Question

I have a class which has three integers to represent it: a serverID, a streamID and an messageID.

I have some HashSet that are small but I do lots of stuff like set intersection on, and others that have 10K+ elements in.

There are only a handful of values for serverID, but they are truly random numbers with a full 32-bits of randomness. Often there is only one serverID for a whole hashtable; other times just a couple of serverIDs.

The streamID is a small number, typically 0 but may be 1 or 2 sometimes.

The messageID is sequentially increasing for each serverID/streamID pair.

I currently have:

(-messageID << 24) ^ messageID ^ serverID ^ streamID

I want to understand that I have a good hash function despite having a sequentially increasing messageID and not a lot of other bits to mix in.

What makes a good hashCode and how can I best mix these three numbers?

Answer 1

I personally always use strategy implemented in java.lang.String:

        for (int i = 0; i < len; i++) {
            h = 31*h + val[off++];
        }

So, in your case I’d use the following: 31 * (31 * messageID + serverID) + streamID