I looking for a correct way to safe get text data from boost::log::sinks::text_ostream_backend. I'm currently getting data only after everything was recorded in sink backend. I want get copy of backend internal buffer (m_ss) while data can still be produced.
using Logger = boost::log::sources::severity_channel_logger_mt<boost::log::trivial::severity_level, std::string>;
class MyClass
{
mutable std::mutex m_mutex;
mutable Logger m_logger {boost::log::keywords::channel = "MyClass"};
std::stringstream m_ss;
bool m_completed {false};
public:
void initLogging(const std::string& id)
{
using namespace boost::log;
using Sink = sinks::synchronous_sink<sinks::text_ostream_backend>;
auto backend = boost::make_shared<sinks::text_ostream_backend>();
backend->add_stream(boost::shared_ptr<std::stringstream>(&m_ss, boost::null_deleter()));
auto sink = boost::make_shared<Sink>(backend);
sink->set_filter(expressions::has_attr<std::string>("ID") && expressions::attr<std::string>("ID") == id);
core::get()->add_sink(sink);
m_logger.add_attribute("ID", attributes::constant<std::string>(id));
}
void doSomething()
{
BOOST_LOG_SEV(logger, boost::log::trivial::severity_level::debug) << "TEST";
BOOST_LOG_SEV(logger, boost::log::trivial::severity_level::error) << "TEST";
// Stop write logs
std::lock_guard<std::mutex> lock(m_mutex);
m_completed = true;
}
void access()
{
std::lock_guard<std::mutex> lock(m_mutex);
if (m_completed) {
// Can read from stream
useStreamData(m_ss.str());
}
}
};
If the stream is only accessed by the sink backend then the easiest way to synchronize access is to use locked_backend
.
auto locked_backend = sink->locked_backend();
Here, the locked_backend
object is a smart pointer, which locks the sink frontend and therefore prevents logging threads from accessing the backend and, consequently, the m_ss
stream. You can use m_ss
safely while locked_backend
exists.
There are also other ways to achieve this. For example, you can implement a custom sink backend or implement a stream buffer that would allow synchronized access to the accumulated data. There are a number of articles and blog posts about writing custom stream buffers, here is one example.