Package madkit.kernel

Class Activator<A extends AbstractAgent>

java.lang.Object
madkit.kernel.Activator<A>
Direct Known Subclasses:
GenericBehaviorActivator

public abstract class Activator<A extends AbstractAgent> extends Object
This class defines a tool for scheduling mechanism. An activator is configured according to a community, a group and a role. It could be used to activate a group of agents on a particular behavior (a method of the agent's class) Subclasses should override execute(List, Object...) for defining how a sequential execution of a list of agents take place. By default, this list corresponds to all the agents in a single core mode or to partial views of the entire list when the multicore mode is used. The multicore mode is set to false by default.
Since:
MaDKit 2.0
Version:
5.0
Author:
Fabien Michel, Olivier Gutknecht
See Also:
  • Constructor Details

    • Activator

      protected Activator(String community, String group, String role)
      Builds a new Activator on the given CGR location of the artificial society with multicore mode set to false. This has the same effect as Activator(community, group, role, false).
      Parameters:
      community -
      group -
      role -
      See Also:
  • Method Details

    • getPriority

      public Integer getPriority()
      The priority of this activator when conflicting with another Activator. A lesser priority means that the activator will be triggered first. When the activator's priority is not set, a default priority is defined using the order in which the activators are added to the simulation engine using Scheduler.addActivator(Activator), that is the first activator has priority 0, the second has 1, and so on. Setting the activator's priority should be done using Scheduler.setActivatorPriority(Activator, int) so that the scheduler can maintain the ordering coherence of the activators list. By default, when two activators have the same priority, the order of activation is undefined.
      Returns:
      the priority of this scheduler.
    • execute

      public void execute(Object... args)
      Parameters:
      args - arguments that could be used by the scheduler to pass information to this activator for an activation
      See Also:
    • executeBehaviorOf

      public void executeBehaviorOf(AbstractAgent agent, String behaviorName, Object... args)
      Executes a specific method on a targeted agent.
      Parameters:
      agent - the targeted agent.
      behaviorName - the name of a method belonging to the agents (even private or inherited ones)
      args - parameters to be passed for the invocation
    • adding

      protected void adding(A agent)
      This method is automatically called when an agent joins the corresponding group and role. This method is empty by default. Override this method when you want to do some initialization when an agent enters the group/role.
      Parameters:
      agent - which has been added to this group/role
    • execute

      public abstract void execute(List<A> agentsList, Object... args)
      This should define what has to be done on the agents for a simulation step. By default, this calls is automatically made using a list containing all the agents for this CGR, i.e. Overlooker.getCurrentAgentsList() is used by default. When the multicore mode is on, the list is only a portion and this method will automatically be distributed over several threads. So, one has to take care about how the activator's fields are used here to avoid a ConcurrentModificationException for instance.
      Parameters:
      agentsList -
    • multicoreExecute

      protected void multicoreExecute(Object... args)
      Executes the behavior on all the agents in a concurrent way, using several processor cores if available. This call decomposes the execution of the activator in nbOfParallelTasks() tasks so that there are independently performed by the available core of the host.

      Default implementation Beware that using this call will produce different outputs for each run unless a concurrent simulation model is used. That is to say, a model supporting concurrent phases in the simulation execution such as the IRM4S model

    • toString

      public String toString()
      returns a string containing the CGR location and the number of monitored agents.
      Returns:
      a string representation of this tool.
    • isMulticoreModeOn

      public boolean isMulticoreModeOn()
      Returns:
      true if the multicore mode is on. I.e. nbOfParallelTasks() > 1. This method could be used by the default behavior of scheduler agents as they test in which mode each activator has to be used.
    • useMulticore

      public void useMulticore(int nbOfParallelTasks)
      Sets the number of tasks which will be used on a multicore architecture. If set to a number greater than 1, the scheduler will automatically use multicoreExecute(Object...) on this activator when execute(Object...) is called. If set to 1, the agents are sequentially activated. Beware that this is the only way to do exact replication of simulations, unless you have clear specifications for your model, see multicoreExecute(Object...).
      Parameters:
      nbOfParallelTasks - the number of simultaneous tasks that this activator will use to make a step. Default is 1 upon creation, so that isMulticoreModeOn() returns false.
    • nbOfParallelTasks

      public int nbOfParallelTasks()
      Returns the number of tasks that will be created by this activator in order to benefit from multicore platforms.
      Returns:
      the number of tasks that will be created.
    • getMadkitServiceExecutor

      public static ExecutorService getMadkitServiceExecutor()
      Returns the MDK ExecutorService that executes task in parallel
      Returns:
      an ExecutorService running task in parallel
    • findMethodOn

      public static <T> Method findMethodOn(Class<T> agentClass, String methodName, Class<?>... parameterTypes) throws NoSuchMethodException
      Returns the agent's method named methodName considering a given agentClass. This also works for private and inherited methods.
      Parameters:
      agentClass - the class wherein the search has to be made
      methodName - the name of the method
      parameterTypes - the parameter types of the targeted method
      Returns:
      the agent's method named methodName
      Throws:
      NoSuchMethodException
    • findMethodOnFromArgsSample

      public static <T> Method findMethodOnFromArgsSample(Class<T> agentClass, String methodName, Object... argsSample) throws NoSuchMethodException
      Returns the agent's method named methodName considering a given agentClass and a sample of the arguments which could be passed to it. The purpose of this method is restricted to a limited number of use cases since findMethodOn(Class, String, Class...) should be preferred if the exact signature of the searched method is known. A typical use case of this method is when the only information available is the arguments which are passed, for instance when overriding the execute(List, Object...) method and the like in Activator subclasses. This also works for private and inherited methods.
      Parameters:
      agentClass - the class wherein the search has to be made
      methodName - the name of the method
      argsSample - a sample of the args which can be passed to the method
      Returns:
      the agent's method named methodName
      Throws:
      NoSuchMethodException - if a matching method cannot be found
    • getSimulationTime

      public Scheduler.SimulationTime getSimulationTime()
      Get the Scheduler.SimulationTime associated with the simulation
      Returns:
      the simulationTime associated with the simulation
    • getCommunity

      public String getCommunity()
      Gets the community to which this activator/probe is binded to.
      Returns:
      a string representing the community's name
    • getGroup

      public String getGroup()
      Gets the group to which this activator/probe is binded to.
      Returns:
      a string representing the group's name
    • getRole

      public String getRole()
      Gets the role to which this activator/probe is binded to.
      Returns:
      a string representing the role's name
    • initialize

      public void initialize()
      Called by the MaDKit kernel when the Activator or Probe is first added. Default behavior is: adding(getCurrentAgentsList());
    • adding

      protected void adding(List<A> agents)
      Called when a list of agents joins the corresponding group and role. This method is automatically called by the MaDKit kernel when agents enter a role due to the use of AbstractAgent.launchAgentBucket(String, int, String...). Override this method when you want to do some initialization on the agents that enter the group/role. Default implementation is:
      
       protected void adding(final List<A> agents) {
              for (A agent : agents) {
                      adding(agent);
              }
       }
       
      Parameters:
      agents - the list of agents which have been added to this group/role at once.
    • removing

      protected void removing(List<A> agents)
      This method is automatically called when a list of agents has leaved the corresponding group and role. This method is empty by default. Override this method when you want to do some initialization on the agents that enter the group/role. Default implementation is:
      
       protected void removing(final List<A> agents) {
              for (A agent : agents) {
                      removing(agent);
              }
       }
       
      Parameters:
      agents - the list of agents which have been removed from this group/role
    • removing

      protected void removing(A agent)
      This method is automatically called when an agent leaves the corresponding group and role. This method is empty by default. Override this method when you want to do some work when an agent leaves the group/role. Note that the role is still handled by the agent when invoked.
      Parameters:
      agent - the agent which is being removed from this group/role
    • size

      public int size()
      Returns the number of the agents handling the group/role couple
      Returns:
      the number of the agents that handle the group/role couple
    • getCurrentAgentsList

      public List<A> getCurrentAgentsList()
      Returns a snapshot at moment t of the agents handling the group/role couple
      Returns:
      a list view (a snapshot at moment t) of the agents that handle the group/role couple (in proper sequence)
      Since:
      MaDKit 3.0
    • getShuffledList

      public List<A> getShuffledList()
      Returns a ListIterator over the agents which is shuffled
      Returns:
      a ListIterator which has been previously shuffled
      Since:
      MaDKit 3.0
    • killAgents

      public void killAgents()
      Kills all the agents which are monitored.
    • allAgentsLeaveRole

      public void allAgentsLeaveRole()
      Makes all the agents leave the corresponding role at once.