trfExe.py

Go to the documentation of this file.

# Copyright (C) 2002-2025 CERN for the benefit of the ATLAS collaboration
 
 
 
import json
import math
import os
import os.path as path
import re
import signal
import subprocess
import sys
import time
 
import logging
from fnmatch import fnmatch
msg = logging.getLogger(__name__)
 
from PyJobTransforms.trfJobOptions import JobOptionsTemplate
from PyJobTransforms.trfUtils import asetupReport, asetupReleaseIsOlderThan, unpackDBRelease, setupDBRelease, \
    cvmfsDBReleaseCheck, forceToAlphaNum, \
    ValgrindCommand, VTuneCommand, isInteractiveEnv, calcCpuTime, calcWallTime, analytic, reportEventsPassedSimFilter
from PyJobTransforms.trfExeStepTools import commonExecutorStepName, executorStepSuffix
from PyJobTransforms.trfExitCodes import trfExit
from PyJobTransforms.trfLogger import stdLogLevels
from PyJobTransforms.trfMPTools import detectAthenaMPProcs, athenaMPOutputHandler
from PyJobTransforms.trfMTTools import detectAthenaMTThreads
 
import PyJobTransforms.trfExceptions as trfExceptions
import PyJobTransforms.trfValidation as trfValidation
import PyJobTransforms.trfArgClasses as trfArgClasses
import PyJobTransforms.trfEnv as trfEnv
import PyJobTransforms.trfMPITools as mpi
 
 
# Depending on the setting of LANG, sys.stdout may end up with ascii or ansi
# encoding, rather than utf-8.  But Athena uses unicode for some log messages
# (another example of Gell-Mann's totalitarian principle) and that will result
# in a crash with python 3.  In such a case, force the use of a utf-8 encoded
# output stream instead.
def _encoding_stream (s):
    enc = s.encoding.lower()
    if enc.find('ascii') >= 0 or enc.find('ansi') >= 0:
        return open (s.fileno(), 'w', encoding='utf-8')
    return s
 
 
 
class executorConfig(object):
 
    
    def __init__(self, argdict={}, dataDictionary={}, firstExecutor=False):
        self._argdict = argdict
        self._dataDictionary = dataDictionary
        self._firstExecutor = firstExecutor
        self._executorStep = -1
        self._totalExecutorSteps = 0
 
    @property
    def argdict(self):
        return self._argdict
    
    @argdict.setter
    def argdict(self, value):
        self._argdict = value
 
    @property
    def dataDictionary(self):
        return self._dataDictionary
    
    @dataDictionary.setter
    def dataDictionary(self, value):
        self._dataDictionary = value
 
    @property
    def firstExecutor(self):
        return self._firstExecutor
    
    @firstExecutor.setter
    def firstExecutor(self, value):
        self._firstExecutor = value
 
    @property
    def executorStep(self):
        return self._executorStep
 
    @executorStep.setter
    def executorStep(self, value):
        self._executorStep = value
 
    @property
    def totalExecutorSteps(self):
        return self._totalExecutorSteps
 
    @totalExecutorSteps.setter
    def totalExecutorSteps(self, value):
        self._totalExecutorSteps = value
 
    
    def setFromTransform(self, trf):
        self._argdict = trf.argdict
        self._dataDictionary = trf.dataDictionary
        
    
    def addToArgdict(self, key, value):
        self._argdict[key] = value
 
    
    def addToDataDictionary(self, key, value):
        self._dataDictionary[key] = value
 
 
 
class transformExecutor(object):
    
    
    def __init__(self, name = 'Dummy', trf = None, conf = None, inData = set(), outData = set()):
        # Some information to produce helpful log messages
        
        self._name = forceToAlphaNum(name)
        # Data this executor can start from and produce
        # Note we transform NULL to inNULL and outNULL as a convenience
        self._inData = set(inData)
        self._outData = set(outData)
        if 'NULL' in self._inData:
            self._inData.remove('NULL')
            self._inData.add('inNULL')
        if 'NULL' in self._outData:
            self._outData.remove('NULL')
            self._outData.add('outNULL')
            
        # It's forbidden for an executor to consume and produce the same datatype
        dataOverlap = self._inData & self._outData
        if len(dataOverlap) > 0:
            raise trfExceptions.TransformSetupException(trfExit.nameToCode('TRF_GRAPH_ERROR'), 
                                                        'Executor definition error, executor {0} is not allowed to produce and consume the same datatypes. Duplicated input/output types {1}'.format(self._name, ' '.join(dataOverlap)))
        
        
        if conf is not None:
            self.conf = conf
        else:
            self.conf = executorConfig()
            if trf is not None:
                self.conf.setFromTransform(trf)
        
        # Execution status
        self._hasExecuted = False
        self._rc = -1
        self._errMsg = None
        
        # Validation status
        self._hasValidated = False
        self._isValidated = False
        
        # Extra metadata
        # This dictionary holds extra metadata for this executor which will be 
        # provided in job reports
        self._extraMetadata = {}
        
        
        self._preExeStart = None
        self._exeStart = self._exeStop = None
        self._valStart = self._valStop = None
        self._memStats = {}
        self._memLeakResult = {}
        self._memFullFile = None
        self._eventCount = None
        self._athenaMP = 0
        self._athenaMT = 0
        self._athenaConcurrentEvents = 0
        self._dbMonitor = None
        self._resimevents = None
        self._alreadyInContainer = None
        self._containerSetup = None
 
        # Holder for execution information about any merges done by this executor in MP mode
        self._myMerger = []
 
        
    
    @property
    def myMerger(self):
        return self._myMerger
 
    @property
    def name(self):
        return self._name
 
    @name.setter
    def name(self, value):
        self._name = value
 
    @property
    def substep(self):
        if '_substep' in dir(self):
            return self._substep
        return None
    
    @property
    def trf(self):
        if '_trf' in dir(self):
            return self._trf
        return None
    
    @trf.setter
    def trf(self, value):
        self._trf = value
        
    @property
    def inData(self):
        
        if '_inData' in dir(self):
            return self._inData
        return None
    
    @inData.setter
    def inData(self, value):
        self._inData = set(value)
        
    def inDataUpdate(self, value):
        
        if '_inData' in dir(self):
            self._inData.update(value)
        else:
            
            self.inData = value
 
 
    @property
    def outData(self):
        
        if '_outData' in dir(self):
            return self._outData
        return None
 
    @outData.setter
    def outData(self, value):
        self._outData = set(value)
    
    def outDataUpdate(self, value):
        
        if '_outData' in dir(self):
            self._outData.update(value)
        else:
            
            self.outData = value
    
    @property
    
    def input(self):
        
        if '_input' in dir(self):
            return self._input
        return None
    
    @property
    
    def output(self):
        
        if '_output' in dir(self):
            return self._output
        return None
    
    @property
    def extraMetadata(self):
        return self._extraMetadata
    
    @property
    def hasExecuted(self):
        return self._hasExecuted
    
    @property
    def rc(self):
        return self._rc
    
    @property
    def errMsg(self):
        return self._errMsg
    
    @property
    def validation(self):
        return self._validation
    
    @validation.setter
    def validation(self, value):
        self._validation = value
        
    @property
    def hasValidated(self):
        return self._hasValidated
    
    @property
    def isValidated(self):
        return self._isValidated
    
    
    @property
    def first(self):
        if hasattr(self, '_first'):
            return self._first
        else:
            return None
 
    @property
    def preExeStartTimes(self):
        return self._preExeStart
    
    @property
    def exeStartTimes(self):
        return self._exeStart
 
    @property
    def exeStopTimes(self):
        return self._exeStop
 
    @property
    def valStartTimes(self):
        return self._valStart
 
    @property
    def valStopTimes(self):
        return self._valStop
    
    @property
    def preExeCpuTime(self):
        if self._preExeStart and self._exeStart:
            return calcCpuTime(self._preExeStart, self._exeStart)
        else:
            return None
 
    @property
    def preExeWallTime(self):
        if self._preExeStart and self._exeStart:
            return calcWallTime(self._preExeStart, self._exeStart)
        else:
            return None
 
    @property
    def cpuTime(self):
        if self._exeStart and self._exeStop:
            return calcCpuTime(self._exeStart, self._exeStop)
        else:
            return None
 
    @property
    def usrTime(self):
        if self._exeStart and self._exeStop:
            return self._exeStop[2] - self._exeStart[2]
        else:
            return None
        
    @property
    def sysTime(self):
        if self._exeStart and self._exeStop:
            return self._exeStop[3] - self._exeStart[3]
        else:
            return None
 
    @property
    def wallTime(self):
        if self._exeStart and self._exeStop:
            return calcWallTime(self._exeStart, self._exeStop)
        else:
            return None
        
    @property
    def memStats(self):
        return self._memStats
 
    @property
    def memAnalysis(self):
        return self._memLeakResult
 
    @property
    def postExeCpuTime(self):
        if self._exeStop and self._valStart:
            return calcCpuTime(self._exeStop, self._valStart)
        else:
            return None
 
    @property
    def postExeWallTime(self):
        if self._exeStop and self._valStart:
            return calcWallTime(self._exeStop, self._valStart)
        else:
            return None
 
    @property
    def validationCpuTime(self):
        if self._valStart and self._valStop:
            return calcCpuTime(self._valStart, self._valStop)
        else:
            return None
    
    @property
    def validationWallTime(self):
        if self._valStart and self._valStop:
            return calcWallTime(self._valStart, self._valStop)
        else:
            return None
 
    @property
    def cpuTimeTotal(self):
        if self._preExeStart and self._valStop:
            return calcCpuTime(self._preExeStart, self._valStop)
        else:
            return None
 
    @property
    def wallTimeTotal(self):
        if self._preExeStart and self._valStop:
            return calcWallTime(self._preExeStart, self._valStop)
        else:
            return None
        
    @property
    def eventCount(self):
        return self._eventCount
 
    @property
    def reSimEvent(self):
        return self._resimevents
 
    @property
    def athenaMP(self):
        return self._athenaMP
 
    @property
    def dbMonitor(self):
        return self._dbMonitor
 
 
    # set start times, if not set already
    def setPreExeStart(self):
        if self._preExeStart is None:
            self._preExeStart = os.times()
            msg.debug('preExeStart time is {0}'.format(self._preExeStart))
 
    def setValStart(self):
        if self._valStart is None:
            self._valStart = os.times()
            msg.debug('valStart time is {0}'.format(self._valStart))
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        msg.info('Preexecute for %s', self._name)
        
    def execute(self):
        self._exeStart = os.times()
        msg.debug('exeStart time is {0}'.format(self._exeStart))
        msg.info('Starting execution of %s', self._name)
        self._hasExecuted = True
        self._rc = 0
        self._errMsg = ''
        msg.info('%s executor returns %d', self._name, self._rc)
        self._exeStop = os.times()
        msg.debug('preExeStop time is {0}'.format(self._exeStop))
        
    def postExecute(self):
        msg.info('Postexecute for %s', self._name)
        
    def validate(self):
        self.setValStart()
        self._hasValidated = True        
        msg.info('Executor %s has no validation function - assuming all ok', self._name)
        self._isValidated = True
        self._errMsg = ''
        self._valStop = os.times()
        msg.debug('valStop time is {0}'.format(self._valStop))
    
    
    def doAll(self, input=set(), output=set()):
        self.preExecute(input, output)
        self.execute()
        self.postExecute()
        self.validate()
 
 
class logscanExecutor(transformExecutor):
    def __init__(self, name = 'Logscan'):
        super(logscanExecutor, self).__init__(name=name)
        self._errorMaskFiles = None
        self._logFileName = None
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        msg.info('Preexecute for %s', self._name)
        if 'logfile' in self.conf.argdict:
            self._logFileName = self.conf.argdict['logfile'].value
        
    def validate(self):
        self.setValStart()
        msg.info("Starting validation for {0}".format(self._name))
        if self._logFileName:
            
            if 'ignorePatterns' in self.conf.argdict:
                igPat = self.conf.argdict['ignorePatterns'].value
            else:
                igPat = []
            if 'ignoreFiles' in self.conf.argdict:
                ignorePatterns = trfValidation.ignorePatterns(files = self.conf.argdict['ignoreFiles'].value, extraSearch=igPat)
            elif self._errorMaskFiles is not None:
                ignorePatterns = trfValidation.ignorePatterns(files = self._errorMaskFiles, extraSearch=igPat)
            else:
                ignorePatterns = trfValidation.ignorePatterns(files = athenaExecutor._defaultIgnorePatternFile, extraSearch=igPat)
            
            # Now actually scan my logfile
            msg.info('Scanning logfile {0} for errors'.format(self._logFileName))
            self._logScan = trfValidation.athenaLogFileReport(logfile = self._logFileName, ignoreList = ignorePatterns)
            worstError = self._logScan.worstError()
    
            # In general we add the error message to the exit message, but if it's too long then don't do
            # that and just say look in the jobReport
            if worstError['firstError']:
                if len(worstError['firstError']['message']) > athenaExecutor._exitMessageLimit:
                    if 'CoreDumpSvc' in worstError['firstError']['message']:
                        exitErrorMessage = "Core dump at line {0} (see jobReport for further details)".format(worstError['firstError']['firstLine'])
                    elif 'G4Exception' in worstError['firstError']['message']:
                        exitErrorMessage = "G4 exception at line {0} (see jobReport for further details)".format(worstError['firstError']['firstLine'])
                    else:
                        exitErrorMessage = "Long {0} message at line {1} (see jobReport for further details)".format(worstError['level'], worstError['firstError']['firstLine'])
                else:
                    exitErrorMessage = "Logfile error in {0}: \"{1}\"".format(self._logFileName, worstError['firstError']['message'])
            else:
                exitErrorMessage = "Error level {0} found (see athena logfile for details)".format(worstError['level'])
            
            # Very simple: if we get ERROR or worse, we're dead, except if ignoreErrors=True
            if worstError['nLevel'] == stdLogLevels['ERROR'] and ('ignoreErrors' in self.conf.argdict and self.conf.argdict['ignoreErrors'].value is True):
                msg.warning('Found ERRORs in the logfile, but ignoring this as ignoreErrors=True (see jobReport for details)')
            elif worstError['nLevel'] >= stdLogLevels['ERROR']:
                self._isValidated = False
                msg.error('Fatal error in athena logfile (level {0})'.format(worstError['level']))
                raise trfExceptions.TransformLogfileErrorException(trfExit.nameToCode('TRF_EXEC_LOGERROR'), 
                                                                       'Fatal error in athena logfile: "{0}"'.format(exitErrorMessage))            
 
        # Must be ok if we got here!
        msg.info('Executor {0} has validated successfully'.format(self.name))
        self._isValidated = True
        self._errMsg = ''        
 
        self._valStop = os.times()
        msg.debug('valStop time is {0}'.format(self._valStop))
 
 
class echoExecutor(transformExecutor):
    def __init__(self, name = 'Echo', trf = None):
        
        # We are only changing the default name here
        super(echoExecutor, self).__init__(name=name, trf=trf)
 
    
    def execute(self):
        self._exeStart = os.times()
        msg.debug('exeStart time is {0}'.format(self._exeStart))
        msg.info('Starting execution of %s', self._name)        
        msg.info('Transform argument dictionary now follows:')
        for k, v in self.conf.argdict.items():
            print("%s = %s" % (k, v))
        self._hasExecuted = True
        self._rc = 0
        self._errMsg = ''
        msg.info('%s executor returns %d', self._name, self._rc)
        self._exeStop = os.times()
        msg.debug('exeStop time is {0}'.format(self._exeStop))
 
 
class dummyExecutor(transformExecutor):
    def __init__(self, name = 'Dummy', trf = None, conf = None, inData = set(), outData = set()):
 
        # We are only changing the default name here
        super(dummyExecutor, self).__init__(name=name, trf=trf, conf=conf, inData=inData, outData=outData)
 
 
    def execute(self):
        self._exeStart = os.times()
        msg.debug('exeStart time is {0}'.format(self._exeStart))
        msg.info('Starting execution of %s', self._name)
        for type in self._outData:
            for k, v in self.conf.argdict.items():
                if type in k:
                    msg.info('Creating dummy output file: {0}'.format(self.conf.argdict[k].value[0]))
                    open(self.conf.argdict[k].value[0], 'a').close()
        self._hasExecuted = True
        self._rc = 0
        self._errMsg = ''
        msg.info('%s executor returns %d', self._name, self._rc)
        self._exeStop = os.times()
        msg.debug('exeStop time is {0}'.format(self._exeStop))
 
 
class scriptExecutor(transformExecutor):
    def __init__(self, name = 'Script', trf = None, conf = None, inData = set(), outData = set(), 
                 exe = None, exeArgs = None, memMonitor = True):
        # Name of the script we want to execute
        self._exe = exe
        
        # With arguments (currently this means paste in the corresponding _argdict entry)
        self._exeArgs = exeArgs
        
        super(scriptExecutor, self).__init__(name=name, trf=trf, conf=conf, inData=inData, outData=outData)
        
        self._extraMetadata.update({'script' : exe})
        
        # Decide if we echo script output to stdout
        self._echoOutput = False
 
        # Can either be written by base class or child   
        self._cmd = None
        
        self._memMonitor = memMonitor
 
    @property
    def exe(self):
        return self._exe
    
    @exe.setter
    def exe(self, value):
        self._exe = value
        self._extraMetadata['script'] = value
 
    @property
    def exeArgs(self):
        return self._exeArgs
    
    @exeArgs.setter
    def exeArgs(self, value):
        self._exeArgs = value
#        self._extraMetadata['scriptArgs'] = value
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        msg.debug('scriptExecutor: Preparing for execution of {0} with inputs {1} and outputs {2}'.format(self.name, input, output))
 
        self._input = input
        self._output = output
        
        
        if self._cmd is None:
            self._buildStandardCommand()
        msg.info('Will execute script as %s', self._cmd)
        
        # Define this here to have it for environment detection messages
        self._logFileName = "log.{0}".format(self._name)
        
        
        if 'TRF_ECHO' in os.environ:
            msg.info('TRF_ECHO envvar is set - enabling command echoing to stdout')
            self._echoOutput = True
        elif 'TRF_NOECHO' in os.environ:
            msg.info('TRF_NOECHO envvar is set - disabling command echoing to stdout')
            self._echoOutput = False
        # PS1 is for sh, bash; prompt is for tcsh and zsh
        elif isInteractiveEnv():
            msg.info('Interactive environment detected (stdio or stdout is a tty) - enabling command echoing to stdout')
            self._echoOutput = True
        elif 'TZHOME' in os.environ:
            msg.info('Tier-0 environment detected - enabling command echoing to stdout')
            self._echoOutput = True
        if self._echoOutput is False:
            msg.info('Batch/grid running - command outputs will not be echoed. Logs for {0} are in {1}'.format(self._name, self._logFileName))
 
        # Now setup special loggers for logging execution messages to stdout and file
        self._echologger = logging.getLogger(self._name)
        self._echologger.setLevel(logging.INFO)
        self._echologger.propagate = False
 
        encargs = {'encoding' : 'utf-8'}
        self._exeLogFile = logging.FileHandler(self._logFileName, mode='w', **encargs)
        self._exeLogFile.setFormatter(logging.Formatter('%(asctime)s %(message)s', datefmt='%H:%M:%S'))
        self._echologger.addHandler(self._exeLogFile)
        
        if self._echoOutput:
            self._echostream = logging.StreamHandler(_encoding_stream(sys.stdout))
            self._echostream.setFormatter(logging.Formatter('%(name)s %(asctime)s %(message)s', datefmt='%H:%M:%S'))
            self._echologger.addHandler(self._echostream)
 
    def _buildStandardCommand(self):
        if self._exe:
            self._cmd = [self.exe, ]
        else:
            raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC_SETUP_FAIL'), 
                                                            'No executor set in {0}'.format(self.__class__.__name__))
        for arg in self.exeArgs:
            if arg in self.conf.argdict:
                # If we have a list then add each element to our list, else just str() the argument value
                # Note if there are arguments which need more complex transformations then
                # consider introducing a special toExeArg() method.
                if isinstance(self.conf.argdict[arg].value, list):
                    self._cmd.extend([ str(v) for v in self.conf.argdict[arg].value])
                else:
                    self._cmd.append(str(self.conf.argdict[arg].value))
 
 
    def execute(self):
        self._hasExecuted = True
        msg.info('Starting execution of {0} ({1})'.format(self._name, self._cmd))
        
        self._exeStart = os.times()
        msg.debug('exeStart time is {0}'.format(self._exeStart))
        if ('execOnly' in self.conf.argdict and self.conf.argdict['execOnly'] is True):
            msg.info('execOnly flag is set - execution will now switch, replacing the transform')
            os.execvp(self._cmd[0], self._cmd)
 
        encargs = {'encoding' : 'utf8'}
        try:
            # if we are already inside a container, then
            # must map /srv of the nested container to /srv of the parent container:
            # https://twiki.atlas-canada.ca/bin/view/AtlasCanada/Containers#Mount_point_summary
            if self._alreadyInContainer and self._containerSetup is not None:
                msg.info("chdir /srv to launch a nested container for the substep")
                os.chdir("/srv")
            p = subprocess.Popen(self._cmd, shell = False, stdout = subprocess.PIPE, stderr = subprocess.STDOUT, bufsize = 1, **encargs)
            # go back to the original working directory immediately after to store prmon in it
            if self._alreadyInContainer and self._containerSetup is not None:
                msg.info("chdir {} after launching the nested container".format(self._workdir))
                os.chdir(self._workdir)
 
            if self._memMonitor:
                try:
                    self._memSummaryFile = 'prmon.summary.' + self._name + '.json'
                    self._memFullFile = 'prmon.full.' + self._name
                    memMonitorCommand = ['prmon', '--pid', str(p.pid), '--filename', 'prmon.full.' + self._name, 
                                         '--json-summary', self._memSummaryFile, '--log-filename', 'prmon.' + self._name + '.log',
                                         '--interval', '30']
                    mem_proc = subprocess.Popen(memMonitorCommand, shell = False, close_fds=True, **encargs)
                    # TODO - link mem.full.current to mem.full.SUBSTEP
                except Exception as e:
                    msg.warning('Failed to spawn memory monitor for {0}: {1}'.format(self._name, e))
                    self._memMonitor = False
            
            while p.poll() is None:
                line = p.stdout.readline()
                if line:
                    self._echologger.info(line.rstrip())
            # Hoover up remaining buffered output lines
            for line in p.stdout:
                self._echologger.info(line.rstrip())
    
            self._rc = p.returncode
            msg.info('%s executor returns %d', self._name, self._rc)
            self._exeStop = os.times()
            msg.debug('exeStop time is {0}'.format(self._exeStop))
        except OSError as e:
            errMsg = 'Execution of {0} failed and raised OSError: {1}'.format(self._cmd[0], e)
            msg.error(errMsg)
            raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC'), errMsg)
        finally:
            if self._memMonitor:
                try:
                    mem_proc.send_signal(signal.SIGUSR1)
                    countWait = 0
                    while (not mem_proc.poll()) and countWait < 10:
                        time.sleep(0.1)
                        countWait += 1
                except OSError:
                    pass
        
        
    def postExecute(self):
        if hasattr(self._exeLogFile, 'close'):
            self._exeLogFile.close()
        if self._memMonitor:
            try:
                memFile = open(self._memSummaryFile)
                self._memStats = json.load(memFile)
            except Exception as e:
                msg.warning('Failed to load JSON memory summmary file {0}: {1}'.format(self._memSummaryFile, e))
                self._memMonitor = False
                self._memStats = {}
 
 
    def validate(self):
        if self._valStart is None:
            self._valStart = os.times()
            msg.debug('valStart time is {0}'.format(self._valStart))
        self._hasValidated = True
        
        
        if self._rc == 0:
            msg.info('Executor {0} validated successfully (return code {1})'.format(self._name, self._rc))
            self._isValidated = True
            self._errMsg = ''
        else:
            # Want to learn as much as possible from the non-zero code
            # this is a bit hard in general, although one can do signals.
            # Probably need to be more specific per exe, i.e., athena non-zero codes
            self._isValidated = False
            if self._rc < 0:
                # Map return codes to what the shell gives (128 + SIGNUM)
                self._rc = 128 - self._rc
            if trfExit.codeToSignalname(self._rc) != "":
                self._errMsg = '{0} got a {1} signal (exit code {2})'.format(self._name, trfExit.codeToSignalname(self._rc), self._rc)
            else:
                self._errMsg = 'Non-zero return code from %s (%d)' % (self._name, self._rc)
            raise trfExceptions.TransformValidationException(trfExit.nameToCode('TRF_EXEC_FAIL'), self._errMsg)
 
        
        if 'checkEventCount' in self.conf.argdict and self.conf.argdict['checkEventCount'].returnMyValue(exe=self) is False:
            msg.info('Event counting for substep {0} is skipped'.format(self.name))
        else:
            if 'mpi' in self.conf.argdict and not mpi.mpiShouldValidate():
                msg.info('MPI mode -- skipping output event count check')
            else:
                checkcount=trfValidation.eventMatch(self)
                checkcount.decide()
                self._eventCount = checkcount.eventCount
                msg.info('Event counting for substep {0} passed'.format(self.name))
 
        self._valStop = os.times()
        msg.debug('valStop time is {0}'.format(self._valStop))
 
 
 
class athenaExecutor(scriptExecutor):
    _exitMessageLimit = 200 # Maximum error message length to report in the exitMsg
    _defaultIgnorePatternFile = ['atlas_error_mask.db']
    
    
    def __init__(self, name = 'athena', trf = None, conf = None, skeletonFile=None, skeletonCA=None, 
                 inData = set(), outData = set(), inputDataTypeCountCheck = None, exe = 'athena.py', exeArgs = ['athenaopts'], 
                 substep = None, inputEventTest = True, perfMonFile = None, tryDropAndReload = True, extraRunargs = {}, runtimeRunargs = {},
                 literalRunargs = [], dataArgs = [], checkEventCount = False, errorMaskFiles = None,
                 manualDataDictionary = None, memMonitor = True, disableMT = False, disableMP = False, onlyMP = False, onlyMT = False, onlyMPWithRunargs = None):
        
        self._substep = forceToAlphaNum(substep)
        self._inputEventTest = inputEventTest
        self._tryDropAndReload = tryDropAndReload
        self._extraRunargs = extraRunargs
        self._runtimeRunargs = runtimeRunargs
        self._literalRunargs = literalRunargs
        self._dataArgs = dataArgs
        self._errorMaskFiles = errorMaskFiles
        self._inputDataTypeCountCheck = inputDataTypeCountCheck
        self._disableMT = disableMT
        self._disableMP = disableMP
        self._onlyMP = onlyMP
        self._onlyMT = onlyMT
        self._onlyMPWithRunargs = onlyMPWithRunargs
        self._skeletonCA=skeletonCA
 
        if perfMonFile:
            self._perfMonFile = None
            msg.debug("Resource monitoring from PerfMon is now deprecated")
        
        # SkeletonFile can be None (disable) or a string or a list of strings - normalise it here
        if isinstance(skeletonFile, str):
            self._skeleton = [skeletonFile]
        else:
            self._skeleton = skeletonFile
            
        super(athenaExecutor, self).__init__(name=name, trf=trf, conf=conf, inData=inData, outData=outData, exe=exe, 
                                             exeArgs=exeArgs, memMonitor=memMonitor)
        
        # Add athena specific metadata
        self._extraMetadata.update({'substep': substep})
 
        # Setup JO templates
        if self._skeleton or self._skeletonCA:
            self._jobOptionsTemplate = JobOptionsTemplate(exe = self, version = '$Id: trfExe.py 792052 2017-01-13 13:36:51Z mavogel $')
        else:
            self._jobOptionsTemplate = None
 
    @property
    def inputDataTypeCountCheck(self):
        return self._inputDataTypeCountCheck
    
    @inputDataTypeCountCheck.setter
    def inputDataTypeCountCheck(self, value):
        self._inputDataTypeCountCheck = value
        
    @property
    def substep(self):
        return self._substep
 
    @property
    def disableMP(self):
        return self._disableMP
    
    @disableMP.setter
    def disableMP(self, value):
        self._disableMP = value
    
    @property
    def disableMT(self):
        return self._disableMT
    
    @disableMT.setter
    def disableMT(self, value):
        self._disableMT = value
 
    @property
    def onlyMP(self):
        return self._onlyMP
    
    @onlyMP.setter
    def onlyMP(self, value):
        self._onlyMP = value
        
    @property
    def onlyMT(self):
        return self._onlyMT
 
    @onlyMT.setter
    def onlyMT(self, value):
        self._onlyMT = value
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        msg.debug('Preparing for execution of {0} with inputs {1} and outputs {2}'.format(self.name, input, output))
        
        # Check we actually have events to process!
        inputEvents = 0
        dt = ""
        if self._inputDataTypeCountCheck is None:
            self._inputDataTypeCountCheck = input
        for dataType in self._inputDataTypeCountCheck:
            if self.conf.dataDictionary[dataType].nentries == 'UNDEFINED':
                continue
 
            thisInputEvents = self.conf.dataDictionary[dataType].nentries
            if thisInputEvents > inputEvents:
                inputEvents = thisInputEvents
                dt = dataType
 
        # Now take into account skipEvents and maxEvents
        if ('skipEvents' in self.conf.argdict and 
            self.conf.argdict['skipEvents'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor) is not None):
            mySkipEvents = self.conf.argdict['skipEvents'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor)
        else:
            mySkipEvents = 0
 
        if ('maxEvents' in self.conf.argdict and 
            self.conf.argdict['maxEvents'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor) is not None):
            myMaxEvents = self.conf.argdict['maxEvents'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor)
        else:
            myMaxEvents = -1
        
        # Any events to process...?
        if (self._inputEventTest and mySkipEvents > 0 and mySkipEvents >= inputEvents):
            raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_NOEVENTS'),
                                                           'No events to process: {0} (skipEvents) >= {1} (inputEvents of {2}'.format(mySkipEvents, inputEvents, dt))
        
        try:
            # Expected events to process
            if (myMaxEvents != -1):
                if (self.inData and next(iter(self.inData)) == 'inNULL'):
                    expectedEvents = myMaxEvents
                else:
                    expectedEvents = min(inputEvents-mySkipEvents, myMaxEvents)
            else:
                expectedEvents = inputEvents-mySkipEvents
        except TypeError:
            # catching type error from UNDEFINED inputEvents count
            msg.info('input event count is UNDEFINED, setting expectedEvents to 0')
            expectedEvents = 0
        
        
        OSSetupString = None
 
        # Extract the asetup string
        asetupString = None
        legacyThreadingRelease = False
        if 'asetup' in self.conf.argdict:
            asetupString = self.conf.argdict['asetup'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor)
            legacyThreadingRelease = asetupReleaseIsOlderThan(asetupString, 22)
        else:
            msg.info('Asetup report: {0}'.format(asetupReport()))
 
        if asetupString is not None:
            legacyOSRelease = asetupReleaseIsOlderThan(asetupString, 24)
            currentOS = os.environ['ALRB_USER_PLATFORM']
            if legacyOSRelease and "centos7" not in currentOS:
                OSSetupString = "centos7"
                msg.info('Legacy release required for the substep {}, will setup a container running {}'.format(self._substep, OSSetupString))
 
 
        # allow overriding the container OS using a flag
        if 'runInContainer' in self.conf.argdict:
            OSSetupString = self.conf.argdict['runInContainer'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor)
            msg.info('The step {} will be performed in a container running {}, as explicitly requested'.format(self._substep, OSSetupString))
            if OSSetupString is not None and asetupString is None:
                raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC_SETUP_FAIL'),
                                                                '--asetup must be used for the substep which requires --runInContainer')
 
        # Conditional MP based on runtime arguments
        if self._onlyMPWithRunargs:
            for k in self._onlyMPWithRunargs:
                if k in self.conf._argdict:
                    self._onlyMP = True
 
        # Check the consistency of parallel configuration: CLI flags + evnironment.
        if ((('multithreaded' in self.conf._argdict and self.conf._argdict['multithreaded'].value) or ('multiprocess' in self.conf._argdict and self.conf._argdict['multiprocess'].value)) and
            ('ATHENA_CORE_NUMBER' not in os.environ)):
            # At least one of the parallel command-line flags has been provided but ATHENA_CORE_NUMBER environment has not been set
            msg.warning('either --multithreaded or --multiprocess argument used but ATHENA_CORE_NUMBER environment not set. Athena will continue in Serial mode')
        else:
            # Try to detect AthenaMT mode, number of threads and number of concurrent events
            if not self._disableMT:
                self._athenaMT, self._athenaConcurrentEvents = detectAthenaMTThreads(self.conf.argdict, self.name, legacyThreadingRelease)
 
            # Try to detect AthenaMP mode and number of workers
            if not self._disableMP:
                self._athenaMP = detectAthenaMPProcs(self.conf.argdict, self.name, legacyThreadingRelease)
 
            # Check that we actually support MT
            if self._onlyMP and self._athenaMT > 0:
                msg.info("This configuration does not support MT, falling back to MP")
                if self._athenaMP == 0:
                    self._athenaMP = self._athenaMT
                self._athenaMT = 0
                self._athenaConcurrentEvents = 0
 
            # Check that we actually support MP
            if self._onlyMT and self._athenaMP > 0:
                    msg.info("This configuration does not support MP, using MT")
                    if self._athenaMT == 0:
                        self._athenaMT = self._athenaMP
                        self._athenaConcurrentEvents = self._athenaMP
                    self._athenaMP = 0
 
        # Small hack to detect cases where there are so few events that it's not worthwhile running in MP mode
        # which also avoids issues with zero sized files. Distinguish from the no-input case (e.g. evgen)
        if not self._disableMP and expectedEvents < self._athenaMP and not self._inData=={'inNULL'}:
            msg.info("Disabling AthenaMP as number of input events to process is too low ({0} events for {1} workers)".format(expectedEvents, self._athenaMP))
            self._disableMP = True
            self._athenaMP = 0
 
        # Handle executor steps
        if self.conf.totalExecutorSteps > 1:
            for dataType in output:
                if self.conf._dataDictionary[dataType].originalName:
                    self.conf._dataDictionary[dataType].value[0] = self.conf._dataDictionary[dataType].originalName
                else:
                    self.conf._dataDictionary[dataType].originalName = self.conf._dataDictionary[dataType].value[0]
                self.conf._dataDictionary[dataType].value[0] += "_{0}{1}".format(executorStepSuffix, self.conf.executorStep)
                msg.info("Updated athena output filename for {0} to {1}".format(dataType, self.conf._dataDictionary[dataType].value[0]))
 
        # And if this is (still) athenaMP, then set some options for workers and output file report
        if self._athenaMP > 0:
            self._athenaMPWorkerTopDir = 'athenaMP-workers-{0}-{1}'.format(self._name, self._substep)
            self._athenaMPFileReport = 'athenaMP-outputs-{0}-{1}'.format(self._name, self._substep)
            self._athenaMPEventOrdersFile = 'athenamp_eventorders.txt.{0}'.format(self._name)
            if 'athenaMPUseEventOrders' in self.conf.argdict and self.conf._argdict['athenaMPUseEventOrders'].value is True:
                self._athenaMPReadEventOrders = True
            else:
                self._athenaMPReadEventOrders = False          
            # Decide on scheduling
            if ('athenaMPStrategy' in self.conf.argdict and 
                (self.conf.argdict['athenaMPStrategy'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor) is not None)):
                self._athenaMPStrategy = self.conf.argdict['athenaMPStrategy'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor)
            else:
                self._athenaMPStrategy = 'SharedQueue'
 
            # See if we have options for the target output file size
            if 'athenaMPMergeTargetSize' in self.conf.argdict:
                for dataType in output:
                    if dataType in self.conf.argdict['athenaMPMergeTargetSize'].value:
                        self.conf._dataDictionary[dataType].mergeTargetSize = self.conf.argdict['athenaMPMergeTargetSize'].value[dataType] * 1000000 # Convert from MB to B
                        msg.info('Set target merge size for {0} to {1}'.format(dataType, self.conf._dataDictionary[dataType].mergeTargetSize))
                    else:
                        # Use a globbing strategy
                        matchedViaGlob = False
                        for mtsType, mtsSize in self.conf.argdict['athenaMPMergeTargetSize'].value.items():
                            if fnmatch(dataType, mtsType):
                                self.conf._dataDictionary[dataType].mergeTargetSize = mtsSize * 1000000 # Convert from MB to B
                                msg.info('Set target merge size for {0} to {1} from "{2}" glob'.format(dataType, self.conf._dataDictionary[dataType].mergeTargetSize, mtsType))
                                matchedViaGlob = True
                                break
                        if not matchedViaGlob and "ALL" in self.conf.argdict['athenaMPMergeTargetSize'].value:
                            self.conf._dataDictionary[dataType].mergeTargetSize = self.conf.argdict['athenaMPMergeTargetSize'].value["ALL"] * 1000000 # Convert from MB to B
                            msg.info('Set target merge size for {0} to {1} from "ALL" value'.format(dataType, self.conf._dataDictionary[dataType].mergeTargetSize))
         
            # For AthenaMP jobs we ensure that the athena outputs get the suffix _000
            # so that the mother process output file (if it exists) can be used directly
            # as soft linking can lead to problems in the PoolFileCatalog (see ATLASJT-317)
            for dataType in output:
                if self.conf.totalExecutorSteps <= 1:
                    self.conf._dataDictionary[dataType].originalName = self.conf._dataDictionary[dataType].value[0]
                if 'eventService' not in self.conf.argdict or 'eventService' in self.conf.argdict and self.conf.argdict['eventService'].value is False:
                    if 'sharedWriter' in self.conf.argdict and self.conf.argdict['sharedWriter'].value:
                        msg.info("SharedWriter: not updating athena output filename for {0}".format(dataType))
                    else:
                        self.conf._dataDictionary[dataType].value[0] += "_000"
                        msg.info("Updated athena output filename for {0} to {1}".format(dataType, self.conf._dataDictionary[dataType].value[0]))
        else:
            self._athenaMPWorkerTopDir = self._athenaMPFileReport = None
 
        
        if 'mpi' in self.conf.argdict:
            msg.info("Running in MPI mode")
            mpi.setupMPIConfig(output, self.conf.dataDictionary)
 
        
        if self._skeleton or self._skeletonCA:
            inputFiles = dict()
            for dataType in input:
                inputFiles[dataType] = self.conf.dataDictionary[dataType]
            outputFiles = dict()
            for dataType in output:
                outputFiles[dataType] = self.conf.dataDictionary[dataType]
 
            # See if we have any 'extra' file arguments
            nameForFiles = commonExecutorStepName(self._name)
            for dataType, dataArg in self.conf.dataDictionary.items():
                if isinstance(dataArg, list) and dataArg:
                    if self.conf.totalExecutorSteps <= 1:
                        raise ValueError('Multiple input arguments provided but only running one substep')
                    if self.conf.totalExecutorSteps != len(dataArg):
                        raise ValueError(f'{len(dataArg)} input arguments provided but running {self.conf.totalExecutorSteps} substeps')
 
                    if dataArg[self.conf.executorStep].io == 'input' and nameForFiles in dataArg[self.conf.executorStep].executor:
                        inputFiles[dataArg[self.conf.executorStep].subtype] = dataArg
                else:
                    if dataArg.io == 'input' and nameForFiles in dataArg.executor:
                        inputFiles[dataArg.subtype] = dataArg
 
            msg.debug('Input Files: {0}; Output Files: {1}'.format(inputFiles, outputFiles))
            
            # Get the list of top options files that will be passed to athena (=runargs file + all skeletons)
            self._topOptionsFiles = self._jobOptionsTemplate.getTopOptions(input = inputFiles, 
                                                                           output = outputFiles)
 
        
        if len(input) > 0:
            self._extraMetadata['inputs'] = list(input)
        if len(output) > 0:
            self._extraMetadata['outputs'] = list(output)
 
        
        dbrelease = dbsetup = None
        if 'DBRelease' in self.conf.argdict:
            dbrelease = self.conf.argdict['DBRelease'].returnMyValue(name=self._name, substep=self._substep, first=self.conf.firstExecutor)
            if path.islink(dbrelease):
                dbrelease = path.realpath(dbrelease)
            if dbrelease:
                # Classic tarball - filename format is DBRelease-X.Y.Z.tar.gz
                dbdMatch = re.match(r'DBRelease-([\d\.]+)\.tar\.gz', path.basename(dbrelease))
                if dbdMatch:
                    msg.debug('DBRelease setting {0} matches classic tarball file'.format(dbrelease))
                    if not os.access(dbrelease, os.R_OK):
                        msg.warning('Transform was given tarball DBRelease file {0}, but this is not there'.format(dbrelease))
                        msg.warning('I will now try to find DBRelease {0} in cvmfs'.format(dbdMatch.group(1)))
                        dbrelease = dbdMatch.group(1)
                        dbsetup = cvmfsDBReleaseCheck(dbrelease)
                    else:
                        # Check if the DBRelease is setup
                        msg.debug('Setting up {0} from {1}'.format(dbdMatch.group(1), dbrelease))
                        unpacked, dbsetup = unpackDBRelease(tarball=dbrelease, dbversion=dbdMatch.group(1))
                        if unpacked:
                            # Now run the setup.py script to customise the paths to the current location...
                            setupDBRelease(dbsetup)
                # For cvmfs we want just the X.Y.Z release string (and also support 'current')
                else:
                    dbsetup = cvmfsDBReleaseCheck(dbrelease)
        
        
        self._envUpdate = trfEnv.environmentUpdate()
        self._envUpdate.setStandardEnvironment(self.conf.argdict, name=self.name, substep=self.substep)
        self._prepAthenaCommandLine() 
        
                
        super(athenaExecutor, self).preExecute(input, output)
        
        # Now we always write a wrapper, because it's very convenient for re-running individual substeps
        # This will have asetup and/or DB release setups in it
        # Do this last in this preExecute as the _cmd needs to be finalised
        msg.info('Now writing wrapper for substep executor {0}'.format(self._name))
        self._writeAthenaWrapper(asetup=asetupString, dbsetup=dbsetup, ossetup=OSSetupString)
        msg.info('Athena will be executed in a subshell via {0}'.format(self._cmd))
        
                
    def postExecute(self):
        super(athenaExecutor, self).postExecute()
        # MPI merging
        if 'mpi' in self.conf.argdict:
            mpi.mergeOutputs()
 
        # Handle executor substeps
        if self.conf.totalExecutorSteps > 1:
            if self._athenaMP > 0:
                outputDataDictionary = dict([ (dataType, self.conf.dataDictionary[dataType]) for dataType in self._output ])
                athenaMPOutputHandler(self._athenaMPFileReport, self._athenaMPWorkerTopDir, outputDataDictionary, self._athenaMP, False, self.conf.argdict)
            if self.conf.executorStep == self.conf.totalExecutorSteps - 1:
                # first loop over datasets for the output
                for dataType in self._output:
                    newValue = []
                    if self._athenaMP > 0:
                        # assume the same number of workers all the time
                        for i in range(self.conf.totalExecutorSteps):
                            for v in self.conf.dataDictionary[dataType].value:
                                newValue.append(v.replace('_{0}{1}_'.format(executorStepSuffix, self.conf.executorStep),
                                                          '_{0}{1}_'.format(executorStepSuffix, i)))
                    else:
                        self.conf.dataDictionary[dataType].multipleOK = True
                        # just combine all executors
                        for i in range(self.conf.totalExecutorSteps):
                            newValue.append(self.conf.dataDictionary[dataType].originalName + '_{0}{1}'.format(executorStepSuffix, i))
                    self.conf.dataDictionary[dataType].value = newValue
 
                    # do the merging if needed
                    if self.conf.dataDictionary[dataType].io == "output" and len(self.conf.dataDictionary[dataType].value) > 1:
                        self._smartMerge(self.conf.dataDictionary[dataType])
 
        # If this was an athenaMP run then we need to update output files
        elif self._athenaMP > 0:
            outputDataDictionary = dict([ (dataType, self.conf.dataDictionary[dataType]) for dataType in self._output ])
            
            skipFileChecks=False
            if 'eventService' in self.conf.argdict and self.conf.argdict['eventService'].value:
                skipFileChecks=True
            athenaMPOutputHandler(self._athenaMPFileReport, self._athenaMPWorkerTopDir, outputDataDictionary, self._athenaMP, skipFileChecks, self.conf.argdict)
            for dataType in self._output:
                if self.conf.dataDictionary[dataType].io == "output" and len(self.conf.dataDictionary[dataType].value) > 1:
                    self._smartMerge(self.conf.dataDictionary[dataType])
        
        if 'TXT_JIVEXMLTGZ' in self.conf.dataDictionary:
            self._targzipJiveXML()
 
        # Summarise events passed the filter ISF_SimEventFilter from log.ReSim
        # This is a bit ugly to have such a specific feature here though
        # TODO
        # The best is to have a general approach so that user can extract useful info from log
        # Instead of hard coding a pattern, one idea could be that user provides a regExp pattern
        # in which the wanted variable is grouped by a name, then transforms could decode the pattern
        # and use it to extract required info and do the summation during log scan.
        if self._logFileName=='log.ReSim' and self.name=='ReSim':
            msg.info('scanning {0} for reporting events passed the filter ISF_SimEventFilter'.format(self._logFileName))
            self._resimevents = reportEventsPassedSimFilter(self._logFileName)
 
        # Remove intermediate input/output files of sub-steps
        # Delete only files with io="temporay" which are files with pattern "tmp*"
        # Some stubs like tmp.RDO_TRIG_000 created in AthenaMP mode or
        # tmp.HIST_ESD_INT, tmp.HIST_AOD_INT as input to DQHistogramMerge.py are not deleted
        # Enable if --deleteIntermediateOutputfiles is set
        if ('deleteIntermediateOutputfiles' in self.conf._argdict and self.conf._argdict['deleteIntermediateOutputfiles'].value):
          inputDataDictionary = dict([ (dataType, self.conf.dataDictionary[dataType]) for dataType in self._input ])
 
          for k, v in inputDataDictionary.items():
            if not v.io == 'temporary':
              continue
            for filename in v.value:
              if os.access(filename, os.R_OK) and not filename.startswith("/cvmfs"):
                msg.info("Removing intermediate {0} input file {1}".format(k, filename))
                # Check if symbolic link and delete also linked file
                if (os.path.realpath(filename) != filename):
                  targetpath = os.path.realpath(filename)
                os.unlink(filename)
                if (targetpath) and os.access(targetpath, os.R_OK):
                  os.unlink(targetpath)
 
 
    def validate(self):
        self.setValStart()
        self._hasValidated = True
        deferredException = None
        memLeakThreshold = 5000
        _hasMemLeak = False
 
        
        try:
            super(athenaExecutor, self).validate()
        except trfExceptions.TransformValidationException as e:
            # In this case we hold this exception until the logfile has been scanned
            msg.error('Validation of return code failed: {0!s}'.format(e))
            deferredException = e
 
        
        if self._memFullFile:
            msg.info('Analysing memory monitor output file {0} for possible memory leak'.format(self._memFullFile))
            self._memLeakResult = analytic().getFittedData(self._memFullFile)
            if self._memLeakResult:
                if self._memLeakResult['slope'] > memLeakThreshold:
                    _hasMemLeak = True
                    msg.warning('Possible memory leak; abnormally high values in memory monitor parameters (ignore this message if the job has finished successfully)')
            else:
                msg.warning('Failed to analyse the memory monitor file {0}'.format(self._memFullFile))
        else:
            msg.info('No memory monitor file to be analysed')
 
        # Logfile scan setup
        # Always use ignorePatterns from the command line
        # For patterns in files, pefer the command line first, then any special settings for
        # this executor, then fallback to the standard default (atlas_error_mask.db)
        if 'ignorePatterns' in self.conf.argdict:
            igPat = self.conf.argdict['ignorePatterns'].value
        else:
            igPat = []
        if 'ignoreFiles' in self.conf.argdict:
            ignorePatterns = trfValidation.ignorePatterns(files = self.conf.argdict['ignoreFiles'].value, extraSearch=igPat)
        elif self._errorMaskFiles is not None:
            ignorePatterns = trfValidation.ignorePatterns(files = self._errorMaskFiles, extraSearch=igPat)
        else:
            ignorePatterns = trfValidation.ignorePatterns(files = athenaExecutor._defaultIgnorePatternFile, extraSearch=igPat)
        
        # Now actually scan my logfile
        msg.info('Scanning logfile {0} for errors in substep {1}'.format(self._logFileName, self._substep))
        self._logScan = trfValidation.athenaLogFileReport(logfile=self._logFileName, substepName=self._substep,
                                                          ignoreList=ignorePatterns)
        worstError = self._logScan.worstError()
        eventLoopWarnings = self._logScan.eventLoopWarnings()
        self._dbMonitor = self._logScan.dbMonitor()
        
 
        # In general we add the error message to the exit message, but if it's too long then don't do
        # that and just say look in the jobReport
        if worstError['firstError']:
            if len(worstError['firstError']['message']) > athenaExecutor._exitMessageLimit:
                if 'CoreDumpSvc' in worstError['firstError']['message']:
                    exitErrorMessage = "Core dump at line {0} (see jobReport for further details)".format(worstError['firstError']['firstLine'])
                elif 'G4Exception' in worstError['firstError']['message']:
                    exitErrorMessage = "G4 exception at line {0} (see jobReport for further details)".format(worstError['firstError']['firstLine'])
                else:
                    exitErrorMessage = "Long {0} message at line {1} (see jobReport for further details)".format(worstError['level'], worstError['firstError']['firstLine'])
            else:
                exitErrorMessage = "Logfile error in {0}: \"{1}\"".format(self._logFileName, worstError['firstError']['message'])
        else:
            exitErrorMessage = "Error level {0} found (see athena logfile for details)".format(worstError['level'])
 
        # If we failed on the rc, then abort now
        if deferredException is not None:
            # Add any logfile information we have
            if worstError['nLevel'] >= stdLogLevels['ERROR']:
                deferredException.errMsg = deferredException.errMsg + "; {0}".format(exitErrorMessage)
            # Add the result of memory analysis
            if _hasMemLeak:
                deferredException.errMsg = deferredException.errMsg + "; Possible memory leak: 'pss' slope: {0} KB/s".format(self._memLeakResult['slope'])
            raise deferredException
        
        
        # Very simple: if we get ERROR or worse, we're dead, except if ignoreErrors=True
        if worstError['nLevel'] == stdLogLevels['ERROR'] and ('ignoreErrors' in self.conf.argdict and self.conf.argdict['ignoreErrors'].value is True):
            msg.warning('Found ERRORs in the logfile, but ignoring this as ignoreErrors=True (see jobReport for details)')
        elif worstError['nLevel'] >= stdLogLevels['ERROR']:
            self._isValidated = False
            msg.error('Fatal error in athena logfile (level {0})'.format(worstError['level']))
            # Add the result of memory analysis
            if _hasMemLeak:
                exitErrorMessage = exitErrorMessage + "; Possible memory leak: 'pss' slope: {0} KB/s".format(self._memLeakResult['slope'])
            raise trfExceptions.TransformLogfileErrorException(trfExit.nameToCode('TRF_EXEC_LOGERROR'), 
                                                                   'Fatal error in athena logfile: "{0}"'.format(exitErrorMessage))
 
        # Print event loop warnings
        if (len(eventLoopWarnings) > 0):
            msg.warning('Found WARNINGS in the event loop, as follows:')
            for element in eventLoopWarnings:
                msg.warning('{0} {1} ({2} instances)'.format(element['item']['service'],element['item']['message'],element['count']))
        
        # Must be ok if we got here!
        msg.info('Executor {0} has validated successfully'.format(self.name))
        self._isValidated = True
 
        self._valStop = os.times()
        msg.debug('valStop time is {0}'.format(self._valStop))
 
    
    def _isCAEnabled(self):
        # CA not present
        if 'CA' not in self.conf.argdict:
            # If there is no legacy skeleton, then we are running with CA
            if not self._skeleton:
                return True
            else:
                return False
 
        # CA present but None, all substeps running with CA
        if self.conf.argdict['CA'] is None:
            return True
 
        # CA enabled for a substep, running with CA
        if self.conf.argdict['CA'].returnMyValue(name=self.name, substep=self.substep) is True:
            return True
 
        return False
 
    
    def _prepAthenaCommandLine(self):
        
        if 'athena' in self.conf.argdict:
            self._exe = self.conf.argdict['athena'].value
        self._cmd = [self._exe]
        
        # Find options for the current substep. Name is prioritised (e.g. RAWtoALL) over alias (e.g. r2a). Last look for 'all'
        currentSubstep = None
        if 'athenaopts' in self.conf.argdict:
            currentName = commonExecutorStepName(self.name)
            if currentName in self.conf.argdict['athenaopts'].value:
                currentSubstep = currentName
                if self.substep in self.conf.argdict['athenaopts'].value:
                    msg.info('Athenaopts found for {0} and {1}, joining options. '
                             'Consider changing your configuration to use just the name or the alias of the substep.'
                             .format(currentSubstep, self.substep))
                    self.conf.argdict['athenaopts'].value[currentSubstep].extend(self.conf.argdict['athenaopts'].value[self.substep])
                    del self.conf.argdict['athenaopts'].value[self.substep]
                    msg.debug('Athenaopts: {0}'.format(self.conf.argdict['athenaopts'].value))
            elif self.substep in self.conf.argdict['athenaopts'].value:
                currentSubstep = self.substep
            elif 'all' in self.conf.argdict['athenaopts'].value:
                currentSubstep = 'all'
 
        # See if there's a preloadlibs and a request to update LD_PRELOAD for athena
        preLoadUpdated = dict()
        if 'LD_PRELOAD' in self._envUpdate._envdict:
            preLoadUpdated[currentSubstep] = False
            if 'athenaopts' in self.conf.argdict:
                if currentSubstep is not None:
                    for athArg in self.conf.argdict['athenaopts'].value[currentSubstep]:
                        # This code is pretty ugly as the athenaopts argument contains
                        # strings which are really key/value pairs
                        if athArg.startswith('--preloadlib'):
                            try:
                                i = self.conf.argdict['athenaopts'].value[currentSubstep].index(athArg)
                                v = athArg.split('=', 1)[1]
                                msg.info('Updating athena --preloadlib option for substep {1} with: {0}'.format(self._envUpdate.value('LD_PRELOAD'), self.name))
                                newPreloads = ":".join(set(v.split(":")) | set(self._envUpdate.value('LD_PRELOAD').split(":")))
                                self.conf.argdict['athenaopts']._value[currentSubstep][i] = '--preloadlib={0}'.format(newPreloads)
                            except Exception as e:
                                msg.warning('Failed to interpret athena option: {0} ({1})'.format(athArg, e))
                            preLoadUpdated[currentSubstep] = True
                        break
            if not preLoadUpdated[currentSubstep]:
                msg.info('Setting athena preloadlibs for substep {1} to: {0}'.format(self._envUpdate.value('LD_PRELOAD'), self.name))
                if 'athenaopts' in self.conf.argdict:
                    if currentSubstep is not None:
                        self.conf.argdict['athenaopts'].value[currentSubstep].append("--preloadlib={0}".format(self._envUpdate.value('LD_PRELOAD')))
                    else:
                        self.conf.argdict['athenaopts'].value['all'] = ["--preloadlib={0}".format(self._envUpdate.value('LD_PRELOAD'))]
                else:
                    self.conf.argdict['athenaopts'] = trfArgClasses.argSubstepList(["--preloadlib={0}".format(self._envUpdate.value('LD_PRELOAD'))])
 
        # Now update command line with the options we have (including any changes to preload)
        if 'athenaopts' in self.conf.argdict:
            if currentSubstep is None and "all" in self.conf.argdict['athenaopts'].value:
                self._cmd.extend(self.conf.argdict['athenaopts'].value['all'])
            elif currentSubstep in self.conf.argdict['athenaopts'].value:
                self._cmd.extend(self.conf.argdict['athenaopts'].value[currentSubstep])
        
        if currentSubstep is None:
            currentSubstep = 'all'
        
        if self._tryDropAndReload:
            if self._isCAEnabled():
                msg.info('ignoring "--drop-and-reload" for CA-based transforms, config cleaned up anyway')
            elif 'valgrind' in self.conf._argdict and self.conf._argdict['valgrind'].value is True:
                msg.info('Disabling "--drop-and-reload" because the job is configured to use Valgrind')
            elif 'athenaopts' in self.conf.argdict:
                athenaConfigRelatedOpts = ['--config-only','--drop-and-reload']
                # Note for athena options we split on '=' so that we properly get the option and not the whole "--option=value" string
                if currentSubstep in self.conf.argdict['athenaopts'].value:
                    conflictOpts = set(athenaConfigRelatedOpts).intersection(set([opt.split('=')[0] for opt in self.conf.argdict['athenaopts'].value[currentSubstep]]))
                    if len(conflictOpts) > 0:
                        msg.info('Not appending "--drop-and-reload" to athena command line because these options conflict: {0}'.format(list(conflictOpts)))
                    else:
                        msg.info('Appending "--drop-and-reload" to athena options')
                        self._cmd.append('--drop-and-reload')
                else:
                    msg.info('No Athenaopts for substep {0}, appending "--drop-and-reload" to athena options'.format(self.name))
                    self._cmd.append('--drop-and-reload')
            else:
                # This is the 'standard' case - so drop and reload should be ok
                msg.info('Appending "--drop-and-reload" to athena options')
                self._cmd.append('--drop-and-reload')
        else:
            msg.info('Skipping test for "--drop-and-reload" in this executor')
            
        if not self._isCAEnabled():  #For CA-jobs, threads and nproc set in runargs file
            # For AthenaMT apply --threads=N if threads have been configured via ATHENA_CORE_NUMBER + multithreaded
            if self._athenaMT > 0 and not self._disableMT:
                if not ('athenaopts' in self.conf.argdict and
                        any('--threads' in opt for opt in self.conf.argdict['athenaopts'].value[currentSubstep])):
                        self._cmd.append('--threads=%s' % str(self._athenaMT))
 
            # For AthenaMP apply --nprocs=N if threads have been configured via ATHENA_CORE_NUMBER + multiprocess
            if self._athenaMP > 0 and not self._disableMP:
                if not ('athenaopts' in self.conf.argdict and
                        any('--nprocs' in opt for opt in self.conf.argdict['athenaopts'].value[currentSubstep])):
                        self._cmd.append('--nprocs=%s' % str(self._athenaMP))
 
        # Add topoptions
        if self._skeleton or self._skeletonCA:
            self._cmd += self._topOptionsFiles
            msg.info('Updated script arguments with topoptions: %s', self._cmd)
 
 
    
    def _writeAthenaWrapper(
        self,
        asetup = None,
        dbsetup = None,
        ossetup = None
        ):
        self._originalCmd        = self._cmd
        self._asetup             = asetup
        self._dbsetup            = dbsetup
        self._containerSetup     = ossetup
        self._workdir            = os.getcwd()
        self._alreadyInContainer = self._workdir.startswith("/srv")
        self._wrapperFile        = 'runwrapper.{name}.sh'.format(name = self._name)
        self._setupFile          = 'setup.{name}.sh'.format(name = self._name)
 
        # Create a setupATLAS script
        setupATLAS = 'my_setupATLAS.sh'
        with open(setupATLAS, 'w') as f:
            print("#!/bin/bash", file=f)
            print("""
if [ -z $ATLAS_LOCAL_ROOT_BASE ]; then
  export ATLAS_LOCAL_ROOT_BASE=/cvmfs/atlas.cern.ch/repo/ATLASLocalRootBase
fi
source ${ATLAS_LOCAL_ROOT_BASE}/user/atlasLocalSetup.sh"""
                  , file=f)
        os.chmod(setupATLAS, 0o755)
 
        msg.debug(
            'Preparing wrapper file {wrapperFileName} with '
            'asetup={asetupStatus} and dbsetup={dbsetupStatus}'.format(
                wrapperFileName = self._wrapperFile,
                asetupStatus    = self._asetup,
                dbsetupStatus   = self._dbsetup
            )
        )
 
        container_cmd = None
        try:
            with open(self._wrapperFile, 'w') as wrapper:
                print('#!/bin/sh', file=wrapper)
                if self._containerSetup is not None:
                    container_cmd = [ os.path.abspath(setupATLAS),
                                     "-c",
                                     self._containerSetup,
                                     "--pwd",
                                     self._workdir,
                                     "-s",
                                     os.path.join('.', self._setupFile),
                                     "-r"]
                    print('echo "This wrapper is executed within a container! For a local re-run, do:"', file=wrapper)
                    print('echo " '+ " ".join(['setupATLAS'] + container_cmd[1:] + [path.join('.', self._wrapperFile)]) + '"', file=wrapper)
                    print('echo "N.B.: if launching a nested container, navigate to /srv before running the above command"',
                          file = wrapper)
                    print('echo "      and use --pwd workdir, where workdir is the transform running directory within /srv"',
                          file=wrapper)
                    print('echo', file=wrapper)
 
                if asetup:
                    wfile = wrapper
                    asetupFile = None
                    # if the substep is executed within a container, setup athena with a separate script
                    # e.g. setupATLAS -c el9 -s setup.RDOtoRDOTrigger.sh -r runwrapper.RDOtoRDOTrigger.sh ...
                    if self._containerSetup is not None:
                        asetupFile = open(self._setupFile, 'w')
                        wfile = asetupFile
                    print(f'source ./{setupATLAS} -q', file=wfile)
                    print(f'asetup {asetup}', file=wfile)
                    print('if [ ${?} != "0" ]; then exit 255; fi', file=wfile)
                if dbsetup:
                    dbroot = path.dirname(dbsetup)
                    dbversion = path.basename(dbroot)
                    print("# DBRelease setup", file=wrapper)
                    print('echo Setting up DBRelease {dbroot} environment'.format(dbroot = dbroot), file=wrapper)
                    print('export DBRELEASE={dbversion}'.format(dbversion = dbversion), file=wrapper)
                    print('export CORAL_AUTH_PATH={directory}'.format(directory = path.join(dbroot, 'XMLConfig')), file=wrapper)
                    print('export CORAL_DBLOOKUP_PATH={directory}'.format(directory = path.join(dbroot, 'XMLConfig')), file=wrapper)
                    print('export TNS_ADMIN={directory}'.format(directory = path.join(dbroot, 'oracle-admin')), file=wrapper)
                    print('DATAPATH={dbroot}:$DATAPATH'.format(dbroot = dbroot), file=wrapper)
                if self._disableMT:
                    print("# AthenaMT explicitly disabled for this executor", file=wrapper)
                if self._disableMP:
                    print("# AthenaMP explicitly disabled for this executor", file=wrapper)
                if self._envUpdate.len > 0:
                    for envSetting in  self._envUpdate.values:
                        if not envSetting.startswith('LD_PRELOAD'):
                            print("export", envSetting, file=wrapper)
                # If Valgrind is engaged, a serialised Athena configuration file
                # is generated for use with a subsequent run of Athena with
                # Valgrind.
                if 'valgrind' in self.conf._argdict and self.conf._argdict['valgrind'].value is True:
                    msg.info('Valgrind engaged')
                    # Define the file name of the serialised Athena
                    # configuration.
                    AthenaSerialisedConfigurationFile = "{name}Conf.pkl".format(
                        name = self._name
                    )
                    # Run Athena for generation of its serialised configuration.
                    print(' '.join(self._cmd), "--config-only={0}".format(AthenaSerialisedConfigurationFile), file=wrapper)
                    print('if [ $? != "0" ]; then exit 255; fi', file=wrapper)
                    # Generate a Valgrind command, suppressing or ussing default
                    # options as requested and extra options as requested.
                    if 'valgrindDefaultOpts' in self.conf._argdict:
                        defaultOptions = self.conf._argdict['valgrindDefaultOpts'].value
                    else:
                        defaultOptions = True
                    if 'valgrindExtraOpts' in self.conf._argdict:
                        extraOptionsList = self.conf._argdict['valgrindExtraOpts'].value
                    else:
                        extraOptionsList = None
                    msg.debug("requested Valgrind command basic options: {options}".format(options = defaultOptions))
                    msg.debug("requested Valgrind command extra options: {options}".format(options = extraOptionsList))
                    command = ValgrindCommand(
                        defaultOptions = defaultOptions,
                        extraOptionsList = extraOptionsList,
                        AthenaSerialisedConfigurationFile = \
                            AthenaSerialisedConfigurationFile
                    )
                    msg.debug("Valgrind command: {command}".format(command = command))
                    print(command, file=wrapper)
                # If VTune is engaged, a serialised Athena configuration file
                # is generated for use with a subsequent run of Athena with
                # VTune.
                elif 'vtune' in self.conf._argdict and self.conf._argdict['vtune'].value is True:
                    msg.info('VTune engaged')
                    # Define the file name of the serialised Athena
                    # configuration.
                    AthenaSerialisedConfigurationFile = "{name}Conf.pkl".format(
                        name = self._name
                    )
                    # Run Athena for generation of its serialised configuration.
                    print(' '.join(self._cmd), "--config-only={0}".format(AthenaSerialisedConfigurationFile), file=wrapper)
                    print('if [ $? != "0" ]; then exit 255; fi', file=wrapper)
                    # Generate a VTune command, suppressing or ussing default
                    # options as requested and extra options as requested.
                    if 'vtuneDefaultOpts' in self.conf._argdict:
                        defaultOptions = self.conf._argdict['vtuneDefaultOpts'].value
                    else:
                        defaultOptions = True
                    if 'vtuneExtraOpts' in self.conf._argdict:
                        extraOptionsList = self.conf._argdict['vtuneExtraOpts'].value
                    else:
                        extraOptionsList = None
                    msg.debug("requested VTune command basic options: {options}".format(options = defaultOptions))
                    msg.debug("requested VTune command extra options: {options}".format(options = extraOptionsList))
                    command = VTuneCommand(
                        defaultOptions = defaultOptions,
                        extraOptionsList = extraOptionsList,
                        AthenaSerialisedConfigurationFile = \
                            AthenaSerialisedConfigurationFile
                    )
                    msg.debug("VTune command: {command}".format(command = command))
                    print(command, file=wrapper)
                else:
                    msg.info('Valgrind/VTune not engaged')
                    # run Athena command
                    print(' '.join(self._cmd), file=wrapper)
            os.chmod(self._wrapperFile, 0o755)
        except OSError as e:
            errMsg = 'error writing athena wrapper {fileName}: {error}'.format(
                fileName = self._wrapperFile,
                error = e
            )
            msg.error(errMsg)
            raise trfExceptions.TransformExecutionException(
                trfExit.nameToCode('TRF_EXEC_SETUP_WRAPPER'),
                errMsg
            )
        self._cmd = [ path.join('.', self._wrapperFile) ]
        if self._containerSetup is not None:
            asetupFile.close()
            self._cmd = container_cmd + self._cmd
 
 
    
    def _smartMerge(self, fileArg):
        
        if 'selfMerge' not in dir(fileArg):
            msg.info('Files in {0} cannot merged (no selfMerge() method is implemented)'.format(fileArg.name))
            return
        
        if fileArg.mergeTargetSize == 0:
            msg.info('Files in {0} will not be merged as target size is set to 0'.format(fileArg.name))
            return
 
        
        mergeCandidates = [list()]
        currentMergeSize = 0
        for fname in fileArg.value:
            size = fileArg.getSingleMetadata(fname, 'file_size')
            if not isinstance(size, int):
                msg.warning('File size metadata for {0} was not correct, found type {1}. Aborting merge attempts.'.format(fileArg, type(size)))
                return
            # if there is no file in the job, then we must add it
            if len(mergeCandidates[-1]) == 0:
                msg.debug('Adding file {0} to current empty merge list'.format(fname))
                mergeCandidates[-1].append(fname)
                currentMergeSize += size
                continue
            # see if adding this file gets us closer to the target size (but always add if target size is negative)
            if fileArg.mergeTargetSize < 0 or math.fabs(currentMergeSize + size - fileArg.mergeTargetSize) < math.fabs(currentMergeSize - fileArg.mergeTargetSize):
                msg.debug('Adding file {0} to merge list {1} as it gets closer to the target size'.format(fname, mergeCandidates[-1]))
                mergeCandidates[-1].append(fname)
                currentMergeSize += size
                continue
            # close this merge list and start a new one
            msg.debug('Starting a new merge list with file {0}'.format(fname))
            mergeCandidates.append([fname])
            currentMergeSize = size
            
        msg.debug('First pass splitting will merge files in this way: {0}'.format(mergeCandidates))
        
        if len(mergeCandidates) == 1:
            # Merging to a single file, so use the original filename that the transform
            # was started with
            mergeNames = [fileArg.originalName]
        else:
            # Multiple merge targets, so we need a set of unique names
            counter = 0
            mergeNames = []
            for mergeGroup in mergeCandidates:
                # Note that the individual worker files get numbered with 3 digit padding,
                # so these non-padded merges should be fine
                mergeName = fileArg.originalName + '_{0}'.format(counter)
                while path.exists(mergeName):
                    counter += 1
                    mergeName = fileArg.originalName + '_{0}'.format(counter)
                mergeNames.append(mergeName)
                counter += 1
        # Now actually do the merges
        for targetName, mergeGroup, counter in zip(mergeNames, mergeCandidates, list(range(len(mergeNames)))):
            msg.info('Want to merge files {0} to {1}'.format(mergeGroup, targetName))
            if len(mergeGroup) <= 1:
                msg.info('Skip merging for single file')
            else:
                
                self._myMerger.append(fileArg.selfMerge(output=targetName, inputs=mergeGroup, counter=counter, argdict=self.conf.argdict))
 
 
    def _targzipJiveXML(self):
        #tgzipping JiveXML files
        targetTGZName = self.conf.dataDictionary['TXT_JIVEXMLTGZ'].value[0]
        if os.path.exists(targetTGZName):
            os.remove(targetTGZName)
 
        import tarfile
        fNameRE = re.compile(r"JiveXML\_\d+\_\d+.xml")
 
        # force gz compression
        tar = tarfile.open(targetTGZName, "w:gz")
        for fName in os.listdir('.'):
            matches = fNameRE.findall(fName)
            if len(matches) > 0:
                if fNameRE.findall(fName)[0] == fName:
                    msg.info('adding %s to %s', fName, targetTGZName)
                    tar.add(fName)
 
        tar.close()
        msg.info('JiveXML compression: %s has been written and closed.', targetTGZName)
 
 
 
class optionalAthenaExecutor(athenaExecutor):
 
    # Here we validate, but will suppress any errors
    def validate(self):
        self.setValStart()
        try:
            super(optionalAthenaExecutor, self).validate()
        except trfExceptions.TransformValidationException as e:
            # In this case we hold this exception until the logfile has been scanned
            msg.warning('Validation failed for {0}: {1}'.format(self._name, e))
            self._isValidated = False
            self._errMsg = e.errMsg
            self._rc = e.errCode
        self._valStop = os.times()
        msg.debug('valStop time is {0}'.format(self._valStop))
 
 
class POOLMergeExecutor(athenaExecutor):
    
    def __init__(self, name = 'hybridPOOLMerge', trf = None, conf = None, skeletonFile=None, skeletonCA='RecJobTransforms.MergePool_Skeleton',
                 inData = set(), outData = set(), exe = 'athena.py', exeArgs = ['athenaopts'], substep = None, inputEventTest = True,
                 perfMonFile = None, tryDropAndReload = True, extraRunargs = {},
                 manualDataDictionary = None, memMonitor = True):
        
        super(POOLMergeExecutor, self).__init__(name, trf=trf, conf=conf, skeletonFile=skeletonFile, skeletonCA=skeletonCA,
                                                      inData=inData, outData=outData, exe=exe, exeArgs=exeArgs, substep=substep,
                                                      inputEventTest=inputEventTest, perfMonFile=perfMonFile, 
                                                      tryDropAndReload=tryDropAndReload, extraRunargs=extraRunargs,
                                                      manualDataDictionary=manualDataDictionary, memMonitor=memMonitor)
    
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        super(POOLMergeExecutor, self).preExecute(input=input, output=output)
 
    
    def execute(self):
        # First call the parent executor, which will manage the athena execution for us
        super(POOLMergeExecutor, self).execute()
 
 
 
class reductionFrameworkExecutor(athenaExecutor):
    
    def preExecute(self, input=set(), output=set()):
        self.setPreExeStart()
        msg.debug('Preparing for execution of {0} with inputs {1} and outputs {2}'.format(self.name, input, output))
        if 'NTUP_PILEUP' not in output:
            # New derivation framework transform uses "formats"
            if 'formats' not in self.conf.argdict:
                raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_REDUCTION_CONFIG_ERROR'),
                                                                'No derivation configuration specified')
 
            if ('DAOD' not in output) and ('D2AOD' not in output):
                raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_REDUCTION_CONFIG_ERROR'),
                                                                'No base name for DAOD output')
 
            formatList = []
            if 'formats' in self.conf.argdict: formatList = self.conf.argdict['formats'].value        
            for reduction in formatList:
                if ('DAOD' in output):
                    dataType = 'DAOD_' + reduction
                    if 'augmentations' not in self.conf.argdict:
                        outputName = 'DAOD_' + reduction + '.' + self.conf.argdict['outputDAODFile'].value[0]
                    else:
                        for val in self.conf.argdict['augmentations'].value:
                            if reduction in val.split(':')[0]:
                                outputName = 'DAOD_' + val.split(':')[1] + '.' + self.conf.argdict['outputDAODFile'].value[0]
                                break
                        else:
                            outputName = 'DAOD_' + reduction + '.' + self.conf.argdict['outputDAODFile'].value[0]
 
                if ('D2AOD' in output):
                    dataType = 'D2AOD_' + reduction
                    outputName = 'D2AOD_' + reduction + '.' + self.conf.argdict['outputD2AODFile'].value[0]
 
                msg.info('Adding reduction output type {0}'.format(dataType))
                output.add(dataType)
                newReduction = trfArgClasses.argPOOLFile(outputName, io='output', runarg=True, type='AOD',
                                                     name=reduction)
                # References to _trf - can this be removed?
                self.conf.dataDictionary[dataType] = newReduction
            
            # Clean up the stub file from the executor input and the transform's data dictionary
            # (we don't remove the actual argFile instance)
            if ('DAOD' in output):
                output.remove('DAOD')
                del self.conf.dataDictionary['DAOD']
                del self.conf.argdict['outputDAODFile']
            if ('D2AOD' in output):
                output.remove('D2AOD')
                del self.conf.dataDictionary['D2AOD']
                del self.conf.argdict['outputD2AODFile']      
 
            msg.info('Data dictionary is now: {0}'.format(self.conf.dataDictionary))
            msg.info('Input/Output: {0}/{1}'.format(input, output))
        
        msg.info('Data dictionary is now: {0}'.format(self.conf.dataDictionary))
        msg.info('Input/Output: {0}/{1}'.format(input, output))
        super(reductionFrameworkExecutor, self).preExecute(input, output)
 
 
 
class DQMergeExecutor(scriptExecutor):
    def __init__(self, name='DQHistMerge', trf=None, conf=None, inData=set(['HIST_AOD', 'HIST_ESD']), outData=set(['HIST']),
                 exe='DQHistogramMerge.py', exeArgs = [], memMonitor = True):
        
        self._histMergeList = 'HISTMergeList.txt'
        
        super(DQMergeExecutor, self).__init__(name=name, trf=trf, conf=conf, inData=inData, outData=outData, exe=exe, 
                                              exeArgs=exeArgs, memMonitor=memMonitor)
 
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        msg.debug('Preparing for execution of {0} with inputs {1} and outputs {2}'.format(self.name, input, output))
 
        super(DQMergeExecutor, self).preExecute(input=input, output=output)
 
        # Write the list of files to be merged
        with open(self._histMergeList, 'w') as DQMergeFile:
            for dataType in input:
                for fname in self.conf.dataDictionary[dataType].value:
                    self.conf.dataDictionary[dataType]._getNumberOfEvents([fname])
                    print(fname, file=DQMergeFile)
            
        self._cmd.append(self._histMergeList)
        
        # Add the output file
        if len(output) != 1:
            raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC_SETUP_FAIL'),
                                                            'One (and only one) output file must be given to {0} (got {1})'.format(self.name, len(output)))
        outDataType = list(output)[0]
        self._cmd.append(self.conf.dataDictionary[outDataType].value[0])
        
        # Set the run_post_processing to True/False
        if (self.conf._argdict.get("run_post_processing",False)):
            self._cmd.append('True')
        else:
            self._cmd.append('False')
 
        if (self.conf._argdict.get("is_incremental_merge",False)):
            self._cmd.append('True')
        else:
            self._cmd.append('False')
 
        for k in ("excludeHist","excludeDir"):
            if k in self.conf._argdict:
                self._cmd.append("--{0}={1}".format(k,self.conf._argdict[k]))
 
 
    def validate(self):
        self.setValStart()
        super(DQMergeExecutor, self).validate()
 
        exitErrorMessage = ''
        # Base class validation successful, Now scan the logfile for missed errors.
        try:
            logScan = trfValidation.scriptLogFileReport(self._logFileName)
            worstError = logScan.worstError()
 
            # In general we add the error message to the exit message, but if it's too long then don't do
            # that and just say look in the jobReport
            if worstError['firstError']:
                if len(worstError['firstError']['message']) > logScan._msgLimit:
                    exitErrorMessage = "Long {0} message at line {1}" \
                                       " (see jobReport for further details)".format(worstError['level'],
                                        worstError['firstError']['firstLine'])
                else:
                    exitErrorMessage = "Logfile error in {0}: \"{1}\"".format(self._logFileName,
                                                                              worstError['firstError']['message'])
        except OSError as e:
            exitCode = trfExit.nameToCode('TRF_EXEC_LOGERROR')
            raise trfExceptions.TransformValidationException(exitCode,
                  'Exception raised while attempting to scan logfile {0}: {1}'.format(self._logFileName, e))
 
        if worstError['nLevel'] == stdLogLevels['ERROR'] and (
                'ignoreErrors' in self.conf.argdict and self.conf.argdict['ignoreErrors'].value is True):
            msg.warning('Found ERRORs in the logfile, but ignoring this as ignoreErrors=True (see jobReport for details)')
 
        elif worstError['nLevel'] >= stdLogLevels['ERROR']:
            self._isValidated = False
            msg.error('Fatal error in script logfile (level {0})'.format(worstError['level']))
            exitCode = trfExit.nameToCode('TRF_EXEC_LOGERROR')
            raise trfExceptions.TransformLogfileErrorException(exitCode, 'Fatal error in script logfile: "{0}"'.format(exitErrorMessage))
 
        # Must be ok if we got here!
        msg.info('Executor {0} has validated successfully'.format(self.name))
        self._isValidated = True
 
        self._valStop = os.times()
        msg.debug('valStop time is {0}'.format(self._valStop))
 
 
class DQMPostProcessExecutor(scriptExecutor):
    def __init__(self, name='DQMPostProcess', trf=None, conf=None, inData=set(['HIST']), outData=set(['HIST']),
                 exe='DQM_Tier0Wrapper_tf.py', exeArgs = [], memMonitor = True):
        
        self._histMergeList = 'HISTMergeList.txt'
        
        super(DQMPostProcessExecutor, self).__init__(name=name, trf=trf, conf=conf, inData=inData, outData=outData, exe=exe, 
                                              exeArgs=exeArgs, memMonitor=memMonitor)
 
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        msg.debug('Preparing for execution of {0} with inputs {1} and outputs {2}'.format(self.name, input, output))
 
        super(DQMPostProcessExecutor, self).preExecute(input=input, output=output)
 
        #build input file list (typically only one):
        dsName=self.conf.argdict["inputHISTFile"].dataset 
        inputList=[]
        for dataType in input:
            for fname in self.conf.dataDictionary[dataType].value:
                #if no dataset name is give, guess it from file name
                if not dsName: dsName=".".join(fname.split('.')[0:4])
                inputList.append("#".join([dsName,fname]))
 
 
        if len(output) != 1:
            raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC_SETUP_FAIL'),
                                                            'One (and only one) output file must be given to {0} (got {1})'.format(self.name, len(output)))
        outDataType = list(output)[0]
        #build argument json:
        #T0 uses keys: 'allowCOOLUpload', 'doWebDisplay', 'incrementalMode', 'inputHistFiles', 'mergeParams', 'outputHistFile', 'postProcessing', 'reverseFileOrder', 'servers'
        #more keys: runNumber, streamName,projectTag,filepaths,productionMode,skipMerge
        wrapperParams={"inputHistFiles" : inputList,
                       "outputHistFile" : dsName+"#"+self.conf.dataDictionary[outDataType].value[0],
                       "incrementalMode": "True" if self.conf._argdict.get("is_incremental_merge",False) else "False",
                       "postProcessing" : "True" if self.conf._argdict.get("run_post_processing",False) else "False",
                       "doWebDisplay"   : "True" if  self.conf._argdict.get("doWebDisplay",False) else "False",
                       "allowCOOLUpload": "True" if  self.conf._argdict.get("allowCOOLUpload",False) else "False",
                       "mergeParams" : "",
 
        }
        if "servers" in self.conf._argdict:
            wrapperParams["server"]=self.conf._argdict["servers"]
 
        for k in ("excludeHist","excludeDir"):
            if k in self.conf._argdict:
                wrapperParams["mergeParams"]+=(" --{0}={1}".format(k,self.conf._argdict[k]))
 
 
        with open("args.json", "w") as f:
            json.dump(wrapperParams, f)
 
        self._cmd.append("--argJSON=args.json")
        
        
 
    def validate(self):
        self.setValStart()
        super(DQMPostProcessExecutor, self).validate()
 
        exitErrorMessage = ''
        # Base class validation successful, Now scan the logfile for missed errors.
        try:
            logScan = trfValidation.scriptLogFileReport(self._logFileName)
            worstError = logScan.worstError()
 
            # In general we add the error message to the exit message, but if it's too long then don't do
            # that and just say look in the jobReport
            if worstError['firstError']:
                if len(worstError['firstError']['message']) > logScan._msgLimit:
                    exitErrorMessage = "Long {0} message at line {1}" \
                                       " (see jobReport for further details)".format(worstError['level'],
                                        worstError['firstError']['firstLine'])
                else:
                    exitErrorMessage = "Logfile error in {0}: \"{1}\"".format(self._logFileName,
                                                                              worstError['firstError']['message'])
        except OSError as e:
            exitCode = trfExit.nameToCode('TRF_EXEC_LOGERROR')
            raise trfExceptions.TransformValidationException(exitCode,
                  'Exception raised while attempting to scan logfile {0}: {1}'.format(self._logFileName, e))
 
        if worstError['nLevel'] == stdLogLevels['ERROR'] and (
                'ignoreErrors' in self.conf.argdict and self.conf.argdict['ignoreErrors'].value is True):
            msg.warning('Found ERRORs in the logfile, but ignoring this as ignoreErrors=True (see jobReport for details)')
 
        elif worstError['nLevel'] >= stdLogLevels['ERROR']:
            self._isValidated = False
            msg.error('Fatal error in script logfile (level {0})'.format(worstError['level']))
            exitCode = trfExit.nameToCode('TRF_EXEC_LOGERROR')
            raise trfExceptions.TransformLogfileErrorException(exitCode, 'Fatal error in script logfile: "{0}"'.format(exitErrorMessage))
 
        # Must be ok if we got here!
        msg.info('Executor {0} has validated successfully'.format(self.name))
        self._isValidated = True
 
        self._valStop = os.times()
        msg.debug('valStop time is {0}'.format(self._valStop))
 
class NTUPMergeExecutor(scriptExecutor):
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        msg.debug('[NTUP] Preparing for execution of {0} with inputs {1} and outputs {2}'.format(self.name, input, output))
 
        # Basic command, and allow overwrite of the output file
        if self._exe is None:
            self._exe = 'hadd'
        self._cmd = [self._exe, "-f"]
        
        
        # Add the output file
        if len(output) != 1:
            raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC_SETUP_FAIL'),
                                                            'One (and only one) output file must be given to {0} (got {1})'.format(self.name, len(output)))
        outDataType = list(output)[0]
        self._cmd.append(self.conf.dataDictionary[outDataType].value[0])
        # Add to be merged to the cmd chain
        for dataType in input:
            self._cmd.extend(self.conf.dataDictionary[dataType].value)
 
        super(NTUPMergeExecutor, self).preExecute(input=input, output=output)
 
 
 
class bsMergeExecutor(scriptExecutor):
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        self._inputBS = list(input)[0]
        self._outputBS = list(output)[0]
        self._maskedFiles = []
        self._useStubFile = False
        if 'maskEmptyInputs' in self.conf.argdict and self.conf.argdict['maskEmptyInputs'].value is True:
            eventfullFiles = []
            for fname in self.conf.dataDictionary[self._inputBS].value:
                nEvents = self.conf.dataDictionary[self._inputBS].getSingleMetadata(fname, 'nentries')
                msg.debug('Found {0} events in file {1}'.format(nEvents, fname))
                if isinstance(nEvents, int) and nEvents > 0:
                    eventfullFiles.append(fname)
            self._maskedFiles = list(set(self.conf.dataDictionary[self._inputBS].value) - set(eventfullFiles))
            if len(self._maskedFiles) > 0:
                msg.info('The following input files are masked because they have 0 events: {0}'.format(' '.join(self._maskedFiles)))
                if len(eventfullFiles) == 0:
                    if 'emptyStubFile' in self.conf.argdict and path.exists(self.conf.argdict['emptyStubFile'].value):
                        self._useStubFile = True
                        msg.info("All input files are empty - will use stub file {0} as output".format(self.conf.argdict['emptyStubFile'].value))
                    else:
                        raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_INPUT_FILE_ERROR'), 
                                                                        'All input files had zero events - aborting BS merge')            
        
        # Write the list of input files to a text file, so that testMergedFiles can swallow it
        self._mergeBSFileList = '{0}.list'.format(self._name)
        self._mergeBSLogfile = '{0}.out'.format(self._name)
        try:
            with open(self._mergeBSFileList, 'w') as BSFileList:
                for fname in self.conf.dataDictionary[self._inputBS].value:
                    if fname not in self._maskedFiles:
                        print(fname, file=BSFileList)
        except OSError as e:
            errMsg = 'Got an error when writing list of BS files to {0}: {1}'.format(self._mergeBSFileList, e)
            msg.error(errMsg)
            raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC_SETUP_WRAPPER'), errMsg)
        
        # Hope that we were given a correct filename...
        self._outputFilename = self.conf.dataDictionary[self._outputBS].value[0]
        if self._outputFilename.endswith('._0001.data'):
            self._doRename = False
            self._outputFilename = self._outputFilename.split('._0001.data')[0]    
        elif self.conf.argdict['allowRename'].value is True:
            # OK, non-fatal, we go for a renaming
            msg.info('Output filename does not end in "._0001.data" will proceed, but be aware that the internal filename metadata will be wrong')
            self._doRename = True
        else:
            # No rename allowed, so we are dead...
            errmsg = 'Output filename for outputBS_MRGFile must end in "._0001.data" or infile metadata will be wrong'
            raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_OUTPUT_FILE_ERROR'), errmsg)
        
        # Set the correct command for execution
        self._cmd = [self._exe, self._mergeBSFileList, '0', self._outputFilename] 
        
        super(bsMergeExecutor, self).preExecute(input=input, output=output)
    
    def execute(self):
        if self._useStubFile:
            # Need to fake execution!
            self._exeStart = os.times()
            msg.debug('exeStart time is {0}'.format(self._exeStart))
            msg.info("Using stub file for empty BS output - execution is fake")
            if self._outputFilename != self.conf.argdict['emptyStubFile'].value:
                os.rename(self.conf.argdict['emptyStubFile'].value, self._outputFilename)            
            self._memMonitor = False
            self._hasExecuted = True
            self._rc = 0
            self._exeStop = os.times()
            msg.debug('exeStop time is {0}'.format(self._exeStop))
        else:
            super(bsMergeExecutor, self).execute()
        
    def postExecute(self):
        if self._useStubFile:
            pass
        elif self._doRename:
            self._expectedOutput = self._outputFilename + '._0001.data'
            msg.info('Renaming {0} to {1}'.format(self._expectedOutput, self.conf.dataDictionary[self._outputBS].value[0]))
            try:
                os.rename(self._outputFilename + '._0001.data', self.conf.dataDictionary[self._outputBS].value[0])
            except OSError as e:
                raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_OUTPUT_FILE_ERROR'), 
                                                                'Exception raised when renaming {0} to {1}: {2}'.format(self._outputFilename, self.conf.dataDictionary[self._outputBS].value[0], e))
        super(bsMergeExecutor, self).postExecute()
 
 
 
class archiveExecutor(scriptExecutor):
 
    def preExecute(self, input = set(), output = set()):
        self.setPreExeStart()
        self._memMonitor = False
 
        #archiving
        if self._exe == 'zip':
            if 'outputArchFile' not in self.conf.argdict:
                raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_ARG_MISSING'), 'Missing output file name')
 
            self._cmd = ['python']
            try:
                with open('zip_wrapper.py', 'w') as zip_wrapper:
                    print("import zipfile, os, shutil", file=zip_wrapper)
                    if os.path.exists(self.conf.argdict['outputArchFile'].value[0]):
                        #appending input file(s) to existing archive
                        print("zf = zipfile.ZipFile('{}', mode='a', allowZip64=True)".format(self.conf.argdict['outputArchFile'].value[0]), file=zip_wrapper)
                    else:
                        #creating new archive
                        print("zf = zipfile.ZipFile('{}', mode='w', allowZip64=True)".format(self.conf.argdict['outputArchFile'].value[0]), file=zip_wrapper)
                    print("for f in {}:".format(self.conf.argdict['inputDataFile'].value), file=zip_wrapper)
                    #This module gives false positives (as of python 3.7.0). Will also check the name for ".zip"
                    #print >> zip_wrapper, "    if zipfile.is_zipfile(f):"
                    print("    if zipfile.is_zipfile(f) and '.zip' in f:", file=zip_wrapper)
                    print("        archive = zipfile.ZipFile(f, mode='r')", file=zip_wrapper)
                    print("        print 'Extracting input zip file {0} to temporary directory {1}'.format(f,'tmp')", file=zip_wrapper)
                    print("        archive.extractall('tmp')", file=zip_wrapper)
                    print("        archive.close()", file=zip_wrapper)
                    # remove stuff as soon as it is saved to output in order to save disk space at worker node
                    print("        if os.access(f, os.F_OK):", file=zip_wrapper)
                    print("            print 'Removing input zip file {}'.format(f)", file=zip_wrapper)
                    print("            os.unlink(f)", file=zip_wrapper)
                    print("        if os.path.isdir('tmp'):", file=zip_wrapper)
                    print("            for root, dirs, files in os.walk('tmp'):", file=zip_wrapper)
                    print("                for name in files:", file=zip_wrapper)
                    print("                    print 'Zipping {}'.format(name)", file=zip_wrapper)
                    print("                    zf.write(os.path.join(root, name), name, compress_type=zipfile.ZIP_STORED)", file=zip_wrapper)
                    print("            shutil.rmtree('tmp')", file=zip_wrapper)
                    print("    else:", file=zip_wrapper)
                    print("        print 'Zipping {}'.format(os.path.basename(f))", file=zip_wrapper)
                    print("        zf.write(f, arcname=os.path.basename(f), compress_type=zipfile.ZIP_STORED)", file=zip_wrapper)
                    print("        if os.access(f, os.F_OK):", file=zip_wrapper)
                    print("            print 'Removing input file {}'.format(f)", file=zip_wrapper)
                    print("            os.unlink(f)", file=zip_wrapper)
                    print("zf.close()", file=zip_wrapper)
                os.chmod('zip_wrapper.py', 0o755)
            except OSError as e:
                errMsg = 'error writing zip wrapper {fileName}: {error}'.format(fileName = 'zip_wrapper.py',
                    error = e
                )
                msg.error(errMsg)
                raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC_SETUP_WRAPPER'),
                    errMsg
                )
            self._cmd.append('zip_wrapper.py')
 
        #unarchiving
        elif self._exe == 'unarchive':
            import zipfile
            for infile in self.conf.argdict['inputArchFile'].value:
                 if not zipfile.is_zipfile(infile):
                     raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_INPUT_FILE_ERROR'),
                                                                     'An input file is not a zip archive - aborting unpacking')
            self._cmd = ['python']
            try:
                with open('unarchive_wrapper.py', 'w') as unarchive_wrapper:
                    print("import zipfile", file=unarchive_wrapper)
                    print("for f in {}:".format(self.conf.argdict['inputArchFile'].value), file=unarchive_wrapper)
                    print("     archive = zipfile.ZipFile(f, mode='r')", file=unarchive_wrapper)
                    print("     path = '{}'".format(self.conf.argdict['path']), file=unarchive_wrapper)
                    print("     print 'Extracting archive {0} to {1}'.format(f,path)", file=unarchive_wrapper)
                    print("     archive.extractall(path)", file=unarchive_wrapper)
                    print("     archive.close()", file=unarchive_wrapper)
                os.chmod('unarchive_wrapper.py', 0o755)
            except OSError as e:
                errMsg = 'error writing unarchive wrapper {fileName}: {error}'.format(fileName = 'unarchive_wrapper.py',
                    error = e
                )
                msg.error(errMsg)
                raise trfExceptions.TransformExecutionException(trfExit.nameToCode('TRF_EXEC_SETUP_WRAPPER'),
                    errMsg
                )
            self._cmd.append('unarchive_wrapper.py')
        super(archiveExecutor, self).preExecute(input=input, output=output)