def main():
if opts.save_jobjson or opts.save_jobgc or opts.get_events:
(workDir, nJobs, jobList) = getWorkJobs(args)
(log, incomplete, splitter, splitInfo) = (None, False, None, {})
(lumiDict, readDict, writeDict) = ({}, {}, {})
try:
splitter = DataSplitter.loadState(os.path.join(workDir, 'datamap.tar'))
except Exception:
pass
jobList = sorted(jobList)
for jobNum in jobList:
del log
log = utils.ActivityLog('Reading job logs - [%d / %d]' % (jobNum, jobList[-1]))
jobInfo = getJobInfo(workDir, jobNum, lambda retCode: retCode == 0)
if not jobInfo:
if not incomplete:
print 'WARNING: Not all jobs have finished - results will be incomplete!'
incomplete = True
continue
if not parameterized:
if splitter:
splitInfo = splitter.getSplitInfo(jobNum)
outputName = splitInfo.get(DataSplitter.Nickname, splitInfo.get(DataSplitter.DatasetID, 0))
else:
outputName = jobInfo['file'].split()[2].replace("_%d_" % jobNum, '_').replace('/', '_').replace('__', '_')
# Read framework report files to get number of events
try:
outputDir = os.path.join(workDir, 'output', 'job_' + str(jobNum))
for fwkXML in getCMSSWInfo(os.path.join(outputDir, 'cmssw.dbs.tar.gz')):
for run in fwkXML.getElementsByTagName('Run'):
for lumi in run.getElementsByTagName('LumiSection'):
run_id = int(run.getAttribute('ID'))
lumi_id = int(lumi.getAttribute('ID'))
lumiDict.setdefault(outputName, {}).setdefault(run_id, set()).add(lumi_id)
for outFile in fwkXML.getElementsByTagName('File'):
pfn = outFile.getElementsByTagName('PFN')[0].childNodes[0].data
if pfn not in writeDict.setdefault(outputName, {}):
writeDict[outputName][pfn] = 0
writeDict[outputName][pfn] += int(outFile.getElementsByTagName('TotalEvents')[0].childNodes[0].data)
for inFile in fwkXML.getElementsByTagName('InputFile'):
if outputName not in readDict:
readDict[outputName] = 0
readDict[outputName] += int(inFile.getElementsByTagName('EventsRead')[0].childNodes[0].data)
except KeyboardInterrupt:
sys.exit(os.EX_OK)
except Exception:
raise
print 'Error while parsing framework output of job %s!' % jobNum
continue
del log
log = utils.ActivityLog('Simplifying lumi sections')
lumis = {}
for sample in lumiDict:
for run in lumiDict[sample]:
for lumi in lumiDict[sample][run]:
lumis.setdefault(sample, []).append(([run, lumi], [run, lumi]))
for sample in lumiDict:
lumis[sample] = mergeLumi(lumis[sample])
del log
for sample, lumis in lumis.items():
print 'Sample:', sample
print '========================================='
print 'Number of events processed: %12d' % readDict[sample]
print ' Number of events written: %12d' % sum(writeDict.get(sample, {}).values())
if writeDict.get(sample, None):
print
head = [(0, ' Output filename'), (1, 'Events')]
utils.printTabular(head, map(lambda pfn: {0: pfn, 1: writeDict[sample][pfn]}, writeDict[sample]))
if opts.save_jobjson:
outputJSON(lumis, open(os.path.join(workDir, 'processed_%s.json' % sample), 'w'))
print 'Saved processed lumi sections in', os.path.join(workDir, 'processed_%s.json' % sample)
if opts.save_jobgc:
print
print 'List of processed lumisections:'
print '-----------------------------------------'
outputGC(lumis)
print
###########################
# Lumi filter manuipulation
###########################
if opts.save_exprgc or opts.save_exprjson or opts.save_exprfull:
if len(args) == 0:
raise Exception('No arguments given!')
try:
lumis = parseLumiFilter(str.join(' ', args))
except Exception:
raise Exception('Could not parse: %s' % str.join(' ', args))
if opts.save_exprgc:
outputGC(lumis)
if opts.save_exprjson:
outputJSON(lumis)
if opts.save_exprfull:
result = {}
for rlrange in lumis:
start, end = rlrange
assert(start[0] == end[0])
llist = result.setdefault(start[0], []).extend(range(start[1], end[1] + 1))
print result
if opts.findrm:
removed = []
utils.eprint = lambda *x: {}
oldDP = DataProvider.loadState(args[0])
for new in args[1:]:
newDP = DataProvider.loadState(new)
(blocksAdded, blocksMissing, blocksChanged) = DataProvider.resyncSources(oldDP.getBlocks(), newDP.getBlocks())
for block in blocksMissing:
tmp = dict(block)
tmp[-1] = new
removed.append(tmp)
oldDP = newDP
utils.printTabular([(DataProvider.Dataset, "Dataset"), (DataProvider.BlockName, "Block"), (-1, "Removed in file")], removed)
if opts.invalid:
splitter = DataSplitter.loadState(opts.invalid)
def getInvalid():
for jobNum in range(splitter.getMaxJobs()):
splitInfo = splitter.getSplitInfo(jobNum)
if splitInfo.get(DataSplitter.Invalid, False):
yield str(jobNum)
print str.join(",", getInvalid())
if opts.jdl:
print job.get("jdl")
if opts.state:
try:
newState = getattr(Job, opts.state)
except:
print "Invalid state: %s", opts.state
def main():
# Set config based on settings from config file or command line
configFile = None
if os.path.exists(args[0]):
configFile = args[0]
config = getConfig(configFile, section = 'global')
config.changeView(setSections = ['jobs']).set('nseeds', '1', '?=')
configParameters = config.changeView(setSections = ['parameters'])
if opts.parameters:
utils.vprint('Provided options:')
for p in opts.parameters:
k, v = p.split('=', 1)
configParameters.set(k.strip(), v.strip().replace('\\n', '\n'), '=')
utils.vprint('\t%s: %s' % (k.strip(), v.strip()))
utils.vprint('')
if not os.path.exists(args[0]):
configParameters.set('parameters', str.join(' ', args).replace('\\n', '\n'))
if opts.dataset:
configParameters.set('default lookup', 'DATASETNICK')
# configParameters.set('parameter adapter', 'BasicParameterAdapter', '=') # Don't track parameter changes
if opts.verbosity > 2:
config.changeView(setSections = None).write(sys.stdout)
# Initialize ParameterFactory
configTask = config.changeView(setSections = [config.get(['task', 'module'], 'DummyTask')])
pm = config.getPlugin('parameter factory', 'SimpleParameterFactory', cls = ParameterFactory).getInstance()
# Create dataset parameter source
class DummySplitter:
def getMaxJobs(self):
return 3
def getSplitInfo(self, pNum):
mkEntry = lambda ds, fl, n, nick: { DataSplitter.Dataset: ds, DataSplitter.Nickname: nick,
DataSplitter.FileList: fl, DataSplitter.NEntries: n }
rndStr = lambda: md5(str(random.random())).hexdigest()[:10]
tmp = [ mkEntry('ds1', ['a', 'b'], 23, 'data_1'), mkEntry('ds1', ['1'], 42, 'data_1'),
mkEntry('ds2', ['m', 'n'], 123, 'data_2'), mkEntry('ds2', ['x', 'y', 'z'], 987, 'data_3') ]
return tmp[pNum]
class DataSplitProcessorTest:
def getKeys(self):
return map(lambda k: ParameterMetadata(k, untracked=True),
['DATASETINFO', 'DATASETID', 'DATASETPATH', 'DATASETBLOCK', 'DATASETNICK'])
def process(self, pNum, splitInfo, result):
result.update({
'DATASETINFO': '',
'DATASETID': splitInfo.get(DataSplitter.DatasetID, None),
'DATASETPATH': splitInfo.get(DataSplitter.Dataset, None),
'DATASETBLOCK': splitInfo.get(DataSplitter.BlockName, None),
'DATASETNICK': splitInfo.get(DataSplitter.Nickname, None),
'DATASETSPLIT': pNum,
})
if opts.dataset.lower() == 'true':
utils.vprint('Registering dummy data provider data')
dataSplitter = DummySplitter()
elif opts.dataset:
dataSplitter = DataSplitter.loadState(opts.dataset)
if opts.dataset:
DataParameterSource.datasetsAvailable['data'] = DataParameterSource(
config.getWorkPath(), 'data', None, dataSplitter, DataSplitProcessorTest())
psource = pm.getSource(config)
if opts.forceiv:
for dp in DataParameterSource.datasetSources:
dp.intervention = (set([1]), set([0]), True)
if opts.listparams:
result = []
needGCParam = False
if psource.getMaxJobs() != None:
countActive = 0
for jobNum in range(psource.getMaxJobs()):
info = psource.getJobInfo(jobNum)
if info[ParameterInfo.ACTIVE]:
countActive += 1
if opts.inactive or info[ParameterInfo.ACTIVE]:
if not info[ParameterInfo.ACTIVE]:
info['GC_PARAM'] = 'N/A'
if str(info['GC_PARAM']) != str(jobNum):
needGCParam = True
result.append(info)
if opts.displaymode == 'parseable':
utils.vprint('Count,%d,%d' % (countActive, psource.getMaxJobs()))
else:
utils.vprint('Number of parameter points: %d' % psource.getMaxJobs())
if countActive != psource.getMaxJobs():
utils.vprint('Number of active parameter points: %d' % countActive)
else:
result.append(psource.getJobInfo(123))
enabledOutput = opts.output.split(',')
output = filter(lambda k: not opts.output or k in enabledOutput, psource.getJobKeys())
stored = filter(lambda k: k.untracked == False, output)
untracked = filter(lambda k: k.untracked == True, output)
if opts.collapse > 0:
result_old = result
result = {}
result_nicks = {}
head = [('COLLATE_JOBS', '# of jobs')]
if 'DATASETSPLIT' in stored:
stored.remove('DATASETSPLIT')
if (opts.collapse == 1):
stored.append('DATASETNICK')
head.append(('DATASETNICK', 'DATASETNICK'))
elif opts.collapse == 2:
head.append(('COLLATE_NICK', '# of nicks'))
for pset in result_old:
if ('DATASETSPLIT' in pset) and (opts.collapse == 1):
pset.pop('DATASETSPLIT')
nickname = None
if ('DATASETNICK' in pset) and (opts.collapse == 2):
nickname = pset.pop('DATASETNICK')
h = md5(repr(map(lambda key: pset.get(key), stored))).hexdigest()
result.setdefault(h, []).append(pset)
result_nicks.setdefault(h, set()).add(nickname)
def doCollate(h):
tmp = result[h][0]
tmp['COLLATE_JOBS'] = len(result[h])
tmp['COLLATE_NICK'] = len(result_nicks[h])
return tmp
result = map(doCollate, result)
else:
head = [('GC_JOB_ID', '#')]
if needGCParam:
head.append(('GC_PARAM', 'GC_PARAM'))
if opts.active:
head.append((ParameterInfo.ACTIVE, 'ACTIVE'))
if opts.visible:
stored = opts.visible.split(',')
head.extend(sorted(zip(stored, stored)))
if opts.untracked:
head.extend(sorted(map(lambda n: (n, '(%s)' % n), filter(lambda n: n not in ['GC_PARAM', 'GC_JOB_ID'], untracked))))
utils.vprint('')
utils.printTabular(head, result)
if opts.save:
utils.vprint('')
ParameterSource.getClass('GCDumpParameterSource').write(opts.save, psource)
utils.vprint('Parameter information saved to ./%s' % opts.save)
if opts.intervention:
utils.vprint('')
tmp = psource.getJobIntervention()
if tmp:
if opts.displaymode == 'parseable':
utils.vprint('R: %s' % str.join(',', map(str, tmp[0])))
utils.vprint('D: %s' % str.join(',', map(str, tmp[1])))
else:
utils.vprint(' Redo: %r' % tmp[0])
utils.vprint('Disable: %r' % tmp[1])
else:
if opts.displaymode == 'parseable':
utils.vprint('NOINT')
else:
utils.vprint('No intervention')
