alexshubin/fallingwords

Babbel Code Challenge

FallingWords

There're 2 scenes in this game:

• Game (where the game happens)
• Results (provides feedback to a user about the results)

Because both screens share the same state (GameScene should count results, and ResultsScene should show that results) I've decided that unidirectional architecture with a shared app state should fit nicely, I've picked RxFeedback and introduced the game state which consists of few vars and could be rendered at any time (AppState).

* how much time was invested

About 7 hours, maybe a little bit less, also I could have saved more time if I had done layout in storyboards. But I don't like storyboards (harder to maintain, codereview, merge, etc.)

* how was the time distributed (concept, model layer, view(s), game mechanics)

I've started from the bottom (services -> views):

• about 1,5 hour was spent on services + models
• about 0,5 hour was spent on some mvp game logic (like start game, show next word etc.)
• about 2 hours were spent on UI layer: layout, localized strings, some boilerplate like coordinator, etc.
• about 1 hour on results tracking + view controller
• about 1 hour - timer + label animation
• 0,5 hour - get back from results screen
• 0,5 pods, project setup etc...

* decisions made to solve certain aspects of the game

The concept seemed quite easy to me - basically we need two things:

1st - dependency to provide data for a game session with a reasonable probability of right translations and some shuffling (RoundsProvider Rounds - I mean 1 word = 1 one round)

2d - something to manage the game logic and easy to test - Reducer is a pure function(easy to test), it mutates the state, which could be rendered any time by anyone who is subscribed (e.g. UIViewController). The system works like that:

• User emits events into reducer by tapping buttons
• Also side effects emit events into reducer (e.g. timer)
• StateConverter translates app state into primitive types(String) which can be rendered by UIKit.

So every step up to UIKit is fully tested 🤓.

Another interesting things:

• RoundsProvider always provides 50% of the right answers and the the other part is random - right/wrong.
• Animation rendering clears previous animations and fires the new one on every render invocation with duration provided by the state.

* decisions made because of restricted time

• WordsLoader should be asynchronous, so DiskLoader implemetation may be easily replaced with for example 'ApiLoader' implemetation. But it would take some additional work on UI (activity indicators, failures, etc...).
• There shouldn't be a singleton. The timer should be a separate dependency, which should be tested with TestScheduler.
• Some edge cases could be handled somehow (like words provider provided not enough words to compose a session, etc...)

* what would be the first thing to improve or add if there had been more time

• UI and design :), add menu
• fix decisions made because of restricted time
• load words from API
• add some multilingual support (there's a lot: swap languages, improve backend to send different languages on demand, etc...)
• maybe results saving

Don't forget to pod install 😉