From: ara howard <ara.t.howard@...>
Date: 2008-03-12T00:48:03+09:00
Subject: Re: setting local variables via eval


On Mar 11, 2008, at 12:23 AM, Joel VanderWerf wrote:

> ara howard wrote:
>>  http://drawohara.com/post/28514698
>
> Nice hack, as usual :)
>
> Isn't
>
>  class Object
>    def scope
>      lambda{  }
>    end
>  end
>
> enough? You don't need to return binding from the lambda, AFAICT.  
> The lambda never gets called.

that's what i head originally - but blows up in 1.9 which requires a  
binding

>
>
> Also, an alternative to mucking with ObjectSpace:
>
>        Thread.current[:val] = v
>        definition = "#{ var } = Thread.current[:val]"
>
> in place of
>
>        value = "ObjectSpace._id2ref #{ v.object_id }"
>        definition = "#{ var } = #{ value }"
>
> Of course, the value of Thread.current[:val] should be saved and  
> restored. Maybe this makes the hack jruby friendly.
>

oh very nice.  yeah, pushing and popping from Thread.current  is very  
good

> There is one case in which the double_quoted_heredoc is not  
> hygienic, when a line of @template begins with __template__:
>
>  view2 = View.new '__template__'
>  view3 = View.new  view2.render
>  puts view3.render
>
> But I don't see any hygienic solution (indenting and using <<-  
> doesn't work, because the interpolated strings might have line  
> breaks). Maybe ruby needs a heredoc variant that runs to eof.
>



yes.  in real code i generate the hdoc in a loop roughly like so

   loop {
     hdoc =  
"__a 
#{ rand(2**30).to_i }__b#{ rand(2**30).to_i }__c#{ rand(2**30).to_i }__"

    break unless template =~ Regexp.escape(hdoc)
   }

etc

great suggestions - the jruby/Thread.current esp.

thx.

a @ http://drawohara.com/
--
sleep is the best meditation.
h.h. the 14th dalai lama