Home | History | Annotate | Download | only in Format 
      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_FORMAT_FORMAT_TOKEN_H
     17 #define LLVM_CLANG_FORMAT_FORMAT_TOKEN_H
     18 
     19 #include "clang/Basic/OperatorPrecedence.h"
     20 #include "clang/Format/Format.h"
     21 #include "clang/Lex/Lexer.h"
     22 #include <memory>
     23 
     24 namespace clang {
     25 namespace format {
     26 
     27 enum TokenType {
     28   TT_ArrayInitializerLSquare,
     29   TT_ArraySubscriptLSquare,
     30   TT_AttributeParen,
     31   TT_BinaryOperator,
     32   TT_BitFieldColon,
     33   TT_BlockComment,
     34   TT_CastRParen,
     35   TT_ConditionalExpr,
     36   TT_ConflictAlternative,
     37   TT_ConflictEnd,
     38   TT_ConflictStart,
     39   TT_CtorInitializerColon,
     40   TT_CtorInitializerComma,
     41   TT_DesignatedInitializerPeriod,
     42   TT_DictLiteral,
     43   TT_FunctionDeclarationName,
     44   TT_FunctionLBrace,
     45   TT_FunctionTypeLParen,
     46   TT_ImplicitStringLiteral,
     47   TT_InheritanceColon,
     48   TT_InlineASMColon,
     49   TT_LambdaLSquare,
     50   TT_LineComment,
     51   TT_ObjCBlockLBrace,
     52   TT_ObjCBlockLParen,
     53   TT_ObjCDecl,
     54   TT_ObjCForIn,
     55   TT_ObjCMethodExpr,
     56   TT_ObjCMethodSpecifier,
     57   TT_ObjCProperty,
     58   TT_OverloadedOperator,
     59   TT_OverloadedOperatorLParen,
     60   TT_PointerOrReference,
     61   TT_PureVirtualSpecifier,
     62   TT_RangeBasedForLoopColon,
     63   TT_RegexLiteral,
     64   TT_SelectorName,
     65   TT_StartOfName,
     66   TT_TemplateCloser,
     67   TT_TemplateOpener,
     68   TT_TrailingAnnotation,
     69   TT_TrailingReturnArrow,
     70   TT_TrailingUnaryOperator,
     71   TT_UnaryOperator,
     72   TT_Unknown
     73 };
     74 
     75 // Represents what type of block a set of braces open.
     76 enum BraceBlockKind {
     77   BK_Unknown,
     78   BK_Block,
     79   BK_BracedInit
     80 };
     81 
     82 // The packing kind of a function's parameters.
     83 enum ParameterPackingKind {
     84   PPK_BinPacked,
     85   PPK_OnePerLine,
     86   PPK_Inconclusive
     87 };
     88 
     89 enum FormatDecision {
     90   FD_Unformatted,
     91   FD_Continue,
     92   FD_Break
     93 };
     94 
     95 class TokenRole;
     96 class AnnotatedLine;
     97 
     98 /// \brief A wrapper around a \c Token storing information about the
     99 /// whitespace characters preceding it.
    100 struct FormatToken {
    101   FormatToken()
    102       : NewlinesBefore(0), HasUnescapedNewline(false), LastNewlineOffset(0),
    103         ColumnWidth(0), LastLineColumnWidth(0), IsMultiline(false),
    104         IsFirst(false), MustBreakBefore(false), IsUnterminatedLiteral(false),
    105         BlockKind(BK_Unknown), Type(TT_Unknown), SpacesRequiredBefore(0),
    106         CanBreakBefore(false), ClosesTemplateDeclaration(false),
    107         ParameterCount(0), BlockParameterCount(0),
    108         PackingKind(PPK_Inconclusive), TotalLength(0), UnbreakableTailLength(0),
    109         BindingStrength(0), NestingLevel(0), SplitPenalty(0),
    110         LongestObjCSelectorName(0), FakeRParens(0),
    111         StartsBinaryExpression(false), EndsBinaryExpression(false),
    112         OperatorIndex(0), LastOperator(false),
    113         PartOfMultiVariableDeclStmt(false), IsForEachMacro(false),
    114         MatchingParen(nullptr), Previous(nullptr), Next(nullptr),
    115         Decision(FD_Unformatted), Finalized(false) {}
    116 
    117   /// \brief The \c Token.
    118   Token Tok;
    119 
    120   /// \brief The number of newlines immediately before the \c Token.
    121   ///
    122   /// This can be used to determine what the user wrote in the original code
    123   /// and thereby e.g. leave an empty line between two function definitions.
    124   unsigned NewlinesBefore;
    125 
    126   /// \brief Whether there is at least one unescaped newline before the \c
    127   /// Token.
    128   bool HasUnescapedNewline;
    129 
    130   /// \brief The range of the whitespace immediately preceding the \c Token.
    131   SourceRange WhitespaceRange;
    132 
    133   /// \brief The offset just past the last '\n' in this token's leading
    134   /// whitespace (relative to \c WhiteSpaceStart). 0 if there is no '\n'.
    135   unsigned LastNewlineOffset;
    136 
    137   /// \brief The width of the non-whitespace parts of the token (or its first
    138   /// line for multi-line tokens) in columns.
    139   /// We need this to correctly measure number of columns a token spans.
    140   unsigned ColumnWidth;
    141 
    142   /// \brief Contains the width in columns of the last line of a multi-line
    143   /// token.
    144   unsigned LastLineColumnWidth;
    145 
    146   /// \brief Whether the token text contains newlines (escaped or not).
    147   bool IsMultiline;
    148 
    149   /// \brief Indicates that this is the first token.
    150   bool IsFirst;
    151 
    152   /// \brief Whether there must be a line break before this token.
    153   ///
    154   /// This happens for example when a preprocessor directive ended directly
    155   /// before the token.
    156   bool MustBreakBefore;
    157 
    158   /// \brief Returns actual token start location without leading escaped
    159   /// newlines and whitespace.
    160   ///
    161   /// This can be different to Tok.getLocation(), which includes leading escaped
    162   /// newlines.
    163   SourceLocation getStartOfNonWhitespace() const {
    164     return WhitespaceRange.getEnd();
    165   }
    166 
    167   /// \brief The raw text of the token.
    168   ///
    169   /// Contains the raw token text without leading whitespace and without leading
    170   /// escaped newlines.
    171   StringRef TokenText;
    172 
    173   /// \brief Set to \c true if this token is an unterminated literal.
    174   bool IsUnterminatedLiteral;
    175 
    176   /// \brief Contains the kind of block if this token is a brace.
    177   BraceBlockKind BlockKind;
    178 
    179   TokenType Type;
    180 
    181   /// \brief The number of spaces that should be inserted before this token.
    182   unsigned SpacesRequiredBefore;
    183 
    184   /// \brief \c true if it is allowed to break before this token.
    185   bool CanBreakBefore;
    186 
    187   bool ClosesTemplateDeclaration;
    188 
    189   /// \brief Number of parameters, if this is "(", "[" or "<".
    190   ///
    191   /// This is initialized to 1 as we don't need to distinguish functions with
    192   /// 0 parameters from functions with 1 parameter. Thus, we can simply count
    193   /// the number of commas.
    194   unsigned ParameterCount;
    195 
    196   /// \brief Number of parameters that are nested blocks,
    197   /// if this is "(", "[" or "<".
    198   unsigned BlockParameterCount;
    199 
    200   /// \brief A token can have a special role that can carry extra information
    201   /// about the token's formatting.
    202   std::unique_ptr<TokenRole> Role;
    203 
    204   /// \brief If this is an opening parenthesis, how are the parameters packed?
    205   ParameterPackingKind PackingKind;
    206 
    207   /// \brief The total length of the unwrapped line up to and including this
    208   /// token.
    209   unsigned TotalLength;
    210 
    211   /// \brief The original 0-based column of this token, including expanded tabs.
    212   /// The configured TabWidth is used as tab width.
    213   unsigned OriginalColumn;
    214 
    215   /// \brief The length of following tokens until the next natural split point,
    216   /// or the next token that can be broken.
    217   unsigned UnbreakableTailLength;
    218 
    219   // FIXME: Come up with a 'cleaner' concept.
    220   /// \brief The binding strength of a token. This is a combined value of
    221   /// operator precedence, parenthesis nesting, etc.
    222   unsigned BindingStrength;
    223 
    224   /// \brief The nesting level of this token, i.e. the number of surrounding (),
    225   /// [], {} or <>.
    226   unsigned NestingLevel;
    227 
    228   /// \brief Penalty for inserting a line break before this token.
    229   unsigned SplitPenalty;
    230 
    231   /// \brief If this is the first ObjC selector name in an ObjC method
    232   /// definition or call, this contains the length of the longest name.
    233   ///
    234   /// This being set to 0 means that the selectors should not be colon-aligned,
    235   /// e.g. because several of them are block-type.
    236   unsigned LongestObjCSelectorName;
    237 
    238   /// \brief Stores the number of required fake parentheses and the
    239   /// corresponding operator precedence.
    240   ///
    241   /// If multiple fake parentheses start at a token, this vector stores them in
    242   /// reverse order, i.e. inner fake parenthesis first.
    243   SmallVector<prec::Level, 4> FakeLParens;
    244   /// \brief Insert this many fake ) after this token for correct indentation.
    245   unsigned FakeRParens;
    246 
    247   /// \brief \c true if this token starts a binary expression, i.e. has at least
    248   /// one fake l_paren with a precedence greater than prec::Unknown.
    249   bool StartsBinaryExpression;
    250   /// \brief \c true if this token ends a binary expression.
    251   bool EndsBinaryExpression;
    252 
    253   /// \brief Is this is an operator (or "."/"->") in a sequence of operators
    254   /// with the same precedence, contains the 0-based operator index.
    255   unsigned OperatorIndex;
    256 
    257   /// \brief Is this the last operator (or "."/"->") in a sequence of operators
    258   /// with the same precedence?
    259   bool LastOperator;
    260 
    261   /// \brief Is this token part of a \c DeclStmt defining multiple variables?
    262   ///
    263   /// Only set if \c Type == \c TT_StartOfName.
    264   bool PartOfMultiVariableDeclStmt;
    265 
    266   /// \brief Is this a foreach macro?
    267   bool IsForEachMacro;
    268 
    269   bool is(tok::TokenKind Kind) const { return Tok.is(Kind); }
    270 
    271   bool isOneOf(tok::TokenKind K1, tok::TokenKind K2) const {
    272     return is(K1) || is(K2);
    273   }
    274 
    275   bool isOneOf(tok::TokenKind K1, tok::TokenKind K2, tok::TokenKind K3) const {
    276     return is(K1) || is(K2) || is(K3);
    277   }
    278 
    279   bool isOneOf(tok::TokenKind K1, tok::TokenKind K2, tok::TokenKind K3,
    280                tok::TokenKind K4, tok::TokenKind K5 = tok::NUM_TOKENS,
    281                tok::TokenKind K6 = tok::NUM_TOKENS,
    282                tok::TokenKind K7 = tok::NUM_TOKENS,
    283                tok::TokenKind K8 = tok::NUM_TOKENS,
    284                tok::TokenKind K9 = tok::NUM_TOKENS,
    285                tok::TokenKind K10 = tok::NUM_TOKENS,
    286                tok::TokenKind K11 = tok::NUM_TOKENS,
    287                tok::TokenKind K12 = tok::NUM_TOKENS) const {
    288     return is(K1) || is(K2) || is(K3) || is(K4) || is(K5) || is(K6) || is(K7) ||
    289            is(K8) || is(K9) || is(K10) || is(K11) || is(K12);
    290   }
    291 
    292   bool isNot(tok::TokenKind Kind) const { return Tok.isNot(Kind); }
    293   bool isStringLiteral() const { return tok::isStringLiteral(Tok.getKind()); }
    294 
    295   bool isObjCAtKeyword(tok::ObjCKeywordKind Kind) const {
    296     return Tok.isObjCAtKeyword(Kind);
    297   }
    298 
    299   bool isAccessSpecifier(bool ColonRequired = true) const {
    300     return isOneOf(tok::kw_public, tok::kw_protected, tok::kw_private) &&
    301            (!ColonRequired || (Next && Next->is(tok::colon)));
    302   }
    303 
    304   /// \brief Determine whether the token is a simple-type-specifier.
    305   bool isSimpleTypeSpecifier() const;
    306 
    307   bool isObjCAccessSpecifier() const {
    308     return is(tok::at) && Next && (Next->isObjCAtKeyword(tok::objc_public) ||
    309                                    Next->isObjCAtKeyword(tok::objc_protected) ||
    310                                    Next->isObjCAtKeyword(tok::objc_package) ||
    311                                    Next->isObjCAtKeyword(tok::objc_private));
    312   }
    313 
    314   /// \brief Returns whether \p Tok is ([{ or a template opening <.
    315   bool opensScope() const {
    316     return isOneOf(tok::l_paren, tok::l_brace, tok::l_square) ||
    317            Type == TT_TemplateOpener;
    318   }
    319   /// \brief Returns whether \p Tok is )]} or a template closing >.
    320   bool closesScope() const {
    321     return isOneOf(tok::r_paren, tok::r_brace, tok::r_square) ||
    322            Type == TT_TemplateCloser;
    323   }
    324 
    325   /// \brief Returns \c true if this is a "." or "->" accessing a member.
    326   bool isMemberAccess() const {
    327     return isOneOf(tok::arrow, tok::period, tok::arrowstar) &&
    328            Type != TT_DesignatedInitializerPeriod;
    329   }
    330 
    331   bool isUnaryOperator() const {
    332     switch (Tok.getKind()) {
    333     case tok::plus:
    334     case tok::plusplus:
    335     case tok::minus:
    336     case tok::minusminus:
    337     case tok::exclaim:
    338     case tok::tilde:
    339     case tok::kw_sizeof:
    340     case tok::kw_alignof:
    341       return true;
    342     default:
    343       return false;
    344     }
    345   }
    346 
    347   bool isBinaryOperator() const {
    348     // Comma is a binary operator, but does not behave as such wrt. formatting.
    349     return getPrecedence() > prec::Comma;
    350   }
    351 
    352   bool isTrailingComment() const {
    353     return is(tok::comment) && (!Next || Next->NewlinesBefore > 0);
    354   }
    355 
    356   prec::Level getPrecedence() const {
    357     return getBinOpPrecedence(Tok.getKind(), true, true);
    358   }
    359 
    360   /// \brief Returns the previous token ignoring comments.
    361   FormatToken *getPreviousNonComment() const {
    362     FormatToken *Tok = Previous;
    363     while (Tok && Tok->is(tok::comment))
    364       Tok = Tok->Previous;
    365     return Tok;
    366   }
    367 
    368   /// \brief Returns the next token ignoring comments.
    369   const FormatToken *getNextNonComment() const {
    370     const FormatToken *Tok = Next;
    371     while (Tok && Tok->is(tok::comment))
    372       Tok = Tok->Next;
    373     return Tok;
    374   }
    375 
    376   /// \brief Returns \c true if this tokens starts a block-type list, i.e. a
    377   /// list that should be indented with a block indent.
    378   bool opensBlockTypeList(const FormatStyle &Style) const {
    379     return Type == TT_ArrayInitializerLSquare ||
    380            (is(tok::l_brace) &&
    381             (BlockKind == BK_Block || Type == TT_DictLiteral ||
    382              !Style.Cpp11BracedListStyle));
    383   }
    384 
    385   /// \brief Same as opensBlockTypeList, but for the closing token.
    386   bool closesBlockTypeList(const FormatStyle &Style) const {
    387     return MatchingParen && MatchingParen->opensBlockTypeList(Style);
    388   }
    389 
    390   FormatToken *MatchingParen;
    391 
    392   FormatToken *Previous;
    393   FormatToken *Next;
    394 
    395   SmallVector<AnnotatedLine *, 1> Children;
    396 
    397   /// \brief Stores the formatting decision for the token once it was made.
    398   FormatDecision Decision;
    399 
    400   /// \brief If \c true, this token has been fully formatted (indented and
    401   /// potentially re-formatted inside), and we do not allow further formatting
    402   /// changes.
    403   bool Finalized;
    404 
    405 private:
    406   // Disallow copying.
    407   FormatToken(const FormatToken &) LLVM_DELETED_FUNCTION;
    408   void operator=(const FormatToken &) LLVM_DELETED_FUNCTION;
    409 };
    410 
    411 class ContinuationIndenter;
    412 struct LineState;
    413 
    414 class TokenRole {
    415 public:
    416   TokenRole(const FormatStyle &Style) : Style(Style) {}
    417   virtual ~TokenRole();
    418 
    419   /// \brief After the \c TokenAnnotator has finished annotating all the tokens,
    420   /// this function precomputes required information for formatting.
    421   virtual void precomputeFormattingInfos(const FormatToken *Token);
    422 
    423   /// \brief Apply the special formatting that the given role demands.
    424   ///
    425   /// Assumes that the token having this role is already formatted.
    426   ///
    427   /// Continues formatting from \p State leaving indentation to \p Indenter and
    428   /// returns the total penalty that this formatting incurs.
    429   virtual unsigned formatFromToken(LineState &State,
    430                                    ContinuationIndenter *Indenter,
    431                                    bool DryRun) {
    432     return 0;
    433   }
    434 
    435   /// \brief Same as \c formatFromToken, but assumes that the first token has
    436   /// already been set thereby deciding on the first line break.
    437   virtual unsigned formatAfterToken(LineState &State,
    438                                     ContinuationIndenter *Indenter,
    439                                     bool DryRun) {
    440     return 0;
    441   }
    442 
    443   /// \brief Notifies the \c Role that a comma was found.
    444   virtual void CommaFound(const FormatToken *Token) {}
    445 
    446 protected:
    447   const FormatStyle &Style;
    448 };
    449 
    450 class CommaSeparatedList : public TokenRole {
    451 public:
    452   CommaSeparatedList(const FormatStyle &Style)
    453       : TokenRole(Style), HasNestedBracedList(false) {}
    454 
    455   void precomputeFormattingInfos(const FormatToken *Token) override;
    456 
    457   unsigned formatAfterToken(LineState &State, ContinuationIndenter *Indenter,
    458                             bool DryRun) override;
    459 
    460   unsigned formatFromToken(LineState &State, ContinuationIndenter *Indenter,
    461                            bool DryRun) override;
    462 
    463   /// \brief Adds \p Token as the next comma to the \c CommaSeparated list.
    464   void CommaFound(const FormatToken *Token) override {
    465     Commas.push_back(Token);
    466   }
    467 
    468 private:
    469   /// \brief A struct that holds information on how to format a given list with
    470   /// a specific number of columns.
    471   struct ColumnFormat {
    472     /// \brief The number of columns to use.
    473     unsigned Columns;
    474 
    475     /// \brief The total width in characters.
    476     unsigned TotalWidth;
    477 
    478     /// \brief The number of lines required for this format.
    479     unsigned LineCount;
    480 
    481     /// \brief The size of each column in characters.
    482     SmallVector<unsigned, 8> ColumnSizes;
    483   };
    484 
    485   /// \brief Calculate which \c ColumnFormat fits best into
    486   /// \p RemainingCharacters.
    487   const ColumnFormat *getColumnFormat(unsigned RemainingCharacters) const;
    488 
    489   /// \brief The ordered \c FormatTokens making up the commas of this list.
    490   SmallVector<const FormatToken *, 8> Commas;
    491 
    492   /// \brief The length of each of the list's items in characters including the
    493   /// trailing comma.
    494   SmallVector<unsigned, 8> ItemLengths;
    495 
    496   /// \brief Precomputed formats that can be used for this list.
    497   SmallVector<ColumnFormat, 4> Formats;
    498 
    499   bool HasNestedBracedList;
    500 };
    501 
    502 } // namespace format
    503 } // namespace clang
    504 
    505 #endif // LLVM_CLANG_FORMAT_FORMAT_TOKEN_H
    506