test_simulation.py fails with the latest version of numpy

[mishinma]

@yeganer @unoebauer @wkerzendorf
In PR #471 the following (a few of them actually) asset statement was added to test_simulation.py:

npt.assert_allclose(
            t_rad, simulation_compare_data['test1/t_rad'], atol=0.0)

It is calculated like this: absolute(a - b) <= (atol + rtol * absolute(b))
The problem here is that t_rad is a Quantity while simulation_compare_data['test1/t_rad'] is just a numpy array. And if somehow works with numpy v 1.9.0, with v 1.10 it fails. An error is raised in the __array_prepare__ method that is overridden in Astropy:

try:
                if _can_have_arbitrary_unit(maybe_arbitrary_arg):
                    converters = [None, None]
                else:
                    raise UnitsError("Can only apply '{0}' function to "
                                     "dimensionless quantities when other "
                                     "argument is not a quantity (unless the "
                                     "latter is all zero/infinity/nan)"
>                                    .format(function.__name__))
E                                    UnitsError: Can only apply 'subtract' function to dimensionless quantities when other argument is not a quantity (unless the latter is all zero/infinity/nan)

[mishinma]

Actually, it would be nice to compare Quantities with a special util function that would validate the units of its arguments. One version of such a function is implemented in Astropy's module tests/helpers.py: assert_quantity_allclose

[yeganer]

I think the right approach should be to use astropy's assert_quantity_allclose. One would have to test if that solution is compatible with different versions of numpy and astropy.

[mishinma]

But units still should be added to the tests. assert_quantity_allclose still raises an error in this case as well, although with meaningful message:

(Pdb) assert_quantity_allclose(t_rad, simulation_compare_data['test1/t_rad'], atol=0.0)
*** UnitsError: Units for 'desired' () and 'actual' (K) are not convertible
(Pdb) npt.assert_allclose(t_rad, simulation_compare_data['test1/t_rad'], atol=0.0)
*** UnitsError: Can only apply 'subtract' function to dimensionless quantities when other argument is not a quantity (unless the latter is all zero/infinity/nan)

[yeganer]

So we have two options. Strip the result of its unit or add the unit to the
compare data. In that case we should do the second option.

Misha [email protected] schrieb am So., 7. Feb. 2016 21:08:

But units still should be added to the tests. assert_quantity_allclose
still raises an error in this case as well, although with more clear
message:

(Pdb) assert_quantity_allclose(t_rad, simulation_compare_data['test1/t_rad'], atol=0.0)
*** UnitsError: Units for 'desired' () and 'actual' (K) are not convertible
(Pdb) npt.assert_allclose(t_rad, simulation_compare_data['test1/t_rad'], atol=0.0)
*** UnitsError: Can only apply 'subtract' function to dimensionless quantities when other argument is not a quantity (unless the latter is all zero/infinity/nan)

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

[yeganer]

Thank you for pointing out numpy's assert_quantity_allclose. I think we should use it whenever possible in our tests.

@wkerzendorf @unoebauer What do you think?

[mishinma]

I made PR #479 to use assert_quantity_allclose in test_simulation.py

[mishinma]

Should the issue be closed now?

[unoebauer]

Yes - closing it now.