[eclipselink-users] Session 'Magic' newb Question

[JimOR <jimorourke@xxxxxxxxxxxxxxxx>]

In the course of the migrating a Toplink JPA RESOURCE_LOCAL DAO project to
EclipseLink, I've hit a fair share of n+1 @ManyToOne(EAGER) DB hits that I'd
like to optimize. As an example; Order.class has 1:n LAZY LineItem which has
EAGER n:1 Product.  So, as you would expect, each access of a LineItem is
generating a DB query for a single Product.

So I set LineItem.product to FetchType.LAZY, and added some
getLinesAndProduct(...) to our OrderDAO API using combinations of
javax.Persistence.Query.setHint(QueryHints.BATCH,...) and
(QueryHints.FETCH,...) to review the generated queries, and wrote a tescase
hoping to generate an error accessing something like
lineitem.getProduct().getVendor() using the original DAO getLines() method.

To my surprise, no testcase errors occur.  Any order's lineitem's product's
vendor name is magically returned (with some trips to the DB) when
requested, regardless of whether I set the dao=null, close/nullify the
EntityManager and/or EMFactory.  The only way I can produce an error is to
serialize and deserialize the list.

My persistence.xml properties contains only jdbc connection stuff, so
everything else is defaults, and our JPA Entities are statically weaved
(woven?) using ant.  We have a static DAOFactory that returns an
implementation of the requested Interface, and each implementation extends
an abstract CommonCRUD class that, among other things, supplies an
EntityManager created from a singleton EntityManagerFactoryProvider class. 
Everything in the project (except the ant weaving and my QueryHints stuff)
is pure javax.persistence API.

Is my deep seated fear of 'LazyInitializationException' been rendered moot
by JPA, or RESOURCE_LOCAL, or EclipseLink sessions, or is this simply a
poorly written testcase?

Any pointers are appreciated.

Jim



-- 
View this message in context: http://www.nabble.com/Session-%27Magic%27-newb-Question-tp25198387p25198387.html
Sent from the EclipseLink - Users mailing list archive at Nabble.com.