Fix ListColumn.to_pandas() to retain list type

python/cudf/cudf/core/column/lists.py

@@ -6,6 +6,7 @@
 from typing import List, Optional, Sequence, Tuple, Union
 
 import numpy as np
+import pandas as pd
 import pyarrow as pa
 from typing_extensions import Self
 
@@ -288,6 +289,23 @@ def _transform_leaves(self, func, *args, **kwargs) -> Self:
             )
         return lc
 
+    def to_pandas(
+        self,
+        *,
+        index: Optional[pd.Index] = None,
+        nullable: bool = False,
+    ) -> pd.Series:
+        # Can't rely on Column.to_pandas implementation for lists.
+        # Need to perform `to_pylist` to preserve list types.
+        if nullable:
+            raise NotImplementedError(f"{nullable=} is not implemented.")
+
+        pd_series = pd.Series(self.to_arrow().to_pylist(), dtype="object")
+
+        if index is not None:
+            pd_series.index = index
+        return pd_series
+
 
 class ListMethods(ColumnMethods):
     """

python/cudf/cudf/tests/test_list.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2020-2023, NVIDIA CORPORATION.
+# Copyright (c) 2020-2024, NVIDIA CORPORATION.
 
 import functools
 import operator
@@ -41,6 +41,8 @@ def test_create_list_series(data):
     expect = pd.Series(data)
     got = cudf.Series(data)
     assert_eq(expect, got)
+    assert isinstance(got[0], type(expect[0]))
+    assert isinstance(got.to_pandas()[0], type(expect[0]))
 
 
 @pytest.mark.parametrize(

python/dask_cudf/dask_cudf/tests/test_groupby.py

@@ -702,13 +702,9 @@ def test_is_supported(arg, supported):
 
 def test_groupby_unique_lists():
     df = pd.DataFrame({"a": [0, 0, 0, 1, 1, 1], "b": [10, 10, 10, 7, 8, 9]})
-    ddf = dd.from_pandas(df, 2)
     gdf = cudf.from_pandas(df)
     gddf = dask_cudf.from_cudf(gdf, 2)
-    dd.assert_eq(
-        ddf.groupby("a").b.unique().compute(),
-        gddf.groupby("a").b.unique().compute(),
-    )
+
     dd.assert_eq(
         gdf.groupby("a").b.unique(),
         gddf.groupby("a").b.unique().compute(),