Re: Proof

On May 14, 2010, at 7:26 PM, Marc Escalier <escalier_at_lal.in2p3.fr> wrote:

> Hi rooters,
>
> may i ask some questions on Proof (Parallel root facility) ?
> (any idea on any of the problems below is welcome. thank you
> just in case, i rank them from the most important to the less
> important)
>
> *Would you know how to get back an histogram that was filled in the
> class selector created by the MakeClass() to the main program ?
> for example, i do :
> {
> #include "myselector.C"
>
> TProof *p=TProof::Open("ccapl0001.in2p3.Fr","workers=50");
> TDSet *myset=new TDSet("TTree","PAUReco");
> myset->Add("root://ccxrasn001:1094//xrootd/users/p/PAU/MonteCarlo/
> GamJet7_108085_r1215/TestLAL/PAU-00-01-60/group10.phys-sm.mc09_7TeV.
> 00105805.
> filtered_minbias6.
> e530_s765_s767_r1205_r1210_FromFilteredAOD_15_6_8_4_v0_NoGRL_NTUP-
> PAU-00-01-60_15_6_8_4_v0.AANT._00001.root");
> ...
> myset->Process("myselector.C");
> //Which fills a histogram : hist_pt_photon
> }
>
> then when i type in the session (or directly to the code) :
> "myset" : i don't have the object anymore
> myset->Print();
> whereas this instruction works if above the Process command (but i
> need after of course if i need to access the histogram that was
> filled by the Process() macro)
> ==>
> Error: Symbol myset is not defined in current scope MinimumProof.C:
> 141:
> Error: Failed to evaluate myset->Print()
> *** Interpreter error recovered ***
>
> so i can't get the histogram since the instance of TDSet no more
> exists after the Process() instruction
> (i wanted to use something like myset->GetOutputList()

The output list is owned by the TProofPlayer and can be accessed via TProof, so you have to use gProof->GetOutputList().

>
> *TDSet : did you succeed to use "jokers" (*) : "MyPath/*root*" ?
> it seems not to work while it was working fine with TChain stuff
>
> so i need to add each file one by one : a bit borying (of course i
> can do a script)
> TDSet *myset=new TDSet("TTree","PAUReco");
> myset->Add("root://ccxrasn001:1094//xrootd/users/p/PAU/MonteCarlo/
> GamJet7_108085_r1215/TestLAL/PAU-00-01-60/group10.phys-sm.mc09_7TeV.
> 00105805.
> filtered_minbias6.
> e530_s765_s767_r1205_r1210_FromFilteredAOD_15_6_8_4_v0_NoGRL_NTUP-
> PAU-00-01-60_15_6_8_4_v0.AANT._00001.root")

Well, Proof was not developed to process local files, so a TDSet constructor from files in local directory was not the first thing to do. And this is the first request we got. A script allows all the flexibility needed to adapt to the various mass storages. This said, you should probably consider using the dataset functionality ( see the web pages) which allows to register and process datasets by name, caching several info to speed up the pre-processing phase.

>
> *bonus question
> in the Terminate() method, if i don't do :
> hist_pt_photon = dynamic_cast<TH1F*>(fOutput->FindObject
> ("hist_pt_photon"));
> the histogram can't be access in the method : would you know why ?

Because the objects are created and filled in a different process than the local client and are returned via the output list. For members of the selector it may be that in the future they automatically connected. But we do not have a time scale yet for that.

G. Ganis
>
Received on Sun May 16 2010 - 09:09:33 CEST