初探：c++异步编程之std::promise和std::future【异步数据获取】

c11以后标准库提供了thread，说起异步可能会第一时间想起thread，线程确实好东西，不过使用线程需要很注意数据同步问题。当然了线程也不能把线程执行结果返回回来，一般都是需要使用全局变量来做数据同步。今天要探讨的是另外一些异步方式如：std::async 、std::packaged_task ，配合以std::promise和std::future来获取执行结果，使用起来有很不错的疗效。

1.std::future获取std::asnyc结果

// 模拟耗时任务
std::string sendRequest()
{
	// 休眠模拟耗时操作3s
	std::this_thread::sleep_for(std::chrono::seconds(3));
	return "good boy";
}

int main(int argc, char* argv[])
{
	// 启动异步任务并获取future
	std::future<std::string> result_future = std::async(std::launch::async, sendRequest);

	std::cout << "doing something ...\n";

	// 获取结果，调用get会阻塞直到结果可用
	std::string result = result_future.get();
	std::cout << "get a result: " << result << std::endl;

	char ch = 0;
	while (std::cin >> ch)
	{
		if (ch == 'q')
		{
			std::cout << "pressed 'q' exit ...\n";
			break;
		}
		std::this_thread::sleep_for(std::chrono::microseconds(10));
	}

	return 0;
}

也可以先调用wait等待结果可用再去获取结果， 这里调用get相当于先调用了wait和get结果获取。

2.模拟一个异步函数接口

i.模拟一个客户端类包含异步请求接口

namespace AClient
{
	struct Position
	{
		float x;
		float y;
		float z;
	};

	struct Request
	{
		// 唯一的一个id
		uint64_t uid;
		std::string params;
	};

	struct Response
	{
		int status;
		Position pos;
	};

	using SharedRequest = std::shared_ptr<Request>;
	using SharedResponse = std::shared_ptr<Response>;

	using Promise = std::promise<SharedResponse>;
	using SharePromise = std::shared_ptr<Promise>;
	using SharedFuture = std::shared_future<SharedResponse>;

	using CallbackType = std::function<void(SharedFuture)>;

	class AsyncRequestClient
	{
	public:
		AsyncRequestClient() = default;
		virtual ~AsyncRequestClient() = default;

		static inline char* getLocalTime()
		{
			static char local[26] = { 0 };
			SYSTEMTIME wtm;
			struct tm tm;

			GetLocalTime(&wtm);
			tm.tm_year = wtm.wYear - 1900;
			tm.tm_mon = wtm.wMonth - 1;
			tm.tm_mday = wtm.wDay;
			tm.tm_hour = wtm.wHour;
			tm.tm_min = wtm.wMinute;
			tm.tm_sec = wtm.wSecond;
			tm.tm_isdst = -1;

			strftime(local, 26, "%Y-%m-%d %H:%M:%S", &tm);

			return local;
		}

		void sendRequest(SharedRequest request)
		{
			std::thread req_thread([request, this] {
				std::this_thread::sleep_for(std::chrono::seconds(3));

				// 假设这里通过远程调用解析得到了数据
				SharedResponse _response = std::make_shared<Response>();
				_response->status = 1;
				_response->pos.x = 1.234567;
				_response->pos.y = 3.141592;
				_response->pos.z = 1.0f;

				std::cout << "Time:[" << getLocalTime() << "] tid:" << std::this_thread::get_id() << " ,request id:" << request->uid << " get resonse." << std::endl;

				std::unique_lock<std::mutex> lck(mutex_);
				if (this->request_list_.count(request->uid) == 0)
				{
					std::cout << "Invalid request id, return." << std::endl;
					return;
				}

				auto req_tuple = this->request_list_[request->uid];
				auto promise = std::get<0>(req_tuple);
				auto callback = std::get<1>(req_tuple);
				auto future = std::get<2>(req_tuple);

				this->request_list_.erase(request->uid);
				lck.unlock();

				promise->set_value(_response);
				callback(future);
			});
			req_thread.detach();
		}

		template<typename CallbackT>
		SharedFuture asyncRequest(SharedRequest request, CallbackT&& cb)
		{
			std::lock_guard<std::mutex> lock(mutex_);
			std::cout << "Time:[" << getLocalTime() << "] tid:" << std::this_thread::get_id() << request->uid << " ,send async request." << std::endl;

			SharePromise promise = std::make_shared<Promise>();
			SharedFuture future(promise->get_future());

			this->request_list_[request->uid] = std::make_tuple(promise, std::forward<CallbackType>(cb), future);

			// 此处只简单的用一个线程来模拟调用服务过程（实际可能是TCP、IPC、RPC）
			sendRequest(request);

			return future;
		}
	private:
		std::map<uint64_t, std::tuple<SharePromise, CallbackType, SharedFuture>> request_list_;
		std::mutex mutex_;
	};
}

ii.调用异步接口获取结果

int main(int argc, char* argv[])
{
	auto request = std::make_shared<AClient::Request>();
	request->uid = 10086;
	request->params = "{\"timestamp\": 1678222981}";

	// 开始异步调用
	AClient::AsyncRequestClient client;
	auto future_result = client.asyncRequest(request, [](AClient::SharedFuture future) {
		auto response = future.get();
		//std::cout << "Time:[" << AClient::AsyncRequestClient::getLocalTime() << "] tid:" << 		std::this_thread::get_id() << ",resonse status:" << response->status
		//			<< ",x:" << response->pos.x << ",y:" << response->pos.y << ",z:" << response->pos.z << std::endl;
	});

	// 等待结果
	future_result.wait();
	std::cout << "Time:[" << ASYClient::AsyncRequestClient::getLocalTime() << "] tid:" <<      std::this_thread::get_id() << " ,wait a result " << "status:" << future_result.get()->status
		<< ",x: " << future_result.get()->pos.x << ",y: " << future_result.get()->pos.y << ",z: " << future_result.get()->pos.z << std::endl;
	char ch = 0;
	while (std::cin >> ch)
	{
		if (ch == 'q')
		{
			std::cout << "pressed 'q' exit ...\n";
			break;
		}
		std::this_thread::sleep_for(std::chrono::microseconds(10));
	}

	return 0;
}

可以看到我们的调用过程和回调过程是在2个不同的线程， 耗时操作同样是用3秒休眠来代替。显而易见是调用3秒后得到了结果。上面例程获取结果有2种方式：1.通过future 等待结果 2.通过回调函数获取。

作者：费码程序猿
欢迎技术交流：QQ：255895056
转载请注明出处，如有不当欢迎指正