More Readable Code in Scala

23 May 2018, 16:33

Scala / clean code

As a Java developer, I ‘ve always had argument with my colleagues to use Scala instead Java. They complained that Scala is a complex language and Java is a more readable and simple one so I never had the opportunity to use Scala well. Fortunately, recently I ‘ve just joined to the startup company called snapptrip.com, and we use Scala for all the backend services. A few days ago, I was working on a peace of code in Scala to register a new user into the database. The code was fairly simple. I had UserRepository object containing CRUD operarions:

object UserRepository {

  def insert(user: User): Future[Int] = {
    db.run( users += user )
  }

  def getByUsername(username: String) : Future[Option[User]] {
   db.run(
     users.filter( users => users.username === username ).
           take(1).
           result.headOption
   )
  }
}

And I had a bunch of services to encapsulate business logic. For example, for registering a user you probably check username uniqueness and password strengths. This is the code that I wrote first:

object UserService {
    def register(user: RegisterRequest) : Future[RegisterResponse] = {
        if(user.username.isEmpty){
          return Future.successful(RegisterResponse(errorMsg = "Please give a username"))
        }
        UserRepository.getByUsername(user.username).flatMap {
          case Some(someUser) => Future.successful(RegisterResponse(errorMsg = "Username already exists"))
          case None => if (user.password.length < 6) {
            Future.successful(RegisterResponse(errorMsg = "Password is too low"))
          } else {
            UserRepository.insert(User(None,user.username,user.password)).map { newId =>
              RegisterResponse(success = true,userID = newId)
            }.recover {
              case _ => RegisterResponse(errorMsg = "Unexpected error has occurred. Please try again later")
            }
          }
        }
    }
}

As a guy that always cares a lot to clean and maintainable code, this code absolutely is not my cup of tea. The code was really complex for such a simple thing. Actully register method should do two things:

a bunch of validation
save to database if validation succeeded

The picture below depicts the idea:

So I decided to refactor this code, and the code transformed to this one:

 def register(newUser: RegisterRequest) : Future[Response] = {
     Validate(newUser) >
          usernameShouldNotBeEmpty >
          passwordShouldNotBeTooShort >
          usernameShouldNotBeTaken andThen
          saveToDatabase
 }

The above code is much more readable and more maintainable, doesn’t it? Moreover, if functional programming is all about using functions and composing them together, I used this idea well. The above code works this way:

‘newUser’ parameter passes to the ‘usernameShouldNotBeEmpty’ function and if the username is empty then ‘register’ will be return early with some type of Response without calling the other functions. On the other hand, if the username is not empty, ‘newUser’ will be passed to the second function. And this flow goes until all validation functions have been executed. If the last function satisfy the condition, then the actual action will be executed, in this case, saving to the database.

Also pay attention to the return type, that I changed from Future[RegisterResponse] to Future[Response]. This way all of the functions can return any type as long as they are subclasses of Response trait:

trait Response
case class ErrorResponse(message:String, success:Boolean=false) extends Response
case class SuccessResponse(message:String, success:Boolean=true) extends Response
case class RegisterUserResponse(userId: Long, success:Boolean=true) extends Response

And all of this workflow managed by Validate class. It defines like this:

class Validate[IN,OUT <: Response](input: IN,
                                   valFunctions: List[ValidateFunc[IN,OUT]]=List.empty) {

  def > (f: ValidateFunc[IN,OUT]) = new Validate(input,valFunctions :+ f)

  def andThen(action: IN => Future[OUT]): Future[OUT] = {
    val it = valFunctions.iterator
    while (it.hasNext){
      val nextFunc = it.next
      nextFunc(input) match {
        case Some(result) => return Future.successful(result)
        case _ => // Do nothing
      }
    }

    action(input)
  }

}

All validation functions are of type ValidateFunc which is basically a function of type IN => Option[OUT]. If something goes wrong, validation functions should return Some(OUT), otherwise None. And OUT is restricted to all subtypes of Response by:

OUT <: Response

As you can see the Validate method > does nothing except appending validations function to a list. All the logic of pipeline execution is in andThen method. It executes all the validation functions sequentially and executes the main function, if and only if all of them return None. You can find the full source code here.

Last world

I wanted to abstract validation in my code so I wrote Validate class for myself, and of course, the next move for me, is to read Cats Validated.

Last world

Share!