JPA2 : N+1 select even with a query with JOIN - java

In my application, I have an entity A with a list of entities B that should be fetched eagerly :
#Entity
public class A
{
...
/* #OrderBy("cValue.id ASC") */
#OneToMany(fetch = FetchType.EAGER, cascade = CascadeType.ALL)
#JoinColumn(name="A_ID", nullable=false)
private List<B> BEntries = new ArrayList<B>();
...
}
#Entity
public class B
{
...
#ManyToOne
#JoinColumn(name = "C_ID", nullable = false)
private C cValue;
...
}
In order to get the list of A, I was first doing this simple query :
CriteriaBuilder critBuilder = em.getCriteriaBuilder();
CriteriaQuery<A> critQuery = critBuilder.createQuery(A.class);
Root<A> critRoot = critQuery.from(A.class);
critQuery.select(critRoot);
But there I saw that Hibernate was doing N+1 select queries on the database, 1 on class A, and N on class B (where N is the number of tuples of A in DB).
I was very surprise that, for eager fetching, Hibernate was not directly doing a LEFT JOIN query.
So I first tried to use the annotation #Fetch(FetchMode.JOIN) of Hibernate, but it was not working as expected.
So I transformed my list query with the following additional instructions:
Join<A,B> joinAB = critRoot.join(A_.BEntries, JoinType.LEFT);
joinAB.join(B_.cValue, JoinType.LEFT);
Ok, now the resulting SQL query contains all the needed LEFT JOIN to build the full A object eagerly... but it's still doing the other N queries on B table!
I first thought it was coming from the #OrderBy annotation I put on the Bentries parameter, but even when removed, it's still doing N+1 selects instead of 1...
Any idea why it's behaving like this?... and even why it's not doing a LEFT JOIN by default on eagerly fetched collections in entities?

Related

Hibernate left join fetch on many returns duplicities

I have a problem, I am using Hibernate via Spring data and I have this data model (model is much bigger but this is the part that causes problems).
Lets assume we have Entity A and Entity B.
Relationship between two entities are Many-to-Many.
I am trying to fetch A records with B recordes fetched (to prevent lazy loading).
The model is connected like this
On entity A:
#ManyToMany(mappedBy = "items")
private List<BEntity> bs = new ArrayList<>();
On entity B
#ManyToMany
#JoinTable(
name = "a_b",
joinColumns = { #JoinColumn(name = "b_id", referencedColumnName = "id") },
inverseJoinColumns = { #JoinColumn(name = "a_id", referencedColumnName = "id") }
)
private Collection<AEntity> as = new ArrayList<>();
So its typical M:N relationship. And I am trying to retrieve A data like following
#Query("SELECT a FROM AEntity a " +
" LEFT JOIN FETCH a.bs" +
"WHERE a.id IN (:aIds)")
List<AEntity> findX(#NonNull #Param("aIds") Collection<Long> aIds);
The resulted SQL is something like
select X -- select fields omitted for simplicity
from item itembo0_
left outer join a_b ab on a.id = ab.a_id
left outer join b b on ab.b_id = b.id
where a.id in (...)
Which is a thing I would expect. The SQL will result duplicites (which I would expect as well cause there might be many B records which each has one result row). But at the end, hibernate does not merge all these rows into a A entity with fetched B fields.
For example, when I pass a 5 IDS into "in" condition, I get a 10 A records. Each one has a 2 B records linked! Thats the weird part.
Is there anyone who can tell me why hibernate does not merge these SQL results by A.id identifier and makes duplicites? Is it because I am asking for a List instead of Set?
Using the "DISTINCT" keyword in your query should be enough to avoid duplicates:
#Query("SELECT DISTINCT a FROM AEntity a " +
" LEFT JOIN FETCH a.bs" +
"WHERE a.id IN (:aIds)")
List<AEntity> findX(#NonNull #Param("aIds") Collection<Long> aIds);
I do not know what kind of data you store, but perhaps a better solution, as you have mentioned, would be to use a Set instead of a List.

DTO projection on a recursive structure

In my data model there is an entity "location" which is recursively. Furthermore there are relations to other entities.
The corresponding JPA (Spring Data JPA) entity looks like:
#Entity
#Table(name = "location")
class Location{
#OneToMany(mappedBy = "parent", orphanRemoval = true)
#OrderBy("name ASC")
Set<Location> children = null
#ManyToOne(fetch = FetchType.EAGER)
#JoinColumn(name = "parent_id")
Location parent = null
#Column(name = "name")
String name = null
#OneToMany(mappedBy = "location", fetch = FetchType.EAGER)
Stops stops = null
...
What is the most performant way to do a read only query? I just need the information inside the entity (table location) with the complete recursive structure but no information from the related entities.
I've read the phrase DTO projection, but nothing about what to do with a recursive structure.
Reading a recursive structure is usually done by making use of what SQL calls a recuresive CTE. JPA does not support that out of the box, because not all RDBMS support it. If you know that your DBMS supports it, you can make use of the following SQL to do this:
WITH RECURSIVE nodes(id, parent_id) AS (
SELECT id, parent_id FROM location l where id = ?
UNION ALL
SELECT l.id, l.parent_id FROM nodes n JOIN location l ON n.parent_id = l.id
)
SELECT id, parent_id FROM nodes
With that you get a list of a specific and all parent location ids as well as their respective parents which is flat. You will have to bring structure into this.
List<Object[]> result = //get the result of the query
Map<Integer, LocationDto> locationMap = new HashMap<>();
result.forEach(r -> locationMap.put(result.get(0), new LocationDto(result[0], result[1])));
locationMap.values().forEach(l -> l.setParent(locaitonMap.get(l.getParentId())));
If you don't want to make use of plain SQL because of portability concerns or just because you don't want to give up on your abstraction, you can make use of Blaze-Persistence which works on top of JPA and adds support for CTEs. Your query with blaze-persistence would look like this
List<LocationCte> result = criteriaBuilderFactory.create(entityManager, LocationCte.class)
.withRecursive(LocationCte.class)
.from(Location.class, "l")
.bind("id").select("l.id")
.bind("parent").select("l.parent.id")
.where("id").eq(initialId)
.unionAll()
.from(Location.class, "l")
.innerJoinOn(LocationCte.class, "cte")
.on("cte.parent").eqExpression("l.id)
.end()
.bind("id").select("l.id")
.bind("parent").select("l.parent.id")
.end()
.from(LocationCte.class)
.getResultList();
You will also need this special entity class
#CTE
#Entity
public class LocationCte {
#Id Integer id;
Integer parent;
}

How can I use Hibernate to load a list of entities and a subset of those entities' related entities?

I have a query I'd like to run against a table, let's call it parent where I'm grabbing all rows that match a certain criteria. In SQL:
select * from parent where status = 'COMPLETE';
I have this table and another related table (child) defined as Hibernate entities such that:
#Entity
#Table(name = "parent")
public class Parent {
//...
#OneToMany(mappedBy = "parent")
private Set<Child> children;
//...
}
#Entity
#Table(name = "child")
public class Child {
//...
#ManyToOne
#JoinColumn(name = "parent_id")
private Parent parent;
#Column(name = "key")
private String key;
//...
}
I'd like my query to ALSO pull two optional child records where key is one of two values. So, in SQL:
select *
from parent p, child ca, child cb
where p.status = 'COMPLETED'
and p.id *= ca.parent_id
and ca.key = 'FIRST_KEY'
and p.id *= cb.parent_id
and cb.key = 'SECOND_KEY';
I could do this in Hibernate by just grabbing the result from the first query and iterating over the children collection looking for the rows I want but that's terribly inefficient: one query that does two outer joins vs one query + an additional query for each looking for the children I care about.
Is there a way to replicate the outer joins in the query above in Hibernate where the objects returned will have the children collection only populated with the two (or less, since they are optional) entities I am interested in?
You don't need two outer joins. You could simply use this HQL and Hibernate will add the right children to the right parent:
List<Parent> parentList = session
.createQuery("from Parent p left join fetch p.children c" +
" where p.status = 'COMPLETE' " +
" and (c.key = 'FIRST_KEY' or c.key = 'SECOND_KEY')")
.setResultTransformer(Criteria.DISTINCT_ROOT_ENTITY)
.list();
for(Parent parent:parentList) {
System.out.println(parent);;
}
Hope that solves your problem.

Hibernate DetachedQuery and Spring's HibernateTemplate with Restriction giving wrong results

My data structure is like this
Department
-> Employees
-> Gender
-> CityID -> Cities
->CityID
->CountryID -> Countries
-> CountryID
Department Class:
public class Department {
#OneToMany(mappedBy = "departmentid", cascade = CascadeType.ALL, fetch = FetchType.EAGER)
private Set<Employee> employees = new HashSet<>();
}
I build Crteria like this:
DetachedCriteria criteria = DetachedCriteria.forClass(Department.class);
DetachedCriteria detlCrit = criteria.createCriteria("employees");
detlCrit.add(Restrictions.eq("gender", "MALE"));
detlCrit.setResultTransformer(Criteria.DISTINCT_ROOT_ENTITY);
I have 1 Department, 2 Employees in the Tables (1 male, 1 female).
When I excecute this criteria iam expecting Hibernate build one 'Department' object, one 'Employee' object, and city, country etc.,
But what iam getting is 1 Department, 2 Employees.
When I see the queries executed by Hibernate in logs, it shows two queries
First Query:
Select * from Department, Employee
Left outer join City on Employee.cityID = City.cityID
Left outer join Country on City.countryID = City.countryID
Where Employee.DeptID = Department.DeptID
AND Employee.Gender = 'MALE';
Second query:
Select * from Employee
Left outer join City on Employee.cityID = City.cityID
Left outer join Country on City.countryID = City.countryID
Where Employee.DeptID = Department.DeptID;
Second query is wrong there is no Restriction applied on Gender='MALE';
What iam doing wrong? any suggestions? how to solve this?
sorry queries may be not exactly correct, but you got the idea.
Any more details needed please ask, I can provide.
Thanks in advance..
Try this,using SessionFactory.
#Autowired
private SessionFactory sessionFactory;
Criteria criteria = sessionFactory.getCurrentSession().createCriteria(Department.class);
criteria.add(Restrictions.eq("gender", "MALE"));
Hope I was useful.
The first query is selecting Department entities and the filtering is applied as you specified in your where clause.
But you cannot truncate associations, you always have to fetch them all eagerly or lazily. That's because Hibernate has to maintain consistency guarantees when flushing back the loaded Department entity and possibly cascading the employees state back to the database.
The second query is most likely because you use a FetchType.EAGER on your employees collection:
#OneToMany(fetch = FetchType.EAGER, cascade = CascadeType.ALL, mappedBy = "department", orphanRemoval = true)
private List<Employee> employees = new ArrayList<>();
Once the Department is fetched, the employee collection is fetched eagerly as well.
Try with an HQL query liken this one:
select distinct d
from Department d
left join fetch d.employees e
where e.gender = :gender

Hibernate Parent/Child SELECT N+1 issue

I jave the following mapped superclass that provides a basic implementation for a parent/child self relationship to create a parent/child list for unlimited nesting of items (i.e. Categories)
#MappedSuperclass
public abstract class ParentChildPathEntity<N extends ParentChild> implements MaterializedPath<N> {
#ManyToOne(fetch=FetchType.LAZY)
#JoinColumn(name = "parent_id")
private N parent;
#Column(name = "name", unique = true)
private String name;
#OneToMany(mappedBy = "parent", fetch = FetchType.LAZY, cascade = CascadeType.ALL)
private Set<N> children = new HashSet<N>();
If I load the entire table with fetch join on the parent and children, a single select loads all the records and i can happily traverse the tree. my problem comes in when i specify to retrieve a node on the tree. i want the node and all its children in a single select. below is the hql for loading the entire table:
hql.append(String.format("tree from %s tree ", tableName));
hql.append("left join fetch tree.parent ");
hql.append("left join fetch tree.children ");
if i specify the node name, i.e.:
where tree.name = :name
then hibernate retrieves the node, but when i access the children i get the SELECT N+1 issue. I realize why this is happening, (because of the tree.name = :name) but is there a way to write the HQL so it loads the specified node and all its children?
I'm just trying to figure out a way to support a simple nested item's list where i can retrieve any parent node and its children with a single select
thanks in advance,
Have you tried using the #BatchSize annotation?
#BatchSize(size = 20)
Ex:
#OneToMany(mappedBy = ..., fetch = FetchType.LAZY)
#BatchSize(size = 20)
public SortedSet<Item> getItems() { ... }
Then, if you specify the join to children in your HQL, you should be able to avoid n+1 select. I am not sure, offhand, if there is a way to specify the batch size in the HQL statement.

Categories

Resources