ENH: Allow select_dtypes("category") to identify ArrowDtype(pa.dictionary))

[nachomaiz]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd

# prepare test data
test_data = pd.DataFrame({"categorical": ["a", "b", "c", "a", "b", "c"]}, dtype="category")

# confirm correct categorical dtype
print(test_data.select_dtypes("category").shape)  # (6, 1)
print(test_data.select_dtypes(pd.CategoricalDtype).shape)  # (6, 1) [warns]
print(test_data["categorical"].dtype)  # category

# parquet round trip with pyarrow backend
test_data.to_parquet("test.parquet")
loaded_data = pd.read_parquet("test.parquet", dtype_backend="pyarrow")

# check categorical dtype
print(loaded_data.select_dtypes("category").shape)  # (6, 0) [doesn't recognize categorical column]
print(loaded_data.select_dtypes(pd.CategoricalDtype).shape)  # (6, 0) [warns and doesn't recognize categorical column]
print(loaded_data["categorical"].dtype)  # dictionary<values=string, indices=int32, ordered=0>[pyarrow]

# this check works
from pandas.core.dtypes.dtypes import CategoricalDtypeType
print(loaded_data.select_dtypes(CategoricalDtypeType).shape)  # (6, 1) [recognizes categorical column]

Issue Description

The read_parquet function with dtype_backend="pyarrow" parameter doesn't seem to be assigning CategoricalDtype correctly to categorical columns.

Instead, they remain as Arrow dictionaries, and are not searchable with df.select_dtypes("category").

Expected Behavior

I expected the round trip of df.to_parquet and pd.read_parquet to maintain the categorical dtype, and for the loaded dtype to be selected when using df.select_dtypes("category").

Installed Versions

INSTALLED VERSIONS

commit : d9cdd2e
python : 3.12.1.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.19045
machine : AMD64
processor : Intel64 Family 6 Model 140 Stepping 1, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : English_United States.1252

pandas : 2.2.2
numpy : 1.26.4
pytz : 2024.1
dateutil : 2.8.2
setuptools : 69.0.3
pip : 24.0
Cython : None
pytest : 8.0.0
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.3
IPython : 8.21.0
pandas_datareader : None
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : None
bottleneck : None
dataframe-api-compat : None
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : 3.8.3
numba : None
numexpr : None
odfpy : None
openpyxl : 3.1.2
pandas_gbq : None
pyarrow : 15.0.2
pyreadstat : 1.2.7
python-calamine : None
pyxlsb : None
s3fs : None
scipy : 1.12.0
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : 2.0.1
zstandard : None
tzdata : 2024.1
qtpy : None
pyqt5 : None

[mroeschke]

Thanks for the issue, but I think this will be the expected behavior to keep the Arrow dictionary type, but I think once there is more support for this type (e.g. #56670) select_dtypes("category") should select ArrowDtype(pa.dictionary(...))

[nachomaiz]

@mroeschke Thank you for your response and for focusing the issue a bit more.

And you're right of course, as long as pandas is able to treat pandas dtypes and Arrow dtypes interchangeably in the frontend, it shouldn't necessarily matter that they are different in the backend.

Very excited for more integration with Arrow! Thanks!