Gather branch(columns) usage from RDataFrame

[simonepigazzini]

Dear ROOT experts,

is there a way to gather from RDataFrame which branches were actually accessed during the evaluation of the computing graph?

Thank you,

simone

[eguiraud]

Hello @simonepigazzini ,

I am afraid that is not something RDF can do.

@pcanal do TTree, TChain or TFile have something like that? (if so, you could run a single-thread RDF on the TFile/TTree/TChain and then interrogate them on what branches have been accessed).

Cheers,
Enrico

[pcanal]

You can seemingly do:

auto c = mytree->GetReadCache(ntuple->GetDirectory()->GetFile());
c->GetCachedBranches()->ls();

You can also use https://root.cern.ch/doc/master/classTTreePerfStats.html