v.what.rast: Sample points on the southern and eastern edge

[Ady0333]

Problem

v.what.rast checks every point twice, and the two checks disagree about the region edges.

First it checks the box:

if (!Vect_point_in_box(Points->x[0], Points->y[0], 0.0, &box)) {
    outside_cnt++;
    continue;
}

Vect_point_in_box() includes all four edges, so a point sitting on the southern or eastern edge passes here.

Then it converts the point to a cell:

row = Rast_northing_to_row(Points->y[0], &window);
col = Rast_easting_to_col(Points->x[0], &window);
if (col < 0 || col >= window.cols || row < 0 || row >= window.rows)
    continue;

On the southern edge row comes out as rows, and on the eastern edge col comes out as cols. Both are one past the end, so the point is thrown away here instead. There is no warning and no counter, so you never find out. The column just stays empty.

r.what already returns a value for the same coordinate, so the two tools disagree.

How I ran into it

I set the region from the points I wanted to sample, which is the normal way to use this tool:

g.region vector=pts
v.what.rast map=pts raster=dem column=v

g.region puts the bounding box exactly on the outermost points, so my southernmost and easternmost points were dropped every time. On a small test with the region taken from a raster, four points out of seven came back empty and nothing was printed.

Fix

I bring those two edges into the last row or column, the same way r.what does.

The equality test is deliberate. Only a point sitting exactly on the edge converts to an index that is out of range, and anything further out is still rejected by the check below.

Tests

I added a pytest file. The expected values come from r.what on the same coordinates, so they are not guesses.

Without the fix, the southern edge, the eastern edge, the corner and the comparison with r.what all fail. Interior points and the northern and western edges pass either way, and so does a point that really is outside the region.

[Ady0333]

/cc @echoix @petrasovaa