I would like to know how to best possible address the following issue:
I have a single ViewController. Its view contains a great number of complex subviews (subclass of UIView). Due to the complexity some of these UIViews initialise their own UIGestureRecognisers and implement the according target actions. As I want to coordinate the gestures of various subviews I have to set the single once ViewController as the gesture's delegate.
There are various possibilities for that.
1) Initialize ALL gestures in the the viewController (this will lead to a massive viewController)
2) defining a protocol in the UIVIews (getViewController), implemented by the ViewController
@protocol CustomViewDelegate <NSObject>
@required
- (UIViewController *)getViewController;
@end
3) customise the init method of the UIViews and using the ViewController as an option.
- (id)initWithFrame:(CGRect)frame andViewController:(UIViewController *)vc;
What is the most elegant possibility to solve this issue? Is it OK to implement target actions inside a UIView object?
Thanks for your thoughts...
If you're defining custom UIView subclasses, you can invest them with as much logic as it makes sense to store local to them, give them delegate protocols to pass anything else up and, as long as you expose the delegate as an IBOutlet, you can wire up your view controller as the relevant delegate directly in Interface Builder or the UI designer part of Xcode 4. I personally think that would be the most natural way forward, since it consolidates any view-specific logic directly in the view and lets you do the wiring up where you would normally do the wiring up.
In terms of overall design, such a scheme conforms to model-view-controller provided your views are doing only view-related logic. So, for example, if you had a custom rectangular view that can take a swipe anywhere on it to reposition a pin, and the 2d position of the pin affects some other system setting, you'd be correct to catch the gesture in the view, reposition the pin and then send updates on its position down to the delegate, which would fulfil the role of controller and push the value to any other views that are affected and out to the model.
Commenting on your suggested solutions directly:
(1) this would focus all logic into the one controller; whether it's correct from a design point-of-view depends on the extent to which you're having to interrogate your custom views (in that you don't want to end up treating them as mostly data that external actors have to know how to manipulate) and the extent to which you want to reuse logic.
(2) I'm not sure I entirely understand the suggestion — what is getViewController defined on and how does it know how to respond? If it's the UIViews themselves and the view controller has to identify itself first then I'd suggest just adopting the delegate pattern wholesale rather than specialising to views and view controllers, e.g. as you may want to build compound views that tie together logic from several subviews.
(3) as a rule of thumb, the sort of things to pass to init are those that the class actually needs to know to be able to initialise; it would probably be better to use a normal property to set the controller after the fact. Or make it an IBOutlet and wire it up so that it happens automatically via the NIB.