Services & Resources / Wolfram Forums
-----
 /
MathGroup Archive
1997
*January
*February
*March
*April
*May
*June
*July
*August
*September
*October
*November
*Archive Index
*Ask about this page
*Print this page
*Give us feedback
*Sign up for the Wolfram Insider

MathGroup Archive 1997

[Date Index] [Thread Index] [Author Index]

Search the Archive

Abs and Sign anomolies

  • To: mathgroup at smc.vnet.net
  • Subject: [mg7398] Abs and Sign anomolies
  • From: Jack Goldberg <jackgold at math.lsa.umich.edu>
  • Date: Fri, 30 May 1997 01:20:09 -0400 (EDT)
  • Organization: Mathematics Department, University of Michigan
  • Sender: owner-wri-mathgroup at wolfram.com

Hi group;

A while ago hucka at eecs.umich.edu posted a question about a strange
result, namely  Abs[I*u*Sign[v]] returns Abs[u] on a Unix system using
Ver 3.  This heart of this anomaly can be traced to the following output
of Mma:

In[1]	Abs[Sign[x]] 
Out[1]  1

which is wrong if  x = 0.  For, Sign[0] = 0 and Abs[0] = 0.  So that the
output of  Abs[Sign[x]] should be unsimplified. Indeed, Mma returns

	Abs[Sign[0]] = 0

Although multiple crashes (hard and soft) and exporting graphics are
extremely annoying, a mathematical "bug" is a far more serious problem.
A programming error like this can lead to disaster and I urge the
newsgroup to post all such anomalies as soon as they are discovered. We
cannot assume that contacting Wolfram Research will result in a warning
in their FAQ site, nor a timely fix.  

Wouldn't it be nice if someone (other than Wolfram Research)) could keep
track of all mathematical "bugs" much the way one of our people keep
track of relative speeds of the kernel on different platforms?  The
advantages for all of us are so clear that no argument need be given.

I would be willing to do this, but - I'm embarrasses to admit - I don't
know how to set up a Web page or whatever so that people could access
the complilations automatically!  If anyone wishes to send me their
discoveries I will be happy to confirm them and keep a list. 

Jack Goldberg
University of Michigan 
Ann Arbor Michigan
e-mail:  jackgold at math.lsa.umich.edu


  • Prev by Date: Re: How do you make Mma assume a parameter is real and positive?
  • Next by Date: Re [mg 7343] How to make Sqrt[a^2]/;(a>
  • Previous by thread: Re: How do you make Mma assume a parameter is real and positive?
  • Next by thread: Re: Abs and Sign anomolies