Weekly Dev Tips

Liskov Substitution

Episode Summary

This is episode 53, on the Liskov Substitution principle. The typical definition used today is that subtypes should always be substitutable for their base types. If this isn't the case, it likely means you have the wrong inheritance relationship, or your interface isn't cohesive enough, or there is some other code smell you should consider addressing through refactoring.

Episode Notes

Hi and welcome back to Weekly Dev Tips. I’m your host Steve Smith, aka Ardalis.

This is episode 53, on the Liskov Substitution principle.

Liskov Substitution Principle

This week's tip is brought to you by devBetter.com.

Sponsor - devBetter Group Career Coaching for Developers

Need to level up your career? Looking for a mentor or a the support of some motivated, tech-savvy peers? devBetter is a group coaching program I started last year. We meet for weekly group Q&A sessions and have an ongoing private Slack channel the rest of the week. I offer advice, networking opportunities, coding exercises, marketing and branding tips, and occasional assignments to help members improve. Interested? Check it out at devBetter.com.

Show Notes / Transcript

We're in the middle of the SOLID principles this week, with the only one who is named after an individual. The Liskov Substitution Principle is named for professor Barbara Liskov, who teaches computer science at MIT. She gave a conference keynote some years ago in which she defined substitutability in object-oriented systems, essentially stating that one type could be considered substitutable for another if there were no situations in which the original type worked but the new one did not. The typical definition used today is that subtypes should always be substitutable for their base types. If this isn't the case, it likely means you have the wrong inheritance relationship, or your interface isn't cohesive enough, or there is some other code smell you should consider addressing through refactoring.

So, although it isn't mentioned in the name, LSP is all about inheritance hierarchies. That's the first thing to remember. Many developers, myself included, learned about inheritance with the aid of the IS-A relationship. That is, you can consider using inheritance to model a problem if there are two related concepts and one IS-A more specific version of the other. You might have a Car class, with an SUV subclass. Or an Animal class with a Mammal subclass. These work because an SUV is a car and a mammal is an animal.

Sometimes, though, you might have an IS-A relationship, but it's not sufficient and it leads you down the wrong OO design path. The classic geometry examples involve squares and rectangles and circles and ellipses. A square is a rectangle; a circle is an ellipse. But take care if you try and use inheritance to model them, since depending on how you do so you could introduce breaking changes into your application. Imagine if you have applications using your Rectangle type that assume they can independently set the height and width of the rectangle. Then someone introduces the Square type that inherits from Rectangle and implements height and width properties so that setting either one will set the other, too. This ensure the resulting type is always a square. But it also means that a method that takes in a rectangle and modifies its height and width to different values will break, because the first value set will be silently overwritten by the second.

Another classic example involves birds. What if your base bird type includes methods like Fly and properties like Altitude? Then, someone introduces subtypes for Ostrich and Penguin. Trying to substitute these in to methods that expect certain behaviors from Fly and Altitude is likely to cause problems. The problem here is that the initial assumption that all birds can fly was inherently flawed.

In business software, we don't usually worry about birds or squares, but we often model things like policies, accounts, orders, etc. and often there are different varieties of these concepts. You'll know when you're violating LSP when you see code that shouldn't care about the specific variety of something, but has to run a type check to do its work properly. A method takes in an Account, and then checks to see if Account Is PremiumAccount, for instance. This signals an LSP violation, and can usually be corrected with a refactoring that shifts behavior up or down the inheritance tree. Pretty much the only place where you might have a conditional on the specific type of an instance of a class is in a factory method that's responsible for creating that specific type. Otherwise, the Replace Conditional with Polymorphism refactoring is your friend. Use it to eliminate excessive conditional logic from your code by making proper use of inheritance in your object model.

Obviously examples of these principles are tougher to describe in a podcast than in video, so to see them in action check out my courses on Pluralsight on the SOLID principles. I just produced a revised course for 2019, but my original SOLID principles of object-oriented design course is out there as well and uses different examples, so you might get benefit from seeing both of them. You'll find links to both in this episode's show notes.

Show Resources and Links

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.