0.15.9 - ParadeDB

Deprecation Notice 🗄️

As of this release, we have deprecated pg_search support for PostgreSQL 13. This decision was taken to enable our team to better serve our existing customers on other PostgreSQL major versions and in anticipation of the upcoming release of PostgreSQL 18 later this year and subsequent deprecation of PostgreSQL 13 later this year.
On April 2, 2025, we will be deprecating pg_analytics, our extension for reading data from external data lakes into Postgres. The pg_analyticsrepository will be archived and will no longer be maintained. The extension will be removed from our Dockerfile and Helm charts. This decision was made because our work on Postgres analytics is being done in our primary extension,pg_search. Additionally, while pg_analytics was an exciting proof of concept, there now exist other tools in the ecosystem that can be used to read from object stores, and we would like to help consolidate usage around those tools rather than compete with them.

We have significantly improved write throughput by optimizing our segment merge policy. As a result, we removed the paradedb.max_mergeable_segment_size and paradedb.segment_merge_scale_factor settings, which are no longer necessary under this new merge policy. We instead introduced a new index WITH option, layer_sizes, which which defaults to 100KB, 1MB, 100MB, but can be set by the user either at CREATE INDEX time or later by ALTER INDEX.
We added support for concurrent merging, which further contributes to write throughput improvements. As a result, we removed the paradedb.is_merging() function. We instead introduced the paradedb.merge_info() and paradedb.vacuum_info() functions.

Fixed an issue where certain queries could come out without a BM25 score
Fixed an issue where REFRESH MATERIALIZED VIEW CONCURRENTLY failed with pg_search installed
Fixed an issue where the max_num_chars parameter was not being respected while querying for highlights using the snippets function
Fixed an issue where certain queries would have missing rows when querying data with NULL values in a string fast field

The full changelog is available here.