Quarkus: Hibernate Panache and Entity lifecycle

[masini]

// programmer is transient and the method is @Transactional
programmer.language = Programmer.Language.Java;
programmer.status = Person.Status.Alive;
programmer.persist();

// now programmer is persistent and still not on the DB, the transaction is not flushed
programmer.language = Programmer.Language.C_Sharp;

When the method finish and transaction is commit I query the DB and found that language is Java and not C_Sharp as expected.

[FroMage]

So it appears that Hibernate is not enhancing my generated accessors for dirty checking. Perhaps it has to do with them being TRANSIENT, I'm checking.

[FroMage]

That's not it. I've also verified that it's running the Hibernate enhancer properly _after_ the Panache enhancer. It's like the Hibernate enhancer does not _see_ the generated accessors.

[FroMage]

So, it turns out that bytebuddy will load entities from the classloader even if we pass it the bytes we want to enhance. That's pretty harsh.

[FroMage]

@Sanne I believe this is a bug in Hibernate:

    @Override
    public byte[] enhance(String className, byte[] originalBytes) throws EnhancementException {
        //Classpool#describe does not accept '/' in the description name as it expects a class name. See HHH-12545
        final String safeClassName = className.replace( '/', '.' );
        try {
// HERE this is going to resolve the type not from the originalBytes but from the classloader
            final TypeDescription typeDescription = typePool.describe( safeClassName ).resolve();

            return byteBuddyState.rewrite( typePool, safeClassName, byteBuddy -> doEnhance(
                    byteBuddy.ignore( isDefaultFinalizer() ).redefine( typeDescription, ClassFileLocator.Simple.of( safeClassName, originalBytes ) ),
                    typeDescription
            ) );
        }
        catch (RuntimeException e) {
            throw new EnhancementException( "Failed to enhance class " + className, e );
        }
    }

[FroMage]

Unless this is fixed in Hibernate I don't know how to work around it. I could perhaps fake a classloader we pass to the Hibernate enhancer context in order to use the byte[] for the type we're enhancing, but that feels really hackish. Naturally all those members are private so I can't reimplement this method outside the class to test how to fix it.

[FroMage]

I can try to generate accessors that call the hibernate methods, but given that I don't have them yet I'm not sure it will work.

[FroMage]

Haha, that actually works.

[FroMage]

Fixed in #1691

[Sanne]

awesome @FroMage , I was starting to look at the ORM angle but that's looking like a very though one.

[FroMage]

Yeah, I figured. Go and accept my PR then ;)

[Sanne]

I will soon :) having a bit of a build queue even locally..

[FroMage]

Fixed now.