A popular feature in Rails called "mixins" came to PHP a while back under the name "traits". Traits let library developers give you a lot of functionality with little to no effort by injecting code into your classes that runs as if originally authored in them.
Just toss a line of code into the class you want to augment and you're off to the races.
The one snag I found when using traits however is that they don't participate in any kind of type system. If desperate enough, you can use reflection to pull out whether a class has one applied or not, but that's only half the battle and a lot of work each time.
This puts me in a slight bind when I'm interacting with classes using a trait I've made as I have no way to type check or hint them in.
It's pretty obvious what I'm doing here. I'm basically writing an interface to add to a class and then creating a trait that fulfils that interface.
Interfaces are a good idea, no question about it. It makes code easier to read and follow when you see classes that use them. Burdening consumers of your library with writing menial code however is not as useful and to a lesser extent so is forcing an implementation with traits.
Using this approach lets someone interested in your library take exactly as much as they need and leave the rest behind. If all they want is to stay in sync with my library and use the defaults, my trait and implementation will always be in sync, they don't have to worry. If however for some reason they need to abide by my library's contract but swap out how it's done? They're fully empowered to do so!
I don't suspect I'm the first person to think of this, however I did come up with it on my own and it seems like a very considerate way to get the best of both worlds. Enjoy!
Just toss a line of code into the class you want to augment and you're off to the races.
The one snag I found when using traits however is that they don't participate in any kind of type system. If desperate enough, you can use reflection to pull out whether a class has one applied or not, but that's only half the battle and a lot of work each time.
This puts me in a slight bind when I'm interacting with classes using a trait I've made as I have no way to type check or hint them in.
But then it dawned on me...
interface MyInterface {
public function thisMethodIsImportant();
}
trait MyInterfaceImpl {
public function thisMethodIsImportant() {
return "Thanks for not forgetting about me!";
}
}
It's pretty obvious what I'm doing here. I'm basically writing an interface to add to a class and then creating a trait that fulfils that interface.
This leaves us with "why?"
I think most of the time when someone is going to use your interface, they're going to wish that they didn't have to actually implement the methods it sets out. Which teaches us a little about how traits came to be: they're purely convenience.Interfaces are a good idea, no question about it. It makes code easier to read and follow when you see classes that use them. Burdening consumers of your library with writing menial code however is not as useful and to a lesser extent so is forcing an implementation with traits.
Using this approach lets someone interested in your library take exactly as much as they need and leave the rest behind. If all they want is to stay in sync with my library and use the defaults, my trait and implementation will always be in sync, they don't have to worry. If however for some reason they need to abide by my library's contract but swap out how it's done? They're fully empowered to do so!
I don't suspect I'm the first person to think of this, however I did come up with it on my own and it seems like a very considerate way to get the best of both worlds. Enjoy!
Comments
Post a Comment