From: Ara.T.Howard@...
Date: 2005-05-05T05:44:49+09:00
Subject: Re: [RCR] Object#inside_metaclass?

On Thu, 5 May 2005, David A. Black wrote:

> Hi --
>
>>  it's sort hard to show in a little code but __logically__ we end up
>>  with something like (greatly simplified)
>>
>>    class C
>>
>>      @trait_defaults = {
>>        'class' => { 'a' => 'default_value' },
>>        'instance' => { 'a' => 'default_value' },
>>      }
>>
>>      def C::a
>
> (Just out of curiosity: do you mean to depart from the usual C.a here?
> And why? :-)

i've got a couple of coding conventions i use including:

   - collections are plural
   - class methods get called with ::

they help me know things like

   p configs                 #=> this is an array of configs

   x::send('method', *args)  #=> x is a class

they are just habbits.

>
>>        @a = @trait_defaults['class']['a'] unless defined? @a
>>        @a
>>      end
>>
>>      def a
>>        @a = @trait_defaults['instance']['a'] unless defined? @a
>>        @a
>>      end
>>    end
>>
>>  so i need to know, at the point of definition if we're inside a
>>  metaclass to generate the code slightly differently.  all the above
>>  is for explanation only and is nothing like real code.
>
> I understand the not real code point, but still, it's showing something I
> can't quite follow, namely the @trait_defaults instance variable in the
> instance method definition (def a).  Do you mean this to refer to the
> class's instance variable?

yes - sorry.  both class methods AND instance methods must look up there
respective defaults in a class instance var.

you can read the real code at

   http://codeforpeople.com/lib/ruby/traits/traits-0.0.0/

> I'm also not sure where inside_metaclass? would go here.  Can you rewrite it
> into pseudo-code looking how it would look if that method existed?

   class Class
     def trait(*args)
       if inside_metaclass?
         define_class_traits(*args)
       else
         define_instance_traits(*args)
       end
     end
     def class_trait(*args)
       class << self
         trait(*args)
       end
     end
   end

make sense?  obviously this would be easy if traits were like 'attr' and just
eval'd some code - but the defaults thing makes it hard since the defaults
must be stored somewhere in the case of instance traits and looked up later -
therefore the method definition of the trait itself must dynamically differ.
this is unlike attr_ methods.  another reason for inside_metaclass? is that
traits works like

   jib:~ > cat a.rb
   require 'traits'
   class C
     class_trait 'a' => 42
     class << self
       trait 'b' => 'forty-two'
     end
     trait 'a' => 42.0
   end

   p C::reader_traits
   p C::class_reader_traits

   jib:~ > ruby a.rb
   ["a"]
   ["a", "b"]

now, 'reader_traits' and 'class_reader_traits' are also implemented like

   def reader_traits(*args)
     ...
   end
   def class_reader_traits(*args)
     class << self
       reader_traits(*args)
     end
   end

eg. the __return__ value of reader_traits is context sensitive to being in/out
of a metaclass.

reason three:

   jib:~ > cat a.rb
   class C
     p ancestors
     class << self
       p ancestors
     end
   end

   jib:~ > ruby a.rb
   [C, Object, Kernel]
   [Class, Module, Object, Kernel]

now imagine you want to implement 'inheritence' for traits that use class @vars
(not @@vars) for default values (for the obvious reason).  you need to look up
the chain of ancestors to find you defaults. problem: your ancestors vary
according to scope.  i solve this by

   klass =
     if inside_metaclass?
       pop_up_to_instance_klass
     else
       self
     end

   klass.ancestors.each{|a| look_for_default a }

the problem all boils down to this : attr_ and friends work exactly the same
whether called at a class or metaclass level.  adding features like having
default values, inheritence of stuff, etc.  makes the distinction of where you
are (class or metaclass) suddenly very important.

i hope i'm making myself clear now - i never meant for this to end up like
this!  ;-)

cheers.

-a
-- 
===============================================================================
| email :: ara [dot] t [dot] howard [at] noaa [dot] gov
| phone :: 303.497.6469
| renunciation is not getting rid of the things of this world, but accepting
| that they pass away. --aitken roshi
===============================================================================