Open eguiraud opened 2 years ago
Step number 1. I just saw that the problem has nothing to do with jitted code. This is also broken:
ROOT::RDataFrame df(1);
auto df2 = df.Define("x", [] { return 1; })
.Vary("x", []() { return ROOT::RVecI(3); }, {}, 3)
.Filter([](int x) { return x > 0; }, {"x"});
auto s = df2.Sum<int>("x");
auto ss = ROOT::RDF::Experimental::VariationsFor(s);
ROOT::RDF::SaveGraph(df, "res.dot");
I am taking a look now.
Reproducer:
yields:
which is of course wrong.
We should instead hide the filters that are replicated for each systematic variation.