Spring Data JPA aggregate functions on an empty resultset

Question

I am working on a project involving Spring and JPA/Hibernate. The database driver used in my development environment is H2. My application has a page that displays statistics, one example of such a statistic is the average age of my users. However, when I try to fetch the average age using JPQL, I receive an exception

Result must not be null!

Assume for simplicity reasons that I store age as an integer on every User object (in my application this is of course not the case, but that's not important for my problem).

User model

@Entity
public class User implements Identifiable<Long> {
    private int age;
    // more fields and methods, irrelevant
}

User repository

@Repository
public interface UserRepository extends CrudRepository<User, Long> {
    @Query("SELECT AVG(u.age) FROM #{#entityName} u")
    long averageAge();
}

I cannot seem to figure out why calling UserRepository#averageAge(); is throwing the exception. I have tried replacing the function AVG in the query by COUNT and this behaves as expected. I have also tried to use an SQL query and setting nativeQuery = true in the annotation, yet to no avail. I can ofcourse solve it by fetching all the users and calculate the average age in plain Java, but this wouldn't be very efficient.

Stacktrace:

Caused by: org.springframework.dao.EmptyResultDataAccessException: Result must not be null!
    at org.springframework.data.repository.core.support.MethodInvocationValidator.invoke(MethodInvocationValidator.java:102)
    at org.springframework.aop.framework.ReflectiveMethodInvocation.proceed(ReflectiveMethodInvocation.java:185)
    at org.springframework.aop.framework.JdkDynamicAopProxy.invoke(JdkDynamicAopProxy.java:212)
    at com.sun.proxy.$Proxy150.averageAge(Unknown Source)
    at my.test.application.StatisticsRunner.run(StatisticsRunner.java:72)
at org.springframework.boot.SpringApplication.callRunner(SpringApplication.java:809)
    ... 30 more

Solved

The exception was caused by the fact that AVG() returns null when performed on an empty table. I fixed it by modifying the query (inspired by the answer to this question) as follows:

@Query("SELECT coalesce(AVG(u.age), 0) FROM #{#entityName} u")
long averageAge();

Answer 1

If you use Spring Data, and if your method returns null when Hibernate can't find a match, make sure you add @org.springframework.lang.Nullable to your method signature:

public interface SomeRepositoryCustom {
    @org.springframework.lang.Nullable
    public Thing findOneThingByAttr(Attribute attr) {
        /* ...your logic here... */
    }
}

This is because Spring Data checks the nullability of your method, and if the annotation is missing, it's going to enforce that you always need to return an object:

/* org.springframework.data.repository.core.support.MethodInvocationValidator */

@Nullable
@Override
public Object invoke(@SuppressWarnings("null") MethodInvocation invocation) throws Throwable {
    /* ...snip... */

    if (result == null && !nullability.isNullableReturn()) {
        throw new EmptyResultDataAccessException("Result must not be null!", 1);
    }

    /* ...snip... */

I used Spring Boot version 2.1.1.RELEASE and Spring Data 2.1.4.RELEASE.

Answer 2

It seems that the EmptyResultDataAccessException exception is thrown when a result from a query was expected to have at least one row (or element) but none was returned.

Related documentation about this can be found here.

I would suggest to run the same query this attempts to run in order to further validate this theory. Now the good question's what to do with this.

You have two options. Either catch the EmptyResultDataAccessException exception at your calling point and handle it directly in there or alternatively you can have an ExceptionHandler which will be tasked with handling such exceptions.

Both ways of handling this, should be OK and you may choose between each depending on your scenario.