I know this seems like a big jump, but I am interested in driving towards doing a 1.0 release of helix sooner rather than later

to me, helix 1.0 would include turning on some of the more experimental parts of helix: • support for react-refresh • inference and auto-filling deps for custom hooks • linting hooks usage at least

these are currently behind feature flags or are otherwise opt in, if they exist at all (inference for custom hooks hasn't been implemented yet)

I've created a helix.edge.core namespace that now represents a more volatile namespace for these features to be tried out

helix.core will stay stable with no breaking changes up until 1.0. helix.edge.core will experience potentially breaking changes

on release of 1.0, helix.edge.core will be renamed to helix.core

my goal is to encourage people to use helix.edge.core to validate those experimental features

how reliable would inference be for hooks deps? for something like effect and layout-effect

it already works, you can pass in :auto-deps instead of a vector of dependencies for use-effect and use-layout-effect

it's reliable

I personally don't like using it for use-effect, but I think it is very good for use-memo and use-callback

the idea is that we could allow it to be used with custom hooks. so if you define a custom hook like:

(defhook use-custom-thing
  ;; annotate with metadata an argument that should be a deps array
  [^:deps deps x]
  ;; do something with deps and x
  (hooks/use-effect
    deps
    (prn x)))

I haven't done this work yet tho

I never use use-memo or use-callback, perhaps because I never had to optimize the render so much. Maybe I should?

It's awesome that auto-detecting stuff works runtime, it's like not even javascript