Re: [boost] Loki SmartPtr study: Policy orthogonality issues

From: Beman Dawes (bdawes_at_[hidden])
Date: 2002-01-14 14:07:55

At 10:10 AM 1/14/2002, Peter Dimov wrote:

 >You can't access SmartPtr::storage_policy in a member declaration since
 >SmartPtr is incomplete at this point.
 >
 >You can access it in the member function body, though.
 >
 >Tying this to my previous question, what is the reason of the
 >'pointer_type'
 >(over)generalization? When is pointer_type not T*? std::allocator 
flashes?

That question also occurred to me several times.

A simple nested inheritance hierarchy works particularly well if 
pointer_type is always T* for the CheckingPolicy:

    template < .... >  // Same as Loki::SmartPtr
    class smart_ptr
      : public OwnershipPolicy<
                   StoragePolicy<
                      CheckingPolicy< T >, ConversionPolicy > >
    {
      ...
    };

Notice that because CheckingPolicy was allowed to assume pointer_type is 
T*, the inheritance relationships are simpler.  StoragePolicy can still 
supply the typedefs, and OwnershipPolicy can use them.  Neat and tidy.

(I made ConversionPolicy a policy of StoragePolicy; that still seems the 
case to me.)

--Beman