archived-code-advanced

[olive totem]

Glad I don't have that in my codebase, much rather have a couple of readable if statements

[misty glade]

Yeah, I think writing good code is making it as obviously simple as possible, not flexing with obscure syntax or use.. there's always fun competitions and stuff for that (IOCCC)

[high lotus]

Hey guys, I'm trying to make a spin wheel, but I want the arrow to stop at one of the four icons. Whats the best way to do this? Right now im just rotating the arrow for a random amount of time and then slowing it down.

https://gyazo.com/ecb11e56e13aed9ce88c3aa2f427a402

[olive totem]

Agree it comes off as a bit pretentious

[misty glade]

A) Determine where you want to stop
B) Spin the wheel for a fixed amount of time
C) Tween the speed from 1.0 to 0.0 between the current direction and end direction

[cedar ledge]

float speed = useWalk
  ? (useSprint
    ? (Input.GetButton(sprintButtonName)
      ? walkSpeed * sprintMultiplier
      : walkSpeed)
     : walkSpeed)
   : (useSprint ? walkSpeed * sprintMultiplier
  : 0f);

[misty glade]

... that's really hard to mentally parse

[cedar ledge]

¯_(ツ)_/¯

better than 1 line

[olive totem]

@high lotus does this help?

int random = Random.Range(0,4);
float x = Mathf.Cos(((2 * Mathf.PI) / 4) * random);
float y = Mathf.Sin(((2 * Mathf.PI) / 4) * random);
Vector2 dialPosition = new Vector2(x, y);

It'll pick a random point from any of the 4 points in the circle

Forgot to add a radius but it gives you a vector in the range 0-1, if it needs to be bigger just multiply the vector

[cedar ledge]

for a sphere use Random.onUnitSphere

[olive totem]

He only has 4 points so not sure why you'd make it harder

[cedar ledge]

just leaving that knowledge there because uniform randomness on the surface of a sphere is unintuitive, ive been told

[olive totem]

Only thing random is one of those 4 points it can land on, again not sure why to make it harder than necessary

[high lotus]

Thanks guys, I'll try this

Nope, I have no idea what to do with that 😂

[olive totem]

How do you move that dial right now

I guess if you use some kind of rotate function you can do something similar in degrees, like dividing 360 by 4 and then multiply that number with a number between 1 and 4. This will give you a rotation that always lands on 1 of the 4 spots

[high lotus]

if (!finishedSpin)
        {
            spinwheelImage.transform.Rotate(0, 0, this.rotationSpeed * SpeedMultiplier * Time.deltaTime);
            if (slowDown)
            {
                this.rotationSpeed *= 0.92f;
            }
        }

In update

Then it counts down between 2 and 4 seconds and slowDown becomes true

[olive totem]

You could keep doing something like you're already doing, and just stop increasing the speed until it fully stops at a desired location

[cedar ledge]

private void Resize()
{
    Vector2 targetMotion;
    Rigidbody2D targetRB;
    if(target.TryGetComponent<Rigidbody2D>(out targetRB))
        targetMotion = targetRB.velocity;
    else if(target.parent.TryGetComponent<Rigidbody2D>(out targetRB))
        targetMotion = targetRB.velocity;
    else if(target.parent.TryGetComponent<Rigidbody2D>(out targetRB))
        targetMotion = targetRB.velocity;
    //for determining delta
    float before = c.orthographicSize;
    c.orthographicSize = Mathf.Lerp(c.orthographicSize,
            Mathf.Lerp(minZoom, maxZoom, targetMotion.sqrMagnitude), zoomStrength * Time.deltaTime);    //super smooth
    float delta = c.orthographicSize - before;
    //warning if camera zooms too fast
    if(Mathf.Log(Mathf.Abs(delta)) > -2)
        Debug.LogWarning("Camera changed size really quickly.");
}

why does it think targetMotion is an undefined local variable on the long line?

i defined it at the very top, so it should be in scope

[distant pivot]

It's declared but not defined--with your if statements, it doesn't appear to have a guaranteed value

[cedar ledge]

ah

(nevermind that)

[distant pivot]

If you have an else at the bottom, you could give it a default value (or before the ifs)

[cedar ledge]

it all makes sense now, thanks

for some reason i thought structs got defaulted to a zero-like state

but thats just with fields

im so happy i got my moving platforms, character controller, and camera follow working absolutely perfectly

the character sticks to the platform perfectly

[native hinge]

Thank you

[cedar ledge]

https://youtu.be/ObGB3N4tlZ8

[native hinge]

Here's a more comprehendible version of my quadro-ternary

float speed = 0f;
if (useWalk)
{
    if (useSprint)
    {
        if (Input.GetButton(sprintButtonName))
            speed = walkSpeed * sprintMultiplier;
        else
            speed = walkSpeed;
    }
    else
        speed = walkSpeed;
}
else if (useSprint)
    speed = walkSpeed * sprintMultiplier;
else
    speed = 0f;

If discord wouldn't auto tab stuff

[cedar ledge]

you could simplify that

if (useWalk)
{
    if (useSprint && Input.GetButton(sprintButtonName))
        speed = walkSpeed * sprintMultiplier;
    else
        speed = walkSpeed;
}
else if (useSprint)
    speed = walkSpeed * sprintMultiplier;
else
    speed = 0f;

i think i can do even more

if (useSprint && Input.GetButton(sprintButtonName))
    speed = walkSpeed * sprintMultiplier;
else if (useWalk)
    speed = walkSpeed;
else
    speed = 0f;

that makes sense to me

@native hinge

[native hinge]

O

h

thanks

There is just one problem

[cedar ledge]

also you should create a sprint axis in the input config instead of hard-coding the name

[native hinge]

I made a button called "Sprint"

and sprintButtonName == "Sprint"

[cedar ledge]

ah ok

i havent used getbutton before

[native hinge]

ah

Anyway, there is only one problem with your simplification

If useWalk == false but useSprint == true I want the player to auto-sprint

In your case it stops movement

[cedar ledge]

so the getbutton doesnt even matter

[native hinge]

In that case, yes

[cedar ledge]

if (useSprint)
    speed = walkSpeed * sprintMultiplier;
else if (useWalk)
    speed = walkSpeed;
else
    speed = 0f;

[native hinge]

Not quite what I meant

If you can both walk and sprint then the button must be held

If you can only sprint but not walk the button does not need to be held

[cedar ledge]

if (useSprint && (Input.GetButton(sprintButtonName) || !useWalk))
    speed = walkSpeed * sprintMultiplier;
else if (useWalk)
    speed = walkSpeed;
else
    speed = 0f;

like that?

[native hinge]

Yeah I think that works

[cedar ledge]

feel free to mess with it however works for you

i was just trying to show it should be possible to simplify

[native hinge]

Mhm, thanks

[cunning dirge]

It's true, but that's fine with me, it might be a little bit, I just want to get paid quickly. I think I was traumatized

[vital breach]

I want to Invoke something from a abstract class anyone an idea how to do that?

[urban warren]

What do you mean? Just invoke it like a normal method..?

[vital breach]

Maybe invoke is no the thing i meant. I want to introduce delay time to a method so that it cant be spammed. But i cant use coroutines

[queen plover]

You can if you get reference to a Mono ehavioe

Then you can use that to call StartCoroutine

[vital breach]

Yeah but this would make my Statemachine to coplex im searching for a simple method

[queen plover]

Without any context it's impossible to know what to do. There isn't a way to start a coroutine without a monobehavior (or maybe scriptable object) though

[vital breach]

hmmm

[urban warren]

You could do async maybe?

[vital breach]

async?

[urban warren]

https://docs.microsoft.com/en-us/dotnet/csharp/programming-guide/concepts/async/

[queen plover]

I remember hearing that doesnt play nice (you can't use unity specific methods or something like that) though I could be wrong

[urban warren]

No, you can. Or at least I haven't had problems yet using unity specific methods.

[queen plover]

I've definitely used it for addressables so I think I'm. Missing context on where I hear that

[vital breach]

Uhh i see. This could work i will give it a try, if not i will just make a coroutine helper and try to not overcomplexificate that Statemachine. Because i will have to run many coroutines

[compact ingot]

You can’t use most engine methods while not on the unity main thread… if you use async without Task.Run…. Your async method will stay on the main thread and unity api calls work

[urban warren]

Ah, got it, that makes sense. I knew about the main thread thing, just didn't realize when async did/didn't run on the main thread. Makes sense why async works fine for me then when I have used it. Thank you for the additional info!

[queen plover]

That was where I got that from then

I always though async was it's own thread though, how else does it not freeze the game with stuff like await

[compact ingot]

Also the unity synchronization context introduces a 1ms thread.sleep whenever an async „thing“ is scheduled I.e. when you await something. That can be a problem but depends on the project.

[queen plover]

Btw, would you really need async for a state machine? Feels thatd be more event bases

[compact ingot]

Furthermore async allocates memory…. So those in terror of the GC don’t like it

[compact ingot]

Async has nothing to do with state machines

[queen plover]

I mean in this user's question

They wanted something akin the coroutines for a statemachine

[compact ingot]

Why not use a proper state machine?

Coroutines are so very Limited

[vital breach]

I actually found a way to fix my problem but its really messy and will make my code a lot longer. With some disgusting if statements

[queen plover]

Tbh I try and avoid state machines for as long as I can. They spook me especially since the simple implementation is very ugly and the more complex one is - more complex to make

[compact ingot]

https://github.com/dotnet-state-machine/stateless

This one has very nice fluent interface and works for gameplay scripting just fine

[vital breach]

I am making a state machine for my player and its quite useful for the many different movement types i am implementing

[cobalt dome]

hi i have a for loop. going through 10 objects. i need an int to count up until half way through the count then start counting down.
so as the for loop i goes 0 1 2 3 4 5 6 7 8 9 the int goes 0 1 2 3 4 4 3 2 1 0
anyone know how to do that?

[queen plover]

Probably a much cleaner way, but an If statement and book that is set when it reaches half way would do the trick

Though it doesn't sound very nice

[compact ingot]

for (int i = 0; i < length; i++)
    int j = length - i - 1;
    int k = (i < length / 2) ? i : j;
}```

[austere jewel]

One for loop, and another reverse for loop. Wow, easier to write, easier to read

[cobalt dome]

thank you. it works

[urban warren]

Do you know how it works?

[compact ingot]

works just as vertx said 😉

[cobalt dome]

yes. i had somthing similar to the k part but the j is what i wasnt getting right

[urban warren]

Just double checking. We are in #archived-code-advanced just want to make sure people aren't copy pasting stuff 🙂

[compact ingot]

that wasn't an advanced code question... was more like a leet code "very easy" question, kinda fizzbuzz level

[austere jewel]

Very much sounds like the kind of question that'd be on a bad quiz; which is why I generally would avoid answering them as given

[compact ingot]

good point

[mighty cipher]

Anyone good in math and geometry to help me with something that might be easy, but i don't know the math to do what i want to do.

thing that i am doing, i'm placing two objects, then spawning a plane on the middle of both of them? (so far everything working like intended)
The object i'm spawning is a plane (as i am working on an AR app) the only thing is i would like the plane to be have it's rotations to make it be in the direction of the vector(just like if i was placing a wall), but i have no idea how to transform my vector dirrection into the rotation needed

        if(placedObjects[0] && placedObjects[1]){
            GameObject objectMiddle = GameObject.CreatePrimitive(PrimitiveType.Plane);
            objectMiddle.transform.localScale = new Vector3(100, 100, 1);
            objectMiddle.transform.LookAt((placedObjects[0].transform.position - placedObjects[1].transform.position));
            Vector3 middlePosition =  (placedObjects[0].transform.position+placedObjects[1].transform.position)/2;    
            objectMiddle.transform.position = middlePosition;
            EventManager.TriggerEvent(ShelfPlaneHandlerData.SpawnPlaneHandler);
        }

[austere jewel]

position it, then use LookAt. LookAt does not take a direction.

[mighty cipher]

the fact is

[austere jewel]

Or use Quaternion.LookRotation, which does take directions

[mighty cipher]

read what i said, i don't want it to look at something, that's not the question

[vital breach]

You are so fast i had not even finished understanding the question wtf

[compact ingot]

a plane doesn't work with simple lookAt because its normal is Y (iirc)

[mighty cipher]

the question is about the math

[austere jewel]

"i would like the plane to be have it's rotations to make it be in the direction of the vector(just like if i was placing a wall)"?

[mighty cipher]

yes

wana me to draw it so it's a bit more understandable ?

hold on

[austere jewel]

it's not difficult to understand

[mighty cipher]

So how ur answer is solving the problem ?

[austere jewel]

You are using LookAt incorrectly. I've corrected your usage, and described another method to orient the plane that takes explicit vectors

[mighty cipher]

that's on the last try, i wasn't using that before

but the fact is

the question is

How to convert a Vector dirrection to a vector of rotation

[austere jewel]

#archived-code-advanced message

[mighty cipher]

....

[austere jewel]

https://docs.unity3d.com/ScriptReference/Quaternion.LookRotation.html you provide it a forward vector, and an optional upwards vector to hint at the orientation, and it returns a rotation built with those directions.

[mighty cipher]

I'm trying it rn

but not sure it's gonna work

i'll keep u posted

https://i.imgur.com/ReOgrBV.png

not working

[austere jewel]

What is it doing?

[mighty cipher]

Nothing els than it was previously doing

[austere jewel]

It will do different things depending on how your plane is configured. As Anikki said, planes usually have Y up, so you'll have to do some cross products to get the correct vectors

[mighty cipher]

that's the thing i need, i am not that good in Math that's the issue

[compact ingot]

Quaternion.FromToRotation is your friend

[austere jewel]

Cross product returns a perpendicular vector from two input vectors https://docs.unity3d.com/ScriptReference/Vector3.Cross.html
FromToRotation could be a friend, but it doesn't well specify the other axes

you could just set Transform's up vector and do what FromToRotation does, as that's all up's setter does

// The green axis of the transform in world space.
public Vector3 up { ... set { rotation = Quaternion.FromToRotation(Vector3.up, value); } }```

[mighty cipher]

ok

i'll try it

ok so i did a few test before trying to change

with this

i have something, but to make it good for me rn with that X scale of my object is Y scale and Z scale is X scale, and Y = Z

Ok it's good, i'll manage with that, Thanks anyway @austere jewel and @compact ingot for the help it was good documentations to use !

[finite pond]

I've noticed that the elements of UnityEvents are re-orderable since Unity 2021. Does that mean the execution order is constant now, meaning from top to bottom? Or is that just a nice-to-have for designers?

[flint sage]

https://github.com/Unity-Technologies/UnityCsReference/blob/master/Runtime/Export/UnityEvent/UnityEvent/UnityEvent_4.cs Pretty sure it's always been

It just iterates through the call array and invokes them one by one

[finite pond]

Oh, really? I've read on the forums that it dealt with delegates and its execution order was random.

That's great to hear!

[flint sage]

Even if it was delegates it's not random

It's just not predictable in 100% of the scenarios

(while it should be in most iirc)

[finite pond]

Hm. Gotcha.

[flint sage]

I don't think Unity guarantees (i.e. in documentation) that they're executed sequentially but I doubt it'll change

[finite pond]

Yeah

[unkempt nova]

Don't crosspost

[shadow seal]

<@&502884371011731486> scam

[hearty garnet]

Hello,

I'm trying to get the spot light in Unity to have no fall-off. I want it to illuminate objects the same amount no matter the distance.

As far as I understand, I need to write a custom shader for that?

[stable spear]

Like a laser light? In a narrow band? Or still expanding like a cone?

[hearty garnet]

Both works, whichever is easier

[stable spear]

have you tried creating a spot light and adjusting the parameters to see if it matches what you want

[hearty garnet]

yessir. so, for objects further than 5 units or so away, i need to increase the intensity a lot. and on that intensity everything closer is super bright. i tried setting the intensity by the distance of whatever we are aiming at, but it's also not very pretty because player can aim to something very far and the super intense light radius will happen to have a closer object on the edge

[stable spear]

This also looks like it might be useful: https://docs.unity3d.com/Manual/ProgressiveLightmapper-CustomFallOff.html

[hearty garnet]

As I understand this is for baked lights, I'm trying to have this behavior on realtime spotlight

[dusky dawn]

hello. I'm trying to get inspector like rotations from code. I see that one quaternion can be equivalent to multiple euler angles, but my question is how do I get the closest euler angle to my current stored euler angle. Any tips? I want a method where I pass a quaternion and the current euler angles and it returns the euler angle values that are closest to the input
Vector3 GetClosestEuler(Quaternion quat, Vector3 currentEuler)

[untold moth]

What do you need it for? Maybe there's a simpler solution than that?

[dusky dawn]

I have an in game animation system and I need to auto key frame getting the euler values from the inspector

unity quaternion to euler conversion is quite bad for this usage

for example, 95, 0, 0 becomes 85, 180, 180

this will lead to a quite different rotation path than the desired

[untold moth]

Can't you set the quaternion value directly?

Also, that example doesn't sound like a realistic scenario

I think quaternion to euler would always return the smallest possible value.

[dusky dawn]

nevermind, you'd have to followed the path I followed to understand the problem

I guess this is a quite complex question for an internet forum

[stable spear]

transform.eulerAngles ? https://docs.unity3d.com/ScriptReference/Transform-eulerAngles.html

transform.localEulerAngles should match the inspector

you can both get and set it

but if anything sets transform.rotation or localRotation, it will "mess up" the euler angles, in the sense that they get set automatically based on the quaternion (which is not a unique transformation)

[dusky dawn]

not it's not transform.eulerAngles

nevermind, it's a too advanced question for here

[hearty garnet]

try UnityEditor.TransformUtils.GetInspectorRotation(transform);

[dusky dawn]

it would work, if it was not editor only

[queen plover]

Perhaps you can look at what GetInspectorRotation does

[daring pelican]

more of a C# question here: how do I write a generic method that would accept a monobehaviour as parameter and then do something like if(GetComponent<T>()) GetComponent<T>().enabled = false;

[stable spear]

https://docs.microsoft.com/en-us/dotnet/csharp/programming-guide/generics/generic-methods

you can add a condition for T, such as:
void EnableComponent<T>(GameObject obj) where T : MonoBehaviour { ...

[tough tulip]

use
bool TryGetComponent<T>(out T s);

[hollow garden]

you already have TryGetComponent
used like ```cs
if (TryGetComponent(out Script script)) {
script.Something();
}

[daring pelican]

Oh dain thanks guys

That's awesome

[queen plover]

Could you also do, GetComponent<Comp>()?.enabled = false

Can't test it right now so not sure

[sturdy tartan]

anyone used AnimationEvent.time in the past?
i'm getting some hasardous numbers with it, and i don't understand how to use it

[dusky dawn]

it will end with this extern method, so it's not possible to read it

[queen plover]

Shame. As you said it can be hard to ask these kinds of questions on Online Forums. Hope you figure it out.

[dusky dawn]

thanks

[warm sonnet]

Hey guys, Im making a damage indicator system that displays where the damage came from, and it all works besides the fact that it does not rotate with my mouse as I look at an enemy/source. If i move to the left of the enemy, it will rotate with movement correctly. It (the canvas element/damage indicator prefab) just does not rotate with my mouse. If anyone could help that'd be amazing, thanks! I also have the video I used to create it.

[daring pelican]

used it in one of the projects, what numbers are you getting?

[sturdy tartan]

i was getting 0.33 while in my Animation that it shows on top 0:10 when the first animationEvent is fired

i basically just want to normalize the value of animationEvent.time

[daring pelican]

Yes you gotta scale it

What's the speed set in the animator state?

For this animation?

[sturdy tartan]

1

[daring pelican]

Hm ok hold on let me open up that old project

Check what's your animationClip.length is saying. I don't remember the solution exactly but I can see that I was using this in my code

You are not setting events via code?

[pastel harbor]

hello

Is there SVT under URP on the horizon already?

[river musk]

I have a list of 1mil gameObjects. Is this ok? Do I need to do it a different way to reference to them?

[warped grove]

what does => (lamda i think?) do?

[thin mesa]

not really an "advanced" topic, but here's your answer:
https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/operators/lambda-operator

[carmine ermine]

im currently spawning game objects randomly on a space of 100 x 100 plane. however, nothing stops them currently from spawning ontop of another.
how would i keep track of what positions are taken and which isnt? should i keep some sort of 2d array holding x and y values and try to keep track like that? or is there any other recommended way

[compact ingot]

unless you have a non-recurring RNG, you will have to waste some cycles or memory on checking for duplicates

you could in theory make a list of all 10k positions, shuffle them, and take N.... those then won't have duplicates... but you need a (relative) lot of memory and cycles

dont use a 2d array btw... use a hashset, the array would be as "wasteful" as the shuffle method

[misty glade]

Don't you hate it when you trip and fall into some corner of your codebase, and accidentally spend 3 days refactoring as a result?

[compact ingot]

best 3 days 😉

[misty glade]

heh

it actually stemmed from a conversation in this channel, and I finally decided to do this refactor and.. obviously.. it was far bigger of a bite than i expected

but it's going to feel SO GOOD when it's done

[compact ingot]

// LCG that produces a full cycle of pseudo random numbers without any duplicates, see https://en.wikipedia.org/wiki/Linear_congruential_generator
uint seed = 0;
uint sampleSize = 10000;
uint generatedNumber = seed % sampleSize;
uint increment = 7; // coprime to sampleSize

for (uint i = 0; i < sampleSize; ++i)
{
    generatedNumber = (generatedNumber + increment) % sampleSize;
}```

[misty glade]

That's cool.

[compact ingot]

if you make a casino game with that, please tell me;-)

[misty glade]

Without knowing Anikki's solution (which is kind of neat!), I would probably make a dictionary with a struct of x/y as the key and a bool as "is occupied" .. or if you need the opposite, a list of available locations with the same struct, and you get at a random index.

Yeah, it's clearly got some shortcomings, but for simple pRNG generation with some fixed criteria (non-repeating, true randomness not required, no moving elements) it's pretty cool

[compact ingot]

nothing like not compiling for two weeks, knowing exactly what you are doing and that it will work, and then it indeed does work

[misty glade]

@carmine ermine I'm doing something like:

        public GridPoint GetOpenSpot()
        {
            if (Items.Count == Width * Height) return null; // grid is full

            // Enumerate available locations.
            Dictionary<(int x, int y), GridPoint> availableLocations = new(Width * Height);
            for (int rows = 0; rows < Height; rows++)
            {
                for (int cols = 0; cols < Width; cols++)
                {
                    availableLocations.Add((cols, rows), new GridPoint(cols, rows)); //initialize
                }
            }

            foreach (GridItem gi in Items.GetAllGridItemsList())
            {
                availableLocations.Remove((gi.X, gi.Y)); // remove filled squares
            }

            int random = NumberUtils.Next(availableLocations.Count); // 0 to count-1
            int i = 0;
            foreach (KeyValuePair<(int x, int y), GridPoint> openSpotKVP in availableLocations)
            {
                if (i == random)
                    return openSpotKVP.Value;
                i++;
            }

            return null; //shouldn't be able to get here.
        }

[compact ingot]

hehe

[misty glade]

that's even only in one of my 3 applications, too

i did "complete" the server side of the refactor but who knows if it'll work yet 😛

[compact ingot]

this is very wasteful

[misty glade]

Basically I was tired of debugging stateful desync issues so I decided to refactor the entire "battle" object that contains everything about a 1v1 battle so that it could be serialized, compared, cloned, and checked for differences

[misty glade]

I'm looking for feedback on that, btw - I wrote that last night in a bit of a haze, and I hate it

(but it works)

I just don't see how it can be immediately/obviously improved, aside from writing it using linq (which isn't my strong suit)

[compact ingot]

if the method should only return one point, its extremely wasteful

[misty glade]

yep, it only returns one point

thoughts on improving it?

    public class GridPoint
    {
        [Key(0)] public int X;
        [Key(1)] public int Y;
        public GridPoint(int x, int y) { X = x; Y = y; }
    }```

GridItem is a fairly bulky item that "exists" on a grid

[compact ingot]

you need to have a grid item for the entire grid i suppose?

[misty glade]

No - a grid item is ... a thing, on the grid

think uh... merge dragons?

Grid items have a bunch of properties and stuff on them, but the grid itself is sparsely populated and items exist at a GridPoint (just a simple class with an x and y)

[carmine ermine]

thank you guys! ill look into it

[misty glade]

            int random = NumberUtils.Next(availableLocations.Count); // 0 to count-1
            int i = 0;
            foreach (KeyValuePair<(int x, int y), GridPoint> openSpotKVP in availableLocations)
            {
                if (i == random)
                    return openSpotKVP.Value;
                i++;
            }

I think this is the most egregiously bad part of the above since this should be O(1) instead of worst case O(n) but I needed the dictionary for random access removal, and converting the dictionary to a list to get index-access to return that value would take best-case O(n) anyway

[compact ingot]

HashSet<Vector2Int> Items;

public GridPoint GetOpenSpot()
{
    cellCount = Width * Height;
    if (Items.Count == cellCount) return null; // grid is full

    Vector2Int candiate;
    for (int i = 0; i < 50; i++)
    {
        int random = NumberUtils.Next(cellCount);
        int row = random / Width;
        int col = random % Width;
        candiate = new Vector2Int(col, row);
        if (!Items.Contains(candidate))
            return candiate;
    }
    return null;
}```

[misty glade]

lemme digest, standby

erm, if I'm understanding that right, you're just attempting to get 50 random locations and then if it fails, return null?

if i understand that right - that can fail

[compact ingot]

depends on how dense you want to fill that grid

[misty glade]

100% dense

[compact ingot]

then this is no good

[misty glade]

🙂

the method should perhaps be named GetOpenSpotGuaranteed()

it only returns null if the grid is 100% dense

[compact ingot]

sec

[misty glade]

np - appreciate it, btw!

[compact ingot]

would just keep that open list around ```cs
List<Vector2Int> Open;

public GridPoint GetOpenSpot()
{
if (Open == null)
{
Open = new List<Vector2Int>();
for (int i = 0; i < length; i++)
{
int random = LCG(Width * Height, i);
Open.Add(new Vector2Int(random % Width, random / Width));
}
Open.Shuffle();
}

return Open.RemoveAt(Open.Count - 1);

}```

[misty glade]

hehe, unfortunately that's a bit difficult since then I'd have to keep that list in sync with the gameplay - grid items can move around

so now moving would require finding the open item in that list, removing it, and adding the from location, etc

it gets complicated since there's lots of game actions that can bounce items back to a grid, remove items, etc - those would all need to get updated

right now the only "truth" is a list of grid items that have their X and Y internally

[compact ingot]

that grid isn't so big right?

[misty glade]

5x5 now, likely not (ever) bigger than 10x10

I don't think my solution is particularly awful in terms of memory/performance

it just... smells

[compact ingot]

then i'd just ignore any efficiency considerations and just avoid unnecessary allocations

[misty glade]

where are the unnecessary allocs?

(creating the Dict when i make the call, versus keeping it around?)

[compact ingot]

yes

[misty glade]

good point.. although moving it to a private member still incurs the memory cost

[compact ingot]

also, just a hunch... but it shouldnt be too difficult to maintain a open-positions list... it might be more effort to include in each game action... but thats basically how efficient data-structures get efficient... by maintaining some invariant on each add/remove/shift/rotate/....

[misty glade]

not that that's a huge issue atm - the server may have several hundred/thousand games at once, but rather the GetOpenSpot method is going to be called probably every turn anyway

it's uhm... super difficult to manage that, I think, unfortunately

[compact ingot]

well, turns arent frames 😉

[misty glade]

there are dozens of game actions, and the elements themselves are serialized/deserialized and two states managed at both ends

[compact ingot]

so i'd probably worry about other things... just make it somewhat nice, so i can sleep

[misty glade]

i'm not saying it's impossible but.. potentially a big piece of work

haha

thank you for the feedback 🙂

[daring pelican]

if you read my now deleted comment, please ignore it, I tested it further and i was wrong 😩

[timber saddle]

Is it possible to write native plugins for use in the Editor with loose C++ or Objective-C files?

[urban warren]

No, they need to be compiled in to a library. https://docs.unity3d.com/Manual/NativePlugins.html

(I am only like 98% sure though so take that as you will)

[timber saddle]

It's a bit ambiguous in the docs, there's no mention of the Editor, and it says:

You can deploy macOS plug-ins as bundles or, if you are using IL2CPP, loose C++ files
But it seems you're right.

[misty glade]

What are "loose" c++ files? You can compile libraries into native DLLs and they work fine..

@timber saddle

Do you just mean .. random files that don't have a corresponding project?

[timber saddle]

Loose as in not compiled into a native binary.
It's how Unity's documentation calls C, C++, Objective-C or Objective-C++ files that are just thrown in there as if they were C# files.

[misty glade]

Ah

[timber saddle]

I use that in Android and iOS. Now I wanted to make something for the Editor when running on a mac, but it seems I need to make an Xcode project and build a .bundle file.
Slight pain in the ass, but I guess it makes sense, as it's not using IL2CPP when running in the Editor.
I guess I'll just have to bite the bullet ¯_(ツ)_/¯

[misty glade]

ha - sorry I couldn't be more helpful 🙂

(I hope I didn't leave you hanging - I typed out the above and then had to go drop kids off at taekwondo and forgot to finish my thought)

[vast meteor]

does anyone know why this bug occurs (when i add the conditional before adding the dict), and then subscribe a object to a group, and try to subscribe a new object to a group, the new object dosent subscribe, however when i dont add that conditional, the objects all subscribe fine, however i cannot resubscribe a given object to any group while that object has been removed from a group and before it is deleted. The outcome i want is one where when i add a object, and then remove it, and then add it again, it adds like normal, and it works for any object i allow it to, basically the oposite of what the bugs do

remove script

video showing the bug

[stable spear]

you need to post the whole file somewhere, that's not enough code to know what's going on

[vast meteor]

https://pastebin.com/xbPBbzHv

i mean thats basically all you need but sure anyways

[austere jewel]

This is a really dodgy way of writing this sort of code for dictionaries of collections.
The usual way would be:

if (!dictionary.TryGetValue(key, out var list))
{
  list = new List<T>();
  dictionary.Add(key, list);
}

if (list.Contains(value))
{
  return;
}
list.Add(value);

Something like that.
If you don't need to perform any more additions or processing in the addition stage you could simplify it to something like:

if (!dictionary.TryGetValue(key, out var list))
{
  dictionary.Add(key, new List<T>() { value });
  return;
}

if(list.Contains(value)) return;
list.Add(value);

Similar thing for removal, use TryGetValue - exit early if there is nothing found.
Call remove, if the collection is empty, remove the key from the dictionary.

[abstract folio]

Trying to do cylindrical UV mapping to an arbitrary, but closed, polyhedron mesh. I’ve got it ALMOST right, but I’m having trouble with the area where triangles cross the UV.x=0 and UV.x=1 boundary.
I THOUGHT I had an algorithm that fixed it, as you can see the before & after results in this image.

Unfortunately, this does NOT work for ALL meshes, - you can see the before & after results with this mesh- clearly it didn’t work so well. And I’m not sure why it is failing in some cases.

The code I’m using takes a mesh as input, and attempts to compute the UV’s based upon the vertex positions, with the assumption that it is closed, and centered around 0,0,0 in modelspace. It’s just a few functions, you can find them here. https://hatebin.com/qnmarqmynf

^ other UV problems remain near the poles, but that's just the nature of cylindrical mapping onto a sphere-like shape. I'll deal with those later, if at all.

oh. I never actually asked a question, lol. Can anyone find the flaw in the linked algorithm?

[long ivy]

Don't be mad, but uh CylindricalUV calculates longitude incorrectly. It's opposite over adjacent. Your components are reversed

[abstract folio]

alas same result, just texture is rotated 90 deg

@long ivy good spot tho!

[sturdy tartan]

the solution for me to normalize was simply taking the animEvent.time / animClipInfo.clip.length

[green rock]

Has anyone ever encountered a bug in which runtime errors aren't displayed on the console even when you explicitly throw an exception at Start() function? Debug.LogError prints okay but not runtime errors. Kinda having a hard time testing/debugging things 😦

[abstract folio]

I HAVE had unity gobble up some exceptions I generated ! usually in GUI stuff.

[green rock]

It should have thrown an exception right away in the end of the console but it didn't

anyone knows how to look into this further? I can assure that this exception is not yet catched in anyway.

[abstract folio]

not caught by YOU.. but if you code is being called by unity itself.. that COULD be withing a TRY block

[green rock]

did you get any solutions for that?

[abstract folio]

I used Debug.Error instead 😦

[green rock]

It happens also with NullReference exceptions when I try to invoke methods from null objects

[green rock]

that's just sad. I hope others can shed a light on why this is happening.

[abstract folio]

@modest lintel it was you that pointed this out to me the other day, no? ^

[echo holly]

does anyone know how to render an RT with defaultHDR format into a png? doing the conversion from HDR rendertexture to texture2d format gives me conversion errors like "Remapping between formats 74 -> 52 is not supported" depending on the format.
i want to encode as a png, if the RT / texture are in LDR, then it all works fine, but when i try to use HDR RT then it breaks down. i came onto this when experimenting with switching the project to linear.

[kindred tusk]

PNG doesn't support HDR, which is likely going to be a problem.

Not an expert, but wouldn't you need to convert the HDR texture to the right depth first?

[echo holly]

yeah, so thats what im trying to do but it wasnt working. trying to convert the hdr RT to a texture and then encode that texture as a png

but not sure of the correct process that allows conversion of an hdr RT into a texture2d. seems like none of the texture formats are compatible

[compact ingot]

Doesn’t RGBAhalf work? At least ARGBhalf is the default HDR render texture format, so aside from byte order you at least have a matching color structure.

[weak agate]

for(int y = 0; y < height; y++){
   for(int x = 0; x < width; x++) { 
       Debug.Log(int2DArray[x,y]);
   }
}

O(n) or O(n^2) complexity?

settle a bet for me 😐

[flint sage]

Big O notation is relative to your array size, so it would be O(n) which would be O(height * width) while O(n^2) would be O((height * width) * (height * width))

[weak agate]

ty Navi, I agree lol

I'd love to know if anyone disagrees with that...

[agile yoke]

I think asking "O(n) or O(n^2)" is meaningless without talking about what "n" is supposed to be, because it does not appear in the code at all, and in this case it's imo not obvious enough to simply omit it.
Just say "O(height * width)" or "O(n) where n = height * width" and there's no need for any arguments.

[weak agate]

Can you give me an example for this case that would result in O(n^2)

[flint sage]

I mean it literally c an't be O(n^2) since n is one variable (unless width and hegiht are equal)😛

[agile yoke]

My point is only that "In O(n), the n refers to the array size" is a convention that is used because it's often convenient and obvious what is meant.
The fact that you asked this question here in the first place is, to me, sufficient evidence that it wasn't clear enough to someone what n is supposed to mean. The right answer is to just clarify, instead of semantic debate about how convenience conventions should be applied to this specific case.

(Other than that: If width and height are related, O(n^2) often becomes a reasonable answer; and that kind of context is obviously always missing from a short snippet like above)

[kindred tusk]

O(n)

It's not really an algorithm though, it's just exactly n steps. (maybe this is debatable maybe, I just mean big O is not necessarily relevant here)

[lapis mica]

i thought it was more about the growth rate as n gets bigger, so if height is a function of width or visa versa it would be O(n^2) but if they're unrelated it would be just O(n).
edit- wait I am wrong unless you take width as n in the O(n^2) case, and array size as n in the O(n) case

[kindred tusk]

Yeah, well O is worst case number of operations in terms of input size.

So a linear search is O(n) even if you're searching for the first element. That would be the best case which is a single comparison, but the algorithm is still O(n).

i.e.

var array = new [] { 0, 1, 2, 3, 4 };
Array.FindIndex(array, 0) // 1 step
Array.FindIndex(array, 4) // 5 (n) steps

But FindIndex is O(n)

Simply stepping through an array to the end is best-case, average-case and worst-case n operations.

So it's just not that useful a thing to discuss.

[lapis mica]

in this case the worst case is also the best case anyway right, since there are no early outs

[kindred tusk]

Exactly.

[weak agate]

yes I agree

just wanted to make sure I'm not losing my mind

[kindred tusk]

So it's O(n) because a 2D array is just a single array with some syntactic sugar

Internally it's just a 1D array with a pretty accessor [a, b]

Your code is just a single linear read of that array (containing width x height elements)

[orchid marsh]

The reasoning is relative to how often the elements are reoccurring. Since you're only iterating through each element once, your cost would be n where n is your number of elements. If you were to iterate each element twice, it would be O(2n). Thrice O(3n).

[compact ingot]

It is neither without definition what is being counted, what n is and what the return condition is. I.e. one could argue it is O(1) space if we ignore string allocations. anyone answering that question with a clear choice has no clue how complexity analysis works so nobody wins the bet

[drifting galleon]

+1

[mossy vine]

Hello this is my current code to zero out the position how can i make it more smooth
transform.localPosition = Vector3.zero;

[sly grove]

Think about it - you'll need to adjust the position a little bit at a time each frame until you reach zero, right?

[remote arrow]

you should use lerp or slerp

[mossy vine]

Yeah i tried this transform.localPosition = Vector3.Lerp(transform.localPosition, Vector3.zero, pickupSmoothTime * Time.deltaTime);
but it didn't work well

[sly grove]

where are you running it?

[mossy vine]

ohh

[sly grove]

Remember you need to do it a little bit each frame, so it needs to happen every frame

[mossy vine]

in a method

[sly grove]

not just once

[mossy vine]

i forgot

thanks

[vast meteor]

oh srry didnt see this i went to sleep last night

so does try get value not return null if the value dosent exist?

that way it dosent break the compilation process

[sly grove]

TrytGetValue returns a bool - true or false - to let you know if the object was found for the given key

the actual value that was found (if it was found) is given to you as an out variable

[vast meteor]

ah

[vast meteor]

the way im doing it would need the list object to be created during the set is added to the dict

nvm

i realized i could literally just plug the list = new list<T> into the dict val

[mossy vine]

How can i stop the Flashlight rigidbody from sliding

[native hinge]

Is it possible to call a method with multiple arguments in the UI system? (UnityEvents)

[compact ingot]

no

[native hinge]

Alright, thanks for the answer

[ember geyser]

i have lots of arrays with 512 floats in them. I want to sum them all, into another array of 512 floats. What's the fastest way to do this?

[compact ingot]

But you can pass in a reference to a scriptable object that can contain many fields

[native hinge]

Might try that

[compact ingot]

for loop with an accumulator

[urban warren]

Yeah, no 'magic' way to do that. Just gotta loop over all of them.

[native hinge]

I was gonna say make an int 'i' and then do foreach(int i in array) { sum += i; }

don't think it would be lol

[urban warren]

Yeah my bad. But it would need to be float not int

[native hinge]

right

[urban warren]

(That is what through me off for a second there)

[compact ingot]

One-liner ```cs
float[] sums;
List<float[]> allMyArrays;
sums = allMyArrays.Select(it => it.Sum()).ToArray();

[urban warren]

That would be less efficient wouldn't it because of the ToArray()?

It would require an additional loop through all of the items.

[compact ingot]

maybe not... since its lazily evaluated item by item... maybe the compiler is smart enough to figure out its a simple map-reduce.

[urban warren]

Nah, I don't think so. The source code doesn't look like it would but I could be wrong.

Also I could be wrong but it sounds like they want to sum by index across all of the arrays (sum all floats at 0, then all at 1, etc.)

[native hinge]

SceneManager.LoadSceneAsync(splitName[0], lsm??); Why does this give me an error?

Doesn't the '??' check if the value is null?

[urban warren]

No, it is the same as doing myInstance != null ? myInstance: otherThing

[native hinge]

ah

thanks

[urban warren]

Sure thing.

[compact ingot]

i wish i could understand that IL vomit that LINQ produces

[urban warren]

Lol I don't either. I was talking about the C# reference. I just mean I would be surprised if the compiler could optimize it like that.

[ember geyser]

my gut feeling is that getting a pointer to each array, and then summing each via to the next via float4 casting will best SIMD it.

I have another suspicion that memcpy of each array onto the stack will make this even faster, as it will reduce memory accessing time.

[compact ingot]

how would that make memory access faster? or do you mean prewarming L1 with the array?

[ember geyser]

pre warming doesn't seem to work nearly as thoroughly as would be nice for this, and memcpy is so fast that two can be put into two stackallocs, and summed, very fast.

and the order seems to matter, too. So creating the target stackalloc first, and then creatiing and copying into the two to be summed, and then doinig the summing, is faster than doing it the other way around. Then copy back out, if necessary, or keep cycling through the remaining arrays to the two input stackallocs, with memcpy

[compact ingot]

probably fewer pipeline flushes that way

[native hinge]

So is this the correct usage of the operator?
agent ??= GetComponent<NavMeshAgent>();

I read up on it

[urban warren]

Yup

[native hinge]

Cool, thanks

And does this check if target equals null?

target ?? null

To me it seems the same as ==

[urban warren]

No it returns the thing on the left unless it is null then returns the thing on the right.

So in that case it is exactly the same as just doing target

Because you are returning null if target is null

[native hinge]

oh thanks

So, if I wrote var c = a ?? b if a is null, it returns b, otherwise returns a?

[long ivy]

correct in syntax, but something you should never do with anything that derives from UnityEngine.Object because of the way Unity overloads the equality operator

[native hinge]

Good to know, thank you

And one last thing

if (TryGetComponent(out agent))
    agent = GetComponent<NavMeshAgent>();

This checks to see if the object has the NavMeshAgent component, and if it does, assigns it, is that right?

[urban warren]

The GetComponent is redundant.

[native hinge]

Oh

[urban warren]

If it has a component it will assign it to the out variable.

[native hinge]

Cool

So I don't even need the if

TryGetComponent(out agent); This auto assigns 'agent' to the NavMeshAgent if the object has it?

[urban warren]

Correct

[native hinge]

Awesome, now I wonder why every tutorial I've seen does not do this

Thanks again

I guess this is safer than a regular GetComponent because it first tries to get it?

It's like try{} catch{} ?

[urban warren]

No, it is like

agent = GetComponent<NavMeshAgent>();
return agent != null;

[native hinge]

ah

[urban warren]

It is more performant in editor too because it does allocate memory if it does not have the component.

[native hinge]

Good to know

[austere jewel]

note that you should not use ?., ??, ??= on UnityEngine.Object types

[native hinge]

Yup, that was mentioned to me earlier

Thanks

[compact ingot]

after some testing in the rabbit hole 😳 ... turns out even a simple SIMD approach with 6 extra lines of code is 10+ times faster than a basic for loop. takes adding 169000000 floats in a parallel.for on 12 cores to reach that SIMD performance.

[urban warren]

I am not familiar with this, are you able to share the code (even if not pretty) so I can take a look at what it does and how it works?

[compact ingot]

https://pastebin.com/gFBma6Ax

Reducing float[8192][8192] to float[8192] via 67100672 float additions ...
Basic          181.907200 ms
Parallel SIMD  25.687800 ms
SIMD           15.813900 ms (vector size = 8)

[urban warren]

Interesting, thank you. Would you be able to explain how the SIMD method works, or at least partly? If not that's fine too, I can google it to learn more too. 🙂

[compact ingot]

Those vector operations get executed with special instructions (single instruction multiple data = SIMD) on the CPU that can add/multiply/divide/... up to 8 single precision floats or integers at once... So you copy the values of a array slice into the vector, then add it to result vector, all in one operation, then shift the slice start index by vector size, and repeat... at the end you collapse the result vector, add the few array entries that didn't evenly fit into a vector and done. @urban warren

static void TestSIMD(int n, float[][] rows)
{
    Stopwatch sw = Stopwatch.StartNew();
    float[] result = new float[n];
    int vectorSize = Vector<float>.Count;
    for (int i = 0; i < n; i++)
    {
        Vector<float> accumulator = Vector<float>.Zero;
        float[] row = rows[i];
        int j;
        // SIMD add
        for (j = 0; j < row.Length - vectorSize; j += vectorSize)
            accumulator = Vector.Add(accumulator, new Vector<float>(row, j));
        result[i] = Vector.Dot(accumulator, Vector<float>.One); // reduce the vector
        for (; j < row.Length; j++) // add the odd tail to the result in case n % row.Lenth != 0
            result[i] += row[i];
    }

    double lap = sw.Elapsed.TotalMilliseconds;
    Console.WriteLine($"SIMD           {lap:0.000000} ms (vector size = {vectorSize})");
}

[ember geyser]

so very very glad you're seeing the same huge gains. I thought I was doing something wrong.

or imagining things.

[compact ingot]

i always thought that those were at best 2X to 4X faster (in practice) ... but this is very nice. with not much extra effort

[ember geyser]

In one endeavour, I got SIMD + stackallocs to be 12x faster than my best efforts with Burst+ Jobs...and thought I was losing my mind.

[compact ingot]

but isnt burst supposed to use SIMD?

[urban warren]

I see, sort of. I will need to write it out my self and mess with it to really understand I think, but you definitely made it a lot clearer for me!

The Vector thing was what was confusing me the most, which you cleared up.
I never knew what SIMD stood for, thanks.
I feel like with such an increase in performance people are really sleeping on this...

[ember geyser]

Indeed, but Burst still has to do all that work of getting things sorted out its way, etc etc... whereas with stackalloc and memcpy and lining stuff up, I could easily beat its way of lining things up to get ripped through by the registers

And my initial tests with float4x4 have turned in StUNNING results.

[compact ingot]

true... but my pitiful attempts at using memcopy/stackallock made the SIMD method 3X slower 😉 so i guess it has a lot of potential but its very non-trivial to achieve

[ember geyser]

it's swings and roundabouts. Sometimes it's needed, sometimes it's just extra overhead.

For me, it's been far easier than ANYTHING i've done with unity in the 2.5 years I've been using it, because it reminds me of my C childhood.

[compact ingot]

tbh, i'd rather write that memory stuff in plain C somehow... but with Span<> and such, i guess one can get rather close in convenient/safe c#

[ember geyser]

I can kind of intuit where and how and what to do to get the SIMD and vectorisations to fly, and it's been super relieving to think outside of FARKING OOP!!!

AGREED! But it feels like having an illicit affair, doing low level in C#. Fun, fun fun!

[compact ingot]

true

OOP is for suits

[ember geyser]

OOP is just so slow. And ECS has so many bloody rules.

By OOP being slow... I mean me. It takes me forever to work out an OOP architecture, for anything. This stuff, I can do everything I want, and have cycles to spare. Literally.

[compact ingot]

my issue with ECS in unity is, that they have to try to make something that works well when fully optimized for a particular application (a specific game/team/pipeline say) and make it generic... so it will end up very low-level and provide little improvement over DIY, or it will have, as you say, so many rules and edge cases that it is out of reach of the normal people that are supposed to benefit from it

[ember geyser]

most of these benefits you can get by 'tricking' the pointer into thinking it's looking at a much bigger type, because they're all 1D arrays. So you can rip through floats as though it's a float4, or even a float4x4

[ember geyser]

I need a billion thumbs up for this comment. You NAILED IT!!!

[compact ingot]

this is my new favourite aspirational video in regard to performance sainthood https://www.youtube.com/watch?v=Ur53sJdS8rQ

[ember geyser]

THANK YOU!!! Will queue up to watch as I nod off a little later tonight

[regal olive]

I'm not able to use Deconstruction syntax with Dictionaries:

Assets\BottleShooter\Shop\Shop.cs(16,56): error CS1061: 'KeyValuePair<string, ShopDataAsset.GunData>' does not contain a definition for 'Deconstruct' and no accessible extension method 'Deconstruct' accepting a first argument of type 'KeyValuePair<string, ShopDataAsset.GunData>' could be found (are you missing a using directive or an assembly reference?)

Assets\BottleShooter\Shop\Shop.cs(16,46): error CS8129: No suitable 'Deconstruct' instance or extension method was found for type 'KeyValuePair<string, ShopDataAsset.GunData>', with 2 out parameters and a void return type.

So far from https://stackoverflow.com/questions/42549491/deconstruction-in-foreach-over-dictionary# I saw that I might need System.ValueTuple which needs to be imported by NuGet, but it's already available in the code.

I'm using VS Code 1.61.1 and C# Extension 1.23.16 which is supposed to support .NET Core.

[gray pulsar]

Default Unity settings use .NET Standard 2.0. The last comment on your stackoverflow link seems to suggest it will be included in .NET Standard 2.1, which presumably means it is not in .NET Standard 2.0. This page confirms that hypothesis: https://docs.microsoft.com/en-us/dotnet/api/system.collections.generic.keyvaluepair-2.deconstruct?view=netcore-2.0

https://docs.unity3d.com/2021.2/Documentation/Manual/dotnetProfileSupport.html

[regal olive]

Indeed, and otherwise Unity uses .NET Framework 4.7.2.

Or 4.x

[wise sandal]

Has anyone here used discord Rich presence, whenever it changes scene it break and gets stuck on whatever the last thing was.

[orchid marsh]

Don't cross post

[onyx blade]

Is there merit in changing my node editor from XNode over to GraphView now that it has slightly matured or is GraphView still heavily subject to change and/or removal

[sage radish]

GraphView will be replaced with Graph Tools Foundation

Which is still a bit too early to use, so....

[onyx blade]

I thought development on GTF had stopped and was replaced by GraphView actually 🤔
Since it is being used by ShaderGraph and the VFX Graph

or is GTF meant to unify the graph API including Bolt?

[sage radish]

I don't know much about GTF apart from what I've read about in the NodeGraphProcessor issues, but it sounds like ShaderGraph will migrate to it in the future. (alelievr is a Unity employee)
https://github.com/alelievr/NodeGraphProcessor/issues/188

[onyx blade]

That does make it sound like that. I should probably wait for more details before making the switch.
Tho I see migration I wanna assume it will just shift the namespace, but thats a risky guess

Unity never rly gives a timestamp anymore, or so i've noticed

The hardest part of the migration will be to migrate existing graphs but I dont think that ll be that big of a deal since my node system is a chain of scriptable objects 🤔

[lusty iron]

Hey friends, we've built out our game for Mac and are encountering some issues. The game runs fine in the main menu and even through some gameplay, but after introducing some effects to the scene (by way of firing weapons) the Mac client crashes.

It gives out a bug report log, but I'm unable to discern anything useful from it. Something I've read online is that pushing the GPU on a Mac too hard will cause this type of crash, but from reviewing the Profiler on PC I can't imagine it's something like that.

To me it feels like a certain element of some of the FX causes this issue, but so far through process of elimination we've been unable to find more clues. We're going to keep doing that, but I wanted to solicit any insight here in case this was a common issue. Cheers

[flint sage]

No one can guess that, you'll to post logs or errors

[lusty iron]

Happy to

[flint sage]

I thin kyou should be able to symbolicate the crash with the symbols from unity

And what about the normal player log?

[lusty iron]

normal player log, like the console?

[flint sage]

Yeah

[lusty iron]

unable to get a read from that prior to the crash, there's no prior warning

I will definintely check that again though

what do you mean by symbolicate the crash?

[flint sage]

4 UnityPlayer.dylib 0x0000000101897c67 0x101235000 + 6696039
These bits point to certain methods, if you have the symbols (dsym for osx iirc) you can match them up and fill them in automatically

[lusty iron]

hmmm

[flint sage]

Unable to read as in you can't read the log or there's nothing in the log?

You can output the log to a file

[lusty iron]

nothing in the log that I saw, I'll do the text output but IIRC there was nothing logged aside from networking logs

could've been there for like a frame and I missed it, will defintiely double check when I can

being able to translate those codes to methods to get a better sense of the specific issue would be really helpful

[flint sage]

I only know how to do it with xcode for mobile projects, which is really annoying to do

Since it relies on xcode finding your dsym, which it fails to do most of the time or just won't do until a few days later

[lusty iron]

very interesting, wasn't aware of that process. googling into it now, thanks for the guidance @flint sage

[ember geyser]

How did you go with this research? Looked at Audio?

[lusty iron]

been tricky trying to get the symbols read correctly, not sure I know what I'm doing

there is indeed audio clips on some of the fx, do you think that could be it?

[ember geyser]

Yes, there's a good chance, if you're doing syncing of effects with audio, that that's part of the problem with Unity and a Mac. There's been a long and weird history of them not playing well together. Best to "pre-warm" all audio sources, and pool.

[lusty iron]

Righteous! We are currently pooling the SFX but have them paired (meaning on the same gObject) with many VFX. Appreciate the angle mate cheers

[ember geyser]

Be very careful moving them around fast (audio sources), it's best to completely disable them, move them, and then add whatever audioclip you need to them, then activate and play. And never use PlayOneShot, as it's a monster.

You can swap audioClips as often as you like, as fast as you like, as this is light and very fast. So you only need just enough AudioSources for the maximum number of sounds you'd ever play.

And there's a hard cap, by default, at 32 Audio Sources, too... I think. set this to 128 or something like that, just so you never hit it.

and the above line is in red, so I'm not sure if it got sent.

I haven't seen that before, haven't used discord a huge amount. The line is about setting audioSource limit to higher than the default 32 (for Mac), to something like 128

Also be very careful with reverb zones, make sure you have only ever got one impacting any audio source, as this gets weird, real fast.

[lusty iron]

hmmm, it very well might be an instance of too many audio sources, that seems very possible

so the scenario you're outlining would involve 2 object pools, one for VFX, and one for SFX

and rather than each SFX object being it's own per sound, you recommend using one pool of X amount of audio sources, and swapping the clips out by use?

[ember geyser]

Yes, exactly, try to use the fewest pooled AudioSources you can, as there's some efficiencies that come from this, and you'll get lots of other strictness in what you program that prevents other audio issues. Mac and Audio are bad, Mac and Unity Audio are very bad together.

Much of what I'm saying does not apply equally to iOS. It's much more of a Mac problem than anything else. This is partly because of the age and lack of updates to the FMOD version in Unity, Unity's ways of handling audio, Unity itself and the nature of security in modern Mac chips with regards Audio

[ember geyser]

The easiest way to solve this is to pool audioSources, in a manager, and have your VFX request audio sources sufficient to perform the entire FX for any given event. I know it's a pain to do this, but the gains are efficiency and stability. and you can be absolutely sure you're pulling deactivated audioSources, positioning them, adding the desired AudioClip and setting all values needed and THEN and ONLY then enabling the AudioSource, so it doesn't teleport through a bunch of reverb zones and other potential issues with rapid moving of them.

[lusty iron]

so right now we have a single VFX pool, with a bunch of different VFX that mostly all have Audio Clips on them

what we do is disable the gameobject, pool it, and turn it on when both the VFX and SFX are needed

I'm confused as to whether you think they should be separate GameObjects, or if the Audio Clip just needs to be specifically disabled

[ember geyser]

I know... I went through all this. It was a pain. It was a lot of work, but the result, for me, solved all problems of audio and Mac/iOS, and gave me better performance, as I used a lot less Audio Sources as a result.

[lusty iron]

so is the true goal to lower the amount of Audio Sources?

and by lower the amount, you mean lower the amount in the scene, not lower how many are active?

[ember geyser]

That's only one of the goals. It's more of a side effect, that you have lower numbers of them, and this is greatly beneficial.

[sturdy tartan]

Hi, is there any way to avoid having a state in the Animator not get stuck and not exiting the current state when the transition condition has been met already ?

[lusty iron]

alrighty, I think we can manage that. It's really just a matter of splitting up the prefabs, splitting the pools, and drastically limiting the pool number of the audio sources

@sturdy tartan not exactly sure what you mean, can you elaborate?

[sturdy tartan]

sure

[lusty iron]

also someone's going to suggest posting in #🏃┃animation instead so I'll do it first lol

[ember geyser]

Yes, the drop in audio sources, although a side effect of this approach, will be a huge performance booster, and remove other issues. But the real goal is to get you absolutely sure that you have a stable movement of deeactivated audio sources to positions, and certain control over their state.

[lusty iron]

it also seems like having audiosources moving around is not optimal? we are doing that in one or two cases to create a doppler effect

[sturdy tartan]

here's an illustration first to explain

[ember geyser]

Use ScriptableObjects to store the state of the AudioSources and their clips, so that you can do this efficiently, or public structs on each VFX unit with all the details needed to set to the audioSources as it pulls them in.

[sturdy tartan]

the animator get stuck at the animation in blue while the transition condition has been met

it can get out of it if i hold the transition condition true for 3 s

[ember geyser]

A few situations where you move at game object moving speeds (audioSources on moving game objects like rigidbodies) is fine. It's the teleporting to a location to play a sound that's the real problem.

[sturdy tartan]

but doing so, ruin my combo

[lusty iron]

and just to make sure I understand, teleporting to a location is fine as long as the AudioSource is 100% disabled?

because I do plan to place the sounds based on location

[urban warren]

As victory said this is not the right channel since this little to nothing to do with scripting. #🏃┃animation is the one you want.

[ember geyser]

And I fully understand what you're doing, the doppler effect, in conjunction with the reverb, is surprisingly VERY good in Unity.

[sturdy tartan]

i'm sry, i'll go there ^^

[lusty iron]

yeah right? we threw it on to just see how it worked and it was pretty nice out of the box

no worries @sturdy tartan they will likely be more helpful, I can't tell what your issue is

[ember geyser]

Yes, fully disabled, and even preferably without any audioclip during its movement to be doubly sure. Just program super, super defensively with regards AudioSources. There are myriad issues with them. There's not just one or two cures, this is a VERY old system that's never had much in the way of bug fixes.

[sturdy tartan]

well the thing is that the character get stuck at a state in the animator as in the picture while it should transition to the next state since the transitioning condition in the image has been met

[ember geyser]

If you crank up the doppler effect, to absurd numbers, it's still good!! Especially with short sounds with a bit of punchy resonance at the start that are in the same reverb zone. Can be quite entertaining.

Also, the Echo Audio Filter, at very low numbers of delay (think 11 through to about 24) with decay set to 0.9 or above, is super interesting, too.

The echo filter has a bug in the delay, you can't animate it. Unity knows about it, but I don't know when they'll fix it. So if you want to animate it, you'll have to do it with code.

[lusty iron]

As we saw with my inability to help with that animator question I'm pretty clueless with the animator so code's fine by me lol

I think this whole rework will help us take a closer look at things like audio effects, we're not using many rn

[ember geyser]

AnimationLegacy is worth a billion times more than Animator! If you can, don't use Animator, for anything!

And since you're pooling effects and stuff, DO NOT USE ANIMATOR!!!

There's huge problems with pooling and restarting Animators, that have never been fixed, that will drive you insane.

[lusty iron]

I'm using Animator

I'm gonna keep using it, our implementation has no issues so far

I'm not using it for effects though, just the main player character

[ember geyser]

Once you get to an issue, consider switching. It'll be quicker than trying to correct the ways ANimator works.

Oh, that's different. if you need timelines for effects, use Legacy Animation.

[urban warren]

Is the legacy animation even still in? I thought they removed it by now?

[ember geyser]

It wont be removed until well after they've made a similarly performant replacement, which isn't on the horizon.

AnimationLegacy is VERY lightweight and very fast. And using Evaluate on it is often quicker than making custom ease maths.

[urban warren]

I don't see it in any menus....

[ember geyser]

just add a component to any object: Animation is the name of it.

[urban warren]

Oh, it is Animation I thought it was previously called something like Animation - Legacy

[ember geyser]

othe rthan the docs, they dono't call it Legacy, because it's not really a legacy system, it's the old system, that's lighter and faster.

Unity's naming conventions are bizarre. The sequencer is called Timeline, the timelines in both Animator and Animation are called timeline.

[fallen halo]

Which is better for Object Pooling, Queue, Stack or List

performance wise

[rocky zephyr]

are you referring to the pooling classes in Unity or roll-your-own?

[fallen halo]

what is pooling classes in Unity?

[fallen halo]

Whatever it is I am talking about rolling my own script

[rocky zephyr]

2021.1 (IIRC) added them: https://docs.unity3d.com/2021.2/Documentation/ScriptReference/Pool.ObjectPool_1.html there are several variants. I use them in my editor extension and they work fine. Performance - unsure.

[fallen halo]

not this, rolling my own

which one should i use for better performance stack, queue or linked list

[compact ingot]

Depends on what exactly you want to do… all of them are potentially linked lists

[fallen halo]

i want to store particle systems that i will reuse as effects

[compact ingot]

I mean algorithmically, and quantitatively

[fallen halo]

30 objects in a minute

[compact ingot]

just use a list

[fallen halo]

i find queue easier. Is it favourable than list

[compact ingot]

it really makes no difference

a list is also a stack, backed by an array

and for pooling, a queue adds no benefit

[fallen halo]

understood

[fresh salmon]

Queue has methods ready for you that do all the job for inserting or taking an item from it

But that's pretty much it

[fallen halo]

let's say i have a character when it moves there are some effects that happens like dust effects on the ground. for things like that i don't need a object pooling right

[fresh salmon]

A simple particle system can do that

[fallen halo]

yep

[fresh salmon]

With enough modules enabled, mainly "rate over speed", you can mimic disabling it, as it will not be emitting if you don't move

[fallen halo]

got it

if i use queue, since it is FIFO(first in first out) then after dequeuing an object won't the program has to reshift all of it's other object indexes

[compact ingot]

Not if it is implemented as a linked list

[fallen halo]

i see

[fresh salmon]

Yeah if it is, it just has to redefine where the "head" is, without shifting

[compact ingot]

Overall, linked lists are terrible for performance though

much more efficient to implement the queue as a list/array with two pointers and resizing it occasionally instead of spreading individual nodes all over the heap

[olive totem]

Lists aren't great for performance either. If you're going to make a pool of objects then use an array and set up enough space so you never need to resize

[compact ingot]

List can do that too, has the same performance as a plain array and has automatic resize built in, just in case. While there are microscopic differences in performance between the two, they aren’t even 10%, whereas list vs linked list can be an order of magnitude faster

[olive totem]

Most information I come across claim that the array is always the fastest option for read and write operations

[lusty iron]

if you're concerned about performance differences between arrays and lists you're likely better off making your own engine. simply by using Unity you're accepting so much overhead that those minor discrepancies aren't really relevant

IMO

[olive totem]

Just an opinion but I think when there's no need to resize then array would be the best option

[lusty iron]

I totally believe that

but I'm not sweating over the times I use a list anyways lol

[compact ingot]

it’s even better to use stackalloc on the array and do math only with simd, but you’ll spend 10 times the time writing the code… and first of all that has to be worth it IMO

[olive totem]

Right but in this particular case using an array over a list isn't that much work, but one should do what makes the most sense to them of course

[urban warren]

I would disagree with this rather heavily.
There are plenty of times when you are making systems that built on top of Unity that either operate on a large number of items or are called very frequently and are relatively expensive where you would want to spend the time to optimize them where you can.
An example could be an AI system. Should you just built your own engine just because you want your AI you have good performance?

Of course I'm not saying you should always spend the time to optimize things, to the level of array vs list, it of course has to be worth it for the time.

[compact ingot]

and array vs list is also just a stopgap towards real optimized systems… hello SIMD, hello Parallelism, hello GPU compute, hello ML, hello pinned memory, hello structs of arrays, hello caching, hello branchless programming, hello better algorithms, hello better architecture …. Endless possibilities for PhDs 🕺

[pliant geyser]

looking for some help regarding pathfinding

im using astar pathfinding and im having trouble with gameobjects running into eachother and getting stuck

is there a way I can have the paths of 2 gameobjects not intersect with eachother?

[urban warren]

Are you using your own implementation?

[pliant geyser]

using the astar pathfinding project

[urban warren]

In that case it comes with local avoidance components. You can look on its doc site for more info.

[pliant geyser]

I think its reserved for pro users only

[urban warren]

Ah, yeah.

[tough knoll]

Hey y'all

I've recently turned to a one-texture-atlas-for-everything solution for my level meshes since it reduces the drawcalls from 2000 to like 20

But as a result of this I have no easy way to differentiate surface types by material anymore.

Are there any tricks that don't involve having seperate colliders per material surface?

This is for things like footsteps or impacts

[tough knoll]

Also, now that I'm reading - what is the advantage of these built-in pooling classes vs. writing a manager that instantiates and manages objects yourself?

[kindred tusk]

https://github.com/Unity-Technologies/UnityCsReference/blob/master/External/MirroredPackageSources/com.unity.ui/Core/ObjectPool.cs
This one doesn't instantiate, it only manages regular C# classes

oh, sorry. wrong one!

It seems 2021 source is not live yet

[austere jewel]

They stopped adding new sources a while back and it's just another "Why the hell, Unity?" moment

It is handy they provide a ListPool, DictionaryPool, and similar more generic versions

[kindred tusk]

oh that's super disappointing.

[austere jewel]

They stopped ages ago

[kindred tusk]

I know I was just poking around

It doesn't make it any less disappointing

More disappointing actually

[austere jewel]

Yeah, it blows hard

Could at least say that Unity's C# source was available, now you can't even say that

[kindred tusk]

I was hoping the C++ would be made public

Yeah, exactly.

UE is fully OSS right?

Wow it is. That's so cool

They even have a guide for building it.

I guess if you have licensing it doesn't really matter if you're open source.

Maybe Unity just do it for the paywall?

[austere jewel]

I'm not sure if I would use OSS, but it's certainly source avaliable

[kindred tusk]

Yeah, not free but open and you're allowed to modify it.

[austere jewel]

I have no idea why Unity does what they do. There's so many amazing things they could provide that'd add so much quality

stuff that the Needle guys provide without even having proper access

[kindred tusk]

UGUI had a live repo.

It was just a mirror too though, they weren't actually working on it.

It would be awesome if we could actually open PRs against their modules. I thought when they started using UPM that everything was going to broken out to GitHub

[austere jewel]

These are things I used actively, and still use albeit with a lot more caution

You can make PRs against stuff like the Graphics repo

Some packages are leading the charge on that sort of thing, but others are just becoming shut away again

[kindred tusk]

Oh, nice. Is that a DOTS repo?

[austere jewel]

SRPs

[kindred tusk]

The good thing about needle is that you can fork at least, and it still adds a nice way to merge it upstream changes.

[austere jewel]

https://github.com/Unity-Technologies/Graphics

[kindred tusk]

Ah that is great.

And they're working straight into this?

1,654 branches!

Holy jesus time to prune

[austere jewel]

I have made one amazing PR https://github.com/Unity-Technologies/Graphics/pull/5347/files/398b6176870ec4776cb4651b5ef65fb612eb42ab

[kindred tusk]

Wow you programmed Unity3d

[austere jewel]

Who knows, I may be in the credits now

[kindred tusk]

It should at least be on your CV 😉

"previous work"

[austere jewel]

"optimisations¹ for vfx graph"

[latent moss]

how can i project a scene onto an object

[real blaze]

hey. can someone please help me out on the two last arguments for the Crossfade animation ?

I wrote their help text in front of them

[compact ingot]

help with what?

[real blaze]

seeing them like this , I can't relate the last two arguments...

as to what they mean

the NormalizedTimeOfsset is the start of the second animation ? ( also start of the fade )

( also sorry for the light mode there. i hope its not night over there )

[final kindle]

I seem to have found a rather anoying compiler difference between VS and Unity. VS thinks this is fine, while Unity doesn't. It complains about casing lambda to lambda. If I remove the ternary so it's a direct setting, it compiles just fine. I'm guessing this is just straight up a bug and I should report it?

Comparison<Vector2Int> comparison = shiftAmount < 0 ?
    (hexA, hexB) => hexA.x.CompareTo(hexB.x) :
    (hexA, hexB) => hexB.x.CompareTo(hexA.x);

[real blaze]

Unity doesn't have a custom compiler. it's just an old version of C#

it doesn't even have the keyword not , for exampel

so if(A is not B) is accepted in newer dotnet , but not in old version that Unity uses

[flint sage]

Likely a unsupported feature in the runtime or compiler

[flint sage]

Unity has a custom version of Roslyn and a custom mono version, it's not just old.

[final kindle]

Maybe. It's weird that the lambda causes a casting problem though.

[real blaze]

if u want the newer fancy C# features, u can use .Net 4.x from player settings

[final kindle]

This makes it work

[real blaze]

I don't know what's the side effects though... so use it at your own risk

[flint sage]

I think it's technically a bug that it's not casting the correct thing

[real blaze]

yeah you could always find alternative ways 👍

[final kindle]

Definately. Just felt like something maybe worth reporting if I've not done something wrong.

[flint sage]

Nah, you haven't done anything wrong. There's no point in reporting it, Unity has to update the compiler/clr to fix it (which they're likely already working on)

[real blaze]

they contain specific syntax changes though ?

@final kindlehey mate. I googled a bit and looks like I was wrong. then sorry. .Net standard 2.0 isn't necessarily older than .Net framework 4.x

the numbers are a bit misleading there

from what I figured, this column is what .Net standard 2.0 means

[final kindle]

Ah. Thanks.

[compact ingot]

netstandard is a compatibility specification that contains all things that work in .net Framework AND .net core

[real blaze]

yes. and now this question still remains. why some syntaxes are not supported ? if it's not really old

[compact ingot]

there is standard 2.1 for newer stuff

[real blaze]

https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/operators/patterns#logical-patterns in this link for example, none of the logical literals are supported ( or and not ) . and it says they're from C# 9.0

[compact ingot]

standard 2.0 is probably C# 7.x

[real blaze]

oh

so it is old...

[compact ingot]

its the most compatible standard that still has sufficiently modern features... 2.1 causes all kinds of incompatibility issues with stuff that is stuck on framework472

[real blaze]

could be useful

[real blaze]

got it. thanks!

[compact ingot]

i'd suspect with 6.0 things will get better... up to now many considered .net core and 5.0 "unfinished" and didnt port their stuff over

[real blaze]

things you could do with C# 8.0 ...

public static RGBColor FromRainbow(Rainbow colorBand) =>
    colorBand switch
    {
        Rainbow.Red    => new RGBColor(0xFF, 0x00, 0x00),
        Rainbow.Orange => new RGBColor(0xFF, 0x7F, 0x00),
        Rainbow.Yellow => new RGBColor(0xFF, 0xFF, 0x00),
        Rainbow.Green  => new RGBColor(0x00, 0xFF, 0x00),
        Rainbow.Blue   => new RGBColor(0x00, 0x00, 0xFF),
        Rainbow.Indigo => new RGBColor(0x4B, 0x00, 0x82),
        Rainbow.Violet => new RGBColor(0x94, 0x00, 0xD3),
        _              => throw new ArgumentException(message: "invalid enum value", paramName: nameof(colorBand)),
    };

[tropic lake]

net core 2.1 is in one of the beta's or alpha's i think

once a stable version of that releases i'm gonna switch over because default methods for interfaces would be nice

[real blaze]

2021 , yeah

[real blaze]

yeah definitely . c# syntax as well as visual scripting ... im so waiting for 2021 LTS

[frank bolt]

Hey everyone! Do you know how to draw the basic inspector for custom PropertyDrawer ? I just want to be able to hide or show a variable when I use an enum

[flint sage]

PropertyField

[tropic lake]

there's #↕️┃editor-extensions

[carmine ermine]

are there any pros for making multiple prefabs for basically the same gameobject with different variations, instead of making 1 base prefab that you can load data into at runtime?
I have right now a creature base gameobject, and all variations of that gameobject should be more or less the same in functionality. they would differ in models, collider size, and a script that defines that creature and loads data(+maybe other things I can't foresee).
right now I feel like making a lot of prefabs is "easier" since you don't have to write extra code to load in data, and some stuff might be annoying to handle.
but I feel like 1 base prefab will be more efficient overall and scale better.

does anybody have some insight on this?
thanks!

[tropic lake]

i think this could be a candidate for scriptable objects

it would mean having a serialized field for the SO, then on Awake taking the values from the SO and turning that into the object

having many prefabs might not be scalable if there are many types and the variations aren't that large

[carmine ermine]

yeah, thats exactly what i thought

[steel leaf]

Does anybody know if async unity messages are now supported? I mean Awake and so on. Unity's Cloud Code example mentions internal async Task Awake() inside MonoBehaviour which seems very suspicious.

https://docs.unity.com/cloud-code/Content/examples.htm

[flint sage]

They are

[compact ingot]

iirc they can also return void and do not have to be Task

[steel leaf]

Does this mean that C# multithreading is a reliable thing now? Or do they use custom scheduling sort of thing. And which Unity version does support this code?

[compact ingot]

mind though that async await without Task.Run stays on the main thread like a coroutine would

[compact ingot]

Yes, custom

it has some issues due to a thread.sleep(1) each time await/yield is called

[steel leaf]

that is great, I mean this is usually not guaranteed with classic async/await

[compact ingot]

I’d like it to be easier to have background threads out of the box

but async await wouldn’t help much with that anyway

[runic tendon]

I have a State machine running in code. When does it make sense to use mecanim?
Like, say I have multiple possible parry States(like sekiro) with the only difference being the parry animation and what attack state it could enter from. Is mecanim better in my case?

[steel leaf]

From my experience it is better to design your system with playables and custom state machine.

We did a system with Animation Controllers at first but it quickly became very unmaintainable and full of bugs.

If you still want to keep using Animation Controllers, try making all of your transitions from Any State, not building complex graphs.

[rocky zephyr]

Sorry for the delay in replying. Others have chimed in but here's my comment: What Unity is providing in Pooling classes isn't a replacement for a comprehensive pooling solution like (for example) Pool Boss (which I use myself for other things). What I found Unity's pool classes good for is when you have a subsystem that needs to create and destroy hashsets, lists, class instances and the like which are created on-the-fly. For example, my tilemap extension has a runtime library that would create a stream of stuff for the GC to handle. Unity's pool classes make it easy to set up myriad small pools to handle each of these - and you can customize the behavior of each pool for reinitializing the old class instance when you get a recycled one from the pool and also you can customize what happens when you return an instance of something to the pool - for example: nulling out old references to GOs etc. So they're useful for many things - you could build a general-purpose pooler for prefabs etc if you like but there's plenty of nice commercial pooling solutions for that already.

[tough knoll]

Hm

Interesting, it seems like such a basic task that it's odd to include it at all, I was wondering if maybe it made use of some secret native code optimizations or something fancy like that

[dusk magnet]

hey yall, I posted this in general earlier but it got buried- I'm trying to find the gradient and/or contour vectors of a noise-based mesh for use with vector and/or tensor fields, and I haven't found any information online about how to do this- I understand that the tangent function does a little bit of what I want, but I don't know how to constrain it to the right direction or if there's a better way to do this

[misty glade]

I have a piece of code to update some UI members of a game object, and I want to centralize it and be able to call it from a variety of places.. I wrote the code but forgot nullable types aren't supported (yet) in Unity. Any suggestions for the following? I'd like to not make one method for each UI element that needs updating, but also be able to only update a part of the UI element:

        public void SetBattleUIStateImmediate(string? statusText = null, Color? battleStateColor = null, Color? iconColor = null, float? percent = null, string? hpText = null)
        {
            if (statusText != null) BattleStateText.text = statusText;
            if (battleStateColor != null) BattleStateText.color = (Color)battleStateColor;
            if (iconColor != null) Icon.color = (Color)iconColor;
            if (percent != null) BattleHealthBar.Percent = (float)percent;
            if (hpText != null) BattleHealthText.text = hpText;
        }

[untold moth]

Nullable are supported by unity though.

[misty glade]

They are? I thought that was a c#9 feature

I mean, I could just wrap this in a nullable region but i thought that wasn't supported..

[untold moth]

Nullable value types seem to be around since C#2

[misty glade]

https://stackoverflow.com/questions/64229527/customizing-csproj-in-unity-enable-nullable-reference-types hm.. appears so, I'll try this out

I suppose what I remembered (incorrectly) is that unity doesn't support the null coalescing operator, but nullable types are fine

[untold moth]

You might need to change null checks to .HasValue though.

Hmmm... A null check should work probably.

[misty glade]

Hm.. MS VS doesn't seem to pick up the directory buildprops file

not sure I want to sprinkle that annotation everywhere i use it

also.. treat warnings as errors gang signs

[untold moth]

You're trying to do nullable reference types. That's not supported in current unity version.

[misty glade]

aye that's what i wanted.. nullable ref types

actually.. what's the difference in

string? nullable;
string nullable;

aren't those both the same thing?

oh well i suppose this works for me

        public void SetBattleUIStateImmediate(string statusText = null, Color? battleStateColor = null, Color? iconColor = null, float? percent = null, string hpText = null)
        {
            if (statusText != null) BattleStateText.text = statusText;
            if (battleStateColor != null) BattleStateText.color = (Color)battleStateColor;
            if (iconColor != null) Icon.color = (Color)iconColor;
            if (percent != null) BattleHealthBar.Percent = (float)percent;
            if (hpText != null) BattleHealthText.text = hpText;
        }

[untold moth]

A string is a reference type, not value type. I'm not completely sure what's the deal with nullable reference types, since reference types can be null as it is.

[compact ingot]

They help the compiler with static analysis to better determine if a value can-be, is or maybe null so you get fewer null reference exceptions at runtime

[misty glade]

I have an item that's de/serialized on the server/client that has a "GridItem" - an item that has some values like X, Y, hit points, etc. When I send the message to the client, it contains a serialized grid item that the client synthesizes (into a new GridItem) using a Clone() method on the serialized copy of the grid item (so that I can GC/discard the rest of the serialized message).

A) Is it leaky to not clone it and just use the entire message's instance of the GridItem (I realize it's going to keep the message around until the GridItem itself is discarded, but that's not leaky, per se, just not really efficient)?

B) If A, is there a way I can enforce cloning of the enclosed GridItem so that I know I've made a mistake forgotten to new() up a local GridItem as I am supposed to? Code below for reference.

    [MessagePackObject]
    public class BattleActionCreateFieldItem : BattleAction
    {
        [Key(0)] public GridItem newFieldItemClone;
        [Key(1)] public Guid whichPlayer;
        ... other stuff unrelated to the GridItem ...
    }

    ... on the server ...
            BattleActionCreateFieldItem bacfi = new();
            GridItem newFieldItem = new()
            {
                Contents = (GridSquareContents)crewUsing.ActiveAbilityValue,
                Owner = actingPlayerGuid,
                MaxHP = GameLogic.GetStartingHP((GridSquareContents)crewUsing.ActiveAbilityValue),
                CurrentHP = GameLogic.GetStartingHP((GridSquareContents)crewUsing.ActiveAbilityValue),
                IsSummoningSick = true,
            };
            bacfi.newFieldItemClone = newFieldItem;

    ... serializes the "bacfi" and sends to client ...

    ... on the client, deserialize and process the "bacfi" ... 
        private void ProcessAddNewFieldItem(BattleActionCreateFieldItem bacfi)
        {
            GridItem newGridItem = GridItem.Clone(bacfi.newFieldItemClone); // This or just reference directly?
            Battle.Fields[bacfi.whichPlayer].Add(newGridItem);
            OnCreateFieldItem?.Invoke(newGridItem);
        }

[flint sage]

I don't remember but doesn't messagepack support deserializing into an existing instance?

We're not cloning the object but we're also not doing something uber performance @misty glade

[drowsy radish]

Hey, so have a bit of a weird problem, I have a camera separate from the main one rendering a weapon, that goes to a render texture to display on my GUI. But as part of the GUI, I want lines to point to certain areas of different weapons. I have all the positions for the points and references and everything, my problem is figuring out how to get the exact point on the screen from one of my transforms, because the actual item and camera are far off screen below the map, and only appear on the screen from a render texture, so something simple like just using WorldToScreenPoint produces inaccurate results. Any ideas what to do about this?

[long ivy]

if your two cameras render to similar size destinations (i.e. the render texture and the screen have the same dimensions), just use the gun camera's transform for the w2s transformation

[stable spear]

Use WorldToScreenPoint on the gun camera to translate the point on the gun to a point on the render texture. And then use the local transform of the render quad (or plane) to find the world point that it is being projected onto, and use WorldToScreenPoint on the main camera to find the exact pixel that corresponds to.

[drowsy radish]

I tried that initially, thought that it might work, but it rendered in the wrong spot, tho it was close. Tried as an experiment fixing it with transform positions of the rendered image etc. But not only did I have a problem when the screen resolution changed, I also had issues with the points being close, but not quite lining up with the actual positions.

[drowsy radish]

Hmmm so obtain the screenpoint from the gun camera to translate it, would it be worldtoviewport because the camera actually isn't rendering anything to the screen? or does screenpoint still work? Cause I tried viewport then multiplying by the width/height of the image renderer because my screenpoint seemed like it was giving odd results.

[stable spear]

That would be easier I guess. Viewport space is (0, 0) to (1,1). Screen space uses pixels.

[drowsy radish]

Ack, having trouble wrapping my head around how to convert a screen space to the right point using the transform of the render texture, to the convert again to screen space using the main camera, I guess time for a break to come at it later.

[stable spear]

is the render texture displaying on a quad?

[drowsy radish]

it is displaying on a GUI image

[stable spear]

oh!

can you show me the element it is displaying on in Inspector.

[drowsy radish]

this?

[stable spear]

hm

i think transform.TransformPoint will still work

but the result might need to be shifted

one thing you could do is compute this point, and then place a tiny cube there to see if it corresponds to the correct spot

[drowsy radish]

hmm I'll check that rn, I have a gizmo for displaying the point

[stable spear]

er, tiny square on the canvas

since it is on a canvas, you want to stay in canvas space

since I assume you'll be drawing the lines, etc on the canvas?

[drowsy radish]

Don't know if I'm using it right, I'm getting odd wild results. I also have the line position thing done so that points to the final position. Cause of issues for now I have it set to scale a few rect images to get the right position, and as far as I can tell, that part is working perfectly.

[stable spear]

start by placing something (like a small rectangle) on the canvas at the same location as the render. And then once that works, add the offset you get from WorldToViewportPoint, scaled by a certain amount. If nothing happens (the rectangle stays in the center of the render), you might need to increase the scale.

[drowsy radish]

I just did a double check, it seems that grabbing the render images transform.position gives me an accurate starter position.

Hmm actually managed to get a semi consistent result as a quick experiment, I'm concerned I brute forced it a bit, and had to scale by over 1700, and then recorrect with a vector3 after

            screenPos = imageRef.transform.position + camRefPos * scaleRef - camRefPositionCorrect;```

hmmm, maybe the extreme scaling came from converting to screen resolution? rather than the rects transform scale

[stable spear]

yea there's some conversion between the camera resolution, canvas resolution, scale. I'm running on no sleep, can't think about it too precisely

[drowsy radish]

hmm so I need like a canvas point to screenpoint conversion essentially

[kindred tusk]

var worldPoint = canvas.transform.TransformPoint(localCanvasPosition);
var screenPoint = camera.WorldToScreenPoint(worldPoint);

[tropic lake]

there are also some methods here for doing the reverse
https://docs.unity3d.com/ScriptReference/RectTransformUtility.html

[rugged sigil]

Hello guys. Instantiating object will also create new Instance of the material and I don't want that. How my objects can have same instance of the material on Instantiating?

[untold moth]

Afaik, you can't really do anything about it. Every renderer in the scene gets an instance of the material created for it.

[rugged sigil]

Hmm, that is less optimized compared to the example where they have same Instance of the material?

[untold moth]

Hmm... Maybe I'm wrong.

Are you accessing the instantiated object material via renderer.material?

[rugged sigil]

Yep, I need to change color

[untold moth]

https://gamedev.stackexchange.com/questions/179664/unity-material-changes-to-material-instance-on-start

That's why it creates an instance

You can do renderer.sharedMaterial to change color to avoid creating an instance of the material. But then all of the objects with that material would get affected.

[rugged sigil]

sharedMaterial still creates instances

And this is what I actually care the most

Since I have like 200 objects in the scene with different instances

[untold moth]

Donno. Maybe you're accessing .material somewhere else..?

[rugged sigil]

Yea, because I have to access them. I will try to make multiple Instances and instead of changing color I will try to assign new Instance of the material. Hope that this can work

Thanks for the help!

But yea, he still creates instance even when its not assigned

[pulsar isle]

I have a list of files I download from a server and instantiate them in the scene as GameObjects. All of them are disabled at start. Next I want to activate and deactivate one by one in a continuous order. with every time only one gameobject being active in the scene. The thing is the ammount of gameobjects is huge as seen in the profiler tab. This causes problems such as crashes when running on mobile. What are my options for object handling so the memory usage is lower/better managed ? Have in mind that every mesh is different from the other.

[raw path]

Hi there
does anyone know if accessing a singleton instance of a script is better by variable or by getter function?
Like if I had this

public static EnemyController instance;
private void Awake(){
  if (!instance) instance = this;
}

is it better to access it at runtime with a variable or with a getter function?

[austere jewel]

A property is better purely because you then cannot set the instance from outside the class

[raw path]

how come I can't?

[onyx blade]

if you dont have a setter, only a getter you cant set the variable by design

[raw path]

what I meant was for that bit of code which is better?
this?

private EnemyController EnemyController => EnemyController.instance;

or this?

private EnemyController _enemyController;
void Start(){
  _enemyController = EnemyController.instance;
}

for access in other scripts

[onyx blade]

you cant access private stuff in other classes tho

[raw path]

not sure if something is going on in the back regarding memory management or something when accessing this dozen of times at runtime

[onyx blade]

so neither will work

[raw path]

@onyx blade this is in a different script, not in EnemyController

the precise question would be, does it make a difference performance wise if I do/dont store the instance of a singleton in another scripts variable?

and instead access the instance with a getter

[untold moth]

It probably costs a tiny bit, unless the compiler inlines the calls. Even if it doesn't you probably not gonna notice any difference.🤔

[raw path]

fair enough

[buoyant spear]

Hey guys, I need some help with a problem the easiest place to start is asking if something is changed about the data when something is stored as a "string" compared to being used as ("say something");

[carmine ermine]

are you trying to understand when string objects are made?

or what

[buoyant spear]

So I have a script that types out text that I have as a queue, I added a list to the script and when I add the strings from the list to the queue, for some reason it doesn't type them.

[carmine ermine]

and i assume you call a function to also type out text from the queue?

like i dont really understand what your asking maybe im missing something

send the code maybe

[buoyant spear]

Okay, I guess I'm terrible at explaining. I don't know how to format script in discord

[carmine ermine]

copy it to hastebin

or something

then send the url here

[buoyant spear]

I was kind of just hoping someone had an answer about strings

[carmine ermine]

we cant know what to answer if you dont explain it well

[olive totem]

You're not being specific at all so nobody can help

[hard lily]

The problem might not be about strings at all, it might be about the logic of the script that is wrong

[buoyant spear]

Well the question was do strings store data differently to straight out typing the string in a variable? Because that is literally the only thing different between the scripts

[olive totem]

It's being stored the same

Something is wrong with the logic you applied but you're not giving any insight into this and I don't have a crystal ball

[minor girder]

Not sure if this is the right channel to ask things 😅
But i've made a chess game and I would like to be able to save the game at the end and then be able to watch a replay of saved games. Is there any good tutorial I can follow to figure this out or is someone willing to help? Thank you in advance!

[olive totem]

There are tutorials on how to save data but the other part of your question is very specific so the chance of finding something about this is very small

If I can make a suggestion, I'd say figure out first how to save things. Then find a way how you're going to represent the data you want to save. In theory you probably want to store information about every piece on the board and the moves it has made in a chronological order

[minor girder]

If this helps, it doesn't have to be like a video. Replay could possibly be like an iteration over the moves

[olive totem]

Yes I was thinking something like that myself and that would be the easiest way imo

[minor girder]

Yeah, that's kind what i want to achieve

[hard lily]

A video would be far more difficult and less intuitive than how replays actually work in video games

[minor girder]

Two simple ⏪ ⏩ buttons and it goes through list, and as you go, the moves change

[olive totem]

Saving data isn't the hard part itself though, it's how you're going to represent it is what might take some time to get down right

[hard lily]

Even games like Dota store replays as commands taken per tick

[minor girder]

thought so, any suggestions? I think the move list would be sufficient in this case?

[buoyant spear]

theres the code, it's pretty basic so it doesn't leave a lot for me to guess at, the only real difference is me adding strings to the queue compared to straight up enqueing("");

[olive totem]

Yeah probably for each move store which piece it was, it's initial position and it's target position, maybe even store how much time it took before the move was made if you want to be able to do a real time replay

[buoyant spear]

both examples that i tried with string variables don't work, for some reason it just doesn't read strings coming from the queue

they get added successfully to the queue, though.

[olive totem]

@buoyant spear You don't show what the typer does

[carmine ermine]

and the list works fine?

[minor girder]

yeah i am basically already doing that, just writing it on the screen

[olive totem]

Alright then you should be ready to find out how exactly you're going to save the data. You can save it as text such as JSON or XML or save it as binary files

[minor girder]

I'll probably return here once i figure the save first 😂 for some help with the loading

[olive totem]

The loading shouldn't be much more difficult than the saving

[minor girder]

I meant replaying 🤦‍♂️

[carmine ermine]

your idea seems solid, honestly

[scarlet copper]

@buoyant spear just an idea from me. Do that with scriptable objects, it will make your live easier.

public class ScriptableDialogue : ScriptableObject
{
    public string DialogueText;
    public string TalkersName; // Or just use another (scriptable)object reference that holds image, name etc.
}

public class ScriptableConversation : ScriptableObject
{
    public ScriptableDialogue[] Conversations;    
}

// your manager script
public Conversation : MonoBehaviour
{
    
    public ScriptableConversation conversation;
    public Queue<ScriptableDialogue> dialogues = new Queue<ScriptableDialogue>();
    
    void Start()
    {
       foreach(var dialogue in conversation.Conversations)
       {
           dialogues.Enqueue(dialogue);
       }
    }
}

OR if you don't reuse dialogues some where else just use normal class:

[Serializeable]
public class Dialogue
{
    public string DialogueText;
    public string TalkersName; // Or just use another (scriptable)object reference that holds image, name etc.
}

public class ScriptableConversation : ScriptableObject
{
    public Dialogue[] Conversations;    
}

// your manager script
public Conversation : MonoBehaviour
{
    
    public ScriptableConversation conversation;
    public Queue<Dialogue> dialogues = new Queue<Dialogue>();
    
    void Start()
    {
       foreach(var dialogue in conversation.Conversations)
       {
           dialogues.Enqueue(dialogue);
       }
    }
}

The pro is you can configure the asset in the asset folder directly without working on the scene itself.

[queen plover]

I like to display dialogue in an editor graph, because when you introduce choices into something like this you'll end up with a horrible list to edit (though a graph can get equally messy)

[scarlet copper]

Should be possible here too I think. I never made a graph but for my own event system I would like. Had no time yet for it.

[queen plover]

I used XNode to make mine (https://github.com/ThePatchWorkDev/DialogGraphs), but It's definitely not the best code that can be found

Especially since it was built for a text-conversation kinda deal- rather than full on animations and storyboard

[regal olive]

Anyone getting this error in 2020.3.21f1?

ArgumentException: Could not find item with id: 0. FindItem assumes complete tree is built. Most likely the item is not allocated because it is hidden under a collapsed item. Check if GetAncestors are overriden for the tree view.

Happens whenever I double click on any item in the Hierarchy. It's triggered by the FindItem call.

Only happens in a specific project.

git clean didn't work.
Affects all scenes.

[kindred tusk]

@regal olive try resetting your layout.

[regal olive]

@kindred tuskokok

just choose anything here?

[kindred tusk]

"reset all layouts"

It's also possible it's coming from some custom editor window

I've never seen it, but I'm not on that specific patch version of 2020.3

[regal olive]

okay, that fixed it
what a weird bug

[kindred tusk]

It's just some bug in the editor UI

[regal olive]

rip

[minor girder]

well hello i have returned. i've achieved the saving and loading and now it's the matter of making the replay/autoplay/rewind w/e functionality.

[leaden jungle]

@minor girder just saving a list of moves should be enough to recreate the game that was played. I've even seen chess games where you could take back moves and change to something else. The only way to recreate it this way though is to do it turn by turn, which should not be difficult.

[minor girder]

yeah, i've saved the move list , now i need to figure out how to make it play move by move

[leaden jungle]

well you could do it by using the same functions you use to move your pieces in the game, but supply the information from the move list

[minor girder]

yeah, i'm looking into that, just thinking should i do it in a separate scene maybe?

[sturdy edge]

probably for loop with a coroutine

[minor girder]

not sure what do you mean?

[leaden jungle]

I would say not a loop because you want to be able to do one step at a time.

I would have an index variable for your moveList, int currentMove and then have it so your move method for when a player moves, make it so it can take parameters that would be supplied moveList[currentMove]

[minor girder]

yeah, i don't think loop would work as I would need to go in both directions

[misty glade]

Thanks for the replies, and sorry for the very late response, had to dip out for an emergency overnight. I couldn't find anything related to deserializing into an existing object, but I'll dig deeper. Thanks

[misty glade]

When I think about it more, it's not terrible to have one of these messages floating around for the life of the item that it's pointing to. The overhead is really tiny (like 10-20 bytes) and they'll get GC'd when the underlying item is discarded through gameplay, so it's ... gonna just be fine, I suspect.

@minor girder : One design pattern I've implemented for my game is the concept of a "BattleAction" (with a number of derived classes, each for one kind of battle action) that, if played in order, could recreate a game state entirely from the starting state. You could probably do something similar to that, if you needed replay/rewind/etc functionality.

My use case is a tiny bit different than yours (I use the battle actions instead of just sending the "game state" because I want to animate the actions one by one) but .. you could have rewind functionality pretty easily with an approach like that.. Additionally (and this is giving me feature ideas for my own game) you could pretty easily save the entire game and play it out later

I do have the ability to send the comprehensive game state but I don't do that often (only at the start of a battle and on-demand while developing and debugging state desync issues)

(the comprehensive game state for my game is 1700 bytes or so - too large to send repeatedly - but each battle action is mostly 20 bytes or less, when serialized)

[minor girder]

could recreate a game state entirely from the starting state
That is something I need! I am saving a list of moves from start to end and I want to be able to "watch" the game move by move

[misty glade]

OK.. lemme prep some source code and toss it in pastebin for you so you can sorta wrap your head around how i'm doing it for my game

[minor girder]

Just one question, would it be possible to modify it to go backwards and forwards in the moves?

[misty glade]

bear in mind I'm using MessagePack for serialization so there's some attribute annotations in there you might not recognize or whatever, but the basic idea is that i'm just serializing it into a byte[] and deserializing it on the receiving end

i mean, probably? i just haven't built that into my game - that's some logic you'd need to figure out and build for yourself, depending on your game

[cursive horizon]

is the game multiplayer?

[minor girder]

my chess? yeah, but local mp, no AI implemented

[cursive horizon]

I meant Sharping's (i'm curious why there's that byte serialization/deserialization phase). You probably won't need that if you're doing local MP, unless it's serving some other purpose which would be interesting!

[misty glade]

yes, my game is multiplayer

i just grabbed these snippets super quick, if you have any questions, don't hesitate to ask

https://pastebin.com/j389Hd4X Battle Actions (things the server has decided "happened")
https://pastebin.com/rvPkFw2g Player Actions (things the player(s) have indicated they want to do)
https://pastebin.com/aWXgQPSc The primary server processing dispatch - this takes the PlayerActions and "does stuff with them"
https://pastebin.com/P9Agw4Z5 The client side processing dispatch - this takes the BattleActions sent by the server and "does stuff" - animates, updates local state, etc

lemme take a quick snap of gameplay so this maybe makes more sense

Basically a player can move grid items around on their grid to try to merge them up into stronger items; move items onto their field where they can attack enemy crew (which is how you win), use active abilities, etc

[cursive horizon]

thanks for sharing this btw! hard to find real implementations for stuff out there

[misty glade]

but it's all based on a series of Player Actions and Battle Actions modifying a Battle state object

theoretically i could rewind/fastforward/replay an entire battle with just those items

pardon the developer art 🚧 we're still hiring our UIUX and art guys

[minor girder]

I will def share mine once I figure mine out

seems like something people could use

I appriciate this a lot kind person! I will look into it later!

[leaden jungle]

@minor girder maybe look into the command pattern with an undo/redo ability

[urban warren]

What would be the 'best' way to access items from a collection in a random order without getting the same item twice?
I can think of several ways, but was hoping for some insight.

I could generate a random int between 0 and Count, and keep a HashSet of the indexes already used, checking if it already contains the generated number, if it does, regenerate it. This feels more like a brute force approach though.

Could generate a random int between 0 and Count, removing that item from the list. However this doesn't let you 'start over', so if I wanted that I would need to keep a second list that the items are added to once they are accessed.

The only other thing I can think of is maybe some type of cool math thing that I don't know about?

[cursive horizon]

I think generally you would make a 'deck' of selections and pull out items as you go. This also lets you add more of certain items to tweak their rarity.

It depends what you need for your game, really. You could also keep track of how often each option has come up and increase or decrease its chances based on that

[urban warren]

So you mean 'pre'randomize the order of the items in the collection, then get them linearly?

[cursive horizon]

this is a good article on it

https://www.ellahoeppner.com/making-randomness-fair/

[cursive horizon]

yeah, build a deck and shuffle it

then draw as many cards as you need

ultimately it's more of a design question than a code questoin

[urban warren]

Right, I'm not looking for probability based randomness. Just straight up plain randomness.

[cursive horizon]

then i'm not sure what you're looking for i guess

[urban warren]

My specific use case is that I have a number of flowers on a GameObjec that can be eaten by a creature, I want to just select which flower to 'eat' randomly.

[cursive horizon]

but you don't want it to eat the same flower twice?

[urban warren]

Exactly.

[cursive horizon]

why not?

[urban warren]

Because it has been eaten...?

[cursive horizon]

ah, so your problem has nothing to do with randomness, and everything to do with keeping track of which flowers are available to be eaten

[urban warren]

No, please reread my original question. I want to randomly get an item from a list, but never the same item.

[cursive horizon]

are you sure?

it's not that you don't want the same flower twice...it's that you only want to eat flowers which aren't already eaten

you should have a list of flowers which are edible and a list of flowers which are eaten, and you can pull from one and place in the other accordingly...or you could have a list of flowers with some kind of eaten boolean which you toggle, and filter them before selecting

regardless, your 'randomness' should just be 'select a flower from the list of edible flowers' and the thing you need to decide on is how you want to manage that list

[urban warren]

That is one of the things that I said was an option in my original question...

[cursive horizon]

right, which is why i said i'm not sure what you're looking for

you asked what the 'best' way is, so we talked about some of the context around what you're doing...but there is no 'best' without some more explicit values to judge against

[carmine ermine]

whats the problem with shuffeling the list once and then getting them in order?

[urban warren]

There is nothing inherently wrong. The problems I see with it is that adds additional overhead to each instance, and if I wanted a new order the items would need to be shuffled.

[cursive horizon]

solutions always have tradeoffs

it's not a difficult implementation to swap out either way, right?

i guess the answer to your question is 'it probably doesn't matter'

[urban warren]

Right, I know there is no actual best way. But I thought it seemed fun to try to make a really efficient system, even if it ultimately doesn't have that much impact on performance really.

[cursive horizon]

that's cool, but the most efficient system is always the one that fits into its context most efficiently

that's kind of what i was getting at originally

the most efficient representation of this is the one that your game already needs to have for gameplay reasons

because you need that anyway

optimization is something you do for specific cases because otherwise...what are you optimizing for?

so we need to build those specific cases first

[carmine ermine]

maybe @compact ingot will have a random formula for your case 😂

[urban warren]

...But this is not a complex situation, it is pretty textbook for 'access random unique index in collection'.

[urban warren]

Haha, maybe they will, they seem to know these types of things.

[carmine ermine]

if you eat a flower, shouldnt you take it out of the collection?

or is it still there

[misty glade]

i just built a quick randomizer for a list if you don't care that the list stays sorted

You could shuffle your list then just eat and RemoveAt(0)

[urban warren]

It is there, but dead, and I think I will want to allow the player to 'regrow' it so I don't want to lose it completely.

[misty glade]

           List<GridItem> ret = Enumerable.ToList(_gridItems.Values);
            var count = ret.Count;
            var last = count - 1;
            for (var i = 0; i < last; ++i)
            {
                var r = NumberUtils.Next(count-i) + i; // 0-9 on first call with count=10; 1-9, 2-9, 3-9
                var tmp = ret[i];
                ret[i] = ret[r];
                ret[r] = tmp;
            }

(replace ret with your list)