Обсуждение: Re: transactions within stored procedures

Hi,

Has there been any progress on nested transactions in the last 10 years?

I am in a situation where I have a script that executes a number of
functions that build and populate a database including a rather large
lookup table (80GB) along with a number of large indexes (10GB).

I am trying to place the contents of this script into a single function
so that the database can be backed up and contain all of the code that
is needed to rebuild it via a single function call.

The problem is that if I were to run all of these functions in a single
transaction, postgresql would run out of memory pretty quickly and it
would all fail. Since you can't set autocommit off and explicitly
control transaction handling in a function (I think..), my dreams of
encapsulating all of the scripting for the project in the database and
simplifying the build procedure are no more.

I would appreciate any thoughts or suggestions..

Cheers,
Tom

On 21 July 2010 10:43, Robot Tom <robotwilcox@googlemail.com> wrote:
> Hi,
>
> Has there been any progress on nested transactions in the last 10 years?
>

PostgreSQL has had subtransactions since version 8.0, if that's what
you mean. If you're experiencing OOM a lot, a high work_mem setting is
often the culprit. This can be set dynamically though.

--
Regards,
Peter Geoghegan