`pygeos.set_precision` creates empty geometries

If you round a LineString until you get a single coordinate value, pygeos.set_precision returns an empty geometry. Shouldn’t it return a single point instead?

geometry = pygeos.linestrings([np.array([0.01, 0.02]), np.array([0.03, 0.04]), np.array([0.05, 0.06])])
# <pygeos.Geometry LINESTRING (0.01 0.02, 0.03 0.04, 0.05 0.06)>

pygeos.set_precision(geometry, 1)
# <pygeos.Geometry LINESTRING EMPTY>

Expected result:

# <pygeos.Geometry POINT (0 0)>

[jorisvandenbossche]

that’s why I was using pygeos.make_valid function afterwards to fix them.

And I see that using make_valid on the LineString with 2 equal vertices, indeed gives a Point instead of an empty LineString (as set_precision does). You could certainly bring it up with the GEOS project if they have a reasoning for collapsing to empty geometries instead of collapsing to a “lower dimension geometry” in set_precision (while having a different logic in make_valid).

[carlosg-m]

In general, “naively” rounding the coordinates has the risk to give identical vertices that don’t get collapsed, resulting in invalid geometries. pygeos.set_precision includes some logic to keep the result as valid geometries.

Yes, in my use case I was able to find some examples of invalid linestring with duplicate coordinates, that’s why I was using pygeos.make_valid function afterwards to fix them.

However I don’t know what kinds of other problems this approach might cause to polygons for example, and if they are fixable. But the idea here was to include as many geometries as possible during analysis.