Learning Flutter’s new navigation and routing system
Note: The sample code in this article is not null-safe and not compatible with Dart 3. For more information on the latest best-practices for navigation and routing in Flutter, go to the Navigation and routing page on docs.flutter.dev.
This article explains how Flutter’s new Navigator
and Router
API works. If you follow Flutter’s open design docs, you might have seen these new features referred to as the Router widget. We’ll explore how these APIs enable more fine-tuned control over the screens in your app and how you can use it to parse routes.
These new APIs are not breaking changes, they simply add a new declarative API. Before Navigator 2.0, it was difficult to push or pop multiple pages, or remove a page underneath the current one. However, if you are happy with how the Navigator
works today, you can keep using it in the same (imperative) way.
The Router
provides the ability to handle routes from the underlying platform and display the appropriate pages. In this article, the Router
is configured to parse the browser URL to display the appropriate page.
This article helps you choose which Navigator
pattern works best for your app, and explains how to use Navigator 2.0 to parse browser URLs and take full control over the stack of pages that are active. The exercise in this article shows how to build an app that handles incoming routes from the platform and manages the pages of your app. The following GIF shows the example app in action:
Navigator 1.0
If you’re using Flutter, you’re probably using the Navigator
and are familiar with the following concepts:
Navigator
— a widget that manages a stack of Route objects.Route
— an object managed by aNavigator
that represents a screen, typically implemented by classes likeMaterialPageRoute
.
Before Navigator 2.0, Routes
were pushed and popped onto the Navigator
’s stack with either named routes or anonymous routes. The next sections are a brief recap of these two approaches.
Anonymous routes
Most mobile apps display screens on top of each other, like a stack. In Flutter, this is easy to achieve by using the Navigator
.
MaterialApp
and CupertinoApp
already use a Navigator
under the hood. You can access the navigator using Navigator.of()
or display a new screen using Navigator.push()
, and return to the previous screen with Navigator.pop()
:
When push()
is called, the DetailScreen
widget is placed on top of the HomeScreen
widget like this:
The previous screen (HomeScreen
) is still part of the widget tree, so any State
object associated with it stays around while DetailScreen
is visible.
Named routes
Flutter also supports named routes, which are defined in the routes
parameter on MaterialApp
or CupertinoApp
:
These routes must be predefined. Although you can pass arguments to a named route, you can’t parse arguments from the route itself. For example, if the app is run on the web, you can’t parse the ID from a route like /details/:id
.
Advanced named routes with onGenerateRoute
A more flexible way to handle named routes is by using onGenerateRoute
. This API gives you the ability to handle all paths:
Here’s the complete example:
Here, settings
is an instance of RouteSettings
. The name and arguments fields are the values that were provided when Navigator.pushNamed
was called, or what initialRoute
is set to.
Navigator 2.0
The Navigator 2.0 API adds new classes to the framework in order to make the app’s screens a function of the app state and to provide the ability to parse routes from the underlying platform (like web URLs). Here’s an overview of what’s new:
Page
— an immutable object used to set the navigator’s history stack.Router
— configures the list of pages to be displayed by the Navigator. Usually this list of pages changes based on the underlying platform, or on the state of the app changing.RouteInformationParser
, which takes theRouteInformation
fromRouteInformationProvider
and parses it into a user-defined data type.RouterDelegate
— defines app-specific behavior of how theRouter
learns about changes in app state and how it responds to them. Its job is to listen to theRouteInformationParser
and the app state and build theNavigator
with the current list ofPages
.BackButtonDispatcher
— reports back button presses to theRouter
.
The following diagram shows how the RouterDelegate
interacts with the Router
, RouteInformationParser
, and the app’s state:
Here’s an example of how these pieces interact:
- When the platform emits a new route (for example, “books/2”) , the
RouteInformationParser
converts it into an abstract data typeT
that you define in your app (for example, a class calledBooksRoutePath
). RouterDelegate
’ssetNewRoutePath
method is called with this data type, and must update the application state to reflect the change (for example, by setting theselectedBookId
) and callnotifyListeners.
- When
notifyListeners
is called, it tells theRouter
to rebuild theRouterDelegate
(using itsbuild()
method) RouterDelegate.build()
returns a newNavigator
, whose pages now reflect the change to the app state (for example, theselectedBookId
).
Navigator 2.0 exercise
This section leads you through an exercise using the Navigator 2.0 API. We’ll end up with an app that can stay in sync with the URL bar, and handle back button presses from the app and the browser, as shown in the following GIF:
To follow along, switch to the master channel, create a new Flutter project with web support, and replace the contents of lib/main.dart
with the following:
Pages
The Navigator has a new pages
argument in its constructor. If the list of Page
objects changes, Navigator
updates the stack of routes to match. To see how this works, we’ll build an app that displays a list of books.
In _BooksAppState
, keep two pieces of state: a list of books and the selected book:
Then in _BooksAppState
, return a Navigator
with a list of Page
objects:
Since this app has two screens, a list of books and a screen showing the details, add a second (detail) page if a book is selected (using collection if
):
Note that the key
for the page is defined by the value of the book
object. This tells the Navigator
that this MaterialPage
object is different from another when the Book
object is different. Without a unique key, the framework can’t determine when to show a transition animation between different Pages
.
Note: If you prefer, you can also extend Page
to customize the behavior. For example, this page adds a custom transition animation:
Finally, it’s an error to provide a pages
argument without also providing an onPopPage
callback. This function is called whenever Navigator.pop()
is called. It should be used to update the state (that determines the list of pages), and it must call didPop
on the route to determine if the pop succeeded:
It’s important to check whether didPop
fails before updating the app state.
Using setState
notifies the framework to call the build()
method, which returns a list with a single page when _selectedBook
is null.
Here’s the full example:
As it stands, this app only enables us to define the stack of pages in a declarative way. We aren’t able to handle the platform’s back button, and the browser’s URL doesn’t change as we navigate.
Router
So far, the app can show different pages, but it can’t handle routes from the underlying platform, for example if the user updates the URL in the browser.
This section shows how to implement the RouteInformationParser
, RouterDelegate
, and update the app state. Once set up, the app stays in sync with the browser’s URL.
Data types
The RouteInformationParser
parses the route information into a user-defined data type, so we’ll define that first:
In this app, all of the routes in the app can be represented using a single class. Instead, you might choose to use different classes that implement a superclass, or manage the route information in another way.
RouterDelegate
Next, add a class that extends RouterDelegate
:
The generic type defined on RouterDelegate
is BookRoutePath
, which contains all the state needed to decide which pages to show.
We’ll need to move some logic from _BooksAppState
to BookRouterDelegate
, and create a GlobalKey
. In this example, the app state is stored directly on the RouterDelegate
, but could also be separated into another class.
In order to show the correct path in the URL, we need to return a BookRoutePath
based on the current state of the app:
Next, the build
method in a RouterDelegate
needs to return a Navigator
:
The onPopPage
callback now uses notifyListeners
instead of setState
, since this class is now a ChangeNotifier
, not a widget. When the RouterDelegate
notifies its listeners, the Router
widget is likewise notified that the RouterDelegate's
currentConfiguration
has changed and that its build
method needs to be called again to build a new Navigator
.
The _handleBookTapped
method also needs to use notifyListeners
instead of setState
:
When a new route has been pushed to the application, Router
calls setNewRoutePath
, which gives our app the opportunity to update the app state based on the changes to the route:
RouteInformationParser
The RouteInformationParser
provides a hook to parse incoming routes (RouteInformation
) and convert it into a user defined type (BookRoutePath
). Use the Uri
class to take care of the parsing:
This implementation is specific to this app, not a general route parsing solution. More on that later.
To use these new classes, we use the new MaterialApp.router
constructor and pass in our custom implementations:
Here’s the complete example:
Running this sample in Chrome now shows the routes as they are being navigated, and navigates to the correct page when the URL is manually edited.
TransitionDelegate
You can provide a custom implementation of TransitionDelegate
that customizes how routes appear on (or are removed from) the screen when the list of pages changes. If you need to customize this, read on, but if you are happy with the default behavior you can skip this section.
Provide a custom TransitionDelegate
to a Navigator
that defines the desired behavior:
For example, the following implementation disables all transition animations:
This custom implementation overrides resolve(
), which is in charge of marking the various routes as either pushed, popped, added, completed, or removed:
markForPush
— displays the route with an animated transitionmarkForAdd
— displays the route without an animated transitionmarkForPop
— removes the route with an animated transition and completes it with a result. “Completing” in this context means that theresult
object is passed to theonPopPage
callback onAppRouterDelegate
.markForComplete
— removes the route without a transition and completes it with aresult
markForRemove
— removes the route with no animated transition and without completing.
This class only affects the declarative API, which is why the back button still displays a transition animation.
How this example works: This example looks at both the new routes and the routes that are exiting the screen. It goes through all of the objects in newPageRouteHistory
and marks them to be added without a transition animation using markForAdd
. Next, it loops through values of the locationToExitingPageRoute
map. If it finds a route marked as isWaitingForExitingDecision
, then it calls markForRemove
to indicate that the route should be removed without a transition and without completing.
Nested routers
This larger demo shows how to add a Router
within another Router
. Many apps require routes for the destinations in a BottomAppBar
, and routes for a stack of views above it, which requires two Navigators. To do this, the app uses an application state object to store app-specific navigation state (the selected menu index and the selected Book
object). This example also shows how to configure which Router
handles the back button.
What’s next
This article explored how to use these APIs for a specific app, but could also be used to build a higher-level API package. We hope that you’ll join us in exploring what a higher-level API built on top of these features can do for users.