SeriesGroupby transform unique not working in Pandas 1.0.1

[Shawe82]

Code Sample, a copy-pastable example if possible

# Your code here
import pandas as pd
df = pd.DataFrame({"title":['a','b','c','c','a'], "asset_id":['a1','b1','c1','c2','c1']})
df.groupby("title").asset_id.transform("unique")

Problem description

Since pandas 1.0.1, unique does not seem to be a whitelisted transform function for SeriesGroupby objects anymore. The above code works with pandas 0.25.3 but not with 1.0.1 It raises a ValueError exception:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-1-ccb0ce21c35f> in <module>
      1 import pandas as pd
      2 df = pd.DataFrame({"title":['a','b','c','c','a'], "asset_id":['a1','b1','c1','c2','c1']})
----> 3 df.groupby("title").asset_id.transform("unique")

~/sandbox/pandas/pandas/core/groupby/generic.py in transform(self, func, *args, **kwargs)
    470         elif func not in base.transform_kernel_whitelist:
    471             msg = f"'{func}' is not a valid function name for transform(name)"
--> 472             raise ValueError(msg)
    473         elif func in base.cythonized_kernels:
    474             # cythonized transform or canned "agg+broadcast"

ValueError: 'unique' is not a valid function name for transform(name)

I tried to find a documentation on that changed but was not succesful. If that is the intended behaviour, I think it would be helpful to document it.

Expected Output

0 [a1, c1]
1 [b1]
2 [c1, c2]
3 [c1, c2]
4 [a1, c1]
Name: asset_id, dtype: object

Output of pd.show_versions()

[paste the output of pd.show_versions() here below this line]
INSTALLED VERSIONS

commit : None
python : 3.8.1.final.0
python-bits : 64
OS : Darwin
OS-release : 18.7.0
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : None
LOCALE : None.UTF-8
pandas : 1.0.1
numpy : 1.18.1
pytz : 2019.3
dateutil : 2.8.1
pip : 19.0.3
setuptools : 40.8.0
Cython : None
pytest : 5.3.5
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : None
pandas_datareader: None
bs4 : None
bottleneck : None
fastparquet : None
gcsfs : None
lxml.etree : None
matplotlib : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pytest : 5.3.5
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : 1.3.13
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
xlsxwriter : None
numba : None

[TomAugspurger]

Thanks. In the future, can you include the full traceback? I've added it

[TomAugspurger]

Last touched in #29672 (cc @jbrockmendel) but I haven't checked if it was working prior to that commit.

[jbrockmendel]

IIRC #29672 was intended to be logic-neutral.

[MarcoGorelli]

If I've done git bisect correctly, it stopped working after this PR: #27467

[WillAyd]

Hmm this seems a little hacky as nothing else in our built-in transform machinery returns a non-scalar (at least that I can think of off the top of my head)

Note that df.groupby("title").asset_id.transform("unique") would also return a different value than df.groupby("title").asset_id.transform(lambda x: x.unique()) which can't be said of other functions

Given that, I'd be -1 on this being considered a regression or even a bug

[MarcoGorelli]

Makes sense. Currently the whitelisted kernels either return a single numeric value (reduction_kernels), or produce a result which has the same shape as the group (transformation_kernels). unique returns a single list, which doesn't fit into either category.

Should this be closed then?

[jreback]

this is not a transformation kernel, you can do this if you want

In [13]: df.groupby('title').asset_id.apply(lambda x: x.unique())                                                                                                                                                                                               
Out[13]: 
title
a    [a1, c1]
b        [b1]
c    [c1, c2]
Name: asset_id, dtype: object