Possible raycast offset solution

[gkjohnson]

Related to #177

Offsets the ray to avoid floating point error on coplanar faces - but since these are being treated as whole triangles there should be no need to test for coplanarity?

TODO

• ~Why are these triangles not being treated as clipped / intersected?~
• Why are the triangles not being marked as coplanar?
  • ~Planes are generated in incorrect direction~
  • Check coplanar detection logic

[gkjohnson]

Triangle planes were not being calculated correctly and coplanar triangles were not being marked as coplanar.

Still some triangles being missed by coplanar count, tho:

[gkjohnson]

Commenting out these lines fixes the issue:

// skip the triangle if we don't intersect with it
if ( ! _splittingTriangle.intersectsTriangle( tri, _edge, true ) ) {

    performCoplanarIncrement( tri );
    continue;

}

[gkjohnson]

Using these functions in place of intersectsTriangle seems to fix the issue, but also increases the amount of triangulation. Why?

const getSide = v => Math.sign( plane.distanceToPoint( v ) );
const isOneSide = t => {

    let s = getSide( t.a ) || getSide( t.b ) || getSide( t.c );
    if ( s !== getSide( t.b ) ) return false;
    if ( s !== getSide( t.c ) ) return false;

    return true;

};

edit - we need to check if the "intersectsTriangle" function properly handles fully encapsulated triangles? Because the "coplanar" function does nothing if the triangles don't intersect at all. Could also try manually processing coplanarity on all triangles afterward.

[gkjohnson]

The intersects triangle function does handle fully encapsulated triangles.

It looks like using the plane is not a great way to determine coplanarity.

TODO

• Check for encapsulation after the fact
• See if it's better to assume everything is coplanar first if a planar triangle arrises and then check what doesn't need the coplanar check.