Add typed versions of Mirror.ProductOf#fromProduct

[mpilquist]

The existing Mirror.ProductOf#fromProduct method is defined for any scala.Product which can lead to some surprising runtime errors:

Welcome to Scala 3.1.2-RC1-bin-SNAPSHOT-nonbootstrapped-git-b636aa9 (17.0.1, Java OpenJDK 64-Bit Server VM).
Type in expressions for evaluation. Or try :help.

scala> case class Point(x: Int, y: Int, z: Int)
// defined case class Point

scala> summon[deriving.Mirror.ProductOf[Point]].fromProduct((1, 2))
java.lang.IndexOutOfBoundsException: 2 is out of bounds (min 0, max 1)
  at scala.Product2.productElement(Product2.scala:43)
  at scala.Product2.productElement$(Product2.scala:40)
  at scala.Tuple2.productElement(Tuple2.scala:24)
  at rs$line$1$Point$.fromProduct(rs$line$1:1)
  at rs$line$1$Point$.fromProduct(rs$line$1:1)
  ... 38 elided

scala> summon[deriving.Mirror.ProductOf[Point]].fromProduct((1, 2, 3.0))
java.lang.ClassCastException: class java.lang.Double cannot be cast to class java.lang.Integer (java.lang.Double and java.lang.Integer are in module java.base of loader 'bootstrap')
  at scala.runtime.BoxesRunTime.unboxToInt(BoxesRunTime.java:99)
  at rs$line$1$Point$.fromProduct(rs$line$1:1)
  at rs$line$1$Point$.fromProduct(rs$line$1:1)
  ... 38 elided

This PR adds fromTuple and fromProductTyped extension methods to Mirror.ProductOf[T], both which ensure the argument is well typed:

scala> summon[deriving.Mirror.ProductOf[Point]].fromTuple((1, 2))
-- [E007] Type Mismatch Error: -------------------------------------------------
1 |summon[deriving.Mirror.ProductOf[Point]].fromTuple((1, 2))
  |                                                   ^^^^^^
  |                                                 Found:    (Int, Int)
  |                                                 Required: (Int, Int, Int)

longer explanation available when compiling with `-explain`
1 error found

scala> summon[deriving.Mirror.ProductOf[Point]].fromTuple((1, 2, 3.0))
-- [E007] Type Mismatch Error: -------------------------------------------------
1 |summon[deriving.Mirror.ProductOf[Point]].fromTuple((1, 2, 3.0))
  |                                                          ^^^
  |                                                 Found:    (3.0d : Double)
  |                                                 Required: Int

longer explanation available when compiling with `-explain`
1 error found

scala> summon[deriving.Mirror.ProductOf[Point]].fromTuple((1, 2, 3))
val res2: Point = Point(1,2,3)

[dwijnand]

Add as experimental?

[mpilquist]

Add as experimental?

Done. Curious when @experimental gets removed from API additions? Any link to docs on that?

[nicolasstucki]

Curious when @experimental gets removed from API additions?

We can only remove experimental when going to a new minor release (3.2.0, 3.3.0, 3.4.0, ...). Before each minor release we need to decide which experimental method will become non-experimental.

[julienrf]

It looks to me that the existing fromProduct method should be fixed instead of adding a new method fromProductTyped. Is this something doable before Scala 4.x?

[OlivierBlanvillain]

What are the benefits of adding these API to the standard library as opposed to defining then in user space? IIUC the Mirror API was not designed to provide a type safe experience out of the box, but rather to be a low-level building block for people to build type-safe frameworks on top...

[mpilquist]

Discoverability & avoidance of dependencies in foundational libraries, which lead to instability & churn in overall library ecosystem. The more type safety we can provide in these foundational APIs, the easier it will be for folks to build their higher level derivations, frameworks, etc.

[OlivierBlanvillain]

That makes sense, although if it's just an extension method away I think we could also do without. But I suppose such argument could be made for any addition to the standard library...

Regardless, I think that PR should also update docs/docs/reference/contextual/derivation.md since it's the main entry point to the Mirror APIs.

[odersky]

I think that addition is quite useful.

[odersky]

@mpilquist Can you rebase on latest master, then we can merge? Thank you!

[mpilquist]

Done, thanks!