tests: fix comparison on numpy 1.25

[alexfikl]

Before comparing two ndarray([DOFArray, DOFArray]) would just return ndarray([False, False]), but it doesn't work anymore. Couldn't find an explanation in the release notes as to why that changed.

I changed it to use actx.np.equal, but I'm not sure if this was supposed to test __eq__?

[inducer]

Thanks for working on this. The following:

import numpy as np


class MyContainer:
    def __eq__(self, other):
        return self

    def __bool__(self):
        raise ValueError("not defined")


a = np.zeros(2, dtype=object)
a[0] = MyContainer()
a[1] = MyContainer()

b = np.zeros(2, dtype=object)
b[0] = MyContainer()
b[1] = MyContainer()

a == b

says

/home/andreas/tmp/numpy-obj-array-equal.py:20: DeprecationWarning: elementwise comparison failed; this will raise an error in the future.
  a == b

so it seems that this may have been an intentional change.

[inducer]

I think this may be the bit about == in https://numpy.org/doc/stable/release/1.25.0-notes.html#expired-deprecations.

[inducer]

This means that == is no longer really a thing for array containers involving numpy object arrays, which I guess is just as well. I think your change is more or less the only option here.

[alexfikl]

I think this may be the bit about == in numpy.org/doc/stable/release/1.25.0-notes.html#expired-deprecations.

Oh yeah, I totally missed that. Thanks for hunting it down!