Class mapping
Information Repositories
Oct-23 [email protected] 2
Mapping scenarios
Green Field
It is a new project
The design is not conditioned by previous
versions
The mapper generates the database structure
Legacy
We upgrade or extend an already existing system
The database is already present and is plenty of
data
The mapping has to be adapted to an existing DB,
probably with a convoluted design
Oct-23 [email protected] 3
Requirements of a class to be
mapped
Plain Java classes can be mapped (POJO)
Default constructor mandatory
Setters/getters etc., as always
Collections over interfaces: Set<>, List<>, etc.
Extra information need for persistence is
added as metadata
@Annotations
xml
Oct-23 [email protected] 4
Oct-23 [email protected] 5
Entity
POJO Ejemplo (entity)
Oct-23 [email protected] 6
Value Type
Oct-23 [email protected] 7
@Id position Access by getter/setters
Access by attributes
The mapper does this to
load the object…
Oct-23 [email protected] 8
Oct-23 [email protected] 9
Metadata with annotations
Add to Java 5
@Entity, @Embeddable, @Id, etc.
They decorate the class to map
Comfortable for the programmer
Compiled, implies early error detection
Oct-23 [email protected] 10
Metadata as XML, example
Oct-23 [email protected] 11
Metadata with XML
On the file orm.xml
This is referenced from persistence.xml
XML override Annotations
At deploy time an administrator can make
adjustments to improve performance
without touching the source code
Oct-23 [email protected] 12
Annotations vs XML
Oct-23 [email protected] 13
Oct-23 [email protected] 14
Categories of annotations
Entity Inheritance
Database Schema Locking
Identity Lifecycle
Direct Mappings Entity Manager
Relationship Queries
mappins
Composition
Oct-23 [email protected] 15
Annotations by category
@AccessType
@OrderColumn
Oct-23 [email protected] 16
Annotations by category
@ElementCollection
Oct-23 [email protected] 17
Oct-23 [email protected] 18
Entities
An entity represents a domain concept
Can be associated with other entities
Has an independent life cycle
Has identity
An entity is always mapped to a table
(or several in rare cases)
Oct-23 [email protected] 19
Entities
@Entity
Marks a class as an entity
@Table (optional)
Sets params for the table
By default the
table will be
called the same
as the class
Oct-23 [email protected] 20
Persistence of attributes
Java types: the mapper knows how to map them
String, Double, Long, Date, LocalDate, Integer, Short, Boolean, Byte…
Domain Value Types:
All the attributes will be embedded in the entity
table
Class marked with @Embeddable, or entities’
attributes as @Embedded
Association ends
Foreign Key to the other entity table
Rest of cases, serialization
The type of the attribute must be Serializable
Oct-23
[email protected] 21
@Colum
Influences the generation of DDL
By default, (without @Column) each attribute in the
class is a column on the table with the same name as
in the class
Oct-23 [email protected] 22
@Column, attributes
Oct-23 [email protected] 23
@Basic
Applies to an attribute
Defines how the mapper will behave regarding that
attribute
Of application to:
Oct-23 [email protected] 24
@Enumerated
How enumerated elements are stored
EnumType.ORDINAL
EnumType.STRING
The column of the
table is of type
INTEGER or
VARCHAR
Oct-23 [email protected] 25
Not needed for the new types
- LocalDate
@Temporal - LocalTime
- LocalDateTime
Qualifies the final database format of
date-time related attributes java.util.Date,
java.util.Calendar, etc.
Options: DATE, TIME, TIMESTAMP
Oct-23 [email protected] 26
Value Types
Support additional domain concepts
Its lifecycle depends on the owning entity
Composition semantics
As attribute of an entity:
All value-type’s attributes are embedded in
the table for the entity
As an aggregate (UML Composition):
It takes an extra table with a foreign key to
entity’s table primary key
Oct-23 [email protected] 27
Example
Oct-23 [email protected] 28
@Embeddable
Marks a class as ValueType
Its attributes can be also set with annotations:
@Basic, @Column, @Lob, @Temporal, @Enumerated
Oct-23 [email protected] 29
@EllementCollection
Marks a collection of ValueType
Oct-23 [email protected] 30
Peculiar case
If there are more than
one attribute on the entity
of the same Value type…
… we have to force the names of
the columns to avoid repeated names
on the DDL
Oct-23 [email protected] 31
@Lob, @Transient
Large Object Binary type
of column to store big
@Lob binary types. For example
images.
Marks an attribute as non
@Transient persistent.
By default all are
persistent
Oct-23 [email protected] 32
Oct-23 [email protected] 33
Identity vs equality
Java identity
Object equality
Database identity
Defined over the primary key of the table
They are mapped with @Id
All entity classes must have al least one @Identifier
The three identities there will not always be equal
The time in which they are is call “Guaranteed identity
context”, or “Persistence context”
Oct-23 [email protected] 34
Database reminder
Requirements of a key
Can never be NULL
Unique for each row
Can never change
Types of keys
Candidate
Natural What is the best as primary key?
Artificial (subrogate)
Oct-23 [email protected] 35
Candidate key
Field (o combinations) that determines uniquely a row
in the table There usually are several.
If none, your table model
is badly designed
Natural key
Candidate keys with significance for the user.
They uses it in their day to day operations (dni, ssn, plate
number, etc.)
Artificial (subrogate) key
Without significance in the domain, but for the system
Always system generated
Oct-23
[email protected] 36
Drawback of natural keys…
Experience shows that they cause long-term
problems if they are used as primary keys in
tables
Always not NULL? What if the user made a
Will never change? typo?
Will never repeat? As primary cannot be
changed …
Oct-23 [email protected] 37
Recommended strategy
Always use artificial keys as primary keys
Possible on a Green Field, but not in Legacy…
Data type for the key
Long forms efficient index
But, can also be String
Anyway, always define a unique
index over the natural key to
avoid repetitions
Oct-23 [email protected] 38
Who generates the value for an
artificial key?
Alternatives:
Provided by the user
… by the application
… by the mapper
… by the database
It is not easy to guarantee a never-repeating value under
any circumstance (rollback, concurrency, etc.)
Oct-23 [email protected] 39
@Id
Marks the attribute that forms the key
It may be a composed key:
multiple @Id and one @IdClass, or
one @EmbeddedId
Oct-23 [email protected] 40
Alternatives
Provided by the user Never!
… by the application Good, write the code…
… by the mapper
@GeneratedValue
… by the database
Oct-23 [email protected] 41
@GeneratedValue
Indicates the key is subrogated and assigned
either by the mapper or the database
Oct-23 [email protected] 42
Subrogated key, generated by the database, a Long
Subrogated key, generated by the mapper
Subrogated key, UUID generated by the application, String
Oct-23 [email protected] 43
@GeneratedValue vs UUID
@GeneratedValue
Advantages
Can be a Long, indexes well
Frees the programmer
Disadvantages
Unstable identity, it is not assigned until the database
FLUSH
hashCode() and equals() must be over the natural
identity (not the generated) natural identity without
setters
Oct-23 [email protected] 44
@GeneratedValue vs UUID
UUID
Advantages
Generated with the new entity, stable identity
hashCode() and equals() over a UUID, generalizable
Unique and universal code, allows external references
Unpredictable, improves security
Disadvantages
36 characters, more space
Indexes worse than a Long
Oct-23 [email protected] 45
@MappedSuperClass
Allows you to set mapping options common
to several Entities in a base class
Oct-23 [email protected] 46
@MappedSuperClass
hashCode()
and equals()
generalized
over UUID
Oct-23 [email protected] 47
Oct-23 [email protected] 48
UML implemented in Java
Essential to keep the references well crossed
Recommendation: Use an specific class
Oct-23 [email protected] 49
Multiplicity with JPA
one-to-one
many-to-many
one-to-many
They are directional
many-to-one
Oct-23 [email protected] 50
Unidirectional many to one
Oct-23 [email protected] 51
Unidirectional one to many
Oct-23 [email protected] 52
Bidirectional one to many
Oct-23 [email protected] 53
Pairing both association ends
With “mappedBy” you tell
the mapper what association
ends are paired
Oct-23 [email protected] 54
If you do not use
“mappedBy” the mapper will
interpret the associations
ends are independent (two
different associations)
Oct-23 [email protected] 55
One to one with a foreign key
“mappedBy” goes
on the class/table
that does not have
the FK
Oct-23 [email protected] 56
Bidirectional many to many
Oct-23 [email protected] 57
Oct-23 [email protected] 58
Strategies to map inheritance
Unique table for all the hierarchy
InheritanceType.SINGLE_TABLE
A table for each non-abstract entity
InheritanceType.TABLE_PER_CLASS
A table per each entity (abstract or not)
InheritanceType.JOINED
Oct-23 [email protected] 59
InheritanceType.SINGLE_TABLE
Table per class hierarchy
Oct-23 [email protected] 60
InheritanceType.SINGLE_TABLE
Table per class hierarchy
All classes are persisted on a unique table with the union
of all possible columns
Extra discriminator column to distinguish the type
Advantages
Simple and efficient
It supports polymorphic queries
Easy implementation
Easy to add/remove attributes on any class
Disadvantages
All non-common columns have to be nullable
Nullable columns might complicate queries and make it error
prone
There will always be empty columns (waste of space)
Might generate tables that waste a lot of space
Oct-23 [email protected] 61
InheritanceType.SINGLE_TABLE
Table per class hierarchy
This is the default strategy of the
mapper
Mapping
On the base class:
@Inheritance
@DiscriminatorColumn (optional)
On each derived class
@DiscriminatorValue (optional)
Recommendation
When derived classes add few properties and you
need efficient polymorphic queries
Oct-23 [email protected] 62
InheritanceType.SINGLE_TABLE
Table per class hierarchy
@DiscriminatorColumn,
@DiscriminatorValue are not
needed, default values are taken
by the mapper if not present
Oct-23 [email protected] 63
InheritanceType.TABLE_PER_CLASS
Table per concrete class
A table for each non-abstract class
Common properties are repeated on each table
Advantages:
Avoid nulls
Disadvantages:
Polymorphic associations (from the baseclass) repeat the
foreign key on each table
Polymorphic queries are less efficient: several SELECT with a
UNION
Changes on the base class are propagated to all the tables
Recommendation
When polymorphic queries are not needed or critical
Oct-23 [email protected] 64
InheritanceType.TABLE_PER_CLASS
Table per concrete class
A table for each non-
abstract class
abstracta
Oct-23 [email protected] 65
InheritanceType.TABLE_PER_CLASS
Table per concrete class
This strategy is optional under
the JPA spec, it might be some
provider does not support it
Oct-23 [email protected] 66
Table per class InheritanceType.JOINED
Each class in the hierarchy, abstract or not, has its own
table
Inheritance solved with foreign keys (joins)
No repetition of columns
Advantages
Relational model totally normalized
No nulls
Supports polymorphism
Changes of the superclass do not affect child tables
Disadvantages
Polymorphic queries needs joins (less performant than single
table)
Oct-23 [email protected] 67
InheritanceType.JOINED
Table per subclass
Oct-23 [email protected] 68
InheritanceType.JOINED
Table per subclass
Recommendation
In most situations, but specially if…
… child classes are very different in the
number of columns,
… cannot tolerate nullable columns and
… you need polymorphism
Oct-23 [email protected] 69
InheritanceType.JOINED
Table per subclass
Oct-23 [email protected] 70
Oct-23 [email protected] 71
Associative class mapping
In Java it is one more class,
but with composed identity
On the DB it is
a table with
extra columns
Oct-23 [email protected] 72
Associative class mapping
Alternatives
Artificial identity
Composed identity
Oct-23 [email protected] 73
With artificial identity
Composed unique index
on the table for extra
safety
Oct-23 [email protected] 74
Con identidad compuesta
Extra class for
composed key
metadata
oct.-23 [email protected] 75
Oct-23 [email protected] 76
Optimistic concurrency control
System transactions vs Application
transactions The optimistic
solution adds an
extra version field to
the entities
Lost update
Oct-23 [email protected] 77
Pessimistic vs Optimistic
ROLLBACK
Lock Lock Lock v1.0 v2.0 v1.0
ROLLBACK
It is no longer v1.0
Pessimistic Optimistic
Extra field for version on entities
Add information to the object to detect changes
between detached and persistent state
Long (counter)
Timestamp (of the last modification)
Extra field for versioning
The field for versioning is
maintained by the mapper:
Updated for each
modification of the row
Before row update it
checks the version on the
table and on the entity, if
they do not match throws
OptimisticLockException
(unchecked)
oct.-23 [email protected] 80
Oct-23 [email protected] 81
One to many with a Bag
If you do not need to keep insertion order
and duplicates are allowed
Use Collection<> instead of Set<>
More performant
There is no need too load the collection to
do a new insertion
Oct-23 [email protected] 82
One to many with a Bag
Oct-23 [email protected] 83
One to many with a List
When the insertion order matters
No mappedBy=“…”
Esto anula actualización de este extremo
Dos @JoinColumn
Oct-23 [email protected] 84
<…>ToMany @OrderBy
List keeps the order after
the sorting
But, on the database it
lost the order
Oct-23 [email protected] 85
Bidirectional many to many
with join table
@JoinTable opcional
Oct-23 [email protected] 86
Oct-23 [email protected] 87
@ElementCollection
Collections of Value Types
Sets, bags, lists, and maps of value
types
Collections of @Embeddable
Similar to OneToMany, but with
valueTypes, no Entities
Oct-23 [email protected] 88
Collections of Value Types,
limitations
They cannot be queried, persisted or updated
independently of its owning class
Strictly privately-owned (dependent) objects.
Semantic of composition, strong owning
There is no cascade option
ValueTypes are always persisted, updated or
removed along with its owning class
Oct-23 [email protected] 89
Map with a Set
IMAGES
The key of ITEM_IMAGE is
composed to avoid
duplicates of the same
ITEM (a set does not admit
duplicates)
Oct-23 [email protected] 90
Map with a List
Keeps the
order
The key is composed
of ITEM_ID +
POSITION to keep
order
Oct-23 [email protected] 91
Map with a Map
Keep the maps
keyset
The key is
PROPERTIES
PROPERTIES_KEY composed of
ITEM_ID +
PROPERTIES_KEY,
duplicates are not
allowed
Oct-23
[email protected] 92
Sorted & ordered collections
Sorted and ordered are different
Sorted, done in memory (JVM)
Comparable o Comparator
Ordered, done on the database
Sorted only for SortedMap and SortedSet
Oct-23 [email protected] 93
Sorted collections
Only for SortedSet and
SortedMap (done by
the JVM)
Oct-23 [email protected] 94
Ordered collections
For any collection (except
List). Done on the
database with order by
Oct-23 [email protected] 95
Oct-23 [email protected] 96
Cascading persistence (or transitive persistence)
In parent/child relationships
(aggregates)
Children depend on the parent
An special case of One to Many
Handle with care!
Or do not use it at all, you will
get the same effect with a few
lines of code…
Oct-23 [email protected] 97
Tipos de cascada JPA
ALL
MERGE
PERSIST
REFRESH
REMOVE
DETACH
Oct-23 [email protected] 98
Delete-orphan cascade
Oct-23 [email protected] 99
