Breaking change consistant hashing

[dennis-tra]

Hi everyone,

the last commit broke our assumption that a randomly ordered slice of nodes will consistently return the same node for the same hash. The following test code fails on the current HEAD of master and works fine with the previous commit:

func TestStableHashring(t *testing.T) {
	nodes := []string{"NODE_1", "NODE_2", "NODE_3", "NODE_4", "NODE_5"}

	hash := "HASH"
	var constantNode string
	for i := 0; i < 100; i++ {

		rand.Shuffle(len(nodes), func(i, j int) { nodes[i], nodes[j] = nodes[j], nodes[i] })

		hr := New(nodes)
		n, ok := hr.GetNode(hash)
		if !ok {
			t.Error("error getting node")
		}

		if constantNode == "" {
			constantNode = n
		}

		if n != constantNode {
			t.Errorf("nodes are not equalt %s != %s", n, constantNode)
		}
	}
}

Is this a falsey assumption or did the latest changes introduce an unintended behaviour?

Cheers,
Dennis

[EmpireJones]

I'm curious of the intention here too; I had to fork with a revert commit, as this broke a desired behavior; I was using this for consensus among multiple servers, but that doesn't work well if they don't agree on anything.

[serialx]

Fixed by #23