Re: Getting the index for an element in mutually referencing containers

[Simon Wright <simon@pushface.org>]

"Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> writes:

> On 2017-03-11 21:05, Simon Wright wrote:
>> "Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> writes:
>>
>>> On 2017-03-11 12:21, Simon Wright wrote:
>>>> "Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> writes:
>>> The semantics of "object deleted" is in question. Usually it is merely
>>> throwing it out of some list (container) and thus decreasing object's
>>> reference count. The object stays alive until nobody else uses it =
>>> nobody holds a strong reference to it.
>>
>> That seems to be a very computer-oriented viewpoint. If your objects
>> represent real-world objects (or roles, or events - OOA view) then you
>> wouldn't want them to hang around! If you decide to abort an engagement
>> and therefore delete the Missile object, whose finalization sends a
>> self-destruct command to the corresponding missile in flight, you want
>> it to happen Right Now.
>
> If you have control over it. In the real world, which includes real
> programs, you might have none and there is no such thing as "now"
> anyway.

Of course not. But there is such a thing as avoiding the possibility
of references preventing deletion when deletion needs to happen as soon
as possible.

> Then I would say that rather your idea is computer-oriented. Objects
> are given us in their public operations. Once you cannot apply any to
> the object, or in other semantics, if Object.Exists yields False,
> there is no object. In terms of missiles, when you see a cloud of
> explosion, the work is done. You drop the last your reference and
> don't care if the object transited into another dimension or some
> object paradise. If you don't see it by any programmatic means, then
> there is none.

I (a software Missile object) have been told to delete myself. In my
finalization, I must send a command to my real-world equivalent telling
it to destroy itself. I only need to send it once, because
communications aren't my responsibility (and if comms are down nothing
can be done anyway).


Basically, I (Simon, now) am having trouble thinking of an application
where reference counting would be an appropriate solution.