Unity game stops while coroutine running - unity3d

Unity(character movement and everything in current game scene) stops while downloading textures from web url.
I'm using Socket.IO for online multiplayer.
request 'map info' with socket.emit in void Start()
get 'map info' and create map entities with Instantiate()
get profile images(textures) from url and change texture of map entities
IEnumerator DownloadImage(string MediaUrl, SpriteRenderer spr) {
// check if url is malformed
if (MediaUrl == null || MediaUrl.Contains("Null")) {
yield break;
}
UnityWebRequest request = UnityWebRequestTexture.GetTexture(MediaUrl);
yield return request.SendWebRequest();
if (request.result == UnityWebRequest.Result.ConnectionError || request.result == UnityWebRequest.Result.DataProcessingError || request.result == UnityWebRequest.Result.ProtocolError)
Debug.Log(request.error);
else {
Texture2D tex = ((DownloadHandlerTexture)request.downloadHandler).texture;
spr.sprite = Sprite.Create(tex, new Rect(0.0f, 0.0f, tex.width, tex.height), new Vector2(0.5f, 0.5f), tex.width / 2f);
}
}
Is there a way to keep game running with user interactions(character movement or button actions) while downloading textures?

Unity co-routines (CR) are a entry-level and sometimes hazardous form of performing work over a series of frames on the Main Unity thread. They are useful for what is essentially a logical lerp across a series of frames such as fading, explode-my-spaceship-after-5-seconds, or deferring cockpit display updates.
CRs are sliced up and multiplexed over the Main Thread so the slowest part of your game is going to be the CR with the largest demand for time in a single frame. Performing anything lengthy during a single yield per frame is going to slow down your game, particularly I/O operations.
Alternative - Unity Jobs
Instead, consider using Unity 'IJob' which allows for operations to be executed on a worker thread thus freeing up the Unity thread to do what it does best and not lengthy I/O operations.
Unity:
Use IJob to schedule a single job that runs in parallel to other jobs and the main thread. When a job is scheduled, the job's Execute method is invoked on a worker thread. More...
Now wasting a whole thread that may spend it's life waiting for I/O to complete is arguably a waste of a perfectly good thread but it is certainly better than blocking the Main Thread.
Coroutines...thar be dragons
So earlier I mentioned "hazardous" with regards to CRs and the reason is three-fold:
When firing off a CR in a method that is called many times per second such as Update() make sure you adequately guard the StartCoroutine() else you could end up with zillions of them running only to run out of memory. This one is quite common on Stack Overflow
Don't make the mistake of thinking they are somehow a worker thread or other piece of magic that won't slow down the Main Thread
In many ways CRs are like DoEvents, Application.DoEvents in Visual Basic and .NET respectively and both leads to a nasty case of re-entrancy (where state is clobbered and unpredictable just like in a multi-threaded app but without the threads)
There's also the whole is Unity promoting a strange use of IEnumerator and yield? debate.
My other tip is, if you need something lerped or delayed and isn't doing I/O, consider just making a class and use Time.
e.g. in my flight sim, I only want to update cockpit displays every 100ms or so. For that I define a Delay class:
public class Delay
{
private float _lastInterval;
/// <summary>
/// The timeout in seconds
/// </summary>
/// <param name="timeout"></param>
private Delay(float timeout)
{
Timeout = timeout;
_lastInterval = Time.time;
}
public float Timeout { get; }
public bool IsTimedOut => Time.time> _lastInterval + Timeout;
public void Reset()
{
_lastInterval = Time.time;
}
public static Delay StartNew(float delayInSeconds)
{
return new Delay(delayInSeconds);
}
}
.
.
.
private void Update()
{
if (!_delay.IsTimedOut)
{
return;
}
// Do something time consuming
_delay.Reset();
}

Related

WebSocketSharp .onMessage on main thread? / handling in Unity?

I find it very difficult to find a simple reliable library to use, in Unity, for a simple websocket client. WebSocketSharp (ie, https://github.com/sta/websocket-sharp ) seems fine but many items are unexplained.
When incoming messages arrive,
wssharp_connection.OnMessage += (sender, e) =>
{
if (!e.IsText) { return; }
Debug.Log(">> OnMessage, " + e.Data);
Handle(e.Data);
// but wait, not the main Unity thread??
};
void Handle(string msg)
{
.. your conventional Unity call, in a monobehavior
}
Testing shows it seems to arrive on (always? sometimes?) not the main Unity thread.
Does anyone know
In fact is it always not the main thread, or can it vary?
If always another thread, in fact is it always the same one thread, or does it vary/many?
Since every single Unity app that uses a websocket client has to do this, I'm surprised that there are not 100s of discussion about how to handle incoming messages, including on the WebSocketSharp page,
but here's what I would do, just in the usual way you handle "threading" in a frame-based engine like Unity:
wssharp_connection.OnMessage += (sender, e) =>
{
if (!e.IsText) { return; }
incoming_messages.Enqueue(e.Data);
};
// and ...
ConcurrentQueue<string> incoming_messages = new ConcurrentQueue<string>();
void Update()
{
if (incoming_messages.TryDequeue(out var message))
{
Handle(message);
}
}
So then
Is that indeed the usual way to handle the threading problem in WebSocketSharp, when using inside Unity?
Given how many billion Unity apps there are with websockets clients, this must be a pretty well-trodden path.

UNITY | Update or InvokeRepeating?

I use many timers in my Unity project, they done like this:
void Update()
{
timer -= Time.deltaTime;
if(timer < 0)
{
DoSomething();
}
}
And in every google link they looks like this.
But today i found (im newbie) InvokeRepeating(); method.
So here is updated timer
int timer = 60;
void Start()
{
InvokeRepeating("Timer", 1f, 1f);
}
void Timer()
{
timer -= 1;
}
So why people change timers in Update()?
And which method less inpact on performance?
Depends!
As usual in programming there are most of the time multiple valid solutions for a problem.
In my eyes the bigest difference is that InvokeRepeating is also working on inactive GameObjects or disabled Components while Update is only called while the object is active and the component enabled.
Note, however, that your current examples do different things. To make them equivalent it should look like either
void Start()
{
timer = 1f;
}
void Update()
{
timer -= Time.deltaTime;
if(timer < 0)
{
timer = 1f;
DoSomething();
}
}
or
void Start()
{
InvokeRepeating(nameof(DoSomething), 1f, 1f);
}
Btw: A third basically equivalent solution would be a Coroutine (which is basically a temporary Update method - in fact the MoveNext call gets executed right after the Update would)
// Yes, Start can be an IEnumertaor and is in this case internally implicitly started as Coroutine!
IEnumerator Start()
{
while(true)
{
yield return new WaitForSeconds(1f);
DoSeomthing();
}
}
As mentioned by Kuruchy: There is also a difference in behavior related to the timeScale.
Update: since it uses Time.deltaTime to decrease the timer it will be affected by the time scale
→ in order to avoid this you would need to use Time.unscaledDeltaTime instead. But afaik still if setting Time.timeScale = 0; then Update isn't called at all.
Coroutine: Similar to update the WaitForSeconds is also timeScale dependent.
→ in roder to avoid this you would need to use WaitForSecondsRealitme but afaik even then setting Time.timeScale = 0; would lead to the routine not getting called at all.
InvokeRepeating: Fromt he docs I can only see that
This does not work if you set the time scale to 0.
not sure how it reacts to the timescale though but I would actually guess the same. Though, here there is no real work around like for the other two.
Performance wise you most probably don't even have to care! I guess it would be over-optimizing.
However my guess would be that Update is actually slightly faster since there you already know the method reference while when using InvokeRepeating you are passing it as a string which is additionally error prone and means that internally Unity has to find that method first.
Without going into too much detail is because of the performance.
The Update() method is invoked internally by Unity and they've done a pretty good job in optimizing it.
InvokeRepeating is much slower in comparison. First of all because the initial method invocation is using the Reflection to find the method you want to start and it's respective calls also take more time than Update. And you want to avoid using Reflection in your code as much as possible.
Here is nice, short article in which the tests were performed between these two methods - http://www.kittehface.com/2017/09/unity-performance-with-invokerepeating.html

Unet Syncing animator layer weight Unity 3D

I'm currently working on a multiplayer fps to learn UNET and Unity. in this project I'm using Network Animator Component to sync my animations.
I have one layer that runs the normal states like walking and jumping and another that just controls the upper body for scoping. This layer gets activated when scoping.
The parameters a syncing fine but not the layer weight.
The script that controls the layer weight gets disabled on the remote player as it is part of the same script that takes care of shooting and we don't want to shoot from the remote player.
So I placed that code on a different script and made Layerweight a Syncvar, still no results.
this is the snippet that sets the weight and fades it back out.
if (isScoped)
{
animator.SetLayerWeight(1, 1);
layerWeight = 1;
}
else
{
layerWeight = Mathf.Lerp(layerWeight, 0, 0.1f);
animator.SetLayerWeight(1, layerWeight);
}
animator.SetBool("Scoped", isScoped);
how can I sync a variable that is updated on a disabled script?
And what are the different custom attributes like [Command] and [ClientRpc] I should use to get this working?
I have read the documentation and manuals of UNET, but struggle at this relatively easy task as I'm having a hard time understanding the difference of these calls.
In general the main problem with [SyncVar] is that they are
synchronized from the server to clients
so not the other way round!
Then
[Command] is a method that is called by any client but only executed on the server
[ClientRpc] is kind of the opposite, it is called by the server and then executed on all clients
so what I would do is have a setter method for the layerweight like
private void SetLayerWeight(float value)
{
// only player with authority may do so
if(!isLocalPlayer) return;
// Tell the server to update the value
CmdSetLayerWeight(value);
}
[Command]
private void CmdSetLayerWeight(float value)
{
// Server may now tell it to all clients
RpcSetLayerWeight(value);
}
[ClientRpc]
private void RpcSetLayerWeight(float value)
{
// called on all clients including server
layerweight = value;
animator.SetLayerWeight(1, layerweight);
}
and call it like
if (isScoped)
{
SetLayerWeight(1);
}
else
{
SetLayerWeight(Mathf.Lerp(layerWeight, 0, 0.1f));
}
Note though that passing this forth and back between server and client will cause some uncanny lag!
You would probably be better syncing the isScoped bool instead and assign the value directly on the calling client:
private void SetScoped(bool value)
{
// only player with authority may do so
if(!isLocalPlayer) return;
// for yourself already apply the value directly so you don't have
// to wait until it comes back from the server
scoped = value;
// Tell the server to update the value and who you are
CmdSetLayerWeight(value);
}
// Executed only on the server
[Command]
private void CmdSetLayerWeight(bool value)
{
// Server may now tell it to all clients
RpcSetLayerWeight(value);
}
// Executed on all clients
[ClientRpc]
private void RpcSetLayerWeight(bool value)
{
// called on all clients including server
// since setting a bool is not expensive it doesn't matter
// if this is done again also on the client who originally initialized this
scoped = value;
}

How to control particles directly while preventing the particle system from doing the same?

I have lots of same simple objects that are affecting gameplay, thousands of them! Well, not thousands, but really many. So if I make them GameObjects, the FPS decreases, especially when spawning them. Even with pooling. I should try a different approach.
You know that particle system in Unity3D can render many particles very fast. It also automatically controls particles, emits and removes them. But in my case, positions and lifetimes of objects are managed by game logic, and particle system is not allowed to do anything without my command, even reorder particles.
I am trying to use SetParticles method to control particles. It works in test project, where I use GetParticles first. I can even remove particles setting lifetime to -1, but can't spawn new ones. Also it does not prevent particle system from controlling particles.
I can disable emission so no particles will be created automatically.
I can set particles speed to 0 so they will not move.
I can set lifetime to huge number so they will not be removed.
I have an pool of Particle instances, to avoid unnessessary GC allocations. Objects receive a reference to particle when they are spawned, change it when updated, and set lifetime -1 and return it to pool when deleted. The pool stores this:
private ParticleSystem.Particle [] _unusedParticles;
private int _unusedCount;
private ParticleSystem.Particle [] _array;
_unused array and counter are needed for pooling, and _array stores all particles, used and unused alike, and is used in SetParticles call.
The main disadvantage of this method is that it doesn't work, probably because SetParticles does not create new particles. Also I guess it doesn't do anything with particles draw order, that's why it's poorly suited for bullet hell games where bullet patterns should look nice.
What should I do to properly disable automatic control of particles and properly setup direct control, with spawning and removing?
What you are looking for might be
List<Matrix4x4> matrixes=new List<Matrix4x4>();
for (...)
{
matrixes.Add(Matrix4x4.TRS( position,rotation,scale));
}
Graphics.DrawMeshInstanced(mesh,0,material, matrixes);
On each frame you can just update positions, rotations and scales, and get all the instances rendered on the GPU in one drawcall (pretty darn fast compared to seperate gameobjects). You can render up to 1000 instances in one call using this way
Create an empty GameObject, then add the ParticleSystem as child. Set playOnAwake to true.
Now when you need it, set GameObject.SetActive to true else false.
To get each of them use ParticleSystem.GetParticles, modify them and ParticleSystem.GetParticles.
I hope this is what you are looking for.
ParticleSystem m_System;
ParticleSystem.Particle[] m_Particles;
public float m_Drift = 0.01f;
private void LateUpdate()
{
InitializeIfNeeded();
// GetParticles is allocation free because we reuse the m_Particles buffer between updates
int numParticlesAlive = m_System.GetParticles(m_Particles);
// Change only the particles that are alive
for (int i = 0; i < numParticlesAlive; i++)
{
m_Particles[i].velocity += Vector3.up * m_Drift;
}
// Apply the particle changes to the Particle System
m_System.SetParticles(m_Particles, numParticlesAlive);
}
void InitializeIfNeeded()
{
if (m_System == null)
m_System = GetComponent<ParticleSystem>();
if (m_Particles == null || m_Particles.Length < m_System.main.maxParticles)
m_Particles = new ParticleSystem.Particle[m_System.main.maxParticles];
}
After we created a particle system in editor, we should disable emission and shape, so only core part and renderer stay active.
The most important part is that Simulation Speed must be zero. A particle system will no longer emit, remove or process particles automatically. Only your code now manages them.
I use this class to control particles. Instead of binding a particle to an object, it has API for registering objects, smoke in this case. Also, it stores a temporary array for particles to avoid GC allocations, and particle count, to avoid using particleCount property of the particle system.
In Update, which is called by my game logic, the following happens:
All objects that were despawned by game logic, are removed from list.
If object count is greater than array length, the array is resized.
If object count is greater than live particle count, _particleSystem.Emit call is made. If we call SetParticles without that, no new particles will appear. We must emit them first.
GetParticles is called. Simple, but probably not the most efficient solution, but it preserves particles data. It may be optimized by setting all particle data on array creation and resizing. If you do so, remove GetParticles call and uncomment the line above. Also, your game logic should manage particles even more carefully.
For each object, we let that object change a particle.
Each particle without object should be removed, so their lifetime is set to negative number.
SetParticles updates particles in system.
public class SmokeSystem {
private ParticleSystem _particleSystem;
private List <Smoke> _smoke = new List <Smoke> ();
private ParticleSystem.Particle [] _particles = new ParticleSystem.Particle[256];
private int _particleCount;
public SmokeSystem (ParticleSystem particleSystem) {
_particleSystem = particleSystem;
}
public void AddSmoke (Smoke smoke) => _smoke.Add (smoke);
public void Update () {
_smoke.RemoveAll (e => e.Despawned);
if (_smoke.Count > _particles.Length) {
int newSize = Max (_smoke.Count, 2 * _particles.Length);
Array.Resize (ref _particles, newSize);
}
int count = _smoke.Count;
if (count > _particleCount) {
_particleSystem.Emit (count - _particleCount);
// _particleCount = count;
}
_particleCount = _particleSystem.GetParticles (_particles);
for (int i = 0; i < count; i++) {
_smoke [i].UpdateParticle (ref _particles [i]);
}
for (int i = count; i < _particleCount; i++) {
_particles [i].remainingLifetime = -1;
}
_particleSystem.SetParticles (_particles, _particleCount);
_particleCount = count;
}
}
It does not depend on GPU instancing support so it will work on WebGL.

Unity: Weird scene transition bug after adding a "Persistent-Scene" with a GameManager

I made a pretty basic 2D game to learn. I have 2 Scenes, and switching between them worked great. I used empty gameObjects as Start/Exit point of the Scene, so that the game would know to put player on point X after exiting through point X (for example exit outside house if I walk out the door).
Then I added a "Scene0", to use for persistent general scripts like GameManager, Sounds, Music, etc. With just one object called "Controller" that I DontDestroyOnLoad().
After adding this Scene and then just switching Scenes right away to my MainScene, all of a sudden the game starts acting really strange;
the first time I move from my MainScene (Scene1), to my secondary Scene (Scene2), it works fine, but then when I leave Scene2 to go back to Scene1, the player spawns in the middle of nowhere.
And this ONLY happens if I launch the game from my Persistent Scene.
I have no idea what is wrong, I don't add anything that interferes with my scene transitions, all I've added so far is playerHealth, for testing.
Scripts attached to my (persistent) Controller:
DDOL:
public class DDOL : MonoBehaviour {
// Use this for initialization
void Awake () {
DontDestroyOnLoad (gameObject);
}
}
GameManager:
public class GameManager : MonoBehaviour {
public static GameManager manager;
public int playerMaxHealth;
public int playerCurrentHealth;
void Awake(){
if (manager == null) {
manager = this;
} else if (manager != this) {
Destroy (gameObject);
}
}
// Use this for initialization
void Start () {
SceneManager.LoadScene("test_scene");
}
// Update is called once per frame
void Update () {
}
}
Scripts attached to my StartPoint:
PlayerStartPoint:
public class PlayerStartPoint : MonoBehaviour {
private PlayerController thePlayer;
private CameraController theCamera;
public Vector2 startDir;
public string pointName;
// Use this for initialization
void Start () {
thePlayer = FindObjectOfType<PlayerController> ();
if (thePlayer.startPoint == pointName) {
thePlayer.transform.position = transform.position;
thePlayer.lastMove = startDir;
theCamera = FindObjectOfType<CameraController> ();
theCamera.transform.position = new Vector3(transform.position.x, transform.position.y, theCamera.transform.position.z);
}
}
}
And finally ExitPoint:
LoadNewArea:
public class LoadNewArea : MonoBehaviour {
public string levelToLoad;
public string exitPoint;
private PlayerController thePlayer;
// Use this for initialization
void Start () {
thePlayer = FindObjectOfType<PlayerController> ();
}
void OnTriggerEnter2D(Collider2D other){
if (other.gameObject.name == "Player")
{
SceneManager.LoadScene(levelToLoad);
thePlayer.startPoint = exitPoint;
}
}
}
EDIT:
After moving all my DDOL gameObject to the Pre-Scene, it worked. So, I can assume the fault is inside Player or Cameras Start() functions since when they start in Scene1 they get called every time I enter the Scene (only DDOL).
I tried adjusting their Start()functions like follows:
Original camera:
void Start () {
Debug.Log("Starting camera");
if (!cameraExists) {
cameraExists = true;
DontDestroyOnLoad (gameObject);}
else{
Destroy (gameObject);
}
}
Changed Camera:
void Start () {
DontDestroyOnLoad (gameObject);
}
The exact same changes was made in Player.
Obviously this doesnt work because it creates a new Camera/Player every time I enter Scene1 (btw why does it not try to create them when I enter Scene2?, is it because they start in Scene1?). HOWEVER, the new player/camera do start at the correct position, and if I zoom out I can see the old player/camera at that same wrong position as before. So something weird happens when their Start() is called a second time it seems.
You've now mentioned that you had code something like this,
void Start () {
Debug.Log("Starting camera");
if (!cameraExists) {
cameraExists = true;
DontDestroyOnLoad (gameObject);}
else{
Destroy (gameObject);
}
}
Note that this is unfortunately just "utterly incorrect", heh :)
The issues you mention in the question (preload scenes etc) are just totally unrelated to the problem here.
In Unity if you have a character C that persists between scenes a, b, c as you load those scenes, you must kick-off C in it's own (perhaps otherwise empty) scene, you can not use "a" as a matter of convenience to kick off C.
The pattern is, in each of a, b, c just have a line of code like p = FindObjectOfType<Player>(); which runs when the scene loads, and position C as you wish.
Now, regarding your specific puzzle about the unusual behavior you are seeing.
I understand that you want to know why you are observing what you do.
It is a combination of confusion over the following issues: 1 - difference between Awake and Start, 2 - confusion over script execution order {but see below1} 3 - confusion about Destroy versus DestroyImmediate 4 - Not using Debug.Log enough, and not using gameObject.name in there (it's a common in Unity to be wildly confused about which object is talking in Debug.Log) 5 - where you mention you see the other object "off to the side", it's common to drastically confuse which one is which in such situations 6 - confusion between the computer programming concept of "instantiation" (ie, of a class or object) and "instantiating" (confusingly, it's the same word - utterly unrelated) game objects in nity.
If you fiddle around with all those issues, you'll discover an explanation for the behavior you're seeing!
But it doesn't amount to much; in Unity in the "C .. a b c" example you have to create C separately beforehand.
1 {aside, never fiddle with the script execution ordering system in Unity in an effort to solve problems; it's only there for R&D purposes; however it could in fact help you investigate the behavior at hand in this problem, if you are particularly keen to fully understand why you're seeing what you're apparently seeing}
Use the debugger. Have breakpoints at the relevant spots, like PlayerStartPoint.Start() and LoadNewArea.OnTriggerEnter2D() and check that they are executed
At the right time
The right number of times
With the expected values
This should make you see where things get out of hand.
If you use Visual Studio, install https://marketplace.visualstudio.com/items?itemName=SebastienLebreton.VisualStudio2015ToolsforUnity to be able to debug Unity from within Visual Studio.
If you are not using Visual Studio, you probably should.
Is player persistent between scenes (does he have DontDestroyOnLoad)? If no then this might be the reason - you can either try loading the scenes by using the additive mode or by instantiating the player on scene load in correct position.