def test_append(self):
"""Also tests layered=False"""
cut = ROOT.TCut('bar<0')
self.splice.make_splice('bar', cut)
cut = ROOT.TCut('bar>=0')
# this should add the rest of the entries
tree = self.splice.make_splice('bar', cut, append=True)
self.assertEqual(self.nentries, nplotted(tree, 'bar'))
def test_make_splice(self):
cut = ROOT.TCut('foo>10')
nexpected = nplotted(self.splice.reset(), 'foo', cut)
tree = self.splice.make_splice('foo_gt_10', cut)
self.assertEqual(nexpected, nplotted(tree, 'foo'))
cut = ROOT.TCut('bar<0')
nexpected = nplotted(self.splice.reset(), 'bar', cut)
tree = self.splice.make_splice('nve_bar', cut)
self.assertEqual(nexpected, nplotted(tree, 'bar'))
def test_elistarray_splice(self):
# entries with 5 elements in data are guaranteed to pass, for
# other entries, it passes when the passing element is < size,
# which happens 3/5 times (when it is b/w 0-2), it is always
# found for size = 5. Therefore, it is found for at least 4/5
# entries.
cut = ROOT.TCut('(data[]%5)>3')
tree = self.splice.make_splice('data_mod_5',
cut,
listtype='entrylistarray')
expected = nplotted(tree, 'data')
# print self.ndata, self.nentries, expected, 4*self.nentries/5
self.assertLess(expected, self.ndata)
self.assertLess(expected, self.nentries)
self.assertGreater(expected, 4 * self.nentries / 5)
def _draw_match(fname, hists):
for pair in hists:
myc = None
for obj in pair:
if fnmatchcase(obj.GetName(), options.dump):
xaxis, yaxis = obj.GetXaxis(), obj.GetYaxis()
xaxis.SetTitle(get_label(xaxis.GetTitle()))
yaxis.SetTitle(get_label(yaxis.GetTitle()))
yaxis.SetTitleOffset(1.25)
if not myc:
myc = ROOT.TCanvas('myc', '', 800, 500)
myc.cd()
obj.Draw()
else:
obj.Draw('same')
if myc:
myc.Print(fname)
del myc
def setUpModule():
rfile = ROOT.TFile.Open('/tmp/testtree.root', 'recreate')
tree = ROOT.TTree('testtree', '')
foo = np.array([0], dtype=np.float32)
tree.Branch("foo", foo, "foo/F")
bar = np.array([0], dtype=np.double)
tree.Branch("bar", bar, "bar/D")
baz = np.array([0], dtype=np.int32)
tree.Branch("baz", baz, "baz/I")
sz = np.array([0], dtype=np.int32)
tree.Branch("sz", sz, "sz/I")
data = np.array([1., 2., 3., 4., 5.], dtype=np.float32)
tree.Branch("data", data, "data[sz]/F")
for i in range(1000):
foo[0] = np.random.lognormal(mean=np.pi)
bar[0] = np.random.normal(loc=0)
baz[0] = np.random.binomial(100, 0.3)
data += i
sz[0] = 3 + i % 3
tree.Fill()
tree.Write()
rfile.Close()
def setUp(self):
"""Make ROOT files with some contents.
Structure:
/tmp/test_Rdir{0,1}.root:/dira/hista
/tmp/test_Rdir{0,1}.root:/dirb/histb
/tmp/test_Rdir{0,1}.root:/dirc/histx
/tmp/test_Rdir{0,1}.root:/dirc/dird/histy
/tmp/test_Rdir{0,1}.root:/dirc/dird/dire/histz
/tmp/test_Rdir{0,1}.root:/hist{0,1,2}
"""
self.fnames, self.rfiles = [], []
for i in range(2):
self.fnames.append('/tmp/test_Rdir{}.root'.format(i))
self.rfiles.append(TFile.Open(self.fnames[-1], 'recreate'))
for f in self.rfiles:
rdir = f.mkdir('dira')
rdir.WriteTObject(ROOT.TH1C('hista', '', 10, 0, 10))
rdir = f.mkdir('dirb')
rdir.WriteTObject(ROOT.TH1C('histb', '', 10, 0, 10))
rdir = f.mkdir('dirc/dird/dire')
rdir.cd()
gDirectory.WriteTObject(ROOT.TH1C('histx', '', 10, 0, 10))
rdir.cd('dird')
gDirectory.WriteTObject(ROOT.TH1C('histy', '', 10, 0, 10))
rdir.cd('dird/dire')
gDirectory.WriteTObject(ROOT.TH1C('histz', '', 10, 0, 10))
for i in range(3):
f.WriteTObject(ROOT.TH1C('hist{}'.format(i), '', 10, 0, 10))
for f in self.rfiles:
f.Write()
f.Close()
ihists = get_hists(['file'], rfileconf, pathtool, robj_t=ROOT.TH1)
# triangular matrix indices for use w/ both cov matrices & scatter plots
import numpy as np
dims = ihists['file'][0].GetXaxis().GetNbins() - 1 # nvars - 1 in corrn matrix
opts = np.empty(shape=(dims, dims), dtype=object)
tril = np.tril_indices(dims)
triu = np.triu_indices(dims)
## covariance matrices
from utils import get_label
if options.lcorrns:
matrices = {}
for transform in transforms:
corrn = [
ROOT.TH2D(transform+'_sig', 'Correlation matrix after {} transform (sig)'
.format(transforms[transform]), dims, 0, dims, dims, 0, dims),
ROOT.TH2D(transform+'_bkg', 'Correlation matrix after {} transform (bkg)'
.format(transforms[transform]), dims, 0, dims, dims, 0, dims)
]
for i, idx in enumerate(zip(*triu)):
hist = [
sig_hists[transform+'_corr'][i*2],
bkg_hists[transform+'_corr'][i*2]
]
if idx[0] == idx[1]: # set bin label using diagonal
for i in xrange(len(hist)):
name = hist[i].GetName()
if i == 0:
name = name[5:name.find('_Signal')]
def setUp(self):
self.hist1 = ROOT.TH1D('hist1', '', 20, -3, 3)
self.hist1.FillRandom('gaus', 1000)
def setUp(self):
self.nbins = 20
self.hist = ROOT.TH1F('hist', '', self.nbins, -3, 3)
self.hist.FillRandom('gaus', 1000)
axes.set_ylabel('Background rejection efficiency')
axes.xaxis.set_label_coords(0.9, -0.05)
for key, hist in roc.iteritems():
info = hist_info(hist)
hist = pycopy(Hist, ROOT.TH1, hist, *info[0], **info[1])
line = rplt.hist(hist, stacked=False)
print type(line), line
axes.legend(fontsize=10, numpoints=1, frameon=False, ncol=3)
if doprint:
pp.savefig()
pp.close()
elif not options.batch:
plt.show()
else:
canvas = ROOT.TCanvas('canvas', '', 800, 600)
if doprint:
canvas.Print('{}_my_ROC_curves.pdf['.format(prefix))
cols = (ROOT.kAzure, ROOT.kRed, ROOT.kBlack)
legend = ROOT.TLegend(0.12, 0.15, 0.8, 0.6)
legend.SetHeader('MVA classifiers')
legend.SetBorderSize(0)
legend.SetFillStyle(0)
ROOT.gStyle.SetOptStat(False)
from utils import th1integral, distance
for i, roc in enumerate(rocs):
coln = 0
for cl, hist in roc.iteritems():
# metrics
if conf.read() > 0:
session = conf.get_session_config(options.session)
else:
sys.exit('No usable sessions in config file')
if not session:
sys.exit('Couldn\'t find session in config file')
print '::: Training {} MVAs: {}\n{}'.format(len(session.methods),
session.methods, '='*50)
print session
print ':::'
from fixes import ROOT
ROOT.gROOT.SetBatch(True)
# files & trees
tree_s = ROOT.TChain(options.sigtree)
map(lambda f: tree_s.Add(f), session.sig_file)
tree_b = ROOT.TChain(options.bkgtree)
map(lambda f: tree_b.Add(f), session.bkg_file)
if not tree_s or not tree_b:
sys.exit('Unable to read input trees.')
nentries_s = tree_s.GetEntries(str(session.cut_sig))
nentries_b = tree_b.GetEntries(str(session.cut_bkg))
# NOTE: normalise: create option like
# nTrain_Signal=num:nTrain_Background=num:...
if options.norm:
size = nentries_b if nentries_s > nentries_b else nentries_s
def get_estimate(mva, cut, tree=tree):
cut = ROOT.TCut('{}>{}'.format(mva, cut))
# NOTE: only process 200,000 entries
tree.Draw('{0}>>h{0}'.format(mva), cut, 'goff', options.maxentries)
hist = ROOT.gDirectory.Get('h{}'.format(mva))
return hist.GetEntries()
def test_get_splice(self):
cut = ROOT.TCut('sz==4')
self.splice.make_splice('sz_eq_4', cut)
nexpected = nplotted(self.splice.reset(), 'sz', cut)
tree = self.splice.get_splice('sz_eq_4')
self.assertEqual(nexpected, nplotted(tree, 'sz'))
def test_set_splice(self):
cut = ROOT.TCut('bar>0')
self.splice.make_splice('pve_bar', cut)
nexpected = nplotted(self.splice.reset(), 'bar', cut)
tree = self.splice.set_splice(self.splice.elists['pve_bar'])
self.assertEqual(nexpected, nplotted(tree, 'bar'))
def test_reset(self):
self.splice.make_splice('sz_gt_5', ROOT.TCut('sz>5')) # zero
self.assertEqual(nplotted(self.splice.reset(), 'sz'), self.nentries)
def test_get_entries(self):
tree = self.splice.make_splice('sz_gt_3', ROOT.TCut('sz>3')) # 2/3
self.assertEqual(self.splice.get_entries(), nplotted(tree, 'sz'))