Alternatives for `EventCounter` removed with`GaudiAlg`

[m-fila]

A number of repositories uses the EventCounter that was part of the deprecated GaudiAlg:

• key4hep/k4RecTracker
• HEP-FCC/k4RecCalorimeter
• HEP-FCC/FCCDIRAC
• HEP-FCC/fcc-tutorials
• HEP-FCC/FCCSW

The EventCounter algorithm itself is simple as it just increments its internal counter on each execution and prints the counter's value on finalize and heartbeat on execution:

• In case the heartbeat printout on execution is desired then a replacement functional algorithm using internal atomic counter could be rather easily written. It should be possible to get the same behavior as in EventCounter
• In case the printout only at the finalize is sufficient then there is a newer mechanism in form of Gaudi::Accumulator::Counter and MessageSvcSink (MessageSvcSink prints the values hold in each counter at the finalize). With it, it's possible to:
  • add Gaudi::Accumulator::Counter member to some algorithms that should be counted (for instance the input/output algorithms or filters) and remove the EventCounter from affected repositories
  • write another EventCounter algorithm that does just the counting (can be copied from GaudiTestSuite)

The solutions involving a separate algorithm doing counting would require to put it in sequential sequencer when used with the AvalancheSchedulerSvc (no data dependencies declared so the algorithm could be executed in any order with the respect to the other algorithms)

Alternatively, the information about the event number as seen from a scheduler can be accesses with EventContext

[tmadlener]

Can we have a sort of combination of things? Something like an EventCounter algorithm that simply prints the event number from the EventContext for the hearbeat with an internal Gaudi::Accumulator::Counter for printing the summary?

[m-fila]

Can we have a sort of combination of things? Something like an EventCounter algorithm that simply prints the event number from the EventContext for the hearbeat with an internal Gaudi::Accumulator::Counter for printing the summary?

We can, no problem

As we discussed it's used mainly for debugging and it's not clear if there are better ways to debug especially in multi-threaded environment