Help us test whether unlogged tables speed up your PostgreSQL test suite

Hi, if you have a test suite with many DB tests that use PostgreSQL, we would like to hear about the impact on suite execution with this feature.

Details here: Experiment: Support for creating unlogged tables to speed up tests by mgibowski · Pull Request #601 · elixir-ecto/ecto_sql · GitHub

1 Like

This came up a few weeks back on twitter, but initial experiments didn‘t seems to provide any tangible speedups. I would expect that‘s because with the ecto sandbox most writes are never committed in the first place with ecto.

Somebody reported some good results:
https://x.com/nelsonmestevao/status/1741097876564984299?s=20
I tried using unlogged tables some time ago on a project I no longer have access to. It had around 400 tests, and the improvement was about 10% / 1.5 seconds.
I tried that branch yesterday on changelog.org and the hexpm codebases, and I cannot tell there is any improvement. Test execution times are sometimes faster and sometimes slower.

I ran it at the time on a ~900 tests project of mine and had similar results.