Date: Sun, 20 Jun 2004 22:35:09 -0700 (PDT)
From: Stephan Szabo <sszabo@megazone.bigpanda.com>
To: ctrl <ctrl@altonsys.com>
Cc: pgsql-sql@postgresql.org
Subject: Re: help with Postgres function
In-Reply-To: <46a31c4d.0406161349.38d3a34b@posting.google.com>
Message-ID: <20040620221537.C69900@megazone.bigpanda.com>
References: <46a31c4d.0406161349.38d3a34b@posting.google.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Wed, 16 Jun 2004, ctrl wrote:

> CREATE OR REPLACE FUNCTION getNextWebsiteForCrawl(integer) RETURNS
> website AS '
> DECLARE
> my_record RECORD;
> w website%rowtype;
> count smallint;

You can't safely use a variable named count and the count(*) expression
below I think, so you'll want to rename this variable.

> BEGIN
>  SELECT id, domain into my_record FROM websites WHERE crawl_status=1
> AND date(last_fetch) > (current_timestamp - interval ''$1 days'')

I don't think that'll get you the interval you want.  I think you want
 current_timestamp - $1 * interval '1 day'
instead since the $1 isn't going to get replaced inside the string.

>  ORDER BY last_fetch LIMIT 1;

>  select count(*) into count from my_record;
>  if count > 0 then

This isn't going to work either.  my_record isn't some kind of recordset,
it's a single variable so I the select count(*) from my_record doesn't
make sense.  Perhaps you could check for my_record.id being non-null.

>  w.id := my_record.id;
>  w.domain := my_record.domain;
>  update websites set crawl_status=2 where id = my_record.id;
>  end IF;
>  return w;

If you couldn't get a value, what do you want this to return? Right now it
returns a website where the values are null I believe.

The function as written is not going to guarantee distinct results if used
from multiple transactions concurrently. How are you planning to use the
function?  Many sequential calls from a single connection, random usage
from multiple, etc?