feat(sse): add @tsed/sse package

[Romakita]

Information

Type	Breaking change
Feature	Yes

---

Server-send event let you push data to the client. It's a simple way to send data from the server to the client. The data is sent as a stream of messages, with an optional event name and id. It's a simple way to send data from the server to the client.

Installation

Before using the Socket.io, we need to install the Socket.io module.

npm install --save @tsed/sse

Then add the following configuration in your Server:

import {Configuration} from "@tsed/common";
import "@tsed/sse"; // import socketio Ts.ED module

@Configuration({
  sse: {}
})
export class Server {}

::: warning
compression middleware should be disabled for correct SSE work
:::

Enable event-stream

To enable the event-stream, you need to use the @EventStream decorator on a method of a controller.

import {Controller} from "@tsed/di";
import {Get} from "@tsed/schema";
import {EventStream, EventStreamCtx} from "@tsed/sse";

@Controller("/sse")
export class MyCtrl {
  @Get("/events")
  @EventStream()
  events(@EventStreamCtx() eventStream: EventStreamCtx) {
    let intervalId: ReturnType<typeof setInterval>;

    eventStream.on("close", () => {
      clearInterval(intervalId);
    });

    eventStream.on("end", () => {
      clearInterval(intervalId);
    });

    intervalId = setInterval(() => {
      // Ts.ED support Model serialization using json-mapper here
      eventStream.emit(new Date());
    }, 1000);
  }
}

Stream events

You can use Node.js stream like EventEmmiter to emit events from your controller to your consumer:

import {EventStream} from "@tsed/sse";
import {Controller} from "@tsed/di";
import {Get} from "@tsed/schema";

@Controller("/sse")
export class MyCtrl {
  private eventEmitter = new EventEmitter();

  $onInit() {
    setInterval(() => {
      this.eventEmitter.emit("message", new Date());
    }, 1000);
  }

  @Get("/events")
  @EventStream()
  events() {
    return this.eventEmitter;
  }
}

Observable

You can also use Observable from rxjs to emit events from your controller to your consumer:

import {EventStream} from "@tsed/sse";
import {Controller} from "@tsed/di";
import {Get} from "@tsed/schema";

@Controller("/sse")
export class MyCtrl {
  @Get("/events")
  @EventStream()
  events() {
    const observable = new Observable((observer) => {
      setInterval(() => {
        observer.next(new Date());
      }, 1000);
    });

    return observable;
  }
}

Todos

• Tests
• Coverage
• Example
• Documentation

[github-actions]

Benchmarks

• Machine: linux x64 | 4 vCPUs | 15.6GB Mem
• Node: v18.20.4
• Run: Fri Jul 26 2024 07:01:58 GMT+0000 (Coordinated Universal Time)
• Method: autocannon -c 100 -d 10 -p 10 localhost:3000 (two rounds; one to warm-up, one to measure)

	Version	Router	Requests/s	Latency	Throughput/Mb
koa	2.13.4	✗	33428.0	29.36	5.96
tsed-koa	7.77.0	✓	11368.4	87.17	9.63
express	4.19.2	✓	10310.4	96.09	1.84
nest	8.4.3	✓	9902.4	100.00	2.38
express-injector	4.19.2	✗	9614.7	103.11	2.19
tsed-express	7.77.0	✓	8521.1	116.30	1.56
express-morgan	4.19.2	✗	5596.8	176.79	1.00
fastify-big-json	3.29.4	✓	N/A	N/A	N/A
fastify-injector	3.29.4	✓	N/A	N/A	N/A
fastify	3.29.4	✓	N/A	N/A	N/A
nest-fastify	8.4.3	✓	N/A	N/A	N/A

Explanation

The benchmark shows a performance difference between the frameworks. We note that Ts.ED is often last. In fact, Ts.ED uses features useful to a production application which reduce its performance.

For example, Ts.ED initializes a sandbox (async_hook) for each request in order to work in an isolated context if necessary.
It also initializes the elements necessary for monitoring requests in a log manager.

All this at a necessary cost that reflects the reality of a production application ;)

[github-actions]

Benchmarks

• Machine: linux x64 | 4 vCPUs | 15.6GB Mem
• Node: v18.20.4
• Run: Mon Jul 29 2024 07:49:04 GMT+0000 (Coordinated Universal Time)
• Method: autocannon -c 100 -d 10 -p 10 localhost:3000 (two rounds; one to warm-up, one to measure)

	Version	Router	Requests/s	Latency	Throughput/Mb
koa	2.13.4	✗	33391.3	29.41	5.95
tsed-koa	7.78.0	✓	11834.5	83.71	10.02
express	4.19.2	✓	10064.0	98.53	1.79
nest	8.4.3	✓	9815.8	100.83	2.36
express-injector	4.19.2	✗	9440.2	105.04	2.15
tsed-express	7.78.0	✓	8828.4	112.29	1.62
express-morgan	4.19.2	✗	5563.1	177.86	0.99
fastify-big-json	3.29.4	✓	N/A	N/A	N/A
fastify-injector	3.29.4	✓	N/A	N/A	N/A
fastify	3.29.4	✓	N/A	N/A	N/A
nest-fastify	8.4.3	✓	N/A	N/A	N/A

Explanation

The benchmark shows a performance difference between the frameworks. We note that Ts.ED is often last. In fact, Ts.ED uses features useful to a production application which reduce its performance.

For example, Ts.ED initializes a sandbox (async_hook) for each request in order to work in an isolated context if necessary.
It also initializes the elements necessary for monitoring requests in a log manager.

All this at a necessary cost that reflects the reality of a production application ;)

[github-actions]

Benchmarks

• Machine: linux x64 | 4 vCPUs | 15.6GB Mem
• Node: v18.20.4
• Run: Thu Aug 01 2024 07:02:23 GMT+0000 (Coordinated Universal Time)
• Method: autocannon -c 100 -d 10 -p 10 localhost:3000 (two rounds; one to warm-up, one to measure)

	Version	Router	Requests/s	Latency	Throughput/Mb
koa	2.13.4	✗	31250.9	31.47	5.57
tsed-koa	7.78.1	✓	11323.3	87.53	9.59
express	4.19.2	✓	10276.2	96.49	1.83
nest	8.4.3	✓	9568.5	103.51	2.30
express-injector	4.19.2	✗	9042.7	109.34	2.06
tsed-express	7.78.1	✓	8171.0	121.14	1.50
express-morgan	4.19.2	✗	5199.9	190.10	0.93
fastify-big-json	3.29.4	✓	N/A	N/A	N/A
fastify-injector	3.29.4	✓	N/A	N/A	N/A
fastify	3.29.4	✓	N/A	N/A	N/A
nest-fastify	8.4.3	✓	N/A	N/A	N/A

Explanation

The benchmark shows a performance difference between the frameworks. We note that Ts.ED is often last. In fact, Ts.ED uses features useful to a production application which reduce its performance.

For example, Ts.ED initializes a sandbox (async_hook) for each request in order to work in an isolated context if necessary.
It also initializes the elements necessary for monitoring requests in a log manager.

All this at a necessary cost that reflects the reality of a production application ;)

[github-actions]

Benchmarks

• Machine: linux x64 | 4 vCPUs | 15.6GB Mem
• Node: v18.20.4
• Run: Thu Aug 01 2024 07:23:23 GMT+0000 (Coordinated Universal Time)
• Method: autocannon -c 100 -d 10 -p 10 localhost:3000 (two rounds; one to warm-up, one to measure)

	Version	Router	Requests/s	Latency	Throughput/Mb
koa	2.13.4	✗	32483.6	30.26	5.79
tsed-koa	7.78.1	✓	11844.0	83.53	10.03
express	4.19.2	✓	10122.4	97.75	1.81
nest	8.4.3	✓	9400.2	105.41	2.26
express-injector	4.19.2	✗	9377.0	105.69	2.14
tsed-express	7.78.1	✓	8428.2	117.69	1.54
express-morgan	4.19.2	✗	5568.2	177.77	0.99
fastify-big-json	3.29.4	✓	N/A	N/A	N/A
fastify-injector	3.29.4	✓	N/A	N/A	N/A
fastify	3.29.4	✓	N/A	N/A	N/A
nest-fastify	8.4.3	✓	N/A	N/A	N/A

Explanation

The benchmark shows a performance difference between the frameworks. We note that Ts.ED is often last. In fact, Ts.ED uses features useful to a production application which reduce its performance.

For example, Ts.ED initializes a sandbox (async_hook) for each request in order to work in an isolated context if necessary.
It also initializes the elements necessary for monitoring requests in a log manager.

All this at a necessary cost that reflects the reality of a production application ;)

[Romakita]

@IvaDey This PR can be review and add officially the sse support to Ts.ED ;)

[Romakita]

🎉 This PR is included in version 7.79.0 🎉

The release is available on:

• v7.79.0
• GitHub release

Your semantic-release bot 📦🚀