DOC: docs for issubdtype doesn't explain why complex64 is not a subdtype of complex or complex_

[david-cortes]

Describe the issue:

Calling issubdtype on a complex64 does not produce the correct output.

Reproduce the code example:

import numpy as np
np.issubdtype(np.array([], dtype="F").dtype, complex)
np.issubdtype(np.array([], dtype="F").dtype, np.complex_)

False
False

Error message:

No response

NumPy/Python version information:

1.21 and current master (1046087)

[eric-wieser]

Are you looking for np.issubdtype(np.array([], dtype="F").dtype, np.complexfloating)?

[eric-wieser]

If you pass a non-numpy type into np.issubtype, it will just call np.dtype on it, and np.dtype(complex) == np.complex_ == np.complex128.

Until #15773, this would give a DeprecationWarning, but then do what you expect.

[david-cortes]

Yes, that would do it, although it would be nice to have that kind of thing as a warning in the docs for dtypes under the part where it says "Built-in Python types" (https://numpy.org/doc/stable/reference/arrays.dtypes.html).

[eric-wieser]

I think this probably belongs as a warning in the docs for issubdtype. I've rephrased the issue as missing documentation, but feel free to contribute a fix yourself.

[tensionhead]

I also stumbled upon this, it is very irritating:

np.issubdtype(np.dtype('complex64'), np.complex)
>>>DeprecationWarning: `np.complex` is a deprecated alias for the builtin `complex`. To silence this warning, use `complex` by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use `np.complex128` here.
>>> False

If I then follow the deprecation warning I still get:

np.issubdtype(np.dtype('complex64'), complex)
>>> False

Same for floats:

np.issubdtype(np.dtype('float32'), float)
>>> False

However:

np.issubdtype(np.dtype('float64'), float)
>>> True

But if I just want to know if I have a complex dtype, the answer by @eric-wieser with np.complexfloating got me what I need.

[eric-wieser]

For completeness, I think:

np.issubdtype(np.dtype('complex128'), complex)
>>> True