Infinite recursive loop with SafeDeleteQueryset

[kursat-ozer]

Hi,

I am having an issue similar to 117

I am using django-treebeard for hierarchical data in DB and deleting an item with descendants should also delete all related descendants. I also want these items to be safely deleted so policy is SOFT_DELETE_CASCADE.

I think I am using the treebeard and safedelete packages in the right way but if you see anything wrong with it please do tell:
Note: The models are slightly different but this is how it is basically.

class FolderQuerySet(MP_NodeQuerySet, SafeDeleteQueryset):
    pass

class Folder(MP_Node, SafeDeleteModel):
    ...
    objects = SafeDeleteManager.from_queryset(FolderQuerySet)()
    all_objects = SafeDeleteAllManager.from_queryset(FolderQuerySet)()
    deleted_objects = SafeDeleteDeletedManager.from_queryset(FolderQuerySet)()

When I try to delete a folder,

• it first finds all descendants
• then should safe delete all items and cascade.
  But in the SafeDeleteQueryset on line 66 it calls _, delete_response = obj.delete(force_policy=force_policy) and this causes the mp_node delete call and starts infinite recursion.
  Updating the line with _, delete_response = obj._delete(force_policy=force_policy) should fix this, right?

class SafeDeleteQueryset(query.QuerySet):
    ...

    def delete(self, force_policy: Optional[int] = None) -> Tuple[int, Dict[str, int]]:
        """Overrides bulk delete behaviour.

        .. note::
            The current implementation loses performance on bulk deletes in order
            to safely delete objects according to the deletion policies set.

        .. seealso::
            :py:func:`safedelete.models.SafeDeleteModel.delete`
        """
        assert self.query.can_filter(), "Cannot use 'limit' or 'offset' with delete."

        if force_policy == NO_DELETE:
            return (0, {})
        elif force_policy == HARD_DELETE:
            return self.hard_delete_policy_action()
        else:
            deleted_counter: Counter = Counter()
            # TODO: Replace this by bulk update if we can
            for obj in self.all():
                _, delete_response = obj.delete(force_policy=force_policy)
                deleted_counter.update(delete_response)
            self._result_cache = None
            return sum(deleted_counter.values()), dict(deleted_counter)

[Gagaro]

I'm not sure we should overpass the delete method.
What's causing the recursion in the mp_node delete method?
You should probably have your own mixin inheriting from both MP_Node and SafeDeleteModel and fix the conflict in its own delete method (or your model directly if you only use it there).

[kursat-ozer]

Sorry for the late response,

What's causing the recursion in the mp_node delete method?

MP_Node just propagates the delete to the next in inheritance which is normal I think.
SafeDelete behaves normal in single object operations. But when a bulk operation is needed, it causes recursive calls. It is not related to MP_Node, it will be the same with any other library that requires modified delete.
As I said in my post it is the same issue with #117
And this is the comment explains the logic:#117 (comment)
And this is the line why an another method created to do the same task:

django-safedelete/safedelete/models.py

Line 164 in 3804385

return self._delete(force_policy, **kwargs)

You should probably have your own mixin inheriting from both MP_Node and SafeDeleteModel and fix the conflict in its own delete method (or your model directly if you only use it there).

What I did is inheriting only SafeDelete and fixing the line I said above and using this model as base.

But this needs a fix :)