How to handle `Reader` monad and `Try`?

Question

I'm reading this great article about dependency injection in scala with Reader monad.

The original example is working well, but I did a little bit change on the return types of the UserRepository.get/find. It was User, but I changed it to Try[User].

Then the code won't be compiled, I had tries many times, but still without lucky.

import scala.util.Try
import scalaz.Reader

case class User(email: String, supervisorId: Int, firstName: String, lastName: String)

trait UserRepository {
  def get(id: Int): Try[User]

  def find(username: String): Try[User]
}

trait Users {

  def getUser(id: Int) = Reader((userRepository: UserRepository) =>
    userRepository.get(id)
  )

  def findUser(username: String) = Reader((userRepository: UserRepository) =>
    userRepository.find(username)
  )
}

object UserInfo extends Users {

  def userEmail(id: Int) = {
    getUser(id) map (ut => ut.map(_.email))
  }

  def userInfo(username: String) =
    for {
      userTry <- findUser(username)
      user <- userTry     // !!!!!!!! compilation error
      bossTry <- getUser(user.supervisorId)
      boss <- bossTry     // !!!!!!!! compilation error
    } yield Map(
      "fullName" -> s"${user.firstName} ${user.lastName}",
      "email" -> s"${user.email}",
      "boss" -> s"${boss.firstName} ${boss.lastName}"
    )
}

The compilation error is:

Error:(34, 12) type mismatch;
 found   : scala.util.Try[Nothing]
 required: scalaz.Kleisli[scalaz.Id.Id,?,?]
      user <- userTry
           ^

and

Error:(36, 12) type mismatch;
 found   : scala.util.Try[scala.collection.immutable.Map[String,String]]
 required: scalaz.Kleisli[scalaz.Id.Id,?,?]
      boss <- bossTry
           ^

I read the document of Kleisli.flatMap (The return type of findUser and getUser is Kleisli), it requires the parameter type is:

B => Kleisli[M, A, C]

Since a Try won't be a Kleisli, there are such errors.

I'm not sure how to handle it. Can I use scala.util.Try here? How can I turn it to a KLeisli type? How can I make this example work?

Answer 1

You can use the ReaderT monad transformer to compose the Reader monad and the Try monad into a single monad that you can use a for-comprehension on, etc.

ReaderT is just a type alias for Kleisli, and you can use Kleisli.kleisli instead of Reader.apply to construct your Reader-y computations. Note that you need scalaz-contrib for the monad instance for Try (or you can write your own—it's pretty straightforward).

import scala.util.Try
import scalaz._, Scalaz._
import scalaz.contrib.std.utilTry._

case class User(
  email: String,
  supervisorId: Int,
  firstName: String,
  lastName: String
)

trait UserRepository {
  def get(id: Int): Try[User]

  def find(username: String): Try[User]
}

trait Users {
  def getUser(id: Int): ReaderT[Try, UserRepository, User] =
    Kleisli.kleisli(_.get(id))

  def findUser(username: String): ReaderT[Try, UserRepository, User] =
    Kleisli.kleisli(_.find(username))
}

Now that that's done, UserInfo is much simpler (and it compiles now, too!):

object UserInfo extends Users {
  def userEmail(id: Int) = getUser(id).map(_.email)

  def userInfo(
    username: String
  ): ReaderT[Try, UserRepository, Map[String, String]] =
    for {
      user <- findUser(username)
      boss <- getUser(user.supervisorId)
    } yield Map(
      "fullName" -> s"${user.firstName} ${user.lastName}",
      "email" -> s"${user.email}",
      "boss" -> s"${boss.firstName} ${boss.lastName}"
    )
}

We can show it works:

import scala.util.{ Failure, Success }

val repo = new UserRepository {
  val bar = User("[email protected]", 0, "Bar", "McFoo")
  val foo = User("[email protected]", 0, "Foo", "McBar")

  def get(id: Int) = id match {
    case 0 => Success(bar)
    case 1 => Success(foo)
    case i => Failure(new Exception(s"No user with id $i"))
  }

  def find(username: String) = username match {
    case "bar" => Success(bar)
    case "foo" => Success(foo)
    case other => Failure(new Exception(s"No user with name $other"))
  }
}

And then:

UserInfo.userInfo("foo").run(repo).foreach(println)
Map(fullName -> Foo McBar, email -> [email protected], boss -> Bar McFoo)

Exactly the same way you'd run a Reader, but you get a Try at the end.