Quality issue: permutations produce the same hashes? about gxhash HOT 4 CLOSED

Hello @Dherse,

The "standard" hash function gxhash64(input: &[u8], seed: i64) -> u64 is backed by smasher tests which has permutation tests and more. However, the Hasher is a different approach, and while it uses the core of the gxhash algorithm, it is not backed by smasher and is covered by very few tests regarding quality (only a few sanity checks).

Given that the rust's default Hasher seems to generate order-dependent hashes, I'd say that the fact that GxHasher is currently order-independent is a bug (I easily reproduced).

Fixing this will only affect GxHasher so I think it can be addressed as part of a hotfix version, I'll see what can be done. Also, some time must be put into adding a few more quality tests on this Hasher side.

from gxhash.

Thank you, I am a large contributor to https://github.com/typst/typst and I am trying to see whether gxhash could replace Siphash 1-3 that we're currently using since it has so much higher performance hence #7 and this issue!

Anyway, looking forward to it 🎉

from gxhash.

My bad, issue is still open, I only tackled the quality bench part for now

from gxhash.

Fixed and published in v2.3.1 🎉

from gxhash.