Re: 'order by' does "wrong" with unicode-chars (german umlauts)

On Friday 19 September 2003 13:11, peter pilsl wrote:
> postgres 7.3.2
>
> I store unicode-data in postgresql. The data is retrieved via
> webinterfaces, processed with perl and then stored in postgresql (and
> viceversa).
>
> All is going nice with one problem. If performing a "select * order by
> field"-query the result is not what I expected.
>
> German umlauts (ie: Ö) are stored as doublechars ("Ö" is "Ö") and only the
> first char seems to be taken into account when sorting.
>
> So it happens that the order is like:
>
> Österreich
> America
> Pakistan
>
> instead of
>
> Amerika
> Österreich
> Pakistan

I'm no expert on locales, but I think you're confusing two things.
Your character-set determines what symbols you can store.
Your locale determines sorting rules. Check the end of the postgresql.conf
file for details of what your current settings are.

> How to deal with this Problem ? Of course converting to latin before
> storing would be a solution but we plan to offer support for many non-latin
> languages later and the meaning of unicode is to get rid of all this
> converting-stuff after all.

What sorting-order do you want? You probably have options like: C, en_GB, de,
it etc.
If you are storing non-latin1 symbols as well as latin1, I can't think of what
a reasonable sort order would be.

Unfortunately, PG only supports one locale at a time, and gets set during
initdb. See the chapter on Localisation in the manuals for details.

--
Richard Huxton
Archonet Ltd