen->Previous->isNot(tok::period) && |
1519 | PreviousNotConst->MatchingParen->Previous->isNot(tok::kw_template); |
1520 | |
1521 | if (PreviousNotConst->is(tok::r_paren) && PreviousNotConst->MatchingParen && |
1522 | PreviousNotConst->MatchingParen->Previous && |
1523 | PreviousNotConst->MatchingParen->Previous->is(tok::kw_decltype)) |
1524 | return true; |
1525 | |
1526 | return (!IsPPKeyword && |
1527 | PreviousNotConst->isOneOf(tok::identifier, tok::kw_auto)) || |
1528 | PreviousNotConst->is(TT_PointerOrReference) || |
1529 | PreviousNotConst->isSimpleTypeSpecifier(); |
1530 | } |
1531 | |
1532 | /// Determine whether ')' is ending a cast. |
1533 | bool rParenEndsCast(const FormatToken &Tok) { |
1534 | // C-style casts are only used in C++ and Java. |
1535 | if (!Style.isCpp() && Style.Language != FormatStyle::LK_Java) |
1536 | return false; |
1537 | |
1538 | // Empty parens aren't casts and there are no casts at the end of the line. |
1539 | if (Tok.Previous == Tok.MatchingParen || !Tok.Next || !Tok.MatchingParen) |
1540 | return false; |
1541 | |
1542 | FormatToken *LeftOfParens = Tok.MatchingParen->getPreviousNonComment(); |
1543 | if (LeftOfParens) { |
1544 | // If there is a closing parenthesis left of the current parentheses, |
1545 | // look past it as these might be chained casts. |
1546 | if (LeftOfParens->is(tok::r_paren)) { |
1547 | if (!LeftOfParens->MatchingParen || |
1548 | !LeftOfParens->MatchingParen->Previous) |
1549 | return false; |
1550 | LeftOfParens = LeftOfParens->MatchingParen->Previous; |
1551 | } |
1552 | |
1553 | // If there is an identifier (or with a few exceptions a keyword) right |
1554 | // before the parentheses, this is unlikely to be a cast. |
1555 | if (LeftOfParens->Tok.getIdentifierInfo() && |
1556 | !LeftOfParens->isOneOf(Keywords.kw_in, tok::kw_return, tok::kw_case, |
1557 | tok::kw_delete)) |
1558 | return false; |
1559 | |
1560 | // Certain other tokens right before the parentheses are also signals that |
1561 | // this cannot be a cast. |
1562 | if (LeftOfParens->isOneOf(tok::at, tok::r_square, TT_OverloadedOperator, |
1563 | TT_TemplateCloser, tok::ellipsis)) |
1564 | return false; |
1565 | } |
1566 | |
1567 | if (Tok.Next->is(tok::question)) |
1568 | return false; |
1569 | |
1570 | // As Java has no function types, a "(" after the ")" likely means that this |
1571 | // is a cast. |
1572 | if (Style.Language == FormatStyle::LK_Java && Tok.Next->is(tok::l_paren)) |
1573 | return true; |
1574 | |
1575 | // If a (non-string) literal follows, this is likely a cast. |
1576 | if (Tok.Next->isNot(tok::string_literal) && |
1577 | (Tok.Next->Tok.isLiteral() || |
1578 | Tok.Next->isOneOf(tok::kw_sizeof, tok::kw_alignof))) |
1579 | return true; |
1580 | |
1581 | // Heuristically try to determine whether the parentheses contain a type. |
1582 | bool ParensAreType = |
1583 | !Tok.Previous || |
1584 | Tok.Previous->isOneOf(TT_PointerOrReference, TT_TemplateCloser) || |
1585 | Tok.Previous->isSimpleTypeSpecifier(); |
1586 | bool ParensCouldEndDecl = |
1587 | Tok.Next->isOneOf(tok::equal, tok::semi, tok::l_brace, tok::greater); |
1588 | if (ParensAreType && !ParensCouldEndDecl) |
1589 | return true; |
1590 | |
1591 | // At this point, we heuristically assume that there are no casts at the |
1592 | // start of the line. We assume that we have found most cases where there |
1593 | // are by the logic above, e.g. "(void)x;". |
1594 | if (!LeftOfParens) |
1595 | return false; |
1596 | |
1597 | // Certain token types inside the parentheses mean that this can't be a |
1598 | // cast. |
1599 | for (const FormatToken *Token = Tok.MatchingParen->Next; Token != &Tok; |
1600 | Token = Token->Next) |
1601 | if (Token->is(TT_BinaryOperator)) |
1602 | return false; |
1603 | |
1604 | // If the following token is an identifier or 'this', this is a cast. All |
1605 | // cases where this can be something else are handled above. |
1606 | if (Tok.Next->isOneOf(tok::identifier, tok::kw_this)) |
1607 | return true; |
1608 | |
1609 | if (!Tok.Next->Next) |
1610 | return false; |
1611 | |
1612 | // If the next token after the parenthesis is a unary operator, assume |
1613 | // that this is cast, unless there are unexpected tokens inside the |
1614 | // parenthesis. |
1615 | bool NextIsUnary = |
1616 | Tok.Next->isUnaryOperator() || Tok.Next->isOneOf(tok::amp, tok::star); |
1617 | if (!NextIsUnary || Tok.Next->is(tok::plus) || |
1618 | !Tok.Next->Next->isOneOf(tok::identifier, tok::numeric_constant)) |
1619 | return false; |
1620 | // Search for unexpected tokens. |
1621 | for (FormatToken *Prev = Tok.Previous; Prev != Tok.MatchingParen; |
1622 | Prev = Prev->Previous) { |
1623 | if (!Prev->isOneOf(tok::kw_const, tok::identifier, tok::coloncolon)) |
1624 | return false; |
1625 | } |
1626 | return true; |
1627 | } |
1628 | |
1629 | /// Return the type of the given token assuming it is * or &. |
1630 | TokenType determineStarAmpUsage(const FormatToken &Tok, bool IsExpression, |
1631 | bool InTemplateArgument) { |
1632 | if (Style.Language == FormatStyle::LK_JavaScript) |
1633 | return TT_BinaryOperator; |
1634 | |
1635 | const FormatToken *PrevToken = Tok.getPreviousNonComment(); |
1636 | if (!PrevToken) |
1637 | return TT_UnaryOperator; |
1638 | |
1639 | const FormatToken *NextToken = Tok.getNextNonComment(); |
1640 | if (!NextToken || |
1641 | NextToken->isOneOf(tok::arrow, tok::equal, tok::kw_const) || |
1642 | (NextToken->is(tok::l_brace) && !NextToken->getNextNonComment())) |
1643 | return TT_PointerOrReference; |
1644 | |
1645 | if (PrevToken->is(tok::coloncolon)) |
1646 | return TT_PointerOrReference; |
1647 | |
1648 | if (PrevToken->isOneOf(tok::l_paren, tok::l_square, tok::l_brace, |
1649 | tok::comma, tok::semi, tok::kw_return, tok::colon, |
1650 | tok::equal, tok::kw_delete, tok::kw_sizeof, |
1651 | tok::kw_throw) || |
1652 | PrevToken->isOneOf(TT_BinaryOperator, TT_ConditionalExpr, |
1653 | TT_UnaryOperator, TT_CastRParen)) |
1654 | return TT_UnaryOperator; |
1655 | |
1656 | if (NextToken->is(tok::l_square) && NextToken->isNot(TT_LambdaLSquare)) |
1657 | return TT_PointerOrReference; |
1658 | if (NextToken->is(tok::kw_operator) && !IsExpression) |
1659 | return TT_PointerOrReference; |
1660 | if (NextToken->isOneOf(tok::comma, tok::semi)) |
1661 | return TT_PointerOrReference; |
1662 | |
1663 | if (PrevToken->is(tok::r_paren) && PrevToken->MatchingParen) { |
1664 | FormatToken *TokenBeforeMatchingParen = |
1665 | PrevToken->MatchingParen->getPreviousNonComment(); |
1666 | if (TokenBeforeMatchingParen && |
1667 | TokenBeforeMatchingParen->isOneOf(tok::kw_typeof, tok::kw_decltype)) |
1668 | return TT_PointerOrReference; |
1669 | } |
1670 | |
1671 | if (PrevToken->Tok.isLiteral() || |
1672 | PrevToken->isOneOf(tok::r_paren, tok::r_square, tok::kw_true, |
1673 | tok::kw_false, tok::r_brace) || |
1674 | NextToken->Tok.isLiteral() || |
1675 | NextToken->isOneOf(tok::kw_true, tok::kw_false) || |
1676 | NextToken->isUnaryOperator() || |
1677 | // If we know we're in a template argument, there are no named |
1678 | // declarations. Thus, having an identifier on the right-hand side |
1679 | // indicates a binary operator. |
1680 | (InTemplateArgument && NextToken->Tok.isAnyIdentifier())) |
1681 | return TT_BinaryOperator; |
1682 | |
1683 | // "&&(" is quite unlikely to be two successive unary "&". |
1684 | if (Tok.is(tok::ampamp) && NextToken && NextToken->is(tok::l_paren)) |
1685 | return TT_BinaryOperator; |
1686 | |
1687 | // This catches some cases where evaluation order is used as control flow: |
1688 | // aaa && aaa->f(); |
1689 | const FormatToken *NextNextToken = NextToken->getNextNonComment(); |
1690 | if (NextNextToken && NextNextToken->is(tok::arrow)) |
1691 | return TT_BinaryOperator; |
1692 | |
1693 | // It is very unlikely that we are going to find a pointer or reference type |
1694 | // definition on the RHS of an assignment. |
1695 | if (IsExpression && !Contexts.back().CaretFound) |
1696 | return TT_BinaryOperator; |
1697 | |
1698 | return TT_PointerOrReference; |
1699 | } |
1700 | |
1701 | TokenType determinePlusMinusCaretUsage(const FormatToken &Tok) { |
1702 | const FormatToken *PrevToken = Tok.getPreviousNonComment(); |
1703 | if (!PrevToken) |
1704 | return TT_UnaryOperator; |
1705 | |
1706 | if (PrevToken->isOneOf(TT_CastRParen, TT_UnaryOperator)) |
1707 | // This must be a sequence of leading unary operators. |
1708 | return TT_UnaryOperator; |
1709 | |
1710 | // Use heuristics to recognize unary operators. |
1711 | if (PrevToken->isOneOf(tok::equal, tok::l_paren, tok::comma, tok::l_square, |
1712 | tok::question, tok::colon, tok::kw_return, |
1713 | tok::kw_case, tok::at, tok::l_brace)) |
1714 | return TT_UnaryOperator; |
1715 | |
1716 | // There can't be two consecutive binary operators. |
1717 | if (PrevToken->is(TT_BinaryOperator)) |
1718 | return TT_UnaryOperator; |
1719 | |
1720 | // Fall back to marking the token as binary operator. |
1721 | return TT_BinaryOperator; |
1722 | } |
1723 | |
1724 | /// Determine whether ++/-- are pre- or post-increments/-decrements. |
1725 | TokenType determineIncrementUsage(const FormatToken &Tok) { |
1726 | const FormatToken *PrevToken = Tok.getPreviousNonComment(); |
1727 | if (!PrevToken || PrevToken->is(TT_CastRParen)) |
1728 | return TT_UnaryOperator; |
1729 | if (PrevToken->isOneOf(tok::r_paren, tok::r_square, tok::identifier)) |
1730 | return TT_TrailingUnaryOperator; |
1731 | |
1732 | return TT_UnaryOperator; |
1733 | } |
1734 | |
1735 | SmallVector<Context, 8> Contexts; |
1736 | |
1737 | const FormatStyle &Style; |
1738 | AnnotatedLine &Line; |
1739 | FormatToken *CurrentToken; |
1740 | bool AutoFound; |
1741 | const AdditionalKeywords &Keywords; |
1742 | |
1743 | // Set of "<" tokens that do not open a template parameter list. If parseAngle |
1744 | // determines that a specific token can't be a template opener, it will make |
1745 | // same decision irrespective of the decisions for tokens leading up to it. |
1746 | // Store this information to prevent this from causing exponential runtime. |
1747 | llvm::SmallPtrSet<FormatToken *, 16> NonTemplateLess; |
1748 | }; |
1749 | |
1750 | static const int PrecedenceUnaryOperator = prec::PointerToMember + 1; |
1751 | static const int PrecedenceArrowAndPeriod = prec::PointerToMember + 2; |
1752 | |
1753 | /// Parses binary expressions by inserting fake parenthesis based on |
1754 | /// operator precedence. |
1755 | class ExpressionParser { |
1756 | public: |
1757 | ExpressionParser(const FormatStyle &Style, const AdditionalKeywords &Keywords, |
1758 | AnnotatedLine &Line) |
1759 | : Style(Style), Keywords(Keywords), Current(Line.First) {} |
1760 | |
1761 | /// Parse expressions with the given operator precedence. |
1762 | void parse(int Precedence = 0) { |
1763 | // Skip 'return' and ObjC selector colons as they are not part of a binary |
1764 | // expression. |
1765 | while (Current && (Current->is(tok::kw_return) || |
1766 | (Current->is(tok::colon) && |
1767 | Current->isOneOf(TT_ObjCMethodExpr, TT_DictLiteral)))) |
1768 | next(); |
1769 | |
1770 | if (!Current || Precedence > PrecedenceArrowAndPeriod) |
1771 | return; |
1772 | |
1773 | // Conditional expressions need to be parsed separately for proper nesting. |
1774 | if (Precedence == prec::Conditional) { |
1775 | parseConditionalExpr(); |
1776 | return; |
1777 | } |
1778 | |
1779 | // Parse unary operators, which all have a higher precedence than binary |
1780 | // operators. |
1781 | if (Precedence == PrecedenceUnaryOperator) { |
1782 | parseUnaryOperator(); |
1783 | return; |
1784 | } |
1785 | |
1786 | FormatToken *Start = Current; |
1787 | FormatToken *LatestOperator = nullptr; |
1788 | unsigned OperatorIndex = 0; |
1789 | |
1790 | while (Current) { |
1791 | // Consume operators with higher precedence. |
1792 | parse(Precedence + 1); |
1793 | |
1794 | int CurrentPrecedence = getCurrentPrecedence(); |
1795 | |
1796 | if (Current && Current->is(TT_SelectorName) && |
1797 | Precedence == CurrentPrecedence) { |
1798 | if (LatestOperator) |
1799 | addFakeParenthesis(Start, prec::Level(Precedence)); |
1800 | Start = Current; |
1801 | } |
1802 | |
1803 | // At the end of the line or when an operator with higher precedence is |
1804 | // found, insert fake parenthesis and return. |
1805 | if (!Current || |
1806 | (Current->closesScope() && |
1807 | (Current->MatchingParen || Current->is(TT_TemplateString))) || |
1808 | (CurrentPrecedence != -1 && CurrentPrecedence < Precedence) || |
1809 | (CurrentPrecedence == prec::Conditional && |
1810 | Precedence == prec::Assignment && Current->is(tok::colon))) { |
1811 | break; |
1812 | } |
1813 | |
1814 | // Consume scopes: (), [], <> and {} |
1815 | if (Current->opensScope()) { |
1816 | // In fragment of a JavaScript template string can look like '}..${' and |
1817 | // thus close a scope and open a new one at the same time. |
1818 | while (Current && (!Current->closesScope() || Current->opensScope())) { |
1819 | next(); |
1820 | parse(); |
1821 | } |
1822 | next(); |
1823 | } else { |
1824 | // Operator found. |
1825 | if (CurrentPrecedence == Precedence) { |
1826 | if (LatestOperator) |
1827 | LatestOperator->NextOperator = Current; |
1828 | LatestOperator = Current; |
1829 | Current->OperatorIndex = OperatorIndex; |
1830 | ++OperatorIndex; |
1831 | } |
1832 | next(/*SkipPastLeadingComments=*/Precedence > 0); |
1833 | } |
1834 | } |
1835 | |
1836 | if (LatestOperator && (Current || Precedence > 0)) { |
1837 | // LatestOperator->LastOperator = true; |
1838 | if (Precedence == PrecedenceArrowAndPeriod) { |
1839 | // Call expressions don't have a binary operator precedence. |
1840 | addFakeParenthesis(Start, prec::Unknown); |
1841 | } else { |
1842 | addFakeParenthesis(Start, prec::Level(Precedence)); |
1843 | } |
1844 | } |
1845 | } |
1846 | |
1847 | private: |
1848 | /// Gets the precedence (+1) of the given token for binary operators |
1849 | /// and other tokens that we treat like binary operators. |
1850 | int getCurrentPrecedence() { |
1851 | if (Current) { |
1852 | const FormatToken *NextNonComment = Current->getNextNonComment(); |
1853 | if (Current->is(TT_ConditionalExpr)) |
1854 | return prec::Conditional; |
1855 | if (NextNonComment && Current->is(TT_SelectorName) && |
1856 | (NextNonComment->isOneOf(TT_DictLiteral, TT_JsTypeColon) || |
1857 | ((Style.Language == FormatStyle::LK_Proto || |
1858 | Style.Language == FormatStyle::LK_TextProto) && |
1859 | NextNonComment->is(tok::less)))) |
1860 | return prec::Assignment; |
1861 | if (Current->is(TT_JsComputedPropertyName)) |
1862 | return prec::Assignment; |
1863 | if (Current->is(TT_LambdaArrow)) |
1864 | return prec::Comma; |
1865 | if (Current->is(TT_JsFatArrow)) |
1866 | return prec::Assignment; |
1867 | if (Current->isOneOf(tok::semi, TT_InlineASMColon, TT_SelectorName) || |
1868 | (Current->is(tok::comment) && NextNonComment && |
1869 | NextNonComment->is(TT_SelectorName))) |
1870 | return 0; |
1871 | if (Current->is(TT_RangeBasedForLoopColon)) |
1872 | return prec::Comma; |
1873 | if ((Style.Language == FormatStyle::LK_Java || |
1874 | Style.Language == FormatStyle::LK_JavaScript) && |
1875 | Current->is(Keywords.kw_instanceof)) |
1876 | return prec::Relational; |
1877 | if (Style.Language == FormatStyle::LK_JavaScript && |
1878 | Current->isOneOf(Keywords.kw_in, Keywords.kw_as)) |
1879 | return prec::Relational; |
1880 | if (Current->is(TT_BinaryOperator) || Current->is(tok::comma)) |
1881 | return Current->getPrecedence(); |
1882 | if (Current->isOneOf(tok::period, tok::arrow)) |
1883 | return PrecedenceArrowAndPeriod; |
1884 | if ((Style.Language == FormatStyle::LK_Java || |
1885 | Style.Language == FormatStyle::LK_JavaScript) && |
1886 | Current->isOneOf(Keywords.kw_extends, Keywords.kw_implements, |
1887 | Keywords.kw_throws)) |
1888 | return 0; |
1889 | } |
1890 | return -1; |
1891 | } |
1892 | |
1893 | void addFakeParenthesis(FormatToken *Start, prec::Level Precedence) { |
1894 | Start->FakeLParens.push_back(Precedence); |
1895 | if (Precedence > prec::Unknown) |
1896 | Start->StartsBinaryExpression = true; |
1897 | if (Current) { |
1898 | FormatToken *Previous = Current->Previous; |
1899 | while (Previous->is(tok::comment) && Previous->Previous) |
1900 | Previous = Previous->Previous; |
1901 | ++Previous->FakeRParens; |
1902 | if (Precedence > prec::Unknown) |
1903 | Previous->EndsBinaryExpression = true; |
1904 | } |
1905 | } |
1906 | |
1907 | /// Parse unary operator expressions and surround them with fake |
1908 | /// parentheses if appropriate. |
1909 | void parseUnaryOperator() { |
1910 | llvm::SmallVector<FormatToken *, 2> Tokens; |
1911 | while (Current && Current->is(TT_UnaryOperator)) { |
1912 | Tokens.push_back(Current); |
1913 | next(); |
1914 | } |
1915 | parse(PrecedenceArrowAndPeriod); |
1916 | for (FormatToken *Token : llvm::reverse(Tokens)) |
1917 | // The actual precedence doesn't matter. |
1918 | addFakeParenthesis(Token, prec::Unknown); |
1919 | } |
1920 | |
1921 | void parseConditionalExpr() { |
1922 | while (Current && Current->isTrailingComment()) { |
1923 | next(); |
1924 | } |
1925 | FormatToken *Start = Current; |
1926 | parse(prec::LogicalOr); |
1927 | if (!Current || !Current->is(tok::question)) |
1928 | return; |
1929 | next(); |
1930 | parse(prec::Assignment); |
1931 | if (!Current || Current->isNot(TT_ConditionalExpr)) |
1932 | return; |
1933 | next(); |
1934 | parse(prec::Assignment); |
1935 | addFakeParenthesis(Start, prec::Conditional); |
1936 | } |
1937 | |
1938 | void next(bool SkipPastLeadingComments = true) { |
1939 | if (Current) |
1940 | Current = Current->Next; |
1941 | while (Current && |
1942 | (Current->NewlinesBefore == 0 || SkipPastLeadingComments) && |
1943 | Current->isTrailingComment()) |
1944 | Current = Current->Next; |
1945 | } |
1946 | |
1947 | const FormatStyle &Style; |
1948 | const AdditionalKeywords &Keywords; |
1949 | FormatToken *Current; |
1950 | }; |
1951 | |
1952 | } // end anonymous namespace |
1953 | |
1954 | void TokenAnnotator::setCommentLineLevels( |
1955 | SmallVectorImpl<AnnotatedLine *> &Lines) { |
1956 | const AnnotatedLine *NextNonCommentLine = nullptr; |
1957 | for (SmallVectorImpl<AnnotatedLine *>::reverse_iterator I = Lines.rbegin(), |
1958 | E = Lines.rend(); |
1959 | I != E; ++I) { |
1960 | bool CommentLine = true; |
1961 | for (const FormatToken *Tok = (*I)->First; Tok; Tok = Tok->Next) { |
1962 | if (!Tok->is(tok::comment)) { |
1963 | CommentLine = false; |
1964 | break; |
1965 | } |
1966 | } |
1967 | |
1968 | // If the comment is currently aligned with the line immediately following |
1969 | // it, that's probably intentional and we should keep it. |
1970 | if (NextNonCommentLine && CommentLine && |
1971 | NextNonCommentLine->First->NewlinesBefore <= 1 && |
1972 | NextNonCommentLine->First->OriginalColumn == |
1973 | (*I)->First->OriginalColumn) { |
1974 | // Align comments for preprocessor lines with the # in column 0 if |
1975 | // preprocessor lines are not indented. Otherwise, align with the next |
1976 | // line. |
1977 | (*I)->Level = |
1978 | (Style.IndentPPDirectives != FormatStyle::PPDIS_BeforeHash && |
1979 | (NextNonCommentLine->Type == LT_PreprocessorDirective || |
1980 | NextNonCommentLine->Type == LT_ImportStatement)) |
1981 | ? 0 |
1982 | : NextNonCommentLine->Level; |
1983 | } else { |
1984 | NextNonCommentLine = (*I)->First->isNot(tok::r_brace) ? (*I) : nullptr; |
1985 | } |
1986 | |
1987 | setCommentLineLevels((*I)->Children); |
1988 | } |
1989 | } |
1990 | |
1991 | static unsigned maxNestingDepth(const AnnotatedLine &Line) { |
1992 | unsigned Result = 0; |
1993 | for (const auto *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) |
1994 | Result = std::max(Result, Tok->NestingLevel); |
1995 | return Result; |
1996 | } |
1997 | |
1998 | void TokenAnnotator::annotate(AnnotatedLine &Line) { |
1999 | for (SmallVectorImpl<AnnotatedLine *>::iterator I = Line.Children.begin(), |
2000 | E = Line.Children.end(); |
2001 | I != E; ++I) { |
2002 | annotate(**I); |
2003 | } |
2004 | AnnotatingParser Parser(Style, Line, Keywords); |
2005 | Line.Type = Parser.parseLine(); |
2006 | |
2007 | // With very deep nesting, ExpressionParser uses lots of stack and the |
2008 | // formatting algorithm is very slow. We're not going to do a good job here |
2009 | // anyway - it's probably generated code being formatted by mistake. |
2010 | // Just skip the whole line. |
2011 | if (maxNestingDepth(Line) > 50) |
2012 | Line.Type = LT_Invalid; |
2013 | |
2014 | if (Line.Type == LT_Invalid) |
2015 | return; |
2016 | |
2017 | ExpressionParser ExprParser(Style, Keywords, Line); |
2018 | ExprParser.parse(); |
2019 | |
2020 | if (Line.startsWith(TT_ObjCMethodSpecifier)) |
2021 | Line.Type = LT_ObjCMethodDecl; |
2022 | else if (Line.startsWith(TT_ObjCDecl)) |
2023 | Line.Type = LT_ObjCDecl; |
2024 | else if (Line.startsWith(TT_ObjCProperty)) |
2025 | Line.Type = LT_ObjCProperty; |
2026 | |
2027 | Line.First->SpacesRequiredBefore = 1; |
2028 | Line.First->CanBreakBefore = Line.First->MustBreakBefore; |
2029 | } |
2030 | |
2031 | // This function heuristically determines whether 'Current' starts the name of a |
2032 | // function declaration. |
2033 | static bool isFunctionDeclarationName(const FormatToken &Current, |
2034 | const AnnotatedLine &Line) { |
2035 | auto skipOperatorName = [](const FormatToken *Next) -> const FormatToken * { |
2036 | for (; Next; Next = Next->Next) { |
2037 | if (Next->is(TT_OverloadedOperatorLParen)) |
2038 | return Next; |
2039 | if (Next->is(TT_OverloadedOperator)) |
2040 | continue; |
2041 | if (Next->isOneOf(tok::kw_new, tok::kw_delete)) { |
2042 | // For 'new[]' and 'delete[]'. |
2043 | if (Next->Next && Next->Next->is(tok::l_square) && Next->Next->Next && |
2044 | Next->Next->Next->is(tok::r_square)) |
2045 | Next = Next->Next->Next; |
2046 | continue; |
2047 | } |
2048 | |
2049 | break; |
2050 | } |
2051 | return nullptr; |
2052 | }; |
2053 | |
2054 | // Find parentheses of parameter list. |
2055 | const FormatToken *Next = Current.Next; |
2056 | if (Current.is(tok::kw_operator)) { |
2057 | if (Current.Previous && Current.Previous->is(tok::coloncolon)) |
2058 | return false; |
2059 | Next = skipOperatorName(Next); |
2060 | } else { |
2061 | if (!Current.is(TT_StartOfName) || Current.NestingLevel != 0) |
2062 | return false; |
2063 | for (; Next; Next = Next->Next) { |
2064 | if (Next->is(TT_TemplateOpener)) { |
2065 | Next = Next->MatchingParen; |
2066 | } else if (Next->is(tok::coloncolon)) { |
2067 | Next = Next->Next; |
2068 | if (!Next) |
2069 | return false; |
2070 | if (Next->is(tok::kw_operator)) { |
2071 | Next = skipOperatorName(Next->Next); |
2072 | break; |
2073 | } |
2074 | if (!Next->is(tok::identifier)) |
2075 | return false; |
2076 | } else if (Next->is(tok::l_paren)) { |
2077 | break; |
2078 | } else { |
2079 | return false; |
2080 | } |
2081 | } |
2082 | } |
2083 | |
2084 | // Check whether parameter list can belong to a function declaration. |
2085 | if (!Next || !Next->is(tok::l_paren) || !Next->MatchingParen) |
2086 | return false; |
2087 | // If the lines ends with "{", this is likely an function definition. |
2088 | if (Line.Last->is(tok::l_brace)) |
2089 | return true; |
2090 | if (Next->Next == Next->MatchingParen) |
2091 | return true; // Empty parentheses. |
2092 | // If there is an &/&& after the r_paren, this is likely a function. |
2093 | if (Next->MatchingParen->Next && |
2094 | Next->MatchingParen->Next->is(TT_PointerOrReference)) |
2095 | return true; |
2096 | for (const FormatToken *Tok = Next->Next; Tok && Tok != Next->MatchingParen; |
2097 | Tok = Tok->Next) { |
2098 | if (Tok->is(tok::l_paren) && Tok->MatchingParen) { |
2099 | Tok = Tok->MatchingParen; |
2100 | continue; |
2101 | } |
2102 | if (Tok->is(tok::kw_const) || Tok->isSimpleTypeSpecifier() || |
2103 | Tok->isOneOf(TT_PointerOrReference, TT_StartOfName, tok::ellipsis)) |
2104 | return true; |
2105 | if (Tok->isOneOf(tok::l_brace, tok::string_literal, TT_ObjCMethodExpr) || |
2106 | Tok->Tok.isLiteral()) |
2107 | return false; |
2108 | } |
2109 | return false; |
2110 | } |
2111 | |
2112 | bool TokenAnnotator::mustBreakForReturnType(const AnnotatedLine &Line) const { |
2113 | assert(Line.MightBeFunctionDecl); |
2114 | |
2115 | if ((Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_TopLevel || |
2116 | Style.AlwaysBreakAfterReturnType == |
2117 | FormatStyle::RTBS_TopLevelDefinitions) && |
2118 | Line.Level > 0) |
2119 | return false; |
2120 | |
2121 | switch (Style.AlwaysBreakAfterReturnType) { |
2122 | case FormatStyle::RTBS_None: |
2123 | return false; |
2124 | case FormatStyle::RTBS_All: |
2125 | case FormatStyle::RTBS_TopLevel: |
2126 | return true; |
2127 | case FormatStyle::RTBS_AllDefinitions: |
2128 | case FormatStyle::RTBS_TopLevelDefinitions: |
2129 | return Line.mightBeFunctionDefinition(); |
2130 | } |
2131 | |
2132 | return false; |
2133 | } |
2134 | |
2135 | void TokenAnnotator::calculateFormattingInformation(AnnotatedLine &Line) { |
2136 | for (SmallVectorImpl<AnnotatedLine *>::iterator I = Line.Children.begin(), |
2137 | E = Line.Children.end(); |
2138 | I != E; ++I) { |
2139 | calculateFormattingInformation(**I); |
2140 | } |
2141 | |
2142 | Line.First->TotalLength = |
2143 | Line.First->IsMultiline ? Style.ColumnLimit |
2144 | : Line.FirstStartColumn + Line.First->ColumnWidth; |
2145 | FormatToken *Current = Line.First->Next; |
2146 | bool InFunctionDecl = Line.MightBeFunctionDecl; |
2147 | while (Current) { |
2148 | if (isFunctionDeclarationName(*Current, Line)) |
2149 | Current->Type = TT_FunctionDeclarationName; |
2150 | if (Current->is(TT_LineComment)) { |
2151 | if (Current->Previous->BlockKind == BK_BracedInit && |
2152 | Current->Previous->opensScope()) |
2153 | Current->SpacesRequiredBefore = Style.Cpp11BracedListStyle ? 0 : 1; |
2154 | else |
2155 | Current->SpacesRequiredBefore = Style.SpacesBeforeTrailingComments; |
2156 | |
2157 | // If we find a trailing comment, iterate backwards to determine whether |
2158 | // it seems to relate to a specific parameter. If so, break before that |
2159 | // parameter to avoid changing the comment's meaning. E.g. don't move 'b' |
2160 | // to the previous line in: |
2161 | // SomeFunction(a, |
2162 | // b, // comment |
2163 | // c); |
2164 | if (!Current->HasUnescapedNewline) { |
2165 | for (FormatToken *Parameter = Current->Previous; Parameter; |
2166 | Parameter = Parameter->Previous) { |
2167 | if (Parameter->isOneOf(tok::comment, tok::r_brace)) |
2168 | break; |
2169 | if (Parameter->Previous && Parameter->Previous->is(tok::comma)) { |
2170 | if (!Parameter->Previous->is(TT_CtorInitializerComma) && |
2171 | Parameter->HasUnescapedNewline) |
2172 | Parameter->MustBreakBefore = true; |
2173 | break; |
2174 | } |
2175 | } |
2176 | } |
2177 | } else if (Current->SpacesRequiredBefore == 0 && |
2178 | spaceRequiredBefore(Line, *Current)) { |
2179 | Current->SpacesRequiredBefore = 1; |
2180 | } |
2181 | |
2182 | Current->MustBreakBefore = |
2183 | Current->MustBreakBefore || mustBreakBefore(Line, *Current); |
2184 | |
2185 | if (!Current->MustBreakBefore && InFunctionDecl && |
2186 | Current->is(TT_FunctionDeclarationName)) |
2187 | Current->MustBreakBefore = mustBreakForReturnType(Line); |
2188 | |
2189 | Current->CanBreakBefore = |
2190 | Current->MustBreakBefore || canBreakBefore(Line, *Current); |
2191 | unsigned ChildSize = 0; |
2192 | if (Current->Previous->Children.size() == 1) { |
2193 | FormatToken &LastOfChild = *Current->Previous->Children[0]->Last; |
2194 | ChildSize = LastOfChild.isTrailingComment() ? Style.ColumnLimit |
2195 | : LastOfChild.TotalLength + 1; |
2196 | } |
2197 | const FormatToken *Prev = Current->Previous; |
2198 | if (Current->MustBreakBefore || Prev->Children.size() > 1 || |
2199 | (Prev->Children.size() == 1 && |
2200 | Prev->Children[0]->First->MustBreakBefore) || |
2201 | Current->IsMultiline) |
2202 | Current->TotalLength = Prev->TotalLength + Style.ColumnLimit; |
2203 | else |
2204 | Current->TotalLength = Prev->TotalLength + Current->ColumnWidth + |
2205 | ChildSize + Current->SpacesRequiredBefore; |
2206 | |
2207 | if (Current->is(TT_CtorInitializerColon)) |
2208 | InFunctionDecl = false; |
2209 | |
2210 | // FIXME: Only calculate this if CanBreakBefore is true once static |
2211 | // initializers etc. are sorted out. |
2212 | // FIXME: Move magic numbers to a better place. |
2213 | |
2214 | // Reduce penalty for aligning ObjC method arguments using the colon |
2215 | // alignment as this is the canonical way (still prefer fitting everything |
2216 | // into one line if possible). Trying to fit a whole expression into one |
2217 | // line should not force other line breaks (e.g. when ObjC method |
2218 | // expression is a part of other expression). |
2219 | Current->SplitPenalty = splitPenalty(Line, *Current, InFunctionDecl); |
2220 | if (Style.Language == FormatStyle::LK_ObjC && |
2221 | Current->is(TT_SelectorName) && Current->ParameterIndex > 0) { |
2222 | if (Current->ParameterIndex == 1) |
2223 | Current->SplitPenalty += 5 * Current->BindingStrength; |
2224 | } else { |
2225 | Current->SplitPenalty += 20 * Current->BindingStrength; |
2226 | } |
2227 | |
2228 | Current = Current->Next; |
2229 | } |
2230 | |
2231 | calculateUnbreakableTailLengths(Line); |
2232 | unsigned IndentLevel = Line.Level; |
2233 | for (Current = Line.First; Current != nullptr; Current = Current->Next) { |
2234 | if (Current->Role) |
2235 | Current->Role->precomputeFormattingInfos(Current); |
2236 | if (Current->MatchingParen && |
2237 | Current->MatchingParen->opensBlockOrBlockTypeList(Style)) { |
2238 | 0", "/home/seafit/code_projects/clang_source/clang/lib/Format/TokenAnnotator.cpp", 2238, __PRETTY_FUNCTION__))" file_link="../../../include/assert.h.html#88" macro="true">assert(IndentLevel > 0); |
2239 | --IndentLevel; |
2240 | } |
2241 | Current->IndentLevel = IndentLevel; |
2242 | if (Current->opensBlockOrBlockTypeList(Style)) |
2243 | ++IndentLevel; |
2244 | } |
2245 | |
2246 | LLVM_DEBUG({ printDebugInfo(Line); }); |
2247 | } |
2248 | |
2249 | void TokenAnnotator::calculateUnbreakableTailLengths(AnnotatedLine &Line) { |
2250 | unsigned UnbreakableTailLength = 0; |
2251 | FormatToken *Current = Line.Last; |
2252 | while (Current) { |
2253 | Current->UnbreakableTailLength = UnbreakableTailLength; |
2254 | if (Current->CanBreakBefore || |
2255 | Current->isOneOf(tok::comment, tok::string_literal)) { |
2256 | UnbreakableTailLength = 0; |
2257 | } else { |
2258 | UnbreakableTailLength += |
2259 | Current->ColumnWidth + Current->SpacesRequiredBefore; |
2260 | } |
2261 | Current = Current->Previous; |
2262 | } |
2263 | } |
2264 | |
2265 | unsigned TokenAnnotator::splitPenalty(const AnnotatedLine &Line, |
2266 | const FormatToken &Tok, |
2267 | bool InFunctionDecl) { |
2268 | const FormatToken &Left = *Tok.Previous; |
2269 | const FormatToken &Right = Tok; |
2270 | |
2271 | if (Left.is(tok::semi)) |
2272 | return 0; |
2273 | |
2274 | if (Style.Language == FormatStyle::LK_Java) { |
2275 | if (Right.isOneOf(Keywords.kw_extends, Keywords.kw_throws)) |
2276 | return 1; |
2277 | if (Right.is(Keywords.kw_implements)) |
2278 | return 2; |
2279 | if (Left.is(tok::comma) && Left.NestingLevel == 0) |
2280 | return 3; |
2281 | } else if (Style.Language == FormatStyle::LK_JavaScript) { |
2282 | if (Right.is(Keywords.kw_function) && Left.isNot(tok::comma)) |
2283 | return 100; |
2284 | if (Left.is(TT_JsTypeColon)) |
2285 | return 35; |
2286 | if ((Left.is(TT_TemplateString) && Left.TokenText.endswith("${")) || |
2287 | (Right.is(TT_TemplateString) && Right.TokenText.startswith("}"))) |
2288 | return 100; |
2289 | // Prefer breaking call chains (".foo") over empty "{}", "[]" or "()". |
2290 | if (Left.opensScope() && Right.closesScope()) |
2291 | return 200; |
2292 | } |
2293 | |
2294 | if (Right.is(tok::identifier) && Right.Next && Right.Next->is(TT_DictLiteral)) |
2295 | return 1; |
2296 | if (Right.is(tok::l_square)) { |
2297 | if (Style.Language == FormatStyle::LK_Proto) |
2298 | return 1; |
2299 | if (Left.is(tok::r_square)) |
2300 | return 200; |
2301 | // Slightly prefer formatting local lambda definitions like functions. |
2302 | if (Right.is(TT_LambdaLSquare) && Left.is(tok::equal)) |
2303 | return 35; |
2304 | if (!Right.isOneOf(TT_ObjCMethodExpr, TT_LambdaLSquare, |
2305 | TT_ArrayInitializerLSquare, |
2306 | TT_DesignatedInitializerLSquare, TT_AttributeSquare)) |
2307 | return 500; |
2308 | } |
2309 | |
2310 | if (Left.is(tok::coloncolon) || |
2311 | (Right.is(tok::period) && Style.Language == FormatStyle::LK_Proto)) |
2312 | return 500; |
2313 | if (Right.isOneOf(TT_StartOfName, TT_FunctionDeclarationName) || |
2314 | Right.is(tok::kw_operator)) { |
2315 | if (Line.startsWith(tok::kw_for) && Right.PartOfMultiVariableDeclStmt) |
2316 | return 3; |
2317 | if (Left.is(TT_StartOfName)) |
2318 | return 110; |
2319 | if (InFunctionDecl && Right.NestingLevel == 0) |
2320 | return Style.PenaltyReturnTypeOnItsOwnLine; |
2321 | return 200; |
2322 | } |
2323 | if (Right.is(TT_PointerOrReference)) |
2324 | return 190; |
2325 | if (Right.is(TT_LambdaArrow)) |
2326 | return 110; |
2327 | if (Left.is(tok::equal) && Right.is(tok::l_brace)) |
2328 | return 160; |
2329 | if (Left.is(TT_CastRParen)) |
2330 | return 100; |
2331 | if (Left.isOneOf(tok::kw_class, tok::kw_struct)) |
2332 | return 5000; |
2333 | if (Left.is(tok::comment)) |
2334 | return 1000; |
2335 | |
2336 | if (Left.isOneOf(TT_RangeBasedForLoopColon, TT_InheritanceColon, |
2337 | TT_CtorInitializerColon)) |
2338 | return 2; |
2339 | |
2340 | if (Right.isMemberAccess()) { |
2341 | // Breaking before the "./->" of a chained call/member access is reasonably |
2342 | // cheap, as formatting those with one call per line is generally |
2343 | // desirable. In particular, it should be cheaper to break before the call |
2344 | // than it is to break inside a call's parameters, which could lead to weird |
2345 | // "hanging" indents. The exception is the very last "./->" to support this |
2346 | // frequent pattern: |
2347 | // |
2348 | // aaaaaaaa.aaaaaaaa.bbbbbbb().ccccccccccccccccccccc( |
2349 | // dddddddd); |
2350 | // |
2351 | // which might otherwise be blown up onto many lines. Here, clang-format |
2352 | // won't produce "hanging" indents anyway as there is no other trailing |
2353 | // call. |
2354 | // |
2355 | // Also apply higher penalty is not a call as that might lead to a wrapping |
2356 | // like: |
2357 | // |
2358 | // aaaaaaa |
2359 | // .aaaaaaaaa.bbbbbbbb(cccccccc); |
2360 | return !Right.NextOperator || !Right.NextOperator->Previous->closesScope() |
2361 | ? 150 |
2362 | : 35; |
2363 | } |
2364 | |
2365 | if (Right.is(TT_TrailingAnnotation) && |
2366 | (!Right.Next || Right.Next->isNot(tok::l_paren))) { |
2367 | // Moving trailing annotations to the next line is fine for ObjC method |
2368 | // declarations. |
2369 | if (Line.startsWith(TT_ObjCMethodSpecifier)) |
2370 | return 10; |
2371 | // Generally, breaking before a trailing annotation is bad unless it is |
2372 | // function-like. It seems to be especially preferable to keep standard |
2373 | // annotations (i.e. "const", "final" and "override") on the same line. |
2374 | // Use a slightly higher penalty after ")" so that annotations like |
2375 | // "const override" are kept together. |
2376 | bool is_short_annotation = Right.TokenText.size() < 10; |
2377 | return (Left.is(tok::r_paren) ? 100 : 120) + (is_short_annotation ? 50 : 0); |
2378 | } |
2379 | |
2380 | // In for-loops, prefer breaking at ',' and ';'. |
2381 | if (Line.startsWith(tok::kw_for) && Left.is(tok::equal)) |
2382 | return 4; |
2383 | |
2384 | // In Objective-C method expressions, prefer breaking before "param:" over |
2385 | // breaking after it. |
2386 | if (Right.is(TT_SelectorName)) |
2387 | return 0; |
2388 | if (Left.is(tok::colon) && Left.is(TT_ObjCMethodExpr)) |
2389 | return Line.MightBeFunctionDecl ? 50 : 500; |
2390 | |
2391 | // In Objective-C type declarations, avoid breaking after the category's |
2392 | // open paren (we'll prefer breaking after the protocol list's opening |
2393 | // angle bracket, if present). |
2394 | if (Line.Type == LT_ObjCDecl && Left.is(tok::l_paren) && Left.Previous && |
2395 | Left.Previous->isOneOf(tok::identifier, tok::greater)) |
2396 | return 500; |
2397 | |
2398 | if (Left.is(tok::l_paren) && InFunctionDecl && |
2399 | Style.AlignAfterOpenBracket != FormatStyle::BAS_DontAlign) |
2400 | return 100; |
2401 | if (Left.is(tok::l_paren) && Left.Previous && |
2402 | (Left.Previous->isOneOf(tok::kw_if, tok::kw_for) || |
2403 | Left.Previous->endsSequence(tok::kw_constexpr, tok::kw_if))) |
2404 | return 1000; |
2405 | if (Left.is(tok::equal) && InFunctionDecl) |
2406 | return 110; |
2407 | if (Right.is(tok::r_brace)) |
2408 | return 1; |
2409 | if (Left.is(TT_TemplateOpener)) |
2410 | return 100; |
2411 | if (Left.opensScope()) { |
2412 | if (Style.AlignAfterOpenBracket == FormatStyle::BAS_DontAlign) |
2413 | return 0; |
2414 | if (Left.is(tok::l_brace) && !Style.Cpp11BracedListStyle) |
2415 | return 19; |
2416 | return Left.ParameterCount > 1 ? Style.PenaltyBreakBeforeFirstCallParameter |
2417 | : 19; |
2418 | } |
2419 | if (Left.is(TT_JavaAnnotation)) |
2420 | return 50; |
2421 | |
2422 | if (Left.isOneOf(tok::plus, tok::comma) && Left.Previous && |
2423 | Left.Previous->isLabelString() && |
2424 | (Left.NextOperator || Left.OperatorIndex != 0)) |
2425 | return 50; |
2426 | if (Right.is(tok::plus) && Left.isLabelString() && |
2427 | (Right.NextOperator || Right.OperatorIndex != 0)) |
2428 | return 25; |
2429 | if (Left.is(tok::comma)) |
2430 | return 1; |
2431 | if (Right.is(tok::lessless) && Left.isLabelString() && |
2432 | (Right.NextOperator || Right.OperatorIndex != 1)) |
2433 | return 25; |
2434 | if (Right.is(tok::lessless)) { |
2435 | // Breaking at a << is really cheap. |
2436 | if (!Left.is(tok::r_paren) || Right.OperatorIndex > 0) |
2437 | // Slightly prefer to break before the first one in log-like statements. |
2438 | return 2; |
2439 | return 1; |
2440 | } |
2441 | if (Left.ClosesTemplateDeclaration) |
2442 | return Style.PenaltyBreakTemplateDeclaration; |
2443 | if (Left.is(TT_ConditionalExpr)) |
2444 | return prec::Conditional; |
2445 | prec::Level Level = Left.getPrecedence(); |
2446 | if (Level == prec::Unknown) |
2447 | Level = Right.getPrecedence(); |
2448 | if (Level == prec::Assignment) |
2449 | return Style.PenaltyBreakAssignment; |
2450 | if (Level != prec::Unknown) |
2451 | return Level; |
2452 | |
2453 | return 3; |
2454 | } |
2455 | |
2456 | bool TokenAnnotator::spaceRequiredBetween(const AnnotatedLine &Line, |
2457 | const FormatToken &Left, |
2458 | const FormatToken &Right) { |
2459 | if (Left.is(tok::kw_return) && Right.isNot(tok::semi)) |
2460 | return true; |
2461 | if (Left.is(Keywords.kw_assert) && Style.Language == FormatStyle::LK_Java) |
2462 | return true; |
2463 | if (Style.ObjCSpaceAfterProperty && Line.Type == LT_ObjCProperty && |
2464 | Left.Tok.getObjCKeywordID() == tok::objc_property) |
2465 | return true; |
2466 | if (Right.is(tok::hashhash)) |
2467 | return Left.is(tok::hash); |
2468 | if (Left.isOneOf(tok::hashhash, tok::hash)) |
2469 | return Right.is(tok::hash); |
2470 | if (Left.is(tok::l_paren) && Right.is(tok::r_paren)) |
2471 | return Style.SpaceInEmptyParentheses; |
2472 | if (Left.is(tok::l_paren) || Right.is(tok::r_paren)) |
2473 | return (Right.is(TT_CastRParen) || |
2474 | (Left.MatchingParen && Left.MatchingParen->is(TT_CastRParen))) |
2475 | ? Style.SpacesInCStyleCastParentheses |
2476 | : Style.SpacesInParentheses; |
2477 | if (Right.isOneOf(tok::semi, tok::comma)) |
2478 | return false; |
2479 | if (Right.is(tok::less) && Line.Type == LT_ObjCDecl) { |
2480 | bool IsLightweightGeneric = Right.MatchingParen && |
2481 | Right.MatchingParen->Next && |
2482 | Right.MatchingParen->Next->is(tok::colon); |
2483 | return !IsLightweightGeneric && Style.ObjCSpaceBeforeProtocolList; |
2484 | } |
2485 | if (Right.is(tok::less) && Left.is(tok::kw_template)) |
2486 | return Style.SpaceAfterTemplateKeyword; |
2487 | if (Left.isOneOf(tok::exclaim, tok::tilde)) |
2488 | return false; |
2489 | if (Left.is(tok::at) && |
2490 | Right.isOneOf(tok::identifier, tok::string_literal, tok::char_constant, |
2491 | tok::numeric_constant, tok::l_paren, tok::l_brace, |
2492 | tok::kw_true, tok::kw_false)) |
2493 | return false; |
2494 | if (Left.is(tok::colon)) |
2495 | return !Left.is(TT_ObjCMethodExpr); |
2496 | if (Left.is(tok::coloncolon)) |
2497 | return false; |
2498 | if (Left.is(tok::less) || Right.isOneOf(tok::greater, tok::less)) { |
2499 | if (Style.Language == FormatStyle::LK_TextProto || |
2500 | (Style.Language == FormatStyle::LK_Proto && |
2501 | (Left.is(TT_DictLiteral) || Right.is(TT_DictLiteral)))) { |
2502 | // Format empty list as `<>`. |
2503 | if (Left.is(tok::less) && Right.is(tok::greater)) |
2504 | return false; |
2505 | return !Style.Cpp11BracedListStyle; |
2506 | } |
2507 | return false; |
2508 | } |
2509 | if (Right.is(tok::ellipsis)) |
2510 | return Left.Tok.isLiteral() || (Left.is(tok::identifier) && Left.Previous && |
2511 | Left.Previous->is(tok::kw_case)); |
2512 | if (Left.is(tok::l_square) && Right.is(tok::amp)) |
2513 | return false; |
2514 | if (Right.is(TT_PointerOrReference)) { |
2515 | if (Left.is(tok::r_paren) && Line.MightBeFunctionDecl) { |
2516 | if (!Left.MatchingParen) |
2517 | return true; |
2518 | FormatToken *TokenBeforeMatchingParen = |
2519 | Left.MatchingParen->getPreviousNonComment(); |
2520 | if (!TokenBeforeMatchingParen || |
2521 | !TokenBeforeMatchingParen->isOneOf(tok::kw_typeof, tok::kw_decltype)) |
2522 | return true; |
2523 | } |
2524 | return (Left.Tok.isLiteral() || |
2525 | (!Left.isOneOf(TT_PointerOrReference, tok::l_paren) && |
2526 | (Style.PointerAlignment != FormatStyle::PAS_Left || |
2527 | (Line.IsMultiVariableDeclStmt && |
2528 | (Left.NestingLevel == 0 || |
2529 | (Left.NestingLevel == 1 && Line.First->is(tok::kw_for))))))); |
2530 | } |
2531 | if (Right.is(TT_FunctionTypeLParen) && Left.isNot(tok::l_paren) && |
2532 | (!Left.is(TT_PointerOrReference) || |
2533 | (Style.PointerAlignment != FormatStyle::PAS_Right && |
2534 | !Line.IsMultiVariableDeclStmt))) |
2535 | return true; |
2536 | if (Left.is(TT_PointerOrReference)) |
2537 | return Right.Tok.isLiteral() || Right.is(TT_BlockComment) || |
2538 | (Right.isOneOf(Keywords.kw_override, Keywords.kw_final) && |
2539 | !Right.is(TT_StartOfName)) || |
2540 | (Right.is(tok::l_brace) && Right.BlockKind == BK_Block) || |
2541 | (!Right.isOneOf(TT_PointerOrReference, TT_ArraySubscriptLSquare, |
2542 | tok::l_paren) && |
2543 | (Style.PointerAlignment != FormatStyle::PAS_Right && |
2544 | !Line.IsMultiVariableDeclStmt) && |
2545 | Left.Previous && |
2546 | !Left.Previous->isOneOf(tok::l_paren, tok::coloncolon)); |
2547 | if (Right.is(tok::star) && Left.is(tok::l_paren)) |
2548 | return false; |
2549 | const auto SpaceRequiredForArrayInitializerLSquare = |
2550 | [](const FormatToken &LSquareTok, const FormatStyle &Style) { |
2551 | return Style.SpacesInContainerLiterals || |
2552 | ((Style.Language == FormatStyle::LK_Proto || |
2553 | Style.Language == FormatStyle::LK_TextProto) && |
2554 | !Style.Cpp11BracedListStyle && |
2555 | LSquareTok.endsSequence(tok::l_square, tok::colon, |
2556 | TT_SelectorName)); |
2557 | }; |
2558 | if (Left.is(tok::l_square)) |
2559 | return (Left.is(TT_ArrayInitializerLSquare) && Right.isNot(tok::r_square) && |
2560 | SpaceRequiredForArrayInitializerLSquare(Left, Style)) || |
2561 | (Left.isOneOf(TT_ArraySubscriptLSquare, |
2562 | TT_StructuredBindingLSquare) && |
2563 | Style.SpacesInSquareBrackets && Right.isNot(tok::r_square)); |
2564 | if (Right.is(tok::r_square)) |
2565 | return Right.MatchingParen && |
2566 | ((Right.MatchingParen->is(TT_ArrayInitializerLSquare) && |
2567 | SpaceRequiredForArrayInitializerLSquare(*Right.MatchingParen, |
2568 | Style)) || |
2569 | (Style.SpacesInSquareBrackets && |
2570 | Right.MatchingParen->isOneOf(TT_ArraySubscriptLSquare, |
2571 | TT_StructuredBindingLSquare)) || |
2572 | Right.MatchingParen->is(TT_AttributeParen)); |
2573 | if (Right.is(tok::l_square) && |
2574 | !Right.isOneOf(TT_ObjCMethodExpr, TT_LambdaLSquare, |
2575 | TT_DesignatedInitializerLSquare, |
2576 | TT_StructuredBindingLSquare, TT_AttributeSquare) && |
2577 | !Left.isOneOf(tok::numeric_constant, TT_DictLiteral)) |
2578 | return false; |
2579 | if (Left.is(tok::l_brace) && Right.is(tok::r_brace)) |
2580 | return !Left.Children.empty(); // No spaces in "{}". |
2581 | if ((Left.is(tok::l_brace) && Left.BlockKind != BK_Block) || |
2582 | (Right.is(tok::r_brace) && Right.MatchingParen && |
2583 | Right.MatchingParen->BlockKind != BK_Block)) |
2584 | return !Style.Cpp11BracedListStyle; |
2585 | if (Left.is(TT_BlockComment)) |
2586 | // No whitespace in x(/*foo=*/1), except for JavaScript. |
2587 | return Style.Language == FormatStyle::LK_JavaScript || |
2588 | !Left.TokenText.endswith("=*/"); |
2589 | if (Right.is(tok::l_paren)) { |
2590 | if ((Left.is(tok::r_paren) && Left.is(TT_AttributeParen)) || |
2591 | (Left.is(tok::r_square) && Left.is(TT_AttributeSquare))) |
2592 | return true; |
2593 | return Line.Type == LT_ObjCDecl || Left.is(tok::semi) || |
2594 | (Style.SpaceBeforeParens != FormatStyle::SBPO_Never && |
2595 | (Left.isOneOf(tok::kw_if, tok::pp_elif, tok::kw_for, tok::kw_while, |
2596 | tok::kw_switch, tok::kw_case, TT_ForEachMacro, |
2597 | TT_ObjCForIn) || |
2598 | Left.endsSequence(tok::kw_constexpr, tok::kw_if) || |
2599 | (Left.isOneOf(tok::kw_try, Keywords.kw___except, tok::kw_catch, |
2600 | tok::kw_new, tok::kw_delete) && |
2601 | (!Left.Previous || Left.Previous->isNot(tok::period))))) || |
2602 | (Style.SpaceBeforeParens == FormatStyle::SBPO_Always && |
2603 | (Left.is(tok::identifier) || Left.isFunctionLikeKeyword() || |
2604 | Left.is(tok::r_paren) || |
2605 | (Left.is(tok::r_square) && Left.MatchingParen && |
2606 | Left.MatchingParen->is(TT_LambdaLSquare))) && |
2607 | Line.Type != LT_PreprocessorDirective); |
2608 | } |
2609 | if (Left.is(tok::at) && Right.Tok.getObjCKeywordID() != tok::objc_not_keyword) |
2610 | return false; |
2611 | if (Right.is(TT_UnaryOperator)) |
2612 | return !Left.isOneOf(tok::l_paren, tok::l_square, tok::at) && |
2613 | (Left.isNot(tok::colon) || Left.isNot(TT_ObjCMethodExpr)); |
2614 | if ((Left.isOneOf(tok::identifier, tok::greater, tok::r_square, |
2615 | tok::r_paren) || |
2616 | Left.isSimpleTypeSpecifier()) && |
2617 | Right.is(tok::l_brace) && Right.getNextNonComment() && |
2618 | Right.BlockKind != BK_Block) |
2619 | return false; |
2620 | if (Left.is(tok::period) || Right.is(tok::period)) |
2621 | return false; |
2622 | if (Right.is(tok::hash) && Left.is(tok::identifier) && Left.TokenText == "L") |
2623 | return false; |
2624 | if (Left.is(TT_TemplateCloser) && Left.MatchingParen && |
2625 | Left.MatchingParen->Previous && |
2626 | (Left.MatchingParen->Previous->is(tok::period) || |
2627 | Left.MatchingParen->Previous->is(tok::coloncolon))) |
2628 | // Java call to generic function with explicit type: |
2629 | // A.<B<C<...>>>DoSomething(); |
2630 | // A::<B<C<...>>>DoSomething(); // With a Java 8 method reference. |
2631 | return false; |
2632 | if (Left.is(TT_TemplateCloser) && Right.is(tok::l_square)) |
2633 | return false; |
2634 | if (Left.is(tok::l_brace) && Left.endsSequence(TT_DictLiteral, tok::at)) |
2635 | // Objective-C dictionary literal -> no space after opening brace. |
2636 | return false; |
2637 | if (Right.is(tok::r_brace) && Right.MatchingParen && |
2638 | Right.MatchingParen->endsSequence(TT_DictLiteral, tok::at)) |
2639 | // Objective-C dictionary literal -> no space before closing brace. |
2640 | return false; |
2641 | return true; |
2642 | } |
2643 | |
2644 | bool TokenAnnotator::spaceRequiredBefore(const AnnotatedLine &Line, |
2645 | const FormatToken &Right) { |
2646 | const FormatToken &Left = *Right.Previous; |
2647 | if (Right.Tok.getIdentifierInfo() && Left.Tok.getIdentifierInfo()) |
2648 | return true; // Never ever merge two identifiers. |
2649 | if (Style.isCpp()) { |
2650 | if (Left.is(tok::kw_operator)) |
2651 | return Right.is(tok::coloncolon); |
2652 | if (Right.is(tok::l_brace) && Right.BlockKind == BK_BracedInit && |
2653 | !Left.opensScope() && Style.SpaceBeforeCpp11BracedList) |
2654 | return true; |
2655 | } else if (Style.Language == FormatStyle::LK_Proto || |
2656 | Style.Language == FormatStyle::LK_TextProto) { |
2657 | if (Right.is(tok::period) && |
2658 | Left.isOneOf(Keywords.kw_optional, Keywords.kw_required, |
2659 | Keywords.kw_repeated, Keywords.kw_extend)) |
2660 | return true; |
2661 | if (Right.is(tok::l_paren) && |
2662 | Left.isOneOf(Keywords.kw_returns, Keywords.kw_option)) |
2663 | return true; |
2664 | if (Right.isOneOf(tok::l_brace, tok::less) && Left.is(TT_SelectorName)) |
2665 | return true; |
2666 | // Slashes occur in text protocol extension syntax: [type/type] { ... }. |
2667 | if (Left.is(tok::slash) || Right.is(tok::slash)) |
2668 | return false; |
2669 | if (Left.MatchingParen && |
2670 | Left.MatchingParen->is(TT_ProtoExtensionLSquare) && |
2671 | Right.isOneOf(tok::l_brace, tok::less)) |
2672 | return !Style.Cpp11BracedListStyle; |
2673 | // A percent is probably part of a formatting specification, such as %lld. |
2674 | if (Left.is(tok::percent)) |
2675 | return false; |
2676 | // Preserve the existence of a space before a percent for cases like 0x%04x |
2677 | // and "%d %d" |
2678 | if (Left.is(tok::numeric_constant) && Right.is(tok::percent)) |
2679 | return Right.WhitespaceRange.getEnd() != Right.WhitespaceRange.getBegin(); |
2680 | } else if (Style.Language == FormatStyle::LK_JavaScript || Style.isCSharp()) { |
2681 | if (Left.is(TT_JsFatArrow)) |
2682 | return true; |
2683 | // for await ( ... |
2684 | if (Right.is(tok::l_paren) && Left.is(Keywords.kw_await) && Left.Previous && |
2685 | Left.Previous->is(tok::kw_for)) |
2686 | return true; |
2687 | if (Left.is(Keywords.kw_async) && Right.is(tok::l_paren) && |
2688 | Right.MatchingParen) { |
2689 | const FormatToken *Next = Right.MatchingParen->getNextNonComment(); |
2690 | // An async arrow function, for example: `x = async () => foo();`, |
2691 | // as opposed to calling a function called async: `x = async();` |
2692 | if (Next && Next->is(TT_JsFatArrow)) |
2693 | return true; |
2694 | } |
2695 | if ((Left.is(TT_TemplateString) && Left.TokenText.endswith("${")) || |
2696 | (Right.is(TT_TemplateString) && Right.TokenText.startswith("}"))) |
2697 | return false; |
2698 | // In tagged template literals ("html`bar baz`"), there is no space between |
2699 | // the tag identifier and the template string. getIdentifierInfo makes sure |
2700 | // that the identifier is not a pseudo keyword like `yield`, either. |
2701 | if (Left.is(tok::identifier) && Keywords.IsJavaScriptIdentifier(Left) && |
2702 | Right.is(TT_TemplateString)) |
2703 | return false; |
2704 | if (Right.is(tok::star) && |
2705 | Left.isOneOf(Keywords.kw_function, Keywords.kw_yield)) |
2706 | return false; |
2707 | if (Right.isOneOf(tok::l_brace, tok::l_square) && |
2708 | Left.isOneOf(Keywords.kw_function, Keywords.kw_yield, |
2709 | Keywords.kw_extends, Keywords.kw_implements)) |
2710 | return true; |
2711 | if (Right.is(tok::l_paren)) { |
2712 | // JS methods can use some keywords as names (e.g. `delete()`). |
2713 | if (Line.MustBeDeclaration && Left.Tok.getIdentifierInfo()) |
2714 | return false; |
2715 | // Valid JS method names can include keywords, e.g. `foo.delete()` or |
2716 | // `bar.instanceof()`. Recognize call positions by preceding period. |
2717 | if (Left.Previous && Left.Previous->is(tok::period) && |
2718 | Left.Tok.getIdentifierInfo()) |
2719 | return false; |
2720 | // Additional unary JavaScript operators that need a space after. |
2721 | if (Left.isOneOf(tok::kw_throw, Keywords.kw_await, Keywords.kw_typeof, |
2722 | tok::kw_void)) |
2723 | return true; |
2724 | } |
2725 | if ((Left.isOneOf(Keywords.kw_let, Keywords.kw_var, Keywords.kw_in, |
2726 | tok::kw_const) || |
2727 | // "of" is only a keyword if it appears after another identifier |
2728 | // (e.g. as "const x of y" in a for loop), or after a destructuring |
2729 | // operation (const [x, y] of z, const {a, b} of c). |
2730 | (Left.is(Keywords.kw_of) && Left.Previous && |
2731 | (Left.Previous->Tok.is(tok::identifier) || |
2732 | Left.Previous->isOneOf(tok::r_square, tok::r_brace)))) && |
2733 | (!Left.Previous || !Left.Previous->is(tok::period))) |
2734 | return true; |
2735 | if (Left.isOneOf(tok::kw_for, Keywords.kw_as) && Left.Previous && |
2736 | Left.Previous->is(tok::period) && Right.is(tok::l_paren)) |
2737 | return false; |
2738 | if (Left.is(Keywords.kw_as) && |
2739 | Right.isOneOf(tok::l_square, tok::l_brace, tok::l_paren)) |
2740 | return true; |
2741 | if (Left.is(tok::kw_default) && Left.Previous && |
2742 | Left.Previous->is(tok::kw_export)) |
2743 | return true; |
2744 | if (Left.is(Keywords.kw_is) && Right.is(tok::l_brace)) |
2745 | return true; |
2746 | if (Right.isOneOf(TT_JsTypeColon, TT_JsTypeOptionalQuestion)) |
2747 | return false; |
2748 | if (Left.is(TT_JsTypeOperator) || Right.is(TT_JsTypeOperator)) |
2749 | return false; |
2750 | if ((Left.is(tok::l_brace) || Right.is(tok::r_brace)) && |
2751 | Line.First->isOneOf(Keywords.kw_import, tok::kw_export)) |
2752 | return false; |
2753 | if (Left.is(tok::ellipsis)) |
2754 | return false; |
2755 | if (Left.is(TT_TemplateCloser) && |
2756 | !Right.isOneOf(tok::equal, tok::l_brace, tok::comma, tok::l_square, |
2757 | Keywords.kw_implements, Keywords.kw_extends)) |
2758 | // Type assertions ('<type>expr') are not followed by whitespace. Other |
2759 | // locations that should have whitespace following are identified by the |
2760 | // above set of follower tokens. |
2761 | return false; |
2762 | if (Right.is(TT_JsNonNullAssertion)) |
2763 | return false; |
2764 | if (Left.is(TT_JsNonNullAssertion) && |
2765 | Right.isOneOf(Keywords.kw_as, Keywords.kw_in)) |
2766 | return true; // "x! as string", "x! in y" |
2767 | } else if (Style.Language == FormatStyle::LK_Java) { |
2768 | if (Left.is(tok::r_square) && Right.is(tok::l_brace)) |
2769 | return true; |
2770 | if (Left.is(Keywords.kw_synchronized) && Right.is(tok::l_paren)) |
2771 | return Style.SpaceBeforeParens != FormatStyle::SBPO_Never; |
2772 | if ((Left.isOneOf(tok::kw_static, tok::kw_public, tok::kw_private, |
2773 | tok::kw_protected) || |
2774 | Left.isOneOf(Keywords.kw_final, Keywords.kw_abstract, |
2775 | Keywords.kw_native)) && |
2776 | Right.is(TT_TemplateOpener)) |
2777 | return true; |
2778 | } |
2779 | if (Left.is(TT_ImplicitStringLiteral)) |
2780 | return Right.WhitespaceRange.getBegin() != Right.WhitespaceRange.getEnd(); |
2781 | if (Line.Type == LT_ObjCMethodDecl) { |
2782 | if (Left.is(TT_ObjCMethodSpecifier)) |
2783 | return true; |
2784 | if (Left.is(tok::r_paren) && canBeObjCSelectorComponent(Right)) |
2785 | // Don't space between ')' and <id> or ')' and 'new'. 'new' is not a |
2786 | // keyword in Objective-C, and '+ (instancetype)new;' is a standard class |
2787 | // method declaration. |
2788 | return false; |
2789 | } |
2790 | if (Line.Type == LT_ObjCProperty && |
2791 | (Right.is(tok::equal) || Left.is(tok::equal))) |
2792 | return false; |
2793 | |
2794 | if (Right.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow) || |
2795 | Left.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow)) |
2796 | return true; |
2797 | if (Right.is(TT_OverloadedOperatorLParen)) |
2798 | return Style.SpaceBeforeParens == FormatStyle::SBPO_Always; |
2799 | if (Left.is(tok::comma)) |
2800 | return true; |
2801 | if (Right.is(tok::comma)) |
2802 | return false; |
2803 | if (Right.is(TT_ObjCBlockLParen)) |
2804 | return true; |
2805 | if (Right.is(TT_CtorInitializerColon)) |
2806 | return Style.SpaceBeforeCtorInitializerColon; |
2807 | if (Right.is(TT_InheritanceColon) && !Style.SpaceBeforeInheritanceColon) |
2808 | return false; |
2809 | if (Right.is(TT_RangeBasedForLoopColon) && |
2810 | !Style.SpaceBeforeRangeBasedForLoopColon) |
2811 | return false; |
2812 | if (Right.is(tok::colon)) { |
2813 | if (Line.First->isOneOf(tok::kw_case, tok::kw_default) || |
2814 | !Right.getNextNonComment() || Right.getNextNonComment()->is(tok::semi)) |
2815 | return false; |
2816 | if (Right.is(TT_ObjCMethodExpr)) |
2817 | return false; |
2818 | if (Left.is(tok::question)) |
2819 | return false; |
2820 | if (Right.is(TT_InlineASMColon) && Left.is(tok::coloncolon)) |
2821 | return false; |
2822 | if (Right.is(TT_DictLiteral)) |
2823 | return Style.SpacesInContainerLiterals; |
2824 | if (Right.is(TT_AttributeColon)) |
2825 | return false; |
2826 | return true; |
2827 | } |
2828 | if (Left.is(TT_UnaryOperator)) |
2829 | return Right.is(TT_BinaryOperator); |
2830 | |
2831 | // If the next token is a binary operator or a selector name, we have |
2832 | // incorrectly classified the parenthesis as a cast. FIXME: Detect correctly. |
2833 | if (Left.is(TT_CastRParen)) |
2834 | return Style.SpaceAfterCStyleCast || |
2835 | Right.isOneOf(TT_BinaryOperator, TT_SelectorName); |
2836 | |
2837 | if (Left.is(tok::greater) && Right.is(tok::greater)) { |
2838 | if (Style.Language == FormatStyle::LK_TextProto || |
2839 | (Style.Language == FormatStyle::LK_Proto && Left.is(TT_DictLiteral))) |
2840 | return !Style.Cpp11BracedListStyle; |
2841 | return Right.is(TT_TemplateCloser) && Left.is(TT_TemplateCloser) && |
2842 | (Style.Standard != FormatStyle::LS_Cpp11 || Style.SpacesInAngles); |
2843 | } |
2844 | if (Right.isOneOf(tok::arrow, tok::arrowstar, tok::periodstar) || |
2845 | Left.isOneOf(tok::arrow, tok::period, tok::arrowstar, tok::periodstar) || |
2846 | (Right.is(tok::period) && Right.isNot(TT_DesignatedInitializerPeriod))) |
2847 | return false; |
2848 | if (!Style.SpaceBeforeAssignmentOperators && |
2849 | Right.getPrecedence() == prec::Assignment) |
2850 | return false; |
2851 | if (Style.Language == FormatStyle::LK_Java && Right.is(tok::coloncolon) && |
2852 | (Left.is(tok::identifier) || Left.is(tok::kw_this))) |
2853 | return false; |
2854 | if (Right.is(tok::coloncolon) && Left.is(tok::identifier)) |
2855 | // Generally don't remove existing spaces between an identifier and "::". |
2856 | // The identifier might actually be a macro name such as ALWAYS_INLINE. If |
2857 | // this turns out to be too lenient, add analysis of the identifier itself. |
2858 | return Right.WhitespaceRange.getBegin() != Right.WhitespaceRange.getEnd(); |
2859 | if (Right.is(tok::coloncolon) && !Left.isOneOf(tok::l_brace, tok::comment)) |
2860 | return (Left.is(TT_TemplateOpener) && |
2861 | Style.Standard == FormatStyle::LS_Cpp03) || |
2862 | !(Left.isOneOf(tok::l_paren, tok::r_paren, tok::l_square, |
2863 | tok::kw___super, TT_TemplateCloser, |
2864 | TT_TemplateOpener)) || |
2865 | (Left.is(tok ::l_paren) && Style.SpacesInParentheses); |
2866 | if ((Left.is(TT_TemplateOpener)) != (Right.is(TT_TemplateCloser))) |
2867 | return Style.SpacesInAngles; |
2868 | // Space before TT_StructuredBindingLSquare. |
2869 | if (Right.is(TT_StructuredBindingLSquare)) |
2870 | return !Left.isOneOf(tok::amp, tok::ampamp) || |
2871 | Style.PointerAlignment != FormatStyle::PAS_Right; |
2872 | // Space before & or && following a TT_StructuredBindingLSquare. |
2873 | if (Right.Next && Right.Next->is(TT_StructuredBindingLSquare) && |
2874 | Right.isOneOf(tok::amp, tok::ampamp)) |
2875 | return Style.PointerAlignment != FormatStyle::PAS_Left; |
2876 | if ((Right.is(TT_BinaryOperator) && !Left.is(tok::l_paren)) || |
2877 | (Left.isOneOf(TT_BinaryOperator, TT_ConditionalExpr) && |
2878 | !Right.is(tok::r_paren))) |
2879 | return true; |
2880 | if (Left.is(TT_TemplateCloser) && Right.is(tok::l_paren) && |
2881 | Right.isNot(TT_FunctionTypeLParen)) |
2882 | return Style.SpaceBeforeParens == FormatStyle::SBPO_Always; |
2883 | if (Right.is(TT_TemplateOpener) && Left.is(tok::r_paren) && |
2884 | Left.MatchingParen && Left.MatchingParen->is(TT_OverloadedOperatorLParen)) |
2885 | return false; |
2886 | if (Right.is(tok::less) && Left.isNot(tok::l_paren) && |
2887 | Line.startsWith(tok::hash)) |
2888 | return true; |
2889 | if (Right.is(TT_TrailingUnaryOperator)) |
2890 | return false; |
2891 | if (Left.is(TT_RegexLiteral)) |
2892 | return false; |
2893 | return spaceRequiredBetween(Line, Left, Right); |
2894 | } |
2895 | |
2896 | // Returns 'true' if 'Tok' is a brace we'd want to break before in Allman style. |
2897 | static bool isAllmanBrace(const FormatToken &Tok) { |
2898 | return Tok.is(tok::l_brace) && Tok.BlockKind == BK_Block && |
2899 | !Tok.isOneOf(TT_ObjCBlockLBrace, TT_LambdaLBrace, TT_DictLiteral); |
2900 | } |
2901 | |
2902 | bool TokenAnnotator::mustBreakBefore(const AnnotatedLine &Line, |
2903 | const FormatToken &Right) { |
2904 | const FormatToken &Left = *Right.Previous; |
2905 | if (Right.NewlinesBefore > 1 && Style.MaxEmptyLinesToKeep > 0) |
2906 | return true; |
2907 | |
2908 | if (Style.Language == FormatStyle::LK_JavaScript) { |
2909 | // FIXME: This might apply to other languages and token kinds. |
2910 | if (Right.is(tok::string_literal) && Left.is(tok::plus) && Left.Previous && |
2911 | Left.Previous->is(tok::string_literal)) |
2912 | return true; |
2913 | if (Left.is(TT_DictLiteral) && Left.is(tok::l_brace) && Line.Level == 0 && |
2914 | Left.Previous && Left.Previous->is(tok::equal) && |
2915 | Line.First->isOneOf(tok::identifier, Keywords.kw_import, tok::kw_export, |
2916 | tok::kw_const) && |
2917 | // kw_var/kw_let are pseudo-tokens that are tok::identifier, so match |
2918 | // above. |
2919 | !Line.First->isOneOf(Keywords.kw_var, Keywords.kw_let)) |
2920 | // Object literals on the top level of a file are treated as "enum-style". |
2921 | // Each key/value pair is put on a separate line, instead of bin-packing. |
2922 | return true; |
2923 | if (Left.is(tok::l_brace) && Line.Level == 0 && |
2924 | (Line.startsWith(tok::kw_enum) || |
2925 | Line.startsWith(tok::kw_const, tok::kw_enum) || |
2926 | Line.startsWith(tok::kw_export, tok::kw_enum) || |
2927 | Line.startsWith(tok::kw_export, tok::kw_const, tok::kw_enum))) |
2928 | // JavaScript top-level enum key/value pairs are put on separate lines |
2929 | // instead of bin-packing. |
2930 | return true; |
2931 | if (Right.is(tok::r_brace) && Left.is(tok::l_brace) && |
2932 | !Left.Children.empty()) |
2933 | // Support AllowShortFunctionsOnASingleLine for JavaScript. |
2934 | return Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_None || |
2935 | Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_Empty || |
2936 | (Left.NestingLevel == 0 && Line.Level == 0 && |
2937 | Style.AllowShortFunctionsOnASingleLine & |
2938 | FormatStyle::SFS_InlineOnly); |
2939 | } else if (Style.Language == FormatStyle::LK_Java) { |
2940 | if (Right.is(tok::plus) && Left.is(tok::string_literal) && Right.Next && |
2941 | Right.Next->is(tok::string_literal)) |
2942 | return true; |
2943 | } else if (Style.Language == FormatStyle::LK_Cpp || |
2944 | Style.Language == FormatStyle::LK_ObjC || |
2945 | Style.Language == FormatStyle::LK_Proto || |
2946 | Style.Language == FormatStyle::LK_TableGen || |
2947 | Style.Language == FormatStyle::LK_TextProto) { |
2948 | if (Left.isStringLiteral() && Right.isStringLiteral()) |
2949 | return true; |
2950 | } |
2951 | |
2952 | // If the last token before a '}', ']', or ')' is a comma or a trailing |
2953 | // comment, the intention is to insert a line break after it in order to make |
2954 | // shuffling around entries easier. Import statements, especially in |
2955 | // JavaScript, can be an exception to this rule. |
2956 | if (Style.JavaScriptWrapImports || Line.Type != LT_ImportStatement) { |
2957 | const FormatToken *BeforeClosingBrace = nullptr; |
2958 | if ((Left.isOneOf(tok::l_brace, TT_ArrayInitializerLSquare) || |
2959 | (Style.Language == FormatStyle::LK_JavaScript && |
2960 | Left.is(tok::l_paren))) && |
2961 | Left.BlockKind != BK_Block && Left.MatchingParen) |
2962 | BeforeClosingBrace = Left.MatchingParen->Previous; |
2963 | else if (Right.MatchingParen && |
2964 | (Right.MatchingParen->isOneOf(tok::l_brace, |
2965 | TT_ArrayInitializerLSquare) || |
2966 | (Style.Language == FormatStyle::LK_JavaScript && |
2967 | Right.MatchingParen->is(tok::l_paren)))) |
2968 | BeforeClosingBrace = &Left; |
2969 | if (BeforeClosingBrace && (BeforeClosingBrace->is(tok::comma) || |
2970 | BeforeClosingBrace->isTrailingComment())) |
2971 | return true; |
2972 | } |
2973 | |
2974 | if (Right.is(tok::comment)) |
2975 | return Left.BlockKind != BK_BracedInit && |
2976 | Left.isNot(TT_CtorInitializerColon) && |
2977 | (Right.NewlinesBefore > 0 && Right.HasUnescapedNewline); |
2978 | if (Left.isTrailingComment()) |
2979 | return true; |
2980 | if (Right.Previous->IsUnterminatedLiteral) |
2981 | return true; |
2982 | if (Right.is(tok::lessless) && Right.Next && |
2983 | Right.Previous->is(tok::string_literal) && |
2984 | Right.Next->is(tok::string_literal)) |
2985 | return true; |
2986 | if (Right.Previous->ClosesTemplateDeclaration && |
2987 | Right.Previous->MatchingParen && |
2988 | Right.Previous->MatchingParen->NestingLevel == 0 && |
2989 | Style.AlwaysBreakTemplateDeclarations == FormatStyle::BTDS_Yes) |
2990 | return true; |
2991 | if (Right.is(TT_CtorInitializerComma) && |
2992 | Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma && |
2993 | !Style.ConstructorInitializerAllOnOneLineOrOnePerLine) |
2994 | return true; |
2995 | if (Right.is(TT_CtorInitializerColon) && |
2996 | Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma && |
2997 | !Style.ConstructorInitializerAllOnOneLineOrOnePerLine) |
2998 | return true; |
2999 | // Break only if we have multiple inheritance. |
3000 | if (Style.BreakInheritanceList == FormatStyle::BILS_BeforeComma && |
3001 | Right.is(TT_InheritanceComma)) |
3002 | return true; |
3003 | if (Right.is(tok::string_literal) && Right.TokenText.startswith("R\"")) |
3004 | // Multiline raw string literals are special wrt. line breaks. The author |
3005 | // has made a deliberate choice and might have aligned the contents of the |
3006 | // string literal accordingly. Thus, we try keep existing line breaks. |
3007 | return Right.IsMultiline && Right.NewlinesBefore > 0; |
3008 | if ((Right.Previous->is(tok::l_brace) || |
3009 | (Right.Previous->is(tok::less) && Right.Previous->Previous && |
3010 | Right.Previous->Previous->is(tok::equal))) && |
3011 | Right.NestingLevel == 1 && Style.Language == FormatStyle::LK_Proto) { |
3012 | // Don't put enums or option definitions onto single lines in protocol |
3013 | // buffers. |
3014 | return true; |
3015 | } |
3016 | if (Right.is(TT_InlineASMBrace)) |
3017 | return Right.HasUnescapedNewline; |
3018 | if (isAllmanBrace(Left) || isAllmanBrace(Right)) |
3019 | return (Line.startsWith(tok::kw_enum) && Style.BraceWrapping.AfterEnum) || |
3020 | (Line.startsWith(tok::kw_typedef, tok::kw_enum) && |
3021 | Style.BraceWrapping.AfterEnum) || |
3022 | (Line.startsWith(tok::kw_class) && Style.BraceWrapping.AfterClass) || |
3023 | (Line.startsWith(tok::kw_struct) && Style.BraceWrapping.AfterStruct); |
3024 | if (Left.is(TT_ObjCBlockLBrace) && !Style.AllowShortBlocksOnASingleLine) |
3025 | return true; |
3026 | |
3027 | if (Left.is(TT_LambdaLBrace)) { |
3028 | if (Left.MatchingParen && Left.MatchingParen->Next && |
3029 | Left.MatchingParen->Next->isOneOf(tok::comma, tok::r_paren) && |
3030 | Style.AllowShortLambdasOnASingleLine == FormatStyle::SLS_Inline) |
3031 | return false; |
3032 | |
3033 | if (Style.AllowShortLambdasOnASingleLine == FormatStyle::SLS_None || |
3034 | Style.AllowShortLambdasOnASingleLine == FormatStyle::SLS_Inline || |
3035 | (!Left.Children.empty() && |
3036 | Style.AllowShortLambdasOnASingleLine == FormatStyle::SLS_Empty)) |
3037 | return true; |
3038 | } |
3039 | |
3040 | // Put multiple C# attributes on a new line. |
3041 | if (Style.isCSharp() && |
3042 | ((Left.is(TT_AttributeSquare) && Left.is(tok::r_square)) || |
3043 | (Left.is(tok::r_square) && Right.is(TT_AttributeSquare) && |
3044 | Right.is(tok::l_square)))) |
3045 | return true; |
3046 | |
3047 | // Put multiple Java annotation on a new line. |
3048 | if ((Style.Language == FormatStyle::LK_Java || |
3049 | Style.Language == FormatStyle::LK_JavaScript) && |
3050 | Left.is(TT_LeadingJavaAnnotation) && |
3051 | Right.isNot(TT_LeadingJavaAnnotation) && Right.isNot(tok::l_paren) && |
3052 | (Line.Last->is(tok::l_brace) || Style.BreakAfterJavaFieldAnnotations)) |
3053 | return true; |
3054 | |
3055 | if (Right.is(TT_ProtoExtensionLSquare)) |
3056 | return true; |
3057 | |
3058 | // In text proto instances if a submessage contains at least 2 entries and at |
3059 | // least one of them is a submessage, like A { ... B { ... } ... }, |
3060 | // put all of the entries of A on separate lines by forcing the selector of |
3061 | // the submessage B to be put on a newline. |
3062 | // |
3063 | // Example: these can stay on one line: |
3064 | // a { scalar_1: 1 scalar_2: 2 } |
3065 | // a { b { key: value } } |
3066 | // |
3067 | // and these entries need to be on a new line even if putting them all in one |
3068 | // line is under the column limit: |
3069 | // a { |
3070 | // scalar: 1 |
3071 | // b { key: value } |
3072 | // } |
3073 | // |
3074 | // We enforce this by breaking before a submessage field that has previous |
3075 | // siblings, *and* breaking before a field that follows a submessage field. |
3076 | // |
3077 | // Be careful to exclude the case [proto.ext] { ... } since the `]` is |
3078 | // the TT_SelectorName there, but we don't want to break inside the brackets. |
3079 | // |
3080 | // Another edge case is @submessage { key: value }, which is a common |
3081 | // substitution placeholder. In this case we want to keep `@` and `submessage` |
3082 | // together. |
3083 | // |
3084 | // We ensure elsewhere that extensions are always on their own line. |
3085 | if ((Style.Language == FormatStyle::LK_Proto || |
3086 | Style.Language == FormatStyle::LK_TextProto) && |
3087 | Right.is(TT_SelectorName) && !Right.is(tok::r_square) && Right.Next) { |
3088 | // Keep `@submessage` together in: |
3089 | // @submessage { key: value } |
3090 | if (Right.Previous && Right.Previous->is(tok::at)) |
3091 | return false; |
3092 | // Look for the scope opener after selector in cases like: |
3093 | // selector { ... |
3094 | // selector: { ... |
3095 | // selector: @base { ... |
3096 | FormatToken *LBrace = Right.Next; |
3097 | if (LBrace && LBrace->is(tok::colon)) { |
3098 | LBrace = LBrace->Next; |
3099 | if (LBrace && LBrace->is(tok::at)) { |
3100 | LBrace = LBrace->Next; |
3101 | if (LBrace) |
3102 | LBrace = LBrace->Next; |
3103 | } |
3104 | } |
3105 | if (LBrace && |
3106 | // The scope opener is one of {, [, <: |
3107 | // selector { ... } |
3108 | // selector [ ... ] |
3109 | // selector < ... > |
3110 | // |
3111 | // In case of selector { ... }, the l_brace is TT_DictLiteral. |
3112 | // In case of an empty selector {}, the l_brace is not TT_DictLiteral, |
3113 | // so we check for immediately following r_brace. |
3114 | ((LBrace->is(tok::l_brace) && |
3115 | (LBrace->is(TT_DictLiteral) || |
3116 | (LBrace->Next && LBrace->Next->is(tok::r_brace)))) || |
3117 | LBrace->is(TT_ArrayInitializerLSquare) || LBrace->is(tok::less))) { |
3118 | // If Left.ParameterCount is 0, then this submessage entry is not the |
3119 | // first in its parent submessage, and we want to break before this entry. |
3120 | // If Left.ParameterCount is greater than 0, then its parent submessage |
3121 | // might contain 1 or more entries and we want to break before this entry |
3122 | // if it contains at least 2 entries. We deal with this case later by |
3123 | // detecting and breaking before the next entry in the parent submessage. |
3124 | if (Left.ParameterCount == 0) |
3125 | return true; |
3126 | // However, if this submessage is the first entry in its parent |
3127 | // submessage, Left.ParameterCount might be 1 in some cases. |
3128 | // We deal with this case later by detecting an entry |
3129 | // following a closing paren of this submessage. |
3130 | } |
3131 | |
3132 | // If this is an entry immediately following a submessage, it will be |
3133 | // preceded by a closing paren of that submessage, like in: |
3134 | // left---. .---right |
3135 | // v v |
3136 | // sub: { ... } key: value |
3137 | // If there was a comment between `}` an `key` above, then `key` would be |
3138 | // put on a new line anyways. |
3139 | if (Left.isOneOf(tok::r_brace, tok::greater, tok::r_square)) |
3140 | return true; |
3141 | } |
3142 | |
3143 | // Deal with lambda arguments in C++ - we want consistent line breaks whether |
3144 | // they happen to be at arg0, arg1 or argN. The selection is a bit nuanced |
3145 | // as aggressive line breaks are placed when the lambda is not the last arg. |
3146 | if ((Style.Language == FormatStyle::LK_Cpp || |
3147 | Style.Language == FormatStyle::LK_ObjC) && |
3148 | Left.is(tok::l_paren) && Left.BlockParameterCount > 0 && |
3149 | !Right.isOneOf(tok::l_paren, TT_LambdaLSquare)) { |
3150 | // Multiple lambdas in the same function call force line breaks. |
3151 | if (Left.BlockParameterCount > 1) |
3152 | return true; |
3153 | |
3154 | // A lambda followed by another arg forces a line break. |
3155 | if (!Left.Role) |
3156 | return false; |
3157 | auto Comma = Left.Role->lastComma(); |
3158 | if (!Comma) |
3159 | return false; |
3160 | auto Next = Comma->getNextNonComment(); |
3161 | if (!Next) |
3162 | return false; |
3163 | if (!Next->isOneOf(TT_LambdaLSquare, tok::l_brace, tok::caret)) |
3164 | return true; |
3165 | } |
3166 | |
3167 | return false; |
3168 | } |
3169 | |
3170 | bool TokenAnnotator::canBreakBefore(const AnnotatedLine &Line, |
3171 | const FormatToken &Right) { |
3172 | const FormatToken &Left = *Right.Previous; |
3173 | |
3174 | // Language-specific stuff. |
3175 | if (Style.Language == FormatStyle::LK_Java) { |
3176 | if (Left.isOneOf(Keywords.kw_throws, Keywords.kw_extends, |
3177 | Keywords.kw_implements)) |
3178 | return false; |
3179 | if (Right.isOneOf(Keywords.kw_throws, Keywords.kw_extends, |
3180 | Keywords.kw_implements)) |
3181 | return true; |
3182 | } else if (Style.Language == FormatStyle::LK_JavaScript) { |
3183 | const FormatToken *NonComment = Right.getPreviousNonComment(); |
3184 | if (NonComment && |
3185 | NonComment->isOneOf( |
3186 | tok::kw_return, Keywords.kw_yield, tok::kw_continue, tok::kw_break, |
3187 | tok::kw_throw, Keywords.kw_interface, Keywords.kw_type, |
3188 | tok::kw_static, tok::kw_public, tok::kw_private, tok::kw_protected, |
3189 | Keywords.kw_readonly, Keywords.kw_abstract, Keywords.kw_get, |
3190 | Keywords.kw_set, Keywords.kw_async, Keywords.kw_await)) |
3191 | return false; // Otherwise automatic semicolon insertion would trigger. |
3192 | if (Right.NestingLevel == 0 && |
3193 | (Left.Tok.getIdentifierInfo() || |
3194 | Left.isOneOf(tok::r_square, tok::r_paren)) && |
3195 | Right.isOneOf(tok::l_square, tok::l_paren)) |
3196 | return false; // Otherwise automatic semicolon insertion would trigger. |
3197 | if (Left.is(TT_JsFatArrow) && Right.is(tok::l_brace)) |
3198 | return false; |
3199 | if (Left.is(TT_JsTypeColon)) |
3200 | return true; |
3201 | // Don't wrap between ":" and "!" of a strict prop init ("field!: type;"). |
3202 | if (Left.is(tok::exclaim) && Right.is(tok::colon)) |
3203 | return false; |
3204 | // Look for is type annotations like: |
3205 | // function f(): a is B { ... } |
3206 | // Do not break before is in these cases. |
3207 | if (Right.is(Keywords.kw_is)) { |
3208 | const FormatToken *Next = Right.getNextNonComment(); |
3209 | // If `is` is followed by a colon, it's likely that it's a dict key, so |
3210 | // ignore it for this check. |
3211 | // For example this is common in Polymer: |
3212 | // Polymer({ |
3213 | // is: 'name', |
3214 | // ... |
3215 | // }); |
3216 | if (!Next || !Next->is(tok::colon)) |
3217 | return false; |
3218 | } |
3219 | if (Left.is(Keywords.kw_in)) |
3220 | return Style.BreakBeforeBinaryOperators == FormatStyle::BOS_None; |
3221 | if (Right.is(Keywords.kw_in)) |
3222 | return Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None; |
3223 | if (Right.is(Keywords.kw_as)) |
3224 | return false; // must not break before as in 'x as type' casts |
3225 | if (Right.isOneOf(Keywords.kw_extends, Keywords.kw_infer)) { |
3226 | // extends and infer can appear as keywords in conditional types: |
3227 | // https://www.typescriptlang.org/docs/handbook/release-notes/typescript-2-8.html#conditional-types |
3228 | // do not break before them, as the expressions are subject to ASI. |
3229 | return false; |
3230 | } |
3231 | if (Left.is(Keywords.kw_as)) |
3232 | return true; |
3233 | if (Left.is(TT_JsNonNullAssertion)) |
3234 | return true; |
3235 | if (Left.is(Keywords.kw_declare) && |
3236 | Right.isOneOf(Keywords.kw_module, tok::kw_namespace, |
3237 | Keywords.kw_function, tok::kw_class, tok::kw_enum, |
3238 | Keywords.kw_interface, Keywords.kw_type, Keywords.kw_var, |
3239 | Keywords.kw_let, tok::kw_const)) |
3240 | // See grammar for 'declare' statements at: |
3241 | // https://github.com/Microsoft/TypeScript/blob/master/doc/spec.md#A.10 |
3242 | return false; |
3243 | if (Left.isOneOf(Keywords.kw_module, tok::kw_namespace) && |
3244 | Right.isOneOf(tok::identifier, tok::string_literal)) |
3245 | return false; // must not break in "module foo { ...}" |
3246 | if (Right.is(TT_TemplateString) && Right.closesScope()) |
3247 | return false; |
3248 | // Don't split tagged template literal so there is a break between the tag |
3249 | // identifier and template string. |
3250 | if (Left.is(tok::identifier) && Right.is(TT_TemplateString)) { |
3251 | return false; |
3252 | } |
3253 | if (Left.is(TT_TemplateString) && Left.opensScope()) |
3254 | return true; |
3255 | } |
3256 | |
3257 | if (Left.is(tok::at)) |
3258 | return false; |
3259 | if (Left.Tok.getObjCKeywordID() == tok::objc_interface) |
3260 | return false; |
3261 | if (Left.isOneOf(TT_JavaAnnotation, TT_LeadingJavaAnnotation)) |
3262 | return !Right.is(tok::l_paren); |
3263 | if (Right.is(TT_PointerOrReference)) |
3264 | return Line.IsMultiVariableDeclStmt || |
3265 | (Style.PointerAlignment == FormatStyle::PAS_Right && |
3266 | (!Right.Next || Right.Next->isNot(TT_FunctionDeclarationName))); |
3267 | if (Right.isOneOf(TT_StartOfName, TT_FunctionDeclarationName) || |
3268 | Right.is(tok::kw_operator)) |
3269 | return true; |
3270 | if (Left.is(TT_PointerOrReference)) |
3271 | return false; |
3272 | if (Right.isTrailingComment()) |
3273 | // We rely on MustBreakBefore being set correctly here as we should not |
3274 | // change the "binding" behavior of a comment. |
3275 | // The first comment in a braced lists is always interpreted as belonging to |
3276 | // the first list element. Otherwise, it should be placed outside of the |
3277 | // list. |
3278 | return Left.BlockKind == BK_BracedInit || |
3279 | (Left.is(TT_CtorInitializerColon) && |
3280 | Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon); |
3281 | if (Left.is(tok::question) && Right.is(tok::colon)) |
3282 | return false; |
3283 | if (Right.is(TT_ConditionalExpr) || Right.is(tok::question)) |
3284 | return Style.BreakBeforeTernaryOperators; |
3285 | if (Left.is(TT_ConditionalExpr) || Left.is(tok::question)) |
3286 | return !Style.BreakBeforeTernaryOperators; |
3287 | if (Left.is(TT_InheritanceColon)) |
3288 | return Style.BreakInheritanceList == FormatStyle::BILS_AfterColon; |
3289 | if (Right.is(TT_InheritanceColon)) |
3290 | return Style.BreakInheritanceList != FormatStyle::BILS_AfterColon; |
3291 | if (Right.is(TT_ObjCMethodExpr) && !Right.is(tok::r_square) && |
3292 | Left.isNot(TT_SelectorName)) |
3293 | return true; |
3294 | |
3295 | if (Right.is(tok::colon) && |
3296 | !Right.isOneOf(TT_CtorInitializerColon, TT_InlineASMColon)) |
3297 | return false; |
3298 | if (Left.is(tok::colon) && Left.isOneOf(TT_DictLiteral, TT_ObjCMethodExpr)) { |
3299 | if (Style.Language == FormatStyle::LK_Proto || |
3300 | Style.Language == FormatStyle::LK_TextProto) { |
3301 | if (!Style.AlwaysBreakBeforeMultilineStrings && Right.isStringLiteral()) |
3302 | return false; |
3303 | // Prevent cases like: |
3304 | // |
3305 | // submessage: |
3306 | // { key: valueeeeeeeeeeee } |
3307 | // |
3308 | // when the snippet does not fit into one line. |
3309 | // Prefer: |
3310 | // |
3311 | // submessage: { |
3312 | // key: valueeeeeeeeeeee |
3313 | // } |
3314 | // |
3315 | // instead, even if it is longer by one line. |
3316 | // |
3317 | // Note that this allows allows the "{" to go over the column limit |
3318 | // when the column limit is just between ":" and "{", but that does |
3319 | // not happen too often and alternative formattings in this case are |
3320 | // not much better. |
3321 | // |
3322 | // The code covers the cases: |
3323 | // |
3324 | // submessage: { ... } |
3325 | // submessage: < ... > |
3326 | // repeated: [ ... ] |
3327 | if (((Right.is(tok::l_brace) || Right.is(tok::less)) && |
3328 | Right.is(TT_DictLiteral)) || |
3329 | Right.is(TT_ArrayInitializerLSquare)) |
3330 | return false; |
3331 | } |
3332 | return true; |
3333 | } |
3334 | if (Right.is(tok::r_square) && Right.MatchingParen && |
3335 | Right.MatchingParen->is(TT_ProtoExtensionLSquare)) |
3336 | return false; |
3337 | if (Right.is(TT_SelectorName) || (Right.is(tok::identifier) && Right.Next && |
3338 | Right.Next->is(TT_ObjCMethodExpr))) |
3339 | return Left.isNot(tok::period); // FIXME: Properly parse ObjC calls. |
3340 | if (Left.is(tok::r_paren) && Line.Type == LT_ObjCProperty) |
3341 | return true; |
3342 | if (Left.ClosesTemplateDeclaration || Left.is(TT_FunctionAnnotationRParen)) |
3343 | return true; |
3344 | if (Right.isOneOf(TT_RangeBasedForLoopColon, TT_OverloadedOperatorLParen, |
3345 | TT_OverloadedOperator)) |
3346 | return false; |
3347 | if (Left.is(TT_RangeBasedForLoopColon)) |
3348 | return true; |
3349 | if (Right.is(TT_RangeBasedForLoopColon)) |
3350 | return false; |
3351 | if (Left.is(TT_TemplateCloser) && Right.is(TT_TemplateOpener)) |
3352 | return true; |
3353 | if (Left.isOneOf(TT_TemplateCloser, TT_UnaryOperator) || |
3354 | Left.is(tok::kw_operator)) |
3355 | return false; |
3356 | if (Left.is(tok::equal) && !Right.isOneOf(tok::kw_default, tok::kw_delete) && |
3357 | Line.Type == LT_VirtualFunctionDecl && Left.NestingLevel == 0) |
3358 | return false; |
3359 | if (Left.is(tok::equal) && Right.is(tok::l_brace) && |
3360 | !Style.Cpp11BracedListStyle) |
3361 | return false; |
3362 | if (Left.is(tok::l_paren) && Left.is(TT_AttributeParen)) |
3363 | return false; |
3364 | if (Left.is(tok::l_paren) && Left.Previous && |
3365 | (Left.Previous->isOneOf(TT_BinaryOperator, TT_CastRParen))) |
3366 | return false; |
3367 | if (Right.is(TT_ImplicitStringLiteral)) |
3368 | return false; |
3369 | |
3370 | if (Right.is(tok::r_paren) || Right.is(TT_TemplateCloser)) |
3371 | return false; |
3372 | if (Right.is(tok::r_square) && Right.MatchingParen && |
3373 | Right.MatchingParen->is(TT_LambdaLSquare)) |
3374 | return false; |
3375 | |
3376 | // We only break before r_brace if there was a corresponding break before |
3377 | // the l_brace, which is tracked by BreakBeforeClosingBrace. |
3378 | if (Right.is(tok::r_brace)) |
3379 | return Right.MatchingParen && Right.MatchingParen->BlockKind == BK_Block; |
3380 | |
3381 | // Allow breaking after a trailing annotation, e.g. after a method |
3382 | // declaration. |
3383 | if (Left.is(TT_TrailingAnnotation)) |
3384 | return !Right.isOneOf(tok::l_brace, tok::semi, tok::equal, tok::l_paren, |
3385 | tok::less, tok::coloncolon); |
3386 | |
3387 | if (Right.is(tok::kw___attribute) || |
3388 | (Right.is(tok::l_square) && Right.is(TT_AttributeSquare))) |
3389 | return true; |
3390 | |
3391 | if (Left.is(tok::identifier) && Right.is(tok::string_literal)) |
3392 | return true; |
3393 | |
3394 | if (Right.is(tok::identifier) && Right.Next && Right.Next->is(TT_DictLiteral)) |
3395 | return true; |
3396 | |
3397 | if (Left.is(TT_CtorInitializerColon)) |
3398 | return Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon; |
3399 | if (Right.is(TT_CtorInitializerColon)) |
3400 | return Style.BreakConstructorInitializers != FormatStyle::BCIS_AfterColon; |
3401 | if (Left.is(TT_CtorInitializerComma) && |
3402 | Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma) |
3403 | return false; |
3404 | if (Right.is(TT_CtorInitializerComma) && |
3405 | Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma) |
3406 | return true; |
3407 | if (Left.is(TT_InheritanceComma) && |
3408 | Style.BreakInheritanceList == FormatStyle::BILS_BeforeComma) |
3409 | return false; |
3410 | if (Right.is(TT_InheritanceComma) && |
3411 | Style.BreakInheritanceList == FormatStyle::BILS_BeforeComma) |
3412 | return true; |
3413 | if ((Left.is(tok::greater) && Right.is(tok::greater)) || |
3414 | (Left.is(tok::less) && Right.is(tok::less))) |
3415 | return false; |
3416 | if (Right.is(TT_BinaryOperator) && |
3417 | Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None && |
3418 | (Style.BreakBeforeBinaryOperators == FormatStyle::BOS_All || |
3419 | Right.getPrecedence() != prec::Assignment)) |
3420 | return true; |
3421 | if (Left.is(TT_ArrayInitializerLSquare)) |
3422 | return true; |
3423 | if (Right.is(tok::kw_typename) && Left.isNot(tok::kw_const)) |
3424 | return true; |
3425 | if ((Left.isBinaryOperator() || Left.is(TT_BinaryOperator)) && |
3426 | !Left.isOneOf(tok::arrowstar, tok::lessless) && |
3427 | Style.BreakBeforeBinaryOperators != FormatStyle::BOS_All && |
3428 | (Style.BreakBeforeBinaryOperators == FormatStyle::BOS_None || |
3429 | Left.getPrecedence() == prec::Assignment)) |
3430 | return true; |
3431 | if ((Left.is(TT_AttributeSquare) && Right.is(tok::l_square)) || |
3432 | (Left.is(tok::r_square) && Right.is(TT_AttributeSquare))) |
3433 | return false; |
3434 | return Left.isOneOf(tok::comma, tok::coloncolon, tok::semi, tok::l_brace, |
3435 | tok::kw_class, tok::kw_struct, tok::comment) || |
3436 | Right.isMemberAccess() || |
3437 | Right.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow, tok::lessless, |
3438 | tok::colon, tok::l_square, tok::at) || |
3439 | (Left.is(tok::r_paren) && |
3440 | Right.isOneOf(tok::identifier, tok::kw_const)) || |
3441 | (Left.is(tok::l_paren) && !Right.is(tok::r_paren)) || |
3442 | (Left.is(TT_TemplateOpener) && !Right.is(TT_TemplateCloser)); |
3443 | } |
3444 | |
3445 | void TokenAnnotator::printDebugInfo(const AnnotatedLine &Line) { |
3446 | llvm::errs() << "AnnotatedTokens(L=" << Line.Level << "):\n"; |
3447 | const FormatToken *Tok = Line.First; |
3448 | while (Tok) { |
3449 | llvm::errs() << " M=" << Tok->MustBreakBefore |
3450 | << " C=" << Tok->CanBreakBefore |
3451 | << " T=" << getTokenTypeName(Tok->Type) |
3452 | << " S=" << Tok->SpacesRequiredBefore |
3453 | << " B=" << Tok->BlockParameterCount |
3454 | << " BK=" << Tok->BlockKind << " P=" << Tok->SplitPenalty |
3455 | << " Name=" << Tok->Tok.getName() << " L=" << Tok->TotalLength |
3456 | << " PPK=" << Tok->PackingKind << " FakeLParens="; |
3457 | for (unsigned i = 0, e = Tok->FakeLParens.size(); i != e; ++i) |
3458 | llvm::errs() << Tok->FakeLParens[i] << "/"; |
3459 | llvm::errs() << " FakeRParens=" << Tok->FakeRParens; |
3460 | llvm::errs() << " II=" << Tok->Tok.getIdentifierInfo(); |
3461 | llvm::errs() << " Text='" << Tok->TokenText << "'\n"; |
3462 | if (!Tok->Next) |
3463 | assert(Tok == Line.Last); |
3464 | Tok = Tok->Next; |
3465 | } |
3466 | llvm::errs() << "----\n"; |
3467 | } |
3468 | |
3469 | } // namespace format |
3470 | } // namespace clang |
3471 | |