1 //== HTMLRewrite.cpp - Translate source code into prettified HTML --*- C++ -*-//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file defines the HTMLRewriter clas, which is used to translate the
11 // text of a source file into prettified HTML.
13 //===----------------------------------------------------------------------===//
15 #include "clang/Lex/Preprocessor.h"
16 #include "clang/Rewrite/Core/Rewriter.h"
17 #include "clang/Rewrite/Core/HTMLRewrite.h"
18 #include "clang/Lex/TokenConcatenation.h"
19 #include "clang/Lex/Preprocessor.h"
20 #include "clang/Basic/SourceManager.h"
21 #include "llvm/ADT/SmallString.h"
22 #include "llvm/ADT/OwningPtr.h"
23 #include "llvm/Support/ErrorHandling.h"
24 #include "llvm/Support/MemoryBuffer.h"
25 #include "llvm/Support/raw_ostream.h"
26 using namespace clang;
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
33 void html::HighlightRange(Rewriter &R, SourceLocation B, SourceLocation E,
34 const char *StartTag, const char *EndTag) {
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!");
41 unsigned BOffset = SM.getFileOffset(B);
42 unsigned EOffset = SM.getFileOffset(E);
44 // Include the whole end token in the range.
45 EOffset += Lexer::MeasureTokenLength(E, R.getSourceMgr(), R.getLangOpts());
48 const char *BufferStart = SM.getBufferData(FID, &Invalid).data();
52 HighlightRange(R.getEditBuffer(FID), BOffset, EOffset,
53 BufferStart, StartTag, EndTag);
56 /// HighlightRange - This is the same as the above method, but takes
57 /// decomposed file locations.
58 void html::HighlightRange(RewriteBuffer &RB, unsigned B, unsigned E,
59 const char *BufferStart,
60 const char *StartTag, const char *EndTag) {
61 // Insert the tag at the absolute start/end of the range.
62 RB.InsertTextAfter(B, StartTag);
63 RB.InsertTextBefore(E, EndTag);
65 // Scan the range to see if there is a \r or \n. If so, and if the line is
66 // not blank, insert tags on that line as well.
67 bool HadOpenTag = true;
69 unsigned LastNonWhiteSpace = B;
70 for (unsigned i = B; i != E; ++i) {
71 switch (BufferStart[i]) {
74 // Okay, we found a newline in the range. If we have an open tag, we need
75 // to insert a close tag at the first non-whitespace before the newline.
77 RB.InsertTextBefore(LastNonWhiteSpace+1, EndTag);
79 // Instead of inserting an open tag immediately after the newline, we
80 // wait until we see a non-whitespace character. This prevents us from
81 // inserting tags around blank lines, and also allows the open tag to
82 // be put *after* whitespace on a non-blank line.
94 // If there is no tag open, do it now.
96 RB.InsertTextAfter(i, StartTag);
100 // Remember this character.
101 LastNonWhiteSpace = i;
107 void html::EscapeText(Rewriter &R, FileID FID,
108 bool EscapeSpaces, bool ReplaceTabs) {
110 const llvm::MemoryBuffer *Buf = R.getSourceMgr().getBuffer(FID);
111 const char* C = Buf->getBufferStart();
112 const char* FileEnd = Buf->getBufferEnd();
114 assert (C <= FileEnd);
116 RewriteBuffer &RB = R.getEditBuffer(FID);
119 for (unsigned FilePos = 0; C != FileEnd ; ++C, ++FilePos) {
121 default: ++ColNo; break;
129 RB.ReplaceText(FilePos, 1, " ");
133 RB.ReplaceText(FilePos, 1, "<hr>");
140 unsigned NumSpaces = 8-(ColNo&7);
142 RB.ReplaceText(FilePos, 1,
143 StringRef(" "
144 " ", 6*NumSpaces));
146 RB.ReplaceText(FilePos, 1, StringRef(" ", NumSpaces));
151 RB.ReplaceText(FilePos, 1, "<");
156 RB.ReplaceText(FilePos, 1, ">");
161 RB.ReplaceText(FilePos, 1, "&");
168 std::string html::EscapeText(const std::string& s, bool EscapeSpaces,
171 unsigned len = s.size();
173 llvm::raw_string_ostream os(Str);
175 for (unsigned i = 0 ; i < len; ++i) {
183 if (EscapeSpaces) os << " ";
190 for (unsigned i = 0; i < 4; ++i)
193 for (unsigned i = 0; i < 4; ++i)
201 case '<': os << "<"; break;
202 case '>': os << ">"; break;
203 case '&': os << "&"; break;
210 static void AddLineNumber(RewriteBuffer &RB, unsigned LineNo,
211 unsigned B, unsigned E) {
212 SmallString<256> Str;
213 llvm::raw_svector_ostream OS(Str);
215 OS << "<tr><td class=\"num\" id=\"LN"
217 << LineNo << "</td><td class=\"line\">";
219 if (B == E) { // Handle empty lines.
221 RB.InsertTextBefore(B, OS.str());
223 RB.InsertTextBefore(B, OS.str());
224 RB.InsertTextBefore(E, "</td></tr>");
228 void html::AddLineNumbers(Rewriter& R, FileID FID) {
230 const llvm::MemoryBuffer *Buf = R.getSourceMgr().getBuffer(FID);
231 const char* FileBeg = Buf->getBufferStart();
232 const char* FileEnd = Buf->getBufferEnd();
233 const char* C = FileBeg;
234 RewriteBuffer &RB = R.getEditBuffer(FID);
236 assert (C <= FileEnd);
239 unsigned FilePos = 0;
241 while (C != FileEnd) {
244 unsigned LineStartPos = FilePos;
245 unsigned LineEndPos = FileEnd - FileBeg;
247 assert (FilePos <= LineEndPos);
248 assert (C < FileEnd);
250 // Scan until the newline (or end-of-file).
252 while (C != FileEnd) {
257 LineEndPos = FilePos++;
264 AddLineNumber(RB, LineNo, LineStartPos, LineEndPos);
267 // Add one big table tag that surrounds all of the code.
268 RB.InsertTextBefore(0, "<table class=\"code\">\n");
269 RB.InsertTextAfter(FileEnd - FileBeg, "</table>");
272 void html::AddHeaderFooterInternalBuiltinCSS(Rewriter& R, FileID FID,
275 const llvm::MemoryBuffer *Buf = R.getSourceMgr().getBuffer(FID);
276 const char* FileStart = Buf->getBufferStart();
277 const char* FileEnd = Buf->getBufferEnd();
279 SourceLocation StartLoc = R.getSourceMgr().getLocForStartOfFile(FID);
280 SourceLocation EndLoc = StartLoc.getLocWithOffset(FileEnd-FileStart);
283 llvm::raw_string_ostream os(s);
284 os << "<!doctype html>\n" // Use HTML 5 doctype
288 os << "<title>" << html::EscapeText(title) << "</title>\n";
290 os << "<style type=\"text/css\">\n"
291 " body { color:#000000; background-color:#ffffff }\n"
292 " body { font-family:Helvetica, sans-serif; font-size:10pt }\n"
293 " h1 { font-size:14pt }\n"
294 " .code { border-collapse:collapse; width:100%; }\n"
295 " .code { font-family: \"Monospace\", monospace; font-size:10pt }\n"
296 " .code { line-height: 1.2em }\n"
297 " .comment { color: green; font-style: oblique }\n"
298 " .keyword { color: blue }\n"
299 " .string_literal { color: red }\n"
300 " .directive { color: darkmagenta }\n"
302 " .expansion { display: none; }\n"
303 " .macro:hover .expansion { display: block; border: 2px solid #FF0000; "
304 "padding: 2px; background-color:#FFF0F0; font-weight: normal; "
305 " -webkit-border-radius:5px; -webkit-box-shadow:1px 1px 7px #000; "
306 "position: absolute; top: -1em; left:10em; z-index: 1 } \n"
307 " .macro { color: darkmagenta; background-color:LemonChiffon;"
308 // Macros are position: relative to provide base for expansions.
309 " position: relative }\n"
310 " .num { width:2.5em; padding-right:2ex; background-color:#eeeeee }\n"
311 " .num { text-align:right; font-size:8pt }\n"
312 " .num { color:#444444 }\n"
313 " .line { padding-left: 1ex; border-left: 3px solid #ccc }\n"
314 " .line { white-space: pre }\n"
315 " .msg { -webkit-box-shadow:1px 1px 7px #000 }\n"
316 " .msg { -webkit-border-radius:5px }\n"
317 " .msg { font-family:Helvetica, sans-serif; font-size:8pt }\n"
318 " .msg { float:left }\n"
319 " .msg { padding:0.25em 1ex 0.25em 1ex }\n"
320 " .msg { margin-top:10px; margin-bottom:10px }\n"
321 " .msg { font-weight:bold }\n"
322 " .msg { max-width:60em; word-wrap: break-word; white-space: pre-wrap }\n"
323 " .msgT { padding:0x; spacing:0x }\n"
324 " .msgEvent { background-color:#fff8b4; color:#000000 }\n"
325 " .msgControl { background-color:#bbbbbb; color:#000000 }\n"
326 " .mrange { background-color:#dfddf3 }\n"
327 " .mrange { border-bottom:1px solid #6F9DBE }\n"
328 " .PathIndex { font-weight: bold; padding:0px 5px; "
329 "margin-right:5px; }\n"
330 " .PathIndex { -webkit-border-radius:8px }\n"
331 " .PathIndexEvent { background-color:#bfba87 }\n"
332 " .PathIndexControl { background-color:#8c8c8c }\n"
333 " .PathNav a { text-decoration:none; font-size: larger }\n"
334 " .CodeInsertionHint { font-weight: bold; background-color: #10dd10 }\n"
335 " .CodeRemovalHint { background-color:#de1010 }\n"
336 " .CodeRemovalHint { border-bottom:1px solid #6F9DBE }\n"
337 " table.simpletable {\n"
341 " border-collapse: collapse; border-spacing: 0px;\n"
344 " text-align:right; font-weight:bold; color:#444444;\n"
345 " padding-right:2ex; }\n"
346 "</style>\n</head>\n<body>";
349 R.InsertTextBefore(StartLoc, os.str());
352 R.InsertTextAfter(EndLoc, "</body></html>\n");
355 /// SyntaxHighlight - Relex the specified FileID and annotate the HTML with
356 /// information about keywords, macro expansions etc. This uses the macro
357 /// table state from the end of the file, so it won't be perfectly perfect,
358 /// but it will be reasonably close.
359 void html::SyntaxHighlight(Rewriter &R, FileID FID, const Preprocessor &PP) {
360 RewriteBuffer &RB = R.getEditBuffer(FID);
362 const SourceManager &SM = PP.getSourceManager();
363 const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
364 Lexer L(FID, FromFile, SM, PP.getLangOpts());
365 const char *BufferStart = L.getBufferStart();
367 // Inform the preprocessor that we want to retain comments as tokens, so we
368 // can highlight them.
369 L.SetCommentRetentionState(true);
371 // Lex all the tokens in raw mode, to avoid entering #includes or expanding
374 L.LexFromRawLexer(Tok);
376 while (Tok.isNot(tok::eof)) {
377 // Since we are lexing unexpanded tokens, all tokens are from the main
379 unsigned TokOffs = SM.getFileOffset(Tok.getLocation());
380 unsigned TokLen = Tok.getLength();
381 switch (Tok.getKind()) {
383 case tok::identifier:
384 llvm_unreachable("tok::identifier in raw lexing mode!");
385 case tok::raw_identifier: {
386 // Fill in Result.IdentifierInfo and update the token kind,
387 // looking up the identifier in the identifier table.
388 PP.LookUpIdentifierInfo(Tok);
390 // If this is a pp-identifier, for a keyword, highlight it as such.
391 if (Tok.isNot(tok::identifier))
392 HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
393 "<span class='keyword'>", "</span>");
397 HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
398 "<span class='comment'>", "</span>");
400 case tok::utf8_string_literal:
401 // Chop off the u part of u8 prefix
404 // FALL THROUGH to chop the 8
405 case tok::wide_string_literal:
406 case tok::utf16_string_literal:
407 case tok::utf32_string_literal:
408 // Chop off the L, u, U or 8 prefix
412 case tok::string_literal:
413 // FIXME: Exclude the optional ud-suffix from the highlighted range.
414 HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
415 "<span class='string_literal'>", "</span>");
418 // If this is a preprocessor directive, all tokens to end of line are too.
419 if (!Tok.isAtStartOfLine())
422 // Eat all of the tokens until we get to the next one at the start of
424 unsigned TokEnd = TokOffs+TokLen;
425 L.LexFromRawLexer(Tok);
426 while (!Tok.isAtStartOfLine() && Tok.isNot(tok::eof)) {
427 TokEnd = SM.getFileOffset(Tok.getLocation())+Tok.getLength();
428 L.LexFromRawLexer(Tok);
431 // Find end of line. This is a hack.
432 HighlightRange(RB, TokOffs, TokEnd, BufferStart,
433 "<span class='directive'>", "</span>");
435 // Don't skip the next token.
440 L.LexFromRawLexer(Tok);
444 /// HighlightMacros - This uses the macro table state from the end of the
445 /// file, to re-expand macros and insert (into the HTML) information about the
446 /// macro expansions. This won't be perfectly perfect, but it will be
447 /// reasonably close.
448 void html::HighlightMacros(Rewriter &R, FileID FID, const Preprocessor& PP) {
449 // Re-lex the raw token stream into a token buffer.
450 const SourceManager &SM = PP.getSourceManager();
451 std::vector<Token> TokenStream;
453 const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
454 Lexer L(FID, FromFile, SM, PP.getLangOpts());
456 // Lex all the tokens in raw mode, to avoid entering #includes or expanding
460 L.LexFromRawLexer(Tok);
462 // If this is a # at the start of a line, discard it from the token stream.
463 // We don't want the re-preprocess step to see #defines, #includes or other
464 // preprocessor directives.
465 if (Tok.is(tok::hash) && Tok.isAtStartOfLine())
468 // If this is a ## token, change its kind to unknown so that repreprocessing
469 // it will not produce an error.
470 if (Tok.is(tok::hashhash))
471 Tok.setKind(tok::unknown);
473 // If this raw token is an identifier, the raw lexer won't have looked up
474 // the corresponding identifier info for it. Do this now so that it will be
475 // macro expanded when we re-preprocess it.
476 if (Tok.is(tok::raw_identifier))
477 PP.LookUpIdentifierInfo(Tok);
479 TokenStream.push_back(Tok);
481 if (Tok.is(tok::eof)) break;
484 // Temporarily change the diagnostics object so that we ignore any generated
485 // diagnostics from this pass.
486 DiagnosticsEngine TmpDiags(PP.getDiagnostics().getDiagnosticIDs(),
487 &PP.getDiagnostics().getDiagnosticOptions(),
488 new IgnoringDiagConsumer);
490 // FIXME: This is a huge hack; we reuse the input preprocessor because we want
491 // its state, but we aren't actually changing it (we hope). This should really
492 // construct a copy of the preprocessor.
493 Preprocessor &TmpPP = const_cast<Preprocessor&>(PP);
494 DiagnosticsEngine *OldDiags = &TmpPP.getDiagnostics();
495 TmpPP.setDiagnostics(TmpDiags);
497 // Inform the preprocessor that we don't want comments.
498 TmpPP.SetCommentRetentionState(false, false);
500 // We don't want pragmas either. Although we filtered out #pragma, removing
501 // _Pragma and __pragma is much harder.
502 bool PragmasPreviouslyEnabled = TmpPP.getPragmasEnabled();
503 TmpPP.setPragmasEnabled(false);
505 // Enter the tokens we just lexed. This will cause them to be macro expanded
506 // but won't enter sub-files (because we removed #'s).
507 TmpPP.EnterTokenStream(&TokenStream[0], TokenStream.size(), false, false);
509 TokenConcatenation ConcatInfo(TmpPP);
511 // Lex all the tokens.
514 while (Tok.isNot(tok::eof)) {
515 // Ignore non-macro tokens.
516 if (!Tok.getLocation().isMacroID()) {
521 // Okay, we have the first token of a macro expansion: highlight the
522 // expansion by inserting a start tag before the macro expansion and
524 std::pair<SourceLocation, SourceLocation> LLoc =
525 SM.getExpansionRange(Tok.getLocation());
527 // Ignore tokens whose instantiation location was not the main file.
528 if (SM.getFileID(LLoc.first) != FID) {
533 assert(SM.getFileID(LLoc.second) == FID &&
534 "Start and end of expansion must be in the same ultimate file!");
536 std::string Expansion = EscapeText(TmpPP.getSpelling(Tok));
537 unsigned LineLen = Expansion.size();
541 // Okay, eat this token, getting the next one.
544 // Skip all the rest of the tokens that are part of this macro
545 // instantiation. It would be really nice to pop up a window with all the
546 // spelling of the tokens or something.
547 while (!Tok.is(tok::eof) &&
548 SM.getExpansionLoc(Tok.getLocation()) == LLoc.first) {
549 // Insert a newline if the macro expansion is getting large.
555 LineLen -= Expansion.size();
557 // If the tokens were already space separated, or if they must be to avoid
558 // them being implicitly pasted, add a space between them.
559 if (Tok.hasLeadingSpace() ||
560 ConcatInfo.AvoidConcat(PrevPrevTok, PrevTok, Tok))
563 // Escape any special characters in the token text.
564 Expansion += EscapeText(TmpPP.getSpelling(Tok));
565 LineLen += Expansion.size();
567 PrevPrevTok = PrevTok;
573 // Insert the expansion as the end tag, so that multi-line macros all get
575 Expansion = "<span class='expansion'>" + Expansion + "</span></span>";
577 HighlightRange(R, LLoc.first, LLoc.second,
578 "<span class='macro'>", Expansion.c_str());
581 // Restore the preprocessor's old state.
582 TmpPP.setDiagnostics(*OldDiags);
583 TmpPP.setPragmasEnabled(PragmasPreviouslyEnabled);