I started learning iOS programming, and I came across something like this: to handle the event when users press return after input text in a UITextField, my view controller implements "textFieldShouldReturn: textField" from UITexTFieldDelegate protocol. But, to handle the event when "Editing Did End", I need to declare an IBAction method in the view controller, and wire up the event to it.
What is the difference between these two types of event handling methods?
The "Editing Did End" is part of the UIControl class, and UITextField inherits it from UIControl. This kind of event is one of the registered touch events this class responds to and the way to do this registration is to assign an action (IBAction using IB) to the particular event.
The delegate protocol part of the implementation is due to the "extension mechanism" provided by delegate protocols in Cocoa design patterns. Essentially the idea is that you can provide special behavior to the object, in this case UITextField, without subclassing. Imagine for example you want some kind of validation before allowing the "return" button. In such case you would be forced to setup a UITextField subclass with this validation code hard-coded in it. And for each different validation code you would be forced to implement a different subclass. Using the delegate mechanism, you delegates something else (typically the text field owner view controller) to perform this "class extension", without subclassing. So you do class customization per-instance.
Note that this approach is coherent: everything that is considered as "event" (touch up, touch down, did end editing, all related to some user interaction with the display and not with the keyboard) is wired to the class using the UIControl target-action mechanism. Everything that is "specific" of the instance (and not the class!) is managed using the delegate.
In protocols (denoted with #protocol //name\ superclass) one can specify methods that the conforming class either can, will, or must take on with #optional, or #required methods. One might use protocols to say, pass data from one view to the next, so long as the second view conformed to the protocol (like #implementation Classname: NSObject ) This is in stark contrast to IBActions and void functions which are local and impassable to other controllers (except with class methods).
Related
I have just been learning iPhone apps development but I have a hard time in understanding what delegate actually means? Can anyone tell me with example what it does and how important it is? Thanks for any helps!
It's a key concept to understand conceptually so getting your head around how to think about it ahead of the technical details is important. Simply put, a delegate is a callback.
Two main scenarios to use delegates:
A class or control wants to abstract out the details on how to do work (like retrieve data).
Allow others to hook code into a pipeline.
Examples:
UITableView - a table view is just a control that knows how to render a list of cells. It handles all the heavy lifting of rendering, scrolling, etc... But, it has no idea how to load your data. So you implement a datasource delegate which has methods to get the cell data for a given row etc... That makes it easy on you. You just use the control and plug in the specifics for your data. The UITableView will do everything for you ... just answer a few specific questions for. A delegate answers those few specific questions.
A text control - you add a text control to your view and voila! you can type in it and alls good. But what if you want to do something when they start typing or when they're done typing? Well, the text control offers a delegate with methods that allow you to hook into the execution pipeline of the text control. It allows the text control to do everything for you and allows you to interject code where you need it. Many times, there's way to interject code to make a decision on whether something is allowed. The control will call back and ask, should I be able to do x? You can interject code and influence the behavior.
If you're creating a control or class, you can create your own protocol, datasource delegates etc... so your control can focus on doing what's advertised. For example, let's say you wanted to create a task control. You could:
First, create a contract. Hey, if you're going to provide data for my control, these are the questions I'm going to ask you. I'll take it from there... In this case, I'm going to ask you the number of tasks and I'm going to have you give me a task given the row number.
#protocol XXTaskBoardDelegate <NSObject>
-(NSInteger*)getTaskCount;
-(XXTask*)getTaskForRow:(NSInteger*)rowNumber;
#end
In the control or class, give the consumer a way to give us the delegate datasource class that will answer the questions the control will ask. At this point, the control is a pure control. It knows nothing about how you get your data. It's asking for an object (id) that implements a contract/protocol. id
#implementation XXTaskBoard
- (void)setDelegate:(id<XXTaskBoardDelegate>)newDelegate
{
// the control stores the delegate so it can callback and ask you questions.
}
Then, for the delegate class, in the header declare you implement that formal protocol
and in the implementation m file you provide the code.
#interface AppController : NSObject<XXTaskBoardDelegate>
{
//...
}
then, implement it in the implementation
#implementation AppController
- (NSInteger*)getTaskCount
{
return [model queryTaskCount];
}
- (XXTask*)getTaskForRow:(NSInteger*)rowNumber
{
return [[model tasks] getItem:(NSInteger*)rowNumber];
}
A delegate is an object that another class can pass messages to. In practice delegate classes have to conform to a delegate protocol.
For instance we will take a subclass of a table view controller. This is a delegate for your table view. First you define that it is a table view delegate by doing this:
MyTableViewController : UITableViewController <UITableViewDelegate>
This says that class MyTableViewController is a subclass of UITableViewController and CONFORMS to the UITableViewDelegate protocol.
Setting [tableView setDelegate:self] (or defining it as such in IB) then passes the self object to the tableview in order for the tableview to send messages to it.
The main message it sends is the didSelectRowAtIndexPath message which tells your class that the user has pressed a table view cell.
So the object that takes the click event (the table view) passes on the message that the cell has been clicked to the delegate object (which in this case is your MyTableViewController object).
Delegate protocols exist so that you can make sure that the delegate object has the necessary methods to deal with your messages. Methods in a delegate protocol can be #optional or enforced. Any methods that are optional don't have to be defined. In your MyTableViewController class the method didSelectRowAtIndexPath is optional - you don't have to have it. If the table view doesn't find the method it just won't call it.
However the cellForRowAtIndexPath is necessary and without it your app won't compile.
I hope this helps and is straightforwards for you. If you need any more info let me know.
Delegates are just way of getting callbacks from something. You pass a delegate (a pointer to an object that conforms to a protocol) to something and when it has new data for you or when an event occurs that something make a method call on the delegate.
For example, when events occur, like your app is put into the background or the app is about to terminate the UIApplication object will call your application delegate to let it know. When a CLLocationManager has a new GPS position is will call your delegate to pass it the new position. UITableViews call their delegates to get UITableViewCells to display in the table. There are many uses of delegates in iOS.
Can any one differentiate when do we use Delegation over category and vice versa. I am clear over this.
Thanks
Category allows to add new methods or overwrite existing methods on a class, thus allows to extend a class without subclassing. Adding methods is the most useful aim, overwriting can go really wrong if you do not know exactly what the class being extended does.
It is more a language feature not a pattern, it works on each class.
Delegate is a pattern not a language feature, the class that is supposed to used must be coded for it, otherwise it won't work.
Usually a delegate will be required to implement a protocol known by the class that is going to receive the delegate. The class will then use the delegate to do stuff it was coded for, some of the most common are sending notifications, using a part of a strategy pattern, that is asking question in certain part of code to make decisions based on the concrete delegate's implementation, letting delegate execute an action or any combination of them.
For example UIApplicationDelegate is a notification (application:didFinishLaunchingWithOptions:) and action (application:openURL:sourceApplication:annotation:) protocol, UITextFieldDelegate is notification (textFieldDidBeginEditing:) and strategy (textField:shouldChangeCharactersInRange:replacementString:).
Actually I can imagine using category to implement delegating in all this cases I stated above: sending notifications, doing actions, taking part in a strategy. But it would require you to very very good know what to extended class is doing, probably to have it code, otherwise you can very easily break the class or be broken by changed class implementation. So this usage would be in my opinion highly wrong.
category: Adds methods to an existing class.
delegate: Modifies behavior of another object by allowing some other object, the delegate, to participate in the object's operation.
Say you've got an iOS application with three tables. Even though they're configured the same way, those tables each may behave differently if they have different delegates.
You can't do that with a category because a category applies equally to all instances of the class it extends. On the other hand, if you want to extend the UITableView class to add some new capability, you need a category*. Delegates are instance-specific and limited to the role envisioned for them by the designer of the delegating class.
*or a subclass, of course.
Categories are used when you need to extend the class without creating a subclass, for example when you need to add a method named isURL to the NSString you can make use of categories as follows, here we not creating a subclass, instead we are extending the implementation.
#interface NSString (Utilities)
- (BOOL) isURL;
#end
Delegates are similar to callback functions,
I have a viewController that's running my game view. The viewController spawns instances of a Dots, which are subclassed from UIView that are added as subviews to the viewController view. Within the dots class I have certain behavior if they are touched (touchesBegan) but I'd also like touching the dots to add to the score which is a property of the viewController, but I can't see any way to communicate that out without ridiculous workarounds. What am I missing here?
Two options (but not the only ones) are NSNotificationCenter and delegate+protocol.
NSNotificationCenter is simpler to implement and use but not ideal in all cases. The viewController would call addObserver to indicate it is interested in some message. The dots would call postNotificationName to send the message to interested objects. Be sure to call removeObserver in the viewController's dealloc.
Delegate+protocol lets you define a more rigid, well-defined, compiler or runtime verifiable interface between objects. For details, see Protocols in the Objective-C Programming Language.
In your case, I think defining a protocol is the better option. The viewController would make itself the delegate of the dot objects. The dot objects would have a delegate property and call the protocol methods through the delegate property.
Why is the score a property of the ViewController in the first place? I can see having a property for the label that displays the score, but the score itself belongs in some kind of model object.
The advantage of this approach is that you could then use Key-Value Observing (KVO), adding your ViewController as an observer of the model object. In fact, this is pretty much what KVO was designed for: an easy way to notify an arbitrary number of observers about a change to a property.
Reference:
Key-Value Coding Protocol Reference
Key-Value Observing Protocol Reference
Model Object Implementation Guide
Here is my dilemma. I would like to have a text box and a button. The user types in text and then presses the button. Once the button is pressed, a text message window (using MFMessageComposeViewController) comes up. I don't know how to set this up. The problem is that the TextBox will require a delegate (UITextFieldDelegate) and the MFMessageComposeViewController will require an MFMessageComposeViewControllerDelegate. How can I have a .h file that declares a view that is more than one delegate?
I'm new to iPhone programming so any help on how to have an interface view that handles more than one delegate (so that I can have multiple types of controls in my view) would be really helpful!
A delegate does not need to be a view. Indeed, in most cases it probably shouldn't be. Often you will make a controller object the delegate, although this depends a lot on what you're doing.
The delegate protocols you need (MFMessageComposeViewControllerDelegate and UITextFieldDelegate) are quite distinct, so a single object can readily implement the methods of both without any confusion. But even if you are the same delegate type for several objects, the methods will be passed a pointer to the calling object so you can decide what to do case-by-case if necessary.
If you just mean how do you declare your class as implementing both protocols, you would do this:
#interface MyDelegate : NSObject <MFMessageComposeViewControllerDelegate, UITextFieldDelegate>
{
...
}
...although this presupposes that the protocols are formally required, which I don't think is the case here. In which case such a protocol list is unnecessary.
Otherwise, I probably am not understanding your question...
EDIT: OK, it seems like what you're looking for is how to link up the delegates at runtime. This varies according to the particular class, but for MFMessageComposeViewController you do this:
MFMessageComposeViewController* composer = ...;
id<MFMessageComposeViewControllerDelegate>* delegate = ...;
composer.messageComposeDelegate = delegate;
Easy, no? In this case the protocol is required, so you'd have to include it in the interface as described previously.
In general, if an object uses a delegate for anything, it will have a property or a method to allow you to set it, which you'll find in the documentation. Eg, in this case: Properties for MFMessageComposeViewController.
Note that delegate properties are conventionally weak references, so the objects in question need to be retained somewhere in your application.
In many situations, such as making the keyboard go away when the use clicks done, there are two options: set the text field's delegate to self, and adopt the UITextFieldDelegate protocol, and then use the method
- (BOOL)textFieldShouldReturn:(UITextField *)textField;
to resignFirstResponder and return YES. But you can also
addTarget:self
action:#selector(myMethod:)
forControlEvent:UIControlEventDidEndOnExit];
or something like that, using the did end on exit event, and then in the method, [sender resignFirstResponder]. So what is the best option in situations like these: the delegate, or the event?
The quick rule of thumb is that delegates are supposed to answer the question of "should I?" on behalf of the object they are a delegate for. Events, on the other hand, are broadcast afterward to let listeners know that something has happened.
In your case, while you could call [sender resignFirstResponder] in response to the event, you're mixing metaphors by doing this. Your delegate should have already made the decision to hide the keyboard (or not) and the event being broadcast is merely to let all the other components know that they keyboard hid.
If you are going to be paired with one other class, where the real type of that class may vary, then it makes a lot of sense to formalize that pairing into a protocol and a delegate arrangement.
If the information you want to send is targeted at a broader set of objects, then it starts to make more sense to use notifications - though now you have somewhat obscured what information is being passed by the notification as there's no central definition for what to expect.
Both are about an equal load to work with - with a delegate you have to set yourself and then remember to unset yourself before you are deallocated. You have to do the same thing with notifications, remember to start listening and then unsubscribe before you are deallocated.
Also, you should try as much as possible to make sure you send notifications out on the main thread as the notices get sent on the same thread they started from. Same goes for delegate methods, it's not very kind to call a delegate method from some other mystery thread!
The delegate makes your objects more reusable they are an adapter that lets any object interact with the defined behaviors of that object and use the object. I would say delegates should be adopted by the object responsible for keeping the state of and defining behavior to actions that will occur in the object that it is using. Events should be used for any other objects that are intersted in particular action that the object that has the protocol does (so objects not responsible for keeping the state of the object that defines the protocol).
For example: A view controller using a textfield will adopt its protocol to dismiss the keyboard and any other behaviors that can occur for a textfield, maybe another controller will do some animation when the keyboard is dismissed, so it will register to the textfield as an event in order to receieve the event of the keyboard being dismissed.