I need to steam incoming data in Protobuf to a corresponding table in Postgres. I deserialized it using the standard Protobuf package and then do executemany to save the data. My CPU load is high, and I suspect it's due to the type conversions from the de-serialized Protobuf data (which is not Python object, is far as I understand) to Postgres. Is there a canonical way of dealing with this? I assume it should be something on the line of #814.