fix: Use group_by in Polars backend aggregation

polars removed DataFrame.groupby in 1.0 in favour of group_by, so PolarsBackend.groupby_agg raised AttributeError and any aggregation run by the local compute engine with backend=polars failed. Signed-off-by: Yihang Chen <yhc0720@berkeley.edu>
haoxu0 · Oct 9, 2026 · Oct 7, 2026 · Oct 9, 2026 · Oct 9, 2026 · Oct 9, 2026
commit dea6862f220e7183f0dbb4b70d92aab556d0c99c
@@ -26,7 +26,7 @@ def groupby_agg(self, df: pl.DataFrame, group_keys, agg_ops) -> pl.DataFrame:
             getattr(pl.col(col), self._POLARS_FUNC_MAP.get(func, func))().alias(alias)
             for alias, (func, col) in agg_ops.items()
         ]
-        return df.groupby(group_keys).agg(agg_exprs)
+        return df.group_by(group_keys).agg(agg_exprs)
 
     def filter(self, df: pl.DataFrame, expr: str) -> pl.DataFrame:
         return df.filter(pl.sql_expr(expr))

@@ -3,6 +3,7 @@
 
 import pandas as pd
 import pyarrow as pa
+import pytest
 
 from feast.infra.compute_engines.backends.pandas_backend import PandasBackend
 from feast.infra.compute_engines.dag.context import ColumnInfo, ExecutionContext
@@ -104,6 +105,34 @@ def test_local_aggregation_node():
     assert result_df["sum_value"].iloc[1] == 70
 
 
+def test_local_aggregation_node_polars_backend():
+    pytest.importorskip("polars")
+    from feast.infra.compute_engines.backends.polars_backend import PolarsBackend
+
+    context = create_context(
+        node_outputs={"source": ArrowTableValue(pa.Table.from_pandas(sample_df))}
+    )
+
+    agg_ops = {
+        "sum_value": ("sum", "value"),
+        "unique_value": ("nunique", "value"),
+    }
+    agg_node = LocalAggregationNode(
+        name="agg",
+        backend=PolarsBackend(),
+        group_keys=["entity_id"],
+        agg_ops=agg_ops,
+    )
+    agg_node.add_input(MagicMock())
+    agg_node.inputs[0].name = "source"
+
+    result = agg_node.execute(context)
+    assert isinstance(result, ArrowTableValue)
+    result_df = result.data.to_pandas().sort_values("entity_id")
+    assert result_df["sum_value"].tolist() == [30, 70]
+    assert result_df["unique_value"].tolist() == [2, 2]
+
+
 def test_local_join_node():
     context = create_context(
         node_outputs={"source": ArrowTableValue(pa.Table.from_pandas(sample_df))}