action_node.h

/* Copyright (C) 2015-2018 Michele Colledanchise -  All Rights Reserved
 * Copyright (C) 2018-2020 Davide Faconti, Eurecat -  All Rights Reserved
*
*   Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"),
*   to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense,
*   and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
*   The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
*
*   THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
*   FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
*   WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/

#ifndef BEHAVIORTREECORE_ACTIONNODE_H
#define BEHAVIORTREECORE_ACTIONNODE_H

#include <atomic>
#include <thread>
#include <future>
#include <mutex>

#include "leaf_node.h"

namespace BT
{
// IMPORTANT: Actions which returned SUCCESS or FAILURE will not be ticked
// again unless resetStatus() is called first.
// Keep this in mind when writing your custom Control and Decorator nodes.

/**
 * @brief The ActionNodeBase is the base class to use to create any kind of action.
 * A particular derived class is free to override executeTick() as needed.
 *
 */
class ActionNodeBase : public LeafNode
{
public:
  ActionNodeBase(const std::string& name, const NodeConfig& config);
  ~ActionNodeBase() override = default;

  virtual NodeType type() const override final
  {
    return NodeType::ACTION;
  }
};

/**
 * @brief The SyncActionNode is an ActionNode that
 * explicitly prevents the status RUNNING and doesn't require
 * an implementation of halt().
 */
class SyncActionNode : public ActionNodeBase
{
public:
  SyncActionNode(const std::string& name, const NodeConfig& config);
  ~SyncActionNode() override = default;

  /// throws if the derived class return RUNNING.
  virtual NodeStatus executeTick() override;

  /// You don't need to override this
  virtual void halt() override final
  {}
};

/**
 * @brief The SimpleActionNode provides an easy to use SyncActionNode.
 * The user should simply provide a callback with this signature
 *
 *    BT::NodeStatus functionName(TreeNode&)
 *
 * This avoids the hassle of inheriting from a ActionNode.
 *
 * Using lambdas or std::bind it is easy to pass a pointer to a method.
 * SimpleActionNode is executed synchronously and does not support halting.
 * NodeParameters aren't supported.
 */
class SimpleActionNode : public SyncActionNode
{
public:
  typedef std::function<NodeStatus(TreeNode&)> TickFunctor;

  // You must provide the function to call when tick() is invoked
  SimpleActionNode(const std::string& name, TickFunctor tick_functor,
                   const NodeConfig& config);

  ~SimpleActionNode() override = default;

protected:
  virtual NodeStatus tick() override final;

  TickFunctor tick_functor_;
};

/**
 * @brief The ThreadedAction executes the tick in a different thread.
 *
 * IMPORTANT: this action is quite hard to implement correctly.
 * Please make sure that you know what you are doing.
 *
 * - In your overriden tick() method, you must check periodically
 *   the result of the method isHaltRequested() and stop your execution accordingly.
 *
 * - in the overriden halt() method, you can do some cleanup, but do not forget to
 *   invoke the base class method ThreadedAction::halt();
 *
 * - remember, with few exceptions, a halted ThreadedAction must return NodeStatus::IDLE.
 *
 * For a complete example, look at __AsyncActionTest__ in action_test_node.h in the folder test.
 *
 * NOTE: when the thread is completed, i.e. the tick() returns its status,
 * a TreeNode::emitWakeUpSignal() will be called.
 */

class ThreadedAction : public ActionNodeBase
{
public:
  ThreadedAction(const std::string& name, const NodeConfig& config) :
    ActionNodeBase(name, config)
  {}

  bool isHaltRequested() const
  {
    return halt_requested_.load();
  }

  // This method spawn a new thread. Do NOT remove the "final" keyword.
  virtual NodeStatus executeTick() override final;

  virtual void halt() override;

private:
  std::exception_ptr exptr_;
  std::atomic_bool halt_requested_;
  std::future<void> thread_handle_;
  std::mutex mutex_;
};

#ifdef USE_BTCPP3_OLD_NAMES
using AsyncActionNode = ThreadedAction;
#endif

/**
 * @brief The StatefulActionNode is the preferred way to implement asynchronous Actions.
 * It is actually easier to use correctly, when compared with ThreadedAction
 *
 * It is particularly useful when your code contains a request-reply pattern,
 * i.e. when the actions sends an asynchronous request, then checks periodically
 * if the reply has been received and, eventually, analyze the reply to determine
 * if the result is SUCCESS or FAILURE.
 *
 * -) an action that was in IDLE state will call onStart()
 *
 * -) A RUNNING action will call onRunning()
 *
 * -) if halted, method onHalted() is invoked
 */
class StatefulActionNode : public ActionNodeBase
{
public:
  StatefulActionNode(const std::string& name, const NodeConfig& config) :
    ActionNodeBase(name, config)
  {}

  /// Method called once, when transitioning from the state IDLE.
  /// If it returns RUNNING, this becomes an asynchronous node.
  virtual NodeStatus onStart() = 0;

  /// method invoked when the action is already in the RUNNING state.
  virtual NodeStatus onRunning() = 0;

  /// when the method halt() is called and the action is RUNNING, this method is invoked.
  /// This is a convenient place todo a cleanup, if needed.
  virtual void onHalted() = 0;

  bool isHaltRequested() const;

protected:
  // do not override this method
  NodeStatus tick() override final;
  // do not override this method
  void halt() override final;

private:
  std::atomic_bool halt_requested_;
};

/**
 * @brief The CoroActionNode class is an a good candidate for asynchronous actions
 * which need to communicate with an external service using an async request/reply interface.
 *
 * It is up to the user to decide when to suspend execution of the Action and resume
 * the parent node, invoking the method setStatusRunningAndYield().
 */
class CoroActionNode : public ActionNodeBase
{
public:
  CoroActionNode(const std::string& name, const NodeConfig& config);
  virtual ~CoroActionNode() override;

  /// Use this method to return RUNNING and temporary "pause" the Action.
  void setStatusRunningAndYield();

  // This method triggers the TickEngine. Do NOT remove the "final" keyword.
  virtual NodeStatus executeTick() override final;

  void tickImpl();

  /** You may want to override this method. But still, remember to call this
    * implementation too.
    *
    * Example:
    *
    *     void MyAction::halt()
    *     {
    *         // do your stuff here
    *         CoroActionNode::halt();
    *     }
    */
  void halt() override;

protected:
  struct Pimpl;   // The Pimpl idiom
  std::unique_ptr<Pimpl> _p;

  void destroyCoroutine();
};


}   // namespace BT

#endif