Added Fancy Indexing for sparse matrix.

theano/sparse/basic.py

@@ -347,7 +347,10 @@ class _sparse_py_operators:
             else:
                 ret = get_item_2d(self, args)
         else:
-            ret = get_item_2d(self, args)
+            if isinstance(args[0], list):
+                ret = get_item_list(self, args)
+            else:
+                ret = get_item_2d(self, args)
         return ret
 
 
@@ -992,7 +995,47 @@ class SparseFromDense(gof.op.Op):
 csr_from_dense = SparseFromDense('csr')
 """Convert a dense matrix to a sparse csr matrix.
 
+<<<<<<< HEAD
 :param x: A dense matrix.
+=======
+# Indexing
+class GetItemList(gof.op.Op):
+
+    def __eq__(self, other):
+        return (type(self) == type(other))
+
+    def __hash__(self):
+        return hash(type(self))
+
+    def make_node(self, x, index):
+        x = as_sparse_variable(x)
+        assert x.format in ["csr", "csc"]
+
+        ind = tensor.as_tensor_variable(index)
+        assert ind.ndim == 2
+        assert 'int' in ind.dtype
+
+        return gof.Apply(self, [x, ind], [x.type()])
+
+    def perform(self, node, inp, (out, )):
+        x = inp[0]
+        y = []
+        indices = inp[1]
+        assert _is_sparse(x)
+
+        for ind in indices:
+            out[0] = x[indices[0][ind]]
+
+    def __str__(self):
+        return self.__class__.__name__
+
+get_item_list = GetItemList()
+
+
+class GetItem2d(gof.op.Op):
+    """Implement a subtensor of sparse variable and that return a
+    sparse matrix.
+>>>>>>> Added Fancy Indexing for sparse matrix.
 
 :return: The same as `x` in a sparse matrix format.
 

theano/sparse/tests/test_basic.py

@@ -15,7 +15,7 @@ from theano import sparse
 from theano import compile, config, gof
 from theano.sparse import enable_sparse
 from theano.gof.python25 import all, any, product
-
+from theano.tensor.basic import _allclose
 
 if not enable_sparse:
     raise SkipTest('Optional package sparse disabled')
@@ -2020,6 +2020,36 @@ class Test_getitem(unittest.TestCase):
     def setUp(self):
         self.rng = numpy.random.RandomState(utt.fetch_seed())
 
+    def test_GetItemList(self):
+        rng = numpy.random.RandomState(utt.fetch_seed())
+        a = sparse.csr_matrix()
+        b = sparse.csc_matrix()
+        y = a[[0, 1, 2, 3, 1]]
+        z = b[[0, 1, 2, 3, 1]]
+
+        fa = theano.function([a], y)
+        fb = theano.function([b], z)
+
+        A = rng.rand(4, 4).astype(theano.config.floatX)
+
+        t_geta = fa(numpy.asarray(A, dtype="float64")).todense()
+        t_getb = fb(numpy.asarray(A, dtype="float64")).todense()
+
+        s_geta = numpy.asarray(scipy.sparse.csr_matrix(A)[[0, 1, 2, 3, 1]].todense(), dtype="float64")
+        s_getb = numpy.asarray(scipy.sparse.csc_matrix(A)[[0, 1, 2, 3, 1]].todense(), dtype="float64")
+
+        utt.assert_allclose(t_geta, s_geta)
+        utt.assert_allclose(t_getb, s_getb)
+
+    def test_GetItemList_wrong_index(self):
+        rng = numpy.random.RandomState(utt.fetch_seed())
+        x = sparse.csr_matrix()
+        y = x[[0, 4]]
+        f = theano.function([x], y)
+
+        A = rng.rand(2, 2).astype(theano.config.floatX)
+        self.assertRaises(IndexError, f, A)
+
     def test_GetItem2D(self):
         sparse_formats = ('csc', 'csr')
         for format in sparse_formats: