MonadError#adaptError could be on ApplicativeError instead

[bplommer]

MonadError#adaptError is defined as

def adaptError[A](fa: F[A])(pf: PartialFunction[E, E]): F[A] =
  flatMap(attempt(fa))(_.fold(e => raiseError(pf.applyOrElse[E, E](e, _ => e)), pure))

But it could be defined on ApplicativeError simply as

def adaptError[A](fa: F[A])(pf: PartialFunction[E, E]): F[A] =
  recoverWith(fa)(pf.andThen(raiseError))

I haven't opened a PR because this would be a breaking change.

[kailuowang]

I think we can at least change the implementation in MonadError and add it to ApplicativeError syntax

[bplommer]

@kailuowang I'm happy to pick this up. I suppose it will need a different name in the ApplicativeError syntax to prevent ambiguity - maybe applicativeAdaptError?

[kailuowang]

@bplommer thanks, yeah, that's a good catch. applicativeAdaptError sounds too long. how about adaptErr or modifyError with a comment on why it can't be the same as the one in MonadError for now? we shall be able to change this in 2.0.

[kubukoz]

Can this be closed? :)

[kailuowang]

I still like to move this method in the type class in 2.1

[kubukoz]

On the >2.11 branch, right?

[kailuowang]

Right, which is master.

[travisbrown]

For what it's worth it should have been possible to fix the syntax in 2.0 by making the MonadErrorOps adaptError package private, but I just tried it for 2.1 and ran into a compiler bug: scala/bug#11787

(private or private[MonadErrorOps] avoid the bug but break bincompat.)