The “don’t repeat yourself” principle is well established, but over-aggressive refactorizarions to extract common code are also widely known for creating hard to maintain code due to the introduction of tight coupling between components that should not be coupled. A passing resemblance between code blocks is reason enough to extract them away, even if that ends up breaking Liskov’s substitution principle.
To mitigate problems caused by DRY fundamentalisms, the “write everything twice” (WET) principle was coined. WET works by postponing aggressive refactorizarions, the kind that introduces complexity and couples unrelated code just because it bears some resemblance, by creating a rule of thumb where similar code blocks showing up twice in the code should not be refactored, and only code that shows up multiple times should be considered for this task. However, this rule ignores context and nuances, and can dissuade developers from cleaning up code.
So, where do you stand on the topic? How do you deal with duplicate code? Do you follow any specific rule of thumb?
If I find myself repeating more than twice, I just ask “Can this be a function”. If yes, I move it there. If not, I just leave it as it is.
Life’s too short to spend all day rewriting your code.
Important/generalized patterns reveal themselves over time. I generally push for people to just write the specific thing they need for the given context and then if the same pattern shows up elsewhere, promote the code to a shared library. It’s really hard to anticipate the correct abstraction right from the start, and it’s easy to include a bunch of “just in case” parameterization that bloats the interface and adds a lot of conceptual overhead.
That being said, with low-leverage/complexity code like html views, repetition isn’t as problematic. Although, I do think that HTML components have matured enough to be the correct unit of abstraction, not CSS classes.
What does the code represent? What does it concern?
Focusing on the code and pattern too much may mislead. My thinking is primarily on composition and concern. The rest follows intuitively - fee with risk, gain, and effort assessment.
I’ve had occasional instances where code duplication is fine or not worth to fix/factor. But I feel like most of the time distinct concerns are easy and often important to factor.
I’ve habe
found the german.
Lol, I guess I habe wrote it on mobile with autocorrect 🙃
- When writing new code: make a good faith attempt to DRY (obviously not fucking with Liskov in the process). First draft is very WET - deleting things is easiet at this stage anyway and accidentally prematurely DRYing causes headaches. Repeat the mantra “don’t let perfect be the enemy of good” to prevent impulsive DRYing.
- When maintaining existing code I wrote: Refactor to DRY things up where it’s clear I’ve made a conceptual misjudgement or oversight. Priority goes to whatever irritates me most in the moment rather than what would be most efficient.
- When altering other people’s code: Assume they had a reason to WET (if you’re lucky, read the docs that discuss the decision) but be a little suspicious and consider DRYing things up if it seems like a misjudgement or oversight. Likely realise 50% of the way through that this is going to take much longer and be way more painful than you hoped because of some esoteric bullshit compatibility issue. Curse yourself for not letting sleeping dogs lie but still start engaging in sunk cost fallacy.
- When reviewing PRs: Attempt to politely influence the writer to DRY it up but don’t take too much offence to WET if it has a decentish reason in context. Throw in an inline one-liner to not make other maintainers question their sanity or competence when they realise they’re reading duplicate code. Also to more reliably establish git blame rather than blaming the next poor fool who comes along and make a minor reformatting revision across the file. Include a date so that someone can stumble across it in 10 years as an archaeological curiosity and their own personal esoteric bullshit compatibility issue.
- Long term maintenance: Live with your irritatingly damp mouldy code. There’s new code that needs to be written!
I prefer the FP approach where I create smaller functions that I compose together in larger functions or methods wich rarely repeat themselves elsewhere identically. Forcing extractions and merging of such functions often leads to weird code acrobatics.
WET/DRY-ness is like a property of code – a metric or smell perhaps, but not something to goal towards. That’s like asking whether you drive fast or slow and whether we should all drive faster or slower.
The implementations mostly don’t matter. The only thing that you need to get right are the interfaces.
I liked this talk on the subject: https://www.deconstructconf.com/2019/dan-abramov-the-wet-codebase
It’s a nice explanation of how it’s less about code that looks the same or currently performs the same operations, and more about what it means.
I don’t think DRY or WET or the “rule of 3” can address the nuances of some abstractions. I think most of the times when I decide to abstract something, it’s not because the code looks repetitive, but because I’ve decomposed the architecture into components with specific responsibilities that work well together.
In other words, I don’t abstract from the bottom up, looking at repetitive code. I abstract from the top down, considering what capabilities are required from a system and decomposing the system to deliver those capabilities.
Of course, repetitive code might be a smell, suggesting that there is room for abstraction, but I don’t design the abstraction based (entirely) on the existing code.
Of course functions still shouldn’t do more than one or two things. Hence they don’t get too complex.
And yeah, duplicates to avoid thight coupling (if still needed) are fine, if kept in moderation.
If the only difference between two classes or structs is hard coded config, rewrite to be a single implementation and pass the configs in.
If it’s more in depth than that it may not be worth refactoring but future copies should be designed more generically.
I can very much recommend taking a look at the AHA principle. Kent C dodds has a short and good presentation on it. That’s what i go by these days
It depends on how much regression testing is in place to test that old and refactored code behaviours have the same outputs, and how much budget there is for writing this tests.
For old financial systems for example, the answer is often to repeat the code again, as there’s likely to be little tests to confirm existing behaviour and writing tests around very complex business domains is prohibitively expensive.
There are 2 metrics that need to be considered:
- easy to read
- easy to modify
The first point is by far the most important. Usually DRY win because less code means less to read so less to put in your head. But if the abstraction is too complicated (for example because there are two many parameteres) then it’s worth considering drying.
And don’t forget the second point, but do not overthing and YAGNI. Sometime a simple comment “don’t forget to update method
foobar()” is enough. Don’t forget either that you can always rewrite the abstraction when you need to modify something (if the original did not fit your new requirements), but for this to be an easy task, the understanding of the original abstraction must be crystal clear. That’s why the first point is so important.I just move any duplicate code into a function, no issues yet. (In face, fixing a single bug often ends up fixing multiple problems)