Popular Movies
A new approach of Popular Movies project from Udacity's Android Developer Nanodegree using Kotlin with Clean Architecture, State Machine design pattern and Test Driven Development.
This project has the same stage 1 features from original project where the data from movies is fetched from TMDB API and the movie posters are displayed in a grid layout and the remaining details are displayed in a separated view. Users can also change the sort order and app keeps fetching API while the main grid is scrolled down.
Architecture Strategies
Please check the post below on Medium with detailed explanation and motivations for this project:
Android Clean Architecture — A “Unicorn” Approach
Highlights:
- Domain uses a State Machine pattern to request data from data layer and contextualize changes in the model.
- Domain also listen to changes in the data or use cases requests and emits updated model
- Domain only exposes use cases related methods to views (that is, there's no state machine related methods being exposed, so views can't control the state machine behavior of domain)
- Presenters listen to model changes and update views accordingly and also controls navigation
- Views are as dumb as it possible with minimum of logic to render data
Technical Strategies
Most part of code uses basic components from framework, so the use of external libraries was reduced to a minimum for better testability, increase of control and possibility to reuse the codebase for different platforms. So there's no use of popular libraries like Jetpack, Retrofit, Dagger, etc.
- App uses Single-Activity architecture
- State machine is initialized in the Application, so it isn't affected by views lifecycle
- For data, was used OkHttp and plain JSONObject to fetch and parse TMDB API responses.
- For image loading and caching was used Picasso
- For threads control was used RxJava only to dispatch tasks to background or main thread (no special operators were used) where a HandlerThread was used to enqueue tasks in the same background thread.
- All dependency injection is manual using a Builder pattern, so data layer and thread dispatcher can be replaced in the tests.
Testing Strategies
- 66 Unit tests - Evaluate each class where dependencies are test doubles (it uses mostly JUnit 4 and Roboletric for Android specific components)
- 9 Component Tests - Evaluate integrated tests inside each view where only data layer and thread handler are test doubles (it uses Roboletric for loading and asserting inside views).
- 5 Integration tests - Evaluate the proper requests to API and data returned (OBS: server is mocked with MockWebServer and parser is a test double).
- 1 End to End test - Evaluate the main use case of loading movies from real API, click in a movie and open its details (it uses Espresso tests).
Copyright
Project developed by Márcio Souza de Oliveira.