[vtkusers] doubles, floats, java

[Karin Faulhaber]

Hi Jeff,

thanks for the quick answer.

>           a) The precision provided is better than what you need. Take
> 		what java gives you and round
>           	and/or format before displaying, printing, or otherwise
> 		presenting it to humans.
> 
>           b) The precision provided is not as good as you need. If
> 		you're using Java, you can use
>           	BigDecimal to get arbitrary precision at the cost of
> 		performance.
> 
> Solution a is sufficient for most, since a Java double provides
> something like 16 digits of precision. I forget the
> formal definition, but it's something like "The absolute value of the
> error will be less than 5 * 10^-17 *
> actual result".

16 digits should be more then enough for my needs, but my numbers _are_
rounded, so I went for solution b. The problem is, even if I use
BigDecimal I have to cast it to get it into vtkDoubleArray, which gives
me the same wrong/rounded numbers as before as result of GetComponent().
Any ideas for that?

Karin





-- 
Karin Faulhaber
Universitaet Karlsruhe
Institut fuer Photogrammetrie und Fernerkundung (IPF)
Englerstr. 7 
76128 Karlsruhe

Tel: 0721/608-3676
E-Mail: faulhaber at ipf.uni-karlsruhe.de