This isn't an answer, but:

>>> np.issubdtype(np.double, np.bool_)
False

also related

>>> np.issubclass_(np.bool, np.generic)
False
>>> np.issubclass_(np.bool_, np.generic)
True

>>> inspect.getmro(np.bool)
(<type 'bool'>, <type 'int'>, <type 'object'>)
>>> inspect.getmro(np.bool_)
(<type 'numpy.bool_'>, <type 'numpy.generic'>, <type 'object'>)

Thanks for the clarification!

Feel free to close the issue, if you think this is the intended behavior.

In [39]: issubdtype(np.bool_, np.int8)
Out[39]: False

In [40]: issubdtype(np.int8, np.bool_)
Out[40]: False

np.bool_ seems to be a type all by itself in the type hierarchy.

Yes, it is. For some reason the picture is hidden on this page:
http://docs.scipy.org/doc/numpy/reference/arrays.scalars.html

I don't know what's going on with the original report, though: dtype(bool)
and dtype(bool_) certainly ought to return the same things...
On Mar 23, 2015 7:49 PM, "Charles Harris" [email protected] wrote:

In [39]: issubdtype(np.bool_, np.int8)
Out[39]: False

In [40]: issubdtype(np.int8, np.bool_)
Out[40]: False

np.bool_ seems to be a type all by itself in the type hierarchy.

—
Reply to this email directly or view it on GitHub
#5711 (comment).

There is this

n [14]: issubdtype(np.object, np.bool)
Out[14]: True

In [15]: issubdtype(np.bool, np.object)
Out[15]: True

Python bool was new in python version 2.2.1. It may be that the current behavior is an accident.

Note that the code in np.issubdtype contains some weird stuff:

    if issubclass_(arg2, generic):
        return issubclass(dtype(arg1).type, arg2)
    mro = dtype(arg2).type.mro()
    if len(mro) > 1:
        val = mro[1]
    else:
        val = mro[0]
    return issubclass(dtype(arg1).type, val)

For example

>>> np.issubdtype(np.float32, np.dtype(float))
True

The commit message in 325aad6 is not very helpful on why this was made so.
One possible motivation could be to make issubdtype(foo, int) and issubdtype(foo, float) do the same thing as if comparing to np.integer_ and np.floating_.

I think #3218 is relevant to this issue.

Duplicate of #2334

Deprecated in #9505