Re: ‘truncate’ on (float-time) causes arith range error on 32bit emacs

[Göktuğ Kayaalp]

On Fri, 29 Jan 2016 10:49:27 +0200, Eli Zaretskii <eliz@gnu.org> writes:
> No.  The usual ways of handling this are either (a) leave the numbers
> as floats, and just be more careful about comparisons; and (b)
> represent large numbers as cons cells.
> 
> Since these are time values, I'd suggest (a) in your case.  What math
> requires integers, exactly, and why?

Hi, I guess (a) will work, I just sent a patch to the user to test (I do
not have a 32bit machine, though the math worked).  This is what the
function does:

  (defun forecast--sun-position-graphic ()
    "Visualise the time since the rise of the sun and the time to the set 
thereof.
  
  E.g.:
  
  Quasi-midday:
  >————————☉———————————<
  Sunrise:
  ☉————————————————————<
  Sunset:
  >————————————————————☉"
    (let* ((today   (aref (forecast--assoca '(daily data) forecast--data) 0))
           (sunrise (forecast--assoca '(sunriseTime) today))
           (sunset  (forecast--assoca '(sunsetTime) today))
           (now     (truncate (float-time)))
           (daylen  (- sunset sunrise))
           (sunsec  (- now sunrise))
           (wwidth  (window-body-width))
           (graph   (concat ">" (make-string (- wwidth 5) ?—) "<"))
           (sun     ?☉)
           (pos    (cond
                    ((< sunrise sunset now) (- wwidth 4))
                    ((> sunrise now) 0)
                    (t (1- (/ sunsec (/ daylen wwidth)))))))
      (aset graph pos sun)
      graph))

The values ‘sunrise’ and ‘sunset’ come  from an API and they are integer
timestamps.

Sorry for the  late reply, I somehow  did not receive the  replies to my
message, I've some  problems with e-mail.  Please CC me  directly if you
reply, I'm not subscribed to the list.

Thanks a lot,
 -gk.

-- 
İ. Göktuğ Kayaalp.
http://gkayaalp.com/