Re: spi/timetravel: unique constraint violation on UPDATE

Den 15. juli 2017 23:15, skrev Tom Lane:
> postgresql(at)get-experience(dot)com writes:
>> I'm having a problem with the timetravel extension. Following simple schema:
>> ...
>> What am I doing wrong here? According to the function comment in
>> timetravel.c it should:
>> a) set last-version row to NEW data; valid_from=now()
>> b) insert a new row with OLD data; valid_to=now() - at this point the
>> old valid_from is already supposed to be updated.
>
> I don't see anywhere in that comment that says "the old valid_from is
> already supposed to be updated". Given that the INSERT of the cloned row
> happens during the BEFORE UPDATE trigger, it would take a lot of magic
> for things to happen that way ;-).
Yeah, good point.

> Perhaps you could make your PK be on (id, valid_from, valid_to).
Doesn't really work because valid_to would change on UPDATE. I'd need to
update foreign relations with another trigger which would be very ugly.

I guess the only solution is to write my own time travel trigger
function and do update before insert...