Python Annotations
Mar 13, 2013 · 2 minute readOne of the most annoying (lack-of, in my opinion) features of Python is it’s dynamic typing. While this is wonderful for quick and dirty things that you don’t want to write lots of code for, it fails pretty much horrifically for large, production-scale code. As I had the opportunity to experience during the summer, when dealing with a project of several thousand interleaving lines of Python, no one can help you figure out what the arguments to a function should actually be (except for docstrings and humans, mostly yourself).
That’s a lot of thinking and storage of data in my head which I think is unnecessary, time-wasting, and potentially dangerous (even good code coverage with tests can still allow a slim corner case to occur and the lack of strong compile-time checks for silly things like type checking keeps me up at night). So when I came across a blog post today on Python type annotations, that sounded just wonderful (how did I miss this before?!). However, after reading more about it, it seems like Python doesn’t go very far to actually do anything with these annotations, especially at compile-time, so you’re left to do the grunt work yourself. But that’s fine! It’s better than nothing.
The link to the blog post and the author’s preliminary implementation of the actual checking (done reasonably elegantly during decorators, which is as good as I guess we can hope for in Python) are below: