RDataFrame not using aliases stored in a file

[apetukho]

Dear ROOT experts,

I’m trying to create histograms using RDataFrame, but I’m unable to use the Aliases for the branches I’ve stored in a file beforehand. Here’s a reproducer

import ROOT

def create_file(file_name, tree_name):
    df = ROOT.ROOT.RDataFrame(10).Define("x", "gRandom->Rndm()")
    df.Snapshot(tree_name, file_name, "")

def create_alias(file_path, tree_name):
    file = ROOT.TFile(file_path, 'update')
    tree = file.Get(tree_name)
    tree.SetAlias("x_alias", "x")
    tree.Write('', ROOT.TObject.kOverwrite)
    file.Close()

def create_hist_tree(file_path, tree_name):
    file = ROOT.TFile(file_path)
    tree = file.Get(tree_name)
    tree.Draw("x_alias")

def create_hist_df(file_path, tree_name):
    df = ROOT.RDataFrame(tree_name, file_path)
    df.Histo1D(("h", "h", 10, 0, 100), "x_alias")

def main():
    file_path = 'f.root'
    tree_name = 'tree'
    print("Creating file")
    create_file(file_path, tree_name)
    print("Creating alias")
    create_alias(file_path, tree_name)
    print("Making hist with TTree")
    create_hist_tree(file_path, tree_name)
    print("Making hist with RDataFrame")
    create_hist_df(file_path, tree_name)

if __name__ == "__main__":
    main()

The output of the program is

Creating file
Creating alias
Making hist with TTree
Info in <TCanvas::MakeDefCanvas>:  created default TCanvas with name c1
Making hist with RDataFrame
Traceback (most recent call last):
  File "rdf_alias.py", line 36, in <module>
    main()
  File "rdf_alias.py", line 33, in main
    create_hist_df(file_path, tree_name)
  File "rdf_alias.py", line 21, in create_hist_df
    df.Histo1D(("h", "h", 10, 0, 100), "x_alias")
  File "/home/alex/root_62806/root/lib/ROOT/_pythonization/_rdataframe.py", line 379, in __call__
    res = self._original_method(model, *args[1:])
TypeError: Template method resolution failed:
  none of the 4 overloaded methods succeeded. Full details:
  ROOT::RDF::RResultPtr<TH1D> ROOT::RDF::RInterface<ROOT::Detail::RDF::RLoopManager,void>::Histo1D(experimental::basic_string_view<char,char_traits<char> > vName) =>
    TypeError: takes at most 1 arguments (2 given)
  ROOT::RDF::RResultPtr<TH1D> ROOT::RDF::RInterface<ROOT::Detail::RDF::RLoopManager,void>::Histo1D(const ROOT::RDF::TH1DModel& model, experimental::basic_string_view<char,char_traits<char> > vName, experimental::basic_string_view<char,char_traits<char> > wName) =>
    TypeError: takes at least 3 arguments (2 given)
  ROOT::RDF::RResultPtr<TH1D> ROOT::RDF::RInterface<ROOT::Detail::RDF::RLoopManager,void>::Histo1D(experimental::basic_string_view<char,char_traits<char> > vName, experimental::basic_string_view<char,char_traits<char> > wName) =>
    TypeError: could not convert argument 1
  ROOT::RDF::RResultPtr<TH1D> ROOT::RDF::RInterface<ROOT::Detail::RDF::RLoopManager,void>::Histo1D(const ROOT::RDF::TH1DModel& model = {"", "", 128U, 0., 0.}, experimental::basic_string_view<char,char_traits<char> > vName = "") =>
    runtime_error: Unknown column: "x_alias"
  Failed to instantiate "Histo1D(ROOT::RDF::TH1DModel*,std::string)"

So the TTree::Draw() approach is able to use the x_alias name for the x branch, but the RDataFrame cannot find such column

  ROOT::RDF::RResultPtr<TH1D> ROOT::RDF::RInterface<ROOT::Detail::RDF::RLoopManager,void>::Histo1D(const ROOT::RDF::TH1DModel& model = {"", "", 128U, 0., 0.}, experimental::basic_string_view<char,char_traits<char> > vName = "") =>
    runtime_error: Unknown column: "x_alias"

Why might that happen and is there a way around it? I would like to avoid using Alias() function of RDataFrame and make aliases before the processing to not complicate the code.

Thanks in advance,
Aleksandr

---

ROOT Version: 6.28/06
Platform: Ubuntu 20.04.6
Compiler: precompiled

---

[mczurylo]

Hi @apetukho,

thanks for reporting the issue together with the reproducer. I can indeed see the same issue, I will take a look into this now and will report back as soon as I have more insight.

Cheers,
Marta

[mczurylo]

Hi @apetukho,

I tried to dig into this, but I’m afraid you’d have to add the df.Alias() in your create_hist_df function as: df.Alias("x_alias", "x").Histo1D(("h", "h", 10, 0, 100), "x_alias"). Unless @eguiraud has some more insights?

Cheers,
Marta