[hibernate-dev] Bytecode enhancement and inheritance

[Luis Barreiro]

Hi Guillaume,

Can you please check the behavior of with the Javassist enhancer in this 
particular case? I believe it's the correct one, but one has to double 
check.

Also, on a related issue, please check the comments I made on 
https://hibernate.atlassian.net/browse/HHH-12593 about property access.

I think we need to sort out both aspects, the interception and enhancing 
as mapped collection, in order to support this case properly. The former 
is a bug but for the latter I don't see a good enough solution.

Let me know what you find so we can proceed on solving this out.

Regards,

Luis Barreiro

Middleware Performance Team

<https://red.ht/sig>

On 06/11/2018 06:34 PM, Guillaume Smet wrote:
> Hi Luis,
>
> This question is related to 
> https://hibernate.atlassian.net/browse/HHH-12601 .
>
> AFAICS, the following case (e.g. having the fields defined in the 
> parent class and the laziness defined in the child class) is not 
> working very well with bytecode enhancement:
> ====
> @Entity
> public class RequestWithLazyEvents extends RelatedToEvents {
>
>     @ManyToMany(fetch = FetchType.LAZY)
>     @Fetch(FetchMode.SELECT)
>     public Set<Event> getEvents() {
>         return events;
>     }
> }
>
> @MappedSuperclass
> public abstract class RelatedToEvents extends Base {
>     protected Set<Event> events = new LinkedHashSet<>();
>
>     public void setEvents(Set<Event> events) {
>         this.events = events;
>     }
> }
> ====
>
> By not working very well, I mean that the interceptor read/write 
> methods are defined in the child class but the getters and setters are 
> not intercepted because we only intercept the getters/setters if the 
> fields are defined in the current class:
>
> See this method in PersistentAttributeTransformer and especially the 
> last test:
> ====
>     private boolean isEnhanced(String owner, String name, String desc) {
>         for ( FieldDescription enhancedField : enhancedFields ) {
>             if ( enhancedField.getName().equals( name )
>                     && enhancedField.getDescriptor().equals( desc )
>                     && 
> enhancedField.getDeclaringType().asErasure().getInternalName().equals( 
> owner ) ) {
>                 return true;
>             }
>         }
>         return false;
>     }
> ====
>
> (Note that supporting this case is not only a matter of removing the test)
>
> I don't know if it's a case we want to support or not but I'm a bit 
> worried that we don't throw any error and the collection is not loaded 
> at all (as the rest of ORM supposes the enhancer will do the work).
>
> You end up saving the collection in the database and when you get it 
> back the collection is always empty.
>
> It can lead to serious data loss if you save the entity back to the 
> database thus I think it's somewhat critical to fix it.
>
> WDYT?
>
> -- 
> Guillaume