Mapping over generic tuples with polymorphic functions - scala

Scala 3 provides polymorphic functions and Tuples similar to shapeless HList:
scala> 1 *: "foo" *: Tuple()
val res0: (Int, String) = (1,foo)
scala> val f: ([T] => T => Option[T]) = [T] => (v: T) => Some(v)
val f: PolyFunction{apply: [T](x$1: T): Option[T]} = <function1>
scala> res0.map(f)
val res1: Option[Int] *: Option[String] *: EmptyTuple = (Some(1),Some(foo))
How could we reimplement the following shapeless example using Scala 3 functionality?
import poly._
object choose extends (Set ~> Option) {
def apply[T](s : Set[T]) = s.headOption
}
scala> val sets = Set(1) :: Set("foo") :: HNil
sets: Set[Int] :: Set[String] :: HNil = Set(1) :: Set(foo) :: HNil
scala> val opts = sets map choose
opts: Option[Int] :: Option[String] :: HNil = Some(1) :: Some(foo) :: HNil
In other words, how could we make something like so compile
scala> val choose: ([T] => Set[T] => Option[T]) = [T] => (s: Set[T]) => s.headOption
val choose: PolyFunction{apply: [T](x$1: Set[T]): Option[T]} = <function1>
scala> val sets = Set(1) *: Set("foo") *: Tuple()
val sets: (Set[Int], Set[String]) = (Set(1),Set(foo))
scala> sets.map(choose)
1 |sets.map(choose)
| ^^^^^^
| Found: (choose : PolyFunction{apply: [T](x$1: Set[T]): Option[T]})
| Required: PolyFunction{apply: [t](x$1: t): Any}

Shapeless map is quite a bit more magical than Scala 3 tuple map, the signature of latter being:
def map[F[_]](f: [t] => (x$1: t) => F[t]): Map[Tuple, F]
Map[Tuple, F] is a special match type that is basically tuple with every argument type wrapped in F[_]. The issue is that shape, t => F[t] that prevents too much fanciness.
Except of course, F can be a match type itself:
type Choose[T] = T match {
case Set[a] => Option[a]
}
def choose[T](t: T): Choose[T] = t match
case set: Set[a] => set.headOption
// messy, but it works
#main def run =
val sets = Set(1) *: Set("foo") *: Tuple()
println(sets.map([T] => (t: T) => choose(t)))
There's currently an issue that compiler won't infer polymorphic function from methods. Match types aren't fully type-safe either, e.g. doing choose("string") will compile but throw a MatchError. I've also ran into problems with inference of a match type from polymorphic function value, hence the usage of def method.

Related

Why eta-expansion of polymorphic method does not result in polymorphic function value?

Scala 2 does not have polymorphic function values so eta-expanding polymorphic methods gives only
scala> def f[A](a: A): A = ???
def f[A](a: A): A
scala> f _
val res0: Nothing => Nothing = $Lambda$7757/1502613782#45af2c1
However Scala 3 does have polymorphic function values so why eta-expanding polymorphic methods does not give more than
scala> def f[A](a: A): A = ???
def f[A](a: A): A
scala> f
val res0: Any => Any = Lambda$7538/1430563609#4a905603
scala> val g: ([A] => A => A) = f
1 |val g: ([A] => A => A) = f
| ^
| Found: Any => Any
| Required: PolyFunction{apply: [A](x$1: A): A}

Can I solve it with Shapeless?

Suppose I have a few functions:
val f1: Int => String
val f2: (Int, Int) => String
val f3: (Int, Int, Int) => String
def fromList1(f: Int => String): List[Int] => Option[String] =
_ match {case x::_ => Some(f(x)); case _ => None}
def fromList2(f: (Int, Int) => String): List[Int] => Option[String] =
_ match {case x::y::_ => Some(f(x, y)); case _ => None}
Now I would like to write one generic fromList to work as follows:
val g1: List[Int] => String = fromList(f1) // as fromList1(f1)
val g2: List[Int] => String = fromList(f2) // as fromList2(f2)
Can I do that with shapeless ?
This may help:
import shapeless._
import syntax.std.traversable._
import shapeless.ops.traversable._
import syntax.std.function._
import ops.function._
def fromList[F, L <: HList, R](f: F)
(implicit fp: FnToProduct.Aux[F, L => R], tr: FromTraversable[L]) =
(p: List[Int]) => p.toHList[L] map f.toProduct
f.toProduct transforms regular function to function that takes HList as parameter - it requires FnToProduct implicit and actually just call it. FnToProduct.Aux is constructor (generated by macro) that creates FnToProduct from dunction F, hlist type HList and result type R. All of them are inferred from f parameter you passed.
Last one, toHList creates Some(HList) from regular List if it's possible, otherwise - None. It uses FromTraversable[L] implicit to do that, where L is already inferred from f. Shapeless2 is smart enough to recognize HList from Tuple (as there probably is implicit conversion).
Example:
scala> val f1: Int => String = _ => "a"
f1: Int => String = <function1>
scala> val f2: (Int, Int) => String = (_, _) => "a"
f2: (Int, Int) => String = <function2>
scala> val g1 = fromList(f1)
g1: List[Int] => Option[String] = <function1>
scala> g1(List(1))
res6: Option[String] = Some(a)
scala> val g2 = fromList(f2)
g2: List[Int] => Option[String] = <function1>
scala> g2(List(1, 2))
res7: Option[String] = Some(a)
scala> g2(List(1))
res8: Option[String] = None
Yes you can
import shapeless._
import shapeless.ops.traversable._
import syntax.std.traversable._
import ops.function._
def fromList[F, I <: HList, O](f: F)(implicit
ftp: FnToProduct.Aux[F, I => O],
ft: shapeless.ops.traversable.FromTraversable[I]): List[Int] => Option[O] =
{ x: List[Int] => x.toHList[I].map(ftp(f)) }
Explanation
We're using FnToProduct to transform any FunctionN to a Function1 that takes an HList as only argument.
So,
Int => String ----> Int :: HNil => String
(Int, Int) => String ----> Int :: Int :: HNil => String
...
Now that we abstracted over the arity of the input parameters for the function, we can simply convert the List[Int] to an HList that suits the transformed function's input.
In order to perform this conversion we need to a FromTraversable[I] in scope.
If everything succeeds we return and Option[O] where O is the return type of the function.
If the input List has the wrong shape, we simply fail returning None.
Usage
# val f1: Int => String = _.toString
f1: Int => String = <function1>
# val f2: (Int, Int) => String = (_, _).toString
f2: (Int, Int) => String = <function2>
# val fromList1 = fromList(f1)
fromList1: List[Int] => Option[String] = <function1>
# val fromList2 = fromList(f2)
fromList2: List[Int] => Option[String] = <function1>
# fromList1(List(1))
res22: Option[String] = Some(1)
# fromList2(List(1, 2))
res23: Option[String] = Some((1,2))
# fromList1(List())
res24: Option[String] = None

Scala match function against variable

When I'm matching value of case classes, such as:
sealed abstract class Op
case class UOp[T, K](f: T => K) extends Op
case class BOp[T, Z, K](f: (T, Z) => K) extends Op
like this:
def f(op: Op): Int =
op match
{
case BOp(g) => g(1,2)
case UOp(g) => g(0)
}
the compiler infers it as
val g: (Nothing, Nothing) => Any
val g: Nothing => Any
Why am I getting Nothing as the type? Is it because of JVM type erasure? Are there elegant ways to match functions against variables?
I came up with this "hackish" solution, maybe there are other ways or cleaner ways to do this still without relying on reflection.
Define a few partial functions which will handle various args:
scala> val f: PartialFunction[Any, String] = { case (x: Int, y: String) => y * x }
f: PartialFunction[Any,String] = <function1>
scala> val g: PartialFunction[Any, String] = { case x: Int => x.toString }
g: PartialFunction[Any,String] = <function1>
scala> def h: PartialFunction[Any, BigDecimal] = { case (a: Int, b: Double, c: Long) => BigDecimal(a) + b + c }
h: PartialFunction[Any,BigDecimal]
scala> val l: List[PartialFunction[Any, Any]] = f :: g :: h :: Nil
l: List[PartialFunction[Any,Any]] = List(<function1>, <function1>, <function1>)
Check which functions can handle different inputs:
scala> l.map(_.isDefinedAt(1))
res0: List[Boolean] = List(false, true, false)
scala> l.map(_.isDefinedAt((1, "one")))
res1: List[Boolean] = List(true, false, false)
Given input find and apply a function:
scala> def applyFunction(input: Any): Option[Any] = {
| l find (_.isDefinedAt(input)) map (_ (input))
| }
applyFunction: (input: Any)Option[Any]
scala> applyFunction(1)
res1: Option[Any] = Some(1)
scala> applyFunction((2, "one"))
res2: Option[Any] = Some(oneone)
scala> applyFunction("one")
res3: Option[Any] = None
scala> applyFunction(1, 1.1, 9L)
res10: Option[Any] = Some(11.1)
This looks quite type unsafe and there must be better ways to do this.
I think magnet pattern should handle this well in more typesafe manner.

Implicit lifting in scala

I want to implicitly convert functions from A => B to List[A] => List[B].
I wrote the following implicit definition:
implicit def lift[A, B](f: A => B): List[A] => List[B] = ...
Unfortunately, when I write the following code, implicit aren't applied:
val plusOne: (List[Int]) => List[Int] = (x: Int) => (x + 1)
If I annotate the function with explicit time, it works fine.
Why? How can I fix it?
UPDATE. It seems that the problem is specific to anonymous functions. Compare:
#Test
def localLiftingGenerics {
implicit def anyPairToList[X, Y](x: (X, Y)): List[X] => List[Y] = throw new UnsupportedOperationException
val v: List[String] => List[Int] = ("abc", 239)
}
#Test
def localLiftingFuns {
implicit def fun2ListFun[X, Y](f: X => Y): List[X] => List[Y] = throw new UnsupportedOperationException
val v: List[String] => List[Int] = ((x: String) => x.length)
}
The first one is compiled well. The second one is marked as error
According to The Scala Language Specification / Expressions / Anonymous Functions (6.23):
If the expected type of the anonymous function is of the form
scala.Functionn[S1, …, Sn, R], the expected type of e is R ...
So, the result type of the function will be inferred as List[Int] unless you separate the function definition from the function value assignment (to get rid of the expected type):
val function = (x: Int) => (x + 1)
val plusOne: (List[Int]) => List[Int] = function
or specify the function type explicitly:
val plusOne: (List[Int]) => List[Int] = ((x: Int) => (x + 1)): Int => Int
(Scala 2.9.1-1 (Java HotSpot(TM) 64-Bit Server VM, Java 1.7.0_05)
A first observation: If you duplicate fun2ListFun and rename it to, e.g.,``fun2ListFun, you'll get
found : String => <error>
required: List[String] => List[Int]
Note that implicit conversions are not applicable because they are ambiguous:
both method fun2ListFun2 of type [X, Y](f: X => Y)List[X] => List[Y]
and method fun2ListFun of type [X, Y](f: X => Y)List[X] => List[Y]
are possible conversion functions from String => <error> to List[String] => List[Int]
val v: List[String] => List[Int] = ((x: String) => x.length)
It looks as if the compiler considers both implicits as applicable.
A second observation:
Splitting
val v: List[String] => List[Int] = ((x: String) => x.length) /* Error*/
into
val f = ((x: String) => x.length)
val v: List[String] => List[Int] = f /* Works */
makes the compiler happy.
The implicit conversion for the input value compiles. So we just have a problem for the output of the anonymous function
def localLiftingFuns {
implicit def fun2ListFun[X, Y](f: X => Y): List[X] => Y = throw new UnsupportedOperationException
val v: List[String] => Int = ((x: String) => x.length)
}
A possible fix using a second implicit conversion:
def localLiftingFuns {
implicit def fun2ListFun[X, Y](f: X => List[Y]): List[X] => List[Y] = throw new UnsupportedOperationException
implicit def type2ListType[X](x:X): List[X] = throw new UnsupportedOperationException
val v: List[String] => List[Int] = ((x: String) => x.length)
}
This version compiles.
Seems the compiler has hard time figuring what is going on with the type of the function. If you will give him a little help, it would work:
scala> implicit def lift[A,B](f: A => B) = (_:List[A]).map(f)
lift: [A, B](f: (A) => B)(List[A]) => List[B]
scala> val f: List[Int] => List[Int] = ((_:Int) + 1):(Int => Int)
f: (List[Int]) => List[Int] = <function1>

How to read the class of a Scala object extending Any but not AnyRef?

I have an heterogeneous List like the following one:
val l = List(1, "One", true)
and I need to filter its objects by extracting only the ones belonging to a given Class. For this purpose I wrote a very simple method like this:
def filterByClass[A](l: List[_], c: Class[A]) =
l filter (_.asInstanceOf[AnyRef].getClass() == c)
Note that I am obliged to add the explicit conversion to AnyRef in order to avoid this compilation problem:
error: type mismatch;
found : _$1 where type _$1
required: ?{val getClass(): ?}
Note that implicit conversions are not applicable because they are ambiguous:
both method any2stringadd in object Predef of type (x: Any)scala.runtime.StringAdd
and method any2ArrowAssoc in object Predef of type [A](x: A)ArrowAssoc[A]
are possible conversion functions from _$1 to ?{val getClass(): ?}
l filter (_.getClass() == c)
However in this way the invocation of:
filterByClass(l, classOf[String])
returns as expected:
List(One)
but of course the same doesn't work, for example, with Int since they extends Any but not AnyRef, so by invoking:
filterByClass(l, classOf[Int])
the result is just the empty List.
Is there a way to make my filterByClass method working even with Int, Boolean and all the other classes extending Any?
The collect method already does what you want. For example to collect all Ints in a collection you could write
xs collect { case x: Int => x }
This of course only works when you hardcode the type but as primitives are handled differently from reference types it is actually better to do so. You can make your life easier with some type classes:
case class Collect[A](collect: PartialFunction[Any,A])
object Collect {
implicit val collectInt: Collect[Int] = Collect[Int]({case x: Int => x})
// repeat for other primitives
// for types that extend AnyRef
implicit def collectAnyRef[A <: AnyRef](implicit mf: ClassManifest[A]) =
Collect[A]({ case x if mf.erasure.isInstance(x) => x.asInstanceOf[A] })
}
def collectInstance[A : Collect](xs: List[_ >: A]) =
xs.collect(implicitly[Collect[A]].collect)
Then you can use it without even passing a Class[A] instance:
scala> collectInstance[Int](l)
res5: List[Int] = List(1)
scala> collectInstance[String](l)
res6: List[String] = List(One)
Using isInstanceOf:
scala> val l = List(1, "One", 2)
l: List[Any] = List(1, One, 2)
scala> l . filter(_.isInstanceOf[String])
res1: List[Any] = List(One)
scala> l . filter(_.isInstanceOf[Int])
res2: List[Any] = List(1, 2)
edit:
As the OP requested, here's another version that moves the check in a method. I Couldn't find a way to use isInstanceOf and so I changed the implementation to use a ClassManifest:
def filterByClass[A](l: List[_])(implicit mf: ClassManifest[A]) =
l.filter(mf.erasure.isInstance(_))
Some usage scenarios:
scala> filterByClass[String](l)
res5: List[Any] = List(One)
scala> filterByClass[java.lang.Integer](l)
res6: List[Any] = List(1, 2)
scala> filterByClass[Int](l)
res7: List[Any] = List()
As can be seen above, this solution doesn't work with Scala's Int type.
The class of an element in a List[Any] is never classOf[Int], so this is behaving as expected. Your assumptions apparently leave this unexpected, but it's hard to give you a better way because the right way is "don't do that."
What do you think can be said about the classes of the members of a heterogenous list? Maybe this is illustrative. I'm curious how you think java does it better.
scala> def f[T: Manifest](xs: List[T]) = println(manifest[T] + ", " + manifest[T].erasure)
f: [T](xs: List[T])(implicit evidence$1: Manifest[T])Unit
scala> f(List(1))
Int, int
scala> f(List(1, true))
AnyVal, class java.lang.Object
scala> f(List(1, "One", true))
Any, class java.lang.Object
This worked for me. Is this what you want?
scala> val l = List(1, "One", true)
l: List[Any] = List(1, One, true)
scala> l filter { case x: String => true; case _ => false }
res0: List[Any] = List(One)
scala> l filter { case x: Int => true; case _ => false }
res1: List[Any] = List(1)
scala> l filter { case x: Boolean => true; case _ => false }
res2: List[Any] = List(true)
Despite my solution could be less elegant than this one I find mine quicker and easier. I just defined a method like this:
private def normalizeClass(c: Class[_]): Class[_] =
if (classOf[AnyRef].isAssignableFrom((c))) c
else if (c == classOf[Int]) classOf[java.lang.Integer]
// Add all other primitive types
else classOf[java.lang.Boolean]
So by using it in my former filterByClass method as it follows:
def filterByClass[A](l: List[_], c: Class[A]) =
l filter (normalizeClass(c).isInstance(_))
the invocation of:
filterByClass(List(1, "One", false), classOf[Int])
just returns
List(1)
as expected.
At the end, this problem reduces to find a map between a primitive and the corresponding boxed type.
Maybe a help can arrive from scala.reflect.Invocation (not included in the final version of 2.8.0), the getAnyValClass function in particular (here slightly edited)
def getAnyValClass(x: Any): java.lang.Class[_] = x match {
case _: Byte => classOf[Byte]
case _: Short => classOf[Short]
case _: Int => classOf[Int]
case _: Long => classOf[Long]
case _: Float => classOf[Float]
case _: Double => classOf[Double]
case _: Char => classOf[Char]
case _: Boolean => classOf[Boolean]
case _: Unit => classOf[Unit]
case x#_ => x.asInstanceOf[AnyRef].getClass
}
With this function the filter is as easy as
def filterByClass[T: Manifest](l:List[Any]) = {
l filter (getAnyValClass(_) == manifest[T].erasure)
}
and the invocation is:
filterByClass[Int](List(1,"one",true))