How to get value of RResultPtr outside a loop?

[ajaydeo]

---

ROOT Version:_ 6.24/02
Platform:Linux/Fedora 34
Compiler: gcc version 11.3.1 20220421

---

I have the following lines in my code. I want to access the elements of vectorPtr outside the for loop.

  ROOT::RDataFrame df(treename, filename);
  std::vector<ROOT::RDF::RResultPtr<TH1D>> histos;
  for (const std::string &col : df.GetColumnNames()) {
    auto h = df.Filter(isPositive, {col}).Histo1D(TH1D(col.c_str(),col.c_str(),8192.,0,8192),col.c_str());
    histos.push_back(h);
    auto vectorPtr = df.Take<double>(col);
  }

How can I access the elements of vectorPtr outside the for loop? I need those for further analysis. I tried to define vectorPtr before/outside the for loop as:

std::vector<ROOT::RDF::RResultPtr<double>> vectorPtr;

and inside the for loop:

vectorPtr = df.Take<double>(col);

which gives the the following error:

/./GetValue.cxx:39:15: error: no viable overloaded '='
    vectorPtr = df.Take<double>(col);
    ~~~~~~~~~ ^ ~~~~~~~~~~~~~~~~~~~~

Please note that I don’t want to trigger event loop at this stage.

[Danilo]

Hi,

Yes, that’s expected. The method RDataFrame::Take is indeed templated with the type of the return collection. By default, that is a std::vector. Therefore, the code would look like this:

  // Define a dummy RDF for this example
  ROOT::RDataFrame df(8);
  auto df2cols = df.Define("a",[](){return 1;}).Define("b",[](){return 2;});
  using vint = std::vector<int>;
  std::vector<ROOT::RDF::RResultPtr<vint>> contents;

  for (const auto &col : df2cols.GetColumnNames()) {
    contents.emplace_back(df2cols.Take<int>(col));
  }

  // Just print the content
  for (auto i : ROOT::TSeqI(8)) {
    std::cout << contents[0]->at(i) << " " << contents[1]->at(i) << std::endl;
  }

I hope this helps.

Cheers,
D

[ajaydeo]

Dear @Danilo,
New year’s greetings! Thank you for the solution.

I have one more question, the answer to that will lead me to the
completion of my first data analysis code using RDataFrame.

I have a ROOT Tree which contains, say 3 branches: x, y, z.
x & y are variable size arrays of type ULong64_t and
z is a simple variable of type Int_t.

I want to generate the following two types of histograms.

1. If the value of z is greater than 1 then I want to find the difference
   (which can be positive or negative) between elements of y.
   And if the absolute difference is say, between 50 and 100, then I want to generate
   2-dimentional histogram of the corresponding elements of x.

2. Also, if the value of z is greater than 1 then for given values of elements of x
   (say “eX1” and “eX2”, which the user can pass as arguments to the code);
   I want to find the difference (which can be positive or negative) between corresponding
   elements of y. In this case, I want to generate 1-dimentional histogram of the difference.

These two tasks will require two different codes.
How can this be achieved? Can you suggest me recipes?

Best regards,

Ajay

[Danilo]

Hi Ajay,

You can easily achieve all this with Defines (that we have already used in this thread), Filters, Histo1D and Histo2D.
Pre-packed examples can be found in the DataFrame tutorials. For a good start you can have a look to this one.

Cheers,
D

[ajaydeo]

Dear @Danilo,

As suggested by you, I worked on the 1. above to generate a 2D histogram.
However, there is a problem with multithreading as mentioned in my new post.

For your reference, I am attaching a sample ROOT file here and the code is
attached (2.4 KB).

I am running the code at the ROOT prompt as: .x rdf_his2D.cxx+

Any help is highly appreciated.

Regards,

Ajay

[ajaydeo]

Dear @Danilo,

As suggested by you, I worked on the 1. above to generate a 2D histogram.
However, there is a problem with multithreading as mentioned in my new post.

For your reference, I am attaching a sample ROOT file here and the code (2.4 KB).

I am running the code at the ROOT prompt as: .x rdf_his2D.cxx+

Any help is highly appreciated.

Regards,

Ajay