Dan Baston

Another possible interface -- return an "intersection result" that makes it possible to get either the overlap fractions or geometries either in bulk or individually by row/col. ``` GEOSGridIntersection* GEOSGrid_intersection(const...

> there will be places where there is no geometry in some grid cells. This is where row/column access could be useful, because otherwise there is no obvious correspondence between...

I can confirm this in `main` using `geosop`, which shows both inputs are valid but throws a `TopologyIntersection` during intersection testing. ``` geosop -a 'POLYGON ((26639.240191093646 6039.3615818717535, 26639.240191093646 5889.361620883223,28000.000095100608 5889.362081553552,...

JTS shows that the first input is invalid, so the error may be in `isValid` rather than `intersects`.


If I update JTS to the latest commit the polygon shows as valid. Unfortunately I wasn't smart enough to record what commit I was on before updating.

Looks like it was 26af9f377 (January 2021) ... and the JTS commit that changed this behavior (isValid false -> true) is https://github.com/locationtech/jts/commit/b520430f425a41961a2da0f09731dcdfbffeb937

Attached XML test case failing in both GEOS and JTS. [geos-issue-786.xml.txt](https://github.com/libgeos/geos/files/10203572/geos-issue-786.xml.txt) If the hole is removed from the larger polygon, the test passes.

I can see that they're different but don't understand why one or the other would be preferable.

Isn't it pretty much the same as ``` MULTILINESTRING ((1 0, 2 0, 2 1, 1 1, 1 0), (2 0, 3 0, 3 1, 2 1), (1 0, 0...