def createNewJob(self):
"""Create any jobs required for this unit"""
j = GPI.Job()
j._impl.backend = self._getParent().backend.clone()
j._impl.application = self._getParent().application.clone()
if self.inputdata:
j.inputdata = self.inputdata.clone()
trf = self._getParent()
task = trf._getParent()
if trf.outputdata:
j.outputdata = trf.outputdata.clone()
elif j.inputdata and j.inputdata._impl._name == "ATLASLocalDataset" and j.application._impl._name != "TagPrepare":
j.outputdata = GPI.ATLASOutputDataset()
elif j.application._impl._name != "TagPrepare":
j.outputdata = GPI.DQ2OutputDataset()
# check for ds name specified and length
if j.outputdata and j.outputdata._impl._name == "DQ2OutputDataset":
max_length = configDQ2['OUTPUTDATASET_NAMELENGTH'] - 11
# merge names need to be shorter
if (j.backend._impl._name == "Panda" or j.backend._impl._name == "Jedi"):
if j.backend.requirements.enableMerge:
max_length -= 12
if j.backend._impl._name == "Jedi":
# go over the outputdata and check for output names that Jedi appends to the outDS name
tmp_len_chg = 8
for o in j.outputdata.outputdata:
if (len(o)+1) > tmp_len_chg:
tmp_len_chg = len(o)+1
max_length -= tmp_len_chg
elif j.backend.individualOutDS:
max_length -= 8
if j.outputdata.datasetname != "":
dsn = [j.outputdata.datasetname, "j%i.t%i.trf%i.u%i" %
(j.id, task.id, trf.getID(), self.getID())]
if len(".".join(dsn)) > max_length:
dsn = [j.outputdata.datasetname[: - (len(".".join(dsn)) - max_length)], "j%i.t%i.trf%i.u%i" %
(j.id, task.id, trf.getID(), self.getID())]
else:
dsn = [trf.getContainerName()[:-1], self.name, "j%i.t%i.trf%i.u%i" %
(j.id, task.id, trf.getID(), self.getID())]
if len(".".join(dsn)) > max_length:
dsn2 = [trf.getContainerName(2 * max_length / 3)[:-1], "", "j%i.t%i.trf%i.u%i" % (j.id, task.id, trf.getID(), self.getID())]
dsn = [trf.getContainerName(2 * max_length / 3)[:-1], self.name[: - (len(".".join(dsn2)) - max_length)], "j%i.t%i.trf%i.u%i" %
(j.id, task.id, trf.getID(), self.getID())]
j.outputdata.datasetname = '.'.join(dsn).replace(":", "_").replace(" ", "").replace(",","_")
j.inputsandbox = self._getParent().inputsandbox
j.outputsandbox = self._getParent().outputsandbox
# check for splitter - TagPrepare and Jedi don't user splitters
if j.application._impl._name == "TagPrepare":
return j
if j.backend._impl._name == "Jedi":
if trf.files_per_job > 0:
j.backend.requirements.nFilesPerJob = trf.files_per_job
elif trf.MB_per_job > 0:
j.backend.requirements.nGBPerJob = trf.MB_per_job / 1000
return j
if not trf.splitter:
# provide a default number of files if there's nothing else given
nfiles = trf.files_per_job
if nfiles < 1:
nfiles = 5
if j.inputdata._impl._name == "ATLASLocalDataset":
j.splitter = AthenaSplitterJob()
if trf.subjobs_per_unit > 0:
j.splitter.numsubjobs = trf.subjobs_per_unit
else:
import math
j.splitter.numsubjobs = int( math.ceil( len(j.inputdata.names) / float(nfiles) ) )
else:
j.splitter = DQ2JobSplitter()
if trf.MB_per_job > 0:
j.splitter.filesize = trf.MB_per_job
elif trf.subjobs_per_unit > 0:
j.splitter.numsubjobs = trf.subjobs_per_unit
else:
j.splitter.numfiles = nfiles
else:
j.splitter = trf.splitter.clone()
# postprocessors
if len(self._getParent().postprocessors.process_objects) > 0:
import copy
j.postprocessors = copy.deepcopy( addProxy(self._getParent()).postprocessors )
return j
