Orange.statistics.util: Add nanunique method

Author: nikicc

Orange/statistics/util.py

@@ -280,6 +280,7 @@ def mean(x):
     n_values = np.prod(x.shape)
     return np.sum(x.data) / n_values
 
+
 def nanmean(x, axis=None):
     """ Equivalent of np.nanmean that supports sparse or dense matrices. """
     def nanmean_sparse(x):
@@ -297,6 +298,7 @@ def nanmean_sparse(x):
     else:
         raise NotImplementedError
 
+
 def unique(x, return_counts=False):
     """ Equivalent of np.unique that supports sparse or dense matrices. """
     if not sp.issparse(x):
@@ -316,3 +318,10 @@ def unique(x, return_counts=False):
         if explicit_zeros:
             return r
         return np.insert(r, 0, 0)
+
+
+def nanunique(x):
+    """ Return unique values while disregarding missing (np.nan) values.
+    Supports sparse or dense matrices. """
+    r = unique(x)
+    return r[~np.isnan(r)]

Orange/tests/test_statistics.py

@@ -5,7 +5,7 @@
 from scipy.sparse import csr_matrix
 
 from Orange.statistics.util import bincount, countnans, contingency, stats, \
-    nanmin, nanmax, unique, mean, nanmean
+    nanmin, nanmax, unique, nanunique, mean, nanmean
 
 
 class TestUtil(unittest.TestCase):
@@ -145,6 +145,13 @@ def test_unique_explicit_zeros(self):
             unique(x2, return_counts=True),
         )
 
+    def test_nanunique(self):
+        x = csr_matrix(np.array([0, 1, 1, np.nan]))
+        np.testing.assert_array_equal(
+            nanunique(x),
+            np.array([0, 1])
+        )
+
     def test_mean(self):
         for X in self.data:
             X_sparse = csr_matrix(X)