Fix union operation in _is_supported()

[charlesbluca]

Changes the _global_set union operation happening in _is_supported() to

_global_set = _global_set.union(set(arg[col]))

Since set.union() doesn't actually modify the set in place. Before this PR, passing something like {"a": ["unsupported_agg"]} into _is_supported() would always return True.

cc @rjzamora

[rjzamora]

Thanks @charlesbluca ! Looks good.

I would normally ask you to add a test that includes a groupby test that is not supported by the optimized code path. However, it looks like any aggregation that is not supported by that path is broken in upstream dask anyway...

[charlesbluca]

Thanks @rjzamora! Would a good compromise be to add some tests for _is_supported() alone to make sure it returns what we're expecting?

[rjzamora]

Thanks @rjzamora! Would a good compromise be to add some tests for _is_supported() alone to make sure it returns what we're expecting?

Yeah - Although I'd rather not add a (possibly fragile) test for a private function, it's probably better than no test coverage at all. Maybe just test that including a completely made-up aggregation name will return False :)

[charlesbluca]

Sure! I added a simple test for returning False. If needed, I can extend it / add another to check for True.

[charlesbluca]

Also I think I need a category label for the rest of the checks to run

[kkraus14]

Also I think I need a category label for the rest of the checks to run

I think gpuci is just currently down due to a network outage.

[charlesbluca]

rerun tests

[charlesbluca]

Is gpuCI up and running for cuDF again? Reran the tests yesterday and looks like they might have timed out

[kkraus14]

rerun tests

[codecov]

Codecov Report

Merging #7959 (51560eb) into branch-0.20 (599f62d) will increase coverage by 0.19%.
The diff coverage is 91.71%.

❗ Current head 51560eb differs from pull request most recent head 71f3b26. Consider uploading reports for the commit 71f3b26 to get more accurate results

@@               Coverage Diff               @@
##           branch-0.20    #7959      +/-   ##
===============================================
+ Coverage        82.30%   82.49%   +0.19%     
===============================================
  Files              101      103       +2     
  Lines            17053    17306     +253     
===============================================
+ Hits             14035    14277     +242     
- Misses            3018     3029      +11     

Impacted Files	Coverage Δ
python/cudf/cudf/core/column/__init__.py	100.00% <ø> (ø)
python/cudf/cudf/utils/cudautils.py	55.04% <ø> (+4.65%)	⬆️
python/cudf/cudf/utils/dtypes.py	83.44% <ø> (-6.45%)	⬇️
python/cudf/cudf/utils/utils.py	89.44% <ø> (+4.37%)	⬆️
python/cudf/cudf/core/column/numerical.py	94.41% <78.57%> (-0.61%)	⬇️
python/cudf/cudf/core/column/lists.py	86.95% <80.00%> (-0.27%)	⬇️
python/cudf/cudf/core/groupby/groupby.py	91.27% <80.95%> (-2.18%)	⬇️
python/cudf/cudf/core/column/column.py	88.48% <85.36%> (+1.04%)	⬆️
python/cudf/cudf/core/column/struct.py	96.15% <86.66%> (-3.85%)	⬇️
python/cudf/cudf/core/index.py	92.62% <89.13%> (-0.40%)	⬇️
... and 31 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 d1c3245...71f3b26. Read the comment docs.

[kkraus14]

@gpucibot merge