BUG: df1.merge(df2, on='ds', how='outer') not convert column ds to the highest resolution datetime type of the two dfs

[seanslma]

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

t1 = '2023-09-01'
t2 = '2023-09-01 01:00:00'
t3 = '2023-09-01 01:30:00'
ds1 = pd.date_range(t1, t2, freq='30T')
ds2 = pd.date_range(t1, t3, freq='30T')
df1 = pd.DataFrame({
    'ds': ds1.astype('datetime64[us]'),
    'y1': range(len(ds1)),
})
df2 = pd.DataFrame({
    'ds': ds2.astype('datetime64[ns]'),
    'y2': range(len(ds2)),
})
df3 = df1.merge(df2, on=['ds'], how='outer') #will only convert df2 `ds` type to df1 `ds` type
df3.dtypes

Issue Description

The column type was not converted from the lower resolution (dtype('<M8[us]')) to the higher resolution (dtype('<M8[ns]')).

It only converted the type in df2 to the type in df2 for the on column.

File ~\conda-envs\dev\lib\site-packages\pandas\core\arrays\_mixins.py:398, in NDArrayBackedExtensionArray._putmask(self, mask, value)
    383 """
    384 Analogue to np.putmask(self, mask, value)
    385 
   (...)
    394     If value cannot be cast to self.dtype.
    395 """
    396 value = self._validate_setitem_value(value)
--> 398 np.putmask(self._ndarray, mask, value)

File <__array_function__ internals>:180, in putmask(*args, **kwargs)

TypeError: Cannot cast array data from dtype('<M8[ns]') to dtype('<M8[us]') according to the rule 'safe'

Expected Behavior

I expect df1.merge(df2, on=['ds'], how='outer') and df2.merge(df1, on=['ds'], how='outer') both should work in this case.

Installed Versions

INSTALLED VERSIONS

commit : ba1cccd
python : 3.9.15.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.19044
machine : AMD64
processor : Intel64 Family 6 Model 58 Stepping 0, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : English_Australia.1252

pandas : 2.1.0
numpy : 1.23.3
pytz : 2022.1
dateutil : 2.8.2
setuptools : 65.5.0
pip : 22.3.1
Cython : 0.29.30
pytest : 7.1.2
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.9.0
html5lib : None
pymysql : None
psycopg2 : 2.9.3
jinja2 : 3.0.3
IPython : 8.4.0
pandas_datareader : None
bs4 : 4.11.1
bottleneck : 1.3.5
dataframe-api-compat: None
fastparquet : None
fsspec : 2022.5.0
gcsfs : None
matplotlib : 3.5.2
numba : 0.53.0
numexpr : 2.7.3
odfpy : None
openpyxl : 3.0.9
pandas_gbq : None
pyarrow : 12.0.0
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.8.1
sqlalchemy : 1.4.46
tables : None
tabulate : None
xarray : 2022.3.0
xlrd : None
zstandard : None
tzdata : 2022.1
qtpy : None
pyqt5 : None

[paulreece]

I have confirmed locally that this bug occurs on the main development branch.

[rhshadrach]

@mroeschke @MarcoGorelli any thoughts on what should happen here?

[mroeschke]

Yeah as mentioned in the OP this should work by casting to the higher resolution. Concat has as similar problem #53641 (comment)

[lithomas1]

I don't think its a regression, I get the same error on 2.0.3, too.

Should have a PR up for 2.1.4, anyways, though.