Godot looks great, but I want to separate game logic and display.
I like most things I see about Godot, and I'm going to try making some games with it.
Whenever I imagine programming a game though, I imagine the game logic and simulation being separate from the display. For instance, if I was to make a game like FTL, I would plan to simulate all the ship interactions and the movement of the characters purely in code, and then write a separate module to render that simulation. The simulation could be rendered with graphics, or with text, or whatever (of course, a text render wouldn't be human friendly, but could act as a dedicated server for some games, or I could use it for machine learning, etc).
I'm not an expert at Godot, but it seems this mindset is not going to fit well into Godot. Is this correct? It seems like the same object that is responsible for tracking the players health is going to also be responsible for drawing that player on the screen and tracking their location on the screen, etc. Will my player class have to end up being a subclass of some complicated Godot class? (Also, I'm a fan of functional programming and don't always use a lot of classes if given the choice.)
What are your thoughts about this. Would you recommend another engine? No other engine seem to be in the same sweet spot that Godot is currently in.
Pretty much every engine out there is built on strong opinions about “how game code should be structured” so while you almost always can bring your own opinions like this, you are likely to make life difficult for yourself…
It very much depends on what kind of game you’re making but I actually don’t think Godot is actually all that unsuitable for the sort of separation of concerns you’re going for, though you’re unlikely to find many tutorials geared towards that style of coding.
I say this as someone with similar instincts about how to structure my code, who spend 2-3 years working against the grain in UE4 to maintain a clear model/view separation in a 2D simultaneous-turn-based strategy game & am currently going down a similar route with Godot.
You can use plain Node & Resource derived classes for a lot of your model-side stuff & only use “things that automatically draw themselves” nodes as a layer on top of that if needs be… but yeah, it very much depends on what kind of game you’re making. For most traditional real-time games (as opposed to turn-based ones etc), especially anything 3D or where you want to use in-built physics stuff, I’d lean towards trying to work with the engine as much as possible, rather than being too fanatical about following a particular design pattern that isn’t fully encouraged by the engine… at least for your first game or two!
I think it is entirely possible in Godot. You don't have to extend any Godot specific class when you create new objects and scripts, just define a new class, and put your game logic there. Then, you can use an instance of that class in your rendering specific logic.
If you want, you can take a look at my pet project, a match three "game", I made it to learn Godot. I had the same concerns as you, so I tried to separate game logic and animation. What I did was to create separate queue for the animations based on game logic steps.
This solution is "ugly" from a programming theory perspective, but it works, especially for a single-player game. You have a global Singleton class that has all the game state information. In the GUI part, you have it check the global Singleton when deciding that to draw.
If you do it this way, you can put the game logic in another thread, or even use a C/C++ GDExtension if you want really high performance.
For example, you keep the player's HP in a global singleton. Then in the GUI where you show the player's HP, you have it check the value in the global singleton.
It seems you want something that can have a good separation of concerns. Maybe something like Bevy? Its using ECS, so you can customize your logic and separation between it however you want.
I'm excited about Bevy and watching it closely (and donating, because I can), but I don't think it has the level of polish that Godot has.
If my goal is really to create a game, then I have a hard to justifying Bevy right now. The famous question being: Do you want to build a game, or do you want to build an engine? It's similar to that. If I start going into the weeds to do text wrapping in Bevy (or similar), then I'm working on an engine more than I am a game.
One thing you could do is to use resources and one or two global scripts to run simulations like I do in my game project (Very heavy calculus and material simulations). There is going to be some interconnectedness with nodes, but that is a strength, not a weakness in my opinion!
It sounds like the thing you want is Functional Reactive Programming (FRP), where you have a state value that represents the current state, an update function which integrates an input and the current state, and a display function which just presents the state for the user. This is not Godot, but you may be able to search for an FRP game engine and find something more than a half-built weekend project from 8 years ago 😅
Godot is not a function-oriented environment at all, though the version of gdscript in Godot 4 does have first-class functions finally. Their entire model is very object-oriented, though. So your player will 100% be a subclass of a Godot class.
That having been said, more than many OOP projects Godot encourages object composition over multiclassing and mega objects. The "right" way to build a character in Godot is to have a health component that just holds health, takes damage, and heals. Then another which does damage to thing, etc. Then the player is a subclass of the physics body, and does have some glue code, but all of the concerns just get added to it and it becomes the sum of its parts. That allows the health component to be added to enemies and trees as well without rewriting it, etc.
FRP is interesting. I could connect a FRP system to Godot's rendering server, as suggested in another comment.
Also, I think there is push-based and pull-based FRP. Pull-based would probably work better, the FRP updates would only calculate as new values are pulled out of the system by the surrounding Godot code.
Overall though, FRP seems like overkill (I'll still consider it), but its moving into building-an-engine territory rather than building-a-game.