hana::experimental::type_name - MSVC support?

[014972304505347]

With the recent fork provided by Microsoft for Boost Hana / MSVC compiler compatibility, Hana now builds with the MSVC compiler. However, the experimental method hana::experimental::type_name has a hardcoded switch in its helper "type_name_impl2()" that only supports Clang:

// Note: We substract the null terminator from the string sizes below.
template <typename T>
constexpr cstring type_name_impl2() {

#if defined(__clang__)
	constexpr char const* pretty_function = __PRETTY_FUNCTION__;
	constexpr std::size_t total_size = sizeof(__PRETTY_FUNCTION__) - 1;
	constexpr std::size_t prefix_size = sizeof("boost::hana::experimental::detail::cstring boost::hana::experimental::detail::type_name_impl2() [T = ") - 1;
	constexpr std::size_t suffix_size = sizeof("]") - 1;
#else
        #error "No support for this compiler."
#endif

There exist pseudo-equivalent MSVC predefined-macros (see this doc) to __PRETTY_FUNCTION__ : __FUNCSIG__, __FUNCDNAME__, __func__, and __FUNCTION__. See this and this for reference. It is also worth looking at BOOST_CURRENT_FUNCTION. These could possibly make type_name compatible with MSVC by replacing the __PRETTY_FUNCTION__ macro.

I tried adding this switch in the type_name.hpp header file, which allows MSVC to hit the switch and use type_name:

	#if defined (__clang__)
	...
	#elif defined(_MSC_VER)
		constexpr char const* pretty_function = __FUNCSIG__;
		constexpr std::size_t total_size = sizeof(__FUNCSIG__) - 1;
		constexpr std::size_t prefix_size = sizeof("boost::hana::experimental::detail::cstring boost::hana::experimental::detail::type_name_impl2() [T = ") - 1;
		constexpr std::size_t suffix_size = sizeof("]") - 1;
	#else
            #error "No support for this compiler."
	#endif

type_name then yields typename strings like this:

        int -> "e_impl2<int>(void"
        "myStruct" user-defined struct -> "e_impl2<struct myStruct>(void"
        std::string -> "e_impl2<class std::basic_string<char,struct std::char_traits<char>,class std::allocator<char> >>(void"

Using BOOST_CURRENT_FUNCTION gives the same result.
This sort of output is messy, but can be interpreted and processed to give a better result.

__FUNCDNAME__, __func__, and __FUNCTION__ all cause compilation errors when used like above, though I suspect they would provide better output "out-of-the-box" if they can be worked out.

Would you consider adding something like the above so that MSVC can have type_name support? I may be interested in writing some code myself for solving the problem, though I think the fix is not complicated.