Bug in hull() about openjscad.org

[Spiritdude]

Note to myself:

function main() {
   return hull(
        CAG.circle({radius:2}).translate([-1,0]),
        CAG.circle().translate([1,0]),
        CAG.circle().translate([0,1.01])
    );
}

works.

Problem: in this.ccw and stack calculation (x1==x2 or y1==y2)

[Spiritdude]

More simple case:

function main() {
   return hull(
        CAG.circle().translate([0,1]),
        CAG.circle().translate([0,-1])
    );
}

gives error, the two circles have a common point, the moment they are apart hull() works.

[Axford]

Perhaps recode the sort to use the faster isLeft() function described for the Graham scan algorithm:

http://geomalgorithms.com/a01-_area.html#isLeft()

via http://geomalgorithms.com/a10-_hull-1.html

Avoids the need for atan... looks partially implemented with the ccw function?

[Spiritdude]

isLeft() = ccw(), the problem lies in the stack operation while(this.ccw(...)<=0) ... and I already have tested prefiltering multiple points with same coords (not sure if it works as I still run the same problem). I did some more testing, it seems with most close (1e-5 or less) but not quite identical x1==x1 and y1==y2

[Axford]

this quick hack to the ccw function sorts it:

this.ccw = function(p1, p2, p3) {
           var tmp = (this.points[p2].x - this.points[p1].x)*(this.points[p3].y - this.points[p1].y) - 
                (this.points[p2].y - this.points[p1].y)*(this.points[p3].x - this.points[p1].x);
           // look for close points
           if (tmp < 0.000001) tmp = 0;
           return tmp;
      }

[Spiritdude]

great, yes, it does it (thanks!)