问题
I am trying to simulate boost::asio::write with timeout. Or you can say, I am trying to use boost::asio::async_write with a timeout.
As I see, boost::asio::write blocks until all data has been written & read on the other side. This kind of functionality certainly requires a timeout.
So, Reading through this simple answer here by Robert Hegner which demostrates how to do a boost::asio::async_read with timeout, I am trying adapt the same logic for write by doing so:
size_t write_data_with_time_out() {
long time_out_secs = 2;
boost::optional<boost::system::error_code> timer_result;
boost::asio::deadline_timer timer(the_socket->get_io_service(), boost::posix_time::seconds(time_out_secs));
timer.expires_from_now();
timer.async_wait([&timer_result] (const boost::system::error_code& error) {
timer_result.reset(error);
});
boost::optional<boost::system::error_code> write_result;
size_t bytes_sent = 0;
boost::asio::async_write(*the_socket, boost::asio::buffer(the_buffer_to_write, the_buffer_to_write.size()), [&write_result, &bytes_sent] (const boost::system::error_code& error, auto size_received) {
write_result.reset(error);
bytes_sent = size_received;
});
the_socket->get_io_service().reset();
while (the_socket->get_io_service().run_one()) {
if (write_result) {
timer.cancel();
}
else if (timer_result) {
the_socket->cancel();
}
}
if (*write_result) {
return 0;
}
return bytes_sent;
}
Problem:
The logic works great for read but does not seem to work for write case. Reason is that while (the_socket->get_io_service().run_one()) gets hung after calling the_socket->cancel() twice.
However, in the read case also the_socket->cancel() is called twice & does not hang on the 3rd loop over of the while & returns out. Hence no issues with read.
Question:
Is my understanding wrong that same timeout logic would work for boost::asio::async_write case? I think this same logic should work. There is something wrong I am doing which is what I need a suggestion on.
Additional info required if possible:
if boost::asio::read & boost::asio::write had a timeout parameter. I would not have been writing this extra. Seems like there have been a lot of requests to asio guys to introduce a timeout in their sync read & write functions. Like this one here.
Is there any scope for asio guys to address this request in near future?
I am doing a sync boost::asio::read & boost::asio::write on the same socket using a worker thread for which this works superb. All I am missing is this timeout functionality.
Environment:
My code runs on Linux & MacOSX with C++ 14 compiler. This question only concerns TCP sockets.
回答1:
I've written the following helper to await any async operation synchronously with a timeout¹:
template<typename AllowTime> void await_operation(AllowTime const& deadline_or_duration) {
using namespace boost::asio;
ioservice.reset();
{
high_resolution_timer tm(ioservice, deadline_or_duration);
tm.async_wait([this](error_code ec) { if (ec != error::operation_aborted) socket.cancel(); });
ioservice.run_one();
}
ioservice.run();
}
I've since also demoed that with a full on TCP client: Boost::Asio synchronous client with timeout
The sample includes write operations and has been completely tested.
Complete Sample:
Taking the "nicer" example from the original post (the FTP client example shows more realistic usage patterns):
Live On Coliru
#ifndef __TCPCLIENT_H__
#define __TCPCLIENT_H__
#include <boost/asio.hpp>
#include <boost/asio/high_resolution_timer.hpp>
#include <iostream>
class TCPClient {
public:
void disconnect();
void connect(const std::string& address, const std::string& port);
std::string sendMessage(const std::string& msg);
private:
using error_code = boost::system::error_code;
template<typename AllowTime> void await_operation(AllowTime const& deadline_or_duration) {
using namespace boost::asio;
ioservice.reset();
{
high_resolution_timer tm(ioservice, deadline_or_duration);
tm.async_wait([this](error_code ec) { if (ec != error::operation_aborted) socket.cancel(); });
ioservice.run_one();
}
ioservice.run();
}
struct raise {
template <typename... A> void operator()(error_code ec, A...) const {
if (ec) throw std::runtime_error(ec.message());
}
};
boost::asio::io_service ioservice { };
boost::asio::ip::tcp::socket socket { ioservice };
};
inline void TCPClient::disconnect() {
using namespace boost::asio;
if (socket.is_open()) {
try {
socket.shutdown(ip::tcp::socket::shutdown_both);
socket.close();
}
catch (const boost::system::system_error& e) {
// ignore
std::cerr << "ignored error " << e.what() << std::endl;
}
}
}
inline void TCPClient::connect(const std::string& address, const std::string& port) {
using namespace boost::asio;
async_connect(socket, ip::tcp::resolver(ioservice).resolve({address, port}), raise());
await_operation(std::chrono::seconds(6));
}
inline std::string TCPClient::sendMessage(const std::string& msg) {
using namespace boost::asio;
streambuf response;
async_read_until(socket, response, '\n', raise());
await_operation(std::chrono::system_clock::now() + std::chrono::seconds(4));
return {std::istreambuf_iterator<char>(&response), {}};
}
#endif
#include <iostream>
//#include "TCPClient.hpp"
int main(/*int argc, char* argv[]*/) {
TCPClient client;
try {
client.connect("127.0.0.1", "27015");
std::cout << "Response: " << client.sendMessage("Hello!") << std::endl;
}
catch (const boost::system::system_error& e) {
std::cerr << e.what() << std::endl;
}
catch (const std::exception& e) {
std::cerr << e.what() << std::endl;
}
}
¹ originally written for this answer https://stackoverflow.com/a/33445833/85371
来源：https://stackoverflow.com/questions/47378022/how-to-simulate-boostasiowrite-with-a-timeout