Dave Challis wrote:
I'll have a look into metaclasses too, haven't stumbled upon those yet
at all.
It's a potentially brain-exploding topic, though, so if the above
solution works for you, you might want to leave it at that ;-)
But very briefly, a metaclass is a something that's responsible for
creating a class, much like an ordinary class is responsible for
creating an object. When Python executes the following statement,
class Spam:
attrib = 1
def func(self):
pass
# <-- end of class statement
it will create a new scope for the class content, execute the class
body, and then, when it reaches the end, call the "metaclass" to create
the actual class object. The metaclass is given the requested name
("Spam" in this case), any base classes, and a dictionary containing
everything from the class scope ("attrib" and "func", in this case).
The thing that's returned is assigned to the "Spam" variable.
The default metaclass ("type") just creates an ordinary class object,
but if you replace that with your own metaclas, you can completely
override that behaviour, or just extend it (e.g. by registering the
subclasses in a common registry). Like, say, this:
registry = [] # list of subclasses
class Plugin(object):
class __metaclass__(type):
def __init__(cls, name, bases, dict):
type.__init__(name, bases, dict)
registry.append((name, cls))
class SpamPlugin(Plugin):
pass
class BaconPlugin(Plugin):
pass
for name, cls in registry:
if cls is not Plugin:
print name, cls
Here, the presence of an inner __metaclass__ class (which is a subclass
of "type") causes Python's class machinery to use that class instead of
"type" when creating class objects for Plugin or any subclass thereof.
The extra code in the __init__ method just all plugins to a list.
For more on this, see e.g.
http://www.python.org/download/releases/2.2/descrintro/#metaclasses
</F>
--
http://mail.python.org/mailman/listinfo/python-list
