Re: Help with view, pushing predicate, and 9i

  • From: "Jonathan Lewis" <jonathan@xxxxxxxxxxxxxxxxxx>
  • To: <oracle-l@xxxxxxxxxxxxx>
  • Date: Thu, 24 Jun 2004 23:47:17 +0100

Have you tried putting a pair of global
hints in the outer query to see if that
makes any difference.

What are the index definitions for
the two indexes hinted ?

Regards

Jonathan Lewis

http://www.jlcomp.demon.co.uk

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

http://www.jlcomp.demon.co.uk/seminar.html
Optimising Oracle Seminar - schedule updated May 1st


----- Original Message ----- 
From: "Bobak, Mark" <Mark.Bobak@xxxxxxxxxxxxxxx>
To: <oracle-l@xxxxxxxxxxxxx>
Sent: Thursday, June 24, 2004 9:38 PM
Subject: Help with view, pushing predicate, and 9i


Hi,

We have a view which is usually accessed like:
select * from vu_xan_comp_agrmt_single where doc_id =3D :some_doc_id;

In 8i (8.1.7.4 on Solaris 8), the 'doc_id=3D ' predicate is pushed into
the view, and it works fine.

We are in the process of converting this app to 9i (9.2.0.5 on Solaris
8), and in 9i, the predicate is not pushed.  This results in a full
table scan on a large table, and, abysmal performance.

This should NOT be a case where the predicate can't be pushed, due to
rownum or analytical or aggregate function, etc.  It works in 8i, so why
not in 9i?  Is this an optimizer bug?  (_push_join_union_view and
_push_join_predicate are both TRUE.)

Any thoughts or suggestions are welcome.

Thanks,

-Mark

PS  Here's the beastie:
CREATE OR REPLACE FORCE VIEW ADDS.VU_XAN_COMP_AGRMT_SINGLE
(DOC_ID, VLAD_ID, CPAG_TEXT, CPAG_TEXT_LENGTH, CPAG_DAY_EFFECTIVE,=20
 CPAG_CALC_DATE, CPAG_FORMAT_MASK)
AS=20
select /*+ index(cpag cpag_indx_pr02) */ distinct doc_id, 0 vlad_id,
  first_value(cpag_text) over (partition by doc_id
  order by cpag_day_effective desc)  cpag_text,
  first_value(length(cpag_text)) over (partition by doc_id
                order by cpag_day_effective desc)  cpag_text_length,
  first_value(cpag_day_effective) over (partition by doc_id
  order by cpag_day_effective desc)  cpag_day_effective,
  first_value(cpag_calc_date) over (partition by doc_id
  order by cpag_day_effective desc)  cpag_calc_date,
  first_value(cpag_format_mask) over (partition by doc_id
                order by cpag_day_effective desc)  cpag_format_mask
 from compressed_agreements cpag
 WHERE cpag.CPAG_DAY_EFFECTIVE <=3D
 DECODE(SIGN(TRUNC(SYSDATE)-cpag.cpag_calc_date),-1,0,
 TRUNC(SYSDATE)-cpag.cpag_calc_date + NVL(cpag.CPAG_DAY_EFFECTIVE,0))
 UNION
 select /*+ index(xpag cpag_indx_pr02) */ distinct doc_id,vlad_id,
  first_value(cpag_text) over (partition by doc_id,vlad_id
  order by vlad_id,cpag_day_effective desc)  cpag_text,
  first_value(length(cpag_text)) over (partition by doc_id,vlad_id
                order by vlad_id,cpag_day_effective desc)
cpag_text_length,
  first_value(cpag_day_effective) over (partition by doc_id,vlad_id
  order by vlad_id,cpag_day_effective desc)  cpag_day_effective,
  first_value(cpag_calc_date) over (partition by doc_id,vlad_id
  order by vlad_id,cpag_day_effective desc)  cpag_calc_date,
  first_value(cpag_format_mask) over (partition by doc_id,vlad_id
                order by vlad_id,cpag_day_effective desc)
cpag_format_mask
 from xan_compressed_agreements xcpag
 WHERE xcpag.CPAG_DAY_EFFECTIVE <=3D
 DECODE(SIGN(TRUNC(SYSDATE)-xcpag.cpag_calc_date),-1,0,
 TRUNC(SYSDATE)-xcpag.cpag_calc_date + NVL(xcpag.CPAG_DAY_EFFECTIVE,0));




Mark J. Bobak
Oracle DBA
ProQuest Company
Ann Arbor, MI
"Post Hoc Ergo Propter Hoc"
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at //www.freelists.org/archives/oracle-l/
FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------


----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at //www.freelists.org/archives/oracle-l/
FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------

Other related posts: