I think a separate traits class for extracting the member type from a pointer to member is the best solution here, lets say it's called "remove_member_pointer" (finding a good name is the main stumbling block really), then you could use:
function_traits<typename remove_member_pointer<member_function_type>::type>::arity etc
O yes, that would be a good solution too. I'm interested in it.
BTW the problem with extending function_traits is that it operates on function-types, and not on pointers or references to functions, but there is no such type as a "member function", only a pointer to one, so we would end up with different behaviors for member functions and regular ones.
Well, in my view the behaviour would be equal, but they would require just some additional specializations (plus the trick to avoid add_pointer).
Actually I don't exactly see why function_traits should not operate on pointers. The function_traits_helper operates on pointers and not on types. So what actually happens is that the user removes the pointer, and it is then added in the function_traits, and specialized on in the details. No problem of course, but you could specialize on function-types AND function-pointers separately to avoid the requirement of using remove_pointer.
Nod.
But, again, a remove_member_pointer, if it can be implemented, would do. Is it already implemented or drafted somewhere? Because we cannot wait to use it ;-)
It's here: https://svn.boost.org/trac/boost/browser/sandbox/type_traits/boost/type_trai... John.