data.util.array_equal: More efficient comparison

Author: janezd

Orange/data/util.py

@@ -99,12 +99,16 @@ def array_equal(a1, a2):
     """array_equal that supports sparse and dense arrays with missing values"""
     if a1.shape != a2.shape:
         return False
-    i1, j1, v1 = sp.find(a1)
-    i2, j2, v2 = sp.find(a2)
-    a1 = v1 if sp.issparse(a1) else a1[i2, j2]
-    a2 = v2 if sp.issparse(a2) else a2[i1, j1]
-    index_equal = set(zip(i1, j1)) == set(zip(i2, j2))
-    return index_equal and np.allclose(a1, a2, equal_nan=True)
+
+    if not (sp.issparse(a1) or sp.issparse(a2)):  # Both dense: just compare
+        return np.allclose(a1, a2, equal_nan=True)
+
+    v1 = np.vstack(sp.find(a1)).T
+    v2 = np.vstack(sp.find(a2)).T
+    if not (sp.issparse(a1) and sp.issparse(a2)):  # Any dense: order indices
+        v1.sort(axis=0)
+        v2.sort(axis=0)
+    return np.allclose(v1, v2, equal_nan=True)
 
 
 def assure_array_dense(a):

Orange/tests/test_util.py

@@ -144,3 +144,17 @@ def test_csc_array_equal(self):
         a2 = sp.csc_matrix(([5, 1, 4], ([1, 0, 0], [2, 0, 2])), shape=(2, 3))
         a2[0, 1] = 0  # explicitly setting to 0
         self.assertTrue(array_equal(a1, a2))
+
+    def test_csc_scr_equal(self):
+        a1 = sp.csc_matrix(([1, 4, 5], ([0, 0, 1], [0, 2, 2])), shape=(2, 3))
+        a2 = sp.csr_matrix(([5, 1, 4], ([1, 0, 0], [2, 0, 2])), shape=(2, 3))
+        self.assertTrue(array_equal(a1, a2))
+
+        a1 = sp.csc_matrix(([1, 4, 5], ([0, 0, 1], [0, 2, 2])), shape=(2, 3))
+        a2 = sp.csr_matrix(([1, 4, 5], ([0, 0, 1], [0, 2, 2])), shape=(2, 3))
+        self.assertTrue(array_equal(a1, a2))
+
+    def test_csc_unordered_array_equal(self):
+        a1 = sp.csc_matrix(([1, 4, 5], [0, 0, 1], [0, 1, 1, 3]), shape=(2, 3))
+        a2 = sp.csc_matrix(([1, 5, 4], [0, 1, 0], [0, 1, 1, 3]), shape=(2, 3))
+        self.assertTrue(array_equal(a1, a2))