Interface FinalizablePyObject

This interface allows PyObjects to have finalizers. Alternatively one can use FinalizableBuiltin.

The difference is that __del__() can be overridden by a new-style subclass's __del__-method on Python-side, while FinalizableBuiltin.__del_builtin__() is always called. If a Python-side finalizer exists, FinalizableBuiltin.__del_builtin__() will be called after the Python-side finalizer has been processed.

One can even implement both interfaces. If both interfaces are implemented, the FinalizeTrigger will call __del__() first and then FinalizableBuiltin.__del_builtin__(). If a new-style subclass has an own, Python-side __del__-method, this overrides the Java-implemented __del__(), but not FinalizableBuiltin.__del_builtin__(), which will be called after the Python-side finalizer.

If you are writing a custom built-in that shall directly extend PyObject or some other not-yet-finalizable builtin and have a finalizer, follow the instructions below.

1. Let your subclass implement FinalizablePyObject (or FinalizableBuiltin).
2. In every constructor call
   FinalizeTrigger.ensureFinalizer(this);
   
3. Write your __del__()-method however you intend it. (or FinalizableBuiltin.__del_builtin__() if FinalizableBuiltin was used)
4. (optional)
   If your finalizer resurrects the object (Python allows this) and you wish the finalizer to run again on next collection of the object:
   In the block where the resurrection occurs, let your __del__()- or FinalizableBuiltin.__del_builtin__()-method call
   FinalizeTrigger.ensureFinalizer(this);. If you implement __del__ in Python and need this functionality, you can simply call someObject.__ensure_finalizer__()
   Note that this is Jython-specific and should be surrounded by a try/except-block to ensure compatibility with other Python implementations.

Note: Regarding to object resurrection, Jython currently behaves like CPython ≥ 3.4. That means the finalizer __del__() or FinalizableBuiltin.__del_builtin__() is called only the first time an object gets gc'ed. If pre-3.4.-behavior is required for some reason (i.e. have the finalizer called repeatedly on every collection after a resurrection), one can achieve this manually via step 5).

The built-in function __ensure_finalizer__ is also useful if a class acquires a finalizer after instances have already been created. Usually only those instances that were created after their class acquired the finalizer will actually be finalized (in contrast to CPython). However, one can manually tell earlier created instances to become finalizable by calling __ensure_finalizer__() on them. As mentioned above, it is recommended to surround this with a try/except-block to ensure compatibility with other Python implementations.

Note that it is not possible to override __ensure_finalizer__ on Python side. If one overrides __ensure_finalizer__ on Python side, Jython will ignore the override-implementation and still call the original one.

It is possible to switch finalization on and off at any desired time for a certain object. This can be helpful if it is only necessary to have __del__() or FinalizableBuiltin.__del_builtin__() called for certain configurations of an object.

To turn off the finalizer, call:

 ((FinalizeTrigger) JyAttribute.getAttr(this, JyAttribute.FINALIZE_TRIGGER_ATTR)).clear();
 

To turn it on again, call

 ((FinalizeTrigger) JyAttribute.getAttr(this, JyAttribute.FINALIZE_TRIGGER_ATTR)).trigger(this);