public void HybridIntersectEqualFiltersTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var bloomFilter2 = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter2.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter2.Add(itm);
}
bloomFilter.Intersect(bloomFilter2);
Assert.AreEqual(addSize, bloomFilter.ItemCount);
Assert.IsTrue(testData.All(bloomFilter.Contains));
}
public void HybridInvertibleBloomFilterEmptySetDiffTest()
{
var addSize = 1000;
var dataSet1 = DataGenerator.Generate().Take(0).ToList();
var dataSet2 = DataGenerator.Generate().Take(addSize).ToList();
var configuration = new HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(10 * addSize, 0.0001F);
foreach (var itm in dataSet1)
{
bloomFilter.Add(itm);
}
var secondBloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
secondBloomFilter.Initialize(10 * addSize, 0.0001F);
foreach (var itm in dataSet2)
{
secondBloomFilter.Add(itm);
}
var changed = new HashSet <long>();
var onlyInFirst = new HashSet <long>();
var onlyInSecond = new HashSet <long>();
var decoded = bloomFilter
.SubtractAndDecode(secondBloomFilter, onlyInFirst, onlyInSecond, changed);
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();
Assert.IsTrue(decoded.HasValue, "Decoding failed");
Assert.IsTrue(onlyInSet1.Length == onlyInFirst.Count, "Incorrect number of changes detected on 'only in set 1'");
Assert.IsTrue(onlyInSet2.Length == onlyInSecond.Count, "Incorrect number of changes detected on 'only in set 2'");
Assert.IsTrue(changed.Count == modified.Length, "Incorrect number of modified items detected");
}
public void HybridIntersectDifferentFiltersTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var bloomFilter2 = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter2.Initialize(2 * size, errorRate);
foreach (var itm in testData.Skip(1000))
{
bloomFilter2.Add(itm);
}
bloomFilter.Intersect(bloomFilter2);
Assert.AreEqual(9000, bloomFilter.ItemCount);
var count = testData.Skip(1000).Count(bloomFilter.Contains);
//Note: intersect introduces a horrible error rate when utilizing XOR, so don't actually use intersect.
//There are however definitions of operations possible where the intersect would not have this horrible effect.
Assert.IsTrue(count > 6000);
Assert.IsTrue(testData.Take(1000).All(d => !bloomFilter.Contains(d)));
}
public void HybridAddDifferentSizesTest()
{
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 HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(4 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var bloomFilter2 = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
//We should create a foldable counter part.
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 HybridAddTest()
{
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 HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var bloomFilter2 = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter2.Initialize(2 * size, errorRate);
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 HybridRemoveKeyTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var contained = testData.Count(item => bloomFilter.Contains(item));
try
{
foreach (var item in testData.Take(addSize / 2))
{
bloomFilter.RemoveKey(item.Id);
}
Assert.Fail("RemoveKey should not be supported by a hybrid invertible Bloom filter");
}
catch (NotSupportedException)
{ };
}
public void HybridRemoveItemTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var contained = testData.Count(item => bloomFilter.Contains(item));
foreach (var item in testData.Take(addSize / 2))
{
bloomFilter.Remove(item);
}
var containedAfterRemove = testData.Count(item => bloomFilter.Contains(item));
//tricky: assuming zero false positives.
Assert.AreEqual(contained, containedAfterRemove * 2, "Wrong item count after removal.");
}
public void HybridSimpleFold()
{
var addSize = 50;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var size = testData.Length;
var configuration = new HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(size, 1024, (uint)3);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var positiveCount = DataGenerator.Generate().Take(500).Count(itm => bloomFilter.Contains(itm));
var folded = bloomFilter.Fold(4);
var positiveCountAfterFold = DataGenerator.Generate().Take(500).Count(itm => bloomFilter.Contains(itm));
Assert.AreEqual(positiveCount, positiveCountAfterFold, "False positive count different after fold");
Assert.AreEqual(256, folded.BlockSize, "Folded block size is unexpected.");
Assert.IsTrue(testData.All(item => bloomFilter.Contains(item)), "False negative found");
}
public void HybridCompressTest()
{
var addSize = 10000;
var errorRate = 0.001F;
var data = DataGenerator.Generate().Take(addSize).ToArray();
var hybridFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(new HybridDefaultBloomFilterConfiguration());
hybridFilter.Initialize(50 * data.Length, errorRate);
Assert.AreEqual(hybridFilter.Capacity, 500000, "Unexpected size of hybrid Bloom filter.");
foreach (var item in data)
{
hybridFilter.Add(item);
}
//check error rate.
var notFoundCount = DataGenerator.Generate().Skip(addSize).Take(10000).Count(itm => hybridFilter.Contains(itm));
Assert.IsTrue(notFoundCount <= errorRate * addSize, "Uncompressed hybrid Bloom filter exceeded error rate.");
hybridFilter.Compress(true);
Assert.AreEqual(hybridFilter.Capacity, 12820, "Unexpected size of compressed hybrid Bloom filter.");
var compressNotFoundCount = DataGenerator.Generate().Skip(addSize).Take(10000).Count(itm => hybridFilter.Contains(itm));
Assert.IsTrue(compressNotFoundCount <= errorRate * addSize, "Compressed hybrid Bloom filter exceeded error rate.");
}
public void HybridFalsePositiveTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new HybridDefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleHybridBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var notFoundCount = testData.Count(itm => !bloomFilter.Contains(itm));
Assert.IsTrue(notFoundCount == 0, "False negative error rate violated");
notFoundCount = testData.Count(itm => !bloomFilter.ContainsKey(itm.Id));
Assert.IsTrue(notFoundCount == 0, "False negative error rate violated on ContainsKey");
notFoundCount = DataGenerator.Generate().Skip(addSize).Take(addSize).Count(itm => bloomFilter.Contains(itm));
Assert.IsTrue(notFoundCount <= errorRate * addSize, "False positive error rate violated");
notFoundCount = DataGenerator.Generate().Skip(addSize).Take(addSize).Count(itm => bloomFilter.ContainsKey(itm.Id));
Assert.IsTrue(notFoundCount <= errorRate * addSize, "False positive error rate violated on ContainsKey");
}
public void HybridIbfDecodePerformance()
{
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($"hibfdecode-{s}.csv",
FileMode.Create)))
{
writer.WriteLine("timeInMs,sizeInBytes,capacity,modCount,detectedModCount,countDiff,countDiffSd,decodeSuccessRate,listADiff,listBDiff,listCDiff");
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 listADiff = new int[50];
var listBDiff = new int[50];
var listCDiff = 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 InvertibleHybridBloomFilter <TestEntity, long, int>(configuration);
bloomFilter1.Initialize((int)(idealCapacity * capacityPercentage), 0.01F);
foreach (var item in dataSet1)
{
bloomFilter1.Add(item);
}
var bloomFilter2 = new InvertibleHybridBloomFilter <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));
listADiff[run] = s1.Count() - onlyInSet1.Count();
listBDiff[run] = s2.Count() - onlyInSet2.Count();
listCDiff[run] = s3.Count() - modified.Count();
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},{listADiff.Average()},{listBDiff.Average()},{listCDiff.Average()}");
}
}
}
}
}
