repeat_parser and std::string attribute combination

[aclex]

I'm trying to create a rule for a string consisting of specified substrings and I seem to fail with it. Here's a simplified example:

#include <iostream>

#include <boost/parser/parser.hpp>

using namespace std;
namespace bp = boost::parser;

namespace
{
	namespace rule
	{
		constexpr boost::parser::rule<struct field_tag, string> field = "field";
		constexpr auto field_def {*bp::string("AB")};
		BOOST_PARSER_DEFINE_RULES(field);
	}
}

int main()
{
	string ret;

	string input{"ABABABAB"};
	const auto result{bp::parse(input, rule::field, ret)};

	cout << "Parsed successfully: " << result << endl;
	cout << "Result: \"" << ret << "\"" << endl;

	return 0;
}

It fails to compile for me both with g++-15 and clang++-20 (I tried both develop and master branches). If I get it right, the repetition (from *) should result in sequence of std::string attributes, which I expect to collapse to one std::string attribute of the rule itself.

Taking a look at the attribute generation rules I see there's no such exception for std::string directly, some similar exception concerns only char:

If C is exactly std::string, and T is either char or char32_t, the combination yields a std::string.

So I just wanted to ask if there's no such exception (for std::string sequence combined to single std::string) intentionally and I just miss something or that's something expected to work? Could you also please suggest a workaround, if possible, for such cases, when the grammar defines the form substrings being repeated? My real-world example is in fact parsing IPv6 address to a string.

[tzlaine]

Yes, the collapsing rules are all intentional -- including the one that says that std::strings should not combine by default. I think you want to use merge[].

https://www.boost.org/doc/libs/latest/doc/html/boost_parser/tutorial.html

[aclex]

Thank you for clarification! Indeed, merge looks like what I wanted, will try to use it.