Snapshot Lock | snaplock

Harness MVCC-lock performance in-memory crates.io

This library provides ReadWriteLock - SnapshotLock, which has lock-free reads powered by Reference Counted clone that is stored near the locked value.

Choose SnapshotLock if:

• Reads vastly outnumber writes (10:1 ratio or higher)
• You need predictable read latency (scales linearly with minimal memory footprint)
• Application is read-performance critical
• You can tolerate slightly slower writes

Writes are protected by std::sync::RwLock. It also provides tokio::sync::RwLock implementation under the async feature.

Features:

• async : Adds new_async constructor. .write() method becomes async

Example:

use snaplock::SnapshotLock;

fn main() {
    let lock = SnapshotLock::new(String::from("Hello"));

    // Multiple readers can access the data concurrently
    let reader1 = lock.read();
    let reader2 = lock.read();
    assert_eq!(*reader1, "Hello");
    assert_eq!(*reader2, "Hello");

    // Writers have exclusive access
    {
        let mut writer = lock.write();
        writer.push_str(", World!");

        // During write readers see old values
        let reader = lock.read();
        assert_eq!(*reader, "Hello");
    }

    // When write is finished, new readers see the updated value
    let reader3 = lock.read();
    assert_eq!(*reader3, "Hello, World!");

    // Old readers still see their old value
    assert_eq!(*reader1, "Hello");
    assert_eq!(*reader2, "Hello");
}

MSRV Policy:

There is none. Current Minimum Supported Rust Version is 1.85.0

Alternatives:

• crossbeam::sync::SharedLock - A sharded reader-writer lock.
• vlock - A fast and scalable multi-version shared state lock with wait-free read access.
• parking_lot - More compact and efficient implementations of the standard synchronization primitives.

Benchmark:

Benchmark data was collected on:

• Windows 10 Version 10.0.19045 Build 19045
• AMD Ryzen 9 7900X3D 12-Core Processor

Results may vary based on hardware architecture and workload characteristics.

To run benchmarks use cargo run --release --example rwlock