Hey, I thought I provide some "realworld" feedback for the Go 1.18 beta and generics.
I recently spend some time implementing a concurrency framework that mimics Scala's Futures and Promises <https://docs.scala-lang.org/overviews/core/futures.html> API in Go as a learning exercise. You can find the source code at https://github.com/joa/go18beta In general it was very easy to get started with Go 1.18 and the added generics. Initially surprising was the fact that member methods can't be further parametrized but given the implementation decisions this makes sense. The absence of type erasure (disclaimer: I'm not a fan of type erasure!) made it harder to implement certain aspects. The code contains a linked list of callback[T] structs and I wanted to have a shared terminal or nil callback so to speak. I did get around this by creating a global variable of callback[any] that I can use with help of unsafe.Pointer in some places. It would be interesting to learn whether this is expected to work at all. I assume it only works here because callback[T] is never converted to an interface tuple and I also never access nilCallback's fields. Another aspect of this are some typical ADTs (like try.Try or option.Option) and for a method FlatMap[T, U](in Option[T])Option[U] I should in theory be able to re-use any Option[T] as an Option[U] as long as the compiler can prove T == U. I was not able to find any information about how to establish such a constraint or whether that's possible at all. Reusing the original instance would obviously crash at runtime when T != U since the interface type is different. In general this was a nice learning exercise and helpful in understanding some of the implementation decisions. While debatable whether a Future[T] would be idiomatic Go I think a type like Try[T] is actually interesting as it ensures that any panic is automatically recovered. While working on this project I did not experience any bugs other other unexpected behavior. I must say though that I did not use the constraints at all. Co- and contravariance would've been useful here but I also understand this is not part of this release. Anyways, that's just some feedback. Maybe you find this helpful or interesting. Best, Joa -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/golang-nuts/0dd15089-d148-43b4-b58a-41dd608eba26n%40googlegroups.com.
