I also open the same question on stackoverflow. For better formatting of code, please refer to https://stackoverflow.com/questions/58353782/attributes-of-sequence-and-list... https://stackoverflow.com/questions/58353782/attributes-of-sequence-and-list... I want to parse something like "{xxxx} {xxxx}" which is separated by eol into a *vector> *: ({xxxx},{xxxx}) so that "{" and "}" stays with internal characters together. Note there are only two elements on the outer vector. My code is: #define BOOST_SPIRIT_UNICODE #include <iostream> #include #include<string> #include<vector> using namespace std; namespace sw=boost::spirit::standard_wide; namespace qi= boost::spirit::qi; using boost::spirit::standard_wide::char_; int main() { wstring s = L"{\"id\":23,\"text\":\"sf\nsf\"}\n{\"id\":23,\"text\":\"sfsf\"}"; qi::rule>(), sw::blank_type> ru; ru = (qi::char_(L"{") >> *(char_-char_(L"}")) >> char_(L"}")) % qi::eol; vector> result; qi::phrase_parse(s.begin(), s.end(), ru, sw::blank, result); for (auto& v : result) { //cout << "Size of string: " << v.size() << endl; for (auto& s : v) { wcout << s; }; cout << endl; }; std::cout << "Size of result"< for the outer vector. Then consider the rule: *ru = (qi::char_(L"{") >> *(char_-char_(L"}")) >> char_(L"}")) % qi::eol;* According to the documentation, *(char_-char_(L"}")) should be vector . And because *a: A, b: vector --> (a >> b): vector *, then I think that *(qi::char_(L"{") >> *(char_-char_(L"}")) >> char_(L"}"))* should be *vector*. This is contracdicted to the result. Where I'm wrong? -- Sent from: http://boost.2283326.n4.nabble.com/Boost-Users-f2553780.html