Save Sweight Factor into RDataFrame Directly

[bshi]

Dear all,

as the title indicates, can we save Sweight factor into RDataFrame directly?

I know classic TTree interfaces can do it. But I am wondering it is faster when using RDataFrame?

Thank you for your attention to my simple question.

Boan

[vpadulan]

Dear @bshi ,

Could you be more specific about your use case? For example, a small snippet of code showing what you would try to do with TTree might help.

Cheers,
Vincenzo

[bshi]

Hi Vpadulan,

Thanks for your reply.

The simple code using TTree interfaces is as follows.

TFile *file = new TFile(out_file, “recreate”);
TTree *result = tree1->CloneTree(0);// tree1 is the mother tree
double sig_sw{0};
result->Branch(“sig_sw”, &sig_sw, “sig_sw/D”);
for(Int_t i=0; i< nevent; ++i){
tree1->GetEntry(i);
sig_sw = sData->GetSWeight(i,“sig”);
result->Fill();
}
file->cd();
result->Write();
file->Close();

I want to convert them into RDataFrame classes. For example,

ROOT::RDataFrame rdf(*tree1);
auto new_rdf = rdf.Define(“sig_sw”,“expression”).Snapshot(“new_tree”,output_file);

Do you have any good ideas?

With best respect,
Boan

[vpadulan]

Hi @bshi ,

One possibility could be

TFile input_f{"original.root"};
std::unique_ptr<TTree> input_t{f.Get<TTree>("original")};


ROOT::RDataFrame df{*input_t};

df.Define("sWeight", 
          [&](std::uint64_t entry){
              return sData->GetSWeight(entry, "sig");
          },
          {"rdfentry_"})
  .Snapshot("out_tree", "out_file.root");

Caveat: the rdfentry_ special column corresponds to the current entry in the original tree traversed sequentially only in single thread. In multithreaded execution it gets shuffled (see the docs).

[bshi]

@vpadulan

uint64_t

Hi, @vpadulan.

I meet this problem when executing your code.

terminate called after throwing an instance of ‘std::runtime_error’
what(): RColumnValue: type specified for column “rdfentry_” is unsigned long but temporary column has type ULong64_t

I don’t understand it. Could you tell me what this is about?

With best respect,
Boan

[vpadulan]

Hi @bshi ,

I may have slightly mispecified the column type, instead of std::uint64_t just type unsigned long and that error should go away.

Cheers,
Vincenzo

[bshi]

Hi @vpadulan .

This problem still remains when type unsigned long.

df.Define("sWeight", 
          [&](unsigned long entry){
              return sData->GetSWeight(entry, "sig");
          },
          {"rdfentry_"})
  .Snapshot("out_tree", "out_file.root");

terminate called after throwing an instance of 'std::runtime_error'
  what():  RColumnValue: type specified for column "rdfentry_" is unsigned long but temporary column has type ULong64_t

Thanks for your help again.

Boan

[vpadulan]

Maybe with ULong64_t :)?

[bshi]

Yes, it is.

Thank you!