graphql_flutter 2.0.0

A GraphQL client for Flutter, bringing all the features from a modern GraphQL client to one easy to use package.

GraphQL Flutter

Table of Contents

• Installation
• Usage
  • GraphQL Provider
  • Offline Cache
    • Normalization
    • Optimism
  • Queries
  • Mutations
    • Mutations with optimism
  • Subscriptions (Experimental)
  • GraphQL Consumer
  • Graphql Upload
• Roadmap

Installation

First, depends on the library by adding this to your packages pubspec.yaml:

dependencies:
  graphql_flutter: ^1.0.0

Now inside your Dart code, you can import it.

import 'package:graphql_flutter/graphql_flutter.dart';

Usage

To use the client it first needs to be initialized with a link and cache. For this example, we will be using an HttpLink as our link and InMemoryCache as our cache. If your endpoint requires authentication you can concatenate the AuthLink, it resolves the credentials using a future, so you can authenticate asynchronously.

For this example we will use the public GitHub API.

...

import 'package:graphql_flutter/graphql_flutter.dart';

void main() {
  final HttpLink httpLink = HttpLink(
    uri: 'https://api.github.com/graphql',
  );

  final AuthLink authLink = AuthLink(
    getToken: () async => 'Bearer <YOUR_PERSONAL_ACCESS_TOKEN>',
    // OR
    // getToken: () => 'Bearer <YOUR_PERSONAL_ACCESS_TOKEN>',
  );

  final Link link = authLink.concat(httpLink as Link);

  ValueNotifier<GraphQLClient> client = ValueNotifier(
    GraphQLClient(
      cache: InMemoryCache(),
      link: link,
    ),
  );

  ...
}

...

GraphQL Provider

In order to use the client, you Query and Mutation widgets to be wrapped with the GraphQLProvider widget.

We recommend wrapping your MaterialApp with the GraphQLProvider widget.

  ...

  return GraphQLProvider(
    client: client,
    child: MaterialApp(
      title: 'Flutter Demo',
      ...
    ),
  );

  ...

Offline Cache

The in-memory cache can automatically be saved to and restored from offline storage. Setting it up is as easy as wrapping your app with the CacheProvider widget.

It is required to place the CacheProvider widget is inside the GraphQLProvider widget, because GraphQLProvider makes client available through the build context.

...

class MyApp extends StatelessWidget {
  @override
  Widget build(BuildContext context) {
    return GraphQLProvider(
      client: client,
      child: CacheProvider(
        child: MaterialApp(
          title: 'Flutter Demo',
          ...
        ),
      ),
    );
  }
}

...

Normalization

To enable apollo-like normalization, use a NormalizedInMemoryCache or OptimisticCache:

ValueNotifier<GraphQLClient> client = ValueNotifier(
  GraphQLClient(
    cache: NormalizedInMemoryCache(
      dataIdFromObject: typenameDataIdFromObject,
    ),
    link: link,
  ),
);

dataIdFromObject is required and has no defaults. Our implementation is similar to Apollo's, requiring a function to return a universally unique string or null. The predefined typenameDataIdFromObject we provide is similar to apollo's default:

String typenameDataIdFromObject(Object object) {
  if (object is Map<String, Object> &&
      object.containsKey('__typename') &&
      object.containsKey('id')) {
    return "${object['__typename']}/${object['id']}";
  }
  return null;
}

However, note that graphql-flutter does not inject __typename into operations the way Apollo does, so if you aren't careful to request them in your query, this normalization scheme is not possible.

Unlike Apollo, we don't have a real client-side document parser and resolver, so operations leveraging normalization can have additional fields not specified in the query. There are a couple of ideas for constraining this (leveraging json_serializable, or just implementing the resolver), but for now, the normalized cache uses a LazyCacheMap, which wraps underlying data with a lazy denormalizer to allow for cyclical references. It has the same API as a normal HashMap, but is currently a bit hard to debug with, as a descriptive debug representation is currently unavailable.

NOTE: A LazyCacheMap can be modified, but this does not affect the underlying entities in the cache. If references are added to the map, they will still dereference against the cache normally.

Optimism

The OptimisticCache allows for optimistic mutations by passing an optimisticResult to RunMutation. It will then call update(Cache cache, QueryResult result) twice (once eagerly with optimisticResult), and rebroadcast all queries with the optimistic cache. You can tell which entities in the cache are optimistic through the .isOptimistic flag on LazyCacheMap, though note that this is only the case for optimistic entities and not their containing operations/maps.

QueryResults also, have an optimistic flag, but I would recommend looking at the data itself, as many situations make it unusable (such as toggling mutations like in the example below). Mutation usage examples

Queries

Creating a query is as simple as creating a multiline string:

String readRepositories = """
  query ReadRepositories(\$nRepositories: Int!) {
    viewer {
      repositories(last: \$nRepositories) {
        nodes {
          id
          name
          viewerHasStarred
        }
      }
    }
  }
""";

In your widget:

// ...
Query(
  options: QueryOptions(
    document: readRepositories, // this is the query string you just created
    variables: {
      'nRepositories': 50,
    },
    pollInterval: 10,
  ),
  // Just like in apollo refetch() could be used to manually trigger a refetch
  builder: (QueryResult result, { VoidCallback refetch }) {
    if (result.errors != null) {
      return Text(result.errors.toString());
    }

    if (result.loading) {
      return Text('Loading');
    }

    // it can be either Map or List
    List repositories = result.data['viewer']['repositories']['nodes'];

    return ListView.builder(
      itemCount: repositories.length,
      itemBuilder: (context, index) {
        final repository = repositories[index];

        return Text(repository['name']);
    });
  },
);
// ...

Mutations

Again first create a mutation string:

String addStar = """
  mutation AddStar(\$starrableId: ID!) {
    addStar(input: {starrableId: \$starrableId}) {
      starrable {
        viewerHasStarred
      }
    }
  }
""";

The syntax for mutations is fairly similar to that of a query. The only difference is that the first argument of the builder function is a mutation function. Just call it to trigger the mutations (Yeah we deliberately stole this from react-apollo.)

...

Mutation(
  options: MutationOptions(
    document: addStar, // this is the mutation string you just created
  ),
  builder: (
    RunMutation runMutation,
    QueryResult result,
  ) {
    return FloatingActionButton(
      onPressed: () => runMutation({
        'starrableId': <A_STARTABLE_REPOSITORY_ID>,
      }),
      tooltip: 'Star',
      child: Icon(Icons.star),
    );
  },
  // you can update the cache based on results
  update: (Cache cache, QueryResult result) {
    return cache;
  },
  // or do something with the result.data on completion
  onCompleted: (dynamic resultData) {
    print(resultData);
  },
);

...

Mutations with optimism

If you're using an OptimisticCache, you can provide an optimisticResult:

...
FlutterWidget(
  onTap: () {
    toggleStar(
      { 'starrableId': repository['id'] },
      optimisticResult: {
        'action': {
          'starrable': {'viewerHasStarred': !starred}
        }
      },
    );
  },
)
...

With a bit more context (taken from the complete mutation example StarrableRepository):

// bool get starred => repository['viewerHasStarred'] as bool;
// bool get optimistic => (repository as LazyCacheMap).isOptimistic;
Mutation(
  options: MutationOptions(
    document: starred ? mutations.removeStar : mutations.addStar,
  ),
  builder: (RunMutation toggleStar, QueryResult result) {
    return ListTile(
      leading: starred
          ? const Icon(
              Icons.star,
              color: Colors.amber,
            )
          : const Icon(Icons.star_border),
      trailing: result.loading || optimistic
          ? const CircularProgressIndicator()
          : null,
      title: Text(repository['name'] as String),
      onTap: () {
        toggleStar(
          { 'starrableId': repository['id'] },
          optimisticResult: {
            'action': {
              'starrable': {'viewerHasStarred': !starred}
            }
          },
        );
      },
    );
  },
  // will be called for both optimistic and final results
  update: (Cache cache, QueryResult result) {
    if (result.hasErrors) {
      print(['optimistic', result.errors]);
    } else {
      final Map<String, Object> updated =
          Map<String, Object>.from(repository)
            ..addAll(extractRepositoryData(result.data));
      cache.write(typenameDataIdFromObject(updated), updated);
    }
  },
  // will only be called for final result
  onCompleted: (dynamic resultData) {
    showDialog<AlertDialog>(
      context: context,
      builder: (BuildContext context) {
        return AlertDialog(
          title: Text(
            extractRepositoryData(resultData)['viewerHasStarred'] as bool
                ? 'Thanks for your star!'
                : 'Sorry you changed your mind!',
          ),
          actions: <Widget>[
            SimpleDialogOption(
              child: const Text('Dismiss'),
              onPressed: () {
                Navigator.of(context).pop();
              },
            )
          ],
        );
      },
    );
  },
);

Subscriptions (Experimental)

The syntax for subscriptions is again similar to a query, however, this utilizes WebSockets and dart Streams to provide real-time updates from a server. Before subscriptions can be performed a global instance of socketClient needs to be initialized.

We are working on moving this into the same GraphQLProvider structure as the http client. Therefore this api might change in the near future.

socketClient = await SocketClient.connect('ws://coolserver.com/graphql');

Once the socketClient has been initialized it can be used by the Subscription Widget

class _MyHomePageState extends State<MyHomePage> {
  @override
  Widget build(BuildContext context) {
    return Scaffold(
      body: Center(
        child: Subscription(
          operationName,
          query,
          variables: variables,
          builder: ({
            bool loading,
            dynamic payload,
            dynamic error,
          }) {
            if (payload != null) {
              return Text(payload['requestSubscription']['requestData']);
            } else {
              return Text('Data not found');
            }
          }
        ),
      )
    );
  }
}

GraphQL Consumer

You can always access the client directly from the GraphQLProvider but to make it even easier you can also use them GraphQLConsumer widget.

  ...

  return GraphQLConsumer(
    builder: (GraphQLClient client) {
      // do something with the client

      return Container(
        child: Text('Hello world'),
      );
    },
  );

  ...

GraphQL Upload

We support GraphQL Upload spec as proposed at https://github.com/jaydenseric/graphql-multipart-request-spec

mutation($files: [Upload!]!) {
  multipleUpload(files: $files) {
    id
    filename
    mimetype
    path
  }
}

import 'dart:io' show File;

// ...

String filePath = '/aboslute/path/to/file.ext';
final QueryResult r = await graphQLClientClient.mutate(
  MutationOptions(
    document: uploadMutation,
    variables: {
      'files': [File(filePath)],
    },
  )
);

Roadmap

This is currently our roadmap, please feel free to request additions/changes.

Feature	Progress
Queries	✅
Mutations	✅
Subscriptions	✅
Query polling	✅
In memory cache	✅
Offline cache sync	✅
GraphQL pload	✅
Optimistic results	✅
Client state management	🔜
Modularity	🔜