def process(self, parentage):
"""
Draws the histogram corresponding to the present cut and optionally the weighting
strategy.
Args:
parentage: the names of the call stack
Returns:
a tuple (Hist, Node.process()) with the projected histogram Hist and the parent Node class
process output.
"""
r = Node.process(self, parentage)
# sys.stdout.write("+")
# sys.stdout.flush()
#Use the cache of the previous CutNode in the call stack.
cache = self.getPreviousCache(parentage)
#Get the total cut string from the call stack.
cut = self.getPrevious(parentage, lambda x: hasattr(x, "getCutsTotal")).\
getCutsTotal(parentage)
#Get the total weight
weights = self.getParents(parentage, lambda x: hasattr(x, "weight"))
wtot = mul([w.weight for w in weights])
if not "var" in self.hist_desc.keys() or not "binning" in self.hist_desc.keys():
raise KeyError("Incorrect hist desc: %s" % str(self.hist_desc))
#Get the sample node
snodes = self.getParents(parentage, lambda x: hasattr(x, "sample"))
if not len(snodes)==1:
raise Exception("More than one parent was a SampleNode, undefined behaviour (probably human error).")
snode = snodes[0]
hi = snode.sample.drawHistogram(self.hist_desc["var"], str(cut), weight=wtot.weight_str, binning=self.hist_desc["binning"], entrylist=cache)
hdir = self.parentsName(parentage)
#Scale MC histograms to 1 inverse picobarn
if snode.sample.isMC:
hi.Scale(snode.sample.lumiScaleFactor(1))
line_to_show = [
"->".join([p.name for p in parentage]+[self.name]),
hi.GetEntries(),
cache.GetN(),
"%.2f" % hi.Integral()
]
self.logger.info(
", ".join(map(lambda x: str(x), line_to_show))
)
hi.SetName(self.name)
snode.saver.save(hdir, hi)
return (hi, r)
def syst_weights(graph):
_syst_weights = []
for lepton, w in weights_lepton.items():
weights_var_by_one = variateOneWeight([x[1] for x in (weights_syst+w)])
# The unvariated weight is taken as the list of the 0th elements of the
# weight tuples
weights_var_by_one.append(
("nominal", [x[1][0] for x in (weights_syst+w)])
)
wtot = []
for wn, s in weights_var_by_one:
j = mul(s) #Multiply together the list of weights
wtot.append((wn, j))
for name, j in wtot:
filter_funcs=[
#Apply the weights separately for the lepton channels
lambda _x,_lepton=lepton: is_chan(_x, _lepton),
#Apply only in MC
is_mc
]
#Apply systematic weights only in case of nominal samples
if name != "nominal":
filter_funcs += [
#Check if the parent was a nominal sample
lambda _x: "/nominal/" in _x[0].name
]
syst = WeightNode(
j, graph, "weight__" + name + "__" + lepton,
[], [], filter_funcs=filter_funcs
)
logger.debug("Appending weight %s" % syst.name)
_syst_weights.append(syst)
#Always produce the unweighted plot
unw = WeightNode(
Weights.no_weight, graph, "weight__unweighted",
[], []
)
_syst_weights.append(unw)
return _syst_weights
