Home | History | Annotate | Download | only in Format 
      1 //===--- FormatTokenLexer.cpp - Lex FormatTokens -------------*- 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 implements FormatTokenLexer, which tokenizes a source file
     12 /// into a FormatToken stream suitable for ClangFormat.
     13 ///
     14 //===----------------------------------------------------------------------===//
     15 
     16 #include "FormatTokenLexer.h"
     17 #include "FormatToken.h"
     18 #include "clang/Basic/SourceLocation.h"
     19 #include "clang/Basic/SourceManager.h"
     20 #include "clang/Format/Format.h"
     21 #include "llvm/Support/Regex.h"
     22 
     23 namespace clang {
     24 namespace format {
     25 
     26 FormatTokenLexer::FormatTokenLexer(const SourceManager &SourceMgr, FileID ID,
     27                                    const FormatStyle &Style,
     28                                    encoding::Encoding Encoding)
     29     : FormatTok(nullptr), IsFirstToken(true), GreaterStashed(false),
     30       LessStashed(false), Column(0), TrailingWhitespace(0),
     31       SourceMgr(SourceMgr), ID(ID), Style(Style),
     32       IdentTable(getFormattingLangOpts(Style)), Keywords(IdentTable),
     33       Encoding(Encoding), FirstInLineIndex(0), FormattingDisabled(false),
     34       MacroBlockBeginRegex(Style.MacroBlockBegin),
     35       MacroBlockEndRegex(Style.MacroBlockEnd) {
     36   Lex.reset(new Lexer(ID, SourceMgr.getBuffer(ID), SourceMgr,
     37                       getFormattingLangOpts(Style)));
     38   Lex->SetKeepWhitespaceMode(true);
     39 
     40   for (const std::string &ForEachMacro : Style.ForEachMacros)
     41     ForEachMacros.push_back(&IdentTable.get(ForEachMacro));
     42   std::sort(ForEachMacros.begin(), ForEachMacros.end());
     43 }
     44 
     45 ArrayRef<FormatToken *> FormatTokenLexer::lex() {
     46   assert(Tokens.empty());
     47   assert(FirstInLineIndex == 0);
     48   do {
     49     Tokens.push_back(getNextToken());
     50     if (Style.Language == FormatStyle::LK_JavaScript) {
     51       tryParseJSRegexLiteral();
     52       tryParseTemplateString();
     53     }
     54     tryMergePreviousTokens();
     55     if (Tokens.back()->NewlinesBefore > 0 || Tokens.back()->IsMultiline)
     56       FirstInLineIndex = Tokens.size() - 1;
     57   } while (Tokens.back()->Tok.isNot(tok::eof));
     58   return Tokens;
     59 }
     60 
     61 void FormatTokenLexer::tryMergePreviousTokens() {
     62   if (tryMerge_TMacro())
     63     return;
     64   if (tryMergeConflictMarkers())
     65     return;
     66   if (tryMergeLessLess())
     67     return;
     68 
     69   if (Style.Language == FormatStyle::LK_JavaScript) {
     70     static const tok::TokenKind JSIdentity[] = {tok::equalequal, tok::equal};
     71     static const tok::TokenKind JSNotIdentity[] = {tok::exclaimequal,
     72                                                    tok::equal};
     73     static const tok::TokenKind JSShiftEqual[] = {tok::greater, tok::greater,
     74                                                   tok::greaterequal};
     75     static const tok::TokenKind JSRightArrow[] = {tok::equal, tok::greater};
     76     // FIXME: Investigate what token type gives the correct operator priority.
     77     if (tryMergeTokens(JSIdentity, TT_BinaryOperator))
     78       return;
     79     if (tryMergeTokens(JSNotIdentity, TT_BinaryOperator))
     80       return;
     81     if (tryMergeTokens(JSShiftEqual, TT_BinaryOperator))
     82       return;
     83     if (tryMergeTokens(JSRightArrow, TT_JsFatArrow))
     84       return;
     85   }
     86 }
     87 
     88 bool FormatTokenLexer::tryMergeLessLess() {
     89   // Merge X,less,less,Y into X,lessless,Y unless X or Y is less.
     90   if (Tokens.size() < 3)
     91     return false;
     92 
     93   bool FourthTokenIsLess = false;
     94   if (Tokens.size() > 3)
     95     FourthTokenIsLess = (Tokens.end() - 4)[0]->is(tok::less);
     96 
     97   auto First = Tokens.end() - 3;
     98   if (First[2]->is(tok::less) || First[1]->isNot(tok::less) ||
     99       First[0]->isNot(tok::less) || FourthTokenIsLess)
    100     return false;
    101 
    102   // Only merge if there currently is no whitespace between the two "<".
    103   if (First[1]->WhitespaceRange.getBegin() !=
    104       First[1]->WhitespaceRange.getEnd())
    105     return false;
    106 
    107   First[0]->Tok.setKind(tok::lessless);
    108   First[0]->TokenText = "<<";
    109   First[0]->ColumnWidth += 1;
    110   Tokens.erase(Tokens.end() - 2);
    111   return true;
    112 }
    113 
    114 bool FormatTokenLexer::tryMergeTokens(ArrayRef<tok::TokenKind> Kinds,
    115                                       TokenType NewType) {
    116   if (Tokens.size() < Kinds.size())
    117     return false;
    118 
    119   SmallVectorImpl<FormatToken *>::const_iterator First =
    120       Tokens.end() - Kinds.size();
    121   if (!First[0]->is(Kinds[0]))
    122     return false;
    123   unsigned AddLength = 0;
    124   for (unsigned i = 1; i < Kinds.size(); ++i) {
    125     if (!First[i]->is(Kinds[i]) ||
    126         First[i]->WhitespaceRange.getBegin() !=
    127             First[i]->WhitespaceRange.getEnd())
    128       return false;
    129     AddLength += First[i]->TokenText.size();
    130   }
    131   Tokens.resize(Tokens.size() - Kinds.size() + 1);
    132   First[0]->TokenText = StringRef(First[0]->TokenText.data(),
    133                                   First[0]->TokenText.size() + AddLength);
    134   First[0]->ColumnWidth += AddLength;
    135   First[0]->Type = NewType;
    136   return true;
    137 }
    138 
    139 // Returns \c true if \p Tok can only be followed by an operand in JavaScript.
    140 bool FormatTokenLexer::precedesOperand(FormatToken *Tok) {
    141   // NB: This is not entirely correct, as an r_paren can introduce an operand
    142   // location in e.g. `if (foo) /bar/.exec(...);`. That is a rare enough
    143   // corner case to not matter in practice, though.
    144   return Tok->isOneOf(tok::period, tok::l_paren, tok::comma, tok::l_brace,
    145                       tok::r_brace, tok::l_square, tok::semi, tok::exclaim,
    146                       tok::colon, tok::question, tok::tilde) ||
    147          Tok->isOneOf(tok::kw_return, tok::kw_do, tok::kw_case, tok::kw_throw,
    148                       tok::kw_else, tok::kw_new, tok::kw_delete, tok::kw_void,
    149                       tok::kw_typeof, Keywords.kw_instanceof, Keywords.kw_in) ||
    150          Tok->isBinaryOperator();
    151 }
    152 
    153 bool FormatTokenLexer::canPrecedeRegexLiteral(FormatToken *Prev) {
    154   if (!Prev)
    155     return true;
    156 
    157   // Regex literals can only follow after prefix unary operators, not after
    158   // postfix unary operators. If the '++' is followed by a non-operand
    159   // introducing token, the slash here is the operand and not the start of a
    160   // regex.
    161   if (Prev->isOneOf(tok::plusplus, tok::minusminus))
    162     return (Tokens.size() < 3 || precedesOperand(Tokens[Tokens.size() - 3]));
    163 
    164   // The previous token must introduce an operand location where regex
    165   // literals can occur.
    166   if (!precedesOperand(Prev))
    167     return false;
    168 
    169   return true;
    170 }
    171 
    172 // Tries to parse a JavaScript Regex literal starting at the current token,
    173 // if that begins with a slash and is in a location where JavaScript allows
    174 // regex literals. Changes the current token to a regex literal and updates
    175 // its text if successful.
    176 void FormatTokenLexer::tryParseJSRegexLiteral() {
    177   FormatToken *RegexToken = Tokens.back();
    178   if (!RegexToken->isOneOf(tok::slash, tok::slashequal))
    179     return;
    180 
    181   FormatToken *Prev = nullptr;
    182   for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; ++I) {
    183     // NB: Because previous pointers are not initialized yet, this cannot use
    184     // Token.getPreviousNonComment.
    185     if ((*I)->isNot(tok::comment)) {
    186       Prev = *I;
    187       break;
    188     }
    189   }
    190 
    191   if (!canPrecedeRegexLiteral(Prev))
    192     return;
    193 
    194   // 'Manually' lex ahead in the current file buffer.
    195   const char *Offset = Lex->getBufferLocation();
    196   const char *RegexBegin = Offset - RegexToken->TokenText.size();
    197   StringRef Buffer = Lex->getBuffer();
    198   bool InCharacterClass = false;
    199   bool HaveClosingSlash = false;
    200   for (; !HaveClosingSlash && Offset != Buffer.end(); ++Offset) {
    201     // Regular expressions are terminated with a '/', which can only be
    202     // escaped using '\' or a character class between '[' and ']'.
    203     // See http://www.ecma-international.org/ecma-262/5.1/#sec-7.8.5.
    204     switch (*Offset) {
    205     case '\\':
    206       // Skip the escaped character.
    207       ++Offset;
    208       break;
    209     case '[':
    210       InCharacterClass = true;
    211       break;
    212     case ']':
    213       InCharacterClass = false;
    214       break;
    215     case '/':
    216       if (!InCharacterClass)
    217         HaveClosingSlash = true;
    218       break;
    219     }
    220   }
    221 
    222   RegexToken->Type = TT_RegexLiteral;
    223   // Treat regex literals like other string_literals.
    224   RegexToken->Tok.setKind(tok::string_literal);
    225   RegexToken->TokenText = StringRef(RegexBegin, Offset - RegexBegin);
    226   RegexToken->ColumnWidth = RegexToken->TokenText.size();
    227 
    228   resetLexer(SourceMgr.getFileOffset(Lex->getSourceLocation(Offset)));
    229 }
    230 
    231 void FormatTokenLexer::tryParseTemplateString() {
    232   FormatToken *BacktickToken = Tokens.back();
    233   if (!BacktickToken->is(tok::unknown) || BacktickToken->TokenText != "`")
    234     return;
    235 
    236   // 'Manually' lex ahead in the current file buffer.
    237   const char *Offset = Lex->getBufferLocation();
    238   const char *TmplBegin = Offset - BacktickToken->TokenText.size(); // at "`"
    239   for (; Offset != Lex->getBuffer().end() && *Offset != '`'; ++Offset) {
    240     if (*Offset == '\\')
    241       ++Offset; // Skip the escaped character.
    242   }
    243 
    244   StringRef LiteralText(TmplBegin, Offset - TmplBegin + 1);
    245   BacktickToken->Type = TT_TemplateString;
    246   BacktickToken->Tok.setKind(tok::string_literal);
    247   BacktickToken->TokenText = LiteralText;
    248 
    249   // Adjust width for potentially multiline string literals.
    250   size_t FirstBreak = LiteralText.find('\n');
    251   StringRef FirstLineText = FirstBreak == StringRef::npos
    252                                 ? LiteralText
    253                                 : LiteralText.substr(0, FirstBreak);
    254   BacktickToken->ColumnWidth = encoding::columnWidthWithTabs(
    255       FirstLineText, BacktickToken->OriginalColumn, Style.TabWidth, Encoding);
    256   size_t LastBreak = LiteralText.rfind('\n');
    257   if (LastBreak != StringRef::npos) {
    258     BacktickToken->IsMultiline = true;
    259     unsigned StartColumn = 0; // The template tail spans the entire line.
    260     BacktickToken->LastLineColumnWidth = encoding::columnWidthWithTabs(
    261         LiteralText.substr(LastBreak + 1, LiteralText.size()), StartColumn,
    262         Style.TabWidth, Encoding);
    263   }
    264 
    265   resetLexer(SourceMgr.getFileOffset(Lex->getSourceLocation(Offset + 1)));
    266 }
    267 
    268 bool FormatTokenLexer::tryMerge_TMacro() {
    269   if (Tokens.size() < 4)
    270     return false;
    271   FormatToken *Last = Tokens.back();
    272   if (!Last->is(tok::r_paren))
    273     return false;
    274 
    275   FormatToken *String = Tokens[Tokens.size() - 2];
    276   if (!String->is(tok::string_literal) || String->IsMultiline)
    277     return false;
    278 
    279   if (!Tokens[Tokens.size() - 3]->is(tok::l_paren))
    280     return false;
    281 
    282   FormatToken *Macro = Tokens[Tokens.size() - 4];
    283   if (Macro->TokenText != "_T")
    284     return false;
    285 
    286   const char *Start = Macro->TokenText.data();
    287   const char *End = Last->TokenText.data() + Last->TokenText.size();
    288   String->TokenText = StringRef(Start, End - Start);
    289   String->IsFirst = Macro->IsFirst;
    290   String->LastNewlineOffset = Macro->LastNewlineOffset;
    291   String->WhitespaceRange = Macro->WhitespaceRange;
    292   String->OriginalColumn = Macro->OriginalColumn;
    293   String->ColumnWidth = encoding::columnWidthWithTabs(
    294       String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding);
    295   String->NewlinesBefore = Macro->NewlinesBefore;
    296   String->HasUnescapedNewline = Macro->HasUnescapedNewline;
    297 
    298   Tokens.pop_back();
    299   Tokens.pop_back();
    300   Tokens.pop_back();
    301   Tokens.back() = String;
    302   return true;
    303 }
    304 
    305 bool FormatTokenLexer::tryMergeConflictMarkers() {
    306   if (Tokens.back()->NewlinesBefore == 0 && Tokens.back()->isNot(tok::eof))
    307     return false;
    308 
    309   // Conflict lines look like:
    310   // <marker> <text from the vcs>
    311   // For example:
    312   // >>>>>>> /file/in/file/system at revision 1234
    313   //
    314   // We merge all tokens in a line that starts with a conflict marker
    315   // into a single token with a special token type that the unwrapped line
    316   // parser will use to correctly rebuild the underlying code.
    317 
    318   FileID ID;
    319   // Get the position of the first token in the line.
    320   unsigned FirstInLineOffset;
    321   std::tie(ID, FirstInLineOffset) = SourceMgr.getDecomposedLoc(
    322       Tokens[FirstInLineIndex]->getStartOfNonWhitespace());
    323   StringRef Buffer = SourceMgr.getBuffer(ID)->getBuffer();
    324   // Calculate the offset of the start of the current line.
    325   auto LineOffset = Buffer.rfind('\n', FirstInLineOffset);
    326   if (LineOffset == StringRef::npos) {
    327     LineOffset = 0;
    328   } else {
    329     ++LineOffset;
    330   }
    331 
    332   auto FirstSpace = Buffer.find_first_of(" \n", LineOffset);
    333   StringRef LineStart;
    334   if (FirstSpace == StringRef::npos) {
    335     LineStart = Buffer.substr(LineOffset);
    336   } else {
    337     LineStart = Buffer.substr(LineOffset, FirstSpace - LineOffset);
    338   }
    339 
    340   TokenType Type = TT_Unknown;
    341   if (LineStart == "<<<<<<<" || LineStart == ">>>>") {
    342     Type = TT_ConflictStart;
    343   } else if (LineStart == "|||||||" || LineStart == "=======" ||
    344              LineStart == "====") {
    345     Type = TT_ConflictAlternative;
    346   } else if (LineStart == ">>>>>>>" || LineStart == "<<<<") {
    347     Type = TT_ConflictEnd;
    348   }
    349 
    350   if (Type != TT_Unknown) {
    351     FormatToken *Next = Tokens.back();
    352 
    353     Tokens.resize(FirstInLineIndex + 1);
    354     // We do not need to build a complete token here, as we will skip it
    355     // during parsing anyway (as we must not touch whitespace around conflict
    356     // markers).
    357     Tokens.back()->Type = Type;
    358     Tokens.back()->Tok.setKind(tok::kw___unknown_anytype);
    359 
    360     Tokens.push_back(Next);
    361     return true;
    362   }
    363 
    364   return false;
    365 }
    366 
    367 FormatToken *FormatTokenLexer::getStashedToken() {
    368   // Create a synthesized second '>' or '<' token.
    369   Token Tok = FormatTok->Tok;
    370   StringRef TokenText = FormatTok->TokenText;
    371 
    372   unsigned OriginalColumn = FormatTok->OriginalColumn;
    373   FormatTok = new (Allocator.Allocate()) FormatToken;
    374   FormatTok->Tok = Tok;
    375   SourceLocation TokLocation =
    376       FormatTok->Tok.getLocation().getLocWithOffset(Tok.getLength() - 1);
    377   FormatTok->Tok.setLocation(TokLocation);
    378   FormatTok->WhitespaceRange = SourceRange(TokLocation, TokLocation);
    379   FormatTok->TokenText = TokenText;
    380   FormatTok->ColumnWidth = 1;
    381   FormatTok->OriginalColumn = OriginalColumn + 1;
    382 
    383   return FormatTok;
    384 }
    385 
    386 FormatToken *FormatTokenLexer::getNextToken() {
    387   if (GreaterStashed) {
    388     GreaterStashed = false;
    389     return getStashedToken();
    390   }
    391   if (LessStashed) {
    392     LessStashed = false;
    393     return getStashedToken();
    394   }
    395 
    396   FormatTok = new (Allocator.Allocate()) FormatToken;
    397   readRawToken(*FormatTok);
    398   SourceLocation WhitespaceStart =
    399       FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace);
    400   FormatTok->IsFirst = IsFirstToken;
    401   IsFirstToken = false;
    402 
    403   // Consume and record whitespace until we find a significant token.
    404   unsigned WhitespaceLength = TrailingWhitespace;
    405   while (FormatTok->Tok.is(tok::unknown)) {
    406     StringRef Text = FormatTok->TokenText;
    407     auto EscapesNewline = [&](int pos) {
    408       // A '\r' here is just part of '\r\n'. Skip it.
    409       if (pos >= 0 && Text[pos] == '\r')
    410         --pos;
    411       // See whether there is an odd number of '\' before this.
    412       unsigned count = 0;
    413       for (; pos >= 0; --pos, ++count)
    414         if (Text[pos] != '\\')
    415           break;
    416       return count & 1;
    417     };
    418     // FIXME: This miscounts tok:unknown tokens that are not just
    419     // whitespace, e.g. a '`' character.
    420     for (int i = 0, e = Text.size(); i != e; ++i) {
    421       switch (Text[i]) {
    422       case '\n':
    423         ++FormatTok->NewlinesBefore;
    424         FormatTok->HasUnescapedNewline = !EscapesNewline(i - 1);
    425         FormatTok->LastNewlineOffset = WhitespaceLength + i + 1;
    426         Column = 0;
    427         break;
    428       case '\r':
    429         FormatTok->LastNewlineOffset = WhitespaceLength + i + 1;
    430         Column = 0;
    431         break;
    432       case '\f':
    433       case '\v':
    434         Column = 0;
    435         break;
    436       case ' ':
    437         ++Column;
    438         break;
    439       case '\t':
    440         Column += Style.TabWidth - Column % Style.TabWidth;
    441         break;
    442       case '\\':
    443         if (i + 1 == e || (Text[i + 1] != '\r' && Text[i + 1] != '\n'))
    444           FormatTok->Type = TT_ImplicitStringLiteral;
    445         break;
    446       default:
    447         FormatTok->Type = TT_ImplicitStringLiteral;
    448         break;
    449       }
    450       if (FormatTok->Type == TT_ImplicitStringLiteral)
    451         break;
    452     }
    453 
    454     if (FormatTok->is(TT_ImplicitStringLiteral))
    455       break;
    456     WhitespaceLength += FormatTok->Tok.getLength();
    457 
    458     readRawToken(*FormatTok);
    459   }
    460 
    461   // In case the token starts with escaped newlines, we want to
    462   // take them into account as whitespace - this pattern is quite frequent
    463   // in macro definitions.
    464   // FIXME: Add a more explicit test.
    465   while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' &&
    466          FormatTok->TokenText[1] == '\n') {
    467     ++FormatTok->NewlinesBefore;
    468     WhitespaceLength += 2;
    469     FormatTok->LastNewlineOffset = 2;
    470     Column = 0;
    471     FormatTok->TokenText = FormatTok->TokenText.substr(2);
    472   }
    473 
    474   FormatTok->WhitespaceRange = SourceRange(
    475       WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength));
    476 
    477   FormatTok->OriginalColumn = Column;
    478 
    479   TrailingWhitespace = 0;
    480   if (FormatTok->Tok.is(tok::comment)) {
    481     // FIXME: Add the trimmed whitespace to Column.
    482     StringRef UntrimmedText = FormatTok->TokenText;
    483     FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f");
    484     TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size();
    485   } else if (FormatTok->Tok.is(tok::raw_identifier)) {
    486     IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText);
    487     FormatTok->Tok.setIdentifierInfo(&Info);
    488     FormatTok->Tok.setKind(Info.getTokenID());
    489     if (Style.Language == FormatStyle::LK_Java &&
    490         FormatTok->isOneOf(tok::kw_struct, tok::kw_union, tok::kw_delete,
    491                            tok::kw_operator)) {
    492       FormatTok->Tok.setKind(tok::identifier);
    493       FormatTok->Tok.setIdentifierInfo(nullptr);
    494     } else if (Style.Language == FormatStyle::LK_JavaScript &&
    495                FormatTok->isOneOf(tok::kw_struct, tok::kw_union,
    496                                   tok::kw_operator)) {
    497       FormatTok->Tok.setKind(tok::identifier);
    498       FormatTok->Tok.setIdentifierInfo(nullptr);
    499     }
    500   } else if (FormatTok->Tok.is(tok::greatergreater)) {
    501     FormatTok->Tok.setKind(tok::greater);
    502     FormatTok->TokenText = FormatTok->TokenText.substr(0, 1);
    503     GreaterStashed = true;
    504   } else if (FormatTok->Tok.is(tok::lessless)) {
    505     FormatTok->Tok.setKind(tok::less);
    506     FormatTok->TokenText = FormatTok->TokenText.substr(0, 1);
    507     LessStashed = true;
    508   }
    509 
    510   // Now FormatTok is the next non-whitespace token.
    511 
    512   StringRef Text = FormatTok->TokenText;
    513   size_t FirstNewlinePos = Text.find('\n');
    514   if (FirstNewlinePos == StringRef::npos) {
    515     // FIXME: ColumnWidth actually depends on the start column, we need to
    516     // take this into account when the token is moved.
    517     FormatTok->ColumnWidth =
    518         encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding);
    519     Column += FormatTok->ColumnWidth;
    520   } else {
    521     FormatTok->IsMultiline = true;
    522     // FIXME: ColumnWidth actually depends on the start column, we need to
    523     // take this into account when the token is moved.
    524     FormatTok->ColumnWidth = encoding::columnWidthWithTabs(
    525         Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding);
    526 
    527     // The last line of the token always starts in column 0.
    528     // Thus, the length can be precomputed even in the presence of tabs.
    529     FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs(
    530         Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth, Encoding);
    531     Column = FormatTok->LastLineColumnWidth;
    532   }
    533 
    534   if (Style.Language == FormatStyle::LK_Cpp) {
    535     if (!(Tokens.size() > 0 && Tokens.back()->Tok.getIdentifierInfo() &&
    536           Tokens.back()->Tok.getIdentifierInfo()->getPPKeywordID() ==
    537               tok::pp_define) &&
    538         std::find(ForEachMacros.begin(), ForEachMacros.end(),
    539                   FormatTok->Tok.getIdentifierInfo()) != ForEachMacros.end()) {
    540       FormatTok->Type = TT_ForEachMacro;
    541     } else if (FormatTok->is(tok::identifier)) {
    542       if (MacroBlockBeginRegex.match(Text)) {
    543         FormatTok->Type = TT_MacroBlockBegin;
    544       } else if (MacroBlockEndRegex.match(Text)) {
    545         FormatTok->Type = TT_MacroBlockEnd;
    546       }
    547     }
    548   }
    549 
    550   return FormatTok;
    551 }
    552 
    553 void FormatTokenLexer::readRawToken(FormatToken &Tok) {
    554   Lex->LexFromRawLexer(Tok.Tok);
    555   Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()),
    556                             Tok.Tok.getLength());
    557   // For formatting, treat unterminated string literals like normal string
    558   // literals.
    559   if (Tok.is(tok::unknown)) {
    560     if (!Tok.TokenText.empty() && Tok.TokenText[0] == '"') {
    561       Tok.Tok.setKind(tok::string_literal);
    562       Tok.IsUnterminatedLiteral = true;
    563     } else if (Style.Language == FormatStyle::LK_JavaScript &&
    564                Tok.TokenText == "''") {
    565       Tok.Tok.setKind(tok::string_literal);
    566     }
    567   }
    568 
    569   if (Style.Language == FormatStyle::LK_JavaScript &&
    570       Tok.is(tok::char_constant)) {
    571     Tok.Tok.setKind(tok::string_literal);
    572   }
    573 
    574   if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format on" ||
    575                                Tok.TokenText == "/* clang-format on */")) {
    576     FormattingDisabled = false;
    577   }
    578 
    579   Tok.Finalized = FormattingDisabled;
    580 
    581   if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format off" ||
    582                                Tok.TokenText == "/* clang-format off */")) {
    583     FormattingDisabled = true;
    584   }
    585 }
    586 
    587 void FormatTokenLexer::resetLexer(unsigned Offset) {
    588   StringRef Buffer = SourceMgr.getBufferData(ID);
    589   Lex.reset(new Lexer(SourceMgr.getLocForStartOfFile(ID),
    590                       getFormattingLangOpts(Style), Buffer.begin(),
    591                       Buffer.begin() + Offset, Buffer.end()));
    592   Lex->SetKeepWhitespaceMode(true);
    593   TrailingWhitespace = 0;
    594 }
    595 
    596 } // namespace format
    597 } // namespace clang
    598