fix: restrict input/search type for vector fields (#2025)

See also: #2018, #2004, #2016

---------

Signed-off-by: yangxuan <[email protected]>

pymilvus/client/entity_helper.py

@@ -291,7 +291,9 @@ def entity_to_array_arr(entity: List[Any], field_info: Any):
     return convert_to_array_arr(entity.get("values", []), field_info)
 
 
-def pack_field_value_to_field_data(field_value: Any, field_data: Any, field_info: Any):
+def pack_field_value_to_field_data(
+    field_value: Any, field_data: schema_types.FieldData, field_info: Any
+):
     field_type = field_data.type
     if field_type == DataType.BOOL:
         field_data.scalars.bool_data.data.append(field_value)
@@ -304,26 +306,51 @@ def pack_field_value_to_field_data(field_value: Any, field_data: Any, field_info
     elif field_type == DataType.DOUBLE:
         field_data.scalars.double_data.data.append(field_value)
     elif field_type == DataType.FLOAT_VECTOR:
-        field_data.vectors.dim = len(field_value)
-        field_data.vectors.float_vector.data.extend(field_value)
+        f_value = field_value
+        if isinstance(field_value, np.ndarray):
+            if field_value.dtype not in ("float32", "float64"):
+                raise ParamError(
+                    message="invalid input for float32 vector, expect np.ndarray with dtype=float32"
+                )
+            f_value = field_value.view(np.float32).tolist()
+
+        field_data.vectors.dim = len(f_value)
+        field_data.vectors.float_vector.data.extend(f_value)
+
     elif field_type == DataType.BINARY_VECTOR:
         field_data.vectors.dim = len(field_value) * 8
         field_data.vectors.binary_vector += bytes(field_value)
+
     elif field_type == DataType.FLOAT16_VECTOR:
-        v_bytes = (
-            bytes(field_value)
-            if not isinstance(field_value, np.ndarray)
-            else field_value.view(np.uint8).tobytes()
-        )
+        if isinstance(field_value, bytes):
+            v_bytes = field_value
+        elif isinstance(field_value, np.ndarray):
+            if field_value.dtype != "float16":
+                raise ParamError(
+                    message="invalid input for float16 vector, expect np.ndarray with dtype=float16"
+                )
+            v_bytes = field_value.view(np.uint8).tobytes()
+        else:
+            raise ParamError(
+                message="invalid input type for float16 vector, expect np.ndarray with dtype=float16"
+            )
 
         field_data.vectors.dim = len(v_bytes) // 2
         field_data.vectors.float16_vector += v_bytes
+
     elif field_type == DataType.BFLOAT16_VECTOR:
-        v_bytes = (
-            bytes(field_value)
-            if not isinstance(field_value, np.ndarray)
-            else field_value.view(np.uint8).tobytes()
-        )
+        if isinstance(field_value, bytes):
+            v_bytes = field_value
+        elif isinstance(field_value, np.ndarray):
+            if field_value.dtype != "bfloat16":
+                raise ParamError(
+                    message="invalid input for bfloat16 vector, expect np.ndarray with dtype=bfloat16"
+                )
+            v_bytes = field_value.view(np.uint8).tobytes()
+        else:
+            raise ParamError(
+                message="invalid input type for bfloat16 vector, expect np.ndarray with dtype=bfloat16"
+            )
 
         field_data.vectors.dim = len(v_bytes) // 2
         field_data.vectors.bfloat16_vector += v_bytes

pymilvus/client/grpc_handler.py

@@ -475,46 +475,49 @@ def _get_info(self, collection_name: str, timeout: Optional[float] = None, **kwa
 
         return fields_info, enable_dynamic
 
-    def _prepare_row_insert_request(
+    @retry_on_rpc_failure()
+    def insert_rows(
         self,
         collection_name: str,
-        entity_rows: List,
+        entities: Union[Dict, List[Dict]],
         partition_name: Optional[str] = None,
+        schema: Optional[dict] = None,
         timeout: Optional[float] = None,
         **kwargs,
     ):
-        if not isinstance(entity_rows, list):
-            raise ParamError(message="None rows, please provide valid row data.")
-
-        fields_info, enable_dynamic = self._get_info(collection_name, timeout, **kwargs)
-        return Prepare.row_insert_param(
-            collection_name,
-            entity_rows,
-            partition_name,
-            fields_info,
-            enable_dynamic=enable_dynamic,
+        request = self._prepare_row_insert_request(
+            collection_name, entities, partition_name, timeout, **kwargs
         )
+        resp = self._stub.Insert(request=request, timeout=timeout)
+        check_status(resp.status)
+        ts_utils.update_collection_ts(collection_name, resp.timestamp)
+        return MutationResult(resp)
 
-    @retry_on_rpc_failure()
-    def insert_rows(
+    def _prepare_row_insert_request(
         self,
         collection_name: str,
-        entities: List,
+        entity_rows: Union[List[Dict], Dict],
         partition_name: Optional[str] = None,
+        schema: Optional[dict] = None,
         timeout: Optional[float] = None,
         **kwargs,
     ):
-        if isinstance(entities, dict):
-            entities = [entities]
-        request = self._prepare_row_insert_request(
-            collection_name, entities, partition_name, timeout, **kwargs
+        if isinstance(entity_rows, dict):
+            entity_rows = [entity_rows]
+
+        if not isinstance(schema, dict):
+            schema = self.describe_collection(collection_name, timeout=timeout)
+
+        fields_info = schema.get("fields")
+        enable_dynamic = schema.get("enable_dynamic_field", False)
+
+        return Prepare.row_insert_param(
+            collection_name,
+            entity_rows,
+            partition_name,
+            fields_info,
+            enable_dynamic=enable_dynamic,
         )
-        rf = self._stub.Insert.future(request, timeout=timeout)
-        response = rf.result()
-        check_status(response.status)
-        m = MutationResult(response)
-        ts_utils.update_collection_ts(collection_name, m.timestamp)
-        return m
 
     def _prepare_batch_insert_request(
         self,
@@ -1376,7 +1379,7 @@ def _wait_for_flushed(
             end = time.time()
             if timeout is not None and end - start > timeout:
                 raise MilvusException(
-                    message=f"wait for flush timeout, collection: {collection_name}"
+                    message=f"wait for flush timeout, collection: {collection_name}, flusht_ts: {flush_ts}"
                 )
 
             if not flush_ret:

pymilvus/client/prepare.py

@@ -364,12 +364,10 @@ def _parse_row_request(
             field["name"]: field for field in fields_info if not field.get("auto_id", False)
         }
 
-        meta_field = (
-            schema_types.FieldData(is_dynamic=True, type=DataType.JSON) if enable_dynamic else None
-        )
-        if meta_field is not None:
-            field_info_map[meta_field.field_name] = meta_field
-            fields_data[meta_field.field_name] = meta_field
+        if enable_dynamic:
+            d_field = schema_types.FieldData(is_dynamic=True, type=DataType.JSON)
+            fields_data[d_field.field_name] = d_field
+            field_info_map[d_field.field_name] = d_field
 
         try:
             for entity in entities:
@@ -390,7 +388,7 @@ def _parse_row_request(
 
                 if enable_dynamic:
                     json_value = entity_helper.convert_to_json(json_dict)
-                    meta_field.scalars.json_data.data.append(json_value)
+                    d_field.scalars.json_data.data.append(json_value)
 
         except (TypeError, ValueError) as e:
             raise DataNotMatchException(message=ExceptionsMessage.DataTypeInconsistent) from e
@@ -400,7 +398,7 @@ def _parse_row_request(
         )
 
         if enable_dynamic:
-            request.fields_data.append(meta_field)
+            request.fields_data.append(d_field)
 
         _, _, auto_id_loc = traverse_rows_info(fields_info, entities)
         if auto_id_loc is not None:
@@ -418,16 +416,18 @@ def row_insert_param(
         collection_name: str,
         entities: List,
         partition_name: str,
-        fields_info: Any,
+        fields_info: Dict,
         enable_dynamic: bool = False,
     ):
         if not fields_info:
             raise ParamError(message="Missing collection meta to validate entities")
 
         # insert_request.hash_keys won't be filled in client.
-        tag = partition_name if isinstance(partition_name, str) else ""
+        p_name = partition_name if isinstance(partition_name, str) else ""
         request = milvus_types.InsertRequest(
-            collection_name=collection_name, partition_name=tag, num_rows=len(entities)
+            collection_name=collection_name,
+            partition_name=p_name,
+            num_rows=len(entities),
         )
 
         return cls._parse_row_request(request, fields_info, enable_dynamic, entities)
@@ -445,9 +445,11 @@ def row_upsert_param(
             raise ParamError(message="Missing collection meta to validate entities")
 
         # upsert_request.hash_keys won't be filled in client.
-        tag = partition_name if isinstance(partition_name, str) else ""
+        p_name = partition_name if isinstance(partition_name, str) else ""
         request = milvus_types.UpsertRequest(
-            collection_name=collection_name, partition_name=tag, num_rows=len(entities)
+            collection_name=collection_name,
+            partition_name=p_name,
+            num_rows=len(entities),
         )
 
         return cls._parse_row_request(request, fields_info, enable_dynamic, entities)
@@ -469,7 +471,7 @@ def _pre_batch_check(
         if not fields_info:
             raise ParamError(message="Missing collection meta to validate entities")
 
-        location, primary_key_loc, auto_id_loc = traverse_info(fields_info, entities)
+        location, primary_key_loc, auto_id_loc = traverse_info(fields_info)
 
         # though impossible from sdk
         if primary_key_loc is None:
@@ -583,16 +585,20 @@ def _prepare_placeholder_str(cls, data: Any):
 
         elif isinstance(data[0], np.ndarray):
             dtype = data[0].dtype
-            pl_values = (array.tobytes() for array in data)
 
             if dtype == "bfloat16":
                 pl_type = PlaceholderType.BFLOAT16_VECTOR
+                pl_values = (array.tobytes() for array in data)
             elif dtype == "float16":
                 pl_type = PlaceholderType.FLOAT16_VECTOR
-            elif dtype == "float32":
+                pl_values = (array.tobytes() for array in data)
+            elif dtype in ("float32", "float64"):
                 pl_type = PlaceholderType.FloatVector
+                pl_values = (blob.vector_float_to_bytes(entity) for entity in data)
+
             elif dtype == "byte":
                 pl_type = PlaceholderType.BinaryVector
+                pl_values = data
 
             else:
                 err_msg = f"unsupported data type: {dtype}"

pymilvus/client/utils.py

@@ -250,7 +250,7 @@ def traverse_rows_info(fields_info: Any, entities: List):
     return location, primary_key_loc, auto_id_loc
 
 
-def traverse_info(fields_info: Any, entities: List):
+def traverse_info(fields_info: Any):
     location, primary_key_loc, auto_id_loc = {}, None, None
     for i, field in enumerate(fields_info):
         if field.get("is_primary", False):
@@ -259,58 +259,7 @@ def traverse_info(fields_info: Any, entities: List):
         if field.get("auto_id", False):
             auto_id_loc = i
             continue
-
-        match_flag = False
-        field_name = field["name"]
-        field_type = field["type"]
-
-        for entity in entities:
-            entity_name, entity_type = entity["name"], entity["type"]
-
-            if field_name == entity_name:
-                if field_type != entity_type:
-                    raise ParamError(
-                        message=f"Collection field type is {field_type}"
-                        f", but entities field type is {entity_type}"
-                    )
-
-                entity_dim, field_dim = 0, 0
-                if entity_type in [
-                    DataType.FLOAT_VECTOR,
-                    DataType.BINARY_VECTOR,
-                    DataType.BFLOAT16_VECTOR,
-                    DataType.FLOAT16_VECTOR,
-                ]:
-                    field_dim = field["params"]["dim"]
-                    entity_dim = len(entity["values"][0])
-
-                if entity_type in [DataType.FLOAT_VECTOR] and entity_dim != field_dim:
-                    raise ParamError(
-                        message=f"Collection field dim is {field_dim}"
-                        f", but entities field dim is {entity_dim}"
-                    )
-
-                if entity_type in [DataType.BINARY_VECTOR] and entity_dim * 8 != field_dim:
-                    raise ParamError(
-                        message=f"Collection field dim is {field_dim}"
-                        f", but entities field dim is {entity_dim * 8}"
-                    )
-
-                if (
-                    entity_type in [DataType.BFLOAT16_VECTOR, DataType.FLOAT16_VECTOR]
-                    and int(entity_dim // 2) != field_dim
-                ):
-                    raise ParamError(
-                        message=f"Collection field dim is {field_dim}"
-                        f", but entities field dim is {int(entity_dim // 2)}"
-                    )
-
-                location[field["name"]] = i
-                match_flag = True
-                break
-
-        if not match_flag:
-            raise ParamError(message=f"Field {field['name']} don't match in entities")
+        location[field["name"]] = i
 
     return location, primary_key_loc, auto_id_loc