The HypoPG extension, which adds support for Hypothetical Indexes so developers can test whether an index would help without actually building it, has reached version 1.4.4. Announced from Tainan, Taiwan, the release remains compatible with PostgreSQL 9.2 and above. This is a maintenance release focused on correctness and robustness, with fixes spanning index naming, memory handling and planner cache invalidation.
The concrete changes in this release:
- Fix the hypothetical index name for Oid greater than or equal to 2^31 (Devrim Gündüz).
- Fix a memory leak in
WITHclause options (Devrim Gündüz). - Check the maximum number of columns when included columns are present too (Devrim Gündüz, Julien Rouhaud).
- Invalidate hypothetical indexes when an invalidation message arrives for the dependent object when possible — covering relations, functions, access methods, operator classes and types — or on any invalidation message for all other object types (Julien Rouhaud, with initial work from Devrim Gündüz).
- Fix the
indcollationandindclasscolumns returned by thehypopg()function (Matt Blewitt). - Prevent a failed
EXPLAINfrom affecting later queries (Matt Blewitt, Julien Rouhaud). - Protect the hypothetical Oid range query against user objects (Julien Rouhaud, reported by Matt Blewitt).
- Additional PostgreSQL 19 compatibility fixes (Julien Rouhaud, with one issue reported by Nacho Alonso Portillo).
Why this matters: HypoPG is a common tool for index tuning, letting teams see how the planner would behave with an index before paying the cost of creating it. The invalidation and hypopg() column fixes make the catalog data and planner behaviour more accurate, while the robustness fixes around EXPLAIN and Oid handling reduce surprising side effects. If you rely on HypoPG in production tuning workflows, upgrading to 1.4.4 is worthwhile, especially if you are preparing for PostgreSQL 19.
Source: https://www.postgresql.org/about/news/hypopg-144-is-out-3421/