see the yesterday's thread about DISTINCT ON (non-standard Postgres feature)
> I have a problem, which I'm not able to solve with a simple query :
>
> I need a resultset with distinct id's, but the max val2 of each id.
> I tried to group by id, but though I need the pk in my resultset
> I have to group it too, which "destroys" the group of val2.