Re: [Chicken-users] Binding identifiers with set!

[felix]

On Sun, 12 Oct 2003 10:19:49 +0100, Category 5 <address@hidden> 
wrote:

> In <address@hidden> on
> comp.lang.scheme, Anton van Straaten pointed out that set! can be used
> to introduce new bindings in Chicken:
>
> % csi
> ; This is the CHICKEN interpreter.
> ; Version 1, Build 17 - netbsd-gnu-ultrasparc
> ; (c)2000-2003 Felix L. Winkelmann
> > > > foo
> Error: unbound variable: foo
> > > > (set! foo 1)
> > > > foo
> 1
> > > >
>
> This clearly doesn't jive with R5RS (5.2.1):
>
> At the top level of a program, a definition
>
> (define <variable> <expression>)
>
> has essentially the same effect as the assignment expression
>
> (set! <variable> <expression>)
>
> if <variable> is bound. If <variable> is not bound, however, then
> the definition will bind <variable> to a new location before
> performing the assignment, whereas it would be an error to perform a
> set! on an unbound variable.

> From a language-lawyerly point of view, "it would be an error" allows
an implementation to do whatever it likes. So this is strictly speaking
not a violation of R5RS.

> How do people feel about this?  I find the behaviour a bit worrying, as
> it amounts to the omission of a basic form of error checking supported
> by R5RS as well as every other implementation I'm aware of.  I can
> imagine situations where this would bite.

Yes, I agree with that. It makes finding bugs harder.

> My first choice would be to see set! changed to support the more
> restrictive semantics, but failing that, I think the current behaviour
> should be explicitly documented in the "Deviations from the standard"
> section of the manual.

Changing the semantics of `set!' is currently not easy to fix. But I have
to think about that some more...
I'll document the behaviour int the manual. Thanks for the suggestion.


cheers,
felix