Received: by 10.52.89.73 with SMTP id bm9mr38294962vdb.3.1335869409651;
        Tue, 01 May 2012 03:50:09 -0700 (PDT)
X-BeenThere: h2-database@googlegroups.com
Received: by 10.52.88.105 with SMTP id bf9ls2182572vdb.8.gmail; Tue, 01 May
 2012 03:50:08 -0700 (PDT)
MIME-Version: 1.0
Received: by 10.52.33.47 with SMTP id o15mr138633vdi.9.1335869408246; Tue, 01
 May 2012 03:50:08 -0700 (PDT)
Authentication-Results: ls.google.com; spf=pass (google.com: domain of
 thomas.ege...@gmail.com designates internal as permitted sender)
 smtp.mail=thomas.ege...@gmail.com; dkim=pass
 header...@gmail.com
Received: by e15g2000vbt.googlegroups.com with HTTP; Tue, 1 May 2012 03:50:08
 -0700 (PDT)
Date: Tue, 1 May 2012 03:50:08 -0700 (PDT)
In-Reply-To: <48627d57-24ad-4100-8e9f-3f7cca58730b@w5g2000vbp.googlegroups.com>
References: <4dca46c4-60c3-41af-9155-f765a76976e9@d4g2000vbn.googlegroups.com>
 <8508563.120.1335433553775.JavaMail.geo-discussion-forums@yndr30>
 <5aad1906-3acf-4edc-bd9c-8ea59f843c30@e42g2000yqa.googlegroups.com>
 <17830687.1959.1335443452762.JavaMail.geo-discussion-forums@vbdx11>
 <863AD90A-BF2E-47ED-B457-56A174471A91@gmail.com> <CACcJavCrzD89RgDqv6PksV_xcX8oFhwXbh-05Rie5=wWPFVeqw@mail.gmail.com>
 <4fef1141-2b6a-43c6-8605-97d9a76af8f9@12g2000vba.googlegroups.com>
 <CAKpgiBbdTishR9ZXRHZZCHO2-1mQXv2v2ChnfQwq3qEo5oOSWQ@mail.gmail.com> <48627d57-24ad-4100-8e9f-3f7cca58730b@w5g2000vbp.googlegroups.com>
User-Agent: G2/1.0
X-HTTP-UserAgent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:11.0)
 Gecko/20100101 Firefox/11.0,gzip(gfe)
Message-ID: <e97a6207-3bd2-49d3-9b17-320277617657@e15g2000vbt.googlegroups.com>
Subject: Re: Huge performance problem when using OR. Much faster(500* +) when
 using two SELECT and UNION
From: Thomas Egense <thomas.ege...@gmail.com>
To: H2 Database <h2-database@googlegroups.com>
Content-Type: text/plain; charset=ISO-8859-1

I see no measurable improvements creating another index over both
columns (LEFT,RIGHT) btw. I already need index for LEFT and RIGHT
individually. Also you have to
consider the column order when creating and index that spawns over
several columns, depending on the data. Ie. have to most restrictive
first etc.
More indexes increases insert-time.

If my table had 1000M instead of 10M rows, maybe the index over
several columns will be faster.