Re: ver 1.9.4 use more decimal digits

["Andreas J. Guelzow" <aguelzow pyrshep ca>]

On Mon, 2009-03-02 at 07:57 +0800, bill lam wrote:
> On Sun, 01 Mar 2009, Morten Welinder wrote:
> > > After upgraded to 1.9.4. Diff the gnumeric file with the previous
> > > 1.8.3
> >
> > This is deliberate.  We need to store enough decimals to ensure
> > that the number we had will come back unchanged.  The number
>
> This is incorrect. The number pairs have the same bit pattern using
> ieee 754 standard so that they are equal.

The ieee 754 does _not_describe a single bit pattern for every number.
You can't even decide whether a number is representable or not in ieee
754 unless you also specify whether you are using ieee 754 single,
double, double extended,... 

> Merely adding more digits
> does no increase the accuracy.

Adding more digits in the stored number allows transfer of these files
from one instance of gnumeric to another that uses a differnet size of
ieee 754 floating point number.
  
>  There is an article by Kahn who is the
> driving force behind the IEEE 754 standard, and he got the Turing
> Award in 1989 for his work on numerical analysis.
>
> How Java's Floating-Point Hurts Everyone Everywhere
> by W. Kahan and J. D. Darcy (March 1998)
>
> http://www.cs.berkeley.edu/~wkahan/JAVAhurt.pdf
>
> The last page "Accuracy < Precision"
>
> From what I can see in your message, previous versions did not
> deliberately find the minimum digit representation of floating point
> number within the limit of ieee 754 standard.

Of course not. Especially since there is no such thing as a "minimum
digit representation of floating point number within the limit of ieee
754 standard" unless you also specify the size of floating point number.

Andreas
>
-- 
Andreas J. Guelzow <aguelzow pyrshep ca>