Indistinguishable hashes in bucket!

[KraftRicardo]

Hi, while using the crate I ran into this console output. Anything I should worry about?

I am not even 100% sure this comes from PtrHash, but I wanted to make sure. Thank you!

part 0: bucket of size 15 with 51 slots: Indistinguishable hashes in bucket!
4e91a1824b73d77 -> slot 47
5e9998ea69bb799 -> slot 26
6e8197728c02ecb -> slot 27
7e898fdaa34a1fd -> slot 26
eff0a930368b480 -> slot 40
fff8a19855d2fb2 -> slot 27
21a4216548d329cf -> slot 43
22a4a0ebca07a0f1 -> slot 13
23a720504c6c1b23 -> slot 7
24a7a7c6ce509255 -> slot 23
37ae30ea6a11b1b6 -> slot 12
38aeb050ec4628d8 -> slot 24
39d137d96daaa30a -> slot 27
3ad1b74fef9f1a3c -> slot 45
47eb8c091dbc0802 -> slot 29
part 0: bucket of size 15 with 51 slots: Indistinguishable hashes in bucket!

[RagnarGrootKoerkamp]

Hey,

Yeah this is because the construction is probabilistic, and can fail at times. But if this is all you're seeing, then a new random seed was chosen and construction was automatically retried.

I'll leave this open for me to improve the log messages here.

Just for my information, how many keys and what parameters were you using here?

[KraftRicardo]

Here you go! This is the code snippet:

  fn build_double(pair_data: PairData, cutoff: usize) -> Self {
      let keys = pair_data.keys;
      let input_values = pair_data.values;
      
      println!("PtrHashDouble num keys: {}", keys.len());

      // Build minimal perfect hash function (mphf)
      let ptr_hash = <PtrHash<usize>>::new(&keys, PtrHashParams::default());
      // Sort values depending on the new mphf
      let mut values: Vec<u16> = vec![0; input_values.len()];
      for (i, key) in keys.iter().enumerate() {
          values[ptr_hash.index(key)] = input_values[i];
      }
// more code

and output:

PtrHashDouble num keys: 2260851
part 0: bucket of size 15 with 51 slots: Indistinguishable hashes in bucket!
4e91a1824b73d77 -> slot 47
5e9998ea69bb799 -> slot 26
6e8197728c02ecb -> slot 27
7e898fdaa34a1fd -> slot 26
eff0a930368b480 -> slot 40
fff8a19855d2fb2 -> slot 27
21a4216548d329cf -> slot 43
22a4a0ebca07a0f1 -> slot 13
23a720504c6c1b23 -> slot 7
24a7a7c6ce509255 -> slot 23
37ae30ea6a11b1b6 -> slot 12
38aeb050ec4628d8 -> slot 24
39d137d96daaa30a -> slot 27
3ad1b74fef9f1a3c -> slot 45
47eb8c091dbc0802 -> slot 29
part 0: bucket of size 15 with 51 slots: Indistinguishable hashes in bucket!

This happens during one of my tests. In this test I call ptrhash several times in a row. This only happens when I run my test in that specific order. I do not get it when I single run the test that causes it. That is why I was worried.

If you need more info please let me know!