xref: /freebsd/contrib/llvm-project/clang/lib/Rewrite/HTMLRewrite.cpp (revision 0fca6ea1d4eea4c934cfff25ac9ee8ad6fe95583)
1 //== HTMLRewrite.cpp - Translate source code into prettified HTML --*- C++ -*-//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 //  This file defines the HTMLRewriter class, which is used to translate the
10 //  text of a source file into prettified HTML.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "clang/Rewrite/Core/HTMLRewrite.h"
15 #include "clang/Basic/SourceManager.h"
16 #include "clang/Lex/Preprocessor.h"
17 #include "clang/Lex/TokenConcatenation.h"
18 #include "clang/Rewrite/Core/Rewriter.h"
19 #include "llvm/ADT/SmallString.h"
20 #include "llvm/Support/ErrorHandling.h"
21 #include "llvm/Support/MemoryBuffer.h"
22 #include "llvm/Support/raw_ostream.h"
23 #include <memory>
24 
25 using namespace clang;
26 using namespace llvm;
27 using namespace html;
28 
29 /// HighlightRange - Highlight a range in the source code with the specified
30 /// start/end tags.  B/E must be in the same file.  This ensures that
31 /// start/end tags are placed at the start/end of each line if the range is
32 /// multiline.
HighlightRange(Rewriter & R,SourceLocation B,SourceLocation E,const char * StartTag,const char * EndTag,bool IsTokenRange)33 void html::HighlightRange(Rewriter &R, SourceLocation B, SourceLocation E,
34                           const char *StartTag, const char *EndTag,
35                           bool IsTokenRange) {
36   SourceManager &SM = R.getSourceMgr();
37   B = SM.getExpansionLoc(B);
38   E = SM.getExpansionLoc(E);
39   FileID FID = SM.getFileID(B);
40   assert(SM.getFileID(E) == FID && "B/E not in the same file!");
41 
42   unsigned BOffset = SM.getFileOffset(B);
43   unsigned EOffset = SM.getFileOffset(E);
44 
45   // Include the whole end token in the range.
46   if (IsTokenRange)
47     EOffset += Lexer::MeasureTokenLength(E, R.getSourceMgr(), R.getLangOpts());
48 
49   bool Invalid = false;
50   const char *BufferStart = SM.getBufferData(FID, &Invalid).data();
51   if (Invalid)
52     return;
53 
54   HighlightRange(R.getEditBuffer(FID), BOffset, EOffset,
55                  BufferStart, StartTag, EndTag);
56 }
57 
58 /// HighlightRange - This is the same as the above method, but takes
59 /// decomposed file locations.
HighlightRange(RewriteBuffer & RB,unsigned B,unsigned E,const char * BufferStart,const char * StartTag,const char * EndTag)60 void html::HighlightRange(RewriteBuffer &RB, unsigned B, unsigned E,
61                           const char *BufferStart,
62                           const char *StartTag, const char *EndTag) {
63   // Insert the tag at the absolute start/end of the range.
64   RB.InsertTextAfter(B, StartTag);
65   RB.InsertTextBefore(E, EndTag);
66 
67   // Scan the range to see if there is a \r or \n.  If so, and if the line is
68   // not blank, insert tags on that line as well.
69   bool HadOpenTag = true;
70 
71   unsigned LastNonWhiteSpace = B;
72   for (unsigned i = B; i != E; ++i) {
73     switch (BufferStart[i]) {
74     case '\r':
75     case '\n':
76       // Okay, we found a newline in the range.  If we have an open tag, we need
77       // to insert a close tag at the first non-whitespace before the newline.
78       if (HadOpenTag)
79         RB.InsertTextBefore(LastNonWhiteSpace+1, EndTag);
80 
81       // Instead of inserting an open tag immediately after the newline, we
82       // wait until we see a non-whitespace character.  This prevents us from
83       // inserting tags around blank lines, and also allows the open tag to
84       // be put *after* whitespace on a non-blank line.
85       HadOpenTag = false;
86       break;
87     case '\0':
88     case ' ':
89     case '\t':
90     case '\f':
91     case '\v':
92       // Ignore whitespace.
93       break;
94 
95     default:
96       // If there is no tag open, do it now.
97       if (!HadOpenTag) {
98         RB.InsertTextAfter(i, StartTag);
99         HadOpenTag = true;
100       }
101 
102       // Remember this character.
103       LastNonWhiteSpace = i;
104       break;
105     }
106   }
107 }
108 
109 namespace clang::html {
110 struct RelexRewriteCache {
111   // These structs mimic input arguments of HighlightRange().
112   struct Highlight {
113     SourceLocation B, E;
114     std::string StartTag, EndTag;
115     bool IsTokenRange;
116   };
117   struct RawHighlight {
118     unsigned B, E;
119     std::string StartTag, EndTag;
120   };
121 
122   // SmallVector isn't appropriate because these vectors are almost never small.
123   using HighlightList = std::vector<Highlight>;
124   using RawHighlightList = std::vector<RawHighlight>;
125 
126   DenseMap<FileID, RawHighlightList> SyntaxHighlights;
127   DenseMap<FileID, HighlightList> MacroHighlights;
128 };
129 } // namespace clang::html
130 
instantiateRelexRewriteCache()131 html::RelexRewriteCacheRef html::instantiateRelexRewriteCache() {
132   return std::make_shared<RelexRewriteCache>();
133 }
134 
EscapeText(Rewriter & R,FileID FID,bool EscapeSpaces,bool ReplaceTabs)135 void html::EscapeText(Rewriter &R, FileID FID,
136                       bool EscapeSpaces, bool ReplaceTabs) {
137 
138   llvm::MemoryBufferRef Buf = R.getSourceMgr().getBufferOrFake(FID);
139   const char* C = Buf.getBufferStart();
140   const char* FileEnd = Buf.getBufferEnd();
141 
142   assert (C <= FileEnd);
143 
144   RewriteBuffer &RB = R.getEditBuffer(FID);
145 
146   unsigned ColNo = 0;
147   for (unsigned FilePos = 0; C != FileEnd ; ++C, ++FilePos) {
148     switch (*C) {
149     default: ++ColNo; break;
150     case '\n':
151     case '\r':
152       ColNo = 0;
153       break;
154 
155     case ' ':
156       if (EscapeSpaces)
157         RB.ReplaceText(FilePos, 1, "&nbsp;");
158       ++ColNo;
159       break;
160     case '\f':
161       RB.ReplaceText(FilePos, 1, "<hr>");
162       ColNo = 0;
163       break;
164 
165     case '\t': {
166       if (!ReplaceTabs)
167         break;
168       unsigned NumSpaces = 8-(ColNo&7);
169       if (EscapeSpaces)
170         RB.ReplaceText(FilePos, 1,
171                        StringRef("&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;"
172                                        "&nbsp;&nbsp;&nbsp;", 6*NumSpaces));
173       else
174         RB.ReplaceText(FilePos, 1, StringRef("        ", NumSpaces));
175       ColNo += NumSpaces;
176       break;
177     }
178     case '<':
179       RB.ReplaceText(FilePos, 1, "&lt;");
180       ++ColNo;
181       break;
182 
183     case '>':
184       RB.ReplaceText(FilePos, 1, "&gt;");
185       ++ColNo;
186       break;
187 
188     case '&':
189       RB.ReplaceText(FilePos, 1, "&amp;");
190       ++ColNo;
191       break;
192     }
193   }
194 }
195 
EscapeText(StringRef s,bool EscapeSpaces,bool ReplaceTabs)196 std::string html::EscapeText(StringRef s, bool EscapeSpaces, bool ReplaceTabs) {
197 
198   unsigned len = s.size();
199   std::string Str;
200   llvm::raw_string_ostream os(Str);
201 
202   for (unsigned i = 0 ; i < len; ++i) {
203 
204     char c = s[i];
205     switch (c) {
206     default:
207       os << c; break;
208 
209     case ' ':
210       if (EscapeSpaces) os << "&nbsp;";
211       else os << ' ';
212       break;
213 
214     case '\t':
215       if (ReplaceTabs) {
216         if (EscapeSpaces)
217           for (unsigned i = 0; i < 4; ++i)
218             os << "&nbsp;";
219         else
220           for (unsigned i = 0; i < 4; ++i)
221             os << " ";
222       }
223       else
224         os << c;
225 
226       break;
227 
228     case '<': os << "&lt;"; break;
229     case '>': os << "&gt;"; break;
230     case '&': os << "&amp;"; break;
231     }
232   }
233 
234   return Str;
235 }
236 
AddLineNumber(RewriteBuffer & RB,unsigned LineNo,unsigned B,unsigned E)237 static void AddLineNumber(RewriteBuffer &RB, unsigned LineNo,
238                           unsigned B, unsigned E) {
239   SmallString<256> Str;
240   llvm::raw_svector_ostream OS(Str);
241 
242   OS << "<tr class=\"codeline\" data-linenumber=\"" << LineNo << "\">"
243      << "<td class=\"num\" id=\"LN" << LineNo << "\">" << LineNo
244      << "</td><td class=\"line\">";
245 
246   if (B == E) { // Handle empty lines.
247     OS << " </td></tr>";
248     RB.InsertTextBefore(B, OS.str());
249   } else {
250     RB.InsertTextBefore(B, OS.str());
251     RB.InsertTextBefore(E, "</td></tr>");
252   }
253 }
254 
AddLineNumbers(Rewriter & R,FileID FID)255 void html::AddLineNumbers(Rewriter& R, FileID FID) {
256 
257   llvm::MemoryBufferRef Buf = R.getSourceMgr().getBufferOrFake(FID);
258   const char* FileBeg = Buf.getBufferStart();
259   const char* FileEnd = Buf.getBufferEnd();
260   const char* C = FileBeg;
261   RewriteBuffer &RB = R.getEditBuffer(FID);
262 
263   assert (C <= FileEnd);
264 
265   unsigned LineNo = 0;
266   unsigned FilePos = 0;
267 
268   while (C != FileEnd) {
269 
270     ++LineNo;
271     unsigned LineStartPos = FilePos;
272     unsigned LineEndPos = FileEnd - FileBeg;
273 
274     assert (FilePos <= LineEndPos);
275     assert (C < FileEnd);
276 
277     // Scan until the newline (or end-of-file).
278 
279     while (C != FileEnd) {
280       char c = *C;
281       ++C;
282 
283       if (c == '\n') {
284         LineEndPos = FilePos++;
285         break;
286       }
287 
288       ++FilePos;
289     }
290 
291     AddLineNumber(RB, LineNo, LineStartPos, LineEndPos);
292   }
293 
294   // Add one big table tag that surrounds all of the code.
295   std::string s;
296   llvm::raw_string_ostream os(s);
297   os << "<table class=\"code\" data-fileid=\"" << FID.getHashValue() << "\">\n";
298   RB.InsertTextBefore(0, os.str());
299   RB.InsertTextAfter(FileEnd - FileBeg, "</table>");
300 }
301 
AddHeaderFooterInternalBuiltinCSS(Rewriter & R,FileID FID,StringRef title)302 void html::AddHeaderFooterInternalBuiltinCSS(Rewriter &R, FileID FID,
303                                              StringRef title) {
304 
305   llvm::MemoryBufferRef Buf = R.getSourceMgr().getBufferOrFake(FID);
306   const char* FileStart = Buf.getBufferStart();
307   const char* FileEnd = Buf.getBufferEnd();
308 
309   SourceLocation StartLoc = R.getSourceMgr().getLocForStartOfFile(FID);
310   SourceLocation EndLoc = StartLoc.getLocWithOffset(FileEnd-FileStart);
311 
312   std::string s;
313   llvm::raw_string_ostream os(s);
314   os << "<!doctype html>\n" // Use HTML 5 doctype
315         "<html>\n<head>\n";
316 
317   if (!title.empty())
318     os << "<title>" << html::EscapeText(title) << "</title>\n";
319 
320   os << R"<<<(
321 <style type="text/css">
322 body { color:#000000; background-color:#ffffff }
323 body { font-family:Helvetica, sans-serif; font-size:10pt }
324 h1 { font-size:14pt }
325 .FileName { margin-top: 5px; margin-bottom: 5px; display: inline; }
326 .FileNav { margin-left: 5px; margin-right: 5px; display: inline; }
327 .FileNav a { text-decoration:none; font-size: larger; }
328 .divider { margin-top: 30px; margin-bottom: 30px; height: 15px; }
329 .divider { background-color: gray; }
330 .code { border-collapse:collapse; width:100%; }
331 .code { font-family: "Monospace", monospace; font-size:10pt }
332 .code { line-height: 1.2em }
333 .comment { color: green; font-style: oblique }
334 .keyword { color: blue }
335 .string_literal { color: red }
336 .directive { color: darkmagenta }
337 
338 /* Macros and variables could have pop-up notes hidden by default.
339   - Macro pop-up:    expansion of the macro
340   - Variable pop-up: value (table) of the variable */
341 .macro_popup, .variable_popup { display: none; }
342 
343 /* Pop-up appears on mouse-hover event. */
344 .macro:hover .macro_popup, .variable:hover .variable_popup {
345   display: block;
346   padding: 2px;
347   -webkit-border-radius:5px;
348   -webkit-box-shadow:1px 1px 7px #000;
349   border-radius:5px;
350   box-shadow:1px 1px 7px #000;
351   position: absolute;
352   top: -1em;
353   left:10em;
354   z-index: 1
355 }
356 
357 .macro_popup {
358   border: 2px solid red;
359   background-color:#FFF0F0;
360   font-weight: normal;
361 }
362 
363 .variable_popup {
364   border: 2px solid blue;
365   background-color:#F0F0FF;
366   font-weight: bold;
367   font-family: Helvetica, sans-serif;
368   font-size: 9pt;
369 }
370 
371 /* Pop-up notes needs a relative position as a base where they pops up. */
372 .macro, .variable {
373   background-color: PaleGoldenRod;
374   position: relative;
375 }
376 .macro { color: DarkMagenta; }
377 
378 #tooltiphint {
379   position: fixed;
380   width: 50em;
381   margin-left: -25em;
382   left: 50%;
383   padding: 10px;
384   border: 1px solid #b0b0b0;
385   border-radius: 2px;
386   box-shadow: 1px 1px 7px black;
387   background-color: #c0c0c0;
388   z-index: 2;
389 }
390 
391 .num { width:2.5em; padding-right:2ex; background-color:#eeeeee }
392 .num { text-align:right; font-size:8pt }
393 .num { color:#444444 }
394 .line { padding-left: 1ex; border-left: 3px solid #ccc }
395 .line { white-space: pre }
396 .msg { -webkit-box-shadow:1px 1px 7px #000 }
397 .msg { box-shadow:1px 1px 7px #000 }
398 .msg { -webkit-border-radius:5px }
399 .msg { border-radius:5px }
400 .msg { font-family:Helvetica, sans-serif; font-size:8pt }
401 .msg { float:left }
402 .msg { position:relative }
403 .msg { padding:0.25em 1ex 0.25em 1ex }
404 .msg { margin-top:10px; margin-bottom:10px }
405 .msg { font-weight:bold }
406 .msg { max-width:60em; word-wrap: break-word; white-space: pre-wrap }
407 .msgT { padding:0x; spacing:0x }
408 .msgEvent { background-color:#fff8b4; color:#000000 }
409 .msgControl { background-color:#bbbbbb; color:#000000 }
410 .msgNote { background-color:#ddeeff; color:#000000 }
411 .mrange { background-color:#dfddf3 }
412 .mrange { border-bottom:1px solid #6F9DBE }
413 .PathIndex { font-weight: bold; padding:0px 5px; margin-right:5px; }
414 .PathIndex { -webkit-border-radius:8px }
415 .PathIndex { border-radius:8px }
416 .PathIndexEvent { background-color:#bfba87 }
417 .PathIndexControl { background-color:#8c8c8c }
418 .PathIndexPopUp { background-color: #879abc; }
419 .PathNav a { text-decoration:none; font-size: larger }
420 .CodeInsertionHint { font-weight: bold; background-color: #10dd10 }
421 .CodeRemovalHint { background-color:#de1010 }
422 .CodeRemovalHint { border-bottom:1px solid #6F9DBE }
423 .msg.selected{ background-color:orange !important; }
424 
425 table.simpletable {
426   padding: 5px;
427   font-size:12pt;
428   margin:20px;
429   border-collapse: collapse; border-spacing: 0px;
430 }
431 td.rowname {
432   text-align: right;
433   vertical-align: top;
434   font-weight: bold;
435   color:#444444;
436   padding-right:2ex;
437 }
438 
439 /* Hidden text. */
440 input.spoilerhider + label {
441   cursor: pointer;
442   text-decoration: underline;
443   display: block;
444 }
445 input.spoilerhider {
446  display: none;
447 }
448 input.spoilerhider ~ .spoiler {
449   overflow: hidden;
450   margin: 10px auto 0;
451   height: 0;
452   opacity: 0;
453 }
454 input.spoilerhider:checked + label + .spoiler{
455   height: auto;
456   opacity: 1;
457 }
458 </style>
459 </head>
460 <body>)<<<";
461 
462   // Generate header
463   R.InsertTextBefore(StartLoc, os.str());
464   // Generate footer
465 
466   R.InsertTextAfter(EndLoc, "</body></html>\n");
467 }
468 
469 /// SyntaxHighlight - Relex the specified FileID and annotate the HTML with
470 /// information about keywords, macro expansions etc.  This uses the macro
471 /// table state from the end of the file, so it won't be perfectly perfect,
472 /// but it will be reasonably close.
SyntaxHighlightImpl(Rewriter & R,FileID FID,const Preprocessor & PP,llvm::function_ref<void (RewriteBuffer &,unsigned,unsigned,const char *,const char *,const char *)> HighlightRangeCallback)473 static void SyntaxHighlightImpl(
474     Rewriter &R, FileID FID, const Preprocessor &PP,
475     llvm::function_ref<void(RewriteBuffer &, unsigned, unsigned, const char *,
476                             const char *, const char *)>
477         HighlightRangeCallback) {
478 
479   RewriteBuffer &RB = R.getEditBuffer(FID);
480   const SourceManager &SM = PP.getSourceManager();
481   llvm::MemoryBufferRef FromFile = SM.getBufferOrFake(FID);
482   const char *BufferStart = FromFile.getBuffer().data();
483 
484   Lexer L(FID, FromFile, SM, PP.getLangOpts());
485 
486   // Inform the preprocessor that we want to retain comments as tokens, so we
487   // can highlight them.
488   L.SetCommentRetentionState(true);
489 
490   // Lex all the tokens in raw mode, to avoid entering #includes or expanding
491   // macros.
492   Token Tok;
493   L.LexFromRawLexer(Tok);
494 
495   while (Tok.isNot(tok::eof)) {
496     // Since we are lexing unexpanded tokens, all tokens are from the main
497     // FileID.
498     unsigned TokOffs = SM.getFileOffset(Tok.getLocation());
499     unsigned TokLen = Tok.getLength();
500     switch (Tok.getKind()) {
501     default: break;
502     case tok::identifier:
503       llvm_unreachable("tok::identifier in raw lexing mode!");
504     case tok::raw_identifier: {
505       // Fill in Result.IdentifierInfo and update the token kind,
506       // looking up the identifier in the identifier table.
507       PP.LookUpIdentifierInfo(Tok);
508 
509       // If this is a pp-identifier, for a keyword, highlight it as such.
510       if (Tok.isNot(tok::identifier))
511         HighlightRangeCallback(RB, TokOffs, TokOffs + TokLen, BufferStart,
512                                "<span class='keyword'>", "</span>");
513       break;
514     }
515     case tok::comment:
516       HighlightRangeCallback(RB, TokOffs, TokOffs + TokLen, BufferStart,
517                              "<span class='comment'>", "</span>");
518       break;
519     case tok::utf8_string_literal:
520       // Chop off the u part of u8 prefix
521       ++TokOffs;
522       --TokLen;
523       // FALL THROUGH to chop the 8
524       [[fallthrough]];
525     case tok::wide_string_literal:
526     case tok::utf16_string_literal:
527     case tok::utf32_string_literal:
528       // Chop off the L, u, U or 8 prefix
529       ++TokOffs;
530       --TokLen;
531       [[fallthrough]];
532     case tok::string_literal:
533       // FIXME: Exclude the optional ud-suffix from the highlighted range.
534       HighlightRangeCallback(RB, TokOffs, TokOffs + TokLen, BufferStart,
535                              "<span class='string_literal'>", "</span>");
536       break;
537     case tok::hash: {
538       // If this is a preprocessor directive, all tokens to end of line are too.
539       if (!Tok.isAtStartOfLine())
540         break;
541 
542       // Eat all of the tokens until we get to the next one at the start of
543       // line.
544       unsigned TokEnd = TokOffs+TokLen;
545       L.LexFromRawLexer(Tok);
546       while (!Tok.isAtStartOfLine() && Tok.isNot(tok::eof)) {
547         TokEnd = SM.getFileOffset(Tok.getLocation())+Tok.getLength();
548         L.LexFromRawLexer(Tok);
549       }
550 
551       // Find end of line.  This is a hack.
552       HighlightRangeCallback(RB, TokOffs, TokEnd, BufferStart,
553                              "<span class='directive'>", "</span>");
554 
555       // Don't skip the next token.
556       continue;
557     }
558     }
559 
560     L.LexFromRawLexer(Tok);
561   }
562 }
SyntaxHighlight(Rewriter & R,FileID FID,const Preprocessor & PP,RelexRewriteCacheRef Cache)563 void html::SyntaxHighlight(Rewriter &R, FileID FID, const Preprocessor &PP,
564                            RelexRewriteCacheRef Cache) {
565   RewriteBuffer &RB = R.getEditBuffer(FID);
566   const SourceManager &SM = PP.getSourceManager();
567   llvm::MemoryBufferRef FromFile = SM.getBufferOrFake(FID);
568   const char *BufferStart = FromFile.getBuffer().data();
569 
570   if (Cache) {
571     auto CacheIt = Cache->SyntaxHighlights.find(FID);
572     if (CacheIt != Cache->SyntaxHighlights.end()) {
573       for (const RelexRewriteCache::RawHighlight &H : CacheIt->second) {
574         HighlightRange(RB, H.B, H.E, BufferStart, H.StartTag.data(),
575                        H.EndTag.data());
576       }
577       return;
578     }
579   }
580 
581   // "Every time you would call HighlightRange, cache the inputs as well."
582   auto HighlightRangeCallback = [&](RewriteBuffer &RB, unsigned B, unsigned E,
583                                     const char *BufferStart,
584                                     const char *StartTag, const char *EndTag) {
585     HighlightRange(RB, B, E, BufferStart, StartTag, EndTag);
586 
587     if (Cache)
588       Cache->SyntaxHighlights[FID].push_back({B, E, StartTag, EndTag});
589   };
590 
591   SyntaxHighlightImpl(R, FID, PP, HighlightRangeCallback);
592 }
593 
HighlightMacrosImpl(Rewriter & R,FileID FID,const Preprocessor & PP,llvm::function_ref<void (Rewriter &,SourceLocation,SourceLocation,const char *,const char *,bool)> HighlightRangeCallback)594 static void HighlightMacrosImpl(
595     Rewriter &R, FileID FID, const Preprocessor &PP,
596     llvm::function_ref<void(Rewriter &, SourceLocation, SourceLocation,
597                             const char *, const char *, bool)>
598         HighlightRangeCallback) {
599 
600   // Re-lex the raw token stream into a token buffer.
601   const SourceManager &SM = PP.getSourceManager();
602   std::vector<Token> TokenStream;
603 
604   llvm::MemoryBufferRef FromFile = SM.getBufferOrFake(FID);
605   Lexer L(FID, FromFile, SM, PP.getLangOpts());
606 
607   // Lex all the tokens in raw mode, to avoid entering #includes or expanding
608   // macros.
609   while (true) {
610     Token Tok;
611     L.LexFromRawLexer(Tok);
612 
613     // If this is a # at the start of a line, discard it from the token stream.
614     // We don't want the re-preprocess step to see #defines, #includes or other
615     // preprocessor directives.
616     if (Tok.is(tok::hash) && Tok.isAtStartOfLine())
617       continue;
618 
619     // If this is a ## token, change its kind to unknown so that repreprocessing
620     // it will not produce an error.
621     if (Tok.is(tok::hashhash))
622       Tok.setKind(tok::unknown);
623 
624     // If this raw token is an identifier, the raw lexer won't have looked up
625     // the corresponding identifier info for it.  Do this now so that it will be
626     // macro expanded when we re-preprocess it.
627     if (Tok.is(tok::raw_identifier))
628       PP.LookUpIdentifierInfo(Tok);
629 
630     TokenStream.push_back(Tok);
631 
632     if (Tok.is(tok::eof)) break;
633   }
634 
635   // Temporarily change the diagnostics object so that we ignore any generated
636   // diagnostics from this pass.
637   DiagnosticsEngine TmpDiags(PP.getDiagnostics().getDiagnosticIDs(),
638                              &PP.getDiagnostics().getDiagnosticOptions(),
639                       new IgnoringDiagConsumer);
640 
641   // FIXME: This is a huge hack; we reuse the input preprocessor because we want
642   // its state, but we aren't actually changing it (we hope). This should really
643   // construct a copy of the preprocessor.
644   Preprocessor &TmpPP = const_cast<Preprocessor&>(PP);
645   DiagnosticsEngine *OldDiags = &TmpPP.getDiagnostics();
646   TmpPP.setDiagnostics(TmpDiags);
647 
648   // Inform the preprocessor that we don't want comments.
649   TmpPP.SetCommentRetentionState(false, false);
650 
651   // We don't want pragmas either. Although we filtered out #pragma, removing
652   // _Pragma and __pragma is much harder.
653   bool PragmasPreviouslyEnabled = TmpPP.getPragmasEnabled();
654   TmpPP.setPragmasEnabled(false);
655 
656   // Enter the tokens we just lexed.  This will cause them to be macro expanded
657   // but won't enter sub-files (because we removed #'s).
658   TmpPP.EnterTokenStream(TokenStream, false, /*IsReinject=*/false);
659 
660   TokenConcatenation ConcatInfo(TmpPP);
661 
662   // Lex all the tokens.
663   Token Tok;
664   TmpPP.Lex(Tok);
665   while (Tok.isNot(tok::eof)) {
666     // Ignore non-macro tokens.
667     if (!Tok.getLocation().isMacroID()) {
668       TmpPP.Lex(Tok);
669       continue;
670     }
671 
672     // Okay, we have the first token of a macro expansion: highlight the
673     // expansion by inserting a start tag before the macro expansion and
674     // end tag after it.
675     CharSourceRange LLoc = SM.getExpansionRange(Tok.getLocation());
676 
677     // Ignore tokens whose instantiation location was not the main file.
678     if (SM.getFileID(LLoc.getBegin()) != FID) {
679       TmpPP.Lex(Tok);
680       continue;
681     }
682 
683     assert(SM.getFileID(LLoc.getEnd()) == FID &&
684            "Start and end of expansion must be in the same ultimate file!");
685 
686     std::string Expansion = EscapeText(TmpPP.getSpelling(Tok));
687     unsigned LineLen = Expansion.size();
688 
689     Token PrevPrevTok;
690     Token PrevTok = Tok;
691     // Okay, eat this token, getting the next one.
692     TmpPP.Lex(Tok);
693 
694     // Skip all the rest of the tokens that are part of this macro
695     // instantiation.  It would be really nice to pop up a window with all the
696     // spelling of the tokens or something.
697     while (!Tok.is(tok::eof) &&
698            SM.getExpansionLoc(Tok.getLocation()) == LLoc.getBegin()) {
699       // Insert a newline if the macro expansion is getting large.
700       if (LineLen > 60) {
701         Expansion += "<br>";
702         LineLen = 0;
703       }
704 
705       LineLen -= Expansion.size();
706 
707       // If the tokens were already space separated, or if they must be to avoid
708       // them being implicitly pasted, add a space between them.
709       if (Tok.hasLeadingSpace() ||
710           ConcatInfo.AvoidConcat(PrevPrevTok, PrevTok, Tok))
711         Expansion += ' ';
712 
713       // Escape any special characters in the token text.
714       Expansion += EscapeText(TmpPP.getSpelling(Tok));
715       LineLen += Expansion.size();
716 
717       PrevPrevTok = PrevTok;
718       PrevTok = Tok;
719       TmpPP.Lex(Tok);
720     }
721 
722     // Insert the 'macro_popup' as the end tag, so that multi-line macros all
723     // get highlighted.
724     Expansion = "<span class='macro_popup'>" + Expansion + "</span></span>";
725 
726     HighlightRangeCallback(R, LLoc.getBegin(), LLoc.getEnd(),
727                            "<span class='macro'>", Expansion.c_str(),
728                            LLoc.isTokenRange());
729   }
730 
731   // Restore the preprocessor's old state.
732   TmpPP.setDiagnostics(*OldDiags);
733   TmpPP.setPragmasEnabled(PragmasPreviouslyEnabled);
734 }
735 
736 /// HighlightMacros - This uses the macro table state from the end of the
737 /// file, to re-expand macros and insert (into the HTML) information about the
738 /// macro expansions.  This won't be perfectly perfect, but it will be
739 /// reasonably close.
HighlightMacros(Rewriter & R,FileID FID,const Preprocessor & PP,RelexRewriteCacheRef Cache)740 void html::HighlightMacros(Rewriter &R, FileID FID, const Preprocessor &PP,
741                            RelexRewriteCacheRef Cache) {
742   if (Cache) {
743     auto CacheIt = Cache->MacroHighlights.find(FID);
744     if (CacheIt != Cache->MacroHighlights.end()) {
745       for (const RelexRewriteCache::Highlight &H : CacheIt->second) {
746         HighlightRange(R, H.B, H.E, H.StartTag.data(), H.EndTag.data(),
747                        H.IsTokenRange);
748       }
749       return;
750     }
751   }
752 
753   // "Every time you would call HighlightRange, cache the inputs as well."
754   auto HighlightRangeCallback = [&](Rewriter &R, SourceLocation B,
755                                     SourceLocation E, const char *StartTag,
756                                     const char *EndTag, bool isTokenRange) {
757     HighlightRange(R, B, E, StartTag, EndTag, isTokenRange);
758 
759     if (Cache) {
760       Cache->MacroHighlights[FID].push_back(
761           {B, E, StartTag, EndTag, isTokenRange});
762     }
763   };
764 
765   HighlightMacrosImpl(R, FID, PP, HighlightRangeCallback);
766 }
767