[C++-sig] Re: Allocating objects on the heap by default.

David Abrahams dave at boost-consulting.com
Mon Jul 7 23:23:24 CEST 2003


Prabhu Ramachandran <prabhu at aero.iitm.ernet.in> writes:

>>>>>> "DA" == David Abrahams <dave at boost-consulting.com> writes:
>
>     >> Hmm, so you are saying none of the functions should have ever
>     >> been written to deal with raw pointers at all and should all
>     >> have used shared_ptr?
>
>     DA> Raw pointers are low-level.  High-level interfaces are easier
>     DA> to use correctly and much easier to wrap.
>
> Indeed.
>
>     >> But at some point using shared_ptr requires one to handle
>     >> reference cycles and figure how to break them
>
>     DA> So does manual new/delete.
>
> Yes, but I've done that already and don't have the time to change all
> my code to use smart pointers now.  Essentially, its not a trivial
> change for me to make (since my library is quite large and uses
> pointers quite a bit) so I'll postpone this and update my to-do list.
>
>     >> (using weak_ptrs).  I'll admit that this is probably rare
>     >> (although I can think of a few candidates in my code) and a lot
>     >> easier to deal with.  It also requires that the users generate
>     >> shared_ptrs when they create a new object.
>
>     DA> I'm not sure what you mean by that.
>
> Its nothing important but what I meant was that a user of the library
> has to use 'shared_ptr<A> p(new A());' and not new A().  

Not if the library supplies a shared_ptr factory function for A.  

    A::create();

I'd tend to hide A's constructor(s) as well.

>     DA> If Pyste wants to supply auto_ptr and shared_ptr objects so
>     DA> that
>
>     DA>         A.held_by(auto_ptr)
>     DA> or
>     DA>         A.held_by(shared_ptr)
>
> Hmm, I am not sure if the A.held_by approach is easy to do

What could be hard about it?

>  but I do know that this can be done easily:
>
>   holder(A, 'std::auto_ptr<test::A>')
>
> Would that do?

It ain't pretty, but it works.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com





More information about the Cplusplus-sig mailing list