JavaScript
Article
By Artem Tabalin

Create Offline Web Apps Using Service Workers & PouchDB

By Artem Tabalin

Two developers using offline web apps on a train

Offline web applications are becoming more and more popular. Offline support is so important that it’s now common to talk about the ‘Offline First’ approach, where it becomes a primary consideration. It’s also gathering popularity with the rise of Progressive Web Apps.

In this post, we will look at how to add offline support to a basic contact list web app by implementing asset caching, client-side data storage and synchronization with a remote data store.

The source code of the app is available on GitHub.

Why Support Offline?

Why should we care about offline support?

I myself spend more than one hour on a train every day. I don’t want to waste this time, so I take my laptop to work a bit along the way. I use the cell network to be online. The connection is not reliable, so I lose it from time to time. My user experience is up to the web app I’m using. Only a few apps, with good offline support, behave as expected and connection loss is transparent. Some behave weirdly, so when I refresh the page, I lose data. Most don’t support offline at all, and I have to wait for a stable connection to be able to use them.

An unreliable connection is not the only use case. We can also talk about situations where you might be offline for several hours, for example, while on an airplane.

Another important advantage of offline support is the performance boost. Indeed, the browser doesn’t need to wait for assets to be loaded from the server. The same for data, once stored on the client.

Thus we need offline:

  1. to be able to use apps even with flaky connection (cell network in a train)
  2. to be able to work without network connection (on an airplane)
  3. to boost the performance

Progressive Web Apps

Google’s Progressive Web Apps (PWA) concept is a methodology aimed at delivering web apps that provide the UX of native mobile apps. PWA includes offline support, but it also covers a lot more:

  • Responsiveness – support for different form factors: mobile, tablet, desktop
  • Web App Manifest – to install an app on the home screen
  • App Shell – a design pattern in which the basic UI app shell is separated from the content loaded afterward
  • Push notifications – to get “instant” updates from the server

Addy Osmani wrote a great intro post about PWA.

In this article, we are going to focus only on a single aspect: offline support.

Defining Offline Support

Let’s clarify what it takes to support offline. We need to take care of two aspects:

  1. app assets – caching HTML, JS scripts, CSS style sheets, images
  2. app data – storing data on client-side

App assets

The first solution in HTML5 to cache offline assets was AppCache. The idea is to provide an app manifest describing which resources should be stored in the browser cache. Thus, the next time an app is loaded, these assets will be taken from the browser cache.

Important: While being simple, there are quite a lot of pitfalls with using AppCache. The standard is now deprecated, although it’s still widely supported by browsers.

Service Workers were introduced to replace AppCache. They provide a flexible solution for the offline support. Service Workers give control over outgoing requests, allowing a script intercept them and return the necessary responses. The caching logic is entirely on the developer’s shoulders. The app code itself can check if an asset is saved in the cache and requests it from the server only if needed.

It’s important to note that Service Workers are supported only via HTTPS (HTTP is allowed for localhost) connections. We will look at how to use Service Workers shortly.

App data

App data can be stored in the offline storage provided by browsers.

There are several options introduced by HTML5:

WebStorage is a key-value storage. This is the simplest cross-browser storage, but there are several pitfalls to be aware of. You have to take care of serialization and deserialization of data that you put inside because the values must be plain strings. You may run up against size limits with larger data sets. Also, it’s possible to get into a race condition, meaning if you have two tabs opened at the same time in the browser you could end up with unexpected behavior.

IndexedDB is much more powerful and seems to be the best way to go with offline storage. It has plenty of space available. It supports transactions and can be safely used in several browser tabs at the same time. It’s also supported by all modern browsers.

WebSQL is literally SQLite in the browser. Full-featured relational DB with ACID on the client. Unfortunately, WebSQL has been deprecated by the standards committee and was never supported in non-Blink/Webkit browsers.

There are several libraries out there which provide an abstraction over offline storage:

  • localForage – simple localStorage-like API
  • IDBWrapper – cross-browser IndexedDB wrapper
  • PouchDB – client-side storage solution inspired by CouchDB. It supports automatic sync with the backend if CouchDB is being used.

The ContactBook App

Now, let’s see how to add offline support to a web application. Our sample app is a basic contact book:

contact book screenshot

We have the list of contacts on the left and a details form on the right used for editing contacts. A contact has three fields: first name, last name, and phone.

You can find the app source code on GitHub. To run the app you’ll need Node.js installed. If you’re not sure about this step, you can follow our beginner’s guide to npm.

Start by downloading the sources and running the following commands from the project folder:

$ npm install
$ npm run serve

What about backend? We are using pouchdb-server to provide a REST API over CouchDB storage, and http-server to serve frontend assets.

Our scripts section of package.json looks like this:

  "scripts": {
    "serve": "npm-run-all -p serve-front serve-backend",
    "serve-front": "http-server -o",
    "serve-backend": "pouchdb-server -d db"
  },

The package npm-run-all allows running several commands in parallel. We start both servers: http-server and pouchdb-server.

Now let’s look at the implementation of offline support for the app assets.

Offline Assets

The directory /public contains all the assets for the application:

  • /css/style.css – the application style sheet
  • /js/ext – directory containing external libs (PouchDB, and Babel to use ES2015 syntax)
  • /js/app.js – the main application script
  • /js/register-service-worker.js – the script that registers the service worker
  • /js/store.js – the adapter class to work with PouchDB storage
  • /contactbook.appcache – the AppCache manifesto
  • /index.html – the application markup
  • /service-worker.js – the source of the service worker

The journey begins with the registration of a service worker. Here is the registration code in register-service-worker.js:

if ('serviceWorker' in navigator) {
    navigator.serviceWorker.register('/service-worker.js', {
        scope: '/'
    }).then(function() {
        // success
    }).catch(function(e) {
        // failed
    });
}

First, we check that serviceWorker is supported by the browser. If yes, we call the register method, providing the URL to the service worker script (in our case /service-worker.js) and extra parameters to specify the scope of the service worker. Parameters are optional and the root / is the scope default value..

Important: to be able to use the root of the app as a scope, the service worker script should be located in the root directory of the app.

The register method returns a Promise.

The life-cycle of a service worker starts with installation. We can handle the install event and put all required resources in the cache:

var CACHE_NAME = 'contact-book-v1';

var resourcesToCache = [
  '/',
  '/css/style.css',
  '/js/ext/babel.min.js',
  '/js/ext/pouchdb.min.js',
  '/js/register-service-worker.js',
  '/js/store.js',
  '/js/app.js'
];

self.addEventListener('install', function(event) {
  event.waitUntil(
    // open the app browser cache
    caches.open(CACHE_NAME)
      .then(function(cache) {
        // add all app assets to the cache
        return cache.addAll(resourcesToCache);
      })
  );
});

The last thing is to handle the fetch event triggered each time a resource from the Service Worker scope is fetched:

self.addEventListener('fetch', function(event) {
  event.respondWith(
    // try to find corresponding response in the cache
    caches.match(event.request)
      .then(function(response) {
        if (response) {
          // cache hit: return cached result
          return response;
        }

        // not found: fetch resource from the server
        return fetch(event.request);
      })
  );
});

That’s it. Let’s test that it’s working:

  1. run the app with npm run serve
  2. open the URL http://127.0.0.1:8080/ in Chrome
  3. stop the web server with Ctrl + C in the console (or use the Chrome dev tools to simulate going offline)
  4. refresh the web page

app assets offline

The app is still available. Awesome!

AppCache

The problem with the solution above is that Service Workers have limited browser support. We can implement a fallback solution using widely supported AppCache. Read more info about AppCache usage here.

The basic usage is straightforward and includes two steps:

  1. Defining the app cache manifest contactbook.appcache:

    CACHE MANIFEST
    # v1 2017-30-01
    
    CACHE:
    index.html
    css/style.css
    js/ext/babel.min.js
    js/ext/pouchdb.min.js
    js/store.js
    js/app.js
    

    For our simple app, we define a single section CACHE and put all the assets there.

  2. Reference the manifest file from the HTML:

    <html manifest="contactbook.appcache" lang="en">
    

That’s it. Let’s open the page in a browser not supporting Service Workers and test it the same way we did before.

app assets offline with appcache

Offline Data

Being able to cache assets is great. But it’s not enough. What makes an app alive is unique data. We are going to use PouchDB as a client-side data storage. It’s powerful, easy-to-use, and provides data sync out-of-box.

If you’re not familiar with it, check out this introduction to PouchDB.

The helper class Store is responsible for the interaction with PouchDB:

class Store {

  constructor(name) {
    this.db = new PouchDB(name);
  }

  getAll() {
    // get all items from storage including details
    return this.db.allDocs({
        include_docs: true
      })
      .then(db => {
        // re-map rows to collection of items
        return db.rows.map(row => {
          return row.doc;
        });
      });
  }

  get(id) {
    // find item by id
    return this.db.get(id);
  }

  save(item) {
    // add or update an item depending on _id
    return item._id ?
      this.update(item) :
      this.add(item);
  }

  add(item) {
    // add new item 
    return this.db.post(item);
  }

  update(item) {
    // find item by id
    return this.db.get(item._id)
      .then(updatingItem => {
        // update item
        Object.assign(updatingItem, item);
        return this.db.put(updatingItem);
      });
  }

  remove(id) {
    // find item by id
    return this.db.get(id)
      .then(item => {
        // remove item
        return this.db.remove(item);
      });
  }
}

The code of the Store class is a typical CRUD implementation, providing a Promise-based API.

Now, our main app component can use the Store:

class ContactBook {

  constructor(storeClass) {
    // create store instance
    this.store = new storeClass('contacts');

    // init component internals
    this.init();

    // refresh the component
    this.refresh();
  }

  refresh() {
    // get all contacts from the store
    this.store.getAll().then(contacts => {
      // render retrieved contacts
      this.renderContactList(contacts);
    });
  }

    ...
}

The Store class is passed to the constructor to decouple the app class from the concrete store. Once the store is created, it’s used in the refresh method to get all the contacts.

The app initialization looks like this:

new ContactBook(Store);

Other app methods interact with the store:

class ContactBook {
  ...

  showContact(event) {
    // get contact id from the clicked element attributes
    var contactId = event.currentTarget.getAttribute(CONTACT_ID_ATTR_NAME);

    // get contact by id
    this.store.get(contactId).then(contact => {
      // show contact details
      this.setContactDetails(contact);

      // turn off editing
      this.toggleContactFormEditing(false);
    })
  }

  editContact() {
    // get id of selected contact
    var contactId = this.getContactId();

    // get contact by id
    this.store.get(this.getContactId()).then(contact => {
      // show contact details
      this.setContactDetails(contact);

      // turn on editing
      this.toggleContactFormEditing(true);
    });
  }

  saveContact() {
    // get contact details from edit form
    var contact = this.getContactDetails();

    // save contact
    this.store.save(contact).then(() => {
      // clear contact details form
      this.setContactDetails({});

      // turn off editing
      this.toggleContactFormEditing(false);

      // refresh contact list
      this.refresh();
    });
  }

  removeContact() {
    // ask user to confirm deletion
    if (!window.confirm(CONTACT_REMOVE_CONFIRM))
      return;

    // get id of selected contact
    var contactId = this.getContactId();

    // remove contact by id
    this.store.remove(contactId).then(() => {
      // clear contact details form
      this.setContactDetails({});

      // turn off editing
      this.toggleContactFormEditing(false);

      // refresh contact list
      this.refresh();
    });
  }

These are the basic operations using the store CRUD methods:

  • showContact – displays contact details once a contact is selected from the list
  • editContact – enables editing of a contact’s details
  • saveContact – saves details of a new or existing contact
  • removeContact – removes the selected contact

Now, if you add contacts while offline and refresh the page, the data won’t be lost.

But, there is a ‘but’…

Data sync

This all works fine, but all data is stored locally in the browser. If we open the app in another browser, we won’t see the changes.

We need to implement data sync with the server. Implementation of the two-way data sync is not a trivial problem. Fortunately, it’s provided by PouchDB if we have CouchDB on the backend.

Let’s change our Store class a bit, so it syncs with the remote data source:

class Store {

  constructor(name, remote, onChange) {
    this.db = new PouchDB(name);

    // start sync in pull mode
    PouchDB.sync(name, `${remote}/${name}`, {
      live: true,
      retry: true
    }).on('change', info => {
      onChange(info);
    });
  }

We’ve added two parameters to the constructor:

  • remote – the URL of the remote server
  • onChange – a callback fired once changes have come from the backend

The PouchDB.sync method does the trick and starts the sync with the backend. The live parameter indicates that it should periodically check for changes, while retry indicates to retry when an error happens (thus if the user goes offline, synchronization won’t stop).

We need to change the app class accordingly and pass required parameters to the Store constructor:

class ContactBook {

  constructor(storeClass, remote) {
    this.store = new storeClass('contacts', remote, () => {
      // refresh contact list when data changed
      this.refresh();
    });

    ...
  }

The main app class constructor now accepts the remote URL which is passed to the store. The onChange callback just calls refresh method to refresh the list of contacts.

The app initialization has to be updated:

new ContactBook(Store, 'http://localhost:5984');

Done! Now our app allows editing the contact list while offline. Once the app has a network connection, the data is synchronized with the backend storage.

Let’s test it:

  1. run the web server with $ npm run serve
  2. open the URL http://127.0.0.1:8080/ in two different browsers
  3. stop the web server clicking Ctrl + C
  4. edit contact list in both browsers
  5. run the web server again with $ npm run serve
  6. check out the contact list in both browsers (it should be up-to-date according to the changes in both browsers)

offline-contactbook-demo

Wonderful, we did it!

Check out the full source code of the app on GitHub.

Conclusion

Providing an offline experience has more and more value today. Being able to use an app with a flaky connection in transport or being offline in the airplane is critical for frequently used apps. It’s also about improving the app performance.

To support offline we needed to take care of:

  • caching app assets – use Service Workers with fallback to AppCache until the former is supported by all modern browsers
  • storing data on the client-side – use browser offline storage, like IndexedDB, with one of the libs available out there

We’ve just looked at how all this can be implemented. I hope you enjoyed reading. Please, share your thoughts about the topic in comments!

This article was peer reviewed by James Kolce and Craig Buckler. Thanks to all of SitePoint’s peer reviewers for making SitePoint content the best it can be!

  • Serkan Sipahi

    Thank you. Very good post… :)

  • Md Kamrul Hasan Pulok

    Thanks. I was looking for this kind of great post. Just found it. Appreciate your effort!

  • Excellent post. Excellent reference. Thanks.

Recommended
Sponsors
Get the latest in JavaScript, once a week, for free.