Currently the category for  the builtin parser  is set to the old-style numeric warnings  ("msg_id") (W0xxx). I think it would be better to use the textual representation by default ( the "symbol"), as these are easily human-readable, while the numeric ones always require a lookup.