As a follow up to my last post about no free lunch I wanted to show more state monad usage and recognize that you can change function return types in the middle of a for-comprehension. This allows each function to return different types of values from the computation along the way.
In the code below, you compose a function by using the for-comprehension. The yield part of the for-comprehension is a State object. When that state object is called, it returns a tuple of (state, value). You can use ._1 or ._2 to obtain the value of interest.
The last composed function shows an example of changing the state type (instead of changing the returned value type mentioned above) in the middle of the for-comprehension using scalaz IndexedState's iPut and iModify. This allows you to change the type of state as you go to adapt to different function's API needs.
import scalaz._
import scalaz.State._
case class Session(version: Int = 0)
// Create a new output string and increment session version.
def func1(arg: String) = State((x: Session) => (x.copy(version = x.version + 1), arg + "-" + x.version))
val composedFunction1: State[Session, String] = for {
// Because we gave this val an explicit type, init does not need the type specification
_ <- init
// Increment the state and return state+value. Essentially, throw away the value.
_ <- func1("joe")
// Modify the state directly.
_ <- modify { s: Session => Session(100) }
// Increment the state and return state+value. Essentially, throw away the value.
_ <- func1("alice")
// Modify the state directly again!
_ <- put(Session(200))
// Increment the state and return state+value. Keep the value this time.
x <- func1("nathan")
} yield x // x is a State object with the type State[Session, String]
println("composedFunction1: " + composedFunction1(Session(0)))
// Now create a composition where the returned value changes. Since the returned
// value does not really matter when the state propagates (the state needs to be the
// same between for steps or have conversions available) so the returned value can
// change each step of the way.
def funcReturnInt(arg: String): State[Session, Int] =
State((x: Session) => (x.copy(version=x.version+1), arg.toInt))
def funcReturnStringNeedsInt(arg: String, i: Int): State[Session, String] =
State((x: Session) => (x.copy(version=x.version+1), arg + "- " + x.version + "." + i))
def funcReturnString(arg: String): State[Session, String] =
State((x: Session) => (x.copy(version=x.version+1), arg + "-" + x.version))
val composedFunction2 = for {
// by using put first, we effectively ignore any input value
_ <- put(Session(0))
_ <- func1("joe")
i <- funcReturnInt("30")
x <- funcReturnStringNeedsInt("alice", i)
} yield x
// Null is used because we have to call the composed function with
// an argument. null is used for illustration purposes.
println("composedFunction2: " + composedFunction2(null))
// We can also compose the two composed functions together. We explicitly
// make the type known (State[Session, String]) for illustration purposes
// but the compiler can figure it out on its own.
def composedFunction1and2 = for {
// Put the argument given by the programmer function into a state object
_ <- init[Session]
_ <- composedFunction1
x <- composedFunction2
} yield x
// Since composedFunction2 resets the state at the start of its function,
// we should expect the saem result as composedFunction2.
println("composedFunction1and2: " + composedFunction1and2(Session(0)))
/** Prints:
composedFunction1: (Session(201),nathan-200)
composedFunction2: (Session(3),alice- 2.30)
composedFunction1and2: (Session(3),alice- 2.30)
**/
// Composed a sequence using for-comprehension but use the values
// returned from the function to do something in the for.
val composedFunction3 = for {
_ <- init[Session]
// The function returns a state object in v but due to the
// way flatMap and map are defined on State objects, the v
// will represent the value when specified on the rhs
v <- funcReturnString("session")
_ = println("Value returned was: " + v)
} yield v // this is the composed function
println("composedFunction3: " + composedFunction3(Session(10)))
/** Prints:
Value returned was: session-10
composedFunction3: (Session(11),session-10)
**/
// Compose a function that grabs the state and makes it explicitly available
// on the rhs of the for-expression line.
val composedFunction4 = for {
_ <- init[Session]
_ <- funcReturnString("session")
// By calling get (which returns a function s => (s, s)), the state because
// accessible as a value. Get is the same as init, both
// put the current state into both the state and value of the returned tuple.
s <- get[Session]
// s maps its input value into both the state and value (the state is now a value) so when its pulled apart
// by map/flatMap, the value is the state. There are other ways to stick the state as a value.
_ = println("This should print the state, because the state was turned into a value: " + s)
} yield s // this is the composed function
println("composedFunction4: " + composedFunction4(Session(10)))
/** Prints:
This should print the state, because the state was turned into a value: Session(11)
composedFunction4: (Session(11),Session(11))
**/
val composedFunction5 = for {
_ <- init[Session]
// We can use gets to generate a value anytime we want without touching the
// the state. This allows us to breakup large functions and use
// the same state across them all in the for-comprehension. This particular
// function takes the current state and returns the version number.
s <- gets{ s: Session => s.version }
_ = println("This should print the original input state's version value. The new value is in the value position: " + s)
} yield s
println("composedFunction5: " + composedFunction5(Session(10)))
/** Prints:
This should print the original input state's version value. The new value is in the value position: 10
composedFunction5: (Session(10),10)
**/
// Don't change the state, but return a new string based on the state.
def funcUsesIntState: State[Int, String] = State( (x: Int) => (x+42, "blah-"+x))
val composedFunction6 = for {
// Put the Session value into a State
_ <- init[Session]
// Now like composedFunction5, keep the current state unchanged, but return a value
// derived from that state.
v <- gets{ s: Session => s.version }
// At this point we have the version number in the value position. So lets change
// the type of the state on the fly and for fun, make that version number the new
// state. This shows that we can change the type of state on the fly.
_ <- iPut[Session, Int](v)
// Now the state is only composed of an Int not the Session object. In this example
// we are ignoring the actual value returned as well. This func does change the state
// by adding 42 to it.
_ <- funcUsesIntState
// We can also change the state type inline with a computation to derive a new state.
// Here we create a new state type that is a Map[String, Int].
_ <- iModify { x: Int => Map("version" -> x) }
// Just so we can return a value, we create one here using a very shortened
// syntax that is not obvious at first glance. We are actually defining a function
// that takes a state of the right type, and returns a String value.
s <- State((m: Map[String, Int]) => (m, "last_value-" + m("version")))
} yield s
println("composedFunction6: " + composedFunction6(Session(10)))
/** Prints:
composedFunction6: (Map(version -> 52),last_value-52)
**/