r5rs section 6.2.3 inexactness and rational sqrt

r5rs section 6.2.3 inexactness and rational sqrt

[Ray Lehtiniemi]

hi folks

it's been a while since i looked at scheme, and i decided to build
a copy of guile-1.6.4 and try writing some simple apps with it. i've
got a few issues to report with numeric handling:

first, i have an issue with (ice-9 format):

  guile> (format #f "~4,48x" (floor 1234))
  "1234.0"

not exactly what i had in mind. maybe it doesn't like the output
of (floor n)?

  guile> (floor 1234)
  1234.0

i'd expect that to be an exact integer, and indeed r5rs section 6.2.3
says that (floor n) should return an exact integer as long as the input
is exact:

  guile> (exact? 1234)
  #t

  guile> (exact? (floor 1234))
  #f

  guile> (exact? (floor (inexact->exact 1234)))
  #f

hmm....

maybe the inexactness from floor is causing the problem in
(ice-9 format)?

  guile> (format #f "~4,48x" (inexact->exact (floor 1234)))
  "04d2"

now that's more like it, but the forced exactness shouldn't be neccessary
according to the standard.


finally, on an unrelated note:

  guile> (rational? (sqrt 6))
  #t

should be #f, i think....



-- 
----------------------------------------------------------------------
     Ray L   <rayl@mail.com>


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-user

Re: r5rs section 6.2.3 inexactness and rational sqrt

[Kevin Ryde]

Ray Lehtiniemi <rayl@mail.com> writes:
>
>   guile> (floor 1234)
>   1234.0
>
> i'd expect that to be an exact integer, and indeed r5rs section 6.2.3
> says that (floor n) should return an exact integer as long as the input
> is exact:

Yep, that looks wrong.  Marius applied a fix not so long ago in the
head, lobby him for the same in 1.6.

>   guile> (rational? (sqrt 6))
>   #t
>
> should be #f, i think....

Alas, sqrt 6 comes out as a flonum, which is rational.


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-user

Re: r5rs section 6.2.3 inexactness and rational sqrt

[Ray Lehtiniemi]

On Sat, Apr 17, 2004 at 09:23:40AM +1000, Kevin Ryde wrote:
> Ray Lehtiniemi <rayl@mail.com> writes:
> >
> >   guile> (floor 1234)
> >   1234.0
> >
> > i'd expect that to be an exact integer, and indeed r5rs section 6.2.3
> > says that (floor n) should return an exact integer as long as the input
> > is exact:
> 
> Yep, that looks wrong.  Marius applied a fix not so long ago in the
> head, lobby him for the same in 1.6.

well, it's not super critical (for me anyway) since i have the workaround.
any idea when the next stable series is due?



> Alas, sqrt 6 comes out as a flonum, which is rational.

bummer.  seems that MIT/GNU scheme 7.7.1 does the same thing too, though....


-- 
----------------------------------------------------------------------
     Ray L   <rayl@mail.com>


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-user