json-mop and inheritance

[jfmcbrayer]

Hi. I'm kind of a CL newbie, but I'm trying to use json-mop and having an issue with inheritance.

Suppose I have two classes:

(defclass parent ()
  ((foo :accessor foo :initarg :foo 
          :initform "Foo"
          :json-key "foo"))
  (:metaclass json-serializable-class))

(defclass child (parent)
    ((bar :accessor bar :initarg :bar
      :json-key "bar"))
    (:metaclass json-serializable-class))

(encode (make-instance 'parent)) prints exactly what I expect: {"foo":"Foo"}

I would expect (encode (make-instance 'child)) to print exactly the same thing. However, it prints {}.

But:

 (describe (make-instance 'child))
#<CHILD {1004ABEC93}>
  [standard-object]

Slots with :INSTANCE allocation:
  FOO  = "Foo"
  BAR  = #<unbound slot>

Maybe I'm misunderstanding a basic CLOS thing? Shouldn't encoding an instance of child include the values it gets from parent's initforms?

[gschjetne]

You're right, it should. I ended up not using this for my own project, and it never occurred to me to test for this case.

I'm afraid I don't have much time to work on it at the moment and I can't promise I'll get to it soon, but you're welcome to give it a try. A good start might be to create a test case for inheritance. And feel free to ask any questions in the thread.

[jfmcbrayer]

I'll give it a try, but as I said, I'm kind of a newbie. A test case is probably within my capabilities, though.

For reference, what did you end up using for your own project, assuming you still needed to generate JSON for CLOS objects?

[gschjetne]

Just defining the method yason:encode-slots on the objects. That way the internal implementation of the classes won't depend on the external JSON representation or vice versa.

[jfmcbrayer]

Ah, thanks. I will probably do that as well. But I guess I will still write you a failing test case for json-mop.