Re: Mysterious FILTER operation ;)

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Mon, 3 Jan 2005 21:05:30 -0000
Message-ID: <021101c4f1d7$f5086fd0$6702a8c0@Primary>

As far as the logical I/O goes, I would guess that the execution plan you are seeing from explain plan is NOT the plan that is actually occurring for CBO - bind variable peeking is probably switching you to a full tablescan.

The filter operation is probably similar to the type of filter that used to precede the parts of partition views which were not
executed - in other words a constant
test that could pre-empt the execution
of the child portions of the plan. If your specific plan has to be generated once with bind variables, then there has to be a mechanism that stops the real work being done when the first variable is larger than the second variable at run-time.

Regards

Jonathan Lewis

http://www.jlcomp.demon.co.uk/faq/ind_faq.html The Co-operative Oracle Users' FAQ

http://www.jlcomp.demon.co.uk/seminar.html Public Appearances - schedule updated Dec 23rd 2004

Original Message ----- From: <J.Velikanovs_at_alise.lv> To: <oracle-l_at_freelists.org> Sent: Monday, January 03, 2005 6:19 PM Subject: Mysterious FILTER operation ;)

Win2000 9.2.0.6 (tested on 9.2.0.4/SPARC Solaris as well) FULL TEST text see at the end of letter. I just trying to understand what FILTER operation doing in case described below.
I have ran the same SQL two times (with and without stats, CBO/RBO)

--

http://www.freelists.org/webpage/oracle-l Received on Mon Jan 03 2005 - 15:05:30 CST