The system I'm working on uses CMP entity beans on a WebLogic platform. Most access is controlled through the use of EJB Select/Finder methods.

Technically, this architecture performs correctly.

However, to increase performance, I'm looking for more control over how often Weblogic flushes data to the database.

During a managed transaction, various inserts and updates will be performed. In order for EJB-QL statements to run correctly against any uncommited data, the changes must be flushed to the database prior to running the EJB-QL statement.

Is it possible for CocoBase to query against an in-memory database (that contains un-committed data)? Ideally I'd like to pre-query the data that may be affected, perform all changes in memory, and then flush the data at the end of the transaction one time.

The issue you have is that Entity Beans don't give the developer access to lifecycle management. For instance, if you issue a query within a current transaction, the entity beans you have changed may or may not yet be saved.

This sort of black box approach is one of the many problems with Entity Beans, and also one of the many reasons Entity Beans are being end of lifed by sun.

Have you considered using POJO persistence instead? The CocoBase EJBQL implementation will run inside the EJB container, but doesn't require CMP or Entity Beans. It also gives you EXPLICIT controls over saves such as:

cbfacade.save(customer); // CAN WRITE TO DB WITHOUT DELAY
cbfacade.executeQLQuery("select..."); // QUERY DB STATE

This seems more like what you WANT to accomplish - and by using POJO persistence you can readily implement it...

THOUGHT Support