def loadProject(self, mcProject, fileStore = None):
self.inGraph = NX.DiGraph()
globTree = mcProject.mcTree
self.maxParallelSubtrees = None
leafEvents = [globTree.getName(i) for i in globTree.getLeaves()]
expMap = None
if fileStore:
expMap = dict()
for name in mcProject.expIDMap:
expMap[name] = fileStore.readGlobalFile(mcProject.expIDMap[name])
else:
expMap = mcProject.expMap
for name, expPath in list(expMap.items()):
exp = ExperimentWrapper(ET.parse(expPath).getroot())
tree = exp.getTree()
self.inGraph.add_node(name)
# Go through the species tree and add the correct
# dependencies (i.e. to the outgroups and the ingroups,
# but not to the other nodes that are just there because
# they are needed to form the correct paths).
for node in tree.postOrderTraversal():
nodeName = tree.getName(node)
# we don't add edges for leaves (in the global tree)
# as they are input sequences and do not form dependencies
# (it would be clever to maybe do the same with existing
# references when --overwrite is not specified but for now
# we just do the leaves)
if nodeName not in leafEvents and tree.isLeaf(node):
self.inGraph.add_edge(name, nodeName)
if fileStore:
configFile = fileStore.readGlobalFile(exp.getConfigID())
else:
# hack from running from cactus-prepare
configFile = exp.getConfigPath()
configElem = ET.parse(configFile).getroot()
conf = ConfigWrapper(configElem)
# load max parellel subtrees from the node's config
if self.maxParallelSubtrees is None:
self.maxParallelSubtrees = conf.getMaxParallelSubtrees()
else:
assert self.maxParallelSubtrees == conf.getMaxParallelSubtrees()
assert NX.is_directed_acyclic_graph(self.inGraph)
def loadProject(self, mcProject):
self.inGraph = NX.DiGraph()
globTree = mcProject.mcTree
self.maxParallelSubtrees = None
leafEvents = [globTree.getName(i) for i in globTree.getLeaves()]
for name, expPath in mcProject.expMap.items():
exp = ExperimentWrapper(ET.parse(expPath).getroot())
tree = exp.getTree()
self.inGraph.add_node(name)
# Go through the species tree and add the correct
# dependencies (i.e. to the outgroups and the ingroups,
# but not to the other nodes that are just there because
# they are needed to form the correct paths).
for node in tree.postOrderTraversal():
nodeName = tree.getName(node)
if not tree.isLeaf(node) and nodeName not in exp.getOutgroupEvents():
# This node is just an internal node added while
# creating the induced tree from the species
# tree. None of the sequence is used, so skip it.
continue
assert tree.hasParent(node)
if nodeName not in exp.getOutgroupEvents() and tree.getName(tree.getParent(node)) != name:
# This leaf isn't an ingroup or an outgroup, it was
# just added to make the species tree
# binary. (Hopefully this will be unnecessary in
# the future.)
continue
# we don't add edges for leaves (in the global tree)
# as they are input sequences and do not form dependencies
# (it would be clever to maybe do the same with existing
# references when --overwrite is not specified but for now
# we just do the leaves)
if nodeName not in leafEvents:
self.inGraph.add_edge(name, nodeName)
configElem = ET.parse(exp.getConfig()).getroot()
conf = ConfigWrapper(configElem)
# load max parellel subtrees from the node's config
if self.maxParallelSubtrees is None:
self.maxParallelSubtrees = conf.getMaxParallelSubtrees()
else:
assert self.maxParallelSubtrees == conf.getMaxParallelSubtrees()
assert NX.is_directed_acyclic_graph(self.inGraph)
def loadProject(self, mcProject, fileStore = None):
self.inGraph = NX.DiGraph()
globTree = mcProject.mcTree
self.maxParallelSubtrees = None
leafEvents = [globTree.getName(i) for i in globTree.getLeaves()]
expMap = None
if fileStore:
expMap = dict()
for name in mcProject.expIDMap:
expMap[name] = fileStore.readGlobalFile(mcProject.expIDMap[name])
else:
expMap = mcProject.expMap
for name, expPath in expMap.items():
exp = ExperimentWrapper(ET.parse(expPath).getroot())
tree = exp.getTree()
self.inGraph.add_node(name)
# Go through the species tree and add the correct
# dependencies (i.e. to the outgroups and the ingroups,
# but not to the other nodes that are just there because
# they are needed to form the correct paths).
for node in tree.postOrderTraversal():
nodeName = tree.getName(node)
# we don't add edges for leaves (in the global tree)
# as they are input sequences and do not form dependencies
# (it would be clever to maybe do the same with existing
# references when --overwrite is not specified but for now
# we just do the leaves)
if nodeName not in leafEvents and nodeName in exp.getSequenceMap():
self.inGraph.add_edge(name, nodeName)
configFile = fileStore.readGlobalFile(exp.getConfigID())
configElem = ET.parse(configFile).getroot()
conf = ConfigWrapper(configElem)
# load max parellel subtrees from the node's config
if self.maxParallelSubtrees is None:
self.maxParallelSubtrees = conf.getMaxParallelSubtrees()
else:
assert self.maxParallelSubtrees == conf.getMaxParallelSubtrees()
assert NX.is_directed_acyclic_graph(self.inGraph)
class ProjectWrapper:
alignmentDirName = 'progressiveAlignment'
def __init__(self, options, seqFile, workingDir):
self.options = options
self.seqFile = seqFile
self.workingDir = workingDir
self.configWrapper = None
self.expWrapper = None
self.processConfig()
self.processExperiment()
def processConfig(self):
# read in the default right out of cactus
if self.options.configFile is not None:
configPath = self.options.configFile
else:
dir = cactusRootPath()
configPath = os.path.join(dir,
"cactus_progressive_config.xml")
configXml = ET.parse(configPath).getroot()
self.configWrapper = ConfigWrapper(configXml)
# here we can go through the options and apply some to the config
self.configWrapper.setBuildHal(True)
self.configWrapper.setBuildFasta(True)
if self.options.outputMaf is not None:
self.configWrapper.setBuildMaf(True)
self.configWrapper.setJoinMaf(True)
# pre-emptively turn down maxParallelSubtree for singleMachine
# mode if not enough threads are provided to support it. Probably
# need to do something for other ?combined? batch systems?
if self.options.batchSystem == 'singleMachine' and \
self.options.database == 'kyoto_tycoon':
if int(self.options.maxThreads) < \
self.configWrapper.getMaxParallelSubtrees() * 3:
self.configWrapper.setMaxParallelSubtrees(
max(1, int(self.options.maxThreads) / 3))
# this is a little hack to effectively toggle back to the
# non-progressive version of cactus (as published in Gen. Res. 2011)
# from the high-level interface.
if self.options.legacy is True:
self.configWrapper.setSubtreeSize(sys.maxint)
def processExperiment(self):
expXml = self.seqFile.toXMLElement()
#create the cactus disk
cdElem = ET.SubElement(expXml, "cactus_disk")
database = self.options.database
assert database == "kyoto_tycoon" or database == "tokyo_cabinet"
confElem = ET.SubElement(cdElem, "st_kv_database_conf")
confElem.attrib["type"] = database
dbElem = ET.SubElement(confElem, database)
self.expWrapper = ExperimentWrapper(expXml)
if self.options.database == "kyoto_tycoon":
self.expWrapper.setDbPort(str(self.options.ktPort))
if self.options.ktHost is not None:
self.expWrapper.setDbHost(self.options.ktHost)
if self.options.ktType == 'memory':
self.expWrapper.setDbInMemory(True)
self.expWrapper.setDbSnapshot(False)
elif self.options.ktType == 'snapshot':
self.expWrapper.setDbInMemory(True)
self.expWrapper.setDbSnapshot(True)
else:
assert self.options.ktType == 'disk'
self.expWrapper.setDbInMemory(False)
self.expWrapper.setDbSnapshot(False)
# sonlib doesn't allow for spaces in attributes in the db conf
# which renders this options useless
# if self.options.ktOpts is not None:
# self.expWrapper.setDbServerOptions(self.options.ktOpts)
if self.options.ktCreateTuning is not None:
self.expWrapper.setDbCreateTuningOptions(
self.options.ktCreateTuning)
if self.options.ktOpenTuning is not None:
self.expWrapper.setDbReadTuningOptions(
self.options.ktOpenTuning)
#set the sequence output directory
outSeqDir = os.path.join(self.workingDir, "sequenceData")
if os.path.exists(outSeqDir) and self.options.overwrite:
system("rm -rf %s" % outSeqDir)
if not os.path.exists(outSeqDir):
system("mkdir %s" % outSeqDir)
self.expWrapper.setOutputSequenceDir(os.path.join(self.workingDir,
"sequenceData"))
def writeXml(self):
assert os.path.isdir(self.workingDir)
configPath = os.path.abspath(
os.path.join(self.workingDir, "config.xml"))
expPath = os.path.abspath(
os.path.join(self.workingDir, "expTemplate.xml"))
self.expWrapper.setConfigPath(configPath)
self.configWrapper.writeXML(configPath)
self.expWrapper.writeXML(expPath)
projPath = os.path.join(self.workingDir,
ProjectWrapper.alignmentDirName)
if os.path.exists(projPath) and self.options.overwrite:
system("rm -rf %s" % projPath)
if self.options.outputMaf is True:
fixNames=1
else:
fixNames=0
if os.path.exists(projPath):
if not self.isSameAsExisting(expPath, projPath, fixNames):
raise RuntimeError("Existing project %s not " % projPath+
"compatible with current input. Please "
"erase the working directory or rerun "
"with the --overwrite option to start "
"from scratch.")
else:
logPath = os.path.join(self.workingDir, 'cactus.log')
logFile = open(logPath, "a")
logFile.write("\nContinuing existing alignment. Use "
"--overwrite or erase the working directory to "
"force restart from scratch.\n")
logFile.close()
else:
cmd = "cactus_createMultiCactusProject.py %s %s --fixNames=%d" % (
expPath, projPath, fixNames)
if len(self.seqFile.outgroups) > 0:
cmd += " --outgroupNames " + ",".join(self.seqFile.outgroups)
if self.options.rootOutgroupDist:
cmd += " --rootOutgroupDist %f" % self.options.rootOutgroupDist
cmd += " --rootOutgroupPath %s" % self.options.rootOutgroupPath
system(cmd)
# create a project in a dummy directory. check if the
# project xml is the same as the current project.
# we do this to see if we should start fresh or try to
# work with the existing project when the overwrite flag is off
def isSameAsExisting(self, expPath, projPath, fixNames):
if not os.path.exists(projPath):
return False
oldPath = os.path.dirname(projPath + "/")
tempPath = "%s_temp" % oldPath
if os.path.exists(tempPath):
system("rm -rf %s" % tempPath)
cmd = "cactus_createMultiCactusProject.py %s %s --fixNames=%d" % (
expPath, tempPath, fixNames)
if len(self.seqFile.outgroups) > 0:
cmd += " --outgroupNames " + ",".join(self.seqFile.outgroups)
if self.options.rootOutgroupDist:
cmd += " --rootOutgroupDist %f" % self.options.rootOutgroupDist
cmd += " --rootOutgroupPath %s" % self.options.rootOutgroupPath
system(cmd)
projFilePathNew = os.path.join(tempPath,'%s_temp_project.xml' %
self.alignmentDirName)
projFilePathOld = os.path.join(oldPath, '%s_project.xml' %
self.alignmentDirName)
newFile = [line for line in open(projFilePathNew, "r")]
oldFile = [line for line in open(projFilePathOld, "r")]
areSame = True
if len(newFile) != len(oldFile):
areSame = False
for newLine, oldLine in zip(newFile, oldFile):
if newLine.replace(tempPath, oldPath) != oldLine:
areSame = False
system("rm -rf %s" % tempPath)
return areSame
def run(self):
logger.info("Progressive Up: " + self.event)
# open up the experiment
# note that we copy the path into the options here
self.options.experimentFile = self.project.expMap[self.event]
expXml = ET.parse(self.options.experimentFile).getroot()
experiment = ExperimentWrapper(expXml)
configXml = ET.parse(experiment.getConfigPath()).getroot()
configWrapper = ConfigWrapper(configXml)
# need at least 3 processes for every event when using ktserver:
# 1 proc to run jobs, 1 proc to run server, 1 proc to run 2ndary server
if experiment.getDbType() == "kyoto_tycoon":
maxParallel = min(len(self.project.expMap),
configWrapper.getMaxParallelSubtrees())
if self.options.batchSystem == "singleMachine":
if int(self.options.maxThreads) < maxParallel * 3:
raise RuntimeError("At least %d threads are required (only %d were specified) to handle up to %d events using kyoto tycoon. Either increase the number of threads using the --maxThreads option or decrease the number of parallel jobs (currently %d) by adjusting max_parallel_subtrees in the config file" % (maxParallel * 3, self.options.maxThreads, maxParallel, configWrapper.getMaxParallelSubtrees()))
else:
if int(self.options.maxCpus) < maxParallel * 3:
raise RuntimeError("At least %d concurrent cpus are required to handle up to %d events using kyoto tycoon. Either increase the number of cpus using the --maxCpus option or decrease the number of parallel jobs (currently %d) by adjusting max_parallel_subtrees in the config file" % (maxParallel * 3, maxParallel, configWrapper.getMaxParallelSubtrees()))
# take union of command line options and config options for hal and reference
if self.options.buildReference == False:
refNode = findRequiredNode(configXml, "reference")
self.options.buildReference = getOptionalAttrib(refNode, "buildReference", bool, False)
halNode = findRequiredNode(configXml, "hal")
if self.options.buildHal == False:
self.options.buildHal = getOptionalAttrib(halNode, "buildHal", bool, False)
if self.options.buildFasta == False:
self.options.buildFasta = getOptionalAttrib(halNode, "buildFasta", bool, False)
# get parameters that cactus_workflow stuff wants
workFlowArgs = CactusWorkflowArguments(self.options)
# copy over the options so we don't trail them around
workFlowArgs.buildReference = self.options.buildReference
workFlowArgs.buildHal = self.options.buildHal
workFlowArgs.buildFasta = self.options.buildFasta
workFlowArgs.overwrite = self.options.overwrite
workFlowArgs.globalLeafEventSet = self.options.globalLeafEventSet
experiment = ExperimentWrapper(workFlowArgs.experimentNode)
donePath = os.path.join(os.path.dirname(workFlowArgs.experimentFile), "DONE")
doneDone = os.path.isfile(donePath)
refDone = not workFlowArgs.buildReference or os.path.isfile(experiment.getReferencePath())
halDone = not workFlowArgs.buildHal or (os.path.isfile(experiment.getHALFastaPath()) and
os.path.isfile(experiment.getHALPath()))
if not workFlowArgs.overwrite and doneDone and refDone and halDone:
self.logToMaster("Skipping %s because it is already done and overwrite is disabled" %
self.event)
else:
system("rm -f %s" % donePath)
# delete database
# and overwrite specified (or if reference not present)
dbPath = os.path.join(experiment.getDbDir(),
experiment.getDbName())
seqPath = os.path.join(experiment.getDbDir(), "sequences")
system("rm -f %s* %s %s" % (dbPath, seqPath,
experiment.getReferencePath()))
if workFlowArgs.configWrapper.getDoTrimStrategy() and workFlowArgs.outgroupEventNames is not None:
# Use the trimming strategy to blast ingroups vs outgroups.
self.addChildTarget(CactusTrimmingBlastPhase(cactusWorkflowArguments=workFlowArgs, phaseName="trimBlast"))
else:
self.addChildTarget(CactusSetupPhase(cactusWorkflowArguments=workFlowArgs,
phaseName="setup"))
logger.info("Going to create alignments and define the cactus tree")
self.setFollowOnTarget(FinishUp(workFlowArgs, self.project))
class ProjectWrapper:
alignmentDirName = 'progressiveAlignment'
def __init__(self, options, seqFile, workingDir):
self.options = options
self.seqFile = seqFile
self.workingDir = workingDir
self.configWrapper = None
self.expWrapper = None
self.processConfig()
self.processExperiment()
def processConfig(self):
# read in the default right out of cactus
if self.options.configFile is not None:
configPath = self.options.configFile
else:
dir = cactusRootPath()
configPath = os.path.join(dir, "cactus_progressive_config.xml")
configXml = ET.parse(configPath).getroot()
self.configWrapper = ConfigWrapper(configXml)
# here we can go through the options and apply some to the config
self.configWrapper.setBuildHal(True)
self.configWrapper.setBuildFasta(True)
if self.options.outputMaf is not None:
self.configWrapper.setBuildMaf(True)
self.configWrapper.setJoinMaf(True)
# pre-emptively turn down maxParallelSubtree for singleMachine
# mode if not enough threads are provided to support it. Probably
# need to do something for other ?combined? batch systems?
if self.options.batchSystem == 'singleMachine' and \
self.options.database == 'kyoto_tycoon':
if int(self.options.maxThreads) < \
self.configWrapper.getMaxParallelSubtrees() * 3:
self.configWrapper.setMaxParallelSubtrees(
max(1,
int(self.options.maxThreads) / 3))
# this is a little hack to effectively toggle back to the
# non-progressive version of cactus (as published in Gen. Res. 2011)
# from the high-level interface.
if self.options.legacy is True:
self.configWrapper.setSubtreeSize(sys.maxint)
def processExperiment(self):
expXml = self.seqFile.toXMLElement()
#create the cactus disk
cdElem = ET.SubElement(expXml, "cactus_disk")
database = self.options.database
assert database == "kyoto_tycoon" or database == "tokyo_cabinet"
confElem = ET.SubElement(cdElem, "st_kv_database_conf")
confElem.attrib["type"] = database
dbElem = ET.SubElement(confElem, database)
self.expWrapper = ExperimentWrapper(expXml)
if self.options.database == "kyoto_tycoon":
self.expWrapper.setDbPort(str(self.options.ktPort))
if self.options.ktHost is not None:
self.expWrapper.setDbHost(self.options.ktHost)
if self.options.ktType == 'memory':
self.expWrapper.setDbInMemory(True)
self.expWrapper.setDbSnapshot(False)
elif self.options.ktType == 'snapshot':
self.expWrapper.setDbInMemory(True)
self.expWrapper.setDbSnapshot(True)
else:
assert self.options.ktType == 'disk'
self.expWrapper.setDbInMemory(False)
self.expWrapper.setDbSnapshot(False)
# sonlib doesn't allow for spaces in attributes in the db conf
# which renders this options useless
# if self.options.ktOpts is not None:
# self.expWrapper.setDbServerOptions(self.options.ktOpts)
if self.options.ktCreateTuning is not None:
self.expWrapper.setDbCreateTuningOptions(
self.options.ktCreateTuning)
if self.options.ktOpenTuning is not None:
self.expWrapper.setDbReadTuningOptions(
self.options.ktOpenTuning)
#set the sequence output directory
outSeqDir = os.path.join(self.workingDir, "sequenceData")
if os.path.exists(outSeqDir) and self.options.overwrite:
system("rm -rf %s" % outSeqDir)
if not os.path.exists(outSeqDir):
system("mkdir %s" % outSeqDir)
self.expWrapper.setOutputSequenceDir(
os.path.join(self.workingDir, "sequenceData"))
def writeXml(self):
assert os.path.isdir(self.workingDir)
configPath = absSymPath(os.path.join(self.workingDir, "config.xml"))
expPath = absSymPath(os.path.join(self.workingDir, "expTemplate.xml"))
self.expWrapper.setConfigPath(configPath)
self.configWrapper.writeXML(configPath)
self.expWrapper.writeXML(expPath)
projPath = os.path.join(self.workingDir,
ProjectWrapper.alignmentDirName)
if os.path.exists(projPath) and self.options.overwrite:
system("rm -rf %s" % projPath)
if self.options.outputMaf is True:
fixNames = 1
else:
fixNames = 0
if os.path.exists(projPath):
if not self.isSameAsExisting(expPath, projPath, fixNames):
raise RuntimeError("Existing project %s not " % projPath +
"compatible with current input. Please "
"erase the working directory or rerun "
"with the --overwrite option to start "
"from scratch.")
else:
logPath = os.path.join(self.workingDir, 'cactus.log')
logFile = open(logPath, "a")
logFile.write("\nContinuing existing alignment. Use "
"--overwrite or erase the working directory to "
"force restart from scratch.\n")
logFile.close()
else:
cmd = "cactus_createMultiCactusProject.py \"%s\" \"%s\" --fixNames=%d" % (
expPath, projPath, fixNames)
if len(self.seqFile.outgroups) > 0:
cmd += " --outgroupNames " + ",".join(self.seqFile.outgroups)
if self.options.rootOutgroupDists:
cmd += " --rootOutgroupDists %s" % self.options.rootOutgroupDists
cmd += " --rootOutgroupPaths %s" % self.options.rootOutgroupPaths
if self.options.root is not None:
cmd += " --root %s" % self.options.root
system(cmd)
# create a project in a dummy directory. check if the
# project xml is the same as the current project.
# we do this to see if we should start fresh or try to
# work with the existing project when the overwrite flag is off
def isSameAsExisting(self, expPath, projPath, fixNames):
if not os.path.exists(projPath):
return False
oldPath = os.path.dirname(projPath + "/")
tempPath = "%s_temp" % oldPath
# Fix for relative directories
if oldPath[0:2] == './':
oldPath = oldPath[2:]
if tempPath[0:2] == './':
tempPath = tempPath[2:]
if os.path.exists(tempPath):
system("rm -rf %s" % tempPath)
cmd = "cactus_createMultiCactusProject.py %s %s --fixNames=%d" % (
expPath, tempPath, fixNames)
if len(self.seqFile.outgroups) > 0:
cmd += " --outgroupNames " + ",".join(self.seqFile.outgroups)
if self.options.rootOutgroupDists:
cmd += " --rootOutgroupDists %s" % self.options.rootOutgroupDists
cmd += " --rootOutgroupPaths %s" % self.options.rootOutgroupPaths
if self.options.root is not None:
cmd += " --root %s" % self.options.root
system(cmd)
projFilePathNew = os.path.join(
tempPath, '%s_temp_project.xml' % self.alignmentDirName)
projFilePathOld = os.path.join(
oldPath, '%s_project.xml' % self.alignmentDirName)
newFile = [line for line in open(projFilePathNew, "r")]
oldFile = [line for line in open(projFilePathOld, "r")]
areSame = True
if len(newFile) != len(oldFile):
areSame = False
for newLine, oldLine in zip(newFile, oldFile):
if newLine.replace(tempPath, oldPath) != oldLine:
areSame = False
system("rm -rf %s" % tempPath)
return areSame
