Merge pull request #564 from nouiz/alloc

NEWS.txt

@@ -23,6 +23,10 @@ Interface changes
    instance, function([x, y], [y]). You can use the kwarg
    ``on_unused_input={'raise', 'warn', 'ignore'}`` to control this.
    (Pascal L.)
+ * tensor.alloc() now raise an error during graph build time
+   when we try to create less dimensions then the number of dimensions
+   the provieded value have. In the past, the error was at run time.
+   (Frederic B.)
 
 New Features
  * debugprint new param ids=["CHAR", "id", "int", ""]

theano/scalar/basic.py

@@ -743,9 +743,14 @@ class ScalarOp(Op):
         if hasattr(self, 'name') and self.name:
             return self.name
         else:
+            param = [(k, v) for k, v in self.__dict__.items()
+                     if k not in ["name", "_op_use_c_code"]]
+            if param:
                 return "%s{%s}" % (self.__class__.__name__,
-                               ", ".join("%s=%s" % (k, v) for k, v in
-                                         self.__dict__.items() if k != "name"))
+                                   ", ".join("%s=%s" % (k, v)
+                                             for k, v in param))
+            else:
+                return self.__class__.__name__
 
     def c_code_cache_version(self):
         return (3,)

theano/tensor/basic.py

@@ -2801,6 +2801,10 @@ class Alloc(gof.Op):
         v = as_tensor_variable(value)
         sh = [as_tensor_variable(s) for s in shape]
         bcast = []
+        if v.ndim > len(sh):
+            raise TypeError("Alloc value to use have more dimensions"
+                            " then the specified dimensions",
+                            v.ndim, len(sh))
         for i, s in enumerate(sh):
             if s.type.dtype[:3] not in ('int', 'uin'):
                 if config.exception_verbosity == 'high':

theano/tensor/opt.py

@@ -1304,7 +1304,7 @@ def local_alloc_elemwise(node):
       -> elemwise(x, y.TensorType(no broadcast flag))
 
     BROADCAST CONDITION: the condition is that the one input that are
-    not to be optimized to have the same braodcast pattern as the
+    not to be optimized to have the same broadcast pattern as the
     output
 
          We can change the alloc by a dimshuffle as the elemwise

theano/tensor/tests/test_basic.py

@@ -1240,10 +1240,16 @@ AllocTester = makeBroadcastTester(
             ),
         bad_runtime = dict(
                     bad_shape12 = (rand(7), numpy.int32(7), numpy.int32(5)),
+                    ),
+        bad_build = dict(
                     too_big32 = (rand(6,2,4), numpy.int32(6), numpy.int32(2)),
-            too_big32b = (rand(6,2,4), numpy.int32(2), numpy.int32(4)),
+                    too_big32b = (rand(6,2,4), numpy.int32(6), numpy.int32(4)),
+                    too_big32c = (rand(6,2,4), numpy.int32(2), numpy.int32(4)),
+                    too_big32d = (rand(6,2,4), numpy.int32(2), numpy.int32(6)),
+                    too_big32e = (rand(6,2,4), numpy.int32(4), numpy.int32(6)),
+                    too_big32f = (rand(6,2,4), numpy.int32(4), numpy.int32(2)),
         ),
-        )
+)
 
 # Since not all inputs of Alloc are differentiable, we need different testers
 s1, s2, s3 = randint_ranged(1, 13, (3,))

theano/tensor/tests/test_opt.py

@@ -3146,23 +3146,27 @@ class T_local_sum(unittest.TestCase):
                 f = theano.function([a],t_like(a).sum(d),mode=mode)
                 assert numpy.allclose(f(input),n_like(input).sum(d))
                 assert len(f.maker.env.nodes)==nb_nodes[1]
-                assert f.maker.env.toposort()[-1].op==T.alloc
-
+                topo = f.maker.env.toposort()
+                assert topo[-1].op == T.alloc
+                assert not any([isinstance(node.op, T.Sum) for node in topo])
             for i in range(3):
                 f = theano.function([a],t_like(a).sum(i),mode=mode)
                 assert numpy.allclose(f(input),n_like(input).sum(i))
                 assert len(f.maker.env.nodes)==nb_nodes[2]
-                assert f.maker.env.toposort()[-1].op==T.alloc
+                topo = f.maker.env.toposort()
+                assert topo[-1].op == T.alloc
+                assert not any([isinstance(node.op, T.Sum) for node in topo])
 
             backup = config.warn.sum_sum_bug
             config.warn.sum_sum_bug = False
             try:
                 for d, dd in [(0,0),(1,0),(2,0),(0,1),(1,1),(2,1)]:
                     f = theano.function([a],t_like(a).sum(d).sum(dd),mode=mode)
-                    print f.maker.env.toposort()
                     assert numpy.allclose(f(input),n_like(input).sum(d).sum(dd))
                     assert len(f.maker.env.nodes)==nb_nodes[3]
-                    assert f.maker.env.toposort()[-1].op==T.alloc
+                    topo = f.maker.env.toposort()
+                    assert topo[-1].op == T.alloc
+                    assert not any([isinstance(node.op, T.Sum) for node in topo])
             finally:
                 config.warn.sum_sum_bug = backup