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), StateStack({LexerState::NORMAL}),
30       Column(0), TrailingWhitespace(0), SourceMgr(SourceMgr), ID(ID),
31       Style(Style), IdentTable(getFormattingLangOpts(Style)),
32       Keywords(IdentTable), Encoding(Encoding), FirstInLineIndex(0),
33       FormattingDisabled(false), MacroBlockBeginRegex(Style.MacroBlockBegin),
34       MacroBlockEndRegex(Style.MacroBlockEnd) {
35   Lex.reset(new Lexer(ID, SourceMgr.getBuffer(ID), SourceMgr,
36                       getFormattingLangOpts(Style)));
37   Lex->SetKeepWhitespaceMode(true);
38 
39   for (const std::string &ForEachMacro : Style.ForEachMacros)
40     ForEachMacros.push_back(&IdentTable.get(ForEachMacro));
41   std::sort(ForEachMacros.begin(), ForEachMacros.end());
42 }
43 
44 ArrayRef<FormatToken *> FormatTokenLexer::lex() {
45   assert(Tokens.empty());
46   assert(FirstInLineIndex == 0);
47   do {
48     Tokens.push_back(getNextToken());
49     if (Style.Language == FormatStyle::LK_JavaScript) {
50       tryParseJSRegexLiteral();
51       handleTemplateStrings();
52     }
53     tryMergePreviousTokens();
54     if (Tokens.back()->NewlinesBefore > 0 || Tokens.back()->IsMultiline)
55       FirstInLineIndex = Tokens.size() - 1;
56   } while (Tokens.back()->Tok.isNot(tok::eof));
57   return Tokens;
58 }
59 
60 void FormatTokenLexer::tryMergePreviousTokens() {
61   if (tryMerge_TMacro())
62     return;
63   if (tryMergeConflictMarkers())
64     return;
65   if (tryMergeLessLess())
66     return;
67 
68   if (Style.Language == FormatStyle::LK_JavaScript) {
69     static const tok::TokenKind JSIdentity[] = {tok::equalequal, tok::equal};
70     static const tok::TokenKind JSNotIdentity[] = {tok::exclaimequal,
71                                                    tok::equal};
72     static const tok::TokenKind JSShiftEqual[] = {tok::greater, tok::greater,
73                                                   tok::greaterequal};
74     static const tok::TokenKind JSRightArrow[] = {tok::equal, tok::greater};
75     // FIXME: Investigate what token type gives the correct operator priority.
76     if (tryMergeTokens(JSIdentity, TT_BinaryOperator))
77       return;
78     if (tryMergeTokens(JSNotIdentity, TT_BinaryOperator))
79       return;
80     if (tryMergeTokens(JSShiftEqual, TT_BinaryOperator))
81       return;
82     if (tryMergeTokens(JSRightArrow, TT_JsFatArrow))
83       return;
84   }
85 }
86 
87 bool FormatTokenLexer::tryMergeLessLess() {
88   // Merge X,less,less,Y into X,lessless,Y unless X or Y is less.
89   if (Tokens.size() < 3)
90     return false;
91 
92   bool FourthTokenIsLess = false;
93   if (Tokens.size() > 3)
94     FourthTokenIsLess = (Tokens.end() - 4)[0]->is(tok::less);
95 
96   auto First = Tokens.end() - 3;
97   if (First[2]->is(tok::less) || First[1]->isNot(tok::less) ||
98       First[0]->isNot(tok::less) || FourthTokenIsLess)
99     return false;
100 
101   // Only merge if there currently is no whitespace between the two "<".
102   if (First[1]->WhitespaceRange.getBegin() !=
103       First[1]->WhitespaceRange.getEnd())
104     return false;
105 
106   First[0]->Tok.setKind(tok::lessless);
107   First[0]->TokenText = "<<";
108   First[0]->ColumnWidth += 1;
109   Tokens.erase(Tokens.end() - 2);
110   return true;
111 }
112 
113 bool FormatTokenLexer::tryMergeTokens(ArrayRef<tok::TokenKind> Kinds,
114                                       TokenType NewType) {
115   if (Tokens.size() < Kinds.size())
116     return false;
117 
118   SmallVectorImpl<FormatToken *>::const_iterator First =
119       Tokens.end() - Kinds.size();
120   if (!First[0]->is(Kinds[0]))
121     return false;
122   unsigned AddLength = 0;
123   for (unsigned i = 1; i < Kinds.size(); ++i) {
124     if (!First[i]->is(Kinds[i]) ||
125         First[i]->WhitespaceRange.getBegin() !=
126             First[i]->WhitespaceRange.getEnd())
127       return false;
128     AddLength += First[i]->TokenText.size();
129   }
130   Tokens.resize(Tokens.size() - Kinds.size() + 1);
131   First[0]->TokenText = StringRef(First[0]->TokenText.data(),
132                                   First[0]->TokenText.size() + AddLength);
133   First[0]->ColumnWidth += AddLength;
134   First[0]->Type = NewType;
135   return true;
136 }
137 
138 // Returns \c true if \p Tok can only be followed by an operand in JavaScript.
139 bool FormatTokenLexer::precedesOperand(FormatToken *Tok) {
140   // NB: This is not entirely correct, as an r_paren can introduce an operand
141   // location in e.g. `if (foo) /bar/.exec(...);`. That is a rare enough
142   // corner case to not matter in practice, though.
143   return Tok->isOneOf(tok::period, tok::l_paren, tok::comma, tok::l_brace,
144                       tok::r_brace, tok::l_square, tok::semi, tok::exclaim,
145                       tok::colon, tok::question, tok::tilde) ||
146          Tok->isOneOf(tok::kw_return, tok::kw_do, tok::kw_case, tok::kw_throw,
147                       tok::kw_else, tok::kw_new, tok::kw_delete, tok::kw_void,
148                       tok::kw_typeof, Keywords.kw_instanceof, Keywords.kw_in) ||
149          Tok->isBinaryOperator();
150 }
151 
152 bool FormatTokenLexer::canPrecedeRegexLiteral(FormatToken *Prev) {
153   if (!Prev)
154     return true;
155 
156   // Regex literals can only follow after prefix unary operators, not after
157   // postfix unary operators. If the '++' is followed by a non-operand
158   // introducing token, the slash here is the operand and not the start of a
159   // regex.
160   // `!` is an unary prefix operator, but also a post-fix operator that casts
161   // away nullability, so the same check applies.
162   if (Prev->isOneOf(tok::plusplus, tok::minusminus, tok::exclaim))
163     return (Tokens.size() < 3 || precedesOperand(Tokens[Tokens.size() - 3]));
164 
165   // The previous token must introduce an operand location where regex
166   // literals can occur.
167   if (!precedesOperand(Prev))
168     return false;
169 
170   return true;
171 }
172 
173 // Tries to parse a JavaScript Regex literal starting at the current token,
174 // if that begins with a slash and is in a location where JavaScript allows
175 // regex literals. Changes the current token to a regex literal and updates
176 // its text if successful.
177 void FormatTokenLexer::tryParseJSRegexLiteral() {
178   FormatToken *RegexToken = Tokens.back();
179   if (!RegexToken->isOneOf(tok::slash, tok::slashequal))
180     return;
181 
182   FormatToken *Prev = nullptr;
183   for (auto I = Tokens.rbegin() + 1, E = Tokens.rend(); I != E; ++I) {
184     // NB: Because previous pointers are not initialized yet, this cannot use
185     // Token.getPreviousNonComment.
186     if ((*I)->isNot(tok::comment)) {
187       Prev = *I;
188       break;
189     }
190   }
191 
192   if (!canPrecedeRegexLiteral(Prev))
193     return;
194 
195   // 'Manually' lex ahead in the current file buffer.
196   const char *Offset = Lex->getBufferLocation();
197   const char *RegexBegin = Offset - RegexToken->TokenText.size();
198   StringRef Buffer = Lex->getBuffer();
199   bool InCharacterClass = false;
200   bool HaveClosingSlash = false;
201   for (; !HaveClosingSlash && Offset != Buffer.end(); ++Offset) {
202     // Regular expressions are terminated with a '/', which can only be
203     // escaped using '\' or a character class between '[' and ']'.
204     // See http://www.ecma-international.org/ecma-262/5.1/#sec-7.8.5.
205     switch (*Offset) {
206     case '\\':
207       // Skip the escaped character.
208       ++Offset;
209       break;
210     case '[':
211       InCharacterClass = true;
212       break;
213     case ']':
214       InCharacterClass = false;
215       break;
216     case '/':
217       if (!InCharacterClass)
218         HaveClosingSlash = true;
219       break;
220     }
221   }
222 
223   RegexToken->Type = TT_RegexLiteral;
224   // Treat regex literals like other string_literals.
225   RegexToken->Tok.setKind(tok::string_literal);
226   RegexToken->TokenText = StringRef(RegexBegin, Offset - RegexBegin);
227   RegexToken->ColumnWidth = RegexToken->TokenText.size();
228 
229   resetLexer(SourceMgr.getFileOffset(Lex->getSourceLocation(Offset)));
230 }
231 
232 void FormatTokenLexer::handleTemplateStrings() {
233   FormatToken *BacktickToken = Tokens.back();
234 
235   if (BacktickToken->is(tok::l_brace)) {
236     StateStack.push(LexerState::NORMAL);
237     return;
238   }
239   if (BacktickToken->is(tok::r_brace)) {
240     if (StateStack.size() == 1)
241       return;
242     StateStack.pop();
243     if (StateStack.top() != LexerState::TEMPLATE_STRING)
244       return;
245     // If back in TEMPLATE_STRING, fallthrough and continue parsing the
246   } else if (BacktickToken->is(tok::unknown) &&
247              BacktickToken->TokenText == "`") {
248     StateStack.push(LexerState::TEMPLATE_STRING);
249   } else {
250     return; // Not actually a template
251   }
252 
253   // 'Manually' lex ahead in the current file buffer.
254   const char *Offset = Lex->getBufferLocation();
255   const char *TmplBegin = Offset - BacktickToken->TokenText.size(); // at "`"
256   for (; Offset != Lex->getBuffer().end(); ++Offset) {
257     if (Offset[0] == '`') {
258       StateStack.pop();
259       break;
260     }
261     if (Offset[0] == '\\') {
262       ++Offset; // Skip the escaped character.
263     } else if (Offset + 1 < Lex->getBuffer().end() && Offset[0] == '$' &&
264                Offset[1] == '{') {
265       // '${' introduces an expression interpolation in the template string.
266       StateStack.push(LexerState::NORMAL);
267       ++Offset;
268       break;
269     }
270   }
271 
272   StringRef LiteralText(TmplBegin, Offset - TmplBegin + 1);
273   BacktickToken->Type = TT_TemplateString;
274   BacktickToken->Tok.setKind(tok::string_literal);
275   BacktickToken->TokenText = LiteralText;
276 
277   // Adjust width for potentially multiline string literals.
278   size_t FirstBreak = LiteralText.find('\n');
279   StringRef FirstLineText = FirstBreak == StringRef::npos
280                                 ? LiteralText
281                                 : LiteralText.substr(0, FirstBreak);
282   BacktickToken->ColumnWidth = encoding::columnWidthWithTabs(
283       FirstLineText, BacktickToken->OriginalColumn, Style.TabWidth, Encoding);
284   size_t LastBreak = LiteralText.rfind('\n');
285   if (LastBreak != StringRef::npos) {
286     BacktickToken->IsMultiline = true;
287     unsigned StartColumn = 0; // The template tail spans the entire line.
288     BacktickToken->LastLineColumnWidth = encoding::columnWidthWithTabs(
289         LiteralText.substr(LastBreak + 1, LiteralText.size()), StartColumn,
290         Style.TabWidth, Encoding);
291   }
292 
293   SourceLocation loc = Offset < Lex->getBuffer().end()
294                            ? Lex->getSourceLocation(Offset + 1)
295                            : SourceMgr.getLocForEndOfFile(ID);
296   resetLexer(SourceMgr.getFileOffset(loc));
297 }
298 
299 bool FormatTokenLexer::tryMerge_TMacro() {
300   if (Tokens.size() < 4)
301     return false;
302   FormatToken *Last = Tokens.back();
303   if (!Last->is(tok::r_paren))
304     return false;
305 
306   FormatToken *String = Tokens[Tokens.size() - 2];
307   if (!String->is(tok::string_literal) || String->IsMultiline)
308     return false;
309 
310   if (!Tokens[Tokens.size() - 3]->is(tok::l_paren))
311     return false;
312 
313   FormatToken *Macro = Tokens[Tokens.size() - 4];
314   if (Macro->TokenText != "_T")
315     return false;
316 
317   const char *Start = Macro->TokenText.data();
318   const char *End = Last->TokenText.data() + Last->TokenText.size();
319   String->TokenText = StringRef(Start, End - Start);
320   String->IsFirst = Macro->IsFirst;
321   String->LastNewlineOffset = Macro->LastNewlineOffset;
322   String->WhitespaceRange = Macro->WhitespaceRange;
323   String->OriginalColumn = Macro->OriginalColumn;
324   String->ColumnWidth = encoding::columnWidthWithTabs(
325       String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding);
326   String->NewlinesBefore = Macro->NewlinesBefore;
327   String->HasUnescapedNewline = Macro->HasUnescapedNewline;
328 
329   Tokens.pop_back();
330   Tokens.pop_back();
331   Tokens.pop_back();
332   Tokens.back() = String;
333   return true;
334 }
335 
336 bool FormatTokenLexer::tryMergeConflictMarkers() {
337   if (Tokens.back()->NewlinesBefore == 0 && Tokens.back()->isNot(tok::eof))
338     return false;
339 
340   // Conflict lines look like:
341   // <marker> <text from the vcs>
342   // For example:
343   // >>>>>>> /file/in/file/system at revision 1234
344   //
345   // We merge all tokens in a line that starts with a conflict marker
346   // into a single token with a special token type that the unwrapped line
347   // parser will use to correctly rebuild the underlying code.
348 
349   FileID ID;
350   // Get the position of the first token in the line.
351   unsigned FirstInLineOffset;
352   std::tie(ID, FirstInLineOffset) = SourceMgr.getDecomposedLoc(
353       Tokens[FirstInLineIndex]->getStartOfNonWhitespace());
354   StringRef Buffer = SourceMgr.getBuffer(ID)->getBuffer();
355   // Calculate the offset of the start of the current line.
356   auto LineOffset = Buffer.rfind('\n', FirstInLineOffset);
357   if (LineOffset == StringRef::npos) {
358     LineOffset = 0;
359   } else {
360     ++LineOffset;
361   }
362 
363   auto FirstSpace = Buffer.find_first_of(" \n", LineOffset);
364   StringRef LineStart;
365   if (FirstSpace == StringRef::npos) {
366     LineStart = Buffer.substr(LineOffset);
367   } else {
368     LineStart = Buffer.substr(LineOffset, FirstSpace - LineOffset);
369   }
370 
371   TokenType Type = TT_Unknown;
372   if (LineStart == "<<<<<<<" || LineStart == ">>>>") {
373     Type = TT_ConflictStart;
374   } else if (LineStart == "|||||||" || LineStart == "=======" ||
375              LineStart == "====") {
376     Type = TT_ConflictAlternative;
377   } else if (LineStart == ">>>>>>>" || LineStart == "<<<<") {
378     Type = TT_ConflictEnd;
379   }
380 
381   if (Type != TT_Unknown) {
382     FormatToken *Next = Tokens.back();
383 
384     Tokens.resize(FirstInLineIndex + 1);
385     // We do not need to build a complete token here, as we will skip it
386     // during parsing anyway (as we must not touch whitespace around conflict
387     // markers).
388     Tokens.back()->Type = Type;
389     Tokens.back()->Tok.setKind(tok::kw___unknown_anytype);
390 
391     Tokens.push_back(Next);
392     return true;
393   }
394 
395   return false;
396 }
397 
398 FormatToken *FormatTokenLexer::getStashedToken() {
399   // Create a synthesized second '>' or '<' token.
400   Token Tok = FormatTok->Tok;
401   StringRef TokenText = FormatTok->TokenText;
402 
403   unsigned OriginalColumn = FormatTok->OriginalColumn;
404   FormatTok = new (Allocator.Allocate()) FormatToken;
405   FormatTok->Tok = Tok;
406   SourceLocation TokLocation =
407       FormatTok->Tok.getLocation().getLocWithOffset(Tok.getLength() - 1);
408   FormatTok->Tok.setLocation(TokLocation);
409   FormatTok->WhitespaceRange = SourceRange(TokLocation, TokLocation);
410   FormatTok->TokenText = TokenText;
411   FormatTok->ColumnWidth = 1;
412   FormatTok->OriginalColumn = OriginalColumn + 1;
413 
414   return FormatTok;
415 }
416 
417 FormatToken *FormatTokenLexer::getNextToken() {
418   if (StateStack.top() == LexerState::TOKEN_STASHED) {
419     StateStack.pop();
420     return getStashedToken();
421   }
422 
423   FormatTok = new (Allocator.Allocate()) FormatToken;
424   readRawToken(*FormatTok);
425   SourceLocation WhitespaceStart =
426       FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace);
427   FormatTok->IsFirst = IsFirstToken;
428   IsFirstToken = false;
429 
430   // Consume and record whitespace until we find a significant token.
431   unsigned WhitespaceLength = TrailingWhitespace;
432   while (FormatTok->Tok.is(tok::unknown)) {
433     StringRef Text = FormatTok->TokenText;
434     auto EscapesNewline = [&](int pos) {
435       // A '\r' here is just part of '\r\n'. Skip it.
436       if (pos >= 0 && Text[pos] == '\r')
437         --pos;
438       // See whether there is an odd number of '\' before this.
439       unsigned count = 0;
440       for (; pos >= 0; --pos, ++count)
441         if (Text[pos] != '\\')
442           break;
443       return count & 1;
444     };
445     // FIXME: This miscounts tok:unknown tokens that are not just
446     // whitespace, e.g. a '`' character.
447     for (int i = 0, e = Text.size(); i != e; ++i) {
448       switch (Text[i]) {
449       case '\n':
450         ++FormatTok->NewlinesBefore;
451         FormatTok->HasUnescapedNewline = !EscapesNewline(i - 1);
452         FormatTok->LastNewlineOffset = WhitespaceLength + i + 1;
453         Column = 0;
454         break;
455       case '\r':
456         FormatTok->LastNewlineOffset = WhitespaceLength + i + 1;
457         Column = 0;
458         break;
459       case '\f':
460       case '\v':
461         Column = 0;
462         break;
463       case ' ':
464         ++Column;
465         break;
466       case '\t':
467         Column += Style.TabWidth - Column % Style.TabWidth;
468         break;
469       case '\\':
470         if (i + 1 == e || (Text[i + 1] != '\r' && Text[i + 1] != '\n'))
471           FormatTok->Type = TT_ImplicitStringLiteral;
472         break;
473       default:
474         FormatTok->Type = TT_ImplicitStringLiteral;
475         break;
476       }
477       if (FormatTok->Type == TT_ImplicitStringLiteral)
478         break;
479     }
480 
481     if (FormatTok->is(TT_ImplicitStringLiteral))
482       break;
483     WhitespaceLength += FormatTok->Tok.getLength();
484 
485     readRawToken(*FormatTok);
486   }
487 
488   // In case the token starts with escaped newlines, we want to
489   // take them into account as whitespace - this pattern is quite frequent
490   // in macro definitions.
491   // FIXME: Add a more explicit test.
492   while (FormatTok->TokenText.size() > 1 && FormatTok->TokenText[0] == '\\' &&
493          FormatTok->TokenText[1] == '\n') {
494     ++FormatTok->NewlinesBefore;
495     WhitespaceLength += 2;
496     FormatTok->LastNewlineOffset = 2;
497     Column = 0;
498     FormatTok->TokenText = FormatTok->TokenText.substr(2);
499   }
500 
501   FormatTok->WhitespaceRange = SourceRange(
502       WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength));
503 
504   FormatTok->OriginalColumn = Column;
505 
506   TrailingWhitespace = 0;
507   if (FormatTok->Tok.is(tok::comment)) {
508     // FIXME: Add the trimmed whitespace to Column.
509     StringRef UntrimmedText = FormatTok->TokenText;
510     FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f");
511     TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size();
512   } else if (FormatTok->Tok.is(tok::raw_identifier)) {
513     IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText);
514     FormatTok->Tok.setIdentifierInfo(&Info);
515     FormatTok->Tok.setKind(Info.getTokenID());
516     if (Style.Language == FormatStyle::LK_Java &&
517         FormatTok->isOneOf(tok::kw_struct, tok::kw_union, tok::kw_delete,
518                            tok::kw_operator)) {
519       FormatTok->Tok.setKind(tok::identifier);
520       FormatTok->Tok.setIdentifierInfo(nullptr);
521     } else if (Style.Language == FormatStyle::LK_JavaScript &&
522                FormatTok->isOneOf(tok::kw_struct, tok::kw_union,
523                                   tok::kw_operator)) {
524       FormatTok->Tok.setKind(tok::identifier);
525       FormatTok->Tok.setIdentifierInfo(nullptr);
526     }
527   } else if (FormatTok->Tok.is(tok::greatergreater)) {
528     FormatTok->Tok.setKind(tok::greater);
529     FormatTok->TokenText = FormatTok->TokenText.substr(0, 1);
530     ++Column;
531     StateStack.push(LexerState::TOKEN_STASHED);
532   } else if (FormatTok->Tok.is(tok::lessless)) {
533     FormatTok->Tok.setKind(tok::less);
534     FormatTok->TokenText = FormatTok->TokenText.substr(0, 1);
535     ++Column;
536     StateStack.push(LexerState::TOKEN_STASHED);
537   }
538 
539   // Now FormatTok is the next non-whitespace token.
540 
541   StringRef Text = FormatTok->TokenText;
542   size_t FirstNewlinePos = Text.find('\n');
543   if (FirstNewlinePos == StringRef::npos) {
544     // FIXME: ColumnWidth actually depends on the start column, we need to
545     // take this into account when the token is moved.
546     FormatTok->ColumnWidth =
547         encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding);
548     Column += FormatTok->ColumnWidth;
549   } else {
550     FormatTok->IsMultiline = true;
551     // FIXME: ColumnWidth actually depends on the start column, we need to
552     // take this into account when the token is moved.
553     FormatTok->ColumnWidth = encoding::columnWidthWithTabs(
554         Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding);
555 
556     // The last line of the token always starts in column 0.
557     // Thus, the length can be precomputed even in the presence of tabs.
558     FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs(
559         Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth, Encoding);
560     Column = FormatTok->LastLineColumnWidth;
561   }
562 
563   if (Style.IsCpp()) {
564     if (!(Tokens.size() > 0 && Tokens.back()->Tok.getIdentifierInfo() &&
565           Tokens.back()->Tok.getIdentifierInfo()->getPPKeywordID() ==
566               tok::pp_define) &&
567         std::find(ForEachMacros.begin(), ForEachMacros.end(),
568                   FormatTok->Tok.getIdentifierInfo()) != ForEachMacros.end()) {
569       FormatTok->Type = TT_ForEachMacro;
570     } else if (FormatTok->is(tok::identifier)) {
571       if (MacroBlockBeginRegex.match(Text)) {
572         FormatTok->Type = TT_MacroBlockBegin;
573       } else if (MacroBlockEndRegex.match(Text)) {
574         FormatTok->Type = TT_MacroBlockEnd;
575       }
576     }
577   }
578 
579   return FormatTok;
580 }
581 
582 void FormatTokenLexer::readRawToken(FormatToken &Tok) {
583   Lex->LexFromRawLexer(Tok.Tok);
584   Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()),
585                             Tok.Tok.getLength());
586   // For formatting, treat unterminated string literals like normal string
587   // literals.
588   if (Tok.is(tok::unknown)) {
589     if (!Tok.TokenText.empty() && Tok.TokenText[0] == '"') {
590       Tok.Tok.setKind(tok::string_literal);
591       Tok.IsUnterminatedLiteral = true;
592     } else if (Style.Language == FormatStyle::LK_JavaScript &&
593                Tok.TokenText == "''") {
594       Tok.Tok.setKind(tok::string_literal);
595     }
596   }
597 
598   if (Style.Language == FormatStyle::LK_JavaScript &&
599       Tok.is(tok::char_constant)) {
600     Tok.Tok.setKind(tok::string_literal);
601   }
602 
603   if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format on" ||
604                                Tok.TokenText == "/* clang-format on */")) {
605     FormattingDisabled = false;
606   }
607 
608   Tok.Finalized = FormattingDisabled;
609 
610   if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format off" ||
611                                Tok.TokenText == "/* clang-format off */")) {
612     FormattingDisabled = true;
613   }
614 }
615 
616 void FormatTokenLexer::resetLexer(unsigned Offset) {
617   StringRef Buffer = SourceMgr.getBufferData(ID);
618   Lex.reset(new Lexer(SourceMgr.getLocForStartOfFile(ID),
619                       getFormattingLangOpts(Style), Buffer.begin(),
620                       Buffer.begin() + Offset, Buffer.end()));
621   Lex->SetKeepWhitespaceMode(true);
622   TrailingWhitespace = 0;
623 }
624 
625 } // namespace format
626 } // namespace clang
627