Using 'IN' with a sub-query in SQL Statements

Question

Are there any performance issues of using \"IN\" keyword in SQL statements in places where we can use JOIN?

SELECT xxx
FROM xxx
WHERE ID IN (SELECT Id FROM x         


        

Answer 1

Rather than a distinct you could use group by. I have had cases where I got better response time using join. Typically when I am joining all the rows via a primary key / foreign key relationship and the where is looking at a non key column. Especially if multiple joins. The IN can SOMETIMES force an index scan and the join will TYPICALLY use a seek if it is going to the PK. When you design you tables line up the primary keys so they are in the same order and explicitly declare the PK / FK relationships. Join are NOT limited to PK / FK. But a common use of a join is to walk the PK / FK relationship and in that case my experience using a join with the keys aligned is the best performance.

Answer 2

No, it's OK to use.

You can write the query above using IN, EXISTS in all RDBMS, some also support INTERSECT.

Semantically this is a semi-join which "give me rows from table A where I have a at least one match in tableB". An INNER JOIN is "give me all matching rows"

So if TableA has 3 rows and TableB has 5 rows that match:

• an INNER JOIN is 15 rows
• a semi-join is 3 rows

This is why IN and EXISTS are pushed by me and the other SQL types here: a JOIN is wrong, requires DISTINCT and will be slower.

EXISTS support multiple column JOINs, IN doesn't in SQL Server (it does in others).

Answer 3

As you can read here, JOINS are faster than sub-selects.