//
// Parse a subchain of input delimited by matching pairs: [],(),{},<>
// [WARNING: no other subchain types are recognized!]. This function expects
// g_cxx.pToken to point to the initial token of the pair ([{<.
// It will parse input until the matching terminator token is found.
// Inner parsing is done by cxxParserParseAndCondenseSubchainsUpToOneOf()
// so this is actually a recursive subchain nesting algorithm.
//
boolean cxxParserParseAndCondenseCurrentSubchain(
unsigned int uInitialSubchainMarkerTypes,
boolean bAcceptEOF
)
{
CXXTokenChain * pCurrentChain = g_cxx.pTokenChain;
g_cxx.pTokenChain = cxxTokenChainCreate();
CXXToken * pInitial = cxxTokenChainTakeLast(pCurrentChain);
cxxTokenChainAppend(g_cxx.pTokenChain,pInitial);
CXXToken * pChainToken = cxxTokenCreate();
pChainToken->iLineNumber = pInitial->iLineNumber;
pChainToken->oFilePosition = pInitial->oFilePosition;
// see the declaration of CXXTokenType enum.
// Shifting by 8 gives the corresponding chain marker
pChainToken->eType = (enum CXXTokenType)(g_cxx.pToken->eType << 8);
pChainToken->pChain = g_cxx.pTokenChain;
cxxTokenChainAppend(pCurrentChain,pChainToken);
// see the declaration of CXXTokenType enum.
// Shifting by 4 gives the corresponding closing token type
unsigned int uTokenTypes = g_cxx.pToken->eType << 4;
if(bAcceptEOF)
uTokenTypes |= CXXTokenTypeEOF;
boolean bRet = cxxParserParseAndCondenseSubchainsUpToOneOf(
uTokenTypes,
uInitialSubchainMarkerTypes
);
g_cxx.pTokenChain = pCurrentChain;
g_cxx.pToken = pCurrentChain->pTail;
return bRet;
}
//
// This function parses input until one of the specified tokens appears.
// The algorithm will also build subchains of matching pairs ([...],(...),{...}).
// Within the subchain analysis of uTokenTypes is completly disabled.
// Subchains do nest.
//
// Please note that this function will skip entire scopes (matching {} pairs)
// unless you pass CXXTokenTypeOpeningBracket to stop at their beginning.
// This is usually what you want, unless you're really expecting a scope to begin
// in the current statement.
//
boolean cxxParserParseUpToOneOf(unsigned int uTokenTypes)
{
return cxxParserParseAndCondenseSubchainsUpToOneOf(
uTokenTypes,
CXXTokenTypeOpeningBracket |
CXXTokenTypeOpeningParenthesis |
CXXTokenTypeOpeningSquareParenthesis
);
}
//
// Parses the <parameters> part of the template prefix.
// Here we are pointing at the initial < (but the token chain has been emptied by cxxParserParseTemplatePrefix())
//
static boolean cxxParserParseTemplatePrefixAngleBrackets(void)
{
CXX_DEBUG_ENTER();
// Here we have the big prolem of <> characters which may be
// template argument delimiters, less than/greater than operators, shift left/right operators.
//
// A well written code will have parentheses around all the ambiguous cases. We handle that
// and we permit any kind of syntax inside a parenthesis.
//
// Without parentheses we still try to handle the << and >> shift operator cases:
// - << is always recognized as shift operator
// - >> is recognized as shift unless it's non-nested. This is what C++11 spec says
// and theoretically it should be also pseudo-compatible with C++03 which treats this
// case as a syntax error.
//
// The 'less-than' and 'greater-than' operators are hopeless in the general case: gcc
// is smart enough to figure them out by looking at the identifiers around but without
// proper state (include headers, macro expansion, full type database etc) we simply can't
// do the same. However, we try to recover if we figure out we (or the programmer?) screwed up.
int iTemplateLevel = 0;
for(;;)
{
// Within parentheses everything is permitted.
if(!cxxParserParseAndCondenseSubchainsUpToOneOf(
CXXTokenTypeGreaterThanSign | CXXTokenTypeSmallerThanSign | CXXTokenTypeOpeningBracket | CXXTokenTypeSemicolon | CXXTokenTypeEOF | CXXTokenTypeAssignment,
CXXTokenTypeOpeningParenthesis | CXXTokenTypeOpeningSquareParenthesis
))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse up to '<>{EOF'");
return FALSE;
}
evaluate_current_token:
switch(g_cxx.pToken->eType)
{
case CXXTokenTypeSmallerThanSign:
if(g_cxx.pToken->pPrev && cxxTokenTypeIsOneOf(g_cxx.pToken->pPrev,CXXTokenTypeIdentifier | CXXTokenTypeKeyword))
{
if(!cxxParserParseNextToken())
{
CXX_DEBUG_LEAVE_TEXT("Syntax error, but tolerate it at this level");
return TRUE;
}
CXX_DEBUG_PRINT("Got token '%s' of type 0x%02x",vStringValue(g_cxx.pToken->pszWord),g_cxx.pToken->eType);
if(cxxTokenTypeIs(g_cxx.pToken,CXXTokenTypeSmallerThanSign))
{
// assume it's an operator
CXX_DEBUG_PRINT("Treating < as shift-left operator");
} else {
CXX_DEBUG_PRINT("Increasing template level");
iTemplateLevel++;
if(cxxTokenTypeIsOneOf(g_cxx.pToken,CXXTokenTypeOpeningParenthesis | CXXTokenTypeOpeningSquareParenthesis))
{
CXX_DEBUG_PRINT("Condensing subchain");
// hmmm.. must condense as subchain
if(!cxxParserParseAndCondenseCurrentSubchain(CXXTokenTypeOpeningParenthesis | CXXTokenTypeOpeningSquareParenthesis,TRUE))
{
CXX_DEBUG_LEAVE_TEXT("Failed to condense current subchain");
return TRUE;
}
if(cxxTokenTypeIs(g_cxx.pToken,CXXTokenTypeEOF))
{
CXX_DEBUG_LEAVE_TEXT("Found EOF, syntax error but we tolerate it");
return TRUE;
}
} else {
// it's ok
CXX_DEBUG_PRINT("No need to condense subchain");
}
}
} else {
// assume it's an operator
CXX_DEBUG_PRINT("Treating < as less-than operator");
}
break;
case CXXTokenTypeGreaterThanSign:
if(iTemplateLevel == 0)
{
// Non-nested > : always a terminator
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
CXX_DEBUG_LEAVE_TEXT("Found end of template");
return TRUE;
}
// Nested > : is is a shift operator?
if(!cxxParserParseNextToken())
{
CXX_DEBUG_LEAVE_TEXT("Syntax error, but tolerate it at this level");
return TRUE;
}
CXX_DEBUG_PRINT("Got token '%s' of type 0x%02x",vStringValue(g_cxx.pToken->pszWord),g_cxx.pToken->eType);
if(cxxTokenTypeIs(g_cxx.pToken,CXXTokenTypeGreaterThanSign))
{
// assume it's an operator
CXX_DEBUG_PRINT("Treating > as shift-right operator");
} else {
CXX_DEBUG_PRINT("Decreasing template level");
iTemplateLevel--;
if(cxxTokenTypeIsOneOf(g_cxx.pToken,CXXTokenTypeOpeningParenthesis | CXXTokenTypeOpeningSquareParenthesis))
{
CXX_DEBUG_PRINT("Condensing subchain");
// hmmm.. must condense as subchain
if(!cxxParserParseAndCondenseCurrentSubchain(CXXTokenTypeOpeningParenthesis | CXXTokenTypeOpeningSquareParenthesis,TRUE))
{
CXX_DEBUG_LEAVE_TEXT("Failed to condense current subchain");
return TRUE;
}
if(cxxTokenTypeIs(g_cxx.pToken,CXXTokenTypeEOF))
{
CXX_DEBUG_LEAVE_TEXT("Found EOF, syntax error but we tolerate it");
return TRUE;
}
} else {
// it's ok
CXX_DEBUG_PRINT("No need to condense subchain");
}
}
break;
case CXXTokenTypeAssignment:
CXX_DEBUG_PRINT("Found assignment, trying to skip up to a 'notable' point");
// try to skip to the next > or , without stopping at < characters.
if(!cxxParserParseUpToOneOf(
CXXTokenTypeGreaterThanSign | CXXTokenTypeComma | CXXTokenTypeOpeningBracket | CXXTokenTypeSemicolon | CXXTokenTypeEOF
))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse up to '}EOF'");
return FALSE;
}
goto evaluate_current_token; // backward jump!
break;
case CXXTokenTypeEOF:
CXX_DEBUG_LEAVE_TEXT("Syntax error, but tolerate it at this level");
return TRUE;
break;
case CXXTokenTypeSemicolon:
cxxParserNewStatement();
CXX_DEBUG_LEAVE_TEXT("Broken template arguments, attempting to continue");
return TRUE;
break;
case CXXTokenTypeOpeningBracket:
CXX_DEBUG_PRINT("Found opening bracket: either syntax error or we screwed up parsing the template parameters (some kind of ugly C++11 syntax?), but we try to recover...");
// skip the whole bracketed part.
if(!cxxParserParseUpToOneOf(CXXTokenTypeClosingBracket | CXXTokenTypeEOF))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse up to '}EOF'");
return FALSE;
}
cxxParserNewStatement();
CXX_DEBUG_LEAVE_TEXT("Broken template arguments recovery complete");
return TRUE;
break;
default:
CXX_DEBUG_ASSERT(FALSE,"Should not end up here");
CXX_DEBUG_LEAVE_TEXT("Found unexpected token type 0x%02x",g_cxx.pToken->eType);
return FALSE;
break;
}
}
// never reached
CXX_DEBUG_LEAVE();
return TRUE;
}