LibSourcey
C++ Networking Evolved
LibSourcey is a collection of cross platform C++11 modules and classes that provide developers with an arsenal for rapidly developing high performance network based p2p and media streaming applications. Think of it as the power and performance of libuv
combined with the features of FFmpeg
, OpenCV
and WebRTC
, all integrated with the ease and readability of the stl
(C++ Standard Library).
- Documentation: http://sourcey.com/libsourcey
- Repository: https://github.com/sourcey/libsourcey
- Licence: LGPL-2.1+
Features
-
Event-based IO — Core modules are built on top of
libuv
(the underlying C library that powersnodejs
) and use event-based asynchronous IO throughout to maximize performance and minimize concurrency reliance for building mission critical native and server side apps. -
Cross platform — The codebase is cross platform and should compile on any system with access to a modern C++11 compiler.
-
Modular libraries — Libraries are modular for easy integration into your existing projects, so you can just "include what you need" without incurring extra incumbent bloat.
-
Well tested — Core modules are well tested with unit tests and stress tested daily in production.
-
Clean and readable code — Modern C++ design principles have been adhered to throughout for clean and readable code.
-
Networking layer — A solid network layer is absolutely essential. The networking layer contains TCP, SSL and UDP socket implementations that combine
libuv
under the hood for blazing fast networking, andopenssl
for security and encryption. -
Media streaming and encoding — The
av
library consists of thin wrappers aroundFFmpeg
andOpenCV
for media capture, encoding, recording, streaming, analysis and more. -
Web servers and clients — A HTTP stack is provided that includes servers, clients, WebSockets, media streaming, file transfers, and authentication. The HTTP parser is based on the super-fast C code used by
nginx
. -
Realtime messaging — LibSourcey aims to bridge the gap between desktop, mobile and web by providing performance oriented messaging solutions that work across all platforms.
- Socket.IO — Socket.IO C++ client that supports the latest protocol revision 4 (>= 1.0). Read more about Socket.IO.
- Symple — Sourcey's home grown realtime messaging protocol that works over the top of Socket.IO to provide rostering, presence and many other features necessary for building online games and chat applications. [More about Symple](<http://sourcey.com/symple).
-
WebRTC support — WebRTC native support allows you to build p2p desktop and server side applications that inherit LibSourcey's realtime messaging and media capabilities. Take a look at the examples for how to stream live webcam and microphone streams to the browser, and also how to record live WebRTC streams on the server side.
Examples
What better way to get acquainted with a new library then with some tasty code examples.
HTTP echo server
Lets start with the classic HTTP echo server, which looks something like this:
http::Server srv{ "127.0.0.1", 1337 };
srv.Connection += [](http::ServerConnection::Ptr conn) {
conn->Payload += [](http::ServerConnection& conn, const MutableBuffer& buffer) {
conn.send(bufferCast<const char*>(buffer), buffer.size());
conn.close();
};
};
srv.start();
Pretty neat right? Its bloody fast too, especially on Linux kernel 3.9 or newer where its optimized to use of kernel level multicore socket load balancing. Don't take our word for it though, here are some benchmarks using wrk
:
LibSourcey httpechoserver
$ wrk -d10s --timeout 2s http://localhost:1337
Running 10s test @ http://localhost:1337
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 265.76us 472.62us 12.42ms 97.51%
Req/Sec 18.84k 1.26k 21.57k 74.50%
375060 requests in 10.01s, 20.39MB read
Requests/sec: 37461.50
Transfer/sec: 2.04MB
Nodejs echo server
$ wrk -d10s --timeout 2s http://localhost:1337
Running 10s test @ http://localhost:1337
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 502.70us 715.11us 14.83ms 97.90%
Req/Sec 11.69k 1.46k 14.52k 70.50%
232667 requests in 10.01s, 21.97MB read
Requests/sec: 23236.33
Transfer/sec: 2.19MB
As you can see the httpechoserver
is almost twice as fast as the nodejs
echo server, which is a massive performance gain over one of the leading web server implementations. Check the httpechoserver
sample for the full code, including the nodejs
echo server we used for profiling.
Processes
Interacting with system processes doesn't have to be painful.
The following code will run the ping sourcey.com
and with stdio
and exit callbacks:
Process proc{ "ping", "sourcey.com" };
proc.onstdout = [](std::string line)
{
// handle process output
};
proc.onexit = [](std::int64_t status)
{
// handle process exit
};
proc.spawn();
// write some random data to the stdin pipe
proc.in() << "random data"
PacketStream
A good starting point is the PacketStream
, which lets you create a dynamic delegate chain for piping, processing and outputting arbitrary data packets. This method of layering packet processors and dynamic functionality makes it possible to develop complex data processing applications quickly and easily.
For example, this is how you would capture a live webcam stream, encode it into H.264, and broadcast it in realtime over the internet:
// Create a PacketStream to pass packets from the
// input device captures -> encoder -> socket
PacketStream stream;
// Setup the encoder options
av::EncoderOptions options;
options.oformat = av::Format("MP4", "mp4",
{ "H.264", "libx264", 640, 480, 25, 48000, 128000, "yuv420p" },
{ "AAC", "aac", 2, 44100, 64000, "fltp" });
// Create a device manager instance to enumerate system devices
av::DeviceManager devman;
av::Device device;
// Create and attach the default video capture
av::VideoCapture::Ptr video;
if (devman.getDefaultCamera(device)) {
video.open(device.id, 640, 480);
video.getEncoderFormat(options.iformat);
stream.attachSource(video, true);
}
// Create and attach the default audio capture
av::AudioCapture::Ptr audio;
if (devman.getDefaultMicrophone(device)) {
audio.open(device.id, 2, 44100);
audio.getEncoderFormat(options.iformat);
stream.attachSource(audio, true);
}
// Create and attach the multiplex encoder
av::MultiplexPacketEncoder::Ptr encoder(options);
stream.attach(encoder);
// Attach the output net::Socket instance (instantiated elsewhere)
// to broadcast encoded packets over the network
stream.attach(socket);
// Start the stream
// This method call will start the device captures and begin
// pumping encoded media into the output socket
stream.start();
There are plenty more demos and sample code to play with over on the examples page.
Installation
See the platform independent installation guides.
Contributors
A BIG thanks to everyone who has contributed to the development of LibSourcey so far:
- Kam Low (@auscaster) — Primary developer
- Yury Shulbn (@yuryshubin) — iOS build toolchain and platform fixes
- Vinci Xu (@VinciShark) — Windows documentation, testing and updates
- Michael Fig (@michael-fig) — Fixed compiler flags to build without FFmpeg
- Hyunuk Kim (@surinkim) — Fixed
std::codecvt
unicode character conversion on Windows - Damian Zelim (@ZelimDamian) — Fixed compiler flags for OS X build
- Norm Ovenseri (@normano) — Added verbose logging output to build system
- Alexey (@deilos) — Fixed cross-platform FFmpeg build script
Contributing
Pull Requests are always welcome, so if you make any improvements please feel free to float them back upstream :)
- Fork LibSourcey on Github
- Create your feature branch (
git checkout -b my-new-feature
) - Commit your changes (
git commit -am 'Add some feature'
) - Push to the branch (
git push origin my-new-feature
) - Create new Pull Request