11 KiB
Modularization learning journey
In this learning journey you will learn about modularization, and the modularization strategy used to create the modules in the Now in Android app.
Overview
Modularization is the practice of breaking the concept of a monolithic, one-module codebase into loosely coupled, self contained modules.
Benefits of modularization
This offers many benefits, including:
Scalability - In a tightly coupled codebase, a single change can trigger a cascade of alterations. A properly modularized project will embrace the separation of concerns principle. This in turn empowers the contributors with more autonomy while also enforcing architectural patterns.
Enabling work in parallel - Modularization helps decrease version control conflicts and enables more efficient work in parallel for developers in larger teams.
Ownership - A module can have a dedicated owner who is responsible for maintaining the code and tests, fixing bugs, and reviewing changes.
Encapsulation - Isolated code is easier to read, understand, test and maintain.
Reduced build time - Leveraging Gradle’s parallel and incremental build can reduce build times.
Dynamic delivery - Modularization is a requirement for Play Feature Delivery which allows certain features of your app to be delivered conditionally or downloaded on demand.
Reusability - Proper modularization enables opportunities for code sharing and building multiple apps, across different platforms, from the same foundation.
Modularization pitfalls
However, modularization is a pattern that can be misused, and there are some gotchas to be aware of when modularizing an app:
Too many modules - each module has an overhead that comes in the form of increased complexity of
the build configuration. This can cause Gradle sync times to increase, and incurs an ongoing
maintenance cost. In addition, adding more modules increases the complexity of the project’s Gradle
setup, when compared to a single monolithic module. This can be mitigated by making use of
convention plugins, to extract reusable and composable build configuration into type-safe Kotlin
code. In the Now in Android app, these convention plugins can be found in
the build-logic
folder.
Not enough modules - conversely if your modules are few, large and tightly coupled, you end up with yet another monolith. This means you lose some benefits of modularization. If your module is bloated and has no single, well defined purpose, you should consider splitting it.
Too complex - there is no silver bullet here. In fact it doesn’t always make sense to modularize your project. A dominating factor is the size and relative complexity of the codebase. If your project is not expected to grow beyond a certain threshold, the scalability and build time gains won’t apply.
Modularization strategy
It’s important to note that there is no single modularization strategy that fits all projects. However, there are general guidelines that can be followed to ensure you maximize its benefits and minimize its downsides.
A barebone module is simply a directory with a Gradle build script inside. Usually though, a module will consist of one or more source sets and possibly a collection of resources or assets. Modules can be built and tested independently. Due to Gradle's flexibility there are few constraints as to how you can organize your project. In general, you should strive for low coupling and high cohesion.
-
Low coupling - Modules should be as independent as possible from one another, so that changes to one module have zero or minimal impact on other modules. They should not possess knowledge of the inner workings of other modules.
-
High cohesion - A module should comprise a collection of code that acts as a system. It should have clearly defined responsibilities and stay within boundaries of certain domain knowledge. For example, the
core:network
module in Now in Android is responsible for making network requests, handling responses from a remote data source, and supplying data to other modules.
Types of modules in Now in Android
Top tip: A module graph (shown above) can be useful during modularization planning for visualizing dependencies between modules.
The Now in Android app contains the following types of modules:
-
The
app
module - contains app level and scaffolding classes that bind the rest of the codebase, such asMainActivity
,NiaApp
and app-level controlled navigation. A good example of this is the navigation setup throughNiaNavHost
and the bottom navigation bar setup throughTopLevelDestination
. Theapp
module depends on allfeature
modules and requiredcore
modules. -
feature:
modules - feature specific modules which are scoped to handle a single responsibility in the app. These modules can be reused by any app, including test or other flavoured apps, when needed, while still keeping it separated and isolated. If a class is needed only by onefeature
module, it should remain within that module. If not, it should be extracted into an appropriatecore
module. Afeature
module should have no dependencies on other feature modules. They only depend on thecore
modules that they require. -
core:
modules - common library modules containing auxiliary code and specific dependencies that need to be shared between other modules in the app. These modules can depend on other core modules, but they shouldn't depend on feature nor app modules. -
Miscellaneous modules - such as
sync
,benchmark
andtest
modules, as well asapp-nia-catalog
- a catalog app for displaying our design system quickly.
Modules
Using the above modularization strategy, the Now in Android app has the following modules:
Name | Responsibilities | Key classes and good examples |
app
|
Brings everything together required for the app to function correctly. This includes UI scaffolding and navigation. | NiaApp, MainActivity App-level controlled navigation via NiaNavHost, NiaAppState, TopLevelDestination
|
feature:1, feature:2 ... |
Functionality associated with a specific feature or user journey. Typically contains UI components and ViewModels which read data from other modules. Examples include:
|
TopicScreen TopicViewModel
|
core:data
|
Fetching app data from multiple sources, shared by different features. | TopicsRepository |
core:designsystem
|
Design system, tokens and icons. | NiaTheme NiaIcons
|
core:ui
|
UI components, composables and resources, used by different features. | NewsResourceCardExpanded
|
core:common
|
Common classes shared between modules. | NiaDispatchers Result
|
core:network
|
Making network requests and handling responses from a remote data source. | RetrofitNiaNetworkApi
|
core:testing
|
Testing dependencies, repositories and util classes. | NiaTestRunner TestDispatcherRule
|
core:datastore
|
Storing persistent data using DataStore. | NiaPreferences UserPreferencesSerializer
|
core:database
|
Local database storage using Room. | NiaDatabase DatabaseMigrations Dao classes
|
core:model
|
Model classes used throughout the app. | Topic Episode NewsResource
|
Modularization in Now in Android
Our modularization approach was defined taking into account the “Now in Android” project roadmap, upcoming work and new features. Additionally, our aim this time around was to find the right balance between overmodularizing a relatively small app and using this opportunity to showcase a modularization pattern fit for a much larger codebase, closer to real world apps in production environments.
This approach was discussed with the Android community, and evolved taking their feedback into account. With modularization however, there isn’t one right answer that makes all others wrong. Ultimately, there are many ways and approaches to modularizing an app and rarely does one approach fit all purposes, codebases and team preferences. This is why planning beforehand and taking into account all goals, problems you’re trying to solve, future work and predicting potential stepping stones are all crucial steps for defining the best fit structure under your own, unique circumstances. Developers can benefit from a brainstorming session to draw out a graph of modules and dependencies to visualize and plan this better.
Our approach is such an example - we don’t expect it to be an unchangeable structure applicable to all cases, and in fact, it could evolve and change in the future. It’s a general guideline we found to be the best fit for our project and offer it as one example you can further modify, expand and build on top of. One way of doing this would be to increase the granularity of the codebase even more. Granularity is the extent to which your codebase is composed of modules. If your data layer is small, it’s fine to keep it in a single module. But once the number of repositories and data sources starts to grow, it might be worth considering splitting them into separate modules.
We are also always open to your constructive feedback - learning from the community and exchanging ideas is one of the key elements to improving our guidance.