Do we generalise from `bool` to `bool | np.bool`, etc.?

[cmp0xff]

In the stubs we usually have Sequence[bool], Sequence[int], etc. In some cases they can actually be Sequence[bool | np.bool], Sequence[int | np.integer], etc., and have been implemented in a few places.

Do we do it across the whole project?

Originally posted by @MarcoGorelli in #1505

[Dr-Irv]

In the stubs we usually have Sequence[bool], Sequence[int], etc. In some cases they can actually be Sequence[bool | np.bool], Sequence[int | np.integer], etc., and have been implemented in a few places.

Do we do it across the whole project?

I'm open to doing this, although we probably should create a type like pandas_int, pandas_bool, etc., that is the union of int and np.integer, bool and np.bool respectively. Then it becomes easier to maintain.

[cmp0xff]

Thank you for supporting the idea. However I think pandas_ can be misleading, since the prefix could also indicate pandas-specific objects. I am having the newly introduced categories BuiltinIntDtypeArg, PandasIntDtypeArg, NumpyIntDtypeArg etc. in my mind.

In my opinion, something like AnyInt: TypeAlias = int | np.integer, AnyBool: TypeAlias = bool | np.bool looks better.

[Dr-Irv]

In my opinion, something like AnyInt: TypeAlias = int | np.integer, AnyBool: TypeAlias = bool | np.bool looks better.

Sure, that makes sense.