added tests for shared variable contract for tensor and sparse

theano/sparse/tests/test_basic.py

@@ -400,5 +400,99 @@ class test_structureddot(unittest.TestCase):
             self.failUnless(numpy.allclose(theano_result, scipy_result))
             self.failIf(theano_time > overhead_rtol*scipy_time + overhead_tol)
 
+
+def test_shared_dont_alias():
+    rng = numpy.random.RandomState([3,5,17])
+    x_lil = random_lil((2,4), 'float64', 5)
+    x = sp.csr_matrix(x_lil) 
+    densifier = rng.randn(*x.shape)
+
+    x_shared = theano.shared(x, borrow = False)
+    prod = theano.sparse.mul_s_d(x_shared,densifier)
+
+    prod_func = theano.function([],prod)
+
+    prod_val = prod_func().todense()
+
+    x_dense_fake = x.todense()
+    x_dense = numpy.asarray(x_dense_fake)
+    scipy_prod_val = x_dense * densifier
+    assert numpy.allclose(scipy_prod_val, prod_val)
+
+    for i in xrange(x.shape[0]):
+        for j in xrange(x.shape[1]):
+            if x[i,j] != 0:
+                x[i,j] += 1
+
+    prod_val_2 = prod_func().todense()
+
+    #value used to construct should not alias with internal
+    assert numpy.allclose(prod_val_2,prod_val)
+
+    x = x_shared.get_value(borrow = False)
+
+    for i in xrange(x.shape[0]):
+        for j in xrange(x.shape[1]):
+            if x[i,j] != 0:
+                x[i,j] += 1
+
+    prod_val_3 = prod_func().todense()
+
+    #value returned by access should not alias with internal
+    assert numpy.allclose(prod_val, prod_val_3)
+
+    #in this case we can alias
+    x = x_shared.get_value(borrow = True)
+    
+    for i in xrange(x.shape[0]):
+        for j in xrange(x.shape[1]):
+            if x[i,j] != 0:
+                x[i,j] += 1
+
+    
+    x_dense_fake = x.todense()
+    x_dense = numpy.asarray(x_dense_fake)
+    scipy_prod_val = x_dense * densifier
+
+
+    #this is not required by the contract but it is a feature we've implemented
+    assert numpy.allclose(scipy_prod_val, prod_func().todense())
+
+def test_shared_do_alias():
+    rng = numpy.random.RandomState([2,4,16])
+    x_lil = random_lil((2,4), 'float64', 5)
+    x = sp.csr_matrix(x_lil)
+
+    x_shared = theano.shared(x, borrow = True)
+
+    densifier = rng.randn(*x.shape)
+
+    prod = theano.sparse.mul_s_d(x_shared,densifier)
+
+    prod_func = theano.function([],prod)
+
+    prod_val = prod_func().todense()
+
+    x_dense_fake = x.todense()
+    x_dense = numpy.asarray(x_dense_fake)
+    scipy_prod_val = x_dense * densifier
+
+    assert numpy.allclose(scipy_prod_val, prod_val)
+
+    for i in xrange(x.shape[0]):
+        for j in xrange(x.shape[1]):
+            if x[i,j] != 0:
+                x[i,j] += 1    
+
+
+    x_dense_fake = x.todense()
+    x_dense = numpy.asarray(x_dense_fake)
+    scipy_prod_val = x_dense * densifier
+
+
+    #not required by the contract but it is a feature we've implemented
+    assert numpy.allclose(scipy_prod_val, prod_func().todense())
+
+
 if __name__ == '__main__':
     unittest.main()

theano/tensor/tests/test_basic.py

@@ -3366,6 +3366,62 @@ def test_dimshuffle_duplicate():
 
     assert success
 
+def test_shared_dont_alias():
+    rng = numpy.random.RandomState([3,5,17])
+    x = rng.uniform(0,1,[2,4])
+    
+    x_shared = theano.shared(x, borrow = False)
+
+    total = theano.tensor.sum(x_shared)
+    
+    total_func = theano.function([],total)
+
+    total_val = total_func()
+
+    assert numpy.allclose(x.sum(), total_val)
+
+    x += 1
+
+    total_val_2 = total_func()
+
+    #value used to construct should not alias with internal
+    assert total_val == total_val_2
+
+    x = x_shared.get_value(borrow = False)
+
+    x += 1
+
+    total_val_3 = total_func()
+
+    #value returned by access should not alias with internal
+    assert total_val == total_val_3
+
+    #in this case we can alias
+    x = x_shared.get_value(borrow = True)
+    x += 1
+
+    #this is not required by the contract but it is a feature we've implemented
+    assert numpy.allclose(x.sum(), total_func())
+
+def test_shared_do_alias():
+    rng = numpy.random.RandomState([2,4,16])
+    x = rng.uniform(1,2,[4,2])
+
+    x_shared = theano.shared(x, borrow = True)
+
+    total = theano.tensor.sum(x_shared)
+
+    total_func = theano.function([],total)
+
+    total_val = total_func()
+
+    assert numpy.allclose(x.sum(), total_val)
+
+    x += 1
+
+    #not required by the contract but it is a feature we've implemented
+    assert numpy.allclose(x.sum(), total_func())
+
 
 if __name__ == '__main__':
     if 1: