Implement groupby collect_set

[ttnghia]

This partially addresses #2973.

This PR implements groupby collect_set aggregation. The idea of this PR is to simply apply drop_list_duplicates (#7528) to the result generated by groupby collect_list, obtaining collect lists without duplicate entries.

Examples:

keys = {1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3};
vals = {10, 11, 10, 10, 20, 21, 21, 20, 30, 33, 32, 31};

keys_output = {1, 2, 3};
vals_output = {{10, 11}, {20, 21}, {30, 31, 32, 33}};

In this PR, a simple, incomplete Python binding for collect_set has been added, and no Java binding is implemented yet. Complete bindings for those Python/Java sides need to be implemented later in some other separate PRs.

[jrhemstad]

Is this only for rolling? If so, can you clarify in the title? Also, the PR description needs to be more complete as it is used in generating the CHANGELOG.

[ttnghia]

Is this only for rolling? If so, can you clarify in the title? Also, the PR description needs to be more complete as it is used in generating the CHANGELOG.

Hi Jake. I just started to work on this thus things are still unclear to put in. I'll update more details for the description along the way.

[mythrocks]

Might I suggest a change to the title and description of this issue? Let's just tackle renaming COLLECT to COLLECT_LIST.
(This is assuming that @shwina has no objection.)

We can address COLLECT_SET in groupby and rolling_window as separate tasks.

[mythrocks]

Also, if this is a work in progress, it would be best to switch this to a draft PR rather than use [WIP] in the title.

[shwina]

Changing COLLECT to COLLECT_LIST sounds absolutely fine to me, although for now, we probably still want the name of the aggregation in Cython/Python to be `"collect"``.

[ttnghia]

Rerun tests.

[ttnghia]

Rerun tests.

[kkraus14]

cmake / conda lgtm, will let @shwina review the Python side 😄

[jlowe]

Java approval

[ttnghia]

Rerun tests.

[ttnghia]

Rerun tests.

[ttnghia]

Rerun tests.

[ttnghia]

Rerun tests.

[ttnghia]

Rerun tests.

[ttnghia]

Rerun tests.

[shwina]

Cython LGTM!

[kkraus14]

@gpucibot merge

[ttnghia]

Thanks all 😃

[codecov]

Codecov Report

Merging #7420 (ac3e9a2) into branch-0.19 (7871e7a) will increase coverage by 0.61%.
The diff coverage is n/a.

❗ Current head ac3e9a2 differs from pull request most recent head 918eeb7. Consider uploading reports for the commit 918eeb7 to get more accurate results

@@               Coverage Diff               @@
##           branch-0.19    #7420      +/-   ##
===============================================
+ Coverage        81.86%   82.47%   +0.61%     
===============================================
  Files              101      101              
  Lines            16884    17402     +518     
===============================================
+ Hits             13822    14353     +531     
+ Misses            3062     3049      -13     

Impacted Files	Coverage Δ
python/cudf/cudf/core/column/categorical.py	91.97% <ø> (+0.58%)	⬆️
python/cudf/cudf/core/column/column.py	87.86% <ø> (+0.10%)	⬆️
python/cudf/cudf/core/column/datetime.py	89.63% <ø> (+0.54%)	⬆️
python/cudf/cudf/core/column/decimal.py	92.75% <ø> (-2.12%)	⬇️
python/cudf/cudf/core/column/lists.py	92.50% <ø> (+1.10%)	⬆️
python/cudf/cudf/core/column/numerical.py	94.83% <ø> (-0.20%)	⬇️
python/cudf/cudf/core/column/string.py	86.79% <ø> (+0.30%)	⬆️
python/cudf/cudf/core/column/timedelta.py	88.57% <ø> (+0.33%)	⬆️
python/cudf/cudf/core/column_accessor.py	95.45% <ø> (+0.14%)	⬆️
python/cudf/cudf/core/dataframe.py	90.90% <ø> (+0.44%)	⬆️
... and 61 more

---

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 d500142...918eeb7. Read the comment docs.