鉴于有std::future::wait_for/until()
,我不明白为什么没有std::future::try_wait()
。我目前正在编写一个生产者 - 消费者示例,我想使用std::future
作为一种方便的方式来通知消费者线程返回。我的消费者代码就像
void consume(std::future<void>& stop) {
while (!stop.try_wait()) { // alas, no such method
// try consuming an item in queue
}
}
我正在考虑使用零持续时间try_wait()
来模拟wait_for()
这真的很难看。作为一个附带问题:任何其他方便消费者线程返回的方式?
答案 0 :(得分:1)
由于std :: future :: wait_for不可用,可以指定自己的超时例程,如代码片段所示:
void even(int n,promise<bool> p)
{
this_thread::sleep_for(chrono::milliseconds(500ms)); //set milliseconds(10ms) to display result
p.set_value( n%2 == 0?true:false);
}
int main()
{
promise<bool> p;
future<bool> f =p.get_future();
int n = 100;
std::chrono::system_clock::time_point tp1 = std::chrono::system_clock::now() ;
thread t([&](){ even(n,move(p)); });
auto span = std::chrono::milliseconds(200ms);
std::future_status s;
do
{
s =f.wait_for(std::chrono::seconds(0));
// do something
}
while( std::chrono::system_clock::now() < (tp1 + span) );
if( s==future_status::ready)
std::cout << "result is " << (f.get()? "Even": "Odd") << '\n';
else
std::cout << "timeout " << '\n';
t.join();
}
答案 1 :(得分:1)
std::experimental::future
添加了.is_ready()
和.then( F )
方法。
is_ready
可能是您的try_wait
(没有超时)。
wait_for
在实践中为您提供了try_wait
的功能。
std::future
并非设计为信号机制,即使它可以作为一个机制使用。如果你想要一个信令mechansim,使用条件变量,互斥和状态来创建一个,它存储信号的状态(可能组合它们)。
struct state {
bool stop = false;
unsigned some_value = 7;
friend auto as_tie( state const& s ) {
return std::tie(s.stop, s.some_value);
}
friend bool operator==( state const& lhs, state const& rhs ) {
return as_tie(lhs)==as_tie(rhs);
}
};
template<class State, class Cmp=std::equal<State>>
struct condition_state {
// gets a copy of the current state:
State get_state() const {
auto l = lock();
return state;
}
// Returns a state that is different than in:
State next_state(State const& in) const {
auto l = lock();
cv.wait( l, [&]{ return !Cmp{}(in, state); } );
return state;
}
// runs f on the state if it changes from old.
// does this atomically in a mutex, so be careful.
template<class F>
auto consume_state( F&& f, State old ) const {
auto l = lock();
cv.wait( l, [&]{ return !Cmp{}(old, state); } );
return std::forward<F>(f)( state );
}
// runs f on the state if it changes:
template<class F>
auto consume_state( F&& f ) const {
return consume_state( std::forward<F>(f), state );
}
// calls f on the state, then notifies everyone to check if
// it has changed:
template<class F>
void change_state( F&& f ) {
{
auto l = lock();
std::forward<F>(f)( state );
}
cv.notify_all();
}
// Sets the value of state to in
void set_state( State in ) {
change_state( [&](State& state) {
state = std::move(in);
} );
}
private:
auto lock() const { return std::unique_lock<std::mutex>(m); }
mutable std::mutex m;
std::condition_variable cv;
State state;
};
举一个例子,假设我们的State
是一个准备好的任务的向量,一个布尔说'#34; abort&#34;:
struct tasks_todo {
std::deque< std::function<void()> > todo;
bool abort = false;
friend bool operator==()( tasks_todo const& lhs, tasks_todo const& rhs ) {
if (lhs.abort != rhs.abort) return false;
if (lhs.todo.size() != rhs.todo.size()) return false;
return true;
}
};
然后我们可以按如下方式编写队列:
struct task_queue {
void add_task( std::function<void()> task ) {
tasks.change_state( [&](auto& tasks) { tasks.todo.push_back(std::move(task)); } );
}
void shutdown() {
tasks.change_state( [&](auto& tasks) { tasks.abort = true; } );
}
std::function<void()> pop_task() {
return tasks.consume_state(
[&](auto& tasks)->std::function<void()> {
if (tasks.abort) return {};
if (tasks.todo.empty()) return {}; // should be impossible
auto r = tasks.front();
tasks.pop_front();
return r;
},
{} // non-aborted empty queue
);
}
private:
condition_state<task_todo> tasks;
};
或某些。