Migration Planning & Safe Schema Evolution Change durable data without treating migrations as generated noise
Phase 10 - Relational Data Modeling Volume 03 - Database Engineering
↓
↓
Phase 10 decides what the durable data should mean before Phase 11 writes application behavior against it.
A schema change is a data change Once production data exists, changing schema can affect stored rows, constraints, application compatibility, indexes, lock duration, and rollback strategy. Migrations deserve code review.
↓
↓
↓
↓
↓
The migration file is part of the application codebase and deployment story.
Additive changes are easiest Adding a nullable column is usually easier than immediately adding a required column to a populated table. For required new data, use a staged migration: add nullable/default, backfill, validate, then enforce NOT NULL.
Example staged change Step Action
An ORM may interpret a field rename as drop-old + add-new, which can lose data. Review generated SQL and use deliberate SQL rename operations or migration edits when preserving existing data matters.
Environment workflow Environment Command intent
A rollback is not automatically "run the migration backward." Many database changes are irreversible after data is transformed or deleted. Production change planning should prefer forward fixes, backups, compatibility windows, and explicit restore strategy.
Phase 10 migration plan • Replace the minimal Todo-only learning model with the TaskFlow production model in a controlled development migration. • Review generated SQL before applying. • Use a development database/shadow workflow only. • Verify tables, foreign keys, uniqueness and indexes after migration. • Do not seed production-like identities until authentication design is introduced.
Prisma improves developer ergonomics, but migrations and generated queries still change a real relational database. You should be able to inspect both sides.
Restate the goal of “Migration Planning and Safe Schema Evolution” as one observable behavior. Write the success signal before touching code.
Trace the boundary map: domain intent → Prisma Client → generated SQL → PostgreSQL → mapped result. Mark the one boundary this lesson changes.
Inspect the current repository state and record the exact files, route, table, or runtime involved. Avoid creating an abstraction before you can name the existing owner.
Implement the smallest vertical slice. Add one coherent change, run the narrowest verification command, and read the output before continuing.
Connect the slice to TaskFlow’s phase artifact: A complete Prisma/ER model for TaskFlow core entities with keys, relation tables, uniqueness, and important indexes.. Verify behavior from the outside as a user or API client would.
Create one controlled failure that is safe to reproduce. Predict the symptom first, capture evidence, locate the earliest broken boundary, then fix only that cause.
Close the reference and rebuild or explain the core change from memory. Record what you would monitor or test before trusting the same change in production.
const tasks = await prisma.task.findMany({
where: { workspaceId },
orderBy: [{ createdAt: "desc" }, { id: "desc" }],
take: 20,
});Do not copy this as a finished answer. Predict what each boundary does, type the smallest useful slice, run it, then compare the result with your prediction.
Verify connection string, host, port, SSL/pooler mode, network reachability, and which environment file the process loaded.
Read the generated SQL before applying it and compare the change against existing production-shaped data.
Without copying the final source, create a second small scenario that applies “Migration Planning and Safe Schema Evolution” to a different TaskFlow resource or input. Write the acceptance criteria first, implement it, and provide the evidence that each criterion is satisfied.
Model boards/lists/tasks and decide which parent IDs should be required.
Use the smallest change that demonstrates this lesson. Stop after it works and capture evidence.JSON arrays replacing relational join tables without reason
Predict the symptom before causing the failure. Then locate the earliest boundary where reality diverges from expectation.Why does this responsibility live here, what input does it trust, what output proves success, and what would go wrong if another layer owned it?
If your explanation is vague, repeat the boundary trace before moving on.Close the reference. Reproduce the core idea from memory or explain the exact request/data flow aloud. Then check your diff or notes for accidental shortcuts that bypass validation, authorization, error handling, or verification earned in earlier phases.