Re: ICU for global collation

On Tue, Jan 04, 2022 at 05:03:10PM +0100, Peter Eisentraut wrote:
> On 04.01.22 03:21, Julien Rouhaud wrote:
>
> > > - if (!lc_collate_is_c(collid) && collid != DEFAULT_COLLATION_OID)
> > > - mylocale = pg_newlocale_from_collation(collid);
> > > + if (!lc_collate_is_c(collid))
> > > + {
> > > + if (collid != DEFAULT_COLLATION_OID)
> > > + mylocale = pg_newlocale_from_collation(collid);
> > > + else if (default_locale.provider == COLLPROVIDER_ICU)
> > > + mylocale = &default_locale;
> > > + }
> >
> > There are really a lot of places with this new code. Maybe it could be some
> > new function/macro to wrap that for the normal case (e.g. not formatting.c)?
>
> Right, we could just put this into pg_newlocale_from_collation(), but the
> comment there says
>
> * In fact, they shouldn't call this function at all when they are dealing
> * with the default locale. That can save quite a bit in hotspots.
>
> I don't know how to assess that.
>
> We could pack this into a macro or inline function if we are concerned about
> this.

Yes that was my idea, just have a new function (inline function or a macro
then since pg_newlocale_from_collation() clearly warns about performance
concerns) that have the whole
is-not-c-collation-and-is-default-collation-or-icu-collation logic and calls
pg_newlocale_from_collation() only when needed.