স্কিপ করে মূল কন্টেন্ট এ যাও

Provider vs Riverpod

This article recaps the defferences and the similarities between Provider and Riverpod.

Defining providers

The primary difference between both packages is how "providers" are defined.

With Provider, providers are widgets and as such placed inside the widget tree, typically inside a MultiProvider:

class Counter extends ChangeNotifier {
...
}

void main() {
runApp(
MultiProvider(
providers: [
ChangeNotifierProvider<Counter>(create: (context) => Counter()),
],
child: MyApp(),
)
);
}

With Riverpod, providers are not widgets. Instead they are plain Dart objects.
Similarly, providers are defined outside of the widget tree, and instead are declared as global final variables.

Also, for Riverpod to work, it is necessary to add a ProviderScope widget above the entire application. As such, the equivalent of the Provider example using Riverpod would be:

// Providers are now top-level variables
final counterProvider = ChangeNotifierProvider<Counter>((ref) => Counter());

void main() {
runApp(
// This widget enables Riverpod for the entire project
ProviderScope(
child: MyApp(),
),
);
}

Notice how the provider definition simply moved up a few lines.

info

Since with Riverpod providers are plain Dart objects, it is possible to use Riverpod without Flutter.
For example, Riverpod can be used to write command line applications.

Reading providers: BuildContext

With Provider, one way of reading providers is to use a Widget's BuildContext.

For example, if a provider was defined as:

Provider<Model>(...);

then reading it using Provider is done with:

class Example extends StatelessWidget {

Widget build(BuildContext context) {
Model model = context.watch<Model>();

}
}

The equivalent in Riverpod would be:

final modelProvider = Provider<Model>(...);

class Example extends ConsumerWidget {

Widget build(BuildContext context, WidgetRef ref) {
Model model = ref.watch(modelProvider);

}
}

Notice how:

  • Riverpod's snippet extends ConsumerWidget instead of StatelessWidget. That different widget type adds one extra parameter to our build function: WidgetRef.

  • Instead of BuildContext.watch, in Riverpod we do WidgetRef.watch, using the WidgetRef which we obtained from ConsumerWidget.

  • Riverpod does not rely on generic types. Instead it relies on the variable created using provider definition.

Notice too how similar the wording is. Both Provider and Riverpod use the keyword "watch" to describe "this widget should rebuild when the value changes".

info

Riverpod uses the same terminology as Provider for reading providers.

  • BuildContext.watch -> WidgetRef.watch
  • BuildContext.read -> WidgetRef.read
  • BuildContext.select -> WidgetRef.watch(myProvider.select)

The rules for context.watch vs context.read applies to Riverpod too:
Inside the build method, use "watch". Inside click handlers and other events, use "read". When in need of filtering out values and rebuilds, use "select".

Reading providers: Consumer

Provider optionally comes with a widget named Consumer (and variants such as Consumer2) for reading providers.

Consumer is helpful as a performance optimization, by allowing more granular rebuilds of the widget tree - updating only the revelant widgets when the state changes:

As such, if a provider was defined as:

Provider<Model>(...);

Provider allows reading that provider using Consumer with:

Consumer<Model>(
builder: (BuildContext context, Model model, Widget? child) {

}
)

Riverpod has the same principle. Riverpod, too, has a widget named Consumer for the exact same purpose.

If we defined a provider as:

final modelProvider = Provider<Model>(...);

Then using Consumer we could do:

Consumer(
builder: (BuildContext context, WidgetRef ref, Widget? child) {
Model model = ref.watch(modelProvider);

}
)

Notice how Consumer gives us a WidgetRef object. This is the same object as we saw in the previous part related to ConsumerWidget.

There is no ConsumerN equivalent in Riverpod

Notice how pkg:Provider's Consumer2, Consumer3 and such aren't needed nor missed in Riverpod.

With Riverpod, if you want to read values from multiple providers, you can simply write multiple ref.watch statements, like so:

Consumer(
builder: (context, ref, child) {
Model1 model = ref.watch(model1Provider);
Model2 model = ref.watch(model2Provider);
Model3 model = ref.watch(model3Provider);
// ...
}
)

When compared to pkg:Provider's ConsumerN APIs, the above solution feels way less heavy and it should be easier to understand.

Combining providers: ProxyProvider with stateless objects

When using Provider, the official way of combining providers is using the ProxyProvider widget (or variants such as ProxyProvider2).

For example we may define:

class UserIdNotifier extends ChangeNotifier {
String? userId;
}

// ...

ChangeNotifierProvider<UserIdNotifier>(create: (context) => UserIdNotifier()),

From there we have two options. We may combine UserIdNotifier to create a new "stateless" provider (typically an immutable value that possibly override ==). Such as:

ProxyProvider<UserIdNotifier, String>(
update: (context, userIdNotifier, _) {
return 'The user ID of the the user is ${userIdNotifier.userId}';
}
)

This provider would automatically return a new String whenever UserIdNotifier.userId changes.

We can do something similar in Riverpod, but the syntax is different.
First, in Riverpod, the definition of our UserIdNotifier would be:

class UserIdNotifier extends ChangeNotifier {
String? userId;
}

// ...

final userIdNotifierProvider = ChangeNotifierProvider<UserIdNotifier>(
(ref) => UserIdNotifier(),
);

From there, to generate our String based on the userId, we could do:

final labelProvider = Provider<String>((ref) {
UserIdNotifier userIdNotifier = ref.watch(userIdNotifierProvider);
return 'The user ID of the the user is ${userIdNotifier.userId}';
});

Notice the line doing ref.watch(userIdNotifierProvider).

This line of code tells Riverpod to obtain the content of the userIdNotifierProvider and that whenever that value changes, labelProvider will be recomputed too. As such, the String emitted by our labelProvider will automatically update whenever the userId changes.

This ref.watch line should feel similar. This pattern was covered previously when explaining how to read providers inside widgets. Indeed, providers are now able to listen to other providers in the same way that widgets do.

Combining providers: ProxyProvider with stateful objects

When combining providers, another alternative use-case is to expose stateful objects, such as a ChangeNotifier instance.

For that, we could use ChangeNotifierProxyProvider (or variants such as ChangeNotifierProxyProvider2).
For example we may define:

class UserIdNotifier extends ChangeNotifier {
String? userId;
}

// ...

ChangeNotifierProvider<UserIdNotifier>(create: (context) => UserIdNotifier()),

Then, we can define a new ChangeNotifier that is based on UserIdNotifier.userId. For example we could do:

class UserNotifier extends ChangeNotifier {
String? _userId;

void setUserId(String? userId) {
if (userId != _userId) {
print('The user ID changed from $_userId to $userId');
_userId = userId;
}
}
}

// ...

ChangeNotifierProxyProvider<UserIdNotifier, UserNotifier>(
create: (context) => UserNotifier(),
update: (context, userIdNotifier, userNotifier) {
return userNotifier!
..setUserId(userIdNotifier.userId);
},
);

This new provider creates a single instance of UserNotifier (which is never re-constructed) and prints a string whenever the user ID changes.

Doing the same thing in provider is achieved differently. First, in Riverpod, the definition of our UserIdNotifier would be:

class UserIdNotifier extends ChangeNotifier {
String? userId;
}

// ...

final userIdNotifierProvider = ChangeNotifierProvider<UserIdNotifier>(
(ref) => UserIdNotifier(),
),

From there, the equivalent to the previous ChangeNotifierProxyProvider would be:

class UserNotifier extends ChangeNotifier {
String? _userId;

void setUserId(String? userId) {
if (userId != _userId) {
print('The user ID changed from $_userId to $userId');
_userId = userId;
}
}
}

// ...

final userNotifierProvider = ChangeNotifierProvider<UserNotifier>((ref) {
final userNotifier = UserNotifier();
ref.listen<UserIdNotifier>(
userIdNotifierProvider,
(previous, next) {
if (previous?.userId != next.userId) {
userNotifier.setUserId(next.userId);
}
},
);

return userNotifier;
});

The core of this snippet is the ref.listen line.
This ref.listen function is a utility that allows listening to a provider, and whenever the provider changes, executes a function.

The previous and next parameters of that function correspond to the last value before the provider changed and the new value after it changed.

Scoping Providers vs .family + .autoDispose

In pkg:Provider, scoping was used for two things:

  • destroying state when leaving a page
  • having custom state per page

Using scoping just to destroy state isn't ideal.
The problem is that scoping doesn't work well over large applications.
For example, state often is created in one page, but destroyed later in a different page after navigation.
This doesn't allow for multiple caches to be active over different pages.

Similarly, the "custom state per page" approach quickly becomes difficult to handle if that state needs to be shared with another part of the widget tree, like you'd need with modals or a with a multi-step form.

Riverpod takes a different approach: first, scoping providers is kind-of discouraged; second, .family and .autoDispose are a complete replacement solution for this.

Within Riverpod, Providers marked as .autoDispose automatically destroy their state when they aren't used anymore.
When the last widget removing a provider is unmounted, Riverpod will detect this and destroy the provider.
Try using these two lifecycle methods in a provider to test this behavior:

ref.onCancel((){
print("No one listens to me anymore!");
});
ref.onDispose((){
print("If I've been defined as `.autoDispose`, I just got disposed!");
});

This inherently solves the "destroying state" problem.

Also it is possible to mark a Provider as .family (and, at the same time, as .autoDispose).
This enables passing parameters to providers, which make multiple providers to be spawned and tracked internally.
In other words, when passing parameters, a unique state is created per unique parameter.



int random(RandomRef ref, {required int seed, required int max}) {
return Random(seed).nextInt(max);
}

This solves the "custom state per page" problem. Actually, there's another advantage: such state is no-longer bound to one specific page.
Instead, if a different page tries to access the same state, such page will be able to do so by just reusing the parameters.

In many ways, passing parameters to providers is equivalent to a Map key.
If the key is the same, the value obtained is the same. If it's a different key, a different state will be obtained.