Re: [Lib2geom-devel] valueAndDerivative() vs. derivative()

[njh <nj...@nj...>]

On Sun, 18 May 2008, Maximilian Albert wrote:

> Hi,
>
> shouldn't valueAndDerivatives() and derivative() do a similar job? The
> following two pieces of code [1] behave very differently for me:

Out of curiosity, do you get the right result when the 2 becomes a 3?  If 
so I think the bug is actually in
     double SBasis::valueAndDerivative(double t, double &der) const {

The problem is that the approach used for the general case in 
SBasis::valueAndDerivatives (iteratively computing the derivative sbasis 
and evaluating at the point) is correct, but relatively slow (involving 
lots of mallocing, e.g) where as the approach in 
SBasis::valueAndDerivative tried to evaluate the derivative 'for free' by 
tacking onto the calcs for the pointAt.  But apparently it is wrong :(

There is a general n derivatives algorithm for polynomials, and I see no 
strong reason that it wouldn't work for sbasis; but nobody ever did it...

I'll sit down and try and work it out now, but no promises.

njh