Re: [Springframework-developer] flush() method on HibernateTemplate

[Colin S. <col...@ex...>]

FLUSH_EAGER is not at all the same, in my opinion. When doing mixed 
Hibernate based code (with JDBC or EJBs) I on a number of occasions had 
to issue flushes to ensure something was safely available for the other 
code, but it was certainly nothing I wanted even most of the time 
(default FLUSH_EAGER)... And your typical DAO based on something like 
HibernateDaoSupport doesn't usually have the concept of working with 
multiple templates.

I was originally kind of against the idea, especially given the fact 
that in non Session bound to thread mode the operation would be 
completely redundant. But at the end of the day, the template is used by 
DAOs, DAOs are very aware of Hibernate specifics, and I think there's 
nothing fundamentally that wrong with allowing an explicit flush to be 
done in an easy fashion. I don't agree this operation is really related 
to transaction state either. Whether that data is in the Session cache 
or has gone out to the db, the transaction can generally still commit or 
rollback with the same effect. And Hibernate itself will on its own 
criteria also decide to flush the cache sometimes, such as when you do a 
query.  In the end, I think it's somewhat contrived if some app has to 
do a half dozen HibernateCallback cases where the only only operation 
inside them is a flush() call.


jürgen höller [werk3AT] wrote:

>Well, that's what HibernateTemplate's FLUSH_EAGER mode is meant for: To automatically flush right after each operation, even if a prebound Session is used. I would generally recommend to use such a preconfigured HibernateTemplate instance here, no matter if a shared one like in a HibernateDaoSupport or a newly instantiated one for a given SessionFactory.
> 
>I'm not keen on a flush operation on HibernateTemplate itself: That's not really a data access operation like the other methods there, but rather a method that just modifies the state of the transaction. If someone doesn't like the approach with a preconfigured HibernateTemplate, there's still the option of a custom HibernateCallback, even if just for a save and a flush.
> 
>Juergen
> 
>
>________________________________
>
>Von: spr...@li... im Auftrag von Colin Sampaleanu
>Gesendet: Mi 01.09.2004 17:34
>An: spr...@li...
>Betreff: [Springframework-developer] flush() method on HibernateTemplate
>
>
>
>Juergen,
>
>We had a request on the forums (and I've seen this before) to add
>flush() as a method on HibernateTemplate. Of course, this would be
>completely useless (or superfluous) in the case there is no existing
>thread bound session (or at least an existing outer transaction to bind
>the new session to and keep it there afterwards). But most people do use
>HibernateTemplate in their DAOs with the expectation that the Session is
>already around or will be created and stay around, and it's probably
>reasonable to allow them to do a flush without having to use
>HibernateCallback, since they could have just done a one-off operation
>like save() and then want to ensure that changes go out, for example
>when combining with JDBC based operations in a DAO.
>
>What do you think?
>  
>