PyTux

Trips of a curious penguin.

Hello, time traveler! You are reading an article that is almost ten years old. The world has changed, and so have I and my opinions. There is a good chance what's below is not current, correct, or secure anymore, and maybe it never was. This page is preserved because I am an archivist at heart, but you have been warned.

Instance monkey-patching in Python

Monkey-patching is the technique of swapping functions or methods with others in order to change a module, library or class behavior.

There are some people with strong opinions about it. I haven’t, but it comes really useful when testing, to simulate side-effecting functions or to silence expected errors and warnings.

Class methods monkey patching in Python is really easy, as you can freely assign function to class method names:

1
2
3
4
5
6
7
8
9
10
11
>>> class Class():
...    def add(self, x, y):
...       return x + y
...
>>> inst = Class()
>>> def not_exactly_add(self, x, y):
...    return x * y
...
>>> Class.add = not_exactly_add
>>> inst.add(3, 3)
9

This way all the instances of the target class will have the method monkey-patched and there is no problem with arguments, bindings… Everything really straight-forward.

We can also call the old existing method, to handle only some cases or to add some functionality while not repeating code (DRY):

1
2
3
4
5
6
7
8
9
10
11
12
13
14
>>> class Class():
...    def add(self, x, y):
...       return x + y
...
>>> old_boring_add = Class.add
>>> def add_is_not_enough(self, x, y):
...    return old_boring_add(self, x, y) + 1
...
>>> inst = Class()
>>> inst.add(3, 3)
6
>>> Class.add = add_is_not_enough
>>> inst.add(3, 3)
7

But what if we wanted to do the same, patching just a single instance?

To recap, the requirements are:

  • we want just the current instance to be patched;
  • we want to build something on top of the existing method, not to replace it entirely;
  • we want each monkey-patch not to rollback all the previous ones (so no super() or class method call);
  • we want to be able to do so also from inside a method.

The trick is to save and use the existing method as we did above, and then bind the new function to the instance with types.MethodType before assigning it to the method name.

The binding is the magic that causes the instance to be passed as first argument (self) each time the method is called. See these two StackOverflow questions to get an idea.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
>>> import types
>>> class Class():
...    def add(self, x, y):
...       return x + y
...    def become_more_powerful(self):
...       old_add = self.add
...       def more_powerful_add(self, x, y):
...          return old_add(x, y) + 1
...       self.add = types.MethodType(more_powerful_add, self)
...
>>> inst = Class()
>>> inst.add(3, 3)
6
>>> inst.become_more_powerful()
>>> inst.add(3, 3)
7
>>> inst.become_more_powerful()
>>> inst.become_more_powerful()
>>> inst.become_more_powerful()
>>> inst.add(3, 3)
10

And here we go!

A practical example

You can see this technique being used in youtube-dl to silence expected warnings in this commit.

The monkey-patching of the instance is done on itself by a method of a testing subclass of the downloader.