Make Subtensor.make_node() detect more broadcastable dimensions.

theano/tensor/subtensor.py

@@ -391,19 +391,6 @@ class Subtensor(Op):
             exception.subtensor_invalid = True
             raise exception
 
-        # infer the broadcasting pattern
-        padded = (idx_list
-                + [slice(None, None, None)] * (x.type.ndim - len(idx_list)))
-        broadcastable = [bc and p.start is None and p.stop is None
-                         # No need to check step when there is only
-                         # one element and start and stop are None,
-                         # the output will always have 1 element.
-                         # We could call get_canonical_form_slice() to
-                         # catch more broadcast case. I let this to
-                         # later.
-                         for p, bc in izip(padded, x.type.broadcastable)
-                         if isinstance(p, slice)]
-
         input_types = Subtensor.collapse(idx_list,
                 lambda entry: isinstance(entry, gof.Type))
         if len(inputs) != len(input_types):
@@ -416,6 +403,31 @@ class Subtensor(Op):
                     "Wrong type for Subtensor template. Expected %s, got %s."
                     % (input.type, expected_type))
 
+        # infer the broadcasting pattern
+        padded = (idx_list
+                + [slice(None, None, None)] * (x.type.ndim - len(idx_list)))
+        broadcastable = []
+        for i, (p, bc) in enumerate(izip(padded, x.type.broadcastable)):
+            if isinstance(p, slice):
+                if bc and p.start in [None, 0]:
+                    # No need to check step when there is only
+                    # one element.
+                    # We could call get_canonical_form_slice() to
+                    # catch more broadcast case. I let this to
+                    # later.
+                    if p.stop is None:
+                        broadcastable.append(bc)
+                        continue
+                    try:
+                        start = theano.tensor.get_scalar_constant_value(p.start)
+                        stop = theano.tensor.get_scalar_constant_value(p.stop)
+                        if stop > start:
+                            broadcastable.append(True)
+                            continue
+                    except theano.tensor.NotScalarConstantError:
+                        pass
+                broadcastable.append(False)
+
         return gof.Apply(self,
                          (x, ) + inputs,
                          [theano.tensor.tensor(dtype=x.type.dtype,