Librum is a Rust crate for cheaply serialising (encoding) and deserialising (decoding) data structures into binary streams
What separates this crate from others such as Bincode or Postcard is that this crate is extensively optimised for just binary encodings (whilst the mentioned crates specifically use Serde and build on a more abstract data model). The original goal of this project was specifically to guarantee size constraints for encodings on a per-type basis at compile-time. Therefore, this crate may be more suited for networking or other cases where many allocations are unwanted.
Keep in mind that this project is still work-in-progress. Until the interfaces are stabilised, different facilities may be replaced, removed, or altered in a breaking way.
This crate is compatible with no_std
.
As Librum is optimised exclusively for a single, binary format, it may outperform other libraries that are more generic in nature.
The librum-benchmarks
binary compares multiple scenarios using Librum and other, similar crates.
According to my runs on an AMD Ryzen 7 3700X, these benchmarks indicate that Librum outperform all of the tested crates -- as demonstrated in the following table:
Benchmark | Bincode | Borsh | Librum | Postcard |
---|---|---|---|---|
encode_u8 |
1.306 | 1.315 | 1.150 | 1.304 |
encode_u32 |
1.321 | 1.317 | 1.146 | 3.016 |
encode_u128 |
2.198 | 2.103 | 1.509 | 6.376 |
encode_struct_unit |
0.000 | 0.000 | 0.000 | 0.000 |
encode_struct_unnamed |
1.362 | 1.448 | 1.227 | 2.659 |
encode_struct_named |
3.114 | 1.530 | 0.969 | 3.036 |
encode_enum_unit |
0.252 | 0.297 | 0.000 | 0.299 |
Total time → | 9.553 | 8.010 | 6.001 | 16.691 |
Total deviation (p.c.) → | +59 | +33 | ±0 | +178 |
All quantities are measured in seconds unless otherwise noted. Please feel free to conduct your own tests of Librum.
Most primitives encode losslessly, with the main exceptions being usize
and isize
.
These are instead first cast as u16
and i16
, respectively, due to portability concerns (with respect to embedded systems).
See specific types' implementations for notes on their data models.
Note that the data model is currently not stabilised, and may not necessarily be in the near future (at least before specialisation). It may therefore be undesired to store encodings long-term.
This crate revolves around the Encode
and Decode
traits which both handle conversions to and from byte streams.
Many standard types come implemented with Librum, including most primitives as well as some standard library types such as Option
and Result
.
Some features enable an extended set of implementations.
It is recommended in most cases to simply derive these two traits for custom types (although this is only supported with enumerations and structures -- not untagged unions). Here, each field is chained according to declaration order:
use librum::{Buf, Decode, Encode};
##[derive(Debug, Decode, Encode, PartialEq)]
struct Ints {
value0: u8,
value1: u16,
value2: u32,
value3: u64,
value4: u128,
}
const VALUE: Ints = Ints {
value0: 0x00,
value1: 0x02_01,
value2: 0x06_05_04_03,
value3: 0x0E_0D_0C_0B_0A_09_08_07,
value4: 0x1E_1D_1C_1B_1A_19_18_17_16_15_14_13_12_11_10_0F,
};
let mut buf = Buf::with_capacity(0x100);
buf.write(VALUE).unwrap();
assert_eq!(buf.len(), 0x1F);
assert_eq!(
buf,
[
0x00, 0x02, 0x01, 0x06, 0x05, 0x04, 0x03, 0x0E,
0x0D, 0x0C, 0x0B, 0x0A, 0x09, 0x08, 0x07, 0x1E,
0x1D, 0x1C, 0x1B, 0x1A, 0x19, 0x18, 0x17, 0x16,
0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0F,
].as_slice(),
);
assert_eq!(buf.read().unwrap(), VALUE);
The Encode
and Decode
traits both rely on streams for carrying the manipulated bytes.
These streams are separated into two type: O-streams (output streams) and i-streams (input streams).
The Buf
type can be used to handle these streams.
To encode an object directly using the Encode
trait, simply allocate a buffer for the encoding and wrap it in an OStream
object:
use librum::{Encode, OStream, SizedEncode};
let mut buf = [0x00; char::MAX_ENCODED_SIZE];
let mut stream = OStream::new(&mut buf);
'Ж'.encode(&mut stream).unwrap();
assert_eq!(buf, [0x00, 0x00, 0x04, 0x16].as_slice());
Streams can also be used to chain multiple objects together:
use librum::{Encode, OStream, SizedEncode};
let mut buf = [0x0; char::MAX_ENCODED_SIZE * 0x5];
let mut stream = OStream::new(&mut buf);
// Note: For serialising multiple characters, the
// `String` and `SizedStr` types are usually
// preferred.
'ل'.encode(&mut stream).unwrap();
'ا'.encode(&mut stream).unwrap();
'م'.encode(&mut stream).unwrap();
'د'.encode(&mut stream).unwrap();
'ا'.encode(&mut stream).unwrap();
assert_eq!(buf, [
0x00, 0x00, 0x06, 0x44, 0x00, 0x00, 0x06, 0x27,
0x00, 0x00, 0x06, 0x45, 0x00, 0x00, 0x06, 0x2F,
0x00, 0x00, 0x06, 0x27
]);
If the encoded type additionally implements SizedEncode
, then the maximum size of any encoding is guaranteed with the MAX_ENCODED_SIZE
constant.
Numerical primitives are encoded in big endian (a.k.a. network order) for... reasons. It is recommended for implementors to follow this convention as well.
Decoding works with a similar syntax to encoding.
To decode a byte array, simply call the decode
method with an IStream
object:
use librum::{Decode, IStream};
let data = [0x45, 0x54];
let mut stream = IStream::new(&data);
assert_eq!(u16::decode(&mut stream).unwrap(), 0x4554);
// Data can theoretically be reinterpretred:
stream = IStream::new(&data);
assert_eq!(u8::decode(&mut stream).unwrap(), 0x45);
assert_eq!(u8::decode(&mut stream).unwrap(), 0x54);
// Including as tuples:
stream = IStream::new(&data);
assert_eq!(<(u8, u8)>::decode(&mut stream).unwrap(), (0x45, 0x54));
A UDP server/client for geographic data:
use librum::{Buf, Encode, Decode, SizedEncode};
use std::io;
use std::net::{SocketAddr, ToSocketAddrs, UdpSocket};
use std::thread::spawn;
// City, region, etc.:
##[derive(Clone, Copy, Debug, Decode, Encode, Eq, PartialEq, SizedEncode)]
enum Area {
AlQuds,
Byzantion,
Cusco,
Tenochtitlan,
// ...
}
// Client-to-server message:
##[derive(Debug, Decode, Encode, PartialEq, SizedEncode)]
enum Request {
AtmosphericHumidity { area: Area },
AtmosphericPressure { area: Area },
AtmosphericTemperature { area: Area },
// ...
}
// Server-to-client message:
##[derive(Debug, Decode, Encode, PartialEq, SizedEncode)]
enum Response {
AtmosphericHumidity(f64),
AtmosphericPressure(f64), // Pascal
AtmosphericTemperature(f64), // Kelvin
// ...
}
struct Party {
pub socket: UdpSocket,
pub request_buf: Buf::<Request>,
pub response_buf: Buf::<Response>,
}
impl Party {
pub fn new<A: ToSocketAddrs>(addr: A) -> io::Result<Self> {
let socket = UdpSocket::bind(addr)?;
let this = Self {
socket,
request_buf: Buf::new(),
response_buf: Buf::new(),
};
Ok(this)
}
}
let mut server = Party::new("127.0.0.1:27015").unwrap();
let mut client = Party::new("0.0.0.0:0").unwrap();
spawn(move || {
let Party { socket, mut request_buf, mut response_buf } = server;
// Recieve initial request from client.
let (len, addr) = socket.recv_from(&mut request_buf).unwrap();
request_buf.set_len(len);
let request = request_buf.read().unwrap();
assert_eq!(request, Request::AtmosphericTemperature { area: Area::AlQuds });
// Handle request and respond back to client.
let response = Response::AtmosphericTemperature(44.4); // For demonstration's sake.
response_buf.write(response).unwrap();
socket.send_to(&response_buf, addr).unwrap();
});
spawn(move || {
let Party { socket, mut request_buf, mut response_buf } = client;
// Send initial request to server.
socket.connect("127.0.0.1:27015").unwrap();
let request = Request::AtmosphericTemperature { area: Area::AlQuds };
request_buf.write(request);
socket.send(&request_buf).unwrap();
// Recieve final response from server.
socket.recv(&mut response_buf).unwrap();
let response = response_buf.read().unwrap();
assert_eq!(response, Response::AtmosphericTemperature(44.4));
});
Librum defines the following features:
- *
alloc
: Enables theBuf
type and implementations for e.g.Box
andArc
- *
proc-macro
: Pulls the procedural macros from thelibrum_macros
crate - *
std
: Enables implementations for types such asMutex
andRwLock
Features marked with * are enabled by default.
Librum has its documentation written in-source for use by rustdoc
.
See Docs.rs for an on-line, rendered instance.
Currently, these docs make use of some unstable features for the sake of readability. The nightly toolchain is therefore required when rendering them.
Librum does not accept source code contributions at the moment. This is a personal choice by the maintainer and may be undone in the future.
Do however feel free to open up an issue on GitLab or (preferably) GitHub if you feel the need to express any concerns over the project.
Copyright 2024 Gabriel Bjørnager Jensen.
This program is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License along with this program. If not, see https://www.gnu.org/licenses/.