And performance ?
And performance ?
It is significantly faster than coreutils in some situations.
They also know cases where it is slower, and thus I doubt it'll remain so.
Long live the Rust Evangelism Strike force. Finally Apache/Linux is now possible.
I was under the impression that GNU Coreutils is weird/hacky in implementation because of 3 reasons (in no particular order): 1. Performance, 2. Portability, 3. To make it blatantly obvious that it's not copied from proprietary UNIX™ code. The last point is... unfortunate at this point in time, if unavoidable given the history. For performance, I'll be interested to see how it goes; Rust should help things compose better and lend itself to saner structuring, but the harder you lean into performance the more unavoidable complexity you incur, so it'll be interesting to find out how well new implementations can do. And portability, I suspect, will be the dump stat of any new implementation. Sure, it'll work on Linux, on x86 and ARM. Beyond that... I mean, last time I looked Rust didn't even support as many CPU architectures as Linux, let alone as many OSs as GNU does. I'm not sure if that's a problem or not; so far the precedent seems to be to just break anything unusual and not care (looking at you, Python cryptography), which is a win for the masses and unfortunate for everything else.
While true, especially with string heavy processing, using the borrow checker with CoW structures to cleanly save allocations does more for boosting performance than some hacky C++ code where you never know when it breaks because there are no guarantees on mutability and lifetime of variables.
Though then you can get the issue that systems don’t really tell you about P/E cores or how they’d want you to use them, and that’s annoying.