Sure, Kevin; that's a fair answer. I know that the root issue lies in the double-purposing of the ClassMetaData.isEmbeddedOnly() method. It is used both to determine whether a class should be an @Embedded, and it is also used to determine whether a class should be a @MappedSuperclass (!). I understand that the latter condition has been solved by the addition of the isAbstract() method in OpenJPA 2.0.x.
Anyhow, this isn't by itself all that bad. What may be bad is that there are several calls made to AnnotationPersistenceMappingSerializer#isMappingMode(ClassMetaData), and its default behavior is to consider whether the passed in ClassMetaData is embedded or not. This drives, among other things, whether field mappings retain all their attributes (like name), but unfortunately it also drives whether mapped superclasses get an @Entity or a @MappedSuperclass tag. So if you set a ClassMapping#isEmbedded() one way, you get mapped superclasses correctly annotated as @MappedSuperclass, but all the field annotations are empty. If you set it the other way, you get all the field annotations populated, but the class gets an @Entity tag. See bug 1360 for more details there. The way that I've almost solved the problem--but boy does it feel wrong!--is to install my own subclass of AnnotationPersistenceMappingSerializer so that isEmbeddedOnly() is simply not considered in the implementation of isMappingMode(ClassMetaData). This is just fine--the only remaining hurdle that I think I'm hearing is probably not going to be fixed anytime soon unless I wade in with a hammer and try to fix it myself :-) is that the @MappedSuperclass gets a @Table annotation. Thanks for the non-silence :-) Best, Laird On Mon, Dec 7, 2009 at 12:45 PM, Kevin Sutter [via OpenJPA] < ml-node+4127373-1333037...@n2.nabble.com<ml-node%2b4127373-1333037...@n2.nabble.com> > wrote: > Hi Laird, > I'm not much better than a cricket with this particular question, but I > have > a couple of observations... > > Since you are attempting to affect the execution of the ReverseMappingTool, > > you may be a little beyond basic usage questions. Thus, I'm going to > cross-copy this question over the dev mailing list. Not guaranteeing > activity because of this since most of us monitor both mailing lists. But, > > it won't hurt in this case... > > One thing to keep in mind is that the ReverseMappingTool is not meant to be > > the end-all tool for generating the Entity classes. It is expected that > tweaking of the generated code will be necessary to make your Entities > "production ready". Now, if the generated code is not syntactically > correct > or complete, then that's another issue. But, personally, I wouldn't expect > > the generated code to be "complete". > > If you continue to only get crickets responding to the request, then this > may indicate a lack of resource or even a lack of expertise on your > particular question. It may also be an excellent opportunity for you to > roll up your sleeves, debug the problem, and contribute a solution back to > the community. We are always looking for new blood. :-) > > Good luck, > Kevin > > > On Mon, Dec 7, 2009 at 10:29 AM, Laird Nelson <[hidden > email]<http://n2.nabble.com/user/SendEmail.jtp?type=node&node=4127373&i=0>> > wrote: > > > Hello; I asked this question before and got crickets. > > > > What is the best way to customize a ReverseMappingTool run such that for > > every table mapped I create a MappedSuperclass that it inherits from? > > > > I am using OpenJPA 1.2.1 and generating annotations, not XML. > > > > So if I have a database table named Person, then I want a > > PersonMappedSuperclass.java and a Person.java class that inherits from > it. > > > > I am so close it is maddening, but cannot cross the final chasm. > > > > The first thing I do is to set my ReverseCustomizer to change the > incoming > > ClassMapping to setEmbeddedOnly(true), and rename it to > > xyzMappedSuperclass. Then I generate a new class, set the > MappedSuperclass > > as its parent, and install the new mapping directly into the repository, > > like so (inside customize(ClassMapping)): > > > > mapping.setEmbeddedOnly(true); > > > > final Class<?> mappingClass = mapping.getDescribedType(); > > final String mappingClassName = mappingClass.getName(); > > > > final String newClassName = mappingClassName.substring(0, > > mappingClassName.length() - "MappedSuperclass".length()); > > final Class<?> newClass = this.tool.generateClass(newClassName, > > mappingClass); > > > > final MappingRepository mappingRepository = this.tool.getRepository(); > > > > > final ClassMapping newMapping = > > (ClassMapping)mappingRepository.addMetaData(newClass); > > newMapping.setPCSuperclass(mappingClass); > > newMapping.setTable(mapping.getTable()); // ? maybe? > > newMapping.setEmbeddedOnly(false); > > > > This does more or less what I want: at the end of this run, I get a > mapped > > superclass and an entity class that extends from it. However, none of > the > > mapped superclass field information is filled in. That is, I get things > > like: > > > > @Basic > > > > Instead of > > > > @Basic > > @Column(name="pmt_no", length=16) > > > > Bug 1360 seems to be related to this ( > > http://issues.apache.org/jira/browse/OPENJPA-1360). It would appear > that > > in > > the 1.2.x line, anyhow, it is impossible to create a MappedSuperclass > whose > > field annotations are properly filled in, thus of course defeating the > > entire purpose of a MappedSuperclass. I would be interested to learn if > > ANYONE has ever generated MappedSuperclasses using this tool. > > > > Best, > > Laird > > > > > ------------------------------ > View message @ > http://n2.nabble.com/MappedSuperclasses-tp4126924p4127373.html > To start a new topic under OpenJPA Users, email > ml-node+208411-553807...@n2.nabble.com<ml-node%2b208411-553807...@n2.nabble.com> > To unsubscribe from OpenJPA Users, click here< (link removed) =>. > > > -- View this message in context: http://n2.nabble.com/MappedSuperclasses-tp4126924p4127860.html Sent from the OpenJPA Users mailing list archive at Nabble.com.
