Async Actions using Coroutines(异步行为使用协程，协同程序)

Behavior.CPP提供了两个方便使用的抽象去创建一个异步的行为，例如：

那些行为可能：

花很长时间来得出结论；

可能返回RUNNING；

可能被停止；

第一类是一个AsyncActionNode，其在单独的线程当中执行方法tick()；

本教程中，引入CoroActionNode，一个不同的行为，其使用coroutines来获得类似的结果；

主要原因是Coroutines不会生成一个新的线程，更有效率；

进一步的是，不需要担心你的代码的线程安全；

在协程中，用户应该现实的调用一个yield（翻译为退出，放弃）方法当用户想让这个行为的执行被挂起的时候；

CoroActionNode封装了yield方法进一个方便的方法setStatusRunningAndYield().

下一个例子可以作为你自己实现的一个模板：

typedef std::chrono::milliseconds Milliseconds;

class MyAsyncAction: public CoroActionNode
{
  public:
    MyAsyncAction(const std::string& name):
        CoroActionNode(name, {})
    {}

  private:
    // This is the ideal skeleton/template of an async action:
    //  - A request to a remote service provider.
    //  - A loop where we check if the reply has been received.
    //  - You may call setStatusRunningAndYield() to "pause".
    //  - Code to execute after the reply.
    //  - A simple way to handle halt().
    NodeStatus tick() override
    {
        std::cout << name() <<": Started. Send Request to server." << std::endl;

        TimePoint initial_time = Now();
        TimePoint time_before_reply = initial_time + Milliseconds(100);

        int count = 0;
        bool reply_received = false;

        while( !reply_received )
        {
            if( count++ == 0)
            {
                // call this only once
                std::cout << name() <<": Waiting Reply..." << std::endl;
            }
            // pretend that we received a reply
            if( Now() >= time_before_reply )
            {
                reply_received = true;
            }

            if( !reply_received )
            {
                // set status to RUNNING and "pause/sleep"
                // If halt() is called, we will NOT resume execution
                setStatusRunningAndYield();
            }
        }

        // This part of the code is never reached if halt() is invoked,
        // only if reply_received == true;
        std::cout << name() <<": Done. 'Waiting Reply' loop repeated "
                  << count << " times" << std::endl;
        cleanup(false);
        return NodeStatus::SUCCESS;
    }

    // you might want to cleanup differently if it was halted or successful
    void cleanup(bool halted)
    {
        if( halted )
        {
            std::cout << name() <<": cleaning up after an halt()
" << std::endl;
        }
        else{
            std::cout << name() <<": cleaning up after SUCCESS
" << std::endl;
        }
    }

    void halt() override
    {
        std::cout << name() <<": Halted." << std::endl;
        cleanup(true);
        // Do not forget to call this at the end.
        CoroActionNode::halt();
    }

    Timepoint Now()
    { 
        return std::chrono::high_resolution_clock::now(); 
    };
};

从上面代码可以看出，行为是假设等待了一个请求消息，在100ms之后请求会到达；

创建一个有两个行为的序列，但是整个序列会在150ms之后被停止；

 <root >
     <BehaviorTree>
        <Timeout msec="150">
            <SequenceStar name="sequence">
                <MyAsyncAction name="action_A"/>
                <MyAsyncAction name="action_B"/>
            </SequenceStar>
        </Timeout>
     </BehaviorTree>
 </root>

int main()
{
    // Simple tree: a sequence of two asycnhronous actions,
    // but the second will be halted because of the timeout.

    BehaviorTreeFactory factory;
    factory.registerNodeType<MyAsyncAction>("MyAsyncAction");

    auto tree = factory.createTreeFromText(xml_text);

    //---------------------------------------
    // keep executin tick until it returns etiher SUCCESS or FAILURE
    while( tree.tickRoot() == NodeStatus::RUNNING)
    {
        std::this_thread::sleep_for( Milliseconds(10) );
    }
    return 0;
}

/* Expected output:

action_A: Started. Send Request to server.
action_A: Waiting Reply...
action_A: Done. 'Waiting Reply' loop repeated 11 times
action_A: cleaning up after SUCCESS

action_B: Started. Send Request to server.
action_B: Waiting Reply...
action_B: Halted.
action_B: cleaning up after an halt()

*/