[boost] Re: smart_ptr suggestion: Support decrementing shared_ptr'scount, forself-references

From: Chuck Messenger (chuckm_at_[hidden])
Date: 2003-05-27 15:49:21

Peter Dimov wrote:
> Chuck Messenger wrote:
>>In general, the abstraction is: you have a group of intra-referential
>>objects.  When any of the group is constructed, they are all
>>constructed (so that the master count is temporarily > 1), and the
>>master count is
>>reset to 1.  When the master count goes to 0, the group is destructed.
>>Hence, the group only remains alive as long as there are any external
>>references (and as long as the intra-group references remain static).
> 
> 
> In the situation you describe, A logically contains a B and B logically
> contains an A, so A and B aren't independent; they actually form an AB
> object. There is nothing to be gained from separating A and B as this
> doesn't eliminate a dependency. If A and B are optional (but still tightly
> coupled) parts of AB, the way I'd express this would probably be
> 
> class AB
> {
>     shared_ptr<A_impl> ai_;
>     shared_ptr<B_impl> bi_;
> };
> 
> There are several ways to allow A_impl and B_impl to communicate: store an
> AB* pointer in each;

Consider that A and B may implement various "interfaces" (i.e. inherit 
from 1+ abstract base classes w/o member variables).  I can't just use 
multiple inheritance (i.e. AB inherits from each interface that either A 
or B needs), because, for one thing, there can be more than one A and/or 
B in the group.  How would you differentiate between them?

> pass a B_impl* to A_impl's methods that need it; keep a
> B_impl* or a weak_ptr<B_impl> in A_impl.

These approaches greatly complicate the code.  Suppose I have a host of 
functions defined at A and B level -- nothing defined at A_impl and 
B_impl.  To do what you're saying, for one thing, I'd have to have a 
parallel set of methods for A_impl and A (and for B_impl and B).  But 
it's worse than that -- I also need a parallel set of external functions 
(methods of other functions, etc), which can accept either an A or an 
A_impl.  It's just not workable.

     - Chuck