len0xx/sequoia

Middleware web framework for Deno written in TypeScript

Sequoia 🦕

A library for handling HTTP requests by using middlewares. Written in TypeScript, works with Deno runtime. Inspired by Oak

Check out the website

Русская версия доступна тут: /README-ru.md

Why Sequoia?

Sequoia was developed to enhance the Developer Experience for those who create APIs with Node.js/Deno. Node.js has a lot of disadvantages compared to Deno. That's why Sequoia was developed first for Deno. Probably it will be adapted to work with Node.js in the future. Read more about Deno on the website

You may consider Sequoia as an advanced and simplified version of Express or Oak. Because for now the libraries have a lot of similarities, but it's a subject to change in the future

Moving to the examples, if you've ever developed an API in Node.js or Deno, this code might look familiar to you:

// express.js
app.get('/', (req, res) => {
    const body = { ok: true, user: { id: 1, name: 'John Doe' } }
    res.status(201).json(body).cookie('access_token', 'a62bc1')
})

A chain of methods like this res.status(201).json(body).cookie('access_token', 'a62bc1') is not a great example of Developer Experience

Sequoia on the other hand provides you a better way to do this:

// Sequoia
app.GET('/', (ctx) => {
    const body = { ok: true, user: { id: 1, name: 'John Doe' } }
    ctx.cookies.set('access_token', 'a62bc1')

    return new HTTPResponse({
        status: HTTPStatus.CREATED,
        type: 'application/json',
        body
    })
})

In Sequoia there is an HTTPResponse class which stands in-between the developer and the actual Response object that is being sent to the user. While using this class you are getting the great intellisense in your IDE and creating APIs become a much more pleasant experience

Usage

Prerequisites

To use this library you obviously need Deno (v1.35.0 or later) to be install on your machine. Check out the installation guide

Getting started

Using Sequoia might seem familiar to those who had some experience with Oak, Koa, Express or any other middleware library.

Warning: The project is still in Beta stage. So the API and the provided functionality might change after achieving v1.0.0

To get started you can use this example:

// example.ts
import { Application, Router, HTTPStatus, HTTPResponse } from 'jsr:@sequoia/sequoia'

const app = new Application({ logging: true })
const router = new Router({ type: 'application/json' })

router.GET('/', (ctx) => {
    const agent = ctx.request.headers.get('User-Agent')
    ctx.response.headers.set('X-Powered-By', 'Sequoia')
    ctx.cookies.set('WebServer', 'Sequoia')

    return new HTTPResponse({
        status: HTTPStatus.SUCCESS,
        body: { ok: true, agent }
    })
})

app.useRouter(router)

const HOST = Deno.env.get('HOST') ?? '0.0.0.0'
const PORT = Number(Deno.env.get('PORT') ?? 80)

app.listen(
    { hostname: HOST, port: PORT },
    () => console.log('The sequoia server is up!')
)

To run this example just use:

deno run --allow-net --allow-env example.ts

To run it on custom IP or port, simply define them as environment variables:

HOST=0.0.0.0 PORT=4000 deno run --allow-net --allow-env example.ts

After running this command the server is running, so you can go to http://localhost:8000 in your browser and there you can see the response which might look like this: { 'ok': true, 'agent': 'curl/7.85.0' }

To see more examples visit the website

Performance

One of the main goals while designing Sequoia was performance. So to measure the performance of our library we ran a simple benchmark (ab -n 1000000 -c 100 127.0.0.1:8000), here are the results for Sequoia, native Deno server and a server on Express:

RPS - Requests per second

Sequoia benchmark (46956 RPS):

Concurrency Level:      100
Time taken for tests:   21.296 seconds
Complete requests:      1000000
Failed requests:        0
Total transferred:      150000000 bytes
HTML transferred:       11000000 bytes
Requests per second:    46956.41 [#/sec] (mean)
Time per request:       2.130 [ms] (mean)
Time per request:       0.021 [ms] (mean, across all concurrent requests)
Transfer rate:          6878.38 [Kbytes/sec] received

Deno benchmark (55536 RPS):

Concurrency Level:      100
Time taken for tests:   18.006 seconds
Complete requests:      1000000
Failed requests:        0
Total transferred:      150000000 bytes
HTML transferred:       11000000 bytes
Requests per second:    55536.75 [#/sec] (mean)
Time per request:       1.801 [ms] (mean)
Time per request:       0.018 [ms] (mean, across all concurrent requests)
Transfer rate:          8135.27 [Kbytes/sec] received

Express (on Node.js) benchmark (11293 RPS):

Concurrency Level:      100
Time taken for tests:   88.543 seconds
Complete requests:      1000000
Failed requests:        0
Total transferred:      211000000 bytes
HTML transferred:       12000000 bytes
Requests per second:    11293.98 [#/sec] (mean)
Time per request:       8.854 [ms] (mean)
Time per request:       0.089 [ms] (mean, across all concurrent requests)
Transfer rate:          2327.18 [Kbytes/sec] received

All of the benchmark were ran on a Debian linux VM with 2 CPU cores (3.5 GHz) and 2 GB or RAM. Feel free to run your own benchmarks and share the results with us

Roadmap

Right now Sequoia is only maintained by its original creator (@len0xx), although anyone on the internet is welcome to contribute. So new features of a library depend directly on the free time of its maintainer. Do not expect releases too often.

The features that are expected in the upcoming releases:

• Allow serving static files via CLI
• Describe examples in Wiki
• Support for CORS headers
• Upgrade to Deno.serve()
• Unit tests and CI/CD through GitHub Actions
• Publish the package to JSR.io
• Support for TLS
• Support for file uploading
• Support for WebSockets
• Better compatibility with svelte-adapter-deno
• Support for compression headers
• Better documentation
• Implement an easy-to-use and highly configurable class for static file serving