Reduce memory footprint of parsed datacard

python/DatacardParser.py

@@ -620,7 +620,7 @@ def parseCard(file, options):
                 raise RuntimeError(
                     "Malformed systematics line %s of length %d: while bins and process lines have length %d" % (lsyst, len(numbers), len(ret.keyline))
                 )
-            errline = dict([(b, {}) for b in ret.bins])
+            errline = {b: {} for b in ret.bins}
             nonNullEntries = 0
             for (b, p, s), r in zip(ret.keyline, numbers):
                 if "/" in r:  # "number/number"
@@ -632,8 +632,11 @@ def parseCard(file, options):
                             raise ValueError('Found "%s" in the nuisances affecting %s for %s. This would lead to NANs later on, so please fix it.' % (r, p, b))
                 else:
                     if r == "-" * len(r):
-                        r = 0.0
-                    errline[b][p] = float(r)
+                        continue
+                    r_float = float(r)
+                    if r_float == 0.0:
+                        continue
+                    errline[b][p] = r_float
                     # values of 0.0 are treated as 1.0; scrap negative values.
                     if pdf not in ["trG", "dFD", "dFD2"] and errline[b][p] < 0:
                         raise ValueError('Found "%s" in the nuisances affecting %s in %s. This would lead to NANs later on, so please fix it.' % (r, p, b))
@@ -674,7 +677,7 @@ def parseCard(file, options):
             syst2.append((lsyst, nofloat, pdf, args, errline))
             continue
         for (b, p, s) in ret.keyline:
-            r = errline[b][p]
+            r = errline[b].get(p, 0.0)
             nullEffect = r == 0.0 or (pdf == "lnN" and r == 1.0)
             if not nullEffect and ret.exp[b][p] != 0:
                 nonNullEntries += 1  # is this a zero background?

python/NuisanceModifier.py

@@ -370,10 +370,10 @@ def doSplitNuisance(datacard, args):
                     for p in datacard.exp[b]:
                         if process == "*" or fullmatch(cprocess, p):
                             foundProc = True
-                            if errline[b][p] not in [0.0, 1.0]:
+                            if errline[b].get(p, 0.0) not in [0.0, 1.0]:
                                 doAddNuisance(datacard, [p, b, newname1, pdf, value1, "overwrite"])
                                 doAddNuisance(datacard, [p, b, newname2, pdf, value2, "overwrite"])
-                            errline[b][p] = 0
+                            del errline[b][p]
 
     if not foundProc and channel != "*":
         if "ifexists" not in opts:
@@ -432,7 +432,7 @@ def doFlipNuisance(datacard, args):
                     for p in datacard.exp[b]:
                         if process == "*" or fullmatch(cprocess, p):
                             foundProc = True
-                            if errline[b][p] not in [0.0, 1.0]:
+                            if errline[b].get(p, 0.0) not in [0.0, 1.0]:
                                 if type(errline[b][p]) is list:
                                     if errline[b][p][0] < 1:
                                         if "p2n" in opts:

scripts/combineCards.py

@@ -230,15 +230,15 @@ def compareParamSystLines(a, b):
                             constraint_terms.append(line)
                     warnings.warn(warning_message, RuntimeWarning)
                     break
-                if type(errline[b][p]) == list:
+                if errline[b].get(p, 0.0) == 0.0:
+                    r = "-"
+                elif type(errline[b][p]) == list:
                     r = "%s/%s" % (
                         FloatToString(errline[b][p][0]),
                         FloatToString(errline[b][p][1]),
                     )
                 elif type in ("lnN", "gmM"):
                     r = "%s" % FloatToString(errline[b][p])
-                if errline[b][p] == 0:
-                    r = "-"
                 if len(r) > cmax:
                     cmax = len(r)  # get max col length, as it's more tricky to do it later with a map
                 systeffect[bout][p] = r

test/datacardDump.py

@@ -111,7 +111,7 @@
             continue
         # end skip systematics
         for p in DC.exp[b].keys():  # so that we get only self.DC.processes contributing to this bin
-            if errline[b][p] == 0:
+            if errline[b].get(p, 0.0) == 0.0:
                 continue
             if pdf == "gmN":
                 exps[p][1].append(1 / sqrt(pdfargs[0] + 1))

test/systematicsAnalyzer.py

@@ -228,7 +228,7 @@ def addTo(dic, key, l):
             if lsyst in list(check_list.keys()):
                 if [p, b] in check_list[lsyst]:
                     continue
-            if (not pdf == "param") and errline[b][p] == 0:
+            if (not pdf == "param") and errline[b].get(p, 0.0) == 0.0:
                 continue
             if pdf == "gmN":
                 numKeysFound += 1
@@ -273,7 +273,7 @@ def addTo(dic, key, l):
 
             elif "shape" in pdf and MB.isShapeSystematic(b, p, lsyst):
 
-                if errline[b][p] == 0:
+                if errline[b].get(p, 0.0) == 0.0:
                     continue
                 systShapeName = lsyst
                 # vals = []