中文 | English
psyche is a non-blocking network library for Linux developed with C++17 and provides several components.
To use psyche you need to #include <psyche/psyche.h> or the header file for the specified component.
psyche supports server and client (todo) based on TCP.
This is a simple echo server.
#include "psyche/psyche.h"
using namespace psyche;
int main() {
Server s(9981);
s.set_read_callback([](Connection con, Buffer buffer)
{
con.send(buffer.retrieve_all());
});
s.start();
}And an echo server which considers the complete lifetime of the connection.
#include "psyche/psyche.h"
using namespace psyche;
int main() {
Server s(9981);
s.set_new_conn_callback([](Connection con)
{
LOG_INFO << "New connect from " << con.peer_endpoint().to_string() << ".";
});
s.set_read_callback([](Connection con, Buffer buffer)
{
const auto msg(buffer.retrieve_all());
LOG_INFO << "Received from (" << con.peer_endpoint().to_string()
<< "):" << msg;
con.send(msg);
con.close();
});
s.set_close_callback([](Connection con)
{
LOG_INFO << con.peer_endpoint().to_string() << " connection closed";
});
s.start();
}Server is a class which owns and maintains connections. Users can set different default callback functions to different situations using the functions like setReadCallback. Then invoke s.start() and the server will start. Server uses a thread pool to invoke read and send callback functions, but only one read callback functions will be invoked at one time for each Connection.
All callback functions will provide Connection as an argument. When the connection is closed, Server will not own Connection anymore. Users can own the Connection as well. When the numbers of Connections instances is zero, the resources will be released (like shared_ptr). When the close callback function is invoked, users should not own the Connection instance anymore.
Read callback provides Buffer as an argument which stores the data received. User can invoke retrieve(num) to retrieve specified length data or retrieveAll() to retrieve all data.
#include <psyche/log.h> to use log library only.
Log library provides std::cout like usage.
LOG_INFO << "New connect from " << con.peer_endpoint().to_string() << ".";
LOG_DEBUG << "Example" << msg;It provides LOG_TRACE, LOG_DEBUG, LOG_INFO, LOG_WARN, LOG_ERROR and LOG_FATAL six levels from low to high.
#include <psyche/thread_pool.h> to use thread pool only.
ThreadPool tp; //Create a thread pool with default count of threads
ThreadPool tp(16); //Create a thread pool with 16 threads
tp.execute([](){ LOG_INFO << "Thread started." ;}); //Invoke lambda function
auto result = tp.execute(Func, arg1, arg2); //Invoke functions with arguments
result.get(); //Get the result of one taskThread pool constructed with default constructor will create threads which quantity related to the hardware (std::thread::hardware_concurrency()). Otherwise, you can specify the number of threads as well.
Invoke Execute() to add tasks. You can pass lambda expression and arguments as well. You can get the return value and call get() to get the return value of your task.
When ThreadPool is destroyed, the tasks in the task queue will not be executed. But the destructor will wait for all the running tasks to finish. You can call stop() to wait for all the tasks to finish explicitly.
You can find more samples in the example folder.
Projects based on psyche:
- psyche-http ：HTTP Server based on psyche
- LixTalk ：Server-side of an IM software
Compile passed with gcc-6 and clang-6 and higher edition in Ubuntu 16.04.
cmake .
make
sudo make install
Or if you want to compile all or specified examples, use the commands below
cmake . -DPSYCHE_BUILD_ALL_EXAMPLES=ON
cmake . -DPSYCHE_BUILD_EXAMPLE="echo(example name)"
The output files will be in the build folder. Header files and lib will be installed into /usr/local.
MIT