The way we interact with TaskContexts during development of an Orocos application is through the deployer . This application consists of the DeploymentComponent which is responsible for creating applications out of component libraries and the DeploymentComponent which is a powerful console tool which helps you to explore, execute and debug componentss in running programs.

The TaskBrowser uses the GNU readline library to easily enter commands to the tasks in your system. This means you can press TAB to complete your commands or press the up arrow to scroll through previous commands.

You can start the deployer in any directory like this:

 $ deployer-gnulinux
				

or in a ROS environment:

 $ rosrun ocl deployer-gnulinux
				

This is going to be your primary tool to explore the Orocos component model so get your seatbelts fastened!

Now let's start the HelloWorld application we just created with orocreate-pkg.

Create an 'helloworld.ops' Orocos Program Script (ops) file with these contents:

 require("print")        // necessary for 'print.ln'
 import("HelloWorld")    // 'HelloWorld' is a directory name to import      

 print.ln("Script imported HelloWorld package:")
 displayComponentTypes() // Function of the DeploymentComponent
 
 loadComponent("Hello", "HelloWorld") // Creates a new component of type 'HelloWorld'
 print.ln("Script created Hello Component with period: " + Hello.getPeriod() )

and load it into the deployer using this command: $ deployer-gnulinux -s helloworld.ops -linfo This command imports the HelloWorld package and any component library in there. Then it creates a component with name "Hello". We call this a dynamic deployment, since the decision to create components is done at run-time.

You could also create your component in a C++ program. We call this static deployment, since the components are fixed at compilation time. The figure below illustrates this difference:

Figure 2.2.  Dynamic vs static loading of components

The 'helloworld' executable is a static deployment of one component in a process, which means it is hard-coded in the helloworld.cpp file. In contrast, using the deployer application allows you to load a component library dynamically.

The output of the deployer should be similar to what we show below. Finally, type cd Hello to start with the exercise.

0.000 [ Info   ][Logger] Real-time memory: 14096 bytes free of 20480 allocated.
0.000 [ Info   ][Logger] No RTT_COMPONENT_PATH set. Using default: .../rtt/install/lib/orocos
0.000 [ Info   ][Logger] plugin 'rtt' not loaded before.

...

0.046 [ Info   ][Logger] Loading Service or Plugin scripting in TaskContext Deployer
0.047 [ Info   ][Logger] Found complete interface of requested service 'scripting'
0.047 [ Info   ][Logger] Running Script helloworld.ops ...
0.050 [ Info   ][DeploymentComponent::import] Importing directory .../HelloWorld/lib/orocos/gnulinux ...
0.050 [ Info   ][DeploymentComponent::import] Loaded component type 'HelloWorld'
Script imported HelloWorld package:
I can create the following component types: 
   HelloWorld
   OCL::ConsoleReporting
   OCL::FileReporting
   OCL::HMIConsoleOutput
   OCL::HelloWorld
   OCL::TcpReporting
   OCL::TimerComponent
   OCL::logging::Appender
   OCL::logging::FileAppender
   OCL::logging::LoggingService
   OCL::logging::OstreamAppender
   TaskContext
0.052 [ Info   ][Thread] Creating Thread for scheduler: 0
0.052 [ Info   ][Hello] Thread created with scheduler type '0', priority 0, cpu affinity 15 and period 0.
HelloWorld constructed !
0.052 [ Info   ][DeploymentComponent::loadComponent] Adding Hello as new peer:  OK.
Script created Hello Component with period: 0
0.053 [ Info   ][Thread] Creating Thread for scheduler: 0
0.053 [ Info   ][TaskBrowser] Thread created with scheduler type '0', priority 0, cpu affinity 15 and period 0.
   Switched to : Deployer
0.053 [ Info   ][Logger] Entering Task Deployer

  This console reader allows you to browse and manipulate TaskContexts.
  You can type in an operation, expression, create or change variables.
  (type 'help' for instructions and 'ls' for context info)
    TAB completion and HISTORY is available ('bash' like)

Deployer [S]> cd Hello
   Switched to : Hello
Hello [S]>
      

The first [ Info ] lines are printed by the Orocos Logger, which has been configured to display informative messages to console with the -linfo program option. Normally, only warnings or worse are displayed by Orocos. You can always watch the log file 'orocos.log' in the same directory to see all messages. After the [Log Level], the [Origin] of the message is printed, and finally the message itself. These messages leave a trace of what was going on in the main() function before the prompt appeared.

Depending on what you type, the TaskBrowser will act differently. The built-in commands cd, help, quit, ls etc, are seen as commands to the TaskBrowser itself, if you typed something else, it tries to execute your command according to the Orocos scripting language syntax.

Hello[R] > 1+1
   = 2
      

A component's interface consists of: Attributes and Properties, Operations, and Data Flow ports which are all public. The class TaskContext groups all these interfaces and serves as the basic building block of applications. A component developer 'builds' these interfaces using the instructions found in this manual.

Figure 2.3.  Schematic Overview of the Hello Component.

Our hello world component.

To display the contents of the current component, type ls, and switch to one of the listed peers with cd, while cd .. takes you one peer back in history. We have two peers here: the Deployer and your component, Hello.

Hello [S]> ls

 Listing TaskContext Hello[S] :

 Configuration Properties: (none)

 Provided Interface:
  Attributes   : (none)
  Operations   : activate cleanup configure error getCpuAffinity getPeriod inFatalError inRunTimeError isActive isConfigured isRunning setCpuAffinity setPeriod start stop trigger update 

 Data Flow Ports: (none)

 Services: 
(none)

 Requires Operations :  (none)
 Requests Services   :  (none)

 Peers        : (none)
Hello [S]>
      

	Note
	To get a quick overview of the commands, type help.

The first line shows the status between square brackets. The [S] here means that the component is in the stopped state. Other states can be 'R' - Running, 'U' - Unconfigured, 'E' - run-time Error, 'F' - Fatal error, 'X' - C++ eXception in user code.

First you get a list of the Properties and Attributes (alphabetical) of the current component. Properties are meant for configuration and can be written to disk. Attributes export a C++ class value to the interface, to be usable by scripts or for debugging and are not persistent.

Next, the operations of this component are listed: each component has some universal functions like activate, start, getPeriod etc.

You can see that the component is pretty empty: no data flow ports, services or peers. We will add some of these right away.

To get an overview of the Task's interface, you can use the help command, for example help this or help this.activate or just short: help activate

Hello [R]> help this

Printing Interface of 'Hello' :

 activate( ) : bool
   Activate the Execution Engine of this TaskContext (= events and commands).
 cleanup( ) : bool
   Reset this TaskContext to the PreOperational state (write properties etc).
...
   Stop the Execution Engine of this TaskContext.

Hello [R]> help getPeriod
 getPeriod( ) : double
Get the configured execution period. -1.0: no thread associated, 0.0: non periodic, > 0.0: the period.

Hello [R]> 
      

Now we get more details about the operations registered in the public interface. We see now that the getPeriod operations takes no arguments You can invoke each operation right away.

Hello [R]> getPeriod()
 = 0
      

Operations are called directly and the TaskBrowser prints the result. The return value of getPeriod() was a double, which is 0. This works just like calling a 'C' function. You can express calling explicitly by writing: getPeriod.call().

When an operation is sent to the Hello component, another thread will execute it on behalf of the sender. Each sent method returns a SendHandle object.

Hello [R]> getPeriod.send()
 = (unknown_t)
      

The returned SendHandle must be stored in a SendHandle attribute to be useful:

Hello [R]> var SendHandle sh
Hello [R]> sh = getPeriod.send()
 = true
Hello [R]> sh.collectIfDone( ret )
 = SendSuccess
Hello [R]> ret
 = 0
      

SendHandles are further explained down the document. They are not required understanding for a first discovery of the Orocos world.

Besides calling or sending component methods, you can alter the attributes of any task, program or state machine. The TaskBrowser will confirm validity of the assignment with the contents of the variable. Since Hello doesn't have any attributes, we create one dynamically:

Hello [R]> var string the_attribute = "HelloWorld"
Hello [R]> the_attribute
 = Hello World
Hello [R]> the_attribute = "Veni Vidi Vici !"
 = "Veni Vidi Vici !"
Hello [R]> the_attribute
 = Veni Vidi Vici !
      

The Data Ports allow seamless communication of calculation or measurement results between components. Adding and using ports is described in Section 3.3, “Data Flow Ports”.

Last but not least, hitting TAB twice, will show you a list of possible completions, such as peers, services or methods.

TAB completion works even across peers, such that you can type a TAB completed command to another peer than the current peer.

In order to quit the TaskBrowser, enter quit:

 Hello [R]> quit

1575.720 [ Info   ][ExecutionEngine::setActivity] Hello is disconnected from its activity.
1575.741 [ Info   ][Logger] Orocos Logging Deactivated.      
      

The TaskBrowser Component is application independent, so that your end user-application might need a more suitable interface. However, for testing and inspecting what is happening inside your real-time programs, it is a very useful tool. The next sections show how you can add properties, methods etc to a TaskContext.

	Note
	If you want a more in-depth tutorial, see the rtt-exercises package which covers each aspect also shown in this manual.

The user application code is filled in by inheriting from the TaskContext and implementing the 'Hook' functions. There are five such functions which are called when a TaskContext's state changes.

The user may insert his configuration-time setup/cleanup code in the configureHook() (read XML, print status messages etc.) and cleanupHook() (write XML, free resources etc.).

The run-time (or: real-time) application code belongs in the startHook(), updateHook() and stopHook() functions.

class MyTask 
    : public TaskContext
  {
  public:
       MyTask(std::string name) 
         : TaskContext(name)
       {
          // see later on what to put here.
       }
  
       /**
        * This function is for the configuration code.
        * Return false to abort configuration.
        */
       bool configureHook() {
          // ...
          return true;
       }

       /**
        * This function is for the application's start up code.
        * Return false to abort start up.
        */
       bool startHook() {
          // ...
          return true;
       }

       /**
        * This function is called by the Execution Engine.
        */
       void updateHook() {
          // Your component's algorithm/code goes in here.
       }

       /**
        * This function is called when the task is stopped.
        */
       void stopHook() {
          // Your stop code after last updateHook()
       }

       /**
        * This function is called when the task is being deconfigured.
        */
       void cleanupHook() {
          // Your configuration cleanup code
       }
  };

	Important
	By default, the TaskContext enters the Stopped state (Figure 2.5, “ TaskContext State Diagram ”) when it is created, which makes configure() an optional call.

If you want to force the user to call configure() of your TaskContext, set the TaskState in your constructor as such:

class MyTask 
    : public TaskContext
  {
  public:
       MyTask(std::string name) 
         : TaskContext(name, PreOperational) // demand configure() call.
       {
          //...
       }
  };

When configure() is called, the configureHook() (which you must implement!) is executed and must return false if it failed. The TaskContext drops to the PreOperational state in that case. When configureHook() succeeds, the TaskContext enters the Stopped state and is ready to run.

A TaskContext in the Stopped state (Figure 2.5, “ TaskContext State Diagram ”) may be start()'ed upon which startHook() is called once and may abort the start up sequence by returning false. If true, it enters the Running state and updateHook() is called (a)periodically by the ExecutionEngine, see below. When the task is stop()'ed, stopHook() is called after the last updateHook() and the TaskContext enters the Stopped state again. Finally, by calling cleanup(), the cleanupHook() is called and the TaskContext enters the PreOperational state.

The functionality of a component, i.e. its algorithm, is executed by its internal Execution Engine. To run a TaskContext, you need to use one of the ActivityInterface classes from the RTT, most likely Activity. This relation is shown in Figure 2.6, “ Executing a TaskContext ”. The Activity class allocates a thread which executes the Execution Engine. The chosen Activity object will run the Execution Engine, which will in turn call the application's hooks above. When created, the TaskContext is assigned the Activity by default. It offers an internal thread which can receive messagse and process events but is not periodicly executing updateHook().

Figure 2.6.  Executing a TaskContext

You can make a TaskContext 'active' by creating an Activity object which executes its Execution Engine.

  #include <rtt/Activity.hpp>

  using namespace RTT;

  TaskContext* a_task = new MyTask("the_task");
  // Set a periodic activity with priority=5, period=1000Hz
  a_task->setActivity( new Activity( 5, 0.001 ));
  // ... start the component:
  a_task->start(); 
  // ...
  a_task->stop();

  // In your TaskContext's configureHook():
  bool configureHook() {
      return this->setPeriod(0.001); // set to 1000Hz execution mode.
  }

  class MyTask 
    : public TaskContext
  {
  public:
       // ...
  
       /**
        * This function is periodically called.
        */
       void updateHook() {
          // Your algorithm for periodic execution goes inhere
          double result;
          if ( inPort.read(result) == NewData )
              outPort.write( result * 2.0 ); // only write if new data arrived.
       }

  };

3.2.2. Default Component Execution Semantics

A TaskContext is run by default by a non periodic RTT:Activity object. This is useful when updateHook() only needs to process data when it arrives on a port or must wait on network connections or does any other blocking operation.

Upon start(), the Execution Engine waits for new methods or data to come in to be executed. Each time such an event happens, the user's application code (updateHook()) is called after the Execution Engine did its work.

An updateHook() function of a non periodic task could look like:

  class MyTask 
    : public TaskContext
  {
  public:
       // ...
  
       /**
        * This function is only called by the Execution Engine 
        * when 'trigger()' is called or an event or command arrives.
        */
       void updateHook() {
            // Your blocking algorithm goes inhere
            char* data;
            double timeout = 0.02; // 20ms
            int rv = my_socket_read(data, timeout);

            if (rv == 0) {
               // process data
               this->stateUpdate(data);
            }
	    
            // This is special for non periodic activities, it makes
            // the TaskContext call updateHook() again after
            // commands and events are processed.
            this->getActivity()->trigger(); 
       }

  };

Warning

Non periodic activities should be used with care and with much thought in combination with scripts (see later). The ExecutionEngine will do absolutely nothing if no asynchronous methods or asynchronous events or no trigger comes in. This may lead to surprising 'bugs' when program scripts or state machine scripts are executed, as they will only progress upon these events and seem to be stalled otherwise.

You can find more detailed information in Section 2, “Activities” in the CoreLib reference.

Purpose

A component has ports in order to send or receive a stream of data. The algorithm writes Output ports to publish data to other components, while input ports allow an algorithm to receive data from other components. A component can be woken up if data arrives at one or more input ports or it can 'poll' for new data on its input ports. Reading and writing data ports is always real-time and thread-safe, on the condition that copying your data (i.e. your operator= ) is as well.

Each component defines its data exchange ports and connections transmit data from one port to another. A Port is defined by a name, unique within that component, the data type it wants to exchange and if its for reading (Input) or writing (Output) data samples. Finally, you can opt that new data on selected Input ports wake up your task. The example below shows all these possibilities.

Each connection between an Output port and an Input port can be tuned for your setup: buffering of data, thread-safety and initialisation of the connection are parameters provided by the user when the connection is created. We call these Connection Policies and use the ConnPolicy object when creating the connection between ports.

Figure 2.7.  Data flow ports are connected with a connection policy

This figure shows that input and output ports can be connected in an N:M way. See Section 4.2, “Setting up the Data Flow” on how to connect ports and which connection policy to choose.

3.3.2. Setting up the Data Flow Interface

Any kind of data can be exchanged (also user defined C/C++ types) but for readability, only the 'double' C type is used here.

  #include <rtt/Port.hpp>
  using namespace RTT;

  class MyTask
    : public TaskContext
  {
    // Input port: We'll let this one wake up our thread
    InputPort<double> evPort;

    // Input port: We will poll this one
    InputPort<double> inPort;

    // Output ports are allways 'send and forget'
    OutputPort<double> outPort;
  public:
    // ...
    MyTask(std::string name)
      : TaskContext(name)
    {
       // an 'EventPort' is an InputPort which wakes our task up when data arrives.
       this->ports()->addEventPort( "evPort", evPort ).doc( "Input Port that raises an event." );

       // These ports do not wake up our task
       this->ports()->addPort( "inPort", inPort ).doc( "Input Port that does *not* raise an event." );
       this->ports()->addPort( "outPort", outPort ).doc( "Output Port, here write our data to." );

       // more additions to follow, see below
     }

     // ...
  };

The example starts with declaring all the ports of MyTask. A template parameter '<double>' specifies the type of data the task wants to exchange through that port. Logically, if input and output are to be connected, they must agree on this type. The name is given in the addPort() function. This name can be used to 'match' ports between connected tasks ( using 'connectPorts', see Section 4, “Connecting Services” ), but it is possible and preferred to connect Ports with different names using the Orocos deployer.

There are two ways to add a port to the TaskContext interface: using addPort() or addEventPort(). In the latter case, new data arriving on the port will wake up ('trigger') the activity of our TaskContext and updateHook() get's executed.

	Note
	Only InputPort can be added as EventPort and will cause your component to be triggered (ie wake up and call updateHook).

  OutputPort<std::vector<double> > myport("name");

  // create an example data sample of size 10:
  std::vector<double> example(10, 0.0);

  // show it to the port (this is a not real-time operation):
  myport.setDataSample( example );
	    
  // Now we are fine ! All items sent into the port of size 10 or less will
  // be passed on in hard real-time.  
  myport.write( example ); // hard real-time.

  #include <rtt/Port.hpp>
  using namespace RTT;

  class MyTask
    : public TaskContext
  {
     // ...Constructor sets up Ports, see above.

     bool startHook() {
       // Check validity of (all) Ports:
       if ( !inPort.connected() ) {
          // No connection was made, can't do my job !
          return false;
       }
       if ( !outPort.connected() ) {
          // ... not necessarily an error, a connection may be
          // made while we are running.
       }
       return true;
     }

     /**
      * Note: use updateHook(const std::vector<PortInterface*>&)
      * instead for having information about the updated event 
      * driven ports.
      */
     void updateHook() {

       double val = 0.0; 

       // Possible return values are: NoData, OldData and NewData.
       if ( inPort.read(val) == RTT::NewData ) {
         // update val...
         outPort.write( val );
       }
     }
     // ...
  };

3.3.5. Using Data Flow in Scripts

When a Port is added, it becomes available to the Orocos scripting system such that (part of) the calculation can happen in a script. Also, the TaskBrowser can then be used to inspect the contents of the DataFlow on-line.

	Note
	In scripting, it is currently not yet possible to know which event port woke your task up.

A small program script could be loaded into MyTask with the following contents:

  program MyControlProgram {
    var double the_K  = K        // read task property, see later.
    var double setp_d

    while ( true ) {
      if ( SetPoint_X.read( setp_d ) != NoData ) {   // read Input Port
        var double in_d = 0.0;
        Data_R.read( in_d )             // read Input Port
        var double out_d = (setp_d - in_d) * the_K  // Calculate
        Data_W.write( out_d )          // write Data Port
      }
      yield       // this is a 'yield' point to avoid inifinite spinning.
    }
  } 

The program "MyControlProgram" starts with declaring two variables and reading the task's Property 'K'. Then it goes into an endless loop, trying to Pop a set point value from the "SetPoint_X" Port. If that succeeds (new or old data present) the "Data_R" Port is read and a simple calculation is done. The result is written to the "Data_W" OutputPort and can now be read by the other end(s). Alternatively, the result may be directly used by the Task in order to write it to a device or any non-task object. You can use methods (below) to send data from scripts back to the C++ implementation.

Remark that the program is executed within the thread of the component. In order to avoid the endless loop, a 'wait' point must be present. The "yield" command inserts such a wait point and is part of the Scripting syntax. If you plan to use Scripting state machines, such a while(true) loop (and hence wait point) is not necessary. See the Scripting Manual for a full overview of the syntax.

	Purpose
	A task's operations define which functions a component offers. Operations are grouped in 'services', much like C++ class methods are grouped in classes. OperationCallers are helper objects for calling operations.

Operations are C/C++ functions that can be used in scripting or can be called from another process or accross a network. They take arguments and return a value. The return value can in return be used as an argument for other Operations or stored in a variable.

To add a C/C++ function to the operation interface, you only need to register it with addOperation(), defined in Service.

  #include <rtt/Operation.hpp>
  using namespace RTT;

  class MyTask
    : public TaskContext
  {
    public:
    void reset() { ... }
    string getName() const { ... }
    double changeParameter(double f) { ... }
    // ...
 
    MyTask(std::string name)
      : TaskContext(name),
    {
       // Add the method objects to the method interface:
       this->addOperation( "reset",  &MyTask::reset, this, OwnThread)
                                   .doc("Reset the system.");
       this->addOperation( "getName", &MyTask::getName, this, ClientThread)
                                   .doc("Read out the name of the system.");
       this->addOperation( "changeParameter", &MyTask::changeParameter, this, OwnThread)
                                   .doc("Change a parameter, return the old value.")
                                   .arg("New Value", "The new value for the parameter."); 

       // more additions to follow, see below
     }
     // ...
  };

In the above example, we wish to add 3 functions to the method interface: reset, getName and changeParameter. You need to pass the name of the function, address (function pointer) of this function and the object on which it must be called (this) to addOperation. Optionally, you may document the operation with .doc("...") and each argument with a .arg() call.

Using this mechanism, any method of any class can be added to a task's method interface, not just functions of a TaskContext You can also add plain C functions, just omit the this pointer.

As the last argument to addOperation, a flag can be passed which can be OwnThread or ClientThread. This allows the component implementer to choose if the operation, when called, is executed in the thread of the ExecutionEngine, or in the thread of the caller (i.e. the Client). This choice is hidden from the user of our operations. It allows us to choose who gets the burden of the execution of the function, but also allows to synchronize operation calls with the execution of updateHook(). Summarized in a table:

The choice of this type is completely up to the implementor of the component and can be made independently of how it will be used by its clients. Clients can indicate the same choice indepenently: they can Call or Send an operation. This is explained in the next two sections.

void(int,double)

  // create a method:
  TaskContext* a_task_ptr = getPeer("ATask");
  OperationCaller<void(void)> my_reset_meth 
       = a_task_ptr->getOperation("reset"); // void reset(void)

  // Call 'reset' of a_task:
  reset_meth();  

  // Send 'reset' of a_task:
  SendHandle<void(void)> handle = reset_meth.send();  

  // Send 'reset' of a_task:
  SendHandle<void(void)> handle = reset_meth.send();  

  // polling for reset() to complete:
  while (handle.collectIfDone() == SendNotReady )
      sleep(1);

  // blocking for reset() to complete:
  handle = reset_meth.send();
  SendStatus ss = handle.collect();
  if (ss != SendSuccess) {
      cout << "Execution of reset failed." << endl;
  }

  // retrieving the return value is not possible for a void(void) method. 

  // used to hold the return value of getName:
  string name;
  OperationCaller<string(void)> name_meth = 
    a_task_ptr->getOperation("getName"); // string getName(void)

  // Call 'getName' of a_task:
  name = name_meth();
  // Equivalent to:
  name = name_meth.call();

  cout << "Name was: " << name << endl; 

  // Send 'getName' to a_task:
  SendHandle<string(void)> nhandle = name.send();

  // collect takes the return value of getName() as first argument and fills it in:
  SendStatus ss = nhandle.collect(name);
  if (ss == SendSuccess) {
    cout << "Name was: " << name << endl; 
  }

  assert( name == nhandle.ret() ); // ret() returns the same as getName() returned. 

  // hold return value of changeParameter:
  double oldvalue;
  OperationCaller<double(double)> mychange =
      a_task_ptr->getOperation("changeParameter"); // double changeParameter(double)

  // Call 'changeParameter' of a_task with argument '1.0'
  oldvalue = mychange( 1.0 );
  // Equivalent to:
  oldvalue = mychange.call( 1.0 );

  // Send 'changeParameter' to a_task:
  SendHandle<double(double)> chandle = changeParameter.send( 2.0 )

  SendStatus ss = chandle.collectIfDone( oldvalue );
  if (ss == SendSuccess) {
      cout << "Oldvalue was: " << oldvalue << endl;
  }  

  OperationCaller<double(double)> mychange = ...;
  assert( mychange.ready() );

 // call:
  var double oldvalue
  ATask.changeParameter( 0.1 )
  // or :
  set oldvalue = ATask.changeParameter( 0.1 ) // store return value 

  // send:
  var SendHandle handle;
  var SendStatus ss;
  handle = ATask.changeParameter.send( 2.0 );

  // collect non-blocking:
  while ( handle.collectIfDone( oldvalue ) ) 
     yield // see text below.
   
  // collect blocking:
  handle.collect( oldvalue ); // see text below.

Purpose

A task's properties are intended to configure and tune a task with certain values. Properties have the advantage of being writable to an XML format, hence can store 'persistent' state. For example, a control parameter. Attributes reflect a C++ class variable in the interface and can be read and written during run-time by a program script, having the same data as if it was a C++ function. Reading and writing properties and attributes is real-time but not thread-safe and should for a running component be limited to the task's own activity.

A TaskContext may have any number of attributes or properties, of any type. They can be used by programs in the TaskContext to get (and set) configuration data. The task allows to store any C++ value type and also knows how to handle Property objects. Attributes are plain variables, while properties can be written to and updated from an XML file.

  #include <rtt/Property.hpp>
  #include <rtt/Attribute.hpp>

  class MyTask
    : public TaskContext
  {
     // we will expose these:
     bool aflag;
     int max;

     double pi;

     std::string param;
     double value;
  public:
    // ...
    MyTask(std::string name)
      : TaskContext(name),
        param("The String"),
        value( 1.23 ),
        aflag(false), max(5), pi(3.14)
    {
       // other code here...

       // attributes and constants don't take a .doc() description.
       this->addAttribute( "aflag", aflag );
       this->addAttribute( "max", max );
      
       this->addConstant( "pi", pi );

       this->addProperty( "Param", param ).doc("Param Description");
       this->addProperty( "Palue", value ).doc("Value Description");
     }
     // ...
  };

  Attribute<bool> the_flag = a_task->getValue("aflag");
  assert( the_flag.ready() );

  bool result = the_flag.get();
  assert( result == false );

  Attribute<int> the_max = a_task->attributes()->getAttribute("max");
  assert( the_max.ready() );
  the_max.set( 10 );
  assert( the_max.get() == 10 );

3.5.3. Accessing Task Attributes in Scripts

A program script can access the above attributes simply by naming them:

  // a program in "ATask" does :
  var double pi2 = pi * 2.
  var int    myMax = 3
  set max = myMax

  set Param = "B Value"

  // an external (peer task) program does :
  var double pi2 = ATask.pi * 2.
  var int    myMax = 3
  set ATask.max = myMax	

When trying to assign a value to a constant, the script parser will throw an exception, thus before the program is run.

	Important
	The same restrictions of Section 3.4.5, “Operation Argument and Return Types” hold for the attribute types, when you want to access them from program scripts.

See also Section 5, “Attributes” in the Orocos CoreLib reference.

In addition to the PreOperational, Stopped and Running TaskContext states, you can use two additional states for more advanced component behaviour: the Exception, FatalError and the RunTimeError states. The first two are shown in Figure 2.8, “ Extended TaskContext State Diagram ”.

Figure 2.8.  Extended TaskContext State Diagram

This figure shows the extended state diagram of a TaskContext. This is Figure 2.5, “ TaskContext State Diagram ” extended with two more states: Exception and FatalError.

The FatalError state is entered whenever the TaskContext's fatal() function is called, and indicates that an unrecoverable error occured. The ExecutionEngine is immediately stopped and no more functions are called. This state can not be left and the only next step is destruction of the component (hence 'Fatal').

When an exception happens in your code, the Exception state is entered. Depending on the TaskState, stopHook() and cleanupHook() will be called to give a chance to cleanup. This state is recoverable with the recover() function which drops your component back to the PreOperational state, from which it needs to be configured again.

It is possible that non-fatal run-time errors occur which may require user action on one hand, but do not prevent the component from performing it's task, or allow degraded performance. Therefor, in the Running state, one can make a transition to theRunTimeError sub-state by calling error(). See Figure 2.9, “ Possible Run-Time failure. ”.

Figure 2.9.  Possible Run-Time failure.

Th

When the application code calls error(), the RunTimeError state is entered and errorHook() is executed instead of updateHook(). If at some moment the component detects that it can resume normal operation, it calls the recover() function, which leads to the Running state again and in the next iteration, updateHook() is called again.

 class MyComponent : public TaskContext
 {
       int faults;
 public:
       MyComponent(const std::string &name) 
         : TaskContext(name), faults(0) 
       {}

 protected:
       // Read data from a buffer.
       // If ok, process data. When to many faults occur,
       // trigger a runtime error.
       void updateHook()
       {
            Data_t data;
            FlowStatus rv = input.read( data );
            if ( rv == NewData ) {
               this->stateUpdate(data);
               faults = 0;
               this->recover(); // may be an external supervisor calls this instead.
            } else {
               faults++;
               if (faults > 4)
                   this->error();
            }

       }

       // Called instead of updateHook() when in runtime error state.
       void errorHook()
       {
            this->updateHook(); // just call updateHook anyway.
       }

       // Called by updateHook()
       void stateUpdate(Data_t data)
       {
            // Check for corrupt data
            if ( checkData(data) == -1 ) {
                 this->fatalError(); // we will enter the FatalError state.
            } else {
                 // data is ok: update internal state...
            }
       }
 };

	Note
	The addPeer and connectPeers functions are used to connect TaskContexts and allow them to use each other's interface. The connectPorts function sets up the data flow between tasks.

We call connected TaskContexts "Peers" because there is no implied hierarchy. A connection from one TaskContext to its Peer can be uni- or bi-directional. In a uni-directional connection (addPeer ), only one peer can use the services of the other, while in a bi-directional connection (connectPeers), both can use each others services. This allows to build strictly hierarchical topological networks as well as complete flat or circular networks or any kind of mixed network.

Peers are connected as such (hasPeer takes a string argument ):

  // bi-directional :
  connectPeers( &a_task, &b_task );
  assert( a_task.hasPeer( &b_task.getName() ) 
          & b_task.hasPeer( &a_task.getName() ) );

  // uni-directional :
  a_task.addPeer( &c_task );
  assert( a_task.hasPeer( &c_task.getName() ) 
          & ! c_task.hasPeer( &a_task.getName() ) );

  // Access the interface of a Peer:
  OperationCaller<bool(void)> m = a_task.getPeer( "CTask" )->getOperation("aOperationCaller");
  // etc. See interface usage in previous sections.

Both connectPeers and addPeer allow scripts or C++ code to use the interface of a connected Peer. connectPeers does this connection in both directions.

From within a program script, peers can be accessed by merely prefixing their name to the member you want to access. A program within "ATask" could access its peers as such :

 // Script:
  var bool result = CTask.aOperation() 

The peer connection graph can be traversed at arbitrary depth. Thus you can access your peer's peers.

	Note
	In typical applications, the DeploymentComponent ('deployer') will form connections between ports using a program script or XML file. The manual method described below is not needed in that case.

Data Flow between TaskContexts can be setup by using connectPorts. The direction of the data flow is imposed by the input/output direction of the ports. The connectPorts(TaskContext* A, TaskContext* B) function creates a connection between TaskContext ports when both ports have the same name and type. It will never disconnect existing connections and only tries to add ports to existing connections or create new connections. The disadvantage of this approach is that you can not specify connection policies.

Instead of calling connectPorts, one may connect individual ports, such that different named ports can be connected and a connection policy can be set. Suppose that Task A has a port a_port, Task B a b_port and Task C a c_port (all are of type PortInterface&). Then connections are made as follows:

  // Create a connection with a buffer of size 10:
  ConnPolicy policy = RTT::ConnPolicy::buffer(10);
  a_port.connectTo( &b_port, policy );
  // Create an unbuffered 'shared data' connection:
  policy = RTT::ConnPolicy::data();
  a_port.connectTo( &c_port, policy );

The order of connections does not matter; the following would also work:

  b_port.connectTo( &a_port, policy ); // ok...
  c_port.connectTo( &a_port, policy ); // fine too.

Note that you can not see from this example which port is input and which is output. For readability, it is recommended to write it as:

output_port.connectTo( &input_port );

ConnPolicy are powerful objects that allow you to connect component ports just like you want them. You can use them to create connections over networks or to setup fast real-time inter-process communication.

Tasks can be disconnected from a network by invoking disconnect() on that task. It will inform all its peers that it has left the network and disconnect all its ports.

As was seen in Section 3.5, “The Attributes and Properties Interface”, Property objects can be added to a task's interface. To read and write properties from or to files, you can use the Marshalling service. It creates or reads files in the XML Component Property Format such that it is human readable and modifiable.

  // ...
  TaskContext* a_task = ...
            mname = ab->getName();
            mname = ab->getName();
  a_task->getProvider<Marshalling>("marshalling")->readProperties( "PropertyFile.cpf" );
  // ...
  a_task->getProvider<Marshalling>("marshalling")->writeProperties( "PropertyFile.cpf" ); 

In order to access a service, we need both the type of the provider, Marshalling and the run-time name of the service, by default "marshalling".

In the example, readProperties() reads the file and updates the task's properties and writeProperties() writes the given file with the properties of the task. Other functions allow to share a single file with multiple tasks or update the task's properties from multiple files.

The PropertyFile.cpf file syntax can be easily learnt by using writeProperties() and looking at the contents of the file. It will contain elements for each Property or PropertyBag in your task. Below is a component with five properties. There are three properties at the top level of which one is a PropertyBag, holding two other properties.

  #include <rtt/TaskContext.hpp>
#include <rtt/Property.hpp>
#include <rtt/PropertyBag.hpp>

  class MyTask
    : public TaskContext
  {
     
     int i_param;
     double d_param;
     PropertyBag sub_bag;
     std::string s_param;
     bool b_param;
  public:
    // ...
    MyTask(std::string name)
      : TaskContext(name),
        i_param(5 ),
        d_param(-3.0),
        s_param("The String"),
        b_param(false)
    {
       // other code here...

       this->addProperty("IParam", i_param ).doc("Param Description");
       this->addProperty("DParam", d_param ).doc("Param Description");
       this->addProperty("SubBag", sub_bag ).doc("SubBag Description");

       // we call addProperty on the PropertyBag object in order to 
       // create a hierarchy
       sub_bag.addProperty("SParam", s_param ).doc("Param Description");
       sub_bag.addProperty("BParam", b_param ).doc("Param Description");
     }
     // ...
  };

Using writeProperties() would produce the following XML file:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE properties SYSTEM "cpf.dtd">
<properties>

  <simple name="IParam" type="short">
    <description>Param Description</description>
    <value>5</value>
    </simple>
  <simple name="DParam" type="double">
    <description>Param Description</description>
    <value>-3.0</value>
  </simple>

  <struct name="SubBag" type="PropertyBag">
    <description>SubBag Description</description>
    <simple name="SParam" type="string">
      <description>Param Description</description>
      <value>The String</value>
    </simple>
    <simple name="BParam" type="boolean">
      <description>Param Description</description>
      <value>0</value>
    </simple>
  </struct>

</properties>

PropertyBags (nested properties) are represented as <struct> elements in this format. A <struct> can contain another <struct> or a <simple> property.

The following table lists the conversion from C++ data types to XML Property types.

Orocos supports two types of scripts:

Both are loaded at run-time into a task. The scripts are parsed to an object tree, which can then be executed by the ExecutionEngine of a task.

  program foo
  {
    var int i = 1
    var double j = 2.0
    changeParameter(i,j)
  }

  TaskContext* a_task = ...

  a_task->getProvider<Scripting>("scripting")->loadPrograms( "ProgramBar.ops" ); 

  foo.start()
  foo.pause()
  foo.step()
  foo.stop()
	

  var bool ret
  ret = foo.isRunning()
  ret = foo.inError()
  ret = foo.isPaused()
	

  set foo.i = 3
  var double oldj = foo.j
	

  scripting::ScriptingService* sa = dynamic_cast<scripting::ScriptingService*>(this->getService("scripting"));
  scripting::ProgramInterface* foo = sa->getProgram("foo");
  if (foo != 0) {
    bool result = foo->start(); // try to start the program !
    if (result == false) {
       // Program could not be started.
       // Execution Engine not running ?
    }
  }

  TaskContext* a_task = ...

  a_task->getProvider<Scripting>("scripting")->loadStateMachines( "StateMachineBar.osd" ); 

  machine.activate()
  machine.start()
  machine.pause()
  machine.step()
  machine.stop()
  machine.deactivate()
  machine.reset()
  machine.reactive()
  machine.automatic() // identical to start()
  machine.requestState("StateName")
	

  set machine.myParam = ...
	

Figure 2.10, “ Component Deployment Levels ” shows the distinction between the three levels of Component Deployment.

Figure 2.10.  Component Deployment Levels

Three levels of using or creating Components can be accomplished in Orocos: Not distributed, embedded distributed and fully distributed.

If your application will not use distributed components and requires a very small footprint, the TaskCore can be used. The Orocos primitives appear publicly in the interface and are called upon in a hard-coded way.

If you application requires a small footprint and distributed components, the C++ Interface of the TaskContext can be used in combination with a Distribution Library which does the network translation. It handles a predefined set of data types (mostly the 'C' types) and needs to be adapted if other data types need to be supported. There is no portable distribution library available.

If footprint is of no concern to your application and you want to distribute any component completely transparently, the TaskContext can be used in combination with a Remoting Library which does the network translation. A CORBA implementation of such a library is being developed on. It is a write-once, use-many implementation, which can pick up user defined types, without requiring modifications. It uses the Orocos Type System to manage user defined types.

A TaskCore is nothing more than a place holder for the Execution Engine and application code functions (configureHook(), cleanupHook(), startHook(), updateHook() and stopHook() ). The Component interface is built up by placing the Orocos primitives as public class members in a TaskCore subclass. Each component that wants to use this TaskCore must get a 'hard coded' pointer to it (or the interface it implements) and invoke the command, method etc. Since Orocos is by no means informed of the TaskCore's interface, it can not distribute a TaskCore.

Instead of putting the Orocos primitives in the public interface of a subclass of TaskCore, one can subclass a TaskContext and register the primitives to the Local C++ Interface. This is a reduced interface of the TaskContext, which allows distribution by the use of a Distribution Library.

The process goes as such: A component inherits from TaskContext and has some Orocos primitives as class members. Instead of calling:

  this->addOperation("name", &foo).doc("Description").arg("Arg1","Arg1 Description");

and providing a description for the primitive as well as each argument, one writes:

  this->addLocalOperation("name", &foo );

This functions does no more than a pointer registration, but already allows all C++ code in the same process space to use the added primitive.

In order to access the interface of such a Component, the user code may use:

  taskA->getLocalOperation("name");

You can only distribute this component if an implementation of a Distribution Library is present. The specification of this library, and the application setup is in left to another design document.

In case you are building your components as instructed in this manual, your component is ready for distribution as-is, given a Remoting library is used. The Orocos CORBA package implements such a Remoting library.

Using the three levels of deployment in one application is possible as well. To save space or execution efficiency, one can use TaskCores to implement local (hidden) functionality and export publicly visible interface using a TaskContext. Figure 2.11, “ Example Component Deployment. ” is an small example of a TaskContext which uses two TaskCores to delegate work to. The Execution Engines may run in one or multiple threads.

Figure 2.11.  Example Component Deployment.

Most projects have define their own task interfaces in C++. Assume you have a class with the following interface :

  class DeviceInterface
  {
  public:
    /**
     * Set/Get a parameter. Returns false if parameter is read-only.
     */
    virtual bool setParameter(int parnr, double value) = 0;
    virtual double getParameter(int parnr) const = 0;

    /**
     * Get the newest data.
     * Return false on error.
     */
    virtual bool updateData() = 0;
    virtual bool updated() const = 0;

    /**
     * Get Errors if any.
     */
    virtual int getError() const = 0;
  };

Now suppose you want to do make this interface available, such that program scripts of other tasks can access this interface. Because you have many devices, you surely want all of them to be accessed transparently from a supervising task. Luckily for you, C++ polymorphism can be transparently adopted in Orocos TaskContexts. This is how it goes.

  #include <rtt/TaskContext.hpp>
  #include <rtt/Operation.hpp>
  #include "DeviceInterface.hpp"

  class TaskDeviceInterface
    : public DeviceInterface,
      public TaskContext
  {
  public:
      TaskDeviceInterface()
        : TaskContext( "DeviceInterface" )
      {
         this->setup();
      }

      void setup() 
      {
         // Add client thread operations :
         this->addOperation("setParameter",
                    &DeviceInterface::setParameter, this, ClientThread)
                        .doc("Set a device parameter.")
	                .arg("Parameter", "The number of the parameter.")
	                .arg("New Value", "The new value for the parameter."); 

         this->addOperation("getParameter",
                     &DeviceInterface::getParameter, this, ClientThread)
                          .doc("Get a device parameter.")
                          .arg("Parameter", "The number of the parameter.");
         this->addOperation("getError",
                     &DeviceInterface::getError, this, ClientThread)
                          .doc("Get device error status.");

         // Add own thread operations :
         this->addOperation("updateData",
                          &DeviceInterface::updateData, this, OwnThread)
                              .doc(&DeviceInterface::updated)
                              .arg("Command data acquisition." );
      }
  };

  #include "TaskDeviceInterface.hpp"

  class MyDevice_1 
    : public TaskDeviceInterface
  {
  public:

    bool setParameter(int parnr, double value) {
      // ...
    }
    double getParameter(int parnr) const {      // ...
    }
    // etc.
  };
	

	  // now add it to the supervising task :
  MyDevice_1 mydev;
  supervisor.addPeer( &mydev, "device" );
	

  program ControlDevice
  {
     const int par1 = 0
     const int par2 = 1
     device.setParameter(par1, supervisor.par1 )
     device.setParameter(par2, supervisor.par2 )

     while ( device.getError() == 0 ) 
     {
        if ( this.updateDevice("device") == true )
           device.updateData()
     }
     this.handleError("device", device.getError() )
  }

  ControlDevice.start()
	

Various sorts of comments are supported in the syntax. Here is a small listing showing the various syntaxes:

  # A perl-style comment, starting at a '#', and running until
  # the end of the line.
  
  // A C++/Java style comment, starting at '//', and running
  // until the end of the line.
  
  /* A C-style comment, starting at '/*', and running until
     the first closing */ /* Nesting is not allowed, that's
     why I have to start a new comment here :-) 
  */
	  

Whitespace is in general ignored, except for the fact that it is used to separate tokens.

Identifiers are names that the user can assign to variables, constants, aliases, labels. The same identifier can only be used once, except that for labels you can use an identifier that has already been used as a variable, constant or alias. However, this is generally a bad idea, and you shouldn't do it.

Some words cannot be used as identifiers, because they are reserved by the Orocos Scripting Framework, either for current use, or for future expansions. These are called keywords. The current list of reserved keywords is included here:

These, and all variations on the (upper- or lower-) case of each of the letters are reserved, and cannot be used as identifiers.

Expressions are a general concept used throughout the Parser system. Expressions represent values that can be calculated at runtime (like a+b). They can be used as arguments to functions, conditions and whatmore. Expressions implicitly are of a certain type, and the Parser system does strong type-checking. Expressions can be constructed in various ways, that are described below...

  // a string with some escaped letters:
  "\"OROCOS rocks, \" my mother said..."
  // a normal integer
  -123
  // a double literal
  3.14159265358979
  // and another one..
  1.23e10  	    

  // define a variable of type int, called counter,
  // and give it the initial value 0.
  var int counter = 0
  // add 1 to the counter variable
  counter = counter + 1
  
  // make the name "counterPlusOne" an alias for the
  // expression counter + 1.  After this, using
  // counterPlusOne is completely equivalent to writing
  // counter + 1
  alias int counterPlusOne = counter + 1
  // you can assign an arbitrarily complex expression
  // to an alias
  alias int reallycomplexalias = ( ( counter + 8 ) / 3 )*robot.position

  // define a constant of type double, with name "pi"
  const double pi = 3.14159265358979
  const double pi2 = 2*pi       // ok, pi2 is 6.28...
  const int turn = counter * pi // warning ! turn will be 0 !

  // define a constant at _parse-time_ !
  const totalParams = table.getNbOfParams()

2.3.3. Strings and Arrays

For convenience, two variable size types have been added to the parser : string and array. They are special because their contents have variable size. For example a string can be empty or contain 10 characters. The same holds for an array, which contains doubles. String and array are thus container types. They are mapped on std::string and std::vector<double>. To access them safely from a task method or command, you need to to pass them by const reference : const std::string& s, const std::vector<double>& v.

Container types can be used in two ways : with a predefined capacity (ie the possibility to hold N items), or with a free capacity, where capacity is expanded as there is need for it. The former way is necessary for real-time programs, the latter can only be used in non real-time tasks, since it may cause a memory allocation when capacity limits are exceeded. The following table lists all available constructors:

Table 3.1. array and string constructors

Copy Syntax (copy done at run-time)	Pre-allocate syntax (init done at parse-time)	Notes
var string x = string()	var string x	Creates an empty string. (std::string)
var string x = string("Hello World")	var string x("Hello World")	Creates a string with contents "Hello World".
var array x = array()	var array x	Creates an empty array. (std::vector<double>)
var array x = array(10)	var array x(10)	Creates an array with 10 elements, all equal to 0.0.
var array x = array(10, 3.0)	var array x(10, 3.0)	Creates an array with 10 elements, all equal to 3.0.
var array x = array(1.0, 2.0, 3.0)	var array x(1.0, 2.0, 3.0)	Creates an array with 3 elements: {1.0, 2.0, 3.0}. Any number of arguments may be given.

	Warning
	The 'Copy Syntax' syntax leads to not real-time scripts because the size is expanded at run-time. See the examples below.

Example 3.1. string and array creation

  // A free string and free array :
  // applestring is expanded to contain 6 characters (non real-time)
  var string applestring = "apples"

  // values is expanded to contain 15 elements (non real-time)
  var array  values      = array(15)         

  // A fixed string and fixed array :
  var string fixstring(10)  // may contain a string of maximum 10 characters
  
  fixstring  = applestring   // ok, enough capacity
  fixstring  = "0123456789x" // allocates new memory (non real-time).


  var array  fixvalues(10)  // fixvalues pre-allocated 10 elements
  var array  morevalues(20) // arrays are initialised with n doubles of value 0.0 

  fixvalues  = morevalues  // will cause allocation in 'fixvalues'
  morevalues = fixvalues   // ok, morevalues has enough capacity, now contains 10 doubles

  fixvalues  = morevalues  // ok, since morevalues only contains 10 items.

  values     = array(20)   // expand values to contain 20 doubles. (non real-time) 

  var array list(1.0, 2.0, 3.0, 4.0) // list contains {1.0, 2.0, 3.0, 4.0}
  var array biglist;                 // creates an empty array
  biglist = list           // 'biglist' is now equal to 'list' (non real-time)
	  

Important

The 'size' value given upon construction (array(10) or string(17)) must be a legal expression at parse time and is only evaluated once. The safest method is using a literal integer ( i.e. (10) like in the examples ), but if you create a Task constant or variable which holds an integer, you can also use it as in : var array example( 5 * numberOfItems ) The expression may not contain any program variables, these will all be zero upon parse time ! The following example is a common mistake also : numberOfItems = 10 var array example( 5 * numberOfItems ) Which will not lead to '50', but to '5 times the value of numberOfItems, being still zero, when the program is parsed.

Another property of container types is that you can index (use []) their contents. The index may be any expression that return an int.

  // ... continued
  // Set an item of a container :
  for (var int i=0; i < 20; i = i+1)
       values[i] = 1.0*i

  // Get an item of a container :
  var double sum
  for (var int i=0; i < 20; i = i+1)
       sum = sum + values[i]  

If an assignment tries to set an item out of range, the assignment will fail, if you try to read an item out of range, the result will return 0.0, or for strings, the null character.

  var mydata d1;

  d1.count = 1;
  d1.name  = "sample";

  var string s1 = "abcdef"

  // retrieve size and capacity of a string :
  var int size = s1.size
  var int cap  = s1.capacity

  var array a1( 10 )
  var array a2(20) = a1

  // retrieve size and capacity of a array :
  var int size = a2.size      // 10
  var int cap  = a2.capacity  // 20

Before we go on describing the details of the programs syntax, we show how you can load a program in your Real-Time Task.

The easiest way is to use the DeploymentComponent where you can specify a script to load in the application's deployment XML file or using the TaskBrowser. You can also do it in C++, as described below.

The example below is for a program script programs.ops which contains a program with the name "progname".

  Component [R]> .provide scripting
  Trying to locate service 'scripting'...
  Service 'scripting' loaded in Component
  Component [R]> scripting.loadPrograms("programs.ops")
    = true
  Component [R]> progname.start()
    = true
      

  #include <rtt/Activity.hpp>
  #include <rtt/TaskContext.hpp>
  #include <rtt/scripting/Scripting.hpp>

  using namespace RTT;

  TaskContext   tc;
  tc.setActivity( new Activity(5, 0.01) );

  // Watch Logger output for errors :
  tc.getProvider<Scripting>("scripting")->loadPrograms("program.ops");

  // start a program :
  tc.getProvider<Scripting>("scripting")->startProgram("progname");

  // Services are always accessed using a shared_ptr
  // cast the "scripting" RTT::Service to an RTT::scripting::ScriptingService shared_ptr:
  RTT::scripting::ScriptingService::shared_ptr ss 
     = boost::dynamic_pointer_cast<ScriptingService>( this->provides()->getService("scripting") );

  ProgramInterfacePtr p = ss->getProgram("progname");

  // start a program :
  p->start();

An Orocos program script is a list of statements, quite similar to a C program. Programs can call C/C++ functions and functions can be loaded into the system, such that other programs can call them. Program scripts are executed by the Execution Engine.

In general, program statements are executed immediately one after the other. However, when the program needs to wait for a result, the Execution Engine temporarily postpones program execution and will try again in the next execution period. This happens typically when yield was called. Calling operations and expressions on the other hand typically do not impose a yield, and thus are executed immediately after each other.

Task attributes/Properties are set in the same way as ordinary script variables.

  // Setting a property named MyProp of type double
  var double d = 5.0
  comp.MyProp = d 

Statements can be grouped in functions. A function can only call another function which is earlier defined. Thus recursive function calling is not allowed in this language.

  // A function only known in the current scripting service
  void func_name( int arg1, double arg2 ) {
    // an arbitrary number of statements
  }
  
  // A function put in the interface of the component
  export double func_name(bool arg) {
    // ...
    if ( arg ) then return +10.0; else return -10.0;
  }

  // A function put in the global service interface of the current process
  global double global_func_name(bool arg) {
    // ...
    if ( arg ) then return +10.0; else return -10.0;
  }

A function can have any number of arguments, which are passed by value, and may return a value.

By default, a function is only known in the scripting service of the current component. You can make this explicit by writing the local keyword in front of the return value. This function will be found as an operation in the 'scripting' Service of the current component. You should not rely on the presence or name of this Operation, since it is considered as 'internal' to the scripting Service. Future releases may relocate or rename this function.

You can add a function to the interface of the current component by using the export keyword. This allows you to extend the interface of a component at run-time.

Finally, the global keyword puts the defined function in the GlobalService, which makes it available for any component or script in the current process.

You may redefine a function with the same name, in which case a warning will be logged and the new function is installed. In case the same function name is in use locally, at the TaskContext interface or globally, the local function is used first, then the TaskContext function and finally the global function.

A function can be called as a regular Operation :

  foo(arg)      // is a global, local or exported function of the current component
  

If one of the statements of the called function throws an exception, an exception is thrown in the current program and the calling program goes into the error state.

	Note
	The 'call' keyword has been deprecated since version 2.5 and should no longer be used.

The return statement behaves like in traditional languages. For programs and functions that do not return a value, the return statement is written like:

  export void foo(int i) {
     // ...
     if ( i < 0 )
        return
     // use i...
  }

When the return statement returns a value, it must be on the same line as the return word:

  export int foo(int i) {
     // ...
     if ( i < 0 )
        return -1  // returned value on same line.
     // use i...
     return i*10
  }

As the examples show, you can return from a function from multiple places.

A special statement 'yield' is provided. It temporarily suspends the execution of the current script and allows the Execution Engine in which it runs to do something else. You will need this in an endless while loop, for example:

  while( true ) { 
    log("Waiting...")
    yield
  }

If the yield statement is omitted, the script would never return and consume all available processor time. Yield suspends the execution of this script until the ExecutionEngine is triggered again, for example, when an asynchronous operation is received or by the expiration of the period in a periodically running component.

A StateMachine is the state machine used in the Orocos system. It contains a collection of states, and each state defines a Program on entry of the state, when it is run and on exit. It also defines all transitions to a next state. Like program scripts, a StateMachine must be loaded in a Task's Execution Engine.

A StateMachine is composed of a set of states. A running StateMachine is always in exactly one of its states. One time per period, it checks whether it can transition from that state to another state, and if so makes that transition. By default, only one transition can be made in one Execution Engine step.

Besides a list of the possible transitions, every state also keeps record of programs to be executed at certain occasions. There can be up to four (all optional) programs in every state: the entry program ( which will be executed each time the state is entered ), the run program ( which will be executed every time the state is the active state ), the handle program ( which will be executed right after run, if no transition succeeds ) and the exit program ( which will be executed when the state is left).

There can be more than one StateMachine. They separately keep track of their own current state, etc.

A StateMachine can have any number of states. It needs to have exactly one "initial state", which is the state that will be entered when the StateMachine is first activated. There is also exactly one final state, which is automatically entered when the StateMachine is stopped. This means that the transition from any state to the final state must always be meaningful.

A State Machine can run in two modes. They are the automatic mode and the reactive (also 'event' or 'request') mode. You can switch from one mode to another at run-time.

5.2.1. Reactive Mode: State Change Semantics

In order to enter the reactive mode, the State Machine must be 'activated'. When active, two possible causes of state transitions can exist: because an event occured or because a transition was requested.

Figure 3.1. State Change Semantics in Reactive Mode

A state can list to which Orocos data flow events it reacts, and under which conditions it will make a transition to another state. A state only reacts to events when its entry program is fully executed (done) and an event may be processed when the run program is executed, thus interrupt the run program. The first event that triggers a transition will 'win' and the state reacts to no more events, executes the event's transition program, then the state's exit program, and finally, the next state is entered and its entry program is executed. The next state now listens for events (if any) to make a transition or just executes its run program.

Another program can request a transition to a particular state as well. When the request arrives, the current state checks its transition conditions and evaluates if a transition to that state is allowed. These conditions are separately listed from the event transitions above. If a transition condition is valid, the exit program of the current state is called, the transition program and then the entry program of the requested state is called and the requested state's run program is executed. If a transition to the current state was requested, only the run program of the current state is executed.

In this mode, it is also possible to request a single transition to the 'best' next state. All transition conditions are evaluated and the first one that succeeds makes a transition to the target state. This mechanism is similar to automatic mode below, but only one transition is made ( or if none, handle is executed ) and then, the state machine waits again. The step() command triggers this behaviour.

5.2.2. Automatic Mode: State Change Semantics

In order to enter automatic mode, the State Machine must be first reactive and then started with the start() command (see later on).

	Note
	This mechanism is in addition to 'reactive' mode. A state machine in automatic mode still reacts to events.

Figure 3.2. State Change Semantics in Automatic Mode

The automatic mode additionally actively evaluates guard conditions. Event reaction remains in effect, but is not shown in this diagram.

In automatic mode, after the run program of the current state finishes, the transition table (to other states) of the current state is evaluated. If a transition succeeds, the transition program is executed, then the exit program of the current state is called and then the entry program of the next state is called. If no transition evaluated to true, the handle program (if any) of the current state is called. This goes on until the automatic mode is left, using the pause, stop or reactive command.

Analogous to the Program section, we first show how you can load a StateMachine in your Real-Time Task. Assume that you have a StateMachine "MachineInstanceName" in a file state-machine.osd.

  Component [R]> .provide scripting
  Trying to locate service 'scripting'...
  Service 'scripting' loaded in Component
  Component [R]> scripting.loadStateMachines("state-machine.osd")
    = true
  Component [R]> MachineInstanceName.activate()
    = true
  Component [R]> MachineInstanceName.start()
    = true
      

  #include <rtt/Activity.hpp>
  #include <rtt/TaskContext.hpp>
  #include <rtt/scripting/Scripting.hpp>

  using namespace RTT;

  TaskContext   tc;
  tc.setActivity( new Activity(5, 0.01) );

  tc.getProvider<Scripting>("scripting")->loadStateMachines( "state-machine.osd" );

  // activate a state machine :
  tc.getProvider<Scripting>("scripting")->activateStateMachine("MachineInstanceName");
  // start a state machine (automatic mode) :
  tc.getProvider<Scripting>("scripting")->startStateMachine("MachineInstanceName");

  // Services are always accessed using a shared_ptr
  // cast the "scripting" RTT::Service to an RTT::scripting::ScriptingService shared_ptr:
  RTT::scripting::ScriptingService::shared_ptr ss 
     = boost::dynamic_pointer_cast<ScriptingService>( this->provides()->getService("scripting") );

  StateMachinePtr sm = ss->getStateMachine("MachineInstanceName");

  // activate and start a StateMachine :
  sm->activate();
  sm->start();

You can think of StateMachines somewhat like C++ classes. You first need to define a type of StateMachine, and you can then instantiate it any number of times afterwards. A StateMachine ( the type ) can have parameters, so every instantiation can work differently based on the parameters it got in its instantiation.

A StateMachine definition looks like this :

  StateMachine MyStateMachineDefinition
  {
    initial state myInit
    {
        // all these programs are optional and can be left out:
        entry {
	    // entry program
	}
	run {
	    // run program
	}
	handle {
	    // handle program
	}
	exit {
	    // exit program
	}
	// Ordered event based and conditional select statements
	transition ... { /* transition program */ } select ...
	transition ...

    }

    final state myExit {
	entry {
	    // put everything in a safe state.
	}
	// leave out empty programs...

	transition ...
    }

    state Waiting {
        // ...
    }

    // ... repeat
  }  

  // See Section 5.5, “Instantiating Machines: SubMachines and RootMachines” :
  RootMachine MyStateMachineDefinition MachineInstanceName

  // In state XYZ :
  // conditionally select the START state
  transition if HMI.startPushed then {
        // (optional)
        // transition program: HMI.startPushed was true
        // when finished, go to START state
  } select START

  // next transition condition, with a transition failure program:
  transition if HMI.waiting then
      select WAIT else {
        // (optional)
        // transition failure program: HMI.startPushed was false
  }

  handle {
     // only executed if no transition above could be made
     // if startPushed and waiting were false:
     // ... 
  } 

  state X {
    // make sure the robot is not moving axis 1 when entering this state
    // and a program is loaded.
    precondition  robot.movingAxis( 1 ) == false
    precondition  programLoaded == true
    // ...
  }

5.4.6. Data Flow Event Transitions

An important property of state machines is that they can react to external (asynchronous) events. Orocos components can define reactions to data arriving on InputPorts. When new data arrives on this port, we speak of an 'event'.

	Note
	A StateMachine can only react to InputPorts which have been added with 'addEventPort' to the TaskContext.

Event transitions are an extension to the transitions above and cite an InputPort name between the transition and the if statement. They are specified as:

  state X {
    var double d
    transition x_in_port(d) if ( d >1.3 ) then {
       // transition succeeds, transition program:
       // ...
    } select ONE_STATE else {
       // transition fails, failure program:
       // ...
    } select OTHER_STATE

    // other events likewise...
  }

Both the transition programs and the the select statements are optional, but at least a program or select statement must be given. The

  transition x_in_port(d) if (d >1.3) then {
	    

short notation statement is equivalent to writing (NOTE: the added .read(d) part) :

  transition if ( x_in_port.read(d) == NewData && d >1.3) then {...
	    

Important

This short notation differs however from the long form as such: if multiple transitions are waiting on the same port, but with a different guard, the short notation will give all transitions a chance to see the NewData return value. If a user would use the long form, only the first transition statement would see NewData, all the others would see OldData as return value. In this case, the short notation is prefered.

When the input port x_in_port receives NewData, our state machine reacts to it and the data is stored in d. The if ... then statement may check this variable and any other state variables and methods to evaluate the transition. If it succeeds, an optional transition program may be given and a target state selected (ONE_STATE). if the transition fails, an optional failure program may be given and an optional select (OTHER_STATE) statement may be given. The number of arguments must match the number of arguments of the used event. The event is processed as an asynchronous callback, thus in the thread or task of the StateMachine's processor.

Event transitions are enabled after the entry program and before the exit program (also in automatic mode). All events are processed in a state until the first event that leads to a valid state transition. In the mean time, the run or handle program may continue, but will be interrupted in a yield statement if an event leads to a transition. Run or handle programs without yield will be allowed to finish normally, before the transition is taken.

  transition if (robotState.read(robot_state) != NoData && Robot.STATE_SAFE == robot_state) select SAFE;
		

5.4.8. Operation Event Transitions

A second source of transitions are the invocation of Operations of the TaskContext the state machine runs in. In order to respond in a state machine to such an event, the operation needs to be added with the addEventOperation function:

  // member variable of 'MyComp':
  Operation<void(int)> requestSafe; // don't forget to initialize with a "name" in the constructor.

  // ...

     // for example: in the TaskContext constructor:
     this->provides()->addEventOperation(requestSafe).doc("This operation does nothing except for requesting the SAFE state");
		

	Note
	A StateMachine can only react to Operations which have been added with 'addEventOperation' to the TaskContext.

	Note
	A StateMachine can only react to Operations of the component it runs in.

Reacting to an Operation call is similar to responding to port events above:

  state X {
    var int reason;

    transition requestSafe(reason) if ( reason == 3 ) then {
       // transition succeeds, transition program:
       // ...
    } select SAFE else {
       // transition fails, failure program:
       // ...
    } select OTHER_STATE

    // other events likewise...
  }

As with Event ports, both the transition programs and the the select statements are optional, but at least a program or select statement must be given.

As mentioned before: you can look at a SubMachine definition as the definition of a C++ class. It is merely the template for its instantiations, and you have to instantiate it to actually be able to do anything with it. There is also a mechanism for passing parameter values to the StateMachines on instantiation.

Note that you always need to write the instantiation after the definition of the StateMachine you're instantiating.

  StateMachine SomeStateMachine
  {
      initial state initState
      {
        // ...
      }
      final state finalState
      {
        // ...
      }
  }

  RootMachine SomeStateMachine someSMinstance 

  StateMachine SomeStateMachine
  {
    // a public constant
    const double pi = 3.1415926535897
    var int counter = 0

    initial state initState
    {
	handle
	{
            // change the value of counter...
            counter = counter + 1
        }
        // ...
    }
    final state finalState
    {
	entry
        {
            someTask.doSomethingWithThisCounter( counter )
        }
        // ...
    }
  }

  Rootmachine SomeStateMachine mymachine 

  var int readcounter = 0
  readcounter = taskname.mymachine.counter

  taskname.mymachine.counter = taskname.mymachine.counter *  2

  StateMachine AxisController
  {
      // a parameter specifying which axis this Controller controls
      param int axisNumber
      initial state init
      {
          entry
          {
              var double power = someTask.getPowerForAxis( axisNumber )
  	    // do something with it...
          }
      }
  }
  
  RootMachine AxisController axiscontroller1( axisNumber = 1 )
  RootMachine AxisController axiscontroller2( axisNumber = 2 )
  RootMachine AxisController axiscontroller3( axisNumber = 3 )
  RootMachine AxisController axiscontroller4( axisNumber = 4 )
  RootMachine AxisController axiscontroller5( axisNumber = 5 )
  RootMachine AxisController axiscontroller6( axisNumber = 6 )

  StateMachine ChildStateMachine
  {
      initial state initState
      {
          // ...
      }
      final state finalState
      {
          // ...
      }
  }
  
  StateMachine ParentStateMachine
  {
      SubMachine ChildStateMachine child1
      SubMachine ChildStateMachine child2
      initial state initState
      {
  	entry
          {
	      // enter initial state :
  	    child1.activate()
  	    child2.activate()
          }
          exit
          {
	      // enter final state :
              child2.stop()
          }
      }
  
      final state finalState
      {
          entry
          {
	      // enter final state :
              child1.stop()
          }
      }
  } 

  StateMachine ChildStateMachine
  {
      param int someValue
      const double pi = 3.1415926535897
      initial state initState
      {
          // ...
      }
      final state finalState
      {
          // ...
      }
  }

  StateMachine ParentStateMachine
  {
      SubMachine ChildStateMachine child1( someValue = 0 )
      SubMachine ChildStateMachine child2( someValue = 0 )
  
      var int counter = 0
      initial state initState
      {
  	  entry
          {
  	      child1.start()
  	      child2.start()
              // set the subcontext's parameter
              child1.someValue = 2
          }
          run
          {
              counter = counter + 1
              // set the subcontext's parameters
              child2.someValue = counter
              // use the subcontext's public variables
              someTask.doSomethingCool( child1.someValue )
          }
          exit
          {
              child2.stop()
          }
      }
  
      final state finalState
      {
          entry
          {
              child1.stop()
          }
      }
  }

someSubMachine.inState( "someStateName" )
	    

Once a state machine is parsed and loaded into the State Machine Processor, it can be manipulated from another script. This can be done through the "states" subtask of the TaskContext in which the state machine was loaded. Assume that you loaded "machine" with subcontexts "axisx" and "axisy" in task "ATask", you can write

  ATask.machine.activate()
  ATask.machine.axisx.activate()
  // now in reactive mode...

  ATask.machine.axisx.start()
  ATask.machine.start()
  // now in automatic mode...

  ATask.machine.stop()
  // again in reactive mode, in final state

  ATask.machine.reset()
  ATask.machine.deactivate()
  // deactivated.
  // etc. 

The first line activates a root StateMachine, thus it enters the initial state and is put in reactive mode , the next line actives its child, the next starts its child, then we start the parent, which bring both in automatic mode. Then the parent is stopped again, reset back to its initial state and finally deactivated.

Thus both RootMachines and SubMachines can be controlled. Some basic properties of the states can be inspected likewise :

  var bool res =  ATask.machine.isActive()      // Active ?
  res = ATask.machine.axisy.isRunning()     // Running ?
  res =  ATask.machine.isReactive()          // Waiting for requests or events? 
  var string current = ATask.machine.getState() // Get current state
  res = ATask.machine.inState( current )    // inState ?

which makes it possible to monitor state machines from other scripts or an operator console.

  StateMachine X {
     // ...
     initial state y {
        entry {
	   // ...
        }
       // guard this transition.
       transition if checkSomeCondition() then
              select z
       transition if checkOtherCondition() then
              select exit
     }
     state z {
       // ...
       // always good to go to state :
       transition select ok_1
	  select ok_1
     }
     state ok_1 {
       // ...
     }
     final state exit {
       // ...
     }
  }

  RootMachine X x

   program interact {
       // First activate x :
       x.activate()  // activate and wait.

       // Request a state transition :
       try x.requestState("z") catch {
           // failed !
       }

       // ok we are in "z" now, try to make a valid transition :
       x.step()

       // enter pause mode :
       x.pause()
       // Different ! Executes a single program statement :
       x.step()

       // unpause, by re-entering reactive Mode :
       x.reactive()

       // we are in ok_1 now, again waiting...
       x.stop()      // go to the final state

       // we are in "exit" now
       reset()

       // back in state "y", handle current state :
       this.x.requestState( this.x.getState() )
       // etc.
    } 

  StateMachine X {
     // ...
     initial state y {
        entry {
	   // ...
        }
        // guard this transition.
        transition if checkSomeCondition() then
              select z
        transition if checkOtherCondition() then
              select exit
     }
     state z {
       // ...
       // always good to go to state :
       transition select ok_1
     }
     state ok_1 {
       // ...
     }
     final state exit {
       // ...
     }
  }

  RootMachine X x

   program interact {
       // First activate x :
       x.activate()  // activate and wait.

       // Enter automatic mode :
       x.start()

       // pause program execution :
       x.pause()

       // execute a single statement :
       x.step()

       // resume automatic mode again :
       x.start()

       // stop, enter final state, in request mode again.
       x.stop()

       // etc...
    } 

The Corba transport aims to make the whole Orocos Component interface available over the network. Consult the Component Builder's Manual for an overview of a Component's interface.

These Component interfaces are available:

The following limitations apply:

Since the multicast service of the CORBA Naming_Server behaves very unpredictable (see this link), you shouldn't use it. Instead, it is better started via some extra lines in /etc/rc.local:

  ################################################################################
  #  Start CORBA Naming Service
  echo Starting CORBA Naming Service
  pidof Naming_Service || Naming_Service -m 0 -ORBListenEndpoints iiop://192.168.246.151:2809 -ORBDaemon
  ################################################################################ 

Where 192.168.246.151 should of course be replaced by your ip adres (using a hostname may yield trouble due to the new 127.0.1.1 entries in /etc/hosts, we think).

All clients (i.e. both your application and the ktaskbrowser) wishing to connect to the Naming_Service should use the environment variable NameServiceIOR

  [user@host ~]$ echo $NameServiceIOR
  corbaloc:iiop:192.168.246.151:2809/NameService 

You can set it f.i. in your .bashrc file or on the command line via

  export NameServiceIOR=corbaloc:iiop:192.168.246.151:2809/NameService

See the orocos website for more information on compiling/running the ktaskbrowser.

As of this writing, MQueues only transport data flow as streams.

You must enable the ENABLE_MQUEUE flag in CMake. This will, depending on your target, try to detect your mqueue.h header file and library. MQueue also requires the boost::serialization library.

Only Gnu/Linux and Xenomai installations which provide this header can be used.

The transport must get to know your data type. There are two options. If your data type is only Plain Old Data (POD), meaning, it does not contain any pointers or dynamically sized objects, the transport can byte-copy your data. If your data type is more complex, it must use the boost::serialization library to transport your type and your type must be known to this framework.

See below on how to do this.

Simple data types without pointers or dynamically sized objects, can be transported quite easily. They are added as such:

  // myapp.cpp
  #include <rtt/types/TemplateTypeInfo.hpp>
  #include <rtt/transports/mqueue/MQTemplateProtocol.hpp>

  using namespace RTT;
  using namespace RTT::mqueue;
  using namespace RTT::types;

  struct MyData {
    double x,y,x;
    int stamp;
  };

  int ORO_main(int argc, char** argv)
  {

     // Add your type to the Orocos type system (see: Writing plugins)
     Types()->addType( new TemplateTypeInfo<MyData, false>("MyData") );

     // New: Install the template protocol for your data type.
     Types()->getType("MyData")->addTransport(ORO_MQUEUE_PROTOCOL_ID, new MQTemplateProtocol<MyData>() );

     // rest of your program can now transport MyData between processes.

  } 

As the code shows, only one line of code is necessary to register simple types to this transport.

In practice, you'll want to write a plugin which contains this code such that your data type is loaded in every Orocos application that you start.

Data types like std::vector or similar can't just be byte-copied. They need special treatment for reading and writing their contents. Orocos uses the boost::serialization library for this. This library already understands the standard containers (vector,list,...) and is easily extendable to learn your types. Adding complex data goes as such:

  // myapp.cpp
  #include <rtt/types/TemplateTypeInfo.hpp>
  #include <rtt/transports/mqueue/MQSerializationProtocol.hpp>

  using namespace RTT;
  using namespace RTT::mqueue;
  using namespace RTT::types;

  struct MyComplexData {
    double x,y,x;
    std::vector<int> stamps;
    MyComplexData() { stamps.resize(10, -1); }
  };

  // New: define the marshalling using boost::serialization syntax:
  namespace boost {
  namespace serialization {

  template<class Archive>
  void serialize(Archive & ar, MyComplexData & d, const unsigned int version)
  {
    ar & d.x;
    ar & d.y;
    ar & d.z;
    ar & d.samps; // boost knows std::vector !
  }
  }
  }

  int ORO_main(int argc, char** argv)
  {

     // Add your type to the Orocos type system (see: Writing plugins). Same as simple case.
     Types()->addType( new TemplateTypeInfo<MyComplexData, false>("MyComplexData") );

     // New: Install the Serialization template protocol for your data type.
     Types()->getType("MyComplexData")->addTransport(ORO_MQUEUE_PROTOCOL_ID, new MQSerializationProtocol<MyComplexData>() );

     // rest of your program can now transport MyComplexData between processes.

  } 

When comparing this to the previous section, only two things changed: We defined a serialize() function, and used the MQSerializationProtocol instead of the MQTemplateProtocol to register our data transport. You can find a tutorial on writing your own serialization function on: The Boost Serialization Website.

If you don't want to use CORBA for setting up a connection, you need to use the createStream function to setup a data flow stream in each process. This requires you to choose a name of the connection and use this name in both processes:

// process1.cpp:
 
  // Your port is probably created in a component:
  OutputPort<MyData> p_out("name");

  // Create a ConnPolicy object:
  ConnPolicy policy = buffer(10); // buffered connection with 10 elements.
  policy.transport = ORO_MQUEUE_PROTOCOL_ID; // the MQueue protocol id
  policy.name_id   = "mydata_conn";          // the connection id

  p_out.createStream( policy );
  // done in proces1.cpp

// process2.cpp:

  // Your port is probably created in a component:
  InputPort<MyData> p_in("indata");

  // Create a ConnPolicy object:
  ConnPolicy policy = ConnPolicy::buffer(10); // buffered connection with 10 elements.
  policy.transport = ORO_MQUEUE_PROTOCOL_ID; // the MQueue protocol id
  policy.name_id   = "mydata_conn";          // the connection id

  p_in.createStream( policy );
  // done in proces2.cpp . We can now transmit data from process1 to
  // process2 .

Both ends must specify the same connection policy. Also, the RTT assumes that the createStream is first done on the output side, and then on the input side. This is because it is an error to connect an input side without an output side producing data. When an output side opens a connection, it will send in a test data sample, which will notify the input side that someone is sending, and that the connection is probably correctly set up.

If either output or input would disapear after the connection has been setup (because their process crashed or did not clean up), the other side will not notice this. You can re-start your component, and the ports will find each other again.

If you want proper connection management, you need to use the CORBA approach below, which keeps track of appearing and disappearing connections.

The CORBA transport supports 'Out-Of-Band' (OOB) connections for data flow. This means that CORBA itself is used to setup the connection between both ports, but the actual data transfer is done using OOB protocol. In our case, CORBA will be used to setup or destroy MQueue streams.

This has several advantages:

So it's more robust, but it requires the CORBA transport.

An Out-Of-Band connection is always setup like this:

  TaskContext *task_a, *task_b;
  // init task_a, task_b...

  ConnPolicy policy = ConnPolicy::buffer(10);

  // override default transport policy to trigger out-of-band:
  policy.transport = ORO_MQUEUE_PROTOCOL_ID;

  // this is the standard way for connecting ports:
  task_a->ports()->getPort("name")->connectTo( task_b->ports()->getPort("outdata"), policy );
      

The important part here is that a policy.transport is set, while using the connectTo function of PortInterface. Normally, setting the transport is not necessary, because the RTT will figure out itself what the best means of transport is. For example, if both ports are in the same process, a direct connection is made, if one or both components are proxies, the transport will use the transport of the proxies, in our case CORBA. However, the transport flag overrides this, and the connection logic will pick this up and use the specified transport.

Overriding the transport parameter even works when you want to test over-CORBA or over-MQueue transport with using two process-local ports. The only thing to do is to set the transport parameter to the protocol ID.

Finally, if you want to use the CORBA IDL interface to connect two ports over the mqueue transport, the workflow is fairly identical. The code below is for C++, but the equivalent can be done in any CORBA enabled language:

  #include <rtt/transports/corba/CorbaConnPolicy.hpp>
  // ...
  using namespace RTT::corba;

  CControlTask_var task_a, task_b;
  // init task_a, task_b...

  CConnPolicy cpolicy = toCORBA( RTT::ConnPolicy::buffer(10) );

  // override default transport policy to trigger out-of-band:
  cpolicy.transport = ORO_MQUEUE_PROTOCOL_ID;

  // this is the standard way for connecting ports in CORBA:
  CDataFlowInterface_var dataflow_a = task_a->ports();
  CDataFlowInterface_var dataflow_b = task_b->ports();

  dataflow_a->createConnection("name", dataflow_b, "outdata", cpolicy );
      

Similar as connectTo above, the createConnection function creates a fully managed connection between two data flow ports. We used the toCORBA function from CorbaConnPolicy.hpp to convert RTT policy objects to CORBA policy objects. Both RTT::ConnPolicy and RTT::corba::CConnPolicy structs are exactly the same, but RTT functions require the former and CORBA functions the latter.

Alternatively, you can use the create streams functions directly from the CORBA interface, in order to create unmanaged streams. In that case, the code becomes:

  #include <rtt/transports/corba/CorbaConnPolicy.hpp>
  // ...
  using namespace RTT::corba;

  CControlTask_var task_a, task_b;
  // init task_a, task_b...

  CConnPolicy cpolicy = toCORBA( RTT::ConnPolicy::buffer(10) );

  // override default transport policy and provide a name:
  cpolicy.transport = ORO_MQUEUE_PROTOCOL_ID;
  cpolicy.name_id = "stream_name";

  // this is the standard way for connecting ports in CORBA:
  CDataFlowInterface_var dataflow_a = task_a->ports();
  CDataFlowInterface_var dataflow_b = task_b->ports();

  dataflow_b->createStream("outdata", cpolicy );
  dataflow_a->createStream("name", cpolicy );
      

Note that creating message queues like this leaves out all management code and will not detect broken connections. It has the same constraints as if the streams were setup in C++, as shown in the previous section.

	Note
	When you use a TaskContext, the ExecutionEngine is the function to be executed periodically and you don't need to write the classes below.

There are two ways to run a function in a periodically. By :

The Activity will detect if it must run an external RunnableInterface. If none was given, it will call its own virtual methods.

If you want to create an activity which reads file-IO, or displays information or does any other possibly blocking operation, the Activity implementation can be used with a period of zero (0). When it is start()'ed, its loop() method will be called exactly once and then it will wait, after which it can be start()'ed again. Analogous to a periodic Activity, the user can implement initialize(), loop() and finalize() functions in a RunnableInterface which will be used by the activity for executing the user's functions. Alternatively, you can reimplement said functions in a derived class of Activity.

  int priority = 5;
  
  RunnableInterface* blocking_activity = ...
  Activity activity( priority, blocking_activity );
  activity.start(); // calls blocking_activity->initialize()

  // now blocking_activity->loop() is called in a thread with priority 5.  
  // assume loop() finished...

  activity.start();  // executes again blocking_activity->loop()

  // calls blocking_activity->breakLoop() if loop() is still executing,
  // when loop() returned, calls blocking_activity->finalize() :
  activity.stop(); 

The Activity behaves differently when being non periodic in the way start() and stop() work. Only the first invocation of start() will invoke initialize() and then loop() once. Any subsequent call to start() will cause loop() to be executed again (if it finished in the first place).

Since the user's loop() is allowed to block the user must reimplement the RunnableInterface::breakLoop() function. This function must do whatever necessary to let the user's loop() function return (mostly set a flag). It must return true on success, false if it was unable to let the loop() function return (the latter is the default implementation's return value). stop() then waits until loop() returns or aborts if breakLoop() returns false. When successful, stop() executes the finalize() function.

There are at least two scheduler types in RTT: The real-time scheduler, ORO_SCHED_RT, and the not real-time scheduler, ORO_SCHED_OTHER. In some systems, both may map to the same scheduler.

When a Activity, it runs in the default 'ORO_SCHED_OTHER' scheduler with the lowest priority. You can specify another priority and scheduler type, by providing an extra argument during construction. When a priority is specified, the Activity selects the the ORO_SCHED_RT scheduler.

  // Equivalent to Activity my_act(OS::HighestPriority, 0.001) :
  Activity my_act(ORO_SCHED_RT, OS::HighestPriority, 0.001);

  // Run in the default scheduler (not real-time):
  Activity other_act ( 0.01 );
      

If none of the above activity schemes fit you, you can always fall back on the SlaveActivity, which lets the user control when the activity is executed. A special function bool execute() is implemented which will execute RunnableInterface::step() or RunnableInterface::loop() when called by the user. Three versions of the SlaveActivity can be constructed:

  #include <rtt/SlaveActivity.hpp>

  // With master
  // a 'master', any ActivityInterface (even SlaveActivity):
  Activity master_one(9, 0.001 );
  // a 'slave', takes over properties (period,...) of 'master_one':
  SlaveActivity slave_one( &master_one );

  slave_one.start();   // fail: master not running.
  slave_one.execute(); // fail: slave not running.

  master_one.start();  // start the master.
  slave_one.start();   // ok: master is running.
  slave_one.execute(); // ok: calls step(), repeat...
  
  // Without master
  // a 'slave' without explicit master, with period of 1KHz.
  SlaveActivity slave_two( 0.001 );
  // a 'slave' without explicit master, not periodic.
  SlaveActivity slave_three;

  slave_two.start();   // ok: start periodic without master
  slave_two.execute(); // ok, calls 'step()', repeat...
  slave_two.stop();

  slave_three.start();   // start not periodic.
  slave_three.execute(); // ok, calls 'loop()', may block !
  // if loop() blocks, execute() blocks as well.
	

Note that although there may be a master, it is still the user's responsibility to get a pointer to the slave and call execute().

There is also a trigger() function for slaves with a non periodic master. trigger() will in that case call trigger() upon the master thread, which will cause it to execute. The master thread is then still responsible to call execute() on the slave. In constrast, calling trigger() upon periodic slaves or periodic activities will always fail. Periodic activities are triggered internally by the elapse of time.

Each Orocos Activity (periodic, non periodic and event driven) type has a thread() method in its interface which gives a non-zero pointer to a ThreadInterface object which provides general thread information such as the priority and periodicity and allows to control the real-timeness of the thread which runs this activity. A non periodic activity's thread will return a period of zero.

A RunnableInterface can get the same information through the this->getActivity()->thread() method calls.

#include "rtt/ActivityInterface.hpp"

using namespace RTT;

ORO_main( int argc, char** argv)
{
  // ... create any kind of Activity like above.

  ActivityInterface* act = ...

  // stop the thread and all its activities:
  act->thread()->stop();
  // change the period:
  act->thread()->setPeriod( 0.01 );

  // ORO_SCHED_RT: real-time  ORO_SCHED_OTHER: not real-time.
  act->thread()->setScheduler(ORO_SCHED_RT);

  act->thread()->start();

  // act is running...

  return 0;
}

 #include <rtt/internal/Signal.hpp>

 using boost::bind;

 class SafetyStopRobot {
 public:
    void handle_now() {
        std::cout << " Putting the robot in a safe state fast !" << std::endl;
    }
 };

 SafetyStopRobot safety;
 

Now we will connect the handler function to a signal. Each event-handler connection is stored in a Handle object, for later reference and connection management.

 // The <..> means the callback functions must be of type "void foo(void)"
 Signal<void(void)> emergencyStop;
 // Use ready() to see if the event is initialised.
 assert( emergencyStop.ready() );
 Handle emergencyHandle;
 Handle notifyHandle;

 // boost::bind is a way to connect the method of an object instance to
 // an event.
 std::cout << "Register appropriate handlers to the Emergency Stop Signal\n";
 emergencyHandle = 
   emergencyStop.connect( bind( &SafetyStopRobot::handle_now, &safety));
 assert( emergencyHandle.connected() );

Finally, we emit the event and see how the handler functions are called:

 std::cout << "Emit/Call the event\n";
 emergencyStop();

The program will output these messages:

     Register appropriate handlers to the Emergency Stop Signal
     Emit the event
      Putting the robot in a safe state fast !
      

If you want to find out how boost::bind works, see the Boost bind manual. You must use bind if you want to call C++ class member functions to 'bind' the member function to an object :

  ClassName object;
  boost::bind( &ClassName::FunctionName, &object)	

Where ClassName::FunctionName must have the same signature as the Signal. When the Signal is called,

  object->FunctionName( args )

is executed by the Signal.

When you want to call free ( C ) functions, you do not need bind :

  Signal<void(void)> event;
  void foo() { ... }
  event.connect( &foo );

You must choose the type of Signal upon construction. This can no longer be changed once the Signal is created. If the type changes, the event() method must given other arguments. For example :

  Signal<void(void)> e_1;
  e_1();

  Signal<void(int)>  e_2;
  e_2( 3 );

  Signal<void(double,double,double)>  positionSignal;
  positionSignal( x, y, z);

Furthermore, you need to setup the connect call differently if the Signal carries one or more arguments :

  SomeClass someclass;

  Signal<void(int, float)> event;

  // notice that for each Signal argument, you need to supply _1, _2, _3, etc...
  event.connect( boost::bind( &SomeClass::foo, someclass, _1, _2 ) );

  event( 1, 2.0 );

	Important
	The return type of callbacks is ignored and can not be recovered.

Signal connections can be managed by using a Handle which both connect() and setup() return :

  Signal<void(int, float)> event;
  Handle eh;

  // store the connection in 'eh'
  eh = event.connect( ... );
  assert( eh.connected() );

  // disconnect the function(s) :
  eh.disconnect();
  assert( !eh.connected() );

  // reconnect the function(s) :
  eh.connect();
  // connected again !
	

Handle objects can be copied and will all show the same status. To have a connection setup, but not connected, one can write :

  Signal<void(int, float)> event;
  Handle eh;

  // setup : store the connection in 'eh'
  eh = event.setup( ... );
  assert( !eh.connected() );

  // now connect the function(s) :
  eh.connect();
  assert( eh.connected() );  // connected !
	

If you do not store the connection of setup(), the connection will never be established and no memory is leaked. If you do not use 'eh' to connect and destroy this object, the connection is also cleaned up. If you use 'eh' to connect and then destroy 'eh', you can never terminate the connection, except by destroying the Signal itself.

The TimeService is implemented using the Singleton design pattern. You can query it for the current (virtual) time in clock ticks or in seconds. The idea here is that it is responsible for synchronising with other (distributed) cores, for doing, for example compliant motion with two robots. This functionality is not yet implemented though.

When the SimulationThread is used and started, it will change the TimeService's clock with each period ( to simulate time progress). Also other threads (!) In the system will notice this change, but time is guaranteed to increase monotonously.

Also take a look at the interface documentation.

  #include <rtt/os/TimeService.hpp>
  #include <rtt/Time.hpp>

  TimeService::ticks timestamp = TimeService::Instance()->getTicks();
  //...

  Seconds elapsed = TimeService::Instance()->secondsSince( timestamp );	

Orocos provides configuration by properties through the Property class. They are used to store primitive data (float, strings,...) in a hierarchies (using PropertyBag). A Property can be changed by the user and has immediate effect on the behaviour of the program. Changing parameters of an algorithm is a good example where properties can be used. Each parameter has a value, a name and a description. The user can ask any PropertyBag for its contents and change the values as they see fit. Java for example presents a Property API. The Doxygen Property API should provide enough information for successfully using them in your Software Component.

Note

Reading and writing a properties value can be done in real-time. Every other transaction, like marshaling (writing to disk), demarshaling (reading from disk) or building the property is not a real-time operation. Example 6.5. Using properties // a property, representing a double of value 1.0: Property<double> myProp("Parameter A","A demo parameter", 1.0); // not real-time ! myProp = 10.9; // real-time double a = myProp.get(); // real-time

Properties are mainly used for two purposes. First, they allow an external entity to browse their contents, as they can form hierarchies using PropertyBags. Second, they can be written to screen, disk, or any kind of stream and their contents can be restored later on, for example after a system restart. The next sections give a short introduction to these two usages.

First of all, a PropertyBag is not the owner of the properties it owns, it merely keeps track of them, it defines a logical group of properties belonging together. Thus when you delete a bag, the properties in it are not deleted, when you clone() a bag, the properties are not cloned themselves. PropertyBag is thus a container of pointers to Property objects.

If you want to duplicate the contents of a PropertyBag or perform recursive operations on a bag, you can use the helper functions we created and which are defined in PropertyBag.hpp (see Doxygen documentation). These operations are however, most likely not real-time.

	Note
	When you want to put a PropertyBag into another PropertyBag, you need to make a Property<PropertyBag> and insert that property into the first bag.

Use add to add Properties to a bag and getProperty(name) to mirror a Property<T>. Mirroring allows you to change and read a property which is stored in a PropertyBag: the property object's value acts like the original. The name and description are not mirrored, only copied upon initialisation:

  PropertyBag bag;
  Property<double> w("Weight", "in kilograms", 70.5 );
  Property<int> pc("PostalCode", "", 3462 );

  struct BirthDate {
     BirthDate(int d, month m, int y) : day(d), month(m), year(y) {}
     int day;
     enum { jan, feb, mar, apr, may, jun, jul, aug, sep, oct, nov, dec} month;
     int year;
  };

  Property<BirthDate> bd("BirthDate", " in 'BirthDate' format", BirthDate(1, apr, 1977));

  bag.add( &w );
  bag.add( &pc );
  bag.add( &bd );

  // setup mirrors:	
  Property<double> weight = bag.getProperty("Weight");
  assert( weight.ready() );

  // values are mirrored:
  assert( weight.get() == w.get() );
  weight.set( 90.3 );
  assert( weight.get() == w.get() );

  Property<BirthDate> bd_bis;
  assert( ! bd_bis.ready() );
  
  bd_bis = bag.getProperty("BirthDate");
  assert( bd_bis.ready() );

  // descriptions and names are not mirrored:
  assert( bd_bis.getName() == bd.getName() );
  bd_bis.setName("Date2");
  assert( bd_bis.getName() != bd.getName() );

Marshalling is converting a property C++ object to a format suitable for transportation or storage, like XML. Demarshalling reconstructs the property again from the stored format. In Orocos, the Marshaller interface defines how properties can be marshalled. The available marshallers (property to file) in Orocos are the TinyMarshaller, XMLMarshaller, XMLRPCMarshaller, INIMarshaller and the RTT::marsh::CPFMarshaller (only if Xerces is available).

The inverse operation (file to property) is currently supported by two demarshallers: TinyDemarshaller and the RTT::marsh::CPFDemarshaller (only if Xerces is available). They implement the Demarshaller interface.

The (de-)marshallers know how to convert native C++ types, but if you want to store your own classes in a Property ( like BirthDate in the example above ), the class must be added to the Orocos type system.

In order to read/write portably (XML) files, use the PropertyMarshaller and PropertyDemarshaller classes which use the default marshaller behind the scenes.

The difference between Buffers and DataObjects is that DataObjects always contain a single value, while buffers may be empty, full or contain a number of values. Thus a DataObject always returns the last value written (and a write always succeeds), while a buffer may implement a FIFO queue to store all written values (and thus can get full).

  #include <rtt/BufferLockFree.hpp>

  // A Buffer may also contain a class, instead of the simple
  // double in this example
  // A buffer with size 10:
  BufferLockFree<double> my_Buf( 10 ); 
  if ( my_Buf.Push( 3.14 ) ) {
     // ok. not full.
  }
  double  contents; 
  if ( my_Buf.Pop( contents ) ) {
     // ok. not empty.
     // contents == 3.14
  }

  #include <rtt/DataObjectInterfaces.hpp>

  // A DataObject may also contain a class, instead of the simple
  // double in this example
  DataObjectLockFree<double> my_Do("MyData"); 
  my_Do.Set( 3.14 ); 
  double  contents; 
  my_Do.Get( contents );   // contents == 3.14
  contents  = my_Do.Get(); // equivalent  

The OS layer makes an abstraction of the operating system on which it runs. It provides C++ interfaces to only the minimal set of operating system primitives that it needs: time reading, mutexes, semaphores, condition variables and threads. The abstraction also allows Orocos users to build their software on all supported systems with only a recompilation step. The OS Abstraction layer is not directly being used by the application writer.

The abstractions cause (almost) no execution overhead, because the wrappers can be called in-line. See the OROBLD_OS_AGNOSTIC option in CMake build tool to control in-lining.

Keeping the Orocos core portable requires an extra abstraction of some operating system (OS) functionalities. For example, a thread can be created, started, paused, scheduled, etc., but each OS uses other function calls to do this. Orocos prefers C++ interfaces, which led to the ThreadInterface which allows control and provides information about a thread in Orocos.

Two thread classes are available in Orocos: Thread houses our thread implementation. The MainThread is a special case as only one such object exists and represents the thread that executes the main() function.

This drawing situates the Operating System abstraction with respect to device driver interfacing (DI) and the rest of Orocos

Figure 7.1. OS Interface overview

RTAI/LXRT is an environment that allows user programs to run with real-time determinism next to the normal programs. The advantage is that the real-time application can use normal system libraries for its functioning, like showing a graphical user interface.

An introduction to RTAI/LXRT can be found in the Porting to LXRT HOWTO, which is a must-read if you don't know what LXRT is.

The common rule when using LXRT is that any user space (GNU/Linux) library can be used and any header included as long as their non-real-time functions are not called from within a hard real-time thread. Specifically, this means that all the RTAI (and Orocos) OS functions, but not the native Linux ones, may be called from within a hard real-time thread. Fortunately these system calls can be done from a not hard real-time thread within the same program.

The OS directory is the only part of the Real-Time Toolkit that needs to be ported to other Operating Systems or processor architectures in case the target supports Standard C++. The os directory contains code common to all OSes. The oro_arch directories contain the architecture dependent headers (for example atomic counters and compare-and-swap ).

In order to start your port, look at the fosi_interface.h and fosi_internal_interface.hpp files in the os directory. These two files list the C/C++ function signatures of all to be ported functions in order to support a new Operating System. The main categories are: time reading, mutexes, semaphores and threads. The easiest way to port Orocos to another operating system, is to copy the gnulinux directory into a new directory and start modifying the functions to match those in your OS.

The following table gives a short overview of the available headers in the os directory.

All tasks in the real-time system have to be performed by some thread. The OS abstraction expects an int ORO_main(int argc, char** argv) function (which the user has written) and will call that after all system initialisation has been done. Inside ORO_main() the user may expect that the system is properly set up and can be used. The resulting orocos-rtt library will contain the real main() function which will call the ORO_main() function.

	Important
	Do not forget to include <rtt/os/main.h> in the main program file, or the linker will not find the ORO_main function.

Note

Using global objects ( or static class members ) which use the OS functions before ORO_main() is entered (because they are constructed before main() ), can come into conflict with an uninitialised system. It is therefor advised not to use static global objects which use the OS primitives. Events in the CoreLib are an example of objects which should not be constructed as global static. You can use dynamically created (i.e. created with new ) global events instead.

Orocos OS only provides a few synchronisation primitives, mainly for guarding critical sections.

  Mutex m;
  void foo() {
     int i;
     MutexLock lock(m);
     // m is locked.
     // ...
  } // when leaving foo(), m is unlocked.  

  Mutex m;
  void bar() {
     int i;
     // non critical section
     {
        MutexLock lock(m);
        // m is locked.
        // critical section
     } //  m is unlocked.
     // non critical section
     //...
  }  

  Semaphore sem(0); // initial value is zero.
  void foo() {
     // Wait on sem, decrement value (blocking ):
     sem.wait()
     // awake : another thread did signal().

     // Signal sem, increment value (non blocking):
     sem.signal();

     // try wait on sem (non blocking):
     bool result = sem.trywait();
     if (result == false ) {
         // sem.value() was zero
     } else {
         // sem.value() was non-zero and is now decremented.
     }
  }  

The Device Interface can be structured in two major parts : physical device interfaces and logical device interfaces. Physical device interfaces can be subdivided in four basic interfaces: AnalogInput, AnalogOutput, DigitalInput, DigitalOutput. Analog devices are addressed with a channel as parameter and write a ranged value, while digital devices are addressed with a bit number as parameter and a true/false value.

Logical device interfaces represent the entities humans like to work with: a drive, a sensor, an encoder, etc. They put semantics on top of the physical interfaces they use underneath. You just want to know the position of a positional encoder in radians for example. Often, the physical layer is device dependent (and thus non-portable) while the logical layer is device independent.

Figure 8.1. Device Interface Overview

An example of the interactions between the logical and the physical layer is the logical encoder with its physical counting card. An encoder is a physical device keeping track of the position of an axis of a robot or machine. The programmer wishes to use the encoder as a sensor and just asks for the current position. Thus a logical encoder might choose to implement the SensorInterface which provides a read(DataType & ) function. Upon construction of the logical sensor, we supply the real device driver as a parameter. This device driver implements for example AnalogInInterface which provides read(DataType & data, unsigned int chan) and allows to read the position of a certain encoder of that particular card.

There are several classes for representing different kinds of IO. Currently there are:

From a logical point of view, the generic SensorInterface<T> is an easy to use abstraction for reading any kind of data of type T.

You need to look in the Orocos Component Library for implementations of the Device Interface. Examples are Axis and AnalogDrive.