http://rubyforge.org/cgi-bin/viewcvs.cgi/evil/lib/evil.rb?cvsroot=evil

Here is a demo using this to get AOP:

require 'evil'

class A
  def foo;"A";end
end

a = A.new
a.foo # => "A"

# first level "cut"
class B < A.clone # need a clone to prevent a loop
  def foo;"["+super+"]";end
end
A.become(B) # replace parent with subclass

a.foo # => "[A]"

# second level "cut"
class C < A.clone
  def foo;"{"+super+"}";end
end
A.become(C)

a.foo # => "{[A]}"

# restore the original functionality
A.become(B.superclass)

a.foo # => "A"

module Brackets
  def foo;"["+super+"]";end
end

# "preclude"-like functionality
class D < A.clone
  include Brackets
end
A.become(D)

a.foo # => "[A]"

Of course if we had this, it would be better to have a safer
Class#replace and that be able to break circular loops
automatically instead of having to use #clone to break the loop
manually.

We have been there you know. There's a reason why we've got a C version of
Florian's Object#become method on the Suby page. The problem with this
version is that A has become B. This means that the code that previously
defined, redefined and undefined methods in the original A, does that in B
now. The problem is that it is not possible to add a wrapper class this
way without breaking existing code, or at least without that code messing
with your advice. The behavior is almost the same, except that code needs
to be adapted to use these wrapper classes, and the basic idea behind AOP
is that that should *not* be so. We used this as an early test, but in the
end it had to go because it is obtrusive.

I'm sorry, but we're two years ahead of you you know.