What is the proper way of doing event handling in C++?

I have an application that needs to respond to certain events in the following manner:

void someMethodWithinSomeClass() {
    while (true) {
        wait for event;
        if (event == SomeEvent) {
            doSomething();
            continue;
        }
        if (event == SomeOtherEvent) {
            doSomethingElse();
            continue;
        }
    } 
}

This would be running is some thread. In some other threads, operations would create and fire the Events.

How do I get these Events to reach the above method/class? What is the proper strategy or architecture for implementing event handling in C++?

Solution 1:

Often, event queues are implemented as command design pattern:

In object-oriented programming, the command pattern is a design pattern in which an object is used to represent and encapsulate all the information needed to call a method at a later time. This information includes the method name, the object that owns the method and values for the method parameters.

In C++, the object that own the method and values for the method parameters is a nullary functor (i.e. a functor that takes no arguments). It can be created using boost::bind() or C++11 lambdas and wrapped into boost::function.

Here is a minimalist example how to implement an event queue between multiple producer and multiple consumer threads. Usage:

void consumer_thread_function(EventQueue::Ptr event_queue)
try {
    for(;;) {
        EventQueue::Event event(event_queue->consume()); // get a new event 
        event(); // and invoke it
    }
}
catch(EventQueue::Stopped&) {
}

void some_work(int n) {
    std::cout << "thread " << boost::this_thread::get_id() << " : " << n << '\n';
    boost::this_thread::sleep(boost::get_system_time() + boost::posix_time::milliseconds(500));
}

int main()
{
    some_work(1);

    // create an event queue that can be shared between multiple produces and multiple consumers
    EventQueue::Ptr queue(new EventQueue);

    // create two worker thread and pass them a pointer to queue
    boost::thread worker_thread_1(consumer_thread_function, queue);
    boost::thread worker_thread_2(consumer_thread_function, queue);

    // tell the worker threads to do something
    queue->produce(boost::bind(some_work, 2));
    queue->produce(boost::bind(some_work, 3));
    queue->produce(boost::bind(some_work, 4));

    // tell the queue to stop
    queue->stop(true);

    // wait till the workers thread stopped
    worker_thread_2.join();
    worker_thread_1.join();

    some_work(5);
}

Outputs:

./test
thread 0xa08030 : 1
thread 0xa08d40 : 2
thread 0xa08fc0 : 3
thread 0xa08d40 : 4
thread 0xa08030 : 5

Implementation:

#include <boost/function.hpp>
#include <boost/thread/thread.hpp>
#include <boost/thread/condition.hpp>
#include <boost/thread/mutex.hpp>
#include <boost/smart_ptr/intrusive_ptr.hpp>
#include <boost/smart_ptr/detail/atomic_count.hpp>
#include <iostream>

class EventQueue
{
public:
    typedef boost::intrusive_ptr<EventQueue> Ptr;
    typedef boost::function<void()> Event; // nullary functor
    struct Stopped {};

    EventQueue()
        : state_(STATE_READY)
        , ref_count_(0)
    {}

    void produce(Event event) {
        boost::mutex::scoped_lock lock(mtx_);
        assert(STATE_READY == state_);
        q_.push_back(event);
        cnd_.notify_one();
    }

    Event consume() {
        boost::mutex::scoped_lock lock(mtx_);
        while(STATE_READY == state_ && q_.empty())
            cnd_.wait(lock);
        if(!q_.empty()) {
            Event event(q_.front());
            q_.pop_front();
            return event;
        }
        // The queue has been stopped. Notify the waiting thread blocked in
        // EventQueue::stop(true) (if any) that the queue is empty now.
        cnd_.notify_all();
        throw Stopped();
    }

    void stop(bool wait_completion) {
        boost::mutex::scoped_lock lock(mtx_);
        state_ = STATE_STOPPED;
        cnd_.notify_all();
        if(wait_completion) {
            // Wait till all events have been consumed.
            while(!q_.empty())
                cnd_.wait(lock);
        }
        else {
            // Cancel all pending events.
            q_.clear();
        }
    }

private:
    // Disable construction on the stack. Because the event queue can be shared between multiple
    // producers and multiple consumers it must not be destroyed before the last reference to it
    // is released. This is best done through using a thread-safe smart pointer with shared
    // ownership semantics. Hence EventQueue must be allocated on the heap and held through
    // smart pointer EventQueue::Ptr.
    ~EventQueue() {
        this->stop(false);
    }

    friend void intrusive_ptr_add_ref(EventQueue* p) {
        ++p->ref_count_;
    }

    friend void intrusive_ptr_release(EventQueue* p) {
        if(!--p->ref_count_)
            delete p;
    }

    enum State {
        STATE_READY,
        STATE_STOPPED,
    };

    typedef std::list<Event> Queue;
    boost::mutex mtx_;
    boost::condition_variable cnd_;
    Queue q_;
    State state_;
    boost::detail::atomic_count ref_count_;
};

Solution 2:

The C++ Standard doesn't address events at all. Usually, however, if you need events you are working within a framework that provides them (SDL, Windows, Qt, GNOME, etc.) and ways to wait for, dispatch and use them.

Aside from that, you may want to look at Boost.Signals2.