Re: Structure mapping using reinterpret_cast.

Hi,

Thank you all for these very complete answers.

In my example, I forgot to point out that class 'A' would follow
certain specific traits :
- Only POD member data.
- No virtual functions.
- No static data.
- No base class.

In the same way, class 'B' and class 'C' would also have specific
traits :
- Only derived from class 'A'
- No virtual functions.
- No member data.

I agree that the standard can sometimes contradict itself, but could
you determine that this is guaranteed to work by the following :

If give two classes 'X' and 'Y' as follows :

     class X
     {
         /* data members, functions, etc ... */
     };

     class Y : public X
     {
         /* data members, functions, etc ... */
     };

And an instance of class 'Y' that we call 'y'

Is there a guarantee that '&y == &((X &)y)' ? I know this can blow up
with multiple inheritance, and I am confining myself to the scope of
the above example. If this is true, and given that a class always has
the same memory layout, in my original example with class 'A', 'B' and
'C' :

     A *a = new A(12);
     B *b = reinterpret_cast<B *>(a);
     C *c = reinterpret_cast<C *>(a);

You should have that 'b' and 'c' are properly mapped over 'a' as if
'a' was the instance of their base class. Anyway, I see that I'll have
to wait for C++0x to be official (and implemented) before performing
those tricks in a guaranteed way :)

Thanks for your time again,

Olivier.

--
      [ See http://www.gotw.ca/resources/clcm.htm for info about ]
      [ comp.lang.c++.moderated. First time posters: Do this! ]