Requesting support for "SELECT TOP n from Table"

[SethuSrinivasan]

In notebook, It looks like if I need to select top N rows, I can rely on "LIMIT" keyword. It would be nice if you can support "TOP" as well

The current approach to select 10 rows:

select * from table1 LIMIT 10

Requesting TOP support:

SELECT TOP 10 * from table1

It is quite common for SQL Server users to use TOP clause in their queries. Can you please add this support to your SQL language?

[Louis_Frolio]

@SethuSrinivasan , Looks like this one slipped through the cracks — apologies if you've long since moved on, but posting anyway in case it helps someone hitting the same wall.

In Databricks SQL, SELECT TOP n doesn't exist. You get the same result with LIMIT, and if you care about which rows you're getting, pair it with ORDER BY.

Three patterns worth knowing:

Just the first N rows in any order: SELECT * FROM table1 LIMIT 10;

Top N by a column — say, highest salary: SELECT * FROM employees ORDER BY salary DESC LIMIT 10;

Top N per group — say, top 3 orders per customer — you'll want a window function: SELECT * FROM (SELECT o.*, ROW_NUMBER() OVER (PARTITION BY customer_id ORDER BY order_total DESC) AS rn FROM orders o) WHERE rn <= 3;

That last one is more flexible than TOP ever was anyway, and it scales well once you start dealing with more complex ranking logic.

Rule of thumb: wherever you'd write SELECT TOP 10 * FROM table1 in SQL Server, swap it for LIMIT 10 in Databricks SQL. Tack on ORDER BY if the order actually matters to you.

Hope this helps, Louis.