Object-Relational Databases

Elisa Bertino CS Department and CERIAS Purdue University

RDBMS

They manage and manipulate simple data (tabular data) They are characterized by a query language (SQL) which is simple, declarative, and standard Tools are available supporting application development (Oracle Developer, Sybase Power Builder, Microsoft Visual Basic)

RDBMS: current situation

RDBMS: current situation

Portable on a large number of platforms Examples of RDBMS: IBM DB2, Oracle, Sybase, Informix, Microsoft SQL Server Very good performance Reliability, transaction management Based on client-server architecture efficiently support a large number of users Provide access control mechanism
3

Today the worldwide market for RDBMS is over 50 billions a year, but RDBMS have also some limitations

RDBMS: problems

RDBMS: problems

Mainly related to the intrinsic characteristics of the relational model:

SQL-92 provides only a limited set of data types Tables have a flat structure and do not provide good support for nested data structures, such as the ones obtained by using the set and array constructors It is not possible to specify sub-typing relations among the objects in a database

The relationships among entities are modeled by value (through the mechanism of foreign keys) and in the case of complex relationships several additional tables and columns may need to be created RDBMS do not exploit current object-oriented software design methodologies that are today widely used

OODBMS: overview

OODBMS: problems

They directly model complex objects and their relationships The object-orientated paradigm to software engineering and programming: uniqueness of the paradigm Good performance for navigational applications Limited support for concurrence, parallelism and distribution
7

Lack of a data model and query language standard and fully accepted Lack of a declarative query language (SQL-like) Simple transaction models Limited access control capabilities Cover a small application market, characterized by applications requiring efficient navigational accesses (like chip design)

RDBMS vs. OODBMS

The object-relational data model

RDBMS provide an efficient and effective support for applications manipulating simple data OODBMS provide an efficient support for some application classes that have to manage complex data, but without many of the positive aspects of RDBMS

The object-relational DBMS have been developed in order to provide all functions and features that RDBMS provide to the management of complex data, typical of OODBMS

ORDBMS: main features

ORDBMS: current situation

New data types

text, images, audio/video, GIS data, etc. user-defined data types collection types

Methods to model the operations on the user-defined types (ex. Java, C) New approaches to represent relationships The data management approach is however still the relational approach:

All data accesses are through SQL All entities of interest are modeled through tables

Today all main producers DBMS (Oracle, Informix, DB2,..) have extended their DBMS with object-relational features Such extensions require extending the SQL language Currently each RDBMS has a proprietary object-relational extension

10

11

ORDBMS: current situation

The SQL-99 standard

The extensions differ for:

The supported functions The approach used to implement such functions The extensions to SQL

And such differences exist even though there is SQL-99

As part SQL-99 an attempt has been made to standardize the object-relational extension of the relational model When SQL-99 was defined the major producers RDBMS already had their versions of objectrelational extensions SQL-99 does not standardize all object-relational features present in commercial DBMS

12

13

The SQL-99 standard

In what follows we will

It is thus a bit early for understanding when and to which extent the standard will be adopted by the commercial products It looks like that a further standard will be required to mediate among the various proprietary extensions

Discuss the general features of an ORDBMS Discuss how such features are defined by the standard

If not otherwise specified, we will use the SQL-99 syntax

Introduce the object-relational features of Oracle

14

15

Extensions to the type system

Type system of SQL92

In SQL-92 the types of relation attributes can be:

Numeric (integers, decimals, etc.) String (fixed length or variable length strings, single characters) Temporal (date, time, datetime, interval) Boolean (true, false) Non structured (BYTE, TEXT, BLOB, CLOB)

16

17

Type system of SQL92

Extension to the type system

For each built-in data type there is a fixed and predefined set of operations that can be defined on the values of this data type Such limitations often make difficult the representation of real-world information

Simple type Abstract data types

User-defined types

Reference types Complex types

record types and collection types

18

19

Simple types

Simple types

The simple types (also called distinct types) are the simplest form of extension to the set of types provided by an ORDBMS They allow the applications to create some new data types, based on a single type (builtin or user-defined) We refer to the type, on which a distinct type is defined, as base type
21

Comparisons with the base type or with other simple types defined on the same base type require the use of cast operations The ORDBMS automatically creates a cast function when a new simple type is created No inheritance or subtyping mechanisms are provided for the simple types The simple types are typically used to define data types that require operations that are different with respect to those available for the types on which the simple types are defined The simple types are considered by the DBMS as totally distinct from their base types The values of a simple type cannot be directly compared with the values of its base type (strong typing)

20

Example

Simple types in SQL-99

Suppose to create a new data type id_employee based on the integer type As the integer type, id_employee is used for recording numeric values but the DBMS will handle these two types as distinct For these two types different operations can be defined (for example the sum of two identifiers does not make sense, whereas a comparison operation could be useful)
23

SQL-99 allows the definition of simple types only on top of built-in types CREATE TYPE <name> AS <built-in type> FINAL

We will see in what follows the meaning of the FINAL clause

22

Example
CREATE TYPE id_employee AS INTEGER FINAL; CREATE TABLE Employees ( id id_employee, name VARCHAR(50), age INTEGER, id_manager id_employee);

Example - assignment
SELECT name FROM Employees WHERE id_manager = 123;

error

24

25

Example - comparison
CREATE TYPE Euro AS Decimal(8,2) FINAL; CREATE TYPE USA_Dollar AS Decimal(8,2) FINAL; CREATE TABLE European_Sales ( cust# INTEGER, order# INTEGER, total Euro); CREATE TABLE USA_Sales ( cust# INTEGER, order# INTEGER, total USA_Dollar);

Example - comparison
SELECT cust#,order# FROM European_Sales ERP, USA_Sales USA WHERE ERP.order# = USA.order# AND ERP.total > USA.total; error!!!

26

27

Casting in SQL-99

Cast functions in SQL-99

CREATE CAST (<source type> AS <target type>) WITH <function signature> [AS ASSIGNMENT]

The DBMS defines two cast functions for each new simple type:

One to translate from the distinct type to the built-in type One to translate from the built-in type to the distinct type

<source type>: input type <target type>: output type At least one between <source type> and <target type> must be a type defined by the application The other can be any type (built-in or simple type)

28

29

Cast functions in SQL-99

Cast functions in SQL-99

< function signature > is the signature of the function implementing the conversion The function must be defined as follows:
FUNCTION <name> (<source type>) RETURNS <target type> code ...

If the AS ASSIGNMENT clause is specified, the casting is implicitly executed when needed For each pair of types only a single casting function can be defined by the user The cast functions for the simple types are automatically created by the system with AS ASSIGNMENT clause

30

31

Casting in SQL-99

Example
SELECT name FROM Employees WHERE id_manager = CAST(123 AS id_employee); SELECT name FROM Employees WHERE id_manager = 123;

The cast function can be invoked:

explicitly CAST(<source type> as <target type>) implicitly, without invoking the CAST function

32

33

Example
SELECT cust#, order# FROM European_Sales ERP, USA_Sales USA WHERE ERP.order# = USA.order# AND CAST(ERP.total AS Decimal(8,2) > CAST(USA.total AS Decimal(8,2));

Example - alternative

To convert from Euro to USA_Dollar one can also define a new cast function

CREATE FUNCTION f(e Euro) RETURNS USA_Dollar BEGIN DECLARE g DECIMAL(8,2); SET g = e*0.9; RETURN g; END; CREATE CAST(Euro AS USA_Dollar) WITH FUNCTION f(Euro);

34

35

Is this correct?
SELECT cust#, order# FROM European_Sales ERP, USA_Sales USA WHERE ERP.order# = USA.order# AND ERP.total > USA.total; It is correct only if the cast function is defined with the AS ASSIGNMENT clause
36 37

ADT

An abstract data type (ADT) includes:

One or more attributes One or more methods

The attributes of an ADT can be specified very much like the table attributes

Default values can be specified It is not possible to specify the NOT NULL constraint We will see better in what follows

An ADT can be instantiable or not

ADT in SQL-99

Example

If there are only attributes (we will complete the definition later on): CREATE TYPE <type name> AS <attribute definition list> [{INSTANTIABLE|NOT INSTANTIABLE}] {FINAL|NOT FINAL} INSTANTIABLE is the default
39

Suppose to have to represent the addresses of the employees in a RDBMS There are two options:

address: VARCHAR(n) Representing each component of address as a separate attribute

38

Example
CREATE TYPE address_t AS number INTEGER, street VARCHAR(50), city CHAR(20), state CHAR(2), zip INTEGER NOT FINAL; address_t is a complex type, the attibutes of which have predefined types

ADT

The ADT can also be nested: CREATE TYPE employee_t AS emp# id_employee, name CHAR(20), resume TEXT, address address_t NOT FINAL;

40

41

ADT

ADT as column types

An ADT can be used as:

The ADT can be used as type of columns

Type of columns in tables (that is, as domains) Type of the tuples in one or more tables (row type)
43

CREATE TABLE Employees ( emp# id_employee, name CHAR(20), resume TEXT, address address_t);

42

ADT as column types

Table Employees
emp# name resume address number street city state zip

Methods

On ADT we can define (signatures of) methods as part of the type definition:
CREATE TYPE book_t AS title CHAR(20), sale_price DECIMAL(9,2), buying_price DECIMAL(9,2) NOT FINAL METHOD revenue() RETURNS DECIMAL(9,2);

44

45

Methods

Methods in SQL-99

The methods are functions defined by the users/applications and associated with the types They can be written in languages that are proprietary of the DBMS or in programming languages (ex. Java) The syntax changes a lot depending on the DBMS used The definition is similar to that of functions

They are created with the command CREATE METHOD

difference: the methods have an implicit parameter denoting the object on which the method is invoked

CREATE METHOD <method name> (parameter list) RETURNS <output data type> FOR <UDT name> <method body>

46

47

Example
CREATE METHOD revenue() RETURNS DECIMAL(9,2) FOR book_t RETURN (SELF.sale_price - SELF.buying_price);

Methods in SQL-99

The methods can be:

methods for instances (INSTANCE)

Can be invoked on an instance of a type Can be invoked on a type

Type methods (STATIC)

CREATE FUNCTION review(l book_t) RETURNS DECIMAL(9,2) RETURN (l. sale_price l.buying_price);

the default is INSTANCE

48

49

Example
CREATE TYPE book_t AS title CHAR(20), sale_price DECIMAL(9,2), buying_price DECIMAL(9,2) NOT FINAL INSTANCE METHOD revenue() RETURNS DECIMAL(9,2), STATIC METHOD max_sale_price() RETURNS DECIMAL(9,2);

Encapsulation

The ADT can be encapsulated In such case, they can only be manipulated through specific functions created by the DBMS upon their creation

50

51

Encapsulation in SQL-99

Constructor methods

Strict encapsulation Three predefined methods

Constructor methods: to create a new ADT instance Observer methods: to query ADT Mutator methods: to modify the values of ADT instances TYPE = CLASS

With each ADT a method (constructor) is associated with has the same name of the type The constructor creates an instance of the type The values to be assigned to the components of the instance to be created are passed as input to the constructor

52

53

Constructors in SQL-99

Example - insertion 1
INSERT INTO Employees VALUES(123,Smith,NULL,address_t(1710,First', Attica,IN', 47890));
Emp# 123 Name Smith Resume NULL Address number street city state zip 14 First Attica IN 47890

for any ADT T, a constructor T( ) exists

address_t() address_t

creates a new instance of the address_t type with the attributes initialized to the default values (such values can also be NULL) address_t(1715, Ashbury Court', Lafayette', 'IN', 47907) creates a new instance of the address_t type with the attributes initialized to the values received in input

54

55

Mutator methods

Example

They are used to modify the instances of an ADT

number(INTEGER) address_t street(VARCHAR(50)) address_t city(CHAR(20)) address_t state(CHAR(2)) address_t zip(INTEGER) address_t

Suppose we would like to make the following insertion in the Employees table:
emp# 123 name Smith resume NULL 14 address number street city state zip First Attica IN 47890

It holds also for SQL-99

57

Two steps: The tuple is created by initializing to null the address column Then the components of the address column are updated

56

Example - insertion 2
INSERT INTO Employees VALUES(123,Smith,NULL,address_t()); UPDATE Employees SET address = address.number(14) WHERE emp# = 123; UPDATE Employees SET address = address.street(First) WHERE emp# = 123; .

Example - insertion 3
BEGIN DECLARE i address_t; SET i = address_t(); SET i = i.number(14); SET i = i.street(First); SET i = i.city(Attica); SET i = i.state(IN); SET i = i.zip(47890); INSERT INTO Employees VALUES (123,Smith,NULL,i); END;

58

59

Observer methods

Examples of selection with observer methods

SELECT name FROM Employees WHERE address.city( ) = Milan OR address.city( ) = Rome; SELECT address.city() FROM Employees WHERE name = Smith;

For each component of an ADT the system automatically creates an observer method with the same name of the component:
number( ) ----> INTEGER street( ) ----> VARCHAR(50) city( ) ----> CHAR(20) state( ) ----> CHAR(2) zip( ) ----> INTEGER

Also in SQL-99

60

61

Selection

Deletion
DELETE FROM Employees WHERE address = address_t(14,First,Attica,IN,47890);

The selection of an ADT column returns an instance of the ADT SELECT address FROM Employees WHERE emp# = 123 The result is address_t(14,First,Attica,IN,47890)

62

63

Update
UPDATE Employees SET address = address.number(18) WHERE emp# = 123; UPDATE Employees SET address = address_t(22,University,West Lafayette,IN,47906);
64 65

Use of methods in queries

CREATE TYPE book_t AS title CHAR(20), sale_price DECIMAL(9,2), buying_price DECIMAL(9,2) NOT FINAL METHOD revenue() RETURNS DECIMAL(9,2); CREATE TABLE bookstore (codB# INTEGER, book book_t); SELECT b.book.revenue( ) FROM bookstore b WHERE b.book.title() = Close Range Wyoming Stories;

Integrity constraints

Operations

It is not possible to define the PRIMARY KEY, UNIQUE, FOREIGN KEY constraints on a column that has as domain an ADT Motivation

Casting defined by the applications for converting from the ADT to other types Possibility of defining ordering and comparison functions

Conceptually everything is OK There are efficiency problems

We do not see them

66

67

Type deletion and update

DROP TYPE <type_name> {CASCADE|RESTRICT}; ALTER TYPE <type_name> <modification_operation>; < modification_operation >::= ADD ATTRIBUTE <attribute_name>| DROP ATTRIBUTE < attribute_name >

Row types

An ADT can also be used to specify the type of the tuples of a table (row type) The tuples of the table are thus instances of the row type, whereas the columns of the table are the same as the attributes of the row type

68

69

Row types

Typed tables in SQL-99

CREATE TABLE <table_name> OF <complex_type_name> [(REF IS <TID_column_name>)]

The row types support:

The definition of a set of tables that share the same structure (typed tables) A direct and simple modeling mechanism for relationships among different tables (referenceable tables) The definition of hierarchies among tables

TUPLE OF A TYPED TABLE = OBJECT each tuple has an identifier representing an additional column for each table; the values of such column are unique in the entire database by default, the identifiers are generated by the system

The REF IS clause denotes the name of an attribute (distinct with respect to the other attribute names) in which the TID (TID - tuple identifier) will be inserted The TID column is always the first column in the table schema If such clause is missing, the column recording the TIDs exists; it is generated by the system but it transparent to the user (that is, the user cannot select it)

70

There are other possibilities, we do not see them

71

Example

Example
CREATE TABLE Projects OF project_t (REF IS my_TID);
Projects
my_TID proj# name 16454 12 description budget 10,000,000

Suppose that we wish to represent information about the projects on which employees work
CREATE TYPE project_t AS proj# INTEGER, name VARCHAR(20), description VARCHAR(50), budget INTEGER NOT FINAL;

Oracle ORDBMS

72

73

Typed tables in SQL-99

Example
CREATE TYPE project_t AS proj# INTEGER, name VARCHAR(20), description VARCHAR(50), budget INTEGER NOT FINAL; CREATE TABLE projects OF project_t (PRIMARY KEY (proj#)); CREATE TABLE projects OF project_t (proj# INTEGER PRIMARY KEY); OK NO

When defining a typed table it is possible to add integrity constraints on the attributes of the ADT on which the table is defined
CREATE TABLE <table_name> OF <complex_type_name> [(REF IS <TID_column_name>, <constraints>)]

74

75

Row type

Selection
SELECT proj# FROM Projects WHERE budget > 1,000,000; SELECT my_TID FROM Projects WHERE budget > 1,000,000;
77

The row types are not encapsulated There is encapsulation only when an ADT is used domain of a column The attributes of the row type are seen as columns of the table (included the TID column, that can be selected) The queries are executed according to the usual approach

76

Insertion
INSERT INTO Projects(Proj#,Name,Description,Budget) VALUES(14,DB development,DB development in Oracle,20,000,000); no value needs to be specified for the identifier column

Reference types

The row types can be combined with reference types (REF type) The REF types directly support the representation of relationships among type instances Such a type allows a column in a tuple in a given table to reference a tuple in another table Each tuple in a table is identified through its TID

78

79

Example

Example

Suppose we wish to record information about which employee works in which project In a RDBMS one would have two tables Employees and Projects The table Employees would include a column indicating the project where each employee works (foreign key)

Employees
emp#
123 proj#

Projects
prj# 12

name
Oracle .

...

12

80

81

Reference types

Reference types in SQL-99

REF (<ADT_type>) [SCOPE <table_name> [<reference_scope_check>]]

In an ORDBMS there are two additional options:

define an ADT project and use this ADT as domain of a column of the Employees relation (however we have data redundancy because the same project may be stored several times) define a table based on a new complex type and refer the tuples of this table from the table Employees

Reference type

82

83

The SCOPE specifies a typed table defined on the type <ADT_type>; it specifies that the admitted values for the reference type are references (that is, pointers) to the tuples stored in table <table_name> If the SCOPE clause is not specified, the scope is implicit and is represented by all references to tuples instances of the type <ADT_type> (independently from the tables where these tuples are stored)

Reference types in SQL-99

Reference types in SQL-99

<reference_scope_check> = REFERENCES ARE [NOT] CHECKED [ON DELETE {CASCADE | SET NULL | SET DEFAULT | RESTRICT | NO ACTION}] The meaning is the same as of the relational model Because the TID is considered immutable, there is no ON UPDATE clause default: RESTRICT

The SCOPE clause represents is similar to the foreign key constraint in the relational model Also in this context there is the referential integrity problem <reference_scope_check> is a clause specifying how integrity has to be maintained, like to what we have seen for the foreign keys

84

85

Example
CREATE TABLE Employees( emp# id_employee, name VARCHAR(50), address address_t, assignment REF(project_t) SCOPE Projects REFERENCES ARE CHECKED ON DELETE CASCADE); We associate an employee with a project The same project can be associated with several employees If a project is deleted, all employees assigned to this project are also deleted

Example
CREATE TABLE Employees( emp# id_employee, name VARCHAR(50), address address_t, assignment REF(project_t) SCOPE Projects REFERENCES ARE CHECKED ON DELETE RESTRICT); A project can be deleted if there are no employees assigned to it

86

87

Example
CREATE TYPE employee_t AS emp# id_employee, name CHAR(20), resume TEXT, address t_address, department REF(department_t) NOT FINAL; CREATE TABLE Employees OF employee_t (REF IS my_tid);

Example
CREATE TYPE department_t AS dept# INTEGER, name VARCHAR(30), manager REF (employee_t) NOT FINAL ;
CREATE TABLE Departments OF department_t (REF IS my_tid);

88

89

Example
Impiegati emp# name ... department

Example

dept# name

manager

The department column for a tuple of Employees references a tuple of the Departments table (that is, the tuple corresponding to the department where the employee works) The manager column for a tuple Departments references a tuple of the Employees table (that is, the tuple corresponding to the employee which is the manager of the department)

Departments

90

91

Reference types in SQL-99

Example
CREATE TABLE Projects OF project_t (proj_ref REF(project_t));
Projects
proj_ref proj# name 12 description budget 10,000,000

Possibility of extending the definition of a typed table with further attributes of REF type

Oracle ORDBMS

92

93

Reference types in SQL-99 manipulation

Reference types in SQL-99 manipulation

The values of reference types can only be compared using the = and <> comparison operators = denotes the identity equality The casting can be defined between a reference type and the target ADT or built-in type

Dereferentiation function DEREF:

It receives in input an expression denoting a value (pointer) for a reference type with a non empty scope It returns the value pointed by the reference (that is, the referenced tuple) It receives in input an expression that denotes a reference and attribute of the ADT to which the reference points It returns the value for that attribute for the referenced tuple

REF function (denoted by )

94

95

Example
SELECT manager FROM Departments WHERE name = DVD; it returns a pointer to an employee (that is, the oid of the employee who is the manager of the DVD department)

Example
SELECT deref(manager) FROM Departments WHERE name = DVD; it returns information on the manager of the DVD department (that is, an entire row of the Employees table)

96

97

Example
SELECT deref(manager).name FROM Departiments WHERE name = DVD; It returns the name of the manager of the DVD department

Example
SELECT manager -> name FROM Department WHERE name = DVD; It returns the name of the manager of the DVD department It is equivalent to the previous query

98

99

Referential integrity

Example
CREATE TABLE Employees( emp# id_employee, name VARCHAR(50), address address_t, assignment REF(project_t) SCOPE Projects REFERENCES ARE CHECKED ON DELETE RESTRICT); CREATE TABLE Projects OF project_t (REF IS My_TID, prog_ref REF(project_t));

The identifiers are generated and assigned by the system The user does not know them beforehand Problem:

How to assure referential integrity for a table that includes a reference type? A subquery is used to determine the identifiers to assign to the new tuples

Solution:

100

101

Referential integrity

Example
INSERT INTO Employees VALUES (2,James, address_t( ),NULL); UPDATE Employees SET assignment = (SELECT my_tid FROM Projects WHERE name = Oracle) WHERE emp# = 2;

When we insert a tuple in the Employees table, we need to specify a project identifier for the attribute assignment; such identifier must reference a tuple of the Projects table Two steps:

The tuple is inserted by assigning a NULL value to the column of reference type Such column is then modified by the UPDATE command

102

103

Reference types in SQL-99 restrictions

Collection types and row types

PRIMARY KEY, UNIQUE, FOREIGN KEY cannot be specified

104

105

Collection types

Collection types in SQL-99

The collection types can be seen as containers for objects with a similar structure There is no set of collection types that is supported by all ORDBMS

The only collection type included in SQL-99 is ARRAY

<attribute_name> <type> ARRAY[<dimension>]

set bag list array

<dimension> is an integer value Constructor:

ARRAY[<value_1>,,<value_n>] <attribute_name>[i] where i is an integer value between 1 and n

access:

106

107

Collection types in SQL-99

Example
CREATE TABLE Employees( emp# id_employee, name VARCHAR(50), skills VARCHAR(20) ARRAY[3]);

The number of elements in an array can be any number between 0 (ARRAY[ ]) and the maximum number of elements specified in the array declaration implicitly, there exists a length parameter directly managed by the system

108

109

Example
INSERT INTO Employees VALUES (2,James,ARRAY[Oracle,Unix,Java]); SELECT * FROM Employees WHERE skills[2] = Unix;

Example
CREATE TYPE employee_t AS emp# id_employee, name VARCHAR(30), address address_t, skills VARCHAR(20) ARRAY[3], manager REF(employee_t), projects REF(project_t) ARRAY[10], children REF(person_t) ARRAY[10], hobbies VARCHAR(20) ARRAY[5] NOT FINAL;

110

111

CREATE TABLE Employees OF employee_t;

Example
UPDATE Employees SET skills = ARRAY[Oracle,Unix]; UPDATE Employees SET skills = ARRAY[SQL Server]; The new array contains a single element (the length changes)
112 113

Collection type in SQL-99 manipulation

Casting

cast on the type of the array elements and reduction in the number of elements usual If the value is truncated an error is generated =, <>

assignment:

comparison:

Collection type in SQL-99 manipulation

Collection type in SQL-99 restrictions

functions

concatenation

CONCATENATE (<array_expression> WITH <array_expression> CARDINALITY(<array_expression>)

For the fields defined as array the UNIQUE, PRIMARY KEY, FOREIGN KEY constraints cannot be specified

cardinality

114

115

Tuple types in SQL-99

Row types in SQL-99

SQL-99 provides the possibility of a nested definition for the row types (we refer to these types of values as tuple types) They do not require the definition of an ADT but their definition can be directly embedded in the definition of a table or another type

Type
ROW (<def component_1>,,<def component_n>)

example: ROW(number
street city state zip

116

117

INTEGER, VARCHAR(50), CHAR(20), CHAR(2), INTEGER)

Example
CREATE TABLE Employees ( emp# id_employee, name CHAR(20), resume TEXT, address ROW( number
street city state zip

Tuple types in SQL-99

INTEGER, VARCHAR(50), CHAR(20), CHAR(2), INTEGER)

Values:

ROW(<value typo_1,,value type_n>) ROW(3,University,West Lafayette,IN,47906)

Example:

);

Also the tuples returned by a query are seen as value of the tuple type The components of a tuple can be accessed by using the dot notation

118

119

Example
INSERT INTO Employees VALUES (3,James,NULL, ROW(3,University,West Lafayette,IN,47906))

Example
CREATE TABLE Addresses ( Iid INTEGER, street VARCHAR(20), city VARCHAR(20), state VARCHAR(20), zip INTEGER); UPDATE Employees SET Address = (SELECT t from Adresses t WHERE Iid = 3) WHERE name = James;

120

121

Example
SELECT Name FROM Employees WHERE address.city = West Lafayette;

Tuple types in SQL-99 manipulation

Assignment: The same number of components The types must be compatible comparison: =, <>, <, <=, >, >= Lexicographic ordering, based on the types of components The values must have the same number of components The presence of NULL can generate UNKNOWN

122

123

Example
ROW(1,1,1) = ROW(1,1,1) ROW(1,1,1) = ROW(1,2,1) ROW(1,NULL,1) = ROW(2,2,1) ROW(1,NULL,1) = ROW(1,2,1) ROW(1,1,1) <> ROW(1,2,1) ROW(2,NULL,2) <> ROW(2,2,1) ROW(2,2,1) <> ROW(2,2,1) ROW(1,NULL,1) <> ROW(1,2,1) ROW(1,1,1) < ROW(1,2,0) ROW(1,NULL,1) < ROW(2,NULL,0) ROW(1,1,1) < ROW(1,1,1) ROW(3,NULL,1) < ROW(2,NULL,0) ROW(1,NULL,1) < ROW(1,2,0) ROW(NULL,1,1) < ROW(2,1,0) TRUE FALSE FALSE UNKNOWN TRUE TRUE FALSE UNKNOWN TRUE TRUE FALSE FALSE UNKNOWN UNKNOWN

Tuple types in SQL-99

The tuple types cannot have PRIMARY KEY, UNIQUE, FOREIGN KEY constraints

124

125

Inheritance

Inheritance

The inheritance mechanism allows one to define supertype/subtype relationships The inheritance mechanism supports the specialization of existing types based on the application requirements

126

127

Inheritance

Inheritance

A subtype inherits all attributes, methods, and constraints defined for its supertypes A subtype can extend the supertype by adding new attributes and methods A subtype can also re-define the inherited methods and attributes

We can distinguish two forms of inheritance

Type inheritance Table inheritance

128

129

Type inheritance
Consider the following entities: Truck: model plate# last_revision weight CHAR(20), INTEGER, DATE, INTEGER, Bus: model plate# CHAR(20), INTEGER,

Type inheritance

last_revision DATE, seats# INTEGER,

next_revision() DATE

next_revision() DATE

In the relational model to represent those entities we need two table and two procedures In a ORDBMS, truck and bus can be considered specializations of a common type of entity: Vehicle One can thus define a type vehicle including all features that are common to the types truck and bus Truck and bus are defined as subtypes of vehicle; each one adds some special features

130

131

Type inheritance in SQL-99

Type inheritance in SQL-99

For ADT Single inheritance

CREATE TYPE <type_name> UNDER <supertype_name> AS additional attributes [NOT] FINAL;

FINAL clause:

If a type has been defined as FINAL, no subtypes can be generated from it Subtypes can be defined

NOT FINAL clause:

The supertype must have been declared with the NOT FINAL clause
133

The NOT FINAL clause is mandatory if no supertype is specified in the definition Otherwise one can choose

132

Example
CREATE TYPE vehicle AS model CHAR(20), plate# INTEGER, last_revision DATE, METHOD next_revision( ) RETURNS DATE NOT FINAL;

Example
CREATE TYPE truck_t UNDER vehicle_t AS weight INTEGER NOT FINAL;

CREATE TYPE bus_t UNDER vehicle_t AS seats# INTEGER NOT FINAL;

134

135

Methods & inheritance

CREATE TYPE person_t AS name CHAR(20), id INTEGER, birth_date DATE, address address_t, METHOD age() RETURNS INTEGER NOT FINAL; CREATE TYPE teacher_t UNDER person_t AS salary DECIMAL(9,2), hiring_date DATE, course course_t NOT FINAL;

Methods & inheritance

The methods are inherited by the subtypes like the attributes:

CREATE TABLE Teachers OF teacher_t; SELECT name, T.age( ) FROM Teachers T WHERE salary > 3000;

136

137

Methods & inheritance

Non instantiable types

It is possible to override an inherited method It is not possible to redefine the attributes Consider a type teacher_t; suppose that with this type a method is associated called age that determines the years of service as teacher (overriding)
CREATE TYPE teacher_t UNDER person_t AS salary DECIMAL(9,2), hiringdate DATE, course course_t OVERRIDING METHOD age() RETURNS INTEGER NOT FINAL;

The declaration of a type specifies whether the type can be instantiated (as thus have its own instances) or not
CREATE TYPE <type_name> AS <attribute_definition_list> [{INSTANTIABLE|NOT INSTANTIABLE}] {FINAL|NOT FINAL}

The default is INSTANTIABLE A type that cannot be instantiated corresponds to an abstract class: it only useful for code reuse

138

139

Substitutability

Table inheritance

In the OODBMS the substitutability principle holds An istance of a type can be used everywhere we can use an instance of an instance of one of its supertypes Such principle does not hold for current ORDBMS To support substitutability :

The typed tables can be organized according to inheritance hierarchies However this is possible only if the types on which the tables are defined are related by the inheritance relation The inheritance among tables is useful in order to extend SQL operations (such as queries) to the instances of a table and to the instances of all the sub-tables of this table

140

CAST functions must be used

141

Example
CREATE TABLE People OF person_t; CREATE TABLE Teachers of teacher_t UNDER people;

Queries

A hierarchy has been created between the People and the Teachers tables

The inheritance relationships holding among tables influences query results A query issued on a table automatically propagates to the sub-tables The same applies to the delete and update operations, whereas an insert operation only applies to a single table If the operation has to be restricted to the instances of a given table, the ONLY option has to be specified

142

143

Example
name id birth_date
16/8/68 3/2/48

Example
address

Smith 74 John 86

People

SELECT name FROM People WHERE birth_date > 1/1/1967;

The result is: Smith and Allen

SELECT name FROM ONLY People WHERE birth_date > 1/1/1967;

name
Allen Mark

id
82 81

birth_date
9/7/67 3/5/58

address

salary .
30K 60K

Teachers

The result is: Smith

144

145

Example
DELETE FROM People WHERE id > 80;

Object-relational Features of Oracle 9i

It deletes John from the People table and Allen and Mark from the Teachers table

146

147

The Oracle type system

Object types

No distinct type Object types (corresponding to ADT) Reference types Collection types Inheritance

Definition of ADT:
Attribute declaration Method specification

specification

body

Code of methods

148

149

Example
CREATE TYPE Complex AS OBJECT( r_part FLOAT, i_part FLOAT, MEMBER FUNCTION sum(x Complex) RETURNS Complex, MEMBER FUNCTION minus(x Complex) RETURNS Complex, MEMBER FUNCTION product(x Complex) RETURNS Complex, MEMBER FUNCTION division(x Complex) RETURNS Complex);

Example
CREATE TYPE BODY Complex AS MEMBER FUNCTION sum(x Complex) RETURN Complex IS BEGIN RETURNS Complex(r_part + x. r_part, i_part + x. i_part ); END sum; MEMBER FUNCTION difference(x Complex) RETURN Complex IS BEGIN RETURNS Complex(r_part - x. r_part, i_part - x. i_part); END difference; . END;

150

151

Object types

Methods

Everything we have said for SQL-99 applies here with the following differences:

They can be functions or procedures Two types:

The notion of body There is no encapsulation: the attributes can be directly accessed through the dot notation

MEMBER

Defined for instances Implicit parameter: SELF Defined for the type

STATIC

152

153

Example
CREATE TYPE Rational AS OBJECT (num INTEGER, den INTEGER, MEMBER PROCEDURE normalize, ... ); CREATE TYPE BODY Rational AS MEMBER PROCEDURE normalize IS g INTEGER; BEGIN g := gcd(SELF.num, SELF.den); g := gcd(num, den); -- equivalent to previous line num := num / g; den := den / g; END normalize; ... END;

Special methods

Constructors

as in SQL-99

MAP methods ORDER methods

154

155

MAP methods

Example
CREATE TYPE Rectangle_typ AS OBJECT ( length NUMBER, width NUMBER, MAP MEMBER FUNCTION area RETURN NUMBER, ... ); CREATE TYPE BODY Rectangle_typ AS MAP MEMBER FUNCTION area RETURN NUMBER IS BEGIN RETURN length * width; END area; ... END;

They allow one to compare ADT instances by mapping them to values of built-in types (DATE, NUMBER, VARCHAR) They thus represent a casting between an ADT and one of the built-in types If for a given ADT, a MAP method exists, the comparisons among instances of this ADT are executed by first converting the instances into the values of the considered built-in type

156

157

Example

ORDER methods

Let o1 and o2 be instances of rectangle_typ type:

o1 < o2 is equivalent to o1.area() < o2.area() The comparison relation is established on two instances of the NUMBER type

These methods implement an order relation on the set of instances of a given type They have always a parameter with the same type of the type for which the method is defined They are useful for comparing complex data types that cannot be easily compared with a MAP method If an ORDER method exists, the method is automatically invoked when the instances of the considered type are compared

158

159

ORDER methods

Example
CREATE TYPE Customer_t AS OBJECT ( id NUMBER, name VARCHAR2(20), address VARCHAR2(30), ORDER MEMBER FUNCTION match (c Customer_t) RETURN INTEGER );

The output of such a method is always an integer which can have one of the following values:

- 1 : SELF < parameter 0 : SELF = parameter +1 : SELF > parameter

For an ADT, at most one MAP or ORDER method can be defined If none of the two is defined, the system supports only = and <>

160

161

Example
CREATE TYPE BODY Customer_t AS ORDER MEMBER FUNCTION match (c Customer_t) RETURN INTEGER IS BEGIN IF id < c.id THEN RETURN -1; ELSIF id > c.id THEN Customers are compared with RETURN 1; respect to their identifiers ELSE RETURN 0; END IF; END; END;

Typed tables

Also in Oracle un ADT type can be used according to two different modes:

as type (i.e. domain) for a table attribute as base type for the definition of a typed table

It is not possible to specify a column for the identifier (no REF IS clause)

The Identifier is not managed as an attribute but as a function

162

163

Object types: access

Example
CREATE TYPE Person_t AS OBJECT (ssn# VARCHAR(20)); CREATE TABLE ptab1 OF Person_t; CREATE TABLE ptab2 (c1 Person_t); SELECT ssn# FROM ptab1 ; SELECT c1.ssn# FROM ptab2 ; SELECT ptab2.c1.ssn# FROM ptab2 ; SELECT p.c1.ssn# FROM ptab2 p ; OK Error Error OK

Access through notation to attributes and methods If the dot notation is used, it is always necessary to use a relation alias for the accessed table

164

165

Syntax of commands
CREATE TYPE typename AS OBJECT (attrname datatype {, attrname datatype}); CREATE OR REPLACE TYPE BODY typename AS method {method}; CREATE TABLE tablename OF typename ([attrname NOT NULL] {,attrname NOT NULL} [,PRIMARY KEY (attrname {,attrname })]); DROP TYPE typename; DROP TABLE tablename; ALTER TYPE typename REPLACE AS OBJECT (new definition of a type) CREATE OR REPLACE TYPE BODY typename IS method {method};

Reference types

It applies what we have seen for SQL-99

The syntax is slightly different REF <type_name> SCOPE IS <table_name>

No possibility of specifiying referential integrity actions Predicate IS DANGLING to establish if a value of type REF references a tuple that does not any longer exists

166

167

Reference types - manipulation

Reference types
CREATE TYPE person_t AS OBJECT( fname VARCHAR2(10), lnames VARCHAR2(15), birth_date DATE, address address_t, mother REF person_t, father REF person_t); CREATE TABLE People OF person_t;

Three main functions:

referencing ref( ): given an object of a given type, it returns the identifier of this object dereferencing deref( ): given an identifier, it returns the object value ( ): given a relation alias, it returns the associated tuple object (by using the proper constructor)

168

169

Reference types
Persone
fname lname

Reference types
father

...

mother

INSERT INTO People VALUES(Thomas,Smith, ,NULL,NULL); INSERT INTO People VALUES(person_t(Joanna,White,,NULL,NULL)); INSERT INTO People VALUES(Robert,Smith,,NULL,NULL);

170

171

Reference types
People
fname Thomas Joanna Robert lname Smith White Smith

Reference types
UPDATE People p SET p.mother = (SELECT ref(d1) FROM People d1 WHERE fname = Joanna and lname = White) WHERE fname = Robert and lname = Smith;

...

mother NULL NULL NULL

father NULL NULL NULL

The mother of Robert Smith is Joanna White

172

173

Reference types
People
fname Thomas Joanna Robert lname Smith White Smith

Reference types
SELECT value(p) FROM People p

...

mother NULL NULL

father NULL NULL NULL

We obtain: person_t(Thomas,Smith,NULL,NULL) person_t(Joanna,White,NULL,NULL) person_t(Robert,Smith,xxxyyywww,NULL) where xxxyyywww is the identifier of the tuple of Joanna White

174

175

Reference types
SELECT deref(p.mother) FROM People p WHERE fname = Robert and lname=Smith; It returns all information contained in the People table related to the mother of Robert Smith It returns

Reference types
DELETE FROM People WHERE fname = Joanna AND lname = Smith; After such operation the tuple corresponding to Robert Smith has a dangling pointer; there is a special predicate to retrieve the tuples with dangling pointers SELECT fname, lname FROM People WHERE mother IS DANGLING; It returns Robert Smith

176

person_t(Joanna,White,NULL,NULL)

177

Collection types

Collection types

Oracle provides two collection types:

nested table varray

The collection types may have elements instances of object types Also an object type can have an attribute the type of which is a collection type
179

178

The nested tables can be considered as a table with a single column Main differences between arrays and nested tables: The arrays have a fixed dimension whereas the nested tables have a variable dimension The arrays are stored inside the table in which are used or as BLOB (the storage strategy is decided by the system) The nested tables are stored in separated tables, with an additional column that identifies the tuple of the table to which they belong

Collection types - varray

Collection types nested tables

Data1 Data2 Data3 Data4 VARRAY_Data[5] . . . . . . . . . . [A11,A12,A13] [B21,B22,B23,B24,B25] [C31,C32] [D41] [E51,E52,E53,E54]

180

181

Collection types

Varrays - creation
CREATE TYPE project_t AS OBJECT( id INTEGER, title VARCHAR2(25), cost NUMBER(7,2)); CREATE TYPE project_list AS VARRAY(50) OF project_t; CREATE TYPE department_t AS OBJECT( dept# INTEGER, name VARCHAR2(15), budget NUMBER(11,2), projects project_list);

They cannot be directly used in the definition of an attribute It is always necessary to assign a name to the collection type before using it For each collection type, a constructor exists To create an istance, it is necessary to pass to the constructor a set of elements of the type on which the collection type is based
183

182

CREATE TABLE Departments OF department_t;

Varray - insertion
INSERT INTO Departments VALUES(30,R&D,1000000000, project_list(project_t(1,DBMS,10000000), project_t (3,C++,20000000))); INSERT INTO Departments VALUES(32,Marketing,1000000000, project_list (project_t(1,New Ads,10000000), project_t(3,Personnel Bonus,20000000)));

Nested tables - creation

CREATE TYPE course_t AS OBJECT( id NUMBER(4), name VARCHAR2(25), credits NUMBER(1)); CREATE TYPE course_list AS TABLE course_t;

184

185

Nested tables - creation

A) CREATE TABLE University_departments ( name VARCHAR2(20), chair VARCHAR2(20), courses course_list) NESTED TABLE courses STORE AS course_tab; B) CREATE TYPE university_department_t AS OBJECT( name VARCHAR2(20), chair VARCHAR2(20), course course_list); CREATE TABLE University_departments OF university_department_t; NESTED TABLE courses STORE AS course_tab;

Nested Table - creation

Notice the NESTED TABLE clause in the previous slide such clause is necessary because the columns defined as nested table are stored as separated tables The general format of the clause is NESTED TABLE colname STORE AS tablename The table tablename is called child-table of the table within which is defined (called parent-table) A child-table can only be accessed through the parent-table The names of the child-tables must be unique in the schema During the manipulation and query operations, the system executes a join between the parent and child tables

186

187

The nested tables are less efficient than the varrays but more flexible

Nested tables: insertion

INSERT INTO University_departments VALUES(Computer Science,Brown, course_list(course_t(1000,Programming I,2), course_t(1001,Mathematical Logics,1), course_t (1002,Database Systems,2), course_t(1003,Computer Graphics,1)));

Collection types - queries

Two possibilities

Select the nested collection

a tuple for each collection a tuple for each element of the collection TABLE function

Select the unnested collection

They apply to both VARRAY and NESTED TABLE

188

189

Nested selection - example

SELECT courses FROM University_departments; The result is
course_list(course_t(1000,Programming I,2), course_t(1001, Mathematical Logics,1), course_t(1002,Database Systems,2), course_t(1003, Computer Graphics,1))

Nested selection - example

SELECT t.* FROM University_departments d, TABLE(d.courses) t; The results is given by a table containing tuples of type course_t
1000 1001 1002 1003 Programming I Mathematical Logics Database Systems Computer Graphics 2 1 2 1

190

191

TABLE function

Example
SELECT d.name, t.* FROM University_departments d, TABLE(d.courses) t; The result is
Computer Science Computer Science Computer Science Computer Science 1000 1001 1002 1003 Programming I Mathematical Logics Database Systems Computer Graphics 2 1 2 1

The TABLE function receives a value of type collection and allows one to use it as a table The previous query performs a join for each tuple of the University_departments table with each element of the collection object The function can also be applied to queries returning a single value of type collection

192

193

Example
SELECT * FROM TABLE(SELECT courses FROM University_departments WHERE name = Psychology); returns all information about the courses of the Psychology department SELECT credits FROM TABLE(SELECT courses FROM University_departments WHERE name = Psychology); returns the credits of all the courses of the Psychology department The queries are correct if the query on which the TABLE function is applied returns a single element

Collection types - DML

Nested tables:

The TABLE function can be used to modify a collection Unlike the nested tables, the elements of a varray cannot be manipulated through SQL instructions To select or update a given element of a VARRAY, PL/SQL must be used There is a large number of functions for manipulating arrays

VARRAYs:

194

195

Nested tables
UPDATE TABLE(SELECT courses FROM University_departments WHERE name = Psychology) SET credits = credits + 1 WHERE id IN (2200,3540); DELETE FROM TABLE(SELECT courses FROM University_departments WHERE name = English) p WHERE p.credits = 2;

Varray
DECLARE my_projects project_list; SELECT projects INTO my_projects. FROM Departments. WHERE id = 30;

IF my_projects(i).Title = DBMS

196

197

Inheritance

Substitutability

Only for types What we said for SQL-99 applies In addition:

Both types and methods can be defined as FINAL/NOT FINAL

It refers to the possibility of using an instance of a type in each context in which an instance of a supertype of this type can be used Substitutability at

FINAL: overriding is not possible

Attribute level Tuple level

Substitutability Late binding

198

199

Substitutability at attribute level

Substitutability at tuple level

An attribute of type REF(<type1>) can be assigned values of type REF(<type2>) if <type2> is a subtype of <type1> An attribute of type ADT <type1> can be assigned values of type ADT <type2> if <type2> is a subtype of <type1> An attribute of type collection defined on type <type1> can include elements of type <type2> if <type2> is a subtype of <type1>
201

A typed table defined on the type < type 1> can contain instances of the type < type 2> if <type2> is a subtype of <type1>

200

Example
CREATE TYPE Person_t AS OBJECT (ssn NUMBER, name VARCHAR2(30), address VARCHAR2(100)) NOT FINAL; CREATE TYPE Student_t UNDER Person_t (deptid NUMBER, major VARCHAR2(30)) NOT FINAL; CREATE TYPE PartTimeStudent_t UNDER Student_t (numhours NUMBER);

Example (on attributes)

CREATE TABLE Departments ( Id INTEGER, name VARCHAR(20), manager Person_t); INSERT INTO Departments VALUES (1,research, Person_t(1243, 'Bob', '121 Front St')); INSERT INTO Departments VALUES (2,development,Student_t(3456, 'Joe', '34 View', 12, 'HISTORY')); INSERT INTO Departments VALUES (3, testing,PartTimeStudent_t(5678, 'Tim', 14 Back St, 13, 'PHYSICS', 20));

202

203

Example (on tuples)

CREATE TABLE People OF Person_t; INSERT INTO People VALUES (Person_t(1243, 'Bob', '121 Front St')); INSERT INTO People VALUES (Student_t(3456, 'Joe', '34 View', 12, 'HISTORY')); INSERT INTO People VALUES (PartTimeStudent_t(5678, 'Tim', 14 Back St, 13, 'PHYSICS', 20));

Limiting substitutability
CREATE TABLE Departments ( Id INTEGER, nome VARCHAR(20), manager Person_t) COLUMN manager NOT SUBSTITUTABLE AT ALL LEVELS; Manager can only be an object of type Person_t (instances of subtypes not admitted) CREATE TABLE People OF Person_t NOT SUBSTITUTABLE AT ALL LEVELS; The table People can only include instances of the type Person_t (and not of its subtypes)

204

205

Limiting substitutability
CREATE TABLE Departmens ( Id INTEGER, name VARCHAR(20), manager Person_t) COLUMN manager IS OF (ONLY Student_t);

Queries

Useful functions:

Only one subtype can be specified

REF DEREF VALUE Expr IS OF type: it determines if the type of Expr is type TREAT(Expr AS type): it modifies the type of Expr into type; type must be subtype of the type of Expr ...

REF, DEREF, VALUE already seen

206

207

Example IS OF

Example IS OF

Consider the University_department table (without substitutability limitation)

Consider the People table (without the substitutability limitations)

SELECT * FROM University_deartment p WHERE p.manager IS OF (Student_t,PartTimeStudent_t) Returns the departments having as manager people who are students or part-time students

SELECT * FROM People p WHERE VALUE(p) IS OF (Student_t,PartTimeStudent_t) It returns the people that are students or part-time students

208

209

Example - TREAT

Main differences with respect to SQL-99

It is useful for accessing the fields of a subtype of the type of a table or an attribute

CREATE TABLE People OF Person_t; INSERT INTO People VALUES (Person_t(1243, 'Bob', '121 Front St')); INSERT INTO People VALUES (Student_t(3456, 'Joe', '34 View', 12, 'HISTORY')); INSERT INTO People VALUES (PartTimeStudent_t(5678, 'Tim', 13, ****,'PHYSICS', 20)); SELECT TREAT(value(p) as PartTimeStudent_t).numhours FROM People p; SELECT p.numhours FROM People p WHERE value(p) IS OF PartTimeStudent_t; ERROR

Collection types

SQL-99: only ARRAY Oracle: VARRAY and NESTED TABLE SQL-99: on types and tables, no substitutability Oracle: only on types, substitutability

inheritance

210

211

Use of ORDB functions from JDBC

Creation of new types

We consider JDBC 3 (but several extensions are already present in DBC 2) New interfaces to map the object relational types into Java types

Because it is a DDL command DDL, the executeUpdate method is used

String type = CREATE TYPE address_t AS number INTEGER, street VARCHAR(50), city CHAR(20), state CHAR(2), zip INTEGER; Statement st = con.createStatement( ); st.executeUpdate(type);

212

213

Manipulation of new type values from Java

Manipulation of new type values from Java

New interfaces (only for the standard types):

STRUCT

To map ADTs To map values of REF types To map values of array types To simplify the mapping of ADT

REF

For each interface, the choosen driver specify a class that implements the interface In such way, it is possible to handle the differences among DBMS

ARRAY

SQLDATA

214

215

Simple types

ADT

No new interfaces They are manipulated by suing the methods of the base type Example

CREATE TYPE id_employee AS INTEGER FINAL;

The instances of ADT types are mapped into instances of the Struct class An object of type Struct contains a value for each attribute of the ADT to which it corresponds The methods for reading and writing values of ADT types are

The getInt and setInt methods are used to read and write fields of type id_employee

ResultSet - getObject(int): to access attributes if type ADT Struct - getAttributes(): to access the components of an instance PreparedStatement - setObject(int, Struct): to set the parameters of type ADT

216

217

Example
CREATE TABLE Employees ( emp# INTEGER PRIMARY KEY, name CHAR(20), address address_t); We would like to assign to Green the address of employee # 12

Example
String query = select address from Employees where emp# = 12; Statement st = con.createStatement(); ResultSet rs = st.executeQuery(query); rs.next(); Struct address = (Struct) rs.getObject(1); //the instance is obtained Object[ ] addr_attr = address.getAttributes(); //the attributes are accessed System.out.print( Street + addr_attr[1] + + addr_attr[0] + + addr_attr[2] + + addr_attr[4] + + addr_attr[3]);

218

219

Example
String up = update Employees set address = ? where name =Green ; PreparedStatement pst = con.prepareStatement(up); pst.setObject(1, address); pst.executeUpdate( );

Reference types

The instances of REF types are mapped into instance of the Ref class The methods to read and write attributes of REF types are

ResultSet - getRef(int): to access attributies of REF types Ref - getObject( ): given a value of REF type, it returns the referenced object (only in JDBC 3) Ref - setObject(Struct): the parameter becomes the new referenced object (only in JDBC 3) PreparedStatement - setRef(int, Ref): to set parameters of REF types

220

221

Example
CREATE TYPE project_t AS proj# INTEGER, name VARCHAR(20), description VARCHAR(50), budget INTEGER; CREATE TABLE Projects of project_t; CREATE TABLE Employees ( emp# INTEGER PRIMARY KEY, name VARCHAR(50), address address_t, assignment REF(project_t));

Example

We want to assign the project of employee #12 to Green and print the information concerning such project

222

223

Example
String query = select assignment from Employees where emp# = 12; Statement st = con.createStatement(); ResultSet rs = st.executeQuery(query); rs.next(); Ref assegn_ref = rs.getRef(1); //the identifier is obtained String update = update Employees set assignment = ? where name = Green; PreparedStatement pst = con.prepareStatement(update); pst.setRef(1,assegn_ref); pst.executeUpdate( );

Example
Struct project = (Struct) assegn_ref.getObject( ); Object [ ] proj_attr = project. getAttributes(); System.out.print(ind_attr[0] + + ind_attr[1] + + ind_attr[2] + + + ind_attr[3]);

224

225

Array types

Example
CREATE TABLE Employees( emp# INTEGER PRIMARY KEY, name VARCHAR(50), address address_t, skills VARCHAR(20) ARRAY[10]); We want to print all skills of employee #12

The instances of the array types are mapped onto instances of the Array class They represent pointers to the database (the content is not copied) the methods to read and write values of attributes of type array are

ResultSet - getArray(): to access attributes of type array (it returns a pointer to the instance) Array - getArray( ): to copy the data contained in the array into structures local to the program Array - getResultSet (): it returns a result set that contains a tuple for each element of the array, with two columns. The first contains the index (starting from 1), the second the value PreparedStatement - setArray(int,Array): to set parameters of type Array

226

227

Example
String query = select skills from Employees where emp# = 12; Statement st = con.createStatement(); ResultSet rs = st.executeQuery(query); rs.next(); Array comp = rs.getArray(1); //returns a pointer to the array String [ ] comp_arr = (String [ ]) comp.getArray( ); for (int i = 0; i < comp_arr.length(); i++) System.out.print(comp_arr[i])

Collection types

As we have seen, the various systems do not conform to the standard for what concerns the collection types the driver can eventually use the Array class also to manage other collection types

in JDBC the nested tables of Oracle are mapped onto objects of type Array

228

229

SQL Data

Interface each class that implements SQLData allows one to define an explicit mapping between an SQL type and a Java class

The casting is thus explicitly defined by the user Reading an object returns an instance of the class associated by the mapping It is not any longer to use Struct and Object

It simplifies the access to structured objects

We do not see this topic

230