archived-code-advanced

[scenic forge]

I'm not sure what it's supposed to do, but it seems like you are looking for the Foo<T> where T : Foo<T> pattern.

[tropic vigil]

The point is that I don't know the T.

[upbeat path]

but you can find the T by examining the returned object

[scenic forge]

Using the pattern, your subclasses can now implement a specific return type.

abstract class AttributeType<T> where T : AttributeType<T>
{
    public abstract T ModifyValue();
}

class IntAttributeType : AttributeType<IntAttributeType>
{
    public override IntAttributeType ModifyValue()
    {
        // ...
    }
}

[tropic vigil]

AttributeType at = new AttributeType<MyClass>();
---
object o = (object)at.ModifyAttributeValue();

This won't work since ModifyAttributeValue is inside AttributeType<T>. at in this case is the type of AttributeType. You can't use it like this unless you cast at to AttributeType<T>, but you don't know the T at the moment.

[tropic vigil]

Gonna check

[upbeat path]

good point, it would need to be on AtributeType so object is your only return type

[stuck plinth]

hmm why do you want to call ModifyAttributeValue without knowing the type? that limits what you can do with the result quite a bit

[tropic vigil]

On second thought this solution misses the point of what I was planning. I wanted to have class IntAttributeType : AttributeType<int>, which would return int value from ModifyValue. But I think I will find a place for your trick in another place.

[tropic vigil]

I wanted to get a value and instantly use it on an object of the same type for some other method.

[scenic forge]

I feel like some context is missing, might be a better idea to share your current code.

As I see it, your current implementation of ModifyValue in your non generic base class AttributeType cannot be safe without knowing T, yet the inherited generic class can make T be anything.

[tropic vigil]

The concept is to create some system for generic attribute scriptable objects. Depending on the attribute, it could be int, float or something else. It worked nicely until I wanted to create an ability that transfers a value of some attribute from one object to another. E.g. you could drain enemy's health or mana. I think implementing such "transfer" functionality directly into AttributeType could do the job, but I was hoping for a solution that wouldn't require modifying AttributeType's code each time I want to create a new kind of effect.

[fallow eagle]

Why does not Burrito's solution work in your case ? I would assume that the operation made by ModifyValue() would change depending on the type of the object

you just want this, no ?

abstract class AttributeType<T>
{
    public abstract T ModifyValue();
}

class IntAttributeType : AttributeType<int>
{
    public override int ModifyValue()
    {
        // ...
    }
}

(Just trying to understand)

[tropic vigil]

I want to be able to do something like this:

void SomeMethod (AttributeType someAttribute)
{
  someAttribute.ModifyValue (someValue);
}

But the method ModifyValue would be inside of AttributeType <T>, which derives from AttributeType, so I can't access it from AttributeType unless I cast it to AttributeType <T>, but I need to know the T in order to cast it.

[fallow eagle]

OK. Thanks, it's more visual that way haha

[stuck plinth]

adding it to AttributeType is probably the route i'd go with, there's not much benefit to making it generic if you're casting back to object for every operation anyway

i don't know exactly what you need but i guess you'd only need to put enough building blocks in AttributeType that you could implement more specific stuff in terms of those?

[scenic forge]

^ This is honestly what I'm interested to see.

[tropic vigil]

Adding more building blocks seems to be the safest way, so I will most likely stick to it.

[dusty wigeon]

In this situation, what is the type of someValue and where it comes from ?

[pure berry]

Hello, I'm trying to implement a StateMachine for my character's controls. I use rigidbody.
I need a basic class or interface from which to derive the various states my character will be in. But I've run into a dilemma.
In your opinion, is it better to use an abstract class or an interface to define a State Machine ?
From what I've seen on the internet, it's the interface that comes up most often. But the tutorials on the iHeartGameDev YouTube channel, which I think are very good, use an abstract class.
What do you think I should choose ?

[upbeat path]

An interface only makes sense if you have multiple characters which each implement their own behaviours

[stuck plinth]

there's really not that much of a difference conceptually, go with whatever is most practical for your code

[pure berry]

Could you elaborate on that ? I want a StateMachine just for my character. So if I go by what you say I should use an abstract class ?

[pure berry]

Got it. So in terms of good practice you think the two are equivalent ?

[upbeat path]

the point is, if you use an interface you MUST implement the methods for each class that inherits it. for an Abstract class this is not a requirement. If you are only implementing a State Machine for one Character then a standard class is more than enough

[stuck plinth]

there's a lot of overlap between them, i guess generally i'd go for interfaces as a preference but they're two ways of achieving more or less the same thing and the differences are stuff like multiple inheritance and whether you need fields in the base class which probably don't matter here

[upbeat path]

@pure berry If you have only one Character that needs a State Machine then the question of Interface or Abstract class does not even arise

[pure berry]

OK, thanks for the reply. I will take this into consideration

[pure berry]

OK, I understand. Thanks for your reply

[pure berry]

Oh so using an abstract class for one character is an insdustry standar ? This is what most game dev use you mean ? Because I've already done both, but never in the context of creating the movement of characters video games.

[upbeat path]

Absolutely not. why would you create either an Interface or an Abstract class which is only going to be used in one place?

[stuck plinth]

this is about a character with multiple different state classes implementing a common interface

[upbeat path]

that is not the way I read the question

[tropic vigil]

The same type as T in someAttribute. I think I could add a Type property that would return it. 🤔 It would look ugly, but would do the job.

[pure berry]

I want to build a StateMachine to define the different states of my character. This is made up of a main class or interface. And from this derives the different states that will redefine the character's movement logic.

[upbeat path]

but, I guess, the character here is irrelevant. you are talking about building the State data structure?

[dusty wigeon]

Yes, but what is the data structure that holds that ? Do you hold your object in a list<object> ?

[pure berry]

Yes, if I decide to work with interfaces, that's what I'll have to do. It's mentioned on this site: https://gamedevbeginner.com/state-machines-in-unity-how-and-when-to-use-them/

[pure berry]

If when you say "State data structure" you're referring to the way we manage the various states of an object, yes that's what I'm talking about. But the link with the character is that depending on the state, the character will follow a different movement logic. I'm talking about movement logic because that's what I'm working on at the moment, but it could be something else.

[stuck plinth]

well really i mean "interface" in the general design sense, an abstract base class can be considered a type of interface in this context 😄 whether you use a C# interface or a base class is still basically down to convenience IMO, i've done it both ways for player state classes

[upbeat path]

OK, I misunderstood what you were trying to do. There is no reason not to use both Abstract and Interface. What I would do is evaluate where the commonalities are. For those states which have much in common implement them in an abstract class, for those with little or no commonality implement them in an Interface. But as @stuck plinth says, there is really very little conceptual; difference

Basically if you find yourself duplicating code you have chosen the wrong path

[pure berry]

Oh ok i understand what you say 😁

[pure berry]

Ok got it. Thank your for your answer.

Thank you very much @stuck plinth and @upbeat path . It was really helpful. Have a nice day ! 😁

[tropic vigil]

Allright, lets do it the hard way:

[CreateAssetMenu(fileName = nameof(TransferAttributesEffect), menuName = "Interactions/Effects/TransferAttributesEffect")]
public class TransferAttributesEffect : InteractableObjectAlterationEffect, ICellInteractHandler
{
[SerializeField] AttributeType [] attributeTypes = new AttributeType[1];
[SerializeField] AttributeType.ValueType valueType = AttributeType.ValueType.currentValue;
[SerializeField] Modifier [] modifiers;
[SerializeField] TargetConverter targetReceiverConverter;

public override void InteractWithTargets(IInteracter selectedSource, InteractableCell[] selectedTargets)
{
  var covertedTargets = GetConvertedTargets(selectedSource, selectedTargets);
  var transferReceivers = targetReceiverConverter.ConvertTargetsToInteractables(selectedSource, selectedTargets);
  foreach (var covertedTarget in covertedTargets)
    foreach (var transferReceiver in transferReceivers)
      foreach (var attributeType in attributeTypes)
        attributeType.ModifyAttributeValue(transferReceiver, valueType, - attributeType.ModifyAttributeValue(selectedSource, covertedTarget, valueType, modifiers, multiplier));
}
}

I just have a bunch of AttributeTypes and I try to use their generic value as an argument for another generic method from the same AttributeType. I don't know what T I'm ending with, but I'm sure it's the same as AttributeType's T.

[dusty wigeon]

Is there any particular reason why you do not use SerializeReference for your array of AttributeType ?

It seem like you handle the type without actually using the type. By using an enum.

[tropic vigil]

AttributeType is a ScriptableObject. I'm not sure if it works well with ScriptableObjects.

[dusty wigeon]

What is the attribute type ? Float, Integer, etc. ?

[tropic vigil]

It can be different for different ScriptableObjects. There is a separate type for each class (e.g. IntAttributeType : AttributeType <int>), which makes it possible to create multiple types of them with the same basic logic.

[dusty wigeon]

Why are you not using something like Modifer<int> ?

[tropic vigil]

Modifiers here are supposed to handle the logic of modification, e.g. "increase by 3", or "reduce by 100% of source's power" or "increase by 50% of target's shields". So Modifiers here don't have a type.

[dusty wigeon]

It really seem strange that they have a type given that you cannot increase by 3 on a boolean type by example.

So, your AttributeType is what define the operation ? Like summation, division, etc. ?

[tropic vigil]

True. I don't have planned adding support for bools so far, but if I made one, I would just leave NotImplementedException here and some comment/log to not modify bool that way. The logic should be implemented for types such as int, float and maybe some other numeric types.

[dusty wigeon]

So, your modifier defines the operation to be executed and the AttributeType is the one actually executing it ?

[tropic vigil]

AttributeType is mostly for directly handling data. Possibility of setting its min and max value, clamping the value accordingly, managing listeners (when it's modified), changing the values in the correct order (e.g. increasing max health before current health during addition, but in other way during reduction).

[tropic vigil]

I would treat it more like a scalar.

[dusty wigeon]

So, attribute type represent Health, Mana, etc. ?

[tropic vigil]

Yep, stuff like that.

[dusty wigeon]

I am not sure why the AttributeType is not directly linked to the Modifiers.

By example, (Obviously, not exactly that)

public class Modifier<int>
{
  [SerializeField] private AttributeType<int> attributeToModify;
  [SerializeField] private int value;
  
  public void Execute(Context context)
  {
    Target target = context.GetTarget();
    Attribute<int> attribute = target.GetAttribute<int>(attributeToModify);

    attribute.Add(value);
  }
}

The issue seem to be that you have two different Type which could be different (Impossible to know at compile time).

You would either need to restrict further at compile time or make runtime validation/cast.

[tropic vigil]

Currently, I'm just casting it if needed (to calculate percentages I need floats anyway). I'm using numeric values, so there are no issues so far. But in the case of other types, I suppose I will simply make separate methods and call them (e.g. in Vector2AttributeType I would call Modifier.Vector2Fraction (IInteracter selectedSource, IInteractable selectedTarget, Vector2AttributeType attributeType) instead of Modifier.FloatFraction (IInteracter selectedSource, IInteractable selectedTarget, AttributeType attributeType)).

[dusty wigeon]

You do you. From what I see, you wont have any choice to use object and cast given how you use your data/store them.

[tranquil terrace]

im writing my own transform / physics system for a fighting game, because i cant use unity's float based transform system for calculations. i was thinking of making a base class "entity" which would store the positions and velocities of an object such as a player or projectile. I'm just wondering what else I should store, if anything?

[soft moon]

What is the problem in using Unity's float based transform system?

[tranquil terrace]

from what i've been told floating point numbers are not deterministic and are subject to errors between different machines. as such i've been recommended using a fixed point system

[silent dagger]

Hello there...

I need some math smart guy because I'm lost 😅

Hello there 🙂
I'm trying to make a sort of solar system, but the game is based on a grid, so I'm using this to calculate the position for each planet

for (int i = 1; i < celestialBodies.Count; i++)
 {
     float angle = CalculateOrbitPosition(celestialBodies[i], currentDate);

     float distance = cellSize * (i);
     float x = centerX + distance * Mathf.Cos(angle);
     float z = centerZ + distance * Mathf.Sin(angle);

     celestialBodies[i].celestialBodies[0].transform.position = new Vector3(x, 0, z);
 }```

But the problem is that I have to keep the planets centered on a 2D Grid... which is "virtual" because I won't use trigger.
So the cellSize defines that theorical cube, and I've stored it in a bidimensional int [,] ... any advice? 🙏😅
Ok... I know the center of the grid, so if  `i` = 1 define also the cell closer to the center which is the sun ...  and here I'm lost 🤔

[sly grove]

I'm really not sure what you're asking to be honest

[silent dagger]

😅

[sly grove]

but if you want a grid based game I would highly recommend Unity's Grid component

it will take care of all conversions to and from grid coordinates/world space and is freely configurable

[silent dagger]

I'm making object orbiting around an object, but I want that are snap to a grid

[sly grove]

that's very confusing

I can think of at least two different ways to do it though.

1. You might have a "real" position with full floating point fidelity, and when drawing the things on screen you just snap the position to a grid for presentation purposes.
2. You reimagine the concept of "orbit" in a grid-based setting which means throwing physics and such out the window

[soft moon]

I see, so would you like to fully recreate Unity's physics with another data type?

[silent dagger]

I need the planets switch from cell to cell, should follow the orbit, but stiked to the center of the cell.

[sly grove]

I'm unclear about what you mean about them "sticking" to the center

if they stick to the center, they will not be orbiting

Is it more that you just want to be able to identify which grid cell they are in?

[silent dagger]

sticked to the center of the cell. 😅

[sly grove]

and let them have a normal orbit?

Because otherwise, this just makes no sense to me

[silent dagger]

yea!

[sly grove]

Well as mentioned, just use the Grid component

You can use https://docs.unity3d.com/ScriptReference/GridLayout.WorldToCell.html this to convert from a world position to the grid coordinate

[silent dagger]

Ok I will investigate on the grid component not even know that exists

[sly grove]

https://docs.unity3d.com/Manual/class-Grid.html

[silent dagger]

Ok I'll start studing it thank you 🙂

Sorry for my bad explanation, but I'm Italian and my English sux a lot 😅

[sly grove]

(this isn't really an advanced code question by the way)

[silent dagger]

I was expecting some complicated array and trigonometry stuff 😛

[tranquil terrace]

not fully, but i need simple things such as collisions and velocity and stuff

[soft moon]

Can your game logic be implemented with Unity's built-in physics?

[tranquil terrace]

it can, but it wouldnt be good for online multiplayer and stuff.

i need to make my own physics for this case

[soft moon]

Why?

[tranquil terrace]

its not deterministic like i said before

[soft moon]

Do you only mean errors when calculating floats?

[tranquil terrace]

yeah

[soft moon]

Do you know how huge they are?

[tranquil terrace]

they are not cross platform deterministic, which means different machines may have different results for a simulation.

[soft moon]

Why does everyone still use Unity's physics?

[tranquil terrace]

for their case its probably fine, but im doing online multiplayer eventually so i need bit-perfect simulations across any platform

apparently 😅

[soft moon]

I think you should reconsider it

I wonder whether you'll be able to recreate Unity's physics, even party, in a better way than Unity has done it. Also, this may take quite some time, especially, if you're asking for the hints from the start

I'm also not sure whether you have already tested this calculation impact across different platforms

[tranquil terrace]

i think i may have made what im trying to do seem more grandiose it is. I'm not trying to make proper rigibody simulations with like 100s of cubes or whatever. at max, 4 things are going to be on screen and stuff. I'm already in part doing physics kinematically in my current system

[soft moon]

There is no need to create your own physics unless you're e.g. dealing with a grid-based game, where Unity's physics may not act accurately

[silent dagger]

Easy as can be Thank you! 🍻

[molten minnow]

If you are doing multiplayer, only one client (or server) will have authority over simulations, so making them deterministic will not matter as much
Unless you are making something like Trackmania where each session must be deterministic and comparable

[tranquil terrace]

its a fighting game, so i'm only streaming inputs to the server, which is why determinism matters for for me

[molten minnow]

So only one machine does the calculation, what will determinism achieve?

[tranquil terrace]

fighting games only send the inputs to the server, and each machine does their own simulation. this is why i need determinism

[molten minnow]

Your server should sync all states, and clients will only display visuals

[rugged radish]

most of fighting games are p2p (if not all of them), and don't involve physics simulation at all

[tranquil terrace]

wdym? stuff like moving up and down? im doing a platform fighter as well, with some projectiles that are affected by gravity

[rugged radish]

oh, a platforming fighting game, not too familiar with those
traditional fighting games operate exclusively on animation frames

[molten minnow]

Still, your server should override all simulations and sync visuals. Without it you are open for cheating

[sly grove]

awesome!

[tall ferry]

That's not really how cheating works

[tranquil terrace]

how? if the server only streams inputs, how would you move the players around and cheat?

you dont have direct control of the character itself, as well as its position

[molten minnow]

Sure, it won't give you advantage but just desync everything completely

[tranquil terrace]

i suppose. thats what rollback is for i guess

[tall ferry]

I feel theres a major difference in what you asked and what's actually being discussed. You're making this for a replay system right?

[molten minnow]

Client will receive a state of the opponent, so it shouldn't be a problem to also get it's own state

[tranquil terrace]

i'm making this for the replay system, and that's going to be a precursor to online multiplayer. I've been recommended to make the replay system, because if I can perfectly recreate a match from inputs alone it apparently makes it viable to for online multiplayer

[molten minnow]

Yeah, it makes sense for replays, but it's not a requirement for multiplayer

[tall ferry]

Well there might be slight differences tbh. Many multiplayer techniques are made specifically to fix desync issues which will happen no matter what. This is just a cause of networking. Meanwhile a replay would exist locally

[tranquil terrace]

i'm in a fighting game development discord server, and they have all attested to using fixed point transform system instead of floating point ones

which is why i wanted to give it a go

[tall ferry]

I'm not saying you shouldnt use this kind of system, just that a replay system alone is gonna have some differences. Even with a fixed system, networking a game will require more to fix desyncs

[tranquil terrace]

thats fine, i understand that. its meant to be a good start, though. i plan to use a networking system called lockstep, where it waits for all inputs to come through before simulating anything

i know im oversimplifying it a bit though 😅

[tall ferry]

Have you looked into what other games, similar to what you're making, are doing for such a system?

A lot of fighting games systems are well known/somewhat documented

[tranquil terrace]

networking or transform system?

[tall ferry]

Their movement and hit detection system, which I believe your original question was for

[tranquil terrace]

oh right. I've not looked, i'll check that out. my original question was "what sort of things should i store in a base 'entity' class for a transform system"

my idea is that i'll calculate the mvoement in fixed point system, and then show the transforms in unity's system

seperating the simulation from the visual or whatever

[solar valve]

Hello everyone, I am currently trying to work on a school project with some friends of mine. We want to recreate a game map on unity HDRP. The issue I am running into is that the base project of HDRP is already 2GB. Which makes it so that when I want to push to unity, to work collaboratively. I get an error from it, for it being too big, I believe the maximum is 50MB or something. I have made a bit of reserch, and there seems to be something called LFS that might help, but it seems that to use it, I need to use Git Krakern's terminal which I am not that experienced in. Anyone that could help with this?

[swift orchid]

In git you usually add to ignore the library folder

[solar valve]

I have that, but there are still files in hdrp that are just too big, or I guess they are not being ignored if that is the case

[swift orchid]

You can sort big files into one folder and ignore that folder, and copy that folder in google disk to share with friends, but it will work only if it is some static things for example 4k texture

[sly grove]

there is no reason you should be pushing "the base of HDRP" into your version control system

show a screenshot of your repository

[bitter crescent]

Hi everyone, can I promote my Unity package here?

[sly grove]

Show the files

just the root directory

[solar valve]

How do i get hdrp in my project without, starting up with the base unity hdrp project?

[sly grove]

I don't care about the commit

[sly grove]

just install HDRP from the package manager

but - show the files in your repo

you can just take a screenshot of the repo from github

[solar valve]

this?

[sly grove]

yeah see - you committed the Library folder

and the Temp folder

so you screwed up making the repo

that's why the repo is so large

[solar valve]

So the git ignore is not doing anything?

[sly grove]

You probably added gitignore AFTER committing the library folder

or you just didn't set up an appropriate gitignore file

[solar valve]

hmmmmm okay okay

[sly grove]

the easiest thing to do here will be to just delete the repository and create it fresh

[solar valve]

but you say that if i just start a normal unity 3d project, and then just add the hdrp package. my project then converts into an hdrp project?

[solar valve]

Yeah I will probably do that

[sly grove]

https://docs.unity3d.com/Packages/com.unity.render-pipelines.high-definition@17.0/manual/install-hdrp.html#install-the-high-definition-rp-package
https://docs.unity3d.com/Packages/com.unity.render-pipelines.high-definition@17.0/manual/Material-Upgrade.html

I mean it's easier to just start with HDRP

[swift orchid]

You will need to setup harp properly

[sly grove]

don't see any reason to do it that way

[wet sail]

hello. why does GPU Instancing not cause any performance boost? im spawning 65k cube, with standard material and "GPU Instancing" ticked on. i get the same 7-10fps.

[sly grove]

Not a code question

[wet sail]

where do i post ?

[sly grove]

not here...
#💻┃unity-talk if nowhere else

[steel snow]

bit confused i tried to replicate some code from unity's documentation and its not working in my test script. im trying to load an asset reference if the field is set like so:

public class LoadWithReference : MonoBehaviour
{
    [SerializeField]
    AssetReferenceT<MeshAsset> _meshAssetReference;
    MeshFilter _meshFilter;
    MeshRenderer _meshRenderer;

    void Awake()
    {
        _meshFilter = GetComponent<MeshFilter>();
        _meshRenderer = GetComponent<MeshRenderer>();
        if (_meshAssetReference != null)
            Set(_meshAssetReference);
    }
    public void Set(in AssetReferenceT<MeshAsset> assetReference)
    {
        if (_meshAssetReference != null) //release if already set
            _meshAssetReference.ReleaseAsset();

        _meshAssetReference = assetReference; //set it again
        AsyncOperationHandle<MeshAsset> handle = _meshAssetReference.LoadAssetAsync<MeshAsset>();
        handle.Completed += OnCompleted;
    }
    private void OnDisable()
    {
        if (_meshAssetReference != null && _meshAssetReference.IsValid())
            _meshAssetReference.ReleaseAsset();
    }

    //...etc
}

currently the field is not set and yet when i press play i get:

UnityEngine.AddressableAssets.InvalidKeyException: Exception of type 'UnityEngine.AddressableAssets.InvalidKeyException' was thrown. No MergeMode is set to merge the multiple keys requested. Keys=, Type=MeshAsset
OperationException : ChainOperation failed because dependent operation failed```


The error points to this line `AsyncOperationHandle<MeshAsset> handle = _meshAssetReference.LoadAssetAsync<MeshAsset>();`

Which i don't understand, why is  that function even occurring since the field is not set so it should be null 🤔

[sly grove]

You'll want to check _meshAssetReference.IsValid() before trying to load it

[steel snow]

but shouldn't != null be sufficient?

the field hasn't been set to anything in inspector

it inherits from Unity's Object so it should be able to be null right?

[sly grove]

AssetReference is a struct

nvm I'm wrong it's not

but still

it's not going to be null

it's just going to be invalid

Unity serializer doesn't leave things null

And it doesn't inherit from UnityEngine.Object.

[steel snow]

oh i misread the source it restricting the generic to object ok that makes more sense whats interesting is unity's doc uses null check https://docs.unity3d.com/Packages/com.unity.addressables@2.2/manual/asset-reference-create.html

[echo venture]

Hey guys, what is the best way to find the root cause of this issue?

[long ivy]

did you extract a stack trace? That would be the place to start

There's a utility under Tools in your screenshot to do it, and instructions in the logcat docs

[honest hull]

Is there a way to automatically call reimport on a bunch of comoute shaders to recompile them every time I change a line In a Cginc they use?
I know unity is supposed to handle it automatically but I’m running into a massive issue that makes compute shaders fail to recompile automatically, forcing me to reimport them all manually: #archived-shaders message

[honest hull]

nvm sorry, figured out that setting the shaders all to use the platform preprocessor fixes it

[vale spindle]

Hello. I'm working on a 3D character controller. Right now I'm not using rigidbody or collider, movenet is only done through scripts. The colliding is done via raycasts (its a cylinder collider). Might sound stupid but I need it this way to have full control over player's surroundings and velocity. So i'm just modifying the transform every frame. It works, but sometimes i clip through objects (as expected). I could easily add the collider and fix it, but i dont want to because i dont need it. What should i do?

The alternative would be to add rigidbody and collider back, but instead not use the raycasts. But i need them to get surface normals. Maybe there's some other way? OnCollisionEnter doesnt cut it

[hushed fable]

https://docs.unity3d.com/ScriptReference/Physics.ContactModifyEvent.html can be used to control the contacts generated by PhysX.

[winged olive]

Hello, since yesterday when I press the ctrl key my scene blinks and it prints this in the console, Im using Unity 6 Preview

[glass anvil]

I'm re-writing my managed voxel renderer to use jobs for both the data and rendering itself.

I populate a collection (I've tried NMHM which required a huuge capacity due to collisions(?) like 10x the known worst case), NHM, now NL)

with this collection I can sample the noise of any voxel values neighboring the chunk border. This way I can clip chunk borders without needing to know anything about/load the neighbor chunk itself.

I schedule that to run and when it's finished I chain the render job to it..

I use .AsParallelWriter when writing the skirt, and AsParallelReader when reading the skirt. I've also tried a normal writer/reader.. but I'm guessing if I want it to work correctly in a IJobFor it needs to be parallel?

The problem is with the attached photo. I've debugged the positions and made sure the skirt values are correctly encompassing the entire chunk.. but for some reason here it's only picking up some values?

        private IEnumerator RegenerateRoutine()
        {
            _chunkDataJob.Skirt.Clear();

            _chunkDataJobHandle = _chunkDataJob.Schedule(ChunkSize, 1);
            
            // bad to call immediately? negates parallelism
            while (!_chunkDataJobHandle.IsCompleted)
                yield return null;

            _chunkDataJobHandle.Complete();

            // Provide render job with the new data
            _chunkMeshJob.voxels = _chunkDataJob.Voxels;
            _chunkMeshJob.skirt = _chunkDataJob.Skirt;
            _chunkMeshJobHandle = _chunkMeshJob.Schedule(dependsOn: _chunkDataJobHandle);

            // bad to call immediately? negates parallelism
            while (!_chunkMeshJobHandle.IsCompleted)
                yield return null;

            _chunkMeshJobHandle.Complete();
            
            // advanced mesh api (much faster(?))
            _chunkMeshFilter.mesh.SetMesh(this);
            _chunkMeshCollider.sharedMesh = _chunkMeshFilter.mesh;
        }

Very new to jobs/burst but I've been learning a lot on github, the chunk itself is rendering fine without using any parallel writing, it's only when I try to write/read do I get this jagged appearance.. I know my face culling logic in the renderer itself works as this is based on a managed version. It definitely seems like the collection isn't writing, or reading the right values

[glass anvil]

it looks like it's being read when the data isn't fully finished writing, but how would that be possible if I'm waiting for completion/chaining dependencies ?

to be clear the mesh job is just an IJob, the data is IJobFor which makes sense for it

[untold moth]

1. There's no point making a dependency if you're waiting for the job to complete anyway.

[glass anvil]

(totally, just trying everything I can at this point)

[untold moth]

What does your job look like?

Just share the whole file correctly. Upload to a paste site. !code

[thorn flint]

Posting code

📃 Large Code Blocks

Use 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 format as C#, add cs to the first line:
```cs
// Your code here
```
Add a comment with a line number if there is an error message.

[untold moth]

A bit too complex to find the issue by just looking at the code. I'd step through the code with a debugger and confirm that all the values and calculations are what you expect.

[rapid owl]

Hi all,

Am using URP and seem to be having some trouble passing depth to my compute shader, I have called:

        _computeShader.SetTextureFromGlobal(_kernel, "depthTexture",  "_CameraDepthTexture");

And also enabled depth in my URP options and end up getting "Property (depthTexture) at kernel index (0) is not set"

Does anything else have to be enabled? I checked the frame debugger and it seems like something exists at least

(Not sure what the right channel for this was so apologies)

[untold moth]

Most likely a race condition

[untold moth]

Then debug it. If there's a racing condition, you're probably writing to the same index ranges from several threads. Confirming the values and indices involved with a debugger would help you identify that.

[untold moth]

Speculating is not gonna help. You should debug properly.

[untold moth]

Why can't a debug log be helpful? It totally can. And If it really doesn't, then stepping through the code, or using log breakpoints can.

Of course you can

This is just code running on separate threads. You just need to disable the burst compiler

And weith some tinkering you can probably even debug burst compiled code

If you don't know how to debug properly, then ask about that, instead of expecting people to run your code in their heads to find the issue.

This video covers some of the questions you might have about debugging a job:
https://www.youtube.com/watch?v=nou6AIHKJz0&ab_channel=LeeHammerton

[visual shard]

hello can someone help me in putting ads in my game i have no experience in it and tried several times from many documentations and videos but never succeded if someone has done it and wants to help, DM me i'd be very gratefull

[soft moon]

Post a specific question, and your possibility of getting helped may increase. Also, I suppose, this refers to another channel. Try #💻┃unity-talk

[oak spire]

Can anyone help me do to damage model for flight sim?

[lament salmon]

If you have specific issues then ask about those

I doubt anyone is going to go "yes sure, I will help you build the entire system"

[oak spire]

I want to know how it works. Any documentation or something.

so I can implement myself

[plucky ridge]

hey all, is there anyway to programmatically add new agentTypes for the NavMeshSurface component baking? I know I can define it using the editor, but i would like to programmatically create agentTypes so i can loop through and re-bake terrains during the runtime.

[long cloud]

Hi everyone. I need help with Coroutines and web requests. I need the script to return data for other scripts to run. Is this possible? If so how?

[soft moon]

Yes. Events.

[long cloud]

Sorry for the late reply. How would I do that exacly?

I also don't want the frame held when doing the web request

[soft moon]

Events are used to be subscribed to, making other scripts be able to retrieve data from the one holding the event. Could you, please, explain your issue more thoroughly?

[long cloud]

I need a script that will request data from my website like version codes, ect. I want this scriptr to be callable by other scripts to get data when needed

[compact ingot]

Did you look at the examples in the documentation?

[soft moon]

I see, web requests is something I cannot help you with

[long cloud]

I did and couldn't find any help

[compact ingot]

Maybe ask a more specific question. The docs have a non-blocking coroutine example for a get request

[long cloud]

They do?

Problem solved thanks to Opera's Aria

[worldly pecan]

Hey guys, I have a C# question;

I have a list of structures, which I want to author and then put onto unmanaged memory(for jobs & stuff). I plan to do this by malloc-ing space for each structure and then maintaining a native array of pointers to these structures. These structures need to implement certain functionalities, defined by an interface. In managed-memory I author this as a list of interfaces. However to convert this to a pointer to unmanaged memory I run into some problems.

1. Firstly, to malloc memory I need to know the underlying type. I guess there might be a way with reflection.
2. Secondly, to get data from the pointer in a job I need to cast it back to the underlying structure. Interfaces are managed objects so I can’t cast it to an interface pointer unless I make the interface unmanaged somehow.

Are there any workarounds to this problem? Perhaps a different way to get structs to inherit functionalities so I can call them without knowing their specific type?

[sly grove]

What's wrong with a regular blittable struct and a NativeArray<MyStruct>?

[worldly pecan]

There's not one type of structure in the array. There's multiple different structs all of different strides that all implement the same methods

Like if I had an interface for a car(ICar) I can have a list of cars called List<ICar> in which I guarantee all of them have the same basic functionality--I just need to translate this to unmanaged memory.

[sly grove]

In that case there's an even bigger problem which is you won't even know how to index your objects in this array

[worldly pecan]

They're pointers to memory scopes I malloced

[sly grove]

unless you keep a separate dictionary of pointers/offsets into it

oh they're pointers

[worldly pecan]

the memory scopes aren't the same size though, but they don't need to be (I don't think so fundamentally)

[sly grove]

do you really need them to be different structures? It might be better to make a struct that's a union of them all and just make an array of those

with an enum flag or something denoting the type

basically - don't try to force polymorphism into the native world

[worldly pecan]

It'd be really wasteful memory-wise for me to do this. There's some structs that require a lot of memory and others that don't.

[sly grove]

how much is "a lot"?

and how many things will there be

and if so - you could store that extra data in a separate collection if those things are rare

[scenic forge]

If size is a concern, your struct can very well be just a tag and a pointer to another struct that contains more data for the specific type.

[worldly pecan]

I'm trying to do this for some entities(like animals). They have different behaviors and different stuff they need to keep track of. A certain animal needs to keep track of different data & different amounts of data than another animal.

I mean... fundamentally I should be able to malloc different sizes and store their pointers in a native array to use them in a job

[sly grove]

yes you fundamentally could but the things that are good for performance are not necessarily things we normally consider "clean"

there are large performance advantages to having all your data in a single colocated array rather than all over the place with pointers and malloc'ed locations

[scenic forge]

It's also just two different ways of writing the same code.

[sly grove]

Basically all the things that make OOP and polymorphism nice are actually anathema to performance and how computers actually work

[rugged radish]

I have something similar in my code. But instead of structs I store values in a big number of arrays for jobs / compute shaders. And in managed code I use a number of "view" classes and manipulators that change the underlying arrays directly.

[worldly pecan]

This is like completely abandoning OOP, I mean I could try something like that

[rugged radish]

yeah, I went pretty much data oriented with oop convenience

[upbeat path]

this seems like you are trying to implement C++ inside of C#, why not just write a native code dll

[worldly pecan]

I would if I knew how lol-- I can do some research if u have some pointers

[upbeat path]

Unity docs have a complete section on it

[worldly pecan]

Seems like a rabbit hole but I'll tell u guys if I figure out a way to do it

[upbeat path]

It's not dificult, I mean Unity devs built Unity that way so how hard could it be?

[solid magnet]

Hi, I have a game with levels and I have an editor inside the unity editor to make them but I don't know where to save them (I don't want/can't save them in different scenes etc...). I want to save them in a json (it's actually just some lists of points), problem is idk how to create a json file inside of the assets folder neither how to access it (the path), how do I do that? Or is there a better way than json to save lists of vector2D/bools/in

[worldly pecan]

Famous last words

[sly grove]

you could save it as json in StreamingAssets or you could use ScriptableObjects instead

[solid magnet]

Yeah I found that out but I have no idea how to use ScriptableObjects

[upbeat path]

if you want to save in the Assets folder look at Application.streamingAssetsPath

[solid magnet]

Thanks

[upbeat path]

then don't you think it's time to learn?

[solid magnet]

That's what I'm doing 🙂

[scenic forge]

It's honestly just two different ways of writing code, if you are deep in OOP you might have only worked with the OOP way:
(The following codes are in TS because C# doesn't have real nice union, but the examples are just to demonstrate how to write the exact same code in two different ways)

type Shape = {
    calculateArea(): number
}

class Circle implements Shape {
    constructor(
        private radius: number
    ) {}

    calculateArea() {
        return Math.PI * this.radius * this.radius
    }
}

class Rect implements Shape {
    constructor(
        private width: number,
        private height: number
    ) {}

    calculateArea() {
        return this.width * this.height
    }
}

const circle = new Circle(42)
const rect = new Rect(42, 69)
circle.calculateArea()
rect.calculateArea()

type Shape =
    | { type: 'circle'; radius: number }
    | { type: 'rect'; width: number; height: number }

function calculateArea(shape: Shape) {
    switch (shape.type) {
        case 'circle':
            return Math.PI * shape.radius * shape.radius
        case 'rect':
            return shape.width * shape.height
    }
}

const circle: Shape = { type: 'circle', radius: 42 }
const rect: Shape = { type: 'rect', width: 42, height: 69 }
calculateArea(circle)
calculateArea(rect)

The first way fundamentally relies on dynamic dispatch, which C++ implements it as virtual function table. If you want to get that in your native code, you will have to basically reinvent it yourself.
The second way however does not require that.

OOP encapsulates data and behavior together into classes, whereas functional separates them into plain data (with no behavior) and free functions operating on those data (behavior). It's not too difficult to change your code between the two styles once you understand them, and the functional style works a lot better with the constraint of Burst.

[worldly pecan]

I'm not very used to that paradigm--and that way is very similar to DOTS I believe. I wanna try to see if I can mimic OOP with some custom code first and then I'll do that if I give up ig.

[scenic forge]

You can, by basically reinventing dynamic dispatch yourself.

Burst supports function pointer, so instead of IFoo with a method DoFooThing(), you would have a Foo struct with two fields, one is a pointer to the data, and one is a function pointer. And calling DoFooThing would be calling the function pointed by the function pointer, with the first argument being the data pointer.

[solid magnet]

levelData is a ScriptableObject class, this script is not from runtime but it's only runned manually from the editor.
How can I make the values be permanents? I want them to still be available when I start the game

    public void SaveLevel(){
        if(levelData.levelPoints.Count >= level){
            levelData.levelPoints[level-1] = path.points;
            levelData.obstacles[level-1] = new List<Vector4>{};
            levelData.end[level-1] = new List<Vector3>{};
            levelData.levelType[level-1] = "default";
            levelData.minimumDriftDuration[level-1] = 0f;
            levelData.isNight[level-1] = false;
            Debug.Log(levelData.levelPoints[0][1]);
        }
        else{
            levelData.levelPoints.Add(path.points);
            levelData.obstacles.Add(new List<Vector4>{});
            levelData.end.Add(new List<Vector3>{});
            levelData.levelType.Add("default");
            levelData.minimumDriftDuration.Add(0f);
            levelData.isNight.Add(false);
        }
    }

[worldly pecan]

I don't recommend using Scriptable Objects to save data, but if you are--I believe it's something along the lines of
EditorUtility.SetDirty(scriptableObj);
AssetDatabase.SaveAssets();//
AssetDatabase.Refresh();

You should really save in another format though. Scriptable Objects are mostly for settings and shouldn't be changed

[solid magnet]

Well they're never gonna be changed at runtime if that's the problem

Anyway, thanks for the help

[worldly pecan]

oh then it's fine

[solid magnet]

Where do I put that piece of code please?

[worldly pecan]

When you want to update your asset. You can think of your instance being a copy of the scriptable object, and when you call that it copies it back. That's not really what it is but it's close enough in effect

[solid magnet]

Ok I'll try that

thanks

[upbeat path]

you will need a using UnityEditor;

[worldly pecan]

put this
using UnityEditor;

[solid magnet]

Thought I had it sorry

Got another problem (kinda related)

It only saves part of the data

I have problems with this part:

public List<List<Vector2>> levelPoints = new List<List<Vector2>>{};

First of all it does not show in the inspector (which might be the default behaviour for lists into lists, idk)
Secondly, this part doesn't seems to save for some reason...
I add things to it like that:

levelData.levelPoints[level-1] = path.points; //path.points is List<Vector2>

[fickle mango]

to show up in the inspector you can create your own class:

[System.Serializable]
public class Path
{
    public List<Vector3> pathPoints;
}```
And then have a List<Path> in your script.

It also makes it more readable and might fix your saving bug.

[solid magnet]

Ok thanks

[solid magnet]

public class LevelData : ScriptableObject
{
    //Vector2: x,y pos of the points
    //public List<List<Vector2>> levelPoints = new List<List<Vector2>>{};
    public ListsOfLevelPoints levelpoints = new ListsOfLevelPoints();
}

public class ListsOfLevelPoints
{
    public List<LevelPoints> list;
}
public class LevelPoints
{
    public List<Vector2> list;
}

Like that?

Sorry I'm still new to this lol

That throws me an error, cannot converts from LevelPoints.levelPoints to Vector2

AAh those errors come from the script saving them

But now it does show up in the inspector

Still trying to find how to access the values

Now how to add things

I figured it out

I don't know if it fixed my problem

But at least now it shows up in the inspector

Thanks a lot

It fixed it

Tysm

[steel snow]

im trying to update normals after editing vertices but i am getting this error:

System.IndexOutOfRangeException: ReadWriteBuffers are restricted to only read & write the element at the job index. You can use double buffering strategies to avoid race conditions due to reading & writing in parallel to the same elements from a job.
This Exception was thrown from a job compiled with Burst, which has limited exception support.

not sure why though, this is the setup:

var vertHandle = new Deform(vertexData, uvs, output, spline, span, scale).Schedule(output.Length, output.Length / 8);

var tris32 = meshData.GetIndexData<int>().AsReadOnly();
var normalHandle = new RecalculateNormals32Bit(output.AsReadOnly(), tris32, normals).Schedule(tris32.Length / 3, tris32.Length / 8, vertHandle);

wondering if passing in output.AsReadOnly() is the issue? though i don't see why because recalc of normals depends on the completion of the vert handle so it should be fine...

ah i have fixed it nevermind 🙂

[worldly pecan]

Ok, after a lot of brainstorming I came up with this solution. We can expose the functions that’s originally defined by the interface by using a function pointer. Then we can define a struct that holds these exposed functions as well as the original pointer which can be put in a native array and used in a job.

To address mallocing enough space for the data we can define a function that queries the size of the object. When we create a new object we need to assign the function pointers to the specific functions defined by the specific structs.

Is this the best way to do this? Probably not—tell me what you guys think though.

[compact ingot]

I don't think following pointers in jobs makes a lot of sense (performance/safety wise).

[worldly pecan]

I mean it's a tradeoff. OOP is often easier to define contained functionality with, and using 2-4 pointers in a job probably isn't the bottleneck

[long ivy]

Out of curiosity, what are you actually trying to do goal-wise? I would shoot this down in a PR so fast

[worldly pecan]

I'm trying to do a bunch of calculations for some agents(mainly pathfinding) in a Unity parallel job where the each execution thread(?) handles an individual entity

There's also other considerations about storage and disposal cause I'm doing this for a procedural world

[scenic forge]

Pointers won't be the thing that's hurting performance the most, but all the cache misses because all of your data is scattered everywhere.

[worldly pecan]

Actually--I don't think so, each thread(?) handles one entity at a time. So all data about that entity will be on the same core's cache so there shouldn't be that many cache misses

[scenic forge]

That's already a performance loss

[worldly pecan]

How so? Is that not how parallel jobs work?

[scenic forge]

No, well, not exactly

Usually your data are packed into a continuous region of memory, and when you schedule a job to run in parallel it has a batch count, so each thread processes batch count amount of work rather than 1 at a time.

It means that a thread when finishes one piece of work, the next piece's data is right there in the CPU cache and no miss.

[worldly pecan]

I understand that--but if each job is a lot of work, the main concern is the cache misses inside a long job rather than between the next one. I plan for each job to do a lot of work, and the cache miss between processing the next entity is far less of a concern than the cache miss inside a single job.

What I mean is that if you have a job that processes 10,000 jobs, and each one takes 0.01 ms to complete, cache misses between them matter a lot, but if you have 100 jobs that take 1 ms each/thread cache misses between them matter less than they do inside it

Also I don't plan to put a lot of memory in these entities either

[scenic forge]

I'm not sure that's quite true, your IntPtr obj will be all over the place in memory, so every work which needs to dereference that, will be an almost guaranteed cache miss.

[worldly pecan]

Yes but each job will only dereference it once and then process a long task

[compact ingot]

what is happening in that long work unit?

[worldly pecan]

Pathfinding mostly I think

[scenic forge]

Sure I guess, but yeah you have just effectively reinvented dynamic dispatch with a vtable and a pointer to the object.

[compact ingot]

and the data of the graph fits into one memory page?

[worldly pecan]

No--but the data is partitioned and it's complicated

I'll worry about cache misses with the graph later

[compact ingot]

you should worry only about them because it sounds like those are your bottleneck

[worldly pecan]

Yeah exactly--but I just wanted to answer my question from earlier; also someone who really wants to stick to OOP can use the reference

[tired dome]

hello all, how do i make this code function properly? i want to make an airship hover in place, and this code works fine if angular drag is set to some arbitrarily high number, but when i set it to 1 the airship flips out and starts spinning weirdly

private void ApplyFloatingPower()
{
Vector3 centerOfMass = rb.centerOfMass;
float upwardForce = -rb.velocity.y; // Invert the velocity vector to cancel out the vertical velocity
rb.AddForceAtPosition(new Vector3(rb.velocity.x, upwardForce, rb.velocity.y), centerOfMass, ForceMode.VelocityChange); // Apply the force as a velocity change

    // Calculate the current up direction of the object
    Vector3 currentUp = transform.up;

    // Calculate the desired up direction (Vector3.up)
    Vector3 desiredUp = Vector3.up;

    // Calculate the rotation required to align currentUp with desiredUp
    Quaternion targetRotation = Quaternion.FromToRotation(currentUp, desiredUp) * rb.rotation;

    // Smoothly interpolate towards the target rotation
    rb.rotation = Quaternion.Slerp(rb.rotation, targetRotation, 1 * Time.deltaTime);

    // Apply damping to the angular velocity to reduce oscillations
    rb.angularVelocity *= (1f - 1 * Time.deltaTime);
}

[flint sage]

Looks like ai

[tired dome]

? ah i did run this through chatgpt a few times, but each iteration it spit out just made things worse lol

[flint sage]

Also use proper code blocks

Some bits are missing

[flint sage]

Man this code is becoming more cursed the more I looked at it

[tired dome]

yeah sorry, physics is not my forte

[upbeat path]

set velocity to zero // Will not move
set gravity off // wil not fall
result - hovering

[untold moth]

That's definitely not an advanced channel issue.

[warm karma]

Hi. Does anyone know if is it even possible to create an attribute that makes it so when the method is called, it automatically adds that method's name to a list? Like for example something like this:

public class RandomClass
{
    [Replayable]
    public void DoSomething()
    {
      //whatever
    }
}

And when I do RandomClass(), as it has the [Replayable] attribute, it should automatically call like another method. Is that possible?

[compact ingot]

you would have to write a source generator that hooks into these attributes at compile time and extends your code to make these additional calls.

[warm karma]

hmmmmm okay. I think I could do something like that. Thanks

[steel snow]

any one here know the algorithm to calculate normals for a mesh ?

ive tried two methods and it doesn't produce the same as unity's built in one i have no idea which algorithm they are using but their solution works

[compact ingot]

Have a look at the implementations here https://github.com/gradientspace/geometry3Sharp

[steel snow]

is this unity's ?

[compact ingot]

No

[steel snow]

ah whats the bet it uses the same one i already use

[compact ingot]

there are many ways to calculate normals

[steel snow]

yup i tried 2 of them

both were not ideal

[compact ingot]

Look at that library

[steel snow]

yet unity's worked exactly how i needed it but their code is closed off

[compact ingot]

It has them all

[steel snow]

i only see one in their code?

[compact ingot]

unity probably is a cross-product average of all incident vertex edges or an interpolation of face normals calculated from triangle edge cross product

[steel snow]

the two i tried was sum the cross products for each vertex then normalize on the next pass, and the other was sum the normal * half length of the normal magnitude and then normalize those

the issue was edge vertices produced wrong normals

since they only join to half as many faces

This is the only normals calc i can see in that geometry project

[regal olive]

You have to intensely guide gpt for it to give you anything useful

[compact ingot]

Does look like what unity does with its area weighted option?

[steel snow]

are you sure thats what unity does though ?

this is what i originally did:

for (int i = 0; i < _triangles.Length; i += 3)
{
    int a = _triangles[i];
    int b = _triangles[i + 1];
    int c = _triangles[i + 2];

    var side1 = _vertices[b] - _vertices[a];
    var side2 = _vertices[c] - _vertices[a];
    var normal = math.normalize(math.cross(side1, side2));

    float area = 0.5f * math.length(normal);

    _normals[a] += normal * area;
    _normals[b] += normal * area;
    _normals[c] += normal * area;
}

then i normalize all the normals on the next pass

but it produced different normals to unity's

[compact ingot]

can't be sure, but it seems to be a reasonably generic algorithm.

[worldly pecan]

Why do you multiply by area if you're going to normalize it again?

Also that is how unity should be calculating normals(idk about multiplying by area)

[steel snow]

i dunno thats just how the algo is ¯_(ツ)_/¯

[compact ingot]

a large area adjacent to a vertex has a larger influence on the normal than a smaller one

[steel snow]

but i have a theory for why mine is not correct

need to run some tests

[worldly pecan]

Oh so the normals are different per vertex... I see

[steel snow]

yes they have different adjoining faces

[compact ingot]

in this approach to normal calculation on a 3d mesh, yes, this should also get around the weird-boundary-edge-vertex-normal-problem

[royal moth]

hey, i'm also trying to figure out how to give a sprite thickness, don't listen to people saying "well use a mesh duhh" because in the game they don't use meshes, it's part of the game's shader, i have NO CLUE how they managed to achieve it but it's probably something to do with parallax shaders, if you figured out how to do this please tell me i'm also trying my best to replicate this effect

i MANAGED to replicate it (poorly) on blender, but i have no clue how to do it on unity

[steel snow]

@compact ingotdont suppose you know how they recalculated the mesh tangents too?

[compact ingot]

add a normal map that your shader uses to do some edge highlights/shadows

[steel snow]

ive got no idea on that algorithm

[compact ingot]

https://docs.unity3d.com/ScriptReference/Mesh-tangents.html its mentioned here how it is calculated.

[steel snow]

hmm that does not seem easy

[compact ingot]

check listing 7.4 here: https://foundationsofgameenginedev.com/FGED2-code.cpp

(this is c++)

[steel snow]

thanks!

[compact ingot]

btw, this has all the answers: https://foundationsofgameenginedev.com/

[steel snow]

neat, once this is done i finally got all the mesh generation applied to job system

[royal moth]

i don't think it's a normal map because i've seen the game's textures and they dont use normal maps for the sprites, if you see how they look in-game it looks like the sprites have some sort of thickness, i SUPPOSE they use a parallax shader but i have no idea how to do it in unity

[compact ingot]

you need a data source for those effects, and a normal map is the richest data-source you can have for them before using a mesh

it doesn't mean its rendered like a regular normal map. Its just input for the custom shader that adds the shadows & highlights

actually it looks like it could also just be a bump + occlusion map, and the normals for the highlights are calculated from the bump

anyway, light edges on sprites are typically done with normal maps, the shadows could be something entirely separate (a post effect), assuming these are skeleton animated?

[royal moth]

okay i see, i'm less than a beginner on shaders and complex stuff like this, so i kinda get what you mean but there are a lot of stuff that i still don't understand, and i wouldn't know how to replicate it on unity

[compact ingot]

have a look here: https://unity.com/how-to/2d-light-shadow-techniques-in-the-universal-render-pipeline

[torn basalt]

Maybe this is the wrong place to ask but does anyone know if it is possible to create a dotween sequence and run it in editor mode? I am woring on an aniumation tool and was curious about testing animations in Edit Mode

[compact ingot]

Here: https://dotween.demigiant.com/documentation.php#editorMethods

[native beacon]

Could likely use the alpha channel as both alpha and height for a flat object with rounded edges like that. You'd need to add some blur to the edge, if course, and use alpha clipping.

Another option would be to use a geometry shader to just draw more planes for each sprite, to provide edge depth

[frozen ravine]

how do I retrieve these values via scripting?

I tried this, but it's giving me way more values than I expected. I want to just return the 3 AnimationCurves

[wide mulch]

Can I serialize and select between delegate functions in editor?

[lethal oxide]

i'd also like to know if there is a good way to go from like a visual graph to a c# function based on a GUI like how they do it in visual effect graph

[lucid crane]

You can do events I think

Well, UnityEvents at least

[frozen ravine]

That's not what I was asking though

I was just asking how to get the animation curve from an animation clip

[wide mulch]

I want something more type-safe, I want to enforce selected functions have a specific signature

[lucid crane]

How about an enumerable variable, as well as an enum/action dictionary?

[wide mulch]

A dictionary lookup doesn't make sense, I could just use the enum values to index an array

[lucid crane]

Fair enough. I prefer dictionaries because I move items around a lot.

Or if you want items blank. Though ig it depends on context and preference.

[lethal oxide]

does anyone know what "=>" is doing here?: (Grid<GridObject> g, int x, int y) => new GridObject(g, x, y). I thought it was related to lambda function but i am not so sure anymore. It looks more like just declaring and initializing a new GridObject

[lucid crane]

Shortcut for return

Or just single line methofs

int AddOne(int x) => x + 1;

[lethal oxide]

so AddOne is a method that takes input x and returns x+1 right?

thanks! I think i understand

[sick flame]

Hey folks, could you mention a scenario, when using fp16 provides visual artifact/degradation compared to fp32?

[brisk pasture]

anytime you have very very large objects or objects very far away from the worlds origin point

floating point numbers are just approximations, the larger the number gets the more precision is lost in that approximation

[frozen ravine]

can someone help me with this? I can't find any documentation online about what a curve binding is and how to use it

all I want to do is get the 3 curves from the animation

but the function is returning 723 items in the array of curves for some reason

I assume it has to do with how curve bindings work, but I'm not sure what to do with the data since I can't find any documentation online about it

why does a curve binding give me an entirely new AnimatorCurve? How does it go from 3 curves in the animation to 723 bindings and how do I use those to get back to 3 curves?

do I have to combine them somehow?

[wide mulch]

16 bits can only store 65,535 unique numbers, while 32 bits can store 4.2 billion unique numbers

With floats, when you go twice as far away, you get half the resolution

So since 32 bit floats have much higher precision than 16 bit floats, you'll have better resolution throughout

[fickle mango]

I have a build error:
MyScript, No suitable method found to override: OnValidate()
That script inherits from LayoutGroup and is inside an Assembly. I added a reference to UnityEngine.UI, but that didn't fix the problem.

[untold moth]

It's probably not implemented in the base class🤷‍♂️

Remove the override keyword and it should work probably

[fickle mango]

nah, it is. The compiler doesn't give me an error, only when I try to build does it show up.

ah I found the problem:

[swift shadow]

What is making him do this?! Red arrow is where I press a button which changes a single integer, which should then be simply updated to GPU? I ain't even rendering here man

[sly grove]

Without seeing the hierarchy view how can we know

Oh it's rendering

Time for the frame debugger

[swift shadow]

here is the CPU view

[untold moth]

What do you see in the GPU usage hierarchy if you select the frame with the spike?

[brisk pasture]

would get the frame debugger out

[sick flame]

Hey thanks!
Why would be more precision lost? 🙂

Ok I got it 😃

[viral sage]

This missing reference is eating my head, Basically this only comes sometimes, if i add some debug or something in the code and run again its not coming, also in the build its not happening, only in the editor its happening, Is there any fix for this or has any one faced this before? please help.

Error:
MissingReferenceException: The object of type 'UI_References' has been destroyed but you are still trying to access it.
Your script should either check if it is null or you should not destroy the object.
UnityEngine.MonoBehaviour.StopCoroutine (UnityEngine.Coroutine routine) (at <2d8783c7af0442318483a199a473c55b>:0)
UI_References.SetLoadingScreen (System.Boolean status, System.String msg) (at Assets/UI_References.cs:111)
LoginManager.Init (System.Boolean isTestUser) (at Assets/LoginManager.cs:102)
GameConfigurationManager.ApplyRemoteSettings (Unity.Services.RemoteConfig.ConfigResponse configResponse) (at Assets/UI Manager/GameConfigurationManager.cs:101)
Unity.Services.RemoteConfig.ConfigManagerImpl.HandleConfigResponse (System.String configType, Unity.Services.RemoteConfig.ConfigResponse configResponse) (at ./Library/PackageCache/com.unity.remote-config-runtime@4.0.2/Runtime/ConfigManagerImpl.cs:553)
Unity.Services.RemoteConfig.ConfigManagerImpl+<>cDisplayClass46_0.<DoRequest>b0 (UnityEngine.AsyncOperation op) (at ./Library/PackageCache/com.unity.remote-config-runtime@4.0.2/Runtime/ConfigManagerImpl.cs:542)
UnityEngine.AsyncOperation.InvokeCompletionEvent () (at <2d8783c7af0442318483a199a473c55b>:0)

[tired fog]

Scriptable Objects professionas, how can I check if an asset has been saved/is being saved? I'm losing some data on a material after an event to save is triggered. This is because Im setting some compute buffers into a material in the scene, and after the event of saving the data gets cleared, so I need to reset it. I would like to subscribe to an event to know if an aset has been saved and reload the textures if thats the case

[stuck plinth]

https://docs.unity3d.com/ScriptReference/AssetModificationProcessor.html how about this?

[tired fog]

But this one is only for before

I need for after it has been saved, to reload the data, since after being saved, material data that doesn't have a property gets flushed (like compute buffers)

I found some issues online but sadly they are all unavailable to be seen in unity forums

[amber moth]

Hello guys,
is there any option for UIToolkit EnumField's Dropdown Position from code

[obsidian glade]

there's not a lot to go on here, but the error is quite clear in what the problem is - are you expecting UI_References to be destroyed at any point? changing scenes? have you explicitly defined code execution order?

[tranquil terrace]

hey guys. I'm writing my own custom collision detection system using SAT. I'm trying to find collision points and I'm struggling to do so here. it works fine in some cases but not in others, it seems. seems to mostly bung up in vertical edges . any idea what might be happening?

and before anyone asks why I'm not using unitys system, it's because I have to write my own for what I'm making.
I'm following this angels tutorial playlist
https://youtube.com/playlist?list=PLSlpr6o9vURwq3oxVZSimY8iC-cdd3kIs&si=io25V7nDvhrNNth7

[sly grove]

Well we haven't seen any code of course but perfectly aligned edges in general present a problem because you don't really have a contact "point" but really a manifold with infinite points along the touching edges

so I'm sure this is a special case in the code somehow

[tranquil terrace]

but how come it's working fine with horizontal along horizontal?

it seems to only be an issue with the vertical edges

[sly grove]

no idea, that will boil down to how your code works

[tranquil terrace]

I can send it when I come back if you want to have a look, I'll send the relevant things

[tranquil terrace]

https://gdl.space/abinogowil.cs

hey heres the code if u wanna have a look

[dusty wigeon]

If you are going to attempt to make your engine you should instead work on your debug skills than relaying on external help.

Why are you not able to find the issue ?

[tranquil terrace]

I'm sort of just following a tutorial, and until now everything has been working fine and I've been able to solve most issues myself, but when it comes to this the math is quite confusing

so i'm struigglign

[dusty wigeon]

What have done to attempt to fix your issue.

Because simply reading code is most of the time not enough.

Did you isolate your issue ?

Did you profile your code ?

If you are saying that your issue with vertical edge, have you printed what is the result of the expected collision.

Are you aware of what is a debugger ?

[tranquil terrace]

yeah, I'll have to try these when I get back from work

[lethal oxide]

let's say I have a class Player which uses interface IAttackable. I then do the following: IAttackable attackable = new Player(); . What would be the benefit of initializing this?

[sly grove]

wdym by "the benefit of initializing this"?

The benefit of the interface is that you could have a piece of code that works with IAttackable. That one piece of code can work with a Player or any other type that implements IAttackable. So you don't have to write a bunch of different code that does the same thing

e.g.

public void DealDamage(IAttackable target, float damage) {
  target.TakeDamage(damage);
}``` as a quick dirty example

[lethal oxide]

so am I specifying the interface of player there?

[sly grove]

wdym? No you're not "specifying the interface" of anything

IAttackable attackable = new Player();```
This code just declares a variable and assigns it to a new instance of the Player class

nothing else
It does illustrate how a variable of the interface type can accept a value of a class that implements the interface though

[upbeat path]

you are creating a player class which only exposes the interface

[sly grove]

creating an instance of the Player class

[lethal oxide]

gotcha. I understand

thank you both! Sorry, I am not up to speed on the lingo. Appreciate the patience

[north osprey]

Hey guys I'm making a SO that's capable of generating new cards for my game but I have one issue, I need to be able to know when the AssetDatabase has refreshed and unity has compiled my new generated code so then and only then I can generate the new SO from that newly created class. The code I'm sending here is still creating the new card with the BaseSO class for testing purposes, but I've already caught on that this issue is gonna happen. Any tips?

    private void GenerateScriptableObject()
    {
        //Find the correct class to inherit from
        
        //Create the ScriptableObject
        BaseCardSO card = ScriptableObject.CreateInstance<BaseCardSO>();
        AssetDatabase.CreateAsset(card, _cardSOPath + _cardName + ".asset");
        
        card.Name = _cardName;
        card.Description = _cardDescription;
        card.DeckType = _deckType;
        card.Image = _cardImage;
        
        //Clear the fields
        _cardName = "";
        _cardDescription = "";
        _deckType = default;
        _cardImage = null;
    }
    
    private void GenerateScript()
    {
        //Class name is going to be the same as the card name, prefixed with "XXX_" where XXX is the deck type
        var className = GetDeckTypePrefix() + _cardName;
        
        //We make sure there are no spaces in the class name and if we remove them and make the next character uppercase, we also remove special characters other than the underscore
        className = CreateValidClassName(className);
        
        //Create the Script
        string script = "Boilerplate Code here..."
        System.IO.File.WriteAllText(_cardScriptPath + className + ".cs", script);
        
        //After creating the script, we need to refresh the asset database so Unity can recognize the new script
        AssetDatabase.Refresh();
    }
}

[compact ingot]

you can trigger a reimport on the new file and just assume that on the next line it was compiled. it shouldn't really matter if it was compiled, i suppose you just want to be sure that it was imported and all the processing for the given asset type has been performed? UnityEditor.AssetDatabase.ImportAsset("path/to/script", ImportAssetOptions.ForceUpdate);

[north osprey]

Actually managed to achieve my goal by using the CompilationPipeline.assemblyCompilationFinished event and then putting my logic there, but thank you anyway! Now my issue is with getting the scriptable object fabricator to know this new type exists xD

[north osprey]

Yup in regards to getting the new type I create, can anyone spot something? I can guarantee that both the namespace and assembly def of BaseCardSO and the dynamicly created types are the same:

private string CreateScriptBoilerplate(string className, string cardName)
    {
        return "using System;\nusing UnityEngine;\n"+"namespace Assets.Project._Scripts.SOArchitechture.Cards{ \n" +
               "\n[CreateAssetMenu(fileName = \"" + cardName + "\", menuName = \"Cards/" + cardName + "\")]\n" +
               "[Serializable]\npublic class " + className + " : BaseCardSO\n{\n    " +
               "public override void ActivateCard()\n    {\n        base.ActivateCard();\n        Debug.Log(\"Activating Card: " + cardName + "\");\n    }\n    " +
               "public override void DeactivateCard()\n    {\n        base.DeactivateCard();\n        Debug.Log(\"Deactivating Card: " + cardName + "\");\n" + "  }\n " +
               "protected override void OnEnable()\n    {\n        Debug.Log(\"Created script of type: \" + GetType() + \" in assembly: \" + GetType().Assembly.GetName().Name);\n" + "   }\n}\n} ";
    }

private void GenerateScriptableObject()
    {
        //Fully qualified name of the class
        var namespaceName = typeof(BaseCardSO).Namespace;
        var assemblyName = typeof(BaseCardSO).AssemblyQualifiedName;
        var className = _newClassName;
        
        string fullyQualifiedName = namespaceName + "." + className + ", " + assemblyName;
        Debug.Log(fullyQualifiedName);
            
        Type newCardType = Type.GetType(fullyQualifiedName);

        if (newCardType == null)  //<--- Always null 
        {
            Debug.LogError("Could not find the class " + _newClassName + " to create the ScriptableObject"); 
            return;
        }
(...)

[serene jetty]

would it be faster to calculate a triangle normal in a compute shader or on the cpu when creating the mesh?

[clever urchin]

It Depends™️

How big is your mesh?

How are you making your mesh, how big is the CPU<->GPU overhead?

etc

[serene jetty]

im generating the tris on the gpu in a compute shader, then on the cpu reading it out and creating a mesh with it in a loop over the tris

just wondering cause if i do the normal calculation on the cpu im transferring less memory and the struct im using will fit evenly on the gpu cahce

[rancid oxide]

WebGL topic but its more on the technical side:
In my browser game I do API calls to an external service, for that I use an api key that is stored as an const string in the code.
I imagine that it is possible to somehow extract that key for malevolent use.
What can I do to avoid that?

[upbeat path]

setup your own backend server and use that as a pass through

[rancid oxide]

Then can't them spoof the communication between the server and unity?

(idk if that is the right term but I guess they could track every message that comes in and out)

[upbeat path]

that is possible, basically nothing is 100% secure you just need to make it very difficult or not worthwhile

[rancid oxide]

Got it, so returning to the pass through, what should I search for? I have no knowledge in this so I'll have to study it

[clever urchin]

Called API shimming

[clever urchin]

Probably best to stick with the CPU then if you've already transferred it off

I don't think it'll be too hard to test both with your setup, but mesh normal calculations are generally pretty fast anyways

[scenic forge]

You should make your own backend, and let the backend call the external service. Ideally the endpoints backend expose are not just a straight proxy, but rather tightly integrated to your game.

Whatever external service you are using, there's a very good chance that it was not designed to be used directly on the client side (but rather intended to be used in your own backend). This means that very likely:

• The service does not have fine grained access control. Getting access to the API key will allow someone to do anything the key can do, eg if the service is a storage layer and you are only using it to update player scores, yet someone getting access to the key can modify not just their own score, but someone else's score, or even delete someone else's account.
• The service does not have any logging for you to know what people are doing with that key.
• The service does not have any rate limiting and someone can spam certain endpoints to bankrupt you.
  With your own backend and you expose for example only a submit score endpoint, that's the only thing client can do, and you can properly protect that endpoint with authentication/logging/rate limiting.

[rancid oxide]

with backend u mean on the website I host the webgl game right?

This service let me get text to speech audio, an it comes via a unity web request directly in form of an AudioClip

I image that in your way I should send the request from the side backend and then somehow transmit the output to unity

[scenic forge]

No, like a web backend. The service you are using is also a backend.

Is that service designed to be used in client directly?

Their documentation should be very clear about that. If the service is designed to be used in client directly, their docs will say so very clearly; if the docs mentions anything about the API key being a secret and should not be exposed, then the service is not designed to be use directly in client.

[rancid oxide]

I think so

here's the service btw
https://elevenlabs.io/docs/api-reference/text-to-speech

[scenic forge]

Hmm not sure, I don't use it and from a quick scan of the docs nowhere it says whether your API key is sensitive or not, and the examples they show use keys client side, which seems to suggest your key isn't sensitive.
On the other hand, a quick Google search shows people who have their ElevenLabs API key compromised and used (Rabbit R1 being one of them which I found hilarious). When in doubt ask their support about your security concerns.

[floral marten]

is there a way to temporarily disable scene view mesh/object outlines? (ideally without force-changing users' currently set "show mesh outline" setting, this is for a plugin!)

[tranquil terrace]

In a strange discovery, the issue with collision point finding seems to be isolated only between the vertices with index 1 and 2...

[tranquil terrace]

https://imgur.com/cYwzial

[jaunty swallow]

Anyone know why having a NavMeshAgent on my object, rotates its transform by -90 on X axis?

I think it has something to do with me using NavMeshPlus, it probably tries to rotate the object to fit the navmesh which rotates to accomodate XY

[sage radish]

This looks like the relevant editor code:
https://github.com/Unity-Technologies/UnityCsReference/blob/77b37cd9f002e27b45be07d6e3667ee53985ec82/Editor/Mono/SceneView/SceneView.cs#L2286C43-L2286C63
Based on this, only AnnotationUtility.showSelectionOutline (internal) determines whether selected objects have outlines, and it's the same property that gets changed by the Gizmo GUI.
I think the only way you can avoid changing that property is using something like Harmony (https://github.com/pardeike/Harmony) to hook into SceneView and modify it.

[floral marten]

hmmmm good point, but also, there is the s_CachedParentRenderersForOutlining and s_CachedChildRenderersForOutlining arrays in there

but alas they seem to be populated on each call

[sage radish]

Only if s_SelectionCacheDirty is true, right?

[floral marten]

hm, true

..this seems cursed

a true, "I could, but, should I?" situation

maybe it's fine to just, reflection edit the user scene view setting just feels kinda cursed

[sage radish]

If clearing the array with Reflection works, I would consider that less cursed, at least user facing. The user would be able to see the setting has been changed if they check the GUI, and try to change it.

[floral marten]

assuming it's not got some cursed dirtying patterns yeah

ill try it

[sage radish]

I hope there's no assert checking if the input array is empty 😔

Otherwise you might need to give it a bogus renderer off screen.

[floral marten]

hmmm my reflection isn't finding the field

oh boy okay might be a version difference

yeah, I'm on 2022 LTS

field isn't there

looks like on 2022 LTS I want s_CachedParentRenderersFromSelection

oh no it works

it shows up for one editor frame, and then disappears

same thing during assembly reloads

but yeah it seems to work otherwise!

[floral marten]

@sage radish behold

public static class SceneViewHacks {
    // theoretically should work from Unity 2022.1 to 6000.0 (I only tested in 2022.3)
    #if UNITY_2023_3_OR_NEWER
    const string FIELD_PARENTS = "s_CachedParentRenderersForOutlining";
    const string FIELD_CHILDREN = "s_CachedChildRenderersForOutlining";
    #elif UNITY_2022_2_OR_NEWER
    const string FIELD_PARENTS = "s_CachedParentRenderersFromSelection";
    const string FIELD_CHILDREN = "s_CachedChildRenderersFromSelection";
    #elif UNITY_2022_1_OR_NEWER
    const string FIELD_PARENTS = "m_CachedParentRenderersFromSelection";
    const string FIELD_CHILDREN = "m_CachedChildRenderersFromSelection";
    #endif
    const BindingFlags bfs = BindingFlags.Static | BindingFlags.NonPublic;
    static readonly FieldInfo sceneOutlineParents = typeof(SceneView).GetField( FIELD_PARENTS, bfs );
    static readonly FieldInfo sceneOutlineChildren = typeof(SceneView).GetField( FIELD_CHILDREN, bfs );

    /// <summary>Hides selection outlines until a new selection is made,
    /// bypassing the user scene view setting</summary>
    public static void HideSelectionOutlinesTemporarily() {
        #if !UNITY_2022_1_OR_NEWER
            throw new NotImplementedException( $"Outline hiding not implemented for unity {Application.unityVersion}" );
        #endif
        sceneOutlineParents.SetValue( null, new int[] {} );
        sceneOutlineChildren.SetValue( null, new int[] {} );
        SceneView.RepaintAll();
    }
}```

[jaunty swallow]

Any ideas why it's firing the complete opposite way?

top transform is the red cube, bottom is the selected object

var raycast = Physics2D.Raycast(transform.position, Hero.Instance.transform.position, 9999, 10);

if(raycast.collider != null && raycast.collider.gameObject == Hero.Instance.gameObject)

the collider is always null

even if I walk into the raycast, it is still null despite the selected object having a collider and being on the right layer

[hushed fable]

You should assume any API you expose will be used by a thirdparty, and consider the potential abuse that could happen.
Wrapping an API is not the perfect solution, but it's way better than exposing your direct API key access to a thirdparty service. Having requests go through your server allows you to serve these requests within the context of your project, like you know a single client (IP, user) can't reasonably be doing thousands of requests per second.

[south ibex]

Is there any guides on implementing IK for articulation bodies? I'm in the process of trying to create physics-based procedural animation.

Attempting to use any physics-independent IK solver (I've tried unity animation rigging (with animator set to animate physics) and multiple github repos) doesn't seem to work

[long ivy]

Your raycast targets layers 1 and 3 and you're providing a position instead of a direction in the second arg

[worldly pecan]

Update to this, I found Unity's already tried to address this issue with Burst-Compiled Function Pointers
https://docs.unity3d.com/Packages/com.unity.burst@1.8/manual/csharp-function-pointers.html

[final steeple]

Do you have a tl;dr on what Burst function pointers solve here? To my knowledge they are just a workaround for not being able to store delegates in jobs, but you could always store actual C# function pointers (+ some state like a GCHandle if you had an instance to invoke on)

[worldly pecan]

You can store a function pointer as a raw pointer like an IntPtr on a unity job, but to call it you have to convert it back to a delegate which is a managed type.

[final steeple]

You can store a managed function pointer delegate

i.e. delegate*<int, void>

[worldly pecan]

No you can't-- delegate* is a managed type

[final steeple]

It's not

[worldly pecan]

I tried--

[final steeple]

If Burst considers it one, that's a bug in Burst

[worldly pecan]

I tried storing it as delegate* unmanaged<void*, void>

it doesn't allow it to be stored

[final steeple]

Yeah this is a Burst bug, thanks Unity

You can cast it at the usage site just fine though

[BurstCompile(CompileSynchronously = true)]
public unsafe struct FunctionPointerJob : IJob
{
    [NativeDisableUnsafePtrRestriction]
    public void* Callback;

    public readonly void Execute()
    {
        ((delegate*<int, void>)Callback)(10);
    }
}

//

static void LogNumber(int value) => Debug.Log(value);

new FunctionPointerJob
{
    Callback = (delegate*<int, void>)&LogNumber
}.Schedule().Complete();```

this works fine

[worldly pecan]

Oh wonders-- Unity's UnsafeUtility doesn't recognize it as unmanaged last I checked.

[final steeple]

Function pointer types in C# are unmanaged, they're just addresses and not reference types
I guess Burst doesn't account for that in its validation code

Yeah they probably only check IsPointer and not IsFunctionPointer

You could probably report it and pray for it to get fixed lol

[worldly pecan]

I wonder what FunctionPointers are for then, if it's just to fix this bug then it seems an awful lot of work to make all this

[final steeple]

Burst FunctionPointers are specifically for managed delegates, not raw function pointers

i.e. public delegate void MyDelegate(int a);

[worldly pecan]

ah ok...

[final steeple]

public unsafe struct EntityFunction
{
    [NativeDisableUnsafePtrRestriction]
    private void* _pointer;

    public EntityFunction(delegate* unmanaged[Cdecl]<Entity*, void> pointer) => _pointer = pointer;

    public readonly delegate* unmanaged[Cdecl]<Entity*, void> Invoke => (delegate* unmanaged[Cdecl]<Entity*, void>)_pointer;
}

you could work around it with this for the time being

Burst should let you store that

Not ideal though, since you'd need a special struct for each unique function signature

[worldly pecan]

I see, didn't know there was an attribute for disabling unsafeptrrestrictions

[final steeple]

There are quite a few niche attributes like that

Some of them have absolutely no documentation and I've never seen them used in any publicly available code

[worldly pecan]

Thanks for the tip--I got it to work! Just gotta fix my saftey now

[final steeple]

np!

[sacred night]

if I make a ScriptableObject instance through code (via the ScriptableObject.CreateInstance function) how does that behave at runtime? does it get saved to the disk with the rest of the game? I'm thinking about using SOs for save files

[upbeat path]

no, you cannot save any SO at runtime without 3rd party assets

[sacred night]

oof. thanks

[solar tangle]

Hi,
Does anyone know how to programatically recenter the world origin(I dont mean to create a new gameobject and then assign everything under it, that would be local recentering). My idea is to recenter to a specified point so it would be recognized as new world origin. This would also help in the case of photon which sends networked data based on world origin(i suppose)

[dusty wigeon]

The easiest would be to use HDRP feature "Camera Relative Rendering" https://docs.unity3d.com/Packages/com.unity.render-pipelines.high-definition@8.0/manual/Camera-Relative-Rendering.html

Obviously, it would not help with photon, but I'm pretty sure you do not need to recenter anything for photon.

[sly grove]

Well at a basic level you would move all the root objects in the scene

[solar tangle]

yes we can do that but photon or anyother networking for the matter wont move the transforms based on that right? it would be local repositioning in that case.

[solar tangle]

yea recentering is one of the options iam exploring for my usecase, so basically i have a passthrough app which would allow users to place a fire-place in their environment and other users spawn around it(not -colocation). So my thought process is, since each user should be able to place a fireplace anywhere in their environment then it should kind of act like a starting point(aka new origin) and photon should spawn and move players relative to that origin?
Any other way do you think would be better?

[dusty wigeon]

I guess you are doing VR. You can send and receive whatever you want. You do not need to take the NetworkTransform.

Simply push the position and interpreted it the way you want on the other device.

Alternatively, you can simply put everything inside a "root"

Also, if you are not in the same physical space, you do not even need that.

[solar tangle]

i'm doing quest passthrough so people will scan the room and place the fireplace in their environment. What i'm confused about is when trying to sync player movements, photon will always sync movement in terms of world transform.
Even if i put everything inside a root, not sure photon will follow the hierarchy, it would still sync transform in world space

atleast thats what i think

[dusty wigeon]

The documentation of Photon state that it sync the local

As the name suggests, this enables syncing of the transform.localScale value.
https://doc.photonengine.com/fusion/current/manual/sync-components/network-transform#

You should head to their forums/social platform if you struggle with that.

[solar tangle]

Got it and Thanks! let me try exploring more on this.

[south ibex]

How can I get a point that's a specific radius away from another point?

Trying to make an enemy whose main mode of movement is rolling. I want it to roll within a specific distance to the player. I tried just moving the object towards the player and stop at a specific distance (using velocity on a rigidbody), but it always ended up getting too close and becoming stuck because it was moving towards the player, not towards a point that was a specific distance from the player.

[devout hare]

origin + direction * distance is a point that's distance away from origin if direction is a normalized direction vector

[south ibex]

👍 thanks

[grand dagger]

yo so i have this door scirpt, but the problem im having is that the door is coming of its hinges, does anyone know how to make it like stay on the hinges?

https://gdl.space/aqaraxiqud.cs

[compact dawn]

Im not sure if this is the right channel to ask this question, but I am helping develop a CAD software for a specific robotics program and we are in the process of adding chains that connect gears together. We are trying to have a system that works by clicking a button and then selecting two gears and a chain automatically generates around the two. Ive recently found a asset that helps alot with the generation of chains on a path and that is the "Splines". The problem im having is i dont know how to automatically generate the spline in the right place on the gears and im not sure if its even possible. I have a picture below of an example that i manually created the points for it to rotate around. To restate the question, how can i have this type of spline generate in a loop around any two gears that i select with the press of a button.

[worldly pecan]

https://docs.unity3d.com/ScriptReference/Selection.html
You need to use this API
Also, you probably should make some EditorGUI to help facilitate this process and organize it

If you need to do it in-game that's harder cause you probably need to build/import your own editor

[dusty wigeon]

You will not be able to rotate the chain around the object from a transform given that the shape is not circular.
You have multiple alternative you could look into that I can think at the moment:

• Using skinned mesh
• Dividing in small section each individual piece of the chain and animating that
• Dividing in two part. Circular and linear.
• Using shader to displace the chain

(In theory, you only have to make a small movement loopable movement to be able to rotate)

[frozen imp]

@halcyon cloak No collab posts here

[halcyon cloak]

If there a channel for that

[normal lagoon]

I have been thinking and want to understand how to properly build the logic with Zenject. I may have MonoBehaviour components, but I think it's not just about using two interfaces and "cluttering" the container with dependencies that always need to be injected into MonoBehaviour through a custom constructor method.In general, I am thinking of starting to create objects from prefabs at the moment of initialization/dynamically. Here, I don't fully understand the correct approach. Yes, there should be a service that handles configuration delivery, and a factory (here, I don't want to use a Placeholder because I think it will require a lot of injections and writing code, which isn't a problem).Suppose I have a player who will have scripts like PlayerSpawnPoint, IPlayer, PlayerProxy (can it be used to hide the implementation? Here, too, I don't understand if it's necessary to use an interface for the factory and proxy implementation - I think not). Some kind of PlayerBehaviour and so on will also be needed, and everything should be covered by a facade.It's interesting to create the player at the level of GameObjectContext, that is, through PlayerInstaller or not.
My main problem is that I don't fully see the application architecture. Of course, many implementations are possible, but I want to finally try dynamically creating objects, attaching components, using an asset provider, and developing smoothly. I sincerely ask for advice, as I'm still far from being a Unity expert.

[compact ingot]

Zenject only makes sense for very large teams for long lived live services. It’s a terrible burden on your process if you are a small team and people on the team don’t fully grok automated dependency injection. Zenject does allow you to do all sorts of stupid hacks and does in no way guarantee a sustainable architecture. Mind also that DI containers are really bad at handling dynamic scopes. The pattern is at its best when all factories and scopes can be defined statically on startup.

[normal lagoon]

I don't work, sorry. I'm learning Unity, that's my current situation. Can you give me some advice? Preferably, write a bit more in detail. I'm not insisting, just asking as a Unity developer to another Unity developer.

[normal lagoon]

hmm, ok interesting

[compact ingot]

Don’t use DI containers in unity would be my advice

[normal lagoon]

how to learn if... :))

[half swan]

There is a very simple DI example in https://unity.huh.how/references
And that one is ok. But anything more is pretty garbage in Unity

[compact ingot]

if you must use one, use VContainer

[normal lagoon]

Well, they don't differ much as far as I know when it comes to proper dependency injection, just smaller memory allocation. Am I wrong?

[half swan]

They differ quite a bit honestly
In terms of usage and set up

[compact ingot]

proper Unity architecture cannot be learned by reading, you need to shoot yourself in the foot 1000 times to actually learn what works and what all the theory you read about means.

[compact ingot]

Zenject allows you to do stupid things, VContainer doesn’t (or at least it doesn’t help you doing it)

I bet you everyone who works in a zenject project that’s older than a year hates life

it’s one of those techs, that once an org adopts them, it gets taken hostage and starts an unhealthy dependence on it, unable to move past it, despite all the health hazards, stuck.

jira being another example

[sage radish]

I've been working in a small team (~4) on a project using Zenject that was released over 4 years ago and still gets occasional updates. I don't hate life 🙋

But we never went all in. Never use factories. Don't have any deep unit testing capabilities, just surface level stuff. We mostly use it instead of singletons and to enable some patterns that would otherwise be difficult.

[compact ingot]

What problem to you want to learn to solve that lead you to zenject?

[compact ingot]

So you probably mostly used it as a fancy service locator?

[sage radish]

Mostly. We use the subcontainer scopes a bunch (project > scene > game object), to easily get references between scripts under the same game object hierarchy. We also use Convention Based Bindings a lot: Container.Bind<IFoo>().To(x => x.AllNonAbstractClasses().DerivingFrom<IFoo>()), to implement some things by just defining a new class.

For example, we have a simple save data system that uses that. If you want to save some data somewhere, all you have to do is define a class:

public class LevelSaveData : SaveData
{
    public int LevelsUnlocked;
}

and inject it where ever you want to access it:

[Inject]
private LevelSaveData _levelSaveData;

private void OnLevelComplete()
{
    _levelSaveData.LevelsUnlocked++;
}

A save manager class injects all the SaveData and manages the serialization and deserialization.

Not sure if a service locator will allow these sorts of use cases. Or if VContainer has something similar.

[compact ingot]

as far as I can recall this should be possible in VContainer too. Seems reasonable usage. I’d still prefer the non-magic (manual) way of it though. My gripe with automatic DI is that it makes certain messaging patterns too easy which in the end produce the same global access & coupling problems as singletons, just with more interfaces to hide behind.

The average dev will just [inject] what they need an never stop for a second to think about the message flow & coupling or where stuff should be routed, whether a new system is needed to handle a case etc. So you end up with ravioli connected by invisible spaghetti and many bugs that were once discovered by static code analysis become runtime bugs.

[sage radish]

True. I would not trust a junior dev to add new bindings, but that's why we have pull requests.

[compact ingot]

i'm currently trying to find an architecture/patterns where juniors feel empowered to contribute at a systems level

[dusty wigeon]

This is really an issue I'm struggle with as well. There is a lot of fancy architecture but they usually are complex for unexperimented developer.

[wide mulch]

How can I interpolate an object that isn't a RigidBody or a CharacterController

[sly grove]

that's a really vague question. It depends on the specific circumstances

generally "interpolation" just means "move a little bit each frame until you reach your destination"

It's also not really an advanced question

[wide mulch]

So I'm writing a custom thing for movement because RigidBodies and CharacterControllers don't have what I want, and right now I'm manually changing transform.position of the object in FixedUpdate, but I have a camera set to follow the player, and that uses LateUpdate, and there's a lot of jitter because of it

If you want I can put the script I made in a pastebin

[sly grove]

right now I'm manually changing transform.position of the object in FixedUpdate
This is a problem

if the object is moving in FixedUpdate there will always be jitter unless you do your own interpolation, yeah

[wide mulch]

Alright

I was just curious if unity had a generalized solution already

[sly grove]

I mean generally if you're not using physics, you would just move your object in Update

[wide mulch]

I mean I am checking for collisions and stuff still, I'm just doing it manually

[wide mulch]

I got interpolation working :3

I'm storing the old and new positions, as well as the current Time.time, and then using the amount elapsed since that to lerp between the old and new positions

[fickle mango]

Is there a way to have async methods as listeners for UnityEvents? If I make my method async I can no longer select it in the inspector, and the way it is right now it doesn't work sometimes:

[compact ingot]

you can use UniTask to do that in a "clean" way by turning them into observables/enumerables.

[sly grove]

make a regular method that calls the async method

[compact ingot]

async event handlers are however a somewhat problematic idea producing all sorts of race conditons and overall debugging hell (thats already bad due to (unity)-event usage)

[scenic forge]

That should work just fine, hover over the green underline, it's not an error but a suggestion.

It's probably one telling you to do _ = UpdatePlayerName(input); for clarity.

[south ibex]

Trying to write a behaviour tree using muse. I want enemy to move towards player if enemy is within a distance of 20 units but do something else if the enemy is outside of that distance.

Currently, the "Move the Rigidbody" action (custom) will cause it to stay in the true branch even if it becomes false because the action is always running.

I tried adding an "Abort if" node but it caused the "Move the Rigidbody" to never run

This is how I had that setup

I also tried without the inverter just to make sure

[lucid crane]

Can I see your script?

Seems pretty simple:
if (playerDistance > 20)
{
moveToPlayer();
}
else
{
// Do something else.
}

[south ibex]

it's not a script, it's a behaviour tree.

[lucid crane]

Oh I thought that was a planning graph. Here may not be the best place to look.

[austere jewel]

To be specific, it's a question for #1202574086115557446

[surreal crane]

Is there a way to reset an AnimatorControllerPlayable through code just like how an Animator behaves when the GameObject it's attached to gets deactivated and reactivated again (restart from entry node)?

[round blade]

Good morning / afternoon/ evening depending on where you are from!

I am looking for some pointers in the right direction on how to set an external display (display 2 for simplicity) and be able to change it's resolution at runtime. I already have a list of resolutions that I can set but when I change the resolution from the dropdown only display 1 is responding and changing:

https://pastebin.com/meGHcB9G

Any settings I'm overlooking for the additional displays?
launch arguments etc

[real blaze]

hi. anyone knows why the _qrcodeImage 's image isn't being updated in the UIToolkit?
its populated by _qrcodeImage = this.Q<Image>("qrcode-image");
and in the debugger I can see the image property is fully set.

for the record, this works in the webgl build, but not in the editor

do I have to call a specific function to force-update it?

can I view the texture in the uitoolkit debugger? to ensure it's not being hidden by an overlay

[upbeat path]

you can try calling

MarkDirtyRepaint();

[real blaze]

didn't work

is there really no way to debug the image at runtime?

(also for the record, the arcode is a valid link and I tested it in other qrcode makers, it works. so the issue here is regarding UITK, not the qr-code)

[upbeat path]

Looking at my UITK Image code, I do not use the image variable. I create a Sprite from the texture and set the sprite variable

[real blaze]

ah, also regarding [your previous suggestion](#archived-code-advanced message), I realized since the script is in a package, my changes didn't work. so maybe MarkDirtyRepaint would work

i'll try that in a editable package

alright, MarkDirtyRepaint didn't work.

so, apparently its the qr-code generator. sorry for the misslead

so, right, the qr code generator was compressing the texture, and I don't know why but this was messing it up, probably because of my compression settings (ASTC)

or rather, this was the author's issue. according to here the compression must occur after applying textures

[sly grove]

On line 13 is it intentional that you started the loop at 1?

[lofty solstice]

Anyone here familiar with the CullingGroup API? I'm having trouble understanding how it works based on the Unity documentation for it.

[steel snow]

im a bit lost on the difference between these functions:
Graphics.RenderMesh which says RenderMesh submits the Mesh for rendering, which means it does not render the Mesh immediately. Unity renders the Mesh as part of normal rendering process. If you want to render a mesh immediately, use Graphics.DrawMeshNow.

why / when would you want to opt to draw immediately? Also how does the render mesh function differ to Graphics.DrawMesh

[sage radish]

If you want to render meshes to be drawn by cameras in the scene, you should use the non-immediate methods. These can be called from any point during the frame, such as Update, as it will just get queued and rendered later.

But if I want to render a mesh immediately, such as when performing your own rendering without using Unity cameras, you need to use immediate methods like Graphics.DrawMeshNow or those found in CommandBuffer.

[steel snow]

not sure why you would want to render without using a camera? also why is there a Draw and a Render mesh function being non immediate

how does unity differentiate between Draw and Render

[sage radish]

The Draw methods are older, the Render meshes are newer. They both do the same things, just with slightly different parameters.

[steel snow]

ah okay

weird they have both but neither were marked deprecated

unnecessary confusion by unity 😄

[sage radish]

Sometimes you don't need a whole camera. Maybe you just want to render one mesh onto a render texture. Instead of creating a temporary camera and messing around with layer masks, you can simply do this:

Graphics.SetRenderTarget(myRenderTexture);
GL.PushMatrix();
GL.LoadProjectionMatrix(myCameraMatrix);
Graphics.DrawMeshNow(...);
GL.PopMatrix();

[steel snow]

ah i see

[sage radish]

But this is a very direct method of rendering, which bypasses Unity's lighting and other rendering systems.

[steel snow]

so like projecting to a billboard etc

so you dont have to use a second camera

[dapper cave]

I'd like to bounce ideas regarding speed vs code readability.
I have this AI checking for friends and foes in vicinity and was briefly tempted to do bitwise test for team check, but it seems that physics is so expensive anyway that i might keep the list.Contains.
teams are held in scriptable objects and team-wide affinity is a reference to other team SO, when team A attacks team C, team C SO gets updated and all of team C counter attack.
what do you guys think, by eperience.

[tall ferry]

I'd look into how u could speed up those physics checks rather than worry about a list as this is probably only comparing a few elements. If you think the bitwise solution would be readable to your future self (assuming solo project) then I'd see no issue with it though dont really think its needed

[sinful meadow]

what methods can i use to make setpixels() and apply() faster on very large textures?

[tropic vigil]

Focus on optimizing the code in places you think it will matter. Make optimizations in other places after seeing it actually has a bad impact on the performance. Optimizing the right segment of the code will usually give you much more than optimizing random code fragments.

[round blade]

I change this around to start from 0 or from 1, works on display 0 which is the main screen. Just doesn't work for the secondary screen

[marble bramble]

I'm instantiating a GameObject from a prefab and GetComponent<>() returns null when trying to access any other script attached to the game object prefab. What gives?

[tender flame]

ive been having an issue for about 3 days asking in the #💻┃code-beginner channel and i am continously getting an issue saying that c# dev kit is targeting the wrong program format x86 instead of x64

thats the error code

[flint sage]

Install the 64x netsdk? https://dotnet.microsoft.com/en-us/download

[stoic otter]

is there a limit as in how much things you can save to the appdata folder?
we have a game what only can crash when the appdata save folder is clean?

[flint sage]

As long as there's disk space

[upbeat path]

There is a limit on individual file size and that depends on your disk format and OS but it is unlikely you are hitting that

[upbeat path]

it is possible that something like OneDrive or Dropbox is causing your problem

[stoic otter]

aa thx steve!

[stoic otter]

so i found a bug in my code where i subscribe 2 times on a event += in the on enable and on disable
the strange thing is the bug happens most of the time for the players who playing it the first time what happen when you 2 times subscribe += to a event

[upbeat path]

it executes twice when you invoke it

[stoic otter]

Yeah but cant it give you a memory leak

right

[upbeat path]

that is not a memory leak, they are completely different beasts

[flint sage]

If you never unsubscribe, those instances never get GC'd though

So technically if you do it a lot you might run out of memory

[stoic otter]

we got another one we dont get this one either

it crashed our entire game

[flint sage]

No idea what that sentence means

[stoic otter]

so i have this error how do you clear the profiller memory ?

[lament salmon]

Probably with the Clear button

[timid grail]

Any articulation body wizards that could take a peek? (thread in code-general)
https://discord.com/channels/489222168727519232/1265591984698163271

TLDR: Instantiating prefabs with arti bodies in them do funky stuff 🤔

[tender flame]

yes ive downloaded that i think ihave the right sdk installed the dev kit js isnt using the right one but i could be wrong

[pearl jackal]

How do I get a shader to render on top of everything?

[fallow echo]

render it last, turn off depth testing for it, I guess 🤷‍♂️

[lucid crane]

Is it a bad practice to store all logic about players in one class? I currently have two classes for each player: One being for the player state machine and one for the player physics. These classes are highly dependent on each other. What I was thinking is because they exchange data so much is to merge them into one class, but I fear that would defy conventions. If they were both in one class it would defy the “single use” convention of SOLID.

[fallow echo]

if both of them know about each other - that's not good, but if only one knows about the other - that's fine (usually)

[lucid crane]

Kinda. PlayerHandler feeds all the parameters that PlayerPhysics needs to know. PlayerPhysics is required an intake of parameters but not from PlayerHandler specifically

[fallow echo]

this seems fine

is it annoying to work with for some reason?

if not, I'd probably leave it as is... if yes - then I'd try to resolve the annoyance in some way other than merging these

[lucid crane]

Kinda. Mainly because PlayerHandler has a parameter called CharacterData, which represents the character the player is playing as. Different characters have different collision boxes and falls speeds which are needed by the PlayerPhysics class. What I currently do is feed those variables to them and assign them as fields within the PlayerPhysics when PlayerHandler starts running.

[fallow echo]

mm... to be fair though - the S in SOLID is also overrated... I use an extended version of it called "less than five responsibilities principle" 😄

[lucid crane]

What would you suggest?

[fallow echo]

does that need to be part of the PlayerHandler in the first place?

[lucid crane]

Does PlayerPhysics?

[fallow echo]

I mean, from what I understand those are params needed by the PPhysics class... can't you just put them there and set them there?

although, injecting params is generally not unreasonable 🤷‍♂️

[lucid crane]

CharacterData holds things needed by the PlayerHandler as well. Such the moves a character can preform.

[fallow echo]

yeah... well then just inject the physics part into the physics one as you do (it could be one struct with the data, so that you won't need to do multiple assignments and can add more data later on, etc.) 🤔

without seeing the code, it seems fine.. (am a bit lazy right now to read code and think hard though, but someone else might feel like it)

[lucid crane]

Perhaps. Though if I merge them I could do something like:
Rect CollisionBox => myCharacterData.CollisionBoxData;

That way I don’t have to keep them as parameters. Though if you said merging is fine, I might just do that.

[fallow echo]

the separation sounds reasonable, tbh, so I'd prefer to keep it... but if you don't mind the potential issue that you may need to separate them again in the future, because something else needs to control the physics thing - then it's not the end of the world
usually it goes the other way around... you start with a merged thing (because it's simpler) until you have a reason to split it... technically split gives you more flexibility

[steel snow]

so im generating a mesh asset, i save it as a subasset to my scriptable object and i assigned it to the field in my scriptable object yet in the inspector i get type mismatch (see image)

the asset file is valid though... when i double click the type mismatch box it loads the mesh inspector and sure enough its correctly showing a mesh so why is it a type mismatch?

this is how im loading it to the field in OnValidate

var path = AssetDatabase.GetAssetPath(this);
var subAssets = AssetDatabase.LoadAllAssetRepresentationsAtPath(path);
foreach (var subAsset in subAssets)
{
    if (subAsset is Mesh mesh && mesh.name == "Plane Mesh")
    {
        _planeMesh = mesh;
        return;
    }
}

[tender flame]

I have found very little information on this specific issue related to vs code and unity. Whenever I launch vs code with the default console template opened it loads for a second and gives me this error. I know I have the latest version of the .NET SDK installed as x64 and I have no idea how to change what the c# dev kit is targeting. I have tried uninstalling the SDK, vs code, and all of the extensions within vs code. I have also tried editing my system environment variables and changing the path settings. I have truly no idea what to do and have been stuck on this issue for days. any help is much appreciated!

[lucid crane]

You need both the vs extension and the vs code extension

Get the unity vs code extension, then in your unity settings set you IDE to vs code

[lethal oxide]

My units are forgetting which healthSystem it has created an instance of during setup. Why is this occuring?

[sage radish]

It is not possible for a class reference to suddenly become null. So that leaves two possibilities: Something is setting the field to null, or the field isn't assigned before Damage is called.

The first one is easy to check since it's a private field, so only this class can assign it. If you're not assigning it to null anywhere, that only leaves the second possibility.

[lethal oxide]

I am not sure what you mean by the field not being set. The totalDamage variable is not null on Debug. But healthSystem is null on Debug

[sage radish]

totalDamage is the parameter, healthSystem is the field. Field not being set means it's null.

[lethal oxide]

healthSystem is a class. A field is a variable stored on class

[sage radish]

HealthSystem is a class. healthSystem is a field of the type HealthSystem.

This is not a #archived-code-advanced topic.

[lethal oxide]

ah i see what you mean

i dont think its either of those issues then. Like you said, nothing should be able to set the field to null since its private. And the field is definitely assigned before Damage is called since the target units are in the scene from the very start and initializes healthSystem on Start()

[sage radish]

Start is called just before the first Update is called. It's possible for the unit to be spawned and immediately get damaged before its first Update, and therefore its Start, is called.

Or if the script is disabled, Start won't be called, but it can still be collided with and Damage can still be called on it.

[tender flame]

i already have that set its not the issue

[lucid crane]

Ah darn. My personal issue was solved because I only had the visual studio code package and not the visual studio package, so I assumed that was this issue.

[opal wigeon]

If I perform some kind of function a few thousand times is it better if variables are declared outside of it?

float b; 
void A()
{
  b *= b;
}

I'm thinking that would free up the garbage collector as the function is reused

[upbeat path]

none of that will affect the GC, it is all on the stack

but it would be damn silly if b were a local variable

[opal wigeon]

I just need a temporary vector3 for something inside the function and was wondering

[upbeat path]

as long as you are not new{} ing it then, yes declaring outside the method is better

[dusty wigeon]

Are you sure of that ? Pretty sure local variable are basically free and it might even be better given that they would be closer.

In any case, it would probably not even be perceptible

I'm not really an expert, but here is the difference between:

public float M(float a) {
    return a*a;
}

and

float a;
public void M() {
    a*=a;
}

https://sharplab.io/#v2:C4LglgNgNAJiDUAfAAgJgIwFgBQyDMABGgQMIEDeOB1BAZhAPYCGwBTA3FTfkQCwEBZABQBKClxo0mAKgC8HCQQC+OJUA===
https://sharplab.io/#v2:C4LglgNgNAJiDUAfAAgJgIwFgBQyDMABGgQMIEDeOB1RhAZhAPYCGwBAsgBQMtvMCUFKjRHIA7AWYAqZgG5h1AL45FQA

[sly grove]

It's not a performance question so much as a "this does a completely different thing" question

[sly grove]

if the variable isn't needed outside the funciton it should be declared inside the function. Always give the most local scope possible

[opal wigeon]

I know but this function needs to have really good performance

[sly grove]

Local variables are almost certainly better for performance too

due to memory locality right on the stack

but - why don't you benchmark it?

[opal wigeon]

I tried to but fps doesn't change and I can't check the memory usage because it's constantly rising for some reason

[sly grove]

talking about benchmarking the function not the game as a whole

[opal wigeon]

I don't know how to do that

[sly grove]

Use benchmark.net or just simply run each function 100,000 times and record how long it took

[honest hull]

So in the unity play mode(not build), there isnt really a way to create a new screen dimension setting and set to it all from a C# script, right?

[tropic vigil]

Do you mean game window? There was a thread about it with solutions, but I'm not sure if it still works.
https://discussions.unity.com/t/add-game-view-resolution-programatically-old-solution-doesnt-work/783954/3

[honest hull]

oooo this looks perfect!! thanks!!

[tender flame]

I ended up getting it fixed thank you for trying to help though!

[earnest geyser]

does anybody know some resources that talk about procedurally animated spiders? I have a working procedurally animated spider but it gets messed up when i try to make it walk on surfaces with steep slopes.

I think I need to cast some horizontal rays to check for obstacles in front since I'm only casting downwards rays from above every foot but I'm not sure what to do

Thi is the current version

[sly grove]

tbh i watched this whole video and don't see any issues

Other than the fact that the body stays at the same height all the time. It would be a bit more realistic if the body adjusted according to the leg heights

[earnest geyser]

ok wait it's not steep enough

and when i rotate the body

[sly grove]

right so - the body isn't adjusting at all

that's a problem

it will need to follow

Also "raycasting straight down" isn't going to cut it for a spider which can walk straight up vertical walls and even upside-down

[earnest geyser]

when i rotate the body and move it, legs at the back just go limp

instead of raycasting straight down

should i raycast down with respect to the body rotation

[sly grove]

definitely, something like that

[earnest geyser]

i think my rays are green

but it should be red

right

[regal olive]

Hello

What RenderParams should I use for Graphics.RenderPrimitives(rp, MeshTopology.Lines, (...))?

https://docs.unity3d.com/ScriptReference/Graphics.RenderPrimitives.html

the example is for MeshTopology.Triangles

[untold moth]

Probably the same params..?🤔
I don't thing there's a difference in params depending on the topology.

[regal olive]

Well I have no triangles, just edges to draw

I'm searching for a way to draw a debug wireframe in ECS

given list of edges

[slate sapphire]

hey so i need help making a game, im struggling to put my ideas into it and need help, i know it kinda seems a little jumpy but i could really use the help.

[regal olive]

I may get the triangles but thought it's possible just with edge list

[untold moth]

I don't think that's enough for the GPU. Pretty sure it needs the triangles regardless of the topology.

[regal olive]

okay, thanks

[untold moth]

Is that really and advanced code issue?

[slate sapphire]

good point, didnt see the other chats at first

[lethal oxide]

I have methods with default input values. i am using a delegate to use these methods but right now its giving me an error because there is nothing within the delegate parameter. How do I specify that the delegate should use default value?

[sly grove]

Your delegate needs to match the expected delegate type of course

Not sure what you mean by default value

Show code?

[lethal oxide]

my gun method has a default input of int damage = 15. I can create an action delegate that takes int called attack but I am not sure how to specify to take the default input: private Action<int> attack;

[sly grove]

You make another function/delegate that calls your method

Show your code if you want specifics

[sage radish]

You can define default values in the delegate. But it sounds like you want to be able to use the default values of the method, without knowing which method you're calling because you're invoking it through a delegate. This is not possible.

You can do this:

public delegate void AttackAction(int damage = 15);

private void Invoke(AttackAction attackAction)
{
    attackAction(); // passes the default 15 damage.
}

But the default will always be 15. It doesn't matter what the defaults are for the actual method assigned to the delegate, or if there are any defaults at all.

[lethal oxide]

that is actually the exact case I am facing. I am just running the methods directly now instead of the delegates. should be okay

thank you both!

[livid zephyr]

Hey everyone! I'm trying to set up a flag football game where flag pulling is the base of the game. The offensive player has to avoid the defensive player and score while the defensive player has to pull the flag to win

The defensive player is controller by Ai and has a navmeshagent and also the football field has Navmesh

Currently the way i have it set up, the ai is supposed to go towards the offensive player to "pull its flag" but the ai stops each time i move my player and then moves once my player is stationary

Here's my code for the ai

Im using Unity and VS 2022

[upbeat path]

<@&502884371011731486> Spam

[austere jewel]

!warn 972587276972462171 don't spam random crap here. There is no off-topic.

[thorn flint]

jamafr has been warned.

[sly grove]

What debugging steps have you taken?

[stray plinth]

Did anyone make a weapons system with scriptableObjects, that hold both the game assets and the logic to spawn, fire them?

It’s probably a well used/represented solution, that is why I am wondering what caveats/issues/etc. were encountered?

[livid zephyr]

I've looked at my navmesh and tried to make adjustments to the settings. I've also looked at my code I pasted and tried different variations

[tropic vigil]

Having logic in ScriptableObjects works pretty nicely for me. The inconvenient thing is that you must pass the source(s) and/or target(s) as arguments, but it's a good price to pay.

[dusty wigeon]

Usually, I have Definition (Weapon, Projectile, Ability, Character, etc.) for many things which have their Instantiate function. I have no particular issue.

[tranquil terrace]

https://imgur.com/taY1n0r

if (IntersectCirclePolygon(closestPoint, radius, polygonVertices, out circlePolygonNormal,
                    out circlePolygonDepth))
            {
                if (circlePolygonDepth < depth)
                {
                    normal = circlePolygonNormal;
                    depth = circlePolygonDepth;
                    return true;
                }
            }
            if (IntersectCirclePolygon(capsuleA, radius, polygonVertices, out circlePolygonNormal,
                    out circlePolygonDepth))
            {
                if (circlePolygonDepth < depth)
                {
                    normal = circlePolygonNormal;
                    depth = circlePolygonDepth;
                    return true;
                }
            }
            if (IntersectCirclePolygon(capsuleB, radius, polygonVertices, out circlePolygonNormal,
                    out circlePolygonDepth))
            {
                if (circlePolygonDepth < depth)
                {
                    normal = circlePolygonNormal;
                    depth = circlePolygonDepth;
                    return true;
                }
            }

I've got an issue with my capsule. my current order is checking the closest point, then the top, then the bottom.
but sometimes the capsule will phase through walls. is my order wrong?

[compact ingot]

scriptable objects are globally accessible instances of classes, and all downsides of that exist with them. If you are disciplined and don't abuse or misuse their nature, they are fine. If you put pure functions on them that don't mutate some internal state or produce side-effects, maybe configured by immutable serialized properties on them, thats all fine.

[livid zephyr]

Do you have any tips I can use?

[sly grove]

standard debugging techniques - logs, attaching the debugger. Make sure the code is running when you expect it to run and with the data you expect.

[steel snow]

does any one know how to update unity navmesh in runtime in specific areas i really don't understand the api at all and theres few example codes provided

i just want to update the navmesh after adding more meshes to the scene but its a bit complicated

[tropic vigil]

You can block access in certain areas by using NavMeshObstacle:
https://docs.unity3d.com/ScriptReference/AI.NavMeshObstacle.html

[steel snow]

i cant use that

i need to rebake the navmesh in an area when i add more geometry to the scene - navmesh obstacle only allows two basic shapes which is totally useless

im trying to use this https://docs.unity3d.com/ScriptReference/AI.NavMeshBuilder.UpdateNavMeshDataAsync.html

its not really obvious to me if when updating via the bounds it will wipe out the areas outside of the bounds or not so it becomes a bit unclear if im using the right thing

[native beacon]

Sounds easy to test though

[steel snow]

i did i tried this:

    private void OnEnable()
    {
        _collider = GetComponent<MeshCollider>();
        _worldBounds = _collider.bounds;

        var center = _surface.transform.TransformPoint(_worldBounds.center);
        _localBounds = new(_worldBounds.center, _worldBounds.size);

        var sources = new List<NavMeshBuildSource>();
        NavMeshBuilder.CollectSources(_worldBounds, NavMesh.AllAreas, _geometrySources, 0, new(0), sources);

        var handle = NavMeshBuilder.UpdateNavMeshDataAsync(_surface.navMeshData, _surface.GetBuildSettings(), sources, _worldBounds);
        handle.completed += OnComplete;
    }

    private void OnComplete(AsyncOperation operation)
    {
        if (operation.isDone)
            Debug.Log("NavMesh update completed successfully.");
        else
            Debug.LogError("NavMesh update failed.");
    }

but this does not do what you would think it does

this is the visual of the navmesh + the bounds area i want to update

but when i run it i get this:

makes no sense to me

[long ivy]

MeshCollider bounds are already worldspace so your TransformPoint line makes no sense

you want the inverse if you want center to be in _surface's local space

[steel snow]

i finally got it working

i can now use abstrast polygon shapes as navmesh obstacles

took forever to figure it out

[sand shore]

checking the capsule's endpoints firs and then the closest point might give you better results because endpoints are often more critical for collision detection. also instead of returning early when collision is detected have you tried to loop over all the points you need to check for collisions and then gather the results and keep track of the smallest penetration depth so you can then return whether any collsion was detected along with the best normal and depth values.

[solid grove]

I must be misunderstanding how RenderTextures work. I'm trying to render two different textures, one that goes to the screen, the other to be used for a graphical effect. I created a secondary camera + render texture and set the target buffers of said camera to the rendertexture, I then set a texture in my compute shader to the render texture. After the the scene renders, when I then read the data of each pixel out to a dummy array just to verify what data is in it, all the values are blank (they should not be). When I open the frame debugger, I can confirm that rendering of the scene is occurring, but it doesn't seem to be writing to the RenderTexture. I feel like I must be missing some fundamental step, anybody know?

In the frame debugger, the steps that draw the secondary texture state the target buffer as a temp buffer, but I don't know if that's unusual.

[lament sphinx]

Whats wrong. I have a camera and will calculate the right upper point dinstance from pivot. With the height all is okay but width is wrong

[lucid crane]

You are multiplying it by 2 just to divide it by 2?

2x/2 = x

(2 * x * y) / 2 = xy

[sly grove]

Unl;ear what you're trying to calculate here

is this a perspective camera or orthographic?

[lament sphinx]

its a perspective

[sly grove]

So are you calculating an angle here?

or what

[lament sphinx]

i try to calculate the half distance from the cameraposition vector to the roght or left end of the view

[sly grove]

but this code seems to only be dealing in angles

so what do you mean by distance

also for a perspective camera distance between parts of the camera view depends also on the depth that you're looking at

[lament sphinx]

the calculation what i postet works in 2d. i dev it for thois . but now i convert to 3d and there this doesnt work with this formular

[sly grove]

Yeah because it makes so sense, you're dealing with angles here

you can't treat an angle as a position

So you have to specify the problem better

[sly grove]

based on your screenshot you want the position on that game plane

for this you should use Plane.Raycast

[sly grove]

e.g.

Plane p = new Plane(Vector3.up, Vector3.zero); // assuming your plane is flat and at y = 0

Ray r = Camera.main.ViewportPointToRay(new (0, 0));
p.Raycast(r, out float distance);
Vector3 topLeft = r.GetPoint(distance);

r = r = Camera.main.ViewportPointToRay(new (1, 1));
p.Raycast(r, out float distance);
Vector3 bottomRight = r.GetPoint(distance);

[lament sphinx]

camera viewport and from the pivot to the corner the distance

[sly grove]

this doesn't explain the 3d space part of it though

that point can be 0 units from the camera or a million units from the camera

and it represents a different position in space depending on where

My example code projects that point to a given Plane in the game world and uses that.

but it depends what you actually want.

[lament sphinx]

okay i try it

i think i must think other in 3d

[sly grove]

yes you have to change your thinking

[wary blaze]

From what I've read online, if Unity is using the Async Upload Pipeline, I should see specific Profile markers in the profiler (AsyncUploadManager.ScheduleAsyncRead, AsyncReadManager.ReadFile, and Async.DirectTextureLoadBegin). I'm assuming these markers will appear on the Main Thread in the CPU Profiler? Can anyone confirm?

[frozen ravine]

is there a way to rename these sub-assets?

whenever I try editting its name with .name and refreshing the asset list, I get errors in the console

[upbeat path]

what are we even looking at here? Those look like ScriptableObjeccts so F2 should be able to rename.
Also just saying ' I get errors in the console' is not very helpful. Show the errors
Also, not a code question, please use the correct channels

[frozen ravine]

well it is a code question because I'm trying to rename them with code

they are scriptable objects

but they're sub-assets

the type you create with this code

the error is that it basically loses a reference to the asset once I try renaming it, which doesn't happen with normal scriptableobjects, only the sub ones

I also tried using AssetDatabase.RenameAsset which isn't doing anything at all on the sub-asset

[exotic trout]

@frozen ravine complete shot in the dark but any chance the so's .name isn't matching it's file name? messed w/ making so's via code but not subasset stuff

[frozen ravine]

well it works with my other scriptable objects, it just doesn't work with that one

Unity's just weird about sub-assets, you can't even rename them in the inspector manually (which is why I'm trying to figure out how to do it with code)

[upbeat path]

Can you rename them in the project view?

[frozen ravine]

no, this is the project view

I can rename everything else just fine

as far as I know, in the project view, not being able to rename them is just an oversight, but I figured there would be a way to rename them via scripting

[final steeple]

If you click on the asset's name, wait half a second, and click again, does it let you rename it?

Unity is weird about exposing renaming

[upbeat path]

I suspect this may be a design flaw in Unity, probably related to meta data

[frozen ravine]

no, I also can't F2 it

[frozen ravine]

zamn

[final steeple]

Certified Unity moment

[frozen ravine]

that really sucks lol, because the sub-assets are really good for organization

[upbeat path]

if you cannot f2 rename then that is actively blocked by code so the chances are you just cannot rename

[frozen ravine]

true

[upbeat path]

very odd behaviour, normally Unity does not care about names at all

[jade veldt]

There has to be a better way of doing this in ECS right? If so let me know, the goal is to make the workers do two seperate tasks repeately after each other. This "works" but seems like a slow way of doing so

also, is there a way to trigger code via an event / pruning rather than just boolean locking with ECS?

[real blaze]

if you're looking for an event, you won't find it (unless you give up on burst-compatibility for every code using that system/component). but you can make it easier for yourself by creating generalized data types that at least make it consistent across your application. example:

public struct Event<T> where T : unmanaged
{
    private T? _args;

    /// <summary>
    /// Gets the event arguments if the event is valid. Also consumes the event.
    /// </summary>
    /// <param name="args">Arguments of the event</param>
    /// <returns><c>true</c> if the event is valid.</returns>
    public bool TryGetArgs(out T? args)
    {
        args = _args;
        _args = null;
        return _args.HasValue;
    }

    public void Invoke(T args) => _args = args;
}

invoking the event:
spawner.spawnWorker.Invoke(...)
using the event (it also consumes it, but you get the idea. change it to accept IDs if you want proper subscription system in it) :

if (spawner.spawnWorker.TryGetArgs(out var args))
{
    // use args...
}

[half swan]

Might wanna ask in #1062393052863414313 next time

[hollow fog]

I watched Codeer’s video on how to make a procedural animation. I need a bit of help with the second step of the video, which involves anchoring the bottom of the leg to the ground.

[azure meadow]

I have a mysterious object that came from a wacky json object... most of the times I can convert it with:

(T)Convert.ChangeType(mysteriousObject, typeof(T), CultureInfo.InvariantCulture)

however, if the object is something that looks like a list, that fails.
my workarround is to reserialize it and parse it back in the correct type 😛

  string json = JsonConvert.SerializeObject(mysteriousObject);
  T value = JsonConvert.DeserializeObject<T>(json);

But this feels really hacky 😅
What is the correct C# syntax for this kind of crazy cast?
(note, just casting (T)mysteriousObject fails)

[sly grove]

that came from a wacky json object

Can you elaborate on this? What does the json look like and how are you parsing it?

[azure meadow]

the json is being parsed as Dictionary<string, Dictionary<string, object>> it's a two key deep json object that can have pretty much anything on the last level

the problem in question is an array of bool

[sly grove]

well.. that's why it's coming in as an object of course

[azure meadow]

does C# has something like void*?

🤔

[sly grove]

object is the closest thing

I'm just curious why you feel you need it

[sly grove]

but when I asked the question I wanted to know things like which Json library you're using and what the actual code was being used to parse it

[azure meadow]

oh, JSON.net (newtonsoft)

[sly grove]

Sounds like you're using Json.NET and JsonConvert. For a structure like this you could consider using the JObject API instead of JsonConvert

That will give you a lot more control here

but a lot of this depends heavily on where the JSON is coming from in the first place and why it has that format

there's probably a better/cleaner way with less ambiguity

[hushed fable]

There are settings to control the amount of type information, if that's the problem you are running into

[azure meadow]

I will investigate the JObject 🤔
The json should always come from another c# serializer... but I am not entirely sure 😅

[sly grove]

right assuming JsonConvert is being used to serialize in the first place. But JSON.Net's type information strategies don't all conform to the actual JSON standard 😉

[azure meadow]

thanks a lot!!! JObject (and the entire family of J* thingies) worked perfectly and are WAY easier than typeguessing in a C# dictionary 🥹

[sly grove]

Awesome glad to hear!

If this wasn't Unity Json.NET also supports dynamic which would probably be good for your use case but Unity IL2CPP doesn't support it!

[rancid bane]

not sure if anybody can help but

please i cant build my apk without fixing

[thin mesa]

you're not even looking at the correct line

[rancid bane]

its the line it sends me to

its highlighted

Lemme check

[crystal relic]

The error says line 71

[rancid bane]

do i delete this

[crystal relic]

Send a snippet of a few lines before and a few lines after

[rancid bane]

before

after

[crystal relic]

Yeah delete what's on line 71.

[rancid bane]

ok let me see

not again...

now can u help fix this 😢