At 12:49 AM 5/2/02 -0400, Tom Lane wrote:
>Lincoln Yeoh <lyeoh@pop.jaring.my> writes:
> > But does Postgresql visit the older tuples first moving to the newer ones,
> > or the newer ones first?
>
>It's going to visit them *all*. Reordering won't improve the
>performance.
Ack! I thought it went through them till the first valid tuple and was just
going the wrong way.
>FWIW I think that with the present implementation of btree, the newer
>tuples actually will be visited first --- when inserting a duplicate
>key, the new entry will be inserted to the left of the equal key(s)
>already present. But it doesn't matter. The only way to speed this
>up is to eliminate some of the visitings, which requires keeping more
>info in the index than we presently do.
OK I'm starting to get it :). Will the index behaviour be changed soon?
Hmm, then what are the row tuple forward links for? Why forward?
Regards,
Link.
