pg-extension-lablisted

# pg-extension-lab — a harness to build, test, scenario, benchmark, and tune a PostgreSQL extension **pg-extension-lab is a harness for validating a PostgreSQL extension's observable behavior in an isolated, reproducible environment, separately from its implementation** — covering **three architecture shapes** and **five reference categories**. Use it not only to build an extension from scratch but to design tuning experiments, write scenarios, and run isolation/regression tests against an existing one. Start with the mental model, identify your shape, then open the category that matches the request — each category has a `README.md` index that links to dense, single-topic detail files (progressive disclosure: this file → category README → detail file). Core mindset: **[`references/mental-models.md`](references/mental-models.md)**. ## Architecture shapes this covers ``` Shape A — pure C / PGXS or Rust/pgrx extension (e.g. a custom index access method) SQL caller → access method / planner hook / pg_extern function → on-disk index pages Shape B — DB extension + external microservice (the service OWNS data) sync: PostgreSQL (SQL function) → HTTP → service (FastAPI/gRPC) → SQL → database async: SQL function → outbox row + NOTIFY → LISTEN worker → service → UPDATE result table Shape C — DB extension + co-located compute sidecar (the daemon owns a costly RESOURCE, not data) N backends → Unix socket/shm → one long-lived daemon holding a device context / model / pool