Instance variable: self vs @

Question

Writing @age directly accesses the instance variable @age. Writing self.age tells the object to send itself the message age, which will usually return the instance variable @age — but could do any number of other things depending on how the age method is implemented in a given subclass. For example, you might have a MiddleAgedSocialite class that always reports its age 10 years younger than it actually is. Or more practically, a PersistentPerson class might lazily read that data from a persistent store, cache all its persistent data in a hash.

The difference is that it is isolating the use of the method from the implementation of it. If the implementation of the property were to change -- say to keep the birthdate and then calculate age based on the difference in time between now and the birthdate -- then the code depending on the method doesn't need to change. If it used the property directly, then the change would need to propagate to other areas of the code. In this sense, using the property directly is more fragile than using the class-provided interface to it.

Be warned when you inherit a class from Struct.new which is a neat way to generate an intializer (How to generate initializer in Ruby?)

class Node < Struct.new(:value)
    def initialize(value)
        @value = value
    end
    def show()
        p @value
        p self.value # or `p value`
    end
end 

n = Node.new(30)
n.show()

will return

30
nil

However, when you remove the initializer, it will return

nil
30

With the class definition

class Node2
    attr_accessor :value
    def initialize(value)
        @value = value
    end
    def show()
        p @value
        p self.value
    end
end

You should provide the constructor.

n2 = Node2.new(30)
n2.show()

will return

30
30

The first answer is entirely correct, but as a relative newbie it wasn't immediately clear to me what it implied (sending messages to self? uh huh...). I think that a short example will help:

class CrazyAccessors
  def bar=(val)
    @bar = val - 20 # sets @bar to (input - 20)
  end
  def bar
    @bar
  end

  def baz=(value)
    self.bar = value # goes through `bar=` method, so @bar = (50 - 20)
  end

  def quux=(value)
    @bar = value     # sets @bar directly to 50
  end
end

obj  = CrazyAccessors.new
obj.baz = 50
obj.bar  # => 30
obj.quux = 50
obj.bar  # => 50

There isn't any difference. I suspect that it was done just for the documentary value of seeing self.age and other_person.age near each other.

I suppose that use does allow for an actual getter to be written in the future, which might do something more complex than just return an instance variable, and in that case the method would not need to change.

But that's an unlikely abstraction to worry about, after all, if the implementation of the object changed it's reasonable to change other methods, at some point a simple reference within the object itself is perfectly reasonable.

In any case, abstraction of the age property still doesn't explain the explicit use of self, as just plain age would also have invoked the accessor.