From: hipster <hipster@...4all.nl>
Date: 2000-12-01T17:46:10+09:00
Subject: [ruby-talk:6697] Re: each/for binding confusion and resolution

On Fri, 01 Dec 2000  03:10:02 +0900, Steven Grady wrote:
> hipster responds:
> >Would, a construct similar to Thread.new,
> >
> >	proc([arg*]) { |args| block } -> aProc
> >
> >reduce the surprise? (I may overlook some binding semantics here, just
> >an idea)
> 
> Interesting.

In the Hacker's Dictionary interpretation, I presume?   :)

> I'm not sure if you meant that the "arg*" list would be passed on
> as the block's "args" list.  If so, there's a problem, because blocks
> already take a set of arguments.  An alternative interpretation is simply
> that the arguments of proc() are to be used locally (i.e. the block
> gets their value, not the variable itself).

The alternative, that's exactly what I meant.

> One of my original questions was whether you could achieve
> my intent using a <for> loop instead of an <each> iterator.
> Using the above syntax (with my alternative semantics), it would
> be something like:
>     procs = []
>     for v in [1, 2]
> 	procs << proc(v) { v }
>     end
> 
> By putting the "v" in the argument list of proc(), I am indicating that
> the v should be derefenced now, rather than when the Proc is called.

Leaving the |'s out is an even better idea. Instead of dereferencing,
I think a reference to v.clone should be stored in the proc's binding
instead of v itself.

> This would also address the scoping problem.  So even if v had been
> previously defined, you could still do this with <each>:
>     v = nil		# Note that this prevents v from being block-local
>     procs = []
>     [1, 2].each {|v|
> 	procs << proc(v) { v }
>     }
> 
> Unfortunately, I'm guessing there is no way to reimplement such a change
> within Ruby; Proc.new would have to be changed inside the interpreter.
> (Perhaps it could be done in Ruby if the Binding class had instance methods
> to modify the bindings, but that's a mere dream..)

The vector [scoping rules, notation, backward compatibility] has been
challenged before... Maybe something for Ruby 1.8?

	Michel