My first experience working with legacy software was on a large business product back in 2015 — my first year of full-time employment. Although I’d worked on teams during internships and in student projects before and started my own side projects, prior to then I’d never touched something with years and years of accumulated technical debt. That experience would shape my career for years to come and helped inform my personal most cherished programming rule: “Don’t Write Bad Code”.
At the time, large swathes of the project were Frankensteinian mismatches of various technical efforts over the years. (To the credit of the engineering team, I believe this is generally no longer the case.) Any given engineering task had a non-trivial probability of being unexpectedly nigh impossible without either a major cleanup & refactor or an excruciating addition of even more technical debt.
As a software developer or adjacent reading this post I hope your reaction contains hope the team often chose the cleanup & refactor. As a realist I can tell you that that over the years, with business goals and deadlines perpetually looming, cleanups often “couldn’t” be prioritized.
Don’t Write Bad Code: A Mantra
I, an annoyingly passionate 22-year-old, came away from the legacy code project with a mantra that I would never write bad code again. I was very proud of that mantra. It has few nuances I’ve honed over the years that I’m excited to go over here. With those nuances I think it might be better phrased as “don’t allow yourself to write bad code” but that doesn’t quite roll off the tongue as well.
The mantra originally used a four-letter word more vulgar than “bad” (I was very angsty). For SEO and approachability purposes I’ve toned it down.
The mantra’s main drive is that I want to always be put in a situation where I can make my best work. I don’t want to be continuously forced to stunt my learning or write sub-optimal code.
Ground Rules
What even are “bad” and “good” code anyway?
I don’t know. Nobody really knows how to define code quality. We can look at a pull request and suggest alternate ways to accomplish some results, but such reviews are always subjective and sometimes wrong.
I fear that someone could read blog posts such as this one and come away with a drive to write perfect code and naysay all other code as inferior. Please don’t do that.
This blog post’s context is that “bad” code makes it harder to accomplish tasks well. Common violating patterns and their results include:
- Copy & pasting many files instead of DRYing: as updates to the code later on would have to involve repeated work
- Creating DRY abstractions too early in development: as it locks code into a pattern that likely will not scale
- Using a framework with patterns it recommends against: as it adds the burden of actively fighting the framework, is a common source of bugs, and makes onboarding new team members more difficult
Still, every tech stack is different, and every project uses its stack differently. Sometimes it’s reasonable to use one of those violating patterns or others, such as copy and pasting code in a project slated to be deprecated. YMMV.
Growth Mindset
To consider yourself finished learning a topic is to say that you have accumulated all its information, including ideas that haven’t been learned yet. That’s rather arrogant, don’t you think?
Carol Dweck’s research into mindsets created one of the few buzzphrases I actually respect: growth mindset. No matter your situation, you generally can and should adopt the resources around you to learn from.
I was once adjacent to a React project that paired several relatively senior backend engineers with several relatively junior frontend engineers. One backend engineer took it on themself to enforce what they saw as proper backend programming principles in the frontend React code. They stubbornly refused to accept inputs from the other developers that doing so was leading the codebase down the wrong path. It was a pity. The perceived backend practices did not apply to React. The rejected frontend inputs actually were canonically correct React best practices. The team suffered from the conflict.
Learn Before Committing
Beyond failing to work effectively with other engineers, the first key mistake the developer in that adjacent project made was to assume that their pre-existing ideas could be copy and pasted into a totally new area.
Don’t be like that one developer. Validate whether you have more to learn before committing to an approach for a project. If you’re working in tech stack you haven’t worked with before then you almost certainly will need to understand best practices for each unfamiliar part of the stack: framework(s), language(s), and all.
I like to use a prerequisite of learning as a personal enforcement strategy for not writing bad code. Until I have validated that I understand the right way to approach a project, I don’t feel comfortable believing the code I write to not be at least sub-optimal. In other words, I assume bad code until I prove otherwise to myself.
Learning Over Time
On a related note, modern programming practices are constantly evolving. What was considered a good practice at the start of a project may be considered outdated -even “bad”- only a year or two later.
Part of not writing bad code is paying attention to the major trends of projects I’m working on. I am a frontend developer, so my areas of code are undergoing frequent and rapid changes in best practices.
I want to always be on a team that gives me time to explore new ideas from the greater industry. It should allow team members to introduce those ideas to a supportive environment. If the ideas are worth adoption, the team should be willing to seriously make an effort.
Leave If Necessary
Nobody likes leaving a team. After putting months or years of thought and time into an area of code and group of people, it can feel like you’re abandoning both the code and the people.
At the end of the day, though, either your team is helping you learn how to write “good” code (whatever that means for your area of programming) or it is not. If it is not, or is but at a slower rate than you are likely to get elsewhere, consider whether you are at a stage in your career where you are ok with not learning.
I left the team with all the legacy code. I liked many of the people on it and still could have learned a great deal from them. The team just wasn’t the right fit for me. I wanted to grow not in work effectively with legacy code with them but instead in understanding modern programming techniques and a more modern tech stack. It was the right decision for me.
You Can Break the Rules
I broke my most cherished programming rule this month. I wrote bad code.
We had a sudden time-sensitive high-reward project come up that happened to touch areas of our systems that already had technical debt the team was hoping to tackle long-term. I again saw the dreaded choice between short-term gain and long-term health, and I chose short-term again.
Shame on me? No! It was a valid decision at the time. Taking on some technical debt allowed the project to be successful within its limited time-frame and with a great deal of benefit to the business.
Sometimes You Have To
Sometimes you have to break your rules. As with a Jewish doctor performing life saving surgeries during Shabbos, or a firefighter breaking down a stranger’s door to save their life, all reasonably written rules contain escape hatches for emergencies. In my much less critical case I broke my “Don’t Write Bad Code” rule because the business opportunity far outweighed the downsides of the added tech debt.
I still like to “punish” myself when I break the rules, though: not because I have to, but because programmer’s penance is cathartic. When I cause a bug on production I like to fix it and another few bugs. When I add tech debt I like to clean it and other existing tech debt after.
This past month I notified stakeholders and technical owners that the systems needed to be overhauled soon, then made sure ownership handoffs preserved that information. I also dedicated time to fixing adjacent bugs and removing a few deprecated components.
When was the last time you wrote something that you knew was truly not the right solution? Why did you do that? Was it justified? What did you do to make up for it?
Liked this post? Thanks! Let the world know:
