What is an N+1 select?

Why can N+1 selects can lead to a performance problem?

Easily detect N+1 selects with QuickPerf

Easily fix N+1 selects with QuickPerf

For the following examples, the test executes the following script before the execution of the test method body:

    INSERT INTO TEAM VALUES (1, 'Manchester United');
    INSERT INTO TEAM VALUES (2, 'Atlético de Madrid');

    INSERT INTO PLAYER VALUES (1, 'Paul', 'Pogba', 1);
    INSERT INTO PLAYER VALUES (2, 'Antoine', 'Griezmann', 2);

Let's suppose that our project contains a Player JPA entity having a many to one association with a Team entity:

    @ManyToOne(targetEntity = Team.class)
    @JoinColumn(name = "team_id")
    private Team team;

The fetch type is not specified. In JPA, the default fetching policy of @ManyToOne is EAGER.

And let's suppose that our application is executing the following "FROM Player" Java Persistence query:

     TypedQuery<Player> fromPlayer = entityManager.createQuery("FROM Player", Player.class);
     List<Player> players = fromPlayer.getResultList();

The following SQL statements are then sent to the database:

    select
        player0_.id as id1_0_,
        player0_.firstName as firstNam2_0_,
        player0_.lastName as lastName3_0_,
        player0_.team_id as team_id4_0_ 
    from
        Player player0_

    select
        team0_.id as id1_1_0_,
        team0_.name as name2_1_0_ 
    from
        Team team0_ 
    where team0_.id=?
    
    Params:[(1)]

    select
        team0_.id as id1_1_0_,
        team0_.name as name2_1_0_ 
    from
        Team team0_ 
    where
        team0_.id=?"
    
    Params:[(2)]

Because of the default EAGER fetch type, the Java code sends a SELECT statement to the database to retrieve each player's team.

Now, let's suppose that our project contains a Player JPA entity having a many to one association with a Team entity:

    @ManyToOne(targetEntity = Team.class, fetch = FetchType.LAZY)
    @JoinColumn(name = "team_id")
    private Team team;

The fetch type is set to LAZY.

And let's suppose that our application is executing the following Java code:

    List<Player> players = fromPlayer.getResultList();

    List<PlayerWithTeamName> playersWithTeamName = players
                                                  .stream()
                                                  .map(player -> new PlayerWithTeamName( player.getFirstName()
                                                                                       , player.getLastName()
                                                                                       , player.getTeam().getName()
                                                                                       )
                                                       )
                                                  .collect(Collectors.toList());

Each time the getName() method is called, a SELECT... FROM Team statement is sent to the database:

    select
        player0_.id as id1_0_,
        player0_.firstName as firstNam2_0_,
        player0_.lastName as lastName3_0_,
        player0_.team_id as team_id4_0_ 
    from
        Player player0_

    select
        team0_.id as id1_1_0_,
        team0_.name as name2_1_0_ 
    from
        Team team0_ 
    where
        team0_.id=?

    Params:[(1)]

    select
        team0_.id as id1_1_0_,
        team0_.name as name2_1_0_ 
    from
        Team team0_ 
    where
        team0_.id=?

   Params:[(2)]

N+1 select antipattern can lead to many JDBC roundtrips in production, and JDBC roundtrips are harmful to performance, as explained in this article.

Hereafter, we show that, with the help of QuickPerf, we can quickly detect and fix the N+1 select antipattern from tests.

QuickPerf live can generate QuickPerf tests that reproduce and detect N+1 select.

We can detect N+1 select by adding @ExpectSelect annotation on a test method to check the number of executed SELECT statements.

    @ExpectSelect(1)
    @Test
    public void should_find_players() {

[PERF] You may think that <1> select statement was sent to the database
       But in fact <3>...

💣  You may have even more select statements with production data.

The outcome of an N+1 select may be to get the same SELECT statements with different values. In the previous examples, the N+1 select outcome was to have additional SELECT statements on Team table. These other SELECT statements are the same apart from the id value of the Team table.

We can systematically detect this pattern by configuring @DisableSameSelectTypesWithDifferentParamValues annotation with a global scope:

package org.quickperf;

import org.quickperf.config.SpecifiableGlobalAnnotations;

import java.lang.annotation.Annotation;
import java.util.Arrays;
import java.util.Collection;

import static org.quickperf.sql.annotation.SqlAnnotationBuilder.*;

public class QuickPerfConfiguration implements SpecifiableGlobalAnnotations {

    public Collection<Annotation> specifyAnnotationsAppliedOnEachTest() {

        return Arrays.asList(
                 disableSameSelectTypesWithDifferentParams()
        );

    }

}

⚠️ The class implementing SpecifiableGlobalAnnotations has to be in the org.quickperf package.

If a test executes several same select types with different parameter values, then it will fail.

Hibernate code examples are available to play with these two ways of detecting N+1 selects:
👉   Hibernate JUnit 4 code example
👉   Hibernate JUnit 5 code example
👉   Hibernate TestNG code example

The use of QuickPerf to detect N+1 selects in a Spring Boot, a Quarkus or a Micronaut application is demonstrated in this repository.

QuickPerf automatically detects that the tested code uses Hibernate or Spring Data JPA. When a test is failing possibly due to an N+1 select, QuickPerf proposes ways to fix an N+1 select with Hibernate or Spring Data JPA:

Perhaps you are facing an N+1 select issue
	* With Hibernate, you may fix it by using JOIN FETCH
	                                       or LEFT JOIN FETCH
	                                       or FetchType.LAZY
	                                       or ...

	* With Spring Data JPA, you may fix it by adding
		@EntityGraph(attributePaths = { "..." }) on repository method.

As suggested by QuickPerf, the first N+1 select example can be fixed can be fixed by specifying a LAZY fetch type:

    @ManyToOne(targetEntity = Team.class, fetch = FetchType.LAZY)
    @JoinColumn(name = "team_id")
    private Team team;

With the second N+1 select example, players and their team can be loaded with one SELECT statement by adding a JOIN FETCH or a LEFT JOIN FETCH in the JPA query:

    List<Player> players = fromPlayer.getResultList();
    TypedQuery<Player> fromPlayer = entityManager.createQuery("FROM Player p LEFT JOIN FETCH p.team"
                                                            , Player.class);

With a LEFT JOIN FETCH, the following SQL query is going to be executed:

    select
        player0_.id as id1_0_0_,
        team1_.id as id1_1_1_,
        player0_.firstName as firstNam2_0_0_,
        player0_.lastName as lastName3_0_0_,
        player0_.team_id as team_id4_0_0_,
        team1_.name as name2_1_1_ 
    from
        Player player0_ 
    left outer join
        Team team1_ 
            on player0_.team_id=team1_.id