Introduction to Dependency Injection

How to understand fancy terminology around dependency injection

7 min readApr 11, 2023

In this article, I want to give insight into dependency injection. The article will explain the dependency injection, inversion of control, composition and separation of concerns. Do not expect any extensive academic terms. This is a simplified version, which I would find helpful in the past.

Little heads up. It is expected, that you have basic knowledge of Java/Kotlin language or similar language to fully understand code snippets.

What is dependency injection?

The funny part is that dependency injection is only a technical term for a simple process. During any programming course, you are shown the creation of classes. Let's create one.

class Boat(private val model: String, private val id: String) {
  
  fun getIdentity(): String {
    return "$model: $identity"
  }

}

In code, there is one class with two constructor variables. This is the simplest way to demonstrate the dependency injection. The act of passing objects to other classes is called dependency injection. The Boat class is dependent on the model type and id strings — dependencies.

In reality, the program manipulates with much more complex dependencies. These dependencies contain programmed behaviours to our liking. Consequently, the program can grow into the required complexity.

Above we can see the simplest version of dependency injection. Let's go further for a more complex example of the Boat, which should get us from point A to point B.

The boat depends on the navigator and engine.

class Navigator{

  fun navigateTo(double lat, double lon) {
    Log.i("Autopilot", "Going to latitude: $lat, longitude: $lon")
  }

}

class Engine {
  
  fun start() {
    Log.i("Engine", "Starting engine")
  }

  fun stop() {
    Log.i("Engine", "Turning off engine")
  } 
  
}

class Boat(private val engine: Engine, private val navigator: Navigator) {

  fun goToNewLocation(double lat, double lon){
    engine.start()
    navigator.navigateTo(lat, lon)
    engine.stop()
  }

}

In the example, the Boat class depends on two other classes: Engine and Navigator. With these components, the Boat can take us to different destinations. Now, our Boat is equipped with only one specific Engine, but what if we would like to change it or have multiple versions of the engine?

The engine is inherited by its subclasses. There can be as many of them as we want.

class Navigator{

  fun navigateTo(double lat, double lon) {
    Log.i("Autopilot", "Going to latitude: $lat, longitude: $lon")
  }

}

interface Engine {
  fun start()
  fun stop()
}

class HightechEngine: Engine {
  
  fun start() {
    Log.i("Hightech Engine", "Starting engine")
  }

  fun stop() {
    Log.i("Hightech Engine", "Turning off engine")
  } 
  
}

class Paddle: Engine {
  
  fun start() {
    Log.i("Paddle", "Paddling")
  }

  fun stop() {
    Log.i("Paddle", "Taking pause")
  } 
  
}

class Boat(private val engine: Engine, private val navigator: Navigator) {

  fun goToNewLocation(double lat, double lon){
    engine.start()
    navigator.navigateTo(lat, lon)
    engine.stop()
  }

}

In the code above, the Engine acts as a generalisation for all engines. Every Engine can be started or turned off. Moreover, thanks to this generalisation, the code can cover as many engines as we want. The code implements a new super high-tech engine class HightechEngine and Paddle(yes, the code can cover the paddle if we want to).

Technically said: These classes inherited the Engine class and they implement the properties of this class. The Boat class is dependent on the Engine class, but it can be any class, which inherits the Enginclass.

To create multiple boats with different engines, it would be enough to:

val expensiveBoat = Boat(HighTechEngine(), Navigator())
val cheapBoat = Boat(Paddle(), Navigator())

From this point, these two boats have the same behaviour from a code structure perspective, but their behaviours during the execution can differ.

Inversion of control

The previous example shown us how to abstract the logic for multiple engines. In other words, the Boat class is unaware of the Engine subclasses. The Boat can call the methods in the interface / abstract class and nothing more. However, the Engine subclass can carry any logic inside of it. Some of them can make the boat fast, and some of them slow. It depends on the Engine itself, not the Boat as a whole.

Technically said: The Boat gives controlls to the Engine and gives out the responsibilities for the logic encapsulated in the Engine class. The Boat decides, when to call the Engine, but what will Engine do is up to the implementation of the Engine itself. The Engine can provide return values or different callbacks, which can Boat use to integrate other components.

If the Boat is called, it perceives any subclass of the Engine as any other engine. But under the abstraction can be hidden any complexity. The Boat can call all the methods, but it cannot modify the definition of the Engine itself.

The inversion was achieved via the interface of the Engine. Specific implementations implement the interface’s methods. The boat can be constructed with different engines via dependency injection.

All the plugins, frameworks and other extensions are built on top of this principle, where the extension gives the programmer publicly available methods. The logic inside of it cannot be changed from the programmer's perspective if he does not have access to the extension code directly.

Composition

Dependency injection is the process of passing dependencies for the new class construction. For example this:

val expensiveBoat = Boat(HighTechEngine(), Navigator())
val cheapBoat = Boat(Paddle(), Navigator())

In the code snippet above, you are the witness of composition. Now, this might be misleading. It could be justified that the dependency injection is composition. However, dependency injection is the process of passing dependencies. The composition is an architectural pattern as a whole. The class is composed of different components to achieve higher complexity.

It can be understood as an alternative to inheritance with its advantages and disadvantages.

Inheritance limits the subclasses by its template. It is beneficial for the generalisation of similar behaviour. In our case for the Engine class, all the engines should have similar behaviour. However, the Boat’s behaviour can differ because it contains multiple components with many options.

I would not compare the inheritance and composition, because both are needed to achieve inversion of control and are essential towards other principles.

Single responsibility principle/separation of concerns

Every module/class in the code should be responsible only for one task. The responsibilities should not be randomly thrown around. It will lead to confusion.

The Engine class should contain only logic about handling the engine's responsibilities. Nothing more. The navigator should contain logic about navigating the boat. Nothin more. If you program some task, ask yourself during the implementation, if the programmed code is the class’s responsibility. If not, move it somewhere else. The option is to move it to another class or create a new one. It depends on the situation.

The advantage is that you can usually fully unit test every created separated unit. The final class is much more robust. If some engine needed to be rewritten, it would not impact other classes because it is independent and self-contained.

Wrapping it together

Firstly, the programmer should learn how to make the code work. Later on, the code should start to shape into something more organized and consistent. It depends on the language, industry, and needs.

In general, these principles around dependency injection apply almost everywhere and usually is taken care of by some kind of framework, which takes on all the heavy lifting.

Plan what are you going to do. Scope down the problem by the single responsibility principle. Plan the classes with the help of composition and inheritance. Construct them with help of dependency injection.

You may ask yourself, why to do all these things and here are the reasons:

testability of the code
robustness
clear architecture
easier troubleshooting
easier refactoring

class Navigator{

  fun navigateTo(double lat, double lon) {
    Log.i("Autopilot", "Going to latitude: $lat, longitude: $lon")
  }

}

// generalises all the engines
// single responsibility item - it takes care of all the engine logic
interface Engine {
  fun start()
  fun stop()
}

// subvariant of the engine, which can be used to build a boat
class HightechEngine: Engine {
  
  fun start() {
    Log.i("Hightech Engine", "Starting engine")
  }

  fun stop() {
    Log.i("Hightech Engine", "Turning off engine")
  } 
  
}

// subvariant of the engine, which can be used to build a boat
class Paddle: Engine {
  
  fun start() {
    Log.i("Paddle", "Paddling")
  }

  fun stop() {
    Log.i("Paddle", "Taking pause")
  } 
  
}

// Boat is contructed with different components with composition
class Boat(private val engine: Engine, private val navigator: Navigator) {

  fun goToNewLocation(double lat, double lon){
    // the controlls are inversed
    // the engine takes care of its tasks
    engine.start()
    // the navigator takes care of its tasks
    navigator.navigateTo(lat, lon)
    // the engine takes care of its tasks
    engine.stop()
  }

}

// with help of the dependency injection, we provide needed components to construct boat
val boat = Boat(HighTechEngine(), Navigator())

This is more of a theoretical overview. In the next couple of articles, I will show best practices from Android development with the Hilt dependency injection framework.

Thanks for reading, and stay tuned for more!

Do not forget to clap if you liked the article, and follow me for more.