Last year, I’d never have called myself a programmer. Sure, I’d dabbled a bit here and there, but I’d never written code daily and I’d never collaborated on a large software project. At best, I was a dilettante.
Nine months later, I’m on a different plane: I’m still no expert, but I’m definitely a programmer (or at least a code monkey). Writing code every day is an important part of my job, and I love it. If you’re interested in learning to code, especially in Python, I hope my story helps you get started faster.
Syntax Is Easy
Oddly enough, learning the syntax of a programming language—the rules for assembling words, curly braces, backslashes, and all the other typographical flotsam in the corners of your keyboard into working code—was relatively easy. Python is particularly straightforward, and Zed Shaw’s Learn Python the Hard Way taught me the basics in a single weekend. Experimenting in an interactive environment was very helpful at this stage, since the interpreter caught all my mistakes, printed a helpful error message, and let me try again right away. Whenever I didn’t understand where I’d gone wrong, pasting the error message verbatim into Google always turned up a relevant answer.
Only a few days in, I was able to exercise my new skills by writing a to-do manager like Steve Losh’s t—and it actually worked. Writing even a simple utility that functioned exactly as I wanted it to was hugely empowering, and using my little program every day kept me excited about learning. Emulating t had an unanticipated, but immensely helpful, side effect: it let me compare my efforts to the actual t source code, which is also written in Python. Taking the time to really understand how Steve’s code worked (and why it was much, much better than mine) accelerated my learning dramatically.
Computer Science Is Practical
Of course, what’s a to-do manager without projects, sub-projects, and maybe even sub-sub-projects? Naively, I assumed that adding this feature to my little utility would be a quick one-hour project. When I sat down to start writing, though, I realized that I had no clue how to handle potentially infinite project nesting. A little research (leavened with some hazy memories from school) revealed that managing data with this sort of parent-child relationship is, unsurprisingly, a well-studied problem in computer science, and after a few nights of experimenting and reading I managed to add projects to my to-do list manager.
Along the way, I realized that I’d always be a few steps behind until I learned some basic computer science: most common problems have established solutions, and my complete ignorance of those solutions was forcing me to reinvent the wheel (poorly) at every turn. It’s a little like trying to organize thousands of loose papers without using a file cabinet; I’d probably come up with some sort of workable system eventually, but why go through all that pain when there’s a good solution that’s widely available? To push the analogy further, file cabinets also have a large ecosystem of standard add-ons designed to solve particular problems—it’s easy to get cabinets with built-in indexing systems, horizontal drawers, or wheels, for example. Similarly, it’s easy to find variations of the basic computer science building blocks that are optimized for particular applications.
Nine months isn’t enough to make anyone an expert in computer science, but there are a lot of excellent courses available free online. I started with MIT’s Introduction to Computer Science and Programming because it’s taught in Python, doesn’t assume any prior knowledge, and moves slowly enough for me to watch while riding the train and keeping an eye out for my stop. I’m not sure what I’ll tackle next, but I’m leaning towards continuing the MIT undergraduate sequence.
Craftsmanship Is Crucial
A few days after I started watching the MIT lectures, I switched jobs and began writing code at work. This was, to say the least, a big transition: I went from being the sole author of a hundred-line program to working with a team of real engineers building a real web application. The standard for the correctness of my code went up, but more importantly, so did the standards for its readability and maintainability.
Even with reams of well-written code (and the inevitable bits of crufty legacy
code) to learn from, it took me months to really understand why anyone would
prioritize anything other than accurate output. “It works right now,” I’d think to
myself, “so why would anyone spend another week just making it look pretty?”
That attitude changed the first time I had to add some functionality to a small
piece of code I’d written a few months earlier. Even as the original author, I
was completely baffled—I’d named every variable x, sbgs,
data (yes, really), or something equally useless, and now I couldn’t make
heads or tails of any of it. I ended up scrapping the whole file, starting from
scratch, and feeling lucky that none of the many bugs in my old code had seen
the light of day.
Needless to say, I’m now a convert. I’m getting better at giving functions and variables meaningful names. I plan my code more carefully. I’m a little more comfortable with object-oriented programming, and I’m keeping global variables to a minimum. More abstractly, I’m slowly learning how to hide complexity so that other people can treat my code as a pluggable component. I want to take pride not just in my code’s functionality, but in its readability, its cleanliness, and its aesthetics—in short, in its craftsmanship.
Taking a few months to focus on something other than basic computer science is a luxury I can afford only because Python provides an off-the-shelf version of most common building blocks. Since I don’t absolutely need to know, for example, how each of the many common sorting algorithms works (though it would probably be helpful), I can afford to put reading on hold for a little while to become a better craftsman. I’ve spent most of this time reading Code Complete, mostly because Jeff Atwood describes it as “the Joy of Cooking for software developers.” I’ll probably still make the same rookie mistakes, but I hope that this early investment will help me learn from my mistakes faster.
What Next?
I’ll finish Code Complete this week, but I’m not sure where I should invest my time next. I could:
- Review linear algebra. I’m hoping to pick up a few machine learning projects soon, so dusting off my math skills would be helpful.
- Start another computer science course or read an introductory book on data structures and algorithms. Along with learning more math, this is probably the best long-term use of my time.
- Learn the Python numerical toolkit in more depth. This is definitely the most practical option, but I think it has the least enduring value.
- Read more about software craftsmanship and project management. I have copies of The Pragmatic Programmer and The Mythical Man-Month, but I’m inclined to focus more on hard skills for the next few months.
If you’ve read this far, I’d appreciate your input! Send me an email and let me know where I went wrong.