orders(order_id, user_id, ordered_at)

Return the user_id of everyone who ordered on three or more consecutive calendar days during 2026, ordered by user_id.

Several orders on one day are still one day, so duplicates have to go before the consecutive check rather than after.

The usual approach is gaps and islands: subtract a row number from the date, and consecutive days collapse to the same value.

Tables

orders

order_id  user_id  ordered_at
--------  -------  -------------------
1         10       2026-03-01 09:00:00
2         10       2026-03-01 18:00:00
3         10       2026-03-02 09:00:00
4         10       2026-03-03 09:00:00
5         11       2026-03-01 09:00:00
6         11       2026-03-02 09:00:00
7         11       2026-03-04 09:00:00
8         12       2026-03-10 09:00:00
... 3 more row(s)

Expected result

user_id
-------
10
12