Monday, February 19, 2007

Pure Functions in the Python Standard Library?

I am toying with a small package for working with and developing pure functions in Python. An important part of this is allowing the pure functions you write to use as much of the standard library as possible, while remaining pure. What functions in the standard library could you call pure? Builtin types, the entire math library, the itertools module, etc. are on my list, but I don't want to miss things. Any suggestions?

So far I have a decorator that ensures a function A) only takes immutables and known-pure functions as arguments, and B) ensures any global names used within the function are pure functions. It is primitive, but a start. I hope to take this further with some actual uses, such as create working processes locally and remotely and pushing pure functions to them. This also has interesting potential for optimizing large computations, like executing functions before they are called when the potential arguments are known.

Pure functions are interesting and might be a nice addition to our Python toolkits.

3 comments:

Chuck said...

I'm not sure how "pure" itertools is, since iterators themselves have side effects -- they rather depend on them.

The kind optimizations you get out of "really pure" FP -- memoization, automatic parallelization, etc -- comes with pretty inflexible demands on referential integrity. I suspect you'd have to have monads or uniqueness types to have functional purity out of anything having to do with iterators and generators.

jto said...

What an interesting question.

I cheated by looking here:

http://www.python.org/doc/current/lib/lib.html

...and it's amazing how little is pure.

I don't know if you can count on anything actually being pure, because almost nothing is immutable in Python.

I think struct qualifies.

Almost all of datetime passes. The only really obvious exception is datetime.now().

re almost qualifies, bit match objects are mutable.

There are other modules you could wrap to be pure, like difflib and hashlib. Also possibly pickle, although behind the scenes it is far from pure, what with __setstate__().

ProfGra said...

Hi,
I'm interested in your decorator. This kind of hack could be the solution to an old problem I have. I use Python to teach algorithmics, and the fact that Python allows reading variables defined outside the «def myfunc():» is annoying.
Thanks for the article!

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