Mailing-List: contact derby-dev-help@db.apache.org; run by ezmlm
Precedence: bulk
Reply-To: <derby-dev@db.apache.org>
Date: Tue, 23 Oct 2012 14:59:13 +0000 (UTC)
From: "Knut Anders Hatlen (JIRA)" <jira@apache.org>
To: derby-dev@db.apache.org
Message-ID: <44224429.16149.1351004353405.JavaMail.jiratomcat@arcas>
In-Reply-To: <1911693874.9962.1350917652313.JavaMail.jiratomcat@arcas>
Subject: [jira] [Commented] (DERBY-5959) Territory-based collation is not
 robust against changes in the collation rules
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable


    [ https://issues.apache.org/jira/browse/DERBY-5959?page=3Dcom.atlassian=
.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=3D1348=
2379#comment-13482379 ]=20

Knut Anders Hatlen commented on DERBY-5959:
-------------------------------------------

Since Derby only supports RuleBasedCollator, one solution could be to store=
 the rules (RuleBasedCollator.getRules()) at database-creation time, and us=
e that value to reconstruct the collator on subsequent boots (using the Rul=
eBasedCollator(String) constructor). This would also allow use of a databas=
e with territory-based collation on a platform that doesn't support the spe=
cific locale, as long as it was available on the platform where the databas=
e was created.

The downside of such an approach is that the database users won't automatic=
ally get the benefit of fixes in the collation rules (such as the above men=
tioned http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=3D6755060 bug) wh=
en upgrading the JRE.

Another approach may be to instruct users to drop and recreate all indexes =
that contain CHAR/VARCHAR columns when switching to another JRE, but that m=
ay be impractical. Also, if the user fails to do this, inconsistencies may =
sneak into the database.
               =20
> Territory-based collation is not robust against changes in the collation =
rules
> -------------------------------------------------------------------------=
-----
>
>                 Key: DERBY-5959
>                 URL: https://issues.apache.org/jira/browse/DERBY-5959
>             Project: Derby
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 10.10.0.0
>            Reporter: Knut Anders Hatlen
>
> When accessing a database with territory-based collation, Derby will use =
the collation rules of the collator returned by Collator.getInstance(databa=
seLocale). However, there is no guarantee that those rules are consistent a=
cross different JVM vendors and versions. This means that the ordering coul=
d vary, and inconsistencies could sneak into the indexes.
> One example is that Oracle's JDK changed the collation rules for Turkish =
between Java 5 and Java 6, so if you run the following script
> connect 'jdbc:derby:memory:db;territory=3Dtr_TR;collation=3DTERRITORY_BAS=
ED;create=3Dtrue';
> create table t(c char(2));
> insert into t values '=C4=B1a', 'Ia', 'ia', '=C4=B0a', '=C4=B1b', 'Ib', '=
ib', '=C4=B0b';
> select * from t order by c;
> you'll get different results on Java 5 and on Java 6 and later.
> Java 5 will order the results like this:
> ij> select * from t order by c;
> C  =20
> ----
> =C4=B1a =20
> Ia =20
> ia =20
> =C4=B0a =20
> =C4=B1b =20
> Ib =20
> ib =20
> =C4=B0b =20
> 8 rows selected
> Java 6 and later order them like this like this:
> ij> select * from t order by c;
> C  =20
> ----
> =C4=B1a =20
> Ia =20
> =C4=B1b =20
> Ib =20
> ia =20
> =C4=B0a =20
> ib =20
> =C4=B0b =20
> 8 rows selected

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrato=
rs
For more information on JIRA, see: http://www.atlassian.com/software/jira