Skip to content

Handling binary messages

Most PartyKit examples you see assume WebSocket messages are strings, and more specifically, JSON strings:

onMessage(message: string) {
const data = JSON.parse(message);

Similarly, you’ll often see the server responding with stringified JSON:

onConnect(connection: Party.Connection) {{ type: "join", id: }));

These examples are simplified, because in reality, both incoming and outgoing messages can be either strings or generic raw binary data buffers:

onMessage(message: string | ArrayBufferLike) {
if (typeof message !== string) {;

Why ArrayBuffers?

Strings are convenient, but they are not a good format for representing non-textual data, such as images, video, WASM bundles, and others. In fact, in JavaScript programs we often use Base64 encode binary data in order to coerce it into a string. This approach is both slow and consumes a lot of memory and network bandwidth.

Even when handling textual data, strings can be inefficient. These days, most web servers compress HTTP responses using compression algorithms like gzip. However, WebSocket messages are not compressed by default. This means that if you are sending over large amounts of textual data (for example large JSON structures), you may be wasting a lot of bytes.

ArrayBuffers allow you to represent data as low-level arrays of binary data: 0s and 1s. This gives you the flexibility to send any data, and represent, encode, and compress it in any format.

Binary encoding JavaScript objects using MessagePack

When it comes to encoding your data in a binary format, there is no one-size-fits-all approach. Depending on the shape of your data, you will find that different binary encodings provide different benefits and trade-offs.

In this example, we’ll use the MessagePack format and the msgpackr library to pack (encode) and unpack (decode) our objects to a binary format.

Encoding and decoding binary data on the server

At its simplest, MessagePack is a drop-in replacement for JSON:

import { pack, unpack } from "msgpackr";
class Server implements Party.Server {
constructor(readonly room: Party.Room) {}
onConnect(connection: Party.Connection) {
const data = { type: "join", id: };
// const message = JSON.stringify(data);
const message = pack(data);;
onMessage(message: string | ArrayBufferLike) {
// if (typeof message === string) {
if (typeof message !== string) {
// const data = JSON.parse(message);
const data = unpack(message);

Decoding binary data on the client

On the client, you can also use the same library to unpack server responses:

import { unpack } from "msgpackr/unpack";
socket.addEventListener((event) => {
//const message = JSON.parse(;
const message = unpack(;

Measure before you pack

MessagePack is easy to use, but that alone doesn’t necessarily make it a good idea.

We chose the msgpackr implementation for this guide because it’s fast. However, given that JSON.parse and JSON.stringify are built natively into the JavaScript runtime, they are very fast too. This means that you may not realize the type of performance gains you want.

Additionally, if you use the msgpackr library on the client size, you are adding nearly 10 kB of code to you JavaScript bundle.

Choosing to use a binary encoding is a trade-off, but typically, generic binary encodings make sense when you are sending large data payloads. For smaller messages, they may not make sense.

Mix both JSON and MessagePack encodings

Because you are in full control of the server and client, you can make assumptions about the data encoding based on the message format.

You could for example assume, that when a message is of type string, it’s going to be JSON, and when it’s a binary blob, it’s encoded with MessagePack:

onMessage(message: string | ArrayBuffer) {
const data =
typeof message !== string
? // byte array -> msgpack
: // string -> json

And do the same on the client:

socket.addEventListener((event) => {
const data = instanceof Blob
? // byte array -> msgpack
: // string -> json

With this approach, you can use MessagePack for large messages, such as syncing over large sets of data at once, and still use JSON for smaller messages.

Validating binary data

Encoding libraries like MessagePack will usually validate that the data conforms to a format, but they are agnostic to the content of the message.

To learn about validating data against a schema, read our guide on validating client inputs guide.