~postgresweekly | Issue 544 - Postgres in WebAssembly (12)

The azure_ai Extension to Azure Database for PostgreSQL — One benefit of Microsoft’s close relationship with OpenAI is that it’s available directly within Azure, and now you can directly work with Azure OpenAI from within Azure Database for PostgreSQL using the azure_ai extension, as well as pgvector to work with any embedding-based results of that work. Claire shows off how this all looks in...

PGlite: Postgres in WebAssembly — People have brought Postgres into WebAssembly contexts before, though usually with a Linux VM layer sat in between. PGlite packages a WASM build of Postgres into a TypeScript library that can be run in the browser or on Node.js or Bun and is only 3.7MB gzipped! An exciting development - we look forward to seeing where it goes next.

The State of the Database Landscape 2024 — Redgate’s latest survey results show that PostgreSQL continues to be in the top four most used databases as organizations continue to embrace multiple database platforms. Find out more about the latest trends in the database landscape.

The Rest is History: Investigations of WAL — Get a deeper insight into just what the WAL files used by Postgres contain as Brian steps through reviewing the Write-Ahead Log history as part of a sample recovery scenario.

11 Lessons to Learn When Using NULLs — NULL represents a missing, unknown or ‘non’-value in SQL and it has some interesting properties and behaviors to be aware of.

The Two Types of Indexes in pgvector — pgvector is a popular extension for working with and querying against vectors, and vectors demand different index types from the norm. Semab compares the IVFFLAT and HNSW options on offer.

The Default Value of fdw_tuple_cost Was Updated to 0.2 – Why? — A quick look at a recent change to Postgres.

When Less is More: Database Connection Scaling — The more connections a Postgres server can handle, the better.. right? Not always. If max_connections is set to an unrealistic number, even if those connections aren’t being used, negative effects can be felt. (A classic article, but potentially worth revisiting.)

pg-lock-tracer: An eBPF-Based Lock Tracer — Use eBPF instrumentation to dig into lock-related bottlenecks in Postgres when running on Linux. Unlike with pg_locks, you get a continuous view of the locking activity and timings, and can produce animated views of locks.

pgagroal 1.6: A High Performance Postgres Connection Pool — Introduces support for CLI/JSON output, client certificates, as well as macOS.

pg_dumpbinary v2.15: Dump a Postgres Database to Binary Format — Rather than produce a SQL dump, this produces a binary one (which can be restored by the accompanying pg_restorebinary tool). There are various edge cases where a binary dump may be preferable, such as when using (very) large bytea columns or custom types that use \0 values that can break standard dumps.