[FSTORE-1728] get_batch_data with spine fg should avoid checking labels

python/hsfs/core/feature_view_engine.py

@@ -335,8 +335,32 @@ def get_batch_query(
                 dataframe_features = engine.get_instance().parse_schema_feature_group(
                     spine.dataframe
                 )
+
+                all_left_feature_group_features = query._left_feature_group.features
+                selected_features = query.features
+                # Using list comprehension
+                selected_left_feature_group_features = [
+                    feature
+                    for feature in all_left_feature_group_features
+                    if feature in selected_features
+                ]
+
+                # remove label features as they are not necessary to check
+                labels = [feat for feat in feature_view_obj.features if feat.label]
+                left_feature_group_features = [
+                    feature
+                    for feature in selected_left_feature_group_features
+                    if not any(
+                        (
+                            feature.name == label.name
+                            and feature._feature_group_id == label._feature_group.id
+                        )
+                        for label in labels
+                    )
+                ]
+
                 spine._feature_group_engine._verify_schema_compatibility(
-                    query._left_feature_group.features, dataframe_features
+                    left_feature_group_features, dataframe_features
                 )
                 query._left_feature_group = spine
             elif isinstance(query._left_feature_group, feature_group.SpineGroup):
@@ -478,7 +502,8 @@ def get_training_data(
                 # forcing dataframe type to default here since dataframe operations are required for training data split.
                 dataframe_type="default"
                 if dataframe_type.lower() in ["numpy", "python"]
-                else dataframe_type,  # forcing dataframe type to default here since dataframe operations are required for training data split.
+                else dataframe_type,
+                # forcing dataframe type to default here since dataframe operations are required for training data split.
             )
         else:
             self._check_feature_group_accessibility(feature_view_obj)

python/hsfs/feature_group.py

@@ -456,8 +456,13 @@ def select_features(
         # Returns
             `Query`. A query object with all features of the feature group.
         """
-        select_features = self.primary_key + self.foreign_key + [self.event_time]
-        if not isinstance(self, ExternalFeatureGroup):
+        select_features = self.primary_key + [self.event_time]
+
+        if not isinstance(self, SpineGroup):
+            select_features = select_features + self.foreign_key
+        if not isinstance(self, ExternalFeatureGroup) and not isinstance(
+            self, SpineGroup
+        ):
             select_features = select_features + self.partition_key
 
         query = self.select_except(select_features)