(iPhone SDK 3.x:) I have a UIControl subclass that creates a different number of subviews depending on the length of an NSArray property. Please take my word for it that this needs to be a UIControl rather than a UIView.
Currently I implement subview management in drawRect, beginning by removing all subviews and then creating the appropriate number based on the property. I don't think this is very good memory management and I'm not sure if drawRect is really the appropriate place to add subviews. Any thoughts on the best way to handle this pattern?
Thank you.
There is a method called layoutSubviews, and like the name already says, that method is thought to layout the subviews. You can call setNeedsLayout and the layoutSubviews method will be called (do not call layoutSubviews directly).
Related
Across my app I have several different subclasses of UIView: UIDatePicker, UIPicker, UIButton, UITableView, UITableViewCell, UITextView, etc. etc... For each of these I'd like to add a very simple drawRect custom implementation that I have working great.
Is there a simple way to get multiple subclasses of UIView to all have the same drawRect implementation without creating a subclass and repeating the same code across each UIPicker, UIButton, etc. etc... ?
I realize the solution to this is probably to write a delegate class for UIView's layer property and do the custom drawing in drawLayer, but I thought I would ask before I go re-working my code.
The answer seems to be no.
Furthermore, it appears that my idea for the workaround also doesn't work: I was thinking I could write a nice little class that implements the drawLayer method from CALayerDelegate and do the drawing in there, and then in each UIView or UIView subclass' init method do a self.layer.delegate = niceLittleCALayerDelegateClass.
My research, however, happened upon this: https://developer.apple.com/library/ios/documentation/GraphicsImaging/Reference/CALayer_class/#//apple_ref/occ/instp/CALayer/delegate which contains the damning sentence: In iOS, if the layer is associated with a UIView object, this property must be set to the view that owns the layer.
So, double nope.
Against rmaddy's advice I'm going to just write a subclass for each of the UIView subclasses I want to implement this drawing behavior in. We'll see how that goes.
Update:
I can't believe I didn't think of this before, but the "right" way to do this (that is to say without subclassing UIView's subclasses and adding a custom drawRect method to each subclass's subclass), from everything I can find, seems to be to either create a subview or a sublayer with a transparent background that does whatever custom drawing you want.
Obviously this is going to draw on top of the UIView you're actually using, so this would get exceedingly complicated if you're trying to draw things that interact with the default elements of the UIView, but for my purposes (just a simple frame drawn with a UIBezierPath) it seems to work great.
At present I'm not sure if the subview or the sublayer approach is more efficient. If anyone can shed light on that, I'd appreciate it.
If I want to initialize views programmatically, where in the viewcontroller lifecycle should this happen?
The initial intuition is loadView. However, here, we don't yet have the frame of the view itself (necessary for calculating the sizes/positions of the views). Ditto for viewDidLoad.
Next intuition is viewWillAppear- here we DO (finally) have a guarantee of the frame of the view. However, this has potential to be called many times throughout the vc lifecycle. Ditto for viewDidAppear, etc...
Finally, I found viewWillLayoutSubviews. This works for the initialization of most static layouts- however, whenever any view moves this gets called again (same problem as viewWillAppear).
I've seen recommendations to init the views in loadView and set their frames in viewWillLayoutSubviews (since setting frames should be idempotent, who cares if it gets called a couple times). But then why does apple so strongly encourage initWithFrame: as the standard initialization method of UIViews (https://developer.apple.com/library/ios/documentation/windowsviews/conceptual/viewpg_iphoneos/CreatingViews/CreatingViews.html)?
Would it be crazy to subclass all my UIViewControllers to have an initWithViewFrame: method? That way I can pass in a frame, manually set it immediately in loadView and be done with it? Or is it better to have a viewHasBeenFormatted flag in viewWillAppear that, if not set, calls the formatting of views and then sets it?
Or is this just apple's way of saying "use interface builder or you're screwed"?
Any help is appreciated!
edit- accidentally wrote loadView where I meant viewWillAppear (in final paragraph)
update- I guess I've come to terms with the fact that there is no place where
The frame is confidently known
The code will only be run once (on setup)
Looks like you're expected to initWithFrame: all your views in viewDidLoad (but then I guess the contents of that view shouldn't treat that frame as even remotely final? because how could it be when it was derived on an assumption? ugh...). Then re-set their frames in layoutSubviews. And make sure to manually handle the differences between initial layout and layout as a result of a moved view there... Man I feel like I've GOT to be missing something... (lol denial...)
I guess that, OR submit and use IB.
update2- viewWillLayoutSubviews WILL get called when one of its subviews is resized. So it is still disqualified as it fails property 2 of the required characteristics that I'm looking for. :(
If you're doing layout with IB, it's fine to do additional view initialization in viewDidLoad (for example, if you need to do stuff that IB doesn't handle well, or if you have UIView subclasses with properties not supported by IB). Alternatively, if you're not using IB, the documentation says you should use loadView to manually initialize your view hierarchy.
You're right, though, that you can't rely on the frame being accurate at that point. So you can accomplish layout via each view's autoResizingMask property, layout constraints (if you're iOS 6 and later), and/or overriding layoutSubviews.
My usual approach is to do layout to some degree in IB, then do anything else I need to (nontrivial layout, custom classes, etc) in viewDidLoad. Then, if I have layout to figure out that autoResizingMask doesn't cover (I'm supporting down to iOS 5), I override viewWillAppear (or layoutSubviews if I'm subclassing UIView) and do some pixel math. I've got a category on UIView to help with this that has things like:
-(void)centerSubviewHorizontally:(UIView *)view pixelsFromTop:(float)pixels;
-(void)centerSubviewHorizontally:(UIView *)view pixelsBelow:(float)pixels siblingView:(UIView *)sibling;
View controllers should not have initWithFrame: methods. What I do in all of my code (I never use IB) is to let the default loadView do its own thing. I create and setup all subviews in viewDidLoad. At this point the view controller's frame has at least a sane value. All subviews can be created with their own sane frames based on the initial size of the view controller's view. With proper autoresizingMask values this may be all you need.
If you need more specific subview layout, put the appropriate layout code in the viewWillLayoutSubviews method. This will deal with any view controller view frame changes including rotation, in-call status bars, etc.
If you don't use interface builder you should override loadView and initialize the views there. If you use autolayout you can also add your constraints there. If you don't use autolayout you can override the layoutSubviews method of your views to adjust the frames.
I have several UITextViews in several ViewControllers. In the past, when I have a couple of instances of needing a custom drawing for a TextView or Label, etc I would just adjust the drawing in viewWillAppear inside the VC that owned the UI object. This time, I will be needed several instances to be customized.
Would it be more appropriate to just create a subclassed UITextView and include the drawing code in drawRect versus having the same drawing code spread around several VC's. I am mainly worried about performance. Code maintainability is a secondary concern though.
To be clear, this is what I would use in drawRect:
- (void)drawRect:(CGRect)rect {
self.layer.cornerRadius = 10;
self.clipsToBounds = YES;
}
So after further testing, initWithFrame doesn't get called, but initWithCoder does. I have also found that initWithCoder is called once and so is drawRect. In my use case (a StaticCellTableView with the UITextView in a cell, what would the difference be?
You can include those 2 lines in init or initWithCoder (In case it's a xib or storyboard) method of the subclassed UITextView.
Basically, this is just properties of the object - the don't need to run every time the UIView needs to refresh itself.
Create a ViewController baseclass, and then use that as the superclass for all your other VCs. You can then add this and other convenience methods to it and share the wealth so to speak.
In this particular case, putting it at the end of viewDidLoad would be a good place, as that is only messaged once, where viewWillAppear may be called multiple times. If you use viewdidLoad the code will be executed once.
I would not put those lines in the view's drawRect since that's called for every refresh. I'd recommend that you subclass UITextView and add those lines to the custom init method.
If I'm creating a UIView programmatically and I wish to change the UIView properties (background, for example, or actually, messing with CALayers), must I place the code outside of UIView such as in the View controller? Can I put the code somewhere inside UIView?
I was checking out the CoreAnimationKioskStyleMenu example, its code is inside UIView but it's loaded from Nib and can be placed at awakeFromNib, so it doesn't seem to apply to my case.
That depends. Obviously, a good way to handle this is to use a xib file, as it is designed to hold data like this, but that isn't always the best answer for every situation.
If the view is meant to be reused frequently (like a button, or some widget) throughout the application, its best to store all that customization in a subclass of the UIView.
If its a single larger view that will always be managed by a UIViewController, you can keep some of the information in the UIViewController. However, if you end up subclassing a UIView anyway it's probably best practice to keep the data in the UIView.
As a general note, I believe its worth your time to push as much of this data into a xib using interface builder. Magic values (like colors or sizes) peppered through your code will always be a problem if you want to modify it. I have found modifying a xib to be much easier.
Actually there are some methods where you could place initialization/ customization code.
(void)willMoveToSuperview:(UIView *)newSuperview;
(void)didMoveToSuperview;
will get called as soon as u add the view as a subview to another view, at which point you already have the frame and all the properties, and you can do further customizing as you wish.
(void)layoutSubviews -- generally used for changing subviews' frames and layout organization.
Will get called each time the view needs to be redrawn by the system, or when you specifically call [self setNeedsLayout] on your UIView.
Hope this helps.
I'm using a container UIView to house a UIImageView and do some custom drawing. At this point I'd like to do some drawing on top of my subview. So overriding drawRect: in my container UIView will only draw below the subviews.
Is there a way to overload drawRect: in my subview without subclassing it?
I think method swizzling may be the answer, but I'm hoping not.
(NOTE: yes, it would have been smarter to have the UIView be the subview of the UIImageView, but unfortunately I'm committed to my mistake now.)
Are you sure you mean overload and not override?
Overloading -- creating a new method with the same basic name but different arguments and therefore a different selector -- could be accomplished by adding a new method category containing your new method to the existing class.
Overriding -- modifying the behavior of an existing method -- would require either monkey-patching the class's method table at runtime (for example, by swizzling) or subclassing.