This is interesting, though it really shouldn't be specific to one RDBMS (and I like PostgreSQL). To be honest, I don't find the status quo particularly diffic

This is interesting, though it really shouldn't be specific to one RDBMS (and I like PostgreSQL). To be honest, I don't find the status quo particularly difficult to deal with. A directory with lots of little files doesn't take up much storage, and nobody cares about the space anyway. That said, improvements are always welcome. I like how it looks at the diff and figures out simple schema changes. However, in my experience, the simple changes are no big deal. The more interesting challenge in migrations is when semantics change and you need to generate data from existing data. Simple diff cannot determine that, that needs to be provided by some set of queries, which easily fits in a migration file. Is there a way to integrate more complex changes and dataset manipulations when needed?