Always convert EventHeader to lcio

[Zehvogel]

BEGINRELEASENOTES

• Modified the edm4hep to lcio converter to always convert theEventHeader to have event number etc. available in lcio.

ENDRELEASENOTES

Does anyone have a good idea for a test?

Resolves: #153

[Zehvogel]

Unfortunately, I cannot reproduce the ci failure locally, for me all tests pass :/

[tmadlener]

Is the EventHeader missing from here?

k4MarlinWrapper/test/gaudi_opts/test_global_converter_maps.py

Line 41 in 2eeb466

podioInput.collections = ["MCParticles"]

[Zehvogel]

Is the EventHeader missing from here?

k4MarlinWrapper/test/gaudi_opts/test_global_converter_maps.py

Line 41 in 2eeb466

podioInput.collections = ["MCParticles"]

I guess so, lets see. Still puzzled why I don't get an error locally :D

[tmadlener]

Maybe the better solution in this case would be to make sure that the PodioInput also reads the EventHeader collection?

[Zehvogel]

hmpf yes I found why it does not error locally for me... the tests also need the install beforehand so if I click on run tests in my editor like a lazy person they run against the k4marlinwrapper from the environment e.g. nightlies

[Zehvogel]

Maybe the better solution in this case would be to make sure that the PodioInput also reads the EventHeader collection?

I agree

[Zehvogel]

Waiting for

• Add EventHeader constant EDM4hep#239 to be able to remove the plain string here
• PodioInput: always read EventHeader k4FWCore#160 to be able to remove the explicit reading of the EventHeader

[tmadlener]

Do we still need the EventHeader after merging key4hep/k4FWCore#160 ? If no, then I think we should remove it here.

[Zehvogel]

we should be able to remove it and that was my plan for today :)

[tmadlener]

It seems we still need it: https://github.com/key4hep/k4MarlinWrapper/actions/runs/6887433198/job/18734625662?pr=162#step:6:2428

I am not entirely sure why though, we pick up the latest version of k4FWCore.

[Zehvogel]

I think I know why, I thought I tested it but as we set the collections to read after calling the constructor we just overwrite the list that has the EventHeaderName in it :(

[Zehvogel]

so appending the collections works, not sure if this is really an improvement :/

[Zehvogel]

I changed the steering files for the tests back again as key4hep/k4FWCore#161 made the change unnecessary.

[tmadlener]

I would still like to have tests for this, but the "simplest" I can think of is to write a Marlin processor that we then run wrapped and check whether the contents of the EventHeader are as expected (first filling them with the EventHeaderCreator algorithm). The processor is almost trivial, but it introduces quite a bit of boiler plate...

[Zehvogel]

anajob will print the real event numbers, so we could write out the lcio file and use a bit of ugly bash :D

        cout << "///////////////////////////////////" << endl;
        cout << "EVENT: " << evt->getEventNumber() << endl
            << "RUN: " << evt->getRunNumber() << endl
            << "DETECTOR: " << evt->getDetectorName() << endl
            << "COLLECTIONS: (see below)" << endl;
        cout << "///////////////////////////////////" << endl << endl;

[tmadlener]

Ah yes, we could do something like

eventHeaderCreator = EventHeaderCreator("eventHeaderCreator", runNumber = 42)

output = MarlinProcessorWrapper("output")
output.ProcessorType = "LCIOOutputProcessor"
output.Parameters({"LCIOOutputFile": ["output.slcio"]})

(plus the usual gaudi boiler plate) and then run this for one event and do anajob output.slcio | grep "EVENT: 42" and fail if that fails.

[tmadlener]

Very nice. Thanks.

[Zehvogel]

@jmcarcell looks like the doctests are failing because jupytext is now missing from the stack?

[tmadlener]

Yeah, see key4hep/key4hep-spack#547

[tmadlener]

As discussed today during the EDM4hep meeting we merge this now without the doctests until we have figured out how to address them properly.