refined VS scribe

Does the Refined library for Scala (at https://github.com/fthomas/refined; "eu.timepit" %% "refined") work in Scala 3? Does it work differently?

Scala likes to do design-by-contract on the type level. You encode your pre- and post- conditions into types. Libraries like iron (scala 3) https://github.com/Iltotore/iron and refined (scala 2) https://github.com/fthomas/refined allow you to do all that without throwing any exceptions and they can even enforce some simple predicates at compile time.

Annotation is not the only way (and probably not the best IMHO) to do refined types. You might be interested in Iron in Scala 3 or Refined in Scala 2/3.

In one respect, nothing. You’re right. Even given refinement types as in Haskell or Scala, there is indeed a necessarily-partial function (refineV in Scala) to refine a value to its refinement type.

This is more or less how derivation works when you want to use something like Refined types (it exposes Validate[Type, Refinement] typeclass if I remember correctly). Enumeratum exposes Enum[A], and newtypes expose Coercible[From, To].

You will probably be interested into Iron or Refined.

You want either a refined type, a newtype, or if you are in Scala 3 an opaque type.

This example uses probably the most popular JSON library for Scala, Circe. There is another Scala library that lets us "be more specific with," or "refine," our types, called Refined. http4s doesn't provide any particular support for Refined, but Circe has a module integrating Refined, meaning that all of Circe's parsing, encoding, and decoding support is integrated with Refined. Putting together http4s' Circe support, Circe's generic codec derivation, and Circe's Refined integration, here's what I came up with:

I believe Scribe also supports what you want.