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 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 __init__(self, path, prefix, size, valDataType='float64', valDim=1, weightDataType='float64', weightDim=1, maxNumEdges=0, maxStrLens={}, allowAppend=True):
assert valDim >= 1 and weightDim >= 1
if valDataType == 'S':
valDataType = 'S' + str(max(2, maxStrLens['val']))
if weightDataType == 'S':
weightDataType = 'S' + str(max(2, maxStrLens['weights']))
self._setup(prefix, 'start', getStart, writeNoSlice, None, 'int32', 1, False)
self._setup(prefix, 'end', getEnd, writeNoSlice, None, 'int32', 1, False)
self._setup(prefix, 'strand', getStrand, writeNoSlice, None, 'int8', 1, False)
self._setup(prefix, 'val', getVal, writeNoSlice, None, valDataType, valDim, True)
self._setup(prefix, 'id', getId, writeNoSlice, None, 'S' + str(maxStrLens.get('id')), 1, False)
self._setup(prefix, 'edges', getEdges, writeSliceFromFront, maxNumEdges, 'S' + str(maxStrLens.get('edges')), 1, False)
self._setup(prefix, 'weights', getWeights, writeSliceFromFront, maxNumEdges, weightDataType, weightDim, True)
self._setup(prefix, 'leftIndex', getNone, writeNoSlice, None, 'int32', 1, False)
self._setup(prefix, 'rightIndex', getNone, writeNoSlice, None, 'int32', 1, False)
if not hasattr(self, '_parseFunc'):
self._geParseClass = GetExtra(prefix)
self._setup(prefix, prefix, self._geParseClass.parse, writeNoSlice, None, 'S' + str(maxStrLens.get(prefix)), 1, False)
# If there is one number in the path, it is the data type dimension.
# Only one value is allowed per element, no extra dimensions are added
# to the array and the element dimension is None.
#
# Example: val.4.float64 contains, per element, a vector of 4 numbers.
# The shape is (n,4) for n elements.
#
# If there are two numbers in the path, the first is the maximal element
# dimension and the second is the data type dimension.
#
# Example: weights.3.4.float64 contains, per element, at most 3 vectors
# of 4 numbers each. The shape is (n,3,4) for n elements.
self._fn = createMemmapFileFn(path, prefix, self._elementDim, self._dataTypeDim, self._dataType)
self._index = 0
shape = [size] + \
([max(1, self._elementDim)] if self._elementDim is not None else []) + \
([self._dataTypeDim] if self._dataTypeDim > 1 else [])
append = os.path.exists(self._fn)
if append:
if not allowAppend:
raise InvalidFormatError('Error: different genome element sources (e.g. different input files) tries to write to index file for the same chromosome (%s). This is probably caused by different files in the same folder containing elements from the same chromosome.' % self._fn)
try:
f = np.memmap( self._fn, dtype=self._dataType, mode='r+' )
self._index = len(f) / product(shape[1:])
del f
existingShape = calcShapeFromMemmapFileFn(self._fn)
self._contents = np.array( np.memmap(self._fn, dtype=self._dataType, mode='r+', shape=tuple(existingShape)) )
self._contents = np.r_[self._contents, np.zeros( dtype=self._dataType, shape=tuple(shape) )]
except Exception:
print 'Error when opening file: ', self._fn
raise
else:
self._contents = np.zeros( dtype=self._dataType, shape=tuple(shape) )
if not append and self._setEmptyVal:
self._contents[:] = findEmptyVal(self._dataType)
def __init__(self, path, prefix, size, valDataType='float64', valDim=1, weightDataType='float64', weightDim=1, maxNumEdges=0, maxStrLens={}, allowAppend=True):
assert valDim >= 1 and weightDim >= 1
if valDataType == 'S':
valDataType = 'S' + str(max(2, maxStrLens['val']))
if weightDataType == 'S':
weightDataType = 'S' + str(max(2, maxStrLens['weights']))
self._setup(prefix, 'start', getStart, writeNoSlice, None, 'int32', 1, False)
self._setup(prefix, 'end', getEnd, writeNoSlice, None, 'int32', 1, False)
self._setup(prefix, 'strand', getStrand, writeNoSlice, None, 'int8', 1, False)
self._setup(prefix, 'val', getVal, writeNoSlice, None, valDataType, valDim, True)
self._setup(prefix, 'id', getId, writeNoSlice, None, 'S' + str(maxStrLens.get('id')), 1, False)
self._setup(prefix, 'edges', getEdges, writeSliceFromFront, maxNumEdges, 'S' + str(maxStrLens.get('edges')), 1, False)
self._setup(prefix, 'weights', getWeights, writeSliceFromFront, maxNumEdges, weightDataType, weightDim, True)
self._setup(prefix, 'leftIndex', getNone, writeNoSlice, None, 'int32', 1, False)
self._setup(prefix, 'rightIndex', getNone, writeNoSlice, None, 'int32', 1, False)
if not hasattr(self, '_parseFunc'):
self._geParseClass = GetExtra(prefix)
self._setup(prefix, prefix, self._geParseClass.parse, writeNoSlice, None, 'S' + str(maxStrLens.get(prefix)), 1, False)
# If there is one number in the path, it is the data type dimension.
# Only one value is allowed per element, no extra dimensions are added
# to the array and the element dimension is None.
#
# Example: val.4.float64 contains, per element, a vector of 4 numbers.
# The shape is (n,4) for n elements.
#
# If there are two numbers in the path, the first is the maximal element
# dimension and the second is the data type dimension.
#
# Example: weights.3.4.float64 contains, per element, at most 3 vectors
# of 4 numbers each. The shape is (n,3,4) for n elements.
self._fn = createMemmapFileFn(path, prefix, self._elementDim, self._dataTypeDim, self._dataType)
self._index = 0
shape = [size] + \
([max(1, self._elementDim)] if self._elementDim is not None else []) + \
([self._dataTypeDim] if self._dataTypeDim > 1 else [])
append = os.path.exists(self._fn)
if append:
if not allowAppend:
raise InvalidFormatError('Error: different genome element sources (e.g. different input files) tries to write to index file for the same chromosome (%s). This is probably caused by different files in the same folder containing elements from the same chromosome.' % self._fn)
try:
f = np.memmap( self._fn, dtype=self._dataType, mode='r+' )
self._index = len(f) / product(shape[1:])
del f
existingShape = calcShapeFromMemmapFileFn(self._fn)
self._contents = np.array( np.memmap(self._fn, dtype=self._dataType, mode='r+', shape=tuple(existingShape)) )
self._contents = np.r_[self._contents, np.zeros( dtype=self._dataType, shape=tuple(shape) )]
except Exception:
print 'Error when opening file: ', self._fn
raise
else:
self._contents = np.zeros( dtype=self._dataType, shape=tuple(shape) )
if not append and self._setEmptyVal:
self._contents[:] = findEmptyVal(self._dataType)
