I've been using pg_regress tests for a while. It's generally a great way to ensure the behavior of your Postgres code works and continues working as expected. However, as my tests became larger, I started getting lost in them; and there are limits as to what you can test by having a psql session.
Postgres extensions are great. They enable fantastic use cases and bring new capabilities to one of the most loved open source databases. But there are edges in some of its features and this can be heard in conversations: limitations of the upgrade system, lack of parameterized initialization, search path/OID resolution issues, hard-wired dependency on .control files, schema droppage 1, etc.
However, the beauty of it is that what we ultimately want from extensions does not need to use CREATE EXTENSION's framework.
At times, representing the variety of types a value can take through multiple tables can taxing, both in terms of development complexity as well as, potentially, performance. You must just need a value to be of any of the given variants. I've set out to build a generalized mechanism for defining these.
As a Rust enthusiast and a contributor to a sister project I am stoked about the release of PL/Rust 1.0.0 that was just announced.
However, its setup instructions are rather long and it takes time to build it. So I took the time to prepare a build for you to try.
Have you ever wondered if Postgres can be a fully self-sufficient platform for your application? Learn how to make it become an application server and win some prizes, too!
Benchmarks and performance claims are attention-grabbers, but that's not what drew me to work on Omnigres. When I first built a prototype of its HTTP server, I didn't foresee the desire to share the numbers. As we all know, getting benchmarks right is hard, and everybody's mileage may vary. But I'll show you some numbers here anyway. It'll be great to validate or invalidate my findings!
Omnigres is a new project to turn Postgres into a complete development and production deployment platform. I've started it to reflect on the complexity and inefficiencies plaguing modern business software development.
As an aging (and sometimes cranky!) developer, I crave simplicity. But that's a topic for another post. Here I wanted to address a common question:
Why didn't you implement this in Rust?