Thread: Re: [GENERAL] Number format problem
Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Tom Lane wrote:
>> This patch broke the regression tests. It seems to change the C-locale
>> behavior on many platforms.
> Oh, so C locale has "" for thousands_sep? Hmm. So there is no way to
> disinguish "" as "don't have" from "" as "don't want". Will revert.
Possibly we could special-case C locale, and allow an empty string only
when not C locale. However, it could be that there are other locales
where the change would surprise people, too :-(
regards, tom lane
Tom Lane wrote: > Bruce Momjian <pgman@candle.pha.pa.us> writes: > > Tom Lane wrote: > >> This patch broke the regression tests. It seems to change the C-locale > >> behavior on many platforms. > > > Oh, so C locale has "" for thousands_sep? Hmm. So there is no way to > > disinguish "" as "don't have" from "" as "don't want". Will revert. > > Possibly we could special-case C locale, and allow an empty string only > when not C locale. However, it could be that there are other locales > where the change would surprise people, too :-( Not sure. Peter stated the French locale was wrong for the Debian they were using, and he should know, so I am going to wait for more feeback. One thing that has me confused is that psql/print.c doesn't check for a zero-length string for thousands_sep in setDecimalLocale, so I am not thinking that is wrong and should skip "" like we do in formatting.c. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
Bruce Momjian wrote: > > > Oh, so C locale has "" for thousands_sep? Hmm. So there is no way to > > > disinguish "" as "don't have" from "" as "don't want". Will revert. > > > > Possibly we could special-case C locale, and allow an empty string only > > when not C locale. However, it could be that there are other locales > > where the change would surprise people, too :-( > > Not sure. Peter stated the French locale was wrong for the Debian they > were using, and he should know, so I am going to wait for more feeback. The baseline glibc sources from gnu.org have (in glibc-2.3.6/localedata/locales/fr_FR): LC_NUMERIC decimal_point "<U002C>" thousands_sep "" grouping 0;0 END LC_NUMERIC So at first sight it's more a glibc issue rather than debian-specific. However, this was filed as a bug in debian long ago. Incidentally, it got fixed just this week, meaning that the future debian-glibc should use a space instead of the empty string, as reported here: http://lists.debian.org/debian-glibc/2006/02/msg00168.html Which is consistent with Peter's statement that it was wrong up to now, and also with what the OP expected in the first place. -- Daniel PostgreSQL-powered mail user agent and storage: http://www.manitou-mail.org
Daniel Verite wrote: > Bruce Momjian wrote: > > > > > Oh, so C locale has "" for thousands_sep? Hmm. So there is no way to > > > > disinguish "" as "don't have" from "" as "don't want". Will revert. > > > > > > Possibly we could special-case C locale, and allow an empty string only > > > when not C locale. However, it could be that there are other locales > > > where the change would surprise people, too :-( > > > > Not sure. Peter stated the French locale was wrong for the Debian they > > were using, and he should know, so I am going to wait for more feeback. > > The baseline glibc sources from gnu.org have > (in glibc-2.3.6/localedata/locales/fr_FR): > > LC_NUMERIC > decimal_point "<U002C>" > thousands_sep "" > grouping 0;0 > END LC_NUMERIC > > So at first sight it's more a glibc issue rather than debian-specific. > > However, this was filed as a bug in debian long ago. Incidentally, it got fixed > just this week, meaning that the future debian-glibc should use a space instead > of the empty string, as reported here: > http://lists.debian.org/debian-glibc/2006/02/msg00168.html > Which is consistent with Peter's statement that it was wrong up to now, and also > with what the OP expected in the first place. Oh, good. I looked and now realize that psql is already checking for a non-"" string: if (*extlconv->thousands_sep) so we are consistent. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073