I want to tokenize my own extension of SQL syntax. This involves recognizing an escaped double quote inside a double quoted string. E.g. in MySQL these two string tokens are equivalent: """" (the second double quote acts as an escape character) and '"'. I have tried different things but I am stuck at how to replace a token's value.
#include <boost/spirit/include/lex_lexertl.hpp>
namespace lex = boost::spirit::lex;
template <typename Lexer>
struct sql_tokens : lex::lexer<Lexer>
{
sql_tokens()
{
string_quote_double = "\\\""; // '"'
this->self("INITIAL")
= string_quote_double [ lex::_state = "STRING_DOUBLE" ] // how to also ignore + ctx.more()?
| ...
;
this->self("STRING_DOUBLE")
= lex::token_def<>("[^\\\"]*") // action: ignore + ctx.more()
| lex::token_def<>("\\\"\\\"") // how to set token value to '"' ?
| lex::token_def<>("\\\"") [ lex::_state = "INITIAL" ]
;
}
lex::token_def<> string_quote_double, ...;
};
So how to set the token's value to " when "" has been found?
Apart from that I have also the following question: I can write a functor for a semantic action to call ctx.more() and ignore the token at the same time (thus combining "low level" tokens into a "high level" string token). But how to elegantly combine this with lex::_state = ".." ?
EDITED in response to comment, see below "UPDATE""
I suggest not trying to solve that in the lexer. Let the lexer yield raw strings:
template <typename Lexer>
struct mylexer_t : lex::lexer<Lexer>
{
mylexer_t()
{
string_quote_double = "\\\"([^\"]|\\\"\\\")*\\\"";
this->self("INITIAL")
= string_quote_double
| lex::token_def<>("[ \t\r\n]") [ lex::_pass = lex::pass_flags::pass_ignore ]
;
}
lex::token_def<std::string> string_quote_double;
};
NOTE That exposing a token attribute like that, requires a modified token typedef:
typedef lex::lexertl::token<char const*, boost::mpl::vector<char, std::string> > token_type;
typedef lex::lexertl::actor_lexer<token_type> lexer_type;
Postprocess in the parser:
template <typename Iterator> struct mygrammar_t
: public qi::grammar<Iterator, std::vector<std::string>()>
{
typedef mygrammar_t<Iterator> This;
template <typename TokenDef>
mygrammar_t(TokenDef const& tok) : mygrammar_t::base_type(start)
{
using namespace qi;
string_quote_double %= tok.string_quote_double [ undoublequote ];
start = *string_quote_double;
BOOST_SPIRIT_DEBUG_NODES((start)(string_quote_double));
}
private:
qi::rule<Iterator, std::vector<std::string>()> start;
qi::rule<Iterator, std::string()> string_quote_double;
};
As you can see, undoubleqoute can be any Phoenix actor that satisfies the criteria for a Spirit semantic action. A brain-dead example implementation would be:
static bool undoublequote(std::string& val)
{
auto outidx = 0;
for(auto in = val.begin(); in!=val.end(); ++in) {
switch(*in) {
case '"':
if (++in == val.end()) { // eat the escape
// end of input reached
val.resize(outidx); // resize to effective chars
return true;
}
// fall through
default:
val[outidx++] = *in; // append the character
}
}
return false; // not ended with double quote as expected
}
But I suggest you write a "proper" de-escaper (as I'm pretty sure MySql will allow \t, \r, \u001e or even more archaic stuff as well).
I have some more complete samples in old answers here:
In fact, as you indicated, it is fairly easy to integrate the attribute value normalization into the lexer itself:
template <typename Lexer>
struct mylexer_t : lex::lexer<Lexer>
{
struct undoublequote_lex_type {
template <typename, typename, typename, typename> struct result { typedef void type; };
template <typename It, typename IdType, typename pass_flag, typename Ctx>
void operator()(It& f, It& l, pass_flag& pass, IdType& id, Ctx& ctx) const {
std::string raw(f,l);
if (undoublequote(raw))
ctx.set_value(raw);
else
pass = lex::pass_flags::pass_fail;
}
} undoublequote_lex;
mylexer_t()
{
string_quote_double = "\\\"([^\"]|\\\"\\\")*\\\"";
const static undoublequote_lex_type undoublequote_lex;
this->self("INITIAL")
= string_quote_double [ undoublequote_lex ]
| lex::token_def<>("[ \t\r\n]") [ lex::_pass = lex::pass_flags::pass_ignore ]
;
}
lex::token_def<std::string> string_quote_double;
};
This reuses the same undoublequote function shown above, but wraps it in Deferred Callable Object (or "polymorphic functor") undoublequote_lex_type that satisfies the criteria for a Lexer Semantic Action.
Here is a fully working proof of concept:
//#include <boost/config/warning_disable.hpp>
//#define BOOST_SPIRIT_DEBUG_PRINT_SOME 80
//#define BOOST_SPIRIT_DEBUG // before including Spirit
#include <boost/spirit/include/lex_lexertl.hpp>
#include <boost/spirit/include/qi.hpp>
#include <fstream>
#ifdef MEMORY_MAPPED
# include <boost/iostreams/device/mapped_file.hpp>
#endif
//#include <boost/spirit/include/lex_generate_static_lexertl.hpp>
namespace /*anon*/
{
namespace phx=boost::phoenix;
namespace qi =boost::spirit::qi;
namespace lex=boost::spirit::lex;
template <typename Lexer>
struct mylexer_t : lex::lexer<Lexer>
{
mylexer_t()
{
string_quote_double = "\\\"([^\"]|\\\"\\\")*\\\"";
this->self("INITIAL")
= string_quote_double
| lex::token_def<>("[ \t\r\n]") [ lex::_pass = lex::pass_flags::pass_ignore ]
;
}
lex::token_def<std::string> string_quote_double;
};
static bool undoublequote(std::string& val)
{
auto outidx = 0;
for(auto in = val.begin(); in!=val.end(); ++in) {
switch(*in) {
case '"':
if (++in == val.end()) { // eat the escape
// end of input reached
val.resize(outidx); // resize to effective chars
return true;
}
// fall through
default:
val[outidx++] = *in; // append the character
}
}
return false; // not ended with double quote as expected
}
template <typename Iterator> struct mygrammar_t
: public qi::grammar<Iterator, std::vector<std::string>()>
{
typedef mygrammar_t<Iterator> This;
template <typename TokenDef>
mygrammar_t(TokenDef const& tok) : mygrammar_t::base_type(start)
{
using namespace qi;
string_quote_double %= tok.string_quote_double [ undoublequote ];
start = *string_quote_double;
BOOST_SPIRIT_DEBUG_NODES((start)(string_quote_double));
}
private:
qi::rule<Iterator, std::vector<std::string>()> start;
qi::rule<Iterator, std::string()> string_quote_double;
};
}
std::vector<std::string> do_test_parse(const std::string& v)
{
char const *first = &v[0];
char const *last = first+v.size();
typedef lex::lexertl::token<char const*, boost::mpl::vector<char, std::string> > token_type;
typedef lex::lexertl::actor_lexer<token_type> lexer_type;
typedef mylexer_t<lexer_type>::iterator_type iterator_type;
const static mylexer_t<lexer_type> mylexer;
const static mygrammar_t<iterator_type> parser(mylexer);
auto iter = mylexer.begin(first, last);
auto end = mylexer.end();
std::vector<std::string> data;
bool r = qi::parse(iter, end, parser, data);
r = r && (iter == end);
if (!r)
std::cerr << "parsing (" << iter->state() << ") failed at: '" << std::string(first, last) << "'\n";
return data;
}
int main(int argc, const char *argv[])
{
for (auto&& s : do_test_parse( "\"bla\"\"blo\""))
std::cout << s << std::endl;
}