FIX-modin-project#5925: Put a sorting-hack into groupby tests to hide m…

…odin-project#6875 bug

Signed-off-by: Dmitry Chigarev <[email protected]>

modin/pandas/test/test_groupby.py

@@ -3230,10 +3230,6 @@ def _apply_transform(df):
 def test_range_groupby_categories(
     observed, func, by_cols, cat_cols, exclude_values, as_index, modify_config
 ):
-    # HACK: there's a bug in range-partitioning impl that can be triggered
-    # here on certain seeds, manually setting the seed so it won't show up
-    # https://github.com/modin-project/modin/issues/6875
-    np.random.seed(0)
     data = {
         "a": ["a", "b", "c", "d", "e", "b", "g", "a"] * 32,
         "b": [1, 2, 3, 4] * 64,
@@ -3248,4 +3244,8 @@ def test_range_groupby_categories(
 
     md_res = func(md_df.groupby(by_cols, observed=observed, as_index=as_index))
     pd_res = func(pd_df.groupby(by_cols, observed=observed, as_index=as_index))
-    df_equals(md_res, pd_res)
+
+    # HACK, FIXME: there's a bug in range-partitioning impl that apparently can
+    # break the order of rows in the result for multi-column groupbys. Placing the sorting-hack for now
+    # https://github.com/modin-project/modin/issues/6875
+    df_equals(md_res.sort_index(axis=0), pd_res.sort_index(axis=0))