log different event counts in event header

[tomeichlersmith]

For EoT estimations, the total number of electrons "thrown" (i.e. events started) is important and currently, there is no storage of that number. Currently, we are basically limited to avoiding use of the maxTriesPerEvent parameter when we wish to estimate the EoT of a generated sample.

We can avoid this limitation by storing some extra integers in the run and event headers. I'm thinking of storing the total counts of number of events thrown and number events processed in the run header and keeping event indices (thrown and accepted) in the event header.

[tomeichlersmith]

We could log the number of tries it took to create a certain event in the EventHeader. This would allow us to then calculate EoT with

$$ N \sum_{j=1}^N n_j / \sum_{i=1}^N w_i $$

where $N$ is the number of events in the output file, $w_i$ is the weight of of the i'th event, and $n_j$ is the number of tries it took to create the j'th event. This is pretty easily "vectorized" as well which is nice and pretty cleanly maps over many files even if they vary in the number of events within them.

I'm trying to think if we could reduce the two sums into one sum and then store an "EoT weight" in the event header which just needs to be summed and multipled by the number of events to get the EoT but I don't know if that's easy or worth it.

[tomeichlersmith]

I actually think we need to move the try count to the RunHeader instead of the EventHeader. As currently configured, if the last event attempted is not succesful, then the count of events leading up to that point is just thrown away. i.e. the sum of the tries for individual events is only equal to the total number of actual tries if the very last event is successful.