DOC: TEST.rst: add example with pytest.mark.parametrize
#24667
Conversation
The "Creating many similar tests" section of TEST.rst did not mention the use of pytest.mark.parametrize, which is a very common and convenient way to run several similar tests. I also noticed the following about the example: - All instructions of `test_single` and `test_double` were identical aside from the dtype. This did not seem consistent with the intent of the example. - The example showed use of `np.matrix`, which is no longer recommended. - The example used a function `imply`, which was not defined. This commit removes the "Creating many similar tests" section and expands the parametric tests section with an example that uses pytest.mark.parametrize. It also mentions NumPy's assert_equal, assert_allclose, and assert_less functions and makes other minor adjustments. [skip actions] [skip azp] [skip cirrus]
| between a result array and a reference, | ||
| - :func:`numpy.testing.assert_allclose` for testing near elementwise equality | ||
| between a result array and a reference (i.e. with specified relative and | ||
| absolute tolerances), and |
There was a problem hiding this comment.
Maybe we should advertise the _nulp versions more?
There was a problem hiding this comment.
Maybe. But actually, does NumPy 2.0 have plans to pare down the testing functions to those that are recommended?
There was a problem hiding this comment.
There is no such plan. The usage of older testing functions is too high to do that I'm afraid. The functions that were deprecated/removed typically had a usage count of a handful, in some cases up to ~30 or so, in SciPy, scikit-learn and Pandas. I just did a quick count for assert_array_almost_equal and it's 2838 in SciPy alone - so we can't touch it really.
This topic can be better addressed in the docs though. This PR is a great step in the right direction.
There was a problem hiding this comment.
OK. Then how about listing assert_array_equal as "not recommended" following the plan suggested in #8457 (comment)?
doc/TESTS.rst
Outdated
| - :func:`numpy.testing.assert_array_less` for testing (strict) elementwise | ||
| ordering between a result array and a reference. | ||
|
|
||
| Note that these assertion functions only compare the numerical vales of the |
There was a problem hiding this comment.
| Note that these assertion functions only compare the numerical vales of the | |
| Note that these assertion functions only compare the numerical values of the |
There was a problem hiding this comment.
We now have strict=True on some (assert_equal?) to check a bit more (mainly the shape and probably the dtype).
There was a problem hiding this comment.
Yes, assert_array_equal. But it's only that one, I think. How bout we add strict to the others and then I can mention it here? Update: see gh-24680.
There was a problem hiding this comment.
That PR got merged, care to update the text too?
There was a problem hiding this comment.
Sure, I'm happy to update this after adding strict to assert_equal.
There was a problem hiding this comment.
Ah missed that there was still another step, merging!
|
This seems like a good improvement, just one comment but if you don't have time to update this is still a significant improvement over the status quo and should be merged. |
[skip ci] Co-authored-by: Sebastian Berg <sebastian@sipsolutions.net>
|
Thanks @mdhaber! |
|
Squash merge (second commit is a one character touch-up without real commit message) would be great, and I'd be happy to come back and mention Comments on #8457 (comment) would be appreciated! |
The "Creating many similar tests" section of
TEST.rstdid not mention the use ofpytest.mark.parametrize, which is a common and convenient way to run several similar tests. I also noticed the following about the example:test_singleandtest_doublewere identical aside from the dtype. This did not seem consistent with the intent of the example.np.matrix, which is no longer recommended.imply, which was not defined.This commit removes the "Creating many similar tests" section and expands the parametric tests section with an example that uses
pytest.mark.parametrize. It also mentions NumPy'sassert_equal,assert_allclose, andassert_lessfunctions and makes other minor adjustments.Closes gh-23283