Thread: Problem with "on conflict"

Hi all. I have some really strange behavior with postgres and 'on conflict'.
I had some error today at a customer and i figured out, the problem is triggered by postgres doing an insert on a table with wrong data (not even provided), when it should insert the data.

After some playing around with the query, the error just went away, magically ….
Here you can see, that the insert is updating stuff:

There also is a trigger on the db:

CREATE OR REPLACE FUNCTION fn_hut_bundle_create_id() returns TRIGGER AS $$
begin
    if new.bundle_id is null or new.bundle_id='' then
        NEW.bundle_id = concat('HB-', nextval('seq_hut_bundle_id'));
    end if;

    return NEW;
END;
$$ LANGUAGE plpgsql;
create trigger tr_hut_bundle_id before insert on hut_bundle for each row EXECUTE FUNCTION fn_hut_bundle_create_id();





now working:


Can someone give me some hint, what the problem could be? After my „playing“ the error can not be reproduced any longer, and the statement is inserting stuff like it should.
i played with null and '' as a primary key(bundle_id)

Postgres is:
PostgreSQL 14.1 (Debian 14.1-1.pgdg110+1) on x86_64-pc-linux-gnu, compiled by gcc (Debian 10.2.1-6) 10.2.1 20210110, 64-bit

Any help would be very welcome ....

Thanks

> On Mar 26, 2025, at 12:15 PM, Udo Polder <udo.polder@gmail.com> wrote:
>
> Hi all. I have some really strange behavior with postgres and 'on conflict'.
> I had some error today at a customer and i figured out, the problem is triggered by postgres doing an insert on a
tablewith wrong data (not even provided), when it should insert the data. 
>
> After some playing around with the query, the error just went away, magically ….
> Here you can see, that the insert is updating stuff:
>
> There also is a trigger on the db:
>
> CREATE OR REPLACE FUNCTION fn_hut_bundle_create_id() returns TRIGGER AS $$
> begin
>     if new.bundle_id is null or new.bundle_id='' then
>         NEW.bundle_id = concat('HB-', nextval('seq_hut_bundle_id'));
>     end if;
>
>     return NEW;
> END;
> $$ LANGUAGE plpgsql;
> create trigger tr_hut_bundle_id before insert on hut_bundle for each row EXECUTE FUNCTION fn_hut_bundle_create_id();
>
> <ti600gsoOaUj04Ng.png>
>
>
>
> now working:
> <U4iCI1t7CX28e3oM.png>
>
> Can someone give me some hint, what the problem could be? After my „playing“ the error can not be reproduced any
longer,and the statement is inserting stuff like it should. 
> i played with null and '' as a primary key(bundle_id)
>
> Postgres is:
> PostgreSQL 14.1 (Debian 14.1-1.pgdg110+1) on x86_64-pc-linux-gnu, compiled by gcc (Debian 10.2.1-6) 10.2.1 20210110,
64-bit
>
> Any help would be very welcome ....
>
> Thanks
>

To me it looks like the record already existed and the on conflict update occurred.  Given that hut_id is not updated
onconflict; the old value remained.  Was the sequence recycled or was 'HB-32’ created manually beforehand? 

Udo Polder <udo.polder@gmail.com> writes:
> I had some error today at a customer and i figured out, the problem is 
> triggered by postgres doing an insert on a table with wrong data (not 
> even provided), when it should insert the data.
> After some playing around with the query, the error just went away, 
> magically ….

Hmmm...

> Postgres is:
> PostgreSQL 14.1 (Debian 14.1-1.pgdg110+1) on x86_64-pc-linux-gnu, 
> compiled by gcc (Debian 10.2.1-6) 10.2.1 20210110, 64-bit

If it's really 14.1, that's quite ancient and full of known bugs.
The first thing you ought to do is update to latest minor release,
14.17.  I did not look at the release notes to see if any fixes
overlap this area, but it'd be silly to spend a lot of time
tracking down a problem only to find it's already fixed.

            regards, tom lane