Re: NEW.* and OLD.* inside trigger function don't seem to contain recently added columns

Hey Tom,

Thank you for taking the time to look at the issue. It appears the ROW() notation was indeed unnecessary in this situation. I tried the methods you described, and they work perfectly.

I consider the primary issue solved. I assume your team will consider whether the original behaviour is considered a bug and if it's worth fixing or not.

Thanks again
Janne Annala
________________________________
From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Sent: Saturday, 23 March 2024 2.41
To: Janne Annala <janne(dot)annala(at)forenom(dot)com>
Cc: pgsql-bugs(at)lists(dot)postgresql(dot)org <pgsql-bugs(at)lists(dot)postgresql(dot)org>
Subject: Re: NEW.* and OLD.* inside trigger function don't seem to contain recently added columns

[You don't often get email from tgl(at)sss(dot)pgh(dot)pa(dot)us(dot) Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

Janne Annala <janne(dot)annala(at)forenom(dot)com> writes:
> Here's the minimum case to reproduce the issue:

> CREATE FUNCTION x()
> RETURNS trigger
> LANGUAGE plpgsql
> AS $function$
> BEGIN
> IF ROW(NEW.*) IS DISTINCT FROM ROW(OLD.*) THEN

Try dropping the ROW() bit, that is just

IF NEW.* IS DISTINCT FROM OLD.* THEN

or even

IF NEW IS DISTINCT FROM OLD THEN

I think what is happening is that the ROW() notation is getting
expanded at parse time to

ROW(NEW.id, NEW.old_column, NEW.updated)

and then there's no dependency on the original rowtype that would
lead to that expansion getting reconsidered. Arguably that's a
bug, but it's not clear to me what the consequences of changing
that behavior would be.

regards, tom lane