Обсуждение: upper() with LATIN1 problem

Hi,

The developers on my company reports me a problem with de upper() function.

The problem is:
- with a text 'magriñá'
- the upper('magriñá') returns MAGRIñà
- NOTE that the characters 'ñ' and 'á' didn't change the case.
- the text are stored in a table in a character varying(50) field
- the database has LATIN1 encoding
- the postgresql version is 7.4.7
- the OS is Debian Linux 2.6.10

Please, can anyone help me to resolve this problem?

Thanks!
--
Guillermo González

=?ISO-8859-1?Q?=22Lic=2E_Guillermo_Gonz=E1lez=22?= <g.gonzalez@cybira.biz> writes:
> The problem is:
> - with a text 'magri��'
> - the upper('magri��') returns MAGRI��
> - NOTE that the characters '�' and '�' didn't change the case.

This is determined by your locale (specifically LC_CTYPE) not by what
the database thinks the encoding is.  You'll probably want to re-initdb
in the desired locale.

            regards, tom lane

Hi,

I look at the variable/parameter LC_CTYPE and it is set to C (I think
this is the default) for performance reason, the installer says.

If I use LATIN1 (spanish) what value is the best option for this
variable/parameter?

Thanks!

--
Guillermo González


Tom Lane wrote:
> =?ISO-8859-1?Q?=22Lic=2E_Guillermo_Gonz=E1lez=22?= <g.gonzalez@cybira.biz> writes:
>
>>The problem is:
>>- with a text 'magriñá'
>>- the upper('magriñá') returns MAGRIñà
>>- NOTE that the characters 'ñ' and 'á' didn't change the case.
>
>
> This is determined by your locale (specifically LC_CTYPE) not by what
> the database thinks the encoding is.  You'll probably want to re-initdb
> in the desired locale.
>
>             regards, tom lane
>

=?ISO-8859-1?Q?=22Lic=2E_Guillermo_Gonz=E1lez=22?= <g.gonzalez@cybira.biz> writes:
> If I use LATIN1 (spanish) what value is the best option for this
> variable/parameter?

Unfortunately locale names are not all that well standardized.
"locale -a" should show you all the names available on your platform.
You probably want some variant of "es_ES", the tricky part is figuring
out which one uses LATIN1 as opposed to any other encoding.

            regards, tom lane