MongoDB Schema Design

Master data modeling and schema patterns.

Quick Start

One-to-One: Embedded

// User with single address - embed if always accessed together
{
  _id: ObjectId('...'),
  name: 'John',
  email: '[email protected]',
  address: {
    street: '123 Main St',
    city: 'New York',
    zip: '10001'
  }
}

One-to-Many: Embed Array

// User with multiple tags - embed if limited size
{
  _id: ObjectId('...'),
  name: 'John',
  tags: ['mongodb', 'database', 'nosql'],
  posts: [
    { _id: 1, title: 'Post 1', content: '...' },
    { _id: 2, title: 'Post 2', content: '...' }
  ]
}

One-to-Many: Reference

// User with many orders - reference if potentially large
{
  _id: ObjectId('user1'),
  name: 'John',
  email: '[email protected]'
}

// Orders collection
{
  _id: ObjectId('order1'),
  customerId: ObjectId('user1'),
  total: 99.99
}

Many-to-Many: Array of References

// Products with categories
{
  _id: ObjectId('product1'),
  name: 'Laptop',
  categoryIds: [
    ObjectId('electronics'),
    ObjectId('computers')
  ]
}

// Categories collection
{
  _id: ObjectId('electronics'),
  name: 'Electronics'
}

Schema Patterns

Attribute Pattern

// Store variant attributes flexibly
{
  _id: ObjectId('...'),
  productName: 'T-Shirt',
  attributes: [
    { key: 'color', value: 'blue' },
    { key: 'size', value: 'L' },
    { key: 'material', value: 'cotton' }
  ]
}

Polymorphic Pattern

// Different document types in same collection
{
  _id: ObjectId('...'),
  type: 'email',
  to: '[email protected]',
  subject: 'Hello'
}

{
  _id: ObjectId('...'),
  type: 'sms',
  phoneNumber: '+1234567890',
  message: 'Hi there'
}

Tree Structures: Adjacency List

// Parent-child relationships
{
  _id: ObjectId('...'),
  name: 'Electronics',
  parent: null
}

{
  _id: ObjectId('...'),
  name: 'Computers',
  parent: ObjectId('electronics')
}

Versioned Pattern

// Track document history
{
  _id: ObjectId('...'),
  name: 'Product',
  description: 'Latest description',
  versions: [
    { v: 1, name: 'Product', description: 'Original', date: ISODate(...) },
    { v: 2, name: 'Product', description: 'Updated', date: ISODate(...) }
  ]
}

Design Principles

Embedding Advantages

• Single query to fetch related data
• Atomic updates for related documents
• No joins needed

Referencing Advantages

• Avoid data duplication
• Smaller documents
• Flexible relationships
• Can grow independently

Decision Tree

Does the related data grow unbounded?
  YES → Use referencing
  NO → Consider embedding

Is the related data frequently accessed separately?
  YES → Use referencing
  NO → Consider embedding

Do updates need to be atomic across documents?
  YES → Use embedding
  NO → Use referencing

Python Design Example

# User with embedded address
users.insert_one({
    'name': 'John',
    'email': '[email protected]',
    'address': {
        'street': '123 Main St',
        'city': 'New York'
    }
})

# User with references to orders
users.insert_one({
    '_id': ObjectId('...'),
    'name': 'John'
})

orders.insert_one({
    'userId': ObjectId('...'),
    'total': 99.99
})

# Query with $lookup
users.aggregate([
    { '$lookup': {
        'from': 'orders',
        'localField': '_id',
        'foreignField': 'userId',
        'as': 'orders'
    }}
])

Best Practices

✅ Embed when data is always accessed together ✅ Reference for unbounded arrays ✅ Keep document size under 16MB ✅ Consider query patterns when designing ✅ Denormalize carefully for performance ✅ Plan for schema evolution ✅ Use validation schemas ✅ Document your design decisions