def master_setup_bulk_subjobs(self, jobs, jdefids):
from Ganga.GPIDev.Lib.Job.Job import Job
master_job = self.getJobObject()
for i in range(len(jdefids)):
j = Job()
j.copyFrom(master_job)
j.splitter = None
j.backend = Panda()
j.backend.id = jdefids[i]
j.id = i
j.status = 'submitted'
j.time.timenow('submitted')
master_job.subjobs.append(j)
return True
def master_setup_bulk_subjobs(self, jobs, jdefids):
from Ganga.GPIDev.Lib.Job.Job import Job
master_job=self.getJobObject()
for i in range(len(jdefids)):
j=Job()
j.copyFrom(master_job)
j.splitter = None
j.backend=Panda()
j.backend.id = jdefids[i]
j.id = i
j.status = 'submitted'
j.time.timenow('submitted')
master_job.subjobs.append(j)
return True
def submit(N, K):
jobs = []
for i in range(K):
j = Job()
j._auto__init__()
j.backend = LCG()
j.backend.middleware = 'GLITE'
j.splitter = GenericSplitter()
j.splitter.attribute = 'application.args'
j.splitter.values = [['x']] * N
j.submit()
jobs.append(j)
import time
def finished():
for j in jobs:
if not j.status in ['failed', 'completed']:
return False
return True
while not finished():
time.sleep(1)
return jobs
def createUnits(self):
"""Create new units if required given the inputdata"""
# call parent for chaining
super(CoreTransform, self).createUnits()
# Use the given splitter to create the unit definitions
if len(self.units) > 0:
# already have units so return
return
if self.unit_splitter == None and len(self.inputdata) == 0:
raise ApplicationConfigurationError(
None,
"No unit splitter or InputData provided for CoreTransform unit creation, Transform %d (%s)"
% (self.getID(), self.name),
)
# -----------------------------------------------------------------
# split over unit_splitter by preference
if self.unit_splitter:
# create a dummy job, assign everything and then call the split
j = Job()
j.backend = self.backend.clone()
j.application = self.application.clone()
if self.inputdata:
j.inputdata = self.inputdata.clone()
subjobs = self.unit_splitter.split(j)
if len(subjobs) == 0:
raise ApplicationConfigurationError(
None,
"Unit splitter gave no subjobs after split for CoreTransform unit creation, Transform %d (%s)"
% (self.getID(), self.name),
)
# only copy the appropriate elements
fields = []
if len(self.fields_to_copy) > 0:
fields = self.fields_to_copy
elif isType(self.unit_splitter, GenericSplitter):
if self.unit_splitter.attribute != "":
fields = [self.unit_splitter.attribute.split(".")[0]]
else:
for attr in self.unit_splitter.multi_attrs.keys():
fields.append(attr.split(".")[0])
# now create the units from these jobs
for sj in subjobs:
unit = CoreUnit()
for attr in fields:
setattr(unit, attr, copy.deepcopy(getattr(sj, attr)))
self.addUnitToTRF(unit)
# -----------------------------------------------------------------
# otherwise split on inputdata
elif len(self.inputdata) > 0:
if self.files_per_unit > 0:
# combine all files and split accorindgly
filelist = []
for ds in self.inputdata:
if isType(ds, GangaDataset):
for f in ds.files:
if f.containsWildcards():
# we have a wildcard so grab the subfiles
for sf in f.getSubFiles(process_wildcards=True):
filelist.append(sf)
else:
# no wildcards so just add the file
filelist.append(f)
else:
logger.warning("Dataset '%s' doesn't support files" % getName(ds))
# create DSs and units for this list of files
fid = 0
while fid < len(filelist):
unit = CoreUnit()
unit.name = "Unit %d" % len(self.units)
unit.inputdata = GangaDataset(files=filelist[fid : fid + self.files_per_unit])
unit.inputdata.treat_as_inputfiles = self.inputdata[0].treat_as_inputfiles
fid += self.files_per_unit
self.addUnitToTRF(unit)
else:
# just produce one unit per dataset
for ds in self.inputdata:
# avoid splitting over chain inputs
if isType(ds, TaskChainInput):
continue
unit = CoreUnit()
unit.name = "Unit %d" % len(self.units)
unit.inputdata = copy.deepcopy(ds)
self.addUnitToTRF(unit)
def createUnits(self):
"""Create new units if required given the inputdata"""
# call parent for chaining
super(CoreTransform, self).createUnits()
# Use the given splitter to create the unit definitions
if len(self.units) > 0:
# already have units so return
return
if self.unit_splitter == None and len(self.inputdata) == 0:
raise ApplicationConfigurationError(
"No unit splitter or InputData provided for CoreTransform unit creation, Transform %d (%s)"
% (self.getID(), self.name))
# -----------------------------------------------------------------
# split over unit_splitter by preference
if self.unit_splitter:
# create a dummy job, assign everything and then call the split
j = Job()
j.backend = self.backend.clone()
j.application = self.application.clone()
if self.inputdata:
j.inputdata = self.inputdata.clone()
subjobs = self.unit_splitter.split(j)
if len(subjobs) == 0:
raise ApplicationConfigurationError(
"Unit splitter gave no subjobs after split for CoreTransform unit creation, Transform %d (%s)"
% (self.getID(), self.name))
# only copy the appropriate elements
fields = []
if len(self.fields_to_copy) > 0:
fields = self.fields_to_copy
elif isType(self.unit_splitter, GenericSplitter):
if self.unit_splitter.attribute != "":
fields = [self.unit_splitter.attribute.split(".")[0]]
else:
for attr in self.unit_splitter.multi_attrs.keys():
fields.append(attr.split(".")[0])
# now create the units from these jobs
for sj in subjobs:
unit = CoreUnit()
for attr in fields:
setattr(unit, attr, copy.deepcopy(getattr(sj, attr)))
self.addUnitToTRF(unit)
# -----------------------------------------------------------------
# otherwise split on inputdata
elif len(self.inputdata) > 0:
if self.files_per_unit > 0:
# combine all files and split accorindgly
filelist = []
for ds in self.inputdata:
if isType(ds, GangaDataset):
for f in ds.files:
if f.containsWildcards():
# we have a wildcard so grab the subfiles
for sf in f.getSubFiles(
process_wildcards=True):
filelist.append(sf)
else:
# no wildcards so just add the file
filelist.append(f)
else:
logger.warning("Dataset '%s' doesn't support files" %
getName(ds))
# create DSs and units for this list of files
fid = 0
while fid < len(filelist):
unit = CoreUnit()
unit.name = "Unit %d" % len(self.units)
unit.inputdata = GangaDataset(
files=filelist[fid:fid + self.files_per_unit])
unit.inputdata.treat_as_inputfiles = self.inputdata[
0].treat_as_inputfiles
fid += self.files_per_unit
self.addUnitToTRF(unit)
else:
# just produce one unit per dataset
for ds in self.inputdata:
# avoid splitting over chain inputs
if isType(ds, TaskChainInput):
continue
unit = CoreUnit()
unit.name = "Unit %d" % len(self.units)
unit.inputdata = copy.deepcopy(ds)
self.addUnitToTRF(unit)
