Why aren't there numpy.testing.assert_array_greater, assert_array_less_equal, assert_array_true, etc?

Question

In numpy.testing, there's assert_array_less and assert_array_equal, but there isn't an assert_array_less_equal function or even an assert_array_greater. I have two questions:

Is there a reason these functions are missing, but assert_array_less is not?
I've written my own versions of these missing functions by using numpy.testing.util.assert_array_compare, e.g.:

def assert_array_greater(aa, bb):
  assert_array_compare(np.greater, aa, bb)

Is this safe? I.e. is there a reason why assert_array_compare is hidden away in numpy.testing.util, rather than living in numpy.testing?

Forgive my paranoia; it just seems weird that these functions don't exist, to the extent that I fear it's for some good reason that I shouldn't be working around.

hpaulj · Accepted Answer · 2017-08-31 19:27:52Z

7

np.testing is a module that collects tests and tools used by various of the numpy unittesting files. So it's designed more for internal use, than for end user use. So the simple answer would be that those extra tests aren't needed.

But looking at the source code for one of those functions:

def assert_array_less(x, y, err_msg='', verbose=True):
    assert_array_compare(operator.__lt__, x, y, err_msg=err_msg,
                         verbose=verbose,
                         header='Arrays are not less-ordered',
                         equal_inf=False)

It looks like it would be easy to write a variation the uses one of the other operator methods.

The 'root' of np.testing is numpy/testing/__init__.py, which is a short file. Looks like its main task is from .utils import *. This is typical subpackage organization. The __init__ collects the necessary imports, but often doesn't have significant code of its own.

edited Aug 31, 2017 at 19:27

answered Aug 31, 2017 at 19:21

hpaulj

233k14 gold badges260 silver badges392 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

SuperElectric Over a year ago

Thanks! Do you have any guesses as to why they use operator.__lt__ rather than numpy.less? (As in, is there a good reason why I should as well?)

hpaulj Over a year ago

That file only uses operator twice. I think operator.__lt__(a,b) delegates to a.__lt__(b), the same as a<b. np.less is ufunc version, which at some compiled level probably ends up calling the same code, but details like that can be hard to trace.

Collectives™ on Stack Overflow

Why aren't there numpy.testing.assert_array_greater, assert_array_less_equal, assert_array_true, etc?

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related