An Angular based Whatsapp clone



Feb 25th (latest)
version 2.0.0
Jan 30th
version 1.4.0
May 28th
version 0.1.0
Chats listing
1 / 13

Our goal for this chapter will be very simple: we want to simply retrieve and display a list of chats from our own server.


We'll start with the server, so let's install a couple of packages first:

yarn add apollo-server-express body-parser cors express graphql
yarn add -D @types/body-parser @types/cors @types/express @types/graphql

Express is a fast, unopinionated, minimalist web framework for node. Cross-Origin Resource Sharing (CORS) is a mechanism that uses additional HTTP headers to let a user agent gain permission to access selected resources from a server on a different origin (domain) than the site currently in use. A user agent makes a cross-origin HTTP request when it requests a resource from a different domain, protocol, or port than the one from which the current document originated. We will need CORS because Webpack's development server used in the client will make use of a different port than the Express server, thus configuring a different origin. GraphQL is a query language for APIs and a runtime for fulfilling those queries with your existing data. GraphQL provides a complete and understandable description of the data in your API, gives clients the power to ask for exactly what they need and nothing more, makes it easier to evolve APIs over time, and enables powerful developer tools. Apollo Server is a community-maintained open-source GraphQL server. It works with pretty much all Node.js HTTP server frameworks. Apollo Server works with any GraphQL schema built with GraphQL.js or with a convenience library such as graphql-tools.

The GraphQL query language is basically about selecting fields on objects. Because the shape of a GraphQL query closely matches the result, you can predict what the query will return without knowing that much about the server. But it's useful to have an exact description of the data we can ask for - what fields can we select? What kinds of objects might they return? What fields are available on those sub-objects? That's where the schema comes in. Every GraphQL service defines a set of types which completely describe the set of possible data you can query on that service. Then, when queries come in, they are validated and executed against that schema.

For the moment let's create some empty schemas and resolvers:

Step 1.1: Create empty Apollo server

Added schema/index.ts
Added schema/resolvers.ts
Added schema/typeDefs.ts

Time to create our index:

Step 1.1: Create empty Apollo server

Added index.ts

Now we want to feed our graphql server with some data. Soon we will need moment, so let's install it:

yarn add moment

Now we can create a fake db:

Step 1.2: Add fake db

Added db.ts

Its' finally time to create our schema

Step 1.3: Add resolvers and schema

Changed schema/typeDefs.ts

and our resolvers

Step 1.3: Add resolvers and schema

Changed schema/resolvers.ts

Out basic server is already done and working. We still have no way to do any kind of mutation, but we already set up several queries to return a list of users or chats. In particular we can choose if we want to return all the chats (and how many messages we want to return for each chat) or if we want to return a single chat. We can also choose which and how many properties we want to return for each query. We can start the server by simply running:

yarn start


Now we can concentrate on the client and bootstrap it using Angular-CLI. First you will need to install it globally with:

yarn global add @angular/cli

Then we can create a new project from scratch:

ng new client --style scss

Time to install a couple of packages:

ng add apollo-angular
yarn add -D @types/graphql

Apollo's Schematics sets everything for us. The only thing missing is to fill the url.

Step 1.1: Add angular-apollo to app module

Added src/app/graphql.module.ts

Let's take a closer look what Schematics prepared for us.

First, it created graphql.module.ts and used ApolloModule and HttpLinkModule.

  • ApolloModule is the center of using GraphQL in your app! It includes all needed services that allows to use ApolloClient’s features.
  • HttpLinkModule makes it easy to fetch data in Angular.

HttpLinkModule is optional, you can replace it with any other Link. Its biggest advantage of all is that it uses Angular's HttpClient internally so it's possible to use it in NativeScript or in combination with any other HttpClient provider. By using HttpLinkModule you get Server-Side Rendering for free, without any additional work.

Next, we got APOLLO_OPTIONS provided to Angular.

We're using the InMemoryCache cache, but there are several options like Redux, Hermes and even ngrx.

  • apollo-cache-inmemory is an official and recommended Apollo Cache.

Cache and Links

These are two main parts of Apollo Stack, both are pluggable and customizable.

  • Cache is responsible for storing data and keeping it consistent.
  • Links describe how to access the source of data (usually a GraphQL endpoint).

Let's take a closer look on what cache means for Apollo, we're going to base it on the official apollo-cache-inmemory, one of many implementations.

But first, imagine, we write our own amazing Apollo Cache implementation.

query getMessages {
  messages {

When we execute a query like one above what we get in return is an array of all messages of logged in user:

  "data": {
    "messages": [
        "id": 0,
        "content": "Hi",
        "likes": 0
        "id": 1,
        "content": "Hello",
        "likes": 0

Our cache stores the result in exact shape as above.

Later on we add more queries and one of them returns a list of messages from one conversation.

Now, a short story. We've got two users, Niccolo and Dotan.

  1. Niccolo opens the app and Apollo fetches all of his messages (I know, it's stupid but stay with me)
  2. One message is to Dotan, Niccolo said "Hi" to him.
  3. Now Dotan opens the app and reacts to the message with a like.
  4. After few minutes Niccolo asks for messages from a conversation with Dotan.
  5. Apollo fetches those messages and there is one in which Niccolo says "Hi" and it's liked by Dotan.
  6. Now we have the same message twice, one has different state that the other, in one place there is no like, in other, we have +1.

As you see, our cache is not smart enough to figure out one query might be a subset of another. That's because it doesn't get enough informations.

Let's go back to InMemoryCache and see how it describes itself:

Thanks to InMemoryCache, it's possible for the results of a query or mutation to update the UI in all the right places. In many cases it's possible for that to happen automatically, whereas in others you need to help the client out a little in doing so.

It means it could solve the issue we have with duplicated messages with different state.

It's very important to understand one concept that InMemoryCache is heavily based on, it's called normalization.

The InMemoryCache splits the result into individual objects, creates a unique identifier for each of them and flattens everything up before saving it to the store. Might seem complicated at first, but as InMemoryCache, let's break everything into smaller parts.

First, "breaking the result into pieces" part. The InMemoryCache tells ApolloClient to add __typename field to your query before passing it to the server. I used server for simplicity, it actually passes it to Apollo Links and they are responsible to make a HTTP request. Every Apollo Cache is able to transform a document, that's as a side note.

query getMessages {
  messages {

The __typename field tells about the type of the object. It's compatible with every GraphQL server as it's part of the specification.

type Message {
  id: ID
  content: String
  likes: Int

type Query {
  messages: [Message]

In our case, __typename returns Message. So now InMemoryCache knows that a message is a Message. Computers are stupid, we have to tell them everything!

Next, let's take a look at "creates a unique identifier" part.

The InMemoryCache iterates through every object and looks for the __typename and two fields, id and _id that usually are used as identifiers. Taking our example, it turns every Message object into Message:1, Message:2, Message:3 and so on. You get the idea.

It's the default behaviour but you can change it as you wish:

const cache = new InMemoryCache({
  dataIdFromObject(result) {
    if (result.__typename) {
      const id = result.uid || || result._id;

      if (id !== undefined) {
        return `${result.__typename}:${id}`;
    return null;

Great, we covered two out of three things InMemoryCache does. Now we're going to explore "flattens everything up".

We're able to understand the type of each individual object and it's unique identifier. It still has to be stored.

What InMemoryCache does is it's saves every element on root level in a plain object where the key contains a unique identifier and the value has a body.

  "Message:1": {
    "id": 1,
    "content": "Hi",
    "likes": 1,
    "__typename": "Message"
  "Message:2": {
    "id": 2,
    "content": "Hello",
    "likes": 0,
    "__typename": "Message"
    "messages": [
        "id": "Message:1",
        "typename": "Message"
        "id": "Message:2",
        "typename": "Message"

Wait, ROOT_QUERY? Yes, it has to somehow store the actual result of every executed query so next time you ask for the same thing, it matches it with the query in store and resolves right away.

You should be aware that the InMemoryCache stringifies variables of a query too, which means the same queries but with different variables are stored separately. Even the order of those variables matters.

As you can see, each element in ROOT_QUERY.message is kind of a reference to a message record.

Let's bring back our short story, a scenario. When Niccolo asks for all messages, the whole three part process is started, each message is saved to the store. When he asks for history of a conversation with Dotan, InMemoryCache starts the process again and if it sees that there is already a massage with the same unique identifier, it knows they both are the same so it merged them. The InMemoryCache can tell what changed and emits the new object to the UI. Thanks to that we no longer have the same message in two places where one has a like and other doesn't.

Uff, Apollo Cache part is done. Thanks for being with us that long, I hope you didn't get bored because there's a lot ahead of us.

Let's talk Apollo Link.

As I mentioned it's kind of a network stack of Apollo. You execute a document with some metadata like variables or a context, Apollo Client receives it and passes it to the chain of Links.

The ApolloLink is based on an Observable (not RxJS) and if you know RxJS you're familiar with pipe. Each Link is basically a function that receives a requested operation and a function that runs the next Link. Just like a pipe!

import { ApolloLink } from 'apollo-link';

const log = new ApolloLink((operation, next) => {
  console.log('Name', operation.operationName);
  return next(operation);

Now when we add it in graphql.module.ts we see the name of each requested operation.

We can even intercept the result.

const intercept = new ApolloLink((operation, next) => {
    return forward(operation).map((data) => {
        // data from a previous link
        console.log('data', data);
        return data;

As you can see below, by network we don't always mean making HTTP requests. It could be Http Link at the end of Links chain or WebSocket Link or even something that works on in-memory data.

import { ApolloLink, Observable } from 'apollo-link';

const inmemory = new ApolloLink(operation => {
    return Observable(observer => {
        // function that executes the operation and returns data
        const result = resolveOperation(operation);;

You can find tons of helpful Links on npm and on Apollo Link documentation.

We highly recommend to take a look at Angular related Links on Apollo Angular documentation.


The gql template tag is what you use to define GraphQL queries in your Apollo apps. It parses your GraphQL query into the GraphQL.js AST format which may then be consumed by Apollo methods. Whenever Apollo is asking for a GraphQL query you will always want to wrap it in a gql template tag.

You can embed a GraphQL document containing only fragments inside of another GraphQL document using template string interpolation. This allows you to use fragments defined in one part of your codebase inside of a query defined in a completely different file.

Step 1.2: Add chats service

Added src/app/services/chats.service.ts
Added src/graphql/fragment.ts
Added src/graphql/getChats.query.ts

Use Apollo service

Let's create a simple service to query the chats from our just created server:

Step 1.2: Add chats service

Added src/app/services/chats.service.ts

We just learned how to use Apollo to attach GraphQL query results to the Angular UI.

The watchQuery method returns a QueryRef object which has the valueChanges property that is an Observable.

Every fetched data is stored in Apollo Client’s cache, so if some other query fetches new information about the chats, this component will update to remain consistent.

It’s also possible to fetch data only once. You simply use query method instead of watchQuery. The query method of Apollo service returns an Observable too that also resolves with the same result as above.

But what is a QueryRef?

As you know, query method returns an Observable that emits a result, just once, then the observable completes. watchQuery works a bit different. It fetches data then stays open and listens to the Cache for updates.

So why watchQuery can not simply expose an Observable and it serves QueryRef instead?

Because Apollo Angular is a bridge between Apollo Client (core library) and Angular framework it has to operate on RxJS based Observables. RxJS Observable has a simple API that cannot be easily extended and it's in general a bad practice.

Why we're talking about those things? Apollo Client exposes an Observable that has bunch of custom method on it. They are super useful and their purpose is to talk to Apollo through them.

Since we have to turn Apollo's Observable to something that is compatible with RxJS we decided to expose an object that has all those methods and valueChanges property that is a regular RxJS Observable.

Make our app pretty

We will use Materials for the UI, so let's install it:

yarn add @angular/cdk @angular/material hammerjs ng2-truncate

Let's configure Material:

Step 1.3: List the chats

Changed src/main.ts
Changed src/styles.scss

We're now creating a shared module where we will define our header component where we're going to project a different content from each component:

Step 1.3: List the chats

Added src/app/shared/components/toolbar/toolbar.component.scss
Added src/app/shared/components/toolbar/toolbar.component.ts
Added src/app/shared/shared.module.ts

Now we want to create the chats-lister module, with a container component called ChatsComponent and a couple of presentational components. Each and every chat item would be presented with a default profile picture as a placeholder, so before we proceed we will have to download it to the src/assets directory:

$ wget src/assets

And now we can proceed to creating the relevant components and their style sheets:

Step 1.3: List the chats

Added src/app/chats-lister/chats-lister.module.ts
Added src/app/chats-lister/components/chat-item/chat-item.component.scss
Added src/app/chats-lister/components/chat-item/chat-item.component.ts
Added src/app/chats-lister/components/chats-list/chats-list.component.scss
Added src/app/chats-lister/components/chats-list/chats-list.component.ts
Added src/app/chats-lister/containers/chats/chats.component.scss
Added src/app/chats-lister/containers/chats/chats.component.ts

Finally let's wire everything up to the main module:

Step 1.3: List the chats

Changed src/app/app.component.ts
Changed src/app/app.module.ts

If you will try to run the frontend you will notice that several messages seems like duplicated, why does it happen?

Since we use NoSQL-like structure in our backend, messages are stored as an array inside each chat so their incremental identifiers are not unique across different chats. We need to normalize them in a way that takes into account both the message id and the chat id:

Step 1.4: Better normalize messages

Changed src/app/graphql.module.ts

That way our application will work even if the backend is a NoSQL. What's even more interesting is that our application will keep working as well even when we will switch our backend to PostgreSQL.