episode 2 - class variables vs. class level instance variables

Hello Dave,

In episode 2 (00:22:15) you mention a “good use” of class << self to quickly implement getter and setter of an instance variable @count. I get a bit confused with this example. Shouldn’t @count be a class variable (@@count), as it reflects the state of class Dave (that is: how many instances of class Dave are instanciated)?

I have checked with a simple:

def get_count
  @count
end

But still, why do class variables (@@count) exist in Ruby then, and how would a @@count variable compare to @count ?

Anyway, a thousand thanks for this enlightening serie!

Remember that instance variables are always accessed in the object referenced by <a href="mailto:self”>self@. While a class is being defined (that is, inside the body of the class itself), self is set to the class object. Thus instance variables set in this context are defined in the class object—they are instance variables, bit of the class, and not of instances of the class.

Thank you so much for your reply. I think I get it now, and when you write ” @@variable style class variables should be avoided”, I feel like I hear your funny reruns of the sixties tv shows… Would they lead to bad coupling by any chance, and also perhaps force you to break the OO encapsulation paradigm?

Cheers, Fred.

In reality, @@class_variables don’t work very well. I’ve never felt the need for them.

Neither did I, but I was just curious. I found an article summing up the subject (class variables vs. class level instance variables), with some interesting metaprogramming techniques that I can now decrypt thanks to your videocast. It seems the troubles arise with inheritance…