Panic in add_constraint_and_split

[Azorlogh]

This test case panics:

    #[test]
    fn edge_intersection_precision_test_3() -> Result<(), InsertionError> {
        let edges = [
            [
                Point2 {
                    x: 7.9910984,
                    y: -59.344463,
                },
                Point2 {
                    x: 15.484455,
                    y: -48.762936,
                },
            ],
            [
                Point2 {
                    x: 15.484455,
                    y: -48.762936,
                },
                Point2 {
                    x: -11.673287,
                    y: -28.37192,
                },
            ],
            [
                Point2 {
                    x: -11.673287,
                    y: -28.37192,
                },
                Point2 {
                    x: -16.214716,
                    y: -43.81278,
                },
            ],
            [
                Point2 {
                    x: -16.214716,
                    y: -43.81278,
                },
                Point2 {
                    x: -3.657666,
                    y: -51.41967,
                },
            ],
            [
                Point2 {
                    x: -3.657666,
                    y: -51.41967,
                },
                Point2 {
                    x: 7.9910984,
                    y: -59.344463,
                },
            ],
            [
                Point2 {
                    x: 7.4022045,
                    y: -51.355137,
                },
                Point2 {
                    x: 7.9910984,
                    y: -59.344463,
                },
            ],
            [
                Point2 {
                    x: 7.4022045,
                    y: -51.355137,
                },
                Point2 {
                    x: 15.484455,
                    y: -48.762936,
                },
            ],
            [
                Point2 {
                    x: 7.4022045,
                    y: -51.355137,
                },
                Point2 {
                    x: -13.92232,
                    y: -36.01863,
                },
            ],
        ];
        let mut cdt: ConstrainedDelaunayTriangulation<Point2<f32>> =
            ConstrainedDelaunayTriangulation::new();
        for edge in edges {
            let point_a = cdt.insert(edge[0])?;
            let point_b = cdt.insert(edge[1])?;
            cdt.cdt_sanity_check();
            cdt.add_constraint_and_split(point_a, point_b, |v| v);
            cdt.cdt_sanity_check();
        }

        Ok(())
    }

thread 'cdt::test::edge_intersection_precision_test_3' panicked at src/cdt.rs:876:17:
assertion `left != right` failed
  left: []
 right: []

It panics on this assertion:

spade/src/cdt.rs

Line 876 in bddafbb

assert_ne!(new_edges, Vec::new());

[Stoeoef]

Thanks for the report! I'll try to look into that soon.

In the meantime, you could consider to fork the repository and remove that assertion if it's a blocker for you - it's probably not too important. I'll know more once I've looked into that.

[Stoeoef]

This one is annoying. When using f64, everything works just fine - the minimized example only requires 2 calls to add_constraint_and_split and looks like this:

Note that vertex 3 and 4 overlap visually but they do not share the same position - it's ever so slightly different. There is also a constraint edge going from 1->4, it's just being visually overlapped by the regular edge 1->3.

This triangulation correctly represents the split and has 4 constraint edges: 0->4, 2->4, 3->4 and 1->4

However, for f32, v3 and v4 do end up on the exact same position after downcasting. In this case, the triangulation should end up with 4 vertices in total and only 3 constraint edges: 0->3,1->3, 2->3

What an annoying edge case 😄 . I'm working on a fix.

[Stoeoef]

Fixed in v2.12.1 (just released) - thanks again for the report 💯