external/bsd/llvm/dist/clang/lib/Format/FormatToken.h

   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 enum TokenType {
  29   TT_ArrayInitializerLSquare,
  30   TT_ArraySubscriptLSquare,
  31   TT_AttributeParen,
  32   TT_BinaryOperator,
  33   TT_BitFieldColon,
  34   TT_BlockComment,
  35   TT_CastRParen,
  36   TT_ConditionalExpr,
  37   TT_ConflictAlternative,
  38   TT_ConflictEnd,
  39   TT_ConflictStart,
  40   TT_CtorInitializerColon,
  41   TT_CtorInitializerComma,
  42   TT_DesignatedInitializerPeriod,
  43   TT_DictLiteral,
  44   TT_FunctionDeclarationName,
  45   TT_FunctionLBrace,
  46   TT_FunctionTypeLParen,
  47   TT_ImplicitStringLiteral,
  48   TT_InheritanceColon,
  49   TT_InlineASMColon,
  50   TT_JavaAnnotation,
  51   TT_LambdaArrow,
  52   TT_LambdaLSquare,
  53   TT_LeadingJavaAnnotation,
  54   TT_LineComment,
  55   TT_ObjCBlockLBrace,
  56   TT_ObjCBlockLParen,
  57   TT_ObjCDecl,
  58   TT_ObjCForIn,
  59   TT_ObjCMethodExpr,
  60   TT_ObjCMethodSpecifier,
  61   TT_ObjCProperty,
  62   TT_OverloadedOperator,
  63   TT_OverloadedOperatorLParen,
  64   TT_PointerOrReference,
  65   TT_PureVirtualSpecifier,
  66   TT_RangeBasedForLoopColon,
  67   TT_RegexLiteral,
  68   TT_SelectorName,
  69   TT_StartOfName,
  70   TT_TemplateCloser,
  71   TT_TemplateOpener,
  72   TT_TrailingAnnotation,
  73   TT_TrailingReturnArrow,
  74   TT_TrailingUnaryOperator,
  75   TT_UnaryOperator,
  76   TT_Unknown
  77 };
  78
  79 // Represents what type of block a set of braces open.
  80 enum BraceBlockKind {
  81   BK_Unknown,
  82   BK_Block,
  83   BK_BracedInit
  84 };
  85
  86 // The packing kind of a function's parameters.
  87 enum ParameterPackingKind {
  88   PPK_BinPacked,
  89   PPK_OnePerLine,
  90   PPK_Inconclusive
  91 };
  92
  93 enum FormatDecision {
  94   FD_Unformatted,
  95   FD_Continue,
  96   FD_Break
  97 };
  98
  99 class TokenRole;
 100 class AnnotatedLine;
 101
 102 /// \brief A wrapper around a \c Token storing information about the
 103 /// whitespace characters preceding it.
 104 struct FormatToken {
 105   FormatToken()
 106       : NewlinesBefore(0), HasUnescapedNewline(false), LastNewlineOffset(0),
 107         ColumnWidth(0), LastLineColumnWidth(0), IsMultiline(false),
 108         IsFirst(false), MustBreakBefore(false), IsUnterminatedLiteral(false),
 109         BlockKind(BK_Unknown), Type(TT_Unknown), SpacesRequiredBefore(0),
 110         CanBreakBefore(false), ClosesTemplateDeclaration(false),
 111         ParameterCount(0), BlockParameterCount(0),
 112         PackingKind(PPK_Inconclusive), TotalLength(0), UnbreakableTailLength(0),
 113         BindingStrength(0), NestingLevel(0), SplitPenalty(0),
 114         LongestObjCSelectorName(0), FakeRParens(0),
 115         StartsBinaryExpression(false), EndsBinaryExpression(false),
 116         OperatorIndex(0), LastOperator(false),
 117         PartOfMultiVariableDeclStmt(false), IsForEachMacro(false),
 118         MatchingParen(nullptr), Previous(nullptr), Next(nullptr),
 119         Decision(FD_Unformatted), Finalized(false) {}
 120
 121   /// \brief The \c Token.
 122   Token Tok;
 123
 124   /// \brief The number of newlines immediately before the \c Token.
 125   ///
 126   /// This can be used to determine what the user wrote in the original code
 127   /// and thereby e.g. leave an empty line between two function definitions.
 128   unsigned NewlinesBefore;
 129
 130   /// \brief Whether there is at least one unescaped newline before the \c
 131   /// Token.
 132   bool HasUnescapedNewline;
 133
 134   /// \brief The range of the whitespace immediately preceding the \c Token.
 135   SourceRange WhitespaceRange;
 136
 137   /// \brief The offset just past the last '\n' in this token's leading
 138   /// whitespace (relative to \c WhiteSpaceStart). 0 if there is no '\n'.
 139   unsigned LastNewlineOffset;
 140
 141   /// \brief The width of the non-whitespace parts of the token (or its first
 142   /// line for multi-line tokens) in columns.
 143   /// We need this to correctly measure number of columns a token spans.
 144   unsigned ColumnWidth;
 145
 146   /// \brief Contains the width in columns of the last line of a multi-line
 147   /// token.
 148   unsigned LastLineColumnWidth;
 149
 150   /// \brief Whether the token text contains newlines (escaped or not).
 151   bool IsMultiline;
 152
 153   /// \brief Indicates that this is the first token.
 154   bool IsFirst;
 155
 156   /// \brief Whether there must be a line break before this token.
 157   ///
 158   /// This happens for example when a preprocessor directive ended directly
 159   /// before the token.
 160   bool MustBreakBefore;
 161
 162   /// \brief Returns actual token start location without leading escaped
 163   /// newlines and whitespace.
 164   ///
 165   /// This can be different to Tok.getLocation(), which includes leading escaped
 166   /// newlines.
 167   SourceLocation getStartOfNonWhitespace() const {
 168     return WhitespaceRange.getEnd();
 169   }
 170
 171   /// \brief The raw text of the token.
 172   ///
 173   /// Contains the raw token text without leading whitespace and without leading
 174   /// escaped newlines.
 175   StringRef TokenText;
 176
 177   /// \brief Set to \c true if this token is an unterminated literal.
 178   bool IsUnterminatedLiteral;
 179
 180   /// \brief Contains the kind of block if this token is a brace.
 181   BraceBlockKind BlockKind;
 182
 183   TokenType Type;
 184
 185   /// \brief The number of spaces that should be inserted before this token.
 186   unsigned SpacesRequiredBefore;
 187
 188   /// \brief \c true if it is allowed to break before this token.
 189   bool CanBreakBefore;
 190
 191   bool ClosesTemplateDeclaration;
 192
 193   /// \brief Number of parameters, if this is "(", "[" or "<".
 194   ///
 195   /// This is initialized to 1 as we don't need to distinguish functions with
 196   /// 0 parameters from functions with 1 parameter. Thus, we can simply count
 197   /// the number of commas.
 198   unsigned ParameterCount;
 199
 200   /// \brief Number of parameters that are nested blocks,
 201   /// if this is "(", "[" or "<".
 202   unsigned BlockParameterCount;
 203
 204   /// \brief A token can have a special role that can carry extra information
 205   /// about the token's formatting.
 206   std::unique_ptr<TokenRole> Role;
 207
 208   /// \brief If this is an opening parenthesis, how are the parameters packed?
 209   ParameterPackingKind PackingKind;
 210
 211   /// \brief The total length of the unwrapped line up to and including this
 212   /// token.
 213   unsigned TotalLength;
 214
 215   /// \brief The original 0-based column of this token, including expanded tabs.
 216   /// The configured TabWidth is used as tab width.
 217   unsigned OriginalColumn;
 218
 219   /// \brief The length of following tokens until the next natural split point,
 220   /// or the next token that can be broken.
 221   unsigned UnbreakableTailLength;
 222
 223   // FIXME: Come up with a 'cleaner' concept.
 224   /// \brief The binding strength of a token. This is a combined value of
 225   /// operator precedence, parenthesis nesting, etc.
 226   unsigned BindingStrength;
 227
 228   /// \brief The nesting level of this token, i.e. the number of surrounding (),
 229   /// [], {} or <>.
 230   unsigned NestingLevel;
 231
 232   /// \brief Penalty for inserting a line break before this token.
 233   unsigned SplitPenalty;
 234
 235   /// \brief If this is the first ObjC selector name in an ObjC method
 236   /// definition or call, this contains the length of the longest name.
 237   ///
 238   /// This being set to 0 means that the selectors should not be colon-aligned,
 239   /// e.g. because several of them are block-type.
 240   unsigned LongestObjCSelectorName;
 241
 242   /// \brief Stores the number of required fake parentheses and the
 243   /// corresponding operator precedence.
 244   ///
 245   /// If multiple fake parentheses start at a token, this vector stores them in
 246   /// reverse order, i.e. inner fake parenthesis first.
 247   SmallVector<prec::Level, 4> FakeLParens;
 248   /// \brief Insert this many fake ) after this token for correct indentation.
 249   unsigned FakeRParens;
 250
 251   /// \brief \c true if this token starts a binary expression, i.e. has at least
 252   /// one fake l_paren with a precedence greater than prec::Unknown.
 253   bool StartsBinaryExpression;
 254   /// \brief \c true if this token ends a binary expression.
 255   bool EndsBinaryExpression;
 256
 257   /// \brief Is this is an operator (or "."/"->") in a sequence of operators
 258   /// with the same precedence, contains the 0-based operator index.
 259   unsigned OperatorIndex;
 260
 261   /// \brief Is this the last operator (or "."/"->") in a sequence of operators
 262   /// with the same precedence?
 263   bool LastOperator;
 264
 265   /// \brief Is this token part of a \c DeclStmt defining multiple variables?
 266   ///
 267   /// Only set if \c Type == \c TT_StartOfName.
 268   bool PartOfMultiVariableDeclStmt;
 269
 270   /// \brief Is this a foreach macro?
 271   bool IsForEachMacro;
 272
 273   bool is(tok::TokenKind Kind) const { return Tok.is(Kind); }
 274   bool is(TokenType TT) const { return Type == TT; }
 275   bool is(const IdentifierInfo *II) const {
 276     return II && II == Tok.getIdentifierInfo();
 277   }
 278   template <typename A, typename B> bool isOneOf(A K1, B K2) const {
 279     return is(K1) || is(K2);
 280   }
 281   template <typename A, typename B, typename C>
 282   bool isOneOf(A K1, B K2, C K3) const {
 283     return is(K1) || is(K2) || is(K3);
 284   }
 285   template <typename A, typename B, typename C, typename D>
 286   bool isOneOf(A K1, B K2, C K3, D K4) const {
 287     return is(K1) || is(K2) || is(K3) || is(K4);
 288   }
 289   template <typename A, typename B, typename C, typename D, typename E>
 290   bool isOneOf(A K1, B K2, C K3, D K4, E K5) const {
 291     return is(K1) || is(K2) || is(K3) || is(K4) || is(K5);
 292   }
 293   template <typename T>
 294   bool isOneOf(T K1, T K2, T K3, T K4, T K5, T K6, T K7 = tok::NUM_TOKENS,
 295                T K8 = tok::NUM_TOKENS, T K9 = tok::NUM_TOKENS,
 296                T K10 = tok::NUM_TOKENS, T K11 = tok::NUM_TOKENS,
 297                T K12 = tok::NUM_TOKENS) const {
 298     return is(K1) || is(K2) || is(K3) || is(K4) || is(K5) || is(K6) || is(K7) ||
 299            is(K8) || is(K9) || is(K10) || is(K11) || is(K12);
 300   }
 301
 302   template <typename T> bool isNot(T Kind) const { return !is(Kind); }
 303
 304   bool isStringLiteral() const { return tok::isStringLiteral(Tok.getKind()); }
 305
 306   bool isObjCAtKeyword(tok::ObjCKeywordKind Kind) const {
 307     return Tok.isObjCAtKeyword(Kind);
 308   }
 309
 310   bool isAccessSpecifier(bool ColonRequired = true) const {
 311     return isOneOf(tok::kw_public, tok::kw_protected, tok::kw_private) &&
 312            (!ColonRequired || (Next && Next->is(tok::colon)));
 313   }
 314
 315   /// \brief Determine whether the token is a simple-type-specifier.
 316   bool isSimpleTypeSpecifier() const;
 317
 318   bool isObjCAccessSpecifier() const {
 319     return is(tok::at) && Next && (Next->isObjCAtKeyword(tok::objc_public) ||
 320                                    Next->isObjCAtKeyword(tok::objc_protected) ||
 321                                    Next->isObjCAtKeyword(tok::objc_package) ||
 322                                    Next->isObjCAtKeyword(tok::objc_private));
 323   }
 324
 325   /// \brief Returns whether \p Tok is ([{ or a template opening <.
 326   bool opensScope() const {
 327     return isOneOf(tok::l_paren, tok::l_brace, tok::l_square,
 328                    TT_TemplateOpener);
 329   }
 330   /// \brief Returns whether \p Tok is )]} or a template closing >.
 331   bool closesScope() const {
 332     return isOneOf(tok::r_paren, tok::r_brace, tok::r_square,
 333                    TT_TemplateCloser);
 334   }
 335
 336   /// \brief Returns \c true if this is a "." or "->" accessing a member.
 337   bool isMemberAccess() const {
 338     return isOneOf(tok::arrow, tok::period, tok::arrowstar) &&
 339            !isOneOf(TT_DesignatedInitializerPeriod, TT_TrailingReturnArrow);
 340   }
 341
 342   bool isUnaryOperator() const {
 343     switch (Tok.getKind()) {
 344     case tok::plus:
 345     case tok::plusplus:
 346     case tok::minus:
 347     case tok::minusminus:
 348     case tok::exclaim:
 349     case tok::tilde:
 350     case tok::kw_sizeof:
 351     case tok::kw_alignof:
 352       return true;
 353     default:
 354       return false;
 355     }
 356   }
 357
 358   bool isBinaryOperator() const {
 359     // Comma is a binary operator, but does not behave as such wrt. formatting.
 360     return getPrecedence() > prec::Comma;
 361   }
 362
 363   bool isTrailingComment() const {
 364     return is(tok::comment) &&
 365            (is(TT_LineComment) || !Next || Next->NewlinesBefore > 0);
 366   }
 367
 368   /// \brief Returns \c true if this is a keyword that can be used
 369   /// like a function call (e.g. sizeof, typeid, ...).
 370   bool isFunctionLikeKeyword() const {
 371     switch (Tok.getKind()) {
 372     case tok::kw_throw:
 373     case tok::kw_typeid:
 374     case tok::kw_return:
 375     case tok::kw_sizeof:
 376     case tok::kw_alignof:
 377     case tok::kw_alignas:
 378     case tok::kw_decltype:
 379     case tok::kw_noexcept:
 380     case tok::kw_static_assert:
 381     case tok::kw___attribute:
 382       return true;
 383     default:
 384       return false;
 385     }
 386   }
 387
 388   prec::Level getPrecedence() const {
 389     return getBinOpPrecedence(Tok.getKind(), true, true);
 390   }
 391
 392   /// \brief Returns the previous token ignoring comments.
 393   FormatToken *getPreviousNonComment() const {
 394     FormatToken *Tok = Previous;
 395     while (Tok && Tok->is(tok::comment))
 396       Tok = Tok->Previous;
 397     return Tok;
 398   }
 399
 400   /// \brief Returns the next token ignoring comments.
 401   const FormatToken *getNextNonComment() const {
 402     const FormatToken *Tok = Next;
 403     while (Tok && Tok->is(tok::comment))
 404       Tok = Tok->Next;
 405     return Tok;
 406   }
 407
 408   /// \brief Returns \c true if this tokens starts a block-type list, i.e. a
 409   /// list that should be indented with a block indent.
 410   bool opensBlockTypeList(const FormatStyle &Style) const {
 411     return is(TT_ArrayInitializerLSquare) ||
 412            (is(tok::l_brace) &&
 413             (BlockKind == BK_Block || is(TT_DictLiteral) ||
 414              (!Style.Cpp11BracedListStyle && NestingLevel == 0)));
 415   }
 416
 417   /// \brief Same as opensBlockTypeList, but for the closing token.
 418   bool closesBlockTypeList(const FormatStyle &Style) const {
 419     return MatchingParen && MatchingParen->opensBlockTypeList(Style);
 420   }
 421
 422   FormatToken *MatchingParen;
 423
 424   FormatToken *Previous;
 425   FormatToken *Next;
 426
 427   SmallVector<AnnotatedLine *, 1> Children;
 428
 429   /// \brief Stores the formatting decision for the token once it was made.
 430   FormatDecision Decision;
 431
 432   /// \brief If \c true, this token has been fully formatted (indented and
 433   /// potentially re-formatted inside), and we do not allow further formatting
 434   /// changes.
 435   bool Finalized;
 436
 437 private:
 438   // Disallow copying.
 439   FormatToken(const FormatToken &) LLVM_DELETED_FUNCTION;
 440   void operator=(const FormatToken &) LLVM_DELETED_FUNCTION;
 441 };
 442
 443 class ContinuationIndenter;
 444 struct LineState;
 445
 446 class TokenRole {
 447 public:
 448   TokenRole(const FormatStyle &Style) : Style(Style) {}
 449   virtual ~TokenRole();
 450
 451   /// \brief After the \c TokenAnnotator has finished annotating all the tokens,
 452   /// this function precomputes required information for formatting.
 453   virtual void precomputeFormattingInfos(const FormatToken *Token);
 454
 455   /// \brief Apply the special formatting that the given role demands.
 456   ///
 457   /// Assumes that the token having this role is already formatted.
 458   ///
 459   /// Continues formatting from \p State leaving indentation to \p Indenter and
 460   /// returns the total penalty that this formatting incurs.
 461   virtual unsigned formatFromToken(LineState &State,
 462                                    ContinuationIndenter *Indenter,
 463                                    bool DryRun) {
 464     return 0;
 465   }
 466
 467   /// \brief Same as \c formatFromToken, but assumes that the first token has
 468   /// already been set thereby deciding on the first line break.
 469   virtual unsigned formatAfterToken(LineState &State,
 470                                     ContinuationIndenter *Indenter,
 471                                     bool DryRun) {
 472     return 0;
 473   }
 474
 475   /// \brief Notifies the \c Role that a comma was found.
 476   virtual void CommaFound(const FormatToken *Token) {}
 477
 478 protected:
 479   const FormatStyle &Style;
 480 };
 481
 482 class CommaSeparatedList : public TokenRole {
 483 public:
 484   CommaSeparatedList(const FormatStyle &Style)
 485       : TokenRole(Style), HasNestedBracedList(false) {}
 486
 487   void precomputeFormattingInfos(const FormatToken *Token) override;
 488
 489   unsigned formatAfterToken(LineState &State, ContinuationIndenter *Indenter,
 490                             bool DryRun) override;
 491
 492   unsigned formatFromToken(LineState &State, ContinuationIndenter *Indenter,
 493                            bool DryRun) override;
 494
 495   /// \brief Adds \p Token as the next comma to the \c CommaSeparated list.
 496   void CommaFound(const FormatToken *Token) override {
 497     Commas.push_back(Token);
 498   }
 499
 500 private:
 501   /// \brief A struct that holds information on how to format a given list with
 502   /// a specific number of columns.
 503   struct ColumnFormat {
 504     /// \brief The number of columns to use.
 505     unsigned Columns;
 506
 507     /// \brief The total width in characters.
 508     unsigned TotalWidth;
 509
 510     /// \brief The number of lines required for this format.
 511     unsigned LineCount;
 512
 513     /// \brief The size of each column in characters.
 514     SmallVector<unsigned, 8> ColumnSizes;
 515   };
 516
 517   /// \brief Calculate which \c ColumnFormat fits best into
 518   /// \p RemainingCharacters.
 519   const ColumnFormat *getColumnFormat(unsigned RemainingCharacters) const;
 520
 521   /// \brief The ordered \c FormatTokens making up the commas of this list.
 522   SmallVector<const FormatToken *, 8> Commas;
 523
 524   /// \brief The length of each of the list's items in characters including the
 525   /// trailing comma.
 526   SmallVector<unsigned, 8> ItemLengths;
 527
 528   /// \brief Precomputed formats that can be used for this list.
 529   SmallVector<ColumnFormat, 4> Formats;
 530
 531   bool HasNestedBracedList;
 532 };
 533
 534 /// \brief Encapsulates keywords that are context sensitive or for languages not
 535 /// properly supported by Clang's lexer.
 536 struct AdditionalKeywords {
 537   AdditionalKeywords(IdentifierTable &IdentTable) {
 538     kw_in = &IdentTable.get("in");
 539     kw_CF_ENUM = &IdentTable.get("CF_ENUM");
 540     kw_CF_OPTIONS = &IdentTable.get("CF_OPTIONS");
 541     kw_NS_ENUM = &IdentTable.get("NS_ENUM");
 542     kw_NS_OPTIONS = &IdentTable.get("NS_OPTIONS");
 543
 544     kw_finally = &IdentTable.get("finally");
 545     kw_function = &IdentTable.get("function");
 546     kw_var = &IdentTable.get("var");
 547
 548     kw_abstract = &IdentTable.get("abstract");
 549     kw_extends = &IdentTable.get("extends");
 550     kw_final = &IdentTable.get("final");
 551     kw_implements = &IdentTable.get("implements");
 552     kw_instanceof = &IdentTable.get("instanceof");
 553     kw_interface = &IdentTable.get("interface");
 554     kw_native = &IdentTable.get("native");
 555     kw_package = &IdentTable.get("package");
 556     kw_synchronized = &IdentTable.get("synchronized");
 557     kw_throws = &IdentTable.get("throws");
 558
 559     kw_option = &IdentTable.get("option");
 560     kw_optional = &IdentTable.get("optional");
 561     kw_repeated = &IdentTable.get("repeated");
 562     kw_required = &IdentTable.get("required");
 563     kw_returns = &IdentTable.get("returns");
 564   }
 565
 566   // ObjC context sensitive keywords.
 567   IdentifierInfo *kw_in;
 568   IdentifierInfo *kw_CF_ENUM;
 569   IdentifierInfo *kw_CF_OPTIONS;
 570   IdentifierInfo *kw_NS_ENUM;
 571   IdentifierInfo *kw_NS_OPTIONS;
 572
 573   // JavaScript keywords.
 574   IdentifierInfo *kw_finally;
 575   IdentifierInfo *kw_function;
 576   IdentifierInfo *kw_var;
 577
 578   // Java keywords.
 579   IdentifierInfo *kw_abstract;
 580   IdentifierInfo *kw_extends;
 581   IdentifierInfo *kw_final;
 582   IdentifierInfo *kw_implements;
 583   IdentifierInfo *kw_instanceof;
 584   IdentifierInfo *kw_interface;
 585   IdentifierInfo *kw_native;
 586   IdentifierInfo *kw_package;
 587   IdentifierInfo *kw_synchronized;
 588   IdentifierInfo *kw_throws;
 589
 590   // Proto keywords.
 591   IdentifierInfo *kw_option;
 592   IdentifierInfo *kw_optional;
 593   IdentifierInfo *kw_repeated;
 594   IdentifierInfo *kw_required;
 595   IdentifierInfo *kw_returns;
 596 };
 597
 598 } // namespace format
 599 } // namespace clang
 600
 601 #endif