Document Stores

You're building a product catalog. Simple, right? Product name, price, description, throw it in a SQL table.

Then reality hits. A laptop has RAM, CPU, screen size, and GPU specs. A T-shirt has size, color, material, and fit type. A book has author, ISBN, page count, and genre. A piece of furniture has dimensions, weight, and assembly instructions.

Every product type has completely different attributes. In a relational database, you have ugly choices:

1. One massive table with hundreds of nullable columns (most are NULL for any given row)
2. Entity-Attribute-Value pattern (a horror show of JOINs and casting)
3. Separate tables for each product type (an explosion of tables and complex UNION queries)

None of these are good. They're all fighting the fundamental mismatch between rigid tabular schemas and data that's inherently varied and nested.

Document stores said: "What if the database just stored your data the way your application already thinks about it, as JSON objects with nested fields that vary per record?"

A document store holds collections of documents. Each document is a self-contained JSON-like object (technically BSON in MongoDB's case. Binary JSON with extra types).

Here's a product document:

{
 "_id": "prod_laptop_001",
 "name": "MacBook Pro 16-inch",
 "price": 2499,
 "category": "laptops",
 "specs": {
 "cpu": "M3 Max",
 "ram_gb": 36,
 "storage_gb": 1000,
 "screen_inches": 16.2,
 "gpu": "38-core GPU"
 },
 "reviews": [
 { "user": "alice", "rating": 5, "text": "Beast of a machine" },
 { "user": "bob", "rating": 4, "text": "Great but expensive" }
 ],
 "variants": [
 { "sku": "MBP16-36-1TB", "color": "Space Black", "in_stock": true },
 { "sku": "MBP16-36-2TB", "color": "Silver", "in_stock": false }
 ]
}

Notice what's happening here. The specs are nested. Reviews are an array of objects. Variants are another array. In a relational database, this would be four separate tables with foreign keys. In a document store, it's one document, self-contained, denormalized, and fetched in a single read.