udokrix
commented
2 years ago issue #42 solved. I misunderstood the comment "of null the normal is calculated automatically" to last parameter "normal" of the function "tessTesselate()". With my own "normal"-parameter it works as it should. "visualdoj" who ported "libtess2" to pascal gave me the solution. Thanks again Udo Krix
Hello Mikko Mononen,
while working with the problem of offsetting polygons in 2d, I came across with your "libtess2" and its pascal-port "fpc-libtess2". When tesselating with winding rule "TESS_WINDING_POSITIVE" and the element type "TESS_BOUNDARY_CONTOURS" I found a strange behavior. Sometimes the correct result was delivered, sometimes not and I do not know why =(. I want to show in 2 examples what I mean. The two polygons
const TESSreal offsetSquare[26] = { 0, 0, 0, 80, 200,80, 200, 0, 120, 0, 120, 200, 200, 200, 200, 120, 0, 120, 0, 200, 80, 200, 80, 0, 0,0 } ;
and
const TESSreal offsetHouse[32]= { 0,0, 0,20, 100,20, 100,0, 80,0, 80,100, 100,100, 86,84, 36,136, 50,150, 64,136, 14,84, 0,100, 20,100, 20,0, 0,0 };
can be obtained by (raw) offsetting to the left 1.) a square in CCR orientation 0,0 200,0 200,200, 0,200, 0,0 with offsetting value=20 2.) a shape of a simple house in CCR orientation 0,0 100,0 100,100, 50,150 0,100 0,0 with offsetting value=20
When tesselating with
1.) tessAddContour(tess, 2, &offsetSquare[0], 2 * sizeof(TESSreal), 13); t=tessTesselate(tess, TESS_WINDING_POSITIVE, TESS_BOUNDARY_CONTOURS, 0, 2, 0);
I got as result the 16 vertices of the 4 outer squares, which all have the negative winding number (-1). The correct result are 4 vertices of the inner square with winding number (+1), the downscaled square. So the delivered result is that of the winding rule TESS_WINDING_NEGATIVE.
This is the output-result (x-,y-component): 80, 0 0 , 0 0 , 80 80 , 80 200 , 0 120 , 0 120 , 80 200 , 80 200 , 120 120 , 120 120 , 200 200 , 200 80 , 120 0 , 120 0 , 200 80, 200
2.) tessAddContour(tess, 2, &offsetHouse[0], 2 * sizeof(TESSreal), 16); t=tessTesselate(tess, TESS_WINDING_POSITIVE, TESS_BOUNDARY_CONTOURS, 0, 2, 0);
I got correct result of 5 vertices of the region with winding number (+1), the downscaled shape of a house. The 5 regions with winding number (-1) are correctly deleted.
This is the output-result (x-, y-component): 80, 90 50 , 121 20 , 90 20 , 20 80 , 20
The fpc-libtes2 delivers the same results as your libtess2.
These are only two examples of many cases I got the correct or false results and I don't know the reason. For me it looks as if in the course of tesselation the orientation of the input polygon is changed from CW -> CCW or vice versa.
I didn't try the GLU tesselator from SGI, that seems to me too complicated with all the callbacks. So I hope you can solve my dilemma.
Udo Krix