Apache Lucene 9.0 Released – New Features and Improvements

Apache Lucene 9.0 has been released. Lucene is a fully Java‑based, high‑performance, feature‑complete full‑text search engine framework that offers a complete query and indexing engine as well as partial text analysis capabilities, aiming to give developers an easy‑to‑use toolkit for adding or building full‑text search functionality.

Main updates

Support for indexing high‑dimensional numeric vectors to perform nearest‑neighbor search using a hierarchical navigable small world graph algorithm.

New analyzers for Serbian, Nepali, and Tamil languages.

IME‑friendly auto‑suggestions for Japanese.

Snowball 2 adds stemming for Hindi, Indonesian, Nepali, Serbian, Tamil, and Yiddish.

New normalization/stemming features for Swedish and Norwegian.

Faceted taxonomy speed increased by 400%.

Indexing speed for multi‑dimensional points improved by 10‑15%.

Sorting speed for point‑based fields improved several‑fold; now an optional feature in 9.0 as it was in later 8.x releases.

ConcurrentMergeScheduler now assumes fast I/O, potentially boosting indexing speed when heuristic detection of modern I/O fails.

Posting list encoding switched from FOR‑delta to PFOR‑delta for further disk‑space savings.

File formats changed from big‑endian to little‑endian order.

Lucene 9 no longer contains split packages, requiring package renaming outside lucene‑core JAR and corresponding import adjustments.

Using Lucene 9 in the module system is considered experimental.