xref: /freebsd/contrib/llvm-project/clang/lib/Rewrite/HTMLRewrite.cpp (revision 770cf0a5f02dc8983a89c6568d741fbc25baa999)
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/RewriteBuffer.h"
20 #include "llvm/Support/ErrorHandling.h"
21 #include "llvm/Support/raw_ostream.h"
22 #include <memory>
23 
24 using namespace clang;
25 using namespace llvm;
26 using namespace html;
27 
28 /// HighlightRange - Highlight a range in the source code with the specified
29 /// start/end tags.  B/E must be in the same file.  This ensures that
30 /// start/end tags are placed at the start/end of each line if the range is
31 /// multiline.
32 void html::HighlightRange(Rewriter &R, SourceLocation B, SourceLocation E,
33                           const char *StartTag, const char *EndTag,
34                           bool IsTokenRange) {
35   SourceManager &SM = R.getSourceMgr();
36   B = SM.getExpansionLoc(B);
37   E = SM.getExpansionLoc(E);
38   FileID FID = SM.getFileID(B);
39   assert(SM.getFileID(E) == FID && "B/E not in the same file!");
40 
41   unsigned BOffset = SM.getFileOffset(B);
42   unsigned EOffset = SM.getFileOffset(E);
43 
44   // Include the whole end token in the range.
45   if (IsTokenRange)
46     EOffset += Lexer::MeasureTokenLength(E, R.getSourceMgr(), R.getLangOpts());
47 
48   bool Invalid = false;
49   const char *BufferStart = SM.getBufferData(FID, &Invalid).data();
50   if (Invalid)
51     return;
52 
53   HighlightRange(R.getEditBuffer(FID), BOffset, EOffset,
54                  BufferStart, StartTag, EndTag);
55 }
56 
57 /// HighlightRange - This is the same as the above method, but takes
58 /// decomposed file locations.
59 void html::HighlightRange(RewriteBuffer &RB, unsigned B, unsigned E,
60                           const char *BufferStart,
61                           const char *StartTag, const char *EndTag) {
62   // Insert the tag at the absolute start/end of the range.
63   RB.InsertTextAfter(B, StartTag);
64   RB.InsertTextBefore(E, EndTag);
65 
66   // Scan the range to see if there is a \r or \n.  If so, and if the line is
67   // not blank, insert tags on that line as well.
68   bool HadOpenTag = true;
69 
70   unsigned LastNonWhiteSpace = B;
71   for (unsigned i = B; i != E; ++i) {
72     switch (BufferStart[i]) {
73     case '\r':
74     case '\n':
75       // Okay, we found a newline in the range.  If we have an open tag, we need
76       // to insert a close tag at the first non-whitespace before the newline.
77       if (HadOpenTag)
78         RB.InsertTextBefore(LastNonWhiteSpace+1, EndTag);
79 
80       // Instead of inserting an open tag immediately after the newline, we
81       // wait until we see a non-whitespace character.  This prevents us from
82       // inserting tags around blank lines, and also allows the open tag to
83       // be put *after* whitespace on a non-blank line.
84       HadOpenTag = false;
85       break;
86     case '\0':
87     case ' ':
88     case '\t':
89     case '\f':
90     case '\v':
91       // Ignore whitespace.
92       break;
93 
94     default:
95       // If there is no tag open, do it now.
96       if (!HadOpenTag) {
97         RB.InsertTextAfter(i, StartTag);
98         HadOpenTag = true;
99       }
100 
101       // Remember this character.
102       LastNonWhiteSpace = i;
103       break;
104     }
105   }
106 }
107 
108 namespace clang::html {
109 struct RelexRewriteCache {
110   // These structs mimic input arguments of HighlightRange().
111   struct Highlight {
112     SourceLocation B, E;
113     std::string StartTag, EndTag;
114     bool IsTokenRange;
115   };
116   struct RawHighlight {
117     unsigned B, E;
118     std::string StartTag, EndTag;
119   };
120 
121   // SmallVector isn't appropriate because these vectors are almost never small.
122   using HighlightList = std::vector<Highlight>;
123   using RawHighlightList = std::vector<RawHighlight>;
124 
125   DenseMap<FileID, RawHighlightList> SyntaxHighlights;
126   DenseMap<FileID, HighlightList> MacroHighlights;
127 };
128 } // namespace clang::html
129 
130 html::RelexRewriteCacheRef html::instantiateRelexRewriteCache() {
131   return std::make_shared<RelexRewriteCache>();
132 }
133 
134 void html::EscapeText(Rewriter &R, FileID FID,
135                       bool EscapeSpaces, bool ReplaceTabs) {
136 
137   llvm::MemoryBufferRef Buf = R.getSourceMgr().getBufferOrFake(FID);
138   const char* C = Buf.getBufferStart();
139   const char* FileEnd = Buf.getBufferEnd();
140 
141   assert (C <= FileEnd);
142 
143   RewriteBuffer &RB = R.getEditBuffer(FID);
144 
145   unsigned ColNo = 0;
146   for (unsigned FilePos = 0; C != FileEnd ; ++C, ++FilePos) {
147     switch (*C) {
148     default: ++ColNo; break;
149     case '\n':
150     case '\r':
151       ColNo = 0;
152       break;
153 
154     case ' ':
155       if (EscapeSpaces)
156         RB.ReplaceText(FilePos, 1, "&nbsp;");
157       ++ColNo;
158       break;
159     case '\f':
160       RB.ReplaceText(FilePos, 1, "<hr>");
161       ColNo = 0;
162       break;
163 
164     case '\t': {
165       if (!ReplaceTabs)
166         break;
167       unsigned NumSpaces = 8-(ColNo&7);
168       if (EscapeSpaces)
169         RB.ReplaceText(FilePos, 1,
170                        StringRef("&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;"
171                                        "&nbsp;&nbsp;&nbsp;", 6*NumSpaces));
172       else
173         RB.ReplaceText(FilePos, 1, StringRef("        ", NumSpaces));
174       ColNo += NumSpaces;
175       break;
176     }
177     case '<':
178       RB.ReplaceText(FilePos, 1, "&lt;");
179       ++ColNo;
180       break;
181 
182     case '>':
183       RB.ReplaceText(FilePos, 1, "&gt;");
184       ++ColNo;
185       break;
186 
187     case '&':
188       RB.ReplaceText(FilePos, 1, "&amp;");
189       ++ColNo;
190       break;
191     }
192   }
193 }
194 
195 std::string html::EscapeText(StringRef s, bool EscapeSpaces, bool ReplaceTabs) {
196 
197   unsigned len = s.size();
198   std::string Str;
199   llvm::raw_string_ostream os(Str);
200 
201   for (unsigned i = 0 ; i < len; ++i) {
202 
203     char c = s[i];
204     switch (c) {
205     default:
206       os << c; break;
207 
208     case ' ':
209       if (EscapeSpaces) os << "&nbsp;";
210       else os << ' ';
211       break;
212 
213     case '\t':
214       if (ReplaceTabs) {
215         if (EscapeSpaces)
216           for (unsigned i = 0; i < 4; ++i)
217             os << "&nbsp;";
218         else
219           for (unsigned i = 0; i < 4; ++i)
220             os << " ";
221       }
222       else
223         os << c;
224 
225       break;
226 
227     case '<': os << "&lt;"; break;
228     case '>': os << "&gt;"; break;
229     case '&': os << "&amp;"; break;
230     }
231   }
232 
233   return Str;
234 }
235 
236 static void AddLineNumber(RewriteBuffer &RB, unsigned LineNo,
237                           unsigned B, unsigned E) {
238   SmallString<256> Str;
239   llvm::raw_svector_ostream OS(Str);
240 
241   OS << "<tr class=\"codeline\" data-linenumber=\"" << LineNo << "\">"
242      << "<td class=\"num\" id=\"LN" << LineNo << "\">" << LineNo
243      << "</td><td class=\"line\">";
244 
245   if (B == E) { // Handle empty lines.
246     OS << " </td></tr>";
247     RB.InsertTextBefore(B, OS.str());
248   } else {
249     RB.InsertTextBefore(B, OS.str());
250     RB.InsertTextBefore(E, "</td></tr>");
251   }
252 }
253 
254 void html::AddLineNumbers(Rewriter& R, FileID FID) {
255 
256   llvm::MemoryBufferRef Buf = R.getSourceMgr().getBufferOrFake(FID);
257   const char* FileBeg = Buf.getBufferStart();
258   const char* FileEnd = Buf.getBufferEnd();
259   const char* C = FileBeg;
260   RewriteBuffer &RB = R.getEditBuffer(FID);
261 
262   assert (C <= FileEnd);
263 
264   unsigned LineNo = 0;
265   unsigned FilePos = 0;
266 
267   while (C != FileEnd) {
268 
269     ++LineNo;
270     unsigned LineStartPos = FilePos;
271     unsigned LineEndPos = FileEnd - FileBeg;
272 
273     assert (FilePos <= LineEndPos);
274     assert (C < FileEnd);
275 
276     // Scan until the newline (or end-of-file).
277 
278     while (C != FileEnd) {
279       char c = *C;
280       ++C;
281 
282       if (c == '\n') {
283         LineEndPos = FilePos++;
284         break;
285       }
286 
287       ++FilePos;
288     }
289 
290     AddLineNumber(RB, LineNo, LineStartPos, LineEndPos);
291   }
292 
293   // Add one big table tag that surrounds all of the code.
294   std::string s;
295   llvm::raw_string_ostream os(s);
296   os << "<table class=\"code\" data-fileid=\"" << FID.getHashValue() << "\">\n";
297   RB.InsertTextBefore(0, os.str());
298   RB.InsertTextAfter(FileEnd - FileBeg, "</table>");
299 }
300 
301 void html::AddHeaderFooterInternalBuiltinCSS(Rewriter &R, FileID FID,
302                                              StringRef title) {
303 
304   llvm::MemoryBufferRef Buf = R.getSourceMgr().getBufferOrFake(FID);
305   const char* FileStart = Buf.getBufferStart();
306   const char* FileEnd = Buf.getBufferEnd();
307 
308   SourceLocation StartLoc = R.getSourceMgr().getLocForStartOfFile(FID);
309   SourceLocation EndLoc = StartLoc.getLocWithOffset(FileEnd-FileStart);
310 
311   std::string s;
312   llvm::raw_string_ostream os(s);
313   os << "<!doctype html>\n" // Use HTML 5 doctype
314         "<html>\n<head>\n";
315 
316   if (!title.empty())
317     os << "<title>" << html::EscapeText(title) << "</title>\n";
318 
319   os << R"<<<(
320 <style type="text/css">
321 body { color:#000000; background-color:#ffffff }
322 body { font-family:Helvetica, sans-serif; font-size:10pt }
323 h1 { font-size:14pt }
324 .FileName { margin-top: 5px; margin-bottom: 5px; display: inline; }
325 .FileNav { margin-left: 5px; margin-right: 5px; display: inline; }
326 .FileNav a { text-decoration:none; font-size: larger; }
327 .divider { margin-top: 30px; margin-bottom: 30px; height: 15px; }
328 .divider { background-color: gray; }
329 .code { border-collapse:collapse; width:100%; }
330 .code { font-family: "Monospace", monospace; font-size:10pt }
331 .code { line-height: 1.2em }
332 .comment { color: green; font-style: oblique }
333 .keyword { color: blue }
334 .string_literal { color: red }
335 .directive { color: darkmagenta }
336 
337 /* Macros and variables could have pop-up notes hidden by default.
338   - Macro pop-up:    expansion of the macro
339   - Variable pop-up: value (table) of the variable */
340 .macro_popup, .variable_popup { display: none; }
341 
342 /* Pop-up appears on mouse-hover event. */
343 .macro:hover .macro_popup, .variable:hover .variable_popup {
344   display: block;
345   padding: 2px;
346   -webkit-border-radius:5px;
347   -webkit-box-shadow:1px 1px 7px #000;
348   border-radius:5px;
349   box-shadow:1px 1px 7px #000;
350   position: absolute;
351   top: -1em;
352   left:10em;
353   z-index: 1
354 }
355 
356 .macro_popup {
357   border: 2px solid red;
358   background-color:#FFF0F0;
359   font-weight: normal;
360 }
361 
362 .variable_popup {
363   border: 2px solid blue;
364   background-color:#F0F0FF;
365   font-weight: bold;
366   font-family: Helvetica, sans-serif;
367   font-size: 9pt;
368 }
369 
370 /* Pop-up notes needs a relative position as a base where they pops up. */
371 .macro, .variable {
372   background-color: PaleGoldenRod;
373   position: relative;
374 }
375 .macro { color: DarkMagenta; }
376 
377 #tooltiphint {
378   position: fixed;
379   width: 50em;
380   margin-left: -25em;
381   left: 50%;
382   padding: 10px;
383   border: 1px solid #b0b0b0;
384   border-radius: 2px;
385   box-shadow: 1px 1px 7px black;
386   background-color: #c0c0c0;
387   z-index: 2;
388 }
389 
390 .num { width:2.5em; padding-right:2ex; background-color:#eeeeee }
391 .num { text-align:right; font-size:8pt }
392 .num { color:#444444 }
393 .line { padding-left: 1ex; border-left: 3px solid #ccc }
394 .line { white-space: pre }
395 .msg { -webkit-box-shadow:1px 1px 7px #000 }
396 .msg { box-shadow:1px 1px 7px #000 }
397 .msg { -webkit-border-radius:5px }
398 .msg { border-radius:5px }
399 .msg { font-family:Helvetica, sans-serif; font-size:8pt }
400 .msg { float:left }
401 .msg { position:relative }
402 .msg { padding:0.25em 1ex 0.25em 1ex }
403 .msg { margin-top:10px; margin-bottom:10px }
404 .msg { font-weight:bold }
405 .msg { max-width:60em; word-wrap: break-word; white-space: pre-wrap }
406 .msgT { padding:0x; spacing:0x }
407 .msgEvent { background-color:#fff8b4; color:#000000 }
408 .msgControl { background-color:#bbbbbb; color:#000000 }
409 .msgNote { background-color:#ddeeff; color:#000000 }
410 .mrange { background-color:#dfddf3 }
411 .mrange { border-bottom:1px solid #6F9DBE }
412 .PathIndex { font-weight: bold; padding:0px 5px; margin-right:5px; }
413 .PathIndex { -webkit-border-radius:8px }
414 .PathIndex { border-radius:8px }
415 .PathIndexEvent { background-color:#bfba87 }
416 .PathIndexControl { background-color:#8c8c8c }
417 .PathIndexPopUp { background-color: #879abc; }
418 .PathNav a { text-decoration:none; font-size: larger }
419 .CodeInsertionHint { font-weight: bold; background-color: #10dd10 }
420 .CodeRemovalHint { background-color:#de1010 }
421 .CodeRemovalHint { border-bottom:1px solid #6F9DBE }
422 .msg.selected{ background-color:orange !important; }
423 
424 table.simpletable {
425   padding: 5px;
426   font-size:12pt;
427   margin:20px;
428   border-collapse: collapse; border-spacing: 0px;
429 }
430 td.rowname {
431   text-align: right;
432   vertical-align: top;
433   font-weight: bold;
434   color:#444444;
435   padding-right:2ex;
436 }
437 
438 /* Hidden text. */
439 input.spoilerhider + label {
440   cursor: pointer;
441   text-decoration: underline;
442   display: block;
443 }
444 input.spoilerhider {
445  display: none;
446 }
447 input.spoilerhider ~ .spoiler {
448   overflow: hidden;
449   margin: 10px auto 0;
450   height: 0;
451   opacity: 0;
452 }
453 input.spoilerhider:checked + label + .spoiler{
454   height: auto;
455   opacity: 1;
456 }
457 </style>
458 </head>
459 <body>)<<<";
460 
461   // Generate header
462   R.InsertTextBefore(StartLoc, os.str());
463   // Generate footer
464 
465   R.InsertTextAfter(EndLoc, "</body></html>\n");
466 }
467 
468 /// SyntaxHighlight - Relex the specified FileID and annotate the HTML with
469 /// information about keywords, macro expansions etc.  This uses the macro
470 /// table state from the end of the file, so it won't be perfectly perfect,
471 /// but it will be reasonably close.
472 static void SyntaxHighlightImpl(
473     Rewriter &R, FileID FID, const Preprocessor &PP,
474     llvm::function_ref<void(RewriteBuffer &, unsigned, unsigned, const char *,
475                             const char *, const char *)>
476         HighlightRangeCallback) {
477 
478   RewriteBuffer &RB = R.getEditBuffer(FID);
479   const SourceManager &SM = PP.getSourceManager();
480   llvm::MemoryBufferRef FromFile = SM.getBufferOrFake(FID);
481   const char *BufferStart = FromFile.getBuffer().data();
482 
483   Lexer L(FID, FromFile, SM, PP.getLangOpts());
484 
485   // Inform the preprocessor that we want to retain comments as tokens, so we
486   // can highlight them.
487   L.SetCommentRetentionState(true);
488 
489   // Lex all the tokens in raw mode, to avoid entering #includes or expanding
490   // macros.
491   Token Tok;
492   L.LexFromRawLexer(Tok);
493 
494   while (Tok.isNot(tok::eof)) {
495     // Since we are lexing unexpanded tokens, all tokens are from the main
496     // FileID.
497     unsigned TokOffs = SM.getFileOffset(Tok.getLocation());
498     unsigned TokLen = Tok.getLength();
499     switch (Tok.getKind()) {
500     default: break;
501     case tok::identifier:
502       llvm_unreachable("tok::identifier in raw lexing mode!");
503     case tok::raw_identifier: {
504       // Fill in Result.IdentifierInfo and update the token kind,
505       // looking up the identifier in the identifier table.
506       PP.LookUpIdentifierInfo(Tok);
507 
508       // If this is a pp-identifier, for a keyword, highlight it as such.
509       if (Tok.isNot(tok::identifier))
510         HighlightRangeCallback(RB, TokOffs, TokOffs + TokLen, BufferStart,
511                                "<span class='keyword'>", "</span>");
512       break;
513     }
514     case tok::comment:
515       HighlightRangeCallback(RB, TokOffs, TokOffs + TokLen, BufferStart,
516                              "<span class='comment'>", "</span>");
517       break;
518     case tok::utf8_string_literal:
519       // Chop off the u part of u8 prefix
520       ++TokOffs;
521       --TokLen;
522       // FALL THROUGH to chop the 8
523       [[fallthrough]];
524     case tok::wide_string_literal:
525     case tok::utf16_string_literal:
526     case tok::utf32_string_literal:
527       // Chop off the L, u, U or 8 prefix
528       ++TokOffs;
529       --TokLen;
530       [[fallthrough]];
531     case tok::string_literal:
532       // FIXME: Exclude the optional ud-suffix from the highlighted range.
533       HighlightRangeCallback(RB, TokOffs, TokOffs + TokLen, BufferStart,
534                              "<span class='string_literal'>", "</span>");
535       break;
536     case tok::hash: {
537       // If this is a preprocessor directive, all tokens to end of line are too.
538       if (!Tok.isAtStartOfLine())
539         break;
540 
541       // Eat all of the tokens until we get to the next one at the start of
542       // line.
543       unsigned TokEnd = TokOffs+TokLen;
544       L.LexFromRawLexer(Tok);
545       while (!Tok.isAtStartOfLine() && Tok.isNot(tok::eof)) {
546         TokEnd = SM.getFileOffset(Tok.getLocation())+Tok.getLength();
547         L.LexFromRawLexer(Tok);
548       }
549 
550       // Find end of line.  This is a hack.
551       HighlightRangeCallback(RB, TokOffs, TokEnd, BufferStart,
552                              "<span class='directive'>", "</span>");
553 
554       // Don't skip the next token.
555       continue;
556     }
557     }
558 
559     L.LexFromRawLexer(Tok);
560   }
561 }
562 void html::SyntaxHighlight(Rewriter &R, FileID FID, const Preprocessor &PP,
563                            RelexRewriteCacheRef Cache) {
564   RewriteBuffer &RB = R.getEditBuffer(FID);
565   const SourceManager &SM = PP.getSourceManager();
566   llvm::MemoryBufferRef FromFile = SM.getBufferOrFake(FID);
567   const char *BufferStart = FromFile.getBuffer().data();
568 
569   if (Cache) {
570     auto CacheIt = Cache->SyntaxHighlights.find(FID);
571     if (CacheIt != Cache->SyntaxHighlights.end()) {
572       for (const RelexRewriteCache::RawHighlight &H : CacheIt->second) {
573         HighlightRange(RB, H.B, H.E, BufferStart, H.StartTag.data(),
574                        H.EndTag.data());
575       }
576       return;
577     }
578   }
579 
580   // "Every time you would call HighlightRange, cache the inputs as well."
581   auto HighlightRangeCallback = [&](RewriteBuffer &RB, unsigned B, unsigned E,
582                                     const char *BufferStart,
583                                     const char *StartTag, const char *EndTag) {
584     HighlightRange(RB, B, E, BufferStart, StartTag, EndTag);
585 
586     if (Cache)
587       Cache->SyntaxHighlights[FID].push_back({B, E, StartTag, EndTag});
588   };
589 
590   SyntaxHighlightImpl(R, FID, PP, HighlightRangeCallback);
591 }
592 
593 static void HighlightMacrosImpl(
594     Rewriter &R, FileID FID, const Preprocessor &PP,
595     llvm::function_ref<void(Rewriter &, SourceLocation, SourceLocation,
596                             const char *, const char *, bool)>
597         HighlightRangeCallback) {
598 
599   // Re-lex the raw token stream into a token buffer.
600   const SourceManager &SM = PP.getSourceManager();
601   std::vector<Token> TokenStream;
602 
603   llvm::MemoryBufferRef FromFile = SM.getBufferOrFake(FID);
604   Lexer L(FID, FromFile, SM, PP.getLangOpts());
605 
606   // Lex all the tokens in raw mode, to avoid entering #includes or expanding
607   // macros.
608   while (true) {
609     Token Tok;
610     L.LexFromRawLexer(Tok);
611 
612     // If this is a # at the start of a line, discard it from the token stream.
613     // We don't want the re-preprocess step to see #defines, #includes or other
614     // preprocessor directives.
615     if (Tok.is(tok::hash) && Tok.isAtStartOfLine())
616       continue;
617 
618     // If this is a ## token, change its kind to unknown so that repreprocessing
619     // it will not produce an error.
620     if (Tok.is(tok::hashhash))
621       Tok.setKind(tok::unknown);
622 
623     // If this raw token is an identifier, the raw lexer won't have looked up
624     // the corresponding identifier info for it.  Do this now so that it will be
625     // macro expanded when we re-preprocess it.
626     if (Tok.is(tok::raw_identifier))
627       PP.LookUpIdentifierInfo(Tok);
628 
629     TokenStream.push_back(Tok);
630 
631     if (Tok.is(tok::eof)) break;
632   }
633 
634   // Temporarily change the diagnostics object so that we ignore any generated
635   // diagnostics from this pass.
636   DiagnosticsEngine TmpDiags(PP.getDiagnostics().getDiagnosticIDs(),
637                              PP.getDiagnostics().getDiagnosticOptions(),
638                              new IgnoringDiagConsumer);
639 
640   // FIXME: This is a huge hack; we reuse the input preprocessor because we want
641   // its state, but we aren't actually changing it (we hope). This should really
642   // construct a copy of the preprocessor.
643   Preprocessor &TmpPP = const_cast<Preprocessor&>(PP);
644   DiagnosticsEngine *OldDiags = &TmpPP.getDiagnostics();
645   TmpPP.setDiagnostics(TmpDiags);
646 
647   // Inform the preprocessor that we don't want comments.
648   TmpPP.SetCommentRetentionState(false, false);
649 
650   // We don't want pragmas either. Although we filtered out #pragma, removing
651   // _Pragma and __pragma is much harder.
652   bool PragmasPreviouslyEnabled = TmpPP.getPragmasEnabled();
653   TmpPP.setPragmasEnabled(false);
654 
655   // Enter the tokens we just lexed.  This will cause them to be macro expanded
656   // but won't enter sub-files (because we removed #'s).
657   TmpPP.EnterTokenStream(TokenStream, false, /*IsReinject=*/false);
658 
659   TokenConcatenation ConcatInfo(TmpPP);
660 
661   // Lex all the tokens.
662   Token Tok;
663   TmpPP.Lex(Tok);
664   while (Tok.isNot(tok::eof)) {
665     // Ignore non-macro tokens.
666     if (!Tok.getLocation().isMacroID()) {
667       TmpPP.Lex(Tok);
668       continue;
669     }
670 
671     // Okay, we have the first token of a macro expansion: highlight the
672     // expansion by inserting a start tag before the macro expansion and
673     // end tag after it.
674     CharSourceRange LLoc = SM.getExpansionRange(Tok.getLocation());
675 
676     // Ignore tokens whose instantiation location was not the main file.
677     if (SM.getFileID(LLoc.getBegin()) != FID) {
678       TmpPP.Lex(Tok);
679       continue;
680     }
681 
682     assert(SM.getFileID(LLoc.getEnd()) == FID &&
683            "Start and end of expansion must be in the same ultimate file!");
684 
685     std::string Expansion = EscapeText(TmpPP.getSpelling(Tok));
686     unsigned LineLen = Expansion.size();
687 
688     Token PrevPrevTok;
689     Token PrevTok = Tok;
690     // Okay, eat this token, getting the next one.
691     TmpPP.Lex(Tok);
692 
693     // Skip all the rest of the tokens that are part of this macro
694     // instantiation.  It would be really nice to pop up a window with all the
695     // spelling of the tokens or something.
696     while (!Tok.is(tok::eof) &&
697            SM.getExpansionLoc(Tok.getLocation()) == LLoc.getBegin()) {
698       // Insert a newline if the macro expansion is getting large.
699       if (LineLen > 60) {
700         Expansion += "<br>";
701         LineLen = 0;
702       }
703 
704       LineLen -= Expansion.size();
705 
706       // If the tokens were already space separated, or if they must be to avoid
707       // them being implicitly pasted, add a space between them.
708       if (Tok.hasLeadingSpace() ||
709           ConcatInfo.AvoidConcat(PrevPrevTok, PrevTok, Tok))
710         Expansion += ' ';
711 
712       // Escape any special characters in the token text.
713       Expansion += EscapeText(TmpPP.getSpelling(Tok));
714       LineLen += Expansion.size();
715 
716       PrevPrevTok = PrevTok;
717       PrevTok = Tok;
718       TmpPP.Lex(Tok);
719     }
720 
721     // Insert the 'macro_popup' as the end tag, so that multi-line macros all
722     // get highlighted.
723     Expansion = "<span class='macro_popup'>" + Expansion + "</span></span>";
724 
725     HighlightRangeCallback(R, LLoc.getBegin(), LLoc.getEnd(),
726                            "<span class='macro'>", Expansion.c_str(),
727                            LLoc.isTokenRange());
728   }
729 
730   // Restore the preprocessor's old state.
731   TmpPP.setDiagnostics(*OldDiags);
732   TmpPP.setPragmasEnabled(PragmasPreviouslyEnabled);
733 }
734 
735 /// HighlightMacros - This uses the macro table state from the end of the
736 /// file, to re-expand macros and insert (into the HTML) information about the
737 /// macro expansions.  This won't be perfectly perfect, but it will be
738 /// reasonably close.
739 void html::HighlightMacros(Rewriter &R, FileID FID, const Preprocessor &PP,
740                            RelexRewriteCacheRef Cache) {
741   if (Cache) {
742     auto CacheIt = Cache->MacroHighlights.find(FID);
743     if (CacheIt != Cache->MacroHighlights.end()) {
744       for (const RelexRewriteCache::Highlight &H : CacheIt->second) {
745         HighlightRange(R, H.B, H.E, H.StartTag.data(), H.EndTag.data(),
746                        H.IsTokenRange);
747       }
748       return;
749     }
750   }
751 
752   // "Every time you would call HighlightRange, cache the inputs as well."
753   auto HighlightRangeCallback = [&](Rewriter &R, SourceLocation B,
754                                     SourceLocation E, const char *StartTag,
755                                     const char *EndTag, bool isTokenRange) {
756     HighlightRange(R, B, E, StartTag, EndTag, isTokenRange);
757 
758     if (Cache) {
759       Cache->MacroHighlights[FID].push_back(
760           {B, E, StartTag, EndTag, isTokenRange});
761     }
762   };
763 
764   HighlightMacrosImpl(R, FID, PP, HighlightRangeCallback);
765 }
766