void l1menu::TriggerRatePlot::addSample( const l1menu::ISample& sample, std::vector<TriggerRatePlot>& ratePlots )
{
float weightPerEvent=sample.eventRate()/sample.sumOfWeights();
// Create cached triggers for each of the rate plots, which depending on the concrete type
// of the ISample may or may not significantly increase the speed at which this next loop happens.
std::vector< std::unique_ptr<l1menu::ICachedTrigger> > cachedTriggers;
for( const auto& ratePlot : ratePlots ) cachedTriggers.push_back( sample.createCachedTrigger( *ratePlot.pTrigger_ ) );
// Now instead of calling addSample() for each TriggerRatePlot individually, get each IEvent from the sample
// and pass that to each rate plot. This is because (depending on the ISample concrete type) getting the
// IEvent can be computationally expensive.
std::vector< std::unique_ptr<l1menu::ICachedTrigger> >::const_iterator iTrigger;
std::vector<TriggerRatePlot>::iterator iRatePlot;
for( size_t eventNumber=0; eventNumber<sample.numberOfEvents(); ++eventNumber )
{
const l1menu::IEvent& event=sample.getEvent(eventNumber);
for( iTrigger=cachedTriggers.begin(), iRatePlot=ratePlots.begin();
iTrigger!=cachedTriggers.end() && iRatePlot!=ratePlots.end();
++iTrigger, ++iRatePlot )
{
iRatePlot->addEvent( event, *iTrigger, weightPerEvent );
}
} // end of loop over events
}
void l1menu::TriggerRatePlot::addSample( const l1menu::ISample& sample )
{
float weightPerEvent=sample.eventRate()/sample.sumOfWeights();
// Create a cached trigger, which depending on the concrete type of the ISample
// may or may not significantly increase the speed at which this next loop happens.
std::unique_ptr<l1menu::ICachedTrigger> pCachedTrigger=sample.createCachedTrigger( *pTrigger_ );
for( size_t eventNumber=0; eventNumber<sample.numberOfEvents(); ++eventNumber )
{
addEvent( sample.getEvent(eventNumber), pCachedTrigger, weightPerEvent );
} // end of loop over events
}