=?ISO-8859-1?Q?Jens_=D8stergaard_Petersen?= <oesterg@hum.ku.dk> writes:
> Thank for the tips, Oliver and Paul! I take it then that there are no
> means of constructing collation tables (like I do in FrontBase), that
> allow me to control which characters are equalled with which characters
> in searches - this is a part of SQL92, I believe.
I'm not sure whether SQL92 allows for user-defined collations, but in
any case we don't yet have any of the SQL-specified facilities in that
area. Postgres just sits atop the libc locale facilities (strcoll()
etc). If you do not like any of the locales available on your machine
then you need to provide your own locale definition. I've never done
it, so I can't say what degree of wizardry is needed... but it's
certainly possible.
BTW, if you do find yourself issuing lots of queries like
select * from ... where lower(col) = lower('foo');
then you'll want to create a functional index on lower(col) to
make this go faster.
regards, tom lane