archived-dots

[rustic rain]

have you used LocalToWorld or any other transforms types?

they have helper methods

[late mural]

but my logic needs to be used on all systems (even if that is only 2 currently...)

[rustic rain]

of all sorts

[rustic rain]

then your architecture is wrong

and you need to figure the other way

[late mural]

they do? i always use local to world and i never knew they had helper methods, how fascinating!

[rustic rain]

take a look at TransformAspect

in manual

it has tons of helpers

[late mural]

then how should i do it, without needing to copy and paste everything like sprinkles, im really lost and confused

i need to have dinner now, ill be back soon though, sorry, and thanks so far for everything

[rustic rain]

or you can store data in a way that is obvious and simple enough

[late mural]

ok i've had dinner and i have thought really hard about this, and i think i get it now, and i think i found a better way than what i was doing previously, thanks so much for all the info, im now going to remove all those static methods, chuck the ones that are actually used (and are a replacement for more than 1 or 2 lines of code) in whatever system uses them most, should another system ever need it then i can just do it manually and not use a method at all! Hopefully that makes some sense, im tired lol, thanks so much for all the info!

[devout prairie]

anyone messed with joint motors at all?
posted a question here - https://forum.unity.com/threads/joint-motor-more-than-one-axis.1379079/
tldr - trying to get motors to work on more than just the x axis

[quasi kernel]

Hi! I'm just getting started with DOTS. I have set up an empty 2D URP project (2022.2.1f1) and it appears I'm already running into an issue. I added the Entity Graphics package (1.0.0-pre.15) to get the required dependencies. However, it appears that transport and 2D packages cause several issues which seem to be related to an old collection library, I believe:

Library\PackageCache\com.unity.2d.animation@9.0.1\Runtime\TransformAccessJob.cs(165,62): error CS1955: Non-invocable member 'NativeHashMap<int, TransformAccessJob.TransformData>.Count' cannot be used like a method.
Library\PackageCache\com.unity.transport@1.3.0\Runtime\IPCManager.cs(79,44): error CS1955: Non-invocable member 'NativeHashMap<ushort, int>.Count' cannot be used like a method.

From my short search, it appears that the NativeHashMap was renamed to NativeParallelHashMap quite some time ago.

Is this a known limitation, i.e. is 2D not supported currently or am I looking at this wrongly?
Thanks!

[balmy thistle]

What version of collections are you using? And how are you including the transport dependency?

[robust scaffold]

the 2d animation package doesnt work with dots, remove that and everything should be fine

[tribal pollen]

So I'm spawning items, and realized I'm setting a new UniformScaleTransform when I spawn them and overwrite the desired scale/rotation from my prefab.
What's the normal way to handle this and only update the position?
Or do I have to get the translate component seperately from the prefab with a componentLookup?

[robust scaffold]

And yea, 2D isnt supported with DOTS in the first place

[quasi kernel]

Thanks for taking the time! Oh well, that settles it then. I was thinking about prototyping a tile-based game with some interacting systems (think Rimworld or similar). Thought that could benefit from this approach. Just thought it'd be a good exercise to get my hands dirty :)... But I get that for most 2D games, this would probably be overengineered. Maybe a GO-based architecture works well enough.

[rustic rain]

oh, Rimworld would benefit a lot from dots

and since RW barely has any animations or graphics, you don't even need neither tile map or skeleton animation to make a clone

[quasi kernel]

Yep, that's what I thought... But does "no 2D support" mean it doesn't render sprites as well? Because I suppose that'd be pretty much the only thing that's needed.

[rustic rain]

there are no 2D tools at all

you use 3d as 2d basically

with z as depth

and there is no 2d rendering, so under the hood it uses hybrid with GO sprite renderers

besides, you can use tilemap, but only in non-interactive way

no physics or such

[quasi kernel]

I think that'd be fine. Maybe I'll have a look. Thanks for the hints! I'm not married to the idea or anything 😄 ... I just want to avoid working against the "intended way of doing things" when I'm still learning.

[robust scaffold]

I've just placed a cube collider on every wall tile. It works decently enough. The stateless physics engine requiring complete physics state regeneration every frame is a bit painful though

[quasi kernel]

So it's good to know the limitations. Maybe I missed it and it's mentioned somewhere in the docs. If not, it'd be nice to add a paragraph to the Entities Graphics documentations.

[tribal pollen]

I have it working with component lookup, but it feels bad to do the lookup, so I could make a new component for it. Just feels bad there isn’t a cleaner way to utilize some of the prefabs transform values, while modifying them

[rustic rain]

you can use static colliders

[robust scaffold]

Static colliders are still regenerated every frame.

[rustic rain]

what's the difference, between static world regeneration once in a while vs dynamic colliders every tick?

[rustic rain]

are you sure?

[robust scaffold]

The BVH isn't rebuilt but the world state still is.

Static colliders only optimize skipping BVH reconstruction. There's an initial world state building that must be done every frame. Not that expensive thankfully, 1,000 colliders only result in 0.05ms parallel.

but that 0.05 adds up with netcode requiring 32+ physics steps for rollback.

[lethal hatch]

I'm trying to implement a nativearray/nativelist of generic structs that calculate faces for my minecraft clone.
The struct implements an interface that gives details on what exactly the mesh should look like (like vertex count, etc), so that it is easier to add new or different shapes other than quads like spikes or flowers in the future. however it seems that unity jobs doesnt like me doing this as it wants me to use a non-nullable value type. is there a good alternative or workaround to this?

ideally, I should be able to have one nativearray or list for a whole grid of differently shaped faces

i suppose i could just store an int or enum which tells the mesh generator what type of mesh it is maybe? but that also makes it harder if I need to do any precalculations to a type of shape before it is sent in to be generated

[robust scaffold]

You can either use a union struct using [StructLayout(Explicit)] or some time of value reinterpretation. Or multiple arrays.

[robust scaffold]

Check out how unity physics does with collider reinterpretation. Colliders are all 1 array but the type of collider and the data within is different. They use a header struct with the type of collider as an enum and reinterprets the data within using that enum.

[lethal hatch]

yeah doing that or similar seems best for my scenario. union structs sound like a headache anyway lmao

[robust scaffold]

It's not too bad. Just keep track of the field offsets and you should be good. Especially if you use structs instead of manually computing the offsets of each field within your array data.

[misty wedge]

unionizing the values for faster netcode serialization is sneaky

Although couldn't you just make it a long at that point? I've never checked how netcode actually serializes everything

[robust scaffold]

The color may change and the properties not. This is equivalent to "composite = true" on the ghost field but with 1 value serialized instead of multiple.

But yea, a ulong would work as well... hrm

[misty wedge]

I thought netcode only serializes the entire component and checks the hashes to check for changes?

Or is it per field

[robust scaffold]

Yea, but when it serializes bytes, it converts all the byte values within Color32 into a uint4.

Which I assume wastes a lot of bits. Depending on how the bitpacking works. Probably doesnt do anything honestly

[misty wedge]

I just thought it would be (very slightly) faster to serialize a single long than 2 uint, but it probably makes no difference

(and like I mentioned I never actually checked how it does it so I might be completely wrong)

[robust scaffold]

same here

[misty wedge]

Yeah pretty much, no idea if it's faster or slower though 😅

[safe lintel]

anyone ever had the Allow unsafe code setting not work? annoying, ive deleted the library but it still thinks its not ticked

[rustic rain]

did you tick it in ProjectSettings though?

[safe lintel]

yeah

[rustic rain]

then I guess just try to restart

[robust scaffold]

make sure the asmref for the script also has it

[safe lintel]

ah wait it was cos it was in the package not the general project.

yeah that was it 🙂

painful to delete library on an older surface pro

[robust scaffold]

Ah, how do you resolve player-player collisions then?

[rotund token]

Why doesn't it just work?

From the players perspective how is the collision different to any other interpolated ghost

[robust scaffold]

Predicted - interpolated collisions (using a owner predicted two sphere collision) results in client-server divergent simulations.

Because on the client, interpolated objects are enforced kinematic, so predicted objects will bounce off them like walls. But on the server, everything is dynamic since there is no such thing as interpolated serverside physics. So objects collide and push normally.

[rotund token]

That's kind of the nature of prediction

[robust scaffold]

I've made everything predicted which solved the issue of collisions but remote player controlled objects move very poorly.

[rotund token]

Your player is not a kinematic?

[robust scaffold]

No? It's a dynamic object

[rotund token]

Most controllers still use kinematics and execute the dynamics itself

[robust scaffold]

I would really hate to implement a full 3D physics stack just to move a single player.

Because, from what I can tell, kinematics dont interact with static objects.

[rotund token]

Controllers often have a dedicated developer or even small team working on it for the entire game development

It's a lot of work

[robust scaffold]

ugh

[rotund token]

Sounds like you should use rival 😄

[robust scaffold]

im looking at it now, im gonna see if i can clamp it to 2d without any particular hacks

[rotund token]

alternatively check out the netcode car demo from unity

since those cars are heavily physics based objects

actually yeah they use dynamic physic object cars on owner predicted

so that seems like a good source of an example of what you wanted to do originally

[devout prairie]

Fwiw, this is the crash i've been getting since switching to 2022.2.1f1 and dots 1.0

[robust scaffold]

I dont know if it's just me but playing the car demo doesnt work for me. But I have been digging through it as reference. That's where I pulled predicted smoothing from which slightly helped the issue.

[rotund token]

hmm worked fine for me

what was issue?

[robust scaffold]

Do cars collides with each other?

Give me a sec, pulling up the repo

[rotund token]

oh i never tried multiplayer

i would have hoped so

because that'd be the whole fun of the demo

[robust scaffold]

out of memory?

[rotund token]

allocating 1TB of memory will do that

[devout prairie]

yeah.. basically it seems fine when i'm actively working on the project - when i walk away from the pc for any length of time i come back to that

seems like probably a memory leak

[rotund token]

hmm not sure

we had a similar bug a few years ago

just randomly try allocate like 40GB of memory

was basically just a bad write on a pointer

[devout prairie]

hmm

[rotund token]

not saying this is you

[devout prairie]

my only real suspect is baking tbh

[rotund token]

graphics has the most issues of the packages for me atm

[devout prairie]

or potentially one of the assets ( i use a fog asset that does graphics related stuff obviously )

[robust scaffold]

Failed to load 'C:/Users/CornFlakes/Documents/ECS-Network-Racing-Sample-master/Assets/Terrain/Terrain_0_0_89adce58-f5ec-4f5a-803b-a54fad3b41f3.asset'. File may be corrupted or was serialized with a newer version of Unity. That's error when I try to play the car sample. And I'm on 2022.2.1F1

[devout prairie]

i haven't really tried to get to the bottom of it yet tbh, i'll maybe try removing that asset from the scene etc.

[rotund token]

odd. what version of unity?

[devout prairie]

there was a breaking bug related to the car project that was mentioned recently, that they have or are going to update with a fix

[rotund token]

yeah physics baking (unless there was something else I missed as well)

but that looks like just a terrain asset?

[robust scaffold]

Yea, just a broken asset I think

[rotund token]

but yeah no issue

just loaded it up then

[robust scaffold]

Hrm, it's 1kb size. I think something just overwrote it. Gonna grab a copy from github

[rotund token]

ah did you pull/clone the repo or just download it rom github?

you have to pull with LFS

[robust scaffold]

Yea, just realized that. Doing that now

huh, I thought they fixed this physics baking problem?

[rotund token]

look at this post

https://forum.unity.com/threads/exception-with-baking-when-upgraded-to-pre-15-physics-version.1373157/

this poor guy, reported it like 2 weeks ago

"update editor, reimport" etc etc

i load up racing sample, get it first thing, report it

they can repo it instantly

but i don't understand, how whoever was developing the racing demo didn't report it -_-

it reminds me of my work, people see a red error, 'clear' - "Someone else must have reported it already"

[robust scaffold]

It's unity, it always throws errors.

[rotund token]

anyway i loaded up rival, it does not seem to use dynamic controllers (seem to be static) which is what I suspected

[robust scaffold]

I thought the backport fix already came through because I can use physics smoothing on my main project with no issue.

ugh, i have to go through all the prefabs to try and find physics bodies with interpolation on just to fix this error

There we go

hrm, doesnt look thin clients work

Yea, collisions are really bad in this

[robust scaffold]

Wow, so many shaders

[misty wedge]

Is Unsafe.NullRef burst compatible?

[rotund token]

i don't even know what that is

[misty wedge]

It returns ref that is "equal" to null

And you can use Unsafe.IsNullRef to check

[rotund token]

from what library

[misty wedge]

Unsafe

it's part of C#

Runtime.CompilerServices

But it's not part of what unity uses

I think you can build something similar with the methods in UnsafeUtility though

[rotund token]

it seems internal

[rotund token]

to the c# library?

    [NonVersionable]
    [MethodImpl((MethodImplOptions) 256)]
    public static bool IsNullRef<T>(ref T source) => ref source == IntPtr.Zero;```

hmm

what am i missing

[misty wedge]

Weird

[rotund token]

wait

that seems to be in .net 7

[rotund token]

oh wait no, .net 5+

[misty wedge]

In a "normal" C# project it's not internal

[rotund token]

regardless i guess question is

why do you need it? but an null check is fine in burst

(but yeah, i thought unity didn't support .net 5 yet so I'm not sure how you're using it 🤷‍♂️)

(or am I mixing up all the confusing versions again)

[misty wedge]

I'm trying to write a TryGetValue that returns by ref

I'm using a BlobHashMap library someone wrote and their methods don't return by ref, which is an issue for me

What happens if you do ref (*(T*) ptr) where ptr == null?

I'm guessing it's a bad idea to use the value, but in theory getting the address of it should also resolve to null, right?

[drowsy pagoda]

Is there a built-in way to disable an entity? (like collider and mesh) in 1.0?

[rotund token]

oh does that library finally work in 1.0?

[misty wedge]

It didn't work before?

[rotund token]

unity broke it in 0.16 from an unneeded safety check

[misty wedge]

Ah yeah, I removed that

😅

(who needs safety amirite)

The blob safety system is so weird, it just doesn't seem to catch a lot of stuff

[rotund token]

think i've done it like this before
ref var r = ref UnsafeUtility.AsRef<int>(null);

[misty wedge]

Yeah that was basically my idea too

Does getting the address of the ref resolve back to null though?

[rotund token]

only 1 way to find out

[misty wedge]

Seems to work

returns true 🤷‍♀️

Using the value even throws a nice NullReferenceException, so also seems pretty "safe" to use

[late mural]

if only 1 entity has a specific component does it automatically get treated like a singleton component, or do i need to specifically mark it somehow?

[rotund token]

automatic

a singleton in entities is simply a unique (set of) component(s)

[late mural]

ok cool, thanks a ton!

[tribal pollen]

How is dots performance on mobile?

[void girder]

Is there a way of updating shared components without forcing a sync point or using an ecb?

[misty wedge]

Nope

By definition they need a sync point for their value to be changed

[void girder]

I thought that since they can be unmanaged now that restriction would no longer exist

[coarse turtle]

for a MultiValuedHashMap, is a value enumerator deterministic/consistent? 🤔
E.g - if I store multiple values for a key and grab a value enumerator, will iterating along the value enumerator have the same order between multiple runs? Or should I not bet on this and actually store a 'sort key' so I can iterate things in the order that they're stored into the MutliValuedHashMap?

[misty wedge]

Normally I think the order for set type containers is undefined

[balmy thistle]

They require a sync point because changing a shared component value potentially repartitions entities and changes chunk layout

[coarse turtle]

alright - yea I think I'll store a sort key for now so I can iterate things in the order that they're stored

[misty wedge]

You can use them inside burst methods now that they are unmanaged

The reason they require a sync point isn't because they weren't unmanaged before though

[void girder]

Right I forgot to mention that I only change scd using entity queries (with scd filter) so it's more of a 'swap', although I guess that still risks a structural change

[misty wedge]

That doesn't just risk a structural change, it is one

[void girder]

Does it force data to be moved?

[misty wedge]

If the SCD value is changed, yes

[void girder]

Even if all of the entities sharing that scd has it changed all together? Ideally no extra work would be done

[misty wedge]

I don't think this is checked for, I might be wrong though

Pretty sure it's always a structural change

[void girder]

Yeah probably honestly

[misty wedge]

The idea is to just not change them that often

[devout prairie]

Been getting this wonderful error..

When i hit play with the subscene open it's fine

[misty wedge]

Everyone gets that error

[devout prairie]

Well it doesn't load the subscene in playmode when the subscene is closed

and it wasn't previously throwing this error 😐

[misty wedge]

It just breaks sometimes

The "quickest" way to "fix" it is recompile, then clear the entity cache

But it will break again at random times

[devout prairie]

yeah i couldn't find the clear entities cache option in 2022/ecs 1.0

[devout prairie]

ah yeah thanks

[rotund token]

Agents = SystemAPI.GetSingletonBuffer<NavMeshAgents>(true),
you request a nice little buffer readonly

but unity don't care, have write access anyway

[misty wedge]

tertle help, my function works with burst off and breaks with burst on 😬

[rotund token]

kekw

[misty wedge]

What even is this error message

System.IndexOutOfRangeException: Index {0} is out of range Length {1}

I guess the formatting broke

And it's not actually 0 and 1

[rotund token]

just clear subscene, restart unity, clear again

[rotund token]

yeah burst doesn't support exceptions

[misty wedge]

Great then how do I debug this

[devout prairie]

clear cache worked for now, thanks

[rotund token]

so that's just the format string

[misty wedge]

Time to print random stuff everywhere

[rotund token]

break point on top of the exception

[devout prairie]

i've had this randomly

[rotund token]

!

[misty wedge]

I thought you can't in burst functions?

[rotund token]

2022.2 and burst 1.8 supports break pointing

[misty wedge]

waht

[rotund token]

(by simply turning burst off automatically where you put a break point)

but yeah there is little reason to turn off burst

[rotund token]

it's deterministic if you don't use parallel writer

[misty wedge]

Weird, if I set it to break on IndexOutOfRangeException, it doesn't

Does that not work with burst?

[rotund token]

is it erroring?

[misty wedge]

Yeah

(I'm talking about the "Break on user exceptions" setting in rider)

[rotund token]

oh that doesn't do anything

you need to set a break point

[misty wedge]

If I set the break point where the exception is thrown, it doesn't get thrown anymore, I'm guessing because it turns off burst

great

[rotund token]

yeah

before you go crazy

have you tried clearing your burst cache

[misty wedge]

How do I do that?

[rotund token]

close unity

delete library/burstcache

restart unity

[misty wedge]

I thought the caching issues were fixed?

I'm on 1.8.2

[rotund token]

no

there is definitely an issue with burst caches i keep running into

(basically a crash or quitting unity while compiling just breaks everything for me atm)

[misty wedge]

Still throws the error :[

I'll just add my own checks and debug that way

[coarse turtle]

O cool. Then I can likely work with this

[misty wedge]

@rotund token I don't think ref pointers resolve back to null in burst

The highlighted line is where a NullReferenceException is being thrown

Kind of a bummer

Yeah, this doesn't log anything

It works outside of burst

Would be nice to add that to the burst docs, since it seems to be missing. Kind of a niche thing anyways

[rotund token]

does this just not mean your IsNullRef line doesn't work

because it's returned the null ref pointer fine

[misty wedge]

🤷‍♀️

[rotund token]

i think I used it something like
ref value TryGetValue(T key, out bool found)

or something

[misty wedge]

That doesn't work if that functions calls a function that also needs to return by ref

Which can be worked around, but it's still really annoying

[rotund token]

what do you mean?

it just changes your code to

ref var result = ref data TryGetFirstValue(key, out var isValid);
if (!isValid)
{
    value = default;
    return false;
}

value = result;
return true;```

[misty wedge]

if TryGetFirstValue uses a function that returns by ref, you will get a compile error

You can't have a function return by ref and have a ref or out parameter

[rotund token]

oh i see what you're saying

[misty wedge]

(and then call another function with the same properties)

It's not a huge issue, since I can just put everything in one function, but it's still kind of annoying

[rotund token]

wait it seems to work fine for me?

[misty wedge]

Let me check, I had issues with that when refactoring earlier

[rotund token]

            where T : unmanaged
        {
            result = true;

            return ref UnsafeUtility.ArrayElementAsRef<T>(array.GetUnsafePtr(), index);
        }```

this compiles fine

(at least in rider)

[misty wedge]

This gives a better explanation of why it's not allowed

[rotund token]

that's because of this line here

comment that out, set hasKey = true

[rotund token]

thats a completely different error

[misty wedge]

What

It's the same one

[misty wedge]

^

[rotund token]

these are not the same error

out _
is 'it'

[misty wedge]

I assume it doesn't show the actual issue, instead only telling you that it's not a valid thing to return by ref

That's why the error message is different

[rotund token]

            where T : unmanaged
        {
            ref var a = ref UnsafeUtility.ArrayElementAsRef<T>(array.GetUnsafePtr(), index);
            result = true;
            return ref a;
        }```
this works fine for me

i don't know what is up with your iterator

but just make a method that doesn't return it?

[misty wedge]

Yes, because you aren't calling a function with an out value in that function

If ArrayElementAsRef had a ref or out parameter, that wouldn't work

both parameters are passed by value

Alternatively call ElementAt in a function that also returns by ref

[rotund token]

i think you're overlooking a really easy solution

get rid of the out

you aren't using

make an overload without it

[misty wedge]

That's why ^

[misty wedge]

I was just trying to explain since ^

Returning null ref is still kind of awkward though since afaik you can't force a function to "receive" the return value as a ref

Overwriting a local value with the null ref value immediately throws a NullReferenceException

[rotund token]

you know it's weird

it's not a compiler error for me

just a warning

[misty wedge]

[misty wedge]

???

[rotund token]

???

[rotund token]

warning not an error

[misty wedge]

That's really odd

hacker

I'm calling the police

It's definitely a compile error for me

[rotund token]

it was a compile error when i did it a different layout

like this

anyway i don't care about, too weird just use pointers 😄

[misty wedge]

Yeah I thought about it, but It feels weird to use pointers for one thing in blobs, and ref for everything else

🤷‍♀️

[cunning perch]

someone call the ref police

[misty wedge]

I think I'm going insane

waht

is going on

[robust scaffold]

The double ref might be a problem.

[misty wedge]

I'm starting to think so too

Maybe burst can't pass a null reference or something

(as ref)

But then I don't see why AddressOf works

[robust scaffold]

If you pass a null reference as a ref property, it may be resolving the wrapper variable which is valid but the reference itself is null.

[misty wedge]

But then AddressOf would also not work

Is there an easy way to print the memory address of a pointer from inside of burst?

I only know the IntPtr string formatter, but that won't work in burst

It has to be pointing to something for it not to evaluate to null

[rotund token]

just do addressof and then get the address of the pointer, cast it to an int

and write it out

x4 if you want to format as hex

[misty wedge]

It's 0

[rotund token]

as expected

[misty wedge]

Yeah, odd that the null check doesn't work

[rotund token]

just compare it to IntPtr.Zero

            return new IntPtr(p) == IntPtr.Zero;```
something like that?

[misty wedge]

Somehow that still doesn't work

I give up

I'm gonna say it's because its 3:51 am

[misty wedge]

It works as long as it's not in the function, so I'm guessing something weird is going in burst when passing it by ref twice for some reason

Time to sleep 🛏️

[rotund token]

you're copying it locally maybe?

[misty wedge]

How? It's passed by ref

[rotund token]

oh wait no

[rotund token]

thought it was a different method

[misty wedge]

Funnily enough it works if I cast the pointer to an int and compare to 0

🤷‍♀️

[calm edge]

I think burst assumes references can't be null and optimises some code out

I've had problems with that in the past

safer to use pointers if it can be null

Is there any way to get entities to render in the scene view?

[rotund token]

they should by default

if they aren't, check your authoring settings

is set to runtime data

[calm edge]

That was it, ta

[late mural]

probably a silly question, but how do i remove a component from an entity?

[robust scaffold]

EntityManager.RemoveComponent<T>(Entity)

[late mural]

thanks a ton!

[safe lintel]

anyone know how ScheduleParallelByRef compares to ScheduleParallel? i see it used in the LocalToWorldSystem

[robust scaffold]

It's for when you have a lot of the same structs required as input into a job. I use it in performance testing to run the same job over and over again.

[safe lintel]

hmm, I guess ill use it then? any downsides if you dont have lots of same structs? is there any doc or forum post on it?

[robust scaffold]

I mean, I dont see anything wrong with just creating the job on create and then setting the fields of the job struct in update just before scheduling. Maybe there might be an issue with long running (multiple frame) jobs but the performance benefit might be minimal.

[rotund token]

i don't believe you could store store private fields anymore

for caching temp allocation etc

i generally find it, not worth the rare error

[late mural]

just checking that this is a good way to get read write for a component on a known entity? ```cs
state.GetComponentLookup<SomeComponent>().GetRefRW(SomeEntity, false);

[rustic rain]

SystemAPI.GetComponentRW

[late mural]

ooh, thanks so much!

[late mural]

how exactly do i use it, as it wants a systemhandle and not an entity?

[rustic rain]

oh yeah

that's a wrong one then

kek

well you can just do

GetComponent

and then SetComponent in the end

[late mural]

ooh good point, thanks for the info!

[rustic rain]

As alternative

        /// <summary>
        /// Provides an unsafe ref to component data directly. Gets invalidated on any structural change.
        /// </summary>
        /// <param name="em"></param>
        /// <param name="entity"></param>
        /// <typeparam name="T"></typeparam>
        /// <returns></returns>
        [GenerateTestsForBurstCompatibility]
        public static unsafe ref T GetComponentReference<T>(this EntityManager em, in Entity entity)
            where T : struct, IComponentData
        {
            var access = em.GetCheckedEntityDataAccess();
            var typeIndex = TypeManager.GetTypeIndex<T>();

            access->EntityComponentStore->AssertEntityHasComponent(entity, typeIndex);
            access->EntityComponentStore->AssertNotZeroSizedComponent(typeIndex);
            if (!access->IsInExclusiveTransaction)
                access->DependencyManager->CompleteWriteDependency(typeIndex);

            // var ptr = access->EntityComponentStore->GetComponentDataRawRW(entity, typeIndex);
            var ptr = access->EntityComponentStore->GetComponentDataWithTypeRW(entity,
                typeIndex,
                em.GlobalSystemVersion);

            return ref UnsafeUtility.AsRef<T>(ptr);
        }

but generally it's better to use SystemAPI

the code I sent is only good if you only have just 1 component of T type to look up for RW in OnUpdate cycle

[late mural]

ok fascinating, thanks for the info

[late mural]

wondering how would i set the name of an entity? (if entities even have names)

[rotund token]

EntityManager.SetName

[late mural]

thanks a ton!

[devout prairie]

any idea if any of the physics devs spoke during the Q&A?

[safe lintel]

I dont think i ever have used private fields and the use by the transform system just recreates every ref job in update(as I do) 🤷‍♂️

[drowsy pagoda]

Does anyone know how to add leaf colliders to a compound collider at runtime?

[rustic rain]

I suggest to bake them and then add already baked collider to entities you want

oh wait

compound...

yeah... nvm me

[drowsy pagoda]

Heck I'll even take how to make a compound collider from scratch during runtime at this point.

[rustic rain]

why is it even required?

sounds like overcomplicating things

[drowsy pagoda]

I have items that can be attached to each other, like assembling something like a chair or a computer, etc. So you can detach a part, and then eventually have compatible parts that can be attached to different assemblies.

[rustic rain]

I think you should look into using joints

[drowsy pagoda]

I did. And i got it to work perfectly.

But the issue was when an item falls to the ground that has fixed joints, it then jumps up wildly.

I tried everything, wrote a bunch of questions in this room. And no results.

[rustic rain]

🤔

[drowsy pagoda]

Now I'm back to combining colliders 😦

[rustic rain]

Ragdolls in sample don't jump wildly

[drowsy pagoda]

But I don't need ragdolls, I need just simple, fixed joints. Basically to replace the traditional Transform Parent Child system.

[rustic rain]

ragdolls are pretty much joints

[drowsy pagoda]

Pretty much.

[rustic rain]

I don't remember them having anything

have you tried to reproduce your bug in isolated environment?

[drowsy pagoda]

Yes, it's actually where the bug occured. I have an environment just for building this behavior.

[rustic rain]

you could report to unity

[safe lintel]

i think if a fixed joint is causing a physics thing to jump around wildly it should be considered a bug, but im not really sure on behaviour of stuff like how collider penetration when fixing two rigidbodies together should be expected with this

[rustic rain]

btw

do your joints have very fixed angle

or some slight movement is allowed?

maybe that could help?

[drowsy pagoda]

Oh yeah I tried that too, allowing some degree of movement, but that gave me other unwanted behavior.

[rustic rain]

oh btw

did you check

PhysicsStep component?

[drowsy pagoda]

Yup, cranked that shit up to extremly high step counts.

Tried all sorts of variations.

Didn't help one bit

Been working on this at every angle I could think of for a week straight lol.

Bout ready to give up on this project.

[rustic rain]

send it to Unity?

[drowsy pagoda]

That I haven't done.

[devout prairie]

Could it be something to do with the mass or inertia of your objects prior to attaching with fixed joint? Have you tried creating an equivalent gameobject setup in a subscene with fixed joint and seeing how it behaves?

[drowsy pagoda]

Yeah I considered that too. So until I got the right behavior all my objects have mass 1.

Inertia I haven’t considered. What’s the rule about inertia?

[devout prairie]

Well I think if your objects are misbehaving maybe probe some of the physics values and compare against for example a simple baked version of your body object

Just wondering if you are for example disconnecting an object that is part of a compound hierarchy for example and then adding physics mass etc to it when creating the jointed setup at runtime ( assuming that's what you're doing? ) maybe they are not initialized properly or something along those lines

[drowsy pagoda]

In authoring, all items are separate. You assemble them at runtime. I never pluck off a child/leaf collider from the body.

The joints are created during bake for items that will have other item(s) attached to them. Then during gameplay, I just assign the other body to the joint data.

[rotund token]

Did you try with havoc out of interest or no access to licence?

[drowsy pagoda]

No access to license. I want to though 😦

[robust scaffold]

In my experience, there's 0 difference from standard unity physics outside of stacking stability.

[viral sonnet]

isn't stability the main appeal of havok?

[robust scaffold]

I guess, yea. That's what they're advertising.

[covert lagoon]

Is the performance about the same?

[devout prairie]

one of the benefits is because Havok is stateful bodies can 'sleep' meaning if they are within a threshold of movement they'll no longer be factored in to simulation calculations, so sims with large numbers of objects will benefit from that where a lot of dynamic objects are stationary..

i think though because Havok is stateful it can also have a little bit more overhead so in some cases be slightly less performant..

[covert lagoon]

Hm.

[devout prairie]

kinda tricky to debug this one

[hushed lichen]

well, need to track down that q I guess 🙂

[calm edge]

Make sure stack traces are enabled for asserts in the player settings

[calm edge]

Is there something I need to do to get a subobject in a subscene to update it's position relative to it's parent?

[robust scaffold]

once an object is converted to an entity at bake in a subscene, you can either rely on the built in parenting transform system or, if it's a physics object, do it manually

[calm edge]

it seems it's not set up properly for the built in parenting

[rotund token]

is the child a physics object?

[calm edge]

the child isn't, but the parent does have physics

[rotund token]

children should just work

[calm edge]

there is no PropagateLocalToWorld which seems like it might be necessary?

in the editor I can see LocalToWorld changing, but not WorldTransform

[rotund token]

WorldTransform should be set to LocalTransform every transform update

are you using rival by any chance?

[calm edge]

I am

[rotund token]

yeah we ( @minor sapphire & I) noticed a bug with rival

rival overrides the transform system with a write group but doesn't update the WorldTransform

so your character is always at 0,0,0

you need a system after rivals own transform system that writes WorldTransfrom = LocalTransform

[calm edge]

hmm, are they cheating for the camera controller then?

cause that works

[rotund token]

i dont use rival

[calm edge]

LocalToWorld targetLocalToWorld = SystemAPI.GetComponent<LocalToWorld>(mainEntityCameraEntity);
hmmm

[rotund token]

i am summoning shinyclef with his fix

[minor sapphire]

oh hi

hold on I'll paste a system for you 😄

    /// <summary>
    /// Rival has a WriteGroup on component <see cref="Rival.CharacterInterpolation"/> that prevents the
    /// TransformSystem from updating the WorldTransform. This system copies LocalTransform to WorldTransform
    /// for any root objects with the CharacterInterpolation component.
    /// </summary>
    [UpdateInGroup(typeof(TransformSystemGroup))]
    [UpdateBefore(typeof(LocalToWorldSystem))]
    [UpdateAfter(typeof(CharacterInterpolationSystem))]
    [WorldSystemFilter(WorldSystemFilterFlags.LocalSimulation | WorldSystemFilterFlags.ClientSimulation)]
    [BurstCompile]
    public partial struct CharacterInterpolationWorldTransformFixSystem : ISystem
    {
        public void OnCreate(ref SystemState state) { }

        public void OnDestroy(ref SystemState state) { }

        [BurstCompile]
        public void OnUpdate(ref SystemState state)
        {
            new CopyTransformJob().Schedule();
        }

        [BurstCompile]
        [WithAll(typeof(CharacterInterpolation))]
        [WithNone(typeof(Parent))]
        private partial struct CopyTransformJob : IJobEntity
        {
            private void Execute(ref WorldTransform worldTransform, in LocalTransform localTransform)
            {
                worldTransform = (WorldTransform)localTransform;
            }
        }
    }

Also @calm edge I mentioned it here: https://forum.unity.com/threads/released-rival-dots-character-controller.1118761/page-3#post-8689680

[calm edge]

thanks, I just hacked up the CharacterInterpolationJob instead of adding a new system 😛

I'll have a read through that thread later

[signal phoenix]

Huh, i should make sure the WorldTransform gets properly updated next time I update the package. I was too used to always using LocalTransform for rigidbodies so I never noticed it.

@minor sapphire I'll be answering your forum questions probably tomorrow

[minor sapphire]

Awesome thanks. I figure I must be missing some fundamental differences between first and third person requirements for a netcode controller. Probably made worse by the fact in my game I can toggle between first and third person views 😄

porting first person sample to a third person context probably leaves some gaps somewhere heh

[calm edge]

The other problem with rivals I had was having 100s of physics objects causes the rivals code to take a massive amount of time doing physics steps

I'm new to ECS physics, but It seems like it's manually stepping the entire physics world?

oooh, I get it now, it's not the rivals code, it's the normal physics step that happens before and then the rivals code is waiting for its job to complete

[rotund token]

this is just a fixed step death spiral

not having burst on doesn't help

[calm edge]

Yeah, but I was just confused at what was responsible

I was just seeing how far I could push it

[late mural]

this warning appeared in the ide im using (visual studio), is it anything to be concerned about, or should i just ignore it?

[rotund token]

interesting error but i wouldn't be too concerned if it's not popping up in unity

[late mural]

ok good!

[safe lintel]

theres a forum post somewhere that shows havok being worse in terms of raw active rigidbodies(that werent sleeping). iirc it wasnt that much worse but kind of surprising

[rotund token]

hmm i'm not sure that's too surprising?

it has to do a lot more to determine state/sleeping/etc

but generally I think you'd have a lot sleeping so in most real world scenarios it's probably going to run a bit better

i tried to add the concept of coming to rest in unity physics, my implementation ended up being so gross

i need to figure out a good way to do it, determining rest was a pain

[late mural]

is it possible to use EntityManager.SetName() in a bursted function or is it impossible?

[rotund token]

it wasn't before, probably still not

[late mural]

ah unfortunate

[late mural]

EntityManager.CreateEntity(); will always put the new entity at 0 0 0 right? Meaning any child entities should have their local position be equal to their world position?

[safe lintel]

makes sense but at the same time i thought for sure havok would be super duper optimized or whatever to ensure it was the fastest and best choice of the two

[rotund token]

on the other hand, unity physics built specifically for burst

i haven't really benchmarked or seen a real world test so if anyone has one would be interesting to look at

[safe lintel]

i do wish they had gone forward with the extra low fidelity 3/6dof rigidbodies like they alluded to a while ago

not sure what the plan for stuff is long term but for many of my own uses just something that bounces/collides but doesnt need a super accurate sim would be fantastic for effects

[gusty comet]

I'm having an issue where I verify completely that I added elements to a DynamicBuffer inside of a Baker, and that during Baking the elements exist in the DynamicBuffer, and that DynamicBuffer is NEVER cleared or has its elements removed elsewhere in the code. Yet, the moment the game is run, the DynamicBuffer is shown as empty. What could be the reason for this?

This is only happening to one of my DynamicBuffers. Everything else works. the chunk size is far from being exceeded by a single entity

The Entities Hierarchy is showing that the elements exist

Both in Edit Mode and Play Mode

I'm printing the length of that DynamicBuffer inside of a job and it's always 0. That's despite the Entities Hierarchy showing that the DynamicBuffer has 2 elements

It's not disabled either

[rotund token]

your code has the funk!

not doing a reinterpret or something on the buffer

so the clear isn't being found from inspection?

ooo

ECB.AddBuffer?

try use the journal - it should tell you everywhere your component is accessed

[gusty comet]

Thanks for the suggestions. No reinterpret. I searched for usages and they're all confined to one file, which I scoured thoroughly. No .Clear(), no Remove(), no ECB.AddBuffer(). The journal didn't reveal anything - printing the length of the buffer from within a job shows 0 despite Entities Hierarchy literally showing the 2 elements it's supposed to have. Here's the access code:

foreach (var eventDone in eventDones.AsNativeArray())
                {
                    EventActionsJob.Execute(ECB, eventDone.Entity);
                }

It's inside an IJobEntity with this

[rotund token]

what is DecisionMakerEventDoneElement

and what does your baker look like

[late mural]

what would be the best way to get an array of all components of a certain type, guessing using an entity query then using ToComponentArray<>()?

[late mural]

hmm it seems ToComponentArray<>() requires a reference to something, how mysterious

[rustic rain]

That's not very good idea

It will copy all data on new memory

[late mural]

ah, should i use a get RW version then?

[rustic rain]

Yoo shoulg use job

that accesses compenets

[late mural]

ok and how would i access the components in a job?

[rustic rain]

i job entity

Or i job chunk

[late mural]

ok thanks a ton!

[rustic rain]

Non job alternative - systemapi query

[late mural]

ok

[lethal cairn]

Is it required entities to be in a subscene so that systems can operate on them?

I have entities existing outside of a subscene and observe that the movement systems won't move them, even though the systems can query them.

[late mural]

trying to use an IJobEntity, how do i complete the job?

[rotund token]

no

from the perspective of a query there is no difference between an entity in a subscene and not

[late mural]

it also seems i can only run, but the docs claim that there are more options

[hushed lichen]

Are you using an EntityCommandBuffer?

[rustic rain]

look at manual

they have examples of IJE implementation

[late mural]

i am, i hope

ah i see, im an idiot, thanks a ton!

[lethal cairn]

That's right. I have just tested with entites in a subscene, my systems won't run too. Don't understand why 🥲

[hushed lichen]

Are you sure your systems are running? Check the DOTS Systems window 🙂

[lethal cairn]

My system is not running. That's what I'm wondering about.

Am I doing it wrong here?

It knows there are 10 entities matching the query

But it doesn't run

[hushed lichen]

Huh, I have no idea

[rotund token]

how do you know it's not running?

did you throw in a debug log?

or a break point?

[rustic rain]

What singleton do I need to get

to run TriggerEventsJob?

SystemAPI.GetSingleton<SimulationSingleton>(?

[rotund token]

yes

[rustic rain]

after what systems those jobs are played back?

[rotund token]

update after PhysicsSimulationGroup

https://github.com/Unity-Technologies/EntityComponentSystemSamples/blob/master/PhysicsSamples/Assets/Demos/2. Setup/2d. Events/Scripts/Stateful/StatefulTriggerEventBufferSystem.cs

good reference

[smoky otter]

Hello, I am using Unity 2021.1.2, I am trying to implement something in the JobSystem, I have all the packages needed and can use NativeArrays, but I cant seem to find "NativeList", anyone familiar with this problem?

[rustic rain]

oh btw, what was that the fastest way to check whether entity matches query (has component)?
Smth about query mask if I remember

[rotund token]

yeah entity query mask

[rustic rain]

so, I just pass it to job and that's it?

[rotund token]

yep

create it from a query with .GetEntityQueryMask()

[rustic rain]

it's part of collections package

[smoky otter]

I had to import the package indeed, guess it is different than using Unity.Collections, thank you! 🙂

[rustic rain]

hmm, I'm guessing for disabled component my only choice is query mathing?

[rotund token]

it should work fine for disabled component

queries can match disabled entities

if you tell it to

[rustic rain]

I'm trying to figure what's better:

1. using IEnablableComponent
2. using buffer that contains list of disabled entities

Since application has very little entities (like probably 5 max disabled entities to track, before they are cleared)

[rotund token]

yes by default queries ignore disabled components

but you can tell them to include them

[rustic rain]

and ignore filter says

no I mean, this is mask

[rotund token]

yes but masks are made from queries no?

all entities in a query should return true if passed to the mask

so if my query is returning disabled entities, the mask should match them

[rustic rain]

it says it won't

look at last screenshot

[rotund token]

        protected override void OnCreate()
        {
            var query = new EntityQueryBuilder(Allocator.Temp)
                .WithAll<TestComponent>()
                .WithOptions(EntityQueryOptions.IncludeDisabledEntities)
                .Build(this)
                .GetEntityQueryMask();

            var entity = this.EntityManager.CreateEntity(typeof(TestComponent));
            this.EntityManager.AddComponent<Disabled>(entity);

            Debug.Log($"Matches Query {query.MatchesIgnoreFilter(entity)}");
        }

why do you make me do this!

[rustic rain]

but Unity description literally says so

lol

oh, btw

YOu specifically included disabled

meanwhile it should be vice versa

[rotund token]

what?

[rustic rain]

.WithOptions(EntityQueryOptions.IncludeDisabledEntities)
For my case I don't want this option

[rotund token]

then don't use it?

[rustic rain]

as I want to filter disabled somehow

[rotund token]

            var query = new EntityQueryBuilder(Allocator.Temp)
                .WithAll<TestComponent>()
                .Build(this)
                .GetEntityQueryMask();

            var entity = this.EntityManager.CreateEntity(typeof(TestComponent));
            this.EntityManager.AddComponent<Disabled>(entity);
            Debug.Log($"Matches Query {query.MatchesIgnoreFilter(entity)}");```

[rustic rain]

🤔

Did you test it with EnablableComponents?

[rotund token]

no

[rustic rain]

then I guess this is where mask won't work

[rotund token]

yes because it doesn't support filtering as the method says

[rustic rain]

well, my question initially was about them 😅

but I appreciate you showed interaction with query options

[rotund token]

Disabled

is a component

[rustic rain]

I guess under the hood Query automatically add Disabled Prefab and etc options

yeah

[rotund token]

which it does work for!

the question you are asking does it work for IEnableableComponent

[rustic rain]

yeah, and it's not. So I'm considering either using query check

or making buffer

which will contain disabled entities

[rotund token]

fastest way is to probably just do a IsEnabled on ComponentLookup

if you care about a specific state

[rustic rain]

in one case I'll have to implement IEnablableComponent pipeline
in other I'll have to implement a singleton buffer

I'm figuring which is less evil 😅

[rotund token]

i assume this is an entity reference?

[rustic rain]

my use case:
if player collided with specific trigger, trigger should get disabled.
Meanwhile trigger should only get disabled, no entity itself.

buffer for such disabled triggers will be very small, max 5 if player is lucky

[rotund token]

that seems straight forward enough

what's the problem?

should the trigger stay perm disabled (i.e. removed from physics simulation?)

[rustic rain]

I guess I'll just make IEnablableComponent

it's time for me to learn how they work at least

😅

[rustic rain]

Hmmmm.

        [BurstCompile]
        public void OnUpdate(ref SystemState state)
        {
            var singleton = SystemAPI.GetSingleton<EndFixedStepSimulationEntityCommandBufferSystem.Singleton>();
            var buffer = singleton.CreateCommandBuffer(state.WorldUnmanaged);

            state.Dependency = new InteractionJob
            {
                ecb = buffer,
                buffer = SystemAPI.GetBuffer<InteractionBuffer>(_interactionBuffer),
                interactableMask = _interactableMask
            }.Schedule(SystemAPI.GetSingleton<SimulationSingleton>(), state.Dependency);


            buffer.SetBuffer<InteractionBuffer>(_interactionBuffer);
        }

InvalidOperationException: The previously scheduled job InteractionSystem:InteractionJob writes to the Unity.Entities.EntityCommandBuffer InteractionJob.UserJobData.ecb. You must call JobHandle.Complete() on the job InteractionSystem:InteractionJob, before you can write to the Unity.Entities.EntityCommandBuffer safely.

did I do smth wrong?

[rotund token]

buffer.SetBuffer<InteractionBuffer>(_interactionBuffer);

you passed the buffer to the job then used the same buffer on the main thread

[rustic rain]

I see

safety system triggered because job handle was not complete

during that write

[rotund token]

just do it before the job and it should be fine

[rustic rain]

yeah, I just wanted to know why exactly it was a problem

hmm

So I need to make a physics object that can move and has physics object children.
It only gets moved by transform, no physics involved.
It's basically just moving trigger

and I need it to have multiple trigger shapes, which will produce different effect on trigger event

I used to do it previously with static colliders

it worked back in 0.51, but not anymore

so I'm wondering if I should make some sort of custom parent system, that will move children accordingly or...

damn

[rotund token]

Compound trigger should work right?

[rustic rain]

does compound allow you to distinguish between collider/entity pair?

for example:
on top you have sphere, on bottom box.
If sphere is triggered - result is A.
If box is triggered - result is B.

[rotund token]

I believe the collision events have the compound collider key included

So you can find the specific collider that hit

I'm in bed so can't look it up

[rustic rain]

yeah, trigger events has this property

I do wonder though, how can I get original entity from this collider key

I guess I will need to make some sort of dictionary on main entity during baking?

[rotund token]

From memory you get rigidbody which has the entity

[rustic rain]

meanwhile collider key is just uint

[rotund token]

Yeah it references a dynamic buffer on the entity with an the children I think?

Been a long time since I looked at this though

[rustic rain]

🤔

[rotund token]

Physics and I are not on good terms atm (well to be fair, more netcode handling of physics)

[rustic rain]

to setup compound

what do I need to change on shapes/bodies?

or do I even need body on children?

[rustic rain]

physics is least documented as well

oh

looks like main entity has this buffer

all right

so maybe this works

[rustic rain]

bruh, smth wrong with my algorithm and Editor crashes if I try to debug it

                if (colliderKeyLookup.TryGetBuffer(e, out var buf))
                {
                    var triggeredEntity = GetTriggeredEntity(buf, key);

            private static Entity GetTriggeredEntity(DynamicBuffer<PhysicsColliderKeyEntityPair> dict, ColliderKey key)
            {
                foreach (var keyEntityPair in dict)
                {
                    if (keyEntityPair.Key == key)
                    {
                        return keyEntityPair.Entity;
                    }
                }

                throw new Exception("GetTriggeredEntity didn't find correct entity.");
            }

I keep gettng exception all the time

what's more curious

with compound trigger events happen even if I don't trigger anything

oooh

looks like my physics body has empty buffer

I see now...

[hushed lichen]

I was going to ask about checking the contents of that buffer 🙂 For loops over a collection behaving weirdly is in my experience almost always because of some unexpected thing with the contents

[rustic rain]

well, it's still doesn't make any sense

my trigger event happens

out of nowhere

also weird, that conversion even creates that buffer for non-compound

considering it has default internal capacity of 16

here's my compound collider

waaait a second

wtf

I can see now why my trigger happens out of nowhere

my physics body is living it's own life before game starts 🤣

[gusty comet]

Here's the Baker

var eventDones = AddBuffer<DecisionMakerEventDoneElement>();
            foreach (var eventDoneList in authoring.EventDoneInstances)
            {
                foreach (var eventDone in eventDoneList.EventList)
                {
                    eventDones.Add(new DecisionMakerEventDoneElement
                    {
                        Entity = GetEntity(eventDone)
                    });
                }
            }

And DecisionMakerEventDoneElement is just an Entity reference to an Entity whose System represents a block of logic. Like an event

Here's the definition

[InternalBufferCapacity(0)]
    public struct DecisionMakerEventDoneElement : IBufferElementData
    {
        public Entity Entity;
    }

[rotund token]

yep it's horrible. in my buffer size override list

[rustic rain]

is that part of your library?

what is it called, so I can look it up later?

[rotund token]

just a window that looks like this that let's you override capacities

backend here
https://gitlab.com/tertle/com.bovinelabs.core/-/blob/master/BovineLabs.Core/Utility/ResizeBufferCapacity.cs

it basically just overrides typemanager data before anything uses it

[rustic rain]

good stuff

[frosty siren]

How can we create entities in baking system?

[rotund token]

with great effort 😄

tldr: you need to add a SceneSection sharedcomponent

to associated with a subscene

[rustic rain]

Don`t

Too much pain

Create all you want in baker instead

[frosty siren]

Ok. By some reason my particular singleton component doesn't exist in build so I looking for different ways of how I can bake my singleton. Have no ideas anymore because I use few singletons and all works fine, but particular one won't appear in build, spent an evening on that

[rustic rain]

How do you make this singleton?

[frosty siren]

Simple baker with zero fields. Works fine in editor

using Unity.Entities;
using UnityEngine;

namespace NSprites
{
    public class AnimationSettingsAuthoring : MonoBehaviour
    {
        private class AnimationSettingsBaker : Baker<AnimationSettingsAuthoring>
        {
            public override void Bake(AnimationSettingsAuthoring authoring)
            {
                AddComponent(new AnimationSettings
                {
                    idleHash = Animator.StringToHash("idle"),
                    walkHash = Animator.StringToHash("walk")
                });
            }
        }
    }
}

[rustic rain]

So, your subscene is not built?

You don`t have any 'editoronly' tsgs??

[frosty siren]

No, all works fine in editor, but particular singleton doesn't exist in build

[frosty siren]

nope

[rustic rain]

Do you have managed stripping??

[frosty siren]

No, I've finally found issue

Seems build crashes because system runs before subscene loaded and with new "all system updates without any require" system just accessing to not yet loaded singleton. I should account this new system behaviour.

[rotund token]

yeah i dislike this behaviour

of having to put require for update on everything for your singletons

[void girder]

So is there no more Job.WidthCode() with ISystem?

Just making sure

[rotund token]

there is not

just use IJob

[thorn trout]

Where can i find tutorial for ecs 1.0?

[calm edge]

pinned messages

[rustic rain]

bruh

what is wrong with physics

my object collides with 1 collider

but some unrelated collider also gets collided

wtf

if I make more copies of collider

they also get trigger event

lol

3 colliders

collision happened here, yet all colliders are affected by ITriggerEventsJob (which removes PhysicsWorldIndex from them and why they don't get drawn AABB anymore)

[frosty siren]

baking runs also before build?

[brave field]

🥲Native containers in IComponentData still not supported yet?

[rustic rain]

only in 1.0

[rustic rain]

yeah

[brave field]

I tried NativeList in IComponentData but it's not working

[rustic rain]

why not?

[brave field]

🤔 Not sure. It just not working and failed to bake into subscene

[rustic rain]

well, it works for me

[viral sonnet]

hm, i don't think it's intended (yet) for bakers

i'm pretty sure bakers/subscenes are not aware of serializing nativelists

[frigid crypt]

I need some help with scene management in entities 1.0

I am able to load a scene but I'm not sure how can I unload the previous one

I have tried to use SceneManager.UnloadScene() but it does not seem to be working

[rustic rain]

I was meaning runtime

[calm edge]

hmm, I'm using rivals character controller and firing slow projectiles so the relative velocity is important, but the character doesn't update physics velocity, is there any way to get velocity out of it?

[signal phoenix]

You can get it with:
KinematicCharacterBody.RelativeVelocity
KinematicCharacterBody.ParentVelocity

ParentVel is the velocity of the moving platform being stood on (if any), and RelativeVel is the character velocity relative to that. So the total velocity is the sum of the 2

[calm edge]

I added this:

                    float3 velocity = float3.zero;

                    if (parent.Value != Entity.Null)
                    {
                        if (SystemAPI.HasComponent<KinematicCharacterBody>(parent.Value))
                        {
                            var body = SystemAPI.GetComponent<KinematicCharacterBody>(parent.Value);
                            velocity = body.RelativeVelocity;
                        }
                    }

but the HasComponent is throwing InvalidOperationException: No suitable code replacement generated, this is either due to generators failing, or lack of support in your current context

[signal phoenix]

Hmm, I haven't encountered that one yet but you could try an editor restart

[calm edge]

no difference 😦

[signal phoenix]

Are you making the call to SystemAPI inside a system, or is it inside a job or a static method?

If you need this in a job, try passing a ComponentLookup<KinematicCharacterBody> instead and replace SystemAPI calls with that

[calm edge]

ah ok

yep that's all working now

[thorn trout]

is using a box collider per tile enough for a ECS voxel engine? It is for a game like v-rising and without colliders i don`t know how to make rival controller collide with chunks

It is for a multiplayer game, i need all colliders spawned

What is the best way?

[stiff mesa]

Hey everyone. A question to folks who have more experience with DOTS than me. When do I use managed systems, and when do I use unmanaged?

[balmy thistle]

use unmanaged whenever possible and unmanaged when you must interact with managed objects/APIs

[rustic rain]

wouldn't it make more sense to still make ISystem?

[balmy thistle]

make ISystem what?

[rustic rain]

unmanaged system, even if you interact with managed components/api

just no burst

[balmy thistle]

if you can restrict it to not interfere with the update, sure

[rustic rain]

why can't I use managed in update?

[rotund token]

if you don't use burst isystem is usually slower to update, also i spent christmas debugging corruptions caused by using an isystem with managed code

[rustic rain]

so, it's better to use SystemBase if it's not bursted?

[rotund token]

from my experience yes

also it's just more flexible so why not

[rustic rain]

quite odd

[rotund token]

obviously you should aim to make everything bursted

[rustic rain]

can't make Animator bursted 😅

[rotund token]

or can you?!

[rustic rain]

🤔

if you know smth, don't hide

[rotund token]

well, much of the animator is just wrapping native calls so i expect it'd be quite possible to call it from burst if you really tried (not in a job)

not that i would suggest doing this, just wait for a new animation system, way too much effort

[rustic rain]

it's 2d too 😅

so not only I'd have to wait for animation package, but also for 2d support

[rotund token]

i'll be honest, i know nothing about 2d animation requirements
i'm surprised the animator is even used though?

[misty wedge]

Just don't animate anything

ezpz

Since I have 0$ graphics budget I just use a really simple state machine for sprite animations

epic

[rotund token]

0x00007FFC8082B20D (RecastNavigation) [I:\Documents\BovineLabs\com.bovinelabs.shattered\Packages\com.bovinelabs.navigation\RecastNavigation~\Navigation.cpp:250] NavMeshGetTileRefAt 
0x00000252EA090876 (Mono JIT Code) (wrapper managed-to-native) BovineLabs.Recast.Interlop.Detour:NavMeshGetTileRefAt (BovineLabs.Recast.Interlop.DTNavMesh,int,int,int)
0x00000252EA090663 (Mono JIT Code) [I:\Documents\BovineLabs\com.bovinelabs.shattered\Packages\com.bovinelabs.navigation\BovineLabs.Recast\Interlop\DTNavMesh.cs:74] BovineLabs.Recast.Interlop.DTNavMesh:GetTileRefAt (int,int,int) ```

i write 2 lines of c++

and i crash my game 😐

[misty wedge]

That's what you get for being unfaithful to C#

[hushed lichen]

C# is a vengeful mistress, for she was created by Microsoft

[oblique cosmos]

Can confirm, I tried coding something in Java once and instantly my bathroom exploded.

(May or may not be a true story)

[hushed lichen]

That's why you go Java->C#
Java doesn't give a shit what you do when you're done with it.

[rustic rain]

animator can be used for anything that needs smth changed over time

[rotund token]

if you do it with playables you have built in support for netcode

[misty wedge]

Is that the animations stuff in netcode?

[rotund token]

playables has been a unity thing for ages https://docs.unity3d.com/Manual/Playables.html

think timeline

netcodes new hybrid / presentation system supports this

[misty wedge]

I meant is that what the animation namespace in netcode is for

I was wondering, since there is no dots animation system yet

Or is that just unused code

[misty wedge]

Also is there any documentation on this? I can't find a thing

[rotund token]

oh yeah

they have a kind of hybrid animations system

completely undocumented

add a
GhostPresentationGameObjectAuthoring script

and then a GhostAnimationController

and you have a networked hybrid entities animation solution

[rustic rain]

is there same but for single player?

[rotund token]

their implementation is very similar to how my old hybrid implementation worked

except they added a layer of netcode for prediction etc via custom playable nodes

[misty wedge]

How does it work / what does it even do?

[rotund token]

let's you animate via an animator / playable graph with netcode prediction support

[safe lintel]

is there an equivalent EndSimulationEntityCommandBufferSystem for baking systems?

[rotund token]

there are no command buffer systems setup in baking world

you could setup one but probably advised to just complete and playback on the spot

[safe lintel]

also for animation ive been spending a lot of time with bendzae's simple dots animation repo, id recommend it if you didnt want to use kinemation

(although its 3d only)

[rotund token]

it still doesn't support blending right?

[safe lintel]

ive added it and rootmotion support on my own fork

tbh rootmotion was kind of breaking my brain, and no guarentees to how good/perfect my implementations of either rootmotion calc or blending is, but it works

https://github.com/thelebaron/unity-dots-animation/tree/exp-rootmotion

[rotund token]

people still use root motion?

hopefully not on the actual player

[safe lintel]

also https://github.com/thelebaron/unity-dots-animation/tree/blending without the rootmotion stuff

[rotund token]

i thought this was mostly avoided these days for inverse techniques like motion matching

because root motion usually feels awful

[safe lintel]

well im fairly certain id have no idea how to do motion matching 🥲

[rotund token]

sadly same

i'm really hoping unitys new animation system isn't just a rewrite

and is a whole new modern implementation of animation

but i fear it'll just be another blend / state machine implementation

[rustic rain]

well, there is a fallback

we can use Latios kek

[rotund token]

(it's fine if that side is also included but i want more)

[safe lintel]

did you look at the old com.unity.animation package?

[rotund token]

it had a burstable timeline implementation!

[safe lintel]

latios is too hard for me to debug to use it

the dataflowgraph setup seemed very different from what exists for mecanim but i havent really used playable graphs to tell if its that different

[rotund token]

i think dfg was flawed

and probably led to the whole thing being rewritten

(not the concept, the implementation)

[safe lintel]

how so? i did see right before they announced 0.5 stuff it was rewritten fully bursted

impression was first release was like conversion, second that we havent seen maybe refined like baking? wild guess though

[rotund token]

didn't know they had rewritten it bursted

so that'd be the first step to fixing it

[safe lintel]

https://forum.unity.com/threads/some-questions-about-dataflowgraph-and-its-roadmap-as-runtime-scripting-tool.1046248/#post-7547824

[rotund token]

good

[safe lintel]

TemporaryBakingType can be used with buffers right?

[rustic rain]

yeah

[robust scaffold]

@signal phoenix Does your character controller allow for the case where characters can push dynamic bodies but not get pushed by them? Does that even make sense?

[rotund token]

isn't that just a kinematic?

[robust scaffold]

In the case where the pushed dynamic object can no longer be pushed, like against a corner, the player character should also stop like a dynamic object. Otherwise, it should act as a kinematic object.

My understanding of physics engines can't see a way this is possible. Maybe a smaller dynamic body within a kinematic body that detects if the kinematic is overlapping a dynamic body, in the case where an opposing dynamic can no longer be pushed, overrides positional data of the complete character. But the issue then arises in cases where two characters collide.

Rival prevents a kinematic object from phasing through walls and other characters on collision and also prevents characters from being pushed, which is good. But what happens when a Character - Dynamic - Character interaction, where two character controllers sandwich a dynamic rigidbody between the two and one attempts to push against the dynamic body.

[rotund token]

I haven't tried rival but this behavior worked fine in the unity physics character controller sample if I recall

And works fine on our controller at work on legacy physics

[robust scaffold]

Yea, Im trying to see if rival supports it

[safe lintel]

is it even worth scheduling a job only to complete it immediately after scheduling vs running it in an isystem?

(in the case of using ecbs/playback)

[rustic rain]

only if you want to thread it I guess

[robust scaffold]

Alright, testing shows that it does work with a C - D - C format. Just that the characters need to have the physics body set to collide. If the characters should be pushed, the body needs to be set as "None" or "Raise Trigger Events"

This is great. Now I'm gonna have to figure out how to implement a 2D top down format.

[rotund token]

There's usually run() for this

To avoid scheduling

[safe lintel]

yeah I figured as much. actually having difficulty finding where baking shows up when profiling the editor

[rotund token]

If it's baking usually you can just systemapi query

You can definitely profile it in editor though

But I think it has to be on an open subscene so it's the main process running it

[safe lintel]

ahh open the subscene

[rotund token]

Closed subscenes bake in a separate process

I usually just turn on editor profiler

Open subscene

Soon as lock disappears pause profiler

[robust scaffold]

Okay, a bit more experimenting with rivals. A character with dynamic physics enabled, physics set to collide, and an infinite mass results in a body that can push dynamics against walls and not through them but not be pushed by them or other characters. Exactly what I want. Now for more testing to see if this set of probably unintended settings causes physics to explode

The main issue I'm seeing right now, dont know if it's resulting from this set of parameters, is that dynamic rigidbody movement (not characters) are very poorly interpolated.

Oh wait, might be because physics baking is broken. Thanks unity.

[light badger]

(just saw your new reply above, but answering anyway)

I think this would be possible:

Make sure SimulatedDynamic is enabled on your character. Then, in your character aspect, customize the OverrideDynamicHitMasses callback so that the character has a characterMass.InverseMass of 0f (infinite mass). The Basic sample character does something similar in its BasicCharacterAspect.OverrideDynamicHitMasses in order to make the vehicle act like it has an infinite mass for pushing the character.

However, with that change, your character will now be pushing things without slowing down at all. If you want pushing to still take masses into account, there might be a way to calculate new InverseMasses for both the character and the other object in OverrideDynamicHitMasses in a way where if the relative velocity between the two would be pushing against the character, the character would be given an infinite mass, but if not, the regular mass ratio would be kept (we'd have to try it in practice, I've never actually tried it)

[robust scaffold]

Huh, havent thought about slowing the char down when pushing. I will need to experiment more with this but once I've implemented a 2d char controller as I think rival is the way forward.

[rotund token]

I finally found some time and got my nav mesh streaming to clients

i should have recorded this in higher resolution

[misty wedge]

Very cool

I fear I'll have to switch to a navmesh approach someday too

Grids are really annoying in a number of situations

[late mural]

wondering is there anyway to make sure an IJobEntity is finished running?

[rotund token]

Complete()

[late mural]

ah thanks!

[late mural]

where would i find this?

[rotund token]

jobs return a jobhandle

by default this can be hidden for ijobentity

but if you do like
state.dependency = yourjob.schedule(state.dependency)

then call state.dependency.complete()

it'll stall mainthread until that job is complete

[late mural]

ok thanks a ton!

[rotund token]

ideally you avoid this as it's not good for performance but sometimes it's necessary

[late mural]

How bad for performance is it, as i have about 10 of a certain type of component, and every now and again it needs to grab these components and use them immediately, but these components never change, so should i instead just store this list of components on a singleton component? (and note i only need to read the component data which should never change)

[robust scaffold]

Profile it but I dont think it'll be that bad. Consider using a blob asset if it doesnt change.

[viral sonnet]

hard to tell. what's bad about it is no further jobs get scheduled.

[late mural]

fascinating, ill definitely profile it, and mess around with it, thanks for the info yall

[robust scaffold]

Do you take suggestions to edit your samples?

[robust scaffold]

For one, in BakedSubSceneReferencePropertyDrawer, I would replace
var sceneField = new PropertyField(property.FindPropertyRelative("SceneAsset"), property.name);
with var sceneField = new PropertyField(property.FindPropertyRelative(nameof(BakedSubSceneReference.SceneAsset)));
which allows for editor side validation of inspector registration. In case someone wants to edit the name of "SceneAsset" and not wonder why the inspector suddenly broke from the renaming.

[robust scaffold]

In fact, I would replace the entire override with: csharp #if UNITY_EDITOR [CustomPropertyDrawer(typeof(BakedSubSceneReference))] public class BakedSubSceneReferencePropertyDrawer : PropertyDrawer { public override VisualElement CreatePropertyGUI(SerializedProperty property) { return new PropertyField(property.FindPropertyRelative(nameof(BakedSubSceneReference.SceneAsset))); } } #endif

[late mural]

wondering how i would make an entity the child of another, guessing using state.EntityManager.GetBuffer<Child>(SomeParentEntity).Add(???), although i cannot workout what to replace the question marks with?

[rotund token]

i think you just add a Parent component to the child with the parent entity in the field

[late mural]

oh ok, thanks a ton!

[rotund token]

(i don't do much in the way of hierarchies though)

oh nice query builder has AddAdditionalQuery

[robust scaffold]

What's a use case for that?

[rotund token]

you've always been able to combine multiple queries into 1

                new EntityQueryDesc { All, Any, None etc },
                new EntityQueryDesc { All, Any, None etc }
            );```

[robust scaffold]

Just an easier way to append components of an existing query onto a new query?

[rotund token]

how you would classically do it

this query would return entities from either subquery

                .WithAll<NavMeshSurface>().WithNone<WorldReference>()
                .AddAdditionalQuery()
                .WithAll<WorldReference>().WithNone<NavMeshSurface>()
                .Build();```

i'm using it like this

the old state.GetEntityQuery( isn't burst compatible because it takes params/array

[robust scaffold]

Oh, a union query rather than intersection if the WithAlls were merged.

[rotund token]

yep

it's a union of queries

[robust scaffold]

Interesting, let me see if that's relevant for anything I'm doing. I can see how that's useful

[rotund token]

it's somewhat of a niche feature that has existed for a long time

but it's useful when you need it

[robust scaffold]

I swear I wanted to do something along those lines, a union query, but I managed to implement the system using a classic intersection with a roundabout way of component querying.

But I cant find it anymore, oh well. Good to know it exists.

[late mural]

im getting this error, but i swear im initializing the random struct correctly: ```cs
[BurstCompile]
public void OnStartRunning(ref SystemState state)
{
ref ChunkMaster MapInfo = ref SystemAPI.GetSingletonRW<ChunkMaster>().ValueRW; // ChunkMaster is my component that contains the random struct aswell as a few other bits and pieces for generating the map

    MapInfo.Seed = (uint)UnityEngine.Random.Range(0, MapInfo.MaxSeed);

    MapInfo.RandStruct = new Unity.Mathematics.Random(MapInfo.Seed);
}

[rotund token]

you can't use UnityEngine random in burst

you have to use mathematics.random

[late mural]

ah, ill just unburst that function for now then...

[rotund token]

oh wait

thats not even your error

your error is you're passing 0 to mathetmatics random

you cant use a 0 seed

.Range(1, MapInfo.MaxSeed);

should be that at minimum

but this also implies your MaxSeed is likely 0 atm

[late mural]

oh ok thanks a ton!

[rotund token]

RandStruct = Random.CreateFromIndex(MapInfo.Seed)

it works for 0 to uint.maxvalue-1

[late mural]

ok ill use that thanks

still getting the error, and MaxSeed is definitely not 0

hmm it seems that seed is always set to 0, weird

by manually changing the rand struct's state to 1 it gets rid of the error message, how odd

[rotund token]

(i still dont think yo ucan use Unityengine random in burst)

[late mural]

i removed the burst compile attribute, that should be right, right?

[rotund token]

hmm

debug out MapInfo.Seed after you write it

and before the other rand call

[late mural]

ok!

the debug log never got called... what???

[rotund token]

yeah so im wondering if you're using this MapInfo rand struct somewhere

and this onstart method is just never called to set it up

[late mural]

no, as the error only occurs during chunk generation, which i have setup to only run when the player moves, the error only pops up then

any ideas why OnStartRunning() aint getting ran?

[rotund token]

sure system signature

: ISystem, ISystemStartStop etc

[late mural]

ISystem

[rotund token]

yeah on start running isn't called unless you also implement ISystemStartStop