Difference between revisions of "OLA Client API"

Revision as of 18:53, 19 April 2013

Warning: This documentation may be out of date. If in doubt check the code in the ola-examples package.

OLA provides a C++ client library which allows other applications to send/receive DMX as well as control the OLA server. When writing a client application, you have a choice between two client APIs. There was a third API, the OlaClient, but that was deprecated in Feb 2013. It will be removed in a future release and should not be used in new code.

Compiler & Linker Flags

To correctly build and link against libola you'll need to pass the right options to your build system. OLA installs a pkg-config file with the right flags. To get the flags run:

$ pkg-config --cflags --libs libola
-I/usr/local/include  -L/usr/local/lib -lola -lolacommon -lprotobuf 

If you're using autotools, you can add

PKG_PROG_PKG_CONFIG
PKG_CHECK_MODULES(libola, [libola >= 0.8.26]) 

to configure.ac

Different Client APIs

If you want to:

Only send DMX

Use StreamingClient.h . This sets up a unidirectional channel between the client and the OLA server.

Do any of the following send DMX, receive DMX, patch ports, control RDM devices

Use the OlaCallbackClient which can be setup using the OlaCallbackClientWrapper

OlaClient - DEPRECATED, DO NOT USE

The OlaClient class provides a SetObserver() method which takes an object of type OlaClientObserver. When calls to the OlaClient object complete, the appropriate method is call on the OlaClientObserver object. The following diagrams attempts to show this (the red parts are the bit the client application provides).

 class Observer: public ola::OlaClientObserver {
   public
     Observer() {}:
     void ForceRDMDiscoveryComplete(unsigned int universe,
                                    const string &error) {
         // something here
     }
 };
 
 Observer observer();
 unsigned int universe = 1;
 client->SetObserver(&observer);
 client->ForceDiscovery(universe);

OlaCallbackClient

On the other hand, the OlaCallbackClient takes a Callback object whenever a method is called.

 void ForceRDMDiscoveryComplete(unsigned int universe, string &error) {
   // fill this in
 }
 
 unsigned int universe = 1;
 client->ForceDiscovery(
     universe,
     ola::NewSingleCallback(&ForceRDMDiscoveryComplete, universe));

The Callbacks used in OlaCallbackClient are much more flexible than the Observer model in OlaClient. It's recommended that all new code use OlaCallbackClient.

Wrapper Classes

The constructors of both OlaClient and OlaCallbackClient take a pointer to a Connected Socket object. While this is useful for testing, most clients don't care about the details of creating a socket, connecting to the OLA server etc. This is where the client wrapper classes come in. They take care of setting up the connection to the server and provide a way to access the OlaClient* object and the underlying SelectServer object.

 #include <ola/OlaClientWrapper.h>
 #include <ola/OlaCallbackClient.h>
 
 int main() {
   ola::OlaCallbackClientWrapper wrapper;
   if (!wrapper.Setup())
     //setup failed
     return -1;
 
   // Get the underlying OlaCallbackClient object
   ola::OlaCallbackClient *client = wrapper.GetClient();
 
   // Get the underlying SelectServer object
   ola::io::SelectServer *ss = client->GetSelectServer();
 
   // do stuff here
 }

Note that for historical reasons, the OlaClientWrapper class is also referred to as SimpleClient.

Example I: Sending an DMX update using the StreamingClient

The simplest interface to OLA is the StreamingClient which just allows an application to send DMX data to the OLA server

#include <stdlib.h>
#include <ola/DmxBuffer.h>
#include <ola/Logging.h>
#include <ola/StreamingClient.h>

#include <iostream>                                                                                                          

using std::cout;
using std::endl;


int main(int argc, char *argv[]) {
  unsigned int universe = 1;  // universe to use for sending data
  unsigned int i;

  // turn on OLA logging
  ola::InitLogging(ola::OLA_LOG_WARN, ola::OLA_LOG_STDERR);

  // Create a new DmxBuffer to hold the data
  ola::DmxBuffer buffer;
  // set all channels to 0
  buffer.Blackout();

  // create a new client and set the Error Closure
  ola::StreamingClient ola_client;

  // Setup the client, this connects to the server
  if (!ola_client.Setup()) {
    cout << "Setup failed" << endl;
    exit(1);
  }
 
  // send the data to the ola server 
  for (i = 0; i < 100; i++) {
    buffer.SetChannel(0, i);
    if (!ola_client.SendDmx(universe, buffer)) {
      cout << "Send DMX failed" << endl;
      exit(1);
    }
    usleep(20000);   // sleep for 20ms between updates                                                                            
  }

  // close the connection
  ola_client.Stop();
  return 0;
}

Note there isn't any need to poll for updates from the server in this case because the communication is all one way (client -> server). This also means that you can't receive DMX data or modify state on the server (like port patching and device parameters) when using the StreamingClient.

Also take note of the call to usleep(). Without this you're likely to fill the network socket buffer at which point the client code will detect an error and close the socket (you may need to increase the number of loop iterations to trigger this).

Example II: Sending an DMX update using the SimpleClient

This example connects to the OLA Server and sends one DMX update just like the example above but instead uses OlaClient & OlaClientWrapper. It's about as basic as you can get because it requires no callbacks and exits immediately.

#include <ola/DmxBuffer.h>
#include <ola/Logging.h>
#include <ola/OlaClientWrapper.h>
#include <ola/OlaClient.h>
 
int main() {
  ola::InitLogging(ola::OLA_LOG_WARN, ola::OLA_LOG_STDERR);
  ola::DmxBuffer buffer;
  buffer.Blackout();
  // some dummy dmx data
  buffer.SetChannel(0, 255);
 
  ola::OlaClientWrapper wrapper;
  if (!wrapper.Setup())
    //setup failed
    return -1;

  // Get the underlying OlaClient object
  ola::OlaClient *client = wrapper.GetClient();

  // Send the DMX data 
  client->SendDmx(1, buffer);
}

Example III: Sending multiple updates

The previous example only sent a single DMX update before exiting. This next example adds a timeout which sends DMX every 50ms. This introduces a new class SelectServer which is used for registering timeouts.

 #include <ola/DmxBuffer.h>
 #include <ola/io/SelectServer.h>
 #include <ola/Logging.h>
 #include <ola/OlaClient.h> 
 #include <ola/OlaClientWrapper.h>
 #include <ola/Callback.h>
 
 // Maximum value of a dmx channel
 static const unsigned int MAX_DMX = 255;
 // How often to send updates
 static const unsigned int TIMEOUT_MS = 50;
 
 class DmxTimeout {
   public:
     DmxTimeout(ola::OlaClient *client): m_tick(0),
                                         m_client(client) {
       m_buffer.Blackout();
       m_buffer.SetChannel(0, MAX_DMX);
     }
 
     // Called on timeout
     bool SendDmx() {
       m_buffer.SetChannel(1, m_tick);
       m_buffer.SetChannel(2, MAX_DMX - m_tick);
       m_client->SendDmx(1, m_buffer);
       m_tick++;
       m_tick %= MAX_DMX + 1;
       // we must return true else we'll get canceled
       return true;
     }
   private:
     unsigned int m_tick;
     ola::DmxBuffer m_buffer;
     ola::OlaClient *m_client;
 };
 
 int main() {
   ola::InitLogging(ola::OLA_LOG_WARN, ola::OLA_LOG_STDERR);
   ola::OlaClientWrapper wrapper;
 
   if (!wrapper.Setup())
     return -1;
 
   // Create a timeout and register it
   DmxTimeout timeout(wrapper.GetClient());
   ola::io::SelectServer *ss = wrapper.GetSelectServer();
   ss->RegisterRepeatingTimeout(TIMEOUT_MS,
                                ola::NewCallback(&timeout, &DmxTimeout::SendDmx));
 
   // Start the main loop
   ss->Run();
 }

In this example we create DmxTimeout class whose SendDmx() method is called every time the timer expires.

The other important part here is the SelectServer. As well as the RegisterRepeatingTimeout method we've used above, this can also be used to register sockets so we can respond to network activity. The Run() method starts the main event processing loop which will halt if an error occurs or Terminate() is called.

Example IV: Receiving DMX data

The forth example shows how to listen and respond to event from the OLA server.

#include <ola/Logging.h>
#include <ola/OlaClientWrapper.h>

static const unsigned int UNIVERSE = 1;

void NewDmx(unsigned int universe,
            const DmxBuffer &data, 
            const std::string &error) {
  OLA_INFO << "Received " << (int) data.Size() <<
    " channels for universe " << (int) universe;
}

int main() {
  ola::InitLogging(ola::OLA_LOG_INFO, ola::OLA_LOG_STDERR);
  ola::OlaCallbackClient wrapper;

  if (!wrapper.Setup())
    return -1;

  ola::OlaClient *client = simple_client.GetClient();
  // Set the observer and register our interest in this universe
  client->SetDmxCallback(UNIVERSE, NewCallback(&NewDmx));
  client->RegisterUniverse(UNIVERSE, ola::REGISTER);
  simple_client.GetSelectServer()->Run();
}

Example V: More complex client

The above is all well and good but what if the main application has it's own event processing loop? An example of this is a GTK/Glib application which uses GMainLoop.

On the other hand, what if you're not connecting to the LLA Server over TCP? Sometimes it may be desirable to embed the LLA server within the main application.

Bypassing the SimpleClient and using OlaClient directly addresses both these problems.

#include <ola/OlaClient.h>
#include <ola/io/SelectServer.h>
#include <ola/network/Socket.h>

using ola::io::PipeSocket;

int main() {
  // Create the select server
  ola::io::SelectServer ss;

  // Create the pipe socket to talk to the server on
  PipeSocket *pipe_socket = new PipeSocket();
  if (!pipe_socket->Init())
    return -1;

  // Remember to add this socket to the SelectServer
  ss.AddSocket(pipe_socket);

  // Setup the OlaClient
  ola::OlaClient client(pipe_socket);
  if (!client.Setup())
    return -1;

  // At this point the client is setup. We then need to setup the LLAServer
  // ...

  // Once that is done we add the pipe as a new connection
  ola_server->NewConnection(m_pipe_socket->OppositeEnd());
}

This example shows how we can create our own instance of a ConnectedSocket (PipeSocket is a subclass of ConnectedSocket) and pass it to the OlaClient to use. This code is very similar to what SimpleClient does under the hood.