Claude Code Plugins

Community-maintained marketplace

Feedback

Database Implementation

@jpicklyk/task-orchestrator
110
0

Database schema design, migrations, query optimization with SQL, Exposed ORM, Flyway. Use for database, migration, schema, sql, flyway tags. Provides migration patterns, validation commands, rollback strategies.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name Database Implementation
description Database schema design, migrations, query optimization with SQL, Exposed ORM, Flyway. Use for database, migration, schema, sql, flyway tags. Provides migration patterns, validation commands, rollback strategies.
allowed-tools Read, Write, Edit, Bash, Grep, Glob

Database Implementation Skill

Domain-specific guidance for database schema design, migrations, and data modeling.

When To Use This Skill

Load this Skill when task has tags:

  • database, migration, schema, sql, flyway
  • exposed, orm, query, index, constraint

Validation Commands

Run Migrations

# Gradle + Flyway
./gradlew flywayMigrate

# Test migration on clean database
./gradlew flywayClean flywayMigrate

# Check migration status
./gradlew flywayInfo

# Validate migrations
./gradlew flywayValidate

Run Tests

# Migration tests
./gradlew test --tests "*migration*"

# Database integration tests
./gradlew test --tests "*Repository*"

# All tests
./gradlew test

Success Criteria (Before Completing Task)

Migration runs without errors on clean database ✅ Schema matches design specificationsIndexes created correctlyConstraints validate as expectedRollback works (if applicable) ✅ Tests pass with new schema

Common Database Tasks

Creating Migrations

  • Add tables with columns, constraints
  • Create indexes for performance
  • Add foreign keys for referential integrity
  • Modify existing schema (ALTER TABLE)
  • Seed data (reference data)

ORM Models

  • Map entities to tables (Exposed, JPA)
  • Define relationships (one-to-many, many-to-many)
  • Configure cascading behavior
  • Define custom queries

Query Optimization

  • Add indexes for frequently queried columns
  • Analyze query plans (EXPLAIN)
  • Optimize N+1 query problems
  • Use appropriate JOIN types

Migration Patterns

Create Table

-- V001__create_users_table.sql
CREATE TABLE users (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    email VARCHAR(255) NOT NULL UNIQUE,
    password_hash VARCHAR(255) NOT NULL,
    name VARCHAR(255) NOT NULL,
    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
);

-- Indexes
CREATE INDEX idx_users_email ON users(email);
CREATE INDEX idx_users_created_at ON users(created_at);

Add Column

-- V002__add_users_phone.sql
ALTER TABLE users
ADD COLUMN phone VARCHAR(20);

-- Add with default value
ALTER TABLE users
ADD COLUMN is_active BOOLEAN NOT NULL DEFAULT true;

Create Foreign Key

-- V003__create_tasks_table.sql
CREATE TABLE tasks (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    title VARCHAR(500) NOT NULL,
    user_id UUID NOT NULL,
    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,

    -- Foreign key with cascade
    CONSTRAINT fk_tasks_user
        FOREIGN KEY (user_id)
        REFERENCES users(id)
        ON DELETE CASCADE
);

CREATE INDEX idx_tasks_user_id ON tasks(user_id);

Create Junction Table (Many-to-Many)

-- V004__create_user_roles.sql
CREATE TABLE user_roles (
    user_id UUID NOT NULL,
    role_id UUID NOT NULL,
    assigned_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,

    PRIMARY KEY (user_id, role_id),

    CONSTRAINT fk_user_roles_user
        FOREIGN KEY (user_id)
        REFERENCES users(id)
        ON DELETE CASCADE,

    CONSTRAINT fk_user_roles_role
        FOREIGN KEY (role_id)
        REFERENCES roles(id)
        ON DELETE CASCADE
);

CREATE INDEX idx_user_roles_user_id ON user_roles(user_id);
CREATE INDEX idx_user_roles_role_id ON user_roles(role_id);

Testing Migrations

Test Migration Execution

@Test
fun `migration V004 creates user_roles table`() {
    // Arrange - Clean database
    flyway.clean()

    // Act - Run migrations
    flyway.migrate()

    // Assert - Check table exists
    val tableExists = database.useConnection { connection ->
        val meta = connection.metaData
        val rs = meta.getTables(null, null, "user_roles", null)
        rs.next()
    }

    assertTrue(tableExists, "user_roles table should exist after migration")
}

Test Constraints

@Test
fun `user_roles enforces foreign key constraint`() {
    // Arrange
    val invalidUserId = UUID.randomUUID()
    val role = createTestRole()

    // Act & Assert
    assertThrows<SQLException> {
        database.transaction {
            UserRoles.insert {
                it[userId] = invalidUserId  // Invalid - user doesn't exist
                it[roleId] = role.id
            }
        }
    }
}

Common Blocker Scenarios

Blocker 1: Migration Fails on Existing Data

Issue: Adding NOT NULL column to table with existing rows

ERROR: column "status" contains null values

What to try:

  • Add column as nullable first
  • Update existing rows with default value
  • Then alter column to NOT NULL

Example fix:

-- Step 1: Add nullable
ALTER TABLE tasks ADD COLUMN status VARCHAR(20);

-- Step 2: Update existing rows
UPDATE tasks SET status = 'pending' WHERE status IS NULL;

-- Step 3: Make NOT NULL
ALTER TABLE tasks ALTER COLUMN status SET NOT NULL;

Blocker 2: Circular Foreign Key Dependencies

Issue: Table A references B, B references A - which to create first?

What to try:

  • Create both tables without foreign keys first
  • Add foreign keys in separate migration after both exist

Example:

-- V001: Create tables without FKs
CREATE TABLE users (...);
CREATE TABLE profiles (...);

-- V002: Add foreign keys
ALTER TABLE users ADD CONSTRAINT fk_users_profile ...;
ALTER TABLE profiles ADD CONSTRAINT fk_profiles_user ...;

Blocker 3: Index Creation Takes Too Long

Issue: Creating index on large table times out

What to try:

  • Use CREATE INDEX CONCURRENTLY (PostgreSQL)
  • Create index during low-traffic period
  • Check if similar index already exists

Blocker 4: Data Type Mismatch

Issue: ORM expects UUID but database has VARCHAR

What to try:

  • Check migration SQL - correct type used?
  • Check ORM mapping - correct type specified?
  • Migrate data type if needed:
    ALTER TABLE tasks ALTER COLUMN id TYPE UUID USING id::uuid;
    

Blocker 5: Missing Prerequisite Table

Issue: Foreign key references table that doesn't exist yet

What to try:

  • Check migration order - migrations run in version order (V001, V002, etc.)
  • Ensure referenced table created in earlier migration
  • Check for typos in table names

If blocked: Report to orchestrator - migration order issue or missing prerequisite

Blocker Report Format

⚠️ BLOCKED - Requires Senior Engineer

Issue: [Specific problem - migration fails, constraint violation, etc.]

Attempted Fixes:
- [What you tried #1]
- [What you tried #2]
- [Why attempts didn't work]

Root Cause (if known): [Your analysis]

Partial Progress: [What work you DID complete]

Context for Senior Engineer:
- Migration SQL: [Paste migration]
- Error output: [Database error]
- Related migrations: [Dependencies]

Requires: [What needs to happen]

Exposed ORM Patterns

Table Definition

object Users : UUIDTable("users") {
    val email = varchar("email", 255).uniqueIndex()
    val passwordHash = varchar("password_hash", 255)
    val name = varchar("name", 255)
    val createdAt = timestamp("created_at").defaultExpression(CurrentTimestamp())
    val updatedAt = timestamp("updated_at").defaultExpression(CurrentTimestamp())
}

Foreign Key Relationship

object Tasks : UUIDTable("tasks") {
    val title = varchar("title", 500)
    val userId = reference("user_id", Users)
    val createdAt = timestamp("created_at").defaultExpression(CurrentTimestamp())
}

Query with Join

fun findTasksWithUser(userId: UUID): List<TaskWithUser> {
    return (Tasks innerJoin Users)
        .select { Tasks.userId eq userId }
        .map { row ->
            TaskWithUser(
                task = rowToTask(row),
                user = rowToUser(row)
            )
        }
}

Rollback Strategies

Reversible Migrations

Good (can rollback):

  • Adding nullable columns
  • Adding indexes
  • Creating new tables (if no data)

Difficult to rollback:

  • Dropping columns (data loss)
  • Changing data types (data transformation)
  • Deleting tables (data loss)

Include Rollback SQL

For complex migrations, document rollback steps:

-- Migration: V005__add_user_status.sql
ALTER TABLE users ADD COLUMN status VARCHAR(20) NOT NULL DEFAULT 'active';

-- Rollback (document in comments):
-- ALTER TABLE users DROP COLUMN status;

Performance Tips

Indexing Strategy

DO create indexes on:

  • Foreign key columns
  • Frequently queried columns (WHERE, JOIN)
  • Columns used in ORDER BY
  • Unique constraints

DON'T create indexes on:

  • Small tables (< 1000 rows)
  • Columns that change frequently
  • Low cardinality columns (gender, boolean)

Query Optimization

-- ❌ BAD - Missing index, full table scan
SELECT * FROM users WHERE email = 'user@example.com';

-- ✅ GOOD - Index on email column
CREATE INDEX idx_users_email ON users(email);

-- ❌ BAD - N+1 query problem
SELECT * FROM users;  -- 1 query
SELECT * FROM tasks WHERE user_id = ?;  -- N queries (one per user)

-- ✅ GOOD - Single query with JOIN
SELECT u.*, t.*
FROM users u
LEFT JOIN tasks t ON t.user_id = u.id;

Common Patterns to Follow

  1. Sequential migration versioning (V001, V002, V003...)
  2. Descriptive migration names (V004__add_user_status.sql)
  3. Idempotent migrations (can run multiple times safely)
  4. Test on clean database before committing
  5. Foreign keys with indexes for performance
  6. NOT NULL with defaults for required fields
  7. Timestamps for audit trail (created_at, updated_at)

What NOT to Do

❌ Don't modify existing migrations (create new one) ❌ Don't drop columns without data backup ❌ Don't forget indexes on foreign keys ❌ Don't use SELECT * in production queries ❌ Don't skip testing migrations on clean database ❌ Don't forget CASCADE behavior on foreign keys ❌ Don't create migrations that depend on data state

Focus Areas

When reading task sections, prioritize:

  • requirements - What schema changes needed
  • technical-approach - Migration strategy
  • data-model - Entity relationships
  • migration - Specific SQL requirements

Remember

  • Test on clean database - always validate migration from scratch
  • Indexes on foreign keys - critical for performance
  • Sequential versioning - V001, V002, V003...
  • Descriptive names - migration filename explains what it does
  • Report blockers promptly - constraint issues, circular dependencies
  • Document rollback - comment how to reverse if needed
  • Validation is mandatory - migration must succeed before completion

Additional Resources

For deeper patterns and examples, see:

  • PATTERNS.md - Complex schema patterns, performance optimization (load if needed)
  • BLOCKERS.md - Detailed database-specific blockers (load if stuck)
  • examples.md - Complete migration examples (load if uncertain)