Topological sort of a graph, allowing better console (and other) outputs.

Uses algorithm borrowed from networkx graph library to sort a graph in
topological order. The method is only used by output plugins, as
internal plumbery does not really care about the node order.

Also includes a bonobo.util.python.require function that helps importing
thing in a package-less context, or when there are conflict with site
package names.

bonobo/util/objects.py

@@ -200,6 +200,9 @@ class ValueHolder:
     def __invert__(self):
         return ~self.value
 
+    def __len__(self):
+        return len(self.value)
+
 
 def get_attribute_or_create(obj, attr, default):
     try:
@@ -207,4 +210,3 @@ def get_attribute_or_create(obj, attr, default):
     except AttributeError:
         setattr(obj, attr, default)
         return getattr(obj, attr)
-

bonobo/util/python.py

@@ -0,0 +1,22 @@
+import inspect
+import os
+import runpy
+
+
+class _RequiredModule:
+    def __init__(self, dct):
+        self.__dict__ = dct
+
+
+class _RequiredModulesRegistry(dict):
+    def require(self, name):
+        if name not in self:
+            bits = name.split('.')
+            pathname = os.path.join(os.getcwd(), os.path.dirname(inspect.getfile(inspect.stack()[1][0])))
+            filename = os.path.join(pathname, *bits[:-1], bits[-1] + '.py')
+            self[name] = _RequiredModule(runpy.run_path(filename, run_name=name))
+        return self[name]
+
+
+registry = _RequiredModulesRegistry()
+require = registry.require