i will try out the caveat approach tomorrow too, i understand it's not recommended, but i'd like to see how it looks. our dataset is quite small so performance may not be an issue.