Heather,
Have you looked into creating a custom sequence generator[1]? Within the
generator, the persistent class and/or fields could be interrogated for
annotations and you could generate a value based on them. This is more of a
runtime vs. init-time solution, though.
Here's an example:
package generator;
import java.lang.reflect.AccessibleObject;
import org.apache.openjpa.jdbc.conf.JDBCConfiguration;
import org.apache.openjpa.jdbc.kernel.JDBCStore;
import org.apache.openjpa.jdbc.meta.ClassMapping;
import org.apache.openjpa.meta.FieldMetaData;
public class MyGenerator extends
org.apache.openjpa.jdbc.kernel.AbstractJDBCSeq
{
@Override
public JDBCConfiguration getConfiguration() {
return null;
}
@Override
protected Object nextInternal(JDBCStore arg0, ClassMapping arg1)
throws Exception {
FieldMetaData fmd = arg1.getFieldMapping("someField");
AccessibleObject fld = (AccessibleObject)fmd.getBackingMember();
MyAnnotation anno =
(MyAnnotation)fld.getAnnotation(MyAnnotation.class);
// Return the value of the annotation. This will get persisted in
the DB column.
return anno.value();
}
}
@Entity
@SequenceGenerator(name="myGenerator",
sequenceName="generator.MyGenerator()")
public class GenValue {
@MyAnnotation("FieldValue")
@GeneratedValue(generator="myGenerator",
strategy=GenerationType.SEQUENCE)
public String someField;
...
}
[1]
http://openjpa.apache.org/builds/latest/docs/manual/ref_guide_sequence.html
-Jeremy
On Fri, May 15, 2009 at 2:06 PM, Kevin Sutter <kwsut...@gmail.com> wrote:
> Hi Heather,
> Interesting use of the OpenJPA metadata processing. Just from your brief
> description, I would be careful on depending on an implementation versus
> defined interfaces. Even if we could get this initialization done earlier
> for you, unless there was some defined mechanism to force this (ie. the
> openjpa.InitializeEagerly property), I wouldn't count on it. In this
> particular case, the metadata repository is meant for OpenJPA processing.
>
> I know with OpenJPA the lines may not be crystal clear between public
> API's,
> provider SPI's, or internal classes. Of course, since this is an
> open-source project, you are welcome to peruse the complete contents of the
> code base and see if a solution is doable or not. I would just be careful
> on being too dependent on a specific implementation.
>
> Kevin
>
> On Fri, May 15, 2009 at 9:30 AM, Heather Sterling <hst...@us.ibm.com>
> wrote:
>
> > Thanks Kevin.
> >
> > Our scenario is that we have added product-specific annotations to the
> > persistent classes. In particular, we have some annotations that we need
> in
> > our perPersist entity listener to generate unique identifiers before the
> > object is persisted. We're currently initializing all of the metadata at
> the
> > beginning. We essentially iterate through the persistent classes and
> create
> > maps that hold this extra information. I guess it's possible for us to
> > lazily load this data in the entity listener if it hasn't been loaded for
> a
> > particular class. I also looked to see if there was another hook we could
> > potentially use, but didn't see any that seemed to fit the problem.
> >
> >
> > Heather Sterling
> > Systems Management Development
> > Phone: 919-254-7163 T/L: 444-7163
> > Cell: 919-423-3143
> > Email: hst...@us.ibm.com
> >
> >
> > [image: Inactive hide details for Kevin Sutter ---05/12/2009 12:53:30
> > PM---Heather,]
> > Kevin Sutter ---05/12/2009 12:53:30 PM---Heather,
> >
> >
> >
> >
> > *Kevin Sutter <kwsut...@gmail.com>*
> >
> > 05/12/2009 12:48 PM
> > Please respond to users
> >
> >
> >
> > To: users@openjpa.apache.org
> > cc:
> > Subject: Re: correct way to load persistent metadata at startup
> >
> >
> > Heather,
> > Good question... :-) This topic has come up recently due to some
> > potential
> > locking issues (serialized access) when reading the metadata repository
> in
> > a
> > multi-threaded environment. At this point, there is not a clear cut
> answer
> > for forcing the initialization of the metadata repository. The code for
> > the
> > new openjpa.InitializeEagerly property was committed to trunk last August
> > (after the 1.2.0 release was created), but we've since determined that it
> > may not be complete for all cases. You are welcome to try it out from
> > either the 1.3.x or trunk branches and see if it helps in your particular
> > situation.
> >
> > Let us know what you find out. Also, can you explain your need for
> getting
> > this data eagerly vs letting the lazy loading process play out? Thanks!
> >
> > Kevin
> >
> >
> >
> > On Mon, May 11, 2009 at 2:40 PM, Heather Sterling <hst...@us.ibm.com>
> > wrote:
> >
> > >
> > >
> > > I am attempting to load all the persistent class metadata eagerly. I
> > > realize this isn't great performance-wise, but I need to do it for the
> > > time-being. I had wanted to call:
> > >
> > > ClassMetaData[] classMetaDatas =
> > > conf.getMetaDataRepositoryInstance().getMetaDatas();
> > >
> > > but realized the data was loaded lazily when nothing came back. I
> > switched
> > > to using:);
> > >
> > > Collection c =
> conf.getMetaDataRepositoryInstance().loadPersistentTypes(.
> > > true, null);
> > >
> > > which returned the classes to me, but getMetaDatas() still returned
> > > nothing. Finally, I resorted to iterating through the class names,
> which
> > > seemed to work.
> > >
> > > Set names = conf.getMetaDataRepositoryInstance
> > > ().getPersistentTypeNames(false, null);
> > > if (names != null) {
> > > for (Iterator it = names.iterator(); it.hasNext();) {
> > > String persistentClassname = (String)it.next();
> > > System.out.println("Pre-loading metadata for: " +
> > > persistentClassname);
> > > try {
> > > ClassMetaData cc =
> conf.getMetaDataRepositoryInstance
> > > ().getMetaData(Class.forName(persistentClassname), null, true);
> > > } catch (ClassNotFoundException e) {
> > > // TODO Auto-generated catch block
> > > e.printStackTrace();
> > > }
> > >
> > > }
> > > }
> > >
> > > I found a link regarding a potential openjpa property called
> > > openjpa.InitializeEagerly (
> > > https://issues.apache.org/jira/browse/OPENJPA-620) but it was never
> > > checked
> > > into a release.
> > >
> > > Given all these options, what is the correct way to load the metadata
> on
> > > startup?
> > >
> > >
> > > Heather Sterling
> > > Systems Management Development
> > > Phone: 919-254-7163 T/L: 444-7163
> > > Cell: 919-423-3143
> > > Email: hst...@us.ibm.com
> >
>
