def _commonReshapeArray(a, delta, appendFunc):
return appendFunc(a, delta, findEmptyVal(str(a.dtype)))
def _getRawTrackView(self, region, borderHandling, allowOverlaps):
assert len(region) == 1
from collections import OrderedDict
from gtrackcore_memmap.track.memmap.CommonMemmapFunctions import findEmptyVal
from gtrackcore_memmap.track.core.TrackView import TrackView
import numpy as np
geSource = ExternalTrackManager.getGESourceFromGalaxyOrVirtualTN(self.trackName, region.genome)
prefixList = geSource.getPrefixList()
valDataType = geSource.getValDataType()
valDim = geSource.getValDim()
weightDataType = geSource.getEdgeWeightDataType()
weightDim = geSource.getEdgeWeightDim()
startList, endList, valList, strandList, idList, edgesList, weightsList = [None] * 7
extraLists = OrderedDict()
tf = TrackFormat.createInstanceFromPrefixList(prefixList, valDataType, valDim, weightDataType, weightDim)
if allowOverlaps and (tf.isDense() or geSource.hasNoOverlappingElements()):
raise IncompatibleTracksError(
prettyPrintTrackName(self.trackName)
+ " with format: "
+ str(tf)
+ " does not satisfy "
+ str(self._trackFormatReq)
)
denseAndInterval = tf.isDense() and tf.isInterval()
numEls = 2 if denseAndInterval else 1
if valDataType == "S":
valDataType = "S2"
if weightDataType == "S":
weightDataType = "S2"
for prefix in prefixList:
if prefix == "start":
startList = np.array([-1], dtype="int32")
elif prefix == "end":
if denseAndInterval:
endList = np.array([0, 1], dtype="int32")
else:
endList = np.array([0], dtype="int32")
elif prefix == "val":
valList = np.array([findEmptyVal(valDataType)] * valDim * numEls, dtype=valDataType).reshape(
(numEls, valDim) if valDim > 1 else numEls
)
elif prefix == "strand":
strandList = np.array([1] * numEls, dtype="int8")
elif prefix == "id":
idList = np.array([""] * numEls, dtype="S1")
elif prefix == "edges":
edgesList = np.array([[""]] * numEls, dtype="S1")
elif prefix == "weights":
weightsList = np.array(
[[[findEmptyVal(weightDataType)]]] * weightDim * numEls, dtype=weightDataType
).reshape((numEls, 1, weightDim) if weightDim > 1 else (numEls, 1))
else:
extraLists[prefix] = np.array([""] * numEls, dtype="S1")
return TrackView(
region,
startList,
endList,
valList,
strandList,
idList,
edgesList,
weightsList,
borderHandling,
allowOverlaps,
extraLists,
)
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)
