assert_tree_match and np.testing ignores array masks

[braingram]

While reviewing #1803 a somewhat related issue was found. See #1803 (comment) for the initial comment.

asdf._tests._helpers.assert_tree_match discards array masks during the comparison.
This is caused by 2 related issues (both covered here) and this line:

asdf/asdf/_tests/_helpers.py

Line 55 in 33e7095

np.testing.assert_array_equal(old.__array__(), new.__array__())

The call to __array__ drops the mask:

import numpy as np
arr = np.arange(5)
m0 = np.ma.masked_array(arr, False)
a2 = m0.__array__()
print(f"{type(a2)}: {a2}")

returns

<class 'numpy.ndarray'>: [0 1 2 3 4]

So the mask never makes it to assert_array_equal. Furthermore:

import numpy as np
arr = np.arange(5)
m0 = np.ma.masked_array(arr, True)
m1 = np.ma.masked_array(arr, False)
np.testing.assert_array_equal(m0, m11)

Does not fail (as it appears assert_array_equal (and assert_equal) ignore masks.

As our test suite contains several masked arrays it seems worthwhile to:

• update assert_tree_match to handle the masks (or remove that test function)
• replace all uses of np.testing that might handle a masked array (I'm not sure if there's a drop-in replacement or some argument we can provide).