Hi there,

i have successfully implemented signed distance fields font creation and rendering and now i am in the planing process for my HUD/UI rendering right now. Normally i would just add a renderHUD() method after all the other rendering stuff and call it and use orthographic projection. Thats how i have done it in the past, but now i have a Entity Component Based Game Engine and i want to do this the "Entity->Component" way if possible.

What i want to achieve is:

- Rendering a bit of text based on triggers (which are components as well) in the 3D scene (Created and destroyed on the fly)

- Rendering simple game Menu UI (Start, Load, Save, Options, Credits, etc.)

- Rendering HUD for the player/game and scene stats

- Rendering tons of debug informations - if required

Do you guys have some tips for a good component based solution?

Thanks in advance,

Finalspace

I faced a similar problem so here's my perspective, though you might not like it.

Not everything has to be framed in the context of an ECS. It's the same problem I see with developers using the MVVM pattern. I myself have at times become so obsessive that everything had to follow that pattern, to the point that things that would have taken 10 seconds to implement in the codebehind required complex XAML-only solutions.

I am building this UI library, which I am using together with my ECS framework. The solution I adopted was to treat the whole HUD as an entity. Therefore the UI system has the only responsibilities of polling input devices and rendering the whole HUD (though I might split those eventually). I assume that you wanted an approach in which every UI element was a different entity? Or something along the lines of having different UI-components, such as a "ButtonComponent", "PanelComponent", "LabelComponent" and so on?

The only advantage I can see is that you would then be able to attach UI elements to 3D objects. The downside is that you'll have to find someway to have different UI-components communicate with each other.

I agree with AvengerDr. Think really hard if you want to shove the UI rendering block into an ECS shaped hole. The ECS concept relates well to game objects, and afaik, the HUD is not a game object. You might try hard to represent it as one, but it really would end up like trying to push a cube through a round hole. ECS is just another tool in your toolbox, so use it where appropriate.

Do things the easy way. Do things the way it makes sense.

I agree that not everything must be done in a ECS-way, but i really like the way how clean it is with all the systems, entities and components.

You have the update systems (like input, behavior, camera, physics engine - in that order) first and then you have the draw systems and every system encapsulates the logic or drawing which are really clean in my opinion. Currently i havent found anything yet which does not fit in the ECS - but these will come definitly i think.

But i will give it more though, to get a simple but good solution - for now i need to fix the damn letter advancement -.-

The downside is that you'll have to find someway to have different UI-components communicate with each other.

Components should already be able to communicate with other components.

I actually think the HUD can easily be made with ECS and even fit with your old way of doing things, you should already have an image component and image render component a UI is just behaviors with those so: a clickable component and system that sends a message to all components in the entity when entered, exited, mouse up, mouse down then a uicomponent with it's system that ties together multiple components in the same group (Like radio buttons).

Components should already be able to communicate with other components.

I actually think the HUD can easily be made with ECS and even fit with your old way of doing things, you should already have an image component and image render component a UI is just behaviors with those so: a clickable component and system that sends a message to all components in the entity when entered, exited, mouse up, mouse down then a uicomponent with it's system that ties together multiple components in the same group (Like radio buttons).

Yes sure, but it would be more time-consuming to have them behave the way you want compared to using an explicit approach. You can't throw a ButtonComponent, a PanelComponent and LabelComponents in an Entity and expect a DropDownBox to appear and work like it should unless you coordinate the logic between all of them. I believe that's easier to do if you have a DropDownBox object where you can do everything you need. After all none of the major UI frameworks use anything even remotely similar to an ECS.

In my approach, I only have a UIComponent for the whole HUD. So it gives me the flexibility to switch HUDs at runtime and that's enough I guess. Further, in my approach I am using Direct2D to render controls from scratch. So they aren't simple textures and thus require a different rendering method than if they were billboards.