I want to transform a List[Option[T]]
into a Option[List[T]]
. The signature type of the function is
def lo2ol[T](lo: List[Option[T]]): Option[List[T]]
The expected behavior is to map a list that contains only Some
s into a Some
containing a list of the elements inside the elements Some
's. On the other hand, if the input list has at least one None
, the expected behavior is to just return None
. For example:
scala> lo2ol(Some(1) :: Some(2) :: Nil)
res10: Option[List[Int]] = Some(List(1, 2))
scala> lo2ol(Some(1) :: None :: Some(2) :: Nil)
res11: Option[List[Int]] = None
scala> lo2ol(Nil : List[Option[Int]])
res12: Option[List[Int]] = Some(List())
An example implementation, without scalaz, would be:
def lo2ol[T](lo: List[Option[T]]): Option[List[T]] = {
lo.foldRight[Option[List[T]]](Some(Nil)){(o, ol) => (o, ol) match {
case (Some(x), Some(xs)) => Some(x :: xs);
case _ => None : Option[List[T]];
}}}
I remember seeing somewhere a similar example, but using Scalaz to simplify the code. How would it look like?
A slightly more succinct version, using Scala2.8 PartialFunction.condOpt
, but still without Scalaz:
import PartialFunction._
def lo2ol[T](lo: List[Option[T]]): Option[List[T]] = {
lo.foldRight[Option[List[T]]](Some(Nil)){(o, ol) => condOpt(o, ol) {
case (Some(x), Some(xs)) => x :: xs
}
}}
There's a function that turns a List[Option[A]]
into an Option[List[A]]
in Scalaz. It's sequence
. To get None
in case any of the elements are None
and a Some[List[A]]
in case all the elements are Some
, you can just do this:
import scalaz.syntax.traverse._
import scalaz.std.list._
import scalaz.std.option._
lo.sequence
This method actually turns F[G[A]
into G[F[A]]
given that there exists an implementation of Traverse[F]
, and of Applicative[G]
(Option
and List
happen to satisfy both and are provided by those imports).
The semantics of Applicative[Option]
are such that if any of the elements of a List
of Option
s are None
, then the sequence
will be None
as well. If you want to get a list of all the Some
values regardless of whether any other values are None
, you can do this:
lo flatMap (_.toList)
You can generalize that for any Monad
that also forms a Monoid
(List
happens to be one of these):
import scalaz.syntax.monad._
def somes[F[_],A](x: F[Option[A]])
(implicit m: Monad[F], z: Monoid[F[A]]) =
x flatMap (o => o.fold(_.pure[F])(z.zero))
For some reason you dislike
if (lo.exists(_ isEmpty)) None else Some(lo.map(_.get))
? That's probably the shortest in Scala without Scalaz.
While the Applicative[Option]
in Scalaz has the wrong behaviour to directly use MA#sequence
, you can also derive an Applicative
from a Monoid
. This is made convenient with MA#foldMapDefault
or MA#collapse
.
In this case, we use a Monoid[Option[List[Int]]
. We first perform an inner map (MA#∘∘
) to wrap the individual Int
s in List
s of one element.
(List(some(1), none[Int], some(2)) ∘∘ {(i: Int) => List(i)}).collapse assert_≟ some(List(1, 2))
(List(none[Int]) ∘∘ {(i: Int) => List(i)}).collapse assert_≟ none[List[Int]]
(List[Option[Int]]() ∘∘ {(i: Int) => List(i)}).collapse assert_≟ none[List[Int]]
Abstracting from List
to any container with instances for Traverse
, Pointed
and Monoid
:
def co2oc[C[_], A](cs: C[Option[A]])
(implicit ct: Traverse[C], cp: Pointed[C], cam: Monoid[C[A]]): Option[C[A]] =
(cs ∘∘ {(_: A).pure[C]}).collapse
co2oc(List(some(1), none[Int], some(2))) assert_≟ some(List(1, 2))
co2oc(Stream(some(1), none[Int], some(2))) assert_≟ some(Stream(1, 2))
co2oc(List(none[Int])) assert_≟ none[List[Int]]
co2oc(List[Option[Int]]()) assert_≟ none[List[Int]]
Sadly, trying to compile this code currently either triggers #2741 or sends the compiler into an infinite loop.
UPDATE
To avoid traversing the list twice, I should have used foldMapDefault
:
(List(some(1), none[Int], some(2)) foldMapDefault (_ ∘ ((_: Int).pure[List])))
This answer was based on the original request that an empty list, or a list containing only None
s, should return a None
. Incidentally, this would be best modeled by the type Option[scalaz.NonEmptyList]
-- NonEmptyList
guarantees at least one element.
If you just want the a List[Int]
, there are many easier ways, given in other answers. Two direct ways that haven't been mentioned:
list collect { case Some(x) => x }
list flatten
This worked for me. I hope this is a correct solution.
It returns None if one of the Options in the List is None, otherwise it returns Option of List[A]
def sequence[A](a: List[Option[A]]): Option[List[A]] = {
a.foldLeft(Option(List[A]())) {
(prev, cur) => {
for {
p <- prev if prev != None
x <- cur
} yield x :: p
}
}
}