Scala: expanded syntax of trampolining function breaks tail recursion - scala

This is an exercise in "Functional Programming in Scala," chapter 13, to implement a trampoline for interpreting tail-recursive functions.
runTrampoline2 is not tail-recursive, and overflows the stack with my test inputs. Furthermore, the tailrec annotation gives a compiler error for runTrampoline2. runTrampoline is tail-recursive and passes the annotation's compile-time check.
My best bet is that the difference between these two trampolines lies in the way a val captures, or doesn't capture, a Unit, like here:
scala> val foo = println("abc")
val foo = println("abc")
abc
foo: Unit = ()
scala> foo
foo
scala> val bar: Int = {println("xyz"); 5}
val bar: Int = {println("xyz"); 5}
xyz
bar: Int = 5
scala> bar
bar
res8: Int = 5
Otherwise these two trampolines look identical to me. I'll include the code for the Free monad and the Suspend, Return, and FlatMap constructors if someone comments they are important for this question, but I don't think they are. Is runTrampoline2's tail-recursion broken by a side effect leaking out of the vals? Thanks!
#annotation.tailrec
def runTrampoline[A](tra: Free[Function0,A]): A =
tra match {
// case Return(A)
case Return(a1) => a1
// case Suspend(Function0[A])
case Suspend(function0A1) => function0A1()
// case FlatMap(Free[Function0[_],A], A=>Free[Function0,B]]
case FlatMap(free1, aFree2) => free1 match {
// case Return(A)
case Return(a2) => runTrampoline(aFree2(a2))
// case Suspend(Function0[A])
case Suspend(function0A) => runTrampoline(aFree2(function0A()))
// case FlatMap(Free[Function0[_],A], A=>Free[Function0,B]]
case FlatMap(a0,g) =>
runTrampoline {
a0 flatMap { a0 => g(a0) flatMap aFree2 }
}
}
}
//#annotation.tailrec
def runTrampoline2[A](tra: Free[Function0,A]): A =
tra match {
// case Return(A)
case Return(a1) => a1
// case Suspend(Function0[A])
case Suspend(function0A1) => {
val a1: A = function0A1()
a1
}
// case FlatMap(Free[Function0[_],A], A=>Free[Function0,A]]
case FlatMap(free1, aFree2) => free1 match {
// case Return(A)
case Return(a2) => {
val free2: Free[Function0,A] = aFree2(a2)
val a3: A = runTrampoline2(free2)
a3
}
// case Suspend(Function0[A])
case Suspend(function0A) => {
val a2: A = function0A()
val free2: Free[Function0,A] = aFree2(a2)
val a3: A = runTrampoline2(free2)
a3
}
// case FlatMap(Free[Function0[_],A], A=>Free[Function0,B]]
case FlatMap(a0,g) =>
runTrampoline2 {
a0 flatMap { a0 => g(a0) flatMap aFree2 }
}
}
}
I asked a similar question a month ago, about type annotations breaking tail-recursion: Scala: type annotations make tail recursion check fail
Solved by Aivean. Here is a corrected version of the trampoline. Each recursive call is at the very end of the case containing it.
#annotation.tailrec
def runTrampoline3[A](tra: Free[Function0,A]): A =
tra match {
case Return(a1) => a1
case Suspend(function0A1) => {
val a1 = function0A1()
a1
}
case FlatMap(free1, aFree2) => free1 match {
case Return(a2) => {
val free2 = aFree2(a2)
runTrampoline3(free2)
}
case Suspend(function0A) => {
val a2 = function0A()
val free2 = aFree2(a2)
runTrampoline3(free2)
}
case FlatMap(a0,g) =>
runTrampoline3 {
a0 flatMap { a0 => g(a0) flatMap aFree2 }
}
}
}

It appears that Scala compiler recognizes tail recursion only when call to itself is actually the last operation in the function.
I decompiled two different examples to check this.
Tail recursion:
scala:
def rec:Int = rec
java:
public final class _$$anon$1 {
private int rec() {
while (true) {}
}
}
No tail recursion:
scala:
def rec:Int = {
val i = rec
i
}
java:
public final class _$$anon$1 {
private int rec() {
final int i = this.rec();
return i;
}
}

Related

How should I get B form A => B

I'm new to Scala, and I'm running into this strange situation.
def bar[A, B](implicit foo: A => B): B = {
// do something
foo
}
And then I got error like
require B but found A => B
How should I get B form A => B
Here's the reason why I did this, I have two functions:
def funcA: String = {
def getStrA: String = "A"
// then there's the same operation in both functions
Try{ } match {
case Success(_) => getStrA
case Failure(_) => // exactlly same error handler in both function
}
}
def funcB: Int = {
def doSomething(x: Int): Int = {
// do something
x / 1
}
val x = 1
Try{ } match {
case Success(_) => doSomething(1)
case Failure(_) => // exactlly same error handler in both function
}
}
Here's what I want to achieve
def funcA: String = {
implicit def getStrA: String = "A"
bar
}
def funcB: Int = {
val x = 1
implicit def doSomething(x: Int): Int = {
// do something
x / 1
}
bar
}
def bar[A, B](implicit foo: A => B): B = {
Try{ } match {
case Success(_) => foo
case Failure(_) => // exactlly same error handler in both function
}
}
You have a conversion from A to B. You need to return B. The only way to do this is to pass A into the function. This signature has an implied assumption that you have some valid A value (most likely hardcoded) that you will always use here.
def bar[A, B](implicit foo: A => B): B = {
val a: A = ... // hmm...
foo(a)
}
Considering, that A is parametric, then you are either missing some information, or this A is impossible to create (it cannot be null because not all types can take null as a value), so you might need to throw exception in such case. Probably you are either missing some A provider or you should always fail this operation.
UPDATE:
There is no need for using implicits at all in your code:
def bar[B](f: onSuccess: A => B) =
Try{ some operations } match {
case Success(value) => onSuccess(value)
case Failure(_) => // error handler
}
def funcA = bar(_ => "A")
def funcB = bar(_ => 1)

How do I call a method that only exists on one of the 2 types in an Either?

I have an array of objects of type Either[A, B]. If I know for a particular element whether it is an A or a B, how do I call a method on it that only exists on one of the 2 types. For example:
import scala.util.Random
object EitherTest extends App {
def newObj(x: Int): Either[A,B] = {
if (x == 0)
Left(new A())
else
Right(new B())
}
val random = new Random()
val randomArray = (0 until 10).map(_ => random.nextInt(2))
val eitherArray = randomArray.map(newObj)
(0 until 10).foreach(x => randomArray(x) match {
case 0 => eitherArray(x).aMethod()
case 1 => eitherArray(x).bMethod()
case _ => println("Error!")
})
}
class A {
def aMethod() = println("A")
}
class B {
def bMethod() = println("B")
}
When I compile this code, the lines
case 0 => eitherArray(x).aMethod()
case 1 => eitherArray(x).bMethod()
both have the error "value aMethod is not a member of Either[A,B]". How can I solve this?
I don't know why fold doesn't get the respect it deserves. It can be so useful.
eitherArray.foreach(_.fold(_.aMethod(), _.bMethod()))
Well, you can do it if you exctract the logic to another method, and do some pattern matching over the value Either, then check if it is Right or Left, and that's it!
object HelloWorld {
import scala.util.Random
def main(args: Array[String]) {
val random = new Random()
val randomArray = (0 until 10).map(_ => random.nextInt(2))
val eitherArray = randomArray.map(EitherTest.newObj)
(0 until 10).foreach(x => randomArray(x) match {
case 0 => EitherTest.callmethod(eitherArray(x))
case 1 => EitherTest.callmethod(eitherArray(x))
case _ => println("Error!")
})
println("Hello, world!")
}
}
class EitherTest
object EitherTest {
def callmethod(ei : Either[A,B]) = {
ei match {
case Left(a) => a.aMethod()
case Right(b) => b.bMethod()
}
}
def newObj(x: Int): Either[A,B] = {
if (x == 0)
Left(new A())
else
Right(new B())
}
}
class A {
def aMethod() = println("A")
}
class B {
def bMethod() = println("B")
}
Will print for you, for one random example:
A
B
A
B
A
A
A
B
B
B
Hello, world!
Basically, the way you do with Either is projections: Either.left gives you the projection of the left type, and Either.right gives you that of the right.
The projections are somewhat similar to options, in that they can be empty (if your Either is a Right, then the left projection is empty and vice versa), and you can use the usual monadic transformations with them, like map, flatMap, foreach, getOrElse etc.
Your example, could look like this:
randomArray.foreach { either =>
either.left.foreach(_.aMethod)
either.right.foreach(_.bMethod)
}
You could also use pattern-matching instead, that's less general, but, perhaps looks a bit clearer in this case:
randomArray.foreach {
case Left(a) => a.aMethod
case Right(b) => b.bMethod
}

Convert match expressions so I can combine them using combinators like orElse

I currently have something that looks like:
Class UserLocation {
def handleUser1(user: User): Unit = user.location match {
case ... => ..
case ... => ..
case _ => ..
}
}
Class UserLocation2 {
def handleUser2(user: User): Unit = user.location match {
case ... => ..
case ... => ..
case _ => ..
}
}
Class UserLocation3 {
def handleUser3(user: User): Unit = user.location match {
case ... => ..
case ... => ..
case _ => ..
}
}
How can I convert the above to a partial function so I can do something like:
def handleUser(user: User): Unit = handleUser1(user) orElse handleUser2(user) orElse handleUser3(user)
There are a few problems with the your layout and your goal. For one thing, all the handleUser methods are hidden in their own class so the goal, as stated, is impossible because they aren't all in scope at the same time. They might be pulled into scope by making the classes implicit, but for that each class needs a constructor parameter.
So here's one possible solution achieved by dropping all the individual classes.
case class User(location: String)
val handleUser1: PartialFunction[User,Unit] = {
case User("Bern") => println("there")
case User("NYC") => println("here")
}
val handleUser2: PartialFunction[User,Unit] = {
case User("Spain") => println("there")
case User("USA") => println("here")
}
val handleUser3: PartialFunction[User,Unit] = {
case User("moon") => println("far")
case User("earth") => println("near")
case User(_) => println("unknown") // the only default
}
val handleUser = handleUser1 orElse handleUser2 orElse handleUser3
handleUser(User("Bern")) // "there"
handleUser(User("moon")) // "far"
handleUser(User("Boon")) // "unknown"
You can define the functions as PartialFunctions and then compose them using orElse:
val f :PartialFunction[Int, Int] = {
case 1 => 1
}
val f2 :PartialFunction[Int, Int] = {
case 2 => 2
}
val f3 :PartialFunction[Int, Int] = {
case 3 => 3
}
def g = f orElse f2 orElse f3
g(1) // 1
g(2) // 2
g(3) // 3
g(4) // scala.MatchError
P.S
#Suma - Thanks for the input.

Can we have an array of by-name-parameter functions?

In Scala we have a by-name-parameters where we can write
def foo[T](f: => T):T = {
f // invokes f
}
// use as:
foo(println("hello"))
I now want to do the same with an array of methods, that is I want to use them as:
def foo[T](f:Array[ => T]):T = { // does not work
f(0) // invokes f(0) // does not work
}
foo(println("hi"), println("hello")) // does not work
Is there any way to do what I want? The best I have come up with is:
def foo[T](f:() => T *):T = {
f(0)() // invokes f(0)
}
// use as:
foo(() => println("hi"), () => println("hello"))
or
def foo[T](f:Array[() => T]):T = {
f(0)() // invokes f(0)
}
// use as:
foo(Array(() => println("hi"), () => println("hello")))
EDIT: The proposed SIP-24 is not very useful as pointed out by Seth Tisue in a comment to this answer.
An example where this will be problematic is the following code of a utility function trycatch:
type unitToT[T] = ()=>T
def trycatch[T](list:unitToT[T] *):T = list.size match {
case i if i > 1 =>
try list.head()
catch { case t:Any => trycatch(list.tail: _*) }
case 1 => list(0)()
case _ => throw new Exception("call list must be non-empty")
}
Here trycatch takes a list of methods of type ()=>T and applies each element successively until it succeeds or the end is reached.
Now suppose I have two methods:
def getYahooRate(currencyA:String, currencyB:String):Double = ???
and
def getGoogleRate(currencyA:String, currencyB:String):Double = ???
that convert one unit of currencyA to currencyB and output Double.
I use trycatch as:
val usdEuroRate = trycatch(() => getYahooRate("USD", "EUR"),
() => getGoogleRate("USD", "EUR"))
I would have preferred:
val usdEuroRate = trycatch(getYahooRate("USD", "EUR"),
getGoogleRate("USD", "EUR")) // does not work
In the example above, I would like getGoogleRate("USD", "EUR") to be invoked only if getYahooRate("USD", "EUR") throws an exception. This is not the intended behavior of SIP-24.
Here is a solution, although with a few restrictions compared to direct call-by-name:
import scala.util.control.NonFatal
object Main extends App {
implicit class Attempt[+A](f: => A) {
def apply(): A = f
}
def tryCatch[T](attempts: Attempt[T]*): T = attempts.toList match {
case a :: b :: rest =>
try a()
catch {
case NonFatal(e) =>
tryCatch(b :: rest: _*)
}
case a :: Nil =>
a()
case Nil => throw new Exception("call list must be non-empty")
}
def a = println("Hi")
def b: Unit = sys.error("one")
def c = println("bye")
tryCatch(a, b, c)
def d: Int = sys.error("two")
def e = { println("here"); 45 }
def f = println("not here")
val result = tryCatch(d, e, f)
println("Result is " + result)
}
The restrictions are:
Using a block as an argument won't work; only the last expression of the block will be wrapped in an Attempt.
If the expression is of type Nothing (e.g., if b and d weren't annotated), the conversion to Attempt is not inserted since Nothing is a subtype of every type, including Attempt. Presumably the same would apply for an expression of type Null.
As of Scala 2.11.7, the answer is no. However, there is SIP-24, so in some future version your f: => T* version may be possible.

Scala: better nested multiple condition check

Lately, I frequently end up writing code like that:
def doSomethingWithLotsOfConditions(arg1, arg2, arg3...) {
arg1.get(arg2) match {
case Some(value1) =>
arg3.get(value1) match {
case Some(value2) =>
arg4.get(arg5, value2) match {
case Some(value3) =>
finallyDoSomethingInside(value3)
case None =>
log("Some excuse for being unable to work with arg4/arg5...")
}
case None =>
log("Some excuse for being unable to work with arg3")
}
case None =>
log("Some excuse for being unable to work with arg1/arg2")
}
}
A somewhat related question seems to heavily advocate for such usage of nested match, although, from my point of view, it hardly seems readable, concise or easy-to-understand: (1) it kind of splits the check itself and its aftermath, (2) it makes code uncontrollably nested without any real rationale for nesting. In these particular cases, I would be glad to structure the code something in lines of:
def doSomethingWithLotsOfConditions(arg1, arg2, arg3...) {
// Step 1
val value1Opt = arg1.get(arg2)
if (value1Opt.isEmpty) {
log("Some excuse for being unable to work with arg1/arg2")
return
}
val value1 = value1Opt.get
// Step 2
val value2Opt = arg3.get(value1)
if (value2Opt.isEmpty) {
log("Some excuse for being unable to work with arg3")
return
}
val value2 = value2Opt.get
// Step 3
val value3Opt = arg4.get(arg5, value2)
if (value3Opt.isEmpty) {
log("Some excuse for being unable to work with arg4/arg5...")
return
}
val value3 = value3Opt.get
// All checked - we're free to act!
finallyDoSomethingInside(value3)
}
However, that pattern (i.e. valueXOpt = (...).get => check isEmpty => value = valueXOpt.get) looks really ugly and is also definitely too verbose. Hell, even Java version would look more concise:
Value1Type value1 = arg1.get(arg2);
if (value1 != null) {
log("Some excuse for being unable to work with arg1/arg2");
return;
}
Is there a better, cleaner alternative, i.e. for getting the value and specifying alternative short escape route (that log a line + return), without going nested with matches?
How about this?
object Options{
implicit class OptionLog[T](val option:Option[T]) extends AnyVal{
def ifNone(body: =>Unit):Option[T] = option.orElse {
body
option
}
}
}
import Options._
def something(arg1:Option[Int], arg2:Option[String], arg3:Option[Long], arg4:Option[Any]){
for{
val1 <- arg1 ifNone(println("arg1 was none"))
val2 <- arg2 ifNone(println("arg2 was none"))
val3 <- arg3 ifNone(println("arg3 was none"))
}{
println(s"doing something with $val1, $val2, $val3")
}
}
Then ...
scala> something(Some(3), Some("hello"), None, Some("blah"))
arg3 was none
scala> something(Some(3), Some("hello"), Some(10l), Some("blah"))
doing something with 3, hello, 10
Maybe you mean, for a condition x:
scala> def f(x: Option[Int]): Int = x orElse { println("nope"); return -1 } map (_ + 1) getOrElse -2
f: (x: Option[Int])Int
scala> f(Some(5))
res3: Int = 6
scala> f(None)
nope
res4: Int = -1
or even
scala> def f(x: Option[Int], y: Option[Int]): Int = (for (i <- x orElse { println("nope"); return -1 }; j <- y orElse { println("gah!"); return -2 }) yield i + j) getOrElse -3
f: (x: Option[Int], y: Option[Int])Int
scala> f(Some(5), None)
gah!
res5: Int = -2
Sorry if I'm oversimplifying.
Don't you want to use the map method?
def doSomethingWithLotsOfConditions(arg1, arg2, arg3...) =
arg1.get(arg2).map(value1 =>
arg3.get(value1).map(value2 =>
arg4.get(arg5, value2).map(value3 =>
finallyDoSomethingInside(value3)).
getOrElse(log("Some excuse for being unable to work with arg4/arg5"))).
getOrElse(log("Some excuse for being unable to work with arg3"))).
getOrElse(log("Some excuse for being unable to work with arg1/arg2"))
It is still nested but it's at least more elegant-looking than your pattern matching above.
Or you can implement your own Functor for this one:
trait Foo[+A] {
def fmap[B](f: A => B): Foo[B]
}
case class Bar[A](value: A) {
def fmap[B](f: A => B): Foo[B] = Bar(f(value))
}
case object Error[Nothing](message: String) {
def fmap[B](f: Nothing => B) = Error(message)
}
def doSomethingWithLotsOfConditions(arg1, arg2, arg3, arg4, arg5) =
arg1.get(arg2).fmap(value1 =>
arg3.get(value1).fmap(value2 =>
arg4.get(arg5, value2).fmap(value3 =>
finallyDoSomethingInside))
def processResult = doSomethingWithLotsOfConditions(...) match {
case Bar(a) => doSomething
case Error(message) => log(message)
}
This code assumes that arg.get returns a Foo.