Return Only What You Need (Projection and Pagination)

Why this matters

Returning full documents when you only need two fields is one of the most common accidental performance problems in MongoDB applications. A document with 40 fields sent over the wire 1,000 times per second is dramatically more expensive than projecting to 3 fields — and it's a one-line fix. Pagination matters because skip(n) performs a full scan of the first n documents every time; cursor-based pagination using a range filter on _id or a timestamp scales to billions of documents without degrading, which is why every high-volume API uses it.

Demo

Project specific fields, sort results, and implement both offset and cursor-based pagination.

Try it yourself

1. Insert 20 documents with a score field. Implement cursor-based pagination using score as the cursor field (instead of _id). Fetch 3 pages of 5 and verify no document appears twice and none are skipped.
2. Use explain('executionStats') on a query with skip(1000).limit(10). Note the docsExamined value. Then implement cursor pagination for the same result. Compare docsExamined between the two.
3. Project with { password: 0 } (exclusion projection) on a users collection. Verify the field is absent. Then try mixing inclusion and exclusion in one projection — observe the error MongoDB returns.

References

• Cursor-based pagination in MongoDB
• MongoDB projection docs