From my (@perilousapricot) notes with @jpivarski:
The idea is to get a TTree from a ROOT file into a Spark dataframe as optimally as possible.
My (@perilousapricot) reason for re-doing existing work:
The existing Root4J interface is targeted towards re-implementing the C++ ROOT API in Java. So, you point it to a ROOT file, and get accessors that let you access ROOT objects from a TFile like they were regular Java objects. So, if you were suitably encouraged, you could modify and use these objects to do an analysis.
Java, however, is a completely different language than C++ (even though ROOT If you want to get to there, you have to combine the I/O, deserialization, object instatiation layers. Once that's done, you have to then add member functions to these instances you get.