Roman Titov
Roman Titov
Alternative solution would be to `split_by_shard` incoming operations, and only apply relevant part of it. Might be a bit cleaner than this one.
I think, alternatively, we can introduce custom [BitStore](https://docs.rs/bitvec/latest/bitvec/store/trait.BitStore.html) implementation that will be layout-compatible with `u64` on LE systems? Seems _a bit_ annoying to do, but I think it should not...
Tbh, I don't see much difference with #4867.
Closing, cause we implemented a different approach in https://github.com/qdrant/qdrant/pull/4928.
The name `ReshardingUpdatePreFilter` is intentionally hideous, cause I can't come up with a good enough one. Please help! 🥲🙏 @generall @timvisee
Closing, cause we implemented a different approach in #4928.
Sorry for a single >500 LoC commit, but this was **so damn annoying** to implement, I didn't have enough energy to do a proper commit history. 😬 Opened as draft,...
@generall I think the only thing left for this PR is to add some documentation to `HashRingCondition` fields, I'll do so once I'm back from a break. Other than that,...
Closing, cause we ended up with a different approach in #4928.
> Otherwise we would need to separate errors for internal and external interfaces, which would mean huge refactoring It will be a moderate refactoring, but I don't think we need...