merged

__init__.py

@@ -4,10 +4,12 @@ import tensor
 import sparse
 import compile
 import gradient
-import opt
+import tensor_opt
+import scalar_opt
 
 from tensor import *
 from compile import *
-from opt import *
+from tensor_opt import *
+from scalar_opt import *
 from gradient import *
 

_test_sparse.py

@@ -5,11 +5,7 @@ import compile
 import gradient
 
 from sparse import _is_dense, _is_sparse, _is_dense_result, _is_sparse_result
-
-""" Types of sparse matrices to use for testing """
-_mtypes = [sparse.csc_matrix, sparse.csr_matrix]
-#_mtypes = [sparse.csc_matrix, sparse.csr_matrix, sparse.dok_matrix, sparse.lil_matrix, sparse.coo_matrix]
-_mtype_to_str = {sparse.csc_matrix: "csc", sparse.csr_matrix: "csr"}
+from sparse import _mtypes, _mtype_to_str
 
 class T_transpose(unittest.TestCase):
     def setUp(self):

elemwise.py

@@ -321,8 +321,15 @@ class Broadcast(Op, Destroyer):
         # the second calling form is used because in certain versions of numpy
         # the first (faster) version leads to segfaults
         ufunc_args = [input.data for input in self.inputs]# + output_storage
-        #self.ufunc(*(ufunc_args+output_storage))
-        output_storage[0][:] = self.ufunc(*ufunc_args)
+        results = self.ufunc(*ufunc_args)
+        if self.ufunc.nout == 1: results = [results]
+        for result, storage in zip(results, output_storage):
+            if storage.shape:
+                storage[:] = result
+            else:
+                storage.itemset(result)
+        # the following should be used instead of the previous loop, unfortunately it tends to segfault
+        # self.ufunc(*(ufunc_args+output_storage))
 
     def _c_all(self, inames, onames, sub):
         _inames = inames

gof/op.py

@@ -89,6 +89,13 @@ class Op(object):
         return self._hash_id
 
     def desc(self):
+        """
+        Description (signature) of this L{Op}. L{Op}s with the same
+        signature may be collapsed by the L{MergeOptimizer}.
+
+        @attention: If your L{Op} has additional options or a different
+        constructor you probably want to override this.
+        """
         return self.__class__
 
     def strdesc(self):
@@ -139,10 +146,10 @@ class Op(object):
         """
         Shallow copy of this L{Op}. The inputs are the exact same, but
         the outputs are recreated because of the one-owner-per-result
-        policy. The default behavior is to call the constructor on
-        this L{Op}'s inputs.
+        policy. The default behavior is to call the constructor on this
+        L{Op}'s inputs.
 
-        To do a bottom-up copy of a graph, use clone_with_new_inputs.
+        To do a bottom-up copy of a graph, use L{clone_with_new_inputs}.
 
         @attention: If your L{Op} has additional options or a different
         constructor you probably want to override this.

scalar.py

@@ -11,6 +11,11 @@ from gof import Result, GuardedOp, Env, utils
 
 
 def as_scalar(x, name = None):
+    if isinstance(x, gof.Op):
+        if len(x.outputs) != 1:
+            raise ValueError("It is ambiguous which output of a multi-output Op has to be fetched.", x)
+        else:
+            x = x.outputs[0]
     if isinstance(x, float):
         s = Scalar('float64', name = name)
         s.data = x
@@ -21,6 +26,7 @@ def as_scalar(x, name = None):
         return s
     if isinstance(x, Scalar):
         return x
+    raise TypeError("Cannot convert %s to Scalar" % x)
 
 def constant(x):
     res = as_scalar(x)
@@ -194,7 +200,7 @@ def _multi(*fns):
         else:
             return [f(name) for name in names]
     if len(fns) == 1:
-        return partial(f2, fns)
+        return partial(f2, fns[0])
     else:
         return [partial(f2, f) for f in fns]
 

sparse.py

@@ -14,6 +14,11 @@ import gof.op, gof.result
 import tensor
 
 
+""" Types of sparse matrices to use for testing """
+_mtypes = [sparse.csc_matrix, sparse.csr_matrix]
+#_mtypes = [sparse.csc_matrix, sparse.csr_matrix, sparse.dok_matrix, sparse.lil_matrix, sparse.coo_matrix]
+_mtype_to_str = {sparse.csc_matrix: "csc", sparse.csr_matrix: "csr"}
+
 
 ## Type checking
 

tensor.py

@@ -304,6 +304,12 @@ s2t.Tensor = Tensor
 # alternate Tensor constructor
 def astensor(data, broadcastable=None, name=None):
     """Return a L{Tensor} containing given data"""
+    if isinstance(data, Op):
+        if len(data.outputs) != 1:
+            raise ValueError("It is ambiguous which output of a multi-output Op has to be fetched.", data)
+        else:
+            data = data.outputs[0]
+    
     if isinstance(data, Tensor):
         if broadcastable is not None and list(data.broadcastable) != list(broadcastable):
             raise TypeError("The data to wrap as a Tensor has the wrong broadcastable pattern. Expected %s, got %s." % (broadcastable, data.broadcastable))
@@ -315,13 +321,17 @@ def astensor(data, broadcastable=None, name=None):
         
     if data is None and broadcastable is None:
         raise TypeError("Cannot make a Tensor out of None.")
-    
+
+    _data = data
     data = numpy.asarray(data)
     if broadcastable is None:
         broadcastable = [s==1 for s in data.shape]
     elif broadcastable in [0, 1]:
         broadcastable = [broadcastable] *  len(data.shape)
-    rval = Tensor(data.dtype, broadcastable, name = name)
+    try:
+        rval = Tensor(data.dtype, broadcastable, name = name)
+    except TypeError:
+        raise TypeError("Cannot convert %s to Tensor." % repr(_data))
     rval.data = data # will raise if broadcastable was mis-specified
     return rval
 s2t.astensor = astensor