Godot Python MMO Part 3

22 Nov 2022

Welcome back to yet another lesson! In the previous lesson, we finished setting up authentication for our chatroom, thus completing our short-term goal.

In this lesson, we will really kick it up a notch by attaching the notion of position and movement to our players, making this feel more like a game.

If you prefer, you can view this lesson on YouTube.

I highly recommend you go through the first and second parts, if you haven’t already. If do you want to start here without viewing the previous lessons, however, you can visit the Releases section of the official GitHub repository, and download the End of lesson 2 code by expanding Assets and downloading Source code (zip).

A sneak peek

Here’s a quick look at what we’ll be finishing up by the end of this lesson:

A demo of what's to come…

Database design

We’re going to want to store information about not only the user, but also their associated actor.

An actor has a user, but it also has a position and a name, etc.

Here’s the most agreeable structure for further down the track when you might want to add items to your game as well as actors, or abstract entities:

Actor Model

The way to read this is every actor has a user and an instanced entity. Every instanced entity has an entity.

If we wanted to extend this later on, we could create something like a loot chest, which needs a position and a name but doesn’t need a user. Loot chests require loot, of course. But this loot might not exist properly in the world so doesn’t need a position:

Loot

That’s just an example of how you might want to extend the database, but for now we will just be adding support for actors. This means adding some new models.

Adding some new models

Let’s add models for our actor in server/models.py, as we discussed, we start with an entity which is the simplest model with just a name.

class Entity(models.Model):
    name = models.CharField(max_length=100)

Next, we will define instanced entities which have a position and an entity.

class InstancedEntity(models.Model):
    x = models.FloatField()
    y = models.FloatField()
    entity = models.ForeignKey(Entity, on_delete=models.CASCADE)

And finally actors, which have a user and an instanced entity.

class Actor(models.Model):
    user = models.OneToOneField(User, on_delete=models.CASCADE)
    instanced_entity = models.OneToOneField(InstancedEntity, on_delete=models.CASCADE)

Note here we are using ForeignKey for the Entity reference in InstancedEntity. This simply describes a many-to-one relationship, whereby many InstancedEntities can reference the same, single Entity. For example, I might have one entity called Book, but I might want to have multiple instances of Book, all with differing positions in the world. This is what ForeignKey allows us to do.

In contrast, you will notice the Actor model has a OneToOneField relationship with the User and InstancedEntity models. This means each Actor can only have exactly one unique User and InstancedEntity.

You will also notice we are specifying on_delete actions here. These are to left Django know what to do if the referenced field gets deleted. The CASCADE option means, if the referenced object is deleted, then so too should the object referencing it.

For example, suppose I had an entity called Book, and perhaps multiple instanced entities of Book. Then suppose I deleted the single Book entity. What should happen to all its instances? Well it would make the most logical sense that they should all get deleted too, which is what CASCADE does, although there are other actions Django offers.

Our first database migration

Let’s run our first database migration! It’s very easy with our manage.py script.

Simply run the following commands:

python manage.py makemigrations
python manage.py migrate

You should get the following output signifying the database has been updated to include some new tables:

(venv) python manage.py makemigrations
Migrations for 'server':
  migrations\0002_entity_instancedentity_actor.py
    - Create model Entity
    - Create model InstancedEntity
    - Create model Actor

(venv) python manage.py migrate
Operations to perform:
  Apply all migrations: server
Running migrations:
  Applying server.0002_entity_instancedentity_actor... OK

Fixing the login and register logic

Now that we’ve re-defined our short-term goals, we need more than just a User when we register, so change the registration logic so we do the following just after we save the user to the database, but just before we send the Ok packet back to the client.

player_entity = models.Entity(name=username)
player_entity.save()
player_ientity = models.InstancedEntity(entity=player_entity, x=0, y=0)
player_ientity.save()
player = models.Actor(instanced_entity=player_ientity, user=user)
player.save()

Let’s fix the login logic too. We need to retrieve more than just the user model, and we don’t really care about holding on to the user model either. So let’s do the following right after we confirm the supplied username and password matches, and right before we set the state to the PLAY state:

user = models.User.objects.get(username=username)
self._actor = models.Actor.objects.get(user=user)

self.send_client(packet.OkPacket())
self.send_client(packet.ModelDataPacket(models.create_dict(self._actor)))

We will need to change the class member declaration in the constructor self._user to self._actor: models.Actor = None.

How do we send all this data?

As our models grow in complexity, you might wonder how is best to send all this data to the client.

The answer, I’ve found, is to pack a model and all its nested foreign relationships up into a big dictionary and send it over the network as a “Model Packet” (later on, we will discuss inefficiencies of this approach and a very neat trick for dealing with that).

In order to convert a model into dictionary, we need a function. Let’s define one now at the top of models.py:

from django.forms import model_to_dict

def create_dict(model: models.Model) -> dict:
    """
    Recursively creates a dictionary based on the supplied model and all its foreign relationships.
    """
    d: dict = model_to_dict(model)
    model_type: type = type(model)
    d["model_type"] = model_type.__name__

    if model_type == Actor:
        d["instanced_entity"] = create_dict(model.instanced_entity)
        # Purposefully don't include user information here.
    elif model_type == InstancedEntity:
        d["entity"] = create_dict(model.entity)
    
    return d

You can see we are leveraging one of Django’s included model_to_dict functions, but that unfortunately won’t recursively dive into each foreign relationship and nest its data into our dictionary for us.

For that, we need to have an if statement for each “composite” model in our game (e.g., Actor is a “composite” model because it is made up of an InstancedEntity and a User). On the other hand, User and Entity are examples of “atomic” models which don’t depend on anything else.

So basically this function works by first saving the atomic models to the overall dictionary, and then it goes through each composite type and recursively adds those to the main dictionary.

Note we don’t touch User information here because we don’t want to send usernames or passwords over the network (except initially from the client to the server).

An example of an actor dictionary would be

{
    "model_type": "actor",
    "instanced_entity": {
        "model_type": "instanced_entity",
        "x": 5.5,
        "y": 10.0,
        "entity": {
            "model_type": "entity",
            "name": "John the destroyer"
        }
    }
}

New packets!

Now that we have support for converting models to dictionaries, we need to actually have a packet for sending these models for the clients to process. We also need a way for the player to tell the server about its intended position target, so we need to add a Target packet.

Let’s go to server/packet.py and add the new packets. Add these new action to our Action enum:

ModelData = enum.auto()
Target = enum.auto()

Also add the packet definitions:

class ModelDataPacket(Packet):
    def __init__(self, model_data: dict):
        super().__init__(Action.ModelData, model_data)

class TargetPacket(Packet):
    def __init__(self, t_x: float, t_y: float):
        super().__init__(Action.Target, t_x, t_y)

If it isn’t already obvious, the ModelData packet is meant to send a dictionary containing all a model’s information to the client. The Target packet is meant to be sent by the client to the server to communicate the player’s next desired position (the payloads represent the coordinates of this position).

A new actor scene in Godot

We should be getting more comfortable in Godot by now, so I won’t provide as explicit instructions on how to do basic things from now on. If you do get stuck, though, remember you can always skip to the relevant section of the YouTube video.

In Godot, create a new scene called Actor.tscn. To this scene, add a Node2D root, rename it to Actor, and add the following other nodes until your scene tree looks like this:

You may get a warning about not having a collision shape yet, but don’t worry about that for now. We will bring collisions into the game later down the track.

Select your Sprite node and choose a texture from the right-hand side inspector panel. I chose the default icon.png that came with the project to start, but later I will show you have to make beautiful animated sprites.

Select your Label node, and drag its box in the 2D editor so it’s just sitting underneath your sprite. You can make it a bit wider to fit more text, and make it align Center in the insector.

Actor Scene

Attach a new script to your root node, unsurprisingly called Actor.gd. Clear out the default code, but we will not be writing new code here just yet.

We want our actors to inherit from a more fundamental Model class, so we need to create that first.

Create a new script called model.gd (not attached to anything), and paste the following code inside:

extends Node

var data: Dictionary = {}

func init(initial_data: Dictionary):
    update(initial_data)
    return self

func update(new_model: Dictionary):
    data = new_model

This is just setting up the functionality for receiving new models and updating them in Godot. Eventually, the update function will be changed a bit when we make it a bit more efficient, but for now it’s quite simple.

Now we can code our Actor.gd script:

extends "res://model.gd"

onready var body: KinematicBody2D = get_node("KinematicBody2D")
onready var label: Label = get_node("KinematicBody2D/Label")
onready var sprite: Sprite = get_node("KinematicBody2D/Sprite")

var server_position: Vector2
var actor_name: String
var velocity: Vector2 = Vector2.ZERO

var is_player: bool = false
var _player_target: Vector2

var speed: float = 70.0

func update(new_model: Dictionary):
    .update(new_model)
    
    var ientity = new_model["instanced_entity"]
    server_position = Vector2(float(ientity["x"]), float(ientity["y"]))
    actor_name = ientity["entity"]["name"]
    
    if label:
        label.text = actor_name

func _physics_process(delta):
    if not body:
        return
        
    var target: Vector2
    if is_player:
        target = _player_target
    elif server_position:
        target = server_position
        
    velocity = (target - body.position).normalized() * speed
    if (target - body.position).length() > 5:
        velocity = body.move_and_slide(velocity)

This script is a little more hefty as it needs to handle updates from the model and turn those into real, meaningful things about the actor, like velocity, position, name, etc.

We can see from the first line, this script extends from model.gd, so will have access to an init and update function. We are also overriding the update function, to additionally interpret some of the information in the model like the position and the actor’s name.

In the physics process, we are updating the velocity according to the current position of the kinematic body and the server’s reported position. For example, if the actor is at position (1, 1) and the server sends an updating saying it should be at (5, 5), then we set the client’s velocity to the vector with length 70, and direction pointing southeast. We keep the velocity here until eventually the client gets close enough to the target position, and then we set the velocity to zero. If the actor is the client’s player character (i.e., not another player), then we choose to follow the local player target position rather than obtain the target from the server. This is to ensure a smooth, responsive experience for the player with no latency. Of course, the server is still keeping track of the player’s position, and later we will see how to correct if the client goes too off course or tries hacking the game.

Player input

That being said, we do need a way to control our player and send our target position to the server, so add the following code to Main.gd:

func _input(event):
    if _player_actor and event.is_action_released("click"):
        var target = _player_actor.body.get_global_mouse_position()
        _player_actor._player_target = target
        var p: Packet = Packet.new("Target", [target.x, target.y])
        _network_client.send_packet(p)

This built-in _input function checks for input events each frame, so here we check if the player released the mouse button (or tap on mobile devices). Then we calculate the player’s intended position target and sends it as a Target packet to the server.

You will notice we are referencing a not-yet-defined _player_actor variable. For now, just add this to the top of Main.gd and we will define it later:

var _player_actor = null

We just need to add an input map for "click" to let Godot know we want to check for the left mouse button (which equates to a tap on mobile devices). In Godot at the top, click Project and then Project Settings. Click the Input Map tab, and type click in the Action input box. Click Add. Now scroll down to where it says click and click the plus button (+) beside it and select Mouse Button. Choose All Devices under Device (or mobile taps won’t work), ensure Left Button is selected and click Add. Now you can close the Project Settings window.

Sending updates to the client

Now we are ready to actually send model updates to the client.

Head on over the protocol.py and add some very simple new logic to the PLAY function:

elif p.action == packet.Action.ModelData:
    self.send_client(p)

elif p.action == packet.Action.Target:
    self._player_target = p.payloads

When our protocol receives a ModelData packet, we simply relay it back to our client for Godot to process. If we receive a Target packet, we just store it for now, and we’ll come back to it later. We should also add this to the __init__ constructor for the protocol just so we know what _player_target is:

self._player_target: list = None

Now we have all the functionality we need to send targets, process actor model updates, and display these in the game. The only thing we’re missing is the ability for the server to update actor positions each tick and broadcast model updates back to the other clients! To do this though, we will need some maths. Create a new file in server/ called utils.py which will serve as a place to throw all our random static functions in so they don’t clutter our more important logic in other files:

import math

def direction_to(current: list, target: list) -> list:
    "Return the vector with unit length pointing in the direction from current to target"
    if target == current:
        return [0, 0]
    
    n_x = target[0] - current[0]
    n_y = target[1] - current[1]

    length = math.dist(current, target)
    return [n_x / length, n_y / length]

This is a function which returns the direction pointing from one vector to another. The returned direction is also a vector with length one. This is useful for calculating the velocity of our actors in the next step.

Open protocol.py and add the following new function:

def _update_position(self) -> bool:
    "Attempt to update the actor's position and return true only if the position was changed"
    if not self._player_target:
        return False
    pos = [self._actor.instanced_entity.x, self._actor.instanced_entity.y]

    now: float = time.time()
    delta_time: float = 1 / self.factory.tickrate
    if self._last_delta_time_checked:
        delta_time = now - self._last_delta_time_checked
    self._last_delta_time_checked = now

    # Use delta time to calculate distance to travel this time
    dist: float = 70 * delta_time
    
    # Early exit if we are already within an acceptable distance of the target
    if math.dist(pos, self._player_target) < dist:
        return False
    
    # Update our model if we're not already close enough to the target
    d_x, d_y = utils.direction_to(pos, self._player_target)
    self._actor.instanced_entity.x += d_x * dist
    self._actor.instanced_entity.y += d_y * dist

    return True

This is the function which updates the position of our player based on its target. Note we are using delta time here (you will need to add self._last_delta_time_checked: float = None in the protocol.py constructor, and we will need to replace the value of 20 in __main__.py with a class member variable called tickrate).

We use delta time to ensure we are moving the correct amount each time this function is called, which is not every tick. Instead, we will call this function any time the packet queue is empty, so we don’t inundate the server with too many calculations if it should be processing more important things.

Some things to note: * The 70 in this file needs to match the player’s speed variable in Godot’s Actor.gd file otherwise the player and server will easily go out of sync. * We return True at the end to indicate we have actually moved the player. At other steps along the way, we return False if there was no movement possible/required. We can take advantage of this to decide whether or not we should re-broadcast our actor’s model later. * This should be the only function using the _last_delta_time_checked variable. If for some reason you need other functions with delta time capabilities, you will need to create other variables. * We are not saving the actor’s new position to the database yet. This will be the homework for this lesson we will discuss at the end 🙂

Finally, let’s call this function. As discussed, we don’t want to call this every tick, so let’s put it at the end of the tick function like this:

# To do when there are no packets to process
elif self._state == self.PLAY: 
    if self._update_position():
        self.broadcast(packet.ModelDataPacket(models.create_dict(self._actor)))

Receiving actor models from the server

So what happens when our client receives a ModelData packet? Currently, nothing, because we are not checking for it in our main state machine. Let’s fix that by opening Main.gd in Godot and adding this to our match statement for the PLAY function:

"ModelData":
    var model_data: Dictionary = p.payloads[0]
    _update_models(model_data)

The _update_models function will be a versatile function intended to update a model of any type passed in. Let’s define this in Main.gd as well:

func _update_models(model_data: Dictionary):
    """
    Runs a function with signature 
    `_update_x(model_id: int, model_data: Dictionary)` where `x` is the name 
    of a model (e.g. `_update_actor`).
    """
    print("Received model data: " + JSON.print(model_data))
    var model_id: int = model_data["id"]
    var func_name: String = "_update_" + model_data["model_type"].to_lower()
    var f: FuncRef = funcref(self, func_name)
    f.call_func(model_id, model_data)

So if I called, for example, _update_models(d) where d is a dictionary containing fields "id": 5 and "model_type": "Actor", then another function called update_actor(model_id, model_data) gets called. Of course, we need to define these functions, but this reflective helper function will really improve the readability and structure of our code once we add many more models. This is another example of a function which may look like overkill for the purpose of our demo, but once the game is extended, it will be a lifesaver.

With that being said, let’s define our update_actor function (also in Main.gd):

func _update_actor(model_id: int, model_data: Dictionary):
    # If this is an existing actor, just update them
    if model_id in _actors:
        _actors[model_id].update(model_data)

    # If this actor doesn't exist in the game yet, create them
    else:
        var new_actor
        
        if not _player_actor: 
            _player_actor = Actor.instance().init(model_data)
            _player_actor.is_player = true
            new_actor = _player_actor
        else:
            new_actor = Actor.instance().init(model_data)
        
        _actors[model_id] = new_actor
        add_child(new_actor)

So this is the function that gets called whenever we receive a ModelData packet containing an actor model’s information. First, we check if we have the actor’s model ID stored in our dictionary of actors (you will need to initialise this at the beginning of Main.gd: var _actors: Dictionary = {}). If we do already have this actor saved, we just call its update function.

If we don’t have this actor’s information stored already, first we need to determine whether it’s the main player or not. We determine this by checking for a variable called _player_actor, which should initially be set to null (so add var _player_actor = null to the beginning of Main.gd).

In either case, we add our new actor to our dictionary of known actors and add it to the game scene.

Let’s try it out!

Ok that should be it! It was a lot to get through, which seems to be a recurring theme with these lessons. I promise the next lesson will be more chill!

Run the server and connect a couple clients in Godot. You will want to register new accounts, because the old accounts won’t have any actors associated with them. Login and start moving around and see if you can see the other actors updating on your screen too!

You should also try using the chat function again to ensure that’s still working.

If any of this doesn’t work for you, and you’ve checked the server output, Godot’s debug log, and still can’t figure it out, I would recommend downloading the official source code for the game up to the end of this lesson and comparing your own code against it. Also don’t be shy to ask around by getting in touch in any of the ways mentioned at the end of this blog post.

There are a couple of things to note:

  1. When you log out and log in again, you will always just start again in the top-right corner.
  2. When a new player joins and starts moving, you will see them pop up right away, but new players won’t see you unless you move. In other words, if you are staying perfectly still when someone new joins, you will be invisible to them.
  3. In addition to point 2, an invisible actor who suddenly moves will appear to come out of the top-left corner and move to their proper position.

Homework

Your task is to fix problem #1 from above. A hint is to use Django’s model.save function somewhere in this code!

In addition to this, think about why problems #2 and #3 might be occurring. You don’t have to fix these (unless you really want to have a go!). It’s good to note that, fixing #2 and #3 requires fixing #1 first. So next lesson we will go through the homework solutions and start with fixing problems #2 and #3.

Conclusion

Thanks again for reading/watching! If you’ve managed to make it through this lesson, you should be seriously proud of yourself. This was a tough one, but as I mentioned earlier, the remainder of this series will be much more chilled out. We’re nearly there, so stay tuned for more!

Get in touch / connect with community

If you have any questions or feedback, I’d love to hear from you! Either drop a comment on the YouTube video, email me (my contact information is in the footer below), or join the Discord to chat with me and other students!