DSL - class-level methods that affect instances

ruby-talk
1

I'm having a problem of design. I've tried a few configurations of class
collaboration and responsibility, but they all end up feeling messy.

I'm looking at these (either or, my original attempt was with a module,
but attr_accessor *looks* like it does what I want):

class C < Translator::Base
  translate 'this', :to => 'that'
end

class C
  include Translator
  translate 'this', :to => 'that'
end

With the expectation that...

c = C.new
c.translations

Would yield a hash of the from and to values.

I have no idea how to break the barrier between class and instance. :frowning:

Any ideas?

···

--
Posted via http://www.ruby-forum.com/.

2

Use class variables (the @@ variety).

class Translator::Base
  def self.translate word, params
    @@translations ||= {}
    @@translations[word]=params[:to]
  end
  def translations
    @@translations ||= {}
  end
end

class C < Translator::Base
  translate 'a', :to => 'b'
  translate 'c', :to => 'd'
end

C.new.translations #=> {"a"=>"b", "c"=>"d"}

···

On Thu, 01 May 2008 18:07:11 -0500, Daniel Waite wrote:

I'm having a problem of design. I've tried a few configurations of class
collaboration and responsibility, but they all end up feeling messy.

I'm looking at these (either or, my original attempt was with a module,
but attr_accessor *looks* like it does what I want):

class C < Translator::Base
  translate 'this', :to => 'that'
end

class C
  include Translator
  translate 'this', :to => 'that'
end

With the expectation that...

c = C.new
c.translations

Would yield a hash of the from and to values.

I have no idea how to break the barrier between class and instance. :frowning:

--
Ken (Chanoch) Bloom. PhD candidate. Linguistic Cognition Laboratory.
Department of Computer Science. Illinois Institute of Technology.
http://www.iit.edu/~kbloom1/

3

Hi --

I'm having a problem of design. I've tried a few configurations of class
collaboration and responsibility, but they all end up feeling messy.

I'm looking at these (either or, my original attempt was with a module,
but attr_accessor *looks* like it does what I want):

class C < Translator::Base
translate 'this', :to => 'that'
end

class C
include Translator
translate 'this', :to => 'that'
end

With the expectation that...

c = C.new
c.translations

Would yield a hash of the from and to values.

I have no idea how to break the barrier between class and instance. :frowning:

I'd say: don't think of it as any more of a barrier than exists
between any other two objects. Every object, including a class object,
has the right to maintain state, and to decide what it will and will
not expose. There's no privileged relationship between c and i just
because i is an instance of c.

(I know that sounds kind of doctrinaire, but I like to lay it on thick
since it usually takes some doing to chip away at the presumptive
class/instance special relationship :slight_smile:

As Ken suggested, class variables break through the class/instance
boundaries. That's one of the reasons I don't like them :slight_smile: You can
certainly do it that way; I'll also show you another possibility,
based on engineering all the objects concerned so that they query each
other where necessary:

module Translator
   attr_reader :translations

   def translate(target, actions)
     @translations ||=
     @translations.push("Translating #{target} to #{actions[:to]}")
   end
end

class C
   extend Translator

   def translations
     self.class.translations
   end

   translate "this", :to => "that"
end

p C.new.translations # ["Translating this to that"]

David

···

On Fri, 2 May 2008, Daniel Waite wrote:

--
Rails training from David A. Black and Ruby Power and Light:
   INTRO TO RAILS June 9-12 Berlin
   ADVANCING WITH RAILS June 16-19 Berlin
   INTRO TO RAILS June 24-27 London (Skills Matter)
See http://www.rubypal.com for details and updates!

4

You might be interested by Active Support's class inheritable attributes:

http://github.com/rails/rails/tree/master/activesupport/lib/active_support/core_ext/class/inheritable_attributes.rb

They do what (I think) you want.

Regards,
Andy Stewart

···

On 2 May 2008, at 00:07, Daniel Waite wrote:

I have no idea how to break the barrier between class and instance. :frowning:

-------
AirBlade Software

5

Ken Bloom wrote:

class C

I have no idea how to break the barrier between class and instance. :frowning:

Use class variables (the @@ variety).

class Translator::Base
  def self.translate word, params
    @@translations ||= {}
    @@translations[word]=params[:to]
  end
  def translations
    @@translations ||= {}
  end
end

class C < Translator::Base
  translate 'a', :to => 'b'
  translate 'c', :to => 'd'
end

C.new.translations #=> {"a"=>"b", "c"=>"d"}

That works! Except it shares translations across all subclasses.

module Translator
  class Base

    def self.translate word, params
      @@translations ||= {}
      @@translations[word]=params[:to]
    end

    def translations
      @@translations ||= {}
    end

  end
end

class C < Translator::Base

  translate 'a', :to => 'b'
  translate 'c', :to => 'd'

end

class D < Translator::Base

  translate 'this', :to => 'that'

end

irb(main):028:0> c = C.new
=> #<C:0x4d4e0>
irb(main):029:0> c.translations
=> {"a"=>"b", "c"=>"d", "this"=>"that"}

While I understand the behavior, its undesirable. I'm going to dust off
Ruby for Rails...

···

On Thu, 01 May 2008 18:07:11 -0500, Daniel Waite wrote:

--
Posted via http://www.ruby-forum.com/\.

6

David A. Black wrote:

I'd say: don't think of it as any more of a barrier than exists
between any other two objects.

That helps, actually. Viscerally, for now, but I'm sure the fullness of
the stance will sink in with time.

(I know that sounds kind of doctrinaire, but I like to lay it on thick
since it usually takes some doing to chip away at the presumptive
class/instance special relationship :slight_smile:

I love saturated opinions. (David West all the way!) I'm re-reading your
book right now starting from chapter 5, and yes, I can already tell I've
lost a forgotten a lot of my understanding of how Ruby works. (I suppose
you could argue I never understood it if I forgot...?)

module Translator
   attr_reader :translations

   def translate(target, actions)
     @translations ||=
     @translations.push("Translating #{target} to #{actions[:to]}")
   end
end

class C
   extend Translator

   def translations
     self.class.translations
   end

   translate "this", :to => "that"
end

p C.new.translations # ["Translating this to that"]

Works beautifully. Amazing. I thought I was going to end up with some
monolithic structure like ActiveRecord. :wink:

@Ken:

Your solution works equally well! I like that yours is class-based so I
don't have to call extend in the body of each inheriting class (of which
there will be quite a few).

I'm not worried about tertiary inheritance; I don't plan on needing it.
(Famous last words, right?)

Thank you both, Ken and David, very much!

···

--
Posted via http://www.ruby-forum.com/\.

7

I couldn't remember why I didn't use @@ variables for another very
similar metaprogramming problem I had this week. Now you've reminded why.

This version uses instance variables on the singleton class (which is
different from using @@ class variables)

class Translator::Base
  def self.translate word, params
    @translations ||= {}
    @translations[word]=params[:to]
  end
  def translations
    self.class.instance_variable_get(:@translations) || {}
  end
end

class C < Translator::Base
  translate 'a', :to => 'b'
  translate 'c', :to => 'd'
end

C.new.translations #=> {"a"=>"b", "c"=>"d"}

class D < Translator::Base
  translate 'e', :to => 'f'
end

D.new.translations #=> {"e"=>"f"}
C.new.translations #=> {"a"=>"b", "c"=>"d"}

#But beware:
class E < C
  translate 'g', :to => 'h'
end

E.new.translations #=> {"g"=>"h"}

There are ways to solve this by explicitly asking for your parent class's
translations, if you'd like to have that functionality.

--Ken

···

On Thu, 01 May 2008 19:55:45 -0500, Daniel Waite wrote:

Ken Bloom wrote:

On Thu, 01 May 2008 18:07:11 -0500, Daniel Waite wrote:

class C

I have no idea how to break the barrier between class and instance. :frowning:

Use class variables (the @@ variety).

class Translator::Base
  def self.translate word, params
    @@translations ||= {}
    @@translations[word]=params[:to]
  end
  def translations
    @@translations ||= {}
  end
end

class C < Translator::Base
  translate 'a', :to => 'b'
  translate 'c', :to => 'd'
end

C.new.translations #=> {"a"=>"b", "c"=>"d"}

That works! Except it shares translations across all subclasses.

module Translator
  class Base

    def self.translate word, params
      @@translations ||= {}
      @@translations[word]=params[:to]
    end

    def translations
      @@translations ||= {}
    end

  end
end

class C < Translator::Base

  translate 'a', :to => 'b'
  translate 'c', :to => 'd'

end

class D < Translator::Base

  translate 'this', :to => 'that'

end

irb(main):028:0> c = C.new
=> #<C:0x4d4e0>
irb(main):029:0> c.translations
=> {"a"=>"b", "c"=>"d", "this"=>"that"}

While I understand the behavior, its undesirable. I'm going to dust off
Ruby for Rails...

--
Ken (Chanoch) Bloom. PhD candidate. Linguistic Cognition Laboratory.
Department of Computer Science. Illinois Institute of Technology.
http://www.iit.edu/~kbloom1/

8

Daniel Waite wrote:

Thank you both, Ken and David, very much!

Here she is, best of both approaches!

module Translator

  module ClassMethods
    attr_reader :translations

    def translate(from_word, options)
      @translations ||= Hash.new
      @translations[from_word] = options[:to]
    end
  end

  def self.included(receiver)
    receiver.extend(ClassMethods)
  end

  def translations
    self.class.translations
  end

  class Base
    include Translator
  end

end

class NewTranslator < Translator::Base

   translate 'customer_profile_id', :to => 'customerProfileId'

end

p NewTranslator.new.translations

I feel a little odd putting the class Base declaration at the end of the
module, but if it's not there it has to be outside it entirely (and it
must still occur _after_ the code inside the module), something I don't
really want to do.

Anywho, thanks again for the awesome help. You guys rock!

···

--
Posted via http://www.ruby-forum.com/\.

9

Hi --

Ken Bloom wrote:

class C

I have no idea how to break the barrier between class and instance. :frowning:

Use class variables (the @@ variety).

class Translator::Base
  def self.translate word, params
    @@translations ||= {}
    @@translations[word]=params[:to]
  end
  def translations
    @@translations ||= {}
  end
end

class C < Translator::Base
  translate 'a', :to => 'b'
  translate 'c', :to => 'd'
end

C.new.translations #=> {"a"=>"b", "c"=>"d"}

That works! Except it shares translations across all subclasses.

module Translator
  class Base

    def self.translate word, params
      @@translations ||= {}
      @@translations[word]=params[:to]
    end

    def translations
      @@translations ||= {}
    end

  end
end

class C < Translator::Base

  translate 'a', :to => 'b'
  translate 'c', :to => 'd'

end

class D < Translator::Base

  translate 'this', :to => 'that'

end

irb(main):028:0> c = C.new
=> #<C:0x4d4e0>
irb(main):029:0> c.translations
=> {"a"=>"b", "c"=>"d", "this"=>"that"}

While I understand the behavior, its undesirable. I'm going to dust off
Ruby for Rails...

I couldn't remember why I didn't use @@ variables for another very
similar metaprogramming problem I had this week. Now you've reminded why.

This version uses instance variables on the singleton class (which is
different from using @@ class variables)

I wish that class variables were $$ instead of @@. They're really a
kind of restricted global, and they have the same qualities of sprawl
that globals do. The @@ always makes it look like they have some kind
of kinship with instance variables, when in fact they're practically
the opposite, in that they open up a shared space for storing state
whereas instance variables specifically make it possible for objects
to control state on an individual, private (give or take the fact that
nothing is really private in Ruby) basis.

David

···

On Fri, 2 May 2008, Ken Bloom wrote:

On Thu, 01 May 2008 19:55:45 -0500, Daniel Waite wrote:

On Thu, 01 May 2008 18:07:11 -0500, Daniel Waite wrote:

--
Rails training from David A. Black and Ruby Power and Light:
   INTRO TO RAILS June 9-12 Berlin
   ADVANCING WITH RAILS June 16-19 Berlin
   INTRO TO RAILS June 24-27 London (Skills Matter)
See http://www.rubypal.com for details and updates!