I have this code in one of my classes:
- (void) processArray
{
NSMutableArray* array = [self getArray];
. . .
[array release];
array = nil;
}
- (NSMutableArray*) getArray
{
//NO 1:
NSMutableArray* array = [[NSMutableArray alloc]init];
//NO 2:
NSMutableArray* array = [NSMutableArray array];
. . .
return array;
}
NO 1: I create an array and return it. In the processArray method I release it.
NO 2: I get an array by simply calling array. As I'm not owner of this, I don't need to release it in the processArray method.
Which is the best alternative, NO 1 or NO 2? Or is there a better solution for this?
The method should return an autoreleased array, NO 2 is the better choice. The reason for this is that when you call the method with
NSMutableArray* array = [self getArray];
you will expect as a matter of convention to not have to release the array. If you want to take ownership of it, you retain it, but otherwise you shouldn't have to care about it. In Cocoa, you only take ownership by explicitly sending alloc or retain or copy (or new). Since the processArray method doesn't do that, it shouldn't have to take care of releasing the array.
So you should use NO 2, and also you should remove these two lines:
[array release];
array = nil;
If the array and its contents use a lot of memory or its used lots of times, you'll want to release them straight away, so use option 1. According to the Objective-C guidelines, you'll want to prefix the word "new" to your subroutine name instead of "get" in that case.
If on the other hand, you want to reduce the number of lines of code that say simply [array release]; or similar then use option 2.
It is simply a balance between reducing lines of code, and reducing unnecessary temporary memory use.
Whilst the autorelease pool will help in reducing memory leaks and make your code smaller, sometimes you need to explicitly release everything as it goes out of use to keep the use of memory down.
HTH
EDIT
Ah - I stand corrected. Reading the iPhone version of the Memory Management Programming Guide for Cocoa I see that the iPhone guidelines are to use a prefix of "new..." so for example "newArray" in this case, if the caller is supposed to manually release and NOT a prefix of "create...". "Creating" can refer either to creation of manually released or of automatically released objects and so would be ambiguous. Text corrected above.
- (void) processArray
{
NSMutableArray* array = [[self getArray] retain];
//Now you are the owner of array, so you take care to release it
. . .
[array release];
array = nil;
}
- (NSMutableArray*) getArray
{
//create a new array
//temporarily the method owns the array
NSMutableArray* array = [[NSMutableArray alloc]init];
//fill in here with elements or what you want
..........
[array autorelease];
//autorelease here says "I don't own the result
//if anyone cares about it, he should retain it himself
return array;
}
So in short when you create new objects you should autorelease them before returning.
Because if the calling method wants to use the result, the calling method should take care
of retaining and releasing the result.
It's always good to run the Klang static analyzer for this issues, when you are not really sure in your retaining/releasing code : http://clang-analyzer.llvm.org/
Related
I'm new to cocoa / objective-c and i'm struggeling with the releases of my objects. I have the following code:
gastroCategoryList = [[NSMutableArray alloc] init];
for (NSDictionary *gastrocategory in gastrocategories) {
NSString *oid = [gastrocategory objectForKey:#"id"];
GastroCategory *gc = [[GastroCategory alloc] initWithId:[oid intValue] name:[gastrocategory objectForKey:#"name"]];
[gastroCategoryList addObject:gc];
}
The analyzer shows me that the "gastrocategory" defined in the for is a potential memory leak. But i'm not sure if i can release this at the end of the for loop?
Also at the following code:
- (NSArray *)eventsForStage:(int)stageId {
NSMutableArray *result = [[NSMutableArray alloc] init];
for (Event *e in eventList) {
if ([e stageId] == stageId) {
[result addObject:e];
}
}
return result;
}
The Analyzer tells me that my "result" is a potential leak. But where should I release this?
Is there also a simple rule to memorize when i should use assign, copy, retain etc. at the #property ?
Another problem:
- (IBAction)showHungryView:(id)sender {
GastroCategoriesView *gastroCategoriesView = [[GastroCategoriesView alloc] initWithNibName:#"GastroCategoriesView" bundle:nil];
[gastroCategoriesView setDataManager:dataManager];
UIView *currentView = [self view];
UIView *window = [currentView superview];
UIView *gastroView = [gastroCategoriesView view];
[window addSubview:gastroView];
CGRect pageFrame = currentView.frame;
CGFloat pageWidth = pageFrame.size.width;
gastroView.frame = CGRectOffset(pageFrame,pageWidth,0);
[UIView beginAnimations:nil context:NULL];
currentView.frame = CGRectOffset(pageFrame,-pageWidth,0);
gastroView.frame = pageFrame;
[UIView commitAnimations];
//[gastroCategoriesView release];
}
I don't get it, the "gastroCategoriesView" is a potential leak. I tried to release it at the end or with autorelease but neither works fine. Everytime I call the method my app is terminating. Thank you very much again!
In your loop, release each gc after adding it to the list since you won't need it in your loop scope anymore:
gastroCategoryList = [[NSMutableArray alloc] init];
for (NSDictionary *gastrocategory in gastrocategories) {
NSString *oid = [gastrocategory objectForKey:#"id"];
GastroCategory *gc = [[GastroCategory alloc] initWithId:[oid intValue] name:[gastrocategory objectForKey:#"name"]];
[gastroCategoryList addObject:gc];
[gc release];
}
In your method, declare result to be autoreleased to absolve ownership of it from your method:
NSMutableArray *result = [[[NSMutableArray alloc] init] autorelease];
// An alternative to the above, produces an empty autoreleased array
NSMutableArray *result = [NSMutableArray array];
EDIT: in your third issue, you can't release your view controller because its view is being used by the window. Setting it to autorelease also causes the same fate, only delayed.
You'll have to retain your GastroCategoriesView controller somewhere, e.g. in an instance variable of your app delegate.
BoltClock's answer is spot-on as to the first part of your question. I'll try to tackle the rest.
Assign is for simple, non-object types such as int, double, or struct. It generates a setter that does a plain old assignment, as in "foo = newFoo". Copy & retain will, as their names imply, either make a copy of the new value ("foo = [newFoo copy]") or retain it ("foo = [newFoo retain]"). In both cases, the setter will release the old value as appropriate.
So the question is, when to copy and when to retain. The answer is... it depends. How does your class use the new value? Will your class break if some other code modifies the incoming object? Say, for example, you have an NSString* property imaginatively named "theString." Other code can assign an NSMutableString instance to theString - that's legal, because it's an NSString subclass. But that other code might also keep its own reference to the mutable string object, and change its value - is your code prepared to deal with that possibility? If not, it should make its own copy, which the other code can't change.
On the other hand, if your own code makes no assumptions about whether theString might have been changed, and works just as well whether or not it was, then you'd save memory by retaining the incoming object instead of unnecessarily making a copy of it.
Basically, the rule, which is unfortunately not so simple sometimes, is to think carefully about whether your own code needs its own private copy, or can correctly deal with a shared object whose value might be changed by other code.
The reason you can release gc after it is added to the gastroCategoryList is that when an object is added to an array, the array retains that object. So, even though you release your gc, it will still be around; retained by the gastroCategoryList.
When you are returning a newly created object from a method, you need to call autorelease. This will cause the object to be released only after the runtime leaves the scope of the calling method, thereby giving the calling method a chance to do something with the returned value.
Note that if your method starts with the word copy or new, then you should not autorelease your object; you should leave it for the calling method to release.
As for copy vs retain vs assign... as a general rule, copy objects that have a mutable version, such as NSArray, NSSet, NSDictionary, and NSString. This will ensure that the object you have a pointer to is not mutable when you don't want it to be.
Otherwise, use retain whenever you want your class to be ensured that an object is still in memory. This will apply to almost every object except for objects that are considered parents of your object, in which case you would use assign. (See the section on retain cycles here).
Also note that you have to use assign for non-object types such as int.
Read through the Memory Management Programming Guide a bit; it's quite helpful.
I have been very confused on how to handle the releasing of an NSMutableArray when I need to return it from a method. I am not even sure if I am suppose to release it or not.
I have the code:
-(NSArray *)methodToCall {
NSMutableArray *mutArray = [[NSMutableArray alloc] initWithCapacity:10];
//Fill the array with stuff here
[mutArray release]; //Am I suppose to have this to keep the memory down?
return mutArray;
}
My question is whether or not I am suppose to have the [mutArray release]; in the code or not. Could someone explain this? I am at a loss and I want to keep my code as clean and leak free as possible.
The caller of your method is going to expect a valid NSArray, but which it doesn't own.
If you release it the way you're doing in your question, you're returning an invalid object (because you've both allocated and released it.) So that's not what you want.
You should "autorelease" the array before returning it. Then you'll be allocating it, but relinquishing ownership (without forcibly releasing it) before returning it. It will be valid until the end of the current event loop when the autorelease pool gets cleaned up, and your caller can retain it or let it go as appropriate:
-(NSArray *)methodToCall {
NSMutableArray *mutArray = [[NSMutableArray alloc] initWithCapacity:10];
//Fill the array with stuff here
return [mutArray autorelease];
}
Your other alternative, since you don't need to keep the reference around either, is to use one of the "autoreleasing" convenience methods to create it in the first place:
-(NSArray *)methodToCall {
// We're not doing an explicit alloc/init here, so...
NSMutableArray *mutArray = [NSMutableArray arrayWithCapacity:10];
// ...no autorelease necessary.
return mutArray;
}
Short answer - No.
As it is now, you are allocating an array and then releasing (freeing) it before the return. So the when you try accessing the return object from your calling method you're going to get a crash.
My suggestion would be to use autorelease or to have the calling method or class ivar be responsible for this array object if it is used often.
An example of the autorelease:
NSMutableArray *mutArray = [[[NSMutableArray alloc] initWithCapacity:10] autorelease];
I also suggest reading the Memory Management from the Developer Documents.
if you go for explicit object allocation by calling alloc and init you are owner of your object, so you are responsible for its object retain value else you do it by implicit you don't need to care about it. it will take care of itself.
-(NSMutableArray *)getArray
{
NSMutableArray *array=[[NSMutableArray alloc] initWithCapacity:10];
//your code
return [array autorelease];
}
in the above code we are the owner of the object so we need to handle its retain count by passing autorelease the autoreleasepool will take care of it.
-(NSMutableArray *)getArray
{
NSMutableArray *array=[NSMutableArray allocWithCapacity:10];
//your code
return array;
}
in the above code we didn't alloc any object we just call class method to define the size of the array.
if you want more details go for the object ownership in Memory management guide from apple library
In general, instead of using alloc/init to create a temporary array, consider using a convenience creation method (+arrayWithCapacity: in this case):
- (NSArray *)methodToCall
{
NSMutableArray *mutableArray = [NSMutableArray arrayWithCapacity:10];
// Fill the array with stuff here
return mutableArray;
}
Convenience creation methods such as +arrayWithCapacity: return an object that the caller is not responsible for, relieving the calling code of the burden of managing memory. Since the calling code is simply returning the reference rather than storing it, that'll simplify things.
I am trying to append objects to my data source and then reload the table. Is this the correct way of approaching it?
self.items is my datasource
//Copy current items
self.itemsCopy = [self.items mutableCopy];//[[NSMutableArray alloc] initWithArray:self.items copyItems:NO];
NSLog(#"Copy Size before append: %d",[itemsCopy count]);
//Get new items
int lastMsgID = [self getLastMessageID];
[self.coreData getMoreMessages:self.title lastID:lastMsgID]; //This will update self.items with 30 objects
//Append new items
[itemsCopy addObjectsFromArray:self.items];
//Empty items and copy itemsCopy
[self.items removeAllObjects];
self.items = [self.itemsCopy mutableCopy];
NSLog(#"Actual Size after append: %d",[self.items count]);
//Reload data
[tableView reloadData];
try
[tableView reloadData];
in the viewWillAppear method....
Yes, looks about right to me.
Depending on how your properties are declared, you may have some memory issues here.
If your properties specify retain as an attribute, then setting assigning a copy of an array to them will result in a retain count of 2 on the object rather than the intended 1.
Eg.
self.myProperty = [myArray mutableCopy];
mutableCopy returns a newly allocated object with a retain count of one. The caller owns this object and is responsible for releasing it. Then you assign it to your property which retains it, increasing the retain count to 2. The object will be leaked when the owner is deallocated because the property will only be released once.
The solution would be to release the mutable copy after assigning and retaining it to to the property like this.
self.myProperty = [myArray mutableCopy];
[self.myProperty release];
Alternatively you could auto release the result of mutableCopy. However there seems to be some controversy surrounding the usage of autorelease on the iPhone. I personally don't see a problem with it unless it's used in a tight loop that iterates many many times in a short period of time.
self.myProperty =[[myArray mutableCopy] autorelease];
If you're using CoreData, then ideally you should be using a NSFetchedResultsController to manage the data for the tableview. (You did ask for the "correct" way :-)
You'll end up with a more memory efficient app.
See the iPhoneCoreDataRecipes Sample for a good example
I try to understand the memory management in ObjectiveC and still some things are a misery for me.
I've got an instance variable:
NSMutableArray *postResultsArray;
when a button is clicked in the UI I create new array:
self.postResultsArray = [NSMutableArray array];
then I add some objects to the array and when the whole operation is done I would like to release the array:
[self.postResultsArray release];
(I assume that all the objects stoed in the array will be released along with the array).
The problem appears when I click the button again and in the code I want to create the array again with:
self.postResultsArray = [NSMutableArray array];
I get:
[CFArray release]: message sent to deallocated instance 0x3d9e390
Can't I initialize the same instance variable for the second time? or maybe I should not release it? and if so, why?
Thanks!
Don't to this:
[self.postResultsArray release];
When you do this, the ivar is still assigned to the old array's memory address. If you want to release the array, there are two safe ways to do it:
[postResultsArray release];
postResultsArray = nil;
Or
self.postResultsArray = nil;
What's happening is that the code for setting the postResultsArray looks like this (paraphrase, not exact):
-(void)setPostResultsArray:(NSMutableArray *)newArray {
[array retain];
[postResultsArray release]; // this is what's causing the the deallocation message in your log
postResultsArray = array;
}
You shouldn't manually release an object that is accessed through a property. Instead of
[self.postResultsArray release];
do
self.postResultsArray = nil;
and all will be fine. The setter will release the old array and then assign nil to the pointer. What is happening now is that the setter is trying to release the old array when you assign a new array to the property, but you have already released that array, hence the error.
Don't do [self.postResultsArray release];, do self.postResultsArray = nil, this has to do with the implementation of properties. They will automatically release whatever is currently stored and retain the new value.
You only need to call release if you create an object with one of the "init" messages or explicitely call retain yourself. Since you do neither you don't need to call release as the code that created the array will be handling the release - most probably by using autorelease.
I'm following a book on iPhone development and there is a particular pattern I keep seeing in the example code that doesn't make much sense to me. Whenever a property is set they first assign a pointer to the new value for the property, then set the property to the pointer, then release the pointer. Example:
Interface:
#interface DoubleComponentPickerViewController : UIViewController {
NSArray *breadTypes;
}
#property(nonatomic, retain) NSArray *breadTypes;
#end
Class method:
- (void)viewDidLoad {
NSArray *breadArray = [[NSArray alloc] initWithObjects:#"White", #"Whole Wheat", #"Rye", #"Sourdough", #"Seven Grain", nil];
self.breadTypes = breadArray;
[breadArray release];
}
Is there any reason to do this instead of just doing the following?
- (void)viewDidLoad {
self.breadTypes = [[NSArray alloc] initWithObjects:#"White", #"Whole Wheat", #"Rye", #"Sourdough", #"Seven Grain", nil];
}
Thanks for the light that will no doubt be shed :)
Let me try and explain it in a different way.
A method that has alloc, copy or new in its name will allocate memory for an object, and gives ownership of that object to the caller, and it is the caller's responsibility to release that memory.
In your viewDidLoad method, you call a method that gives you ownership of an object. It is your method's responsibility to release it. However, before you do that, you want to do something with it - after all, that's why you alloc'ed it, not to just release it, but to do something useful with it.
Regardless of what it is that you want to do with it, you have to release it (or autorelease it*). In this case your use of the object is to pass it to self.breadTypes. self.breadTypes may not look like a method, but it is (it is a setter). You pass it breadArray. It does what it needs to with it. It might retain it for use later, or it might copy some info out of it, or make a copy of the entire thing. Your viewDidLoad doesn't really care. It assumes that self.breadTypes does what it needs to and when it returns, it doesn't care what you do with breadArray.
And what you do with it, is what you have to do with anything that you own - release (or autorelease* it).
That's why you have to use the temp variable, breadArray. You can't quite release the results from alloc on the same line, since the object would get released before self.breadTypes can have at it:
self.breadTypes = [[[NSArray alloc] initWithObjects:#"White", ..., nil] release];
Thus you are forced to assign to a temp variable, pass it to self.breadTypes, and then release the object that is saved in breadArray.
Now, you could try to do it this way so you don't use a temp variable:
- (void)viewDidLoad {
self.breadTypes = [[NSArray alloc] initWithObjects:#"White", #..., nil];
[self.breadTypes release];
}
but that is not very efficient since you are calling yet another method (self.breadTypes as a getter) that you didn't really need to if you have just stored the value in a temp variable.
*Now, as a responder said, you could use autorelease for an alternative version:
- (void)viewDidLoad {
self.breadTypes = [[[NSArray alloc] initWithObjects:#"White", ..., nil]
autorelease];
}
Apple urges us to think twice about whether we want to use autorelease vs. release. Autorelease may not be the best choice for every situation. I personally like to clean up after myself as soon as I possibly can, and not use autorelease needlessly. Autoreleased objects get released at the end of the execution of the run loop, for example soon after viewDidLoad returns. You should read up a bit more about autorelease (and memory management on the iPhone which is slightly different than MacOS X Cocoa), as I am oversimplifying it all.
BTW: If you retain an object, you are assuming ownership of it, and you will have the same responsibility again, to release it after you are done with it.
Yes. Those methods are alloc'ing the variables so they must be released. The fact that the property has a retain attribute means that when you say #synthesize breadTypes; the compiler is actually generating a setBreadTypes that properly releases the current breadType member and retains the new one. Thus your function must not retain the variable it alloc'ed.
You could, however write:
- (void)viewDidLoad {
self.breadTypes = [[[NSArray alloc] initWithObjects:#"White",
#"Whole Wheat", #"Rye", #"Sourdough",
#"Seven Grain", nil]
autorelease];
}
You'll want to brush up on Cocoa Memory Management