Re: boolean short-circuiting in plpgsql

Kev  writes:
> ...because the case should force it to only evaluate 'old' when TG_OP
> =  'UPDATE' and otherwise ('INSERT') skip through to 't'.  But this
> causes the same error on insert.  I suspect it's because the select
> query gets parameterized and at that point the 'old' is missing,
> before the case even gets to be parsed.

Got it in one.

> How do I get around this
> without having two 'perform' statements?

What you need is two nested IF statements.  The PERFORM in your example
is not relevant to the problem.

			regards, tom lane