Guy Rouillier wrote:
> This approach will not work reliably. Between your insert of nextval and
> your select of currval, someone else may have inserted another row, such
> that your currval will actuall retrieve the wrong currval. I suppose you
> could address this with synchronized, but a better way (IMO) is to retrieve
> the nextval into your code. You can then use it for the insert and also no
> the value for any further use you may have, and not have to worry about
> multi-user issues.
I believe that Oracle guarentees isolation of sequence values between
transactions - the value you get on from 'seq.nextval' will be what you
(in the same transaction) get from the next 'seq.curval' regardless of
what anybody does in another transaction.
>
> We always wrap the sequence with a getter function, and this has saved our
> hides many times. For example, let's say you start by using the sequence
> number as your primary key. Then someone comes along and decides that they
> want the IDs to look like "CS00001". You may have to change code in
> several places if you just access the sequence directly, but if you wrap it
> in a getter function, you don't have to change anything but that getter
> function.
>
Having said the above, I agree with your approach completely - I always
get the sequence value first, within a RequiresNew transaction. This is
also for portability - I'm not reliant on Oracle's semantics, and it's
easy to port to PostgreSQL's sequence implementation or whatever you
need to do in other databases.
-danch
_______________________________________________
JBoss-user mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/jboss-user
