Pass on skipped v2 tests

[ioanaif]

No description provided.

[codecov]

Codecov Report

Merging #1445 (72dee09) into main (edfce38) will increase coverage by 1.24%.
The diff coverage is 59.04%.

Impacted Files	Coverage Δ
src/awkward/_v2/_connect/cuda/__init__.py	0.00% <0.00%> (ø)
src/awkward/_v2/_connect/jax/__init__.py	0.00% <0.00%> (ø)
...c/awkward/_v2/_connect/rdataframe/to_rdataframe.py	0.00% <0.00%> (ø)
src/awkward/_v2/contents/recordarray.py	82.47% <ø> (+0.97%)	⬆️
src/awkward/_v2/operations/convert/ak_from_cupy.py	50.00% <0.00%> (+23.33%)	⬆️
src/awkward/_v2/operations/convert/ak_from_iter.py	93.75% <ø> (ø)
src/awkward/_v2/operations/convert/ak_from_jax.py	50.00% <0.00%> (-25.00%)	⬇️
src/awkward/_v2/operations/convert/ak_to_cupy.py	33.33% <0.00%> (+23.95%)	⬆️
src/awkward/_v2/operations/convert/ak_to_jax.py	33.33% <0.00%> (-41.67%)	⬇️
src/awkward/_v2/operations/describe/ak_backend.py	10.00% <0.00%> (-2.50%)	⬇️
... and 105 more

[jpivarski]

Thanks for this! There must be a lot fewer skips now, though I'd have to check There are no unqualified skips now, apart from the "Missing check for overridden __repr__"! (There will always be skipif, for situations like "ROOT was compiled without C++17 support".) This is awesome!

% fgrep pytest.mark.skip tests/v2/test_*.py
tests/v2/test_0021-emptyarray.py:@pytest.mark.skip(
tests/v2/test_0025-record-array.py:# @pytest.mark.skip(reason="FIXME: recordtype requires field argument")
tests/v2/test_0032-replace-dressedtype.py:@pytest.mark.skip(reason="Missing check for overridden __repr__")
tests/v2/test_0032-replace-dressedtype.py:@pytest.mark.skip(reason="Missing check for overridden __repr__")
tests/v2/test_0032-replace-dressedtype.py:@pytest.mark.skip(reason="Missing check for overridden __repr__")
tests/v2/test_0032-replace-dressedtype.py:@pytest.mark.skip(reason="Missing check for overridden __repr__")
tests/v2/test_0049-distinguish-record-and-recordarray-behaviors.py:@pytest.mark.skip(reason="Missing check for overridden __repr__")
tests/v2/test_0080-flatpandas-multiindex-rows-and-columns.py:@pytest.mark.skipif(
tests/v2/test_0401-add-categorical-type-for-arrow-dictionary.py:@pytest.mark.skip(reason="Missing check for overridden __repr__")
tests/v2/test_0652-tests-of-complex-numbers.py:@pytest.mark.skip(reason="Remember to implement sorting for complex numbers.")
tests/v2/test_0914-types-and-forms.py:@pytest.mark.skipif(
tests/v2/test_0914-types-and-forms.py:@pytest.mark.skipif(
tests/v2/test_0959-_getitem_array-implementation.py:@pytest.mark.skipif(
tests/v2/test_1125-to-arrow-from-arrow.py:@pytest.mark.skipif(
tests/v2/test_1294-to-and-from_parquet.py:@pytest.mark.skipif(
tests/v2/test_1300-awkward-to-cpp-converter-with-cling.py:@pytest.mark.skipif(not cpp17, reason="ROOT was compiled without C++17 support")
tests/v2/test_1300-awkward-to-cpp-converter-with-cling.py:@pytest.mark.skipif(not cpp17, reason="ROOT was compiled without C++17 support")
tests/v2/test_1300-awkward-to-cpp-converter-with-cling.py:@pytest.mark.skipif(not cpp17, reason="ROOT was compiled without C++17 support")
tests/v2/test_1300-awkward-to-cpp-converter-with-cling.py:@pytest.mark.skipif(not cpp17, reason="ROOT was compiled without C++17 support")

Below, I have just one suggested change (click on it and it will be committed). I had a question about a removed test, but I've answered it: it's okay to remove. So just take the suggested change and then turn on auto-merge (squash).

[jpivarski]

Yeah, we did just remove record-sorting (and argsorting was never there in v1), so I'm surprised that there's an explicit test for it. I somewhat remember this one—Lindsey contributed it, it was something that came up in Coffea. But I can't see how we ever claimed to sort electron objects, instead of argsorting on electron pt and then applying the integer array to the electron objects.

@lgray, does this test look familiar? Is the record-sorting deprecation that we're talking about in #1451 impact Coffea?

[jpivarski]

I did some digging; the first v1 version of this test came in PR #812, which was a bug-fix for issue #811, which @lgray opened, having discovered the bug through Coffea. That's where the references to electrons come from.

In the original issue, only leptons.pt is passed to argsort: numbers, not the particle objects. The original issue was about leptons being a union array, calling .pt on it and getting a "union of numbers and numbers", calling argsort on that and getting a "union of integers and integers", and then not being able to use a union array as a slice. That would be a case of a missing simplify_uniontype, which I think you've addressed in another test here.

Yes, your fix of src/awkward/_v2/contents/unionarray.py is putting a simplify_uniontype in _getitem_field (and _getitem_fields), which would fix issue #811 by making leptons.pt return a non-union array of numbers.

So I think it's fine to delete this test; it's being tested elsewhere.