Failure of boost::geometry::within for polygons

[Mitsuru]

Dear experts

As this title, I failed boost::geometry::within, some are succeeded but the others are failed.

I wonder if boost::geometry::within has specification for polygons, such as ordering (clockwise, or anti-clockwise), start & end points must be the same.

If you have the same experience, or have how to cope with this problem,
I would appreciate it if you could help me.

Kind regards

[JLBorges]

Polygon concept: https://www.boost.org/doc/libs/1_76_0/libs/geometry/doc/html/geometry/reference/concepts/concept_polygon.html

We can check the validity of the polygon with is_valid() https://www.boost.org/doc/libs/1_76_0/libs/geometry/doc/html/geometry/reference/algorithms/is_valid.html

correct() can fix some kinds of invalidities
https://www.boost.org/doc/libs/1_76_0/libs/geometry/doc/html/geometry/reference/algorithms/correct.html

[lastchance]

@Mitsuru, when you say "failed", what do you mean?

Can you give examples of which ones "failed"?

Note that any test point "on" the boundary of a polygon is necessarily going to raise questions about
- floating-point accuracy
- the definition of "inside" or "outside" for something which is "on the boundary of"

If you are worried about clockwise/anticlockwise traversal then simply try what happens with both. Similarly for closed/non-closed.

I suspect that non-convex polygon inside non-convex polygon is a bit of a nightmare anyway. You could have all vertices of the child polygon inside the parent, yet some parts of its area outside. Personally, I would try to avoid non-convex polygons: you can break them down into convex components, ultimately into triangles.