From: Austin Ziegler <halostatue@...>
Date: 2005-04-25T21:53:00+09:00
Subject: Re: How are Ruby iterators implemented?

On 4/25/05, Roshan James <roshanj@microsoft.com> wrote:
> How are iterators implemented in Ruby?
> 
> Sometime back I had asked a questuin about invoking iterators in
> locksteps and I was told that iterators in ruby are implemented as
> closures that are passed to the called method. If iterators are
> passed as closures then, 1) Every consumer of the iterator should
> undergo closure conversion in some way 2) How is "break"
> implemented ? This similar to how it would be done in Scheme or
> similar language that inherently supports closures.
> 
> I was reading about CLU (the parent language for the iterator
> concept), and they implemnted iterators on a single stack. I
> believe that that would involve have caller and callee function
> frames to live on the stack at the same time. The iteration would
> be implemnetd as goto between these functions with a corresposding
> shift in stock location to refernce locals. This implementation
> has the inherent limitation of not being able to do multiple
> itetarors in lock step.
> 
> Another alternative is that the itetror is converted to an object
> that maintains state - a classical generator. This is what C# and
> Py do.
> 
> How does Ruby do iterators ?

What I'm getting ready to tell you is as far as I understand it; it
may not be completely correct -- and the most authoritative way to
see this is the source.

That said, simple iterators like #each implement a loop through the
container and yield the object in the iterator to the provided code
block. #each suspends execution until the block (closure) finishes
executing. That would generally indicate a change in stack frame.

Ruby does have generators available, but they're overkill for ~95%
of cases.

-austin
-- 
Austin Ziegler * halostatue@gmail.com
               * Alternate: austin@halostatue.ca