Fix: DataFrameGroupBy.get_group was raising with length>1 tuples (#17653

)

#17216 added similar logic to what's in pandas https://github.com/pandas-dev/pandas/blob/602ae10f3d0d599ebbdd151e8a09f0baf20b4637/pandas/core/groupby/groupby.py#L787-L794, but missed one crucial ingredient: checking that the length of the keys is `1` before raising

Authors:
  - Marco Edward Gorelli (https://github.com/MarcoGorelli)
  - Bradley Dice (https://github.com/bdice)

Approvers:
  - Bradley Dice (https://github.com/bdice)

URL: #17653

python/cudf/cudf/core/groupby/groupby.py

@@ -641,7 +641,7 @@ def get_group(self, name, obj=None):
                 "instead of ``gb.get_group(name, obj=df)``.",
                 FutureWarning,
             )
-        if is_list_like(self._by):
+        if is_list_like(self._by) and len(self._by) == 1:
             if isinstance(name, tuple) and len(name) == 1:
                 name = name[0]
             else:

python/cudf/cudf/tests/test_groupby.py

@@ -4076,6 +4076,13 @@ def test_get_group_list_like():
         df.groupby(["a"]).get_group([1])
 
 
+def test_get_group_list_like_len_2():
+    df = cudf.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6], "c": [3, 2, 1]})
+    result = df.groupby(["a", "b"]).get_group((1, 4))
+    expected = df.to_pandas().groupby(["a", "b"]).get_group((1, 4))
+    assert_eq(result, expected)
+
+
 def test_size_as_index_false():
     df = pd.DataFrame({"a": [1, 2, 1], "b": [1, 2, 3]}, columns=["a", "b"])
     expected = df.groupby("a", as_index=False).size()