when defining method in Scala, I found this
def method1: Int => Int = (j: Int) => j // works
def method2: Int => Int = j => j // works
def method3: Int => Int = j: Int => j // error
def method4: Int => Int = {j: Int => j} // works
Can anyone explain why method3 does not work? Is there any ambiguity in it?
One possible explanation is indeed that this restriction avoids ambiguity: x: A => B could be understood as an anonymous function that takes a parameter x of type A and returns the object B. Or it could be understood as "casting" a variable x to the type A => B. It is very rare that both of these would be valid programs, but not impossible. Consider:
class Foo(val n: Int)
val Foo = new Foo(0)
val j: Int => Foo = new Foo(_)
def method1: Int => Foo = (j: Int) => Foo
def method2: Int => Foo = j: Int => Foo
println(method1(1).n)
println(method2(1).n)
This actually compiles and prints:
0
1
All of these variants are covered by Anonymous Functions section of the specification. The relevant part is
In the case of a single untyped formal parameter,
(x) => ecan be abbreviated tox => e. If an anonymous function(x : T) => ewith a single typed parameter appears as the result expression of a block, it can be abbreviated tox: T => e.
In method3, the function isn't the result expression of a block; in method4 it is.
EDIT: oops, presumably you meant why the limitation is there. I'll leave this answer for now as stating what the limitation is exactly, and remove it if anyone gives a better answer.
{ }. The result expression of a block is its last expression.