public void ParallelInvertibleAddDifferentSizesTest()
{
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 DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(4 * size, errorRate);
Parallel.ForEach(Partitioner.Create(testData, true), d => bloomFilter.Add(d));
var bloomFilter2 = new InvertibleBloomFilter <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);
Parallel.ForEach(Partitioner.Create(testData2, true), d => bloomFilter2.Add(d));
//add the bloom filters.
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 InvertibleRemoveKeyTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <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.RemoveKey(item.Id);
}
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 InvertibleIntersectDifferentFiltersTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var bloomFilter2 = new InvertibleBloomFilter <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 > 7800);
}
public void InvertibleIntersectEqualFiltersTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var bloomFilter2 = new InvertibleBloomFilter <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 InvertibleAddTest()
{
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 DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(2 * size, errorRate);
foreach (var itm in testData)
{
bloomFilter.Add(itm);
}
var bloomFilter2 = new InvertibleBloomFilter <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 InvertibleBloomFilterQuasiDecodeTest()
{
var size = 100000;
var data = DataGenerator.Generate().Take(size).ToList();
var errorRate = 0.001F;
var configuration = new DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(size, errorRate);
foreach (var itm in data)
{
bloomFilter.Add(itm);
}
data = DataGenerator.Generate().Skip(500).Take(8000).ToList();
data.Modify(1000);
var estimate = bloomFilter.QuasiDecode(data);
Assert.IsTrue(estimate > 90500 && estimate < 97000, "Unexpected estimate for difference.");
}
public void InvertibleSimpleFold()
{
var addSize = 50;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var size = testData.Length;
var configuration = new DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <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.Extract().BlockSize);
Assert.IsTrue(testData.All(item => bloomFilter.Contains(item)), "False negative found");
}
public void InvertibleBloomFilterSetDiffTest()
{
var addSize = 1000;
var modCount = 50;
var dataSet1 = DataGenerator.Generate().Take(addSize).ToList();
var dataSet2 = DataGenerator.Generate().Take(addSize).ToList();
dataSet2.Modify(modCount);
var configuration = new DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <TestEntity, long, sbyte>(configuration);
bloomFilter.Initialize(10 * modCount, 0.0001F);
foreach (var itm in dataSet1)
{
bloomFilter.Add(itm);
}
var secondBloomFilter = new InvertibleBloomFilter <TestEntity, long, sbyte>(configuration);
secondBloomFilter.Initialize(10 * modCount, 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'");
//very bad at recognizing changes.
Assert.IsTrue(changed.Count <= modified.Length, "Incorrect number of modified items detected");
}
public void InvertibleCompressTest()
{
var addSize = 10000;
var errorRate = 0.001F;
var data = DataGenerator.Generate().Take(addSize).ToArray();
var filter = new InvertibleBloomFilter <TestEntity, long, sbyte>(new DefaultBloomFilterConfiguration());
filter.Initialize(50 * data.Length, errorRate);
Assert.AreEqual(filter.Capacity, 500000, "Unexpected size of Bloom filter.");
foreach (var item in data)
{
filter.Add(item);
}
//check error rate.
var notFoundCount = DataGenerator.Generate().Skip(addSize).Take(10000).Count(itm => filter.Contains(itm));
Assert.IsTrue(notFoundCount <= errorRate * addSize, "Uncompressed Bloom filter exceeded error rate.");
filter.Compress(true);
Assert.AreEqual(filter.Capacity, 12820, "Unexpected size of compressed Bloom filter.");
var compressNotFoundCount = DataGenerator.Generate().Skip(addSize).Take(10000).Count(itm => filter.Contains(itm));
Assert.IsTrue(compressNotFoundCount <= errorRate * addSize, "Compressed Bloom filter exceeded error rate.");
}
public void InvertibleFalsePositiveTest()
{
var addSize = 10000;
var testData = DataGenerator.Generate().Take(addSize).ToArray();
var errorRate = 0.001F;
var size = testData.Length;
var configuration = new DefaultBloomFilterConfiguration();
var bloomFilter = new InvertibleBloomFilter <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");
}
/// <summary>
/// Create filters
/// </summary>
/// <param name="estimatorData">Filter data to rehydrate.</param>
private void CreateFilters(IStrataEstimatorData <int, TCount> estimatorData = null)
{
var configuration = Configuration.ConvertToEstimatorConfiguration();
HashFunctionCount = configuration.BestHashFunctionCount(BlockSize, ErrorRate);
for (var idx = 0; idx < StrataFilters.Length; idx++)
{
if (idx >= MaxStrata)
{
StrataFilters[idx] = null;
continue;
}
var filterData = estimatorData.GetFilterForStrata(idx);
//lazily create Strata filters.
StrataFilters[idx] = new Lazy <InvertibleBloomFilter <KeyValuePair <int, int>, int, TCount> >(() =>
{
var res = new InvertibleBloomFilter <KeyValuePair <int, int>, int, TCount>(configuration);
//capacity doesn't really matter, the capacity is basically the block size.
res.Initialize(BlockSize, BlockSize, HashFunctionCount);
res.Rehydrate(filterData);
return(res);
});
}
}