Router<T> class Null safety

The dispatcher for opening and closing pages of an application.

This widget listens for routing information from the operating system (e.g. an initial route provided on app startup, a new route obtained when an intent is received, or a notification that the user hit the system back button), parses route information into data of type T, and then converts that data into Page objects that it passes to a Navigator.

Each part of this process can be overridden and configured as desired.

The routeInformationProvider can be overridden to change how the name of the route is obtained. The RouteInformationProvider.value is used as the initial route when the Router is first created. Subsequent notifications from the RouteInformationProvider to its listeners are treated as notifications that the route information has changed.

The backButtonDispatcher can be overridden to change how back button notifications are received. This must be a BackButtonDispatcher, which is an object where callbacks can be registered, and which can be chained so that back button presses are delegated to subsidiary routers. The callbacks are invoked to indicate that the user is trying to close the current route (by pressing the system back button); the Router ensures that when this callback is invoked, the message is passed to the routerDelegate and its result is provided back to the backButtonDispatcher. Some platforms don't have back buttons (e.g. iOS and desktop platforms); on those platforms this notification is never sent. Typically, the backButtonDispatcher for the root router is an instance of RootBackButtonDispatcher, which uses a WidgetsBindingObserver to listen to the popRoute notifications from SystemChannels.navigation. Nested Routers typically use a ChildBackButtonDispatcher, which must be provided the BackButtonDispatcher of its ancestor Router (available via Router.of).

The routeInformationParser can be overridden to change how names obtained from the routeInformationProvider are interpreted. It must implement the RouteInformationParser interface, specialized with the same type as the Router itself. This type, T, represents the data type that the routeInformationParser will generate.

The routerDelegate can be overridden to change how the output of the routeInformationParser is interpreted. It must implement the RouterDelegate interface, also specialized with T; it takes as input the data (of type T) from the routeInformationParser, and is responsible for providing a navigating widget to insert into the widget tree. The RouterDelegate interface is also Listenable; notifications are taken to mean that the Router needs to rebuild.

Concerns regarding asynchrony

Some of the APIs (notably those involving RouteInformationParser and RouterDelegate) are asynchronous.

When developing objects implementing these APIs, if the work can be done entirely synchronously, then consider using SynchronousFuture for the future returned from the relevant methods. This will allow the Router to proceed in a completely synchronous way, which removes a number of complications.

Using asynchronous computation is entirely reasonable, however, and the API is designed to support it. For example, maybe a set of images need to be loaded before a route can be shown; waiting for those images to be loaded before RouterDelegate.setNewRoutePath returns is a reasonable approach to handle this case.

If an asynchronous operation is ongoing when a new one is to be started, the precise behavior will depend on the exact circumstances, as follows:

If the active operation is a routeInformationParser parsing a new route information: that operation's result, if it ever completes, will be discarded.

If the active operation is a routerDelegate handling a pop request: the previous pop is immediately completed with "false", claiming that the previous pop was not handled (this may cause the application to close).

If the active operation is a routerDelegate handling an initial route or a pushed route, the result depends on the new operation. If the new operation is a pop request, then the original operation's result, if it ever completes, will be discarded. If the new operation is a push request, however, the routeInformationParser will be requested to start the parsing, and only if that finishes before the original routerDelegate request completes will that original request's result be discarded.

If the identity of the Router widget's delegates change while an asynchronous operation is in progress, to keep matters simple, all active asynchronous operations will have their results discarded. It is generally considered unusual for these delegates to change during the lifetime of the Router.

If the Router itself is disposed while an asynchronous operation is in progress, all active asynchronous operations will have their results discarded also.

No explicit signals are provided to the routeInformationParser or routerDelegate to indicate when any of the above happens, so it is strongly recommended that RouteInformationParser and RouterDelegate implementations not perform extensive computation.

Application architectural design

An application can have zero, one, or many Router widgets, depending on its needs.

An application might have no Router widgets if it has only one "screen", or if the facilities provided by Navigator are sufficient. This is common for desktop applications, where subsidiary "screens" are represented using different windows rather than changing the active interface.

A particularly elaborate application might have multiple Router widgets, in a tree configuration, with the first handling the entire route parsing and making the result available for routers in the subtree. The routers in the subtree do not participate in route information parsing but merely take the result from the first router to build their sub routes.

Most applications only need a single Router.

URL updates for web applications

In the web platform, keeping the URL in the browser's location bar up to date with the application state ensures that the browser constructs its history entry correctly, allowing its back and forward buttons to function as the user expects.

If an app state change leads to the Router rebuilding, the Router will retrieve the new route information from the routerDelegate's RouterDelegate.currentConfiguration method and the routeInformationParser's RouteInformationParser.restoreRouteInformation method.

If the location in the new route information is different from the current location, this is considered to be a navigation event, the PlatformRouteInformationProvider.routerReportsNewRouteInformation method calls SystemNavigator.routeInformationUpdated with replace = false to notify the engine, and through that the browser, to create a history entry with the new url. Otherwise, PlatformRouteInformationProvider.routerReportsNewRouteInformation calls SystemNavigator.routeInformationUpdated with replace = true to update the current history entry with the latest RouteInformation.

One can force the Router to report new route information as navigation event to the routeInformationProvider (and thus the browser) even if the RouteInformation.location has not changed by calling the Router.navigate method with a callback that performs the state change. This causes Router to call the RouteInformationProvider.routerReportsNewRouteInformation with RouteInformationReportingType.navigate, and thus causes PlatformRouteInformationProvider to push a new history entry regardlessly. This allows one to support the browser's back and forward buttons without changing the URL. For example, the scroll position of a scroll view may be saved in the RouteInformation.state. Using Router.navigate to update the scroll position causes the browser to create a new history entry with the RouteInformation.state that stores this new scroll position. When the user clicks the back button, the app will go back to the previous scroll position without changing the URL in the location bar.

One can also force the Router to ignore a navigation event by making those changes during a callback passed to Router.neglect. The Router calls the RouteInformationProvider.routerReportsNewRouteInformation with RouteInformationReportingType.neglect, and thus causes PlatformRouteInformationProvider to replace the current history entry regardlessly even if it detects location change.

To opt out of URL updates entirely, pass null for routeInformationProvider and routeInformationParser. This is not recommended in general, but may be appropriate in the following cases:

• The application does not target the web platform.

• There are multiple router widgets in the application. Only one Router widget should update the URL (typically the top-most one created by the WidgetsApp.router, MaterialApp.router, or CupertinoApp.router).

• The application does not need to implement in-app navigation using the browser's back and forward buttons.

In other cases, it is strongly recommended to implement the RouterDelegate.currentConfiguration and RouteInformationParser.restoreRouteInformation APIs to provide an optimal user experience when running on the web platform.

State Restoration

The Router will restore the current configuration of the routerDelegate during state restoration if it is configured with a restorationScopeId and state restoration is enabled for the subtree. For that, the value of RouterDelegate.currentConfiguration is serialized and persisted before the app is killed by the operating system. After the app is restarted, the value is deserialized and passed back to the RouterDelegate via a call to RouterDelegate.setRestoredRoutePath (which by default just calls RouterDelegate.setNewRoutePath). It is the responsibility of the RouterDelegate to use the configuration information provided to restore its internal state.

To serialize RouterDelegate.currentConfiguration and to deserialize it again, the Router calls RouteInformationParser.restoreRouteInformation and RouteInformationParser.parseRouteInformation, respectively. Therefore, if a restorationScopeId is provided, a routeInformationParser must be configured as well.