Add arrays of decimal128 to join tests

[kuhushukla]

Fixes #2134
Signed-off-by: Kuhu Shukla kuhus@nvidia.com

Could also add this test additionally but I think the changes below suffice.

# local sort because of https://github.com/NVIDIA/spark-rapids/issues/84
# After 3.1.0 is the min spark version we can drop this
@ignore_order(local=True)
@pytest.mark.parametrize('data_gen', single_array_gens_sample_with_decimal128, ids=idfn)
@pytest.mark.parametrize('join_type', ['Inner', 'Left', 'Right', 'Cross', 'LeftSemi', 'LeftAnti'], ids=idfn)
def test_broadcast_join_right_arr_dec128(data_gen, join_type):
    def do_join(spark):
        left = two_col_df(spark, data_gen, int_gen, length=500)
        right = two_col_df(spark, data_gen, int_gen, length=100)
        return left.join(broadcast(right), left.b == right.b, join_type)
    assert_gpu_and_cpu_are_equal_collect(do_join, conf=allow_negative_scale_of_decimal_conf)

[razajafri]

should we add [databricks] to the title?

[kuhushukla]

build

[revans2]

@razajafri how insistent are you on databricks? If you are okay without it we can merge it in as is. If you really want it we can change the title and rerun

[kuhushukla]

@razajafri Are we good on this change?

[razajafri]

@razajafri how insistent are you on databricks? If you are okay without it we can merge it in as is. If you really want it we can change the title and rerun

Sorry, I didn't mean to hold this up just for databricks. we can merge it