Re: Forward references and recursion

• From: muchan <muchan@xxxxxxxxxxx>
• Date: Fri, 09 Dec 2005 11:29:28 +0100

First let me say that someone will say that I should re-consider my solution
and just not do this.  If I don't find an easy solution I will, but it will
mean cutting and pasting to duplicate a lot of code, and isn't that what the
C++ and the STL are supposed to avoid?  At any rate:

I have a two classes, each contains a list of the other, like so:

class A
{
,,,
    list<B> Blist;
}

class B
{
...
    list<A> Alist;
}

The declaration for class B is, of course, fine, but the list<B> in class A
generates many errors.  I've tried a simple forward declaration

(snipped)

The basic problem, IMO, is that a class need to know the size of the classes
it "contains".

It needs not only the name, but it need to know the declaration, to calculate
the size of that member, to calculate the size of itself...

So, philosophically speaking ;) you need not only "forward declaration", but
"forward definition" of class structure with all members of known size,
which is not possible in recursive way, as you have tried.

A workaround, is to change the member as the pointer to the list,
since the size of pointer for any object is known to compiler, it can deside
the size of the class structure with only forward declaration of the name.
That is,

class A;
class B;

 class A
 {
 ...
     list<B>* pBlist;
 };

 class B
 {
 ...
     list<A>* Alist;
 };

and provide the constractor which initialises the pointer to
either newly allocate or already existing list<>.

You may decide to change only class A or class B to contain a pinter,
but in your place I'd probably decide for symmetry and similar interfaces.
Or rather, I'm not very sure if they really should cross linked together,

 (Can you draw down the relation of actual object A and B, and the contents
  of each A's and B's inside the lists insides A's and B's?
  Sorry, I can't imagine it well.

  Probably you needs just a outside lists of plain data classes, and some
  ober classes to define the relations to each of them...
  In short, I think it's better take some hours to think about the designs
  of real objects your program should represent.

  Think about some "atom" classes of data, and manager classes to manage them...
  without putting the data and mechanism inside just two classes.
  Well, just IMHO, or IMH Imagination.)

muchan

.

• Follow-Ups:
  • Re: Forward references and recursion
    • From: James Curran

• References:
  • Forward references and recursion
    • From: Corey Cooper

• Prev by Date: Re: Forward references and recursion
• Next by Date: Re: Forward references and recursion
• Previous by thread: Re: Forward references and recursion
• Next by thread: Re: Forward references and recursion
• Index(es):
  • Date
  • Thread

• Windows
• Science
• Usenet

• Archive
• About
• Privacy
• Search
• Imprint