컨텐츠로 건너뛰기
Bloc

Migration Guide

이 내용은 아직 번역본이 없습니다.

v9.0.0

package:hydrated_bloc

✨ Reintroduce HydratedBloc.storage API

Rationale

Refer to the rationale for reintroducing the Bloc.observer and Bloc.transformer overrides.

v8.x.x

Future<void> main() async {
  final storage = await HydratedStorage.build(
    storageDirectory: kIsWeb
        ? HydratedStorage.webStorageDirectory
        : await getTemporaryDirectory(),
  );
  HydratedBlocOverrides.runZoned(
    () => runApp(App()),
    storage: storage,
  );
}

v9.0.0

Future<void> main() async {
  WidgetsFlutterBinding.ensureInitialized();
  HydratedBloc.storage = await HydratedStorage.build(
    storageDirectory: kIsWeb
        ? HydratedStorage.webStorageDirectory
        : await getTemporaryDirectory(),
  );
  runApp(App());
}

v8.1.0

package:bloc

✨ Reintroduce Bloc.observer and Bloc.transformer APIs

Rationale

The BlocOverrides API was introduced in v8.0.0 in an attempt to support scoping bloc-specific configurations such as BlocObserver, EventTransformer, and HydratedStorage. In pure Dart applications, the changes worked well; however, in Flutter applications the new API caused more problems than it solved.

The BlocOverrides API was inspired by similar APIs in Flutter/Dart:

Problems

While it wasn’t the primary reason for these changes, the BlocOverrides API introduced additional complexity for developers. In addition to increasing the amount of nesting and lines of code needed to achieve the same effect, the BlocOverrides API required developers to have a solid understanding of Zones in Dart. Zones are not a beginner-friendly concept and failure to understand how Zones work could lead to the introduction of bugs (such as uninitialized observers, transformers, storage instances).

For example, many developers would have something like:

void main() {
  WidgetsFlutterBinding.ensureInitialized();
  BlocOverrides.runZoned(...);
}

The above code, while appearing harmless, can actually lead to many difficult to track bugs. Whatever zone WidgetsFlutterBinding.ensureInitialized is initially called from will be the zone in which gesture events are handled (e.g. onTap, onPressed callbacks) due to GestureBinding.initInstances. This is just one of many issues caused by using zoneValues.

In addition, Flutter does many things behind the scenes which involve forking/manipulating Zones (especially when running tests) which can lead to unexpected behaviors (and in many cases behaviors that are outside the developer’s control — see issues below).

Due to the use of the runZoned, the transition to the BlocOverrides API led to the discovery of several bugs/limitations in Flutter (specifically around Widget and Integration Tests):

which affected many developers using the bloc library:

v8.0.x

void main() {
  BlocOverrides.runZoned(
    () {
      // ...
    },
    blocObserver: CustomBlocObserver(),
    eventTransformer: customEventTransformer(),
  );
}

v8.1.0

void main() {
  Bloc.observer = CustomBlocObserver();
  Bloc.transformer = customEventTransformer();


  // ...
}

v8.0.0

package:bloc

❗✨ Introduce new BlocOverrides API

Rationale

The previous API used to override the default BlocObserver and EventTransformer relied on a global singleton for both the BlocObserver and EventTransformer.

As a result, it was not possible to:

  • Have multiple BlocObserver or EventTransformer implementations scoped to different parts of the application
  • Have BlocObserver or EventTransformer overrides be scoped to a package
    • If a package were to depend on package:bloc and registered its own BlocObserver, any consumer of the package would either have to overwrite the package’s BlocObserver or report to the package’s BlocObserver.

It was also more difficult to test because of the shared global state across tests.

Bloc v8.0.0 introduces a BlocOverrides class which allows developers to override BlocObserver and/or EventTransformer for a specific Zone rather than relying on a global mutable singleton.

v7.x.x

void main() {
  Bloc.observer = CustomBlocObserver();
  Bloc.transformer = customEventTransformer();


  // ...
}

v8.0.0

void main() {
  BlocOverrides.runZoned(
    () {
      // ...
    },
    blocObserver: CustomBlocObserver(),
    eventTransformer: customEventTransformer(),
  );
}

Bloc instances will use the BlocObserver and/or EventTransformer for the current Zone via BlocOverrides.current. If there are no BlocOverrides for the zone, they will use the existing internal defaults (no change in behavior/functionality).

This allows allow each Zone to function independently with its own BlocOverrides.

BlocOverrides.runZoned(
  () {
    // BlocObserverA and eventTransformerA
    final overrides = BlocOverrides.current;


    // Blocs in this zone report to BlocObserverA
    // and use eventTransformerA as the default transformer.
    // ...


    // Later...
    BlocOverrides.runZoned(
      () {
        // BlocObserverB and eventTransformerB
        final overrides = BlocOverrides.current;


        // Blocs in this zone report to BlocObserverB
        // and use eventTransformerB as the default transformer.
        // ...
      },
      blocObserver: BlocObserverB(),
      eventTransformer: eventTransformerB(),
    );
  },
  blocObserver: BlocObserverA(),
  eventTransformer: eventTransformerA(),
);

❗✨ Improve Error Handling and Reporting

Rationale

The goal of these changes is:

  • make internal unhandled exceptions extremely obvious while still preserving bloc functionality
  • support addError without disrupting control flow

Previously, error handling and reporting varied depending on whether the application was running in debug or release mode. In addition, errors reported via addError were treated as uncaught exceptions in debug mode which led to a poor developer experience when using the addError API (specifically when writing unit tests).

In v8.0.0, addError can be safely used to report errors and blocTest can be used to verify that errors are reported. All errors are still reported to onError, however, only uncaught exceptions are rethrown (regardless of debug or release mode).

❗🧹 Make BlocObserver abstract

Rationale

BlocObserver was intended to be an interface. Since the default API implementation are no-ops, BlocObserver is now an abstract class to clearly communicate that the class is meant to be extended and not directly instantiated.

v7.x.x

void main() {
  // It was possible to create an instance of the base class.
  final observer = BlocObserver();
}

v8.0.0

class MyBlocObserver extends BlocObserver {...}


void main() {
  // Cannot instantiate the base class.
  final observer = BlocObserver(); // ERROR


  // Extend `BlocObserver` instead.
  final observer = MyBlocObserver(); // OK
}

❗✨ add throws StateError if Bloc is closed

Rationale

Previously, it was possible to call add on a closed bloc and the internal error would get swallowed, making it difficult to debug why the added event was not being processed. In order to make this scenario more visible, in v8.0.0, calling add on a closed bloc will throw a StateError which will be reported as an uncaught exception and propagated to onError.

❗✨ emit throws StateError if Bloc is closed

Rationale

Previously, it was possible to call emit within a closed bloc and no state change would occur but there would also be no indication of what went wrong, making it difficult to debug. In order to make this scenario more visible, in v8.0.0, calling emit within a closed bloc will throw a StateError which will be reported as an uncaught exception and propagated to onError.

❗🧹 Remove Deprecated APIs

Summary
  • mapEventToState removed in favor of on<Event>
  • transformEvents removed in favor of EventTransformer API
  • TransitionFunction typedef removed in favor of EventTransformer API
  • listen removed in favor of stream.listen

package:bloc_test

✨ MockBloc and MockCubit no longer require registerFallbackValue

Summary

registerFallbackValue is only needed when using the any() matcher from package:mocktail for a custom type. Previously, registerFallbackValue was needed for every Event and State when using MockBloc or MockCubit.

v8.x.x

class FakeMyEvent extends Fake implements MyEvent {}
class FakeMyState extends Fake implements MyState {}
class MyMockBloc extends MockBloc<MyEvent, MyState> implements MyBloc {}


void main() {
  setUpAll(() {
    registerFallbackValue(FakeMyEvent());
    registerFallbackValue(FakeMyState());
  });


  // Tests...
}

v9.0.0

class MyMockBloc extends MockBloc<MyEvent, MyState> implements MyBloc {}


void main() {
  // Tests...
}

package:hydrated_bloc

❗✨ Introduce new HydratedBlocOverrides API

Rationale

Previously, a global singleton was used to override the Storage implementation.

As a result, it was not possible to have multiple Storage implementations scoped to different parts of the application. It was also more difficult to test because of the shared global state across tests.

HydratedBloc v8.0.0 introduces a HydratedBlocOverrides class which allows developers to override Storage for a specific Zone rather than relying on a global mutable singleton.

v7.x.x

void main() async {
  HydratedBloc.storage = await HydratedStorage.build(
    storageDirectory: await getApplicationSupportDirectory(),
  );


  // ...
}

v8.0.0

void main() {
  final storage = await HydratedStorage.build(
    storageDirectory: await getApplicationSupportDirectory(),
  );


  HydratedBlocOverrides.runZoned(
    () {
      // ...
    },
    storage: storage,
  );
}

HydratedBloc instances will use the Storage for the current Zone via HydratedBlocOverrides.current.

This allows allow each Zone to function independently with its own BlocOverrides.

v7.2.0

package:bloc

✨ Introduce new on<Event> API

Rationale

The on<Event> API was introduced as part of [Proposal] Replace mapEventToState with on<Event> in Bloc. Due to an issue in Dart it’s not always obvious what the value of state will be when dealing with nested async generators (async*). Even though there are ways to work around the issue, one of the core principles of the bloc library is to be predictable. The on<Event> API was created to make the library as safe as possible to use and to eliminate any uncertainty when it comes to state changes.

Summary

on<E> allows you to register an event handler for all events of type E. By default, events will be processed concurrently when using on<E> as opposed to mapEventToState which processes events sequentially.

v7.1.0

abstract class CounterEvent {}
class Increment extends CounterEvent {}


class CounterBloc extends Bloc<CounterEvent, int> {
  CounterBloc() : super(0);


  @override
  Stream<int> mapEventToState(CounterEvent event) async* {
    if (event is Increment) {
      yield state + 1;
    }
  }
}

v7.2.0

abstract class CounterEvent {}
class Increment extends CounterEvent {}


class CounterBloc extends Bloc<CounterEvent, int> {
  CounterBloc() : super(0) {
    on<Increment>((event, emit) => emit(state + 1));
  }
}

If you want to retain the exact same behavior as in v7.1.0 you can register a single event handler for all events and apply a sequential transformer:

import 'package:bloc/bloc.dart';
import 'package:bloc_concurrency/bloc_concurrency.dart';


class MyBloc extends Bloc<MyEvent, MyState> {
  MyBloc() : super(MyState()) {
    on<MyEvent>(_onEvent, transformer: sequential())
  }


  FutureOr<void> _onEvent(MyEvent event, Emitter<MyState> emit) async {
    // TODO: logic goes here...
  }
}

You can also override the default EventTransformer for all blocs in your application:

import 'package:bloc/bloc.dart';
import 'package:bloc_concurrency/bloc_concurrency.dart';


void main() {
  Bloc.transformer = sequential<dynamic>();
  ...
}

✨ Introduce new EventTransformer API

Rationale

The on<Event> API opened the door to being able to provide a custom event transformer per event handler. A new EventTransformer typedef was introduced which enables developers to transform the incoming event stream for each event handler rather than having to specify a single event transformer for all events.

Summary

An EventTransformer is responsible for taking the incoming stream of events along with an EventMapper (your event handler) and returning a new stream of events.

typedef EventTransformer<Event> = Stream<Event> Function(Stream<Event> events, EventMapper<Event> mapper)

The default EventTransformer processes all events concurrently and looks something like:

EventTransformer<E> concurrent<E>() {
  return (events, mapper) => events.flatMap(mapper);
}

v7.1.0

@override
Stream<Transition<MyEvent, MyState>> transformEvents(events, transitionFn) {
  return events
    .debounceTime(const Duration(milliseconds: 300))
    .flatMap(transitionFn);
}

v7.2.0

/// Define a custom `EventTransformer`
EventTransformer<MyEvent> debounce<MyEvent>(Duration duration) {
  return (events, mapper) => events.debounceTime(duration).flatMap(mapper);
}


MyBloc() : super(MyState()) {
  /// Apply the custom `EventTransformer` to the `EventHandler`
  on<MyEvent>(_onEvent, transformer: debounce(const Duration(milliseconds: 300)))
}

⚠️ Deprecate transformTransitions API

Rationale

The stream getter on Bloc makes it easy to override the outbound stream of states therefore it’s no longer valuable to maintain a separate transformTransitions API.

Summary

v7.1.0

@override
Stream<Transition<Event, State>> transformTransitions(
  Stream<Transition<Event, State>> transitions,
) {
  return transitions.debounceTime(const Duration(milliseconds: 42));
}

v7.2.0

@override
Stream<State> get stream => super.stream.debounceTime(const Duration(milliseconds: 42));

v7.0.0

package:bloc

❗ Bloc and Cubit extend BlocBase

Rationale

As a developer, the relationship between blocs and cubits was a bit awkward. When cubit was first introduced it began as the base class for blocs which made sense because it had a subset of the functionality and blocs would just extend Cubit and define additional APIs. This came with a few drawbacks:

  • All APIs would either have to be renamed to accept a cubit for accuracy or they would need to be kept as bloc for consistency even though hierarchically it is inaccurate (#1708, #1560).

  • Cubit would need to extend Stream and implement EventSink in order to have a common base which widgets like BlocBuilder, BlocListener, etc. can be implemented against (#1429).

Later, we experimented with inverting the relationship and making bloc the base class which partially resolved the first bullet above but introduced other issues:

  • The cubit API is bloated due to the underlying bloc APIs like mapEventToState, add, etc. (#2228)
    • Developers can technically invoke these APIs and break things
  • We still have the same issue of cubit exposing the entire stream API as before (#1429)

To address these issues we introduced a base class for both Bloc and Cubit called BlocBase so that upstream components can still interoperate with both bloc and cubit instances but without exposing the entire Stream and EventSink API directly.

Summary

BlocObserver

v6.1.x

class SimpleBlocObserver extends BlocObserver {
  @override
  void onCreate(Cubit cubit) {...}


  @override
  void onEvent(Bloc bloc, Object event) {...}


  @override
  void onChange(Cubit cubit, Object event) {...}


  @override
  void onTransition(Bloc bloc, Transition transition) {...}


  @override
  void onError(Cubit cubit, Object error, StackTrace stackTrace) {...}


  @override
  void onClose(Cubit cubit) {...}
}

v7.0.0

class SimpleBlocObserver extends BlocObserver {
  @override
  void onCreate(BlocBase bloc) {...}


  @override
  void onEvent(Bloc bloc, Object event) {...}


  @override
  void onChange(BlocBase bloc, Object? event) {...}


  @override
  void onTransition(Bloc bloc, Transition transition) {...}


  @override
  void onError(BlocBase bloc, Object error, StackTrace stackTrace) {...}


  @override
  void onClose(BlocBase bloc) {...}
}

Bloc/Cubit

v6.1.x

final bloc = MyBloc();
bloc.listen((state) {...});


final cubit = MyCubit();
cubit.listen((state) {...});

v7.0.0

final bloc = MyBloc();
bloc.stream.listen((state) {...});


final cubit = MyCubit();
cubit.stream.listen((state) {...});

package:bloc_test

❗seed returns a function to support dynamic values

Rationale

In order to support having a mutable seed value which can be updated dynamically in setUp, seed returns a function.

Summary

v7.x.x

blocTest(
  '...',
  seed: MyState(),
  ...
);

v8.0.0

blocTest(
  '...',
  seed: () => MyState(),
  ...
);

❗expect returns a function to support dynamic values and includes matcher support

Rationale

In order to support having a mutable expectation which can be updated dynamically in setUp, expect returns a function. expect also supports Matchers.

Summary

v7.x.x

blocTest(
  '...',
  expect: [MyStateA(), MyStateB()],
  ...
);

v8.0.0

blocTest(
  '...',
  expect: () => [MyStateA(), MyStateB()],
  ...
);


// It can also be a `Matcher`
blocTest(
  '...',
  expect: () => contains(MyStateA()),
  ...
);

❗errors returns a function to support dynamic values and includes matcher support

Rationale

In order to support having a mutable errors which can be updated dynamically in setUp, errors returns a function. errors also supports Matchers.

Summary

v7.x.x

blocTest(
  '...',
  errors: [MyError()],
  ...
);

v8.0.0

blocTest(
  '...',
  errors: () => [MyError()],
  ...
);


// It can also be a `Matcher`
blocTest(
  '...',
  errors: () => contains(MyError()),
  ...
);

❗MockBloc and MockCubit

Rationale

To support stubbing of various core APIs, MockBloc and MockCubit are exported as part of the bloc_test package. Previously, MockBloc had to be used for both Bloc and Cubit instances which was not intuitive.

Summary

v7.x.x

class MockMyBloc extends MockBloc<MyState> implements MyBloc {}
class MockMyCubit extends MockBloc<MyState> implements MyBloc {}

v8.0.0

class MockMyBloc extends MockBloc<MyEvent, MyState> implements MyBloc {}
class MockMyCubit extends MockCubit<MyState> implements MyCubit {}

❗Mocktail Integration

Rationale

Due to various limitations of the null-safe package:mockito described here, package:mocktail is used by MockBloc and MockCubit. This allows developers to continue using a familiar mocking API without the need to manually write stubs or rely on code generation.

Summary

v7.x.x

import 'package:mockito/mockito.dart';


...


when(bloc.state).thenReturn(MyState());
verify(bloc.add(any)).called(1);

v8.0.0

import 'package:mocktail/mocktail.dart';


...


when(() => bloc.state).thenReturn(MyState());
verify(() => bloc.add(any())).called(1);

Please refer to #347 as well as the mocktail documentation for more information.

package:flutter_bloc

❗ rename cubit parameter to bloc

Rationale

As a result of the refactor in package:bloc to introduce BlocBase which Bloc and Cubit extend, the parameters of BlocBuilder, BlocConsumer, and BlocListener were renamed from cubit to bloc because the widgets operate on the BlocBase type. This also further aligns with the library name and hopefully improves readability.

Summary

v6.1.x

BlocBuilder(
  cubit: myBloc,
  ...
)


BlocListener(
  cubit: myBloc,
  ...
)


BlocConsumer(
  cubit: myBloc,
  ...
)

v7.0.0

BlocBuilder(
  bloc: myBloc,
  ...
)


BlocListener(
  bloc: myBloc,
  ...
)


BlocConsumer(
  bloc: myBloc,
  ...
)

package:hydrated_bloc

❗storageDirectory is required when calling HydratedStorage.build

Rationale

In order to make package:hydrated_bloc a pure Dart package, the dependency on package:path_provider was removed and the storageDirectory parameter when calling HydratedStorage.build is required and no longer defaults to getTemporaryDirectory.

Summary

v6.x.x

HydratedBloc.storage = await HydratedStorage.build();

v7.0.0

import 'package:path_provider/path_provider.dart';


...


HydratedBloc.storage = await HydratedStorage.build(
  storageDirectory: await getTemporaryDirectory(),
);

v6.1.0

package:flutter_bloc

❗context.bloc and context.repository are deprecated in favor of context.read and context.watch

Rationale

context.read, context.watch, and context.select were added to align with the existing provider API which many developers are familiar and to address issues that have been raised by the community. To improve the safety of the code and maintain consistency, context.bloc was deprecated because it can be replaced with either context.read or context.watch dependending on if it’s used directly within build.

context.watch

context.watch addresses the request to have a MultiBlocBuilder because we can watch several blocs within a single Builder in order to render UI based on multiple states:

Builder(
  builder: (context) {
    final stateA = context.watch<BlocA>().state;
    final stateB = context.watch<BlocB>().state;
    final stateC = context.watch<BlocC>().state;


    // return a Widget which depends on the state of BlocA, BlocB, and BlocC
  }
);

context.select

context.select allows developers to render/update UI based on a part of a bloc state and addresses the request to have a simpler buildWhen.

final name = context.select((UserBloc bloc) => bloc.state.user.name);

The above snippet allows us to access and rebuild the widget only when the current user’s name changes.

context.read

Even though it looks like context.read is identical to context.bloc there are some subtle but significant differences. Both allow you to access a bloc with a BuildContext and do not result in rebuilds; however, context.read cannot be called directly within a build method. There are two main reasons to use context.bloc within build:

  1. To access the bloc’s state
@override
Widget build(BuildContext context) {
  final state = context.bloc<MyBloc>().state;
  return Text('$state');
}

The above usage is error prone because the Text widget will not be rebuilt if the state of the bloc changes. In this scenario, either a BlocBuilder or context.watch should be used.

@override
Widget build(BuildContext context) {
  final state = context.watch<MyBloc>().state;
  return Text('$state');
}

or

@override
Widget build(BuildContext context) {
  return BlocBuilder<MyBloc, MyState>(
    builder: (context, state) => Text('$state'),
  );
}
  1. To access the bloc so that an event can be added
@override
Widget build(BuildContext context) {
  final bloc = context.bloc<MyBloc>();
  return ElevatedButton(
    onPressed: () => bloc.add(MyEvent()),
    ...
  )
}

The above usage is inefficient because it results in a bloc lookup on each rebuild when the bloc is only needed when the user taps the ElevatedButton. In this scenario, prefer to use context.read to access the bloc directly where it is needed (in this case, in the onPressed callback).

@override
Widget build(BuildContext context) {
  return ElevatedButton(
    onPressed: () => context.read<MyBloc>().add(MyEvent()),
    ...
  )
}

Summary

v6.0.x

@override
Widget build(BuildContext context) {
  final bloc = context.bloc<MyBloc>();
  return ElevatedButton(
    onPressed: () => bloc.add(MyEvent()),
    ...
  )
}

v6.1.x

@override
Widget build(BuildContext context) {
  return ElevatedButton(
    onPressed: () => context.read<MyBloc>().add(MyEvent()),
    ...
  )
}

?> If accessing a bloc to add an event, perform the bloc access using context.read in the callback where it is needed.

v6.0.x

@override
Widget build(BuildContext context) {
  final state = context.bloc<MyBloc>().state;
  return Text('$state');
}

v6.1.x

@override
Widget build(BuildContext context) {
  final state = context.watch<MyBloc>().state;
  return Text('$state');
}

?> Use context.watch when accessing the state of the bloc in order to ensure the widget is rebuilt when the state changes.

v6.0.0

package:bloc

❗BlocObserver onError takes Cubit

Rationale

Due to the integration of Cubit, onError is now shared between both Bloc and Cubit instances. Since Cubit is the base, BlocObserver will accept a Cubit type rather than a Bloc type in the onError override.

v5.x.x

class MyBlocObserver extends BlocObserver {
  @override
  void onError(Bloc bloc, Object error, StackTrace stackTrace) {
    super.onError(bloc, error, stackTrace);
  }
}

v6.0.0

class MyBlocObserver extends BlocObserver {
  @override
  void onError(Cubit cubit, Object error, StackTrace stackTrace) {
    super.onError(cubit, error, stackTrace);
  }
}

❗Bloc does not emit last state on subscription

Rationale

This change was made to align Bloc and Cubit with the built-in Stream behavior in Dart. In addition, conforming this the old behavior in the context of Cubit led to many unintended side-effects and overall complicated the internal implementations of other packages such as flutter_bloc and bloc_test unnecessarily (requiring skip(1), etc…).

v5.x.x

final bloc = MyBloc();
bloc.listen(print);

Previously, the above snippet would output the initial state of the bloc followed by subsequent state changes.

v6.x.x

In v6.0.0, the above snippet does not output the initial state and only outputs subsequent state changes. The previous behavior can be achieved with the following:

final bloc = MyBloc();
print(bloc.state);
bloc.listen(print);

?> Note: This change will only affect code that relies on direct bloc subscriptions. When using BlocBuilder, BlocListener, or BlocConsumer there will be no noticeable change in behavior.

package:bloc_test

❗MockBloc only requires State type

Rationale

It is not necessary and eliminates extra code while also making MockBloc compatible with Cubit.

v5.x.x

class MockCounterBloc extends MockBloc<CounterEvent, int> implements CounterBloc {}

v6.0.0

class MockCounterBloc extends MockBloc<int> implements CounterBloc {}

❗whenListen only requires State type

Rationale

It is not necessary and eliminates extra code while also making whenListen compatible with Cubit.

v5.x.x

whenListen<CounterEvent,int>(bloc, Stream.fromIterable([0, 1, 2, 3]));

v6.0.0

whenListen<int>(bloc, Stream.fromIterable([0, 1, 2, 3]));

❗blocTest does not require Event type

Rationale

It is not necessary and eliminates extra code while also making blocTest compatible with Cubit.

v5.x.x

blocTest<CounterBloc, CounterEvent, int>(
  'emits [1] when increment is called',
  build: () async => CounterBloc(),
  act: (bloc) => bloc.add(CounterEvent.increment),
  expect: const <int>[1],
);

v6.0.0

blocTest<CounterBloc, int>(
  'emits [1] when increment is called',
  build: () => CounterBloc(),
  act: (bloc) => bloc.add(CounterEvent.increment),
  expect: const <int>[1],
);

❗blocTest skip defaults to 0

Rationale

Since bloc and cubit instances will no longer emit the latest state for new subscriptions, it was no longer necessary to default skip to 1.

v5.x.x

blocTest<CounterBloc, CounterEvent, int>(
  'emits [0] when skip is 0',
  build: () async => CounterBloc(),
  skip: 0,
  expect: const <int>[0],
);

v6.0.0

blocTest<CounterBloc, int>(
  'emits [] when skip is 0',
  build: () => CounterBloc(),
  skip: 0,
  expect: const <int>[],
);

The initial state of a bloc or cubit can be tested with the following:

test('initial state is correct', () {
  expect(MyBloc().state, InitialState());
});

❗blocTest make build synchronous

Rationale

Previously, build was made async so that various preparation could be done to put the bloc under test in a specific state. It is no longer necessary and also resolves several issues due to the added latency between the build and the subscription internally. Instead of doing async prep to get a bloc in a desired state we can now set the bloc state by chaining emit with the desired state.

v5.x.x

blocTest<CounterBloc, CounterEvent, int>(
  'emits [2] when increment is added',
  build: () async {
    final bloc = CounterBloc();
    bloc.add(CounterEvent.increment);
    await bloc.take(2);
    return bloc;
  }
  act: (bloc) => bloc.add(CounterEvent.increment),
  expect: const <int>[2],
);

v6.0.0

blocTest<CounterBloc, int>(
  'emits [2] when increment is added',
  build: () => CounterBloc()..emit(1),
  act: (bloc) => bloc.add(CounterEvent.increment),
  expect: const <int>[2],
);

package:flutter_bloc

❗BlocBuilder bloc parameter renamed to cubit

Rationale

In order to make BlocBuilder interoperate with bloc and cubit instances the bloc parameter was renamed to cubit (since Cubit is the base class).

v5.x.x

BlocBuilder(
  bloc: myBloc,
  builder: (context, state) {...}
)

v6.0.0

BlocBuilder(
  cubit: myBloc,
  builder: (context, state) {...}
)

❗BlocListener bloc parameter renamed to cubit

Rationale

In order to make BlocListener interoperate with bloc and cubit instances the bloc parameter was renamed to cubit (since Cubit is the base class).

v5.x.x

BlocListener(
  bloc: myBloc,
  listener: (context, state) {...}
)

v6.0.0

BlocListener(
  cubit: myBloc,
  listener: (context, state) {...}
)

❗BlocConsumer bloc parameter renamed to cubit

Rationale

In order to make BlocConsumer interoperate with bloc and cubit instances the bloc parameter was renamed to cubit (since Cubit is the base class).

v5.x.x

BlocConsumer(
  bloc: myBloc,
  listener: (context, state) {...},
  builder: (context, state) {...}
)

v6.0.0

BlocConsumer(
  cubit: myBloc,
  listener: (context, state) {...},
  builder: (context, state) {...}
)

v5.0.0

package:bloc

❗initialState has been removed

Rationale

As a developer, having to override initialState when creating a bloc presents two main issues:

  • The initialState of the bloc can be dynamic and can also be referenced at a later point in time (even outside of the bloc itself). In some ways, this can be viewed as leaking internal bloc information to the UI layer.
  • It’s verbose.

v4.x.x

class CounterBloc extends Bloc<CounterEvent, int> {
  @override
  int get initialState => 0;


  ...
}

v5.0.0

class CounterBloc extends Bloc<CounterEvent, int> {
  CounterBloc() : super(0);


  ...
}

?> For more information check out #1304

❗BlocDelegate renamed to BlocObserver

Rationale

The name BlocDelegate was not an accurate description of the role that the class played. BlocDelegate suggests that the class plays an active role whereas in reality the intended role of the BlocDelegate was for it to be a passive component which simply observes all blocs in an application.

v4.x.x

class MyBlocDelegate extends BlocDelegate {
  ...
}

v5.0.0

class MyBlocObserver extends BlocObserver {
  ...
}

❗BlocSupervisor has been removed

Rationale

BlocSupervisor was yet another component that developers had to know about and interact with for the sole purpose of specifying a custom BlocDelegate. With the change to BlocObserver we felt it improved the developer experience to set the observer directly on the bloc itself.

?> This changed also enabled us to decouple other bloc add-ons like HydratedStorage from the BlocObserver.

v4.x.x

BlocSupervisor.delegate = MyBlocDelegate();

v5.0.0

Bloc.observer = MyBlocObserver();

package:flutter_bloc

❗BlocBuilder condition renamed to buildWhen

Rationale

When using BlocBuilder, we previously could specify a condition to determine whether the builder should rebuild.

BlocBuilder<MyBloc, MyState>(
  condition: (previous, current) {
    // return true/false to determine whether to call builder
  },
  builder: (context, state) {...}
)

The name condition is not very self-explanatory or obvious and more importantly, when interacting with a BlocConsumer the API became inconsistent because developers can provide two conditions (one for builder and one for listener). As a result, the BlocConsumer API exposed a buildWhen and listenWhen

BlocConsumer<MyBloc, MyState>(
  listenWhen: (previous, current) {
    // return true/false to determine whether to call listener
  },
  listener: (context, state) {...},
  buildWhen: (previous, current) {
    // return true/false to determine whether to call builder
  },
  builder: (context, state) {...},
)

In order to align the API and provide a more consistent developer experience, condition was renamed to buildWhen.

v4.x.x

BlocBuilder<MyBloc, MyState>(
  condition: (previous, current) {
    // return true/false to determine whether to call builder
  },
  builder: (context, state) {...}
)

v5.0.0

BlocBuilder<MyBloc, MyState>(
  buildWhen: (previous, current) {
    // return true/false to determine whether to call builder
  },
  builder: (context, state) {...}
)

❗BlocListener condition renamed to listenWhen

Rationale

For the same reasons as described above, the BlocListener condition was also renamed.

v4.x.x

BlocListener<MyBloc, MyState>(
  condition: (previous, current) {
    // return true/false to determine whether to call listener
  },
  listener: (context, state) {...}
)

v5.0.0

BlocListener<MyBloc, MyState>(
  listenWhen: (previous, current) {
    // return true/false to determine whether to call listener
  },
  listener: (context, state) {...}
)

package:hydrated_bloc

❗HydratedStorage and HydratedBlocStorage renamed

Rationale

In order to improve code reuse between hydrated_bloc and hydrated_cubit, the concrete default storage implementation was renamed from HydratedBlocStorage to HydratedStorage. In addition, the HydratedStorage interface was renamed from HydratedStorage to Storage.

v4.0.0

class MyHydratedStorage implements HydratedStorage {
  ...
}

v5.0.0

class MyHydratedStorage implements Storage {
  ...
}

❗HydratedStorage decoupled from BlocDelegate

Rationale

As mentioned earlier, BlocDelegate was renamed to BlocObserver and was set directly as part of the bloc via:

Bloc.observer = MyBlocObserver();

The following change was made to:

  • Stay consistent with the new bloc observer API
  • Keep the storage scoped to just HydratedBloc
  • Decouple the BlocObserver from Storage

v4.0.0

BlocSupervisor.delegate = await HydratedBlocDelegate.build();

v5.0.0

HydratedBloc.storage = await HydratedStorage.build();

❗Simplified Initialization

Rationale

Previously, developers had to manually call super.initialState ?? DefaultInitialState() in order to setup their HydratedBloc instances. This is clunky and verbose and also incompatible with the breaking changes to initialState in bloc. As a result, in v5.0.0 HydratedBloc initialization is identical to normal Bloc initialization.

v4.0.0

class CounterBloc extends HydratedBloc<CounterEvent, int> {
  @override
  int get initialState => super.initialState ?? 0;
}

v5.0.0

class CounterBloc extends HydratedBloc<CounterEvent, int> {
  CounterBloc() : super(0);


  ...
}