archived-dots

[light mason]

what if child is offset to parent

it should just return the wp

no matter what

[zenith wyvern]

As far as I know it does

[light mason]

something is wrong

so im have a simple grid

im just query a child in the grid

never find

but when i un parent it does

[zenith wyvern]

What's the query?

[light mason]

the exact same entity un-parented is a hit

[zenith wyvern]

So that job doesn't run when it's a child?

[light mason]

it runs but the child localToWorld and Translate is not the world pos

[zenith wyvern]

Translate is relative to the parent

Aka local position

But LocalToWorld.Position should always be world position

[light mason]

that what assumed

let me place the parent so that one of the children is at 00 and see what i get

odd

its a hit

[zenith wyvern]

Are you sure you should be rounding there

If you're dealing in a grid it should probably be floor

[light mason]

do i need to do something special when i instantiate the parent

ok ill do that

but i have ui pringint as mouse move and that seems correct

mmm

i just notice that child from other parent get a hit that are not at 00

[zenith wyvern]

And comparing equality on floats is never a good idea. You should convert them to int3s before comparing. No idea if that's the issue but

[light mason]

ok

let me place an entity in scene and not instance it

[native trail]

Thanks ! This was the most helpful advice i've ever heard about ecs :D

[light mason]

ok looks like i was mixing localToWord and translate

[pulsar jay]

I am curious: Did anybody here try or manage to combine addressables with entities? I am currently loading prefabs via addressables and then converting them into prefabs in multiple systems.

[gusty comet]

Ok I did some reading and figured out that my collision detection method I was trying to program is never going to work because it's "discrete" not "continuous". Reading for anyone who is curious about custom collision detection (since the character controller in dots is crap and rigidbodies don't work for everything) https://digitalrune.github.io/DigitalRune-Documentation/html/138fc8fe-c536-40e0-af6b-0fb7e8eb9623.htm

[pulsar jay]

I am currently thinking about moving the conversion step into the addressables system and later on possible convert dureing edit time and remap the keys to load the already converted prefabs

[gusty comet]

Meaning? I know Unity's rigidbody uses the same method as the article I linked but it doesnt go into good detail https://docs.unity3d.com/Manual/ContinuousCollisionDetection.html

[pliant pike]

I think I did it with scriptable objects

[gusty comet]

Oh you replied to the wrong person initially. Had me really confused lol

[pliant pike]

yeah sorry

[pulsar jay]

Yeah I am also using scriptable objects for some stuff already. But then there are some things that are just prefabs (e.g. VFX effects) that I would like to load and instantiate

[pliant pike]

I'm not sure if you can convert sound effects, but if you can get the raw data you can convert it however you want

[pulsar jay]

doesnt exactly matter what it is atm. Couls also be an enemy prefab for example

[pliant pike]

actually I think I came up against problems using addressables with Systems

[pulsar jay]

It would be great to just call Addressables.LoadAsset<Entity>("Enemy"); and get the already converted prefab

[pliant pike]

I found and used this code that worked

public static int TryGetUnityObjectsofTypefromPath<T>(string path, List<T> assetsFound) where T : UnityEngine.Object
    {
        string[] filepaths = System.IO.Directory.GetFiles(path);
 
        int countfound = 0;
 
        Debug.Log("Filepaths length " + filepaths.Length);
 
        if(filepaths != null && filepaths.Length > 0)
        {
            for(int i = 0; i < filepaths.Length; i++)
            {
                UnityEngine.Object obj = AssetDatabase.LoadAssetAtPath(filepaths[i], typeof(T));
                if(obj is T asset)
                {
                    countfound++;
                    if(!assetsFound.Contains(asset))
                    {
                        assetsFound.Add(asset);
                    }                        
                }
            }
        }
        return countfound;
    }```

[pulsar jay]

But loading from AssetDatabase only works in editor doesnt it?

[pliant pike]

actually you might be fine in a system I had to use that because addressables wouldn't work in a GameObjectConversnionSystem

yeah

its if you want things converted before run time

[pulsar jay]

Did you try converting prefabs at edit time?

I figured there are methods to store and load entities as binaries which I thought about using

[pliant pike]

if you put them in a subscene then they are basically converted at edit time

[pulsar jay]

yeah I thought about that. But thats not really extensible and would basically replace the addressables system

[karmic basin]

If you're worried about missing collisions between simulation steps, you can add your own failsafe system which cast a ray (or shape collider) between old and new position. If you hit something, means the engine missed a collision, so force new position.
Now you're talking about that, I can't remember if com.unity.physics does DCD or CCD 🤔 Could be interesting to check

[pliant pike]

weird theory but would it not be more optimal to have collision check on the environment objects, say if you know the max speed of the moving objects check the distance surrounding the evironment objects at that max distance, if they are close enough then you do the sphere casts or raycasts whatever, I'm curious if anyone has done it that way before 🤔

[toxic crest]

is it valid and safe to declare const members in a job struct?

of a blittable type of course

[coarse turtle]

yea - const members are just readonly data - I've done that without issues

[light mason]

wondering if someone could help me out

i have a Lambda that drags an abject around

work with Translation

but as soon as I start using LocalToWorld is stops working

i need to use localToWorld because the obj are parented

im not sure what im doing wrong here

im setting to 2 ,1, 2 just to eliminate the input

but nothing is moving

or if this looks correct that would be good to know too

[gusty comet]

@karmic basin I dont think the package has any method for CCD or DCD - I think the character controller does this manually using raycasts and whatnot. I tried the failsafe approach but started having weird issues

[stone osprey]

Quick theory question... kinda targets the ecs and its persistence. I basically managed to persist a bunch of entities in a database. That includes their relations. But i have one issue... my persisted player entity for example. This one references a bunch of other entities in the database as items ( item - entities ). So in order to make the player work properly, i need to load him and all his referenced item-entities. But what if those item-entities reference other entities again ? And those referenced reference other ones ? Do i always need to load the whole reference-tree ? Does every game load all references between entities ? Or are there tricks to determine how far i need to load something ? Is this even possible in an ecs ?

[gusty comet]

Or at least if there's any CCD implementation in the package itself I'm not aware of it (without setting up a rigidbody or copying the character controller from the physics example)

[ocean tundra]

@light mason you should use translation

You could grab localtoparent and use that to figure out the world position to local position

But you should only write to Translation, local to world will be overridden by the transform systems

[dark cypress]

You can just write to it after the transform systems run.

[ocean tundra]

Oh yeah that would work

[light mason]

@ocean tundra do you have an example .. this is driven me crazy

[ocean tundra]

I dont sorry, im at work and dont have my projects here

but try Liburia's idea

just UpdateAfter TransformSystem

[light mason]

i just was @dark cypress chat , so i have to set my system to update after transformsys?

ah

typed at the same time let me try

[light mason]

ok so i have [UpdateAfter(typeof(TransformSystemGroup))]

and this in the system

so i can drag and move

but as soon as i delete NodeMoveComponent the node jumps back to it original position

its driving me bananas

[dark cypress]

try setting both translation and ltw

[light mason]

if i ref translation and ltw the job never runs

at this point im not sure if this is a bug or not

ok so it does run but i get double transform on children

[light mason]

how do i go from ws to local space

[ocean tundra]

@light mason pretty sure theres a LocalToParent component that has the matrix toconvert to local

@light mason https://docs.unity3d.com/Packages/com.unity.entities@0.17/manual/transform_system.html#section-2-hierarchical-transforms-basic

[light mason]

I guess where im hittting a wall is the coord i was to set the entity too in in ws

[ocean tundra]

actaully i think if you used that you would take world pos and get a local pos relative to parent

not sure if thats waht your after

[light mason]

yea i think what makes sense

i just have no idea how to get that ws to ls

[ocean tundra]

take your position and devide? by the matrix

[light mason]

so i add my ws point to a matrix

and devide my LocatToWorld ?

[ocean tundra]

@light mason i really hope you understand matrixes

but im pretty sure you should never touch matrix values directly

theres some helper methods in the new maths library

[frosty chasm]

Hi, i wanted to use DOTS but i can't see Entity in package manager

[ocean tundra]

@frosty chasm https://docs.unity3d.com/Packages/com.unity.entities@0.17/manual/install_setup.html

[sly jasper]

If I want to copy a very large float4x4 array into Matrix4x4 array, are there any good ways to do so?

[frosty chasm]

ty

[light mason]

ok got it working

adding the mul of the inverse of locatToWorld and the worldPoint

yea i have the basic down of matrices but im still finding my why around the API

api

[lusty otter]

I need to generate a waveform texture from audio clip data and I'm using Job system to speed it up, but AudioClip.GetData doesn't have a native container variant.

Is there a way to make a native array of a managed array so I can pass it to the job without making another copy and taking up twice the memory?

The job only needs read access.

[coarse turtle]

You can try to pin the array and grab a pointer to pass to your jobs or use the fixed statement;

// Fixed
float[] blah = new float[x] { ... };

fixed (float* ptr = blah) {
}

// Pinning
GCHandle handle = GCHandle.Alloc(blah, GCHandleType.Pinned);
IntPtr intPtr = handle.AddrOfPinnedObject();

// Pass this to job
float* ptr = (float*)intPtr.ToPointer();

// Free the pinned resource
handle.Free();

// Job Example
unsafe struct SomeJob : IJob {
  [NativeDisableUnsafePtrRestriction]
  public float* Ptr;

  public int Length;

  public void Execute() {
    for (int i = 0; i < Length; i++) {
      float content = Ptr[i];
      // if you want to access the pointer
      float* f = Ptr + i;
    }
  }
}

[ocean tundra]

I know we are not meant to change the state of a struct from a Method but is there any issues if we do so anyway?

I'm reading data from a network stream, i then create the struct and pass in the DataStreamReader to load up the rest of the values

        {
            EntityId = reader.ReadInt();```

[ocean tundra]

Also anyone know how to use DataStreamReader.ReadBytes

i have no idea how unsafe pointers work

this seems like crazyness

            fixed (double* timePointer = &Time)
            {
                reader.ReadBytes( (byte* )timePointer, sizeof(double));
            }

and to write if anyone is curious
writer.WriteBytes((byte*) &time, sizeof(double));

[sly jasper]

I managed to convert it using UnsafeUtility.MemMove

[ocean tundra]

Well it seems that the read/write bytes I'm doing doesn't work

Anyone have any ideas?

[quaint bear]

Hello to everyone or anyone who can point me to a general direction of code..
So I want to make an editor window which will be populated with entites in my world with their entity index.

I want to select a button in that list and then that entity should be highlighted inside entity debugger.

Something like

selection.activeGameObject= gameobject. 

//Similarly,
selection.activeEntity= entityList[entityIndexFromList] 
//The above should select the entity from the entity debugger and show the components in inspector as usual.. 
//Just need an easier way to find entites, will expand this to use specific components later.. (Dots Editor package is not that friendly)

Please ping me if you have any ideas. I'd love to hear some suggestions.. thanks!

[whole gyro]

@quaint bear you'll probably want to look into EntitySelectionProxy for selecting the entity: https://docs.unity3d.com/Packages/com.unity.entities@0.17/api/Unity.Entities.Editor.EntitySelectionProxy.html

[lusty otter]

Thanks, seems just like what I need.

[quaint bear]

Hey, thanks!
I'll take a look :thumbsup:

[ocean tundra]

Does anyone have any ideas how to get a nice unique ID for every entity in a subscene. The ID needs to be saved in the scene (inside a monobehaviour) so it persists

If it helps this ID is the entities network id, both client and server load the same subscene but cause of entity remapping the entity index and version can change so I can't use that. So need a ID to help link them

[karmic basin]

So a UUID ?

Something like that https://docs.microsoft.com/en-us/dotnet/api/system.guid

[pliant pike]

I think I got a a unique id from AssetDatabase.AssetPathToGUID("Assets/Scr....

[karmic basin]

Nice one, though only for assets 😛

[charred thistle]

I wanted to learn DOTS again, but I haven't played with him for so long that I don't even know where to start =/

[agile dome]

I found samples very useful when I was exploring Unity Tiny. Pinned post has all the links

[odd ridge]

how do I implement collision detection in DOTS? say I have thousands of characters, I want to detect when a character is close enough to start fighting another

[coarse turtle]

sounds like you need a spatial data structure to reduce the # of distance checks, otherwise if you want a naive approach iterate through everything (n^2 operation)

[warm osprey]

How does EntityManager.SetEnabled(entity, true) know, which of entity's children acquired the Disabled tag by EntityManager.SetEnabled(entity, false), and which of them acquired the Disabled tag by other means? Because EntityManager.SetEnabled(entity, true) will not remove the Disabled tag from children that acquired it by other means.

I'm asking because I'd like to create a large multitude of hierarchies on a worker thread using ECB, which are disabled at creation time. I'd like to enable them at a later time with EntityManager.SetEnabled (so I don't need to traverse the hierarchy myself to remove the Disabled tag from the children and grandchildren). But I just can't find a way to do that correctly.

[ocean tundra]

@pliant pike @karmic basin I've considered guids but size wise they are kinda big 128? Bytes where a int is just 32

[amber flicker]

Once you have a unique id you can e.g. build a runtime hashmap or something if you want to less data over the wire. I would recommend something like Mr K mentioned. You can check out how DOTS Timeline does it for their assets.

(or buy my tagging solution on the store - but that sounds a little overkill 😉 )

[ocean tundra]

Hmmm so everything has a nice easy guid, but at runtime the server maps that guid to a nice small int and keeps that list in sync over the network

[amber flicker]

I mean.. like all of these solutions, maintaining a table/database always ends up kind of fiddly. Especially when you factor in old versions of a client, addressables etc etc. Versioning is probably quite a large thing to think about.

[ocean tundra]

Yea there's lots of annoyance to this id thing

Putting in a app version check is on the todo list

And honestly don't think I can use addressables, many of my older versions did. But I've embrased subscenes and I don't think we can make a subscene addressable

[amber flicker]

I haven't tried but it seems like it should be quite possible? A ScriptableObject could reference a subscene no? At any rate, not saying it's trivial 🙂

[whole gyro]

Has anyone been able to serialize entities from a subscene? I recently switched from GameObjectConversionUtility to subscenes for converting my prefab entities. Its all working great in game, however, when I try to serialize the world using SerializeUtility, it fails to serialize any entities from the subscene. I believe this is because SerializeUtility doesn't support remapping and serializing entity references in SharedComponents, but the subscene is automatically adding a SceneTag shared component with a reference to the subscene entity. Do I need to manually remove these subscene components before serialization?

[ocean tundra]

@whole gyro wait there's a component on each subscene entity with the original subscene entity?

How was it not automaticly remapped?

As for serialize untility I've used it lots without subscenes and it was super annoying but never with subscenes

[whole gyro]

Yeah, these components are on all of my subscene entities:

I was able to get serialize utility working before I switched to subscenes (after doing some mesh -> ID substitution). Only difference now are these extra components.

Maybe I need to use SerializeUtilityHybrid... I just really don't understand what I'm supposed to do with the ReferencedUnityObjects that it gives me back when I call Serialize().

Yeah okay, removing the SceneTag component before calling SerializeUtility.SerializeWorld() works fine. Only difference from what I can tell is the entity window no longer groups the entity under the subscene.

[whole gyro]

Oh actually I misread your question. No its not an entity reference to the original subscene entity. Its an entity reference to a special selection entity for the entire subscene. Not really sure how its used but my guess is that it has something to do with livelink and loading/unloading subscenes in the editor.

[ocean tundra]

@whole gyro thanks for clarifying. Having the original entity doesn't actually solve my issue. I'm thinking some sort of editor script that can create nice int IDs and save it into a mb with IConvert

[shut hare]

How can you get a entity through a gameobject with the new hybrid workflow ?

[ocean tundra]

@shut hare I think it works the other way, Entity => gameobject

you could have a MB that you add to the entity, then in a system set a public Entity field on the MB

[shut hare]

I made the logic the other way around with the old workflow GameObjectEntity

[ocean tundra]

the MB would need to be added as a hybrid component

[shut hare]

what do you mean by MB ?

[ocean tundra]

MonoBehaviour

[shut hare]

ok so the MB would be empty ?

[ocean tundra]

basilcy, just a entity field

[shut hare]

isn't it the same thing as the GameObjectEntity class ?

[ocean tundra]

No clue sorry

[shut hare]

its ok thx i'll tinker on what you said

[plush meadow]

As far as I understand, entities with different meshes are split into different chunks to reduce drawcalls. I just want to make sure that if each of those entities all have different transform matrices (like scaled differently, sheared and such) it will still have the same effect right?

I'm not very knowledgeable about graphics stuff so sorry if this is too obvious of a question

[coarse turtle]

not to my knowledge, because the LocalToWorld is not a shared component data, it can fall within the same chunk

Mesh 1 -> chunk has LocalToWorld matrices, A, B, C with different values
Mesh 2 -> chunk has LocalToWorld matrices, W, X, Y with different values

So A, B, C should fall within the same chunk because they share the same archetype + SharedComponentData, same goes for W, X, Y - it will fall into the same chunk that Mesh B is a part of

[plush meadow]

I see, so that's still useful at least

But will having different matrices have any effects on number of drawcalls?

[coarse turtle]

depends on how it is drawn, I assume you are using hybrid renderer - which I have little experience in using

[plush meadow]

I am using hybrid

[coarse turtle]

generally I don't think so, since hybrid renderer probably uses indirect drawing/instancing to reduce the drawcalls. Someone else whose worked with it more actively can provide a much better answer than me

[plush meadow]

I see, your answer still helped a lot though

[worthy rampart]

generally speaking no one would ever be wanting to render the same mesh with the same transform

[sly jasper]

From my understanding, it should work for different transform matrices. You need to do some tricks only if they have different material properties.

And I think the renderer is using direct instancing. There is a batch limit of 1023

[sage mulch]

Is there any way to access a material property for an entity without using ShaderGraph?

[ocean tundra]

yup

[sage mulch]

Could you link me to documentation please? Everything I've found is for ShaderGraph

[ocean tundra]

Oh wait

https://docs.unity3d.com/Packages/com.unity.rendering.hybrid@0.11/manual/material_overrides.html

im pretty sure that needs shadergraph as well

[sage mulch]

Yike

Ok ty

Just spent ages learning how to write shaders lmao

Should have done my research

[ocean tundra]

😛

i remember skimming over some fourm posts AGES ago about how to enable the overrides for custom shaders

no idea if they succcessed

[sage mulch]

Ah ok

Wonder if that's planned for the future

Or if they're just leaving custom shaders behind

[ocean tundra]

honestly if your in URP or HDRP i would just use shader graph

[sage mulch]

Is there a performance impact?

I figured it would be slower than writing one tbh

[ocean tundra]

well the whole old render pipline is being replaced with those

[sage mulch]

Ah

[ocean tundra]

umm not too sure, I think URP is faster

but HDRP is prettier 😛

[sage mulch]

This is a very general question, but would you say anything that is possible via written shader is possible via ShaderGraph?

My shader is super weird and relies on the camera direction, idk if ShaderGraph even supports that

[ocean tundra]

URP is meant to match features of built in, but be faster, HDRP is meant for the crazy HD graphics

not sure sorry, my shader graph skills are super low

[sage mulch]

Np, ty for the help!

[ocean tundra]

i mainly use URP as it is, little to no customization

eventually i want to get better at it but graphics always seems to be lower down on my list

[sage mulch]

Yeah I'm just trying to do some multi-directional sprite nonsense

2.5d

[ocean tundra]

sounds cool tho

maybe do them in code, with jobs + burst it could be super fast

and you get more maintainable c# code instead of shader magic

[sage mulch]

That's good advice, ty

I'll def look into it

[ocean tundra]

good luck, sorry i couldnt help with the shader stuff

[safe lintel]

no you dont need shadergraph

but you do obviously need to make your shader be compatible with one of the renderpipelines and hybrid renderer compatible

theres a bit in the shader that goes like CBUFFER_START(UnityPerMaterial) and this is where you declare that properties and then a little below another entry where you declare again that they are instancing compatible UNITY_DOTS_INSTANCING_START(MaterialPropertyMetadata) ie UNITY_DOTS_INSTANCED_PROP(float4, _BaseColor)

[ocean tundra]

Anyone tried getting DOTS and hybrid renderer working over 2 screens?

I need a hook during the Conversion flow on a MonoBehaviour, anyone know if that exists

Before conversion, a component attached to the SubScene needs a callback like OnEnable

[light mason]

have a question i was wondering if someone could help me our with

if i have an Enitites.ForEach( DynamicBuffer<Foo>

and i use and cmdBuffer to create a new Instantiate of foo

can i use add(foo) inside the lambda

as in Add to the buffer of and entity i instantiated with ecb

hope that question makes sense

or is there an command buffer method i should use to add the new element

ok i found AppendToBuffer()

[vagrant surge]

probably interesting for the ECS fans https://github.com/bevyengine/bevy/pull/1525

thats from the Rust gamedev community, for a new engine project. The interesting part is that the ecs is a hybrid ECS

components are stored in archetypes and data chunk by default, but its possible to remove them from there and put them into sparse sets like in entt or entitas for some component types

[ocean tundra]

hmmm its a interesting idea

being able to specify if the component should be optimised for iteration speed or add/remove performance

honestly sounds like a good idea, more controls to us

[vagrant surge]

indeed

he also implements some of the ideas from Flecs like archetype graphs

that act as a big optimization around add/remove component for archetypes

one of the interesting things about such an approahc for the hybrid

is that you can go even more hybrid

not just storing the components into sparse sets

but for example something like a hashmap

or other storage that could store entityID->component

[ocean tundra]

but also does optimising tag components to have little/no cost and adding the enabled component optimisation/feature also give us the same things

[vagrant surge]

not exactly

being able to use sparse sets for some components makes the typical "state machine" pattern work a lot better

important to know that this ecs also makes tag components much faster and has component enable/disable

[ocean tundra]

very cool

[vagrant surge]

ecs libraries are starting to go quite a bit beyond the feature set of unity ecs

on the cpp side we have Flecs which is highly advanced, and then this now for rust

[ocean tundra]

maybe after we get those 2 features in DOTS (in 2 years 😛 ) they could consider something like this

[vagrant surge]

btw, Rust works incredible for ECS

[ocean tundra]

yea i bet

[vagrant surge]

its just a perfect fit for the language

rust is a functional-ish highly parallel language

where memory management is a huge pain due to the borrowchecker

buuut

if you use a ecs to manage the memory for you, thats a total headache you remove

[ocean tundra]

i doubt ill be learning rust anytime soon

work is all mobile/web/c#/.netcore

and free time is all unity 😛

i really like the idea of us being able to choose that optmization

and it would be quite simple too, unity could probably make it into a Attribute we add on our IComponents

[deft stump]

C# forever

[gusty comet]

shows the priority of the engine. They don't have a pipeline, or even proper shadow rendering, but a rewrite of the ecs is more important than those anyway 😎

thats like godot adding dynamic gi before they had occlusion culling, or even bare minimum acceptable frustum culling

in bevy's case it shouldn't be that surprising I guess, because the whole engine came into being because the author didn't like amethyst's ecs (amethyst itself is upgrading their ecs btw, at the cost of killing all their momentum!)

[safe lintel]

no editor is a big stopper for me 🙂

do wish unity put more of their development in the open though

[deft stump]

I wish Unity makes their engine open source

Or other than that... have their trello board open for the public to see

[safe lintel]

probably only the second is really feasible

[dark cypress]

How can I get the entity itself in IJobChunk?

[coarse turtle]

You pass in an EntityTypeHandle to the job struct

struct ChunkJob : IJobChunk {
  [ReadOnly] public EntityTypeHandle EntityHandle;
  
  public void Execute(ArchetypeChunk chunk, ...) {
    var entities = chunk.GetNativeArray(EntityHandle);
    for (int i = 0; i < chunk.Count; i++) {
      var e = entities[i];
    }
  }
}

also on a separate related note - I recently saw this post from godot on ecs architecture: https://godotengine.org/article/why-isnt-godot-ecs-based-game-engine Might be interesting to check out on what their concerns are 👀

[dark cypress]

Where do I get the EntityTypeHandle?

[coarse turtle]

very similar to how you get a ComponentTypeHandle, from SystemBase it's calling the GetEntityTypeHandle() function

from ISystemBase it's usually systemState.GetEntityTypeHandle()

[pliant pike]

no matter how many times I've seen the word polymorphism and it explained I still never understand it

[dark cypress]

Thank you.

[light mason]

Can I have Entity field on an GeneateAuthoringComponent and set it later in a system?

or is that not a allowed

[safe lintel]

what do you mean?

[light mason]

I have public Entity on in a [GenerateAuthoringComponent]

the entity i want to ref will be created in a sys

or should i spilt that into a seperate comp and create and set in a sys

[dark cypress]

[GenerateAuthoringComponent] just makes a monobehavior for your component.

[safe lintel]

I dont see why not. GenerateAuthoringComponent really only relates to the editor workflow

[dark cypress]

The component itself functions as normal.

[light mason]

so like a node lib you create in editor

you instantiate but you want to set a ref go other entity

node are prefabs in the example above

Bad example above but hopefully it makes sense

[ocean tundra]

Is there a good way to get a NativeHashMap to auto grow as it fills up?

[worthy rampart]

write an extension method that grows it automatically on insertion 🤣

[ocean tundra]

dont think i can do that from a job tho

[worthy rampart]

I don't think ya can either

[ocean tundra]

hmmm

maybe it already auto grows

From the comments: The number of items that the container can hold before it resizes its internal storage.

nope guess not

[timber ivy]

@ocean tundra you were suggesting a* to me, does a* allow you to tell the agents to follow a path at some offset?

[viral comet]

How often can I bump a thread on the Unity dots forum?

Cant find any info/rules

[timber ivy]

Do i need to use capture variables inside entities.foreach that ref the component?

[ocean tundra]

@timber ivy thats not really a a* (or any pathfinding) thing

[timber ivy]

it kinda is, because the built in pathfinding always finds the shortest path between two points

and it results in the ai moving in a perfectly straight line

[ocean tundra]

with pathfinding it helps to break it into 2 parts

1. the acutal pathfinding, this is finding the best path between 2 points
   2, the actual following of the path, this can be many different options depending on the game type

i belive that built in pathfinding has both concepts mixed together

[timber ivy]

Any ideas how i can get the ai to spread out and not fall in a straight line?

[ocean tundra]

but for me, i've been messing with RTS games, so a space game pathfinding and medevial pathfinding both works the same for step 1, but following is completly different

[timber ivy]

this is what i got going on now

and it just looks like shit

[ocean tundra]

so again depends on the game, for my space one i had units in a 'squadron' the squad would pathfind (step 1) and then use flocking steering behaviours to follow

[timber ivy]

the navmesh is litterally the whole sidewalk

but they utilize only the center

(this is unities built in navmesh and pathfinding btw)

i tried offsetting the destination

but that didn't change anything

[ocean tundra]

so for crowds i would use pathfinding + (steering behaviours OR RVO (also known as local avoidance)

so yes the path would always be the same for all those people

but the actual following is different

[timber ivy]

Ah so id have to code my own navmeshagent system basically

instead of using unities?

[ocean tundra]

i've never used unity's but i would guess yes

[worthy rampart]

I'd also vary the speed on em slightly

And make sure they can avoid collisions with each other

[ocean tundra]

yea thats a great idea

[worthy rampart]

As no one literally walks in sync

[timber ivy]

yeah

[worthy rampart]

People walk slower/past each other

[timber ivy]

it looks so robotic rn

[worthy rampart]

It's the little things like that

That take it from looking robotic to not

[ocean tundra]

i'm trying to find a good article on steering behaviours, there was a good one that had cool animations

[worthy rampart]

Also do they move their arms

[ocean tundra]

local avoidance is a MASSIVE pain 😛

[worthy rampart]

Or anything like that

[timber ivy]

nah there is just an idle animation rn

I haven't really messed with animating it yet

just wanna spread them out a little for now but apparently that isn't possible with built in navmeshagents

[ocean tundra]

i think these were the ones that made the most sense to mee

https://gamedevelopment.tutsplus.com/series/understanding-steering-behaviors--gamedev-12732

[timber ivy]

ig this just gives me an excuse to make a dots compatible pathing system

[ocean tundra]

cant find animations

https://gamedevelopment.tutsplus.com/tutorials/understanding-steering-behaviors-collision-avoidance--gamedev-7777

https://gamedevelopment.tutsplus.com/tutorials/understanding-steering-behaviors-path-following--gamedev-8769

those are the 2 ones most fitting for you i think

[timber ivy]

Awesome

[ocean tundra]

(damn embedded links taking up the whole page, sorry everyone)

[timber ivy]

awesome thank you roy

[ocean tundra]

not sure how fitting for dots they are

will likly need to rebuld loads

[timber ivy]

I will be able to make something happen surely but

for now this will get me started at least

its actually nuts to me that something like pathfinding for AI is so limited with the built in shit

[ocean tundra]

yup

i've always used https://arongranberg.com/astar/features

but sadly compatibility with DOTS is lacking

There is also https://github.com/dotsnav/dotsnav but its very new, looks really promising tho

[timber ivy]

Thats cool

prolly a little heavy for me though

[ocean tundra]

a star pathfinding project sure

but dotsnav looks really simple

nice and clean

[ocean tundra]

Please use code tags

[light mason]

wondering what im missing here

for the image?

[ocean tundra]

for code

[light mason]

what code

[ocean tundra]

any code

dont take pics copy paste it into code tags

[light mason]

it wont let me add code

[ocean tundra]

are you using code tags?

[light mason]

yea

it just dissapears

[ocean tundra]

                            networkDriver.BeginSend(connection.Value, out DataStreamWriter writer);
                            writer.WriteByte((byte) GameNetworkProtocol.CurveData);
                            //todo generate a nicer typeId
                            var typeId = 1;
                            writer.WriteInt(typeId);

[light mason]

that why i was wondering if you saw any code from me

[ocean tundra]

weird works fine for me

using `?

3 of them

start and end of code

[light mason]

just `

ooo ok

[ocean tundra]

but yea that code looked fine

whats the issue

[light mason]

so the data doesnt appear to get set

[ocean tundra]

you mean the data getcomponent?

you will need to call setcomponent somewhere

[timber ivy]

Wonder what would happen if i parented my ai to an empty object

and offset the child by a value

[light mason]

                var data = GetComponent<NodeBezierMasterComponent>(connectEntity);
                // Set EndNode Entity  Hit target node
                Entities
                    .WithoutBurst()
                    .ForEach((Entity e, int entityInQueryIndex, ref NodeBaseComponent s, in LocalToWorld l) =>
                    {
                        data.EndNode = e;
                        data.CurveDotsMax = 10;
                        ecb.SetComponent( connectEntity, new NodeBezierMasterComponent()
                        {
                            CurveDotsMax = data.CurveDotsMax,
                            StartNode = data.StartNode,
                            EndNode = data.EndNode
                        });
   
                    }).Schedule();

                
                
                
                // Create Dots 
                Entities
                    .WithoutBurst()
                    .ForEach((Entity e, ref NodeConnectComponent c, in LocalToWorld l) =>
                    {
                        Debug.Log("____0");
                        Debug.Log(data.CurveDotsMax);
                        for (int d = 0; d < 10; d++)
                        {
                            Debug.Log("Creating curve dots like never before son!");
                            Entity curveDotEntity = ecb.Instantiate(nodeLibBuffer[3]);
                            ecb.SetComponent(curveDotEntity, new Translation {Value = new float3(inputRecordData.MouseGridX, d, inputRecordData.MouseGridZ)});

                        }
                        Debug.Log("_____1");
                        ecb.RemoveComponent<NodeConnectComponent>(connectEntity);
                    }).Schedule();```

Debug.Log(data.CurveDotsMax); is 0

what am i missing here

[ocean tundra]

soo

                        data.EndNode = e;
                        data.CurveDotsMax = 10;``` will never survive

your not saving data again

also these are both jobs, thats not how you share data between jobs

[light mason]

ok so if NodeBezierMasterComponent is a singleton how do i write data into i from different entitie in a forEach

[ocean tundra]

well you shouldnt

[light mason]

because I cant capture either

[ocean tundra]

as it will constanly be replaced

your looping over every entity with .ForEach((Entity e, int entityInQueryIndex, ref NodeBaseComponent s, in LocalToWorld l) =>

theres no guarantee of ordering

whatever the last entity you loop over will be the one your saving to the singleton

not really sure what your trying to achieve

[light mason]

im have a nighmare getting tick to work

having

[ocean tundra]

what are you trying todo?

[light mason]

trying to add the code but it not working

it on a cell hit w node

[ocean tundra]

no dont add code 😛

explain the high level steps your wanting

[light mason]

what i added before was not correct

I have a singleton

when i hit a node or enity of NodeBase I want to update the singleton

[ocean tundra]

what do you mean by 'i' and 'hit'?

as in player collides?

[light mason]

no it mouse in a grid

[timber ivy]

@worthy rampart you're a legend man. By making each ai having a randomized speed set, it made the offsets happen naturually cause it causes the path to be recalculated when the agent has to steer past slower moving agents

[ocean tundra]

@timber ivy keep in mind the agents will be recalculating every X (frame? seconds?) which is not great

[timber ivy]

Huh?

[ocean tundra]

performance wide

[timber ivy]

i mean

i don't think it will be an issue

i got 1200 rn

30 fps on a business laptop

[light mason]

wait.. i just need to GetComoponent in the lambda

right

[timber ivy]

and the ai stream in and out based on camera position, so i don't really think there will ever be more than a few hundred on the screen at any time tops

[ocean tundra]

@timber ivy sweet

my performance goals are crazy high so that would never work for me 😛

[timber ivy]

how many are you trying to have on screen at once?

[worthy rampart]

Glad I could help :)

Have fun

[ocean tundra]

@light mason still lost at what your trying to do sorry
when i did mouse grid stuff i did the following:
raycast to world
take Hit and devide by grid size, this would give me the grid int x and y
i had a native hashmap of int2 (x y) to entity to find the entity for that grid

@timber ivy ALL of them 😛

[timber ivy]

I wonder if giving each one a random "width" for the agent collider would add more variety as well

[ocean tundra]

i usually try to make management/strategy games and want to have no unit caps (limited only by CPU speed 😛 )

[timber ivy]

yeah you can't really get away with streaming them in and out

and reusing them like i can

with rts type games

[ocean tundra]

yup 😦

[timber ivy]

Speaking of streaming

i wonder when unity will get built in level streamer

like ue4 has LOL

[ocean tundra]

maybe, i think subscenes are a step in that direction

[timber ivy]

Yeah I agree

[ocean tundra]

they are kinda painful still, but their super quick / background loading seems solid

[timber ivy]

is it better than LoadLevelAsync

[ocean tundra]

i dont think it would be that hard to build the next layer with them as they are

all you really need is a way to map int2 (chunk X and Y) to a scene GUID

[light mason]

@ocean tundra ok let me regroup here , thank for talking it over with me

[timber ivy]

I have one i wrote a while ago that works that way

but it uses LoadLevelAsync and asset bundles

the GUID is just the asset bundle name

worked ok i guess but deff had some frame drops

are subscenes usable with DOTS?

[ocean tundra]

@timber ivy what do you mean by usable 😛

[timber ivy]

can I use them in parralel jobs?

or are they only exposed to main thread

[ocean tundra]

umm both??

😛

theres a SceneSystem you use

call methods on that from main thread

OR create a entity with the right components using a ECB

(the system hepler thread does that for you)

then it all automagicly happens

eg

 var sceneEntity = world1.GetExistingSystem<SceneSystem>().LoadSceneAsync(data.SceneGUID,
                    new SceneSystem.LoadParameters()
                    {
                        Flags = SceneLoadFlags.NewInstance
                    });

internally that does some stuff to find the scene reference and create a entity with a few components

for me seems to work fine

worst part is I cant find any good hooks into the conversion process

specifily i would like to convert a subscene 2ce, into a Server and Client version

[timber ivy]

Hmm since it converts everything to entities

does that mean some of the built in components in unity

aren't compatible with subscenes?

[ocean tundra]

theres that hybrid/companion thing

that does work with subscenes

you have to manually call something but you can choose to persist old MB's and unity components into the subscene

but by default i dont think it does that with anything

[timber ivy]

Ah right

got ya

thanks!

[ocean tundra]

and this is all the docs

https://docs.unity3d.com/Packages/com.unity.entities@0.17/manual/conversion.html

@timber ivy oh and livelink is kinda a pain

[timber ivy]

livelink?

[ocean tundra]

most of my issues have been around that, as the concept is great, edit your authoing GO and it will just 'magicly' update the entity version

during runtime

or even during a build

but in practice its a pain to keep working

livelink basicly will stream over your editing changes during runtime, you can even attach to builds and do it

[timber ivy]

interesting

[ocean tundra]

in concept its great

and it actualy works, in smaller demos/samples

but im struggling with it. again mainly cause server client.

i want it to work, but only into the server world, my standard networking stuff can take care of sending the data to a client

but you cant edit or control livelink at all

no way to turn it off for a world either

Question for everyone
What built in components do people think should have networking support by default?
So far I have Translation, Rotation, Scale
What else do people expect to be synchronised?

[timber ivy]

Are you creating a net library?

[ocean tundra]

yup

[timber ivy]

Tbh man

I don't think they should be automatically synced

idk what you're going for

[ocean tundra]

no its up to the user to decide if it gets synced

[timber ivy]

but it sounds like you're trying to make something high level

[worthy rampart]

Roycon I think you should give the models a tiny scaling

Russel

Sorry

Like 5%

[ocean tundra]

sorry the question is more like, Which built in types should have support for networking

[worthy rampart]

Random

[timber ivy]

ah

thats a good idea too derek

and roy something I always hated having to manually sync back when I used pun was animations

[worthy rampart]

I'd give it as much variance as you can

Without it looking wierd

Eventually you'll probably want some different models too tho

[timber ivy]

Yeah i will have some diff ones soon

there is still like

[worthy rampart]

There's always 100 other things to do

[timber ivy]

9k free modelsa nd animations on mixamos website

that i am going to grab here soon

to use with this

[ocean tundra]

@timber ivy Ooo i need to dig into the animations package to see what animation components i need to add for syncing

good idea TY

[timber ivy]

no problem

but please make this optional

[ocean tundra]

yup

[timber ivy]

and give the user the ability to serialize their own data for transport

[ocean tundra]

its 2 parts

[timber ivy]

what are you using to handle the networking itself?

c# sockets?

[ocean tundra]

Code gen, which creates the types to allow networking

[timber ivy]

nice thats cool

but what are you using to transport data cause

if you're doing an rts

i really wouldn't use C# sockets

[ocean tundra]

then on component you select what data types to sync

[timber ivy]

or anything that uses c# sockets internally

[ocean tundra]

nope im using Unity.Transport atm

[timber ivy]

Idk what it uses internally

[ocean tundra]

but it should be easily swapable

some very low level udp c++ socket i think

[timber ivy]

might want to check into

[ocean tundra]

yup

[timber ivy]

https://github.com/nxrighthere/ENet-CSharp

league of legends uses enet

so its proven at least

[ocean tundra]

but it should be swappable, all i need todo is create some different code gen templates and put some toggles into the code generator

yup my last prototypes used enet

[timber ivy]

yeah that guy has some

Really good resources on his github

[ocean tundra]

the only reason ive started with transport this time is i want Full job support

[timber ivy]

oh someone used it with jobs

and ecs

[ocean tundra]

where with eNet i found integrating with jobs a pain and just gave up 😛

[timber ivy]

there was a post in the dots forum

about how he did it

https://forum.unity.com/proxy.php?image=https%3A%2F%2Fi.imgur.com%2F19gLolB.png&hash=4343145dd600d0eca45968636d5c2b34

So yeah

unity.transport deff way easier to use if you want full job support

[ocean tundra]

interesting

[timber ivy]

He used io threads

and a lockless q instead

so there wasn't the constant overhead of spinning up a new job

[ocean tundra]

yea makes sense. thats what my first prototype did too

only 1 tho, the network thread

oh and a serialisation thread 😛 so 2 threads

[timber ivy]

yeah so it was the same as his 😉

2 threads + main thread for game code

I did something similar in this

https://github.com/russDevAcc/Simulnet

still need to finish that 😦

[ocean tundra]

server to server communication sounds cool

[timber ivy]

Yeah i was working on

handing over a player

[ocean tundra]

the ability to break apart the server to have more resources...

[timber ivy]

from one server to another when I got side tracked with work

[ocean tundra]

could have sooooo many entities

[timber ivy]

this is what i was going for https://www.youtube.com/watch?v=gq1vDG-st1k

[ocean tundra]

i feel like to do that right the gameplay needs to kinda enable that

[timber ivy]

yeah

tbh

what i was doing was

connecting to the new server

then disconnecting from existing

after you get too far

so you were not constantly jumping servers if you walked back in fourth

The thing that kinda fucked with me though was making the handover seamless

[ocean tundra]

yea..

[timber ivy]

like

if you were aiming your gun

and walked across

your entire player state got reset

[ocean tundra]

for something like that, i would look at actor frameowkrs, something like project orleans would give you that whole dynamic server adding and removing

[timber ivy]

yeah

those things are so complex

tho

[ocean tundra]

you would need a bit of custom code to handle the position based server

yea very

[timber ivy]

idk if you ever really messed with akka

but that shit is so damn complex

[ocean tundra]

orleans would get you like 50% there tho

orleans and akka are very simlar

[timber ivy]

but even then

youd have to get the gameplay code right

[ocean tundra]

yea the other 50% is still huge 😛

[timber ivy]

to handle server handovers

[ocean tundra]

with orleans you wouldnt handover

[timber ivy]

what would you do?

just keep adding new servers?

[ocean tundra]

the indivudial actor would move but your connection to orleans is to the clustor

[timber ivy]

ah really?

[ocean tundra]

it will eventually notice that all your messages are now going to server x instead and move your connection

somehow you could probably bump that to do it early

the biggest issue is orleans recommends something else inbetween, like web apis to handle auth and things

[timber ivy]

between the user and the cluster?

[ocean tundra]

yea, the cluster is meant for 'backend'

[timber ivy]

could just have a relay server inbetween

that just bounces messages back and fourth

[ocean tundra]

yea probably

[timber ivy]

but not really suitable for some games as that would just make latency

[ocean tundra]

but again more to build

[timber ivy]

but for an mmo

it would be fine

[minor sapphire]

has anyone had issues with ICustomBootstrap not loading when using a custom build configuration, but it works if you do the default unity build (ctrl-shift-b)?

boostrap just not getting called 😐

[ocean tundra]

oh no

dont say that

i use a Custom Bootstrap

but i havnt started any builds yet

@minor sapphire got any links for how to setup the builds? and ill give it a try

we need to use that new build pipleine thing right?

[minor sapphire]

yeah

https://forum.unity.com/threads/cant-get-icustombootstrap-to-work-in-build.968544/

take a look at this thread, he has the same issue

but

it looks like it might be a package issue

if you have your own ICustomBootstrap (not in a package) you might be fine

I'm about to test moving the bootstrap out of the package

[ocean tundra]

well thats annoying, i eventually want to distribute as a package

[minor sapphire]

I haven't confirmed yet

[timber ivy]

what on earth is icustombootstrap

[ocean tundra]

its some 'magic' you can use to control where/when your systems get created

[minor sapphire]

it's a way to override the default bootstrap, the bootstrap sets up the world and default systems

[ocean tundra]

i use it to configure a Server and Client world with different Systems

[minor sapphire]

yes which is my situation too

[ocean tundra]

and also to exclude my systems from the default world

are you building some network tech too?

[minor sapphire]

no I'm trying out dotsnet

it has bootstrap for server/client...

[ocean tundra]

Is there a way to tell what packages are installed in a project?

[minor sapphire]

from code?

[ocean tundra]

Ideally I only want to add Types that are actaully installed to a project, eg no point adding Animation networking support if no animation package installed

yea from code

i guess i could just read the manifest json

seems like a terriable idea tho

[minor sapphire]

Hmm if you know what packages you are looking for potentially you could use reflection

[ocean tundra]

yea probably refelction 😦

[minor sapphire]

if it's an editor only situation, why not read the manifest? 😄

oh but I suppose they could have local packages

maybe reflection

[ocean tundra]

didnt think of local packages

yea refelct over every loaded assembly seems like the best way

then over every single type 😛

[ocean tundra]

When people want to send a network command (from client to server) (eg player move) which sounds better?

1. Find the player entity and add a item to the command buffer
2. Add a command component to any entity (not a buffer so can only add 1 per entity)

[minor sapphire]

2 seems non-viable. One command per frame?

[ocean tundra]

well you could create more then 1 entity

or if its a unit in a rts add it to many units eg every unit you have selected

im starting to think that both approchs might be right, per entity commands for things like unit movements and a more 'general' command pattern for other things, eg chat or building construction

[ocean tundra]

oh burst isnt on by default interesting

It works!!

Thats about 650 worst case entitys syncing translation and rotation every frame

The delay is cause I've hardcoded a 1sec delay in untill i get a proper network time working

[fallen ridge]

Cool one wowie

[sage mulch]

Is it possible to access a ShaderGraph shader's property through a System?

URP

[ocean tundra]

@sage mulch Did we already have a discussion about this?

Maybe I sent the wrong link

https://docs.unity3d.com/Packages/com.unity.rendering.hybrid@0.11/manual/index.html#built-in-material-property-overrides-v2-only

Does ECS clean up entitys with no components on them?

[deft stump]

no

they get reused

[ocean tundra]

wait so ECS does deal with them

[deft stump]

yeah.
They get reused.

[north bay]

You need to explicitly destroy them. They do not get destroyed for you once they have no components

[ocean tundra]

How do I query for them. Is there a empty query somewhere?

[north bay]

I don't think there is, how do you end up with empy entities?

[deft stump]

and why would you need to get empty entities anyways?

to fill them up again?

[hollow sorrel]

asmdef have conditionals based on what other asmdefs are present and what versions

[sage mulch]

@ocean tundra yeah we did, was hoping for some more input on it. Thanks again for the link but I can't seem to see the "Hybrid instanced (experimental)" toggle. Is the "Exposed" toggle that I do see the same thing, just renamed?

[frosty siren]

Can someone clarify what ISystemBase is for? What can we do with it for now? I have read that ISystemBase is to write main threaded job with burst.

[zenith wyvern]

It lets you burst your OnUpdate method, so job scheduling itself will be bursted. It's currently very limited though and only supports IJobChunk.

[frosty siren]

Oh, i see. You know, lambda style code for dots is here for at least 2 years and i completely forgot how it feels to write all logic in self defined jobs. So Entities.ForEach is something default for me now)

[karmic basin]

Yes

Wait no

I don't remember sorry for the ping xD

But I think it changed name yeah at some point

[sage mulch]

all good haha, appreciate the response

i'll just keep trying

[karmic basin]

I don't have it in 2020.2.5

Oh but you're not interested in GPU instanced, you want to override property, right ?

Then yeah check that box and follow the manual

And double-check if you're in v1 or v2

[karmic basin]

It's called "override property declaration" on my version

[sage mulch]

Thanks! I went ahead and enabled v2, enabled "Override Property Declaration -> Hybrid Per Instance" on my ShaderGraph's Property. I then wrote an IComponentData struct and System to interact with that struct per instructions. I then attached the AuthoringComponent of the struct to my GameObject and dropped a ConvertToEntity on it. The System seems to be running correctly (per the Entity Debugger) but the property doesn't change on the shader. Any thoughts on this?

(attaching some screenshots)

Also there has been no console output regarding a missing IComponentData struct (which the instructions says would happen) so I think it is being recognized, just not changed? nvm I think I'm wrong about this part, don't believe there would be any output

[timber ivy]

can someone explain the authoring workflow to me please?

the documentation is kinda vague and either it j ust magically works on its own or they are leaving some shit out

[slim nebula]

you put it in a subscene and everything in the subscene magically gets converted to entities

or you put a convert to component on it, and that converts it

[timber ivy]

what about [GenerateAuthoringComponent]

does that just generate a monobehavior you can use on prefabs

[slim nebula]

GenerateAuthoringComponent creates a monobehavior with the same fields as the component and adds the component to the converted entity from the fields

sometimes the generation doesn't work well and you need to write authoring manually but most times you dont

[timber ivy]

any benefits of using authoring over createarchetype if you dont need the editor serialization to edit the fields values?

[slim nebula]

the idea of authoring is converting developer editable/readable data into a format that's more efficient for the computer to process.

if the efficient way for the computer to process the data is just inherently readable then no you don't really need authoring

the idea is to allow fast development and fast runtime simultaneously by adding a small amount of boilerplate code

[timber ivy]

Thank you for all the info simoyd

appreciate you

[slim nebula]

[ocean tundra]

Is there any way to 'ask' if a entity has other components?
Eg If (EntityManager.HasOtherComponents<Component1, Component2, Component3>(entity)?

maybe something todo with archetype?

[slim nebula]

just .hascomponent for each component I think

pass in the entity that has the one component

[ocean tundra]

the issue is i'm building a tool/library, so i wont know about all the possiable user components

i only know about my components

[slim nebula]

hascomponent<YourComponent>(entity) will return true if YourComponent is on entity regardless if there are other components on entity

[ocean tundra]

yup, but i need to know if there are other user components on it

[slim nebula]

you want to iterate through components of unknown type?

[ocean tundra]

yea i guess

[slim nebula]

I think you'd need the architype for that yea

uhhhhhhh lemme look I think I have a vidja that had it

[ocean tundra]

all good

ill get digging into the archetype

can we use them in jobs?

[slim nebula]

uhhh I forget if you can use them directly or you might need to create a copied array of the data

[ocean tundra]

hmm

ill get digging in a bit

not urgently needed

[slim nebula]

yeah nope the code I remembered still requires the type to be specified. I dunno

why do you need to know if other components exist? Seems like some paradigm is being broken here

typically you'd be querrying data that you know exists

[ocean tundra]

ok so im working on networking stuff 😛
one of the things users will need todo is send commands (examples include, move unit command, build building, sync time)
some of those can be added directly to the player entity (those are nice and easy)
some will be added to entities a player control (eg unit move)
but some might be pure throwaway (eg chat, create new entity, attach chat command, attach target players)

the throwaway ones (event entities) should be cleaned up by my systems, BUT the commands attached to player entities (eg units) should not

[slim nebula]

you could make your components system components. then when an entity is deleted, the other components go away and only your system components are present. you could add one additional normal component that gets deleted that you can detect WithNone<>(). then you delete your throwaway components entities, and the other entities continue working fine

it sounds like a destructor pattern. system components do that

[ocean tundra]

hmmm maybe

[north bay]

Are you only looking for a way to clean them up?

[ocean tundra]

yea for now

ill try the system components

but if that dosnt work looks like EntityArchetype has a calculate difference method that might work too

ty for the idea

havnt really used many system components yet

[slim nebula]

I've been using them a bunch in pairs like that to handle clean up.

each system can have a pair of tag components to handle cleaning up instances of random junk, tracked by an entitiy

I have some animator instance stuff and some room instance stuff. I don't even known what all components are on there anymore. everything only worries about what it itself touches

[ocean tundra]

yea thats the best way

[karmic basin]

I only made a quick test in the past with the color, and it worked. Not all types are supported yet. SO I don't know for FLoat

[acoustic spire]

Interesting finding
I replaced DynamicBuffers with components containing UnsafeLists (of size 4096) in my Minecraft clone and actually got performance boost.
I moved everything - blockTypes, vertices, indices, uvs

Storing blocks in DynamicBuffer:
 - Generating :
     total    9.7354 ms in 100 iterations
     avg    0.0973 ms
     max    3.6375 ms
     min    0.046 ms
 - Meshing :
     total    7710.938 ms in 100 iterations
     avg    77.1093 ms
     max    581.9719 ms
     min    50.339 ms
Storing blocks in UnsafeList:
 - Generating :
     total    8.455 ms in 100 iterations
     avg    0.0845 ms
     max    3.3248 ms
     min    0.0388 ms
 - Meshing :
     total    2053.0682 ms in 100 iterations
     avg    20.5306 ms
     max    82.3683 ms
     min    10.5827 ms

[ocean tundra]

with the dynamic buffers you were testing in a build?

well with both tests were in a build?

[acoustic spire]

No. Why I should?

[ocean tundra]

there are a TON of safety checks everywhere

but in builds they are all gone

[karmic basin]

@sage mulch they have samples scenes, you could try to compare with your setup, check everything is the same

[acoustic spire]

Ah. They can be turned off in editor. One moment

[slim nebula]

there's still stuff you can't turn off. build is much faster than editor

[ocean tundra]

yea true, but i wouldnt be surprised if theres still some unity safety checks somewhere that you cant disable in editor

[acoustic spire]

Is it possible to build Play mode tests too?

[north bay]

Huh, I haven't seen any checks that aren't guarded by enable_unity_collections_checks on the C# side so far

[ocean tundra]

im pretty sure theres some around burst

as in builds a exception inside burst code kills your app right? in editor it dosnt

[slim nebula]

i disabled every check I could find in the menus and it's still much slower than my standalone build

[north bay]

I thought that's also handled by enable_unity_collections_checks but it's been a long time since I disabled safety checks inside the editor

[ocean tundra]

not sure

[acoustic spire]

so, without safe checks and jobs leak detection

Storing blocks in DynamicBuffer:
 - Generating :
     total    10.6013 ms in 100 iterations
     avg    0.106 ms
     max    5.152 ms
     min    0.0386 ms
 - Meshing :
     total    3466.3829 ms in 100 iterations
     avg    34.6638 ms
     max    176.7178 ms
     min    19.2338 ms
Storing blocks in UnsafeList:
 - Generating :
     total    8.4996 ms in 100 iterations
     avg    0.0849 ms
     max    3.4231 ms
     min    0.0374 ms
 - Meshing :
     total    1971.4737 ms in 100 iterations
     avg    19.7147 ms
     max    104.2451 ms
     min    10.2631 ms

[ocean tundra]

My thoughts are for any performance stuff profile in builds, editor just adds too much stuff 😛

[north bay]

True, but the editor should be good enough for estimations

[slim nebula]

i mean. does that generating and meshing process change the architype of the entity that the buffer is on?

like a dynamic buffer will have it's initialization size in the chunk data

whereas your unsafe list would always exist outside of the chunk data right?

I think dynamic buffer tries to balance between those two extremes: 1) second memory allocation outside of chunk vs 2) limited size in chunk

whereas you know the exact requirements of your list

so I feel like maybe it makes some sense

[ocean tundra]

theres a attribute you put on dynamic buffers to tell unity where to put your data

[acoustic spire]

It'll take some for me to make this project be built but I got interested in that too. There is the bigger problem with UnsafeList that makes it unusable - I can't assign its data to Mesh since it has no API for pointer nor converting to nativearray

[ocean tundra]

InternalBufferCapacityAttribute

[north bay]

You can create a NativeArray from an unsafe buffer

[acoustic spire]

any reference? I didn't find that

[north bay]

Let me look it up really quick

[ocean tundra]

    chunk along with other component types in the same archetype. When the number of elements in the buffer exceeds
 this limit, the entire buffer is moved outside the chunk.```

so a fixed size unsafe list is probably all inside the chunk right?

[north bay]

NativeArrayUnsafeUtility.ConvertExistingDataToNativeArray<T>() inside Unity.Collections.LowLevel.Unsafe

[acoustic spire]

Thanks

[timber ivy]

so you guys think dots vehicle physics is even remotely possible yet?

[slim nebula]

i mean all the math operations are available. What else would it need? project tiny showcases a racing game.

[timber ivy]

Well there are no wheel colliders in the physics api yet

how would I even begin to replicate that?

[ocean tundra]

dont lots of people just custom make those anyway

[timber ivy]

no idea

[ocean tundra]

what is there to a wheel collider? just keeping the main body up off the ground (suspension) and then they are the spots to apply forces from (which you can then use some math to work out things like grip)

[slim nebula]

that's very game specific

[ocean tundra]

oh and i guess turning, not really sure how to do that but again i expect thats just maths

[timber ivy]

prob pretty complicated maths tbh

[ocean tundra]

turns out you shouldnt be turning the wheel colliders anyway

[timber ivy]

you rotate them when you turn no?

also

[ocean tundra]

so i guess its just about applying some rotational force

[timber ivy]

https://github.com/unity-car-tutorials/Unity5-WheelColliderSource/blob/master/Assets/GTWheelColliderSource/Scripts/WheelColliderSource/WheelColliderSource.cs

ima look into this

idk anytime i've made cars i just rotated the front wheels to turn

[ocean tundra]

You do not turn or roll WheelCollider objects to control the car

right in the docs 😛

[timber ivy]

hmm

[ocean tundra]

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

[timber ivy]

it works if you do it tho

LOL

or rather

worked in early unity 5 days

ig that ws quiet some time ago

[ocean tundra]

😛

[timber ivy]

oh rofl

thats right

there is a steerangle property

on wheel collider

so the docs are correct

can an entity have multiple of the same component on it?

[ocean tundra]

no it cant

can be a buffer component tho

[timber ivy]

yee

[ocean tundra]

are you still talking about wheels?

i would probably make them separate entities

[timber ivy]

really?

[ocean tundra]

yea

[timber ivy]

4 entities per vehicle wouldnt be an issue?

[ocean tundra]

i dont think so

they wouldnt be colliders, i would raycast down from each wheel, do some maths and assign the result into a component

then the car systems update after, read the result from every wheel, do something 😛

also if each wheel is a child entity its SUPER easy to have more or less of them

eg a bike or making a massive truck with 20 wheels

[timber ivy]

So

Do I have a wheel component as well as a wheel entity?

will the authoring workflow take care of this for me?

[ocean tundra]

of course

how else would you query for the wheels?

[light mason]

wondering if someone could help me with a question.

Is there anyway to know the entity count inside a lanbda

kind the equivalent of creating an EntityQuery and CalculateEntityCount per frame

basically i only want to run a Entities.ForEach if i have an entity count over 1

[ocean tundra]

@light mason ecs will already limit your On update to needing more then 1 entity

But you can do that calculate entity could and wrap the entities.foreach will a if

[light mason]

ok good to know

thank you @ocean tundra

[karmic basin]

@timber ivy Check the physics samples from unity

They made some vehicles work

like a truck

https://github.com/Unity-Technologies/EntityComponentSystemSamples/blob/master/UnityPhysicsSamples/Documentation/images/vehicles.png

I think they tried a tank also if I remember well one of their Unite videos 1 or 2 years ago

[gusty comet]

i have question, its possible to use dots in Mirror/UNet networking? i mean if i have networkbehaviovur and i will convert that thing to entity and idk if that thing will break some things

[gusty comet]

or someone knows some networking which supports DOTS already

[rare dew]

there's DOTS Networking

https://youtu.be/P_-FoJuaYOI

[rare dew]

@gusty comet ^

[gusty comet]

its worth using that?

or its not stable af

[deft stump]

it's still unstable

just like the rest of dots

[karmic basin]

Use World.DefaultGameObjectInjectionWorld to access the default world. @gusty comet

https://docs.unity3d.com/Packages/com.unity.entities@latest

[karmic basin]

-_- I'm not your IDE dude 😉

you can find scripting API here https://docs.unity3d.com/Packages/com.unity.entities@0.17/api/Unity.Entities.GameObjectConversionUtility.ConvertGameObjectHierarchy.html#Unity_Entities_GameObjectConversionUtility_ConvertGameObjectHierarchy_UnityEngine_GameObject_Unity_Entities_GameObjectConversionSettings_

[gusty comet]

i mean old input values had Gameobject, World

[karmic basin]

now it's GameObject, GameObjectConversionSettings

Have a read on the manual to refresh your knowledge ?

[gusty comet]

what about blob asset store

[karmic basin]

• this : https://docs.unity3d.com/Packages/com.unity.entities@0.17/manual/conversion.html

[gusty comet]

also, any methods made in monobehaviovur should use ECS?

[karmic basin]

No ? But I'm not really sure I get where you're going with that question

[gusty comet]

i mean like if method is executed every update

nvm

[karmic basin]

Yeah I don't get it

Systems have their own SystemBase.onUpdate() method. That's something else than MB.Update()

If that was the question 🤷‍♂️

[gusty comet]

i speaking about something like, its 1 method and that thing is doing for example, get all entity and change value or something, then that thing should be done in ecs than just normal monobehaviovur?

[karmic basin]

Oh, you can access Entities package from MBs. But better do it from ECS systems

let's say you want to draw debug gizmos on entities transform, yeah you could plug a quick dirty code on a MB OnDrawGizmos, query entities from here to access their position and draw your gizmos.

But if you have entities, you want to go all in and process their components through systems

Otherwise what would be the point of entities in the first place ?

Mainly you would work on entities from MBs for the conversion system

or for authoring them easily from the editor before that

authoring their components* I mean

Hope that answers your question

Simplest way to iterate through entities and change value would be with a ForEach query first.

https://docs.unity3d.com/Packages/com.unity.entities@0.17/manual/ecs_entities_foreach.html

So again, read that manual before going further

If you're still lost, check the Unity official github repo, they have an "ECSsamples" that showcases main concepts gradually

[merry fossil]

hey y'all what's the current state of the DOTS visual scripting?
i heard it was abandoned in favor of bolt? hopefully not?

[karmic basin]

There's a forum post on that matter

Officially it's "still worked on" but they're gonna stop pushing updates and releases to us

[merry fossil]

i just looked at literally 5 forum threads and did not find that answer straightforwardly

[merry fossil]

that's just sad. wtf

[karmic basin]

Yeah and maybe a blog post for the future of visual tools

the idea is they want to merge ALL their visual graph tools to look the same (and have the same backend I guess)

I think they gonna keep both for some time, eventually merging at some point

But no real precise roadmap I'm aware of

[merry fossil]

jesus christ. so they're pulling an "okay we don't want to abandon or release anything, really, so we'll just put out 3 different versions" again?

like with the 3 different UI systems we now have

[karmic basin]

What killed it for me, is they stopped codegen since a few drops :/

[zenith wyvern]

There won't be any more dots specific vs solution afaik

Right now they're focused on bolt, eventually that solution will have dots support

[merry fossil]

for visual scripting or for DOTS in general?

[karmic basin]

codegen FROM VS, sorry

[gusty comet]

visual scripting can make dots easily or what?

but from comments about bolt its just not too good for performance

[karmic basin]

Well next Bolt version planned before Unity acquired them was improving performance and close to release, but they cancelled that to integrate it with Unity

People are angry about that :/

[karmic basin]

Use case I intended at first was, build systems from visual scripting, then codegen, so it writes all the boilerplate code for me while I learn it.

[merry fossil]

i wish they would just open source their engine and let people extend themselves. we would have such benefits 😦

[gusty comet]

ohh no no

imagine open source

[merry fossil]

i do imagine open source, literally every day as a programmer. 99% of what i use is based on open source software (and so does everyone else)

[gusty comet]

not always open source means it will be good

[merry fossil]

of course there are "bad" open source projects with bad leadership. but there are extremely good benefits to open source. also: unreal literally did just that and it made their engine wayyyy more attractive to many devs

[gusty comet]

GhostAuthoringComponent got removed from netcode or what?

[vagrant surge]

the original visual script, before-canning version, codegened into burst C# systems

now they have two blueprint ripoffs

[gusty comet]

idk, in docs i see that thing exists

[deft stump]

go?

GameObject?

Why?

[gusty comet]

GhostAuthoringComponent

[acoustic spire]

Should these 2 blocks of code be equivalent? In both cases it will wait for all jobs completed, right?

// 1
_handle = Entities.ForEach((int entityInQueryIndex, Entity e) => {
// ...
}).ScheduleParallel(Dependency);
_handle.Complete();
// 2
Entities.ForEach((int entityInQueryIndex, Entity e) => {
// ...
}).ScheduleParallel()

Works differently for me

[north bay]

Those are not equivalent. The first says Schedule this and right after stall the main thread till it's completed.
While the second says Schedule this but let the main thread do other stuff.

[acoustic spire]

But I call Complete so it should block until jobs in foreach finish

ah, one min

So, I thought the 2nd actually wait for completion

[acoustic spire]

Can't get it. I'm trying to skip scheduling new jobs if there are still active jobs from previous update

void OnUpdate() {
  // Skip this update if we have active running jobs from previous update
  if (!_handle.IsCompleted) return;
  _handle.Complete();
  _handle = Entities.ForEach(...).ScheduleParallel(Dependency);
}

But getting error like it doesn't skip scheduling

The previously scheduled job MeshingSystem:<>c__DisplayClass_BuildingMeshData writes to the BufferTypeHandle<Game.ChunkMeshVertices>

[sage mulch]

this was good advice, thank you. I went ahead and checked out all of their sample scenes and I suppose something may be wrong with my setup. their examples with color worked, but not when I do the exact same setup in my scene haha. kinda odd

From some more experiments it may be because I'm using trying to override the color of a SpriteRenderer. Maybe this isn't supported yet

[safe lintel]

dont think spriterenderer is supported by the hybridrenderer yet

[zenith wyvern]

Combine _handle with Dependency before you schedule

[sage mulch]

That would explain it! thanks

[acoustic spire]

The problem was that I forgot that I accessed dynamicbuffers later in the system. I was using ref in these jobs and request them though in later

[karmic basin]

@safe lintel @sage mulch Sprite renderer is supported as a hybrid component, but I don't know if that support extends to overriding its properties 🤔

[safe lintel]

If its a hybrid component its just gameobject land, hybrid renderer wont touch it. afaik material overrides arent part of either renderpipeline yet

[pliant pike]

is there a way of checking if a singleton buffer exists 🤔

[dark cypress]

What are best practices for RNG?

I've seen a blog that suggested using a singleton array with 1k random instances, and throw them into parallel jobs based on thread index, but that feels weird.
Another approach is to dump random instances into components. And since IIRC it's just an int it shouldn't take much.

[ocean tundra]

@pliant pike HasComponent works for buffers i think

[pliant pike]

ok, yeah that means I have to first get the entity, I guess that's the only way

thanks

[timber ivy]

Would anyone mind taking a quick look at my github and see if im making any huge no nos with dots?

Also what is the steps to create a package out of your code?

[ocean tundra]

@timber ivy Did you find any steps/guides?

[gusty comet]

ECS already have animations/physics systems?

[ocean tundra]

check out the pinned message

yes to both

physics is in agood state

animations still very early

[gusty comet]

while looking at pure ecs soo, ECS means every thing should use entity system?

like static object in game

or dynamic

[ocean tundra]

yes to both

they are both 'pure' ecs/entities

[gusty comet]

soo if i want ECS then everything should use ECS?

[ocean tundra]

yes

you dont have too tho

theres hybrid options

[gusty comet]

yes i see, changing gameobject to entity and using still mb

[ocean tundra]

check out the docs in entities, conversion workflow, hybrid

[gusty comet]

i think while i have rooms in my game and if i tried doing pure ecs then uhm idk if that will work with multiple textures/meshes

[ocean tundra]

it will

thats the hybrid renderer