Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

3d hull should allow coplanar points #157

Open
simonlindholm opened this issue Feb 16, 2020 · 3 comments
Open

3d hull should allow coplanar points #157

simonlindholm opened this issue Feb 16, 2020 · 3 comments

Comments

@simonlindholm
Copy link
Member

https://codeforces.com/blog/entry/73366 links to https://github.com/Laakeri/contestlib/blob/master/src/geometry/3dconvexhull.cpp which seems like a good implementation. Need to test/minify it.

@Chillee
Copy link
Collaborator

Chillee commented Feb 16, 2020

That implementation does add an extra log n factor though. Could be worth it to replace it regardless, depending on benchmark results.

@simonlindholm
Copy link
Member Author

golfed a bit:

bool sameHalf(P3 a, P3 b, P3 p1, P3 p2, P3 p3) {
	P3 hlp = (p2-p1).cross(p3-p1);
	return (hlp.dot(a-p1))*(hlp.dot(b-p1))>=0;
}

vector<tuple<int, int, int>> hull3d_2(vector<P3>& ps) {
	map<tuple<int, int, int>, int> hull;
	hull[{0,1,2}]=3;
	hull[{0,1,3}]=2;
	hull[{0,2,3}]=1;
	hull[{1,2,3}]=0;
	rep(i,4,sz(ps)) {
		vector<array<int, 4>> ch;
		for (auto [tri, q] : hull) if (auto [a,b,c] = tri;
				!sameHalf(ps[i], ps[q], ps[a], ps[b], ps[c])) {
			ch.push_back({a,b,i,c});
			ch.push_back({a,c,i,b});
			ch.push_back({b,c,i,a});
			ch.push_back({a,b,c,i});
		}
		for (auto [a,b,c,q] : ch) {
			auto [it,r] = hull.emplace(make_tuple(a,b,c), q);
			if (!r) hull.erase(it);
		}
	}
	vector<tuple<int, int, int>> ret;
	for (auto [tri, q] : hull) {
		auto [a,b,c] = tri;
		if ((ps[q] - ps[a]).dot((ps[b] - ps[a]).cross(ps[c] - ps[a])) > 0)
			swap(a, b);
		ret.emplace_back(a, b, c);
	}
	return ret;
}

Should be possible to get rid of the log factor by the same method as the current 3d hull implementation -- keeping an O(n^2)-sized map from edges to faces -- though it would add some code.

I wonder how much more code O(n log n) randomized 3d hull would be, it doesn't sound too bad from http://www.inf.fu-berlin.de/lehre/SS13/AG/randinc3d.pdf...

@simonlindholm
Copy link
Member Author

http://www.inf.fu-berlin.de/lehre/SS13/AG/randinc3d.pdf

I tried to code this, and it ended up at ~70 lines (buggy wip implementation)

https://github.com/bqi343/USACO/blob/master/Implementations/content/geometry%20(13)/3D/Hull3D.h looks much better. Would be interesting to benchmark Delaunay triangulation based on that implementation, including with concyclic points.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants