I'm Brett Slatkin and this is where I write about programming and related topics. You can contact me here or view my projects.

30 August 2016

How you learn to become a better programmer

I received a nice email last week from a professor at a major university, who asked:

I'm wondering if you can share with me some advice on how to train inexperienced graduate students to be productive in writing quality code in a short period of time.

First: I'm flattered that this person thought I have any advice worth giving! My second thought was: It's impossible! But after thinking it over, I came up with one suggestion. In hopes that someone else finds my reply useful, here's what I wrote in return (with a bit more editing):

I think the best thing for improving code quality is to require your students to write a corresponding set of unittests along with any code they write. Python has a wonderful unittesting module that's built-in, as well as a good mocking library. I would encourage your students to also run a test coverage tool periodically to understand what parts of their code isn't tested.

Unittests are like an insurance policy for change. They're especially important for languages like Python that have no other way to verify correctness at even a superficial level. When your code has unittests, it becomes a lot easier to modify and expand a program over time while preserving confidence in the functionality you've already built.

Even when I write code in my free time for fun or personal projects — here's a recent example — I still write tests because they help me build functionality faster overall. The time it takes to write tests is far less than the time it takes to fix all of the bugs introduced unknowingly when you don't have tests. It's worth acknowledging that testing may feel like a waste of effort sometimes, but it's important for your students to understand that taking the time to write tests will be more efficient overall for any program that's non-trivial.

My last piece of advice is this: Sometimes people say that something was too hard to test, so they just skipped writing a test for it. This is exactly the wrong conclusion. If your code is too hard to test, that means your code is bad. The solution to every problem — no matter how hard the problem is — can be easy to test. If your program is not easy to test, then your code needs to be refactored or rewritten to make it easy. Doing that is how you learn to become a better programmer.
© 2009-2024 Brett Slatkin