harvester.py

 
 
from tracking.harvest.harvesting import HarvestingModule
from tracking.harvest import refiners
 
from ROOT import Belle2
 
import numpy as np
 
 
class ReconstructionPositionHarvester(HarvestingModule):
 
    """ Harvester module to check for the reconstructed positions """
 
    def __init__(self, output_file_name, tracks_store_vector_name="CDCTrackVector"):
        """ Initiialize with the output file name of the root file and the store obj with the
        CDCTrack vector to use. MC track cands are needed. """
        super(
            ReconstructionPositionHarvester,
            self).__init__(
            foreach=tracks_store_vector_name,
            output_file_name=output_file_name)
 
        
        self.mc_hit_lookupmc_hit_lookup = Belle2.TrackFindingCDC.CDCMCHitLookUp()
 
    def peel(self, track_cand):
        """ Extract the information """
        mc_hit_lookup = self.mc_hit_lookupmc_hit_lookup
        mc_hit_lookup.fill()
 
        sum_of_difference_norms_stereo = 0
        num_norms_stereo = 0
        sum_of_difference_norms_axial = 0
        num_norms_axial = 0
 
        number_of_wrong_rl_infos = 0
 
        for reco_hit3D in list(track_cand.items()):
            underlaying_cdc_hit = reco_hit3D.getWireHit().getHit()
            hit_difference = mc_hit_lookup.getRecoPos3D(underlaying_cdc_hit) - reco_hit3D.getRecoPos3D()
            sum_of_difference_norms_axial += hit_difference.xy().norm()
            num_norms_axial += 1
 
            if reco_hit3D.getStereoType() != 0:  # AXIAL
                sum_of_difference_norms_stereo += abs(hit_difference.z())
                num_norms_stereo += 1
 
                correct_rl_info = mc_hit_lookup.getRLInfo(underlaying_cdc_hit)
 
                if correct_rl_info != reco_hit3D.getRLInfo():
                    number_of_wrong_rl_infos += 1
 
        return dict(sum_of_difference_norms_axial=sum_of_difference_norms_axial,
                    num_norms_axial=num_norms_axial,
                    mean_of_difference_norms_axial=np.true_divide(sum_of_difference_norms_axial, num_norms_axial),
                    sum_of_difference_norms_stereo=sum_of_difference_norms_stereo,
                    num_norms_stereo=num_norms_stereo,
                    mean_of_difference_norms_stereo=np.true_divide(sum_of_difference_norms_stereo, num_norms_stereo),
                    number_of_wrong_rl_infos=number_of_wrong_rl_infos,
                    mean_wrong_rl_infos=np.true_divide(number_of_wrong_rl_infos, num_norms_stereo))
 
    
    save_tree = refiners.save_tree(
        
        folder_name="tree"
        
    )
 
 
class WrongRLInfoCounter(HarvestingModule):
 
    """
    Harvester module to check for the correct reconstructed orientation.
    """
 
    def __init__(self, output_file_name, tracks_store_vector_name="TrackCands"):
        """
        Initialize with the root output file name and the store array name of the track cands to use.
        """
        HarvestingModule.__init__(self,
                                  foreach=tracks_store_vector_name,
                                  output_file_name=output_file_name)
 
        
        self.mc_hit_lookupmc_hit_lookup = Belle2.TrackFindingCDC.CDCMCHitLookUp()
 
    def peel(self, track_cand):
        """ Extract the information. """
        cdc_hits = Belle2.PyStoreArray("CDCHits")
 
        mc_hit_lookup = self.mc_hit_lookupmc_hit_lookup
        mc_hit_lookup.fill()
 
        number_of_wrong_rl_infos = 0
        number_of_wrong_rl_infos_stereo_only = 0
 
        number_of_hits = 0
        number_of_stereo = 0
 
        for hitID in range(track_cand.getNHits()):
            hit = track_cand.getHit(hitID)
            if not hit.__class__.__name__ == "genfit::WireTrackCandHit":
                continue
 
            underlaying_cdc_hit = cdc_hits[hit.getHitId()]
 
            correct_rl_info = mc_hit_lookup.getRLInfo(underlaying_cdc_hit)
 
            number_of_hits += 1
 
            if underlaying_cdc_hit.getISuperLayer() % 2 != 0:
                number_of_stereo += 1
 
            if correct_rl_info != hit.getLeftRightResolution():
                if underlaying_cdc_hit.getISuperLayer() % 2 != 0:
                    number_of_wrong_rl_infos_stereo_only += 1
                number_of_wrong_rl_infos += 1
 
        return dict(number_of_wrong_rl_infos=number_of_wrong_rl_infos,
                    number_of_wrong_rl_infos_stereo_only=number_of_wrong_rl_infos_stereo_only,
                    number_of_hits=number_of_hits,
                    number_of_stereo=number_of_stereo,
                    mean_wrong_rl=np.true_divide(number_of_wrong_rl_infos, number_of_hits),
                    mean_wrong_rl_stereo=np.true_divide(number_of_wrong_rl_infos_stereo_only, number_of_stereo))
 
    
    save_tree = refiners.save_tree(
        
        folder_name="tree"
        
    )
 
 
class SegmentFakeRatesModule(HarvestingModule):
 
    """
    Harvesting module to check for basic matching information of the found segments.
    If you want to have matching information, please use the MC matcher module to make all possible
    combinations between legendre, local and MC track candidates.
    Also, the Segments must be transformed into GF track cands.
    """
 
    def __init__(
            self,
            output_file_name,
            local_track_cands_store_array_name="LocalTrackCands",
            mc_track_cands_store_array_name="MCTrackCands",
            legendre_track_cand_store_array_name="LegendreTrackCands"):
        """
        Initialize the harvesting module with
        Arguments
        ---------
          output_file_name Name of the root file
          local_track_cands_store_array_name Name of the StoreArray for local track cands
          mc_track_cands_store_array_name Name of the StoreArray for MC track cands
          legendre_track_cand_store_array_name Name of the StoreArray for legendre track cands
        """
        super(
            SegmentFakeRatesModule,
            self).__init__(
            foreach=local_track_cands_store_array_name,
            output_file_name=output_file_name)
 
        
        self.mc_track_cands_store_array_namemc_track_cands_store_array_name = mc_track_cands_store_array_name
        
        self.legendre_track_cand_store_array_namelegendre_track_cand_store_array_name = legendre_track_cand_store_array_name
 
        
        self.mc_track_matcher_localmc_track_matcher_local = Belle2.TrackMatchLookUp(self.mc_track_cands_store_array_namemc_track_cands_store_array_name, self.foreachforeach)
        
        self.mc_track_matcher_legendremc_track_matcher_legendre = Belle2.TrackMatchLookUp(
            self.mc_track_cands_store_array_namemc_track_cands_store_array_name,
            legendre_track_cand_store_array_name)
 
    def prepare(self):
        """ Prepare the CDC lookup. """
        
        self.cdcHitscdcHits = Belle2.PyStoreArray("CDCHits")
        return HarvestingModule.prepare(self)
 
    def peel(self, local_track_cand):
        """ Extract the information from the segments. """
        mc_track_matcher_local = self.mc_track_matcher_localmc_track_matcher_local
        mc_track_matcher_legendre = self.mc_track_matcher_legendremc_track_matcher_legendre
 
        cdc_hits = self.cdcHitscdcHits
 
        is_background = mc_track_matcher_local.isBackgroundPRRecoTrack(local_track_cand)
        is_ghost = mc_track_matcher_local.isGhostPRRecoTrack(local_track_cand)
        is_matched = mc_track_matcher_local.isAnyChargeMatchedPRRecoTrack(local_track_cand)
        is_clone = mc_track_matcher_local.isAnyChargeClonePRRecoTrack(local_track_cand)
        is_clone_or_matched = is_matched or is_clone
        hit_purity = abs(mc_track_matcher_local.getRelatedPurity(local_track_cand))
 
        # Stereo Track?
        first_cdc_hit_id = local_track_cand.getHitIDs(Belle2.Const.CDC)[0]
        first_cdc_hit = cdc_hits[first_cdc_hit_id]
        is_stereo = first_cdc_hit.getISuperLayer() % 2 == 1
        superlayer_of_segment = first_cdc_hit.getISuperLayer()
 
        has_partner = np.NaN
        hit_purity_of_partner = np.NaN
        hit_efficiency_of_partner = np.NaN
        mc_track_pt = np.NaN
        mc_track_dist = np.NaN
        number_of_new_hits = local_track_cand.getNHits()
        number_of_hits = local_track_cand.getNHits()
        number_of_hits_in_same_superlayer = np.NaN
        partner_has_stereo_information = np.NaN
 
        if is_clone_or_matched:
            related_mc_track_cand = mc_track_matcher_local.getRelatedMCRecoTrack(local_track_cand)
            has_partner = (mc_track_matcher_legendre.isAnyChargeMatchedMCRecoTrack(related_mc_track_cand) or
                           mc_track_matcher_legendre.isAnyChargeMergedMCRecoTrack(related_mc_track_cand))
            mc_track_pt = related_mc_track_cand.getMomSeed().Pt()
            mc_track_dist = related_mc_track_cand.getPosSeed().Mag()
            if has_partner:
                partner_legendre_track_cand = mc_track_matcher_legendre.getRelatedPRRecoTrack(related_mc_track_cand)
                hit_purity_of_partner = abs(mc_track_matcher_legendre.getRelatedPurity(partner_legendre_track_cand))
                hit_efficiency_of_partner = abs(mc_track_matcher_legendre.getRelatedEfficiency(related_mc_track_cand))
 
                # Count number of new hits
                legendre_hits = set(list(partner_legendre_track_cand.getHitIDs()))
                local_hits = set(list(local_track_cand.getHitIDs()))
 
                local_hits_new = local_hits - legendre_hits
                number_of_new_hits = len(local_hits_new)
 
                # Count number of hits in this superlayer
                partner_has_stereo_information = 0
                number_of_hits_in_same_superlayer = 0
                for cdc_hit_ID in legendre_hits:
                    cdc_hit = cdc_hits[cdc_hit_ID]
                    if cdc_hit.getISuperLayer() == superlayer_of_segment:
                        number_of_hits_in_same_superlayer += 1
 
                    if cdc_hit.getISuperLayer() % 2 == 1:
                        partner_has_stereo_information = 1
 
        return dict(superlayer_of_segment=superlayer_of_segment,
                    mc_track_dist=mc_track_dist,
                    is_background=is_background,
                    is_ghost=is_ghost,
                    is_matched=is_matched,
                    is_clone=is_clone,
                    is_clone_or_matched=is_clone_or_matched,
                    is_stereo=is_stereo,
                    mc_track_pt=mc_track_pt,
                    hit_purity=hit_purity,
                    has_partner=has_partner,
                    hit_purity_of_partner=hit_purity_of_partner,
                    hit_efficiency_of_partner=hit_efficiency_of_partner,
                    number_of_new_hits=number_of_new_hits,
                    number_of_hits=number_of_hits,
                    number_of_hits_in_same_superlayer=number_of_hits_in_same_superlayer,
                    partner_has_stereo_information=partner_has_stereo_information)
 
    
    save_tree = refiners.save_tree(
        
        folder_name="tree"
        
    )
 
 
class SegmentFinderParameterExtractorModule(HarvestingModule):
 
    """
    Harvester module to extract momentum and position information from the found segments and tracks.
    You need the apply MC track matcher module.
    """
 
    def __init__(self, local_track_cands_store_array_name, mc_track_cands_store_array_name, output_file_name):
        """ Init the harvester with the local track candidates StoreArray name and the one for MC track cands
        and the output file name for the result root file.
        """
        super(
            SegmentFinderParameterExtractorModule,
            self).__init__(
            foreach=local_track_cands_store_array_name,
            output_file_name=output_file_name)
 
        
        self.mc_track_cands_store_array_namemc_track_cands_store_array_name = mc_track_cands_store_array_name
 
        
        self.mc_track_matchermc_track_matcher = Belle2.TrackMatchLookUp(self.mc_track_cands_store_array_namemc_track_cands_store_array_name,
                                                        self.foreachforeach)
 
    def peel(self, local_track_cand):
        """ Extract the information from the local track candidate. """
        mc_track_matcher = self.mc_track_matchermc_track_matcher
 
        is_matched = mc_track_matcher.isAnyChargeMatchedPRRecoTrack(local_track_cand)
        is_background = mc_track_matcher.isBackgroundPRRecoTrack(local_track_cand)
        is_ghost = mc_track_matcher.isGhostPRRecoTrack(local_track_cand)
 
        related_mc_track_cand = mc_track_matcher.getRelatedMCRecoTrack(local_track_cand)
 
        track_momentum = np.NaN
        # track_position = np.NaN
        track_pt = np.NaN
        track_phi = np.NaN
 
        segment_momentum = np.NaN
        # segment_position = np.NaN
        segment_pt = np.NaN
        segment_phi = np.NaN
 
        if is_matched:
            track_momentum = related_mc_track_cand.getPosSeed()
            # track_position = related_mc_track_cand.getPosSeed()
            track_pt = track_momentum.Pt()
            track_phi = track_momentum.Phi()
            # trajectory_track = \
            #     Belle2.TrackFindingCDC.CDCTrajectory3D(Belle2.TrackFindingCDC.Vector3D(track_position),
            #                                            Belle2.TrackFindingCDC.Vector3D(track_momentum),
            #                                            related_mc_track_cand.getChargeSeed())
 
            segment_momentum = local_track_cand.getMomSeed()
            # segment_position = local_track_cand.getPosSeed()
            segment_pt = segment_momentum.Pt()
            segment_phi = segment_momentum.Phi()
            # trajectory_segment = \
            #     Belle2.TrackFindingCDC.CDCTrajectory3D(Belle2.TrackFindingCDC.Vector3D(segment_position),
            #                                            Belle2.TrackFindingCDC.Vector3D(segment_momentum),
            #                                            local_track_cand.getChargeSeed())
 
        return dict(is_matched=is_matched,
                    is_background=is_background,
                    is_ghost=is_ghost,
                    segment_pt=segment_pt,
                    track_pt=track_pt,
                    difference_pt=segment_pt - track_pt,
                    segment_phi=segment_phi,
                    track_phi=track_phi,
                    difference_phi=segment_phi - track_phi)
 
    
    save_tree = refiners.save_tree(
        
        folder_name="tree"
        
    )
 
 
class SeedsAnalyser(HarvestingModule):
 
    """
    Extract the momentum seeds of the first and the last hit of the track candidates (when use_vxd_hits is true),
    the helx and the mc momentum.
    """
 
    def __init__(self, output_file_name, track_cands, use_vxd_hits=True):
        """
        Init with the root output file name and the name of the store array of the track cands to use.
        """
        HarvestingModule.__init__(self, foreach=track_cands, output_file_name=output_file_name)
 
        
        self.use_vxd_hitsuse_vxd_hits = use_vxd_hits
 
    def peel(self, legendre_track_cand):
        """
        Extract the information.
        """
        if legendre_track_cand.getChargeSeed() > 0:
            helix = Belle2.Helix(legendre_track_cand.getPosSeed(), legendre_track_cand.getMomSeed(), 1, 1.5)
        if legendre_track_cand.getChargeSeed() < 0:
            helix = Belle2.Helix(legendre_track_cand.getPosSeed(), legendre_track_cand.getMomSeed(), -1, 1.5)
 
        matcher = Belle2.TrackMatchLookUp("MCTrackCands", self.foreachforeach)
        mc_track_cand = matcher.getAnyChargeMatchedMCRecoTrack(legendre_track_cand)
 
        pxd_clusters = Belle2.PyStoreArray("PXDClusters")
        svd_clusters = Belle2.PyStoreArray("SVDClusters")
 
        if mc_track_cand:
            mc_x = mc_track_cand.getPosSeed().X()
            mc_y = mc_track_cand.getPosSeed().Y()
            mc_z = mc_track_cand.getPosSeed().Z()
            mc_mom_x = mc_track_cand.getMomSeed().X()
            mc_mom_y = mc_track_cand.getMomSeed().Y()
            mc_mom_z = mc_track_cand.getMomSeed().Z()
        else:
            mc_x = np.NaN
            mc_y = np.NaN
            mc_z = np.NaN
            mc_mom_x = np.NaN
            mc_mom_y = np.NaN
            mc_mom_z = np.NaN
 
        return_dict = dict(
            helix_z=helix.getZ0(),
            helix_x=helix.getPerigeeX(),
            helix_y=helix.getPerigeeY(),
            helix_mom_z=helix.getMomentumZ(1.5),
            helix_mom_x=helix.getMomentumX(1.5),
            helix_mom_y=helix.getMomentumY(1.5),
            mc_x=mc_x,
            mc_y=mc_y,
            mc_z=mc_z,
            mc_mom_x=mc_mom_x,
            mc_mom_y=mc_mom_y,
            mc_mom_z=mc_mom_z)
 
        if self.use_vxd_hitsuse_vxd_hits:
            first_hit = legendre_track_cand.getHit(0)
            if first_hit.getDetId() == Belle2.Const.PXD:
                first_cluster = pxd_clusters[first_hit.getHitId()]
                first_true_hit = first_cluster.getRelated("PXDTrueHits")
            elif first_hit.getDetId() == Belle2.Const.SVD:
                first_cluster = svd_clusters[first_hit.getHitId()]
                first_true_hit = first_cluster.getRelated("SVDTrueHits")
 
            vxdID = first_cluster.getSensorID()
            sensorInfoBase = Belle2.VXD.GeoCache.getInstance().getSensorInfo(vxdID)
            first_momentum = sensorInfoBase.vectorToGlobal(first_true_hit.getMomentum(), True)
 
            last_hit = legendre_track_cand.getHit(-1)
            if last_hit.getDetId() == Belle2.Const.PXD:
                last_cluster = pxd_clusters[last_hit.getHitId()]
                last_true_hit = last_cluster.getRelated("PXDTrueHits")
            elif last_hit.getDetId() == Belle2.Const.SVD:
                last_cluster = svd_clusters[last_hit.getHitId()]
                last_true_hit = last_cluster.getRelated("SVDTrueHits")
 
            vxdID = first_cluster.getSensorID()
            sensorInfoBase = Belle2.VXD.GeoCache.getInstance().getSensorInfo(vxdID)
            last_momentum = sensorInfoBase.vectorToGlobal(last_true_hit.getMomentum(), True)
 
            return_dict.update(dict(first_hit_mom_x=first_momentum.X(),
                                    first_hit_mom_y=first_momentum.Y(),
                                    first_hit_mom_z=first_momentum.Z(),
                                    last_hit_mom_x=last_momentum.X(),
                                    last_hit_mom_y=last_momentum.Y(),
                                    last_hit_mom_z=last_momentum.Z()))
 
        return return_dict
 
    
    save_tree = refiners.SaveTreeRefiner()