(npartitions, split_every, split_out)
| 3999 | @pytest.mark.parametrize("split_every", [2, 5]) |
| 4000 | @pytest.mark.parametrize("split_out", [1, 5, 20]) |
| 4001 | def test_hash_split_unique(npartitions, split_every, split_out): |
| 4002 | from string import ascii_lowercase |
| 4003 | |
| 4004 | s = pd.Series(np.random.choice(list(ascii_lowercase), 1000, replace=True)) |
| 4005 | ds = dd.from_pandas(s, npartitions=npartitions) |
| 4006 | |
| 4007 | dropped = ds.unique(split_every=split_every, split_out=split_out) |
| 4008 | |
| 4009 | dsk = dropped.__dask_optimize__(dropped.dask, dropped.__dask_keys__()) |
| 4010 | from dask.core import get_deps |
| 4011 | |
| 4012 | dependencies, dependents = get_deps(dsk) |
| 4013 | |
| 4014 | assert dropped.npartitions == (split_out or 1) |
| 4015 | assert sorted(dropped.compute(scheduler="sync")) == sorted(s.unique()) |
| 4016 | |
| 4017 | |
| 4018 | @pytest.mark.parametrize("split_every", [None, 2]) |
nothing calls this directly
no test coverage detected
searching dependent graphs…