Broken output when a multidimensional NumPy array fails to be approximate something

[Wrzlprmft]

Consider the following example comparing two multidimensional arrays using approx:

from pytest import approx
import numpy as np

def test_something():
	shape = (5,4)
	X = np.ones(shape)
	Y = np.ones(shape)*2
	assert X==approx(Y)

The error thrown by PyTest is:

E    assert array([[1., 1., 1., 1.],\n       [1., 1., 1., 1.],\n       [1., 1., 1., 1.],\n       [1., 1., 1., 1.],\n       [1., 1., 1., 1.]]) == <[TypeError("only size-1 arrays can be converted to Python scalars") raised in repr()] ApproxNumpy object at 0x7f63bda42a58>
E     +  where <[TypeError("only size-1 arrays can be converted to Python scalars") raised in repr()] ApproxNumpy object at 0x7f63bda42a58> = approx(array([[2., 2., 2., 2.],\n       [2., 2., 2., 2.],\n       [2., 2., 2., 2.],\n       [2., 2., 2., 2.],\n       [2., 2., 2., 2.]]))

This is confusing since at first glance it seems that the problem is that I tried to convert some array in an inappropriate manner, while actually the assertion failed.

My PyTest version is 3.3.2.

[pytestbot]

GitMate.io thinks possibly related issues are #1164 (test_something fails), #213 (Issue with numpy arrays and pytest.mark.parametrize), #2172 (Capturing output causes test to fail.), #2314 (capsys fails to capture the output by stdout), and #10 (assert has strange behavior with numpy arrays).

[kalekundert]

Thanks for catching this. I'm pretty sure I know what the problem is. I'll try to make a fix this weekend.

[nicoddemus]

Thanks @kalekundert, that would be great!

[kalekundert]

Actually, this was fixed by #3615. Update to the most recent version of pytest and you should be fine.