Channels
#arrow-contributors
- e
Eugenio
07/22/2018, 12:44 PMwhich means that, theoretically, you could traverse the
hierarchy, only to find that the "id" is NOT of the right type, and there's no link to the real typeKind - p
pakoito
07/22/2018, 12:44 PMyeah but that’s not unsafer than what we have right now - e
Eugenio
07/22/2018, 12:44 PMtrue - e
Eugenio
07/22/2018, 12:44 PMand we can still do that at runtime - e
Eugenio
07/22/2018, 12:45 PMperhaps it could help with autocomplete though? - e
Eugenio
07/22/2018, 12:45 PMor to guide users in building kinds - e
Eugenio
07/22/2018, 12:45 PMbecause it wouldn't even compile if you're not using a "kind id" or a kind - p
pakoito
07/22/2018, 12:45 PMwhat are you going to use the link between the witness and the class for? - e
Eugenio
07/22/2018, 12:46 PMin my original snippet it was `forType`:Copy codeabstract class KindId(val forType: KClass<out Kind<*, *>>) : KindOrId() - e
Eugenio
07/22/2018, 12:47 PMit could almost even be
, so that we could strip it at runtimeprivate - e
Eugenio
07/22/2018, 12:48 PMor we could use a generic type I guess - e
Eugenio
07/22/2018, 12:48 PMit's just annoying/confusing that you have to give some types, even if they're wildcards - e
Eugenio
07/22/2018, 12:49 PMlike instead ofCopy code
it woud beobject ForEither : KindId(Either::class)Copy codeobject ForEither : KindId<Either<*,*>>() - e
Eugenio
07/22/2018, 12:49 PMusing
is a nice trick to avoid having to specify generic type args::class - p
pakoito
07/22/2018, 12:50 PMfair - p
pakoito
07/22/2018, 12:51 PMstill, what do you need those for in the codegen? I can smelelelelelelelel what the Eugenio is cooking - e
Eugenio
07/22/2018, 12:52 PMhehehehe - e
Eugenio
07/22/2018, 12:52 PMeasy: - e
Eugenio
07/22/2018, 12:54 PMso first I search for all the
annotations, say that I find this:@instanceExtensionsCopy code@instanceExtensions interface EitherFunctorInstance<L> : Functor<EitherPartialOf<L>> { ... } - e
Eugenio
07/22/2018, 12:54 PMI need to go from
toFunctor<EitherPartialOf<L>>
, somehowEither - e
Eugenio
07/22/2018, 12:56 PMso • I expand
->EitherPartialOf<L>
• I look atKind<ForEither, A>
->ForEither
->ForEither : KindId(Either::class)
💥Either - p
pakoito
07/22/2018, 12:57 PM👍 - p
pakoito
07/22/2018, 12:57 PMgotcha - p
pakoito
07/22/2018, 12:57 PMI would suggest - p
pakoito
07/22/2018, 12:57 PMdon’t do the last expansion - p
pakoito
07/22/2018, 12:58 PMput the extfuns onKind2<ForEither, L, R> - p
pakoito
07/22/2018, 12:58 PMEitherOf<L, R> - p
pakoito
07/22/2018, 12:58 PMthat’s what we do when we do it manually - p
pakoito
07/22/2018, 12:59 PMbecause it works even when abstracted