ideas

[SEAPUNK]

database initialization:

1. /schema.sql or /schema/ directory (schema .sql files are executed in alphabetic order)
2. /seeds.sql or /seeds/ directory (alphabetic order)

migrations:

• /migrations/ folder, format is TIMESTAMP_NAME.sql
  • TIMESTAMP is the current UTC time, in YYYYMMDDHHMMSS, like Rails' ActiveMigration
  • TIMESTAMP is also essentially the id of the migration, so there cannot be multiple migrations with the same timestamp
  • NAME is the description of the migration (no format in particular)
• schema migrations are kept track of in schema_migrations table of the database
• grater migrate [TIMESTAMP] - migrates schemas. runs all migrations that have not been run (load schema_migrations folder). if timestamp is provided, then that specific migration is run (and error thrown if grater notices that the migration has been run before, saying you gotta run it with --force)

[SEAPUNK]

now the question here is:

what do you do if there is no schema_migrations table?

[SEAPUNK]

A: do not do anything, just exit out with an error

have multiple commands to resolve this:

• grater init new - creates the new database with the new schema and seeds (if database exists, require the --drop-existing flag)
• grater init fresh - just creates the schema_migrations table, and populates it with the current migrations, to indicate that the current database is already completely migrated
• grater init stale - creates the schema_migrations table, and then runs migrations. --interactive indicates that the user wants to choose which migrations to run, mark as completed without running, or skip.

[SEAPUNK]

should migrations have a "check" function, to make sure that the migration is applicable?

[SEAPUNK]

mmmmmaybe. i'll think about it when the time comes to consider it