archived-code-advanced

[pseudo onyx]

Also, i dont understand why if the dot product is greater than 0, it means the character is facing left, and if the dot product is less than 0, it means the character is facing right

[sly grove]

IDK seems really overcomplicated. I would have just done this:
private int FacingSign => (int)Mathf.Sign(transform.forward.z);

[pseudo onyx]

What does this code do?

[sly grove]

it checks if you're facing forward or back

if you want left/right then it's private int FacingSign => (int)Mathf.Sign(transform.forward.x);

[pseudo onyx]

ok thanks

[pseudo onyx]

When I use this line of code:
private int FacingSign => return (int) Mathf.Sign(transform.forward.x)
I get the following error "The type or namespace transform could not be found"
'Cowboy.Sign(transform.forward.x)' must declare a body because it is not marked abstract, extern, or partial

[sly grove]

1. the other code you had also used transform so I can only assume you didn't put this code in a MonoBehaviour if it's complaining about that.
2. get rid of return (that typo was my bad)

also you're missing a semicolon... not seeming very #archived-code-advanced right now tbh

[pseudo onyx]

Ok thanks

[strange blade]

Im sorry if this isnt in the right section, I wasnt sure where this would go. Im trying to enter play mode faster, and Ive already messed with the enter play mode settings. My scene has a fairly large object pool, but other than that its basically empty.

Why is unity calling SaveScene() everytime I enter playmode? I already saved my game, I dont need it to do that. Can I turn this off? I have failed to find any information about how to control this or why its even called when I press play.

[brisk spruce]

I think it's gotta save so the scene editor is synced with playmode

[strange blade]

I dont think I understand why it would need to sync the two though, as no changes are really made during play mode and everything should be pre-set for testing via the editor

[brisk spruce]

I don't think saving a non-dirty project should have any impact either though

[strange blade]

Im hoping im wrong with this, but I ran into an answer on Unity:
https://issuetracker.unity3d.com/issues/editor-unselects-elements-when-entering-and-then-exiting-the-play-mode

One of these edge cases is with prefab instances, which were not getting fully backed up to disk even though they were dirty. This mistake was causing them to be different when coming back out of playmode, and also causing their selected state to not be properly restored.

Unfortunately, for historical design reasons that we are now somewhat stuck with, this optimization cannot work robustly, so we have decided to remove it.```

If im understanding this correctly, it sounds like.. I cant avoid it? As unity used to actual have a ignore save button..

[sleek terrace]

anyone know why my shader produces no color whatsoever? It seems makes the entire image invisible

Shader "Unlit/test" {
    Properties{
        _MainTex("Main Texture", 2D) = "white" {}
    }

        SubShader
        {
            Pass {
        
                CGPROGRAM
                #pragma vertex vert
                #pragma fragment frag

                #include "UnityCG.cginc"

                struct appdata {
                    float4 vertex : POSITION;
                    float2 uv : TEXCOORD0;
                };

                struct v2f {
                    float2 uv : TEXCOORD0;
                    float4 vertex : SV_POSITION;
                };

                sampler2D _MainTex;

                v2f vert(appdata v) {
                    v2f o;
                    o.vertex = UnityObjectToClipPos(v.vertex);
                    o.uv = v.uv;
                    return o;
                }

                fixed4 frag(v2f i) : SV_Target {
                    fixed4 col = tex2D(_MainTex, i.uv);
                    return col;
                }

                ENDCG
            }
    }
}

[grave fossil]

https://discord.com/channels/489222168727519232/1128942517023416370
Anyone have experience with downloading ASTC or DTX files instead of png?

why texture = ((DownloadHandlerTexture)request.downloadHandler)?.texture; returns me null even if data is not null and it is correct ASTC image?

[hardy sentinel]

nope, looks alright..

[sleek terrace]

I ended up using shader graph but I'm still not sure why that shader makes the tile map transparent

[gleaming rock]

How should I do packet delta compression when player position data is sent to clients via UDP?

So far I am thinking of creating custom packet for each client per tick of all player positions (delta compressed)

But that is O(n^2) operation and with just 32 players would result in 1024 iterations per tick (timed and on my pc it is about 2ms per tick)

Currently my code is something like (called every tick and is Unreliable packet)

void SendPositionUpdateOfAllPlayers()
{
  // create message
  // add every players' position to the message
  // send the message to all clients
}

This works fine except I don't want to send non changing state of players or projectiles.

Should I use job system with burst for this or would it even work with riptide (to build messages in jobs)

Or maybe send states as reliable packet every second and send deltas relative to that (simple approach but doesn't yield as great bandwidht savings)?

[azure moss]

whats the best way to make manual animator transition?

think of it transition without condition in animator. animator will play animation automatically after end reached or sometime before it

[azure moss]

how about sending only and only when needed

you might have a situation where player isnt in your view

server or p2p implementation?

[gleaming rock]

Dedicated server

[plush hare]

Each input packet, the client should tell the server the last state he received

and the server computes the delta based off of that.

You'll need to variable width encode your values, and quantize your floating points before packing the data together.

You can use a bit flag to denote whether a state has changed or not.

that's about all there is to it.

[gleaming rock]

@plush hare thanks. However what im worried about is iterating players with O(n^2) complexity per tick to build the delta compressed packets

[plush hare]

Test it in a build. Not in the editor.

Your game server should only be running one core anyways, so jobs is of no use

if you're lucky, burst will SIMD your stuff. Otherwise you'll be writing your own AVX2 parallelized code

[gleaming rock]

I tested on il2cpp development build on windows

2ms per 1024 iterations

[plush hare]

yeah thats a lot

sounds like you're doing something wrong

[gleaming rock]

Assuming 32 players

[plush hare]

thats too much time for 32 players. you're making a mistake somewhere

[gleaming rock]

Alright I will debug

[plush hare]

actually depending on the CPU, that might be a reasonable time frame. It's 4 in the morning and my brain isn't working rn

30hz should give you about 33ms per tick

again, your server should only be running one core per instance. Sometimes it hurts performance to try to parallelize everything on to different cores

[gleaming rock]

Yeah I was hoping that burst could optimize the nested loops

[plush hare]

The only real recommendation I can give right now is to just make sure you've got nice linear data

minimize cache misses

make it nice and predictable for the hardware prefetcher

[gleaming rock]

Okok

[tropic stag]

posted in the vfx channel, but maybe someone here will recognize what it might be? #✨┃vfx-and-particles message

[gleaming rock]

@plush hare Appreciate the help. Not sure what I was doing earlier, now I'm getting now about 0,15ms per tick

[wet burrow]

What's the best place to talk about packages?
I created a custom package that I can download from GitHub into my project using the package manager (which I'm super pleased with) but I want to make custom classes in my actual project that inherit from one of the base classes in the package, yet it won't recognise the namespace so I can't reference those scripts in my project.
Thoughts on how to deal with this?

[flint sage]

You got to create an AsmDef and reference the one in your package

[dusty wigeon]

Stop cross posting...

[wet burrow]

I'll check that later. I'm sure I did that and it didn't work, but I might have misremembered

[flint sage]

Then your class has to be private otherwise it would work

[tropic stag]

coming from a programming background I spent way too much time trying to figure out a good way to do something instead of just making a game... I implemented a state machine using components subclassing monobehaviour, and when switching states I just set enabled = false on the prev state, and enabled = true on the new one. Instead of abstract state enter and state exit behaviours I just use onenabled and ondisabled.. is tnere anything wrong with this approach?

[flint sage]

If it works, it works

[plucky laurel]

the only wrong thing about it is that if your fsm is nested it will receive its parents OnEnable/OnDisable

to it wont have a true "state enter" event

[sly grove]

A simple switch statement and an enum is more lightweight. Unless your state machine is complex it might not be worthwhile to build out a class structure like that, and will probably be less performant.

[brisk spruce]

If by on-x you mean using event subroutines, they can be tricky and extra work if you want your events to play nice with a Task based Async architecture (the default eventargs stuff doesn't/can't run async)

Forces you to make your own type of async event args and async subroutine to bind to, it's a pain

[white pulsar]

i'm currently implementing A* on a grid based aproach. currently my system obtains the nearest available node from the start position and the world position by converting the world position to a grid cell

However, my code is a bit faulty, as it seems to be shifted a bit, i'm unsure what's causing the shifting itself
https://hastebin.com/share/jeturohuti.java

the hastebin itself contains what i've done so far, sure its a bit cut down to avoid sharing the entire class, but it should contain the general idea for it.

the issue is basically that the GetNodeIndexFromPos is nnnot really working as i expect it to, considering the fact that its shifted as the picture above shows

apologies if this isnt precisely the correct place to ask, i've never triedd to implement A* until recently and it seems a complex topic :b

[brisk spruce]

Could you circle a "expected" vs "actual" in that pic

[plain abyss]

I have a 3D environment made out of runtime-generated thin-walled meshes, and I want an orthographic overhead view to display those walls as lines. I have a line renderer as a child of the meshes set to have points along the vertices of the mesh, set to only be visible to the overhead camera. This allows the overhead camera to see the "walls" with the line renderer despite them being vertical planes (First image shows this working as expected). This works fine for simple shapes like squares, but for more complicated shapes it forms a lattice of lines across the vertices (Second image). I want to only outline the "outside" of the mesh, but I can't use bounds because some of the meshes are angled or circular and a box is not always the best fit. How could I determine the outermost edge of vertices in a mesh to pass to the line renderer, or, barring that, is there another way I can display lines where flat vertical planes would be?

[white pulsar]

Will do, I'll just make the start and end gizmos the same color as the start and end transforms

[brisk spruce]

Is it possible to hook into your runtime generated logic to keep track of walls at that point?

Or is it 3rd party?

[plain abyss]

They're generated in an entirely different program, saved as GLTF meshes, streamed to the project and loaded at runtime. I could make an additional mesh that just has the outlines, but that would be more data that has to pass between them

[brisk spruce]

Prolly would be better performance at the cost of game size.

Aside from that, can you get the normals of the lines? A vertical polygon has a normal in the horizontal plane right, it's vert component should be zero

[white pulsar]

@lethal flint
https://i.gyazo.com/176310eb79bfc093e71d38aeeeed8c3f.mp4
here you go, the red represents the start, black represents the path while green represents the destination

here's a still pic with the grid itself showing, the white wireframes being the available grids for the algorithm

[brisk spruce]

Looks like global vs local position issue, do they belong to a parent object perchance?

[white pulsar]

no, all objects are at the root of the hierarchy

[brisk spruce]

Or the uh, whole grid object parent

[plain abyss]

Using the normal could work, it'd at least be better, but there might still be some internal geometry I don't need

[white pulsar]

sorry, the GameObject AStar has both components, the AStarNodeGrid and the Pathfinding System

while the End and Start game objects are the ones i'm using for testing purposes

doing a search of localPosition on my project doesnt reveal anything, so i doubt it's that specifically, might be with the method i wrote that should transform the world position to node index

[torpid smelt]

Alright, quick one. is there any way to see what the full function name is there? this is in a profiler window which is zoomed in all the way, and I need to know what's getting called here

naturally I can't zoom in any more

[sly grove]

you can switch to hierarchy view instead of the flame graph

[sage radish]

Does clicking it help? Or hovering over it to show a tooltip?

[torpid smelt]

No & No, unfortunately. Gonna see if I can find it now in the hierachy

[sage radish]

I think you can right click and show it in the hierarchy, or select it and press a button somewhere

[torpid smelt]

Hierachy doesn't have it, despite it being there in the Timeline view.

[torpid smelt]

Right clicking is a solid nada here, not sure if I need to toggle something to allow that or not. this is in the standalone profiler.

[sage radish]

What Unity version are you on?

[torpid smelt]

2020.3.21f1, coworker found it on a previous frame. thanks for the help

fucking photon

[sleek terrace]

Does anyone know why my the default custom unlit shader isnt working with tilemaps?

[sleek terrace]

figured it out, it seems by default tile maps are facing backwards, thus you have to set shaders setting "Cull" to off

[sleek terrace]

How is that even possible?

Im generating at most 25 objects

Funny, but every time I instantiate the object, its instantiates itself...causing the start to be called, causing more instantiating...

🙂

[shadow seal]

Ah, recursion

[sleek terrace]

wait, that shouldnt occur tho

because, Im not instantiating the hexgridmap object

If I instantiated the Hex shape in HexGridMap, should the instantiate call the Awake and Start methods in Hexgrid map?

[sturdy frost]

how IJobEntity get nativeThreadIndex

[sleek terrace]

If the hex shape and hex grid map are in thesame object, instantiating the hex shape will also instantiate all components associated with the game object, just recalling the awake methods in said componenets

[untold moth]

Probably passed to it when the job is started on one of the threads.

[warped pasture]

Hey, is there a way to draw a batch of Sprites in a performant way?
I need to have some graphical image between each tile division in a hexagonal tilemap, some sort of "wall", that may or may not appear depending on some gameplay rules.
I know this kinda sound vague but I looked at Graphics.DrawTexture but it doesn't seem to work on Update or take camera into account and I need something that can be zoomed in and out too. Did some research around and couldn't find a satisfying solution and I'm worried about spawning a hundred sprite renderers in a scene for a WebGL game.

[long ivy]

are you doing something that breaks batching? That would be the easiest solution. If your image uses the same material and texture (or textures are packed into the same atlas), you can manipulate sprite geometry directly to avoid extra renderers

[warped pasture]

No i'm doing nothing out of the ordinary. Just read this and yeah, it's gonna be the same image and material, so apparently is fine.
https://docs.unity3d.com/Manual/DrawCallBatching.html

[storm tartan]

Hi! I am searching for a tutorial to make a dynamic blocking animation system for melee combat. I want the hand and the sword to be positioned according to the collision point, by using Two Bone IK Constraint and other dynamic bone components. Any recommendation or info about this context?

[timber shadow]

Where do people go to talk about unity's localization package?

[spare violet]

Unity forum has a dedicated localization section. One of the Unity devs is quite active there.

[urban warren]

Right, so I am doing some reflection for a editor tool, and running in to an issue with events.
There is an event that takes a internal class as a param. I need a property from that param class, and I need to be able to subscribe and unsubscibe. So I can't just use a lambda to 'convert' from one to another.

internal class UnityParameter
{
  public int foo;
}

public class UnityClass
{
  public void Register(Action<UnityParameter> action);
  public void Unregister(Action<UnityParameter> action);
}

This is the method I want to add to pass as a parameter to the Register and Unrigster methods.

public void MyCallback(int bar);

Any ideas?

All that is coming to mind for me is to have a 'converter' method. And a static dictionary of Dictionary<UnityClass, Action<int>> which the converter method calls maybe. And I subscibe MyCallback to the action in the dictionary.

That feels really ugly, so I was hoping there would be a nicer way...

[flint sage]

Generate some code that takes UnityParameter and casts it to object (or gets the field or w/e) and calls your function

[urban warren]

Generate some code?

[flint sage]

Linq.expressions for example

[urban warren]

Ahhh

[flint sage]

Or Pure IL

Or abuse one of the InternalsVisibleTo assembly names

[urban warren]

Shoot I gotta go re-re-re learn/brush up on expression trees, but that seems like a good way to do it! Thanks!

[flint sage]

It's not that difficult for such a simple case

And you can find examples

[urban warren]

Yeah I remember seeing examples of it.

[flint sage]

If you got until monday I can send you a similar example 😛

[urban warren]

I just don't use them enough, so how exactly to build one starts to slip my mind.

[urban warren]

Depends on if I can figure it out or not depends if I have until monday 😛

[flint sage]

😂

[urban warren]

All this just to support SerializedProperty array binding in UITk

[urban warren]

Wait, wouldn't this have the same issue as a anonymous lambda where you couldn't unsubscribe it?

[flint sage]

Maybe

Does that matter

[urban warren]

Unfortunatly

I might be able to go without, not sure how Unity will handle it

[flint sage]

Uhh

You might be able to get a methodInfo from it and assign that

and use that to unassign as well

[urban warren]

Hmm, I think that would be the same thing

[flint sage]

Tbh, I'm not sure on the details of assign events reflection

You can always set it to null

[urban warren]

Well it isn't an event event. It is the UITK visualElement.RegisterCallback method

[dusty wigeon]

How the event is exposed if UnityParameter is internal ?

[flint sage]

Oh right

[urban warren]

It is actually a visualElement.RegisterCallback which is generic and takes anything that inherits from EventBase<T>. And one of the subclasses (SerializedPropertyBindEvent) is internal

[dusty wigeon]

You are trying to directly register to a SerializedPropertyBindEvent event which has a T of UnityParameter ?

[urban warren]

No, this is what I am trying to do. But SerializedPropertyBindEvent is internal

visualElement.RegisterCallback(MyMethod);

public void MyMethod(SerializedPropertyBindEvent evt)
{

}

[dusty wigeon]

Maybe you could somehow instantiate a System.Action<UnityParameter> with Activator.CreateInstance through getting the type by Reflection.

[dusty wigeon]

Then, the issue is that you cannot declare public void MyMethod(SerializedPropertyBindEvent evt) ?

You cast pointer of function to other type.

[urban warren]

Yes sort of. The event has a SerializedProperty Property { get; } which I want to access, so I still need it, or I guess, need the value from it.

I guess I could do void MyMethod(object evt) instead. But that passes the responsibility of getting the property off to the method itself.

[dusty wigeon]

This would be what I would try.

With the use of maybe Convert.ChangeType.

[urban warren]

Basically I want to create a extension method void RegisterPropertyCallback(this VisualElement element, Action<SerializedProperty> callback)

[dusty wigeon]

You can define the extension itself ? What is public void MyMethod(SerializedPropertyBindEvent evt) in relation to this extension ?

[urban warren]

That is not possible due to SerializedPropertyBindEvent being internal. So I would change it to public void MyMethod(SerializedProperty property) and pass it to the extension method as the callback

[dusty wigeon]

So you want to convert public void MyMethod(SerializedProperty property) to something like public void MyMethod(SerializedPropertyBindEvent evt)

While SerializedProperty is a field of SerializedPropertyBindEvent

[urban warren]

Other way around

internal class SerializedPropertyBindEvent
{
  public SerializedProperty Property {get;}
}

This is not possible as SerializedPropertyBindEvent is internal.

visualElement.RegisterCallback(MyMethod);

public void MyMethod(SerializedPropertyBindEvent evt) { }

So this is what I want to do.

visualElement.RegisterPropertyCallback(MyMethod);

public void MyMethod(SerializedProperty property) { }


public static void RegisterPropertyCallback(this VisualElement element, Action<SerializedProperty> callback)
{
  // Somehow register the `callback` to the `RegisterCallback` and get the `Property` value from `SerializedPropertyBindEvent`
}

[dusty wigeon]

I'll look at it latter if you want to continue the conversation.

[urban warren]

Sure thing, thanks

[plucky trellis]

I've been trying to use a compute buffer in a URP render pass for a while, but no matter what, it always generates a memory leak, or other problems, no matter how I use it. (For reference, the buffer is of a fixed size, known at compile time) Instanciating in the pass constructor, and destroying it when the parent render feature's Dispose() method seems to leak (why?). Creating the buffer in the OnCameraSetup() method, and subsequently destroying it every frame in OnCameraCleanup() doesn't leak at least, but the created buffer doesn't seem to function (why?)

Is there any advice on how to go about this?

[dusty wigeon]

It seems to me that you want three things.
1- Create a System.Action<SerializedPropertyBindEvent> that invoke a Action<SerializedProperty> with the property inside the Action<SerializedProperty>.
2- Register the newly created System.Action in RegisterCallback(System.Action<SerializedPropertyBindEvent>);
3- Have the capacity to unregister the previously registred event.

For the first part, I believe you can simply create the required System.Action through the usage of Reflection with Delegate.CreateDelegate. There is multiple example online that seem plausible that could potentially be adapted to your need.

The second part should be pretty easy if you have a little knowledge about Reflection.

For the 3rd part, you will need to either use a Dictionary to hold the conversion between Action<SerializedProperty> and System.Action<SerializedPropertyBindEvent> or return an object to represent the newly created derived which you would be able to use to unregister.

class Program
{
    public static void Perform<T>(T value)
    {
        Console.WriteLine("Called = " + value);
    }

    public static Delegate CreateAction(Type type)
    {
        var methodInfo = typeof (Program).GetMethod("Perform").MakeGenericMethod(type);
        var actionT = typeof (Action<>).MakeGenericType(type);
        return Delegate.CreateDelegate(actionT, methodInfo);
    }

    static void Main(string[] args)
    {
        CreateAction(typeof (int)).DynamicInvoke(5);
        Console.ReadLine();
    }
}

https://stackoverflow.com/questions/12131301/how-can-i-dynamically-create-an-actiont-at-runtime

[urban warren]

haha funny enough that is actually what I was asking if there was a way to avoid (the dictionary I mean)

[dusty wigeon]

I've little to no experience with URP Render Pass, however everything that I have done seem to work correctly. I've follow tutorial and use the open source code to found out my way.

[dusty wigeon]

I do not think you can avoid it. Except if you use a return object.

[urban warren]

Thanks, good to know I was right that if I wanna unregister the only way really is to use a dictionary to store them.

[dusty wigeon]

If you find a way to map System.Action<SerializedPropertyBindEvent> into Action<SerializedProperty> directly without instancing an intermediate you could, however I doubt it is possible.

[urban warren]

Yeah, too bad you can't do like an implicit conversion for delegates like you can for structs and classes. Guess that would feel a bit cursed though haha

[dusty wigeon]

Maybe the dynamic keyword could be use there. I've yet to experience the full capacity of the this particular keyword.

[urban warren]

Yeah nor have I, though it could be a possability

[dusty wigeon]

Implicit conversion would not necessary help here because you would need to instantiate a new value to convert one from an other.

[urban warren]

Ahh you're right, I would need compile time access to the internal SerializedPropertyBindEvent

[onyx blade]

is there a difference in using actions in a list, or defining a delegate to fire?
Are there significant things I should be aware of?```cs
namespace Utils
{
public class Timer
{
private float _currentTime = 0;
private float _maxTime = 0;

    private List<Action> _actions = new();

    public Timer(float time, Action a)
    {
        _actions.Add(a);
        _maxTime = time;
    }

    public Timer(float time, Action[] a)
    {
        _actions.AddRange(a);
        _maxTime = time;
    }

    public void Tick()
    {
        _currentTime += Time.deltaTime;

        if (_currentTime > _maxTime)
        {
            _actions.ForEach(x => x.Invoke());
            _currentTime = 0;
        }
    }
}

}

[sly stag]

No need for a List. A single Action variable should be subscribable as a multicast delegate, in just the same way you'd subscribe to an event

ie _action += MyMethod

[onyx blade]

oh

oops

[onyx blade]

I may have completely glossed over event registrations on actions. that saves me quite a bit of overhead

much appreciated

[sly stag]

No worries

That one took me a bit to discover too!

[onyx blade]

if only there was a way I didnt need to use Update to make it tick xd ```cs
public class TimerController : MonoBehaviour
{
private List<Timer> _timers = new();

    private void Update()
    {
        foreach (var timer in _timers)
        {
            timer.Tick();
        }
    }
}

altho tbh, one update loop for all my timers isnt too bad

[sly stag]

Yeah, one update loop is not bad. I wouldn't get worried about the overhead of using Update unless you've got 100+ gameobjects in the scene using it at a time, in which case it's time to use some sort of manager for it

[onyx blade]

some form of frustum culling based update checker might not be bad 🤔

make a scene updater behaviour that updates a bunch of others from its Update, based on if they are visible in the camera.
can reduce everything into a single update loop haha

actually, that sounds like an interesting idea, albeit heavily scope creeping

[sly stag]

Just because things are off-screen doesn't usually mean you don't want them to do anything

[onyx blade]

true

[celest depot]

Hello guys. I dunno why I get this error message from my unity project. Each time I install the 2D package in my project, the burst one make this error message

[brisk spruce]

@urban warren whats your end goal for your efforts above? It sort of looks like you are trying to do Event Delegation

[swift orchid]

@onyx blade you can also use Asynk or Coroutine instead of Update. And also you can do not call it each frame it is to expensive. You can call it each millisecond for example.

[onyx blade]

wouldnt calling it in update be more efficient than every millisecond given update runs dependent on the framerate?

or am I missunderstanding something here?

[urban warren]

To add a callback to the SerializedPropertyBindEvent event

[onyx blade]

I do understand the use of the coroutine here, as I can just infinite loop a coroutine and work from there

with the whole WaitForSeconds logic

[swift orchid]

Ye, do infinite loop as coroutine or as Async but maybe update is the same result.

[onyx blade]

iirc, a coroutine has an update loop going on in the background.
dont quote me on that tho

I vaguely recall reading about that somewhere

[swift orchid]

"wouldnt calling it in update be more efficient than every millisecond given update runs dependent on the framerate?" about this one

[onyx blade]

ah

[swift orchid]

Did you read my answer I deleted?

[onyx blade]

yeah i did

[swift orchid]

nice sry I typed message then acsidantely delet it

[onyx blade]

no worries

[swift orchid]

I think Update the same as Looped Asynk or Corroutine you can use any option.

[onyx blade]

I also do kinda think im getting into optimization that may be irrelevant. still good to know.
but the difference in performance is most likely negligable, especially for my usecase.
im not planning to have 500 timers xd

if I have like 10, I tihnk I just about covered things

[sly stag]

Async functions can't call Unity API, so they're not a replacement for Update, and Coroutines run every frame so are no performance improvement (and they also have a tendency to make code messier)

Don't make up performance issues. Unity provides you with the best profiler in the industry. Identify bottlenecks through profiling, then address those.

The idea isn't to make a game that maximizes performance, the idea is to finish a game...if you waste all your time blindly looking for things to micro-optimize, you will not release

[stiff hornet]

just to be more specific, it's being off the main thread which is not allowed for most of the Unity API.
Async can be used fine if you use UniTask (package) or Awaitable: https://docs.unity3d.com/2023.1/Documentation/ScriptReference/Awaitable.html

[onyx blade]

this one hurts me, I havent finished a game in ages

for exactly this reason, and the reason that I suck at doing art and need to rely on others that may or may not help

[sly stag]

Yeah, it's easy to get caught up in, premature optimization can lead you down interesting roads, it's easy to put in weeks of work only to realize you've gained just a few milliseconds

That said, shameless self-plug, here's my solution so people don't have to go down that path with Update:

https://github.com/DameonL/MonoInjection

[onyx blade]

when people ask me what I do, I used to say game developer. now I say tool developer most of the time.
I build tools I will need to make a game, and never finish the game

[onyx blade]

it's not shameless if it's relevant IMO

[sly stag]

Damn, 6 years ago, I need to update that repo

[onyx blade]

im just gonna star that for later, I dont need it (yet) but it is VERY interesting

I must say @sly stag im a big fan of your website xd

[sly stag]

Thank you! The design was fun to put together

[night drift]

Anyone know how to add animations to this code

https://hatebin.com/maectjkwyf

[onyx blade]

time for my own shameless plug! https://torbenvanassche.dev/

[night drift]

Like idle animations and walking animations

[sly stag]

Fancy!

[onyx blade]

the dog is an actual native 3D element btw! not a game engine of any kind 😄

[fresh salmon]

You've posted the same question in 3 channels now. Don't do that, stick to #💻┃code-beginner

[night drift]

My questions is getting flooded by the guy with the mw2 pfp

[fresh salmon]

And you're breaking the server rules

[onyx blade]

bruh

im going to assume gifs in general are a no-go right?

[fresh salmon]

<@&502884371011731486> Shitposting

[onyx blade]

well, meme gifs*

[night drift]

Are you serious rn?

[night drift]

Why are you so serious

[fresh salmon]

Yeah not sure how that gif went through

[night drift]

Its a gif of a goofy cat

Chill

[onyx blade]

server rules are here for professional conduct...

they exist for a reason

[fresh salmon]

No gif here lol, have you actually read the rules?

I mean you broke two in less than a minute, so that probably means no

[night drift]

Well your being very rude

SO uhh

Thats harrasment right there

[onyx blade]

@fresh salmon lets not feed the troll and let one of the mods handle the situation

[hushed fable]

@night drift Read the rules and quit spamming the channel

[night drift]

Im not spamming

It was one gif

Dont ban me tho

[night drift]

Does anyone know how to still help me tho?

[fresh salmon]

In #🏃┃animation probably

[onyx blade]

casually going to drop that I dont like unity's animation controller

xD

[fresh salmon]

You don't do it all from the code. Setup an Animator controller with transitions and parameters, and control said parameters from the code

[night drift]

Yeah

I did that but

wAit

no

[fresh salmon]

Haha

[night drift]

It looked like this

[onyx blade]

ok that was pretty funny ngl

[fresh salmon]

Recording a video of yourself breaking the rules, very smart, especially when mods can see deleted messages

[night drift]

It was a GIF OF A CAT

Please

Ill leave this as soon as i fix my code

I simply want animations

[sly stag]

Is this some sort of weak blackmail attempt?

[night drift]

FUCK
FUCK

[fresh salmon]

They pasted the wrong video yeah

Lol

[night drift]

Huh?

[fresh salmon]

Go to #💻┃code-beginner
Adding animation to a script isn't an advanced issue

[night drift]

I did it

But i looked glitched asf

and the #💻┃code-beginner Channel is filled with the ghost mw2 dude

[fresh salmon]

Repost later then

[onyx blade]

I was about to say, that doesn't really change the difficulty of your question

[sly stag]

#💻┃unity-talk

[night drift]

Yes

[little sluice]

I want to copy pixel data from one texture to another.
But not in a rectangle area, but in a triangle. Any smart ideas for that? In Winforms i would use a GraphicsPath and in Halcon i would use a region to limit the domain, but Unity Texture2D does not have that. Anything similar?

[onyx blade]

do you know the points beforehand?

or is the Texture the only input

if you know the triangle's points, you can itterate the pixels and figure out where stuff is using the same principles for mesh collision detection 🤔

if you dont know the points, I have nothing to offer.
but im very curious how to solve that

[sacred dew]

yeah that's kind of a nightmare to do, there are algorithms for looping over triangle areas in a 2D context, but nothing I know from the unity API.

[onyx blade]

I would think the only real way of doing it is to check if its inside the triangle or not and doing something based on that information. by simply checking the pixel coordinates in relation to the triangle

which, depending on the resolution of the texture, could be a fkn performance nightmare

[sacred dew]

you can speed this up a fair amount by using a triangle drawing algorithm, which are well established

so you only iterate the pixels you need

[onyx blade]

could you also um

generate a triangle mesh?

like

if you take the square you have and the triangle,
you could use the texture on a square.

if you then slice the the triangle from the square,
n the above case you would get 9 or so triangles.
then you search the vertices that are not part of the triangle and remove them
then you sample the triangle from the texture according to the UV's and the resulting texture is what you need

thats probably way roundabout of doing it, but im weirdly interested in seeing how that performs

[little sluice]

I know the points. Its the triangles of a mesh.

[little sluice]

triangle drawing algorithm is a great searchterm, thanks!

[brisk spruce]

is there any quick and easy way to just show a small widget for each of these points for my script, in the editor? Just like a single dot or whatever should be plenty, at global world position

or is this better asked in #↕️┃editor-extensions

[onyx blade]

you can use the OnDrawGizmos API for that

it has a function like start and update on monobehaviours, which you can use for editor only drawing of stuff

[brisk spruce]

perfect, this looks exactly like what I need, thanks!

[onyx blade]

not a problem! I'm currently debating whether or not I need tk define a grid for my game or just manually place the tiles. considering they don't change placing them is fine, but it would be much easier to drag them into a grid they snap into or generate them via code...

[brisk spruce]

Does this logic seem sane? I think its working

Dictionary<Vector3, HashSet<Vector3>> Waypoints = new();
private void TryAddWaypoint(in PolygonCollider2D collidor, in Vector2 from, in Vector2 to)
{
    if (Waypoints[from].Contains(to))
        return;

    var closest = collidor.ClosestPoint(to);
    if (!(to == closest))
        return;

    Waypoints[from].Add(to);

    if (!Waypoints.ContainsKey(to))
        Waypoints[to] = new();

    TryAddWaypoint(collidor, to, to + Vector2.up);
    TryAddWaypoint(collidor, to, to + Vector2.right);
    TryAddWaypoint(collidor, to, to + Vector2.down);
    TryAddWaypoint(collidor, to, to + Vector2.left);
}

trying to not only get all the points but track the connected From/To points as a dictionary of Point -> ConnectedPoints, should make doing the shortest path stuff easy later

[onyx blade]

oh man, I'm not the best at pathfinding... I bought an asset to specifically avoid having to do it xd

[brisk spruce]

code above is purely just to pre-cache the list of all the points contained inside the collidor

and their points they connect to

ah shoot I need to do 8 directions not just 4

[onyx blade]

depends on if you need diagonals, but I assume so

[brisk spruce]

yeh I do :x

TryAddWaypoint(collidor, to, to + Vector2.up);
TryAddWaypoint(collidor, to, to + Vector2.up + Vector2.right);
TryAddWaypoint(collidor, to, to + Vector2.right);
TryAddWaypoint(collidor, to, to + Vector2.down + Vector2.right);
TryAddWaypoint(collidor, to, to + Vector2.down);
TryAddWaypoint(collidor, to, to + Vector2.down + Vector2.left);
TryAddWaypoint(collidor, to, to + Vector2.left);
TryAddWaypoint(collidor, to, to + Vector2.up + Vector2.left);

that outta do er

[onyx blade]

yeah that looks like all 8

I think the initial logic looks fine at a glance, hard to tell from my phone to be certain tho

[tall ferry]

u might wanna check if the dict contains [from] at the beginning and create it if it doesnt, rather than doing the same on [to]. Otherwise when you call this initially, you might have to assign the first point manually

[brisk spruce]

yeh just encountered that haha

Aight so I think I almost got this, but my value for a var I am storing the lines in for the gizmo to read off is is becoming null inside of OnDrawGizmoSelected

I currently just have it declared as a private variable on the class, do I need to apply an attribute to it or make it public or something so it actually persists? Im guessing atm its being treated as ephemeral by the scene editor

[tall ferry]

where are u assigning _lines? because if its on awake/start, itll just always be null until you play

[brisk spruce]

inside of the Setter of Waypoints

Im seeing even Waypoints is considered null though using a complex getter/setter, lets see if I can just function with it being a normal public var and just calc the lines on draw

nah its still null

[tall ferry]

well when is waypoints being set

please dont share extremely cropped screenshots of code, it really shows nothing

[brisk spruce]

its being set by an import script

SuperTiled2Unity has "post processor" scripts you can run to manipulate game objects, so I add this component to one of the output game objects and set Waypoints on it

I dunno if there is an easy way to make a Dictionary serializable in the editor but the data should be getting set, prior it was just a list and the list was populated post import in the scene editor

but it just gets set like so, pretty straightforward

var waypointComponent = grnd.GetComponent<WaypointData>();
waypointComponent ??= grnd.AddComponent<WaypointData>();

waypointComponent.Waypoints = Waypoints;

oh I think I have an idea of whats different, I think because Waypoints (local var) is declared in the scope of the Import script, its dereferencing once the import script finishes

I need to instead do this, to make a new list and fill it with the contents of the Import scripts one so it doesnt de-reference...

waypointComponent.Waypoints ??= new();
waypointComponent.Waypoints.Clear();
waypointComponent.Waypoints.AddRange(Waypoints);

nope nevermind that aint it, still null

[tall ferry]

you shouldnt use ?? on unity objects, check if its null with ==

https://docs.unity3d.com/ScriptReference/Object.html

This class doesn't support the null-conditional operator (?.) and the null-coalescing operator (??).

[brisk spruce]

anywho

does the value need to be actually serializable for OnDrawGizmosSelected to use it, and if it isnt it is null in that context?

Cuz I did a test here and added a string field, and then I set it right after I set Waypoints in the same spot, and then hit a breakpoint inside of OnDrawGizmosSelected annnd...

the string (which is serializable and shows up in the editor) does have its value, but the Dictionary (which is not serialized and shown in the editor) is treated as null

so Im guessing I need a serialized one?

[tall ferry]

no OnDrawGizmosSelected has no requirement for the variables being serializable, your Waypoints is just null in the editor since no code like awake/start has ran

although you are just showing somewhat random snippets, its hard to help. I dont know your system so most of the code shown really means nothing to me, since i dont know when its ran

[brisk spruce]

then why is DidIGetSet not null

[placid ocean]

because its set in inspector

[brisk spruce]

it got set in the same way

var waypointComponent = grnd.GetComponent<WaypointData>();

if (waypointComponent == null)
    waypointComponent = grnd.AddComponent<WaypointData>();

if (waypointComponent.Waypoints == null)
    waypointComponent.Waypoints = new();

waypointComponent.Waypoints.Clear();
waypointComponent.Waypoints.AddRange(Waypoints);

waypointComponent.DidIGetSet = "Yes";

the fact DidIGetSet has a value of Yes there means this code did indeed succeed in setting it

My gut says this has to do with serialization or something, because thats the only difference I am seeing between the two

[tall ferry]

is it serialized in the inspector now?

[brisk spruce]

it also was working earlier when Waypoints was a List<Vector2> and not a Dictionary<...>

And List is serializable but vanilla Dictionary is not

[brisk spruce]

yee

[tall ferry]

so therefore its set in the inspector.

[placid ocean]

why do you not justcreate new Dictionary inline

and check

[brisk spruce]

what do you mean here

[placid ocean]

public Dictionary<Vector3, Hashset<Vector3>> Waypoints = new Dictionary<Vector3, Hashset<Vector3>>();

then it wont be null, just empty.

Im not sure if unity will cache waypoints added at editor time to dictionary

[coral schooner]

Nope

[thin mesa]

correct, unity does not serialize Dictionary by default. there are assets that can do it (or fake it), but unity itself can not

[placid ocean]

There you go

Or use a List of structs if in doubt (for gizmo anyway)

[brisk spruce]

yeah my assumption was because it wasnt serializable, it wasnt persisting to be seen by OnDrawGizmosSelected which seems to indeed be the case

not hard to fix though, just gotta use one of the SerializableDictionary extensions

[coral schooner]

Just populate the dictionary on initialization with Linq

[brisk spruce]

nope, its not being "seen" by the event, it looks like it needs to be an actual serializable value that the editor understands for OnDrawGizmosSelected to persist it in that calling context

[tall ferry]

the issue isnt with it being serializable, its just that the data does not exist yet because your code hasnt ran.
notice in this example, i did not make my class serializable

[coral schooner]

Gizmos don't require any serializable data

[brisk spruce]

you gave it a default value of not null

so of course it wont be null

[tall ferry]

that is exactly my point..

OnDrawGizmosSelected does not require serializable

[brisk spruce]

make it null/empty by default and then set its value via an importer script

[coral schooner]

Can you post the full code? I'm curious why this is an issue

[brisk spruce]

this code is being run inside of an importer script that builds the component during import of a custom asset (SuperTiled2Unity stuff)

var waypointComponent = grnd.GetComponent<WaypointData>();

if (waypointComponent == null)
    waypointComponent = grnd.AddComponent<WaypointData>();

waypointComponent.Waypoints.Clear();
waypointComponent.Waypoints.AddRange(Waypoints);

waypointComponent.DidIGetSet = "Yes";

and then the component itself is literally just

public class WaypointData : MonoBehaviour
{
    public Dictionary<Vector3, HashSet<Vector3>> Waypoints = new ();

    public string DidIGetSet;

    private void OnDrawGizmosSelected() { ... }
}

I had that before, if the importer code creates the dictionary, then its value is null inside of OnDrawGizmosSelected

if I default it to empty so I dont need to set it in the importer, it is en empty dict in OnDrawGizmosSelected

but interestingly enough, the value of .DidIGetSet does get set there and it is present inside of OnDrawGizmosSelected

only difference between the two is one is serializable and the other is not

[placid ocean]

I think youre right, it wont populate waypoints as it cant be serialized?

[coral schooner]

in AddRange(Waypoints); what is Waypoints in that context? KVPs?

[brisk spruce]

yeah I just tested with a List<int> and that is also persisting correctly with the exact same code

only difference between the List and the Dict is one is serializable and the other is not

[tall ferry]

yes because unity saves its value in the inspector, so it is not empty/null. My example was pointing out that OnDrawGizmosSelected does not require the data to be serialized. The reason Waypoints is null is because its not saved by anything

[brisk spruce]

Im pretty sure now if I just swap to a serializable dict this will start working

[tall ferry]

as that will save the values in the inspector, yes

but if you change something from Waypoints, you'll need to have it adjust the values in the inspector or the draw gizmos wont be updated since itll still be referencing that

[brisk spruce]

aight, so far so good, but looks like collidor.ClosestPoint isnt working quite right D:

ah I think I see what is going on, its offset by a bit yeah, aight

[tired fog]

Is 370 kb of data on a command buffer too much in unity?

Or would it be better to separate that into multiple command buffers?

[sage radish]

I don't think spreading it over multiple buffers will help. How many commands are you putting into it?

[tired fog]

Not sure, but many, It's a recursive function that assigns it, good thing, sometimes it weights 120Bytes, and sometimes 380Kb

A rough count, 1228 commands for weight around 31.6Kb

So, a rough count, for that picture I sent there might be around 14620 Commands

😅

Well, on the frame debugger I have this

for 31Kb

[exotic trout]

Hey @brisk spruce, Little late to the party but as someone who also had Waypoint related stuff in a dictionary I just wanted to +1 AYellowPaper's - Serialized Dictionary package (It's free). I've used a ton of third party implementations of serialized dictionaries ("Odin Inspector", "SerializeDictionary", "Serialized Dictionary Lite") but AYellowPaper's is by far the best experience i've had. Your mileage may vary though 🙂

[brisk spruce]

I'll check it out! What was the difference you found that made it stand out?

[exotic trout]

Other cheap or free versions just felt a little clunky with their inspector implementation and odin inspector's implementation doesn't support use in prefabs

[stark crag]

I'm just getting started with DOTS, I created a subscene that has items on a little island in my scene, it is set to Auto Load, on startup the objects all appear, but I'm not seeing log messages from one of my scripts in the subscene (I am debugging why it isn't spawning a GameObject, so I added Debug.Log onto its Awake / Update methods, but not seeing anything). I'm sure I am doing something wrong, but I don't know what.

[north marsh]

If I spawn several un-awaited Tasks to modify a list should I put locks on the list?

Is lock ever needed in Unity tbh?

[brisk spruce]

Tasks dont need locking

but you should be awaiting the Task at some point somewhere

[devout coyote]

Forget what you know about gameobjects with DOTS, do everything in systems instead

Also #1062393052863414313

[tiny pewter]

no await then it runs synchronously and you are read-write the list, you have to lock the critical section

[north marsh]

Can you elaborate on why task won't need locking?

[brisk spruce]

if you dont await it, it runs async but you drop all error logging and whatnot, the stack trace goes kapoof and really bad stuff happens if you drop the task itself before it complete due to exiting context

if it errors or breaks itll just silently fail

but it will still silently fail but in an async way XD

[north marsh]

Yes I noticed this.
All errors are silent within unawaited tasks

[brisk spruce]

fuckin magic, but uh, theres literature on it but effectively Microsoft has some very smart people that made a very smart API

and I have only the smallest inkling of why they magically "just work" and Ive been using them for years

[north marsh]

but wouldn't the several tasks race against each other?

[brisk spruce]

something something Task Manager Engine or whatever

[brisk spruce]

theres a big giant invisible machine underneath the hood ensuring 2 tasks dont step on each others toes at any point

[north marsh]

Ok, and it is not related to Unity's running everything on main thread?

[brisk spruce]

unity doesnt run everything on a main thread, just uh, most of things

but as soon as you use Tasks (correctly), it is inherently auto-magically multithreaded

[north marsh]

List<int> lst = new List<int>(){1,2,3,4};

async Task modifyLst() {
    //remove first element
    //add it to the back
}

void Start(){
    _ = modifyLst();
    _ = modifyLst();
}

So I am for sure getting 3, 4, 1, 2 in this case?

[brisk spruce]

nope itll be a race condition

[tiny pewter]

no

[brisk spruce]

they wont error, but it will be random

or wait sorry in your case uh

[north marsh]

why would this be race

you just said task won't interleave with each other?

[brisk spruce]

itll be a race condition for which fires first but each task doesnt have any awaits in them, so youll still have them do their individual tasks one then the other

[tiny pewter]

i have said you are read-writing the list and you havent await (wait until previous one finish)

[north marsh]

that is intended

[brisk spruce]

so your end result will be the same, but which one did the first step and which did the second technically is a race condition but its very likely to be A then B

[north marsh]

ok I see, that makes sense.

[brisk spruce]

the reason is purely because you're function is Idempotent

[north marsh]

yes

[brisk spruce]

such that A then B produces same result as B then A

if however you modified it slightly to have this:

async Task modifyLst() {
//remove first element
// wait 1 second
//add it to the back
}

then you'll have a variety of results

[tiny pewter]

the execution order maybe first task remove first->interrupt->second task remove then insert->first task insert

[north marsh]

wait 1 second as in thread.sleep?

[brisk spruce]

await Task.Delay(1000)

dont use thread.sleep for Task stuff, really dont need to use it at all anymore, Tasks and coroutines replace thread stuff

[north marsh]

ok so if you delay, then task now can interleave with each other, correct?

[tiny pewter]

actually i believe even you await 1000 the result is still random the interrupt maybe occur before the first one remove

[brisk spruce]

and anything you can await counts here

[north marsh]

but didn't Pixxel say that interruption is not possible without a await in the task itself?

[brisk spruce]

an async method without any awaits just compiles to run synchronously

[north marsh]

oh yea ture

[brisk spruce]

youll notice in Visual Studio if you have async but no awaits within, it greys out and gets underlined as a warning

[north marsh]

For context I am spawning four players in a for loop.

Don't really care who spawns first, but I do use four unawaited tasks to spawn them, instead of awaiting one to spawn then spawn another. Task has await inside because it needs to load addressables etc.

But inside the spawning task I do have to access a common resource (list of spawnpoints).

That's why I asked.

[north marsh]

if you wait 1 sec before removing or after adding to back result is still deterministic, yes?

[brisk spruce]

yeah that would make it Idempotent again

[north marsh]

what about nested tasks. Is result still deterministic?

My understanding is yes, because the two runs of ModifyLst still cannot interleave.

List<int> lst = new List<int>(){1,2,3,4};

async Task ModifyLst() {
    //remove first element
    //add it to the back
}

async Task Wrap(){
    return await ModifyLst();
}

void Start(){
    _ = Wrap();
    _ = Wrap();
}

[austere jewel]

You need to await every Task object you create

else you are going to eat any exception that gets thrown in one

[tiny pewter]

oh i mess up, task without await will run senquentially

[brisk spruce]

yeah the crux of it is there is no await happening between
//remove first element
and
//add it to the back
so they will always run back to back without an interrupt

[north marsh]

what should I do if I just want to run something in the background without caring about its outcome, also don't want it block main thread? In that case I always use unawaited Task and coroutines

[austere jewel]

use async void

but you also need to actually put the Task on a thread unless what you are waiting does that already

because it won't do that on its own

[brisk spruce]

coroutines are meant for that, I think coroutines still keep their errors and stack trace intact

[brisk spruce]

do coroutines keep their stack trace/errors intact? async void drops everything and messes up your stack trace something fierce, makes debugging and break points faceplant

[austere jewel]

Async void works fine in Unity

Any exception thrown in one is thrown to the Unity runtime and logged as usual

Outside of Unity I would not advise it

[north marsh]

can coroutine do Task.WhenAll(

[austere jewel]

No

[north marsh]

spawn that many coroutines, then?

but if a function is async void then you cannot await it at all.

That's fine for fire-and-forget only (which is what I asked for) but I sometimes do want to await for them and I dont want to make a second async Taskfucntion with exact same code

[brisk spruce]

Honestly both seem equally annoying but what can you do, such is the life of coroutines :x

[brisk spruce]

can just wrap it in async void wrapper 1 liner

async void MyMethodCoroutine(params) => await MeMethodNormal(params)

[austere jewel]

yeah I usually would go MyMethod, MyMethodAsync

[north marsh]

yeah Probably yes yes

[austere jewel]

(it's sometimes a practice to append Async to every task-producing method)

[north marsh]

I kept doing void DoTask() => _ = DoTaskAsync(); in the past.

okay ty so much

[brisk spruce]

yeah dereferencing a Task can cause bad things to happen, better to async void wrapper it

Spooky Garbage Collection At A Distance

[north marsh]

feels like I never truly learnt C#

lol

[north marsh]

Just to confirm, if you

void Start(){
   DoTask(2);
   Debug.Log(1);
}

async void DoTask(int val) => await DoTaskAsync(val);

async Task DoTaskAsync(int val) {
  await Task.Delay(1000);
  Debug.Log(val);
}

Will always get 1 then 2?

async void called this way won't be blocking right?

[flint sage]

Yeah but it's not recommended since it'll swallow exceptions

[north marsh]

I thought async void won't swallow exeception, and exception swallow happens only when you dont await a Task?

nvm, it does still swallow exception

then there is no point to use async void wrapper

_ = DoTaskAsync() does the same thing

[fresh salmon]

Yeah it's the Task that contains info on whether an exception was thrown. No Task, no report

[compact ingot]

any exception in the DoTaskAsync above will be handled as a top level, unhandled exception by unity, the exception only gets swallowed if DoTask were to return a Task an that task was not awaited

[north marsh]

Do you mean the async void DoTask(int val) => await DoTaskAsync(val); won't actually swallow?

I gave it a shot and it seems it still swallows exception.

[fresh salmon]

Might be a Unity quirk, but in a regular app, DoTask swallows as it can't be awaited

[compact ingot]

yes, it only works because unity catches everything

it leads to people writing terrible async code just because it works in unity and never learn to do it properly

if you want to write somewhat proper async in Unity (that makes the hackery explicit), use UniTask

[north marsh]

No I am saying it WILL swallow

[compact ingot]

public class Test : MonoBehaviour {
    void Start () {
        try {
            Debug.Log ( "START" );
            DoTask ( 2 );
            Debug.Log ( "START AFTER DO" );
        }
        catch ( Exception e ) {
            Debug.Log ( e.Message, this );
        }
    }

    async void DoTask ( int i ) {
        await DoTaskAsync ( i );
        throw new Exception ( "DoTask EXCEPTION" );
    }

    async Task DoTaskAsync ( int i ) {
        Debug.Log ( $"TASK PRE DELAY {i} {Time.frameCount}" );
        await Task.Delay ( 100 );
        Debug.Log ( $"TASK POST DELAY {i} {Time.frameCount}" );
        throw new Exception ( "DoTaskAsync EXCEPTION" );
    }

    void Update () {
        Debug.Log ( $"UPDATE {Time.frameCount}" );
    }
}

the exception in async void DoTask is swallowed, the exception inside DoTaskAsync is not

[north marsh]

Why the try catch?

[compact ingot]

to show that you cannot catch it

[north marsh]

also how does UniTask help with excpetion handling in this case?

I am acutally using it, but it feels more like a better syntax sugar for me which makes a lot more thing await-able (e.g. a coroutine)

[compact ingot]

It’s has unitaskvoid to explicitly express fire and forget so you never have to use plain void

in a Team with people of different expertise regarding async it makes it much easier to write proper code

[north marsh]

Thank you 🙂

[north marsh]

I think the exception in async void DoTask is not swallowed but we just never reach that line because of the previous exception in DoTaskAsync?

[lament vine]

Hi there, is anyone around for me to bounce a quick question off of?

I would really appreciate any insight you guys&girls might have, but I don't want to paste a giant script into here.

Edit: Here we go. https://hatebin.com/yusrlqlvuv

I'm working on a procedural generation script for a game i'm working on, and I'm having issues generating my roads.

[thorn flint]

Posting code

📃 Large Code Blocks
Large code blocks should be posted as links to services like:
https://gdl.space/, https://paste.ofcode.org/, https://hatebin.com/
https://paste.myst.rs/, https://hastebin.com/

📃 Inline Code
Surround code with three backquotes. Not quotation marks.
To get C# formatting the first line should only contain cs or csharp.
Add a comment with a line number if there is an error message.
```cs
// Your code here
```
Do not share screenshots of code unless requested.

[quick stirrup]

Hey guys, I need help with C# Attributes

[compact ingot]

true, 😄

[glad badger]

When creating complex assembly hierarchies, I usually see people nest script folders within one another as shown in the image on the left with their asmdefs kind of all over the place really because of the structuring. Is there any problem creating assembly hierarchies as shown in the image on the right?

I just never see people use the approach on the right and wondering does Unity have a problem with that way?

[craggy spear]

1- read https://dontasktoask.com/
2- post your question in #💻┃code-beginner as it sounds like a beginner level question

[glad badger]

command line tool problems perhaps?

[devout coyote]

Don't think there should be a rpoblem

Just looks weird in the editor I guess

[quick stirrup]

I have multiple singleton managers that serve for multiple interfaces.
For example, UpgradeManager handles some upgrade operations for "IUpgradeable"s. I am basically passing the IUpgradeable to the UpgradeManager. Upgrade method and doing the business there. I want to call the Upgrade method like an extension method for IUpgradeable but I also do not want to create static extension methods for each type of managers and interfaces that those managers serve.

I want to implement a generic solution using an attribute for these kind of methods. The attribute will get the target type from the first parameter of the method target type (for our example typeof(IUpgradeable)), then objects that inherits IUpgradeable interface will be able to call a method with same name (by probably using some Reflection) as the manager's one.

In summary, if the manager has a method like:

[MyAttribute]
DoSomething(ITargetType targetObject, int someParameter)

Then I want to be able to call like:
(myObject as ITargetType).DoSomething(5)

And it will work like as it was called like:
TheManager.instance.DoSomething(myObject , 5)

[urban warren]

So I have a hex terrain grid. Each hex cell just samples a stepped height map to get the height. My issue is, I am trying to figure out which cell the mouse is over.
My thinking was to get a ScreenToWorld raycast against a ground Plane at 0 facing up, that way I have a start and end point. Get all of the cells along the ray (from the ray origin to the point on the plane). And I thought I could just test the height of the cell vs the height of the ray at that position. But that doesn't really work intuitively, and getting the height of the ray at a point is kind of a pain.
So I was wondering if anyone had thoughts of how I could approach it differently. I guess I could make a hex mesh collider, and move it to each cell and raycast against it. But that feels really ugly to me.

[sly stag]

You should be able to get the click position at a height of 0 and then use math to figure out which hex that click would be in using only the dimensions of the hexes and your mouse x,y

I don't know the formula for hexes, but the same concept as if you were trying to work out the cell for a click on a grid, but obviously the formula for a hex will be a little longer

[urban warren]

I mean, that is what my thinking was. But struggling with how to get there.

Getting a cell at a position is easy if it is just a flat grid. But the height is what is messing me up

[sly stag]

https://math.stackexchange.com/questions/3058657/converting-a-point-to-a-hexagon-coordinate#3059458

Why does height come into it?

[urban warren]

Because the cells are 3D?

[sly stag]

Are you trying to align the hexes with terrain topology?

[urban warren]

No, the hexes are the terrain topology. If you look at the screenshot you can see the terrain & hexes. I am trying to figure out which one the mouse is over.

I have a 2D logical hex grid, and then when I generate the terrain I use a height map to give each cell a height as well.

[quick stirrup]

I think the approach of measuring ray height you mentioned can be tried. As you said, you raycast a plane and you can stop at the first "higher than the ray" you come across by running a while loop from the nearest to the farthest one among the hexes on the horizontal path.

Additionally, I don't know if this is less costly than using a hex collider. Seems like a lot of computation as you get farther away from the hex you're aiming for.

[urban warren]

That approach kind of falls apart when the mouse is not over the top of a cell and instead the side. Maybe just moving a hex collider around would be the better solution. But it is just so much uglier for what is needed.

[dusty wigeon]

The easiest way, would be to simply add a collider on every tile. Depending, on your platform and type of game, I suggest you use that. It might be useful later. You always have the option to come back latter.

You can also use https://docs.unity3d.com/ScriptReference/RaycastCommand.html if performance is an issue.

[urban warren]

I think for now I will just see if I can go with the collider that I move and test against. I don't really need colliders for anything else.

[dusty wigeon]

You might not need it at the moment, however it takes a lot less time to setup all collider and they might come useful later.

You are over-engineering.

Which is not a good thing.

[urban warren]

It would probably take more work actually as I would need to handle instantiating and pooling them and updating their positions as the terrain is infinit. And right now I don't have any GOs actually spawning, just GPI instancing which later will be replaced with chunks of meshes probably

[dusty wigeon]

... I see that you have been over-engineering for a while.

Terrible habit.

[urban warren]

Lol, well I don't think I am, but I'm open to suggestions if ya have them. The terrain/grid is infinite, and will probably use a dual grid for the actual mesh generation. In the game all you do is place tiles on cells, and sometimes the tiles affect the surrounding cells.

[dusty wigeon]

You do not need to create 1 mesh. You can use multiple mesh.

1 per hexagon.

[urban warren]

That is what I am doing right now

[dusty wigeon]

Then, why you said you dont have GO ?

[urban warren]

Because I am using GPU instancing

[dusty wigeon]

You can use GPU instancing on a material ?

Not exactly sure which type you are talking.

The Graphics command ?

[urban warren]

What? Just like Graphics.RenderMeshInstanced(renderParams, _hexPillarMesh, 0, _pillarMatrixes);

Though I am having trouble figuring out how to set per instance data :/

[dusty wigeon]

You can simply use a material and mark it as instancing.

While having a standard renderer.

Also, I guess your game is gonna have a somehow isometric view or top down ?

[urban warren]

Yeah top down, like a city builder type angle

[dusty wigeon]

Yeah, even more reason to not care about performance.

Top down game are hyper easy to optimize.

Because the camera is limited to the amount of things it see.

Personnaly, I would use a GameObject per tile with a collider.

Everything would be way simpler.

[urban warren]

Huh, ya know I never realized that....

[dusty wigeon]

Give it a try.

In case you did not look at it yet.

https://docs.unity3d.com/Manual/GPUInstancing.html

[urban warren]

Yeah I went and checked that my self just to see haha

Anyway, that will only work until I switch out for a generated mesh. Then I will just be using the GOs for the colliders

[dusty wigeon]

Why would you generate a mesh ?

[urban warren]

Because otherwise I would need 6 Meshes for every hex. Are you familar with duel grid based mesh generation?

[dusty wigeon]

Not particularly

Nothing is coming up on my search.

[urban warren]

Basically you have a second grid that is offset and you generate meshes from segments. It results in having smoother transitions and requires very few different meshes
https://twitter.com/OskSta/status/1448248658865049605

[dusty wigeon]

Like cube marching then.

Not exactly sure where is the issue though ?

[urban warren]

Yeah similar

[dusty wigeon]

You have finite amount of mesh.

Which means a finite amount of instanced draw call

And you have a limited amount of object in view.

I do not think you would gain much with mesh generation in your context.

[urban warren]

Yeah I'm just not sure at what point the trade off would be. Because if it is generated, you can collapse overlapping vertices, and have fewer meshes to draw. But could no longer use instancing. And with instancing you would have to render 6x the number of meshes per hex, but you can use instancing.

[dusty wigeon]

You are still over-engineering. Let's me ask you this: Do you need that ?

Is your game not gonna work/take too much performance because of that ?

[urban warren]

What is "that" in this case?

[dusty wigeon]

Mesh generation.

How many draw call you expect without using any optimization ?

In other words, how many renderer in view you would have at a given time.

[north marsh]

any way to extend AreaEffector2D so that it could selectively affect gameobject according to my custom rule (not just based on layer?)

[urban warren]

Not 100% sure atm as I am still playing with the hex size/scale atm. But 200k+ maybe?

[dusty wigeon]

200k renderer ?

What are you doing lol.

1:1 earth ?

You gonna need more than mesh generation to be able to sustain something like that.

The amount of detail you would get with 200k renderer in view would be ridiculous.

[urban warren]

200k cells. This is 287k cells render with the position gizmo being the center of the terrain/area

[dusty wigeon]

Your view isnt what I'm expecting it to be

You said top down/isometric

[urban warren]

Not sure what you would call it really? To me this is still a 'top down game'. But sorry I wasn't clearer on the angle!
You can title/rotate the camera so it looks from a lower angle or from a higher angle, and zoom in/out

[sleek marsh]

if you have a fixed perspective you probably want to have at least a few separate objects so the camera can cull most of them and save on tris

[dusty wigeon]

Having a fix angle make everything a LOT more easy to optimize.

[urban warren]

Not a fixed angle unfortunately

[dusty wigeon]

Everything is dynamic (Remove tiles/add tiles) ?

[urban warren]

No the terrain tiles/cells are static

or I guess better put, are not edited at runtime

[dusty wigeon]

So you would need to generate your mesh once ?

[urban warren]

Infinite terrain though

[dusty wigeon]

Once per chunk

[urban warren]

Yeah, it could be generated once and kept in memory

Though might need to dispose of them as the memory could get too large.

Mesh gen using burst and jobs is fast

[sage radish]

So you'll only ever need to render this terrain and won't need to deal with collisions other than mouse clicking?

[dusty wigeon]

Relatively to CPU yes. It can be still a huge percentage of a frame.

[urban warren]

Yuuup

[sage radish]

I've seen games do mouse picking on the GPU. In fact, I believe Unity does that to know what you're clicking on in the scene (where not all meshes are guaranteed to have colliders).

I don't know what the go-to approach is, but a simple one would be to sample a depth texture of the scene at the mouse screen position, use that to convert to a world position and then you know exactly the world position the mouse is over.

[dusty wigeon]

It seem to me that if you intend to make such a huge terrain with the possibility of viewing everything you gonna need to integrate the notion of tessalation somewhere.

[urban warren]

oooh, good thinking. I you could get a depth texture with just the terrain right? Though I guess just moving a single collider around would probably be easier to code up and maintain. At least for me.

[dusty wigeon]

I'm looking at the tiles in the back and my heart is crisping.

[urban warren]

Possibly, not too sure how I would go about it. LODs might be enough. Right now it is taking about 24ms to render. And that is with an unoptimized mesh, and rendering all 6 sides of the mesh even if only the top is visible

[dusty wigeon]

How would you make your hexagon lower in in amount of vertices ? You would need to combine multiple hexagon together.

For a standard terrain, it is kinda easy. however for something like that it is going to be a pain in the ass.

[urban warren]

Yeah, I mean if do mesh gen instead of instancing

[dusty wigeon]

I mean, you gonna need for sure to do Mesh Generation to be able to manage that much information.

I though you were doing a small top down city builder.

Maybe something like Civilization.

[urban warren]

The big kicker is the the cell size. Civilization's is much large so that helps a lot. I mean I might end up clamping the view down a bit so you see less.

[dusty wigeon]

I mean, how would you sit a building on a cell that small.

[urban warren]

Trees and stuffs, again, still working on the scale. I will probably make the cells 2 - 4x the size they are right now.

[dusty wigeon]

Also, the cell are gonna be really small to manipulate easily.

[dusty wigeon]

You see, this is WAY more managable.

And from this angle, you would be more than fine with renderer only.

You have what, 400 visible renderers ?

If we say that every tiles is 3 draw calls, that is 1200 draw calls which run on almost every PC.

[urban warren]

This is what is right now for angle and 'gameplay'

[dusty wigeon]

I feel like it is unnecessary to add the possibility to change the pitch.

[urban warren]

Maybe, but it feels nice to me and I like it when games have it. Though right now it probably goes down too far allowing you to see too far.

[dusty wigeon]

Anyway, if you have something close to what you are showing me. I feel that mesh generation is not necessary.

If it more in scope of what you were showing previously.

You gonna need to make a lot of optimization.

[urban warren]

Yeah I will have to see how instancing handles it with more instances but less geometry when I am doing the duel grid gen

[sage radish]

This is a perfect scenario for GPU mesh generation and rendering, because you only ever need to render it. The CPU doesn't need the mesh. Jobs and Burst are fast, but a GPU is bonkers. An RTX 3060 has 3584 freaking cores. No CPU comes close to that.

[urban warren]

Ooooh, I totally forgot about Unity adding GPU support! I think the example repo that Aras made has some examples of doing it if memory serves.

[sage radish]

GPU support to what?

[urban warren]

For mesh gen?

Unless I am thinking of something different than you

[dusty wigeon]

You were planning on doing it on CPU O.o

[sage radish]

Drawing procedural meshes on the GPU has been possible for a long time. Graphics.DrawProcedural has existed since Unity 5.2.

[urban warren]

Oooh you mean that

I was thinking of the VertexBuffer stuff Unity added https://docs.unity3d.com/ScriptReference/Mesh.GetVertexBuffer.html

[sleek marsh]

i have to ask is your performance really struggling with the stuff you have going on? it seems like a scene like this should still give you 300+ frames easily

unless you’re developing for mobile

[urban warren]

Right now with about 100k cells it is taking around 4ms, though the WaitForPresentOnGfxThread takes 8ms sometimes. Not sure if that is directly related or not. Brings it down to about 50 FPS for me

[sage radish]

That means the CPU is waiting for the GPU to finish.

So you're bottlenecked by the GPU.

[urban warren]

Makes sense Probably RenderMeshInstanced I would imagine

[brisk spruce]

aight, I think this seems like a very solid skeleton for the core of my game engine, anyone wanna give their 2 cents on it? Anything I should consider as a potential gotcha?

Game is a pseudo turn based system where everything occurs in "rounds", but the player input can queue up more than 1 "round" worth of stuff happening (IE if they click to move 3 squares, thats 3 rounds of stuff thatll happen back to back)

https://hatebin.com/tzibjqiliq

[viscid blaze]

Can anyone help with gyro?

So it was working and doing that before

but now my remote might not be working, I unplugged and re plugged and checked in project setiings and iphone is still connected

at the start now it changes the cube to rotate a lil but for my phone its just not showing up, I also restarted the app and nothing

Thats what its doing now

Something with the remote 5 because I reopened and now my phone isnt showing up

sorry

[dusty wigeon]

Good:

• You have a separation between Input and Action which marvelous. It can look like a small things, however it really helps in the management of the flow of the game such as game state, player control, UI, etc.
• You attempted to manage the game with the StateMachine Pattern which can really help in game that are turn base.
• I see that you though of dividing the management of the action and the actual effect of the action.

Bad:

• The usage of Task seem to be a bit forced and out of place. The usually pattern I see is more close to having a "current" task which is update by function call instead of a literal Task.
• It is kinda hard to understand the purpose of RoundEvent. It may be related to the fact that you seem to have a peculiar way of doing the "rounds".
• The StateMachine Pattern is not correctly implemented. In fact, what you have done is closer to the Command Pattern.

In object-oriented programming, the command pattern is a behavioral design pattern in which an object is used to encapsulate all information needed to perform an action or trigger an event at a later time.

Neutrals:

• Your design of "rounds" is non-orthodox given that a single move is a full rounds worth. There is also the RoundEvent enum that is defined as a Bitmask and is pretty specific.
• You are doing the waterfall methodology. It means that you design everything first, then you implement it all at the same time. This is not the most optimal strategy in the elaboration of any application or video game. Instead, you should opt for a methodology that build small iteration at a time. In your situation, that would be the implementation of a single Action such as move (Selection of "Unit", Highlight of available displacement, etc.). There is no need for any architecture at this stage.

https://en.wikipedia.org/wiki/State_pattern
https://en.wikipedia.org/wiki/Command_pattern

[crisp zealot]

how would I go about syncing a firing animation clip in my animator component to my weapon firerate?

[dusty wigeon]

You change the speed of the animation.

[crisp zealot]

how would I go about that? I'm using a bool parameter, but im not sure if thats the best approach

    {
        WeaponAnimator.SetBool(_Shot, true);
    }```

[dusty wigeon]

You would need to expose a parameter from the Animator directly.

Something like that.

[crisp zealot]

here is my current parameter setup, not sure where to go with what you showed me

[brisk spruce]

1. I am using Tasks to leverage IAsyncEnumerable and provide cascading support "downstream" in case anything lower in the tree needs to be properly async (IO ops and whatnot)

2. Rounds can have multiple queued up, a user can click to do multiple moves of rounds (ie click 3 tiles over), but I wanna be able to short-circuit out in case partway through something interrupts the queue (their character Encounters an object of interest part way through, triggers cutscene, gets threatened by an enemy). The RoundEvent has flags for these possibilities.

3. It will effectively be a StateMachine, the state machine parts just weren't included in the example there.

4. I already did a bunch of proof of concepts for individual pieces of my game, starting to bring things together now and this is the first big piece. I have a solid idea of all the pieces I intend to build.

[crisp zealot]

im trying to find this, my animator setup in my inspector doesnt look like this

[dusty wigeon]

This is the inspector from an Animation State inside the Animator. (What I have shown)

[crisp zealot]

mine looks like this

[dusty wigeon]

The upper part

This is a transition

You need to click on a state and look at the upper part*

[crisp zealot]

ah now i see

[dusty wigeon]

1. Overengineering. You can achieve whatever you are planning to do in simpler ways. Keep It Simple and Stupid (KISS).
2. As far as I know, you do not need to have the concept at this level. You only need the concept of "interrupting". Also, the fact that it is a bitmask is strange and seem unnecessary. (I'm not asking you to explain it to me. I'm saying that the current way that you defined and named your structure (Notably the fact that it is a bitmask and that is not an Event) was not obvious for me. You take it how you want. Code should be self descriptive. )
3. You should try to keep your StateMachine as a StateMachine and not merge it with other behavior.
4. When I say iterative implementation, I do not say to do proof of concept, but literally build on top of previous iteration. The important is to always have something functional, take small steps to be able to rapidly come back if needed and then slowly refactor.

[brisk spruce]

Overengineering. You can achieve whatever you are planning to do in simpler ways. Keep It Simple and Stupid (KISS).
I mean, if you have an alternative Id love to hear it, but this seems actually pretty simple to me actually. IAsyncEnumerable is pretty easy to use
As far as I know, you do not need to have the concept at this level. You only need the concept of "interrupting"
Well I need to know what kind of interruption it is, I may add further metadata to it but yes, thats the big thing about IAsyncEnumerable in modern C#, for all intents and purposes it is the successor to Events. Events are pretty terrible as soon as you want them to play nice with anything async (there's a reason you pretty much never see Tasks and Events working together at the same time, you cant += Tasks together the way you expect Events to work. IAsyncEnumerable solves this.
You should try to keep your StateMachine as a StateMachine and not merge it with other behavior.
Im not. This is one of the 3-4 states of the game fundamentally (Menus, Cutscene, Ready, and Busy)
When I say iterative implementation, I do not say to do proof of concept, but literally build on top of previous iteration.
Honestly workflow wasnt something I wanted to chat about, Im familiar with waterfall, but I usually operate on much larger scales for my projects. A single sprint for me on my personal projects usually is on the scale of weeks to months, I have the time/energy for it.

... I should add a post to my blog demo'ing how largely speaking IAsyncEnumerable can just be a straight up replacement for Events, its largely the same principle but easier to work with, maintains stack traces better, and best of all it works with Async stuff inherently which is ❤️

[jolly token]

Not sure what is the point of IAsyncEnumerable if there is no code that will react to its returning value

[brisk spruce]

Eh sort of, you still have the concept of "the pipe is now closed" but usually you can have some kind of useful output on the listening end. In my case my output from the pipe is the roundEvent I |= together

[jolly token]

But if you are just trying to | (which look wrong in the code but I assume it's pseudo) it then it could be just Task<RoundEvent> and let the handler do | 😄

[brisk spruce]

That would return a single event and not multiple

Task<RoundEvent> is to IAsyncEnumerable<RoundEvent> in the same way that RoundEvent is to IEnumerable<RoundEvent> in syncronous code

You can yield return and whatnot inside of IAsyncEnumerable methods

[dusty wigeon]

Events is actually a means of synchronisation between thread. Not sure what you mean by we never see Event and thread together. Event are not return values.
Your StateMachine contains directly the function PerformRounds which is not a responsibility of the state machine.
When I say KISS, I mean something like the following. There is no need for the usage of async process here as far as I understand the way your game works. (It might actually cause issue down the line)

private class PerformRoundsState : State {
  private Queue<Action> actions;
  
  private void Update(){
    if(actions.Count == 0)
      return;
    
    if(!actions.Peek().Update()){
      action.Dequeue();
    }
  }
}

[brisk spruce]

and what do you do with current though

[jolly token]

I know, I'm saying that you can just return RoundEvent.Okay | RoundEvent.Interest; instead of yield return each flags because it doesn't seem like you are doing something with that.

[dusty wigeon]

My bad, current is unused there.

[brisk spruce]

Ah, but what if you want an actual count of how many things got returned?

[dusty wigeon]

There is probably a more elegant way of doing this.

[brisk spruce]

Now you turned 2 things into 1, which can mess up your counts 😉

[jolly token]

Why do you need that count for?

[brisk spruce]

tracking expirations on stuff, etc, can be a wide variety of things.

Even just debugging, if Im wanting the 7th event, but the 4th and 5th event have been smooshed together by some called downstream, now my 7th is my 6th

but also yield return is just very nice to have overall, usually when you are Guard Clausing, yield return is slick

instead of

var myEvent = RoundEvent.None;
if (case)
  myEvent |= RoundEvent.Okay;
if (case)
  myEvent |= RoundEvent.Threaten;
if (case)
  myEvent |= RoundEvent.Interest;

....

return myEvent;

You get this instead

if (case)
  yield return RoundEvent.Okay;
if (case)
  yield return RoundEvent.Threaten;
if (case)
  yield return RoundEvent.Interest;

and thats it, done

Note how it removes your need to keep doing a "init a var, track it for the entire method, then finally return it at the very end" which gets very tiresome very quick for every single one of your implementations to be forced to do

Same reason why usually its a lot nicer to return an IEnumerable<T> instead of a List<T>, the latter you gotta new up a list, start filling it up, then return it at the end. Whereas the former you just yield return as you please and usually remove a few lines of code

[jolly token]

I doubt that you'll need that for generalized code.
Also not sure with the intention of your Linq expression, probably SelectMany use case?

[brisk spruce]

Which one?

[jolly token]

This one

var handler = RoundHandlers
      .Select(rh => rh.Run()).Aggregate((a,b) =>  a.Join(b));

[brisk spruce]

Ah you cant SelectMany IAsyncEnumerables

in fact normally you cant even Join em but I have a specific extension method that handles that

Theres a very popular mainstream library though that lotsa of folks use that lets you get all your usual Linq stuff on IAsyncEnumerables, though the methods are all async as well of course

I only need Join though so I have just 1 file for that, I dont need the whole lib for it

[hardy sentinel]

IAsyncEnumerable ain't that bad, why is it overengineering? it's like IEnumerable but async. Would you call usage of IEnumerable overengineering?

[dusty wigeon]

Because there is no need of async.

[hardy sentinel]

hm well right no Linq 😛 but still, decent for plain iterations

[dusty wigeon]

Why would you turn something async if there is no need.

[brisk spruce]

what gives you the impression there is no need

[hardy sentinel]

saying assuming there's a need, of course

[brisk spruce]

as I stated earlier, I expect I have need of async there, IO operations and other such stuff

[dusty wigeon]

No need of async for IO and other stuff as well.

[jolly token]

Sure, name Join was quite misleading to me when using within Linq

[brisk spruce]

yeah I kept the linq method name still hehe, its pretty much the same principle, joins the two lists together but in a ~fancy~ way haha

[jolly token]

Join should work like database join. I'd name it Concat

[brisk spruce]

I believe thats Intersect for Linq IIRC?

[hardy sentinel]

that's partly true as well, but might not go along with all architectures. I'm fine with having some dispatcher when I need to sync to main thread, but some apps or orgs may not be

could complicate stuff a lot for junior devs as well

[jolly token]

Intersect works like you're putting in HashSet one so bit different from Join

[dusty wigeon]

It is strange to complicated the whole architecture on the pretext of maybe one action will actually need async.

[brisk spruce]

Im not sure why you think its complicated

[dusty wigeon]

I say that it is more complicated. Not complicated.

There is a difference.

[brisk spruce]

I dont consider it more complicated either

the code doesnt change, even if it wasnt async Id still have pretty much the same code

[dusty wigeon]

I'm sorry, but this is a fact. If you can remove the usage of a whole concept it only make it simpler.

You ask opinion, and this what I think. If you do not need that stuff, do not put it there.

[brisk spruce]

my await foreach would turn into .Aggregate() and thats.... about it

[dusty wigeon]

From my perspective, it is not needed.

[brisk spruce]

and then Id have some fewer awaits in my code, and instead of Task id return void, aside from that, itd all be the exact same logic

[dusty wigeon]

However, I'm not the one making the game, I'm just speaking from what I saw.

[brisk spruce]

I guess mostly when you talk about complicated, I view that through the lense of the actual logic.

Id be writing the exact same logic for all my code, Id just be yield returning to an IEnumerable instead of an IAsyncEnumerable

[dusty wigeon]

And it would be better.

You could even turn it down way more.

[brisk spruce]

because I saved... five characters?

[dusty wigeon]

No, because you removed a whole dimension.

[hardy sentinel]

sorry, I just read the script. it's quite complicated if it just intends to process a turn 😅

[dusty wigeon]

Threading can cause issue.

[brisk spruce]

process n turns, which many things can happen in 1 turn (every single entity that is active takes its turn simultaneously)

[hardy sentinel]

best scripts are simplest scripts -- try to read it as a junior, and count the WTFs 😄

[dusty wigeon]

Bad idea to have simultaneously turn.

[brisk spruce]

thats how it is and how the game works

[dusty wigeon]

You gonna have a lot of issue with synchronisation and undefined behavior.

[tall ferry]

i may have missed it, but what does the async actually provide in the current implementation? Is this for performance

[brisk spruce]

the actions that will occur for the turn are all pre-calced, that wont be an issue

[dusty wigeon]

How many action you intend to process ?

[hardy sentinel]

I get this, and if you intend to work alone that's totally fine! But if you intend to work with a team, might wanna consider how intuitive would be for more junior guys to process.
All depends on the context, of course. Code is code and as long as it works and won't shoot you in the foot later it's fair game and up to preference.

Personally some of my code is intentionally made complex and/or optimized for extensibility & performance. It's just the important high-level stuff I keep 100% clean.

[brisk spruce]

I do find a lot of people struggle with Tasks and async stuff for awhile, its a bit of a different way to look at code, its one of those things that once you get comfortable with how they work, its not bad, but I wouldnt be writing stuff like this for a junior to consume from naw

[dusty wigeon]

There is a difference between struggle and keeping things simple.

You seem to take this the wrong way.

[brisk spruce]

No people just genuingly struggle with Tasks and Async for awhile, Ive had to train plenty of em myself personally, its a bit to grok the whole concept of it when you are used to Synchronous code

[dusty wigeon]

Do you ever consider the fact that you may be over using Async and Task ?

[jolly token]

Is it intended to be simultaneous? It looks half sequential 😄

[brisk spruce]

they're pretty much the default now for modern asp.net stuff and whatnot

[dusty wigeon]

We are not in ASP.net environment.

I could not talk for ASP.net.

[brisk spruce]

this part right here will fire off everything simultaneously unblocked

var handler = RoundHandlers
    .Select(rh => rh.Run()).Aggregate((a, b) => a.Join(b));

[dusty wigeon]

However, as far, as I experienced Unity, I use rarely async.

[jolly token]

It will start simultaneously, But it won't proceed after first yield until the loop consume the value

[brisk spruce]

Nah they all run independently of each other writing to their pipes

oh sorry I see what you mean, mmm yes I think you are right? but the blocker is merely:

roundEvent |= roundEvent;
which wont exactly block for a notable amount of time

it should definitely be pulling off the pipe way faster than the pipe is getting filled up

[jolly token]

The blocker is the other actions.

yield stops IAsyncEnumerator til Current is consumed

[brisk spruce]

Oh you mean here?

foreach (var action in actions)

[jolly token]

Nah await foreach(var e in handler)

And that too probably

[brisk spruce]

gotcha yeah, should be fine because as the pipe should always be close to, if not completely, empty

so it will write to the pipe and pretty much instantly get picked up and consumed

I think I see what your concern is, if I do like

foreach (var enemy in enemies) {
    var roundEvent = await DoEnemyAction(enemy);
    yield return roundEvent;
}

then I will end up still only processing enemies one at a time right?

[jolly token]

Similar, but it's worse because it starts simultaneously at first but not after first yield 😄

[brisk spruce]

yeh so, the trick is you actually would wanna do something like this

// ToList is actually important here to hydrate the tasks, I have learned this lesson the hard way before D:
var enemyTasks = enemies.Select(DoEnemyAction).ToList(); 

foreach (var enemyTask in enemyTasks) {
   var roundEvent = await enemyTask;
   yield return roundEvent;
}

psuedo code but yeah the key is to fire off the events together (if you wanna get fancy it can be good to batch em too, rather than just fire em all, sometimes), and then do your yield returns after you got everything spun up and going

[jolly token]

Well yes for regular Task? But you now have IAsyncEnumerable involved

[brisk spruce]

nah thats handled here:
yield return roundEvent;

[jolly token]

Handler1() { await Task.Delay(100); }
Handler2() { yield return val1; await DoSomeWork(); }

In this case your Handler2 will not invoke DoSomeWork until Handler1 is done

Because val1 is not consumed

It's lazy evaluation

[brisk spruce]

right, change Handler2 to this

Handler2() {
   var workTask = DoSomeWork();
   yield return val1;
   await workTask;
}

make sure your tasks are spun up before you start yielding

[jolly token]

Sounds bad, what if you have dependency between tasks? What if you need to yield return or yield break depending on task's return value? :/

[brisk spruce]

what if you have dependency between tasks
Whats the issue?
What if you need to yield return or yield break depending on task's return value?
Still no issue, you can await before the yield, just know that any new tasks you spin up after a yield can be blocked

[jolly token]

So you are using IAsyncEnumerable just to return value at the end? That sounds very error prone

[brisk spruce]

you can return multiple, no issue

[jolly token]

The main problem would be await's behavior will be different depending on where you put it (before yield or after)

Well Task's behavior to be specific

And that is very confusing

[brisk spruce]

yeah its not await that you care about, its making a new Task

as long as you try to avoid, as much as possible, making new Tasks after the yield, itll work very smooth

[hardy sentinel]

personally I think something like this might be quite simpler, without any performance drawbacks:

public class TurnManager {
    // ...

    public void QueueAction(TurnAction action) { .. }
    public void UnqueueAction(TurnAction action { .. }

    public void EndTurn() {
        var consumedActions = actions.Where(x => x.turnID == currentTurnID);
        onTurnEnd?.Invoke(consumedActions, onTurnStart);
        actions.RemoveRange(consumedActions);
    }
}

public class TurnProcessor {   
    public TurnProcessor() { TurnManager.instance.onTurnEnd += ProcessTurn; }

    async void ProcessTurn(IEnumerable<TurnAction> turnActions, System.Action callback) {
        // potentially filter or sort actions. e.g based on initiative etc.
        foreach (var action in turnActions) { // could even 'try to perform' them in parallel and process the results on main thread (e.g.: anims)
            if (!await action.TryPerform(out var result)) { ProcessResult(result); }
        }
        callback?.Invoke();
    }
}

[jolly token]

You can't expect everyone would understand and follow that, If it is just your personal project then well you do what you want

[hardy sentinel]

yeah, they already said it wouldn't be for the eyes of juniors, but I agree it's not great to complicate core stuff like turn processing right off the bat 😄

you should also take in account you'll need to process stuff like animations, position changes, and UI in main thread anyway, and as it is, everything would have to be called on main thread (including IO)

I personally find IAsyncEnumerable great, but it might be a bit overkill for this considering it needs boilerplate to make functional 😅

[brisk spruce]

so I just checked this here, doesnt seem to be the case unless I missed something

https://dotnetfiddle.net/o57fhm

actually since I cant see the console live, maybe it is the case, lemme retest but on my own machine to see if it actually halts and then spits all 10 out at once

nope no problem, the longer Task.Await doesnt block the other enumerable at all

[hardy sentinel]

it should work because linq does lazy evaluation

[brisk spruce]

What do you mean? which line of code?

[hardy sentinel]

        var handler = RoundHandlers
            .Select(rh => rh.Run()).Aggregate((a,b) =>  a.Join(b));

        // Process all the events and aggregate them into a single flagged enum
        await foreach(var e in handler)
            roundEvent |= roundEvent;

your code should work because linq does lazy evaluation, so nothing is populated or aggregated until you actually iterate on handlers (for which you use await)
but actually might have to do .Select(async rh => await rh.Run()) or smth xD

[brisk spruce]

.Aggregate wont be lazy, thats a greedy linq function, it hydrates the query 🙂

[hardy sentinel]

really 👀

[brisk spruce]

its like .Max, .Min, and .Sum and etc

yeh, cuz it returns just a T, not an IEnumerable

its just a fancy .Sum function when you gotta do something fancier than just A+B

[hardy sentinel]

this might be a problem then 😛 rh.Run() should be executing async code

[brisk spruce]

it does :3

We are aggregating effectively a list of Tasks, not the results of the tasks

its uh, a sort of fancy .SelectMany for Tasks, for all intents and purposes

[jolly token]

I see your Join is not just Concat so yeah then it's fine. I'd say it is doing quite inefficient operations, but that's different argument.

[brisk spruce]

Ive tested it under some hefty loads, it doesnt seem to have much overhead, a little bit but an acceptable amount for my purposes

I could write a better version that actually can take multiple AsyncEnumerables, rather than repeatedly combining A+B=C, C+D=E, E+F=G, but instead just do A+B+C+D+E+F=G

[jolly token]

Yeah, I guess it is more of Rx Merge

[brisk spruce]

 Busy = true;
 var roundEvent = RoundEvent.None;
 var handlers = RoundHandlers.Select(rh => rh.Run()).ToArray();
 // Aggregate all AsyncEnumerables together
 var handler = IAsyncEnumerableExtensions.Concact(handlers);

 // Process all the events and aggregate them into a single flagged enum
 await foreach (var e in handler)
     roundEvent |= roundEvent;

 return roundEvent;

Aight, now it looks a little bit cleaner :3

[hardy sentinel]

riight, so it would need to be rh.Run()() to execute the actual async code, mb

[brisk spruce]

nah .Run() returns an IAsyncEnumerable

Honestly maybe I should call the method .SelectMany lol, thats the closest to what it is for Linq

Im turning
IEnumerable<IAsyncEnumerable<T>> into IAsyncEnumerable<T> effectively

[jolly token]

Should be roundEvent |= e ?

[brisk spruce]

oh nice catch thanks!

Im surprised that VS didnt highlight the fact I wasnt using e there O_o

[jolly token]

I think it should be called Merge, pretty much https://github.com/dotnet/reactive/blob/main/Ix.NET/Source/System.Interactive.Async/System/Linq/Operators/Merge.cs

[dusty glacier]

can someone help me with this, I have a point (red one) on a slope and another one at the same height (blue one) how do I calculate the third point's (yellow one) position if i only know the position of red and blue dot and have the surface normal of the slope, im working in 3 dimensions

[dusty wigeon]

Use Vector3.Dot to project the Red/Blue vector on the Red/Anypoint point on vector.

Oh nvm, that would give you an other point.

[dusty glacier]

you mean something like this

[dusty wigeon]

You might want to use trigonemetry here.

[dusty glacier]

tried it, but it doesn't provide correct results

[dusty wigeon]

Function I did the other day.

private static bool IntersectLineLine(Vector3 lineAPositionA, Vector3 lineAPositionB, Vector3 B, out Vector3 result)
    {
        Vector3 A = lineAPositionB - lineAPositionA;
        float AMagnitude = A.magnitude;
        float BMagnitude = B.magnitude;

        if (BMagnitude == 0 || AMagnitude == 0)
        {
            result = Vector3.zero;
            return false;
        }

        float dot = Vector3.Dot(A, B) / (AMagnitude * BMagnitude);

        if (dot == 0)
        {
            result = Vector3.zero;
            return false;
        }

        result = lineAPositionA + B.normalized * (AMagnitude / dot);
        return true;
    }

Not exactly sure of the validity, I did not use it a lot.

[dusty glacier]

I'm trying to boxcast the position for the feet ik system and the boxcast hit point is not at the same xz position as the legIK, and im trying to add or remove the needed height for it to be on the ground

[dusty wigeon]

Can you just raycast under the feet ?

[dusty glacier]

no, beacuze raycast checks a single point and there might be nothing under that point

[dusty wigeon]

You can use boxcast or sphere cast also.

[dusty glacier]

i do

wait a sec

do you see the problem, the sphere on the right is the hit position of the boxcast

i am aplying a correct feet height offset

[dusty wigeon]

Yeah, this is usually why you use raycast directly.

[dusty glacier]

and this happesn with raycasts

[dusty wigeon]

Not really an issue.

[dusty glacier]

?

[flint sage]

(or just do 2 raycasts)

[dusty wigeon]

I mean, you do not need to solve it at 100%.

You could always fall back to boxcat if raycast fail.

[dusty glacier]

that can cause twiching if im not mistaken

[flint sage]

The argument of ray/boxcast is pretty useless since it won't change the fact that it's not working for them

[dusty wigeon]

I was just pointing out something. I'm not forcing anything here.

This is what I have done and it was working pretty well.

[dusty glacier]

raycast are just not a good option rn

I have been hammering at this issue for the past 2 days so I'm really clueless

[dusty wigeon]

If you use Vector3.up always you can simply take the point at the location on your vector.

[dusty glacier]

can you explain that a bit?

[dusty wigeon]

Let's say your blue dot is (X,Y,Z), then find the coordinate on the vector at (X, ?, Z)

[dusty glacier]

the vector from red to the yellow dot?

[dusty wigeon]

Yeah

[dusty glacier]

you are right only Y is changed

but how do i find it

also take note that the slope I drew is actually a plane

[dusty wigeon]

Let's say that you have a vector that has the point (A,B,C) and the direction (X, Y, Z).
You can define the equation of the line by:
x - A = X * t
y - B = Y * t
z - C = Z * t

You can find t by using x - A = X * t or any other equation that you know the coordinate.

[dusty glacier]

I'll try this, thanks

[dusty wigeon]

You could also use trigo. (The law of sin)

[dusty glacier]

one thing, what are the lowrercase x, y, z for?

are they the slope normal?

[dusty wigeon]

x, y, z are the variable.

in the equation you have x and t as a variable.

It would be the position of your Vector3.up.

[dusty glacier]

Vector3 slopeNormal = hit.normal;
Vector3 HitPoint = hit.point;
Vector3 PointToProject = leg_transform.position;

[dusty glacier]

this is all the info I have

[dusty glacier]

I will try it and let you know

thx

[dusty wigeon]

You could also just do intersection line/line I gave you earlier.

Even better intersection plane line.

[dusty glacier]

im affraid that is a bit more than i can understand

[dusty wigeon]

 private static bool LinePlaneIntersection(Vector3 linePositionA, Vector3 linePositionB, Vector3 planePositionA, Vector3 planePositionB, Vector3 planePositionC, out Vector3 result)
    {
        Vector3 P01 = planePositionB - planePositionA;
        Vector3 P02 = planePositionC - planePositionA;
        Vector3 P03 = planePositionC - planePositionB;
        Vector3 IAB = linePositionB - linePositionA;
        float sqrMagnitudeIAB = IAB.sqrMagnitude;

        Vector3 normal = Vector3.Cross(P01, P02);
        float denominator = Vector3.Dot(-IAB, normal);
        if (Mathf.Abs(denominator) < 0.001f) { result = new Vector3(); return false; }

        result = linePositionA + IAB * Vector3.Dot(Vector3.Cross(P01, P02), linePositionA - planePositionA) / denominator;

        return true;
    }

@dusty glacier

Even better, just use: https://docs.unity3d.com/ScriptReference/Plane.Raycast.html

[dusty glacier]

wait that might be excatly what im looking for

[dusty wigeon]

You could even use directly the general equation of a plane given that you know X, and Z.

normal.x * (rigidbody.x - point.x) + normal.y * (y - point.y) + normal.z * (rigidbody.z - point.z) = 0
y = (- normal.x * (rigidbody.x - point.x) - normal.z * (rigidbody.z - point.z) - normal.y * point.y) / normal.y

[dusty glacier]

using the plane raycast still gives wrong results

hit is the hit information of the boxcast it is marked as a white sphere in the first image

'leg_transform.position' is marked as the white box

[dusty wigeon]

Why you do not directly use hitPoint ?

[dusty glacier]

it applies some movement on the x or z axis and the leg starts flying around

and still produces incorrect results

[dusty wigeon]

To be honest, I feel that it is what you do with the result that is wrong.

Can you put a gizmo where the point is ?

[dusty glacier]

which point

the raycast hit

[dusty wigeon]

The point you are trying to find

[dusty glacier]

one sec

it is the white sphere

[dusty wigeon]

So the point is correct

[dusty glacier]

see it is not in line with the leg

it is moved to the right

[dusty wigeon]

Or is it not align with the other one ?

[dusty glacier]

it should be right below the box

[dusty wigeon]

What is transform.leg at this heigh ?

[dusty glacier]

in line with it on global Y axis

[dusty glacier]

something is wrong

as soon as the angle of the plane changes from 0 the position of the leg snaps to something else

[dusty wigeon]

What is transform.leg at this height ?

In relation to the hitpoint.

[dusty glacier]

do you mean the distance

?

[dusty wigeon]

Put two gizmo

[dusty glacier]

.

the leg transform is the white box

hit point is the sphere

they are not in line as they should be

[dusty wigeon]

Pretty sure there is something wrong in your entry data.

Try it in a more isolate environment.

[tired fog]

Hi

[dusty glacier]

hold on

[dusty glacier]

hi

[tired fog]

I want to render a mesh at local position 0, any idea why this wouldn't render it?

        var param =  new RenderParams(GrassMaterial);
        Graphics.RenderMesh(in param, GrassMeshes[0], 0,  Matrix4x4.Translate(Vector3.zero));

GrassMeshes[0] is just a mesh, trying to bug fix and getting to the root of the problem

Nvm, it was working, I just couldn't find it

[wanton adder]

Hlo guys, I am working on a top down shooter game. It's my first game that i probably will get finished. I want to figure out how do i add recoil and make a modular gun script

[dusty glacier]

You were absolutley right, the way i added feet offset messed with everything

thanks for the help, in the end i used the sin method

[soft elm]

damn that hurts my foot lol

[little sluice]

Question:
I'm creating a mesh and saving it via AssetDatabase.CreateAsset().
When i do that again and overwrite it, all existing references to that asset no longer work. (Missing mesh)
However: On a file level, it stays exactly the same. Same meta-file content, same guid, same file id. Nothing changed. I change the scriptableObject to reference the new asset and it is exactly the same as before.

Whats the correct way to overwrite Assets?

Not a bit of difference according to git. Mhhh

[jolly token]

LoadAssetAtPath then modify the returned object

Then SetDirty & SaveAssets

[little sluice]

Overwrite Asset, Open ScriptableObject -> Missing ( Mesh)
Overwrite Asset, Close Unity, Open Unity, Open ScriptableObject -> All good

why easy when it can be weird

[jolly token]

Don’t overwrite

[brisk spruce]

does unity have a way to combine IEnumerators together simultaneously for coroutines?

A sort of "When All" operator

(not just yield returning A then B then C, as my understanding is B wont start til A finishes)

or something like

yield return Coroutines.WhenAll(A, B, C)

[jolly token]

Run them as independent coroutine and yield coroutine

[brisk spruce]

what do you mean by the last part? yield coroutine?

[jolly token]

Coroutine work1 = StartCoroutine(Work1());
Coroutine work2 = StartCoroutine(Work2());
yield return work1;
yield return work2;

[brisk spruce]

I see, hmm, wish there was a way to combine em into 1 easily

I guess I could make a simple extension method

public static IEnumerator WhenAll(params IEnumerator[] coroutines) 
{
   foreach (var coroutine in coroutines)
      yield return coroutine;
}

[regal olive]

I've got a really bad memory leak that persists even when I end play mode.

It says to account for the 21gb leak I need to use a platform specific profiler? But I don't understand what it wants / how to do what it wants

[jolly token]

That's different

[regal olive]

Does that mean a platform built version with the profiler in it?

[jolly token]

You need to StartCoroutine so it can run separately

[dusty wigeon]

You got issue with native object.

What is your platform ?

[regal olive]

Windows 11

[dusty wigeon]

Then you should use Memory Profler that are specific for Windows.

[regal olive]

You mean outside of unity or inside it? If inside it then I'm not sure how to get / use a windows specific one

[dusty wigeon]

I'm not exactly sure what to use with Windows Specific, however each platform, PS, Xbox, Android, IOS, Switch has their own profiler.

I would guess you should look what tool there is with Visual Studio.

Also, you can simply guess what your issue.

If your game is not large.

[regal olive]

It's pretty early on right now but if the leak persists when not in play mode then it's an issue with the editor?

[dusty wigeon]

You should always profile builds*

[regal olive]

There hasn't been a "build" yet

Unless you are just refering to build as the current version. It's not been built into an exe I mean

So from what I gather, I need to look into memory profilers in my IDE / Rider

[dusty wigeon]

Not necessary your IDE.

[dusty wigeon]

You should build it to be able to analyze correctly memory issue.

Otherwise you gonna have a lot of noise coming from the Editor.

[regal olive]

I guess it makes sesne to do that first

[brisk spruce]

right, hmm

the issue is you cant for sure say if you have already started the coroutine or not, I actually dislike that API

[dusty wigeon]

If I were you, I would look into: https://devblogs.microsoft.com/visualstudio/analyze-cpu-memory-while-debugging/

(I cannot guarantee you that it is what you need)

[brisk spruce]

theoretically StartCoroutine should return something distinct from an IEnumerator

the fact that a started coroutine and a non started one cannot really be distinguished is sort of a problem, design wise

[regal olive]

For the build I'm using Deep Profiling support and autoconnect profiler?

[dusty wigeon]

Not necessary.

[brisk spruce]

if you get handed an IEnumerator, you dunno if its the coroutine itself, or the coroutine starter

[dusty wigeon]

Usually, you manage the logic with a wrapper. (Started or not Started)