LCIO to EDM4hep conversion still loses collection parameters

[Zehvogel]

Check duplicate issues.

• Checked for duplicates

Goal

In lcio, it was possible (and used to be) to attach parameters to a collection (not just the PID stuff). As far as I can tell, these are not converted at the moment.

So the info stored by e.g. the thrust calculation in ILD miniDST files gets lost when converting to edm4hep.

Operating System and Version

any

compiler

any

The version of the key4hep stack

any

Package Version

any

Reproducer

• Obtain minidst.slcio
• convert it:

check_missing_cols --minimal minidst.slcio > colls.txt
lcio2edm4hep minidst.slcio minidst.edm4hep.root colls.txt

• compare

dumpevent minidst.slcio 1 | grep Thrust
podio-dump -d minidst.edm4hep.root | grep Thrust

Additional context

I think this was discussed a couple of times in the edm4hep meetings but this part was maybe missed as there are just too many places in lcio to store stuff...

[tmadlener]

This might also be one of the things that change fundamentally / conceptually between the two worlds. In principle (i.e. currently) there are no collection parameters at event level any longer in EDM4hep. Either they are valid for all collections in a file, in which case they become collection metadata or they would simply become event level parameters.

Having said that: Since we could just re-use the naming convention for the collection metadata on the event level, effectively we would be in a similar place as for LCIO. The key difference would be the interface to access and store this data.

Also, I don't think there is currently a mechanism for storing such event level parameters easily through k4FWCore, but @jmcarcell might know better.

[Zehvogel]

Either they are valid for all collections in a file, in which case they become collection metadata or they would simply become event level parameters.

So far I have only encountered the latter case, which would also be what I would suggest. I.e. I would be happy if any found lcio collection parameters would be converted to edm4hep event parameters.

I don't know what we would want to do about the edm4hep to lcio direction. Personally, I don't care too much. There might be some older analyses Marlin processors out there, but I think we could solve that when there is actual demand ;)