Re: Isolation level in a function - Mailing list pgsql-general

From Bruno Wolff III
Subject Re: Isolation level in a function
Date
Msg-id 20060124102258.GA18870@wolff.to
Whole thread Raw
In response to Isolation level in a function  ("bgolda" <e9syuk002@sneakemail.com>)
List pgsql-general
On Thu, Jan 19, 2006 at 02:05:41 -0800,
  bgolda <e9syuk002@sneakemail.com> wrote:
> Hello, this is my first post, please don't shoot...
>
> I was just experimenting with transactions (PG 8.1), and there is
> something which puzzles me. If i write 'SET TRANSACTION ISOLATION LEVEL
> SERIALIZABLE;' in my function, it breaks. Error informs me, that it was
> executed after some query, while it is a first command in the function
> after declares and begin!

Because there is already a transaction which the function is executing in
and for which a query has been started (e.g. the one that calls the function).

> However, it seems to work perfectly well if I change in the same place
> the system variable, responsible for the transaction level. I used
> set_config, if I remember well. The function is only for a DBA task, so
> it seems to be all right to use set_config, isn't it?
>
> And my questions are:
> 1) Why in the first case it is not possible? Is it an error or am I
> just doing something wrong? Should have set something before?
> 2) Are there any downsides of the mentioned solution (changing
> transaction_isolation value), except of the fact that it is an
> administration command and should not be used :)?
>
> Thanks,
>
> _-_-_-_
> Bart Golda
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 2: Don't 'kill -9' the postmaster

pgsql-general by date:

Previous
From: "R, Rajesh (STSD)"
Date:
Subject: Re: [HACKERS] [PATCH] Better way to check for getaddrinfo function.
Next
From: "FERREIRA, William (VALTECH)"
Date:
Subject: execution plan : Oracle vs PostgreSQL