🚀 Announcing beve-go: Blazing-Fast, GC-Friendly Serialization is Here!

[meftunca]

Hey everyone,

The day has finally come! We are incredibly excited and proud to announce that beve-go, the official Go implementation of the Beve specification, is complete, stable, and ready for production use.

Our mission was to create not just another serialization format, but a truly modern, high-performance library engineered for today's hardware and the demands of low-latency applications. After countless hours of development and rigorous testing, we believe we have achieved just that.

But we won't just ask you to take our word for it. The results speak for themselves.

---

🏆 Benchmark Highlights: The Proof is in the Performance

We've benchmarked beve-go against the most popular serialization libraries in the Go ecosystem, including encoding/json, bytedance/sonic, vmihailenco/msgpack, and fxamacker/cbor. The full, detailed results are available right here in our repository, but here are the key takeaways:

⚡ Jaw-Dropping Unmarshal Speed This is where Beve truly shines. beve-go is astoundingly fast at deserialization, making it perfect for read-heavy workloads like API gateways, data pipelines, and configuration loading.

• Over 16x faster than Go's standard encoding/json.
• Significantly outperforms other high-performance binary formats like Sonic and MessagePack.

🚀 Dominant Marshal Performance beve-go leads the pack in serialization speed across all tested data sizes—small, medium, and large. Whether you're sending a small API response or a large data record, Beve ensures minimal overhead.

🧠 Engineered for Go's Garbage Collector Performance isn't just about speed; it's also about predictable latency. We meticulously designed beve-go to be incredibly memory-efficient.

• Marshalling operations complete with just 2 allocations, regardless of data size.
• This drastically reduces GC pressure, preventing performance stutters and making beve-go ideal for high-throughput, low-latency services.

📦 A Perfect Balance of Speed and Size While hyper-optimized for speed, Beve also produces a compact binary output, consistently smaller than JSON. It strikes the perfect balance, saving you network bandwidth without compromising on its best-in-class performance.

---

What Makes beve-go Different?

• Developer-Friendly: Beve is a tagged (schemaless) format. There's no code generation step and no .proto files to manage. Your Go structs are the source of truth, making development fast and flexible.
• Modern by Design: Beve is built to leverage the capabilities of modern CPUs, forming a solid foundation for data-intensive applications.

Get Started Today!

We built Beve to be the high-performance, efficient, and enjoyable serialization library we always wanted for our own Go projects. Now, we're thrilled to share it with the world.

• ➡️ Explore the repository and benchmarks:https://github.com/beve-org/beve-go
• 👩‍💻 Give it a try! It's easy to get started. Just go get github.com/beve-org/beve-go and check out the examples in our documentation.
• 💬 Join the conversation! Have questions, ideas, or feedback? This is the place for it. We'd love to hear from you.

A huge thank you to everyone who contributed and provided feedback along the way. We truly can't wait to see what you build with Beve!

Best,

The Beve Team

[meftunca]

@stephenberry I'm waiting for you to test beni-go when you have a chance, my friend.

[stephenberry]

Nice! I'll have to look over it a bit more, but this looks like a discrepancy in the README:

CBOR: 1,225 bytes ← Smallest
BEVE: 955 bytes ← 22% smaller than JSON!

It looks like BEVE is actually the smallest

[meftunca]

Along with my latest revisions, I will share a more optimized version tomorrow. I have also set up a GitHub Actions configuration to enable performance testing across multiple systems. Additionally, I will share full support for BEVE on the Go side and examples of its usage.