Tactical-RPG-Test

General Explanations

here is a simple video that shows how game is running: https://youtu.be/MPgrugWhruU

How Features Implemented

Game Scene Mangement

• there is only one GamePlayLogic scene that all logic of offline match is imlemented in there
• theme scenes are added additively to the logic scene to have different theme maps
• right now we have Desert and Jungle maps

GameFlow Management

• there is GameStateManager that handles the state of the game. there is a GameState class that holdes state of the match all needed data to run the match is in there.
• any input of player is just a query to GameState and UI and Match Logic are seperated.

Game Data Management

• State of the whole game is wrapped in GameState model. things like what settings has player selected in the game
• player progress like experience and level is in PlayerData class and deserializes to the class from file to be used
• above classes are serializable and IFileUtility as different implemetations like encrypted version for devices nd row version for the editor and development

Game Rules And GameState

• there is a class type named ActionQuery these are the actions sent to the IUtilityMatchQueries and functions of the class decide that the input is valid or not and apply it to the MatchState

Opponent And Ai

• Opponent is a high level class that can implement AIOpponent or OnlineOpponent. ApplyAction function can have different implementations. for current we only have ai

best approaches for making AI For turn based games are

• Minimax: as state space is to big, recursive minimax simply cause overflow. also making itrative version is challenging and we cant add enemy class or difficulty level.

• Data Driven: time consuming for a test project to define possible rules but for AI with series of action works great.

• Goal Oriented Action Planning. also great but needs some effort to make a planner integrated with such space.

• Utility Theorem also great that on different conditions gives score to the actions and selects action with highest score. its time consuming to select best multipliers.

• Chain Of Responsiblity:(used in this project and is fast to implement) its a design pattern rather than being a technique for Ai. list of actions are defined and ai agent selects the first action that suits the condition. there are different implementations for this design pattern but its a very simple version. we can make different classes of opponents by just changing the order of these condition-actions

• there are 4 classes of Ai including : defensive, rational(tested), unitDestroyer, Finisher

Conventions

• there are many rules for naming and formating the code but i used rider/resharper default conventions
• there is good use of oop and classes are short
• used fail-fast approach and tried dont placing any exception-handler or null checking to be able to solve problem fast.
• seperated assemblies to seperate team or programmer codes form other plugins
• basically functions that return null are bad but i used many times. but its not bad by fail-fast approach.

Design Patterns

Dependency Injection

• best for testablity(writing unit tests)
• loosely coupled and changability and flexiblity(can simply change implementations of an interface)
• cleaner than singletons
• better use of OOP and polymophism

MVVM

• very good for ui based apps and parts of the code that use ui
• it seperates binding of UI in other classes called #View-Model# and seperates UI logic from bindings so its great for testablity
• most mvvm plugins for unity are over complicated so made a simple framework to use mvvm in this project

chain of responsibility

• explained in AI part

Plugins Used In Project

Zenject(Extenject)

a dependency injection container for unity that supports unity contexts.

UniRx

a plugin to use reactive extensions and use reactive programming in unity. its a great tool for making sequential events and making web requests and coroutines cleaner.

Other:

• environment and tanks 3d models assets
• addressables to manage memory and asset loading

Things Could be better

• including a localization plugin(is its not a complete simulation its ignored).can simply write a component to read value of current language reference to a key(guid) and update it or implement an interface and implement different languages(not good idea, causes discrepancy)
• make shared functionality in some function in seperate functions to have more DRY code.
• could use pooling but as object get instantiate at the start and there are not too many, it was not needed.
• tanks and towers are not injected so they are not accessible to DI framework.
• could not find good audio effects but by just placing good audio files, they should work as there are audio sources on objects and they get called by events
• some long functions on matchQueryUtility but they are understandable.
• better use domain based namespaces instead of category based namespaces.