Simon Hawkins

can confirm that works as intended for regular Index and data columns. There is [another issue](https://github.com/pandas-dev/pandas/issues/21136) where a level of a MultiIndex appears to be not sorted. Maybe related. see...

> Looks like this error occurs with pandas 1.4.0 and 1.4.2 but not 1.3.5 first bad commit: [9bf09adcdab1a5babaed9f3315f2572cebf9f1f2] REF: fast_xs avoid object dtype (#43203) cc @jbrockmendel

> We'll have to re-write something; not yet clear to me where the appropriate fix is. @jbrockmendel i've opened a PR https://github.com/pandas-dev/pandas/pull/48246 that effectively restores the 1.3.5 code path, but...

Thanks @aberres for the report > Just reporting it here if this is an unintended change. Maybe I missed it, but I did not see this mentioned in the changelog....

changing milestone to 1.3.5

Looking at this issue from purely a backport fix perspective, I doesn't appear that we have any solutions here (for 1.3.x). Changing the behavior of apply would not be suitable...

removing issue from 1.3.x milestone. Any changes to address the apply issue would not be backported.

closed by https://github.com/MacPython/pandas-wheels/pull/188

I'll milestone this as 1.5 from https://github.com/pandas-dev/pandas/issues/46512#issue-1181203886 > The behavior for the inplace=False case has changed from 1.4.1 to main since https://github.com/pandas-dev/pandas/pull/45168 allows other na values in the StringArray Constructor....

> If you want `pandas` to treat `pd.NA` and `np.nan` differently, your best bet would be `object` type data (or perhaps nullable float data depending on how #32265 goes) adding...