On Mon, Apr 11, 2011 at 1:08 AM, Charles Oliver Nutter
<headius / headius.com> wrote:

> Am I misunderstanding the original question?

Since MRI inserts proxies in the ancestor chain that point to the
actual modules and adjusts the parent pointers to have a linear
ancestor chain (by linear I mean that you follow the pointers up and
you're done), you get this

    module M
    end

    class C
      include M
    end

    module N
      def doesnt
        p :doesnt
      end
    end

    module M
      include N

      def works
        p :works
      end
    end

    p C.ancestors # => [C, M, Object, Kernel, BasicObject]
    p M.ancestors # => [M, N]

That linear ancestor chain is not updated if the ancestors chain of
involved modules is modified. That's what I meant by "caching".

Both the #works method and the N module augment the M module after
inclusion in C. Thanks to the pointer to M in the ancestor chain (via
the proxy) instances of C respond to #works. But because of the
implementation N does not belong to the ancestor chain of C, while it
belongs to the ancestor chain of M. In consequence

    C.new.works
    C.new.doesnt

raises an exception in the second line.

I think that module proxies are implementation, and that conceptually
they do not belong to the object model (please correct me if that's
wrong!). So, conceptually, mixin methods are not copied into classes,
modules are "linked" to classes and inspected at call time to offer a
method dispatch that supports dynamic stuff as shown above. But that
metaphor does not extend to modules mixed-in later.

Question was whether this was a conscious design/implementation
decision on behalf of speed.

Charles I know you know all of this, just wanted to clarify my question :).