Drop case in Refined

[dwijnand]

.. with before-and-after (newly created) spec changes

[dwijnand]

Btw, happy to iterate and tweak any minor detail, the intent of the change is my only goal.

[dwijnand]

Let me know what you want to do in terms of bincompat..

[fthomas]

MiMa complains about a lot of stuff that case brought in. Of those things, I think I only care about scala.Serializable. What do you think of the others?

[dwijnand]

I'm happy to recover Serializable.

The other one that's a shot-call on your part - do you want to provide Refined.unapply?

[fthomas]

Yes, Refined.unapply would be useful too.

[dwijnand]

Thanks to @ceedubs it seems this was already decided against in #57. Perhaps this alternative fix is more acceptable.

[dwijnand]

@fthomas Added Refined.unapply, squashed and pushed.

[fthomas]

Thanks @dwijnand! LGTM. So the only unsafe operations are now Refined.unsafeApply and asInstanceOf, right?

[dwijnand]

Refined.unsafeApply - yes. What do you mean by asInstanceOf? Can you show a code/repl snippet?

[fthomas]

scala> val x: Int Refined Positive = 5
x: eu.timepit.refined.api.Refined[Int,eu.timepit.refined.numeric.Positive] = 5

scala> x.asInstanceOf[Int Refined Negative]
res0: eu.timepit.refined.api.Refined[Int,eu.timepit.refined.numeric.Negative] = 5

[dwijnand]

Right.. yeah, that one too :)

[dwijnand]

Custom value classes don't have that problem.

[fthomas]

I'm wondering if we should make Refined a case class again by making it a sealed abstract case class. With this combination of keywords we don't get unsafe apply and copy methods.

[dwijnand]

Can't have sealed value classes, as value classes are final.

[fthomas]

Good catch, @dwijnand. I've heard that the value class aspect of Refined was a reason for not using refined in the first place since value classes can't be nested. Maybe extends AnyVal should be dropped from Refined.

[dwijnand]

Maybe value classes should be allowed to nest... :D

[fthomas]

I heard that is an easy change in scalac ;-)