Embeddable Common-Lisp

On Feb 3, 2008 12:06 AM, Stelian Ionescu <sionescu@...> wrote:
> When compiling this function from KMRCL
>
> (defun remove-keyword (key arglist)
>   (loop for sublist = arglist then rest until (null sublist)
>      for (elt arg . rest) = sublist
>      unless (eq key elt) append (list elt arg)))
>
> ECL complains that "Iteration in LOOP follows body code" but the way I
> interpret section 6.1.4 of the Hyperspec is that termination clauses are
> *not* considered to be body clauses, so I think that LOOP shouldn't
> signal an error here

The hyperspec is rather clear about this issue:

loop [name-clause] {variable-clause}* {main-clause}* => result*

all for statements are variable-clause, hence they cannot appear after
a main-clause (until, unless, etc)

http://www.lisp.org/HyperSpec/Body/mac_loop.html

> SBCL, CMUCL, Clisp, GCL and ABCL all accept that and only Clisp signals a
> warning about it

LOOP is an ugly beast. I just stole that code and I doubt I can fix
that. But I insist it is not a bug, but a commonly accepted practice.

Juanjo

-- 
Facultad de Fisicas, Universidad Complutense,
Ciudad Universitaria s/n Madrid 28040 (Spain)
http://juanjose.garciaripoll.googlepages.com

On Feb 3, 2008 12:12 PM, Nicolas Neuss <neuss@...> wrote:
> "Juan Jose Garcia-Ripoll" <jjgarcia@...> writes:
>
> >> SBCL, CMUCL, Clisp, GCL and ABCL all accept that and only Clisp signals a
> >> warning about it
> >
> > LOOP is an ugly beast. I just stole that code and I doubt I can fix
> > that. But I insist it is not a bug, but a commonly accepted practice.
>
> As much as I remember, also Allegro did not accept this construct, and I
> had to change my code to remain portable.

It is quite possible that both compilers drink from the same sources.
Most implementations of LOOP come from the same original one which
motivated the spec.

In my opinion the strict format is useful: it shows that the iteration
statements are implemented and executed before the termination
conditions and that these can be evaluated in a different order.

Juanjo

-- 
Facultad de Fisicas, Universidad Complutense,
Ciudad Universitaria s/n Madrid 28040 (Spain)
http://juanjose.garciaripoll.googlepages.com