You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
dask/dataframe/tests/test_dataframe.py::test_apply: Failed: DID NOT WARN. No warnings of type (<class 'FutureWarning'>,) were emitted.
The list of emitted warnings is: [].
A couple of new pandas-related deprecations have popped up
FutureWarning: DataFrameGroupBy.dtypes is deprecated and will be removed in a future version. Check the dtypes on the base object instead
FutureWarning: The 'axis' keyword in DataFrame.rolling is deprecated and will be removed in a future version. Call the method without the axis keyword instead.
dask/dataframe/tests/test_groupby.py::test_groupby_aggregate_categorical_observed[8 failing variants]: AssertionError: DataFrame.iloc[:, 0] (column name="cat_2") are different
dask/dataframe/tests/test_groupby.py::test_groupby_aggregate_categorical_observed[4 failing variants]: AssertionError: DataFrame.iloc[:, 0] (column name="cat_2") are different
pandas-dev/pandas#52151 was fixed in Pandas, I'll recheck if we're still seeing errors with nightly.
Update: we're seeing different errors now - looks like a new deprecation. The groupby/categorical issue is fixed.
I'll look into the new failures.
dask/dataframe/tests/test_dataframe.py::test_apply: FutureWarning: the convert_dtype parameter is deprecated and will be removed in a future version. Do ``ser.astype(object).apply()`` instead if you want ``convert_dtype=False``.
dask/dataframe/tests/test_dataframe.py::test_apply_infer_columns: FutureWarning: the convert_dtype parameter is deprecated and will be removed in a future version. Do ``ser.astype(object).apply()`` instead if you want ``convert_dtype=False``.
Okay, @j-bennet's change for convert_dtype in Series.apply is is. We've now got a couple of new deprecations:
dask/dataframe/tests/test_accessors.py::test_dt_accessor: FutureWarning: The behavior of DatetimeProperties.to_pydatetime is deprecated, in a future version this will return a Series containing python datetime objects instead of an ndarray. To retain the old behavior, call `np.array` on the result
dask/dataframe/tests/test_dataframe.py::test_apply: FutureWarning: Returning a DataFrame from Series.apply when the supplied functionreturns a Series is deprecated and will be removed in a future version.
dask/dataframe/tests/test_dataframe.py::test_apply_infer_columns: FutureWarning: Returning a DataFrame from Series.apply when the supplied functionreturns a Series is deprecated and will be removed in a future version.
@phofl is this something you have bandwidth to fix?
Looks like there are some lingering upstream failures. Most (all?) are related to deprecations. @j-bennet@phofl is this something you have bandwidth for?
Activity
jrbourbeau commentedon Mar 20, 2023
Looks like some deprecations were added to
pandasmainover the weekend. cc @j-bennet for visibilityjrbourbeau commentedon Mar 20, 2023
#10094 should take care of the
axis=-related failures here. @j-bennet mentioned offline that she's working on theobserved=-related failuresj-bennet commentedon Mar 20, 2023
Draft PR for
observeddeprecations here: #10095.jrbourbeau commentedon Mar 22, 2023
A couple of new
pandas-related deprecations have popped up@charlesbluca @j-bennet do either of you have bandwidth to knock those out?
j-bennet commentedon Mar 22, 2023
@jrbourbeau I'll take care of those.
j-bennet commentedon Mar 24, 2023
Found the reason for these:
The reason is this PR:
I filed an issue in pandas to clarify the behavior:
groupby/minandobserved=Falseon categoricals between 2.0 and 2.1 pandas-dev/pandas#52151From the discussion, it looks like the change in behavior was not intentional.
j-bennet commentedon Mar 31, 2023
pandas-dev/pandas#52151 was fixed in Pandas, I'll recheck if we're still seeing errors with nightly.
j-bennet commentedon Mar 31, 2023
Update: we're seeing different errors now - looks like a new deprecation. The groupby/categorical issue is fixed.
I'll look into the new failures.
jrbourbeau commentedon Mar 31, 2023
Thanks for following up here @j-bennet 👍
jrbourbeau commentedon Apr 10, 2023
Okay, @j-bennet's change for
convert_dtypeinSeries.applyis is. We've now got a couple of new deprecations:@phofl is this something you have bandwidth to fix?
phofl commentedon Apr 24, 2023
There is a new groupby deprecation in pandas. I want to engage upstream before we adjust, since it is very very noisy
jrbourbeau commentedon May 1, 2023
Looks like there are some lingering upstream failures. Most (all?) are related to deprecations. @j-bennet @phofl is this something you have bandwidth for?
phofl commentedon May 1, 2023
Can take a look tomorrow. Most of them seem familiar
jrbourbeau commentedon May 1, 2023
Sounds good -- thanks @phofl. FWIW I went ahead and pushed up dask/distributed#7813 for one of the tests
j-bennet commentedon May 1, 2023
I can take a look at those.
j-bennet commentedon May 1, 2023
#10243
j-bennet commentedon May 5, 2023
Closed via #10256.
For now. :)
jrbourbeau commentedon May 5, 2023
Thanks @j-bennet @phofl for your work here