The optimizer will have to remember this best plan even when it
considers a different join order for the outer query, because it's
possible that the subquery's best plan for that optimization will be
the best one overall.
It occurs to me that the code structure already supports this. Every
time the optimizer figures out the best access path for an
optimizable in the context of a join order it's considering, it tells
the optimizable to remember that access path as the "best," and when
it finds a complete plan that's the best so far, it tells the
optimizable to remember it as "truly the best." When optimizing an
outer query containing a table subquery, the "best" and "truly the
best" access paths for the subquery should be the complete plan for
the subquery from the current (or latest) optimization of it.
- Jeff Lichtman
[EMAIL PROTECTED]
Check out Swazoo Koolak's Web Jukebox at
http://swazoo.com/
