Geometry Splitter with splitting on self-intersecting end-point of a line

[jorisvandenbossche]

A bit related to #1486, but in this case it is not about "not splitting" at self-intersections (that don't intersect with the splitter), but about splitting a line at a point that also happens to be a self-intersection, and where this point is also an endpoint of the line. (cc @dbaston)

Self-intersection in the interior works fine:

$ geosop -a "LINESTRING (0 -1, 0 0, 0 1, 1 1, 1 0, 0 0, -1 0)" -b "POINT (0 0)" split
GEOMETRYCOLLECTION (LINESTRING (0 -1, 0 0), LINESTRING (0 0, 0 1, 1 1, 1 0, 0 0, -1 0))

But when the self-intersection point is an endpoint of the linestring, splitting on it does not do anything:

$ geosop -a "LINESTRING (0 0, 0 1, 1 1, 1 0, 0 0, -1 0)" -b "POINT (0 0)" split
GEOMETRYCOLLECTION (LINESTRING (0 0, 0 1, 1 1, 1 0, 0 0, -1 0))

Should that also split? (I would expect so)

From shapely/shapely#1569

[dbaston]

The problem here is that the splitter first identifies the segment that contains the split point, and then splits the line within that segment. It doesn't consider the possibility that, in a non-simple case, multiple segments may intersect the split point. I would argue that the result of your first example is also incorrect - the output should have three lines, not two.

[jorisvandenbossche]

I would argue that the result of your first example is also incorrect - the output should have three lines, not two.

Ah, yes of course ..

[dbaston]

The first step in fixing this is probably to augment DistanceOp::nearestLocations with a DistanceOp::allNearestLocations that can return > 1 location pairs having the same distance.