On Sat, 22 Jan 2005 17:09:42 -0500, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > This time setlocale() was needed to get the behaviour
> > I needed (database initdb'ed to 'C', my order set to 'pl_PL',
> > or whatever locale I need at given moment).
> I would imagine that the performance is spectacularly awful :-(.
> Have you benchmarked it? A large sort on a unitext column,
> for instance, would be revealing.
True. Yet it would be still better than nothing ("C"). Actually
I was thinking that maybe functional indexes could be
used to boost the speed (at least for ordering).
> > ...but I would like to force ORDER BY using operators
> > provided by me without this 'USING <' clause.
> Hmm, the existence of the default btree operator class should be
> sufficient.
If You (or anyone) could try that SQL file and try to find
missing clause... :)
I guess that the case is that DOMAIN unitext is not quite
another type, so text's default operators sometimes take
precedence over unitext's own. :)
> > CREATE OR REPLACE FUNCTION lower(unitext) RETURNS unitext AS $$
> > utf8::decode($_[0]);
> > return lc($_[0]);
> > $$ LANGUAGE plperlu IMMUTABLE;
>
> AFAIK upper/lower cannot be considered to be locale-independent
> (see Turkish I/i business for a counterexample).
I imagine it is not possible to make 'one size fits all' lower(),
yet perl's uc()/lc() in my opinion for some cases is still
better than choosing one locale or using "C" locale.
Regards, Dawid