MathGroup Archive 1999

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

Search the Archive

Re: bug in 4.0

  • To: mathgroup at smc.vnet.net
  • Subject: [mg18585] Re: [mg18566] bug in 4.0
  • From: David Withoff <withoff at wolfram.com>
  • Date: Tue, 13 Jul 1999 01:01:23 -0400
  • Sender: owner-wri-mathgroup at wolfram.com

>Can anybody duplicate the following error?
>Wolfram say they cannot replicate it.
> 
> In[1]:=
> $Version
> 
> Out[1]=
> "4.0 for Power Macintosh (April 20, 1999)"
> 
> In[2]:=
> LegendreP[2, -Cos[10*N[Degree, 16]]]
> 
> Out[2]=
> 0.954769
> 
> In[3]:=
> LegendreP[2, -Cos[10*N[Degree, 17]]]
> 
> Out[3]=
> 0.47738473279471564
> 
> Thank you
> jonathan connor
> --
>******************************************************************************
> Professor J.N.L. Connor,     Phone(direct line): 0161-275-4693 (national)
> Department of Chemistry,                     :+44-161-275-4693 (international)> University of Manchester,
> Manchester M13 9PL,            Phone(secretary): 0161-275-4686 or 4600
> England.                                    Fax: 0161-275-4734 or 4598 
>******************************************************************************

I don't know why there was any difficulty reproducing this behavior.
It is easy to reproduce, and has already been reported to the people
at Wolfram Research who are responsible for this function.  Perhaps
the person who tried to reproduce it initially made a typo or something
or accidentally tried it in a different version of Mathematica.  In any
case, this behavior has already been reproduced and forwarded to the
appropriate people.

One workaround is to avoid doing this calculation with machine arithmetic.
If you wish you could add a rule to LegendreP to automatically convert
machine number arguments into variable-precision arguments.

Dave Withoff
Wolfram Research


  • Prev by Date: Re: Simplifying constants
  • Next by Date: Re: LaTex on a Mac
  • Previous by thread: bug in 4.0
  • Next by thread: Parametric curve fitting