Message-Id: <199906221413.XAA11243@srapc451.sra.co.jp>
To: David Warnock <david@sundayta.co.uk>
cc: t-ishii@sra.co.jp,
	"pgsql-interfaces@postgreSQL.org" <pgsql-interfaces@postgreSQL.org>
Subject: Re: [INTERFACES] JDBC and character sets 
From: Tatsuo Ishii <t-ishii@sra.co.jp>
Reply-To: t-ishii@sra.co.jp
In-reply-to: Your message of Tue, 22 Jun 1999 14:54:11 +0100.
	<376F9583.BFB85129@sundayta.co.uk>
Date: Tue, 22 Jun 1999 23:13:46 +0900
Sender: t-ishii@srapc451.sra.co.jp

>I think I saw from the list of developers that you wrote a lot of the
>multiu-byte code. Is that correct? If so my grateful thanks.

Yes, I'm responsible for the code multi-byte.

>Are there any limitations or gotchas about using unicode everywhere?
>
>Specifically
>
>1. Column length. Is this measured in unicode characters or do I need to
>increase the length of Varchars? ie is a varchar(10) certain to hold 10
>unicode characters?

When you define varchar(n), n should be counted in bytes, not
characters.
We assume Unicode is input as UTF-8 encoding. In UTF-8, 10 ASCII chars
take 10 bytes. So varchar(10) will hold 10 Unicode chars if they are
all ASCII. However, if you use ISO8859 chars they will take 2 bytes
for each letter. If you use KANJI, 3 bytes for each letter. You could
use octet_length() to measure the size of a Unicode string in bytes.

>2. Indexing. What sort order will I get from an index or an order by for
>unicode characters.

It will sorted in the order of Unicode code point.

>Can this be customised.

Currently no.

>Generally I try to do any
>really important sorting in Java where I can use the correct sort order
>for the locale.

>3. Upper/lowercase. I have been using separate columns for uppercase
>versions of names etc again so that the case changes can be done by the
>client which will know the correct rules for the locale where the data
>is entered. What do upper/lower case functions in Postgresql do with
>unicode?

I think it will related to locale. I'm not sure but I've heard about
Unicode locale. If it really exists, you could do:

configure --with-mb=UNICODE --with-locale

so that upper/lower works for Unicode.

>4. Are there any limitations on what I use to write triggers? Can all
>the different ways work reliably with unicode?

I'm not sure but it should work with triggers.
--
Tatsuo Ishii