#bevy_mod_scripting

2116 messages · Page 3 of 3 (latest)

ornate pond
#

okay, nah

#

I figured out how to invoke a callback, waited for the "reached final state", then got the same error on add_system

#

at least I know register_callback is working

ornate pond
#

now is this of any relevance? I tried printing the system names in the part where it looks for the created system and got this

DEBUG bevy_mod_scripting_core::script_system: Found a system with name "<Enable the debug feature to see the name>"
DEBUG bevy_mod_scripting_core::script_system: Found a system with name "<Enable the debug feature to see the name>"
DEBUG bevy_mod_scripting_core::script_system: Found a system with name "<Enable the debug feature to see the name>"
#

at crates/bevy_mod_scripting_core/src/script_system.rs:173 to be clear

ornate pond
#

so I guess that was it? it worked after enabling the debug feature

#

although for some reason it can't find the Update schedule

#

but sure enough I got a bunch of prints from lua after setting it to PostUpdate

#

oh and it doesn't ever run if I set to Startup or PostStartup, I guess it's too late at that point?

#

even though I'm not calling it from the callback anymore

old trail
#

Ohh

#

That sounds right

#

Damn, that must have started happening last bevy update

#

Ill have a look at fixing tonight

amber glen
#

Hey, I'm using luajit and am noticing that the performance is really bad. Is this expected? I'm gonna try changing to a different Lua version in a bit and I'll report back.

local N = 10000

-- 1) Perf.noop() calls: 9.61ms
for i = 1, N do
    Perf.noop()
end

-- 2) Pure Lua empty loop: 9.294ms
local x = 0
for i = 1, N do
    x = x + 1
end

LuaJIT seems to be setup:

LuaJIT 2.1.ROLLING — JIT ON, arch=arm64, os=OSX
JIT flags: fold cse dce fwd dse narrow loop abc sink fuse
amber glen
#

Changing to Lua54 seems to have fixed the issue.

-- 1) Perf.noop() calls: 0.301ms
-- 2) Pure Lua empty loop: 0.059ms
old trail
old trail
#

that big of a difference happening even in pure lua code is strange

#

what happens if you increase the N value to something like 1e7 ?

amber glen
#

Perf.noop was a noop rust binding

amber glen
#

I did more debugging, and it seemed like the JIT compiler was failing all traces. I've unfortunately moved on and don't really remember the specifics

ornate pond
# old trail This should fix this: https://github.com/makspll/bevy_mod_scripting/pull/531

first impression 😅

TRACE bevy_mod_scripting_lua: ~/.local/share/cargo/git/checkouts/bevy_mod_scripting-5f706eeba1b2b94f/f6ac743/crates/languages/bevy_mod_scripting_lua/src/lib.rs:345: Context EntityScript(entity: 24v0, script: path scripts/main.lua) is not subscribed to callback Enablethedebugfeaturetoseethename

and my registered PostUpdate system doesn't run

#

BUT

#

I'm trying to figure out why enabling the debug feature actually makes it work

#

it seems the actual error that causes this comes from mlua?
FromLuaConversionError { from: "nil", to: "function", message: None }

#

oh nvm it's probably trying to find a function called "Enablethedebugfeaturetoseethename" in the script's globals

#

yep hilarious

function Enablethedebugfeaturetoseethename(time)
    print(string.format("[SCRYPT SYSTEM] update system ran, dt=%f", time.delta_secs))
end
old trail
#

oh did I miss a reference to the system name somewhere

#

ah yeah one sec

#

yeah basically, bevy doesn't really expose the SystemKey easilly in this area, the callback name it uses is based on the name of the system, which was using DebugName.. not a fan of that struct being such a footgun!

#

anyhow should be fixed now

ornate pond
#

yeap, working beautifully

ornate pond
#

I think there may be some other semi-random inconsistencies, maybe I missed a feature flag or something?

like, for some reason I'm now able to add the system to regular ol Update, but no longer to FixedUpdate

I can get types["ButtonInput<GamepadButton>"], but not types["ButtonInput<KeyCode>"]

old trail
#

As for schedules, this will depend on where youre adding the system, like you cant add systems to FixedUpdate if you are currently running in FixedUpdate

vast dock
#

Things have progressed.

After lua actions, scenarios, I now have selection based lua actions:

Apr-08-2026_21-02-30.gif
vast dock
#

Is there a way to make lua error panic instead of a warning in the logs? I was thinking of using that for testing purposes.

old trail
vast dock
#

Ah i'll look into that!

vast dock
#

@old trail That MR you just did is a godsend ahahaha

#

I was getting back into some save/loading stuff and fighting with timing again...

old trail
#

Nice! Theres at least one more coming related to the pipeline

#

I am discovering stuff as i am building my proof of concept

#

Turns out using the thing yourself gives you perspective haha

vast dock
#

Oh, you've been working on your demo game thing?

vast dock
#

Auto loading lua scenarios script on launch using command lines (I'll need that for dedicated servers and other things later)

Apr-15-2026_22-31-15.gif
vast dock
vast dock
#

Well, I just did and now I need to create those lad files and definitions....

old trail
#

I am curious what people think about ScriptComponent how does the fact that it stores multiple scripts in a list play with the ECS for you ? I was considering re-thinking that and having each script live in a child entity of the target entity, to play nicer with change detection

vast dock
vast dock
#

I am planning in inspiring myself from the factorio modding scene for structure, distribution and version control.

vast dock
old trail
old trail
vast dock
#

Or the question might be, is it expensive to maintain it in that way now that it's set up?

vast dock
old trail
#

I am mostly just curious if people generally use the vec or create children entities etc, not doing anything with it at the moment, but generally BMS is already massive so any simplifications are always good

urban talon
# old trail I am curious what people think about `ScriptComponent` how does the fact that it...

Hmm, I want to use the multiple script handles, and I've added PRs to ensure that they're loaded in the order they're given, so that's definitely something I'd want to preserve; however, despite me wanting to do this, when I've tried it, it hasn't worked on whim, so I've fell back on my old concatenation code. I'm sure with some effort I could make it work, but last time I was in the midst of a gamejam, so it wasn' t the right time to try.

old trail
#

Thanks for the input! That helps!

pale fern
#

I've noticed that in 0.19, global functions and generic types don't get translated into the lua declaration file. I can understand the generic one, but why are the functions ignored?

old trail
#

I think they are also experimental in LLS

old trail
#

Writing my own game is giving me a lot of ideas, one thing I am noticing is that performance seems to be splendid so far. For example currently I am spawning a static script per spell type in the game, and that spell receives events like on_spell_spawned(entity) on_spell_collision(entity, otherEntity), and with that setup, I was able to spawn hundreds of fully physics simulated projectiles with lua running on those hooks, setting sprites, emitting sounds etc, and the framerate was not getting hit.

Now I am not sure how a VM per spell will scale, but there is always the possibility of sharing contexts between those and using stored callbacks instead.

#

I am thinking about expanding event callbacks to support more pipelining style workflows, for example on_player_input might benefit from explicit ordering between scripts as well as the option to stop further scripts from processing the callback based on return value

#

I think it might be easy to start tanking the frame rate if you start running things per update across many entities with non trivial amounts of ECS interaction. Although I don't know where that limit is

vast dock
#

I would tend to run things on a message or a event to limit execution ? Or does all script run always every frame ?

old trail
#

Nah scripts purely execute if you send a message + on attach detach

vast dock
#

Makes sense !

So you would add more native conditional triggers ?

old trail
#

kinda yeah, for example CallbackLabel might also specify some settings, like ExplicitOrderingEnabled , and ScriptCallbackEvent might have priority set on it, and that way for some callbacks all events will be sorted first. On top of that you could have escape hatches, maybe something like world.stop_processing_attachment(), or just a special return value, that tells the engine to stop handling any more events, either for the specific attachment or say for all events containing an entity etc.

#

that way you could say build a player controller, with default ordering priority, and then modders could insert something with higher priority and say modify either the payload itself or some state prior to it reaching your controller, or straight up fully override it

vast dock
#

I haven't played with that part of the library much so far.

But it does make me think of what they do in BEI

#

Maybe a BEI based example would avoid developing a whole section of library ?

old trail
#

what's BEI ?

brave garden
old trail
#

Ah I see

old trail
#

this year is the year of ergonomics

#

Thoughts on making ScriptCallbackResponseEvent observable ?

vast dock
#

For me this would be useful when saving:

  • I ask lua to send me the script values that I need to save
  • I then the save lua calls a rust function that indicates it's finished
  • Then I can write to file.
#

Having script end event would allow me to not have all this mechanism anymore.

#

It's a bit the same for load:

  • I send the save data to lua
  • I need to wait until it has been executed
  • I can then resume game
#

Ha... Going through your stuff I just realise I can already expose the event...

molten pine
#

@old trail Any chance to see some kind of a demo of arcane_assembly at some point? 😍

old trail
#

Is there still a scriptcallbackresponse when the function is void?

Yeah, you configure if the response exists per callback event, but whatever ScriptValue is returned will be present on the response

#

@Gedes Any chance to see some kind of a demo of arcane_assembly at some point? 😍
Aye would love to, atm there is not that much going on, although it's probably enough to showcase a basic modding setup:

  • Mods get a top level script, either attached to the player or as a static script
  • Spell components all share a context, and use registered callbacks to listen to events like on_cast on_expire
  • Spells are parsed from dotgraphs into a DFS graph the spell executor runs in that order

I've got a small bevy_hanabi particle binding layer as well which is pretty neat:

vast dock
molten pine
#

Wow.

vast dock
#

I've been fighting with some flaky tests loading scripts in my environment for a while and i'm now trying to tackle them to get my code coverage up.
Run time never fails and always load all 3 scripts correctly.
However in test sometimes I load 0, 1 or 2 of my files depending on my PC's mood despite running app.update() a whole bunch.

Would you have an idea why this would fail sometimes? I read the files correctly, I get the correct relative path, but then my scripts do not get attached. Or they don't run? Unsure.

/// Resource tracking all loaded scripts
#[derive(Debug, Resource, Default)]
pub struct LoadedScripts(pub Vec<Handle<ScriptAsset>>);

/// Initial setup
fn setup(
    mut commands: Commands,
    asset_server: Res<AssetServer>,
    mut loaded_scripts: ResMut<LoadedScripts>,
) {
    let script_dir = Path::new("assets/lua/GMActions");

    let entries = fs::read_dir(script_dir).unwrap();

    for entry in entries.flatten() {
        let path = entry.path();

        // Accept only `.lua` and `.luau` files
        if let Some(ext) = path.extension().and_then(|e| e.to_str())
            && (ext == "lua" || ext == "luau")
            && let Some(relative_path) = path.strip_prefix("assets").ok().and_then(|p| p.to_str())
        {
            let relative_path = relative_path.to_string();
            let handle = asset_server.load(relative_path);
            loaded_scripts.0.push(handle.clone());

            commands.queue(AttachScript::<LuaScriptingPlugin>::new(
                ScriptAttachment::StaticScript(handle),
            ));
        }
    }
}

I suspect that something is up with these lines when in a test configuration and/or bevy's asset server.

let handle = asset_server.load(relative_path);
loaded_scripts.0.push(handle.clone());

commands.queue(AttachScript::<LuaScriptingPlugin>::new(
    ScriptAttachment::StaticScript(handle),
));

Somehow the script I attach are not run, only sometimes. It's very odd...

old trail
#

Hmm, can you see if theres any clues if you run with trace log level for bms core?

vast dock
#

In this example my polygon GM action is never executed for instance.
That file looks like this:

local Template = require("./../library/spawnTemplate")
local simple = require("./../templates/FirstTemplates")

local Types = require("./../library/GMActions")
type GMActionPoints = Types.GMActionPoints

local gm_action_name = "polygon"
register_gm_action(gm_action_name, "Polygon", false, "Digit3")

function on_gm_action(name: string, template: string, data: GMActionPoints, scroll_value: number)
  -- do stuff
end
#

Normally register_gm_action is called and therefore bevy knows about the polygon game master action. And that always runs flawlessly during runtime.

old trail
#

How do you trigger thay callback?

vast dock
#

Sometimes all of the luau files will be executed, sometimes none, somtimes a few.

#

But perhaps my structure is not great and I need to rework it...

old trail
#

I can see its running in the logs:

01T20:05:12.480896Z TRACE bevy_mod_scripting_lua: Context StaticScript(script: path lua/GMActions/polygon.luau) is not subscribed to callback on_script_loaded

#

Hmm

#

If you add a print do you see it?

vast dock
#

Here is another example of failure where the print did not occur

📎 message.txt
#

The print itself is print("TEST_PRINT")

#

And I have a override to make it print in bevy.

#

(I tried removing the override also and i have the same results)

#

One thing that I am suspecting is that maybe something is missing in my bevy test setup that messes with BMS?

old trail
#

Are you potentially queuing callbacks in the same frame youre attaching scripts or is this purely relying on the on load code?

vast dock
#

I have two different category of static scripts:

  • Gm actions will load and stay loaded, they are never called by bevy in any way before the failure they are just run.
  • Scenario will load, wait X amount of frames for the load to complete and then unload.

In the case they successfully load and execute the register functions, I then start sending them callback through bevy.
For the gm actions, I do not send anything to them until the test runs (it always succeeds once the load is successful).
For the scenario i shouldn't either. I do send fixed updates when they are running.

vast dock
#

There we go, networked lua script actions using remote procedure calls.

May-21-2026_18-17-29.gif
grim jetty
#

Hi folks !
Really enjoying bevy_mod_scripting so far ! Have been able to figure out & solve most issues so far , except this one :
How can values insideOption<>s / Results be retrieved inside Lua scripts ?
Specifically: I have this component

#[derive(Clone, Default, Component, Reflect)]
pub(crate) struct TemperatureSensor {
    pub(crate) temperature: Option<f32>,
}

I can see that the value is displayed as TemperatureSensor { temperature: Some(-270.0) } inside lua scripts, but I need to actually get the value itself.
Thanks in advance !