Recap: the other way

In the previous post we built offset pagination ( ?page=2 ). The server runs LIMIT 10 OFFSET 10 . Simple, but it has two pains:

Deep pages are slow. OFFSET 1000000 makes the database walk and throw away a million rows.

New rows shift the pages. If someone inserts a row while you scroll, page 2 might repeat a row from page 1 or skip one.

Cursor pagination fixes both. Twitter, Instagram, GitHub, Slack — almost every infinite-scroll feed uses it.

Cute mental model

You are reading a long book. You do not say "give me page 47". You put a bookmark on the last word you read. Next time you open the book, you flip straight to the bookmark and keep going.

The bookmark = the cursor

"Keep going" = WHERE id > bookmark

You never count pages. You just move forward from where you stopped.

That is the whole idea.

Step 1: what exactly is a cursor?

A cursor is just a value that points to a specific row.…