[djwatson24]

Cool to see explicit annotations for how to make rc fast when you have access to the ir. But doesn’t solve some major issues, like cycle collection:

“ In practice, mutable references are the main way to con- struct cyclic data. Since mutable references are uncommon in our setting, we leave the responsibility to the programmer to break cycles by explicitly clearing a reference cell that may be part of a cycle. ”

[bjourne]

Always a catch with these zero-overhead refcounting schemes. If the programmer has to use some equivalent to weakrefs then how is this an improvement over existing schemes? The possibility of cyclic garbage to accumulate is a very big drawback in many applications.

[studius]

const is really popular in JS... (i.e. could they surreptitiously be taking on Orinoco?)

[djwatson24]

I don’t think const or non const is the main issue- some data structures just require cycles like graphs or circular linked lists.

[Jweb_Guru]

I don't disagree, but I also suspect that a very large number of real world JavaScript applications have few to no application-induced cycles. I might even go further and say this extends beyond JavaScript--the most common reason I even define recursive data structures in a language like Rust is for AST-like data that almost invariably forms a tree.

[studius]

I supposed you could construct as dynamic and then make it constant? I wonder why that feature isn't seen in common languages, even though I know it sounds crazy.

[vnorilo]

Clojure has transient mutable collections that are designed to become persistent once "complete". [1] Some systems have made use of using reference count of one (unique owner) as permission to mutate, something that TFA generalizes as memory reuse.

1: https://clojure.org/reference/transients