Fix Index.difference to match with pandas

python/cudf/cudf/core/_base_index.py

@@ -30,7 +30,7 @@
 from cudf.core.column import ColumnBase, column
 from cudf.core.column_accessor import ColumnAccessor
 from cudf.utils import ioutils
-from cudf.utils.dtypes import is_mixed_with_object_dtype
+from cudf.utils.dtypes import can_convert_to_column, is_mixed_with_object_dtype
 from cudf.utils.utils import _is_same_name
 
 
@@ -935,13 +935,21 @@ def difference(self, other, sort=None):
         >>> idx1.difference(idx2, sort=False)
         Int64Index([2, 1], dtype='int64')
         """
+        if not can_convert_to_column(other):
+            raise TypeError("Input must be Index or array-like")
+
         if sort not in {None, False}:
             raise ValueError(
                 f"The 'sort' keyword only takes the values "
                 f"of None or False; {sort} was passed."
             )
 
-        other = cudf.Index(other)
+        other = cudf.Index(other, name=getattr(other, "name", self.name))
+
+        if not len(other):
+            return self._get_reconciled_name_object(other)
+        elif self.equals(other):
+            return self[:0]._get_reconciled_name_object(other)
 
         res_name = _get_result_name(self.name, other.name)
 

python/cudf/cudf/core/index.py

@@ -724,6 +724,29 @@ def _intersection(self, other, sort=False):
 
         return new_index
 
+    @_cudf_nvtx_annotate
+    def difference(self, other, sort=None):
+        if isinstance(other, RangeIndex) and self.equals(other):
+            return self[:0]._get_reconciled_name_object(other)
+
+        return self._try_reconstruct_range_index(
+            super().difference(other, sort=sort)
+        )
+
+    def _try_reconstruct_range_index(self, index):
+        if index.dtype.kind == "f":
+            return index
+        # Evenly spaced values can return a
+        # RangeIndex instead of Int64Index
+        unique_diffs = index.to_frame(name="None").diff()["None"].unique()
+        if len(unique_diffs) == 2 and (
+            unique_diffs[0] is cudf.NA and unique_diffs[1] != 0
+        ):
+            diff = unique_diffs[1]
+            new_range = range(index[0], index[-1] + diff, diff)
+            return type(self)(new_range, name=index.name)
+        return index
+
     def sort_values(
         self,
         return_indexer=False,

python/cudf/cudf/tests/test_index.py

@@ -788,6 +788,10 @@ def test_index_to_series(data):
         ["5", "6", "2", "a", "b", "c"],
         [1.0, 2.0, 3.0, 4.0, 5.0, 6.0],
         [1.0, 5.0, 6.0, 0.0, 1.3],
+        ["ab", "cd", "ef"],
+        pd.Series(["1", "2", "a", "3", None], dtype="category"),
+        range(0, 10),
+        [],
     ],
 )
 @pytest.mark.parametrize(
@@ -798,8 +802,11 @@ def test_index_to_series(data):
         [10, 20, 30, 40, 50, 60],
         ["1", "2", "3", "4", "5", "6"],
         ["5", "6", "2", "a", "b", "c"],
+        ["ab", "ef", None],
         [1.0, 2.0, 3.0, 4.0, 5.0, 6.0],
         [1.0, 5.0, 6.0, 0.0, 1.3],
+        range(2, 4),
+        pd.Series(["1", "a", "3", None], dtype="category"),
         [],
     ],
 )
@@ -817,6 +824,7 @@ def test_index_difference(data, other, sort, name_data, name_other):
 
     expected = pd_data.difference(pd_other, sort=sort)
     actual = gd_data.difference(gd_other, sort=sort)
+
     assert_eq(expected, actual)
 
 

python/cudf/cudf/utils/dtypes.py

@@ -426,6 +426,11 @@ def get_min_float_dtype(col):
 
 
 def is_mixed_with_object_dtype(lhs, rhs):
+    if cudf.api.types.is_categorical_dtype(lhs.dtype):
+        return is_mixed_with_object_dtype(lhs.dtype.categories, rhs)
+    elif cudf.api.types.is_categorical_dtype(rhs.dtype):
+        return is_mixed_with_object_dtype(lhs, rhs.dtype.categories)
+
     return (lhs.dtype == "object" and rhs.dtype != "object") or (
         rhs.dtype == "object" and lhs.dtype != "object"
     )