Clean Software Architecture

October 20, 2021

👋 Hello there!!

Why does clean code matter?

Writing a software program that works does not require much knowledge or skill. On the other hand, getting software right is hard. When software is written clean, you no longer need many programmers to keep it working or have a massive list of bugs whenever the software adds new features. Changes are quick and straightforward, and bugs are few. This may seem farfetched, but I’ve experienced this most recently.

Harmful code is way too common in our industry, but it doesn’t have to be. Some companies believe that throwing more resources into poorly coded software will solve the problems. Still, statistics show, it might eventually get better, but if the cleanup of the code doesn’t happen, productivity stales or, even worse, it just halts.

So why does this happen?

I believe that the culprit for this is having the following belief:

“We can clean it up later; we just have to get to market first!”

Of course, we all know that the cleanup never happens, and developers get pushed to get the next feature done.

Instead, we should embrace these beliefs:

“Slow and steady wins the race.”

“The only way to go fast, is to go well”

How can we avoid this?

So far, we’ve talked in the blog about the importance of Test Driven Development, but today we will talk about Clean Architecture.

Most people believe that Architecture is about high level, but this is far from the truth. Software Architecture deals with an immense number of low-level details. The low-level details and high-level structure are part of the whole system in a software design.

To avoid the common mishaps in poorly written code, we must follow this diagram.

Our outermost layer contains our input-output devices. The innermost layer contains our application business rules. In between, we have our use case layer, which purpose is to allow input and output layers to interact with our business rules.

All the dependencies point towards our business rules to ensure our code is clean and our architecture is flexible. This means that our business modules do not depend on lower-level modules. For example, we wouldn’t want our tic-tac-toe game rules to depend on the UI code or the server rendering code. A good rule of thumb to recognize what a high-level module is vs. a lower level module is to see how close to the input/output it is, the closer to the inputs/outputs, the lower level it is.

Remember that higher-level modules should never depend on lower-level modules. Therefore, arrows always point towards the center.

Another rule to keep our clean architecture is to keep the flow of control in the opposite direction.

What does this mean?

It means that our entities/business rules should control the flow of the application. But how can our high-level modules control the lower-level modules without depending on them?

The answer to this is using the Dependency Inversion Principle.

Dependency Inversion

A. HIGH-LEVEL MODULES SHOULD NOT DEPEND UPON LOW-LEVEL MODULES. BOTH SHOULD DEPEND UPON ABSTRACTIONS.

B. ABSTRACTIONS SHOULD NOT DEPEND UPON DETAILS. DETAILS SHOULD DEPEND UPON ABSTRACTIONS.

The way we execute this in code is to use of Polymorphic Interfaces.

Most programming languages provide a way to create some abstractions. We can create those abstractions in our higher-level modules and implement them on lower-level modules like our controllers. This way, our higher-level modules don’t depend on the details but instead on this interface’s abstraction. We achieve clean architecture once our high-level modules use abstract interfaces to control the flow of the application outwards while also not depending on lower-level modules.

This principle alone will drastically improve the quality of your code and its ability to change with minimum negative impacts.

Tomorrow we’ll look into some parts of our tic-tac-toe code that violate this principle and how we fix this.

Cheers! ❤️

Written by Edgardo Carreras.

• ← Persisting Tic-Tac-Toe State on Server Restarts
• Persistence Refactoring - Polymorphic Interface →