Async Operations

Flutter apps regularly talk to the network, disk, and other asynchronous sources. This guide shows how to keep those flows reactive when you build with Flutter Compositions.

Goals

• Keep loading/error/data state co-located with the widget that needs it.
• Share asynchronous state with descendants without prop drilling.
• Make retries and refresh flows explicit.

Choosing the Right Composable

Use case	Recommended tool	Notes
One-off future with simple UI	useFuture	Returns a Ref<AsyncValue<T>> that you can switch on.
Future that depends on other reactive values	useAsyncData	Automatically re-runs when the watch input changes.
Stream subscriptions	useStream	Keeps the latest event in a Ref<T?>.
Exposing loading/error/data separately	useAsyncValue	Splits an AsyncValue<T> into individual refs.
Imperative refresh button	useAsyncData	The returned refresh() callback can be exposed to the UI.

Example: Loading a Profile

class ProfilePage extends CompositionWidget {
  const ProfilePage({super.key, required this.userId});

  final String userId;

  @override
  Widget Function(BuildContext) setup() {
    final props = widget();

    final (status, refresh) = useAsyncData<User, String>(
      (id) => api.fetchUser(id),
      watch: () => props.value.userId,
    );

    return (context) => switch (status.value) {
          AsyncLoading() => const Center(child: CircularProgressIndicator()),
          AsyncError(:final errorValue) => ErrorView(
              message: '$errorValue',
              onRetry: refresh,
            ),
          AsyncData(:final value) => ProfileView(user: value),
          AsyncIdle() => const SizedBox.shrink(),
        };
  }
}

Key points:

• useAsyncData recalculates when props.value.userId changes.
• The refresh callback can be wired to pull-to-refresh or a retry button.
• Returning early for AsyncIdle covers the first frame before the request starts.

Example: Streaming Updates

class NotificationsPanel extends CompositionWidget {
  @override
  Widget Function(BuildContext) setup() {
    final notifications = useStream<NotificationEvent>(
      notificationBus.events,
      initialData: null,
    );

    final recent = computed(() {
      final event = notifications.value;
      if (event == null) return <NotificationEvent>[];
      return [...previousEvents.value, event]
          .take(10)
          .toList(growable: false);
    });

    return (context) => NotificationList(events: recent.value);
  }
}

• useStream disposes the subscription automatically.
• Combine the latest emission with other refs (previousEvents here) to build derived state.

Sharing Async State with Descendants

const todosKey = InjectionKey<ReadonlyRef<AsyncValue<List<Todo>>>>('todos');

class TodosProvider extends CompositionWidget {
  @override
  Widget Function(BuildContext) setup() {
    final (todos, refresh) = useAsyncData<List<Todo>, void>(
      (_) => api.fetchTodos(),
    );

    provide(todosKey, todos);

    return (context) => Column(
          children: [
            ElevatedButton(onPressed: refresh, child: const Text('Refresh')),
            const TodosList(),
          ],
        );
  }
}

class TodosList extends CompositionWidget {
  const TodosList({super.key});

  @override
  Widget Function(BuildContext) setup() {
    final todos = inject(todosKey);

    return (context) => switch (todos.value) {
          AsyncLoading() => const CircularProgressIndicator(),
          AsyncError(:final errorValue) => Text('Error: $errorValue'),
          AsyncData(:final value) => ListView(
              children: [for (final todo in value) TodoTile(todo: todo)],
            ),
          _ => const SizedBox.shrink(),
        };
  }
}

• Provide the entire Ref<AsyncValue<T>> to descendants to keep reactivity.
• Consumers stay lightweight—they just switch on the async state.

Handling Parallel Requests

When you need to run multiple futures simultaneously, use Future.wait inside useAsyncData:

final (status, _) = useAsyncData<_DashboardPayload, void>(
  (_) async {
    final [user, stats, notifications] = await Future.wait([
      api.fetchUser(),
      api.fetchStats(),
      api.fetchNotifications(),
    ]);
    return _DashboardPayload(user, stats, notifications);
  },
);

Wrap heavy requests with caching (e.g., keep the last result in a Ref and early-return) to avoid unnecessary network calls during rebuilds.

Testing Tips

• Pass in fake services through InjectionKeys so that composables stay deterministic.
• Pump frames with tester.pump() until the async request completes.
• For useStream, emit test events through a StreamController that you control inside the test.

Next Steps

• Learn how dependency injection keeps async code modular in Dependency Injection.
• Explore a larger end-to-end example in Best Practices.