Add getOrRaise for OptionT, EitherT and IorT

[geirolz]

I've recently opened and merged a PR for mouse adding getOrRaise method to F[Option[A]] and F[Either[A, B]].
typelevel/mouse#323

Since I found this method useful I'd like to introduce it to OptionT, EitherT and IorT.

What do you think ?

---

P.S. For mouse, I've added getOrRaiseMsg as well which accepts a String and lifts it to RuntimeException.
Honestly, I've always missed this feature in cats since in 99% of the cases I've seen this:

IO.raiseError(new RuntimeException("ERROR"))

It would be great to have either

• raiseErrorMsg which accepts a string
• An overload of raiseError which accepts a string
• A string interpolator to lift a string to an Exception like val ex: RuntimeException = ex"ERROR" in order to keep all other API unchanged

[armanbilge]

Thanks for the PR! Is this syntax the same ideas as orRaise?
https://typelevel.org/cats/api/cats/syntax/ApplicativeErrorOps.html#orRaise(other:=%3EE)(implicitF:cats.ApplicativeError[F,E]):F[A]

def orRaise(other: => E)(implicit F: ApplicativeError[F, E]): F[A]

scala> import cats._, implicits._

scala> val fa: Either[String, Int] = Left("wrong")

scala> fa.orRaise("wronger")
res1: Either[String,Int] = Left(wronger)

scala> val fb: Either[String, Int] = Right(42)

scala> fb.orRaise("wrongest")
res2: Either[String,Int] = Right(42)

---

Edit: right, I see, you need the get to unpack it from the Option/Either/etc. It reminds me of this blogpost.
https://typelevel.org/blog/2018/04/13/rethinking-monaderror.html

[geirolz]

Edit: right, I see, you need the get to unpack it from the Option/Either/etc. It reminds me of this blogpost. https://typelevel.org/blog/2018/04/13/rethinking-monaderror.html

@armanbilge MonadBlunder looks a very idiomatic way to express/handle errors, or at least simplify complex cases where you have to deal with "always succeed" monad. At the moment unpacking and switching between Option, Either, Ior and their MT is still too complex and verbose IMHO.
The post is dated 2018 are you still thinking about it or are you working on another solution?

[armanbilge]

I'm not currently working on anything, but it's an interesting idea and your PR reminded me of this problem :)

I'm not sure if such a new idea could land in Cats due to the burden of guaranteeing long-term compatibility. However, I wonder if a 3rd party library can explore this idea by creating a Parallel-like typeclass that relates a MonadError datatype (e.g. IO) to its error-free Monad counterpart (e.g., some IO newtype that only implements Monad, but that seems rather strange 🤔 ). For Either or Option it would simply be Id I believe.

[armanbilge]

Ha, that seems to be this: https://github.com/LukaJCB/cats-uio

Maybe that repo just needs to be refreshed by an interested person :)

[geirolz]

Ha, that seems to be this: https://github.com/LukaJCB/cats-uio

Maybe that repo just needs to be refreshed by an interested person :)

I'll try to have a look, even just for Option and Either would be great! Especially when you have to deal with nested monads 😊 I see @LukaJCB already reached a great point there!

MonadBlunder aside, @armanbilge @satorg what do think about getOrRaise ?

[satorg]

lgtm, thanks!