Forum: Ruby Extending object instances with <<

Announcement (2017-05-07): www.ruby-forum.com is now read-only since I unfortunately do not have the time to support and maintain the forum any more. Please see rubyonrails.org/community and ruby-lang.org/en/community for other Rails- und Ruby-related community platforms.
Tim B. (Guest)
on 2007-02-24 14:53
(Received via mailing list)
Hi,

I'm trying to write a class with a method to extend instances of
itself to contain additional accessors. I thought using `class <<
self` would be the most elegant way to go about it, but I'm running
into some problems. To illustrate:

class Test
  # takes an array of symbols to add to the instance.
  def add syms
    syms.each { |sym|
      @@__tmp = sym
      $__tmp = sym
      class << self
        #attr_accessor sym  # this would be my preferance, but sym
isn't in scope here
        #attr_accessor $__tmp  # this works, but uses globals
        attr_accessor @@__tmp   # this is nearly as bad as using globals
      end # <<
    } # each
  end # add
end # Test

t = Test.new
t.add [:thingie, :thingie2]

t.thingie="whatever"
t.thingie2="bla"
puts t.thingie
puts t.thingie2



I don't like the idea of using globals to transport the symbol
information and the class members approach is nearly as bad
(synchronization issues mainly, apart from elegance). But I can't
think of another way to transport dynamic data into the `class<<self`
block.

Alternatives would be to handle this using `method_missing` though
that wouldn't just affect a single instance or using `eval` which
would involve executing strings I'm banging together.

Another thing I tried was:

...
self.class.attr_accessor sym
...

but that doesn't work because `attr_accessor` is private (contrary to
what it says in the documentation...)

Any ideas? Am I missing something?
   -tim
Robert K. (Guest)
on 2007-02-24 15:11
(Received via mailing list)
On 24.02.2007 13:52, Tim B. wrote:
>    syms.each { |sym|
> end # Test
> information and the class members approach is nearly as bad
> (synchronization issues mainly, apart from elegance). But I can't
> think of another way to transport dynamic data into the `class<<self`
> block.

There is:

irb(main):017:0> class Bar
irb(main):018:1>   def add(*syms)
irb(main):019:2>     cl = class<<self;self;end
irb(main):020:2>     cl.instance_eval { attr_accessor *syms }
irb(main):021:2>   end
irb(main):022:1> end
=> nil
irb(main):023:0> f=Bar.new
=> #<Bar:0x3c1a40>
irb(main):024:0> f.add :bar
=> nil
irb(main):025:0> f.bar=10
=> 10
irb(main):026:0> f.bar
=> 10

> but that doesn't work because `attr_accessor` is private (contrary to
> what it says in the documentation...)
>
> Any ideas? Am I missing something?

See above.  Apart from that you could simply use OpenStruct or inherit
OpenStruct which does all this for you already automagically:

irb(main):013:0> require 'ostruct'
=> true
irb(main):014:0> f=OpenStruct.new
=> #<OpenStruct>
irb(main):015:0> f.bar=10
=> 10
irb(main):016:0> f.bar
=> 10

Major difference is that you do not explicitly control accessor creation
but automatically get *all* - even spelling errors.

Kind regards

  robert
Ken B. (Guest)
on 2007-02-26 04:10
(Received via mailing list)
On Sat, 24 Feb 2007 14:08:19 +0100, Robert K. wrote:

>>    syms.each { |sym|
>> end # Test
> => 10
> irb(main):016:0> f.bar
> => 10
>
> Major difference is that you do not explicitly control accessor creation
> but automatically get *all* - even spelling errors.

Test=Struct.new(*syms)
or
Test=Struct.new(:foo,:bar,:baz)

This will guard you against spelling errors.

--Ken
Robert K. (Guest)
on 2007-02-26 12:10
(Received via mailing list)
On 26.02.2007 02:38, Ken B. wrote:
> On Sat, 24 Feb 2007 14:08:19 +0100, Robert K. wrote:
>> Major difference is that you do not explicitly control accessor creation
>> but automatically get *all* - even spelling errors.
>
> Test=Struct.new(*syms)
> or
> Test=Struct.new(:foo,:bar,:baz)
>
> This will guard you against spelling errors.

That's true.  However, I interpreted the OP's posting that he needs to
to the extension on a per instance basis.  That does not work with the
approach you presented.

Kind regards

  robert
Giles B. (Guest)
on 2007-02-27 02:13
(Received via mailing list)
I think you're correct. The original poster's problem was to add
accessors to instances, splats are handy in the general case but
instance_eval is what you need here.
This topic is locked and can not be replied to.