Fix: Import problem numba, prints removed

README.rst

@@ -115,6 +115,9 @@ categories:
   - for binary probability forecasts: ``brier_score``
   - for threshold exceedances with an ensemble forecast: ``threshold_brier_score``
 
+* Mean CRPS with decomposition into reliability, resolution, uncertainty:
+  - for an ensemble via the Hersbach method: ``mean_crps``
+
 All functions are robust to missing values represented by the floating
 point value ``NaN``.
 

properscoring/_gufuncs.py

@@ -124,34 +124,28 @@ def _mean_crps_rel_pot(observations, forecasts, mean_crps, reliability, crps_pot
         for i in range(N):
             e = forecasts[i, :]
             x = observations[i]
-            print('i', i, 'e', e, 'x', x, 'Alpha', alpha_k, 'Beta', beta_k)
+            #print('i', i, 'e', e, 'x', x, 'Alpha', alpha_k, 'Beta', beta_k)
             if (k > 0) and (k < M):
                 if e[k] < x:
-                    print('case 1')
                     alpha_k += e[k] - e[k-1]
                     beta_k += 0
                 elif (e[k-1] < x) and (x <= e[k]):
-                    print('case 2')
                     alpha_k += x - e[k-1]
                     beta_k += e[k] - x
                 elif x <= e[k-1]:
-                    print('case 3')
                     alpha_k += 0
                     beta_k += e[k] - e[k-1]
             elif (k == 0):
                 if x < e[0]:
-                    print('case 4')
                     alpha_k += 0
                     beta_k += e[0] - x
             elif (k == M):
                 if e[M-1] < x:
-                    print('case 5')
                     alpha_k += x - e[M-1]
                     beta_k += 0
         alpha_k /= N
         beta_k /= N
 
-        print('k', k, 'p_k', p_k, 'Alpha', alpha_k, 'Beta', beta_k)
         mea += alpha_k * p_k**2 + beta_k * (1 - p_k)**2
 
         g_k = alpha_k + beta_k

properscoring/_mean_crps.py

@@ -8,7 +8,7 @@ Created on Wed Nov 13 16:37:32 2019
 import numpy as np
 from scipy.stats import rankdata
 
-from properscoring._gufuncs import _uncertainty_comp, _mean_crps_rel_pot
+
 
 
 #def ranks(obs, ensemble):#, mask=):
@@ -33,7 +33,14 @@ from properscoring._gufuncs import _uncertainty_comp, _mean_crps_rel_pot
 #        ranks[ties==tie[i]]=[np.random.randint(index[j],index[j]+tie[i]+1,tie[i])[0] for j in range(len(index))]
 #    return ranks
 
+try:
+    from properscoring._gufuncs import _uncertainty_comp, _mean_crps_rel_pot
+except ImportError as exc:
+    def _make_import_error(a):
+        raise ImportError('Numba is not installed.')
+    _uncertainty_comp = lambda x: _make_import_error(x)
 
+    
 def _mean_crps_hersbach(observations, forecasts):
     """mean CRPS with reliability, resolution and uncertainty