BUG: intersection of polygons is not symmetric

[brendan-ward]

First reported in Shapely #1623.

Given a.wkt:

POLYGON ((11174.451451065055 13710.473363871335, -14730.97396732842 -5570.416909945428, -24371.419104236804 7382.295799251309, 1534.0063141566734 26663.18607306807, 11174.451451065055 13710.473363871335))

and b.wkt:

POLYGON ((-2129.8301268088653 3808.3625460804574, -13.419399941060902 5383.564741760034, 2589.8327945394485 5383.564741760034, 2589.8327945394485 2421.1656459793585, 760.529880485728 -6043.318375123083, -2129.8301268088653 -6043.318375123083, -2129.8301268088653 3808.3625460804574))

The intersection of a and b is not equivalent to the intersection of b and a (is sensitive to order).

❯ geosop -a a.wkt -b b.wkt -f wkt intersection
POLYGON ((-2129.8301268088653 3808.3625460804574, -13.419399941060902 5383.564741760034,
 -13.419399941060874 5383.564741760034, -2129.8301268088653 3808.3625460804574))

❯ geosop -a b.wkt -b a.wkt -f wkt intersection
POLYGON ((-13.419399941060902 5383.564741760034, -13.419399941060874 5383.564741760034, 
2589.8327945394485 5383.564741760034, 2589.8327945394485 2421.1656459793585, 
760.529880485728 -6043.318375123083, -2129.8301268088653 -6043.318375123083, 
-2129.8301268088653 3808.3625460804574, -13.419399941060902 5383.564741760034))

[dr-jts]

It's simpler to say that the intersection computed for B and A is incorrect. But may be useful to know that it's order-sensitive.

[dr-jts]

This does not seem to be a problem in JTS.

[pramsey]

If you flip JTS to using an unordered map, as GEOS did, do you generate the bug in JTS?

[dr-jts]

I'll try and check this. Although the difference here is that this works in JTS, whereas in #827 JTS also fails.

[dr-jts]

This seems to work in GEOS 3.14. The results have different vertex order, but they are topologically identical.

@brendan-ward are you able to test?