Drizzle Migration Squash

1. Inspect the current migration chain.

   • Read server/drizzle/meta/_journal.json.
   • Read the SQL files in server/drizzle/.
   • Inspect the latest snapshot in server/drizzle/meta/.

2. If the schema change also moves live data, apply that transition before squashing.

   • Generate or hand-write a normal migration on the current chain that creates the new table or column, backfills existing data, and removes the legacy shape.
   • Run just migrate so the local database is already on the target schema before rewriting history.
   • Keep that backfill logic out of the final squashed 0000_*.sql; the squashed baseline should describe the final schema for fresh databases, not legacy transition steps.

3. Fold the latest schema into the baseline migration.

   • Update server/drizzle/0000_*.sql so it directly creates the latest schema.
   • If later migrations added enum values, indexes, or constraints, move those into the baseline SQL.
   • Prefer direct CREATE TABLE constraints when possible, and keep separate CREATE INDEX statements when needed.

4. Collapse the metadata to one baseline snapshot.

   • Keep only server/drizzle/meta/0000_snapshot.json.
   • Ensure that snapshot reflects the latest schema, not the old baseline.
   • Set prevId in 0000_snapshot.json to 00000000-0000-0000-0000-000000000000.
   • Reduce server/drizzle/meta/_journal.json to one entry for the 0000_* migration.

5. Remove later migration files.

   • Delete server/drizzle/0001_*.sql and above.
   • Delete server/drizzle/meta/0001_snapshot.json and above.

6. Recompute the baseline migration hash.

   • Run shasum -a 256 server/drizzle/0000_*.sql.
   • Drizzle stores the SHA-256 of the SQL file in drizzle.__drizzle_migrations.hash.

7. Rewrite the local Drizzle migration journal in Postgres.

   • Inspect the table with:
     • docker compose exec -T kowalski_db psql -U kowalski_user -d kowalski -c "\d drizzle.__drizzle_migrations"
   • Replace its rows with the single new baseline:
     • TRUNCATE TABLE drizzle.__drizzle_migrations RESTART IDENTITY;
     • INSERT INTO drizzle.__drizzle_migrations (hash, created_at) VALUES ('<new_sha256>', <journal_when_value>);
   • This preserves the current schema and data while aligning Drizzle’s bookkeeping with the squashed files.

8. Verify.

   • Run just migrate and confirm it is a no-op or succeeds cleanly.
   • Run just ready.

Safety Notes

• Prefer rewriting drizzle.__drizzle_migrations over dropping the DB when local data should be preserved.
• If the local database is disposable, just clean-db followed by just migrate is a simpler fallback, but it destroys data.
• Do not squash migrations after they have been shared broadly or applied in environments you cannot rewrite safely.

Expected Outcome

• One baseline SQL migration in server/drizzle/
• One baseline snapshot in server/drizzle/meta/
• One row in drizzle.__drizzle_migrations
• just ready passes against the squashed migration history