There are a few other uses. The documentation simply uses the bool documentation for bool_, if we are trying to discourage use we should add a note for that. bool_ is also defined in _type_aliases.py, and used in numpy/_core/src/multiarray/scalartypes.c.src. My understanding is that we want to remove all internal usage, is that correct?

EDIT: or are we reserving bool_ for Python bool?

My understanding is that we want to remove all internal usage, is that correct?

I'd say yes. But that doesn't need to be backported to 2.0.x, and isn't urgent so I'd suggest doing that separately.

EDIT: or are we reserving bool_ for Python bool?

No, np.bool_ is an alias to np.bool and should disappear in the far future. We should not have any aliases for Python builtin types.

@charris I think all redundant usages of bool_ are already removed. bool_ is still present as an alias and can be used as np.bool_ or np.dtype("bool_"). The former requires "bool_" to be defined in _type_aliases.py. Once we drop bool_, both can be removed.

In scalartypes.c.src I agree, "|O:bool_" can be changed to "|O:bool". And in docs we can have another note about the bool -> bool_ change. I will fix both.