Le 04/08/2015 22:37, clipka a écrit :

>
> Missing support for bezier in spline functions and sphere_sweep is just
> a matter of incomplete implementation, not of a philosophy.
>

the spline in povray support 4 kinds: natural_spline, linear_spline, 
quadratic_spline and cubic_spline. The latter three matching the 
equations given in https://en.wikipedia.org/wiki/B%C3%A9zier_curve for
linear/quadratic/cubic Bézier curves.

sphere_sweep, in regard to Bézier, is limited to linear_spline and 
cubic_spline. (b_spline is a more general curve, which can or cannot be 
a Bèzier curve according to the spacing of the points)



>
> One reason (the primary one in my book) for missing NURBS support has
> been the lack of affordable user interfaces to define NURBS surfaces and
> export the coefficients to a format suitable for POV-Ray. I wouldn't
> want to have to input NURBS coefficients directly to test such a
> feature, especially if I have no way of verifying the resulting shape.

Yep, I even get tired of the syntax of poly{} object which was not 
reader & writer friendly at all. It took maybe twenty years to get the 
polynomial syntax (same object, just easier to handle).

>
> I didn't know Blender supported NURBS surfaces. If it does, you could
> help by providing a very basic export function that just writes the
> coefficients to an array.
>
> I only have a rudimentary understanding of the underlying math though,
> and there's other stuff high on my agenda at present, so don't hold your
> breath unless Jerome is willing to pick up the glove.
>
> @Jerome: Even if the implementation would be similar to that of bicubic
> patch type 1, i.e. an internal translation to something mesh-like, so
> that in a sense the whole job could also be done by the exporter by
> generating a mesh2 object, I think a NURBS primitive would be beneficial
> in order to minimize file I/O in the workflow.

I have some other items on my to-do-list (from a long time), and I might 
get stalled in a hen-egg situation: without an array of valid values for 
a nurbs, no incentive to parse and implement such object.
(hint: please provide an interesting nurbs, not just a sphere or a 
torus, something that is unique to nurbs)

If I get something right, the array would be a 2D rectangular array of 
4D values (3D positions of control points and weight).
A global tunning of the nurbs would be its order (2 or 4 ? can it be 
anything ? blender's doc states from 2 to 6), and wether each axis (of 
array) is endpoint or uniform (which i do not understand yet, seems tied 
to the knots setting, allowing only 2 knots vectors)

The dimensions of the array can be anything, each dimension being at 
least the order.


>
>
>> If I wrote the AutodeskMaya2Povray exporter, I would meet the same
>> problems?
>> :(
>
> I don't think so. I'm pretty sure Maya provides mechanisms for export
> filters to access mesh representations of NURBS surfaces rather than the
> NURBS coefficients themselves, so that would be an option.
>

Post a reply to this message