v10 Cluster3D points are not saved

[bear-is-asleep]

I'm submitting a ...

• bug report

Cluster3D points are not saved, so no truth pointwise analysis can be done.

Checklist

• I've added at least 1 label under Labels
• Assigned a reviewer under Assignees

Expected Behavior

clust_label in the larcv files should be occupied with labeled voxels.

Current Behavior

clust_label is empty

Possible Solution (optional)

I'm hoping this is just due to some labels being changed. The product and module labels are identical in v9 and v10. Something fundamental happened in Supera (or upstream) which needs to be debugged.

How to reproduce

Produce a single event with any generator. I used particle gun

lar -c pgun.fcl -n 1
lar -c g4_sce.fcl -s prod*.root
lar -c detsim_sce.fcl -s prod*_G4*.root
lar -c reco1_mpvmpr.fcl -s prod*_G4*_DetSim*.root

On a machine with spine (you just need the repo for the driver)

import sys
import yaml

SOFTWARE_DIR = '/sdf/data/neutrino/software/spine' # Change this path to your software install if you are not on S3DF
DATA_DIR = '/sdf/data/neutrino/bearc/larcv/v10_03_01_pgun_uniform/' #Or your path if local

# Set software directory
sys.path.insert(0, SOFTWARE_DIR)

#Spine imports
from spine.driver import Driver
from spine.utils.globals import *

cfg_path = 'mc.cfg'

#Data
DATA_PATH = DATA_DIR + 'larcv*.root'
cfg_data = yaml.safe_load(open(cfg_path, 'r').read().replace('DATA_PATH', DATA_PATH))

# prepare function configures necessary "handlers"
driver_data = Driver(cfg_data)

data = driver_data.process()

#Collect one image
entry = 0
clust_label = data['clust_label'][entry]

len(clust_label) #this is 0, but should be non-zero

#mc.cfg
io:
  loader:
    batch_size: 64
    shuffle: False
    num_workers: 4
    collate_fn: all
    dataset:
      name: LArCVDataset
      file_keys: DATA_PATH
      schema:
        input_data:
          parser: sparse3d
          #sparse_event: sparse3d_pcluster
          args:
            sparse_event_list:
              - sparse3d_reco
              - sparse3d_reco_chi2
              - sparse3d_reco_hit_charge0
              - sparse3d_reco_hit_charge1
              - sparse3d_reco_hit_charge2
              - sparse3d_reco_hit_key0
              - sparse3d_reco_hit_key1
              - sparse3d_reco_hit_key2
        seg_label:
          parser: sparse3d
          sparse_event: sparse3d_pcluster_semantics_ghost
        clust_label:
          parser: cluster3d
          cluster_event: cluster3d_pcluster
          particle_event: particle_corrected
          sparse_semantics_event: sparse3d_pcluster_semantics
          add_particle_info: true
          clean_data: true
        ppn_label:
          parser: particle_points
          particle_event: particle_corrected
          sparse_event: sparse3d_pcluster
        meta:
          parser: meta
          sparse_event: sparse3d_pcluster
        run_info:
          parser: run_info
          sparse_event: sparse3d_pcluster
        sed:
          parser: cluster3d
          args: {cluster_event: cluster3d_sed, add_particle_info: false, clean_data: false}

[bear-is-asleep]

Something I've noticed is the track IDs of MCParticles are now negative.

[bear-is-asleep]

Resolved with PR #645