Fix .loc of dataframe with nullable boolean dtype

[m-rossi]

I use the Nullable Boolean data type with Dask. It worked fine since the recent version 2021.11.0. Let me show an example:

import dask.dataframe as dd
import pandas as pd

s1 = pd.Series([0, 1, 2])
s2 = pd.Series([True, False, pd.NA], dtype="boolean")

ddf1 = dd.from_pandas(s1, npartitions=1)
ddf2 = dd.from_pandas(s2, npartitions=1)

While this indexing

ddf1[ddf2]

works fine, this indexing

ddf1.loc[ddf2]

fails with the following error

KeyError: 'Cannot index with non-boolean dask Series. Try passing computed values instead (e.g. ``ddf.loc[iindexer.compute()]``)'

This PR adds the pd.BooleanDtype() to list of boolean-dask Series and adds a test which fails before my code changes.

• Tests added / passed
• Passes pre-commit run --all-files

[GPUtester]

Can one of the admins verify this patch?

[quasiben]

add to allowlist

[jsignell]

Thanks for pushing this change so quickly. It looks like this doesn't quite work with older versions of pandas. I think you can use is_bool_dtype like we do in

dask/dask/dataframe/core.py

Line 2568 in 11f52ed

if is_bool_dtype(data._meta):

[jsignell]

Thanks for sticking with this @m-rossi!