Has anyone considered adding some sort of higher kinded type (HKT) capability to Carbon?

[travis-leith]

I suspect HKT would be easier to design in conjunction with generics, rather than trying to make it work after generics are stabilised. Just wondering if HKT has been considered or possibly ruled out?

https://serokell.io/blog/kinds-and-hkts-in-haskell

[geoffromer]

I'm not aware of any discussions of it, although of course there could have been some that I wasn't there for. Are there problems that will come up in Carbon that you think HKTs would solve?

[travis-leith]

I'm not thinking of problems "in" Carbon itself. Just that HKTs increase the range of ideas that are succinctly expressible in the language itself. Particularly for building generic libraries. On the other hand, HKTs have a reputation for being arcane, but that might not be an issue if it is mainly used by library builders. My point is that if consideration is not given to the syntax early on, it might be awkward to adopt later on. So perhaps it would be best to consider them now, and explicitly decide if they should be part of the language or not.

[geoffromer]

One of our main goals with Carbon is to make it easier to evolve, so that we can confidently wait to add features until we have clear motivating use cases for them. Having those use cases not only makes it much easier to justify investing the resources to design a feature, it also makes the feature itself easier to design, because we can get immediate feedback about whether the design works by applying it to those use cases.