feat: Making feature view source optional (#6074)

nickquinn408 · nickquinn408 · commit 432c74ab7eb7 · 2026-03-06T16:54:09.000-08:00
diff --git a/docs/getting-started/concepts/batch-feature-view.md b/docs/getting-started/concepts/batch-feature-view.md
@@ -27,7 +27,7 @@ class BatchFeatureView(FeatureView):
     def __init__(
         *,
         name: str,
-        source: Union[DataSource, FeatureView, List[FeatureView]],
+        source: Optional[Union[DataSource, FeatureView, List[FeatureView]]] = None,
         sink_source: Optional[DataSource] = None,
         schema: Optional[List[Field]] = None,
         entities: Optional[List[Entity]] = None,
@@ -142,6 +142,7 @@ See:
 ## 🛑 Gotchas
 
 - `sink_source` is **required** when chaining views (i.e., `source` is another FeatureView or list of them).
+- `source` is optional; if omitted (`None`), the feature view has no associated batch data source.
 - Schema fields must be consistent with `sink_source`, `batch_source.field_mapping` if field mappings exist.
 - Aggregation logic must reference columns present in the raw source or transformed inputs.
 
diff --git a/protos/feast/core/FeatureView.proto b/protos/feast/core/FeatureView.proto
@@ -61,13 +61,15 @@ message FeatureViewSpec {
     google.protobuf.Duration ttl = 6;
 
     // Batch/Offline DataSource where this view can retrieve offline feature data.
+    // Optional: if not set, the feature view has no associated batch data source (e.g. purely derived views).
     DataSource batch_source = 7;
 
     // Whether these features should be served online or not
     // This is also used to determine whether the features should be written to the online store
     bool online = 8;
 
     // Streaming DataSource from where this view can consume "online" feature data.
+    // Optional: only required for streaming feature views.
     DataSource stream_source = 9;
 
     // Description of the feature view.
diff --git a/sdk/python/feast/feature_view.py b/sdk/python/feast/feature_view.py
@@ -73,9 +73,8 @@ class FeatureView(BaseFeatureView):
         ttl: The amount of time this group of features lives. A ttl of 0 indicates that
             this group of features lives forever. Note that large ttl's or a ttl of 0
             can result in extremely computationally intensive queries.
-        batch_source: The batch source of data where this group of features
-            is stored. This is optional ONLY if a push source is specified as the
-            stream_source, since push sources contain their own batch sources.
+        batch_source: Optional batch source of data where this group of features
+            is stored. If no source is provided, this will be None.
         stream_source: The stream source of data where this group of features is stored.
         schema: The schema of the feature view, including feature, timestamp, and entity
             columns. If not specified, can be inferred from the underlying data source.
@@ -97,7 +96,7 @@ class FeatureView(BaseFeatureView):
     name: str
     entities: List[str]
     ttl: Optional[timedelta]
-    batch_source: DataSource
+    batch_source: Optional[DataSource]
     stream_source: Optional[DataSource]
     source_views: Optional[List["FeatureView"]]
     entity_columns: List[Field]
@@ -115,7 +114,7 @@ def __init__(
         self,
         *,
         name: str,
-        source: Union[DataSource, "FeatureView", List["FeatureView"]],
+        source: Optional[Union[DataSource, "FeatureView", List["FeatureView"]]] = None,
         sink_source: Optional[DataSource] = None,
         schema: Optional[List[Field]] = None,
         entities: Optional[List[Entity]] = None,
@@ -133,8 +132,9 @@ def __init__(
 
         Args:
             name: The unique name of the feature view.
-            source: The source of data for this group of features. May be a stream source, or a batch source.
-                If a stream source, the source should contain a batch_source for backfills & batch materialization.
+            source (optional): The source of data for this group of features. May be a stream source,
+                a batch source, a FeatureView, or a list of FeatureViews. If None, the feature view
+                has no associated data source.
             schema (optional): The schema of the feature view, including feature, timestamp,
                 and entity columns.
             # TODO: clarify that schema is only useful here...
@@ -170,7 +170,9 @@ def __init__(
         self.data_source: Optional[DataSource] = None
         self.source_views: List[FeatureView] = []
 
-        if isinstance(source, DataSource):
+        if source is None:
+            pass  # data_source remains None, source_views remains []
+        elif isinstance(source, DataSource):
             self.data_source = source
         elif isinstance(source, FeatureView):
             self.source_views = [source]
@@ -199,11 +201,14 @@ def __init__(
         elif self.data_source:
             # Batch source definition
             self.batch_source = self.data_source
-        else:
+        elif self.source_views:
             # Derived view source definition
             if not sink_source:
                 raise ValueError("Derived FeatureView must specify `sink_source`.")
             self.batch_source = sink_source
+        else:
+            # source=None - no batch source
+            self.batch_source = None
 
         # Initialize features and entity columns.
         features: List[Field] = []
diff --git a/sdk/python/tests/unit/test_feature_views.py b/sdk/python/tests/unit/test_feature_views.py
@@ -23,6 +23,18 @@
 from feast.utils import _utc_now, make_tzaware
 
 
+def test_create_feature_view_without_source():
+    fv = FeatureView(name="test_no_source", ttl=timedelta(days=1))
+    assert fv.batch_source is None
+    assert fv.stream_source is None
+
+    proto = fv.to_proto()
+    assert not proto.spec.HasField("batch_source")
+
+    fv_roundtrip = FeatureView.from_proto(proto)
+    assert fv_roundtrip.batch_source is None
+
+
 def test_create_feature_view_with_conflicting_entities():
     user1 = Entity(name="user1", join_keys=["user_id"])
     user2 = Entity(name="user2", join_keys=["user_id"])
