Computes a tiled Euclidean distance raster over a very large, dense point set.
Given a very large, dense point set—on the order of 10s to 100s of millions
of points or more—this apply method provides the means to relatively
efficiently produce a Euclidean distance raster layer for those points.
This operator assumes that there is a layout definition for the space in
qustion, and that the input points have been binned according to that
layout's mapTransform object.
There are caveats to using this function. The assumption is that the
incoming point set is _dense_ relative to the layout. An empty SpatialKey
in isolation is not probematic. However, large areas devoid of points
which translate to many adjacent spatial keys in the grid which are empty
will likely cause problems. Additionally, tiles whose Euclidean distance
values are defined by points that are not in the current tile or in one of
the 8 neighboring tiles will produce incorrect values. This is
particularly true in regions outside the mass of points where the medial
axis of the point set intrudes (these will be generated in large interior
voids or in areas where the point set boundary has nonconvex pockets).