I'm writing an app in which the user takes a photo of them self, and then goes through a series of views to adjust the image using a navigation controller. This works perfectly fine if the user takes the photo with the front camera (set as default on devices that support it), but when I repeat the process I get about half way through and it crashes after throwing a memory warning.
After profiling in Instruments I see that my apps memory footprint holds at about 20-25 MB when using the lower resolution front camera image, but when using the back camera every view change adds another 33 MB or so until it crashes at about 350 MB (on a 4S)
Below is the code I'm using to handle saving the photo to the documents directory, and then reading that file location to set the image to a UIImageView. The "read" portion of this code is recycled through several view controllers (viewDidLoad) to set the image I saved as the background image in each view as I go.
I have removed all my image modification code to strip this down to the bear minimum attempting to isolate the problem, and I can't seem to find it. As it stands right now, all the app does is take a photo in the first view and then use that photo as the background image for about 10 more views, allocating as the user navigates through the view stack.
Now obviously the higher resolution photos would use more memory, but what I don't understand is that why the low resolution photos don't seem to be using more and more memory as I go, whereas the high resolution photos continuously use more and more until a crash.
How I am saving and reading the image:
- (void) imagePickerController:(UIImagePickerController *)picker didFinishPickingMediaWithInfo:(NSDictionary *)info
{
UIImage *image = [info objectForKey:#"UIImagePickerControllerOriginalImage"];
jpgData = UIImageJPEGRepresentation(image, 1);
NSArray *paths = NSSearchPathForDirectoriesInDomains(NSDocumentDirectory, NSUserDomainMask, YES);
NSString *documentsPath = [paths objectAtIndex:0];
filePath = [documentsPath stringByAppendingPathComponent:#"image.jpeg"];
[jpgData writeToFile:filePath atomically:YES];
[self dismissModalViewControllerAnimated:YES];
[disableNextButton setEnabled:YES];
jpgData = [NSData dataWithContentsOfFile:filePath];
UIImage *image2 = [UIImage imageWithData:jpgData];
[imageView setImage:image2];
}
Now I know that I could try scaling the image before I save it, which I plan on looking into next, but I don't see why this doesn't work as is. Maybe I was falsely under the impression that ARC automatically deallocated views and their subviews when they leave the top of the stack.
Can anyone shed some light on why I'm stock piling my devices memory? (Hopefully something simple I'm completely overlooking) Did I somehow manage to throw ARC out the window?
EDIT: How I call for the image in my other views
- (void)loadBackground
{
NSArray *paths = NSSearchPathForDirectoriesInDomains(NSDocumentDirectory, NSUserDomainMask, YES);
NSString *documentsPath = [paths objectAtIndex:0];
NSString *filePath = [documentsPath stringByAppendingPathComponent:#"image.jpeg"];
UIImage *image = [UIImage imageWithContentsOfFile:filePath];
[backgroundImageView setImage:image];
}
How navigation between my view controllers is established:
EDIT 2:
What my basic declarations look like:
#import <UIKit/UIKit.h>
#import <AVFoundation/AVFoundation.h>
#interface PhotoPickerViewController : UIViewController <UIImagePickerControllerDelegate, UINavigationControllerDelegate>
{
IBOutlet UIImageView *imageView;
NSData *jpgData;
NSString *filePath;
UIImagePickerController *imagePicker;
IBOutlet UIBarButtonItem *disableNextButton;
}
#end
If relevant, how I call up my image picker:
- (void)callCameraPicker
{
if ([UIImagePickerController isSourceTypeAvailable:UIImagePickerControllerSourceTypeCamera] == YES)
{
NSLog(#"Camera is available and ready");
imagePicker.sourceType = UIImagePickerControllerSourceTypeCamera;
imagePicker.delegate = self;
imagePicker.allowsEditing = NO;
imagePicker.cameraCaptureMode = UIImagePickerControllerCameraCaptureModePhoto;
NSArray *devices = [AVCaptureDevice devicesWithMediaType:AVMediaTypeVideo]; for (AVCaptureDevice *device in devices)
{
if([[UIScreen mainScreen] respondsToSelector:#selector(scale)] && [[UIScreen mainScreen] scale] == 2.0)
{
imagePicker.cameraDevice = UIImagePickerControllerCameraDeviceFront;
}
}
imagePicker.modalTransitionStyle = UIModalTransitionStyleCoverVertical;
[self presentModalViewController:imagePicker animated:YES];
}
else
{
NSLog(#"Camera is not available");
UIAlertView *cameraAlert = [[UIAlertView alloc] initWithTitle:#"Error"
message:#"Your device doesn't seem to have a camera!"
delegate:self cancelButtonTitle:#"Dismiss"
otherButtonTitles:nil];
[cameraAlert show];
}
}
EDIT 3: I logged viewDidUnload, and it was in fact not being called so I'm now calling loadBackground in viewWillAppear and making my backgroundImageView nil in viewDidDisappear. I expected this to help but it made no difference.
- (void)viewWillAppear:(BOOL)animated
{
[self loadBackground];
}
- (void)viewDidDisappear:(BOOL)animated
{
NSLog(#"ViewDidDisappear");
backgroundImageView = nil;
}
The relationship between UIImage and UIImageView is not necessarily intuitive for everyone.
UIImage is a high level representation of your image data - alone, it does nothing in terms of displaying the data.
UIImageView works with UIImage to allow you to display an image.
There is no reason why multiple instances of UIImageView cannot display the same UIImage. This is nice and efficient, because there is only one in-memory representation of the image data, being shared by multiple views.
What you seem to be doing is creating a new UIImage for each one of your views, by loading it from disk. So this is a poor general design in two respects: instantiating what is effectively the same UIImage over and over again, and re-loading the same image data from disk repeatedly.
Your memory problem is really a separate issue, where you are not properly releasing the image data you keep loading into UIImage objects and UIImageViews.
In theory, you should be able to take the very first UIImage you're getting from UIImagePickerController and simply pass that reference around to your views, without reloading from disk.
If you need to be saving and reloading from disk because of higher level functional requirements (e.g. because the image is being changed by the user and you want to keep saving it), you'll need to be sure you are fully tearing down the previous UIView, by removing it from the it's view hierarchy. It is helpful to setup a breakpoint in the dealloc method for the view to confirm it is being removed and dealloced, and make sure you set any iVar references to sub-views (it appears your backgroundImageView is an iVar) are set to nil. If you are not properly tearing down that backgroundImageView, it is continuing to hold a reference to the UIImage you set to it's image property.
There are a couple of things that are curious about the code you posted:
None of your view-callback implementations call super. That’s bad! Make extra sure that you are calling super in viewDidUnload and (if you implemented it) didReceiveMemoryWarning.
Make sure you implement didReceiveMemoryWarning in a meaningful way!
You really should not be re-creating that image over and over again! I assume you are not editing the actual image because you use JPEG compression on it which — even at 100% quality — will deteriorate your image with every save…
Check your implementation of viewDidUnload make sure to set every of your IBOutlets to nil.
ARC is not Pixie Dust™! It just saves you a bit of typing, it does not free you from designing and maintaining your object graphs!
From your question, I see at the very least these graphs that refer to your image:
image 1 <- image-view 1 <- view-controller 1 <- navigation-controller <- key window <- application
image 1 <- image-view 1 <- view 1 <- view-controller 1 <- navigation-controller <- key window <- application
This is repeated for every view-controller with an index shift on the view-controller, view, image view and image. While you have to have separate views, image-views for your view-controllers, I cannot think of a reason why you would want several copies of the same image.
So the first axe on your memory consumption clearly is to no longer create all those copies of the same image data — I’d estimate that this will get you half of the low-hanging memory savings.
The next thing is that ARC can only free the memory consumed by your objects if it is no longer referenced.
Memory wise, views are not exactly lightweight objects and when you build up a deep navigation stack you end up with gobs of them.
So you need to axe any unneeded strong references to those views, as well.
The level, at which this has to happen is the view-controller. The latest time at which this should happen is in the view-controller’s implementation of viewDidUnload.
Why the view-controller?
From what you described, the image itself is only referenced by the UIImageView — this is a bad choice, IMHO, but I digress…
UIViewController is designed to “know”, when its view is needed and when it’s safe to dispose of it — that’s why it implements didReceiveMemoryWarning and viewDidUnload:
If the memory pressure gets to high and the view-controller’s view is not “on screen” the root implementation of didReceiveMemoryWorning will let go of its view and call viewDidUnload upon itself, afterwards.
This is why you must call through to super in your implementations of both of those methods.
In addition, this is why if you have strong IBOutlets that refer to subviews of the view-controller’s view, you must nil them in viewDidUnload or the system cannot reclaim the memory they occupy.
At its heart UIViewController is a big-ass finite state-machine. All of those “something-will/did-whatever” callbacks are used to transition between those states and most of the default implementations do some very important book-keeping to keep all that state in order.
If you are not invoking them in your overrides, you˚ll end up in inconsistent states and bad things — like this out of memory crasher — happen.
Just create separate folder and save your Capture images in it. After your successful operation clear that folder data(or)folder.using the nsfilemager.
Related
As weird as it sounds...
App loaded through test flight or iTunes does not show image on the screen, but it does when it is loaded through Xcode. I believe the UIImageView is there, and the UIImage is loaded.
App loaded through the Xcode works.
If the device is connected to the Xcode and debugging, device connected with USB cable. App works despite the load method...
Core data seems to works in all cases.
Any idea, somebody has seen something similar. I am quite lost.
Also if you have any tip on how to debug this on the device I will appreciate it.
The few code I think it is involve.
I am loading the UIImage from disk using
UIImage * previewImage = [UIImage imageWithContentsOfFile:path];
The image is located in Documents Directory fileName is the input
NSURL *documentsFolderURL = [[[NSFileManager defaultManager] URLsForDirectory:NSDocumentDirectory inDomains:NSUserDomainMask] lastObject];
NSString *documentsFolderPath = [documentsFolderURL path];
if (![_documentsDirectory isEqualToString:documentsFolderPath]) [self generateLocaFolderPath];
NSString *returnedPath = [_documentsDirectory stringByAppendingPathComponent:fileName];
return returnedPath;
and setting the image in the imageView with
previewView = [[UIImageView alloc] initWithFrame:CGRectMake(0, 0, frameImageThumbnail.size.width, frameImageThumbnail.size.height)];
[previewView setImage:previewImage];
Everything else seems to work well...
My first question!!! Thanks for reading!
_ edit _
I had the strong believe now that the problem is not in the UIImage, I have use the same methods in other parts of the app and it is working fine.
But now I believe the problem is in the animation I perform to the UIView that contains the UIImageView.
[_view setFrame:frameImageThumbnail];
[_view setClipsToBounds:YES];
[_view setAutoresizesSubviews:YES];
[_image setAutoresizingMask:UIViewAutoresizingFlexibleWidth | UIViewAutoresizingFlexibleHeight];
[_view addSubview:_image];
[UIView animateWithDuration:0.4
delay:0.1
options:UIViewAnimationOptionCurveEaseInOut
animations:^(void){
CGRect destination1Frame = _view.frame;
//... some complicated calculations with the rect ...
[_view setFrame:destination1Frame];
[UIView transitionFromView:_customView toView:_view duration:0 options: UIViewAnimationOptionShowHideTransitionViews completion:^(BOOL finisehdCompletion){
//stop showing video in _customView
}];
}];
I am using NSAutolayout in the storyboard. Though my application only allows one landscape orientation. No idea if all this meshing up with the frames it is making.
dizzy to the poor auto layout guy...
Make sure you have / your image is set in Copy Bundle Resources. You can check if your image path is valid also. But I think the image is not set to copy at Build Phases.
I had the same problem.. Haven't find a correct solution.
But I deleted the image(check if it's deleted in ur Bundle resources), then put the image again, with ANOTHER name.
It worked for me. It's a temporary solution, idk if it's an apple bug or what.
Ok After more than 20 ipa files build and installed in my devices. In all this process to get rid of the problem I removed the _view from the storyboard, and initiallize the view programmatically. I remember that _view was originally set to (weak, nonatomic) property because was made with the storyboard, this super nice ctrl+.
That was I change I made, weak to Strong, and works now.
I don't know why it didn't work only when the program was install through iTunes with the ipa file or through the test flight app. That is still a mystery to me.
On one hand, I'm just curious.
On the other, I have to count somehow the amount of images created (to trigger cleanup at some quantity). Unfortunately memory warnings notify me too late, the app gets terminated before I have any chance to react, so thatswhy I want to work this issue around.
I could scatter around my project with a global counter, that increments before I make any UIImage operation, but obviously this is exactly what I want to avoid, that would result in an unmaintainable design.
So I was thinking about subclass UIImage initializers (some kind of MYImage class), than use MYImage troughout the project, but UIImageView's would use UIImages anyway (maybe I can set UIImageView's setImage method, but I bet UIKit uses actual _image instance variable at many cases).
The best would be to count pixels amount from UIImage sizes, so I could flush image memory above a given pixelcount.
Any notification? Something trough categories? Or some kind of NSObject KVO observing (with literal checkings on class/method names)? Do you have ideas/experiences?
There a a lot of ways you could accomplish this. One way would be to override one (or more) of the image creation methods in a subclass, get the size of the image created, and add that to a user default to keep track of the cumulative sizes of all the images. For example, I overrode imageNamed like this:
#implementation RDImage
+(UIImage *)imageNamed:(NSString *) name {
UIImage *anImage = [super imageNamed:name];
double oldSize = [[NSUserDefaults standardUserDefaults] doubleForKey:#"SizeAsWxH"];
[[NSUserDefaults standardUserDefaults] setDouble:(anImage.size.width * anImage.size.height) + oldSize forKey:#"SizeAsWxH"];
if ([[NSUserDefaults standardUserDefaults] doubleForKey:#"SizeAsWxH"] > 6912) {
NSLog(#"Too big -- do something");
}
return anImage;
}
In the applicationDidFinishLaunchingWithOptions: method, I set the value of that key to zero.
In the view controller when I create an image, I just use my class to make a new one:
UIImage *anImage = [RDImage imageNamed:....
Another way would be to override the setImage: method of UIImageView, and do the same kind of thing inside that method.
I wanted to create a gallery. It loads different images based on the category, that a user selects. I used to populate images in UIImageViews.
My when selecting different categories is that it does not clear the previously selected images. This is my code for populating images.
-(void)refresh{
categoryLabel.text = treatment.treatmentName;
NSMutableArray *galleryImages =[NSMutableArray alloc] ;
galleryImages = [[PatientImage alloc]find:treatment.treatmentId];
int imgCount = [galleryImages count];
for(int i=0;i<imgCount;i++){
PatientImage *apatientImage = [galleryImages objectAtIndex:i];
UIImage *img1 = [UIImage imageNamed:apatientImage.imageBefore];
UIImageView *myImageView = [[UIImageView alloc] initWithImage:img1];
myImageView.contentMode = UIViewContentModeTopRight;
myImageView.frame = CGRectMake(120+i*240,120.0,100.0, 100.0);
[self.view addSubview:myImageView];
UIImage *img2 = [UIImage imageNamed:apatientImage.imageAfter];
UIImageView *myImageView2 = [[UIImageView alloc] initWithImage:img2];
myImageView2.contentMode = UIViewContentModeTopRight;
myImageView2.frame = CGRectMake(120+i*240+300,120.0,100.0, 100.0);
[self.view addSubview:myImageView2];
}
}
First things first, You have some memory leaks there. You are allocating UIImageViews but are not releasing them anywhere, after you have added them to your view. I don't know if that applies to ARC, though. Same applies to your Mutable array, but I suppose you are releasing it after the 'for' loop somewhere, since it seems you posted code after omitting some of it.
As far as your actual question is concerned, I wouldn't do this this way. I would make the mutable array an object variable, and then fill it with my image views. When calling refresh again, I would first call -removeFromSuperview on each image view, then empty the array, then repopulate it and add the new subviews to my view. That is the simple way.
I don't know if you are using ARC, but you should be careful about memory management when using dynamically loaded views. Each time you add a view to another one, you increase its retain counter. You must then call release to remove ownership, and let the iOS runtime handle the rest.
Also note that operations such as this using views are expensive in terms of memory. So, another way of repopulating the gallery view is to just change the image an imageView holds. That will save you some memory, and time. In case the view doesn't have a constant number of images to be displayed, you can refine your algorithm to change the images on the already created image views, and then add more image views if necessary or delete the remaining ones, if any.
I hope I helped.
try at the start of refresh call
[[self subviews] makeObjectsPerformSelector: #selector(removeFromSuperview)];
or
for (id imageView in self.subviews){
if([imageView isKindOfClass:[UIImageView class]]) [imageView removeFromSuperview];
}
call [tableview reloadData] if You are using tableview to show your gallery images
or call view's
[self.view setNeedsDisplay] method for refreshing the view.
I’ve read all the iOS memory allocation/deallocation basics, but can’t find anything on the following:
I’ve created a small app, that pretty much lists a grid of UIButtons and clicking on any of those, it adds a UIScrollView to the current controller view (and adds a bunch of UIView and UIWebView’s inside etc).
On adding the UIScrollVIew, I also add a UIButton, that takes me back to the "home grid" and then removes the UIScrollView from the superview.
I release all the things I’m retaining/allocating etc and when I check the app with Instruments, it doesn’t show any memory leak.
But every time I tap on any of the UIButton objects, I allocate more memory (according to Instruments) and it keeps growing – and "reopening" the same kind of UIScrollView from a button always adds more memory allocations.
If I simulate a memory warning in the simulator, it deallocated a bit of the memory and I can then keep growing it again.
So here’s my question: should I bother with trying to deallocate this somehow manually? And if so, where should I actually do this? I’m quite new to Obj-C, so I think I have most of the basics covered, but more advanced topics still require some help.
Creating the grid:
UIScrollView *grid = [[UIScrollView alloc] initWithFrame:CGRectMake(0, 0, 1024, 768)];
[[self view] addSubview:grid];
[grid release];
Adding buttons (in a for loop):
UIButton *slotItem = [[UIButton alloc] initWithFrame:CGRectMake(((float) slotWidth * j), ((float) slotHeight * i), (float) slotWidth, (float)slotHeight)];
[grid addSubview:slotItem];
UIWebView *buttonWebThumb = [[UIWebView alloc] initWithFrame:CGRectMake(0, 0, slotWidth, slotHeight)];
[buttonWebThumb setBackgroundColor:[UIColor clearColor]];
[buttonWebThumb setOpaque:NO];
[buttonWebThumb loadHTMLString:[NSString stringWithFormat:#"%# %# %#", htmlTop, [[row objectAtIndex:j] objectAtIndex:1], htmlBottom] baseURL:[NSURL fileURLWithPath:[[NSBundle mainBundle] bundlePath]]];
buttonWebThumb.scalesPageToFit = YES;
[slotItem addSubview:buttonWebThumb];
buttonWebThumb.userInteractionEnabled = NO;
buttonWebThumb.exclusiveTouch = NO;
[buttonWebThumb release];
[slotItem addTarget:self action:#selector(showPages:) forControlEvents:UIControlEventTouchUpInside];
[slotItem release];
showPages method then creates another UIScrollView and adds 1-10 separate UIWebView’s in there and adds a “close” button to the new UIScrollView.
You're probably seeing this because you aren't removing all elements from the view before leaving it.
For example, if you exit the view with a method like this, do this:
- (id) viewExitAction: (id) sender {
id elem;
for(elem in buttons) {
[elem removeSelfFromView];
}
// etc
}
This will automatically decrease the retain counts of these objects down to zero. (Provided you added them to the view and then released them, as you're supposed to.)
Otherwise, the viewDidUnload and dealloc methods never get called.
Dealloc is invoked indirectly through the release method, and you should never call it directly.NSObject Manual on dealloc
I'm trying to use the select and copy feature of the IKImageView. If all you want to do is have an app with an image, select a portion and copy it to the clipboard, it's easy. You set the copy menu pick to the first responder's copy:(id) method and magically everything works.
However, if you want something more complicated, like you want to copy as part of some other operation, I can't seem to find the method to do this.
IKImageView doesn't seem to have a copy method, it doesn't seem to have a method that will even tell you the selected rectangle!
I have gone through Hillegass' book, so I understand how the clipboard works, just not how to get the portion of the image out of the view...
Now, I'm starting to think that I made a mistake in basing my project on IKImageView, but it's what Preview is built on (or so I've read), so I figured it had to be stable... and anyway, now it's too late, I'm too deep in this to start over...
So, other than not using IKImageView, any suggestions on how to copy the select region to the clipboard manually?
EDIT actually, I have found the copy(id) method, but when I call it, I get
<Error>: CGBitmapContextCreate: unsupported parameter combination: 8 integer bits/component; 16 bits/pixel; 1-component color space; kCGImageAlphaPremultipliedLast; 2624 bytes/row.
Which obviously doesn't happen when I do a normal copy through the first-responder... I understand the error message, but I'm not sure where it's getting those parameters from...
Is there any way to trace through this and see how this is happening? A debugger won't help for obvious reasons, as well as the fact that I'm doing this in Mozilla, so a debugger isn't an option anyway...
EDIT 2 It occurs to me that the copy:(id) method I found may be copying the VIEW rather than copying a chunk of the image to the clipboard, which is what I need.
The reason I thought it was the clipboard copy is that in another project, where I'm copying from an IKImageView to the clipboard straight from the edit menu, it just sends a copy:(id) to the firstResponder, but I'm not actually sure what the firstresponder does with it...
EDIT 3 It appears that the CGBitmapContextCreate error is coming from [imageView image] which, oddly enough, IS a documented method.
It's possible that this is happening because I'm putting the image in there with a setImage:(id) method, passing it an NSImage*... Is there some other, more clever way of getting an NSImage into an IKImageView?
The -copy: method in IKImageView does what every other -copy: method does: it copies the current selection to the clipboard. It is, however, implemented as a private method in IKImageView for some reason.
You can just call it directly:
[imageView copy:nil];
This will copy whatever is currently selected to the clipboard.
I don't think there's a way to directly access the image content of the current selection in IKImageView using public methods, this is a good candidate for a bug report/feature request.
You can, however, use the private method -selectionRect to get a CGRect of the current selection and use that to extract the selected portion of the image:
//stop the compiler from complaining when we call a private method
#interface IKImageView (CompilerSTFU)
- (CGRect)selectionRect
#end
#implementation YourController
//imageView is an IBOutlet connected to your IKImageView
- (NSImage*)selectedImage
{
//get the current selection
CGRect selection = [imageView selectionRect];
//get the portion of the image that the selection defines
CGImageRef selectedImage = CGImageCreateWithImageInRect([imageView image],(CGRect)selection);
//convert it to an NSBitmapImageRep
NSBitmapImageRep* bitmap = [[[NSBitmapImageRep alloc] initWithCGImage:selectedImage] autorelease];
CGImageRelease(selectedImage);
//create an image from the bitmap data
NSImage* image = [[[NSImage alloc] initWithData:[bitmap TIFFRepresentation]] autorelease];
//in 10.6 you can skip converting to an NSBitmapImageRep by doing this:
//NSImage* image = [[NSImage alloc] initWithCGImage:selectedImage size:NSZeroSize];
return image;
}
#end
Ok, so the copy: nil fails, and the [imageView image] fails, but it turns out that I have another copy of the NSImage from when I added it into the view in the first place, so I could that. Also, CGImageCreateWithImageInRect expects a CGImageRef not an NSImage*, so I had to do some conversions.
In addition, for some reason the selection rectangle is flipped, either it's bottom origined, and the image is top, or the other way around, so I had to flip it.
And for some reason, the compiler suddenly started complaining that NSRect isn't the same type as CGRect (Which implies that it suddenly went from 32 to 64 bit or something... not sure why...)
Anyway, here is my copy of selectedImage:
- (NSImage*)selectedImage
{
//get the current selection
CGRect selection = flipCGRect(imageView, [imageView selectionRect]);
//get the portion of the image that the selection defines
struct CGImage * full = [[doc currentImage] CGImageForProposedRect: NULL context: NULL hints: NULL];
CGImageRef selectedImage = CGImageCreateWithImageInRect( full, selection);
//convert it to an NSBitmapImageRep
NSBitmapImageRep* bitmap = [[[NSBitmapImageRep alloc] initWithCGImage:selectedImage] autorelease];
CGImageRelease(selectedImage);
// //create an image from the bitmap data
NSImage* image = [[[NSImage alloc] initWithData:[bitmap TIFFRepresentation]] autorelease];
// //in 10.6 you can skip converting to an NSBitmapImageRep by doing this:
//NSImage* image = [[NSImage alloc] initWithCGImage:selectedImage size:NSZeroSize];
return image;
}
I wrote flipCGRect, and [doc currentImage] returns an NSImage*...