static Symbols tokenize(const QByteArray &input, int lineNum = 1, TokenizeMode mode = TokenizeCpp)
{
Symbols symbols;
const char *begin = input;
const char *data = begin;
while (*data) {
if (mode == TokenizeCpp) {
int column = 0;
const char *lexem = data;
int state = 0;
Token token = NOTOKEN;
for (;;) {
if (static_cast<signed char>(*data) < 0) {
++data;
continue;
}
int nextindex = keywords[state].next;
int next = 0;
if (*data == keywords[state].defchar)
next = keywords[state].defnext;
else if (!state || nextindex)
next = keyword_trans[nextindex][(int)*data];
if (!next)
break;
state = next;
token = keywords[state].token;
++data;
}
// suboptimal, is_ident_char should use a table
if (keywords[state].ident && is_ident_char(*data))
token = keywords[state].ident;
if (token == NOTOKEN) {
// an error really
++data;
continue;
}
++column;
if (token > SPECIAL_TREATMENT_MARK) {
switch (token) {
case QUOTE:
data = skipQuote(data);
token = STRING_LITERAL;
// concatenate multi-line strings for easier
// STRING_LITERAAL handling in moc
if (!Preprocessor::preprocessOnly
&& !symbols.isEmpty()
&& symbols.last().token == STRING_LITERAL) {
QByteArray newString = symbols.last().unquotedLexem();
newString += input.mid(lexem - begin + 1, data - lexem - 2);
newString.prepend('\"');
newString.append('\"');
symbols.last() = Symbol(symbols.last().lineNum,
STRING_LITERAL,
newString);
continue;
}
break;
case SINGLEQUOTE:
while (*data && (*data != '\''
|| (*(data-1)=='\\'
&& *(data-2)!='\\')))
++data;
if (*data)
++data;
token = CHARACTER_LITERAL;
break;
case LANGLE_SCOPE:
// split <:: into two tokens, < and ::
token = LANGLE;
data -= 2;
break;
case DIGIT:
while (is_digit_char(*data))
++data;
if (!*data || *data != '.') {
token = INTEGER_LITERAL;
if (data - lexem == 1 &&
(*data == 'x' || *data == 'X')
&& *lexem == '0') {
++data;
while (is_hex_char(*data))
++data;
}
break;
}
token = FLOATING_LITERAL;
++data;
// fall through
case FLOATING_LITERAL:
while (is_digit_char(*data))
++data;
if (*data == '+' || *data == '-')
++data;
if (*data == 'e' || *data == 'E') {
++data;
while (is_digit_char(*data))
++data;
}
if (*data == 'f' || *data == 'F'
|| *data == 'l' || *data == 'L')
++data;
break;
case HASH:
if (column == 1) {
mode = PreparePreprocessorStatement;
while (*data && (*data == ' ' || *data == '\t'))
++data;
if (is_ident_char(*data))
mode = TokenizePreprocessorStatement;
continue;
}
break;
case NEWLINE:
++lineNum;
continue;
case BACKSLASH:
{
const char *rewind = data;
while (*data && (*data == ' ' || *data == '\t'))
++data;
if (*data && *data == '\n') {
++data;
continue;
}
data = rewind;
} break;
case CHARACTER:
while (is_ident_char(*data))
++data;
token = IDENTIFIER;
break;
case C_COMMENT:
if (*data) {
if (*data == '\n')
++lineNum;
++data;
if (*data) {
if (*data == '\n')
++lineNum;
++data;
}
}
while (*data && (*(data-1) != '/' || *(data-2) != '*')) {
if (*data == '\n')
++lineNum;
++data;
}
token = WHITESPACE; // one comment, one whitespace
// fall through;
case WHITESPACE:
if (column == 1)
column = 0;
while (*data && (*data == ' ' || *data == '\t'))
++data;
if (Preprocessor::preprocessOnly) // tokenize whitespace
break;
continue;
case CPP_COMMENT:
while (*data && *data != '\n')
++data;
continue; // ignore safely, the newline is a separator
default:
continue; //ignore
}
}
#ifdef USE_LEXEM_STORE
if (!Preprocessor::preprocessOnly
&& token != IDENTIFIER
&& token != STRING_LITERAL
&& token != FLOATING_LITERAL
&& token != INTEGER_LITERAL)
symbols += Symbol(lineNum, token);
else
#endif
symbols += Symbol(lineNum, token, input, lexem-begin, data-lexem);
} else { // Preprocessor
const char *lexem = data;
int state = 0;
Token token = NOTOKEN;
if (mode == TokenizePreprocessorStatement) {
state = pp_keyword_trans[0][(int)'#'];
mode = TokenizePreprocessor;
}
for (;;) {
if (static_cast<signed char>(*data) < 0) {
++data;
continue;
}
int nextindex = pp_keywords[state].next;
int next = 0;
if (*data == pp_keywords[state].defchar)
next = pp_keywords[state].defnext;
else if (!state || nextindex)
next = pp_keyword_trans[nextindex][(int)*data];
if (!next)
break;
state = next;
token = pp_keywords[state].token;
++data;
}
// suboptimal, is_ident_char should use a table
if (pp_keywords[state].ident && is_ident_char(*data))
token = pp_keywords[state].ident;
switch (token) {
case NOTOKEN:
++data;
break;
case PP_IFDEF:
symbols += Symbol(lineNum, PP_IF);
symbols += Symbol(lineNum, PP_DEFINED);
continue;
case PP_IFNDEF:
symbols += Symbol(lineNum, PP_IF);
symbols += Symbol(lineNum, PP_NOT);
symbols += Symbol(lineNum, PP_DEFINED);
continue;
case PP_INCLUDE:
mode = TokenizeInclude;
break;
case PP_QUOTE:
data = skipQuote(data);
token = PP_STRING_LITERAL;
break;
case PP_SINGLEQUOTE:
while (*data && (*data != '\''
|| (*(data-1)=='\\'
&& *(data-2)!='\\')))
++data;
if (*data)
++data;
token = PP_CHARACTER_LITERAL;
break;
case PP_DIGIT:
while (is_digit_char(*data))
++data;
if (!*data || *data != '.') {
token = PP_INTEGER_LITERAL;
if (data - lexem == 1 &&
(*data == 'x' || *data == 'X')
&& *lexem == '0') {
++data;
while (is_hex_char(*data))
++data;
}
break;
}
token = PP_FLOATING_LITERAL;
++data;
// fall through
case PP_FLOATING_LITERAL:
while (is_digit_char(*data))
++data;
if (*data == '+' || *data == '-')
++data;
if (*data == 'e' || *data == 'E') {
++data;
while (is_digit_char(*data))
++data;
}
if (*data == 'f' || *data == 'F'
|| *data == 'l' || *data == 'L')
++data;
break;
case PP_CHARACTER:
if (mode == PreparePreprocessorStatement) {
// rewind entire token to begin
data = lexem;
mode = TokenizePreprocessorStatement;
continue;
}
while (is_ident_char(*data))
++data;
token = PP_IDENTIFIER;
break;
case PP_C_COMMENT:
if (*data) {
if (*data == '\n')
++lineNum;
++data;
if (*data) {
if (*data == '\n')
++lineNum;
++data;
}
}
while (*data && (*(data-1) != '/' || *(data-2) != '*')) {
if (*data == '\n')
++lineNum;
++data;
}
token = PP_WHITESPACE; // one comment, one whitespace
// fall through;
case PP_WHITESPACE:
while (*data && (*data == ' ' || *data == '\t'))
++data;
continue; // the preprocessor needs no whitespace
case PP_CPP_COMMENT:
while (*data && *data != '\n')
++data;
continue; // ignore safely, the newline is a separator
case PP_NEWLINE:
++lineNum;
mode = TokenizeCpp;
break;
case PP_BACKSLASH:
{
const char *rewind = data;
while (*data && (*data == ' ' || *data == '\t'))
++data;
if (*data && *data == '\n') {
++data;
continue;
}
data = rewind;
} break;
case PP_LANGLE:
if (mode != TokenizeInclude)
break;
token = PP_STRING_LITERAL;
while (*data && *data != '\n' && *(data-1) != '>')
++data;
break;
default:
break;
}
if (mode == PreparePreprocessorStatement)
continue;
#ifdef USE_LEXEM_STORE
if (token != PP_IDENTIFIER
&& token != PP_STRING_LITERAL
&& token != PP_FLOATING_LITERAL
&& token != PP_INTEGER_LITERAL)
symbols += Symbol(lineNum, token);
else
#endif
symbols += Symbol(lineNum, token, input, lexem-begin, data-lexem);
}
}
symbols += Symbol(); // eof symbol
return symbols;
}
Symbols Preprocessor::macroExpandIdentifier(Preprocessor *that, SymbolStack &symbols, int lineNum, QByteArray *macroName)
{
Symbol s = symbols.symbol();
// not a macro
if (s.token != PP_IDENTIFIER || !that->macros.contains(s) || symbols.dontReplaceSymbol(s.lexem())) {
Symbols syms;
syms += s;
syms.last().lineNum = lineNum;
return syms;
}
const Macro ¯o = that->macros.value(s);
*macroName = s.lexem();
Symbols expansion;
if (!macro.isFunction) {
expansion = macro.symbols;
} else {
bool haveSpace = false;
while (symbols.test(PP_WHITESPACE)) {
haveSpace = true;
}
if (!symbols.test(PP_LPAREN)) {
*macroName = QByteArray();
Symbols syms;
if (haveSpace)
syms += Symbol(lineNum, PP_WHITESPACE);
syms += s;
syms.last().lineNum = lineNum;
return syms;
}
QList<Symbols> arguments;
while (symbols.hasNext()) {
Symbols argument;
// strip leading space
while (symbols.test(PP_WHITESPACE)) {}
int nesting = 0;
bool vararg = macro.isVariadic && (arguments.size() == macro.arguments.size() - 1);
while (symbols.hasNext()) {
Token t = symbols.next();
if (t == PP_LPAREN) {
++nesting;
} else if (t == PP_RPAREN) {
--nesting;
if (nesting < 0)
break;
} else if (t == PP_COMMA && nesting == 0) {
if (!vararg)
break;
}
argument += symbols.symbol();
}
arguments += argument;
if (nesting < 0)
break;
}
// empty VA_ARGS
if (macro.isVariadic && arguments.size() == macro.arguments.size() - 1)
arguments += Symbols();
if (arguments.size() != macro.arguments.size() &&
// 0 argument macros are a bit special. They are ok if the
// argument is pure whitespace or empty
(macro.arguments.size() != 0 || arguments.size() != 1 || !arguments.at(0).isEmpty()))
that->error("Macro argument mismatch.");
// now replace the macro arguments with the expanded arguments
enum Mode {
Normal,
Hash,
HashHash
} mode = Normal;
for (int i = 0; i < macro.symbols.size(); ++i) {
const Symbol &s = macro.symbols.at(i);
if (s.token == HASH || s.token == PP_HASHHASH) {
mode = (s.token == HASH ? Hash : HashHash);
continue;
}
int index = macro.arguments.indexOf(s);
if (mode == Normal) {
if (index >= 0) {
// each argument undoergoes macro expansion if it's not used as part of a # or ##
if (i == macro.symbols.size() - 1 || macro.symbols.at(i + 1).token != PP_HASHHASH) {
Symbols arg = arguments.at(index);
int idx = 1;
expansion += macroExpand(that, arg, idx, lineNum, false);
} else {
expansion += arguments.at(index);
}
} else {
expansion += s;
}
} else if (mode == Hash) {
if (index < 0)
that->error("'#' is not followed by a macro parameter");
const Symbols &arg = arguments.at(index);
QByteArray stringified;
for (int i = 0; i < arg.size(); ++i) {
stringified += arg.at(i).lexem();
}
stringified.replace('"', "\\\"");
stringified.prepend('"');
stringified.append('"');
expansion += Symbol(lineNum, STRING_LITERAL, stringified);
} else if (mode == HashHash) {
if (s.token == WHITESPACE)
continue;
while (expansion.size() && expansion.last().token == PP_WHITESPACE)
expansion.pop_back();
Symbol next = s;
if (index >= 0) {
const Symbols &arg = arguments.at(index);
if (arg.size() == 0) {
mode = Normal;
continue;
}
next = arg.at(0);
}
if (!expansion.isEmpty() && expansion.last().token == s.token) {
Symbol last = expansion.last();
expansion.pop_back();
if (last.token == STRING_LITERAL || s.token == STRING_LITERAL)
that->error("Can't concatenate non identifier tokens");
QByteArray lexem = last.lexem() + next.lexem();
expansion += Symbol(lineNum, last.token, lexem);
} else {
expansion += next;
}
if (index >= 0) {
const Symbols &arg = arguments.at(index);
for (int i = 1; i < arg.size(); ++i)
expansion += arg.at(i);
}
}
mode = Normal;
}
if (mode != Normal)
that->error("'#' or '##' found at the end of a macro argument");
}
return expansion;
}