Skip to content

Swift Package for a Use Case centric architecture as proposed by Robert C. Martin and others.

License

Notifications You must be signed in to change notification settings

AppCron/ACInteractor

Repository files navigation

ACInteractor

License License

Swift Package for a Use Case centric architecture as proposed by Robert C. Martin and others.

Overview

ACInteractor is a Swift Package that supports a Use Case centric architecture and TDD in Swift projects. The basic idea is that one Use Case, and only one Use Case, is executed by a single class. As proposed by Robert C. Martin, these kind of classes are called Interactors.

  • Each Interactor has a Request model.
  • Each Interactor has a Response model.
  • Each Interactor has an Execute function that takes the Request model and returns the Response model.
class LoginViewController: UIViewController {
    ...
    func login() {
        let request = LoginUserInteractor.Request()
        
        request.email = "[email protected]"
        request.password = "1234"
        
        request.onComplete = { response in
            self.userLabel.text = response.username
        }
        
        request.onError = { error in
            self.displayError(error.message)
        }
        
        Logic.executor.execute(request)
    }
}

Consumers, like ViewControllers, can easily execute InteractorRequests with the help of the InteractorExecutor. Therefore an initialized instance of each Interactor has to be registered on the InteractorExecutor. Interactors should be stateless, since all Requests of a given Interactor are handled by the same instance of that Interactor.

ACInteractor adds no constraints to dependency management. It's up to you how to initialize your Interactor instances. I'd recommend Dependency Injection with a custom initializer. More details can be found at the section "Dependency Injection".

ACInteractor was built with TDD in mind. Each Interactor has a single execution function, a defined request and response, a stateless implementation and injected dependencies. This helps writing isolated Unit Tests for each Interactor. See section "Unit Testing" for more details about writing tests.

License

Apache License, Version 2.0, January 2004

Must watch

If you're new to Clean Code and TDD, Robert C. Martin's Talk at Ruby Midwest is highly recommended. It's fun to watch and worth the time. The entire architecture around ACInteractor and the goals of it are based on this talk.

Content

ACInteractor currently consists of the following files:

ACInteractor
├── LICENSE
├── Package.swift   // The Swift Package description
├── README.md
├── Sources         // The source files
└── Tests           // The unit test files

Setup

Since Swift Package is not supporting iOS Xcode projects yet, it's recommended to add the files of the Sources/ACInteractor folder directly to your project.

via Git Submodule

You can add the entire ACInteractor project as a Git Submodule to your repository.

  1. Navigate to the root directory of your Git repository in Terminal.
  2. Run the following command:
git submodule add https://github.com/AppCron/ACInteractor.git
  1. Add the files of the Sources/ACInteractor folder to your project.

via Download

Alternatively you can just download the files directly from Github and add the files of the Sources folder to your project.

as Swift Package Dependency

You can also add it as Swift Package Dependency to another Swift Package.

dependencies: [
    .package(url: "https://github.com/AppCron/ACInteractor.git", from: "1.0.0")
]

Terminal Commands

You can run the tests directly from the command line with swift test.
To create an Xcode project use swift package generate-xcodeproj.

Writing Interactors

class LoginUserInteractor: Interactor {
    class Request: InteractorRequest<Response> {...}
    class Response {...}
    
    func execute(request: Request) {...}
}

Let's write our first Interactor. It should handle a user login and accepts all logins as long as the user provides a password.

Each Interactor has to implement the Interactor protocol, which requires the Interactor to have an execute() function. Since each Interactor handles exactly one Use Case, only one execute function is necessary.

Usually, it contains two nested classes. The Request and the Response.

The Request

class LoginUserInteractor: Interactor {
    class Request: InteractorRequest<Response> {
      var email: String?
      var password: String?
    }
    ...
}

The Request contains all the parameters that are necessary to execute it. In our case the email and the password.

It has to be derived from the generic InteractorRequest<ResponseType> and should by typed with the ResponseType of the Interactor. This will help us execute the Request later. See section "The Execute Function" for details.

The Response

class LoginIntactor: Interactor {
    ...
    class Response {
    	var username: String?
    	var sessionToken: String?
	}
    ...
}

The Response is the result of the Use Case. In our example, it contains the username and a session token.

The Response can be of any type, as long as it is the same type the InteractorRequest is typed with. It can even be a common type like String or Bool.

The Execute Function

class LoginUserInteractor: Interactor {
	...
    func execute(request: Request) {
        if (request.password?.characters.count > 0) {
            
            // Create Response
            let response = Response()
            
            // Set values
            response.username = request.email?.componentsSeparatedByString("@").first
            response.sessionToken = "123456910"
            
            // Call completion handler
            request.onComplete?(response)
            
        } else {
            // Do error handling
        }
    }
}

The Execute Function is the part where the business logic of the Use Case is implemented. It takes the Request as an argument and returns the Response to the completion handler.

The onComplete closure is already defined in the InteractorRequest and already typed with the Response. That's why the Request has to be a subclass of InteractorRequest.

Registering Interactors

class Logic {
    
    static let executor = InteractorExecutor()
    
    static func registerInteractors() {
        let loginInteractor = LoginUserInteractor()
        let loginRequest = LoginUserInteractor.Request()
        
        executor.registerInteractor(loginInteractor, request: loginRequest)
    }

}

To enable consumers, like ViewControllers, to easily execute InteractorRequests it's necessary to register the corresponding Interactor first.

In our case we use a helper class called Logic. It contains a static function registerInteractors() that creates a LoginInteractor instance and registers it with the corresponding Request at the InteractorExecutor.

Besides that, the Logic contains a static property with a global Executor instance. This makes it easier for the consumer to access the given instance.

Executing Requests

class LoginViewController: UIViewController {
    ...
    func login() {
        let request = LoginUserInteractor.Request()
        
        request.email = "[email protected]"
        request.password = "1234"
        
        request.onComplete = { (response: LoginUserInteractor.Response) in
            self.userLabel.text = response.username
        }
        
        Logic.executor.execute(request)
    }
}

To execute a Request you can simply call the executeMethod() on the InteractorExecutor. Just make sure you have registered the that Interactor class with its Request on the same InteractorExecutor instance.

In our example this instance is stored in the static executor property on the Logic class, as shown above.

Error Handling

Basic error handling is already part of ACInteractor. Each InteractorRequest has an onError property that stores a closure for the error handling.

On the Interactor Implementation

class LoginUserInteractor: Interactor {
	...
    func execute(request: Request) {
        if (request.password?.characters.count > 0) {
            ...        
        } else {
            let error = InteractorError(message: "Empty password!")
            request.onError?(error)
        }
    }
}

On the Interactor all you have to do is create an instance of InteractorError, supply an error message and call the error handler with the error object.

On the Execute Call

class LoginViewController: UIViewController {
    ...
    func login() {
        let request = LoginUserInteractor.Request()
        
		...
        
        request.onError = { (error: InteractorError) in
            self.displayError(error.message)
        }
        
        Logic.executor.execute(request)
    }
}

When executing a request make sure to set a closure for error handling on its onError property.

Extended Completion Handlers

It is not necessary to use the default completion and error handlers. You can add custom completion closures, like onUpdate(UpdateResponse), or custom error closures, like onExtendedError(ExtendedError). This can be either done by adding them as properties to a specific request or by subclassing InteractorRequest.

Be aware when adding custom error handlers. ACInteractor uses the default onError closure to report internal errors, like not finding an Interactor for a given request. See section "Troubleshooting" for Details.

Asynchronous Requests

Since ACInteractor uses closures for result handling, you can easily switch between synchronous and asynchronous behavior without the need to adjust your Interactor's interface.

When making asynchronous callbacks from your Interactor, it's recommended to dispatch your onCompletion and onError closure calls to the thread the Interactor's execute() method has been called from. Whether to use background threads and when or how to dispatch back to the caller's thread is an implementation detail of your Interactor, that should be hidden from the caller. It is not the responsibility of a ViewController to dispatch your asynchronous stuff back on the main thread. Maybe it can be done with dispatch_async, maybe dispatch_sync is necessary, the ViewController can't know.

Dependency Injection

In a real world example the LoginInteractor would call a webservice to verify the login credentials and store the session token in a local database. Since we don't want all these technical details in our Interactor we encapsulate them in two separate classes.

protocol WebservicePlugin
class HttpWebservicePlugin: WebservicePlugin { }

protocol UserEntityGateway
class UserCoreDataGateway: UserEntityGateway { }

The WebservicePlugin handles the webservice calls and calls the onCompletion closure once finished.

The UserEntityGateway has functions to create new users and to save users. It is responsible for creating and saving new entities. So our LoginInteractor does not need to now how we persist data. It can be a CoreData-, a Realm- or just an In-Memory-Database.

Additionally it is useful to define a protocol for each dependency. This lets us replace them easily with mocks when writing unit tests.

On the Interactor Implementation

class LoginUserInteractor: Interactor {
	let webservicePlugin: WebservicePlugin
	let userGateway: UserEntityGateway

	init(webservicePlugin: WebservicePlugin, userGateway: UserEntityGateway) {
		self.webservicePlugin = webservicePlugin
		self.userGateway = userGateway
	}

    func execute(request: Request) {
    	...
    	self.webservicePlugin.doStuff()
    	self.userEntityGateway.doStudd()
    	...
    }
}

On the Interactor we need a custom init function that takes the dependencies as parameters and stores them in properties. We can then use them in the execute function.

Register with Dependencies

class Logic {
    
    static let executor = InteractorExecutor()
    
    static func registerInteractors() {
    	let webservicePlugin = HttpWebservicePlugin()
    	let userGateway = UserCoreDataGateway()

        let loginInteractor = LoginUserInteractor(webservicePlugin: webservicePlugin, userGateway: userGateway)
        let loginRequest = LoginUserInteractor.Request()
        
        executor.registerInteractor(loginInteractor, request: loginRequest)
    }
}

When registering the LoginInteractor we use the custom init method to supply the matching implementations. At this point you can easily replace the concrete implementations of the Plugin and the Gateway with other implementations as long as they conform to the specified protocols. For example the UserCoreDataEntityGateway could be replaced with an UserInMemoryGateway.

On the Execute Call

class LoginViewController: UIViewController {
    ...
    func login() {
        let request = LoginUserInteractor.Request()
       	...
        Logic.executor.execute(request)
    }
}

Nothing changes :) The Interactor is still executed with the same Request on the InteractorExecutor. This means you can easily refactor Interactors behind the scenes and extract technical details in EntityGateways and Plugins without breaking the API that is used by the caller.

Unit Testing

ACInteractor was build with TDD in mind. Each Interactor has a single execution function, a defined request and response, a stateless implementation and injected dependencies. This helps writing isolated Unit Tests for each Interactor.

If you use the Dependency Injection approach described above you can easily mock the dependencies in your Unit Tests. For example this means you don't need to execute a real webservice call or setup a real database to run your tests. The mocks can just return the values that support you current test. Mocks can also help you to test edge cases that would be otherwise hard to simulate, like a long taking webservice request or a full database.

Troubleshooting

  1. I can not register my Interactor at the InteractorExecutor. I get a Compiler Error.

    • Make sure the interactor implements the Interactor protocol
    • Make sure the request is a subclass of InteractorRequest<Response> and is correctly typed.
    • Make sure the request is an initialized object instance.
  2. Calling execute on the InteractorExecutor does not call the execute method of my Interactor.

    • Make sure you have registered the Interactor with the corresponding InteractoRequest by calling the registerInteractor() function on the InteractorExecutor.
    • Make sure you have called the registerInteractor() and the execute() function on the same instance of InteractorExecutor.
    • Make sure you have set an onError closure on the request. It might provide additional details in the error message about what went wrong.

Credits

Special thanks from Florian Rieger to the people that helped getting ACInteractor together.

Happy Coding 🐳