From: Ananda Kumar
Date: August 17 2010 12:21pm
Subject: Re: Extremly slow Join with 'OR'
List-Archive: http://lists.mysql.com/mysql/222548
Message-Id: <AANLkTi=Taov+1Bn2vcgKWu_gLSQ7S=ex-xUfKqipY+7a@mail.gmail.com>
MIME-Version: 1.0
Content-Type: multipart/alternative; boundary=001485f78c0a263070048e03fecf

--001485f78c0a263070048e03fecf
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

use UNION ALL ..instead of UNION for better performance...

On Tue, Aug 17, 2010 at 5:01 PM, Johan De Meersman <vegivamp@stripped>wrot=
e:

> You may want to split of your or conditions into a separate query, and us=
e
> UNION.
>
> On Tue, Aug 17, 2010 at 11:22 AM, =D0=92=D0=BB=D0=B0=D0=B4 =D0=A0 <vuliad=
@gmail.com> wrote:
>
> > The main problem - if add in Join on `OR`-condition, select become
> > VERY slow. I realy
> > have to use this condition.
> >
> > -- --------------------------------------------------------
> > --
> > -- `tree_data`
> > --
> >
> > CREATE TABLE IF NOT EXISTS `tree_data` (
> >  `id` int(11) NOT NULL AUTO_INCREMENT,
> >  `pid` int(11) NOT NULL,
> >  PRIMARY KEY (`id`),
> > ) ENGINE=3DMyISAM  DEFAULT CHARSET=3Dcp1251 AUTO_INCREMENT=3D869 ;
> >
> > testiong on 800 rows
> >
> > Simple tree join query
> > First QUERY is:
> >
> > SELECT
> > `l0`.`id` AS  'l0id',
> > `l0`.`pid` AS  `l0pid` ,
> > `l1`.`id` AS  'l1id',
> > `l1`.`pid` AS  `l1pid` ,
> > `l2`.`id` AS  'l2id',
> > `l2`.`pid` AS  `l2pid` ,
> > `l3`.`id` AS  'l3id',
> > `l3`.`pid` AS  `l3pid` ,
> > `l4`.`id` AS  'l4id',
> > `l4`.`pid` AS  `l4pid`
> > FROM  `tree_data` AS  `l0`
> > LEFT JOIN `tree_data` as `l1` on (`l1`.`pid`=3D`l0`.`id`)
> > LEFT JOIN `tree_data` as `l2` on (`l2`.`pid`=3D`l1`.`id`)
> > LEFT JOIN `tree_data` as `l3` on (`l3`.`pid`=3D`l2`.`id`)
> > LEFT JOIN `tree_data` as `l4` on (`l4`.`pid`=3D`l3`.`id`)
> > WHERE  `l0`.`pid` =3D0
> >
> > id      select_type     table   type    possible_keys   key     key_len
> ref
> >     rows    Extra
> > 1       SIMPLE  l0      ref     pid     pid     4       const   4
> > 1       SIMPLE  l1      ref     pid     pid     4       test.l0.id
>  9
> > 1       SIMPLE  l2      ref     pid     pid     4       test.l1.id
>  9
> > 1       SIMPLE  l3      ref     pid     pid     4       test.l2.id
>  9
> > 1       SIMPLE  l4      ref     pid     pid     4       test.l3.id
>  9
> >
> > time execution is 0.0069.  result 207 rows
> >
> > looking at other query
> >
> > SELECT
> > `l0`.`id` AS  'l0id',
> > `l0`.`pid` AS  `l0pid` ,
> > `l1`.`id` AS  'l1id',
> > `l1`.`pid` AS  `l1pid` ,
> > `l2`.`id` AS  'l2id',
> > `l2`.`pid` AS  `l2pid` ,
> > `l3`.`id` AS  'l3id',
> > `l3`.`pid` AS  `l3pid` ,
> > `l4`.`id` AS  'l4id',
> > `l4`.`pid` AS  `l4pid`
> > FROM  `tree_data` AS  `l0`
> > LEFT JOIN `tree_data` as `l1` on (`l1`.`pid`=3D`l0`.`id` or `l1`.`pid`=
=3D100)
> > LEFT JOIN `tree_data` as `l2` on (`l2`.`pid`=3D`l1`.`id` or `l2`.`pid`=
=3D200)
> > LEFT JOIN `tree_data` as `l3` on (`l3`.`pid`=3D`l2`.`id` or `l3`.`pid`=
=3D300)
> > LEFT JOIN `tree_data` as `l4` on (`l4`.`pid`=3D`l3`.`id` or `l4`.`pid`=
=3D400)
> > WHERE  `l0`.`pid` =3D0
> >
> > And now execution time is 0.1455!!!
> > profiling gives: Sending data   0.137295
> > And analize:
> > id  select_type table   type  possible_keys     key     key_len ref
> > rows    Extr
> > 1       SIMPLE  l0      ref     pid     pid     4       const   4
> > 1       SIMPLE  l1      ALL     pid     NULL    NULL    NULL    800
> > 1       SIMPLE  l2      ALL     pid     NULL    NULL    NULL    800
> > 1       SIMPLE  l3      ALL     pid     NULL    NULL    NULL    800
> > 1       SIMPLE  l4      ALL     pid     NULL    NULL    NULL    800
> >
> > And what will happen if there will be 10000 rows?
> >
> > (In real i using this `or` condition for selecting to the tree nodes
> > for current element id and id can be id from other nodes.)
> >
>
>
>
> --
> Bier met grenadyn
> Is als mosterd by den wyn
> Sy die't drinkt, is eene kwezel
> Hy die't drinkt, is ras een ezel
>

--001485f78c0a263070048e03fecf--