MongoDB Fundamentals

What we're doing: Show one document holding data that a relational schema would split across two tables.

{
  "_id": "65f1...",
  "title": "Dune",
  "author": { "name": "Frank Herbert", "country": "USA" },
  "tags": ["sci-fi", "classic"]
}

1

The whole value is one document — one self-contained record.

4

author is an embedded document: a relational schema would put this in a separate authors table and join on a foreign key.

Why this works: A relational design normalizes author into its own table so the name is not repeated per book, then joins the two at query time. A document design instead asks whether a book is ever read without its author — if the two are always fetched together, nesting author removes the join and the query becomes a single document read.

What we're doing: Show the four nesting levels and a field whose value is itself an embedded document.

// db "shop" > collection "books" > this document
{
  "_id": ObjectId("65f1a2b3c4d5e6f7a8b9c0d1"),
  "title": "Dune",
  "author": { "name": "Frank Herbert" },
  "tags": ["sci-fi", "classic"]
}

2

_id is generated automatically as an ObjectId — nothing supplied one here.

4

author is an embedded document, not a reference — its value is itself field-value pairs.

Why this works: The nesting from server down to document is what lets a single write or read touch everything about one thing — author does not need its own collection unless it is looked up or updated independently of the book.

What we're doing: Show why "update" and "replace" are two different verbs, not one.

// document before: { "_id": 1, "status": "pending", "total": 42 }

updateOne({ _id: 1 }, { $set: { status: "shipped" } })
// -> { "_id": 1, "status": "shipped", "total": 42 }   total survives

replaceOne({ _id: 1 }, { status: "shipped" })
// -> { "_id": 1, "status": "shipped" }                total is gone

3

update touches only the field named in $set — every other field on the document is left exactly as it was.

6

replace treats the new document as the WHOLE new document — anything not included in it is dropped, except _id.

Why this works: update and replace both "change" a document, but they answer different questions: "adjust these fields" versus "this is now the entire document." Reaching for replace when only one field should change silently deletes every other field the caller did not think to repeat.

What we're doing: Show that an insert missing a field silently succeeds unless a validator is in place.

// No validator on "orders"
db.orders.insertOne({ total: 42 })   // succeeds — "status" was never required

db.createCollection("orders_v2", {
  validator: { $jsonSchema: { required: ["status"] } }
})
db.orders_v2.insertOne({ total: 42 })   // rejected — "status" is missing

2

With no validator, MongoDB accepts any shape — the missing "status" field is not an error.

7

The validator turns "status" from an unenforced convention into a database-level requirement.

Why this works: Nothing about the document model itself forces a field to exist — that has to be chosen deliberately, either in application code or with a validator. Skipping both does not mean the data has no real shape, only that nothing is checking it.