Re: [Scheme-reports] when and unless John Cowan (19 May 2011 21:42 UTC)
Re: [Scheme-reports] when and unless Pierpaolo Bernardi (20 May 2011 09:32 UTC)
Re: [Scheme-reports] when and unless Alex Queiroz (20 May 2011 09:38 UTC)
Re: [Scheme-reports] when and unless Alaric Snell-Pym (23 May 2011 10:36 UTC)
Re: [Scheme-reports] when and unless Eli Barzilay (23 May 2011 10:41 UTC)
Re: [Scheme-reports] when and unless Aaron W. Hsu (23 May 2011 22:50 UTC)
Re: [Scheme-reports] when and unless Pierpaolo Bernardi (24 May 2011 08:25 UTC)
Re: [Scheme-reports] when and unless Emmanuel Medernach (24 May 2011 08:37 UTC)
Re: [Scheme-reports] when and unless Alaric Snell-Pym (24 May 2011 09:06 UTC)
Re: [Scheme-reports] when and unless Aaron W. Hsu (24 May 2011 18:57 UTC)
Re: when and unless Arthur A. Gleckler (24 May 2011 19:35 UTC)
Re: [Scheme-reports] when and unless leppie (24 May 2011 08:51 UTC)
Re: [Scheme-reports] when and unless Pierpaolo Bernardi (24 May 2011 09:06 UTC)
Re: [Scheme-reports] when and unless leppie (24 May 2011 09:23 UTC)
Re: [Scheme-reports] when and unless John Cowan (24 May 2011 14:51 UTC)
Re: [Scheme-reports] when and unless Pierpaolo Bernardi (24 May 2011 14:57 UTC)
Re: [Scheme-reports] when and unless Aaron W. Hsu (24 May 2011 18:56 UTC)

Re: [Scheme-reports] when and unless leppie 24 May 2011 08:48 UTC

On Tue, May 24, 2011 at 10:25 AM, Pierpaolo Bernardi
<olopierpa@gmail.com> wrote:

> I use them, sometimes, when rewriting in scheme an algorithm
> written in imperative style in a foreign language.  For example,
> when taking an algorithm from a book.
>
> Sometimes it's worth the time to translate the code in good style,
> and sometimes the time is missing.  Luckily, with scheme we have
> the choice.
>
> Nevertheless, I'd prefer not to have these macros in the standard.
> And I'd strongly prefer not to have them in WG1 scheme.

There are good reasons to use UNLESS and WHEN, especially in terms of
error checking.

Take for example:

(define (foo bar)
  (unless (list? bar) (error 'foo "not a list" foo))
  (map values bar))

Now applying the above with say (foo 1) will raise an error. If I set
my debugger to break inside ERROR, I get a nice complete stacktrace,
including the caller, IOW FOO.

Now take what is suggested:

(define (foo bar)
  (if (not (list? bar)) (error 'foo "not a list" foo)
                           (map values bar)))

Now inside ERROR, you do not have 'lost' the caller (FOO) as the
application of ERROR was a tail call. Given the context of FOO's
application that may also be a tail call, and so on.

In my specific case (and might be the same with other implementations)
I generate a stacktrace (if enabled for the session) for every raised
error.

Using the first approach allows me to go up a (stack)frame and inspect
the FOO procedure accordingly.

My 2 cents.

Cheers

leppie

--
http://codeplex.com/IronScheme
http://xacc.wordpress.com

_______________________________________________
Scheme-reports mailing list
Scheme-reports@scheme-reports.org
http://lists.scheme-reports.org/cgi-bin/mailman/listinfo/scheme-reports