b100m-filter

The fastest bloom filter in Rust. No accuracy compromises. Use any hasher.

Usage

# Cargo.toml
[dependencies]
b100m-filter = "0.3.0"

use b100m_filter::BloomFilter;

let num_blocks = 4; // by default, each block is 512 bits
let values = vec!["42", "🦀"];

let filter = BloomFilter::builder(num_blocks).items(values.iter());
assert!(filter.contains("42"));
assert!(filter.contains("🦀"));

use b100m_filter::BloomFilter;
use ahash::RandomState;

let num_blocks = 4; // by default, each block is 512 bits

let filter = BloomFilter::builder(num_blocks)
    .hasher(RandomState::default())
    .items(["42", "🦀"].iter());

Background

Bloom filters are a space efficient approximate membership set data structure. False positives from contains are possible, but false negatives are not, i.e. contains for all items in the set is guaranteed to return true, while contains for all items not in the set probably return false.

Blocked bloom filters are supported by an underlying bit vector, chunked into 512, 256, 128, or 64 bit "blocks", to track item membership. To insert, a number of bits, based on the item's hash, are set in the underlying bit vector. To check membership, a number of bits, based on the item's hash, are checked in the underlying bit vector.

Once constructed, neither the bloom filter's underlying memory usage nor number of bits per item change.

Implementation

b100m-filter is blazingly fast because it uses L1 cache friendly blocks and efficiently derives many index bits from only one hash per value. Compared to traditional implementations, b100m-filter is 2-5 times faster for small sets of items, and hundreds of times faster for larger item sets. In all cases, b100m-filter maintains competitive false positive rates.

Runtime Performance

Runtime comparison to other bloom filter crates:

• Bloom memory size = 16Kb
• 1000 contained items
• 364 hashes per item

*fastbloom-rs uses XXHash, which is faster than SipHash.

False Positive Performance

b100m-filter does not compromise accuracy. Below is a comparison false positive rate with other bloom filter crates:

Scaling

b100m-filter scales very well.

As the number of bits and set size increase, traditional bloom filters need to perform more hashes per item to keep false positive rates low. However, b100m-filter's optimal number of hashes is bounded while keeping near zero rates even for many items:

Bloom filter speed is directly proportional to number of hashes.

References

• Bloom Filter
• Less hashing, same performance: Building a better Bloom filter
• A fast alternative to the modulo reduction

License

Licensed under either of

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.