Merge pull request #1962 from weaviate/pr/1942

Fix empty list filter raising IndexError instead of WeaviateInvalidInputError

Author: dirkkul

integration/test_rbac.py

@@ -389,7 +389,7 @@
         (
             Permissions.alias(alias="myCAR", collection="*", read=True, delete=True),
             Role(
-                name="AlliasRole",
+                name="AlliasRole3",
                 alias_permissions=[
                     AliasPermissionOutput(
                         alias="MyCAR",  # capitalized the first letter.

test/collection/test_filter.py

@@ -9,8 +9,10 @@
     _FilterAnd,
     _FilterNot,
     _FilterOr,
+    _FilterValue,
     _Operator,
 )
+from weaviate.collections.filters import _FilterToGRPC
 from weaviate.proto.v1 import base_pb2
 
 
@@ -26,6 +28,43 @@ def test_empty_input_contains_all() -> None:
         wvc.query.Filter.by_property("test").contains_all([])
 
 
+def test_empty_list_equal() -> None:
+    with pytest.raises(weaviate.exceptions.WeaviateInvalidInputError):
+        wvc.query.Filter.by_property("test").equal([])
+
+
+def test_empty_list_not_equal() -> None:
+    with pytest.raises(weaviate.exceptions.WeaviateInvalidInputError):
+        wvc.query.Filter.by_property("test").not_equal([])
+
+
+def test_empty_list_less_than() -> None:
+    with pytest.raises(weaviate.exceptions.WeaviateInvalidInputError):
+        wvc.query.Filter.by_property("test").less_than([])
+
+
+def test_empty_list_less_or_equal() -> None:
+    with pytest.raises(weaviate.exceptions.WeaviateInvalidInputError):
+        wvc.query.Filter.by_property("test").less_or_equal([])
+
+
+def test_empty_list_greater_than() -> None:
+    with pytest.raises(weaviate.exceptions.WeaviateInvalidInputError):
+        wvc.query.Filter.by_property("test").greater_than([])
+
+
+def test_empty_list_greater_or_equal() -> None:
+    with pytest.raises(weaviate.exceptions.WeaviateInvalidInputError):
+        wvc.query.Filter.by_property("test").greater_or_equal([])
+
+
+def test_empty_list_grpc_conversion() -> None:
+    """Ensure the gRPC converter raises WeaviateInvalidInputError for empty lists."""
+    fv = _FilterValue(target="test", value=[], operator=_Operator.EQUAL)
+    with pytest.raises(weaviate.exceptions.WeaviateInvalidInputError):
+        _FilterToGRPC.convert(fv)
+
+
 def test_filter_lists() -> None:
     f1 = wvc.query.Filter.by_property("test").equal("test")
     f2 = wvc.query.Filter.by_creation_time().greater_or_equal(datetime.datetime.now())

weaviate/collections/classes/filters.py

@@ -220,18 +220,42 @@ def contains_none(self, val: FilterValuesList) -> _Filters:
 
     def equal(self, val: FilterValues) -> _Filters:
         """Filter on whether the property is equal to the given value."""
+        if isinstance(val, list) and len(val) == 0:
+            raise WeaviateInvalidInputError(
+                "Filtering on empty lists is not supported by Weaviate. "
+                "To filter by property length, use "
+                "Filter.by_property('prop', length=True).equal(0)"
+            )
         return _FilterValue(target=self._target_path(), value=val, operator=_Operator.EQUAL)
 
     def not_equal(self, val: FilterValues) -> _Filters:
         """Filter on whether the property is not equal to the given value."""
+        if isinstance(val, list) and len(val) == 0:
+            raise WeaviateInvalidInputError(
+                "Filtering on empty lists is not supported by Weaviate. "
+                "To filter by property length, use "
+                "Filter.by_property('prop', length=True).equal(0)"
+            )
         return _FilterValue(target=self._target_path(), value=val, operator=_Operator.NOT_EQUAL)
 
     def less_than(self, val: FilterValues) -> _Filters:
         """Filter on whether the property is less than the given value."""
+        if isinstance(val, list) and len(val) == 0:
+            raise WeaviateInvalidInputError(
+                "Filtering on empty lists is not supported by Weaviate. "
+                "To filter by property length, use "
+                "Filter.by_property('prop', length=True).equal(0)"
+            )
         return _FilterValue(target=self._target_path(), value=val, operator=_Operator.LESS_THAN)
 
     def less_or_equal(self, val: FilterValues) -> _Filters:
         """Filter on whether the property is less than or equal to the given value."""
+        if isinstance(val, list) and len(val) == 0:
+            raise WeaviateInvalidInputError(
+                "Filtering on empty lists is not supported by Weaviate. "
+                "To filter by property length, use "
+                "Filter.by_property('prop', length=True).equal(0)"
+            )
         return _FilterValue(
             target=self._target_path(),
             value=val,
@@ -240,6 +264,12 @@ def less_or_equal(self, val: FilterValues) -> _Filters:
 
     def greater_than(self, val: FilterValues) -> _Filters:
         """Filter on whether the property is greater than the given value."""
+        if isinstance(val, list) and len(val) == 0:
+            raise WeaviateInvalidInputError(
+                "Filtering on empty lists is not supported by Weaviate. "
+                "To filter by property length, use "
+                "Filter.by_property('prop', length=True).equal(0)"
+            )
         return _FilterValue(
             target=self._target_path(),
             value=val,
@@ -248,6 +278,12 @@ def greater_than(self, val: FilterValues) -> _Filters:
 
     def greater_or_equal(self, val: FilterValues) -> _Filters:
         """Filter on whether the property is greater than or equal to the given value."""
+        if isinstance(val, list) and len(val) == 0:
+            raise WeaviateInvalidInputError(
+                "Filtering on empty lists is not supported by Weaviate. "
+                "To filter by property length, use "
+                "Filter.by_property('prop', length=True).equal(0)"
+            )
         return _FilterValue(
             target=self._target_path(),
             value=val,

weaviate/collections/filters.py

@@ -40,6 +40,12 @@ def convert(weav_filter: Optional[_Filters]) -> Optional[base_pb2.Filters]:
 
     @staticmethod
     def __value_filter(weav_filter: _FilterValue) -> base_pb2.Filters:
+        if isinstance(weav_filter.value, list) and len(weav_filter.value) == 0:
+            raise WeaviateInvalidInputError(
+                "Filtering on empty lists is not supported by Weaviate. "
+                "To filter by property length, use "
+                "Filter.by_property('prop', length=True).equal(0)"
+            )
         operator = weav_filter.operator._to_grpc()
         target = _FilterToGRPC.__to_target(weav_filter.target)
         if isinstance(weav_filter.value, bool):
@@ -116,7 +122,9 @@ def __filter_to_text(value: FilterValues) -> Optional[str]:
 
     @staticmethod
     def __filter_to_text_list(value: FilterValues) -> Optional[base_pb2.TextArray]:
-        if not isinstance(value, list) or not (
+        if not isinstance(value, list) or len(value) == 0:
+            return None
+        if not (
             isinstance(value[0], TIME)
             or isinstance(value[0], str)
             or isinstance(value[0], uuid_lib.UUID)
@@ -135,14 +143,14 @@ def __filter_to_text_list(value: FilterValues) -> Optional[base_pb2.TextArray]:
 
     @staticmethod
     def __filter_to_bool_list(value: FilterValues) -> Optional[base_pb2.BooleanArray]:
-        if not isinstance(value, list) or not isinstance(value[0], bool):
+        if not isinstance(value, list) or len(value) == 0 or not isinstance(value[0], bool):
             return None
 
         return base_pb2.BooleanArray(values=cast(List[bool], value))
 
     @staticmethod
     def __filter_to_float_list(value: FilterValues) -> Optional[base_pb2.NumberArray]:
-        if not isinstance(value, list) or not isinstance(value[0], float):
+        if not isinstance(value, list) or len(value) == 0 or not isinstance(value[0], float):
             return None
 
         return base_pb2.NumberArray(values=cast(List[float], value))
@@ -152,6 +160,7 @@ def __filter_to_int_list(value: FilterValues) -> Optional[base_pb2.IntArray]:
         # bool is a subclass of int in Python, so the check must ensure it's not a bool
         if (
             not isinstance(value, list)
+            or len(value) == 0
             or not isinstance(value[0], int)
             or isinstance(value[0], bool)
         ):
@@ -224,6 +233,12 @@ def __parse_filter(value: FilterValues) -> Dict[str, Any]:
         if isinstance(value, float):
             return {"valueNumber": value}
         if isinstance(value, list):
+            if len(value) == 0:
+                raise WeaviateInvalidInputError(
+                    "Filtering on empty lists is not supported by Weaviate. "
+                    "To filter by property length, use "
+                    "Filter.by_property('prop', length=True).equal(0)"
+                )
             if isinstance(value[0], str):
                 return {"valueTextArray": value}
             if isinstance(value[0], uuid_lib.UUID):