SeriesGroupBy.agg does not accept Aggregates when passed as kwargs

[flying-sheep]

It uses pandas’ validation function, which doesn’t recognize dask’s Aggregate class:

dask/dask/dataframe/groupby.py

Line 2279 in 99036c4

columns, arg = validate_func_kwargs(kwargs)

[hendrikmakait]

@flying-sheep, thanks for creating this issue. Please provide a minimal reproducer (see http://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports for guidelines) to allow us to investigate your problem.

[flying-sheep]

Not really necessary, you can directly see the problem in the linked line, but sure.

import pandas as pd
import dask.dataframe as dd

custom_sum = dd.Aggregation(
    name='custom_sum',
    chunk=lambda s: s.sum(),
    agg=lambda s0: s0.sum()
)

df = pd.DataFrame(dict(a=[1, 1, 1, 1, 1], g=[5, 6, 6, 6, 7]))
ddf = dd.from_pandas(df, npartitions=2)

ddf.groupby('g')['a'].agg(sum="sum").compute()  # works
ddf.groupby('g')['a'].agg(sum=custom_sum).compute()  # broken

[hendrikmakait]

Thanks for adding the reproducer! Please keep in mind that the time of contributors is limited, so having a reproducer ready allows us to move more quickly. It is also a great starting point for a regression test.