I have an in progress recursive directory walk crate jwalk. It’s distinguishing features are:
- Walk is performed in parallel using rayon
- Results are streamed in sorted order
It is inspired by both
ignore. I attempts to match (or better) the performance of
ignore::WalkParallel while also providing streamed sorted results like
cargo bench to see relative performance for walking linux repo. Thanks @BurntSushi!
It’s incomplete, missing many of the details provided by those crates. In particular it isn’t filtering results based on ignore rules. But I “think” the code is flexible enough to support these things cleanly.
I’m looking for feedback:
Would anyone want to use this over what’s already available? Should I make the effort to complete and publish it.
This is my first maybe public crate, I’d love for anyone to look through the code and tell me how to make it better.
I’m particularly interested in getting best possible performance while using rayon. Performance is good right now on my computer. But I don’t really understand all the details in the rayon stacks when profiling… is utilization good or time being wasted somewhere?