Re: varchar sort ordering ignore blanks

On Jan 15, 2008 6:37 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> Luca Arzeni <l(dot)arzeni(at)amadego(dot)com> writes:
> > That is: the sort order in postgres 8.1.9 seems to ignore the blank.
>
> This is expected behavior in most non-C locales.
>
> > In all cases I'm using locale LATIN9 during DB creation, but I tested
> also
> > with ASCII, UTF8 and LATIN1 encoding.
>
> LATIN9 isn't a locale, it's an encoding. Try "initdb --locale=C".
>
> regards, tom lane

--------------------
I guess this has nothing to do with the encoding, but with the collation
rules used, which is governed by "lc_collate" parameter. See what you
get on both DBs for:

SHOW lc_collate ;

HTH,
Csaba.
-------------------

Thanks Tom, and Csaba

both of you hit the problem: actually Postgres 7.4.7 has a C locale and
Postgres 8.1 has US.UTF8 locale. Setting locale to locale=C or locale=POSIX
for release 8.1 solved this issue, but it opens another one: if I use
locale=C, I get

XXXX A
XXXX C
XXXXB

as sort order, but this setting gives me an error when it cames to:

XXXX d
XXXX e
XXXX f
XXXX è

because the right sort ordering should be:

XXXX d
XXXX e
XXXX è
XXXX f

So the problem is:

- C or POSIX locale is OK with blanks but fails on locale specific vowels
- LATIN9 locale is OK with vowels but ignores blanks

Is there any way to consider blanks meaningfull AND sort properly locale
specific vowels ?

I don't know what SQL standard says about this issue, but I'm sure that in
Italy you sort names considering vowels AND blanks!

Thanks, Luca