Skip to content


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Python monkey-patching for Humans
branch: master

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.


Python monkey-patching for Humans

Altered States tries to take the concept of Python for Humans to the world of monkey patching. A search On PyPI shows that there are already many often technically sophisticated packages that does this. The thing with them is that they are all making a procedure that should be relatively simple complicated.

With Altered States you get to choose between two functions, one that can either manipulate your world via a with statement:

>>> from altered import state
>>> class Anon(object): pass
>>> o = Anon()
>>> = 'foo'
>>> with state(o, foo='bar'):
...     print

or using a decorator:

>>> from altered import state
>>> struct = {'a': 1}
>>> @state(struct, a=3)
... def fn():
...     return struct['a']
>>> fn()

This example also shows how .state() can be applied to dict's as well as objects.

From version 0.8.5, you can also make changes in two steps (e.g. if you need setup/teardown steps in a test) using the .alter() function:

>>> from altered import alter, E
>>> o = E(foo='foo')
>>> restore = alter(o, foo='bar')
>>> print(
>>> restore()
>>> print(

Just like .state(), .alter() works on dict's too:

>>> from altered import alter
>>> struct = {'a': 1}
>>> restore = alter(struct, a=3)
>>> print(struct)
{'a': 3}
>>> restore()
>>> print(struct)
{'a': 1}

.alter() can of course not keep track of downstream Exceptions like .state() can, so if you need to guarantee state restoration, it's up to you to ensure that the returned restoration function is actually called.

Altered States has been verified to run on Python 2.5, 2.6 and 27.


I implemented the code that later became Altered States beacuse I was unhappy about how fixture setup was obscuring actual test code in test suites, and that is also the recommended primary use of Altered States. But there are other usecases where a reversible change of the environment can be useful (switching between authenticated users, temporary I/O redirection, probably more).

Expando objects

Altered States also contains an optional feature called Expando objects. It's a simple object that can be used to create replacement structures easily. It's basically an empty object that you can add any extra attributes to, with a conceptual implementation along the lines of:

class Expando(object):
    def __init__(self, *args, **kw):

Full source is marginally more complex, see here. So if you need an object with another object embedded that has a method you can create that with:

>>> from altered import Expando
>>> faked_ctx = Expando(user=Expando(get_name=lambda: 'Foo Bar'))
>>> faked_ctx.user.get_name()
'Foo Bar'

If we revisit the second example, it can be expressed with an Expando object like this:

>>> from altered import Expando, state
>>> obj = Expando(a=1)
>>> @state(obj, a=3)
... def fn():
...     return obj.a
>>> fn()

Expando is also aliased to the shorter name E for terser nesting:

>>> from altered import E
>>> fakey = E(substruct=E(membr='foo', method=lambda: 'x'))

Reference documentation

More detailed (but incomplete as of yet) documentation can be found at Read The Docs here.

What was the name again?

You mean you haven't seen the movie? Go see it, it's a trip! And when you see it, take note of the implicit warning in the film of what happens if you take your usage too far:

The whole idea of Altered States is to create side-effects. Please use Altered States responsibly.

Something went wrong with that request. Please try again.