def getClone(tfile, root_dir, name):
th1 = getHist(tfile, root_dir, name)
cloned = th1.Clone()
cloned.SetDirectory(0)
cloned.Sumw2()
ROOT.SetOwnership(cloned, 1)
return cloned
def getTemplate(infile, config, section):
# so far we need only one template per sample, we do not differentiate plus/minus etc.
if config.has_option(section, 'template_hist'):
if config.has_option(section, 'template_file'):
# we have to open some external file
extrafile = ROOT.TFile(
config.get(section, 'input_folder') + '/' +
config.get(section, 'template_file'), 'READONLY')
if extrafile.IsZombie():
raise RuntimeError, 'No file %s in %s' % \
(config.get(section,'template_file'),
config.get(section,'input_folder'))
template = getClone(extrafile, '', '',
config.get(section, 'template_hist'))
template.SetDirectory(infile.GetDirectory(''))
extrafile.Close()
else:
# read the template from input_file
template = getClone(infile, '', '',
config.get(section, 'template_hist'))
template.Scale(1.0 / template.Integral())
return template
else:
return None
def dumpOutput(filename, outputPrefix, usePreQCD = False, combineWJets = False):
# TODO fix this
inputFile = filename
outputFile = "%s_split.root" % inputFile
templateFile = inputFile + "_templates.root"
metaFile = inputFile + "_meta.json"
temp = TFile(templateFile)
output = TFile(outputFile, "RECREATE")
output.cd()
meta = json.loads(open(metaFile).read())
dataHist = temp.Get('newdata;1')
mcHists = {'Data':dataHist}
for key in temp.GetListOfKeys():
keyString = key.GetName()
if keyString.endswith('_updated'):
mcHists[keyString.split('_')[0]] = temp.Get(keyString)
reverseIndex = dict((v, k) for (k, v) in meta['groupIndexMap'].items())
lowerEdgeIdx = 0
for (binName, binIndex) in meta['groupIndexMap'].items():
newBins = meta['numBinsVec'][binIndex]
lowerBin = meta['lowerEdgeBinVec'][binIndex]
upperBin = meta['upperEdgeBinVec'][binIndex]
binVec = [float(x) for x in meta['binLowerEdges'][lowerBin - 1:upperBin]]
binArray = array.array('f', binVec)
for mc in mcHists:
newName = "%s%s" % (mc, binName)
# TODO propagate xmin/xmax from the fit
newHist = ROOT.TH1F(newName, newName, newBins - 1, binArray)
for histIndex in range(newBins):
oldContent = mcHists[mc].GetBinContent(lowerBin + histIndex)
oldError = mcHists[mc].GetBinError(lowerBin + histIndex)
newHist.SetBinContent(histIndex, oldContent)
newHist.SetBinError(histIndex, oldError)
newHist.Write()
lowerEdgeIdx += newBins
output.Write()
temp.Close()
temp = output
discList = []
for key in temp.GetListOfKeys():
keyString = key.GetName()
keySplit = keyString.split('_')
currDisc = "_".join(keySplit[2:])
if currDisc == 'svm':
continue
if not currDisc in discList:
discList.append(currDisc)
discList.sort()
# split the histograms out, so plot them now
defaultColors = {'SingleTop' : ROOT.kCyan,
'SingleTT' : ROOT.kCyan - 3,
'SingleTS' : ROOT.kCyan - 2,
'SingleTTW' : ROOT.kCyan - 1,
'SingleTbarT' : ROOT.kCyan + 1,
'SingleTbarS' : ROOT.kCyan + 2,
'SingleTbarTW' : ROOT.kCyan + 3,
'WJets' : ROOT.kGreen,
'Wqq' : ROOT.kGreen + 1,
'Wcx' : ROOT.kGreen + 2,
'Wbx' : ROOT.kGreen + 3,
'WZ' : ROOT.kYellow +3,
'WW' : ROOT.kYellow -6,
'DiBoson' : ROOT.kOrange,
'EWK' : ROOT.kOrange,
'Top' : ROOT.kRed,
'Zinv' : ROOT.kBlue-3,
'ZJets' : ROOT.kBlue,
'ZZ' : ROOT.kPink -6,
'Stop450' : ROOT.kYellow,
'QCDpre' : ROOT.kMagenta,
'QCD' : ROOT.kMagenta}
xlabelsPerDisc = {'lepPt' : 'GeV/c',
'MET' : 'GeV',
'hT' : 'GeV',
'std' : 'GeV/c',
'stdt' : 'GeV/c',
'sumEt' : 'GeV',
'wrp' : 'rad',
'wMT' : 'GeV/c^{2}',
'hT' : 'GeV',
}
sampleToCategory = { 'WZ' : 'Diboson',
'WW' : 'Diboson',
'ZZ' : 'Diboson',
'SingleTT' : 'SingleTop',
'SingleTS' : 'SingleTop',
'SingleTTW' : 'SingleTop',
'SingleTbarT' : 'SingleTop',
'SingleTbarS' : 'SingleTop',
'SingleTbarTW' : 'SingleTop',
'Zinv' : 'ZJets',
'ZJets' : 'ZJets' }
def iterateJetBottomTau():
return itertools.product((0,1,2,3,4,5),(0,1,2),(0,1))
integralLookup = []
inputKeyLookup = []
for key in temp.GetListOfKeys():
keyString = key.GetName()
if not keyString in inputKeyLookup:
inputKeyLookup.append(keyString)
integralLookup.append((keyString, temp.Get(keyString).Integral()))
htmlTarget = "%s.html" % outputPrefix
relBase = os.path.dirname(htmlTarget)
relSuffix = os.path.basename(outputPrefix)
with open(htmlTarget,'w+') as idx:
idx.write("<html><head><title>Fit Results</title></head><body><h1>SHyFT Fit Results</h1>")
sampleCounts = {}
# FIXME: This should really really be automagic
for njet, nbjet, ntau in iterateJetBottomTau():
if (nbjet + ntau) > njet:
continue
if not njet in sampleCounts:
sampleCounts[njet] = {}
if not nbjet in sampleCounts[njet]:
sampleCounts[njet][nbjet] = {}
if not ntau in sampleCounts[njet][nbjet]:
sampleCounts[njet][nbjet][ntau] = {}
plot = StackedPlot()
# FIXME should have serialization/deserialization fixed by now
targetSuffix = "_%sj_%sb_%st" % (njet, nbjet, ntau)
shouldSave = False
sizeListing = []
def customCompare(val):
return -1*val[1]
foundBin = False
for keyString,_ in sorted(integralLookup, key=customCompare, reverse=True):
sampleString = keyString.split('_')[0]
if keyString.endswith(targetSuffix):
foundBin = True
hist = temp.Get(keyString)
if hist.Integral() != 0:
shouldSave = True
color = defaultColors.get(sampleString, None)
currSample = keyString.split('_')[0]
plot.addQuantity(currSample, temp.Get(keyString), color = color)
tableCategory = sampleToCategory.get(sampleString, sampleString)
sampleCounts[njet][nbjet][ntau][tableCategory] = \
sampleCounts[njet][nbjet][ntau].get(tableCategory, 0) + \
temp.Get(keyString).Integral()
disc = keyString.split('_')[1]
if not foundBin:
print "Got nothing for %s? (probably okay)" % targetSuffix
continue
if disc in xlabelsPerDisc:
plot.xAxisTitle = xlabelsPerDisc[disc]
plot.title = "%s (%sj, %sb, %st)" % (disc, njet, nbjet, ntau)
if shouldSave:
target = os.path.join(relBase,'%sfit%s' % (relSuffix, targetSuffix))
plot.draw(target + ".svg")
plot.draw(target + ".pdf")
plot.draw(target + ".png")
idx.write('<img src="%sfit%s.svg" alt="%s" />' % (relSuffix, targetSuffix, plot.title))
idx.write('<table><tr><th>Bin</th><th>Data</th><th>Total Pred</th><th>SF</th>')
headers = {}
totalPred = {}
sampleSum = {}
allPredictions = 0
for njet in sampleCounts:
totalPred[njet] = {}
for nbjet in sampleCounts[njet]:
totalPred[njet][nbjet] = {}
for ntau in sampleCounts[njet][nbjet]:
totalPred[njet][nbjet][ntau] = 0
for sample in sampleCounts[njet][nbjet][ntau]:
sampleSum[sample] = sampleSum.get(sample, 0) + sampleCounts[njet][nbjet][ntau][sample]
if sample != 'Data':
headers[sample] = 1
totalPred[njet][nbjet][ntau] += sampleCounts[njet][nbjet][ntau][sample]
allPredictions += sampleCounts[njet][nbjet][ntau][sample]
headerList = sorted(headers.keys())
for header in headerList:
idx.write('<th>%s</th>' % header)
idx.write('</tr>\n')
for njet in sorted(sampleCounts):
for nbjet in sorted(sampleCounts[njet]):
for ntau in sorted(sampleCounts[njet][nbjet]):
if sampleCounts[njet][nbjet][ntau] == {}:
continue
if totalPred[njet][nbjet][ntau] > 0:
frac = (sampleCounts[njet][nbjet][ntau]['Data']/totalPred[njet][nbjet][ntau])
else:
frac = 0.0
idx.write('<tr><th>%sj_%sb_%st</th><td>%.2e</td><td>%.2e</td><td>%.2f</td>' %\
(njet,
nbjet,
ntau,
sampleCounts[njet][nbjet][ntau]['Data'],
totalPred[njet][nbjet][ntau], frac))
for sample in headerList:
idx.write('<td>%.2e</td>' % sampleCounts[njet][nbjet][ntau].get(sample, 0))
idx.write('</tr>\n')
if allPredictions > 0.0:
idx.write('<tr><th>Total</th><th>%.2e</th><th>%.2e</th><th>%.2f</th>' % (sampleSum['Data'], allPredictions, sampleSum['Data']/allPredictions))
else:
idx.write('<tr><th>Total</th><th>%.2e</th><th>%.2e</th><th>INF</th>' % (sampleSum['Data'], allPredictions))
for header in headerList:
idx.write('<th>%.2e</th>' % sampleSum[header])
idx.write('</tr></table>')
idx.write("</body></html>")
def draw(self, filename, nostack=False, dropPattern=""):
"""
Draw the current image
"""
if not nostack:
drawOpt = "HIST"
else:
drawOpt = "nostack,HIST"
myCanvas = ROOT.TCanvas("c1", "stacked hists", 10, 10, 900, 700)
oldPad = ROOT.gPad
myCanvas.cd()
hs = ROOT.THStack('hs', self.title)
r = 0
if nostack:
for shape in self.getFilteredShapes(dropPattern):
integral = shape['shape'].Integral()
shape['oldshape'] = shape['shape']
shape['shape'] = shape['shape'].Clone()
if integral:
shape['shape'].Scale(1 / integral)
for shape in self.getFilteredShapes(dropPattern):
shape['shape'].SetOption('')
if self.isData(shape):
continue
# 2 is red. Go figure.
if 'color' in shape:
if nostack:
shape['shape'].SetLineColor(shape['color'])
shape['shape'].SetFillColor(ROOT.kWhite)
shape['shape'].SetLineWidth(4)
else:
shape['shape'].SetLineColor(ROOT.kBlack)
shape['shape'].SetFillColor(shape['color'])
#shape['shape'].SetLineWidth(1.0)
else:
shape['shape'].SetFillColor(2 + r)
if nostack:
shape['shape'].SetLineColor(2 + r)
else:
shape['shape'].SetLineColor(ROOT.kBlack)
r += 1
hs.Add(shape['shape'])
#if not nostack:
# hs.Draw(drawOpt)
#else:
# first = True
# for shape in self.shapes:
# if first:
# shape['shape'].Draw()
# else:
# shape['shape'].Draw('same')
hs.SetDrawOption(drawOpt)
hs.Draw(drawOpt)
hs.SetDrawOption(drawOpt)
for shape in self.getFilteredShapes(dropPattern):
if self.isData(shape):
if shape['shape'].GetMaximum() > hs.GetMaximum():
hs.SetMaximum(shape['shape'].GetMaximum())
pass
if not nostack:
shape['shape'].Draw('esame')
else:
shape['shape'].Draw('HIST,same')
if self.xAxisTitle:
axis = hs.GetXaxis()
axis.SetTitle(self.xAxisTitle)
myCanvas.RedrawAxis()
leg = ROOT.TLegend(0.65, 0.7, 0.99, 0.99)
#leg = ROOT.TLegend()
for shape in reversed([x
for x in self.getFilteredShapes(dropPattern)]):
if self.isData(shape):
#shape['shape'].SetLineColor(ROOT.kBlack)
if not nostack:
leg.AddEntry(shape['shape'], shape['name'], "pl")
else:
leg.AddEntry(shape['shape'], shape['name'], "l")
for shape in reversed([x
for x in self.getFilteredShapes(dropPattern)]):
if not self.isData(shape):
#shape['shape'].SetLineColor(ROOT.kBlack)
if not nostack:
leg.AddEntry(shape['shape'], shape['name'], 'f')
else:
leg.AddEntry(shape['shape'], shape['name'], 'l')
leg.Draw()
myCanvas.SaveAs(filename)
myCanvas.Close()
ROOT.gPad = oldPad
if nostack:
for shape in self.getFilteredShapes(dropPattern):
shape['shape'] = shape['oldshape']
reverseIndex = dict((v, k) for (k, v) in meta['groupIndexMap'].items())
lowerEdgeIdx = 0
for (binName, binIndex) in meta['groupIndexMap'].items():
newBins = meta['numBinsVec'][binIndex]
lowerBin = meta['lowerEdgeBinVec'][binIndex] - 1
upperBin = meta['upperEdgeBinVec'][binIndex] - 1
binVec = [float(x) for x in meta['binLowerEdges'][lowerBin:upperBin + 1]]
binVec.append(
float(binVec[-1] + meta['binLowerEdges'][lowerBin + 1] -
meta['binLowerEdges'][lowerBin]))
binArray = array.array('f', binVec)
for mc in mcHists:
newName = "%s%s" % (mc, binName)
# TODO propagate xmin/xmax from the fit
newHist = ROOT.TH1F(newName, newName, newBins, binArray)
for histIndex in range(newBins):
# Stupid zero-based indices
oldBin = lowerBin + histIndex + 1
newBin = histIndex + 1
oldContent = mcHists[mc].GetBinContent(oldBin)
oldError = mcHists[mc].GetBinError(oldBin)
newHist.SetBinContent(newBin, oldContent)
newHist.SetBinError(newBin, oldError)
newHist.Write()
lowerEdgeIdx += newBins
output.Write()
temp.Close()
temp = output
discList = []
for key in temp.GetListOfKeys():
from array import array
import math
from SHyFT.ROOTWrap import ROOT
import sys
xvals = []
yvals = []
with open(sys.argv[1]) as f:
for line in f:
line = line.rstrip('\n')
splitted = line.split(' ')
if (len(splitted) == 2):
xvals.append(float(splitted[0]))
yvals.append(float(splitted[1]))
c = ROOT.TCanvas("c", "Log likelihood scan", 200, 10, 700, 500)
g = ROOT.TGraph(len(xvals), array('d', xvals), array('d', yvals))
f = ROOT.TF1("fun", "[2] * x * x + [1] * x + [0]")
h = ROOT.TF1("fun", "[2] * (x - [0])^2 / ([1] ^ 2)")
g.Fit(f)
g.Draw("AL")
g.Fit(h)
g.Draw("A")
c.Update()
c.Print("fitpara.pdf")
print len(xvals)
print len(yvals)
peak = f.GetMinimum(xvals[0], xvals[-1], 1E-10, 100000)
maxx = f.GetMinimumX(xvals[0], xvals[-1], 1E-10, 100000)
print "f(%.4f) = %.4f" % (maxx, peak)
derivative2 = f.Derivative2(maxx)
# strong reference the TFile to keep ROOT from reaping it
fileRefs = []
for arg in args:
argSplit = arg.split(':')
if len(argSplit) == 1:
fileGlob = arg
plotRegex = '.*'
elif len(argSplit) == 2:
fileGlob = argSplit[0]
plotRegex = argSplit[1]
fileList = glob.glob(fileGlob)
plotRegex = re.compile(plotRegex)
for oneFile in fileList:
if debug:
print "Examining %s" % oneFile
rootFile = ROOT.TFile(oneFile, 'READONLY')
ROOT.gDirectory.Cd("")
fileRefs.append(rootFile)
for key in rootFile.GetListOfKeys():
keyString = key.GetName()
if plotRegex.search(keyString) != None:
if debug:
print " Examining %s" % keyString
plotsToPrint.append((keyString, rootFile.Get(keyString)))
totalSum = 0.0
maxKeyLen = 0
for key, _ in plotsToPrint:
maxKeyLen = max(maxKeyLen, len(key))
def getHist(tfile, root_dir, name):
readname = ''
if root_dir != '':
readname = root_dir + '/'
readname += name
#print 'getting ',readname
th1 = tfile.Get(readname)
try:
if th1.IsZombie():
print 'no such histogram!'
return None
except ReferenceError, e:
print "Failed to read %s" % readname
return None
ROOT.SetOwnership(th1, 1)
return th1
# get a clone of the histogram, considering its location
def getClone(tfile, root_dir, name):
th1 = getHist(tfile, root_dir, name)
cloned = th1.Clone()
cloned.SetDirectory(0)
cloned.Sumw2()
ROOT.SetOwnership(cloned, 1)
return cloned
# substitute the leading part of the name (Top, WJets, etc) by user-defined prefix
# will help to deal with multiple single top templates, for example