การข้ามและการจำกัด: การแบ่งหน้าแบบออฟเซ็ต

Offset pagination—also called page-number pagination—divides results into fixed-size pages and uses a page number to determine how far into the result set to start. Page 1 shows items 1-20, page 2 shows items 21-40, and so on. In MongoDB this is implemented with skip() to jump over earlier results and limit() to restrict how many documents are returned per page.

limit(N) tells the cursor to return at most N documents. skip(N) tells MongoDB to skip the first N documents before returning any results. Together they implement page-number pagination: to get page P with size S items per page, use skip((P-1)*S) and limit(S).

Offset pagination usually requires a total document count to display page numbers in the UI. Use countDocuments(filter) to count matching documents before applying pagination. Run the count and the paged query in parallel to avoid adding latency. The total count divided by the page size (rounded up) gives the total number of pages.

MongoDB implements skip() by scanning and discarding the first N documents. Even with an index, MongoDB must walk through and count N index entries before returning results. On page 1, skip is 0—fast. On page 500 with 20 items per page, skip is 9980—MongoDB must traverse nearly 10,000 entries just to find where page 500 starts. This is the core performance problem with offset pagination.

The time to execute a skip/limit query grows linearly with the skip amount. The query cost is O(skip + limit)—the server must examine skip documents before returning limit documents. For page 1 this is O(20); for page 1000 with 20 items per page it is O(20020). As users navigate to higher page numbers, queries get progressively slower, often going from milliseconds to seconds on large collections.

Despite its performance limitations, offset pagination is acceptable in these scenarios: (1) the collection has fewer than a few thousand documents; (2) users rarely navigate beyond the first few pages; (3) the feature requires jumping directly to a page number (e.g., 'go to page 47'). Many admin dashboards and search results with low page depth fit this profile. Use keyset pagination for infinite scroll or large data sets.

A typical REST list endpoint accepts page and limit query parameters, validates them, and applies skip/limit accordingly. Always cap the maximum limit to prevent clients from requesting thousands of documents in a single call, which would exhaust server memory.

Offset pagination has a subtle correctness problem: if documents are inserted or deleted between page requests, items can shift positions in the sorted result set. A document inserted between page 1 and page 2 pushes every subsequent document forward, causing one item to appear on both pages (duplicate) or be skipped entirely. This is called the page drift problem and is inherent to offset pagination.

For very large collections, countDocuments(filter) can be slow because it scans the index. An alternative is estimatedDocumentCount(), which is O(1) but counts all documents in the collection without applying a filter. For simple cases where you want the total without filtering, the estimated count is much faster. For filtered counts on large collections, consider caching the count or using Atlas's faceted search.

Always combine pagination with a tight projection for list endpoints. Fetching all fields while paginating defeats the purpose: you're still paying to transfer full document payloads for every page. A projection that returns only summary fields (name, price, thumbnail) reduces bandwidth by 80-95% compared to fetching full documents, making pagination viable at larger page offsets.

Switch from offset to keyset (cursor) pagination when: users scroll infinitely through results (no page numbers needed), the collection has more than 100,000 documents, page load times increase noticeably for higher page numbers, or data changes frequently between page requests. Keyset pagination is always O(log n) regardless of position in the result set, because it uses a range query on an indexed field instead of skip.

Test your understanding of MongoDB & NoSQL Databases concepts from this lesson.

In this lesson you learned: offset pagination uses skip((page-1)*size) and limit(size) to fetch a page, skip() scans and discards documents so deep pages become progressively slower, and offset pagination is acceptable for small collections or shallow page depths but keyset pagination is better at scale. Next up we implement keyset pagination with range queries for consistent O(log n) performance.