jerry-024 commented on code in PR #72:
URL:
https://github.com/apache/paimon-vector-index/pull/72#discussion_r3746985963
##########
core/src/index.rs:
##########
@@ -1843,18 +1846,19 @@ impl<R: SeekRead> VectorIndexReader<R> {
let total_vectors = usize::try_from(reader.total_vectors)
.map_err(|_| invalid_input("negative IVF vector count"))?;
let nprobe = params.resolve_ivf_nprobe(reader.nlist,
total_vectors, None)?;
- progressive_ivf_search(
+ progressive_ivf_batch_search(
params,
reader.nlist,
nprobe,
+ queries,
query_count,
params.top_k,
total_vectors,
- |nprobe| {
+ |active_queries, active_query_count, nprobe| {
Review Comment:
**[minor] Preserve statistics for the logical batch operation**
Each automatic retry calls `search_batch_ivfrq_reader` with
`active_query_count`. That function initializes its aggregate `query_count`
from this subset and overwrites `reader.last_search_stats`, so after
progressive expansion `ivfrq_search_stats()` describes only the final retry and
omits work performed for queries completed in earlier rounds.
Please aggregate IVF-RQ statistics across the internal rounds and publish
one record with the original batch query count. The roaring-filter IVF-RQ path
needs the same treatment.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
