From: Tim Pease <tim.pease@...>
Date: 2011-07-14T13:52:37+09:00
Subject: Re: Issue with: instance_variable_defined?


On Jul 13, 2011, at 7:03 PM, Hal Fulton wrote:

> I expect the third line to be true rather than false.
> 
> Comments?
> 
> Hal
> 
> class SomeClass
>      attr_reader :a1, :a2    # Creates @a1, a1, @a2, a2
>      attr_writer :b1, :b2    # Creates @b1, b1=, @b2, b2=
>      attr_accessor :c1, :c2  # Creates @c1, c1, c1=, @c2, c2, c2=
>      # ...
>  def initialize
>    @c2 = 99
>  end
> end
> 
> obj = SomeClass.new
> p obj.instance_variable_defined?(:@c2)   # true
> p obj.instance_variable_defined?(:@d2)   # false
> p obj.instance_variable_defined?(:@a2)   # false  :-/

The behavior seems correct. The "rb_attr" method in eval.c only creates a reader method and a writer method. It cannot initialize an instance variable since it operates at the class level, not the instance level.

In order to create the behavior you expect, the attr_* methods would have to inject "magic" into the object initialization sequence. That is the only way a class level method could initialize an instance variable for every object instantiated from the class. I do not see any evidence of such magic in the C-level "rb_attr" method.

I believe you are getting a false value correctly.

  p obj.instance_variable_defined?(:@b2)    # false
  obj.b2 = 'foo'
  p obj.instance_variable_defined?(:@b2)    # true

Since the :a2 accessor is only creating a reader, you will never be able to create an instance variable in the normal fashion ...

  obj.instance_variable_set(:@a2, 'bar')
  p obj.instance_variable_defined?(:@a2)    # true

But that is an obvious example to the point of absurdity.

Blessings,
TwP