Re: problem with custom_variable_classes

Поиск
Список
Период
Сортировка
От hubert depesz lubaczewski
Тема Re: problem with custom_variable_classes
Дата
Msg-id 20080924213825.GA17162@depesz.com
обсуждение исходный текст
Ответ на problem with custom_variable_classes  (Malcolm Studd <mstudd@recognia.com>)
Ответы Re: problem with custom_variable_classes  ("Taras Kopets" <tkopets@gmail.com>)
Re: problem with custom_variable_classes  (Tom Lane <tgl@sss.pgh.pa.us>)
Список pgsql-general
On Wed, Sep 24, 2008 at 12:15:41PM -0400, Malcolm Studd wrote:
> I have a pl/pgSQL function[1] to calculate row numbers (based on [2]).
> It uses a custom variable. This was working earlier, but is breaking now
> saying it can't recognise the variable. The custom_variable_classes is
> set in the postgresql.conf.
> pgdb001=> select rownum('') from generate_series(1,10);
> ERROR:  unrecognized configuration parameter "olap.rownum_name"
> pgdb001=> show custom_variable_classes;
>  custom_variable_classes
> -------------------------
>  olap
> (1 row)
>         settings_id := current_setting('olap.rownum_name');

it looks like a bug to me.
it did work in 8.2, and it doesn't in 8.3, or in 8.4.

while manual in 8.4 still states:
"When custom_variable_classes is set, the server will accept definitions of arbitrary variables within each specified
class.".

so it should work, but it doesn't:
# show custom_variable_classes ;
 custom_variable_classes
-------------------------
 depesz
(1 row)
# select current_setting('depesz.xxx');
ERROR:  unrecognized configuration parameter "depesz.xxx"

of course i still can set a value, and then get it:
# select set_config('depesz.xxx', 'xx', false);
 set_config
------------
 xx
(1 row)
# select current_setting('depesz.xxx');
 current_setting
-----------------
 xx
(1 row)

but it means that there is currently no way to check if there is value in this
- i.e. if it has been initialized.

i would suggest to return "NULL" when calling current_setting on unknown variable.

this is not how it worked in 8.2 (it returned empty string), but NULL is (in my opinion) better choice.

strangely - name of "current_setting" function nor "custom_variable_classes"
doesn't show in any release docs, which makes the change somewhat mysterious.

Best regards,

depesz

--
Linkedin: http://www.linkedin.com/in/depesz  /  blog: http://www.depesz.com/
jid/gtalk: depesz@depesz.com / aim:depeszhdl / skype:depesz_hdl / gg:6749007

В списке pgsql-general по дате отправления:

Предыдущее
От: Malcolm Studd
Дата:
Сообщение: problem with custom_variable_classes
Следующее
От: Garry Saddington
Дата:
Сообщение: Re: case expression