#decode #encode #byte

nightly ed

Encode and Decode traits

11 releases

0.3.0 Feb 16, 2022
0.2.2 Sep 4, 2021
0.2.1 Aug 31, 2021
0.1.6 Mar 15, 2021
0.1.2 Jul 26, 2020

#819 in Rust patterns

Download history 85/week @ 2022-08-14 77/week @ 2022-08-21 44/week @ 2022-08-28 115/week @ 2022-09-04 159/week @ 2022-09-11 151/week @ 2022-09-18 68/week @ 2022-09-25 117/week @ 2022-10-02 161/week @ 2022-10-09 107/week @ 2022-10-16 184/week @ 2022-10-23 129/week @ 2022-10-30 371/week @ 2022-11-06 422/week @ 2022-11-13 376/week @ 2022-11-20 326/week @ 2022-11-27

1,499 downloads per month
Used in 2 crates

MIT license

652 lines

ed (encode/decode)

A minimalist crate for deterministic binary encodings in Rust.

CI codecov Crate API


This crate provides Encode and Decode traits which can be implemented for any type that can be converted to or from bytes, and implements these traits for many built-in Rust types. It also provides derive macros so that Encode and Decode can be easily derived for structs.

ed is far simpler than serde because it does not attempt to create an abstraction which allows arbitrary kinds of encoding (JSON, MessagePack, etc.), and instead forces focuses on binary encodings. It is also significantly faster than bincode, the leading binary serde serializer.

One aim of ed is to force top-level type authors to design their own encoding, rather than attempting to provide a one-size-fits-all encoding scheme. This lets users of ed be sure their encodings are as effiient as possible, and makes it easier to understand the encoding for compatability in other languages or libraries (contrasted with something like bincode, where it is not obvious how a type is being encoded without understanding the internals of bincode).

Another property of this crate is a focus on determinism (important for cryptographically hashed types) - built-in encodings are always big-endian and there are no provided encodings for floating point numbers or usize.


use ed::{Encode, Decode};

// traits are implemented for built-in types
let bytes = 123u32.encode()?; // `bytes` is a Vec<u8>
let n = u32::decode(bytes.as_slice())?; // `n` is a u32

// derive macros are available
#[derive(Encode, Decode)]
struct Foo {
  bar: (u32, u32),
  baz: Vec<u8>

// encoding and decoding can be done in-place to reduce allocations
let mut bytes = vec![0xba; 40];
let mut foo = Foo {
  bar: (0, 0),
  baz: Vec::with_capacity(32)

// in-place decode, re-using pre-allocated `foo.baz` vec
assert_eq!(foo, Foo {
  bar: (0xbabababa, 0xbabababa),
  baz: vec![0xba; 32]

// in-place encode, into pre-allocated `bytes` vec
foo.encode_into(&mut bytes)?;


~15K SLoC