Re: [HACKERS] Bug #659: lower()/upper() bug on

> I think it is really not hard to do this for UTF-8. I don't have to know the
> relation between the locale and the encoding. Look at this:
> We can use the LC_CTYPE from pg_controldata or alternatively the LC_CTYPE
> at server startup. For nearly every locale (de_DE, ja_JP, ...) there exists
> also a locale *.utf8 (de_DE.utf8, ja_JP.utf8, ...) at least for the actual Linux glibc.

My Linux box does not have *.utf8 locales at all. Probably not so many
platforms have them up to now, I guess.

> We don't need to know more than this. If we call
> setlocale(LC_CTYPE, <value of LC_CTYPE extended with .utf8 if not already given>)
> then glibc is aware of doing all the conversions. I attach a small demo program
> which set the locale ja_JP.utf8 and is able to translate german umlaut A (upper) to
> german umlaut a (lower).

Interesting idea, but the problem is we have to decide to use exactly
one locale before initdb. In my understanding, users willing to use
Unicode (UTF-8) tend to use multiple languages. This is natural since
Unicode claims it can handle several languages. For example, user
might want to have a table like this in a UTF-8 database:

create table t1(
english text, -- English message
germany text, -- Germany message
japanese text -- Japanese message
);

If you have set the local to, say de_DE, then:

select lower(japanese) from t1;

would be executed in de_DE.utf8 locale, and I doubt it produces any
meaningfull results for Japanese.
--
Tatsuo Ishii