[hibernate-dev] envers / new metamodel

Strong Liu stliu at hibernate.org
Wed Aug 22 00:38:45 EDT 2012 

On Aug 22, 2012, at 8:11 AM, Steve Ebersole <steve at hibernate.org> wrote:

> Also, I think how you were expecting beforeMetadataProcessing to happen can't really happen, but I suspect it is the more direct way to port envers.
> 
> I *think* what you were thinking is to have envers pass along its extra EntitySource/EntityHierarchy stuff.  You cannot just simply "add the custom sources into {@MetadataSources} and get it processed by Hibernate Metamodel" if you mean adding EntitySource/EntityHierarchy.  That is actually not the role of MetadataSources.  MetadataSources is a collection of sources for find metadata information; its the collection of mapping files, annotated classes, annotated packages, etc. org.hibernate.metamodel.spi.MetadataSourceProcessor implementations (there is one for HBM and one for Annotations) are responsible for taking the MetadataSources and building EntityHierarchy instances and passing those off to binder.
> 
> So I need to understand exactly what y'all want here in terms of Envers since that is our primary use case for this.  Are you wanting to "contribute" additional EntityHierarchy structures?  IIRC, old envers code used to build in-memory DOM structures (or am I way off base in my memroy there?); if that is the target, maybe contributing JaxbRoots to MetadataSources is the way to go.

Envers still builds  in-memory DOM, and this is why I want a "beforeMetadataProcessing" ( the name is under discussion ) so we can call org.hibernate.metamodel.MetadataSources#addDocument

but the problem here is, I'm not sure if envers could (easier) build a DOM without knowing metadata info ( like hibernate type of user provided mapping entity property )

> 
> Also, I wonder if we ought to rename MetadataSources to remove the ambiguity, maybe MetadataOrigins or MetadataSourcesOrigins?
> 
> On 08/21/2012 08:16 AM, Steve Ebersole wrote:
>> Yes, thats what you said in the javadocs ;)
>> 
>> What I mean is that we have a use case for one of those.  But not the
>> other.  So why develop something that we do not have a use case for?
>> 
>> On Mon 20 Aug 2012 11:23:14 PM CDT, Strong Liu wrote:
>>> reuse the IntegratorService / ServiceLoader ?
>>> 
>>> On Aug 21, 2012, at 2:14 AM, Steve Ebersole <steve at hibernate.org
>>> <mailto:steve at hibernate.org>> wrote:
>>> 
>>>> Also, I have been thinking that there really ought to be different
>>>> types of "integrators".  Not sure what we gain by forcing
>>>> MetadataContributingIntegrator, ServiceContributingIntegrator,
>>>> TypeContributingIntegrator, etc to extend Integrator
>>>> 
>>>> 
>>>> On Sat 18 Aug 2012 05:50:29 PM CDT, Steve Ebersole wrote:
>>>>> The general idea is good.  But not really getting the point/purpose of
>>>>> having both before and after hooks.
>>> 
>>> BEFORE:
>>> 
>>> modules can choose extend the MetadataSources with its own hbm/ann and
>>> get Metadata (Binder) process it
>>> 
>>> this is pretty like what envers does today, or what i'm hoping it
>>> could do, so we could reuse lots of envers code
>>> 
>>> AFTER:
>>> 
>>> modify existing EntityBindings etc or create its own EntityBindings
>>> based on the existing ones
>>> 
>>> 
>>> 
>>> 
>>>>> 
>>>>> On 08/17/2012 01:12 AM, Strong Liu wrote:
>>>>>> I'm thinking add the interface below, with it, modules like envers
>>>>>> can choose either before or after the metamodel get processed to hook
>>>>>> into its own extending mappings
>>>>>> 
>>>>>> 
>>>>>> public interface MetadataContributingIntegrator extends Integrator {
>>>>>>   /**
>>>>>>    * Allow the integrator to alter the {@link MetadataImplementor}
>>>>>> BEFORE {@link MetadataSources} get processed.
>>>>>>    * <p/>
>>>>>>    *
>>>>>>    * At this stage, metamode ( like {@link
>>>>>> org.hibernate.metamodel.spi.binding.EntityBinding} etc. ) is not
>>>>>> available yet.
>>>>>>    * This is a good time to add the custom sources into
>>>>>> {@MetadataSources} and get it processed by Hibernate Metamodel.
>>>>>>    *
>>>>>>    * @param metadata The Metamodel which is going to be completed
>>>>>> by processing MetadataSources.
>>>>>>    * @param source  Meta data sources to be processed.
>>>>>>    */
>>>>>>   public void beforeMetadataProcessing(MetadataImplementor
>>>>>> metadata, MetadataSources source);
>>>>>> 
>>>>>>   /**
>>>>>>    * Allow the itnegrator to alter the {@link MetadataImplementor}
>>>>>> AFTER @{@link MetadataSources} get processed.
>>>>>>    * <p/>
>>>>>>    *
>>>>>>    * At this stage, metamode ( like {@link
>>>>>> org.hibernate.metamodel.spi.binding.EntityBinding} etc. ) is bindded.
>>>>>>    * So, it depends on the integrator to manually create metamodel
>>>>>> and add it to {@link MetadataImplementor} or modifiy the
>>>>>>    * existing one.
>>>>>>    *
>>>>>>    * @param metadata
>>>>>>    * @param source
>>>>>>    */
>>>>>>   public void afterMetadataProcessing(MetadataImplementor metadata,
>>>>>> MetadataSources source);
>>>>>> }
>>>>>> 
>>>>>> downside is, envers with old metamodel calls
>>>>>> Configuration.buildMappings() again during the integration phase, so,
>>>>>> its created hbm xml can be processed (again) before SF create, but
>>>>>> with new metamodel, it is hard to do that I think ( or too time
>>>>>> consuming )
>>>>>> 
>>>>>> and to create envers own metamodel ( hbm ) it needs to know the
>>>>>> hibernate type of entity property, which can be easily get from
>>>>>> Entitybinding but hard / duplicated to resolve from source by envers
>>>>>> itself, so it seems we should choose "afterMetadataProcessing"
>>>>>> 
>>>>>> but envers' hbm creation involves lots of code and pretty complicated
>>>>>> ( to me :) it's better to reuse those code and choose
>>>>>> "beforemetadataProcessing" and let new metamodel to take care of that.
>>>>>> 
>>>>>> thoughts?
>>>>>> 
>>>>>> 
>>>>>> On Aug 2, 2012, at 12:17 AM, Hardy Ferentschik <hardy at hibernate.org
>>>>>> <mailto:hardy at hibernate.org>>
>>>>>> wrote:
>>>>>> 
>>>>>>> On 31 Jan 2012, at 10:55 PM, Sanne Grinovero wrote:
>>>>>>> 
>>>>>>>> Why is the "annotation indexing" discussion part of the metamodel?
>>>>>>> Why not? We are using Jandex in the new metamodel which is a
>>>>>>> annotation index/repository
>>>>>>> 
>>>>>>>> I initially understood that a replacement of commons-annotations was
>>>>>>>> being developed, which would be nice for Search too as Search
>>>>>>>> does not
>>>>>>>> and should not depend on Hibernate ORM.
>>>>>>> as Strong already said, there is no replacement module for
>>>>>>> commons-annotations. There is no
>>>>>>> need for it.
>>>>>>> 
>>>>>>> And yes, Search should imo also switch to Jandex, however, it can
>>>>>>> initially just create its own index.
>>>>>>> Of course it might be nice to be able to use a Jandex index passed
>>>>>>> to it via the integrator spi. Different story though.
>>>>>>> 
>>>>>>> --Hardy
>>>>>>> _______________________________________________
>>>>>>> hibernate-dev mailing list
>>>>>>> hibernate-dev at lists.jboss.org <mailto:hibernate-dev at lists.jboss.org>
>>>>>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>>>>>> -------------------------
>>>>>> Best Regards,
>>>>>> 
>>>>>> Strong Liu <stliu at hibernate.org <http://hibernate.org>>
>>>>>> http://about.me/stliu/bio
>>>>>> 
>>>>>> _______________________________________________
>>>>>> hibernate-dev mailing list
>>>>>> hibernate-dev at lists.jboss.org
>>>>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>>>>> 
>>>>> 
>>>> 
>>>> --
>>>> steve at hibernate.org <mailto:steve at hibernate.org>
>>>> http://hibernate.org
>>> 
>>> -------------------------
>>> Best Regards,
>>> 
>>> Strong Liu <stliu at hibernate.org <http://hibernate.org/>>
>>> http://about.me/stliu/bio
>>> 
>> 
>> --
>> steve at hibernate.org
>> http://hibernate.org
> 
> -- 
> steve at hibernate.org
> http://hibernate.org

-------------------------
Best Regards,

Strong Liu <stliu at hibernate.org>
http://about.me/stliu/bio

More information about the hibernate-dev mailing list