0.24.2

​

New Features 🎉

• Expose index build time via pdb.index_created_at().
• Add nfkc and reading_form options to the Lindera tokenizer.
• Add minimum_should_match parameter to paradedb.boolean().
• The Docker image will automatically choose sensible defaults for CPU and memory on first use.

​

Performance Improvements 🚀

• Allow ORDER BY score Top-K queries to get the Block-Max WAND optimization in the presence of trailing tiebreaker columns.
• Significantly reduce the size of unicode_words indexes by improving the density of positions.

​

Stability Improvements 💪

• Gate use of the aggregate and join scans on the presence of a partial index predicate.
• Fix an issue with heap filters representing constants in the aggregate scan.
• Fix missed timestamp-rewriting sites and add time/timetz parsing support.
• Fix potential crash when receiving a UNIX signal during mutable segment reading.
• Fix ORDER BY score Top-K queries with alternate join shapes in the join scan.
• Fix incorrect mutability classification of pdb.alias, preventing its use in read-only queries.
• Soften the datetime_fields deprecation to a warning.