Apollo Server

Apollo Server is the part of the Apollo suite that we use to create the backend of a GraphQL project: a GraphQL server.

It is able to do the following:

• Receive an incoming GraphQL query from a client
• Validate that query against the server schema
• Populate the queried schema fields
• Return the fields as a JSON response object

Example schema

We will use the following schema in the examples.

// schema.js

const typeDefs = gql`
  " Our schema types will be nested here
`;
module.exports = typeDefs;

type Query {
  tracksForHome: [Track!]!
}

type Track {
  id: ID!
    author: Author!
  thumbnail: String
  length: Int
  modulesCount: Int
}

type Author {
  id: ID!
  name: String!
  photo: String
}

Setting up the server

We instantiate an ApolloServer instance and pass our schema to it. We then subscribe to it with a listener.

// index.js

const { ApolloServer } = require("apollo-server");
const typeDefs = require("./schema");
const server = new ApolloServer({ typeDefs });

server.listen().then(() => {
  console.log(`
        Server is running!
        Listening on port 4000
        Query at http://localhost:4000
  `);
});

When we access the local URL we are able to access the Apollo server using the Explorer GUI. This automatically loads our schema and is basically a more fancy version of GraphiQL:

It makes it easy to read descriptions of the dataypes and to construct queries by clicking to insert fields.

Adding some mock data

We are not connected to a database yet but we can create a mock that will enable us to run test queries.

We do this just by updating the Apollo Server options. We can either use generic dummy data or provide our own mock.

Generic mock

const server = new ApolloServer({ typeDefs, mocks: true });

Custom mock

const mocks = {
  Track: () => ({
    id: () => "track_01",
        author: () => {
      return {
        name: "Grumpy Cat",
        photo:
          "https://res.cloudinary.com/dety84pbu/image/upload/v1606816219/kitty-veyron-sm_mctf3c.jpg",
      };
    },
    thumbnail: () =>
      "https://res.cloudinary.com/dety84pbu/image/upload/v1598465568/nebula_cat_djkt9r.jpg",
    length: () => 1210,
    modulesCount: () => 6,
  }),
};

const server = new ApolloServer({ typeDefs, mocks });

We can now run queries against our server.

Implementing resolvers

A resolver is a function that populates data for a given query. It should have the same name as the field for the query. So far we have one query in our schema: tracksForHome which returns the tracks to be listed on the home page. We must therefore also name our resolver for this query tracksForHome.

It can fetch data from a single data source or multiple data sources (other servers, databases, REST APIs) and present this as a single integrated resource to the client, matching the shape requested.

As per the generic example, you write write your resolvers as keys on a resolvers object, e.g:

const resolvers = {};

The resolvers object’s keys will correspond to the schema’s types and fields. You distinguish resolvers which directly correspond to a query in the schema from other resolver types by wraping them in Query {}.

const resolvers = {
  Query: {
    tracksForHome: () => {},
  },
};

Resolver parameters

Each resolver function has the same standard parameters that you can invoke when implementing the resolution: resolverFunction(parent, args, context, info).

• parent
  • Used with resolver chains ---add example
• args
  • an object comprising arguments provided for the given field by the client. For instance if the client requests a field with an accompanying id argument, id can be parsed via the args object
• context
  • shared state between different resolvers that contains essential connection parameters such as authentication, a database connection, or a RESTDataSource (see below). This will be typically instantiated via a class which is then invoked within the ApolloServer instance under the dataSources key.
• info
  • not used so frequently but employed as part of caching

Typically you won’t use every parameter with every resolver. You can ommit them with _, __; the number of dashes indicating the argument placement.

RESTDataSource

A resolver can return data from multiple sources. One of the most common sources is a RESTful endpoint. Apollo provides a specific class for handling REST endpoints in your resolvers: RESTDataSource.

REST APIs fall victim to the “n + 1” problem: say you want to get an array of one resource type, then for each element returned you need to send another request using one of its properties to fetch a related resource.

This is implicit in the case of the Track type in the schema. Each Track has an author key but the Author type isn’t embedded in Track it has to be fetched using an id. In a REST API, this would require a request to a separate end point for each Track returned, increasing the time complexity of the request.

Here is an example of RESTDataSource being used. It is just a class that can be extended and which provides inbuilt methods for running fetches against a REST API:

const { RESTDataSource } = require("apollo-datasource-rest");

class TrackAPI extends RESTDataSource {
  constructor() {
    super();
    this.baseURL = "https://odyssey-lift-off-rest-api.herokuapp.com/";
  }

  getTracksForHome() {
    return this.get("tracks");
  }

  getAuthor(authorId) {
    return this.get(`author/${authorId}`);
  }
}

Using our RESTDataSource in our resolver

As our GraphQL server is sourcing data from a REST API, we can now integrate the RESTDataSource class with our resolver.

First thing, we need to instantiate an instance of our TrackApi class, otherwise we won’t be able to use any of its methods in the resolver.

We will create an instance of this class and pass it into ApolloServer object we established at the beginning. We will pass it to the dataSources key. This will allow us to access it from within the context parameter in our resolver function

We can also get rid of the mocks object since we don’t need it any more. We will replace it with our resolvers constant:

const server = new ApolloServer({
  typeDefs,
-  mocks,
+  resolvers,
+  dataSources: () => {
+  return {
+    trackApi: new TrackApi()
+  }
  }
})

Now we can complete our resolver:

const resolvers = {
  Query: {
    tracksForHome: (_, __, {dataSources}) => {},
      return dataSources.trackApi.getTracksForHome()
  },
};

So we destructure the dataSources object from the parent Apollo Server instance (in the place of the context parameter) which gives us access to our trackApi class. This resolver will now make the API request and return the tracks.

The tracksForHome query returns Track objects and these have a required author key that returns an Author type. So we are also going to need a resolver that can return the author data that will be populated along with Track.

We already have this functionality in our class: getAuthor so we just need to integrate it:

const resolvers = {
  Query: {
    tracksForHome: (_, __, { dataSources }) => {
      return dataSources.trackApi.getTracksForHome();
    },
  },
  Track: {
    author: ({ authorId }, _, { dataSources }) => {
      return dataSources.trackApi.getAuthor(authorId);
    },
  },
};

• Just as we nest the tracksForHome resolver under Query, we must nest author under Track to match the structure of the schema. This resolver doesn’t respond to a query that is exposed to the client so it shouldn’t go under Query.

• We invoke the context again when we destructure dataSources from the ApolloServer instance.
• This time we utilise the args parameter in the resolver since an id will be provided as a client-side argument to return a specific author.

The useMutation hook

We invoke the useMutation hook to issue mutations from the client-side.

As with queries and query constants