In http://rubykaigi.org/2013/talk/S58 @Shogu describes how refinements work, but says he would rather they work with a block-like syntax. I asked why he couldn't just implement these extensions in pure ruby using the same techniques that stub and mock libraries use. The translated answer was "Because I want to refinement only to apply inside the…

block_refinement.rb

class BlockRefinement

  def self.refines(*classes_refined)
    @classes_refined = classes_refined
  end

  def self.refines_method(method_refined, &new_method)
    @classes_refined.each do |class_refined|
      old_method = class_refined.instance_method(method_refined) 
      @reverts ||= []
      @reverts << Proc.new do
        class_refined.instance_eval do
          define_method(method_refined, old_method)
        end
      end
      @applies ||= []
      @applies << Proc.new do
        class_refined.instance_eval do
          define_method(method_refined, new_method)
        end
      end
    end
  end

  def self.using_refinement(&context)
    @applies.each(&:call)
    yield
  ensure
    @reverts.each(&:call)
  end

end

class MathRefinement < BlockRefinement
  refines Float, String
  refines_method(:*) do |arg|
    "(#{self.to_s}) * (#{arg.to_s})"
  end
  refines_method(:+) do |arg|
    "(#{self.to_s}) + (#{arg.to_s})"
  end
end

class MyDSL

end

class Object
  def refined_instance_eval(using, &block)
    using.using_refinement do
      instance_eval(&block)
    end
  end
end

puts 3.14159 * 5 + 10

MyDSL.refined_instance_eval(MathRefinement) do
  puts 3.14159 * 5 + 10
end

puts 3.14159 * 5 + 10

# OUTPUT:
# 25.70795
# ((3.14159) * (5)) + (10)
# 25.70795