2009/6/11 hubert depesz lubaczewski :
> On Thu, Jun 11, 2009 at 12:45:56PM +0200, Pavel Stehule wrote:
>> generally - modification of cycle's control variable isn't good
>> technique, because it's should be broken by some optimizations. When
>
> i would argue then that these optimizations are broke
On Thu, Jun 11, 2009 at 12:45:56PM +0200, Pavel Stehule wrote:
> generally - modification of cycle's control variable isn't good
> technique, because it's should be broken by some optimizations. When
i would argue then that these optimizations are broken, then.
> you would to modify this some var
2009/6/11 hubert depesz lubaczewski :
> On Wed, Jun 10, 2009 at 04:51:44PM -0400, Tom Lane wrote:
>> It's the new implementation. Depending on unspecified implementation
>> details is a good way to have broken code.
>
> i'm not sure if it's good change. there might be perfectly good reasons
> to i
On Wed, Jun 10, 2009 at 04:51:44PM -0400, Tom Lane wrote:
> It's the new implementation. Depending on unspecified implementation
> details is a good way to have broken code.
i'm not sure if it's good change. there might be perfectly good reasons
to increment idx from within loop.
Best regards,
Hello
2009/6/10 Tom Lane :
> "Atul Chojar" writes:
>> So in 8.3.7, the incrementing of the for-loop variable "idx" is being
>> ignored; that is not the case in 8.2.7. Is this a new feature of 8.3.7 or a
>> bug?
>
> It's the new implementation. Depending on unspecified implementation
> details is
"Atul Chojar" writes:
> So in 8.3.7, the incrementing of the for-loop variable "idx" is being
> ignored; that is not the case in 8.2.7. Is this a new feature of 8.3.7 or a
> bug?
It's the new implementation. Depending on unspecified implementation
details is a good way to have broken code.
We recently upgraded from postgres version 8.2.7 to 8.3.7. The below
pl/pgsql test function behaves differently in the 2 versions.
The code of the function is :-
CREATE OR REPLACE FUNCTION "public"."testloop" () RETURNS varchar AS
$body$
BEGIN
FOR idx IN 1..10 LOOP
raise not
