Re: Odd join selectivity

From: Greg Rahn <>
Date: Thu, 6 Mar 2008 14:55:54 -0800
Message-ID: <>

What is the problem with this plan? The estimate of 54k is probably close enough to 273k not to cause a plan change.

The question of "Where is 1.0170e-04 coming from?" is probably not something that a support person would be able to answer. It probably takes a developer to look at the code, and it not something they are just going to do because you ask. If there is a problem with the plan, I'm sure they would help, but they are not really there to answer reverse engineering questions.

On 3/6/08, Charles Schultz <> wrote:
> I filed a case with Oracle Support and ended up with a carpet salesman.
> :-( I asked him specifically where this number is coming from and he threw
> metalink note 68992.1 at me. When I told him that 1/NDV does not match any
> of our indexes, he wants me to reset all my statistics back to default, run
> a few 10046 traces, download and install sqlt (the next version of
> traceanalzyer, apparently) and upload all the results. Sound familiar to
> anyone?


Greg Rahn

Received on Thu Mar 06 2008 - 16:55:54 CST

Original text of this message