Android 4.4 APIs

API Level: 19

Android 4.4 (KITKAT) is a new release for the Android platform that offers new features for users and app developers. This document provides an introduction to the most notable new APIs.

As an app developer, you should download the Android 4.4 system image and SDK platform from the SDK Manager as soon as possible. If you don't have a device running Android 4.4 on which to test your app, use the Android 4.4 system image to test your app on the Android emulator. Then build your apps against the Android 4.4 platform to begin using the latest APIs.

Update your target API level

To better optimize your app for devices running Android 4.4, you should set your targetSdkVersion to "19", install it on an Android 4.4 system image, test it, then publish an update with this change.

You can use APIs in Android 4.4 while also supporting older versions by adding conditions to your code that check for the system API level before executing APIs not supported by your minSdkVersion. To learn more about maintaining backward compatibility, read Supporting Different Platform Versions.

For more information about how API levels work, read What is API Level?

Important Behavior Changes

If you have previously published an app for Android, be aware that your app might be affected by changes in Android 4.4.

If your app reads from external storage...

Your app can not read shared files on the external storage when running on Android 4.4, unless your app has the READ_EXTERNAL_STORAGE permission. That is, files within the directory returned by getExternalStoragePublicDirectory() are no longer accessible without the permission. However, if you need to access only your app-specific directories, provided by getExternalFilesDir(), then you do not need the READ_EXTERNAL_STORAGE permission.

If your app uses WebView...

Your app might behave differently when running on Android 4.4, especially when you update your app's targetSdkVersion to "19" or higher.

The code underlying the WebView class and related APIs has been upgraded to be based on a modern snapshot of the Chromium source code. This brings a variety of improvements for performance, support for new HTML5 features, and support for remote debugging of your WebView content. The scope of this upgrade means that if your app uses WebView, its behavior may be impacted in some cases. Although known behavior changes are documented and mostly affect your app only when you update your app's targetSdkVersion to "19" or higher—the new WebView operates in "quirks mode" to provide some legacy functionality in apps that target API level 18 and lower—it's possible that your app depends on unknown behaviors from the previous version of WebView.

So if your existing app uses WebView, it's important that you test on Android 4.4 as soon as possible and consult Migrating to WebView in Android 4.4 for information about how your app might be affected when you update your targetSdkVersion to "19" or higher.

If your app uses AlarmManager...

When you set your app's targetSdkVersion to "19" or higher, alarms that you create using either set() or setRepeating() will be inexact.

To improve power efficiency, Android now batches together alarms from all apps that occur at reasonably similar times so the system wakes the device once instead of several times to handle each alarm.

If your alarm is not associated with an exact clock time, but it's still important that your alarm be invoked during a specific time range (such as between 2pm and 4pm), then you can use the new setWindow() method, which accepts an "earliest" time for the alarm and a "window" of time following the earliest time within which the system should invoke the alarm.

If your alarm must be pinned to an exact clock time (such as for a calendar event reminder), then you can use the new setExact() method.

This inexact batching behavior applies only to updated apps. If you've set the targetSdkVersion to "18" or lower, your alarms will continue to behave as they have on previous versions when running on Android 4.4.

If your app syncs data using ContentResolver...

When you set your app's targetSdkVersion to "19" or higher, creating a sync with addPeriodicSync() will perform your sync operations within a default flex interval of approximately 4% of the period you specify. For example, if your poll frequency is 24 hours, then your sync operation may occur within roughly a one-hour window of time each day, instead of at exactly the same time each day.

To specify your own flex interval for sync operations, you should begin using the new requestSync() method. For more details, see the section below about Sync Adapters.

This flex interval behavior applies only to updated apps. If you've set the targetSdkVersion to "18" or lower, your existing sync requests will continue to behave as they have on previous versions when running on Android 4.4.

Printing Framework

Android now includes a complete framework that allows users to print any document using a printer connected over Wi-Fi, Bluetooth, or other services. The system handles the transaction between an app that wants to print a document and the services that deliver print jobs to a printer. The android.print framework provides all the APIs necessary to specify a print document and deliver it to the system for printing. Which APIs you actually need for a given print job depends on your content.

Printing generic content

If you want to print content from your UI as a document, you need to first create a subclass of PrintDocumentAdapter. Within this class, you must implement a few callback methods, including onLayout() to establish your layout based on the provided printing properties, and onWrite() to serialize your printable content into a ParcelFileDescriptor.

In order to write your content to the ParcelFileDescriptor you must pass it a PDF. The new PdfDocument APIs offer a convenient way to do this by providing a Canvas from getCanvas(), on which you can draw your printable content. Then write the PdfDocument to the ParcelFileDescriptor using the writeTo() method.

Once you've defined your implementation for PrintDocumentAdapter, you can execute print jobs upon the user's request using the PrintManager method, print(), which takes the PrintDocumentAdapter as one of its arguments.

Printing images

If you want to print just a photo or other bitmap, then the helper APIs in the support library do all the work for you. Simply create a new instance of PrintHelper, set the scale mode with setScaleMode(), then pass your Bitmap to printBitmap(). That's it. The library handles all the remaining interaction with the system to deliver the bitmap to the printer.

Building print services

As a printer OEM, you can use the android.printservice framework to provide interoperability with your printers from Android devices. You can build and distribute print services as APKs, which users can install on their devices . A print service app operates primarily as a headless service by subclassing the PrintService class, which receives print jobs from the system and communicates the jobs to its printers using the appropriate protocols.

For more information about how to print your app content, read Printing Content.

SMS Provider

The Telephony content provider (the "SMS Provider") allows apps to read and write SMS and MMS messages on the device. It includes tables for SMS and MMS messages received, drafted, sent, pending, and more.

Beginning with Android 4.4, the system settings allow users to select a "default SMS app." Once selected, only the default SMS app is able to write to the SMS Provider and only the default SMS app receives the SMS_DELIVER_ACTION broadcast when the user receives an SMS or the WAP_PUSH_DELIVER_ACTION broadcast when the user receives an MMS. The default SMS app is responsible for writing details to the SMS Provider when it receives or sends a new message.

Other apps that are not selected as the default SMS app can only read the SMS Provider, but may also be notified when a new SMS arrives by listening for the SMS_RECEIVED_ACTION broadcast, which is a non-abortable broadcast that may be delivered to multiple apps. This broadcast is intended for apps that---while not selected as the default SMS app---need to read special incoming messages such as to perform phone number verification.

For more information, read the blog post, Getting Your SMS Apps Ready for KitKat.

Wireless and Connectivity

Host card emulation

Android apps can now emulate ISO14443-4 (ISO-DEP) NFC cards that use APDUs for data exchange (as specified in ISO7816-4). This allows an NFC-enabled device running Android 4.4 to emulate multiple NFC cards at the same time, and allows an NFC payment terminal or other NFC reader to initiate a transaction with the appropriate NFC card based on the application identifier (AID).

If you want to emulate an NFC card that is using these protocols in your app, create a service component based on the HostApduService class. Whereas if your app instead uses a secure element for card emulation, you need to create a service based on the OffHostApduService class, which will not directly be involved in the transactions but is necessary to register the AIDs that should be handled by the secure element.

For more information, read the NFC Card Emulation guide.

NFC reader mode

A new NFC reader mode allows an activity to restrict all NFC activity to only reading the types of tags the activity is interested in while in the foreground. You can enable reader mode for your activity with enableReaderMode(), providing an implementation of NfcAdapter.ReaderCallback that receives a callback when new tags are detected.

This new capability, in conjunction with host card emulation, allows Android to operate on both ends of a mobile payment interface: One devices operates as the payment terminal (a device running a reader mode activity) and another device operates as the payment client (a device emulating an NFC card).

Infrared transmitters

When running on a device that includes an infrared (IR) transmitter, you can now transmit IR signals using the ConsumerIrManager APIs. To get an instance of ConsumerIrManager, call getSystemService() with CONSUMER_IR_SERVICE as the argument. You can then query the device's supported IR frequencies with getCarrierFrequencies() and transmit signals by passing your desired frequency and signal pattern with transmit().

You should always first check whether a device includes an IR transmitter by calling hasIrEmitter(), but if your app is compatible only with devices that do have one, you should include a <uses-feature> element in your manifest for "android.hardware.consumerir" (FEATURE_CONSUMER_IR).

Multimedia

Adaptive playback

Support for adaptive video playback is now available with the MediaCodec APIs, enabling seamless change in resolution during playback onto a Surface—you can feed the decoder input frames of a new resolution and the resolution of the output buffers change without a significant gap.

You can enable adaptive playback by adding two keys to MediaFormat that specify the maximum resolution that your app requires from the codec: KEY_MAX_WIDTH and KEY_MAX_HEIGHT. With these added to your MediaFormat, pass the MediaFormat to your MediaCodec instance with configure().

The codec will transition between resolutions that are the same as or smaller than these values in a seamless fashion. The codec may also support resolutions larger than the specified maximums (as long as it is within the limits of supported profiles), but transitions to larger resolutions may not be seamless.

To change the resolution while decoding H.264 video, continue to queue frames using MediaCodec.queueInputBuffer(), but be certain that you provide the new Sequence Parameter Set (SPS) and Picture Parameter Set (PPS) values together with the Instantaneous Decoder Refresh (IDR) frame in a single buffer.

However, before you attempt to configure your codec for adaptive playback, you must verify that the device supports adaptive playback by calling isFeatureSupported(String) with FEATURE_AdaptivePlayback.

Note: Support for adaptive playback is vendor specific. Some codecs may require more memory for larger resolution hints. Therefore, you should set the resolution maximums based on the source material you are decoding.

On-demand audio timestamps

To facilitate audio-video synchronization, the new AudioTimestamp class provides timeline details about a specific "frame" in an audio stream handled by AudioTrack. To get the most recent timestamp available, instantiate an AudioTimestamp object and pass it to getTimestamp(). If the request for the timestamp succeeds, the AudioTrack instance is filled in with a position in frame units, together with the estimated time when that frame either was presented or is committed to be presented.

You can use the value of nanoTime in the AudioTimestamp (which is monotonic) to find the closest associated video frame compared to framePosition so you can drop, duplicate, or interpolate video frames to match the audio. Alternatively, you can determine the delta time between the value of nanoTime and a future video frame’s expected time (with consideration for the sample rate) to predict which audio frame is expected at the same moment as a video frame.

Surface image reader

The new ImageReader API provides you direct access to image buffers as they are rendered into a Surface. You can acquire an ImageReader with the static method newInstance(). Then call getSurface() to create a new Surface and deliver your image data with a producer such as MediaPlayer or MediaCodec. To be notified when new images are available from the surface, implement the ImageReader.OnImageAvailableListener interface and register it with setOnImageAvailableListener().

Now as you draw content to your Surface, your ImageReader.OnImageAvailableListener receives a call to onImageAvailable() as each new image frame becomes available, providing you with the corresponding ImageReader. You can use the ImageReader to acquire the frame's image data as an Image object by calling acquireLatestImage() or acquireNextImage().

The Image object provides direct access to the image's timestamp, format, dimensions, and pixel data in a ByteBuffer. However, in order for the Image class to interpret your images, they must be formatted according to one of the types defined by constants in either ImageFormat or PixelFormat.

Peak and RMS measurement

You can now query the peak and RMS of the current audio stream from Visualizer by creating a new instance of Visualizer.MeasurementPeakRms and passing it to getMeasurementPeakRms(). When you call this method, the peak and RMS values of the given Visualizer.MeasurementPeakRms are set to the latest measured values.

Loudness enhancer

The LoudnessEnhancer is a new subclass of AudioEffect that allows you to increase the audible volume of your MediaPlayer or AudioTrack. This can be especially useful in conjunction with the new getMeasurementPeakRms() method mentioned above, in order to increase the volume of spoken audio tracks while other media is currently playing.

Remote controllers

Android 4.0 (API level 14) introduced the RemoteControlClient APIs that allow media apps to consume media controller events from remote clients such as media controls on the lock screen. Now the new RemoteController APIs allow you to build your own remote controller, enabling the creation of innovative new apps and peripherals that can control the playback of any media app that integrates with RemoteControlClient.

To build a remote controller, you can implement your user interface any way you want to, but to deliver the media button events to the user's media app you must create a service that extends the NotificationListenerService class and implements the RemoteController.OnClientUpdateListener interface. Using the NotificationListenerService as the basis is important because it provides the appropriate privacy restrictions, which require users to enable your app as a notification listener within the system security settings.

The NotificationListenerService class includes a couple abstract methods you must implement, but if you are only concerned with the media controller events for handling media playback, you can leave your implementation for those empty and instead focus on the RemoteController.OnClientUpdateListener methods.

Ratings from remote controllers

Android 4.4 builds upon the existing capabilities for remote control clients (apps that receive media control events with the RemoteControlClient) by adding the ability for users to rate the current track from the remote controller.

The new Rating class encapsulates information about a user rating. A rating is defined by its rating style (either RATING_HEART, RATING_THUMB_UP_DOWN, RATING_3_STARS, RATING_4_STARS, RATING_5_STARS or RATING_PERCENTAGE) and the rating value that's appropriate for that style.

To allow users to rate your tracks from a remote controller:

To receive a callback when the user changes the rating from the remote controller, implement the new RemoteControlClient.OnMetadataUpdateListener interface and pass an instance to setMetadataUpdateListener(). When the user changes the rating, your RemoteControlClient.OnMetadataUpdateListener receives a call to onMetadataUpdate(), passing RATING_KEY_BY_USER as the key and a Rating object as the value.

Closed captions

VideoView now supports WebVTT subtitle tracks when playing HTTP Live Stream (HLS) videos, displaying the subtitle track according to the closed caption preferences the user has defined in system settings.

You can also provide VideoView with your WebVTT subtitle tracks using the addSubtitleSource() method. This method accepts an InputStream that carries the subtitle data and a MediaFormat object that specifies the format for the subtitle data, which you can specify using createSubtitleFormat(). These subtitles also appear over the video according to the user's preferences.

If you do not use VideoView to display your video content, you should make your subtitle overlay match the user's closed captioning preference as closely as possible. A new CaptioningManager API allows you to query the user’s closed captioning preferences, including styles defined by CaptioningManager.CaptionStyle, such as typeface and color. In case the user adjusts some preferences once your video has already started, you should listen for changes to the preferences by registering an instance of CaptioningManager.CaptioningChangeListener to receive a callback when any of the preferences change, then update your subtitles as necessary.

Animation & Graphics

Scenes and transitions

The new android.transition framework provides APIs that facilitate animations between different states of your user interface. A key feature is the ability for you to define distinct states of your UI, known as "scenes," by creating a separate layout for each one. When you want to animate from one scene to another, execute a "transition," which calculates the necessary animation to change the layout from the current scene to the next scene.

To transition between two scenes, you generally need to perform the following:

  1. Specify the ViewGroup containing the UI components you want to change.
  2. Specify the layout representing the end-result of the change (the next scene).
  3. Specify the type of transition that should animate the layout change.
  4. Execute the transition.

You can use a Scene object to accomplish steps 1 and 2. A Scene contains metadata describing the properties of a layout that are necessary to perform a transition, including the scene's parent view and the scene's layout. You can create a Scene using a class constructor or the static method getSceneForLayout().

You must then use the TransitionManager to accomplish steps 3 and 4. One way is to pass your Scene to the static method go(). This finds the scene's parent view in the current layout and performs a transition on the child views in order to reach the layout defined by the Scene.

Alternatively, you don't need to create a Scene object at all, but can instead call beginDelayedTransition(), specifying a ViewGroup that contains the views you want to change. Then add, remove, or reconfigure the target views. After the system lays out the changes as necessary, a transition starts to animate all the affected views.

For additional control, you can define sets of transitions that should occur between pre-defined scenes, using an XML file in your project res/transition/ directory. Inside a <transitionManager> element, specify one or more <transition> tags that each specify a scene (a reference to a layout file) and the transition to apply when entering and/or exiting that scene. Then inflate this set of transitions using inflateTransitionManager(). Use the returned TransitionManager to execute each transition with transitionTo(), passing a Scene that is represented by one of the <transition> tags. You can also define sets of transitions programmatically with the TransitionManager APIs.

When specifying a transition, you can use several predefined types defined by subclasses of Transition, such as Fade and ChangeBounds. If you don't specify a transition type, the system uses AutoTransition by default, which automatically fades, moves, and resizes views as necessary. Additionally, you can create custom transitions by extending any of these classes to perform the animations however you'd like. A custom transition can track any property changes you'd like, and create any animation you want to based on those changes. For example, you could provide a subclass of Transition that listens for changes to the "rotation" property of a view then animate any changes.

For more information, see the TransitionManager documentation.

Animator pausing

The Animator APIs now allow you to pause and resume an ongoing animation with methods pause() and resume().

To track the state of an animation, you can implement the Animator.AnimatorPauseListener interface, which provides callbacks when an animation is paused and resumed: pause() and resume(). Then add the listener to an Animator object with addPauseListener().

Alternatively, you can subclass the AnimatorListenerAdapter abstract class, which now includes empty implementations for the pause and resume callbacks defined by Animator.AnimatorPauseListener.

Reusable bitmaps

You can now reuse any mutable bitmap in BitmapFactory to decode any other bitmap—even when the new bitmap is a different size---as long as the resulting byte count of the decoded bitmap (available from getByteCount()) is less than or equal to the allocated byte count of the reused bitmap (available from getAllocationByteCount(). For more information, see inBitmap.

New APIs for Bitmap allow similar reconfiguration for reuse outside of BitmapFactory (for manual bitmap generation or custom decoding logic). You can now set a bitmap’s dimensions with methods setHeight() and setWidth(), and specify specify a new Bitmap.Config with setConfig() without affecting the underlying bitmap allocation. The reconfigure() method also provides a convenient way to combine these changes with one call.

However, you should not reconfigure a bitmap that's currently used by the view system, because the underlying pixel buffer will not be remapped in a predictable way.

User Content

Storage access framework

On previous versions of Android, if you want your app to retrieve a specific type of file from another app, it must invoke an intent with the ACTION_GET_CONTENT action. This action is still the appropriate way to request a file that you want to import into your app. However, Android 4.4 introduces the ACTION_OPEN_DOCUMENT action, which allows the user to select a file of a specific type and grant your app long-term read access to that file (possibly with write access) without importing the file to your app.

If you're developing an app that provides storage services for files (such as a cloud save service), you can participate in this unified UI for picking files by implementing a content provider as a subclass of the new DocumentsProvider class. Your subclass of DocumentsProvider must include an intent filter that accepts the PROVIDER_INTERFACE action ("android.content.action.DOCUMENTS_PROVIDER"). You must then implement the four abstract methods in the DocumentsProvider:

queryRoots()
This must return a Cursor that describes all the root directories of your document storage, using columns defined in DocumentsContract.Root.
queryChildDocuments()
This must return a Cursor that describes all the files in the specified directory, using columns defined in DocumentsContract.Document.
queryDocument()
This must return a Cursor that describes the specified file, using columns defined in DocumentsContract.Document.
openDocument()
This must return a ParcelFileDescriptor representing the specified file. The system calls this method once the user selects a file and the client app requests access to it by calling openFileDescriptor().

For more information, see the Storage Access Framework guide.

External storage access

You can now read and write app-specific files on secondary external storage media, such as when a device provides both emulated storage and an SD card. The new method getExternalFilesDirs() works the same as the existing getExternalFilesDir() method except it returns an array of File objects. Before reading or writing to any of the paths returned by this method, pass the File object to the new getStorageState() method to verify the storage is currently available.

Other methods for accessing your app-specific cache directory and OBB directory also now have corresponding versions that provide access to secondary storage devices: getExternalCacheDirs() and getObbDirs(), respectively.

The first entry in the returned File array is considered the device's primary external storage, which is the same as the File returned by existing methods such as getExternalFilesDir().

Note: Beginning with Android 4.4, the platform no longer requires that your app acquire the WRITE_EXTERNAL_STORAGE or READ_EXTERNAL_STORAGE when you need to access only your app-specific regions of the external storage using the methods above. However, the permissions are required if you want to access the shareable regions of the external storage, provided by getExternalStoragePublicDirectory().

Sync adapters

The new requestSync() method in ContentResolver simplifies some of the procedure for defining a sync request for your ContentProvider by encapsulating requests in the new SyncRequest object, which you can create with SyncRequest.Builder. The properties in SyncRequest provide the same functionality as the existing ContentProvider sync calls, but adds the ability to specify that a sync should be dropped if the network is metered, by enabling setDisallowMetered().

User Input

New sensor types

The new TYPE_GEOMAGNETIC_ROTATION_VECTOR sensor provides rotation vector data based on a magnetometer, which is a useful alternative to the TYPE_ROTATION_VECTOR sensor when a gyroscope is not available or when used with batched sensor events to record the device's orientation while the phone is sleeping. This sensor requires less power than TYPE_ROTATION_VECTOR, but may be prone to noisy event data and is most effective while the user is outdoors.

Android also now supports built-in step sensors in hardware:

TYPE_STEP_DETECTOR
This sensor triggers an event each time the user takes a step. Upon each user step, this sensor delivers an event with a value of 1.0 and a timestamp indicating when the step occurred.
TYPE_STEP_COUNTER
This sensor also triggers an event upon each detected step, but instead delivers the total accumulated number of steps since this sensor was first registered by an app.

Be aware that these two step sensors don't always deliver the same results. The TYPE_STEP_COUNTER events occur with a higher latency than those from TYPE_STEP_DETECTOR, but that's because the TYPE_STEP_COUNTER algorithm does more processing to eliminate false positives. So the TYPE_STEP_COUNTER may be slower to deliver events, but its results should be more accurate.

Both step sensors are hardware dependent (Nexus 5 is the first device to support them), so you should check for availability with hasSystemFeature(), using the FEATURE_SENSOR_STEP_DETECTOR and FEATURE_SENSOR_STEP_COUNTER constants.

Batched sensor events

To better manage device power, the SensorManager APIs now allow you to specify the frequency at which you'd like the system to deliver batches of sensor events to your app. This doesn't reduce the number of actual sensor events available to your app for a given period of time, but instead reduces the frequency at which the system calls your SensorEventListener with sensor updates. That is, instead of delivering each event to your app the moment it occurs, the system saves up all the events that occur over a period of time, then delivers them to your app all at once.

To provide batching, the SensorManager class adds two new versions of the registerListener() method that allow you to specify the "maximum report latency." This new parameter specifies the maximum delay that your SensorEventListener will tolerate for delivery of new sensor events. For example, if you specify a batch latency of one minute, the system will deliver the recent set of batched events at an interval no longer than one minute by making consecutive calls to your onSensorChanged() method—once for each event that was batched. The sensor events will never be delayed longer than your maximum report latency value, but may arrive sooner if other apps have requested a shorter latency for the same sensor.

However, be aware that the sensor will deliver your app the batched events based on your report latency only while the CPU is awake. Although a hardware sensor that supports batching will continue to collect sensor events while the CPU is asleep, it will not wake the CPU to deliver your app the batched events. When the sensor eventually runs out of its memory for events, it will begin dropping the oldest events in order to save the newest events. You can avoid losing events by waking the device before the sensor fills its memory then call flush() to capture the latest batch of events. To estimate when the memory will be full and should be flushed, call getFifoMaxEventCount() to get the maximum number of sensor events it can save, and divide that number by the rate at which your app desires each event. Use that calculation to set wake alarms with AlarmManager that invoke your Service (which implements the SensorEventListener) to flush the sensor.

Note: Not all devices support batching sensor events because it requires support by the hardware sensor. However, beginning with Android 4.4, you should always use the new registerListener() methods, because if the device does not support batching, then the system gracefully ignores the batch latency argument and delivers sensor events in real time.

Controller identities

Android now identifies each connected controller with a unique integer that you can query with getControllerNumber(), making it easier for you to associate each controller to a different player in a game. The number for each controller may change due to controllers being disconnected, connected, or re-configured by the user, so you should track which controller number corresponds to each input device by registering an instance of InputManager.InputDeviceListener. Then call getControllerNumber() for each InputDevice when a change occurs.

Connected devices also now provide product and vendor IDs that are available from getProductId() and getVendorId(). If you need to modify your key mappings based on the available set of keys on a device, you can query the device to check whether certain keys are available with hasKeys(int...).

User Interface

Immersive full-screen mode

To provide your app with a layout that fills the entire screen, the new SYSTEM_UI_FLAG_IMMERSIVE flag for setSystemUiVisibility() (when combined with SYSTEM_UI_FLAG_HIDE_NAVIGATION) enables a new immersive full-screen mode. While immersive full-screen mode is enabled, your activity continues to receive all touch events. The user can reveal the system bars with an inward swipe along the region where the system bars normally appear. This clears the SYSTEM_UI_FLAG_HIDE_NAVIGATION flag (and the SYSTEM_UI_FLAG_FULLSCREEN flag, if applied) so the system bars remain visible. However, if you'd like the system bars to hide again after a few moments, you can instead use the SYSTEM_UI_FLAG_IMMERSIVE_STICKY flag.

Translucent system bars

You can now make the system bars partially translucent with new themes, Theme.Holo.NoActionBar.TranslucentDecor and Theme.Holo.Light.NoActionBar.TranslucentDecor. By enabling translucent system bars, your layout will fill the area behind the system bars, so you must also enable fitsSystemWindows for the portion of your layout that should not be covered by the system bars.

If you're creating a custom theme, set one of these themes as the parent theme or include the windowTranslucentNavigation and windowTranslucentStatus style properties in your theme.

Enhanced notification listener

Android 4.3 added the NotificationListenerService APIs, allowing apps to receive information about new notifications as they are posted by the system. In Android 4.4, notification listeners can retrieve additional metadata for the notification and complete details about the notification's actions:

The new Notification.extras field includes a Bundle to deliver your notification builder additional metadata such as EXTRA_TITLE and EXTRA_PICTURE. The new Notification.Action class defines the characteristics of an action attached to the notification, which you can retrieve from the new actions field.

Drawable mirroring for RTL layouts

On previous versions of Android, if your app includes images that should reverse their horizontal orientation for right-to-left layouts, you must include the mirrored image in a drawables-ldrtl/ resource directory. Now, the system can automatically mirror images for you by enabling the autoMirrored attribute on a drawable resource or by calling setAutoMirrored(). When enabled, the Drawable is automatically mirrored when the layout direction is right-to-left.

Accessibility

The View class now allows you to declare "live regions" for portions of your UI that dynamically update with new text content, by adding the new accessibilityLiveRegion attribute to your XML layout or calling setAccessibilityLiveRegion(). For example, a login screen with a text field that displays an "incorrect password" notification should be marked as a live region, so the screen reader will recite the message when it changes.

Apps that provide an accessibility service can now also enhance their capabilities with new APIs that provide information about view collections such as list or grid views using AccessibilityNodeInfo.CollectionInfo and AccessibilityNodeInfo.CollectionItemInfo.

App Permissions

The following are new permissions that your app must request with the <uses-permission> tag to use certain new APIs:

INSTALL_SHORTCUT
Allows an application to install a shortcut in Launcher
UNINSTALL_SHORTCUT
Allows an application to uninstall a shortcut in Launcher
TRANSMIT_IR
Allows an application to use the device's IR transmitter, if available

Note: Beginning with Android 4.4, the platform no longer requires that your app acquire the WRITE_EXTERNAL_STORAGE or READ_EXTERNAL_STORAGE when you want to access your app-specific regions of the external storage using methods such as getExternalFilesDir(). However, the permissions are still required if you want to access the shareable regions of the external storage, provided by getExternalStoragePublicDirectory().

Device Features

The following are new device features that you can declare with the <uses-feature> tag to declare your app requirements and enable filtering on Google Play or check for at runtime:

FEATURE_CONSUMER_IR
The device is capable of communicating with consumer IR devices.
FEATURE_DEVICE_ADMIN
The device supports device policy enforcement via device admins.
FEATURE_NFC_HOST_CARD_EMULATION
The device supports host- based NFC card emulation.
FEATURE_SENSOR_STEP_COUNTER
The device includes a hardware step counter.
FEATURE_SENSOR_STEP_DETECTOR
The device includes a hardware step detector.

For a detailed view of all API changes in Android 4.4, see the API Differences Report.