Disruptor-cpp is a fully functional C++ port of the LMAX disruptor. Implements all the features available in java Disruptor v3.3.7.
- Clang 3.8 or newer
- GCC 5.0 or newer
- Microsoft Visual C++ 2015 or newer
Boost must be available on your machine. You can install it using your favorite package manager or build it on your own. If Boost has been installed into standard system locations the following commands will start the build:
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=release
makeIf Boost has been installed into a custom location you will probably need to specify BOOST_ROOT variable. Please refer to the Find boost documentation for details.
Optionally you may want to compile and run the unit tests and benchmarks. The tests compilation is activated by means of DISRUPTOR_BUILD_TESTS flag:
cmake .. -DCMAKE_BUILD_TYPE=release -DDISRUPTOR_BUILD_TESTS=trueThe simplest way to compile the library on Windows is to use the provided Visual Studio solution files:
- Disruptor-lib.sln - the solution only includes the library.
- Disruptor-all.sln - the solution includes the library, benchmarks and unit tests.
Boost must be available on your machine. The boost.props file already included into the solution but you may need to modify the headers and libraries directory according to your boost location and folder structure.
To give you a taste of C++ disruptor let us consider a very basic example where the event is passed from producer to consumer. The event will carry a single long value:
struct LongEvent
{
long value;
};The consumer will print the event value to standard output and also notify the publishing thread when everything is processed:
struct PrintingEventHandler : Disruptor::IEventHandler< LongEvent >
{
explicit PrintingEventHandler(int toProcess) : m_actuallyProcessed(0), m_toProcess(toProcess)
{}
void onEvent(LongEvent& event, int64_t, bool) override
{
std::cout << "Event: " << event.value << std::endl;
if (++m_actuallyProcessed == m_toProcess)
m_allDone.notify_all();
}
void waitEndOfProcessing()
{
std::unique_lock<decltype(m_mutex)> lk(m_mutex);
m_allDone.wait(lk);
}
private:
std::mutex m_mutex;
std::condition_variable m_allDone;
int m_toProcess;
int m_actuallyProcessed;
};Now we can wire all the things together:
auto const ExpectedNumberOfEvents = 10000;
auto const RingBufferSize = 1024;
// Instantiate and start the disruptor
auto eventFactory = []() { return LongEvent(); };
auto taskScheduler = std::make_shared< Disruptor::ThreadPerTaskScheduler >();
auto disruptor = std::make_shared< Disruptor::disruptor<LongEvent> >(eventFactory, RingBufferSize, taskScheduler);
auto printingEventHandler = std::make_shared< PrintingEventHandler >(ExpectedNumberOfEvents);
disruptor->handleEventsWith(printingEventHandler);
taskScheduler->start();
disruptor->start();
// Publish events
auto ringBuffer = disruptor->ringBuffer();
for (auto i = 0; i<ExpectedNumberOfEvents; ++i)
{
auto nextSequence = ringBuffer->next();
(*ringBuffer)[nextSequence].value = i;
ringBuffer->publish(nextSequence);
}
// Wait for the end of execution and shutdown
printingEventHandler->waitEndOfProcessing();
disruptor->shutdown();
taskScheduler->stop();For more details, please refer the original Java Disruptor documentation.