[BUG] Series unique should return an array rather than a dataframe

[beckernick]

Pandas's unique method returns an array, while ours returns a Series.

import cudf
print(cudf.__version__)
​
df = cudf.DataFrame({
    "a": [0,1,2,0,1,2,0]
})
type(df['a'].unique()), type(df['a'].to_pandas().unique())
0.20.0a+248.g7623f3978b
(cudf.core.series.Series, numpy.ndarray)

[kkraus14]

Hmm, this is problematic for us in handling things like strings or decimal values which can't be supported by arrays as of now. @shwina is scoping building cudf.arrays similar to pandas.arrays which would potentially give a path forward in the future.

[github-actions]

This issue has been labeled inactive-90d due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.

[vyasr]

I'm not sure what we should do about this. I don't think extension arrays are on our roadmap for any time soon. @galipremsagar WDYT? We could match the type for types supported by cupy, but returning a different object based on the data type seems even worse than the status quo. Fundamentally I don't know that we'll ever be able to solve this since we have other types that cupy will probably never support (list and struct, for instance).