This implementation is robust in the sense that the answers returned are exact, but it is not fully adaptive. Basically an initial test is computed, and if the accuracy of this is too low then an exact version of the test is executed. Compared to Shewchuk's original C code this is slower, but eventually I hope to make improvements that bring the performance closer in line to his version.

Example

var orientation =require("robust-orientation")

console.log(orientation([0,0],[1e-64,0],[0,1]))

Install

npm install robust-orientation

API

var orient =require("robust-orientation")

orient(p0, p1, p2, ...)

Exactly computes the orientation of a collection of (n+1) points in n-dimensions.

p0,p1,p2,... is a list of points

Returns The orientation of the point set:

<0 if the tuple of points is positively oriented

>0 if the tuple of points is negatively oriented

=0 if the points are coplanar

Note For up to 5 points, you can directly call an optimized orientation routine, thus avoiding an extra dispatch/switch statement by calling orient[d], where d is the number of points. Eg.