Store number of simulated events

[gipert]

Storing the number of simulated events is crucial for post-processing. At the moment one could look for unique evtids in the vertices table, but this require some computation.

evtids = lh5.read_as("stp/vertices/evtid", "output.lh5", "np")
n_g4ev = len(np.unique(evtids))

Why not then just store a Scalar with the number of simulated events?

[tdixon97]

I think its already helpful for post-processing, but maybe we need to even store a little bit more information, ie the tcm like grouping as we discussed?

[EricMEsch]

Vertices should already contain one unique entry per event (at least given if one event is represented by one primary. I guess this will be different for multiple primaries per event). I am not sure about the .lh5 format, but if you use the .hdf5 files there already is an entry that stores the amount of entries: f["hit"]["vertices"]["evtid"]["entries"] . Which means the number exists and is stored somewhere. I am sure .lh5 has something similar.

In case there are multiple primaries per event the last entry of evtids should correspond to the number of simulated events (-1), because primary vertices will always be stored. That should at least be faster than len(np.unique()) i assume.

[ManuelHu]

I am not sure about the .lh5 format, but if you use the .hdf5 files there already is an entry that stores the amount of entries: f["hit"]["vertices"]["evtid"]["entries"] . Which means the number exists and is stored somewhere. I am sure .lh5 has something similar.

no, those entries are removed when converting to LH5.

In case there are multiple primaries per event the last entry of evtids should correspond to the number of simulated events (-1), because primary vertices will always be stored. That should at least be faster than len(np.unique()) i assume.

not in the case of multithreading. There the distribution of event ids between threads is - unfortunately - quite complex.

[gipert]

ie the tcm like grouping as we discussed?

maybe, we need to think about whether it's the right place. But let's discuss this in legend-exp/reboost#16