Versions Compared

Old Version 1

changes.mady.by.user Mikhail Dubrovin

Saved on

compared with

New Version Current

changes.mady.by.user Mikhail Dubrovin

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Migration of unmigrated content due to installation of a new plugin

Abstract

We discuss a PSTuple package, which is intended for creation, accumulation and storage of histograms and tuples in psana PSHist - a histogramming package for PSANA project.

Objectives

In psana project PSANA framework we need in package which allows to accumulate data in form of histograms and tuples and save them in file for further analysis. Though it might be based on well-known underlying packages like ROOT, HBOOK, ROOT, or HippoTuple, we prefer to use an uniform interface with abstract base class, substituting the direct interaction with underlying methods. This intermediate abstract layer provides a flexibility in implementation of new things, for example multithreding in analysis, which algorithm is not yet defined.

Interface

An example of the program interface for this histogram-tuple-management package can be presented as follows. First, the underlying (derived) package needs to be choisen and instantiated. For example for ROOT

Code Block
      StoreManagerHManager *sManagerhMan = new RootStoreManagerRootHist("my-data-file.root"); // for ROOT

Then later in the program we may create ntuple(s) without knowing what is the underlying histogram-tuple-storage management system

Code Block
      Tuple *nt = sManagerhMan->ntuple("EXP Data");

and histograms

Code Block
     H1 *pHis1i = hMan->his1i("His1 int    title",100,0.,1.);
     H1 *pHis1pHis1f = sManagerhMan->his1>his1f("His1 float  Titletitle",100,0.,1.);
     H1 *pHis1d = hMan->his1d("His1 double title",100,0.,1.);
     H2 *pHis2pHis2d = sManagerhMan->his2>his2d("His2 double Titletitle",100,0.,1.,100,0.,1.);

When the ntuple is created, its parameters need to be defined once,

Code Block
      TuplePar *pBeamEnergy  = nt->parameter("beamEnergy");  // minValue, maxValue, dtype ?
      TuplePar *pBeamCurrent = nt->parameter("beamCurrent");

Then, in data processing stage, for example for each event, we may accumulate data in created histogram and ntuple objects ,using pointers

Code Block
      pBeamEnergy ->fill(E);
      pBeamCurrent->fill(I);
      nt          ->addRow();

or parameter names

Code Block

      nt         ->fill("beamEnergy", E);
      nt         ->fill("beamCurrent", I);
      nt         ->addRow();

Histograms can be accumulated using their pointers

Code Block

      pHis1ipHis1      ->fill(x,[weight]);
      ...
      pHis2      ->fill(x,y,[weight]);

where we assume thap that all parameters like x, y, E, and I and optional weight were earlier defined.

When you are all done, write the data into a file:

Code Block
      sManagerhMan->write();
      delete sManagerhMan;

Structure and content of the package

Package PSTuple PSHist contains the base abstract class and methods, which have to be re-implemented in derived classes for HBOOK, ROOT, HippoTuple, etc., i.e. in packages HBookTuplelike HBookHist, RootTupleRootHist, HippoTuple, etc.HippoHist, etc.

QUESTIONS

  • What kind of interface is preferable for tuple parameters: (1) name-based, (2) pointer-based, (3) index-based?
  • Do we really have to care about different types of data for histograms and tuple parameters (int, float, double)?
  • Arrays in tuples?
  • What should be the right mapping between PSHist and Root classes?
    Code Block
    
HManager  <-> TFile
H1        <-> TH1D
Tuple     <-> TTree or TBranch
Parameter <-> TBranch or TBranch-item

APPENDIX

Histogramming in Root

Code Block

// Base Class Headers --
#include "root/TROOT.h"
#include "root/TFile.h"
#include "root/TH1D.h"
#include "root/TTree.h"
#include "root/TBranch.h"
#include "root/TRandom.h"

Histograms

Code Block

// Initialization
if(!TROOT::Initialized()) {static TROOT root( "RootManager", "RootManager ROOT God" );}

// Open output file
TFile* pfile = new TFile("file.root", "RECREATE", "Created for you by RootManager" );

// Create histograms
TH1D *pHis1 = new TH1D("pHis1","My comment to TH1D", 100, 0.5, 100+0.5);
TH2D *pHis2 = new TH2D("pHis2","My comment to TH1D", 100, 0.5, 100+0.5,
                                                     200, 0.5, 200+0.5);

// Fill histograms in each event
pHis1 -> Fill( value, [weight] );
pHis2 -> Fill( x, y, [weight] );

// Write histograms in file in the very end
pHis1 -> Write();
pHis2 -> Write();

//Close file
pfile -> Close();

NTuple

Assuming that file is already open,

Code Block

// Define some structures for TTree
typedef struct {float x,y,z;} POINT;
static POINT point;
float new_v;

// Create TTree
TTree* ptree = new TTree("ptree", "My comment to TTree");

// Create branches
TBranch *pbranch = ptree->Branch("new_v", &new_v, "new_v/F");
                   ptree->Branch("point",&point,"x:y:z");

// Fill data structures for each event
      new_v   = gRandom->Gaus(0, 1);
      point.x = gRandom->Gaus(1, 1);
      point.y = gRandom->Gaus(2, 1);
      point.z = gRandom->Gaus(3, 1);

// Add an event record to the tree
      ptree->Fill();

// Write the tree to the file
ptree -> Write();

Histogramming in BABAR

All histograms in Workbook examples are based on plane Root...

Histograms

Code Block

     HepHistogram myPlot("The Title",100,0.,1.);
     float x, weight;
       ...
     myPlot.accumulate( x, weight );

NTuple

Code Block

#include "HepTuple/Histogram.h"
#include "HepTuple/TupleManager.h"

    HepTupleManager* manager = gblEnv->getGen()->ntupleManager();

    HepTuple *_ntuple = manager->ntuple("file-name.root");

   _ntuple->column("run",   eventID->run(), -99, "Event" );
   _ntuple->column("event", eventCounter,   -99, "Event" );

   _ntuple->dumpData(); // for each event

Histogramming in CLEO

Histograms

Declare Histogram in your .h file

Code Block

#include "HistogramInterface/HistogramPackage.h’’
...
HIHist1D *m_trackMom;

Define Histogram in your Processor

Code Block

m_trackMom = iHistoManager.histogram("track momentum (GeV)",50, 0, 1.5);

Fill Histogram in event() method of your Processor

Code Block

m_trackMom->fill(momentum);

Load Conversion Module in .tcl script

Code Block

module sel RootHistogramModule
root file root_suez_style_example_data31.root
root init

NTuple

Declare Ntuple in your Processor’s .h file

Code Block

#include "HistogramInterface/HistogramPackage.h’’
#include ‘‘HistogramInterface/HINtupleVarNames.h’’
...
HINtuple *m_trackTuple;

Make an enum of variable names

Code Block

enum{kpx, kpy, kpz, kd0, kz0, knum_vars};

Define Ntuple in hist_book()
// make an object that holds the names of the variables in your ntuple

Code Block

HINtupleVarNames knames(knum_vars);
knames.addVar(kpx,"px");
knames.addVar(kpy,"py");
knames.addVar(kpz,"pz");
knames.addVar(kd0,"d0");
knames.addVar(kz0,"z0");
m_trackTuple = iHistoManager.ntuple(1,"tracktuple",knum_vars,50000,knames.names());

Fill Ntuple in event() method

Code Block

trackTuple[kpx] = px;
trackTuple[kpy] = py;
trackTuple[kpz] = pz;
trackTuple[kd0] = d0;
trackTuple[kz0] = z0;
m_trackTuple->fill(trackTuple);