stacking StateT in scalaz

2020-04-02 07:31发布

问题:

I'm trying to understand Monad Transformers in Scala by porting some examples from this tutorial by Dan Piponi: http://blog.sigfpe.com/2006/05/grok-haskell-monad-transformers.html

I did a couple of easy ones:

import Control.Monad.State
import Control.Monad.Identity

test1 = do
    a <- get
    modify (+1)
    b <- get
    return (a,b)

test2 = do
    a <- get
    modify (++"1")
    b <- get
    return (a,b)

go1 = evalState test1 0
go2 = evalState test2 "0" 

becomes:

import scalaz._, Scalaz._

val test1 = for {
  a <- get[Int]
  _ <- modify[Int](1+)
  b <- get
} yield (a,b)

val test2 = for {
  a <- get[String]
  _ <- modify[String](_ + "1")
  b <- get
} yield (a,b)

val go1 = test1.eval(0)
val go2 = test2.eval("0")

But how the heck can I port this next example to Scala?

test3 = do
    modify (+ 1)
    lift $ modify (++ "1")
    a <- get
    b <- lift get
    return (a,b)

go3 = runIdentity $ evalStateT (evalStateT test3 0) "0"

I've gotten this far using scalaz 7.1.0-M6:

type SST[F[_],A] = StateT[F,String,A]
type IST[F[_],A] = StateT[F,Int,A]

val p1: StateT[Id,Int,Unit] = modify[Int](1+)
val p2: StateT[Id,String,Unit] = modify[String](_ + "1")
val p3: StateT[({type l[a]=StateT[Id,String,a]})#l,Int,Unit] = p2.liftM[IST]

but that's not even close yet, and may even be backwards for all I can tell.

Of course I can do this:

import scalaz.Lens._
val test3 = for {
  _ <- firstLens[Int,String] lifts (modify (1+))
  _ <- secondLens[Int,String] lifts (modify (_ + "1"))
  a <- firstLens[Int,String] lifts get
  b <- secondLens[Int,String] lifts get
} yield (a,b)

val go3 = test3.eval(0,"0")

but then I'm not using stacked StateT at all, so it doesn't answer the question.

Thanks in advance!

回答1:

The problem is that the modify you get with the usual imports is from State, and isn't going to help you with StateT.

It's a good idea to start with the Haskell type signature:

test3
  :: (MonadState [Char] m, MonadState s (t m), MonadTrans t,
      Num s) =>
     t m (s, [Char])

Which you should be able to translate into something like this:

import scalaz._, Scalaz._

def test3[M[_]: Monad](implicit
  inner: MonadState[({ type T[s, a] = StateT[M, s, a] })#T, String],
  outer: MonadState[
    ({
      type T[s, a] = StateT[({ type L[y] = StateT[M, String, y] })#L, s, a ]
    })#T,
    Int
  ],
  mt: MonadTrans[({ type L[f[_], a] = StateT[f, Int, a] })#L]
) = for {
  _ <- outer.modify(_ + 1)
  _ <- mt.liftMU(inner.modify(_ + "1"))
  a <- outer.get
  b <- mt.liftMU(inner.get)
} yield (a, b)

It's hideous, but it's a fairly straightforward rewording of the Haskell. For some reason the compiler doesn't seem to find the outer instance, though, so you have to help it a little:

def test3[M[_]: Monad](implicit
  inner: MonadState[({ type T[s, a] = StateT[M, s, a] })#T, String],
  mt: MonadTrans[({ type L[f[_], a] = StateT[f, Int, a] })#L]
) = {
  val outer =
    StateT.stateTMonadState[Int, ({ type L[y] = StateT[M, String, y] })#L]

  for {
    _ <- outer.modify(_ + 1)
    _ <- mt.liftMU(inner.modify(_ + "1"))
    a <- outer.get
    b <- mt.liftMU(inner.get)
  } yield (a, b)
}

Now you can write the following, for example:

scala> test3[Id].eval(0).eval("0")
res0: (Int, String) = (1,01)

Exactly as in the Haskell example.

Footnote

You can clean this up a bit if you're happy with committing to Id as the monad of the inner state transformer (as your comment suggests):

def test3 = {
  val mt = MonadTrans[({ type L[f[_], a] = StateT[f, Int, a] })#L]
  val outer = StateT.stateTMonadState[Int, ({ type L[y] = State[String, y] })#L]
  for {
    _ <- outer.modify(_ + 1)
    _ <- mt.liftMU(modify[String](_ + "1"))
    a <- outer.get
    b <- mt.liftMU(get[String])
  } yield (a, b)
}

It's a little less generic, but it may work for you.