d3.geoStitch should ignore unpaired points on the antimeridian.

[mbostock]

Consider the stitching of a single polygon where a single edge (DE…GH) lies on the antimeridian:

C-----D
|     |
|     E
|    /
|   F
|    \
|     G
|     |
I-----H

During stitching, two fragments are considered:

C-----D
|      
|     E
|    /
|   F
|    \
|     G
|      
I-----H

After stitching, this produces two rings:

C-----D
|     |
|     |
|     |
|     |
|     |
|     |
|     |
I-----H

  E
 /|
F |
 \|
  G

This is because the points that lie on the antimeridian (D, E, G and H) do not have a matching fragment on the other side of the antimeridian to join to. So the fragments DCIH and EFG are each treated as “standalone rings”, resulting in separate polygons.

I employed this workaround to stitch contours:

var shared = {};

// Record the y-intersections with the antimeridian.
p.coordinates.forEach(function(ring) {
  ring.forEach(function(p) {
    if (p[0] === -180 || p[0] === 180) {
      shared[p[1]] |= p[0] === -180 ? 1 : 2;
    }
  });
});

// Offset any unshared antimeridian points to prevent their stitching.
p.coordinates.forEach(function(ring) {
  ring.forEach(function(p) {
    if ((p[0] === -180 || p[0] === 180) && shared[p[1]] !== 3) {
      p[0] = p[0] === -180 ? -179.9995 : 179.9995;
    }
  });
});

But it seems to me that d3.geoStitch could do this by default, and then simple cases like this wouldn’t get broken by d3.geoStitch:

{
  "type": "Polygon",
  "coordinates": [
    [[180, 10], [170, 11], [170, 19], [180, 20], [180, 19], [179, 15], [180, 11], [180, 10]]
  ]
}