Ordering and unicode

On Thursday 10 November 2005 21:31, Michael Schuerig wrote:

I have a database on PostgreSQL 8.0.3 with unicode (utf-8) encoding,
client encoding is set to unicode, too. LC_COLLATE for the cluster is
de_DE.iso885915@euro. I noticed that that collation doesn't work for
two-byte characters, apparently they are ordered bytewise.

My current conjecture is that I'd have to re-initialize the cluster
with a utf-8 collation. Is this correct? Right now I don't have
access to the machine and can't check this.

Yes, it is the case. Just checked on another installation. Thanks for
listening...

Michael

--
Michael Schuerig Thinking is trying to make up
mailto:michael@schuerig.de for a gap in one's education.
http://www.schuerig.de/michael/ --Gilbert Ryle

On 10.11.2005, at 21:31 Uhr, Michael Schuerig wrote:

My current conjecture is that I'd have to re-initialize the cluster
with
a utf-8 collation. Is this correct? Right now I don't have access to
the machine and can't check this.

Yes, as far as I know there is no other way of changing the locale
settings. Hopefully you are on Linux! If you deploy on Mac OS X or
*BSD it won't work even with a change.

cug

On 11.11.2005, at 9:33 Uhr, Guido Neitzer wrote:

Yes, as far as I know there is no other way of changing the locale
settings. Hopefully you are on Linux! If you deploy on Mac OS X or
*BSD it won't work even with a change.

I have to correct me: Hopefully you are not on Mac OS X. On Mac OS X,
locale support is not yet available for UTF-8, so you will not get
correct ordering. This is from a statement of an Apple engineer --
they have this on the to do list.

As the locale support in Mac OS X comes directly from BSD I assume,
it's not better there but I have tested this only on one BSD
plattform and looked through the provided locale files (in the given
cvs directories) for others where it seems similar.

This is not a problem of PostgreSQL, only a problem for PostgreSQL
when running on the "wrong" plattform.

cug