1 //===--- FormatToken.h - Format C++ code ------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 /// 10 /// \file 11 /// \brief This file contains the declaration of the FormatToken, a wrapper 12 /// around Token with additional information related to formatting. 13 /// 14 //===----------------------------------------------------------------------===// 15 16 #ifndef LLVM_CLANG_LIB_FORMAT_FORMATTOKEN_H 17 #define LLVM_CLANG_LIB_FORMAT_FORMATTOKEN_H 18 19 #include "clang/Basic/IdentifierTable.h" 20 #include "clang/Basic/OperatorPrecedence.h" 21 #include "clang/Format/Format.h" 22 #include "clang/Lex/Lexer.h" 23 #include <memory> 24 25 namespace clang { 26 namespace format { 27 28 #define LIST_TOKEN_TYPES \ 29 TYPE(ArrayInitializerLSquare) \ 30 TYPE(ArraySubscriptLSquare) \ 31 TYPE(AttributeParen) \ 32 TYPE(BinaryOperator) \ 33 TYPE(BitFieldColon) \ 34 TYPE(BlockComment) \ 35 TYPE(CastRParen) \ 36 TYPE(ConditionalExpr) \ 37 TYPE(ConflictAlternative) \ 38 TYPE(ConflictEnd) \ 39 TYPE(ConflictStart) \ 40 TYPE(CtorInitializerColon) \ 41 TYPE(CtorInitializerComma) \ 42 TYPE(DesignatedInitializerPeriod) \ 43 TYPE(DictLiteral) \ 44 TYPE(ForEachMacro) \ 45 TYPE(FunctionAnnotationRParen) \ 46 TYPE(FunctionDeclarationName) \ 47 TYPE(FunctionLBrace) \ 48 TYPE(FunctionTypeLParen) \ 49 TYPE(ImplicitStringLiteral) \ 50 TYPE(InheritanceColon) \ 51 TYPE(InlineASMBrace) \ 52 TYPE(InlineASMColon) \ 53 TYPE(JavaAnnotation) \ 54 TYPE(JsComputedPropertyName) \ 55 TYPE(JsFatArrow) \ 56 TYPE(JsTypeColon) \ 57 TYPE(JsTypeOptionalQuestion) \ 58 TYPE(LambdaArrow) \ 59 TYPE(LambdaLSquare) \ 60 TYPE(LeadingJavaAnnotation) \ 61 TYPE(LineComment) \ 62 TYPE(MacroBlockBegin) \ 63 TYPE(MacroBlockEnd) \ 64 TYPE(ObjCBlockLBrace) \ 65 TYPE(ObjCBlockLParen) \ 66 TYPE(ObjCDecl) \ 67 TYPE(ObjCForIn) \ 68 TYPE(ObjCMethodExpr) \ 69 TYPE(ObjCMethodSpecifier) \ 70 TYPE(ObjCProperty) \ 71 TYPE(ObjCStringLiteral) \ 72 TYPE(OverloadedOperator) \ 73 TYPE(OverloadedOperatorLParen) \ 74 TYPE(PointerOrReference) \ 75 TYPE(PureVirtualSpecifier) \ 76 TYPE(RangeBasedForLoopColon) \ 77 TYPE(RegexLiteral) \ 78 TYPE(SelectorName) \ 79 TYPE(StartOfName) \ 80 TYPE(TemplateCloser) \ 81 TYPE(TemplateOpener) \ 82 TYPE(TemplateString) \ 83 TYPE(TrailingAnnotation) \ 84 TYPE(TrailingReturnArrow) \ 85 TYPE(TrailingUnaryOperator) \ 86 TYPE(UnaryOperator) \ 87 TYPE(Unknown) 88 89 enum TokenType { 90 #define TYPE(X) TT_##X, 91 LIST_TOKEN_TYPES 92 #undef TYPE 93 NUM_TOKEN_TYPES 94 }; 95 96 /// \brief Determines the name of a token type. 97 const char *getTokenTypeName(TokenType Type); 98 99 // Represents what type of block a set of braces open. 100 enum BraceBlockKind { BK_Unknown, BK_Block, BK_BracedInit }; 101 102 // The packing kind of a function's parameters. 103 enum ParameterPackingKind { PPK_BinPacked, PPK_OnePerLine, PPK_Inconclusive }; 104 105 enum FormatDecision { FD_Unformatted, FD_Continue, FD_Break }; 106 107 class TokenRole; 108 class AnnotatedLine; 109 110 /// \brief A wrapper around a \c Token storing information about the 111 /// whitespace characters preceding it. 112 struct FormatToken { 113 FormatToken() {} 114 115 /// \brief The \c Token. 116 Token Tok; 117 118 /// \brief The number of newlines immediately before the \c Token. 119 /// 120 /// This can be used to determine what the user wrote in the original code 121 /// and thereby e.g. leave an empty line between two function definitions. 122 unsigned NewlinesBefore = 0; 123 124 /// \brief Whether there is at least one unescaped newline before the \c 125 /// Token. 126 bool HasUnescapedNewline = false; 127 128 /// \brief The range of the whitespace immediately preceding the \c Token. 129 SourceRange WhitespaceRange; 130 131 /// \brief The offset just past the last '\n' in this token's leading 132 /// whitespace (relative to \c WhiteSpaceStart). 0 if there is no '\n'. 133 unsigned LastNewlineOffset = 0; 134 135 /// \brief The width of the non-whitespace parts of the token (or its first 136 /// line for multi-line tokens) in columns. 137 /// We need this to correctly measure number of columns a token spans. 138 unsigned ColumnWidth = 0; 139 140 /// \brief Contains the width in columns of the last line of a multi-line 141 /// token. 142 unsigned LastLineColumnWidth = 0; 143 144 /// \brief Whether the token text contains newlines (escaped or not). 145 bool IsMultiline = false; 146 147 /// \brief Indicates that this is the first token. 148 bool IsFirst = false; 149 150 /// \brief Whether there must be a line break before this token. 151 /// 152 /// This happens for example when a preprocessor directive ended directly 153 /// before the token. 154 bool MustBreakBefore = false; 155 156 /// \brief The raw text of the token. 157 /// 158 /// Contains the raw token text without leading whitespace and without leading 159 /// escaped newlines. 160 StringRef TokenText; 161 162 /// \brief Set to \c true if this token is an unterminated literal. 163 bool IsUnterminatedLiteral = 0; 164 165 /// \brief Contains the kind of block if this token is a brace. 166 BraceBlockKind BlockKind = BK_Unknown; 167 168 TokenType Type = TT_Unknown; 169 170 /// \brief The number of spaces that should be inserted before this token. 171 unsigned SpacesRequiredBefore = 0; 172 173 /// \brief \c true if it is allowed to break before this token. 174 bool CanBreakBefore = false; 175 176 /// \brief \c true if this is the ">" of "template<..>". 177 bool ClosesTemplateDeclaration = false; 178 179 /// \brief Number of parameters, if this is "(", "[" or "<". 180 /// 181 /// This is initialized to 1 as we don't need to distinguish functions with 182 /// 0 parameters from functions with 1 parameter. Thus, we can simply count 183 /// the number of commas. 184 unsigned ParameterCount = 0; 185 186 /// \brief Number of parameters that are nested blocks, 187 /// if this is "(", "[" or "<". 188 unsigned BlockParameterCount = 0; 189 190 /// \brief If this is a bracket ("<", "(", "[" or "{"), contains the kind of 191 /// the surrounding bracket. 192 tok::TokenKind ParentBracket = tok::unknown; 193 194 /// \brief A token can have a special role that can carry extra information 195 /// about the token's formatting. 196 std::unique_ptr<TokenRole> Role; 197 198 /// \brief If this is an opening parenthesis, how are the parameters packed? 199 ParameterPackingKind PackingKind = PPK_Inconclusive; 200 201 /// \brief The total length of the unwrapped line up to and including this 202 /// token. 203 unsigned TotalLength = 0; 204 205 /// \brief The original 0-based column of this token, including expanded tabs. 206 /// The configured TabWidth is used as tab width. 207 unsigned OriginalColumn = 0; 208 209 /// \brief The length of following tokens until the next natural split point, 210 /// or the next token that can be broken. 211 unsigned UnbreakableTailLength = 0; 212 213 // FIXME: Come up with a 'cleaner' concept. 214 /// \brief The binding strength of a token. This is a combined value of 215 /// operator precedence, parenthesis nesting, etc. 216 unsigned BindingStrength = 0; 217 218 /// \brief The nesting level of this token, i.e. the number of surrounding (), 219 /// [], {} or <>. 220 unsigned NestingLevel = 0; 221 222 /// \brief Penalty for inserting a line break before this token. 223 unsigned SplitPenalty = 0; 224 225 /// \brief If this is the first ObjC selector name in an ObjC method 226 /// definition or call, this contains the length of the longest name. 227 /// 228 /// This being set to 0 means that the selectors should not be colon-aligned, 229 /// e.g. because several of them are block-type. 230 unsigned LongestObjCSelectorName = 0; 231 232 /// \brief Stores the number of required fake parentheses and the 233 /// corresponding operator precedence. 234 /// 235 /// If multiple fake parentheses start at a token, this vector stores them in 236 /// reverse order, i.e. inner fake parenthesis first. 237 SmallVector<prec::Level, 4> FakeLParens; 238 /// \brief Insert this many fake ) after this token for correct indentation. 239 unsigned FakeRParens = 0; 240 241 /// \brief \c true if this token starts a binary expression, i.e. has at least 242 /// one fake l_paren with a precedence greater than prec::Unknown. 243 bool StartsBinaryExpression = false; 244 /// \brief \c true if this token ends a binary expression. 245 bool EndsBinaryExpression = false; 246 247 /// \brief Is this is an operator (or "."/"->") in a sequence of operators 248 /// with the same precedence, contains the 0-based operator index. 249 unsigned OperatorIndex = 0; 250 251 /// \brief Is this the last operator (or "."/"->") in a sequence of operators 252 /// with the same precedence? 253 bool LastOperator = false; 254 255 /// \brief Is this token part of a \c DeclStmt defining multiple variables? 256 /// 257 /// Only set if \c Type == \c TT_StartOfName. 258 bool PartOfMultiVariableDeclStmt = false; 259 260 /// \brief If this is a bracket, this points to the matching one. 261 FormatToken *MatchingParen = nullptr; 262 263 /// \brief The previous token in the unwrapped line. 264 FormatToken *Previous = nullptr; 265 266 /// \brief The next token in the unwrapped line. 267 FormatToken *Next = nullptr; 268 269 /// \brief If this token starts a block, this contains all the unwrapped lines 270 /// in it. 271 SmallVector<AnnotatedLine *, 1> Children; 272 273 /// \brief Stores the formatting decision for the token once it was made. 274 FormatDecision Decision = FD_Unformatted; 275 276 /// \brief If \c true, this token has been fully formatted (indented and 277 /// potentially re-formatted inside), and we do not allow further formatting 278 /// changes. 279 bool Finalized = false; 280 281 bool is(tok::TokenKind Kind) const { return Tok.is(Kind); } 282 bool is(TokenType TT) const { return Type == TT; } 283 bool is(const IdentifierInfo *II) const { 284 return II && II == Tok.getIdentifierInfo(); 285 } 286 bool is(tok::PPKeywordKind Kind) const { 287 return Tok.getIdentifierInfo() && 288 Tok.getIdentifierInfo()->getPPKeywordID() == Kind; 289 } 290 template <typename A, typename B> bool isOneOf(A K1, B K2) const { 291 return is(K1) || is(K2); 292 } 293 template <typename A, typename B, typename... Ts> 294 bool isOneOf(A K1, B K2, Ts... Ks) const { 295 return is(K1) || isOneOf(K2, Ks...); 296 } 297 template <typename T> bool isNot(T Kind) const { return !is(Kind); } 298 299 bool isStringLiteral() const { return tok::isStringLiteral(Tok.getKind()); } 300 301 bool isObjCAtKeyword(tok::ObjCKeywordKind Kind) const { 302 return Tok.isObjCAtKeyword(Kind); 303 } 304 305 bool isAccessSpecifier(bool ColonRequired = true) const { 306 return isOneOf(tok::kw_public, tok::kw_protected, tok::kw_private) && 307 (!ColonRequired || (Next && Next->is(tok::colon))); 308 } 309 310 /// \brief Determine whether the token is a simple-type-specifier. 311 bool isSimpleTypeSpecifier() const; 312 313 bool isObjCAccessSpecifier() const { 314 return is(tok::at) && Next && (Next->isObjCAtKeyword(tok::objc_public) || 315 Next->isObjCAtKeyword(tok::objc_protected) || 316 Next->isObjCAtKeyword(tok::objc_package) || 317 Next->isObjCAtKeyword(tok::objc_private)); 318 } 319 320 /// \brief Returns whether \p Tok is ([{ or a template opening <. 321 bool opensScope() const { 322 return isOneOf(tok::l_paren, tok::l_brace, tok::l_square, 323 TT_TemplateOpener); 324 } 325 /// \brief Returns whether \p Tok is )]} or a template closing >. 326 bool closesScope() const { 327 return isOneOf(tok::r_paren, tok::r_brace, tok::r_square, 328 TT_TemplateCloser); 329 } 330 331 /// \brief Returns \c true if this is a "." or "->" accessing a member. 332 bool isMemberAccess() const { 333 return isOneOf(tok::arrow, tok::period, tok::arrowstar) && 334 !isOneOf(TT_DesignatedInitializerPeriod, TT_TrailingReturnArrow, 335 TT_LambdaArrow); 336 } 337 338 bool isUnaryOperator() const { 339 switch (Tok.getKind()) { 340 case tok::plus: 341 case tok::plusplus: 342 case tok::minus: 343 case tok::minusminus: 344 case tok::exclaim: 345 case tok::tilde: 346 case tok::kw_sizeof: 347 case tok::kw_alignof: 348 return true; 349 default: 350 return false; 351 } 352 } 353 354 bool isBinaryOperator() const { 355 // Comma is a binary operator, but does not behave as such wrt. formatting. 356 return getPrecedence() > prec::Comma; 357 } 358 359 bool isTrailingComment() const { 360 return is(tok::comment) && 361 (is(TT_LineComment) || !Next || Next->NewlinesBefore > 0); 362 } 363 364 /// \brief Returns \c true if this is a keyword that can be used 365 /// like a function call (e.g. sizeof, typeid, ...). 366 bool isFunctionLikeKeyword() const { 367 switch (Tok.getKind()) { 368 case tok::kw_throw: 369 case tok::kw_typeid: 370 case tok::kw_return: 371 case tok::kw_sizeof: 372 case tok::kw_alignof: 373 case tok::kw_alignas: 374 case tok::kw_decltype: 375 case tok::kw_noexcept: 376 case tok::kw_static_assert: 377 case tok::kw___attribute: 378 return true; 379 default: 380 return false; 381 } 382 } 383 384 /// \brief Returns actual token start location without leading escaped 385 /// newlines and whitespace. 386 /// 387 /// This can be different to Tok.getLocation(), which includes leading escaped 388 /// newlines. 389 SourceLocation getStartOfNonWhitespace() const { 390 return WhitespaceRange.getEnd(); 391 } 392 393 prec::Level getPrecedence() const { 394 return getBinOpPrecedence(Tok.getKind(), true, true); 395 } 396 397 /// \brief Returns the previous token ignoring comments. 398 FormatToken *getPreviousNonComment() const { 399 FormatToken *Tok = Previous; 400 while (Tok && Tok->is(tok::comment)) 401 Tok = Tok->Previous; 402 return Tok; 403 } 404 405 /// \brief Returns the next token ignoring comments. 406 const FormatToken *getNextNonComment() const { 407 const FormatToken *Tok = Next; 408 while (Tok && Tok->is(tok::comment)) 409 Tok = Tok->Next; 410 return Tok; 411 } 412 413 /// \brief Returns \c true if this tokens starts a block-type list, i.e. a 414 /// list that should be indented with a block indent. 415 bool opensBlockOrBlockTypeList(const FormatStyle &Style) const { 416 return is(TT_ArrayInitializerLSquare) || 417 (is(tok::l_brace) && 418 (BlockKind == BK_Block || is(TT_DictLiteral) || 419 (!Style.Cpp11BracedListStyle && NestingLevel == 0))); 420 } 421 422 /// \brief Same as opensBlockOrBlockTypeList, but for the closing token. 423 bool closesBlockOrBlockTypeList(const FormatStyle &Style) const { 424 return MatchingParen && MatchingParen->opensBlockOrBlockTypeList(Style); 425 } 426 427 private: 428 // Disallow copying. 429 FormatToken(const FormatToken &) = delete; 430 void operator=(const FormatToken &) = delete; 431 }; 432 433 class ContinuationIndenter; 434 struct LineState; 435 436 class TokenRole { 437 public: 438 TokenRole(const FormatStyle &Style) : Style(Style) {} 439 virtual ~TokenRole(); 440 441 /// \brief After the \c TokenAnnotator has finished annotating all the tokens, 442 /// this function precomputes required information for formatting. 443 virtual void precomputeFormattingInfos(const FormatToken *Token); 444 445 /// \brief Apply the special formatting that the given role demands. 446 /// 447 /// Assumes that the token having this role is already formatted. 448 /// 449 /// Continues formatting from \p State leaving indentation to \p Indenter and 450 /// returns the total penalty that this formatting incurs. 451 virtual unsigned formatFromToken(LineState &State, 452 ContinuationIndenter *Indenter, 453 bool DryRun) { 454 return 0; 455 } 456 457 /// \brief Same as \c formatFromToken, but assumes that the first token has 458 /// already been set thereby deciding on the first line break. 459 virtual unsigned formatAfterToken(LineState &State, 460 ContinuationIndenter *Indenter, 461 bool DryRun) { 462 return 0; 463 } 464 465 /// \brief Notifies the \c Role that a comma was found. 466 virtual void CommaFound(const FormatToken *Token) {} 467 468 protected: 469 const FormatStyle &Style; 470 }; 471 472 class CommaSeparatedList : public TokenRole { 473 public: 474 CommaSeparatedList(const FormatStyle &Style) 475 : TokenRole(Style), HasNestedBracedList(false) {} 476 477 void precomputeFormattingInfos(const FormatToken *Token) override; 478 479 unsigned formatAfterToken(LineState &State, ContinuationIndenter *Indenter, 480 bool DryRun) override; 481 482 unsigned formatFromToken(LineState &State, ContinuationIndenter *Indenter, 483 bool DryRun) override; 484 485 /// \brief Adds \p Token as the next comma to the \c CommaSeparated list. 486 void CommaFound(const FormatToken *Token) override { 487 Commas.push_back(Token); 488 } 489 490 private: 491 /// \brief A struct that holds information on how to format a given list with 492 /// a specific number of columns. 493 struct ColumnFormat { 494 /// \brief The number of columns to use. 495 unsigned Columns; 496 497 /// \brief The total width in characters. 498 unsigned TotalWidth; 499 500 /// \brief The number of lines required for this format. 501 unsigned LineCount; 502 503 /// \brief The size of each column in characters. 504 SmallVector<unsigned, 8> ColumnSizes; 505 }; 506 507 /// \brief Calculate which \c ColumnFormat fits best into 508 /// \p RemainingCharacters. 509 const ColumnFormat *getColumnFormat(unsigned RemainingCharacters) const; 510 511 /// \brief The ordered \c FormatTokens making up the commas of this list. 512 SmallVector<const FormatToken *, 8> Commas; 513 514 /// \brief The length of each of the list's items in characters including the 515 /// trailing comma. 516 SmallVector<unsigned, 8> ItemLengths; 517 518 /// \brief Precomputed formats that can be used for this list. 519 SmallVector<ColumnFormat, 4> Formats; 520 521 bool HasNestedBracedList; 522 }; 523 524 /// \brief Encapsulates keywords that are context sensitive or for languages not 525 /// properly supported by Clang's lexer. 526 struct AdditionalKeywords { 527 AdditionalKeywords(IdentifierTable &IdentTable) { 528 kw_final = &IdentTable.get("final"); 529 kw_override = &IdentTable.get("override"); 530 kw_in = &IdentTable.get("in"); 531 kw_CF_ENUM = &IdentTable.get("CF_ENUM"); 532 kw_CF_OPTIONS = &IdentTable.get("CF_OPTIONS"); 533 kw_NS_ENUM = &IdentTable.get("NS_ENUM"); 534 kw_NS_OPTIONS = &IdentTable.get("NS_OPTIONS"); 535 536 kw_finally = &IdentTable.get("finally"); 537 kw_function = &IdentTable.get("function"); 538 kw_import = &IdentTable.get("import"); 539 kw_let = &IdentTable.get("let"); 540 kw_var = &IdentTable.get("var"); 541 542 kw_abstract = &IdentTable.get("abstract"); 543 kw_assert = &IdentTable.get("assert"); 544 kw_extends = &IdentTable.get("extends"); 545 kw_implements = &IdentTable.get("implements"); 546 kw_instanceof = &IdentTable.get("instanceof"); 547 kw_interface = &IdentTable.get("interface"); 548 kw_native = &IdentTable.get("native"); 549 kw_package = &IdentTable.get("package"); 550 kw_synchronized = &IdentTable.get("synchronized"); 551 kw_throws = &IdentTable.get("throws"); 552 kw___except = &IdentTable.get("__except"); 553 554 kw_mark = &IdentTable.get("mark"); 555 556 kw_extend = &IdentTable.get("extend"); 557 kw_option = &IdentTable.get("option"); 558 kw_optional = &IdentTable.get("optional"); 559 kw_repeated = &IdentTable.get("repeated"); 560 kw_required = &IdentTable.get("required"); 561 kw_returns = &IdentTable.get("returns"); 562 563 kw_signals = &IdentTable.get("signals"); 564 kw_qsignals = &IdentTable.get("Q_SIGNALS"); 565 kw_slots = &IdentTable.get("slots"); 566 kw_qslots = &IdentTable.get("Q_SLOTS"); 567 } 568 569 // Context sensitive keywords. 570 IdentifierInfo *kw_final; 571 IdentifierInfo *kw_override; 572 IdentifierInfo *kw_in; 573 IdentifierInfo *kw_CF_ENUM; 574 IdentifierInfo *kw_CF_OPTIONS; 575 IdentifierInfo *kw_NS_ENUM; 576 IdentifierInfo *kw_NS_OPTIONS; 577 IdentifierInfo *kw___except; 578 579 // JavaScript keywords. 580 IdentifierInfo *kw_finally; 581 IdentifierInfo *kw_function; 582 IdentifierInfo *kw_import; 583 IdentifierInfo *kw_let; 584 IdentifierInfo *kw_var; 585 586 // Java keywords. 587 IdentifierInfo *kw_abstract; 588 IdentifierInfo *kw_assert; 589 IdentifierInfo *kw_extends; 590 IdentifierInfo *kw_implements; 591 IdentifierInfo *kw_instanceof; 592 IdentifierInfo *kw_interface; 593 IdentifierInfo *kw_native; 594 IdentifierInfo *kw_package; 595 IdentifierInfo *kw_synchronized; 596 IdentifierInfo *kw_throws; 597 598 // Pragma keywords. 599 IdentifierInfo *kw_mark; 600 601 // Proto keywords. 602 IdentifierInfo *kw_extend; 603 IdentifierInfo *kw_option; 604 IdentifierInfo *kw_optional; 605 IdentifierInfo *kw_repeated; 606 IdentifierInfo *kw_required; 607 IdentifierInfo *kw_returns; 608 609 // QT keywords. 610 IdentifierInfo *kw_signals; 611 IdentifierInfo *kw_qsignals; 612 IdentifierInfo *kw_slots; 613 IdentifierInfo *kw_qslots; 614 }; 615 616 } // namespace format 617 } // namespace clang 618 619 #endif 620