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