Direct Cache Access

Apollo Client normalizes all of your data so that if any data you previously fetched from your GraphQL server is updated in a later data fetch from your server then your data will be updated with the latest truth from your server.

This normalization process is constantly happening behind the scenes when you call watchQuery but this process is often not enough to describe the updates to your data model as the result of a mutation. For example, if you wanted to add an item to the end of an array fetched by one of your queries. You also might want to read data from the normalized Apollo Client store at a specific id without making another GraphQL server fetch.

To interact directly with your data in the Apollo Client store you may use the methods readQuery, readFragment, writeQuery, and writeFragment that are accessible from the ApolloClient class. This article will teach you how to use these methods to control your data.

If you would like a better understanding of the data normalization process then we recommend reading the 'How it works' documentation article. Knowledge around how Apollo Client works is not a prerequisite for using the methods described here, but it may be helpful.

All the methods we will discuss can be called from the ApolloClient class. Any code demonstration in this article will assume that we have already initialized an instance of ApolloClient and that we have imported the gql tag from graphql-tag.

You can read more about Caching here

Updating the Cache after a Mutation

Being able to read and write to the Apollo cache from anywhere in your application gives you a lot of power over your data. However, there is one place where we most often want to update our cached data: after a mutation. As such, Apollo Client has optimized the experience for updating your cache with the read and write methods after a mutation with the update function. Let us say that we have the following GraphQL mutation:

mutation TodoCreateMutation($text: String!) {
  createTodo(text: $text) {
    id
    text
    completed
  }
}

We may also have the following GraphQL query:

query TodoAppQuery {
  todos {
    id
    text
    completed
  }
}

At the end of our mutation we want our query to include the new todo like we had sent our TodoAppQuery a second time after the mutation finished without actually sending the query. To do this we can use the update function provided as an option of the Apollo.mutate method. To update your cache with the mutation just write code that looks like:

// We assume that the GraphQL operations `TodoCreateMutation` and
// `TodoAppQuery` have already been defined using the `gql` tag.
 
const text = 'Hello, world!';
 
@Component({ ... })
class AppComponent {
  createTodo() {
    this.apollo.mutate({
      mutation: TodoCreateMutation,
      variables: {
        text,
      },
      update: (proxy, { data: { createTodo } }) => {
        // Read the data from our cache for this query.
        const data = proxy.readQuery({ query: TodoAppQuery });
 
        // If you are using the Query service (TodoAppGQL) instead of defining your GQL as a constant, you can reference the query as:
        // const data = proxy.readQuery({ query: this.todoAppGQL.document });
 
        // Add our todo from the mutation to the end.
        data.todos.push(createTodo);
 
        // Write our data back to the cache.
        proxy.writeQuery({ query: TodoAppQuery, data });
 
        // alternatively when using Query service:
        // proxy.writeQuery({ query: this.todoAppGQL.document, data });
      },
    }).subscribe();
  }
}

The first proxy argument is an instance of DataProxy (opens in a new tab) has the same for methods that we just learned exist on the Apollo Client: readQuery, readFragment, writeQuery, and writeFragment. The reason we call them on a proxy object here instead of on our client instance is that we can easily apply optimistic updates (which we will demonstrate in a bit). The proxy object also provides an isolated transaction which shields you from any other mutations going on at the same time, and the proxy object also batches writes together until the very end.

If you provide an optimisticResponse option to the mutation then the update function will be run twice. Once immediately after you call apollo.mutate with the data from optimisticResponse. After the mutation successfully executes against the server the changes made in the first call to update will be rolled back and update will be called with the actual data returned by the mutation and not just the optimistic response.

Putting it all together:

const text = 'Hello, world!';
 
@Component({ ... })
class AppComponent {
  createTodo() {
    this.apollo.mutate({
      mutation: TodoCreateMutation,
      variables: {
        text,
      },
      optimisticResponse: {
        id: -1, // -1 is a temporary id for the optimistic response.
        text,
        completed: false,
      },
      update: (proxy, { data: { createTodo } }) => {
        const data = proxy.readQuery({ query: TodoAppQuery });
        data.todos.push(createTodo);
        proxy.writeQuery({ query: TodoAppQuery, data });
      },
    });
  }
}

As you can see the update function on apollo.mutate provides extra change management functionality specific to the use case of a mutation while still providing you the powerful data control APIs that are available on Apollo service.

The update function is not a good place for side-effects as it may be called multiple times. Also, you may not call any of the methods on proxy asynchronously.

Resources

Error Handling Optimistic UI