Re: strange infinite loop in plpgsql - Mailing list pgsql-general

From rihad
Subject Re: strange infinite loop in plpgsql
Date
Msg-id 4735F628.6040006@mail.ru
Whole thread Raw
In response to Re: strange infinite loop in plpgsql  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-general
Tom Lane wrote:
> rihad <rihad@mail.ru> writes:
>>      LOOP
>>        SELECT date+1 INTO day FROM days WHERE date=day OR EXTRACT(dow
>> FROM day) IN (0,6);
>>        EXIT WHEN NOT FOUND;
>>        timeout := timeout + 86400;
>>      END LOOP;
>
> If the EXTRACT condition is true, then the SELECT will always succeed.

Isn't the new "day" re-evaluated on every loop iteration? I'm totally
confused.

> This code will get even more whacko once you have more than one row
> in "days", because it'll pick a random one of the rows in that case
> (in practice, the physically first one).  I think you need something
> more like
>
>     LOOP
>       IF EXTRACT(dow FROM day) IN (0,6) THEN
>         -- don't bother to consult table on weekends
>         day := day + 1;
>       ELSE
>         SELECT date+1 INTO day FROM days WHERE date=day;
>         EXIT WHEN NOT FOUND;
>       END IF;
>       timeout := timeout + 86400;
>     END LOOP;
>
> BTW, you forgot to initialize "timeout".
>
Sorry, I hand-cooked this fast from the working code. I guess it
defaults to NULL instead of "random bits", which of course wouldn't save
me either, but the real (somewhat bigger) code eventually does RETURN
LEAST(timeout, expiration_timeout); skipping any nulls.

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: (Never?) Kill Postmaster?
Next
From: Tony Caduto
Date:
Subject: Verison 8.3 PL/pgSQL debugger Question