docs/changelog/0.21.6.mdx
ORDER BY lower() was not getting an optimized Top K scan if lower() was literal normalizedu32 overflow issue that could cause queries to fail over large segmentsGROUP BY queries over datetime fields would error with paradedb.enable_aggregate_custom_scan enabledImproved the filtering performance of partial indexes. For example, in the following query deleted_at IS NULL
gets fully pushed down to the index.
CREATE INDEX idx ON profiles USING bm25 (...) WHERE deleted_at IS NULL;
SELECT * FROM profiles WHERE headline ||| 'recruteur' AND deleted_at IS NULL;
The full changelog is available on the GitHub Release.