Introduction to Databases
Extended ER Model and other Modelling Languages
Prof. Beat Signer
Department of Computer Science
Vrije Universiteit Brussel
[Link]
2 December 2005
Extended Entity-Relationship (EER) Model
Over the years the ER model has been extended with
additional concepts to increase its expressiveness
specialisation and generalisation (ISA relationship)
ISA constraints
aggregation
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
Specialisation and Generalisation
ISA
An entitiy set can be further subgrouped
or specialised into distinctive entity sets
an entity subset may for example have additional attributes
top-down approach
Parts of multiple entity sets may be generalised in a
higher level entity set
bottom-up approach
Specialisations as well as generalisations are
represented by the same ISA relationship construct
represented by a triangle in the EER model
A ISA B implies that every A entity is also a B entity
generalisation can be seen as the inversion of specialisation
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
ISA Relationship
ISA
id
name
0..1
Persons
LivesAt
0..*
address
ISA
studentID
Students
Locations
Teachers
teaching
hours
ISA
research
topic
February 22, 2013
PhD
Students
Professors
#projects
Beat Signer - Department of Computer Science - bsigner@[Link]
ISA Relationship ...
ISA
The attributes of a higher level entity set (superclass) are
inherited by the lower level entity set (subclass) via the
ISA relationship
e.g. students inherit the id attribute from persons
A subclass also inherits any relationship participation
from its superclass
e.g. students, teachers etc. can participate in the LivesAt
relationship
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
ISA Constraints
ISA
To model an application domain more precisely, specific
constraints can be defined over an ISA relationship
Membership constraints
a condition-defined membership constraint assigns an entity to a
specific subclass based on one or more attribute values
a user-defined membership constraint can be used to manually
assign entities to a specific subclass (manual classification)
Disjointness constraints
a disjoint constraint specifies that an entity can only belong to
exactly one subclass
- e.g. a teacher can be either a PhD student or a professor (but not both)
- expressed by writing the word disjoint next to the triangle
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
ISA Constraints ...
ISA
Disjointness constraints ...
an overlapping constraint over subclasses specifies that an entity
may belong to multiple subclasses
- e.g. a person may be a teacher as well as a student
- by default, subclasses participate in an overlapping constraint
Completeness constraints
a total specialisation (generalisation) constraint defines that each
superclass entity must belong to a subclass
- e.g. each teacher must belong to the PhD Students or Professors subclass
(or to both of them)
- represented by a double line connection to the triangle (like total participation)
a partial specialisation (generalisation) constraint specifies that
some superclass entities may not belong to any subclass
- default completeness constraint
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
ISA Constraints ...
ISA
name
ISA
id
Adults
birthdate
0..1
Persons
LivesAt
0..*
birthdate <
23.02.1996
address
ISA
studentID
Students
Teachers
ISA
research
topic
Locations
PhD
Students
teaching
hours
disjoint
Professors
#projects
disjointness and completeness constraints are independent of each other
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
Aggregation
A limitation of the ER model is that relationships cannot
be modelled over other relationship sets
What if we would like to introduce a manager for each
(employee, company, duration) combination?
from
to
Durations
Employees
id
February 22, 2013
Companies
WorksFor
name
name
address
Beat Signer - Department of Computer Science - bsigner@[Link]
Aggregation ...
We cannot just create a
4-ary relationship WorksFor
not all combinations have
a manager
Introduce a new
4-ary Manages
relationship
from
Durations
Employees
id
problem: we
introduce some
redundancy!
Companies
WorksFor
name
name
address
Manages
- (employee, company, duration)
are replicated in (manager,
employee, company, duration)
February 22, 2013
to
Managers
Beat Signer - Department of Computer Science - bsigner@[Link]
10
Aggregation ...
from
to
Durations
Employees
id
name
An aggregation treats the
relationship as a higher
level entity
Companies
WorksFor
name
address
Manages
Managers
can be used in another relationship
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
11
Part-Of Hierarchies
Cars
3..4
1..1
CPartOf
TPartOf
1..1
1..1
Chassis
Tires
PartOf
subpart
superpart
0..1
Products
0..*
general Part-Of
PartOf
PartOf
particular product
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
12
ER Design Issues
When do we model something as an attribute
and when as an entity set?
there is no general answer and the choice depends on the
specific application domain to be modelled
When do we model something as a relationship set and
when as an entity set?
a relationship set often corresponds to an action between entities
In general we should try to avoid higher level n-ary
relationship sets
3-ary relationship sets should be the maximum and even these
should be used carefully
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
13
Attribute or Entity?
Should the address in the example be modelled
as an attribute or as a separate entity?
This depends on the intended use of the address
if we need an address for different entities (e.g. not just for
employees) it is better to model the address as separate entity
phone
age
#offices
birthday
Employees
id
name
0..1
0..*
Offices
address
street
February 22, 2013
LocatedAt
city
Beat Signer - Department of Computer Science - bsigner@[Link]
14
Attribute or Entity?
Should the duration be modelled as a from and to
attribute or as a separate entity?
Depends again on the use
if we foresee that the same duration period is used by other entity
sets, it should be modelled as a separate entity
from
to
Durations
Employees
id
February 22, 2013
Companies
WorksFor
name
name
address
Beat Signer - Department of Computer Science - bsigner@[Link]
15
N-ary or Binary Relationships?
Basically any n-ary relationship can be replaced
by a number of binary relationship sets
"real" 3-ary relationships should not be replaced
- e.g. WorksFor presented earlier
the Parents relationship is a example of a
non-real 3-ary relationship
- have to use null values if there is either no father or mother
- better to model it as two separate binary relationships
father
Persons
child
mother
father
Parents
HasFather
child
Persons
child
HasMother
mother
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
16
ER Design Principles
Make sure that you pay attention to all the entities, their
relationships and any other constraints mentioned in the
description (requirements) of a specific application
domain
e.g. show cardinality constraints for all relationships
Avoid redundancy!
The resulting conceptual model should be as simple as
possible, but not simpler
introduce only necessary entities and relationships
Some constraints (e.g. functional dependencies) cannot
be expressed in the ER model
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
17
ER Design Principles ...
In two weeks we will see how the conceptual ER model
can be mapped to a logical database schema
(e.g. relational database schema)
The resulting relational schema can be further analysed
by using functional dependencies and normalisation
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
18
ER Summary
Entities and attributes
Relationships
cardinality constraints
participation constraints
Keys
Weak entities
ISA hierarchies
Aggregation
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
19
Classification and Typing
Aristotle (384 BC-322 BC) introduced
a biological classification (ontology)
artificial classification system
organisms with the two subclasses plants
and animals
animals further subclassified into animals
with blood and animals without blood
animals also classified into walking animals,
flying animals and swimming animals (habitat)
Aristotle
Characteristics (problems) of Aristotle's classification
based on behaviour and not on similarities (non-evolutionary)
animals can belong to multiple classes
- e.g. a duck belongs to the walking animals as well as swimming animals class
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
20
Classification and Typing ...
Bertrand Russell defined the
type of a given predicate P as
the entity set spanned by P
intensional notion of type
The notion of type for classifying entities based on a set
of conditions was questionned
by Wittgenstein
Bertrand Russell
Ludwig Wittgenstein
there exists no precise definition (type) for some general concepts
- e.g. no definition for the general concept of a game
We can have a definitional (e.g. intensional) or
prototypical definition of a type
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
21
OM Model
Data model that integrates concepts from
the entity-relationship and object-oriented
data models
based on objects, collections and associations
(binary collections)
- supports associations over associations
different types of collections
Moira C. Norrie
- sets, bags, rankings and sequences
Separates typing and classification (role modelling)
Operational part (algebra) over objects and collections
A family of object database platforms have been realised
based on the OM model (e.g. OMS Java)
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
22
OM Typing and Classification
Classification
Entities
Roles
Employees
employee
Semantic
Grouping
Designers
designer
p1
represented by
subtype
Typing
Females
p2
{ p1,p2 } p1,p2,p3,p4 }
subcollection
Collections
Objects
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
23
OM Example (RSL Link Model)
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
24
The Unified Modelling Language (UML)
UML is a standard
developed by the
Object Management Group (OMG)
used to specify the
components of a
software system
James Rumbaugh
Grady Booch
Ivar Jacobson
based on earlier
work including the object-modelling technique (OMT) by
Rumbaugh et al., the Booch method by Booch and objectoriented software engineering (OOSE) by Jacobson
UML supports different types of diagrams
class diagrams, use case diagrams, package diagrams,...
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
25
UML Class Diagrams
A UML class diagram is similar to an ER model
A UML class is "like" an ER entity set
contains attributes as well as methods
phone
birthday
Employees
id
name
ER entity set
February 22, 2013
employee
age
address
id
name
birthday
phone
address
age()
UML class
Beat Signer - Department of Computer Science - bsigner@[Link]
26
UML Class Diagrams ...
UML relationship class with attributes can be used like
an ER aggregation in other UML relationships
e.g. locatedAt in the example
Note that the position of the cardinality constraints are
switched in UML
Employees
employee
0..1
LocatedAt
office
0..*
Offices
employee
employee
locatedAt
office
0..1
0..*
office
locatedAt
years
years
Employees
employee
LocatedAt
ER relationships
February 22, 2013
office
Offices
employee
employee
office
office
UML relationships
Beat Signer - Department of Computer Science - bsigner@[Link]
27
UML Class Diagrams ...
Teachers
teacher
ISA
PhD
Students
overlapping
Professors
Teachers
ISA
phdStudent
professor
teacher
disjoint
disjoint
PhD
Students
Professors
ER specialisation
February 22, 2013
phdStudent
professor
UML specialisation
Beat Signer - Department of Computer Science - bsigner@[Link]
28
Modelling Exercises
Model the RSL link model in ER
Define an ER model for the following application
You are going to develop a database for the Chamber of Commerce and Industry of Brussels (CVO) to manage information
about language courses in Dutch, Spanish, French and Italian.
The database should store information about the teachers and
students as well as the individual language courses. A semester
lasts for 13 weeks. Regular courses are scheduled for two
sessions (each 3 hours) per week; advanced courses for three
sessions (each 4 hours) per week. Each course has a difficulty
level (module) of 1, 2 or 3 and costs a certain amount of money.
A teacher may not give more than 4 courses and not teach in
more than 2 languages. Each student has to do a final exam and
will get a grade for the exam.
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
29
Homework
Study the following chapter of the
Database System Concepts book
chapter 7
- sections 7.8-7.10
- Extended ER Model (EER)
- Unified Modelling Language (UML)
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
30
Exercise 2
Entity-Relationship (ER) model
Extended Entity-Relationship (EER) model
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
31
References
A. Silberschatz, H. Korth and S. Sudarshan,
Database System Concepts (Sixth Edition),
McGraw-Hill, 2010
M.C. Norrie, An Extended Entity-Relationship Approach
to Data Management in Object-Oriented Systems,
Proceedings of ER 93, 12th International Conference on
Conceptual Modeling, Arlington, USA, December 1993
B. Signer and M.C. Norrie, As We May Link: A General
Metamodel for Hypermedia Systems, Proceedings of ER
2007, 26th International Conference on Conceptual
Modeling, Auckland, New Zealand, November 2007
February 22, 2013
Beat Signer - Department of Computer Science - bsigner@[Link]
32
Next Lecture
Relational Model and Relational Algebra
2 December 2005