fix: add project_id filter to SnowflakeRegistry UPDATE path

_apply_object had a missing project_id clause in the UPDATE branch's WHERE condition. In a shared Snowflake registry, this allowed a feast apply in one project to silently overwrite a same-named object belonging to a different project. The SELECT path already scoped correctly by project_id. The DELETE path (_delete_object) also scoped correctly. This commit brings the UPDATE path in line with both. Fixes #6208 Signed-off-by: Abhishek8108 <87538407+Abhishek8108@users.noreply.github.com>
ntkathole · Apr 11, 2026 · Apr 8, 2026 · Apr 10, 2026 · Apr 11, 2026 · Apr 11, 2026
commit bb4eca05f68b558f67c21024c8dd0616c6c1c638
@@ -373,7 +373,8 @@ def _apply_object(
                             {proto_field_name} = TO_BINARY({proto}),
                             last_updated_timestamp = CURRENT_TIMESTAMP()
                         WHERE
-                            {id_field_name.lower()} = '{name}'
+                            project_id = '{project}'
+                            AND {id_field_name.lower()} = '{name}'
                 """
                 execute_snowflake_statement(conn, query)
 

@@ -0,0 +1,116 @@
+# Copyright 2024 The Feast Authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     https://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from unittest.mock import MagicMock, patch
+
+import pandas as pd
+import pytest
+
+from feast.entity import Entity
+from feast.infra.registry.snowflake import SnowflakeRegistry, SnowflakeRegistryConfig
+
+
+@pytest.fixture
+def registry():
+    config = SnowflakeRegistryConfig(
+        database="TEST_DB",
+        schema="PUBLIC",
+        account="test_account",
+        user="test_user",
+        password="test_password",
-        password="test_password",
+        password="test_password",  #pragma: allowlist secret
-        password="test_password",
+        password="test_password",  #pragma: allowlist secret
+    )
+    mock_conn = MagicMock()
+    mock_conn.__enter__ = MagicMock(return_value=MagicMock())
+    mock_conn.__exit__ = MagicMock(return_value=False)
+
+    # execute_snowflake_statement during __init__ creates tables; return empty cursor
+    mock_cursor = MagicMock()
+    mock_cursor.fetch_pandas_all.return_value = pd.DataFrame()
+
+    with patch(
+        "feast.infra.registry.snowflake.GetSnowflakeConnection", return_value=mock_conn
+    ), patch(
+        "feast.infra.registry.snowflake.execute_snowflake_statement",
+        return_value=mock_cursor,
+    ):
+        reg = SnowflakeRegistry(config, "test_project", None)
+    return reg
+
+
+def test_apply_object_update_includes_project_id_in_where_clause(registry):
+    """
+    Regression test for feast-dev/feast#6208.
+
+    _apply_object UPDATE path was missing project_id in the WHERE clause,
+    allowing a same-named object in one project to silently overwrite the
+    same-named object in a different project when a Snowflake registry is shared.
+
+    The SELECT path correctly scopes by project_id; the UPDATE must do the same.
+    """
+    entity = Entity(name="driver", join_keys=["driver_id"])
+    project = "project_b"
+
+    # Non-empty DataFrame tells _apply_object the row already exists → UPDATE path
+    existing_row = pd.DataFrame({"project_id": [project]})
+
+    captured_queries = []
+
+    def capture_and_respond(conn, query):
+        normalised = " ".join(query.split())
+        captured_queries.append(normalised)
+        cursor = MagicMock()
+        cursor.fetch_pandas_all.return_value = (
+            existing_row if "SELECT" in normalised else pd.DataFrame()
+        )
+        return cursor
+
+    mock_conn = MagicMock()
+    mock_conn.__enter__ = MagicMock(return_value=MagicMock())
+    mock_conn.__exit__ = MagicMock(return_value=False)
+
+    with patch(
+        "feast.infra.registry.snowflake.GetSnowflakeConnection", return_value=mock_conn
+    ), patch(
+        "feast.infra.registry.snowflake.execute_snowflake_statement",
+        side_effect=capture_and_respond,
+    ), patch.object(
+        registry, "_maybe_init_project_metadata"
+    ), patch.object(
+        registry, "_initialize_project_if_not_exists"
+    ), patch.object(
+        registry, "get_project", return_value=MagicMock()
+    ), patch.object(
+        registry, "apply_project"
+    ), patch.object(
+        registry, "_set_last_updated_metadata"
+    ):
+        registry._apply_object(
+            "ENTITIES",
+            project,
+            "ENTITY_NAME",
+            entity,
+            "ENTITY_PROTO",
+        )
+
+    update_queries = [q for q in captured_queries if q.startswith("UPDATE")]
+    assert len(update_queries) == 1, (
+        f"Expected exactly 1 UPDATE query, got {len(update_queries)}: {update_queries}"
+    )
+
+    update_query = update_queries[0]
+    assert f"project_id = '{project}'" in update_query, (
+        f"feast#6208: UPDATE WHERE clause is missing project_id filter — "
+        f"cross-project overwrites are possible in shared Snowflake registries.\n"
+        f"Query was: {update_query}"
+    )