[boost] Re: shifted_ptr review

From: Philippe A. Bouchard (philippe_at_[hidden])
Date: 2003-09-30 16:27:52

David Abrahams wrote:

[...]

> No, I was suggesting making the constructor private, making
> make_shifted_ptr a friend, and have make_shifted_ptr<T>(...) return a
> shifted_ptr<T>.
>
> Your approach seems very complicated.  Is there a reason to do it
> that way?

make_shifted_ptr<T>() is the best solution, but it would be possible also to
create a new type explicitly allocated with the standard operator new, if we
are to use infinite overloads:

template <typename T>
    class stype
    {
        shifted_header h;
        char s[sizeof(T)];

    public:
            stype();

        template<class Arg1>
            stype(Arg1 &);

        template<class Arg1, class Arg2>
            stype(Arg1 &, Arg2 &);

    ...
    };

template<typename T>
    class  shifted_ptr
    {
    public:
            template <typename U>
                shifted_ptr(shifted_ptr::details::stype<U> &);

    ...
    };

This way it would be possible to force the following syntax:

shifted_ptr<int> p = new stype<int>(9);

Eventually it would be possible to implement something similar to the
garbage collector's black listing of pointers.  It uses fast heap page
detection to isolate pointer allocations that differs:
http://www.hpl.hp.com/personal/Hans_Boehm/gc/gcdescr.html

Philippe