@hundreds-breakfast-49010 I didn’t catch up on this entire thread, so sorry if I’m repeating stuff, but… I think the reason we’re not using `ignore`’s

Walk

itself is that we want to cache subsets of the walk independently. That said, if there’s a nice way to use `ignore`’s

Walk

implementation and still have per-directory caching, that sounds great… I guess that would probably require

ignore

to (optionally) take a custom

readdir

function (which we could inject a caching version of), and because of how generics and dispatch work in rust, that would probably require dynamic dispatch for every

readdir

call, which I don’t imagine Andrew would love…