Tracking issue for a DevOps Daily post on what's new in Postgres 19, with a benchmark-driven angle.
The hook
PG18 landed with checksums-on-by-default and the async I/O work. PG19 is the next major (beta over 2026, GA expected around Sept/Oct 2026). Bobby's specific curiosity: the graph query work and whether it actually improves query performance, "like a kind of relation index."
That most likely points at SQL/PGQ (Property Graph Queries, SQL:2023) landing in Postgres, letting you run graph-style pattern matches over existing relational tables via GRAPH_TABLE(...) instead of hand-rolled recursive CTEs. Confirm what actually made the PG19 cut before writing (feature freeze / release notes) rather than assuming.
Questions to answer in the post
- What graph-query support actually shipped in 19? Is it full SQL/PGQ
GRAPH_TABLE, or a subset? What's still missing vs Neo4j-style engines?
- Is it a real execution improvement or just nicer syntax over the same recursive-CTE plans? i.e. does it add a genuine index/traversal strategy, or does it compile down to the same joins?
- Benchmarks: multi-hop traversal (friends-of-friends, dependency graphs, org charts) as
GRAPH_TABLE vs the equivalent WITH RECURSIVE. Latency + plan shape at a few depths and graph sizes.
- Other notable 19 features worth a section: whatever's headline in the release notes (skip-scan / index work, planner changes, logical replication, async I/O follow-ups). Pick 2-3 that matter operationally.
Angle / differentiation
Not a vendor-blog feature list. Lead with the graph-query question and back it with our own numbers (we have the Neon/Supabase bench harness + a Neon Postgres we can point at a 19 preview if available). Show the query both ways, the EXPLAIN plans, and a chart. Be honest if it turns out to be syntax sugar rather than a speedup.
Assets
chart fence for the traversal benchmark (bar or line by depth), terminal fence for the psql session, maybe a diagram for the graph shape.
- OG image via the usual
npm run images:* flow.
Prereqs before drafting
- Verify PG19 status + whether SQL/PGQ is in (release notes / commitfest), so we don't publish claims that get cut.
- Check
content/posts/ for any existing Postgres 18/19 post to cross-link and avoid overlap.
Category: likely cloud or a DB-leaning one. Not started.
Tracking issue for a DevOps Daily post on what's new in Postgres 19, with a benchmark-driven angle.
The hook
PG18 landed with checksums-on-by-default and the async I/O work. PG19 is the next major (beta over 2026, GA expected around Sept/Oct 2026). Bobby's specific curiosity: the graph query work and whether it actually improves query performance, "like a kind of relation index."
That most likely points at SQL/PGQ (Property Graph Queries, SQL:2023) landing in Postgres, letting you run graph-style pattern matches over existing relational tables via
GRAPH_TABLE(...)instead of hand-rolled recursive CTEs. Confirm what actually made the PG19 cut before writing (feature freeze / release notes) rather than assuming.Questions to answer in the post
GRAPH_TABLE, or a subset? What's still missing vs Neo4j-style engines?GRAPH_TABLEvs the equivalentWITH RECURSIVE. Latency + plan shape at a few depths and graph sizes.Angle / differentiation
Not a vendor-blog feature list. Lead with the graph-query question and back it with our own numbers (we have the Neon/Supabase bench harness + a Neon Postgres we can point at a 19 preview if available). Show the query both ways, the EXPLAIN plans, and a chart. Be honest if it turns out to be syntax sugar rather than a speedup.
Assets
chartfence for the traversal benchmark (bar or line by depth),terminalfence for the psql session, maybe adiagramfor the graph shape.npm run images:*flow.Prereqs before drafting
content/posts/for any existing Postgres 18/19 post to cross-link and avoid overlap.Category: likely
cloudor a DB-leaning one. Not started.