> Why did you use row_number instead of rank?
Because I assumed he only wanted one row in the event of ties.
Hmmm, although with that schema, there won't be ties. So it's pretty
much arbitrary then.
> I am now curious how the speed compares though. I still think the
> DISTINCT ON will be faster, but it would be a great surprise.
Hopefully we'll find out! The windowing functions are usually much
faster for me. I think in 9.0 or 9.1 someone replumbed DISTINCT ON to
use a bunch of the window function internals, at which point it'll cease
to matter.
--
-- Josh Berkus
PostgreSQL Experts Inc.
http://www.pgexperts.com