Drizzle ORM Migrations

Manage database schema changes using Drizzle ORM with SQLite through decision-driven migration strategies.

DECISION POINTS

Schema Change Strategy Decision Tree

Is this a production database?
├─ YES → Always use generate + migrate workflow
│  ├─ Breaking change (column removal, type change)?
│  │  ├─ YES → Multi-step migration with data preservation
│  │  └─ NO → Standard migration generation
│  └─ Performance impact (large table, complex index)?
│     ├─ YES → Plan maintenance window, test on replica
│     └─ NO → Standard migration during low traffic
└─ NO (development) → Can use push for speed
   ├─ Breaking existing local data?
   │  ├─ YES → Backup first, consider fresh DB
   │  └─ NO → Safe to push directly
   └─ Collaborating with team?
      ├─ YES → Use migrations for consistency
      └─ NO → Push is acceptable

Column Type Selection Matrix

Data Type	Use	Drizzle Type	Index Recommended
User ID	Primary/Foreign keys	text('id').primaryKey()	Auto on PK/FK
Email	Unique identifiers	text('email').unique()	YES
Timestamps	Created/updated times	text('created_at').default(sql\CURRENT_TIMESTAMP`)`	If filtered
JSON Data	Flexible schemas	text('data', { mode: 'json' })	NO (use computed)
Booleans	Status flags	integer('active', { mode: 'boolean' })	If filtered
Counters	Stats, quantities	integer('count').default(0)	If aggregated

Migration Conflict Resolution

Migration fails to apply?
├─ Schema conflict detected
│  ├─ Column exists → Check if type matches expected
│  ├─ Table exists → Verify schema matches or rename
│  └─ Index exists → Check if definition identical
├─ Foreign key violation
│  ├─ Referenced table missing → Create dependency first
│  ├─ Referenced column missing → Fix reference or add column
│  └─ Circular reference → Break into separate migrations
└─ Data type mismatch
   ├─ Incompatible conversion → Create transformation migration
   ├─ NULL constraint violation → Add default or migrate data
   └─ Unique constraint violation → Clean duplicate data first

FAILURE MODES

1. "Push to Production" Anti-Pattern

Symptoms: Using npm run db:push on production database Detection: If environment is production AND using push command Fix:

• Stop immediately
• Generate proper migration: npm run db:generate
• Review SQL in /drizzle folder
• Test on staging first
• Apply via migration runner

2. "Missing Index Cascade" Anti-Pattern

Symptoms: Query performance degrades after adding foreign keys Detection: Foreign key columns without corresponding indexes Fix:

// Add index for FK columns
export const checkIns = sqliteTable('check_ins', {
  userId: text('user_id').notNull().references(() => users.id),
}, (table) => ({
  userIdIdx: index('idx_checkins_user_id').on(table.userId), // Add this
}));

3. "Destructive Migration Drift" Anti-Pattern

Symptoms: Data loss during column rename or type changes Detection: Migration contains DROP COLUMN or incompatible type changes Fix:

• Create multi-step migration
• Step 1: Add new column
• Step 2: Migrate data
• Step 3: Remove old column

4. "Schema Bloat" Anti-Pattern

Symptoms: Tables with 15+ columns, deeply nested JSON in text fields Detection: Single table handling multiple concerns Fix:

• Break into related tables with proper foreign keys
• Extract JSON objects into separate tables
• Use relations to maintain referential integrity

5. "Missing WHERE Clause Disaster" Anti-Pattern

Symptoms: Accidental bulk deletes or updates Detection: DELETE or UPDATE statements without .where() clause Fix:

// WRONG: Deletes all records
await db.delete(users);

// RIGHT: Targeted deletion
await db.delete(users).where(eq(users.id, userId));

WORKED EXAMPLES

Adding a New Table with Relations

Scenario: Add user preferences table to existing user system

Step 1: Analyze Impact

• New table, no breaking changes
• Will need FK relationship to users
• Should include index on user_id for lookups

Step 2: Schema Definition

// src/db/schema.ts
export const userPreferences = sqliteTable('user_preferences', {
  id: text('id').primaryKey(),
  userId: text('user_id').notNull().references(() => users.id, {
    onDelete: 'cascade',  // Delete prefs when user deleted
  }),
  theme: text('theme', { enum: ['light', 'dark', 'auto'] }).default('auto'),
  notifications: integer('notifications', { mode: 'boolean' }).default(true),
  timezone: text('timezone').default('UTC'),
  updatedAt: text('updated_at').notNull().default(sql`CURRENT_TIMESTAMP`),
}, (table) => ({
  userIdIdx: index('idx_user_preferences_user_id').on(table.userId),
}));

export const userPreferencesRelations = relations(userPreferences, ({ one }) => ({
  user: one(users, {
    fields: [userPreferences.userId],
    references: [users.id],
  }),
}));

export const usersRelations = relations(users, ({ one, many }) => ({
  preferences: one(userPreferences),  // Add this line
  checkIns: many(checkIns),
}));

Step 3: Generate and Review

npm run db:generate

Step 4: Verify Generated SQL

-- drizzle/0001_create_user_preferences.sql
CREATE TABLE `user_preferences` (
  `id` text PRIMARY KEY NOT NULL,
  `user_id` text NOT NULL,
  `theme` text DEFAULT 'auto',
  `notifications` integer DEFAULT true,
  `timezone` text DEFAULT 'UTC',
  `updated_at` text DEFAULT CURRENT_TIMESTAMP NOT NULL,
  FOREIGN KEY (`user_id`) REFERENCES `users`(`id`) ON UPDATE no action ON DELETE cascade
);

CREATE INDEX `idx_user_preferences_user_id` ON `user_preferences` (`user_id`);

What Expert Catches: FK cascade behavior is correct, index is present, defaults are appropriate What Novice Misses: Would forget the index, might not consider cascade behavior

QUALITY GATES

• Migration generates without errors (npm run db:generate)
• Generated SQL reviewed for correctness in /drizzle folder
• Foreign key constraints include proper onDelete behavior
• Indexes added for all foreign key columns
• Test database successfully migrates (npm run db:migrate)
• Sample queries execute without performance issues
• No data loss verified if modifying existing tables
• Migration is reversible or rollback plan documented
• Schema matches TypeScript types (no type/runtime mismatches)
• Relations properly defined for new tables

NOT-FOR Boundaries

❌ DO NOT use this skill for:

• PostgreSQL/Supabase projects → Use supabase-admin skill instead
• Prisma ORM migrations → Different syntax and CLI commands
• Raw SQL database management → Use database-specific tools
• Database design theory → Use architecture planning skills
• Performance tuning → Use database optimization skills
• Backup/restore operations → Use database administration skills

✅ Instead delegate to:

• supabase-admin for PostgreSQL schema management
• database-design for entity relationship modeling
• sql-optimization for query performance issues