Merge pull request #166 from ExpediaGroup/reuse_connections

fix: Reuse online store for materialization writes

sdk/python/feast/infra/materialization/contrib/spark/spark_materialization_engine.py

@@ -233,6 +233,7 @@ def _map_by_partition(
     iterator,
     spark_serialized_artifacts: _SparkSerializedArtifacts,
 ):
+    feature_view, online_store, repo_config = spark_serialized_artifacts.unserialize()
     """Load pandas df to online store"""
     for pdf in iterator:
         pdf_row_count = pdf.shape[0]
@@ -244,13 +245,6 @@ def _map_by_partition(
 
         table = pyarrow.Table.from_pandas(pdf)
 
-        # unserialize artifacts
-        (
-            feature_view,
-            online_store,
-            repo_config,
-        ) = spark_serialized_artifacts.unserialize()
-
         if feature_view.batch_source.field_mapping is not None:
             # Spark offline store does the field mapping during pull_latest_from_table_or_query
             # This is for the case where the offline store is not spark