From: Alex Young <alex@...>
Date: 2007-12-14T18:45:24+09:00
Subject: Re: attr_accessor for question methods (?)

Charles Oliver Nutter wrote:
> Trans wrote:
>> Like to make a small suggestion for 1.9. It would be great if we could
>> do:
>>
>>   attr_accessor :foo?
>>
>> With the effect of:
>>
>>   def foo?
>>     @foo
>>   end
>>
>>   def foo=(x)
>>     @foo=x
>>   end
>>
>> This is much nice then having to define one attribute plus a
>> supplemental "?" method. attr_reader :foo? would work too, of course.
>> While it would be rather meaningless for attr_writer, it could work in
>> any case --really no need to raise an error.
>>
>> A minor detail, but one that would be convenient to have.
> 
> Perhaps an additional method would be the way to do it?
> 
> question_accessor
> predicate_accessor
> pred_accessor
> boolean_accessor
> bool_accessor

How about a single method that handles all possible accessor methods, 
rather than a proliferation of attr_*s and *_accessors?

   attr_set :foo, [:test, :read, :write]

There's room in there to make it user-extensible, too, so if I want (for 
example) a generic attribute method that resets my attribute to nil, I 
could do (something like):

   def_attr_setter :nullify, "%s_to_nil",
     Proc.new{|obj,sym| obj.instance_variable_set("@#{sym}", nil)}

then later:

   attr_set :bar, [:nullify]

which would give me something equivalent to this method:

   def bar_to_nil
     self.class.__attr_setters__[:nullify].call(:bar)
   end

Thoughts?  It's pretty ugly internally the way I've sketched it here, 
but I kind of like the API...

-- 
Alex