TST (string dtype): clean-up assorted xfails #60354
Conversation
simonjayhawkins
changed the title
pandas/tests/reshape/test_pivot.py
Outdated
| def test_pivot_columns_is_none(self): | ||
| # GH#48293 | ||
| df = DataFrame({None: [1], "b": 2, "c": 3}) | ||
| df = DataFrame([[1, 2, 3]], columns=Index([None, "b", "c"], dtype="object")) |
There was a problem hiding this comment.
I think we do need to figure out how this should work with the new string data type; the larger discussion to that is being had in #60329 (comment)
| def test_get_dummies_with_str_dtype(any_string_dtype): | ||
| @pytest.mark.parametrize("use_string_repr", [True, False]) | ||
| def test_get_dummies_with_any_string_dtype( | ||
| request, any_string_dtype, any_string_dtype2, use_string_repr, using_infer_string |
There was a problem hiding this comment.
this is a new feature for 3.0.0
(that doesn't appear to be fully tested and consistent - hence additional parameterisation addded).
so using using_infer_string is probably not needed as we won't need backwards compatibility once inference is enable by default for 3.0.0
also any changes won't need backporting.
There was a problem hiding this comment.
I commented about this on the PR adding this test, see #59577 (comment), because I don't think it makes much sense to allow this. I see there is a PR following up on it to disallow dtype=str in get_dummies (#59786), but we should follow-up on it to ensure it gets into 3.0
(but so conclusion: I would leave this alone for this PR, and focus on fixing it in #59786)
| @@ -13,6 +13,10 @@ | |||
| from pandas._libs import missing as libmissing | |||
| from pandas._libs.sparse import IntIndex | |||
|
|
|||
| from pandas.core.dtypes.cast import ( | |||
There was a problem hiding this comment.
this commit has code changes but have pushed to this PR while we discuss
@jorisvandenbossche @WillAyd opening as draft and will add more to this PR over time (as I find solvable ones).