Using fragments

Learn how to use fragments to share fields across queries

A GraphQL fragment is a shared piece of query logic.

1
fragment NameParts on Person {
2
  firstName
3
  lastName
4
}
5

6
query GetPerson {
7
  people(id: "7") {
8
    ...NameParts
9
    avatar(size: LARGE)
10
  }
11
}

It's important to note that the component after the on clause is designated for the type we are selecting from. In this case, people is of type Person and we want to select the firstName and lastName fields from people(id: "7").

There are two principal uses for fragments in Apollo:

Sharing fields between multiple queries, mutations or subscriptions.
Breaking your queries up to allow you to co-locate field access with the places they are used.

In this document we'll outline patterns to do both; we'll also make use of utilities in the graphql-anywhere and graphql-tag packages which aim to help us, especially with the second problem.

Reusing fragments

The most straightforward use of fragments is to reuse parts of queries (or mutations or subscriptions) in various parts of your application. For instance, in GitHunt on the comments page, we want to fetch the same fields after posting a comment as we originally query. This way we can be sure that we render consistent comment objects as the data changes.

To do so, we can simply share a fragment describing the fields we need for a comment:

1
import gql from 'graphql-tag';
2

3
CommentsPage.fragments = {
4
  comment: gql`
5
    fragment CommentsPageComment on Comment {
6
      id
7
      postedBy {
8
        login
9
        html_url
10
      }
11
      createdAt
12
      content
13
    }
14
  `,
15
};

We put the fragment on CommentsPage.fragments.comment by convention, and use the familiar gql helper to create it.

When it's time to embed the fragment in a query, we simply use the ...Name syntax in our GraphQL, and embed the fragment inside our query GraphQL document:

1
const SUBMIT_COMMENT_MUTATION = gql`
2
  mutation SubmitComment($repoFullName: String!, $commentContent: String!) {
3
    submitComment(repoFullName: $repoFullName, commentContent: $commentContent) {
4
      ...CommentsPageComment
5
    }
6
  }
7
  ${CommentsPage.fragments.comment}
8
`;
9

10
export const COMMENT_QUERY = gql`
11
  query Comment($repoName: String!) {
12
    # ...
13
    entry(repoFullName: $repoName) {
14
      # ...
15
      comments {
16
        ...CommentsPageComment
17
      }
18
      # ...
19
    }
20
  }
21
  ${CommentsPage.fragments.comment}
22
`;

You can see the full source code to the CommentsPage in GitHunt here.

Colocating fragments

A key advantage of GraphQL is the tree-like nature of the response data, which in many cases mirrors your rendered component hierarchy. This, combined with GraphQL's support for fragments, allows you to split your queries up in such a way that the various fields fetched by the queries are located right alongside the code that uses the field.

Although this technique doesn't always make sense (for instance it's not always the case that the GraphQL schema is driven by the UI requirements), when it does, it's possible to use some patterns in Apollo client to take full advantage of it.

In GitHunt, we show an example of this on the FeedPage, which constructs the following view hierarchy:

1
FeedPage
2
└── Feed
3
    └── FeedEntry
4
        ├── RepoInfo
5
        └── VoteButtons

The FeedPage conducts a query to fetch a list of Entrys, and each of the subcomponents requires different subfields of each Entry.

The graphql-anywhere package gives us tools to easily construct a single query that provides all the fields that each subcomponent needs, and allows to easily pass the exact field that a component needs to it.

Creating fragments

To create the fragments, we again use the gql helper and attach to subfields of ComponentClass.fragments, for example:

1
VoteButtons.fragments = {
2
  entry: gql`
3
    fragment VoteButtons on Entry {
4
      score
5
      vote {
6
        vote_value
7
      }
8
    }
9
  `,
10
};

If our fragments include sub-fragments then we can pass them into the gql helper:

1
FeedEntry.fragments = {
2
  entry: gql`
3
    fragment FeedEntry on Entry {
4
      commentCount
5
      repository {
6
        full_name
7
        html_url
8
        owner {
9
          avatar_url
10
        }
11
      }
12
      ...VoteButtons
13
      ...RepoInfo
14
    }
15
    ${VoteButtons.fragments.entry}
16
    ${RepoInfo.fragments.entry}
17
  `,
18
};

Filtering with fragments

We can also use the graphql-anywhere package to filter the exact fields from the entry before passing them to the subcomponent. So when we render a VoteButtons, we can simply do:

1
import { filter } from 'graphql-anywhere';
2

3
<VoteButtons
4
  entry={filter(VoteButtons.fragments.entry, entry)}
5
  canVote={loggedIn}
6
  onVote={type => onVote({
7
    repoFullName: full_name,
8
    type,
9
  })}
10
/>

The filter() function will grab exactly the fields from the entry that the fragment defines.

Importing fragments when using Webpack

When loading .graphql files with graphql-tag/loader, we can include fragments using import statements. For example:

1
#import "./someFragment.graphql"

Will make the contents of someFragment.graphql available to the current file. See the Webpack Fragments section for additional details.

Fragments on unions and interfaces

This is an advanced feature that Apollo Boost does not support. Learn how to set Apollo Client up manually in our Apollo Boost migration guide.

By default, Apollo Client doesn't require any knowledge of the GraphQL schema, which means it's very easy to set up and works with any server and supports even the largest schemas. However, as your usage of Apollo and GraphQL becomes more sophisticated, you may start using fragments on interfaces or unions. Here's an example of a query that uses fragments on an interface:

1
query {
2
  all_people {
3
    ... on Character {
4
      name
5
    }
6
    ... on Jedi {
7
      side
8
    }
9
    ... on Droid {
10
      model
11
    }
12
  }
13
}

In the query above, all_people returns a result of type Character[]. Both Jedi and Droid are possible concrete types of Character, but on the client there is no way to know that without having some information about the schema. By default, Apollo Client's cache will use a heuristic fragment matcher, which assumes that a fragment matched if the result included all the fields in its selection set, and didn't match when any field was missing. This works in most cases, but it also means that Apollo Client cannot check the server response for you, and it cannot tell you when you're manually writing invalid data into the store using update, updateQuery, writeQuery, etc. Also, the heuristic fragment matcher will not work accurately when using fragments with unions or interfaces. Apollo Client will let you know this with a console warning (in development), if it attempts to use the default heuristic fragment matcher with unions/interfaces. The IntrospectionFragmentMatcher is the solution for working with unions/interfaces, and is explained in more detail below.

The section below explains how to pass the necessary schema knowledge to the Apollo Client cache so unions and interfaces can be accurately matched and results validated before writing them into the store.

To support result validation and accurate fragment matching on unions and interfaces, a special fragment matcher called the IntrospectionFragmentMatcher can be used. If there are any changes related to union or interface types in your schema, you will have to update the fragment matcher accordingly.

We recommend setting up a build step that extracts the necessary information from the schema into a JSON file, where it can be imported from when constructing the fragment matcher. To set it up, follow the three steps below:

Query your server / schema to obtain the necessary information about unions and interfaces and write it to a file.

You can automate this or set this as a script to run at build time.

If you want to auto-generate the introspection result, there's a tool called GraphQL Code Generator that does it. Define where your GraphQL Schema is available and where to write the file:

1
# codegen.yml
2
schema: YOUR_API
3
overwrite: true
4
generates:
5
  ./fragmentTypes.json:
6
    plugins:
7
      - fragment-matcher

With all of that, you simply run:

1
gql-gen

To learn more, you can read the "Fragment Matcher" chapter.

In order to introspect the server manually, set this as a script to run at build time.

1
const fetch = require('node-fetch');
2
const fs = require('fs');
3

4
fetch(`${YOUR_API_HOST}/graphql`, {
5
  method: 'POST',
6
  headers: { 'Content-Type': 'application/json' },
7
  body: JSON.stringify({
8
    variables: {},
9
    query: `
10
      {
11
        __schema {
12
          types {
13
            kind
14
            name
15
            possibleTypes {
16
              name
17
            }
18
          }
19
        }
20
      }
21
    `,
22
  }),
23
})
24
  .then(result => result.json())
25
  .then(result => {
26
    // here we're filtering out any type information unrelated to unions or interfaces
27
    const filteredData = result.data.__schema.types.filter(
28
      type => type.possibleTypes !== null,
29
    );
30
    result.data.__schema.types = filteredData;
31
    fs.writeFile('./fragmentTypes.json', JSON.stringify(result.data), err => {
32
      if (err) {
33
        console.error('Error writing fragmentTypes file', err);
34
      } else {
35
        console.log('Fragment types successfully extracted!');
36
      }
37
    });
38
  });

Create a new IntrospectionFragment matcher by passing in the fragmentTypes.json file you just created. You'll want to do this in the same file where you initialize the cache for Apollo Client.

1
import { IntrospectionFragmentMatcher } from 'apollo-cache-inmemory';
2
import introspectionQueryResultData from './fragmentTypes.json';
3

4
const fragmentMatcher = new IntrospectionFragmentMatcher({
5
  introspectionQueryResultData
6
});

Pass in the newly created IntrospectionFragmentMatcher to configure your cache during construction. Then, you pass your newly configured cache to ApolloClient to complete the process.

1
import ApolloClient from 'apollo-client';
2
import { InMemoryCache } from 'apollo-cache-inmemory';
3
import { HttpLink } from 'apollo-link-http';
4

5
// add fragmentMatcher code from step 2 here
6

7
const cache = new InMemoryCache({ fragmentMatcher });
8

9
const client = new ApolloClient({
10
  cache,
11
  link: new HttpLink(),
12
});

Pagination

Error handling