Re: Commit for a single table?

What do you mean "for the sake of transactions I don't want many commits"? Commits are almost 0 overhead, all the work is done, all that happens in a commit is the rollback data is marked as no longer needed, and the transaction is marked committed. Just go ahead and commit what you want. Don't mess with trying to control this, if you have performance issues, look elsewhere.
The entire basis of RDBM'S is predicated on the "all or none" notion of a transaction, and all uncommitted data being visible only to the current transaction.
ed

Hannes Lechner wrote:

> Hi!
>
> is there any way of performing a COMMIT on a single table only (i.e. w/o
> also commiting changes to the rest of the tables)?
>
> or is it possible to maybe have two (or more) independent table-spaces (or
> simmilar) in a database so that a connection to a specific db-space only
> commits all tables in this table-space - leaving all others untouched?
>
> i need this to code some sort of DB-based IPC (a message queue implemented
> in a table; whenever adding a message (=row) this row must be visible to all
> connections/sessions - i.e. i do have to perform a commit after inserting
> the row... but for the sake of tranactions i don't want to many commits...
>
> any suggestions????
>
> Thanks,
> Hannes
Received on Fri Apr 16 1999 - 07:17:13 CDT