public void ReverseAddDifferentSizesTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var testData2 = DataGenerator.Generate().Skip(addSize).Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new KeyValueBloomFilterConfiguration();
var bloomFilter = new InvertibleReverseBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(4 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var bloomFilter2 = new InvertibleReverseBloomFilter <TestEntity, long, sbyte>(configuration);
//We have to create a foldable version.
var data = bloomFilter.Extract();
var foldFactor = configuration.FoldingStrategy.GetAllFoldFactors(data.BlockSize).Where(f => f > 1).OrderBy(f => f).First();
bloomFilter2.Initialize(addSize, data.BlockSize / foldFactor, data.HashFunctionCount);
foreach (var itm in testData2)
{
bloomFilter2.Add(itm);
}
bloomFilter.Add(bloomFilter2);
var contained = testData.Union(testData2).Count(item => bloomFilter.Contains(item));
Assert.AreEqual(contained, 2 * addSize, "Not all items found in added Bloom filters");
}
public void SplitRibfDecodePerformance()
{
var configuration = new KeyValueLargeBloomFilterConfiguration();
var size = new[] { 1000, 10000, 100000 };
var modPercentage = new[] { 0, 0.01D, 0.1D, 0.2D, 0.5D, 1.0D };
foreach (var s in size)
{
using (
var writer =
new StreamWriter(File.Open($"splitribfdecode-{s}.csv",
FileMode.Create)))
{
writer.WriteLine("timeInMs,sizeInBytes,capacity,modCount,detectedModCount,countDiff,countDiffSd,decodeSuccessRate");
foreach (var mod in modPercentage)
{
foreach (var capacityPercentage in new[] { 0.5, 1, 2, 5, 10, 100 })
{
var sizeInBytes = new long[100];
var timeSpan = new long[50];
var countAggregate = new int[50];
var modCountResultAggregate = new int[50];
var decodeResult = new int[50];
for (var run = 0; run < 50; run++)
{
var dataSet1 = DataGenerator.Generate().Take(s).ToList();
var dataSet2 = DataGenerator.Generate().Take(s).ToList();
dataSet2.Modify((int)(s * mod));
var onlyInSet1 = dataSet1.Where(d => dataSet2.All(d2 => d2.Id != d.Id)).Select(d => d.Id).OrderBy(id => id).ToArray();
var onlyInSet2 = dataSet2.Where(d => dataSet1.All(d1 => d1.Id != d.Id)).Select(d => d.Id).OrderBy(id => id).ToArray();
var modified = dataSet1.Where(d => dataSet2.Any(d2 => d2.Id == d.Id && d2.Value != d.Value)).Select(d => d.Id).OrderBy(id => id).ToArray();
var idealCapacity = Math.Max(15, onlyInSet1.Count() + onlyInSet2.Count() + modified.Count());
var stopWatch = new Stopwatch();
stopWatch.Start();
var bloomFilter1 = new InvertibleReverseBloomFilter <TestEntity, long, int>(configuration);
bloomFilter1.Initialize((int)(idealCapacity * capacityPercentage), 0.01F);
foreach (var item in dataSet1)
{
bloomFilter1.Add(item);
}
var bloomFilter2 = new InvertibleReverseBloomFilter <TestEntity, long, int>(configuration);
bloomFilter2.Initialize((int)(idealCapacity * capacityPercentage), 0.01F);
foreach (var item in dataSet2)
{
bloomFilter2.Add(item);
}
var s1 = new HashSet <long>();
var s2 = new HashSet <long>();
var s3 = new HashSet <long>();
var success = bloomFilter1.SubtractAndDecode(bloomFilter2, s1, s2, s3);
stopWatch.Stop();
using (var stream = new MemoryStream())
{
_protobufTypeModel.Serialize(stream, bloomFilter1.Extract());
stream.Position = 0;
sizeInBytes[run] = stream.Length;
}
using (var stream = new MemoryStream())
{
_protobufTypeModel.Serialize(stream, bloomFilter2.Extract());
stream.Position = 0;
sizeInBytes[50 + run] = stream.Length;
}
timeSpan[run] = stopWatch.ElapsedMilliseconds;
countAggregate[run] = onlyInSet1.Count() + onlyInSet2.Count() + modified.Count();
modCountResultAggregate[run] = s1.Union(s2).Union(s3).Count(v => onlyInSet1.Contains(v) ||
onlyInSet2.Contains(v) || modified.Contains(v));
decodeResult[run] = success == true ? 1 : 0;
}
var countAvg = (long)countAggregate.Average();
var modCountResult = (long)modCountResultAggregate.Average();
var differenceResult =
modCountResultAggregate.Select((r, i) => r - countAggregate[i]).ToArray();
var differenceSd = Math.Sqrt(differenceResult.Variance());
writer
.WriteLine($"{timeSpan.Average()},{sizeInBytes.Average()},{Math.Max(15, capacityPercentage * (int)(s * mod))},{countAvg},{modCountResult},{(long)differenceResult.Average()},{differenceSd},{1.0D * decodeResult.Sum() / 50}");
}
}
}
}
}
public void RibfDecodePerSizePerformance()
{
var configuration = new KeyValueLargeBloomFilterConfiguration();
var size = new[] { 1000, 10000, 100000 };
var modPercentage = new[] { 0, 0.01D, 0.1D, 0.2D, 0.5D, 1.0D };
foreach (var s in size)
{
using (
var writer =
new StreamWriter(File.Open($"ribfdecodespersize-{s}.csv",
FileMode.Create)))
{
writer.WriteLine("capacity,modCount,estimatedModCount,size,decodesPerSize,decodesPerSizeSd,decodeSuccessRate");
foreach (var mod in modPercentage)
{
foreach (var capacity in new[] { 10, 100, 500, 1000, 2000, 5000, 10000 })
{
var countSize = 0;
var decodesPerSize = new double[50];
var decodeResult = new int[50];
var modCountResultAggregate = new int[50];
for (var run = 0; run < 50; run++)
{
var dataSet1 = DataGenerator.Generate().Take(s).ToList();
var dataSet2 = DataGenerator.Generate().Take(s).ToList();
dataSet2.Modify((int)(s * mod));
var onlyInSet1 = dataSet1.Where(d => dataSet2.All(d2 => d2.Id != d.Id)).Select(d => d.Id).OrderBy(id => id).ToArray();
var onlyInSet2 = dataSet2.Where(d => dataSet1.All(d1 => d1.Id != d.Id)).Select(d => d.Id).OrderBy(id => id).ToArray();
var modified = dataSet1.Where(d => dataSet2.Any(d2 => d2.Id == d.Id && d2.Value != d.Value)).Select(d => d.Id).OrderBy(id => id).ToArray();
var bloomFilter1 = new InvertibleReverseBloomFilter <TestEntity, long, int>(configuration);
bloomFilter1.Initialize(capacity, 0.01F);
foreach (var item in dataSet1)
{
bloomFilter1.Add(item);
}
var bloomFilter2 = new InvertibleReverseBloomFilter <TestEntity, long, int>(configuration);
bloomFilter2.Initialize(capacity, 0.01F);
foreach (var item in dataSet2)
{
bloomFilter2.Add(item);
}
var s1 = new HashSet <long>();
var s2 = new HashSet <long>();
var s3 = new HashSet <long>();
decodeResult[run] = bloomFilter1.SubtractAndDecode(bloomFilter2, s1, s2, s3) == true ? 1 : 0;
var mods = s1.Union(s2).Union(s3).ToArray();
modCountResultAggregate[run] = mods.Count(v => onlyInSet1.Contains(v) ||
onlyInSet2.Contains(v) || modified.Contains(v));
decodesPerSize[run] = 1.0D * modCountResultAggregate[run] / bloomFilter1.Extract().Counts.Length;
countSize = bloomFilter1.Extract().Counts.Length;
}
writer
.WriteLine($"{capacity},{s * mod},{modCountResultAggregate.Average()},{countSize},{decodesPerSize.Average()},{Math.Sqrt(decodesPerSize.Variance())},{decodeResult.Average()}");
}
}
}
}
}