[FIX] Merge: work with sparse

Orange/data/table.py

@@ -57,15 +57,15 @@ def __init__(self, table, row_index):
         self.id = table.ids[row_index]
         self._x = table.X[row_index]
         if sp.issparse(self._x):
-            self.sparse_x = self._x
+            self.sparse_x = sp.csr_matrix(self._x)
             self._x = np.asarray(self._x.todense())[0]
         self._y = table._Y[row_index]
         if sp.issparse(self._y):
-            self.sparse_y = self._y
+            self.sparse_y = sp.csr_matrix(self._y)
             self._y = np.asarray(self._y.todense())[0]
         self._metas = table.metas[row_index]
         if sp.issparse(self._metas):
-            self.sparse_metas = self._metas
+            self.sparse_metas = sp.csr_matrix(self._metas)
             self._metas = np.asarray(self._metas.todense())[0]
 
     @property

Orange/widgets/data/owmergedata.py

@@ -4,9 +4,11 @@
 from AnyQt.QtWidgets import QApplication, QStyle, QSizePolicy
 
 import numpy as np
+import scipy.sparse as sp
 
 import Orange
 from Orange.data import StringVariable, ContinuousVariable
+from Orange.data.util import hstack
 from Orange.widgets import widget, gui, settings
 from Orange.widgets.utils import itemmodels
 from Orange.widgets.utils.sql import check_sql_input
@@ -362,20 +364,29 @@ def _join_table_by_indices(self, reduced_extra, indices):
     def _join_array_by_indices(left, right, indices, string_cols=None):
         """Join (horizontally) two arrays, taking pairs of rows given in indices
         """
-        tpe = object if object in (left.dtype, right.dtype) else left.dtype
-        left_width, right_width = left.shape[1], right.shape[1]
-        arr = np.full((indices.shape[1], left_width + right_width), np.nan, tpe)
-        if string_cols:
-            arr[:, string_cols] = ""
-        for indices, to_change, lookup in (
-                (indices[0], arr[:, :left_width], left),
-                (indices[1], arr[:, left_width:], right)):
-            known = indices != -1
-            to_change[known] = lookup[indices[known]]
-        return arr
-
-
-def test():
+        def prepare(arr, inds, str_cols):
+            try:
+                newarr = arr[inds]
+            except IndexError:
+                newarr = np.full_like(arr, np.nan)
+            else:
+                empty = np.full(arr.shape[1], np.nan)
+                if str_cols:
+                    assert arr.dtype == object
+                    empty = empty.astype(object)
+                    empty[str_cols] = ''
+                newarr[inds == -1] = empty
+            return newarr
+
+        left_width = left.shape[1]
+        str_left = [i for i in string_cols or () if i < left_width]
+        str_right = [i - left_width for i in string_cols or () if i >= left_width]
+        res = hstack((prepare(left, indices[0], str_left),
+                      prepare(right, indices[1], str_right)))
+        return res
+
+
+def main():
     app = QApplication([])
     w = OWMergeData()
     data = Orange.data.Table("tests/data-gender-region")
@@ -388,4 +399,4 @@ def test():
 
 
 if __name__ == "__main__":
-    test()
+    main()

Orange/widgets/data/tests/test_owmergedata.py

@@ -3,6 +3,7 @@
 from itertools import chain
 
 import numpy as np
+import scipy.sparse as sp
 
 from Orange.data import Table, Domain, DiscreteVariable, StringVariable
 from Orange.widgets.data.owmergedata import OWMergeData, INSTANCEID, INDEX
@@ -425,3 +426,28 @@ def test_best_match(self):
         self.assertEqual(self.widget.attr_merge_extra, zoo_images.domain[-1])
         self.assertEqual(self.widget.attr_combine_data, zoo.domain[-1])
         self.assertEqual(self.widget.attr_combine_extra, zoo_images.domain[-1])
+
+    def test_sparse(self):
+        """
+        Merge should work with sparse.
+        GH-2295
+        GH-2155
+        """
+        data = Table("iris")[::25]
+        data_ed_dense = Table("titanic")[::300]
+        data_ed_sparse = Table("titanic")[::300]
+        data_ed_sparse.X = sp.csr_matrix(data_ed_sparse.X)
+        self.send_signal("Data", data)
+
+        self.send_signal("Extra Data", data_ed_dense)
+        output_dense = self.get_output("Data")
+        self.assertFalse(sp.issparse(output_dense.X))
+        self.assertFalse(output_dense.is_sparse())
+
+        self.send_signal("Extra Data", data_ed_sparse)
+        output_sparse = self.get_output("Data")
+        self.assertTrue(sp.issparse(output_sparse.X))
+        self.assertTrue(output_sparse.is_sparse())
+
+        output_sparse.X = output_sparse.X.toarray()
+        self.assertTablesEqual(output_dense, output_sparse)