All Blogs

Check this out: http://ubuntulinuxhelp.com/can-we-talk-here/ It's very heartfelt and compassionate; and really does say what Ubuntu really is about.

In Python, closure capture the current frame (a reference to it in fact), not the current binding. So in when you define the local _f function, the current frame is captured, and when it is called, the local variable l is searched in this frame, and since the frame is not copied, but merely referenced, the value is the last value bound to the name l in the frame.

A simple example :


>>> def test():
... def f():
... print i
... i = 4
... f()
... i = 5
... f()
...
>>> test()
4
5


If you want to capture the current value, you need to change your function to take an additional parameter with a default value initialized with l[0], as default value of parameter are evaluated when compiling the function.

Your code should be changed to (only including the function declaration) :


>>> for l in list:
... if not hasattr(Klass, l[1]):
... print '%s should print %d' % (l[1], l[0])
... def _f(self, v = l[0]):
... print v
... _f.__doc__ = 'Get %s' % l[1]
... setattr(Klass, l[1], _f)


One thing you need to remember is that the nested function _f is not storing a reference to "l" from the parent scope: it has a reference to the parent scope itself. So when you rebind "l" in the parent scope, it affects what _f() sees.

There are a few solutions to the problem:

1. bind "l" as a default value to an argument of _f().

2. define _f() in a scope where "l" does not change (i.e. call another function that creates _f() inside your loop).

3. Go the object oriented route, and use an instance rather than a function. By making use of the descriptor __get__() call, you can do method-style binding to instances.

http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/502271

Here's how I was thinking: "list" is an array of pointers to some "tupple" structure. I iterate through all these pointers, and assign the location of the current tupple to "l". Then I generate some function, _f, which uses the object at location given by l. If later on I change l (so it points to something else), the memory location used into the previously defined function should not change *with* l... It's like

char *foo = "bar";
char *baz = foo;
foo = "bat";
printf("%s %s\n", foo, baz);

which prints "bat bar", not "bat bat".

I know about the "list" name, stupid, wouldn't do that in production code :-)

(Note: the code in this comment uses > for indentation, because you don't allow the pre tag.)

The problem comes when you define the function _f. Your definition references l by reference, not by value. Your loop changes l.

This way of binding variables from outer scopes provides a lot of power; for example, you can define a callback function that references variables from your outer scope directly, rather than needing a magic "caller context" argument as you might find in C.

To do what you want, try this:

def make_print(v, doc):
>def _f():
>>print v
>_f.__doc__ = doc
>return f

for l in list:
>if not hasattr(Klass, l[1]):
>>print '%s should print %d' % (l[1], l[0])
>>setattr(Klass, l[1], make_print(l[0], 'Get %s' % l[1]))


Alternatively, reference the original tuple item, which won't change:

for i in enumerate(list):
>if not hasattr(Klass, list[i][1]):
>>print '%s should print %d' % (list[i][1], list[i][0])
>>def _f(self):
>>>print list[i][0]
>>_f.__doc__ = 'Get %s' % list[i][1]
>>setattr(Klass, list[i][1], _f)


As an aside, don't call your lists list, because Python
has a built-in function named list. Worse yet, don't call your tuples list. :)

I am very interested in this algorithm. Could you send me a copy of the source code? Thanks a lot.

I was gonna ask you about source code, but then I read comments...

Cool post.
I also have a python implementation (many people posted interesting ideas on my post).

Is there a way to allow the softphone to pop when the screen saver has kicked in?
without having to enter the unlock credentials?

Yeah, it's cool. I've just updated my clutter build and will try to do some examples when I get home. I can't run glx stuff over vnc :)

I repeat: I know this is nothing really "new", it's just a playground.

Oh and, my antispamsystem dislikes blogspot.

Certainly not to rain on anything here but someone in the gnome world has done something similar:
njpatel dot blogspot dot com/2007/06/flickr-clutter-fluttr.html

(it won't let me submit that in link form!)

Clutter looks increasingly cool, and I plan on looking at your implementation here to find out how to use it a bit better. It seems to be in a constant flux though, so I'm not sure learning it now would really be useful.

Matt: glad it's useful to at least someone :-)
Corey: IIRC lowfat is written using more low-level OpenGL, which I don't know (yet). I did not intend to write a full-fledged slideshow app whatsoever, actually the main reason I started to look at Clutter was to write some (non-interactive) "public information display", easy to manage for non-geeks, software.

Err, not really... Doing the "wall" thing didn't work out, so I gave up.

You should take a look at Macslow's lowfat. It needs work and having combined effort would actually produce something faster!

Thanks man. I am sort of dabbling in learning C/C++ (mainly out of curiosity), this has been a big help with clutter. Cheers.

Any progress on this? I've been trying to do this myself, but I can't figure it out.

I know the TinyCA package, but it's not suited for this scenario as we work using an console-only server, and don't want to have our CA key anywhere outside it.

Have you ever tried TinyCA for this ?, it is a Perl/GTK interface for openssl. I use it to manage my certificates.