Zod Sockets

Socket.IO solution with I/O validation and the ability to generate AsyncAPI specification and a contract for consumers.

How it works

Demo Chat

Technologies

• Typescript first.
• Sockets — Socket.IO, using WebSocket for transport.
• Schema validation — Zod 3.x.
• Generating documentation according to AsyncAPI 3.0 specification.
• Generating client side types — inspired by zod-to-ts.
• Supports any logger having info(), debug(), error() and warn() methods.

Concept

The library distinguishes between incoming and outgoing events. The first are called Actions, and the second — Emission. Emission is configured first, representing the schemas for validating the outgoing data, as well as optionally received acknowledgements. Based on this configuration, an Actions Factory is created, where Actions are produced that have schemas for checking the incoming data and an optionally sent acknowledgement, and a handler. This handler is aware of the Emission types and is equipped with the emission and broadcasting methods, while its returns become an acknowledgement for the Action. This configuration is used to validate the input and output data against the specified schemas, it can be exported to frontend side, thus ensuring that the established contract is followed.

Quick start

Installation

Install the package and its peer dependencies.

yarn add zod-sockets zod socket.io typescript

Set up config

import { createSimpleConfig } from "zod-sockets";

const config = createSimpleConfig(); // shorthand for root namespace only

Create a factory

import { ActionsFactory } from "zod-sockets";

const actionsFactory = new ActionsFactory(config);

Create an action

import { z } from "zod";

const onPing = actionsFactory.build({
  event: "ping",
  input: z.tuple([]).rest(z.unknown()),
  output: z.tuple([z.literal("pong")]).rest(z.unknown()),
  handler: async ({ input }) => ["pong", ...input] as const,
});

Create a server

import http from "node:http";
import { Server } from "socket.io";
import { attachSockets } from "zod-sockets";

attachSockets({
  /** @see https://socket.io/docs/v4/server-options/ */
  io: new Server(),
  config: config,
  actions: [onPing],
  target: http.createServer().listen(8090),
});

Try it

Start the application and execute the following command:

curl "http://localhost:8090/socket.io/?EIO=4&transport=polling"

The expected response should be similar to:

{
  "sid": "***",
  "upgrades": ["websocket"],
  "pingInterval": 25000,
  "pingTimeout": 20000,
  "maxPayload": 1000000
}

Then consider using Postman for sending the ping event to ws://localhost:8090 with acknowledgement.

Basic features

Emission

The outgoing events should be configured using z.tuple() schemas. Those tuples describe the types of the arguments supplied to the Socket::emit() method, excluding an optional acknowledgment, which has its own optional schema being a callback function having ordered arguments. The schemas may also have transformations. This declaration establishes the constraints on your further implementation as well as payload validation and helps to avoid mistakes during the development. Consider the following examples of two outgoing events, with and without acknowledgment:

import { z } from "zod";
import { createSimpleConfig } from "zod-sockets";

const config = createSimpleConfig({
  emission: {
    // enabling Socket::emit("chat", "message", { from: "someone" })
    chat: {
      schema: z.tuple([z.string(), z.object({ from: z.string() })]),
    },
    // enabling Socket::emit("secret", "message", ([readAt]: [Date]) => {})
    secret: {
      schema: z.tuple([z.string()]),
      ack: z.tuple([
        z
          .string()
          .datetime()
          .transform((str) => new Date(str)),
      ]),
    },
  },
});

Server compatibility

With HTTP(S)

You can attach the sockets server to any HTTP or HTTPS server created using native Node.js methods.

import { createServer } from "node:http"; // or node:https
import { attachSockets } from "zod-sockets";

attachSockets({ target: createServer().listen(port) });

With Express

For using with Express.js, supply the app as an argument of the createServer() (avoid app.listen()).

import express from "express";
import { createServer } from "node:http"; // or node:https
import { attachSockets } from "zod-sockets";

const app = express();
attachSockets({ target: createServer(app).listen(port) });

With Express Zod API

For using with express-zod-api, take the httpServer or httpsServer returned by the createServer() method and assign it to the target property.

import { createServer } from "express-zod-api";
import { attachSockets } from "zod-sockets";

const { httpServer, httpsServer } = await createServer();
attachSockets({ target: httpsServer || httpServer });

Logger compatibility

Customizing logger

The library supports any logger having info(), debug(), error() and warn() methods. For example, pino logger with pino-pretty extension:

import pino, { Logger } from "pino";
import { attachSockets } from "zod-sockets";

const logger = pino({
  transport: {
    target: "pino-pretty",
    options: { colorize: true },
  },
});
attachSockets({ logger });

// Setting the type of logger used
declare module "zod-sockets" {
  interface LoggerOverrides extends Logger {}
}

With Express Zod API

If you're using express-zod-api, you can reuse the same logger from the returns of the createServer() method.

import { createServer } from "express-zod-api";
import { attachSockets } from "zod-sockets";
import type { Logger } from "winston";

const { logger } = await createServer();
attachSockets({ logger });

// Setting the type of logger used
declare module "zod-sockets" {
  interface LoggerOverrides extends Logger {}
}

Receiving events

Making actions

Actions (the declarations of incoming events) are produced on an ActionFactory which is an entity made aware of the Emission types (possible outgoing events) by supplying the config to its constructor. This architecture enables you to keep the produced Actions in separate self-descriptive files.

import { ActionsFactory } from "zod-sockets";

const actionsFactory = new ActionsFactory(config);

Produce actions using the build() method accepting an object having the incoming event name, the input schema for its payload (excluding acknowledgment) and a handler, which is a function where you place your implementation for handling the event. The argument of the handler in an object having several handy entities, the most important of them is input property, being the validated event payload:

const onChat = actionsFactory.build({
  // ns: "/", // optional, root namespace is default
  event: "chat",
  input: z.tuple([z.string()]),
  handler: async ({ input: [message], client, all, withRooms, logger }) => {
    /* your implementation here */
    // typeof message === "string"
  },
});

Acknowledgements

Actions may also have acknowledgements that are basically direct and immediate responses to the one that sent the incoming event. Acknowledgement is acquired from the returns of the handler and being validated against additionally specified output schema. When the number of payload arguments is flexible, you can use rest() method of z.tuple(). When the data type is not important at all, consider describing it using z.unknown(). When using z.literal(), Typescript may assume the type of the actually returned value more loose, therefore the as const expression might be required. The following example illustrates an action acknowledging "ping" event with "pong" and an echo of the received payload:

const onPing = actionsFactory.build({
  event: "ping",
  input: z.tuple([]).rest(z.unknown()),
  output: z.tuple([z.literal("pong")]).rest(z.unknown()),
  handler: async ({ input }) => ["pong" as const, ...input],
});

Dispatching events

In Action context

The Emission awareness of the ActionsFactory enables you to emit and broadcast other events due to receiving the incoming event. Depending on your application's needs and architecture, you can choose different ways to send events. The emission methods have constraints on emission types declared in the configuration. The input is available for processing the validated payload of the Action.

actionsFactory.build({
  handler: async ({ input, client, withRooms, all }) => {
    // sending to the sender of the received event:
    await client.emit("event", ...payload);
    // sending to everyone except the client:
    await client.broadcast("event", ...payload);
    // sending to everyone except the client in a room:
    withRooms("room1").broadcast("event", ...payload);
    // sending to everyone except the client within several rooms:
    withRooms(["room1", "room2"]).broadcast("event", ...payload);
    // sending to everyone everywhere including the client:
    all.broadcast("event", ...payload);
  },
});

In Client context

The previous example illustrated the events dispatching due to or in a context of an incoming event. But you can also emit events regardless the incoming ones by setting the onConnection property within hooks of the config, which has a similar interface except input and fires for every connected client:

import { createSimpleConfig } from "zod-sockets";

const config = createSimpleConfig({
  // emission: { ... },
  hooks: {
    onConnection: async ({ client, withRooms, all }) => {
      /* your implementation here */
    },
  },
});

Independent context

Moreover, you can emit events regardless the client activity at all by setting the onStartup property within hooks of the config. The implementation may have a setInterval() for recurring emission.

import { createSimpleConfig } from "zod-sockets";

const config = createSimpleConfig({
  hooks: {
    onStartup: async ({ all, withRooms }) => {
      // sending to everyone in a room:
      withRooms("room1").broadcast("event", ...payload);
      // sending to everyone within several rooms:
      withRooms(["room1", "room2"]).broadcast("event", ...payload);
      // sending to everyone everywhere:
      all.broadcast("event", ...payload);
      // sending to some particular user by familiar id:
      (await all.getClients())
        .find(({ id }) => id === "someId")
        ?.emit("event", ...payload);
    },
  },
});

Rooms

Available rooms

Rooms are the server side concept. Initially, each newly connected Client is located within a room having the same identifier as the Client itself. The list of available rooms is accessible via the getRooms() method of the all handler's argument.

const handler = async ({ all, logger }) => {
  logger.debug("All rooms", all.getRooms());
};

Distribution

The client argument of a handler (of a Client or an Action context) provides methods join() and leave() in order to distribute the clients to rooms. Those methods accept a single or multiple room identifiers and may be async depending on adapter, therefore consider calling them with await anyway.

const handler = async ({ client }) => {
  await client.leave(["room2", "room3"]);
  await client.join("room1");
};

Who is where

Regardless the context, each handler has withRooms() argument accepting a single or multiple rooms identifiers. The method returns an object providing the getClients() async method, returning an array of clients within those rooms. Those clients are also equipped with distribution methods join() and leave().

const handler = async ({ withRooms }) => {
  await withRooms("room1").getClients();
  await withRooms(["room1", "room2"]).getClients();
};

Alternatively, you can request getClients() method of the all argument, which returns an array of all familiar clients having rooms property, being an array of the room identifiers that client is located.

const handler = async ({ all, logger }) => {
  const clients = await all.getClients();
  for (const client of clients) {
    logger.debug(`${client.id} is within`, client.rooms);
  }
};

Subscriptions

In order to implement a subscription service you can utilize the rooms feature and make two Actions: for subscribing and unsubscribing. Handlers of those Actions can simply do client.join() and client.leave() in order to address the client to/from a certain room. A simple setInterval() function within an Independent Context (onStartup hook) can broadcast to those who are in that room. Here is a simplified example:

import { createServer } from "express-zod-api";
import { attachSockets, createSimpleConfig, ActionsFactory } from "zod-sockets";
import { Server } from "socket.io";
import { z } from "zod";

const { logger, httpsServer, httpServer } = await createServer();

const config = createSimpleConfig({
  emission: {
    time: { schema: z.tuple([z.date()]) }, // constraints
  },
  hooks: {
    onStartup: async ({ withRooms }) => {
      setInterval(() => {
        withRooms("subscribers").broadcast("time", new Date());
      }, 1000);
    },
  },
});

const factory = new ActionsFactory(config);
await attachSockets({
  config,
  logger,
  io: new Server(),
  target: httpsServer || httpServer,
  actions: [
    factory.build({
      event: "subscribe",
      input: z.tuple([]),
      handler: async ({ client }) => client.join("subscribers"),
    }),
    factory.build({
      event: "unsubscribe",
      input: z.tuple([]),
      handler: async ({ client }) => client.leave("subscribers"),
    }),
  ],
});

Metadata

Metadata is a custom object-based structure for reading and storing additional information about the clients. Initially it is an empty object.

Defining constraints

You can specify the schema of the metadata in config. Please avoid transformations in those schemas since they are not going to be applied.

import { z } from "zod";
import { createSimpleConfig } from "zod-sockets";

const config = createSimpleConfig({
  metadata: z.object({
    /** @desc Number of messages sent to the chat */
    msgCount: z.number().int(),
  }),
});

Reading

In every context you can read the client's metadata using the getData() method. Since the presence of the data is not guaranteed, the method returns an Partial<> object of the specified schema.

const handler = async ({ client, withRooms }) => {
  client.getData();
  withRooms("room1")
    .getClients()
    .map((someone) => someone.getData());
};

Writing

Within a client context you can use setData() method to store the metadata on the client. The method provides type assistance of its argument and may throw ZodError if it does not pass the validation against the specified schema.

const handler = async ({ client }) => {
  client.setData({ msgCount: 4 });
};

Namespaces

Namespaces allow you to separate incoming and outgoing events into groups, in which events can have the same name, but different essence, payload and handlers. For using namespaces replace the createSimpleConfig() method with new Config(), then use its .addNamespace() method for each namespace. Namespaces may have emission, examples, hooks and metadata. Read the Socket.IO documentation on namespaces.

import { Config } from "zod-sockets";

const config = new Config()
  .addNamespace({
    // The namespace "/public"
    emission: { chat: { schema } },
    examples: {}, // see Generating documentation section
    hooks: {
      onStartup: () => {},
      onConnection: () => {},
      onDisconnect: () => {},
      onAnyIncoming: () => {},
      onAnyOutgoing: () => {},
    },
    metadata: z.object({ msgCount: z.number().int() }),
  })
  .addNamespace({
    path: "private", // The namespace "/private" has no emission
  });

Integration

Exporting types for frontend

In order to establish constraints for events on the client side you can generate their Typescript definitions.

import { Integration } from "zod-sockets";

const integration = new Integration({ config, actions });
const typescriptCode = integration.print(); // write this to a file

Check out the generated example.

You can adjust the naming of the produced functional arguments by applying the .describe() method to the schemas.

There is also a special handling for the cases when event has both .rest() on the payload schema and an acknowledgement, resulting in producing overloads, because acknowledgement has to go after ...rest which is prohibited. You can adjust the number of the those overloads by using the maxOverloads option of the Integration constructor. The default is 3.

Then on the frontend side you can create a strictly typed Socket.IO client.

import { io } from "socket.io-client";
import { Root } from "./generated/backend-types.ts"; // the generated file

const socket: Root.Socket = io(Root.path);

Alternatively, you can avoid installing and importing socket.io-client module by making a standalone build having serveClient option configured on the server.

Generating documentation

You can generate the AsyncAPI specification of your API and write it into a file, that can be used as the documentation:

import { Documentation } from "zod-sockets";

const yamlString = new Documentation({
  config,
  actions,
  version: "1.2.3",
  title: "Example APP",
  servers: { example: { url: "https://example.com/socket.io" } },
}).getSpecAsYaml();

See the example of the generated documentation on GitHub or open in Studio.

Adding examples to the documentation

You can add Action examples using its .example() method, and Emission examples you can describe in the examples property of namespace config.

import { createSimpleConfig, ActionsFactory } from "zod-sockets";

// Examples for outgoing events (emission)
const config = createSimpleConfig({
  emission: {
    event1: { schema },
    event2: { schema, ack },
  },
  examples: {
    event1: { schema: ["example payload"] }, // single example
    event2: [
      // multiple examples
      { schema: ["example payload"], ack: ["example acknowledgement"] },
      { schema: ["example payload"], ack: ["example acknowledgement"] },
    ],
  },
});

// Examples for incoming event (action)
const factory = new ActionsFactory(config);
const action = factory
  .build({
    input: payloadSchema,
    output: ackSchema,
  })
  .example("input", ["example payload"])
  .example("output", ["example acknowledgement"]);

Adding security schemas to the documentation

You can describe the security schemas for the generated documentation both on server and namespace levels.

// Single namespace
import { createSimpleConfig } from "zod-sockets";

const config = createSimpleConfig({
  security: [
    {
      type: "httpApiKey",
      description: "Server security schema",
      in: "header",
      name: "X-Api-Key",
    },
  ],
});

// Multiple namespaces
import { Config } from "zod-sockets";

const config = new Config({
  security: [
    {
      type: "httpApiKey",
      description: "Server security schema",
      in: "header",
      name: "X-Api-Key",
    },
  ],
}).addNamespace({
  security: [
    {
      type: "userPassword",
      description: "Namespace security schema",
    },
  ],
});