Optimize AbstractKnnVectorQuery#createBitSet with intoBitset

[gf2121]

This tries to implement intoBitset to speed up the building of BitSet.

[github-actions]

This PR does not have an entry in lucene/CHANGES.txt. Consider adding one. If the PR doesn't need a changelog entry, then add the skip-changelog-check label to it and you will stop receiving this reminder on future updates to the PR.

[jpountz]

Instead of implementing intoBitSet on the iterator, we could do BitSet bitSet = BitSet.of(iterator, maxDoc); and then clear deleted docs with liveDocs.applyMask(bitSet, 0);? Ie. there wouldn't be a FilterDocIdSetIterator anymore?

[gf2121]

Thanks for feedback!

This is what i had in mind initially. The problem here is applyMask is requiring a FixedBitSet while BitSet#of may return a SparseFixedBitSet. So i have to fork the Bits#applyMask code for the sparse case, which could look like:

BitSet bitSet = BitSet.of(iterator, maxDoc);
if (liveDocs != null) {
  if (liveDocs instanceof FixedBitSet fixedBitSet) {
    liveDocs.applyMask(fixedBitSet, 0);
  } else {
    for (int i = bitSet.nextSetBit(0);
         i != DocIdSetIterator.NO_MORE_DOCS;
         i = i + 1 >= bitSet.length() ? DocIdSetIterator.NO_MORE_DOCS : bitSet.nextSetBit(i + 1)) {
      if (bitSet.get(i) == false) {
        bitSet.clear(i);
      }
    }
  }
}
return bitSet;

Or another potential way i'm thinking is making Bits#applyMask accept a BitSet. Default impl works fine but specialized impls need an instanceof FixedBitset check.

What do you think?

[jpountz]

Oh, this is annoying... It's not great either but I guess I'd like to do the SparseFixedBitSet/FixedBitSet decision up-front, something like:

int threshold = maxDoc >> 7; // like BitSet#of
if (it.cost() >= threshold) {
  // Create a FixedBitSet in an optimized way
} else {
  // existing code
}

[jpountz]

It looks like a similar change can be made to AbstractVectorSimilarityQuery.

[github-actions]

This PR does not have an entry in lucene/CHANGES.txt. Consider adding one. If the PR doesn't need a changelog entry, then add the skip-changelog-check label to it and you will stop receiving this reminder on future updates to the PR.

[github-actions]

This PR does not have an entry in lucene/CHANGES.txt. Consider adding one. If the PR doesn't need a changelog entry, then add the skip-changelog-check label to it and you will stop receiving this reminder on future updates to the PR.

[msokolov]

if liveDocs == null we could return Bitset.of(iterator, maxDoc), no, and avoid the internal logic in the filtered iterator, which probably goes away with branch prediction 🤷