Alembic Migrations Roadmap

[pmdevita]

• Upgrade
  • Upgrade only a specific component
  • Upgrade to a specific checkpoint
  • Improve command verbosity
• Downgrade
  • Downgrade to a specific checkpoint
  • Improve command verbosity
• Make Migrations
  • Detect table/column renames and ask user for input
  • Detect new non-null column being added and get a default value
  • Improve command verbosity
    • Add/Remove models
    • Add/Modify/Remove columns
• Testing?

Some things will just need to be determined through usage:

• Atsume Alembic currently creates migrations by simulating all previous migrations on an in memory database and then doing the diff with that, whereas vanilla Alembic expects to be running against your already in use database. Does this cause issues or result in poor performance after a lot of migrations?
  • We can look into squashing migrations if this happens

[pmdevita]

Regarding the in memory database, Prisma is an example of another ORM with migrations generating an in-memory database to detect schema differences, although they do it just for error checking (the feature is called Shadow DB). This gives me a bit more confidence in the viability of this approach