This is episode 55, on the Interface Segregation principle, or ISP. ISP is the I in the SOLID Principles of object-oriented design.
Hi and welcome back to Weekly Dev Tips. I’m your host Steve Smith, aka Ardalis.
This is episode 55, on the Interface Segregation principle.
This week's tip is brought to you by devBetter.com.
You can find help and advice all over today, from twitter to Stack Overflow to various Slack communities. These are great, but if you need more, if you need someone you can count on when you reach out for advice on code or your career, check out devBetter.com. Read the testimonials. Join risk free and see if it's for you. Not only will you learn from me, but if you stay for a while you'll grow and begin to be a mentor for others, gaining valuable confidence and experience. I've watched it happen.
Returning back to the next SOLID principle, this week we're on the I as in Interface. Yes, it's the Interface Segregation Principle! Or ISP for short. Let's start by defining what this principle says. ISP states that no client should be forced to depend on methods it does not use. That's it. You can think of ISP as basically suggesting that you should prefer small, cohesive interfaces to large interfaces. In episode 49, which hopefully you already heard because you're listening to these in sequential order, we talked about the Single Responsibility Principle. These two principles are related! If you have classes that have only one responsibility, they're likely to expose a small, cohesive interface.
That might be worth diving into a bit more. I'm guessing that when some developers, especially C# or Java developers, hear about ISP, they immediately map the word 'interface' to the interface type and keyword in these languages. That's not wrong, it's just not fully right, either. The public methods and properties a class exposes are its interface, even if it doesn't implement any interface defined separately. And ISP applies to implicit class interfaces just as much as it does to explicit interface definitions declared with the interface type. If you keep this in mind, it may make it easier to see how SRP and ISP relate, since SRP doesn't make any mention of interface types, only classes.
The previous principle we discussed, Liskov Substitution, also relates to ISP. If you have a small, cohesive interface on a class, then when you decide to inherit from it you only have a few methods to worry about. If you have a massive number of methods, and your subtype is only concerned with a subset of them, it's much more likely that you won't fully implement all of the class's methods, potentially leaving some methods to throw NotImplementedException. Thus, it's easier to follow LSP if you first follow ISP.
What's the point of having a smaller interface instead of a larger one, though? What's the down side to having fewer, larger classes with many methods on them? Certainly there's an argument to be made, and many have, that it makes it easier to find methods when you don't have so many different classes to search through to find them. The goal of ISP is to reduce dependency between types and especially between components in a system. The more types that depend on a particular type, the more difficult that type is to change. The more inertia it has in the system. If you have a class in your application that's used by 100 other classes, how likely are you to make major changes to its design, compared to one that you just wrote that has literally no other classes using it yet? Of course it's easier and WAY less risky to change classes that have few references to them.
Let's say you want to change something fundamental about how that class that's used by 100 other classes works. One way you could do that would be to create a new instance of that class's interface or abstract base class, but then you'd have to implement all of its methods. Odds are, there are a bunch of them. That's not going to be easy - so instead you might just add some more conditional logic to some of the methods. Maybe add a few more method parameters to some methods. Maybe add a couple of new method overloads. And guess what? The problem just got bigger. The class just got bigger. The inertia just grew.
By having a small interface and being focused on a small set of functionality, ISP-following classes are easier to change an easier to replace. By being small, they're easier to subtype or implement, making it easier to build systems that are modular. You find a class that has two methods and you want to change how it works in some situations? Just write a new class, implement the two methods, and swap it in where appropriate. Prefer writing new code to change legacy systems rather than modifying existing code. By following ISP, it's easier to replace classes with different variants that implement the same interface, which results in more loosely coupled and modular designs. These designs are easier to maintain and easier to test. You can learn more about ISP and the rest of the SOLID principles from my courses on Pluralsight, linked from this episode's show notes on weeklydevtips.com/episodes/55.
That’s it for this week. If you want to hear more from me, go to ardalis.com/tips to sign up for a free tip in your inbox every Wednesday. I'm also streaming programming topics on twitch.tv/ardalis most Fridays at noon Eastern Time. Thank you for subscribing to Weekly Dev Tips, and I'll see you next week with another great developer tip.