M5.BloomFilterBefore you search the whole collection, ask a better first question: could this key be here?
M5.BloomFilter checks set membership with a compact bit array. Rule out missing keys. Verify possible matches where your data lives.
A MILL5 source-available library for .NET 10
dotnet add package M5.BloomFilter --version 2.0.1Version 2.0.1 requires .NET 10. Upgrading from 1.x? Rebuild saved filters from the original keys.
Source available under PolyForm Strict 1.0.0. The LICENSE terms govern. Uses outside the license’s permitted purposes require separate MILL5 authorization.
Check a key to see which bits it needs.
Illustration: 64 bits, 3 hash positions. Cells show 0 or 1; outlines mark the selected key’s positions. A simple browser hash demonstrates the idea; the C# default is Murmur3KirschMitzenmacher.
One missing bit rules a key out. Skip the unnecessary lookup.
All bits are set? Check your database or original set to be certain.
The filter stores evidence of keys, not the keys themselves.
Choose an expected item count and a target false-positive probability. Add keys, then use Contains to decide whether to do an exact lookup.
This example plans for 100,000 keys with a 0.01 target error rate. That is a sizing input, not a measured guarantee.
Browse the C# sourceusing M5.BloomFilter;
var filter = FilterBuilder.Build(
expectedElements: 100_000,
errorRate: 0.01);
filter.Add("customer:42");
// true: possibly present
bool maybe = filter.Contains("customer:42");
// false means definitely absent.
// Confirm true results in your source of truth.FilterBuilder.Build(count, errorRate) creates an IBloomFilter. An overload accepts a HashMethod.
Add, Contains and All accept keys or collections through the provided extensions. Strings use UTF-8; preserve the same encoding and representation when checking keys.
ComputeHash exposes bit positions. Statistics exposes M (bits), K (hash count), N (expected items) and P (calculated ideal false-positive estimate).
ToImmutable() returns a detached ReadOnlyFilter snapshot for querying or serialization. Later additions to the original filter do not update the snapshot.
Built-in additions and membership checks synchronize access to filter bits. HashBits returns a detached copy; modifying it does not change the filter.
Async methods return completed tasks around in-memory work; they do not move hashing to a background thread.
False positives are allowed. False negatives are not part of the Bloom filter contract when bits, hashing and encoding remain consistent.
Overfilling increases false positives. Size for your workload and rebuild when your data outgrows that plan. There is no delete operation.
A Bloom filter is not an exact set, a security boundary, or a replacement for your original data.
Use the companion packages for JSON and protobuf. Both carry the bit array and the sizing and hash metadata needed to query the snapshot.
Add M5.BloomFilter.Serialization.Json. Register BloomFilterConverter before serializing a ReadOnlyFilter. Its nested bit-array converter is applied automatically.
using System.Text.Json;
using System.Text.Json.Serialization;
using M5.BloomFilter;
var options = new JsonSerializerOptions();
options.Converters.Add(new BloomFilterConverter());
var snapshot = (ReadOnlyFilter)filter.ToImmutable();
var json = JsonSerializer.Serialize(snapshot, options);
var restored = JsonSerializer
.Deserialize<ReadOnlyFilter>(json, options);Newtonsoft.Json is supported by converters in Newtonsoft.Json. Register its BloomFilterConverter with JsonSerializerSettings; nested bit arrays use their annotated converter.
Add M5.BloomFilter.Serialization.ProtoBuf. Wrap a filter in the serialization model, then restore it with AsBloomFilter().
using ProtoBuf;
var payload = new M5.BloomFilter
.Serialization.BloomFilter(filter);
using var stream = new MemoryStream();
Serializer.Serialize(stream, payload);
stream.Position = 0;
var decoded = Serializer.Deserialize<
M5.BloomFilter.Serialization.BloomFilter>(stream);
var restored = decoded.AsBloomFilter();Standalone bit-array packages are also available for JSON and protobuf. They serialize bits; a complete Bloom filter needs its hash and sizing metadata too.
M5.BloomFilter is a source-available C# Bloom filter library for .NET 10. It tests whether a key might belong to a set, using a compact bit array instead of storing the original keys.
A positive result means possibly present and can be a false positive. A negative result means definitely absent when the same filter, hash function, and key encoding are used correctly and the bits are intact. Confirm positive matches against your source of truth.
No. The requested error rate is a sizing target based on the expected number of elements and the usual hash-distribution assumptions. Inserting more items than planned increases the false-positive probability.
FilterBuilder.Build uses Murmur3KirschMitzenmacher by default. The browser illustration uses a separate simple teaching hash and does not run the C# library.
Version 2 payloads identify the current hash and encoding contract. Older or unversioned payloads must be rebuilt from original keys; changing a version number cannot migrate existing bits.
M5.BloomFilter is a MILL5 source-available library, built by Richard Crane, Microsoft MVP and founder of MILL5. The source code and issue tracker are on GitHub.
From the MILL5 developer tools family: practical building blocks for developers, alongside FastSearch, Pineapple and yocto.