​
Stability Improvements 💪

  • Fixed an issue where the custom scan would throw an error instead of gracefully falling back to an index scan when boolean predicates were used
  • Fixed an issue where joins with the @@@ operator could exclude results from the output
  • Fixed an issue where “Top N” queries could enter an endless loop if enough documents in a segment had been deleted
  • Fixed an issue where the custom scan would solve an incorrect boolean condition if there were parentheses involved in the query
  • Fixed an issue where max_parallel_workers_per_gather setting was not being respected by parallel scans

​
Performance Improvements 🚀

  • Reduced index size and improved write throughput by eliminating some extra write overhead
  • The custom scan can now use fast fields when both string and numeric fields are present in the output columns

​
New Features 🎉

  • Introduced paradedb.snippet_positions which returns the byte offsets of highlighted text
  • Introduced support for the Jieba tokenizer

​
New Contributors 👋

​
Full Changelog

The full changelog is available here.