MongoDB Indexes: Compound, Sparse, Partial, TTL

Q: MongoDB Indexes: Compound, Sparse, Partial, TTL

MongoDB indexes are **B-tree structures** that let the query planner skip collection scans. **Compound indexes** cover multiple fields and follow the **ESR rule** (Equality → Sort → Range) so the leading fields are the most selective. **Sparse indexes** include only documents where the field exists — great for optional fields and for sparse-unique constraints. **Partial indexes** are stricter: a `partialFilterExpression` picks exactly which documents to index (active users only, unshipped orders). **TTL indexes** auto-delete documents after `expireAfterSeconds`, perfect for sessions and log data. In Spring Data MongoDB you declare them with `@Indexed`, `@CompoundIndex`, or programmatically via `MongoTemplate.indexOps()`.

Intermediate·2 min read·

MongodbShopifyAtlassianEbayHubspot

Fast recall

Quick revision

Core answer

MongoDB indexes are B-tree structures that let the query planner skip collection scans. Compound indexes cover multiple fields and follow the ESR rule (Equality → Sort → Range) so the leading fields are the most selective. Sparse indexes include only documents where the field exists — great for optional fields and for sparse-unique constraints. Partial indexes are stricter: a partialFilterExpression picks exactly which documents to index (active users only, unshipped orders). TTL indexes auto-delete documents after expireAfterSeconds, perfect for sessions and log data. In Spring Data MongoDB you declare them with @Indexed, @CompoundIndex, or programmatically via MongoTemplate.indexOps().

Interview-ready guide

Interview answer

Explain the index model

B-tree structures that avoid collection scans
Without them, every query is O(N)

Cover the core types

Compound + ESR: Equality, Sort, Range order
Sparse: skip docs missing the field (also sparse-unique trick)
Partial: stricter filter expression — prefer this for new indexes
TTL: auto-expire after expireAfterSeconds, single-field date
Multikey: automatic on arrays, one array per compound index

Mention the costs

Every index slows writes
Unused indexes still pay the tax — audit with $indexStats
Use explain('executionStats') to confirm IXSCAN and the key-to-doc ratio

Wrap up with Spring Data

@Indexed, @CompoundIndex, or MongoTemplate.indexOps() programmatic
Use migration scripts for reviewable index changes in production

03 · Deep dive