Hints (was: Index Tuning Features)

On Wed, Oct 11, 2006 at 03:08:42PM -0700, Ron Mayer wrote:
> Is one example is the table of addresses clustered by zip-code
> and indexes on State, City, County, etc?

No.

> Now I'm not saying that a more advanced statistics system
> couldn't one-day be written that sees these patterns in the
> data -- but it doesn't seem likely in the near term.  DBA-based
> hints could be a useful interim work-around.

Some others in the hints thread seem to be suggesting additional ways
of teaching the optimiser what to do.  _That_ seems to me to be a
good idea (but I don't think that qualifies as what people usually
think of as hints).  A sufficiently general system of hints sprinkled
on the SQL is a lot of work, and doesn't seem to me to be a whole lot
easier than working out how to make second-order relationship
discovery (of the sort you're talking about) cheaper and automatic. 
Certainly, there's plenty of statistics math kicking around that
allows one to discover such relationships, and they have the benefit
of not being by definition a way to work around the optimiser.

A

-- 
Andrew Sullivan  | ajs@crankycanuck.ca
Users never remark, "Wow, this software may be buggy and hard 
to use, but at least there is a lot of code underneath."    --Damien Katz