def ncrcat(input,output,kwargs={}):
'''Call `ncrcat` via task_manager
Parameters
----------
input : list
list of files to concatenate
output : str
output file
kwargs : dict, optional
dictionary of keyword arguments to task_manager.submit
'''
kwargs['modules'] = ['nco']
kwargs['module_purge'] = False
if 'memory' not in kwargs:
kwargs['memory'] = '100GB'
(fid,tmpfile) = tempfile.mkstemp('.filelist')
with open(tmpfile,'w') as fid:
for f in input:
fid.write('%s\n'%f)
jid = tm.submit(['cat',tmpfile,'|','ncrcat','-o',output],**kwargs)
return jid
def ncop_chunktime(script,
kwargs,
chunk_size,
start=0,
stop=None,
clobber=False,
cleanup=True):
'''
run script on time segments within a file and concatenate results
'''
jid_list = []
def op_one_chunk(tnx):
#-- intermediate output file
file_out_i = file_out + '.tnx.%d-%d' % (tnx)
#-- update input arguments
kwargs.update({'dimsub': {'time': tnx}, 'file_out': file_out_i})
#-- submit
print '\'{0}\''.format(json.dumps(kwargs))
if not os.path.exists(file_out_i) or clobber:
jid = tm.submit([script, '\'{0}\''.format(json.dumps(kwargs))])
jid_list.extend(jid)
return file_out_i
file_out = copy.copy(kwargs['file_out'])
if not os.path.exists(file_out) or clobber:
#-- get time chunks
if stop is None:
ds = xr.open_dataset(kwargs['file_in'],
decode_times=False,
decode_coords=False)
stop = len(ds.time)
time_chunks = gen_time_chunks(start, stop, chunk_size)
#-- operate on each chunk
file_cat = [op_one_chunk(tnx) for tnx in time_chunks]
#-- concatenate files
jid = ncrcat(file_cat, file_out, depjob=jid_list)
if cleanup:
tm.submit(['rm', '-f', ' '.join(file_cat)], depjob=jid)
return jid_list
def ncrcat(input, output, depjob=[]):
'''
call ncrcat
'''
(fid, tmpfile) = tempfile.mkstemp('.filelist')
with open(tmpfile, 'w') as fid:
for f in input:
fid.write('%s\n' % f)
jid = tm.submit(['cat', tmpfile, '|', 'ncrcat', '-o', output],
depjob=depjob)
return jid
def op_one_chunk(tnx):
#-- intermediate output file
file_out_i = file_out + '.tnx.%d-%d' % (tnx)
#-- update input arguments
kwargs.update({'dimsub': {'time': tnx}, 'file_out': file_out_i})
#-- submit
print '\'{0}\''.format(json.dumps(kwargs))
if not os.path.exists(file_out_i) or clobber:
jid = tm.submit([script, '\'{0}\''.format(json.dumps(kwargs))])
jid_list.extend(jid)
return file_out_i
file_out_i = file_out+'.tnx.%d-%d'%(tnx)
if os.path.exists(file_out_i) and not clobber:
return file_out_i
#-- update input arguments
if isel in kwargs:
kwargs['isel'] = kwargs['isel'].update({'time':tnx})
else:
kwargs['isel'] = {'time':tnx}
kwargs.update('file_out': file_out_i})
#-- submit
print(json_cmd(kwargs))
jid = tm.submit([script,json_cmd(kwargs)],**submit_kwargs_i)
jid_list.append(jid)
return file_out_i
#-- get stopping index
if stop is None:
stop = len(xr.open_dataset(kwargs['file_in'],
decode_times=False,
decode_coords=False).time)
time_chunks = gen_time_chunks(start,stop,chunk_size)
#-- operate on each chunk
file_cat = [_apply_one_chunk(tnx) for tnx in time_chunks]
'output_dir': odir,
'max_depth': max_depth,
'clobber': clobber
}
script = 'assemble_wod_database_nc.py'
files_out = []
for file_in in files:
file_out = os.path.join(odir, os.path.basename(file_in) + '.nc')
files_out.append(file_out)
if not os.path.exists(file_out) or clobber:
kwargs = settings.copy()
kwargs.update({'file_in': file_in, 'file_out': file_out})
jid = tm.submit([script, '\'{0}\''.format(json.dumps(kwargs))])
tm.wait()
#-- if 2 phases, resubmit this script and exit
if split_phase and tm.total_elapsed_time() > 1:
print('resubmitting myself')
call('bsub < run_global_req_temp_salt_o2.py', shell=True)
sys.exit(0)
#-----------------------------------------------------------------
#-- make gridded
#-----------------------------------------------------------------
clobber = False
'q': 'caldera',
'm': '12000'
}
bset_lg = {
'q': 'geyser',
'm': '300000'
}
# stop script (graceful exit) and wait on jobs after:
tm.QUEUE_MAX_HOURS = 1. / 3600.
# max number of jobs in queue
tm.MAXJOBS = 50
# loop and submit
jobs = []
for i in range(1, 3):
jid = tm.submit(['sleep', '10'])
jobs.extend(jid)
time.sleep(2)
# submit dependent job
print('dep job')
jid = tm.submit(['ls', '-1'], depjob=jobs)
# wait on all jobs
ok = tm.wait()
# report status
print 'OK (False indicates an error): ' + str(ok)
