Rails - Mar 2006 - Changes to class caching between Rails 0.14.3 and 1.0?

Hi all,

We''re having a rather unique problem with class caching. We are 
implementing the strategy pattern in the following manner:

1. Set an accessor for the class, let''s call it :strategy

  class Widget
    cattr_accessor :strategy
  end

2. In environment.rb, we instantiate an object that implements the 
strategy and use the accessor to set it for the class

  strat = MyStrategy.new
  Widget.strategy = strat

3. In the regular flow of the application, that strategy gets accessed 
so it can affect some behavior:

  ...
  Widget.strategy.do(:foo)
  ...

OK, now in development mode, class caching is off. In Rails 0.14.3 this 
didn''t seem to matter -- the class variable stayed set. But now 
mysteriously in Rails 1.0 I will get an exception when trying to call 
.do() on nil... ie, Widget.strategy becomes nil at some point. I''m 
assuming the class gets unloaded and then reloaded without :strategy 
ever getting set again. That''s likely because the environment never got
to set up the class variable again.

So, in 0.14.3, did the environment get evaluated again? Or did the class 
never really get unloaded?

Does anyone know a way I can avoid this problem? Is there a way to 
control class caching on a per-class basis? Looking at "module 
Dependencies" makes me believe there is not.

Thanks for any help,
rusty

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

> Does anyone know a way I can avoid this problem? Is there a way to
> control class caching on a per-class basis? Looking at "module
> Dependencies" makes me believe there is not.
>
> Thanks for any help,
> rusty
Try this on edge rails...

class Widget < AR::Base
  def self.reloadable?() false; end
end

--
Rick Olson
http://techno-weenie.net

> Try this on edge rails...
>
> class Widget < AR::Base
>   def self.reloadable?() false; end
> end
>
Doh, sorry, I read that as 1.1.

Rick Olson wrote:
>> Try this on edge rails...
>>
>> class Widget < AR::Base
>>   def self.reloadable?() false; end
>> end
>>
> 
> Doh, sorry, I read that as 1.1.
Hi Rick,

Thanks for the reply! Are you saying that this problem can be addressed 
if I use Rails 1.1?

Thanks,
rusty

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