We have good coverage for the happy case on numbparse but we do not check for errors nicely - we could improve here.