Crates.io | rbase64 |
lib.rs | rbase64 |
version | 2.0.3 |
source | src |
created_at | 2022-11-11 20:29:26.645151 |
updated_at | 2022-11-28 02:09:06.024589 |
description | A fast multi-threaded base64 encoding library and CLI tool |
homepage | https://github.com/uhmarcel/rbase64 |
repository | https://github.com/uhmarcel/rbase64 |
max_upload_size | |
id | 713040 |
size | 140,115 |
A fast multi-threaded base64 encoding / decoding library and CLI tool, made in Rust.
Have you ever wanted to base64 encode tens of gigabytes of data, but felt your encoder was not fast enough? Me neither. Either way, rbase64 is up to the task.
The goal of rbase64 is to provide a fast implementation of base64 encoding. The library is thoroughly tested and optimized for high throughput. As of v2.0.3, rbase64 is able to achieve up to 11.133 GiB/s encoding and 14.251 GiB/s decoding rates (measured in MacBook Air M1 2020 laptop, 10~20 MiB batches).
The project also comes with a command line interface powered by the library encoder. While limited by I/O, it still achieves high performance compared to other alternatives (for example, GNU base64 by a factor of 11x).
For details, see Performance.
Add rbase64 to your Cargo.toml dependencies:
[dependencies]
rbase64 = "2.0.3"
Sample usage:
use rbase64;
fn main() {
let a = b"Hello world";
let b = "SGVsbG8gd29ybGQ=";
assert_eq!(rbase64::encode(a), b);
assert_eq!(&rbase64::decode(b).unwrap(), a);
}
Installation:
$ cargo install rbase64 --features cli
Usage: rbase64 [OPTIONS]
Options | Description |
---|
-d, --decode | Decode input (default: false) -i, --input <INPUT> | Input file (stdin if missing) -o, --output <OUTPUT> | Output file (stdout if missing) -h, --help | Print help information -V, --version | Print version information
Rbase64 CLI is designed to have the same API as classic GNU base64, however, it does not append newline characters to file or stdout output.
Basic usage:
$ echo -n "Hello world" | rbase64
SGVsbG8gd29ybGQ=
$ echo -n "SGVsbG8gd29ybGQ=" | rbase64 --decode
Hello world
Performance is measured using benchmarks with Criterion. Each bench measures encoding/decoding speed and throughput at varying input sizes in bytes. Measurements below were taken by running benchmarks on a MacBook Air M1 2020 laptop. Results are compared against three of the top most widely used Rust base64 encoding libraries in cargo.io.
Bench (type / bytes) | rbase64 (v2.0.3) | base64 (v0.20.0) | data-encoding (v2.3.2) | rustc-serialize (v0.3.24) |
---|---|---|---|---|
encode/3 | 186.21 MiB/s | 77.240 MiB/s | 77.208 MiB/s | 91.615 MiB/s |
encode/50 | 1.3780 GiB/s | 904.74 MiB/s | 882.87 MiB/s | 806.29 MiB/s |
encode/100 | 1.7103 GiB/s | 1.2844 GiB/s | 1.3009 GiB/s | 1.0334 GiB/s |
encode/500 | 2.5775 GiB/s | 1.8897 GiB/s | 1.9288 GiB/s | 1.2626 GiB/s |
encode/3072 | 2.7913 GiB/s | 2.3780 GiB/s | 2.4211 GiB/s | 1.4501 GiB/s |
encode/51200 | 2.8454 GiB/s | 2.4486 GiB/s | 2.4678 GiB/s | 1.4765 GiB/s |
encode/102400 | 2.8644 GiB/s | 2.4247 GiB/s | 2.4756 GiB/s | 1.4799 GiB/s |
encode/512000 | 6.1009 GiB/s | 2.5105 GiB/s | 2.5174 GiB/s | 1.4623 GiB/s |
encode/1048576 | 8.0095 GiB/s | 2.5115 GiB/s | 2.5182 GiB/s | 1.4506 GiB/s |
encode/5242880 | 10.042 GiB/s | 2.1806 GiB/s | 2.2009 GiB/s | 1.3416 GiB/s |
encode/10485760 | 11.133 GiB/s | 2.0549 GiB/s | 2.1037 GiB/s | 1.3438 GiB/s |
encode/20971520 | 10.804 GiB/s | 2.0814 GiB/s | 2.0638 GiB/s | 1.3138 GiB/s |
Bench (type / bytes) | rbase64 (v2.0.3) | base64 (v0.20.0) | data-encoding (v2.3.2) | rustc-serialize (v0.3.24) |
---|---|---|---|---|
decode/3 | 235.57 MiB/s | 52.127 MiB/s | 55.164 MiB/s | 96.523 MiB/s |
decode/50 | 1.9261 GiB/s | 540.87 MiB/s | 728.58 MiB/s | 550.51 MiB/s |
decode/100 | 2.5807 GiB/s | 942.93 MiB/s | 1.1012 GiB/s | 579.12 MiB/s |
decode/500 | 3.3135 GiB/s | 1.8286 GiB/s | 1.7942 GiB/s | 563.76 MiB/s |
decode/3072 | 3.4941 GiB/s | 2.9708 GiB/s | 2.2418 GiB/s | 609.78 MiB/s |
decode/51200 | 3.7189 GiB/s | 3.0924 GiB/s | 2.3634 GiB/s | 618.20 MiB/s |
decode/102400 | 3.7623 GiB/s | 3.0878 GiB/s | 2.4015 GiB/s | 618.18 MiB/s |
decode/512000 | 7.0065 GiB/s | 3.1235 GiB/s | 2.4229 GiB/s | 618.47 MiB/s |
decode/1048576 | 9.3497 GiB/s | 3.1195 GiB/s | 2.4073 GiB/s | 617.00 MiB/s |
decode/5242880 | 12.148 GiB/s | 2.8267 GiB/s | 2.2475 GiB/s | 600.88 MiB/s |
decode/10485760 | 13.424 GiB/s | 2.7416 GiB/s | 2.2384 GiB/s | 599.09 MiB/s |
decode/20971520 | 14.251 GiB/s | 2.6512 GiB/s | 2.1692 GiB/s | 596.77 MiB/s |
Manual benchmarks using randomized binary files. Compared against classic GNU base64, and race64 (C).
$ time (cat random-1gb.bin | base64 | pv -a > /dev/null)
[ 164MiB/s]
( cat random-1gb.bin | base64 | pv -a > /dev/null; ) 7.89s user 0.96s system 106% cpu 8.317 total
$ time (cat random-1gb.b64 | base64 --decode | pv -a > /dev/null)
[ 105MiB/s]
( cat random-1gb.b64 | base64 --decode | pv -a > /dev/null; ) 9.16s user 1.01s system 104% cpu 9.699 total
$ time (cat random-1gb.bin | ./race64 | pv -a > /dev/null)
[ 898MiB/s]
( cat random-1gb.bin | ./race64 | pv -a > /dev/null; ) 0.88s user 1.14s system 128% cpu 1.566 total
$ time (cat random-1gb.b64 | ./race64 -d | pv -a > /dev/null)
[ 723MiB/s]
( cat random-1gb.b64 | ./race64 -d | pv -a > /dev/null; ) 0.87s user 0.95s system 127% cpu 1.426 total
$ time (cat random-1gb.bin | rbase64 | pv -a > /dev/null)
[1.71GiB/s]
( cat random-1gb.bin | rbase64 | pv -a > /dev/null; ) 0.56s user 0.40s system 121% cpu 0.788 total
$ time (cat random-1gb.b64 | rbase64 --decode | pv -a > /dev/null)
[1.16GiB/s]
( cat random-1gb.b64 | rbase64 --decode | pv -a > /dev/null; ) 0.59s user 0.46s system 119% cpu 0.871 total
$ time (cat random-10gb.bin | base64 | pv -a > /dev/null)
[ 154MiB/s]
( cat random-10gb.bin | base64 | pv -a > /dev/null; ) 78.74s user 10.96s system 101% cpu 1:28.34 total
$ time (cat random-10gb.b64 | base64 --decode | pv -a > /dev/null)
[ 107MiB/s]
( cat random-10gb.b64 | base64 --decode | pv -a > /dev/null; ) 91.02s user 10.00s system 105% cpu 1:35.39 total
$ time (cat random-10gb.bin | ./race64 | pv -a > /dev/null)
[ 818MiB/s]
( cat random-10gb.bin | ./race64 | pv -a > /dev/null; ) 8.56s user 13.17s system 128% cpu 16.917 total
$ time (cat random-10gb.b64 | ./race64 -d | pv -a > /dev/null)
[ 724MiB/s]
( cat random-10gb.b64 | ./race64 -d | pv -a > /dev/null; ) 8.42s user 9.05s system 123% cpu 14.152 total
$ time (cat random-10gb.bin | rbase64 | pv -a > /dev/null)
[1.64GiB/s]
( cat random-10gb.bin | rbase64 | pv -a > /dev/null; ) 5.30s user 4.22s system 117% cpu 8.125 total
$ time (cat random-10gb.b64 | rbase64 --decode | pv -a > /dev/null)
[1.18GiB/s]
( cat random-10gb.b64 | rbase64 --decode | pv -a > /dev/null; ) 5.58s user 4.40s system 117% cpu 8.491 total
For reference, here's the throughput of piping a file via cat
with no processing:
$ time (cat random-10gb.b64 | pv -a > /dev/null)
[2.31GiB/s]
( cat random-10gb.b64 | pv -a > /dev/null; ) 0.88s user 4.72s system 95% cpu 5.858 total
This project is dual-licensed under MIT and Apache 2.0.