Re: AW: [HACKERS] OR clause status report

Vadim Mikheev wrote:
>
> Andreas Zeugswetter wrote:
> >
> > Vadim wrote:
> > >Bruce Momjian wrote:
> > >>
> > >> I have succeeded in making OR clauses use indexes.  I have not dealt
> > >> with some of the more complex issues like x=3 or y=5 as Vadim mentioned,
> > >> but at least it works.
> > >>
> > >>         test=> select * from test where x=102532 or x=102533;
> > >
> > >But did you care about the case when two indices (on test(x) and
> > >on test(y)) exist ?
> >
> > Do you mean using two indices in one access plan,
>               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> This.

Actually, I meant NOT USING indices if both exist, because of
currently there is no check that tuples returned by second index
were not returned by first.

> > or the decision which index to use ?

Either both indices should be used or no one...

Also, Bruce, did you test the case (x = 5 or x > 4) ?
What about (x = 5 or x = 5)? - I'm not sure does cnfify()
get rid of duplicates or not...

Vadim