API: update Series.apply to follow changes in DataFrame.apply #19866
Description
As a follow-up to the changes in DataFrame.apply (#18577), we should make sure Series.apply behaves consistently with that as well.
I think the default behaviour is OK (it already did what we now do by default for DataFrame: expand Series, keep lists or arrays as scalars, see below).
But we could in principle also add the same control over the output type by adding a similar result_type keyword.
Current behaviour:
In [42]: s = pd.Series([1, 2, 3, 4], index=['a', 'b', 'c', 'd']) In [44]: s.apply(lambda x: [1, 2]) Out[44]: a [1, 2] b [1, 2] c [1, 2] d [1, 2] dtype: object In [45]: s.apply(lambda x: pd.Series([1, 2])) Out[45]: 0 1 a 1 2 b 1 2 c 1 2 d 1 2 In [46]: s.apply(lambda x: pd.Series([1, 2], index=['A', 'B'])) Out[46]: A B a 1 2 b 1 2 c 1 2 d 1 2 In [47]: s.apply(lambda x: np.array([1, 2])) Out[47]: a [1, 2] b [1, 2] c [1, 2] d [1, 2] dtype: object