JairusSW / as-json

The only JSON library you'll need for AssemblyScript. SIMD enabled

    __  _____  _____  _____       _____  _____ 
 __|  ||   __||     ||   | | ___ |  _  ||   __|
|  |  ||__   ||  |  || | | ||___||     ||__   |
|_____||_____||_____||_|___|     |__|__||_____|
v0.9.14

Installation

npm install json-as

Add the transform to your asc command (e.g. in package.json)

--transform json-as/transform

Alternatively, add it to your asconfig.json

{
  // ...
  "options": {
    "transform": ["json-as/transform"]
  }
}

If you'd like to see the code that the transform generates, run with JSON_DEBUG=true

Usage

import { JSON } from "json-as/assembly";

// @json or @serializable work here
@json
class Vec3 {
  x: f32 = 0.0;
  y: f32 = 0.0;
  z: f32 = 0.0;
}

@json
class Player {
  @alias("first name")
  firstName!: string;
  lastName!: string;
  lastActive!: i32[];
  // Drop in a code block, function, or expression that evaluates to a boolean
  @omitif("this.age < 18")
  age!: i32;
  @omitnull()
  pos!: Vec3 | null;
  isVerified!: boolean;
}

const player: Player = {
  firstName: "Emmet",
  lastName: "West",
  lastActive: [8, 27, 2022],
  age: 23,
  pos: {
    x: 3.4,
    y: 1.2,
    z: 8.3
  },
  isVerified: true
};

const stringified = JSON.stringify<Player>(player);

const parsed = JSON.parse<Player>(stringified);

Classes can even have inheritance. Here's a nasty example

@json
class Base {}

@json
class Vec1 extends Base {
  x: f32 = 1.0;
}
@json
class Vec2 extends Vec1 {
  y: f32 = 2.0;
}
@json
class Vec3 extends Vec2 {
  z: f32 = 3.0;
}

const arr: Base[] = [
  new Vec1(),
  new Vec2(),
  new Vec3()
];

const serialized = JSON.stringify(arr);
// [{"x":1.0},{"x":1.0,"y":2.0},{"y":2.0,"x":1.0,"z":3.0}]
const parsed = JSON.parse<Base[]>(serialized);

Classes can even have inheritance. Here's a nasty example

@json
class Base {}

const serialized = JSON.stringify(arr);
// [{"x":1.0},{"x":1.0,"y":2.0},{"y":2.0,"x":1.0,"z":3.0}]
const parsed = JSON.parse<Base[]>(serialized);

It also supports arbitrary-length integers with @hypercubed/as-mpz

import { MpZ } from "@hypercubed/as-mpz";

const a = MpZ.from(18448972);
const b = MpZ.from('7881297289452930');
const c = a.add(b);

const serialized = JSON.stringify(c);
// 7881297307901902
const parsed = JSON.parse<MpZ>(serialized);

It also supports nullable primitive types, something that AssemblyScript usually can't do

@json
class Vec2 {
  x: f64 | null = 1.0;
  y: f32 | null = 2.0;
}

const vec: Vec2 = {
  x: 1.0,
  y: null
}

vec.y = 2.0;

if (vec.y) {
  // do something
}

NOTE: There are a few quirks to using nullable primitives. First, you'll get a warning about usize (ignore it, its fine) and secondly, you'll get a wasm validation error if you do not have a ! operator after accessing an element. Eg. console.log(vec.y.toString()) fails, but console.log(vec.y!.toString()) works.

You can also add it to your asconfig.json

{ // ... "options": { "transform": ["json-as/transform"], "disableWarning": [226] } }

If you use this project in your codebase, consider dropping a star. I would really appreciate it!

Notes

If you want a feature, drop an issue (and again, maybe a star). I'll likely add it in less than 7 days.

Contact

Contact me at:

Email: [e-mail] GitHub: JairusSW Discord: jairussw

Performance

Run or view the benchmarks here

Below are benchmark results comparing JavaScript's built-in JSON implementation and JSON-AS

My library beats JSON (written in C++) on all counts and, I see many places where I can pull at least a 60% uplift in performance if I implement it.

Note: SIMD is in-development and only available on the v1 branch on GitHub

Serialization Benchmarks:

Value	JavaScript (ops/s)	JSON-AS (ops/s)	JSON-AS (Pages)	JSON-AS (SIMD+Pages)	Max Throughput
"hello world"	7,124,361	44,290,480 (6.2x)	73,601,235 (10.3x)	NOT IMPLEMENTED	1.91 GB/s
12345	9,611,677	66,900,642 (6.9x)	145,924,333 (15.2x)	NOT IMPLEMENTED	0.58 GB/s
1.2345	7,227,259	20,322,939 (2.8x)	NOT IMPLEMENTED	NOT IMPLEMENTED	0.16 GB/s
[[],[[]],[[],[[]]]]	5,655,429	34,453,102 (6.0x)	NOT IMPLEMENTED	NOT IMPLEMENTED	1.32 GB/s
{ x: f64, y: f64, z: f64 }	3,878,604	44,557,996 (11.5x)	113,203,242 (29.2x)	172,023,231 (44.4x)	8.61 GB/s

Deserialization Benchmarks:

Value	JavaScript (ops/s)	JSON-AS (ops/s)	Difference
"hello world"	12,210,131	24,274,496	+ 98%
"12345"	21,376,873	254,640,930	+ 1,191%
1.2345	23,193,902	221,869,840	+ 987%
[[],[[]],[[],[[]]]]	4,777,227	74,921,123	+ 1,568%
{ x: f64, y: f64, z: f64 }	10,973,723	25,214,019	+ 230%

And my PC specs:

Component	Specification
Wasmer Version	v4.3.0
CPU	AMD Ryzen 7 7800x3D @ 6.00 GHz
Memory	T-Force DDR5 6000 MHz
OS	Ubuntu WSL2

Issues

Please submit an issue to https://github.com/JairusSW/as-json/issues if you find anything wrong with this library