Hash Algorithm for Point3D

[Michael Morris]

Does anyone know of a good hashCode() algorithm for a 3D point with doubles for X, Y and Z? The equals() method for the class returns true if X, Y and Z are equal for both objects and the compared object passes an instanceof test.

Thanks in advance,
Michael Morris

[Jim Yingst]

Well, being lazy myself, I'd probably just use the basic algorithm recommended in Effective Java pp. 38-39:

    public int hashCode() {
        long x = Double.toLongBits(X);
        long y = Double.toLongBits(Y);
        long z = Double.toLongBits(Z);
        int result = 17;
        result = 37 * result + (int) (x ^ (x >> 32));
        result = 37 * result + (int) (y ^ (y >> 32));
        result = 37 * result + (int) (z ^ (z >> 32));
        return result;
    }

Bloch advocates this sort of thing as generally yielding "reasonably good" hash functions, though usually not state-of-the-art. I'm not sure what might be better in this case, aside from one obvious possible improvement by precalculating the result 17 * 37 and thus skipping one multiplication. And the numbers 17 and 37 may be replaced by many other possibilities, preferably prime.

[Michael Morris]

I had done something similar to that. I probably shouldn't fret so much and could probably even use something involving distance to origin and individual coordinates. I will never have a coordinate over about 15000' and the granularity will be about 1/64".
Michael Morris