When writing a function whose sole purpose is to c...

# arrow

[CLOVIS]

When writing a function whose sole purpose is to check some invariant, is it best practice to declare it as:
EffectScope<Failure>.() -> Unit
or:
() -> Either<Failure, Unit>
I think the first one is clearer that "it's just validation code", what do you think? Is there any official recommendation?
The first one also doesn't need the user to call bind , which they may forget since there is no proper result.

[simon.vergauwen]

Hey @CLOVIS,

There is not official recommendation at this time, you can find a more lengthy answer here. https://stackoverflow.com/questions/72277426/which-should-we-choose-between-effect-and-either-as-a-return-type-of-our-busines/72277572#72277572

The recommended way going forward will probably become context receivers, as you mention this won’t suffer from the bind issue and it’ll also be the most generic version.

With Arrow 1.x.x:

context(EffectScope<Failure>)
suspend fun validate(): Unit = TODO()

With Arrow 2.x.x suspend will become optional, depending on whether you use suspend inside or not.

context(Raise<Failure>)
(suspend) fun validate(): Unit = TODO()

In Arrow 2.x.x you’ll be able to do either { validate() } to turn it into () -> Either<Failure, Unit>. Where in Arrow 1.x.x you still need to be explicit about either.eager and EagerEffectScope<Failure>.

I expect Arrow 2.x.x to become stable sooner than context receivers since they’re still JVM experimental atm.

You can now use extension functions instead of context receivers