Remove any2stringadd.

[markhibberd]

Following on from a related thread Paul has started this here https://github.com/paulp/policy/commit/d5c06772d

How feasible it is to support from behind a flag may be questionable - but this would be a big win if it can be done.

[ijuma]

This would be cool indeed.

[non]

I had an idea of putting in a -Z flag that basically detects these sorts of usages of + (and makes them errors). This way new projects could avoid producing this kind of code, but the old projects would still work. Sort of like a built-in wart-remover, but without having to use ambiguous implicits to do it (and presumably with more specific error messages).

What do you think? The other alternative is to do something similar to what Paul did with Policy recently and modify the stdlib/compiler. I don't see a clear way for us to do that without drastically revising our compatibility guarantees.

[puffnfresh]

This was some of the motivation for #26 (any2stringadd is just an implicit in Predef) - sadly my PR misses a case and so doesn't give a nice error.

[paulp]

@puffnfresh You know the implicit never kicks on expressions like 1 + "bob" because a (String)String method has been synthetically added to Int/Long/etc, yes? See paulp/policy@4bb74b2bb .

[puffnfresh]

Also, -Yno-predef is already a valid solution to this problem. I think we should also focus on making -Yno-imports a sensible thing to use.

[puffnfresh]

@paulp I am sadly very aware of that.

[non]

Independent of -Yno-xyz solutions, we would need something else to disable the methods on Int &c. I'll try to prepare a flag-based branch to see how it feels.

[puffnfresh]

@non absolutely. That'd be great.

[puffnfresh]

I updated #32 to fix the problem with my implicitAmbiguous annotation. Now this code:

object Nope {
  @annotation.implicitAmbiguous("any2string is disabled")
  implicit def any2stringaddamb1[A](self: A): any2stringadd[A] = ???

  List(1, 2, 3) + "two"
}

Gives this:

/tmp/Nope.scala:5: error: any2string is disabled
  List(1, 2, 3) + "two"
      ^

Which makes me pretty happy.

🍰

[propensive]

Nice! Horrible use case, though... ;)

[milessabin]

To resurrect this issue, please rework it as an issue/PR against Lightbend Scala (ie. scala/scala).

[SethTisue]

related: https://issues.scala-lang.org/browse/SI-194, scala#5235