1. Getting Started

This is how you use Cling:

package ...;

import org.teleal.cling.model.message.header.STAllHeader;
import org.teleal.cling.model.meta.LocalDevice;
import org.teleal.cling.model.meta.RemoteDevice;
import org.teleal.cling.registry.Registry;
import org.teleal.cling.registry.RegistryListener;

/**
 * Runs a simple UPnP discovery procedure.
 */
public class Main {

    public static void main(String[] args) throws Exception {

        // UPnP discovery is asynchronous, we need a callback
        RegistryListener listener = new RegistryListener() {

            public void remoteDeviceDiscoveryStarted(Registry registry,
                                                     RemoteDevice device) {
                System.out.println(
                        "Discovery started: " + device.getDisplayString()
                );
            }

            public void remoteDeviceDiscoveryFailed(Registry registry,
                                                    RemoteDevice device,
                                                    Exception ex) {
                System.out.println(
                        "Discovery failed: " + device.getDisplayString() + " => " + ex
                );
            }

            public void remoteDeviceAdded(Registry registry, RemoteDevice device) {
                System.out.println(
                        "Remote device available: " + device.getDisplayString()
                );
            }

            public void remoteDeviceUpdated(Registry registry, RemoteDevice device) {
                System.out.println(
                        "Remote device updated: " + device.getDisplayString()
                );
            }

            public void remoteDeviceRemoved(Registry registry, RemoteDevice device) {
                System.out.println(
                        "Remote device removed: " + device.getDisplayString()
                );
            }

            public void localDeviceAdded(Registry registry, LocalDevice device) {
                System.out.println(
                        "Local device added: " + device.getDisplayString()
                );
            }

            public void localDeviceRemoved(Registry registry, LocalDevice device) {
                System.out.println(
                        "Local device removed: " + device.getDisplayString()
                );
            }

            public void beforeShutdown(Registry registry) {
                System.out.println(
                        "Before shutdown, the registry has devices: "
                        + registry.getDevices().size()
                );
            }

            public void afterShutdown() {
                System.out.println("Shutdown of registry complete!");

            }
        };

        // This will create necessary network resources for UPnP right away
        System.out.println("Starting Cling...");
        UpnpService upnpService = new UpnpServiceImpl(listener);

        // Send a search message to all devices and services, they should respond soon
        upnpService.getControlPoint().search(new STAllHeader());

        // Let's wait 10 seconds for them to respond
        System.out.println("Waiting 10 seconds before shutting down...");
        Thread.sleep(10000);

        // Release all resources and advertise BYEBYE to other UPnP devices
        System.out.println("Stopping Cling...");
        upnpService.shutdown();
    }
}

You need cling-core.jar and teleal-common.jar on your classpath to build and run this code.

2. A first UPnP service and control point

The most basic UPnP service imaginable is the binary light. This device has one service, the power switch, turning the light on and off. In fact, the SwitchPower:1 service and the BinaryLight:1 device are standardized templates you can download here.

In the following sections we'll implement this UPnP service and device with the Cling Core library as a simple Java console application.

2.1. The SwitchPower service implementation

This is the source of the SwitchPower:1 service - note that although there are many annotations in the source, no runtime dependency on Cling exists:

package example.binarylight;

import org.teleal.cling.binding.annotations.*;

@UpnpService(
        serviceId = @UpnpServiceId("SwitchPower"),
        serviceType = @UpnpServiceType(value = "SwitchPower", version = 1)
)
public class SwitchPower {

    @UpnpStateVariable(defaultValue = "0", sendEvents = false)
    private boolean target = false;

    @UpnpStateVariable(defaultValue = "0")
    private boolean status = false;

    @UpnpAction
    public void setTarget(@UpnpInputArgument(name = "NewTargetValue")
                          boolean newTargetValue) {
        target = newTargetValue;
        status = newTargetValue;
        System.out.println("Switch is: " + status);
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "RetTargetValue"))
    public boolean getTarget() {
        return target;
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "ResultStatus"))
    public boolean getStatus() {
        return status;
    }

}

To compile this class the Cling Core library has to be available on your classpath. However, once compiled this class can be instantiated and executed in any environment, there are no dependencies on any framework or library code.

The annotations are used by Cling to read the metadata that describes your service, what UPnP state variables it has, how they are accessed, and what methods should be exposed as UPnP actions. You can also provide Cling metadata in an XML file or programmatically through Java code - both options are discussed later in this manual. Source code annotations are usually the best choice.

You might have expected something even simpler: After all, a binary light only needs a single boolean state, it is either on or off. The designers of this service also considered that there might be a difference between switching the light on, and actually seeing the result of that action. Imagine what happens if the light bulb is broken: The target state of the light is set to true but the status is still false, because the SetTarget action could not make the switch. Obviously this won't be a problem with this simple demonstration because it only prints the status to standard console output.

2.2. Binding a UPnP device

Devices (and embedded devices) are created programmatically in Cling, with plain Java code that instantiates an immutable graph of objects. The following method creates such a device graph and binds the service from the previous section to the root device:

LocalDevice createDevice()
        throws ValidationException, LocalServiceBindingException, IOException {

    DeviceIdentity identity =
            new DeviceIdentity(
                    UDN.uniqueSystemIdentifier("Demo Binary Light")
            );

    DeviceType type =
            new UDADeviceType("BinaryLight", 1);

    DeviceDetails details =
            new DeviceDetails(
                    "Friendly Binary Light",
                    new ManufacturerDetails("ACME"),
                    new ModelDetails(
                            "BinLight2000",
                            "A demo light with on/off switch.",
                            "v1"
                    )
            );

    Icon icon =
            new Icon(
                    "image/png", 48, 48, 8,
                    getClass().getResource("icon.png")
            );

    LocalService<SwitchPower> switchPowerService =
            new AnnotationLocalServiceBinder().read(SwitchPower.class);

    switchPowerService.setManager(
            new DefaultServiceManager(switchPowerService, SwitchPower.class)
    );

    return new LocalDevice(identity, type, details, icon, switchPowerService);

    /* Several services can be bound to the same device:
    return new LocalDevice(
            identity, type, details, icon,
            new LocalService[] {switchPowerService, myOtherService}
    );
    */
    
}

Let's step through this code. As you can see, all arguments that make up the device's metadata have to be provided through constructors, because the metadata classes are immutable and hence thread-safe.

DeviceIdentity

Every device, no matter if it is a root device or an embedded device of a root device, requires a unique device name (UDN). This UDN should be stable, that is, it should not change when the device is restarted. When you physically unplug a UPnP appliance from the network (or when you simply turn it off or put it into standby mode), and when you make it available later on, it should expose the same UDN so that clients know they are dealing with the same device. The UDN.uniqueSystemIdentifier() method provides exactly that: A unique identifier that is the same every time this method is called on the same computer system. It hashes the network cards hardware address and a few other elements to guarantee uniqueness and stability.

DeviceType

The type of a device also includes its version, a plain integer. In this case the BinaryLight:1 is a standardized device template which adheres to the UDA (UPnP Device Architecture) specification.

DeviceDetails

This detailed information about the device's "friendly name", as well as model and manufacturer information is optional. You should at least provide a friendly name value, this is what UPnP applications will display primarily.

Icon

Every device can have a bunch of icons associated with it which similar to the friendly name are shown to users when appropriate. You do not have to provide any icons if you don't want to, use a constructor of LocalDevice without an icon parameter.

Service

Finally, the most important part of the device are its services. Each Service instance encapsulates the metadata for a particular service, what actions and state variables it has, and how it can be invoked. Here we use the Cling annotation binder to instantiate a Service, reading the annotation metadata of the SwitchPower class.

Because a Service instance is only metadata that describes the service, you have to set a ServiceManager to do some actual work. This is the link between the metadata and your implementation of a service, where the rubber meets the road. The DefaultServiceManager will instantiate the given SwitchPower class when an action which operates on the service has to be executed (this happens lazily, as late as possible). The manager will hold on to the instance and always re-use it as long as the service is registered with the UPnP stack. In other words, the service manager is the factory that instantiates your actual implementation of a UPnP service.

Also note that LocalDevice is the interface that represents a UPnP device which is "local" to the running UPnP stack on the host. Any device that has been discovered through the network will be a RemoteDevice with RemoteService's, you typically do not instantiate these directly.

A ValidationException will be thrown when the device graph you instantiated was invaild, you can call getErrors() on the exception to find out which property value of which class failed an integrity rule. The local service annotation binder will provide a LocalServiceBindingException if something is wrong with your annotation metadata on your service implementation class. An IOException can only by thrown by this particular Icon constructor, when it reads the resource file.

2.3. Running the server

The Cling Core main API entry point is a thread-safe and typically single shared instance of UpnpService:

package example.binarylight;

import org.teleal.cling.UpnpService;
import org.teleal.cling.UpnpServiceImpl;
import org.teleal.cling.binding.*;
import org.teleal.cling.binding.annotations.*;
import org.teleal.cling.model.*;
import org.teleal.cling.model.meta.*;
import org.teleal.cling.model.types.*;

import java.io.IOException;

public class BinaryLightServer implements Runnable {

    public static void main(String[] args) throws Exception {
        // Start a user thread that runs the UPnP stack
        Thread serverThread = new Thread(new BinaryLightServer());
        serverThread.setDaemon(false);
        serverThread.start();
    }

    public void run() {
        try {

            final UpnpService upnpService = new UpnpServiceImpl();

            Runtime.getRuntime().addShutdownHook(new Thread() {
                @Override
                public void run() {
                    upnpService.shutdown();
                }
            });

            // Add the bound local device to the registry
            upnpService.getRegistry().addDevice(
                    createDevice()
            );

        } catch (Exception ex) {
            System.err.println("Exception occured: " + ex);
            ex.printStackTrace(System.err);
            System.exit(1);
        }
    }

}

(The createDevice() method from the previous section should be added to this class.)

As soon as the UPnPServiceImpl is created, the stack is up and running. You always have to create a UPnPService instance, no matter if you write a client or a server. The UpnpService maintains a registry of all the discovered remote device on the network, and all the bound local devices. It manages advertisements for discovery and event handling in the background.

You should shut down the UPnP service properly when your application quits, so that all other UPnP systems on your network will be notified that bound devices which are local to your application are no longer available. If you do not shut down the UpnpService when your application quits, other UPnP control points on your network might still show devices as available when they are in fact already gone.

The createDevice() method from the previous section is called here, as soon as the Registry of the local UPnP service stack is available.

You can now compile and start this server, it should print some informational messages to your console and then wait for connections from UPnP control points. Use the Cling Workbench if you want to test your server immediately.

2.4. Creating a control point

The client application has the same basic scaffolding as the server, it also uses a shared single instance of UpnpService:

package example.binarylight;

import org.teleal.cling.UpnpService;
import org.teleal.cling.UpnpServiceImpl;
import org.teleal.cling.controlpoint.*;
import org.teleal.cling.model.action.*;
import org.teleal.cling.model.message.*;
import org.teleal.cling.model.message.header.*;
import org.teleal.cling.model.meta.*;
import org.teleal.cling.model.types.*;
import org.teleal.cling.registry.*;

public class BinaryLightClient implements Runnable {

    public static void main(String[] args) throws Exception {
        // Start a user thread that runs the UPnP stack
        Thread clientThread = new Thread(new BinaryLightClient());
        clientThread.setDaemon(false);
        clientThread.start();

    }

    public void run() {
        try {

            UpnpService upnpService = new UpnpServiceImpl();

            // Add a listener for device registration events
            upnpService.getRegistry().addListener(
                    createRegistryListener(upnpService)
            );

            // Broadcast a search message for all devices
            upnpService.getControlPoint().search(
                    new STAllHeader()
            );

        } catch (Exception ex) {
            System.err.println("Exception occured: " + ex);
            System.exit(1);
        }
    }

}

Typically a control point sleeps until a device with a specific type of service becomes available on the network. The RegistryListener is called by Cling when a remote device has been discovered - or when it announced itself automatically. Because you usually do not want to wait for the periodic announcements of devices, a control point can also execute a search for all devices (or devices with certain service types or UDN), which will trigger an immediate discovery announcement from those devices that match the search query.

You can already see the ControlPoint API here with its search(...) method, this is one of the main interfaces you interact with when writing a UPnP client with Cling.

If you compare this code with the server code from the previous section you can see that we are not shutting down the UpnpService when the application quits. This is not an issue here, because this application does not have any local devices or service event listeners (not the same as registry listeners) bound and registered. Hence, we do not have to announce their departure on application shutdown and can keep the code simple for the sake of the example.

Let's focus on the registry listener implementation and what happens when a UPnP device has been discovered on the network.

2.5. Executing an action

The control point we are creating here is only interested in services that implement SwitchPower. According to its template definition this service has the SwitchPower service identifier, so when a device has been discovered we can check if it offers that service:

RegistryListener createRegistryListener(final UpnpService upnpService) {
    return new DefaultRegistryListener() {

        ServiceId serviceId = new UDAServiceId("SwitchPower");

        @Override
        public void remoteDeviceAdded(Registry registry, RemoteDevice device) {

            Service switchPower;
            if ((switchPower = device.findService(serviceId)) != null) {

                System.out.println("Service discovered: " + switchPower);
                executeAction(upnpService, switchPower);

            }

        }

        @Override
        public void remoteDeviceRemoved(Registry registry, RemoteDevice device) {
            Service switchPower;
            if ((switchPower = device.findService(serviceId)) != null) {
                System.out.println("Service disappeared: " + switchPower);
            }
        }

    };
}

If a service becomes available we immediately execute an action on that service. When a SwitchPower device disappears from the network a log message is printed. Remember that this is a very trivial control point, it executes a single a fire-and-forget operation when a service becomes available:

void executeAction(UpnpService upnpService, Service switchPowerService) {

        ActionInvocation setTargetInvocation =
                new SetTargetActionInvocation(switchPowerService);

        // Executes asynchronous in the background
        upnpService.getControlPoint().execute(
                new ActionCallback(setTargetInvocation) {

                    @Override
                    public void success(ActionInvocation invocation) {
                        assert invocation.getOutput().length == 0;
                        System.out.println("Successfully called action!");
                    }

                    @Override
                    public void failure(ActionInvocation invocation,
                                        UpnpResponse operation,
                                        String defaultMsg) {
                        System.err.println(defaultMsg);
                    }
                }
        );

}

class SetTargetActionInvocation extends ActionInvocation {

    SetTargetActionInvocation(Service service) {
        super(service.getAction("SetTarget"));
        try {

            // Throws InvalidValueException if the value is of wrong type
            setInput("NewTargetValue", true);

        } catch (InvalidValueException ex) {
            System.err.println(ex.getMessage());
            System.exit(1);
        }
    }
}

The Action (metadata) and the ActionInvocation (actual call data) APIs allow very fine-grained control of how an invocation is prepared, how input values are set, how the action is executed, and how the output and outcome is handled. UPnP is inherently asynchronous, so just like the registry listener, executing an action is exposed to you as a callback-style API.

It is recommended that you encapsulate specific action invocations within a subclass of ActionInvocation, which gives you an opportunity to further abstract the input and output values of an invocation. Note however that an instance of ActionInvocation is not thread-safe and should not be executed in parallel by two threads.

The ActionCallback has two main methods you have to implement, one is called when the execution was successful, the other when it failed. There are many reasons why an action execution might fail, read the API documentation for all possible combinations or just print the generated user-friendly default error message.

2.6. Starting the application

Compile the binary light demo application:

$ javac -cp /path/to/teleal-common.jar:/path/to/cling-core.jar \
        -d classes/ \
        src/example/binarylight/BinaryLightServer.java \
        src/example/binarylight/BinaryLightClient.java \
        src/example/binarylight/SwitchPower.java

Don't forget to copy your icon.png file into the classes output directory as well, into the right package from which it is loaded as a reasource (the example.binarylight package if you followed the previous sections verbatim).

You can start the server or client first, which one doesn't matter as they will discover each other automatically:

$ java -cp /path/to/teleal-common.jar:/path/to/cling-core.jar:classes/ \
        example.binaryLight.BinaryLightServer
$ java -cp /path/to/teleal-common.jar:/path/to/cling-core.jar:classes/ \
        example.binaryLight.BinaryLightClient

You should see discovery and action execution messages on each console. You can stop and restart the applications individually (press CTRL+C on the console).

2.7. Debugging and logging

Although the binary light is a very simple example, you might run into problems. Cling Core helps you resolve most problems with extensive logging. Internally, Cling Core uses Java JDK logging, also known as java.util.logging or JUL. There are no wrappers, logging frameworks, logging services, or other dependencies.

By default, the implementation of JUL in the Sun JDK will print only messages with level INFO, WARNING, or SEVERE on System.out, and it will print each message over two lines. This is quite inconvenient and ugly, so your first step is probably to configure one line per message. This requires a custom logging handler.

Next you want to configure logging levels for different logging categories. Cling Core will output some INFO level messages on startup and shutdown, but is otherwise silent during runtime unless a problem occurs - it will then log messages at WARNING or SEVERE level.

For debugging, usually more detailed logging levels for various log categories are required. The logging categories in Cling Core are package names, e.g the root logger is available under the name org.teleal.cling. The following tables show typically used categories and the recommended level for debugging:

Network/Transport 
org.teleal.cling.transport.spi.DatagramIO (FINE)
org.teleal.cling.transport.spi.MulticastReceiver (FINE)
UDP communication
org.teleal.cling.transport.spi.DatagramProcessor (FINER)
UDP datagram processing and content
org.teleal.cling.transport.spi.UpnpStream (FINER)
org.teleal.cling.transport.spi.StreamServer (FINE)
org.teleal.cling.transport.spi.StreamClient (FINE)
TCP communication
org.teleal.cling.transport.spi.SOAPActionProcessor (FINER)
SOAP action message processing and content
org.teleal.cling.transport.spi.GENAEventProcessor (FINER)
GENA event message processing and content
org.teleal.cling.transport.impl.HttpHeaderConverter (FINER)
HTTP header processing

UPnP Protocol 
org.teleal.cling.protocol.ProtocolFactory (FINER)
org.teleal.cling.protocol.async (FINER)
Discovery (Notification & Search)
org.teleal.cling.protocol.ProtocolFactory (FINER)
org.teleal.cling.protocol.RetrieveRemoteDescriptors (FINE)
org.teleal.cling.protocol.sync.ReceivingRetrieval (FINE)
org.teleal.cling.binding.xml.DeviceDescriptorBinder (FINE)
org.teleal.cling.binding.xml.ServiceDescriptorBinder (FINE)
Description
org.teleal.cling.protocol.ProtocolFactory (FINER)
org.teleal.cling.protocol.sync.ReceivingAction (FINER)
org.teleal.cling.protocol.sync.SendingAction (FINER)
Control
org.teleal.cling.model.gena (FINER)
org.teleal.cling.protocol.ProtocolFactory (FINER)
org.teleal.cling.protocol.sync.ReceivingEvent (FINER)
org.teleal.cling.protocol.sync.ReceivingSubscribe (FINER)
org.teleal.cling.protocol.sync.ReceivingUnsubscribe (FINER)
org.teleal.cling.protocol.sync.SendingEvent (FINER)
org.teleal.cling.protocol.sync.SendingSubscribe (FINER)
org.teleal.cling.protocol.sync.SendingUnsubscribe (FINER)
org.teleal.cling.protocol.sync.SendingRenewal (FINER)
GENA

Core 
org.teleal.cling.transport.Router (FINER)
Message Router
org.teleal.cling.registry.Registry (FINER)
org.teleal.cling.registry.LocalItems (FINER)
org.teleal.cling.registry.RemoteItems (FINER)
Registry
org.teleal.cling.binding.annotations (FINER)
org.teleal.cling.model.meta.LocalService (FINER)
org.teleal.cling.model.action (FINER)
org.teleal.cling.model.state (FINER)
org.teleal.cling.model.DefaultServiceManager (FINER)
Local service binding & invocation
org.teleal.cling.controlpoint (FINER)
Control Point interaction

One way to configure JUL is with a properties file. For example, create the following file as mylogging.properties:

# Enables a one-message-per-line handler (shipping in teleal-common.jar)
handlers=org.teleal.common.logging.SystemOutLoggingHandler

# The default (root) log level
.level=INFO

# Extra settings for various categories

org.teleal.cling.level=INFO

org.teleal.cling.protocol.level=FINEST

org.teleal.cling.registry.Registry.level=FINER
org.teleal.cling.registry.LocalItems.level=FINER
org.teleal.cling.registry.RemoteItems.level=FINER 

You can now start your application with a system property that names your logging configuration:

$ java -cp /path/to/teleal-common.jar:/path/to/cling-core.jar:classes/ \
        -Djava.util.logging.config.file=/path/to/mylogging.properties \
        example.binaryLight.BinaryLightServer

You should see the desired log messages printed on System.out.

3. The Cling Core API

The programming interface of Cling is fundamentally the same for UPnP clients and servers. The single entry point for any program is the UpnpService instance. Through this API you access the local UPnP stack, and either execute operations as a client (control point) or provide services to local or remote clients through the registry.

The following diagram shows the most important interfaces of Cling Core:

API Overview

You'll be calling these interfaces to work with UPnP devices and interact with UPnP services. Cling provides a fine-grained meta-model representing these artifacts:

Metamodel Overview

In this chapter we'll walk through the API and metamodel in more detail, starting with the UpnpService.

3.1. Working with a UpnpService

The UpnpService is an interface:

public interface UpnpService {

    public UpnpServiceConfiguration getConfiguration();
    public ProtocolFactory getProtocolFactory();
    public Router getRouter();

    public ControlPoint getControlPoint();
    public Registry getRegistry();

    public void shutdown();

}

An instance of UpnpService represents a running UPnP stack, including all network listeners, background maintenance threads, and so on. Cling Core bundles a default implementation which you can simply instantiate as follows:

UpnpService upnpService = new UpnpServiceImpl();

With this implementation, the local UPnP stack is ready immediately, it listens on the network for UPnP messages. You should call the shutdown() method when you no longer need the UPnP stack. The bundled implementation will then cut all connections with remote event listeners and also notify all other UPnP participants on the network that your local services are no longer available. If you do not shutdown your UPnP stack, remote control points might think that your services are still available until your earlier announcements expire.

The bundled implementation offers two additional constructors:

UpnpService upnpService =
    new UpnpServiceImpl(RegistryListener... registryListeners);

This constructor accepts your custom RegistryListener instances, which will be activated immediately even before the UPnP stack listens on any network interface. This means that you can be notified of all incoming device and service registrations as soon as the network stack is ready. Note that this is rarely useful, you'd typically send search requests after the stack is up and running anyway - after adding listeners to the registry.

The second constructor supports customization of the UPnP stack configuration:

UpnpService upnpService =
    new UpnpServiceImpl(new DefaultUpnpServiceConfiguration(8081));

This example configuration will change the TCP listening port of the UPnP stack to 8081, the default being an ephemeral (system-selected free) port. The UpnpServiceConfiguration is also an interface, in the example above you can see how the bundled default implementation is instantiated.

The following section explain the methods of the UpnpService interface and what they return in more detail.

3.1.1. Customizing configuration settings

This is the configuration interface of the default UPnP stack in Cling Core, an instance of which you have to provide when creating the UpnpServiceImpl:

public interface UpnpServiceConfiguration {

    // NETWORK
    public NetworkAddressFactory createNetworkAddressFactory();

    public DatagramProcessor getDatagramProcessor();
    public SOAPActionProcessor getSoapActionProcessor();
    public GENAEventProcessor getGenaEventProcessor();

    public StreamClient createStreamClient();
    public MulticastReceiver createMulticastReceiver(NetworkAddressFactory naf);
    public DatagramIO createDatagramIO(NetworkAddressFactory naf);
    public StreamServer createStreamServer(NetworkAddressFactory naf);

    public Executor getMulticastReceiverExecutor();
    public Executor getDatagramIOExecutor();
    public Executor getStreamServerExecutor();

    // DESCRIPTORS
    public DeviceDescriptorBinder getDeviceDescriptorBinderUDA10();
    public ServiceDescriptorBinder getServiceDescriptorBinderUDA10();

    // PROTOCOL
    public Executor getAsyncProtocolExecutor();
    public Executor getSyncProtocolExecutor();

    // REGISTRY
    public Namespace getNamespace();
    public Executor getRegistryMaintainerExecutor();
    public Executor getRegistryListenerExecutor();

}

This is quite an extensive SPI but you typically won't implement it from scratch. Overriding and customizing the bundled DefaultUpnpServiceConfiguration should suffice in most cases.

The configuration settings reflect the internal structure of Cling Core:

Network

The NetworkAddressFactory provides the network interfaces, ports, and multicast settings which are used by the UPnP stack. At the time of writing, the following interfaces and IP addresses are ignored by the default configuration: any IPv6 interfaces and addresses, interfaces whose name starts with "vmnet", and the local loopback. Otherwise, all interfaces and their TCP/IP addresses are used and bound.

You can set the system property org.teleal.cling.network.useInterfaces to provide a comma-separated list of network interfaces you'd like to bind excusively. Additionally, you can restrict the actual TCP/IP addresses to which the stack will bind with a comma-separated list of IP address provided through the org.teleal.cling.network.useAddresses system property.

Furthermore, the configuration declares the network-level message listeners and senders, that is, the implementations used by the network Router.

Stream messages are TCP HTTP requests and responses, the bundled implementation will use the Sun JDK 6.0 webserver to listen for HTTP requests, and it sends HTTP requests through the standard JDK HttpURLConnection.

UDP unicast and multicast datagrams are received, parsed, and send by a custom implementation bundled with Cling Core that does not require any particular Sun JDK classes.

Descriptors

Reading and writing UPnP XML descriptors is handled by binders, you can provide alternative implementations if necessary.

Executors

The Cling UPnP stack is multi-threaded, thread creation and execution is handled through java.util.concurrent executors. The default configuration uses a pool of threads with a maximum size of 64 concurrently running threads, which should suffice for even very large installations. Executors can be configured fine-grained, for network message handling, actual UPnP protocol execution (handling discovery, control, and event procedures), and local registry maintenance and listener callback execution. Most likely you will not have to customize any of these settings.

If you are trying to use Cling Core in a runtime container such as Tomcat, JBoss AS, or Glassfish, you might run into an error on startup. The error tells you that Cling couldn't use the Java JDK's HTTPURLConnection for HTTP client operations. This is an old and badly designed part of the JDK: Only "one application" in the whole JVM can configure it. If your container is already using the HTTPURLConnection, you have to switch Cling to an alternative HTTP client, e.g. the other bundled implementation based on Apache HTTP Core.

Furthermore, as mentioned earlier, Cling Core provides HTTP services using, by default, the SUN JDK 6.x bundled webserver. If you are not deploying with the Sun JDK, or if you also switched the HTTP client already to Apache HTTP Core, you probably also have or want to switch the HTTP server.

This is how you override the default network transport configuration:

...
import org.teleal.cling.transport.impl.apache.StreamClientConfigurationImpl;
import org.teleal.cling.transport.impl.apache.StreamClientImpl;
import org.teleal.cling.transport.impl.apache.StreamServerConfigurationImpl;
import org.teleal.cling.transport.impl.apache.StreamServerImpl;

public class MyUpnpServiceConfiguration extends DefaultUpnpServiceConfiguration {

    @Override
    public StreamClient createStreamClient() {
        return new StreamClientImpl(new StreamClientConfigurationImpl());
    }

    @Override
    public StreamServer createStreamServer(NetworkAddressFactory networkAddressFactory) {
        return new StreamServerImpl(
                new StreamServerConfigurationImpl(
                        networkAddressFactory.getStreamListenPort()
                )
        );
    }

}

Don't forget to add the Apache HTTP Core libraries to your classpath or as dependencies in your pom.xml!

3.1.2. The protocol factory

Cling Core internals are modular and any aspect of the UPnP protocol is handled by an implementation (class) which can be replaced without affecting any other aspect. The ProtocolFactory provides implementations, it is always the first access point for the UPnP stack when a message which arrives on the network or an outgoing message has to be handled:

public interface ProtocolFactory {

    public ReceivingAsync createReceivingAsync(IncomingDatagramMessage message)
                            throws ProtocolCreationException;;

    public ReceivingSync createReceivingSync(StreamRequestMessage requestMessage);
                            throws ProtocolCreationException;;

    public SendingNotificationAlive createSendingNotificationAlive(LocalDevice ld);
    public SendingNotificationByebye createSendingNotificationByebye(LocalDevice ld);
    public SendingSearch createSendingSearch(UpnpHeader searchTarget);
    public SendingAction createSendingAction(ActionInvocation invocation, URL url);
    public SendingSubscribe createSendingSubscribe(RemoteGENASubscription s);
    public SendingRenewal createSendingRenewal(RemoteGENASubscription s);
    public SendingUnsubscribe createSendingUnsubscribe(RemoteGENASubscription s);
    public SendingEvent createSendingEvent(LocalGENASubscription s);
    
}

This API is a low-level interface that allows you to access the internals of the UPnP stack, in the rare case you need to manually trigger a particular procedure.

The first two methods are called by the networking code when a message arrives, either multicast or unicast UDP datagrams, or a TCP (HTTP) stream request. The default protocol factory implementation will then pick the appropriate receiving protocol implementation to handle the incoming message.

The local registry of local services known to the UPnP stack naturally also sends messages, such as alive and byebye notifications. Also, if you write a UPnP control point various search, control, and eventing messages are send by the local UPnP stack. The protocol factory decouples the message sender (registry, control point) from the actual creation, preparation, and transmission of the messages.

Transmission and reception of messages at the lowest-level is the job of the network Router.

3.1.3. Accessing low-level network services

The reception and sending of messages, that is, all message transport, is encapsulated through the Router interface:

public interface Router {

    public void received(IncomingDatagramMessage msg);
    public void received(UpnpStream stream);

    public void send(OutgoingDatagramMessage msg);
    public StreamResponseMessage send(StreamRequestMessage msg);

    public void broadcast(byte[] bytes);

}

UPnP works with two types of messages: Multicast and unicast UDP datagrams which are typically handled asynchronously, and request/response TCP messages with an HTTP payload. The Cling Core bundled RouterImpl will instantiate and maintain the listeners for incoming messages as well as transmit any outgoing messages.

The actual implementation of a message receiver which listens on the network or a message sender is provided by the UpnpServiceConfiguration, which we have introduced earlier. You can access the Router directly if you have to execute low-level operations on the network layer of the UPnP stack.

Most of the time you will however work with the ControlPoint and Registry interfaces to interact with the UPnP stack.

3.2. Client operations with ControlPoint

Your primary API when writing a UPnP client application is the ControlPoint. An instance is available with getControlPoint() on the UpnpService.

public interface ControlPoint {

    public void search(UpnpHeader searchType);
    public void execute(ActionCallback callback);
    public void execute(SubscriptionCallback callback);

}

A UPnP client application typically wants to:

  • Search the network for a particular service which it knows how to utilize. Any response to a search request will be delivered asynchronously, so you have to listen to the Registry for device registrations, which will occur when devices respond to your search request.
  • Execute actions which are offered by services. Action execution is processed asynchronously in Cling Core, and your ActionCallback will be notified when the execution was a success (with result values), or a failure (with error status code and messages).
  • Subscribe to a service's eventing, so your SubscriptionCallback is notified asynchronously when the state of a service changes and an event has been received for your client. You also use the callback to cancel the event subscription when you are no longer interested in state changes.

Let's start with searching for UPnP devices on the network.

3.2.1. Searching the network

When your control point joins the network it probably won't know any UPnP devices and services that might be available. To learn about the present devices it can broadcast - actually with UDP multicast datagrams - a search message which will be received by every device. Each receiver then inspects the search message and decides if it should reply directly (with notification UDP datagrams) to the sending control point.

Search messages carry a search type header and receivers consider this header when they evaluate a potential response. The Cling ControlPoint API accepts a UpnpHeader argument when creating outgoing search messages.

Most of the time you'd like all devices to respond to your search, this is what the dedicated STAllHeader is used for:

upnpService.getControlPoint().search(
        new STAllHeader()
);

Notification messages will be received by your control point and you can listen to the Registry and inspect the found devices and their services. (By the way, if you call search() without any argument, that's the same.)

On the other hand, when you already know the unique device name (UDN) of the device you are searching for - maybe because your control point remembered it while it was turned off - you can send a message which will trigger a response from only a particular device:

upnpService.getControlPoint().search(
        new UDNHeader(udn)
);

This is mostly useful to avoid network congestion when dozens of devices might all respond to a search request. Your Registry listener code however still has to inspect each newly found device, as registrations might occur independently from searches.

You can also search by device or service type. This search request will trigger responses from all devices of type "urn:schemas-upnp-org:device:BinaryLight:1":

UDADeviceType udaType = new UDADeviceType("BinaryLight");
upnpService.getControlPoint().search(
        new UDADeviceTypeHeader(udaType)
);

If the desired device type is of a custom namespace, use this variation:

DeviceType type = new DeviceType("org-mydomain", "MyDeviceType", 1);
upnpService.getControlPoint().search(
        new DeviceTypeHeader(type)
);

Or, you can search for all devices which implement a particular service type:

UDAServiceType udaType = new UDAServiceType("SwitchPower");
upnpService.getControlPoint().search(
        new UDAServiceTypeHeader(udaType)
);
ServiceType type = new ServiceType("org-mydomain", "MyServiceType", 1);
upnpService.getControlPoint().search(
        new ServiceTypeHeader(type)
);
3.2.2. Invoking an action

UPnP services expose state variables and actions. While the state variables represent the current state of the service, actions are the operations used to query or maniuplate the service's state. You have to obtain a Service instance from a Device to access any Action. The target device can be local to the same UPnP stack as your control point, or it can be remote of another device anywhere on the network. We'll discuss later in this chapter how to access devices through the local stack's Registry.

Once you have the device, access the Service through the metadata model, for example:

Service service = device.findService(new UDAServiceId("SwitchPower"));
Action getStatusAction = service.getAction("GetStatus");

This method will search the device and all its embedded devices for a service with the given identififer and returns either the found Service or null. The Cling metamodel is thread-safe, so you can share an instance of Service or Action and access it concurrently.

Invoking an action is the job of an instance of ActionInvocation, note that this instance is NOT thread-safe and each thread that wishes to execute an action has to obtain its own invocation from the Action metamodel:

ActionInvocation getStatusInvocation = new ActionInvocation(getStatusAction);

ActionCallback getStatusCallback = new ActionCallback(getStatusInvocation) {

    @Override
    public void success(ActionInvocation invocation) {
        ActionArgumentValue status  = invocation.getOutput("ResultStatus");

        assert status != null;

        assertEquals(status.getArgument().getName(), "ResultStatus");

        assertEquals(status.getDatatype().getClass(), BooleanDatatype.class);
        assertEquals(status.getDatatype().getBuiltin(), Datatype.Builtin.BOOLEAN);

        assertEquals((Boolean) status.getValue(), Boolean.valueOf(false));
        assertEquals(status.toString(), "0"); // '0' is 'false' in UPnP
    }

    @Override
    public void failure(ActionInvocation invocation,
                        UpnpResponse operation,
                        String defaultMsg) {
        System.err.println(defaultMsg);
    }
};

upnpService.getControlPoint().execute(getStatusCallback);

Execution is asynchronous, your ActionCallback has two methods which will be called by the UPnP stack when the execution completes. If the action is successful, you can obtain any output argument values from the invocation instance, which is conveniently passed into the success() method. You can inspect the named output argument values and their datatypes to continue processing the result.

Action execution doesn't have to be processed asynchronously, after all, the underlying HTTP/SOAP protocol is a request waiting for a response. The callback programming model however fits nicely into a typical UPnP client, which also has to process event notifications and device registrations asynchronously. If you want to execute an ActionInvocation directly, within the current thread, use the empty ActionCallback.Default implementation:

new ActionCallback.Default(getStatusInvocation, upnpService.getControlPoint()).run();

When invocation fails you can access the failure details through invocation.getFailure(), or use the shown convenience method to create a simple error message. See the Javadoc of ActionCallback for more details.

When an action requires input argument values, you have to provide them. Like output arguments, any input arguments of actions are also named, so you can set them by calling setInput("MyArgumentName", value):

Action action = service.getAction("SetTarget");

ActionInvocation setTargetInvocation = new ActionInvocation(action);

setTargetInvocation.setInput("NewTargetValue", true); // Can throw InvalidValueException

// Alternative:
//
// setTargetInvocation.setInput(
//         new ActionArgumentValue(
//                 action.getInputArgument("NewTargetValue"),
//                 true
//         )
// );

ActionCallback setTargetCallback = new ActionCallback(setTargetInvocation) {

    @Override
    public void success(ActionInvocation invocation) {
        ActionArgumentValue[] output = invocation.getOutput();
        assertEquals(output.length, 0);
    }

    @Override
    public void failure(ActionInvocation invocation,
                        UpnpResponse operation,
                        String defaultMsg) {
        System.err.println(defaultMsg);
    }
};

upnpService.getControlPoint().execute(setTargetCallback);

This action has one input argument of UPnP type "boolean". You can set a Java boolean primitive or Boolean instance and it will be automatically converted. If you set an invalid value for a particular argument, such as an instance with the wrong type, an InvalidValueException will be thrown immediately.

Empty values and null in Cling
There is no difference between empty string "" and null in Cling, because the UPnP specification does not address this issue. The SOAP message of an action call or an event message must contain an element <SomeVar></SomeVar> for all arguments, even if it is an empty XML element. If you provide an empty string or a null value when preparing a message, it will always be a null on the receiving end because we can only transmit one thing, an empty XML element. If you forget to set an input argument's value, it will be null/empty element.
3.2.3. Receiving events from services

The UPnP specification defines a general event notification system (GENA) which is based on a publish/subscribe paradigm. Your control point subscribes with a service in order to receive events. When the service state changes, an event message will be delivered to the callback of your control point. Subscriptions are periodically refreshed until you unsubscribe from the service. If you do not unsubscribe and if a refresh of the subscription fails, maybe because the control point was turned off without proper shutdown, the subscription will timeout on the publishing service's side.

This is an example subscription on a service that sends events for a state variable named Status (e.g. the previously shown SwitchPower service). The subscription's refresh and timeout period is 600 seconds:

SubscriptionCallback callback = new SubscriptionCallback(service, 600) {

    @Override
    public void established(GENASubscription sub) {
        System.out.println("Established: " + sub.getSubscriptionId());
    }

    @Override
    protected void failed(GENASubscription subscription,
                          UpnpResponse responseStatus,
                          Exception exception,
                          String defaultMsg) {
        System.err.println(defaultMsg);
    }

    @Override
    public void ended(GENASubscription sub,
                      CancelReason reason,
                      UpnpResponse response) {
        assert reason == null;
    }

    public void eventReceived(GENASubscription sub) {

        System.out.println("Event: " + sub.getCurrentSequence().getValue());

        Map<String, StateVariableValue> values = sub.getCurrentValues();
        StateVariableValue status = values.get("Status");

        assertEquals(status.getDatatype().getClass(), BooleanDatatype.class);
        assertEquals(status.getDatatype().getBuiltin(), Datatype.Builtin.BOOLEAN);

        System.out.println("Status is: " + status.toString());

    }

    public void eventsMissed(GENASubscription sub, int numberOfMissedEvents) {
        System.out.println("Missed events: " + numberOfMissedEvents);
    }

};

upnpService.getControlPoint().execute(callback);

The SubscriptionCallback offers the methods failed(), established(), and ended() which are called during a subscription's lifecycle. When a subscription ends you will be notified with a CancelReason whenever the termination of the subscription was irregular. See the Javadoc of these methods for more details.

Every event message from the service will be passed to the eventReceived() method, and every message will carry a sequence number. Variable values are not transmitted individually, each message contains StateVariableValue instances for all evented variables of a service. You'll receive a snapshot of the state of the service at the time the event was triggered.

Whenever the receiving UPnP stack detects an event message that is out of sequence, e.g. because some messages were lost during transport, the eventsMissed() method will be called before you receive the event. You then decide if missing events is important for the correct behavior of your application, or if you can silently ignore it and continue processing events with non-consecutive sequence numbers.

You end a subscription regularly by calling callback.end(), which will unsubscribe your control point from the service.

3.3. The Registry

The Registry, which you access with getRegistry() on the UpnpService, is the heart of a Cling Core UPnP stack. The registry is responsible for:

  • Maintaining discovered UPnP devices on your network. It also offers a management API so you can register local devices and offer local services. This is how you expose your own UPnP devices on the network. The registry handles all notification, expiration, request routing, refreshing, and so on.
  • Managing GENA (general event & notification architecture) subscriptions. Any outgoing subscription to a remote service is known by the registry, it is refreshed periodically so it doesn't expire. Any incoming eventing subscription to a local service is also known and maintained by the registry (expired and removed when necessary).
  • Providing the interface for the addition and removal of RegistryListener instances. A registry listener is used in client or server UPnP applications, it provides a uniform interface for notification of registry events. Typically, you write and register a listener to be notified when a service you want to work with becomes available on the network - on a local or remote device - and when it disappears.
3.3.1. Browsing the Registry

Although you typically create a RegistryListener to be notified of discovered and disappearing UPnP devices on your network, sometimes you have to browse the Registry manually.

The following call will return a device with the given unique device name, but only a root device and not any embedded device. Set the second parameter of registry.getDevice() to false if the device you are looking for might be an embedded device.

Registry registry = upnpService.getRegistry();
Device foundDevice = registry.getDevice(udn, true);

assertEquals(foundDevice.getIdentity().getUdn(), udn);

If you know that the device you need is a LocalDevice - or a RemoteDevice - you can use the following operation:

LocalDevice localDevice = registry.getLocalDevice(udn, true);

Most of the time you need a device that is of a particular type or that implements a particular service type, because this is what your control point can handle:

DeviceType deviceType = new UDADeviceType("MY-DEVICE-TYPE", 1);
Collection<Device> devices = registry.getDevices(deviceType);
ServiceType serviceType = new UDAServiceType("MY-SERVICE-TYPE-ONE", 1);
Collection<Device> devices = registry.getDevices(serviceType);
3.3.2. Listening to registry changes

The RegistryListener is your primary API when discovering devices and services with your control point. UPnP operates asynchronous, so advertisements (either alive or byebye) of devices can occur at any time. Responses to your network search messages are also asynchronous.

This is the interface:

public interface RegistryListener {

    public void remoteDeviceDiscoveryStarted(Registry registry, RemoteDevice device);
    public void remoteDeviceDiscoveryFailed(Registry registry, RemoteDevice device, Exception ex);

    public void remoteDeviceAdded(Registry registry, RemoteDevice device);
    public void remoteDeviceUpdated(Registry registry, RemoteDevice device);
    public void remoteDeviceRemoved(Registry registry, RemoteDevice device);

    public void localDeviceAdded(Registry registry, LocalDevice device);
    public void localDeviceRemoved(Registry registry, LocalDevice device);

}

Typically you don't want to implement all of these methods. Some are only useful if you write a service or a generic control point. Most of the time you want to be notified when a particular device with a particular service appears on your network. So it is much easier to extend the DefaultRegistryListener, which has empty implementations for all methods of the interface, and only override the methods you need.

The remoteDeviceDiscoveryStarted() and remoteDeviceDiscoveryFailed() methods are completely optional but useful on slow machines (such as Android handsets). Cling will retrieve and initialize all device metadata for each UPnP device before it will announce it on the Registry. UPnP metadata is split into several XML descriptors, so retrieval via HTTP of these descriptors, parsing, and validating all metadata for a complex UPnP device and service model can take several seconds. These two methods allow you to access the device as soon as possible, after the first descriptor has been retrieved and parsed. At this time the services metadata is however not available:

public class QuickstartRegistryListener extends DefaultRegistryListener {

    @Override
    public void remoteDeviceDiscoveryStarted(Registry registry, RemoteDevice device) {

        // You can already use the device here and you can see which services it will have
        assertEquals(device.findServices().length, 3);

        // But you can't use the services
        for (RemoteService service: device.findServices()) {
            assertEquals(service.getActions().length, 0);
            assertEquals(service.getStateVariables().length, 0);
        }
    }

    @Override
    public void remoteDeviceDiscoveryFailed(Registry registry, RemoteDevice device, Exception ex) {
        // You might want to drop the device, its services couldn't be hydrated
    }
}

This is how you register and activate a listener:

QuickstartRegistryListener listener = new QuickstartRegistryListener();
upnpService.getRegistry().addListener(listener);

Most of the time, on any device that is faster than a cellphone, your listeners will look like this:

public class MyListener extends DefaultRegistryListener {

    @Override
    public void remoteDeviceAdded(Registry registry, RemoteDevice device) {
        Service myService = device.findService(new UDAServiceId("MY-SERVICE-123"));
        if (myService != null) {
            // Do something with the discovered service
        }
    }

    @Override
    public void remoteDeviceRemoved(Registry registry, RemoteDevice device) {
        // Stop using the service if this is the same device, it's gone now
    }
}

The device metadata of the parameter to remoteDeviceAdded() is fully hydrated, all of its services, actions, and state variables are available. You can continue with this metadata, writing action invocations and event monitoring callbacks. You also might want to react accordingly when the device disappears from the network.

4. Creating and binding services

Out of the box, any Java class can be a UPnP service. Let's go back to the first example of a UPnP service in chapter 1, the SwitchPower:1 service implementation, repeated here:

package example.binarylight;

import org.teleal.cling.binding.annotations.*;

@UpnpService(
        serviceId = @UpnpServiceId("SwitchPower"),
        serviceType = @UpnpServiceType(value = "SwitchPower", version = 1)
)
public class SwitchPower {

    @UpnpStateVariable(defaultValue = "0", sendEvents = false)
    private boolean target = false;

    @UpnpStateVariable(defaultValue = "0")
    private boolean status = false;

    @UpnpAction
    public void setTarget(@UpnpInputArgument(name = "NewTargetValue")
                          boolean newTargetValue) {
        target = newTargetValue;
        status = newTargetValue;
        System.out.println("Switch is: " + status);
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "RetTargetValue"))
    public boolean getTarget() {
        return target;
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "ResultStatus"))
    public boolean getStatus() {
        return status;
    }

}

This class depends on the org.teleal.cling.annotation package at compile-time. The metadata encoded in these source annotations is preserved in the bytecode and Cling will read it at runtime when you bind the service ("binding" is just a fancy word for reading and writing metadata). You can load and execute this class without accessing the annotations, in any environment and without having the Cling libraries on your classpath. This is a compile-time dependency only.

Cling annotations give you much flexibility in designing your service implementation class, as shown in the following examples.

4.1. Annotating a service implementation

The previously shown service class had a few annotations on the class itself, declaring the name and version of the service. Then annotations on fields were used to declare the state variables of the service and annotations on methods to declare callable actions.

Your service implementation might not have fields that directly map to UPnP state variables.

4.1.1. Mapping state variables

The following example only has a single field named power, however, the UPnP service requires two state variables. In this case you declare the UPnP state variables with annotations on the class:

@UpnpService(
        serviceId = @UpnpServiceId("SwitchPower"),
        serviceType = @UpnpServiceType(value = "SwitchPower", version = 1)

)
@UpnpStateVariables(
        {
                @UpnpStateVariable(
                        name = "Target",
                        defaultValue = "0",
                        sendEvents = false
                ),
                @UpnpStateVariable(
                        name = "Status",
                        defaultValue = "0"
                )
        }
)
public class SwitchPowerAnnotatedClass {

    private boolean power;

    @UpnpAction
    public void setTarget(@UpnpInputArgument(name = "NewTargetValue")
                          boolean newTargetValue) {
        power = newTargetValue;
        System.out.println("Switch is: " + power);
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "RetTargetValue"))
    public boolean getTarget() {
        return power;
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "ResultStatus"))
    public boolean getStatus() {
        return power;
    }
}

The power field is not mapped to the state variables and you are free to design your service internals as you like. Did you notice that you never declared the datatype of your state variables? Also, how can Cling read the "current state" of your service for GENA subscribers or when a "query state variable" action is received? Both questions have the same answer.

Let's consider GENA eventing first. This example has an evented state variable called Status, and if a control point subscribes to the service to be notified of changes, how will Cling obtain the current status? If you'd have used @UpnpStateVariable on your fields, Cling would then directly access field values through Java Reflection. On the other hand if you declare state variables not on fields but on your service class, Cling will during binding detect any JavaBean-style getter method that matches the derived property name of the state variable.

In other words, Cling will discover that your class has a getStatus() method. It doesn't matter if that method is also an action-mapped method, the important thing is that it matches JavaBean property naming conventions. The Status UPnP state variable maps to the status property, which is expected to have a getStatus() accessor method. Cling will use this method to read the current state of your service for GENA subscribers and when the state variable is manually queried.

If you do not provide a UPnP datatype name in your @UpnpStateVariable annotation, Cling will use the type of the annotated field or discovered JavaBean getter method to figure out the type. The supported default mappings between Java types and UPnP datatypes are shown in the following table:

Java TypeUPnP Datatype
java.lang.Booleanboolean
booleanboolean
java.lang.Shorti2
shorti2
java.lang.Integeri4
inti4
org.teleal.cling.model.types.UnsignedIntegerOneByteui1
org.teleal.cling.model.types.UnsignedIntegerTwoBytesui2
org.teleal.cling.model.types.UnsignedIntegerFourBytesui4
java.lang.Floatr4
floatr4
java.lang.Doublefloat
doublefloat
java.lang.Characterchar
charchar
java.lang.Stringstring
java.util.Calendardatetime
Byte[]bin.base64
java.net.URIuri

Cling tries to provide smart defaults. For example, the previously shown service classes did not name the related state variable of action output arguments, as required by UPnP. Cling will automatically detect that the getStatus() method is a JavaBean getter method (its name starts with get or is) and use the JavaBean property name to find the related state variable. In this case that would be the JavaBean property status and Cling is also smart enough to know that you really want the uppercase UPnP state variable named Status.

4.1.2. Explicitly naming related state variables

If your mapped action method does not match the name of a mapped state variable, you have to provide the name of (any) argument's related state variable:

@UpnpAction(
        name = "GetStatus",
        out = @UpnpOutputArgument(
                name = "ResultStatus",
                stateVariable = "Status"
        )
)
public boolean retrieveStatus() {
    return status;
}

Here the method has the name retrieveStatus, which you also have to override if you want it be known as a the GetStatus UPnP action. Because it is no longer a JavaBean accessor for status, it explicitly has to be linked with the related state variable Status. You always have to provide the related state variable name if your action has more than one output argument.

The "related statevariable" detection algorithm in Cling has one more trick up its sleeve however. The UPnP specification says that a state variable which is only ever used to describe the type of an input or output argument should be named with the prefix A_ARG_TYPE_. So if you do not name the related state variable of your action argument, Cling will also look for a state variable with the name A_ARG_TYPE_[Name Of Your Argument]. In the example above, Cling is therefore also searching (unsuccessfully) for a state variable named A_ARG_TYPE_ResultStatus. (Given that direct querying of state variables is already deprecated in UDA 1.0, there are NO state variables which are anything but type declarations for action input/output arguments. This is a good example why UPnP is such a horrid specification.)

For the next example, let's assume you have a class that was already written, not necessarily as a service backend for UPnP but for some other purpose. You can't redesign and rewrite your class without interrupting all existing code. Cling offers some flexibility in the mapping of action methods, especially how the output of an action call is obtained.

4.1.3. Getting an output value from another method

In the following example, the UPnP action has an output argument but the mapped method is void and does not return any value:

public boolean getStatus() {
    return status;
}

@UpnpAction(
        name = "GetStatus",
        out = @UpnpOutputArgument(
                name = "ResultStatus",
                getterName = "getStatus"
        )
)
public void retrieveStatus() {
    // NOOP in this example
}

By providing a getterName in the annotation you can instruct Cling to call this getter method when the action method completes, taking the getter method's return value as the output argument value. If there are several output arguments you can map each to a different getter method.

Alternatively, and especially if an action has several output arguments, you can return multiple values wrapped in a JavaBean from your action method.

4.1.4. Getting output values from a JavaBean

Here the action method does not return the output argument value directly, but a JavaBean instance is returned which offers a getter method to obtain the output argument value:

@UpnpAction(
        name = "GetStatus",
        out = @UpnpOutputArgument(
                name = "ResultStatus",
                getterName = "getWrapped"
        )
)
public StatusHolder getStatus() {
    return new StatusHolder(status);
}

public class StatusHolder {
    boolean wrapped;

    public StatusHolder(boolean wrapped) {
        this.wrapped = wrapped;
    }

    public boolean getWrapped() {
        return wrapped;
    }
}

Cling will detect that you mapped a getter name in the output argument and that the action method is not void. It now expects that it will find the getter method on the returned JavaBean. If there are several output arguments, all of them have to be mapped to getter methods on the returned JavaBean.

An important piece is missing from the SwitchPower:1 implementation: It doesn't fire any events when the status of the power switch changes. This is in fact required by the specification that defines the SwitchPower:1 service. The following section explains how you can propagate state changes from within your UPnP service to local and remote subscribers.

4.2. Providing events on service state changes

The standard mechanism in the JDK for eventing is the PropertyChangeListener reacting on a PropertyChangeEvent. Cling utilizes this API for service eventing, thus avoiding a dependency between your service code and proprietary APIs.

Consider the following modification of the original SwitchPower:1 implementation:

package example.localservice;

import org.teleal.cling.binding.annotations.*;
import java.beans.PropertyChangeSupport;

@UpnpService(
        serviceId = @UpnpServiceId("SwitchPower"),
        serviceType = @UpnpServiceType(value = "SwitchPower", version = 1)
)
public class SwitchPowerWithPropertyChangeSupport {

    private final PropertyChangeSupport propertyChangeSupport;

    public SwitchPowerWithPropertyChangeSupport() {
        this.propertyChangeSupport = new PropertyChangeSupport(this);
    }

    public PropertyChangeSupport getPropertyChangeSupport() {
        return propertyChangeSupport;
    }

    @UpnpStateVariable(defaultValue = "0", sendEvents = false)
    private boolean target = false;

    @UpnpStateVariable(defaultValue = "0")
    private boolean status = false;

    @UpnpAction
    public void setTarget(@UpnpInputArgument(name = "NewTargetValue") boolean newTargetValue) {

        boolean targetOldValue = target;
        target = newTargetValue;
        boolean statusOldValue = status;
        status = newTargetValue;

        // These have no effect on the UPnP monitoring but it's JavaBean compliant
        getPropertyChangeSupport().firePropertyChange("target", targetOldValue, target);
        getPropertyChangeSupport().firePropertyChange("status", statusOldValue, status);

        // This will send a UPnP event, it's the name of a state variable that sends events
        getPropertyChangeSupport().firePropertyChange("Status", statusOldValue, status);
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "RetTargetValue"))
    public boolean getTarget() {
        return target;
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "ResultStatus"))
    public boolean getStatus() {
        return status;
    }

}

The only additional dependency is on java.beans.PropertyChangeSupport. Cling detects the getPropertyChangeSupport() method of your service class and automatically binds the service management on it. You will have to have this method for eventing to work with Cling. You can create the PropertyChangeSupport instance in your service's constructor or any other way, the only thing Cling is interested in are property change events with the "property" name of a UPnP state variable.

Consequently, firePropertyChange("NameOfAStateVariable") is how you tell Cling that a state variable value has changed. It doesn't even matter if you call firePropertyChange("Status") or firePropertyChange("Status", oldValue, newValue). Cling only cares about the state variable name; if it knows the state variable is evented it will pull the data out of your service implementation instance by accessing the appropriate field or a getter.

The reason for this behavior is that in UPnP an event message has to include all evented state variable values, not just the one that changed. So Cling will read all of your evented state variable values from your service implementation when you fire a single relevant change. It does not care about a single state variable's old and new value. You can add those values when you fire the event if you also want to listen to state changes in your code, and you require the old and new value.

Note that most of the time a JavaBean property name is not the same as UPnP state variable name. For example, the JavaBean status property name is lowercase, while the UPnP state variable name is uppercase Status. The Cling eventing system ignores any property change event that doesn't exactly name a service state variable. This allows you to use JavaBean eventing independently from UPnP eventing, e.g. for GUI binding (Swing components also use the JavaBean eventing system).

More advanced mappings are possible and often required, as shown in the next examples. We are now leaving the SwitchPower service behind, as it is no longer complex enough.

4.3. Converting string action argument values

The UPnP specification defines no framework for custom datatypes. The predictable result is that service designers and vendors are overloading strings with whatever semantics they consider necessary for their particular needs. For example, the UPnP A/V specifications often require lists of values (like a list of strings or a list of numbers), which are then transported between service and control point as a single string - the individual values are represented in this string separated by commas.

Cling supports these conversions and it tries to be as transparent as possible.

4.3.1. String value converters

Consider the following service class with all state variables of string UPnP datatype - but with a much more specific Java type:

import org.teleal.cling.model.types.csv.CSV;
import org.teleal.cling.model.types.csv.CSVInteger;

@UpnpService(
        serviceId = @UpnpServiceId("MyService"),
        serviceType = @UpnpServiceType(namespace = "mydomain", value = "MyService"),
        stringConvertibleTypes = MyStringConvertible.class
)
public class MyServiceWithStringConvertibles {

    @UpnpStateVariable
    private URL myURL;

    @UpnpStateVariable
    private URI myURI;

    @UpnpStateVariable(datatype = "string")
    private List<Integer> myNumbers;

    @UpnpStateVariable
    private MyStringConvertible myStringConvertible;

    @UpnpAction(out = @UpnpOutputArgument(name = "Out"))
    public URL getMyURL() {
        return myURL;
    }

    @UpnpAction
    public void setMyURL(@UpnpInputArgument(name = "In") URL myURL) {
        this.myURL = myURL;
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "Out"))
    public URI getMyURI() {
        return myURI;
    }

    @UpnpAction
    public void setMyURI(@UpnpInputArgument(name = "In") URI myURI) {
        this.myURI = myURI;
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "Out"))
    public CSV<Integer> getMyNumbers() {
        CSVInteger wrapper = new CSVInteger();
        if (myNumbers != null)
            wrapper.addAll(myNumbers);
        return wrapper;
    }

    @UpnpAction
    public void setMyNumbers(
            @UpnpInputArgument(name = "In")
            CSVInteger myNumbers
    ) {
        this.myNumbers = myNumbers;
    }

    @UpnpAction(out = @UpnpOutputArgument(name = "Out"))
    public MyStringConvertible getMyStringConvertible() {
        return myStringConvertible;
    }

    @UpnpAction
    public void setMyStringConvertible(
            @UpnpInputArgument(name = "In")
            MyStringConvertible myStringConvertible
    ) {
        this.myStringConvertible = myStringConvertible;
    }
}

The state variables are all of UPnP datatype string because Cling knows that the Java type of the annotated field is "string convertible". This is always the case for java.net.URI and java.net.URL.

Any other Java type you'd like to use for automatic string conversion has to be named in the @UpnpService annotation on the class, like the MyStringConvertible. Note that these types have to have an appropriate toString() method and a single argument constructor that accepts a java.lang.String ("from string" conversion).

The List<Integer> is the collection you'd use in your service implementation to group several numbers. Let's assume that for UPnP communication you need a comma-separated representation of the individual values in a string, as is required by many of the UPnP A/V specifications. First, tell Cling that the state variable really is a string datatype, it can't infer that from the field type. Then, if an action has this output argument, instead of manually creating the comma-separated string you pick the appropriate converter from the classes in org.teleal.cling.model.types.csv.* and return it from your action method. These are actually java.util.List implementations, so you could use them instead of java.util.List if you don't care about the dependency. Any action input argument value can also be converted from a comma-separated string representation to a list automatically - all you have to do is use the CSV converter class as an input argument type.

4.3.2. Working with enums

Java enum's are special, unfortunately: You can't instantiate an enum value through reflection. So Cling can convert your enum value into a string for transport in UPnP messages, but you have to convert it back manually from a string. This is shown in the following service example:

@UpnpService(
        serviceId = @UpnpServiceId("MyService"),
        serviceType = @UpnpServiceType(namespace = "mydomain", value = "MyService"),
        stringConvertibleTypes = MyStringConvertible.class
)
public class MyServiceWithEnum {

    public enum Color {
        Red,
        Green,
        Blue
    }

    @UpnpStateVariable
    private Color color;

    @UpnpAction(out = @UpnpOutputArgument(name = "Out"))
    public Color getColor() {
        return color;
    }

    @UpnpAction
    public void setColor(@UpnpInputArgument(name = "In") String color) {
        this.color = Color.valueOf(color);
    }

}

Cling will automatically assume that the datatype is a UPnP string if the field (or getter) or getter Java type is an enum. Furthermore, an <allowedValueList> will be created in your service descriptor XML, so control points know that this state variable has in fact a defined set of possible values.

5. Cling on Android

Cling Core provides a UPnP stack for Android applications. Typically you'd write control point applications, as most Android systems today are small hand-held devices. You can however also write UPnP server applications on Android, all features of Cling Core are supported.

Cling on the Android emulator
At the time of writing, receiving UDP Multicast datagrams was not supported by the Android emulator. The emulator will send (multicast) UDP datagrams, however. You will be able to send a multicast UPnP search and receive UDP unicast responses, therefore discover existing running devices. You will not discover devices which have been turned on after your search, and you will not receive any message when a device is switched off. Other control points on your network will not discover your local Android device/services at all. All of this can be confusing when testing your application, so unless you really understand what works and what doesn't, you might want to use a real device instead.

This chapter explains how you can integrate Cling with your Android application as a shared Android application service component.

5.1. Configuring the application service

You can instantiate the Cling UpnpService in your Android application's main activity. On the other hand, if several activities in your application require access to the UPnP stack, a better design would utilize a background android.app.Service. Any activity that wants to access the UPnP stack can then bind and unbind from this service as needed.

The interface of this service component is org.teleal.cling.android.AndroidUpnpService:

public interface AndroidUpnpService {
    public UpnpService get();
    public UpnpServiceConfiguration getConfiguration();
    public Registry getRegistry();
    public ControlPoint getControlPoint();
}

An activity typically accesses the Registry of known UPnP devices or searches for and controls UPnP devices with the ControlPoint.

You have to configure the built-in implementation of this service component in your AndroidManifest.xml:

<manifest ...>

    <uses-permission android:name="android.permission.INTERNET"/>
    <uses-permission android:name="android.permission.ACCESS_WIFI_STATE"/>
    <uses-permission android:name="android.permission.CHANGE_WIFI_MULTICAST_STATE"/>
    <uses-permission android:name="android.permission.ACCESS_NETWORK_STATE"/>

    <application ...>

        <activity ...>
            ...
        </activity>

        <service android:name="org.teleal.cling.android.AndroidUpnpServiceImpl"/>

    </application>

</manifest>

The Cling UPnP service requires access to the WiFi interface on the device, this is in fact the only network interface on which it will bind. The service will automatically detect when the WiFi interface is switched off and handle this situation gracefully: Any client operation will result in a "no response from server" state, which your code has to expect anyway.

The service component starts and stops the UPnP system when the service component is created and destroyed. This depends on how you access the service component from within your activities.

5.2. Accessing the service from an activity

The lifecycle of services in Android is well defined. The first activity which binds to a service will start the service if it is not already running. When no activity is bound to the service any more, the operating system will destroy the service.

Let's write a simple UPnP browsing activity. It shows all devices on your network in a list and it has a menu option which triggers a search action. The activity connects to the UPnP service and then listens to any device additions or removals in the registry, so the displayed list of devices is kept up-to-date.

This is the skeleton of the activity class:

import android.app.ListActivity;
import android.content.ComponentName;
import android.content.Context;
import android.content.Intent;
import android.content.ServiceConnection;
import android.os.Bundle;
import android.os.IBinder;
import android.view.Menu;
import android.view.MenuItem;
import android.widget.ArrayAdapter;
import android.widget.Toast;
import org.teleal.cling.android.AndroidUpnpService;
import org.teleal.cling.android.AndroidUpnpServiceImpl;
import org.teleal.cling.model.meta.Device;
import org.teleal.cling.model.meta.LocalDevice;
import org.teleal.cling.model.meta.RemoteDevice;
import org.teleal.cling.registry.DefaultRegistryListener;
import org.teleal.cling.registry.Registry;

public class UpnpBrowser extends ListActivity {

    private ArrayAdapter<DeviceDisplay> listAdapter;

    private AndroidUpnpService upnpService;

    private ServiceConnection serviceConnection = ...

    private RegistryListener registryListener = new BrowseRegistryListener();

    @Override
    public void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);

        listAdapter =
            new ArrayAdapter(
                this,
                android.R.layout.simple_list_item_1
            );
        setListAdapter(listAdapter);

        getApplicationContext().bindService(
            new Intent(this, AndroidUpnpServiceImpl.class),
            serviceConnection,
            Context.BIND_AUTO_CREATE
        );
    }

    @Override
    protected void onDestroy() {
        super.onDestroy();
        if (upnpService != null) {
            upnpService.getRegistry().removeListener(registryListener);
        }
        getApplicationContext().unbindService(serviceConnection);
    }

    ...

}

We utilize the default layout provided by the Android runtime and the ListActivity superclass. Note that this activity can be your applications main activity, or further up in the stack of a task. The listAdapter is the glue between the device additions and removals on the Cling Registry and the list of items shown in the user interface.

The upnpService variable is null when no backend service is bound to this activity. Binding and unbinding occurs in the onCreate() and onDestroy() callbacks, so the activity is bound to the service as long as it is alive.

Pausing the UPnP service in the background
When an activity is no longer active (in stopped or paused state) it will still be bound to the UPnP service. The UPnP service will continue running, even if your application is not visible anymore. This will consume CPU time and power on your device, as the UPnP service's registry has to periodically maintain discovered devices, refresh announcements of local devices, remove expired GENA event subscriptions, and so on. You can tell the UPnP service to no longer maintain its registry by calling Registry#pause() when your activity's onPause() or onStop() method is called. You can then resume the background service maintenance with Registry#resume(), or check the status with Registry#isPaused(). Please read the Javadoc of these methods for more details and what consequences pausing registry maintenance has on devices, services, and GENA subscriptions.

Binding and unbinding the service is handled with this ServiceConnection:

private ServiceConnection serviceConnection = new ServiceConnection() {

    public void onServiceConnected(ComponentName className, IBinder service) {
        upnpService = (AndroidUpnpService) service;

        // Refresh the list with all known devices
        listAdapter.clear();
        for (Device device : upnpService.getRegistry().getDevices()) {
            registryListener.deviceAdded(device);
        }

        // Getting ready for future device advertisements
        upnpService.getRegistry().addListener(registryListener);

        // Search asynchronously for all devices
        upnpService.getControlPoint().search();
    }

    public void onServiceDisconnected(ComponentName className) {
        upnpService = null;
    }
};

First, all UPnP devices that are already known can be queried and displayed (there might be none if the UPnP service was just started and no device has so far announced its presence.)

Next a listener is registered with the Registry of the UPnP service. This listener will process additions and removals of devices as they are discovered on your network, and update the items shown in the user interface list. The BrowseRegistryListener is removed when the activity is destroyed.

Finally, you start asynchronous discovery by sending a search message to all UPnP devices, so they will announce themselves. Note that this search message is NOT required every time you connect to the service. It is only necessary once, to populate the registry with all known devices when your (main) activity and application starts.

This is the BrowseRegistryListener, its only job is to update the displayed list items:

class BrowseRegistryListener extends DefaultRegistryListener {

    @Override
    public void remoteDeviceDiscoveryStarted(Registry registry, RemoteDevice device) {
        deviceAdded(device);
    }

    @Override
    public void remoteDeviceDiscoveryFailed(Registry registry, final RemoteDevice device, final Exception ex) {
        runOnUiThread(new Runnable() {
            public void run() {
                Toast.makeText(
                        BrowseActivity.this,
                        "Discovery failed of '" + device.getDisplayString() + "': " +
                                (ex != null ? ex.toString() : "Couldn't retrieve device/service descriptors"),
                        Toast.LENGTH_LONG
                ).show();
            }
        });
        deviceRemoved(device);
    }

    @Override
    public void remoteDeviceAdded(Registry registry, RemoteDevice device) {
        deviceAdded(device);
    }

    @Override
    public void remoteDeviceRemoved(Registry registry, RemoteDevice device) {
        deviceRemoved(device);
    }

    @Override
    public void localDeviceAdded(Registry registry, LocalDevice device) {
        deviceAdded(device);
    }

    @Override
    public void localDeviceRemoved(Registry registry, LocalDevice device) {
        deviceRemoved(device);
    }

    public void deviceAdded(final Device device) {
        runOnUiThread(new Runnable() {
            public void run() {
                DeviceDisplay d = new DeviceDisplay(device);
                int position = listAdapter.getPosition(d);
                if (position >= 0) {
                    // Device already in the list, re-set new value at same position
                    listAdapter.remove(d);
                    listAdapter.insert(d, position);
                } else {
                    listAdapter.add(d);
                }
            }
        });
    }

    public void deviceRemoved(final Device device) {
        runOnUiThread(new Runnable() {
            public void run() {
                listAdapter.remove(new DeviceDisplay(device));
            }
        });
    }
}

For performance reasons, when a new device has been discovered, we don't wait until a fully hydrated (all services retrieved and validated) device metadata model is available. We react as quickly as possible and don't wait until the remoteDeviceAdded() method will be called. We display any device even while discovery is still running. You'd usually not care about this on a desktop computer, however, Android handheld devices are slow and UPnP uses several bloated XML descriptors to exchange metadata about devices and services. Sometimes it can take several seconds before a device and its services are fully available. The remoteDeviceDiscoveryStarted() and remoteDeviceDiscoveryFailed() methods are called as soon as possible in the discovery process. By the way, devices are equal (a.equals(b)) if they have the same UDN, they might not be identical (a==b).

Note that the Registry will call the listener methods in a separate thread. You have to update the displayed list data in the thread of the user interface.

The following two methods on the activity add a menu with a search action, so a user can refresh the list manually:

@Override
public boolean onCreateOptionsMenu(Menu menu) {
    menu.add(0, 0, 0, R.string.search_lan)
        .setIcon(android.R.drawable.ic_menu_search);
    return true;
}

@Override
public boolean onOptionsItemSelected(MenuItem item) {
    if (item.getItemId() == 0 && upnpService != null) {
        upnpService.getRegistry().removeAllRemoteDevices();
        upnpService.getControlPoint().search();
    }
    return false;
}

Finally, the DeviceDisplay class is a very simple JavaBean that only provides a toString() method for rendering the list. You can display any information about UPnP devices by changing this method:

class DeviceDisplay {
    Device device;

    public DeviceDisplay(Device device) {
        this.device = device;
    }

    public Device getDevice() {
        return device;
    }

    @Override
    public boolean equals(Object o) {
        if (this == o) return true;
        if (o == null || getClass() != o.getClass()) return false;
        DeviceDisplay that = (DeviceDisplay) o;
        return device.equals(that.device);
    }

    @Override
    public int hashCode() {
        return device.hashCode();
    }

    @Override
    public String toString() {
        // Display a little star while the device is being loaded
        return device.isFullyHydrated() ? device.getDisplayString() : device.getDisplayString() + " *";
    }
}

We have to override the equality operations as well, so we can remove and add devices from the list manually with the DeviceDisplay instance as a convenient handle.

5.3. Optimizing service behavior

The UPnP service consumes memory and CPU time while it is running. Although this is typically not an issue on a regular machine, this might be a problem on an Android handset. You can preserve memory and handset battery power if you disable certain features of the Cling UPnP service, or if you even pause and resume it when appropriate.

5.3.1. Tuning registry maintenance

There are several things going on in the background while the service is running. First, there is the registry of the service and its maintenance thread. If you are writing a control point, this background registry maintainer is going to renew your outbound GENA subscriptions with remote services periodically. It will also expire and remove any discovered remote devices when the drop off the network without saying goodbye. If you are providing a service, your device announcements will be refreshed by the registry maintainer and inbound GENA subscriptions will be removed if they haven't been renewed in time. Effectively, the registry maintainer prevents stale state on the UPnP network, so all participants have an up-to-date view of all other participants, and so on.

By default the registry maintainer will run every second and check if there is something to do (most of the time there is nothing to do, of course). The default Android configuration however has a default sleep interval of five seconds, so it is already consuming less background CPU time - while your application might be exposed to somewhat outdated information. You can further tune this setting by overriding the getRegistryMaintenanceIntervalMillis() in the UpnpServiceConfiguration. On Android, you have to subclass the service implementation to provide a new configuration:

public class MyUpnpService extends AndroidUpnpServiceImpl {

    @Override
    protected AndroidUpnpServiceConfiguration createConfiguration(WifiManager wifiManager) {
        return new AndroidUpnpServiceConfiguration(wifiManager) {

            @Override
            public int getRegistryMaintenanceIntervalMillis() {
                return 7000;
            }

        };
    }
}

Don't forget to configure MyUpnpService in your AndroidManifest.xml now instead of the original implementation. You also have to use this type when binding to the service in your activities.

5.3.2. Pausing and resuming registry maintenance

Another more effective but also more complex optimization is pausing and resuming the registry whenever your activities no longer need the UPnP service. This is typically the case when an activity is no longer in the foreground (paused) or even no longer visible (stopped). By default any activity state change has no impact on the state of the UPnP service unless you bind and unbind from and to the service in your activities lifecycle callbacks.

In addition to binding and unbinding from the service you can also pause its registry by calling Registry#pause() when your activity's onPause() or onStop() method is called. You can then resume the background service maintenance (thread) with Registry#resume(), or check the status with Registry#isPaused().

Please read the Javadoc of these methods for more details and what consequences pausing registry maintenance has on devices, services, and GENA subscriptions. Depending on what your application does, this rather minor optimization might not be worth dealing with these effects. On the other hand, your application should already be able to handle failed GENA subscription renewals, or disappearing remote devices!

5.3.3. Configuring discovery

The most effective optimization is selective discovery of UPnP devices. Although the UPnP service's network transport layer will keep running (threads are waiting and sockets are bound) in the background, this feature allows you to drop discovery messages selectively and quickly.

For example, if you are writing a control point, you can drop any received discovery message if it doesn't advertise the service you want to control - you are not interested in any other device. On the other hand if you only provide devices and services, all discovery messages (except search messages for your services) can probably be dropped, you are not interested in any remote devices and their services at all.

Discovery messages are selected and potentially dropped by Cling as soon as the UDP datagram content is available, so no further parsing and processing is needed and CPU time/memory consumption is significantly reduced while you keep the UPnP service running even in the background on an Android handset.

To configure which services are supported by your control point application, override the service implementation as shown in the previous section and provide an array of ServiceType instances:

public class MyUpnpService extends AndroidUpnpServiceImpl {

    @Override
    protected AndroidUpnpServiceConfiguration createConfiguration(WifiManager wifiManager) {
        return new AndroidUpnpServiceConfiguration(wifiManager) {

            @Override
            public ServiceType[] getExclusiveServiceTypes() {
                return new ServiceType[] {
                        new UDAServiceType("SwitchPower")
                };
            }

        };
    }
}

This configuration will ignore any advertisement from any device that doesn't also advertise a schemas-upnp-org:SwitchPower:1 service. This is what our control point can handle, so we don't need anything else. If instead you'd return an empty array (the default behavior), all services and devices will be discovered and no advertisements will be dropped.

If you are not writing a control point but a server application, you can return null in the getExclusiveServiceTypes() method. This will disable discovery completely, now all device and service advertisements are dropped as soon as they are received.

This manual has been created with Lemma from tested source code and Javadoc. Try it, you will like it.
org.teleal.cling:cling-core:1.0.5

Creative Commons License