view SMART/Java/Python/ncList/Benchmark.py @ 68:85e80c21b1f7 draft

Uploaded
author m-zytnicki
date Mon, 16 Nov 2015 12:00:32 -0500
parents 94ab73e8a190
children
line wrap: on
line source

import os, os.path, random, glob, subprocess, threading, time, resource
from optparse import OptionParser
from SMART.Java.Python.misc.Progress import *
from SMART.Java.Python.getRandomRegions import RandomRegionsGenerator
from commons.core.writer.TranscriptWriter import TranscriptWriter
from SMART.Java.Python.structure.Transcript import Transcript
from commons.core.parsing.GffParser import GffParser

#TYPES = ("bin", "has", "seg", "fj", "nc", "new")
TYPES = ("new", )

class RunCmd(threading.Thread):
	def __init__(self, cmd, out, err, time, memory):
		threading.Thread.__init__(self)
		self._cmd    = cmd
		self._out    = out
		self._err    = err
		self._time   = time
		self._memory = memory
		self._id	 = os.getpid()
		self._mem    = 0.0
		self._outputFileName = "tmp_%d.out" % (self._id)

	def run(self):
		self._p = subprocess.Popen(self._cmd, stdout = self._out, stderr = self._err, shell = True)
		#self._p.wait()

	def _runShellCommand(self, command):
		p = subprocess.call("%s > %s" % (command, self._outputFileName), shell=True)
		handle = open(self._outputFileName)
		data   = [line.split() for line in handle.readlines()[1:] if line]
		handle.close()
		os.remove(self._outputFileName)
		return data

	def _getPid(self):
		self._pid	  = None
		cpt           = 1
		while True:
			commandsFound = []
			for line in self._runShellCommand("ps -o pid,cmd"):
				if line[1:] == self._cmd.split(" "):
					self._pid = int(line[0])
				commandsFound.append(" ".join(line[1:]))
			if self._pid != None:
				return True
			time.sleep(1)
			if cpt % 100 == 0:
				print "pid of '%s' not found after %d seconds. Found: %s" % (self._cmd, cpt, " --- ".join(commandsFound))
			cpt += 1
			if cpt > 300:
				return False

	def _fetchMemory(self):
		lines = self._runShellCommand("ps u -p %d" % (self._pid))
		for line in lines:
			self._mem = max(self._mem, float(line[3]))
			return self._mem >= self._memory
		#print "Cannot find the memory of the current PID (%d) in: %s" % (self._pid, " --- ".join([" ".join(line) for line in lines]))
		return False
	
	def getMemory(self):
		return self._mem

	def _abort(self):
		try:
			self._p.terminate()
		except Exception:
			pass
		self._killSubThreads()
	
	def _killSubThreads(self):
		for line in self._runShellCommand("ps --ppid %d -o pid" % (self._pid)):
			self._runShellCommand("kill %s" % (line[0]))
		self._runShellCommand("kill %s" % (self._pid))

	def go(self):
		startTime = time.time()
		self.run()
		#self.start()
		while not self._getPid():
			#self.start()
			self.run()
		while True:
			if self._time != None and time.time() - startTime > self._time:
				print "\nCommand '%s' did not finish in time. Aborting it." % (self._cmd)
				self._abort()
				break
			if self._memory != None and self._fetchMemory():
				print "\nCommand '%s' required too much memory (%f). Aborting it." % (self._cmd, self._mem)
				self._abort()
				break
			#self.join(0.1)
			time.sleep(0.1)
			#if not self.isAlive():
			if self._p.poll() != None:
				return True
		return False


class DataStructure(object):
	def __init__(self):
		self._structure = {}

	def addData(self, data):
		if data._nbRefs not in self._structure:
			self._structure[data._nbRefs] = {}
		if data._nbQueries not in self._structure[data._nbRefs]:
			self._structure[data._nbRefs][data._nbQueries] = {}
		if data._genomeSize not in self._structure[data._nbRefs][data._nbQueries]:
			self._structure[data._nbRefs][data._nbQueries][data._genomeSize] = {}
		if data._type not in self._structure[data._nbRefs][data._nbQueries][data._genomeSize]:
			self._structure[data._nbRefs][data._nbQueries][data._genomeSize][data._type] = []
		self._structure[data._nbRefs][data._nbQueries][data._genomeSize][data._type].append(data._group)

	def export(self):
		outputString = "#refs\t#queries\tgenome size\ttype\t# written\t# overlaps\tbuild t.\trun t.\tmem\n"
		for nbRefs in sorted(self._structure.keys()):
			for nbQueries in sorted(self._structure[nbRefs].keys()):
				for genomeSize in sorted(self._structure[nbRefs][nbQueries].keys()):
					for type in TYPES:
						if type not in self._structure[nbRefs][nbQueries][genomeSize]:
							outputString += "NA\tNA\tNA\t%s\tNA\tNA\tNA\tNA\tNA\tNA\n" % (type)
						else:
							for group in self._structure[nbRefs][nbQueries][genomeSize][type]:
								outputString += "%d\t%d\t%d\t%s\t%d\t%d\t%f\t%f\t%f\n" % (nbRefs, nbQueries, genomeSize, type, group._nbOutputs, group._nbOverlaps, group._buildTime, group._runTime, group._mem)
		return outputString


class Data(object):
	def __init__(self, type, buildTime, runTime, mem, nbRefs, nbQueries, nbOutputs, nbOverlaps, genomeSize):
		self._type	   = type
		self._nbRefs	 = nbRefs
		self._nbQueries  = nbQueries
		self._genomeSize = genomeSize
		self._group	  = Group(nbOutputs, nbOverlaps, buildTime, runTime, mem)

	def checkConsistency(self, data):
		return self._group.checkConsistency(data._group)


class Group(object):
	def __init__(self, nbOutputs, nbOverlaps, buildTime, runTime, mem):
		self._buildTime  = buildTime
		self._runTime	 = runTime
		self._mem    	 = mem
		self._nbOutputs	 = nbOutputs
		self._nbOverlaps = nbOverlaps

	def checkConsistency(self, group):
		if (self._buildTime == "NA" or group._buildTime == "NA"):
			return True
		return (self._nbOutputs == group._nbOutputs and self._nbOverlaps == group._nbOverlaps)
		

class Benchmark(object):

	def __init__(self, verbosity = 1):
		self._verbosity = verbosity
		self._checkEnvironmentVariable()
		self._toolName	= {"bin":  os.path.join(os.environ["SMARTPATH"], "ncList", "FindOverlapsWithSeveralIntervalsBin.py"), \
							"has": os.path.join(os.environ["SMARTPATH"], "ncList", "FindOverlapsWithSeveralIntervalsHashBin.py"), \
							"seg": os.path.join(os.environ["SMARTPATH"], "ncList", "FindOverlapsWithSeveralIntervalsBinSegment.py"), \
							"fj":  os.path.join(os.environ["SMARTPATH"], "ncList", "FindOverlapsFJoin.py"), \
							"nc":  os.path.join(os.environ["SMARTPATH"], "ncList", "FindOverlapsWithSeveralIntervals.py"), \
							"new": os.path.join(os.environ["SMARTPATH"],		   "FindOverlapsOptim.py")}
		self._structure = DataStructure()
		self._pid	   = os.getpid()
		self._count		= 0
		self._time	  = None
		self._memory	= None

	def _checkEnvironmentVariable(self):
		if "SMARTPATH" not in os.environ:
			raise Exception("'SMARTPATH' is not set. Please set it to '<installation-direction>/S-mart/Java/Python'.")

	def _createTmpFileName(self, name, extension):
		self._count += 1
		return "tmp_%d_%d_%s.%s" % (self._pid, self._count, name, extension)

	def _dumpAndReturn(self, fileName, exception):
		handle = open(fileName)
		print "Error in parsing file '%s':" % (fileName)
		for line in handle:
			print line.strip()
		print "Command is: '%s'" % (self._command)
		raise exception

	def setNbReferences(self, nbReferences):
		self._nbReferences = nbReferences

	def setNbQueries(self, nbQueries):
		self._nbQueries = nbQueries

	def setGenomeSizes(self, nbGenomeSizes):
		self._nbGenomeSizes = nbGenomeSizes

	def setNbReplicates(self, nbReplicates):
		self._nbReplicates = nbReplicates

	def setChromosomeName(self, chromosome):
		self._chromosomeName = chromosome

	def setSizes(self, minSize, maxSize):
		self._minSize = minSize
		self._maxSize = maxSize

	def setOutputFileName(self, fileName):
		self._outputFileName = fileName

	def setLimits(self, time, memory):
		self._time   = time
		self._memory = memory

	def _generateIntervals(self, nbElements, genomeSize):
		fileName = self._createTmpFileName("intervals", "gff3")
		iRR	  = RandomRegionsGenerator(0)
		iRR.setMinSize(self._minSize)
		iRR.setMaxSize(self._maxSize)
		iRR.setGenomeSize(genomeSize)
		iRR.setChromosomeName(self._chromosomeName)
		iRR.setStrands(False)
		iRR.setNumber(nbElements)
		iRR.setOutputFile(fileName)
		iRR.run()
		return fileName

	def _startTool(self, type, refFileName, queryFileName, nbReferences, nbQueries, genomeSize):
		outputFileName = self._createTmpFileName("output", "gff3")
		outFileName	   = self._createTmpFileName("out", "out")
		errFileName	   = self._createTmpFileName("err", "err")
		outHandle	   = open(outFileName, "w")
		errHandle	   = open(errFileName, "w")
		self._command  = "python %s -i %s -f gff3 -j %s -g gff3 -o %s -v 3" % (self._toolName[type], queryFileName, refFileName, outputFileName)
		thread		   = RunCmd(self._command, outHandle, errHandle, self._time, self._memory)
		over		   = thread.go()
		self._mem      = thread.getMemory()
		if os.path.exists(outputFileName):
			os.remove(outputFileName)
		outHandle.close()
		errHandle.close()
		errData = open(errFileName).readlines()
		if errData:
			print "Error output: \n%s\n" % ("\n".join(errData))
		if not over:
			errHandle = open(errFileName, "r")
			error = errHandle.readlines()
			errHandle.close()
			if error:
				for line in error:
					print line.strip()
				print "Previous process failed"
		os.remove(errFileName)
		if not over:
			return False
		return outFileName

	def _parseTrace(self, type, fileName, genomeSize):
		handle	  = open(fileName)
		buildTime = 0
		try:
			for line in handle:
				line = line.strip()
				if "time spent" in line:
					buildTime += float(line.split()[-1][:-1])
				elif "done" in line:
					buildTime += float(line.split("(")[1][:-2])
				elif "# queries" in line:
					nbQueries = int(line.split()[-1])
				elif "# refs" in line:
					nbRefs = int(line.split()[-1])
				elif "# written" in line:
					nbOutputs	= int(line.split()[2])
					nbOverlaps = int(line.split()[3][1:])
				elif "time" in line:
					runTime = float(line.split()[-1][:-1])
		except Exception, e:
			handle.close()
			self._dumpAndReturn(fileName, e)
		handle.close()
		try:
			return Data(type, buildTime, runTime, self._mem, nbRefs, nbQueries, nbOutputs, nbOverlaps, genomeSize)
		except Exception, e:
			handle.close()
			self._dumpAndReturn(fileName, e)
	
	def _cleanTmpFiles(self, really = False):
		files = glob.glob("tmp_%d_*.pkl" % (self._pid)) + glob.glob("tmp_%d_*.bin" % (self._pid))
		if really: 
			files += glob.glob("tmp_%d_*.gff3" % (self._pid)) + glob.glob("tmp_%d_*.out" % (self._pid))
		for fileName in files:
			os.remove(fileName)
		
	def run(self):
		progress = Progress(len(self._nbReferences) * len(self._nbQueries) * len(self._nbGenomeSizes) * self._nbReplicates, "Processing", self._verbosity)
		for nbReferences in self._nbReferences:
			for queriesRatio in self._nbQueries:
				nbQueries = int(nbReferences * queriesRatio)
				for genomeSizeRatio in self._nbGenomeSizes:
					genomeSize = int(nbReferences * genomeSizeRatio)
					for replicate in range(self._nbReplicates):
						refFileName	  = self._generateIntervals(nbReferences, genomeSize)
						queryFileName = self._generateIntervals(nbQueries,	genomeSize)
						data		  = {}
						for type in TYPES:
							fileName   = self._startTool(type, refFileName, queryFileName, nbReferences, nbQueries, genomeSize)
							if not fileName:
								data[type] = Data(type, "NA", "NA", "NA", nbReferences, nbQueries, "NA", "NA", genomeSize)
							else:
								data[type] = self._parseTrace(type, fileName, genomeSize)
								self._structure.addData(data[type])
								os.remove(fileName)
							self._cleanTmpFiles()
						self._cleanTmpFiles(True)
						firstType = TYPES[0]
						for type in TYPES[1:]:
							if not data[firstType].checkConsistency(data[type]):
								raise Exception("Outputs are not consistent.\n  # outputs: %d vs %d.\n  # overlaps: %d vs %d.\n  %s: %f + %f; %s: %f + %f.\n  Files are %s and %s." % (data[firstType]._group._nbOutputs, data[type]._group._nbOutputs, data[firstType]._group._nbOverlaps, data[type]._group._nbOverlaps, firstType, data[firstType]._group._buildTime, data[firstType]._group._runTime, data[firstType]._group._mem, type, data[type]._group._buildTime, data[type]._group._runTime, data[type]._group._mem, refFileName, queryFileName))
						for fileName in (queryFileName, refFileName):
							if os.path.exists(fileName):
								os.remove(fileName)
						progress.inc()
		progress.done()
		handle = open(self._outputFileName, "w")
		handle.write(self._structure.export())
		handle.close()



if __name__ == "__main__":
	
	description = "Benchmark v1.0.2: Compare NC-List with other tools. Only work under Linux. [Category: Other]"
	parser = OptionParser(description = description)
	parser.add_option("-r", "--nbReferences", dest="nbReferences",   action="store", default=None,   type="string", help="number of references (list of integers separated by commas) [compulsory] [format: string]")
	parser.add_option("-q", "--nbQueries",    dest="nbQueries",		 action="store", default=None,   type="string", help="number of queries as a factor of the number of references (list of floats separated by commas) [compulsory] [format: string]")
	parser.add_option("-R", "--nbReplicates", dest="nbReplicates",   action="store", default=None,   type="int",	help="number of replicates [compulsory] [format: int]")
	parser.add_option("-s", "--genomeSizes",  dest="genomeSizes",    action="store", default=None,   type="string", help="genome size as a factor of the number of references (list of floats separated by commas) [compulsory] [format: string]")
	parser.add_option("-c", "--chromosome",   dest="chromosome",	 action="store", default="chr1", type="string", help="name of the chromosome [default: chr1] [format: string]")
	parser.add_option("-z", "--minSize",      dest="minSize",        action="store", default=None,   type="int",	help="minimum size of the reads [compulsory] [format: int]")
	parser.add_option("-Z", "--maxSize",      dest="maxSize",        action="store", default=None,   type="int",	help="maximum size of the reads [compulsory] [format: int]")
	parser.add_option("-o", "--output",       dest="outputFileName", action="store",				 type="string", help="output file [compulsory] [format: output file in TXT format]")
	parser.add_option("-t", "--time",         dest="time",           action="store", default=None,   type="int",	help="maximum time to wait (in seconds) [default: None] [format: int]")
	parser.add_option("-m", "--memory",	      dest="memory",		 action="store", default=None,   type="float",	help="maximum memory usage (in %) [default: None] [format: float]")
	parser.add_option("-v", "--verbosity",    dest="verbosity",      action="store", default=1,	     type="int",	help="trace level [format: int]")
	(options, args) = parser.parse_args()

	benchmark = Benchmark(options.verbosity)
	benchmark.setNbReferences(map(int, options.nbReferences.split(",")))
	benchmark.setNbQueries(map(float, options.nbQueries.split(",")))
	benchmark.setGenomeSizes(map(float, options.genomeSizes.split(",")))
	benchmark.setNbReplicates(options.nbReplicates)
	benchmark.setChromosomeName(options.chromosome)
	benchmark.setSizes(options.minSize, options.maxSize)
	benchmark.setLimits(options.time, options.memory)
	benchmark.setOutputFileName(options.outputFileName)
	benchmark.run()