Re: [Modeling-users] inheritance and relations confusion

[Sebastien B. <sbi...@us...>]

Mario Ruggier <ma...@ru...> wrote:
> I have an entity (B) inheriting from another (A), with the
> constraint that any instance of either must be related
> to one instance of B. The corresponding model is therefore:
>=20
> model.entities =3D [
>    Entity('A', ...),
>    Entity('B', parent=3D'A', ...),
> ]
> model.associations=3D[
>    # A is always related to a B
>     Association('A','B',
>      relations=3D['b','as'],
>      multiplicity=3D[ [1,1], [0,None] ],
>      delete=3D['nullify','deny'],
>      keys=3D['fkBId','id'],
>    ),
> ]

Okay, so you have a kind of a reflexive relationship on the inheritance
tree of A, joining A to Bs (and thus, B to Bs) --your 'b' should then be
related to a 'B' itself, maybe himself, or you have a cycle
there. That's fine, I just want to make it clear so that we understand
each other.

> I can create the instances and rels in the expected way,
> but when I come to delete them, there is a problem:
> I try to get to the b for the a in question, to nullify the
> relation (respecting the "deny" delete rule). Thus I do:
>=20
>  > > > a =3D ec.fetch('A',qualifier=3D'...',isDeep=3D1)[0] # gives correc=
t a
>  > > > b =3D a.getB()
>  > > > b
> .<pympack.A.A instance at 0x95ba60>
>  > > > for related_a in b.getAs():
>                b.removeFromAs(related_a)
> Traceback (most recent call last):
>      for related_a in b.getAs():
> AttributeError: A instance has no attribute 'getAs'
>=20
> Thus, he insists that the b instance is of type A,
> and does not find the getAs() rel-method on it (but the
> id and other b-only non-rel attributes are correctly
> present!).

I made some tries but really could not get into this situation: whatever
I try, getB() always returns 'B' instances, never 'A' instances. Could
you please give the exact sequence of operations leading to this
particular situation? I'm probably too tired now for this, but I cannot
find of a way to make this happen, so a complete sequence of
instructions revealing the bug would be a great help. Thanks!

> But, removing known objects "explicitly", with :
>=20
>      a.removeObjectFromBothSidesOfRelationshipWithKey(b,'b')
>=20
> as many times as necessary (for each a related to this b) works
> just fine.
>=20
> Any enlightenment?
>
> Cheers, mario
>=20
> ps: ah, but when did relations ever behave nicely when inheritance is
> involved ;-?

...and how come you find some many bugs in one day? I thought this was
somehow regulated by an international agreement!?!!

  Thanks for reporting, and looking forward to see what makes this
  happen,

-- S=E9bastien.