-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 03 May 1999 17:04:39 UT, Erik Naggum wrote:

>  (INTEGER 0 10) is actually a very specific subtype of FIXNUM.

Yes, you're right, of course.

>  try (setq excl::*loop-real-data-type* 'fixnum) and see if you win.

Yes, that works for fixnum loop counters (as is the case in "repeat"
loops). Thanks.

>  if you have ACL 5.0, the sources for LOOP macro should be in
>  #p"sys:src;loop.cl", or in #p"sys:src;cl;code;loop.cl" if you
>  have signed the support agreement with limited source code.

Yes, I've looked into it. What I find surprising is that I'd suppose

(loop for i from 1 to n do...)

or

(loop repeat n do...)

with a (fixnum-sized) integer "n" are a lot more usual than

(loop for i from 1.0 to 10.0 do...)

or

(loop repeat 10.0 do...)

or similar. So I would've expected for the loop macro to detect if
it's using statically defined fixnum constants and act as appropriate.
With "dotimes" we get:

USER(14): :ma (dotimes (i 10.0) (do-something))
(LET ((#:G116 10.0) (I 0))
  (DECLARE)
  (BLOCK NIL
    (TAGBODY
     #:|Tag22| (COND ((>= I #:G116) (RETURN-FROM NIL (PROGN))))
             (TAGBODY (DO-SOMETHING))
             (PSETQ I (1+ I))
             (GO #:|Tag22|))))

USER(15): :ma (dotimes (i 10) (do-something))
(LET ((I 0))
  (DECLARE (TYPE (INTEGER 0 10) I))
  (BLOCK NIL
    (TAGBODY
     #:|Tag23| (COND ((>= I 10) (RETURN-FROM NIL (PROGN))))
             (TAGBODY (DO-SOMETHING))
             (PSETQ I (1+ I))
             (GO #:|Tag23|))))

which at least seems more logical.

                                                       /L/e/k/t/u


-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.0.2i

iQA/AwUBNy660f4C0a0jUw5YEQLL4wCg5Y2Y0R6N+wBTk1pfFyxkJl2Xej4AnRJy
ngAV8B9Nh8wQbh8EvF6ko0Xh
=5JMV
-----END PGP SIGNATURE-----