What's New in Postgres 18 & 19
summary
Postgres 18/19 add concrete, usable features: UUID v7 with timestamp-ordering lets UUID primary keys be b-tree indexed; Virtual generated columns appear as computed non-stored fields the ORM can map; a built-in repack (online table rewriting) fixes bloat by copying live rows and swapping with minimal locks; async I/O claims 2–3× faster reads (benchmarks often show ~1.5×); index skip scanning lets multi-column indexes match on later columns; OAuth2 auth and SCRAM-256 replace MD5; early graph and object-storage extensions (pg_lake, pg_duckdb, Apache AGE) appear.
AI-written summary · about
chapters (18)
- 0:00Introduction and welcome
- 1:15Postgres version overview
- 2:26UUID improvements in Postgres
- 4:42Virtual generated columns
- 6:07Returning old and new values on update
- 7:33Insert on conflict select
- 7:55Temporal constraints with b-tree gist
- 10:05Database constraints and window functions
- 11:15Async I/O and performance upgrades
- 12:21Index skip scanning for multi-column indexes
- 13:31Parallel index rebuilds
- 13:52Re-pack feature for table bloat
- 16:47Parallel vacuum workers and OAuth2
- 18:13Password hashing updates
- 18:54Query plan hints in Postgres 19
- 20:22Postgres extensions for analytics
- 21:53Graph queries in Postgres 19
- 23:45Summary and upgrade advice
