Walter is right about transitive readonly - here's the alternative

[Bill Baxter]

Sean Kelly wrote:
> Frits van Bommel wrote:
>> Sean Kelly wrote:
>>> Though the lack of a dot operator overload may complicate the syntax 
>>> a tad.
>>
>> Isn't that what 'alias this' would be for?
> 
> Hrm... so something like this?
> 
>     struct SmartPtr( T ) {
>         T* val;
>         static if( is( T == class ) ||
>                    is( T == struct ) )
>         alias val this;
> 
>         void opAssign( T* v ) {
>             val = v;
>         }
>     }
> 
>     struct S {
>         void foo() {}
>     }
> 
>     SmartPtr!(S) p = new S;
>     p.foo();
> 
> 
> Seems like it would work, huh :-)  I hadn't really thought about 'alias 
> this' in this context.

That is interesting.  But if inheritance is such a great way to 
implement smart pointers, then why isn't it regularly used to implement 
smart pointers in C++?

oh wait... you can't inherit from a pointer in C++.
Or dynamically change your base object.

With copy constructors and destructors this could be a very slick way to 
do smart pointers.  Nice!

--bb