Re: [Chicken-users] eval and local variables

[felix winkelmann]

On 5/6/07, Chicken Monk <address@hidden> wrote:
> Hi,
>
> I noticed that the following doesn't work in Chicken:
>
> (define a 42)
>
> (define (foo b)
>      (let ((c (+ b 1)))
>          (let ((env (interaction-environment)))
>              (eval '(+ a b c) env))))
>              ; doesn't work with or without the env.
>
> (print (foo 3))
>
> Error: unbound variable: b
>
> I would expect eval to "know" about b and c as well.  The manual,
> however, points out that it uses (interactive-environment) by default.
>
> A few questions:
>
> 1. Why is (interactive-environment) the default?  (Or, why doesn't eval
> have access to all the variables that "regular" code would have?)

It does, but only to global variables.

> 2. Is it possible to get the "local environment" (which should have
> access to b and c)?  Is there even such a thing, or am I seeing this
> completely wrong?

This is not possible in general, unless you really want to peek
into the internal representation of a closure.

> 3. Are there other ways to make this work?  I am aware of the
> environments egg, but having to manually add values to an environment
> does not sound very appealing. :-(

That's understandable, but the environments egg is generally intended
to handle custom
environments for top-level variables (and that only in interpreted code).

> Of course, it's possible that I completely misunderstand the subject, in
> which case I would love to be enlightened. :-)

Well, actually it would be mightily handy! See for example:

http://www.bcl.hamilton.ie/~qobi/map-closure/

But the implementation of such a facility severely constrains the
implementation,
and thus is quite expensive (I believe, but who knows - people said this
of continuations, once...)


cheers,
felix