Choose timezone
Your profile timezone:
45th ROOT Parallelism, Performance and Programming Model
Present: Kim, Danilo, Enric, Javi, Guilherme, Philippe, Jim
Kim:
- Question by Axel: can the TTree event number be used rather than the TMVA EventId? Kim needs to look it up since in the options we are not using TTreeFormula but TFormula.
Discussion:
----
PyROOT: We were on the right track. Inline with what people requested. For example interoperability with NumPy. Some people even mentioned Pandas. We need to continue with the effort in this direction.
Some other things need to be discussed. Do we marry NumPy so much that we pythonise signatures to use numpy.
A good type which can mirror NumPy in C++ is the std::array because it does not resize. Perhaps even the span is better since it does not need to own the memory. We agree that we need more time to see what the modern C++ will do in 5y from now.
----
Documentation:
- Doc of PyROOT: we need to document in general PyROOT and the single pieces, for example the Pythonisation. We want to keep the doc unified and have a special section in the class doc in C++ : this will lure in casual users and satisfy the PyROOT users. Also the style of the code examples must be clearly distinguishable from the C++ ones.
----
Easier way to install ROOT:
The impression is that people sometimes try to install something other than ROOT because ROOT "is difficult". What can we do about this?
- It can help to install packages with "apt". But this makes almost necessary to have a PPA but then it's clumsy. The binaries are clumsy in tarballs because they do not have deps. Probably the best idea is to have deb and rpm. The problem is that the cost of achieving it is not negligible.
- Involve Oksana for the topic about mini-builds (core+rio+cling+pyroot with -DNano) or nano+RDF (-Dana... inventing names here...)
There are minutes attached to this event.
Show them.