I have a spring 4 app where I'm trying to delete an instance of an entity from my database. I have the following entity:
#Entity
public class Token implements Serializable {
#Id
#SequenceGenerator(name = "seqToken", sequenceName = "SEQ_TOKEN", initialValue = 500, allocationSize = 1)
#GeneratedValue(strategy = GenerationType.SEQUENCE, generator = "seqToken")
#Column(name = "TOKEN_ID", nullable = false, precision = 19, scale = 0)
private Long id;
#NotNull
#Column(name = "VALUE", unique = true)
private String value;
#ManyToOne(fetch = FetchType.EAGER)
#JoinColumn(name = "USER_ACCOUNT_ID", nullable = false)
private UserAccount userAccount;
#Temporal(TemporalType.TIMESTAMP)
#Column(name = "EXPIRES", length = 11)
private Date expires;
...
// getters and setters omitted to keep it simple
}
I have a JpaRepository interface defined:
public interface TokenRepository extends JpaRepository<Token, Long> {
Token findByValue(#Param("value") String value);
}
I have a unit test setup that works with an in memory database (H2) and I am pre-filling the database with two tokens:
#Test
public void testDeleteToken() {
assertThat(tokenRepository.findAll().size(), is(2));
Token deleted = tokenRepository.findOne(1L);
tokenRepository.delete(deleted);
tokenRepository.flush();
assertThat(tokenRepository.findAll().size(), is(1));
}
The first assertion passes, the second fails. I tried another test that changes the token value and saves that to the database and it does indeed work, so I'm not sure why delete isn't working. It doesn't throw any exceptions either, just doesn't persist it to the database. It doesn't work against my oracle database either.
Edit
Still having this issue. I was able to get the delete to persist to the database by adding this to my TokenRepository interface:
#Modifying
#Query("delete from Token t where t.id = ?1")
void delete(Long entityId);
However this is not an ideal solution. Any ideas as to what I need to do to get it working without this extra method?
Most probably such behaviour occurs when you have bidirectional relationship and you're not synchronizing both sides WHILE having both parent and child persisted (attached to the current session).
This is tricky and I'm gonna explain this with the following example.
#Entity
public class Parent {
#Id
#GeneratedValue(strategy = IDENTITY)
#Column(name = "id", unique = true, nullable = false)
private Long id;
#OneToMany(cascade = CascadeType.PERSIST, mappedBy = "parent")
private Set<Child> children = new HashSet<>(0);
public void setChildren(Set<Child> children) {
this.children = children;
this.children.forEach(child -> child.setParent(this));
}
}
#Entity
public class Child {
#Id
#GeneratedValue(strategy = IDENTITY)
#Column(name = "id", unique = true, nullable = false)
private Long id;
#ManyToOne
#JoinColumn(name = "parent_id")
private Parent parent;
public void setParent(Parent parent) {
this.parent = parent;
}
}
Let's write a test (a transactional one btw)
public class ParentTest extends IntegrationTestSpec {
#Autowired
private ParentRepository parentRepository;
#Autowired
private ChildRepository childRepository;
#Autowired
private ParentFixture parentFixture;
#Test
public void test() {
Parent parent = new Parent();
Child child = new Child();
parent.setChildren(Set.of(child));
parentRepository.save(parent);
Child fetchedChild = childRepository.findAll().get(0);
childRepository.delete(fetchedChild);
assertEquals(1, parentRepository.count());
assertEquals(0, childRepository.count()); // FAILS!!! childRepostitory.counts() returns 1
}
}
Pretty simple test right? We're creating parent and child, save it to database, then fetching a child from database, removing it and at last making sure everything works just as expected. And it's not.
The delete here didn't work because we didn't synchronized the other part of relationship which is PERSISTED IN CURRENT SESSION. If Parent wasn't associated with current session our test would pass, i.e.
#Component
public class ParentFixture {
...
#Transactional(propagation = Propagation.REQUIRES_NEW)
public void thereIsParentWithChildren() {
Parent parent = new Parent();
Child child = new Child();
parent.setChildren(Set.of(child));
parentRepository.save(parent);
}
}
and
#Test
public void test() {
parentFixture.thereIsParentWithChildren(); // we're saving Child and Parent in seperate transaction
Child fetchedChild = childRepository.findAll().get(0);
childRepository.delete(fetchedChild);
assertEquals(1, parentRepository.count());
assertEquals(0, childRepository.count()); // WORKS!
}
Of course it only proves my point and explains the behaviour OP faced. The proper way to go is obviously keeping in sync both parts of relationship which means:
class Parent {
...
public void dismissChild(Child child) {
this.children.remove(child);
}
public void dismissChildren() {
this.children.forEach(child -> child.dismissParent()); // SYNCHRONIZING THE OTHER SIDE OF RELATIONSHIP
this.children.clear();
}
}
class Child {
...
public void dismissParent() {
this.parent.dismissChild(this); //SYNCHRONIZING THE OTHER SIDE OF RELATIONSHIP
this.parent = null;
}
}
Obviously #PreRemove could be used here.
I had the same problem
Perhaps your UserAccount entity has an #OneToMany with Cascade on some attribute.
I've just remove the cascade, than it could persist when deleting...
You need to add PreRemove function ,in the class where you have many object as attribute e.g in Education Class which have relation with UserProfile
Education.java
private Set<UserProfile> userProfiles = new HashSet<UserProfile>(0);
#ManyToMany(fetch = FetchType.EAGER, mappedBy = "educations")
public Set<UserProfile> getUserProfiles() {
return this.userProfiles;
}
#PreRemove
private void removeEducationFromUsersProfile() {
for (UsersProfile u : usersProfiles) {
u.getEducationses().remove(this);
}
}
One way is to use cascade = CascadeType.ALL like this in your userAccount service:
#OneToMany(cascade = CascadeType.ALL)
private List<Token> tokens;
Then do something like the following (or similar logic)
#Transactional
public void deleteUserToken(Token token){
userAccount.getTokens().remove(token);
}
Notice the #Transactional annotation. This will allow Spring (Hibernate) to know if you want to either persist, merge, or whatever it is you are doing in the method. AFAIK the example above should work as if you had no CascadeType set, and call JPARepository.delete(token).
This is for anyone coming from Google on why their delete method is not working in Spring Boot/Hibernate, whether it's used from the JpaRepository/CrudRepository's delete or from a custom repository calling session.delete(entity) or entityManager.remove(entity).
I was upgrading from Spring Boot 1.5 to version 2.2.6 (and Hibernate 5.4.13) and had been using a custom configuration for transactionManager, something like this:
#Bean
public HibernateTransactionManager transactionManager(EntityManagerFactory entityManagerFactory) {
return new HibernateTransactionManager(entityManagerFactory.unwrap(SessionFactory.class));
}
And I managed to solve it by using #EnableTransactionManagement and deleting the custom
transactionManager bean definition above.
If you still have to use a custom transaction manager of sorts, changing the bean definition to the code below may also work:
#Bean
public PlatformTransactionManager transactionManager(EntityManagerFactory entityManagerFactory) {
return new JpaTransactionManager(entityManagerFactory);
}
As a final note, remember to enable Spring Boot's auto-configuration so the entityManagerFactory bean can be created automatically, and also remove any sessionFactory bean if you're upgrading to entityManager (otherwise Spring Boot won't do the auto-configuration properly). And lastly, ensure that your methods are #Transactional if you're not dealing with transactions manually.
I was facing the similar issue.
Solution 1:
The reason why the records are not being deleted could be that the entities are still attached. So we've to detach them first and then try to delete them.
Here is my code example:
User Entity:
#Entity
public class User {
#OneToMany(cascade = CascadeType.ALL, fetch = FetchType.LAZY, mappedBy = "user")
private List<Contact> contacts = new ArrayList<>();
}
Contact Entity:
#Entity
public class Contact {
#Id
private int cId;
#ManyToOne
private User user;
}
Delete Code:
user.getContacts().removeIf(c -> c.getcId() == contact.getcId());
this.userRepository.save(user);
this.contactRepository.delete(contact);
Here we are first removing the Contact object (which we want to delete) from the User's contacts ArrayList, and then we are using the delete() method.
Solution 2:
Here we are using the orphanRemoval attribute, which is used to delete orphaned entities from the database. An entity that is no longer attached to its parent is known as an orphaned entity.
Code example:
User Entity:
#Entity
public class User {
#OneToMany(cascade = CascadeType.ALL, fetch = FetchType.LAZY, mappedBy = "user", orphanRemoval = true)
private List<Contact> contacts = new ArrayList<>();
}
Contact Entity:
#Entity
public class Contact {
#Id
private int cId;
#ManyToOne
private User user;
}
Delete Code:
user.getContacts().removeIf(c -> c.getcId() == contact.getcId());
this.userRepository.save(user);
Here, as the Contact entity is no longer attached to its parent, it is an orphaned entity and will be deleted from the database.
I just went through this too. In my case, I had to make the child table have a nullable foreign key field and then remove the parent from the relationship by setting null, then calling save and delete and flush.
I didn't see a delete in the log or any exception prior to doing this.
If you use an newer version of Spring Data, you could use deleteBy syntax...so you are able to remove one of your annotations :P
the next thing is, that the behaviour is already tract by a Jira ticket:
https://jira.spring.io/browse/DATAJPA-727
#Transactional
int deleteAuthorByName(String name);
you should write #Transactional in Repository extends JpaRepository
Your initial value for id is 500. That means your id starts with 500
#SequenceGenerator(name = "seqToken", sequenceName = "SEQ_TOKEN",
initialValue = 500, allocationSize = 1)
And you select one item with id 1 here
Token deleted = tokenRepository.findOne(1L);
So check your database to clarify that
I've the same problem, test is ok but on db row isn't deleted.
have you added the #Transactional annotation to method? for me this change makes it work
In my case was the CASCADE.PERSIST, i changed for CASCADE.ALL, and made the change through the cascade (changing the father object).
CascadeType.PERSIST and orphanRemoval=true doesn't work together.
Try calling deleteById instead of delete on the repository. I also noticed that you are providing an Optional entity to the delete (since findOne returns an Optional entity). It is actually strange that you are not getting any compilation errors because of this. Anyways, my thinking is that the repository is not finding the entity to delete.
Try this instead:
#Test
public void testDeleteToken() {
assertThat(tokenRepository.findAll().size(), is(2));
Optional<Token> toDelete = tokenRepository.findOne(1L);
toDelete.ifExists(toDeleteThatExists -> tokenRepository.deleteById(toDeleteThatExists.getId()))
tokenRepository.flush();
assertThat(tokenRepository.findAll().size(), is(1));
}
By doing the above, you can avoid having to add the #Modifying query to your repository (since what you are implementing in that #Modifying query is essentially the same as calling deleteById, which already exists on the JpaRepository interface).
Related
I have an SpringBoot application in which I have defined an entity as given below
#Entity
public class Organisation {
#Id
#GeneratedValue
#JsonIgnore
private Long id;
private String entityId;
#OneToMany(mappedBy = "parent")
#Where(clause = "active_ind=true")
#JsonIgnore
private Set<Organisation> activeSubOrgs = new HashSet<>();
#ManyToOne(fetch = FetchType.LAZY)
#JoinColumn(name = "parentId")
private Organisation parent;
public Set<Organisation> getActiveSubOrgs() {
return activeSubOrgs;
}
In my service class I have a function to fetch the children
public Set<Organisation> getChildrenForEntity(String entityId) {
Organisation parent = organisationRepository.findByEntityIdAndActiveInd(entityId, true);
return parent.getActiveSubOrgs();
}
This works fine and gets the children when called from rest controller, but when I use the same function to test in junit, it always returns empty. In my sql trace log I see that query is not being triggred when getActiveSubOrgs() is being called. My junit test is as given below
#SpringBootTest
#RunWith(SpringRunner.class)
#Transactional
public class OrgServiceTest {
#Autowired
private OrganisationService organisationService;
#Before
public void setup() {
Organisation company = new Organisation("c", true);
company = organisationRepository.save(company);
Organisation circle = new Organisation("circle1", true);
circle.setParent(company);
circle = organisationRepository.save(circle);
Organisation div1 = new Organisation("div1", true);
div1.setParent(circle);
div1 = organisationRepository.save(div1);
}
#Test
public void getChildrenForEntitySuccessTest() {
Set<Organisation> children = organisationService.getChildrenForEntity("c");
System.out.println(children.iterator().next().getEntityId());
assertEquals("circle1", children.iterator().next().getEntityId());
}
The children Set in the test is empty when it should actually have circle1. I have tried calling Hibernate.initialize() on children, but that did not work either.
The issue is that bidirectional relationships have to be updated on both sides, i.e. the parent and the children have to know about each other. In your function setup(), you just define the parent of the child. Therefore, each child knows about its parent. However, the parent does not know about its children.
For bidirectional relationships, a good way of handling this is to define a function for one class to set/add a property and automatically update the other one. In case of OneToMany relationships, this can be handled nicely with add(entity) functions.
public void addActiveSubOrg(Organisation activeSubOrg) {
this.activeSubOrgs.add(activeSubOrgs);
activeSubOrg.setParent(activeSubOrg);
}
I am using Spring Data and #Transactional annotation(for automatic rollback after tests).
I have simple bidirectional relation between account and user(owning side):
#Entity
#Table(name = "ACCOUNT_T")
public class AccountEntity {
#Id
#GeneratedValue(strategy = GenerationType.IDENTITY)
private Long id;
private String email;
private String password;
private String verificationCode;
private Boolean active = false;
#OneToOne(mappedBy = "account", fetch = FetchType.EAGER,
cascade = {CascadeType.MERGE, CascadeType.PERSIST,
CascadeType.DETACH, CascadeType.REFRESH})
private UserEntity user;
}
#Entity
#Table(name = "USER_T")
public class UserEntity {
#Id
#GeneratedValue(strategy = GenerationType.IDENTITY)
private Long id;
private String name;
private String surname;
private String phone;
private LocalDate birthDate;
#OneToOne(cascade = CascadeType.ALL, orphanRemoval = true, fetch = FetchType.EAGER)
#JoinColumn(name = "account_id")
private AccountEntity account;
}
I am using JpaRepositories and fetching is set to eager.
Why sometimes when I get objects from database I can't get their child
objects-null is returned. It depends on from which side I add objects.
I have written simple test using Junit5:
#ExtendWith(SpringExtension.class)
#SpringBootTest
#Transactional
class UserAndAccountRepositoriesTest {
void testA() {
UserEntity userEntity = new UserEntity();
setUserProperties(userEntity);
AccountEntity accountEntity = new AccountEntity();
setAccountProperties(accountEntity); //just setting values for fields
accountEntity.setUser(userEntity);
accountRepository.save(accountEntity);
accountRepository.findAll().get(0).getUser(); //returns user
userRepository.findAll().get(0).getAccount(); //returns null,but should return account related to that user
}
void testB() {
UserEntity userEntity = new UserEntity();
setUserProperties(userEntity);
AccountEntity accountEntity = new AccountEntity();
setAccountProperties(accountEntity);
accountEntity.setUser(userEntity);
accountRepository.save(accountEntity);
accountRepository.findAll().get(0).getUser(); //again returns null,but shouldn't
userRepository.findAll().get(0).getAccount(); //returns account
}
}
Without #Transactional everything works fine - I am not getting null.
What am I doing wrong?
You'd need to set both sides of a relationship for explicitly defining it.
Try adding userEntity.setAccount(accountEntity) during your setup case, this would resolve the issue.
Hibernate won't help you and assume just because you set a -> b, it would set b <- a for you within the other entity.
The reason why it might work without #Transactional is that, without the annotation you are committing your setup data into whatever datasource you are using, and nothing is rollbacked at the end, and since you are selecting data without any id with findAll, you are getting previous user/account entites that have already been committed, some with relationship & some without, thus the random error you are getting.
It is because you are not setting account in userEntity. Please try like following:
userEntity.setAccount(accountEntity);
I will explain why the behavior is different depending on whether your are in a transaction or not :
When you are in a transaction :
a) Any get to fetch an entity A you have created prior to this transaction (so which is already in DB) will return a new object in term of memory adress, and hibernate will resolve its bidirectional relationship, even if you did not set it explicitly.
b) Any get to fetch an entity B you have created earlier in this transaction (so which is not yet in DB) will return the same object in term of memory adress, so it really is the same object, thus if you did not set its bidirectional relationship explicitly, it will not exist until you set it or until the transaction is over (as it will effectively persist B in DB) and you fetch B again.
When you are not in a transaction :
Any get to fetch any entity will behave like described in case a).
Conclusion :
The author was in case b).
I'm saving two entity related to each other. After it, I can get the first entity, but I get a NullPointerException when I try to get the second entity from the first entity. This is the example:
#Entity
#Table(name = "PARAMETRIZACION")
public class Parametrizacion {
#Id
#Column(name = "id_param", unique = true, nullable = false)
private Integer idParam;
#OneToMany(fetch = FetchType.LAZY)
private List<Arreglo> listArreglo;
}
And
#Entity
#Table(name = "ARREGLO")
public class Arreglo {
#Id
#Column(name = "id_arreglo", unique = true, nullable = false)
private Integer idArreglo;
}
And my Service:
#Service
#Repository
public class TestServiceImpl implements TestService {
#Override
#Transactional(rollbackFor = Exception.class)
public void methodTest(){
...
parametrizacionRepository.saveAndFlush(parametrizacion);//Id=1
...
arregloRepository.saveAndFlush(listArreglo);//Id=1
Parametrizacion paramFetch = parametrizacionRepository.findById(1);
Log.info("Param.Id=" + paramFetch.getIdParam());
Log.info("Size=" + paramFetch.getListArreglo().size());
}
}
The result for first log is: Param.Id=1
The result for second log is: NullPointerException
How can I get the full entity including his childrens? Only If I do this query after commit transaction I can get the data but I need Save data, Update data and Find data before do Commit on finish transaction.
Maybe there is a problem with the unidirectional relationship. Try adding some #ManyToOne field in the Arreglo class and declare how should they match by adding mappedBy="" to the #OneToMany annotation.
There are some nice examples how the relations should look like:
https://en.wikibooks.org/wiki/Java_Persistence/OneToMany
What you are doing is saving parametrizacion and listArreglo separately. And this don't set any relation for parametrizacion with Arreglo. You have to set listArreglo to parametrizacion's listArreglo variable and save only parametrizacion.
I have the following simple application
Users Entity
#Entity
public class Users implements Serializable {
#Id
#GeneratedValue
private long id;
private String name;
#OneToMany(mappedBy = "user", fetch = FetchType.EAGER, cascade = {CascadeType.ALL})
private Set<UserRoleUser> userRoleUser;
// GETTERS AND SETTERS
}
UserRole Entity
#Entity
public class UserRole implements Serializable {
#Id
#GeneratedValue
private long id;
private String roleName;
#OneToMany(mappedBy = "userrole", fetch = FetchType.LAZY, cascade = CascadeType.ALL)
private Set<UserRoleUser> userRoleUser;
// GETTERS AND SETTERS
}
UserRoleUser Many to many resolver class
#Entity
public class UserRoleUser implements Serializable {
#Id
#GeneratedValue
private long id;
#ManyToOne
#JoinColumn(name = "fk_userId")
private Users user;
#ManyToOne
#JoinColumn(name = "fk_userroleId")
private UserRole userrole;
// GETTERS AND SETTERS
}
UserRoleUserRepository
#Repository
#Transactional
public interface UserRoleUserRepository extends JpaRepository<UserRoleUser, Long>, QueryDslPredicateExecutor<UserRoleUser>{
}
Main Application class
#SpringBootApplication
#Configuration
public class Application {
public static void main(String[] args) {
ConfigurableApplicationContext context = SpringApplication.run(Application.class, args);
UserRoleUserRepository userRoleUserRepository = context.getBean(UserRoleUserRepository.class);
Iterable<UserRoleUser> findAll = userRoleUserRepository.findAll(QUserRoleUser.userRoleUser.id.gt(0));
for (UserRoleUser userRoleUser : findAll) {
userRoleUserRepository.delete(userRoleUser);
}
}
}
On running the main application, the database records in the UserRoleUser table are not being deleted. What could be the issue? I am using Spring Data and QueryDsl.
I have also tried putting the delete functionality on a Controller but still doesn't work.
#RestController
#RequestMapping("/api")
public class DeleteController {
#Autowired
UserRoleUserRepository userRoleUserRepository;
#GetMapping("/delete")
public String delete() {
Iterable<UserRoleUser> findAll = userRoleUserRepository.findAll(QUserRoleUser.userRoleUser.id.gt(0));
for (UserRoleUser userRoleUser : findAll) {
userRoleUserRepository.delete(userRoleUser);
}
return new Date().toString();
}
}
If you need to use the given methods provided by CrudRepository, use the JpaRepository.deleteInBatch(). This solves the problem.
The problem is the entities are still attached and will not be deleted until they become detached. If you delete by their id instead of the entity itself, it will delete them.
One thing I noticed is you are deleting the users one at a time which could lead to a database performance hit as the query will be recreated each time. The easiest thing to do is to add all the ids to a set then delete the set of ids. Something like this:
Set<Integer> idList = new HashSet<>();
for (UserRoleUser userRoleUser : findAll) {
idList.add(userRoleUser.getId());
}
if (!idList.isEmpty()) {
userRoleUserRepository.delete(idList);
}
then in your repository add the delete method
#Modifying
#Query("DELETE FROM UserRoleUser uru WHERE uru.id in ?1")
#Transactional
void delete(Set<Integer> id);
The reason why the child objects (UserRoleUser) are not being deleted upon userRoleUserRepository.delete(userRoleUser) call is that each UserRoleUser points to a Users which in turn holds a #OneToMany reference Set<UserRoleUser> userRoleUser.
As described in this StackOverflow answer, what your JPA implementation (e.g. Hibernate) effectively does is:
The cache takes note of the requested child exclusion
The cache however does not verify any changes in Set<UserRoleUser>
As the parent #OneToMany field has not been updated, no changes are made
A solution would go through first removing the child element from Set<UserRoleUser> and then proceed to userRoleUserRepository.delete(userRoleUser) or userRepository.save(user)
In order to avoid this complication two answers have been provided:
Remove element by Id, by calling userRoleUserRepository.deleteById(userRoleUser.getId()) : in this case the entity structure (and therefore the parent) is not checked before deletion. In the analog case of deleteAll something more convoluted such as userRoleUserRepository.deleteByIdIn(userRoleUserList.stream().map(UserRoleUser::getId).collect(Collectors.toList())) would have to be employed
Convert your CrudRepository to a JpaRepository and use its deleteInBatch(userRoleUserList) method. As explained in this article and this StackOverflow answer the deleteInBatch method tries to delete all records at once, possibly generating a StackOverflow error in the case the number of records is too large. As repo.deleteAll() removes one record at a time this error it minimizes this risk (unless the call is itself inside a #Transactional method)
According to this StackOverflow answer, extra care should be used when recurring to deleteInBatch as it:
Does not cascade to other entities
Does not update the persistence context, requiring it to be cleared (the method bypasses the cache)
Finally , as far as I know , there is no way this could be done by simply calling userRoleUserRepository.delete(userRoleUser) without first updating the parent object. Any updates on this (whether by allowing such behaviour through annotations, configuration or any other means) would be a welcome addition to the answer.
Page entity.
#Entity
#Table(name = "pages", schema = "admin")
public class Page implements Serializable {
private static final long serialVersionUID = 1L;
#Id
#GeneratedValue(strategy = GenerationType.IDENTITY)
#Column(name = "id", unique = true)
private Integer id;
#Column(name = "name")
private String name;
#ManyToOne(targetEntity = Partition.class, fetch = FetchType.LAZY)
private Partition partition;
#Column(name = "is_startable")
private Boolean isStartable;
#Column(name = "priority")
private Integer priority;
#Column(name = "prefix_granted_authority")
private String prefixGrantedAuthority;
#OneToMany(mappedBy = "page", cascade = CascadeType.ALL, fetch = FetchType.LAZY, orphanRemoval = true)
private List<Permission> permissions;
#Column(name = "link", unique = true)
private String link;
PageRepository
List<Page> findByPermissionsGroupsOrderByPartitionNameAscNameAsc(#Param(value = "group") Group group);
PageServiceImpl
#Override
public Collection<Page> getAccessedPages(Group group) {
try {
List<Page> pages = pageRepository.findByPermissionsGroupsOrderByPartitionNameAscNameAsc(group);
return pages;
} catch (Exception ex) {
logger.error("getPage error", ex);
return null;
}
}
getAccessedPages return real List of page entities(not null), but all fields in entities are null.
Why?
I also encounter this problem while ago, it looks like spring data does some kind lazy instantiation.
So if you not access this fields inside of your transaction, they will stay null. Add annotation #Transactional on method where are you calling this request and problem will be solved.
I wanted to expand on #user902383's answer, which ultimately also solved my issue, but it was too long for a comment.
In my case, I had repository method fetching an entity, Helper, called inside a #PostLoad listener that used Helper for calculations for filling a field in another entity, Child. The listener method was already annotated with org.springframework.transaction.annotation.Transactional.
When called by Child's repository it fetched a Helper entity with all fields filled, but when called by the repository of an entity Parent which had a child Child, it fetched an empty Helper object with only the id filled even though it was properly annotated.
The issue was that I was using this hack to access the repository outside of a Spring #Component (I couldn't make the listener a component). I suspect that the Spring magic for detecting when a field is dereferenced in a #Transactional method does not work when the repository was not properly #Autowired. I still do not know why it worked in Child's repository but not in Parent.
My solution to this particular problem was moving the repository call and dereferencing to a #Service, which properly #Autowires the repository, and doing the hackish static call for getting that service instead, which makes for better code structure anyway.