Roadmap

In progress

Accept an expression through --expr for rcl evaluate.
Add a --follow for rcl fmt that follows imports.
Add to_lower_ascii, to_upper_ascii functions. Maybe even to_snake_case.
Add integer division and modulo methods (or operators, but probably methods are nicer here).

Yaml output mode.
Preserve insertion order in dicts and sets, GC'd runtime.
Evaluate whether to add a tuple type to describe heterogeneous lists of statically known length.
Add a spread operator for inside collections, and evaluate whether it should supersede the dict/set union operator. It would solve some typing issues, and it's easier to format in a pleasant way.
Add nested fields to record syntax, i.e. {x.y = "z"} == {x = {y = "z" }}.
Mitigate Trojan Source pitfall (https://trojansource.codes/). We can't ban Bidi code points from strings because it would break json compatibility, but we can ban it from comments, and we can make the formatter turn it into escape sequences in strings.

Add a more efficient runtime system, possibly with a garbage collector.
Multithreaded evaluation. Due to strict evaluation and being purely functional without side effects, this is actually feasible: when we evaluate f(a, b, c), we can evaluate a, b, and c in parallel. Similarly in operator chains like a + b + c. It might use a work-stealing scheduler, where the current thread puts something to evaluate up for grabs, continues to evaluate the next thing, and once it cannot continue, it either waits for the next result or evaluates it itself.