Catch the highlights of GraphQLConf 2023! Click for recordings. Or check out our recap blog post.
v2
Local State
Managing State with Field Policies

Managing State with Field Policies

Your Apollo Client queries can include local-only fields that aren't defined in your GraphQL server's schema. The values for these fields are calculated locally using any logic you want, such as reading data from localStorage.

A single query can include both local-only fields and fields that are fetched from your GraphQL server.

Defining

Let's say we're building an e-commerce application. Most of a product's details are stored on our back-end server, but we want to define an isInCart boolean field that's local to the client. First, we create a field policy for isInCart.

A field policy specifies custom logic for how a single GraphQL field is fetched from and written to your Apollo Client cache. You can define field policies for both local-only fields and remotely fetched fields.

💡

Field policies are primarily documented in Customizing the behavior of cached fields. This article specifically describes how to use them with local-only fields.

You define your application's field policies in a map that you provide to the constructor of Apollo Client's InMemoryCache. Each field policy is a child of a particular type policy (much like the corresponding field is a child of a particular type).

Here's a sample InMemoryCache constructor that defines a field policy for Product.isInCart:

const cache = new InMemoryCache({
  typePolicies: {
    // Type policy map
    Product: {
      fields: {
        // Field policy map for the Product type
        isInCart: {
          // Field policy for the isInCart field
          read(_, { variables }) {
            // The read function for the isInCart field
            return localStorage.getItem('CART').includes(variables.productId);
          },
        },
      },
    },
  },
});

The field policy above defines a read function for the isInCart field. Whenever you query a field that has a read function, the cache calls that function to calculate the field's value. In this case, the read function returns whether the queried product's ID is in the CART array in localStorage.

You can use read functions to perform any sort of logic you want, including:

  • Manually executing other cache operations
  • Calling helper utilities or libraries to prepare, validate, or sanitize data
  • Fetching data from a separate store
  • Logging usage metrics
💡

If you query a local-only field that doesn't define a read function, Apollo Client performs a default cache lookup for the field. See Storing local state in the cache for details.

Querying

Now that we've defined a field policy for isInCart, we can include the field in a query that also queries our back-end server, like so:

const GET_PRODUCT_DETAILS = gql`
  query ProductDetails($productId: ID!) {
    product(id: $productId) {
      name
      price
      isInCart @client
    }
  }
`;

That's it! The @client directive tells Apollo Client that isInCart is a local-only field. Because isInCart is local-only, Apollo Client omits it from the query it sends to our server to fetch name and price. The final query result is returned only after all local and remote fields are populated.

Storing

You can use Apollo Client to query local state, regardless of how you store that state. Apollo Client provides a couple of optional but helpful mechanisms for representing local state:

Storing local state in reactive variables

Apollo Client reactive variables are great for representing local state:

  • You can read and modify reactive variables from anywhere in your application, without needing to use a GraphQL operation to do so.
  • Unlike the Apollo Client cache, reactive variables don't enforce data normalization, meaning you can store data in any format you want.
  • If a field's value depends on the value of a reactive variable, and that variable's value changes, every active query that includes the field automatically refreshes.

Example

Returning to our e-commerce application, let's say we want to fetch a list of the item IDs in a user's cart, and this list is stored locally. The query to do that looks like this:

export const GET_CART_ITEMS = gql`
  query GetCartItems {
    cartItems @client
  }
`;

Let's initialize a reactive variable to store our local list of cart items, like so:

export const cartItemsVar = makeVar([]);

This initializes a reactive variable that contains an empty array. We can get this variable's current value by calling cartItemsVar(), and we can set a new value by calling cartItemsVar(newValue).

Next, let's define the field policy for cartItems. As always, we pass this to the constructor of InMemoryCache:

export const cache = new InMemoryCache({
  typePolicies: {
    Query: {
      fields: {
        cartItems: {
          read() {
            return cartItemsVar();
          },
        },
      },
    },
  },
});

This read function returns the value of our reactive variable whenever cartItems is queried.

Now, let's create a button component that enables the user to add a product to their cart:

import { cartItemsVar } from './cache';
// ... other imports
 
@Component({ ... })
export class AddToCardComponent {
  @Input() productId: number;
 
  submit() {
    const cartItems = cartItemsVar();
    cartItemsVar([...cartItems, this.productId])
  }
}

On submit call, the value of cartItemsVar is updated by appending productId. When this happens, Apollo Client notifies every active query that includes the cartItems field.

Here's a Cart component that uses the GET_CART_ITEMS query and therefore refreshes automatically whenever the value of cartItemsVar changes:

export const GET_CART_ITEMS = gql`
  query GetCartItems {
    cartItems @client
  }
`;
 
@Component({...})
export class CartComponent {
  data: Observable<any>;
 
  constructor(private apollo: Apollo) {
    // Emits a new result when cartItemsVar changes its value
    this.data = this.apollo.watchQuery({
      query: GET_CART_ITEMS,
    }).valueChanges;
  }
}

Storing local state in the cache

Storing local state directly in the Apollo Client cache provides some advantages, but usually requires more code than using reactive variables:

  • You don't have to define a field policy for local-only fields that are present in the cache. If you query a field that doesn't define a read function, by default Apollo Client attempts to fetch that field's value directly from the cache.
  • When you modify a cached field with writeQuery or writeFragment, every active query that includes the field automatically refreshes.

Example

Let's say our application defines the following query:

const IS_LOGGED_IN = gql`
  query IsUserLoggedIn {
    isLoggedIn @client
  }
`;

The isLoggedIn field of this query is a local-only field. We can use the writeQuery method to write a value for this field directly to the Apollo Client cache, like so:

cache.writeQuery({
  query: IS_LOGGED_IN,
  data: {
    isLoggedIn: !!localStorage.getItem('token'),
  },
});

This writes a boolean value according to whether localStorage includes a token item, indicating an active session.

Now our application's components can render according to the value of the isLoggedIn field, without our needing to define a read function for it:

@Component({
  template: `
    <div *ngIf="isLoggedIn | async; else loginBlock">
      <app-pages />
    </div>
 
    <ng-template #loginBlock>
      <app-login />
    </ng-template>
  `,
})
export class AppComponent {
  isLoggedIn: Observable<boolean>;
 
  constructor(private apollo: Apollo) {
    this.isLoggedIn = this.apollo
      .watchQuery({
        query: IS_LOGGED_IN,
      })
      .valueChanges.pipe(map(result => result.data.isLoggedIn));
  }
}

Note that even if you do store local data as fields in the Apollo Client cache, you can (and probably should!) still define read functions for those fields. A read function can execute helpful custom logic, such as returning a default value if a field isn't present in the cache.

Modifying

The way you modify the value of a local-only field depends on how you store that field:

  • If you're using a reactive variable, all you need to do is set the reactive variable's new value. Apollo Client automatically detects this change and triggers a refresh of every active operation that includes an affected field.

  • If you're using the cache directly, call one of writeQuery, writeFragment, or cache.modify (all documented here) to modify cached fields. Like reactive variables, all of these methods trigger a refresh of every affected active operation.

  • If you're using another storage method, such as localStorage, set the field's new value in whatever method you're using. Then, you can force a refresh of every affected operation by calling cache.evict. In your call, provide both the id of your field's containing object and the name of the local-only field.

Using local-only fields as GraphQL variables

If your GraphQL query uses variables, the local-only fields of that query can provide the values of those variables.

To do so, you apply the @export(as: "variableName") directive, like so:

const GET_CURRENT_AUTHOR_POST_COUNT = gql`
  query CurrentAuthorPostCount($authorId: Int!) {
    currentAuthorId @client @export(as: "authorId")
    postCount(authorId: $authorId)
  }
`;

In the query above, the result of the local-only field currentAuthorId is used as the value of the $authorId variable that's passed to postCount.

You can do this even if postCount is also a local-only field (i.e., if it's also marked as @client).

Considerations for using @export

  • To use the @export directive, a field must also use the @client directive. In other words, only local-only fields can be used as variable values.

  • A field that @exports a variable value must appear before any fields that use that variable.

  • If multiple fields in an operation use the @export directive to assign their value to the same variable, the field listed last takes precedence. When this happens in development mode, Apollo Client logs a warning message.

  • At first glance, the @export directive appears to violate the GraphQL specification's requirement that the execution order of an operation must not affect its result:

💡

…the resolution of fields other than top‐level mutation fields must always be side effect‐free and idempotent, the execution order must not affect the result, and hence the server has the freedom to execute the field entries in whatever order it deems optimal.

However, all @exported variable values are populated before an operation is sent to a remote server. Only local-only fields can use the @export directive, and those fields are stripped from operations before they're transmitted.