Persistence Layers In An Enterprise Application:

1
Persistence Layers In An Enterprise Application
An Evolution From SQL to OJB to Hibernate

• Scott Delap, Amit Gollapudi, Scott GelbDemand
  Management, Inc.
• www.demandsolutions.com

2
Imagine If Your Application

• Instantly supported all relational databases
• Wrote its own SQL
• Modified your database schema to support its
  object model

3
Presentation Goal
Explore the need for persistence layers and
identify their common features through practical
examples.
4
Agenda

• The Beginning Entity Beans Are Painful
• Practical Persistence Layers
• OJB The Transition
• Practical OJB
• Hibernate The Transition
• Practical Hibernate

5
Agenda

• The Beginning Entity Beans Are Painful
• Practical Persistence Layers
• OJB The Transition
• Practical OJB
• Hibernate The Transition
• Practical Hibernate

6
The Beginning Entity Beans Are Painful
Early persistence options 2001

• EJBs were marketed as the better way
• Container Managed Persistence
• Bean Managed Persistence

7
The Beginning Entity Beans Are Painful
BMP with DAOs and JDBC

• Each object/table had a Data Access Object (DAO)
• Each DAO used JDBC
• Multiple DAO implementations were often needed to
  support multiple databases

8
The Beginning Entity Beans Are Painful
Sample schema Scott/Tiger
9
The Beginning Entity Beans Are Painful
Example DAO JDBC code

• String sqlString insert into emp(ename, mgr,
  job) values (?, ?, ?)
• PreparedStatement ps connection.prepareStatement
  (sqlString)
• ps.setString(1, e.getName())
• ps.setString(2, e.getJob())
• ps.setInt(3, e.getManager().getId)
• ps.executeUpdate()

10
Id rather go to the dentist
Soon it was discovered that this repetitive
process was less fun than going to the dentist.
11
Agenda

• The Beginning Entity Beans Are Painful
• Practical Persistence Layers
• OJB The Transition
• Practical OJB
• Hibernate The Transition
• Practical Hibernate

12
Practical Persistence Layers
Why a persistence layer?

• Refocus on business logic
• Avoid monotonous SQL coding
• Support for multiple databases
• Object-oriented approach
• Relationships easily defined and maintained

13
Practical Persistence Layers
Problem Generated IDs

• Database independence
• External processes inserting data
• Performance
• Clustering

14
Practical Persistence Layers
Solution Generated IDs

• Native generation
• Sequence tables
• In-memory sequence

15
Practical Persistence Layers
Problem Querying data

• Complex queries
• Aggregate functions
• Multiple table joins
• Arithmetic and string manipulation
• Null value handling

16
Practical Persistence Layers
Solution Hand-written SQL

• Write SQL against persistence layernot database
• Reduces SQL changes if objects change
• Leverage persistence layer to handle SQL syntax
  differences
• Example coalesce() vs. nvl()
• Common rules still apply
• PreparedStatements
• StringBuffers

17
Practical Persistence Layers
Problem Caching decisions

• Minimize database access
• Dirty caches
• Clustering

18
Practical Persistence Layers
Solution Caching decisions

• No cache
• Local cache
• Distributed cache
• Cacheable object vs. non-cacheable (often
  updated)

19
Practical Persistence Layers
Problem Manual schema creation

• Complexity in supporting multiple databases
• Maintenance costs

20
Practical Persistence Layers
Solution Automate it!

• Multiple database DDLs easy to maintain
• Object changes instantly reflected in the schema
• Saves time
• Reduces logic error

21
Agenda

• The Beginning Entity Beans Are Painful
• Practical Persistence Layers
• OJB The Transition
• Practical OJB
• Hibernate The Transition
• Practical Hibernate

22
OJB The Transition
Why OJB 2002?

• Mature
• Large user base
• Performance
• Most feature-rich API
• Strong project momentum

23
OJB The Transition
Lines of code removed

• Greater than 90 for some DAOs
• Almost 20 (6000 lines) in one month

24
OJB The Transition
Basic changes

• Domain object refactor
• Added foreign key ID fields

25
OJB The Transition
Basic changes (cont.)

• DAOs simplified
• Several obsolete
• Database-dependent code removed

26
OJB The Transition
Implementation choices

• PersistenceBroker API
• Flexibility
• Performance
• Database level transactions
• ODMG API
• Built on top of PersistenceBroker
• 3.0 compliant
• Object level transactions

27
OJB The Transition
Example changes Before

• String sqlString insert into emp(ename, mgr,
  job) values (?, ?, ?)
• PreparedStatement ps connection.prepareStatement
  (sqlString)
• ps.setString(1, e.getName())
• ps.setString(2, e.getJob())
• ps.setInt(3, e.getManager().getId)
• ps.executeUpdate()

28
OJB The Transition
Example changes After

• PersistenceBroker broker DSPersistenceBrokerFact
  ory.createPersistenceBroker()
• broker.store(e)
• --------------------------------------------------
  -----------
• ltclass-descriptor idemployee
• classjavaone.example.Employee
• tableEmployeegt
• ltfield-descriptor id1
• nameid columnid
• length38 jdbc-typeINTEGER
• java-typeint primarykeytrue/gt
• ltfield-descriptor id2
• namenamecolumnname
• jdbc-typeVARCHAR/gt

29
Agenda

• The Beginning Entity Beans Are Painful
• Practical Persistence Layers
• OJB The Transition
• Practical OJB
• Hibernate The Transition
• Practical Hibernate

30
Practical OJB
Generated IDs

• Store in database table
• Optimize GrabSize
• No nice native concept
• Synchronization with non-OJB insertions
• Clear table

31
Practical OJB
Querying data Report query

• String fields new String "name",
  "hireDate"
• Criteria criteria new Criteria()
• criteria.addEqualTo("departmentId", new
  Integer(7))
• Query query QueryFactory.newReportQuery(Employe
  e.class, fields, criteria, true)
• Iterator it broker.getReportQueryIteratorByQuer
  y(query)
• while (it.hasNext())
• Object row (Object) it.next()
• System.out.println("Name " row0 "
  Hire Date " row1)

32
Practical OJB
Querying data Writing custom SQL

• ClassDescriptor descriptor broker.getClassDescr
  iptor(Employee.class)
• EmployeeSqlGenerator.employeeTable
  descriptor.getFullTableName()
• EmployeeSqlGenerator.id descriptor.getF
  ieldDescriptorByName(Employee.ID).getColumnName()
  
• EmployeeSqlGenerator.name descriptor.getFie
  ldDescriptorByName(Employee.NAME).getColumnName()
  
• StringBuffer buffer new StringBuffer()
• buffer.append(select ).append(id)
• buffer.append(, ).append(name)
• buffer.append( from ).append(employeeTable)

33
Practical OJB
Caching Choices

• Default (ObjectCacheDefaultImpl)
• SoftReferences
• Timeout
• Optionally transaction aware
• Short-life cache (ObjectCachePerBrokerImpl)
• Flushed upon broker release
• Flushed upon transaction completion

34
Practical OJB
Caching Choices (cont.)

• No cache (ObjectCacheEmptyImpl)
• Complex implementations (JCSImpl/OSCacheImpl)
• Clustering

35
Practical OJB
Caching Tuning

• CacheFilter
• Cacheable property per class-descriptor

36
Practical OJB
Schema creation OJB and Torque

• Leverage Torques DDL generation capabilities
• Repository.xml already contained most of the
  needed information
• Convert repository.xml to Torque
• Submitted patch

37
Practical OJB
Problem OJB?

• Monolithic repository.xml
• Debugging
• All mapped classes are required in classpath
• Custom query support
• Lacking good OQL implementation

38
Practical OJB
Problem OJB? (cont.)

• O/R support limitations
• Collections (ex ordered list, maps)
• Redundant field declarations for referential
  integrity
• Discriminator-less polymorphism

39
Practical OJB
Problem OJB? (cont.)

• Project momentum
• Direction
• Doesnt match our organizations needs
• Lacking in cooperation
• Patches
• Mailing list

40
Agenda

• The Beginning Entity Beans Are Painful
• Practical Persistence Layers
• OJB The Transition
• Practical OJB
• Hibernate The Transition
• Practical Hibernate

41
Hibernate The Transition
Why Hibernate 2004?

• Mature API
• Strong development community
• Comprehensive documentation
• Variety of conversion tools (including MiddleGen
  support)
• Powerful query language (HQL)
• Future JDO implementation

42
Hibernate The Transition
Example changes Before (OJB)

• ltclass-descriptor idemployee
• classjavaone.example.Employee
• tableEmployeegt
• ltfield-descriptor id1
• nameid
• columnid
• length38 jdbc-typeINTEGER
• java-typeint primarykeytrue/gt
• ltfield-descriptor id2
• namename
• columnname
• jdbc-typeVARCHAR/gt

43
Hibernate The Transition
Example changes After (Hibernate)

• ltclass namejavaone.example.Employee
  tableEmpgt
• ltid nameid columnEmpNo
• typejava.lang.Integergt
• ltgenerator classnative /gt
• lt/idgt
• ltproperty namename type"java.lang.Stringgt
  
• ltcolumn nameeName length10
• not-nullfalse /gt
• lt/propertygt

44
Hibernate The Transition
Example changes Before/After

• OJB
• PersistenceBroker broker DSPersistenceBrokerFact
  ory.createPersistenceBroker()
• broker.store(e)
• Hibernate
• Session session configuration.buildSessionFactor
  y().openSession()
• session.save(e)

45
Hibernate The Transition
Using the session

• Implicit database/memory synchronization
• Session-level cache
• Contains(), evict(), clear()

46
Hibernate The Transition
Basic changes

• Default (no-argument) constructor required
• Composite IDs require
• equals()
• hashCode()
• Serializable
• Unchecked versus checked exceptions

47
Hibernate The Transition
Basic changes (cont.)

• Foreign key fields no longer required

48
Hibernate The Transition
Problem hbm generation

• Class to hbm
• Object relationships/relational relationships
• Type conversion
• DDL to hbm (Middlegen support)
• Complex relationships
• Composite keys
• Optimal collection mapping

49
Hibernate The Transition
Solution OJB2Hbm

• Convert OJB repository.xml
• Minimize code changes
• Preserve DDL
• Indexes retained
• Optimal collection representation
• Available on SourceForge
• http//ojb2hbm.sf.net

50
Agenda

• The Beginning Entity Beans Are Painful
• Practical Persistence Layers
• OJB The Transition
• Practical OJB
• Hibernate The Transition
• Practical Hibernate

51
Practical Hibernate
Generate IDs

• Native generation
• Identity column for MS SQL
• Sequences for Oracle/PostgreSQL
• Hi/Lo if database does not provide native
  generation
• Specify unsaved values for transient objects

52
Practical Hibernate
Querying data HQL with query substitution

• ltproperty name"hibernate.query.substitutions"gt
• toLowercase lower
• lt/propertygt
• --------------------------------------------------
  ----
• select emp.id, emp.name, emp.hireDate from
  Employee emp where toLowercase(emp.name) name
• select employee0_.EmpNo as x0_0_,
  employee0_.eName as x1_0_, employee0_.hireDate as
  x2_0_ from Emp employee0_ where
  (lower(employee0_.eName)? )

53
Practical Hibernate
Querying data Report query

• String sql "select emp.name, emp.hireDate "
• "from Employee emp "
• "where emp.departmentId ?"
• List employees session.find(sql, new
  Integer(7), Hibernate.INTEGER)
• Iterator it employees.iterator()
• while (it.hasNext())
• Object row (Object) it.next()
• System.out.println("Name " row0 " Hire
  Date " row1)

54
Practical Hibernate
Caching decisions

• Second level cache
• Specify by class and/or collection
• Multiple usage strategies
• Supported implementations
• EHCache (default)
• HashtableCacheProvider (non-production use)
• JCSCache (deprecated)
• OSCache
• SwarmCache
• JBossTreeCache
• Query cache

55
Practical Hibernate
Schema creation

• SchemaExport
• Converts hbm mappings to DDL
• Available as ANT task
• SchemaUpdate
• Incremental updates
• Depends on JDBC metadata

56
Summary
What have we learned

• Dont believe the hype
• OJB is a viable option
• Hibernate is a viable option
• Keep an eye on JDO
• Reduce coupling between your application and
  persistence layer
• Hand-written SQL is often necessary
• Automated schema creation rocks!

57
Persistence Layers In An Enterprise Application
An Evolution From SQL to OJB to Hibernate

• Scott Delap, Amit Gollapudi, Scott GelbDemand
  Management, Inc.
• www.demandsolutions.com