Fix Index.difference to handle duplicate values when one of the inp…

…uts is empty (#15016)

This PR removes duplicate values in two short-circuit code-paths of `Index.difference` which is already fixed in `pandas-2.2`

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - Matthew Roeschke (https://github.com/mroeschke)

URL: #15016

python/cudf/cudf/core/_base_index.py

@@ -1096,12 +1096,12 @@ def difference(self, other, sort=None):
         other = cudf.Index(other, name=getattr(other, "name", self.name))
 
         if not len(other):
-            res = self._get_reconciled_name_object(other)
+            res = self._get_reconciled_name_object(other).unique()
             if sort:
                 return res.sort_values()
             return res
         elif self.equals(other):
-            res = self[:0]._get_reconciled_name_object(other)
+            res = self[:0]._get_reconciled_name_object(other).unique()
             if sort:
                 return res.sort_values()
             return res

python/cudf/cudf/tests/test_index.py

@@ -15,7 +15,7 @@
 import cudf
 from cudf.api.extensions import no_default
 from cudf.api.types import is_bool_dtype
-from cudf.core._compat import PANDAS_GE_200
+from cudf.core._compat import PANDAS_GE_200, PANDAS_GE_220
 from cudf.core.index import (
     CategoricalIndex,
     DatetimeIndex,
@@ -797,9 +797,26 @@ def test_index_to_series(data):
     "name_data,name_other",
     [("abc", "c"), (None, "abc"), ("abc", pd.NA), ("abc", "abc")],
 )
-def test_index_difference(data, other, sort, name_data, name_other):
+def test_index_difference(request, data, other, sort, name_data, name_other):
     pd_data = pd.Index(data, name=name_data)
     pd_other = pd.Index(other, name=name_other)
+    request.applymarker(
+        pytest.mark.xfail(
+            condition=PANDAS_GE_220
+            and isinstance(pd_data.dtype, pd.CategoricalDtype)
+            and not isinstance(pd_other.dtype, pd.CategoricalDtype)
+            and pd_other.isnull().any(),
+            reason="https://github.com/pandas-dev/pandas/issues/57318",
+        )
+    )
+    request.applymarker(
+        pytest.mark.xfail(
+            condition=not PANDAS_GE_220
+            and len(pd_other) == 0
+            and len(pd_data) != len(pd_data.unique()),
+            reason="Bug fixed in pandas-2.2+",
+        )
+    )
 
     gd_data = cudf.from_pandas(pd_data)
     gd_other = cudf.from_pandas(pd_other)