Have you looked at creating a function in perl and creating a new connection? Or using a dblink query which can create a new connection? These two methods work. I have used them to insert to a log table regardless of the parent transaction being commited or rolled back.
The key is opening a new session which using dblink or pl/perl dbi connection will do. This is not ideal or efficient. It would be nice if you could just do autonomous transactions natively in pl/pgsql, but I find this method works for the cases where you need it(logging, huge batch processing tasks where it's not ideal to process everything in one transaction).
The same can be done with plProxy which is quite efficient but yes opening connections is not. So if used extensively it would be clever to use pgBouncer to reuse connections. Thanks for interesting idea.
Bob
"Hi all.
Is there a way to have (sub)transactions within a function body? I'd like to execute some code (a transaction!) inside a function and later decide whether that transaction is to be committed or not.
> I have seen this feature being asked for, and this work-around suggested so > many times. If plpgql does it internally, why not provide a clean interface > for this? Is there some road-block, or that nobody has ever tried it?
Initially we aimed at just exposing SAVEPOINT and ROLLBACK TO in functions, but ran into the problem that the SPI stack needs to be dealt with appropriately and you can't do it if the user is able to modify it arbitrarily by calling transaction-modifying commands. That's when the EXCEPTION idea came up. We never went back and studied whether we could have fixed the SPI limitation, but it's not trivial.