feat(bigquery): Support DATE-type event timestamp columns (#6362)

* feat(bigquery): Support DATE-type event timestamp columns When the event_timestamp column in BigQuery is a DATE type, the generated SQL wraps comparison values in TIMESTAMP(), causing a type mismatch error. This adds a timestamp_field_type parameter to BigQuerySource that, when set to "DATE", generates DATE() comparisons instead. Closes #2530 (part 2) Signed-off-by: Jonathan Wrede <wrede.jonathan00@gmail.com> * fix(bigquery): Use protobuf 4.25.x compatible generated code The proto files were regenerated with protobuf 6.31.1 / grpcio-tools 1.80.0, which imports runtime_version -- a module that does not exist in protobuf 4.25.x used by the project. Revert generated code to 4.25.1 format while keeping the new timestamp_field_type field. Signed-off-by: Jonathan Wrede <wrede.jonathan00@gmail.com> * fix(bigquery): Add Literal type annotation for cast_style Mypy infers str from the ternary expression; annotate with the exact Literal union so the call to get_timestamp_filter_sql passes type checking. Signed-off-by: Jonathan Wrede <wrede.jonathan00@gmail.com> * fix: Make timestamp_field_type default to None in FeatureViewQueryContext Callers that do not use DATE-typed timestamp fields (e.g. Spark offline store tests) should not be forced to pass timestamp_field_type. Adding a default keeps the new field backward-compatible. Signed-off-by: Jonathan Wrede <wrede.jonathan00@gmail.com> * fix: Keep timestamp_field_type required in FeatureViewQueryContext A default value on timestamp_field_type breaks the SparkFeatureViewQueryContext subclass because its non-default fields (min_date_partition, max_date_partition) would follow a field with a default. Instead, keep it required and update the Spark test to pass it. Signed-off-by: Jonathan Wrede <wrede.jonathan00@gmail.com> * fix: regenerate protos matching upstream mypy-protobuf style Reset all non-DataSource generated files to match master. Only DataSource_pb2.py and DataSource_pb2.pyi contain our timestamp_field_type additions (field 28). The .pyi stub is hand-edited to match the existing import style used on master. Signed-off-by: Jonathan Wrede <wrede.jonathan00@gmail.com> --------- Signed-off-by: Jonathan Wrede <wrede.jonathan00@gmail.com> Signed-off-by: RutujaPathade <73137503+RutujaPathade@users.noreply.github.com>
feast-dev · ntkathole · May 23, 2026 · May 13, 2026 · May 8, 2026 · May 15, 2026
commit 2014acd8b7905e0c66b439ad4476e676c1cd966d
diff --git a/protos/feast/core/DataSource.proto b/protos/feast/core/DataSource.proto
@@ -29,7 +29,7 @@ import "feast/types/Value.proto";
 import "feast/core/Feature.proto";
 
 // Defines a Data Source that can be used source Feature data
-// Next available id: 28
+// Next available id: 29
 message DataSource {
   // Field indexes should *not* be reused. Not sure if fields 6-10 were used previously or not,
   // but they are going to be reserved for backwards compatibility.
@@ -81,6 +81,10 @@ message DataSource {
   // Must specify creation timestamp column name
   string created_timestamp_column = 5;
 
+  // (Optional) Type of the timestamp_field column ("TIMESTAMP" or "DATE").
+  // When set to "DATE", SQL generation uses date-only comparisons.
+  string timestamp_field_type = 28;
+
   // This is an internal field that is represents the python class for the data source object a proto object represents.
   // This should be set by feast, and not by users.
   // The field is used primarily by custom data sources and is mandatory for them to set. Feast may set it for

diff --git a/sdk/python/feast/data_source.py b/sdk/python/feast/data_source.py
@@ -205,6 +205,7 @@ class DataSource(ABC):
     tags: Dict[str, str]
     owner: str
     date_partition_column: str
+    timestamp_field_type: str
     created_timestamp: Optional[datetime]
     last_updated_timestamp: Optional[datetime]
 
@@ -219,6 +220,7 @@ def __init__(
         tags: Optional[Dict[str, str]] = None,
         owner: Optional[str] = "",
         date_partition_column: Optional[str] = None,
+        timestamp_field_type: Optional[str] = None,
     ):
         """
         Creates a DataSource object.
@@ -237,6 +239,9 @@ def __init__(
             owner (optional): The owner of the data source, typically the email of the primary
                 maintainer.
             date_partition_column (optional): Timestamp column used for partitioning. Not supported by all stores
+            timestamp_field_type (optional): Type of the timestamp_field column.
+                Defaults to "TIMESTAMP". Set to "DATE" when the event timestamp column
+                is a DATE type, so SQL generation uses date-only comparisons.
         """
         self.name = name
         self.timestamp_field = timestamp_field or ""
@@ -257,6 +262,7 @@ def __init__(
         self.date_partition_column = (
             date_partition_column if date_partition_column else ""
         )
+        self.timestamp_field_type = timestamp_field_type if timestamp_field_type else ""
         now = _utc_now()
         self.created_timestamp = now
         self.last_updated_timestamp = now
@@ -280,6 +286,7 @@ def __eq__(self, other):
             or self.created_timestamp_column != other.created_timestamp_column
             or self.field_mapping != other.field_mapping
             or self.date_partition_column != other.date_partition_column
+            or self.timestamp_field_type != other.timestamp_field_type
             or self.description != other.description
             or self.tags != other.tags
             or self.owner != other.owner

diff --git a/sdk/python/feast/infra/offline_stores/bigquery.py b/sdk/python/feast/infra/offline_stores/bigquery.py
@@ -161,13 +161,18 @@ def pull_latest_from_table_or_query(
             project=project_id,
             location=config.offline_store.location,
         )
+        cast_style: Literal["date_func", "timestamp_func"] = (
+            "date_func"
+            if data_source.timestamp_field_type == "DATE"
+            else "timestamp_func"
+        )
         timestamp_filter = get_timestamp_filter_sql(
             start_date,
             end_date,
             timestamp_field,
             date_partition_column=data_source.date_partition_column,
             quote_fields=False,
-            cast_style="timestamp_func",
+            cast_style=cast_style,
         )
         query = f"""
             SELECT
@@ -220,13 +225,18 @@ def pull_all_from_table_or_query(
             + BigQueryOfflineStore._escape_query_columns(feature_name_columns)
             + timestamp_fields
         )
+        cast_style: Literal["date_func", "timestamp_func"] = (
+            "date_func"
+            if data_source.timestamp_field_type == "DATE"
+            else "timestamp_func"
+        )
         timestamp_filter = get_timestamp_filter_sql(
             start_date,
             end_date,
             timestamp_field,
             date_partition_column=data_source.date_partition_column,
             quote_fields=False,
-            cast_style="timestamp_func",
+            cast_style=cast_style,
         )
         query = f"""
             SELECT {field_string}
@@ -938,10 +948,17 @@ def arrow_schema_to_bq_schema(arrow_schema: pyarrow.Schema) -> List[SchemaField]
                 {% if loop.last %}{% else %}, {% endif %}
             {% endfor %}
         FROM {{ featureview.table_subquery }}
+        {% if featureview.timestamp_field_type == "DATE" %}
+        WHERE {{ featureview.timestamp_field }} <= DATE('{{ featureview.max_event_timestamp[:10] }}')
+        {% if featureview.ttl == 0 %}{% else %}
+        AND {{ featureview.timestamp_field }} >= DATE('{{ featureview.min_event_timestamp[:10] }}')
+        {% endif %}
+        {% else %}
         WHERE {{ featureview.timestamp_field }} <= '{{ featureview.max_event_timestamp }}'
         {% if featureview.ttl == 0 %}{% else %}
         AND {{ featureview.timestamp_field }} >= '{{ featureview.min_event_timestamp }}'
         {% endif %}
+        {% endif %}
         {% if featureview.date_partition_column %}
         AND {{ featureview.date_partition_column | backticks }} <= '{{ featureview.max_event_timestamp[:10] }}'
         {% if featureview.min_event_timestamp %}

diff --git a/sdk/python/feast/infra/offline_stores/bigquery_source.py b/sdk/python/feast/infra/offline_stores/bigquery_source.py
@@ -35,6 +35,7 @@ def __init__(
         created_timestamp_column: Optional[str] = "",
         field_mapping: Optional[Dict[str, str]] = None,
         date_partition_column: Optional[str] = None,
+        timestamp_field_type: Optional[str] = None,
         query: Optional[str] = None,
         description: Optional[str] = "",
         tags: Optional[Dict[str, str]] = None,
@@ -54,6 +55,9 @@ def __init__(
             field_mapping (optional): A dictionary mapping of column names in this data source to feature names in a feature table
                 or view. Only used for feature columns, not entities or timestamp columns.
             date_partition_column (optional): Timestamp column used for partitioning.
+            timestamp_field_type (optional): Type of the timestamp_field column.
+                Set to "DATE" when the event timestamp column is a DATE type,
+                so SQL generation uses date-only comparisons instead of TIMESTAMP().
             query (optional): The query to be executed to obtain the features. When both 'table'
                 and 'query' are provided, 'query' takes priority for reads.
             description (optional): A human-readable description.
@@ -81,6 +85,7 @@ def __init__(
             created_timestamp_column=created_timestamp_column,
             field_mapping=field_mapping,
             date_partition_column=date_partition_column,
+            timestamp_field_type=timestamp_field_type,
             description=description,
             tags=tags,
             owner=owner,
@@ -121,6 +126,7 @@ def from_proto(data_source: DataSourceProto):
             timestamp_field=data_source.timestamp_field,
             created_timestamp_column=data_source.created_timestamp_column,
             date_partition_column=data_source.date_partition_column,
+            timestamp_field_type=data_source.timestamp_field_type or None,
             query=data_source.bigquery_options.query,
             description=data_source.description,
             tags=dict(data_source.tags),
@@ -139,6 +145,7 @@ def _to_proto_impl(self) -> DataSourceProto:
             timestamp_field=self.timestamp_field,
             created_timestamp_column=self.created_timestamp_column,
             date_partition_column=self.date_partition_column,
+            timestamp_field_type=self.timestamp_field_type,
         )
 
         return data_source_proto

diff --git a/sdk/python/feast/infra/offline_stores/offline_utils.py b/sdk/python/feast/infra/offline_stores/offline_utils.py
@@ -98,6 +98,7 @@ class FeatureViewQueryContext:
     date_partition_column: Optional[
         str
     ]  # this attribute is added because partition pruning affects Athena's query performance.
+    timestamp_field_type: Optional[str]
 
 
 def get_feature_view_query_context(
@@ -160,6 +161,10 @@ def get_feature_view_query_context(
             feature_view.batch_source.date_partition_column,
         )
 
+        timestamp_field_type = getattr(
+            feature_view.batch_source, "timestamp_field_type", ""
+        )
+
         max_event_timestamp = to_naive_utc(entity_df_timestamp_range[1]).isoformat()
         min_event_timestamp = None
         if feature_view.ttl:
@@ -181,6 +186,7 @@ def get_feature_view_query_context(
             min_event_timestamp=min_event_timestamp,
             max_event_timestamp=max_event_timestamp,
             date_partition_column=date_partition_column,
+            timestamp_field_type=timestamp_field_type or None,
         )
         query_context.append(context)
 
@@ -340,7 +346,7 @@ def get_timestamp_filter_sql(
     date_partition_column: Optional[str] = None,
     tz: Optional[timezone] = None,
     cast_style: Literal[
-        "timestamp", "timestamp_func", "timestamptz", "raw"
+        "timestamp", "timestamp_func", "timestamptz", "raw", "date_func"
     ] = "timestamp",
     date_time_separator: str = "T",
     quote_fields: bool = True,
@@ -355,10 +361,11 @@ def get_timestamp_filter_sql(
         date_partition_column: optional partition column (for pruning)
         tz: optional timezone for datetime inputs
         cast_style: one of:
-            - "timestamp": TIMESTAMP '...'                  → Common Sql engine Snowflake, Redshift etc.
+            - "timestamp": TIMESTAMP '...'              → Common Sql engine Snowflake, Redshift etc.
             - "timestamp_func": TIMESTAMP('...')         → BigQuery, Couchbase etc.
             - "timestamptz": '...'::timestamptz          → PostgreSQL
             - "raw": '...'                               → no cast, string only
+            - "date_func": DATE('...')                   → BigQuery DATE columns
         date_time_separator: separator for datetime strings (default is "T")
             (e.g. "2023-10-01T00:00:00" or "2023-10-01 00:00:00")
         quote_fields: whether to quote the timestamp and partition column names
@@ -384,6 +391,9 @@ def format_casted_ts(val: Union[str, datetime]) -> str:
             return f"TIMESTAMP '{val_str}'"
         elif cast_style == "timestamp_func":
             return f"TIMESTAMP('{val_str}')"
+        elif cast_style == "date_func":
+            date_str = val_str[:10] if len(val_str) >= 10 else val_str
+            return f"DATE('{date_str}')"
         elif cast_style == "timestamptz":
             return f"'{val_str}'::{cast_style}"
         else: