fix: group-by len (#330)

narwhals/_pandas_like/group_by.py

@@ -118,9 +118,12 @@ def agg_pandas(  # noqa: PLR0913
             assert expr._depth == 1
             assert expr._root_names is not None
             assert expr._output_names is not None
+            function_name = remove_prefix(expr._function_name, "col->")
+            function_name = POLARS_TO_PANDAS_AGGREGATIONS.get(
+                function_name, function_name
+            )
             for root_name, output_name in zip(expr._root_names, expr._output_names):
-                name = remove_prefix(expr._function_name, "col->")
-                simple_aggregations[output_name] = (root_name, name)
+                simple_aggregations[output_name] = (root_name, function_name)
 
         aggs = collections.defaultdict(list)
         name_mapping = {}

tests/group_by_test.py

@@ -35,3 +35,12 @@ def test_group_by_iter(constructor: Any) -> None:
     for key, _df in df.group_by(["a", "b"]):
         keys.append(key)
     assert sorted(keys) == sorted(expected_keys)
+
+
+@pytest.mark.parametrize("constructor", [pd.DataFrame, pl.DataFrame])
+def test_group_by_len(constructor: Any) -> None:
+    result = (
+        nw.from_native(constructor(data)).group_by("a").agg(nw.col("b").len()).sort("a")
+    )
+    expected = {"a": [1, 3], "b": [2, 1]}
+    compare_dicts(result, expected)