Selector-by-object feature

ROOT
1

Hi all,

I am very much interested in the new Selector-by-object feature. I have just compiled the 5.34-rc1 version on my laptop to use it. Is there any documentation or any hints on how to use it?

Also, is this functionality offered for standalone running (ie processing a chain using a TSelector but only on one core)? Eventually I am planning to use my code on a proof cluster, but for testing and quick results I want to run on a small subset of my data first. I would be interested in using proof-lite, but I am spending too much time straightening package load issues with that (whereas loading packages on the cluster seems to work) , so I would be more willing on running standalone root with the TSelector, and in the end keepeing ownership of the TSelector object, much like I used to run with the “MakeClass” derived classes with standalone root.

I was very happy to see a feature like this developed, since the proof sessions can become truly interactive by accessing the TSelector members after the processing is completed.

Cheers,
Nik

2

Hi,
Sorry for the later reply.
I have created a new doc page at root.cern.ch/drupal/content/basic-processing summarizing the usage of TProof::Process and some information about processing by object.

Basically you can create your object and pass it to Process instead of the selector file name or the selector name.
The selector must be streamable, because that’s the way it is transferred to the workers.

Btw, this functionality is available for local TChain processing since a long time; not sure since when, but it is certainly available in all ROOT versions currently around.

I would be interested to hear about your difficulties loading code in PROOF-Lite.

G. Ganis