'fromPoints' is O(n), 'simpleFromPoints' is O(n log n).

[lemmih]

Hi @noinia. How do you feel about this naming scheme where simple* functions have an (Ord r, Fractional r) constraint and check for self-intersections (ie if the polygons are actually simple)?

[noinia]

Hi @noinia. How do you feel about this naming scheme where simple* functions have an (Ord r, Fractional r) constraint and check for self-intersections (ie if the polygons are actually simple)?

Sounds good to me. Note that I do think we can actually check if the polygon has self-intersectsions with only (Ord r, Num r) actually (but they would still run in O(n log n) time.

[lemmih]

Sounds good to me. Note that I do think we can actually check if the polygon has self-intersectsions with only (Ord r, Num r) actually (but they would still run in O(n log n) time.

The way I did it before (by basically using Data.Ratio without the Integral constraint) had robustness issues when used with Doubles. Some polygons would trigger the intersection check when I used my trick but then had no intersections when using Doubles directly.

[noinia]

Sounds good to me. Note that I do think we can actually check if the polygon has self-intersectsions with only (Ord r, Num r) actually (but they would still run in O(n log n) time.

The way I did it before (by basically using Data.Ratio without the Integral constraint) had robustness issues when used with Doubles. Some polygons would trigger the intersection check when I used my trick but then had no intersections when using Doubles directly.

We should be able to use the 4 ccw tests to test for intersection. But maybe we can let that solve itself after fixing #77.

A separate discussion is what to do about Doubles/Floating point numbers. With the better onSegment test in the visibility-polygon branch some of the isSelfIntersecting Poygon checks with doubles now fail. I'm not entirely sure what to do about issues like this. On one hand it would be nice if the algorithms could do something reasonable with Doubles, but I'm not sure how much effort to put in this since it is clear that we won't be able to get correct results for all situations anyway.