Does any one know when is the best time to stop an NSTimer that is held reference inside of a UIViewController to avoid retain cycle between the timer and the controller?
Here is the question in more details: I have an NSTimer inside of a UIViewController.
During ViewDidLoad of the view controller, I start the timer:
statusTimer = [NSTimer scheduledTimerWithTimeInterval: 1 target: self selector: #selector(updateStatus) userInfo: nil repeats: YES];
The above causes the timer to hold a reference to the view controller.
Now I want to release my controller (parent controller releases it for example)
the question is: where can I put the call to [statusTimer invalidate] to force the timer to release the reference to the controller?
I tried putting it in ViewDidUnload, but that does not get fired until the view receives a memory warning, so not a good place. I tried dealloc, but dealloc will never get called as long as the timer is alive (chicken & egg problem).
Any good suggestions?
You could avoid the retain cycle to begin with by, e.g., aiming the timer at a StatusUpdate object that holds a non-retained (weak) reference to your controller, or by having a StatusUpdater that is initialized with a pointer your controller, holds a weak reference to that, and sets up the timer for you.
You could have the view stop the timer in -willMoveToWindow: when the target window is nil (which should handle the counterexample to -viewDidDisappear: that you provided) as well as in -viewDidDisappear:. This does mean your view is reaching back into your controller; you could avoid reaching in to grab the timer by just send the controller a -view:willMoveToWindow: message or by posting a notification, if you care.
Presumably, you're the one causing the view to be removed from the window, so you could add a line to stop the timer alongside the line that evicts the view.
You could use a non-repeating timer. It will invalidate as soon as it fires. You can then test in the callback whether a new non-repeating timer should be created, and, if so, create it. The unwanted retain cycle will then only keep the timer and controller pair around till the next fire date. With a 1 second fire date, you wouldn't have much to worry about.
Every suggestion but the first is a way to live with the retain cycle and break it at the appropriate time. The first suggestion actually avoids the retain cycle.
One way around it is to make the NStimer hold a weak reference to your UIViewController. I created a class that holds a weak reference to your object and forwards the calls to that:
#import <Foundation/Foundation.h>
#interface WeakRefClass : NSObject
+ (id) getWeakReferenceOf: (id) source;
- (void)forwardInvocation:(NSInvocation *)anInvocation;
#property(nonatomic,assign) id source;
#end
#implementation WeakRefClass
#synthesize source;
- (id)init{
self = [super init];
// if (self) {
// }
return self;
}
+ (id) getWeakReferenceOf: (id) _source{
WeakRefClass* ref = [[WeakRefClass alloc]init];
ref.source = _source; //hold weak reference to original class
return [ref autorelease];
}
- (NSMethodSignature *)methodSignatureForSelector:(SEL)aSelector {
return [[self.source class ] instanceMethodSignatureForSelector:aSelector];
}
- (void)forwardInvocation:(NSInvocation *)anInvocation
{
[anInvocation invokeWithTarget:self.source ];
}
#end
and you use it like this:
statusTimer = [NSTimer scheduledTimerWithTimeInterval: 1 target: [WeakRefClass getWeakReferenceOf:self] selector: #selector(updateStatus) userInfo: nil repeats: YES];
Your dealloc method gets called (unlike before) and inside it you just call:
[statusTimer invalidate];
You can try with - (void)viewDidDisappear:(BOOL)animated and then you should validate it again in - (void)viewDidAppear:(BOOL)animated
More here
For #available(iOS 10.0, *) you could also use:
Timer.scheduledTimer(
withTimeInterval: 1,
repeats: true,
block: { [weak self] _ in
self?.updateStatus()
}
)
The -viewDidDisappear method may be what you're looking for. It's called whenever the view is hidden or dismissed.
invalidate timer inside - (void)viewWillDisappear:(BOOL)animated did work for me
I wrote a "weak reference" class for exactly this reason. It subclasses NSObject, but forwards all methods that NSObject doesn't support to a target object. The timer retains the weakref, but the weakref doesn't retain its target, so there's no retain cycle.
The target calls [weakref clear] and [timer invalidate] or so in dealloc. Icky, isn't it?
(The next obvious thing is to write your own timer class that handles all of this for you.)
If the timer.REPEAT is set to YES, the owner of the timer (e.g. view controller or view) will not be deallocated until the timer is invalidated.
The solution to this question is to find some trigger point to stop your timer.
For example, I start a timer to play animated GIF images in a view, and the trigger point would be:
when the view is added to the superview, start the timer
when the view is removed from the superview, stop the timer
so I choose the UIView's willMoveToWindow: method as such:
- (void)willMoveToWindow:(UIWindow *)newWindow {
if (self.animatedImages && newWindow) {
_animationTimer = [NSTimer scheduledTimerWithTimeInterval:_animationInterval
target:self selector:#selector(drawAnimationImages)
userInfo:nil repeats:YES];
} else {
[_animationTimer invalidate];
_animationTimer = nil;
}
}
If your timer is owned by a ViewController, maybe viewWillAppear: and viewWillDisappear: are a good place for you to start and stop the timer.
I had exactly the same issue and in the end I decided to override the release method of the View Controller to look for the special case of the retainCount being 2 and my timer running. If the timer wasn't running then this would have caused the release count to drop to zero and then call dealloc.
- (oneway void) release {
// Check for special case where the only retain is from the timer
if (bTimerRunning && [self retainCount] == 2) {
bTimerRunning = NO;
[gameLoopTimer invalidate];
}
[super release];
}
I prefer this approach because it keeps it simple and encapsulated within the one object, i.e., the View Controller and therefore easier to debug. I don't like, however, mucking about with the retain/release chain but I cannot find a way around this.
Hope this helps and if you do find a better approach would love to hear it too.
Dave
EDIT: Should have been -(oneway void)
You can write this code in dealloc function of view controller
for eg.
-(void)dealloc
{
if([statusTimer isValid])
{
[statusTimer inValidate];
[statustimer release];
statusTimer = nil;
}
}
this way the reference counter of statustimer will automatically decrement by 1
& also the data on the allocated memory will also erase
also you can write this code in - (void)viewDidDisappear:(BOOL)animated function
Related
I run my animations in a UITAbleViewCell.
Each cell has its own animation and the cells are reusable.
I use [mView performSelectorInBackground:#selector(layoutSubview) withObject:nil];
There in the background thread I initiate the runLoop to perform tasks like this:
- (void)startAnimation
{
NSRunLoop *mLoop = [NSRunLoop currentRunLoop];
self.animationTimer = [NSTimer scheduledTimerWithTimeInterval:animationInterval target:self selector:#selector(setNeedsDisplay) userInfo:nil repeats:YES];
mRunLoop = YES;
while (mRunLoop == YES && [mLoop runMode:NSDefaultRunLoopMode beforeDate:[NSDate dateWithTimeIntervalSinceNow:0.01]]);
}
and stop it:
- (void)stopAnimation
{
if (![NSThread isMainThread]) {
[[NSThread currentThread] cancel];
}
mRunLoop = NO;
self.animationTimer = nil;
CFRunLoopStop(CFRunLoopGetCurrent());
}
I run into problems when I fast scroll through table, because on the first cell initiation I begin the animation, so the first runLoop call occures which performs a setNeedDisplay and all the methods from it. But before finishing the first runLoop cycle the cell disappears from the view and is already available for reuse. So I begin clearing it, while the cycle is still performing operations and here I meet situations like
message sent to deallocated instance
So could you please give me some hints of how should I correctly stop performing the operations in that thread? I mean if I want to realese for example an object, which is performing some actions how to immediately stop'em?
Hope I gave enough info.
Thanks
UPDATE: No ideas at all?
I'll take a completely different stab on it:
Get rid of the cell's timers and background threads altogether!
Animation is not something where NSTimer is a good fit in the first place and having multiple timers won't help much, either.
UITableView has a method visibleCells and a method indexPathsForVisibleRows. I'd suggest to use a single CADisplayLink — which is suited for animation, as it calls you back with the actual refresh rate of the display or a fraction thereof — in your tableview-controller and in the callback of that display-link iterate over the visible cells.
If you want to schedule the display-link on the run-loop of a secondary thread, feel free to do so, but I'd check if you can get away without extra threading first.
Some code:
#interface AnimatedTableViewController ()
#property (strong, nonatomic) CADisplayLink *cellAnimator;
- (void)__cellAnimatorFired:(CADisplayLink *)animator;
#end
#implementation AnimatedTableViewController
#synthesize cellAnimator = cellAnimator_;
- (void)setCellAnimator:(CADisplayLink *)animator
{
if (animator == cellAnimator_)
return;
[cellAnimator_ invalidate];
cellAnimator_ = animator;
[cellAnimator_ addToRunLoop:[NSRunLoop currentRunLoop] forMode:NSCommonRunLoopModes];
}
- (void)viewDidAppear:(BOOL)animated
{
[super viewDidAppear:animated];
self.cellAnimator = [CADisplayLink displayLinkWithTarget:self selector:#selector(__cellAnimatorFired:)];
...
}
- (void)viewWillDisappear:(BOOL)animated
{
self.cellAnimator = nil;
...
[super viewWillDisappear:animated];
}
- (void)__cellAnimatorFired:(CADisplayLink *)animator
{
NSArray *visibleCells = [self.tableView visibleCells];
[visibleCells enumerateObjectsUsingBlock:^(UITableViewCell *cell, NSUInteger unused, BOOL *stop){
[cell setNeedsDisplay];
}];
}
...
#end
NSTimer has a -cancel method that stops the timer from firing. Calling it in -prepareForReuse (and, for that matter, in -stopAnimation) may help.
However, this code looks rather dangerous. Nesting run loops like this is almost never a good idea—and moreover, as far as I can tell it's totally unnecessary. If you let -startAnimation return, your animation timer will still get run on the main run loop. And if you're doing it this way because there's some code after -startAnimation that you want to delay, you should restructure your code so this isn't needed.
(If you drop the runloop stuff in -startAnimation, don't stop the runloop in -stopAnimation either.)
Something like the approach danyowdee recommends would be even better, but at least get rid of this runloop stuff. It's just asking for trouble.
I think you can use this method for your problem
[NSObject cancelPreviousPerformRequestsWithTarget:yourTarget selector:aSelector object: anArgument];
I think that the best way to avoid that behavior is assigning the delegate that receives the cancel method in other class that won't be reused. For example, you can have a private array of instances that process all the cancel methods, each row mapped into an array element.
I recommend you the lazy tables example provided by Apple in Xcode documentation. It's a great example of how to load images asynchroniously in background with a table. I think that also it would be useful for you for the scrolling subjects (decelerating and paging).
Only one more consideration, i don't recommend messing up with several cfrunloopstop, test it hard!
I've read up a lot about NSTimers, but I must be doing something very wrong with them, because it's practically all the leaks that show up in the Leaks Instrument. The "Responsible Frame" column says -[NSCFTimer or +[NSTimer(NSTimer).
So here's how I have an NSTimer set up in my main menu. I shortened it up to just show how the timer is setup.
.h -
#interface MainMenu : UIView {
NSTimer *timer_porthole;
}
#end
#interface MainMenu ()
-(void) onTimer_porthole:(NSTimer*)timer;
#end
.m -
(in initWithFrame)
- (id)initWithFrame:(CGRect)frame {
self = [super initWithFrame:frame];
if (self) {
timer_porthole = [[NSTimer scheduledTimerWithTimeInterval:.05
target:self
selector:#selector(onTimer_porthole:)
userInfo:nil
repeats:YES] retain];
}
return self;
}
When leaving the view, it kills the timers:
-(void) kill_timers{
[timer_porthole invalidate];
timer_porthole=nil;
}
And of course, dealloc:
- (void)dealloc {
[timer_porthole invalidate];
[timer_porthole release];
timer_porthole = nil;
[super dealloc];
}
Don't call retain on your NSTimer!
I know it sounds counter-intuitive but when you create the instance it's automatically registered with the current (probaby main) threads run loop (NSRunLoop). Here's what Apple have to say on the subject...
Timers work in conjunction with run
loops. To use a timer effectively, you
should be aware of how run loops
operate—see NSRunLoop and Threading
Programming Guide. Note in particular
that run loops retain their timers, so
you can release a timer after you have
added it to a run loop.
Once scheduled on a run loop, the
timer fires at the specified interval
until it is invalidated. A
non-repeating timer invalidates itself
immediately after it fires. However,
for a repeating timer, you must
invalidate the timer object yourself
by calling its invalidate method.
Calling this method requests the
removal of the timer from the current
run loop; as a result, you should
always call the invalidate method from
the same thread on which the timer was
installed. Invalidating the timer
immediately disables it so that it no
longer affects the run loop. The run
loop then removes and releases the
timer, either just before the
invalidate method returns or at some
later point. Once invalidated, timer
objects cannot be reused.
Quotes are sourced from Apple's NSTimer class reference.
So your instantiation becomes...
timer_porthole = [NSTimer scheduledTimerWithTimeInterval:.05
target:self
selector:#selector(onTimer_porthole:)
userInfo:nil
repeats:YES];
And now that you're no longer holding the reference to the instance you wont want the release call in your dealloc method.
I've seen you already accepted an answer but there are two things here that I wanted to rectify:
It's not needed to retain a scheduled timer but it doesn't do any harm (as long as you release it when it's no longer needed). The "problematic" part of a timer/target relationship is that...
a timer retains its target. And you've decided to set that target to self.
That means — retained or not — the timer will keep your object alive, as long as the timer is valid.
With that in mind, let's revisit your code from bottom to top:
- (void)dealloc {
[timer_porthole invalidate]; // 1
[timer_porthole release];
timer_porthole = nil; // 2
[super dealloc];
}
1 is pointless:
If timer_porthole was still a valid timer (i.e. scheduled on a runloop) it would retain your object, so this method wouldn't be called in the first place...
2 no point here, either:
This is dealloc! When [super dealloc] returns, the memory that your instance occupied on the heap will be freed. Sure you can nil out your part of the heap before it gets freed. But why bother?
Then there is
-(void) kill_timers{
[timer_porthole invalidate];
timer_porthole=nil; // 3
}
3 given your initializer (and as others have pointed out) you are leaking your timer here; there should be a [timer_porthole release] before this line.
PS:
If you think it all over, you'll see that retaining the timer (at least temporarily) creates a retain-cycle. In this particular case that happens to be a non-issue which is resolved as soon as the timer is invalidated...
You missed [timer_porthole release]; call in your kill_timers method. If you call kill_timers before dealloc method is called, you set timer_porthole to nil, but you did not release it.
I thought I would put this out here as a separate question from my previous
retaining-repeating-nstimer-for-later-access as the discussion has moved forward making a new question clearer than yet another EDIT:
The scenario is an object creates a repeating NSTimer, lets say in viewDidLoad, once created the NSTimer needs to stay around so it can be accessed by other methods.
NSTimer *ti = [NSTimer scheduledTimerWithTimeInterval:1
target:self
selector:#selector(updateDisplay:)
userInfo:nil
repeats:YES];
I understand that when created the runloop takes ownership of the NSTimer and ultimately stops, removes and releases the NSTimer when [ti invalidate]; is called.
By virtue of the fact that we need to access the NSTimer in more than one method we need some way to hold a reference for future use, the revised question is:
// (1) Should the NSTimer be held using an owning reference (i.e.)
#property(nonatomic, retain) NSTimer *walkTimer;
[self setWalkTimer: ti];
...
...
// Cancel method
[[self walkTimer] invalidate;
[self setWalkTimer:nil];
...
...
// dealloc method
[walkTimer release];
[super dealloc];
.
// (2) Should the NSTimer be held using a weak reference (i.e.)
#property(nonatomic, assign) NSTimer *walkTimer;
[self setWalkTimer: ti];
...
...
// Cancel method
[[self walkTimer] invalidate];
[self setWalkTimer:nil];
...
...
// dealloc method
[super dealloc];
.
// (3) Use an iVar and rely on the runLoop holding (i.e. retaining) the timer
NSTimer *walkTimer;
NSTimer *walkTimer = [NSTimer scheduledTimerWithTimeInterval:1
target:self
selector:#selector(updateDisplay:)
userInfo:nil
repeats:YES];
...
...
// Cancel method
[walkTimer invalidate];
walkTimer = nil;
.
// (4) Something not listed above ...
I am happy for just (1) (2) (3) or (4) as a lot of discussion regarding which is best has already been written on the Other thread. There does seem to be a lot of conflicting answers so I hope this more specific question will help focus on what might be best practice in this situation.
EDIT:
As a side note in the Apple NSTimer Class Reference 4 out of 5 of the sample code projects use NSTimers that are assigned** to a retained property. Here is an example of what the class reference examples show:
#property (nonatomic, retain) NSTimer *updateTimer;
updateTimer = [NSTimer scheduledTimerWithTimeInterval:.01 target:self selector:#selector(updateCurrentTime) userInfo:p repeats:YES];
...
...
// Cancel
[updateTimer invalidate];
updateTimer = nil;
...
...
// Dealloc method
[super dealloc];
[updateTimer release];
** It should be noted that in the examples Apple are assigning the iVar directly and not using the property setter.
After giving it all some more thought and finding an important flaw in my reasoning, I've come to a different conclusion:
It doesn't matter much, whether you hold an owning or a non-owning reference to a timer that you need to invalidate. It is completely a matter of taste.
The deal breaker is, what the target of the timer is:
If the object that creates a timer is its target, managing that object's lifetime becomes more fragile: it cannot simply be retain/release managed, instead you need to ensure that the client that holds the last reference to this object makes it invalidate the timer before it disposes of it.
Let me illustrate the situation with a couple of sort-of-object-graphs:
You start in a state from which you setup the timer and set yourself as the target. Setup of the Timer: yourObject is owned by someClientObject. In parallel exists the current run-loop with an array of scheduledTimers. the setupTimer method is called upon yourObject:
The result is the following initial state. In addition to the former state yourObject now has a reference (owned or not) to the workTimer, which in turn owns yourObject. Furthermore, workTimer is owned by the run-loops scheduledTimers array:
So now you'll use the object, but when you're done with it and simply release it, you'll end up with simple release leak: after someClientObject disposes of yourObject through a simple release, yourObject is disassociated from the object-graph but kept alive by workTimer. workTimer and yourObject are leaked!
Where you leak the object (and the timer) because the runloop keeps the timer alive, which — in turn — keeps an owning reference to your object.
This can be avoided if yourObject is only ever owned by one single instance at a time, when it is properly disposed of proper disposal through cancellation: before disposing of yourObject through release, someClientObject calls the cancelTimer method on yourObject. Within that method, yourObject invalidates workTimer and (if it owned workTimer) disposes of workTimer through release:
But now, how do you resolve the following situation?
Multiple Owners: Setup like in the initial state, but now with multiple independent clientObjects that hold references to yourObject
There is no easy answer, I am aware of! (Not that the latter has to say much, but...)
So my advice is...
Don't make your timer a property/don't provide accessors for it! Instead, keep it private (with the modern runtime I think you could go so far as to define the ivar in a class extension) and only deal with it from one single object. (You may retain it, if you feel more comfortable doing so, but there is absolutely no need for it.)
Caveat: If you absolutely need to access the timer from another object, make the property retain the timer (as that is the only way to avoid crashes caused by clients that directly invalidated the timer they accessed) and provide your own setter. Rescheduling a timer is — in my opinion — not a good reason to break encapsulation here: provide a mutator if you need to do that.
Set the timer up with a target other than self. (There are plenty of ways doing so. Maybe through writing a generic TimerTarget class or — if you can use it — through a MAZeroingWeakReference?)
I apologize for being a moron in the first discussion and want to thank Daniel Dickison and Rob Napier for their patience.
So here is the way I am going to handle timers from now on:
// NSTimer+D12WeakTimerTarget.h:
#import <Foundation/NSTimer.h>
#interface NSTimer (D12WeakTimerTarget)
+(NSTimer *)D12scheduledTimerWithTimeInterval:(NSTimeInterval)ti weakTarget:(id)target selector:(SEL)selector userInfo:(id)userInfo repeats:(BOOL)shouldRepeat logsDeallocation:(BOOL)shouldLogDealloc;
#end
// NSTimer+D12WeakTimerTarget.m:
#import "NSTimer+D12WeakTimerTarget.h"
#interface D12WeakTimerTarget : NSObject {
__weak id weakTarget;
SEL selector;
// for logging purposes:
BOOL logging;
NSString *targetDescription;
}
-(id)initWithTarget:(id)target selector:(SEL)aSelector shouldLog:(BOOL)shouldLogDealloc;
-(void)passthroughFiredTimer:(NSTimer *)aTimer;
-(void)dumbCallbackTimer:(NSTimer *)aTimer;
#end
#implementation D12WeakTimerTarget
-(id)initWithTarget:(id)target selector:(SEL)aSelector shouldLog:(BOOL)shouldLogDealloc
{
self = [super init];
if ( !self )
return nil;
logging = shouldLogDealloc;
if (logging)
targetDescription = [[target description] copy];
weakTarget = target;
selector = aSelector;
return self;
}
-(void)dealloc
{
if (logging)
NSLog(#"-[%# dealloc]! (Target was %#)", self, targetDescription);
[targetDescription release];
[super dealloc];
}
-(void)passthroughFiredTimer:(NSTimer *)aTimer;
{
[weakTarget performSelector:selector withObject:aTimer];
}
-(void)dumbCallbackTimer:(NSTimer *)aTimer;
{
[weakTarget performSelector:selector];
}
#end
#implementation NSTimer (D12WeakTimerTarget)
+(NSTimer *)D12scheduledTimerWithTimeInterval:(NSTimeInterval)ti weakTarget:(id)target selector:(SEL)selector userInfo:(id)userInfo repeats:(BOOL)shouldRepeat logsDeallocation:(BOOL)shouldLogDealloc
{
SEL actualSelector = #selector(dumbCallbackTimer:);
if ( 2 != [[target methodSignatureForSelector:aSelector] numberOfArguments] )
actualSelector = #selector(passthroughFiredTimer:);
D12WeakTimerTarget *indirector = [[D12WeakTimerTarget alloc] initWithTarget:target selector:selector shouldLog:shouldLogDealloc];
NSTimer *theTimer = [NSTimer scheduledTimerWithTimeInterval:ti target:indirector selector:actualSelector userInfo:userInfo repeats:shouldRepeat];
[indirector release];
return theTimer;
}
#end
Original (for full disclosure):
You know my opinion from your other post:
There is little reason for an owning reference of a scheduled timer (and bbum seems to agree).
That said, your options 2, and 3 are essentially the same. (There is additional messaging involved in [self setWalkTimer:nil] over walkTimer = nil but I'm not sure if the compiler won't optimize that away and access the ivar directly, but well...)
I generally manage the invalidate inside of the accessor so that you never get surprised by a timer accessing you after you think you got rid of it:
#property(nonatomic, retain) NSTimer *walkTimer;
[self setWalkTimer: ti];
- (void)setWalkTimer:(NSTimer *)aTimer
{
if (aTimer != walkTimer_)
{
[aTimer retain];
[walkTimer invalidate];
[walkTimer release];
walkTimer = aTimer;
}
}
...
...
// Cancel method
[self setWalkTimer:nil];
...
...
// Make a new timer, automatically invalidating the old one
[self setWalkTimer:[... a new timer ...]]
...
...
// dealloc method
[walkTimer_ invalidate];
[walkTimer_ release];
[super dealloc];
What is the correct way to release a NSTimer in my dealloc method ? It was created with the following code ?
-(void)mainTimerLoop {
mainTimer = [NSTimer scheduledTimerWithTimeInterval:1/10
target:self
selector:#selector(gameLoop)
userInfo:nil
repeats:YES];
}
Thanks
The way you're doing it, you won't ever hit dealloc. A timer retains its target. In this case, that means the timer has retained you. It will not release you until it is invalidated. Since you created the timer, you must also invalidate it at some point prior to dealloc, because the timer's retain will prevent your object's being dealloced.
You have two options:
find another place to invalidate the timer (view goes offscreen, application is terminating, what have you)
set something else as the timer's target.
As an example of the latter:
#interface GameLoopTimerTarget : NSObject {
id owner; /* not retained! */
}
- (id)initWithOwner:(id)owner;
- (void)timerDidFire:(NSTimer *)t;
#end
#implementation GameLoopTimerTarget
- (id)initWithOwner:(id)owner_ {
self = [super init];
if (!self) return nil;
owner = owner_;
return self;
}
- (void)timerDidFire:(NSTimer *)t {
#pragma unused (t)
[owner performSelector:#selector(gameLoop)];
}
#end
/* In your main object… */
/* assume synthesized:
#property (retain, NS_NONATOMIC_IPHONE_ONLY) GameLoopTimer *mainTimerTarget; */
- (void)mainTimerLoop {
self.mainTimerTarget = [[[GameLoopTimerTarget alloc] initWithOwner:self] autorelease];
mainTimer = [NSTimer scheduledTimerWithTimeInterval:1.0/10.0 target:self.mainTimerTarget selector:#selector(timerDidFire:) userInfo:nil repeats:YES];
}
- (void)dealloc {
/* other stuff */
[timer invalidate], timer = nil;
[mainTimerTarget release], mainTimerTarget = nil;
/* more stuff */
[super dealloc];
}
Notice how the time interval is 1.0/10.0 - this could also be written 0.1, but it cannot be written 1/10, as that division will truncate to 0.0.
Also notice how this breaks the retain cycle:
Both you and your timer retain the timer target.
You hit dealloc at the normal time.
You then invalidate the timer and release the timer target.
The timer target is then deallocated.
A valid NSTimer is retained by the run loop, which, if it is repeating, will be forever or until you invalidate it. You shouldn't release it, since, in your example code, you did not explicitly retain it. If you invalidate it, it will no longer be retained by the run loop, and will be autoreleased.
This might be OK for a repeating timer, but is dangerous for a one-shot timer, since it might end being released before you ever access it to see if it's valid and/or try to invalidate it (which would lead to a bad-access app crash). Therefore if you plan on, in any way, looking at a timer variable after it's creation (including to check it, invalidate it and/or release it), it might be a good practice to explicitly retain it somewhere in your app, and then release it and set it to nil after it's invalid and you are done with it.
You can release it and set it to nil in one statement if you declare it as a retain property. Then you can write:
self.timer = nil;
you have a really good answer about NSTimer here How do I use NSTimer? there they talk about stoping a repeating NSTimer doing
[myTimer invalidate];
I think the best advice here is -
Do not retain the NSTimer instance, and do not release it.
As soon as it is scheduled on an NSRunloop (current runloop in the OP's example, an NSTimer is retained by the runloop until being invalidated, or until the runloop stops.
What you should be doing, is to invalidate your timer at the right time - and on the same thread where you created and scheduled it.
Keep in mind, also, that NSTimer retains its target, and won't let the target "die" before it dies itself. design your code so that you don't have a retain cycle that will prevent the releasing of both your object (holding the timer) and the timer (holding you object).
You don't need to release it because it will be autoreleased. Anything created by a convenience method (i.e. you don't call alloc yourself) is the responsibility of the called function to memory manage, which usually means that it will call autorelease on the object it creates before it returns it.
I would assign the timer to a property with the retain keyword though to make sure it doesn't get deallocated on you. Generally autoreleased objects are deallocated in the event loop if they don't have any retains.
I have a Class with a NSTimer *myTimer; variable. At some point I do:
myTimer = [NSTimer scheduledTimerWithTimeInterval:20 target:self selector:#selector(doStuff) userInfo:nil repeats: YES];
further, I have a method:
- (void)doStuff
{
if(myTimer)
{
//do stuff
}
}
and I stop my timer when the class is released through:
- (void)dealloc
{
if (myTimer) { //if myTimer==nil it already has been stopped in the same way
[myTimer invalidate];
myTimer = nil;
}
}
Now, the problem is that when I release the class the timer goes on and on and on firing the event anyway. Am I doing something wrong? It seems the dealloc method is never called, otherwise myTimer would be nil and even if the selector is fired it would not go into the if(myTimer)
This will never work, because timers retain their target, which means your dealloc method will never get invoked until after you've invalidated the timer.
For more info, see the NSTimer documentation and this blog post on "Dangerous Cocoa Calls"
Have you tried the handy debugger tools at your disposal? What happens if you set a breakpoint in your dealloc method? Also, you should post more context around your creation. Is it possible you're creating the timer more than once, thereby replacing the original (but not invalidating it) and leaving it out there to fire at will?