A Sliver of Ruby #3: Nesting Scopes and Constant Name Resolution

I was plodding my way along The Path To Enlightenment, this morning (ruby koans), and I was rather stumped on the following:

class Animal

  LEGS = 4

  def legs_in_animal

    LEGS

  end

end

class MyAnimals

  LEGS = 2

  class Bird < Animal

    def legs_in_bird

      LEGS

    end

  end

end

class MyAnimals::Oyster < Animal

  def legs_in_oyster

    LEGS

  end

end

def test_who_wins_with_both_nested_and_inherited_constants

  assert_equal 2, MyAnimals::Bird.new.legs_in_bird

end

def test_who_wins_with_explicit_scoping_on_class_definition

  assert_equal 4, MyAnimals::Oyster.new.legs_in_oyster

end

Notice that in the first test, the lexical scope wins, but in the second test, the inheritance hierarchy wins...and I was baffled as to why...

Turns out that, constant names are resolved in the following order (according to zetetic):

1. the enclosing scope
2. the outer scope (if any)
3. included modules
4. the superclass
5. the class Object
6. the class Kernel

And also referenced this complete example:

http://coderrr.wordpress.com/2008/03/11/constant-name-resolution-in-ruby/

The bottom-line answer to my question is: the syntax used to declare Oyster (i.e. using the qualified name as opposed to re-opening MyAnimals and nesting Oyster within) specifically means that it does not participate in the lexical scope of MyAnimals, it's merely sharing namespace.  There's no outer scope for Oyster and therefore the name resolution makes it all the way down to the 4th source.

It's the difference between this:

class MyAnimals::Oyster < Animal

  def legs_in_oyster

    LEGS

  end

end

and this:

class MyAnimals

  class Oyster < Animal

    def legs_in_oyster

      LEGS

    end

  end

end