> > One other thing about our particular setup is that we use separate
> > schema's for all user data and the functions go in the public schema.
> > So before executing this function we issue something like 'set_path to
> > username,public'.
>
> Mph. Are you expecting the function to work for more than one such path
> value over the life of a connection? Maybe you need to do the selection
> part with an EXECUTE not only the update. As-is, the first execution
> will latch down which copy of recurbilling_transactions will be used
> for the selection, regardless of later changes in search_path.
>
Argh... I knew better. Yes i need to use EXECUTE because it is used
across multiple paths.
Chris