ROOT I/O Meeting

Friday 7 Dec 2018, 16:00 → 17:40 Europe/Zurich

32/1-A24 (CERN)

Brian Paul Bockelman (University of Nebraska-Lincoln (US)), Philippe Canal (Fermi National Accelerator Lab. (US))

Attendees: Brian, Guilherme, Jakob, Philippe, Danilo, Oksana, Axel.

Axel: Now have a yellow master and flooding jenkins with PRs.  Hoping to tag v6.16 by the end of next week.

Brian:  Traveling week so RAS.  Will be working on bulk I/O.

Danilo: Discover a deficiency in TDirectoryFile::GetObjectChecked (TDirectoryFile::GetDirectory actually) and after Philippe fixed the problem, I added test for it.

Guilherme:  Not much I/O related things.  Working on warnings removal and removal of deprecated features.  Have submitted the CHEP paper for the buffer merger talk.

Jakob:  I have improve the RDataFrame sqlite linking to Davix and I am working on fixing it ‘properly’ by introducing RRawFile.

Discussion: should RRawFile throw exception or return error code.

Philippe:  Exceptions harder to deal with and requires exception safe calling code.

Axel: Users never check error code so throwing is much nicer to the users.

Brian: CMS migrated to use exception in order to increase the amount of information.

Brian: Will send example of use.

Jakob: Should we have a common base class for exception.

Axel/Brian: yes …. we need to have an ‘exception framework’.

Axel: so for the moment RRawFile should send ‘experimental exceptions’.

Oksana: Converged on a solution for the new enums.  It is now merged in master and v6.16.  Write the slides the I/O plan of work.   I have submitted an abstract to ACAT about compression algorithm.

Zhe: Working on benchmarking RDataSource with bulk I/O.  Adding different type of variable to the benchmark.   So far I am using JITted DataFrame.  I want to move to templates it is failing due to a deficiency in bulk I/O I am looking into it.

Philippe: RooFit schema evolution problem seems related to some type parsing error somewhere in Core/Meta.

 

 

16:00 → 16:20 Round Table