BUG: datetime64 can't be compared with pyarrow-backed variant

[MarcoGorelli]

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

df = pd.DataFrame({'a': pd.to_datetime(pd.Series(['2020-01-01']))})
df['b'] = df['a'].convert_dtypes(dtype_backend='pyarrow')
print(df['b'] < df['a'])

Issue Description

Should the above be supported? it currently raises

Expected Behavior

i think

0    False
dtype: bool[pyarrow]

Installed Versions

INSTALLED VERSIONS

commit : 0691c5c
python : 3.12.6
python-bits : 64
OS : Linux
OS-release : 5.15.167.4-microsoft-standard-WSL2
Version : #1 SMP Tue Nov 5 00:21:55 UTC 2024
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : C.UTF-8
LOCALE : C.UTF-8

pandas : 2.2.3
numpy : 2.2.3
pytz : 2025.1
dateutil : 2.9.0.post0
pip : None
Cython : None
sphinx : None
IPython : 8.31.0
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : None
blosc : None
bottleneck : None
dataframe-api-compat : None
fastparquet : None
fsspec : 2025.2.0
html5lib : None
hypothesis : 6.125.2
gcsfs : None
jinja2 : 3.1.5
lxml.etree : None
matplotlib : None
numba : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
psycopg2 : None
pymysql : None
pyarrow : 19.0.0
pyreadstat : None
pytest : 8.3.4
python-calamine : None
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlsxwriter : None
zstandard : None
tzdata : 2025.1
qtpy : None
pyqt5 : None

[abhigyan631]

I think this fix can be used

import pandas as pd

df = pd.DataFrame({'a': pd.to_datetime(pd.Series(['2020-01-01']))})

df['b'] = df['a'].convert_dtypes(dtype_backend='pyarrow')

df['b'] = df['b'].astype('datetime64[ns]') # Convert 'b' back to datetime64[ns]

print(df['b'] < df['a'])

We need to convert the PyArrow-backed column (df['b']) back to a NumPy-backed datetime64[ns] type before performing the comparison to get the expected result.

[rhshadrach]

I think the discussion in #60639 implies that yes, this should be supported.

[AshleySonny]

take