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