We are seeing some geometry intersection test failures in CI. Originally we thought it was related to pandas 2.1 (issue #124, PR #125) but it isn't. What is happening is that our code is behaving the same as ever, but we compare some geometry results with that produced by geopandas and we have started getting slightly different results for intersection tests with a rectangle of zero area. The geopandas code for such either calls shapely or pygeos or handles the test itself.
We are seeing some
geometry
intersection test failures in CI. Originally we thought it was related topandas 2.1
(issue #124, PR #125) but it isn't. What is happening is that our code is behaving the same as ever, but we compare some geometry results with that produced bygeopandas
and we have started getting slightly different results for intersection tests with a rectangle of zero area. Thegeopandas
code for such either callsshapely
orpygeos
or handles the test itself.I have a fix for this in progress.