Channels
#arrow-contributors
- d
domfox
04/24/2018, 7:07 AMI'm also curious about how the typeclasses work in general - is
being defined as an extension method onbind
locally within the block?Any - d
domfox
04/24/2018, 7:08 AMMy general level of knowledge: has done some basic MTL-style Haskell programming, pored over the internals of Scalaz a couple of years ago. - j
jereksel
04/24/2018, 7:09 AMNo, binds are defined for Either, Option, etc. I'm at work so I can't give you exact locations, but from GH search I found this: https://github.com/arrow-kt/arrow/blob/ee18d155ca24f378224520f8131db7ef72bb587c/modules/core/arrow-typeclasses/src/main/kotlin/arrow/typeclasses/MonadContinuations.kt#L50 When I'm looking at this I don't see any hacks to be honest. - d
domfox
04/24/2018, 7:15 AMThe import of
suggests some entanglement in internal implementation, but perhaps it's harmless.kotlin.coroutines.experimental.intrinsics.COROUTINE_SUSPENDED - d
domfox
04/24/2018, 7:26 AM(On "how typeclasses are implemented", I've found the blogpost series on Typeclassless style, reading that now) - j
jereksel
04/24/2018, 7:27 AMWell we're writing our own coroutine type, so I think it's inevitable. But it looks like the way standard library does it, so I'm not worried about that. - j
jereksel
04/24/2018, 7:27 AMhttps://github.com/JetBrains/kotlin/blob/master/libraries/stdlib/coroutines/src/kotlin/coroutines/intrinsics/Intrinsics.kt#L61 - also, since 1.3 it looks like it's not experimental anymore - j
jereksel
04/24/2018, 9:03 AMSo I got AP processing done, but there are still some issues with ScriptEngine - j
jereksel
04/24/2018, 9:03 AMCopy codepackage com.example.domain val networkResult: NetworkResult = HttpError("boom!") val f: (String) -> String = String::toUpperCase when (networkResult) { is HttpError -> networkResult.copy(f(networkResult.message)) else -> networkResult } //null - j
jereksel
04/24/2018, 9:04 AM
doesn't return value - it always returns nullwhen - p
pakoito
04/24/2018, 9:50 AM@domfox this is what you’re looking for - p
pakoito
04/24/2018, 9:51 AMthat’s the reflection access to fields of the coroutineimpl class - p
pakoito
04/24/2018, 9:51 AMto enable multiple returns - p
pakoito
04/24/2018, 9:52 AMthe MonadContinuation class @jereksel linked before is quite tame, as you see we only need to copy those stack labels as the monad progresses through the bind/await - p
pakoito
04/24/2018, 9:53 AMkotlin.coroutines.experimental.intrinsics.COROUTINE_SUSPENDED
That’s the expected usage of the API. All the machinery is marked “experimental” even after 3 minor versions of the language since they were released. - p
pakoito
04/24/2018, 9:54 AMThe implementation of coroutines is in itself a labyrinth, like why would they check that COROUTINE_SUSPENDED marker by reference to assert if a block returned a value or not - p
pakoito
04/24/2018, 9:55 AMprobably performance - p
pakoito
04/24/2018, 9:56 AMand it’s split between the core, implemented in the compiler and barebones, and the framework, which contains the actual coroutine abstractions for reification, cancellation etc - p
pakoito
04/24/2018, 9:56 AMwe only use the core so we have to rewrite parts of the framework. Luckily for us, we don’t need a general solution but one that adapts to typeclasses - p
pakoito
04/24/2018, 9:59 AMas for whether this is a stable solution or not, this implementation and the intrinsics are jvm-only and part of the stdlib. The moment they change the library will break, we’ll update, and that’s that - d
domfox
04/24/2018, 10:11 AMThank you, that is indeed exactly what I was looking for. - p
pakoito
04/24/2018, 10:24 AMhow nasty would you say it is? - p
pakoito
04/24/2018, 10:24 AM😄 - d
domfox
04/24/2018, 10:30 AMWell, it's doing magic by using an already fairly magical language feature in an unintended way 😉 On the other hand, there's a deep affinity between monads and continuations, so I can't say it's an unreasonable approach...🎉 1 - p
pakoito
04/24/2018, 1:44 PM@domfox if you're interested on helping me writing the next generation of continuations, I will take any help I can get - p
pakoito
04/24/2018, 1:44 PMI've already failed 3 attempts of rewriting them, so 4th time is the charm I guess 🙂💪 1 - d
domfox
04/24/2018, 3:05 PMWhat's the major sticking point? - p
pakoito
04/24/2018, 8:31 PMthe scope/intentions for the code - p
pakoito
04/24/2018, 8:31 PMi wanted to have something like
that executed the code in the common pool and blocked the current thread for a responseEither.binding(CommonPool) { ... }