How to set the splitlevel for branches in the output root file

[kratsg]

Hi,

I'd like to use uproot (and columnar analysis) to analyze Delphes outputs. Right now, all the branches (e.g. Jet.Mass) seem to be set at splitlevel=0 ("unsplit") which means it takes some time to deserialize the branch before one can load it up.

Is there a way we can configure the splitlevel when Delphes writes these branches out? (This is Snowmass2021-related).

Thanks,

Giordon

[Pavel Demin]

Delphes uses the default splitlevel value. According to the ROOT documentation at ​this link, the default value is 1.

All the branches are created by the code on line 41 in external/ExRootAnalysis/ExRootTreeBranch.cc:
​https://github.com/delphes/delphes/blob/master/external/ExRootAnalysis/ExRootTreeBranch.cc#L41

The splitlevel parameter is not currently configurable. So please set it in the code and let me know if it solves the problem. If setting splitlevel to a different value solves the problem with uproot, I can make splitlevel configurable or just set it to the value that works with uproot.

[kratsg]

Hi Pavel, thanks for the quick response. I noted in the documentation that the default splitlevel in all the ROOT::TTree::Branch() functions are 99 (fully split) so I'm wondering how the Delphes output seems to indicate a splitlevel of 0. Maybe I'm misreading things or looking at the branches incorrectly:

root [1] TBranch* branch = Delphes->GetBranch("Jet.Mass")
(TBranch *) 0x5bd3980
root [2] branch->GetSplitLevel()
(int) 0

I'll play around with the parameter and see how things work. Perhaps we can close the issue at this point and if I find that a configurable one improves uproot, I can open up another ticket and reference this one.

[Pavel Demin]

As far as I know (from a ROOT user point of view) Jet.Mass is a leaf not a branch. Only Jet is a branch.

The splitlevel of the Jet branch is indeed 99:

root [1] TBranch* branch = Delphes->GetBranch("Jet")
(TBranch *) 0x55b92a912a10
root [2] branch->GetSplitLevel()
(int) 99

So, when Delphes creates the Jet branch the splitlevel the Jet branch is the default 99 value (fully split). Jet.Mass is a result of the splitting (a leaf or a subbranch).

Have you tried to point uproot to the Jet branch instead of the Jet.Mass leaf/subbranch?

[kratsg]

Indeed, you're right, the splitlevel seems ok -- and so this is probably not the issue I'm seeing. I'll explore further. Thanks!