Wrong interaction between VariationsFor and no variations in distributed RDataFrame

[vepadulano]

Check duplicate issues.

• Checked for duplicates

Description

If a branch of the computation graph has no variations and the user calls VariationsFor on an action, the resulting map should have only the nominal value available. In distributed mode, a wrong interaction leads to segfaults and errors such as:

  File "/home/vpadulan/programs/rootproject/rootbuild/distrdf-debug/lib/DistRDF/Backends/Base.py", line 99, in distrdf_mapper
    mergeables = get_mergeable_values(rdf_plus.rdf, current_range.id, computation_graph_callable,
    ^^^^^^^^^^^
  File "/home/vpadulan/programs/rootproject/rootbuild/distrdf-debug/lib/DistRDF/Backends/Base.py", line 59, in get_mergeable_values
    mergeables = [Utils.get_mergeablevalue(action) for action in actions]
      ^^^^^^^^^^^^^^^^^
  File "/usr/lib64/python3.13/functools.py", line 929, in wrapper
    return dispatch(args[0].__class__)(*args, **kw)
    ^^^^^^^^^^^^^^^
  File "/home/vpadulan/programs/rootproject/rootbuild/distrdf-debug/lib/DistRDF/Backends/Utils.py", line 164, in get_mergeablevalue
    return ROOT.Detail.RDF.GetMergeableValue(resultptr)
      ^^^^^^^^^^^^^^^^^
TypeError: Template method resolution failed:
  unique_ptr<ROOT::Detail::RDF::RMergeableValue<TH1D>,default_delete<ROOT::Detail::RDF::RMergeableValue<TH1D> > > ROOT::Detail::RDF::GetMergeableValue(ROOT::RDF::RResultPtr<TH1D>& rptr) =>
    TypeError: could not convert argument 1
  unique_ptr<ROOT::Detail::RDF::RMergeableVariations<TH1D>,default_delete<ROOT::Detail::RDF::RMergeableVariations<TH1D> > > ROOT::Detail::RDF::GetMergeableValue(ROOT::RDF::Experimental::RResultMap<TH1D>& rmap) =>
    SegmentationViolation: segfault in C++; program state was reset
  Failed to instantiate "GetMergeableValue(ROOT::RDF::Experimental::RResultMap<TH1D>*)"
  Failed to instantiate "GetMergeableValue(ROOT::RDF::Experimental::RResultMap<TH1D>)"

FYI @gpetruc

Reproducer

import ROOT
from dask.distributed import Client, LocalCluster

RunGraphs = ROOT.RDF.Experimental.Distributed.RunGraphs
VariationsFor = ROOT.RDF.Experimental.Distributed.VariationsFor
RDataFrame = ROOT.RDF.Experimental.Distributed.Dask.RDataFrame


def create_connection() -> Client:
    cluster = LocalCluster(n_workers=1, threads_per_worker=1, processes=True)
    client = Client(cluster)
    return client


def run(conn):
    h = RDataFrame(1, daskclient=conn).Define("x", "42").Histo1D(("h","h",1,0,10),"x")
    vars = VariationsFor(h)
    h.GetValue()

    print(f"{h.GetEntries()=},{vars.GetKeys()=}")


if __name__ == "__main__":
    with create_connection() as conn:
        run(conn)

ROOT version

Any

Installation method

Any

Operating system

Any

Additional context

No response