My experiment on having _Use_bitmap to always true or always false

I had the assumption that _Special is the case where we definitely have one of the standard character types. But apparently char_traits can be specialized for custom character types, so _Special may be true for something not behaving as a character.

It seems that we should fix _Special for all _Traits_meow algorithms. Perhaps in another PR?
(Also, it's possibly better to make _Special not a template parameter, which would shorten mangled names.)

The strategy may be like this:

template <class>
constexpr bool _Is_std_char_traits_specialization = false;
template <class _CharT>
constexpr bool _Is_std_char_traits_specialization<char_traits<_CharT>> = _Is_EcharT<_CharT>;

It seems that we should fix _Special for all _Traits_meow algorithms. Perhaps in another PR?

Yes, I think I can do this some time later (If you or someone else won't do it before me).
I'd prefer keeping the current PR draft, until it can be merged with main containing the changes of that PR.
This way basic_string won't become more broken for odd character types.

(Also, it's possibly better to make _Special not a template parameter, which would shorten mangled names.)

Good idea.

The strategy may be like this:

Also agree.

I'm marking this PR as ready again.

Turns out that basic_string::find_meow_of rejects unicorns here:

Since it is compile-time error already, vectorization doesn't make this case any more broken.

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

🧶 🐈‍⬛ 🎉