Christian Leclerc wrote:
> Hello Tom,
>
> Thanks for your reply and advise. I understand in your email that the
> use of a sequence object will increase the performance of my trigger.
> Therefore I'm going to replace that in my code immediately. Anyway, my
> problem is not exactly the performances; it's more the insertion time
> growing. Indeed I don't understand why the insertion time grows
> "exponentially" with my single transaction. If I split my objects
> insertion into several transactions (instead of one), the problem seems
> to disappear. Would you see any reasons linked to the Postgres
> transactions explaining this insertion time growing ?
>
I think that the insertion time growing is related to the way in which
you were doing your inserts; using a sequence object will likely fix the
issue and then you will have the answer to your question.
Sean