public void TestAddMulModHash()
{
var h = new AddMulModHash(new MersenneTwister(1215125));
var list = Enumerable.Range(0, 80000).Select(i => h.ComputeHash((uint)i)).ToList();
var set = new HashSet <uint>(list);
Assert.Equal(set.Count(), list.Count());
}
public void TestAddMulModHashBucketCollisions()
{
var h = new AddMulModHash(new MersenneTwister(1215125));
var maxCollisions = Enumerable.Range(0, 80000)
.Select(i => h.ComputeHash((uint)i) % 1000)
.GroupBy(i => i, i => 1)
.Max(g => g.Count());
Assert.True(maxCollisions < 100);
}
public void TestMinSketch()
{
var rand = new MersenneTwister(1095807143);
uint slots = 5113;
uint algs = 5;
//uint slots = 1279;
//uint algs = 3;
var inputs = Enumerable.Range(0, (int)slots).Select(ix => (uint)rand.NextUint32()).Distinct().ToList();
var uniqueInputs = inputs.Count();
var medianOvercountExpected = (double)uniqueInputs / slots;
var peakOvercountPercentageAllowed = 1.03;
var actualCount = 1;
var allowedOvercount = Math.Max(actualCount,
actualCount * medianOvercountExpected * peakOvercountPercentageAllowed);
var s = new CountMinSketch <AddMulModHash>(slots, algs, AddMulModHash.DeterministicDefault());
foreach (var i in inputs)
{
s.InterlockedAdd(i, actualCount);
}
var errors = new List <KeyValuePair <uint, long> >(100);
for (uint i = 0; i < uniqueInputs; i++)
{
var found = s.Estimate(i);
if (found > allowedOvercount)
{
errors.Add(new KeyValuePair <uint, long>(i, found));
}
}
Assert.Equal(9, errors.Count());
}
