algebra VS rust_serialization_benchmark

Love seeing the work on modes! There's definitely a use for this in cryptography, where you might want to serialize things containing elliptic curve points, which can be serialized in both "compressed" form and "uncompressed" form. We make extensive use of this in our serialization framework in arkworks: https://github.com/arkworks-rs/algebra/tree/master/serialize

Rust takes the cake in the blockchain space: Substrate, Cosmos (CosmWasm), and Solana. All of the zero knowledge cryptography libraries used for layer 2 solutions are written in Rust, compiling to Wasm (see arkworks, Risc0). Ethereum's next version of smart contracts will even use a restricted subset of Wasm ("Ewasm") instead of EVM.

Is ark a prefix Embark is using for all their Rust crates, or is it a one-off name for your future crate? If so, it might collide with our naming convention in the arkworks ecosystem: arkworks.rs

You can take a look at our approach in the arkworks library: https://github.com/arkworks-rs/algebra/blob/920070c60d481a29fb3c262ef9579f34cbb053a6/ff/src/fields/macros.rs#L103

https://github.com/djkoloski/rust_serialization_benchmark

Apache/arrow-rs: https://github.com/apache/arrow-rs

From https://arrow.apache.org/faq/ :

> How does Arrow relate to Flatbuffers?

> Flatbuffers is a low-level building block for binary data serialization. It is not adapted to the representation of large, structured, homogenous data, and does not sit at the right abstraction layer for data analysis tasks.

> Arrow is a data layer aimed directly at the needs of data analysis, providing a comprehensive collection of data types required to analytics, built-in support for “null” values (representing missing data), and an expanding toolbox of I/O and computing facilities.

> The Arrow file format does use Flatbuffers under the hood to serialize schemas and other metadata needed to implement the Arrow binary IPC protocol, but the Arrow data format uses its own representation for optimal access and computation

Nice that comfy gets even easier. Also, if serde's compile time is an issue, then there's nanoserde which is usually much much faster according to benchmarks

A note on performance and size: Some benchmarks and statistics are included in the README. But only because people will be curious. I've based my methodology on rust_serialization_benchmark, but decided to not extend it (for now) since it seems to exclude any Rust types which are not widely supported by all formats being tested (like HashMap's and 128-bit numbers). The test suite is already quite nice if you want to take it for a spin.

While we haven't benchmarked either of those ourselves. You can checkout rust_serialization_benchmark which has protobuf under the name prost.

Update: Benchmark PR submitted: https://github.com/djkoloski/rust_serialization_benchmark/pull/37

Here is a speed and size benchmark of different rust binary serialization formats: https://github.com/djkoloski/rust_serialization_benchmark Warning: I think the creator of this benchmark is also the creator of rkyv, one of the best positioned formats in the benchmark.

So I assume you're deserializing them before processing the book? If so then if you want an easy speed-up you could also take a look at these benchmarks and pick a faster serialization crate. (: (Although you might or might not get a big speedup; depends on what exactly you're deserializing and how much you are deserializing.)

You can add the rust serialization benchmark to that list

Hey! Similar to bincode, it provides a very similar, compact binary format. The rkyv benchmark is the most comprehensive I'm aware of, but compared to bincode, postcard is generally a similar speed for serialization or deserialization (maybe a touch slower), but generally produces a slightly smaller "on the wire" size.

You should add your crate to these benchmarks. (Which are, AFAIK, the most comprehensive set of benchmarks currently available for Rust serialization libraries.)