Activity Lifecycle
Activities are a fundamental building block of Android applications and they can exist in a number of different states. The activity lifecycle begins with instantiation and ends with destruction, and includes many states in between. When an activity changes state, the appropriate lifecycle event method is called, notifying the activity of the impending state change and allowing it to execute code to adapt to that change. This article examines the lifecycle of activities and explains the responsibility that an activity has during each of these state changes to be part of a well-behaved, reliable application.
Activity Lifecycle Overview
Activities are an unusual programming concept specific to Android. In traditional application development there is usually a static main method, which is executed to launch the application. With Android, however, things are different; Android applications can be launched via any registered activity within an application. In practice, most applications will only have a specific activity that is specified as the application entry point. However, if an application crashes, or is terminated by the OS, the OS can try to restart the application at the last open activity or anywhere else within the previous activity stack. Additionally, the OS may pause activities when they're not active, and reclaim them if it is low on memory. Careful consideration must be made to allow the application to correctly restore its state in the event that an activity is restarted, especially if that activity depends on data from previous activities.
The activity lifecycle is implemented as a collection of methods the OS calls throughout the lifecycle of an activity. These methods allow developers to implement the functionality that is necessary to satisfy the state and resource management requirements of their applications.
It is extremely important for the application developer to analyze the requirements of each activity to determine which methods exposed by the activity lifecycle need to be implemented. Failure to do this can result in application instability, crashes, resource bloat, and possibly even underlying OS instability.
This chapter examines the activity lifecycle in detail, including:
- Activity States
- Lifecycle Methods
- Retaining the State of an Application
This section also includes a walkthrough that provide practical examples on how to efficiently save state during the Activity lifecycle. By the end of this chapter you should have an understanding of the Activity lifecycle and how to support it in an Android application.
Activity Lifecycle
The Android activity lifecycle comprises a collection of methods exposed within the Activity class that provide the developer with a resource management framework. This framework allows developers to meet the unique state management requirements of each activity within an application and properly handle resource management.
Activity States
The Android OS arbitrates Activities based on their state. This helps Android identify activities that are no longer in use, allowing the OS to reclaim memory and resources. The following diagram illustrates the states an Activity can go through during its lifetime:
These states can be broken into 4 main groups as follows:
Active or Running – Activities are considered active or running if they are in the foreground, also known as the top of the activity stack. This is considered the highest priority activity in Android, and as such will only be killed by the OS in extreme situations, such as if the activity tries to use more memory than is available on the device as this could cause the UI to become unresponsive.
Paused – When the device goes to sleep, or an activity is still visible but partially hidden by a new, non-full-sized or transparent activity, the activity is considered paused. Paused activities are still alive, that is, they maintain all state and member information, and remain attached to the window manager. This is considered to be the second highest priority activity in Android and, as such, will only be killed by the OS if killing this activity will satisfy the resource requirements needed to keep the Active/Running Activity stable and responsive.
Stopped/Backgrounded – Activities that are completely obscured by another activity are considered stopped or in the background. Stopped activities still try to retain their state and member information for as long as possible, but stopped activities are considered to be the lowest priority of the three states and, as such, the OS will kill activities in this state first to satisfy the resource requirements of higher priority activities.
Restarted – It is possible for an activity that is anywhere from paused to stopped in the lifecycle to be removed from memory by Android. If the user navigates back to the activity it must be restarted, restored to its previously saved state, and then displayed to the user.
Activity Re-Creation in Response to Configuration Changes
To make matters more complicated, Android throws one more wrench in the mix called Configuration Changes. Configuration changes are rapid activity destruction/re-creation cycles that occur when the configuration of an activity changes, such as when the device is rotated (and the activity needs to get re-built in landscape or portrait mode), when the keyboard is displayed (and the activity is presented with an opportunity to resize itself), or when the device is placed in a dock, among others.
Configuration changes still cause the same Activity State changes that would occur during stopping and restarting an activity. However, in order to make sure that an application feels responsive and performs well during configuration changes, it's important that they are handled as quickly as possible. Because of this, Android has a specific API that can be used to persist state during configuration changes. We'll cover this later in the Managing State Throughout the Lifecycle section.
Activity Lifecycle Methods
The Android SDK and, by extension, the Xamarin.Android framework provide a powerful model for managing the state of activities within an application. When an activity's state is changing, the activity is notified by the OS, which calls specific methods on that activity. The following diagram illustrates these methods in relation to the Activity Lifecycle:
As a developer, you can handle state changes by overriding these methods within an activity. It's important to note, however, that all lifecycle methods are called on the UI thread and will block the OS from performing the next piece of UI work, such as hiding the current activity, displaying a new activity, etc. As such, code in these methods should be as brief as possible to make an application feel well performing. Any long-running tasks should be executed on a background thread.
Let's examine each of these lifecycle methods and their use:
OnCreate
OnCreate
is the first method to be called when an activity is created.
OnCreate
is always overridden to perform any startup initializations
that may be required by an Activity such as:
- Creating views
- Initializing variables
- Binding static data to lists
OnCreate
takes a
Bundle
parameter, which is a dictionary for storing and passing state
information and objects between activities If the bundle is not null,
this indicates the activity is restarting and it should restore its
state from the previous instance. The following code illustrates how to
retrieve values from the bundle:
protected override void OnCreate(Bundle bundle)
{
base.OnCreate(bundle);
string intentString;
bool intentBool;
if (bundle != null)
{
intentString = bundle.GetString("myString");
intentBool = bundle.GetBoolean("myBool");
}
// Set our view from the "main" layout resource
SetContentView(Resource.Layout.Main);
}
Once OnCreate
has finished, Android will call OnStart
.
OnStart
OnStart
is always called by the system after OnCreate
is finished. Activities
may override this method if they need to perform any specific tasks
right before an activity becomes visible such as refreshing current
values of views within the activity. Android will call OnResume
immediately after this method.
OnResume
The system calls OnResume when the Activity is ready to start interacting with the user. Activities should override this method to perform tasks such as:
- Ramping up frame rates (a common task in game development)
- Starting animations
- Listening for GPS updates
- Display any relevant alerts or dialogs
- Wire up external event handlers
As an example, the following code snippet shows how to initialize the camera:
protected override void OnResume()
{
base.OnResume(); // Always call the superclass first.
if (_camera==null)
{
// Do camera initializations here
}
}
OnResume
is important because any operation that is done in OnPause
should be un-done in OnResume
, since it's the only lifecycle method
that is guaranteed to execute after OnPause
when bringing the
activity back to life.
OnPause
OnPause is called when the system is about to put the activity into the background or when the activity becomes partially obscured. Activities should override this method if they need to:
Commit unsaved changes to persistent data
Destroy or clean up other objects consuming resources
Ramp down frame rates and pausing animations
Unregister external event handlers or notification handlers (i.e. those that are tied to a service). This must be done to prevent Activity memory leaks.
Likewise, if the Activity has displayed any dialogs or alerts, they must be cleaned up with the
.Dismiss()
method.
As an example, the following code snippet will release the camera, as the Activity cannot make use of it while paused:
protected override void OnPause()
{
base.OnPause(); // Always call the superclass first
// Release the camera as other activities might need it
if (_camera != null)
{
_camera.Release();
_camera = null;
}
}
There are two possible lifecycle methods that will be called after OnPause
:
OnResume
will be called if the Activity is to be returned to the foreground.OnStop
will be called if the Activity is being placed in the background.
OnStop
OnStop is called when the activity is no longer visible to the user. This happens when one of the following occurs:
- A new activity is being started and is covering up this activity.
- An existing activity is being brought to the foreground.
- The activity is being destroyed.
OnStop
may not always be called in low-memory situations, such as
when Android is starved for resources and cannot properly background
the Activity. For this reason, it is best not to rely on OnStop
getting called when preparing an Activity for destruction. The next
lifecycle methods that may be called after this one will be OnDestroy
if the Activity is going away, or OnRestart
if the Activity is coming
back to interact with the user.
OnDestroy
OnDestroy
is the final method that is called on an Activity instance before it's
destroyed and completely removed from memory. In extreme situations
Android may kill the application process that is hosting the Activity,
which will result in OnDestroy
not being invoked. Most Activities
will not implement this method because most clean up and shut down has
been done in the OnPause
and OnStop
methods. The OnDestroy
method
is typically overridden to clean up long running tasks that might
leak resources. An example of this might be background threads that
were started in OnCreate
.
There will be no lifecycle methods called after the Activity has been destroyed.
OnRestart
OnRestart
is called after your activity has been stopped, prior to it being
started again. A good example of this would be when the user presses
the home button while on an activity in the application. When this
happens OnPause
and then OnStop
methods are called, and the
Activity is moved to the background but is not destroyed. If the user
were then to restore the application by using the task manager or a
similar application, Android will call the OnRestart
method of the
activity.
There are no general guidelines for what kind of logic should be
implemented in OnRestart
. This is because OnStart
is always invoked
regardless of whether the Activity is being created or being restarted,
so any resources required by the Activity should be initialized in
OnStart
, rather than OnRestart
.
The next lifecycle method called after OnRestart
will be OnStart
.
Back vs. Home
Many Android devices have two distinct buttons: a "Back" button and a "Home" button. An example of this can be seen in the following screenshot of Android 4.0.3:
There is a subtle difference between the two buttons, even though they appear to have the same effect of putting an application in the background. When a user clicks the Back button, they are telling Android that they are done with the activity. Android will destroy the Activity. In contrast, when the user clicks the Home button the activity is merely placed into the background – Android will not kill the activity.
Managing State Throughout the Lifecycle
When an Activity is stopped or destroyed the system provides an opportunity to save the state of the Activity for later rehydration. This saved state is referred to as instance state. Android provides three options for storing instance state during the Activity lifecycle:
Storing primitive values in a
Dictionary
known as a Bundle that Android will use to save state.Creating a custom class that will hold complex values such as bitmaps. Android will use this custom class to save state.
Circumventing the configuration change lifecycle and assuming complete responsibility for maintaining state in the activity.
This guide covers the first two options.
Bundle State
The primary option for saving instance state is to use a key/value
dictionary object known as a
Bundle.
Recall that when an Activity is created that the OnCreate
method is
passed a bundle as a parameter, this bundle can be used to restore the
instance state. It is not recommended to use a bundle for more complex
data that won't quickly or easily serialize to key/value pairs (such as
bitmaps); rather, it should be used for simple values like strings.
An Activity provides methods to help with saving and retrieving the instance state in the Bundle:
OnSaveInstanceState – This is invoked by Android when the activity is being destroyed. Activities can implement this method if they need to persist any key/value state items.
OnRestoreInstanceState – This is called after the
OnCreate
method is finished, and provides another opportunity for an Activity to restore its state after initialization is complete.
The following diagram illustrates how these methods are used:
OnSaveInstanceState
OnSaveInstanceState
will be called as the Activity is being stopped. It will receive a
bundle parameter that the Activity can store its state in. When a
device experiences a configuration change, an Activity can use the
Bundle
object that is passed in to preserve the Activity state by
overriding OnSaveInstanceState
. For example, consider the following
code:
int c;
protected override void OnCreate (Bundle bundle)
{
base.OnCreate (bundle);
this.SetContentView (Resource.Layout.SimpleStateView);
var output = this.FindViewById<TextView> (Resource.Id.outputText);
if (bundle != null) {
c = bundle.GetInt ("counter", -1);
} else {
c = -1;
}
output.Text = c.ToString ();
var incrementCounter = this.FindViewById<Button> (Resource.Id.incrementCounter);
incrementCounter.Click += (s,e) => {
output.Text = (++c).ToString();
};
}
The code above increments an integer named c
when a button named
incrementCounter
is clicked, displaying the result in a TextView
named output
. When a configuration change happens - for example, when
the device is rotated - the above code would lose the value of c
because the bundle
would be null
, as shown in the figure below:
To preserve the value of c
in this example, the Activity can
override OnSaveInstanceState
, saving the value in the bundle as shown
below:
protected override void OnSaveInstanceState (Bundle outState)
{
outState.PutInt ("counter", c);
base.OnSaveInstanceState (outState);
}
Now when the device is rotated to a new orientation, the integer is saved in the bundle and is retrieved with the line:
c = bundle.GetInt ("counter", -1);
Note
It is important to always call the base implementation of
OnSaveInstanceState
so that the state of the view hierarchy can also be saved.
View State
Overriding OnSaveInstanceState
is an appropriate mechanism for saving
transient data in an Activity across orientation changes, such as the
counter in the above example. However, the default implementation of
OnSaveInstanceState
will take care of saving transient data in the UI
for every view, so long as each view has an ID assigned. For example,
say an application has an EditText
element defined in XML as follows:
<EditText android:id="@+id/myText"
android:layout_width="fill_parent"
android:layout_height="wrap_content"/>
Since the EditText
control has an id
assigned, when
the user enters some data and rotates the device, the data is still displayed,
as shown below:
OnRestoreInstanceState
OnRestoreInstanceState
will be called after OnStart
. It provides an activity the opportunity
to restore any state that was previously saved to a Bundle during the
previous OnSaveInstanceState
. This is the same bundle that is
provided to OnCreate
, however.
The following code demonstrates how state can be restored in
OnRestoreInstanceState
:
protected override void OnRestoreInstanceState(Bundle savedState)
{
base.OnRestoreInstanceState(savedState);
var myString = savedState.GetString("myString");
var myBool = savedState.GetBoolean("myBool");
}
This method exists to provide some flexibility around when state should
be restored. Sometimes it is more appropriate to wait until all
initializations are done before restoring instance state. Additionally,
a subclass of an existing Activity may only want to restore certain
values from the instance state. In many cases, it's not necessary to
override OnRestoreInstanceState
, since most activities can restore
state using the bundle provided to OnCreate
.
For an example of saving state using a Bundle
, refer to the
Walkthrough - Saving the Activity state.
Bundle Limitations
Although OnSaveInstanceState
makes it easy to save transient
data, it has some limitations:
It is not called in all cases. For example, pressing Home or Back to exit an Activity will not result in
OnSaveInstanceState
being called.The bundle passed into
OnSaveInstanceState
is not designed for large objects, such as images. In the case of large objects, saving the object from OnRetainNonConfigurationInstance is preferable, as discussed below.Data saved by using the bundle is serialized, which can lead to delays.
Bundle state is useful for simple data that doesn't use much memory,
whereas non-configuration instance data is useful for more complex
data, or data that is expensive to retrieve, such as from a web service
call or a complicated database query. Non-configuration instance data
gets saved in an object as needed. The next section introduces
OnRetainNonConfigurationInstance
as a way of preserving more complex
data types through configuration changes.
Persisting Complex Data
In addition to persisting data in the bundle, Android also supports
saving data by overriding
OnRetainNonConfigurationInstance
and returning an instance of a Java.Lang.Object
that contains the
data to persist. There are two primary benefits of using
OnRetainNonConfigurationInstance
to save state:
The object returned from
OnRetainNonConfigurationInstance
performs well with larger, more complex data types because memory retains this object.The
OnRetainNonConfigurationInstance
method is called on demand, and only when needed. This is more economical than using a manual cache.
Using OnRetainNonConfigurationInstance
is suitable for scenarios
where it is expensive to retrieve the data multiple times, such as in web
service calls. For example, consider the following code that searches
Twitter:
public class NonConfigInstanceActivity : ListActivity
{
protected override void OnCreate (Bundle bundle)
{
base.OnCreate (bundle);
SearchTwitter ("xamarin");
}
public void SearchTwitter (string text)
{
string searchUrl = String.Format("http://search.twitter.com/search.json?" + "q={0}&rpp=10&include_entities=false&" + "result_type=mixed", text);
var httpReq = (HttpWebRequest)HttpWebRequest.Create (new Uri (searchUrl));
httpReq.BeginGetResponse (new AsyncCallback (ResponseCallback), httpReq);
}
void ResponseCallback (IAsyncResult ar)
{
var httpReq = (HttpWebRequest)ar.AsyncState;
using (var httpRes = (HttpWebResponse)httpReq.EndGetResponse (ar)) {
ParseResults (httpRes);
}
}
void ParseResults (HttpWebResponse httpRes)
{
var s = httpRes.GetResponseStream ();
var j = (JsonObject)JsonObject.Load (s);
var results = (from result in (JsonArray)j ["results"] let jResult = result as JsonObject select jResult ["text"].ToString ()).ToArray ();
RunOnUiThread (() => {
PopulateTweetList (results);
});
}
void PopulateTweetList (string[] results)
{
ListAdapter = new ArrayAdapter<string> (this, Resource.Layout.ItemView, results);
}
}
This code retrieves results from the web formatted as JSON, parses them, and then presents the results in a list, as shown in the following screenshot:
When a configuration change occurs - for example, when a device is
rotated - the code repeats the process. To reuse the
originally retrieved results and not cause needless, redundant network
calls, we can use OnRetainNonconfigurationInstance
to save the
results, as shown below:
public class NonConfigInstanceActivity : ListActivity
{
TweetListWrapper _savedInstance;
protected override void OnCreate (Bundle bundle)
{
base.OnCreate (bundle);
var tweetsWrapper = LastNonConfigurationInstance as TweetListWrapper;
if (tweetsWrapper != null) {
PopulateTweetList (tweetsWrapper.Tweets);
} else {
SearchTwitter ("xamarin");
}
public override Java.Lang.Object OnRetainNonConfigurationInstance ()
{
base.OnRetainNonConfigurationInstance ();
return _savedInstance;
}
...
void PopulateTweetList (string[] results)
{
ListAdapter = new ArrayAdapter<string> (this, Resource.Layout.ItemView, results);
_savedInstance = new TweetListWrapper{Tweets=results};
}
}
Now when the device is rotated, the original results are retrieved from
the LastNonConfiguartionInstance
property. In this example, the
results consist of a string[]
containing tweets. Since
OnRetainNonConfigurationInstance
requires that a Java.Lang.Object
be returned, the string[]
is wrapped in a class that subclasses
Java.Lang.Object
, as shown below:
class TweetListWrapper : Java.Lang.Object
{
public string[] Tweets { get; set; }
}
For example, attempting to use a TextView
as the object returned from
OnRetainNonConfigurationInstance
will leak the Activity, as
illustrated by the code below:
TextView _textView;
protected override void OnCreate (Bundle bundle)
{
base.OnCreate (bundle);
var tv = LastNonConfigurationInstance as TextViewWrapper;
if(tv != null) {
_textView = tv;
var parent = _textView.Parent as FrameLayout;
parent.RemoveView(_textView);
} else {
_textView = new TextView (this);
_textView.Text = "This will leak.";
}
SetContentView (_textView);
}
public override Java.Lang.Object OnRetainNonConfigurationInstance ()
{
base.OnRetainNonConfigurationInstance ();
return _textView;
}
In this section, we learned how to preserve simple state data with the
Bundle
, and persist more complex data types with
OnRetainNonConfigurationInstance
.
Summary
The Android activity lifecycle provides a powerful framework for state management of activities within an application but it can be tricky to understand and implement. This chapter introduced the different states that an activity may go through during its lifetime, as well as the lifecycle methods that are associated with those states. Next, guidance was provided as to what kind of logic should be performed in each of these methods.