2.2.1 Key-Value Databases

Key-Value stores are the simplest NoSQL data models. Every item is stored as an attribute name (key) together with its value.

• Key: A unique identifier (e.g., "user:123").
• Value: Opaque to the database—it can be text, a blob, or an object.
• Best for: Session data, user profiles, shopping carts, and caching.
• Examples: Redis, Riak, Amazon DynamoDB.

2.2.2 Document Databases

Document databases are similar to Key-Value stores, but the value is "transparent" to the database. It is typically stored in a format like JSON or BSON.

• Transparency: Because the DB understands the document structure, you can query specific fields (e.g., "Find all users where city = 'Mumbai'").
• Nested Structures: Documents can contain arrays and sub-documents, allowing for complex data representation.
• Examples: MongoDB, Couchbase.

2.2.3 Deep Dive into BSON (Binary JSON)

While JSON is human-readable, it is inefficient for computers to parse. MongoDB uses BSON.

• Speed: BSON is designed specifically for fast traversal (it includes the length of fields so the DB can "skip" them without reading every byte).
• Data Types: Unlike JSON, BSON includes native support for Dates, Binary data (images/files), and floating-point numbers.

2.2.4 Indexing in Document Databases

To avoid "Collection Scans" (reading every document), we use Indexes.

1. Compound Indexes: Indexing multiple fields (e.g., { city: 1, age: -1 }). Useful for complex queries.
2. Geospatial Indexes: Allowing for "Find the nearest coffee shop" queries based on latitude and longitude.
3. TTL (Time-To-Live) Indexes: Automatically deleting documents after a certain period (perfect for logs or session data).

Comparison: Key-Value vs. Document

2.2.3 Handling Relationships in NoSQL

NoSQL doesn't have "Foreign Keys" in the traditional sense. Relationships are handled in two ways:

1. Embedding: Putting the related data inside the aggregate (e.g., putting 'Tags' inside a 'Post' document).
2. Linking: Storing an ID that refers to another document manually (Programmatic join).