Fix feature name consistency between online & historical apis (#1434)

* Fix feature name consistency between online & historical apis

Signed-off-by: Tsotne Tabidze <tsotne@tecton.ai>

* Fix integration tests

Signed-off-by: Tsotne Tabidze <tsotne@tecton.ai>

sdk/python/feast/feature_store.py

@@ -468,13 +468,13 @@ def _get_online_features(
 
                 if feature_data is None:
                     for feature_name in requested_features:
-                        feature_ref = f"{table.name}:{feature_name}"
+                        feature_ref = f"{table.name}__{feature_name}"
                         result_row.statuses[
                             feature_ref
                         ] = GetOnlineFeaturesResponse.FieldStatus.NOT_FOUND
                 else:
                     for feature_name in feature_data:
-                        feature_ref = f"{table.name}:{feature_name}"
+                        feature_ref = f"{table.name}__{feature_name}"
                         if feature_name in requested_features:
                             result_row.fields[feature_ref].CopyFrom(
                                 feature_data[feature_name]

sdk/python/tests/test_e2e_local.py

@@ -74,12 +74,12 @@ def test_e2e_local() -> None:
                 entity_rows=[{"driver_id": 1001}],
             )
 
-            assert "driver_hourly_stats:avg_daily_trips" in result.to_dict()
+            assert "driver_hourly_stats__avg_daily_trips" in result.to_dict()
 
-            assert "driver_hourly_stats:conv_rate" in result.to_dict()
+            assert "driver_hourly_stats__conv_rate" in result.to_dict()
             assert (
                 abs(
-                    result.to_dict()["driver_hourly_stats:conv_rate"][0]
+                    result.to_dict()["driver_hourly_stats__conv_rate"][0]
                     - _get_last_feature_row(driver_df, 1001)["conv_rate"]
                 )
                 < 0.01

sdk/python/tests/test_materialize_from_bigquery_to_datastore.py

@@ -81,13 +81,13 @@ def test_bigquery_table_to_datastore_correctness(self):
         response_dict = fs.get_online_features(
             [f"{fv.name}:value"], [{"driver_id": 1}]
         ).to_dict()
-        assert abs(response_dict[f"{fv.name}:value"][0] - 0.3) < 1e-6
+        assert abs(response_dict[f"{fv.name}__value"][0] - 0.3) < 1e-6
 
         # check prior value for materialize_incremental()
         response_dict = fs.get_online_features(
             [f"{fv.name}:value"], [{"driver_id": 3}]
         ).to_dict()
-        assert abs(response_dict[f"{fv.name}:value"][0] - 4) < 1e-6
+        assert abs(response_dict[f"{fv.name}__value"][0] - 4) < 1e-6
 
         # run materialize_incremental()
         fs.materialize_incremental(
@@ -98,7 +98,7 @@ def test_bigquery_table_to_datastore_correctness(self):
         response_dict = fs.get_online_features(
             [f"{fv.name}:value"], [{"driver_id": 3}]
         ).to_dict()
-        assert abs(response_dict[f"{fv.name}:value"][0] - 5) < 1e-6
+        assert abs(response_dict[f"{fv.name}__value"][0] - 5) < 1e-6
 
     def test_bigquery_query_to_datastore_correctness(self):
         # create dataset
@@ -151,4 +151,4 @@ def test_bigquery_query_to_datastore_correctness(self):
         response_dict = fs.get_online_features(
             [f"{fv.name}:value"], [{"driver_id": 1}]
         ).to_dict()
-        assert abs(response_dict[f"{fv.name}:value"][0] - 0.3) < 1e-6
+        assert abs(response_dict[f"{fv.name}__value"][0] - 0.3) < 1e-6

sdk/python/tests/test_online_retrieval.py

@@ -71,9 +71,9 @@ def test_online() -> None:
             entity_rows=[{"driver": 1}, {"driver": 123}],
         )
 
-        assert "driver_locations:lon" in result.to_dict()
-        assert result.to_dict()["driver_locations:lon"] == ["1.0", None]
-        assert result.to_dict()["driver_locations_2:lon"] == ["2.0", None]
+        assert "driver_locations__lon" in result.to_dict()
+        assert result.to_dict()["driver_locations__lon"] == ["1.0", None]
+        assert result.to_dict()["driver_locations_2__lon"] == ["2.0", None]
 
         # invalid table reference
         with pytest.raises(ValueError):
@@ -99,7 +99,7 @@ def test_online() -> None:
             feature_refs=["driver_locations:lon", "driver_locations_2:lon"],
             entity_rows=[{"driver": 1}, {"driver": 123}],
         )
-        assert result.to_dict()["driver_locations:lon"] == ["1.0", None]
+        assert result.to_dict()["driver_locations__lon"] == ["1.0", None]
 
         # Rename the registry.db so that it cant be used for refreshes
         os.rename(store.config.registry, store.config.registry + "_fake")
@@ -122,7 +122,7 @@ def test_online() -> None:
             feature_refs=["driver_locations:lon", "driver_locations_2:lon"],
             entity_rows=[{"driver": 1}, {"driver": 123}],
         )
-        assert result.to_dict()["driver_locations:lon"] == ["1.0", None]
+        assert result.to_dict()["driver_locations__lon"] == ["1.0", None]
 
         # Create a registry with infinite cache (for users that want to manually refresh the registry)
         fs_infinite_ttl = FeatureStore(
@@ -141,7 +141,7 @@ def test_online() -> None:
             feature_refs=["driver_locations:lon", "driver_locations_2:lon"],
             entity_rows=[{"driver": 1}, {"driver": 123}],
         )
-        assert result.to_dict()["driver_locations:lon"] == ["1.0", None]
+        assert result.to_dict()["driver_locations__lon"] == ["1.0", None]
 
         # Wait a bit so that an arbitrary TTL would take effect
         time.sleep(2)
@@ -154,7 +154,7 @@ def test_online() -> None:
             feature_refs=["driver_locations:lon", "driver_locations_2:lon"],
             entity_rows=[{"driver": 1}, {"driver": 123}],
         )
-        assert result.to_dict()["driver_locations:lon"] == ["1.0", None]
+        assert result.to_dict()["driver_locations__lon"] == ["1.0", None]
 
         # Force registry reload (should fail because file is missing)
         with pytest.raises(FileNotFoundError):