Introduction

In EFP 2, a plan to review TerraModulus Relictus branch has been made. This EFP describes about the report of the entire review process. In the following, the report will be presented section by section according to the previously made plan.

Codebase Commentary

The source code was originally reviewed with an order but later reorganized by their respective topics and themes. The review has been conducted as of commit 2ba2b09bf94110efbdc36cd95ce8d74a20835205. In the following, various content will be discussed following their headers. Please note that the organization of content in this section has been aided by Artificial Intelligence (AI).

This section presents commentary on the Minicraft+ codebase, reviewed in preparation for the new TerraModulus project. Observations include evaluations of the existing implementation as well as suggestions aligned with upcoming development directions. Where applicable, proposed changes for TerraModulus are distinguished from recommendations for refining the existing codebase.

Application Initialization and Entrypoint

The application framework would include a dedicated Main.kt file, designed separately for both client and server implementations. This file contains the entry point main function, which is responsible for initializing the application, processing command-line arguments, and launching the primary handlers essential for execution.

Core Structure and System Architecture

Within the core package, several critical components define the foundation of the application's architecture.

The Game component should be instantiated rather than implemented solely through static members or as a singleton. Its designated class name would be TerraModulus, with the NAME field defined as a top-level constant. Meanwhile, the VERSION field would be managed through the application profile, ensuring loading during early runtime and later accessibility via a getter method.

Initializer class would be incorporated within Main.kt file, facilitating the initialization of all core threads required for application functionality, such as rendering and ticking threads.

The io package would no longer reside within core, as its scope is overly generic. At a minimum, several parts would be grouped into audio package and the Localization class would be separated from it to ensure modular organization. Furthermore, all display elements would be registered directly to the menu, allowing a simple reload operation to refresh all elements without reliance on external hooks.

Ticking operations would be centrally managed within the TerraModulus class. Specifically, menu ticking would be executed there, necessitating that MenuManager be placed outside RenderSystem, although rendering would still take it as a parameter. World ticking would be handled by its respective manager, while audio processing would operate in a separate thread under the control of the audio system.

To enhance clarity, Display would be renamed to Screen. Instead of retaining the original Screen class, a Canvas may be introduced to manage display elements more effectively. Screen transitions would continue to follow a Last-In-First-Out (LIFO) linked list structure; however, improvements could be made to the handling of screen initialization and exit procedures. Rather than relying on an initializing function, a constructor reference could be utilized, particularly since the parent instance parameter is primarily intended for rendering. Passing a constructor function and managing parent screens within MenuManager would offer a more streamlined approach.

Window Management and Handling

Similarly, the Initializer class should avoid the use of static members. Window management would be handled through a dedicated window handler, which could also oversee error popups when necessary. FPS (Frames Per Second) and TPS (Ticks Per Second) counters would be regulated within their respective handler classes, ensuring proper thread management.

Additionally, the application window would be launched at an early stage, following the initialization of essential basic elements. This approach eliminates the necessity of a splash screen without a visible window. The window would always be centered upon launch to ensure a consistent user experience.

Argument parsing would be handled within the Main.kt file, reinforcing a structured approach to processing runtime configurations. The run function would reside within the TerraModulus class. Moreover, the getExceptionTrace function could potentially be repurposed for detailed error reporting, contributing to improved debugging and system reliability.

Launching Arguments and Options

The launching arguments would be systematically structured, with all options named using kebab case, each prefixed with double hyphens. Hardware acceleration would be enabled by default; however, a specific launch option to disable it would also be available. The fullscreen parameter would function as a toggle, determining whether the application starts in fullscreen mode. Regardless of the fullscreen setting, the window size configuration would always be applied. If no specific window dimensions were provided, default values would be used to ensure a consistent display.

Internal Assets

Certain assets that are not intended for modularization or customization may be designated as internal components. Unless explicitly required, these resources, including logos, would remain at the root level rather than being organized into separate directories.

Input Handling and Management Architecture

The ClipboardHandler would be managed within the broader InputHandler, as all input interactions, including keyboard and mouse operations, would be handled via GLFW (Graphics Library Framework). Consequently, there would be no necessity to implement separate interfaces for key input processing. The menu structure would also undergo modifications to ensure that key inputs are detected only by the relevant menu elements, following a prioritized approach.

The InputHandler oversees several functionalities, including input processing, key mapping, controller states, and key state handling. However, controller management would be separated into a dedicated ControllerManager. The capability of the originally used gamepad library is relatively limited, necessitating enhancements in controller support. The abstract Key class and its subclasses provide essential functionality, and repeated key states would no longer be required. Consequently, PhysicalKey could be simplified, and ControllerKey would be deemed unnecessary. Additionally, logical operations such as "or" and "and" could be incorporated within the Key class. String-based expressions would be deprecated, and the PhysicalKey#stayDown function would likely be disregarded. Moreover, if key state processing results in a single Key state, subclasses such as CompoundedKey and ORKey would no longer be required. Key mapping would employ a distinct class rather than relying on the Key class.

To improve organization and efficiency, input handling would be structured into four distinct classes: KeyInputHandler, TextInputHandler, GamepadInputHandler, and MouseInputHandler. Among these, TextInputHandler would also manage clipboard operations. Text input queues would be processed only upon request and reset at the start of each tick cycle. A regex filter could be implemented for validation, and backspace key operations would follow the same structured processing method. The releaseAll method would be omitted, while key hints or hotkeys, despite being complex, would likely follow a similar implementation method. Modifier keys would not be handled separately within the handler class. Controller vibration feedback would be managed by ControllerManager or another dedicated class, but controller detection and management would remain within ControllerManager.

Key Input Event Processing

Key input events triggered by GLFW would be received and queued within the main ticking thread. Text inputs would be restricted exclusively to text input fields, while keyboard inputs would remain universally active. Each keystroke transitions through three states: "just pressed," "down," and "just released," represented as boolean values. A press event sets "just pressed" and "down" to true, while a release event sets "down" to false and "just released" to true. At the start of each input tick, both "just pressed" and "just released" states would be reset to false. Subsequently, queued input events would be processed, updating key states accordingly, followed by window ticking operations that handle input. The ticking frequency might be set to 1.5 times the maximum FPS. If necessary, an additional ticking thread operating at a constant frequency could be introduced.

For events such as controller port interactions, listener callbacks would be used instead of polling events each tick. These events occur infrequently and would be best handled asynchronously to prevent unnecessary processing overhead. Furthermore, GLFW-generated controller events inherently include the target controller information, eliminating the need for additional search operations. GamepadInputHandler would likely be activated only when controllers are enabled.

Input Mapping and Key Assignments

Input mapping would be managed separately from the input handler, allowing configurations based on user preferences while keeping certain mappings hardcoded. This includes the F3 debug key and advanced modifier functionalities involving Shift, Ctrl, and Alt modifiers. In-game inputs and controllers would follow similar processing principles and be handled within the same class structure. Mapped hotkeys would remain inactive when input fields are engaged, allowing standard printed key inputs while ensuring hardcoded keys such as escape and enter receive the appropriate handling, especially in one-line field or with an Input Method Editor (IME).

Key mapping would employ predefined sets with the keys of constants rather than arbitrary string-based assignments, ensuring clarity in configuration and eliminating the need for loops. Mapping would strictly serve logical purposes, while a separate display mapping would exist for User Interface (UI) representation. Since the mapping configurations would only be used internally in the program, the necessity to use string representation would be ineffective. The mapping would be hardcoded, so a loop would not be used. The final input states resulting from triggered events would be logically combined using "or" operations to ensure all inputs are properly accounted for. However, no sticky key would be realized for this system.

Keyboard-Based Movement for Tile Selection

To optimize keyboard interactions for movement and tile selection, two potential input schemes may be considered. One approach involves using movement keys exclusively, with the selected tile automatically aligning with the tile positioned in front of the player. Alternatively, a dual-set input method could be implemented, where one set of keys (e.g., WASD) controls movement, while another set (e.g., arrow keys) allows independent tile selection by directional input.

Controller Management and On-Screen Keyboard Integration

Controller management would be enhanced through GLFW, allowing users to select their preferred controller. This configuration option would be adjustable via settings but would not be stored persistently on disk. Additionally, the On-Screen Keyboard (OSK) for controllers would remain available, with notable improvements to its underlying functionality.

The OnScreenKeyboardMenu would likely retain its existing name but would include a constructor parameter accepting an input field object. A menu handle for the screen might also be passed, although the exact implementation remains uncertain. Alternatively, the OSK could function as an "invisible" menu or component, meaning it would be excluded from ticking and rendering processes unless actively required.

Given the potential variation in key sizes, keys should be mapped through reference linking as components rather than predefined layouts. If a key reference is null, it would indicate that no accessible keys exist in that direction. Navigation would still be managed via the D-Pad, ensuring seamless controller-based interaction. The sticky key's state would be stored within the key object itself, which would also be linked to its final property. A grid layout may be employed to organize key positioning efficiently.

All key states including whether a key is pressed or selected, would be maintained within the key object. The OSK would operate independently of traditional keyboard input and would automatically hide itself whenever no controller is detected during a tick cycle. Key event triggers from the OSK would be directed to the associated input field for processing.

To ensure flexibility, an abstract class would be introduced, eliminating the need for \0 as a placeholder for sticky keys. This structure would also allow for specialized keys that perform actions beyond text entry, such as the switch key to the layout of numeric and symbolic keys. Positioning would not rely on x and y coordinates since they are unnecessary for this implementation. Exception should not be abused, but managed directly within the input field instance and corresponding menu components. Visibility checks within the ticking function would be unnecessary, and rendering could be simplified by incorporating sprites into components.

Key Mapping Configuration and Restriction

During key mapping configuration, upon the first key release event, all keys pressed concurrently would be recorded as a key combination. However, specific keys, including escape, menu, and Windows keys, would be restricted from inclusion due to contextual compatibility requirements. The escape key would also function as a cancellation trigger during key detection. Notably, the escape key would remain hardcoded for menu control operations.

Hotkey Configurations and Key Sets

Dropping an entire item stack would be executed using the sneaking modifier alongside the drop hotkey. If the configured drop key sequence includes the sneak key, it would be considered a conflict. Each mapping would define a set of permitted keys within a predefined constant.

Permitted key sets:

Typing Keys: Letters, numbers, and punctuation characters on the main keypad, excluding modifiers
Control Keys: Ctrl, Alt, Shift, Tab, Enter, Esc, Enter, Backspace, Pause/Break, Insert, Delete
Function Keys: F1 to F12
Navigation Keys: Home, End, Page Up, Page Down
Numpad Keys: All numpad keys except Num Lock
Forbidden Keys: System-wide reserved keys, including lock keys, Win, Context, Print Screen

Although Tab could be categorized as a navigation key, it has been excluded from that set to avoid redundancy. Unlisted keys may be disregarded. While checking conflicts, hardcoded keys would be taken into account to ensure compatibility.

Debugging Mechanisms and Interface Adjustments

Debugging functionality would be refined to ensure consistency and ease of access. Hotkeys related to F3, primarily used for toggling the debug screen, would remain hardcoded rather than being configurable. To maintain code consistency, an invisible menu session would be introduced to process both standalone F3 inputs and key combinations involving F3. The session would activate upon pressing F3 and deactivate upon sending a command or releasing the key. It would exclusively handle key inputs, triggering commands when applicable. For most keys, a direct command would be dispatched to a target, such as toggling debugging functions. Otherwise, upon release, the system would verify whether an advanced input session had been engaged. If no such session was detected, the debug screen toggle would be executed.

Additionally, the DebugPanelDisplay, functioning as a Graphical User Interface (GUI) for commands may be deprecated if command support is fully integrated. Maintaining this panel would introduce unnecessary complexity, and using in-game chat for commands could provide a more efficient approach. Many entries currently depend on this menu, and transitioning to command-based interactions would simplify the interface.

Likewise, the ListItemSelectDisplay could be restructured as a menu dialog. However, since it is exclusively utilized by the debug panel, its implementation would be unnecessary. Consequently, associated entries within screen.entry.commands would also be removed.

Logging System and Configuration

All application events would be logged irrespective of severity levels, eliminating the need for separate logging, debugging, or tracing launch options. Debugging functionality would be universally available, with selective toggling permitted in specific areas. Given the ease of creating user-generated content, debugging would be equally accessible. By structuring each class with a single responsibility, logging targets would be clearly identifiable via fully qualified class names, eliminating the necessity for arbitrary behavior-specific "title names". Log4j 2, SLF4J, and kotlin-logging would serve as the primary logging utilities.

The logging system would transition to a different library for improved idiomatic usage. This change would introduce extensive logging configuration capabilities supported by both the library and its public interface. In adherence to Kotlin’s idiomatic principles, loggers would not be manually instantiated in external classes but would be defined based on descriptive class names. This approach ensures syntactic consistency and eliminates the need for custom target names. Since the console would typically be hidden from users, log output would remain as comprehensive as file-appended logs, without filtering. Additionally, logging configurations would be exclusively managed via programmed code, negating the need for customizable configurations. Notably, the Rust components of the application would not incorporate any logging mechanisms.

Localization and Language Settings

The debug locale would no longer be necessary, as untranslated strings would already be logged, ensuring careful programming without requiring translation key displays on-screen. Localization files would be stored within assets, with detailed implementation depending on the new asset system. Instead of a dedicated debug locale, a hotkey-based feature could highlight untranslated UI components dynamically. Since all entries would already be loaded into memory, switching languages would not require additional loading processes.

Language settings would be structured as an inner class within the main settings system, aligning with preferences. With the debug locale removed, sorting would be further simplified. GUI improvements would eliminate the need for entry highlighting; instead, selected options could be bordered, bolded, or tinted in a distinguishable color. Upon exiting the menu, the chosen entry would be applied immediately, negating the need for a "save" button. The translation disclaimer could remain, while an option to force Unicode font would not be needed.

Settings Organization and UI Integration

The Settings system would transition to Preferences, ensuring that setting values remain independent of UI components. UI elements should be designed to reference and link to these settings rather than being treated as direct entries. World generation configurations would exist solely within the world generation options rather than being global entries. Additionally, all settings would be structured as type-specified property entries

The settings screen would be unified as SettingsScreen rather than maintaining separate OptionsMainMenuDisplay and OptionsWorldDisplay. There would be no functional distinction between the settings accessible from the main menu and those available during gameplay. Each option would store both its original and current values, allowing for comparison during saving and providing a notification if changes have been made. Further settings menus would be instantiated only upon access, reducing unnecessary memory usage. The layout structure would primarily follow two models: Submenus or Tabs navigation, depending on stylistic choices. Changes would only be applied upon exiting the settings menu. Additionally, submenu-related buttons would end with an ellipsis when following a submenu-based layout.

Applying changes only when menus are closed would be a more effective approach for managing settings. As multiple submenus may exist within the settings interface, UI elements linked to settings values should update exclusively upon menu closure. This approach would prevent frequent and potentially disruptive effects when users modify settings multiple times within a session.

To establish clear distinctions, the following definitions apply:

Settings: The overarching system within an application that governs configurations and operational adjustments, controlling various features and functionalities
Options: Specific choices within a context, often presented as a list or menu, allowing users to customize their experience or performing specific actions
Preferences: User-defined selections that determine application behavior, stored persistently
Configurations: The structural setup of an application’s parameters and options determining behaviors and structures of application or system, specifying how components interact.

Within this context, "Settings" encapsulates the entire configurable system, encompassing various "Options", while "Preferences" store individual user selections on disk. "Configurations" would not be used in this context.

Error Handling and Management

To ensure robust error management, the default uncaught exception handler should be initialized as early as possible. This approach guarantees that any uncaught exceptions are properly handled and users receive relevant notifications. Instead of referencing a separate function, a lambda expression should be used, allowing for greater flexibility in error processing. The error message content should be distinct from other caught exceptions to facilitate clear differentiation.

Errors should trigger informational pop-ups only under specific conditions: if they occur before rendering initialization is complete, or if they fall within or outside the scope of the application. For error catching mechanisms, only documented errors should be considered, unless they are managed within the primary exception handler or a designated error processing block.

Screen System and UI Management

The RenderingLimitingModel is an ineffective implementation, as scaling the UI and managing menus more efficiently would eliminate the need for excessive stretching and scrolling of entries. Components such as ScreenEntry and EntryRenderingUnit function more as workarounds rather than solutions, making them visually and structurally inadequate. A more refined approach using OpenGL would be pursued.

The screen package, excluding its subclasses, originally contained only Display, Menu, Toast, TutorialDisplayHandler, and RelPos. The term "display" by definition in dictionary refers to a monitor display, whereas "screens" typically denote sets of visual components presented on the window canvas.

The title screen would be enhanced with a background, with selectable entries featuring highlights rather than arrow-based indicators. It would include a logo, menu entries, a version string and customizable title texts via asset packs instead of hardcoded conditions. "Splashes" may be renamed to "title texts" as they may no longer visually splash. A help message at the bottom would be unnecessary when mouse support is enabled by default. The version string would be expanded with additional details. Converting UI elements into distinct components, similar to Java Abstract Window Toolkit (AWT) or Swing, would eliminate redundant position recalculations each frame, improving code quality and scope management. The initfunction in TitleDisplay should not handle elements related to screen appearance. The "Play" menu entry would be replaced with separate "Singleplayer" and "Multiplayer" options, followed by "Settings". Achievements would not be included at this stage. The compact "Help" menu, with only four entries, would function as a dialog menu for simplicity. Any layout that completely replaces or overlays existing canvas components would be categorized as a screen with content as menus and components, while entry-based structures would be grouped within a menu. The title screen entries would include "Help" and "Quit". The PlayDisplayhas already been deprecated, with its two entries merged into theTitleDisplay as aforementioned.

Many screen menus would restrict the Esc key from returning to the previous screen. When screens permit navigation back to prior screens, a dedicated return button would be provided to prevent users from unintentionally losing progress.

Only the top-most screen would be interactive. If an upper screen fully covers lower screens, those lower screens would not be rendered. The gameplay canvas would also function as a Screen to allow for unified management via ScreenManager. When the gameplay screen is instantiated, it would become the base screen, with all other screens removed. Similarly, initiating a world-loading screen would clear all other screens.

Screen Structure and Component-Based Menu Redesign

Display would be redefined as an abstract class and renamed Screen. However, the parent reference would not be managed within this class but by the screen manager. Additional functionality would be introduced to enhance menu-related features. clearScreen would be removed, and canExit could be replaced by an overridable function or a callable screen handle for the screen. The class structure would resemble a panel rather than a conventional Display. Menu selection would primarily support keyboard controls, but tab-based navigation would likely suffice. Arrow-key-based selection would align with layout structures, such as AWT-style grid or border layouts, rather than relying on manual position adjustments. onSelectionChange may not be necessary, except in cases where it facilitates access to hidden menus through key selection or mouse hovering when only selectable components are involved. The rendering mechanism would be revised, removing dependencies on selection index for rendering depth.

Each menu would have a distinct content structure. Instead of ListEntry, component-based implementations would be adapted, making removeSelectedEntry obsolete. The system would adopt a model similar to AWT or Swing. The stack-handling approach in InventoryMenu requires further refinement.

As ItemListMenu transitions to slot-based design, display length constraints would no longer be necessary. Likewise, cutting and aligning entries manually would become obsolete, allowing the removal of the positioning parameter. The menu construction process would prioritize flexibility through component-based design, including padding adjustments.

Menu would evolve into a generalized component container rather than an entry-based structure. It would not contain entries by default, and container-like components managing entries (such as ListEntry) would be handled separately. A foundational abstract menu would exclude properties such as size, position, title, spacing, frame, selectability, rendering, and searching, maintaining minimal structural dependencies. The menu would encapsulate a container, passed as a protected final parameter via constructor delegation. Components would be managed independently of menus, which can exclusively handle components. Screen would follow a similar concept but would extend its functionality to directly accommodate both components and menus. The builder and copy constructor patterns currently in use are inefficient and would be reconsidered.

The searching functionality would be independent of Menu, incorporated only when a container requires it. This mechanism would be powered by an input field and contextual filtering logic rather than fragile highlighting and excessive scrolling adjustments. Searching would be performed only upon Enter activation and would deactivate upon Escape. Component rendering would be should on relative positioning, eliminating the need for additional translation functions. Visibility states may be introduced, and some containers may support listeners. Scrolling animations introduce unnecessary complexity and would be removed, along with dynamic entry updates, builders, and MenuListEntry.

Input Masking and Interactivity Management

An input masking session would be introduced for each tick to regulate input state checking in menus. This session would be managed by MenuManager, allowing specific menu interactions to suppress or fully override queued inputs dynamically. Such modifications would occur on demand without affecting physical key detection outside the menu ticking session. While its necessity and practicality need further evaluation, this system could improve interaction consistency across menus and screens.

Components would not reference a menu or screen handle when ticking, while menus should always receive the screen handle upon screen ticking. Handles and functionality would be managed within screens and menus, even for locally defined components. Some menus may require fixed frames, though not all would. Rendering functions should be consistently present across screens, menus, and components.

Inventory System and Menu Organization

An abstract InventoryScreen class would be introduced, serving as the foundation for all menu screens incorporating player inventory functionality. This class would encompass player inventory menus, crafting interfaces, and container menus, but not book-related screens. Every inventory menu would include an item counter, distinct from the item list menu, which would remain designated for the creative mode item list. Additionally, the OSK menu should be controlled by an input field rather than parent menus, enabling flexible dynamic menu additions with prioritized handling.

For screens such as ContainerDisplay with dual-sided menus, directional controls should be used without any cyclic traversal across menus. Vertical menus, however, may require different handling due to their inherent connectivity, potentially using Tab for navigation. Notably, slot selection and stack transfers between menus would be available exclusively when the mouse is disabled for ketboard controls, preventing conflicts with mouse-driven interactions, similar to tile selection via keyboard controls. The getOtherIdx function lacks relevance under this system, and determining control behavior based on cursor position introduces unnecessary complexity. A standardized control mechanism would be implemented, with a refined naming convention replacing "heaviness". Additionally, color fading should be implemented via the alpha channel.

Crafting and Recipe System

The CraftingDisplay would be integrated into PlayerInvDisplay, which would subsequently be renamed PlayerInventoryScreen. Depending on the implementation approach, further specialization may be introduced via subclasses such as SurvivalPlayerInventoryScreen and CreativePlayerInventoryScreen, ensuring logical consistency with the overall code style. While the recipe system remains undecided, relevant crafting menus and recipe unlocking mechanics would be addressed in conjunction with its development. The achievement acquisition method would be revised, as the current implementation is excessively hardcoded. Additionally, inventory change listeners could be introduced if necessary to refine the refresh functionality.

A new recipe system would be introduced, recipe menus and crafting menus would be greatly changed, since all the crafting mechanisms are too similar. Most critically, crafting menus would integrate player storage visibility to enhance inventory interaction.

Inventory Structure and Management

The InventoryMenu would accommodate a holder, though not necessarily limited to an entity. Inventory contents would be linked to the holder’s inventory, ensuring bidirectional synchronization of updates. Slot alignment would not be enforced, allowing menu slots to correspond directly with inventory slots without rigid structuring. The creative inventory would not be managed within this system, as previously discussed. Inventory menu slots would register listeners at either the inventory or menu level, depending on their nature: temporary slots, such as those in crafting interfaces, would be handled within the menu, whereas persistent slots (i.e., inventory slots) would be registered within the inventory. Since inventory modifications can occur outside direct menu interactions, listener registration at the inventory level, potentially via the holder, would be necessary.

Item stack operations would be managed by the item list menu, with a well-defined structure supporting both non-inventory item collections and inventory-based stacks. A wrapper class may be introduced for streamlined handling, though further implementation details remain to be determined. The refresh function may be unnecessary for menu frame updates, while the description menu would become a tooltip or a separate menu aside.

Inventory Screen and Slot Interaction

Some aspects of PlayerInvDisplay align with previously discussed elements. The creative mode item list would be accessible through an independent menu or a tab. The selected slot would be bordered or highlighted, with item stacks appearing at the cursor while hovering over inventory menus, leaving the source slot empty, or remaining in the source slot as a reference for interaction. No help text should be shown, eliminating the method used to hide the item list menu.

The RepairBenchDisplay would retain its inventory menu structure, though storage slots would be preferable to temporary ones, replacing RepairBenchCarrier. While this approach introduces a new paradigm within the codebase, SlotEntry would transition from an entry-based system to a direct slot representation. Menus managing stack transfers currently follow repetitive processes, which could be simplified through an optimized transfer model. Each slot component would maintain an association with its backing inventory slot instance. Disabled buttons and empty slots would remain selectable but non-interactive. Storage spaces would feature attributes or properties defining their slot types, such as crafting input slots, product slots, and other specialized configurations. Fishing rod durability tracking could be simplified to simplify tool mechanisms. The functionalities of slot entries would predominantly depend on their associated backing slots. RepairInfo would be regarded as a context generated by the input items, while the carrier component primarily delineates operational scope within the display class. The current method of using leading spaces for positioning within strings is problematic but common within the codebase; replacing slot labels with intuitive graphical elements would improve clarity. The onAction function, currently used only once within a specific scope, could be refactored into a lambda. The redundant "checking again" mechanism should be converted into assertions, as its execution is intentional. Stack movement would primarily target destination slots, though additional interactions such as mouse scrolling or L1-R1 slot switching would also be supported, behaving similarly to addToCarrier operations. If slots are storage-based, stacks would persist on exit without being automatically dropped. The code structure of RepairBench is actually neat, and there could be slots stored in the entity.

Inventory and Item Management

The inventory system, while well-structured, should be moved to a distinct package separate from item. If inventories incorporate slots, they would use mapped slot structures instead of generic lists, potentially featuring distinct slot sets per storage unit. To support stacking, item metadata excluding quantity must remain consistent, ensuring compatibility with a slot-count-driven inventory model. If slots originate from underlying storage mechanisms, primary inventory capacities would be derived from these foundational quantity limits.

Item picking functionalities should be generalized for all mobs using inventory slots.

Item stacks would be managed through an inventory provider responsible for constructing inventories and item managers. Item configurations would align with provider specifications, including stacking criteria and limitations. Consequently, stack size would not be hardcoded within individual items or stacks. While items themselves would not be cloned, stacks should retain this capability. Item depletion should be validated upon each interaction, invoking the inventory manager listener upon stack modifications. Interaction methods should return true upon successful interactions, potentially reflecting state changes accordingly.

The RewardChestInventory could be generalized to a class next to the UnlimitedInventory.

Recipe System and Resource Handling

The recipe system should be relocated into a separate package. Recipe instantiation should originally be handled through constructors or factory methods, eliminating the need for string parsing during initialization. Direct instance referencing should be used instead of strings or names, ensuring efficiency and reliability. Recipes must be unlocked prior to usage, maintaining integration with the advancement system.

The main recipe processing method should be replaced with script-based execution. However, due to the inherent complexity of scripts, their implementation must align with standardized execution environments rather than directly mirroring the core code structure. Recipes should be treated as resources rather than being hardcoded within the system.

Miscellaneous Message Displaying Screens

The InfoDisplay would be removed, with statistics integrated into the pause menu. Each player would maintain their own set of statistics, as a part of the server or world data, but not as a unit of server or world, although it is still possible to have server or world statistics. World-related information may not be presented here but could be displayed elsewhere as needed.

The MessageDisplay class serves only two purposes: displaying link error messages and resource pack descriptions. The former should be replaced with standard application notifications, while the latter could be merged into the main menu layout or incorporated via tooltips. As a result, this class would become obsolete and should be deprecated.

The TempDisplay component offers little functionality. When opening an Internet address in a browser, a confirmation popup should always be displayed, but there is no need to automatically close the screen indicating the browser action. Instead, only the confirmation popup should close immediately.

Popup Management and Option Dialogs

Several functionalities currently handled by PopupDisplay could be streamlined through tooltips or full-canvas options menus. Submenus with limited components may be implemented as simplified panes, akin to anonymous classes, possibly replacing this class. Popups containing only informational messages could transition into notifications, while those featuring basic button options could be refactored into a dedicated class, such as OptionMenu. This structure would consist of a display message alongside defined options with names and fallback settings.

Confirmation popups for quitting may not be necessary. Options dialogs that do not exit or transition screens could be full-canvas, while those requiring menu changes could incorporate a frame. Full-canvas screens would feature a lightly dimmed and blurred background or a dedicated background image, complementing the components. When listening for key sequences in the settings menu, a non-full-canvas banner would display active key inputs, ensuring clarity before any key is released. For instance, the confirmation dialog menu for resetting key bindings would include a frame, with its style determined later.

Buttons should not rely on hotkeys, and this structure should be removed entirely, even for Keyboard-Only controls, with a simplified set of selectable options. Additionally, screen resolution may be leveraged for better horizontal arrangement of options, improving usability. Selection entries or button would provide a more intuitive alternative to direct key-listening mechanisms. The options dialog menu would not support Esc for exiting; users must choose an available option instead. The OSK would not require specific handling within this system.

Entity System and Interaction Model

The player entity is categorized as a living entity, distinct from mobs due to its unique control implementation. The primary distinction lies in the control system, allowing entity behaviors to be generalized without requiring excessive subclassing or repetitive implementations. Since player take action would be governed by specific abilities, such functionality should not be integrated into general entities. A revised classification of entity types will be proposed at a later stage.

Entity interactions follow structured event-based processing:

Actors initiate actions on targets, which receive those actions as events
Targets represent entities affected by an actor's interaction
Controllers regulate entity behavior based on triggered events, with multiple controllers potentially managing a single entity
Events dictate method execution within the code
Global states refer to the collective conditions across the entire world

All actions are event-driven, meaning that interactions occur only when events are emitted and processed. This includes regular ticking events and discrete updates. Unlike spontaneous occurrences of attacking or using, entities operate through receiver interfaces responsible for damage and usage handling. When a mob or entity initiates an attack or use event against a target, this event is registered within the main event, forming a hierarchical structure of linked actions. Passive consequences stemming from primary events leverage their provided parameters alongside commonly observed global states.

Attack events initiated by controllers should be structured efficiently. They must focus solely on responding to the primary event with predefined arguments and contextual global states. Attack initialization should remain minimal, calculating only the foundational aspects of an attack, while relevant parameters should be assigned by the controller. The main entity structure should not directly manage event responses. Attack actions may instead be managed through independent controllers, with the corresponding attack response solely processing the event rather than managing indirect state changes by the event. Thus, only the damage reception interface should reside within an entity's main structure.

Similarly, use interactions require clarification. Events related to "use" actions may involve tiles or entities as subjects, while "used on" events primarily concern items as subjects. The subject may shift based on the specific event type, but by default, the target would be the subject. Until a refined solution resolves this ambiguity, usage actions should not be split into separate categories. Using the middle mouse button may provide an alternative interaction mechanism, with further details subject to future discussion.

Mob Intelligence and Damage Handling

Movement capabilities, including swimming and traversal across surfaces, would be dictated by AI-driven mov controls, ensuring proper contextual application. Additionally, the "on fire" visual effect would be treated as a transient metadata attribute, with entities retaining the ability to reject such properties upon assignment. Fire and burn damage would be processed through the damage source, acting as an argument within the damage-handling system. For the controls, there might be a method that even mods could add by the interface with the default values; this may require Typed Anchor Dynamic Mapping, which will be proposed later.

Damage evaluation would be managed by controls, eliminating redundant checks such as isAttackable. Since entities block interactions regardless of their nature, preliminary validation steps should be omitted, allowing actions to proceed without additional verification layers. Similar simplifications also apply to isInvulnerableTo and isUsable, reducing inconsistencies and streamlining API usage.

Interactive box sizing would be adjusted with greater precision, rather than relying on approximate sprite dimensions. The current implementation using tickTime would be deprecated. Temporary timers for event-driven effects, such as burning damage, would be pooled within dedicated tick-based handlers per entity, minimizing unnecessary class-level bloat while retaining separation of responsibilities. Death detection would reside within health-related systems, while speed modifications would be managed via move controls, particularly when factoring in speed-altering effects.

The MobAi subclass would be dismantled, with individual components such as despawning mechanisms, intelligent behaviors, automation logic, rendering, and loot processing transferred to specialized controls and handlers.

For EnemyMob, behavioral intelligence would be dictated by control mechanisms. The tracking system requires refinement to enhance targeting efficiency. The attack range should extend beyond direct physical contact, ensuring broader interaction capacity. All attack-related methods should be consolidated into a single function, accepting a damage source parameter for streamlined execution.

Player-Specific Considerations

Several exposed fields in Player would be concealed behind dedicated handlers, including delay counters and GUI-related elements, which would not reside within a general mob class. Potion effects would be managed via specialized control structures. Tile condition validation for event handling should be processed within tile logic rather than mob systems unless explicitly tied to mob behaviors.

Target selection for interactions should prioritize proximity-based identification. If no valid target is within range, no interaction should be initiated. Cursor-based selection should allow for enhanced precision. If an entity obstructs selection within the active range, the tile beneath should not be selectable, even the entity gives no interaction.

Damage System and Tile Interaction

The DamageSource class should incorporate an argument inferred from DamageType. The argument would support multiple data structures applicable across different scenarios. DamageType would transition into a standard class rather than an enumerated type to facilitate extensibility and improve interface accessibility.

Tile interactions, including effects triggered upon stepping or collision, would be processed through predefined standard methods within tile classes. If interactions require continuous updates, associated tile entities would be returned. Factory methods such as onTileSet should be contained within relevant initialization scopes where tiles are set and defined. Additional configurations would be determined by tile properties.

Fluid tiles would be categorized separately, incorporating dedicated fluid mechanisms. Tile connection-based texture rendering should be exclusively handled within texture definitions rather than within tile logic. Tile destruction callbacks, such as those triggered by mining or explosions, would follow a standardized method structure, accepting destruction sources or types as parameters.

Finally, getData methods in tile implementations would be eliminated, as relevant data would be processed through chunk or tile managers.

Furniture and Tile System

The movement mechanisms for Furniture would be reconsidered in light of the decimal coordinate system. Lightweight objects may remain entities, whereas heavier objects, such as Furnace and Oven, would transition into tile-based structures. Additionally, item sprites should not be stored within entity instances, ensuring separation of visual assets from entity logic. The use of reflection for copying remains fragile and should be avoided.

The inclusion of a die function within furniture structures is conceptually awkward and should be restructured.

Furniture item implementation should align with furniture mechanisms. If all furniture can be held, players might be able to pick them up using bare hands or a specialized item. However, selectively allowing only certain furniture items to be held would introduce inconsistencies. If furniture items associate data via properties, a corresponding resource key must be registered explicitly, rather than through dynamic registration. Furthermore, furniture without item representations should remain unregistered for its item form unless explicitly required. As a result, not all furniture may be freely accessible to players.

The PowerGloveItem currently serves as a placeholder for furniture pickup. Its removal or revision should be determined based on future inventory system developments.

Tile Management and Placement Behavior

The TileItem class would persist, maintaining its function as the intermediary for direct tile placement. However, the approach depends on broader decisions regarding tile mechanics. Tiles could transition into items reserved for creative mode, facilitating expedited placements. TileModel would act as a default data structure rather than preserving original tile data, as stacked tiles do not require retained tile data. If tiles lack item-based representations, an item type containing a single resource key with metadata encapsulating tile attributes would be introduced. The primary determinant for placement eligibility would be the tile type.

Tile Classification and Logical Consistency

If TorchTile remains an item that can be picked up, classifying it as furniture may be more appropriate. Furniture is inherently designed for mobility, allowing players to relocate lightweight objects easily, whereas tiles are fixed structural components. This misalignment in mechanisms creates inconsistencies in both logical behavior and expected interactions. Additionally, small decorative tiles such as wall-mounted torches would be designed explicitly for placement on wall tiles, reinforcing logical coherence within the system.

Inventory Drop Handling

The mechanisms governing item drops upon player death remain undecided, with various implementation approaches offering distinct properties. Ultimately, the chosen method would align with the game's thematic direction. If the design philosophy prioritizes realism, the system may resemble Rust, where a fallen player's body retains inventory items. In such a case, the DeathChest concept would become unnecessary.

Loot System

Inventory randomization should be managed within generation-handling systems rather than being embedded in Chest. Container inventories should remain publicly accessible as standardized public interfaces. Since dungeon mechanisms are no longer relevant, previous implementations can be disregarded despite potential improvements.

Tile loot drops may be regulated via tile properties. Damage multipliers could be controlled through game rules, ensuring they affect players in appropriate scenarios. The concept of automatic damage revival lacks logical coherence, particularly for tiles without health attributes, as they are primarily influenced by interactions such as mining.

Resource tiles like OreTile and other rock formations would yield loot upon each mining action or complete destruction. Their visual design would avoid exaggerated spiky appearances, adhering to a more natural aesthetic.

World System and Dimension Management

The World class should be instantiated independently rather than residing within core. Additionally, level management should be delegated to world instances rather than being handled by unrelated classes. Several generator configurations, including dimensions, would be defined within assets rather than hardcoded into the program. World settings such as the seed and world name would be encapsulated within a dedicated configuration object.

All world-related randomization processes should be specific to instances and potentially replaceable by the backend engine as needed. Certain fields, such as the level change listener, last world entry and exit timestamps, lack practical utility. Dimension transitions should be governed by actual implementations rather than relying on an unused listener. Moreover, tracking entry and exit times through direct comparisons are unnecessary when initialization and termination can be managed via instance creation and destruction. The approach of resetting sprite animations within these processes is flawed; sprite animations should instead be handled separately. Maintaining instances remains more effective than invoking resetGame. Additionally, the logic involving Player#respawn and Level#add within the same block introduces unnecessary complexity; a more semantic and logically structured code approach should be adopted. While WorldSettings aligns conceptually with this logic, the broader implementation still requires improvement.

Under the new entity system, an entity cannot simply be removed from a dimension; it must be eliminated entirely. However, entities may transition to a different dimension by modifying coordinates, ensuring no null entries. The getClosestPlayer function should be maintained as a general-purpose helper method, accessible across various processes including commands. Other relevant methods may also be retained.

The execution of world events, activities, and commands must be carefully evaluated to avoid bypassing critical processes. Improper handling could lead to unintended execution omissions. This includes aspects such as teleportation across dimensions; all necessary unprocessed events should be considered.

Difference in Tile System

There would be several differences due to the new tile system to be proposed.

FenceTile serves merely as an overlay for ground tiles, but the simplified tile system resulted in verbose code that should be optimized.

Since HoleTile lacks functional significance, it would be removed. Instead, an "air tile" designation would be introduced, also applicable to InfiniteFallTile.

Both DirtTile and SandTile would require foundational support. In the absence of underlying tiles or in cases where only air tiles are present, these tiles would collapse. Additionally, footsteps on such tiles would generate extra dust particle effects.

Since the conventional depth-based level system would be retired in favor of independent dimensions with no direct geographical continuity, StairsTile would be deprecated. However, stair-like tiles could still be incorporated as mobility aids for mobs. These tiles could be composed of various materials and may exist as fragments or incomplete sections.

Chunk-Based Dimension System

The original fixed-size level structure would evolve into an unlimited world consisting of chunk-based dimensions. Given that world generation would be no longer constrained by predefined size limitations, structure generation would operate independently of world size constraints. Each dimension would retain a display name and receive a seed derived from the overarching world seed but would not possess specific size parameters. Even if the generator produces a finite world size, additional chunks should remain accessible, albeit empty.

During dimension construction, essential metadata including mob count limits, would be defined. Entity container management would be confined within dimensions and chunks, overseeing entity counts, additions, removals, and transfers. Server-side entity queues for additions and removals would use linked lists within the entity manager. Chunk-level entity managers may be synchronized with dimension-level managers, ensuring consistency across modifications. Entities might be organized within a quadtree structure to facilitate efficient chunk management.

As part of the proposed world system rework, depth indices for dimensions would be eliminated. The notion of "levels" would be discarded in favor of a pure dimension-based structure. The concept of "level" does not currently exist within the revised framework.

Entity Indexing and Searching Mechanisms

Entities would be stored across two distinct mappings. Each chunk would manage its entities using either a quadtree or k-d tree structure. Dimensions would oversee all loaded chunks, featuring an indexing mechanism for querying entities within chunks. The world would handle all loaded dimensions, implementing an indexing system for querying entities across dimensions while maintaining a global map that associates entity classes with collections of entities.

For nearest-entity searches within a specified range, only selected chunks would be considered. Nearest-neighbor search algorithms would be employed, or entities would be filtered by type using a mapped collection before looping through results. Additionally, when selecting all entities within a given range, including those positioned atop a tile, temporary geometries in the physics engine may be used to identify collided objects rather than exclusively relying on centers of entities.

Game Time Management and Ticking System

The in-game ticks per second (TPS) would be adjusted to 20 instead of 60. Additionally, the length of a full-day cycle would be set to 20 minutes rather than 18, ensuring a reasonable duration while maintaining alignment with 60-minute mechanism. To preserve performance consistency, the game speed should remain constant and may not be subject to pausing. Several fields would be introduced for tracking game time values:

dayTime: Represents the total number of ticks (time) elapsed during a single game day
dayCount: Tracks the number of days that have passed within the game world
gameTime: Records the cumulative ticks (time) elapsed throughout gameplay

A backup routine may be implemented to replace auto-saving, requiring a temporary suspension of game ticking for data synchronization. However, this would not constitute an actual pause in gameplay.

Time enums representing different phases of the day may not follow uniform quarter divisions, particularly since "0" would correspond to either midnight or midday. Seasonal cycles and variable day lengths may be introduced in future updates, requiring definitive constants to define distinct time phases. Minecraft's approach could serve as a reference, though its daytime ticking structure does not strictly conform to real-world time conventions for "0". Further implementation details would be discussed and finalized later.

Entity Ticking and UI Separation

Player entity ticking should be entirely independent of UI ticking, as entities exist within the world system, whereas UI elements function solely on the client side. Current implementations may handle menu interactions within entity ticking functions, which is impractical. Instead, such operations should be delegated to the MenuManager or a similar management class, ensuring proper class isolation and separation of responsibilities.

Sleeping Mechanisms

With the introduction of the new sleeping mechanisms, tracking sleeping player counts would no longer be necessary. Additionally, Bed instances would not store color enums; item, entity, and tile display names managed via translation keys and registry keys rather than defined within constructors. Removing the player entity from the level upon sleeping is an illogical design choice and should be revised accordingly.

Future sleeping mechanisms would be reconsidered to maintain consistent game speed and ensure continuous events are not disrupted by time-skipping or acceleration. However, extended nighttime durations may be problematic under certain circumstances. To mitigate this, an alternate dream world dimension could be introduced during sleep, functioning as a single-player environment. If a player sustains significant damage or dies within this dimension, they would instantly wake up in bed. Correspondingly, a biological health system would be developed to integrate this mechanism.

Players would be able to sleep at any time, provided their character reaches a fatigue threshold. This approach allows gameplay to continue seamlessly while preserving in-game consistency. Detailed implementation specifics would be determined later. Additionally, this method avoids performance inefficiencies that could arise from artificially accelerating game ticks.

Ticking Order and Parallel Processing

The sequence in which tick events are executed within a world is crucial. Since tick events can affect entity positions and even their existence, entity existence-related events should be processed at the end of a ticking cycle. Furthermore, players include both local and remote participants, meaning tick operations must rely on signal transmissions while maintaining synchronized entity management. All entities, including their controls and existence states, should be processed simultaneously to ensure consistency.

Chunk managers would refresh entity positions according to updated coordinates, following structured stages for optimized management. If parallel processing of entities is introduced, tick-dependent events should be segregated and queued instead of being executed immediately, allowing for accurate state synchronization.

Client and Server System Considerations

Users would be assigned unique client IDs, though detailed implementation specifics are yet to be determined.

The server architecture requires comprehensive planning, encompassing serialization, routing, and network communication. Safeguards must be implemented to mitigate non-synchronization issues and data loss. For consistency and compatibility, there should be no fundamental distinction in implementation between single-player and multiplayer, particularly within the server-client framework. Transmission data may be identified using numeric IDs, though the protocol would not support forward or backward compatibility.

Server Analytics and Data Collection

Server-side analytics could track active clients based on Launcher activity and periodic pings, similar to Minicraft+. However, this process should not be classified as general telemetry, which collects device information and usage reports. Data collection must occur only with user consent, and users should retain the ability to enable or disable tracking via preferences after the initial launch (as identified by local save data).

Both data collection methods are entirely optional, with accompanying user agreements and terms. Server-side counting would be managed via database records, whereas pings would remain anonymous. Descriptive telemetry and automatic crash reporting, however, may not be anonymous, as they could include device details or account identifiers. Implementation specifics are yet to be finalized, but they represent an important consideration.

Multiplayer Management and Account System

The MultiplayerDisplay offered little meaningful reference. However, login authentication should be managed by the Launcher, with client ID, user ID, authorization token, and session ID transmitted via launch arguments. Additional data may be retrieved by the application based on these parameters.

Since skins would be managed by the platform’s account system, they should not be processed within the application. Consequently, there would be no in-game preview or customization screen for skins. Instead, skin data would be transmitted only from the server to the client, where local comparison would occur. The watcher thread for skin tracking would no longer be necessary. The Mob Sprite handling process would differ significantly under this structure.

Client IDs would rely on Universally Unique IDs (UUIDs) with specialized modifications. On the server side, all client IDs would be recorded and associated with account IDs. Account IDs, unique UUIDs maintained by the server database, would serve as server player entity IDs (EIDs). These EIDs would be distinct within the server or world, appended with identifiers differentiating players from other entities.

Client IDs would remain unique, permitting multiple Launcher instances across various client devices. However, two clients cannot connect to the same server using identical client IDs, if attempted, a new client ID would be generated. Additionally, multiple clients cannot access the same server under a single account.

In single-player mode, offline gameplay would be supported via a default hardcoded player EID, mapped to account IDs to facilitate guest modes without requiring an account. Although guest mode simplifies implementation, maintaining a system where account-based player EIDs are universally used ensures that server worlds remain playable offline. These safeguards assumed ideal data integrity, with unauthorized modifications ignored.

While EIDs are not strictly network-related, they were originally categorized under Network. Instead, they should be reorganized into utility functions or a dedicated classification.

Client synchronization decisions would be dictated by the server, meaning any invalid player actions could be ignored or reversed at the server’s discretion.

Pause Menu and Player Death Screen

The pause menu structure would introduce SingleplayerPauseScreen and MultiplayerPauseScreen, with respective versions allocated based on game mode. Single-player sessions and LAN hosts would use the single-player version, whereas multiplayer participants would engage the multiplayer variant. The menu title would be "Game Menu", featuring:

Universal options: "Back to Game", "Settings", "Statistics" and "Advancements"
Singleplayer exclusive options: "Open to LAN", "World Options" and "Save and Quit"
Multiplayer exclusive options: "Disconnect", with additional functionalities extendable via plugins or mods

A potential "Save" or "Backup" option for single-player remains undecided. Implementing these features via commands would be simpler, though both actions require suspending gameplay for synchronized data storage. Game saving and backups should occur only when all server-side ticking threads are suspended, rendering them internal server operations rather than in-game commands.

Dedicated servers have predefined server commands, whereas single-player mode lacks such functionality. The scope of in-game commands is subject to future discussion. The most possible implementation method would be within the pause menu, pending further review.

The background of the pause menu would display a dimmed gameplay screen, which would remain interactive in terms of rendering but not input, clicking the dimmed screen would not close the menu.

The PlayerDeathDisplay would transition into PlayerDeathScreen, visually similar to the pause menu but with the gameplay screen frozen at the last frame before death. The header "You Died!" would accompany the death message. Upon death, the player entity ceases to exist in the world, yet the player remains present, albeit in an unspawned state, preventing any interaction with world events.

At this stage, the player may either respawn or quit, with the world continuing to function normally. If instant respawn is enabled, the death screen would be bypassed, and the player would respawn immediately. The quit button functions identically to its counterpart in the pause menu. No additional menus or information would be accessible during this screen.

Save System and World Loading Management

A SaveManager or equivalent class would assume the upper-level functionalities previously managed by FileHandler. Basic file helper functions, such as folder deletion and copying, are already available in Kotlin through Path.deleteRecursively and File.copyRecursively, eliminating the need for additional implementations. Future asset management should not rely on filesystem-based listing, as an advanced asset system would be expected to replace this approach.

World loading would be optimized to provide a structured and visually informative process. For instance, loading sessions should incorporate progress bars for both asset loading and world loading. Since resource loading logic operates outside menus, the interface operations should reside within a dedicated loading menu rather than the actual loading operations being embedded in the interface elements. The loading screen should be instantiated alongside the loading session, dynamically updating and refreshing progress as resources are processed. Additionally, logging world names is redundant when proper class structure organization provides sufficient clarity. Resetting the world before initialization is unnecessary when instances are properly instantiated; once loaded, reset values should be disregarded.

The exceptions encountered during world loading are generally justified, as rare failures require immediate termination or cancellation of the process. While existing code structures may be somewhat disorganized, the fundamental approach remains practical.

The save and load system should operate as independent components rather than existing as direct Save and Load classes. Each aspect, including timing, system handling, and individual responsibilities, should be distinctly isolated. Preference management should remain autonomous, with no direct association with world save data.

If a loader fails to locate the intended object, an explicit error should be generated, signaling the absence of the requested resource. As a result, unknown objects should not be loaded. The UnknownItem class should be deprecated for consistency. Additionally, strict mode should always enforce immediate loading termination upon encountering errors.

Level Transition and Loading Screen

The LevelTransitionDisplay should function purely as a UI component, delegating the actual dimension transition process to a dedicated manager. Since precise progress tracking may not always be possible, the loading screen should display a standard background alongside status messages. Moreover, dimensions might not be preloaded in advance; at a minimum, the target chunk must be fully generated before transition completion.

The LoadingDisplay would be renamed asWorldLoadingScreen, preventing users from bypassing it via the escape key. Given that world loading occurs chunk by chunk, progress tracking would be segmented accordingly, reflecting the individual status of each chunk. Corrupted chunks and regions should be regenerated, with automatic backups created before deletion from the loaded pool. Missing spawn-area chunks should be dynamically generated and monitored by WorldCreatingScreen, which would also handle new world creation via WorldCreateScreen. The existing LoadingDisplay implementation was somewhat disorderly. Loading operations should be managed by a dedicated manager class, ensuring exception handling is properly delegated to menu-driven dialogs rather than the loading screen itself.

Automatic data fixing would occur within structured loading schemas, assuming previously reported and resolved issues. Instead of menu-based message popups, application notifications would relay relevant alerts. Progress would be handled specially.

Compatibility with Legacy Save Formats

Both LegacyLoad and HistoricLoad were introduced to maintain compatibility with older world save formats. Initially, LegacyLoad facilitated handling of old save data, whereas HistoricLoad was subsequently introduced to distinguish between the period when version tracking was not originally considered and the time when versioning became included among old versions. Significant structural shifts following the legacy phase warranted this separation.

Additionally, HistoricLoad introduced a stricter conditional checking methodology in the codebase, moving away from previous implementations that relied predominantly on the simple if-else statements and switch-case logics. This change was driven by the reliance on text-based comma-separated values (CSV) format, which required extensive string parsing, while the following approach of using JavaScript Object Notation (JSON) also conforms to this. By using this approach, potential errors could be minimized while maintaining essential support for older saves.

Schema-Based Save System

Future iterations of the save system would adhere to a schema-driven framework, ensuring robust backward compatibility. This approach would enable progressive fixes for outdated saves, applying predefined schemas that gradually improve structural integrity. These data fixers should be static and immutable, implemented as singletons within the codebase. Exception management would incorporate contextual detail, offering different behaviors based on operational mode:

Compatible mode permits special accommodations for older formats and corrupted data.
Strict mode enforces immediate failure upon encountering inconsistencies.

There is a potential possibility for recovery mode for corrupted data, but such functionality may be provided by the Launcher instead of game. File parsing and format handling would reside within low-level components, particularly when processing custom serialization formats. Where applicable, Java Virtual Machine (JVM) objects would be instantiated in Rust parts via parsed Rust objects, though preference settings may remain exempt from this approach. Serialization functionalities may also be integrated systematically.

Save operations should be independently structured rather than housed within a unified Save class. Serialization would be modular, tailored to different data objects such as entities, ensuring all dynamic tags are properly processed during save operations. A dedicated WorldSave class would be introduced to oversee world-specific save data, facilitating logical and structural serialization processes. Preference settings could use general serialization, given their inherent key-value structure. The Rust component would oversee encoding of serialized objects into save files.

It is important to distinguish save versioning from game or public interface versions. Instead of aligning with standard game updates, world saves would be tracked through an independent version number, incremented sequentially for identification.

World Naming and Filesystem

World names should remain unrestricted, requiring an alternate approach for directory naming. This would involve generating unique identifiers using a randomized alphanumeric character table indexed by an incrementing counter. Both the table and counter would be maintained as internal data within the client application. Under normal operation, world directories should remain unmodified, as an internal lookup table stores directory names for preference-related local user data, such as the last played world. Since unexpected name collisions are possible, warnings would be issued when duplicates arise, prompting the system to skip conflicting identifiers. The identifier format would be base-36 and may be randomized. Server-related data would not be stored on the client side but transmitted on demand, with only primary metadata and configurations retained locally. Given this approach, there would be no need for regular expression-based name validation.

Players should primarily manage worlds through the application rather than a file manager, using advanced editors where applicable. The application should facilitate world import, export, and backup creation, features that might be supported within the launcher itself. Duplicate world names would be permitted, though warnings should inform users when identical names are detected. The application should preserve user-defined names rather than modifying them for uniqueness.

World Creation and Listing

The WorldCreateDisplay would be renamed and accessed through the world list screen, which displays all locally stored worlds. This interface would solely process user inputs and configurations without saving changes until the world creation process begins. If any fields remain empty or set to "default", predefined values would be assigned automatically. The seed input field would accept any characters and hash them through a custom hashing function, even when numeric values are entered. An option may be provided to input a true numeric seed separately near the input field. Fields containing invalid characters or unrecognized values should be flagged as "unparsable", disabling the completion button. Users would not be shown the actual directory name assigned to the world.

Settings fields would be initialized within this screen but not retained post-world creation. They should not persist as global states within places like the Settings class. Instead, the WorldSettings class should act as a dedicated record class storing all values, isolated from the screen logic.

Within the world list screen, users would have access to operations such as importing, exporting, copying, deleting, and creating worlds, with the middle three restricted to existing worlds. The WorldInfo class would resemble a lightweight record class, along with world data size information. Users would be able to customize world icons, although the default icon remains uniform across all entries. The displayed world version would reflect the last accessed game version, rather than the actual data format version.

Directory names would not be visible but should be accessible via a button on the selected entry, with a folder icon labelled "Open Folder", avoiding disabled button being shown. Users must select a world and click the play button to launch it, though an optional double-click quick access action could be enabled in preferences. Given a more refined and expansive UI, world information would be embedded within each entry, rather than appearing in a separate menu. If stored preferences reference a nonexistent world key, those entries should be designated as unknown.

While the world list screen remains open, a watcher thread would monitor filesystem and operating system (OS) notifications for updates. Any operations performed within the screen menu should transmit selected world data as arguments, avoiding reliance on static fields. Redundant code in the ticking method should be minimized through refactoring and splitting.

World Loading and Operations

World loading procedures would undergo significant restructuring. The existing methods setWorldName and hasLoadedWorld should be removed, as their implementations are conceptually flawed. getValidWorldName would no longer be necessary.

Copying a world may function as a straightforward duplication, with "Copy of" prefixed to the original world name. Users would be required to manually edit the name post-duplication or enter a custom name via a dialog, similar to the IntelliJ rename prompt.

World Boundaries and Tiles

The ConnectTile class should be eliminated. Given the theoretically infinite world size, visible or discoverable boundaries should not exist under normal gameplay conditions. Players should be constrained within soft world boundaries, preventing standard access beyond defined limits.

Certain rock tiles may contain varying concentrations of ores and minerals.

Chunk-Based World Generation

Only the chunk generator within a dimension would use the dimension seed or world seed during initialization. Other randomized events should rely on thread-specific random number generators. Generation processes by covered generators created during dimension initialization should begin only after explicit generation commands, ensuring controlled execution rather than immediate world filling. Arbitrary placement of structures at specific locations should be discouraged unless dictated by a larger generative pattern.

World generation would be governed by a manager, which maintains registries of generation parameters, functions, and configurations. A random number generator with preset seed values would drive the generation process. The generator should remain active throughout world loading for continuous chunk generation. Generation configurations should be modular, stored separately to allow for public customization rather than being hardcoded. The interface should provide customization options for generator configurations, including sampling functions and randomization algorithms.

If the world size is theoretically unlimited, validation mechanisms should be unnecessary, assuming robust generator design. Testing functions may pose challenges due to fragmented content generation, complicating validation methodologies.

Advanced World Generation Techniques

If world generation incorporates high variability, conventional techniques such as switch-case structures, loops, and if-else conditions may prove insufficient. Even when fully programmatic approaches are used, constant values should be effectively managed at the field or top level. Additionally, cave structures would not be exclusively planar.

Dungeon generation exemplifies non-standard dimension creation, dictated by large-scale patterns rather than general algorithms. Boss rooms and dungeon gate structures are generated toward the center, requiring specialized logic. To accommodate such designs, scripting may be necessary, instead of conventional data structures like simple lists of attributes and properties formatted in JSON.

Early world generation including terrains would be handled by the Rust part, with chunk data subsequently transmitted to Kotlin for secondary population processes.

Features are small-scale decorative elements, typically tile placements.
Structures consist of large-scale templates with complex logic, including preconfigured tile and entity placements.
Jigsaws serve as presets for templates used in structures.

Most generation elements may leverage custom data formats, while generator configurations, features, and structure definitions may incorporate scripting support. This framework should ensure that procedural generation remains extensive, flexible, customizable, modular, and standardized.

World Generation Testing

Map generation testing functions should be excluded from production code, as they introduce redundancy within released builds and remain inaccessible without enabling hidden launch parameters. A separate module like demo would allow manual parameter input for visual validation of configurations and algorithms.

Entity Management and Behaviors

When an entity is introduced into the world, a unique identifier would be assigned through its constructor, ensuring uniqueness and immutability within the world instance. Entity removal would be governed by the world’s ticking system per update, eliminating the need for a dedicated boolean field tracking entity existence. Since an entity’s presence cannot be guaranteed beyond the current tick, references to target entities may be transient, avoiding persistence beyond a single tick cycle. If there exists any component storing entity references, such behavior should be supported by the controls managed by the target entities, ensuring appropriate destruction; thus, omnidirectional incoming reference may not be recommended. Entity models should be concealed from direct external interactions, relying solely on internal state transitions. A potential EntityPrototype record class may be introduced to store control instances before entity construction. Additionally, color field in entity is redundant and should be excluded.

Despawning would be facilitated by a dedicated control module, ensuring entities that do not support despawning retain null control references. This approach extends to both removal and entity death scenarios. Methods "delegated" to control operations could be structured as inline extension functions, serving as lightweight utility methods. Since controls can be dynamically assigned, their scope would remain unrestricted. Similar logic referred from ItemHolder, with control classes potentially leveraging extension functions for master classes like Entity.

There could be more details in the mechanisms of Composter, but it might still depend on the direction of the game. Therefore, it would be more complex and fancy.

The walkDist field in Mob is exclusively tied to sprite animations and serves no additional purpose. Animation-related data should be consolidated into an isolated object, ensuring responsibility is managed independently of the master class. Walking distance tracking for statistics should shift to move control modules, reinforcing modularity. Another dedicated object should store all geological data, such as position and direction, rather than scattering related fields within the mob structure. Attack cooldown logic should be carefully balanced, ensuring fairness in both offensive and defensive engagements, rather than a short-period immunity system that lacks strategic coherence. A Measure data monitoring class may be introduced to store both bounds and counter values for statistical tracking.

Passive mob behaviors, such as grazing, should incorporate realistic action countdowns rather than arbitrary event triggers. Both hostile and neutral mobs should feature enemy-tracking references, allowing intelligent behavior across different models. Current code indicates that the spawners are immune to creepers, though its chaotic implementation should be refined. Many mobs exhibit overly simplistic mechanisms, failing to convey clear behavioral differentiation

Spawning and Generation Mechanisms

Mobs may be primarily generated during world creation, foregoing natural despawning mechanics for added realism. The noActionTime field in Mob merely functions as a despawning check and lacks broader significance, and the speed field similarly lacks meaningful utility. Explosion damage should be applied directly to affected entities instead of being processed through an event interface like onExploded. While the mob level mechanism remains a potential feature, universal implementation is unlikely.

Spawning and despawning conditions should be adaptable to gameplay design and thematic direction. Since mobs are expected to spawn only during world generation, natural spawning systems should not be implemented, leaving chunk managers uninvolved in ongoing mob population dynamics. This means that spawner-like devices would not exist, with all standard structures spawned exclusively within desired areas during generation. Village generation, for example, should conform to standardized structure generation rules rather than ad hoc placement. Certain rock tiles may contain varying levels of ore or mineral content, supporting resource diversity within the terrain.

The Spawner mechanism is overly "reflective", making its implementation in TerraModulus improbable. If mob-generation devices are required, a more robust interface design should replace low-level structural dependencies. Additionally, daylight detection mechanisms for spawn prevention may require optimization if the lighting system undergoes revision.

Collision and Entity Movement

Having both isSolid and blocks for collision conditions is redundant. A more effective approach would be relying solely on blocks while dynamically assessing entity data upon collision when necessary.

Since movement would be primarily governed by the physics engine, explicit move method calls for entities are unnecessary. However, velocity and acceleration control remain crucial, and should be managed through dedicated move control modules. The previous sub-stepping implementation attempted to divide movement into discrete units, but this approach becomes impractical within a decimal coordinate system. Prior to the introduction of the improved algorithm, tunneling artifacts occurred at higher speeds, which was a known issue that had been previously reported.

Entity Behaviors and Interaction

The handling of Arrow had been previously discussed, and its integration into the new entity system would significantly improve collision detection. The logic referencing tile ID 16 (lapis ore) appears misplaced, with no clear rationale for its inclusion in entity behavior checks.

While ItemEntity would not feature inherent acceleration, initial velocity and ground friction would still be considered. The current implementation misinterprets velocity and acceleration principles, leading to confusion. Ideally, each ItemEntity instance should represent a stack of items, with distinct properties for total elapsed time and remaining time until deletion. If the remaining time is null, the entity should persist indefinitely. The two fields currently designated for synchronization serve little practical purpose.

With a 3D world system, vertical motion calculations for animations would be directly handled via physics, while shadow rendering would be standardized across all entities. Entity initialization should leverage factory methods, rather than relying on Level for instantiation. Additionally, initial direction and velocity assignments should be handled through specialized factory methods for, e.g., item and loot dropping.

Explosions and Advanced Motion Physics

The Tnt class would remain classified as Furniture, but with enhanced visual effects for explosion interactions. Explosion mechanism for both entities and tiles should conform to generalized explosion algorithms, considering exposure factors and interaction ranges. Calculation methods for entities and tiles should remain distinct, though shared foundational attributes could streamline consistency. Explosion mechanism may support different explosion types, such as gas-based or fire-based detonations, although it may just be using different attribute values.

Character movement should follow gradual velocity increases, simulating realistic acceleration rather than initiating motion instantaneously; friction calculations may be involved. Knockback effects from external forces should be processed as impact events, converting external force into a constant velocity, counteracted by environmental resistance (ground friction or air drag).

Item and Entity Property Management

The burn command for mobs should be structured as a dynamic property assignment, allowing flexibility rather than implementing it as a static function.

Item data management has encountered structural issues, requiring separations between metadata, properties, and display information. Each category should be used differently based on contextual needs. For example, descriptionsshould be recognized as part of display information, rather than intrinsic item metadata. They were originally not clear for their categories.

Some item attributes may be hardcoded, while others would be dynamically assigned. Internal attributes would be categorized as item properties, whereas externally accessible attributes would be classified as item tags. Both properties and tags may be static or dynamic, depending on their intended behavior.

Tiles should incorporate structured lists of properties, including but not limited to:

Flammability
Material type
Hardness
Blast resistance
Luminosity

Tile metadata and models should remain independent of tile positions, ensuring proper decoupling. The current handling of tile variants within DirtTile does not adhere to this principle and requires revision.

Food and Potion Systems

If a new food system incorporating various recipes is introduced, materials should maintain minimal attribute definitions, given their shared item-type similarities. Consuming food would still deplete stamina, ensuring balance within gameplay mechanics.

Should potion effects be integrated, they should function as universal interfaces applicable to all living mobs. A dedicated registry should store all potion effect types and their handlers, allowing structured public access and customization.

Item and Tile Systems

When items are linked via resource keys, manual association of sprites or models for individual items should become unnecessary.

Instead of relying on Tile$Material, tools should derive behavior from tile properties rather than being constrained by an enum. Identifiers should reside in the registry rather than within individual tile instances, though a resource entry may be returned, containing both the resource key and object. Other enum-dependent implementations should transition toward tile properties or metadata, since using enums as function arguments limits tile variation flexibility.

Using an enum-based approach in DecorTile provides little advantage over conventional constructors and fields. Instead, tiles should be registered as fields, with arguments directly passed into constructors during initialization.

Current implementations of Tiles and Items classes exhibit low-level, disorganized structures, requiring structural improvements for maintainability.

If FarmTile serves merely as a variant of DirtTile, their mechanisms could be merged rather than introducing a separate tile classification. Dirt tiles would dynamically transition into farmable states based on moisture and nutrient levels, allowing crops to grow naturally on them. Rather than defining a new tile, crop interactions would be tied to soil conditions, ensuring a more seamless agricultural system.

Rendering System and UI Management

No class should extend the core game class TerraModulus. The RenderSystem would exclusively oversee high-level operations, with UI components delegated to a dedicated manager, such as UIManager or MenuManager. Sprite management should be merged into the registry system, ensuring unified asset handling. Rendering performance should be hardware-accelerated, relying on native libraries and APIs, particularly for physics-based computations. Specifics regarding graphics and physics implementations will be addressed later. The major render system would manage the world renderer states, separating UI and world graphics to ensure clarity and modularity. Various fields should be decoupled accordingly.

The window size would remain dynamic, with no fixed dimensions. Instead of a canvas scale, an adjustable UI scale would be introduced, functioning as a simple multiple of base dimensions, allowing users to modify UI scaling based on window size without storing these preferences. The entire render system would be backed by an optimized rendering engine.

The canvas should be fully occupied, with UI elements such as the status bar, Heads-Up Display (HUD) components (e.g., hot bar, boss bar, debug screen) structured as menu components rather than static overlays. Level rendering should integrate world background and sky rendering as part of comprehensive world rendering.

Fullscreen functionality and screenshot handling should be managed directly within the render system rather than external methods.

The canvas should aggregate rendering results from both UI and world graphics, efficiently processing composited graphical elements. Internally buffered arrays might be employed within the rendering engine, improving performance and consistency.

Rendering queues present a valuable optimization strategy, streamlining render request processing. However, clearing request validation introduces unnecessary overhead. Instead, hidden elements should be skipped within rendering functions, rather than processing functions for rendering requests. Maintaining well-defined responsibilities prevents inefficiencies.

A separate overlay canvas for light rendering should be eliminated, with a single unified canvas handling all composited graphical elements via the rendering engine. The current workaround approach in Screen, aimed at optimizing rendering, is not the most practical industry standard. Furthermore, sprite rendering and inverted color line rendering have yet to capitalize on Graphics2D optimizations, which could be enhanced through OpenGL hardware acceleration.

Several menu components require updates, including the hot bar and various menu frames, though these adjustments align with graphics discussions scheduled for later stages.

The rendering system will likely be custom-implemented for optimal compatibility and usability. However, it remains beneficial to explore existing meshing, modeling, or sprite libraries as supplementary tools.

The toggle mechanism for hardware acceleration remains an open question. Under normal circumstances, there should be no need to disable hardware acceleration, but certain edge cases, such as graphics card incompatibility, may require turning it off. If compatibility issues arise before the application fully launches, in-app toggling becomes impractical. Instead, hardware acceleration settings should be configured via launch arguments or managed internally (by detection), preventing post-launch usability issues.

Loading Screen and Splash Screen

The loading screen and possibly a splash screen may include a hardcoded embedded image within the application. The loading screen would feature a progress bar, initializing full loading of system components in a manner similar to Minecraft: simple, clean, and iconic, after the appearance of the splash screen with a logo. Additionally, customized splash screens could be inserted pre-title screen, or during world initialization phases (early, middle, after).

Initialization timing for tiles, entities, and items remains under discussion, contingent on registry implementation methods. The title screen should appear only after initialization completes, ensuring all essential resources are preloaded. If fullscreen mode is enabled at launch, the window should transition immediately upon initialization, rather than employing a delayed separate switch.

Font System and Text Rendering

The Font class might be consolidated within a single package, likely hidden inside the graphics module. Character mappings (originally as chars field) would be fully configurable through asset packs. Multiple text rendering modes should be available, including plain, formatted, and pre-styled options.

An optional configuration object, such as FontStyle, may be utilized when drawing text, supporting attributes like background color and foreground color. Features about rendering boundaries should be deprecated, including horizontal scrolling function. Horizontal scrolling should be reserved exclusively for animated automatically scrolling text, rather than general user interactable menu elements.

Text rendering modes could be consolidated into a single configuration attribute, ensuring streamlined customization: plain text remains unstyled, while enabled attributes apply respective formatting styles. The stack-based "redo" mechanism remains excessively complex, warranting removal. Width calculations may leverage advanced looping techniques, eliminating invalid formatting codes from processing logic. A display component architecture may mitigate such issues. Centered text rendering may also be discarded, with paragraph alignment and shadow effects serving as refined alternatives. Current FontStyle implementation is inadequate and could benefit from a transition to UI component-based solutions.

Texture Atlas and Sprite Architecture

A texture atlas system would be implemented to improve memory efficiency, particularly in lower-level engines handling many sprites. All textures would use a shared set of atlases, with new atlases created only when the existing ones reach capacity. However, sprite dimensions may not always conform to square proportions or power-of-two sizing, requiring an efficient algorithm. A useful reference for this approach is David Colson’s rectangle packing article . Since most textures will be rectangular or square, all assets should be fully loaded from relevant asset packs. The MinicraftImage class serves as a simplified alternative to existing image handling structures but may not be used in this framework.

A high-level wrapper would be introduced to manage low-level sprite instances. When sprites are incorporated into the centralized texture atlas, each Sprite should correspond to a data pointer containing information about pixel locations within the atlas. These references would then be registered appropriately. Additionally, Px would be replaced with OpenGL functionalities, ensuring greater performance and flexibility.

The SpriteAnimation class should be restructured to function as animated instances of sprites, storing only a pointer with minimal additional data. Animated sprites may exist as either UI components or animated texture segments, following the specification outlined in the asset pack system. Internally, its current implementation is overly complex, rather than aligning with external timing mechanisms, animation should operate using a tick counter within the world ticking function. Borders should not be managed here, but delegated to a separate layer. Since sprite instances should be recreated rather than "relinked", the system should undergo a complete redesign for efficiency.

Registry-Based Sprite Management

The codebase currently lacks structure, requiring major improvements. SpriteManager should be governed by registries. If skins are directly managed via accounts, their in-game management should be simplified, avoiding unnecessary record-based associations. "Missing" texture management should be handled at a higher level, rather than simple functions. With a fully implemented texture atlas system, sprite sheets should no longer require inefficient separate asset loading. The SpriteMeta structure is redundant and should be discarded and rebuilt. Additionally, SpriteType is messy and inefficient, and would function better under a registry-based approach.

The interaction between Sprites and SpriteLink remains overly convoluted. The logic behind mirroring tiles individually without considering relative positional adjustments within the sprite system suggests a lack of proper filtering mechanisms. OpenGL already provides these functionalities, making a manual extra implementation unnecessary.

Entity Rendering and Texture Models

The entity rendering process should primarily rely on texture models, though appearance may vary based on entity states and properties. Instead of using method overrides, an external control element should manage state-dependent visual changes. Interactive models should remain distinct, affecting only physics mechanics, while both elements remain stored within entity data. Interactive models should be simpler than texture models, as textures will be preloaded within the atlas system.

Mob sprite compilation should be generalized into standardized models, due to the new texture model system. If mob models follow a consistent framework, there should be no need for separate sprite sets to manage player "carrying" textures. Player rendering would remain straightforward, covering motions and postures such as swimming, item holding, and falling. Item holding visuals should be refined to better distinguish held items, rather than exclusively displaying items when actively in use.

Item Rendering and GUI

Items are primarily rendered within HUD components or inventory menus, with additional instances appearing as item entities. When rendered within the GUI, supplementary attributes such as durability should be displayed alongside item models. To ensure modular structure, models and GUI rendering functions should be cleanly separated, preventing unnecessary dependencies.

Particle Animation and Tile Interaction

Using ExplosionTileTicker for animation is inefficient, as leveraging tiles for extended animations is a fundamentally flawed design. Given the tile system, where tile ticking is random while tile entity ticking is regular, introducing transitional tiles disrupts system consistency. Instead of relying on temporary tile states, particle animations with immediate tile destruction would maintain structural integrity within the system. As a result, ExplodedTile should be removed entirely.

Each particle entity should represent a single sprite unit, ensuring visual and mechanical consistency. Additionally, particles should not actively interact with other world objects. TextParticle, in particular, does not behave like a particle but rather serves as a distinct visual effect, making its implementation overly complex for a basic particle system. Likewise, WaterParticle is excessively intricate, its code structure could be simplified significantly if initial velocities were integrated as standard entity mechanisms.

The glint particle used in WateringCanItem should be formally registered rather than arbitrarily managed. Additionally, splashing particles should be handled within a more generalized framework, preventing fragmented particle logic. The current implementation of the watering can is close to be a magical element in the gameplay, thus should be rewritten. The item should target an area near the player similar to how the particle effect works, rather than focusing exclusively on the selected tile, ensuring broader interaction across surrounding tiles.

Shaders and Light System

Darkness rendering should be refined based on light sources and light levelsstored within chunk data. The rendered light intensity would be influenced by the gamma value setting. Mirroring settings of textures could be streamlined into two separate boolean values, while sprite transformations should be handled through alternative methods, such as filtering functions or similar optimizations.

The Lantern could be considered one of many advanced artificial light sources. The lighting system requires significant improvement, particularly in handling light obstruction by objects, which would enhance gameplay immersion. Currently, the system does not account for blocked light, and addressing this limitation could make lighting mechanisms more dynamic and interactive. Additionally, enum-based if-else chains should transition to switch-case structures if the Java Development Kits (JDK) version is updated, while Kotlin implementations should utilize when expressions for better readability.

World background rendering should be governed by shaders, ensuring that scenery remains dynamic rather than static. Entity rendering distance should be adjustable via client settings, allowing users to configure their viewing range accordingly. Lighting effects, including blending and raytracing, should be processed through shaders, using tile light values as input data. However, entity-based lighting information should remain excluded from mechanical logic, meaning player sight obstruction would not interfere with shader computations.

Resource Pack Management

Resources and assets should no longer be loaded before canvas initialization, but rather loaded simultaneously with vanilla assets.

Resource packs would no longer rely on directory structures, as assets would be categorized separately within the system. Only packs designed for the current application version should be fully supported, with older packs receiving minimal compatibility support where possible. Vanilla assets, if fully configurable, should be bundled into a single package, either embedded within the binary or provided as an external resource, though modifications should be prohibited, ensuring they are only loaded once on startup. Traversing embedded assets manually would no longer be necessary.

The resource pack selection screen would maintain a two-tab structure, unloaded packs on the left and loaded packs on the right. Higher screen resolution eliminates the need to combine both lists into a single view, as that could reduce discoverability. Only brief descriptions of packs should be displayed, avoiding unnecessary detail within the UI. A watcher system should actively monitor all packs, ensuring real-time updates. Upon exiting the pack selection screen, all selected packs should be loaded into memory, though they should remain unlocked, allowing dynamic reloading of all packs except vanilla assets, even when no changes occur within the UI.

When new items are added to the pack list, the selected entry should remain active but shift accordingly. If an entry is removed, the selection should be cleared. Entries would not be selected by mouse but hovered; other control methods would select entries. The file watcher should function as an independent utility class, not an inner class of the display system. During loading, packs should be locked, preventing modifications until the process completes. Movement of packs via keyboard or controller should require the entry to be actively selected, reinforcing structured navigation. The entry list might use a linked list. Soft-read locks should secure pack metadata retrieval, ensuring safe concurrent access. Logo dimensions should remain unrestricted, though optimal sizing should always be enforced. Stream handling remains low-level, so a better manager should be made; filtering assets for retrieval should no longer be necessary since it would no longer contain a directory structure. Only embedded assets may invoke readStringFromInputStream, though usage might be minimized. Metadata structures should transition away from JSON, with asset content organized across multiple files instead of relying on single metadata sources.

Pack Loading Steps

Initial Pack Loading during Application Startup

Locate the directory containing packs.
Match available packs with user preferences.
Select compatible packs; issue warnings for unsupported versions or missing entries.
Apply a hard lock on all packs, including vanilla assets.
Verify vanilla assets via checksum validation; emit errors if inconsistencies arise.
Load selected packs sequentially into memory.
Release all loaded packs upon successful completion.
Finalize initialization.

Pack Reloading after Saving Preferences

Unload all assets except vanilla assets.
Identify matching packs and apply hard locks; issue errors if needed.
If recoverable errors occur, revert to the pack list screen.
If a pack updates during the process, return to the pack list screen.
Load selected packs sequentially into memory.
Release loaded packs.

The built-in loading screen before loading other assets would use embedded assets in the binary. Although vanilla assets are managed differently, their format and specifications should remain consistent with custom packs. The key distinction lies in how the pack manager aligns operations within the standardized system.

Pack Management and Initialization Framework

Since pack loading requires distinct stages for initialization and configuration, a single pack manager may not suffice. Post-construction initialization is not considered design-safe, requiring multiple utility components to streamline pack management workflows. Individual packs should remain self-contained, ensuring safe execution structures.

All pack loading operations should be handled through a function within the manager system. Skins should be processed separately from other assets. If assets are loaded at a lower level, their processing logic might not reside within the Kotlin part.

Exceptions should be handled at the individual pack level, meaning that strict mode should terminate loading upon encountering errors, while compatible mode should bypass minor exceptions, allowing support for older versions.

Book Display and Text Management

The BookDisplay implementation likely stemmed from previous attempts to introduce editable books, but text field handling requires improvement. Further refinements should be discussed separately. Additionally, books without a title and page count should not be classified as books, meaning they should never use this class.

The PagedDisplay class is redundant, as its functionality can be more effectively replaced by a scroll pane, which offers greater ease of use. Unless the system explicitly manages book-like content, pages are unnecessary, making the class unnecessary.

Book Content Handling and Translation Compatibility

Since books function as items with stored content, translating them dynamically is discouraged, except for special book types. Translation discrepancies between server-side and client-side configurations could lead to inconsistent content display. For game-generated static books (such as "Antidious"), these should be categorized separately from regular book items to enable client-side translation support.

Thus, BookData should be considered an outdated structure, as static book content should be managed via assets rather than hardcoded within the code.

Utilities

In Kotlin, Action should be replaced with a typealias for the function type () -> Unit, declared in the utils file. Similarly, Condition could be defined as () -> Boolean, or enhanced using a functional interface for additional capabilities. MonoCondition would be better replaced by Predicate, improving clarity and modularity.

Dimension, Insets, Point, and Rectangle should transition into immutable value classes. These graphical elements should be encapsulated within a dedicated graphics module, restricting use to client-side operations. Additional functions should support transformations and interactions with these classes. Insets may be redundant if Dimension suffices, though certain functionalities could be directly merged into Rectangle.

The Ellipsis class likely functions as a dot-based loading animation, a format that is rarely used today. Modern UI favors progress bars, spinners, or more dynamic and informative animations. The dot animation system should be retired. SequentialEllipsis seems to loop a single dot, while SmoothEllipsis mimics a looping effect similar to a progress bar. The DotUpdater updating functions, though varied, lack significant justification for inclusion and should be reconsidered.

Direction-based positions should follow standardized enum class structures, supporting configurations with two, four, five, eight, or nine directions. RelPos exemplifies a set of nine directional positions in a rectangle, while border layouts require sets of five directions. xIndex and yIndex should be computed dynamically based on ordinal, rather than being explicitly stored. To enhance value safety, methods such as getPos should be avoided, with getOpposite optimized through switch-case structures in Java or when blocks in Kotlin. If anchors must be defined based on Point and RelPos, an Anchor class should be introduced. Some classes may support interchanging, while "to" functions are preferred over "from" functions.

Both Tickable and ClientTickable serve no real purpose, lacking clear functional necessity, and the reason behind this cannot be found. Direction represents entity activity directions, but its enum constraints limit extensibility when additional directional systems are required. A spherical coordinate-based vector system with unit length definitions would provide better flexibility for direction systems without a limited set of directions.

Positional coordinates (e.g., x and y) should not be treated as separate parameters, but consolidated into unified structures, particularly when representing vectors or object positions. This ensures a single-point reference model, applicable to simple vectors, tile positions within dimensions, or other models. Returning entries, such as tiles from chunk managers, should yield data objects containing positional references, tile types, and metadata, allowing direct modifications without extra method calls. Entity instantiations and removals may adhere to a similar management approach.

Most functions in MyUtils can be replaced with helper functions from existing utility libraries. Any functions explicitly designed for this codebase should be declared as top-level functions for improved simplicity.

Due to limitations in coordinate system design, Vector2 remains underutilized in the codebase. Multiple vector classes should be introduced, accommodating different parameter configurations and vector properties. These classes should maintain record-class structures with immutable fields or as value classes, ensuring primitive-like value integrity.

Certain utility classes could be introduced for synchronization. Thread-safe maps and collections require simplified syntax, resembling Rust’s Mutex and ReadWriteLock implementations. Note that existing libraries may already have been available for this purpose. A similar class could be introduced for managing internally held instances, though ensuring proper design remains the developer’s responsibility.

Toast and Notification System

Toast notifications should replace on-screen notification text, with world events communicated via chat and client-side game events displayed as toasts. Each toast notification should maintain its own defined lifetime, ensuring structured appearance and expiration timing.

The current toast usage hierarchy lacks consistency. AppToast extends Toast, yet is handled at the same level as the superclass, leading to inconsistencies in field usage. Toasts should either be defined as separate classes or treated as structured subclasses within a unified system, depending on implementation needs. The behavioral distinctions between toast categories suggest that a simple classification approach may be not enough.

Application (system) toasts and gameplay toasts share similar structural properties but serve different use cases. Separating them visually by placing application notifications in the bottom-right corner (similar to Windows system notifications) and gameplay notifications in the top-right corner would provide clear distinction and accessibility. Additionally, players should be able to review gameplay notifications via the inventory menu and access application notifications through the pause menu or other non-gameplay screens.

Toast content should support formatted text, allowing bold titles (pre-styled) and detailed descriptions via external functions responsible for formatted output generation. Instance creation may not be overly frequent, permitting full instantiation of frames without singleton constraints.

The toast constructor design should maintain similar parameter structures, with subclasses deciding primary attributes while allowing optional extra parameters. The sprite property should be treated as an abstract property rather than a constructor parameter, enabling consistent code pattern generalization. Toasts should support nullable values for custom rendering, yet favor sprite-based rendering for simplicity. A separate class may be introduced to align both code quality and flexibility of rendering, allowing sprite rendering and advanced animated rendering.

Toast animations could be introduced but require separate discussion regarding implementation specifics. AchievementUnlockToast is not optimally used in its current form, better structural management for content, frame design, and animation handling should be considered.

Fishing System

All fishing-related code should be contained within the item class, ensuring logical structuring and consistency.

The FishingData system could be converted into resources linked via resource keys, streamlining data management. For realism, the fishing line could have durability, with the potential to break if stretched excessively, requiring players to replace it. Fishing rods should be restricted to catching fish rather than other item types while remaining usable anywhere.

Tool Type System and Mining Mechanisms

ToolType should transition from an enum-based system to a class structure, with all instances stored as constants. ToolItem instances should reference these tool types directly during construction, avoiding loop-based initialization. The tool type should be stored as an item property, enhancing modularity.

If tiles incorporate a health system, mining level mechanisms for tools may become obsolete. However, a dual health meter system could be introduced, where using an appropriate tool activates a mining meter, accelerating resource extraction. Additionally, single-action harvesting tools could bypass the mining meter entirely, eliminating the concern altogether.

Tool durability should remain exclusive to item instances, rather than being defined by tool types. Some materials may not have a dedicated tool type, requiring alternative mechanisms for extraction. Attack damage bonuses for tools may be managed via mappings, associating attacker and target attributes before linking them to specific tools. These collections could be registered to the specific tools systematically, ensuring consistency across combat interactions.

Bucket and Plant Systems

The BucketItem should store fluids rather than tiles, ensuring alignment with its intended purpose as a liquid container. The concept of a "hole" tile value is illogical; it should instead be null, emphasizing the absence of fluid rather than an arbitrary placeholder. Additionally, the offset field, once relevant for sprite sheet slicing, has lost its necessity and can be replaced by ordinal() for cleaner indexing.

Some systems like plants, should not depend exclusively on tile instances, but rather on plant class registrations within the game system. This registry-driven approach allows tiles, items, and other related elements to be registered based on the core plant classifications. This includes but not limited to trees, crops, flowers and grass. Given this structure, mechanisms for TreeTile, FlowerTile, and CropTile should be unified, providing shared functionality while still allowing specific behaviors per plant type, such as fertilization.

Both real-life and fantasy plant species could be supported within the system, introducing unique mechanisms like growth, harvesting, nutrition, diseases, and life stages. Examples of fantasy species include heavenly berries and hellish berries. Each plant type should undergo systematic development, ensuring its growth mechanisms align with environmental factors.

Color System

The legacy formatted color values are no longer relevant and should not be maintained. Instead of relying on the tint function, the system should support expanded predefined colors and alternative utility functions. Formatting codes should not be handled within Color, but rather assigned to a dedicated class responsible for display string formatting or another appropriate system. Additionally, original format codes may contain control characters, which cannot be directly typed, limiting their usability.

The limit function is redundant, as it essentially mirrors the existing clamp function, warranting removal. Tinting functions, along with most related functions, should also be deprecated, though the lightness calculation method may still be repurposed elsewhere.

The ClothingItem class has faced long-standing issues, primarily due to its lack of meaningful functionality. Its mechanisms should undergo a complete rewrite, aligning with a system similar to Leather Armor in Minecraft, allowing for greater adaptability and refinement.

The dye system should transition from a limited color selection to a more extensive color range, such as a full 32-bit RGB spectrum. However, dyed objects may present modeling challenges, requiring a blending or color mapping function for proper visualization. A 1-bit image may facilitate tinting textures, using a dedicated tinting function to ensure consistent rendering across different dyed materials.

Wool and Carpet Tiles

The canWool mechanism appears illogical and should be considered for removal from the game. Replacing wool tiles with carpet tiles may be a better alternative, ensuring cleaner mechanisms and usability.

Audio System

The audio system would be redesigned using OpenAL, supporting multiple new features. However, mixing algorithms will be excluded, opting instead for simple additive methods to maintain audio consistency and quality.

Non-game environments should feature only UI sounds and background music (BGM).
In-game audio should exclusively originate from world-based sound sources while an additional ambience track may be available for environmental audio like raining, ensuring immersive interactions.
In-game sound sources should remain unrestricted, while non-game elements should be limited to the two primary sources.

Further implementation will be integrated within the future asset system, allowing flexibility and refinement in managing sound behaviors.

Achievements, Advancements and Quests

If achievements transition to an account-based system, all related data should be sourced from the server rather than application assets. This setup would resemble Xbox achievements in Minecraft: Bedrock Edition, enhancing cross-device consistency.

To improve code organization, achievement logic and non-UI components should be decoupled from display-related classes, ensuring separation between achievement management and loading mechanisms. Achievement notifications could be redesigned using toast notifications, delivering more streamlined and engaging alerts.

The existing Achievement unit class is lightweight, containing only a few fields, while other logic resides within the display class, hindering clarity. Reorganizing these elements would greatly improve structure and maintainability.

The quest management system should be separated from QuestsDisplay, similar to how advancements were intended to be structured. The quest tree display is complex, requiring keyboard navigation in keyboard-only mode, with mouse-based drag controls as an alternative. The components could be linked like how OSK keys work.

Several methods for node description presentation could be adopted:

Tooltips displaying descriptions on hover
A dedicated panel showing hovered-node content
Selecting a node via click and displaying content in a separate menu

The selected method would depend on feature support and available UI capabilities.

The JSON-formatted data structure of the quest data would no longer be important when a new data format will be proposed. Entry and menu data synchronization with internal states should always be preferred over full refreshing of the status and lists to prevent unnecessary overheads and ensure consistencies and simplicities. For example, refreshDisplayableQuests and reloadEntries should be used upon data loading.

Inner subclasses within Display subclasses remain overly complex, but they could be converted into structured submenus.

Quests should not exist as singletons—instead, but structured in series, with individual quests unlocking sequentially. The quest menu layout could display series lists on the left, while the quest tree visualization occupies the right-hand menu. Series descriptions should be positioned either at the top or bottom of the tree layout.

Quests and advancements should remain distinct, as quests rely on a more abstract progression systems than advancements. In abstraction of them, they are called "Progression Systems". Locked quests should always remain hidden to prevent spoilers, while advancements may be selectively obscured based on advancement levels.

The SeriesQuestViewerDisplay implementation is notably complex. A configurable shape-drawing component could simplify tree-node visualization, allowing quest nodes to function as discrete UI components. If geometric shapes could be defined as components, direct placement within component pools would remove the need for graphical drawing algorithms, significantly reducing system complexity. Elements could be aligned as rows and columns.

The quest tree structure should follow a linked list format, eliminating the need for quest relationship mapping, as node-to-node connectivity should be handled within the component instance itself. Auto-scrolling should apply to keyboard and controller users, while mouse-based users can drag the tree view freely.

If a quest manager is instantiated, functions such as resetGameQuests and load should be deprecated, with saving handled via the world save manager.

A select dropdown should replace separate tabs for unlocked and completed series, allowing users to filter quests by "All," "Unlocked," or "Completed", with an empty right-hand display when no selection is active.

Given shared properties between advancements and quests, a common superclass should be introduced for simplification and structured inheritance. Most modular vanilla components would allow easy toggling by using same public interfaces.

Advancements should be grouped into topic categories, while quests retain series-based organization. Multiple progress trees may exist within a single category or series, ensuring flexible progression tracking.

Progress elements should require both prerequisite elements and completion criteria, ensuring tasks marked as completed remain fulfilled upon later prerequisite unlocking. All elements should form tree-linked structures, with preceding elements acting as unlocking requirements for succeeding tasks.

To streamline structural definitions, JSON should be replaced with more efficient and structural formats, simplifying creation and parsing.

When the preceding elements have already to be completed, the succeeding elements should be unlocked unless they are necessarily hidden; if completing an element requiring completing another element, it should be included as a kind of criteria instead of being a "requirement" as both "criterion" and "requirement" have similar meanings. Also, rewards should be abstract and with a list for each element. Each criterion may follow with description and the boolean value of whether to be visible.

The existing AdvancementTrigger should be renamed to AdvancementTracker, ensuring clarity in role designation.

Trackers should be registered systematically, facilitating progress detection. They should also support anonymous class implementations, where active instances reference both metadata and tracking status, whether global or player-specific.

Trackers should incorporate generic types, reducing verbosity in type checking. Additionally:

Rangeable should be split into Minimum, Maximum, and Range components.
ItemConditions should transition from stack-based tracking to count-based evaluation.
No-op conditions and trackers should be removed, with empty conditions used for structure simplification

During the processing of elements, the game should not be saved for progress synchronization. During tracker evaluation, elements should be bundled into record objects, preventing mutual inclusion complexity.

Given the complexity of AdvancementElement, separating responsibilities into distinct classes in a separate package would enhance organization. Recipe-locking functionalities should exist within a public interface, enabling advancements or other game systems to control recipe access restrictions. Advancement categories may be user-specific or global, depending on the nature of the tracked progress.

Unlike advancements, quests may require prerequisites to be unlocked before tracking completing criteria, preventing pre-unlock tracking. However, the quest system should align structurally with the advancement system, as its current format is outdated.

Help and Tutorials

The HowToPlayDisplay screen may be retained, with key rendering improved using images, displaying mapped keys rather than textual key names. Paragraphs should be consolidated into a single translation entry, simplifying content management, though minor inconsistencies in line breaks may be unavoidable.

Control schemes should be categorized into three types:

Keyboard-Only
Mouse & Keyboard
Controller-Only
Controller & Keyboard

Pure controller scheme may not be recommended if controller connection is unstable. If the controller-only option is selected, the keyboard remains unusable even when the controller disconnects, so it is essential to carefully select the options. The order of the names follows the major-first, minor-second pattern. Each keybinding set does not share settings to another scheme.

Each control scheme should feature structured guides, with subtopics categorized under individual control settings. General gameplay guides should be separated from control-specific guidance, while advanced guides may be referenced externally, such as in a wiki. The layout might align with the structure used in Minecraft: Bedrock Edition.

Tutorials should be restructured, merging into the How-To-Play screen or menu-based guides, akin to RimWorld’s guide system. Completion detection should occur in real-time, tracking interactive elements and input events dynamically. Minecraft’s tutorial approach is not applicable here.

General Code Style

In this section, code style is mostly about the Object-Oriented Programming (OOP) practices in Kotlin.

Data Structures and Debugging Practices

Prefer static fields when runtime lists offer minimal advantages, as static fields can enhance debugging accessibility.

Avoid overly concise code for performance assumptions. Reducing code size does not necessarily improve runtime efficiency, nor does using loops inherently indicate a better or smarter approach.

Constructor and Property Optimization in Kotlin

Constructor parameters should be simplified into abstract properties when applicable.

Not all initializations need to be placed directly in constructors. Factory helper methods can often provide better structuring.

Low-Level Programming Considerations

Extra caution is required when dealing with low-level programming, even in Kotlin, to ensure logical correctness without over-reliance on compiler warnings or errors.

When porting lower-level APIs to higher-level components, developers must remain vigilant to prevent overlooking critical details.

Utility Classes and Code Idioms

Internal logic in utility classes often operates at a lower level, requiring careful adherence to coding best practices.

Resultant syntax should remain idiomatic to Kotlin, leveraging practical code patterns for improved clarity.

Event Listener Management

Listeners must be properly registered and unregistered when their holders are destroyed or deemed unnecessary.

"Zombie" listeners should be avoided, as they introduce memory leaks and unintended behaviors. Compilers may not be able to check this.

Using try-finally within try-catch blocks can ensure proper listener cleanup.

Class Visibility and Scope Management

If a class’s construction is restricted through private or internal constructors, it can be safely marked public. External code cannot instantiate it directly, keeping control within its intended scope.

Some classes serve specific roles within their domain. Even when publicly exposed, they primarily function within their designated boundaries, ensuring their responsibilities remain intact.

When data moves across domains using publicly available class definitions, safety is maintained as long as the system is well-designed. If the original domain constrains data creation, the integrity remains protected even when other domains use the data.

Code Entity Visibility

General-purpose functions or constants should be moved outside their class definitions for broader accessibility.

Restrict visibility of specific members to prevent unintended modifications and access that might break mechanisms.

Counter Mechanisms and Event Timing

Numerical countdowns should only transition to count-up models when multiple events require tracking.

If an event dominates counter functionality, both countdown (for the event) and count-up (for secondary triggers) may coexist.

Increment/decrement operations should occur at the end of ticking functions, while conditions regarding counters should be processed beforehand.

Avoidance of Unnecessary Singletons

Singletons should be avoided unless absolutely necessary, as they introduce additional verbosity and unused code overhead.

Class Nesting

A single file should contain no more than two levels of inner classes, ensuring better readability, organization, and simplicity of class structures, even preventing unnecessary verbosity.

Classes referenced exclusively by another class within a multi-component package should be nested within their primary referencing class, particularly when thematic consistency applies.

Extension Function Usage

Extension functions should be used thoughtfully, avoiding excessive pollution that negatively affects coding experience.

Field Initialization Guidelines

If inline initialization is enough, fields should be kept inlined rather than delegated to an initialization block.

String Key Avoidance

String keys should be avoided whenever possible. Instead, defined constants or enums should be used to prevent runtime-only errors and potential bugs and reduce unnecessary parsing overhead.

String parsing is error-prone. In fact, when it is possible to use a way to use OOP features such as classes and builders, those should be used instead of string representations, prevent unnecessary parsing overheads and errors, even subtle bugs. This way, compiler may also aid debugging when the OOP way is used.

UI Control and Data

The Model-View-Controller (MVC) principle is an important pattern to separate responsibilities of different components in the application. For example, the UI instances of settings only view and control the internal setting values of the settings, but are not persistent or acting as container for internal code.

Development Roadmap and Direction

The review of the Minicraft+ codebase serves as the basis for the development of TerraModulus, focusing on both refining the existing project and strategically planning for the new one. The roadmap emphasizes a transition towards a more structured, modular, and efficient architecture.

A key aspect of this direction is a clear separation of concerns, reflected in the proposed organization of the core package and input handling. The application framework will feature a dedicated Main.kt file for client and server implementations. The Game component will be instantiated as TerraModulus, with versioning managed through the application profile. Input handling will be streamlined into distinct classes for key, text, gamepad, and mouse inputs, while controller management will be handled by a dedicated ControllerManager.

The user interface will undergo significant enhancements, moving towards a component-based design similar to AWT or Swing. This will improve code quality and scope management, with screens and menus managed more effectively. The settings system will be refactored into a Preferences system, ensuring settings values are independent of UI components. Furthermore, the logging system will be transitioned to a different library for improved idiomatic usage, and the save system will adopt a schema-driven framework to ensure backward compatibility.

The roadmap also outlines significant advancements in several other areas. Asset and resource management will be modernized to handle game resources more efficiently. Code quality is a central concern, with a drive towards cleaner, more maintainable, and idiomatic Kotlin code. A new framework is being developed to provide a robust foundation for TerraModulus. There is consideration for using Rust for lower-level engine components to potentially enhance performance. A brand new world system is planned, promising more dynamic and engaging environments. Enhancements to physics are also in scope, alongside the ambition to introduce a 3D world. The project aims to leverage more libraries to expand functionalities and features.

World generation will be a major focus, with a shift towards more flexible, customizable, and standardized procedural generation techniques. This includes the use of scripting and custom data formats to allow for extensive control over world generation parameters, features, and structures. The roadmap also addresses the importance of robust error handling, efficient memory management, and a consistent user experience.

Overall, the development roadmap for TerraModulus outlines a comprehensive overhaul of the Minicraft+ codebase, aimed at creating a more maintainable, extensible, and feature-rich game. This involves a focus on improved architecture, UI design, input handling, and world generation, alongside significant advancements in asset management, code quality, and the underlying engine, with an emphasis on modularity, clarity, and adherence to Kotlin's idiomatic principles.

The TerraModulus EFP Book