Re: [Plib-users] triangles from branches

[Paolo S. <ax...@ti...>]

Steve Baker wrote:
...
> Take a look at the TORCS project - it uses PLIB and ODE.

I will. Until now I saw the nick's solution and I tried to implement it, 
  since it doesn't need to access the triangle data.

...
> Well, you need to walk the tree structure of the model using things
> like getNumKids() and getKid(n) at each node and using isAKindOf to
> recognise whether you have a leaf node or a branch node...for example:
> 
> 
> void processObj ( ssgEntity *n )
> {
>   if ( n == NULL ) return ;
> 
>   if ( n -> isAKindOf ( ssgTypeLeaf() ) )
>   {
>     ssgLeaf *l = (ssgLeaf *) n ;
>     ....DO SOMETHING WITH THE TRIANGLES...
>     return ;
>   }
> 
>   ssgBranch *b = (ssgBranch *) n ;
> 
>   for ( int i = 0 ; i < b -> getNumKids () ; i++ )
>     processObj ( b -> getKid ( i ) ) ;
> }
> 
> ssgLeaf nodes contain the triangles - there are member functions
> to get the indices of the three vertices of the N'th triangle
> (getTriangle) and to retrieve the vertex, colour, normal and
> texture coordinate given a vertex index:  getVertex()/getColour()/
> getNormal()/getTexCoord().

I wrote a function that uses the algorithm that you and Wolfram suggest:

static void getTriangleData( ssgEntity * ent, dReal * vertices, int * 
indices, int *vert_offset, int *ind_offset ) {
	int i,j;
	if (ent->getNumKids() == 0){
		ssgLeaf * le = (ssgLeaf *) ent;
		int ind_start = *ind_offset;
		int ind_max = 0;
		for (i=0; i<le->getNumTriangles(); i++) {
			short int ind[3];
			le->getTriangle(i, &ind[0], &ind[1], &ind[2]);
			for (j=0; j<3; j++){
				indices[*ind_offset+j] = ind[j];
				if (indices[*ind_offset+j]>ind_max)
					ind_max=indices[*ind_offset+j];
				indices[*ind_offset+j] += ind_start;	
			}
			(*ind_offset)+=3;
		}
		for (i = 0; i < ind_max; i++) {
			float *vert;
			vert = le->getVertex(i);
			for (j=0; j<3; j++){
				vertices[*vert_offset+i*3+j] = (dReal) vert[j];
			}
			(*vert_offset)+=3;
		}
	}
	else {
		ssgBranch * br = (ssgBranch *) ent;
		
		for (i=0; i<br->getNumKids(); i++){
			getTriangleData(br->getKid(i), vertices, indices, vert_offset, 
ind_offset);
		}
	}
}

but I fear that I have to handle also the transformation nodes to get 
right vertices positions. And I have no idea how to calculate the exact 
dimensions of the arrays I should pass to this function. Just for the 
indices I used:

static int getNumIndices(ssgEntity * ent) {
	int i, num = 0;
	if (ent->getNumKids() == 0){
		ssgLeaf * le = (ssgLeaf *) ent;
		return (le->getNumTriangles())*3;
	}
	else {
		ssgBranch * br = (ssgBranch *) ent;
		
		for (i=0; i<br->getNumKids(); i++){
			num += getNumIndices(br->getKid(i));
		}
		return num;
	}
}

Since I didn't find any simple way to check if the triangle data I get 
with this method are correct (of course when I use this data I get wreid 
collisions :)) I gave up...

Maybe torcs people have found a solution.
--
Paolo Sacconier