Add MessagePack's encode and decode

[idranme]

Is your feature request related to a problem? Please describe.

encode or decode MessagePack

Describe the solution you'd like

import { encode, decode } from "https://deno.land/std/encoding/magpack.ts";

const object = {
  nil: null,
  integer: 1,
  float: Math.PI,
  string: "Hello, world!",
  binary: Uint8Array.from([1, 2, 3]),
  array: [10, 20, 30],
  map: { foo: "bar" },
  timestampExt: new Date(),
};

const encoded= encode(object);
const object = decode(encoded);

Describe alternatives you've considered

[lino-levan]

This would be pretty huge for socketio/socket.io-deno#7. It looks like https://deno.land/x/[email protected] does exist, but having this be part of std would be really nice. Does this make sense to go in encoding?

[kt3k]

It'd be also worth considering using rust implementation via wasm

[lino-levan]

Could be a pretty huge selling point over /x/msgpackr... I don't plan on picking this up right now, but I might get back to it in a few ~weeks-months if no one else picks this up.

[idranme]

...

[lino-levan]

@iuioiua I know recently you've been against wasm modules in std. Thoughts on this one?

[iuioiua]

I was previously against a Wasm implementation of std/encoding/varint. The number of moving parts required to operate and maintain it was disproportionally large compared to the issue it solved. It was easily solvable using only TypeScript.

On the other hand, the Wasm implementation of std/crypto makes sense as it's perhaps not easily solvable using only TypeScript and robust Rust implementations for algorithms already exist.

I'm unfamiliar with MessagePack. Either way, I'd urge any possible implementor to consider whether a Wasm implementation is worth the added complexity or can be done well enough with just TypeScript, which would be far simpler.

[lino-levan]

On the other hand, the Wasm implementation of std/crypto makes sense as it's perhaps not easily solvable using only TypeScript and robust Rust implementations for algorithms already exist.

IMO the main advantage of std/crypto using wasm is perf, with the implementation itself being secondary. MessagePack is an interesting case because it's not immediately obvious whether performance or simplicity is more valuable. MessagePack as a format is supposed to be a "more performant + smaller JSON with more types". I think we might want a typescript implementation to start and move over to WASM if we see it as limiting.

[idranme]

On the other hand, the Wasm implementation of std/crypto makes sense as it's perhaps not easily solvable using only TypeScript and robust Rust implementations for algorithms already exist.

IMO the main advantage of std/crypto using wasm is perf, with the implementation itself being secondary. MessagePack is an interesting case because it's not immediately obvious whether performance or simplicity is more valuable. MessagePack as a format is supposed to be a "more performant + smaller JSON with more types". I think we might want a typescript implementation to start and move over to WASM if we see it as limiting.

👍👍

[lino-levan]

Just finished the decode implementation and writing some tests. Will open a PR with both encode + decode later today.