The Clean Architecture is the system architecture guideline proposed by Robert C. Martin (Uncle Bob) derived from many architectural guidelines like Hexagonal Architecture, Onion Architecture, etc… over the years.
This is one of the guidelines adhered to by software engineers to build scalable, testable, and maintainable software.
Why do we need to architect?
“The goal of software architecture is to minimize the human resources required to build and maintain the required system.” ― Robert C. Martin, Clean Architecture
Advantages of Proper Architecture
- Easy to Develop
- Easy to Deploy
The Clean Architecture
Here’s the clean architecture illustration created by Robert Martin:
We can see there are four layers in the diagram. Blue layer, Green layer, Red layer, and Yellow layer.
Each circle represents different areas of the software. The outermost layer is the lowest level of the software and as we move in deeper, the level will be higher. In general, as we move in deeper, the layer is less prone to change.
The Dependency Rule
The Dependency Rule states that the source code dependencies can only point inwards.
This means nothing in an inner circle can know anything at all about something in an outer circle. i.e. the inner circle shouldn’t depend on anything in the outer circle. The Black arrows represented in the diagram show the dependency rule.
This is the important rule that makes this architecture work. Also, this is hard to understand. So I’m gonna break this rule at first to let you understand what problems it brings and then explain and let’s see how to keep up with this rule. So please bear with me.
First of all, this circular representation might be confusing for many. So let’s try to represent it vertically.
The colors represented here are the same as the colors represented in the clean architecture diagram.
Remember, the arrow should be read as “depend on”. i.e.
Frameworks and Drivers should depend on
Interface Adapters, which depend on
Application Business Rules which depend on
Enterprise Business Rules.
Nothing in the bottom layer should depend on the top layer.
Frameworks and Drivers
Software areas that reside inside this layer are
- User Interface
- External Interfaces (eg: Native platform API)
- Web (eg: Network Request)
- Devices (eg: Printers and Scanners)
This layer holds
Presenters(UI Logic, States)
Controllers(Interface that holds methods needed by the application which is implemented by Web, Devices or External Interfaces)
Gateways(Interface that holds every CRUD operation performed by the application, implemented by DB)
Application Business Rules
Rules which are not Core-business-rules but essential for this particular application come under this. This layer holds
Use Cases. As the name suggests, it should provide every use case of the application. i.e. it holds each and every functionality provided by the application.
Also, this is the layer that determines which
Gateway to be called for the particular use case. Sometimes we need controllers from different modules.
This is where different modules are coordinated. For instance, we want to apply a discount for the user who purchased for x amount within a month.
Here we need to get the amount the user has spent on this month from the
purchase module and then with the result we need to apply the discount for the user in the
checkout module. Here
applyDiscountUseCase calls the purchase module’s controller for the data and then applies the discount in the checkout module.
Enterprise Business Rules
This is the layer that holds core-business rules or domain-specific business rules. Also, this layer is the least prone to change.
Change in any outer layer doesn’t affect this layer. Since
Business Rules won’t change often, the change in this layer is very rare. This layer holds Entities.
An entity can either be a core data structure necessary for the business rules or an object with methods that hold business logic in it.
For example: calculating
Interest module in the banking application is the core business logic that should be inside this layer.
Let’s look at a simple example to understand this well.
The example demonstrates a simple application that has only one network request.
How can we architect an app that translates the sentence given by the user using a translation API? let’s try to architect.
Each layer does a specific thing. Looks good right? Let’s check the dependency flow for this above architecture to know if anything is wrong.
Remember Dependency Rule? “The Dependency Rule states that the source code dependencies can only point inwards”.
UI → Presenter (✅ Not Violating)
Presenter → Translate Usecase (✅ Not Violating)
Translate Usecase → Translate Controller (❌ Violating)
Translate Controller → Web (❌ Violating)
But it seems correct, right?
UI requests data from
Presenter which requests data from
Use Case which should request data from
Controller which should request data from
After all, how can we expect the
webto throw some data to the
Controllerbeing dependent on it? Also, how can we expect the
Use Caseto get the proper data from the
Controllerwithout depending on it?
But the Dependency Rule strictly says dependencies can only point inwards. It adds up by saying this is the rule that makes the architecture work.
In order to pass this rule, we need to invert the arrow to the opposite direction. Is that possible? Here comes Polymorphism. When we include some Polymorphism here, something magic happens.
Simply by having an
Interface between these 2 layers, we could invert the dependency. This is known as The Dependency Inversion Principle.
Let’s implement the Dependency Inversion Principle in the cases where the Dependency Rule is violated.
Thus the flow becomes:
Let’s check the dependency flow now to know if anything violates it.
Now we can see that no inner layer depends on any outer layer. Rather, the outer layer depends on the inner layer.
So why should the outer layer depend on the inner layer but not the other way around?
Imagine you’re in a hotel. We want the hotel to serve us what we want, but not what they offer right?. The same thing is happening here, we want the DB to give the data the application needs but not the data it has.
Application orders what data it wants and it doesn’t care how DB or API prepares the data. This way, the application doesn’t depend on DB or API. If we need/want to change the DB or API Schema in the future, we can simply change it. As far as it gives what the application asks for, the application doesn’t even know the change in DB or API.
Also, the single-way dependency rule saves the application from the deadlock state. i.e. imagine in a 2 layer architecture, the first layer depends on the second layer, and the second layer depends on the first layer. In such a case, If we need to change anything in the first layer, it breaks the second layer. If we need to change anything in the second layer, it breaks the first layer. This can be rejected by following the deadlock state.
This is the clean architecture described by Uncle Bob.
We are yet to see how to move the data across the boundaries and how to handle errors. We’ll do so in future articles.
Thanks for reading.