On Programming


I’ve recently realized that I don’t really know what most jobs entail. Like, I vaguely know what an accountant is, but what do they do on a day-to-day basis? How about lawyers? I feel like those guys do a lot of reading, and maybe they go to court and argue sometimes, but surely that can’t take all day? Probably? What do doctors do when nobody is sick? Is there ever a time when nobody is sick? I get the distinct impression that maybe Dr. House isn’t an accurate forecasting model for answering this question.

As an engineer it’s easy to talk shit about other people’s jobs, I mean, after all, theirs probably didn’t require five years of school. I really do feel like probably I could do most jobs out there (though granted, not necessarily could I do them well), but, I’m starting to feel like maybe that’s just me being an asshole. Probably, come to think of it. Anyway, that’s where this question originally stemmed from. And that got me thinking: what does my job entail, really, when you get down to it? Since I don’t know the answers about other people’s jobs, I thought I’d spend today writing about mine.

I get into work around 9:00 sharp if I’m lucky, and the bus is feeling cooperative. Usually it’s not. 9:15 is more reasonable, and on bad days it’s closer to 10:00. I grab a snack and some coffee, and get to my desk where I spend the next few minutes going through the daily log-into-everything routine. While I’m working my way through my snack, I’ll halfheartedly check my email. There’s never anything directly for me, just newsgroup digests and things that I’m cc’d on. If there are more than 5 emails in my inbox, I seriously consider making some new email filters to keep them away from me. I don’t have time for emails that aren’t useful. I get maybe a thousand emails a week, of which there are maybe ten useful ones. You imagine how many email filters I have.

With that out of the way, I’ll fire up my text editor – a program which is really good at editing text. It’s like notepad, but on crack – and open up the project I was last working on. When I get in in the morning, I’m usually chalked full of ideas for how to solve my issue. I don’t actively come up with them, it’s just my brain having churned through the problem during the time I wasn’t at work. I’d say, honestly, most of my best work gets done when I’m not actually at the office. I’ll make the changes I thought of overnight, and run it through the test I’m currently trying to make pass. This takes around 5 minutes, which is too short a time to start something else, but too long to wait for it to finish. Usually I check Facebook.

When you study computer science in school, you learn about all of these really cool approaches for solving problems. Knowing these cool approaches well enough to write them yourself is how you get hired in this industry. A lot of it is trying to come up with a clever way of structuring the information you have so that it makes it easy to do the thing you want to do. I know that’s abstract and not very helpful – I apologize – so let me make it a little more concrete.

The problem: you have a million people in a city, all of whom have a telephone, and you want to be able to quickly find one so you can call them. The answer, of course, is you look in the phone book, which, if you consider it, is a pretty clever way of structuring information. It’s sorted by person’s name, so you can pretty easily track someone down. But imagine if it weren’t sorted – you’d have to go through the entire book and hope that the person you were looking for fell somewhere near the front. And you’d have to hope that you didn’t get bored and start skimming, or you might miss your friend without even realizing it. It’s a pretty obvious solution, but you’d be surprised how often the answer is “sort the bugger and then look through it”.

I know other things like this too. Most of them are general-purpose “shapes” to put data in, all of which are good for certain things and bad for others. In technical interviews, most of the marks come from picking the right shapes. These shapes are things called “data structures”, and as you might expect, they are used to structure data. Generally, if you pick the right data structure in an interview for a non-elite company, you will solve the problem and they will give you the job. It’s as easy as that.

There’s a data structure called a “stack” which corresponds in real life to those stacks of plates at the cafeteria. If you restrict yourself to only dealing with one plate at a time, the only things you can do to that stack are take a plate off the top, or put a plate onto the top. The plates have an ordering to them: first in, last out. As a result, the stack data structure is good for keeping track of where you are in a problem. Like in a good discussion with a friend, if your attention is sharp, you will make segues and digressions, but eventually meander back to the topic before your digression. And then you’ll finish that topic, which turned out to be a digression from the thing you were talking about earlier, and you’ll return to that. And so on.

There’s another one, called an “array”, which is like a shopping list. It’s easy to put things on the end of it, but it’s sorta hard to put something into the middle, because you didn’t leave any room to write there. But it’s nice for finding things in particular places (but it isn’t good at finding particular things). An array is like athletes finishing a race at the Olympics – what’s important to the Olympic Committee is that it’s easy to figure out whom to give the gold to: whomever is at the first spot in the array.

I know lots of things like that. Good ways to structure information if you want to do some things quickly, and don’t care about other things. I also know what it means for these things to be fast or slow, even when computers are trillions of times faster than we humans could ever be. This is the interesting stuff in my life. I spend an embarrassing amount of time thinking about these kinds of things.

But like I said these are things you learn in school. Unfortunately, in the industry, you get really excited when you get a chance to use these things. In the industry, these interesting decisions and trade-offs have been made by people with more seniority than you. My day-to-day job is spent wandering through huge swathes of code I’ve never seen before, trying to read enough to get a basic gist of it. What I’m looking for is enough of an understanding to see if it could possibly be causing my problem or not. This is known as doing maintanence programming, and nobody likes doing it.

Maintanence programming is like doing triage, or science, maybe. You start with some hypothesis of what is causing your program to misbehave, and then you try to confirm or deny that. Denying it can be as easy as chopping out 20% of the code and seeing if your problem is still there, or more involved like trying to find other problems that must be present if that were indeed the problem. It’s rarely the first hypothesis you consider.

Eventually, sometimes after days of hunting, you find where the problem is. Note that this is the effort just to locate where the problem is. Finding it is necessary to fix the issue, but it’s sadly insufficient. If you’re lucky, the problem is that your coworker forgot to consider what would happen in the case of that number over there being 0. If you’re unlucky, just because you’ve found the problem doesn’t mean you have any idea about how to fix it. If you’re really unlucky, you spend a few hours figuring out how to fix it, and that just reveals a related problem somewhere else. Back to step one. And if you’re cursed, you realize that the guys with seniority who did the conceptual designing of this system didn’t do their job very well, and fixing this tiny problem is going to require a huge architectural renovation.

That happens depressingly often.

This is what I spend my days doing. I’d say maybe 80% of my actual productive work day is spent tracking down and subsequently thinking about how to fix these bugs in our giant codebase. The good news is that it’s always a new problem you’re working on. The bad news is that it constantly feels like you’re drowning in things you don’t understand. Luckily it gets better. If you can hack it for a year, you start getting really quick at finding and fixing these problems. Then because management decides your efficiency is discouraging to the intern, they’ll switch you to building a system of your own. This is the everlasting dream of all programmers: to spend all their time writing code, and leave fixing it to the interns. I’m glad I’m not an intern anymore.

My favorite part about being a programmer is building tools for myself to automate tedious parts of my job. For example, yesterday I needed to test the same thing with 30 different, tiny modifications, and see if any of my tests worked. Each edit took about 15 seconds, but running the tests took about 5 minutes. After two iterations of this, I decided to write a program that would make the changes for me, run the tests, and start flashing at me if it found the right answer. Writing this tool took maybe ten minutes, and saved me two hours of tedious, can’t-do-anything-else-ness. I got lucky on that one, usually writing the tool takes as long as doing the manual labor, and doesn’t work very well, but at least it’s mentally stimulating. Writing tools is like intellectual masturbation for programmers, I swear to god.

To be a good programmer, you need to be able to focus. Intently. Remember, you’re trying figure out why a computer program doesn’t work by running it on your brain, really, really slowly. This is why programmers are stereotypically hyper-pedantic; it’s a useful skill when your livelihood depends on not making any mental missteps when you’re pretending to be a computer. Imagine pretending to be a computer. It’s not a very appealing past-time, is it?

That’s not to say it’s all bad. Programming gets you really good at pattern-matching. We see the world in a way that can’t possibly make sense to anyone else. In the same way that plates at the cafeteria and discussions with your friend and gridlock are all just stacks when you squint a little bit, sequentiality and the future and uncertainty and incremental changes are all just monads if you look hard enough.

Over time, as a programmer, your perception of reality lifts itself into the same abstractions you use all the time at work. I no longer see the woman at the counter at Subway as a person; for most of my intents and purposes, she’s just an abstract class which implements the IFoodProvider interface which can give me back a Future[Food], which if I block on, will eventually turn into real food. I don’t mean to do it, and I don’t feel like a good person when I realize it’s happening, but like it not, that’s my reality. Sometimes I need to work really hard to realize there are real things like people or intentions behind these visages I see in terms of functions and patterns.

This way of the world doesn’t really paint me as being a warm and cuddly person, but I think it’s helpful. It’s a buffer between me and making bad decisions. It helps distill interactions to their constituent parts, and filter out the irrelevant ones. I’m often told that I’m good at asking questions, and I take this to be a direct consequence of my being good at seeing the problem at hand and ignoring the bullshit.

I’ve spent the last decade of my life doing this stuff. I feel like if properly motivated, I would never need to stop writing this essay. There are lots of things I want to say; lots of things I feel are misunderstood. But I won’t, because this is already long enough, and scratches enough an itch for understanding that I’ll leave it be.

If you’re going to come away with anything from this essay, I’d suggest it not be “Sandy sees me as a mathematical construct,” but instead “I should talk more about the intricacies of my job because it’s misunderstood and I really like it,” or, y’know, even “wow, Sandy must be really smart” would be acceptable. I wouldn’t complain about that one.