Comments on: Constraints, Inserts and bind http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/ Just another Oracle weblog Wed, 22 May 2013 12:40:08 +0000 hourly 1 http://wordpress.com/ By: latch: row cache objects « Oracle http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-44670 Wed, 25 Jan 2012 22:07:08 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-44670 [...] As usual my master comes to the rescue with this note – Constraints, Inserts and bind [...]

]]> By: Jonathan Lewis http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-13785 Thu, 28 Jun 2007 10:06:42 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-13785 Robert, since 9i the optimizer has been able to combine existing predicates with constraints to generate new predicates. I suspect you have a constraint on column n1 that is being read each time you optimise a new query. The scale of activity is much less for queries, though, as it only applies to the where clauses, not to every column in the table.

]]> By: Robert Winter http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-13769 Thu, 28 Jun 2007 02:30:48 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-13769 I tried a similar test with SELECT and saw the same thing going on with cdef$. I can understand needing the constraint during an insert, but why does Oracle need it during a query? I’m running 10.2.0.3 on Solaris SPARC.
declare
        n1 number;
        v1 varchar2(10);
        v_cmd varchar2(256);
begin
        for i in 1..1000 loop
            v_cmd := ‘select n1, v1 from t1 where n1 = ‘ || i;
                execute immediate v_cmd into n1, v1;
        end loop;
end;
/

]]> By: Jason Bucata http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-573 Mon, 18 Dec 2006 15:42:34 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-573 I generally agree that it’s a worthwhile use of memory. My current job has me thinking a lot about the shared pool, since we’ve got a lot of shared pool pressure, so that’s just something I’ve started glomming onto…

]]> By: Syed Jaffar Hussain http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-493 Fri, 15 Dec 2006 19:01:01 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-493 Thanks for clarification. I got it.

Jaffar

]]> By: Jonathan Lewis http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-491 Fri, 15 Dec 2006 18:30:47 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-491 Jaffar, the long column in question is the definition column of cdef$. It has nothing to do with the type of the column in the user’s table.

]]> By: Syed Jaffar Hussain http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-489 Fri, 15 Dec 2006 18:22:44 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-489 Jonathan,

A small confusion in your example, you define column v1 datatype as varchar, but, you say,

>>Oracle doesn’t keep long columns cached in the dictionary cache (rowcache)

Is this same with column have other datatype than long?

Jaffar

]]> By: globelovers http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-487 Fri, 15 Dec 2006 18:00:58 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-487 support!

]]> By: Jonathan Lewis http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-466 Fri, 15 Dec 2006 07:37:55 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-466 Jason, Good question.

create table t1 (n1 number check (n1 = 1), …
for 255 columns.
insert /*+ tagged */ into t1 values (1,1,1…)
for 255 columns.
Track down the SQL statement in query v$db_object_cache to check the shared memory:

select sharable_mem, name
from v$db_object_cache
where name like ‘%tagged%’
and name not like ‘%db_object_cache%’
;

A test with 255 constraints in place used 170KB of memory, repeated without constraints the demand was 80KB.

Worth knowing about the extra memory requirements – on the other hand I’d rather use the memory than allow the data to be wrong.

]]> By: Jason Bucata http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-460 Thu, 14 Dec 2006 23:03:53 +0000 http://jonathanlewis.wordpress.com/2006/12/14/constraints-inserts-and-bind/#comment-460 So does that imply that the parsed version of the SQL statement has keep the information about all the constraints that that statement could possibly trip–if it doesn’t look at cdef$ again each time it’s run?

For a table with a lot of constraints, couldn’t that get big?

]]>