Ambrose Bonnaire-Sergeant

I take that back, recursive schemas seem to work. I forgot to add a `:ref` in recursive places.

Reproduced [here](https://github.com/metosin/malli/pull/1068). My initial assessment is that because `m/-function-info` returns `:varargs` for `[:=> [:cat :int [:? :int] :string] :keyword]`, the `cljs$core$IFn$_invoke$arity$variadic` arity is wrapped instead of the `cljs$core$IFn$_invoke$arity$` arity.

This was added for 0.15.0-SNAPSHOT, but it was added as an optional final child to `:=>` instead of properties. I think this prevents any future extension of `:=>` children. In...

There's something bugging me about the dynamism being lost here. For example, does this interfere with dynamically scoped registries in a way that matters?

> I would assume that I can recreate a schema validator with current bindings when calling -validator on it, no caching. I'm not sure that holds even today, the caching...

I needed to change the directory for the script to work (Sonoma 14.6.1, M1). ``` #!/bin/bash set -e cd /Applications/SuperCollider.app/Contents/MacOS/ exec ./sclang "$@" ```

> Maybe it makes more sense to move next_in_place back into sequenceable. Yeah probably. We can get rid of this template and delegate to implementations. ``` template requires(behavior::sequenceable && !behavior::sequenceable_in_place)...

Pressing enter now throws instead of `nil`. ``` clojure.core=> Exception: unable to map file /var/folders/s5/309818k92r92zn9c_nzvmnzc0000gn/T/jank-repl-1ca7.jank due to error: unable to map file ```

Fixed by https://github.com/metosin/malli/pull/1040

This looks great. Do you know if we can combine this with pprinting the rest of the forms?