archived-code-advanced

[jolly token]

Much smaller than single texture atlas

So you don't really have to worry about having them on memory
What you can worry about is GC if you let them unreferenced

[sterile snow]

It probably matters a lot less than my mind assumed it would be

But aside from the character and stage that were selected by the player, I imagine it'd be a good idea to get these instances cleanly gotten rid of

At the very least tbh

[jolly token]

You probably need them later in game, don't worry about it.
You can unload the unused assets, but the game data is fine.

[sterile snow]

Fair enough, least I got some advice on the matter. Much appreciated.

[jolly token]

Okay I reduced it to tolerable level 😌

[undone coral]

this is fine

if you feel more comfortable describing this stuff in code than in the editor - i would have done it in the editor - well this is #archived-code-advanced , you are capable, you can do that

[sterile snow]

Well, good to know it's fine!

[muted ridge]

I have, and im using windows 10. the dll works fine is visual studio. im just trying to use it with unity. the available assets for kinect are rare and very old.

do u have one that uses unity? (sry for the late answer)

[young wraith]

Hi anyone

I'm making a clean up game

but line 147-150 makes the game slow

https://codebeautify.org/alleditor/y22250370

How can I enhance that line of code

[flint sage]

Do one getpixels call instead of so many individual ones

But the real winner is using a shader and graphics.blit

[undone coral]

this is a tough one, sorry.

you can't just copy and paste that one DLL

i think you'll get a little more experienced with this stuff and it will be easier

[muted ridge]

well its for a school project, internet says its as easy as copy pasting the dll but we got decieved once again xdd

[undone coral]

you can try to copy the right SDK content into unity

it's not going to be that one DLL

the DLL itself has references to other DLLs, and the runtime tries to find those via a variety of strategies

it's telling you it can't find something

i think if you configured the error messages correctly, it would tell you what that is

it may already be telling you in the editor

that said, there is an official, modern Azure Kinect device, and a modern nuget package, which may be easier for you to use

[muted ridge]

thats what i was thinking, could there be a way where i can check which other dlls its connected to and then paste them in the editor aswell

this is the error im getting

[muted ridge]

i might try this as well tonight

[undone coral]

there are tools that can tell you what managed and native DLLs something references

it might be easier to copy the entire SDK directory into your Assets/ folder for now though

that might Just Work

you should also use .net framework 4 as your ... whatever it's called

i hardly remember anymore, but you will change it from .net standard 2 in your player settings

[muted ridge]

thanks a lot, ill try it out when i got time !

[exotic gull]

Trying to use Newtonsoft Json in my project. Any idea what's causing this? ```cs
class Control
{
[JsonProperty("look_x")]
public float look_x { get; set; }
[JsonProperty("look_y")]
public float look_y { get; set; }
[JsonProperty("jump")]
public bool jump { get; set; }
[JsonProperty("arm_left")]
public bool arm_left { get; set; }
[JsonProperty("arm_right")]
public bool arm_right { get; set; }
[JsonProperty("forward")]
public bool forward { get; set; }
[JsonProperty("backward")]
public bool backward { get; set; }
[JsonProperty("left")]
public bool left { get; set; }
[JsonProperty("right")]
public bool right { get; set; }
[JsonProperty("dead")]
public bool dead { get; set; }
[JsonProperty("reload")]
public bool reload { get; set; }
}

class HFFInput
{
public static Control hffControl;

public static void init()
{
    hffControl = new Control();
}

public static void unpackJson(string json)
{
    hffControl = JsonConvert.DeserializeObject<Control>(json);
}

}```

[maiden turtle]

provide all of the other context

[exotic gull]

Working with the BepInEx plugin injector to inject code into a game at runtime (modding). To the best of my knowledge, the game uses Unity 2017. I've included the Newtonsoft Json and System.Numerics dlls in the plugins folder to be injected along with my code.

And in case you're wondering, the mod is really only applicable in a singleplayer setting, and a lot of the community is supportive of the development of this mod. (The only reason I say this is because the last time I asked for modding help here, people asked to make sure I wasn't doing so for nefarious reasons before they provided help)

[regal glade]

i'm trying to figure out multi-threading, especially using Parallel.For from System.Threading.Tasks. i just wanted to ask since it's running way worse than a normal for loop, and before I waste 20 hours reading, is this the way to do it in Unity, or does Unity prefer some other method for running a Parallel.For loop? I'm seeing some stuff about ParallelFor jobs but not sure that's what I should be looking at.

[compact ingot]

How is it way worse? In unity you can use the Job System to schedule parallel workloads with full engine optimization, but parallel tasks work too. Generally running parallel code introduces some overhead, so it will always be slower than single threaded code but it can do more work in the same amount of real time.

[regal glade]

i'm trying to learn it so i'm not sure i'm doing it right, and the for loops i'm attempting it on are long and complicated, but it simply runs way worse when i run it so probably something is not done right.

[compact ingot]

Can you specify what ‘way worse’ means? How do you profile the performance?

[regal glade]

simply checking the frame rate in both the editor and in a build, it drops from 120-300 fps to 80-120 fps

[compact ingot]

so you run those tasks every frame?

[regal glade]

yes

[compact ingot]

well that’s not going to be performant indeed

you should run worker tasks independent of the update loop

[regal glade]

why is it so bad to run a Parallel in the update? wouldn't you do something similar for multi-threading pathfinding for example?

[compact ingot]

let them finish whenever they do and use the result when they are done, launching new tasks every frame is wasteful

[sly grove]

Unity prefers the job system where possible

what exactly are you trying to multi-thread?

[regal glade]

oh boy. well, i've created a 3D tile system where each tile is represented by a 3D array of Tiles, and each frame where the camera moves I run a for loop to iterate through the visible tiles and create a "slice" of the world represented by values in a Color32 array that I then send to a shader via a Texture3D

i say 3D but this is still very much 2D

[compact ingot]

That’s really a longer discussion about how multitasking works on multicore CPUs

[regal glade]

so i have a 3 for loop deep nested loop (x, y, z) that does some checks to look for walls, doors, etc to determine render states

it runs fine even on old CPUs but it's still by far my heaviest process so I wanted to attempt to multi-thread it if possible

[compact ingot]

This will be quite a challenge to multithread effectively

[regal glade]

yea maybe

maybe i should look at this job system first

[compact ingot]

the job system will guide you a bit more

but you might end up figuring out that true performance improvement would require very specific data structure that drives the computation and parallelization, which ultimately leads to data oriented programming/ dots

[regal glade]

yea just gotta figure out what that is first 😂 sorry multi-threading is completely new to me

[compact ingot]

It remains that way for most people

[regal glade]

been reading all day but still very confusing

[compact ingot]

not something anyone can master in a few days

[regal glade]

just need to find some good resources that isn't MS docs

[compact ingot]

How theoretical/generic do you want it to be?

[regal glade]

i don't mind as long as it's not MS docs. MS docs only works for people who already know most of it, it's like reading math on wikipedia.

any resources are good resources, can always learn bits and pieces here and there

if you have any recommendations i am all ears

[compact ingot]

I’d work through a mooc on the subject, there is a quite nice and often used method of incrementally teaching it

in textbooks that are written to accompany such a course this same explain would also occur

The keywords to look for are concurrent and parallel programming

[regal glade]

i don't really like going through courses but i'll consider it, thanks for the tip

[jade jackal]

☝️

they are extremely hard to navigate unless you already know where everything is

[tribal pivot]

Have you profiled this right now, why do you think multithreading is the way to go? Is it even possible to multithread

[regal glade]

@tribal pivot i did profile it and the function takes like 5-6ms, and it's iterating through 3 axes of tiles in a matrix so it might be like 80*60*50 entries. like i said it runs ok at the moment but i wanted to look into multithreading it, it just felt like it made sense to offload some of the 240 000 loops to other cores

because i am also going to add more on top later

i say "matrix" but it's a 3D array

all the iterations of the for loop are independent really so i don't need them to return in any specific order, i just don't know why it runs badly yet

maybe they're fighting over access or something, or the parallel overhead is actually simply heavier than the loop

that's what im trying to find out

not to mention thread safety

[north thorn]

hey, when using the xr rig, is there a way of getting possibly position (vec3) and rotation (qt) for each eye? i know that when using ovr, the eyes were already there named as left and right camera anchor. here with xri i get just the center camera and for my project i kinda need both right and left ones.. help pls

[plain gate]

Could anyone help me converting a counting sort algorithm to be in place?

[sly grove]

why not just look up an in-place algorithm

[compact ingot]

iirc that’s not really possible, best you can do is fake in-place where you get the results in the input array but you aren’t saving space since you definitely need 2 arrays

[plain gate]

What two arrays are you referring to? The count array is 1 and the elements to be sorted is another. I am just trying to work together to understand.

[compact ingot]

Those two

[plain gate]

But if it is not in place you would need 3 arrays. The input array, the count array, and the temp output array.

[compact ingot]

in-place means that you don’t need extra space besides the input array

[plain gate]

correct this is not "true" in place but my count array is very small and doesn't really impact.

[compact ingot]

sure, but you asked for in-place and it sounded like a theoretical question which generally has little to do with what matters in practice

[sly grove]

this is not Unity related

[young wraith]

the problem with calling one getpixels is that it will not count all the cleaned amount

[long ivy]

you can't make an in-place counting sort unless you want to do something hacky (like using upper word/lower word if your value range is small enough)

[hollow socket]

I can't tell if this goes in advanced or beginner but I'm struggling to find a way to find the difference in the way an object is facing to an object

like a number of how far off the rotation is to facing the object

[brisk pasture]

the difference in what unit

[hollow socket]

degrees

[brisk pasture]

Vector3.Angle should work

compare your transform.Forward with a vector you create by subtracting position from the target object position

[hollow socket]

Ok

Thanks

[brisk pasture]

https://docs.unity3d.com/ScriptReference/Vector3.Angle.html

the example is exactly what you are trying to do

[hollow socket]

wow

[brisk pasture]

if the angle should only be around the Y axis you can use Vector3.SignedAngle the exact same way too

jut pass it Vector3.up for its 3rd arg

this will also let you know which direction the angle differs in as well

[hollow socket]

Thanks

[young wraith]

Hi can someone help me https://stackoverflow.com/questions/74069108/getpixel-makes-the-game-slow?noredirect=1#comment130777290_74069108

[compact ingot]

you are uploading a new texture to the GPU each frame, this is very inefficient

[young wraith]

actually the getpixels are the problem

When I try to remove the code Color pixelDirtMask = _templateDirtMask.GetPixel(pixelXOffset + x, pixelYOffset + y);

[compact ingot]

you should also not get/set individual pixels

it says so in the docs

[young wraith]

How can I know the remaining dirt to be cleaned if I don't get individual pixels?

[compact ingot]

you'd load the whole texture into a buffer, and process the changes in that buffer, or use the raw data to do the same

[regal glade]

I've managed to get pixel getting and setting running pretty well every frame, but it depends on the size of the texture. The best method I have found so far is using get and setpixels32 and then adjusting the color32 array

[compact ingot]

^this (multi pixel methods) also works

but editing textures like that is very inefficient regardless

what would be 100x better is doing all this with a render texture

or a compute shader

[young wraith]

can render texture compute the remaining dirt to be cleaned?

[young wraith]

can it compute the remaining dirt to be cleaned?

[compact ingot]

do you need pixel accurate info on what is dirt and what is not?

typically you'd have one datastructure that holds the info on what has been cleaned (a low resolution texture or plain array) and another one that handles the display of the progress in a nice way

[young wraith]

yes I need to know the remaining dirt

[compact ingot]

--> screen pixel accurate <-- is the keyword here

[young wraith]

ok i'll try to research about dat

[regal glade]

Is this 2d or 3d

[young wraith]

2d

[regal glade]

Then even if you have a 1080p resolution a 2d array should be fine to determine what's on the screen. You can make the game logic alter the array contents and only draw it every frame with no get calls.

Alternatively store the status of the dirt in a color32 array only and alter that

Then you can just do setpixels32 directly

[young wraith]

Thank you for ur inputs I love u all

[regal glade]

Since the screen has 2 axis and the color32 array is 1d you have to do something like "x + (y * xSize)" to access it right in a nested for loop if that makes sense

On phone

[young wraith]

ok i'll study about color32 array

[copper steeple]

Hello! idk if this is a code-advanced question but i didn't find anything useful in google so yeah. Anyway, the question is: How can i create custom properties on the inspector using fieldInfo?

            foreach (FieldInfo fieldInfo in myCustomType.GetFields(BindingFlags.NonPublic | BindingFlags.Instance))
            {
                EditorGUILayout.PropertyField(fieldInfo/*do magic with fieldInfo?*/)  ;
            }

[flint sage]

Use SerializedProperties instead of fieldinfo

[copper steeple]

but how? idk how to, all i can get is fieldInfo or propertyInfo

in fact, that's the main problem

[flint sage]

CustomEditor?

this.serializedObject.FindProperty

[copper steeple]

i can't, i don't have an instance of the object i want to get the fields of

and please tag me so i can respond faster

[flint sage]

Show more context then, what are you actually doing

[north thorn]

probably nobody right?

[copper steeple]

i have a custom class that contains information an entity should have: property. Let's say Shield. A shield property should have : maxShield, shieldSprite, ShieldColor etc. I want to tweak the values of the shield property for each prefab; i also want to do that in one place: the entityBuilder AND i want to make it generic so i don't have to change the code each time i add a different property. That's why i'm searching for all types that inherit from the abstract Property class. For each iteration, i want to get the property fields and draw them in my entityBuilder so i can edit and store the data

maybe i'm doing it wrong, but at this point i just wan't to know if my solution is even possible

bellow should be a SaveProperty Button

[modern storm]

Is there a way to force an update of a transform's position and rotation so it is reflected in the hierarchy immediately?

[young wraith]

whats the xSize? is it the brush width?

[winter yacht]

How do you know it's not reflected in the hierarchy?

[regal glade]

Since the screen is both x and y but you are using a 2d color array you need to multiply the y value by the maximum size of the x axis, so 1920 for a 1080p resolution

So when y goes up to 1 we add a full loop of x to the index

[modern storm]

@winter yacht I use an active ragdoll and when I enable the gameobject I play an animation, the animation rig's transform are immediatly set to the right positions, but the active ragdoll's rig takes an extra frame to set itself even though I go through all the transforms to set them to match the animation rig's position and rotations.

[regal glade]

This is within a nested for loop

[winter yacht]

Just a thought... maybe have the meshrenderer disabled and only enable after setting the positions?

Basically hide the first frame somehow.

[modern storm]

@winter yacht If I Debug.Break right after enabling the object, starting the animation and setting the transforms, I see the transforms of the ragdoll are not at the positions I set them. The renderer is enabled.

Yes I do this already but it causes other issues. Those ragdolls replace entities rendered with DrawMeshInstancedIndirect when I need them to interact with physics. I was hoping to find a way to get them ready to receive a beat up on the same frame I enabled them.

I was hoping to find something like Animator.Update() for transforms to force an update

Does anyone know when changes to a transform is updated in the hierarchy? Is it during FixedUpdate ?

[compact ingot]

what exactly are you trying to assert? changes happen immediately, effects of a change may appear only during the next update/fixed update depending on what you are looking for

[modern storm]

@compact ingot This shows the problem well, In this image the green rig are the transforms used by my active ragdoll, the red one represents the animator's rig. This screenshot is taken just after I enable the gameobject, play an animation with animator.Play() and run through all the transforms of the green rig and set their position and rotation to match the animation rig (red). Then I did Debug.Break to take a shot.

Somehow the animator manages to set the transform of it's rig on the same frame the animation starts playing but the changes to the green rig are not reflected untill the next frame

[pale pier]

Anyone have any idea how this audio mixer group selection UI is made? I want to make UI like it to be able to nest items inside each other

[winter yacht]

You are taking into account that setting some transforms in the green rig will affect their children?

[modern storm]

@winter yacht yes. The position of the green rig at that moment is the position it was at when the gameobject was previously disabled.

as in the animation that was playing at that time

Since I'm replacing entities with the dolls, I need it to be ready on the frame it's enabled. Thats why im trying to figure out how to force the transform update

somehow the animator manages to force that update

[lament salmon]

hey I didnt read your whole convo but theres this setting
https://docs.unity3d.com/ScriptReference/Physics-autoSyncTransforms.html

might help/be related

[modern storm]

I will check it out ty

[lament salmon]

actually that doc kinda says that it shouldn't be used... maybe this method is more relevant
https://docs.unity3d.com/ScriptReference/Physics.SyncTransforms.html

[modern storm]

That's worth a try!

[modern storm]

@lament salmon unfortunatly that doesnt work 😦

[reef lance]

Anyone familiar with CIL, can I use call instruction but pass a lambda instead of a static method and if yes how?

[twin vector]

https://cdn.discordapp.com/attachments/883435300880261120/1030851573087227964/4465d352-8629-4f00-8177-fc9aac4350ea_15-10-2022.png

by looking at the IL it looks like call is used for methods
and lambdas are used with callvirt

https://cdn.discordapp.com/attachments/883435300880261120/1030852016601305088/2a67603e-166f-4ff0-b8aa-d2a090e70b78_15-10-2022.png

You could generate a display class and call that if u really want calls but that might be counter to what you'd want if you're looking at CIL

[flint sage]

If you're generating IL yourself then iirc you can but I doubt you can coerce the compiler to generate it for you

[hot gale]

Hey so, I'm wondering if something like this is possible. I have a Height Map and a Texture Map for a Sphere that are mapped Cylindrically

Is there any way to project the heightmap onto a sphere and then apply the texture so that I can create a planet type object?

[flint sage]

Sure it's just math

[hot gale]

Yeah, I meant more with Unity's built in stuff.

[light falcon]

Hi, quick question; can you save json in a custom file type? Some games have save games called 'Savegame.data' or some other custom extension. I would like to know if it's as easy as changing the file extension

[flint sage]

Yeah sure

Just read it as text and it doesn't matter what the extension is

[devout hare]

Yes, you can name the file whatever you want. The file extension isn't the same as file type.

[flint sage]

Extensions don't really mean anything, it just helps the OS to decide what application to open

[jolly token]

Saving raw json is just changing file extension. However other game's custom format probably have more obfuscation and encryption than raw json.

[light falcon]

Thanks folks!

[safe vapor]

What would be a good way to go about making a "terrain smoothing" brush? I'm making an in-game level editor. I want to make something that has the same basic behavior as the Unity Editor terrain smoothing brush (I already have all the other editor brushes implemented).

anyway, I guess one good way is to make a loop that starts on the outside of the selection circle and goes in a spiral. For each vertex, it raises or lowers the loop iterated vertex in whatever direction the average height of those 5 vertices is. That's probably the best way but to make it go in a circular spiral pattern, that would involve some complicated trig that might introduce rounding bugs

If there was a way to use the editor's actual smooth height function, that would be great but I haven't found it, if its possible

[undone coral]

the smoothing brush measures the average height within the brush's borders, then raises or lowers terrain to that measurement within the brush

[safe vapor]

oh is that all it does? Wow ok I was trying to make it more complicated than it needs to be

[atomic void]

Hello,

What do I do when I have big list of positions and I want to draw instanced meshes on these points?

Can I just send a Vector3 list for positions to the gpu?

I am currently sending transform matrices and using Graphics.DrawMeshInstanced

[sly grove]

positions are part of the Matrix4x4s

DrawMeshInstanced only takes matrices iirc

[atomic void]

Yes, thats the case

I know what transform matrices do
DrawMeshInstanced didnt let me have more than 1024 matrices

[sly grove]

correct 1023 at a time

You have to batch them

[atomic void]

And I was thinking that it is not efficient sending any more information other than positions to the gpu

Because rotations and scale dont change in my case

[sly grove]

If you want to get more efficient you'll want to look into DrawMeshInstancedIndirect

In which case the position data will come from a compute buffer

[atomic void]

Will I be able to have more than 1024 with a standard shader?

Now that I think, I'll hit the materials input buffer limit

[sly grove]

If you use DrawMeshInstancedIndirect you'll need your own shader which reads position data from your compute buffer

[atomic void]

If that 1024 is because of the materials input buffer size limit

[sly grove]

This is a really good guide on it:
https://toqoz.fyi/thousands-of-meshes.html

[atomic void]

@sly grove allright thanks

[kind sigil]

I want to create a custom attribute [RadiusGizmo] that I can slap on any float field and as a result I want it to draw a circle with the given radius as a gizmo in the scene view around the selected game object. Where do I start?

[dapper nest]

I cannot wrap my head around how to implement this. I'm trying to implement movement so that when the player is on a slope that is "too steep", the input direction should be translated so that it does not move up it. As if you set the movement direction into the slope (even with 0 at Y axis), the physics system will still push it up (unless it is a 90 degree wall). Any ideas how to implement it so that the movement on a steep slope is the same as if it were a 90 degree wall? I made a very artistic visual representation in case I can't explain it well enough in words

Here is the current logic for sloped movement, but because I want my air acceleration high, it just shoots up on steep slopes

private void SlopeMovement()
        {
            // Calculate desired velocity based on direction
            Vector3 desiredVelocity = m_desiredDirection * m_airSettings.MaxSpeed;
            desiredVelocity = Vector3.ProjectOnPlane(desiredVelocity, Vector3.up);

            // Determine if we should accelerate or decelerate
            float speedChange = (m_desiredDirection == Vector3.zero) ? m_airSettings.Deceleration : m_airSettings.Acceleration;

            // Find new velocity and apply it
            Vector3 newVelocity = Vector3.MoveTowards(Motor.Velocity, desiredVelocity, speedChange);
            newVelocity.y = Motor.VerticalSpeed;
            Motor.SetVelocity(newVelocity);
        }

sorry if this doesn't go under "advanced" code, but no one seems to have the answer elsewhere 😅

[silver pendant]

How about adding an invisible 90° wall? It's kinda dumb but it could be an easy solution.

[dapper nest]

And collide with the air every time there is a steep slope... ?

I am asking here because there has to be some way to calculate it, not ruin the game experience

[frosty violet]

Hello everybody. Does anyone know how I can convert a value from a linear range [1, 0] to a value in a logarithmic range [1, 0]? thank you

[forest isle]

Seems you are working against the physics engine. If you want to do this manually, perhaps when the player collides with the slope, check the slope contact.normal. If Vector3.Dot(normal, player.up) > 0.8f , limit y movement. https://docs.unity3d.com/ScriptReference/ContactPoint-normal.html

[dapper nest]

This really does not solve anything though. As for limiting velocity added to y, I am already doing this. And you cannot limit the y movement either, as any time you come into contact with a steep slope, it will stop your y velocity instantly, which feels very bad.

[flint wraith]

Math.log10(value) + 20 i think

[flint wraith]

Check this video

https://youtu.be/pbuJUaO-wpY

[livid kraken]

https://assetstore.unity.com/packages/tools/physics/kinematic-character-controller-99131 with this asset being now free I can not recomend it enough. Look at the code even for learning purposes.

[forest isle]

Oh, I see what you meant now. Your rigidbody drag is so high, the player won't fall down easily? Seems like this would be solved by playing with the physics parameters. Maybe instead of limiting the y movement / velocity, just increase the rigidbody gravity then, when it's in contact with a slope. It kind of depends on how you actually are moving the player and the physics setup

[frosty violet]

Thank you.

[flint wraith]

Dont know exact minute but look for slider part in code also might need to modify it

[frosty violet]

Yes, I got it. I never used audio mixer, so I simulated the exact behavior of this component by hand. By watching this video first, I would have saved a lot of time.

[flint wraith]

Oh you were doing just that thing?

Hahahahaha

[frosty violet]

Yes, instead of using a slider value, I am using the player distance with the audio source to create spatial audio.

[dapper nest]

Thank you for this resource @livid kraken. Managed to find this line when looking through the code which is exactly what I needed. Thank you

Vector3 perpenticularObstructionNormal = Vector3.Cross(Vector3.Cross(Motor.CharacterUp, Motor.GroundingStatus.GroundNormal), Motor.CharacterUp).normalized;

[wet burrow]

Any of you guys know how to import a .GIF file as texture assets in runtime? I want a way to play them in-game, but I want to be able to load them from a file that's already on the user's device. I found one method that seems to only be set up for downloading GIFs from a URL, and another that requires System.Drawing.Imaging which seems to be incompatible with my project.

[proven venture]

if you want to write your own you can use an imagemagick wrapper and get the raw pixel data for each gif frame and call LoadRawTextureData

else just google one

[ember peak]

how to export to android?

[wet burrow]

nvm I figured out how to refactor that URL-based method to work from a file (although I haven't tested it), thanks though

[wet burrow]

go to build settings and set the target to Android. Make sure you've downloaded the necessary NDK and JDK, and that you've enabled USB debugging on your device.

[misty glade]

I have a client-server game that is a unity client along with a C# console app (containerized and deployed in Azure so I can eventually scale it). I've been a little bit lazy with some of the networking messages, opting to occasionally send the entire player model back and forth instead of cherry picking data and merging it on both ends.

As development has progressed, I'm finding that the model is getting larger - it's about 1kb - and many normal game operations send the model back and forth.

How should I plan for load here? Where do I spend my efforts? It's pretty time consuming to pick out only the data I need and re-assemble it on both ends, but I'm starting to worry that I won't be able to scale unless I cut down on the bandwidth. Should I base my efforts on some sort of "normal" kb/sec average and use that to guide my efforts?

Thanks for your thoughts. Screenshot is of console app running locally while I click around the app.

It's probably on the order of like 5kb/minute for normal gameplay..? But I was hoping that my server would be able to scale to several thousand concurrents or more - 5MB/minute seems low, yeah?

[undone coral]

I was just about to reference the asset Uri did. you should be using it

[undone coral]

5MB/minute is low

[misty glade]

In terms of total server throughput - what kind of numbers do you think I should start thinking about it? I'm planning to add quite a few more features and I find just sending the entire player model back and forth is convenient

If I were able to get 10k concurrent users (I mean, first of all, hooray, right?) that'd be around 50MB/min - I feel like that'd be pretty dangerous

Spikes could easily be 10MB/sec

[undone coral]

usually the thing that falls over in managed runtimes is memory usage

[misty glade]

I'll keep an eye on that. I hadn't really thought that'd be a bottleneck. I was always assuming it would be bandwidth - which feels harder to solve than cpu/ram/disk issues

[undone coral]

on public clouds, bandwidth is always bad. you'll have to contact them directly with your support account to get advice on specifically which instance where has good bandwidth

for example the outbound bandwidth of a M6 instance on AWS is about 25MB/s

last i measured

which is quite low.

10MB/s is fine

[misty glade]

I don't speak AWS instance sizing 🙂 But my azure container instances are tiny, but I have no idea what kind of networking costs/throughput I get with them

[undone coral]

hmm

well it's going to be the RAM that stops you from getting to even 100 concurrents

although, 100 concurrents is like a 150,000 DAU game

[misty glade]

I think I have to pay for all these metrics

maybe not

Anyway, I'm probably just wasting time.. I should get back to the game dev. (btw, beta launch in tier 3 markets scheduled for nov 15 woohoo)

[undone coral]

for spellsource i have a benchmarking system with a dummy client

to measure how much RAM each user takes

yeah

[misty glade]

I spent the whole weekend ripping out the networking library and replacing it with a different/(better?) one

[undone coral]

that's good

[undone coral]

lol

no that stuff is a little low yield

[misty glade]

I was showstopped by not having nat punchthrough support on the older one

you ever work on something really gnarly for like 72 hours and then magically it works and it's done and you're kind of in a daze after? that's me

(including a 2+ hour support call with ms azure support.. which was awful, they literally couldn't figure out Zoom so they all had the wrong audio devices and were playing my voice through their speakers with horrible feedback/echo)

[undone coral]

lol

[misty glade]

"dude trust me you aren't using your headset I can hear myself clearly"

memory usage seems....... really low

In broad terms - what's going to soak a lot of memory? Is it the overhead on connections? My game is pretty lightweight - I think an active user is gonna use something like 2-4kb

those are my container sizes - tiny

[undone coral]

really depends what you simulate on the server

nothing? is it just a relay essentially?

[misty glade]

virtually nothing.. there's no "world" per se, it's mostly just .. hm.. how do i describe it.. it's an idle game, so i track occasional things like currencies, speed, distance, but most of the calculation is done on the client

[undone coral]

i think it's probably fine then

sometimes there is a per-thread memory cost that is high on managed runtimes

[misty glade]

ie, we trust the client to calculate distance and taps and stuff, and then the client just says "ok i reached level 2, here's my player model" and I do some light validation on it and persist it, that's mostly it

[undone coral]

if you do not do thread-per-connection/thread-per-user you are fine

[misty glade]

Yeah, I don't ever send anything to all connected clients (aside from shutting down gracefully, which honestly doesn't count) 🙂

[undone coral]

the bandwidth willb e fine i think

[misty glade]

(battles are turn based battles so even the ProcessBattles() call is virtually empty - it just checks to see if it's any computers' turns and if so, takes those turns)

[undone coral]

does the bot create copies of the game state?

[misty glade]

Yeah, but it's small

maybe a few hundred bytes

All of those dicts and lists contain objects with only a couple of fields, and the collections themselves will only have <10 items

etc

I still have to figure out how I'm going to load test this application.. I suppose I'm going to have to write some app that can fake a player for each module of the game and .. click around and stuff

[undone coral]

i have a client that plays randomly authored in the server tests

https://github.com/hiddenswitch/Spellsource/blob/9bca0a14ac64dbbbed3e9e919b0a5df30030b564/spellsource-server/src/test/java/com/hiddenswitch/framework/tests/MatchmakingTests.java#L398

this is with 400 concurrents playing as quickly as possible

it's a way more complex game

[misty glade]

Do you have your tests in your CI/CD pipeline?

[undone coral]

yes everything is tested

[misty glade]

I probably ought to do that. That's one of my biggest gaps I think but... you know, indie things. (don't hate me)

[undone coral]

it makes more sense for me because randos were contributing

and can break things in surprising ways

[misty glade]

Yeah. I suspect I have a huge ass pile of unfun work once I start friends and family beta testing

[oak sinew]

hi there wanted to ask how can i make controls like these? i want my players to scatter inside that area like in this video
I've tried using navmesh and assigning random destinations but it isn't giving a scattering feel like in this video

[glass anvil]

first of all this is something to ask in #🤖┃ai-navigation not advanced coding.
It looks to me like they're just moving towards it, look how the first few are running into the wall.. does not look like proper pathfinding to me.. could even be a simple character controller

the scattering is just them bumping into eachother

but maybe there is more to it than that have no idea where this is from.

[ember peak]

it shows errors tho

[white pecan]

Does anyone know how I can get the IntPtr of an IEnumerator

[brisk pasture]

think you would need to go through GCHandle for that

var handle = GCHandle.Alloc(items);

then cast the handle to a IntPtr, and be sure to free the handle when you are done with it

think it has a ToIntPtr method on it as well

now i would question why, IEnumerator is not a regular collection of stuff in continuous memory you can easily pass off to something else or memory map

[undone coral]

why...

[timber flame]

What is the best common name for storage classes and persistent data? (save/load persistent data)
Also what about mapper/converter to map object to persistent data
For example there is a world object (voxel game) and I want to convert it to WorldPersistentData
WorldObjectToPersistentConverter?

   public class WorldObjectToPersistentConverter
    {
        public WorldPersistentData Convert(World world)
        {
            
        }
    }
    public class WorldPersistentToObjectConverter
    {
        public World Convert(WorldPersistentData worldData)
        {
            
        }
    }

A world consists of chunks and each chunk has voxels.
Also, copying data is time consuming. Is it better to have one data structure for logic and persistent data?

[lament salmon]

To me it also looks like a bunch of low friction rigidbodies just naturally squishing in there, physics driven

[dapper cave]

when you dispatch a compute shader, does it immediately start execution ?

[pliant crest]

"copying" data?

shouldn't you be streaming in data

and streaming out data

assuming its works on same basis as any other 3d chunk loader

[real plume]

i have a somewhat general question, but i could post code if necessary: i'm running a tilemap game that has a separate tilemap for the water tiles, which are running a kind of cellular automata process to simulate fluidity. the code works well enough, and if there's not a lot else going on there's no issue. however, after a time, i start getting major cpu hits when other things happen during runtime like instantiation and destroying of gameobjects. my question is mainly if doing the water tilemap running calculations on where to move next update is the type of thing that would be worth trying to run on a separate thread, or if i should be looking at other optimizations first?

[lament salmon]

always analyze your performance with the profiler before you start optimizing further

[real plume]

i've done that and it seems like the major fps spikes are from "animation", which presumably is from the tilemap because it drops when i disable the automata script. also the stack use seems to just rise and rise, as well as memory use. i changed my tile data from classes to structs and that helped with memory but the cpu usage and memory usage is still bad

[real plume]

i think what is happening is that the water calculation process is hogging the main thread so when other things want to happen they have to wait. but i guess my other question would be is if this "animation" cpu usage looking like a spike would necessarily translate into that same spike occurring even if the calculations were magically done with zero cpu usage (just theoretically). if that were the case, i would need to look at reducing the animation, but if it weren't the case, then i could look into reducing the main thread load from that calculation process.

[long ivy]

you need to determine if it's animation or calculation that is causing your performance problem. How you optimize will go from there. If it's animation, a separate thread won't help because you can't interact with the tilemap outside the main thread

[real plume]

that's what i was thinking. and i understood that second statement somewhat lol (just watched a good video on the subject). do you have any suggestions as to how i might start diagnosing which is which? i guess this might require getting into specifics but are there any general techniques for helping to parse this? i only first opened the performance monitor thing yesterday so there's a lot about it i don't understand...

[long ivy]

profiler, augmented with some https://docs.unity3d.com/ScriptReference/Profiling.Profiler.BeginSample.html/EndSample pairs where needed to identify exactly which fragments are slow

[real plume]

so if i'm understanding i want to use this for a part of code that i suspect is a cpu hog?

[real plume]

that seems to be the case. okay thanks a lot for the info.

[dapper nest]

I mean I don't really think telling someone that they "should be using" an asset instead of learning is a good idea imo. Sure it is a great controller and is free, but definitely not "should be using" and instead a good way to learn

[timber flame]

I mean when I want to convert a voxel (struct) to voxel persistent data, I have to copy it. Using ref assignment?

[formal falcon]

I'm making a game for a gamejam, and I've run into an implementation problem involving audio...
So there are two scenes, one scene is a 2D game where the camera renders to a render texture, and a scene where the game is being played (render texture on a TV model.)
The problem I'm facing is getting the audio to route from the 2D game to the "real life" scene in a way that audio from the 2D game seems like it's coming from the TV.
The obvious hack is to fade volume of the 2D game master mixer based on distance, but is there a more elegant way to achieve this?

The idea is that you can leave the 2D game on the TV playing if you want, and walking away from the TV changes the audio (volume, doppler, etc.)

[ruby hare]

Wnat to ask is there new feature other than asset bundle to help with dynamic content ?

[brisk pasture]

addressables

#📦┃addressables

its built ontop of asset bundles but makes them a lot easier to use

also the scriptable build pipeline if you just need to make existing asset bundles more flexible

[timber flame]

I get this error when using message pack csharp serialization

FormatterNotRegisteredException: Bitzooma.MiniCity.Core.WorldPersistentData is not registered in resolver: MessagePack.Unity.UnityResolver

I have used Key() attribute and MessagePackObject for classes

  private static readonly MessagePackSerializerOptions MessagePackSerializerOptions = new(UnityResolver.Instance);

Unity 2021.3.6 .NetFramework + Mono
and if I use MessagePackSerializerOptions.Standard I get the error below

MessagePackDynamicObjectResolverException: can't find matched constructor. type:Bitzooma.MiniCity.Voxel

What does it mean? I should have ctor in data structures?!

[flint sage]

IIRC you need a compound resolver that registers your own classes and you should have a ctor that has the same order of parameters as the properties defined in the model

I don't have reference code anymore so can't give you concrete details

[timber flame]

I see some examples without any ctor in its github

[flint sage]

Could be but that's how I did it

[timber flame]

Thanks

and why the errors are different when I change the options

It worked

 [SerializationConstructor]
        public Voxel(byte id, byte rotationType)
        {
            Id = id;
            RotationType = rotationType;
            _flags = 0;
        }

[undone coral]

So there are two scenes
do you mean two scenes or two Scenes? it sounds like you have one Scene

audio to route from the 2D game to the "real life" scene in a way that audio from the 2D game seems like it's coming from the TV.
unfortunately you can only have 1 audio listener. my suggestion is to place the 2d game content right where the TV is, and hide it from the main camera using rendering layers. then it will all work idiomatically

[formal falcon]

Why the particulates on capitalization? One scene (the file type) is loaded, then on request a second scene (also the file type) is loaded. You can have multiple scenes loaded at once. Anyway, if I did that, I'm sure I would need to scale down the 2D content which might have unintended implications on rendering, not to mention it would make it annoying to debug in the scene view.

[undone coral]

i suggest using only one Scene

i say that because it wasn't clear from your language

if you meant scenes metaphorically, or a literal unity scene

i don't think the scaling will matter. as long you use a game object to achieve scaling and positioning whose local space is still "normal" coordinates for the 2d metaphorical scene

it should work well

[formal falcon]

Why? it works perfectly the way I have it now

[undone coral]

hmm

anyway that is my advice, i think your game sounds great

all the other solutions are more complicated. you could try duplicate the audio sources into the tv, at the right positions, which is a lot harder than moving the metaphorical 2d game scene

you can also ignore audio

which i think is what you are probably going to do

[formal falcon]

that's actually not a bad idea, I'll try that

[undone coral]

really depends what the goals are. immersion? is it meant to be played in a browser? etc. etc.

[undone coral]

it's usually a bad sign when someone takes the suggestion i say is the worse one lol

but i've been striking out lately with trying to get people to do good stuff

[formal falcon]

I could extend the AudioSource class and create logic to move the audio source to the TV on start

BUT, changing volume in the mixer based on distance would be more performant, just not as immersive. but it's a gamejam game so that's okay

[undone coral]

i just think this is a really, really bad idea

when it would take you all of 5 minutes to move the 2d game scene into the tv, and set up the camera render layers correctly

it's a game jam

you're talking about performance...

it's a built in render pipeline game right?

[formal falcon]

Yeah

[ruby hare]

thank you

[scenic forge]

I have a source generator project specific to the Unity project, I only want to source control the SG source but not the built SG dll. Is there a way to make Unity execute a command when opening the project so I can use it to build the SG project and copy the dll over (it has to happen early enough in the opening process otherwise I'm under the impression that SG wouldn't work) rather than having to tell people that clone the repo to "please build the SG project, copy the dll over, before opening the Unity project"?

[undone coral]

imo you should check in the built dll

normally you could achieve this using an asset preprocessor, but it will always be glitchy to generate code which then modifies the player assemblies on launch

[rain wolf]

I do not have any advice but am in a very similar situation so i'm happy to see that you have posted this! Additionally, do you happen to have found a good way to specify AdditionalFiles in Unity (since you can't edit the csproj directly)?

[undone coral]

i have never gained anything from generating sources at launch, and have gained a lot from checking it in, and i've been using source generators for years

that said, you probably mean a msbuild feature called Source Generators

and not something like protobufs

you should check in the DLL

[scenic forge]

I was wondering about additional files too. There's a hook to modify generated csproj but I'm under the impression that csproj isn't actually used for building, only for IDE (so your SG will work in IDE properly but not when you are building the game)

[rain wolf]

I'm (obvs) not Burrito, and this may be the best solution but working on a large team, PR noise can be a bummer -_-

[scenic forge]

I don't think it's ideal to check in the SG dll, it's really just artifact of the SG project and that should serve as the single source of truth

[undone coral]

yes but it might provide zero ROI

[proven venture]

InitializeOnLoadAttrib?

[undone coral]

to generate this on demand

what is the project?

[scenic forge]

If someone makes changes to the SG project but forgets to rebuild and commit the new dll, that's a mistake that will be pretty hard to spot.

They can get out of sync due to human error basically.

[undone coral]

in my experience with protobuffers, the kind of person who has the knowledge to make modifications to them can handle checking in the updated generated artifacts

[undone coral]

that's what i'm saying. nobody who has that level of knowledge will make such a mistake

[rain wolf]

i completely disagree with this take 😅

[undone coral]

and someone who lacks that knowledge, who can't modify e.g. protobuf files, will never need to think about source generation in either scenario

i don't know. unity is a build tool, it's not a Build Tool

i use gradle as my build tool, and i do have a single command for all source generation

it really comes down to what the project is

[undone coral]

you can write a piece of C# code, as this user suggested, that runs in an editor script

and executes the generation step

i guarantee you that it will be really buggy

it will be running all the time

it will be hard to engineer it to detect changes, etc.

because you will be reinventing a build tool

you can try to use an asset importer class, but that is going to be

[proven venture]

then FileSystemWatcher and rebuild on change

[undone coral]

i will just say that i can tell you haven't done this before

it is not that simple

[scenic forge]

Ideally Unity would just support project reference so you can just reference the SG project like you normally do.

[undone coral]

well on the flip side, vendoring has a lot of benefits too

what is the project?

[undone coral]

this is correct, the .csproj files are only for IDEs

i have a separate solution that, via a build tool, generates assembles that i check in

i do that because when an artist does need my repo, they will not comprehend any steps beyond "download a zip" and "open the project"

they're not going to understand git anyway

they're not going to be using this thing. they don't reall yhave a reason to be in unity, honestly.

so it's tricky

really comes down to what the project is

[rain wolf]

Yeah, I think that's correct. But is there a hook or something to know when the csproj has been regenerated (or a reliable plugin or something to make sure edits stick around?)

[scenic forge]

You can just modify the csproj straight in that hook, don't need to edit manually

[rain wolf]

i was thinking about this too. is it custom or do you have a link to something?

[undone coral]

you shouldn't make any edits to the csproj files

[scenic forge]

But the bigger problem remains that those changes won't affect actual building so it's pointless.

[rain wolf]

Yeah, not directly, but I guess how else can one (for SG) specify something like AdditionalFiles

[undone coral]

https://github.com/hiddenswitch/Spellsource/blob/9bca0a14ac64dbbbed3e9e919b0a5df30030b564/spellsource-protos/build.gradle#L86

i don't know how meaningful this will be to you

[undone coral]

is SG referring to Source Generators, an MSBuild feature?

you do not use them directly from unity

[rain wolf]

Source Generators.

[undone coral]

you create a separate solution that is an ordinary dotnet solution

and use dotnet publish

and cp your build artifacts into unity

[rain wolf]

the consumer is who has to specify the files for the generator.

[scenic forge]

That's exactly the problem we are trying to solve, to make that process automatic instead of manual for every time SG is changed.

[undone coral]

unity isn't a Build Tool, it cannot know.

[rain wolf]

yeah, i'm not trying to make it one

[undone coral]

yeah

you definitely cannot modify Unity player code

using a source generator

is that what you are trying to do?

for code weaving?

[rain wolf]

getting the source generator imported works fine, but the source generator needs ot know where files live (and this is usually specified as AdditionalFiles in the CSProj)

nah

[undone coral]

okay

[rain wolf]

we are generating an enum for localization keys

[undone coral]

honestly i would ismply avoid using the feature

[rain wolf]

which works

[undone coral]

well

that is honestly kind of low ROI

this is a very very burdensome, buggy feature to use

for that

@scenic forge hasn't said what the projec tis

so i don't know

sorry i can't give better advice

[rain wolf]

cool

[undone coral]

you can just go and make the enum

the mistake would be obvious

nobody really fucks that up

and when they do it takes seconds to fix and is obvious

so i don't know

[scenic forge]

I mean it doesn't matter what the feature is, it's already determined SG is a good way to solve our problem, simply to make SG nicer to use than having to do all these manual labor of copying and syncing.

[undone coral]

besides, there are approaches to localization that obviate the need for that

[rain wolf]

pretty cool to hear that you work with devs that don't make mistakes.

[undone coral]

i don't know, almost* every person who has ever said that has been wrong

[rain wolf]

i make mistakes and have caused bugs

[misty glade]

I have something busted in my build, and I'm not sure how to fix it.

I recently refactored a script to rename it from ISGSceneManager to IsgSceneManager. Works fine, but now only on Android, I get these error messages in my build:

How 2 haz fix

[undone coral]

did you try the rename trick

[misty glade]

(I've deleted the DDOL object, readded a new one, readded it)

[undone coral]

where you change it so that it doens't differ only in casing?

[misty glade]

What's the rename trick? To force unity to generate a new guid for it?

[undone coral]

try changing it back to ISGSceneManager

then rename to ISGSceneManager2

then rename it to IsgSceneManager

[misty glade]

IIRC unity refers to the script objects by a guid or something?

k

[undone coral]

refactor rename*

[undone coral]

yes but case sensitivity is fucky

[misty glade]

Yeah, no problem.. I was thinking of mucking with the meta files and making sure they're correct

[undone coral]

yeah

[misty glade]

Actually...... I can probably at least verify the problem is there by looking at the guids and making sure they line up

[undone coral]

i suggest using gradle then

gradle or bazel

they are good build tools to solve problems like these

otherwise you can share what you are using the source generator feature for, maybe we can brainstorm an alternative

[scenic forge]

I do have something similar as well (unrelated to the current SG discussion), we have 1 i18n json as single source of truth so we can pass that to localization platforms, and need to generate code to handle stuffs. Due to the lack of additional files support we did it with a regular editor script and it just reruns whenever it detects changes, it works alright but every time you change anything in i18n your commit includes like 4 other files 🙁

[undone coral]

handle stuffs
hmm... well i have bad news for you

you're trying to Microsoft Visual Studio & Windows your way into a multi-platform, not microsoft visual studio world

it will not work

this will not be the first nor last problem

if you want to make Microsoft Visual Studio applications, do not use Unity

those skills are not really translatable

the fewer editor scripts the better

[scenic forge]

I don't see what this has to do with VS or Windows, the game needs localization.

[undone coral]

it is hard to come from a place where you are comfortable, Microsoft Visual Studio, to a place you are not

[rain wolf]

actually kinda the same. We are using Smartling and have had to do (Unity yaml <=> csv <=> smartling) which we quarantined to a separate (orphan) branch. Smartling reads/writes to CSV branch, and we have a job that turns the CSV back into unity assets when it sees changes from smartling

[undone coral]

that is a really bonkers workflow

[scenic forge]

Yeah which is exactly why we have 1 i18n json only as the single source of truth, to prevent workflow getting out of sync due to some human error.

[rain wolf]

Yeah, if the rest of the company was not already using Smartling it would probably not have been our choice because it does not play nice with Unity

[scenic forge]

I don't see your point, I don't even use VS.

[undone coral]

do you use Rider?

i'm not sure then where you would have gotten introduced to source generators

maybe you are a rare C# backend developer... using Rider...

[scenic forge]

No I use VS Code, and I fail to see how localization has anything to do with what my IDE is.

[undone coral]

okay i see

[misty glade]

OK, the files are completely cursed. Let this be a lesson, don't refactor file/classes in windows/unity to just change the casing. 😐

[undone coral]

anyway i use the unity localization package, and i always do localization at the end, when it makes the most sense to

[undone coral]

yeah

if you can roll back in source control

[misty glade]

Renamed it and git kept the original casing 😐

[undone coral]

okay

[misty glade]

Nah it was months ago

[undone coral]

that's unfortunate

[misty glade]

I'm just renaming it to something else

[undone coral]

okay

[misty glade]

"IsgSceneManager" RIP

:p

[undone coral]

yeah that's tough, it doesn't really make sense to integrate with smartling, the best time to do localization is at the end

then it doesn't matter

how it is integrated

[undone coral]

and you can work with unity localization throughout 99% of your product lifecycle harmlessly

it doesn't really make sense to iteratively translate

i can see how someone can think it does, but it doesn't

[rain wolf]

this is getting continued updates

[undone coral]

if you are in that kind of jeopardy

[rain wolf]

so that really is no end

[undone coral]

well you know what i mean

right before a release

not during development

like literally the last thing before you release something

[rain wolf]

we release weekly 😄

[undone coral]

well it's not like this can't be done in unity localization

safely

it certainly can

it's a very good package

that apes all the best stuff from everyone else

[rain wolf]

truly in awe of your experience

wow

sage

[undone coral]

okay well, i don't know, i agree it's tough

it's tough to know. i don't know what these games are

i find out bits and pieces

[scenic forge]

Is Unity localization type safe? Eg prevent referencing strings that don't exist, or knowing a string takes exactly 2 interpolation arguments.

[undone coral]

you can't use source generators in unity code really, end of story

[scenic forge]

You very well can, the integration is just not seamless.

[undone coral]

i think if your game is making money and you've got at hing working, you know, go for it

you got something

i don't really know

i've offered my perspective

you probably have a support contract, go ahead and use it

once i heard "stuffs" and "csv branch", it signaled to me a certain set of processes that require support contracts and paid advice, which i cannot give, i am sorry

[proven venture]

well they wrote a fluent api-esque system on top of ugui, so... right

what i meant to say was he does some pretty wild stuff

[undone coral]

who does?

[proven venture]

burrito

[undone coral]

i see

i mean this stuff is really interesting

it is sort of subjective what the yield is on doing weird source code features

making things intellectually stimulating there means more game code is written, even if the intellectually stimulating stuff is low yield

so i do not mean to begrudge experiments in this approach at all - go ahead, use source generators and figure it out for the rest of us 🙂

[scenic forge]

SG is very powerful even not for something specific to Unity

(And they are also a pain to debug so its usage isn't widespread even in the wider C# community)

There are tons of uses for SG, practically anytime you have boilerplatey/repetitive code it may be a good SG use case.

[undone coral]

yeah i agree

i use source generation just not Source Generators

listen i mean to be supportive

i guess i am not familiar enough with the feature

i got the highest yield from embracing real build tools like Gradle

to solve problems

[scenic forge]

You can think of source generators as source generation but always up to date with the input (doesn't need to be run manually and invisible)

Source generators don't have the issue of "I have to manually run any time changes occur so they are reflected in the generated code", "I must either gitignore generated code but give instruction to how to generate that code so my project even runs", or "I must commit the generated code and keep it in sync all the time"

[undone coral]

yeah but macros are always a little radioactive

it makes sense to me to use them as an IDL, but then the reality is it's a network artifact, like protobufs, it's going to be out of date somewhere always

and in my unity streaming service, it is really high ROI to have done everything using grpc / protobufs, because it really did ensure backwards compatibility across my API

[scenic forge]

Source generators are invisible to the consumer, as long as you have it set up, you do not need to manually handle them ever and the generated code will always be in sync with input.

[undone coral]

it just sounds like they would have, in the general sense, the same problems Macros do

unreal engine uses a lot of macros

out of necessity

[scenic forge]

You mean C++ templates?

[proven venture]

ue c++ has a lot of macros for all i know

[scenic forge]

I don't know much about Unreal or C++ templates, but they are very different from C# SG

[undone coral]

they use proper macros yes

hmm

when i looked at the source generator docs just now

it looked like macros by another name

[scenic forge]

C++ templates directly modify and transforms your source code, SG does not.

[undone coral]

source generators can achieve "code weaving" aka bytecode instrumentation

which i have also seen from the java end, and takes a lot of expertise to do correctly

[scenic forge]

It cannot.

[undone coral]

hmm

[scenic forge]

It can only append, not modify.

Which makes it a lot less powerful, but a lot more sane to work with.

The only thing C# SG can ever do is "here's a new file I generated, include it in the project"

It cannot modify anything.

[proven venture]

il weaving on the other hand

they kind of achieve the same thing except one generates and one changes the source dll

[scenic forge]

IL weaving or any sort of modification can do a lot more than SG but yeah.

[proven venture]

but you'll need a weaving library since source generators wont do it

[scenic forge]

Yep, and it's very easy to blow things up completely and descend into madness, so it's very niche.

[summer spindle]

I REALLY need help with Local Multiplayer, anyone that have done a Local Multiplayer game that can help me or give me a link to a working Youtube Tutorial? Please!

[sly grove]

do not spam/crosspost. keep it in #💻┃code-beginner

[potent sandal]

Hey guys, hopefully this belongs in the advanced category. I'm building a small space colonizer game where you can build on planets. While I like how it's shaping up I was considering switching my current building system (place anywhere) to a tile based one to have more precise control on how buildings are placed and whatnot. The problem is the following:
All gameplay happens on actual spherical planets! I have no idea on how to begin to map a tilemap of squares onto a sphere.
Does anyone have any tips OR know any online resource on the topic? Anything is appreciated!

[sly grove]

I can point you towards one of my favorite resources on such matters:
https://www.redblobgames.com/x/1938-square-tiling-of-sphere/

I'm going to say right out that Unity's build in Grid/Tilemap etc are probably not going to be of use to you, and if you haven't already, you'll probably need to make your own solution.

[potent sandal]

This is amazing thanks. I just gave it a quick peek, but I'll read it properly once i've got a bit of time!
From the interactive things I've seen tho it's seems like I should be generating a second sphere around the planet with uniform quads and use those for my tiles?
Probably missed a few things but nothing that reading it properly wont fix.

[gray pulsar]

you probably don't need actual meshes for each tile. What you want is a mathematical formula for converting back and forth between a point on the sphere and tile coordinates

[timber flame]

Which one lead to more compression?
Storing the entire byte array or only filled values + indices.
I think the first one because the consecutive values can be repeated and it can be compressed more

[sly grove]

it depends on the nature of the data.
If the data has high entropy, storing the entire array will be smaller than values + indices.
On the other extreme, If the entire array is all zeroes or all ones, then it's very highly compressible via this technique

[timber flame]

DataStruct1[] // byte rotation, byte id --> Entire Array
DataStruct2[] // byte rotation, byte id, index --> Only Filled values

Yes, it depends

but suppose it is a world with many empty spaces close together probably

I use LZ4 compression as well

Also, I do not know LZ4 can handle repeated patterns in bytes or not
23,2,23,2,23,2
data(id,rotation),data(id,rotation),...

[midnight violet]

Why is Unity so poorly implementing Native stuff from iOS... anyone has some trick how to get the lastError for ReplayKit without constantly recieving the same error? ReplayKit.lastError will stay forever... no matter if you restart recording...

[jolly token]

Compressing algorithm does the something similar for you, so I don't think filtering will help making result smaller..

[timber flame]

Can LZ4 compress data structures efficiently if they are the same?

public struct S{
   public byte Id;
   public byte Rot;
}

[jolly token]

They should

[timber flame]

So, adding indices can ruin that compression :/

[jolly token]

Yeah it could be just adding more things to compress

[regal olive]

how do i get the width and height of 2d camera??

[sly grove]

orthographicHeight gives you half the height of the camera

the width is based on that height and the aspect ratio

float height = cam.orthographicHeight * 2;
float width = height * cam.aspect;```

[regal olive]

bruh i used orthographicSize not orthographicHeight

thx

[sly grove]

always helps to check the docs

[regal olive]

how?

orthographicHeight dont exist

[sly grove]

https://docs.unity3d.com/ScriptReference/Camera-orthographicSize.html

you're right

like I said - always check the docs

I just typed that code above off the top of my head

[onyx vault]

I want to make a GameMode system in unity for a multiplayer game where I can override base functions to create new modes easily. Would it better design practice to have an event based system where I subscribe and unsubscribe from game events(eg player dies, player joins late) and invoke them when they occur or have a base GameModeClass and override functions that are called when the events are invoked? Also what would be the best way to manage different player prefabs for each gamemode?

[regal olive]

The best way is the way that uses least of resources the computer has

[onyx vault]

I'm asking more from a software design standpoint of view

[regal olive]

Oh could have said that at the start 🙉

[fresh salmon]

Would it be better design practice...
Second sentence

And why not? Event based is extremely versatile as it's loosely coupled, your game mode "drivers" could subscribe to the events they need from a game manager object

Watch out for subscription spaghetti though, try not to link events back and forth too much

[regal olive]

Hello what happens when object is destroyed using destroy
. Does it free the space it used quick

[midnight violet]

It is flagged to be garbage collected if there is stuff inside it filling up memory and will be removed on next GC if not done manually

[regal olive]

How would i fully Remove it?

Like it never existed

[midnight violet]

If its collected by GC, it is removed.

[austere jewel]

To fully collect the managed side of the objects you need to also set references to them to null. Same as any C# object.

[real plume]

want to ask about that. if i'm instantiating 50 'collectible' objects because the player did X thing, then i soon after destroy them because the player picked them up, is this the type of thing that probably would have leftover references? or do you mean explicit references by variable declaration and things along those lines? i'm just now peeking my head into garbage collection and memory so sorry if this question isn't the best phrased...

[midnight violet]

It depends on what is in the obejct. Its not just gameobject are being handled equally all over the game runtime. If you load textures for example and create texture2Ds, all of those things can affect if GC is kicking in or things never get disposed

[real plume]

in this case it's a prefab to represent a dirt tile after it's destroyed (very similar to terraria or minecraft)

[austere jewel]

If you Destroy something the native C++ representation is freed. As with any other C# object, references to it keep it in memory. Destroy cannot set references to null.
The references just evaluate to null via Unity's overridden equals comparison.
To have the managed C# memory ready to be garbage collected all references to those objects must be null.

You're likely to come along and set those references to null later by just operating on the same lists, overwriting the variables, etc. So usually you don't have to think about it

[real plume]

ok, that actually was in the vein of my next question. i'm making a game that includes a tilemap running a cellular automata process and both the memory use, heap use, and cpu use seem to rise over time so i'm trying to diagnose. the cellular automata script uses extended use of lists and dicts so i was wondering if that was part of the issue. the code isn't cleaned up for review but if you have a moment to look to see if there's any obvious memory collection or leak issues or whatever i'd appreciate it a lot.... http://pastie.org/p/3MDJEtV9NL5wFvX7Bp5ya1 - not the shortest script but a lot is repeating logic or commented out...

and please don't make fun of my ifs lol

[austere jewel]

https://docs.unity3d.com/Packages/com.unity.memoryprofiler@1.0/index.html
Use the memory profiler to diagnose the issue

[real plume]

i'm trying to learn but i think the issue is that there's several things going on at once. anyway i just though i'd ask but i get it's a big ask, nbd...

thanks for the answers though i appreciate it

[midnight violet]

If several things are going on, take apart your system and profile single mechanics, loadings, instantiations, whatever is going on. And dig into it to find out

[fringe sentinel]

gonna ask here but anyone know how to fix this?

its all from unity default scripts
all of them being com.unity.ugui

while this comes from a game decompile, i first deleted the decompiled versions, then reinstalled them in package manager

version is 2019.2.13f1

[misty glade]

I have some frame stuttering when my client starts. In profiling it, though, I have this spike - a single frame that takes 201ms to process, but there's no single obvious low hanging fruit. Any ideas on this?

Is this useful? What's all this GC alloc going on?

[wooden cedar]

In the editor, in builds, or in both

[misty glade]

both

I'm not an expert with the profiler, but I can't seem to identify what's actually taking up all the time in this frame

[austere jewel]

JIT is compiling a whole bunch of stuff

[misty glade]

hm... I still get stuttering on Android, which my understanding is that I'm not using any JIT at all there

[wooden cedar]

Just curious, I saw something like that once (EventSystem to taking a while) with complex UI size fitters and with best fit text. But that was mobile only.

[misty glade]

I mean, I have some fitters, but nothing here is doing any layout/UI work.. It's likely the parsing of the settings json I'm getting from the server, but ... I can't seem to prove that

[austere jewel]

There's a bunch of logging mentioned. Turning off stack traces for anything but errors is always a good first step

[misty glade]

OK, lemme see, brb.

Remind me where that setting is? That's a console setting? I haven't played with that setting in months

This stuff?

[wooden cedar]

This mayyy help
https://answers.unity.com/questions/1264528/insane-eventsystemupdate-lag-spike-no-graphic-rayc.html

[misty glade]

i'll try with deep profiling.. not sure my problem is what this guy is mentioning..

AHA, deep profiling uncovered it

and ... I'm stupid.

(I don't even want to share the culprit)

I forgot that I added some geolocation service thingy and meant to make it non-blocking but.... just forgot about it. Yes, that's a blocking HTTP request right in my game. 😐

[austere jewel]

That certainly seems more relevant

and yes, what you had above sadly does affect both the stack traces in builds and in the editor.
I wish they were separate things. I've had to make custom build steps that set those sort of things just for builds and then reset them

[misty glade]

Yeah. I mean, it's one thing to have a clue and chase it down (and still have the hiccup), it's another to know and realize that you are, in fact, dumb today. 🙂

All good - I just wrote it that way because I was debugging some other thing (as a completely related aside, fuck microsoft azure container instances) and just.... forgot to do it properly with async or a coroutine

I've also added a little bit of pseudo-profiling and .. my MessagePack object takes 93ms to deserialize, and my initialization in unity (once it's deserialized and I set up my game world) is another 140ms, so there's still some other low hanging fruit

ISG: <color=#6666FFFF>Debug: LNL: Connected. Connection took 110.5901 ms</color>
ISG: <color=#6666FFFF>Debug: Deserialization took 93.0114 ms.</color>
ISG: <color=#ccccccff>Verbose: Finished initalizing. It took 144.6879 ms.</color>

(all of those steps are sync atm)

Anyway, I'm rambling. Thanks for the help.

[blissful bay]

Hi, Im looking for a high-performance way to find a closest point to another point, right now Im using this https://gdl.space/feqegiyidu.cpp
but Im worried that this wont perform well enough without sacrificing accuracy
there has to be a better way to do this, right?

[midnight violet]

So can you explain your code a bit? What is pathCreator.path.GetPointAtDistance

[blissful bay]

sure thing, pathCreator is an asset for creating bezier curves in unity, this function returns the position of a point on a predefined curve

it takes floats from 0 to 1

[midnight violet]

Alright, and the ontriggerstay is on one object or on multiple? what is ClosestPoitnFound doing

[blissful bay]

the idea is that the player is supposed to "drag" an object along the curve, this script is on the object they are supposed to drag, when their finger is in its collider, the ontrigger stay is looking for the closest point from the finger on the curve, ClosestPointFound then moves this gameObject to the point that has been found

[sly grove]

well one thing to do is certainly not call GetPointAtDistance three times

save the result in a variable and reuse it.

Same with other.transform.position

Actually this whole thing:

dist.x = pathCreator.path.GetPointAtDistance(f).x - other.transform.position.x;
                dist.y = pathCreator.path.GetPointAtDistance(f).y - other.transform.position.y;
                dist.z = pathCreator.path.GetPointAtDistance(f).z - other.transform.position.z;```
Could just be ```cs
Vector3 dist = pathCreator.path.GetPointAtDistance(f) - other.transform.position;```No?

[blissful bay]

I have read that calculating the x,y and z separately was faster than using the vectors as a whole

but I havent tested it yet

[sly grove]

maybe, but not the way you're doing it it where you're recalculating the whole vec3 on either end 3 times 😱

[blissful bay]

needless to say, calling the method 3 times is a mistake nonetheless

[blissful bay]

aye, now I realize it, thats certainly a thing to fix

is there a way to do the entire logic better tho? I dont really have a good feeling about the way it is now

but I wasnt able to come up with anything better myself yet

[sly grove]

Doesn't the bezier library you're using have a ClosestPointOnCurve?

[blissful bay]

not that I am aware of

I will go through it once again, gimme a while

[sly grove]

Might want to look into Unity's first party splines library. They have such a method:
https://docs.unity3d.com/Packages/com.unity.splines@2.0/api/UnityEngine.Splines.SplineUtility.html#UnityEngine_Splines_SplineUtility_GetNearestPoint__1___0_Unity_Mathematics_float3_Unity_Mathematics_float3__System_Single__System_Int32_System_Int32_

looks like it might do some kind of binary search under the hood too

which is promising for performance

[blissful bay]

it actually does, I just missed it the first time

I dont really understand how it works, but it probably does

https://gdl.space/picarevoxu.cpp if anybody would be willing to help me understand Id be glad, but it isnt really needed for what I need to do

[blissful bay]

thanks for pointing out that there should be a function for that! I thought of it, but I looks like I wasnt thorough enough

[analog nexus]

How come HttpMethod.Patch exists and works for some people but not others

if you look at https://github.com/Unity-Technologies/mono/blob/unity-main/mcs/class/System.Net.Http/System.Net.Http/HttpMethod.cs#L136

it just throws on access

and yet - it works for me

and my colleague

[alpine adder]

what is the best practice for creating a currency system in my game? I see a lot about playerprefs but I was thinking more saving to a file? especially as a long term goal would to get this on aws for multiplayer

[regal olive]

Saving to file makes it easy to modify and is not secure. Here is good tutorial https://www.sitepoint.com/saving-and-loading-player-game-data-in-unity/

[alpine adder]

I see so this will work and eventually I just move this over to aws?

[stuck onyx]

anybody debugging on device on a mac? can't attach visual studio to the unity process of the phone

[regal olive]

Are you think about serve side!

[alpine adder]

no sure never done multiplayer saving before tbh

I have no idea how that works lol

but the thought of an api call everytime gold updates does seem extreme

[regal olive]

For serve side saving use PHP and MySQL

[alpine adder]

how does saving work though? is it really a server call for every input?

[flint sage]

It depends on how you approach things

[regal olive]

You could save data on when they exist game

[flint sage]

If you want the minimum amount of cheating possible then you need to validate everything on the server

[stuck onyx]

that would not guarante the data seve

save*

[flint sage]

E.g. user buys something? Server validates and executes everything and only tells the client to update the inventory with x items and x gold change

[regal olive]

Yeah or he could save it on som secure file first and move it to the server

[stuck onyx]

I dont know about your project, but for the moment i recommend you to ignore possible cheating

if people cheats on your game means you have become succesful

if you have become succesful a few cheaters will be the least of your problems, believe me

dont waste your time on that

i tell you by experience

[regal olive]

Yeah they come when the game gets population

[stuck onyx]

yep

think about more important things like scalability

game mechanics

marketing

[regal olive]

Also fighting cheaters is pain in the ass

[stuck onyx]

leave the anticheating for when you become succesful, if youdo

there will be ALWAYS cheaters, thats unavoidable

[alpine adder]

so im making just a simple multiplayer td

although nothing multiplayer is simple lol

so think just get gold build towers

[stuck onyx]

but wasting your time creating a super safe anti cheating game its stupid.

i mean wasting it now

[alpine adder]

just trying to think through the currency system for the time being

and how that can be done

[stuck onyx]

what are you using as database?

[regal olive]

Just send it to your database

[flint sage]

You should probbaly just do peer 2 peer networking instead of some centralized system for some simple TD game

[alpine adder]

I have nothing set up 😄

I just have normcore atm

and got people to join database and the rest is still all in the works

learnign about this thank you all!

going to hook it up to a firebase db?

does that sound correct?

a SO wouldn't work right

[gloomy mica]

Playable API question. How to set an existing AnimationClipPlayable's clip to null (without recreate the playable)?

there's only a GetAnimationClip() function, but when I inspect the Animator's Graph via PlayableGraphVisualizer, the playable can set clip to null without recreate itself

[undone coral]

i think there's a photon sample for multiplayer tower defense

on phones?

[alpine adder]

I see thank you! I did some research and came to azure playfab. Im curious to try it out

im sure the learning curve will be fast hahaha

[undone coral]

is this for a phone?

[misty glade]

I have some weird and difficult to diagnose problems with UGS in app advertising.. Randomly and without any discernable error messages or logging, the sound devices don't resume when coming back from an interstitial. The sound plays before, the sound plays from the interstitial ad, but coming back to the client .. no sound. Any ideas? Edit: On Android devices

[undone coral]

why are you using PATCH?

is this for an http server you control, or something else?

in unity, you use unitywebrequest, not system.net.httpclient

[undone coral]

that sounds like a red herring. serialization does not take 93ms. if unity objects are creating during deserialization maybe

if your deserialization involves huge, huge huge strings

[misty glade]

this is strictly my messagepack deserialization.. not sure ..

[undone coral]

maybe

[misty glade]

they are :p

19kb json

[undone coral]

that's fine

but i mean

inside the json

is it 18.999kB of strings

isn't msgpack a binary format..

[misty glade]

nah it's mostly small strings and primitives

yeah

[undone coral]

hmm

[misty glade]

it's 19kb packed

[undone coral]

it doesn't really take 93ms to deserialize 19kB of data

it might take 0.2ms, or 1ms in the worst case

or less

[misty glade]

                    d("Received configuration info from game. Unpacking settings and initializing systems.");
                    _startedConfiguration = DateTime.Now;
                    ServerSendConfiguration ssc = UnityNetworkingUtils.Deserialize<ServerSendConfiguration>(msg.Message);
                    TimeSpan ts = DateTime.Now - _startedConfiguration;
                    d($"Deserialization took {ts.TotalMilliseconds} ms.");



        public static T Deserialize<T>(byte[] value)
        {
            return MessagePackSerializer.Deserialize<T>(value);
        }

[undone coral]

can you ask msgpack for a dictionary instead of your object?

[misty glade]

that's the 93ms call

[undone coral]

hmm

i would maybe use stopwatch

or Profiler

profiler is the most reliable

[misty glade]

it's pretty much only MessagePackSerializer.Deserialize

[undone coral]

i wouldn't use datetime.now

a call to datetime.now can take a long time lol

on a device

believe it or not

[misty glade]

Yeah, I mean, that's fair, I was just looking for magnitude of order issues

[undone coral]

because it needs to load timezone data

[misty glade]

really? lol, didn't know

[undone coral]

yeah

[misty glade]

I use UtcNow almost everywhere, fwiw - does that avoid that timezone overhead?

[undone coral]

in webgl, at some point, calling datetime.now in a mobile browser took 2.7s

you shouldn't use datetime ever

if you can avoid it

[misty glade]

or does that still need to .. "unconvert" my perfect time (pacific time best time) to utc

that's crazy, TIL

I use it extensively on my server.. 😐

[undone coral]

that's fine

i mean in unity

[misty glade]

ah

[undone coral]

i don't totally comprehend why unity's implementation of datetime is so bad

it is the Built In Render Pipeline of time functionality

[misty glade]

so, the object that msgpack is de/serializing here is .... quite large, i don't think 93mb ms is out of bounds

[undone coral]

you're going to use Profiler.BeginSample

it's going to be like 0.2ms

[misty glade]

ok lemme try, standby a sec

wait profiler is a unity thing? should i use stopwatch or profiler? i just want to see the ms delay in console output, not have to fire up the profiler to dig into it

[undone coral]

you should use profiler

because it will give you valuable context

[misty glade]

Profiler.BeginSample is conditionally compiled away using ConditionalAttribute. Thus it will have zero overhead, when it is deployed in non-Development Build.

Oh, actually, yeah, this is nice

[undone coral]

yes

it will be absolutely correct in editor for small stuff

it will be relatively correct for big picture performance

so you can use it to optimize in editor, but if you need absolute performance, you should measure big stuff like total render time in standalone builds

[analog nexus]

It is working well for us so far. Patch is the right verb for the action performed

[misty glade]

Is there any way to get a profiler frame data out to the console? or do I have to enable the profiler to view this info

[undone coral]

i would enable the profiler view at this stage

[misty glade]

I'm mostly just interested in this chunk of init code

[undone coral]

you will be able to find it easily

[flint sage]

Is this first call or are all calls the same speed?

[misty glade]

like, i did that just to get roughly what's failing

[undone coral]

yeah

[misty glade]

This only happens once in the game lifetime

[undone coral]

unfortunately that will give you wrong answers because of the imprecision of DateTime

[misty glade]

"ssc" in the snippet above is rather large

[flint sage]

Chances are then that msgpack is generating more optimized systems to deserialize in the first call

[undone coral]

i don't think msgpack in unity emits

and reflection is slow but it's not that slow

i don't know if it caches the results of attributes. i don't use attribute annotation based stuff

[misty glade]

https://pastebin.com/DSfxDEaU if you wanna see the size/nesting of the data

[undone coral]

you can also try marking classes as sealed if it is properly caching the reflection results

[misty glade]

all of the object classes are sealed and .. pretty plain, but I don't think that makes a difference, since messagepack is generating its own code for deserializing.. but i can't read that particular code since it's a source-generated-mess

7700 lines

[undone coral]

i see

if it's generated sources* (not a .net source generator)

[misty glade]

yeah, required for android

[undone coral]

i guarantee you it does not take 92ms

it's going to take closer to 0.1ms or less

[misty glade]

hm.. ok, doing the profiler thing

[undone coral]

it's going to be immeasurable

[misty glade]

standby 🙂

[undone coral]

it will probably create a lot of GARBAGE which almost all deserialization does

but that will not show up in the time it takes to deserialize

[undone coral]

being stubborn about stuff like this is a guaranteed way to ship a game very very late

it really depends on your objectives

[misty glade]

that's fine, I don't mind garbage since.. theoretically that can be cleaned up later (and is hopefully optimized a bit not to lock up rendering threads)

what do I do with this profiler frame? do I just look for it in the profiler? it'll have it's own line item?

(sorry, trying to read and digest the API for Profiler. really quickly)

that's it?

[undone coral]

you should try to find which frame it was called on. the simplest way to do that is to type in the name of the sample in the search

then scrub the timeline

around when it is called

[misty glade]

oh i can get it in the profiler no problem since it happens on demand (when I click "login" in the client and run the server locally)

[undone coral]

okay great

[misty glade]

it'll be a spike i'm pretty sure

standby

hm... where do i find it..? i think i have the frame but I don't see the profile sample I made

[undone coral]

you should see the list of methods at the bottom

that are being called

one view is the hierarchy, another is the timeline

you can also see it shows you only Main Thread

you can change which thread if it's not being called on the main thread at all

but the best way to od it is to type SSC into the search box

[misty glade]

k

[undone coral]

then scrub gently around when you clicked login

and look carefully at the hierarchy view, it should appear

[misty glade]

crap "SSC" is found in "IsScene"

[undone coral]

lol

you can also write MessagePack

[misty glade]

"Deserializer" is probably the unique search

[undone coral]

sure

[misty glade]

messagepack was my next guess but obv that failed

hundreds of mp calls :p

[undone coral]

and gently scrub

hmm, just to confirm you do not have Deep Profiling turned on correct?

it should be turned off

[misty glade]

oh, i have it on

[undone coral]

turn it off and run it again

[misty glade]

i couldn't find the call anywhere

[analog nexus]

We already have a workaround, I am just asking how it could be explained that we need the workaround for sind people but not others

[misty glade]

k - i'm going to name the key "pangloss" because you deserve the credit 😉

[undone coral]

is it not being called on the main thread? turning off deep profiling will reduce the noise significantly

[undone coral]

because httpclient isn't supported in unity

[undone coral]

lol

[misty glade]

it's being called on the main thread, yes (it's on my TODO to offload this but.. for now it's on the main thread)

[undone coral]

do you mean in some essential sense? because httpclient isn't fully implemented in unity

and unitywebrequest, on most platforms, is using libcurl or the platform http client

you shouldn't ever use httpclient

it's a major sticking point for people

[analog nexus]

Says who

[undone coral]

listen i'm not a content creator and i don't aspire to be one

relying on content creators for programming knowledge is a guaranteed way to ship a game very very late

[misty glade]

You owe me a beer

[undone coral]

so the sooner you start using unitywebrequest the better your game will perform and the fewer bugs you'll experience, on all devices

[analog nexus]

Again, it is understandable that mono is not implementing everything, this is also not my question. The question is why it is working for some people, possibly hinting at a bigger problem of framework version mismatch between people

[undone coral]

lol

i think at this stage you can now enable deep profiling and see what exactly is taking so long

[analog nexus]

I don't know what you're on about about content creators

[misty glade]

Yeah, I mean.. I could but.. I'm a bit frazzled with launch timeline - I think I'm just going to offload the MP deserialization to another thread or coroutine it or something

it works just fine and literally it's one little burp during startup.. I feel like people are mostly ok with a little jankiness during load screens

[undone coral]

because the big difference between self and time is looking not great

[misty glade]

the music plays smoothly during the loading anyway, it's just a little bit of framerate stuttering

[undone coral]

yeah it's going to be fine

it is easy to offload

[undone coral]

you asked me "says who" and what do you want me to link to? a youtube video? a blog post?

unity certainly says use unitywebrequest

anyway

this isn't going anywhere

[analog nexus]

Well you appealed to authority without backing up your claim

[undone coral]

i told you the facts of the situation

that httpclient isn't fully implemented

[undone coral]

there isn't more of a story than that

[analog nexus]

I didn't ask for facts about UWR vs .net

[undone coral]

yeah i think that's fine

lol

the stars

you can easily offload it... with unitask 🙂

[analog nexus]

I asked about how some people have different .net implementations

[undone coral]

they don't

unity standalone player doesn't fully implement httpclient

so you cannot rely on it 100%

[analog nexus]

But they do. It works for me and doesn't for others

[undone coral]

it's okay

you are welcome to bust your head against this bug

you can ship your game late

it's your game

this is #archived-code-advanced

[analog nexus]

Are you saying it stops working once you go from editor to a build

[undone coral]

if you want to die on the hill of "muh visual studio" go fo rit

[undone coral]

i am saying it isn't fully implemented

[analog nexus]

I'm not even using VS

[undone coral]

there are methods that are empty

or buggy or incomplete.

it's that simple

[misty glade]

Dr.P when we start our next game I'm going to have a big long chat with you about UniRX because .. I'm too far in to refactor it (remember when we talked about it like a year ago and I said "this looks cool but I'm going to stay the course" - I wish I hadn't!). Next project is going to be rx based for sure.

[undone coral]

lol

The Faith of UniRx

[misty glade]

I occasionally see you promoting it in answers and I always have a little giggle and sad head shake that I didn't listen earlier 😛

[undone coral]

i am just teasing

[misty glade]

We're doing animation bugs and stuff and ... gosh, what a headache. It would have been so nice to just drop animation events into a stream

[undone coral]

i am trying to emphasize i don't know why you feel so strongly about using an API that is not supported in Unity versus one that is

it would take you an hour to rewrite everything to use the working API. depending on what the game is it might even be 5 minutes

[analog nexus]

Well the one that is supported by unity is not implemented in the same why as mono

[misty glade]

I can't recall if I told you my non-rx solution.. I literally copy a data structure on the server and client, then copy it again on the client, one for the "true" data and another for the "animated" data, and then merge them all up at the end. It works, but man, so fiddly..

[undone coral]

UnityWebRequest has a simple enough API

[undone coral]

yeah

that's reinventing UniRx

[misty glade]

cursed comment

[undone coral]

yeah it took me a while to figure out how to "Delay out" animation events

like at an intellectual level

[misty glade]

I mean, it's only now that I realize that this fundamentally IS rx

[undone coral]

what is the game?

[misty glade]

and so all the work I did... doing this... was just wasted effort because my implementation is not as good as UniRX (probably is)