[CM] scoping in scheme
Wed, 29 Oct 2003 16:59:13 -0500
You can get this effect in Scheme using fluid-let (if your
implementation of Scheme supports fluid-let, or if you define it
(set! i (+ i 2))))
(define i -3)
(list (fluid-let ((i 1))
=> (3 -3)
Note the the define was moved to toplevel. According to the R5RS
specification, a define with in the body of a begin (or other lambda
expression) is going to be local to that body only. So in a strict
R5RS implementation you would get an error when calling test-proc
since i would be unbound. (See the section "Internal definitions" in
Guile seems to let you get away with this. Chez scheme does not.
MzScheme doesn't seem to like the internal define right after begin
at all. Except this works in MzScheme:
(define i -3)
(let ((i 1))
Anyway to complicate matters fluid-let is NOT part of R5RS. But it is
available in Chez scheme and MzScheme. And you can add fluid let to
Guile with a syntax expansion maco. See
Not sure what Scheme you are using.
Anyway the mechanism of fluid-let essentially creates a temporary
binding for the toplevel value of i. So it does not actually create
a new lexical variable. Instead it assigns the desired value to the
current lexical value, and then it restores the old value upon
exiting from the body of the fluid-let.
>hope this is not too off topic: Is it possible to define a top-level
>closure in scheme which acesses (and changes) variables from the
>dynamically scoped environment it is called in?
>In the following example i is bound at define time and its value in a
>function call is always taken from the definition's lexical context:
> (lambda ()
> (set! i (+ i 2))))
> (define i -3)
> (let ((i 1))
>=> (1 -1)
>It would be nice if the result could be made to be (3 -3) without
>having to supply i as arguments to test-func. Test-func is only
>evaluated for its side effects which should affect the variable's
>meanings in the current dynamic context of the function call.
>It might be either trivial or the wrong way to go about the whole
>thing but I thought I'd ask anyway.
>Cmdist mailing list