Re: [Chicken-users] Weird scope of variable defined inside cond

[Mario Domenech Goulart]

Hi Vitor,

On Tue, 12 May 2015 02:44:30 -0300 Vítor De Araújo <address@hidden> wrote:

> I'm seeing some weird behavior regarding scope of a variable defined
> with 'define' inside a cond clause. For instance:
>
>   (define (foo x)
>     (define bar 23)
>     (cond [(number? x)
>            (define baz 42)
>            (list baz)]
>           [else 'whatever]))
>
>   (foo 3)
>   (print baz)     ; prints 42!
>   (print bar)     ; unbound variable, as expected
>
> Additionally, if I type the above definition of 'foo' directly in csi,
> I get a message:
>
>   Note: the following toplevel variables are referenced but unbound:
>
>     baz (in foo)
>
> Is this a bug, or I'm missing something? I'm running Chicken 4.9.0.1 on
> Debian GNU/Linux x86-64. I also tried it with the version from the git
> repository and got the same result.

That's a little can of worms. :-)

The definition of `baz' is not valid in that context according to R5RS,
since `define' forms are not expressions (they are definitions), and
`cond' clauses should contain expressions:
http://www.schemers.org/Documents/Standards/R5RS/HTML/r5rs-Z-H-7.html#%_sec_4.2.1

CHICKEN implements `define' with `set!', and `set!' in CHICKEN can be
applied to unbound variables, in that case creating toplevel ones:
http://wiki.call-cc.org/man/4/Extensions%20to%20the%20standard#set

So, that code ends up creating a toplevel binding for `baz':

$ csc -debug 2 foo.scm
[canonicalized]
(##core#callunit "library")

(##core#callunit "eval")

(##core#callunit "chicken_2dsyntax")

(##core#undefined)

(set! foo
  (##core#lambda
    (x1)
    (let ((bar6 (##core#undefined)))
      (let ((t9 (set! bar6 '23)))
        (if (number? x1) (let ((t8 (set! baz '42))) (list baz)) 'whatever)))))

(foo '3)

(print baz)

((##sys#implicit-exit-handler))

(##core#undefined)


Best wishes.
Mario
-- 
http://parenteses.org/mario