/// LexRawTokensFromMainFile - Lets all the raw tokens from the main file into
/// the specified vector.
static void LexRawTokensFromMainFile(Preprocessor &PP,
std::vector<Token> &RawTokens) {
SourceManager &SM = PP.getSourceManager();
// Create a lexer to lex all the tokens of the main file in raw mode. Even
// though it is in raw mode, it will not return comments.
const llvm::MemoryBuffer *FromFile = SM.getBuffer(SM.getMainFileID());
Lexer RawLex(SM.getMainFileID(), FromFile, SM, PP.getLangOpts());
// Switch on comment lexing because we really do want them.
RawLex.SetCommentRetentionState(true);
Token RawTok;
do {
RawLex.LexFromRawLexer(RawTok);
// If we have an identifier with no identifier info for our raw token, look
// up the indentifier info. This is important for equality comparison of
// identifier tokens.
if (RawTok.is(tok::raw_identifier))
PP.LookUpIdentifierInfo(RawTok);
RawTokens.push_back(RawTok);
} while (RawTok.isNot(tok::eof));
}
/// HighlightMacros - This uses the macro table state from the end of the
/// file, to re-expand macros and insert (into the HTML) information about the
/// macro expansions. This won't be perfectly perfect, but it will be
/// reasonably close.
void html::HighlightMacros(Rewriter &R, FileID FID, const Preprocessor& PP) {
// Re-lex the raw token stream into a token buffer.
const SourceManager &SM = PP.getSourceManager();
std::vector<Token> TokenStream;
const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
Lexer L(FID, FromFile, SM, PP.getLangOpts());
// Lex all the tokens in raw mode, to avoid entering #includes or expanding
// macros.
while (1) {
Token Tok;
L.LexFromRawLexer(Tok);
// If this is a # at the start of a line, discard it from the token stream.
// We don't want the re-preprocess step to see #defines, #includes or other
// preprocessor directives.
if (Tok.is(tok::hash) && Tok.isAtStartOfLine())
continue;
// If this is a ## token, change its kind to unknown so that repreprocessing
// it will not produce an error.
if (Tok.is(tok::hashhash))
Tok.setKind(tok::unknown);
// If this raw token is an identifier, the raw lexer won't have looked up
// the corresponding identifier info for it. Do this now so that it will be
// macro expanded when we re-preprocess it.
if (Tok.is(tok::raw_identifier))
PP.LookUpIdentifierInfo(Tok);
TokenStream.push_back(Tok);
if (Tok.is(tok::eof)) break;
}
// Temporarily change the diagnostics object so that we ignore any generated
// diagnostics from this pass.
DiagnosticsEngine TmpDiags(PP.getDiagnostics().getDiagnosticIDs(),
&PP.getDiagnostics().getDiagnosticOptions(),
new IgnoringDiagConsumer);
// FIXME: This is a huge hack; we reuse the input preprocessor because we want
// its state, but we aren't actually changing it (we hope). This should really
// construct a copy of the preprocessor.
Preprocessor &TmpPP = const_cast<Preprocessor&>(PP);
DiagnosticsEngine *OldDiags = &TmpPP.getDiagnostics();
TmpPP.setDiagnostics(TmpDiags);
// Inform the preprocessor that we don't want comments.
TmpPP.SetCommentRetentionState(false, false);
// We don't want pragmas either. Although we filtered out #pragma, removing
// _Pragma and __pragma is much harder.
bool PragmasPreviouslyEnabled = TmpPP.getPragmasEnabled();
TmpPP.setPragmasEnabled(false);
// Enter the tokens we just lexed. This will cause them to be macro expanded
// but won't enter sub-files (because we removed #'s).
TmpPP.EnterTokenStream(TokenStream, false);
TokenConcatenation ConcatInfo(TmpPP);
// Lex all the tokens.
Token Tok;
TmpPP.Lex(Tok);
while (Tok.isNot(tok::eof)) {
// Ignore non-macro tokens.
if (!Tok.getLocation().isMacroID()) {
TmpPP.Lex(Tok);
continue;
}
// Okay, we have the first token of a macro expansion: highlight the
// expansion by inserting a start tag before the macro expansion and
// end tag after it.
std::pair<SourceLocation, SourceLocation> LLoc =
SM.getExpansionRange(Tok.getLocation());
// Ignore tokens whose instantiation location was not the main file.
if (SM.getFileID(LLoc.first) != FID) {
TmpPP.Lex(Tok);
continue;
}
assert(SM.getFileID(LLoc.second) == FID &&
"Start and end of expansion must be in the same ultimate file!");
std::string Expansion = EscapeText(TmpPP.getSpelling(Tok));
unsigned LineLen = Expansion.size();
Token PrevPrevTok;
Token PrevTok = Tok;
// Okay, eat this token, getting the next one.
TmpPP.Lex(Tok);
// Skip all the rest of the tokens that are part of this macro
// instantiation. It would be really nice to pop up a window with all the
// spelling of the tokens or something.
while (!Tok.is(tok::eof) &&
SM.getExpansionLoc(Tok.getLocation()) == LLoc.first) {
// Insert a newline if the macro expansion is getting large.
if (LineLen > 60) {
Expansion += "<br>";
LineLen = 0;
}
LineLen -= Expansion.size();
// If the tokens were already space separated, or if they must be to avoid
// them being implicitly pasted, add a space between them.
if (Tok.hasLeadingSpace() ||
ConcatInfo.AvoidConcat(PrevPrevTok, PrevTok, Tok))
Expansion += ' ';
// Escape any special characters in the token text.
Expansion += EscapeText(TmpPP.getSpelling(Tok));
LineLen += Expansion.size();
PrevPrevTok = PrevTok;
PrevTok = Tok;
TmpPP.Lex(Tok);
}
// Insert the expansion as the end tag, so that multi-line macros all get
// highlighted.
Expansion = "<span class='expansion'>" + Expansion + "</span></span>";
HighlightRange(R, LLoc.first, LLoc.second,
"<span class='macro'>", Expansion.c_str());
}
// Restore the preprocessor's old state.
TmpPP.setDiagnostics(*OldDiags);
TmpPP.setPragmasEnabled(PragmasPreviouslyEnabled);
}
/// SyntaxHighlight - Relex the specified FileID and annotate the HTML with
/// information about keywords, macro expansions etc. This uses the macro
/// table state from the end of the file, so it won't be perfectly perfect,
/// but it will be reasonably close.
void html::SyntaxHighlight(Rewriter &R, FileID FID, const Preprocessor &PP) {
RewriteBuffer &RB = R.getEditBuffer(FID);
const SourceManager &SM = PP.getSourceManager();
const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID);
Lexer L(FID, FromFile, SM, PP.getLangOpts());
const char *BufferStart = L.getBuffer().data();
// Inform the preprocessor that we want to retain comments as tokens, so we
// can highlight them.
L.SetCommentRetentionState(true);
// Lex all the tokens in raw mode, to avoid entering #includes or expanding
// macros.
Token Tok;
L.LexFromRawLexer(Tok);
while (Tok.isNot(tok::eof)) {
// Since we are lexing unexpanded tokens, all tokens are from the main
// FileID.
unsigned TokOffs = SM.getFileOffset(Tok.getLocation());
unsigned TokLen = Tok.getLength();
switch (Tok.getKind()) {
default: break;
case tok::identifier:
llvm_unreachable("tok::identifier in raw lexing mode!");
case tok::raw_identifier: {
// Fill in Result.IdentifierInfo and update the token kind,
// looking up the identifier in the identifier table.
PP.LookUpIdentifierInfo(Tok);
// If this is a pp-identifier, for a keyword, highlight it as such.
if (Tok.isNot(tok::identifier))
HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
"<span class='keyword'>", "</span>");
break;
}
case tok::comment:
HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
"<span class='comment'>", "</span>");
break;
case tok::utf8_string_literal:
// Chop off the u part of u8 prefix
++TokOffs;
--TokLen;
// FALL THROUGH to chop the 8
case tok::wide_string_literal:
case tok::utf16_string_literal:
case tok::utf32_string_literal:
// Chop off the L, u, U or 8 prefix
++TokOffs;
--TokLen;
// FALL THROUGH.
case tok::string_literal:
// FIXME: Exclude the optional ud-suffix from the highlighted range.
HighlightRange(RB, TokOffs, TokOffs+TokLen, BufferStart,
"<span class='string_literal'>", "</span>");
break;
case tok::hash: {
// If this is a preprocessor directive, all tokens to end of line are too.
if (!Tok.isAtStartOfLine())
break;
// Eat all of the tokens until we get to the next one at the start of
// line.
unsigned TokEnd = TokOffs+TokLen;
L.LexFromRawLexer(Tok);
while (!Tok.isAtStartOfLine() && Tok.isNot(tok::eof)) {
TokEnd = SM.getFileOffset(Tok.getLocation())+Tok.getLength();
L.LexFromRawLexer(Tok);
}
// Find end of line. This is a hack.
HighlightRange(RB, TokOffs, TokEnd, BufferStart,
"<span class='directive'>", "</span>");
// Don't skip the next token.
continue;
}
}
L.LexFromRawLexer(Tok);
}
}
/// HighlightMacros - This uses the macro table state from the end of the
/// file, to re-expand macros and insert (into the HTML) information about the
/// macro expansions. This won't be perfectly perfect, but it will be
/// reasonably close.
void html::HighlightMacros(Rewriter &R, FileID FID, Preprocessor& PP) {
// Re-lex the raw token stream into a token buffer.
const SourceManager &SM = PP.getSourceManager();
std::vector<Token> TokenStream;
Lexer L(FID, SM, PP.getLangOptions());
// Lex all the tokens in raw mode, to avoid entering #includes or expanding
// macros.
while (1) {
Token Tok;
L.LexFromRawLexer(Tok);
// If this is a # at the start of a line, discard it from the token stream.
// We don't want the re-preprocess step to see #defines, #includes or other
// preprocessor directives.
if (Tok.is(tok::hash) && Tok.isAtStartOfLine())
continue;
// If this is a ## token, change its kind to unknown so that repreprocessing
// it will not produce an error.
if (Tok.is(tok::hashhash))
Tok.setKind(tok::unknown);
// If this raw token is an identifier, the raw lexer won't have looked up
// the corresponding identifier info for it. Do this now so that it will be
// macro expanded when we re-preprocess it.
if (Tok.is(tok::identifier)) {
// Change the kind of this identifier to the appropriate token kind, e.g.
// turning "for" into a keyword.
Tok.setKind(PP.LookUpIdentifierInfo(Tok)->getTokenID());
}
TokenStream.push_back(Tok);
if (Tok.is(tok::eof)) break;
}
// Temporarily change the diagnostics object so that we ignore any generated
// diagnostics from this pass.
IgnoringDiagClient TmpDC;
Diagnostic TmpDiags(&TmpDC);
Diagnostic *OldDiags = &PP.getDiagnostics();
PP.setDiagnostics(TmpDiags);
// Inform the preprocessor that we don't want comments.
PP.SetCommentRetentionState(false, false);
// Enter the tokens we just lexed. This will cause them to be macro expanded
// but won't enter sub-files (because we removed #'s).
PP.EnterTokenStream(&TokenStream[0], TokenStream.size(), false, false);
TokenConcatenation ConcatInfo(PP);
// Lex all the tokens.
Token Tok;
PP.Lex(Tok);
while (Tok.isNot(tok::eof)) {
// Ignore non-macro tokens.
if (!Tok.getLocation().isMacroID()) {
PP.Lex(Tok);
continue;
}
// Okay, we have the first token of a macro expansion: highlight the
// instantiation by inserting a start tag before the macro instantiation and
// end tag after it.
std::pair<SourceLocation, SourceLocation> LLoc =
SM.getInstantiationRange(Tok.getLocation());
// Ignore tokens whose instantiation location was not the main file.
if (SM.getFileID(LLoc.first) != FID) {
PP.Lex(Tok);
continue;
}
assert(SM.getFileID(LLoc.second) == FID &&
"Start and end of expansion must be in the same ultimate file!");
std::string Expansion = PP.getSpelling(Tok);
unsigned LineLen = Expansion.size();
Token PrevTok = Tok;
// Okay, eat this token, getting the next one.
PP.Lex(Tok);
// Skip all the rest of the tokens that are part of this macro
// instantiation. It would be really nice to pop up a window with all the
// spelling of the tokens or something.
while (!Tok.is(tok::eof) &&
SM.getInstantiationLoc(Tok.getLocation()) == LLoc.first) {
// Insert a newline if the macro expansion is getting large.
if (LineLen > 60) {
Expansion += "<br>";
LineLen = 0;
}
LineLen -= Expansion.size();
// If the tokens were already space separated, or if they must be to avoid
// them being implicitly pasted, add a space between them.
if (Tok.hasLeadingSpace() ||
ConcatInfo.AvoidConcat(PrevTok, Tok))
Expansion += ' ';
// Escape any special characters in the token text.
Expansion += EscapeText(PP.getSpelling(Tok));
LineLen += Expansion.size();
PrevTok = Tok;
PP.Lex(Tok);
}
// Insert the expansion as the end tag, so that multi-line macros all get
// highlighted.
Expansion = "<span class='expansion'>" + Expansion + "</span></span>";
HighlightRange(R, LLoc.first, LLoc.second,
"<span class='macro'>", Expansion.c_str());
}
// Restore diagnostics object back to its own thing.
PP.setDiagnostics(*OldDiags);
}