I agree that the HSQLDB method of using sequences is vague, but an easy solution is not terrible difficult- override the HSQLDialect. getSequenceNextValString() method to return an SQL statement of your choice.

try to use
SELECT TOP1 NEXTVALUE FOR SEQUENCE FROM TABLE
instead of
SELECT NEXTVALUE FOR SEQUENCE FROM TABLE

where TABLE – name of any table in the DB (for example it can be any system table like INFORMATION_SCHEMA.SYSTEM_COLUMNS) and SEQUENCE – the name of your sequance.

it’s worked

- If you are not going to use the Objects for update afterwards you can include the seq.nextval in the insert statement in Oracle9 as well.

I have built a persistence layer called “Ebean” which you can check out at http://www.avaje.org. There I also raise the Architectural and API issues of EJB3.

I certainly think Ebean would be a good fit for your problem but I’m a little biased

Cheers, Rob.

I do agree that comparing Ibatis and Hibernate is a bit like comparing apples and oranges. The blog is by no means a comparision. We felt that we had applied our (golden) hammer Hibernate to the wrong problem and expected (or hoped?) that Ibatis would do a better job. In the end we were not even able to decide whether Ibatis would do a better job, because we saw serious issues with the framework itself (not in comparison with Hibernate) with, as stated above, testability and extendability, important features to any framework.

The “TypeHandlerCallback” interface is indeed a means to tweak reading and writing objects from a ResultSet. This is one of the three client-extensions. I think they’re rightly named “client”, since they allow adaption of the outward behaviour of the framework. What I miss though is “enigine”-extensions that allow adapting the behaviour of Ibatis itself, like the possibility to tweak the behaviour and usage of the SelectKeyStatement, as suggested in the blog.

With regard to the read consistency; this is indeed very important in many situations, but not in ours. It would be helpful if we could somehow toggle this behaviour, but regretfully Oracle does not have this feature and our only option is to read in small chunks.

I do not quite understand. The “TypeHandlerCallback” interface provides you with a means to make just those tweaks – or am I missing something?

On a further note with respect to:

“Oracle cannot handle this, due to its read consistency functionality. After quite some time a ORA-01555: snapshot too old (rollback segment too small) will inevitably crash the long running transaction.”

This might look like an obstacle to you, but this “read consistency” is functionality you cannot do without in a highly concurrent and highly transactional environment.

I haven’t done any testing from a single iBatis mapping file and multiple, different type databases. Maybe BerkeleyDB behaves ‘better’ with respect to sequences, or maybe there is a way to use one of the ways to get the sequence nextvalue using a Dynamic Mapped Statements.

IMO you can only compare Hibernate to iBatis to a certain degree. iBatis is typically well suited for environments where database/SQL savvy developers want to map their records onto objects (ROM) whereas Hibernate is typically used in top-down scenarios – the classmodel taking precedence over the data model (ORM).