free-control

What is it?

This project provides control structures for the free monad in Scala. See below for usage and examples.

Usage

The control flow algebra can be composed into an existing free algebra as described in cats documentation (see section called Composing Free monads ADTs):

For example, assuming an existing MyAlgebra and MyInterpreter:

import control.free._

type MyApp[A] = ControlFlowAppp[MyAlegbra, A]

def program(implicit A: MyAlegbra[MyApp], C: ControlFlow[MyApp]): Free[MyApp, ???] = {
  import A._, C._
  
  // your program goes here
}

// The control flow interpreter wraps the other interpreter(s). This is necessary for implementing some control flow features.
val interpreter: MyApp ~> Id = ControlFlowInterpreter(MyInterpreter)

// Run it!
program.foldMap(interpreter)

Features

Here is a tour of the supported features. All of this code (and more) is tested in ControlFlowSpec.

For the below examples, assume we have a Counter free monad with the following algebra:

sealed trait CounterA[A]
final case class Set(n: Int) extends CounterA[Unit]  // Sets the current value
final case class Add(n: Int) extends CounterA[Unit]  // Adds a value to the currently stored value
final case object Get extends CounterA[Int]          // Gets the currently stored value
final case object Reset extends CounterA[Unit]       // Sets the current value to 0

Note there is state associated with this algebra: it stores a single Int value.

Now let's see what we can do with our control flow structures.

While loops

Loops while a condition is true.

for {
  _ <- whileF(get.map(_ < 100)) {
    add(1)
  }
  res <- get
} yield res
// returns 100

Repeat loops

Repeats an action n times.

for {
  _ <- repeatF(constF(100)) {
    add(1)
  }
  res <- get
} yield res
// returns 100

Conditionals

Conditionals are specified by calling condF with multiple cases, followed by the otherwise case. The first case which evaluates to true is executed, and no further cases are executed.

for {
  _ <- set(5)
  res <- condF(
    get.map(_ < 2)  -> constF("less than 2"),
    get.map(_ < 5)  -> constF("less than 5"),
    get.map(_ < 10) -> constF("less than 10"))(
    otherwise        = constF("greater than or equal to 10")
} yield res
// returns "less than 10"

Functions

We can put some of the above concepts together and define a function for multiplication.

for {
  mult <- function1F[Int, Unit] { x =>
    for {
      a <- get
      _ <- condF(
        constF(x == 0) -> reset,
        constF(x == 1) -> stayF)(
        otherwise = repeatF(constF(x - 1))(add(a)))
    } yield ()
  }
  _ <- set(5)
  _ <- mult(3)
  res <- get
} yield res
// returns 15

Scope

If your free monad contains state (as is the case with our counter), we can control state changes with scope. Note when you push a scope, that scope has access to state in its parent scope, but any state changes are local to the child scope.

for {
  _ <- set(1)
  _ <- pushScopeF
  _   <- add(42)
  a   <- get
  _ <- popScopeF
  b <- get
} yield (a, b)
// returns (43, 1)

Mutable Variables

Sometimes scope is not sufficient to control state, and we want to store things in mutable variables. Variables are named, and can be set and read.

for {
  _ <- setVarF("v1", constF(3))
  _ <- setVarF("v2", constF(5))
  res1 <- getVarF[Int]("v1")
  res2 <- getVarF[Int]("v2")
  res3 <- getVarF[Int]("bogus")
} yield (res1, res2, res3)
// returns (Some(3), Some(5), None)

Goto

What kind of control flow would we have without goto? This is just for fun, I don't recommend you use this (really this goes for the entire project, but more so goto).

for {
  _ <- set(100)
  _ <- labelF("lbl")
  _ <- add(-1)
  _ <- condF(
    get.map(_ > 10) -> gotoF("lbl"))(
    otherwise = stayF)
  res <- get
} yield res
// returns 10

rtitle/free-control