= Indexes
:description: Keep filters and sorts fast as a collection grows: create, inspect and delete MongoDB indexes from the console, with the unique, TTL and full-text patterns.
:keywords: MongoDB index, unique index, TTL index, text index, compound index, query performance
:group: Your data
:order: 20

Without an index, a filter scans every document in the collection. That is fine for a thousand documents and slow for a million. An index keeps the values of a field sorted, so a filter or a sort on that field jumps straight to the documents it needs. Indexes also give you two rules for free: a **unique** index refuses duplicates, and a **TTL** index deletes documents after a time.

Every collection has an index on `_id` already. The rest you add under **Indexes**.

image::/assets/docs-images/indexes.png[The Indexes page: each collection with the number of indexes it has, ready to expand]

== Create an index

. Open **Indexes** and expand the collection.
. Click **New Index** and give it a name, such as `email-unique` or `category-price`.
. Type the **Keys**: which fields, and `1` for ascending, `-1` for descending, or `"text"` for full-text search.
. Add **Options** if you need them, and **Save**.

Both fields are JSON, checked as you type. Building an index on a collection that already has data takes a moment.

[cols="1,3"]
|===
| Option | What it does

| `unique`
| Refuses a document whose value already exists in another. Fails with `406` if duplicates exist when you create it: clean them up first.

| `sparse`
| Indexes only documents that have the field. With `unique`, documents without the field do not count as duplicates.

| `expireAfterSeconds`
| Makes it a TTL index: a document is deleted that many seconds after the date in the indexed field. `0` means at that date. The check runs about once a minute.

| `name`
| The index's name. The console sets it from the field you type.
|===

== The patterns that cover most apps

[cols="2,3"]
|===
| Keys and options | Use it for

| `{"status": 1}`
| A filter on one field, such as `?filter={"status": "pending"}`.

| `{"category": 1, "price": -1}`
| A filter on the first field and a sort on the second, such as `?filter={"category": "shoes"}&sort={"price": -1}`. The index serves any query on a prefix of its fields.

| `{"email": 1}` with `{"unique": true}`
| No two documents with the same email.

| `{"taxId": 1}` with `{"unique": true, "sparse": true}`
| Unique when present, optional otherwise.

| `{"expiresAt": 1}` with `{"expireAfterSeconds": 0}`
| Sessions, tokens, logs: gone when `expiresAt` passes. The field must be a date.

| `{"title": "text", "body": "text"}`
| Keyword search with `?filter={"$text": {"$search": "running shoes"}}`. One text index per collection; put every field you want searched in it.
|===

== Change or delete an index

An index cannot be edited. To change one, delete it and create it again with the new definition. The **Delete** action asks for confirmation and is immediate; until you create a replacement, queries that used it scan the collection. The `_id` index cannot be deleted.

== From a script

[cols="2,3"]
|===
| Operation | API

| List | `GET /<collection>/_indexes`
| Create | `PUT /<collection>/_indexes/<name>` with `{ "keys": {...}, "ops": {...} }`
| Delete | `DELETE /<collection>/_indexes/<name>`
|===

A `406 Not Acceptable` on create means the definition was refused: duplicates under a unique index, or a name already used by a different index. The response says which. With xref:cli.adoc[`rhc`], indexes are part of the setup file and are created with the rest of the service.

== Related pages

* xref:managing-data.adoc[Collections and documents]: the filters and sorts the indexes speed up.
* xref:schemas.adoc[Schemas]: rules on one document. A unique index is a rule across documents.
* xref:constraints.adoc[Constraints]: rules across the whole collection that an index cannot express.