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 "clang/Basic/TokenKinds.h" 19 #include "llvm/ADT/SmallPtrSet.h" 20 #include "llvm/Support/Debug.h" 21 22 #define DEBUG_TYPE "format-token-annotator" 23 24 namespace clang { 25 namespace format { 26 27 namespace { 28 29 /// Returns \c true if the token can be used as an identifier in 30 /// an Objective-C \c \@selector, \c false otherwise. 31 /// 32 /// Because getFormattingLangOpts() always lexes source code as 33 /// Objective-C++, C++ keywords like \c new and \c delete are 34 /// lexed as tok::kw_*, not tok::identifier, even for Objective-C. 35 /// 36 /// For Objective-C and Objective-C++, both identifiers and keywords 37 /// are valid inside @selector(...) (or a macro which 38 /// invokes @selector(...)). So, we allow treat any identifier or 39 /// keyword as a potential Objective-C selector component. 40 static bool canBeObjCSelectorComponent(const FormatToken &Tok) { 41 return Tok.Tok.getIdentifierInfo() != nullptr; 42 } 43 44 /// With `Left` being '(', check if we're at either `[...](` or 45 /// `[...]<...>(`, where the [ opens a lambda capture list. 46 static bool isLambdaParameterList(const FormatToken *Left) { 47 // Skip <...> if present. 48 if (Left->Previous && Left->Previous->is(tok::greater) && 49 Left->Previous->MatchingParen && 50 Left->Previous->MatchingParen->is(TT_TemplateOpener)) 51 Left = Left->Previous->MatchingParen; 52 53 // Check for `[...]`. 54 return Left->Previous && Left->Previous->is(tok::r_square) && 55 Left->Previous->MatchingParen && 56 Left->Previous->MatchingParen->is(TT_LambdaLSquare); 57 } 58 59 /// Returns \c true if the token is followed by a boolean condition, \c false 60 /// otherwise. 61 static bool isKeywordWithCondition(const FormatToken &Tok) { 62 return Tok.isOneOf(tok::kw_if, tok::kw_for, tok::kw_while, tok::kw_switch, 63 tok::kw_constexpr, tok::kw_catch); 64 } 65 66 /// A parser that gathers additional information about tokens. 67 /// 68 /// The \c TokenAnnotator tries to match parenthesis and square brakets and 69 /// store a parenthesis levels. It also tries to resolve matching "<" and ">" 70 /// into template parameter lists. 71 class AnnotatingParser { 72 public: 73 AnnotatingParser(const FormatStyle &Style, AnnotatedLine &Line, 74 const AdditionalKeywords &Keywords) 75 : Style(Style), Line(Line), CurrentToken(Line.First), AutoFound(false), 76 Keywords(Keywords) { 77 Contexts.push_back(Context(tok::unknown, 1, /*IsExpression=*/false)); 78 resetTokenMetadata(); 79 } 80 81 private: 82 bool parseAngle() { 83 if (!CurrentToken || !CurrentToken->Previous) 84 return false; 85 if (NonTemplateLess.count(CurrentToken->Previous)) 86 return false; 87 88 const FormatToken &Previous = *CurrentToken->Previous; // The '<'. 89 if (Previous.Previous) { 90 if (Previous.Previous->Tok.isLiteral()) 91 return false; 92 if (Previous.Previous->is(tok::r_paren) && Contexts.size() > 1 && 93 (!Previous.Previous->MatchingParen || 94 !Previous.Previous->MatchingParen->is(TT_OverloadedOperatorLParen))) 95 return false; 96 } 97 98 FormatToken *Left = CurrentToken->Previous; 99 Left->ParentBracket = Contexts.back().ContextKind; 100 ScopedContextCreator ContextCreator(*this, tok::less, 12); 101 102 // If this angle is in the context of an expression, we need to be more 103 // hesitant to detect it as opening template parameters. 104 bool InExprContext = Contexts.back().IsExpression; 105 106 Contexts.back().IsExpression = false; 107 // If there's a template keyword before the opening angle bracket, this is a 108 // template parameter, not an argument. 109 Contexts.back().InTemplateArgument = 110 Left->Previous && Left->Previous->Tok.isNot(tok::kw_template); 111 112 if (Style.Language == FormatStyle::LK_Java && 113 CurrentToken->is(tok::question)) 114 next(); 115 116 while (CurrentToken) { 117 if (CurrentToken->is(tok::greater)) { 118 // Try to do a better job at looking for ">>" within the condition of 119 // a statement. Conservatively insert spaces between consecutive ">" 120 // tokens to prevent splitting right bitshift operators and potentially 121 // altering program semantics. This check is overly conservative and 122 // will prevent spaces from being inserted in select nested template 123 // parameter cases, but should not alter program semantics. 124 if (CurrentToken->Next && CurrentToken->Next->is(tok::greater) && 125 Left->ParentBracket != tok::less && 126 (isKeywordWithCondition(*Line.First) || 127 CurrentToken->getStartOfNonWhitespace() == 128 CurrentToken->Next->getStartOfNonWhitespace().getLocWithOffset( 129 -1))) 130 return false; 131 Left->MatchingParen = CurrentToken; 132 CurrentToken->MatchingParen = Left; 133 // In TT_Proto, we must distignuish between: 134 // map<key, value> 135 // msg < item: data > 136 // msg: < item: data > 137 // In TT_TextProto, map<key, value> does not occur. 138 if (Style.Language == FormatStyle::LK_TextProto || 139 (Style.Language == FormatStyle::LK_Proto && Left->Previous && 140 Left->Previous->isOneOf(TT_SelectorName, TT_DictLiteral))) 141 CurrentToken->setType(TT_DictLiteral); 142 else 143 CurrentToken->setType(TT_TemplateCloser); 144 next(); 145 return true; 146 } 147 if (CurrentToken->is(tok::question) && 148 Style.Language == FormatStyle::LK_Java) { 149 next(); 150 continue; 151 } 152 if (CurrentToken->isOneOf(tok::r_paren, tok::r_square, tok::r_brace) || 153 (CurrentToken->isOneOf(tok::colon, tok::question) && InExprContext && 154 !Style.isCSharp() && Style.Language != FormatStyle::LK_Proto && 155 Style.Language != FormatStyle::LK_TextProto)) 156 return false; 157 // If a && or || is found and interpreted as a binary operator, this set 158 // of angles is likely part of something like "a < b && c > d". If the 159 // angles are inside an expression, the ||/&& might also be a binary 160 // operator that was misinterpreted because we are parsing template 161 // parameters. 162 // FIXME: This is getting out of hand, write a decent parser. 163 if (CurrentToken->Previous->isOneOf(tok::pipepipe, tok::ampamp) && 164 CurrentToken->Previous->is(TT_BinaryOperator) && 165 Contexts[Contexts.size() - 2].IsExpression && 166 !Line.startsWith(tok::kw_template)) 167 return false; 168 updateParameterCount(Left, CurrentToken); 169 if (Style.Language == FormatStyle::LK_Proto) { 170 if (FormatToken *Previous = CurrentToken->getPreviousNonComment()) { 171 if (CurrentToken->is(tok::colon) || 172 (CurrentToken->isOneOf(tok::l_brace, tok::less) && 173 Previous->isNot(tok::colon))) 174 Previous->setType(TT_SelectorName); 175 } 176 } 177 if (!consumeToken()) 178 return false; 179 } 180 return false; 181 } 182 183 bool parseUntouchableParens() { 184 while (CurrentToken) { 185 CurrentToken->Finalized = true; 186 switch (CurrentToken->Tok.getKind()) { 187 case tok::l_paren: 188 next(); 189 if (!parseUntouchableParens()) 190 return false; 191 continue; 192 case tok::r_paren: 193 next(); 194 return true; 195 default: 196 // no-op 197 break; 198 } 199 next(); 200 } 201 return false; 202 } 203 204 bool parseParens(bool LookForDecls = false) { 205 if (!CurrentToken) 206 return false; 207 FormatToken *Left = CurrentToken->Previous; 208 assert(Left && "Unknown previous token"); 209 FormatToken *PrevNonComment = Left->getPreviousNonComment(); 210 Left->ParentBracket = Contexts.back().ContextKind; 211 ScopedContextCreator ContextCreator(*this, tok::l_paren, 1); 212 213 // FIXME: This is a bit of a hack. Do better. 214 Contexts.back().ColonIsForRangeExpr = 215 Contexts.size() == 2 && Contexts[0].ColonIsForRangeExpr; 216 217 if (Left->Previous && Left->Previous->is(TT_UntouchableMacroFunc)) { 218 Left->Finalized = true; 219 return parseUntouchableParens(); 220 } 221 222 bool StartsObjCMethodExpr = false; 223 if (FormatToken *MaybeSel = Left->Previous) { 224 // @selector( starts a selector. 225 if (MaybeSel->isObjCAtKeyword(tok::objc_selector) && MaybeSel->Previous && 226 MaybeSel->Previous->is(tok::at)) 227 StartsObjCMethodExpr = true; 228 } 229 230 if (Left->is(TT_OverloadedOperatorLParen)) { 231 // Find the previous kw_operator token. 232 FormatToken *Prev = Left; 233 while (!Prev->is(tok::kw_operator)) { 234 Prev = Prev->Previous; 235 assert(Prev && "Expect a kw_operator prior to the OperatorLParen!"); 236 } 237 238 // If faced with "a.operator*(argument)" or "a->operator*(argument)", 239 // i.e. the operator is called as a member function, 240 // then the argument must be an expression. 241 bool OperatorCalledAsMemberFunction = 242 Prev->Previous && Prev->Previous->isOneOf(tok::period, tok::arrow); 243 Contexts.back().IsExpression = OperatorCalledAsMemberFunction; 244 } else if (Style.isJavaScript() && 245 (Line.startsWith(Keywords.kw_type, tok::identifier) || 246 Line.startsWith(tok::kw_export, Keywords.kw_type, 247 tok::identifier))) { 248 // type X = (...); 249 // export type X = (...); 250 Contexts.back().IsExpression = false; 251 } else if (Left->Previous && 252 (Left->Previous->isOneOf(tok::kw_static_assert, tok::kw_while, 253 tok::l_paren, tok::comma) || 254 Left->Previous->isIf() || 255 Left->Previous->is(TT_BinaryOperator))) { 256 // static_assert, if and while usually contain expressions. 257 Contexts.back().IsExpression = true; 258 } else if (Style.isJavaScript() && Left->Previous && 259 (Left->Previous->is(Keywords.kw_function) || 260 (Left->Previous->endsSequence(tok::identifier, 261 Keywords.kw_function)))) { 262 // function(...) or function f(...) 263 Contexts.back().IsExpression = false; 264 } else if (Style.isJavaScript() && Left->Previous && 265 Left->Previous->is(TT_JsTypeColon)) { 266 // let x: (SomeType); 267 Contexts.back().IsExpression = false; 268 } else if (isLambdaParameterList(Left)) { 269 // This is a parameter list of a lambda expression. 270 Contexts.back().IsExpression = false; 271 } else if (Line.InPPDirective && 272 (!Left->Previous || !Left->Previous->is(tok::identifier))) { 273 Contexts.back().IsExpression = true; 274 } else if (Contexts[Contexts.size() - 2].CaretFound) { 275 // This is the parameter list of an ObjC block. 276 Contexts.back().IsExpression = false; 277 } else if (Left->Previous && Left->Previous->is(TT_ForEachMacro)) { 278 // The first argument to a foreach macro is a declaration. 279 Contexts.back().IsForEachMacro = true; 280 Contexts.back().IsExpression = false; 281 } else if (Left->Previous && Left->Previous->MatchingParen && 282 Left->Previous->MatchingParen->is(TT_ObjCBlockLParen)) { 283 Contexts.back().IsExpression = false; 284 } else if (!Line.MustBeDeclaration && !Line.InPPDirective) { 285 bool IsForOrCatch = 286 Left->Previous && Left->Previous->isOneOf(tok::kw_for, tok::kw_catch); 287 Contexts.back().IsExpression = !IsForOrCatch; 288 } 289 290 // Infer the role of the l_paren based on the previous token if we haven't 291 // detected one one yet. 292 if (PrevNonComment && Left->is(TT_Unknown)) { 293 if (PrevNonComment->is(tok::kw___attribute)) { 294 Left->setType(TT_AttributeParen); 295 } else if (PrevNonComment->isOneOf(TT_TypenameMacro, tok::kw_decltype, 296 tok::kw_typeof, tok::kw__Atomic, 297 tok::kw___underlying_type)) { 298 Left->setType(TT_TypeDeclarationParen); 299 // decltype() and typeof() usually contain expressions. 300 if (PrevNonComment->isOneOf(tok::kw_decltype, tok::kw_typeof)) 301 Contexts.back().IsExpression = true; 302 } 303 } 304 305 if (StartsObjCMethodExpr) { 306 Contexts.back().ColonIsObjCMethodExpr = true; 307 Left->setType(TT_ObjCMethodExpr); 308 } 309 310 // MightBeFunctionType and ProbablyFunctionType are used for 311 // function pointer and reference types as well as Objective-C 312 // block types: 313 // 314 // void (*FunctionPointer)(void); 315 // void (&FunctionReference)(void); 316 // void (&&FunctionReference)(void); 317 // void (^ObjCBlock)(void); 318 bool MightBeFunctionType = !Contexts[Contexts.size() - 2].IsExpression; 319 bool ProbablyFunctionType = 320 CurrentToken->isOneOf(tok::star, tok::amp, tok::ampamp, tok::caret); 321 bool HasMultipleLines = false; 322 bool HasMultipleParametersOnALine = false; 323 bool MightBeObjCForRangeLoop = 324 Left->Previous && Left->Previous->is(tok::kw_for); 325 FormatToken *PossibleObjCForInToken = nullptr; 326 while (CurrentToken) { 327 // LookForDecls is set when "if (" has been seen. Check for 328 // 'identifier' '*' 'identifier' followed by not '=' -- this 329 // '*' has to be a binary operator but determineStarAmpUsage() will 330 // categorize it as an unary operator, so set the right type here. 331 if (LookForDecls && CurrentToken->Next) { 332 FormatToken *Prev = CurrentToken->getPreviousNonComment(); 333 if (Prev) { 334 FormatToken *PrevPrev = Prev->getPreviousNonComment(); 335 FormatToken *Next = CurrentToken->Next; 336 if (PrevPrev && PrevPrev->is(tok::identifier) && 337 Prev->isOneOf(tok::star, tok::amp, tok::ampamp) && 338 CurrentToken->is(tok::identifier) && Next->isNot(tok::equal)) { 339 Prev->setType(TT_BinaryOperator); 340 LookForDecls = false; 341 } 342 } 343 } 344 345 if (CurrentToken->Previous->is(TT_PointerOrReference) && 346 CurrentToken->Previous->Previous->isOneOf(tok::l_paren, 347 tok::coloncolon)) 348 ProbablyFunctionType = true; 349 if (CurrentToken->is(tok::comma)) 350 MightBeFunctionType = false; 351 if (CurrentToken->Previous->is(TT_BinaryOperator)) 352 Contexts.back().IsExpression = true; 353 if (CurrentToken->is(tok::r_paren)) { 354 if (MightBeFunctionType && ProbablyFunctionType && CurrentToken->Next && 355 (CurrentToken->Next->is(tok::l_paren) || 356 (CurrentToken->Next->is(tok::l_square) && Line.MustBeDeclaration))) 357 Left->setType(Left->Next->is(tok::caret) ? TT_ObjCBlockLParen 358 : TT_FunctionTypeLParen); 359 Left->MatchingParen = CurrentToken; 360 CurrentToken->MatchingParen = Left; 361 362 if (CurrentToken->Next && CurrentToken->Next->is(tok::l_brace) && 363 Left->Previous && Left->Previous->is(tok::l_paren)) { 364 // Detect the case where macros are used to generate lambdas or 365 // function bodies, e.g.: 366 // auto my_lambda = MACRO((Type *type, int i) { .. body .. }); 367 for (FormatToken *Tok = Left; Tok != CurrentToken; Tok = Tok->Next) 368 if (Tok->is(TT_BinaryOperator) && 369 Tok->isOneOf(tok::star, tok::amp, tok::ampamp)) 370 Tok->setType(TT_PointerOrReference); 371 } 372 373 if (StartsObjCMethodExpr) { 374 CurrentToken->setType(TT_ObjCMethodExpr); 375 if (Contexts.back().FirstObjCSelectorName) { 376 Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = 377 Contexts.back().LongestObjCSelectorName; 378 } 379 } 380 381 if (Left->is(TT_AttributeParen)) 382 CurrentToken->setType(TT_AttributeParen); 383 if (Left->is(TT_TypeDeclarationParen)) 384 CurrentToken->setType(TT_TypeDeclarationParen); 385 if (Left->Previous && Left->Previous->is(TT_JavaAnnotation)) 386 CurrentToken->setType(TT_JavaAnnotation); 387 if (Left->Previous && Left->Previous->is(TT_LeadingJavaAnnotation)) 388 CurrentToken->setType(TT_LeadingJavaAnnotation); 389 if (Left->Previous && Left->Previous->is(TT_AttributeSquare)) 390 CurrentToken->setType(TT_AttributeSquare); 391 392 if (!HasMultipleLines) 393 Left->setPackingKind(PPK_Inconclusive); 394 else if (HasMultipleParametersOnALine) 395 Left->setPackingKind(PPK_BinPacked); 396 else 397 Left->setPackingKind(PPK_OnePerLine); 398 399 next(); 400 return true; 401 } 402 if (CurrentToken->isOneOf(tok::r_square, tok::r_brace)) 403 return false; 404 405 if (CurrentToken->is(tok::l_brace)) 406 Left->setType(TT_Unknown); // Not TT_ObjCBlockLParen 407 if (CurrentToken->is(tok::comma) && CurrentToken->Next && 408 !CurrentToken->Next->HasUnescapedNewline && 409 !CurrentToken->Next->isTrailingComment()) 410 HasMultipleParametersOnALine = true; 411 bool ProbablyFunctionTypeLParen = 412 (CurrentToken->is(tok::l_paren) && CurrentToken->Next && 413 CurrentToken->Next->isOneOf(tok::star, tok::amp, tok::caret)); 414 if ((CurrentToken->Previous->isOneOf(tok::kw_const, tok::kw_auto) || 415 CurrentToken->Previous->isSimpleTypeSpecifier()) && 416 !(CurrentToken->is(tok::l_brace) || 417 (CurrentToken->is(tok::l_paren) && !ProbablyFunctionTypeLParen))) 418 Contexts.back().IsExpression = false; 419 if (CurrentToken->isOneOf(tok::semi, tok::colon)) { 420 MightBeObjCForRangeLoop = false; 421 if (PossibleObjCForInToken) { 422 PossibleObjCForInToken->setType(TT_Unknown); 423 PossibleObjCForInToken = nullptr; 424 } 425 } 426 if (MightBeObjCForRangeLoop && CurrentToken->is(Keywords.kw_in)) { 427 PossibleObjCForInToken = CurrentToken; 428 PossibleObjCForInToken->setType(TT_ObjCForIn); 429 } 430 // When we discover a 'new', we set CanBeExpression to 'false' in order to 431 // parse the type correctly. Reset that after a comma. 432 if (CurrentToken->is(tok::comma)) 433 Contexts.back().CanBeExpression = true; 434 435 FormatToken *Tok = CurrentToken; 436 if (!consumeToken()) 437 return false; 438 updateParameterCount(Left, Tok); 439 if (CurrentToken && CurrentToken->HasUnescapedNewline) 440 HasMultipleLines = true; 441 } 442 return false; 443 } 444 445 bool isCSharpAttributeSpecifier(const FormatToken &Tok) { 446 if (!Style.isCSharp()) 447 return false; 448 449 // `identifier[i]` is not an attribute. 450 if (Tok.Previous && Tok.Previous->is(tok::identifier)) 451 return false; 452 453 // Chains of [] in `identifier[i][j][k]` are not attributes. 454 if (Tok.Previous && Tok.Previous->is(tok::r_square)) { 455 auto *MatchingParen = Tok.Previous->MatchingParen; 456 if (!MatchingParen || MatchingParen->is(TT_ArraySubscriptLSquare)) 457 return false; 458 } 459 460 const FormatToken *AttrTok = Tok.Next; 461 if (!AttrTok) 462 return false; 463 464 // Just an empty declaration e.g. string []. 465 if (AttrTok->is(tok::r_square)) 466 return false; 467 468 // Move along the tokens inbetween the '[' and ']' e.g. [STAThread]. 469 while (AttrTok && AttrTok->isNot(tok::r_square)) 470 AttrTok = AttrTok->Next; 471 472 if (!AttrTok) 473 return false; 474 475 // Allow an attribute to be the only content of a file. 476 AttrTok = AttrTok->Next; 477 if (!AttrTok) 478 return true; 479 480 // Limit this to being an access modifier that follows. 481 if (AttrTok->isOneOf(tok::kw_public, tok::kw_private, tok::kw_protected, 482 tok::comment, tok::kw_class, tok::kw_static, 483 tok::l_square, Keywords.kw_internal)) 484 return true; 485 486 // incase its a [XXX] retval func(.... 487 if (AttrTok->Next && 488 AttrTok->Next->startsSequence(tok::identifier, tok::l_paren)) 489 return true; 490 491 return false; 492 } 493 494 bool isCpp11AttributeSpecifier(const FormatToken &Tok) { 495 if (!Style.isCpp() || !Tok.startsSequence(tok::l_square, tok::l_square)) 496 return false; 497 // The first square bracket is part of an ObjC array literal 498 if (Tok.Previous && Tok.Previous->is(tok::at)) 499 return false; 500 const FormatToken *AttrTok = Tok.Next->Next; 501 if (!AttrTok) 502 return false; 503 // C++17 '[[using ns: foo, bar(baz, blech)]]' 504 // We assume nobody will name an ObjC variable 'using'. 505 if (AttrTok->startsSequence(tok::kw_using, tok::identifier, tok::colon)) 506 return true; 507 if (AttrTok->isNot(tok::identifier)) 508 return false; 509 while (AttrTok && !AttrTok->startsSequence(tok::r_square, tok::r_square)) { 510 // ObjC message send. We assume nobody will use : in a C++11 attribute 511 // specifier parameter, although this is technically valid: 512 // [[foo(:)]]. 513 if (AttrTok->is(tok::colon) || 514 AttrTok->startsSequence(tok::identifier, tok::identifier) || 515 AttrTok->startsSequence(tok::r_paren, tok::identifier)) 516 return false; 517 if (AttrTok->is(tok::ellipsis)) 518 return true; 519 AttrTok = AttrTok->Next; 520 } 521 return AttrTok && AttrTok->startsSequence(tok::r_square, tok::r_square); 522 } 523 524 bool parseSquare() { 525 if (!CurrentToken) 526 return false; 527 528 // A '[' could be an index subscript (after an identifier or after 529 // ')' or ']'), it could be the start of an Objective-C method 530 // expression, it could the start of an Objective-C array literal, 531 // or it could be a C++ attribute specifier [[foo::bar]]. 532 FormatToken *Left = CurrentToken->Previous; 533 Left->ParentBracket = Contexts.back().ContextKind; 534 FormatToken *Parent = Left->getPreviousNonComment(); 535 536 // Cases where '>' is followed by '['. 537 // In C++, this can happen either in array of templates (foo<int>[10]) 538 // or when array is a nested template type (unique_ptr<type1<type2>[]>). 539 bool CppArrayTemplates = 540 Style.isCpp() && Parent && Parent->is(TT_TemplateCloser) && 541 (Contexts.back().CanBeExpression || Contexts.back().IsExpression || 542 Contexts.back().InTemplateArgument); 543 544 bool IsCpp11AttributeSpecifier = isCpp11AttributeSpecifier(*Left) || 545 Contexts.back().InCpp11AttributeSpecifier; 546 547 // Treat C# Attributes [STAThread] much like C++ attributes [[...]]. 548 bool IsCSharpAttributeSpecifier = 549 isCSharpAttributeSpecifier(*Left) || 550 Contexts.back().InCSharpAttributeSpecifier; 551 552 bool InsideInlineASM = Line.startsWith(tok::kw_asm); 553 bool IsCppStructuredBinding = Left->isCppStructuredBinding(Style); 554 bool StartsObjCMethodExpr = 555 !IsCppStructuredBinding && !InsideInlineASM && !CppArrayTemplates && 556 Style.isCpp() && !IsCpp11AttributeSpecifier && 557 !IsCSharpAttributeSpecifier && Contexts.back().CanBeExpression && 558 Left->isNot(TT_LambdaLSquare) && 559 !CurrentToken->isOneOf(tok::l_brace, tok::r_square) && 560 (!Parent || 561 Parent->isOneOf(tok::colon, tok::l_square, tok::l_paren, 562 tok::kw_return, tok::kw_throw) || 563 Parent->isUnaryOperator() || 564 // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen. 565 Parent->isOneOf(TT_ObjCForIn, TT_CastRParen) || 566 (getBinOpPrecedence(Parent->Tok.getKind(), true, true) > 567 prec::Unknown)); 568 bool ColonFound = false; 569 570 unsigned BindingIncrease = 1; 571 if (IsCppStructuredBinding) { 572 Left->setType(TT_StructuredBindingLSquare); 573 } else if (Left->is(TT_Unknown)) { 574 if (StartsObjCMethodExpr) { 575 Left->setType(TT_ObjCMethodExpr); 576 } else if (InsideInlineASM) { 577 Left->setType(TT_InlineASMSymbolicNameLSquare); 578 } else if (IsCpp11AttributeSpecifier) { 579 Left->setType(TT_AttributeSquare); 580 } else if (Style.isJavaScript() && Parent && 581 Contexts.back().ContextKind == tok::l_brace && 582 Parent->isOneOf(tok::l_brace, tok::comma)) { 583 Left->setType(TT_JsComputedPropertyName); 584 } else if (Style.isCpp() && Contexts.back().ContextKind == tok::l_brace && 585 Parent && Parent->isOneOf(tok::l_brace, tok::comma)) { 586 Left->setType(TT_DesignatedInitializerLSquare); 587 } else if (IsCSharpAttributeSpecifier) { 588 Left->setType(TT_AttributeSquare); 589 } else if (CurrentToken->is(tok::r_square) && Parent && 590 Parent->is(TT_TemplateCloser)) { 591 Left->setType(TT_ArraySubscriptLSquare); 592 } else if (Style.Language == FormatStyle::LK_Proto || 593 Style.Language == FormatStyle::LK_TextProto) { 594 // Square braces in LK_Proto can either be message field attributes: 595 // 596 // optional Aaa aaa = 1 [ 597 // (aaa) = aaa 598 // ]; 599 // 600 // extensions 123 [ 601 // (aaa) = aaa 602 // ]; 603 // 604 // or text proto extensions (in options): 605 // 606 // option (Aaa.options) = { 607 // [type.type/type] { 608 // key: value 609 // } 610 // } 611 // 612 // or repeated fields (in options): 613 // 614 // option (Aaa.options) = { 615 // keys: [ 1, 2, 3 ] 616 // } 617 // 618 // In the first and the third case we want to spread the contents inside 619 // the square braces; in the second we want to keep them inline. 620 Left->setType(TT_ArrayInitializerLSquare); 621 if (!Left->endsSequence(tok::l_square, tok::numeric_constant, 622 tok::equal) && 623 !Left->endsSequence(tok::l_square, tok::numeric_constant, 624 tok::identifier) && 625 !Left->endsSequence(tok::l_square, tok::colon, TT_SelectorName)) { 626 Left->setType(TT_ProtoExtensionLSquare); 627 BindingIncrease = 10; 628 } 629 } else if (!CppArrayTemplates && Parent && 630 Parent->isOneOf(TT_BinaryOperator, TT_TemplateCloser, tok::at, 631 tok::comma, tok::l_paren, tok::l_square, 632 tok::question, tok::colon, tok::kw_return, 633 // Should only be relevant to JavaScript: 634 tok::kw_default)) { 635 Left->setType(TT_ArrayInitializerLSquare); 636 } else { 637 BindingIncrease = 10; 638 Left->setType(TT_ArraySubscriptLSquare); 639 } 640 } 641 642 ScopedContextCreator ContextCreator(*this, tok::l_square, BindingIncrease); 643 Contexts.back().IsExpression = true; 644 if (Style.isJavaScript() && Parent && Parent->is(TT_JsTypeColon)) 645 Contexts.back().IsExpression = false; 646 647 Contexts.back().ColonIsObjCMethodExpr = StartsObjCMethodExpr; 648 Contexts.back().InCpp11AttributeSpecifier = IsCpp11AttributeSpecifier; 649 Contexts.back().InCSharpAttributeSpecifier = IsCSharpAttributeSpecifier; 650 651 while (CurrentToken) { 652 if (CurrentToken->is(tok::r_square)) { 653 if (IsCpp11AttributeSpecifier) 654 CurrentToken->setType(TT_AttributeSquare); 655 if (IsCSharpAttributeSpecifier) 656 CurrentToken->setType(TT_AttributeSquare); 657 else if (((CurrentToken->Next && 658 CurrentToken->Next->is(tok::l_paren)) || 659 (CurrentToken->Previous && 660 CurrentToken->Previous->Previous == Left)) && 661 Left->is(TT_ObjCMethodExpr)) { 662 // An ObjC method call is rarely followed by an open parenthesis. It 663 // also can't be composed of just one token, unless it's a macro that 664 // will be expanded to more tokens. 665 // FIXME: Do we incorrectly label ":" with this? 666 StartsObjCMethodExpr = false; 667 Left->setType(TT_Unknown); 668 } 669 if (StartsObjCMethodExpr && CurrentToken->Previous != Left) { 670 CurrentToken->setType(TT_ObjCMethodExpr); 671 // If we haven't seen a colon yet, make sure the last identifier 672 // before the r_square is tagged as a selector name component. 673 if (!ColonFound && CurrentToken->Previous && 674 CurrentToken->Previous->is(TT_Unknown) && 675 canBeObjCSelectorComponent(*CurrentToken->Previous)) 676 CurrentToken->Previous->setType(TT_SelectorName); 677 // determineStarAmpUsage() thinks that '*' '[' is allocating an 678 // array of pointers, but if '[' starts a selector then '*' is a 679 // binary operator. 680 if (Parent && Parent->is(TT_PointerOrReference)) 681 Parent->setType(TT_BinaryOperator); 682 } 683 // An arrow after an ObjC method expression is not a lambda arrow. 684 if (CurrentToken->getType() == TT_ObjCMethodExpr && 685 CurrentToken->Next && CurrentToken->Next->is(TT_LambdaArrow)) 686 CurrentToken->Next->setType(TT_Unknown); 687 Left->MatchingParen = CurrentToken; 688 CurrentToken->MatchingParen = Left; 689 // FirstObjCSelectorName is set when a colon is found. This does 690 // not work, however, when the method has no parameters. 691 // Here, we set FirstObjCSelectorName when the end of the method call is 692 // reached, in case it was not set already. 693 if (!Contexts.back().FirstObjCSelectorName) { 694 FormatToken *Previous = CurrentToken->getPreviousNonComment(); 695 if (Previous && Previous->is(TT_SelectorName)) { 696 Previous->ObjCSelectorNameParts = 1; 697 Contexts.back().FirstObjCSelectorName = Previous; 698 } 699 } else { 700 Left->ParameterCount = 701 Contexts.back().FirstObjCSelectorName->ObjCSelectorNameParts; 702 } 703 if (Contexts.back().FirstObjCSelectorName) { 704 Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = 705 Contexts.back().LongestObjCSelectorName; 706 if (Left->BlockParameterCount > 1) 707 Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = 0; 708 } 709 next(); 710 return true; 711 } 712 if (CurrentToken->isOneOf(tok::r_paren, tok::r_brace)) 713 return false; 714 if (CurrentToken->is(tok::colon)) { 715 if (IsCpp11AttributeSpecifier && 716 CurrentToken->endsSequence(tok::colon, tok::identifier, 717 tok::kw_using)) { 718 // Remember that this is a [[using ns: foo]] C++ attribute, so we 719 // don't add a space before the colon (unlike other colons). 720 CurrentToken->setType(TT_AttributeColon); 721 } else if (Left->isOneOf(TT_ArraySubscriptLSquare, 722 TT_DesignatedInitializerLSquare)) { 723 Left->setType(TT_ObjCMethodExpr); 724 StartsObjCMethodExpr = true; 725 Contexts.back().ColonIsObjCMethodExpr = true; 726 if (Parent && Parent->is(tok::r_paren)) 727 // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen. 728 Parent->setType(TT_CastRParen); 729 } 730 ColonFound = true; 731 } 732 if (CurrentToken->is(tok::comma) && Left->is(TT_ObjCMethodExpr) && 733 !ColonFound) 734 Left->setType(TT_ArrayInitializerLSquare); 735 FormatToken *Tok = CurrentToken; 736 if (!consumeToken()) 737 return false; 738 updateParameterCount(Left, Tok); 739 } 740 return false; 741 } 742 743 bool couldBeInStructArrayInitializer() const { 744 if (Contexts.size() < 2) 745 return false; 746 // We want to back up no more then 2 context levels i.e. 747 // . { { <- 748 const auto End = std::next(Contexts.rbegin(), 2); 749 auto Last = Contexts.rbegin(); 750 unsigned Depth = 0; 751 for (; Last != End; ++Last) 752 if (Last->ContextKind == tok::l_brace) 753 ++Depth; 754 return Depth == 2 && Last->ContextKind != tok::l_brace; 755 } 756 757 bool parseBrace() { 758 if (CurrentToken) { 759 FormatToken *Left = CurrentToken->Previous; 760 Left->ParentBracket = Contexts.back().ContextKind; 761 762 if (Contexts.back().CaretFound) 763 Left->setType(TT_ObjCBlockLBrace); 764 Contexts.back().CaretFound = false; 765 766 ScopedContextCreator ContextCreator(*this, tok::l_brace, 1); 767 Contexts.back().ColonIsDictLiteral = true; 768 if (Left->is(BK_BracedInit)) 769 Contexts.back().IsExpression = true; 770 if (Style.isJavaScript() && Left->Previous && 771 Left->Previous->is(TT_JsTypeColon)) 772 Contexts.back().IsExpression = false; 773 774 unsigned CommaCount = 0; 775 while (CurrentToken) { 776 if (CurrentToken->is(tok::r_brace)) { 777 assert(Left->Optional == CurrentToken->Optional); 778 Left->MatchingParen = CurrentToken; 779 CurrentToken->MatchingParen = Left; 780 if (Style.AlignArrayOfStructures != FormatStyle::AIAS_None) { 781 if (Left->ParentBracket == tok::l_brace && 782 couldBeInStructArrayInitializer() && CommaCount > 0) 783 Contexts.back().InStructArrayInitializer = true; 784 } 785 next(); 786 return true; 787 } 788 if (CurrentToken->isOneOf(tok::r_paren, tok::r_square)) 789 return false; 790 updateParameterCount(Left, CurrentToken); 791 if (CurrentToken->isOneOf(tok::colon, tok::l_brace, tok::less)) { 792 FormatToken *Previous = CurrentToken->getPreviousNonComment(); 793 if (Previous->is(TT_JsTypeOptionalQuestion)) 794 Previous = Previous->getPreviousNonComment(); 795 if ((CurrentToken->is(tok::colon) && 796 (!Contexts.back().ColonIsDictLiteral || !Style.isCpp())) || 797 Style.Language == FormatStyle::LK_Proto || 798 Style.Language == FormatStyle::LK_TextProto) { 799 Left->setType(TT_DictLiteral); 800 if (Previous->Tok.getIdentifierInfo() || 801 Previous->is(tok::string_literal)) 802 Previous->setType(TT_SelectorName); 803 } 804 if (CurrentToken->is(tok::colon) || Style.isJavaScript()) 805 Left->setType(TT_DictLiteral); 806 } 807 if (CurrentToken->is(tok::comma)) { 808 if (Style.isJavaScript()) 809 Left->setType(TT_DictLiteral); 810 ++CommaCount; 811 } 812 if (!consumeToken()) 813 return false; 814 } 815 } 816 return true; 817 } 818 819 void updateParameterCount(FormatToken *Left, FormatToken *Current) { 820 // For ObjC methods, the number of parameters is calculated differently as 821 // method declarations have a different structure (the parameters are not 822 // inside a bracket scope). 823 if (Current->is(tok::l_brace) && Current->is(BK_Block)) 824 ++Left->BlockParameterCount; 825 if (Current->is(tok::comma)) { 826 ++Left->ParameterCount; 827 if (!Left->Role) 828 Left->Role.reset(new CommaSeparatedList(Style)); 829 Left->Role->CommaFound(Current); 830 } else if (Left->ParameterCount == 0 && Current->isNot(tok::comment)) { 831 Left->ParameterCount = 1; 832 } 833 } 834 835 bool parseConditional() { 836 while (CurrentToken) { 837 if (CurrentToken->is(tok::colon)) { 838 CurrentToken->setType(TT_ConditionalExpr); 839 next(); 840 return true; 841 } 842 if (!consumeToken()) 843 return false; 844 } 845 return false; 846 } 847 848 bool parseTemplateDeclaration() { 849 if (CurrentToken && CurrentToken->is(tok::less)) { 850 CurrentToken->setType(TT_TemplateOpener); 851 next(); 852 if (!parseAngle()) 853 return false; 854 if (CurrentToken) 855 CurrentToken->Previous->ClosesTemplateDeclaration = true; 856 return true; 857 } 858 return false; 859 } 860 861 bool consumeToken() { 862 FormatToken *Tok = CurrentToken; 863 next(); 864 switch (Tok->Tok.getKind()) { 865 case tok::plus: 866 case tok::minus: 867 if (!Tok->Previous && Line.MustBeDeclaration) 868 Tok->setType(TT_ObjCMethodSpecifier); 869 break; 870 case tok::colon: 871 if (!Tok->Previous) 872 return false; 873 // Colons from ?: are handled in parseConditional(). 874 if (Style.isJavaScript()) { 875 if (Contexts.back().ColonIsForRangeExpr || // colon in for loop 876 (Contexts.size() == 1 && // switch/case labels 877 !Line.First->isOneOf(tok::kw_enum, tok::kw_case)) || 878 Contexts.back().ContextKind == tok::l_paren || // function params 879 Contexts.back().ContextKind == tok::l_square || // array type 880 (!Contexts.back().IsExpression && 881 Contexts.back().ContextKind == tok::l_brace) || // object type 882 (Contexts.size() == 1 && 883 Line.MustBeDeclaration)) { // method/property declaration 884 Contexts.back().IsExpression = false; 885 Tok->setType(TT_JsTypeColon); 886 break; 887 } 888 } else if (Style.isCSharp()) { 889 if (Contexts.back().InCSharpAttributeSpecifier) { 890 Tok->setType(TT_AttributeColon); 891 break; 892 } 893 if (Contexts.back().ContextKind == tok::l_paren) { 894 Tok->setType(TT_CSharpNamedArgumentColon); 895 break; 896 } 897 } 898 if (Line.First->isOneOf(Keywords.kw_module, Keywords.kw_import) || 899 Line.First->startsSequence(tok::kw_export, Keywords.kw_module) || 900 Line.First->startsSequence(tok::kw_export, Keywords.kw_import)) { 901 Tok->setType(TT_ModulePartitionColon); 902 } else if (Contexts.back().ColonIsDictLiteral || 903 Style.Language == FormatStyle::LK_Proto || 904 Style.Language == FormatStyle::LK_TextProto) { 905 Tok->setType(TT_DictLiteral); 906 if (Style.Language == FormatStyle::LK_TextProto) { 907 if (FormatToken *Previous = Tok->getPreviousNonComment()) 908 Previous->setType(TT_SelectorName); 909 } 910 } else if (Contexts.back().ColonIsObjCMethodExpr || 911 Line.startsWith(TT_ObjCMethodSpecifier)) { 912 Tok->setType(TT_ObjCMethodExpr); 913 const FormatToken *BeforePrevious = Tok->Previous->Previous; 914 // Ensure we tag all identifiers in method declarations as 915 // TT_SelectorName. 916 bool UnknownIdentifierInMethodDeclaration = 917 Line.startsWith(TT_ObjCMethodSpecifier) && 918 Tok->Previous->is(tok::identifier) && Tok->Previous->is(TT_Unknown); 919 if (!BeforePrevious || 920 // FIXME(bug 36976): ObjC return types shouldn't use TT_CastRParen. 921 !(BeforePrevious->is(TT_CastRParen) || 922 (BeforePrevious->is(TT_ObjCMethodExpr) && 923 BeforePrevious->is(tok::colon))) || 924 BeforePrevious->is(tok::r_square) || 925 Contexts.back().LongestObjCSelectorName == 0 || 926 UnknownIdentifierInMethodDeclaration) { 927 Tok->Previous->setType(TT_SelectorName); 928 if (!Contexts.back().FirstObjCSelectorName) 929 Contexts.back().FirstObjCSelectorName = Tok->Previous; 930 else if (Tok->Previous->ColumnWidth > 931 Contexts.back().LongestObjCSelectorName) 932 Contexts.back().LongestObjCSelectorName = 933 Tok->Previous->ColumnWidth; 934 Tok->Previous->ParameterIndex = 935 Contexts.back().FirstObjCSelectorName->ObjCSelectorNameParts; 936 ++Contexts.back().FirstObjCSelectorName->ObjCSelectorNameParts; 937 } 938 } else if (Contexts.back().ColonIsForRangeExpr) { 939 Tok->setType(TT_RangeBasedForLoopColon); 940 } else if (CurrentToken && CurrentToken->is(tok::numeric_constant)) { 941 Tok->setType(TT_BitFieldColon); 942 } else if (Contexts.size() == 1 && 943 !Line.First->isOneOf(tok::kw_enum, tok::kw_case, 944 tok::kw_default)) { 945 FormatToken *Prev = Tok->getPreviousNonComment(); 946 if (!Prev) 947 break; 948 if (Prev->isOneOf(tok::r_paren, tok::kw_noexcept)) 949 Tok->setType(TT_CtorInitializerColon); 950 else if (Prev->is(tok::kw_try)) { 951 // Member initializer list within function try block. 952 FormatToken *PrevPrev = Prev->getPreviousNonComment(); 953 if (!PrevPrev) 954 break; 955 if (PrevPrev && PrevPrev->isOneOf(tok::r_paren, tok::kw_noexcept)) 956 Tok->setType(TT_CtorInitializerColon); 957 } else 958 Tok->setType(TT_InheritanceColon); 959 } else if (canBeObjCSelectorComponent(*Tok->Previous) && Tok->Next && 960 (Tok->Next->isOneOf(tok::r_paren, tok::comma) || 961 (canBeObjCSelectorComponent(*Tok->Next) && Tok->Next->Next && 962 Tok->Next->Next->is(tok::colon)))) { 963 // This handles a special macro in ObjC code where selectors including 964 // the colon are passed as macro arguments. 965 Tok->setType(TT_ObjCMethodExpr); 966 } else if (Contexts.back().ContextKind == tok::l_paren) { 967 Tok->setType(TT_InlineASMColon); 968 } 969 break; 970 case tok::pipe: 971 case tok::amp: 972 // | and & in declarations/type expressions represent union and 973 // intersection types, respectively. 974 if (Style.isJavaScript() && !Contexts.back().IsExpression) 975 Tok->setType(TT_JsTypeOperator); 976 break; 977 case tok::kw_if: 978 case tok::kw_while: 979 if (Tok->is(tok::kw_if) && CurrentToken && 980 CurrentToken->isOneOf(tok::kw_constexpr, tok::identifier)) 981 next(); 982 if (CurrentToken && CurrentToken->is(tok::l_paren)) { 983 next(); 984 if (!parseParens(/*LookForDecls=*/true)) 985 return false; 986 } 987 break; 988 case tok::kw_for: 989 if (Style.isJavaScript()) { 990 // x.for and {for: ...} 991 if ((Tok->Previous && Tok->Previous->is(tok::period)) || 992 (Tok->Next && Tok->Next->is(tok::colon))) 993 break; 994 // JS' for await ( ... 995 if (CurrentToken && CurrentToken->is(Keywords.kw_await)) 996 next(); 997 } 998 if (Style.isCpp() && CurrentToken && CurrentToken->is(tok::kw_co_await)) 999 next(); 1000 Contexts.back().ColonIsForRangeExpr = true; 1001 next(); 1002 if (!parseParens()) 1003 return false; 1004 break; 1005 case tok::l_paren: 1006 // When faced with 'operator()()', the kw_operator handler incorrectly 1007 // marks the first l_paren as a OverloadedOperatorLParen. Here, we make 1008 // the first two parens OverloadedOperators and the second l_paren an 1009 // OverloadedOperatorLParen. 1010 if (Tok->Previous && Tok->Previous->is(tok::r_paren) && 1011 Tok->Previous->MatchingParen && 1012 Tok->Previous->MatchingParen->is(TT_OverloadedOperatorLParen)) { 1013 Tok->Previous->setType(TT_OverloadedOperator); 1014 Tok->Previous->MatchingParen->setType(TT_OverloadedOperator); 1015 Tok->setType(TT_OverloadedOperatorLParen); 1016 } 1017 1018 if (!parseParens()) 1019 return false; 1020 if (Line.MustBeDeclaration && Contexts.size() == 1 && 1021 !Contexts.back().IsExpression && !Line.startsWith(TT_ObjCProperty) && 1022 !Tok->is(TT_TypeDeclarationParen) && 1023 (!Tok->Previous || !Tok->Previous->isOneOf(tok::kw___attribute, 1024 TT_LeadingJavaAnnotation))) 1025 Line.MightBeFunctionDecl = true; 1026 break; 1027 case tok::l_square: 1028 if (!parseSquare()) 1029 return false; 1030 break; 1031 case tok::l_brace: 1032 if (Style.Language == FormatStyle::LK_TextProto) { 1033 FormatToken *Previous = Tok->getPreviousNonComment(); 1034 if (Previous && Previous->getType() != TT_DictLiteral) 1035 Previous->setType(TT_SelectorName); 1036 } 1037 if (!parseBrace()) 1038 return false; 1039 break; 1040 case tok::less: 1041 if (parseAngle()) { 1042 Tok->setType(TT_TemplateOpener); 1043 // In TT_Proto, we must distignuish between: 1044 // map<key, value> 1045 // msg < item: data > 1046 // msg: < item: data > 1047 // In TT_TextProto, map<key, value> does not occur. 1048 if (Style.Language == FormatStyle::LK_TextProto || 1049 (Style.Language == FormatStyle::LK_Proto && Tok->Previous && 1050 Tok->Previous->isOneOf(TT_SelectorName, TT_DictLiteral))) { 1051 Tok->setType(TT_DictLiteral); 1052 FormatToken *Previous = Tok->getPreviousNonComment(); 1053 if (Previous && Previous->getType() != TT_DictLiteral) 1054 Previous->setType(TT_SelectorName); 1055 } 1056 } else { 1057 Tok->setType(TT_BinaryOperator); 1058 NonTemplateLess.insert(Tok); 1059 CurrentToken = Tok; 1060 next(); 1061 } 1062 break; 1063 case tok::r_paren: 1064 case tok::r_square: 1065 return false; 1066 case tok::r_brace: 1067 // Lines can start with '}'. 1068 if (Tok->Previous) 1069 return false; 1070 break; 1071 case tok::greater: 1072 if (Style.Language != FormatStyle::LK_TextProto) 1073 Tok->setType(TT_BinaryOperator); 1074 if (Tok->Previous && Tok->Previous->is(TT_TemplateCloser)) 1075 Tok->SpacesRequiredBefore = 1; 1076 break; 1077 case tok::kw_operator: 1078 if (Style.Language == FormatStyle::LK_TextProto || 1079 Style.Language == FormatStyle::LK_Proto) 1080 break; 1081 while (CurrentToken && 1082 !CurrentToken->isOneOf(tok::l_paren, tok::semi, tok::r_paren)) { 1083 if (CurrentToken->isOneOf(tok::star, tok::amp)) 1084 CurrentToken->setType(TT_PointerOrReference); 1085 consumeToken(); 1086 if (CurrentToken && CurrentToken->is(tok::comma) && 1087 CurrentToken->Previous->isNot(tok::kw_operator)) 1088 break; 1089 if (CurrentToken && CurrentToken->Previous->isOneOf( 1090 TT_BinaryOperator, TT_UnaryOperator, tok::comma, 1091 tok::star, tok::arrow, tok::amp, tok::ampamp)) 1092 CurrentToken->Previous->setType(TT_OverloadedOperator); 1093 } 1094 if (CurrentToken && CurrentToken->is(tok::l_paren)) 1095 CurrentToken->setType(TT_OverloadedOperatorLParen); 1096 if (CurrentToken && CurrentToken->Previous->is(TT_BinaryOperator)) 1097 CurrentToken->Previous->setType(TT_OverloadedOperator); 1098 break; 1099 case tok::question: 1100 if (Style.isJavaScript() && Tok->Next && 1101 Tok->Next->isOneOf(tok::semi, tok::comma, tok::colon, tok::r_paren, 1102 tok::r_brace)) { 1103 // Question marks before semicolons, colons, etc. indicate optional 1104 // types (fields, parameters), e.g. 1105 // function(x?: string, y?) {...} 1106 // class X { y?; } 1107 Tok->setType(TT_JsTypeOptionalQuestion); 1108 break; 1109 } 1110 // Declarations cannot be conditional expressions, this can only be part 1111 // of a type declaration. 1112 if (Line.MustBeDeclaration && !Contexts.back().IsExpression && 1113 Style.isJavaScript()) 1114 break; 1115 if (Style.isCSharp()) { 1116 // `Type?)`, `Type?>`, `Type? name;` and `Type? name =` can only be 1117 // nullable types. 1118 // Line.MustBeDeclaration will be true for `Type? name;`. 1119 if ((!Contexts.back().IsExpression && Line.MustBeDeclaration) || 1120 (Tok->Next && Tok->Next->isOneOf(tok::r_paren, tok::greater)) || 1121 (Tok->Next && Tok->Next->is(tok::identifier) && Tok->Next->Next && 1122 Tok->Next->Next->is(tok::equal))) { 1123 Tok->setType(TT_CSharpNullable); 1124 break; 1125 } 1126 } 1127 parseConditional(); 1128 break; 1129 case tok::kw_template: 1130 parseTemplateDeclaration(); 1131 break; 1132 case tok::comma: 1133 if (Contexts.back().InCtorInitializer) 1134 Tok->setType(TT_CtorInitializerComma); 1135 else if (Contexts.back().InInheritanceList) 1136 Tok->setType(TT_InheritanceComma); 1137 else if (Contexts.back().FirstStartOfName && 1138 (Contexts.size() == 1 || Line.startsWith(tok::kw_for))) { 1139 Contexts.back().FirstStartOfName->PartOfMultiVariableDeclStmt = true; 1140 Line.IsMultiVariableDeclStmt = true; 1141 } 1142 if (Contexts.back().IsForEachMacro) 1143 Contexts.back().IsExpression = true; 1144 break; 1145 case tok::identifier: 1146 if (Tok->isOneOf(Keywords.kw___has_include, 1147 Keywords.kw___has_include_next)) 1148 parseHasInclude(); 1149 if (Style.isCSharp() && Tok->is(Keywords.kw_where) && Tok->Next && 1150 Tok->Next->isNot(tok::l_paren)) { 1151 Tok->setType(TT_CSharpGenericTypeConstraint); 1152 parseCSharpGenericTypeConstraint(); 1153 } 1154 break; 1155 default: 1156 break; 1157 } 1158 return true; 1159 } 1160 1161 void parseCSharpGenericTypeConstraint() { 1162 int OpenAngleBracketsCount = 0; 1163 while (CurrentToken) { 1164 if (CurrentToken->is(tok::less)) { 1165 // parseAngle is too greedy and will consume the whole line. 1166 CurrentToken->setType(TT_TemplateOpener); 1167 ++OpenAngleBracketsCount; 1168 next(); 1169 } else if (CurrentToken->is(tok::greater)) { 1170 CurrentToken->setType(TT_TemplateCloser); 1171 --OpenAngleBracketsCount; 1172 next(); 1173 } else if (CurrentToken->is(tok::comma) && OpenAngleBracketsCount == 0) { 1174 // We allow line breaks after GenericTypeConstraintComma's 1175 // so do not flag commas in Generics as GenericTypeConstraintComma's. 1176 CurrentToken->setType(TT_CSharpGenericTypeConstraintComma); 1177 next(); 1178 } else if (CurrentToken->is(Keywords.kw_where)) { 1179 CurrentToken->setType(TT_CSharpGenericTypeConstraint); 1180 next(); 1181 } else if (CurrentToken->is(tok::colon)) { 1182 CurrentToken->setType(TT_CSharpGenericTypeConstraintColon); 1183 next(); 1184 } else { 1185 next(); 1186 } 1187 } 1188 } 1189 1190 void parseIncludeDirective() { 1191 if (CurrentToken && CurrentToken->is(tok::less)) { 1192 next(); 1193 while (CurrentToken) { 1194 // Mark tokens up to the trailing line comments as implicit string 1195 // literals. 1196 if (CurrentToken->isNot(tok::comment) && 1197 !CurrentToken->TokenText.startswith("//")) 1198 CurrentToken->setType(TT_ImplicitStringLiteral); 1199 next(); 1200 } 1201 } 1202 } 1203 1204 void parseWarningOrError() { 1205 next(); 1206 // We still want to format the whitespace left of the first token of the 1207 // warning or error. 1208 next(); 1209 while (CurrentToken) { 1210 CurrentToken->setType(TT_ImplicitStringLiteral); 1211 next(); 1212 } 1213 } 1214 1215 void parsePragma() { 1216 next(); // Consume "pragma". 1217 if (CurrentToken && 1218 CurrentToken->isOneOf(Keywords.kw_mark, Keywords.kw_option)) { 1219 bool IsMark = CurrentToken->is(Keywords.kw_mark); 1220 next(); // Consume "mark". 1221 next(); // Consume first token (so we fix leading whitespace). 1222 while (CurrentToken) { 1223 if (IsMark || CurrentToken->Previous->is(TT_BinaryOperator)) 1224 CurrentToken->setType(TT_ImplicitStringLiteral); 1225 next(); 1226 } 1227 } 1228 } 1229 1230 void parseHasInclude() { 1231 if (!CurrentToken || !CurrentToken->is(tok::l_paren)) 1232 return; 1233 next(); // '(' 1234 parseIncludeDirective(); 1235 next(); // ')' 1236 } 1237 1238 LineType parsePreprocessorDirective() { 1239 bool IsFirstToken = CurrentToken->IsFirst; 1240 LineType Type = LT_PreprocessorDirective; 1241 next(); 1242 if (!CurrentToken) 1243 return Type; 1244 1245 if (Style.isJavaScript() && IsFirstToken) { 1246 // JavaScript files can contain shebang lines of the form: 1247 // #!/usr/bin/env node 1248 // Treat these like C++ #include directives. 1249 while (CurrentToken) { 1250 // Tokens cannot be comments here. 1251 CurrentToken->setType(TT_ImplicitStringLiteral); 1252 next(); 1253 } 1254 return LT_ImportStatement; 1255 } 1256 1257 if (CurrentToken->Tok.is(tok::numeric_constant)) { 1258 CurrentToken->SpacesRequiredBefore = 1; 1259 return Type; 1260 } 1261 // Hashes in the middle of a line can lead to any strange token 1262 // sequence. 1263 if (!CurrentToken->Tok.getIdentifierInfo()) 1264 return Type; 1265 switch (CurrentToken->Tok.getIdentifierInfo()->getPPKeywordID()) { 1266 case tok::pp_include: 1267 case tok::pp_include_next: 1268 case tok::pp_import: 1269 next(); 1270 parseIncludeDirective(); 1271 Type = LT_ImportStatement; 1272 break; 1273 case tok::pp_error: 1274 case tok::pp_warning: 1275 parseWarningOrError(); 1276 break; 1277 case tok::pp_pragma: 1278 parsePragma(); 1279 break; 1280 case tok::pp_if: 1281 case tok::pp_elif: 1282 Contexts.back().IsExpression = true; 1283 next(); 1284 parseLine(); 1285 break; 1286 default: 1287 break; 1288 } 1289 while (CurrentToken) { 1290 FormatToken *Tok = CurrentToken; 1291 next(); 1292 if (Tok->is(tok::l_paren)) 1293 parseParens(); 1294 else if (Tok->isOneOf(Keywords.kw___has_include, 1295 Keywords.kw___has_include_next)) 1296 parseHasInclude(); 1297 } 1298 return Type; 1299 } 1300 1301 public: 1302 LineType parseLine() { 1303 if (!CurrentToken) 1304 return LT_Invalid; 1305 NonTemplateLess.clear(); 1306 if (CurrentToken->is(tok::hash)) 1307 return parsePreprocessorDirective(); 1308 1309 // Directly allow to 'import <string-literal>' to support protocol buffer 1310 // definitions (github.com/google/protobuf) or missing "#" (either way we 1311 // should not break the line). 1312 IdentifierInfo *Info = CurrentToken->Tok.getIdentifierInfo(); 1313 if ((Style.Language == FormatStyle::LK_Java && 1314 CurrentToken->is(Keywords.kw_package)) || 1315 (Info && Info->getPPKeywordID() == tok::pp_import && 1316 CurrentToken->Next && 1317 CurrentToken->Next->isOneOf(tok::string_literal, tok::identifier, 1318 tok::kw_static))) { 1319 next(); 1320 parseIncludeDirective(); 1321 return LT_ImportStatement; 1322 } 1323 1324 // If this line starts and ends in '<' and '>', respectively, it is likely 1325 // part of "#define <a/b.h>". 1326 if (CurrentToken->is(tok::less) && Line.Last->is(tok::greater)) { 1327 parseIncludeDirective(); 1328 return LT_ImportStatement; 1329 } 1330 1331 // In .proto files, top-level options and package statements are very 1332 // similar to import statements and should not be line-wrapped. 1333 if (Style.Language == FormatStyle::LK_Proto && Line.Level == 0 && 1334 CurrentToken->isOneOf(Keywords.kw_option, Keywords.kw_package)) { 1335 next(); 1336 if (CurrentToken && CurrentToken->is(tok::identifier)) { 1337 while (CurrentToken) 1338 next(); 1339 return LT_ImportStatement; 1340 } 1341 } 1342 1343 bool KeywordVirtualFound = false; 1344 bool ImportStatement = false; 1345 1346 // import {...} from '...'; 1347 if (Style.isJavaScript() && CurrentToken->is(Keywords.kw_import)) 1348 ImportStatement = true; 1349 1350 while (CurrentToken) { 1351 if (CurrentToken->is(tok::kw_virtual)) 1352 KeywordVirtualFound = true; 1353 if (Style.isJavaScript()) { 1354 // export {...} from '...'; 1355 // An export followed by "from 'some string';" is a re-export from 1356 // another module identified by a URI and is treated as a 1357 // LT_ImportStatement (i.e. prevent wraps on it for long URIs). 1358 // Just "export {...};" or "export class ..." should not be treated as 1359 // an import in this sense. 1360 if (Line.First->is(tok::kw_export) && 1361 CurrentToken->is(Keywords.kw_from) && CurrentToken->Next && 1362 CurrentToken->Next->isStringLiteral()) 1363 ImportStatement = true; 1364 if (isClosureImportStatement(*CurrentToken)) 1365 ImportStatement = true; 1366 } 1367 if (!consumeToken()) 1368 return LT_Invalid; 1369 } 1370 if (KeywordVirtualFound) 1371 return LT_VirtualFunctionDecl; 1372 if (ImportStatement) 1373 return LT_ImportStatement; 1374 1375 if (Line.startsWith(TT_ObjCMethodSpecifier)) { 1376 if (Contexts.back().FirstObjCSelectorName) 1377 Contexts.back().FirstObjCSelectorName->LongestObjCSelectorName = 1378 Contexts.back().LongestObjCSelectorName; 1379 return LT_ObjCMethodDecl; 1380 } 1381 1382 for (const auto &ctx : Contexts) 1383 if (ctx.InStructArrayInitializer) 1384 return LT_ArrayOfStructInitializer; 1385 1386 return LT_Other; 1387 } 1388 1389 private: 1390 bool isClosureImportStatement(const FormatToken &Tok) { 1391 // FIXME: Closure-library specific stuff should not be hard-coded but be 1392 // configurable. 1393 return Tok.TokenText == "goog" && Tok.Next && Tok.Next->is(tok::period) && 1394 Tok.Next->Next && 1395 (Tok.Next->Next->TokenText == "module" || 1396 Tok.Next->Next->TokenText == "provide" || 1397 Tok.Next->Next->TokenText == "require" || 1398 Tok.Next->Next->TokenText == "requireType" || 1399 Tok.Next->Next->TokenText == "forwardDeclare") && 1400 Tok.Next->Next->Next && Tok.Next->Next->Next->is(tok::l_paren); 1401 } 1402 1403 void resetTokenMetadata() { 1404 if (!CurrentToken) 1405 return; 1406 1407 // Reset token type in case we have already looked at it and then 1408 // recovered from an error (e.g. failure to find the matching >). 1409 if (!CurrentToken->isOneOf( 1410 TT_LambdaLSquare, TT_LambdaLBrace, TT_AttributeMacro, TT_IfMacro, 1411 TT_ForEachMacro, TT_TypenameMacro, TT_FunctionLBrace, 1412 TT_ImplicitStringLiteral, TT_InlineASMBrace, TT_FatArrow, 1413 TT_LambdaArrow, TT_NamespaceMacro, TT_OverloadedOperator, 1414 TT_RegexLiteral, TT_TemplateString, TT_ObjCStringLiteral, 1415 TT_UntouchableMacroFunc, TT_ConstraintJunctions, 1416 TT_StatementAttributeLikeMacro, TT_FunctionLikeOrFreestandingMacro, 1417 TT_RecordLBrace)) 1418 CurrentToken->setType(TT_Unknown); 1419 CurrentToken->Role.reset(); 1420 CurrentToken->MatchingParen = nullptr; 1421 CurrentToken->FakeLParens.clear(); 1422 CurrentToken->FakeRParens = 0; 1423 } 1424 1425 void next() { 1426 if (!CurrentToken) 1427 return; 1428 1429 CurrentToken->NestingLevel = Contexts.size() - 1; 1430 CurrentToken->BindingStrength = Contexts.back().BindingStrength; 1431 modifyContext(*CurrentToken); 1432 determineTokenType(*CurrentToken); 1433 CurrentToken = CurrentToken->Next; 1434 1435 resetTokenMetadata(); 1436 } 1437 1438 /// A struct to hold information valid in a specific context, e.g. 1439 /// a pair of parenthesis. 1440 struct Context { 1441 Context(tok::TokenKind ContextKind, unsigned BindingStrength, 1442 bool IsExpression) 1443 : ContextKind(ContextKind), BindingStrength(BindingStrength), 1444 IsExpression(IsExpression) {} 1445 1446 tok::TokenKind ContextKind; 1447 unsigned BindingStrength; 1448 bool IsExpression; 1449 unsigned LongestObjCSelectorName = 0; 1450 bool ColonIsForRangeExpr = false; 1451 bool ColonIsDictLiteral = false; 1452 bool ColonIsObjCMethodExpr = false; 1453 FormatToken *FirstObjCSelectorName = nullptr; 1454 FormatToken *FirstStartOfName = nullptr; 1455 bool CanBeExpression = true; 1456 bool InTemplateArgument = false; 1457 bool InCtorInitializer = false; 1458 bool InInheritanceList = false; 1459 bool CaretFound = false; 1460 bool IsForEachMacro = false; 1461 bool InCpp11AttributeSpecifier = false; 1462 bool InCSharpAttributeSpecifier = false; 1463 bool InStructArrayInitializer = false; 1464 }; 1465 1466 /// Puts a new \c Context onto the stack \c Contexts for the lifetime 1467 /// of each instance. 1468 struct ScopedContextCreator { 1469 AnnotatingParser &P; 1470 1471 ScopedContextCreator(AnnotatingParser &P, tok::TokenKind ContextKind, 1472 unsigned Increase) 1473 : P(P) { 1474 P.Contexts.push_back(Context(ContextKind, 1475 P.Contexts.back().BindingStrength + Increase, 1476 P.Contexts.back().IsExpression)); 1477 } 1478 1479 ~ScopedContextCreator() { 1480 if (P.Style.AlignArrayOfStructures != FormatStyle::AIAS_None) { 1481 if (P.Contexts.back().InStructArrayInitializer) { 1482 P.Contexts.pop_back(); 1483 P.Contexts.back().InStructArrayInitializer = true; 1484 return; 1485 } 1486 } 1487 P.Contexts.pop_back(); 1488 } 1489 }; 1490 1491 void modifyContext(const FormatToken &Current) { 1492 if (Current.getPrecedence() == prec::Assignment && 1493 !Line.First->isOneOf(tok::kw_template, tok::kw_using, tok::kw_return) && 1494 // Type aliases use `type X = ...;` in TypeScript and can be exported 1495 // using `export type ...`. 1496 !(Style.isJavaScript() && 1497 (Line.startsWith(Keywords.kw_type, tok::identifier) || 1498 Line.startsWith(tok::kw_export, Keywords.kw_type, 1499 tok::identifier))) && 1500 (!Current.Previous || Current.Previous->isNot(tok::kw_operator))) { 1501 Contexts.back().IsExpression = true; 1502 if (!Line.startsWith(TT_UnaryOperator)) { 1503 for (FormatToken *Previous = Current.Previous; 1504 Previous && Previous->Previous && 1505 !Previous->Previous->isOneOf(tok::comma, tok::semi); 1506 Previous = Previous->Previous) { 1507 if (Previous->isOneOf(tok::r_square, tok::r_paren)) { 1508 Previous = Previous->MatchingParen; 1509 if (!Previous) 1510 break; 1511 } 1512 if (Previous->opensScope()) 1513 break; 1514 if (Previous->isOneOf(TT_BinaryOperator, TT_UnaryOperator) && 1515 Previous->isOneOf(tok::star, tok::amp, tok::ampamp) && 1516 Previous->Previous && Previous->Previous->isNot(tok::equal)) 1517 Previous->setType(TT_PointerOrReference); 1518 } 1519 } 1520 } else if (Current.is(tok::lessless) && 1521 (!Current.Previous || !Current.Previous->is(tok::kw_operator))) { 1522 Contexts.back().IsExpression = true; 1523 } else if (Current.isOneOf(tok::kw_return, tok::kw_throw)) { 1524 Contexts.back().IsExpression = true; 1525 } else if (Current.is(TT_TrailingReturnArrow)) { 1526 Contexts.back().IsExpression = false; 1527 } else if (Current.is(TT_LambdaArrow) || Current.is(Keywords.kw_assert)) { 1528 Contexts.back().IsExpression = Style.Language == FormatStyle::LK_Java; 1529 } else if (Current.Previous && 1530 Current.Previous->is(TT_CtorInitializerColon)) { 1531 Contexts.back().IsExpression = true; 1532 Contexts.back().InCtorInitializer = true; 1533 } else if (Current.Previous && Current.Previous->is(TT_InheritanceColon)) { 1534 Contexts.back().InInheritanceList = true; 1535 } else if (Current.isOneOf(tok::r_paren, tok::greater, tok::comma)) { 1536 for (FormatToken *Previous = Current.Previous; 1537 Previous && Previous->isOneOf(tok::star, tok::amp); 1538 Previous = Previous->Previous) 1539 Previous->setType(TT_PointerOrReference); 1540 if (Line.MustBeDeclaration && !Contexts.front().InCtorInitializer) 1541 Contexts.back().IsExpression = false; 1542 } else if (Current.is(tok::kw_new)) { 1543 Contexts.back().CanBeExpression = false; 1544 } else if (Current.is(tok::semi) || 1545 (Current.is(tok::exclaim) && Current.Previous && 1546 !Current.Previous->is(tok::kw_operator))) { 1547 // This should be the condition or increment in a for-loop. 1548 // But not operator !() (can't use TT_OverloadedOperator here as its not 1549 // been annotated yet). 1550 Contexts.back().IsExpression = true; 1551 } 1552 } 1553 1554 static FormatToken *untilMatchingParen(FormatToken *Current) { 1555 // Used when `MatchingParen` is not yet established. 1556 int ParenLevel = 0; 1557 while (Current) { 1558 if (Current->is(tok::l_paren)) 1559 ++ParenLevel; 1560 if (Current->is(tok::r_paren)) 1561 --ParenLevel; 1562 if (ParenLevel < 1) 1563 break; 1564 Current = Current->Next; 1565 } 1566 return Current; 1567 } 1568 1569 static bool isDeductionGuide(FormatToken &Current) { 1570 // Look for a deduction guide template<T> A(...) -> A<...>; 1571 if (Current.Previous && Current.Previous->is(tok::r_paren) && 1572 Current.startsSequence(tok::arrow, tok::identifier, tok::less)) { 1573 // Find the TemplateCloser. 1574 FormatToken *TemplateCloser = Current.Next->Next; 1575 int NestingLevel = 0; 1576 while (TemplateCloser) { 1577 // Skip over an expressions in parens A<(3 < 2)>; 1578 if (TemplateCloser->is(tok::l_paren)) { 1579 // No Matching Paren yet so skip to matching paren 1580 TemplateCloser = untilMatchingParen(TemplateCloser); 1581 if (!TemplateCloser) 1582 break; 1583 } 1584 if (TemplateCloser->is(tok::less)) 1585 ++NestingLevel; 1586 if (TemplateCloser->is(tok::greater)) 1587 --NestingLevel; 1588 if (NestingLevel < 1) 1589 break; 1590 TemplateCloser = TemplateCloser->Next; 1591 } 1592 // Assuming we have found the end of the template ensure its followed 1593 // with a semi-colon. 1594 if (TemplateCloser && TemplateCloser->Next && 1595 TemplateCloser->Next->is(tok::semi) && 1596 Current.Previous->MatchingParen) { 1597 // Determine if the identifier `A` prior to the A<..>; is the same as 1598 // prior to the A(..) 1599 FormatToken *LeadingIdentifier = 1600 Current.Previous->MatchingParen->Previous; 1601 1602 // Differentiate a deduction guide by seeing the 1603 // > of the template prior to the leading identifier. 1604 if (LeadingIdentifier) { 1605 FormatToken *PriorLeadingIdentifier = LeadingIdentifier->Previous; 1606 // Skip back past explicit decoration 1607 if (PriorLeadingIdentifier && 1608 PriorLeadingIdentifier->is(tok::kw_explicit)) 1609 PriorLeadingIdentifier = PriorLeadingIdentifier->Previous; 1610 1611 return (PriorLeadingIdentifier && 1612 PriorLeadingIdentifier->is(TT_TemplateCloser) && 1613 LeadingIdentifier->TokenText == Current.Next->TokenText); 1614 } 1615 } 1616 } 1617 return false; 1618 } 1619 1620 void determineTokenType(FormatToken &Current) { 1621 if (!Current.is(TT_Unknown)) 1622 // The token type is already known. 1623 return; 1624 1625 if ((Style.isJavaScript() || Style.isCSharp()) && 1626 Current.is(tok::exclaim)) { 1627 if (Current.Previous) { 1628 bool IsIdentifier = 1629 Style.isJavaScript() 1630 ? Keywords.IsJavaScriptIdentifier( 1631 *Current.Previous, /* AcceptIdentifierName= */ true) 1632 : Current.Previous->is(tok::identifier); 1633 if (IsIdentifier || 1634 Current.Previous->isOneOf( 1635 tok::kw_namespace, tok::r_paren, tok::r_square, tok::r_brace, 1636 tok::kw_false, tok::kw_true, Keywords.kw_type, Keywords.kw_get, 1637 Keywords.kw_set) || 1638 Current.Previous->Tok.isLiteral()) { 1639 Current.setType(TT_NonNullAssertion); 1640 return; 1641 } 1642 } 1643 if (Current.Next && 1644 Current.Next->isOneOf(TT_BinaryOperator, Keywords.kw_as)) { 1645 Current.setType(TT_NonNullAssertion); 1646 return; 1647 } 1648 } 1649 1650 // Line.MightBeFunctionDecl can only be true after the parentheses of a 1651 // function declaration have been found. In this case, 'Current' is a 1652 // trailing token of this declaration and thus cannot be a name. 1653 if (Current.is(Keywords.kw_instanceof)) { 1654 Current.setType(TT_BinaryOperator); 1655 } else if (isStartOfName(Current) && 1656 (!Line.MightBeFunctionDecl || Current.NestingLevel != 0)) { 1657 Contexts.back().FirstStartOfName = &Current; 1658 Current.setType(TT_StartOfName); 1659 } else if (Current.is(tok::semi)) { 1660 // Reset FirstStartOfName after finding a semicolon so that a for loop 1661 // with multiple increment statements is not confused with a for loop 1662 // having multiple variable declarations. 1663 Contexts.back().FirstStartOfName = nullptr; 1664 } else if (Current.isOneOf(tok::kw_auto, tok::kw___auto_type)) { 1665 AutoFound = true; 1666 } else if (Current.is(tok::arrow) && 1667 Style.Language == FormatStyle::LK_Java) { 1668 Current.setType(TT_LambdaArrow); 1669 } else if (Current.is(tok::arrow) && AutoFound && Line.MustBeDeclaration && 1670 Current.NestingLevel == 0 && 1671 !Current.Previous->isOneOf(tok::kw_operator, tok::identifier)) { 1672 // not auto operator->() -> xxx; 1673 Current.setType(TT_TrailingReturnArrow); 1674 } else if (Current.is(tok::arrow) && Current.Previous && 1675 Current.Previous->is(tok::r_brace)) { 1676 // Concept implicit conversion constraint needs to be treated like 1677 // a trailing return type ... } -> <type>. 1678 Current.setType(TT_TrailingReturnArrow); 1679 } else if (isDeductionGuide(Current)) { 1680 // Deduction guides trailing arrow " A(...) -> A<T>;". 1681 Current.setType(TT_TrailingReturnArrow); 1682 } else if (Current.isOneOf(tok::star, tok::amp, tok::ampamp)) { 1683 Current.setType(determineStarAmpUsage( 1684 Current, 1685 Contexts.back().CanBeExpression && Contexts.back().IsExpression, 1686 Contexts.back().InTemplateArgument)); 1687 } else if (Current.isOneOf(tok::minus, tok::plus, tok::caret)) { 1688 Current.setType(determinePlusMinusCaretUsage(Current)); 1689 if (Current.is(TT_UnaryOperator) && Current.is(tok::caret)) 1690 Contexts.back().CaretFound = true; 1691 } else if (Current.isOneOf(tok::minusminus, tok::plusplus)) { 1692 Current.setType(determineIncrementUsage(Current)); 1693 } else if (Current.isOneOf(tok::exclaim, tok::tilde)) { 1694 Current.setType(TT_UnaryOperator); 1695 } else if (Current.is(tok::question)) { 1696 if (Style.isJavaScript() && Line.MustBeDeclaration && 1697 !Contexts.back().IsExpression) { 1698 // In JavaScript, `interface X { foo?(): bar; }` is an optional method 1699 // on the interface, not a ternary expression. 1700 Current.setType(TT_JsTypeOptionalQuestion); 1701 } else { 1702 Current.setType(TT_ConditionalExpr); 1703 } 1704 } else if (Current.isBinaryOperator() && 1705 (!Current.Previous || Current.Previous->isNot(tok::l_square)) && 1706 (!Current.is(tok::greater) && 1707 Style.Language != FormatStyle::LK_TextProto)) { 1708 Current.setType(TT_BinaryOperator); 1709 } else if (Current.is(tok::comment)) { 1710 if (Current.TokenText.startswith("/*")) 1711 if (Current.TokenText.endswith("*/")) 1712 Current.setType(TT_BlockComment); 1713 else 1714 // The lexer has for some reason determined a comment here. But we 1715 // cannot really handle it, if it isn't properly terminated. 1716 Current.Tok.setKind(tok::unknown); 1717 else 1718 Current.setType(TT_LineComment); 1719 } else if (Current.is(tok::r_paren)) { 1720 if (rParenEndsCast(Current)) 1721 Current.setType(TT_CastRParen); 1722 if (Current.MatchingParen && Current.Next && 1723 !Current.Next->isBinaryOperator() && 1724 !Current.Next->isOneOf(tok::semi, tok::colon, tok::l_brace, 1725 tok::comma, tok::period, tok::arrow, 1726 tok::coloncolon)) 1727 if (FormatToken *AfterParen = Current.MatchingParen->Next) { 1728 // Make sure this isn't the return type of an Obj-C block declaration 1729 if (AfterParen->Tok.isNot(tok::caret)) { 1730 if (FormatToken *BeforeParen = Current.MatchingParen->Previous) 1731 if (BeforeParen->is(tok::identifier) && 1732 !BeforeParen->is(TT_TypenameMacro) && 1733 BeforeParen->TokenText == BeforeParen->TokenText.upper() && 1734 (!BeforeParen->Previous || 1735 BeforeParen->Previous->ClosesTemplateDeclaration)) 1736 Current.setType(TT_FunctionAnnotationRParen); 1737 } 1738 } 1739 } else if (Current.is(tok::at) && Current.Next && !Style.isJavaScript() && 1740 Style.Language != FormatStyle::LK_Java) { 1741 // In Java & JavaScript, "@..." is a decorator or annotation. In ObjC, it 1742 // marks declarations and properties that need special formatting. 1743 switch (Current.Next->Tok.getObjCKeywordID()) { 1744 case tok::objc_interface: 1745 case tok::objc_implementation: 1746 case tok::objc_protocol: 1747 Current.setType(TT_ObjCDecl); 1748 break; 1749 case tok::objc_property: 1750 Current.setType(TT_ObjCProperty); 1751 break; 1752 default: 1753 break; 1754 } 1755 } else if (Current.is(tok::period)) { 1756 FormatToken *PreviousNoComment = Current.getPreviousNonComment(); 1757 if (PreviousNoComment && 1758 PreviousNoComment->isOneOf(tok::comma, tok::l_brace)) 1759 Current.setType(TT_DesignatedInitializerPeriod); 1760 else if (Style.Language == FormatStyle::LK_Java && Current.Previous && 1761 Current.Previous->isOneOf(TT_JavaAnnotation, 1762 TT_LeadingJavaAnnotation)) 1763 Current.setType(Current.Previous->getType()); 1764 } else if (canBeObjCSelectorComponent(Current) && 1765 // FIXME(bug 36976): ObjC return types shouldn't use 1766 // TT_CastRParen. 1767 Current.Previous && Current.Previous->is(TT_CastRParen) && 1768 Current.Previous->MatchingParen && 1769 Current.Previous->MatchingParen->Previous && 1770 Current.Previous->MatchingParen->Previous->is( 1771 TT_ObjCMethodSpecifier)) { 1772 // This is the first part of an Objective-C selector name. (If there's no 1773 // colon after this, this is the only place which annotates the identifier 1774 // as a selector.) 1775 Current.setType(TT_SelectorName); 1776 } else if (Current.isOneOf(tok::identifier, tok::kw_const, tok::kw_noexcept, 1777 tok::kw_requires) && 1778 Current.Previous && 1779 !Current.Previous->isOneOf(tok::equal, tok::at) && 1780 Line.MightBeFunctionDecl && Contexts.size() == 1) { 1781 // Line.MightBeFunctionDecl can only be true after the parentheses of a 1782 // function declaration have been found. 1783 Current.setType(TT_TrailingAnnotation); 1784 } else if ((Style.Language == FormatStyle::LK_Java || 1785 Style.isJavaScript()) && 1786 Current.Previous) { 1787 if (Current.Previous->is(tok::at) && 1788 Current.isNot(Keywords.kw_interface)) { 1789 const FormatToken &AtToken = *Current.Previous; 1790 const FormatToken *Previous = AtToken.getPreviousNonComment(); 1791 if (!Previous || Previous->is(TT_LeadingJavaAnnotation)) 1792 Current.setType(TT_LeadingJavaAnnotation); 1793 else 1794 Current.setType(TT_JavaAnnotation); 1795 } else if (Current.Previous->is(tok::period) && 1796 Current.Previous->isOneOf(TT_JavaAnnotation, 1797 TT_LeadingJavaAnnotation)) { 1798 Current.setType(Current.Previous->getType()); 1799 } 1800 } 1801 } 1802 1803 /// Take a guess at whether \p Tok starts a name of a function or 1804 /// variable declaration. 1805 /// 1806 /// This is a heuristic based on whether \p Tok is an identifier following 1807 /// something that is likely a type. 1808 bool isStartOfName(const FormatToken &Tok) { 1809 if (Tok.isNot(tok::identifier) || !Tok.Previous) 1810 return false; 1811 1812 if (Tok.Previous->isOneOf(TT_LeadingJavaAnnotation, Keywords.kw_instanceof, 1813 Keywords.kw_as)) 1814 return false; 1815 if (Style.isJavaScript() && Tok.Previous->is(Keywords.kw_in)) 1816 return false; 1817 1818 // Skip "const" as it does not have an influence on whether this is a name. 1819 FormatToken *PreviousNotConst = Tok.getPreviousNonComment(); 1820 1821 // For javascript const can be like "let" or "var" 1822 if (!Style.isJavaScript()) 1823 while (PreviousNotConst && PreviousNotConst->is(tok::kw_const)) 1824 PreviousNotConst = PreviousNotConst->getPreviousNonComment(); 1825 1826 if (!PreviousNotConst) 1827 return false; 1828 1829 bool IsPPKeyword = PreviousNotConst->is(tok::identifier) && 1830 PreviousNotConst->Previous && 1831 PreviousNotConst->Previous->is(tok::hash); 1832 1833 if (PreviousNotConst->is(TT_TemplateCloser)) 1834 return PreviousNotConst && PreviousNotConst->MatchingParen && 1835 PreviousNotConst->MatchingParen->Previous && 1836 PreviousNotConst->MatchingParen->Previous->isNot(tok::period) && 1837 PreviousNotConst->MatchingParen->Previous->isNot(tok::kw_template); 1838 1839 if (PreviousNotConst->is(tok::r_paren) && 1840 PreviousNotConst->is(TT_TypeDeclarationParen)) 1841 return true; 1842 1843 // If is a preprocess keyword like #define. 1844 if (IsPPKeyword) 1845 return false; 1846 1847 // int a or auto a. 1848 if (PreviousNotConst->isOneOf(tok::identifier, tok::kw_auto)) 1849 return true; 1850 1851 // *a or &a or &&a. 1852 if (PreviousNotConst->is(TT_PointerOrReference)) 1853 return true; 1854 1855 // MyClass a; 1856 if (PreviousNotConst->isSimpleTypeSpecifier()) 1857 return true; 1858 1859 // const a = in JavaScript. 1860 return (Style.isJavaScript() && PreviousNotConst->is(tok::kw_const)); 1861 } 1862 1863 /// Determine whether ')' is ending a cast. 1864 bool rParenEndsCast(const FormatToken &Tok) { 1865 // C-style casts are only used in C++, C# and Java. 1866 if (!Style.isCSharp() && !Style.isCpp() && 1867 Style.Language != FormatStyle::LK_Java) 1868 return false; 1869 1870 // Empty parens aren't casts and there are no casts at the end of the line. 1871 if (Tok.Previous == Tok.MatchingParen || !Tok.Next || !Tok.MatchingParen) 1872 return false; 1873 1874 FormatToken *LeftOfParens = Tok.MatchingParen->getPreviousNonComment(); 1875 if (LeftOfParens) { 1876 // If there is a closing parenthesis left of the current 1877 // parentheses, look past it as these might be chained casts. 1878 if (LeftOfParens->is(tok::r_paren) && 1879 LeftOfParens->isNot(TT_CastRParen)) { 1880 if (!LeftOfParens->MatchingParen || 1881 !LeftOfParens->MatchingParen->Previous) 1882 return false; 1883 LeftOfParens = LeftOfParens->MatchingParen->Previous; 1884 } 1885 1886 // The Condition directly below this one will see the operator arguments 1887 // as a (void *foo) cast. 1888 // void operator delete(void *foo) ATTRIB; 1889 if (LeftOfParens->Tok.getIdentifierInfo() && LeftOfParens->Previous && 1890 LeftOfParens->Previous->is(tok::kw_operator)) 1891 return false; 1892 1893 // If there is an identifier (or with a few exceptions a keyword) right 1894 // before the parentheses, this is unlikely to be a cast. 1895 if (LeftOfParens->Tok.getIdentifierInfo() && 1896 !LeftOfParens->isOneOf(Keywords.kw_in, tok::kw_return, tok::kw_case, 1897 tok::kw_delete)) 1898 return false; 1899 1900 // Certain other tokens right before the parentheses are also signals that 1901 // this cannot be a cast. 1902 if (LeftOfParens->isOneOf(tok::at, tok::r_square, TT_OverloadedOperator, 1903 TT_TemplateCloser, tok::ellipsis)) 1904 return false; 1905 } 1906 1907 if (Tok.Next->is(tok::question)) 1908 return false; 1909 1910 // `foreach((A a, B b) in someList)` should not be seen as a cast. 1911 if (Tok.Next->is(Keywords.kw_in) && Style.isCSharp()) 1912 return false; 1913 1914 // Functions which end with decorations like volatile, noexcept are unlikely 1915 // to be casts. 1916 if (Tok.Next->isOneOf(tok::kw_noexcept, tok::kw_volatile, tok::kw_const, 1917 tok::kw_requires, tok::kw_throw, tok::arrow, 1918 Keywords.kw_override, Keywords.kw_final) || 1919 isCpp11AttributeSpecifier(*Tok.Next)) 1920 return false; 1921 1922 // As Java has no function types, a "(" after the ")" likely means that this 1923 // is a cast. 1924 if (Style.Language == FormatStyle::LK_Java && Tok.Next->is(tok::l_paren)) 1925 return true; 1926 1927 // If a (non-string) literal follows, this is likely a cast. 1928 if (Tok.Next->isNot(tok::string_literal) && 1929 (Tok.Next->Tok.isLiteral() || 1930 Tok.Next->isOneOf(tok::kw_sizeof, tok::kw_alignof))) 1931 return true; 1932 1933 // Heuristically try to determine whether the parentheses contain a type. 1934 auto IsQualifiedPointerOrReference = [](FormatToken *T) { 1935 // This is used to handle cases such as x = (foo *const)&y; 1936 assert(!T->isSimpleTypeSpecifier() && "Should have already been checked"); 1937 // Strip trailing qualifiers such as const or volatile when checking 1938 // whether the parens could be a cast to a pointer/reference type. 1939 while (T) { 1940 if (T->is(TT_AttributeParen)) { 1941 // Handle `x = (foo *__attribute__((foo)))&v;`: 1942 if (T->MatchingParen && T->MatchingParen->Previous && 1943 T->MatchingParen->Previous->is(tok::kw___attribute)) { 1944 T = T->MatchingParen->Previous->Previous; 1945 continue; 1946 } 1947 } else if (T->is(TT_AttributeSquare)) { 1948 // Handle `x = (foo *[[clang::foo]])&v;`: 1949 if (T->MatchingParen && T->MatchingParen->Previous) { 1950 T = T->MatchingParen->Previous; 1951 continue; 1952 } 1953 } else if (T->canBePointerOrReferenceQualifier()) { 1954 T = T->Previous; 1955 continue; 1956 } 1957 break; 1958 } 1959 return T && T->is(TT_PointerOrReference); 1960 }; 1961 bool ParensAreType = 1962 !Tok.Previous || 1963 Tok.Previous->isOneOf(TT_TemplateCloser, TT_TypeDeclarationParen) || 1964 Tok.Previous->isSimpleTypeSpecifier() || 1965 IsQualifiedPointerOrReference(Tok.Previous); 1966 bool ParensCouldEndDecl = 1967 Tok.Next->isOneOf(tok::equal, tok::semi, tok::l_brace, tok::greater); 1968 if (ParensAreType && !ParensCouldEndDecl) 1969 return true; 1970 1971 // At this point, we heuristically assume that there are no casts at the 1972 // start of the line. We assume that we have found most cases where there 1973 // are by the logic above, e.g. "(void)x;". 1974 if (!LeftOfParens) 1975 return false; 1976 1977 // Certain token types inside the parentheses mean that this can't be a 1978 // cast. 1979 for (const FormatToken *Token = Tok.MatchingParen->Next; Token != &Tok; 1980 Token = Token->Next) 1981 if (Token->is(TT_BinaryOperator)) 1982 return false; 1983 1984 // If the following token is an identifier or 'this', this is a cast. All 1985 // cases where this can be something else are handled above. 1986 if (Tok.Next->isOneOf(tok::identifier, tok::kw_this)) 1987 return true; 1988 1989 // Look for a cast `( x ) (`. 1990 if (Tok.Next->is(tok::l_paren) && Tok.Previous && Tok.Previous->Previous) { 1991 if (Tok.Previous->is(tok::identifier) && 1992 Tok.Previous->Previous->is(tok::l_paren)) 1993 return true; 1994 } 1995 1996 if (!Tok.Next->Next) 1997 return false; 1998 1999 // If the next token after the parenthesis is a unary operator, assume 2000 // that this is cast, unless there are unexpected tokens inside the 2001 // parenthesis. 2002 bool NextIsUnary = 2003 Tok.Next->isUnaryOperator() || Tok.Next->isOneOf(tok::amp, tok::star); 2004 if (!NextIsUnary || Tok.Next->is(tok::plus) || 2005 !Tok.Next->Next->isOneOf(tok::identifier, tok::numeric_constant)) 2006 return false; 2007 // Search for unexpected tokens. 2008 for (FormatToken *Prev = Tok.Previous; Prev != Tok.MatchingParen; 2009 Prev = Prev->Previous) 2010 if (!Prev->isOneOf(tok::kw_const, tok::identifier, tok::coloncolon)) 2011 return false; 2012 return true; 2013 } 2014 2015 /// Return the type of the given token assuming it is * or &. 2016 TokenType determineStarAmpUsage(const FormatToken &Tok, bool IsExpression, 2017 bool InTemplateArgument) { 2018 if (Style.isJavaScript()) 2019 return TT_BinaryOperator; 2020 2021 // && in C# must be a binary operator. 2022 if (Style.isCSharp() && Tok.is(tok::ampamp)) 2023 return TT_BinaryOperator; 2024 2025 const FormatToken *PrevToken = Tok.getPreviousNonComment(); 2026 if (!PrevToken) 2027 return TT_UnaryOperator; 2028 2029 const FormatToken *NextToken = Tok.getNextNonComment(); 2030 if (!NextToken || 2031 NextToken->isOneOf(tok::arrow, tok::equal, tok::kw_noexcept) || 2032 NextToken->canBePointerOrReferenceQualifier() || 2033 (NextToken->is(tok::l_brace) && !NextToken->getNextNonComment())) 2034 return TT_PointerOrReference; 2035 2036 if (PrevToken->is(tok::coloncolon)) 2037 return TT_PointerOrReference; 2038 2039 if (PrevToken->is(tok::r_paren) && PrevToken->is(TT_TypeDeclarationParen)) 2040 return TT_PointerOrReference; 2041 2042 if (PrevToken->isOneOf(tok::l_paren, tok::l_square, tok::l_brace, 2043 tok::comma, tok::semi, tok::kw_return, tok::colon, 2044 tok::kw_co_return, tok::kw_co_await, 2045 tok::kw_co_yield, tok::equal, tok::kw_delete, 2046 tok::kw_sizeof, tok::kw_throw, TT_BinaryOperator, 2047 TT_ConditionalExpr, TT_UnaryOperator, TT_CastRParen)) 2048 return TT_UnaryOperator; 2049 2050 if (NextToken->is(tok::l_square) && NextToken->isNot(TT_LambdaLSquare)) 2051 return TT_PointerOrReference; 2052 if (NextToken->is(tok::kw_operator) && !IsExpression) 2053 return TT_PointerOrReference; 2054 if (NextToken->isOneOf(tok::comma, tok::semi)) 2055 return TT_PointerOrReference; 2056 2057 if (PrevToken->Tok.isLiteral() || 2058 PrevToken->isOneOf(tok::r_paren, tok::r_square, tok::kw_true, 2059 tok::kw_false, tok::r_brace) || 2060 NextToken->Tok.isLiteral() || 2061 NextToken->isOneOf(tok::kw_true, tok::kw_false) || 2062 NextToken->isUnaryOperator() || 2063 // If we know we're in a template argument, there are no named 2064 // declarations. Thus, having an identifier on the right-hand side 2065 // indicates a binary operator. 2066 (InTemplateArgument && NextToken->Tok.isAnyIdentifier())) 2067 return TT_BinaryOperator; 2068 2069 // "&&(" is quite unlikely to be two successive unary "&". 2070 if (Tok.is(tok::ampamp) && NextToken->is(tok::l_paren)) 2071 return TT_BinaryOperator; 2072 2073 // This catches some cases where evaluation order is used as control flow: 2074 // aaa && aaa->f(); 2075 if (NextToken->Tok.isAnyIdentifier()) { 2076 const FormatToken *NextNextToken = NextToken->getNextNonComment(); 2077 if (NextNextToken && NextNextToken->is(tok::arrow)) 2078 return TT_BinaryOperator; 2079 } 2080 2081 // It is very unlikely that we are going to find a pointer or reference type 2082 // definition on the RHS of an assignment. 2083 if (IsExpression && !Contexts.back().CaretFound) 2084 return TT_BinaryOperator; 2085 2086 return TT_PointerOrReference; 2087 } 2088 2089 TokenType determinePlusMinusCaretUsage(const FormatToken &Tok) { 2090 const FormatToken *PrevToken = Tok.getPreviousNonComment(); 2091 if (!PrevToken) 2092 return TT_UnaryOperator; 2093 2094 if (PrevToken->isOneOf(TT_CastRParen, TT_UnaryOperator)) 2095 // This must be a sequence of leading unary operators. 2096 return TT_UnaryOperator; 2097 2098 // Use heuristics to recognize unary operators. 2099 if (PrevToken->isOneOf(tok::equal, tok::l_paren, tok::comma, tok::l_square, 2100 tok::question, tok::colon, tok::kw_return, 2101 tok::kw_case, tok::at, tok::l_brace, tok::kw_throw, 2102 tok::kw_co_return, tok::kw_co_yield)) 2103 return TT_UnaryOperator; 2104 2105 // There can't be two consecutive binary operators. 2106 if (PrevToken->is(TT_BinaryOperator)) 2107 return TT_UnaryOperator; 2108 2109 // Fall back to marking the token as binary operator. 2110 return TT_BinaryOperator; 2111 } 2112 2113 /// Determine whether ++/-- are pre- or post-increments/-decrements. 2114 TokenType determineIncrementUsage(const FormatToken &Tok) { 2115 const FormatToken *PrevToken = Tok.getPreviousNonComment(); 2116 if (!PrevToken || PrevToken->is(TT_CastRParen)) 2117 return TT_UnaryOperator; 2118 if (PrevToken->isOneOf(tok::r_paren, tok::r_square, tok::identifier)) 2119 return TT_TrailingUnaryOperator; 2120 2121 return TT_UnaryOperator; 2122 } 2123 2124 SmallVector<Context, 8> Contexts; 2125 2126 const FormatStyle &Style; 2127 AnnotatedLine &Line; 2128 FormatToken *CurrentToken; 2129 bool AutoFound; 2130 const AdditionalKeywords &Keywords; 2131 2132 // Set of "<" tokens that do not open a template parameter list. If parseAngle 2133 // determines that a specific token can't be a template opener, it will make 2134 // same decision irrespective of the decisions for tokens leading up to it. 2135 // Store this information to prevent this from causing exponential runtime. 2136 llvm::SmallPtrSet<FormatToken *, 16> NonTemplateLess; 2137 }; 2138 2139 static const int PrecedenceUnaryOperator = prec::PointerToMember + 1; 2140 static const int PrecedenceArrowAndPeriod = prec::PointerToMember + 2; 2141 2142 /// Parses binary expressions by inserting fake parenthesis based on 2143 /// operator precedence. 2144 class ExpressionParser { 2145 public: 2146 ExpressionParser(const FormatStyle &Style, const AdditionalKeywords &Keywords, 2147 AnnotatedLine &Line) 2148 : Style(Style), Keywords(Keywords), Current(Line.First) {} 2149 2150 /// Parse expressions with the given operator precedence. 2151 void parse(int Precedence = 0) { 2152 // Skip 'return' and ObjC selector colons as they are not part of a binary 2153 // expression. 2154 while (Current && (Current->is(tok::kw_return) || 2155 (Current->is(tok::colon) && 2156 Current->isOneOf(TT_ObjCMethodExpr, TT_DictLiteral)))) 2157 next(); 2158 2159 if (!Current || Precedence > PrecedenceArrowAndPeriod) 2160 return; 2161 2162 // Conditional expressions need to be parsed separately for proper nesting. 2163 if (Precedence == prec::Conditional) { 2164 parseConditionalExpr(); 2165 return; 2166 } 2167 2168 // Parse unary operators, which all have a higher precedence than binary 2169 // operators. 2170 if (Precedence == PrecedenceUnaryOperator) { 2171 parseUnaryOperator(); 2172 return; 2173 } 2174 2175 FormatToken *Start = Current; 2176 FormatToken *LatestOperator = nullptr; 2177 unsigned OperatorIndex = 0; 2178 2179 while (Current) { 2180 // Consume operators with higher precedence. 2181 parse(Precedence + 1); 2182 2183 int CurrentPrecedence = getCurrentPrecedence(); 2184 2185 if (Precedence == CurrentPrecedence && Current && 2186 Current->is(TT_SelectorName)) { 2187 if (LatestOperator) 2188 addFakeParenthesis(Start, prec::Level(Precedence)); 2189 Start = Current; 2190 } 2191 2192 // At the end of the line or when an operator with higher precedence is 2193 // found, insert fake parenthesis and return. 2194 if (!Current || 2195 (Current->closesScope() && 2196 (Current->MatchingParen || Current->is(TT_TemplateString))) || 2197 (CurrentPrecedence != -1 && CurrentPrecedence < Precedence) || 2198 (CurrentPrecedence == prec::Conditional && 2199 Precedence == prec::Assignment && Current->is(tok::colon))) 2200 break; 2201 2202 // Consume scopes: (), [], <> and {} 2203 if (Current->opensScope()) { 2204 // In fragment of a JavaScript template string can look like '}..${' and 2205 // thus close a scope and open a new one at the same time. 2206 while (Current && (!Current->closesScope() || Current->opensScope())) { 2207 next(); 2208 parse(); 2209 } 2210 next(); 2211 } else { 2212 // Operator found. 2213 if (CurrentPrecedence == Precedence) { 2214 if (LatestOperator) 2215 LatestOperator->NextOperator = Current; 2216 LatestOperator = Current; 2217 Current->OperatorIndex = OperatorIndex; 2218 ++OperatorIndex; 2219 } 2220 next(/*SkipPastLeadingComments=*/Precedence > 0); 2221 } 2222 } 2223 2224 if (LatestOperator && (Current || Precedence > 0)) { 2225 // LatestOperator->LastOperator = true; 2226 if (Precedence == PrecedenceArrowAndPeriod) { 2227 // Call expressions don't have a binary operator precedence. 2228 addFakeParenthesis(Start, prec::Unknown); 2229 } else { 2230 addFakeParenthesis(Start, prec::Level(Precedence)); 2231 } 2232 } 2233 } 2234 2235 private: 2236 /// Gets the precedence (+1) of the given token for binary operators 2237 /// and other tokens that we treat like binary operators. 2238 int getCurrentPrecedence() { 2239 if (Current) { 2240 const FormatToken *NextNonComment = Current->getNextNonComment(); 2241 if (Current->is(TT_ConditionalExpr)) 2242 return prec::Conditional; 2243 if (NextNonComment && Current->is(TT_SelectorName) && 2244 (NextNonComment->isOneOf(TT_DictLiteral, TT_JsTypeColon) || 2245 ((Style.Language == FormatStyle::LK_Proto || 2246 Style.Language == FormatStyle::LK_TextProto) && 2247 NextNonComment->is(tok::less)))) 2248 return prec::Assignment; 2249 if (Current->is(TT_JsComputedPropertyName)) 2250 return prec::Assignment; 2251 if (Current->is(TT_LambdaArrow)) 2252 return prec::Comma; 2253 if (Current->is(TT_FatArrow)) 2254 return prec::Assignment; 2255 if (Current->isOneOf(tok::semi, TT_InlineASMColon, TT_SelectorName) || 2256 (Current->is(tok::comment) && NextNonComment && 2257 NextNonComment->is(TT_SelectorName))) 2258 return 0; 2259 if (Current->is(TT_RangeBasedForLoopColon)) 2260 return prec::Comma; 2261 if ((Style.Language == FormatStyle::LK_Java || Style.isJavaScript()) && 2262 Current->is(Keywords.kw_instanceof)) 2263 return prec::Relational; 2264 if (Style.isJavaScript() && 2265 Current->isOneOf(Keywords.kw_in, Keywords.kw_as)) 2266 return prec::Relational; 2267 if (Current->is(TT_BinaryOperator) || Current->is(tok::comma)) 2268 return Current->getPrecedence(); 2269 if (Current->isOneOf(tok::period, tok::arrow)) 2270 return PrecedenceArrowAndPeriod; 2271 if ((Style.Language == FormatStyle::LK_Java || Style.isJavaScript()) && 2272 Current->isOneOf(Keywords.kw_extends, Keywords.kw_implements, 2273 Keywords.kw_throws)) 2274 return 0; 2275 } 2276 return -1; 2277 } 2278 2279 void addFakeParenthesis(FormatToken *Start, prec::Level Precedence) { 2280 Start->FakeLParens.push_back(Precedence); 2281 if (Precedence > prec::Unknown) 2282 Start->StartsBinaryExpression = true; 2283 if (Current) { 2284 FormatToken *Previous = Current->Previous; 2285 while (Previous->is(tok::comment) && Previous->Previous) 2286 Previous = Previous->Previous; 2287 ++Previous->FakeRParens; 2288 if (Precedence > prec::Unknown) 2289 Previous->EndsBinaryExpression = true; 2290 } 2291 } 2292 2293 /// Parse unary operator expressions and surround them with fake 2294 /// parentheses if appropriate. 2295 void parseUnaryOperator() { 2296 llvm::SmallVector<FormatToken *, 2> Tokens; 2297 while (Current && Current->is(TT_UnaryOperator)) { 2298 Tokens.push_back(Current); 2299 next(); 2300 } 2301 parse(PrecedenceArrowAndPeriod); 2302 for (FormatToken *Token : llvm::reverse(Tokens)) 2303 // The actual precedence doesn't matter. 2304 addFakeParenthesis(Token, prec::Unknown); 2305 } 2306 2307 void parseConditionalExpr() { 2308 while (Current && Current->isTrailingComment()) 2309 next(); 2310 FormatToken *Start = Current; 2311 parse(prec::LogicalOr); 2312 if (!Current || !Current->is(tok::question)) 2313 return; 2314 next(); 2315 parse(prec::Assignment); 2316 if (!Current || Current->isNot(TT_ConditionalExpr)) 2317 return; 2318 next(); 2319 parse(prec::Assignment); 2320 addFakeParenthesis(Start, prec::Conditional); 2321 } 2322 2323 void next(bool SkipPastLeadingComments = true) { 2324 if (Current) 2325 Current = Current->Next; 2326 while (Current && 2327 (Current->NewlinesBefore == 0 || SkipPastLeadingComments) && 2328 Current->isTrailingComment()) 2329 Current = Current->Next; 2330 } 2331 2332 const FormatStyle &Style; 2333 const AdditionalKeywords &Keywords; 2334 FormatToken *Current; 2335 }; 2336 2337 } // end anonymous namespace 2338 2339 void TokenAnnotator::setCommentLineLevels( 2340 SmallVectorImpl<AnnotatedLine *> &Lines) { 2341 const AnnotatedLine *NextNonCommentLine = nullptr; 2342 for (AnnotatedLine *Line : llvm::reverse(Lines)) { 2343 assert(Line->First); 2344 bool CommentLine = true; 2345 for (const FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) { 2346 if (!Tok->is(tok::comment)) { 2347 CommentLine = false; 2348 break; 2349 } 2350 } 2351 2352 // If the comment is currently aligned with the line immediately following 2353 // it, that's probably intentional and we should keep it. 2354 if (NextNonCommentLine && CommentLine && 2355 NextNonCommentLine->First->NewlinesBefore <= 1 && 2356 NextNonCommentLine->First->OriginalColumn == 2357 Line->First->OriginalColumn) { 2358 // Align comments for preprocessor lines with the # in column 0 if 2359 // preprocessor lines are not indented. Otherwise, align with the next 2360 // line. 2361 Line->Level = 2362 (Style.IndentPPDirectives != FormatStyle::PPDIS_BeforeHash && 2363 (NextNonCommentLine->Type == LT_PreprocessorDirective || 2364 NextNonCommentLine->Type == LT_ImportStatement)) 2365 ? 0 2366 : NextNonCommentLine->Level; 2367 } else { 2368 NextNonCommentLine = Line->First->isNot(tok::r_brace) ? Line : nullptr; 2369 } 2370 2371 setCommentLineLevels(Line->Children); 2372 } 2373 } 2374 2375 static unsigned maxNestingDepth(const AnnotatedLine &Line) { 2376 unsigned Result = 0; 2377 for (const auto *Tok = Line.First; Tok != nullptr; Tok = Tok->Next) 2378 Result = std::max(Result, Tok->NestingLevel); 2379 return Result; 2380 } 2381 2382 void TokenAnnotator::annotate(AnnotatedLine &Line) { 2383 for (auto &Child : Line.Children) 2384 annotate(*Child); 2385 2386 AnnotatingParser Parser(Style, Line, Keywords); 2387 Line.Type = Parser.parseLine(); 2388 2389 // With very deep nesting, ExpressionParser uses lots of stack and the 2390 // formatting algorithm is very slow. We're not going to do a good job here 2391 // anyway - it's probably generated code being formatted by mistake. 2392 // Just skip the whole line. 2393 if (maxNestingDepth(Line) > 50) 2394 Line.Type = LT_Invalid; 2395 2396 if (Line.Type == LT_Invalid) 2397 return; 2398 2399 ExpressionParser ExprParser(Style, Keywords, Line); 2400 ExprParser.parse(); 2401 2402 if (Line.startsWith(TT_ObjCMethodSpecifier)) 2403 Line.Type = LT_ObjCMethodDecl; 2404 else if (Line.startsWith(TT_ObjCDecl)) 2405 Line.Type = LT_ObjCDecl; 2406 else if (Line.startsWith(TT_ObjCProperty)) 2407 Line.Type = LT_ObjCProperty; 2408 2409 Line.First->SpacesRequiredBefore = 1; 2410 Line.First->CanBreakBefore = Line.First->MustBreakBefore; 2411 } 2412 2413 // This function heuristically determines whether 'Current' starts the name of a 2414 // function declaration. 2415 static bool isFunctionDeclarationName(bool IsCpp, const FormatToken &Current, 2416 const AnnotatedLine &Line) { 2417 auto skipOperatorName = [](const FormatToken *Next) -> const FormatToken * { 2418 for (; Next; Next = Next->Next) { 2419 if (Next->is(TT_OverloadedOperatorLParen)) 2420 return Next; 2421 if (Next->is(TT_OverloadedOperator)) 2422 continue; 2423 if (Next->isOneOf(tok::kw_new, tok::kw_delete)) { 2424 // For 'new[]' and 'delete[]'. 2425 if (Next->Next && 2426 Next->Next->startsSequence(tok::l_square, tok::r_square)) 2427 Next = Next->Next->Next; 2428 continue; 2429 } 2430 if (Next->startsSequence(tok::l_square, tok::r_square)) { 2431 // For operator[](). 2432 Next = Next->Next; 2433 continue; 2434 } 2435 if ((Next->isSimpleTypeSpecifier() || Next->is(tok::identifier)) && 2436 Next->Next && Next->Next->isOneOf(tok::star, tok::amp, tok::ampamp)) { 2437 // For operator void*(), operator char*(), operator Foo*(). 2438 Next = Next->Next; 2439 continue; 2440 } 2441 if (Next->is(TT_TemplateOpener) && Next->MatchingParen) { 2442 Next = Next->MatchingParen; 2443 continue; 2444 } 2445 2446 break; 2447 } 2448 return nullptr; 2449 }; 2450 2451 // Find parentheses of parameter list. 2452 const FormatToken *Next = Current.Next; 2453 if (Current.is(tok::kw_operator)) { 2454 if (Current.Previous && Current.Previous->is(tok::coloncolon)) 2455 return false; 2456 Next = skipOperatorName(Next); 2457 } else { 2458 if (!Current.is(TT_StartOfName) || Current.NestingLevel != 0) 2459 return false; 2460 for (; Next; Next = Next->Next) { 2461 if (Next->is(TT_TemplateOpener)) { 2462 Next = Next->MatchingParen; 2463 } else if (Next->is(tok::coloncolon)) { 2464 Next = Next->Next; 2465 if (!Next) 2466 return false; 2467 if (Next->is(tok::kw_operator)) { 2468 Next = skipOperatorName(Next->Next); 2469 break; 2470 } 2471 if (!Next->is(tok::identifier)) 2472 return false; 2473 } else if (Next->is(tok::l_paren)) { 2474 break; 2475 } else { 2476 return false; 2477 } 2478 } 2479 } 2480 2481 // Check whether parameter list can belong to a function declaration. 2482 if (!Next || !Next->is(tok::l_paren) || !Next->MatchingParen) 2483 return false; 2484 // If the lines ends with "{", this is likely a function definition. 2485 if (Line.Last->is(tok::l_brace)) 2486 return true; 2487 if (Next->Next == Next->MatchingParen) 2488 return true; // Empty parentheses. 2489 // If there is an &/&& after the r_paren, this is likely a function. 2490 if (Next->MatchingParen->Next && 2491 Next->MatchingParen->Next->is(TT_PointerOrReference)) 2492 return true; 2493 2494 // Check for K&R C function definitions (and C++ function definitions with 2495 // unnamed parameters), e.g.: 2496 // int f(i) 2497 // { 2498 // return i + 1; 2499 // } 2500 // bool g(size_t = 0, bool b = false) 2501 // { 2502 // return !b; 2503 // } 2504 if (IsCpp && Next->Next && Next->Next->is(tok::identifier) && 2505 !Line.endsWith(tok::semi)) 2506 return true; 2507 2508 for (const FormatToken *Tok = Next->Next; Tok && Tok != Next->MatchingParen; 2509 Tok = Tok->Next) { 2510 if (Tok->is(TT_TypeDeclarationParen)) 2511 return true; 2512 if (Tok->isOneOf(tok::l_paren, TT_TemplateOpener) && Tok->MatchingParen) { 2513 Tok = Tok->MatchingParen; 2514 continue; 2515 } 2516 if (Tok->is(tok::kw_const) || Tok->isSimpleTypeSpecifier() || 2517 Tok->isOneOf(TT_PointerOrReference, TT_StartOfName, tok::ellipsis)) 2518 return true; 2519 if (Tok->isOneOf(tok::l_brace, tok::string_literal, TT_ObjCMethodExpr) || 2520 Tok->Tok.isLiteral()) 2521 return false; 2522 } 2523 return false; 2524 } 2525 2526 bool TokenAnnotator::mustBreakForReturnType(const AnnotatedLine &Line) const { 2527 assert(Line.MightBeFunctionDecl); 2528 2529 if ((Style.AlwaysBreakAfterReturnType == FormatStyle::RTBS_TopLevel || 2530 Style.AlwaysBreakAfterReturnType == 2531 FormatStyle::RTBS_TopLevelDefinitions) && 2532 Line.Level > 0) 2533 return false; 2534 2535 switch (Style.AlwaysBreakAfterReturnType) { 2536 case FormatStyle::RTBS_None: 2537 return false; 2538 case FormatStyle::RTBS_All: 2539 case FormatStyle::RTBS_TopLevel: 2540 return true; 2541 case FormatStyle::RTBS_AllDefinitions: 2542 case FormatStyle::RTBS_TopLevelDefinitions: 2543 return Line.mightBeFunctionDefinition(); 2544 } 2545 2546 return false; 2547 } 2548 2549 void TokenAnnotator::calculateFormattingInformation(AnnotatedLine &Line) { 2550 for (AnnotatedLine *ChildLine : Line.Children) 2551 calculateFormattingInformation(*ChildLine); 2552 2553 Line.First->TotalLength = 2554 Line.First->IsMultiline ? Style.ColumnLimit 2555 : Line.FirstStartColumn + Line.First->ColumnWidth; 2556 FormatToken *Current = Line.First->Next; 2557 bool InFunctionDecl = Line.MightBeFunctionDecl; 2558 bool AlignArrayOfStructures = 2559 (Style.AlignArrayOfStructures != FormatStyle::AIAS_None && 2560 Line.Type == LT_ArrayOfStructInitializer); 2561 if (AlignArrayOfStructures) 2562 calculateArrayInitializerColumnList(Line); 2563 2564 while (Current) { 2565 if (isFunctionDeclarationName(Style.isCpp(), *Current, Line)) 2566 Current->setType(TT_FunctionDeclarationName); 2567 const FormatToken *Prev = Current->Previous; 2568 if (Current->is(TT_LineComment)) { 2569 if (Prev->is(BK_BracedInit) && Prev->opensScope()) 2570 Current->SpacesRequiredBefore = 2571 (Style.Cpp11BracedListStyle && !Style.SpacesInParentheses) ? 0 : 1; 2572 else 2573 Current->SpacesRequiredBefore = Style.SpacesBeforeTrailingComments; 2574 2575 // If we find a trailing comment, iterate backwards to determine whether 2576 // it seems to relate to a specific parameter. If so, break before that 2577 // parameter to avoid changing the comment's meaning. E.g. don't move 'b' 2578 // to the previous line in: 2579 // SomeFunction(a, 2580 // b, // comment 2581 // c); 2582 if (!Current->HasUnescapedNewline) { 2583 for (FormatToken *Parameter = Current->Previous; Parameter; 2584 Parameter = Parameter->Previous) { 2585 if (Parameter->isOneOf(tok::comment, tok::r_brace)) 2586 break; 2587 if (Parameter->Previous && Parameter->Previous->is(tok::comma)) { 2588 if (!Parameter->Previous->is(TT_CtorInitializerComma) && 2589 Parameter->HasUnescapedNewline) 2590 Parameter->MustBreakBefore = true; 2591 break; 2592 } 2593 } 2594 } 2595 } else if (Current->SpacesRequiredBefore == 0 && 2596 spaceRequiredBefore(Line, *Current)) { 2597 Current->SpacesRequiredBefore = 1; 2598 } 2599 2600 Current->MustBreakBefore = 2601 Current->MustBreakBefore || mustBreakBefore(Line, *Current); 2602 2603 if (!Current->MustBreakBefore && InFunctionDecl && 2604 Current->is(TT_FunctionDeclarationName)) 2605 Current->MustBreakBefore = mustBreakForReturnType(Line); 2606 2607 Current->CanBreakBefore = 2608 Current->MustBreakBefore || canBreakBefore(Line, *Current); 2609 unsigned ChildSize = 0; 2610 if (Prev->Children.size() == 1) { 2611 FormatToken &LastOfChild = *Prev->Children[0]->Last; 2612 ChildSize = LastOfChild.isTrailingComment() ? Style.ColumnLimit 2613 : LastOfChild.TotalLength + 1; 2614 } 2615 if (Current->MustBreakBefore || Prev->Children.size() > 1 || 2616 (Prev->Children.size() == 1 && 2617 Prev->Children[0]->First->MustBreakBefore) || 2618 Current->IsMultiline) 2619 Current->TotalLength = Prev->TotalLength + Style.ColumnLimit; 2620 else 2621 Current->TotalLength = Prev->TotalLength + Current->ColumnWidth + 2622 ChildSize + Current->SpacesRequiredBefore; 2623 2624 if (Current->is(TT_CtorInitializerColon)) 2625 InFunctionDecl = false; 2626 2627 // FIXME: Only calculate this if CanBreakBefore is true once static 2628 // initializers etc. are sorted out. 2629 // FIXME: Move magic numbers to a better place. 2630 2631 // Reduce penalty for aligning ObjC method arguments using the colon 2632 // alignment as this is the canonical way (still prefer fitting everything 2633 // into one line if possible). Trying to fit a whole expression into one 2634 // line should not force other line breaks (e.g. when ObjC method 2635 // expression is a part of other expression). 2636 Current->SplitPenalty = splitPenalty(Line, *Current, InFunctionDecl); 2637 if (Style.Language == FormatStyle::LK_ObjC && 2638 Current->is(TT_SelectorName) && Current->ParameterIndex > 0) { 2639 if (Current->ParameterIndex == 1) 2640 Current->SplitPenalty += 5 * Current->BindingStrength; 2641 } else { 2642 Current->SplitPenalty += 20 * Current->BindingStrength; 2643 } 2644 2645 Current = Current->Next; 2646 } 2647 2648 calculateUnbreakableTailLengths(Line); 2649 unsigned IndentLevel = Line.Level; 2650 for (Current = Line.First; Current != nullptr; Current = Current->Next) { 2651 if (Current->Role) 2652 Current->Role->precomputeFormattingInfos(Current); 2653 if (Current->MatchingParen && 2654 Current->MatchingParen->opensBlockOrBlockTypeList(Style) && 2655 IndentLevel > 0) 2656 --IndentLevel; 2657 Current->IndentLevel = IndentLevel; 2658 if (Current->opensBlockOrBlockTypeList(Style)) 2659 ++IndentLevel; 2660 } 2661 2662 LLVM_DEBUG({ printDebugInfo(Line); }); 2663 } 2664 2665 void TokenAnnotator::calculateUnbreakableTailLengths(AnnotatedLine &Line) { 2666 unsigned UnbreakableTailLength = 0; 2667 FormatToken *Current = Line.Last; 2668 while (Current) { 2669 Current->UnbreakableTailLength = UnbreakableTailLength; 2670 if (Current->CanBreakBefore || 2671 Current->isOneOf(tok::comment, tok::string_literal)) { 2672 UnbreakableTailLength = 0; 2673 } else { 2674 UnbreakableTailLength += 2675 Current->ColumnWidth + Current->SpacesRequiredBefore; 2676 } 2677 Current = Current->Previous; 2678 } 2679 } 2680 2681 void TokenAnnotator::calculateArrayInitializerColumnList(AnnotatedLine &Line) { 2682 if (Line.First == Line.Last) 2683 return; 2684 auto *CurrentToken = Line.First; 2685 CurrentToken->ArrayInitializerLineStart = true; 2686 unsigned Depth = 0; 2687 while (CurrentToken != nullptr && CurrentToken != Line.Last) { 2688 if (CurrentToken->is(tok::l_brace)) { 2689 CurrentToken->IsArrayInitializer = true; 2690 if (CurrentToken->Next != nullptr) 2691 CurrentToken->Next->MustBreakBefore = true; 2692 CurrentToken = 2693 calculateInitializerColumnList(Line, CurrentToken->Next, Depth + 1); 2694 } else { 2695 CurrentToken = CurrentToken->Next; 2696 } 2697 } 2698 } 2699 2700 FormatToken *TokenAnnotator::calculateInitializerColumnList( 2701 AnnotatedLine &Line, FormatToken *CurrentToken, unsigned Depth) { 2702 while (CurrentToken != nullptr && CurrentToken != Line.Last) { 2703 if (CurrentToken->is(tok::l_brace)) 2704 ++Depth; 2705 else if (CurrentToken->is(tok::r_brace)) 2706 --Depth; 2707 if (Depth == 2 && CurrentToken->isOneOf(tok::l_brace, tok::comma)) { 2708 CurrentToken = CurrentToken->Next; 2709 if (CurrentToken == nullptr) 2710 break; 2711 CurrentToken->StartsColumn = true; 2712 CurrentToken = CurrentToken->Previous; 2713 } 2714 CurrentToken = CurrentToken->Next; 2715 } 2716 return CurrentToken; 2717 } 2718 2719 unsigned TokenAnnotator::splitPenalty(const AnnotatedLine &Line, 2720 const FormatToken &Tok, 2721 bool InFunctionDecl) { 2722 const FormatToken &Left = *Tok.Previous; 2723 const FormatToken &Right = Tok; 2724 2725 if (Left.is(tok::semi)) 2726 return 0; 2727 2728 if (Style.Language == FormatStyle::LK_Java) { 2729 if (Right.isOneOf(Keywords.kw_extends, Keywords.kw_throws)) 2730 return 1; 2731 if (Right.is(Keywords.kw_implements)) 2732 return 2; 2733 if (Left.is(tok::comma) && Left.NestingLevel == 0) 2734 return 3; 2735 } else if (Style.isJavaScript()) { 2736 if (Right.is(Keywords.kw_function) && Left.isNot(tok::comma)) 2737 return 100; 2738 if (Left.is(TT_JsTypeColon)) 2739 return 35; 2740 if ((Left.is(TT_TemplateString) && Left.TokenText.endswith("${")) || 2741 (Right.is(TT_TemplateString) && Right.TokenText.startswith("}"))) 2742 return 100; 2743 // Prefer breaking call chains (".foo") over empty "{}", "[]" or "()". 2744 if (Left.opensScope() && Right.closesScope()) 2745 return 200; 2746 } 2747 2748 if (Right.is(tok::identifier) && Right.Next && Right.Next->is(TT_DictLiteral)) 2749 return 1; 2750 if (Right.is(tok::l_square)) { 2751 if (Style.Language == FormatStyle::LK_Proto) 2752 return 1; 2753 if (Left.is(tok::r_square)) 2754 return 200; 2755 // Slightly prefer formatting local lambda definitions like functions. 2756 if (Right.is(TT_LambdaLSquare) && Left.is(tok::equal)) 2757 return 35; 2758 if (!Right.isOneOf(TT_ObjCMethodExpr, TT_LambdaLSquare, 2759 TT_ArrayInitializerLSquare, 2760 TT_DesignatedInitializerLSquare, TT_AttributeSquare)) 2761 return 500; 2762 } 2763 2764 if (Left.is(tok::coloncolon) || 2765 (Right.is(tok::period) && Style.Language == FormatStyle::LK_Proto)) 2766 return 500; 2767 if (Right.isOneOf(TT_StartOfName, TT_FunctionDeclarationName) || 2768 Right.is(tok::kw_operator)) { 2769 if (Line.startsWith(tok::kw_for) && Right.PartOfMultiVariableDeclStmt) 2770 return 3; 2771 if (Left.is(TT_StartOfName)) 2772 return 110; 2773 if (InFunctionDecl && Right.NestingLevel == 0) 2774 return Style.PenaltyReturnTypeOnItsOwnLine; 2775 return 200; 2776 } 2777 if (Right.is(TT_PointerOrReference)) 2778 return 190; 2779 if (Right.is(TT_LambdaArrow)) 2780 return 110; 2781 if (Left.is(tok::equal) && Right.is(tok::l_brace)) 2782 return 160; 2783 if (Left.is(TT_CastRParen)) 2784 return 100; 2785 if (Left.isOneOf(tok::kw_class, tok::kw_struct)) 2786 return 5000; 2787 if (Left.is(tok::comment)) 2788 return 1000; 2789 2790 if (Left.isOneOf(TT_RangeBasedForLoopColon, TT_InheritanceColon, 2791 TT_CtorInitializerColon)) 2792 return 2; 2793 2794 if (Right.isMemberAccess()) { 2795 // Breaking before the "./->" of a chained call/member access is reasonably 2796 // cheap, as formatting those with one call per line is generally 2797 // desirable. In particular, it should be cheaper to break before the call 2798 // than it is to break inside a call's parameters, which could lead to weird 2799 // "hanging" indents. The exception is the very last "./->" to support this 2800 // frequent pattern: 2801 // 2802 // aaaaaaaa.aaaaaaaa.bbbbbbb().ccccccccccccccccccccc( 2803 // dddddddd); 2804 // 2805 // which might otherwise be blown up onto many lines. Here, clang-format 2806 // won't produce "hanging" indents anyway as there is no other trailing 2807 // call. 2808 // 2809 // Also apply higher penalty is not a call as that might lead to a wrapping 2810 // like: 2811 // 2812 // aaaaaaa 2813 // .aaaaaaaaa.bbbbbbbb(cccccccc); 2814 return !Right.NextOperator || !Right.NextOperator->Previous->closesScope() 2815 ? 150 2816 : 35; 2817 } 2818 2819 if (Right.is(TT_TrailingAnnotation) && 2820 (!Right.Next || Right.Next->isNot(tok::l_paren))) { 2821 // Moving trailing annotations to the next line is fine for ObjC method 2822 // declarations. 2823 if (Line.startsWith(TT_ObjCMethodSpecifier)) 2824 return 10; 2825 // Generally, breaking before a trailing annotation is bad unless it is 2826 // function-like. It seems to be especially preferable to keep standard 2827 // annotations (i.e. "const", "final" and "override") on the same line. 2828 // Use a slightly higher penalty after ")" so that annotations like 2829 // "const override" are kept together. 2830 bool is_short_annotation = Right.TokenText.size() < 10; 2831 return (Left.is(tok::r_paren) ? 100 : 120) + (is_short_annotation ? 50 : 0); 2832 } 2833 2834 // In for-loops, prefer breaking at ',' and ';'. 2835 if (Line.startsWith(tok::kw_for) && Left.is(tok::equal)) 2836 return 4; 2837 2838 // In Objective-C method expressions, prefer breaking before "param:" over 2839 // breaking after it. 2840 if (Right.is(TT_SelectorName)) 2841 return 0; 2842 if (Left.is(tok::colon) && Left.is(TT_ObjCMethodExpr)) 2843 return Line.MightBeFunctionDecl ? 50 : 500; 2844 2845 // In Objective-C type declarations, avoid breaking after the category's 2846 // open paren (we'll prefer breaking after the protocol list's opening 2847 // angle bracket, if present). 2848 if (Line.Type == LT_ObjCDecl && Left.is(tok::l_paren) && Left.Previous && 2849 Left.Previous->isOneOf(tok::identifier, tok::greater)) 2850 return 500; 2851 2852 if (Left.is(tok::l_paren) && Style.PenaltyBreakOpenParenthesis != 0) 2853 return Style.PenaltyBreakOpenParenthesis; 2854 if (Left.is(tok::l_paren) && InFunctionDecl && 2855 Style.AlignAfterOpenBracket != FormatStyle::BAS_DontAlign) 2856 return 100; 2857 if (Left.is(tok::l_paren) && Left.Previous && 2858 (Left.Previous->is(tok::kw_for) || Left.Previous->isIf())) 2859 return 1000; 2860 if (Left.is(tok::equal) && InFunctionDecl) 2861 return 110; 2862 if (Right.is(tok::r_brace)) 2863 return 1; 2864 if (Left.is(TT_TemplateOpener)) 2865 return 100; 2866 if (Left.opensScope()) { 2867 // If we aren't aligning after opening parens/braces we can always break 2868 // here unless the style does not want us to place all arguments on the 2869 // next line. 2870 if (Style.AlignAfterOpenBracket == FormatStyle::BAS_DontAlign && 2871 (Left.ParameterCount <= 1 || Style.AllowAllArgumentsOnNextLine)) 2872 return 0; 2873 if (Left.is(tok::l_brace) && !Style.Cpp11BracedListStyle) 2874 return 19; 2875 return Left.ParameterCount > 1 ? Style.PenaltyBreakBeforeFirstCallParameter 2876 : 19; 2877 } 2878 if (Left.is(TT_JavaAnnotation)) 2879 return 50; 2880 2881 if (Left.is(TT_UnaryOperator)) 2882 return 60; 2883 if (Left.isOneOf(tok::plus, tok::comma) && Left.Previous && 2884 Left.Previous->isLabelString() && 2885 (Left.NextOperator || Left.OperatorIndex != 0)) 2886 return 50; 2887 if (Right.is(tok::plus) && Left.isLabelString() && 2888 (Right.NextOperator || Right.OperatorIndex != 0)) 2889 return 25; 2890 if (Left.is(tok::comma)) 2891 return 1; 2892 if (Right.is(tok::lessless) && Left.isLabelString() && 2893 (Right.NextOperator || Right.OperatorIndex != 1)) 2894 return 25; 2895 if (Right.is(tok::lessless)) { 2896 // Breaking at a << is really cheap. 2897 if (!Left.is(tok::r_paren) || Right.OperatorIndex > 0) 2898 // Slightly prefer to break before the first one in log-like statements. 2899 return 2; 2900 return 1; 2901 } 2902 if (Left.ClosesTemplateDeclaration) 2903 return Style.PenaltyBreakTemplateDeclaration; 2904 if (Left.is(TT_ConditionalExpr)) 2905 return prec::Conditional; 2906 prec::Level Level = Left.getPrecedence(); 2907 if (Level == prec::Unknown) 2908 Level = Right.getPrecedence(); 2909 if (Level == prec::Assignment) 2910 return Style.PenaltyBreakAssignment; 2911 if (Level != prec::Unknown) 2912 return Level; 2913 2914 return 3; 2915 } 2916 2917 bool TokenAnnotator::spaceRequiredBeforeParens(const FormatToken &Right) const { 2918 if (Style.SpaceBeforeParens == FormatStyle::SBPO_Always) 2919 return true; 2920 if (Right.is(TT_OverloadedOperatorLParen) && 2921 Style.SpaceBeforeParensOptions.AfterOverloadedOperator) 2922 return true; 2923 if (Style.SpaceBeforeParensOptions.BeforeNonEmptyParentheses && 2924 Right.ParameterCount > 0) 2925 return true; 2926 return false; 2927 } 2928 2929 bool TokenAnnotator::spaceRequiredBetween(const AnnotatedLine &Line, 2930 const FormatToken &Left, 2931 const FormatToken &Right) { 2932 if (Left.is(tok::kw_return) && Right.isNot(tok::semi)) 2933 return true; 2934 if (Style.isJson() && Left.is(tok::string_literal) && Right.is(tok::colon)) 2935 return false; 2936 if (Left.is(Keywords.kw_assert) && Style.Language == FormatStyle::LK_Java) 2937 return true; 2938 if (Style.ObjCSpaceAfterProperty && Line.Type == LT_ObjCProperty && 2939 Left.Tok.getObjCKeywordID() == tok::objc_property) 2940 return true; 2941 if (Right.is(tok::hashhash)) 2942 return Left.is(tok::hash); 2943 if (Left.isOneOf(tok::hashhash, tok::hash)) 2944 return Right.is(tok::hash); 2945 if ((Left.is(tok::l_paren) && Right.is(tok::r_paren)) || 2946 (Left.is(tok::l_brace) && Left.isNot(BK_Block) && 2947 Right.is(tok::r_brace) && Right.isNot(BK_Block))) 2948 return Style.SpaceInEmptyParentheses; 2949 if (Style.SpacesInConditionalStatement) { 2950 if (Left.is(tok::l_paren) && Left.Previous && 2951 isKeywordWithCondition(*Left.Previous)) 2952 return true; 2953 if (Right.is(tok::r_paren) && Right.MatchingParen && 2954 Right.MatchingParen->Previous && 2955 isKeywordWithCondition(*Right.MatchingParen->Previous)) 2956 return true; 2957 } 2958 2959 // auto{x} auto(x) 2960 if (Left.is(tok::kw_auto) && Right.isOneOf(tok::l_paren, tok::l_brace)) 2961 return false; 2962 2963 // operator co_await(x) 2964 if (Right.is(tok::l_paren) && Left.is(tok::kw_co_await) && Left.Previous && 2965 Left.Previous->is(tok::kw_operator)) 2966 return false; 2967 // co_await (x), co_yield (x), co_return (x) 2968 if (Left.isOneOf(tok::kw_co_await, tok::kw_co_yield, tok::kw_co_return) && 2969 Right.isNot(tok::semi)) 2970 return true; 2971 // requires clause Concept1<T> && Concept2<T> 2972 if (Left.is(TT_ConstraintJunctions) && Right.is(tok::identifier)) 2973 return true; 2974 2975 if (Left.is(tok::l_paren) || Right.is(tok::r_paren)) 2976 return (Right.is(TT_CastRParen) || 2977 (Left.MatchingParen && Left.MatchingParen->is(TT_CastRParen))) 2978 ? Style.SpacesInCStyleCastParentheses 2979 : Style.SpacesInParentheses; 2980 if (Right.isOneOf(tok::semi, tok::comma)) 2981 return false; 2982 if (Right.is(tok::less) && Line.Type == LT_ObjCDecl) { 2983 bool IsLightweightGeneric = Right.MatchingParen && 2984 Right.MatchingParen->Next && 2985 Right.MatchingParen->Next->is(tok::colon); 2986 return !IsLightweightGeneric && Style.ObjCSpaceBeforeProtocolList; 2987 } 2988 if (Right.is(tok::less) && Left.is(tok::kw_template)) 2989 return Style.SpaceAfterTemplateKeyword; 2990 if (Left.isOneOf(tok::exclaim, tok::tilde)) 2991 return false; 2992 if (Left.is(tok::at) && 2993 Right.isOneOf(tok::identifier, tok::string_literal, tok::char_constant, 2994 tok::numeric_constant, tok::l_paren, tok::l_brace, 2995 tok::kw_true, tok::kw_false)) 2996 return false; 2997 if (Left.is(tok::colon)) 2998 return !Left.is(TT_ObjCMethodExpr); 2999 if (Left.is(tok::coloncolon)) 3000 return false; 3001 if (Left.is(tok::less) || Right.isOneOf(tok::greater, tok::less)) { 3002 if (Style.Language == FormatStyle::LK_TextProto || 3003 (Style.Language == FormatStyle::LK_Proto && 3004 (Left.is(TT_DictLiteral) || Right.is(TT_DictLiteral)))) { 3005 // Format empty list as `<>`. 3006 if (Left.is(tok::less) && Right.is(tok::greater)) 3007 return false; 3008 return !Style.Cpp11BracedListStyle; 3009 } 3010 return false; 3011 } 3012 if (Right.is(tok::ellipsis)) 3013 return Left.Tok.isLiteral() || (Left.is(tok::identifier) && Left.Previous && 3014 Left.Previous->is(tok::kw_case)); 3015 if (Left.is(tok::l_square) && Right.is(tok::amp)) 3016 return Style.SpacesInSquareBrackets; 3017 if (Right.is(TT_PointerOrReference)) { 3018 if (Left.is(tok::r_paren) && Line.MightBeFunctionDecl) { 3019 if (!Left.MatchingParen) 3020 return true; 3021 FormatToken *TokenBeforeMatchingParen = 3022 Left.MatchingParen->getPreviousNonComment(); 3023 if (!TokenBeforeMatchingParen || !Left.is(TT_TypeDeclarationParen)) 3024 return true; 3025 } 3026 // Add a space if the previous token is a pointer qualifier or the closing 3027 // parenthesis of __attribute__(()) expression and the style requires spaces 3028 // after pointer qualifiers. 3029 if ((Style.SpaceAroundPointerQualifiers == FormatStyle::SAPQ_After || 3030 Style.SpaceAroundPointerQualifiers == FormatStyle::SAPQ_Both) && 3031 (Left.is(TT_AttributeParen) || Left.canBePointerOrReferenceQualifier())) 3032 return true; 3033 if (Left.Tok.isLiteral()) 3034 return true; 3035 // for (auto a = 0, b = 0; const auto & c : {1, 2, 3}) 3036 if (Left.isTypeOrIdentifier() && Right.Next && Right.Next->Next && 3037 Right.Next->Next->is(TT_RangeBasedForLoopColon)) 3038 return getTokenPointerOrReferenceAlignment(Right) != 3039 FormatStyle::PAS_Left; 3040 return ( 3041 (!Left.isOneOf(TT_PointerOrReference, tok::l_paren) && 3042 (getTokenPointerOrReferenceAlignment(Right) != FormatStyle::PAS_Left || 3043 (Line.IsMultiVariableDeclStmt && 3044 (Left.NestingLevel == 0 || 3045 (Left.NestingLevel == 1 && Line.First->is(tok::kw_for))))))); 3046 } 3047 if (Right.is(TT_FunctionTypeLParen) && Left.isNot(tok::l_paren) && 3048 (!Left.is(TT_PointerOrReference) || 3049 (getTokenPointerOrReferenceAlignment(Left) != FormatStyle::PAS_Right && 3050 !Line.IsMultiVariableDeclStmt))) 3051 return true; 3052 if (Left.is(TT_PointerOrReference)) { 3053 // Add a space if the next token is a pointer qualifier and the style 3054 // requires spaces before pointer qualifiers. 3055 if ((Style.SpaceAroundPointerQualifiers == FormatStyle::SAPQ_Before || 3056 Style.SpaceAroundPointerQualifiers == FormatStyle::SAPQ_Both) && 3057 Right.canBePointerOrReferenceQualifier()) 3058 return true; 3059 // & 1 3060 if (Right.Tok.isLiteral()) 3061 return true; 3062 // & /* comment 3063 if (Right.is(TT_BlockComment)) 3064 return true; 3065 // foo() -> const Bar * override/final 3066 if (Right.isOneOf(Keywords.kw_override, Keywords.kw_final) && 3067 !Right.is(TT_StartOfName)) 3068 return true; 3069 // & { 3070 if (Right.is(tok::l_brace) && Right.is(BK_Block)) 3071 return true; 3072 // for (auto a = 0, b = 0; const auto& c : {1, 2, 3}) 3073 if (Left.Previous && Left.Previous->isTypeOrIdentifier() && Right.Next && 3074 Right.Next->is(TT_RangeBasedForLoopColon)) 3075 return getTokenPointerOrReferenceAlignment(Left) != 3076 FormatStyle::PAS_Right; 3077 return !Right.isOneOf(TT_PointerOrReference, TT_ArraySubscriptLSquare, 3078 tok::l_paren) && 3079 (getTokenPointerOrReferenceAlignment(Left) != 3080 FormatStyle::PAS_Right && 3081 !Line.IsMultiVariableDeclStmt) && 3082 Left.Previous && 3083 !Left.Previous->isOneOf(tok::l_paren, tok::coloncolon, 3084 tok::l_square); 3085 } 3086 // Ensure right pointer alignment with ellipsis e.g. int *...P 3087 if (Left.is(tok::ellipsis) && Left.Previous && 3088 Left.Previous->isOneOf(tok::star, tok::amp, tok::ampamp)) 3089 return Style.PointerAlignment != FormatStyle::PAS_Right; 3090 3091 if (Right.is(tok::star) && Left.is(tok::l_paren)) 3092 return false; 3093 if (Left.is(tok::star) && Right.isOneOf(tok::star, tok::amp, tok::ampamp)) 3094 return false; 3095 if (Right.isOneOf(tok::star, tok::amp, tok::ampamp)) { 3096 const FormatToken *Previous = &Left; 3097 while (Previous && !Previous->is(tok::kw_operator)) { 3098 if (Previous->is(tok::identifier) || Previous->isSimpleTypeSpecifier()) { 3099 Previous = Previous->getPreviousNonComment(); 3100 continue; 3101 } 3102 if (Previous->is(TT_TemplateCloser) && Previous->MatchingParen) { 3103 Previous = Previous->MatchingParen->getPreviousNonComment(); 3104 continue; 3105 } 3106 if (Previous->is(tok::coloncolon)) { 3107 Previous = Previous->getPreviousNonComment(); 3108 continue; 3109 } 3110 break; 3111 } 3112 // Space between the type and the * in: 3113 // operator void*() 3114 // operator char*() 3115 // operator void const*() 3116 // operator void volatile*() 3117 // operator /*comment*/ const char*() 3118 // operator volatile /*comment*/ char*() 3119 // operator Foo*() 3120 // operator C<T>*() 3121 // operator std::Foo*() 3122 // operator C<T>::D<U>*() 3123 // dependent on PointerAlignment style. 3124 if (Previous) { 3125 if (Previous->endsSequence(tok::kw_operator)) 3126 return (Style.PointerAlignment != FormatStyle::PAS_Left); 3127 if (Previous->is(tok::kw_const) || Previous->is(tok::kw_volatile)) 3128 return (Style.PointerAlignment != FormatStyle::PAS_Left) || 3129 (Style.SpaceAroundPointerQualifiers == 3130 FormatStyle::SAPQ_After) || 3131 (Style.SpaceAroundPointerQualifiers == FormatStyle::SAPQ_Both); 3132 } 3133 } 3134 const auto SpaceRequiredForArrayInitializerLSquare = 3135 [](const FormatToken &LSquareTok, const FormatStyle &Style) { 3136 return Style.SpacesInContainerLiterals || 3137 ((Style.Language == FormatStyle::LK_Proto || 3138 Style.Language == FormatStyle::LK_TextProto) && 3139 !Style.Cpp11BracedListStyle && 3140 LSquareTok.endsSequence(tok::l_square, tok::colon, 3141 TT_SelectorName)); 3142 }; 3143 if (Left.is(tok::l_square)) 3144 return (Left.is(TT_ArrayInitializerLSquare) && Right.isNot(tok::r_square) && 3145 SpaceRequiredForArrayInitializerLSquare(Left, Style)) || 3146 (Left.isOneOf(TT_ArraySubscriptLSquare, TT_StructuredBindingLSquare, 3147 TT_LambdaLSquare) && 3148 Style.SpacesInSquareBrackets && Right.isNot(tok::r_square)); 3149 if (Right.is(tok::r_square)) 3150 return Right.MatchingParen && 3151 ((Right.MatchingParen->is(TT_ArrayInitializerLSquare) && 3152 SpaceRequiredForArrayInitializerLSquare(*Right.MatchingParen, 3153 Style)) || 3154 (Style.SpacesInSquareBrackets && 3155 Right.MatchingParen->isOneOf(TT_ArraySubscriptLSquare, 3156 TT_StructuredBindingLSquare, 3157 TT_LambdaLSquare)) || 3158 Right.MatchingParen->is(TT_AttributeParen)); 3159 if (Right.is(tok::l_square) && 3160 !Right.isOneOf(TT_ObjCMethodExpr, TT_LambdaLSquare, 3161 TT_DesignatedInitializerLSquare, 3162 TT_StructuredBindingLSquare, TT_AttributeSquare) && 3163 !Left.isOneOf(tok::numeric_constant, TT_DictLiteral) && 3164 !(!Left.is(tok::r_square) && Style.SpaceBeforeSquareBrackets && 3165 Right.is(TT_ArraySubscriptLSquare))) 3166 return false; 3167 if (Left.is(tok::l_brace) && Right.is(tok::r_brace)) 3168 return !Left.Children.empty(); // No spaces in "{}". 3169 if ((Left.is(tok::l_brace) && Left.isNot(BK_Block)) || 3170 (Right.is(tok::r_brace) && Right.MatchingParen && 3171 Right.MatchingParen->isNot(BK_Block))) 3172 return Style.Cpp11BracedListStyle ? Style.SpacesInParentheses : true; 3173 if (Left.is(TT_BlockComment)) 3174 // No whitespace in x(/*foo=*/1), except for JavaScript. 3175 return Style.isJavaScript() || !Left.TokenText.endswith("=*/"); 3176 3177 // Space between template and attribute. 3178 // e.g. template <typename T> [[nodiscard]] ... 3179 if (Left.is(TT_TemplateCloser) && Right.is(TT_AttributeSquare)) 3180 return true; 3181 // Space before parentheses common for all languages 3182 if (Right.is(tok::l_paren)) { 3183 if (Left.is(TT_TemplateCloser) && Right.isNot(TT_FunctionTypeLParen)) 3184 return spaceRequiredBeforeParens(Right); 3185 if (Left.is(tok::kw_requires)) 3186 return spaceRequiredBeforeParens(Right); 3187 if ((Left.is(tok::r_paren) && Left.is(TT_AttributeParen)) || 3188 (Left.is(tok::r_square) && Left.is(TT_AttributeSquare))) 3189 return true; 3190 if (Left.is(TT_ForEachMacro)) 3191 return (Style.SpaceBeforeParensOptions.AfterForeachMacros || 3192 spaceRequiredBeforeParens(Right)); 3193 if (Left.is(TT_IfMacro)) 3194 return (Style.SpaceBeforeParensOptions.AfterIfMacros || 3195 spaceRequiredBeforeParens(Right)); 3196 if (Line.Type == LT_ObjCDecl) 3197 return true; 3198 if (Left.is(tok::semi)) 3199 return true; 3200 if (Left.isOneOf(tok::pp_elif, tok::kw_for, tok::kw_while, tok::kw_switch, 3201 tok::kw_case, TT_ForEachMacro, TT_ObjCForIn)) 3202 return Style.SpaceBeforeParensOptions.AfterControlStatements || 3203 spaceRequiredBeforeParens(Right); 3204 if (Left.isIf(Line.Type != LT_PreprocessorDirective)) 3205 return Style.SpaceBeforeParensOptions.AfterControlStatements || 3206 spaceRequiredBeforeParens(Right); 3207 3208 // TODO add Operator overloading specific Options to 3209 // SpaceBeforeParensOptions 3210 if (Right.is(TT_OverloadedOperatorLParen)) 3211 return spaceRequiredBeforeParens(Right); 3212 // Function declaration or definition 3213 if (Line.MightBeFunctionDecl && (Left.is(TT_FunctionDeclarationName))) { 3214 if (Line.mightBeFunctionDefinition()) 3215 return Style.SpaceBeforeParensOptions.AfterFunctionDefinitionName || 3216 spaceRequiredBeforeParens(Right); 3217 else 3218 return Style.SpaceBeforeParensOptions.AfterFunctionDeclarationName || 3219 spaceRequiredBeforeParens(Right); 3220 } 3221 // Lambda 3222 if (Line.Type != LT_PreprocessorDirective && Left.is(tok::r_square) && 3223 Left.MatchingParen && Left.MatchingParen->is(TT_LambdaLSquare)) 3224 return Style.SpaceBeforeParensOptions.AfterFunctionDefinitionName || 3225 spaceRequiredBeforeParens(Right); 3226 if (!Left.Previous || Left.Previous->isNot(tok::period)) { 3227 if (Left.isOneOf(tok::kw_try, Keywords.kw___except, tok::kw_catch)) 3228 return Style.SpaceBeforeParensOptions.AfterControlStatements || 3229 spaceRequiredBeforeParens(Right); 3230 if (Left.isOneOf(tok::kw_new, tok::kw_delete)) 3231 return Style.SpaceBeforeParens != FormatStyle::SBPO_Never || 3232 spaceRequiredBeforeParens(Right); 3233 } 3234 if (Line.Type != LT_PreprocessorDirective && 3235 (Left.is(tok::identifier) || Left.isFunctionLikeKeyword() || 3236 Left.is(tok::r_paren) || Left.isSimpleTypeSpecifier())) 3237 return spaceRequiredBeforeParens(Right); 3238 return false; 3239 } 3240 if (Left.is(tok::at) && Right.Tok.getObjCKeywordID() != tok::objc_not_keyword) 3241 return false; 3242 if (Right.is(TT_UnaryOperator)) 3243 return !Left.isOneOf(tok::l_paren, tok::l_square, tok::at) && 3244 (Left.isNot(tok::colon) || Left.isNot(TT_ObjCMethodExpr)); 3245 if ((Left.isOneOf(tok::identifier, tok::greater, tok::r_square, 3246 tok::r_paren) || 3247 Left.isSimpleTypeSpecifier()) && 3248 Right.is(tok::l_brace) && Right.getNextNonComment() && 3249 Right.isNot(BK_Block)) 3250 return false; 3251 if (Left.is(tok::period) || Right.is(tok::period)) 3252 return false; 3253 // u#str, U#str, L#str, u8#str 3254 // uR#str, UR#str, LR#str, u8R#str 3255 if (Right.is(tok::hash) && Left.is(tok::identifier) && 3256 (Left.TokenText == "L" || Left.TokenText == "u" || 3257 Left.TokenText == "U" || Left.TokenText == "u8" || 3258 Left.TokenText == "LR" || Left.TokenText == "uR" || 3259 Left.TokenText == "UR" || Left.TokenText == "u8R")) 3260 return false; 3261 if (Left.is(TT_TemplateCloser) && Left.MatchingParen && 3262 Left.MatchingParen->Previous && 3263 (Left.MatchingParen->Previous->is(tok::period) || 3264 Left.MatchingParen->Previous->is(tok::coloncolon))) 3265 // Java call to generic function with explicit type: 3266 // A.<B<C<...>>>DoSomething(); 3267 // A::<B<C<...>>>DoSomething(); // With a Java 8 method reference. 3268 return false; 3269 if (Left.is(TT_TemplateCloser) && Right.is(tok::l_square)) 3270 return false; 3271 if (Left.is(tok::l_brace) && Left.endsSequence(TT_DictLiteral, tok::at)) 3272 // Objective-C dictionary literal -> no space after opening brace. 3273 return false; 3274 if (Right.is(tok::r_brace) && Right.MatchingParen && 3275 Right.MatchingParen->endsSequence(TT_DictLiteral, tok::at)) 3276 // Objective-C dictionary literal -> no space before closing brace. 3277 return false; 3278 if (Right.getType() == TT_TrailingAnnotation && 3279 Right.isOneOf(tok::amp, tok::ampamp) && 3280 Left.isOneOf(tok::kw_const, tok::kw_volatile) && 3281 (!Right.Next || Right.Next->is(tok::semi))) 3282 // Match const and volatile ref-qualifiers without any additional 3283 // qualifiers such as 3284 // void Fn() const &; 3285 return getTokenReferenceAlignment(Right) != FormatStyle::PAS_Left; 3286 3287 return true; 3288 } 3289 3290 bool TokenAnnotator::spaceRequiredBefore(const AnnotatedLine &Line, 3291 const FormatToken &Right) { 3292 const FormatToken &Left = *Right.Previous; 3293 3294 // If the token is finalized don't touch it (as it could be in a 3295 // clang-format-off section). 3296 if (Left.Finalized) 3297 return Right.hasWhitespaceBefore(); 3298 3299 if (Right.Tok.getIdentifierInfo() && Left.Tok.getIdentifierInfo()) 3300 return true; // Never ever merge two identifiers. 3301 3302 // Leave a space between * and /* to avoid C4138 `comment end` found outside 3303 // of comment. 3304 if (Left.is(tok::star) && Right.is(tok::comment)) 3305 return true; 3306 3307 if (Style.isCpp()) { 3308 // Space between import <iostream>. 3309 // or import .....; 3310 if (Left.is(Keywords.kw_import) && Right.isOneOf(tok::less, tok::ellipsis)) 3311 return true; 3312 // Space between `module :` and `import :`. 3313 if (Left.isOneOf(Keywords.kw_module, Keywords.kw_import) && 3314 Right.is(TT_ModulePartitionColon)) 3315 return true; 3316 // No space between import foo:bar but keep a space between import :bar; 3317 if (Left.is(tok::identifier) && Right.is(TT_ModulePartitionColon)) 3318 return false; 3319 // No space between :bar; 3320 if (Left.is(TT_ModulePartitionColon) && 3321 Right.isOneOf(tok::identifier, tok::kw_private)) 3322 return false; 3323 if (Left.is(tok::ellipsis) && Right.is(tok::identifier) && 3324 Line.First->is(Keywords.kw_import)) 3325 return false; 3326 // Space in __attribute__((attr)) ::type. 3327 if (Left.is(TT_AttributeParen) && Right.is(tok::coloncolon)) 3328 return true; 3329 3330 if (Left.is(tok::kw_operator)) 3331 return Right.is(tok::coloncolon); 3332 if (Right.is(tok::l_brace) && Right.is(BK_BracedInit) && 3333 !Left.opensScope() && Style.SpaceBeforeCpp11BracedList) 3334 return true; 3335 if (Left.is(tok::less) && Left.is(TT_OverloadedOperator) && 3336 Right.is(TT_TemplateOpener)) 3337 return true; 3338 } else if (Style.Language == FormatStyle::LK_Proto || 3339 Style.Language == FormatStyle::LK_TextProto) { 3340 if (Right.is(tok::period) && 3341 Left.isOneOf(Keywords.kw_optional, Keywords.kw_required, 3342 Keywords.kw_repeated, Keywords.kw_extend)) 3343 return true; 3344 if (Right.is(tok::l_paren) && 3345 Left.isOneOf(Keywords.kw_returns, Keywords.kw_option)) 3346 return true; 3347 if (Right.isOneOf(tok::l_brace, tok::less) && Left.is(TT_SelectorName)) 3348 return true; 3349 // Slashes occur in text protocol extension syntax: [type/type] { ... }. 3350 if (Left.is(tok::slash) || Right.is(tok::slash)) 3351 return false; 3352 if (Left.MatchingParen && 3353 Left.MatchingParen->is(TT_ProtoExtensionLSquare) && 3354 Right.isOneOf(tok::l_brace, tok::less)) 3355 return !Style.Cpp11BracedListStyle; 3356 // A percent is probably part of a formatting specification, such as %lld. 3357 if (Left.is(tok::percent)) 3358 return false; 3359 // Preserve the existence of a space before a percent for cases like 0x%04x 3360 // and "%d %d" 3361 if (Left.is(tok::numeric_constant) && Right.is(tok::percent)) 3362 return Right.hasWhitespaceBefore(); 3363 } else if (Style.isJson()) { 3364 if (Right.is(tok::colon)) 3365 return false; 3366 } else if (Style.isCSharp()) { 3367 // Require spaces around '{' and before '}' unless they appear in 3368 // interpolated strings. Interpolated strings are merged into a single token 3369 // so cannot have spaces inserted by this function. 3370 3371 // No space between 'this' and '[' 3372 if (Left.is(tok::kw_this) && Right.is(tok::l_square)) 3373 return false; 3374 3375 // No space between 'new' and '(' 3376 if (Left.is(tok::kw_new) && Right.is(tok::l_paren)) 3377 return false; 3378 3379 // Space before { (including space within '{ {'). 3380 if (Right.is(tok::l_brace)) 3381 return true; 3382 3383 // Spaces inside braces. 3384 if (Left.is(tok::l_brace) && Right.isNot(tok::r_brace)) 3385 return true; 3386 3387 if (Left.isNot(tok::l_brace) && Right.is(tok::r_brace)) 3388 return true; 3389 3390 // Spaces around '=>'. 3391 if (Left.is(TT_FatArrow) || Right.is(TT_FatArrow)) 3392 return true; 3393 3394 // No spaces around attribute target colons 3395 if (Left.is(TT_AttributeColon) || Right.is(TT_AttributeColon)) 3396 return false; 3397 3398 // space between type and variable e.g. Dictionary<string,string> foo; 3399 if (Left.is(TT_TemplateCloser) && Right.is(TT_StartOfName)) 3400 return true; 3401 3402 // spaces inside square brackets. 3403 if (Left.is(tok::l_square) || Right.is(tok::r_square)) 3404 return Style.SpacesInSquareBrackets; 3405 3406 // No space before ? in nullable types. 3407 if (Right.is(TT_CSharpNullable)) 3408 return false; 3409 3410 // No space before null forgiving '!'. 3411 if (Right.is(TT_NonNullAssertion)) 3412 return false; 3413 3414 // No space between consecutive commas '[,,]'. 3415 if (Left.is(tok::comma) && Right.is(tok::comma)) 3416 return false; 3417 3418 // space after var in `var (key, value)` 3419 if (Left.is(Keywords.kw_var) && Right.is(tok::l_paren)) 3420 return true; 3421 3422 // space between keywords and paren e.g. "using (" 3423 if (Right.is(tok::l_paren)) 3424 if (Left.isOneOf(tok::kw_using, Keywords.kw_async, Keywords.kw_when, 3425 Keywords.kw_lock)) 3426 return Style.SpaceBeforeParensOptions.AfterControlStatements || 3427 spaceRequiredBeforeParens(Right); 3428 3429 // space between method modifier and opening parenthesis of a tuple return 3430 // type 3431 if (Left.isOneOf(tok::kw_public, tok::kw_private, tok::kw_protected, 3432 tok::kw_virtual, tok::kw_extern, tok::kw_static, 3433 Keywords.kw_internal, Keywords.kw_abstract, 3434 Keywords.kw_sealed, Keywords.kw_override, 3435 Keywords.kw_async, Keywords.kw_unsafe) && 3436 Right.is(tok::l_paren)) 3437 return true; 3438 } else if (Style.isJavaScript()) { 3439 if (Left.is(TT_FatArrow)) 3440 return true; 3441 // for await ( ... 3442 if (Right.is(tok::l_paren) && Left.is(Keywords.kw_await) && Left.Previous && 3443 Left.Previous->is(tok::kw_for)) 3444 return true; 3445 if (Left.is(Keywords.kw_async) && Right.is(tok::l_paren) && 3446 Right.MatchingParen) { 3447 const FormatToken *Next = Right.MatchingParen->getNextNonComment(); 3448 // An async arrow function, for example: `x = async () => foo();`, 3449 // as opposed to calling a function called async: `x = async();` 3450 if (Next && Next->is(TT_FatArrow)) 3451 return true; 3452 } 3453 if ((Left.is(TT_TemplateString) && Left.TokenText.endswith("${")) || 3454 (Right.is(TT_TemplateString) && Right.TokenText.startswith("}"))) 3455 return false; 3456 // In tagged template literals ("html`bar baz`"), there is no space between 3457 // the tag identifier and the template string. 3458 if (Keywords.IsJavaScriptIdentifier(Left, 3459 /* AcceptIdentifierName= */ false) && 3460 Right.is(TT_TemplateString)) 3461 return false; 3462 if (Right.is(tok::star) && 3463 Left.isOneOf(Keywords.kw_function, Keywords.kw_yield)) 3464 return false; 3465 if (Right.isOneOf(tok::l_brace, tok::l_square) && 3466 Left.isOneOf(Keywords.kw_function, Keywords.kw_yield, 3467 Keywords.kw_extends, Keywords.kw_implements)) 3468 return true; 3469 if (Right.is(tok::l_paren)) { 3470 // JS methods can use some keywords as names (e.g. `delete()`). 3471 if (Line.MustBeDeclaration && Left.Tok.getIdentifierInfo()) 3472 return false; 3473 // Valid JS method names can include keywords, e.g. `foo.delete()` or 3474 // `bar.instanceof()`. Recognize call positions by preceding period. 3475 if (Left.Previous && Left.Previous->is(tok::period) && 3476 Left.Tok.getIdentifierInfo()) 3477 return false; 3478 // Additional unary JavaScript operators that need a space after. 3479 if (Left.isOneOf(tok::kw_throw, Keywords.kw_await, Keywords.kw_typeof, 3480 tok::kw_void)) 3481 return true; 3482 } 3483 // `foo as const;` casts into a const type. 3484 if (Left.endsSequence(tok::kw_const, Keywords.kw_as)) 3485 return false; 3486 if ((Left.isOneOf(Keywords.kw_let, Keywords.kw_var, Keywords.kw_in, 3487 tok::kw_const) || 3488 // "of" is only a keyword if it appears after another identifier 3489 // (e.g. as "const x of y" in a for loop), or after a destructuring 3490 // operation (const [x, y] of z, const {a, b} of c). 3491 (Left.is(Keywords.kw_of) && Left.Previous && 3492 (Left.Previous->Tok.is(tok::identifier) || 3493 Left.Previous->isOneOf(tok::r_square, tok::r_brace)))) && 3494 (!Left.Previous || !Left.Previous->is(tok::period))) 3495 return true; 3496 if (Left.isOneOf(tok::kw_for, Keywords.kw_as) && Left.Previous && 3497 Left.Previous->is(tok::period) && Right.is(tok::l_paren)) 3498 return false; 3499 if (Left.is(Keywords.kw_as) && 3500 Right.isOneOf(tok::l_square, tok::l_brace, tok::l_paren)) 3501 return true; 3502 if (Left.is(tok::kw_default) && Left.Previous && 3503 Left.Previous->is(tok::kw_export)) 3504 return true; 3505 if (Left.is(Keywords.kw_is) && Right.is(tok::l_brace)) 3506 return true; 3507 if (Right.isOneOf(TT_JsTypeColon, TT_JsTypeOptionalQuestion)) 3508 return false; 3509 if (Left.is(TT_JsTypeOperator) || Right.is(TT_JsTypeOperator)) 3510 return false; 3511 if ((Left.is(tok::l_brace) || Right.is(tok::r_brace)) && 3512 Line.First->isOneOf(Keywords.kw_import, tok::kw_export)) 3513 return false; 3514 if (Left.is(tok::ellipsis)) 3515 return false; 3516 if (Left.is(TT_TemplateCloser) && 3517 !Right.isOneOf(tok::equal, tok::l_brace, tok::comma, tok::l_square, 3518 Keywords.kw_implements, Keywords.kw_extends)) 3519 // Type assertions ('<type>expr') are not followed by whitespace. Other 3520 // locations that should have whitespace following are identified by the 3521 // above set of follower tokens. 3522 return false; 3523 if (Right.is(TT_NonNullAssertion)) 3524 return false; 3525 if (Left.is(TT_NonNullAssertion) && 3526 Right.isOneOf(Keywords.kw_as, Keywords.kw_in)) 3527 return true; // "x! as string", "x! in y" 3528 } else if (Style.Language == FormatStyle::LK_Java) { 3529 if (Left.is(tok::r_square) && Right.is(tok::l_brace)) 3530 return true; 3531 if (Left.is(Keywords.kw_synchronized) && Right.is(tok::l_paren)) 3532 return Style.SpaceBeforeParensOptions.AfterControlStatements || 3533 spaceRequiredBeforeParens(Right); 3534 if ((Left.isOneOf(tok::kw_static, tok::kw_public, tok::kw_private, 3535 tok::kw_protected) || 3536 Left.isOneOf(Keywords.kw_final, Keywords.kw_abstract, 3537 Keywords.kw_native)) && 3538 Right.is(TT_TemplateOpener)) 3539 return true; 3540 } 3541 if (Left.is(TT_ImplicitStringLiteral)) 3542 return Right.hasWhitespaceBefore(); 3543 if (Line.Type == LT_ObjCMethodDecl) { 3544 if (Left.is(TT_ObjCMethodSpecifier)) 3545 return true; 3546 if (Left.is(tok::r_paren) && canBeObjCSelectorComponent(Right)) 3547 // Don't space between ')' and <id> or ')' and 'new'. 'new' is not a 3548 // keyword in Objective-C, and '+ (instancetype)new;' is a standard class 3549 // method declaration. 3550 return false; 3551 } 3552 if (Line.Type == LT_ObjCProperty && 3553 (Right.is(tok::equal) || Left.is(tok::equal))) 3554 return false; 3555 3556 if (Right.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow) || 3557 Left.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow)) 3558 return true; 3559 if (Left.is(tok::comma) && !Right.is(TT_OverloadedOperatorLParen)) 3560 return true; 3561 if (Right.is(tok::comma)) 3562 return false; 3563 if (Right.is(TT_ObjCBlockLParen)) 3564 return true; 3565 if (Right.is(TT_CtorInitializerColon)) 3566 return Style.SpaceBeforeCtorInitializerColon; 3567 if (Right.is(TT_InheritanceColon) && !Style.SpaceBeforeInheritanceColon) 3568 return false; 3569 if (Right.is(TT_RangeBasedForLoopColon) && 3570 !Style.SpaceBeforeRangeBasedForLoopColon) 3571 return false; 3572 if (Left.is(TT_BitFieldColon)) 3573 return Style.BitFieldColonSpacing == FormatStyle::BFCS_Both || 3574 Style.BitFieldColonSpacing == FormatStyle::BFCS_After; 3575 if (Right.is(tok::colon)) { 3576 if (Line.First->isOneOf(tok::kw_default, tok::kw_case)) 3577 return Style.SpaceBeforeCaseColon; 3578 if (!Right.getNextNonComment() || Right.getNextNonComment()->is(tok::semi)) 3579 return false; 3580 if (Right.is(TT_ObjCMethodExpr)) 3581 return false; 3582 if (Left.is(tok::question)) 3583 return false; 3584 if (Right.is(TT_InlineASMColon) && Left.is(tok::coloncolon)) 3585 return false; 3586 if (Right.is(TT_DictLiteral)) 3587 return Style.SpacesInContainerLiterals; 3588 if (Right.is(TT_AttributeColon)) 3589 return false; 3590 if (Right.is(TT_CSharpNamedArgumentColon)) 3591 return false; 3592 if (Right.is(TT_BitFieldColon)) 3593 return Style.BitFieldColonSpacing == FormatStyle::BFCS_Both || 3594 Style.BitFieldColonSpacing == FormatStyle::BFCS_Before; 3595 return true; 3596 } 3597 // Do not merge "- -" into "--". 3598 if ((Left.isOneOf(tok::minus, tok::minusminus) && 3599 Right.isOneOf(tok::minus, tok::minusminus)) || 3600 (Left.isOneOf(tok::plus, tok::plusplus) && 3601 Right.isOneOf(tok::plus, tok::plusplus))) 3602 return true; 3603 if (Left.is(TT_UnaryOperator)) { 3604 if (!Right.is(tok::l_paren)) { 3605 // The alternative operators for ~ and ! are "compl" and "not". 3606 // If they are used instead, we do not want to combine them with 3607 // the token to the right, unless that is a left paren. 3608 if (Left.is(tok::exclaim) && Left.TokenText == "not") 3609 return true; 3610 if (Left.is(tok::tilde) && Left.TokenText == "compl") 3611 return true; 3612 // Lambda captures allow for a lone &, so "&]" needs to be properly 3613 // handled. 3614 if (Left.is(tok::amp) && Right.is(tok::r_square)) 3615 return Style.SpacesInSquareBrackets; 3616 } 3617 return (Style.SpaceAfterLogicalNot && Left.is(tok::exclaim)) || 3618 Right.is(TT_BinaryOperator); 3619 } 3620 3621 // If the next token is a binary operator or a selector name, we have 3622 // incorrectly classified the parenthesis as a cast. FIXME: Detect correctly. 3623 if (Left.is(TT_CastRParen)) 3624 return Style.SpaceAfterCStyleCast || 3625 Right.isOneOf(TT_BinaryOperator, TT_SelectorName); 3626 3627 auto ShouldAddSpacesInAngles = [this, &Right]() { 3628 if (this->Style.SpacesInAngles == FormatStyle::SIAS_Always) 3629 return true; 3630 if (this->Style.SpacesInAngles == FormatStyle::SIAS_Leave) 3631 return Right.hasWhitespaceBefore(); 3632 return false; 3633 }; 3634 3635 if (Left.is(tok::greater) && Right.is(tok::greater)) { 3636 if (Style.Language == FormatStyle::LK_TextProto || 3637 (Style.Language == FormatStyle::LK_Proto && Left.is(TT_DictLiteral))) 3638 return !Style.Cpp11BracedListStyle; 3639 return Right.is(TT_TemplateCloser) && Left.is(TT_TemplateCloser) && 3640 ((Style.Standard < FormatStyle::LS_Cpp11) || 3641 ShouldAddSpacesInAngles()); 3642 } 3643 if (Right.isOneOf(tok::arrow, tok::arrowstar, tok::periodstar) || 3644 Left.isOneOf(tok::arrow, tok::period, tok::arrowstar, tok::periodstar) || 3645 (Right.is(tok::period) && Right.isNot(TT_DesignatedInitializerPeriod))) 3646 return false; 3647 if (!Style.SpaceBeforeAssignmentOperators && Left.isNot(TT_TemplateCloser) && 3648 Right.getPrecedence() == prec::Assignment) 3649 return false; 3650 if (Style.Language == FormatStyle::LK_Java && Right.is(tok::coloncolon) && 3651 (Left.is(tok::identifier) || Left.is(tok::kw_this))) 3652 return false; 3653 if (Right.is(tok::coloncolon) && Left.is(tok::identifier)) 3654 // Generally don't remove existing spaces between an identifier and "::". 3655 // The identifier might actually be a macro name such as ALWAYS_INLINE. If 3656 // this turns out to be too lenient, add analysis of the identifier itself. 3657 return Right.hasWhitespaceBefore(); 3658 if (Right.is(tok::coloncolon) && 3659 !Left.isOneOf(tok::l_brace, tok::comment, tok::l_paren)) 3660 // Put a space between < and :: in vector< ::std::string > 3661 return (Left.is(TT_TemplateOpener) && 3662 ((Style.Standard < FormatStyle::LS_Cpp11) || 3663 ShouldAddSpacesInAngles())) || 3664 !(Left.isOneOf(tok::l_paren, tok::r_paren, tok::l_square, 3665 tok::kw___super, TT_TemplateOpener, 3666 TT_TemplateCloser)) || 3667 (Left.is(tok::l_paren) && Style.SpacesInParentheses); 3668 if ((Left.is(TT_TemplateOpener)) != (Right.is(TT_TemplateCloser))) 3669 return ShouldAddSpacesInAngles(); 3670 // Space before TT_StructuredBindingLSquare. 3671 if (Right.is(TT_StructuredBindingLSquare)) 3672 return !Left.isOneOf(tok::amp, tok::ampamp) || 3673 getTokenReferenceAlignment(Left) != FormatStyle::PAS_Right; 3674 // Space before & or && following a TT_StructuredBindingLSquare. 3675 if (Right.Next && Right.Next->is(TT_StructuredBindingLSquare) && 3676 Right.isOneOf(tok::amp, tok::ampamp)) 3677 return getTokenReferenceAlignment(Right) != FormatStyle::PAS_Left; 3678 if ((Right.is(TT_BinaryOperator) && !Left.is(tok::l_paren)) || 3679 (Left.isOneOf(TT_BinaryOperator, TT_ConditionalExpr) && 3680 !Right.is(tok::r_paren))) 3681 return true; 3682 if (Right.is(TT_TemplateOpener) && Left.is(tok::r_paren) && 3683 Left.MatchingParen && Left.MatchingParen->is(TT_OverloadedOperatorLParen)) 3684 return false; 3685 if (Right.is(tok::less) && Left.isNot(tok::l_paren) && 3686 Line.startsWith(tok::hash)) 3687 return true; 3688 if (Right.is(TT_TrailingUnaryOperator)) 3689 return false; 3690 if (Left.is(TT_RegexLiteral)) 3691 return false; 3692 return spaceRequiredBetween(Line, Left, Right); 3693 } 3694 3695 // Returns 'true' if 'Tok' is a brace we'd want to break before in Allman style. 3696 static bool isAllmanBrace(const FormatToken &Tok) { 3697 return Tok.is(tok::l_brace) && Tok.is(BK_Block) && 3698 !Tok.isOneOf(TT_ObjCBlockLBrace, TT_LambdaLBrace, TT_DictLiteral); 3699 } 3700 3701 // Returns 'true' if 'Tok' is a function argument. 3702 static bool IsFunctionArgument(const FormatToken &Tok) { 3703 return Tok.MatchingParen && Tok.MatchingParen->Next && 3704 Tok.MatchingParen->Next->isOneOf(tok::comma, tok::r_paren); 3705 } 3706 3707 static bool 3708 isItAnEmptyLambdaAllowed(const FormatToken &Tok, 3709 FormatStyle::ShortLambdaStyle ShortLambdaOption) { 3710 return Tok.Children.empty() && ShortLambdaOption != FormatStyle::SLS_None; 3711 } 3712 3713 static bool isAllmanLambdaBrace(const FormatToken &Tok) { 3714 return (Tok.is(tok::l_brace) && Tok.is(BK_Block) && 3715 !Tok.isOneOf(TT_ObjCBlockLBrace, TT_DictLiteral)); 3716 } 3717 3718 // Returns the first token on the line that is not a comment. 3719 static const FormatToken *getFirstNonComment(const AnnotatedLine &Line) { 3720 const FormatToken *Next = Line.First; 3721 if (!Next) 3722 return Next; 3723 if (Next->is(tok::comment)) 3724 Next = Next->getNextNonComment(); 3725 return Next; 3726 } 3727 3728 bool TokenAnnotator::mustBreakBefore(const AnnotatedLine &Line, 3729 const FormatToken &Right) { 3730 const FormatToken &Left = *Right.Previous; 3731 if (Right.NewlinesBefore > 1 && Style.MaxEmptyLinesToKeep > 0) 3732 return true; 3733 3734 if (Style.isCSharp()) { 3735 if (Left.is(TT_FatArrow) && Right.is(tok::l_brace) && 3736 Style.BraceWrapping.AfterFunction) 3737 return true; 3738 if (Right.is(TT_CSharpNamedArgumentColon) || 3739 Left.is(TT_CSharpNamedArgumentColon)) 3740 return false; 3741 if (Right.is(TT_CSharpGenericTypeConstraint)) 3742 return true; 3743 if (Right.Next && Right.Next->is(TT_FatArrow) && 3744 (Right.is(tok::numeric_constant) || 3745 (Right.is(tok::identifier) && Right.TokenText == "_"))) 3746 return true; 3747 3748 // Break after C# [...] and before public/protected/private/internal. 3749 if (Left.is(TT_AttributeSquare) && Left.is(tok::r_square) && 3750 (Right.isAccessSpecifier(/*ColonRequired=*/false) || 3751 Right.is(Keywords.kw_internal))) 3752 return true; 3753 // Break between ] and [ but only when there are really 2 attributes. 3754 if (Left.is(TT_AttributeSquare) && Right.is(TT_AttributeSquare) && 3755 Left.is(tok::r_square) && Right.is(tok::l_square)) 3756 return true; 3757 3758 } else if (Style.isJavaScript()) { 3759 // FIXME: This might apply to other languages and token kinds. 3760 if (Right.is(tok::string_literal) && Left.is(tok::plus) && Left.Previous && 3761 Left.Previous->is(tok::string_literal)) 3762 return true; 3763 if (Left.is(TT_DictLiteral) && Left.is(tok::l_brace) && Line.Level == 0 && 3764 Left.Previous && Left.Previous->is(tok::equal) && 3765 Line.First->isOneOf(tok::identifier, Keywords.kw_import, tok::kw_export, 3766 tok::kw_const) && 3767 // kw_var/kw_let are pseudo-tokens that are tok::identifier, so match 3768 // above. 3769 !Line.First->isOneOf(Keywords.kw_var, Keywords.kw_let)) 3770 // Object literals on the top level of a file are treated as "enum-style". 3771 // Each key/value pair is put on a separate line, instead of bin-packing. 3772 return true; 3773 if (Left.is(tok::l_brace) && Line.Level == 0 && 3774 (Line.startsWith(tok::kw_enum) || 3775 Line.startsWith(tok::kw_const, tok::kw_enum) || 3776 Line.startsWith(tok::kw_export, tok::kw_enum) || 3777 Line.startsWith(tok::kw_export, tok::kw_const, tok::kw_enum))) 3778 // JavaScript top-level enum key/value pairs are put on separate lines 3779 // instead of bin-packing. 3780 return true; 3781 if (Right.is(tok::r_brace) && Left.is(tok::l_brace) && Left.Previous && 3782 Left.Previous->is(TT_FatArrow)) { 3783 // JS arrow function (=> {...}). 3784 switch (Style.AllowShortLambdasOnASingleLine) { 3785 case FormatStyle::SLS_All: 3786 return false; 3787 case FormatStyle::SLS_None: 3788 return true; 3789 case FormatStyle::SLS_Empty: 3790 return !Left.Children.empty(); 3791 case FormatStyle::SLS_Inline: 3792 // allow one-lining inline (e.g. in function call args) and empty arrow 3793 // functions. 3794 return (Left.NestingLevel == 0 && Line.Level == 0) && 3795 !Left.Children.empty(); 3796 } 3797 llvm_unreachable("Unknown FormatStyle::ShortLambdaStyle enum"); 3798 } 3799 3800 if (Right.is(tok::r_brace) && Left.is(tok::l_brace) && 3801 !Left.Children.empty()) 3802 // Support AllowShortFunctionsOnASingleLine for JavaScript. 3803 return Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_None || 3804 Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_Empty || 3805 (Left.NestingLevel == 0 && Line.Level == 0 && 3806 Style.AllowShortFunctionsOnASingleLine & 3807 FormatStyle::SFS_InlineOnly); 3808 } else if (Style.Language == FormatStyle::LK_Java) { 3809 if (Right.is(tok::plus) && Left.is(tok::string_literal) && Right.Next && 3810 Right.Next->is(tok::string_literal)) 3811 return true; 3812 } else if (Style.Language == FormatStyle::LK_Cpp || 3813 Style.Language == FormatStyle::LK_ObjC || 3814 Style.Language == FormatStyle::LK_Proto || 3815 Style.Language == FormatStyle::LK_TableGen || 3816 Style.Language == FormatStyle::LK_TextProto) { 3817 if (Left.isStringLiteral() && Right.isStringLiteral()) 3818 return true; 3819 } 3820 3821 // Basic JSON newline processing. 3822 if (Style.isJson()) { 3823 // Always break after a JSON record opener. 3824 // { 3825 // } 3826 if (Left.is(TT_DictLiteral) && Left.is(tok::l_brace)) 3827 return true; 3828 // Always break after a JSON array opener. 3829 // [ 3830 // ] 3831 if (Left.is(TT_ArrayInitializerLSquare) && Left.is(tok::l_square) && 3832 !Right.is(tok::r_square)) 3833 return true; 3834 // Always break after successive entries. 3835 // 1, 3836 // 2 3837 if (Left.is(tok::comma)) 3838 return true; 3839 } 3840 3841 // If the last token before a '}', ']', or ')' is a comma or a trailing 3842 // comment, the intention is to insert a line break after it in order to make 3843 // shuffling around entries easier. Import statements, especially in 3844 // JavaScript, can be an exception to this rule. 3845 if (Style.JavaScriptWrapImports || Line.Type != LT_ImportStatement) { 3846 const FormatToken *BeforeClosingBrace = nullptr; 3847 if ((Left.isOneOf(tok::l_brace, TT_ArrayInitializerLSquare) || 3848 (Style.isJavaScript() && Left.is(tok::l_paren))) && 3849 Left.isNot(BK_Block) && Left.MatchingParen) 3850 BeforeClosingBrace = Left.MatchingParen->Previous; 3851 else if (Right.MatchingParen && 3852 (Right.MatchingParen->isOneOf(tok::l_brace, 3853 TT_ArrayInitializerLSquare) || 3854 (Style.isJavaScript() && Right.MatchingParen->is(tok::l_paren)))) 3855 BeforeClosingBrace = &Left; 3856 if (BeforeClosingBrace && (BeforeClosingBrace->is(tok::comma) || 3857 BeforeClosingBrace->isTrailingComment())) 3858 return true; 3859 } 3860 3861 if (Right.is(tok::comment)) 3862 return Left.isNot(BK_BracedInit) && Left.isNot(TT_CtorInitializerColon) && 3863 (Right.NewlinesBefore > 0 && Right.HasUnescapedNewline); 3864 if (Left.isTrailingComment()) 3865 return true; 3866 if (Left.IsUnterminatedLiteral) 3867 return true; 3868 if (Right.is(tok::lessless) && Right.Next && Left.is(tok::string_literal) && 3869 Right.Next->is(tok::string_literal)) 3870 return true; 3871 // Can break after template<> declaration 3872 if (Left.ClosesTemplateDeclaration && Left.MatchingParen && 3873 Left.MatchingParen->NestingLevel == 0) { 3874 // Put concepts on the next line e.g. 3875 // template<typename T> 3876 // concept ... 3877 if (Right.is(tok::kw_concept)) 3878 return Style.BreakBeforeConceptDeclarations; 3879 return (Style.AlwaysBreakTemplateDeclarations == FormatStyle::BTDS_Yes); 3880 } 3881 if (Style.PackConstructorInitializers == FormatStyle::PCIS_Never) { 3882 if (Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeColon && 3883 (Left.is(TT_CtorInitializerComma) || Right.is(TT_CtorInitializerColon))) 3884 return true; 3885 3886 if (Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon && 3887 Left.isOneOf(TT_CtorInitializerColon, TT_CtorInitializerComma)) 3888 return true; 3889 } 3890 if (Style.PackConstructorInitializers < FormatStyle::PCIS_CurrentLine && 3891 Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma && 3892 Right.isOneOf(TT_CtorInitializerComma, TT_CtorInitializerColon)) 3893 return true; 3894 // Break only if we have multiple inheritance. 3895 if (Style.BreakInheritanceList == FormatStyle::BILS_BeforeComma && 3896 Right.is(TT_InheritanceComma)) 3897 return true; 3898 if (Style.BreakInheritanceList == FormatStyle::BILS_AfterComma && 3899 Left.is(TT_InheritanceComma)) 3900 return true; 3901 if (Right.is(tok::string_literal) && Right.TokenText.startswith("R\"")) 3902 // Multiline raw string literals are special wrt. line breaks. The author 3903 // has made a deliberate choice and might have aligned the contents of the 3904 // string literal accordingly. Thus, we try keep existing line breaks. 3905 return Right.IsMultiline && Right.NewlinesBefore > 0; 3906 if ((Left.is(tok::l_brace) || (Left.is(tok::less) && Left.Previous && 3907 Left.Previous->is(tok::equal))) && 3908 Right.NestingLevel == 1 && Style.Language == FormatStyle::LK_Proto) { 3909 // Don't put enums or option definitions onto single lines in protocol 3910 // buffers. 3911 return true; 3912 } 3913 if (Right.is(TT_InlineASMBrace)) 3914 return Right.HasUnescapedNewline; 3915 3916 if (isAllmanBrace(Left) || isAllmanBrace(Right)) { 3917 auto FirstNonComment = getFirstNonComment(Line); 3918 bool AccessSpecifier = 3919 FirstNonComment && 3920 FirstNonComment->isOneOf(Keywords.kw_internal, tok::kw_public, 3921 tok::kw_private, tok::kw_protected); 3922 3923 if (Style.BraceWrapping.AfterEnum) { 3924 if (Line.startsWith(tok::kw_enum) || 3925 Line.startsWith(tok::kw_typedef, tok::kw_enum)) 3926 return true; 3927 // Ensure BraceWrapping for `public enum A {`. 3928 if (AccessSpecifier && FirstNonComment->Next && 3929 FirstNonComment->Next->is(tok::kw_enum)) 3930 return true; 3931 } 3932 3933 // Ensure BraceWrapping for `public interface A {`. 3934 if (Style.BraceWrapping.AfterClass && 3935 ((AccessSpecifier && FirstNonComment->Next && 3936 FirstNonComment->Next->is(Keywords.kw_interface)) || 3937 Line.startsWith(Keywords.kw_interface))) 3938 return true; 3939 3940 return (Line.startsWith(tok::kw_class) && Style.BraceWrapping.AfterClass) || 3941 (Line.startsWith(tok::kw_struct) && Style.BraceWrapping.AfterStruct); 3942 } 3943 3944 if (Left.is(TT_ObjCBlockLBrace) && 3945 Style.AllowShortBlocksOnASingleLine == FormatStyle::SBS_Never) 3946 return true; 3947 3948 // Ensure wrapping after __attribute__((XX)) and @interface etc. 3949 if (Left.is(TT_AttributeParen) && Right.is(TT_ObjCDecl)) 3950 return true; 3951 3952 if (Left.is(TT_LambdaLBrace)) { 3953 if (IsFunctionArgument(Left) && 3954 Style.AllowShortLambdasOnASingleLine == FormatStyle::SLS_Inline) 3955 return false; 3956 3957 if (Style.AllowShortLambdasOnASingleLine == FormatStyle::SLS_None || 3958 Style.AllowShortLambdasOnASingleLine == FormatStyle::SLS_Inline || 3959 (!Left.Children.empty() && 3960 Style.AllowShortLambdasOnASingleLine == FormatStyle::SLS_Empty)) 3961 return true; 3962 } 3963 3964 if (Style.BraceWrapping.BeforeLambdaBody && Right.is(TT_LambdaLBrace) && 3965 Left.isOneOf(tok::star, tok::amp, tok::ampamp, TT_TemplateCloser)) 3966 return true; 3967 3968 // Put multiple Java annotation on a new line. 3969 if ((Style.Language == FormatStyle::LK_Java || Style.isJavaScript()) && 3970 Left.is(TT_LeadingJavaAnnotation) && 3971 Right.isNot(TT_LeadingJavaAnnotation) && Right.isNot(tok::l_paren) && 3972 (Line.Last->is(tok::l_brace) || Style.BreakAfterJavaFieldAnnotations)) 3973 return true; 3974 3975 if (Right.is(TT_ProtoExtensionLSquare)) 3976 return true; 3977 3978 // In text proto instances if a submessage contains at least 2 entries and at 3979 // least one of them is a submessage, like A { ... B { ... } ... }, 3980 // put all of the entries of A on separate lines by forcing the selector of 3981 // the submessage B to be put on a newline. 3982 // 3983 // Example: these can stay on one line: 3984 // a { scalar_1: 1 scalar_2: 2 } 3985 // a { b { key: value } } 3986 // 3987 // and these entries need to be on a new line even if putting them all in one 3988 // line is under the column limit: 3989 // a { 3990 // scalar: 1 3991 // b { key: value } 3992 // } 3993 // 3994 // We enforce this by breaking before a submessage field that has previous 3995 // siblings, *and* breaking before a field that follows a submessage field. 3996 // 3997 // Be careful to exclude the case [proto.ext] { ... } since the `]` is 3998 // the TT_SelectorName there, but we don't want to break inside the brackets. 3999 // 4000 // Another edge case is @submessage { key: value }, which is a common 4001 // substitution placeholder. In this case we want to keep `@` and `submessage` 4002 // together. 4003 // 4004 // We ensure elsewhere that extensions are always on their own line. 4005 if ((Style.Language == FormatStyle::LK_Proto || 4006 Style.Language == FormatStyle::LK_TextProto) && 4007 Right.is(TT_SelectorName) && !Right.is(tok::r_square) && Right.Next) { 4008 // Keep `@submessage` together in: 4009 // @submessage { key: value } 4010 if (Left.is(tok::at)) 4011 return false; 4012 // Look for the scope opener after selector in cases like: 4013 // selector { ... 4014 // selector: { ... 4015 // selector: @base { ... 4016 FormatToken *LBrace = Right.Next; 4017 if (LBrace && LBrace->is(tok::colon)) { 4018 LBrace = LBrace->Next; 4019 if (LBrace && LBrace->is(tok::at)) { 4020 LBrace = LBrace->Next; 4021 if (LBrace) 4022 LBrace = LBrace->Next; 4023 } 4024 } 4025 if (LBrace && 4026 // The scope opener is one of {, [, <: 4027 // selector { ... } 4028 // selector [ ... ] 4029 // selector < ... > 4030 // 4031 // In case of selector { ... }, the l_brace is TT_DictLiteral. 4032 // In case of an empty selector {}, the l_brace is not TT_DictLiteral, 4033 // so we check for immediately following r_brace. 4034 ((LBrace->is(tok::l_brace) && 4035 (LBrace->is(TT_DictLiteral) || 4036 (LBrace->Next && LBrace->Next->is(tok::r_brace)))) || 4037 LBrace->is(TT_ArrayInitializerLSquare) || LBrace->is(tok::less))) { 4038 // If Left.ParameterCount is 0, then this submessage entry is not the 4039 // first in its parent submessage, and we want to break before this entry. 4040 // If Left.ParameterCount is greater than 0, then its parent submessage 4041 // might contain 1 or more entries and we want to break before this entry 4042 // if it contains at least 2 entries. We deal with this case later by 4043 // detecting and breaking before the next entry in the parent submessage. 4044 if (Left.ParameterCount == 0) 4045 return true; 4046 // However, if this submessage is the first entry in its parent 4047 // submessage, Left.ParameterCount might be 1 in some cases. 4048 // We deal with this case later by detecting an entry 4049 // following a closing paren of this submessage. 4050 } 4051 4052 // If this is an entry immediately following a submessage, it will be 4053 // preceded by a closing paren of that submessage, like in: 4054 // left---. .---right 4055 // v v 4056 // sub: { ... } key: value 4057 // If there was a comment between `}` an `key` above, then `key` would be 4058 // put on a new line anyways. 4059 if (Left.isOneOf(tok::r_brace, tok::greater, tok::r_square)) 4060 return true; 4061 } 4062 4063 // Deal with lambda arguments in C++ - we want consistent line breaks whether 4064 // they happen to be at arg0, arg1 or argN. The selection is a bit nuanced 4065 // as aggressive line breaks are placed when the lambda is not the last arg. 4066 if ((Style.Language == FormatStyle::LK_Cpp || 4067 Style.Language == FormatStyle::LK_ObjC) && 4068 Left.is(tok::l_paren) && Left.BlockParameterCount > 0 && 4069 !Right.isOneOf(tok::l_paren, TT_LambdaLSquare)) { 4070 // Multiple lambdas in the same function call force line breaks. 4071 if (Left.BlockParameterCount > 1) 4072 return true; 4073 4074 // A lambda followed by another arg forces a line break. 4075 if (!Left.Role) 4076 return false; 4077 auto Comma = Left.Role->lastComma(); 4078 if (!Comma) 4079 return false; 4080 auto Next = Comma->getNextNonComment(); 4081 if (!Next) 4082 return false; 4083 if (!Next->isOneOf(TT_LambdaLSquare, tok::l_brace, tok::caret)) 4084 return true; 4085 } 4086 4087 return false; 4088 } 4089 4090 bool TokenAnnotator::canBreakBefore(const AnnotatedLine &Line, 4091 const FormatToken &Right) { 4092 const FormatToken &Left = *Right.Previous; 4093 // Language-specific stuff. 4094 if (Style.isCSharp()) { 4095 if (Left.isOneOf(TT_CSharpNamedArgumentColon, TT_AttributeColon) || 4096 Right.isOneOf(TT_CSharpNamedArgumentColon, TT_AttributeColon)) 4097 return false; 4098 // Only break after commas for generic type constraints. 4099 if (Line.First->is(TT_CSharpGenericTypeConstraint)) 4100 return Left.is(TT_CSharpGenericTypeConstraintComma); 4101 // Keep nullable operators attached to their identifiers. 4102 if (Right.is(TT_CSharpNullable)) 4103 return false; 4104 } else if (Style.Language == FormatStyle::LK_Java) { 4105 if (Left.isOneOf(Keywords.kw_throws, Keywords.kw_extends, 4106 Keywords.kw_implements)) 4107 return false; 4108 if (Right.isOneOf(Keywords.kw_throws, Keywords.kw_extends, 4109 Keywords.kw_implements)) 4110 return true; 4111 } else if (Style.isJavaScript()) { 4112 const FormatToken *NonComment = Right.getPreviousNonComment(); 4113 if (NonComment && 4114 NonComment->isOneOf( 4115 tok::kw_return, Keywords.kw_yield, tok::kw_continue, tok::kw_break, 4116 tok::kw_throw, Keywords.kw_interface, Keywords.kw_type, 4117 tok::kw_static, tok::kw_public, tok::kw_private, tok::kw_protected, 4118 Keywords.kw_readonly, Keywords.kw_override, Keywords.kw_abstract, 4119 Keywords.kw_get, Keywords.kw_set, Keywords.kw_async, 4120 Keywords.kw_await)) 4121 return false; // Otherwise automatic semicolon insertion would trigger. 4122 if (Right.NestingLevel == 0 && 4123 (Left.Tok.getIdentifierInfo() || 4124 Left.isOneOf(tok::r_square, tok::r_paren)) && 4125 Right.isOneOf(tok::l_square, tok::l_paren)) 4126 return false; // Otherwise automatic semicolon insertion would trigger. 4127 if (NonComment && NonComment->is(tok::identifier) && 4128 NonComment->TokenText == "asserts") 4129 return false; 4130 if (Left.is(TT_FatArrow) && Right.is(tok::l_brace)) 4131 return false; 4132 if (Left.is(TT_JsTypeColon)) 4133 return true; 4134 // Don't wrap between ":" and "!" of a strict prop init ("field!: type;"). 4135 if (Left.is(tok::exclaim) && Right.is(tok::colon)) 4136 return false; 4137 // Look for is type annotations like: 4138 // function f(): a is B { ... } 4139 // Do not break before is in these cases. 4140 if (Right.is(Keywords.kw_is)) { 4141 const FormatToken *Next = Right.getNextNonComment(); 4142 // If `is` is followed by a colon, it's likely that it's a dict key, so 4143 // ignore it for this check. 4144 // For example this is common in Polymer: 4145 // Polymer({ 4146 // is: 'name', 4147 // ... 4148 // }); 4149 if (!Next || !Next->is(tok::colon)) 4150 return false; 4151 } 4152 if (Left.is(Keywords.kw_in)) 4153 return Style.BreakBeforeBinaryOperators == FormatStyle::BOS_None; 4154 if (Right.is(Keywords.kw_in)) 4155 return Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None; 4156 if (Right.is(Keywords.kw_as)) 4157 return false; // must not break before as in 'x as type' casts 4158 if (Right.isOneOf(Keywords.kw_extends, Keywords.kw_infer)) { 4159 // extends and infer can appear as keywords in conditional types: 4160 // https://www.typescriptlang.org/docs/handbook/release-notes/typescript-2-8.html#conditional-types 4161 // do not break before them, as the expressions are subject to ASI. 4162 return false; 4163 } 4164 if (Left.is(Keywords.kw_as)) 4165 return true; 4166 if (Left.is(TT_NonNullAssertion)) 4167 return true; 4168 if (Left.is(Keywords.kw_declare) && 4169 Right.isOneOf(Keywords.kw_module, tok::kw_namespace, 4170 Keywords.kw_function, tok::kw_class, tok::kw_enum, 4171 Keywords.kw_interface, Keywords.kw_type, Keywords.kw_var, 4172 Keywords.kw_let, tok::kw_const)) 4173 // See grammar for 'declare' statements at: 4174 // https://github.com/Microsoft/TypeScript/blob/main/doc/spec-ARCHIVED.md#A.10 4175 return false; 4176 if (Left.isOneOf(Keywords.kw_module, tok::kw_namespace) && 4177 Right.isOneOf(tok::identifier, tok::string_literal)) 4178 return false; // must not break in "module foo { ...}" 4179 if (Right.is(TT_TemplateString) && Right.closesScope()) 4180 return false; 4181 // Don't split tagged template literal so there is a break between the tag 4182 // identifier and template string. 4183 if (Left.is(tok::identifier) && Right.is(TT_TemplateString)) 4184 return false; 4185 if (Left.is(TT_TemplateString) && Left.opensScope()) 4186 return true; 4187 } 4188 4189 if (Left.is(tok::at)) 4190 return false; 4191 if (Left.Tok.getObjCKeywordID() == tok::objc_interface) 4192 return false; 4193 if (Left.isOneOf(TT_JavaAnnotation, TT_LeadingJavaAnnotation)) 4194 return !Right.is(tok::l_paren); 4195 if (Right.is(TT_PointerOrReference)) 4196 return Line.IsMultiVariableDeclStmt || 4197 (getTokenPointerOrReferenceAlignment(Right) == 4198 FormatStyle::PAS_Right && 4199 (!Right.Next || Right.Next->isNot(TT_FunctionDeclarationName))); 4200 if (Right.isOneOf(TT_StartOfName, TT_FunctionDeclarationName) || 4201 Right.is(tok::kw_operator)) 4202 return true; 4203 if (Left.is(TT_PointerOrReference)) 4204 return false; 4205 if (Right.isTrailingComment()) 4206 // We rely on MustBreakBefore being set correctly here as we should not 4207 // change the "binding" behavior of a comment. 4208 // The first comment in a braced lists is always interpreted as belonging to 4209 // the first list element. Otherwise, it should be placed outside of the 4210 // list. 4211 return Left.is(BK_BracedInit) || 4212 (Left.is(TT_CtorInitializerColon) && 4213 Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon); 4214 if (Left.is(tok::question) && Right.is(tok::colon)) 4215 return false; 4216 if (Right.is(TT_ConditionalExpr) || Right.is(tok::question)) 4217 return Style.BreakBeforeTernaryOperators; 4218 if (Left.is(TT_ConditionalExpr) || Left.is(tok::question)) 4219 return !Style.BreakBeforeTernaryOperators; 4220 if (Left.is(TT_InheritanceColon)) 4221 return Style.BreakInheritanceList == FormatStyle::BILS_AfterColon; 4222 if (Right.is(TT_InheritanceColon)) 4223 return Style.BreakInheritanceList != FormatStyle::BILS_AfterColon; 4224 if (Right.is(TT_ObjCMethodExpr) && !Right.is(tok::r_square) && 4225 Left.isNot(TT_SelectorName)) 4226 return true; 4227 4228 if (Right.is(tok::colon) && 4229 !Right.isOneOf(TT_CtorInitializerColon, TT_InlineASMColon)) 4230 return false; 4231 if (Left.is(tok::colon) && Left.isOneOf(TT_DictLiteral, TT_ObjCMethodExpr)) { 4232 if (Style.Language == FormatStyle::LK_Proto || 4233 Style.Language == FormatStyle::LK_TextProto) { 4234 if (!Style.AlwaysBreakBeforeMultilineStrings && Right.isStringLiteral()) 4235 return false; 4236 // Prevent cases like: 4237 // 4238 // submessage: 4239 // { key: valueeeeeeeeeeee } 4240 // 4241 // when the snippet does not fit into one line. 4242 // Prefer: 4243 // 4244 // submessage: { 4245 // key: valueeeeeeeeeeee 4246 // } 4247 // 4248 // instead, even if it is longer by one line. 4249 // 4250 // Note that this allows allows the "{" to go over the column limit 4251 // when the column limit is just between ":" and "{", but that does 4252 // not happen too often and alternative formattings in this case are 4253 // not much better. 4254 // 4255 // The code covers the cases: 4256 // 4257 // submessage: { ... } 4258 // submessage: < ... > 4259 // repeated: [ ... ] 4260 if (((Right.is(tok::l_brace) || Right.is(tok::less)) && 4261 Right.is(TT_DictLiteral)) || 4262 Right.is(TT_ArrayInitializerLSquare)) 4263 return false; 4264 } 4265 return true; 4266 } 4267 if (Right.is(tok::r_square) && Right.MatchingParen && 4268 Right.MatchingParen->is(TT_ProtoExtensionLSquare)) 4269 return false; 4270 if (Right.is(TT_SelectorName) || (Right.is(tok::identifier) && Right.Next && 4271 Right.Next->is(TT_ObjCMethodExpr))) 4272 return Left.isNot(tok::period); // FIXME: Properly parse ObjC calls. 4273 if (Left.is(tok::r_paren) && Line.Type == LT_ObjCProperty) 4274 return true; 4275 if (Left.ClosesTemplateDeclaration || Left.is(TT_FunctionAnnotationRParen)) 4276 return true; 4277 if (Right.isOneOf(TT_RangeBasedForLoopColon, TT_OverloadedOperatorLParen, 4278 TT_OverloadedOperator)) 4279 return false; 4280 if (Left.is(TT_RangeBasedForLoopColon)) 4281 return true; 4282 if (Right.is(TT_RangeBasedForLoopColon)) 4283 return false; 4284 if (Left.is(TT_TemplateCloser) && Right.is(TT_TemplateOpener)) 4285 return true; 4286 if (Left.isOneOf(TT_TemplateCloser, TT_UnaryOperator) || 4287 Left.is(tok::kw_operator)) 4288 return false; 4289 if (Left.is(tok::equal) && !Right.isOneOf(tok::kw_default, tok::kw_delete) && 4290 Line.Type == LT_VirtualFunctionDecl && Left.NestingLevel == 0) 4291 return false; 4292 if (Left.is(tok::equal) && Right.is(tok::l_brace) && 4293 !Style.Cpp11BracedListStyle) 4294 return false; 4295 if (Left.is(tok::l_paren) && 4296 Left.isOneOf(TT_AttributeParen, TT_TypeDeclarationParen)) 4297 return false; 4298 if (Left.is(tok::l_paren) && Left.Previous && 4299 (Left.Previous->isOneOf(TT_BinaryOperator, TT_CastRParen))) 4300 return false; 4301 if (Right.is(TT_ImplicitStringLiteral)) 4302 return false; 4303 4304 if (Right.is(TT_TemplateCloser)) 4305 return false; 4306 if (Right.is(tok::r_square) && Right.MatchingParen && 4307 Right.MatchingParen->is(TT_LambdaLSquare)) 4308 return false; 4309 4310 // We only break before r_brace if there was a corresponding break before 4311 // the l_brace, which is tracked by BreakBeforeClosingBrace. 4312 if (Right.is(tok::r_brace)) 4313 return Right.MatchingParen && Right.MatchingParen->is(BK_Block); 4314 4315 // We only break before r_paren if we're in a block indented context. 4316 if (Right.is(tok::r_paren)) { 4317 if (Style.AlignAfterOpenBracket == FormatStyle::BAS_BlockIndent) { 4318 return Right.MatchingParen && 4319 !(Right.MatchingParen->Previous && 4320 (Right.MatchingParen->Previous->is(tok::kw_for) || 4321 Right.MatchingParen->Previous->isIf())); 4322 } 4323 4324 return false; 4325 } 4326 4327 // Allow breaking after a trailing annotation, e.g. after a method 4328 // declaration. 4329 if (Left.is(TT_TrailingAnnotation)) 4330 return !Right.isOneOf(tok::l_brace, tok::semi, tok::equal, tok::l_paren, 4331 tok::less, tok::coloncolon); 4332 4333 if (Right.is(tok::kw___attribute) || 4334 (Right.is(tok::l_square) && Right.is(TT_AttributeSquare))) 4335 return !Left.is(TT_AttributeSquare); 4336 4337 if (Left.is(tok::identifier) && Right.is(tok::string_literal)) 4338 return true; 4339 4340 if (Right.is(tok::identifier) && Right.Next && Right.Next->is(TT_DictLiteral)) 4341 return true; 4342 4343 if (Left.is(TT_CtorInitializerColon)) 4344 return Style.BreakConstructorInitializers == FormatStyle::BCIS_AfterColon; 4345 if (Right.is(TT_CtorInitializerColon)) 4346 return Style.BreakConstructorInitializers != FormatStyle::BCIS_AfterColon; 4347 if (Left.is(TT_CtorInitializerComma) && 4348 Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma) 4349 return false; 4350 if (Right.is(TT_CtorInitializerComma) && 4351 Style.BreakConstructorInitializers == FormatStyle::BCIS_BeforeComma) 4352 return true; 4353 if (Left.is(TT_InheritanceComma) && 4354 Style.BreakInheritanceList == FormatStyle::BILS_BeforeComma) 4355 return false; 4356 if (Right.is(TT_InheritanceComma) && 4357 Style.BreakInheritanceList == FormatStyle::BILS_BeforeComma) 4358 return true; 4359 if ((Left.is(tok::greater) && Right.is(tok::greater)) || 4360 (Left.is(tok::less) && Right.is(tok::less))) 4361 return false; 4362 if (Right.is(TT_BinaryOperator) && 4363 Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None && 4364 (Style.BreakBeforeBinaryOperators == FormatStyle::BOS_All || 4365 Right.getPrecedence() != prec::Assignment)) 4366 return true; 4367 if (Left.is(TT_ArrayInitializerLSquare)) 4368 return true; 4369 if (Right.is(tok::kw_typename) && Left.isNot(tok::kw_const)) 4370 return true; 4371 if ((Left.isBinaryOperator() || Left.is(TT_BinaryOperator)) && 4372 !Left.isOneOf(tok::arrowstar, tok::lessless) && 4373 Style.BreakBeforeBinaryOperators != FormatStyle::BOS_All && 4374 (Style.BreakBeforeBinaryOperators == FormatStyle::BOS_None || 4375 Left.getPrecedence() == prec::Assignment)) 4376 return true; 4377 if ((Left.is(TT_AttributeSquare) && Right.is(tok::l_square)) || 4378 (Left.is(tok::r_square) && Right.is(TT_AttributeSquare))) 4379 return false; 4380 4381 auto ShortLambdaOption = Style.AllowShortLambdasOnASingleLine; 4382 if (Style.BraceWrapping.BeforeLambdaBody && Right.is(TT_LambdaLBrace)) { 4383 if (isAllmanLambdaBrace(Left)) 4384 return !isItAnEmptyLambdaAllowed(Left, ShortLambdaOption); 4385 if (isAllmanLambdaBrace(Right)) 4386 return !isItAnEmptyLambdaAllowed(Right, ShortLambdaOption); 4387 } 4388 4389 return Left.isOneOf(tok::comma, tok::coloncolon, tok::semi, tok::l_brace, 4390 tok::kw_class, tok::kw_struct, tok::comment) || 4391 Right.isMemberAccess() || 4392 Right.isOneOf(TT_TrailingReturnArrow, TT_LambdaArrow, tok::lessless, 4393 tok::colon, tok::l_square, tok::at) || 4394 (Left.is(tok::r_paren) && 4395 Right.isOneOf(tok::identifier, tok::kw_const)) || 4396 (Left.is(tok::l_paren) && !Right.is(tok::r_paren)) || 4397 (Left.is(TT_TemplateOpener) && !Right.is(TT_TemplateCloser)); 4398 } 4399 4400 void TokenAnnotator::printDebugInfo(const AnnotatedLine &Line) { 4401 llvm::errs() << "AnnotatedTokens(L=" << Line.Level << "):\n"; 4402 const FormatToken *Tok = Line.First; 4403 while (Tok) { 4404 llvm::errs() << " M=" << Tok->MustBreakBefore 4405 << " C=" << Tok->CanBreakBefore 4406 << " T=" << getTokenTypeName(Tok->getType()) 4407 << " S=" << Tok->SpacesRequiredBefore 4408 << " F=" << Tok->Finalized << " B=" << Tok->BlockParameterCount 4409 << " BK=" << Tok->getBlockKind() << " P=" << Tok->SplitPenalty 4410 << " Name=" << Tok->Tok.getName() << " L=" << Tok->TotalLength 4411 << " PPK=" << Tok->getPackingKind() << " FakeLParens="; 4412 for (unsigned i = 0, e = Tok->FakeLParens.size(); i != e; ++i) 4413 llvm::errs() << Tok->FakeLParens[i] << "/"; 4414 llvm::errs() << " FakeRParens=" << Tok->FakeRParens; 4415 llvm::errs() << " II=" << Tok->Tok.getIdentifierInfo(); 4416 llvm::errs() << " Text='" << Tok->TokenText << "'\n"; 4417 if (!Tok->Next) 4418 assert(Tok == Line.Last); 4419 Tok = Tok->Next; 4420 } 4421 llvm::errs() << "----\n"; 4422 } 4423 4424 FormatStyle::PointerAlignmentStyle 4425 TokenAnnotator::getTokenReferenceAlignment(const FormatToken &Reference) { 4426 assert(Reference.isOneOf(tok::amp, tok::ampamp)); 4427 switch (Style.ReferenceAlignment) { 4428 case FormatStyle::RAS_Pointer: 4429 return Style.PointerAlignment; 4430 case FormatStyle::RAS_Left: 4431 return FormatStyle::PAS_Left; 4432 case FormatStyle::RAS_Right: 4433 return FormatStyle::PAS_Right; 4434 case FormatStyle::RAS_Middle: 4435 return FormatStyle::PAS_Middle; 4436 } 4437 assert(0); //"Unhandled value of ReferenceAlignment" 4438 return Style.PointerAlignment; 4439 } 4440 4441 FormatStyle::PointerAlignmentStyle 4442 TokenAnnotator::getTokenPointerOrReferenceAlignment( 4443 const FormatToken &PointerOrReference) { 4444 if (PointerOrReference.isOneOf(tok::amp, tok::ampamp)) { 4445 switch (Style.ReferenceAlignment) { 4446 case FormatStyle::RAS_Pointer: 4447 return Style.PointerAlignment; 4448 case FormatStyle::RAS_Left: 4449 return FormatStyle::PAS_Left; 4450 case FormatStyle::RAS_Right: 4451 return FormatStyle::PAS_Right; 4452 case FormatStyle::RAS_Middle: 4453 return FormatStyle::PAS_Middle; 4454 } 4455 } 4456 assert(PointerOrReference.is(tok::star)); 4457 return Style.PointerAlignment; 4458 } 4459 4460 } // namespace format 4461 } // namespace clang 4462