Dzung Bui

Make FSTPostingFormat to build FST off-heap

10

### Description This is an attempt to make FSTPostingFormat to write the FST off-heap. Instead of write it on-heap then save to disk, we configure the compiler to write the...

Make Lucene90 postings format to write FST off heap

13

### Description Only the root block will be written off-heap while the sub blocks won't be (They are small so it might not be worth it: We would need to...

Fix Cohere dataset command in README

1

Seems like the README is outdated.

Add Query for reranking KnnFloatVectorQuery with full-precision vectors

32

### Description fixes #13564 Added a new Query which wraps around KnnFloatVectorQuery and does re-ranking for quantized index using full precision vectors. The idea is to first run KnnFloatVectorQuery with...

Can we support vectors to be loaded with direct I/O for full precision re-ranking?

1

### Description Spin-off from discussion in https://github.com/apache/lucene/pull/14708. One of the concern with with full precision (FP) re-ranking (for quantized vectors) is that if we use off-heap vector reader it will...

Optimize prefix sum

3

### Description Follow-up of https://github.com/apache/lucene/pull/15156, more optimizations for other places in codebase.

Add profiling support for DoubleValuesSource

1

### Description Lucene now has great [profiling support](https://github.com/apache/lucene/blob/main/lucene/sandbox/src/java/org/apache/lucene/sandbox/search/QueryProfilerIndexSearcher.java) for understanding search-time costs. It provides detailed breakdowns of time spent across sub-queries and scorers. I'm wondering if it makes sense to...

### Describe the blog post In this blogpost, we will describe recent substantial improvements to Lucene's FST (Finite State Transducer) data structure inspired by the Tantivy Rust search engine's own...