1 //===--- FormatToken.h - Format C++ code ------------------------*- C++ -*-===// 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 contains the declaration of the FormatToken, a wrapper 11 /// around Token with additional information related to formatting. 12 /// 13 //===----------------------------------------------------------------------===// 14 15 #ifndef LLVM_CLANG_LIB_FORMAT_FORMATTOKEN_H 16 #define LLVM_CLANG_LIB_FORMAT_FORMATTOKEN_H 17 18 #include "clang/Basic/IdentifierTable.h" 19 #include "clang/Basic/OperatorPrecedence.h" 20 #include "clang/Format/Format.h" 21 #include "clang/Lex/Lexer.h" 22 #include <memory> 23 #include <unordered_set> 24 25 namespace clang { 26 namespace format { 27 28 #define LIST_TOKEN_TYPES \ 29 TYPE(ArrayInitializerLSquare) \ 30 TYPE(ArraySubscriptLSquare) \ 31 TYPE(AttributeColon) \ 32 TYPE(AttributeMacro) \ 33 TYPE(AttributeParen) \ 34 TYPE(AttributeSquare) \ 35 TYPE(BinaryOperator) \ 36 TYPE(BitFieldColon) \ 37 TYPE(BlockComment) \ 38 TYPE(CastRParen) \ 39 TYPE(ConditionalExpr) \ 40 TYPE(ConflictAlternative) \ 41 TYPE(ConflictEnd) \ 42 TYPE(ConflictStart) \ 43 TYPE(ConstraintJunctions) \ 44 TYPE(CtorInitializerColon) \ 45 TYPE(CtorInitializerComma) \ 46 TYPE(DesignatedInitializerLSquare) \ 47 TYPE(DesignatedInitializerPeriod) \ 48 TYPE(DictLiteral) \ 49 TYPE(FatArrow) \ 50 TYPE(ForEachMacro) \ 51 TYPE(FunctionAnnotationRParen) \ 52 TYPE(FunctionDeclarationName) \ 53 TYPE(FunctionLBrace) \ 54 TYPE(FunctionTypeLParen) \ 55 TYPE(ImplicitStringLiteral) \ 56 TYPE(InheritanceColon) \ 57 TYPE(InheritanceComma) \ 58 TYPE(InlineASMBrace) \ 59 TYPE(InlineASMColon) \ 60 TYPE(InlineASMSymbolicNameLSquare) \ 61 TYPE(JavaAnnotation) \ 62 TYPE(JsComputedPropertyName) \ 63 TYPE(JsExponentiation) \ 64 TYPE(JsExponentiationEqual) \ 65 TYPE(JsPipePipeEqual) \ 66 TYPE(JsPrivateIdentifier) \ 67 TYPE(JsTypeColon) \ 68 TYPE(JsTypeOperator) \ 69 TYPE(JsTypeOptionalQuestion) \ 70 TYPE(JsAndAndEqual) \ 71 TYPE(LambdaArrow) \ 72 TYPE(LambdaLBrace) \ 73 TYPE(LambdaLSquare) \ 74 TYPE(LeadingJavaAnnotation) \ 75 TYPE(LineComment) \ 76 TYPE(MacroBlockBegin) \ 77 TYPE(MacroBlockEnd) \ 78 TYPE(NamespaceMacro) \ 79 TYPE(NonNullAssertion) \ 80 TYPE(NullCoalescingEqual) \ 81 TYPE(NullCoalescingOperator) \ 82 TYPE(NullPropagatingOperator) \ 83 TYPE(ObjCBlockLBrace) \ 84 TYPE(ObjCBlockLParen) \ 85 TYPE(ObjCDecl) \ 86 TYPE(ObjCForIn) \ 87 TYPE(ObjCMethodExpr) \ 88 TYPE(ObjCMethodSpecifier) \ 89 TYPE(ObjCProperty) \ 90 TYPE(ObjCStringLiteral) \ 91 TYPE(OverloadedOperator) \ 92 TYPE(OverloadedOperatorLParen) \ 93 TYPE(PointerOrReference) \ 94 TYPE(PureVirtualSpecifier) \ 95 TYPE(RangeBasedForLoopColon) \ 96 TYPE(RegexLiteral) \ 97 TYPE(SelectorName) \ 98 TYPE(StartOfName) \ 99 TYPE(StatementAttributeLikeMacro) \ 100 TYPE(StatementMacro) \ 101 TYPE(StructuredBindingLSquare) \ 102 TYPE(TemplateCloser) \ 103 TYPE(TemplateOpener) \ 104 TYPE(TemplateString) \ 105 TYPE(ProtoExtensionLSquare) \ 106 TYPE(TrailingAnnotation) \ 107 TYPE(TrailingReturnArrow) \ 108 TYPE(TrailingUnaryOperator) \ 109 TYPE(TypeDeclarationParen) \ 110 TYPE(TypenameMacro) \ 111 TYPE(UnaryOperator) \ 112 TYPE(UntouchableMacroFunc) \ 113 TYPE(CSharpStringLiteral) \ 114 TYPE(CSharpNamedArgumentColon) \ 115 TYPE(CSharpNullable) \ 116 TYPE(CSharpNullConditionalLSquare) \ 117 TYPE(CSharpGenericTypeConstraint) \ 118 TYPE(CSharpGenericTypeConstraintColon) \ 119 TYPE(CSharpGenericTypeConstraintComma) \ 120 TYPE(Unknown) 121 122 /// Determines the semantic type of a syntactic token, e.g. whether "<" is a 123 /// template opener or binary operator. 124 enum TokenType : uint8_t { 125 #define TYPE(X) TT_##X, 126 LIST_TOKEN_TYPES 127 #undef TYPE 128 NUM_TOKEN_TYPES 129 }; 130 131 /// Determines the name of a token type. 132 const char *getTokenTypeName(TokenType Type); 133 134 // Represents what type of block a set of braces open. 135 enum BraceBlockKind { BK_Unknown, BK_Block, BK_BracedInit }; 136 137 // The packing kind of a function's parameters. 138 enum ParameterPackingKind { PPK_BinPacked, PPK_OnePerLine, PPK_Inconclusive }; 139 140 enum FormatDecision { FD_Unformatted, FD_Continue, FD_Break }; 141 142 /// Roles a token can take in a configured macro expansion. 143 enum MacroRole { 144 /// The token was expanded from a macro argument when formatting the expanded 145 /// token sequence. 146 MR_ExpandedArg, 147 /// The token is part of a macro argument that was previously formatted as 148 /// expansion when formatting the unexpanded macro call. 149 MR_UnexpandedArg, 150 /// The token was expanded from a macro definition, and is not visible as part 151 /// of the macro call. 152 MR_Hidden, 153 }; 154 155 struct FormatToken; 156 157 /// Contains information on the token's role in a macro expansion. 158 /// 159 /// Given the following definitions: 160 /// A(X) = [ X ] 161 /// B(X) = < X > 162 /// C(X) = X 163 /// 164 /// Consider the macro call: 165 /// A({B(C(C(x)))}) -> [{<x>}] 166 /// 167 /// In this case, the tokens of the unexpanded macro call will have the 168 /// following relevant entries in their macro context (note that formatting 169 /// the unexpanded macro call happens *after* formatting the expanded macro 170 /// call): 171 /// A( { B( C( C(x) ) ) } ) 172 /// Role: NN U NN NN NNUN N N U N (N=None, U=UnexpandedArg) 173 /// 174 /// [ { < x > } ] 175 /// Role: H E H E H E H (H=Hidden, E=ExpandedArg) 176 /// ExpandedFrom[0]: A A A A A A A 177 /// ExpandedFrom[1]: B B B 178 /// ExpandedFrom[2]: C 179 /// ExpandedFrom[3]: C 180 /// StartOfExpansion: 1 0 1 2 0 0 0 181 /// EndOfExpansion: 0 0 0 2 1 0 1 182 struct MacroExpansion { 183 MacroExpansion(MacroRole Role) : Role(Role) {} 184 185 /// The token's role in the macro expansion. 186 /// When formatting an expanded macro, all tokens that are part of macro 187 /// arguments will be MR_ExpandedArg, while all tokens that are not visible in 188 /// the macro call will be MR_Hidden. 189 /// When formatting an unexpanded macro call, all tokens that are part of 190 /// macro arguments will be MR_UnexpandedArg. 191 MacroRole Role; 192 193 /// The stack of macro call identifier tokens this token was expanded from. 194 llvm::SmallVector<FormatToken *, 1> ExpandedFrom; 195 196 /// The number of expansions of which this macro is the first entry. 197 unsigned StartOfExpansion = 0; 198 199 /// The number of currently open expansions in \c ExpandedFrom this macro is 200 /// the last token in. 201 unsigned EndOfExpansion = 0; 202 }; 203 204 class TokenRole; 205 class AnnotatedLine; 206 207 /// A wrapper around a \c Token storing information about the 208 /// whitespace characters preceding it. 209 struct FormatToken { 210 FormatToken() 211 : HasUnescapedNewline(false), IsMultiline(false), IsFirst(false), 212 MustBreakBefore(false), IsUnterminatedLiteral(false), 213 CanBreakBefore(false), ClosesTemplateDeclaration(false), 214 StartsBinaryExpression(false), EndsBinaryExpression(false), 215 PartOfMultiVariableDeclStmt(false), ContinuesLineCommentSection(false), 216 Finalized(false), BlockKind(BK_Unknown), Decision(FD_Unformatted), 217 PackingKind(PPK_Inconclusive), Type(TT_Unknown) {} 218 219 /// The \c Token. 220 Token Tok; 221 222 /// The raw text of the token. 223 /// 224 /// Contains the raw token text without leading whitespace and without leading 225 /// escaped newlines. 226 StringRef TokenText; 227 228 /// A token can have a special role that can carry extra information 229 /// about the token's formatting. 230 /// FIXME: Make FormatToken for parsing and AnnotatedToken two different 231 /// classes and make this a unique_ptr in the AnnotatedToken class. 232 std::shared_ptr<TokenRole> Role; 233 234 /// The range of the whitespace immediately preceding the \c Token. 235 SourceRange WhitespaceRange; 236 237 /// Whether there is at least one unescaped newline before the \c 238 /// Token. 239 unsigned HasUnescapedNewline : 1; 240 241 /// Whether the token text contains newlines (escaped or not). 242 unsigned IsMultiline : 1; 243 244 /// Indicates that this is the first token of the file. 245 unsigned IsFirst : 1; 246 247 /// Whether there must be a line break before this token. 248 /// 249 /// This happens for example when a preprocessor directive ended directly 250 /// before the token. 251 unsigned MustBreakBefore : 1; 252 253 /// Set to \c true if this token is an unterminated literal. 254 unsigned IsUnterminatedLiteral : 1; 255 256 /// \c true if it is allowed to break before this token. 257 unsigned CanBreakBefore : 1; 258 259 /// \c true if this is the ">" of "template<..>". 260 unsigned ClosesTemplateDeclaration : 1; 261 262 /// \c true if this token starts a binary expression, i.e. has at least 263 /// one fake l_paren with a precedence greater than prec::Unknown. 264 unsigned StartsBinaryExpression : 1; 265 /// \c true if this token ends a binary expression. 266 unsigned EndsBinaryExpression : 1; 267 268 /// Is this token part of a \c DeclStmt defining multiple variables? 269 /// 270 /// Only set if \c Type == \c TT_StartOfName. 271 unsigned PartOfMultiVariableDeclStmt : 1; 272 273 /// Does this line comment continue a line comment section? 274 /// 275 /// Only set to true if \c Type == \c TT_LineComment. 276 unsigned ContinuesLineCommentSection : 1; 277 278 /// If \c true, this token has been fully formatted (indented and 279 /// potentially re-formatted inside), and we do not allow further formatting 280 /// changes. 281 unsigned Finalized : 1; 282 283 private: 284 /// Contains the kind of block if this token is a brace. 285 unsigned BlockKind : 2; 286 287 public: 288 BraceBlockKind getBlockKind() const { 289 return static_cast<BraceBlockKind>(BlockKind); 290 } 291 void setBlockKind(BraceBlockKind BBK) { 292 BlockKind = BBK; 293 assert(getBlockKind() == BBK && "BraceBlockKind overflow!"); 294 } 295 296 private: 297 /// Stores the formatting decision for the token once it was made. 298 unsigned Decision : 2; 299 300 public: 301 FormatDecision getDecision() const { 302 return static_cast<FormatDecision>(Decision); 303 } 304 void setDecision(FormatDecision D) { 305 Decision = D; 306 assert(getDecision() == D && "FormatDecision overflow!"); 307 } 308 309 private: 310 /// If this is an opening parenthesis, how are the parameters packed? 311 unsigned PackingKind : 2; 312 313 public: 314 ParameterPackingKind getPackingKind() const { 315 return static_cast<ParameterPackingKind>(PackingKind); 316 } 317 void setPackingKind(ParameterPackingKind K) { 318 PackingKind = K; 319 assert(getPackingKind() == K && "ParameterPackingKind overflow!"); 320 } 321 322 private: 323 TokenType Type; 324 325 public: 326 /// Returns the token's type, e.g. whether "<" is a template opener or 327 /// binary operator. 328 TokenType getType() const { return Type; } 329 void setType(TokenType T) { Type = T; } 330 331 /// The number of newlines immediately before the \c Token. 332 /// 333 /// This can be used to determine what the user wrote in the original code 334 /// and thereby e.g. leave an empty line between two function definitions. 335 unsigned NewlinesBefore = 0; 336 337 /// The offset just past the last '\n' in this token's leading 338 /// whitespace (relative to \c WhiteSpaceStart). 0 if there is no '\n'. 339 unsigned LastNewlineOffset = 0; 340 341 /// The width of the non-whitespace parts of the token (or its first 342 /// line for multi-line tokens) in columns. 343 /// We need this to correctly measure number of columns a token spans. 344 unsigned ColumnWidth = 0; 345 346 /// Contains the width in columns of the last line of a multi-line 347 /// token. 348 unsigned LastLineColumnWidth = 0; 349 350 /// The number of spaces that should be inserted before this token. 351 unsigned SpacesRequiredBefore = 0; 352 353 /// Number of parameters, if this is "(", "[" or "<". 354 unsigned ParameterCount = 0; 355 356 /// Number of parameters that are nested blocks, 357 /// if this is "(", "[" or "<". 358 unsigned BlockParameterCount = 0; 359 360 /// If this is a bracket ("<", "(", "[" or "{"), contains the kind of 361 /// the surrounding bracket. 362 tok::TokenKind ParentBracket = tok::unknown; 363 364 /// The total length of the unwrapped line up to and including this 365 /// token. 366 unsigned TotalLength = 0; 367 368 /// The original 0-based column of this token, including expanded tabs. 369 /// The configured TabWidth is used as tab width. 370 unsigned OriginalColumn = 0; 371 372 /// The length of following tokens until the next natural split point, 373 /// or the next token that can be broken. 374 unsigned UnbreakableTailLength = 0; 375 376 // FIXME: Come up with a 'cleaner' concept. 377 /// The binding strength of a token. This is a combined value of 378 /// operator precedence, parenthesis nesting, etc. 379 unsigned BindingStrength = 0; 380 381 /// The nesting level of this token, i.e. the number of surrounding (), 382 /// [], {} or <>. 383 unsigned NestingLevel = 0; 384 385 /// The indent level of this token. Copied from the surrounding line. 386 unsigned IndentLevel = 0; 387 388 /// Penalty for inserting a line break before this token. 389 unsigned SplitPenalty = 0; 390 391 /// If this is the first ObjC selector name in an ObjC method 392 /// definition or call, this contains the length of the longest name. 393 /// 394 /// This being set to 0 means that the selectors should not be colon-aligned, 395 /// e.g. because several of them are block-type. 396 unsigned LongestObjCSelectorName = 0; 397 398 /// If this is the first ObjC selector name in an ObjC method 399 /// definition or call, this contains the number of parts that the whole 400 /// selector consist of. 401 unsigned ObjCSelectorNameParts = 0; 402 403 /// The 0-based index of the parameter/argument. For ObjC it is set 404 /// for the selector name token. 405 /// For now calculated only for ObjC. 406 unsigned ParameterIndex = 0; 407 408 /// Stores the number of required fake parentheses and the 409 /// corresponding operator precedence. 410 /// 411 /// If multiple fake parentheses start at a token, this vector stores them in 412 /// reverse order, i.e. inner fake parenthesis first. 413 SmallVector<prec::Level, 4> FakeLParens; 414 /// Insert this many fake ) after this token for correct indentation. 415 unsigned FakeRParens = 0; 416 417 /// If this is an operator (or "."/"->") in a sequence of operators 418 /// with the same precedence, contains the 0-based operator index. 419 unsigned OperatorIndex = 0; 420 421 /// If this is an operator (or "."/"->") in a sequence of operators 422 /// with the same precedence, points to the next operator. 423 FormatToken *NextOperator = nullptr; 424 425 /// If this is a bracket, this points to the matching one. 426 FormatToken *MatchingParen = nullptr; 427 428 /// The previous token in the unwrapped line. 429 FormatToken *Previous = nullptr; 430 431 /// The next token in the unwrapped line. 432 FormatToken *Next = nullptr; 433 434 /// The first token in set of column elements. 435 bool StartsColumn = false; 436 437 /// This notes the start of the line of an array initializer. 438 bool ArrayInitializerLineStart = false; 439 440 /// This starts an array initializer. 441 bool IsArrayInitializer = false; 442 443 /// If this token starts a block, this contains all the unwrapped lines 444 /// in it. 445 SmallVector<AnnotatedLine *, 1> Children; 446 447 // Contains all attributes related to how this token takes part 448 // in a configured macro expansion. 449 llvm::Optional<MacroExpansion> MacroCtx; 450 451 bool is(tok::TokenKind Kind) const { return Tok.is(Kind); } 452 bool is(TokenType TT) const { return getType() == TT; } 453 bool is(const IdentifierInfo *II) const { 454 return II && II == Tok.getIdentifierInfo(); 455 } 456 bool is(tok::PPKeywordKind Kind) const { 457 return Tok.getIdentifierInfo() && 458 Tok.getIdentifierInfo()->getPPKeywordID() == Kind; 459 } 460 bool is(BraceBlockKind BBK) const { return getBlockKind() == BBK; } 461 bool is(ParameterPackingKind PPK) const { return getPackingKind() == PPK; } 462 463 template <typename A, typename B> bool isOneOf(A K1, B K2) const { 464 return is(K1) || is(K2); 465 } 466 template <typename A, typename B, typename... Ts> 467 bool isOneOf(A K1, B K2, Ts... Ks) const { 468 return is(K1) || isOneOf(K2, Ks...); 469 } 470 template <typename T> bool isNot(T Kind) const { return !is(Kind); } 471 472 bool isIf(bool AllowConstexprMacro = true) const { 473 return is(tok::kw_if) || endsSequence(tok::kw_constexpr, tok::kw_if) || 474 (endsSequence(tok::identifier, tok::kw_if) && AllowConstexprMacro); 475 } 476 477 bool closesScopeAfterBlock() const { 478 if (getBlockKind() == BK_Block) 479 return true; 480 if (closesScope()) 481 return Previous->closesScopeAfterBlock(); 482 return false; 483 } 484 485 /// \c true if this token starts a sequence with the given tokens in order, 486 /// following the ``Next`` pointers, ignoring comments. 487 template <typename A, typename... Ts> 488 bool startsSequence(A K1, Ts... Tokens) const { 489 return startsSequenceInternal(K1, Tokens...); 490 } 491 492 /// \c true if this token ends a sequence with the given tokens in order, 493 /// following the ``Previous`` pointers, ignoring comments. 494 /// For example, given tokens [T1, T2, T3], the function returns true if 495 /// 3 tokens ending at this (ignoring comments) are [T3, T2, T1]. In other 496 /// words, the tokens passed to this function need to the reverse of the 497 /// order the tokens appear in code. 498 template <typename A, typename... Ts> 499 bool endsSequence(A K1, Ts... Tokens) const { 500 return endsSequenceInternal(K1, Tokens...); 501 } 502 503 bool isStringLiteral() const { return tok::isStringLiteral(Tok.getKind()); } 504 505 bool isObjCAtKeyword(tok::ObjCKeywordKind Kind) const { 506 return Tok.isObjCAtKeyword(Kind); 507 } 508 509 bool isAccessSpecifier(bool ColonRequired = true) const { 510 return isOneOf(tok::kw_public, tok::kw_protected, tok::kw_private) && 511 (!ColonRequired || (Next && Next->is(tok::colon))); 512 } 513 514 bool canBePointerOrReferenceQualifier() const { 515 return isOneOf(tok::kw_const, tok::kw_restrict, tok::kw_volatile, 516 tok::kw___attribute, tok::kw__Nonnull, tok::kw__Nullable, 517 tok::kw__Null_unspecified, tok::kw___ptr32, tok::kw___ptr64, 518 TT_AttributeMacro); 519 } 520 521 /// Determine whether the token is a simple-type-specifier. 522 bool isSimpleTypeSpecifier() const; 523 524 bool isObjCAccessSpecifier() const { 525 return is(tok::at) && Next && 526 (Next->isObjCAtKeyword(tok::objc_public) || 527 Next->isObjCAtKeyword(tok::objc_protected) || 528 Next->isObjCAtKeyword(tok::objc_package) || 529 Next->isObjCAtKeyword(tok::objc_private)); 530 } 531 532 /// Returns whether \p Tok is ([{ or an opening < of a template or in 533 /// protos. 534 bool opensScope() const { 535 if (is(TT_TemplateString) && TokenText.endswith("${")) 536 return true; 537 if (is(TT_DictLiteral) && is(tok::less)) 538 return true; 539 return isOneOf(tok::l_paren, tok::l_brace, tok::l_square, 540 TT_TemplateOpener); 541 } 542 /// Returns whether \p Tok is )]} or a closing > of a template or in 543 /// protos. 544 bool closesScope() const { 545 if (is(TT_TemplateString) && TokenText.startswith("}")) 546 return true; 547 if (is(TT_DictLiteral) && is(tok::greater)) 548 return true; 549 return isOneOf(tok::r_paren, tok::r_brace, tok::r_square, 550 TT_TemplateCloser); 551 } 552 553 /// Returns \c true if this is a "." or "->" accessing a member. 554 bool isMemberAccess() const { 555 return isOneOf(tok::arrow, tok::period, tok::arrowstar) && 556 !isOneOf(TT_DesignatedInitializerPeriod, TT_TrailingReturnArrow, 557 TT_LambdaArrow, TT_LeadingJavaAnnotation); 558 } 559 560 bool isUnaryOperator() const { 561 switch (Tok.getKind()) { 562 case tok::plus: 563 case tok::plusplus: 564 case tok::minus: 565 case tok::minusminus: 566 case tok::exclaim: 567 case tok::tilde: 568 case tok::kw_sizeof: 569 case tok::kw_alignof: 570 return true; 571 default: 572 return false; 573 } 574 } 575 576 bool isBinaryOperator() const { 577 // Comma is a binary operator, but does not behave as such wrt. formatting. 578 return getPrecedence() > prec::Comma; 579 } 580 581 bool isTrailingComment() const { 582 return is(tok::comment) && 583 (is(TT_LineComment) || !Next || Next->NewlinesBefore > 0); 584 } 585 586 /// Returns \c true if this is a keyword that can be used 587 /// like a function call (e.g. sizeof, typeid, ...). 588 bool isFunctionLikeKeyword() const { 589 switch (Tok.getKind()) { 590 case tok::kw_throw: 591 case tok::kw_typeid: 592 case tok::kw_return: 593 case tok::kw_sizeof: 594 case tok::kw_alignof: 595 case tok::kw_alignas: 596 case tok::kw_decltype: 597 case tok::kw_noexcept: 598 case tok::kw_static_assert: 599 case tok::kw__Atomic: 600 case tok::kw___attribute: 601 case tok::kw___underlying_type: 602 case tok::kw_requires: 603 return true; 604 default: 605 return false; 606 } 607 } 608 609 /// Returns \c true if this is a string literal that's like a label, 610 /// e.g. ends with "=" or ":". 611 bool isLabelString() const { 612 if (!is(tok::string_literal)) 613 return false; 614 StringRef Content = TokenText; 615 if (Content.startswith("\"") || Content.startswith("'")) 616 Content = Content.drop_front(1); 617 if (Content.endswith("\"") || Content.endswith("'")) 618 Content = Content.drop_back(1); 619 Content = Content.trim(); 620 return Content.size() > 1 && 621 (Content.back() == ':' || Content.back() == '='); 622 } 623 624 /// Returns actual token start location without leading escaped 625 /// newlines and whitespace. 626 /// 627 /// This can be different to Tok.getLocation(), which includes leading escaped 628 /// newlines. 629 SourceLocation getStartOfNonWhitespace() const { 630 return WhitespaceRange.getEnd(); 631 } 632 633 prec::Level getPrecedence() const { 634 return getBinOpPrecedence(Tok.getKind(), /*GreaterThanIsOperator=*/true, 635 /*CPlusPlus11=*/true); 636 } 637 638 /// Returns the previous token ignoring comments. 639 FormatToken *getPreviousNonComment() const { 640 FormatToken *Tok = Previous; 641 while (Tok && Tok->is(tok::comment)) 642 Tok = Tok->Previous; 643 return Tok; 644 } 645 646 /// Returns the next token ignoring comments. 647 const FormatToken *getNextNonComment() const { 648 const FormatToken *Tok = Next; 649 while (Tok && Tok->is(tok::comment)) 650 Tok = Tok->Next; 651 return Tok; 652 } 653 654 /// Returns \c true if this tokens starts a block-type list, i.e. a 655 /// list that should be indented with a block indent. 656 bool opensBlockOrBlockTypeList(const FormatStyle &Style) const { 657 // C# Does not indent object initialisers as continuations. 658 if (is(tok::l_brace) && getBlockKind() == BK_BracedInit && Style.isCSharp()) 659 return true; 660 if (is(TT_TemplateString) && opensScope()) 661 return true; 662 return is(TT_ArrayInitializerLSquare) || is(TT_ProtoExtensionLSquare) || 663 (is(tok::l_brace) && 664 (getBlockKind() == BK_Block || is(TT_DictLiteral) || 665 (!Style.Cpp11BracedListStyle && NestingLevel == 0))) || 666 (is(tok::less) && (Style.Language == FormatStyle::LK_Proto || 667 Style.Language == FormatStyle::LK_TextProto)); 668 } 669 670 /// Returns whether the token is the left square bracket of a C++ 671 /// structured binding declaration. 672 bool isCppStructuredBinding(const FormatStyle &Style) const { 673 if (!Style.isCpp() || isNot(tok::l_square)) 674 return false; 675 const FormatToken *T = this; 676 do { 677 T = T->getPreviousNonComment(); 678 } while (T && T->isOneOf(tok::kw_const, tok::kw_volatile, tok::amp, 679 tok::ampamp)); 680 return T && T->is(tok::kw_auto); 681 } 682 683 /// Same as opensBlockOrBlockTypeList, but for the closing token. 684 bool closesBlockOrBlockTypeList(const FormatStyle &Style) const { 685 if (is(TT_TemplateString) && closesScope()) 686 return true; 687 return MatchingParen && MatchingParen->opensBlockOrBlockTypeList(Style); 688 } 689 690 /// Return the actual namespace token, if this token starts a namespace 691 /// block. 692 const FormatToken *getNamespaceToken() const { 693 const FormatToken *NamespaceTok = this; 694 if (is(tok::comment)) 695 NamespaceTok = NamespaceTok->getNextNonComment(); 696 // Detect "(inline|export)? namespace" in the beginning of a line. 697 if (NamespaceTok && NamespaceTok->isOneOf(tok::kw_inline, tok::kw_export)) 698 NamespaceTok = NamespaceTok->getNextNonComment(); 699 return NamespaceTok && 700 NamespaceTok->isOneOf(tok::kw_namespace, TT_NamespaceMacro) 701 ? NamespaceTok 702 : nullptr; 703 } 704 705 void copyFrom(const FormatToken &Tok) { *this = Tok; } 706 707 private: 708 // Only allow copying via the explicit copyFrom method. 709 FormatToken(const FormatToken &) = delete; 710 FormatToken &operator=(const FormatToken &) = default; 711 712 template <typename A, typename... Ts> 713 bool startsSequenceInternal(A K1, Ts... Tokens) const { 714 if (is(tok::comment) && Next) 715 return Next->startsSequenceInternal(K1, Tokens...); 716 return is(K1) && Next && Next->startsSequenceInternal(Tokens...); 717 } 718 719 template <typename A> bool startsSequenceInternal(A K1) const { 720 if (is(tok::comment) && Next) 721 return Next->startsSequenceInternal(K1); 722 return is(K1); 723 } 724 725 template <typename A, typename... Ts> bool endsSequenceInternal(A K1) const { 726 if (is(tok::comment) && Previous) 727 return Previous->endsSequenceInternal(K1); 728 return is(K1); 729 } 730 731 template <typename A, typename... Ts> 732 bool endsSequenceInternal(A K1, Ts... Tokens) const { 733 if (is(tok::comment) && Previous) 734 return Previous->endsSequenceInternal(K1, Tokens...); 735 return is(K1) && Previous && Previous->endsSequenceInternal(Tokens...); 736 } 737 }; 738 739 class ContinuationIndenter; 740 struct LineState; 741 742 class TokenRole { 743 public: 744 TokenRole(const FormatStyle &Style) : Style(Style) {} 745 virtual ~TokenRole(); 746 747 /// After the \c TokenAnnotator has finished annotating all the tokens, 748 /// this function precomputes required information for formatting. 749 virtual void precomputeFormattingInfos(const FormatToken *Token); 750 751 /// Apply the special formatting that the given role demands. 752 /// 753 /// Assumes that the token having this role is already formatted. 754 /// 755 /// Continues formatting from \p State leaving indentation to \p Indenter and 756 /// returns the total penalty that this formatting incurs. 757 virtual unsigned formatFromToken(LineState &State, 758 ContinuationIndenter *Indenter, 759 bool DryRun) { 760 return 0; 761 } 762 763 /// Same as \c formatFromToken, but assumes that the first token has 764 /// already been set thereby deciding on the first line break. 765 virtual unsigned formatAfterToken(LineState &State, 766 ContinuationIndenter *Indenter, 767 bool DryRun) { 768 return 0; 769 } 770 771 /// Notifies the \c Role that a comma was found. 772 virtual void CommaFound(const FormatToken *Token) {} 773 774 virtual const FormatToken *lastComma() { return nullptr; } 775 776 protected: 777 const FormatStyle &Style; 778 }; 779 780 class CommaSeparatedList : public TokenRole { 781 public: 782 CommaSeparatedList(const FormatStyle &Style) 783 : TokenRole(Style), HasNestedBracedList(false) {} 784 785 void precomputeFormattingInfos(const FormatToken *Token) override; 786 787 unsigned formatAfterToken(LineState &State, ContinuationIndenter *Indenter, 788 bool DryRun) override; 789 790 unsigned formatFromToken(LineState &State, ContinuationIndenter *Indenter, 791 bool DryRun) override; 792 793 /// Adds \p Token as the next comma to the \c CommaSeparated list. 794 void CommaFound(const FormatToken *Token) override { 795 Commas.push_back(Token); 796 } 797 798 const FormatToken *lastComma() override { 799 if (Commas.empty()) 800 return nullptr; 801 return Commas.back(); 802 } 803 804 private: 805 /// A struct that holds information on how to format a given list with 806 /// a specific number of columns. 807 struct ColumnFormat { 808 /// The number of columns to use. 809 unsigned Columns; 810 811 /// The total width in characters. 812 unsigned TotalWidth; 813 814 /// The number of lines required for this format. 815 unsigned LineCount; 816 817 /// The size of each column in characters. 818 SmallVector<unsigned, 8> ColumnSizes; 819 }; 820 821 /// Calculate which \c ColumnFormat fits best into 822 /// \p RemainingCharacters. 823 const ColumnFormat *getColumnFormat(unsigned RemainingCharacters) const; 824 825 /// The ordered \c FormatTokens making up the commas of this list. 826 SmallVector<const FormatToken *, 8> Commas; 827 828 /// The length of each of the list's items in characters including the 829 /// trailing comma. 830 SmallVector<unsigned, 8> ItemLengths; 831 832 /// Precomputed formats that can be used for this list. 833 SmallVector<ColumnFormat, 4> Formats; 834 835 bool HasNestedBracedList; 836 }; 837 838 /// Encapsulates keywords that are context sensitive or for languages not 839 /// properly supported by Clang's lexer. 840 struct AdditionalKeywords { 841 AdditionalKeywords(IdentifierTable &IdentTable) { 842 kw_final = &IdentTable.get("final"); 843 kw_override = &IdentTable.get("override"); 844 kw_in = &IdentTable.get("in"); 845 kw_of = &IdentTable.get("of"); 846 kw_CF_CLOSED_ENUM = &IdentTable.get("CF_CLOSED_ENUM"); 847 kw_CF_ENUM = &IdentTable.get("CF_ENUM"); 848 kw_CF_OPTIONS = &IdentTable.get("CF_OPTIONS"); 849 kw_NS_CLOSED_ENUM = &IdentTable.get("NS_CLOSED_ENUM"); 850 kw_NS_ENUM = &IdentTable.get("NS_ENUM"); 851 kw_NS_OPTIONS = &IdentTable.get("NS_OPTIONS"); 852 853 kw_as = &IdentTable.get("as"); 854 kw_async = &IdentTable.get("async"); 855 kw_await = &IdentTable.get("await"); 856 kw_declare = &IdentTable.get("declare"); 857 kw_finally = &IdentTable.get("finally"); 858 kw_from = &IdentTable.get("from"); 859 kw_function = &IdentTable.get("function"); 860 kw_get = &IdentTable.get("get"); 861 kw_import = &IdentTable.get("import"); 862 kw_infer = &IdentTable.get("infer"); 863 kw_is = &IdentTable.get("is"); 864 kw_let = &IdentTable.get("let"); 865 kw_module = &IdentTable.get("module"); 866 kw_readonly = &IdentTable.get("readonly"); 867 kw_set = &IdentTable.get("set"); 868 kw_type = &IdentTable.get("type"); 869 kw_typeof = &IdentTable.get("typeof"); 870 kw_var = &IdentTable.get("var"); 871 kw_yield = &IdentTable.get("yield"); 872 873 kw_abstract = &IdentTable.get("abstract"); 874 kw_assert = &IdentTable.get("assert"); 875 kw_extends = &IdentTable.get("extends"); 876 kw_implements = &IdentTable.get("implements"); 877 kw_instanceof = &IdentTable.get("instanceof"); 878 kw_interface = &IdentTable.get("interface"); 879 kw_native = &IdentTable.get("native"); 880 kw_package = &IdentTable.get("package"); 881 kw_synchronized = &IdentTable.get("synchronized"); 882 kw_throws = &IdentTable.get("throws"); 883 kw___except = &IdentTable.get("__except"); 884 kw___has_include = &IdentTable.get("__has_include"); 885 kw___has_include_next = &IdentTable.get("__has_include_next"); 886 887 kw_mark = &IdentTable.get("mark"); 888 889 kw_extend = &IdentTable.get("extend"); 890 kw_option = &IdentTable.get("option"); 891 kw_optional = &IdentTable.get("optional"); 892 kw_repeated = &IdentTable.get("repeated"); 893 kw_required = &IdentTable.get("required"); 894 kw_returns = &IdentTable.get("returns"); 895 896 kw_signals = &IdentTable.get("signals"); 897 kw_qsignals = &IdentTable.get("Q_SIGNALS"); 898 kw_slots = &IdentTable.get("slots"); 899 kw_qslots = &IdentTable.get("Q_SLOTS"); 900 901 // C# keywords 902 kw_dollar = &IdentTable.get("dollar"); 903 kw_base = &IdentTable.get("base"); 904 kw_byte = &IdentTable.get("byte"); 905 kw_checked = &IdentTable.get("checked"); 906 kw_decimal = &IdentTable.get("decimal"); 907 kw_delegate = &IdentTable.get("delegate"); 908 kw_event = &IdentTable.get("event"); 909 kw_fixed = &IdentTable.get("fixed"); 910 kw_foreach = &IdentTable.get("foreach"); 911 kw_implicit = &IdentTable.get("implicit"); 912 kw_internal = &IdentTable.get("internal"); 913 kw_lock = &IdentTable.get("lock"); 914 kw_null = &IdentTable.get("null"); 915 kw_object = &IdentTable.get("object"); 916 kw_out = &IdentTable.get("out"); 917 kw_params = &IdentTable.get("params"); 918 kw_ref = &IdentTable.get("ref"); 919 kw_string = &IdentTable.get("string"); 920 kw_stackalloc = &IdentTable.get("stackalloc"); 921 kw_sbyte = &IdentTable.get("sbyte"); 922 kw_sealed = &IdentTable.get("sealed"); 923 kw_uint = &IdentTable.get("uint"); 924 kw_ulong = &IdentTable.get("ulong"); 925 kw_unchecked = &IdentTable.get("unchecked"); 926 kw_unsafe = &IdentTable.get("unsafe"); 927 kw_ushort = &IdentTable.get("ushort"); 928 kw_when = &IdentTable.get("when"); 929 kw_where = &IdentTable.get("where"); 930 931 // Keep this at the end of the constructor to make sure everything here 932 // is 933 // already initialized. 934 JsExtraKeywords = std::unordered_set<IdentifierInfo *>( 935 {kw_as, kw_async, kw_await, kw_declare, kw_finally, kw_from, 936 kw_function, kw_get, kw_import, kw_is, kw_let, kw_module, kw_readonly, 937 kw_set, kw_type, kw_typeof, kw_var, kw_yield, 938 // Keywords from the Java section. 939 kw_abstract, kw_extends, kw_implements, kw_instanceof, kw_interface}); 940 941 CSharpExtraKeywords = std::unordered_set<IdentifierInfo *>( 942 {kw_base, kw_byte, kw_checked, kw_decimal, kw_delegate, kw_event, 943 kw_fixed, kw_foreach, kw_implicit, kw_in, kw_interface, kw_internal, 944 kw_is, kw_lock, kw_null, kw_object, kw_out, kw_override, kw_params, 945 kw_readonly, kw_ref, kw_string, kw_stackalloc, kw_sbyte, kw_sealed, 946 kw_uint, kw_ulong, kw_unchecked, kw_unsafe, kw_ushort, kw_when, 947 kw_where, 948 // Keywords from the JavaScript section. 949 kw_as, kw_async, kw_await, kw_declare, kw_finally, kw_from, 950 kw_function, kw_get, kw_import, kw_is, kw_let, kw_module, kw_readonly, 951 kw_set, kw_type, kw_typeof, kw_var, kw_yield, 952 // Keywords from the Java section. 953 kw_abstract, kw_extends, kw_implements, kw_instanceof, kw_interface}); 954 } 955 956 // Context sensitive keywords. 957 IdentifierInfo *kw_final; 958 IdentifierInfo *kw_override; 959 IdentifierInfo *kw_in; 960 IdentifierInfo *kw_of; 961 IdentifierInfo *kw_CF_CLOSED_ENUM; 962 IdentifierInfo *kw_CF_ENUM; 963 IdentifierInfo *kw_CF_OPTIONS; 964 IdentifierInfo *kw_NS_CLOSED_ENUM; 965 IdentifierInfo *kw_NS_ENUM; 966 IdentifierInfo *kw_NS_OPTIONS; 967 IdentifierInfo *kw___except; 968 IdentifierInfo *kw___has_include; 969 IdentifierInfo *kw___has_include_next; 970 971 // JavaScript keywords. 972 IdentifierInfo *kw_as; 973 IdentifierInfo *kw_async; 974 IdentifierInfo *kw_await; 975 IdentifierInfo *kw_declare; 976 IdentifierInfo *kw_finally; 977 IdentifierInfo *kw_from; 978 IdentifierInfo *kw_function; 979 IdentifierInfo *kw_get; 980 IdentifierInfo *kw_import; 981 IdentifierInfo *kw_infer; 982 IdentifierInfo *kw_is; 983 IdentifierInfo *kw_let; 984 IdentifierInfo *kw_module; 985 IdentifierInfo *kw_readonly; 986 IdentifierInfo *kw_set; 987 IdentifierInfo *kw_type; 988 IdentifierInfo *kw_typeof; 989 IdentifierInfo *kw_var; 990 IdentifierInfo *kw_yield; 991 992 // Java keywords. 993 IdentifierInfo *kw_abstract; 994 IdentifierInfo *kw_assert; 995 IdentifierInfo *kw_extends; 996 IdentifierInfo *kw_implements; 997 IdentifierInfo *kw_instanceof; 998 IdentifierInfo *kw_interface; 999 IdentifierInfo *kw_native; 1000 IdentifierInfo *kw_package; 1001 IdentifierInfo *kw_synchronized; 1002 IdentifierInfo *kw_throws; 1003 1004 // Pragma keywords. 1005 IdentifierInfo *kw_mark; 1006 1007 // Proto keywords. 1008 IdentifierInfo *kw_extend; 1009 IdentifierInfo *kw_option; 1010 IdentifierInfo *kw_optional; 1011 IdentifierInfo *kw_repeated; 1012 IdentifierInfo *kw_required; 1013 IdentifierInfo *kw_returns; 1014 1015 // QT keywords. 1016 IdentifierInfo *kw_signals; 1017 IdentifierInfo *kw_qsignals; 1018 IdentifierInfo *kw_slots; 1019 IdentifierInfo *kw_qslots; 1020 1021 // C# keywords 1022 IdentifierInfo *kw_dollar; 1023 IdentifierInfo *kw_base; 1024 IdentifierInfo *kw_byte; 1025 IdentifierInfo *kw_checked; 1026 IdentifierInfo *kw_decimal; 1027 IdentifierInfo *kw_delegate; 1028 IdentifierInfo *kw_event; 1029 IdentifierInfo *kw_fixed; 1030 IdentifierInfo *kw_foreach; 1031 IdentifierInfo *kw_implicit; 1032 IdentifierInfo *kw_internal; 1033 1034 IdentifierInfo *kw_lock; 1035 IdentifierInfo *kw_null; 1036 IdentifierInfo *kw_object; 1037 IdentifierInfo *kw_out; 1038 1039 IdentifierInfo *kw_params; 1040 1041 IdentifierInfo *kw_ref; 1042 IdentifierInfo *kw_string; 1043 IdentifierInfo *kw_stackalloc; 1044 IdentifierInfo *kw_sbyte; 1045 IdentifierInfo *kw_sealed; 1046 IdentifierInfo *kw_uint; 1047 IdentifierInfo *kw_ulong; 1048 IdentifierInfo *kw_unchecked; 1049 IdentifierInfo *kw_unsafe; 1050 IdentifierInfo *kw_ushort; 1051 IdentifierInfo *kw_when; 1052 IdentifierInfo *kw_where; 1053 1054 /// Returns \c true if \p Tok is a true JavaScript identifier, returns 1055 /// \c false if it is a keyword or a pseudo keyword. 1056 /// If \c AcceptIdentifierName is true, returns true not only for keywords, 1057 // but also for IdentifierName tokens (aka pseudo-keywords), such as 1058 // ``yield``. 1059 bool IsJavaScriptIdentifier(const FormatToken &Tok, 1060 bool AcceptIdentifierName = true) const { 1061 // Based on the list of JavaScript & TypeScript keywords here: 1062 // https://github.com/microsoft/TypeScript/blob/master/src/compiler/scanner.ts#L74 1063 switch (Tok.Tok.getKind()) { 1064 case tok::kw_break: 1065 case tok::kw_case: 1066 case tok::kw_catch: 1067 case tok::kw_class: 1068 case tok::kw_continue: 1069 case tok::kw_const: 1070 case tok::kw_default: 1071 case tok::kw_delete: 1072 case tok::kw_do: 1073 case tok::kw_else: 1074 case tok::kw_enum: 1075 case tok::kw_export: 1076 case tok::kw_false: 1077 case tok::kw_for: 1078 case tok::kw_if: 1079 case tok::kw_import: 1080 case tok::kw_module: 1081 case tok::kw_new: 1082 case tok::kw_private: 1083 case tok::kw_protected: 1084 case tok::kw_public: 1085 case tok::kw_return: 1086 case tok::kw_static: 1087 case tok::kw_switch: 1088 case tok::kw_this: 1089 case tok::kw_throw: 1090 case tok::kw_true: 1091 case tok::kw_try: 1092 case tok::kw_typeof: 1093 case tok::kw_void: 1094 case tok::kw_while: 1095 // These are JS keywords that are lexed by LLVM/clang as keywords. 1096 return false; 1097 case tok::identifier: { 1098 // For identifiers, make sure they are true identifiers, excluding the 1099 // JavaScript pseudo-keywords (not lexed by LLVM/clang as keywords). 1100 bool IsPseudoKeyword = 1101 JsExtraKeywords.find(Tok.Tok.getIdentifierInfo()) != 1102 JsExtraKeywords.end(); 1103 return AcceptIdentifierName || !IsPseudoKeyword; 1104 } 1105 default: 1106 // Other keywords are handled in the switch below, to avoid problems due 1107 // to duplicate case labels when using the #include trick. 1108 break; 1109 } 1110 1111 switch (Tok.Tok.getKind()) { 1112 // Handle C++ keywords not included above: these are all JS identifiers. 1113 #define KEYWORD(X, Y) case tok::kw_##X: 1114 #include "clang/Basic/TokenKinds.def" 1115 // #undef KEYWORD is not needed -- it's #undef-ed at the end of 1116 // TokenKinds.def 1117 return true; 1118 default: 1119 // All other tokens (punctuation etc) are not JS identifiers. 1120 return false; 1121 } 1122 } 1123 1124 /// Returns \c true if \p Tok is a C# keyword, returns 1125 /// \c false if it is a anything else. 1126 bool isCSharpKeyword(const FormatToken &Tok) const { 1127 switch (Tok.Tok.getKind()) { 1128 case tok::kw_bool: 1129 case tok::kw_break: 1130 case tok::kw_case: 1131 case tok::kw_catch: 1132 case tok::kw_char: 1133 case tok::kw_class: 1134 case tok::kw_const: 1135 case tok::kw_continue: 1136 case tok::kw_default: 1137 case tok::kw_do: 1138 case tok::kw_double: 1139 case tok::kw_else: 1140 case tok::kw_enum: 1141 case tok::kw_explicit: 1142 case tok::kw_extern: 1143 case tok::kw_false: 1144 case tok::kw_float: 1145 case tok::kw_for: 1146 case tok::kw_goto: 1147 case tok::kw_if: 1148 case tok::kw_int: 1149 case tok::kw_long: 1150 case tok::kw_namespace: 1151 case tok::kw_new: 1152 case tok::kw_operator: 1153 case tok::kw_private: 1154 case tok::kw_protected: 1155 case tok::kw_public: 1156 case tok::kw_return: 1157 case tok::kw_short: 1158 case tok::kw_sizeof: 1159 case tok::kw_static: 1160 case tok::kw_struct: 1161 case tok::kw_switch: 1162 case tok::kw_this: 1163 case tok::kw_throw: 1164 case tok::kw_true: 1165 case tok::kw_try: 1166 case tok::kw_typeof: 1167 case tok::kw_using: 1168 case tok::kw_virtual: 1169 case tok::kw_void: 1170 case tok::kw_volatile: 1171 case tok::kw_while: 1172 return true; 1173 default: 1174 return Tok.is(tok::identifier) && 1175 CSharpExtraKeywords.find(Tok.Tok.getIdentifierInfo()) == 1176 CSharpExtraKeywords.end(); 1177 } 1178 } 1179 1180 private: 1181 /// The JavaScript keywords beyond the C++ keyword set. 1182 std::unordered_set<IdentifierInfo *> JsExtraKeywords; 1183 1184 /// The C# keywords beyond the C++ keyword set 1185 std::unordered_set<IdentifierInfo *> CSharpExtraKeywords; 1186 }; 1187 1188 } // namespace format 1189 } // namespace clang 1190 1191 #endif 1192