I was asked to check if RDataFrame supports correctly a feature of Tree/Chains that we have been using for a while in our analysis setups. In this setup we friend an auxiliary tree to our main tree, where the aux tree has fewer entries in it than the main tree, but has an index built on it for a branch that has the same name as another branch in the main tree. The result of this setup mimics an SQL join.
I tried such a setup out on an RDataFrame and unfortunately the behaviour is incorrect. Please see the following notebook which contains a pretty minimal demonstration of the difference in behaviour between a simple TTree Draw and the RDataFrame Histo1D ... https://cernbox.cern.ch/index.php/s/TRygKfYwboH1OXR