def _createPreProcFiles(self):
geSource = self._geSourceManager.getGESource()
genome = geSource.genome
collector = PreProcMetaDataCollector(genome, self._trackName)
collector.updateMetaDataForFinalization(geSource.getFileSuffix(), geSource.getPrefixList(), \
geSource.getValDataType(), geSource.getValDim(), \
geSource.getEdgeWeightDataType(), geSource.getEdgeWeightDim(), \
geSource.hasUndirectedEdges(),
geSource.getVersion(), PreProcessUtils.constructId(geSource), \
self._geSourceManager.getNumElements(), \
self._geSourceManager.getBoundingRegionTuples(), \
self._geSourceManager.getValCategories(), \
self._geSourceManager.getEdgeWeightCategories(), \
self._allowOverlaps)
if self._geSourceManager.getNumElements() == 0:
return
if self._mode != 'Real':
for ge in geSource:
pass
return
output = OutputManager(genome, self._trackName, self._allowOverlaps, self._geSourceManager)
writeFunc = output.writeRawSlice if geSource.isSliceSource() else output.writeElement
for ge in geSource:
writeFunc(ge)
collector.flagChrsAsPreProcessed(self._allowOverlaps, self._geSourceManager.getAllChrs())
output.close()
def merge(genome, trackName, allowOverlaps):
path = createDirPath(trackName, genome, allowOverlaps=allowOverlaps)
collector = PreProcMetaDataCollector(genome, trackName)
chrList = collector.getPreProcessedChrs(allowOverlaps)
if not collector.getTrackFormat().reprIsDense():
chrList = sorted(chrList)
existingChrList = [
chr
for chr in ChrMemmapFolderMerger._existingChrIter(path, chrList)
]
if len(existingChrList) == 0:
raise EmptyGESourceError(
'No data lines has been read from source file (probably because it is empty).'
)
firstChrTrackData = TrackSource().getTrackData(trackName,
genome,
existingChrList[0],
allowOverlaps,
forceChrFolders=True)
arrayList = firstChrTrackData.keys()
for arrayName in arrayList:
mergedArray = firstChrTrackData[arrayName][:]
elementDim, dtypeDim = parseMemmapFileFn(
firstChrTrackData[arrayName].filename)[1:3]
del firstChrTrackData[arrayName]
for chr in existingChrList[1:]:
chrTrackData = TrackSource().getTrackData(trackName,
genome,
chr,
allowOverlaps,
forceChrFolders=True)
mergedArray = ChrMemmapFolderMerger.mergeArrays(
mergedArray, np.array(chrTrackData[arrayName][:]))
elementDimNew, dtypeDimNew = parseMemmapFileFn(
chrTrackData[arrayName].filename)[1:3]
elementDim = max(elementDim, elementDimNew)
dtypeDim = max(dtypeDim, dtypeDimNew)
del chrTrackData[arrayName]
mergedFn = createMemmapFileFn(path, arrayName, elementDim,
dtypeDim, str(mergedArray.dtype))
f = np.memmap(mergedFn,
dtype=mergedArray.dtype,
mode='w+',
shape=mergedArray.shape)
f[:] = mergedArray
f.flush()
del f
del mergedArray
def _findTrackInfoBasedMetaData(self):
if not self._foundTrackInfoBasedMetaData:
if PreProcMetaDataCollector.hasKey(self._genome, self._trackName):
collector = PreProcMetaDataCollector(self._genome, self._trackName)
self._fileSuffix = collector.getFileSuffix()
self._preProcVersion = collector.getPreProcVersion()
self._id = collector.getId()
self._undirectedEdges = True if collector.hasUndirectedEdges() else False
else:
ti = TrackInfo(self._genome, self._trackName)
self._fileSuffix = ti.fileType
self._preProcVersion = ti.preProcVersion
self._id = ti.id
self._undirectedEdges = True if ti.undirectedEdges else False
def _allGESourceManagers(self, trackName, allowOverlaps):
collector = PreProcMetaDataCollector(self._genome, trackName)
if allowOverlaps == False and collector.overlapRuleHasBeenFinalized(
True):
for i in range(1):
self._status = 'Trying to prepare preprocessing for track "%s"' % ':'.join(trackName) + \
(' (allowOverlaps: %s)' % allowOverlaps)
yield self._getGESourceManagerFromTrack(trackName)
else:
for geSource in self._allGESources(trackName):
if allowOverlaps == True:
tf = TrackFormat.createInstanceFromGeSource(geSource)
if tf.isDense() or geSource.hasNoOverlappingElements():
return
self._status = 'Trying to prepare preprocessing for track "%s"' % ':'.join(trackName) + \
(' (filename: "%s")' % geSource.getFileName() if geSource.hasOrigFile() else '') + \
(' (allowOverlaps: %s)' % allowOverlaps)
if PreProcessUtils.shouldPreProcessGESource(
trackName, geSource, allowOverlaps):
yield self._getGESourceManagerFromGESource(geSource)
def process(self):
assert self._genome is not None, 'Error: genome must be specified when preprocessing tracks.'
atLeastOneFinalized = False
for trackName in self._allTrackNames():
assert trackName != ['']
overlapRulesProcessedForTrackName = []
collector = PreProcMetaDataCollector(self._genome, trackName)
try:
trackName = self._renameTrackNameIfIllegal(trackName)
for allowOverlaps in [True, False]:
anyGeSourceManagers = False
for geSourceManager in self._allGESourceManagers(
trackName, allowOverlaps):
anyGeSourceManagers = True
# PreProcess if needed
if self._shouldPreProcess():
PreProcessUtils.removeOutdatedPreProcessedFiles(
self._genome, trackName, allowOverlaps,
self._mode)
if self._shouldPrintProcessMessages(
) and allowOverlaps not in overlapRulesProcessedForTrackName:
self._printProcessTrackMessage(
trackName, allowOverlaps)
overlapRulesProcessedForTrackName.append(
allowOverlaps)
self._status = 'Trying to preprocess geSource...'
geSourceJob = PreProcessGeSourceJob(
trackName, geSourceManager, allowOverlaps,
self._mode)
anyWarnings = geSourceJob.process()
if self._raiseIfAnyWarnings and anyWarnings and trackName not in self._warningTrackNames:
self._warningTrackNames.append(trackName)
collector.updatePreProcDirtyStatus(
geSourceJob.hasModifiedData())
# Finalize overlapRule output if needed
if anyGeSourceManagers and self._shouldFinalize(
) and collector.preProcIsDirty():
if self._mode == 'Real' and self._shouldMergeChrFolders(
):
self._status = 'Trying to combine chromosome vectors into combined vectors.'
PreProcessUtils.createBoundingRegionShelve(
self._genome, trackName, allowOverlaps)
ChrMemmapFolderMerger.merge(
self._genome, trackName, allowOverlaps)
self._status = 'Trying to remove chromosome folders'
PreProcessUtils.removeChrMemmapFolders(
self._genome, trackName, allowOverlaps)
self._status = 'Trying to check whether 3D data is correct'
PreProcessUtils.checkIfEdgeIdsExist(
self._genome, trackName, allowOverlaps)
PreProcessUtils.checkUndirectedEdges(
self._genome, trackName, allowOverlaps)
PreProcessUtils.checkUndirectedEdges(
self._genome, trackName, allowOverlaps)
collector.markOverlapRuleAsFinalized(allowOverlaps)
# Finalize track if needed
if self._shouldFinalize():
if collector.preProcIsDirty():
self._status = 'Trying to finalize.'
collector.finalize(self._username,
self._shouldPrintProcessMessages())
if not atLeastOneFinalized:
atLeastOneFinalized = True
else:
collector.removeEntry()
except NotSupportedError, e:
collector.removeEntry()
if self.PASS_ON_EXCEPTIONS:
raise
else:
self._printExceptionMsg(e, trackName, Error=False)
except Exception, e:
collector.removeEntry()
if self.PASS_ON_EXCEPTIONS:
raise
else:
self._printExceptionMsg(e, trackName, Error=True)
def _getGESourceManagerFromTrack(self, trackName):
origBrTuples = PreProcMetaDataCollector(self._genome, trackName).\
getBoundingRegionTuples(allowOverlaps=True)
return OverlapClusteringGESourceManager(self._genome, trackName,
origBrTuples)