Add support for inhomogenous error breakdowns (#265)

* add support for inhomogenous errors
* Disable Pylint too-many-branches for make_dict
* Update type in docstring for sanitize_value

---------

Co-authored-by: Graeme Watt <[email protected]>

docs/usage.rst

@@ -370,5 +370,7 @@ multiple dependent variables and a (different) subset of the bins has missing co
 In this case the uncertainties should be set to zero for the missing bins with a non-numeric central value like ``'-'``.
 The warning message can be suppressed by passing an optional argument ``zero_uncertainties_warning=False`` when
 defining an instance of the ``Variable`` class.
+Furthermore, note that `None` can be used to suppress the uncertainty for individual bins in cases where
+the uncertainty components may only apply to a subset of the values.
 
-.. _`Uncertainties`: https://hepdata-submission.readthedocs.io/en/latest/data_yaml.html#uncertainties
+.. _`Uncertainties`: https://hepdata-submission.readthedocs.io/en/latest/data_yaml.html#uncertainties

hepdata_lib/__init__.py

@@ -235,6 +235,7 @@ def add_uncertainty(self, uncertainty):
         self.uncertainties.append(uncertainty)
 
     def make_dict(self):
+        # pylint: disable=too-many-branches
         """
         Return all data in this Variable as a dictionary.
 
@@ -277,6 +278,8 @@ def make_dict(self):
             # if at least one of the uncertainties is not zero.
             if nonzero_uncs[i]:
                 for unc in self.uncertainties:
+                    if unc.values[i] is None:
+                        continue
                     if unc.is_symmetric:
                         valuedict['errors'].append({
                             "symerror":

hepdata_lib/helpers.py

@@ -258,7 +258,8 @@ def any_uncertainties_nonzero(uncertainties, size):
     for unc in uncertainties:
 
         # Treat one-sided uncertainties as
-        values = np.array(unc.values)
+        tmp = 0 if unc.is_symmetric else (0,0)
+        values = np.array([tmp if v is None else v for v in unc.values])
         values[values.astype(str)==''] = 0
         values = values.astype(float)
 
@@ -273,15 +274,17 @@ def sanitize_value(value):
     Handle conversion of input types for internal storage.
 
     :param value: User-side input value to sanitize.
-    :type value: string, int, or castable to float
+    :type value: string, int, NoneType, or castable to float
 
-    Strings and integers are left alone,
+    Strings, integers and None are left alone,
     everything else is converted to float.
     """
     if isinstance(value,str):
         return value
     if isinstance(value,int):
         return value
+    if value is None:
+        return value
     return float(value)
 
 

tests/test_uncertainty.py

@@ -115,3 +115,18 @@ def test_zero_uncertainties(self):
         # Check that 'errors' key is missing only if zero uncertainties
         self.assertTrue(all('errors' in dictionary['values'][i] for i in [0, 1, 3]))
         self.assertTrue('errors' not in dictionary['values'][2])
+
+    def test_inhomogenous_uncertainties(self):
+        '''Test cases where an uncertainty only applies to a subset of the bins'''
+        var = Variable("testvar", is_independent=False, is_binned=False, values=[1,2,3],
+                       zero_uncertainties_warning=False)
+        unc_a = Uncertainty('errorA', is_symmetric=True)
+        unc_a.values = [ 0.1, 0.2, None ]
+        var.add_uncertainty(unc_a)
+        unc_b = Uncertainty('errorB', is_symmetric=True)
+        unc_b.values = [ 0.1, 0.2, 0.3 ]
+        var.add_uncertainty(unc_b)
+        dictionary = var.make_dict()
+        self.assertTrue(len([ errs['label'] for i in [0,1,2] \
+                                            for errs in dictionary['values'][i]['errors'] \
+                                            if errs['label'] == 'errorA'])==2)