[CI][Python] Nigthly dask tests are failing due to type and conversion differences #49083
Description
Describe the bug, including details regarding any error messages, version, and platform.
The nightly job for dask (test-conda-python-3.11-dask-upstream_devel) has been failing during the last ~10 days. This matches the release of pandas 3.0.0.
There are 10 test failures. From what I can see they aren't all related to the same cause, some examples:
if check_dtype and str(adt) != str(bdt):
> raise AssertionError(f"a and b have different dtypes: (a: {adt}, b: {bdt})")
E AssertionError: a and b have different dtypes: (a: str, b: object)
or
pyarrow/error.pxi:155: in pyarrow.lib.pyarrow_internal_check_status
???
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
> ???
E pyarrow.lib.ArrowNotImplementedError: Function 'quantile' has no kernel matching input types (large_string)
or
if any(blk.dtype.kind in "mM" for blk in self._mgr.blocks):
msg = (
"DataFrame contains columns with dtype datetime64 "
"or timedelta64, which are not supported for cov."
)
> raise TypeError(msg)
E TypeError: DataFrame contains columns with dtype datetime64 or timedelta64, which are not supported for cov.
but all of them seem to be related to differences of conversion. All this seems related to the pandas 3 version. I think we've seen some of those issues with pandas tests integrations before.
Component(s)
Continuous Integration, Python