Scala在使用函数时出现“类型不匹配错误”，但对于方法来说效果很好

Question

def stringToIntMethod(input:String):Option[Int] = {
  try{
    Some(Integer.parseInt(input.trim()))
  }
  catch{
    case e:Exception => None
  }
}

val stringToIntFunction: (String) => Option[Int] = (in:String) => {
  try{
    Some(Integer.parseInt(in.trim()))
  }
  catch{
    case e:Exception => None
  }
}

val stringAndIntArray = Array("Hello", "1","2","Hi") //Input

println("with Method is: " + stringAndIntArray.flatMap(stringToIntMethod))
println("with functon is: " + stringAndIntArray.flatMap(stringToIntFunction))

getting a type mismatch error while using stringToIntFunction in flatMap

type mismatch;
  found   : String => Option[Int]
  required: String => scala.collection.GenTraversableOnce[?]
println("with functon is: " + stringAndIntArray.flatMap(stringToIntFunction))
                                                    ^

Why is it?

Answer 1

flatMap requires lambda, you are passing it a normal method

Here is the fix

stringAndIntArray.flatMap(stringToIntMethod _)

Scala REPL

scala> def toInt(s: String): Option[Int] = Some(s.toInt)
toInt: (s: String)Option[Int]

scala> Array("1", "2", "3").flatMap(toInt _)
res1: Array[Int] = Array(1, 2, 3)

Normal method can be converted to lambda using underscore

More examples for clarity

scala> def foo(s: String, i: Int): Double = 1
foo: (s: String, i: Int)Double

scala> foo _
res2: (String, Int) => Double = $$Lambda$1162/1477996447@62faf77

scala> foo(_, _)
res3: (String, Int) => Double = $$Lambda$1168/1373527802@30228de7


scala> foo(_: String, _: Int)
res5: (String, Int) => Double = $$Lambda$1183/477662472@2adc1e84


scala> foo("cow", _: Int)
res7: Int => Double = $$Lambda$1186/612641678@146add7b

scala> foo("Cow is holy", _: Int)
res8: Int => Double = $$Lambda$1187/1339440195@7d483ebe

Also adding a comment from lambda. xy. x

f _ is syntactic sugar for f(_) which is again syntactic sugar for x => f(x)

Answer 2

I'm still puzzled myself about why this example does not work. I assume it has to do with Scala's type inference. If you look at the error message

 found   : String => Option[Int]
 required: String => scala.collection.GenTraversableOnce[?]

it says that the return value of stringToIntFunction does not fit the argument value of flatMap . Indeed, the type test

Some(1).isInstanceOf[TraversableOnce[Int]]

leads to:

<console>:138: warning: fruitless type test: a value of type Some[Int] cannot also be a scala.collection.TraversableOnce[Int] (the underlying of TraversableOnce[Int])
  Some(1).isInstanceOf[TraversableOnce[Int]]
                      ^
res24: Boolean = false

Funnily enough, when I change the return type of your function to TraversableOnce[Int] , it works:

def stringToIntFunction : String => TraversableOnce[Int] = (in:String) => {
   //...
}

leads to

scala> stringAndIntArray.flatMap(stringToIntFunction)
res28: Array[Int] = Array(1, 2)

The reason is that even though Option does not derive from TraversableOnce , there is an implicit conversion:

 scala> def f(to : TraversableOnce[Int]) = to.size
 f: (to: TraversableOnce[Int])Int
 scala> f(Some(1))
 res25: Int = 1

This has also been noticed in a different question before.

My theory is that for a method the return value is explicitly known to the compiler which allows it to detect the presence of an implicit conversion from Option[Int] => TraversableOnce[Int] . But in the case of a function value the compiler would only look for an implicit conversion between (String => Option[Int]) => (String => TraversableOnce[Int]) . When the lambda application stringToIntFunction _ is passed instead, the compiler seems to see that it can apply the implicit conversion again.