Re: [Sbcl-devel] bug in format - fixed-format floating point?

[Dave K. <ip...@gm...>]

Applies fine and works great.  Thanks.

On 4/25/06, Raymond Toy <ray...@er...> wrote:
> >>>>> "Dave" =3D=3D Dave Kern <ip...@gm...> writes:
>
>     Dave> In sbcl-0.9.11 running under linux on an x86 machine:
>     Dave> When I use format with "~F" and specify the number of fraction =
digits,
>     Dave> an extra digit is output if the input is a "small number". Hope=
fully
>     Dave> the following makes it clear:
>
>     Dave> (format t "~,2F" 0.001) =3D> 0.001   ;; should be 0.00
>     Dave> (format t "~,2F" 0.005) =3D> 0.005   ;; should be 0.00 or 0.01
>     Dave> (format t "~,2F" 0.006) =3D> 0.01     ;; correct
>
> CMUCL has a similar problem.  One possible fix is the following
> patch.  This should probably apply cleanly to sbcl, since they both
> have very similar fp printing routines.
>
> Ray
>
> In function flonum-to-string:
> Index: print.lisp
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> RCS file: /project/cmucl/cvsroot/src/code/print.lisp,v
> retrieving revision 1.108
> diff -u -r1.108 print.lisp
> --- print.lisp  17 Feb 2006 15:45:40 -0000      1.108
> +++ print.lisp  25 Apr 2006 13:24:15 -0000
> @@ -1539,7 +1539,11 @@
>                    (write-string "." stream)
>                    (dotimes (i (- e))
>                      (write-char #\0 stream))
> -                  (write-string string stream)
> +                  ;; If we're out of room (because fdigits is too
> +                  ;; small), don't print out our string.
> +                  (when (or (null fdigits)
> +                            (plusp (+ e fdigits)))
> +                    (write-string string stream))
>                    (when fdigits
>                      (dotimes (i (+ fdigits e (- (length string))))
>                        (write-char #\0 stream)))))
>
>