Stability Improvements 💪

  • In ParadeDB Enterprise, fixed an issue with physical replication where read replicas would sometimes error if a WAL record was being concurrently applied
  • Deprecated the raw tokenizer in favor of the new keyword tokenizer, which fixes an issue where raw was lowercasing tokens
  • Fixed an issue where segments created by CREATE INDEX could get merged in the next INSERT/UPDATE if they were under the max layer size, which could be a very expensive operation

Performance Improvements 🚀

  • Landed some initial work on improving the query performance of partitioned indexes by pushing down ORDER BY...LIMIT

Full Changelog

The full changelog is available here.