[Fluxus] pdata size behavior

[Dave Griffiths]

Hi Glauber,

This is due to the fact that (build-seg-plane) returns a primitive which
is an un-indexed quad list - the vertices are not shared between the
faces.

The best way to deal with this is to make an indexed polygon object
yourself with build-polygons, and stitch the vertices together with the
index - for example:

(define (build-mesh-plane w h)
    (let ((p (build-polygons (* w h) 'quad-list)))
        (with-primitive p
            (pdata-index-map!
                (lambda (i p)
                    (vector (modulo i w) (quotient i w) 0))
                "p")            
            (poly-set-index     
                (build-list
                    (* (- w 1) (- h 1) 4)
                    (lambda (i)
                        (let ((f (+ (quotient i (* 4 (- w 1))) (quotient i 4))))
                            (list-ref (list f (+ f 1) (+ f w 1) (+ f w))
                                (modulo i 4))))))
            (recalc-normals 1)) p))


(clear)
(hint-wire)
(wire-colour 0)
(define p (build-mesh-plane 10 10))

(with-primitive p
    (pdata-map!
        (lambda (p)
            (vadd p (vmul (srndvec) 0.2)))
        "p")            
        (recalc-normals 1))

On Mon, 2010-05-24 at 19:41 -0300, Glauber Alex Dias Prado wrote:
> Hey ppl,
> 
> I was figuring out how to make a kohonen map
> <http://en.wikipedia.org/wiki/Kohonen_map> using pdata to represent my
> err data, but this is a bit tricky as there are much more pdata than
> vertices, and preferably i would like a one-to-one mapping:
> 
> (clear)
> (hint-none)
> (hint-wire)
> (point-width 4)
> (hint-points)
> ;(translate (vector 0 0 9))
> 
> (define grid
>     (build-seg-plane 10 10))
> 
> above is just the grid which i would like to manipulate as if it were a
> matrix of vertices, but:
> 
> (with-primitive grid
> 	(pdata-size))
> returns 400
> 
> and i would like something like 100 at least for vertex position
> manipulation, so is those 400 related to all (t p n c) pdatas and there
> are only 100 for "p"?
> 
> I guess not, so i might be using the wrong primitive type for this
> maybe?
> 
> does someone know what kind of primitive i could have a
> one-vertex-per-pdata representation?