Fix DataFrame.merge(Series, how="left"/"right") on column and index n…

…ot resulting in a RangeIndex (rapidsai#17739)

Left or right merging on a DataFrame column with a Series index (or vice versa) would not result in a `cudf.RangeIndex` in the result like pandas.

IMO not the most ideal fix, but I think this is the most best location for the fix given the flow of this merge code.

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

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

URL: rapidsai#17739

python/cudf/cudf/core/join/join.py

@@ -1,6 +1,7 @@
 # Copyright (c) 2020-2025, NVIDIA CORPORATION.
 from __future__ import annotations
 
+import itertools
 from typing import Any
 
 import pylibcudf as plc
@@ -188,6 +189,23 @@ def __init__(
             self._using_right_index = any(
                 isinstance(idx, _IndexIndexer) for idx in self._right_keys
             )
+            if self.how in {"left", "right"} and not (
+                all(
+                    isinstance(idx, _IndexIndexer)
+                    for idx in itertools.chain(
+                        self._left_keys, self._right_keys
+                    )
+                )
+                or all(
+                    isinstance(idx, _ColumnIndexer)
+                    for idx in itertools.chain(
+                        self._left_keys, self._right_keys
+                    )
+                )
+            ):
+                # For left/right merges, joining on an index and column should result in a RangeIndex
+                self._using_left_index = False
+                self._using_right_index = False
         else:
             # if `on` is not provided and we're not merging
             # index with column or on both indexes, then use

python/cudf/cudf/tests/test_joining.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 
 from itertools import combinations, product, repeat
 
@@ -2275,3 +2275,18 @@ def test_merge_timedelta_types(dtype1, dtype2):
         else True,
         check_dtype=len(actual) > 0,
     )
+
+
+def test_merge_index_on_opposite_how_column_reset_index():
+    df = pd.DataFrame({"a": [1, 2, 3, 4, 5]}, index=[1, 3, 5, 7, 9])
+    ser = pd.Series([1, 2], index=pd.Index([1, 2], name="a"), name="b")
+    df_cudf = cudf.DataFrame.from_pandas(df)
+    ser_cudf = cudf.Series.from_pandas(ser)
+
+    expected = pd.merge(df, ser, on="a", how="left")
+    result = cudf.merge(df_cudf, ser_cudf, on="a", how="left")
+    assert_eq(result, expected)
+
+    expected = pd.merge(ser, df, on="a", how="right")
+    result = cudf.merge(ser_cudf, df_cudf, on="a", how="right")
+    assert_eq(result, expected)