"Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote in message
news:cd3pmm$ah5$1_at_titan.btinternet.com...
> The (main) reason why both queries show exactly the same
> logical I/O is that the resulting join is based on PK to PK,
> so the semi-join's strategy of stopping after one row is never
> relevant.
This just emphases how volatile state of the art in SQL optimization is.
This little trick - stopping after finding first match - was the only
advantage semijoin tuple iteration semantics had over nested loops. I was
told lately (but didn't verify) that this optimization has been added to NL
as well.
Received on Thu Jul 15 2004 - 15:58:14 CDT
