def test_particle_queueio(tmpdir, batch_size, n_reads=10):
queueio_name = "queueio_{}".format(uuid.uuid4())
# First, create a file to write to:
file_name = str(tmpdir + "/test_queueio_particles_{}.h5".format(queueio_name))
# Next, write some particles to that file:
create_particle_file(file_name, rand_num_events=25)
# Now, let's get the configuration of a queueio object:
default_config = larcv.QueueProcessor.default_config()
default_config["InputFiles"].append(file_name)
# Add the batch filler to the default config:
default_config["Verbosity"] = 0
default_config["ProcessDriver"]["ProcessName"].append(f"test_{queueio_name}")
default_config["ProcessDriver"]["ProcessType"].append("BatchFillerParticle")
if default_config["ProcessDriver"]["ProcessList"] is None:
default_config["ProcessDriver"]["ProcessList"] = {
f"test_{queueio_name}":
{
"ParticleProducer": "test",
}
}
else:
default_config["ProcessDriver"]["ProcessList"].append(
{f"test_{queueio_name}":
{
"ParticleProducer": "test",
}
})
queue_proc = larcv.QueueProcessor()
queue_proc.configure(default_config)
indexes = numpy.arange(batch_size*n_reads*2) % queue_proc.get_n_entries()
queue_proc.set_next_batch(indexes[0:batch_size])
queue_proc.prepare_next()
while queue_proc.is_reading():
print("Sleeping")
time.sleep(0.1)
queue_proc.pop_current_data()
for i in range(n_reads):
batch=indexes[i*batch_size:(i+1)*batch_size]
queue_proc.set_next_batch(batch)
queue_proc.prepare_next()
while queue_proc.is_reading():
print("Sleeping")
time.sleep(0.1)
def test_sparsetensor2d_queueio(tmpdir, batch_size, n_projections, n_reads=10):
queueio_name = "queueio_{}".format(uuid.uuid4())
# First, create a file to write to:
file_name = str(tmpdir +
"/test_queueio_sparsetensor2d_{}.h5".format(queueio_name))
# Next, write some sparsetensor2ds to that file:
create_sparsetensor2d_file(file_name,
rand_num_events=25,
n_projections=n_projections)
# Now, let's get the configuration of a queueio object:
default_config = larcv.QueueProcessor.default_config()
default_config["InputFiles"].append(file_name)
# Add the batch filler to the default config:
default_config["Verbosity"] = 0
default_config["ProcessDriver"]["ProcessName"].append(
f"test_{queueio_name}")
default_config["ProcessDriver"]["ProcessType"].append(
"BatchFillerSparseTensor2D")
process_list = {
f"test_{queueio_name}": {
"TensorProducer": "test",
"MaxVoxels": 100,
"UnfilledVoxelValue": -999,
"Channels": list(range(n_projections)),
}
}
if default_config["ProcessDriver"]["ProcessList"] is None:
default_config["ProcessDriver"]["ProcessList"] = process_list
else:
default_config["ProcessDriver"]["ProcessList"].append(process_list)
queue_proc = larcv.QueueProcessor()
queue_proc.configure(default_config)
indexes = numpy.arange(
batch_size * n_reads * 2) % queue_proc.get_n_entries()
queue_proc.set_next_batch(indexes[0:batch_size])
queue_proc.prepare_next()
while queue_proc.is_reading():
print("Sleeping")
time.sleep(0.1)
queue_proc.pop_current_data()
for i in range(n_reads):
batch = indexes[i * batch_size:(i + 1) * batch_size]
queue_proc.set_next_batch(batch)
queue_proc.prepare_next()
while queue_proc.is_reading():
print("Sleeping")
time.sleep(0.1)
def configure(self,cfg, color=0):
# if "this" was configured before, reset it
if self._name: self.reset()
# get name
if not cfg['filler_name']:
sys.stderr.write('filler_name is empty!\n')
raise ValueError
# ensure unique name
if self.__class__.exist(cfg['filler_name']) and not self.__class__.instance_by_name(cfg['filler_name']) == self:
sys.stderr.write('filler_name %s already running!' % cfg['filler_name'])
return
self._name = cfg['filler_name']
# get QueueProcessor config file
self._cfg_file = cfg['filler_cfg']
# if not self._cfg_file or not os.path.isfile(self._cfg_file):
# sys.stderr.write('filler_cfg file does not exist: %s\n' % self._cfg_file)
# raise ValueError
# set verbosity
if 'verbosity' in cfg:
self._verbose = bool(cfg['verbosity'])
# configure thread processor
self._proc = larcv.QueueProcessor(self._name)
self._proc.configure(self._cfg_file, color)
# fetch batch filler info
self._storage = {}
# Making a map between process names and data outputs:
for pid, datatype in zip(self._proc.batch_fillers(), self._proc.batch_types()):
# What's the name from the config?
name = self._proc.storage_name(pid)
datatype = larcv.BatchDataTypeName(datatype)
self._storage[name]=batch_pydata(datatype)
if 'make_copy' in cfg and cfg['make_copy']:
self._storage[name]._make_copy = True
# all success?
# register *this* instance
self.__class__._instance_m[self._name] = self
def test_tensor3d_queueio(tmpdir, make_copy, batch_size, from_dense, n_reads=2):
queueio_name = "queueio_{}".format(uuid.uuid4())
# First, create a file to write to:
file_name = str(tmpdir + "/test_queueio_tensor3d_{}.h5".format(queueio_name))
# Next, write some tensor3ds to that file:
if from_dense:
create_dense_tensor3d_file(file_name, rand_num_events=25)
else:
create_tensor3d_file(file_name, rand_num_events=25)
if from_dense:
tensor_type = "dense"
else:
tensor_type = "sparse"
# Now, let's get the configuration of a queueio object:
default_config = larcv.QueueProcessor.default_config()
default_config["InputFiles"].append(file_name)
# Add the batch filler to the default config:
default_config["Verbosity"] = 0
default_config["ProcessDriver"]["ProcessName"].append(f"test_{queueio_name}")
default_config["ProcessDriver"]["ProcessType"].append("BatchFillerTensor3D")
process_list = {f"test_{queueio_name}":
{
"TensorProducer": "test",
"TensorType": tensor_type,
"Channels": [0,],
"EmptyVoxelValue": -999.,
"Augment": True,
}
}
if default_config["ProcessDriver"]["ProcessList"] is None:
default_config["ProcessDriver"]["ProcessList"] = process_list
else:
default_config["ProcessDriver"]["ProcessList"].append(process_list)
queue_proc = larcv.QueueProcessor()
queue_proc.configure(default_config)
indexes = numpy.arange(batch_size*n_reads*2) % queue_proc.get_n_entries()
queue_proc.set_next_batch(indexes[0:batch_size])
print(queue_proc.is_reading())
queue_proc.prepare_next()
time.sleep(2)
while queue_proc.is_reading():
print("Sleeping")
time.sleep(0.5)
queue_proc.pop_current_data()
for i in range(n_reads):
batch=indexes[i*batch_size:(i+1)*batch_size]
queue_proc.set_next_batch(batch)
queue_proc.prepare_next()
while queue_proc.is_reading():
print("Sleeping")
time.sleep(0.1)
