1 //== HTMLRewrite.cpp - Translate source code into prettified HTML --*- 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 //  This file defines the HTMLRewriter class, which is used to translate the
11 //  text of a source file into prettified HTML.
12 //
13 //===----------------------------------------------------------------------===//
14 
15 #include "clang/Rewrite/Core/HTMLRewrite.h"
16 #include "clang/Basic/SourceManager.h"
17 #include "clang/Lex/Preprocessor.h"
18 #include "clang/Lex/TokenConcatenation.h"
19 #include "clang/Rewrite/Core/Rewriter.h"
20 #include "llvm/ADT/SmallString.h"
21 #include "llvm/Support/ErrorHandling.h"
22 #include "llvm/Support/MemoryBuffer.h"
23 #include "llvm/Support/raw_ostream.h"
24 #include <memory>
25 using namespace clang;
26 
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.
HighlightRange(Rewriter & R,SourceLocation B,SourceLocation E,const char * StartTag,const char * EndTag)32 void html::HighlightRange(Rewriter &R, SourceLocation B, SourceLocation E,
33                           const char *StartTag, const char *EndTag) {
34   SourceManager &SM = R.getSourceMgr();
35   B = SM.getExpansionLoc(B);
36   E = SM.getExpansionLoc(E);
37   FileID FID = SM.getFileID(B);
38   assert(SM.getFileID(E) == FID && "B/E not in the same file!");
39 
40   unsigned BOffset = SM.getFileOffset(B);
41   unsigned EOffset = SM.getFileOffset(E);
42 
43   // Include the whole end token in the range.
44   EOffset += Lexer::MeasureTokenLength(E, R.getSourceMgr(), R.getLangOpts());
45 
46   bool Invalid = false;
47   const char *BufferStart = SM.getBufferData(FID, &Invalid).data();
48   if (Invalid)
49     return;
50 
51   HighlightRange(R.getEditBuffer(FID), BOffset, EOffset,
52                  BufferStart, StartTag, EndTag);
53 }
54 
55 /// HighlightRange - This is the same as the above method, but takes
56 /// decomposed file locations.
HighlightRange(RewriteBuffer & RB,unsigned B,unsigned E,const char * BufferStart,const char * StartTag,const char * EndTag)57 void html::HighlightRange(RewriteBuffer &RB, unsigned B, unsigned E,
58                           const char *BufferStart,
59                           const char *StartTag, const char *EndTag) {
60   // Insert the tag at the absolute start/end of the range.
61   RB.InsertTextAfter(B, StartTag);
62   RB.InsertTextBefore(E, EndTag);
63 
64   // Scan the range to see if there is a \r or \n.  If so, and if the line is
65   // not blank, insert tags on that line as well.
66   bool HadOpenTag = true;
67 
68   unsigned LastNonWhiteSpace = B;
69   for (unsigned i = B; i != E; ++i) {
70     switch (BufferStart[i]) {
71     case '\r':
72     case '\n':
73       // Okay, we found a newline in the range.  If we have an open tag, we need
74       // to insert a close tag at the first non-whitespace before the newline.
75       if (HadOpenTag)
76         RB.InsertTextBefore(LastNonWhiteSpace+1, EndTag);
77 
78       // Instead of inserting an open tag immediately after the newline, we
79       // wait until we see a non-whitespace character.  This prevents us from
80       // inserting tags around blank lines, and also allows the open tag to
81       // be put *after* whitespace on a non-blank line.
82       HadOpenTag = false;
83       break;
84     case '\0':
85     case ' ':
86     case '\t':
87     case '\f':
88     case '\v':
89       // Ignore whitespace.
90       break;
91 
92     default:
93       // If there is no tag open, do it now.
94       if (!HadOpenTag) {
95         RB.InsertTextAfter(i, StartTag);
96         HadOpenTag = true;
97       }
98 
99       // Remember this character.
100       LastNonWhiteSpace = i;
101       break;
102     }
103   }
104 }
105 
EscapeText(Rewriter & R,FileID FID,bool EscapeSpaces,bool ReplaceTabs)106 void html::EscapeText(Rewriter &R, FileID FID,
107                       bool EscapeSpaces, bool ReplaceTabs) {
108 
109   const llvm::MemoryBuffer *Buf = R.getSourceMgr().getBuffer(FID);
110   const char* C = Buf->getBufferStart();
111   const char* FileEnd = Buf->getBufferEnd();
112 
113   assert (C <= FileEnd);
114 
115   RewriteBuffer &RB = R.getEditBuffer(FID);
116 
117   unsigned ColNo = 0;
118   for (unsigned FilePos = 0; C != FileEnd ; ++C, ++FilePos) {
119     switch (*C) {
120     default: ++ColNo; break;
121     case '\n':
122     case '\r':
123       ColNo = 0;
124       break;
125 
126     case ' ':
127       if (EscapeSpaces)
128         RB.ReplaceText(FilePos, 1, "&nbsp;");
129       ++ColNo;
130       break;
131     case '\f':
132       RB.ReplaceText(FilePos, 1, "<hr>");
133       ColNo = 0;
134       break;
135 
136     case '\t': {
137       if (!ReplaceTabs)
138         break;
139       unsigned NumSpaces = 8-(ColNo&7);
140       if (EscapeSpaces)
141         RB.ReplaceText(FilePos, 1,
142                        StringRef("&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;"
143                                        "&nbsp;&nbsp;&nbsp;", 6*NumSpaces));
144       else
145         RB.ReplaceText(FilePos, 1, StringRef("        ", NumSpaces));
146       ColNo += NumSpaces;
147       break;
148     }
149     case '<':
150       RB.ReplaceText(FilePos, 1, "&lt;");
151       ++ColNo;
152       break;
153 
154     case '>':
155       RB.ReplaceText(FilePos, 1, "&gt;");
156       ++ColNo;
157       break;
158 
159     case '&':
160       RB.ReplaceText(FilePos, 1, "&amp;");
161       ++ColNo;
162       break;
163     }
164   }
165 }
166 
EscapeText(StringRef s,bool EscapeSpaces,bool ReplaceTabs)167 std::string html::EscapeText(StringRef s, bool EscapeSpaces, bool ReplaceTabs) {
168 
169   unsigned len = s.size();
170   std::string Str;
171   llvm::raw_string_ostream os(Str);
172 
173   for (unsigned i = 0 ; i < len; ++i) {
174 
175     char c = s[i];
176     switch (c) {
177     default:
178       os << c; break;
179 
180     case ' ':
181       if (EscapeSpaces) os << "&nbsp;";
182       else os << ' ';
183       break;
184 
185     case '\t':
186       if (ReplaceTabs) {
187         if (EscapeSpaces)
188           for (unsigned i = 0; i < 4; ++i)
189             os << "&nbsp;";
190         else
191           for (unsigned i = 0; i < 4; ++i)
192             os << " ";
193       }
194       else
195         os << c;
196 
197       break;
198 
199     case '<': os << "&lt;"; break;
200     case '>': os << "&gt;"; break;
201     case '&': os << "&amp;"; break;
202     }
203   }
204 
205   return os.str();
206 }
207 
AddLineNumber(RewriteBuffer & RB,unsigned LineNo,unsigned B,unsigned E)208 static void AddLineNumber(RewriteBuffer &RB, unsigned LineNo,
209                           unsigned B, unsigned E) {
210   SmallString<256> Str;
211   llvm::raw_svector_ostream OS(Str);
212 
213   OS << "<tr><td class=\"num\" id=\"LN"
214      << LineNo << "\">"
215      << LineNo << "</td><td class=\"line\">";
216 
217   if (B == E) { // Handle empty lines.
218     OS << " </td></tr>";
219     RB.InsertTextBefore(B, OS.str());
220   } else {
221     RB.InsertTextBefore(B, OS.str());
222     RB.InsertTextBefore(E, "</td></tr>");
223   }
224 }
225 
AddLineNumbers(Rewriter & R,FileID FID)226 void html::AddLineNumbers(Rewriter& R, FileID FID) {
227 
228   const llvm::MemoryBuffer *Buf = R.getSourceMgr().getBuffer(FID);
229   const char* FileBeg = Buf->getBufferStart();
230   const char* FileEnd = Buf->getBufferEnd();
231   const char* C = FileBeg;
232   RewriteBuffer &RB = R.getEditBuffer(FID);
233 
234   assert (C <= FileEnd);
235 
236   unsigned LineNo = 0;
237   unsigned FilePos = 0;
238 
239   while (C != FileEnd) {
240 
241     ++LineNo;
242     unsigned LineStartPos = FilePos;
243     unsigned LineEndPos = FileEnd - FileBeg;
244 
245     assert (FilePos <= LineEndPos);
246     assert (C < FileEnd);
247 
248     // Scan until the newline (or end-of-file).
249 
250     while (C != FileEnd) {
251       char c = *C;
252       ++C;
253 
254       if (c == '\n') {
255         LineEndPos = FilePos++;
256         break;
257       }
258 
259       ++FilePos;
260     }
261 
262     AddLineNumber(RB, LineNo, LineStartPos, LineEndPos);
263   }
264 
265   // Add one big table tag that surrounds all of the code.
266   RB.InsertTextBefore(0, "<table class=\"code\">\n");
267   RB.InsertTextAfter(FileEnd - FileBeg, "</table>");
268 }
269 
AddHeaderFooterInternalBuiltinCSS(Rewriter & R,FileID FID,const char * title)270 void html::AddHeaderFooterInternalBuiltinCSS(Rewriter& R, FileID FID,
271                                              const char *title) {
272 
273   const llvm::MemoryBuffer *Buf = R.getSourceMgr().getBuffer(FID);
274   const char* FileStart = Buf->getBufferStart();
275   const char* FileEnd = Buf->getBufferEnd();
276 
277   SourceLocation StartLoc = R.getSourceMgr().getLocForStartOfFile(FID);
278   SourceLocation EndLoc = StartLoc.getLocWithOffset(FileEnd-FileStart);
279 
280   std::string s;
281   llvm::raw_string_ostream os(s);
282   os << "<!doctype html>\n" // Use HTML 5 doctype
283         "<html>\n<head>\n";
284 
285   if (title)
286     os << "<title>" << html::EscapeText(title) << "</title>\n";
287 
288   os << "<style type=\"text/css\">\n"
289       " body { color:#000000; background-color:#ffffff }\n"
290       " body { font-family:Helvetica, sans-serif; font-size:10pt }\n"
291       " h1 { font-size:14pt }\n"
292       " .code { border-collapse:collapse; width:100%; }\n"
293       " .code { font-family: \"Monospace\", monospace; font-size:10pt }\n"
294       " .code { line-height: 1.2em }\n"
295       " .comment { color: green; font-style: oblique }\n"
296       " .keyword { color: blue }\n"
297       " .string_literal { color: red }\n"
298       " .directive { color: darkmagenta }\n"
299       // Macro expansions.
300       " .expansion { display: none; }\n"
301       " .macro:hover .expansion { display: block; border: 2px solid #FF0000; "
302           "padding: 2px; background-color:#FFF0F0; font-weight: normal; "
303           "  -webkit-border-radius:5px;  -webkit-box-shadow:1px 1px 7px #000; "
304           "position: absolute; top: -1em; left:10em; z-index: 1 } \n"
305       " .macro { color: darkmagenta; background-color:LemonChiffon;"
306              // Macros are position: relative to provide base for expansions.
307              " position: relative }\n"
308       " .num { width:2.5em; padding-right:2ex; background-color:#eeeeee }\n"
309       " .num { text-align:right; font-size:8pt }\n"
310       " .num { color:#444444 }\n"
311       " .line { padding-left: 1ex; border-left: 3px solid #ccc }\n"
312       " .line { white-space: pre }\n"
313       " .msg { -webkit-box-shadow:1px 1px 7px #000 }\n"
314       " .msg { -webkit-border-radius:5px }\n"
315       " .msg { font-family:Helvetica, sans-serif; font-size:8pt }\n"
316       " .msg { float:left }\n"
317       " .msg { padding:0.25em 1ex 0.25em 1ex }\n"
318       " .msg { margin-top:10px; margin-bottom:10px }\n"
319       " .msg { font-weight:bold }\n"
320       " .msg { max-width:60em; word-wrap: break-word; white-space: pre-wrap }\n"
321       " .msgT { padding:0x; spacing:0x }\n"
322       " .msgEvent { background-color:#fff8b4; color:#000000 }\n"
323       " .msgControl { background-color:#bbbbbb; color:#000000 }\n"
324       " .mrange { background-color:#dfddf3 }\n"
325       " .mrange { border-bottom:1px solid #6F9DBE }\n"
326       " .PathIndex { font-weight: bold; padding:0px 5px; "
327         "margin-right:5px; }\n"
328       " .PathIndex { -webkit-border-radius:8px }\n"
329       " .PathIndexEvent { background-color:#bfba87 }\n"
330       " .PathIndexControl { background-color:#8c8c8c }\n"
331       " .PathNav a { text-decoration:none; font-size: larger }\n"
332       " .CodeInsertionHint { font-weight: bold; background-color: #10dd10 }\n"
333       " .CodeRemovalHint { background-color:#de1010 }\n"
334       " .CodeRemovalHint { border-bottom:1px solid #6F9DBE }\n"
335       " table.simpletable {\n"
336       "   padding: 5px;\n"
337       "   font-size:12pt;\n"
338       "   margin:20px;\n"
339       "   border-collapse: collapse; border-spacing: 0px;\n"
340       " }\n"
341       " td.rowname {\n"
342       "   text-align:right; font-weight:bold; color:#444444;\n"
343       "   padding-right:2ex; }\n"
344       "</style>\n</head>\n<body>";
345 
346   // Generate header
347   R.InsertTextBefore(StartLoc, os.str());
348   // Generate footer
349 
350   R.InsertTextAfter(EndLoc, "</body></html>\n");
351 }
352 
353 /// SyntaxHighlight - Relex the specified FileID and annotate the HTML with
354 /// information about keywords, macro expansions etc.  This uses the macro
355 /// table state from the end of the file, so it won't be perfectly perfect,
356 /// but it will be reasonably close.
SyntaxHighlight(Rewriter & R,FileID FID,const Preprocessor & PP)357 void html::SyntaxHighlight(Rewriter &R, FileID FID, const Preprocessor &PP) {
358   RewriteBuffer &RB = R.getEditBuffer(FID);
359 
360   const SourceManager &SM = PP.getSourceManager();
361   const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
362   Lexer L(FID, FromFile, SM, PP.getLangOpts());
363   const char *BufferStart = L.getBuffer().data();
364 
365   // Inform the preprocessor that we want to retain comments as tokens, so we
366   // can highlight them.
367   L.SetCommentRetentionState(true);
368 
369   // Lex all the tokens in raw mode, to avoid entering #includes or expanding
370   // macros.
371   Token Tok;
372   L.LexFromRawLexer(Tok);
373 
374   while (Tok.isNot(tok::eof)) {
375     // Since we are lexing unexpanded tokens, all tokens are from the main
376     // FileID.
377     unsigned TokOffs = SM.getFileOffset(Tok.getLocation());
378     unsigned TokLen = Tok.getLength();
379     switch (Tok.getKind()) {
380     default: break;
381     case tok::identifier:
382       llvm_unreachable("tok::identifier in raw lexing mode!");
383     case tok::raw_identifier: {
384       // Fill in Result.IdentifierInfo and update the token kind,
385       // looking up the identifier in the identifier table.
386       PP.LookUpIdentifierInfo(Tok);
387 
388       // If this is a pp-identifier, for a keyword, highlight it as such.
389       if (Tok.isNot(tok::identifier))
390         HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
391                        "<span class='keyword'>", "</span>");
392       break;
393     }
394     case tok::comment:
395       HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
396                      "<span class='comment'>", "</span>");
397       break;
398     case tok::utf8_string_literal:
399       // Chop off the u part of u8 prefix
400       ++TokOffs;
401       --TokLen;
402       // FALL THROUGH to chop the 8
403     case tok::wide_string_literal:
404     case tok::utf16_string_literal:
405     case tok::utf32_string_literal:
406       // Chop off the L, u, U or 8 prefix
407       ++TokOffs;
408       --TokLen;
409       // FALL THROUGH.
410     case tok::string_literal:
411       // FIXME: Exclude the optional ud-suffix from the highlighted range.
412       HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
413                      "<span class='string_literal'>", "</span>");
414       break;
415     case tok::hash: {
416       // If this is a preprocessor directive, all tokens to end of line are too.
417       if (!Tok.isAtStartOfLine())
418         break;
419 
420       // Eat all of the tokens until we get to the next one at the start of
421       // line.
422       unsigned TokEnd = TokOffs+TokLen;
423       L.LexFromRawLexer(Tok);
424       while (!Tok.isAtStartOfLine() && Tok.isNot(tok::eof)) {
425         TokEnd = SM.getFileOffset(Tok.getLocation())+Tok.getLength();
426         L.LexFromRawLexer(Tok);
427       }
428 
429       // Find end of line.  This is a hack.
430       HighlightRange(RB, TokOffs, TokEnd, BufferStart,
431                      "<span class='directive'>", "</span>");
432 
433       // Don't skip the next token.
434       continue;
435     }
436     }
437 
438     L.LexFromRawLexer(Tok);
439   }
440 }
441 
442 /// HighlightMacros - This uses the macro table state from the end of the
443 /// file, to re-expand macros and insert (into the HTML) information about the
444 /// macro expansions.  This won't be perfectly perfect, but it will be
445 /// reasonably close.
HighlightMacros(Rewriter & R,FileID FID,const Preprocessor & PP)446 void html::HighlightMacros(Rewriter &R, FileID FID, const Preprocessor& PP) {
447   // Re-lex the raw token stream into a token buffer.
448   const SourceManager &SM = PP.getSourceManager();
449   std::vector<Token> TokenStream;
450 
451   const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
452   Lexer L(FID, FromFile, SM, PP.getLangOpts());
453 
454   // Lex all the tokens in raw mode, to avoid entering #includes or expanding
455   // macros.
456   while (1) {
457     Token Tok;
458     L.LexFromRawLexer(Tok);
459 
460     // If this is a # at the start of a line, discard it from the token stream.
461     // We don't want the re-preprocess step to see #defines, #includes or other
462     // preprocessor directives.
463     if (Tok.is(tok::hash) && Tok.isAtStartOfLine())
464       continue;
465 
466     // If this is a ## token, change its kind to unknown so that repreprocessing
467     // it will not produce an error.
468     if (Tok.is(tok::hashhash))
469       Tok.setKind(tok::unknown);
470 
471     // If this raw token is an identifier, the raw lexer won't have looked up
472     // the corresponding identifier info for it.  Do this now so that it will be
473     // macro expanded when we re-preprocess it.
474     if (Tok.is(tok::raw_identifier))
475       PP.LookUpIdentifierInfo(Tok);
476 
477     TokenStream.push_back(Tok);
478 
479     if (Tok.is(tok::eof)) break;
480   }
481 
482   // Temporarily change the diagnostics object so that we ignore any generated
483   // diagnostics from this pass.
484   DiagnosticsEngine TmpDiags(PP.getDiagnostics().getDiagnosticIDs(),
485                              &PP.getDiagnostics().getDiagnosticOptions(),
486                       new IgnoringDiagConsumer);
487 
488   // FIXME: This is a huge hack; we reuse the input preprocessor because we want
489   // its state, but we aren't actually changing it (we hope). This should really
490   // construct a copy of the preprocessor.
491   Preprocessor &TmpPP = const_cast<Preprocessor&>(PP);
492   DiagnosticsEngine *OldDiags = &TmpPP.getDiagnostics();
493   TmpPP.setDiagnostics(TmpDiags);
494 
495   // Inform the preprocessor that we don't want comments.
496   TmpPP.SetCommentRetentionState(false, false);
497 
498   // We don't want pragmas either. Although we filtered out #pragma, removing
499   // _Pragma and __pragma is much harder.
500   bool PragmasPreviouslyEnabled = TmpPP.getPragmasEnabled();
501   TmpPP.setPragmasEnabled(false);
502 
503   // Enter the tokens we just lexed.  This will cause them to be macro expanded
504   // but won't enter sub-files (because we removed #'s).
505   TmpPP.EnterTokenStream(&TokenStream[0], TokenStream.size(), false, false);
506 
507   TokenConcatenation ConcatInfo(TmpPP);
508 
509   // Lex all the tokens.
510   Token Tok;
511   TmpPP.Lex(Tok);
512   while (Tok.isNot(tok::eof)) {
513     // Ignore non-macro tokens.
514     if (!Tok.getLocation().isMacroID()) {
515       TmpPP.Lex(Tok);
516       continue;
517     }
518 
519     // Okay, we have the first token of a macro expansion: highlight the
520     // expansion by inserting a start tag before the macro expansion and
521     // end tag after it.
522     std::pair<SourceLocation, SourceLocation> LLoc =
523       SM.getExpansionRange(Tok.getLocation());
524 
525     // Ignore tokens whose instantiation location was not the main file.
526     if (SM.getFileID(LLoc.first) != FID) {
527       TmpPP.Lex(Tok);
528       continue;
529     }
530 
531     assert(SM.getFileID(LLoc.second) == FID &&
532            "Start and end of expansion must be in the same ultimate file!");
533 
534     std::string Expansion = EscapeText(TmpPP.getSpelling(Tok));
535     unsigned LineLen = Expansion.size();
536 
537     Token PrevPrevTok;
538     Token PrevTok = Tok;
539     // Okay, eat this token, getting the next one.
540     TmpPP.Lex(Tok);
541 
542     // Skip all the rest of the tokens that are part of this macro
543     // instantiation.  It would be really nice to pop up a window with all the
544     // spelling of the tokens or something.
545     while (!Tok.is(tok::eof) &&
546            SM.getExpansionLoc(Tok.getLocation()) == LLoc.first) {
547       // Insert a newline if the macro expansion is getting large.
548       if (LineLen > 60) {
549         Expansion += "<br>";
550         LineLen = 0;
551       }
552 
553       LineLen -= Expansion.size();
554 
555       // If the tokens were already space separated, or if they must be to avoid
556       // them being implicitly pasted, add a space between them.
557       if (Tok.hasLeadingSpace() ||
558           ConcatInfo.AvoidConcat(PrevPrevTok, PrevTok, Tok))
559         Expansion += ' ';
560 
561       // Escape any special characters in the token text.
562       Expansion += EscapeText(TmpPP.getSpelling(Tok));
563       LineLen += Expansion.size();
564 
565       PrevPrevTok = PrevTok;
566       PrevTok = Tok;
567       TmpPP.Lex(Tok);
568     }
569 
570 
571     // Insert the expansion as the end tag, so that multi-line macros all get
572     // highlighted.
573     Expansion = "<span class='expansion'>" + Expansion + "</span></span>";
574 
575     HighlightRange(R, LLoc.first, LLoc.second,
576                    "<span class='macro'>", Expansion.c_str());
577   }
578 
579   // Restore the preprocessor's old state.
580   TmpPP.setDiagnostics(*OldDiags);
581   TmpPP.setPragmasEnabled(PragmasPreviouslyEnabled);
582 }
583