Can LongDistanceFeatureQuery benefit from better sampling technique to evaluate iterator for competitive hits [LUCENE-10567]
- Dominant language
- Java
- Stars
- 3.6k
- Forks
- 1.4k
- Avg merge
- 2d 11h
- Merged PRs (30d)
- 88
Description
#11532 introduced an improvement in sampling technique to evaluate if we can iterate over a subset of points instead of doc values in sorting. This original code was inspired by how [LongDistanceFeatureQuery](https://github.com/apache/lucene/blob/main/lucene/core/src/java/org/apache/lucene/document/LongDistanceFeatureQuery.java#L361) computes competitive hits.
We should investigate if the same improvement in sampling technique can benefit LongDistanceFeatureQuery as well.
---
Migrated from [LUCENE-10567](https://issues.apache.org/jira/browse/LUCENE-10567) by Mayya Sharipova (@mayya-sharipova)
Contributor guide
Research direction
Start with LongDistanceFeatureQuery.java around line 361 and compare its competitive-hit sampling with the improvement introduced by issue #11532. Determine whether that sampling technique benefits LongDistanceFeatureQuery; the work is done when the investigation reaches a supported conclusion and any resulting behavior is validated.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- java
- Domain
- search
- Issue type
- Refactor
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100