Re: BUG #3847: plpython trigger caches table structure - doesn't see new / changed columns

The trigger function does not recognize the "test4" column the second time
it is added - the update throws an error.

On Jan 1, 2008 11:17 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> "Mark Reid" <reid(dot)write(at)gmail(dot)com> writes:
> > If a column is added, dropped, then re-added (all within a transaction),
> a
> > plpython trigger function loses track of the column and throws an error
> when
> > trying to access it. Here is the best minimal test case I could come up
> > with:
>
> The cases you are saying work and don't work are exactly the same:
>
> > -- This works
> > alter table clarence add column test4 varchar;
> > update clarence set test4=12 where pick_id=1454;
> > alter table clarence drop column test4;
>
> > -- This does not work
> > alter table clarence add column test4 varchar;
> > update clarence set test4=12 where pick_id=1454; -- this creates a
> > problem... plpgsql seems to work fine.
> > alter table clarence drop column test4;
>
> Please be clearer.
>
> regards, tom lane
>