Mixed usage of the sequence parser and the expectation parser

Question

I want to parse the following string

    "***: @a_-091 , *** 1"

to a struct defined as

    using type = boost::fusion::vector<char, int, double>;

When the parser

    *qi::omit[qi::char_ - '@'] >> '@' >> qi::char_ >> '_' >> qi::int_ >> *qi::omit[qi::char_ - qi::digit] >> qi::double_

is used, the result is OK. However, the result is totally different with the following parser

    *qi::omit[qi::char_ - '@'] >> '@' >> qi::char_ >> '_' >> qi::int_ >> *qi::omit[qi::char_ - qi::digit] > qi::double_

The following is the sample code.

    #include <vector>
    #include <sstream>
    #include <iostream>
    #include <boost/spirit/include/qi.hpp>
    #include <boost/spirit/include/qi_match.hpp>
    #include <boost/fusion/include/io.hpp>

    using type = boost::fusion::vector<char, int, double>;

    int main() {
        std::istringstream istr{
            "***: @a_-091 , *** 1"
        };

        std::vector<type> data;

        namespace qi = boost::spirit::qi;
        istr >> std::noskipws >> qi::match(*(
            *qi::omit[qi::char_ - '@'] >> '@' >> qi::char_ >> '_' >> qi::int_ >> *qi::omit[qi::char_ - qi::digit] > qi::double_
            ), data);
        // istr >> std::noskipws >> qi::match(*(
            *qi::omit[qi::char_ - '@'] >> '@' >> qi::char_ >> '_' >> qi::int_ >> *qi::omit[qi::char_ - qi::digit] >> qi::double_
            ), data);

        for (size_t i = 0; i != data.size(); ++i) {
            std::cerr << data[i] << "\n";
        }

        return 0;
    }

PS: The problem has been solved by the link proved by cv_and_he in the comment. It is caused by the mixed usage of ">>" (the sequence parser) and ">" (the expectation parser).

Two suggestions: one question per question, and code must be in the question. And describe the code, describe what it's doing (or not doing) and what you expected instead. (And comments go into the comment section, not in the question.) — Mat, May 30 '15 at 10:29
You talk about two things that do different stuff. Yet you show only one piece of code and describe neither. — Mat, May 30 '15 at 10:40
@Mat I don't quite understand. As you suggested, I remove a problem asked. Now, the problem is the different behavior betwwen ">>" and ">". — cqdjyy01234, May 30 '15 at 10:44
Why would you expect the same thing when using two different parsers? What behavior exactly don't you understand? What results are you getting and what were you expecting? — Mat, May 30 '15 at 10:49
@Mat Yes, they are different parsers. When ">>" is used, the double is parsed succesfully. Thus, ">" should successed as well. I think the only difference between the two parsers is that ">>" allows failure, but ">" does not (throw an exception instead). — cqdjyy01234, May 30 '15 at 10:53
I believe your problem is similar to the one described [here](http://boost.2283326.n4.nabble.com/Unexpected-behavior-using-versus-tp4662087p4662134.html). — llonesmiz, May 30 '15 at 11:06
[This](http://coliru.stacked-crooked.com/a/705d921eaf86a031) also works, but it's awful. — llonesmiz, May 30 '15 at 11:24

score 0 · Answer 1 · answered May 30 '15 at 13:28

0

As a quick hint, perhaps you can rewrite using >>.

To still have the expectation points you can perhaps write

a >> b > c

like

a >> b >> (c > eps)

I will test this later

answered May 30 '15 at 13:28

sehe

374,641
47
450
633

Mixed usage of the sequence parser and the expectation parser

1 Answers1