ROOT Team Meeting
Zoom link in announcement email; please contact rootdev@cern.ch if you did not receive it!
Meeting notes 16/01/2023
Shift report
Bertrand outgoing, Javier incoming. No urgent items to report, nightlies are posted on shush
Meetings
IO
C++ attributes for IO annotations
RooFit/TMVA
Discussion of plan of work for RooFit, nothing special to report.
Planning/godparents
No report
LIM
No report
PoW 2023
Link: https://indico.cern.ch/event/1241914/contributions/5220094/attachments/2576510/4442936/ROOT-2023-Team.pdf
Change in ratio of supervisors/supervisees. Significantly increase need for supervision in 2023.
Good chance to focus on smaller-scale innovation, backlog of PRs and feature requests.
Review of 2022 PoW, what's done and what's carried over to 2023
Core I/O
- MT and TBufferFile: fix writing from multiple threads scaling.
- TBufferFile > 1 GB: this is getting upsetting for many people
- lossy compression: library has undefined license, we want to make this available.
Mentions of supporting shared_ptr in I/O. Difficult to implement, needs discussion, moved to offline.
RNTuple
- backfilling --> late schema extension. ATLAS and CMS need to make changes to the schema of a file while filling for certain cases.
- lossy compression: bring it on-par with TTree functionality.
- metadata API: what metadata should we store? Things that are close to the data (e.g. scale factor), but also data provenance information.
RDataFrame
- live histogram: needed to showcase true interactive use case.
- Batch processing: strong synergy with RNTuple, need to synchronise the efforts.
RooFit
- AD effort is split in two:
- C++ code generation
-
Interfaces to ROOT minimizer
-
Consolidate JSON: deliver with ATLAS, analysis wants to publish their likelihood in this format. Motivation for JSON format (not ROOT file): theorists can reinterpret, the likelihoods can be reused, provides more confidence in conveners to make this publications
- Pythonic interface to write RooFit models with JSON dictionaries. RooWorkSpace in terms of dictionaries is better than the current string-based approach.
- More benchmarks for fitting tools in HEP (zfit, pyhf).
Math
- benchmarks for minuit against scipy.
Discussion regarding benchmarks against other tools It is important that we publish these studies the first time they're done. But there should also be a repository to keep track of this type of benchmarks (different from example from the rootbench benchmarks). Then, it would be nice to integrate this as part of our CI or the release procedure, to keep track of the performance over time.
Interpreters
- status of numba support in cppyy: currently at a stop, project proposal should be written soon. Progress stopped also upstream, lack of funding and unclear use cases.
Build
- pip install ROOT: important for users, developers, packagers of the Python ecosystem. At the same time, it will surely be a burden on the team, so we must be careful.
- superbuilds: e.g. ROOT xrootd plugin built on top of existing ROOT build because you want to build a new xrootd version.
Summer student proposals
Prepare at https://docs.google.com/document/d/1wW43LQDz-iIffiRAz6-ql0ogO5ovlB5ldrU4HDABt_Y/edit?pli=1#heading=h.7oshzctzr4o5
Final discussion next monday
CHEP 2023
Need to see who is able to attend, expected 6-7 slots
6.28 release schedule
Finish all work by friday this week, take next week to fix build issues. Release soon afterwards.