Jeff Flinn wrote:
I haven't had a chance to look at the library yet, but I'd have thought there would be a corresponding path concept/class. The std::string in the interface sounds overly restrictive. Just as the string algorithm is not restricted to std::(w)string.
Hmm ... something along the lines of this? struct ptree_traits { typedef std::string key_type; typedef std::string path_type; typedef char separator_type; static const separator_type default_separator = '.'; std::pair<key_type, path_type> split_path(const path_type &p, separator_type s) { path_type::size_type offset = p.find(s); if(offset == path_type::npos) { // Do whatever is to do in this situation. } else { return std::make_pair(p.substr(0, offset), p.substr(offset+1)); } } }; struct radixtree_traits { typedef int key_type; typedef int path_type; typedef int separator_type; static const separator_type default_separator = 10; std::pair<key_type, path_type> split_path(const path_type &p, separator_type s) { return std::make_pair(p % s, p / s); } }; Other path types (e.g. a vector of ints) would just ignore the separator. Sounds good to me. Not as complicated as I expected either. Sebastian Redl