A terse, human-readable, URL-safe encoding for JSONish strings.

Overview

Use this when you need to encode a string to make it URL-safe, but you also want to keep it as small and readable as possible (unlike base64). For example:

| Input string | base64 | utf64 | | ----------------- | ------------------------ | ------------------ | | Hello | SGVsbG8= | YHello | | "Hello!" | IkhlbGxvISI= | AYHelloGA | | {"Hello":"world"} | eyJIZWxsbyI6IndvcmxkIn0= | MAYHelloAFAworldAN |

I made this because I wanted to build a web API with a nice JSON schema that could also be cached by a CDN. To make it cacheable, I had to use the GET method; but GET can't (portably) have a request body, so this means all the API parameters need to be packed into the URL. UTF-64 is a fire-and-forget way to solve this problem.

UTF-64 uses the very permissive 0BSD licence so you can freely use this code & spec anywhere. I picked 0BSD as it seems to be the "public domain equivalent" most widely accepted by corporations, e.g. Google has a specific exception permitting the use of 0BSD.

Installation & usage

JavaScript

npm install utf64

import * as utf64 from "utf64";

console.log(utf64.encode("Hello!"));
console.log(utf64.decode("YHelloG"));

Python

pip install utf64

import utf64

print(utf64.encode("Hello!"))
print(utf64.decode("YHelloG"))

Go

go get utf64.moreplease.com

package main

import (
	"fmt"
	"utf64.moreplease.com"
)

func main() {
	fmt.Println(utf64.Encode("Hello!"))
	result, err := utf64.Decode("YHelloG")
	if err != nil {
		panic(err)
	}
	fmt.Println(result)
}

Rust

Kindly contributed by Mark Musante (@mjmusante)

cargo add utf64

use utf64::*;

fn main() {
    println!("{}", "Hello!".encode_utf64().unwrap());
    match "YHelloG".decode_utf64() {
        Ok(result) => println!("{result}"),
        Err(e) => panic!("{e}"),
    }
}

Command-line tool

The JS package includes a utf64 command-line tool:

npm install -g utf64

utf64 "Hello\!"
utf64 -d YHelloG

Specification

Output is encoded using base64url-compatible characters: _ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-

| utf64 | Decoded as | | ----------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | _ | As-is | | a to z | As-is | | 0 to 9 | As-is | | ABCDEFGHIJKLMNOPQRSTU | Mapped to: "',.;:!?()[]{}#=+-*/\ | | V | Newline | | W | Space | | X | Prefix for Unicode 0-63. For example, "Xk" is "%" (U+0025) | | Y | Prefix for Unicode 64-127. For example, "Y_" is "@" (U+0040) | | Z | Prefix for Unicode 128+. The following characters are interpreted as UTF-8, reduced to 6-bit bytes by stripping the redundant top two bits. For example, "ZhBr" is "€" (UTF-8 [11]100010 [10]000010 [10]101100) |

See test.json for tests that (hopefully) cover all the edge cases, for both valid and invalid encodings.