Does the cats-mtl design support for-comprehensions with multiple Monad constraints? Looks like it does!

cats-mtl.scala

package cats

import cats.data.{StateT, WriterT}
import cats.implicits._

/**
 * TypeLevel is working on some alternative encodings of the Monad Transformer
 * Library (MTL). The existing solution in cats was broken for a few reasons.
 * One annoying issue made it imposisible to use for-comprehension with more
 * than one Monad$FOO constraint owing to implicit resolution ambiguities.
 *
 * I was curious if the new encoding that Edmund Nobel is working on caused
 * (https://github.com/edmundnoble/cats-mtl)
 * for-comphrenesion issues. It looks like it doesn't.
 *
 * (FYI: my implmenetation is based on this cats-mtl code and other issues
 * discussed in the cats project but doens't necessarily reflect the current or
 * final design.)
 */
object MTL {

  /**
   * MonadState
   */
  trait MTLMonadState[F[_], S] {
    val monad: Monad[F]

    def get: F[S]

    def set(s: S): F[Unit]

    def modify(f: S => S): F[Unit]
  }

  /**
   * MonadWriter
   */
  trait MTLMonadWriter[F[_], W] {
    val monad: Monad[F]

    def tell(w: W): F[Unit]
  }

  /**
   * StateT is an MTLMonadState
   */
  implicit def stateTMTLMonadState[F[_], S](
    implicit
    applicative: Applicative[F],
    stateMonad: Monad[StateT[F,S,?]]
  ) = new MTLMonadState[StateT[F,S,?], S] {

    val monad: Monad[StateT[F,S,?]] = stateMonad

    def get = StateT.get

    def set(s: S) = StateT.set(s)

    def modify(f: S => S) = StateT.modify(f)
  }

  /**
   * WriterT is a MTLMonadWriter
   */
  implicit def writerTMTLMonadWriter[F[_], W](
    implicit
    monoid: Monoid[W],
    applicative: Applicative[F],
    writerMonad: Monad[WriterT[F,W,?]]
  ) = new MTLMonadWriter[WriterT[F,W,?],W] {
    val monad = writerMonad

    def tell(w: W) = WriterT.tell(w)
  }

  /**
   * If (F,W) is a monad-writer, then (StateT[F,S,_], W) is a monad-writer
   *
   * MTLMonadWriter[F,W] => MonadWriter[StateT[F,S,?],W]
   */
  implicit def monadWriterImpliesMWState[F[_],W,S](
    implicit
    monoid: Monoid[W],
    mw: MTLMonadWriter[F,W],
    applicative: Applicative[F],
    stateMonad: Monad[StateT[F,S,?]]
  ) = new MTLMonadWriter[StateT[F,S,?],W] {

    val monad = stateMonad

    def tell(w: W) = StateT.lift(mw.tell(w))
  }

  /**
   * If (F,S) is a monad-state, then (WriterT[F,W,_], S) is a monad-state
   *
   * MTLMonadState[F,S]  => MonadState[WriterT[F,W,?],S]
   */
  implicit def monadStateImpliesMSWriter[F[_],W,S](
    implicit
    monoid: Monoid[W],
    ms: MTLMonadState[F,S],
    applicative: Applicative[F],
    writerMonad: Monad[WriterT[F,W,?]]
  ) = new MTLMonadState[WriterT[F,W,?],S] {
    val monad = writerMonad

    def get = WriterT.lift(ms.get)

    def set(s: S) = WriterT.lift(ms.set(s))

    def modify(f: S => S) = WriterT.lift(ms.modify(f))

  }

  /**
   * A small program that only requires a monad writer
   */
  def smallProgram[F[_]](
    implicit
    monad: Monad[F],
    mWriter: MTLMonadWriter[F,String]
  ): F[String] = for {
    _ <- mWriter.tell("hello")
    _ <- mWriter.tell("world")
  } yield "!"

  /**
   * a function that makes some assumption about F without ever specifying it
   *
   * Specifically let's ensure that for-comprehension still works
   */
  def program[F[_]](
    implicit
    monad: Monad[F],
    mState: MTLMonadState[F,Int],
    mWriter: MTLMonadWriter[F,String]
  ): F[Int] = for {
    x <- monad.pure(10)
    state <- mState.get
    _ <- mState.set(11)
    _ <- mWriter.tell("hi")
    s <- smallProgram[F] // <-- can use smallprogram here, has less constraints
  } yield x + state + s.length

  /**
   * let's make a concret F which is a monad stack with StateT, WriterT and Option.
   */
  type Stack1[A] = StateT[WriterT[Option,String,?],Int, A]
  type Stack2[A] = WriterT[StateT[Option,Int,?],String, A]

  /**
   * force the implicit resolution and make sure it compiles
   */
  val stack1SmallProgram = smallProgram[Stack1]
  val stack2SmallProgram = smallProgram[Stack2]
  val stack1Program = program[Stack1]
  val stack2Program = program[Stack2]

  /**
   * Let's evaluate the programs
   */
  val i1: Option[Int] = stack1Program.run(10).run.map{ case (log, (state, value)) => value }
  // i1: Option[Int] = Some(21)

  val i2: Option[Int] = stack2Program.run.run(10).map{ case (log, (state, value)) => value }
  // i2: Option[Int] = Some(21)
}