I just became aware of the existence of SamplingInput_source. It samples events at random but starts at the first event of a file unless the skipToEvent parameter is used. That parameter is not practical to set for grid jobs. So the start of each file is likely to be over-sampled compared to the end. Very similar to issue #109. Ideally one wants to sample events from each dataset uniformly.
Random access isn't required for current Mu2e use cases of this source. An interface to allow a 'random' starting point as Andrei specifies would be useful.
Hello,
I just became aware of the existence of SamplingInput_source. It samples events at random but starts at the first event of a file unless the skipToEvent parameter is used. That parameter is not practical to set for grid jobs. So the start of each file is likely to be over-sampled compared to the end. Very similar to issue #109. Ideally one wants to sample events from each dataset uniformly.
Andrei