[SIP-71] `into` modifier doesn't work with enums

[prolativ]

Compiler version

3.7.3-RC1-bin-20250619-b1da8fc-NIGHTLY

Minimized code

import scala.language.experimental.into

into enum MyEnum:
  case MyString(s: String)
  case MyInt(i: Int)

import MyEnum.*

given Conversion[String, MyEnum] = MyString(_)
given Conversion[Int, MyEnum] = MyInt(_)

def takeMyEnum(f: MyEnum) = f

@main def test = 
  val f1 = takeMyEnum(123)
  val f2 = takeMyEnum("abc")

Output

[error] Only access modifiers are allowed on enum definitions
[error] into enum MyEnum:
[error]           ^^^^^^

Expectation

This snippet should compile without an error or a warning, similarly to the non-enum version of the code

import scala.language.experimental.into

into sealed trait MyEnum

object MyEnum:
  case class MyString(s: String) extends MyEnum
  case class MyInt(i: Int) extends MyEnum

import MyEnum.*

given Conversion[String, MyEnum] = MyString(_)
given Conversion[Int, MyEnum] = MyInt(_)

def takeMyEnum(f: MyEnum) = f

@main def test = 
  val f1 = takeMyEnum(123)
  val f2 = takeMyEnum("abc")

[odersky]

But should it? I am now coming around the opinion that it's better to leave it like this.

Only access modifiers are allowed on enum definitions

That's a nice simple rule. It would be a lot uglier if we said

Only access modifiers and into are allowed on enum definitions

And since the stated objective of into on classes is for migration from Scala 2, enums don't come up since they don't exist in Scala 2.

[sjrd]

IMO it makes sense to allow into on enums. enums desugar into classes, so why should be forbid a modifier that can make sense? (of course, we forbid things like abstract on enums because it cannot possibly make sense)

Only access modifiers are allowed on enum definitions

That's a nice simple rule. It would be a lot uglier if we said

Hum ... was it intended to be this nice, or did it just happen that everything we rejected so far was everything-but-access-modifers? If it is intended as coming from first principles, that is of course a valid argument. But if it just happened to be that way, IMO that's not a very strong argument.

[odersky]

Hum ... was it intended to be this nice, or did it just happen that everything we rejected so far was everything-but-access-modifers?

Yes, I think it was intended. You could have other modifiers such as transparent on enums but we want to keep it simple.

But if it just happened to be that way, IMO that's not a very strong argument.

It's still a valid argument. Even if something is nice by sheer serendipity there's still a reason to keep it nice.