Stop 07 · NoSQL DatabasesTopics · 6/12

Indexing in MongoDB

In short

Without an index, MongoDB checks every document to answer a query. Fine with a thousand documents, painful with ten million. Indexes fix that.

Like SQL indexes, MongoDB indexes speed up queries on specific fields — critical once a collection grows beyond a few thousand documents.

The blob in a white cap sorts tiffin boxes by coloured dots on their lids.

Think of it like this

Dabbawala colour codes

Dabbawalas don't open every tiffin to find where it goes. Coloured marks on the lid show the station and building, so sorting is quick. A MongoDB index is that mark on the lid: a sorted guide on certain fields that points straight at the right documents.

Watch it happen

Using { city: 1, year: 1 }

Delhi, 1

Delhi, 3

Pune, 1

Pune, 2

Pune, 4

Surat, 3

1/5This is a compound index on city, then year. Entries are sorted by city first, then by year.

The big ideas

What you need to know

Single-field indexes

createIndex({ email: 1 }) speeds up queries on email. 1 means ascending, -1 descending.

Compound indexes

One index on several fields. Order matters: { city: 1, year: 1 } helps city queries, not year-only ones.

Unique and TTL indexes

unique: true blocks duplicates. A TTL index deletes documents automatically after a set time.

Check with explain()

COLLSCAN means every document was read. IXSCAN means your index did the work.

See it in code

Walk through the code

1// Fast lookups by email, and no duplicate accounts2db.users.createIndex({ email: 1 }, { unique: true });3 4// Compound: filter by city, then year5db.students.createIndex({ city: 1, year: 1 });6 7// Delete sessions about an hour after createdAt8db.sessions.createIndex({ createdAt: 1 }, { expireAfterSeconds: 3600 });9 10// Did the query use an index?11db.students.find({ city: "Pune", year: 2 }).explain("executionStats");

1/4A unique index on email: fast logins, and duplicate signups get rejected.

Every index uses memory and slows writes a little. Add indexes for the queries your app actually runs.

Save yourself the pain

Common mistakes

Oops

Creating { year: 1, city: 1 } and then querying by city alone.

Do this instead

A compound index helps from its leftmost fields. Put the field you always filter on first.

Oops

Indexing every field just in case.

Do this instead

Each index is updated on every write and uses memory. Find slow queries with explain() and index those.

Your mission

COLLSCAN to IXSCAN

~25 min

Quick check

You have an index { city: 1, year: 1 }. Which query can use it best?

Go deeper

1Create an index on a frequently queried field and compare query performance with `explain()`.
2Learn compound indexes for queries that filter on multiple fields together.
3Understand the write-performance cost of adding too many indexes.