BUG: DataFrame.to_parquet() throws an exception when the dtype of an integer column is changed to 'category' using .astype() (pyarrow backend)

[s184012]

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

from io import StringIO
import pandas as pd

input = StringIO(
    """,ID,Team,Puzzler
0,ID_1,10,1
1,ID_2,11,0
2,ID_3,10,0
3,ID_4,11,1""")

df = pd.read_csv(input, dtype_backend='pyarrow', index_col=0)
df = df.astype({'Team': 'category'})
df.to_parquet('example.parquet')

Issue Description

I get the error message

pyarrow.lib.ArrowInvalid: ('Could not convert <pyarrow.Int64Scalar: 10> with type pyarrow.lib.Int64Scalar: did not recognize Python value type when inferring an Arrow data type', 'Conversion failed for column Team with type category')

when trying to call .to_parquet on the dataframe. I only get the error when setting the type with .astype() function. If i instead load the dataframe as df = pd.read_csv(input, dtype_backend='pyarrow', index_col=0, dtype={'Team': 'category') it works fine.

Expected Behavior

To create a parquet file named 'example.parquet' without error.

Installed Versions

/home/simon/Code/bugreport/.venv/lib/python3.10/site-packages/_distutils_hack/init.py:33: UserWarning: Setuptools is replacing distutils.
warnings.warn("Setuptools is replacing distutils.")

INSTALLED VERSIONS

commit : aa1f96b
python : 3.10.6.final.0
python-bits : 64
OS : Linux
OS-release : 5.19.0-40-generic
Version : #41~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Fri Mar 31 16:00:14 UTC 2
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_GB.UTF-8
LOCALE : en_GB.UTF-8

pandas : 2.1.0.dev0+627.gaa1f96bc6b
numpy : 1.24.3
pytz : 2023.3
dateutil : 2.8.2
setuptools : 67.7.2
pip : 23.1.1
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : None
pandas_datareader: None
bs4 : None
bottleneck : None
brotli : None
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : None
numba : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 11.0.0
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : None
snappy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
zstandard : None
tzdata : 2023.3
qtpy : None
pyqt5 : None

[mroeschke]

Here's a simpler reproducer; this looks to be an issue with pyarrow not being able to convert Categoricals with ArrowDtype

In [9]: pa.Table.from_pandas(pd.DataFrame(pd.Categorical(np.array([1], dtype=object))))
Out[9]: 
pyarrow.Table
0: dictionary<values=int64, indices=int8, ordered=0>
----
0: [  -- dictionary:
[1]  -- indices:
[0]]

In [10]: pa.Table.from_pandas(pd.DataFrame(pd.Categorical(pd.Series([1], dtype="int64[pyarrow]"))))
ArrowInvalid: ('Could not convert <pyarrow.Int64Scalar: 1> with type pyarrow.lib.Int64Scalar: did not recognize Python value type when inferring an Arrow data type', 'Conversion failed for column 0 with type category')

[mroeschke]

I think this will need to be fully addressed in pyarrow so closing. Feel free to report this in Arrow

[phofl]

This should be fixed on the next arrow release, we already have an issue about this