def convert_junc_txt_component(juncFilePath, uncFile):
name, layout, pars, nBinnedVars, \
nBinColumns, nEvalVars, formula, \
nParms, columns, dtypes = _parse_jme_formatted_file(juncFilePath,
interpolatedFunc=True,
parmsFromColumns=True,
jme_f=uncFile)
temp = _build_standard_jme_lookup(name, layout, pars, nBinnedVars, nBinColumns,
nEvalVars, formula, nParms, columns, dtypes,
interpolatedFunc=True)
wrapped_up = {}
for key, val in temp.items():
newkey = (key[0], 'jec_uncertainty_lookup')
vallist = list(val)
vals, names = vallist[-1]
knots = vals[0:len(vals):3]
downs = vals[1:len(vals):3]
ups = vals[2:len(vals):3]
downs = np.array([down.flatten() for down in downs])
ups = np.array([up.flatten() for up in ups])
for knotv in knots:
knot = np.unique(knotv.flatten())
if knot.size != 1:
raise Exception('Multiple bin low edges found')
knots = np.array([np.unique(k.flatten())[0] for k in knots])
vallist[2] = ({'knots': knots, 'ups': ups.T, 'downs': downs.T}, vallist[2][-1])
vallist = vallist[:-1]
wrapped_up[newkey] = tuple(vallist)
return wrapped_up
def _evaluate(self, *args):
""" uncertainties = f(args) """
bin_vals = {
argname: args[self._dim_args[argname]]
for argname in self._dim_order
}
eval_vals = {
argname: args[self._eval_args[argname]]
for argname in self._eval_vars
}
#lookup the bins that we care about
dim1_name = self._dim_order[0]
dim1_indices = np.clip(
np.searchsorted(
self._bins[dim1_name], bin_vals[dim1_name], side='right') - 1,
0, self._bins[dim1_name].size - 2)
#get clamp values and clip the inputs
outs = np.ones(shape=(args[0].size, 2), dtype=np.float)
for i in np.unique(dim1_indices):
mask = np.where(dim1_indices == i)
vals = np.clip(eval_vals[self._eval_vars[0]][mask],
self._eval_knots[0], self._eval_knots[-1])
outs[:, 0][mask] += self._eval_ups[i](vals)
outs[:, 1][mask] -= self._eval_downs[i](vals)
return outs
def masked_bin_eval(dim1_indices, dimN_bins, dimN_vals):
dimN_indices = np.empty_like(dim1_indices)
for i in np.unique(dim1_indices):
idx = np.where(dim1_indices == i)
dimN_indices[idx] = np.clip(
np.searchsorted(dimN_bins[i], dimN_vals[idx], side='right') - 1, 0,
len(dimN_bins[i]) - 2)
return dimN_indices
def convert_btag_csv_file(csvFilePath):
btag_f = open(csvFilePath)
nameandcols = btag_f.readline().split(';')
btag_f.close()
name = nameandcols[0].strip()
columns = nameandcols[1].strip()
columns = [column.strip() for column in columns.split(',')]
corrections = np.genfromtxt(csvFilePath,
dtype=None,
names=tuple(columns),
converters={1: lambda s: s.strip(),
2: lambda s: s.strip(),
10: lambda s: s.strip(' "')},
delimiter=',',
skip_header=1,
unpack=True,
encoding='ascii'
)
all_names = corrections[[columns[i] for i in range(4)]]
labels = np.unique(corrections[[columns[i] for i in range(4)]])
wrapped_up = {}
for label in labels:
etaMins = np.unique(corrections[np.where(all_names == label)][columns[4]])
etaMaxs = np.unique(corrections[np.where(all_names == label)][columns[5]])
etaBins = np.union1d(etaMins, etaMaxs)
ptMins = np.unique(corrections[np.where(all_names == label)][columns[6]])
ptMaxs = np.unique(corrections[np.where(all_names == label)][columns[7]])
ptBins = np.union1d(ptMins, ptMaxs)
discrMins = np.unique(corrections[np.where(all_names == label)][columns[8]])
discrMaxs = np.unique(corrections[np.where(all_names == label)][columns[9]])
discrBins = np.union1d(discrMins, discrMaxs)
vals = np.zeros(shape=(len(discrBins) - 1, len(ptBins) - 1, len(etaBins) - 1),
dtype=corrections.dtype[10])
for i, eta_bin in enumerate(etaBins[:-1]):
for j, pt_bin in enumerate(ptBins[:-1]):
for k, discr_bin in enumerate(discrBins[:-1]):
this_bin = np.where((all_names == label) &
(corrections[columns[4]] == eta_bin) &
(corrections[columns[6]] == pt_bin) &
(corrections[columns[8]] == discr_bin))
vals[k, j, i] = corrections[this_bin][columns[10]][0]
label_decode = []
for i in range(len(label)):
label_decode.append(label[i])
if isinstance(label_decode[i], bytes):
label_decode[i] = label_decode[i].decode()
else:
label_decode[i] = str(label_decode[i])
str_label = '_'.join([name] + label_decode)
feval_dim = btag_feval_dims[label[0]]
wrapped_up[(str_label, 'dense_evaluated_lookup')] = (vals, (etaBins, ptBins, discrBins), tuple(feval_dim))
return wrapped_up
def extract_json_histo_structure(parselevel, axis_names, axes):
if 'value' in parselevel.keys():
return
name = list(parselevel)[0].split(':')[0]
bins_pairs = [
key.split(':')[-1].strip('[]').split(',') for key in parselevel.keys()
]
bins = []
for pair in bins_pairs:
bins.extend([float(val) for val in pair])
bins.sort()
bins = np.unique(np.array(bins))
axis_names.append(name.encode())
axes[axis_names[-1]] = bins
extract_json_histo_structure(parselevel[list(parselevel)[0]], axis_names,
axes)
def _build_standard_jme_lookup(name,
layout,
pars,
nBinnedVars,
nBinColumns,
nEvalVars,
formula,
nParms,
columns,
dtypes,
interpolatedFunc=False):
#the first bin is always usual for JECs
#the next bins may vary in number, so they're jagged arrays... yay
bins = {}
offset_col = 0
offset_name = 1
bin_order = []
for i in range(nBinnedVars):
binMins = None
binMaxs = None
if i == 0:
binMins = np.unique(pars[columns[0]])
binMaxs = np.unique(pars[columns[1]])
bins[layout[i + offset_name]] = np.union1d(binMins, binMaxs)
else:
counts = np.zeros(0, dtype=np.int)
allBins = np.zeros(0, dtype=np.double)
for binMin in bins[bin_order[0]][:-1]:
binMins = np.unique(pars[np.where(
pars[columns[0]] == binMin)][columns[i + offset_col]])
binMaxs = np.unique(pars[np.where(
pars[columns[0]] == binMin)][columns[i + offset_col + 1]])
theBins = np.union1d(binMins, binMaxs)
allBins = np.append(allBins, theBins)
counts = np.append(counts, theBins.size)
bins[layout[i + offset_name]] = awkward.JaggedArray.fromcounts(
counts, allBins)
bin_order.append(layout[i + offset_name])
offset_col += 1
#skip nvars to the variable columns
#the columns here define clamps for the variables defined in columns[]
# ----> clamps can be different from bins
# ----> if there is more than one binning variable this array is jagged
# ----> just make it jagged all the time
binshapes = tuple([bins[thebin].size - 1 for thebin in bin_order])
clamp_mins = {}
clamp_maxs = {}
var_order = []
offset_col = 2 * nBinnedVars + 1
offset_name = nBinnedVars + 2
jagged_counts = np.ones(bins[bin_order[0]].size - 1, dtype=np.int)
if len(bin_order) > 1:
jagged_counts = np.maximum(
bins[bin_order[1]].counts - 1,
0) #need counts-1 since we only care about Nbins
for i in range(nEvalVars):
var_order.append(layout[i + offset_name])
if not interpolatedFunc:
clamp_mins[layout[i +
offset_name]] = awkward.JaggedArray.fromcounts(
jagged_counts,
np.atleast_1d(pars[columns[i + offset_col]]))
clamp_maxs[layout[i +
offset_name]] = awkward.JaggedArray.fromcounts(
jagged_counts,
np.atleast_1d(pars[columns[i + offset_col +
1]]))
offset_col += 1
#now get the parameters, which we will look up with the clamped values
parms = []
parm_order = []
offset_col = 2 * nBinnedVars + 1 + (interpolatedFunc
== False) * 2 * nEvalVars
for i in range(nParms):
parms.append(
awkward.JaggedArray.fromcounts(jagged_counts,
pars[columns[i + offset_col]]))
parm_order.append('p%i' % (i))
wrapped_up = {}
wrapped_up[(name, 'jme_standard_function')] = (formula, (bins, bin_order),
(clamp_mins, clamp_maxs,
var_order), (parms,
parm_order))
return wrapped_up
def convert_effective_area_file(eaFilePath):
ea_f = open(eaFilePath, 'r')
layoutstr = ea_f.readline().strip().strip('{}')
ea_f.close()
name = eaFilePath.split('/')[-1].split('.')[0]
layout = layoutstr.split()
if not layout[0].isdigit():
raise Exception(
'First column of Effective Area File Header must be a digit!')
#setup the file format
nBinnedVars = int(layout[0])
nBinColumns = 2 * nBinnedVars
nEvalVars = int(layout[nBinnedVars + 1])
minMax = ['Min', 'Max']
columns = []
dtypes = []
offset = 1
for i in range(nBinnedVars):
columns.extend(['%s%s' % (layout[i + offset], mm) for mm in minMax])
dtypes.extend(['<f8', '<f8'])
offset += nBinnedVars + 1
for i in range(nEvalVars):
columns.append('%s' % (layout[i + offset]))
dtypes.append('<f8')
pars = np.genfromtxt(eaFilePath,
dtype=tuple(dtypes),
names=tuple(columns),
skip_header=1,
unpack=True,
encoding='ascii')
bins = {}
offset_col = 0
offset_name = 1
bin_order = []
for i in range(nBinnedVars):
binMins = None
binMaxs = None
if i == 0:
binMins = np.unique(pars[columns[0]])
binMaxs = np.unique(pars[columns[1]])
bins[layout[i + offset_name]] = np.union1d(binMins, binMaxs)
else:
counts = np.zeros(0, dtype=np.int)
allBins = np.zeros(0, dtype=np.double)
for binMin in bins[bin_order[0]][:-1]:
binMins = np.unique(pars[np.where(
pars[columns[0]] == binMin)][columns[i + offset_col]])
binMaxs = np.unique(pars[np.where(
pars[columns[0]] == binMin)][columns[i + offset_col + 1]])
theBins = np.union1d(binMins, binMaxs)
allBins = np.append(allBins, theBins)
counts = np.append(counts, theBins.size)
bins[layout[i + offset_name]] = awkward.JaggedArray.fromcounts(
counts, allBins)
bin_order.append(layout[i + offset_name])
offset_col += 1
# again this is only for one dimension of binning, fight me
# we can figure out a 2D EA when we get there
offset_name += 1
wrapped_up = {}
lookup_type = 'dense_lookup'
dims = bins[layout[1]]
for i in range(nEvalVars):
ea_name = '_'.join([name, columns[offset_name + i]])
values = pars[columns[offset_name + i]]
wrapped_up[(ea_name, lookup_type)] = (values, dims)
return wrapped_up
def __init__(self, runs=None, lumis=None):
self.array = np.zeros(shape=(0, 2))
if runs is not None:
self.array = np.unique(np.c_[runs, lumis], axis=0)
def __call__(self, runs, lumis):
mask = np.zeros(dtype='bool', shape=runs.shape)
for run in np.unique(runs):
if run in self._masks:
mask |= (np.searchsorted(self._masks[run], lumis)%2==1) & (runs==run)
return mask
