Add dseries.struct.explode

[isVoid]

Closes #8660

Per discussions in thread #8872 , this PR adds a struct-accessor member function to provide a lateral view to a struct type series.

Example:

>>> import cudf, dask_cudf as dgd
>>> ds = dgd.from_cudf(cudf.Series(
...     [{'a': 42, 'b': 'str1', 'c': [-1]},
...      {'a': 0,  'b': 'str2', 'c': [400, 500]},
...      {'a': 7,  'b': '',     'c': []}]), npartitions=2)
>>> ds.struct.explode().compute()
    a     b           c
0  42  str1        [-1]
1   0  str2  [400, 500]
2   7                []

[rjzamora]

Thanks for putting this together @isVoid ! Nice work - the implementation is nice and clean. I just have a minor suggestion for the new test (and a docstring nit-pick).

[isVoid]

I noticed one of the cases (for nested struct columns) failed because the field names were not properly reconstructed. Investigating.

[isVoid]

I think it's the nested field names got dropped upon constructing a nested type dask cudf object, as this behavior is observed from plain constructing a dask_cudf object:

>>> ds = dask_cudf.from_cudf(cudf.Series([{'a': 123, 'b':{'c': 456}}]), 2)
>>> ds.compute()
0    {'a': 123, 'b': {'0': 456}}
dtype: struct

This reminds me of a similar issue we had in cudf:

• [BUG] Selecting struct column drops field names #7249
• [BUG] cuDF cannot create list of struct dataframe using dict or from pandas #7561
  Both were resolved by applying the dtype after construction.

[isVoid]

rerun tests

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (branch-21.10@8075199). Click here to learn what that means.
The diff coverage is n/a.

❗ Current head 5c7f72e differs from pull request most recent head 61bc39d. Consider uploading reports for the commit 61bc39d to get more accurate results

@@               Coverage Diff               @@
##             branch-21.10    #9086   +/-   ##
===============================================
  Coverage                ?   10.84%           
===============================================
  Files                   ?      116           
  Lines                   ?    18781           
  Branches                ?        0           
===============================================
  Hits                    ?     2037           
  Misses                  ?    16744           
  Partials                ?        0           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8075199...61bc39d. Read the comment docs.

[rjzamora]

LGTM - Thanks!

[galipremsagar]

rerun tests

[galipremsagar]

@gpucibot merge

[galipremsagar]

rerun tests

[galipremsagar]

Not re-triggering the CI so that @rapidsai/ops can look into the logs here.

[galipremsagar]

rerun tests

[galipremsagar]

rerun tests

[galipremsagar]

@gpucibot merge

[galipremsagar]

rerun tests

[galipremsagar]

rerun tests