ROOT vecops deltaphi type error in RDataFrames

[jd550179]

Check duplicate issues.

• Checked for duplicates

Description

When calculating the deltaPhi between two scalars we get the following error:

df = df.Define("dPhiHbbHtautau", f"ROOT::VecOps::DeltaPhi<float, float>(HbbFourVector_{systematic.suffix}.Phi(), LLFourVector_{systematic.suffix}.Phi() )")

raceback (most recent call last):
  File "/user/jdegens/analysis/dihiggs/plotting/HHARD/bin/bbttManager.py", line 26, in <module>
    analysis_manager.run_analyses_local()  # runs all rdf simultaneously
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/user/jdegens/analysis/dihiggs/plotting/HHARD/python/utils/wrappers.py", line 12, in timeit_wrapper
    result = func(*args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^^^
  File "/user/jdegens/analysis/dihiggs/plotting/HHARD/python/managers/AnalysisManager.py", line 199, in run_analyses_local
    R.RDF.RunGraphs(dfs)
cppyy.gbl.std.runtime_error: unsigned int ROOT::RDF::RunGraphs(vector<ROOT::RDF::RResultHandle>) =>
    runtime_error: RDataFrame: type mismatch: column "dPhiHbbHtautau" is being used as float but the Define or Vary node advertises it as double

I am not exactly sure what the error is but if I do:

ROOT::VecOps::DeltaPhi<float, float>(HbbFourVector_{systematic.suffix}.Phi(), LLFourVector_{systematic.suffix}.Phi() )"

it runs without an error. Also this fixes the error:

df = df.Define("deltaphi", f"(float) ROOT::VecOps::DeltaPhi<float, float>(HbbFourVector_{systematic.suffix}.Phi(), LLFourVector_{systematic.suffix}.Phi() )")

There is probably something wrong with the type of the output of deltaPhi. We also use deltaR calls and we dont see this problem. It was observed in v32.0.08 and was not there previously in v30.

Reproducer

I dont have an easy reproducer at the moment...

ROOT version

ROOT Version: 6.32.08
Built for linuxx8664gcc on Dec 03 2024, 17:12:25
From tags/6-32-08@6-32-08

Installation method

views

Operating system

linux alma9

Additional context

No response

[vepadulano]

Dear @jd550179 ,

Thanks for reaching out! The DeltaPhi function return type is the common type between the types of the operands. Judging by your calls to .Phi() in the expression, I'm inferring both operands are of type double, so also DeltaPhi returns a double. The error then is quite clear:

column "dPhiHbbHtautau" is being used as float but the Define or Vary node advertises it as double

It means that the column you are defining with the return value of DeltaPhi has type double but somewhere else in the computation graph it's being used as float. The most probable cause is that you are using this column as parameter to some other expression and there the type mismatch happens. Could you check where else in your analysis you're using that column? Then we could reason about it together better.

And just for confirmation, calling ROOT::VecOps::DeltaPhi<float, float> forces the return type to be float, so clearly that error will disappear, although it doesn't remove the underlying cause.

[jd550179]

Sorry i noticed i made a copy paste error when i created the issue. The following gives the error:

df = df.Define("dPhiHbbHtautau", f"ROOT::VecOps::DeltaPhi<float>(HbbFourVector_{systematic.suffix}.Phi(), LLFourVector_{systematic.suffix}.Phi() )")

this fixes it:

df = df.Define("dPhiHbbHtautau", f"ROOT::VecOps::DeltaPhi<float, float>(HbbFourVector_{systematic.suffix}.Phi(), LLFourVector_{systematic.suffix}.Phi() )")

I noticed that the template of the function has changed from root version 6.30 to 6.32 so I am quite certain it has something to do with that.
v32, line 2934: https://root.cern.ch/doc/v632/RVec_8hxx_source.html
v30, line 2876: https://root.cern.ch/doc/v630/RVec_8hxx_source.html

(the v32 way is the same as in the master)

[vepadulano]

Hi @jd550179 ,

Yes you are right there was a change in signature, but it does not invalidate my comment from before. The call to ROOT::VecOps::DeltaPhi<float>(HbbFourVector_{systematic.suffix}.Phi(), LLFourVector_{systematic.suffix}.Phi() ) will infer that the type of Phi() is double, so the common type will be double (e.g. https://godbolt.org/z/7zof766rc). The error you see is still valid, so as I was mentioning you need to understand where else in your computation graph you are trying to use a double value as a float. If you want you can paste the relevant part of your code here. Another tip is you could enable the verbose logging and that might tell you more about the functions you are defining and which could be the responsible, see https://root.cern/doc/master/classROOT_1_1RDataFrame.html#rdf-logging