How do I explicitly send key-value observing notifications from an NSResponder? - objective-c

I'm trying to make a NSTouchBar in an SDL application and I need to attach a responder to the NSWindow object (that's the only access SDL gives into the Cocoa windowing system).
https://developer.apple.com/reference/appkit/nstouchbar
If you explicitly adopt the NSTouchBarProvider protocol in an object,
you must also explicitly send the associated key-value observing
notifications within NSTouchBar methods; this lets the system respond
appropriately to changes in the bar.
What does that mean and how do I do it? I see lots of documentation about how to subscribe to the notifications, but not how to send them?
Right now I have:
#interface MyTouchBarResponder : NSResponder <NSTouchBarDelegate>
- (id)init;
- (NSTouchBar *)makeTouchBar;
- (nullable NSTouchBarItem *)touchBar:(NSTouchBar *)touchBar makeItemForIdentifier:(NSTouchBarItemIdentifier)identifier;
#property(strong, readonly) NSTouchBar *touchBar;
#end
and I'm attaching it to the window with the code from a previous question I asked here: How to create an NSTouchBar from an NSWindow object?
touchBarResponder.nextResponder = window.nextResponder;
window.nextResponder = touchBarResponder;
but my callbacks aren't ever being called (I put exit(0) in them to make it very obvious). When I hack the code directly into the SDL library, things work as expected, but that's not a viable permanent solution.
Thank you.

First, your custom responder should conform to NSTouchBarProvider (in the above, you declare the touchBar property, but not the explicit conformance)
Second, you want to make sure that your custom responder is in the responder chain of the window (whether the first responder or just later in the chain). After adjusting the responder chain with your above code, you want to call -makeFirstResponder: and pass in some view in the window (if you need that view to be first responder) or with the custom responder object. You should then verify that the window's firstResponder is that object.
With these in place, you should get at least one call to touchBar after the window is shown and made key.
To answer the question on key-value observing notifications, that is needed for when you want to change the actual NSTouchBar object being returned from touchBar. In the general case this isn't necessary, since it's unnecessary in the static touch bar case, and even in the dynamic case, you can rely on just setting the defaultItemIdentifiers on the previously created touch bar and it will update. However, should you need to change the touch bar object, you need to ensure that -willChangeValueForKey: and -didChangeValueForKey: are sent for touchBar when you change the return value. This developer documentation on KVO goes into much more detail.

Related

Using `valueForKey` to access view in UIBarButtonItem, private API violation?

Since UIBarButtonItem doesn't subclass UIView, it's impossible to get at the normal characteristics like its frame.
One way to do this is [barButtonItem valueForKey:#"view"]
This works perfectly, and allows you to add a GestureRecognizer (for instance) to the underlying UIView.
However, is this a private UIKit API violation?
This is not private in terms of immediate rejection upon validation, but it's private enough to be considered fragile (that is, new iOS version can break your existing app in the app store that's using the code).
I can say, that a similar code (fetching backgroundView ivar of UIToolbar via KVC) has passed app store validation and is being used in production.
In case of possible bad things, you must wrap the method in #try { ... } #catch, so that you intercept KVC possibly failing in newer iOS release.
Five Pieces of Evidence for "It's Not Private"
It's a property that you can get to in other ways. Try this and one of those views is, in fact, the _view ivar of the UIBarButtonItem in question. This indicates that access to this UIView is not prohibited itself, though the KVO way in might be questionable (but I doubt it).
NSArray *array = self.toolBar.subviews;
for (UIView *view in array) {
view.backgroundColor = UIColor.greenColor;
}
They actually trigger the KVO for this property. ivars do not have to trigger the KVO API, right?
#Farcaller mentions a similar case which is for sale in the App Store. Since he/she answered within the first 20 minutes of the question being up there, it's reasonable (but not safe!) to assume that there might be thousands of apps in the App Store that do this.
This UIView gets subbed out each time the button is pressed, so you cannot just, for example, set a gesture recognizer on it and be done. You can, however, keep setting the same gesture recognizer every time the view gets replaced. To me, this is actually more evidence that it's not a private API thing, but rather you have to be very careful when using it (and use KVO to make sure you have the latest one).
My app is for sale in the App Store and does this.

Sending an event up the UIResponder chain?

I just created a custom UIViewController with some user actions like touch. I would like to handle the user interaction in the parentObject. In other words the one that created the ViewController.
From other languages I am used to use Events that are pushed up. So my parent object would have some kind of listener on the reference of the ViewController object it can react to.
How would that type of interaction handled by Objective C?
This can be done by 1) responder chain, 2) notifications and 3) delegates.
All UI objects form the responder chain, starting from the currently focused element, then it's parent view and so on, usually until the application object. By sending action to the special First Responder object in your nib, you'll throw it down the responder chain until someone handles it. You can use this mechanism for firing events without knowing who and when will handle them. This is similar to HTML event model.
Notifications send by NSNotificationCenter can be received by any number of listeners. This is the closest analog to e.g. C# events.
Delegates is the simplest mechanism of sending event to one single object. The class declares weak property named delegate that can be assigned with any object, and a protocol that this object is supposed to implement. Many classes use this approach; the main problem is that you can't have more than one listener this way.
you should look into delegations/delegate for interactions between two viewControllers. You need to understand how it works first.
https://developer.apple.com/library/mac/#documentation/General/Conceptual/DevPedia-CocoaCore/Delegation.html
It sounds like you need to implement a delegate protocol, which will allow your 'child' view controller to communicate back to it's 'parent'
See http://developer.apple.com/library/ios/#documentation/General/Conceptual/DevPedia-CocoaCore/Delegation.html

multiple delegates in objective-c

i have been working on moving one of my apps away from the "shared appdelegate" process which seems to be frowned up, despite its over whelming use. i have been attempting to setup protocol methods for what i want to do but am having zero luck. my question is, can you even have lets say a single viewcontroller send delegate requests to multiple classes? from what im finding out it doesn't seem like you can. which doesn't make sense because i thought that was the whole point of delegates and protocols with mvc. now just to clarify, i know you can have a single viewcontroller act as the delegate for multiple other viewcontrollers. but that's not what i am asking. for a simple example, lets say you have apples flip-utility template. the "done" button just calls a delegate method to the mainvc to dismiss it. now lets say we added a new class called...
#interface NewClass : NSObject <TheOtherDelegate>
and it had a delegate method...
- (void)doSomething
{
NSLog(#"The Delegate did something...");
}
can we have a button on the flipsideviewcontroller, that we wanted to call that delegate method, but still keep the "done" button call to the delegate method on the mainviewcontroller that dismisses it?
that being said, i put together a quicky project just to see if it would work and it doesn't. i came across an "answer" that says you have to instantiate the class first you want to be the delegate...
NewClass *myDelegate = [NewClass alloc] init]
[fillInMethodHere setDelegate:myDelegate];
not sure why it got a correct answer check, because needless to say it doesn't work. is there something i am missing? i scoured ib to see if there is some "delegate" connection somewhere but i couldn't find anything.
on a side note, as i was working in my working project, i read a suggestion about removing the #import and adding #class. again, that broke all kinds of things. the strange thing is before doing that, what i had so far was working and building fine. when i removed the new #class and un-commented the #import. xcode all of a sudden gave me an error "cannot find protocol deceleration for..." but yet, it worked seconds earlier. i would up having to remove the protocol code and re-add it for it to work again. very starge.
any help would be appreciated. everything iv read in docs, google, stack, etc that say something should work, don't in an actual project.
A "delegate" isn't some fancy object. It's simply a synthesized property of type id called delegate. If you wanted to, you could have an arbitrary number of properties that all conformed to the same protocol. Then when you wanted to issue a callback, you would just address all of them:
[self.mydelegateA doSomething];
[self.mydelegateB doSomething];
etc.
You could also have an NSMutableArray property that you could add objects to, and then use [self.myMutableArrayOfDelegates makeObjectsPerformSelector:#selector(doSomething)].
Finally, there's always the route of NSNotificationCenter (not to be confused with push notifications) is a class that provides an inter-object messaging system. Many objects can register for a message that any other object can send.
Please see the Apple's documentation for more information. Click Here.
Regardless of the fact that this is OS X documentation, it's still quite good at explaining things visually: click here.
Here's an example of simply changing the name of the delegate property: click here
And here's an example of adding another protocol and a second delegate: click here
Finally, here's an example that builds on the previous two and has a third delegate that also conforms to the same protocol: click here

Using Protocols in Objective C to Transfer Data Between Different Objects?

Hey guys, I currently have a root table view which has a toolbar at the bottom and has labels and a refresh button within it, much like the Mail app's toolbar. This root table view controller obtains data from a server by allocating and initializing a DataUpdater class. Within this class are the NSURLConnection delegate methods that are called while communicating with the server.
As you can probably guess, I need to know when certain (delegate) functions are called within the DataUpdater class and the values of the parameters passed to these delegate functions so that I can update the labels on the toolbar accordingly (i.e. Connecting..., Updated, etc).
The problem I am having is determining how to notify the root table view controller of what is going on in these delegate methods. Would I use protocols, if so how? I have been skimming the documentation and don't quite see how I would get this effect. Or would you suggest I implement my program another way?
Thanks in advance!
A protocol is a kind of contract that says: I promise to provide the non-optional methods defined in the protocol, and maybe even the optional ones. It's purpose is like Java interfaces: to work around missing multiple-inheritence.
The delegate pattern in Objective-C normally works like this: you define a protocol, and then in your class, you define a variable like id<MyProtocol> myDelegate; and define a setter and maybe getter (either via normal methods, e.g. - (void)setDelegate:(id<MyProtocol>)aDelegate; or via properties.
Note that the delegate is not retained ! So if you work with a property, you need the assign option, not retain.
Now back in your class, you check whether myDelegate is nil and if not, you can directly call its non-optional methods. If you want to call an optional method, you first need to verify its presence via respondsToSelector:.
So if you decide to use the delegate pattern, you need to define a protocol, add that protocol to your root table view controller, implement the necessary methods there, and make sure to call [foo setDelegate:self]; or something similar to inform your other class that the root table view controller is the delegate. And of course implement the delegate calls in your class.
Edit:
An alternative might be to use NSNotifications, BTW. The advantage of notifications is that you can have multiple objects listen and react to them. The disadvantage is that you cannot (directly) pass values back. For example, you can define a delegate method that asks the delegate whether to do something or not. That's not possible with notifications, it's more like shouting into a room instead of having a one-to-one conversation.
DarkDust's answer about protocols is fine but I would like to add some things to it.
One underlying thing that is often forgotten when it comes to delegation is object ownership. When a program is running it creates a tree of objects. Its root object is the application delegate and for example it owns a navigation controller, which owns the individual view controllers, which own the view and the view owns its subviews and so on.
Often the question comes up: "Why is the delegate not retained, just assigned?" The problem is that if you send a message to a deallocated object the program crashes. So how do you make sure the delegate stays around? The answer is object ownership.
I give you an example: a UITableView and its data source which is the TableViewController which is nothing but a delegate. The TableViewController holds a reference with its view property to the UITableView, so it owns the TableView. That means when the tableView is alive there must also be its parent object present, which is the UITableView's delegate. So there is no danger that the delegate goes away somehow.
In the end it is again all about memory management.
Take home message is: think upfront about object ownership will make your program mode modular, easier to maintain and will lead to a looser coupling between individual objects.

Where do FirstResponder methods come from?

I'm looking at IKImageDemo supplied by Apple, the rotate round-slider is linked to a setRotation: method in the FirstResponder. However, none of the objects in the project seem to HAVE such a method, and yet the code works.
I'm trying copy this into my own project, and MY FirstResponder doesn't have a setRotation: method, so I'm not sure where it lives. Google has been unhelpful...
thanks.
Well, the first responder in the app happens to be an instance of IKImageView. IKImageView responds to the setRotation: selector (which can be seen by passing respondsToSelector:#selector(setRotation:) to any instance of IKImageView), although I cannot find where in documentation it mentions the setRotation: method
First Responder methods aren't magic. What happens when a message is sent to the first responder is that the app's current first responder (this is usually the focused view/control) is asked whether or not it implements the method. If it does, the method is called. If it doesn't, the next responder up the chain is asked, and so on until the top level (the NSApplication instance) is reached. The object must actually implement the method for it to be called, it can't just declare it.
In this case IKImageView implements -setRotation: as a private method. This means that the method is present (which is why the IKImageView accepts the message sent to the First Responder) but its use is not documented or supported. It seems odd that Apple would ship an example using a private method but there you go. It's definitely the case that sometimes methods are accidentally left out of the public headers when their use is supported, however it's generally wise to avoid private methods unless someone from Apple has specifically told you it's OK to use one.
You can generate headers for all methods of an Objective-C object, including private methods, from the binary using class-dump.
IKImageView has a public method -setRotationAngle: which is probably the way to go if you want to change the rotation.
I've found a way of resolving this annoyance. Even in the original Apple example, once you remove the binding for setRotation in the First Responder, you cannot put it back, unless doing this trick: simply use the Attributes Inspector for the First Responder and add a User Defined action "setRotation:" with type "id". Now even the yellow triangle in the First Responder binding for setRotation: in the Apple example disappears, and it shows up also in my own IKImageView instance.