The Main Scene

Let’s prepare the dungeon generator’s main scene to focus on the algorithm in the next lesson.

Create a new scene. Select 2D Scene from the Scene docker and rename the Node2D to MSTDungeon. Add the following child nodes:

• Camera2D: make sure to have Current switched on in the Inspector and set Zoom to Vector2(8, 8) so we have enough room on the screen for the dungeon.
• TileMap: rename to Level. This node stores and draws the dungeon data.
• Node2D: rename to Rooms. This node is used as a group node for the rooms we’ll spawn during the build process.

Your scene tree should look like the following.

With the Level node selected, assign tileset-prototype.tres to the Tile Set property in the Inspector. Also set Cell > Size to Vector2(60, 60) as this is our tiles’ size.

Setting up the main script

Attach a script to your MSTDungeon node and remove its contents. We’ll start fresh with the following code:

extends Node2D


# Emitted when all the rooms stabilized.
signal rooms_placed

const Room := preload("Room.tscn")

# Maximum number of generated rooms.
export var max_rooms := 60
# Controls the number of paths we add to the dungeon after generating it,
# limiting player backtracking.
export var reconnection_factor := 0.025

var _rng := RandomNumberGenerator.new()
var _data := {}
var _path: AStar2D = null
var _sleeping_rooms := 0
var _mean_room_size := Vector2.ZERO

onready var rooms: Node2D = $Rooms
onready var level: TileMap = $Level

Since we can’t predict when the physics simulation stabilizes, we need signals to coordinate our actions. For this, we introduce the rooms_placed signal, which we emit after all the rooms stabilized and aren’t colliding any longer. We do so to know when to begin the next phase.

Next, we load the Room scene because we want to instantiate it from code as we require many rooms.

We then have a block of exported variables that appear in the Inspector as properties for easier manipulation. We define max_rooms to play with the number of rooms we want. reconnection_factor will make sense when we go over the algorithm. In short, we use it to bring back some edge connections after calculating the MST graph. This means more connections between rooms, which improves gameplay as the player doesn’t always have to backtrack.

Since this is a PCG lesson, we have to deal with random numbers, so we define _rng as our RandomNumberGenerator. Then we define _data as a dictionary since it’s faster than arrays at inserting and deleting operations. We use the AStar2D _path variable to store and manipulate the Minimum Spanning Tree graph. We use the variable _sleeping_rooms to keep track of how many rooms entered the sleep state. We use _main_room_size to store the mean size of the rooms. Our algorithm will select “main” rooms that are close to this size. We’ll see how this works in the next part.

Finally, we store references to the relevant nodes.