archived-code-advanced

[midnight violet]

Just a suggestion, you know yourself best 🙂

[leaden sandal]

Okay, I'll go over my knowledge to see what I'm missing.

but in the mean time why won't it get what I mean by game input, there is a script named game input so why wont it recognize it?

[midnight violet]

did you rename the input asset of the c#?

[leaden sandal]

nvm I fixed it

the file name space thing was still input

The input is working now

[midnight violet]

yeah, thought so

and you get the right mousepos now?

[leaden sandal]

yeah

It actually works like yours now if I make the width super tiny

[midnight violet]

yeah this is something you gotta test around now between nearclipplane and the width of the line. you can set that codewise too to not have to fight around with that graph editor

[leaden sandal]

It works great until I rotate my camera then it gets really glitchy?

[midnight violet]

glitchy means?

[leaden sandal]

ill send a vid 1 sec

See what I mean?

[midnight violet]

ah, I see. You might have to generate the points based on the cameras.forward vector instead then. But I got no time to test around with that, sorry. probably later

[leaden sandal]

All good, I don't expect you to write the code for me.

I could just have a separate camera that only handles the line drawing and switch between them whenever i have to use the line drawing

[midnight violet]

I mean you can use the overlay option of the cameras and as you said, just use layers to render that line with your overlay camera

[leaden sandal]

Yeah but I scrapped that idea as the changing canvas size means the camera position has to be changed and that requires a lot of math

It has to be in the center of the canvas which is easy, but getting it the right distance away so that it matches the mouse position is annoying as it changes per screen size.

my first attempt was a ui object averaged between origin and mouse with the length being the hypotenuse of the mouses current position and it worked perfectly but changing aspect ratios ducked it up.

[midnight violet]

I am just wondering why the linerenderer disappears. Is it maybe just a rendering issue of the linerenderer? Could you use Debug.DrawLine to draw a debug line output , so you see if the positions are still correct even if rotating?

[leaden sandal]

There may have been something else I changed but I am very sure that all I changed was the material

It was probably something to do with that material not rendering too render textures because it had the wrong properties or smthn idk

[midnight violet]

Nah just because the linerenderer is a mesh, it might just rotate and therefor not being seen by the camera or what not

Oh so I am talking about the current state, not your first rendertexture approach 😉

[leaden sandal]

Oh right

I'll look into it in the morning, I'll lyk the problem when I have fixed it. (if I can)

[hybrid belfry]

Sorry about the delay. Yes, so I change the values in Inspector (Gridsize), this triggers the Redraw of the Grid by checking the bool. I was using Serialize to save the Grid data which Unity can't serialize

[lusty basalt]

Is there a simple way or pattern to opt out of a property function in c#? perhaps a way of making a property setter take multiple parameters? set the underlying variable to public and then assign via a lambda that conditially returns the underlying variable or the assignment?

I can think of ways to do it, but not any that seem clean or simple

[hardy sentinel]

example? pseudocode would help

mind that properties generate actual functions: e.g. public int myProp { get => x; set => x = value; } creates this under the hood:

public int myProp { get => myProp_get(); set => myProp_set(value); }

int myProp_get() => x;
void myProp_set(int value) => x = value;

[cosmic garnet]

Hi! How should android game be debugged if it crashes immediately? I press icon - black screen appears - game exits. This happens in like a 1-2 second timeframe. Everything works fine in editor and I get no errors. ADB Logcat shows:
Process com.CompanyName.GameName has crashed too many times: killing! Unity
This started happening when I installed Google Ads into the project.

[lusty basalt]

so my problem is that we are a very small team developing in a huge codebase
So while something like
health set(aDmgNr)
{
//prevent from going negative,
_health -= aDmgNr
}

might be a reasonable implementation at inception, it is often obtuse that Health is even a property when coming back to the same file way later and then health is used like -= Dmg instead of = Dmg like the property intends. this is ideally solved by naming and project code standards, I agree, but It got me thinking whether there migh be some way that I could structure a setter function that allowed

health set(aDmgNr, aTransformOrSimple = PropertyInput.Transform)
{
if(aTransformOrSimple = PropertyInput.Transform)
//prevent from going negative,
_health -= aDmgNr
else
{
_Health +=
}
}

While still retaining the overloaded interface of that properties provide, so that assignment can simply be done with = for example

[craggy sierra]

https://hatebin.com/dyctnilpvc
Could someone look at this .gitattributes? I wanna make sure this makes sense, took someone else's and a bit of my own understanding and put this together, wanna make sure this is good before I keep working on my project lol

also, does the following make sense to re-process all my files?

git rm --cached .
git add .
git commit

[hardy sentinel]

int _health;
public int Health {
  get => _health;
  set {
      _health = Mathf.Max(0, value);
      if (_health == 0) { OnDeath?.Invoke(); }
  }
}

[lusty basalt]

yeah I was just using the health thing as a simple example

but thanks!

[hardy sentinel]

you can't use multiple params on the property itself, but wrapper get/set methods aren't that uncommon

int _health;
public int GetHealth() => _health;
public void SetHealth(int amount, HealthSetMode mode) {
    if (mode == HealthSetMode.StandardMode) {
      _health = Mathf.Max(0, value);
      if (_health == 0) { OnDeath?.Invoke(); }
    }
    else if (mode == HealthSetMode.AllowNegative) {
      _health = amount;
    }
    else if (mode == HealthSetMode.Add) {
      _health += amount;
    }
  }
}

I assume u know all this, but going through it anyway 😄

as for the = part, if you REALLY WANT TO, you could do something with Tuples 🙂

(int health, HealthSetMode mode) Health_ {
    set => SetHealth(value.health, value.mode); // or could embed the logic here
}

// from another script
Health_ = (10, HealthSetMode.StandardMode);

[lusty basalt]

Thats actually very clean. Probably still to complex to use in production when we can just move to having some prefix/suffix that indicates property(so you remember to push f12) but honestly this was more about personal curiosity than actual implementation haha

thanks!

[hardy sentinel]

haha yeah wouldn't go with that in production 😄

[craggy sierra]

@dusty wigeon ok, they found an issue apparently, they labelled the issue as "Player Crashes on AtomicList::Load when when rerendering the scene"
apparently they couldn't reproduce on win10 or in a variety of other unity versions, for varying reasons between the other unity versions.

thanks for telling me to report!

[hardy sentinel]

a cleaner (architecturally) implementation would be adding a bunch of boilerplate and look more complex, lol.. but it would break less often

public interface IHealthEntity {    int Health { get; set; } Action OnDeath { get; set; }     } // Simple.. only has health and dies

public interface IHealthChanger {
    Func<int> HealthChangeAmountFunc { get; }
    HealthSetMode HealthChangeMode { get; }
}

public static class HealthManager {
    public static void ChangeHealth(IHealthChanger changer, IHealthEntity target) {
        // do all logic here
    }
}

if not with IHealthChanger, there should be a way to retrieve the Amount and the Mode from another class, and keep IHealthEntity as simple as possible

it's an example, but it's the ideal setup for all systems -- manager, gets the data from an object, makes changes to another object, interfaces are involved for loose coupling

[lusty basalt]

A really stupid way to do it that would introduce other risks

//globalaccess
public static GlobalPropertySetMode = PropertySetMode.Local;

//declaration
int _health;
public int Health {
  get => _health;
  set {
      if(GlobalPropertySetMode = PropertySetMode.Local){
      _health = Mathf.Max(0, value);
      if (_health == 0) { OnDeath?.Invoke(); }
      else{//...
      Global.GlobalPropertySetMode = PropertySetMode.Local;
      }
    }
  }
}

//application, if you want to opt out
GlobalPropertySetMode = PropertSetMode.Variant;
Health -= numbertoadd;

[hardy sentinel]

oh yeah no way xD

I would hardly even make the Health int public.. just

int IHealthEntity.Health { get; set; }
Action IHealthEntity.OnDeath { get; set; }

whatever systems want to access it can do so by only being interested in the IHealthEntity of the class (loose coupling)

in my code I usually do that with ICombatEntity -- that has all the info a combat system may need (hp, damage, armor, whatever)

but yeah the CombatManager only accesses the ICombatEntity parts of the objects.. never casts to e.g.: Player or Enemy

[lusty basalt]

If you need loose coupling for unit testing reasons i agree, but In smaller teams I am strongly of the opinion that his introduces waaay to much complexity(not saying you said you claim it is the best way haha)

[hardy sentinel]

it just requires a senior person that fully understands the pros & cons of these and is willing to communicate them to more junior members

but overall just the fact that it'd help keep the relevant code contained and reduce the spaghetti potential, would make it worth it

[lusty basalt]

we use a publicly accessible stupid global state struct blob for all communication between objects(Ideally, people do cheat for closesly coupled objects sometimes, making me fume out my ears...) so for example noone tells the healthbar to animate a healthchange, it simply watches the HP-value in the global state and reacts accordingly. Ideally this state should be so stupid that it has literally no logic other than copy constructors, but In practice sometimes it simply makes sense to ofload assignment logic into a propery like the health thing. But for example checking wheter something is dead is never done inside the state

so data wise there is no difference at all between a player and an enemy, the player is just an enemy that has different operations done on its data and vice versa for enemies

My general opinion is that it is better to have less complex(whatever that means) code that require the developer to be responsible in their implementations than it is to have more complex code that prevents the developer from making irresponsible implementations. Atleast for smaller teams. But I'm aware that is a very non-standard opinion.^^

[lusty basalt]

For the record Im not saying this would be good haha

[undone coral]

also, does the following make sense to re-process all my files?
no

[craggy sierra]

ok then how do i do it

[undone coral]

it's a really bad .gitattributes file

[craggy sierra]

w h y

[undone coral]

do you know what the instructions mean

[craggy sierra]

mostly yeah

not entirely sure what LFS does or what the unity yaml merger does internally

but i vaguely understand what its saying

[undone coral]

did you configure git to use unityyamlmerge

[craggy sierra]

isnt that what the gitattributes tells it to

maybe i need to install it

no clue how though

[undone coral]

the file is one part of configuring it

will you be working with terrains?

[craggy sierra]

yes very much so

thats why i change my gitattributes

i put that in a comment

[hardy sentinel]

Haha, I totally agree, but 'simple' depends on where you draw the line for what is simple and what's not 😄
For me personally training a new dev for 2-3 months to be able to read and produce code with my preferred architecture would be a no-brainer.
..but if time was pressuring and I wanted a finished product in 2-3 months, anything would go, really 🙂

especially if project was estimated to go on for more than a year.. imagine the amount of times it would break if architecture was weak.. better sort it in the beginning

and if using interfaces is your line where code goes beyond simple, you can expect some issues in the coming months xD

[undone coral]

@craggy sierra this .gitattributes snippet will deal with lfs files. you should never specify line endings for any file format, with the strict exception of a .bat file

#
# https://help.github.com/articles/dealing-with-line-endings/
#
* text=auto
# These are explicitly windows files and should use crlf
*.bat text eol=crlf
# 3D models
*.bin filter=lfs diff=lfs merge=lfs -text
*.3dm filter=lfs diff=lfs merge=lfs -text
*.3ds filter=lfs diff=lfs merge=lfs -text
*.blend filter=lfs diff=lfs merge=lfs -text
*.c4d filter=lfs diff=lfs merge=lfs -text
*.collada filter=lfs diff=lfs merge=lfs -text
*.dae filter=lfs diff=lfs merge=lfs -text
*.dxf filter=lfs diff=lfs merge=lfs -text
*.fbx filter=lfs diff=lfs merge=lfs -text
*.jas filter=lfs diff=lfs merge=lfs -text
*.lws filter=lfs diff=lfs merge=lfs -text
*.lxo filter=lfs diff=lfs merge=lfs -text
*.ma filter=lfs diff=lfs merge=lfs -text
*.max filter=lfs diff=lfs merge=lfs -text
*.mb filter=lfs diff=lfs merge=lfs -text
*.vrmesh filter=lfs diff=lfs merge=lfs -text
*.swatch filter=lfs diff=lfs merge=lfs -text
*.obj filter=lfs diff=lfs merge=lfs -text
*.mesh filter=lfs diff=lfs merge=lfs -text
*.ply filter=lfs diff=lfs merge=lfs -text
*.skp filter=lfs diff=lfs merge=lfs -text
*.stl filter=lfs diff=lfs merge=lfs -text
*.ztl filter=lfs diff=lfs merge=lfs -text
# Audio
*.aif filter=lfs diff=lfs merge=lfs -text
*.aiff filter=lfs diff=lfs merge=lfs -text
*.it filter=lfs diff=lfs merge=lfs -text
*.mod filter=lfs diff=lfs merge=lfs -text
*.mp3 filter=lfs diff=lfs merge=lfs -text
*.ogg filter=lfs diff=lfs merge=lfs -text
*.s3m filter=lfs diff=lfs merge=lfs -text
*.wav filter=lfs diff=lfs merge=lfs -text
*.xm filter=lfs diff=lfs merge=lfs -text
# Fonts
*.otf filter=lfs diff=lfs merge=lfs -text
*.ttf filter=lfs diff=lfs merge=lfs -text
# Images
*.bmp filter=lfs diff=lfs merge=lfs -text
*.exr filter=lfs diff=lfs merge=lfs -text
*.gif filter=lfs diff=lfs merge=lfs -text
*.hdr filter=lfs diff=lfs merge=lfs -text
*.iff filter=lfs diff=lfs merge=lfs -text
*.jpeg filter=lfs diff=lfs merge=lfs -text
*.jpg filter=lfs diff=lfs merge=lfs -text
*.pict filter=lfs diff=lfs merge=lfs -text
*.png filter=lfs diff=lfs merge=lfs -text
*.psd filter=lfs diff=lfs merge=lfs -text
*.tga filter=lfs diff=lfs merge=lfs -text
*.tif filter=lfs diff=lfs merge=lfs -text
*.tiff filter=lfs diff=lfs merge=lfs -text
*.bundle filter=lfs diff=lfs merge=lfs -text
*.so filter=lfs diff=lfs merge=lfs -text
*.a filter=lfs diff=lfs merge=lfs -text
*.dylib filter=lfs diff=lfs merge=lfs -text
*.dll filter=lfs diff=lfs merge=lfs -text
*.exe filter=lfs diff=lfs merge=lfs -text
# Large files
*.abc filter=lfs diff=lfs merge=lfs -text
*.terrainlayer filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
*.tar.gz filter=lfs diff=lfs merge=lfs -text
*.tgz filter=lfs diff=lfs merge=lfs -text
*.gz filter=lfs diff=lfs merge=lfs -text
*.bz2 filter=lfs diff=lfs merge=lfs -text
*.tar.bz2 filter=lfs diff=lfs merge=lfs -text
*.sbsar filter=lfs diff=lfs merge=lfs -text
*.sbs filter=lfs diff=lfs merge=lfs -text
*.psb filter=lfs diff=lfs merge=lfs -text
*.ptx filter=lfs diff=lfs merge=lfs -text

@craggy sierra to set up unityyamlmerge follow my instructions here https://gist.github.com/Ikalou/197c414d62f45a1193fd?permalink_comment_id=3269891#gistcomment-3269891

[craggy sierra]

" you should never specify line endings for any file format"
but if i do that git tries to write everything as one format

[undone coral]

you misconfigured git. your line ending settings should be checkout system, commit lf

[craggy sierra]

it literally just spams "git will replace X with X next time it touches it"

[undone coral]

that is normal

but that's also not what it says

[craggy sierra]

but unity hates it

"your line ending settings should be checkout system, commit lf" wdym?

[undone coral]

you should use git config --global core.autocrlf true

are you developing on windows?

[craggy sierra]

yes

# Configure Git to ensure line endings in files you checkout are correct for Windows.

uhhhh

but its for unity

unity has weird line ending preferences according to what ive read

[undone coral]

i have spoken

you are welcome to break your project in mysterious ways

if this is what you want to do

is your goal to work with other people?

it's always hard

unityyamlmerge is the hardest to configure, because even the docs are wrong

@craggy sierra you've made way too many configuration choices and they're interfering with each other. use the .gitattributes i shared. to correctly do terrains, do

Assets/Terrains/*.asset filter=lfs diff=lfs merge=lfs -text

instead of marking all assets as lfs

it's up to you guys to agree to put terrains in the terrains folder

[craggy sierra]

there are apparently some other weird assets that are binary

[undone coral]

extremely few. only lightmap settings, and you don't need to put every binary file into lfs

you can safely check in binary files into git

you can't check in LARGE files into GITHUB

you can of course commit large files into git. this is a github limitation, not a git limitation

you can check in binary files into github too

you can use my unityyamlmerge instructinos if you want it to work

trust me, i have probably been doing this 10-100x longer than you have

never git rm your whole folder and readd it

don't do stuff like that

[craggy sierra]

i severely doubt 100x but ive just never had problems lol

[undone coral]

if you have configured unityyamlmerge yourself

it was probably never working

[craggy sierra]

i havent configured it at all

[undone coral]

okay

[craggy sierra]

i havent even used unityyamlmerge in the git attributes

[craggy sierra]

ynot

how do i reprocess my files then?

[undone coral]

you can retrieve the logs other ways. did you use your google ads development id/key or a production id?

[undone coral]

it doesn't work that way

do you mean "how do i use lfs in an existing project"?

[craggy sierra]

thats only part of the issue

[undone coral]

what is your goal?

[craggy sierra]

when i add stuff to my gitignore, it doesnt get removed next commit

[undone coral]

are you trying to use git to work on something with other people?

[craggy sierra]

so i figured its the same with gitattribute

[craggy sierra]

no

[undone coral]

👀 if you git add .gitignore first, it will apply immediately

[cosmic garnet]

What are the other ways? I used development id. All I did was installed the ads package from official site, then went to samples tab, which opened GitHub and took the AdsController from there and did some minor tweaks

[craggy sierra]

but its already added in my tests

[undone coral]

i think you'll have to google it i'm not sure

[craggy sierra]

like existing project

[undone coral]

alright well i think you can start with a lot of recipes online

to help you deal with these problems

i've shared all the unity specific knowledge

[craggy sierra]

alright

[cosmic garnet]

I did Google it. Only answer I found that would apply to my case was through console with adb. And output from that didn't help me. That's why I am asking here

[craggy sierra]

so should i replace the lfs part of my git attributes, then handle terrains specifically with .asset, and then just use that?

oh and configure unityyamlmerge

[undone coral]

you have to remove all your line ending specifiers

[craggy sierra]

yeah

[undone coral]

with the exception of the two in my snippet

[craggy sierra]

i was going to mention that too

i mean i dont use bat files

[undone coral]

well you have to configur eline endings the way i said if you want things to Just Work

yes, you will receive a message from git

that it is converting the line endings in the commit for you

but keeping them as is on the file system

that is the correct behavior

the reason you need the settings i shared

is because some asset store packages are authored by users on macOS

so even though you think you're not "working" with anyone, you sort of are

[craggy sierra]

OH

ohhh

ok

do i need some magic to apply this to my existing project

or does it not matter

[undone coral]

i'm not 100% sure

the safest thing to do

is to make the gitattributes * text=auto change

then the git config global change
push those commits

then clone your repository anew

however, i'm not sure how git deals with checking files that have been marked as a certain line ending in the past

i have no idea because i've never put an existing project into this kind of jeopardy lol

[craggy sierra]

whatever ill just set it the way you said, then just leave it probably

ive never had issues

except for the terrains

which i figured out why because i treated terrain as text

@undone coral wait it says .asset is treated as -text do i need to add a specifier for terrain then? I guess I should still do that because its not yaml, just binary data

https://hatebin.com/czyrbveazi

ok how does this look?

[undone coral]

this looks good to me. you can probably remove the .sln and adjacent lines

[craggy sierra]

Alright

[jagged fiber]

Hello everyone! 🙂
Have a question about Scriptable Objects and this article:
https://unity.com/how-to/architect-game-code-scriptable-objects

The proposed idea there is to store both default and runtime state in Scriptable Objects.
(e.g. [SerializableField] initial HP and [NonSerialized] current HP), so that even the runtime state can be shared without breaking modularity.

How do I do this, when I have a single EnemyHP scriptable object, while having many enemies?
How to efficiently share runtime state in this case, which will differ from enemy to enemy?

[compact ingot]

you have two fields on the SO, one stores the serialized initial value and the other any change to that value which is not serialized, ideally however, those would be separate objects and you would not store your individual object's state in SOs.

What you do instead is use the SO as an intermediary between any two objects that hold its reference to communicate, only sending values through it but not storing them.

[hybrid plover]

I'm trying to make this code multithreaded, but 'GetPixel' needs to be called from main thread. Any way on how to get around this?

[undone coral]

what is your goal

[compact ingot]

you would ideally only write the pixels to the texture once your calculations are complete and have the parallel code work on threadsafe/thread-usable structures

[undone coral]

is this editor code?

[hybrid plover]

yes

[undone coral]

are you trying to pack the channels

as a utility

[hybrid plover]

was meant for this

yes

[undone coral]

or write a converter

what are you converting from

VRay?

are you writing a materials converter

[hybrid plover]

Yeah material converter from URP to HDRP

[undone coral]

oh..

hmm

[hybrid plover]

packing smoothness, metallic, ao to one mask map

[undone coral]

the packing thing exists on github already

i forget what it's called

[hybrid plover]

It works, but for large files (4k or larger) it takes around 17 sec per map and i'd like to make it faster

[hybrid plover]

If you could find it that would be nice

[compact ingot]

just don't use getPixel, read the entire texture into a buffer, process that, then write all of them back in one go via SetPixels

[hardy sentinel]

or, evn better, use gpu

[compact ingot]

yeah, a conversion shader would make it practically instant

[hybrid plover]

Hmm yeah I'll give that a try

[hardy sentinel]

check out compute shaders

[compact ingot]

doesn't have to be a compute shader

just render to a temporary render texture

the bottleneck will be the encoding to a JPG/PNG

[hybrid plover]

right now, encoding is less than one sec

[compact ingot]

yes, that will become the bottleneck

if you use a shader

[hybrid plover]

oh okay, i'll look into it. Thanks guys!

I'm going to try reading it into a buffer first

[flint quest]

Has anyone figured out how to successfully switch unity scenes when using hybrid ECS?

[hardy sentinel]

xD

[dapper cave]

my conclusion too 👍 requests are naturally staggered. i'll keep an eye on the profiler, it's easy to replace. magnitude by .sqrmag and if need be with kd tree

looks like kd would be needed for boids, i'll keep that for later.

[hardy sentinel]

what screen setup do u guys have/recommend? Been with 2x24'' monitors for a while. Curious what else works fine
(need to ask here to avoid troll answers in unity-talk)

[fresh salmon]

await sender.RedirectTo(channels.Single(c => c.Name == "#unity-talk"))
No escaping!

[jovial totem]

I have a problem that's hard to describe. I am estimating the amount of light at a single point in space, but the visible luminance of a surface seems to increase logarithmically with the amount of light hitting it. How do I transform the value to become linear? I know this is a math question, but I can't wrap my head around it.

[craggy sierra]

can you explain what you are trying to do atm

like what are you doing to got that luminance

etc

[jovial totem]

taking a snapshot of an octahedron and downscaling the result to 1x1 px

[compact ingot]

explain the purpose/goal i think

[jovial totem]

get the visual light level at a point for stealth mechanics

[compact ingot]

i think this discussion has happened before

[compact ingot]

well, if its truly growing on a log scale, you just Mathf.Exp(level) it

[hazy epoch]

Maybe I'm not understanding the (limited) documentation...but should there not be a way to iterate through the Array of Bindings on a PlayableDirector? The only method I'm seeing that would help with that is PlayableDirector.GetGenericBinding, but it takes an object...

[dusty wigeon]

You do not want to use the visible light for this because you will not be able to detect what is the state of your character if he is occluded. Instead, you should use Physics to detect if the player is "visible".

[hazy epoch]

It doesn't seem very code safe to have to hardcode in what object you're looking for. You should be able to iterate over the Array and find out what component that Binding is expecting.

[dusty wigeon]

Do you use PlayableDirector.SetGenericBinding or other form to set the binding ?

[hazy epoch]

I'm not using either yet, because I'm trying to figure out how to even "see" the binding.

[jovial totem]

im using both

[dusty wigeon]

Just use Physics. You do not want to deal with luminescence. There is nothing to gain, you are going to get pretty decent result without it.

[dusty wigeon]

Do you have any binding ?

[hazy epoch]

Essentially we have created timelines in a generic scene, and are then Instantiating them as needed. Because the links from the components to the bindings in the track won't exist anymore, we need to go in and re-establish them after Instantiation.

Something along the lines of...

foreach(PlayableBinding b in PlayableDirector.Bindings)
if(b.sourceObject.GetType() == typeof(Animator) )
b.sourceObject = GetComponent<Animator>();

Problem is... PlayableDirector.Bindings isn't a thing.

[dusty wigeon]

I do know a lot about Playable, but are you sure you need to work directly from the PlayableDirector ? Does the Graph contains the information you are looking for ? https://docs.unity3d.com/ScriptReference/Playables.PlayableDirector-playableGraph.html

[vernal furnace]

What should I use code wise to toggle vsync? The unity docs are telling me to use two different things. Am confused
Application.targetFrameRate = -1;
QualitySettings.vSyncCount = 1;

[hazy epoch]

That's probably adding to my confusion. The documentation is just so unspecific. So in terms of a timeline...the PlayableGraph.GetOutput() would return one of these? https://i.imgur.com/h8PVwYq.png

[dusty wigeon]

vSync and targetFrameRate does the same thing. They both limit the frameRate.

Depending on your target platform, use either vSyncCount or Application.targetFrameRate to set your application frame rate.

https://docs.unity3d.com/ScriptReference/QualitySettings-vSyncCount.html

It depends on what plateform you are using, but some support, some do not, while some force arbitrary value.

[compact ingot]

use target frame rate if you want to cap the frame rate without locking it to the monitor refresh rate

[vernal furnace]

Vsync is off by default, but if they toggle it on should I be changing the target frame rate or uncapping it?

[dusty wigeon]

I am expecting that you might need to iterate on the Graph first then retrieve the binding associated with the actual node of the graph.

[compact ingot]

you should always set a reasonable target frame rate or at least allow your users to set one, if you don't you are just gonna waste energy for no gain

[undone coral]

what is your objective

to make a thing in your graphics settings?

[vernal furnace]

Toggle/untoggle vsync in a settings menu

Yes.

[undone coral]

on windows and macos?

only windows?

[vernal furnace]

Currently only windows.

Mac is another monster.

[dusty wigeon]

`Mobile platforms ignore vSyncCount. On mobile devices, when a frame is ready, it displays the next time the screen refreshes. You can use Application.targetFrameRate to control the frame rate on mobile platforms.

VR platforms ignore vSyncCount. Instead, the VR SDK controls the frame rate.

On all other platforms, Unity ignores the value of targetFrameRate if you set vSyncCount.`

[undone coral]

Application.targetFrameRate = -1;
QualitySettings.vSyncCount = m_IsVsyncOn ? 1 : 0;

[vernal furnace]

Okay so I was confusing the two. Ty.

[vernal furnace]

Yes, but that didn't answer my question. I had the two backwards.

[dusty wigeon]

What do you mean ? You do not have to set Application.targetFrameRate.

[vernal furnace]

So I just don't touch it than

From what I understand from you and from doctorpangloss, all I want to change is
QualitySettings.vSyncCount = m_IsVsyncOn ? 1 : 0;

[burnt egret]

Hi there, i realize that this might not be something super advanced but i can't seem to get it to work for the life of me.

I have a tree of nodes that is rendered on a worldspace canvas. I'm trying to draw a line in between the parent node and all of its child nodes recursively. The problem is, the line renderer just seems like it doesn't want to work no matter what I do. It just sets the position of the line segments to 0,0,0. Is there an alternative, or should i start making my own line rendering solution that uses Images rather than whatever the line renderer is using?

[compact ingot]

there is no good way to draw arbitrary lines on a canvas

[burnt egret]

unfortunate

thank you

i'll update if theres any interesting solution i come up with

[regal olive]

hello

im trying to make a spaceship with the behaviours of this:

https://www.youtube.com/watch?v=tMRxBtzufU4

https://www.youtube.com/watch?v=GwkfU4i9Ja4

the second video start clip at 5:57

for the best example

the first video

is a good

but really awful looking

this is my code so far

warning

its terrible

https://hatebin.com/yfghfwtlve

[analog bane]

One small tip on a first glance, would be to name that flight mode function ToggleHoverMode, and you can just do hoverMode = !hoverMode;

[regal olive]

okay

but what else

my code doesnt work aswell lol

but i think u get the idea

[timber flame]

Should we use Dispose IDisposable zenject to unsubscribe events for non mono classes injected through zenject?
What if both (subscriber/listener and event raiser) have same life time? Both of them are destroyed when unloading a scene

public class Subscriber : IDisposable
{
   private readonly EventSystemA _eventSystem;
   public Subscriber(EventSystemA eventSystem){
      _eventSystem = eventSystem;
      _eventSystem.EventB += OnEvent;
   }
   public void Dispose()
   {
     //here
   }
}
public class EventSystemA{
}

[analog bane]

I'd also honestly unify the booleans, otherwise you'll be able to have flight mode and hover mode be true

[regal olive]

wdym

[analog bane]

As far as I understand, VTOLs are either driving on the ground, in VTOL mode, or plane mode. That's 3 states, but you have 4 possible states with those booleans

[regal olive]

when will a spaceship drive on the ground

or a plane

i just want to get into the spaceship

have it go into hover mode

and then go up

and then go flying

outside the atmosphere

[analog bane]

If you don't need the other boolean, then I'd operate without it

If it's not in hover mode, you can then know it must be in flight mode

There's also that standalone transform that is inside it's own curly brackets, but I'm not sure how c # handles that

I would normally expect an else statement for that to work

[thin mesa]

brackets not attached to any statement just create another scope, the code inside will still run since there's no condition/branch preventing it from doing so, it's just another scope. in the case of their code creating that scope is basically pointless and likely just leftover from some if statement they've since removed

[analog bane]

Gotcha

[burnt egret]

thanking chat-gpt for this one, this script lets you draw a line on a canvas with a linerenderer

using UnityEngine;
using UnityEngine.UI;
[ExecuteAlways]
public class LineDrawer : MonoBehaviour
{
    public RectTransform lineRectTransform;
    public LineRenderer lineRenderer;
    public RectTransform pointA;
    public RectTransform pointB;

    private Vector3[] positions = new Vector3[2];

    void Start()
    {
        lineRenderer.positionCount = 2;
    }

    void LateUpdate()
    {
        // Get the positions of the child objects in world space
        Vector3 worldPositionA = pointA.position;
        Vector3 worldPositionB = pointB.position;

        // Convert the world space positions to screen space
        Vector3 screenPositionA = RectTransformUtility.WorldToScreenPoint(null, worldPositionA);
        Vector3 screenPositionB = RectTransformUtility.WorldToScreenPoint(null, worldPositionB);

        // Convert the screen space positions to the local space of the canvas
        RectTransformUtility.ScreenPointToLocalPointInRectangle(lineRectTransform, screenPositionA, null, out Vector2 localPosition0);
        RectTransformUtility.ScreenPointToLocalPointInRectangle(lineRectTransform, screenPositionB, null, out Vector2 localPosition1);

        // Set the line positions in the LineRenderer
        lineRenderer.SetPosition(0, localPosition0);
        lineRenderer.SetPosition(1, localPosition1);
    }
}

[crystal parrot]

Is there any mesh combine studio for ecs?

[ancient talon]

I would recommend looking into Raymarching for stuff that involves procedural mesh combination at runtime.

[crystal parrot]

No, i need it because my city is all modular

it has lot of

(really lot) of shadow caster

even sinty assets from polygon are super slow because this

[dusk bloom]

https://hastebin.com/share/mapicizoje.csharp

Essentially what is going on is that I am moving the gameObject that the script is attached to back and forth between a undefined set loop of transforms using a horizontal input,

when going forward through the transforms, it will select the next one in line as the one to go to, however when going backwards for some reason, it doesn't select the one before it as the one it should be heading towards?

(when it should)
any ideas of why this is happening? (Unity 2D URP)

[patent bear]

Only took as quick look but line 55 is my quick guess as to the issue. You seem to be using two different values to track the current index (currentPointIndex and currey) and checking both in the of statement.

What is the point of the currey variable rather than using a single index?

This topic also would probably be better posted in general rather than advanced in the future as it's not a particularly advanced issue.

[lavish meteor]

Hi, is there a way to completely simulate ingame processes? Basically I want to let my players train AI but the PCs in question are low spec. Because of that the training can't run in realtime but should rather be a complete simulation of the training process... (e.g. a simulation of the car driving but without actually loading game objects and rendering stuff)

[untold moth]

"without loading gameObjects" would imply that you have the simulation logic separate from unity. In this case, why do you even need unity? You could just run a console app.

[lavish meteor]

Yes that would be the most complicated option for which I'm hoping to find a workaround for... There are things like the headless mode which would batch those values without it really running in realtime but this isn't viable since I want the player to train the AI by setting values etc.

[dusty wigeon]

I am not sure a low PC can train an AI for car driving in a relatively short amount of time.

If you want to train your AI with the best performance and that you do not need Unity GameObject, I would suggest that you externalise the training. Removing the "Graphics" won't be enough.

[dusty wigeon]

You can still set values at runtime.

[untold moth]

I don't understand. What do you need the engine for? Can you explain?

Since you mentioned "without loading gameObjects", it means that you're not using any of the engine features.

Or are you using ecs?

[lavish meteor]

I need the engine for the game. Basically the player sets his values (like learning rate, discount factor etc) and presses on "train ai" or smth. Then the AI would be trained without any visual feedback but every output would be stored in some kind of List so the player can see the learning process after the training. That way he could speed it up, skip it etc. It could run externally but even with this simple of a game there is so much to simulate

And I can't simplify the training environment because I would have to simplify the game environment too

So I was hoping for some kind of build-in simulation mode or smth

Things like collisions would basically be impossible without a proper physics simulation

[untold moth]

Well, gameObjects and components are the core of unity engine. Without using them, you're not using any of the engine features.
As for simulation speed, you can controll it for example via Time.scale or your own logic.
The rendering can also be disabled - just remove cameras from your scene.

[untold moth]

You can't have collisions without colliders and rb. And you can't have those without gameObjects. Which is why I was so confused.

[lavish meteor]

This seems to be a viable option. Any idea how that would be performance wise? Would it come near completely simulating the environment?

[untold moth]

What do you mean by "simulating the environment"?
It will disable rendering and free up the time that would be spent on it, so it really depends on how much was your rendering taking and wether there are other performance heavy things in the project.

If you need specific numbers, profile your game and see what's taking the frame time.🤷‍♂️

[lavish meteor]

Complete external physical simulation of the game environment. In theory it would be the same thing as not rendering the actual process?

[untold moth]

Wdym by "rendering the process" and "external physics simulation"?
Rendering is one system that runs each frame. Physics system is another one. Then there're scripts that have update. Some other components could also be updating.

You can see everything that's being processed in the profiler. Then you can decide what systems you could stop if needed.

[lavish meteor]

By external physics simulation I mean a script that does all calculations done in the unity physics system and therefore simulating the rotation, position, velocity etc of the car after the AI has provided certain outputs

[untold moth]

Okay, so what else aside from rendering do you need to disable?

[lavish meteor]

I would just need the player objects which interact with the track boundaries and the corresponding learning algorithms... The question is probably if Unity could handle for example 100 instances of an object, all with different neural networks... (or for my reinforcement algorithm, the continuous actor-critic learning in the background)

Since unity isn't really made for this kind of stuff right?

[untold moth]

There isn't really anything unity specific in that equation. Aside from maybe physics. Physics probably can handle 100 vehicles, I'm not sure what numbers you'll get exactly though. Gotta test that.
The rest of it is your logic and neural networks. None of which is unity specific. It would have the same overhead as in a console app.

[dusty wigeon]

Training an driving Agent, depending on how the environment is, could take 1000 to 100'000 iterations.

[lament salmon]

@lavish meteor You can also add the cars in a separate physics scene and call this
https://docs.unity3d.com/ScriptReference/PhysicsScene.Simulate.html

[lavish meteor]

Exactly what I was looking for. Is there a performance difference to just not rendering the objects instead?

[lament salmon]

Not sure about that, but you can simulate it whenever/however fast or slow you want by calling Simulate

So the other scene doesnt necessarily have to use FixedUpdate, Update etc.

[lavish meteor]

Okay. How do I make an instance of this scene and add objects to it? Can't find that in the docs

[lament salmon]

I think you make a scene and get the physicsscene from that. Look here
https://learn.unity.com/tutorial/multi-scene-physics#

[lavish meteor]

Thank you

[untold moth]

It would be the same as just adjusting the fixed timestep of the default physics scene.

[regal olive]

Hello, rn I have a vtol aircraft that has a specific hover height. But that is in the world. How would I make the hover height be a specific distance from the ground

[modern shell]

Well you need to do a raycast in the down direction and check where the ground is

[undone coral]

"train ai" or smth.
how serious are you about trying to achieve this?

[undone coral]

train AI but the PCs in question are low spec.
then you will be greatly limited in what you can do

[rocky grove]

Hey there 👋 I'm currently building developer-local environment configuration into our project. Previously, we had a globally injected component that was used to configure the game's server environment (production, staging, local fake environment, etc.). On release branches, this had to be always set to "production" which was cumbersome (people constantly touched and committed changes to the prefab while testing). The new system I'm building now takes the final environment configuration from the build pipeline (via command line arguments passed to Unity while building), which prevents that our build will ever run with undesired environment configurations.

To stay flexible, I am trying to add a user preference screen that allows our team to locally select an environment to run against. The selected environment is not committed to the project Git repo, preventing people from messing up the assets while testing. I can inject this configuration when running a custom build pipeline script. However, I wonder if there's some way to inject an editor-only asset into the build when only using the default Unity Editor build window? I want to make sure that if people build the normal way, that their local environment selection is regarded. Any thoughts on how I could achieve this are appreciated.

[lavish meteor]

I was referring to the name of the button... I actually already have an actor-critic learning algorithm in place and I'm onto this project well over a few months. It's a school project and therefore the performance is limited to the school computers (which are pretty low spec)...

[undone coral]

what is the goal?

it sounds like there's a game, and also "an AI"

[lavish meteor]

Playfully visualizing two learning algorithms and designing it to be interactive and accessible

[undone coral]

inject an editor-only asset into the build when only using the default Unity Editor build window
i believe you can add whatever build steps you want to the default build pipeline

take a look at interfaces like this https://docs.unity3d.com/ScriptReference/Build.IPreprocessBuildWithReport.html @rocky grove

[undone coral]

doesn't ml-agents already do this

isn't there a mature openai gym interface to unity? i am pretty sure ml-agents has a pre-authored way to record what is happening in training

i'm not sure why you are talking about running multiple agents in parallel yet. are you saying this is important for the task? what is it? it sounds like you want to have racing

[lavish meteor]

It does, but I want to do it myself. The machine learning itself isn't the problem here

[undone coral]

but it also sounds like you are trying to achieve some kind of parallel computing inside of a unity environment

[lavish meteor]

Genetic algorithm

[undone coral]

The machine learning itself isn't the problem here
it sounds like it is the huge, looming problem

knowing how many details there are, even if you were to achieve what you want - running N unity scenes within 1 unity process simultaneous, to fit your ill-architected approach - it will be so full of bugs that you will have no idea if it's working

are you saying you want to show a genetic algorithm process to generate & select a car driving algorithm, and you want the "playful visualization" of seeing 100 ghost cars driving around and then 10 ghost cars driving the furthest being selected, etc. etc.?

@lavish meteor why not record traces from the appropriate environment, and visualize the traces as a data format? it's so much simpler

then your ga fitness step can be 100 unity processes for 100 agents (the only viable architecture)

[lavish meteor]

This is literally what I was asking about and what I'm trying to do

[undone coral]

and at the end of the method in python you've authored to do the fitness step, you can write a json file, and read it separately, in a different application, perhaps the same one that is running the training, to visualize all 100 vehicles at 1 time

what difference does it make if it's 10ms or 60s of time between when the simulation occurred and when you visualize it? it makes no product sense anyway, so don't stress about it

[lavish meteor]

Then the AI would be trained without any visual feedback but every output would be stored in some kind of List so the player can see the learning process after the training

[undone coral]

you can't run 100 agents in 1 scene

do you see that?

that's a bad architecture

[undone coral]

my feedback to you is that you are poor at communication

you are getting preoccupied with telling me i'm wrong, instead of working on the underlying issue, a bad architecture

focus on that

you cannot run 100 agents in 1 scene

1 candidate per 1 process in your fitness function

then you can chew through N-cores-per-machine candidates at a time in the fitness function

you don't need physics for your traces

[lavish meteor]

Because you still think I'm trying to run it in realtime... The problem is solved and you don't seem like you want to help/give feedback but bash on my project?

[undone coral]

are you trying to simulate (in the sense of interact with physics) more than 1 candidate / agent / actor per unity process, in any way?

simple yes or no

visualizing the traces can occur with physics turned off

you're not giving me a simple yes or no

which is worrying

it's okay if it's just "no"

[lavish meteor]

What do you mean by 'per unity process'? Each timestep? Update function?

[undone coral]

okay, so that's a "yes"

well, you can't do that

is this a race track? is the idea the candidates are cars trying to run a lap the fastest?

simple yes, or what is the idea?

are they supposed to interact with each other? like are there 10 racers in 1 course, and they can bump each other?

or they do not collide?

[shadow dagger]

Guys, who can help with IK?

[undone coral]

What do you mean by 'per unity process'?
for every mything.exe in your task manager built with unity, are there 2 or more candidates being simulated in mything.exe?

[devout coyote]

NGL solving a car driving problem using a genetic algorithm feels like hitting a nail with a screwdriver

But that's a whole different conversation

[undone coral]

the very best thing you can do

is stay focused on my questions

[lavish meteor]

yes. I don't get why I couldn't do that?

[undone coral]

okay

is this because the cars interact with each other?

i don't get it. you have to just state what the environment is

and take your fingers and strike the keys on your keyboard 100x faster

because while i'm sure you're a good programmer and you've achieved a lot so far, another piece of feedback to you is, the people who are truly really good at this talk and write 10x faster than you do right now

[lavish meteor]

The problem is, that you are just disrespectful

[devout coyote]

[undone coral]

that is my feedback to you. i'm sorry i couldn't be more helpful

i am trying to help you thrive

[lavish meteor]

I appreciate any help I can get. But if your only way of helping me is bashing on my project and writing disrespectfully I'm rather gonna look for help elsewhere

[undone coral]

i don't know. i can tell you are really good at this, and usually, when someone shows an interest in your work, it is the opposite of bashing

most people would clock out and stop paying attention

[lavish meteor]

yes

[undone coral]

like i said, my feedback is you should try to answer the questions because it will help you

[undone coral]

do they interact with each other?

like is it important that the cars collide with each other?

[lavish meteor]

not in the learning process. In the actual game yes (If I make the ais race each other)

[devout coyote]

He's supplied you with a bunch of questions that you need to answer before he can give you any insightful information and help you out.. he's raised some very valid concerns and I don't think you took the time to answer any of his questions with the detail required to really habe a conversation

[undone coral]

okay

then there should be just 1 candidate per unity process in the fitness step

the solution found by GA will not work well with collisions unless it is specifically trained on collisions

not work well is an understatement, it will behave catastrophically / chaotically

there's too many bugs with trying to get 10 candidates computed simultaneously (in parallel) in 1 unity process .exe

however, you can obviously visualize as many as you want simultaneously

[lavish meteor]

So Unity can't handle that many gameobjects?

[undone coral]

the code that implements the fitness function should not live inside C#

[undone coral]

unity built in physics cannot correctly run well enough to produce the results needed for a GA fitness function

it can handle lots of game objects

it can barely handle 1 physics scene correctly. if you didn't configure things correctly, such as by using Physics.Simulate and running at a user controlled timestep like ml-agents does

your results will not be valid the performance benefit of parallel fitness is outweighed by the costs of noise in the simulation

you will not know if it converges at all

because you're introducing too many buggy situations

[lavish meteor]

So you think that I either have to completely simulate it outside of unity? Or do you think I should simulate each agent at a time?

[undone coral]

you simulate 1 agent at a time

is your fitness function authored in c#, living inside unity?

meaning

is the whole schedule of steps for GA happening inside unity

[lavish meteor]

yes

[undone coral]

or is it running somewhere else, and delegating parts of the fitness function to unity

okay

the performance benefit of parallel computation by simulating multiple candidates in a fitness function at a time* is outweighed by the costs of noise in the simulation
do you understand what i am saying here?

lemme improve that

[lavish meteor]

But why is the fitness function the bad thing? By fitness function you mean the function that determines how fit an agent is by calculating how far it came and how fast it got there? I was more worried about the neural networks, controlling the car

[undone coral]

hmm, in a GA architecture, the fitness function is putting the car on the track, running its neural network, letting it make decisions, until a certain amount of time has elapsed

and then measuring the distance it traveled in that amount of time (really the distance left from the goal, if you are minimizing)

and you choose a very long amount of time in the beginning, and make it gradually shorter, for the sake of solving this particular problem

"calculating how far it came and how fast it got there" - in abstract, that could mean running the neural network and driving the cars. is that what you mean?

or is the content of your fitness function

double Fitness(Scene scene) {
 return (scene.GetGameObjectByName("goal").transform.position - 
  scene.GetGameObjectByName("agent").transform.position).magnitude;
}

[lavish meteor]

So everything besides running the neural network is called before the simulation... Previous cars will be despawned, their fitness will be calculated by my custom progress script, All cars will be spawned, their parameters will be altered

[undone coral]

because it should be

double Fitness(Genotype<BitGene> thisAgent) {
 var scene = new Scene();
 Instantiate(Car)...
 // run the simluation
 ...
 return (scene.GetGameObjectByName("goal").transform.position - 
  scene.GetGameObjectByName("agent").transform.position).magnitude;
}

[lavish meteor]

No. I have a progress script which calculates how far the car drove along the track by using trigonometry

[undone coral]

okay well a fitness function has a signature of the form static double Fitness(Genotype oneAgent)

do you have that?

[lavish meteor]

I'm not done with my genetic algorithm function. I'm still spawning in the cars which proves to be more difficult than I thought

So no

[undone coral]

gotchyu

okay have you looked at https://github.com/giacomelli/GeneticSharp

or the java jenetics library

this is how you express a genetic algorithm setup

it sounds like you should look at openai gym too

you haven't gotten far enough to not start over 🙂

i think you should delegate the fitness step to unity

and do everything else outside of unity

because you are going to have multiple unity processes - only for the sake of parallel computing of fitness

there is no other reason

since you are trying to use neural agents... i'm not sure if you have an opinion on how to "mix" neural agents

or what their inputs are

there's a lot of open questions here. i don't know if you're referencing specific literature

@lavish meteor i would make my priority trying to get this running at least once. both the uppermost thing - the definition of your GA problem, in a form that resembles genetics, and proving for a simple problem that you can get something to converge

[lavish meteor]

Their inputs are raycasts which determine the distance to the edge of the track

[undone coral]

then, for your actual racing problem, correctly implement double Fitness(IChromosome car) and prove that it runs correctly

[sand acorn]

Hello there, I have a voxel generation system / minecraft like terrain generator which works fine, but I would like to render a face on the chunk side if lets say the loop in x value is equal to a coordinate / the chunk is a border chunk; The problem I have here is that since all the chunks are generated independently and all chunk have the same block ranges (0 to 16) ,iso f I want to check if X is equal to 0, it will make a face where all the other chunk value is 0 too making unwanted faces, so how can I over come this? Here is an example of what happens

[undone coral]

i'm not sure if GA makes sense here

since you're using an "actor-critic learning algorithm", you would only be running 1 agent at a time. it sounds like you have that working. does it converge?

do you have a Unity Recording package recording of the car racing the track?

btw, i believe PPO and more is in ml-agents & openai's gym RL framework already

and it would be worth comparing your result to theirs for correctness

[lavish meteor]

Genetic algorithms is the only thing missing. As I said I wanted to provide an own algorithm for both genetic and actor-critic which may or may not be 'ill-architected'. My solutions probably aren't as good as GeneticSharp or MLAgents but if I don't try to solve the problems myself, there is nothing to be gained here

[rocky grove]

Hah, that might be what I was looking for. Thanks, I'll give it a shot!

[undone coral]

well i think you'll figure it out

gotta go

[undone coral]

what i am really saying is, based on this answer, you are not 100% serious about achieving this, you are at like 20-50% seriousness

like if it does or doesn't work doesn't really matter to you

that was the very first question i asked for a reason

[lavish meteor]

Actor critic will be just one agent. Genetic could be altered so there is also just 1 agent at a time but the player later gets shown multiple agents

[undone coral]

it just sounds like you don't know what GA is yet

or how to use it

and actor critic, like i said - do you have a video of such a model trained and working?

you veered off again

[lavish meteor]

I do... So much so that I want to achieve it myself

[undone coral]

just answer the actual questions

no more of this, coping by veering off. you made some big architectural mistakes

so it goes

i make mistakes too

do you have a video of the thing completing a lap correctly?

[lavish meteor]

No

[undone coral]

so

it sounds like it doesn't work yet

that's the honest reality

[lavish meteor]

It doesn't

[undone coral]

okay

[lavish meteor]

Not saying I don't have that algorithm. It just doesn't work

[undone coral]

yeah

well if i were trying to learn this

i would study the existing thing, or use the existing thing

i would at least use it for the high level, how is this supposed to look when it is right

[lavish meteor]

That's you

[undone coral]

in the same way that looking at the snippets from Genetics will not implement GA for you

but it will at least tell you what the signatures should be, which i know is important to people who are good programmers

it's all made way harder by trying to do this in C#

the libraries for implementing a neural network are not mature enough

[lavish meteor]

I know. I could probably use hundreds of libraries if I were to do this in python

[undone coral]

i would try to author it in python first and operate unity via an RPC

which incidentally is exactly how ml agents works

[lavish meteor]

But I really don't want to

[undone coral]

okay i really gotta go

[lavish meteor]

👍

[undone coral]

when i was implementing stuff like this, 3,000 years ago, i was so worried about screwing up the transformation from the math in the paper to the math in python that i didn't even work in logspace.

[lavish meteor]

Yes, it's by far the hardest part. A few weeks ago I was just figuring out the math behind it (got stuck at the gradient descent) when I realised it wasn't applicable to my problem. I was trying to do everything realtime (So I wasn't working with episodes but rather calculated the loss every timestep) which of course wont work for school computers. Instead of deleting my progress (Like I did when I tried with Q-Learning) I rather started coding a Genetic Algorithm and then try to apply new things I hope I'll learn doing so... Look, I really respect your knowledge about this but I want to learn this via trial and error

[lavish meteor]

I do. What I was talking about is, instead of spawning all cars at once I would rather only spawn a small amount or even just one at a time and then put it all together and look for the fittest agent. That way I would avoid the bugs you were talking about and by simulating the whole thing I would achieve equal times

[devout coyote]

It would 10x your run time though

Using a GA to optimize the parameters of a neural network is quite an interesting approach, I am really curious to what results you will get

[undone coral]

you're just biting off a lot of problems all at once

if you want to author neural networks from scratch, which is what it sounds like

that alone is a problem set you should expect to spend 10-20h on

i wouldn't try to implement them inside of unity & C#

it's too janky

[lavish meteor]

The neural networks are working. Even backpropagation is working. I did need a lot of time for that, that's true. Doing it in C# was a bold choice but I didn't want to commit to a new programming language

[devout coyote]

Since speed is of concern (goal is to run it on school pcs right) are your implementations fast?

[lavish meteor]

Which implementations?

[undone coral]

are you doing this within a framework where you can prove it works?

like do you have a successful MNIST result?

[lavish meteor]

No. I did calculate it via a calculator to ensure I have the outputs I should have. Regarding backpropagation I have no idea... If I wanted my change my weights with respect to certain outputs that should occur on certain inputs it works like it should

I have tested both functions. I don't have a clue if they are as accurate as they should be

I also made a bigger batch where I made ChatGPT code a function which outputs every calculation done to ensure there were no errors

[timber flame]

Do you prefer to use Single instead of First, SingleOrDefault, FirstOrDefault when you want to get a specific element/item from a collection while it should be in?
For instance, for assets, collections with specific elemenets in (editor). They are immutable.

[undone coral]

I did calculate it via a calculator
what do you mean?

are you doing this within the context of a class?

or some other learning material?

[lavish meteor]

Your question was if my nn and backpropagation have accurate outputs right?

[lavish meteor]

Via a school calculator... blame me all you want but I wanted to be sure about it working as it should

[dusk bloom]

Currey was my attempt to debug. It doesn't work only using the currentPointIndex because that checks the travelPoint ahead of the player if they were going forward, but Currey Was meant to check the one they just were at so that if they were to go backwards through it, it would catch it too, and decrease the travelPoint to the one before that, but this also didn't work.
I am going to put this in general now

[undone coral]

you gotta fight this instinct to not answer questions

if you want to succeed at this

[undone coral]

are you doing this within the context of a class?

or some other learning material?

a school calculator
i just don't understand what this means. what concretely is a school calculator?

[lavish meteor]

Yes

[undone coral]

the class had a problem set about implementing a neural network? can you link me to it?

traditionally you test your implementation against MNIST, where you should get like 99% accuracy

[lavish meteor]

Something you use for math

Like in school

[undone coral]

okay well i think i've reached the end of my support here

i don't really know what the context is, but it doesn't sound very serious

[lavish meteor]

Like I said I also did it in the context of my class

Via logging all calculations

[undone coral]

the traditional way to test if your implementation of a neural network is correct is by solving the MNIST digit recognition problem

[lavish meteor]

I will do that and tell you what I got

[undone coral]

it just sounds like the class has flaws

i am surprised this is how the class taught the implementation of a neural network

this is in a university?

[lavish meteor]

No its 11th grade in germany

[undone coral]

hmm

even then it would be more serious

[lavish meteor]

We currently have html in our class

[undone coral]

as long as it's not an online course

well i think it's all very interesting and you will have a long journey ahead of you

[lavish meteor]

The class teaches us nothing...

I tried learning it all by myself... we even skipped machine learning

[dusty wigeon]

Machine Learning is something you usually learn in University. So I would not blame them to not teach you that.

[devout coyote]

I didn't start learning about ML until my masters degree :s

And I think grade 11 is like highschool ish?

[lavish meteor]

I don't blame em. It's just kind of hard to not get preyed on because my knowledge in this field isn't as deep as someone who does this for his masters degree

[lavish meteor]

Yes. We have elementary school from 1st-4th grade and secondary school (In my case the 'Gymnasium') from 5th to 12th grade

[devout coyote]

You can honestly be very proud of yourself for figuring all of this stuff out, I had to take quite a couple of extra math classes before I could grasp the basics of building my own ML algos

[dusty wigeon]

The issue, is that you are trying to get to the "juicy" part without understanding the underlaying concept that you require. You should start with simple lesson and build from there while also search for the knowledge you are missing. Otherwise, you are going to think you know a lot where in fact you little to nothing because you did not take the time to truly understand what you are doing.

One advice that I can give you is to NOT use Unity as your framework because you now need to understand how Unity works and how Machine Learning work while also not being able to follow any tutorial on ML because most of them are done in Python. Also, you should try to make it work with a know library before doing your own implementation so you can compare correctly.

[lavish meteor]

I do understand the underlaying concept... I'm just bad at finding and understanding the right terms for certain concepts. I'm learning ml since longer than a year and I did start at finding minimizing functions etc. This whole thread about my project is completely misinterpreted. I don't have any problems with my algorithms and if I did I would get help in associated forums. I just wanted to know whether or not there is a built in way to simulate physics behaviour, so I can minimize performance issues...

But yes I didn't know about things like MNIST...

[dusty wigeon]

I mean, you are expecting to train an AI live. I just hope your environment is super minimalist, because no amount of optimization will make you able to get to a relatively fast convergence.

[lavish meteor]

It is... There are actually a lot of yt videos using a similar approach and setup so I'm confident that it'll work if I implement it correctly. Also not using python wasn't that bad in regards of properly learning and understanding ml... If I wanted to implement something I couldn't reference it from existing code but had to understand core concepts and put them into code. If my goal was to make progress as fast as possible it would have definitely been the wrong choice

And I'm not expecting to train it live (if you mean realtime?). That's what my original question is all about

[dusty wigeon]

I need the engine for the game. Basically the player sets his values (like learning rate, discount factor etc) and presses on "train ai" or smth. Then the AI would be trained without any visual feedback but every output would be stored in some kind of List so the player can see the learning process after the training. That way he could speed it up, skip it etc. It could run externally but even with this simple of a game there is so much to simulate

Is there anything I did not understood there ? Pretty such you meant to train your AI inside the Unity Player.

How much time do you expect your training to take ? Most of the video on youtube are made with trained agent off camera. Then they show you the result with visualization tool. It can takes only 100 iterations for simple environment, but some can take more than 100'000 to get to convergence.

[lavish meteor]

That's pretty much what I'm trying to do. If it is too long of a waiting time I'll have to pregenerate a few samples or let the teacher do it, in which case the project failed. I'm just trying to minimize this waiting time as much as possible

[waxen adder]

Can anyone explain to me what a Quaternion.Euler is

[dusty wigeon]

Anyway, good luck. And for your performance, you basically just need to remove the camera.

[sly grove]

It's a function which returns an equivalent Quaternion for the provided Euler angles

[waxen adder]

xRotation = 0f;

cam.transform.localRotation = Quaternion.Euler(xRotation, 0, 0):

What would this return

[sly grove]

The 0,0,0 rotation

Aka Quaternion.identity

[waxen adder]

So what is Quaternion though

[sly grove]

It's a variable that represents a rotation

That's all you really need to care about

[waxen adder]

Oh ok thanks

[fickle inlet]

Hey guys, I am doing a research on Time.deltaTime. Since Time.Del

Since Time.deltaTime varies every frame, what issues does it bring?

Like, what could go wrong with Time.deltaTime when moving an object, animating something, etc

I am trying to have this deep understanding of Time.deltaTIme

Also, another question that I have regarding the deltatime, since delta is changing every frame, how does the object end up at the same distance, if delta time is changed every frame?

[alpine sinew]

@exotic elm https://docs.unity3d.com/ScriptReference/Time-deltaTime.html

it's literally just the time that passed since the last frame

as to what can go wrong -- it depends on how you use it.

[fickle inlet]

I understand how it works per frame, but I am looking more for limitations that devs experience when using time.deltatime, stuff that you can't find in the documentation

Like for an example, using it in multiplayer and each client is sending different time interval and stuff like that

[alpine sinew]

oh yeah, so you've already got an understanding then. in that scenario, the server or host would definitely want to be using FixedUpdate() instead of Update() so that all of the clients are forced to synchronize to one clock...

there's not really any "limitations". It just depends on what you want to use it for. It's kind of like asking "what are the limitations of using my computer to program?" like, the answer would be "well... compared to what else? I don't know of any alternatives... and also, what's wrong with the computer in the first place?"

[fickle inlet]

But do you know some common mistakes that new devs do when they use Time.deltaTime. Like some bad practices that they do unaware?

I can list a few, like objects going through walls when low fps and the wall is too thin. Then the networking issue that I wrote above and also animation looking jittery on low fps. But I wanna know if there are more that I don't know

Just listing things so that you know I did some research on my own, but wanna expand it even further

[dusty wigeon]

To be honest, there is not a lot of magic going with Time.deltaTime. The only thing you need to ask yourself is if you are working with the correct unit whenever you are doing an operation.

By example, if you do transform.positon += velocity, you should realize that it is wrong because you are adding a velocity to a position. Instead, you should do transform.position += velocity * Time.deltaTime. The only bad "pratice" is whenever someone does not use the correct unit for the given task.

Note that in mutliplayer, you do not need to synchronize the players through time (with FixedDeltaTime). In fact, that would be a bad "pratice" from my experience.

[hexed meteor]

Would someone be able to tell me why the 'execute' bool is mysteriously set to true when I'm exiting play mode? Makes absolutely no sense. Did I break Unity again?

[dusty wigeon]

What is the serialize value of the execute boolean ?

[hexed meteor]

false

[dusty wigeon]

are you sure ?

[hexed meteor]

It's always false... it's very spooky

yeah!

[dusty wigeon]

Is it serialize as false though

[hexed meteor]

It's false in the inspector

I guess I'm not sure what serialized means

in the script above it's false as per default, and in my inspector it's also false

[dusty wigeon]

I suspect that the value is serialize as true than override because of the [ExecuteInEditMode]

[alpine sinew]

am I wrong to assume that the [SerializeField] tag stores the value in memory? And therefore, whatever the value was when you exited, it will be the same when you enter? Regardless of assigning the parameter in the top of the class...

@hexed meteor I'm pretty sure, unless i'm wrong on that (^) that you'll just need to assign it each and every time you run the script.

so assign it inside Awake() or OnValidate()

(I'm pretty sure OnValidate will work here)

[alpine sinew]

Serialized means that the value is stored in memory outside of runtime

[undone coral]

don't use a checkbox to make something happen this way

the whole script is flawed

[patent bear]

This isn't quite corrent on a MonoBehaviour. [SerializeField] marks a property to be initialized to a value saved to disk. The initial value is saved to the scene/prefab file (or sits in memory waiting to be flushed). This is a separate value than the actual runtime value of the variable. When exiting playmode, Unity resets the value

[undone coral]

if you want to do something in the editor from a component, and you don't want to write a custom editor

[patent bear]

That's why changes in play mode don't get saved

[undone coral]

use [ContextMenu], then click the ... button in the component's title

i don't know what you guys are discussing regarding serializefield here, because it's all irrelevant in the context of the flaws of this script

you can find an example of [ContextMenu] in google

[patent bear]

Agreed, just wanted to correct the statement that the [SerializedField] is saved during play mode which isn't true or, as you mention, relevant in this case.

[plush hare]

In multiplayer, each client is sending stuff at a fixed interval. Aka the tickrate. So I'm not sure what you're getting at

Generally you make a clock, which uses the deltaTime like you were asking about earlier.

timeAccumulator += time.deltaTime;
while (timeAccumulator >= tickDelta) {
timeAccumulator -= tickDelta;
// tick
}

[fickle inlet]

I am no expert in multiplayer, but is the code example something that is provided by default by developers who make the networking, or is it something that a beginner has to build himself and has to understand? Could a beginner developer not know what is tickrate and then he just uses time.deltattime that he synces over the network and then faces the inaccurate issue.

[plush hare]

I don't know what framework you're using, but there shouldn't be a reason to use time.deltatime in a network simulation user-code assuming it's running at a fixed frequency

I'm assuming we're talking about server authoritative simulations

with rollback and such

[fickle inlet]

Yes, I understand that, but is it a common mistake that a beginner does not know about fixed tick rate and he just goes with time.deltatime?

[plush hare]

those two dont correlate

so your question doesn't make sense

When we're doing normal movement code in a unity simulation, most noobs are going to put it in the update() loop. This update loop is highly variable, and therefore if I moved a character by 5 meters every frame, his speed would be dependent on the frame rate

A fixed simulation loop is not variable

it always ticks n times per second unless your hardware cant keep up

[fickle inlet]

by fixed simulation loop, are you talking about something custom made by networking frameworks, or do you mean fixedupdate loop?

[plush hare]

network engine loop

[fickle inlet]

does unet have that?

[plush hare]

no idea

This is generally why it's good to understand how these things work under the hood

at least on a basic level.

before you go dedicating your time making a game in it

[fickle inlet]

Yes that is what I am doing, I am trying to find common mistakes with Time.deltaTime

[plush hare]

UNet is deprecated btw

and you probably shouldn't use it

[fickle inlet]

Mirror is based on UNet

I don't know if it is still maintained by an open source community

haven't used it in years

[plush hare]

What type of game are you making

[fickle inlet]

But anyway my questions were more regarding time.deltatime than networking.

I am researching basic unity stuff and trying to learn every bit of it

I used time.deltatime for years

but never went into depth with it. I know it makes moving fps indenpendent, but still wanted to research it deeply and be able to explain it to someone

[plush hare]

It's really not that deep

If I want my player to move 5 meters per second, and my simulation is already ticking at 30hz (30 times per second), then I can just multiply 5 by the delta time

5 * 0.0333

that way it only moves a fraction of the velocity every simulation tick

until it builds up to 100% after 1 second

[fickle inlet]

Right now I am searching when Time.deltaTime could be used incorrectly as it has variable time intervals and could effect things in the game

[plush hare]

Using time.deltaTime in a fixed update loop would be improper use.

You're spending too much time on things that really aren't important

there are better things to learn

like writing netcode

[dusty wigeon]

The correct way to interpret this would be so that you "convert" velocity into displacement. (m/s * s = m)

[plush hare]

All deltaTime is doing, is returning the time since the last render frame. It's really not that deep

[alpine sinew]

deltaTime is such a basic concept without really much to know about it... only thing to know is, like, when to use it. Like, you're trying to build a house but obsessing about what type of hammer to use

[regal olive]

If I wanted to measure the distance from an object to the ground using a raycast, what would be the code?

[fickle inlet]

When you put time.deltatime into a fixedupdate it returns time.fixeddeltatime

[plush hare]

no

it does not

it still returns deltatime

[dusty wigeon]

Both value are equals in this situation.

[plush hare]

wrong

[alpine sinew]

burt is on fire lol

[regal olive]

If I wanted to measure the distance from an object to the ground using a raycast, what would be the code?

[plush hare]

one is the render time, other is the fixed simulation time

[dusty wigeon]

Both return the same value.

[plush hare]

they do not

[dusty wigeon]

If you use it in FixedUpdate

[plush hare]

if they do, then that's really dumb and shouldn't happen

[alpine sinew]

if i have a fixedupdate of 0.01sec, but a LOT of shit happened inbetween, there's the chance that it could lag and my deltaTime != 0.01 ... no?

[plush hare]

Generally you put all of your simulation code inside of a fixed clock, and then interpolate the visual state between simulation ticks

[regal olive]

If I wanted to measure the distance from an object to the ground using a raycast, what would be the code?

[dusty wigeon]

Pretty sure they do. There is no reason to have different value.

[plush hare]

I don't think you understand how this works

[fickle inlet]

Write this Debug.Log(Time.deltaTime) inside a fixedUpdate and tell me what value it returns?

[alpine sinew]

#archived-code-advanced message

[regal olive]

If I wanted to measure the distance from an object to the ground using a raycast, what would be the code?

[alpine sinew]

#archived-code-advanced message

[dusty wigeon]

I sure do. There is no reason to have different value betwen Time.deltaTime and Time.fixedDeltaTime inside FixedUpdate();

[fickle inlet]

Cam you elaborate more?

[dusty wigeon]

Both represent the same concept

[plush hare]

Not for the advanced section, but just use the distance property of the hitResult

[plush hare]

Gimme a sec to test this out

[regal olive]

how do i do that

[plush hare]

If you're asking this, then you shouldn't be in the advanced section

nobody here is going to write code for you

[regal olive]

im making a vtol aircraft and the hover height needs to be 2 units from the ground

[alpine sinew]

check out #archived-code-general

[regal olive]

what would i ask

[dusty wigeon]

Not sure I understand your question. If you use FixedUpdate, Unity is going to make more FixedUpdate than there is Update. (If you lag)

[alpine sinew]

#archived-code-advanced is for people to argue about the differences between deltaTime

[plush hare]

Ok. time.deltaTime returns the fixed delta clock inside fixedupdate

[fickle inlet]

These are discussions I like, I love these specific things that devs don't normally talk about.

[plush hare]

don't know what the purpose of that is

but I guess its there if you need it

[alpine sinew]

this is so dumb

what if I WANT the un-fixed deltaTime within the fixed update?

[fickle inlet]

Test it

[alpine sinew]

like if i need to determine the lag

visual* lag

[dusty wigeon]

It is not about the purpose. It just make sense. Time.deltaTime is the duration between two frame. So, if you call it in FixedUpdate, this is the duration between two FixedUpdate.

[plush hare]

fixedupdate isn't always a render frame

time.fixedDeltaTime exists for a reason

[fickle inlet]

FixedUpdate is a physics timestep which by default is 0.02

[dusty wigeon]

Time.deltaTime is not about render frame. It is about the difference between two Update.

[fickle inlet]

so it updates every 0.02 secds

[plush hare]

When someone says "frame" I assume render frame

not a simulation tick

but alright

it is what it is

[fickle inlet]

Time.deltaTime is difference between last and current frame

[plush hare]

I'm aware buddy

[fickle inlet]

Inside an Update loop

[alpine sinew]

When you call Time.deltaTime in FixedUpdate, it automatically calls Time.fixedDeltaTime instead. Just a little syntactic sugar to help.

[fickle inlet]

Btw, official unity documentation

https://docs.unity3d.com/ScriptReference/Time-deltaTime.html

[alpine sinew]

so i can call deltaTime, or fixedDeltaTime and get the same result --- how the hell do I get the deltaTime within the FixedUpdate then?

[plush hare]

you write your own clock

[dusty wigeon]

It makes no sense to call deltaTime in FixedUpdate

That is what I am trying to say.

[alpine sinew]

I'm sure there's some use-case

[dusty wigeon]

You can have 3 FixedUpdate in the same frame.

[fickle inlet]

fixed update doesnt run based on frame cycles

[dusty wigeon]

No, there is no use case for that.

[plush hare]

It does.

[dusty wigeon]

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

[plush hare]

Unity's fixed update runs on a render frame, just not always every render frame.

[dusty wigeon]

If you were to use the Time.deltaTime of the current frame in FixedUpdate, you will be using the wrong time value. (x times the same value which can be larger or smaller than the actual real value of the simulation)

[plush hare]

Everyone understands that

It's just weird to force that kind of thing

[alpine sinew]

EXACTLY. so maybe you want that to determine how much your simulation differs from the updates

[austere jewel]

What's the actual conflict here, this is a long discussion over something that is pretty factual, though I understand it's a convoluted topic when it comes to specifics

[plush hare]

no conflict

just trying to figure out why unity forces time.deltaTime to return time.fixedDeltaTime in a fixed update loop

[austere jewel]

I just see a few claims, Time.deltaTime in FixedUpdate returns fixedDeltaTime, you can test it yourself

[plush hare]

you'd assume the programmer knows the difference

[austere jewel]

It's for ease of use, you can share code between the two and nothing weird will happen

[alpine sinew]

no conflict. Just want to know how to get deltaTime (non-fixed) within the FixedUpdate() now

[dusty wigeon]

People are upset that Time.deltaTime = Time.fixedDeltaTime in FixedUpdate while I think it makes prefect sense.

[austere jewel]

you don't need to detect which type of update you're running

[austere jewel]

Cache it from Update

[plush hare]

I mean I guess that works

as an answer

[austere jewel]

Though realise it's not a very sensible thing to be doing

[dusty wigeon]

Update is Run after FixedUpdate, so you will need a bit more of work.

[fickle inlet]

Look what I found

`Variable time step:

I’ll put this in here as an option in the solution space with the caveat that most game developers I know recommend against it. It’s good to remember why it’s a bad idea, though.

It adapts to playing both too slowly and too fast. If the game can’t keep up with real time, it will just take larger and larger time steps until it does.

It makes gameplay non-deterministic and unstable. And this is the real problem, of course. Physics and networking in particular become much harder with a variable time step.`

Is this true for Unity too?

[plush hare]

You've just posted what you already know

If you don't run your code at a fixed timestep, or scale it using deltaTime, then it won't be deterministic

determinism means that it "always behaves the same" in laymans terms

You're thinking way too hard about this

[fickle inlet]

Just learning how game engine and game loops works

[plush hare]

Tell me what you're confused about

and I'll do my best to explain it to you

[plush hare]

    Using time.DeltaTime for movement:
        - Player will always move at same rate. Movement is "deterministic".

    Not using time.DeltaTime:
        - Players with lower framerates will move slower. Players with faster framerates will move faster.

Fixed update loop (30fps (never changes))

    Using time.DeltaTime for movement:
        - Player will always move at same rate. 

    Not using time.DeltaTime for movement:
        - Player will always move at same rate. 
        - Player's "velocity" will not be broken up over 1 second. Will probably move faster than above, but still "deterministic" movement. Will always move at the same rate.```

[fickle inlet]

I understand how it works

[plush hare]

I don't get what you're asking then

This is just pure mathematics and logic

[fickle inlet]

Ah nothing, it just clicked to me now 😄

I was searching for cases where Time.deltaTime would not good idea to use. But only those cases where it is not obvious at first. I know I am not gonna use Time.deltaTime if I want something to occur every frame, because that is something that you can easily see. But I am looking more for those invisible cases, where it looks all god, but is causing an issue that you can see only by strict logging.

I don't know if this is a type of a question appropriate to ask here

[plush hare]

Mouse movements, for example, should not use time.deltaTime

you should just use the raw values

I know I am not gonna use Time.deltaTime if I want something to occur every frame,
You also SHOULD use time.deltaTime for something every render frame movement related

I'm trying to think of other scenarios

[fickle inlet]

The only I know if you are using collision it can go over the wall if fps is lower because it is doing fewer steps

and the distance(time) between steps is bigger

[plush hare]

yeah if the distance is massive then you'll tunnel through walls

[fickle inlet]

what happens when you do use delta time for mouse?

did you try it?

[plush hare]

No, I just understand how delta time works

which you seem to be struggling with

Mouse movements are updated every frame. It gives us the delta from the last render frame

so if I move my mouse 10 degrees left, I'm going to instantly update my rotation or whatever in 1 single frame, 10 degrees left

The point of delta time, mainly, is to split up a velocity into multiple chunks

When I say "my player movement is 5", that generally means he's going to move 5 meters in 1 second

[fickle inlet]

But mu question was more like, what would happen if you did apply timedeltatime to mouse position? would mouse move slower?

[plush hare]

yes

and would probably be very unpredictable

and feel awful

[tiny pewter]

Time.deltatime is applied for continuous movement by normalize it, so for something that expected to have a suddenly change, dont use it

[plush hare]

Applying deltaTime to your mouse movement would do the opposite to applying it in movement.

Lower FPS would make your mouse faster. Higher FPS would make your mouse slower

you definitely do not want that

at least I think that's what would happen

[torn shuttle]

https://gdl.space/ocafenujuj.cpp
I press w in the air while jumping then I go across my entire 20x1x10 plane and fall. How do I stop this?

[plush hare]

You quite literally just posted this in beginner

[fickle inlet]

Just wanted to know what would happen since you pointed it out 😄

[torn shuttle]

no one answering it

[austere jewel]

Do not cross-post, this is also not an advanced question. If you're not getting an answer, reply to your message with more context.

[fickle inlet]

Do you know what advantages would time.deltatime have over time.fixeddeltatime?

Like some case where it's just no brainer

When I think about it fixed is better for everything because it is more stable/precise

[plush hare]

You're still not understanding

time.deltaTime and time.fixedDeltaTime are two different things

They are only the same when called inside the FixedUpdate() function

Time.deltaTime = time since last render frame
time.fixedDeltaTime = time since last FixedUpdate call

[fickle inlet]

I do understand differences

[plush hare]

If you're only calling code inside FixedUpdate(), you don't even need to realistically scale by delta time

it will still work properly

Aside from that, it really has nothing to do with your game being stable or precise

Writing a stable game involves writing good code that doesn't crash

[austere jewel]

(The value to scaling by a delta in FixedUpdate is that if you changed the fixed time interval, your logic would stay the same, and that your values would reflect real-world distances, but often physics methods already take this into account)

[fickle inlet]

I do understand differences, but I am thinking about use cases. Moving any object would be better in fixed update than in update, even though you can do it an update too? Is this true?

[plush hare]

It's easier on the CPU

rather than forcing your CPU to perform collision resolution etc 144+ times per second, you can do it 30 times per second

and interpolate on the render frames in between

Whether or not you want to go through the work to do this, is up to you

because it can be a lot of work. You'll probably end up writing your own library to use

https://docs.unity3d.com/ScriptReference/Physics.Simulate.html

This is an example of a clock. It goes inside the Update() loop.

It does the same thing as FixedUpdate does

ignore the Physics.Simulate(Time.fixedDeltaTime); part for now. Just look at the code example

[fickle inlet]

That is out of my scope for now 🙂

[plush hare]

yeah

its alright

Give your brain a rest. When you wake up, deltaTime should make more sense.

[fickle inlet]

😄

I truly do understand it, just needed more depth

This was a really good read if anyone needs it https://gameprogrammingpatterns.com/game-loop.html

[plush hare]

float timer;
float tickRate = 30; // Execute our custom loop 30 times per second.
float myFixedDeltaTime = //  (1 / tickRate) will be 0.033 at 30hz tickrate. This is the amount of seconds between each tick.

public void Update() {

        // Timer starts at 0. Every frame, we add the seconds since last frame. Like below.
        timer += Time.deltaTime;

        // Is our clock ready to tick? While loop prevents the clock from falling behind. Could execute multiple times in a single update() just like unity's fixedupdate().
        while (timer >= myFixedDeltaTime )
        {
            // Take away the delta time that we just consumed
            timer -= myFixedDeltaTime ;
            MyCustomFixedUpdate();
        }
}

public void MyCustomFixedUpdate() {
 Debug.Log("I'm running 30 times per second!");
}```

This is how unity's fixed update works btw.

Let me know if you need me to explain anything

You can compute the delta time by doing 1 / tickRate

so 1 / 30 = 0.033... that means every 0.033 seconds, we should run MyCustomFixedUpdate

[fickle inlet]

Looks interesting

Didn't know you could do something like that

is unity fixedupdate working similarly?

[plush hare]

You'll notice that I'm using both time.deltaTime and time.fixedDeltaTime

yes

[fickle inlet]

I did notice

[plush hare]

there's nothing special about their fixed update

When you think of a "clock" you probably think of a perfect clock. Like down to the nanoseconds

but it doesn't matter in this example. The fixed update loop doesn't need to be that precise

[fickle inlet]

I will debug this tomorrow

Really good for learning

can you use if statement too?

[plush hare]

Where?

if timer >= fixedtime... ?

[fickle inlet]

while (timer >= Time.fixedDeltaTime)
{
// Take away the delta time that we just consumed for the tick so we're not stuck in an infinite loop
timer -= Time.fixedDeltaTime;
MyCustomFixedUpdate();
}

[plush hare]

So you're asking if we can replace the while with if?

[fickle inlet]

yes

[plush hare]

You wouldn't. If you get a big lagspike, and your deltaTime was like 2 or something

then you'd have trouble

        {
            // Take away the delta time that we just consumed for the tick so we're not stuck in an infinite loop
            timer -= Time.fixedDeltaTime;
            MyCustomFixedUpdate();
        }```

If the timer was at 2 here, and our fixedDeltaTime is 0.033

it will keep running the loop in order to catch up

timer -= Time.fixedDeltaTime;

remember that we're decrementing on every tick

This is very simple btw. I feel like you've been up for hours and your brain is mush

[fickle inlet]

I'll run logs tomorrow, too tired now

[plush hare]

yep. Ping me tomorrow if you need help

[fickle inlet]

Sure, thanks!

[regal olive]

Do you know any asset or code to compile code in runtime in iOS? In Windows I do that by having an asset bundle of the code and doing assembly.Load but it doesnt work for iOS because the ahead of time compilation of il2cpp. Has someone tried Roslyn C# in iOS?

[craggy kindle]

I need some help regarding C# namespaces.
I'm making an online game that has some common code between the client and the server, so I have a common dir that is synchronized between the two. I'm using Jetbrains Rider which suggests that the namespace in each file should correspond to the path of the file, relative the root of the project. The problem is that this shared code appears in 2 different paths (in the client and in the server), which means that the namespace must not match its path in atleast one of them.
What is the correct way to approach this? I don't want to add // ReSharper disable once CheckNamespace in every new file in the common directory.

[dusty wigeon]

Make a third namespace for Common/Core/Shared for every file that is shared between both ?

[fierce kayak]

i got this error and i cant do anything with this here

[craggy kindle]

But how do I make Rider not warn me about that path?

[austere jewel]

Did you comment out the other line of code with the issue? Or is that the only error now

[fierce kayak]

oh i havent done that yet but the entry point error is something different

it doesnt include a script in the error

[austere jewel]

please keep to one channel. We were going over that there

[fierce kayak]

ok

[dusty wigeon]

I do not think it is possible to compile at runtime on device that only support Ahead of Time. If you want to add additional content, you should use DLL/LIB/SO/A files. Interpreted language is also a possibility.

[sage radish]

iOS doesn't allow it. It's the reason IL2CPP exists in the first place. The most you can do is interpreted languages, like Lua. There are some .NET interpreters out there too, but they are slow.

[dusty wigeon]

As I said, make a third folder that contains all the code that is shared between your Server/Client. Then rename the namespace to reflect the path of your folder.

[craggy kindle]

Do you mean that the folder should be outside of these projects?

because as long as it's inside then the path of the files will be different

[dusty wigeon]

You should have a third project. If you really want that. Alternatively, you can have 1 single project name X where you have X.Client, X.Server and X.Common. (Which would probably make a lot more sense)

[craggy kindle]

I believe a single project would be problematic. I'll try to use a third project, thanks 🙂

[regal olive]

correct me if Im wrong, but does are windows files, and linux. Is there anything similar to a dll in Windows for iOS?

[dusty wigeon]

A quick google search give me: *.dylib and *.a as the corresponding format.

[iron torrent]

Hi Peeps. I am in dire need of some help here.

I have been working on a side project concept for sometime now, and I have the idea working in a non URP project, and I have it working in a URP project. However, I can not get it to work in URP with Cinemachine, and I do not believe Unity Support are looking at this issue correctly.

The idea.

I am trying to create a solution where the player appears to be in a side scrolling world, where the world wraps and the player is under the impression it is continuous. I have it working with 3 Cameras. The main Camera, a left and right camera. That move without clearing the Depth Flag, and gives the left and right camera the illusion that they overlap on the two ends.

Now, if I move the Main Camera myself so that it follows the player, in either a Non URP or URP project, I have no issues. The player is moving smoothly and the Cameras are all working as it should be working. However, I when I put Cinemachine into the mix, and remove the script to follow the player manually. It doesn't work.

Unity Support is saying this is by design, I am questioning how this can be by design.

[obsidian glade]

this seems a bit vague, do you have some example images/videos for what "working" and "not working" means?

[dusty wigeon]

I am not sure I understand the concept of your Application. Like Deynai, you will need to provide Video of what is "working" and what is "not working".

Also:

• What is by design ? What is the actual conversation with the support ? You may have miss some critical information,
• What Cinemachine does that influence your mechanic ?
• Do you need Cinemachine ?
• What URP have to do with any of this ?

[undone coral]

it sounds like you need ot learn how cinemachine works

in any case, you probably want to create a hierarchy like

Static (real) World
 - Other objects that repeat
Wrapping (modulo) World
 - Player
 - Camera
 - Other objects that don't repeat

let's say the region of the world that repeats is from position x=0 (lower left pivot), size 0 to B, for border.
create prefabs to cover x=-B..0 and x=B..2B
then, the world space position of Wrapping (modulo) world is
wrapping world.x = B*Mathf.floor(player local position.x / B)
that's it

it doesn't make sense to use 3 cameras, a depth flag, etc.

the position of everything in the "wrapping" world in that local space is always in natural, "always going forward" or "always going backward" coordinates.

which is what you need anyway

then you can use virtual cameras inside that hierarchy correctly

so if the player is holding right arrow to move right in this side scroller, the character's localposition.x will keep going up and up and up and up, which is what you expect

same with the camera and other objects that aren't supposed to repeat / wrap

@iron torrent does that make sense?

if you need to place an enemy that should "Repeat" it has to be in the "static (real) world" hierarchy

the tricky part about this is you probably think it's the other way around

and that's how you wound up with 3 cameras

you can still place 2d colliders and such in the static world. when you move the wrapped world transform, you have to sequence everything correctly and make an adjustment if you are using physics

you can alternatively move the static world duplicates. it's the same idea

[undone coral]

you can use an interpreter based solution like moonsharp for C#, embed lua, or try to roll your own thing using javascriptcore.

[alpine sinew]

What software do you guys use to manage Git repositories? I've been using github desktop so far and it's a hindrance because I can't revert to old commits without creating new branches.

I'd rather use something visual than managing the git in the terminal.

or tell me in which channel i should post this question

[undone coral]

the only time i interact with a UI to use git is a merge tool. i use the one built into Rider

i also use Rider's UI to visualize diffs and history

and i use it to do an interactive rebase, i.e., to write commit messages. but i also do interactive rebases in the terminal

otherwise, the whole point of git is that its cli hasn't changed since it was gifted to us by linus torvalds, maybe one of the greatest software architects who ever lived. so if he felt it should have a GUI, he'd make one

so you better start using git in the terminal

[alpine sinew]

lol i'm not mad at it

[undone coral]

lol

[alpine sinew]

i'm just lazy.

[undone coral]

yeah

understandable

it will really free you

then you will have this skill that will last forever

isn't it interesting that, compared to 10 years ago, i use git the same exact way, but i can't say the same for e.g. Unity?

there's a reason it is the way it is

[alpine sinew]

i've used git professionally when i worked with linux systems ... just, like i said, too lazy right now lol

[undone coral]

also the guy thought really long and hard about how multiple people work together on code

well

just do it

Rider's is nice

[alpine sinew]

but you said Rider - so i thought "i wonder if visual studio has a git visual" and it totally does

[undone coral]

the jetbrain's ides have good UIs

i can't speak for that one

i think you are bound to screw something up using them unfortunately

[alpine sinew]

visual studio also doesnt' support checking out old commits. I guess i'll go look into git from the terminal again 😦

thanks @undone coral

[undone coral]

i can definitely do it in rider

[dusty wigeon]

I use TortoiseGit/TortoiseSVN.

[fierce kayak]

hey guys i know this is probably super easy to ressolve, but how would i set a bool in one script to whatever the bool is in the other script? cuz i want to detect whenever it's gameover, which does all of that work in my playerscript but i want that same bool to be updated in my logic script so i can detect if whenever i try to pause the game it will let me or not depending on if its game over or not, and all of that pause stuff is in my logic script which is a different script to where the bool gameover is being done. tag me if you reply

[undone coral]

#💻┃code-beginner

[tulip cradle]

Do quaternion questions go to code or physics?

[iron torrent]

@obsidian glade @dusty wigeon @undone coral

I have attached a screen shot, you can see the location of the 3 cameras as well as the affect that they have when not clearing the depth flag. The red square is the representation of the player.

So when the player moves left and right, the Cameras will adjust to where they need to go. Keeping the Game View wrapped like it is.

The issue I am having, is that the Left and Right Camera are both Overlays in a URP project. And the main Cameras has a Brain attached to it. And the Virtual Camera is setup to just follow the player with no damping, which works. But the issue is that when the player is in the position as depicted in the Game View, you can every so slightly see a lag or gap between the two screens. The cameras Left and Right are being controlled by a Script that moves with the main Camera.

As far as wrapping the player goes, it works, enemies wrap fine. The issue is the Cameras itself, it is either not synced or they Glitch, meaning that they flash the background every now and then.

[undone coral]

okay

i don't get it. why are you using these side cameras

succinctly

and do you think there is a way to achieve this

[iron torrent]

Why?

So the Game View shows what is shows

[undone coral]

without using those cameras?

you could duplicate the object with 4 written on it

and likewise with 1

and put them on the ends

[iron torrent]

no, you can't

[undone coral]

wouldn't that achieve the same thing?

if i wanted to show what hte game view shows, can't i duplicate "4" into x = 0?

i.e. its rightmost border is at x = 0?

yes or no

[iron torrent]

If you duplicate it, you would need to duplicate everything that is in that section

[undone coral]

okay sure

but you agree that duplicating it Just Works

right?

[iron torrent]

no it doesnt

[undone coral]

okay, why not

[iron torrent]

Because it won't work

[undone coral]

i don't understand. if i duplicate the rectangle with 4, and put it in world space before the "1"

i will get what i see in the game view, with 1 camera

are you saying i wouldn't?

[iron torrent]

the rectangle represents the background of that scene area

[undone coral]

if you want me to help you, help me by answering my questions

i am asking for a simple yes or no

i'm not asking for caveats

[iron torrent]

Dude, the image of 1,2,3,4 is one image it is to represent the screen play area.

[undone coral]

okay, that's fine

so sure, if i duplicate that one image

[iron torrent]

the entire image is one image

[undone coral]

that wasn't obvious to me - if i duplicate the background into the appropriate position

it will recreate the appearance, right? but with just 1 camera

yes or no

[iron torrent]

no

[undone coral]

really?

[iron torrent]

I don't think you understand the problem

[undone coral]

well, that's why i'm trying to ask very simple questions