feat: add .toEitherT to Validated

[jeengbe]

Similar to .toEither but with EitherT

[jeengbe]

I believe this is Scala 3 syntax, please replace it with the Scala 2 equivalent (which I assume doc comments are written in)

[rossabaker]

Didn't compile it myself, but I think this would do it:

Suggested change

	* scala> given ExecutionContext = ExecutionContext.global
	* scala> import scala.concurrent.ExecutionContext.Implicits.global

[jeengbe]

Hm, the errors correctly mark:

[error] /home/runner/work/cats/cats/core/src/main/scala/cats/data/Validated.scala:286:7: covariant type E occurs in invariant position in type [F[_]](implicit evidence$1: cats.Applicative[F]): cats.data.EitherT[F,E,A] of method toEitherT
[error]   def toEitherT[F[_]: Applicative]: EitherT[F, E, A] =
[error]       ^

yet with an extension, this variance difference is no issue:

extension [E, A](validated: Validated[E, A])
  def toEitherT[F[_]: Applicative]: EitherT[F, E, A] = validated match
    case Validated.Invalid(e) => EitherT.leftT(e)
    case Validated.Valid(a)   => EitherT.rightT(a)

I have unfortunately not the slightest idea how to fix this

[jeengbe]

And perhaps

Suggested change

	this match {
	case Invalid(e) => EitherT.leftT(e)
	case Valid(a) => EitherT.rightT(a)
	}
	EitherT.fromEither(this.toEither)

is cleaner

[rossabaker]

Thanks for the contribution! The compilation error can be fixed with:

  def toEitherT[F[_]: Applicative, EE >: E, AA >: A]: EitherT[F, EE, AA] =

... but that's not what we want, because then .toEitherT needs three parameters, and it loses its convenience. I thought Ior.toIorT might serve as prior art, but it doesn't exist. But IorT.fromIor does! Maybe what we want is EitherT.fromValidated?

[rossabaker]

Didn't compile it myself, but I think this would do it:

Suggested change

	* scala> given ExecutionContext = ExecutionContext.global
	* scala> import scala.concurrent.ExecutionContext.Implicits.global

[rossabaker]

The next sentence says we have an Either. Perhaps it's better to make the EitherT a separate example?

[satorg]

Maybe what we want is EitherT.fromValidated?

Agree. Adding a new constructor to EitherT seems more coherent taking into account that it already offers constructors like fromEither, fromOption and some others.

Another thought is that EitherT is more complex type comparing to Validated, so I personally feel that making more complex abstractions depending on more basic ones is usually preferable that doing the opposite.

[jeengbe]

Thanks for outlining the reasoning, I agree.