Mailing-List: contact ojb-dev-help@db.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "OJB Developers List" <ojb-dev@db.apache.org>
Content-Type: text/plain;
  charset="us-ascii"
From: Oleg Nitz <on@ukr.net>
Reply-To: on@ukr.net
To: ojb-dev@db.apache.org
Subject: Batch retrieval 
Date: Mon, 25 Aug 2003 23:36:33 +0000
User-Agent: KMail/1.4.2
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Message-Id: <200308252336.33912.on@ukr.net>

Hi!

I have commited some changes to the OJB core, but if they are not in time=
,=20
they can be easily turned off via the 'autoPrefetch' field in=20
PersistenceBrokerImpl.java
The idea is to make retrieval more efficient via gathering them into batc=
hes=20
(less queries =3D> less network and database hits).
For the "broker" junit tests the number of SELECT queries is reduces by 9=
%,=20
for "odmg" junit tests - by 18%, but I expect much more gain in the real =
life=20
complicated object structures.
How does it works? During every getCollectionByQuery() call of=20
PersistenceBroker all objects that should be retrieved via relations with=
=20
auto-retrieve=3D"true" (without proxies and without QueryCustomizer),
are not retrieved immediately, but the tasks for their retrieval are=20
collected. They are gathered into groups of similar tasks and executed at=
 the=20
end of getCollectionByQuery().=20
Also this means that during retrieval of one object of class A which has =
1:N=20
relation to class B which has 1:N relation to class C and N:1 relation to=
=20
class D, all C's and and D's will be loaded by one query for each relatio=
n.
This algoritm uses Jackob's prefetching classes, but methods that modify=20
metadata are not used.=20

Regards,
 Oleg

---------------------------------------------------------------------
To unsubscribe, e-mail: ojb-dev-unsubscribe@db.apache.org
For additional commands, e-mail: ojb-dev-help@db.apache.org