@jcubic/wayne

v0.18.3

Published

3 months ago

Service Worker Routing for in browser HTTP requests

Downloads

Service Worker Routing library for in-browser HTTP requests

It's like an Express inside Service Worker.

Most of the time Service Worker is used for caching HTTP requests and making the app work when there is no internet (mostly for PWA), but in fact, you can create completely new responses to requests that never leave the browser. This library makes that easier by adding a simple API similar to Express.

Usage

Installation from npm:

npm install @jcubic/wayne

yarn add @jcubic/wayne

The standard way of installing the service worker

if ('serviceWorker' in navigator) {
    const scope = location.pathname.replace(/\/[^\/]+$/, '/');
    navigator.serviceWorker.register('sw.js', { scope, type: 'module' })
             .then(function(reg) {
                 reg.addEventListener('updatefound', function() {
                     const installingWorker = reg.installing;
                     console.log('A new service worker is being installed:',
                                 installingWorker);
                 });
                 // registration worked
                 console.log('Registration succeeded. Scope is ' + reg.scope);
             }).catch(function(error) {
                 // registration failed
                 console.log('Registration failed with ' + error);
             });
}

If you want to support browsers that don't support ES Modules in Service Worker use this instead:

if ('serviceWorker' in navigator) {
    const scope = location.pathname.replace(/\/[^\/]+$/, '/');
    navigator.serviceWorker.register('sw.js', { scope })
             .then(function(reg) {
                 reg.addEventListener('updatefound', function() {
                     const installingWorker = reg.installing;
                     console.log('A new service worker is being installed:',
                                 installingWorker);
                 });
                 // registration worked
                 console.log('Registration succeeded. Scope is ' + reg.scope);
             }).catch(function(error) {
                 // registration failed
                 console.log('Registration failed with ' + error);
             });
}

Inside the same file you can send AJAX requests with standard fetch API.

function get(url) {
    fetch(url)
      .then(res => res.text())
      .then(text => output.innerHTML = text);
}

input.addEventListener('click', () => {
    get(`./user/${user_id.value}`);
});

error.addEventListener('click', () => {
    get(`./error`);
});

Service worker - sw.js file

Importing Wayne module:

when worker created as ES Module

import { Wayne } from 'https://cdn.jsdelivr.net/npm/@jcubic/wayne';

const app = new Wayne();

When the Service Worker created as normal script

importScripts('https://cdn.jsdelivr.net/npm/@jcubic/wayne/index.umd.min.js');

const app = new wayne.Wayne();

When using bundlers like Vite:

import { Wayne } from '@jcubic/wayne';

Using the library

const users = {
  1: 'Jakub T. Jankiewicz',
  2: 'John Doe',
  3: 'Jane Doe'
};

app.get('/user/{id}', function(req, res) {
  const user = users[req.params.id];
  if (user) {
    res.json({result: user});
  } else {
    res.json({error: 'User Not Found'});
  }
});

app.get('/error', function(req, res) {
  nonExisting();
});

app.get('/redirect', function(req, res) {
  res.redirect(301, '/message');
});

app.get('/message', function(req, res) {
  res.text('Lorem Ipsum');
});

app.get('/404', function(req, res) {
    res.text('Not Found', { status: 404, statusText: 'Not Found' });
});

app.get('/external', function(req, res) {
  // lorem ipsum API
  res.redirect('https://api.buildable.dev/@62d55492951509001abc363e/live/lorem-ipsum');
});

Handle the same extension for all requests

importScripts(
  'https://cdn.jsdelivr.net/npm/@jcubic/wayne/index.umd.min.js',
  'https://cdn.jsdelivr.net/gh/jcubic/static@master/js/path.js'
);

const app = new Wayne();

app.get('*', function(req, res) {
  const url = new URL(req.url);
  const extension = path.extname(url.pathname);
  const accept = req.headers.get('Accept');
  if (extension === '.js' && accept.match(/text\/html/)) {
    res.text('// Sorry no source code for you');
  } else {
    res.fetch(req);
  }
});

This code will show the comment // Sorry no source code for you for every request to JavaScript files from the browser (if open in a new tab). When you want to view the file the browser sends Accept: text/html HTTP header.

File system middleware

import { Wayne, FileSystem } from 'https://cdn.jsdelivr.net/npm/@jcubic/wayne';
import FS from "https://cdn.skypack.dev/@isomorphic-git/lightning-fs";
import mime from "https://cdn.skypack.dev/mime";
import path from "https://cdn.skypack.dev/path-browserify";

const { promises: fs } = new FS("__wayne__");

const app = new Wayne();

app.use(FileSystem({ path, fs, mime, prefix: '__fs__' }));

When not using a module the code will be similar. When you access URLs with the prefix __fs__ like ./__fs__/foo it will read files from the indexedDB file system named __wayne__. See Lightning-FS repo for details about the library.

From version 0.12 you can use test callback option to check if the file should serve from the filesystem. Note that it will receive URLs from all domains.

From version 0.13.0 you can use dir callback function that allow to dynamically change directory of served files.

const test = url => {
    const path = url.pathname;
    // return true if pathname should go to filesystem
    return path.match(/__fs__/);
};

const dir = () => '/';

app.use(wayne.FileSystem({ path, fs, mime, test, dir }));

From version 0.14.0 both functions dir and test can be async. So you can use data from IndexedDB e.g. using idb-keyval by Jake Archibald.

A patch in 0.14.3 allow putting interceptors to inject something into output HTML from FileSystem middleware. You do this by adding middleware before FileSystem and patch res.send method:

function fs_interecept(callback) {
    return function(req, res, next) {
        const send = res.send.bind(res);
        res.send = function(data, ...rest) {
            const url = new URL(req.url);
            if (test(url)) {
                data = callback(data);
            }
            return send(data, ...rest);
        };
        next();
    };
}

app.use(fs_interecept(function(html) {
    return html.replace(/<\/body>/, `<script>console.log('intercepted')</script></body>`);
}));

You should use the same test function to make sure that you patch only those requests that came from FS.

RPC mechanism

In Service Worker, you create a generic route that sends data to the BroadcastChannel:

import { send } from 'https://cdn.jsdelivr.net/npm/@jcubic/wayne';

const channel = new BroadcastChannel('__rpc__');

app.get('/rpc/{name}/*', async (req, res) => {
    const args = req.params[0].split('/');
    const method = req.params.name;
    try {
        const data = await send(channel, method, args);
        res.json(data);
    } catch(e) {
        res.text(e.message);
    }
});

and in the main thread, you create the other side of the channel and the remote methods:

import { rpc } from 'https://cdn.jsdelivr.net/npm/@jcubic/wayne';

const channel = new BroadcastChannel('__rpc__');

rpc(channel, {
    ping: function() {
        return 'pong';
    },
    sin: function(x) {
        return Math.sin(x);
    },
    random: function() {
        return Math.random();
    },
    json: function() {
        return fetch('https://api.npoint.io/8c7cc24b3fd405b775ce').then(res => res.json());
    }
});

When you send a request /rpc/ping you will get a response from methods.ping function.

fetch('./rpc/ping')
  .then(res => res.text())
  .then(text => {
     console.log({ text });
  });

With this setup, you can create new functions/methods that will map to HTTP requests.

The demo below uses random requests:

let index = 0;
const requests = [
    './rpc/ping/',
    './rpc/json/',
    './rpc/random/',
    './rpc/sin/10'
];

rpc.addEventListener('click', () => {
    get(random_request() );
});

function random_request() {
    const next_index = index++ % requests.length;
    return requests[next_index];
}

[!WARNING] Above code will only work when main page is open, so you can't navigate away. It may freeze the browser, becasue it will not able to find other side of the BroadcastChannel. If you want to send data persistently, it's better to use IndexedDB to share data between main thread and Service Worker. You can use library like idb-keyval by Jake Archibald for this.

Server-Sent Events

Server-Sent Events is the way to stream data in the browser. It's a native browser implementation of Long Polling. Here is an example of how to use SSE with Wayne:

Service Worker

app.get('/sse', function(req, res) {
  const stream = res.sse({
    onClose() {
      clearInterval(timerId);
    }
  });
  var timerId = setInterval(function() {
    const now = (new Date()).toString();
    stream.send({ data: now });
  }, 1000);
});

Main tread

let see_source;

sse_start.addEventListener('click', () => {
    see_source = new EventSource("./sse");
    see_source.onmessage = event => {
        console.log(event.data);
    };
});

sse_stop.addEventListener('click', () => {
    if (see_source) {
        see_source.close();
        see_source = null;
    }
});

3rd party URL

Service Worker allows intercepting everything that origineted from the page that has service worker inclding URLs from different origin. From version 0.15.0 Wayne allow to inrecept such URLs. You just use full URL instead of just path as a route:

app.get(`https://github.com/{user}/{repo}`, (req, res) => {
    res.text(`Sorry, you can't fetch ${req.params.user} repo named ${req.params.repo}`);
});

If you run fetch in browser:

await fetch('https://github.com/jcubic/wayne').then(res => res.text());

you will get the string:

"Sorry, you can't fetch jcubic repo named wayne"

If you want to restrict the request to only same origin you can do this with filter option:

const app = new Wayne({
    filter: req => {
        const url = new URL(req.url);
        return url.host === self.location.host;
    }
});

You can also use middleware for this

app.use((req, res, next) => {
    const url = new URL(req.url);
    if (url.origin === location.origin) {
        next();
    } else {
        res.fetch(req);
    }
});

With middleware you can also intercept all requests to remote servers and block those that are not on the list:

function is_chrome_extension(url) {
  return url.match(/chrome-extension:/);
}

function is_valid_request(url) {
  if (is_chrome_extension(url)) {
    return true;
  }
  const valid_hosts = ['localhost', 'cdn.jsdelivr.net'];
  const host = new URL(url).host;

  return valid_hosts.includes(host);
}

app.use((req, res, next) => {
    const url = new URL(req.url);
    if (url.origin === location.origin) {
        next();
    } else {
        if (is_valid_request(req.url)) {
          res.fetch(req);
        } else {
          res.html(`<!DOCTYPE HTML><html><body><h1>This request is blocked</h1></body></html>`, { status: 403 });
        }
    }
});

If you want to set the valid URLS in your main script, you can save them in indexedDB. As with filesystem, you can use idb-keyval by Jake Archibald, for this.

Overwrite HTML code

app.get('*', async (req, res) => {
  const _res = await fetch(req);
  const type = _res.headers.get('Content-Type');
  let response;
  if (type.match(/text\/html/i)) {
      const html = await _res.text();
      response = html.replace(/<\/body>/, '<p>Patched by Wayne<p></body>');
  } else {
      response = await _res.arrayBuffer();
  }
  res.send(response, { type });
});

This code will intercept every page and if it's HTML it will add HTML code at the end (before closing body tag).

Using with ES Modules

You can intercept the import of ES Module with Wayne. Here is example:

Main tread

<script>
window.ready = navigator.serviceWorker.register('./sw.js', { scope: location.pathname })
</script>
<script type="module">
// wait for Service Woker
await window.ready;
// next tick delay is require for the worker to intitialize properly
await new Promise(resolve => setTimeout(resolve, 0));

// static imports works only when you install and refresh the browser
// they probbaly run just after the code is parsed
const { default: $ } = await import('./@jquery');

$('body').css('background', 'rebeccapurple');
</script>

Service Worker

app.get('*', (req, res) => {
  const url = new URL(req.url);
  const name = req.url.replace(/.*@/, '');
  if (url.pathname.match(/\+esm/)) {
    res.fetch(`https://cdn.jsdelivr.net${url.pathname}`);
  } else if (url.pathname.match(/@/)) {
     if (name.match(/css/)) {
        res.fetch(`https://cdn.jsdelivr.net/npm/${name}`);
     } else {
        res.fetch(`https://esm.run/${name}`);
     }
  } else {
     res.fetch(req);
  }
});

The code checks if the URL contain @ in the path and redirect them to https://esm.run. If the script import other scrips they usually look like this:

import require$$0 from"/npm/[email protected]/+esm"

And needs to be imported from jsDelivr, the same if you import CSS file. See example of loading jQuery Terminal where this code is used.

PWA

To use with PWA and cache you need to use a custom middleware:

const app = new Wayne();

app.use(async (req, res, next) => {
    const url = new URL(req.url);

    const cache = await get_cache();
    const cache_response = await cache.match(req);

    if (cache_response) {
        if (navigator.onLine) {
            const net_response = await fetch(req);
            cache.put(req, net_response.clone());
            res.respond(net_response);
        } else {
            res.respond(cache_response);
        }
    } else {
       next();
    }
});

const cache_url = [
  '/'
  /* add any files you want to cache here */
];

self.addEventListener('install', (event) => {
   event.waitUntil(cache_all());
});

function get_cache() {
    return caches.open('pwa-assets');
}

async function cache_all() {
    const cache = await get_cache();
    return cache.addAll(cache_url);
}

This approach is recommended by the answer to this StackOverflow question:

Service-worker force update of new assets

It always fetch a new value of the assets when there is internet connection and serve cached value when user is offline. When there are no cached value it do default action (which can be normal fetch outside of cache or Wayne route).

This is just example of using cache with Wayne middleware, you can use different caching approach.

First load

According to the spec, the default behavior of the Service Worker is to control the HTTP requests after reloading the page. To make the SW always in control use this code in your SW:

// take control of uncontrolled clients on first load
// ref: https://web.dev/service-worker-lifecycle/#clientsclaim
self.addEventListener('activate', (event) => {
  event.waitUntil(clients.claim());
});

You can read more in the article The service worker lifecycle by Jake Archibald.

Demo

All in one demo.
Proof of Concept of ReactJS application inside Service Worker.
Filesystem demo.
Server-Sent Events Proxy demo.
Offline demo.
Download demo.
Source Code Syntax highlight demo.
Using with React and Vite
PWA/Cache

The source code for the demos can be found in the docs' directory at the gh-pages branch.

API reference

Wayne constrcutor accept object with options:

filter - a function that is called with request object, and should return false if the request should not be proxied with Service Worker.

Wayne object has those methods that correspond to HTTP methods

get
post
put
delete
patch

Each method accepts a URL with markers inside curly brackets, those markers will be available from Request.params object. The request object is the browser native object of a given request see MDN for details. The only change to the native API is that the object has property params.

Here are a few most important Request properties:

headers - Headers object to get key/value pairs use Object.fromEntires(req.headers.entries()).
method - request method as a string.
url - string with full URL.
referrer - HTTP referer.
arrayBuffer() - Returns a promise that resolves with an ArrayBuffer representation of the request body.
blob() - Returns a promise that resolves with a Blob representation of the request body.
formData() - Returns a promise that resolves with a FormData representation of the request body.
json() - Returns a promise that resolves with the result of parsing the request body as JSON.
text() - Returns a promise that resolves with a text representation of the request body.

Response object is an instance of HTTPResponse those have methods:

html()
json()
text()
send()

each of those methods accepts string as the first argument. The second argument is options:

headers - any headers as key-value pairs or you can pass Headers object.
statusText - The status message associated with the status code, e.g., OK.
status - The status code for the response, e.g., 200.
type - Content-Type of the response (MIME).

Additional methods:

redirect() - accept URL or optional first argument that is the number of HTTP code
respond(res) - accept Response object in case you want to use a different Response object.
sse([options]) - function creates Server-Sent Event stream, the return object has a method send that sends a new event.
fetch(url | Request) - method will send a normal HTTP request to the server and return the result to the client. You can use the default Request object from the route.
download(data, { filename }) - a method that can be used to trigger file download. The data can be a string or arrayBuffer you can use native fetch API and call await res.text() or await res.arrayBuffer() and pass the result as data.

The application also has middleware as in Express.js

use(function(err, req, res, next) {}) 4 parameters it's an error handler
use(function(req, res, next) {}) 3 parameters it's a middleware

Additional exported functions:

FileSystem({ path: string, fs: <FS Module>, prefix: string }) - a function that creates a middleware for the file system. You should use FS that supports Service Worker like the one that uses IndexedDB e.g. BrowserFS or LightingFS.
rpc(channel, object) - a function that should be used in the main thread that creates an RPC-like mechanism. The first argument is an instance of a broadcast channel and the second is an object with remote functions.
send(channel, method: string, args: any[]) - function sends remote procedure to the main thread.

Story

The idea of using a Service worker to serve pure in-browser HTTP requests has a long history. I first used this technique for my Git Web Terminal and described the usage of it in the article from 2018: How to create Web Server in Browser. In June 2022, I came up with a cool new way of using this technique. While creating PoC for the article I'm going to write (will update this story when ready), I realized that I can extract all the logic of creating those fake HTTP requests into a library. This is how Wayne was born.

The name of the library was inspired by the scene in Wayne's World 2 in which Wayne dresses up as a construction worker.

I highly recommend both movies if you haven't seen them already.

Contribution

If you have any ideas for an improvement, don't hesitate to create an issue. Code contributions are also welcome.

Working on your first Pull Request? You can learn how from this free series How to Contribute to an Open Source Project on GitHub

Published

Vulnerabilities

Links

Maintainers

Keywords

Readme