PostgreSQL: Ground Truth for Query Plans and Table Scans
Today's commits center on correctness fixes for planner and foreign-data statistics, plus groundwork—new test coverage and internal refactors—for an upcoming change to how index scans fetch heap data. No pull requests landed; all ten items are direct commits.
Duration: PT2M29S
Episode overview
This episode is a short developer briefing from PostgreSQL.
It explains recent repository work in plain language.
- Show: PostgreSQL
- Published: 2026-09-04T13:04:03Z
- Audio duration: PT2M29S
Transcript excerpt
This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.
Good morning. It's September 4th, and today's postgres activity is all about making sure the planner and executor trust accurate numbers.
The clearest thread is statistics correctness. Etsuro Fujita fixed a bug in postgres_fdw, commit 6a85715, where importing table statistics from a partitioned or inherited remote table produced different results than the normal sampling method. That mismatch could feed bad numbers into the planner and produce worse…
Related, Richard Guo's commit 55fa1b0 closes a planner correctness gap around lateral references and outer joins. Since a change last year removed an older safety check, the planner could, in rare cases, approve a join order that could never actually be satisfied—wasting planning effort or, worse, causing the…
The second theme is preparation for a bigger architectural change: moving index scan heap fetches behind a new table access method interface. Peter Geoghegan laid groundwork with three commits—a new test module for non-MVCC index scans (f7a563c), new coverage for index statistics counters in pg_stat_all_indexes…
A few smaller but useful items: Nathan Bossart tuned varbit bitwise operators for auto-vectorization, and separately…
Botto…
Nearby episodes from PostgreSQL
- Hardening Edge Cases in Replication and Test Stability
- Query Planner Correctness Fixes and Version 19 Cleanup
- Integer Overflow Cleanup Across the Tree
- Closing the Coverage Gap on Custom Scans
- Weekly Recap - Correctness Fixes and Shared Memory Hardening
- Dictionary Parsing Gets Hardened
- Correctness Fixes for Query Planning and Shared Memory
- Widening Fixes and Planner Corrections Land as Backpatches