def __setitem__(self, indexes_or_slice, v):
if self.errors() is None: raise NotImplementedError , "__setitem__: errors is None"
# if indexes_or_slice is iterable,
# we should assume that it is trying to do slicing.
# This also means no axis can use iterables as values.
if self.dimension() == 1 and '__iter__' not in dir(indexes_or_slice):
indexes_or_slice = (indexes_or_slice,)
# We also allow use of dictionary. This is a convenient and flexible way
# to get or set a slice.
if isinstance(indexes_or_slice, dict):
s = _slicingInfosFromDictionary(indexes_or_slice, self.axes())
else:
s = _makeSlicingInfos( indexes_or_slice, self.dimension() )
pass # end if
#at this point, s is a tuple of SlicingInfo instances.
slicingInfos = s
# check sanity of inputs
if not isinstance(s, tuple) or len(s) != self.dimension():
raise NotImplementedError , "Histogram[ %s ]. my dimension: %s" % (
s, self.dimension())
mydatasets = self.datasets()
# slicingInfo tuple --> a tuple of index slices
indexSlices = self.slicingInfos2IndexSlices( slicingInfos )
#now we want to know if user is requesting for a real slice
#or just an element
#this is done by trying to get a slice of the dataset self._data
aslice = self._data[indexSlices]
#1. element, not slice
if isNumber(aslice) and isNumberList(v) :
if len(v) == 2:
self._data[indexSlices] = v[0]
self._errors[indexSlices] = v[1]
pass
elif len(v) == len(mydatasets):
for i, ds in enumerate(mydatasets): ds[indexSlices] = v[i]
pass
else:
raise RuntimeError , \
"shape mismatch in histogram[ indexes ] = value tuple. "\
"len(value tuple) = %s, but histogram has %s datasets" \
% (len(v), len(mydatasets))
return v
#2. slice
#shape = aslice.shape() #get shape
# for slice, we would require the right hand side to be a list of datasets
if isHistogram( v ): v = v.data(), v.errors()
# try to set slice, defer to dataset's __setitem__
# but first we must assert length of arrays match
assert len(v) == len(mydatasets), \
"rhs must be a %s-tuple of datasets, "\
"instead of a %s-tuple" % (
len( mydatasets ), len(v) )
for lhs, rhs in zip(mydatasets, v):
if rhs is not None : lhs[ indexSlices ] = rhs
else: debug.log( 'indefinite behavior: setting to None' )
continue
return self[ indexes_or_slice ]
def isNumberList( l ):
return reduce(operator.and_, [isNumber(i) for i in l])
def __getitem__(self, s):
"""Slicing
h[ (3.0, 4.0), () ]
h[ 3.0, () ]
h[ (), () ]
h[ (None, 4.0), (999., None ) ]
"""
if self.errors() is None:
raise NotImplementedError , "__getitem__: errors is None"
# if s is iterable, we should assume that it is trying to do slicing.
# This also means no axis can use iterables as values.
if self.dimension() == 1 and '__iter__' not in dir(s): s = (s,)
# We also allow use of dictionary. This is a convenient and flexible way
# to get or set a slice.
if isinstance(s, dict):
s = _slicingInfosFromDictionary(s, self.axes())
else:
s = _makeSlicingInfos( s, self.dimension() )
#at this point, s is a tuple of SlicingInfo instances.
# check sanity of inputs
if not isinstance(s, tuple) or len(s) != self.dimension():
raise NotImplementedError , "Histogram[ %s ]. my dimension: %s" % (
s, self.dimension())
# a more meaningful name
slicingInfos = s
# slicingInfo tuple --> a tuple of index slices
indexSlices = self.slicingInfos2IndexSlices( slicingInfos )
#the following line will fail if a dataset is None
#should define a special NoneDataset to solve this problem
newdatasets = [dataset[indexSlices] for dataset in self.datasets()]
#if requested for item instead of slice, return the item now.
if isNumber(newdatasets[0]) or isDimensional(newdatasets[0]):
return newdatasets
#meta data need to be passed to the new histogram
newAttrs = self._attributes.copy()
#axes of new histogram
newAxes = []
for slicingInfo, name in zip(slicingInfos, self.axisNameList()):
axis = self.axisFromName( name )
if not isSlicingInfo( slicingInfo ):
# if it is not a slicingInfo instance,
# it must be a value indexable in this axis.
# This is already tested in method "slicingInfo2IndexSlices"
value = '%s' % slicingInfo
name = axis.name()
newAttrs[name] = value
else:
newAxes.append( axis[ slicingInfo ] )
pass
continue
#name of new histogram
newName = "%s in %s" % (
self.name(),
["%s(%s)"%(axisName, slicingInfo) for axisName, slicingInfo \
in zip(self.axisNameList(), slicingInfos)])
#new histogram
new = Histogram(
name = newName, unit = self.unit(),
data = newdatasets[0], errors = newdatasets[1],
axes = newAxes, attributes = newAttrs, slice = True)
#addtional datasets. This is not tested yet!!!
#probably we should really limit histogram to have only two datasets!!!
for i in range(2, len(newdatasets)):
ds = newdatasets[i]
new.addDataset( ds.name(), ds )
continue
return new