BUG: DataFrame.apply with axis=1 and EA dtype

[jbrockmendel]

• closes REGR: Row series are broken after applying to_datetime() #36785
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[simonjayhawkins]

Thanks @jbrockmendel for the PR.

The code sample in #36785 still fails, but now with ValueError: Length of passed values is 2, index implies 4.

but happy to close #36785 if we feel that the underlying issue is solved.

[simonjayhawkins]

can you add a release note to 1.1.5 in the regression fixes section

[simonjayhawkins]

but happy to close #36785 if we feel that the underlying issue is solved.

looks like there maybe still an issue.

[jbrockmendel]

ill take another look this afternoon

[simonjayhawkins]

looks like there maybe still an issue.

thanks @jbrockmendel. code sample in OP is now passing.

before the edit, the existing test_invert_array was failing on ci which had more than 1 row. so that case is covered.

The sample in the OP was still failing and the dataframe had more than 1 column. Since only the one test was failing, it appears that this case is not covered by existing tests.

maybe we should an additional test case to test_apply_axis1_with_ea, maybe...

        df = DataFrame(date_range("2020", periods=3, tz="UTC")).T
        result = df.apply(lambda x: x, axis=1)
        tm.assert_frame_equal(result, df)

[jreback]

ok, obviously we want to sync these up in a followon

[simonjayhawkins]

@meeseeksdev backport 1.1.x

[jbrockmendel]

@simonjayhawkins im falling behind on GH notifications. do i need to make a PR adding more tests or something here?

[simonjayhawkins]

The PR was merged so I assume not.