source: git/modules/VertexFinder.cc@ e01b141

/** \class VertexFinder
 *
 *  Cluster vertices from tracks
 *
 *  \authors A. Hart, M. Selvaggi
 *
 */

#include "modules/VertexFinder.h"
#include "classes/DelphesClasses.h"
#include "classes/DelphesFactory.h"
#include "classes/DelphesFormula.h"
#include "classes/DelphesPileUpReader.h"

#include "ExRootAnalysis/ExRootClassifier.h"
#include "ExRootAnalysis/ExRootFilter.h"
#include "ExRootAnalysis/ExRootResult.h"

#include "TDatabasePDG.h"
#include "TFormula.h"
#include "TLorentzVector.h"
#include "TMath.h"
#include "TMatrixT.h"
#include "TObjArray.h"
#include "TRandom3.h"
#include "TString.h"
#include "TVector3.h"

#include <algorithm>
#include <iostream>
#include <map>
#include <stdexcept>
#include <string>
#include <utility>
#include <vector>

using namespace std;

static const Double_t mm = 1.;
static const Double_t m = 1000. * mm;
static const Double_t ns = 1.;
static const Double_t s = 1.e+9 * ns;
static const Double_t c_light = 2.99792458e+8 * m / s;

//------------------------------------------------------------------------------

VertexFinder::VertexFinder() :
  fSigma(0), fMinPT(0), fMaxEta(0), fSeedMinPT(0), fMinNDF(0), fGrowSeeds(0)
{
}

//------------------------------------------------------------------------------

VertexFinder::~VertexFinder()
{
}

//------------------------------------------------------------------------------

void VertexFinder::Init()
{
  fSigma = GetDouble("Sigma", 3.0);
  fMinPT = GetDouble("MinPT", 0.1);
  fMaxEta = GetDouble("MaxEta", 10.0);
  fSeedMinPT = GetDouble("SeedMinPT", 5.0);
  fMinNDF = GetInt("MinNDF", 4);
  fGrowSeeds = GetInt("GrowSeeds", 1);

  fInputArray = ImportArray(GetString("InputArray", "TrackSmearing/tracks"));
  fItInputArray = fInputArray->MakeIterator();

  fOutputArray = ExportArray(GetString("OutputArray", "tracks"));
  fVertexOutputArray = ExportArray(GetString("VertexOutputArray", "vertices"));
}

//------------------------------------------------------------------------------

void VertexFinder::Finish()
{
  if(fItInputArray) delete fItInputArray;
}

//------------------------------------------------------------------------------

static Bool_t secondAscending(pair<UInt_t, Double_t> pair0, pair<UInt_t, Double_t> pair1)
{
  return (pair0.second < pair1.second);
}

static Bool_t secondDescending(pair<UInt_t, Double_t> pair0, pair<UInt_t, Double_t> pair1)
{
  return (pair0.second > pair1.second);
}

//------------------------------------------------------------------------------

void VertexFinder::Process()
{
  Candidate *candidate;

  // Clear the track and cluster maps before starting
  trackIDToDouble.clear();
  trackIDToInt.clear();
  trackIDToBool.clear();
  clusterIDToDouble.clear();
  clusterIDToInt.clear();
  clusterIDToBool.clear();
  trackPT.clear();
  clusterSumPT2.clear();

  // Create the initial cluster seeds
  createSeeds();

  // In order of descending seed pt, grow each cluster. If a cluster ends up with
  // fewer than MinNDF tracks, release the tracks for other clusters to claim.
  sort(clusterSumPT2.begin(), clusterSumPT2.end(), secondDescending);
  for(vector<pair<UInt_t, Double_t> >::const_iterator cluster = clusterSumPT2.begin(); cluster != clusterSumPT2.end(); cluster++)
  {
    // Skip the cluster if it no longer has any tracks
    if(!clusterIDToInt.at(cluster->first).at("ndf"))
      continue;

    // Grow the cluster if GrowSeeds is true
    if(fGrowSeeds)
      growCluster(cluster->first);

    // If the cluster still has fewer than MinNDF tracks, release the tracks;
    // otherwise, mark the seed track as claimed

    if((Int_t)clusterIDToInt.at(cluster->first).at("ndf") < fMinNDF)
    {
      for(map<UInt_t, map<string, Int_t> >::iterator track = trackIDToInt.begin(); track != trackIDToInt.end(); track++)
      {
        if(track->second.at("clusterIndex") != (Int_t)cluster->first)
          continue;
        track->second["clusterIndex"] = -1;
        trackIDToBool[track->first]["claimed"] = false;
      }
    }
    else
      trackIDToBool[clusterIDToInt.at(cluster->first).at("seed")]["claimed"] = true;
  }

  // Add tracks to the output array after updating their ClusterIndex.
  fItInputArray->Reset();
  while((candidate = static_cast<Candidate *>(fItInputArray->Next())))
  {
    if(candidate->Momentum.Pt() < fMinPT || fabs(candidate->Momentum.Eta()) > fMaxEta)
      continue;
    candidate->ClusterIndex = trackIDToInt.at(candidate->GetUniqueID()).at("clusterIndex");
    fOutputArray->Add(candidate);
  }

  // Add clusters with at least MinNDF tracks to the output array in order of
  // descending sum(pt**2).
  clusterSumPT2.clear();
  for(map<UInt_t, map<string, Int_t> >::const_iterator cluster = clusterIDToInt.begin(); cluster != clusterIDToInt.end(); cluster++)
  {

    if(cluster->second.at("ndf") < fMinNDF)
      continue;
    clusterSumPT2.push_back(make_pair(cluster->first, clusterIDToDouble.at(cluster->first).at("sumPT2")));
  }
  sort(clusterSumPT2.begin(), clusterSumPT2.end(), secondDescending);

  for(vector<pair<UInt_t, Double_t> >::const_iterator cluster = clusterSumPT2.begin(); cluster != clusterSumPT2.end(); cluster++)
  {
    DelphesFactory *factory = GetFactory();
    candidate = factory->NewCandidate();

    candidate->ClusterIndex = cluster->first;
    candidate->ClusterNDF = clusterIDToInt.at(cluster->first).at("ndf");
    candidate->ClusterSigma = fSigma;
    candidate->SumPT2 = cluster->second;
    candidate->Position.SetXYZT(0.0, 0.0, clusterIDToDouble.at(cluster->first).at("z"), 0.0);
    candidate->PositionError.SetXYZT(0.0, 0.0, clusterIDToDouble.at(cluster->first).at("ez"), 0.0);

    fVertexOutputArray->Add(candidate);
  }
}

//------------------------------------------------------------------------------

void VertexFinder::createSeeds()
{
  Candidate *candidate;
  UInt_t clusterIndex = 0, maxSeeds = 0;

  // Loop over all tracks, initializing some variables.
  fItInputArray->Reset();
  while((candidate = static_cast<Candidate *>(fItInputArray->Next())))
  {
    if(candidate->Momentum.Pt() < fMinPT || fabs(candidate->Momentum.Eta()) > fMaxEta)
      continue;

    trackIDToDouble[candidate->GetUniqueID()]["pt"] = candidate->Momentum.Pt();
    trackIDToDouble[candidate->GetUniqueID()]["ept"] = candidate->ErrorPT ? candidate->ErrorPT : 1.0e-15;
    ;
    trackIDToDouble[candidate->GetUniqueID()]["eta"] = candidate->Momentum.Eta();

    trackIDToDouble[candidate->GetUniqueID()]["z"] = candidate->DZ;
    trackIDToDouble[candidate->GetUniqueID()]["ez"] = candidate->ErrorDZ ? candidate->ErrorDZ : 1.0e-15;

    trackIDToInt[candidate->GetUniqueID()]["clusterIndex"] = -1;
    trackIDToInt[candidate->GetUniqueID()]["interactionIndex"] = candidate->IsPU;

    trackIDToBool[candidate->GetUniqueID()]["claimed"] = false;

    trackPT.push_back(make_pair(candidate->GetUniqueID(), candidate->Momentum.Pt()));
  }

  // Sort tracks by pt and leave only the SeedMinPT highest pt ones in the
  // trackPT vector.
  sort(trackPT.begin(), trackPT.end(), secondDescending);
  for(vector<pair<UInt_t, Double_t> >::const_iterator track = trackPT.begin(); track != trackPT.end(); track++, maxSeeds++)
  {
    if(track->second < fSeedMinPT)
      break;
  }
  // If there are no tracks with pt above MinSeedPT, create just one seed from
  // the highest pt track.
  if(!maxSeeds)
    maxSeeds++;
  if(trackPT.size() > maxSeeds)
  {
    trackPT.erase(trackPT.begin() + maxSeeds, trackPT.end());
  }

  // Create the seeds from the SeedMinPT highest pt tracks.
  for(vector<pair<UInt_t, Double_t> >::const_iterator track = trackPT.begin(); track != trackPT.end(); track++, clusterIndex++)
  {
    addTrackToCluster(track->first, clusterIndex);
    clusterSumPT2.push_back(make_pair(clusterIndex, track->second * track->second));
  }
}

//------------------------------------------------------------------------------

void VertexFinder::growCluster(const UInt_t clusterIndex)
{
  Bool_t done = false;
  UInt_t nearestID;
  Int_t oldClusterIndex;
  Double_t nearestDistance;
  vector<UInt_t> nearTracks;
  nearTracks.clear();

  // Grow the cluster until there are no more tracks within Sigma standard
  // deviations of the cluster.
  while(!done)
  {
    done = true;
    nearestID = 0;
    nearestDistance = -1.0;

    // These two loops are for finding the nearest track to the cluster. The
    // first time, the ID of each track within 10*Sigma of the cluster is
    // saved in the nearTracks vector; subsequently, to save time, only the
    // tracks in this vector are checked.
    if(!nearTracks.size())
    {

      for(map<UInt_t, map<string, Double_t> >::const_iterator track = trackIDToDouble.begin(); track != trackIDToDouble.end(); track++)
      {
        if(trackIDToBool.at(track->first).at("claimed") || trackIDToInt.at(track->first).at("clusterIndex") == (Int_t)clusterIndex)
          continue;

        Double_t distance = fabs(clusterIDToDouble.at(clusterIndex).at("z") - track->second.at("z")) / hypot(clusterIDToDouble.at(clusterIndex).at("ez"), track->second.at("ez"));
        if(nearestDistance < 0.0 || distance < nearestDistance)
        {
          nearestID = track->first;
          nearestDistance = distance;
        }
        if(distance < 10.0 * fSigma)
          nearTracks.push_back(track->first);
      }
    }

    else
    {
      for(vector<UInt_t>::const_iterator track = nearTracks.begin(); track != nearTracks.end(); track++)
      {
        if(trackIDToBool.at(*track).at("claimed") || trackIDToInt.at(*track).at("clusterIndex") == (Int_t)clusterIndex)
          continue;
        Double_t distance = fabs(clusterIDToDouble.at(clusterIndex).at("z") - trackIDToDouble.at(*track).at("z")) / hypot(clusterIDToDouble.at(clusterIndex).at("ez"), trackIDToDouble.at(*track).at("ez"));
        if(nearestDistance < 0.0 || distance < nearestDistance)
        {
          nearestID = *track;
          nearestDistance = distance;
        }
      }
    }

    // If no tracks within Sigma of the cluster were found, stop growing.
    done = nearestDistance > fSigma || nearestDistance < 0.0;
    if(done)
    {
      continue;
    }

    // Add the nearest track within Sigma to the cluster. If it already
    // belonged to another cluster, remove it from that cluster first.
    if(nearestDistance < fSigma)
    {
      oldClusterIndex = trackIDToInt.at(nearestID).at("clusterIndex");
      if(oldClusterIndex >= 0)
        removeTrackFromCluster(nearestID, oldClusterIndex);

      trackIDToBool[nearestID]["claimed"] = true;
      addTrackToCluster(nearestID, clusterIndex);
    }
  }
}

//------------------------------------------------------------------------------

Double_t VertexFinder::weight(const UInt_t trackID)
{
  return ((trackIDToDouble.at(trackID).at("pt") / (trackIDToDouble.at(trackID).at("ept") * trackIDToDouble.at(trackID).at("ez"))) * (trackIDToDouble.at(trackID).at("pt") / (trackIDToDouble.at(trackID).at("ept") * trackIDToDouble.at(trackID).at("ez"))));
}

//------------------------------------------------------------------------------

void VertexFinder::removeTrackFromCluster(const UInt_t trackID, const UInt_t clusterID)
{
  Double_t wz = weight(trackID);

  trackIDToInt[trackID]["clusterIndex"] = -1;
  clusterIDToInt[clusterID]["ndf"]--;

  clusterIDToDouble[clusterID]["sumZ"] -= wz * trackIDToDouble.at(trackID).at("z");
  clusterIDToDouble[clusterID]["errorSumZ"] -= wz * trackIDToDouble.at(trackID).at("ez") * trackIDToDouble.at(trackID).at("ez");
  clusterIDToDouble[clusterID]["sumOfWeightsZ"] -= wz;
  clusterIDToDouble[clusterID]["z"] = clusterIDToDouble.at(clusterID).at("sumZ") / clusterIDToDouble.at(clusterID).at("sumOfWeightsZ");
  clusterIDToDouble[clusterID]["ez"] = sqrt((1.0 / clusterIDToInt.at(clusterID).at("ndf")) * (clusterIDToDouble.at(clusterID).at("errorSumZ") / clusterIDToDouble.at(clusterID).at("sumOfWeightsZ")));
  clusterIDToDouble[clusterID]["sumPT2"] -= trackIDToDouble.at(trackID).at("pt") * trackIDToDouble.at(trackID).at("pt");
}

//------------------------------------------------------------------------------

void VertexFinder::addTrackToCluster(const UInt_t trackID, const UInt_t clusterID)
{
  Double_t wz = weight(trackID);

  if(!clusterIDToInt.count(clusterID))
  {
    clusterIDToInt[clusterID]["ndf"] = 0;
    clusterIDToInt[clusterID]["seed"] = trackID;
    clusterIDToDouble[clusterID]["sumZ"] = 0.0;
    clusterIDToDouble[clusterID]["errorSumZ"] = 0.0;
    clusterIDToDouble[clusterID]["sumOfWeightsZ"] = 0.0;
    clusterIDToDouble[clusterID]["sumPT2"] = 0.0;
  }

  trackIDToInt[trackID]["clusterIndex"] = clusterID;
  clusterIDToInt[clusterID]["ndf"]++;

  clusterIDToDouble[clusterID]["sumZ"] += wz * trackIDToDouble.at(trackID).at("z");
  clusterIDToDouble[clusterID]["errorSumZ"] += wz * trackIDToDouble.at(trackID).at("ez") * trackIDToDouble.at(trackID).at("ez");
  clusterIDToDouble[clusterID]["sumOfWeightsZ"] += wz;
  clusterIDToDouble[clusterID]["z"] = clusterIDToDouble.at(clusterID).at("sumZ") / clusterIDToDouble.at(clusterID).at("sumOfWeightsZ");
  clusterIDToDouble[clusterID]["ez"] = sqrt((1.0 / clusterIDToInt.at(clusterID).at("ndf")) * (clusterIDToDouble.at(clusterID).at("errorSumZ") / clusterIDToDouble.at(clusterID).at("sumOfWeightsZ")));
  clusterIDToDouble[clusterID]["sumPT2"] += trackIDToDouble.at(trackID).at("pt") * trackIDToDouble.at(trackID).at("pt");
}

//------------------------------------------------------------------------------