API Design

MonoTouch.ObjCRuntime

The MonoTouch.ObjCRuntime (http://www.go-mono.com/docs/index.aspx?tlink=20@N:MonoTouch.ObjCRuntime) namespace allows developers to bridge the worlds between C# and Objective-C. This is a new binding designed specifically for the iPhone based on the experience from Cocoa# and Gtk#.

MonoTouch.Foundation

The MonoTouch.Foundation (http://www.go-mono.com/docs/index.aspx?link=N:MonoTouch.Foundation) namespace provides the basic data types designed to interoperate with the Objective-C Foundation framework that is part of the iPhone and it is the base for object oriented programming in Objective-C.

MonoTouch mirrors in C# the hierarchy of classes from Objective-C. For example, the Objective-C base class NSObject (http://developer.apple.com/iphone/library/documentation/Cocoa/Reference/Foundation/Classes/NSObject_Class/Reference/Reference.html) is usable from C# via MonoTouch.Foundation.NSObject (http://www.go-mono.com/docs/index.aspx?link=T:MonoTouch.Foundation.NSObject).

Although this namespace provides bindings for the underlying Objective-C Foundation types, in a few cases we have mapped the underlying types to .NET types. For example:

• Instead of dealing with NSString (http://developer.apple.com/iphone/library/documentation/Cocoa/Reference/Foundation/Classes/NSString_Class/Reference/NSString.html) and NSArray (http://developer.apple.com/iphone/library/documentation/Cocoa/Reference/Foundation/Classes/NSArray_Class/Reference/Reference.html) the runtime instead exposes these as C# strings (http://www.go-mono.com/docs/index.aspx?link=T:System.String) and strongly typed arrays (http://www.go-mono.com/docs/index.aspx?link=T:System.Array) throughout the API.
• Various helper APIs are exposed here to allow developers to bind third party Objective-C APIs, other iPhone APIs or APIs that are not currently bound by MonoTouch.

For more details on binding APIs see the MonoTouch Binding Generator section.

MonoTouch.UIKit

The MonoTouch.UIKit (http://www.go-mono.com/docs/index.aspx?link=N:MonoTouch.UIKit) namespace contains a one to one mapping to all of the UI components that make up CocoaTouch in the form of C# classes. The API has been modified to follow the conventions used in the C# language.

C# delegates are provided for common operations. See the delegates section for more information.

OpenGLES

For OpenGLES we distribute a modified version (http://www.go-mono.com/docs/index.aspx?link=N:OpenTK) of the OpenTK (http://www.opentk.com) API, an object-oriented binding to OpenGL that has been modified to use CoreGraphics data types and structures as well as only exposing the functionality that is available on the iPhone.

OpenGLES 1.1 functionality is available through the GL type.

OpenGLES 2.0 functionality is available through the GL type.

Types

Where it made sense, we exposed C# types instead of low-level MonoTouch.Foundation types to the C# universe. This means that the API uses the C# "string" type instead of NSString and it uses strongly typed C# arrays instead of exposing NSArray.

Additionally, instead of exposing CGRect, CGPoint and CGSize from the CoreGraphics API, we replaced those with the System.Drawing implementations RectF, PointF and SizeF as they would help developers preserve existing OpenGL code that uses OpenTK.

Inheritance

The MonoTouch API design allows developers to extend native Objective-C types in the same way that they would extend a C# type, using the "override" keyword on a derived class as well as chaining up to the base implementation using the "base" C# keyword.

This design allows developers to avoid dealing with Objective-C selectors as part of their development process as the entire Objective-C system is wrapped already by the MonoTouch libraries.

Types and Interface Builder

When you create .NET classes that are instances of types created by Interface Builder, you need to provide a constructor that takes a single IntPtr parameter.   This is required to bind the managed object instance with the unmanaged object.    The code consists of a single line, like this:

public partial class void MyView : UIView { 

   // This is the constructor that you need to add.

   public MyView (IntPtr handle) : base (handle) {}

}

Delegates

Objective-C and C# have different meanings for the words delegate in each language.

In the Objective-C world and in the documentation that you will find online about CocoaTouch, a delegate is typically an instance of a class that will respond to a set of methods. This is very similar to a C# interface with the difference that the methods are not always mandatory.

These delegates play an important role in UIKit and other CocoaTouch APIs. They are used for various things:

• To provide notifications to your code (Similar to event delivery in C# or Gtk+)
• To implement models for data visualization controls.
• To drive the behavior of a control.

The programming pattern was designed to minimize the creation of derived classes to alter behavior for a control. This solution is similar in spirit to what other GUI toolkits have done over the years: Gtk's signals, Qt slots, Winforms events, WPF/Silverlight events and so on. To avoid having hundreds of interfaces (one for each action) or requiring developers to implement too many methods they do not need, Objective-C supports optional method definitions. This is different than C# interfaces that require all methods to be implemented.

In Objective-C classes, you will see that classes that use this programming pattern expose a property, usually called delegate which is required to implement the mandatory parts of the interface and zero or more of the optional parts.

In MonoTouch we offer three mutually exclusive mechanisms to bind to these delegates:

1. Via events.
2. Strongly typed via a Delegate property.
3. Loosely typed via a WeakDelegate property.

For example, consider the UIWebView (http://developer.apple.com/iphone/library/documentation/UIKit/Reference/UIWebView_Class/Reference/Reference.html) class. This dispatches to a UIWebViewDelegate (http://developer.apple.com/iphone/library/documentation/UIKit/Reference/UIWebViewDelegate_Protocol/Reference/Reference.html) instance which is assigned to the delegate (http://developer.apple.com/iphone/library/documentation/UIKit/Reference/UIWebView_Class/Reference/Reference.html#//apple_ref/occ/instp/UIWebView/delegate) property.

DateTime startTime, endTime;
var web = new UIWebView (new CGRect (0, 0, 200, 200));
web.LoadStarted += (o, e) => startTime = DateTime.Now;
web.LoadFinished += (o, e) => endTime = DateTime.Now;

void SetupTextField (UITextField tf)
{
    tf.ShouldReturn = delegate (textfield) {
        textfield.ResignFirstResponder ();
        return true;
    }
}

class Notifier : UIWebViewDelegate  {
    DateTime startTime, endTime;
 
    public override LoadStarted (UIWebView webview)
    {
        startTime = DateTime.Now;
    }
 
    public override LoadingFinished (UIWebView webView)
    { 
        endTime= DateTime.Now;
    }
}

     var web = new UIWebView (new CGRect (0, 0, 200, 200));
     web.Delegate = new Notifier ();

class Notifier : NSObject  {
    DateTime startTime, endTime;
 
    [Export ("webViewDidStartLoad:")]
    public void LoadStarted (UIWebView webview)
    {
        startTime = DateTime.Now;
    }
 
    [Export ("webViewDidFinishLoad:")]
    public void LoadingFinished (UIWebView webView)
    { 
        endTime= DateTime.Now;
    }
}
[...]
     var web = new UIWebView (new CGRect (0, 0, 200, 200));
     web.WeakDelegate = new Notifier ();

foo.delegate = [[SomethingDelegate] alloc] init]

    foo.Delegate = new SomethingDelegate ();

public class SampleTitleModel : UIPickerViewTitleModel {
 
    public override string TitleForRow (UIPickerView picker, int row, int component) 
    { 
        return String.Format ("At {0} {1}", row, component);
    } 
}
 
[...]
 
pickerView.Model = new MyPickerModel ();

   UIAccelerometer.SharedAccelerometer.Acceleration += (sender, args) => {
      UIAcceleration acc = args.Acceleration;
      Console.WriteLine ("Time={0} at {1},{2},{3}", acc.Time, acc.X, acc.Y, acc.Z);
   }

var web = new UIWebView (new CGRect (0, 0, 200, 200));
web.LoadStarted += () => { startTime = DateTime.Now; }
web.LoadFinished += () => { endTime = DateTime.Now; }

Responding to Events

In Objective-C code sometimes event handlers for multiple controls and provides for information for multiple controls will be hosted in the same class. This is possible because classes respond to messages, and as long as a class responds to that message it is possible to link objects together.

As previously detailed, MonoTouch supports both the Objective-C delegate pattern, when you can create a new class that implements the delegate and override the desired methods as well as the C# event-based programming model.

It is also possible to support Objective-C's pattern where responders for multiple different operations are all hosted in the same instance of a class. To do this though, you will have to use low-level features of the MonoTouch binding.

For example, if you wanted your class to respond both to the UITextFieldDelegate.textFieldShouldClear: message as well as the UIWebViewDelegate.webViewDidStartLoad: in the same instance of a class you would have to use the [Export] attribute declaration. It would look like this:

public class MyCallbacks : NSObject {
    [Export ("textFieldShouldClear:"]
    public bool should_we_clear (UITextField tf)
    {
        return true;
    }
 
    [Export ("webViewDidStartLoad:")] 
    public void OnWebViewStart (UIWebView view)
    {
        Console.WriteLine ("Loading started");
    }
}

The C# names for the methods are not important, all that matters are the strings passed to the [Export] attribute.

When using this style of programming you are responsible for making sure that the C# parameters match the actual types that the runtime engine will pass.

Models

In UIKit storage facilities or responders are implemented using helper classes, these are usually referred in the Objective-C code as delegates and they are implemented as protocols.

Objective-C protocols are like interfaces, but they support optional methods, that is, not all of the methods need to be implemented for the protocol to work.

There are two ways of implementing a model, you can either implement it manually or use the existing strongly typed definitions.

The manual mechanism is necessary when you try to implement a class that has not been bound by MonoTouch. It is very easy to do, all you have to do is:

• Flag your class for registration with the runtime
• Apply the [Export] attribute with the actual selector name on each method you want to override
• Instantiate the class, and pass it.

For example, the following implement only one of the optional methods in the UIApplicationDelegate protocol definition:

public class MyAppController : NSObject {
        [Export ("applicationDidFinishLaunching:")]
        public void FinishedLaunching (UIApplication app)
        {
                SetupWindow ();
        }
}

The Objective-C selector name ("applicationDidFinishLaunching:") is declared with the Export attribute and the class is registered with the [Register] attribute.

MonoTouch provides already strongly typed declarations ready to use that do not require manual binding. To support this programming model, the MonoTouch runtime supports the [Model] attribute on a class declaration. This informs the runtime that it should not wire up all the methods in the class, unless the method is explicitly implemented.

This means that in MonoTouch.UIKit the classes that represent a protocol with optional methods are written like this:

[Model]
public class SomeViewModel : NSObject {
    [Export ("someMethod:")]
    public virtual int SomeMethod (TheView view) {
       throw new ModelNotImplementedException ();
    }
    ...
}

When you want to implement a model that only implements some of the methods all you have to do is to override the methods that you are interested in, and ignore the other methods. The runtime will only hook up the overwritten methods, not the original methods to the Objective-C world.

The equivalent to the previous manual sample is:

public class AppController : UIApplicationDelegate {
    public override void FinishedLaunching (UIApplication uia)
    {
     ...
    }
}

The advantage is that there is no need to dig into the Objective-C header files to find the selector, the types of the arguments, the mapping to C# and that you get intellisense from MonoDevelop and strong types.

Interface Builder Outlets and C#

This is a low-level description of how Outlets integrate with C# and is provided for advanced users of MonoTouch. When using MonoDevelop the mapping is done automatically behind the scenes using generated code on the flight for you.

When you design your user interface with Interface Builder, you will only be designing the look of the application and will establish some default connections. If you want to programatically fetch information, alter the behavior of a control at runtime or modify the control at runtime, it is necessary to bind some of the controls to your managed code.

This is done in a few steps:

1. Add the outlet declaration to your "File's owner".
2. Connect your control to the "File's owner".
3. Store the UI plus the connections into your XIB/NIB file.
4. Load the NIB file at runtime.
5. Access the outlet variable.

The steps (1) through (3) are covered in Apple's documentation for building interfaces with Interface Builder.

When using MonoTouch your application will need to create a class that derives from UIViewController and implement it like this:

public class MyViewController : UIViewController {
    public MyViewController (string nibName, NSBundle bundle) : base (nibName, bundle)
    {
        // You can have as many arguments as you want, but you need to call
        // the base constructor with the provided nibName and bundle.
    }
}

Then to load your ViewController from a NIB file, you do this:

		var controller = new MyViewController ("HelloWorld", NSBundle.MainBundle, this);

This loads the user interface from the NIB. Now, to access the outlets, it is necessary to inform the runtime that we want to access them. To do this, the UIViewController subclass needs to declare the properties and annotate them with the [Connect] attribute. Like this:

[Connect]
UITextField UserName { 
    get {
        return (UITextField) GetNativeField ("UserName");
    }
    set {
        SetNativeField ("UserName", value);
    }
}

The property implementation is the one that actually fetches and stores the value for the actual native type.

You do not need to worry about this when using MonoDevelop and InterfaceBuilder. MonoDevelop automatically mirrors all the declared outlets with code in a partial class that is compiled as part of your project.

Selectors

A core concept of Objective-C programming are selectors. You will often come across APIs that require you to pass a selector, or expects your code to respond to a selector.

Creating new selectors in C# is very easy, you just create a new instance of the MonoTouch.ObjCRuntime.Selector class and use the result in any place in the API that requires it. For example:

    var selector_add = new Selector ("add:plus:");

To make a C# method respond to a selector call, you need to inherit from the NSObject type and decorating your C# method with the selector name using the [Export] attribute, for example:

public class MyMath : NSObject {
    [Export ("add:plus:")]
    int Add (int first, int second)
    {
         return first + second;
    }
}

Note that selector names must match exactly, including all intermediate and trailing colons (":") if present.

NSObject Constructrors

Most classes in MonoTouch that derive from NSObject will expose constructors specific to the functionality of the object, but they will also expose various constructors that are not immediately obvious.

The constructors are used as follows:

public Foo (IntPtr handle)

This constructor is used to instantiate your class when the runtime needs to map your class to an unmanaged class.   This happens when you load a XIB/NIB file.   The Objective-C runtime will have at this point created an object in the unmanaged world and this constructor will be called to initialize the managed side.  
Typically all you need to do is call the base constructor with the handle parameter and in your body do any initialization that is necessary.

public Foo ()

This is the default constructor for a class, and in MonoTouch provided classes this initializes the MonoTouch.Foundation.NSObject class and all of the classes in between and at the end chains this to Objective's C "init" method on the class.

public Foo (NSObjectFlag x)

This constructor is used to initialize the instance, but prevent the code from calling the Objective-C "init" method at the end.   You typically use this when you already have registered for initialization (when you use [Export] on your constructor) or when you have already done your initialization through another mean.

public Foo (NSCoder coder)

This constructor is provided for the cases where the object is being initialized from an NSCoding instance.   For more information see Apple's Archives and Serialization Programming Guide.

Exceptions

The MonoTouch API design does not raise Objective-C exceptions as C# exceptions.  The design enforces that no garbage be sent to the Objective-C world in the first place and that any exceptions that must be produced are produced by the binding itself before invalid data is ever passed to the Objective-C world.