Saturday, October 14, 2006

Identity Comparison vs Comparing Identities

An explaination of the virtues of foo is not None over foo != None lead to an explaination of identity comparison with the is operator. A fellow equated this to, roughly, id(a.x) == id(b.x), which I told him was roughly correct but probably not actually correct. It only took a little bit of through to see how uncorrect it was.

The following code creates a simple class with one property descriptor (read-only). It solves the requirements that with to instances of this class, a and b, id(a.x)==id(b.x) can be true while a.x is not b.x! How does this happen?

class foo(object):
x = property(lambda s: id(s))
a = foo()
b = foo()
assert id(a.x) == id(b.x)
assert a.x is not b.x

How on earth does this code prove what it does? a.x and b.x are created on the fly, passed to the id function, and then destroyed with no references left. Because they are both created in the right order with the id(a.x)==id(b.x) expression, they just happen to get the same memory addresses, which in CPython is used for the id function's result. This leads to misleading results, so don't rely on them in such ways. Identification is what it is, and you shouldn't try to break it down.


Anonymous said...

Pretty neat experiment there.

Anonymous said...

wow, ironfroggy. for a complex problem, that's a pretty clean and elegant example. nice.

I write here about programming, how to program better, things I think are neat and are related to programming. I might write other things at my personal website.

I am happily employed by the excellent Caktus Group, located in beautiful and friendly Carrboro, NC, where I work with Python, Django, and Javascript.

Blog Archive