Don't DRY Your Code Prematurely
Don't DRY Your Code Prematurely
This is another post in our Code Health series. A version of this post originally appeared in Google bathrooms worldwide as a Google Tes...
DRY = Don't repeat yourself
Ultimate DRY: just keep refactoring the one method to accept hundreds of parameters and do everything.
Add two numbers? DoIt(1, 2);
Subtract? DoIt(null, null, 3, 1);
Etc.
invokeOperation(new Object[]("multiply", 2, 5))This guy seniors
I’ve always understood DRY to be about not duplicating concepts rather than not duplicating code.
In the example here, you have separate concepts that happen to use very similar code right now. It’s not repeating yourself as the concepts are not the same. The real key is understanding that, which to be fair, is mentioned in the article.
IMO, this is where techniques like Domain-Driven Design really shine as they put the business concepts at the forefront of things.
IMO, this is where techniques like Domain-Driven Design really shine as they put the business concepts at the forefront of things.
Do you have a resource on where to learn DDD? I feel like I never understood the concept well.
As already mentioned, the blue book by Evic Evans is a good reference, but it's a ittle dry. Vaughn Vernon has a book, "Implementing Domain-Driven Design" that is a little easier to get into.
Personally, I found that I only really grokked it when I worked on a project that used event-sourcing a few years back. When you don't have the crutch of just doing CRUD with a relational database, you're forced to think about business workflows - and that's really the key to properly understanding Domain-Driven Design.
If you can work with python cosmic python (http://www.cosmicpython.com/book/preface.html) is a great resource
"Domain Driven Design" by Eric Evans, aka the blue book. It's very dense however and very object oriented, but concepts apply even if you dont work with object oriented languages, you might have to do more footwork to get from a domain model to services that adhere to the model.
"Head first Software Architecture" might be an easier on ramp and touches on simmiliar concepts.
That’s how DRY is described in Pragmatic Programmer, where DRY was first coined. They’re clear that just because code look similar, doesn’t necessarily mean it’s the same.
yes, this is exactly what you have to think about. the left example even aknowledges that deadlines for "tasks" might be different from deadlines for "payments", which suggests that the abstraction is not "clean".
It should be about concepts but it's more often applied to duplicate algorithms by inexperienced people (which is a huge mistake).
I guess I never thought of it like this, but it resonates.
This is silly. Everyone knows that DRY is telling you that if you do the same sequence of mouse clicks three times in a row, you should spend the day writing a script to automate the task instead of quickly finishing what you were doing by doing the same sequence of clicks a fourth time. If you are supposed to apply it to the code you write, then there'd never be boilerplate-heavy languages like Java.
You are wrong.
DRY has been applied to written code at least since the book Clean Code, which has been widely recommended to programmers for nearly two decades.
I'm not making a stance here about whether or not that is the right recommendation, but saying that DRY was never meant to be applied to "the code you write" is simply wrong.
(I remember DRY being pushed before that book, but it's the only primary source I have at this moment.)
First off, I generally don't worry about DRY until there are 3 instances, not 2. With only 2, it's really easy to over-generalize or have a bad structure for the abstraction.
But otherwise, I disagree with the article. If it's complicated enough to bother abstracting the logic, the worst that can happen in the above situation is that you just duplicate that whole class once you discover that it's not the same. And if that never happens, you only have 1 copy to maintain.
The code in the article isn't complicated enough that I'd bother. It even ends up with about the same number of lines of code, hinting that you probably haven't simplified things much.
I personally factorize as soon as there are two copies, but do not hesitate to inline the code and redo the abstraction when there is a 3rd use if it doesn't fit. I find it much easier to inline and re-abstact a bad abstraction, than check if two copies are indeed identical.
The exception is business logic. Usually I want all of them to be dupplicates because there is a very high chance that it's just accidental that part of the logic is similar. I take great care to have good primitives but the actual business logic that glue those primitives together is written as many time as needed.
Yeah, I'm reminded of how Germanic languages used to have singular, dual and plural. If we'd still had dual, we'd probably also be talking about not abstracting until we actually have a plural.
The code in the article isn't complicated enough that I'd bother. It even ends up with about the same number of lines of code, hinting that you probably haven't simplified things much.
I think it's a good example of the problem though. People take that same idea and apply it too liberally. The point isn't that specific code, it's about not apply DRY to code that's coincidentally identical.
But otherwise, I disagree with the article. If it's complicated enough to bother abstracting the logic, the worst that can happen in the above situation is that you just duplicate that whole class once you discover that it's not the same. And if that never happens, you only have 1 copy to maintain.
That's... Not at all true in practice. What often happens with these "DRY" abstractions when they've been improperly applied is you end up with an inheritance hierarchy or a crazy template or some other thing. You're really lucky if you can just copy some code and find your way out of the weeds.
There are plenty of bad abstractions in the wild and novices applying DRY is a common source of them.
There are plenty of bad abstractions in the wild and novices applying DRY is a common source of them.
You're both saying the same thing though. Novices aggressively apply DRY the moment a second bit of identical code appears, while experienced developers often wait for a third copy and then think about whether DRY fits.
That said, I think "don't apply DRY too aggressively is the whole point of this discussion, and the person you're replying to was kind of needlessly disagreeing.
Another older blog post saying the same: https://sandimetz.com/blog/2016/1/20/the-wrong-abstraction