//
//  main.cpp
//  thread_pool v0.2a
//
//  Created by Kristof Toth on 26/06/15.
//
//  Copyright © 2015 Kristof Toth <mrtoth@strongds.hu>
//  This program is free software. It comes without any warranty, to
//  the extent permitted by applicable law. You can redistribute it
//  and/or modify it under the terms of the Do What The Fuck You Want
//  To Public License, Version 2, as published by Sam Hocevar. See
//  http://www.wtfpl.net/ for more details.
//
//  Generalized thread pool class.
//  ( READ BEFORE USING IN PRODUCTION CODE )
//  User's notes:
//                  - the recommended way to add a task to the queue is to use the
//                    add_task(F&& f, Args&&... args) format, where f is any kind of
//                    functor and (args...) are the parameters to the functor (any number/kind).
//                  - copying & moving not allowed   (not sure if they'd make sense)
//                  - if a task is added to the pool via add_task(std::packaged_task<R()>& ),
//                    the client must grant, that the packaged_task object is NOT
//                    destructed before the task is finished. (otherwise it's undefined behaviour)
//                    this way of adding tasks is only recommended for micro-operations,
//                    where the overhead of add_task(F&& f, Args&&... args)'s use of operator new
//                    is too much (this is rarely the case).
//

/* necessary includes */
#include <iostream>
#include <thread>
#include <future>
#include <mutex>
#include <atomic>
#include <vector>
#include <deque>
#include <condition_variable>
#include <functional>
#include <memory>

/* for the tests */
#include <boost/multiprecision/cpp_int.hpp>
#include <boost/type_index.hpp>
//#include <boost/asio/io_service.hpp>
//#include <boost/bind.hpp>
//#include <boost/thread/thread.hpp>



class thread_pool
{
    /* under the hood */
    std::mutex mu;
    std::condition_variable cond;
    std::vector<std::thread> workers;
    std::deque<std::function<void()> > queue;
    std::atomic<bool> fin;

    /* the main loop of the threads */
    void loop();

public:

    /* construction & destruction */
    explicit thread_pool(size_t);
    ~thread_pool();

    /* disallowed operations */
    thread_pool(const thread_pool&) = delete;
    thread_pool& operator=(const thread_pool&) = delete;
    thread_pool(thread_pool&&) = delete;
    thread_pool& operator=(thread_pool&&) = delete;

    template <typename R> void add_task(std::packaged_task<R()>&& ) = delete;
    template <typename R> void priority_task(std::packaged_task<R()>&& ) = delete;

    /* adding tasks to the queue */
    // TODO: use template-based policies to deal with priority tasks to avoid code-dupe.
    template <typename F, typename... Args>
    auto add_task(F&& f, Args&&... args) -> std::future<decltype(f(args...))>;

    template <typename F, typename... Args>
    auto priority_task(F&& f, Args&&... args) -> std::future<decltype(f(args...))>;

    template <typename R> void add_task(std::packaged_task<R()>& );
    template <typename R> void priority_task(std::packaged_task<R()>& );
//	template <typename R> void add_task(std::function<R()> );      << think about this
    void add_task(std::function<void()> );

    /* other operations */
    void add_thread(size_t);

    /* means of getting information */
    inline size_t get_thread_num() const { return workers.size(); }
    inline size_t get_queue_size() const { return queue.size(); }

};

thread_pool::thread_pool(size_t thcount):
        fin(false)
{
    for (size_t i = 0; i < thcount ; ++i)
        workers.emplace_back(&thread_pool::loop, this);
}

thread_pool::~thread_pool()
{
    fin = true;
    cond.notify_all();

    for (auto& i : workers)
        i.join();
}

template <typename F, typename... Args>
auto thread_pool::add_task(F&& f, Args&&... args)
-> std::future<decltype(f(args...))>
{
    auto pckgd_tsk = std::make_shared<std::packaged_task<decltype(f(args...))()> >
            (std::bind(std::forward<F>(f), std::forward<Args>(args)...));

    {
        std::lock_guard<std::mutex> lock(mu);
        queue.emplace_back([pckgd_tsk](){ (*pckgd_tsk)(); });
    }
    cond.notify_one();

    return pckgd_tsk->get_future();
}

template <typename F, typename... Args>
auto thread_pool::priority_task(F&& f, Args&&... args)
-> std::future<decltype(f(args...))>
{
    auto pckgd_tsk = std::make_shared<std::packaged_task<decltype(f(args...))()> >
            (std::bind(std::forward<F>(f), std::forward<Args>(args)...));

    auto ret_val= pckgd_tsk->get_future();

    {
        std::lock_guard<std::mutex> lock(mu);
        queue.emplace_front([pckgd_tsk](){ (*pckgd_tsk)(); });
    }
    cond.notify_one();

    return ret_val;
}

template <typename R>
void thread_pool::add_task(std::packaged_task<R()>& arg)
{
    {
        std::lock_guard<std::mutex> lock(mu);
        queue.emplace_back([&arg](){ arg(); });
    }
    cond.notify_one();
}

template <typename R>
void thread_pool::priority_task(std::packaged_task<R()>& arg)
{
    {
        std::lock_guard<std::mutex> lock(mu);
        queue.emplace_front([&arg](){ arg(); });
    }
    cond.notify_one();
}

void thread_pool::add_task(std::function<void()> func)
{
    {
        std::lock_guard<std::mutex> lock(mu);
        queue.push_back(std::move(func));
    }
    cond.notify_one();
}

void thread_pool::add_thread(size_t num = 1)
{
    for (size_t i = 0; i < num; ++i)
    {
        workers.emplace_back(&thread_pool::loop, this);
    }
}

void thread_pool::loop()
{
    std::function<void()> fun;
    while (true)
    {
        {
            std::unique_lock<std::mutex> lock(mu);

            while (!fin && queue.empty())
                cond.wait(lock);

            if (fin)
                return;

            fun = std::move(queue.front());
            queue.pop_front();
        }
        fun();
    }
}

///////////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////                 functions for testing & test zone                 ////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////////

boost::multiprecision::cpp_int mp_factorial(unsigned long long int number)
{
    boost::multiprecision::cpp_int num = number;
    for (unsigned long long int i = number; i > 1; --i)
    {
        num *=(--number);
    }

    return num;
}

void simpleTask()
{
    std::this_thread::sleep_for(std::chrono::milliseconds(1));
}

void function_vs_packaged_task_test(thread_pool &tp, const unsigned short TIMES = 10000)
{
    namespace ch = std::chrono;
    namespace mp = boost::multiprecision;


    auto fstart = ch::high_resolution_clock::now();
    for (int j = 0; j < TIMES; ++j)
    {
        tp.add_task(simpleTask);
    }
    auto fdur = ch::high_resolution_clock::now() - fstart;

    auto pt = std::packaged_task<void()>(simpleTask);
    auto pstart = ch::high_resolution_clock::now();
    for (int j = 0; j < TIMES; ++j)
    {
        tp.add_task(pt);
    }
    auto pdur = ch::high_resolution_clock::now() - pstart;

    std::cout << "Function: " << ch::duration_cast<ch::microseconds>(fdur).count() <<
    std::endl << "Packaged task: " << ch::duration_cast<ch::microseconds>(pdur).count() << std::endl;
}

void factorial_test(thread_pool& tp, unsigned long long number)
{
    std::packaged_task<boost::multiprecision::cpp_int()> pt(std::bind(mp_factorial, number-1));
    tp.add_task(pt);
    auto fu1 = tp.add_task(mp_factorial, number);
    auto fu2 = pt.get_future();

    std::cout << "Results for n & n-1: " << fu1.get() << " " << fu2.get() << std::endl;
    std::cout << "With " << tp.get_thread_num() << " threads" << std::endl;
    std::cout << "Type of future: " << boost::typeindex::type_id_with_cvr<decltype(fu1)>().pretty_name() << std::endl;
}

void boost_vs_me_test(thread_pool& tp)
{ /* TODO */ }


int main(void)
{
    thread_pool tp(4);

    factorial_test(tp, 6);
    function_vs_packaged_task_test(tp, 10000);


    return 0;
}