The document discusses various approaches to improve performance of Java enterprise applications using object-relational mapping (ORM), including: 1) Using data transfer objects (DTOs) and constructor expressions to fetch data from multiple tables into flat POJOs in a single query, avoiding the N+1 select problem. 2) For some use cases, native SQL queries with constructor mapping or libraries like QLRM or jOOQ may perform better than ORM for all queries. 3) The document recommends using entities for command/write operations and DTOs with techniques like constructor expressions for read/query operations when performance is important.

1. PERFORMANTE
JAVA ENTERPRISE APPLIKATIONEN
TROTZ O/R-MAPPING
Simon Martinelli, simas GmbH
@simas_ch | about.me/simas_ch
https://github.com/simasch/orders

2. DAS PROBLEM

3. DAS MODELL

4. N+1 SELECT PROBLEM
• Orders und OrderItems = FetchType.LAZY
@OneToMany(mappedBy = "customer", cascade = CascadeType.ALL, orphanRemoval =
true)
private Set<Order> orders;
@OneToMany(mappedBy = "order", cascade = CascadeType.ALL, orphanRemoval = true)
private List<OrderItem> items;
• Ein Query für alle Customers
• Pro Customer 1 Query für die Orders
• Pro Order 1 Query für die OrderItems

5. LÖSUNGSANSÄTZE
• FetchType.EAGER oder EntityGraph (JPA 2.1)
– Nur ein Hinweis für JPA
– != SQL JOIN
– Hibernate erlaubt nur eine List mit
FetchType.EAGER pro Entity
org.hibernate.HibernateException: cannot simultaneously fetch
multiple bags
• JOIN FETCH
– Achtung vor kartesischen Produkten

6. DTO
Quelle: Martin Fowler, http://martinfowler.com/eaaCatalog/dataTransferObject.html

7. CustomerInfoDTO
public class CustomerInfoDTO {
private final Long id;
private final String lastname;
private final String firstname;
private final double revenue;
public CustomerInfoDTO(Long id, String lastname, String firstname,
double revenue) {
this.id = id;
this.lastname = lastname;
this.firstname = firstname;
this.revenue = revenue;
}
...

8. CONSTRUCTOR EXPRESSION
Query q = em.createQuery(
"SELECT " +
"NEW control.CustomerInfoDTO(c.id, c.lastname, c.firstname," +
"SUM(i.product.price)) " +
"FROM Customer c " +
"JOIN c.orders o " +
"JOIN o.items i " +
"GROUP BY c.lastnamem, c.firstname" +
"ORDER BY c.lastname, c.firstname");
List<CustomerInfoDTO> list = q.getResultList();

9. DAS DATENMODELL IM ZENTRUM
• Ein 1-1 Abbild der Datenbank in Entities oft
unnötig
X

10. WAS IST MIT JPQL OHNE DIE
BEZIEHUNG CUSTOMER->ORDER?
NEU IN JPA 2.1
Query q = em.createQuery(
"SELECT " +
"NEW control.CustomerInfoDTO(c.id, c.lastname, c.firstname, " +
"SUM(i.product.price)) " +
"FROM Customer c " +
"JOIN c.orders o ON o.customerId = c.id " +
"JOIN o.items i " +
"GROUP BY c.lastname, c.firstname" +
"ORDER BY c.lastname, c.firstname");
List<CustomerInfoDTO> list = q.getResultList();

11. ORM FÜR ALLES?
Just because you're using Hibernate, doesn't
mean you have to use it for everything.
A point I've been making for about ten years
now.
Gavin King, 10.12.2013

12. SQL?
• VORTEILE
– Testbar mit SQL Developer/TOAD/IDE
– Optimierbar (Execution Plan)
• ABER SQL VON HAND? NEIN!
– JPA 2.1 ConstructorResult
– QLRM
– jOOQ

13. JPA 2.1 CONSTRUCTOR RESULT
Query q = em.createNativeQuery(
"SELECT C.ID, C.LASTNAME, C.FIRSTNAME, SUM(P.PRICE) AS REVENUE" +
"FROM CUSTOMERS C " +
"JOIN ORDERS O ON O.CUSTOMER_ID = C.ID " +
"JOIN ORDERITEMS I ON I.ORDER_ID = O.ID " +
"JOIN PRODUCTS P ON P.ID = I.PRODUCT_ID " +
"GROUP BY C.ID, C.LASTNAME, C.FIRSTNAME " +
"ORDER BY C.LASTNAME, C.FIRSTNAME", "CustomerInfoDTO");
@SqlResultSetMapping(name="CustomerInfoDTO",
classes={
@ConstructorResult(targetClass=CustomerInfoDTO.class,
columns={@ColumnResult(name="ID"),
@ColumnResult(name="LASTNAME"),
@ColumnResult(name="FIRSTNAME"),
@ColumnResult(name="REVENUE", type=Double.class)})
})

14. QLRM
Query q = em.createNativeQuery(
"SELECT C.ID, C.LASTNAME, C.FIRSTNAME, SUM(P.PRICE) AS REVENUE" +
"FROM CUSTOMERS C " +
"JOIN ORDERS O ON O.CUSTOMER_ID = C.ID " +
"JOIN ORDERITEMS I ON I.ORDER_ID = O.ID " +
"JOIN PRODUCTS P ON P.ID = I.PRODUCT_ID " +
"GROUP BY C.ID, C.LASTNAME, C.FIRSTNAME " +
"ORDER BY C.LASTNAME, C.FIRSTNAME");
JpaResultMapper mapper = new JpaResultMapper();
List<CustomerInfoDTO> list =
jpaResultMapper.list(q, CustomerInfoDTO.class);
QLRM: https://github.com/simasch/qlrm

15. jOOQ
List<CustomersInfoDTO> list = create.
select(CUSTOMERS.ID, CUSTOMERS.LASTNAME, CUSTOMERS.FIRSTNAME,
sum(PRODUCTS.PRICE)).
from(CUSTOMERS).
join(ORDERS).on(ORDERS.CUSTOMER_ID.eq(CUSTOMERS.ID)).
join(ORDERITEMS).on(ORDERITEMS.ORDER_ID.eq(ORDERS.ID)).
join(PRODUCTS).on(PRODUCTS.ID.eq(ORDERITEMS.PRODUCT_ID)).
groupBy(CUSTOMERS.ID, CUSTOMERS.NAME).
orderBy(CUSTOMERS.NAME).
fetchInto(CustomersInfoDTO.class);
jOOQ: http://www.jooq.org

16. CQRS
Quelle: Martin Fowler, http://martinfowler.com/bliki/CQRS.html

17. CQRS IM KLEINEN
• QUERIES
– JPA Constructor Expression
– JPA ConstructorResult
– QLRM
– jOOQ
• COMMANDS
– JPA Entities

18. EMPFEHLUNGEN
• Entities zum Erstellen und Ändern der Daten
• DTO für lesende Zugriffe
– Constructor Expression
– SQL mit QLRM oder jOOQ

19. BONUS MATERIAL

20. RANDOM-DATA-GENERATOR
RandomDataGenerator randomDataGenerator = new RandomDataGenerator();
List<Customer> customers = randomDataGenerator.generateList(
400,
new GenConfig()
.name(Name.Firstname, "firstname")
.name(Name.Lastname, "lastname"),
Customer.class);

21. LOG4JDBC
Anpassungen im META-INF/persistence.xml
<property name="javax.persistence.jdbc.url"
value="jdbc:log4jdbc:derby://localhost:1527/orders"/>
<property name="javax.persistence.jdbc.driver"
value="net.sf.log4jdbc.sql.jdbcapi.DriverSpy"/>
log4jdbc-log4j2: https://code.google.com/p/log4jdbc-log4j2/