Boost Xpressive ignores capture placeholders

[Evgeny Yashin <johny5.coder@…>]

This code:

#include "boost/xpressive/xpressive.hpp"

int main()
{
	using namespace boost::xpressive;

	cregex scheme = (s1 = +_w) >> "://";
	cregex host_ipv6 = (s2 = ('[' >> +(xdigit | ':') >> ']'));
	cregex uri_re = scheme >> host_ipv6;

	cmatch m;
	if( regex_match( "http://[fe::]", m, uri_re ) )
	{
		//	[first, last) pair of iterators, with implicit operator string()
		std::string protocol = m[1];
		protocol = protocol;
	}

    return 0;
}

It runs and it matches but protocol will contain empty string. Visual Studio 2015, in debug/release, 32/64bit.

[Eric Niebler]

After the match, m has no sub_matches. It does, however, have nested results that correspond to the nested regexes in uri_re. Please read [^1] for more information.

[^1]: http://www.boost.org/doc/libs/1_64_0/doc/html/xpressive/user_s_guide.html#boost_xpressive.user_s_guide.grammars_and_nested_matches.nested_regexes_and_sub_match_scoping

[Evgeny Yashin <johny5.coder@…>]

Thank you, overlooked indeed. A bit unexpected though.

Any way to flatten the matching results? Breaking into set of regexps, here, should be considered as simplifying final expression for human reading, not declaring any recursive grammars.

[Evgeny Yashin <johny5.coder@…>]

Addition, I see that iterating through tree gives the flatten view over nested results, I am mostly about keeping original index based syntax, where s1= .. match will be accessed via m[1] index later on.

[Eric Niebler]

It can't. There could be a 0^th submatch in two nested regexes. They would stomp each other. The nested_results keeps them separate.