“Bad abstraction is worse than duplication”
Oh shit, thank you so much for this part ! I don’t even count number of time I had to face enthusiasts developers saying "These lines of code are very similar, let’s factorize them ! "
And that’s how MathManagerHelper class is born…
God yeah - this hits hard with me. I work in a place where the DRY principle is chanted like a mantra and I have to push back against it constantly. It’s like the one principle jr. devs learn (along with “never hardcode anything”) that just gets applied to everything whether it makes sense or not.
"These lines of code are very similar, let’s factorize them ! "
“We should take three simple functions that run similar (but different) things and make them call a single large complicated function that takes a parameter that tells it how to execute!”
::shudder::
You just reminded me of the early days at a company I worked where the factory pattern was absolute. Their were interfaces and factories for fucking DTOs. It was insanity! That was the place where I really learned the concept of “cargo cult”
What pattern should be used instead?
“The one that makes sense”
Patterns and principles are guidelines, not rules. You don’t want to just apply them blindly.
You need to balance principles. Over-applying the DRY principle can lead to more complex logic that’s harder to understand than if you just wrote code in-line.
The rule I see is functions should be fairly atomic and almost obvious what they do in context of the code.
At least for my small brain that’s how I like it. I can understand some complex abstractions but rembering that actually this function behaves in three different ways depending on what flag is set is awful. It means you could look at one example and be totally wrong in another. Ideally you could guess the functions purpose even in a black box setting based on inputs outputs and the name should then make it obvious.
The example that the other commenter gave did not require the user to input the flags. As far as I understand, they mean there would be a number of secondary functions that will call the other with the correct parameter.
Fair point. My point still stands on it breaking the black box test. Where the input can wildly effect the logic that creates the output.
+++ I cannot stress this enough.
Programmers tend to get over obsessive about DRY code because it seems like it should be easier to maintain fewer functions and writing less code should be easier right? But it tends to ignore the actual subtle differences between the different consumers of that abstraction, and that leads to massively branching functions that couple unrelated parts of the codebase together. These then become impossible to read and change without worrying about breaking something else.
When I was at Thoughtworks (Martin Fowler’s current company), we were also taught the rule of three for refactoring, and were taught to always optimize your code for readability and maintainability first and foremost. Refactoring and improving easy to read/maintain code is always relatively trivial, by nature of it being clear, concise and decoupled, and that can be done once you have a reason to do so (i.e. performance issue etc), so there’s no point over optimizing for anything else up front.
This talk from Dan Abramov (one of the lead devs of React) does a really really good job explaining some of these issues and is entirely worth the full watch/listen (do it on company time, it’s professional development):
https://www.deconstructconf.com/2019/dan-abramov-the-wet-codebasets
My rule of three: consider rewriting any article, blog, exhortation, argument, leaflet, or wiki page that uses more than three fonts, sizes, colors, or styles.
