#​621 — October 23, 2025

Web Version
Together with  TigerData logo

Postgres Weekly

Is Your Database Read Heavy or Write Heavy? — Most databases will tend to be heavier on reads than writes, but it depends on your workload. Is it worth knowing for sure? When it comes to tuning, yes. David shows off an epic SQL query to figure out the read:write ratio for your tables.

David and Elizabeth Christensen

💡 The query in the post above works fine, but if you want to also include tables that are read-or-write only, change the AND on line 19 to OR.

Tiger Lake: Bridging Postgres and The Lakehouse — Tiger Lake unifies Postgres with your lakehouse in real time. No pipelines, no orchestration. Stream, analyze, and serve data continuously using open formats. Build dashboards, agents, and analytics powered by one integrated, Postgres-native architecture.

TigerData sponsor

Tiger Data Introduces a Free Postgres Plan — Tiger Data, formerly Timescale, has offered a Postgres platform for some time now, but never a free tier. Now you can get up to two services for free with up to 750MB storage on each. The pitch is being made towards AI and agent developers, due to the state of the industry right now, but could be used by anyone for small projects or experiments.

TigerData

💡 Other providers like Aiven, Neon and Supabase offer similar free tiers.

🔎 Hybrid Search in Postgres: The Missing Manual — An explanation of how full text search with BM25 scoring (powered by ParadeDB) coupled with vector similarity search (powered by pgvector) trumps Postgres’s native full text search when it comes to relevance ranking.

James Blackwood-Sewell

The Benefits of a DESCending Index — “The well-known use case for the DESC clause in an index definition is support for a mixed ORDER BY clause. But there are corner cases in which PostgreSQL can populate a descending index more space-efficiently, and a forward scan on a descending index can perform better than a backward scan on an ascending index.”

Laurenz Albe

▶  14x Faster with 12x Less Compute: Sometimes Postgres Really is All You Need — How James’ team went from an unreliable 12 server HBase/OpenTSDB cluster to two servers with 100% uptime with Postgres/Timescale for its time series workload.

James Udiljak

📄 Getting a Portable Setup for Postgres 18 on Windows – A handy walkthrough. Mohit Sindhwani

📺 Proper Postgres Parameters to Prevent Poor Performance – An hour-long webinar recording. Greg Dostatni

📄 PostGIS Performance: pg_stat_statements and Postgres Tuning Paul Ramsey

RELEASES AND CODE:

NoraSearch: Extension to Quickly Find Substring Match Offsets and Counts — For example, searching for 'abra' in 'abracadabra' would return {{0,4},{7,4}}. The author cites one use case as being for analyzing DNA sequences.

Lemmer EL ASSAL

pg_ivm 1.13: Incremental View Maintenance (IVM) Extension — A more efficient way to update materialized views incrementally, applying only changes instead of fully recomputing the view like REFRESH MATERIALIZED VIEW does. v1.13 adds support for outer joins.

IVM Development Group

PL/Haskell 5.0: Use Haskell in Your SQL Functions — Haskell is a functional language rather than a procedural one, but you can nonetheless use it to define procedures in a way Postgres can work with. v5 adds support for date and time types.

Edward F. Behn, Jr.