parsed string attribute contains \0

Hi,

for my (spare time) project I have to implement e.g. the following BNF rule:

integer ::= digit { [ underline ] digit }

The separator underline rises up on several production rules, e.g. hex 
values (16'A_FF_E),bit pattern (2"0111_0100) or others 
(10*42.47_11*e-6_6). After some (negative) experience with polluting the 
grammar rules with lots of semantic actions I decided to store the 
underline pruned literal strings into the AST, tag them with the context 
from BNF, and evaluate them later (e.g. parse with a 2nd pass explicit 
for int, double, hex, etc with the knowledge of the tags added before). 
IIRC this is the recommanded way (avoid SA).

At this point I run into the problem with the 'flattended'/'joined' 
string - there is a \0 at the point of underline occurence.

---8<----
#include <iostream>
#include <boost/spirit/home/x3.hpp>

int main()
{
    namespace x3 = boost::spirit::x3;
    using x3::char_;

    auto const integer =  x3::rule<struct _, std::string>{} =
        x3::lexeme[ +char_("0-9") >> *(char_("0-9") | "_") ];

    for(std::string const str: {
            "4711", "123_456"
    }) {
      auto iter = str.begin(), end = str.end();

      std::string attr;
      bool r = x3::phrase_parse(iter, end, integer, x3::space, attr);

      std::cout << "parse '" << str << "': ";
      if (r && iter == end) {
        std::cout << "succeeded: '" << attr << "'\n";
        if(attr.find('\0') != std::string::npos) {
            std::cout << "found '\\0'\n";
        }
      } else {
        std::cout << "*** failed ***\n";
      }
    }

    return 0;
}

--->8----

Depending on the capabilities of the terminal used you can see, e.g.:

parse '4711': succeeded: '4711'
parse '123_456': succeeded: '123456'
found '\0'

or even only '123  (then the \0 strikes).
The rule written here looks odd, I know. Normally I would write
x3::lexeme[ +char_("0-9") >> -(*char_("0-9_") ];
(which still isn't sufficent I guess) but this contains the unwanted 
underline. This ignores the literal in attribute but gives the problems 
mentioned above.

How can I fix this? How to write the rule that trailing underlines fails 
to parse?

Thanks,
Olaf



------------------------------------------------------------------------------
Announcing the Oxford Dictionaries API! The API offers world-renowned
dictionary content that is easy and intuitive to access. Sign up for an
account today to start using our lexical data to power your apps and
projects. Get started today and enter our developer competition.
http://sdm.link/oxford