def write_hist(histfile: Path, name: str, hist: np.array,
bin_edges: np.array) -> None:
if histfile.exists():
raise Exception(f"Error writing {histfile}, already exists.")
if histfile.suffix == ".npz":
np.savez_compressed(histfile, **{
name: hist,
"bin_edges": bin_edges,
})
elif histfile.suffix == ".txt":
np.savetxt(
histfile,
hist,
header="bin edges:\n" + str(bin_edges) + f"\n{name}:",
)
elif histfile.suffix == ".hdf5":
try:
import h5py
with h5py.File(histfile, "w") as f:
f.create_dataset(name,
data=hist,
compression="gzip",
compression_opts=9)
f.create_dataset(
"bin_edges",
data=bin_edges,
compression="gzip",
compression_opts=9,
)
except ModuleNotFoundError:
raise Exception("Please install h5py to write hdf5 files")
elif histfile.suffix == ".root":
import uproot
# TODO: Discard sumw2?
if hist.ndim == 1:
from uproot_methods.classes.TH1 import from_numpy
h = from_numpy([hist, bin_edges])
else:
from uproot_methods.classes.TH2 import from_numpy
h = from_numpy([hist, np.arange(0, hist.shape[0] + 1), bin_edges])
with uproot.create(histfile) as f:
f[name] = h
else:
raise Exception(f"Unknown output format: {histfile.suffix}")
def test_th1(self):
from uproot_methods.classes.TH1 import Methods, _histtype, from_numpy
edges = np.array((0., 1., 2.))
values = np.array([2, 3])
h = from_numpy((values, edges))
assert h.name is None
assert h.numbins == 2
assert h.title == b""
assert h.low == 0
assert h.high == 2
assert h.underflows == 0
assert h.overflows == 0
np.testing.assert_equal(h.edges, edges)
np.testing.assert_equal(h.values, values)
np.testing.assert_equal(h.variances, values**2)
np.testing.assert_equal(h.alledges, [-np.inf] + list(edges) + [np.inf])
np.testing.assert_equal(h.allvalues, [0] + list(values) + [0])
np.testing.assert_equal(h.allvariances, [0] + list(values**2) + [0])
np.testing.assert_equal(h.bins, ((0, 1), (1, 2)))
np.testing.assert_equal(h.allbins,
((-np.inf, 0), (0, 1), (1, 2), (2, np.inf)))
assert h.interval(0) == (-np.inf, 0)
assert h.interval(1) == (0, 1)
assert h.interval(2) == (1, 2)
assert h.interval(3) == (2, np.inf)
assert h.interval(-1) == h.interval(3)
def test_histogram(self):
np = TestHistogram.NUMPY_LIB
data = np.array([2,3,4,5,6,7], dtype=np.float32)
data[data<2] = 0
weights = np.ones_like(data, dtype=np.float32)
w, w2, e = self.ha.histogram_from_vector(data, weights, np.array([0,1,2,3,4,5], dtype=np.float32))
npw, npe = np.histogram(data, np.array([0,1,2,3,4,5]))
hr = from_numpy((w, e))
f = uproot.recreate("test.root")
f["hist"] = hr
data = np.random.normal(size=10000)
data = np.array(data, dtype=np.float32)
weights = np.ones_like(data, dtype=np.float32)
w, w2, e = self.ha.histogram_from_vector(data, weights, np.linspace(-1,1,100, dtype=np.float32))
hr = from_numpy((w, e))
f["hist2"] = hr
f.close()
def to_th1(hdict, name):
content = np.array(hdict.contents)
content_w2 = np.array(hdict.contents_w2)
edges = np.array(hdict.edges)
#remove inf/nan just in case
content[np.isinf(content)] = 0
content_w2[np.isinf(content_w2)] = 0
content[np.isnan(content)] = 0
content_w2[np.isnan(content_w2)] = 0
#update the error bars
centers = (edges[:-1] + edges[1:]) / 2.0
th1 = from_numpy((content, edges))
th1._fName = name
th1._fSumw2 = np.array(hdict.contents_w2)
th1._fTsumw2 = np.array(hdict.contents_w2).sum()
th1._fTsumwx2 = np.array(hdict.contents_w2 * centers).sum()
return th1
def save_shapes(var, hist, edges, args):
def get_vwname(v, w):
vwname = ''
if 'nominal' in v:
if 'off' in w: return ()
elif 'nominal' in w:
vwname = 'nominal'
elif '_up' in w:
vwname = w.replace('_up', 'Up').replace('wgt_', '')
elif '_down' in w:
vwname = w.replace('_down', 'Down').replace('wgt_', '')
else:
if 'nominal' not in w: return ()
elif '_up' in v:
vwname = v.replace('_up', 'Up')
elif '_down' in v:
vwname = v.replace('_down', 'Down')
return vwname
hist = hist[var.name]
centers = (edges[:-1] + edges[1:]) / 2.0
bin_columns = [c for c in hist.columns if 'bin' in c]
sumw2_columns = [c for c in hist.columns if 'sumw2' in c]
data_names = [n for n in hist.s.unique() if 'data' in n]
for c in args['channels']:
for r in args['regions']:
out_fn = f'combine_new/shapes_{c}_{r}_{args["year"]}_{args["label"]}.root'
out_file = uproot.recreate(out_fn)
data_obs_hist = np.zeros(len(bin_columns), dtype=float)
data_obs_sumw2 = np.zeros(len(sumw2_columns), dtype=float)
for v in hist.v.unique():
for w in hist.w.unique():
vwname = get_vwname(v, w)
if vwname == '': continue
if vwname == 'nominal':
data_obs = hist[hist.s.isin(data_names) & (hist.r == r)
& (hist.c == c)]
data_obs_hist = data_obs[bin_columns].sum(
axis=0).values
data_obs_sumw2 = data_obs[sumw2_columns].sum(
axis=0).values
mc_hist = hist[~hist.s.isin(data_names) & (hist.v == v) &
(hist.w == w) & (hist.r == r) &
(hist.c == c)]
for s in mc_hist.s.unique():
if s in grouping.keys():
mc_hist.loc[hist.s == s, 'group'] = grouping[s]
mc_hist = mc_hist.groupby('group').aggregate(
np.sum).reset_index()
for g in mc_hist.group.unique():
histo = mc_hist[mc_hist.group ==
g][bin_columns].values[0]
if len(histo) == 0: continue
sumw2 = mc_hist[mc_hist.group ==
g][sumw2_columns].values[0]
rname = r.replace('-', '_')
name = f'{rname}_{g}_{vwname}'
th1 = from_numpy([histo, edges])
th1._fName = name
th1._fSumw2 = np.array(sumw2)
th1._fTsumw2 = np.array(sumw2).sum()
th1._fTsumwx2 = np.array(sumw2 * centers).sum()
out_file[f'{g}_{vwname}'] = th1
th1_data = from_numpy([data_obs_hist, edges])
th1_data._fName = 'data_obs'
th1_data._fSumw2 = np.array(data_obs_sumw2)
th1_data._fTsumw2 = np.array(data_obs_sumw2).sum()
th1_data._fTsumwx2 = np.array(data_obs_sumw2 * centers).sum()
out_file['data_obs'] = th1_data
out_file.close()
