Make RefinedTypeOps definition more ergonomic

[Iltotore]

Currently, declaring a new type requires to declare the constraint in the opaque type and in RefinedTypeOps's second argument which can be problematic for long constraints. Example derived from Valentin Bergeron and Raphaël Lemaitre's talk

opaque type Sats = Long :| GreaterEqual[0] & LessEqual[100000000 * 21000000]
object Sats extends RefinedTypeOps[Long, GreaterEqual[0] & LessEqual[100000000 * 21000000], Sats]

An alias can be used to mitigate the boilerplate:

private type SatsConstraint =
  GreaterEqual[0] & LessEqual[100000000 * 21000000]

opaque type Sats <: Long = Long :| SatsConstraint
object Sats extends RefinedTypeOps[Long, SatsConstraint, Sats]

Ideally, the redundancy should be removed either by passing only the third parameter in RefinedTypeOps:

opaque type Sats = Long :| GreaterEqual[0] & LessEqual[100000000 * 21000000]
object Sats extends RefinedTypeOps[Sats]

or defining the type alias from the companion object (like Neotype):

opaque type Sats = Sats.Type
object Sats extends RefinedTypeOps[Long, GreaterEqual[0] & LessEqual[100000000 * 21000000]]

I think defining the ops from the opaque type (first solution) is more natural. It was not possible before but might be doable easily enough since TypeRepr#dealias in Scala 3.4+ also dealiases opaque types. Thus, a mirror-like mechanism can (need to be tested) be used internally:

trait RefinedTypeOps[T](using mirror: IronType.Mirror, constraint: RuntimeConstraint[mirror.BaseType, mirror.ConstraintType])

On the other hand, solution 2 is pretty easy to do:

trait RefinedTypeOps[A, C]:

  type Type = A :| C

Users' opinions need to be collected in order to make a choice.

[vbergeron-ledger]

I 👍 the solution from the opaque type definition is cleaner : more readable and the companion object fulfill more its role of, well, companion.

[Iltotore]

Unfortunately the second solution does not seem doable at the moment. Scala is missing inline constructor parameters.

Since this feature will probably not land until long time, maybe it's worth to implement the first proposal instead?

What do you think @vbergeron-ledger?