[gregplaysguitar]

IMO the best way to use this is as a way to determine what’s changed between your canonical dB schema (well organised and easily understood DDL) and migrations (which are what gets deployed, but don’t work well as a dB reference). So you edit the schema, then use migra to generate a candidate migration for the changes which you review before deploying.

Tusker (mentioned in other comment) does it this way; my company has used the technique very successfully for several years via some lightweight tooling built around migra.