From: Xavier Noria <fxn@...>
Date: 2006-11-11T09:52:46+09:00
Subject: questions about Class#remove_class implementation

I am reading Rails's Class#remove_class:

   def remove_class(*klasses)
     klasses.flatten.each do |klass|
       # Skip this class if there is nothing bound to this name
       next unless defined?(klass.name)

       basename = klass.to_s.split("::").last
       parent = klass.parent

       # Skip this class if it does not match the current one bound  
to this name
       next unless parent.const_defined?(basename) && klass =  
parent.const_get(basename)

       parent.send :remove_const, basename unless parent == klass
     end
   end

and have a few questions:

1. Do you think there's a reason for using Module#name
    first and Module#to_s afterwards? Looks to me that
    they are only different for singleton classes, and
    for those ones split does not provide a meaningful
    chunk anyway:

    irb(main):002:0> klass = (class << ActiveRecord::Base; self; end)
    => #<Class:ActiveRecord::Base>
    irb(main):003:0> klass.name
    => ""
    irb(main):004:0> klass.to_s.split("::").last
    => "Base>"

2. In which situation do you think parent.const_defined?(basename)
    && klass = parent.const_get(basename) could return false?

3. Does parent.send :remove_const effectively unload the
    class definition from the interpreter? If yes, what
    would happen to some dangling instance afterwards if
    some method of that class was invoked? Would the object
    still hold some kind of relationship to the now non-
    existent class that could result in some fatal error?

4. When could parent == klass hold at the end of the block?

-- fxn

PS: Rails defines Module#parent as

   def parent
     parent_name = name.split('::')[0..-2] * '::'
     parent_name.empty? ? Object : parent_name.constantize
   end