Data Modeling and Relational Database Design Instructor Guide PDF

Data Modeling and Relational
Database Design
Volume 1 • Instructor Guide
...........................................................................................
Course Code 20000GC10

Edition 1.0
M08988
®
Authors Copyright  Oracle Corporation, 1998, 1999. All rights reserved.
This documentation contains proprietary information of Oracle Corporation. It is

Jan Speelpenning
provided under a license agreement containing restrictions on use and disclosure
Patrice Daux and is also protected by copyright law. Reverse engineering of the software is
Jeff Gallus prohibited. If this documentation is delivered to a U.S. Government Agency of the
Department of Defense, then it is delivered with Restricted Rights and the
following legend is applicable:
Restricted Rights Legend

Technical Contributors
Use, duplication or disclosure by the Government is subject to restrictions for
and Reviewers commercial computer software and shall be deemed to be Restricted Rights
software under Federal law, as set forth in subparagraph (c) (1) (ii) of DFARS
Simmie Kastner 252.227-7013, Rights in Technical Data and Computer Software (October 1988).
Stijn Vanbrabant
This material or any portion of it may not be copied in any form or by any means
Joni Lounsberry without the express prior written permission of the Worldwide Education Services
Kate Heap group of Oracle Corporation. Any other copying is a violation of copyright law and
may result in civil and/or criminal penalties.
Gabriella Varga
If this documentation is delivered to a U.S. Government Agency not within the
Department of Defense, then it is delivered with “Restricted Right,” as defined in
FAR 52.227-14, Rights in Data-General, including Alternate III (June 1987).
Publishers The information in this document is subject to change without notice. If you find
any problems in the documentation, please report them in writing to Education
Avril Price-Budgen Products, Oracle Corporation, 500 Oracle Parkway, Box 659806, Redwood
Fiona Simpson Shores, CA 94065. Oracle Corporation does not warrant that this document is
error-free.
Don Griffin
Oracle, SQL*Plus, SQL*Net, Oracle Developer, Oracle7, Oracle8, Oracle
Designer and PL/SQL are trademarks or registered trademarks of Oracle
Corporation.
All other products or company names are used for identification purposes only,
and may be trademarks of their respective owners.
Contents
.....................................................................................................................................................
Contents
Lesson 1: Introduction to Entities, Attributes, and Relationships
Introduction 1-2
Why Conceptual Modeling? 1-4
Entity Relationship Modeling 1-7
Goals of Entity Relationship Modeling 1-8
Database Types 1-9
Entities 1-10
Entities and Sets 1-12
Attributes 1-13
Relationships 1-15
Entity Relationship Models and Diagrams 1-17
Representation 1-18
Attribute Representation 1-19
Relationship Representation 1-20
Data and Functionality 1-23
Types of Information 1-24
Other Graphical Elements 1-27
Summary 1-28
Practice 1—1: Instance or Entity 1-29
Practice 1—2: Guest 1-30
Practice 1—3: Reading 1-31
Practice 1—4: Read and Comment 1-32
Practice 1—5: Hotel 1-33
Practice 1—6: Recipe 1-34
General Instructor Notes 1-35
Practices 1-38
Suggested Timing 1-41
Workshop Interviewing 1-42
Instructor Notes 1-44
Instructor Post Lesson Teaching Evaluation 1-53
Suggested Graphics 1-55
.....................................................................................................................................................
®
iii
Contents
.....................................................................................................................................................
Lesson 2: Entities and Attributes in Detail

Introduction 2-2
Data Compared to Information 2-4
Data 2-5
Tracking Entities 2-7
Electronic Mail Example 2-9
Evolution of an Entity Definition 2-11
Functionality 2-13
Tracking Attributes 2-14
Subtypes and Supertypes 2-17
Summary 2-20
Practice 2—1: Books 2-21
Practice 2—2: Moonlight 2-22
Practice 2—3: Shops 2-23
Practice 2—4: Subtypes 2-24
Practice 2—5: Schedule 2-25
Practice 2—6: Address 2-26
Practice 2—6: Address (continued) 2-27
Lesson 3: Relationships in Detail

Introduction 3-2
Establishing a Relationship 3-4
Relationship Types 3-9
Relationships and Attributes 3-16
Attribute Compared to Relationship 3-18
Relationship Compared to Attribute 3-19
m:m Relationships May Hide Something 3-20
Resolving Relationships 3-25
Summary 3-28
Practice 3—1: Read the Relationship 3-29
Practice 3—2: Find a Context 3-30
Practice 3—3: Name the Intersection Entity 3-31
Practice 3—4: Receipt 3-32
.....................................................................................................................................................
iv Data Modeling and Relational Database Design
Contents
.....................................................................................................................................................
Practice 3—5: Moonlight P&O 3-33

Practice 3—6: Price List 3-35
Practice 3—7: E-mail 3-36
Practice 3—8: Holiday 3-37
Lesson 4: Constraints
Introduction 4-2
Identification 4-4
Unique Identifier 4-6
Arcs 4-12
Arc or Subtypes 4-16
More About Arcs and Subtypes 4-17
Hidden Relationships 4-18
Domains 4-19
Some Special Constraints 4-20
Summary 4-24
Practice 4—1: Identification Please 4-25
Practice 4—2: Identification 4-26
Practice 4—3: Moonlight UID 4-28
Practice 4—4: Tables 4-29
Practice 4—5: Modeling Constraints 4-30
Lesson 5: Modeling Change

Introduction 5-2
Time 5-4
Date as Opposed to Day 5-5
Entity DAY 5-6
Modeling Changes Over Time 5-7
A Time Example: Prices 5-10
Current Price 5-16
Journalling 5-17
.....................................................................................................................................................
®
v
Contents
.....................................................................................................................................................
Summary 5-19
Practice 5—1: Shift 5-20
Practice 5—2: Strawberry Wafer 5-21
Practice 5—3: Bundles 5-22
Practice 5—4: Product Structure 5-24
Lesson 6: Advanced Modeling Topics

Introduction 6-2
Patterns 6-4
Master Detail 6-5
Basket 6-6
Classification 6-7
Hierarchy 6-8
Chain 6-10
Network 6-11
Symmetric Relationships 6-13
Roles 6-14
Fan Trap 6-15
Data Warehouse 6-16
Drawing Conventions 6-17
Generic Modeling 6-19
Generic Models 6-20
More Generic Models 6-21
Most Generic Model 6-22
Summary 6-23
Practice 6—1: Patterns 6-24
Practice 6—2: Data Warehouse 6-25
Practice 6—3: Argos and Erats 6-26
Practice 6—4: Synonym 6-27
.....................................................................................................................................................
®
vi
Contents
.....................................................................................................................................................
Lesson 7: Mapping the ER Model

Introduction 7-2
Why Create a Database Design? 7-4
Transformation Process 7-6
Naming Convention 7-8
Basic Mapping 7-12
Relationship Mapping 7-14
Mapping of Subtypes 7-20
Subtype Implementation 7-23
Summary 7-29
Practice 7—1: Mapping Supertype 7-30
Practice 7—2: Quality Check Subtype Implementation 7-31
Practice 7—3: Quality Check Arc Implementation 7-32
Practice 7—4: Mapping Primary Keys and Columns 7-33
Lesson 8: Denormalized Data

Introduction 8-2
Why and When to Denormalize 8-4
Storing Derivable Values 8-6
Pre-Joining Tables 8-8
Hard-Coded Values 8-10
Keeping Details With Master 8-12
Repeating Single Detail with Master 8-14
Short-Circuit Keys 8-16
End Date Columns 8-18
Current Indicator Column 8-20
Hierarchy Level Indicator 8-22
Denormalization Summary 8-24
Practice 8—1: Name that Denormalization 8-25
Practice 8—2: Triggers 8-26
Practice 8—3: Denormalize Price Lists 8-29
Practice 8—4: Global Naming 8-30
.....................................................................................................................................................
®
vii
Contents
.....................................................................................................................................................

Lesson 9: Database Design Considerations

Introduction 9-2
Reconsidering the Database Design 9-4
Oracle Data Types 9-5
Most Commonly-Used Oracle Data Types 9-6
Column Sequence 9-7
Primary Keys and Unique Keys 9-8
Artificial Keys 9-11
Sequences 9-13
Indexes 9-16
Choosing Columns to Index 9-19
When Are Indexes Used? 9-21
Views 9-23
Use of Views 9-24
Old-Fashioned Design 9-25
Distributed Design 9-27
Benefits of Distributed Design 9-28
Oracle Database Structure 9-29
Summary 9-31
Practice 9—1: Data Types 9-32
Practice 9—2: Artificial Keys 9-34
Practice 9—3: Product Pictures 9-35
Appendix A: Solutions
Introduction to Solutions A-2
Practice 1—1 Instance or Entity: Solution A-4
Practice 1—2 Guest: Solution A-5
Practice 1—3 Reading: Solution A-6
Practice 1—4 Read and Comment: Solution A-7
Practice 1—5 Hotel: Solution A-8
Practice 1—6 Recipe: Solution A-9
.....................................................................................................................................................
®
viii
Contents
.....................................................................................................................................................
Practice 2—1 Books: Solution A-11

Practice 2—2 Moonlight: Solution A-12
Practice 2—3 Shops: Solution A-13
Practice 2—4 Subtypes: Solution A-14
Practice 2—5 Schedule: Solution A-15
Practice 2—6 Address: Solution A-16
Practice 3—1 Read the Relationship: Solution A-18
Practice 3—2 Find a Context: Solution A-19
Practice 3—3 Name the Intersection Entity: Solution A-20
Practice 3—4 Receipt: Solution A-21
Practice 3—5 Moonlight P&O: Solution A-23
Practice 3—6 Price List: Solution A-27
Practice 3—7 E-mail: Solution A-28
Practice 3—8 Holiday: Solution A-30
Practice 4—1 Identification Please: Solution A-32
Practice 4—2 Identification: Solution A-34
Practice 4—3 Moonlight UID: Solution A-37
Practice 4—4 Tables: Solution A-38
Practice 4—5 Constraints: Solution A-39
Practice 5—1 Shift: Solution A-40
Practice 5—2 Strawberry Wafer: Solution A-41
Practice 5—3 Bundles: Solution A-42
Practice 5—4 Product Structure: Solution A-44
Practice 6—1 Patterns: Solution A-45
Practice 6—2 Data Warehouse: Solution A-46
Practice 6—3 Argos and Erats: Solution A-47
Practice 6—4 Synonym: Solution A-48
Practice 7—1 Mapping Supertype: Solution A-49
Practice 7—2 Quality Check Subtype Implementation: Solution A-50
Practice 7—3 Quality Check Arc Implementation: Solution A-51
Practice 7—4 Primary Keys and Columns: Solution A-52
Practice 8—1 Name that Denormalization: Solution A-53
Practice 8—2 Triggers: Solution A-54
Practice 8—3 Denormalize Price Lists: Solution A-56
Practice 8—4 Global Naming: Solution A-58
Practice 9—1 Data Types: Solution A-59
.....................................................................................................................................................
®
ix
Contents
.....................................................................................................................................................
Practice 9—2 Artificial Keys: Solution A-61

Practice 9—3 Product Pictures: Solution A-62
Appendix B: Normalization
Introduction B-2
Normalization and its Benefits B-3
First Normal Form B-7
Second Normal Form B-9
Third Normal Form B-11
Normalization During Data Modeling B-13
Summary B-16
Instructor Notes B-17
.....................................................................................................................................................
®
x
1
.................................
Introduction to
Entities, Attributes, and
Relationships
.........................................................................................................................................
Introduction
Lesson Aim
This lesson explains the reasons for conceptual modeling and introduces the key role
players: entities, attributes, and relationships.
Schedule
Overview See Page 44
About the slide

See Page 44
• Why conceptual modeling?
• Introduction of the Key role players:
– Entities
– Attributes
– Relationships
1-2
Topic See Page

Introduction 2
Why Conceptual Modeling? 4
Entity Relationship Modeling 7
Goals of Entity Relationship Modeling 8
Database Types 9
Entities 10
Entities and Sets 12
Attributes 13
Relationships 15
Entity Relationship Models and Diagrams 17
Representation 18
Attribute Representation 19
Relationship Representation 20
Data and Functionality 23
.............................................................................................................................................
1-2 Data Modeling and Relational Database Design
Introduction
.........................................................................................................................................
Topic See Page

Types of Information 24
Other Graphical Elements 27
Summary 28
Practice 1—1: Instance or Entity 29
Practice 1—2: Guest 30
Practice 1—3: Reading 31
Practice 1—4: Read and Comment 32
Practice 1—5: Hotel 33
Practice 1—6: Recipe 34
Objectives
At the end of this lesson, you should be able to do the following:
• Explain why conceptual modeling is important
• Describe what an entity is and give examples
• Describe what an attribute is and give examples
• Describe what a relationship is and give examples
• Draw a simple diagram
• Read a simple diagram
.........................................................................................................................................
®
1-3
.........................................................................................................................................
Why Conceptual Modeling?

This is a course on conceptual data modeling and physical data modeling. Why do you
need to learn this? Why invest time in creating entity models when you need tables?
Why bother about business functionality and interviews and feedback sessions when
you need programs? In this course you learn why. You learn why it is a wise decision
to spend time in modeling and why it is a good investment. You will learn even more,
including how to create, read, and understand models and how to check them, as well
as how to derive table and key definitions from them.
About the slide

Why Create a Conceptual Model? See Page 44
• It describes exactly the information needs of the

business
• It facilitates discussion
• It helps to prevent mistakes, misunderstanding
• It forms important “ideal system” documentation
• It forms a sound basis for physical database
design
• It is a very good practice with many practitioners
1-3
This list shows the reasons for creating a conceptual model. The most important
reason is that a conceptual model facilitates the discussion on the shape of the future
system. It helps communication between you and your sponsor as well as you and your
colleagues. A model also forms a basis for the default design of the physical database.
Last but not least, it is relatively cheap to make and very cheap to change.
What You Learn in This Course

In this course you learn how to analyze the requirements of a business, how to
represent your findings in an entity relationship diagram and how to define and refine
the tables and various other database objects from that model.
In summary, as a result of what you learn in this course you will know:
• How to model the information needs of a business and the rules that apply.
• Which tables you need in your database, and why.
• Which columns you need in your tables, and why.
.............................................................................................................................................
Why Conceptual Modeling?
.........................................................................................................................................
• Which constraints and other database objects you require.

You will also know how to explain this to:
• Your sponsors.
• The developers.
• Your fellow designers.
The House Building Metaphor

Imagine someone who wants to have a house built. Initially, the house only exists in
the minds of the future home owners as ideas, or as pieces of various dreams.
Sometimes the future inhabitants may not even know what they want, or know if what
they want is even feasible. Dreams may be full of internal contradictions and
impossibilities.This is not a problem in the dream world, but in the physical realm any
inconsistencies and obstacles need to be resolved before someone can construct the
house.
About the slide

Between Dream and Reality... See Page 44
1-4
A building contractor needs a solid plan, a set of blueprints of the house with a
description of the materials to be used, the size of the roof beams, the capacity of the
plumbing and many, many other things. The contractor follows the plan, and has the
knowledge to construct what is on the blueprint. But how do the ideas of the home
owner become the blueprint for contractor? This is where the architect becomes
involved.
.........................................................................................................................................
®
1-5
.........................................................................................................................................
The Architect
The architects are the intermediary between sponsor and constructor. They are trained
in the skills of translating ideas into models. The architect listens to the description of
the ideas and asks all kinds of questions. The architect’s skills in extracting the ideas,
putting it down in a format that allows discussion and analysis, giving advice,
describing sensible options, documenting it, and confirming it with the home owners,
are the cornerstones to providing the future home-owner with a plan of the home they
want.
Sketches
The architect’s understanding of the dreams is transformed into sketches of the new
house—only sketches! These consist of floor plans and several artist’s impressions,
and show the functional requirements of the house, not the details of the construction.
This is a conceptual model, the first version.
Easy Change
If parts of the model are not satisfactory or are misunderstood, the model can easily be
changed. Such a change would only need a little time and an eraser, or a fresh sheet of
paper. Remember, it is only changing a model. The cost of change at this stage is very
low. Certainly it is far less costly than making changes to the floor plan or roof
dimensions after construction has started. The house model is then reviewed again,
and further changes are made. The architect continues to explore and clarify the
dreams and make alternative suggestions until all controversial issues are settled, and
the model is stable and ready for the final approval by the sponsor.
Technical Design
Then the architect converts the model into a technical design, a plan the contractor can
use to build the house. Calculations are made to determine, for example, the number of
doors, how thick the walls and floor beams must be, the dimensions of the plumbing,
and the exact construction of the roof. These are technical issues that need not involve
the customer.
What? as Opposed to How?

While the conceptual model addresses the What? phase in the process, the design
addresses the question of How? it is to be constructed.
Conceptual modeling is similar to the work of an architect—transforming things that
only exist in people’s minds into a design that is sufficiently substantial to be created
physically.
.............................................................................................................................................
Entity Relationship Modeling
.........................................................................................................................................

PRICE LEVEL
# CODE
* DESCRIPTION
defined by applied to
part of
ORGANIZATION
•
o EMAIL
* NAME
Models business,
o POSTAL CODE
at o REGION
o STREET
o TOWN parent organization o
o TELEPHONE NUMBER
TITLE MOVIE o CONTACT NAME
# PRODUCT CODE* CATEGORY o CONTACT EXTENSION
* TITLE o AGE RATING
o DESCRIPTION * DURATION the
* MONOCHROME GAME
not implementation
o AUDIO * CATEGORY
o PREVIEW * MEDIUM SUPPLIER
o MINIMUM MEMO # SUPPLIER CODE
o EMAIL
available * APPROVED
for * REFERENCE
the source of
PRICE HISTORY
# EFFECTIVE DATE reviewed inavailable as on
* PRICE
•
* DEFAULT DAYS
* OVERDUE RATE
Is a well-established
OTHER ORGANIZATION
PUBLICATION
# REFERENCE
* TITLE
CATALOG the holder o
# REFERENCE
technique
o VOLUME o CATALOG DAT
o ISSUE o DESCRIPTION CUSTOMER
o PUBLISH DATE o EMAIL
managed b
* DESIGNATION
EMPLOYEE the manager o * FIRST NAME
the source of the source of * POSITION
* LAST NAME
o OTHER INITIALS
* LAST NAME * STREET
o FIRST NAME * TOWN
o OTHER INITIALS * POSTAL CODE
o EMAIL * REGION
o HOME PHONE
•
acquired fro o WORK EXTENSION
the cancellor of responsible
Has a robust syntax

o WORK PHONE
m responsible o PHOTOGRAPH
of o STAFF REMARKS
for for
COPY
* ACQUIRE DATE
* PURCHASE COST
* SHELF CODE
the holder of
o CONDITION responsible for
o CUSTOMER REMARKS
•
... MEMBERSHIP TYPE
# CODE
Results in easy-to-
* DESCRIPTION
rented on reserved on * DISCOUNT PERCENTAGE
o STANDARD FEE
held by the type o

held by
read diagrams… REVIEW

in
# SEQUENCE
* ARTICLE
* HOT
in
of
approved by
MEMBERSHIP
# NUMBER
o TERMINATION REASON
o TERMINATION DATE
of
o AUTHOR
o URL
renewed fo used fo
the reservation for r r
cancelled by
for
requested authorized by of
...although they may

against
the
BOOKING
* BOOK DATE requestor #MEMBERSHIP
START DATE
PERIOD
o EXPIRE DATE o ACTUAL FEE PAID
o NOTIFY DATE of
o RESERVE DATE
o STAFF REMARKS
approved by
look rather complex

for
fulfilled as
RENTAL
* RENTAL DATE
o STAFF REMARKS
o COMPLETED
at first sight for

RENTAL ITEM
the rental for
composed of
part of
# LINE NO
* RENTAL PERIOD
* PRICE PAID
o RETURN DATE
o STAFF REMARKS
1-5
What is Involved in Modeling?

Entity Relationship modeling is about modeling a business. To be more precise: it is
about modeling the data requirements for a business based on the current or desired
functionality of the future system.
To model a business you have to understand to a fair degree of detail what the business
is about.
Entity Relationship modeling is a technique used to describe the shared understanding
of the information needs of a business. It is a well-established technique that leads to
diagrams which are quite easy to read and therefore also easy to check.
.........................................................................................................................................
®
1-7
.........................................................................................................................................
Goals of Entity Relationship Modeling

About the slide
Goals of Entity Relationship Modeling See Page 45
• Capture all required information

• Information appears only once
• Model no information that is derivable from other
information already modeled
• Information is in a predictable, logical place
1-6
The goals of conceptual data modeling are to ensure that:

• All pieces of information that are required to run a business properly are
recognized.
Models should be complete. Requirements should be known before you start
implementing. Dependencies must be clear.
• Every single piece of required information appears only once in the model.
This is an important goal. As soon as a system stores particular information twice,
you run into the possibility that this information is not the same in both places. If
you are a user of an information system and discover inconsistencies in the data,
which information would you to trust?
This goal implies that an ideal system does not contain derivable information.
• In the future system, the information is made available in a predictable, logical
place; related information is kept together.
• A proper Entity Relationship model leads to a set of logically coherent tables.
.............................................................................................................................................
Database Types
.........................................................................................................................................
Database Types
About the slide
Database Types See Page 45
ER Model
Hierarchical Network
Relational
1-7
Entity Relationship modeling is independent of the hardware or software used for

implementation. Although you can use an Entity Relationship model as a basis for
hierarchical databases, network databases, and relational databases, it is strongly
connected to the latter.
.........................................................................................................................................
®
1-9
.........................................................................................................................................
Entities
This section gives definitions and examples.
About the slide
Entity See Page 45
• An Entity is:
– “Something” of significance to the business
about which data must be known.
– A name for the things that you can list.
– Usually a noun.
• Examples: objects, events
• Entities have instances.
1-8
Definition of an Entity
There are many definitions and descriptions of an entity. Here are a few; some are
quite informal, some are very precise.
• An entity is something of interest.
• An entity is a category of things that are important for a business, about which
information must be kept.
• An entity is something you can make a list of, and which is important for the
business.
• An entity is a class or type of things.
• An entity is a named thing, usually a noun.
Two important aspects of an entity are that it has instances and that the instances of the
entity somehow are of interest to the business.
Note the difference between an entity and an instance of an entity.
.............................................................................................................................................
Entities
.........................................................................................................................................
More on Entities
About the slide

Entities and Instances See Page 46
PERSON Mahatma Gandhi

PRODUCT 2.5 x 35 mm copper nail
PRODUCT TYPE nail
EMPLOYMENT CONTRACT my previous contract
JOB violinist
SKILL LEVEL fluent
TICKET RESERVATION tonight: Hamlet in the Royal
PURCHASE the CD I bought yesterday
ELECTION for parliament next fall
PRINTER PREFERENCE …
DOCUMENT VERSION ...
1-9
The illustration shows examples of entities and examples of instances of those entities.
Note:
• There are many entities.
• Some entities have many instances, some have only a few.
• Entities can be:
– Tangible, like PERSON or PRODUCT.
– Nontangible, like REQUIRED SKILL LEVEL.
– An event, like ELECTION.
• An instance of one entity may be an entity in its own right: the instance “violinist”
of entity JOB could be the name of another entity with instances like “David
Oistrach”, “Kyung-Wha Chung.”
.........................................................................................................................................
®
1-11
.........................................................................................................................................
Entities and Sets

About the slide
Entities and Sets See Page 46
• An entity represents a set of instances that are of

interest to a particular business.
JOB
manager
cook
waitress
dish washer
financial controller
porter
waiter
piano player
1-10
You can regard entities as sets. The illustration shows a set JOB and the set shows
some of its instances. At the end of the entity modeling process entities are
transformed into tables; the rows of those tables represent an individual instance.
During entity modeling you look for properties and rules that are true for the whole
set. Often you can decide on the rules by thinking about example instances. The
following lessons contain many examples of this.
Set Theory
Entity relationship modeling and the theory of relational databases are both based on a
sound mathematical theory, that is, set theory.
.............................................................................................................................................
Attributes
.........................................................................................................................................
Attributes
About the slide
Attribute See Page 46
• Also represents something of significance to the

business
• Is a single valued property detail of an entity
• Is a specific piece of information that:
– Describes
– Quantifies
– Qualifies
– Classifies
– Specifies
an entity.
1-11
What is an Attribute?
An attribute is a piece of information that in some way describes an entity. An attribute
is a property of the entity, a small detail about the entity.
Entities Have Attributes

For now, assume that all entities have at least one attribute. Later, you discover
exceptions to this assumption. The attribute describes, quantifies, qualifies, classifies,
and specifies an entity. Usually, there are many attributes for an entity, but again, we
are only interested in those attributes that are of importance to the business.
Values and Data Types

Attributes have values. An attribute value can be a number, a character string, a date,
an image, a sound, and even more. These are called data types or formats. Usually the
values for a particular attribute of the instances of an entity all have the same data type.
Every attribute has a data type.
Attribute is Single Valued

An attribute for an entity must be single valued. In more precise terms, an entity
instance can have only one value for that attribute at any point in time. This is the most
important characteristic of an attribute.
The attribute value, however, may change over time.
.........................................................................................................................................
®
1-13
.........................................................................................................................................
Attribute Examples
About the slide

Attribute Examples See Page 46
Practices
See Page 47
Entity Attribute
This is a suitable
EMPLOYEE Family Name, Age, Shoe Size, place to do
Town of Residence, Email, ... practices 1-1 and
CAR Model, Weight, Catalog Price, … 1-2
ORDER Order Date, Ship Date, …
JOB Title, Description, ...
TRANSACTION Amount, Transaction Date, …
EMPLOYMENT Start Date, Salary, ...
CONTRACT
1-12
Note:
• Attribute Town of Residence for EMPLOYEE is an example of an attribute that is
quite likely to change, but is probably single valued at any point in time.
• Attribute Shoe Size may seem to be of no importance, but that depends on the
business: if the business supplies industrial clothing to its employees, this may be a
very sensible attribute to take.
• Attribute Family Name may not seem to be single-valued for someone with a
double name. This double name, however, can be regarded as a single string of
characters that forms just one name.
Volatile Attributes
Some attributes are volatile (unstable). An example is the attribute Age. Always look
for nonvolatile, stable, attributes. If there is a choice, use the nonvolatile one. For
example, use the attribute Birth Date instead of Age.
.............................................................................................................................................
Relationships
.........................................................................................................................................
Relationships
Relationships
• Also represent something of significance to the

business
• Express how entities are mutually related
• Always exist between two entities (or one entity twice)
• Always have two perspectives
• Are named at both ends
1-13
Entities usually have relationships. Here are some examples.
About the slide

Relationship Examples See Page 47
EMPLOYEES have JOBS

JOBS are held by EMPLOYEES
PRODUCTS are classified by a PRODUCT TYPE

PRODUCT TYPE is a classification for a PRODUCT
PEOPLE make TICKET RESERVATIONS

TICKET RESERVATIONS are made by PEOPLE
1-14
.........................................................................................................................................
®
1-15
.........................................................................................................................................
A relationship connects two entities. A relationship represents a significant

dependency of two entities—always two entities.
A particular relationship can be worded in many ways: An EMPLOYEE has a JOB, or
an EMPLOYEE performs a JOB, or an EMPLOYEE holds a JOB.
An EMPLOYEE applies for a JOB expresses a different relationship. Note that this
example shows that two entities can have more than one relationship.
About the slide

Employees have Jobs See Page 47
JOB manager
EMPLOYEE cook
Shintaro waitress
dish washer
Jill financial controller
Adam
Ahmed porter
waiter
Maria
piano player
Numerical observation:
• All EMPLOYEES have a JOB
• No EMPLOYEE has more than one JOB
• Not all JOBS are held by an EMPLOYEE
• Some JOBS are held by more than one EMPLOYEE
1-15
Based on what you know about instances of the entities, you can decide on four
questions:
• Must every employee have a job?
In other words, is this a mandatory or optional relationship for an employee?
• Can employees have more than one job?
and
• Must every job be done by an employee?
In other words, is this a mandatory or optional relationship for a job?
• Can a job be done by more than one employee?
Later on we will see why these questions are important and why (and how) the
answers have an impact on the table design.
.............................................................................................................................................
Entity Relationship Models and Diagrams
.........................................................................................................................................
Entity Relationship Models and Diagrams

An Entity Relationship Model (ER Model) is a list of all entities and attributes as well
as all relationships between the entities that are of importance. The model also
provides background information such as entity descriptions, data types and
constraints. The model does not necessarily include a picture, but usually a diagram of
the model is very valuable.
An Entity Relationship Diagram (ER Diagram) is a picture, a representation of the
model or of a part of the model. Usually one model is represented in several diagrams,
showing different business perspectives.
Graphical Elements
Entity Relationship diagramming uses a number of graphical elements. These are
discussed in the next pages.
Unfortunately, there is no ISO standard representation of ER diagrams. Oracle has its
own convention. In this course we use the Oracle diagramming technique, which is
built into the Oracle Designer tool.
.........................................................................................................................................
®
1-17
.........................................................................................................................................
Representation
Entity
About the slide

Entity Representation in Diagram See Page 48
• Drawn as a “softbox”
• Name singular
EMPLOYEE JOB
• Name inside
ELECTION
• Neither size,
nor position
has a special TICKET
meaning ORDER
RESERVATION
JOB ASSIGNMENT
During design, entities usually lead to tables.
1-16
In an ER diagram entities are drawn as soft boxes with the entity name inside. Borders
of the entity boxes never cross each other. Entity boxes are always drawn upright.
Throughout this book, entity names are printed in capitals. Entity names are preferably
in the singular form; you will find that diagrams are easier to read this way.
Box Size
Neither the size of an entity, nor its position, has a special meaning. However, a reader
might construe a larger entity to be of more importance than a smaller one.
Where Entities Lead
During the design for a relational database, an entity usually leads to a table.
.............................................................................................................................................
Attribute Representation
.........................................................................................................................................
Attribute Representation
About the slide
Attributes in Diagrams See Page 48
EMPLOYEE JOB
* Family Name * Title
* Address o Description
o Birth Date
o Shoe Size
o Email
Mandatory attribute, that is, known and available

* for every instance
o Optional attribute, that is, unknown or unimportant
to know for some instances
During design, attributes lead to columns.

1-17
Attributes are listed within the entity box. They may be preceded by a * or an O. These
symbols mean that the attribute is mandatory or optional, respectively. Throughout
this book attributes are printed in Initial Capital format.
* Mandatory: It is realistic to assume that for every instance of the entity the
attribute value is known and available when the entity instance is recorded and that
there is a business need to record the value.
o Optional: The value of the attribute for an instance of the entity may be unknown
or unavailable when that instance is recorded or the value may be known but of no
importance.
Not all attributes of an entity need to be present in the diagram, but all attributes must
be known before making the table design. Often only a few attributes are shown in a
diagram, for reasons of clarity and readability. Usually you choose those attributes that
help understanding of what the entity is about and which more or less “define” the
entity.
Where Attributes Lead
During design an attribute usually leads to a column. A mandatory attribute leads to a
not null column.
.........................................................................................................................................
®
1-19
.........................................................................................................................................
Relationship Representation
Relationships are represented by a line, connecting the entities. The name of the
relationship, from either perspective, is printed near the starting point of the
relationship line.
The shape of the end of the relationship line represents the degree of the relationship.
This is either one or many. One means exactly one; many means one or more.
About the slide

Relationship in Diagrams See Page 48
An employee has exactly one job
EMPLOYEE JOB
has
held by
Jobs are held by one or more employees
During design, relationships lead to foreign keys.
1-18
In the above example, it is assumed that JOBS are held by one or more EMPLOYEES.
This is shown by the tripod (or crowsfoot), at EMPLOYEE.
An EMPLOYEE, on the other hand, is assumed here to have exactly one JOB. This is
represented by the single line at JOB.
The relationship line may be straight, but may also be curved; curves have no special
meaning, nor does the position of the starting point of the relationship line. The
diagram below represents exactly the same model, but arguably less clearly.
has
JOB
EMPLOYEE
held by
.............................................................................................................................................
Relationship Representation
.........................................................................................................................................
Mandatory and Optional Relationships

Relationships can be mandatory or optional, in the same way as attributes. Mandatory
relationships are drawn as a solid line; optional relationships as dotted lines.
mandatory: optional:
Relationship and Relationship Ends

Here, the relationship between EMPLOYEE and JOB is modeled using the optional
relationship end and mandatory relationship end notation.
About the slide

See Page 48
EMPLOYEE has JOB
held by
When you read the relationship, imagine it split into two perspectives:
About the slide

JOB See Page 48
EMPLOYEE has
held by
Every EMPLOYEE has exactly one JOB or, alternatively:

An EMPLOYEE must have exactly one JOB.
EMPLOYEE has JOB
held by
A JOB may be held by one or more EMPLOYEES.
.........................................................................................................................................
®
1-21
.........................................................................................................................................
Reading a Relationship End
About the slide

See Page 49
Reading a Relationship End
Practices
See Page 49
P split into Q This is a suitable
part of
place to do
practices 1-1 and
1-2
“Each“Each must be
P P may be split intosplit
may be
or more Qs”
oneinto
exactly one Q
one or more Qs
”
“Each “QEachmust
Q
be
must be partpart
may be
of
of exactly
exactly one P
one P”
one or more Ps
”
1-24
A relationship from entity1 to entity2 must be read:
Each entity1 {must be | may be}

relationship_name
{one or more | exactly one} entity2
Where Relationships Lead

During design relationships lead to foreign keys and foreign key columns. An optional
relationship leads to non mandatory foreign key columns.
Relationship Name in the Diagram

Throughout this book relationship names in the diagrams are printed in lower case
italics.
For reasons of space and readability of the diagrams in this book, relationship names
are sometimes kept very short, and sometimes only a preposition is used.
.............................................................................................................................................
Data and Functionality
.........................................................................................................................................
Data and Functionality

About the slide
Functions Drive Data See Page 50
• Business functions are always present.

– Explicit
– Assumed
• Business functions need data.
• An entity, attribute, or relationship may be
modeled because:
– It is used by a business function.
– The business need may arise in the near
future.
1-25
Functions Drive the Conceptual Data Model

Although this course does not cover the method of function modeling, functions are
present at any time, in any discussion on a conceptual data model. You cannot talk
about, nor judge a conceptual data model without knowing or assuming the desired
functionality of the future system.
Often a conceptual data model discussion may seem to be about the data structure but
actually is about functionality, usually unclear or undetermined pieces of functionality.
The language used is that of the conceptual data model, the representation used is that
of the entity relationship diagram, but the discussion in fact is about functionality.
Functions drive the conceptual data model. The question “Do we need to take Shoe
Size for an employee?” can only be answered by answering positively the question “Is
there a business function that needs it?”
Consider the conceptual data model as the shadow of the functions of a system.
Most of the time during this course, functionality is only briefly sketched, or merely
assumed, to prevent you from reading page after page of functional descriptions.
.........................................................................................................................................
®
1-23
.........................................................................................................................................
Types of Information
About the slide
Weather Forecast See Page 50
-DQXDU\
.¡EHQKDYQ
%UHPHQ
%HUOLQ
0QFKHQ

$PVWHUGDP
%UX[HOOHV
3DULV
%RUGHDX[
1-26
What Information is Available?

The illustration shows a piece of a weather forecast torn from a European newspaper,
showing various types of information. What are the types of information? One of the
first things you will see are, for example, “København”, “Bremen”. These are cities, or
more precisely, names of cities. The little drawings represent the type of weather;
these drawings are icons. The next columns are temperatures, probably maximum and
minimum; the arrows indicate wind direction and the number next to it is the wind
force. Then there is a date on top which is the forecast date. Therefore we have:
• City
• Name of the city (such as “København”)
• Weather type (such as “cloudy with rain”)
• Icon of the weather type
• Minimum temperature
• Maximum temperature
• Wind direction arrow
• Wind force
• Forecast date
Is this all?
No, you can find out even more information. To do this you have to have some
“business” knowledge. In this case it is geographical knowledge.
.............................................................................................................................................
Types of Information
.........................................................................................................................................
About the slide

See Page 50
DK København
(Copenhagen)
IR UK
NL Bremen
Amsterdam
Berlin
BE Bruxelles DE
(Brussels)
LU München
Paris (Munich)
FR
CH
Bordeaux IT
1-27
You may notice that the cities in the weather forecast are not printed in a random order.
The German cities (Bremen, Berlin and Munchen) are grouped together, just as the
French cities are. Moreover, the cities are not ordered alphabetically by name but seem
to be ordered North-South. Apparently this report “knows” something to facilitate the
grouping and sorting. This could be:
• Country of the city
• Geographical position of the city
and maybe even
• Geographical position of the country
Next Step
Try to identify which of the above types of information is probably an entity, which is
an attribute and which is a relationship.
City and Country are easy. These are entities, both with, at least, attribute Name and
Geographical Position. Weather Type could also be an entity as there is an attribute
available: Icon. For the same reason there could be an entity Wind Direction. Now,
where does this leave the temperatures and forecast date? These cannot be attributes of
City as the forecast date is not single value for a City: there can be many forecast dates
for a city. This is how you discover that there is still one entity missing, such as
Forecast, with attributes Date, Minimum and Maximum Temperature, Wind Force.
.........................................................................................................................................
®
1-25
.........................................................................................................................................
There are likely to be relationships between:

• COUNTRY and CITY
• CITY and FORECAST
• FORECAST and WEATHER TYPE
• FORECAST and WIND DIRECTION.
About the slide

Weather Forecast, a Solution See Page 50
CITY COUNTRY
* Name located in * Name
o Geographical o Geographical
Position having Position
subject of
about
FORECAST referring to WEATHER TYPE

* Date * Icon
o Minimum referred in * Description
Temperature
o Maximum referring to WIND DIRECTION
Temperature * Icon
o Wind Force
referred in * Description
1-28
In this entity relationship diagram some assumptions are made about the relationships:
• Every FORECAST must be about one CITY, and
not all CITIES must be in a FORECAST—but may be in many
• Every CITY is located in a COUNTRY, and
every COUNTRY has one or more CITIES
• A FORECAST must not always contain a WEATHER TYPE, and
not all WEATHER TYPES are in a FORECAST—but may be in many
• A FORECAST must not always contain a WIND DIRECTION, and
not all WIND DIRECTIONS are in a FORECAST—but may be in many
The rationale behind these assumptions is that we consider an incomplete FORECAST
still to be a FORECAST, unless we do not know the date or the CITY the FORECAST
refers to.
.............................................................................................................................................
Other Graphical Elements
.........................................................................................................................................
Other Graphical Elements

About the slide
See Page 50
Graphical Elements of ER Diagram
Entity
Attribute ** **
o
Relationship
Subtype
Unique identifier
Arc
Nontransferability
#o #
1-29
The illustration shows all graphical elements you can encounter in a ER diagram. You
saw earlier how to represent an entity, an attribute, and a relationship.
The lessons following this one discuss the remaining four types of elements:
• Subtype, represented as an entity within the boundary of another entity
• Unique identifier, represented as a # in front of an attribute or as a bar across a
relationship line
• Arc, represented as an arc-shaped line across two or more relationship lines
• Nontransferability symbol, represented as a diamond across a relationship line
Limited Set of Graphical Elements

As you can see, the set of graphical elements in ER diagramming is very limited. The
complexity of ER modeling is clearly not in the representation. The main complexity
of ER modeling lies in the understanding of the business, in the recognition of the
entities that play a role in that business, the relevant attributes that describe the
entities, and the relationships that connect them.
.........................................................................................................................................
®
1-27
.........................................................................................................................................
Summary
Conceptual models are created to model the functional and information needs of a
business. These models may be based on the current needs but can also be a reflection
of future needs. This course is about modeling the information needs. Functional
needs cannot be ignored while modeling data, as these form the only legitimate basis
for the data model. Ideally, the conceptual models are created free of any consideration
of the possible technical problems during implementation. Consequently the model is
only concerned with what the business does and needs and not with how it can be
realized.
Summary
• ER Modeling models information conceptually

• Based on functional business needs
• “What”, not “How”
• Diagrams provide easy means of communication
• Detailed, but not too much
1-30
Entity Relationship modeling is a well-established technique for catching the

information needs. The ER model forms the basis for the technical data model.
Technical considerations take place at that level.
Entity Relationship diagrams provide an easy-to-read and relatively easy-to-create
diagrammatic representation of the ER model. These diagrams initially form the
foundation for the discussion of business needs. Later they provide the best possible
map of a future system.
The diagrams show a fair amount of detail, but are not too detailed to become
cluttered.
.............................................................................................................................................
Practice 1—1: Instance or Entity
.........................................................................................................................................
Practice 1—1: Instance or Entity

Goal (See Page 47)
The goal of this practice is to learn to make a distinction between an entity, an
attribute, and an instance of an entity.
Your Assignment
List which of the following concepts you think is an Entity, Attribute, or Instance. If
you mark one as an entity, then give an example instance. If you mark one as an
attribute or instance, give an entity. For the last three rows, find a concept that fits.
Practice: Instance or Entity?
Concept E/A/I? Example Instance or Entity

PRESIDENT
ELLA FITZGERALD
DOG
ANIMAL
HEIGHT
E CAR
A CAR
I CAR
1-32
.........................................................................................................................................
®
1-29
.........................................................................................................................................
Practice 1—2: Guest

Goal (See Page 47)
The goal of this practice is to recognize attributes for an entity.
Scenario
On the left side of the illustration are three entities that play a role in a hotel
environment: GUEST, HOTEL, and ROOM. On the right is a choice of attributes.
Your Assignment
Draw a line between the attribute and the entity or entities it describes.
Practice: Guest
Address
Arrival Date
Family Name
GUEST
Room Number
HOTEL
Floor Number
ROOM
Number of Beds
Number of Parking Lots
Price
TV set available?
1-33
.............................................................................................................................................
Practice 1—3: Reading
.........................................................................................................................................
Practice 1—3: Reading

Goal (See Page 49)
The goal of this practice is to read a relationship.
Your Assignment
Which text corresponds to the diagram?
Practice: Reading
EMPLOYEE assigned to DEPARTMENT

responsible for
A Each EMPLOYEE may be assigned to one or more DEPARTMENTS

Each DEPARTMENT must be responsible for one or more EMPLOYEES
B Each EMPLOYEE must be assigned to one or more DEPARTMENTS

Each DEPARTMENT may be responsible for one or more EMPLOYEES
C Each EMPLOYEE must be assigned to exactly one DEPARTMENT

Each DEPARTMENT may be responsible for exactly one EMPLOYEE
1-34
.........................................................................................................................................
®
1-31
.........................................................................................................................................
Practice 1—4: Read and Comment

Your Assignment (See Page 49)
1 Read each of the relationships in the model presented here.
2 Next, comment on the relationship you just read. Use your knowledge of people
and towns.
Practice: Read and Comment
PERSON born in TOWN

birthplace of
living in
home town of
visitor of
visited by
mayor of
with mayor
1-35
.............................................................................................................................................
Practice 1—5: Hotel
.........................................................................................................................................
Practice 1—5: Hotel

1 Comment on the relationships of the model presented here.
Practice: Hotel
HOTEL
* Address
the lodging host of
for
ROOM
* Room Number
with
in in guest in
STAY of PERSON
* Arrival Date with * Name
1-36
2 Make up two more possible relationships between PERSON and HOTEL that
might be of some use for the hotel business.
.........................................................................................................................................
®
1-33
.........................................................................................................................................
Practice 1—6: Recipe

Goal (See Page 51)
The goal of this practice is to discover the various types of information that are present
in a given source of information.
Scenario
You work as an analyst for a publishing company that wants to make recipes available
on the Web. It wants the public to be able to search for recipes in a very easy way.
Your ideas about easy ways are highly esteemed.
Your Assignment
1 Analyze the example page from Ralph’s famous Raving Recipes book and list as
many different types of information that you can find that seem important.
Ralph’s Raving Recipes
Soups Açorda alentejana

bread soup from Portugal
vegetarian for 4 persons:
15 min 1 onion
easy 4 cloves of garlic
1 red pepper
1 liter of vegetable broth
4 tablespoons of olive oil
4 fresh eggs
1 handful of parsley or coriander
salt, pepper
9-12 slices of (old) bread
preparation Cut the onion into small pieces and fry together
with the garlic. Wash the red pepper, cut it in
half, remove the seeds and fry it for at least 15
page 127
1-37
2 Group the various types of information into entities and attributes.

3 Name the relationships you discover and draw a diagram.
.............................................................................................................................................
General Instructor Notes
.........................................................................................................................................

Course characteristics
Data modeling is a topic that has been taught by Oracle in many countries for a
considerable time. It is probably a course with the least number of changes and the
longest history in Oracle Education.
It is also a course with many different local flavors that have grown over time. In some
countries the “modeling course” currently is a three-day event, with a lot of emphasis
on database issues, in other countries the “modeling course” is a four-day course and
only concerned with Entity Relationship Modeling.
With this new course we had to provide material for both a three-day class and a four-
day class, with centre of gravity in ER Modeling.
Audience
The audience for an ER modeling course today is different from the audience ten years
ago. Some of Entity Relationship modeling is common knowledge today. That means
this course can go beyond what a modeling course could achieve in the old days. As
there is not more ER modeling to teach, the extension is mostly in the area of more
(and more complex) business contexts, and that reflects changes in businesses of today
as well.
There are three main groups of attendees that form a very heterogeneous audience:
• (Future) database administrators
• (Future) analysts
• (Future) all-round Oracle specialists, focussed on developing applications
Global perspective
The course is intentionally a course with an international, global, perspective.
Examples come from various parts of the world; entities like LANGUAGE and
CURRENCY are made explicit in examples and practices.
Database Design
This course covers the initial transformation of the ER model into relational objects, of
all elements of a ER model. Most of the elements have a clear counterpart in the
physical world; sometimes there is choice, sometimes there is no general relational
translation. This is partially related to relational databases in general, partially to
Oracle in particular.
In the chapter on Design Considerations we cover some of the design considerations
that are specific to an Oracle environment.
.........................................................................................................................................
®
1-35
.........................................................................................................................................
Normalization
Normalization sits uneasily within the context of this course. We had long discussions
about where to position the chapter in the text. Some wanted the topic removed from
this course completely, others wanted it in the modeling part of the book, again others
see it as a design topic. To cut the knot in a pragmatic way we have decided to put the
chapter on Normalization in the appendix. One reason for this decision is that the
material is already quite sizeable without this chapter. Moreover, this gives ample
freedom to ignore it altogether or to teach it at the time you consider best (our
suggestion would be to do it just before the chapter on Denormalization).
Moonlight Coffees
Moonlight Coffees is the name of the fictional global company that forms the central
context in the practices. Many practices focus on various aspects of Moonlight. The
case is built up in many steps. Be aware that you may know much more about
Moonlight than the students do initially. Moonlight is a multinational like Starbucks
and Seattle’s First. Its internal organization resembles the way Oracle is organized.
ElectronicMail
ElectronicMail is the name of another fictional company that forms the major context
in the lessons and electronic presentation. Wherever it made sense we have used this
context to illustrate modeling issues. In addition to that, we used many other smaller
contexts that need no special introduction.
Educational remarks
This Data Modeling and Relational Database Design course is different from the other
courses in the Designer area. This course is about a theory, about a method, about
considerations and not about the use of software. This determines the different nature
of the course.
This course is typically a course that is based on interaction, particularly discussion.
This course, and in particular the part about Entity Relationship modeling, is about
gaining insights, rather than collecting knowledge.
In many situations a student would not be helped at all by being presented with just the
solution to a practice. They would probably even be further away from mastering the
content of a practice (because they were not initially aware of the problem). They
would not go through the mental process of, for example, analyzing, proposing,
weighing-up, rejecting, and refining. This process, however, forms the basis of the
skill of performing an analysis. Giving out a solution is killing this process.
Some wheels have to be reinvented over and over again.
Most practices win in effect, if they are discussed in the group. These discussions can
take place at any time: at the start, at the end, and during the practice.
.............................................................................................................................................
.........................................................................................................................................
Although students start working individually, they usually soon start discussing the
work amongst themselves. After some time you should try to encourage everyone to
participate in the discussion.
Modeling is not a bag of tricks. Modeling is more a state of mind, an attitude. You
learn modeling while you do it, and this is also true for someone who has already done
a lot of modeling. This means that it really helps delegates to do as many practices as
possible during the course when there is an environment for discussion and feedback.
The total of theory on ER modeling is limited. This can be taught in, say, two hours,
but that would only be touching the surface. Applying the modeling technique in real
life (or almost real life) situations is the best way to learn it.
Solutions
Solutions to all practices are supplied at the end of the book. That means that delegates
are able to cheat—and some will do this. There is not much we can do about that.
However, do not view this negatively. When students get stuck, the solutions help to
get them moving again, but they will have learned less than they could have done.
Tell your class cheating is allowed, but let them promise you that they ask you first
before they cheat. That, probably, is the best way to handle it.
.........................................................................................................................................
®
1-37
.........................................................................................................................................
Practices
There are many practices in this course. You are invited to make your selection. Use
practices that you feel comfortable with, where you like the specific context, where
you have additional experience.
Classroom Discussion
We strongly advise you to discuss every practice you do with the whole group. This
will lead to more discussion and this type of discussion is very valuable. These
discussions are probably as near to real life data modeling as you can get in a
classroom situation.
When to Do the Practices

We strongly advise you not to postpone all practices till the end of the lesson (except
for the last lessons). This table below suggests where to place the practices. You can
use most of the suggested mid-lesson practices to check if the students have
understood what you have just taught.
Suggested Use
In the table below practices are classified with one of “Yes”, “Opt” or “Cha”.
Yes means we advise you to do this practice. Most of these are fairly short and are
suggested as mid-lesson practice.
Opt means we suggest to see this practice as optional. We advise you to select at least
one of the optional practices per chapter.
Cha means we think this practice is more challenging. Consider these optional as well.
The challenging practices are mostly open-ended, more likely to end in a discussion
about possible solutions than in “the” solution. These practices require some more
insight and ask to apply things that were taught in other lessons than the last one or
were not yet taught at all. Some may require problem-solving skills.
Ch No Practice When 3 day 4 day

event event
1 1 Instance or Entity After Slide 1-12 Yes Yes
2 Guest After Slide 1-12 Yes Yes
3 Reading After Slide 1-24 Yes Yes
4 Read and Comment After Slide 1-24 Yes Yes
5 Hotel End of Lesson Yes Yes
6 Recipe End of Lesson Cha Cha
2 1 Books After Slide 2-12 Yes Yes
2 Moonlight Coffees After Slide 2-17 Yes Yes
3 Shops After Slide 2-17 Yes Yes
4 Subtypes End of Lesson Yes Yes
5 Schedule End of Lesson Cha Cha
.............................................................................................................................................
Practices
.........................................................................................................................................
Ch No Practice When 3 day 4 day

event event
6 Address End of Lesson Opt Yes
3 1 Read the Relationship After Slide 3-11 Yes Yes
2 Find a Context After Slide 3-19 Cha Cha
3 Name the Intersection Entity After Slide 3-34 Yes Yes
4 Receipt End of Lesson Yes Yes
5 Moonlight P&O End of Lesson Opt Cha
6 Price List End of Lesson Yes Yes
7 E-Mail End of Lesson Opt Cha
8 Holiday End of Lesson Opt Cha
4 1 Identification Please After Slide 4-5 Yes Yes
2 Identification After Slide 4-10 Yes Yes
3 Moonlight UID End of Lesson Opt Yes
4 Tables End of Lesson Opt Cha
5 Constraints End of Lesson Cha Yes
5 1 Shift End of Lesson Yes Yes
2 Strawberry Wafer End of Lesson Opt Yes
3 Bundles End of Lesson Cha Cha
4 Product Structure End of Lesson Opt Cha
6 1 Patterns After Slide 6-17 Yes Yes
2 Data Warehouse End of Lesson Opt Yes
3 Argos and Erats End of Lesson Cha Cha
4 Synonym End of Lesson Opt Cha
7 1 Mapping Supertype End of Lesson Yes Yes
2 Quality Check Subtype Implementation End of Lesson Yes Yes
3 Quality Check Arc Implementation End of Lesson Opt Yes
4 Primary Keys and Columns End of Lesson Opt Yes
8 1 Name that Denormalization End of Lesson Yes Yes
2 Triggers End of Lesson Yes Yes
3 Denormalize Price Lists End of Lesson Opt Opt
4 Global Naming End of Lesson Opt Cha
9 1 Data Types End of Lesson Yes Yes
2 Artificial Keys End of Lesson Opt Cha
3 Product Pictures End of Lesson Opt Opt
Practice slides
We have added slides for most of the practices, wherever it seemed practical. These
often help to do practices as a classroom activity. The practice slides are located at the
end of the presentation. You may want to reposition the practice slides within the
presentation, but we strongly advise you not to do that. By changing the position of the
.........................................................................................................................................
®
1-39
.........................................................................................................................................
slides, you change the slide numbers as they are visible during the show. Students use
the slide numbers as a means to synchronize their book with your presentation and get
very confused if the primary key has changed.
.............................................................................................................................................
Suggested Timing
.........................................................................................................................................
Suggested Timing
If you teach the three- day event, please tell your students that you will only do a
selection of the practices and why.
The schedule is based on a 6 hour teaching day.
Three-Day Event
Day Three-Day Event Approximate
Duration
(min.)
1 Chapter 1 180
Chapter 2 (partial) 180
2 Chapter 2 (rest) 60
Chapter 3 210
Chapter 5 60
Chapter 6 60
Chapter 7 60
Chapter 8 60
Chapter 9 60
The Appendix on Normalization may take 30 minutes.
Four-Day Event
Day Four-Day Event Approximate
Duration
(min.)
1 Chapter 1 180
Chapter 3 300
3 Chapter 4 240
Chapter 5 120
4 Chapter 6 90
Chapter 7 90
Chapter 8 90
Chapter 9 90
The Appendix on Normalization may take 30 minutes.
.........................................................................................................................................
®
1-41
.........................................................................................................................................
Workshop Interviewing
Some of the practices can be done in different ways. In particular, the practice on
Moonlight P&O in Lesson 3 can easily be transformed into a interview session. Other
elements of Moonlight can be used in the same way.
Usually delegates appreciate this kind of session very much, not only because it is fun
to do, but certainly also because it opens their eyes to something that seems to be
straightforward.
This section describes a possible setup for an interview session.
Goal
The goal of this way of working is modest: to let delegates experience some of the
problems that may occur when you have to acquire information through an interview.
Your role
You as the instructor play the role of the head of Moonlight Personnel and
Organization (P&O) Department, Mr. or Mrs. Shadow.
Prepare yourself using the text, and use whatever you know about Oracle’s internal
organization as if it were the Moonlight organization.
During the interviews, volunteer only a little information. Anything the students don’t
ask don’t tell them! Talk about irrelevancies and about specific cases in much detail
After the interviews let the groups work for some time without your interference. Start
moving around doing some quality checks on the model when you see groups think
they are ready. Search for similarities and distinctions between the solutions.
Finally, the groups write their model on a large sheet of paper. Hang these on the wall
for discussion. Ask one or two of the groups (draw lots if they are all eager to do it) to
explain their solution.
After the presentation you should do a short evaluation of what happened. Examples
of things you can focus on:
• The type of questions they asked you
• The questions they did not ask you
• How well they used their time
• If they were carried away by a Moonlight Shadow
• The assumed information in the data models (you know what you told them!)
Delegates’ role
The delegates work in small teams of three or four and play the role of analysts. They
are in the early stages of an analysis and they need to collect information. After data
collection each group makes a model.
.............................................................................................................................................
Workshop Interviewing
.........................................................................................................................................
Instruction for the delegates

This is your chance to practice interviewing. You will meet the head of Moonlight
P&O, Mr. or Mrs. Shadow, and have a first opportunity to ask questions.
You will be divided into groups. One group is randomly selected to start interviewing,
getting 15 minutes interviewing time. The other groups may listen, but may not ask
questions or give comments. The second group gets 10 minutes, the third and other
groups 5 minutes each. Groups will have time to prepare the interview. The groups
should make sure that responsibilities are clear. Who conducts the interview? Who
takes notes? Who guards the time?
After the interviews every group creates a model, only one solution per group, even if
everyone in the group does not agree. After some time (about 20 minutes) you have a
second chance to ask Mr. or Mrs. Shadow two questions per group to clarify what is
still unclear. This interview time is your time. Use it well.
What is needed?
For the delegates
• Pencil
• Eraser
• Large sheets of paper for the presentation
• Felt tip pens
• Adhesive tape
For you
• A cap or hat to wear when you play the role of Mr. or Mrs. Shadow
.........................................................................................................................................
®
1-43
.........................................................................................................................................
Instructor Notes
Topic Timings
Lecture 60 minutes
11
Practice 120 minutes
Introduc tion to Entities, A ttribu tes , and
R elatio nsh ips Total 180 minutes
This chapter introduces entities, attributes and

O verview
relationships, their representation and the
• W h y co n c ep tu al m o d elin g ? mapping information of the database objects
• Intro d u c tio n o f th e K ey role p lay ers :
– E n tities they map to. ER modeling in a mini-nutshell.
– A ttrib u tes
– R e la tio ns h ip s Lesson 1 is a “First Class” chapter to the
“Focus” chapters that follow. We believe this is
a good approach from a educational point of
1 -2
view. Participants get an overall picture in an
early stage of the course and can add details
into that framework during the rest of the
course.
Creating conceptual models is relatively cheap
W hy C reate a C on ceptua l M odel?
as it has a great impact on the efficiency during
• It d e s crib e s ex ac tly th e in fo rm a tio n n ee d s o f th e
b u s in es s
the actual development of a system. For
•
•
It fa cilitates d is cu s s io n
It h e lp s to p rev e nt m ista ke s, m is u n d ers tan d in g
example, far fewer ambiguities, errors, and
•
•
It fo rm s im p o rtan t “id e al sy ste m ” d o cu m en tatio n
It fo rm s a s ou n d ba s is fo r p h ys ica l da tab a se
frustration. At the other hand: far better
•
d e sig n
It is a v ery g o o d p ra c tic e w ith m a n y p rac titio n ers
documentation, better understanding of what
the business really wants, better decision
1 -3
processes. This is, of course, specially true
when you use a Case tool like Oracle Designer
where the transformers and generators do a lot
of the work after analysis.
The house metaphor is a metaphor for a
B etw ee n D ream and R eality...
method, not for modeling only. The analyst is
the person between the dream and reality.
1 -4
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Entity R ela tion ship M o delin g

PRICE LEVEL
# CODE
* DESCRIPTION
defined by applied to
part of
ORGANIZATION
•
o EMAIL
* NAME
M o d e ls b u s in e ss ,
o POSTAL CODE
at o REGION
o STREET
o TOWN parent organization of
o TELEPHONE NUMBER
TITLE MOVIE o CONTACT NAME
# PRODUCT CODE * CATEGORY o CONTACT EXTENSION
* TITLE o AGE RATING
o DESCRIPTION * DURATION the
* MONOCHROME GAME
n o t im plem en tatio n
o AUDIO * CATEGORY SUPPLIER
o PREVIEW * MEDIUM
o MINIMUM MEMORY # SUPPLIER CODE
o EMAIL
available * APPROVED
for * REFERENCE
PRICE HISTORY
the source of
# EFFECTIVE DATE reviewed in available as on
* PRICE
•
* DEFAULT DAYS
* OVERDUE RATE
Is a w e ll-e stab lish e d

OTHER ORGANIZATION
PUBLICATION
# REFERENCE
* TITLE
CATALOG the holder o
tec h n iq u e
# REFERENCE
o VOLUME o CATALOG DATE
o ISSUE o DESCRIPTION CUSTOMER
o PUBLISH DATE managed by o EMAIL
* DESIGNATION
EMPLOYEE the manager of * FIRST NAME
the source of the source of * POSITION * LAST NAME
o OTHER INITIALS
* LAST NAME * STREET
o FIRST NAME * TOWN
o OTHER INITIALS * POSTAL CODE
o EMAIL * REGION
o HOME PHONE
•
acquired fro o WORK EXTENSION
the cancellor of
H a s a ro b u s t sy n ta x
o WORK PHONE
m responsible responsible o PHOTOGRAPH
of o STAFF REMARKS
for for
COPY
* ACQUIRE DATE
* PURCHASE COST
* SHELF CODE
the holder of
o CONDITION responsible for
o CUSTOMER REMARKS
•
MEMBERSHIP TYPE
...
# CODE
R e su lts in e as y -to -
* DESCRIPTION
rented on reserved on * DISCOUNT PERCENTAGE
o STANDARD FEE
held by the type o

held by
re ad d iag ram s … REVIEW

in
# SEQUENCE
* ARTICLE
* HOT
in
of
approved by
MEMBERSHIP
# NUMBER
o TERMINATION REASON
o TERMINATION DATE
of
o AUTHOR
o URL
renewed fo used fo
the reservation for r r
cancelled by
for
requested authorized by of
...a lth ou g h th e y m ay
against
the
BOOKING MEMBERSHIP PERIOD
* BOOK DATE requestor # START DATE
o EXPIRE DATE o ACTUAL FEE PAID
o NOTIFY DATE of
o RESERVE DATE
o STAFF REMARKS
approved by
loo k ra the r c o m p le x
for
fulfilled as
RENTAL
* RENTAL DATE
o STAFF REMARKS
o COMPLETED
a t first s ig h t for
RENTAL ITEM
the rental for
composed of
part of
# LINE NO
* RENTAL PERIOD
* PRICE PAID
o RETURN DATE
o STAFF REMARKS
1 -5
The “only once” criterion is important, maybe

G oa ls o f En tity R ela tion ship M o delin g
even more important than the “all”. You could
• C a ptu re a ll req u ired info rm atio n refer to the experience many delegates will
• Info rm atio n a p p ears o nly on ce
• M o d e l n o in fo rm a tio n th at is d eriv ab le fro m o th er
have when searching for information on the
info rm atio n a lre ad y m o d e led
• Info rm atio n is in a p red ic ta b le , lo g ica l p la ce
Web. There are many Web sites with many
internal information inconsistencies, sometimes
even within one and the same web page. A
1 -6
normal and sensible reaction is to distrust all
information on that site.
A classical picture that tells that an ER model
D atabas e Ty pes
captures business information that can be used
E R M odel
as a basis for any database, as the model tells
something about the business, not about the
implementation. Having said that, in practice
99.99% of the ER models lead to relational
H ierarchical
R elational
N etw ork
database implementations.
1 -7
Mention that an “Entity” was originally called

En tity
an instance of an “Entity Type” which was a
• A n E n tity is : more accurate naming. That is history now. In
– “ S o m e th in g” o f sig n ific an ce to th e b u sin es s
–
a b o u t w h ic h d ata m u s t be k n o w n .
A n am e fo r th e th in g s th at y o u ca n list.
this book we use the commonly-used words
•
– U s u ally a n o u n .
E xa m p le s: o b je cts , e ve n ts
“entity” and “instance”.
• E n tities h av e in stan ce s.
1 -8
.........................................................................................................................................
®
1-45
.........................................................................................................................................
Some entities are simple, tangible concepts, like

E ntities and Instanc es
PERSON or COMPANY.
P ER S O N
PRODUCT
M ah atm a G a nd hi
2 .5 x 35 m m co pp er nail
Some entities are almost tangible such as the
P R O D U C T TY P E
E M P LO Y M E N T C O N T R A C T
n ail
m y previo us co ntrac t
event ELECTION. Some can be surprisingly
JO B v iolinist
S K ILL LE V E L flue nt complex concepts, such as a DOCUMENT
TIC K E T R ES E R VA TIO N ton igh t: H am let in the Ro ya l
P U R C H A SE the C D I b oug h t yes te rd ay VERSION.
E L EC T IO N for pa rliam ent ne xt fall
P R IN TE R P R E FE R E N C E
D O C U M E N T V ER S IO N
…
...
There is no clear line between entity and
1 -9
instance.
Entity can be seen as the name of a set. Set

E ntities and S ets
theory is the basic mathematical concept behind
• A n e n tity re p re se nts a s et o f in sta n ce s th a t are o f
inte res t to a p a rtic u lar b u sin e ss .
ER modeling and Relational Databases.
JO B
m anager
cook
w aitres s
dish w ash er
financial controller
porter
w aiter
piano p layer
1 -1 0
The crucial issue here is the fact that an

A ttribu te
attribute must be single valued at any point in
• A ls o re p re sen ts so m eth in g o f s ig n ifica n ce to th e
b u s in es s
time. If that is not the case, the attribute hides a
•
•
Is a s in g le va lu e d p ro p erty d eta il o f a n en tity
Is a s p ec ific piec e o f in fo rm atio n th a t:
relationship.
– D e sc rib e s
– Q u an tifie s
– Q u alifies
– C las sifie s
– S p ec ifie s
a n e n tity .
1 -1 1
There are, on purpose, some disputable

A ttrib ute E xam ples
attributes here. You could ask delegates to
E n tity A ttrib u te comment on them in the light of the previous
EMPLOYEE F a m ily N a m e , A g e, S h o e S ize,
T o w n o f R e s id en c e, E m a il, ... slide.
C AR M o d e l, W eig h t, C atalo g P rice , …
ORDER
JO B
O rd er D ate, S h ip D ate, …
T itle, D esc riptio n , ...
You might ask delegates to make up more
T R A N S A C T IO N
EM PLOYM ENT
A m ou n t, Tran s ac tio n D ate, …
S tart D a te , S ala ry , ...
attributes per entity.
C O N T R A CT
It is likely the implicit definitions of the entities
1 -1 2
start to move after adding attributes, for
example, when someone suggests Color or
Licence Number as attribute for CAR.
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Practice
Practice: Instan ce or E ntity?
Good practice to discuss jointly. Initially, let the
C o nc ep t E /A /I? E xam ple Ins tan ce or E ntity
P R E S ID E N T students find examples individually.
E LLA FITZG E R A LD
DOG
A N IM A L
H E IG H T
E CAR
A CAR
I CAR
1 -3 2
Practice
Prac tice: G u est
A d d re ss
Good practice to do jointly. Allow the students
A rriva l D a te some time to think about it.
F a m ily N a m e
GUEST
R o o m N u m be r
H O TE L
F lo o r N um b er
RO O M
N u m b e r o f B ed s
N u m b e r o f P a rk in g L o ts
P ric e
T V s et av a ilab le?
1 -3 3
R elatio nsh ips
• A ls o re p re sen t s o m e th in g o f sig n ific an ce to th e

b u s in es s
• E xp re ss h o w en titie s a re m u tu ally re la ted
• A lw ay s e x ist b e tw ee n tw o e n tities (o r o n e en tity tw ice )
• A lw ay s h a v e tw o p e rs p ec tive s
• A re n a m ed at b o th en d s
1 -1 3
Note that the wording is not yet standardized.

R elatio nship Ex am ples
E M P L O Y E E S h av e JO B S
J O B S are h e ld b y E M P L O Y E E S
P R O D U C TS a re cla ss ifie d b y a P R O D U C T T Y P E
P R O D U C T T Y P E is a c la ss ific atio n fo r a P R O DU C T
P E O P L E m ak e T IC K E T R E S E R V AT IO N S
T IC KE T R E S E R V A T IO N S a re m a de b y P E O P L E
1 -1 4
The numerical observations form the basis for

E m ployees h ave Job s
JOB
the optionality and degree of the relationship.
m anager
EM P LO Y E E
S hintaro w aitre ss
cook
dish w as her
The next step is to generalize from the
Jill financial controller
A hm ed
A dam
M aria
w aiter
porter observation: “is it always the case that no
piano p layer
N u m e ric al o b se rv atio n :
employee has more than one job”?
• A ll E M P L O Y E E S h av e a J O B
• N o E M P L O Y E E h as m o re th an on e J O B
• N o t a ll J O B S are h e ld b y an E M P L O Y E E
• S o m e JO B S are he ld b y m o re th an on e E M P L O Y E E
1 -1 5
.........................................................................................................................................
®
1-47
.........................................................................................................................................
Do not mention any details now, but you should

Entity R epres entatio n in D iag ra m
disclose the fact that entities lead to tables.
• D ra w n a s a “s o ftb o x ”
• N a m e sin gu la r
E M P LO Y E E JO B
• N a m e in s id e
ELECTION
• N e ith er s ize ,
n o r p os itio n
h a s a s p ec ia l TIC K E T
m e an in g ORDER
R E S E R V A TIO N
JO B A S S IG N M E N T
D u rin g d es ig n , e n tities u su ally lea d to ta b le s.
1 -1 6
We use an initial capital for attributes

A ttribu tes in D iag ra m s
throughout the book. This is different from
EM PLOYEE JO B
* Fam ily N am e * T itle
what Designer shows. It takes, however, far less
* A dd ress o D e scriptio n
o
o
B irth D ate
S ho e S ize space to print them and it is easier to write in
o E m ail
M an d ato ry a ttrib u te , th at is , kn o w n a n d a va ila b le

longhand.
* for ev ery in stan c e
o O p tio n al attrib u te, th at is , u n k n ow n o r un im p o rtan t
to kn o w fo r so m e in s tan c es
D u rin g d es ig n , a ttrib u te s lea d to c olu m n s.

1 -1 7
The build in this slide shows step by step how

R elationsh ip in D iagram s
the degree of a relationship is represented. The
ex
A n e m p loy ee h as e actly
xac tly o
onne
e jo b
first step of the build turns the word “has” to
E M P L O YE E JO B
italic and green.
has
held by
This is not immediately obvious so point it out
to the class.
Job s a re he ld b y o ne or m o
held re em ploye es
ore
D u rin g d es ig n , re la tio n sh ip s le a d to fo re ig n k ey s.
1 -1 8
Degree of the relationship, from two

Tw o Persp ective s
perspectives.
This is only touched here. More details in the
chapter on Relationships
E M P L O YE E has JO B
held by
1 -2 1
A relationship end (this is Oracle Designer

One Way
jargon) is drawn in three pieces:
• The name, near the entity
• The optionality, near the entity
held by
• The degree, near the opposite entity
1 -2 2
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
The O th er W ay
held by
1 -2 3
The reading formula.

R ea ding a R elatio nsh ip En d
P split in to Q
pa rt o f
“ E a ch“ EPach Pmmust be

m ay be sp lit intosplit
ay be
r m o re Q s ”
o neinoto
e xa ctly o ne Q
o n e o r m ore Q s
”
“ E a ch “QE a chmmuQayst mbb eeu st b e pa rtpao frteofxactly onon

e xactly on e P
e P”
e or m ore P s
”
1 -2 4
Practice
Prac tice: R ead in g
Good practice to do jointly.
E M P LO YE E assigne d to D E PA R T M E N T
responsible for
A E ach E M P LO Y E E m ay be assigned to one or m ore D E P A R TM E N TS

E ach D E P A R TM E N T m ust be resp onsible for one or m ore E M P LO Y E E S
B E ach E M P LO Y E E m ust be assigne d to one or m ore D E P A R TM E N TS

E ach D E P A R TM E N T m ay be respo nsible for one or m ore E M P LO Y E E S
C E ach E M P LO Y E E m ust be assigne d to exactly one D E P A R TM E N T

E ach D E P A R TM E N T m ay be respo nsible for exactly one E M P LO Y E E
1 -3 4
Practice
Prac tice: R ead and C o m m ent
born in
Good practice to do jointly.
PERSON TOW N
birthplace of
Let the delegates read a relationship end aloud.
living in
hom e tow n of This is much more difficult for a novice than
visitor of
visite d by
you and they are likely to be aware of. It helps
m ayor of
to overcome the embarrassment when everyone
w ith m ayor
has some problems.
1 -3 5
It also sets the scene for many joint discussions

and activities.
.........................................................................................................................................
®
1-49
.........................................................................................................................................
Functionality drives data. This course is not

Fu nctions D rive D ata
about function modeling. Nevertheless,
• B u sin es s fu n c tio ns a re alw ay s p re sen t. functions are always in the background. It does
– E xp licit
– A s su m ed not make sense to create a data model when
• B u sin es s fu n c tio ns n e ed da ta.
• A n e n tity , attrib u te , o r re latio n s hip m a y be
m o d ele d b ec au s e:
there is no information about the functions that
– It is u se d b y a b u s in e ss fu nc tio n . define the scope.
– T he b u s in es s ne e d m a y a ris e in th e ne ar
fu tu re .
1 -2 5
This example introduces the idea of types of

W ea th er F orecas t
information. Almost everything on this
-D Q X D U \
. ¡ E H Q K DY Q overview can be named as a particular type of

% UH P HQ
% H U OLQ
information. Some clearly are entities, some
0 QFKHQ
$ P V WH U G D P

clearly are attributes, some are unclear. Let the
% U X [ H OOH V
3 D U LV

delegates make suggestions.

% RU G HD X [
They will be surprised by the number of
1 -2 6 different types of information.
The hidden information is also important. This
DK K øbenhavn
may become clear when the map is shown and
( C openhagen )
IR UK people discover the North/South sorting.
NL B rem en
A m sterdam
B erlin
BE B ruxelles DE
( B russels )
LU M ünchen
P aris ( M unich )
FR
CH
B ordeaux IT
1 -2 7
The model is, of course, only one possible

W eathe r Fo rec ast, a So lution
solution. It is made quite overwhelming on
C IT Y C O U N TR Y
* N am e
o G e ographical
P osition
loc ated in
having
* N am e
o G eographica l
purpose. The simple clipping from the paper
P osition
subject of can contain 5 entities, 10 attributes and 4

ab out
FO R E C A S T referring to W E A TH E R TY P E
relationships. In other words: there is a danger
* D ate * Icon
o M inim u m
Tem perature
o M axim um referring to
referre d in * D escription
W IN D D IR E C TIO N
of underestimating.
Tem perature * Icon
referre d in * D esc ription
o W ind F orce
1 -2 8
This slide is to show that the set of graphical

G raph ical Elem ents of E R D iag ra m elements of ER diagrams is very limited.
E n tity Modeling is not about learning a lot of syntax.
A ttrib u te ** **
R elatio n s h ip
o
Modeling is about understanding business
S u b typ e
U n iq u e id en tifie r
contexts, about analyzing real world things.
A rc
N o n tra n sfe rab ility
#o #
1 -2 9
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Sum m ary
• E R M o de ling m o d els info rm a tio n c o nc e p tu a lly

• B a se d o n fu n ctio n a l b u s in es s ne e ds
• “W h a t”, n o t “H ow ”
• D ia g ra m s p ro vid e e as y m ea n s of c o m m u n ic a tio n
• D e ta ile d , b u t n o t to o m u c h
1 -3 0
Suggested Use of Practices

Prac tices
• Ins tan c e o r E n tity Practice 3day 4day

• G ue s t
• R e ad in g Instance or Entity Yes Yes
• H o tel
• R e cipe Guest Yes Yes
Reading Yes Yes
Read and Comment Yes Yes
1 -3 1 Hotel Yes Yes
Recipe Cha Cha
Practice
Pra ctice : H o tel
Do not explicitly go into the notion of an
H O T EL
* A ddress intersection entity, such as STAY. This will be
the lodgin g host of
fo r
ROOM
covered in Chapter 3.
* R o om N um ber
w ith
in in guest in
S TA Y of P ER S O N
* A rrival D ate w ith * N am e
1 -3 6
Practice
R alph’s R avin g R ecipes
So ups A çord a a len tejan a

b read soup from P ortugal
Allow the students some time (say ten minutes)
vegetarian
15 m in
fo r 4 p erso ns:
1 on ion
to find types of information individually. Next,
easy 4 clo ves of garlic
1 red pepper
1 lite r of vegetable broth
collect the results on the white board, making a
4 tab lespoons of olive o il
4 fre sh eggs
1 ha ndful of parsley or c oriander
long list. Accept whatever is given for now, and
salt, pepper
preparation
9-12 slices of (old) brea d
C ut the onio n into sm all pieces and fry tog ether
discuss later.
w ith the garlic. W ash the red pepper, cut it in
half, rem ove the seeds and fry it for at least 15
page 127
1 -3 7
.........................................................................................................................................
®
1-51
.........................................................................................................................................
.............................................................................................................................................
Instructor Post Lesson Teaching Evaluation
.........................................................................................................................................

Name: Full Email:
Course Name: Data Modeling and Relational Database Design

Lesson number: 1
Number of teaches before filling this form: 1 - 2 - 3 to 5 - >5
Global comments: (circle all that apply)

• Lesson content: Trivial - Too easy - OK - Difficult - Too difficult
• Slide content: Too many slides - Too few slides - O.K
• Text content.Too much text - Not enough text - Unclear - O.K.
• Practice content: Too difficult - Too easy - Problems - O.K.
Detail comments
Content type: Slide - Text - Practice - Instructor notes
Note: 1:needs animation - 2:too much animation - 3:needs more text - 4:too much text
- 5:Unclear - 6:not necessary - 7:Other
Content Page
Note Comments/Suggestions
Type Number
Photocopy this page and fax to: Oracle Designer Education Products @ +(44) 118.924.5181
Additional sheets are available at the end of the instructor’s guide.
If you draw additional diagrams on white board use the Graphic sheet in the Instructor Evaluation
section at the end of this book.
.........................................................................................................................................
®
1-53
.........................................................................................................................................
.............................................................................................................................................
Suggested Graphics
.........................................................................................................................................
Suggested Graphics
Instructor name: Full email:
Course Name:Data Modeling and Relational Database Design

Lesson No: 1 Page No:
Please sketch your additional diagram below.
.........................................................................................................................................
®
1-55
.........................................................................................................................................
Oracle Designer Education Products

Curriculum Development
520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.............................................................................................................................................
2
.................................
Entities and Attributes

in Detail
.........................................................................................................................................
Introduction
Lesson Aim
This lesson provides you with a detailed discussion about entities and attributes and
how you can track these in various sources of information. The lesson looks at the
evolution of an entity definition and the concept of subtype and supertype entity. The
lesson also introduces the imaginary business of ElectronicMail Inc.which is used in
many examples throughout this book.
Schedule
About this slide

See Page 28
• Data compared to information
• Entities and how to track them down
• Attributes
• Subtypes and supertypes
2-2
Topic See Page

Introduction 2
Data Compared to Information 4
Data 5
Tracking Entities 7
Electronic Mail Example 9
Evolution of an Entity Definition 11
Functionality 13
Tracking Attributes 14
Subtypes and Supertypes 17
Summary 20
Practice 2—1: Books 21
.............................................................................................................................................
Introduction
.........................................................................................................................................
Topic See Page

Practice 2—2: Moonlight 22
Practice 2—3: Shops 23
Practice 2—4: Subtypes 24
Practice 2—5: Schedule 25
Practice 2—6: Address 26
Objectives
• Track entities from various sources
• Track attributes from various sources
• Decide when you should model a piece of information as an entity or an attribute
• Model subtypes and supertypes
.........................................................................................................................................
®
2-3
.........................................................................................................................................
Data Compared to Information

About this slide
Data Compared to Information See Page 28
• Data
– Facts given from which other facts may be
inferred
– Raw material
Example: Telephone Directory
• Information
– Knowledge, intelligence
Example: Telephone number of florist
2-3
The words data and information are often used as if they are synonyms. Nevertheless,
they have a different meaning.
Data: Raw material, from which you can draw conclusions. Facts from which you
can infer new facts. A typical example is a telephone directory. This is a huge
collection of facts with some internal structure.
Information: Knowledge, intelligence, a particular piece of data with a special

meaning or function. Often information leads to data. In reverse, information is often
the result of the deriving process from data—this may be a particular piece of data. If
data is structured in some way, this is very helpful in the process of finding
information. To expand the telephone directory data example, information is the
telephone number of your dentist or the home address of a colleague.
.............................................................................................................................................
Data
.........................................................................................................................................
Data
About this slide
Data See Page 28
Data~
• Modeling, Conceptual
Structuring data concepts into logical, coherent,
and mutually related groups
• Modeling, Physical
Modeling the structure of the (future) physical
database
• Base
A set of data, usually in a variety of formats, such
as paper and electronically-based
• Warehouse
A huge set of organized information
2-4
Conceptual Data Modeling

Conceptual data modeling is the examination of a business and business data in order
to determine the structure of business information and the rules that govern it. This
structure can later be used as the basis for the definition of the storage of the business
data. Conceptual data modeling is independent of possible technical implementations.
For that reason, a conceptual data model is relatively stable over longer periods of
time, as businesses change, often only gradually, over a period of time. Conceptual
Data modeling is also called Information Engineering.
Physical Data Modeling

Physical data modeling is concerned with implementation in a given technical
software and hardware environment. The physical implementation is highly dependent
on the current state of technology and is subject to change as available technologies
rapidly change. A technical design made five years ago is likely to be quite outdated
today.
By distinguishing between the conceptual and physical models, you separate the rather
stable from the rather unstable parts of a design. This is true for both data models and
functional specifications.
.........................................................................................................................................
®
2-5
.........................................................................................................................................
Database
A database is a set of data. The various parts of the data are usually available in
different forms, such as paper and electronic-based. The electronic-based data may
reside, for example, in spreadsheets, in all kinds of files, or in a regular data base.
Today, relational databases are very common; but many older systems are still around.
The older systems are mostly hierarchical databases and network databases. Systems
of more recent date are semantic databases and object oriented databases.
Data Warehouse
A data warehouse is composed of data from multiple sources placed into one logical
database. This data warehouse database, (or, more correctly, this database structure), is
optimized for Online Analytical Processing (OLAP) actions.
Often a data warehouse contains summarized data from day-to-day transaction
systems with additional information from other sources. An example is a phone
company that tracks the traffic load for a routing system. The system does not store the
individual telephone calls, but stores the data summarized by hour.
From a data analysis point of view a data warehouse is just a database, like any other,
only with very specific and characteristic functional requirements.
.............................................................................................................................................
Tracking Entities
.........................................................................................................................................
Tracking Entities
The nouns in, for example, the texts, notes, brochures, and screens you see concerning
a business often refer to entities, attributes of entities, or instances of entities.
About this slide

Entities See Page 29
• Give the entity a unique name

• Create a formal description of
the entity
• Add a few attributes, if possible
• Be aware of homonyms
• Check entity names and descriptions regularly
• Avoid use of reserved words
• Remove relationship name from entity name
2-5
Naming an Entity Uniquely

First distinguish an entity by outlining the concept in your mind. Next, try to find a
unique and clear name for an entity. This is not always easy as there are far more
concepts than clear names. Use your imagination. Use a dictionary. Use a combination
of words, use ‘X’ if necessary, but do not let the lack of a good name stop you from
modeling. Good names evolve over time.
Check the names you used every now and then. The implicit definition of an entity
may change during analysis, for instance, as a result of adding an attribute or changing
the optionality of a relationship.
Creating a Formal Description

Create a formal description of the entity. This is usually not difficult and the writing
helps clarify your thinking about what you are talking about. Check this description
regularly. Sometimes concepts evolve during the modeling process. The definitions, of
course, should follow that evolution.
Be Aware of Synonyms
In many business contexts one and the same concept is known under different names.
Select one and mention the synonyms in the description: “...also known as ...”.
.........................................................................................................................................
®
2-7
.........................................................................................................................................
Avoid Homonyms
Often in a business one word is used for different concepts. Sometimes even the same
person will use the same word but with different meanings as you can see in the next
example.
“The data modeling course you attend now was written in 1999 and requires modeling
skills to teach.” In this sentence the word “course” refers to three different concepts: a
course event (like the one you are attending today), a course text (which was written in
1999) and the course type (that apparently needs particular skills).
Avoid Reserved Words

Although you are free to use any name you want for an entity, try to avoid database
and programming terms as entity names if possible. This may prevent naming
problems and confusion later on in the design stage.
Remove Relationship Name from Entity Name

Often you can select entity names in a more or less generic way. In the example, both
diagrams model the same context. In the first the “guest” aspect is part of the entity
name as well as part of the relationship name.
About this slide

See Page 29
GUEST HOTEL
guest of
host of
PERSON guest of
ACCOMMODATION
host of
The second model is more general in its naming. There a guest is seen as a PERSON
playing the role of being a guest.
As a rule, if there is choice take the more general name. It allows, for example, for the
addition of a second relationship between the same entities that shows, for example,
person is working for or is owning shares in the accommodation. The first model
would require new entities.
This subject is closely related to the concept of subtypes and roles. You find more on
this later in this lesson and when we discuss Patterns.
.............................................................................................................................................
Electronic Mail Example
.........................................................................................................................................
Electronic Mail Example

In this course we investigate various business contexts. One is that of ElectronicMail,
a company that supplies an e-mail service. Here is some background information.
About this slide

Some Background Information See Page 29
“ElectronicMail (EM) wants to provide an attractive and user- friendly

Web-based e-mail system. Important concepts are user and message.
An EM user has a unique address of 30 characters at most and a
password supplied by the person who set up the EM user. Who the
person really is, we do not know, although we ask for some additional
information, such as the name, country, birth date, line of business, and
a few more things.
Users must be able to send and receive mail messages. A mail
message is usually a piece of straight text. A message may have
attached files. An attachment is a file, like a spreadsheet, that is sent
and kept with the message, but not created with our software.
Messages are kept in folders. Every user has three folders to start with:
Inbox, Outbox, and Wastebasket. Additional folders can be created by
the user.”
2-7
About this slide

(0
ORJR DGYHUWLVHPHQWDUHD s See Page 29
&RPSRVH &RPSRVH ge
7HPSODWH default
sa
)ROGHUV
es
il m
6XEMHFW test 6HQG
$GGUHVVHV
7R bipi, giovanni_papini@yahoo.com
m a 6DYH'UDIW
3UHIHUHQFHV
e
*HW1HZ0DLO
&F myself
pos 6DYH7HPSODWH
%FF
m
co
&DQFHO
([LW
0HVVDJH
to
this is a test
n
ree
DGYHUWLVHPHQW
WH[W and a text as well

tralalalala
c
fs
pompidom
ho
DUHD
.HHS
ke tc &RS\
s
$WWDFKPHQWV 7\SH
$GG
abc.html Hypertext
6LJQDWXUH
xyz.doc Word document
2-8
.........................................................................................................................................
®
2-9
.........................................................................................................................................
The screenshots may give an idea of how the Compose a Mail Message screen and the
Maintain Addresses screen will look like.
About this slide
(0 DGYHUWLVHPHQWDUHD
See Page 29
ORJR
&RPSRVH $GGUHVVHV s se
s
dre
)ROGHUV
1LFNQDPHV
$GGUHVVHV $OLDV (PDLODGGUHVV

a d
3UHIHUHQFHV
apple
ta in
w.j.appletree@cats.com
bipi
in
sabine_papini @yahoo.com
ma
*HW1HZ0DLO
joe j.suspender@last.com
myself jtiddlywink@em.com
o
nt
([LW
*URXS
ree
DGYHUWLVHPHQW
c
o fs
IULHQGV
h bipi
e tc
DUHD
joe
sk giovanni_papini@yahoo.com
p.g.m.papini@em.com
2-9
About this slide

Some Desired Functionality See Page 30
Practice
“Users of ElectronicMail must be able to address messages to a mail
list, for example, a group of e-mail addresses. The system should only See Page 30
keep one copy of the message sent (to save database space) plus This is a suitable
information about whom the message was sent to.
place to do
Users must be able to create templates for their messages. A template practices 2-1.
must be named and may contain everything a real message contains.
A template may be used for new messages.
Users must be able to reply to a message. By replying the user creates
a new message to which the old message is added.
Users must be able to create an alias for an e-mail address, to hide the
often complex addresses behind an easy-to-remember nickname.”
2-10
.............................................................................................................................................
Evolution of an Entity Definition
.........................................................................................................................................
Evolution of an Entity Definition

To illustrate the evolution of a concept, consider ElectronicMail’s entity MESSAGE.
The first intuitive description of this entity may be:
A message is a piece of text sent by a user.
Any user? Well, no.
A message is a piece of text sent by an EM user.
Must every message contain text? Should it not be possible to send a message that
only transports an attachment, without additional text?
A message is a note that is sent by an EM user. A

message does not necessarily contain text, nor a
subject, etc.
And what about a message that comes from an external source and is received by a EM
user? Should those not be kept as well?
A message is a note that is sent by an EM user or

received by an EM user or both. A message does not
necessarily contain text, nor a subject, etc.
Now suppose a message is sent by an EM user to an external e-mail address only.

Suppose the EM user does not want to keep a copy of the mail message. In that case
there is no need for the system to keep the message as there is no internal EM user that
needs the message. In fact, it is not important at all to keep messages that were sent by
a EM user; only those that were actually received by an EM user are of interest.
A message is a note that is received by an EM user. A

message does not necessarily contain text, nor a
subject, etc.
The thinking process shown here is typical for the change of a definition from the first
idea to something that is much more well thought-out—though this does not mean that
the definition is final.
.........................................................................................................................................
®
2-11
.........................................................................................................................................
Entity Life Cycle

It often helps to make things clear if you think about the life cycle of an entity. The life
cycle refers to the functional steps of the entity. For example, how can the entity
instance come into existence? How can it change? How does it disappear?
In case of entity MESSAGE the questions are:
• When does “something” become a message?
• When does a message change?
• When can a message be removed?
Creating a Message
When I type in some text in the compose screen, is that text a message? You will
probably agree that it does not make much sense to consider it as a message until some
fields are completed, such as the To or Subject field. The checks must take place after
I hit the send key. Only after all checks have been made is the message born.
Removing a Message
When can the system remove a message? When a user hits the delete key? But what
should the system do when there are other receivers of that same message? It is better
to consider the deleting of a message as the signal to the system that you no longer
need the right to read the message. When all users that did receive the same message
have done this, then the message can be deleted. Apparently, for a message to exist it
must have receivers that still need the message.
Changing a Message
Changing a message? As long as the text is not sent, it is no problem as it is not yet
considered to be a message. Changing it after sending it? Changing something that is
history? This cannot be done. Changing the text should lead to a new message.
Draft
What about a message that is not yet ready for sending? Suppose a user wants to finish
a message at a later date. Is there a place for this? Do we want an unsent, or draft,
message in the system? Is a DRAFT a special case of entity MESSAGE, or should we
treat a DRAFT as a separate entity?
Template
What about the templates? A template is about everything a message can be, but a
template is only used as a kind of stamp for a message. Templates are named,
messages are not. Is TEMPLATE a special case of entity MESSAGE, or should we
look upon it as a separate entity?
.............................................................................................................................................
Functionality
.........................................................................................................................................
Functionality
In the previous evolution of the entity definition, the definition changes were invoked
by thinking and rethinking the functionality of the system around messaging. This
illustrates the statement made earlier: functions drive the conceptual data model.
Business Functions
“Users of ElectronicMail must be able to address messages to a mail

list, for example, a group of e-mail addresses. The system should only
keep one copy of the message sent (to save data base space) plus
information about whom the message was sent to.
Users must be able to create templates for their messages. A template
must be named and may contain everything a real message contains.
A template may be used for new messages.
Users must be able to reply to a message. By replying the user
creates a new message to which the old message is added.
Users must be able to create an alias for an e-mail address, to hide
the often complex addresses behind an easy-to-remember nickname.”
2-11
The first idea of the functionality of a system, or desired functionality, can be derived
from the verbs in, for example, descriptive texts and interview notes. In the above text
the functionality is expressed at a high level, without much detail. Nevertheless, you
can probably imagine more detailed functionality.
In this course functionality is always present, often implicitly assumed, sometimes in
detail.
.........................................................................................................................................
®
2-13
.........................................................................................................................................
Tracking Attributes
About the slide
An Attribute... See Page 30
• Always answers “of what?”

• Is the property of entity, not of relationship
• Must be single valued
• Has format, for example:
– Character string
– Number
– Date
– Picture
– Sound
• Is an elementary piece of data
2-13
As discussed earlier, the nouns in, for example, the texts, notes, brochures, and screens
you see used in a business often refer to entities, attributes of entities, or instances of
entities. You can usually easily recognize attributes by asking the questions “Of
what?” and “Of what format?”. Attributes describe, quantify, qualify, classify, specify
or give a status of the entity they belong to. We define an attribute as a property of an
entity; this implies there is no concept of a standalone attribute.
In the background information text on ElectronicMail that is shown below, the first
occurrence of the (probable) entities are capitalized, the attributes are boxed and
instances are shown in italics.
“ElectronicMail (EM) wants to provide an attractive and user friendly About the slide
Web-based email system. Important concepts are user and message. See Page 30
An EM USER has a unique address of 30 characters at most and a
password supplied by the PERSON who set up the EM user. Who the
person really is, we do not know, although we ask for some additional
information, like the name, COUNTRY, birth date, line of business, and
a few things more.
Users must be able to send and receive mail MESSAGES. A mail
message is usually a piece of straight text. A message may have
attached files. An ATTACHMENT is a file, like a spreadsheet, that is
sent and kept with the message, but not created with our software.
Messages are kept in FOLDERS. Every user has three folders to start
with: Inbox, Outbox and Wastebasket. Additional folders can be created
by the user.”
.............................................................................................................................................
Tracking Attributes
.........................................................................................................................................
List the types of information, distinguish the probable entities and attributes and group
them. Add attributes, if necessary, (like Name of COUNTRY) in the example. Distill
one or more attributes from the instances (like Name of FOLDER).
EM Entities and Attributes
Nouns Entities/Attributes/ Entities with their

Instances Attributes
user USER USER
address Address - Address
password Password - Password
person PERSON PERSON
name Name - Name
country COUNTRY - Birth Date
birth date Birth Date - Occupation
occupation Occupation COUNTRY
message MESSAGE - Name
text Text MESSAGE
attachment ATTACHMENT - Text
file File ATTACHMENT
folder FOLDER - Filename
inbox Inbox FOLDER
outbox Outbox - Name
wastebasket Wastebasket
Naming Attributes
Attribute names become the candidate column names at a later stage. Column names
must follow conventions. Try to name attributes avoiding the use of reserved words.
Do not use abbreviations, unless these were decided beforehand. Examples of
frequently-used abbreviations are Id, No, Descr, Ind(icator).
Do not use attribute names like Amount, Value, Number. Always add an explanation
of the meaning of the attribute name: Amount Paid, Estimated Value, Licence No.
Always put frequently-used name components, such as “date” or “indicator”, of
attribute names in the same position, for example, at the end —Start Date, Creation
Date, and Purchase Date.
Do not use underscores in attribute names that consist of more than one word. Keep in
mind that attribute names, like entity names, must be as clear and understandable as
possible.
.........................................................................................................................................
®
2-15
.........................................................................................................................................
Entities Compared to Attributes

Sometimes a piece of information that is an attribute in one context is an entity in
another context. This is purely specific to the business. A typical attribute, like Name,
may need to be modeled as an entity. This happens, for example, when the model
needs an extra dimension, such as the language. If product names must be kept in
several languages and prices must be kept in various currencies, you may suddenly
find one product has several names. For example: “This particular article of clothing is
named ‘Acapulco swimming trunks’ in English, and ‘Akapulko Badehose’ in
German.”
A commonly encountered dimension is time. This is discussed later.
About the slide

Attribute and Entity See Page 31
GARMENT
Name
Price
• Attributes in one model can be entities in another.
GARMENT
CURRENCY PRICE NAME LANGUAGE
2-16
Redundancy
You should take special care to prevent using redundant attributes, that is, attribute
values that can be derived from the values of others. An example is shown below.
Using derivable information is typically a physical design decision. This is also true
for audit type attributes such as Date Instance Created, and User Who Modified.
About the slide

COMMODITY See Page 31
* Name Practice
* Price exclusive VAT
* Price inclusive VAT
See Page 31 This
* VAT % is a suitable place
to do practices 2-2
and 2-3
.............................................................................................................................................
Subtypes and Supertypes
.........................................................................................................................................

Sometimes it makes sense to subdivide an entity X into subtypes. This may be the case
when a group of instances has special properties, such as attributes or relationships
that only exist for that group, or a fixed value for one of the attributes, or when there is
some functionality that only applies to the group. Such a group is called a subtype of
X. Entity X is called the supertype as a consequence. Subtypes are also modeled when
particular constraints apply to the subtype only. This is discussed further in the lesson
on Constraints.
About the slide

See Page 32
A Subtype ...
• Inherits all attributes of supertype

• Inherits all relationships of supertype
• Usually has its own attributes or
relationships or business functions
• Is drawn within supertype
• Never exists alone ADDRESS
• May have subtypes of its own USER
• Is also known as “Subentity”
LIST
2-18
Subtypes have all properties of X and usually have additional ones. In the example,
supertype ADDRESS is divided into two subtypes, USER and LIST. One thing USER
and LIST have in common is an attribute NAME and the functional fact that they can
both be used in the To field when writing a message.
Inheritance
In the next illustration, is a new entity, COMPOSITION, as a supertype of
MESSAGE, DRAFT, and TEMPLATE. The subtypes have several attributes in
common. These common attributes are listed at the supertype level. The same applies
to relationships. Subtypes inherit all attributes and relationships of the supertype
entity.
.........................................................................................................................................
®
2-17
.........................................................................................................................................
COMPOSITION
o Subject
o Cc
o Bcc DRAFT
o Text * Name
MESSAGE TEMPLATE
* Name
Read the diagram as:

Every MESSAGE (DRAFT, or TEMPLATE) is a COMPOSITION
and thus has attributes like Subject and Text. Conversely:
Every COMPOSITION is either a MESSAGE, a DRAFT, or a TEMPLATE
Always More Than One Subtype

Entity relationship modeling prescribes that when an ER model is complete subtypes
never stand alone. In other words, if an entity has a subtype, there should always be at
least a second subtype. This makes sense. What use would there be for distinguishing
between an entity and the single subtype? This idea leads to the two subtype rules.
About the slide

See Page 32
Subtype: Rules
Subtypes of the same entity must be:

• Exhaustive:
Every instance of a supertype is also instance of
one of the subtypes
and
• Mutually exclusive:
Every instance of the supertype is of one and only
one subtype
Name subtypes A
adequately:
B C NON B OTHER A
2-20
.............................................................................................................................................
.........................................................................................................................................
Nested Subtypes
You can nest Subtypes. For readability, you would not usually subtype to more than
two levels, but there is no major reason not to do so. Reconsider the placement of the
attributes and relationships after creating a new level.
About the slide

See Page 33
COMPOSITION OTHER
o Subject COMPOSITION
o Cc * Name
o Bcc DRAFT
o Text *DRAFT
Name
MESSAGE TEMPLATE
TEMPLATE
* Name
Subtypes Always Exist

Every entity can always be subtyped. You can always make up a rule to subdivide the
instances in groups, but that is not the issue. The reason for subtyping should always
be that there is a business need to show similarities and differences at the same time.
About the slide

See Page 33
More on Subtypes
Subtypes always exist...
EMPLOYEE
CURRENT OTHER
EMPLOYEE EMPLOYEE
... but do not all make sense
EMPLOYEE
EMPLOYEE WITH OTHER
SHOE SIZE > 45 EMPLOYEE
2-22
Implementing Subtypes
You can implement subtype entities in various ways, for example, as separate tables or
as a single table, based on the super entity.
.........................................................................................................................................
®
2-19
.........................................................................................................................................
Summary
Entities can often be recognized as nouns in texts that functionally describe a business.
Entities can be tangible, intangible, and events. Subtypes of an entity share all
attributes and relationships of that entity, but may have additional ones.
Summary
• Entities
– Nouns in texts
– Tangible, intangible, events
• Attributes
– Single-valued qualifiers of entities
• Subtypes
– Inherit all attributes and relationships of
supertype
– May have their own attributes and relationships
2-23
Attributes are single-valued elementary pieces of information that describe, qualify,

quantify, classify, specify or give a status of the entity they belong to.
Most entities have attributes.
Every attribute can be promoted to a separate entity which is related to the entity the
attribute initially belonged to. You must do this when you discover that the attribute is
not single valued, for example, when names must be kept in multiple languages or
values in multiple currencies.
.............................................................................................................................................
Practice 2—1: Books
.........................................................................................................................................
Practice 2—1: Books

Goal (See Page 30)
The goal of this practice is to differentiate between various meanings of a word used in
a text.
Your Assignment
1 In this text the word book is used with several meanings. These meanings are
different entities in the context of a publishing company or a book reseller. Try to
distinguish the various entities, all referred to as book. Give more adequate names
for these entities and make up one or two attributes to distinguish them.
1. I have just finished writing a book. It’s a novel about justice and
power.
2. We have just published this book. The hard cover edition is available
now.
3. Did you read that new book on Picasso? I did. It's great!
4. If you like you can borrow my book.
5. I have just started translating this book into Spanish. I use the modern
English text as a basis and not the original, which is 16th century.
6. I ordered that book for my parents.
7. Yes, we have that book available. You should find it in Art books.
8. A second printing of the book War and Peace is very rare.
9. I think My name is Asher Lev is one of the best books ever written.
Mine is autographed.
10. I want to write a book on entity relationship modeling when I retire.
2-25
2 Create an ER model based on the text. Put the most general entity at the top of your
page and the most specific one at the bottom. Fit the others in between. Do not
worry about the relationship names.
.........................................................................................................................................
®
2-21
.........................................................................................................................................
Practice 2—2: Moonlight

Moonlight Coffees
Scenario (See Page 31)
You work as a contractor for Moonlight Coffees Inc. One of your
colleagues, who is a business analyst, has prepared some
documentation. Below you find an extract from the summary
document.
Your Assignment
1 Make a list of about 15 different entities that you think are important for Moonlight
Coffees. Use your imagination and common sense and, of course, use what you
find in the summary that is printed below.
Moonlight Coffees
Summary
Moonlight Coffees is a fast growing chain of high quality coffee shops with currently
over 500 shops in 12 countries of the world. Shops are located at first-class
locations, such as major shopping, entertainment and business areas, airports,
railway stations, museums. Moonlight Coffees has some 9,000 employees.
Products
All shops serve coffees, teas, soft drinks, and various kinds of pastries. Most shops
sell nonfoods, like postcards and sometimes even theater tickets.
Financial
Shop management reports sales figures on a daily basis to Headquarters, in local
currency. Moonlight uses an internal exchange rates list that is changed monthly.
Since January 1, 1999, the European Community countries must report in Euros.
Stock
Moonlight Coffees is a public company; stock is traded at NASDAQ, ticker symbol
MLTC. Employees can participate in a stock option plan.
2-26
2 Write a formal definition of the entity that represents:

– The coffee shops.
– The Moonlight employees.
.............................................................................................................................................
Practice 2—3: Shops
.........................................................................................................................................
Practice 2—3: Shops

Moonlight Coffees
You work as a contractor for Moonlight Coffees. Your task is to
create a conceptual data model for their business. You have
collected all kinds of documents about Moonlight. Below is a page
of a shop list.
Your Assignment
Use the information from the list as a basis for an ER model. Pay special attention to
find all attributes.
Shop List
Moonlight Coffees
Shoplist, ordered to date opened page 4
181 The Flight, JFK Airport terminal 2, New York, USA, 212.866.3410, Airport, 12-oct-97
182 Hara, Kita Shinagawa,Tokyo, JP, 3581.3603/4, Museum, 25-oct-97
183 Phillis, 25 Phillis Rd, Atlanta, USA, 405.867.3345, Shopping Centre, 1-nov-97
184 JFK, JFK Airport terminal 4, New York, USA, 212.866.3766, Airport, 1-nov-97
185 VanGogh, Museumplein 24, Amsterdam, NL, 76.87.345, Museum, 10-nov-97
186 The Queen, 60 Victoria Street, London, UK, 203.75.756, Railway Station, 25-nov-97
187 Wright Bros, JFK Airport terminal 1, New York, USA, 212.866.9852, Airport, 6-jan-98
188 La Lune, 10 Mont Martre, Paris, FR, 445 145 20, Entertainment, 2-feb-98
189
2-27
.........................................................................................................................................
®
2-23
.........................................................................................................................................
Practice 2—4: Subtypes

Goal (See Page 34)
The goal of this practice is to determine correct and incorrect subtyping.
Your Assignment
Find all incorrect subtyping in the illustration. Explain why you think the subtyping is
incorrect. Adjust the model to improve it.
Subtypes
DISABLED CAR
PERSON
STATION WAGON
DEAF
SEDAN
BLIND
OTHER DISABLED
PERSON BUILDING HOUSE
HOTEL DOG
ROOM WITH BATH DOMESTIC
ANIMAL
OTHER ROOM MAMMAL
2-28
.............................................................................................................................................
Practice 2—5: Schedule
.........................................................................................................................................
Practice 2—5: Schedule

Moonlight Coffees
You work as a contractor for Moonlight Coffees.
Your Assignment
Use the schedule that is used in one of the shops in Amsterdam as
a basis for an entity relationship model. The schedule shows, for example, that in the
week of 12 to 18 October Annet B is scheduled for the first shift on Monday, Friday,
and Saturday.
van Gogh, Museumplein, Amsterdam
Schedule Oct 12 - Oct 18 prepared by Janet
Shift Mon Tue Wed Thu Fri Sat Sun

Annet S 2 2 2 1
Annet B 1 1 1
Dennis 2 2 1 2 3
Jürgen 5 4
Kiri 3 4 4
Wil
2-29
The scheme suggests there is only one shift per person per day.
.........................................................................................................................................
®
2-25
.........................................................................................................................................
Practice 2—6: Address

Goal (See Page 34)
The goal of this practice is to sort out various ways of modeling addresses.
Your Assignment
An entity, possibly PERSON (or ADDRESS) may have attributes that describe the
address as in the examples below.
Practice: Address (1/2)
Rheingasse 123 34 Oxford Road

53111 Bonn Reading
Germany Berkshire RG1 8JS
UK
1020 Maple Drive
Kirkland WA 98234
USA
1 How would you model the address information if the future system is required to
produce accurate international mailings?
.............................................................................................................................................
Practice 2—6: Address (continued)
.........................................................................................................................................
Practice 2—6: Address (continued)

2 Would your model from the previous practice also accept the addresses below?
P.O. Box 66708 c/o Mrs Smith

Nairobi Maude Street
Kenya Sandton
Johannesburg 2144
South Africa
3 Check if your model would be different if the system is also required to have
facilities to search addresses in the following categories. Make the necessary
changes, if any.
All addresses:
• In Kirkland
• With postal code 53111 in Bonn
• That are P.O. Boxes
• On:
– Oxford Road or
– Oxford Rd or
– OXFORD ROAD or
– OXFORD RD
in Reading
.........................................................................................................................................
®
2-27
.........................................................................................................................................
Instructor Notes
Instructor Note
Topic Timings
22
Lecture 60 minutes
E ntities and A ttrib utes in D etail
Practice 180 minutes
Total 240 minutes
This chapter addresses how to track entities, the

O verview
importance of naming and clear definitions.
• D a ta c o m p are d to in fo rm a tio n Then attributes are discussed and the thin line
• E n tities an d h o w to trac k th e m d o w n
• A ttrib u te s between entity and attribute. Subtypes are
• S u bty p es a nd su p e rtyp e s
introduced and the reasons for modeling
subtypes. This is followed by a large number of
practices.
2 -2
There is a distinction between data and

D ata C om pared to Info rm a tion
information, although both terms are often used
• D a ta interchangeably.
– F acts g iv en fro m w h ic h o the r fa cts m ay b e
in fe rre d
– R a w m aterial
E xa m p le : T elep h o n e D irec to ry
• Info rm atio n
– K n o w led g e , in te llig e n ce
E xa m p le : T elep h o n e n u m be r of flo rist
2 -3
Some definitions.
D ata
D ata ~
• M o d elin g , C o nc ep tu al
S tru c tu rin g d ata c o n ce p ts in to lo g ic al, co h e re nt,
a nd m u tu a lly re lated g ro up s
• M o d elin g , P h y sic al
M o d elin g th e stru ctu re o f the (fu tu re ) ph y sic a l
d ata ba se
• B as e
A s et o f d ata, u su a lly in a va rie ty o f fo rm ats, s u ch
a s p a p er a n d e lec tro n ica lly -b a sed
• W are h ou s e
A h u g e s e t of o rg a nized in fo rm a tio n
2 -4
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
A working description of an entity. The formal

E ntities
definition is important. Attributes help define
• G ive th e e n tity a u n iq u e n a m e the entity.
• C re ate a fo rm al d es cription o f
th e en tity
• A d d a few attrib u tes , if p os sib le
• B e a w are o f h o m o n y m s
• C h ec k en tity n am e s an d de sc rip tio n s re g u larly
• A v oid u se o f re se rv ed w o rd s
• R e m o ve re la tio ns h ip n a m e fro m e n tity n am e
2 -5
Although it is not always possible to avoid it,

R elation ship N am e in Entity N am e
try not to have the relationship as part of the
GUEST
guest of
HOTEL
entity name. Typical examples that are hard to
host of get rid of:
PERSON gu est of
A C C O M M O D A T IO N EMPLOYEE should be a PERSON that is
host of employed by ...
This is related to the concept of roles.
2 -6
In the lesson texts the company ElectronicMail

S om e B a ckg ro und Inform ation
plays a prominent role. Many examples refer to
“E lectronicM ail (E M ) w ants to provide an attractive and user- friendly
W e b-based e-m ail syste m . Im portant concepts are user and m essage .
A n E M user has a uniq ue address of 30 c haracters at m ost and a
their mail business. The context of e-mail is a
pa ssw ord supplied by the person w ho set up the E M user. W ho the
pe rson really is, w e do not know , although w e ask for som e addition al
info rm ation, such as th e nam e, country, birth da te, line of business, and
very convenient one for this purpose as almost
a few m ore things.
U s ers m ust be able to send and receive m ail m e ssages. A m ail
m e ssage is usually a p iece of straight text. A m e ssage m ay have
every student has a fairly good understanding of
attached files. A n attac hm ent is a file, like a sp readsheet, that is sen t
an d kept w ith the m es sage, but not create d w ith our softw are. e-mail from the perspective of a user. This
M e ssages are kept in folders. E very user has th ree folders to start w ith:
Inb ox, O utbox, and W astebasket. A ddition al folders can be created by
the user.”
context is 100% similar in various places
2 -7
around the world.
More on ElectronicMail. This screen is for

(0
ORJR DGYHUWLVHP HQWDUHD es composing messages and templates, for
ag
&RPSRVH &RPSRVH ss
me
) R OG HUV
$ G GU HVVHV
6 X E MHF W te st
7 HP S OD WH d efault
ma
il 6 HQ G example. You can see the possibility of adding
7R
o se
bipi, giovan ni_ pa pin i@ ya ho o.com 6 D YH ' UD IW
3UH IH UHQ F HV
* HW1 HZ 0 D LO
& F
% FF
m yself
co
mp 6D YH7 HP S OD WH
& D Q F HO
attachments to a mail.
( [LW
0 H VVD J H th isn
to
e isa tex
a test
etra Note the single line field for multiple addresses.
DGYHUWLVHPHQW
and t as we ll
W H[ W
c r lalala la
s
of
pom p ido m
tc h
DUHD
. HH S
sk
e &RS\
$ WWD F K P H Q W V 7 \ S H
As often: buttons represent functionality.
$ GG
a bc.htm l Hyp ertext
6 LJ Q D WX U H
xyz.do c W ord d ocum e nt
2 -8
Screen to create aliases and to create and

(0
ORJR DGYHUWLVHP HQWDUHD maintain user lists. These concepts will be used
es s
&RP SRVH $GGUHVVHV dr
es
) R OG HUV
1 LF N Q D P H V
ad later on in many examples. Make sure
$GGUHVVHV
a in
$ OLD V ( P D LOD G G UHVV
t cats.com
inapini
everybody knows these ideas.
3UH IH UHQ F HV
ap ple w.j.ap pletree@
bipi sab ina e_p @ yaho o.com
o jtidmdlyw ink@ em .com

* HW1 HZ 0 D LO
jo e j.susp en der@ last.com
m yself
t
( [LW
re en
sc
* UR X S
DGYHUWLVHPHQW
of
IULH Q G V
h
etc
bipi
DUHD
jo e
sk giova nni_pa pini@ yaho o.com
p.g.m .pa pin i@ e m .co m
2 -9
.........................................................................................................................................
®
2-29
.........................................................................................................................................
This slide shows clearly the evolution of a

E volutio n of an Entity D efin ition
definition. The initial definition seemed simple
A m es s ag e is a p iec e o f te xt s en t b y a us e r.
A m es s ag e is a p iec e o f te xt s en t b y an E M u se r.
and to the point, but after some iterations the
A m es s ag e is a n ote th at is s en t b y an E M u se r. A
m e ss a g e d o e s n ot n e ce ss arily co n ta in tex t, n o r a
definition seems much better. This process is
s u b jec t, e tc .
A m es s ag e is a n ote th at is s en t b y an E M u se r o r
important.
re ce iv ed b y a n E M u s er o r b o th . A m e ss a g e d o e s n ot
n e ce ss arily co n tain tex t, n o r a su b jec t, etc . Later, during practices, you should often ask
A m es s ag e is a n ote th at is re ce iv ed b y a n E M u s er. A
m e ss a g e d o e s n ot n e ce ss arily co n ta in tex t, n o r a students what their definition is of an entity that
s u b jec t, e tc .
2 -1 2 seems odd, instead of showing what (you think)
is wrong.
Practice
This is quite a complex practice. It forces
1. I have just finished w riting a book. It’s a nove l about justice and
pow er.
2. W e have just publis hed this book. The ha rd cover edition is available
students to think about implicit definitions and
now .
3. D id you read that n ew book on P icasso ? I did. It's great!
4. If you like you can borrow m y book.
the differences between them. A good question
5. I have just started translating this book into S panish. I use the m odern
E nglish text as a ba sis and not the orig inal, w hich is 16th century . to start the discussion: which text lines use
6. I ordered that book for m y parents.
7. Y es, w e have that book available. Y ou should find it in A rt boo ks.
8. A second printing o f the book W ar and P eace is very rare.
“book” for the same concept?
9. I think M y nam e is A sher Lev is one of the best books ever w ritte n.
M ine is autographe d.
10. I w ant to w rite a bo ok on entity relationship m odeling w hen I retire.
2 -2 5
The third bullet: “single-valued” is probably the

A n A ttribu te...
most important. You will often use phrase this
• A lw ay s a n s w e rs “o f w ha t? ” to show that something cannot be an attribute of
• Is th e p ro p erty o f en tity, n o t o f rela tio n s hip
• M u st be s in g le va lu e d a suggested entity
• H a s fo rm a t, fo r e xa m p le :
– C haracter string
– N um ber
– D ate
– P icture
– S ound
• Is a n e lem en tary p ie ce o f d a ta
2 -1 3
Nouns in (spoken) text often lead to

N oun s, En titie s, A ttrib ute s
“E lectronicM ail (E M ) w ants to provide an attractive and user friendly

• Entities
W e b-based em ail system . Im portant concepts a re user and m essage .
A n E M U S E R has a uniq ue address of 30 chara cters at m ost and a
pa ssw ord supplied by the P E R S O N w ho set up the E M user. W ho the
• Attributes
pe rson really is, w e do not know , although w e ask for som e addition al
info rm ation, like the na m e, C O U N TR Y , birth date , line of business, and
a few things m ore. • Instances of entities
U s ers m ust be able to send and receive m ail M E S S A G E S . A m ail
m e ssage is usually a p iece of straight text. A m e ssage m ay have
attached files. A n A TT A C H M E N T is a file, like a spreadsheet , that is
sen t and kept w ith the m essage, but not c reated w ith our softw are.
M e ssages are kept in FO LD E R S . E very user has three folders to sta rt
w ith: Inbox , O utbox and W as tebasket . A dditional fo lders can be created
by the user.”
2 -1 4
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
E M E ntitie s a nd A ttribu tes
N o uns E n tities/A ttributes/ E n tities w ith their

Ins tances A ttributes
u ser U SE R U SE R
ad d ress A d dress - A d d ress
p asswo rd Passw ord - P assw o rd
p erso n PE R SO N PE R SO N
n am e N am e - N am e
co u ntry COUNTRY - B irth D ate
b irth date B irth D ate - O ccup atio n
o ccup atio n O ccu patio n COUNTRY
m essag e M E SS A G E - N am e
text Text M ESS A G E
attach m en t A T TA C H M E N T - T ext
file File A T TA C H M E N T
fo lder FO L D ER - F ilen am e
in b ox Inb ox FO L D ER
o u tb ox O utb ox - N am e
w astebasket W asteb as ket
2 -1 5
If a business is language-dependent, the model

A ttribute and E ntity
will contain many entities that would have been
GARMENT
N am e
P rice
an attribute in a single language environment.
• A ttrib u te s in o n e m o d el c an b e e n titie s in a n o th e r.
When you do an analysis for a system that may
be used in other countries, be aware of this
GARMENT
language “layer”. Currency is another layer, but
CU RRENCY P R IC E NAME LANGUAGE usually of less impact.
2 -1 6
Redundancy usually leaves you with a choice.

R edu ndan cy
Which of the attributes can remain, which is
C O M M O D IT Y
removed?
* N am e
* P ric e exclu sive V AT
* P ric e inclu sive V AT
Of course, redundancy can be more complex
* VAT %
than the example, when several entities are
involved.
2 -1 7
M o on lig h t C o ffees
Practice
S um m ary
M oonlight C offe es is a fast grow ing chain of high quality coffee shops w ith currently
Introduction of Moonlight Coffees, the chain of
over 500 sho ps in 12 countries of th e w orld. S hops are located at first-clas s
locations, such as m ajor shopping, entertainm ent and business areas, airp orts,
railw ay stations , m useum s. M oonlig ht C offees has som e 9,000 em ployees. coffeeshops that we will visit in many practices.
P roducts
A ll shops serve coffees, teas, soft d rinks, and various k inds of pastries. M ost sho ps
sell nonfoo ds, like postcards and som etim es even theater tick ets.
This text provides a general overview, shows
Financial
S hop m ana gem ent reports sales fig ures on a daily bas is to H eadquarters, in local
some of the business areas. List the entities on
currency. M oon light uses an interna l exchange rates lis t that is changed m onthly.
S ince Janua ry 1, 1999, the E urope an C om m unity coun tries m ust report in E uros.
S tock
the white board and leave it there for the rest of
M oonlight C offe es is a public com p any; stock is traded at N A S D A Q , ticker sym b ol
M LTC . E m ploye es can participate in a stock option pla n. the week. During the next days you can mark
2 -2 6
the entities that have been investigated more
closely and the entities that have been replaced
by others.
.........................................................................................................................................
®
2-31
.........................................................................................................................................
Practice
Sh op List
Straightforward practice.
M oo n lig h t Co ffee s
Shoplist, ordered to date opened page 4
181 T he Flight, JF K Airport term inal 2, N ew York, U SA , 212.866.3410, Airport, 12-oct-97

182 H ara, Kita Shinagaw a,Tokyo, JP, 3581.3603/4, M useum , 25-oct-97
183 P hillis, 25 Phillis R d, A tlanta, U SA , 405.867.3345, Shopping C entre, 1-nov-97
184 JF K , JF K A irport term inal 4, N ew Y ork, U S A, 212.866.3766, Airport, 1-nov-97
185 V anGogh, M useum plein 24, Am sterdam , N L, 76.87.345, M useum , 10-nov-97
186 T he Q ueen, 60 V ictoria Street, London, U K, 203.75.756, R ailw ay S tation, 25-nov-97
187 W right Bros, JFK Airport term inal 1, N ew York, U SA , 212.866.9852, A irport, 6-jan-98
188 La Lune, 10 M ont M artre, Paris, F R , 445 145 20, E ntertainm ent, 2-feb-98
189
2 -2 7
Be aware that the word “inheritance” is a

A Sub type ...
loaded term in Object Modeling.
• Inh e rits all a ttribu te s o f s u p erty pe
• Inh e rits all re la tion s h ip s o f su p e rtyp e
• U s ua lly h a s its o w n attrib u tes o r
re la tio n sh ip s o r b u s in es s fun ctio n s
• Is d raw n w ith in s up ertyp e
• N e ve r e xists a lo n e A DD RE S S
• M ay h a ve s u bty p es o f its o w n USER
• Is a ls o kn o w n as “S u be n tity”
L IS T
2 -1 8
S ubtyp e: E xam ple
C O M P O S ITIO N
o S ubject
o Cc
o B cc D R A FT
o Text * N am e
MESSAGE TE M P LA TE
* N am e
2 -1 9
Often the second subtype is called NON-B.

Sub typ e: R ules
This naming is fine as long as there are only
S u b ty p es o f th e s a m e en tity m u st b e :
• E xh a u stiv e: two subtypes. When there is a third, the name
E ve ry in s ta n ce o f a su p e rty p e is also in s tan c e o f
and
o n e o f th e s u b ty p es
should be replaced by, for example, OTHER A.
• M u tu ally e xc lu s ive :
E ve ry in s ta n ce o f th e s u p erty p e is o f o n e an d on ly
o n e s ub ty p e
Na m e s u bty p es A
ad e q ua tely:
B C NON B O TH E R A
2 -2 0
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
There is no formal limit to the number of

Sub typ es: T hree Le vels
subtypes you can or should use, but more than
the 3 levels is for most people difficult to
C O M P O S ITIO N O TH E R
o S ubject
comprehend. The models should help you to
C O M P O S IT IO N
o Cc
o B cc
o Text
* Name
D R A FT
*DNRam e
A FT
understand the business.
MESSAGE TE
TEMMPPLA
LATE
* N am e
TE
Note the move of attribute Name during the
build of the slide. Introducing subtypes usually
2 -2 1 means a reshuffle of attributes (and a
redirection of relationships).
The rule is: model subtypes not because they
M ore on S ubtype s
exist, but because there is a business need to
S u b ty p es a lw a ys e xis t... distinguish them.
E M P L O YE E
CURRENT O TH E R
E M P LO Y E E E M P LO Y E E
... b u t d o n o t all m a ke s en s e
EM P LO Y E E
E M P LO Y E E W ITH O TH E R
S H O E S IZE > 4 5 E M P LO Y E E
2 -2 2
Sum m ary
• E n tities
– N o u n s in te xts
– T an g ib le , in ta n g ib le, e ve n ts
• A ttrib u te s
– S in g le -valu ed q ua lifie rs o f en titie s
• S u bty p es
– In h erit all attrib u tes a nd relatio n sh ip s o f
s u p erty p e
– M a y h av e th eir o w n a ttrib u tes an d rela tio n sh ips
2 -2 3
Suggested use of practices

Prac tices
• Books Practice 3day 4day

• M o o n lig h t C o ffe es
• S h op s Books Yes Yes
• S u bty p es
• S ch e d u le Moonlight Coffees Yes Yes
• A d d re ss
Shops Yes Yes
Subtypes Yes Yes
2 -2 4 Schedule Cha Cha
Address Opt Yes
.........................................................................................................................................
®
2-33
.........................................................................................................................................
Practice
S ubtyp es
D IS AB LE D C AR
You’ll find that the HOTEL example, in
P ER S O N
DEAF
S TA TIO N W A G O N
SEDAN
particular, is difficult. The DOG example is
B LIN D
O TH E R D IS A B LE D
more difficult to explain.
PERSON B U IL D IN G HOUSE
H O TE L DOG
R O O M W ITH B A TH D O M E S TIC
A N IM A L
O TH E R R O O M MAMMAL
2 -2 8
Practice
When the quicker working students have
van G ogh, M useum plein, A m s terdam
S ched ule O c t 12 - O c t 18 prepared by J an e t

completed this and while others are still
A n ne t S
S hift M on Tu e W ed
2
T hu Fri
2
S at
2
S un
1
working on it, ask if the model would change
A n ne t B
D e nn i s
1
2 2 1 2
1
3
1
when someone could do both shift 2 and 3 on a
J ü rg e n
K i ri 3 4
5
4
4
particular day.
Wil
2 -2 9
Practice
Practice: A ddre ss ( 1 /2 )
Every analyst will run into this address
problem. It has been solved many times in
R he in gasse 123
53 111 Bo nn
G erm a ny
3 4 O xfo rd R o ad
R e ading
B erkshire R G 1 8 JS
many different ways (some of which should be
10 20 M aple D rive
UK
on the black list of awful solutions!).
K irklan d W A 982 34
USA
2 -3 0
Practice
Practice: A ddre ss ( 2 /2 )
Did the previous solution include PO boxes and
other uncommon but not impossible addresses?
P .O . Box 667 08 c/o M rs S m ith
N airo bi M aude S tree t
K eny a S an dton
Jo hann esbu rg 214 4
S ou th A frica
2 -3 1
.............................................................................................................................................
.........................................................................................................................................

Name: Full Email:

Lesson number: 2

Detail comments
Content Page
Type Number
.........................................................................................................................................
®
2-35
.........................................................................................................................................
.............................................................................................................................................
Suggested Graphics
.........................................................................................................................................
Suggested Graphics

.........................................................................................................................................
®
2-37
.........................................................................................................................................

520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.............................................................................................................................................
3
.................................
Relationships
in Detail
.........................................................................................................................................
Introduction
Lesson Aim
This lesson discusses in detail how to establish a relationship between two entities.
You meet the ten types of relationship and examples of the less frequent types. This
lesson looks at nontransferable relationships and discusses the differences and
similarities between relationships and attributes. It also provides a solution for the
situation where a relationship seems to have an attribute.
Schedule
About the slide

See Page 38
• Relationships
• Ten different relationship types
• Nontransferability
• Relationships that seem to have attributes
3-2
Topic See Page

Introduction 2
Establishing a Relationship 4
Relationship Types 9
Relationships and Attributes 16
Attribute Compared to Relationship 18
Relationship Compared to Attribute 19
m:m Relationships May Hide Something 20
Resolving Relationships 25
Summary 28
Practice 3—1: Read the Relationship 29
Practice 3—2: Find a Context 30
.............................................................................................................................................
Introduction
.........................................................................................................................................
Topic See Page

Practice 3—3: Name the Intersection Entity 31
Practice 3—4: Receipt 32
Practice 3—5: Moonlight P&O 33
Practice 3—6: Price List 35
Practice 3—7: E-mail 36
Practice 3—8: Holiday 37
Objectives
• Create a well-defined relationship between entities
• Identify which relationship types are common and which are not
• Give real-life examples of uncommon relationship types
• Choose between using an attribute or a relationship to model particular
information
• Resolve a m:m relationship into an intersection entity and two relationships
• Resolve other relationships and know when to do so
.........................................................................................................................................
®
3-3
.........................................................................................................................................
Establishing a Relationship
About the slide
Establishing a Relationship See Page 38
• Determine the existence of a relationship

• Choose a name for the relationship from both
perspectives
• Determine optionality
• Determine degree
• Determine nontransferability
3-3
Determining the Existence of a Relationship

• Ask, for each of your entities, if it is somehow related to one or more of the entities
in your model, and, if so, draw a dotted “skeleton” relationship line.
• Usually all entities in a model are related to at least one other entity. Exceptions are
rare, but they do exist.
• Two entities can be related more than once. For example, in the ElectronicMail
system there are two relationships between entities MESSAGE and USER, one is
about who is sending a MESSAGE and one about who receives a MESSAGE.
• An entity can be related to itself. This is called a recursive relationship. For
example, a MESSAGE can be a reply to another MESSAGE. See the paragraph on
recursive relationships for more details on this.
About the slide

USER See Page 38
MESSAGE sending
receiving
replying
.............................................................................................................................................
.........................................................................................................................................
Choosing Name for the Relationship

• Sometimes the relationship name for the second perspective is simply the passive
tense of the other one, such as is owner of and is owned by. Sometimes there are
distinct words for both concepts, such as parent of / child of or composed of / part
of.
• Try to use names that end in a preposition.
• If you cannot find a name, you may find these relationship names useful:
– Consists of / is part of
– Is classified as / is classification for
– Is assigned to / is assignment of
– Is referred to / referring to
– Responsible for / the responsibility of
• Sometimes a very short name is sufficient, for example, with, in, of, for, by, about,
at, into.
About the slide

See Page 39
Relationship Names
sender
MESSAGE sent by USER
of
sent to
receiver
reply of of
replied
to by
3-5
Are sent to and receiver of really opposite? If so, the assumption is that if a
MESSAGE is sent to a USER, it also arrives. Maybe it is safer to name the
relationship received by / receiver of...
.........................................................................................................................................
®
3-5
.........................................................................................................................................
Determining Optionality of Both the Relationship Ends

• Answer the questions:
– Must every MESSAGE be sent by a USER?
– Must every USER be sender of an MESSAGE?
– Must every MESSAGE be sent to a USER?
– Must every USER be addressed in a MESSAGE?
When an answer is Yes the relationship end is mandatory, otherwise it is optional.
• Be careful at this point. Often a relationship end seems to be mandatory, but
actually it is not. In the ElectronicMail example it seems that every MESSAGE
must be sent by a USER. But a MESSAGE that was sent by an external user to an
internal USER has no relationship to a USER, unless the system were to keep
external users as well.
• Sometimes a relationship is ultimately mandatory, but not initially. Such a
relationship should be modeled as optional.
Optionality
author USER
MESSAGE written by of
received by
receiver
reply of of
replied
to by
3-7
.............................................................................................................................................
.........................................................................................................................................
Determining Degree of Both the Relationship Ends

• Answer the questions:
– Can a MESSAGE be written by more than one USER?
– Can a USER be author of more than one MESSAGE?
If the answer is No the degree is called “1”.
If the answer is Yes the degree is called “many” or just “m”.
• This must be determined for all relationship ends.
• Note that a mandatory “many” relationship end from A to B does not mean that it
is mandatory for A to be split into more than one B. One B is fine. Read it as:
every A must be split into at least one B.
About the slide

A split into B See Page 39
part of
• An optional “many” relationship end means zero, one or more. In the e-mail
example a USER can be author of 0,1 or more MESSAGES.
• Sometimes the degree is a fixed value, or there is a maximum number. Assume a
MESSAGE may be containing one or more ATTACHMENTS, but for some
business reason, the number of ATTACHMENTS per MESSAGE may not exceed
4. The degree then is <5. The diagram, however, shows a crowsfoot.
Practice
Degree See Page 39
This is a suitable
place to do
practice 3-1.
author
MESSAGE written by of USER
received by
receiver
reply of of
containing replied
to by
with <5
ATTACHMENT
3-10
.........................................................................................................................................
®
3-7
.........................................................................................................................................
Determine Nontransferability of Both the Relationship Ends

• When a MESSAGE is created, the USER who is the author of the MESSAGE is a
fact. It would be strange if a mail system allowed you to change the author after the
MESSAGE is completed.
• Often relationships have the following property: you cannot change the
connection, once made. That property is called nontransferability.
Nontransferability leads to nonupdatable foreign keys. Nontransferability is shown
in the diagram with a little diamond shaped symbol through the line of the
relationship end.
About the slide

Nontransferability See Page 40
FOLDER
containing
filed in
author
MESSAGE written by USER
of
received by
receiver
reply of of
replied
to by
3-12
• Not all relationships are nontransferable. Assume the mail system allows a user to
file a MESSAGE in a FOLDER. This is only a valuable functionality if the user is
allowed to change the FOLDER in which a MESSAGE is filed.
.............................................................................................................................................
Relationship Types
.........................................................................................................................................
Relationship Types
There are three main groups of relationships, named after their degrees:
• One to many (1:m)
• Many to many (m:m)
• One to one (1:1)
This paragraph discusses the various types and gives some examples of their variants.
Relationships—1:m
The various types of 1:m relationships are most common in an ER Model. You have
seen several examples already.
About the slide

Relationship Types See Page 40
1:m
(a)
(b)
(c)
(d)
3-13
a Mandatory at both ends. This type of relationship typically models entities that
cannot exist without each other. Often the existence of mandatory details for a
master is more wishful thinking than a strict business rule. Often the
relationship expresses that an entity is always split into details. Seen from the
other perspective, it often expresses an entity that is always classified,
assigned.
.........................................................................................................................................
®
3-9
.........................................................................................................................................
Circumventing Mandatory 1 to Mandatory m Usually you would try to avoid

relationship type (a) in favor of type (b), by taking a different perspective on the
subject. For example, suppose an order is defined as something with at least one
order item. In other words, an order is regarded as a composed concept. You can
avoid modeling order as an entity as you can decide to model a slightly different
concept instead, say ORDER HEADER. Next, define an ORDER HEADER to
have zero, one or more ORDER ITEMS. An order would then be a thing
composed of two entities: any ORDER HEADER with one or more ORDER
ITEMS. Empty headers would not be considered to be an order.
Why Circumvent? Implementing a 1:m relationship that is mandatory at both

ends causes technical problems. In particular it is difficult to make sure details
exist for a newly-created record. In most relational database environments it is
even impossible.
b Optional 1: mandatory m. This is a very common type of relationship, together
with (d). Normally, at least 90% all relationships are of type (b) and (d). The
relationship expresses that the entity at the 1-end can stand alone, whereas the
entity at the many end can only exist in the context of the other entity.
c Mandatory 1: optional m. This is not common. You will see it only when the
relationship expresses that an entity instance only exists when it is a non-empty
set, and where the elements of the set can exist independently. In the example
below a PRODUCT may be part of one BUNDLE. According to the model, a
BUNDLE is of no interest if it is empty.
PRODUCT
part
of BUNDLE
consists
of
d Optional at both ends. See remarks for (b).
.............................................................................................................................................
Relationship Types
.........................................................................................................................................
Relationships—m:m
The various types of m:m relationships are common in a first version of an ER Model.
In later stages of the model most m:m relationships, and possibly all, will disappear.
About the slide

m:m
(e)
(f)
(g)
3-15
e Mandatory at both sides is very uncommon in normal circumstances. This

relationship seems to mean that an entity instance can only be created if it is
immediately assigned to an instance of the other entity, and vice versa. But
how can this occur when we do not have an instance of either entity? Enforcing
the mandatory rule from scratch leads to a conflict.
The relationship can, however, be part of a model of a theoretical nature, like
the mathematical: a LINE always consists of many POINTS and a POINT is
always part of many LINES. It can also describe an existing situation: a
DEPARTMENT always has EMPLOYEES and an EMPLOYEE is always
assigned to a DEPARTMENT. Here the question may arise if it is guaranteed
that the situation will always remain this way.
A m:m relationship that is mandatory at both sides can occur when the
relationship is part of an arc. See the lesson on Constraints for more details.
.........................................................................................................................................
®
3-11
.........................................................................................................................................
f Mandatory at one end is not uncommon in early versions of a model although

they usually disappear at a later stage.
About the slide

See Page 41
USER
part
of LIST
consists
of
g Optional at both ends is common in early versions of a model. These also

usually disappear at a later stage.
.............................................................................................................................................
Relationship Types
.........................................................................................................................................
Relationships—1:1
Usually you will find just a few of the various types of 1:1 relationships in every ER
Model.
About the slide

1:1
(h)
( i)
(j)
3-17
h A 1:1 relationship, mandatory at both ends, tightly connects two entities: when
you create an instance of one entity there must be exactly one dedicated
instance for the other simultaneously; for example, entity PERSON and entity
BIRTH. This leads to the question why you want to make a distinction between
the two entities anyway. The only acceptable answer is: only if there is a
functional need.
If you have this relationship in your model, it is often, possibly always, part of
an arc.
i Mandatory at one end is often in a model where roles are modeled, for
example, in this hospital model.
About the slide

PERSON acting as PATIENT See Page 41
* Name role of * Blood Type
acting as EMPLOYEE
role of * Job
Note: These role-based relationships are often named is / is type of or simply

is / is.
.........................................................................................................................................
®
3-13
.........................................................................................................................................
Both PATIENT and EMPLOYEE are roles played by a PERSON. The attribute
BLOOD TYPE is, according to this model, only of interest when this person is
a PATIENT. Note that PATIENT and EMPLOYEE cannot be modeled as
subtypes of PERSON, as a PERSON may play both roles. You meet the
concept of roles again in a later lesson.
j Optional at both ends is uncommon. However, they can occur, for example,
when there is a relationship between two entities that are conceptually the
same but exist in different systems. An example of this is entity EMPLOYEE
in one system and entity PERSON in a different, possibly a third-party, system.
Many 1:1 relationships (of all three variants) do occur when some of the
entities represent various stages in a process, such as in the next example.
Relationship names in this case can always be leads to / result of or something
similar.
Practice
See Page 42
MESSAGE
DRAFT
This is a suitable
place to do
basis for
practice 3-2.
result of
If you consider a person to be a process as well, the earlier example of BIRTH

and PERSON fit nicely into this general idea.
.............................................................................................................................................
Relationship Types
.........................................................................................................................................
Redundancy
Like attributes, relationships can be redundant.
About the slide

Redundant Relationships See Page 42
COUNTRY COUNTRY
location of location of birth
of of located of
located
in in
TOWN TOWN
hometown hometown
of living living of living born
in in in in
PERSON PERSON
3-20
In the left-hand example the relationship from PERSON to COUNTRY can be derived
from the other two relationships and should be removed from the model.
This is a semantic issue and cannot be concluded from the structure alone, as the right-
hand example shows.
.........................................................................................................................................
®
3-15
.........................................................................................................................................
Relationships and Attributes

Attributes can hide a relationship. In fact, any attribute can hide a relationship.
About the slide

Relationships and Attributes See Page 43
• An attribute can hide a relationship

• Relationship can be “downgraded” to attribute
ATTACHMENT TYPE
* Name
of
with
ATTACHMENT
* Type ATTACHMENT
* Content * Content
3-21
In the example, attribute TYPE of entity ATTACHMENT can be replaced by an entity

ATTACHMENT TYPE plus a relationship from ATTACHMENT to ATTACHMENT
TYPE.
You would have no choice other than to model it this way as soon as you need to keep
extra attributes for ATTACHMENT TYPE. But, if there are no important attributes for
ATTACHMENT TYPE to keep other than the Name of the type, you could consider
removing the entity and take Type as an attribute of ATTACHMENT.
You could also consider using the left-hand option when the number of types is a fixed
and small amount, such as in the context of a chain of hotels where there are only three
types of rooms: single, double, and suite.
.............................................................................................................................................
Relationships and Attributes
.........................................................................................................................................
Attribute Compared to Relationship
• Easy model • Value control

• Fewer tables • List of values
• No join • Other relationships
ATTACHMENT TYPE
* Name
of
with
ATTACHMENT
* Type ATTACHMENT
* Content * Content
3-22
The table based on entity ATTACHMENT would contain the same columns in both
situations, but the Attachment Type Name column would be a foreign key column in
the second implementation. This would mean that an Attachment Type Name entered
for an ATTACHMENT can only be taken from the types listed in the table based on
entity ATTACHMENT TYPE. The list serves as a pick list and spelling check.
There are advantages and disadvantages for both models.
The one entity model is somewhat easier to read because it is less packed with lines. In
the table implementation you would need no joins to get the required information.
However, a two-entity model is usually far more flexible. It leaves the option open to
create relationships from other entities to the new entity. You would have control over
the values entered as they are checked against a given set. Usually, the two-table
implementation takes less (sometimes even much less) space in the database.
Use your common sense when you select the attributes and entities.
About the slide

NAME EMPLOYEE JOB See Page 43
* Id
SALARY BADGE
GENDER NATIONALITY
TEAM ADDRESS
.........................................................................................................................................
®
3-17
.........................................................................................................................................
Attribute Compared to Relationship

About the slide
Attribute Compared to Relationship See Page 43
• There is no such thing as a foreign key attribute.

• Usually, the attribute name should not contain an
entity name.
FOLDER
* Name
containing
placed in
MESSAGE
* Message Id
* Text
* Folder Name
3-24
Nonexistence of Foreign Key Attributes

Be aware of foreign key attributes such as attribute Folder Name of entity MESSAGE
in the example. In ER modeling there is no such thing as a foreign key attribute. The
future foreign key is represented by the relationship between MESSAGE and
FOLDER. A foreign key column (or columns) will result from the primary unique
identifier of the entity FOLDER. See the lesson on CONSTRAINTS for more details
on unique identifiers.
No Entity Name in Attribute Name

When an attribute name contains an entity name, it usually comes from one of the
following situations:
• The attribute hides a relationship to an entity, as in the above example. The second
entity was probably added in a later stage.
• The attribute hides an entity. A typical example is an attribute Employment Date of
entity EMPLOYEE. This might hide the entity EMPLOYMENT, as there is
probably no rule that an employee may be employed by the same company only
once.
• The entity name in the attribute name is redundant. A typical example is attribute
Message Id of entity MESSAGE. The name “Id” would suffice.
• The attribute is the result of a one-to-one relationship that is not modeled, for
example, attributes Birth Date and Birthplace of entity EMPLOYEE. These are in
fact attributes of an entity BIRTH that is not (and probably will never be) modeled.
.............................................................................................................................................
Relationship Compared to Attribute
.........................................................................................................................................
Relationship Compared to Attribute

About the slide
Relationship Compared to Attribute See Page 44
MESSAGE addressed to USER

addressee of
MESSAGE USER
* Addressee
MESSAGE addressed to USER

o Addressee
addressee of
3-25
Sometimes a piece of information looks like a relationship between entities, but

actually is not a relationship.
In ElectronicMail’s Compose Message screen there is a field labeled “To” where the
user is supposed to enter the names of the addressees. Initially you may want to model
that as a relationship addressed to / addressee of between MESSAGE and USER, but
this is a questionable approach. If a message is sent to an external user would it make
sense for ElectronicMail to keep track of all external user addresses that were used to
send messages to, just for the sake of maintaining the relationship? Would this be
possible?
In this case it would be a better choice to see the Addressee as an attribute of the
MESSAGE. This attribute may contain a value that is also known as a USER. In other
words, entity USER contains only suggestions for addressees.
Another possibility is to do both—model an optional relationship and an optional
attribute that cooperatively handle the addressee. An extra constraint (which cannot be
shown in the diagram) must then make sure that at least one of the attribute and the
relationship is actually given a value for a MESSAGE.
.........................................................................................................................................
®
3-19
.........................................................................................................................................
m:m Relationships May Hide Something

During the process of modeling you will find many relationships to be of type m:m.
Often this is a temporary thing. After you have been able to add more details to the
model, a lot of the m:m relationships will disappear as, after consideration, they
simply do not model the business properly.
A typical example is about the CUSTOMER/PRODUCT relationship.
About the slide

m:m Relationships May Hide Something See Page 44
CUSTOMER PRODUCT
* Id buyer of * Code
* Name * Name
bought by
3-26
Suppose you make a model for a retail company that sells PRODUCTS. A
CUSTOMER buys PRODUCTS. Suppose future customers are accepted into the
system as well. This would mean:
A CUSTOMER may buy one or more PRODUCTS
A PRODUCT may be bought by one or more CUSTOMERS
A typical event for this company would be customer Nick Sanchez buying two shirts.
“Nick Sanchez” is a CUSTOMER Name, “shirt” is a PRODUCT Name. This leaves
the question of where to put the “two”, the quantity information.
.............................................................................................................................................
.........................................................................................................................................
Quantity is Attribute of ...
CUSTOMER
? PRODUCT
buyer of
* Id * Code
* Name bought by * Name
Quantity
CUSTOMER
? PRODUCT
buyer of
* Id * Code
Quantity
3-27
It is clear that Quantity is neither a property of CUSTOMER nor of PRODUCT.

Quantity seems to be an attribute of the relationship between CUSTOMER and
PRODUCT.
CUSTOMER PRODUCT
buyer of
* Id * Code
Quantity
Relationships do not and cannot have attributes. Apparently an entity of which

quantity is a property is missing. For that reason we need to change the model. Entity
ORDER (or SALE or PURCHASE) enters the scene.
.........................................................................................................................................
®
3-21
.........................................................................................................................................
About the slide

New Entity ORDER See Page 45
CUSTOMER
* Id with
* Name of ORDER
PRODUCT with *Quantity Sold

* Code
for
* Name
CUSTOMERS PRODUCTS ORDERS

Id Name Code Name Ctr_id Pdt_code Quantity_sold
1 Sanchez 1 Jeans 1 2 2
2 Lowitch 2 Shirt 1 3 2
3 Yomita 3 Tie 2 2 1
4 4 3
3-29
The table design here is the default design for implementing the model. Note the two
foreign key columns in the ORDERS table, Ctr_id (foreign key to CUSTOMERS) and
Pdt_code (to PRODUCTS).
Now suppose Pepe Yomita enters the store and buys one pair of jeans, two shirts, and
one silk tie. Given the current model this would mean that Pepe places three orders:
one for the jeans, one for the shirts and one for the tie. Three orders, all at the same
time, from one and the same customer. No problem so far, the model allows for this.
Now suppose the store wants to automate the billing of the orders. (This is probably
one of the reasons for making the model anyway.) Using the above model, this would
mean three orders and, as a consequence, three bills as the system has no way to know
these three orders somehow belong to each other.
It is better to change the model in such a way that one order can be for more than one
product. That means we should have a m:m relationship between ORDER and
PRODUCT, which we should investigate next.
.............................................................................................................................................
.........................................................................................................................................
Multiple PRODUCTS for an ORDER
CUSTOMER
* Id with
* Name of ORDER
* Id
PRODUCT with
* Date
* Code
for
* Name
Qu
ant
?
ity
3-30
Then there is the question again: where do you put quantity? Quantity can now no
longer be an attribute of an order because the attribute must be single-valued and
cannot contain three values 1, 2 and 1 at the same time. Quantity has become a
property of the m:m relationship between PRODUCT and ORDER.
.........................................................................................................................................
®
3-23
.........................................................................................................................................
This leads to:
About the slide

See Page 45
Another New Entity: ORDER ITEM
CUSTOMER
* Id with
ORDER
* Name of HEADER
* Id
PRODUCT
* Date
* Code
* Name
with
with
for for
ORDER ITEM
*Quantity Sold
3-31
Note the name change from ORDER to ORDER_HEADER, to avoid the 1: m

relationship that is mandatory at both ends. The set of tables for this model could be:
About the slide

Tables See Page 45
CUSTOMERS
Id Name ORDER_HEADERS
1 Sanchez Id Ctr_id Date_ordered
2 Lowitch 1 1 25-MAY-1999
3 Yomita 2 2 25-MAY-1999
4 3 1 25-MAY-1999
4
ORDER_ITEMS PRODUCTS
Ohd_id Pdt_code Quantity_sold Code Name
1 2 2 1 Jeans
2 2 2 2 Shirt
3 1 1 3 Tie
4 4
3-32
.............................................................................................................................................
Resolving Relationships
.........................................................................................................................................
Relationships and Intersection Entities
In the previous pages you have seen a typical example where relationships seem to
have attributes. The relationships in the example were many to many relationships.
This situation was handled by creating a new entity, an intersection entity, that replaces
the relationship and can hold attributes.
This leads to the following questions:
• What are the steps in resolving a relationship in general?
• Should every m:m relationship be resolved?
• Can other relationships than m:m be resolved?
Resolving a Relationship
Suppose we want to resolve the m:m relationship between entities A and B.
About the slide

Resolving m:m Relationship See Page 45
in Practice
A A
See Page 46
of A/B
COMBINATION
This is a suitable
xxx place to do
practice 3-3.
yyy
B B of
in
• Create new intersection entity

• Create two m:1 relationships, derive optionality
• Remove m:m relationship
3-33
1 First create a new intersection entity. You will experience that sometimes there is
no good word available for the concept you are modeling. The new entity can
always be named with the neologism “A/B COMBINATION”, or a name that is
somehow derived from the name of the original m:m relationship. Do not let the
unavailability of a proper name for the entity stop you from modeling it.
2 Next create two new m:1 relationships from entity A/B COMBINATION, one to A
and one to B. Initially, draw these as mandatory at A/B COMBINATION, as you
will probably only be interested in complete pairs of A and B. If the original m:m
.........................................................................................................................................
®
3-25
.........................................................................................................................................
relationship was optional (or mandatory) at A’s side, then the new relationship
from A to A/B COMBINATION is also optional (or mandatory).
3 Name the relationships. You can often name both relationships “in / of”.
4 The next step is to remove the m:m relationship you started with.
5 Finally, reconsider the newly-drawn relationships. Given a second thought they
may be optional at the A/B COMBINATION side. Also, they may turn out to be of
type m:m and need to be resolved, as you have seen in the example on customers
buying products.
Should Every m:m Relationship be Resolved?

The answer depends on a number of factors.
Given the usual scenario, when you start creating an ER model you will discover that
many of the relationships you draw are of type m:m. Most of these will appear to hide
entities that you need in a later stage as you need to have a place to put specific
attributes. Finally, you will have only a few “genuine m:m” relationships left.
No Purely from a conceptual data modeling point of view, there is no need to resolve
these genuine m:m relationships. The model is rich enough to be the basis for table
design. A m:m relationship will transform into a binary table; this is a table that
consists of the columns of two foreign keys only. This is exactly the same table as the
one that would result from the intersection entity when you resolved the m:m
relationship.
A m:m relationship in a conceptual data diagram needs less space than a separate
entity plus two relationships. For this reason a diagram with unresolved m:m
relationships is more transparent and easier to read.
Yes From a function modeling point of view the answer is different. If your model
contains a true m:m relationship there is apparently a business need to keep
information on the combinations of, say, entity A and B. In other words, the system
would contain at least one business function that creates the relationship. This “create
relationship” cannot be expressed as a usage of entities of attributes, although this is
usually what design tools require of the functional model. Oracle Designer is no
exception. This means that when you create an ER model in Oracle Designer you
would always resolve the m:m relationships in order to create a fully-defined
functional model with all data usages included.
Resolving Other Relationships

Can relationships other than m:m be resolved? Yes. Every relationship, even a 1:1, can
be resolved into an intersection entity and two relationships, just like a m:m
relationship. When would you want to do this? It is quite rare to find a situation where
you have to. A typical situation where you may like to resolve a non m:m relationship
is when one entity represents something that is outside your system, for example,
when the entity is part of a third-party package.
.............................................................................................................................................
.........................................................................................................................................
Suppose you need your system to create a m:1 relationship from external entity
PERSON to CUSTOMER TYPE, one of your internal entities:
Resolving m:1 Relationship
external
classified
PERSON as
CUSTOMER
classification
of TYPE
internal
3-35
This would result later on in a change of the table structure of the third-party
PERSONS table. This is undesirable (third parties often ask you to you sign a contract
that simply forbids you to do that) and sometimes even impossible if you have no
authority over that table.
About the slide

external See Page 46
PERSON CUSTOMER
TYPE
with in
for with
CLASSIFICATION
internal
The above model leaves the external entity PERSON as is and does the referencing
from inside. The m:1 relationship is replaced by an entity CLASSIFICATION and two
relationships.
.........................................................................................................................................
®
3-27
.........................................................................................................................................
Summary
Relationships connect entities and express how they are connected. There are ten types
of relationships, 4 of type1:m, 3 of type m:m and 3 of type 1:1.
The m:1 relationship that is optional at the 1 side is by far the most common type in
finished ER models. This one is very easy to implement in a relational database.
At the beginning of the process of creating an ER model, there are often many m:m
relationships. Many of these disappear after closer investigation.
Relationships cannot have attributes. If this seems to be the case, you need to resolve
the relationship into an intersection entity plus two relationships.
The other types are less common—some express more a desired situation rather than
reality, such as the m:1 relationship that is mandatory at both ends.
Summary
• Relationships express how entities are connected.

• Initially relationships often seem to be of type m:m.
• Finally relationships are most often of type m:1.
• Relationships can be resolved into:
– Two new relationships.
– One intersection entity.
3-37
.............................................................................................................................................
Practice 3—1: Read the Relationship
.........................................................................................................................................
Practice 3—1: Read the Relationship

Goal (See Page 39)
The goal of this practice is to learn to read relationships from an ER diagram.
Your Assignment
Read the diagrams aloud, from both perspectives. Make sentences that can be
understood and verified by people who know the business area, but do not know how
to read ER models.
Practice: Read the Relationship
ALU of
BRY
with
PUR bazooned in
YOK
bazooned by
KLO bilought in
HAR
glazoed with
3-39
.........................................................................................................................................
®
3-29
.........................................................................................................................................
Practice 3—2: Find a Context

Goal (See Page 42)
The purpose of this practice is to use your modeling skills.
Your Assignment
Given the following ER diagrams, find a context that fits the model.
1
.............................................................................................................................................
Practice 3—3: Name the Intersection Entity
.........................................................................................................................................
Practice 3—3: Name the Intersection Entity

Goal (See Page 46)
The goal of this practice is to find a proper name for the intersection entity after
resolving the m:m relationship.
Your Assignment
1 Resolve the following m:m relationships. Find an acceptable name for the
intersection entity.
Practice: Name the Intersection Entity
PRODUCT sold by DEPARTMENT

STORE
selling
PERSON crewing SAILBOAT

crewed by
INTERPRETER fluent in LANGUAGE

spoken by
3-44
2 Invent at least one attribute per intersection entity that could make sense in some
serious business context. Give it a clear name.
.........................................................................................................................................
®
3-31
.........................................................................................................................................
Practice 3—4: Receipt

Moonlight Coffees
Goal (See Page 47)
The purpose of this practice is to use a simple source of real life
data as a basis for a conceptual data model.
Scenario
You work as a contractor for Moonlight Coffees. Your task is to create a conceptual
data model for their business. You have collected all kinds of documents about
Moonlight. Below you see an example of a receipt given at one of the shops.
Your Assignment
Use the information from the receipt and make a list of entities and attributes.
Served by: Dennis

Till: 3 Dec 8, 4:35 pm
-----------------------
CAPPUCC M 3.60
* 2 7.20
CREAM .75
* 2 1.50
APPLE PIE 3.50
BLACKB MUF 4.50
<SUB> 16.70
tax 12% 2.00
<TOTAL> 18.70
=======
CASH 20.00
RETURN 1.30
-----------------------
Hope to serve you again
@MOONLIGHT COFFEES
25 Phillis Rd, Atlanta
3-45
.............................................................................................................................................
Practice 3—5: Moonlight P&O
.........................................................................................................................................
Practice 3—5: Moonlight P&O

Moonlight Coffees
Goal (See Page 47)
The purpose of this practice is to create a ER model iteratively,
based on new pieces of information and new requirements.
Scenario
You are still working as a contractor for Moonlight Coffees—apparently you are doing
very well!
Your Assignment
1 Create a entity relationship model based on the following personnel and
organization information:
All Moonlight Coffee employees work for a department such as

“Global Pricing” or “HQ”, or for a shop. All employees are at the
payroll of one of our country organizations. Jill, for example,
works as a shop manager in London; Werner is a financial
administrator working for Accounting and is located in Germany.
2 Extend or modify the diagram based on this information:
All shops belong to one country organization (“the countries”).

There is only one country organization per country. All countries
and departments report to HQ, except HQ itself.
3 And again:
Employees can work part time. Lynn has had an 80%

assignment for Product Development since the 1st September.
Before that she had a full-time position.
.........................................................................................................................................
®
3-33
.........................................................................................................................................
4 Change the model—if necessary and if possible—to allow for the following new
information.
a Jan takes shifts in two different shops in Prague.
b Last year Tess resigned in Brazil as a shop manager and moved to Toronto.
Recently she joined the shop at Toronto Airport.
c To reduce the number of direct reports, departments and country organizations
may also report to another department instead of Headquarters.
d The shops in Luxembourg report to Belgium.
e To prevent conflicting responsibilities, employees are not allowed to work for
a department and for a shop at the same time.
5 Would your model be able to answer the next questions?
a Who is currently working for Operations?
b Who is currently working for Moonlight La Lune at the Mont Martre, France?
c Are there currently any employees working for Marketing in France?
d What is the largest country in terms of number of employees? In terms of
managers? In terms of part-timers?
e When can we celebrate Lynn’s fifth year with the company? When can we do
the same with Tess’ fifth year with Moonlight?
f What country has the lowest number of resignations?
.............................................................................................................................................
Practice 3—6: Price List
.........................................................................................................................................
Practice 3—6: Price List

Moonlight Coffees
Goal (See Page 47)
The purpose of this practice is to use a simple source of real life
data as a basis for a conceptual data model.
Scenario
You work as a contractor for Moonlight Coffees.
Your Assignment
Make a ER model based on the pricelist from one of the Moonlight Coffee Stores.
SULFHOLVW 3KLOOLV5RDG$WODQWD
Practice:
List
Practice:
Price List
YLVLWXVDWZZZPRRQOLJKWFRP
small medium large

regular coffee 2.25 2.90 3.50
Price
cappuccino 2.90 3.60 4.20

café latte 2.60 3.20 3.90
special coffee 3.10 3.70 4.40
espresso 2.25 2.90 3.50
coffee of the day 2.00 2.50 3.00
decaffeinated .25 .50 .75 extra
black tea 2.25 2.90 3.50
infusions 2.60 3.20 3.90
herbal teas 2.90 3.60 4.20
tea of the day 2.00 2.50 3.00
decaffeinated .25 .50 .75 extra
milk 1.25 1.90 2.50
soft drinks 2.25 2.90 3.50
soda water 2.25 2.90 3.50
mineral water 2.90 3.60 4.20
G apple pie 3.50
H
G strawberry cheesecake 3.50
X
O
F

whole wheat oats muffin with almonds 3.90
Q
L
U blackberry muffin 4.50
[ H
D E
7 P
fruitcake 4.50
H
V W cake of the day 4.00
H
O S
D H
6 6 additional whipped cream .75
3-47
.........................................................................................................................................
®
3-35
.........................................................................................................................................
Practice 3—7: E-mail

Goal (See Page 48)
The goal of this practice is to extend an existing conceptual data model.
Scenario.
FOLDER
containing
placed in
author
USER
COMPOSITION written by of
part
MESSAGE received by of LIST
receiver consists
OTHER of of
COMPOSITION
reply of
containing replied
to by
with <5
ATTACHMENT ATT. TYPE
3-48
Your Assignment
Take the given model as starting point. Add, delete, or change any entities, attributes,
and relationships so that you can facilitate the following functionality:
1 A user must be able to create nick names (aliases) for other users.
2 A folder may contain other folders.
3 A user must be able to forward a composition. A forward is a new message that is
automatically sent together with the forwarded message.
4 All folders and lists are owned by a user.
Challenge:
5 A mail list may contain both users and other lists.
6 A mail list may contain external addresses, like “giovanni_papini@yahoo.com”.
7 A nickname may be an alias for an external address.
.............................................................................................................................................
Practice 3—8: Holiday
.........................................................................................................................................
Practice 3—8: Holiday

Goal (See Page 48)
The purpose of this practice is to do a quality check on a conceptual data model.
Scenario
“Paul and I hiked in the USA. Eric and I hiked in France and we rented a car in the
USA last year”.
Your Assignment
Comment on the model given below that was based on the scenario text.
COUNTRY TRANSPORT
France Boots
USA Boots
USA Car
C TRANSPORT RT
F r OU COUNTRY
O
US an NT
C o ts S P
US A ce RY
Bo AN
A
TR
C
Er OM
Bo ar
s
COMPANION
ot
ON
i
Er c PA
Pa c i N
NI
IO
E ic PA
ul N
Er OM
C
Pa ric
ul
3-49
.........................................................................................................................................
®
3-37
.........................................................................................................................................
Instructor Notes
Instructor Note
Topic Timings
33
Lecture 45- 60 minutes
R elationsh ip s in D e ta il
Practice 165- 240 minutes
Total 210- 300 minutes
This chapter goes into the details of

O verview
Relationships, covering naming issues,
• R e la tio n s h ip s optionality and degree, nontransferability. It
• T e n d iffe re n t relatio n sh ip typ e s
• N o n tran s fera b ility discusses the 10 types of relationships with
• R e la tio n s h ip s th a t se em to h av e a ttrib u tes
examples. Attributes compared to
Relationships. Resolving m:m relationships.
Resolving 1:m relationships. Again, lots of
3 -2
practices.
This slide and the following are a recap of what

Es tablish in g a R elatio nship
was earlier discussed in lesson 1, but now in
• D e te rm in e th e ex iste n ce o f a relatio n sh ip more detail.
• C h o o s e a n am e fo r th e re latio n s h ip fro m bo th
p e rs pe ctives
• D e te rm in e o p tio n ality
• D e te rm in e d e g ree
• D e te rm in e n o n tran s fera b ility
3 -3
In step one the existence of the relationships is

Es tab lishing the R elations hip
identified.
M E S SA G E sending U S ER
receiving New is the recursive relationship.
replyin g
3 -4
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Note that the next few illustrations in the book

N am ing the R ela tion ship
are different from the ones in the presentation.
receiving This is because of the builds in the presentation.
M E S SA G E received by
receiver of
USER
The slide titles correspond.
A M E S S A G E is rec eive d b y a U S E R
A U S E R is re c eiv e r o f a M E S S A G E
3 -6
O ptio nality
N o: Y es:
M E S SA G E received by USER
receiver of
• M u s t ev ery M E S S A G E b e re ceive d b y a U S E R ? Y es
• M u s t ev ery U S E R b e rec eive r o f a M E S S A G E ? No
3 -8
ER diagrams can only represent 0, 1 or more as

M andatory 1: M an datory m
the values of the degree of a relationship. If a
value is known, like <5, you should add that to
A split into B
the description of the entity.
part of
Oracle Designer allows you to enter any value:

• E ve ry A m u st be s p lit in to a t lea st on e B you can set the cardinality to 4.
• E ve ry B m u st be p a rt o f ex ac tly o n e A
3 -9
D eg re e
O n e: O n e or m ore:
M E S SA G E received by USER
receiver of
• C an a M E S S A G E b e rec eive d b y m o re
tha n o n e U S E R? Y es
• C a n a US E R b e th e rec eive r o f m o re th an o n e
M ESSAG E ? Y es
3 -1 1
Practice
Practice: R ead the R ela tions hip
Read the diagrams without a possible
of
AL U
w ith
BRY
interpretation.
PU R bazoone d in
YOK
bazo oned by
KL O bilought in
HAR
glaz oed w ith
3 -3 9
.........................................................................................................................................
®
3-39
.........................................................................................................................................
Another example of nontransferability:

N ontran sferability
TICKET - PASSENGER. Once a ticket is
F O LD E R assigned, the name of the passenger is printed
c ontaining
filed in
author
on the ticket and can not be transferred to
U S ER
M E S SA G E w ritten by
receiv ed by
of
another passenger. The ticket must be destroyed
reply of
receiver
of
and a new ticket must be created in order to
replie d
to by
make the change.
3 -1 2 In theory nontransferability can occur at both
sides of a relationship. In practice they can only
be implemented at the side where the future
foreign key will come. As a “1-end of a
relationship” can also be implemented as a
foreign key, you cannot say in general that the
diamond can only occur on the m-ends of a
relationship.
Usually the concept of a 1:m relationship that is
R elatio nship Typ es
1:m mandatory at the 1-end leads to a lot of
discussion. Such a relationship is perfectly all
(a) right in business situations, but is difficult (and
(b)
(c)
used to be impossible) to implement and used to
(d) be avoided for that reason.
More on this in the Mapping chapter.
3 -1 3

m :1
PRODUCT
part
of B UN DL E
consists
of
3 -1 4
A relationship type that leads to much

m :m discussion is the m:m that is mandatory at both
ends. The discussion is usually about
(e) implementation issues. In business contexts this
(f)
(g)
relationship is uncommon, but not impossible.
Another discussion is about the question
whether every m:m relationship should be
3 -1 5 resolved. See slide 3-33
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Some mail systems do not allow empty mail

m :m lists.
U SE R
pa rt
of LIS T
consists
of
3 -1 6
Entities that have a 1:1 relationship always lead

1:1 to the question: why make the distinction
anyway? Entities that have a 1:1 relationship
(h) can always be merged into one new entity,
( i)
(j)
maybe with loss of knowledge such as the
attributes that now become optional.
Often the distinction has to do with
3 -1 7 functionality of the system or with completely
different responsibilities.
This cannot be modeled with both PATIENT
1 :1 R elations hip s
R oles and EMPLOYEE as subtypes of PERSON as
there is potentially overlap between these two
PERSON
* Name
acting as P A TIE N T
role of * B lood Typ e
sets.
acting as EM PL O Y E E
role of
* Job
3 -1 8
1 :1 R elations hip s
P rocess
M E SS A G E
D RA F T
basis for
result of
3 -1 9
.........................................................................................................................................
®
3-41
.........................................................................................................................................
Practice
Find a C on te xt (1)
These practices seem to be easy, but they are
not. You will often discover students
misconceptions here.
3 -4 0
Practice
3 -4 1
Practice
3 -4 2
Practice
The first reaction of some will be: this is
impossible. When delegates assume a 1:m
relationship to stand for something like “is
divided into” they think the entity at the many
end represents something smaller than the other
entity.
3 -4 3
To check for redundancies, you must read and

R edu ndan t R ela tion ship s
understand the model, use good common sense
C O U N TR Y C O U N TR Y
and have a good long term memory.
location of location of birth
of of located of
located
in in
TOW N TO W N
ho m etow n hom e tow n
of liv ing living of living born
in in in in
PE R S O N P ER S O N
3 -2 0
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
A useful guideline is to upgrade every attribute

R elatio nsh ips a nd A ttribu te s
with a limited set of allowable values as a
• A n a ttrib u te ca n h id e a re latio n s h ip relationship to an entity and decide its
• R e la tio n s h ip c an b e “d o w n g rad ed ” to attrib u te
implementation at design time.
A T TA C H M E N T TY P E
* N am e
of
Using Oracle Designer, you have the option to
A TT AC H M EN T
w ith create a domain and connect the attribute to the
* T ype A TTA C H M E N T
* C onte nt * C onte nt
domain. Such a domain, however, is not a
3 -2 1 concept in Relational Theory.
A ttribu te C om pared to R e lations hip
• E as y m o d el • V alu e c o n tro l
• F e w e r tab le s • L ist o f v alu e s
• N o jo in • O the r rela tio n sh ip s
A T TA C H M E N T TY P E
* N am e
of
w ith
A TT AC H M EN T
* T ype A TTA C H M E N T
* C onte nt * C onte nt
3 -2 2
Not all make sense here (although this is hard to

A ttrib ute or E ntity
judge in the absence of a business context). Let
the students judge.
NAME
S A LA R Y
E M P LO YE E
* Id
JO B
BADGE
NAME, SALARY and ADDRESS do not
GENDER N A TIO N AL ITY (seem to) make much sense.
T EA M A D D R E SS
JOB, BADGE and TEAM may have attributes

of their own and seem to be solid entities.
3 -2 3
GENDER and NATIONALITY seem to be
potential domains.
This may lead to a discussion on naming, in
A ttribu te C om pared to R e lations hip
particular the suggestion that attribute Message
•
•
T h e re is n o su c h th in g a s a fo reig n ke y attrib u te.
U s ua lly, th e a ttrib u te n am e s h o u ld n o t co n ta in an
Id should be replaced by Id. Other entity names
e n tity n a m e.
F O LD E R
* N am e
in attribute names very often hide an entity or
containing relationship.
placed in
M E S SA G E
* M essa ge Id
* T ext
* F old er N a m e
3 -2 4
.........................................................................................................................................
®
3-43
.........................................................................................................................................
Using a single attribute Addressee where a

R elationsh ip C om pared to A ttrib ute
string like “alison, fred,
M E SS A G E addressed to U SE R giovanni_papini@yahoo.com” can be entered
a ddressee of
seems relationally incorrect. That string clearly

M E S SA G E
* A dd resse e
U SE R contains multiple addresses. At the other hand,
the message only has one addressee string,
M E S SA G E addressed to U SE R
o A dd resse e
addressee of which, after parsing, can contain only unknown
3 -2 5
users. The middle model is most likely to be
used for Addressee. For MESSAGES that are
received by USERS, the story is different. Then
a m:m relationship is the best model.
The next slides show an example where close
m :m R e lationsh ip s M ay H id e Som ething
investigation of a m:m relationship leads to the
conclusion that there in something missing in
C U S TO M E R PRODUCT
the model. Here the model misses a place to put
* Id buyer of * C o de
* N am e
bought by
* Name Quantity.
This is not the same as resolving a m:m
relationship.
3 -2 6
Q ua ntity is A ttrib ute of ...
C U S TO M E R
? PRODUCT
buyer of
* Id * C o de
* N am e bought by * Name
Q u antity
C U ST O M E R
? P RO D U C T
buyer of
* Id * C ode
* Name bou ght by * N am e
Q uan tity
3 -2 7
A ttrib ute of R ela tion ship?
C U S TO M E R PRODUCT
buyer of
* Id * C od e
* N am e bought by * N am e
Q ua ntity
3 -2 8
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Although you are probably used to a model

N ew E ntity O R D E R
with ORDER ITEM this model is fine.
C U S TO M ER
* Id
* N am e
w ith
O R D ER
Everything that is ordered can be recorded. An
of
PRODUCT
* C ode
w ith *Q ua ntity S old additional business need, as in this example, to
for
* N am e
create one bill for several products bought,
C U S TO M E R S P R O D U C TS ORDERS
Id N am e
1 S anchez
C ode N am e
1 Jeans
C tr_id P dt_code Q uantity_s old
1 2 2
leads to the ITEM entity
2 Low itch 2 S hirt 1 3 2
3 Y om ita 3 Tie 2 2 1
4 4 3
3 -2 9
M ultip le P R O D U C T S fo r an O R D ER
C US T O M E R
* Id w ith
* N am e of
ORDER
* Id
P RO D U C T w ith
* D ate
* C o de
for
* N am e
Qu
a n ti
?ty
3 -3 0
ORDER ITEM is introduced.

A noth er N ew E ntity: O R D E R IT EM
ORDER disappeared and becomes ORDER
C U S TO M ER
* Id
* N am e
w ith
of
O R D ER
H E A D ER
HEADER.
* Id
PRODUCT
* C ode
* N am e
* D ate
Of course, a model with entity ORDER and
w ith
w ith
ORDER ITEM connected with a m:m
for
O R D ER ITE M
for
relationship that is mandatory at both ends is
*Q u antity So ld
correct as well.
3 -3 1
Do not touch the naming of tables and columns

Tab les
now. This is discussed in the chapter on
C U S TO M E R S
Id N am e O R D E R _H E A D E R S
Id C tr_id D ate_o rdered
Mapping.
1 S anche z
2 Low itch 1 1 25-M A Y -1999
3 Y om ita 2 2 25-M A Y -1999
4 3 1 25-M A Y -1999
4
O R D E R _ITE M S P R O D U C TS
O hd_id P dt_code Q uantity_s old C ode N am e
1 2 2 1 Jeans
2 2 2 2 S hirt
3 1 1 3 Tie
4 4
3 -3 2
The animation shows the steps in resolving a

R eso lv ing m :m R elatio nship
in
m:m relationship.
A A
x xx
of A /B
C O M B IN A TIO N
Underline the need to check the new
yyy relationships. These may be of type m:m after
B B
in
of
closer inspection.
• C re ate n ew in ters ec tio n e n tity
• C re ate tw o m :1 relatio n s h ip s, d erive o p tio n ality
• R e m o ve m :m rela tio n sh ip
3 -3 3
.........................................................................................................................................
®
3-45
.........................................................................................................................................
Practice
P ractice : N a m e the Intersec tion En tity
Some intersection entities are easy to find as
PRODUCT sold b y
selling
D E P AR T M E N T
S TO RE
there is a good name available. Others are more
difficult as there is no appropriate word around.
P E R SO N crew ing SA IL BO AT
crew ed by Then you have to make up a name for the
IN TE R P RE T ER fluent in LA N G U A G E
concept.
spoken by
3 -4 4
R es olvin g m :1 R e lationsh ip
extern al
c lassified
P E R SO N as
C U S TO M ER
clas sification
of T YP E
in ternal
3 -3 5
Delegates may want to see the consequences of

R es olvin g m :1 R e lationsh ip
this solution in terms of tables. The table based
extern al
on CLASSIFICATION would be a binary table
P E R SO N C U S TO M ER
TY P E
where the foreign key column(s) to the table
w ith
for
in based on PERSON are primary key columns as
w ith
C LA SS IF IC A TIO N well.
in ternal
3 -3 6
Sum m ary
• R elatio n sh ips e xp re ss h o w e ntitie s are co n n e cted .

• In itia lly re latio n s hip s often s ee m to b e of typ e m :m .
• F in a lly relatio n sh ip s a re m o st o fte n o f ty p e m :1 .
• R elatio n sh ips c an b e re so lv ed in to :
– T w o n e w re la tion s h ip s.
– O n e in tersec tio n en tity.
3 -3 7
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................

Prac tices
• R e ad th e R ela tio n sh ip Practice 3day 4day

• F in d a C o n tex t
• N a m e th e In terse ctio n E n tity Read the Relationship Yes Yes
• R e ce ip t
• M o o n lig h t P &O Find a Context Cha Cha
• P ric e L ist
• E M a il Name the Intersection Yes Yes
• H o lid ay
Receipt Yes Yes
3 -3 8 Moonlight P&O Opt Cha
Price List Yes Yes
E-mail Opt Cha
Holiday Opt Cha
Practice
S e rv e d b y: De n ni s
T il l : 3 D e c 8 , 4 :3 5 p m
- - -- - -- - -- - -- - -- - -- -- - -
If delegates want to know what BLACKB MUF
C AP P UC C M 3 . 60
C RE A M
* 2
. 75
7. 2 0 etc. stand for, you can tell them to take a look at
A PP L E P IE
B LA C KB MU F
* 2 1. 5 0
3. 5 0
4. 5 0
the price list that is printed in one of the next
< S UB > 1 6. 7 0
t a x 1 2%
< T OT A L>
2. 0 0
1 8. 7 0
practices.
= == == = =
C A SH
R E TU R N
2 0. 0 0
1. 3 0
- - -- - -- - -- - -- - -- - -- -- - -
A hidden attribute is the name of the a product;
H o pe to se r ve yo u a ga i n
@M O ON L IG H T C OF F EE S
2 5 P hi l li s R d , A tl an t a
clearly the short name is used here.
3 -4 5
Practice
A ll M oo nlight C o ffee em ployee s w ork for a dep artm en t such as
This practice can very easily be replaced by an
“G lob al P ricing ” or “H Q ”, or for a sh op. A ll e m ploye es a re a t the
payroll of one of ou r co untry organiz ations . Jill, for exam ple,
w orks a s a s hop m an age r in L ondo n; W erne r is a finan cia l
interview session. See the remarks about
adm inistra tor w orkin g for A ccoun ting a nd is lo cate d in G erm any.
interview sessions.
A ll sh ops belon g to one coun try o rga nizatio n (“the coun tries”).
Th ere is only o ne c ountry organ ization per coun try. A ll co untries
and dep artm ents re port to H Q , excep t H Q itse lf. A complexity here is the idea of a department
E m ploy ees ca n w ork pa rt tim e . Lyn n ha s ha d an 80%
assign m en t for P rod uct D eve lopm ent sin ce th e 1 st Se pte m be r. that is not located anywhere. Departments are
B efore tha t she ha d a fu ll-tim e position.
“virtual” and should probably be defined as a
3 -4 6
set of tasks instead of a set of people. The
departments are a functional split up of the
company, the country organizations a
geographical split up of the company (forced by
the tax laws in most or all countries).
3K LOOLV5 RDG$WODQ WD
Practice
Practice:
S ULF HOLVW
List
Practice:
Price List
YLVLWXVD WZ Z Z P RRQOLJKWFRP
reg ular coffee

sm all
2 .2 5
m edium
2 .90
large
3.5 0
An important question here is: do you want to
Price
ca p pu ccin o 2 .9 0 3 .60 4.2 0

ca fé la tte 2 .6 0 3 .20 3.9 0
sp e cial coffee
e sp ress o
co ffee o f the da y
d e caffeina ted
3 .1 0
2 .2 5
2 .0 0
.2 5
3 .70
2 .90
2 .50
.50
4.4 0
3.5 0
3.0 0
.7 5 e xtra
see regular coffee or regular coffee small as a
b la ck te a
infu sion s
h e rba l te as
te a o f th e d ay
2 .2 5
2 .6 0
2 .9 0
2 .0 0
2 .90
3 .20
3 .60
2 .50
3.5 0
3.9 0
4.2 0
3.0 0
product? And how about decaffeinated? Or is
d e caffeina ted
m ilk
so ft d rin ks
so d a w a ter
.2 5
1 .2 5
2 .2 5
2 .2 5
.50
1 .90
2 .90
2 .90
.7 5
2.5 0
3.5 0
3.5 0
e xtra
regular coffee small decaffeinated a product?
m ine ra l w a ter 2 .9 0 3 .60 4.2 0
G a p ple pie 3 .5 0
H
G stra w b e rry ch ee se cake 3 .5 0
X
O
F w ho le w he a t o ats m u ffin w ith alm o nd s 3 .9 0
Q

L
U b la ckbe rry m uffin 4 .5 0
[ H
D E
fru itca ke 4 .5 0
7 P
H
V W ca ke o f th e d a y 4 .0 0
H
O S
D H a d ditio n al w hip p ed cre am .7 5
6 6
3 -4 7
.........................................................................................................................................
®
3-47
.........................................................................................................................................
Practice
F O LD E R
Typical situation: new functional wishes may or
co n taining
p laced in
autho r
U SE R
may not affect a model.
C O M P O S ITIO N w ritten by of
part
M ESSAGE received b y of LIS T
receiver co n sists
O TH E R of of
C O M P O S ITIO N
rep ly of
co ntainin g replied
to by
w ith <5
A T TA C H M E N T A TT . T YP E
3 -4 8
Practice
This type of situation is called a “fan trap”. The
C O U N TR Y TR A N S P O R T
France
USA
B oots
B oots
tables presented produce more combinations
USA C ar
than the given scenario contains.
RT
C TRANSPORT
F OU C O U N TR Y
O
U ran N T
C o ts S P
S c
The solution can be described as resolving the
Bo A N
U R
SA A e Y
TR
C
E OM
Bo r
s
a
C O M P A N IO N
ot
N
E ri c P A
relationship between 3 entities. This is an
IO
P r ic N
N
au IO
PA
l N
Er M
CO
example of an even more general n-entities

E r ic
P a ic
ul
3 -4 9
situation.
.............................................................................................................................................
.........................................................................................................................................

Name: Full Email:

Lesson number: 3

Detail comments
Content Page
Type Number
.........................................................................................................................................
®
3-49
.........................................................................................................................................
.............................................................................................................................................
Suggested Graphics
.........................................................................................................................................
Suggested Graphics

.........................................................................................................................................
®
3-51
.........................................................................................................................................

520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.............................................................................................................................................
4
.................................
Constraints
.........................................................................................................................................
Introduction
This lesson is about constraints that apply to a business. Constraints are also known as
business rules. Some of these constraints can be easily modeled. Some can be
diagrammed but the resulting decreased clarity may not be acceptable. Some
constraints cannot be modeled at all. These should be listed in a separate document.
Schedule
About the slide

See Page 31
• Unique Identifiers
• Arcs
• Domains
• Various other constraints
4-2
Topic See Page

Introduction 2
Identification 4
Unique Identifier 6
Arcs 12
Arc or Subtypes 16
More About Arcs and Subtypes 17
Hidden Relationships 18
Domains 19
Some Special Constraints 20
Summary 24
Practice 4—1: Identification Please 25
Practice 4—2: Identification 26
Practice 4—3: Moonlight UID 28
.............................................................................................................................................
Introduction
.........................................................................................................................................
Topic See Page

Practice 4—4: Tables 29
Practice 4—5: Modeling Constraints 30
Objectives
• Describe the problem of identification in the real world
• Add unique identifiers to your model and know how they are represented
• Recognize correct and incorrect unique identifiers
• Decide when an arc is needed in your model
• Describe the similarities between arcs and subtypes
• Describe various types of business constraints that cannot be represented in an ER
diagram
.........................................................................................................................................
®
4-3
.........................................................................................................................................
Identification
What Are We Talking About?
It is not unreasonable to assume everybody knows Rembrandt was born in the
Netherlands. What most people probably do not know is that Rembrandt was born on a
farm as the son of Pajamas and an unknown father. Rembrandt had a twin sister.
Although Rembrandt never married, he was the father of numerous children. You can
easily recognize Rembrandt and his offspring as they all have four white stripes at the
end of their tails.
Identification is about knowing what or who you are talking about. Obviously, the
name Rembrandt is not unique to the famous painter; other human beings and even
cats have the same name.
In day-to-day conversations, you can usually assume that you and the people you talk
to share enough of the same context and know enough about each other’s jobs and
interests, to understand what you are both talking about. Language is always a rather
nonspecific way to communicate, with lots of ambiguities, but people are very capable
of interpretation. Computers must communicate in a more specific way that is not
open to much interpretation. It would help a system to be told “Rembrandt the painter”
or “Rembrandt van Rijn, born in 1606” or maybe even the combination of all:
“Rembrandt van Rijn, the painter, born in 1606”, to distinguish this Rembrandt from
the other famous creatures with the same name.
The Problem of Identification

There are three sides to the problem of identification. One is identification in the real
world—how do we distinguish two real world things that have very similar properties?
This is the most difficult side. The second is identification within a database system—
how do we distinguish rows in tables? This one is far less complex. A third issue deals
with representation: how do we know what real world thing a row in a table
represents?
.............................................................................................................................................
Identification
.........................................................................................................................................
Identification in the Real World Many things in the real world are difficult, if not
impossible, to identify—distinguishing between two cabs, two customers, two versions
of a contract, or two performances of the fourth string quartet by Shostakovich. As a
general rule, real world things cannot be identified with certainty. You have to live
with a substantial level of ambiguity. For example, how can I be sure that the car at the
other side of the street with license plate MN4606 is the same car as the one I saw last
week with that number? I cannot even be sure it is the same license plate. In normal
circumstances there in no reason for doubt, but that is not the same as certainty.
Sometimes people have their reasons for creating confusion.
Fortunately, some things in the real world are easier as they are within your reach.
There you can define the rules. When a company sends out, for example, invoices, it
can give every single invoice a unique number. When a business lets people create
ElectronicMail usernames (identities), they can force these names to be unique.
Identification Within a Database Usually, database systems can make sure that a
row is not stored twice, or, to be more exact, that a particular combination of values is
not stored twice, within the same table. The technical problem is solved for you by the
standard software you use.
Representation The remaining problem is to make sure that you can always know
what real world thing is represented by a particular row in a table. The solution to this
problem depends highly on the context. How likely do you consider it to be that two
different employees for the same company have the same family name, or the same
family name plus initials, or the same family name plus initials plus birthdate?
About the slide

G. Papini, please? See Page 31
EMPLOYEES
Name Initials Birthdate
PAPINI G. 02-FEB-1954
HIDE T.M. 11-JUN-1961
PAPINI G. 02-FEB-1945
BAKER S.J.T. 24-SEP-1958
Clearly, the answer could be different when your company employs five or 50,000
employees.
Be aware that adding a new identifying attribute for EMPLOYEE, say, Id, only
partially solves the above problem. It would be very useful within the database. It
would not help much in the real world where employees usually would not know their
IDs, let alone the IDs of others. This kind of Id attribute often works only as an
internal, but not as an external identification.
.........................................................................................................................................
®
4-5
.........................................................................................................................................
Unique Identifier
To know what you are talking about, you need to find, for every entity, a value, or a
combination of values, that uniquely identifies the entity instance. This value or
combination is called the Unique Identifier for the entity.
About the slide

Unique Identifier Examples See Page 31
Practice
JOB Name
See Page 32
COMPUTER IN NETWORK IP Address
This is a suitable
TELEPHONE Country code, place to do
Area code, practice 4-1.
Telephone number
EMPLOYEE Employee number or
Name,
Initials,
Birth Date
MAIL LIST Name,
Owner
4-5
The MAIL LIST example shows that a unique identifier is not necessarily a
combination of attributes: the owner of a MAIL LIST is actually represented by a
relationship.
UID Representation
In an ER diagram, the components of the UID of an entity are marked:
• # for attributes.
• With a small bar across the relationship end for relationships (a barred
relationship).
About the slide

Indicates Unique Identifier CUSTOMER See Page 32
# Family Name
ORDER by o Initials
# Date # Address
responsible o Telephone
for
Indicates Unique Identifier
.............................................................................................................................................
Unique Identifier
.........................................................................................................................................
About the slide

USER See Page 32
# Name
part of owner
of
contains owned by
MAIL LIST
# Name
Single Attribute UID

The model shows that a USER of ElectronicMail is identified by attribute Name only.
Many entities will be identified by a single attribute. Typical candidate attributes, if
available, for single attribute UIDs are: Id, Code, Name, Description, Reference.
Multiple Attribute UID

An entity may have a UID that consists of multiple attributes; for example, a software
package can usually be identified by its Name and its Version, such as Oracle
Designer, version 7.0.
Composed UID
A MAIL LIST, illustrated above, is identified by the Name of the LIST plus the USER
that owns the LIST. That means that the combination of OWNER and a Name of a list
must be a unique pair.
This means that every USER must name their LIST instances uniquely, but need not
worry about names given by other users. It also means that the system may have many
LIST instances with the same name, as long as they are owned by different USERS.
You may argue that a USER also has a composed UID, as the Name must be unique,
within this mail system. To show this, you could add an extra high level entity, MAIL
PROVIDER, plus a relationship form USER to PROVIDER. The relationship then is
part of the UID of a USER.
Cascade Composed UID

It is not uncommon that an entity has a barred relationship to another entity that has a
barred relationship to a third entity, and so on.
About the slide

ROOM FLOOR HOTEL See Page 32
# No # No # Name
.........................................................................................................................................
®
4-7
.........................................................................................................................................
UID: Relationships Only
About the slide

Multiple Relationship UID See Page 32
USER USER
# Name # Name
owner is owner
part of of
of referred to
contains owned by owned by
LIST LIST
# Name # Name
contains
referring
contained
to
in
LIST ITEM
4-8
A Unique Identifier can also consist of relationships only.

At the lower right side of the diagram, entity LIST ITEM is shown, which resulted
from the resolved m:m relationship between LIST and USER.
The model shows that a LIST ITEM is identified by the combination of the USER and
the LIST. In other words, the model says that a LIST may contain as many ITEMS as
you like, as long as they refer to different USERS.
This results in the next definition:
A Unique Identifier (UID) of an entity is a constraint that declares the uniqueness
of values; a UID is composed of one or more attributes, one or more relationships,
or a combination of attributes and relationships of the entity.
Consequently, not all components of the UID may be optional.
Indirect Identification
Identification regularly takes place using an indirect construction, that is, when the
instance of an entity is identified only by the instance of another entity it refers to.
Examples
• In many office buildings employees are identified by their badge, which is
identified by a code.
• Around the world a person is identified by the picture on their passport.
.............................................................................................................................................
Unique Identifier
.........................................................................................................................................
• All cows in the European Community are identified by the number of the tag they
are supposed to wear in their ear.
• When you park a car at Amsterdam International Airport you enter the parking lot
by inserting a credit card into a slot at the gate. The parking event is identified by
the credit card of the person that parked the car. This is a double indirect
identification.
Clearly, these identification constructions are not 100% reliable, but are probably as
far as you can go in a situation.
The model of these indirect identifications is shown in the next illustration, at the right
bottom corner. An instance of S is identified by the single instance of T it refers to. In
other words, the UID consists of one relationship only.
Multiple UIDs
Entities may have multiple UIDs. Earlier, you saw the example of entity EMPLOYEE
that can be identified by an Employee Number, and possibly by a combination of, for
example, Name, Initials and Birth Date.
At some point in time, usually at the end of your analysis, you promote one of the
UIDs to be the primary UID. All the other UIDs are called secondary UIDs.
You would usually select the UID that is most compact or easy to remember to become
primary UID. The reason, of course, is that the UID leads to one or more foreign key
columns in related tables. These columns should not be too sizeable. Preferably, the
primary UID of an entity does not consist of optional elements.
UID in Diagram
Only the primary UID is shown in ER Diagrams.
Where UIDs Lead

Unique Identifiers lead to Primary Key and Unique Key constraints.
.........................................................................................................................................
®
4-9
.........................................................................................................................................
Unique Identifier Examples
About the slide

Well-defined Unique Identifiers See Page 32
Z Q P
# Z1 # Q1 # P1
o Z2
o Z3 Y
# Y1 K
# Z4
# Y2 L
# L1
X
# X1 M
# M1
XY R T S
# R1 # T1
4-9
Examples of Incorrect Unique Identifiers
About the slide

Incorrect Unique Identifiers See Page 33
Practice
L F G
K # G1 See Page 33
# L1 # F1
# K1 This is a suitable
place to do
P R practice 4-2.
# P1 # R1
KL
T
o # T1 Q
# Q1
G H
# G1
4-10
.............................................................................................................................................
Unique Identifier
.........................................................................................................................................
Information-Bearing Identifiers
When things in the real world are coded, you need to be especially careful. Codes that
have been used for some time are often information bearing. An example is a company
that uses product codes like 54.0.093.81, where 54 refers to the product group, 0
shows that the product is still in production, 093 identifies the factory where the
product is made and 81 is a sequence number. These codes come from the time when a
maximum amount of information had to be squeezed into a minimum number of bits.
The example above would be modeled conceptually:
About the slide

See Page 35
Information-Bearing Codes
54.0.093.81
Product Group
In Production?
Factory
Sequence Number
PRODUCT PRODUCT GROUP

# Code
# Code
* In Production?
* Sequence No FACTORY
# Id
4-11
The Code attribute would contain the same codes, for reasons of compatibility, but
now without meaning, as the old meaning is transferred to the attributes and
relationships. Product 54.0.093.81 may now be produced by factory 123 and may no
longer be in Product Group 54.
.........................................................................................................................................
®
4-11
.........................................................................................................................................
Arcs
Suppose ElectronicMail rents the Advertisement Areas that are located in their various
mail screens on the Web. This renting is controlled by contracts; contracts consist of
one or more standard conditions and customized conditions. This can be modeled with
four entities: CONTRACT, CONTRACT COMPONENT, STANDARD CONDITION
and CUSTOMIZED CONDITION. See the model below. How do we model the
following constraint: every instance of CONTRACT COMPONENT refers to either a
STANDARD CONDITION or a CUSTOMIZED CONDITION, but not to both at the
same time?
An arc is a constraint about two or more relationships of an entity. An arc indicates
that any instance of that entity can have only one valid relationship of the relationships
in the arc at any one time. An arc models an exclusive or across the relationships. An
arc is therefor also called exclusive arc.
There is no similar constraint construct for attributes of an entity.
About the slide

Arcs See Page 35
Contract “A contract consists of contract

components; these are standard
Conditions Std?
1 conditions or customized conditions”
2
3
4
5 CONTRACT STANDARD basis for
6
CONDITION
based
consists in on
of CUSTOMIZED
CONDITION Arc
in
Indicates
relationship part of
in arc referring to referring to
CONTRACT COMPONENT
4-12
Arc Representation
The arc is drawn as an arc-shaped line, around an entity. Where the arc crosses a
relationship line a small circle is drawn, but only if the relationship participates in the
arc.
.............................................................................................................................................
Arcs
.........................................................................................................................................
Mandatory Compared to Optional Relationships in an Arc

When the arc is drawn across two mandatory relationships, as in the example above, it
means that every instance of CONTRACT COMPONENT must have one valid
relationship. When the arc is drawn across two optional relationships, it would mean
that an instance may have one valid relationship.
Another Arc Example
About the slide

Exclusive Arc See Page 35
USER owner
of
owned
by
LIST
is referred to container is referred
of to
contained
referring to in referring to
LIST ITEM
4-13
Suppose a MAIL LIST may contain USERS as well as other MAIL LISTS. This
means that a particular LIST ITEM may refer to a USER or a LIST. To be more
precise, it must be a reference to a USER or to a LIST, but not to both at the same time.
Note
• The relationship contained in/container of from LIST ITEM to LIST (the one that
is printed in gray) is not part of the arc as there is no small circle at the intersection
with the arc.
• A relationship that is part of a UID may also be part of an arc.
• The constraint that a LIST may only contain LISTS other than itself cannot be
shown in the model.
Where Arcs Lead

An arc is normally implemented as a check constraint in an Oracle database. Note that
a check constraint is not an ISO standard relational database object. In other words, an
arc must be implemented differently in other database systems.
.........................................................................................................................................
®
4-13
.........................................................................................................................................
About the slide

Possible Arc Constructs See Page 36
4-14
Some Rules About Arcs

• An arc always belongs to one entity.
• Arcs can include more than two relationships.
• Not all relationships of an entity need to be included in an arc.
• An entity may have several arcs.
• An arc should always consist of relationships of the same optionality:
all relationships in an arc must be mandatory or all must be optional.
• Relationships in an arc may be of different degree, although this is rare.
Tips About Arcs

• Do not include a relationship in more than one arc, for clarity reasons.
• Consider modeling subtypes instead of arcs (see the next paragraph).
.............................................................................................................................................
Arcs
.........................................................................................................................................
Incorrect Arcs
About the slide

Some Incorrect Arc Constructs See Page 36
• The arc “belongs” to one

entity
• Relationships in the arc

must be of the same
optionality
• Arcs must contain at least
two relationships
An arc may be correct, but is

quite difficult to implement ...
4-15
You cannot capture all possible relationship constraints with arcs. For example, if two
out of three relationships must be valid, this cannot be represented. The table below
shows what an arc can express.
Number of Valid Relationships in Arc

Per Entity Instance Minimum Maximum
}n n n
}n 1 1
}n 0 n
}n 0 1
4-16
.........................................................................................................................................
®
4-15
.........................................................................................................................................
Arc or Subtypes
Relationships within an arc are often of a very similar nature. They frequently carry
exactly the same names. If that is the case, an arc can often be replaced by a subtype
construction, as the illustration shows. On the left you see the arc that contains both
referring to relationships of LIST ITEM. In the model on the right there is only one
relationship left, now connected to an entity ADDRESS, a new supertype entity of
USER and LIST.
Both models are equivalent.
About the slide

Arc or Subtype See Page 36
USER ADDRESS
owner owner
of USER of
owned owned
by by
LIST LIST
is referred is referred contains
to contains to
referring referring is referred
to to to
in referring to in
LIST ITEM LIST ITEM
4-17
The model on the left emphasizes the difference between USER and LIST, which
clearly exists; the other model emphasizes the commonality. This commonality is
mainly a functional issue. Both USERS and LISTS can be part of a LIST and both can
be used as the address in the To, Cc or Bcc field in the screen for composing a
message.
Generally speaking, you can replace every arc with a supertype/subtype construction
and every supertype/subtype construction with an arc.
.............................................................................................................................................
More About Arcs and Subtypes
.........................................................................................................................................
More About Arcs and Subtypes

Arcs and Subtypes are similar notions. The five models that are printed below all show
the same context.
Model 1 and 2 are equivalent models to what you have seen before.
If every instance of A is related to a P or a Q, then you could say there are P-related-
A’s and Q-related-A’s. These two subtypes of A are shown in model 3.
Model 4 goes one step beyond this and shows subtypes of entity A and a supertype R
of P and Q.
Though models 3 and 4 are completely correct, it is likely they both model something
twice.
About the slide

Arc and Subtypes See Page 36
A A
1 2
R
Q P Q
P
A C
A C B
A C B
B
3 4 5
R R
Q Q
P Q P P
4-18
Note that only model 5 does not present the same information. In model 5, an instance
of B may be related to an instance of Q, unlike that which is modeled in 3 and 4.
.........................................................................................................................................
®
4-17
.........................................................................................................................................
Hidden Relationships
Every subtype hides a relationship between the subtype and its supertype. Moreover,
the relationships are in an arc, as the next illustration shows. Both relationships are
mandatory 1:1 is/is relationships.
About the slide

Subtypes Hide Relationships in Arc See Page 37
A A
is B
B is
is C
C
is
• Every A • Every A must

is either a B or a C be a B or be a C
• Every B is an A • Every B must be an A
• Every C is an A • Every C must be an A
4-19
.............................................................................................................................................
Domains
.........................................................................................................................................
Domains
A very common type of attribute constraint is a set of values that shows the possible
values an attribute can have. Such a set is called a domain.
Very common domains are, for example:
• Yesno: Yes, No
• Gender: Male, Female, Unknown
• Weekday: Sun, Mon, Tue, Wed, Thu, Fri, Sat
In a conceptual data model you can recognize these as entities with, usually, only two
attributes: Code and Description. These domain entities are referred to frequently but
do not have any “many” relationships of their own, (see model A below). Typically,
you would know all the values before the system is built. The number of values is
normally low. Often you would deliver such a system with non-empty code tables
An alternative model for the (sometimes many) code entities is a more generic, two-
entity approach: CODE and CODE TYPE, model B.
Model A has the advantage of fewer relationships per entity as well as easy-to-
understand entities; B has obviously fewer entities and therefore will lead to fewer
tables.
.
About the slide

Value sets See Page 37
CODE TYPE
# Id
YESNO * Name
# Code
A * Max Length
* Description of Description
B
GENDER
# Code
* Description CODE
# Code
WEEKDAY * Description
# Code
* Description
4-20
Domains that have a large number of values, such as all positive integers up to a
particular value, are usually not modeled.
You should list and describe such a constraint in a separate document.
.........................................................................................................................................
®
4-19
.........................................................................................................................................
Some Special Constraints

Although an entity relationship model can express many of the constraints that are not
too complex, there are many types of constraints left that cannot be modeled. These
constraints must be listed on a separate document and often need to be handled
programmatically.
Categories: Examples
• Conditional domain: The domain for an attribute depends on the value of one or
more attributes of the same entity.
• State value transition: The set of values an attribute may be changed to depends
on the current value of that attribute.
• Range check: A numeric attribute must be between attribute values of a related
instance.
• Front door check: A valid relationship must only exist at creation time.
• Conditional relationship: A relationship must exist or may not exist, if an
attribute (of a related entity) has a special value.
• State value triggered check: A check must take place when an attribute is given a
value that indicates a certain state.
• There are also combinations of the above.
Range Check: Example
About the slide

See Page 37
EMPLOYEE JOB
* Name * Title
* Address * Minimum Salary between
with * Maximum Salary
of
for referring to
EMPLOYMENT
* Start Date
o End Date
* Salary
Constraint: Employee salary must be within the salary range of the job of the
employee.
.............................................................................................................................................
.........................................................................................................................................
State Value Transition: Example
About the slide

See Page 38
Possible
Wid
Mar
to
Sin
Div
DP
Marital Status
EMPLOYEE
Transitions from
* Name
* Address Single
* Current Marital Status Married
Widowed
Divorced
Domestic Partnership
Constraint: Marital Status of employees cannot change from any value to all other
values.
Conditional Relationship: Example
About the slide

See Page 38
CONTRACT
# Id
* Standard Indicator STANDARD basis for
CONDITION based
consists on
in
of CUSTOMIZED
CONDITION
in
part of
referring to referring to
CONTRACT COMPONENT
4-23
Constraint: If a CONTRACT has Standard Indicator set to Yes, the CONTRACT

COMPONENT may not refer to a CUSTOMIZED CONDITION.
.........................................................................................................................................
®
4-21
.........................................................................................................................................
Derived Attribute?
You may argue that the attribute Standard Indicator of CONTRACT is derivable. If the
contract contains CUSTOMIZED CONDITIONS, it is, by consequence, not a
standard CONTRACT. This may be true, but it is not necessarily so. Suppose the
contract is created in various steps, by various people with different responsibilities.
Then, the creation of a CONTRACT is a process that may take days. The Standard
Indicator, then, is an attribute of that process. Only when the CONTRACT is finalized,
should a check be made that the Indicator corresponds with the actual STANDARD
and CUSTOMIZED CONDITIONS. In those situations, the entity CONTRACT will
usually have an attribute Completed Indicator that triggers the check when set to Yes.
Rules May Lead to Attributes

If you cannot capture a constraint in the model, the best you can do within the model is
make the model rich enough so that a program for constraint checking performs well.
Consider the rule:
If the Standard Indicator is set to No, and there is no CUSTOMIZED
CONDITION, then the CONTRACT is not yet ready for being sent to the
CUSTOMER.
This rule deals with a procedure and cannot be modeled as such, but it calls for an
indicator at entity CONTRACT to indicate something like a Ready To Send status.
Model for Overview

An analyst often runs into constraints that cannot be modeled and thus must be
documented separately. This is not a weakness of the model. An important goal of a
diagram is to give an overall picture, not all the details. The model should let you view
the key areas clearly.
.............................................................................................................................................
.........................................................................................................................................
Boundaries
More than once the checking of constraints or special rules needs to use information
that is not directly related to one of the entities in the model.
Typical examples are rules and boundaries set by external sources, like a mother
company or national legislation. If reasonably possible, these rules should be part of
your conceptual data model, and should not be hard coded in your programs. The
reason is obvious: if the rule changes, which is beyond your power, there is a chance
you do not have to make changes to your programs. Only an update of a value in a
table would be necessary. The time spent developing a complete model is fully
justified by the programming time saved.
About the slide

See Page 38
Boundaries
EXTERNAL
unrelated entity # Id
* Description
* Value
and possible implementation

EXTERNALS
Id Description Value
1 Value added tax % 15
2 Maximum available Space per Mail User in Mbyte 500
3 Maximum level of Nested Mail Folders 3
4 Maximum level of Nested Mail Lists 16
4-24
.........................................................................................................................................
®
4-23
.........................................................................................................................................
Summary
Entities in the real world must be individually identified before they can be
represented in a database. You would not know what you are talking about, otherwise.
Some entities are really difficult to identify, such as people and paintings. Some are
more easy, especially when they are part of the domain as you can make up the rules,
such as a unique number for each of the invoices you send to your customers. Some
unique identifiers are already present in the real world, often as a combination of
attributes and relationships of the entity.
Summary
• Identification
– Can be a real problem in the real world
– Models cannot overcome this
• Entities must have at least one Unique Identifier
• Unique Identifiers consist of attributes or
relationships or both
• Arcs
• Many types of constraint are not represented
in ER model
4-25
Arcs in a diagram represent a particular type of constraint for the relationships of one
entity.
Many business constraints cannot be represented in a diagram and must be listed
separately. This way the model remains clear and not too full of graphical elements.
.............................................................................................................................................
Practice 4—1: Identification Please
.........................................................................................................................................
Practice 4—1: Identification Please

Describe how you would identify the following entities, making up any attributes and
relationships you consider appropriate.
Practice: Identification Please
• A city
• A contact person for a customer
• A train
• A road
• A financial transaction
• An Academy Award (Oscar)
• A painting
• A T.V. show
4-27
.........................................................................................................................................
®
4-25
.........................................................................................................................................
Practice 4—2: Identification

Are the entities in the next diagrams identifiable?
1
A B C
# Xx * Yy # Zz
A B
C # Id
# Code
A
* Xx
B C with D
# Yy # Zz # Id
of
P Q
# Id
.............................................................................................................................................
Practice 4—2: Identification
.........................................................................................................................................
P
# Name
6
Note: the next model describes a context that may be different from the world you are
familiar with.
PERSON
FEMALE
MALE son of
# Name
# Seqno mother of # Birth Date
partner in partner in
with husband with wife

MARRIAGE
# Start Date
7 Given the above model, answer the following questions.

a Can person A marry twice?
b Can person A marry twice on the same day?
c Can person A marry with person B twice?
d Can person A marry with person B twice on the same day?
e Can person A be married to person B and person C simultaneously?
f Can person A be married to person A?
.........................................................................................................................................
®
4-27
.........................................................................................................................................
Practice 4—3: Moonlight UID

Moonlight Coffees
Goal (See Page 39)
The purpose of this practice is to define UIDs for given entities.
Scenario
Moonlight Coffees, organization model.
Your Assignment
Use what you know about Moonlight Coffees by now, and, most importantly, use your
imagination.
1 Given the model below, indicate UIDs for the various entities. Add whatever
attributes you consider appropriate. Country organizations have a unique “tax
registration number” in their countries.
2 Are there any arcs missing?
reporting to report of
DEPARTMENT report of
HQ
reporting to
OTHER
DEPARTMENT COUNTRY in COUNTRY
ORGANIZATION
with of
with with with
EMPLOYEE belongs to in
with SHOP
of for for
PAYROLL with
ENTRY
to for to
ASSIGNMENT as JOB
in
4-34
.............................................................................................................................................
Practice 4—4: Tables
.........................................................................................................................................
Practice 4—4: Tables

Goal (See Page 39)
The purpose of this practice is to match a given context with a ER model.
Your Assignment
Read the text on ISO Relational tables.
Do a quality check on the ER model based on the quoted text and what you know
about this subject. Also list constraints that are mentioned in the text but not modeled.
Practice: Table 1
“In a relational database system, data is stored in tables. Tables of a

database user must have a unique name. A table must have at least
one column. A column has a unique name within the table. A column
must have a data type and may be Not Null.
Tables can have one primary key and any number of unique keys. A
key contains one or more columns of the table. A column can be part
of more than one key.
A table can have foreign keys. A foreign key always connects one
table with another. A foreign key consists of one or more columns of
the one table that refers to key columns of the other table.
The sequence of columns within the key and foreign key is important.”
from with
FOREIGN KEY TABLE KEY
# Name # Name # Name
with
to PRIMARY
referenced of
with in with UNIQUE
with
for in
from in COLUMN for
ASSOCIATION in USAGE
# Seqno # Name
in * Data Type of # Seqno
to o Not Null
.........................................................................................................................................
®
4-29
.........................................................................................................................................
Practice 4—5: Modeling Constraints

Goal (See Page 39)
The purpose of this practice is to learn what constraints can be modeled and how, and
which cannot be modeled.
Your Assignment
Change the diagrams to model the constraint given.
EMPLOYEE
# Name managed by
manager of
1 Every EMPLOYEE must have a manager, except the Chief Executive Officer.
owned
USER by LIST
owner of
# Name # Name
owned
owner of by NICKNAME
# Alias
2 A user may not use the same name for both NICKNAME and LIST name.
with
subfolder USER
# Name
FOLDER within
Name owner
of
owned
by
3 A top level FOLDER must have a unique name per user; sub folders must have a
unique name within the folder where they are located.
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Instructor Notes
Instructor Note
Topic Timings
44
Lecture 60 minutes
C o nstrain ts
Practice 90 - 180 minutes
Total 150 - 240 minutes
This lesson describes business constraints and

O verview
how these may be captured in an ER model.
• U n iq u e Id e n tifiers Unique identifiers and arcs are discussed in
• A rc s
• D o m a in s detail. Of course, many types of business
• V ario u s o th er co n s train ts
constraints cannot be modeled at all. Examples
of these are discussed in the second part of this
lesson.
4 -2
If you saw this information in a table, would

R em b randt
you suspect its correctness? Possibly the system
had checked the unicity of the combination of
Name, Initials and Birthdate. For the table there
are two employees called G. Papini. This is an
example of the problem of identification. The
bad news is: you cannot prevent this happening.
4 -3
The identification of employee is very

U nique Iden tifier Exa m ples
disputable. Although unlikely, the combination
JO B N am e
C O M P U T E R IN N E T W O R K IP A d d re ss may have duplicates. Students may suggest
TELEPHONE C o u n try co d e,
A re a c o d e, using a Social Security Number. This, however,
T elep h o n e n u m b e r
EM PLOYEE E m p lo ye e n u m b er or is not a safe identifier within some countries
N am e,
In itials , (like the USA) and is not available in every
B irth D a te
M A IL L IS T N am e, country around the world.
O w n er
4 -5 The safest identifiers are the abstract and
technical ones, like those for JOB and
TELEPHONE.
.........................................................................................................................................
®
4-31
.........................................................................................................................................
Practice 4-1
Prac tice: Id entification Ple ase
Good practice for joint class activity and
•
•
A c ity
A c o n tact pe rs o n fo r a cu s to m er
discussion.
• A tra in
• A ro a d
• A fin a n cia l tra n sa ction
• A n A c ad e m y A w ard (O sc ar)
• A p a in tin g
• A T .V . sh o w
4 -2 7
The drawing technique suggests that you do not

U nique Identifier
need to show the optionality of an attribute that
Indicates U nique Identifier
is part of a UID. This wrongly suggests that the
C U S TO M E R
O R D ER by
# F am ily N am e
o Initials
attributes in a UID are always mandatory.
# D ate # A dd ress
responsible
for
o Telepho ne
Indicates U nique Identifier

Columns in a primary key in an Oracle RDBMS
must be mandatory, but that is another issue.
4 -6
Entities can have several relationships. The

U n iq ue Ide ntifie rs
USER
barred relationship is not an arbitrary choice.
# N am e
part of o w ner
of
contains ow ned by
M AIL LIS T Avoid the discussion about the implementation
# N am e
of cascade barred relationships. This is covered
ROOM
# No
F LO O R
# No
H O T EL
# N am e in the chapter on Mapping.
4 -7
Resolved relationships usually lead to

M ultip le R ela tion ship U ID
intersection entities that are identified by their
U S ER U SE R
# Name
ow
o w ner
# N am e
is ow ner
relationships.
part of of
of
of refe rred to
contain
containss ow ned by ow ne
nedd by
LIS T LIS T
# Name # Name
contains
referring
contained
to
in
LIS T ITE M
4 -8
Entity P is well identified as it refers to Q which

W ell-de fine d U nique Iden tifiers
is well identified as it refers to R which is well
Z
# Z1
Q
# Q1
P
# P1
identified.
o Z2
o Z3
# Z4
Y
# Y1 K Entity K is well identified as each of its
# Y2 L
X
# L1 subtypes is well identified.
# X1 M
# M1
Entity S is well identified by a single
XY R T
# R1 # T1
S
relationship.
4 -9
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
This slide can be used as a class discussion:

In correct U nique Iden tifie rs
explain why the entity is not well defined.
L F G
K
# K1
# L1 # F1 # G1
KL in not well identified as there are possibly
KL
P
# P1
R
# R1 KL instances that are not related to a K nor an
L.
T
# T1 Q
o
# Q1 G is not well identified as you would not know
G H
# G1
which F it refers to. You cannot model that a set
4 -1 0
is identified by its elements.
T is not well identified because the attribute is
optional.
P, Q and R are not well identified as the
identifiers have a circular structure.
H is not well identified because two instances
of H can refer to the same instance of G. The
identifier of that instance is therefore not unique
for H.
Practice 4.2.1
P ractice : Identification 1
discussion
A B C
# Xx * Yy # Zz
4 -2 8
Practice 4.2.2
A
discussion.
B
C # Id
# C o de
4 -2 9
Practice 4.2.3
A
discussion.
* Xx
B C w ith D
# Yy # Zz # Id
of
4 -3 0
.........................................................................................................................................
®
4-33
.........................................................................................................................................
Practice 4.2.4
discussion.
P Q
# Id
4 -3 1
Practice 4.2.5
P
discussion.
# N am e
4 -3 2
Practice 4.2.6, 4.2.7

P E R SO N
M A LE
# S eqn o
son of
FE M A LE
# N am e discussion.
m other of # B irth D ate
partner in partner in
Do not forget the extra questions (4.2.7) as
w ith hu sband w ith w ife
these really focus on the understanding of what
M A R R IA G E
# S ta rt D ate
is in the model and what not.
4 -3 3
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
There is a a great deal to explain about the

In form ation -B earing C o des
problems of information-bearing numbers
P rod u c t G ro u p
54.0.093.81
(codes). The older the codes the more
In P ro d u ctio n ?
F ac tory
S e q u en c e N u m b er
suspicious you should be, but these codes are an
undeniable part of the world today.
PRODUCT PRODUCT GROUP
# C ode
* In Produ ction?
# C o de
Some examples:
* S equ ence N o F A C TO R Y
# Id
• Social Security numbers that have gender
4 -1 1
coded into them. What if someone has a
change of gender?
• Bank account numbers that have a branch
office coded into them. What should a bank
do when their customers do not accept the
change in their bank account number when
they move to another city?
• An e-mail account that carries the name of
the internet provider. What do you do when
you want to change to another provider?
Guidelines:
• Do not introduce new information-bearing
codes
• Try to bury existing ones
• If that is impossible, try to neutralize old
ones by treating them as codes that do not
bear information
Give the delegate time to absorb this contract
A rcs
context as it contains much detail.
C on tra ct “A co n tra ct c o ns is ts o f c o n tra ct
Co nditions
1
2
S td?
co m p o n e n ts ; th e s e a re sta n da rd
co n d itio n s o r c u sto m iz ed c o n ditio n s”
Explain that an arc “softens” a constraint that
3
4
5
6
C O N TR A C T S T A N D A R D ba
C O N D ITIO N
sis for
basis
was modeled earlier: where you had two
based
consists
of
in on
C U S TO M IZE D mandatory relationships now one is enough.
C O N D ITIO N A rc
in
In dic ate s
re latio n ship part of
in arc refe rring to referring to
C O N TR A C T C O M P O N E N T
4 -1 2
The example shows why the little circles are of

Exc lus ive A rc
value. Moreover, it shows that relationships in
an arc may also be barred.
USER ow ner
of
ow ned
by
LIS T
is referred to container is referred
of to
conta ined
contained
referring to in referring to
LIS T ITE M
4 -1 3
.........................................................................................................................................
®
4-35
.........................................................................................................................................
Arcs may:
Pos sible A rc C on stru cts
• Contain more than two relationships
• Contain optional relationships, but then
they must all be optional
• Cross 1-ends of relationships
• Cross relationships of different degree.
4 -1 4
Some schools of thought say the degree
must be the same, but there is no reason for
this. As long as both relationships will
result into foreign keys for the same table,
the implementation is no problem.
An entity can have several arcs.
The most common mistakes about arcs.
Som e Inco rre ct A rc C ons truc ts
For Oracle Designer users: in Oracle Designer
• T h e a rc “b e lo n g s ” to o n e
e n tity
it is possible to create an arc across one
• R elatio n s h ip s in the a rc relationship, by drawing a regular one and then
m u s t b e o f th e sa m e
•
o p tio n ality
A rcs m us t c on ta in a t lea st
deleting one of the arced relationships...
tw o re la tio n sh ip s
A n arc m a y be c o rrec t, b u t is
q u ite d ifficu lt to im p le m en t ...
4 -1 5
Arcs can always be replaced by a subtype/

A rc or S ubtyp e
supertype construction, if you are allowed to
U SE R A DD R E S S create the supertype. This is allowed if the
ow ner
ne r ow ner
of
of
ow ned
by
USER of
ow ned
by
(future) subtypes are mutually exclusive. And
is referred
to contains
co ntains
LIS T
is referred
LIS T
conta ins
this should be the case when you were
to
referring
to
referring
to
is referred
to modeling with one of the goals of entity
in referring to in
LIS T IT EM LIS T IT EM modeling in mind: create a model where the
4 -1 7
important real world things are represented only
once.
This can be put in other words: a correct entity
model should always allow any two entities to
become subtypes of a new super entity.
In model 3 and 4 entity B and C show expicitly
A rc a nd Sub types
what is sometimes referred to as implicit
A
1 2
A
subtypes of A, as entity B (C) can be defined as
R
P
Q P Q
those instances of A that refer to P (Q).
A C
A C B
A C B
B
3 4 5
R R
Q Q
P Q P P
4 -1 8
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Knowledge of this subtype/arc model is

Sub ty pes H ide R elation ships in A rc
important as one of the possible table
A A
is B
implementations (“arc implementation”) of the
B is
is
left-hand model is based on the right-hand
C C
is model.
• E v ery A • E v ery A m us t
is eith e r a B o r a C be a B or be a C
• E v ery B is a n A • E v ery B m us t b e an A
• E v ery C is a n A • E v ery C m us t b e an A
4 -1 9
Domains in the sense used by Oracle Designer

Va lu e se ts
C O D E TY P E
are only “known” within the context of the tool.
YESNO
# C ode
A
# Id
* N am e
* M ax Length
Domains are no ISO object, nor an Oracle
* D escription
GENDE R
B of D escription
database object.
# C ode
W EEKDAY
* D esc ription CODE
# C ode
* D escription
When you use Oracle Designer, you would very
# C ode
* D escription likely model the code/description entities as
domains. The important property of a domain is
4 -2 0 the fact that the values are known and set before
the system goes live.
You would model all the other ones as entities
as every now and then new values are
introduced.
Using the generic model leads to the need of
additional, though simple, coding. The multiple
table variant gives some loss of performance,
but may reduce the number of tables and
maintenance applications dramatically.
This comes down to a non-equijoin at table
O ther C on stra in ts: R an ge C heck
level. This constraint cannot be diagrammed.
E M P LO Y E E JO B
* N am e * Title
* A ddress * M inim um S alary betw een
w ith * M axim um S alary
of
for referring to
E M P LO Y M E N T
* S tart D ate
o E n d D ate
* S alary
4 -2 1
.........................................................................................................................................
®
4-37
.........................................................................................................................................
Note that you can model the Allowed

O ther C on stra in ts: State Value Tran sition
Transitions, as status pairs. This, however, does
not check the transitions. It only keeps
P ossible
to
information about what is allowed and what
Mar
Wid
Div
Sin
DP
M arital S tatu s
E M P LO Y E E
Transitions from
* N am e
* A ddress
* C urrent M arital S ta tus
S ingle
M arrie d
W idow ed
not. You must code an additional check to
D ivorc ed
D om estic P a rtnership compare the old and new value after an update
of Current Marital Status.
4 -2 2
In general, no constraints can be modeled that

C ond itio nal R elatio nship
depend on attribute values. ER models do not
C O N TR A C T
# Id
have a representation of values as the models
* S tandard In dicator S TA N D A R D basis for
consists
C O N D ITIO N
in
based
on
only deal with data structure.
of C U S TO M IZE D
C O N D ITIO N
in
part of
referring to referring to
C O N TR A C T C O M P O N E N T
4 -2 3
This picture is, of course, not 100% safe,

B oun daries
particularly when the nature of a rule changes,
u n re late d en tity
E X TE R N A L
# Id
* D escription
for example, if new legislation determines that
a n d p o ss ib le im p lem e n tatio n
* V alue
the tax on added value (sales tax) is a fixed
E X TE R N A LS
Id D e scription V a lue
value and not a percentage or when the
1
2
3
V a lue added tax %
M a xim um available S p ace per M ail U ser in M by te
M a xim um level of N ested M ail Folders
15
500
3
percentage depends on the nature of the
4 M a xim um level of N ested M ail Lists 16
product.
4 -2 4
Sum m ary
• Ide n tifica tio n

– C a n b e a rea l p rob le m in th e re a l w o rld
– M o d e ls ca n n ot o v erco m e th is
• E n tities m u st h a ve a t lea s t o n e U n iq u e Id e ntifie r
• U n iq u e Id e n tifiers c o ns is t o f attrib u te s o r
re la tio n sh ip s o r b o th
• A rc s
• M an y typ e s o f co n s tra in t a re n ot rep re se n te d
in E R m o d e l
4 -2 5
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................

Prac tices
• Ide n tifica tio n P lea se Practice 3day 4day

• Ide n tifica tio n
• M o o n lig h t U ID Identification Please Yes Yes
• T a b le s
• M o d e lin g C o n stra in ts Identification Yes Yes
Moonlight UID Opt Yes
Tables Opt Cha
4 -2 6 Constraints Cha Yes
Practice 4.3
rep ortin g to rep ort o f
D E P A RT M E N T
HQ rep ort o f The model is based on the result of practice 3-5:
repo rtin g to
OTHER
D E P A RT M E N T COUNTRY
O R G AN IZ A TIO N
in COUNTRY Moonlight P&O. If the delegates did not
of
w ith
EM P LO Y E E
w ith w ith
b elon g s to
w ith
in
perform that practice, spend a little time reading
of
w ith
fo r for
SHOP the model to make sure they understand the
to
P AY R O L L
E N TR Y
w ith
business context.
fo r to
AS S IG N M E N T as JO B
in
4 -3 4
Practice 4.4
Prac tice: Table 1
When your students do not know much about
“In a relational database sy stem , data is stored in tables. Tables of a
databas e user m ust have a unique nam e. A ta ble m ust have at lea st
one colum n. A colum n has a unique nam e w ith in the table. A colu m n tables and keys, this is a good practice.On one
m ust ha ve a data type and m ay be N ot N ull.
Tables can have one prim a ry key and any num ber of unique keys. A
key con tains one or m ore c olum ns of the table . A colum n can be part
hand it gives “business” information about
of m ore than one key.
A table can have foreign ke ys. A foreign key a lw ays connects one
table w ith another. A foreig n key consists of on e or m ore colum ns of
tables, columns, and keys, but on the other hand
the one table that refers to key colum ns of the other table.
The seq uence of colum ns w ithin the key and fo reign key is im portant.” the practice is about identification and
constraints.
4 -3 5
Practice 4.5.1
Pra ctice : C o nstrain ts 1
You can model constraints in 4.5.1 - 4.5.3 using
subtypes, UIDs, arcs or combinations of these.
EM PL O Y E E
# N am e m anag ed by
m anager of
4 -3 7
.........................................................................................................................................
®
4-39
.........................................................................................................................................
Practice 4.5.2
Typical example of supertype modeling where
ow ned
various entities share the same namespace.
USER by LIS T
ow ner of
# N am e # N am e
ow n ed
ow ner of by N IC K N A M E
# A lias
4 -3 8
Practice 4.5.3
This is the most complex of the three.
w ith
subfolder
w ithin
U SE R
# N am e
The endless loop (see 4.2.5) can be broken
F O LD E R
N am e ow ner
of because the top level folder is identified
ow ned
by
differently.
4 -3 9
.............................................................................................................................................
.........................................................................................................................................

Name: Full Email:

Lesson number: 4

Detail comments
Content Page
Type Number
.........................................................................................................................................
®
4-41
.........................................................................................................................................
.............................................................................................................................................
Suggested Graphics
.........................................................................................................................................
Suggested Graphics

.........................................................................................................................................
®
4-43
.........................................................................................................................................

520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.............................................................................................................................................
5
.................................
Modeling Change
.........................................................................................................................................
Introduction
Every update of an attribute or transfer of a relationship means loss of information.
Often that information is no longer of use, but some systems need to keep track of
some or all of the old values of an attribute. This may lead to an explicit time
dimension in the model which is usually quite a complicated issue.
Lesson Aim
Time is often present in a business context, as many entities are in fact a representation
of an event. This lesson discusses the possibilities and difficulties that arise when you
incorporate time in your entity model.
Schedule
About the slide

See Page 25
• Date and time
• Modeling change over time
• Prices change
• Journalling
5-2
Topic See Page

Introduction 2
Time 4
Date as Opposed to Day 5
Entity DAY 6
Modeling Changes Over Time 7
A Time Example: Prices 10
Journalling 17
Summary 19
Practice 5—1: Shift 20
.............................................................................................................................................
Introduction
.........................................................................................................................................
Topic See Page

Practice 5—2: Strawberry Wafer 21
Practice 5—3: Bundles 22
Practice 5—4: Product Structure 24
Objectives
• Make a well considered decision about using entity DATE or attribute Date
• Model life cycle attributes to all entities that need them
• List all constraints that arise from using a time dimension
• Cope with journalling
.........................................................................................................................................
®
5-3
.........................................................................................................................................
Time
Modeling Time
In many models time plays a role. Often entities that are essentially events are part of a
model, for example, PURCHASE, ASSIGNMENT. One of the properties you record
about these entities is the date or date and time of the event. Often the date and time
are part of a unique identifier.
About the slide
Change and Time See Page 25
• Every update means loss of information.

• Time in your model makes the model more
complex.
• There are often complex join conditions.
• Users can work in advance.
• When do you model date/time as an entity?
• What constraints do arise?
• How do you handle journalling?
A second time-related issue often helps to increase the usability of a system

dramatically. By adding dates like Start, Expiry, End Date, to data in the system, you
allow users to work in advance. Suppose a particular value, say the price of gas or
diesel, will change as of January 1. It is very useful to be able to tell the system the
new value long before New Year’s Eve. By adding a time dimension to the model you
make the system independent of the now.
As always, there is a price for adding things such as this. Adding a time dimension to
your conceptual data model makes the model considerably more complex. In
particular, the number of constraints and business rules that must be checked will
increase.
A third time-related issue in conceptual data models is connected to the concept of
logging or journalling. Suppose you allow values to be updated, but you want to keep
track of some of the old values. In other words, what do you do when you need to keep
a record of the history of attribute values, of relationships, of entire entities?
The following issues arise:
• When do you model date/time as an entity, and when as an attribute?
• How do you handle the constraints that arise in systems that deal with time-related
data?
• How do you handle journalling?
.............................................................................................................................................
Date as Opposed to Day
.........................................................................................................................................
Date as Opposed to Day

Probably all current operating systems and database systems have types “date” and
“time” available that know, for example, that 29-OCT-1983 was a Saturday in the 10th
month, called October, of 1983.
Some database systems, like Oracle, see time as a component of a date and store them
in one. Knowing that, you are likely to decide that dates can be modeled as attributes
with the format date.
About the slide

Entity DAY or Attribute Date See Page 26
PURCHASE
on
Single attribute entity without m:1 relationships
is usually replaced by attribute
of
PURCHASE
DAY * Date
# Date
5-4
A day, however, is not just a date. My great-grand father was born on a day in 1852,
but the exact date is unknown. A Genealogical Register System should therefore be
able to store fragments of a date, such as “1852”, or even a description, such as
“around 1765”.
Systems that store historical information often have to deal with several dates for one
event, according to multiple sources with nonidentical information.
Some systems have to take dates in conjunction with the reliability of that date.
Clearly, in these cases a simple attribute would not suffice.
Loosely speaking, when you are interested in the date only, and these dates are known
to the user, model an attribute; on the other hand, when you are interested in the day,
model it as an entity with attribute Date, which is possibly a unique identifying
attribute.
.........................................................................................................................................
®
5-5
.........................................................................................................................................
Entity DAY
It is not only systems that deal with historical information that struggle with dates.
Sometimes a system needs to know more about a day than can be derived from its
date. A planning system, for example, often needs to know if a particular day is a
public holiday. Many data warehouse systems use a calendar that is different from the
normal one, for example, where a year is divided into four-week periods or 30 day
Months or Quarters where Q1 starts in the middle of May.
Some warehouses need weather information about days in order to do statistical
analysis about the influence of the weather on, for example, their sales. In these cases
a day has attributes or relationships of its own and should be modeled as entity DAY.
About the slide

Entity DAY See Page 26
DAY
# Date
* Public Holiday Indicator
first day of
starts on
TASK for TASK

ASSIGNMENT in # Id
* Duration in Hours
of EMPLOYEE
with # Name
5-5
The above model shows part of a planning system where tasks are assigned to
employees. Tasks may take from a few hours to, at maximum, several days.
Based on this model, table TASK_ASSIGNMENTS will contain a date column that is
a foreign key column to the DAYS table.
Date and Time

As stated earlier, an Oracle DATE column always contains date and time. This needs
some special attention as two DATE columns may apparently contain the same date
but they are not equal because of a difference in their time component.
While modeling, always make explicitly clear when time of the day is an issue, for
instance, by naming the attribute DateTime. As soon as hours and minutes play a role,
the concepts of “time zone” and “daylight saving time” may become important.
.............................................................................................................................................
Modeling Changes Over Time
.........................................................................................................................................

Date and Time in your models may substantially increase the complexity of your
system, as the next example shows.
The context for this example is that of an Embassy Information System, but could
have been chosen from almost any business area.
Embassy employees have an assignment for a country, but, of course, the assignments
may change over time. Therefore, the model would need an entity ASSIGNMENT
with a mandatory attribute Start Date and an optional End Date. Start Date is modeled
as part of the UID for ASSIGNMENT. This means that the model allows an employee
to have two assignments in the same country, as long as they start on different days. It
also allows the employee to have two assignments that start on the same day, as long
as these are for different countries.
Suppose we know today that Jacqueline will switch from Chili to Morocco on the first
of next month. This fact can be fed into the system immediately, by creating a new
instance of ASSIGNMENT with a Start Date that is still in the future at create time.
The future users will appreciate this kind of functionality.
EMPLOYEE COUNTRY
# Id # Name
of in
for as
ASSIGNMENT
# Start Date
o End Date
End Date Redundant?

You may argue that attribute End Date of ASSIGNMENT is redundant because
Jacqueline’s assignments follow each other: the End Date of Jacqueline’s assignment
in Chili matches the Start Date of the one in Morocco. This may be true, but it does not
take into consideration that embassy people may take a leave and return after a couple
of years. In other words, if you do not model attribute End Date you ignore the
possibility that the assigned periods of a person are not contiguous.
Note that the model does allow an employee to have two assignments in, for example,
Honduras, that overlap! The unique identifier does not protect the data against
overlapping periods. Adding End Date to the UID does not help.
You would need a whole series of extra constraints to cope with this.
.........................................................................................................................................
®
5-7
.........................................................................................................................................
Countries Have a Life Cycle Too

Suppose the Embassy Information System contains data that goes back to at least the
late eighties. In those days the USSR and Zaire were still countries. Suppose there are
ASSIGNMENTS that refer to the USSR and Zaire. In the case of Zaire, you could
consider an update of the Name of the COUNTRY: Democratic Republic Congo is
essentially just the new name for Zaire. In case of the USSR this would not make
sense. There is not a new name for the old country. The old country simply ceased to
exist when it broke into several countries. Although the concept of a country seems
very stable, countries may change fundamentally during the lifetime of the
information system.
This leads to the next model.
About the slide

Even a Country Has a Life Cycle See Page 26
COUNTRY
# Name
EMPLOYEE # Start Date
# Id * End Date
of in
life cycle
for
attributes
as
ASSIGNMENT
# Start Date
o End Date
5-7
Time-related Constraints
Be aware of the numerous constraints that result from the time dimension! Here is a
selection:
• An ASSIGNMENT may only refer to a COUNTRY that is valid at the Start Date
of the ASSIGNMENT.
• The obvious one: End Date must be past Start Date.
• A business rule: ASSIGNMENT periods may not overlap. The Start Date of an
ASSIGNMENT for an EMPLOYEE may not be between any Start Date and End
Date of an other ASSIGNMENT for the same EMPLOYEE.
• As for the previous constraint, but for End Date.
.............................................................................................................................................
.........................................................................................................................................
• You would probably not allow an ASSIGNMENT to be transferred to another

COUNTRY, unless the ASSIGNMENT has not yet started, that is, the Start Date of
the ASSIGNMENT is still in the future.
This is an example of conditional nontransferability.
For updates of the attribute Start Date here are some possible constraints:
• A Start Date of an ASSIGNMENT may be updated to a later date, unless this date
is later then the End Date (if any) of the COUNTRY it refers to.
• A Start Date of an ASSIGNMENT may be updated to a later date, if the current
Start Date is still in the future.
• A Start Date of an ASSIGNMENT may be updated to an earlier date, unless this
date is earlier than the Start Date of the COUNTRY it refers to.
• A Start Date of an ASSIGNMENT may be updated to an earlier date, if this new
date is still in the future.
• A Start Date of a COUNTRY may be updated to a later date, if there are no
ASSIGNMENTS that would get disconnected.
Similar constraints apply to attribute End Date.
Referential Logic
Note that, except for two, these constraints result from referential logic only. There
may be more additional business constraints.
Imagine the sheer number of constraints if a time-affected entity is related to several
other time-affected entities! Fortunately, these constraints all have a similar pattern;
these result from the referential, time related, logic.
Not in Diagram
You cannot model any of these constraints in the diagram as they all have to be listed
separately.
Implementation
In an Oracle environment, one of these constraints can be implemented as a check
constraint, (End Date must be later than Start Date). All the others will be
implemented as database triggers.
.........................................................................................................................................
®
5-9
.........................................................................................................................................
A Time Example: Prices

About the slide
Products and Prices See Page 27
PRODUCT
# Id
* Name
with
PRICE =
of PRICED PRODUCT=
PRICE HISTORICAL PRICE
* Price in $
# Start date
o End Date
5-8
Products have a price. Prices change. Old prices are probably of interest. That leads to
a model with entities PRODUCT and PRICE. The latter entity contains the prices and
the time periods they are applicable. In real-life situations you find the concept of
PRICE also named PRICED PRODUCT, HISTORICAL PRICE (and less appropriate:
price list or price history); all these names more or less describe the concept.
You may argue the need for an End Date attribute. If the various periods of a product
price are contiguous, End Date is obsolete. If, on the other hand, the products are not
always available, as in the fruit and vegetable market, the periods should have an
explicit End Date.
.............................................................................................................................................
.........................................................................................................................................
Introducing Order Header and Order Item
About the slide

What Price to Pay? See Page 27
ORDER HEADER
PRODUCT referred # Id
# Id by
* Order Date
* Name
n
ee
with
with
tw
be
of
of
ORDER ITEM
PRICE
* Quantity Ordered
* Price in $ referring
# Start date to
o End Date
5-9
Here, entities ORDER HEADER and ORDER ITEM are introduced. An ORDER
HEADER holds the information that applies to all items, like the Order Date and the
relationship to the CUSTOMER that placed the order or the EMPLOYEE that handled
it. (For clarity, these relationships are not drawn here.) The ORDER ITEM holds the
Quantity Ordered and refers to the PRODUCT ordered. The price that must be paid
can be found by matching the Order Date between Start Date and End Date of PRICE.
Note that you cannot model this “between relationship”.
This model is a fairly straightforward product pricing model and is often used.
Order
Note that the concept of an order in this model is composed of ORDER HEADER and
ORDER ITEM.
To find the order total for an order, it would need a join over four tables.
.........................................................................................................................................
®
5-11
.........................................................................................................................................
Price List
A variant on the above model is often used when prices as a group are usually changed
at the same time. The period that prices are valid is the same for many prices; that
would lead to this model:
Price List Search
between
PRICE LIST
ORDER HEADER
# Id
PRODUCT # Id
* Start Date
o End Date
# Id * Order Date
* Name referred with
with with by
on of
of
ORDER ITEM
PRICED PRODUCT
* Quantity
Ordered
to
5-10
Entity PRICE LIST represents the set of prices for the various products; PRICED
PRODUCT represents the price list items. To know the price paid for an ordered item,
you take the Order Date of the ORDER HEADER, and take the PRICE LIST that is
applicable at that date. Next, you go from ORDER ITEM to the PRODUCT that is
referred to and from there to the PRICED PRODUCT of the PRICE LIST you have
just found. To find the order total for an order, it would need a join over five tables.
.............................................................................................................................................
.........................................................................................................................................
Buying a PRODUCT or a PRICED PRODUCT?

Another variant of a pricing model is shown here.
About the slide

See Page 27
Order for Priced Products
PRICE LIST
ORDER HEADER
# Id
PRODUCT # Id
* Start Date
o End Date
# Id * Order Date
* Name with
with with
on of
of
referred by ORDER ITEM
PRICED PRODUCT
* Quantity
Ordered
to
5-11
Here an ORDER ITEM refers directly to a PRICED PRODUCT. At create time of the
ORDER ITEM the constraint is applied that the Order Date must mach the correct
PRICE LIST period. To find the order total for an order now only requires three tables.
.........................................................................................................................................
®
5-13
.........................................................................................................................................
Negotiated Prices
Negotiated Prices
PRICE LIST
ORDER HEADER
# Id
PRODUCT # Id
* Start Date
o End Date
# Id * Order Date
* Name referred by with
with with
on of
of
ORDER ITEM
PRICED PRODUCT
* Quantity Ordered
* Price in $ referring * Negotiated Price
to
5-12
When prices are subject to negotiation, the model becomes simpler. Negotiated Price
is now an attribute of entity ORDER ITEM; ORDER ITEM refers to PRODUCT.
Every referential constraint can be modeled.
This model may seem to hold derivable information, but this is not true. Even in the
case that almost all Negotiated Prices are equal to the current product price, you have
to model Negotiated Price at ORDER ITEM level, just because of the small chance of
an exception. To find the order total you require only two tables. You can imagine that
many analysts choose this variant of the model as a safeguard, even if there is nothing
to negotiate at present.
.............................................................................................................................................
.........................................................................................................................................
Which Variant to Use and When?

Typically, the model with the negotiated prices will occur where the number of
ORDER ITEMS per ORDER HEADER is low, often just a single one, and where the
value is high, as, for example, in the context of a used car business.
You see ORDER ITEM referring to a PRODUCT most often in the situation where
prices do not change frequently. The number of items per ORDER HEADER is often
well over one, and the overall value limited. Typical examples are the fashion industry
and grocery stores.
The model with ORDER ITEM referring to PRICED PRODUCT is often used in
businesses where prices often change, as in the fresh fruit and vegetable markets.
Prices there may even change during the day.
The model with attribute Current Price for a PRODUCT is typically the model for the
supermarket environment where instant availability of prices at the checkouts is vital.
As stated earlier, the best model for a particular context depends on functional needs.
See more on this in the chapters on Denormalized Data and Design Considerations.
.........................................................................................................................................
®
5-15
.........................................................................................................................................
Current Price
About the slide
Current Prices See Page 28
PRODUCT PRODUCT PRODUCT

# Id # Id # Id
* Name * Name * Name
* Current Price * Current Price with
with with
old
of of of
PRICE PRICE PRICE
* Price in $ * Price in $ * Price in $
# Start Date # Start Date # Start date
* End Date o End Date o End Date
o Current Indicator
5-13
These models are variants on the PRODUCT-PRICE model you have seen before.
In the left-hand model the 1:m relationship between PRODUCT and PRICE shows the
real historical prices only. You can guess that only historical prices are kept because
attribute End Date is mandatory; an additional constraint is that this value should
always be in the past. The Current Price of a PRODUCT is represented as an attribute.
This model does not have any redundancies.
In many situations it would be a good design decision to keep the current product
prices as well as the old prices in one table based on entity PRICE. The middle model
is an ER representation of that situation. Note that End Date is now optional.
The right-hand model is another model that contains a subtle redundancy. See more on
this type of redundancy in the lesson on Denormalized Data.
.............................................................................................................................................
Journalling
.........................................................................................................................................
Journalling
When a system allows a user to modify or remove particular information, the question
should arise if the old values must be kept on record. This is called logging or
journalling. You will often encounter this when the information is of a financial
nature.
Consequences for the Model
About the slide

Journalling See Page 28
by
PAYMENT to
by o Date Paid
PAYMENT to * Amount in $
*Date Paid
with
*Amount in $
of
AMOUNT
MODIFICATION
* Old Amount in $
* Modified by
* Date Modification
5-14
A journal usually consists of both the modified value and the information about who
did the modification and when it was done. This extra information can, of course, be
expanded if you wish.
Apart from the consequences for the conceptual data model, the system needs special
journalling functionality: any business function that allows an update of Amount In
should result in the requested update, plus the creation of an entity instance AMOUNT
MODIFICATION with the proper values. Of course, the system would need special
functions as well in order to do something with the logged data.
No Journal Entity
When several, or all, attributes of an entity need to be journalled, it is often
implemented by maintaining a full shadow table that has the same columns as the
original plus some extra to store information about the who, when, and what of the
change. This table does not result from a separate entity; it is just a second, special,
implementation of one and the same entity.
.........................................................................................................................................
®
5-17
.........................................................................................................................................
Journalling Registers Only

Note that logging does not prevent a user from making updates. Preventing updates
entirely is a functional issue and is invisible in the conceptual data model. Be aware
that preventing updates altogether would also block the possibility to change typos or
other mistakes.
At this stage, decisions must be made about the behavior of the system with respect to
updates; sometimes this leads to modifications in the conceptual data model.
For example, suppose that in a particular business context a certain group of users is
allowed to create instances of PAYMENT but is not allowed to change them. Changes
can only be made by, say, a financial manager. Suppose you just created a PAYMENT
instance and you discover you made a mistake. For those cases the business would
need some mechanism to stop the erroneous instance. One mechanism would be to ask
one of the financial managers to make the change. A far better mechanism would be to
add functionality so that a payment can be neutralized. This may be represented in the
model as an attribute Neutralized Indicator that users can set to Yes.
.............................................................................................................................................
Summary
.........................................................................................................................................
Summary
Every update in a system means loss of information. To avoid that you can create your
model to keep a history of the old situations. Sometimes relationships refer to a time-
dependent state of an entity. In other words, the updated entity is in fact a new instance
of the entity and not an updated existing instance. If this is the case, the time-
dependent referential constraints cannot be modeled by a relationship only.
Time in your model is a complicated issue. Many models have some time-related
entities.
Summary
• Consider the need for keeping old values.

• Time in your model is complicated:
– Implicit versions
– References
• Journalling
5-15
.........................................................................................................................................
®
5-19
.........................................................................................................................................
Practice 5—1: Shift

Moonlight Coffees
Goal (See Page 29)
The purpose of this practice is to model various aspects of time.
Scenario
Some shops are open 24 hours a day, seven days a week. Others
close at night. Employees work in shifts. Shifts are subject to local legislation. Below
you see the shifts that are defined in one of the shops in Amsterdam.
Your Assignment
List the various date/time elements you find in this Shift scheme and make a
conceptual data model.
Practice: Shift
Museumplein, Amsterdam, March 21
Shift 1 2 3 4 5
Mon 6:30 11:30 16:00 20:30 -

11:30 16:00 20:30 23:00
7:00 11:30 16:00 20:30 -
Tue
11:30 16:00 20:30 23:00
7:00 11:30 16:00 20:30 -
Wed
11:30 16:00 20:30 23:00
7:00 11:30 16:00 20:30 -
Thu
11:30 16:00 20:30 23:00
7:00 11:30 16:00 20:30 -
Fri 11:30 16:00 20:30 24:00
8:00 11:30 15:00 18:00 21:00
Sat/Sun
11:30 15:00 18:00 21:00 24:00
5-17
.............................................................................................................................................
Practice 5—2: Strawberry Wafer
.........................................................................................................................................
Practice 5—2: Strawberry Wafer

Moonlight Coffees
You have modeled a price list in an earlier lesson. Now some new
information is available.
Your Assignment
Revisit your model and make changes, if necessary, given this extra information.
Prices are at the same level within a country; prices are determined
by the Global Pricing Department. Usually the prices for regular,
global products are re-established once a year.
Prices and availability for local specialties are determined by the
individual shops. For example, the famous Norwegian Vafler med
Jordbær (a delicious wafer with fresh strawberries) is only available
in summer. Its price depends on the current local market price of
fresh strawberries.
SULMVOLMVW GH.H\]HU.H\]HUOHL$QWZHUSHQ
EH]RHNWRQVRS¶W:HEZZZPRRQOLJKWFRP
klein middel groot

gewone koffie 60 90 120
cappuccino 90 110 140
koffie verkeerd 75 100 130
speciale koffies 99 125 150
espresso 60 95 110
koffie van de dag 45 75 100
caffeine vrij 5 10 15 toeslag
zwarte thees 60 100 120
vruchten thees 75 110 130
kruiden thees 80 120 140
dag thee 50 85 100
caffeine vrij 5 10 15 toeslag
frisdranken 60 100 130
diverse sodas 60 100 130
mineraal water 75 120 140
U appel taart 180
: H
7
%
E brusselse wafel 150
P
I H
H
L
W
S
portie chocolade bonbons 150
V
H
X
O
F
6
koekje van eigen deeg 120

Q
L portie slagroom 30
5-19
.........................................................................................................................................
®
5-21
.........................................................................................................................................
Practice 5—3: Bundles

Moonlight Coffees
Goal (See Page 30)
The purpose of this practice is to expand the concept of an old
entity.
Scenario
As a test, Moonlight sells bundled products in some shops, for a special price. Here are
some examples.
A SweetTreat(tm) consists of a large soft drink plus cake of

the day.
A BigBox(tm) consist of a large coffee of the day plus two
cakes of the day.
A SuperSweetTreat(tm) consists of a SweetTreat(tm) plus
whipped cream (on the cake).
A FamilyFeast(tm) consists of two BigBoxes(tm) plus two
SweetTreats™ plus a small surprise.
Bundles sell very well; all kinds of new bundles are expected to come.
The system should know how all these products are composed, in order to complete
various calculations.
Your Assignment
1 Modify the product part of the model in such a way that the desired calculations
can be completed.
PRODUCT GROUP
# Name
classification
for
classified
as
PRODUCT
# Id
* Name
.............................................................................................................................................
Practice 5—3: Bundles
.........................................................................................................................................
2 Change the model in such a way that it allows for:
A DecafPunch(tm) consists of a regular decaffeinated coffee

or a regular decaffeinated tea, plus a blackberry muffin.
.........................................................................................................................................
®
5-23
.........................................................................................................................................
Practice 5—4: Product Structure

Moonlight Coffees
Goal (See Page 30)
The purpose of this practice is to model a hierarchical structure.
Scenario
Moonlight needs to make sales information available as a tool to
optimize its business. A hierarchical product structure is being developed to be able to
report on different summary levels. This hierarchical structure should replace the
single level product group classification. Below you see the current idea about a
product structure. This structure is far from complete, but it should give you an idea of
the shape the structure will take. The + signs mean that the structure will be expanded
at that point.
Your Assignment
1 Create a model for a product classification structure.
+ Products
+ Drinks
+ Coffees
Regular
Cappuccino
Café Latte
+ Special Coffee
Teas
+ Black
Chinese
Indian
English
+ Infusions
+ Herbal
Soft drinks
Juices
Orange
Grape
+ Waters
+ Sodas
+ Dairy Products
+Foods
+ Pastry
+ Candy Bars
+ Local Specialties
+Non Foods
Merchandise
CDs
+ Stationary
Other
+ Tickets
+ Art
5-22
2 (Optional) How would you treat the bundled products?
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Instructor Notes
Topic Timings
Lecture 15-20 minutes
55
Practice 45-100 minutes
M odelin g C h ange
Total 60-120 minutes
Before you start with lesson 5

It is a good idea to summarize what you have
covered so far. A nice way to do that is to create
an ER model about the concepts they have seen:
entities, attributes, relationships, unique
identifiers, arcs. You will find this a very
valuable activity. You may use
ERSUMMARY.ppt that contains a build of the
model. You find this in the PowerPoint
presentation as a separate file.
In many ways change (and therefore, time) is
O verview
part of an ER model. This is often a hugely
• D a te a n d tim e complicating factor. Suddenly you must be able
• M o d e lin g c h an g e o ve r tim e
• P ric es ch an g e to cope with information such as “in her
• J o u rn allin g
capacity as senior counsellor, Jane worked
between April ’95 and July’97 for ABC Inc.
that has been part of XYZ International since
5 -2
May ’99.
The difference between Day and Date is
discussed as well as the need for an entity DAY.
Various examples of modeling price changes
are presented and discussed.
Time in a model is a complication, but you do
C h ange an d Tim e
not have much choice other than to accept it.
•
•
E ve ry u pd ate m ea n s lo ss o f in fo rm a tio n .
T im e in yo u r m o d el m a ke s th e m o d e l m o re
At the other hand, a well-designed time-related
c o m p le x.
• T h e re are o fte n c o m p le x jo in c on d ition s. system can be very user friendly and functional,
•
•
U s ers c an w o rk in a d van ce .
W h e n do yo u m o de l da te/tim e a s an e n tity?
coping with simple things such as working in
•
•
W h a t c o n stra in ts d o a ris e ?
H o w d o yo u h a n d le jo u rn a llin g ?
advance (like being able to enter a new price or
code that will be effective as of tomorrow) to
5 -3 very complex calculations over periods of time.
.........................................................................................................................................
®
5-25
.........................................................................................................................................
A date is the identifying attribute of entity DAY.

E ntity D A Y or A ttribu te D ate
When you can use a nice rich data type like
P U R C H A SE
Oracle’s DATE you do not need to worry about,
for example, the weekday, month name, or
on
S ingle attribute entity w ithout m :1 rela tionships
is usually repla ced by attribute quarter of a particular day. Choosing not to
of
DAY
P U R C H AS E
* D ate
model this means you assume this will be
# D a te
available at implementation time. If not, you
5 -4
should model all these date-related pieces of
information as separate entities and attributes!
You need entity DAY when nonderivable
Entity D A Y
attributes are needed, such as whether or not a
DAY
# D ate
* P ublic Ho liday In dicato r
day is a public holiday or a bank holiday.
first day of
starts on
Many planning systems use a list of days, for
T AS K
A SS IG N M E N T
for
in
TA S K
# Id
instance to facilitate reporting on individual
* D ura tion in H ours
of E M PL O Y EE schedules.
w ith # N am e
5 -5
M odelin g C h ange
E M P LO Y E E COUNTRY
# Id # N am e
of in
for as
A SS IG N M E N T
# S tart D a te
o E nd D a te
5 -6
Be aware that because that particular things

Eve n a C oun try H as a L ife C y cle
have not changed during your lifetime does not
COUNTRY mean they will never change.
# N am e
# Start Da te
E M P LO Y E E
# Id * E nd Da te Countries that cease to exist, currencies that
of in
life cycle
attribu tes
change, telephone numbers that are no longer
for as
A SS IG N M E N T
# S tart D a te
related to the physical hardware, are all
o E nd D a te
examples of change that many systems did not
5 -7 foresee.
Referential logic is only concerned with a
reference to a record in another table, not with a
reference to a valid record (according to some
rule) in another table. This is why you need to
make many extra checks in these situations.
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
The next few slides show all variants of

P roduc ts a nd Pric es
product-price models. These are used here as a
PRODUCT real-life metaphor of a time-related problem.
# Id
* N am e
w ith
P RICE =
Another name for the PRICE entity you run into
P R IC E
of P RICE D P R O D U C T =
H IS TO R IC A L PR IC E quite often is PRICE HISTORY although this
* P rice in $
# Start date
o E nd D ate
name is, strictly speaking, not correct (similar
to using the name “company” for the table of
5 -8 employees).
None of the time-related constraints can be
W h at P rice to Pay ?
modeled.
O R D ER H EA D E R
PR O D U C T referred # Id
# Id by
* O rde r D ate
* N am e
en
w ith
e
w ith
tw
be
of
of
O R D ER IT E M
P RICE
* Q u antity O rde red
# S tart da te to
o E nd Da te
5 -9
Price List S earch
b etw ee n
P R IC E L IS T
O R D E R H E AD E R
# Id
PRODUCT # Id
* S tart D a te
o E nd D ate
# Id * O rd er D a te
* N am e referred w ith
w ith w ith by
on of
of
O R D ER ITE M
P RICE D P R O D U C T
* Q uantity
O rdere d
to
5 -1 0
Sometimes it makes sense to model things in

O rd er fo r Priced Pro duc ts
such a way that you do not order a product but a
P R IC E L IS T
priced product as it is called here. It is a
# Id
* S tart D a te
o E nd D ate
PRODUCT
# Id
# Id
* O rd er D a te
different way of seeing the same thing happen.
* N am e w ith
w ith w ith The consequence during implementation is
of
on
of
referred by O R D ER ITE M
* Q uantity
different tables and different join conditions.
O rdere d
to
5 -1 1
N eg otiate d P rice s
P R IC E L IS T
# Id
PRODUCT # Id
* S tart D a te
o E nd D ate
# Id * O rd er D a te
* N am e referred by w ith
w ith w ith
on of
of
O R D ER ITE M
* Q u antity O rd ered
* Price in $ referring * N eg otiate d P rice
to
5 -1 2
.........................................................................................................................................
®
5-27
.........................................................................................................................................
A few more models to be more or less

C urrent Pric es
complete. The last one is used in the lesson on
PR O DU C T PR O D U C T P RO DU CT
# Id
* Name
# Id
* N am e
# Id
* Name
Denormalized Data.
* C u rre nt P rice * C u rre nt P rice w ith
w ith w ith
old
of of of
P RIC E P RIC E P R IC E
* Price in $ * Price in $ * P rice in $
# S ta rt D ate # S ta rt D ate # S tart date
* En d D ate o E nd D ate o E nd D ate
o C urren t Ind icator
5 -1 3
Journalling is often seen as a functional issue;

J ourna llin g
with Oracle Designer it is very easy to get the
by
P A YM E N T
general journalling done.
by to
o D a te P aid
PAYMENT
*D ate P aid
to * A m ou nt in $ However, if you only need partial journalling, it
w ith
*A m ou nt in $
of
makes sense to explicitly model it. This allows
AMOUNT
M O D IFIC A T IO N
* O ld A m o unt in $
you to describe constraints that apply (for
* M od ified b y
* D ate M odification example, only log when the amount is
5 -1 4 increased), and the desired functionalities and
responsibilities.
Sum m ary
• C o n s id er th e n e ed fo r k e ep in g o ld va lu e s.
• T im e in yo u r m o d el is co m p lic ated :
– Im p lic it v ers io n s
– R e fere n ce s
• J o u rn allin g
5 -1 5

Prac tices
• S h ift Practice 3day 4day

• S tra w b e rry W afer
• B u n d le s Shift Yes Yes
• P ro d u ct S tru c tu re
Strawberry Wafer Opt Cha
Bundles Cha Cha
Product Structure Opt Cha
5 -1 6
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Practice 5.1
Pra ctice : Sh ift
M useum plein, A m sterdam , M arch 2 1
Time and weekday here are both independent of
S hift
M on
1 2 3
6:30 11:3 0 1 6:00 20: 30

4 5
-
date. The date at the top of the list, March 21, is
probably a date, but could be a day without a
1 1:30 16:0 0 2 0:30 23: 00
7:00 11:3 0 1 6:00 20: 30 -
Tue
1 1:30 16:0 0 2 0:30 23: 00
W ed 7:00 11:3 0 1 6:00 20: 30

1 1:30 16:0 0 2 0:30 23: 00
7:00 11:3 0 1 6:00 20: 30
-
-
year (such as the new schedule at the start of
Thu
Fri
1 1:30 16:0 0 2 0:30 23: 00
7:00 11:3 0 1 6:00 20: 30
1 1:30 16:0 0 2 0:30 24: 00
-
each new season).
8:00 11:3 0 1 5:00 18: 00 21 :00
S at/S un 1 1:30 15:0 0 1 8:00 21: 00 24 :00
5 -1 7
Practice 5.2
Prac tice: Straw berry W afer
The global products and local products behave
P rices are at the sam e level w ithin a country; p rices are determ ine d
in the same way the same from a selling point
b y the G lobal P ricing D epartm ent. U sually the prices for regular,
g lobal products are re-established once a year.
P rices and availability for local specialties are determ ined by the
of view but act differently from the perspective
in dividual shops. For exam ple, the fam ous N orw egian V afler m ed
J ordbæ r (a delicious w afe r w ith fresh straw berries) is only available
in sum m er. Its price dep ends on the current lo cal m arket price of
of pricing.
fresh straw berries.
This practice is again another variant of a

pricing model, with a different method for the
5 -1 8
global and local prices.
GH.H\]HU.H\]HUOHL$QWZHUSHQ
This price list is intentionally a version other
S ULMVOLMVW E H]RHNWRQVRS¶W: HEZZZP RRQOLJKWFRP
gew one koffie

klein
60
m iddel
90
groot
120
than the one that was printed in practice 3.6.
cappuccino 90 110 140
koffie verkeerd
speciale koffies
espresso
75
99
60
100
125
95
130
150
110
This implicitly describes something about
koffie van de dag
caffeine vrij
zw arte thees
45
5
60
75
10
100
100
15
120
toeslag currencies and language.
vruchten thees 75 110 130
kruiden thees 80 120 140
dag thee
caffeine vrij
50
5
85
10
100
15 toeslag
The question arises if there are basic differences
frisdranken 60 100 130
U
H
diverse sodas
m ineraal w ater
appel taart
60
75
100
120
130
140
180
between the two lists, or if it is just a translation
:
brusselse w afel 150
issue.
7 E
%
P
I H
H
L W
S
portie chocolade bonbons 150
V H
X
O
F
6
koekje van eigen deeg 120
Q
L portie slagroom 30
5 -1 9
.........................................................................................................................................
®
5-29
.........................................................................................................................................
Practice 5.3.1
This practice results in a model for a Bill of
A Sw ee tT reat (tm ) c o n sists of a larg e so ft d rin k p lu s ca k e o f
the d a y.
A B ig B o x (tm ) co n s ist o f a la rg e co ffee o f th e d a y p lu s tw o
Material.
ca ke s o f th e d a y.
A S u p e rSw ee tT re at (tm ) c o n sists of a S w e etT re at (tm ) p lu s
w h ip p e d crea m (o n th e c ak e).
Let the students sweat!
A F am ily F ea st (tm ) co n s is ts o f tw o B ig B o x es (tm ) p lu s tw o
S w e e tT re ats ™ p lu s a sm all su rp rise . Many will model a recursive relationship of
A D ec afP un c h (tm ) co n s is ts o f a reg ular d eca ffein a te d c o ffee
o r a re g u lar d ec affein a te d tea , plu s a b lac kb e rry m u ffin.
PRODUCT or will have a separate BUNDLE
entity.
5 -2 0
Why is this all incorrect?

• A product can be part of various bundles in
different quantities (like cake of the day that
occurs once in a StweetTreat and twice in a
BigBox).
• An instance of BUNDLE can consist of
other instances of BUNDLE (like a
SuperSweetTreat), so BUNDLE should
have a recursive relationship as well.
Practice 5.3.2
This is a hard one. Modeling the or makes it
much more complex. Do not let them spend too
much time on this one.
+ P rod u cts
+ D rin ks
Practice
+ Co ffees
R eg u la r
T eas
C ap p uc cin o
C afé L atte
+ S p ec ia l C offe e
Typical classification for desired decision
+ B la ck
Ch in ese
Ind ian
E n g lish
support functionality. How did Herbal teas do it
+ In fus io n s
+ H erb al
S o ft d rin ks
J u ic es
O ra ng e
compared to Black teas? How did teas do it
G ra pe
+ W ate rs
+ S o d as
+ Dairy P ro du cts
+Fo o d s
compared to coffees?
+ P astry
+ Can d y Ba rs
+ Lo ca l S p ec ia lties
+No n F o od s
M erch an dise
C Ds
+ S ta tio n ary
O the r
+ T icke ts
+ A rt
5 -2 2
.............................................................................................................................................
.........................................................................................................................................

Name: Full Email:

Lesson number: 5

Detail comments
Content Page
Type Number
.........................................................................................................................................
®
5-31
.........................................................................................................................................
.............................................................................................................................................
Suggested Graphics
.........................................................................................................................................
Suggested Graphics

.........................................................................................................................................
®
5-33
.........................................................................................................................................

520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.............................................................................................................................................
6
.................................
Advanced Modeling
Topics
.........................................................................................................................................
Introduction
Lesson Aim
This lesson gives an overview of patterns you can discover in data models. This lesson
introduces some generic models. You can use these to make your model withstand
future changes that are predictable but not yet known.
Objectives
Schedule
About the slide

See Page 28
• Patterns
• Drawing conventions
• Generic modeling
6-2
Topic See Page

Introduction 2
Patterns 4
Master Detail 5
Basket 6
Classification 7
Hierarchy 8
Chain 10
Network 11
Symmetric Relationships 13
Roles 14
Fan Trap 15
.............................................................................................................................................
Introduction
.........................................................................................................................................
Topic See Page

Data Warehouse 16
Drawing Conventions 17
Generic Modeling 19
Generic Models 20
Summary 23
Practice 6—1: Patterns 24
Practice 6—2: Data Warehouse 25
Practice 6—3: Argos and Erats 26
Practice 6—4: Synonym 27

• Recognize common patterns in conceptual data models
• Know the general behavior, such as the common constraints, of these patterns
• Use particular drawing conventions
• Create a more generic model for selected sections of a conceptual data model
.........................................................................................................................................
6-3
®
.........................................................................................................................................
Patterns
Similar Structure
Many models contain parts that have a similar structure, although the context may be
completely different. For example, the structure of a conceptual data model in the
context of a dictionary that deals with concepts such as headword, entry, meaning,
synonym is, surprisingly, almost identical to the structure of a railroad with track,
station, connection, and also to the structure of a baseball or soccer competition.
Easier to see are the similarities between, for example, ORDER HEADER with
ORDER ITEM and QUOTATION HEADER with QUOTATION ITEM, or between
MARRIAGE and JOB ASSIGNMENT.
• Similar structure
• Similar rules and constraints?
Why Search for Similarities?

The main reason why it is important to look for similarities is that it will save you
time. If you have solved a problem in a particular context and you can apply the
solution to another, it obviously saves time. Moreover, you will feel confident that you
know about the situation. It will help you to ask the right questions. It will help you
identify the really complex and unpleasant things and will prevent you from making
the same mistakes twice.
Are there similarities between marriage and job assignment? Of course, the business
rules in the context of a marriage are different, because they are determined differently
compared to those of a job assignment. But when you are aware of the similarities, you
can easily check if business rules of the first context apply in the second, by asking, for
example:
• Can an assignment be for more than one job?
• Can someone have two assignments simultaneously? Unofficially?
• How does an assignment start? How does it end?
The following paragraphs discuss a series of patterns that you will encounter while
creating your models. For all these patterns you will see the characteristics and the
rules that usually apply.
.............................................................................................................................................
Master Detail
.........................................................................................................................................
Master Detail
About the slide
Patterns: Master–Detail See Page 28
consists
of
part of
B
• Characteristic: consists of
An instance of B only exists in the context of an A
• Metaphor: Master–Detail
6-3
Master-detail constructions are very common, as 1:m relationships are very common.
Distinguish between a 1:m relationship that is typically directed from the 1 to the
many and a relationship that is directed the other way around (see below). Master-
detail is characterized by the fact that the master A is divided into B’s. B’s do not exist
alone; they are always in the context of an A.
It is very rare that these relationships are transferable; if an instance of B is connected
to the wrong instance of A, it is far more likely that the instance of B is deleted and
then recreated in the context of the correct A.
Typical master-detail relationship names:
• Consists of
• Divided into
• Made of
• (Exists) With
Often a master A is of no value when it has no B’s, for example, the relationship is
mandatory at the 1 side. This mandatory relationship end can usually be circumvented,
as you have seen before.
Implementation
The tables that come from this master-detail pattern should be considered as clustered.
.........................................................................................................................................
6-5
®
.........................................................................................................................................
Basket
About the slide
Pattern: Basket See Page 28
A X
A Y
Z
X B
consists
of Y
part of
B Z
• Characteristic:
container for various types of items
• Items may be of different types
• Metaphor: Shopping Basket
6-4
A basket construction is a special case of a master-detail pattern. A basket can contain

one or more things, but these things (often named: items) can be of different types. A
single item is always of one type only. That is the reason for the arc. The arc shows
that an item must be of one and only one of the types.
.............................................................................................................................................
Classification
.........................................................................................................................................
Classification
Patterns: Classification
classifying
classified by
P
• Characteristic: classified by, grouped by

Q exists independently, may be related
• Metaphor: EMPLOYEE–DEPARTMENT
6-5
This is again a 1:m relationship, but now the main orientation is from P to Q.
This is typically the case when Q can exist independently from P. Q acts as a class for
P, something with which to group P’s.
Usually entities in a conceptual data model have several of these classes.
Typical classification-type relationship names:
• Classified by
• Grouped by
• Assigned to
• (Exists) In
The relationship is usually transferable as classifications may change over time.
.........................................................................................................................................
6-7
®
.........................................................................................................................................
Hierarchy
About the slide
Patterns: Hierarchy See Page 29
A
# Id
• Characteristic: manager of / subordinate of

• Additional constraints to guard hierarchical nature
• Metaphor: Mother–Child
6-6
Most hierarchical structures have a known limit for the maximum number of levels. If
that is the case and the limit is a low number of 5, for example, then usually the best
model is the one that is shown in the left of the illustration, one entity per level.
Model the structure with the recursive relationship if:
• The structure has no known level limit.
• The structure has a level limit, but the limit is high, say six or more.
• An instance of the structure can easily have a change of position, thus changing its
level.
• You like maintaining constraints.
Disputable or False Hierarchies

Often structures should be hierarchical but you cannot be sure. Sometimes they seem
hierarchical but actually are not so. You can have, for example, the is owner of
relationship between companies. Suppose company C1 owns company C2, company
C2 owns company C3, could it be that company C3 owns the shares of company C1?
Even if legislation would prohibit such strange constructions, would you be sure?
Many people see the parent/child relationship as a metaphor for a hierarchical
relationship. Clearly this is wrong as a child usually has two parents and can have step-
parents as well.
.............................................................................................................................................
Hierarchy
.........................................................................................................................................
Also the hierarchical structure of a FILE SYSTEM with files and folders, which are
files of a particular type, is a disputable hierarchy when you think of the concept of a
shortcut in Windows (or a Link in UNIX). These shortcuts transform the hierarchy
conceptually into a network although technically a shortcut and a link are just files
with a special role.
Recursive Relationship and Optionality

Recursive relationships that describe a real hierarchy are usually optional at both ends,
as the hierarchy must start or end somewhere.
Constraints Applying to a Hierarchy

The recursive model, as you see in the centre of the illustration, only requires an
instance of A to refer to a valid instance of A. A1 referring to A1 is fine, according to
the model. A2 referring to A3 and A3 referring to A2 is fine as well. These are the only
obvious diversions from a real hierarchy.
Constraints that apply in a hierarchical structure deal with safeguarding the hierarchy
and should prevent the table from containing the above kind of data.
Implementation
The first constraint, A1 may not refer to A1, and you can easily check this with an
Oracle check constraint. The others need some programming and lead to database
triggers.
Possibly you may have to check extra business rules, for example, when the number of
levels may not exceed a given value.
.........................................................................................................................................
6-9
®
.........................................................................................................................................
Chain
About the slide
Patterns: Chain See Page 29
preceded
by B
BEAD
# Id
followed CHAIN
by
A
BEAD
# Seqno
• Characteristic: preceded by / followed by

• Sequence is important
• Metaphor: Elephants
6-7
A Chain (of beads) can be regarded as a special kind of hierarchy. A chain is a

recursive relationship of an entity. The relationship of the chain is a 1:1 relationship as
a chain is characterized by the fact that an object in the chain is preceded and followed
by one object at most.
A chain is a structure where sequence is of importance, for example, the sequence of
the pages in a chapter and of the chapters in a document, of the critical path in a
procedure, of the preferred road from A to B.
A chain can also be modeled as a master-detail. The recursive model allows an easy
insertion in the chain. The right-hand model with entity CHAIN and BEAD may need
to change the sequence numbers of all the beads behind the inserted one.
.............................................................................................................................................
Network
.........................................................................................................................................
Network
About the slide
Patterns: Network See Page 29
A
A A
• Characteristic: pairs
Every A can be connected to every A
(sometimes: to every other A)
• Metaphor: Web Document with Hyperlinks
6-8
Network structures typically describe pairs of things of the same type, for example,
marriage, railroad track (pair of start and end stations), synonyms (two words with the
same meaning), and Web documents with hyperlinks to other Web documents.
Characteristics
Often:
• The m:m relationship must be resolved to hold specific information about the pair
such as the date of the marriage, or the length of the railroad track.
• The two relationships of the intersection entity form the unique identifier.
• Time-related constraints apply in networks that must guard, for example, the kind
of rules that deal with “sequentially monogamous”.
• The two relationships refer to different subtypes of the entity:
Note that a hierarchy is a network where a particular set of business rules apply.
.........................................................................................................................................
6-11
®
.........................................................................................................................................
Bill of Material
A special example of a network structure is a Bill of Material (BOM). A BOM
describes the way things are composed of other things, and how many of these other
things (here it is instances of PRODUCT) are needed. Entity COMPOSITION is the
intersection entity with attribute Quantity Needed.
About the slide

Bill of Material See Page 29
product of
PRODUCT COMPOSITION
# Code with * Quantity Needed
part in
in
PRODUCTS COMPOSITIONS
Code Name Prod_code Part_code Quantity
914.53 AAAAAAAAA 854.01 604.18 1
914.54 AA 854.01 604.19 1
914.55 BBBBBBBBB 854.01 914.54 2
914.56 CCCCCCC 914.54 914.55 1
DDDDD 914.54 914.56 1
934.76 915.12 3
6-10
About the slide

See Page 30
854.01
914.54 914.54
604.18 914.55
914.56
604.19
6-11
.............................................................................................................................................
Symmetric Relationships
.........................................................................................................................................
Symmetric Relationships
Symmetric recursive relationships cause a very special kind of problem which is more
complex than you would assume.
In most contexts a record of a pair (A1, A2) has a different meaning when referred to as
(A2, A1). For example, if the model is about entity PERSON and the relationship is
mother of /daughter of, then the existence of person pair (P1, P2) would mean the
exclusion of the possibility of pair (P2, P1).
The recursive relationship of PERSON and family of / family of. Here, if (P1, P2) is
true, then (P2, P1) is equally true. This is called a symmetric relationship. There are
other symmetric recursive relationships such as: STATION directly connected by rail
with STATION,
Symmetric Relationships: Problem

When in a symmetric relationship the pair (S1, S2) is valid, the pair (S2, S1) must be
valid as well. Nevertheless, it would not make much sense to record both pairs as that
would essentially store the same information twice—which would oppose one of the
basic principles of database design.
But if we record only one pair, which should we record? And how would you know
which of the two pairs was used if someone else had recorded it?
Symmetric Relationships: Solution

A way which is often used to model these symmetric situations is based on the
following idea: think of (S1, S2) as Group1, (S3, S4) as Group2 and so on. Looking at
the relationship this way, you can say that a GROUP always consists of exactly two
instances of S. The model and the table implementation are shown below.
About the slide

GROUP Group_id S
See Page 30
# Id 1 S1
consists of 2 1 S2
in 2 S3
2 S4
S
3 S5
3 S6
.........................................................................................................................................
6-13
®
.........................................................................................................................................
Roles
About the slide
Patterns: Roles See Page 30
A P
• Characteristic: is / is 1:m (or 1:1) relationships

• Metaphor: Person–Many Hats
(not necessarily concurrent...)
6-13
Roles often occur when a system needs to know more about people than the basic
Name/Address/City information. Modeling the roles as separate entities offers the
possibility to show which attributes are mandatory for a particular role, and, if
necessary, to show relationships between the various roles. The example below shows
that a person in their role as president of a country can appoint a person in the role of
minister of a department. Possibly the words “presidency” and “ministership” are
closer to the concepts than the ones in the diagram.
About the slide

PERSON ROLE
TYPE
See Page 30
ROLE
roles
PERSON PRESIDENT COUNTRY
appointing
appointed by
MINISTER DEPARTMENT
PARTY PARTY
LEADER
.............................................................................................................................................
Fan Trap
.........................................................................................................................................
Fan Trap
About the slide
Fan Trap See Page 30
A B
A B
C
AB
AC BC
• Characteristic: ring of m:m related entities

• Metaphor: ABC Combination
6-15
A Fan Trap (named after the characteristic shape of the solution) occurs when three or
more entities are related through m:m relationships and form a ring. Usually you
should replace the relationships with a central entity having several m:1 relationships.
Preventing a fan trap is similar to resolving a m:m relationship between two entities.
Why Traps Occur

Resolving the three m:m relationships results into three intersection entities, AB, BC
and AC. These will contain related pairs. Joining AB and BC may, however, result in
different information to what AC contains which you may have seen in practice 3-8.
Note there are various ways of avoiding the trap, as is shown in the illustration. All can
be correct, depending on the context.
About the slide

A B C A B C A B C See Page 31
AB BC
ABC ABC ABC
.........................................................................................................................................
6-15
®
.........................................................................................................................................
Data Warehouse
About the slide
Patterns: Data Warehouse See Page 31
July Practice
B C
See Page 31
A X D
This is a suitable
place to do
practices 6-1.
F E
• Characteristic: multidimensional, many, many

detail instances
• Metaphor: star model
Stars may be strangely shaped:
• Snowflake model
6-17
A data warehouse system can be modeled as any system. Data warehouses contain the
same sort of information as any straightforward transaction processing information
system. Data warehouses usually contain less detailed, summarized, information as
warehouses are mainly built for overview and statistical analysis. However, Data
warehouses in general receive the input from online transaction systems that do
contain details.
Data warehouses often have a star-shaped model: this is made up of one central entity
(the facts) containing the condensed, summarized, information, and several
dimensions that classify and group the details.
Common dimensions represent entities such as:
• Time
• Geography
• Actor (for example, salesperson, patient, customer, instructor)
• Product (for example, article, medical treatment, course)
Often the dimensions are classified as well. Time may be structured in day, week,
month, quarter, year. You can classify products in various ways as you have seen in
earlier examples. If this is the case, the model is usually described as the Snowflake
model, as it looks like the crystal shape of a snowflake.
.............................................................................................................................................
Drawing Conventions
.........................................................................................................................................
Drawing Conventions
About the slide
Drawing Conventions See Page 31
high volumes
high volumes
Not important which convention you choose,

as long as you follow one of them
6-18
Two drawing conventions are widely in use: one that positions the entities with the
high volumes at the top of the paper and one that does the opposite. Both try to avoid
crossing relationship lines, partially overlapping entities, and relationship lines that
cross entities. Whatever convention you choose, choose one and use it consistently.
This will prevent errors and make the reading of large diagrams much easier.
Keep the overall structure of the layout unchanged during the modeling project as
many people are disoriented when you change the structure.
Make separate diagrams for every business area. These may have a different layout;
these diagrams are mainly used for communication with subject matter experts.
At the end of this course, you should be able to read models created in any drawing
convention, and you should be able to complete a model following any convention
used.
.........................................................................................................................................
6-17
®
.........................................................................................................................................
Use Conventions Sensibly
But:
Readability first
The major goal of creating the diagram (but not the model) is to give a representation
of the model that can be used for communication purposes. This means that you must
never let a convention interfere with readability and clarity. Do not be concerned that
readability takes space. Usually an entity model is represented by several diagrams
that show only the entities and relationships that deal with a particular functional part
of the future system. Splitting the model over various diagrams adds to the readability.
About the slide

Model Readability See Page 32
B A B
A
C
E C
D
F E
D
• Takes space
• Subject to taste F
6-20
.............................................................................................................................................
Generic Modeling
.........................................................................................................................................
Generic Modeling
Generic Modeling
MANUFACTURER MANUFACTURER ARTICLE
* Name * Name TYPE
FILM ARTICLE
* AsaTRIPOD o Weight
* Height
LENS o Focal Distance
CAMERA o Height
* Focal
BODY o Asa Number
Distance
* Weight o ...
6-21
What is Generic Modeling?

Generic modeling is looking at the same context from another, more distant
perspective. From a distance many things looks the same.
Suppose you are to make a model for a photographer’s shop. The business typically
sells many different articles, for example, camera bodies, compact cameras, lenses,
films. For each type of article, there are between, say, 10 and 500 different types. You
can model every type as an entity, for example, CAMERA BODY, LENS, FILM.
You could also model them all as subtypes of the entity ARTICLE, or all as just
ARTICLE, without the subtypes.
This, however, would not work. For example, there is the fact that every now and then
new kinds of articles are stocked in the shop. Every time this happens it leads to a new
entity with its own attributes in the model.
The model with entity ARTICLE would only be a workable model if there were no (or
possibly only very few) new instances of ARTICLE TYPE during the life cycle of the
system.
.........................................................................................................................................
6-19
®
.........................................................................................................................................
Generic Models
More generic models are shown below. They may be useful in particular situations.
About the slide

ARTICLE TYPE
ARTICLE See Page 32
* Definition Prop1 o Property1
o Definition Prop2
o Property2
o Definition Prop3
o Property3
o Definition Prop4
o Property4
...
o Property5
o Property6
o Property7
MANUFACTURER
o Property8
* Name
Recycling of Attributes You can use this model if it is safe to assume the articles
will have a limited number of attributes. This limit may be a high number but must be
set beforehand. Property1 may contain the Asa Number for instances of ARTICLE of
TYPE Film and may contain Weight for instances of ARTICLE of TYPE Camera
Body and so on. The major advantage of this model is the possibility of adding new
instances of ARTICLE TYPE without the need to change the model.
The type of information that should be entered for Property1, Property2, and so on can
be described by using, for example, the Definition Prop1, attributes of ARTICLE
TYPE. Here you can also store information about the data type of these properties.
About the slide

ARTICLE TYPE See Page 32
ARTICLE PROPERTY
ARTICLE PROPERTY VALUE

o Value
Attributes Modeled as PROPERTY Instance This model takes another approach.

Every value for a PROPERTY of an ARTICLE is stored separately. This model gives a
lot of freedom to define new articles and properties during the life cycle of the system.
.............................................................................................................................................
More Generic Models
.........................................................................................................................................
More Generic Models

Everything is a “Thing”
The world is full of things that may be related to things:
having some kind of

relationship with
THING
having some kind of
relationship with
Resolving the m:m relationship:
THING
ASSOCIATION
Now add some definition information:
About the slide

See Page 33
THING
TYPE ASSOCIATION
TYPE
THING
ASSOCIATION
6-26
This is a rather generic model. In fact, it is a model of the universe and beyond. Note
that the number of attributes for entity THING may be substantial.
.........................................................................................................................................
6-21
®
.........................................................................................................................................
Most Generic Model
THING TYPE ASSOCIATION

TYPE
THING
ASSOCIATION
PROPERTY
THING PROPERTY VALUE
6-27
This model combines the concepts of “thing” and the property/property value and thus
allows everything to be represented with a free number of properties per type.
Value of Generic Modeling

The use of generic modeling is mainly to reduce to a minimum the number of possible
future changes of the conceptual data model. This can be an enormous advantage as it
cuts maintenance costs during the lifetime of a system. The other side of the coin is
that the initial coding of the programs is more complex as the entities are not “down-
to-earth” things.
Best of Two Worlds

In many models you would use a mix of the easy-to-understand, straightforward
entities and the more generic thing-like entities.
About the slide

CUSTOMER ‘generic’ See Page 33
ARTICLE TYPE
ORDER HEADER
ARTICLE PROPERTY
ORDER ITEM
‘down to earth’ ARTICLE PROPERTY VALUE
.............................................................................................................................................
Summary
.........................................................................................................................................
Summary
Summary
• Patterns
– Show similarities
– Invent your wheel only once
• Generic models
– Reduce the number of entities dramatically
– Are more complex to implement
– Are very flexible
– Are usually the best choice in unstable
situations
6-29
Thinking in terms of patterns forms a valuable way of doing quality checks on a

conceptual data model. Often constraints and considerations in one context can be
transferred to the other context with a simple translation.
Using a drawing convention in your models helps to improve readability and clarity.
This may prevent mistakes and inaccuracies.
Generic modeling can prevent the need to change data structures in the future and can
reduce the number of tables and programs dramatically. The price is increased
complexity in both data model and programs.
.........................................................................................................................................
6-23
®
.........................................................................................................................................
Practice 6—1: Patterns

Goal (See Page 31)
The purpose of this practice is to predict the main pattern in a given context.
Your Assignment
What pattern do you expect to find in the given contexts? If you do not see it, make a
quick sketch of the model. Use your imagination and common sense.
Practice: Patterns
• Model of moves in a chess game

• Model of tenders (quotations)
• Model of recipes
• Model of all people involved in college: students,
teachers, parents, …
• Rentals in a video shop
• Model of phases in a process
6-31
.............................................................................................................................................
Practice 6—2: Data Warehouse
.........................................................................................................................................
Practice 6—2: Data Warehouse

Moonlight Coffees
Goal (See Page 34)
In this practice you create a conceptual data model for a data
warehouse for Moonlight Coffees Inc.
Scenario
Moonlight wants to build a data warehouse based on the detailed sales figures the
shops report back on a daily basis. Examples of questions Moonlight wants the data
warehouse to answer are printed below.
•What is the sales volume in $ of coffee last month compared with the coffee sales
volume same month last year?
•What is the sales volume in $ of coffee per head in Japan compared with the
average coffee sales volume in the Moonlight countries around the world?
•What is the growth of the sales volume in $ of coffee in Sweden compared with the
growth of sales volume of all products in the same geographical area? What is the
growth in local currency?
•What was the total sales volume in $ of coffee last month, compared with the total
coffee sales volume in the same month last year, for the shops that have been open
for at least 18 months?
•What is the growth of the sales volume in $ of nonfoods compared to that of foods?
•What is the best day of the week for total sales in the various countries? How is that
related to the average? Is the best day of the week dependent on the type of
location?
•What products are most profitable per country? Globally?
•Does the service level (#employees per 1000 items sold) have influence on sales?
6-32
Your Assignment
1 Check the Moonlight models you created so far. Do they cater for answering the
listed questions. If not, make the appropriate changes.
2 For a data warehouse data model, suggest the central “facts” entity.
.........................................................................................................................................
6-25
®
.........................................................................................................................................
Practice 6—3: Argos and Erats

Goal (See Page 35)
When you model information, you make a lot of assumptions, often without being
aware of this. Most of these assumptions are likely to be correct as they are usually
based on experience in similar contexts or common.
This practice helps to increase your awareness of this.
Scenario
The scenario for this practice is Stranger in a Strange Land. Lost in Darkness. The
Wanderer in the Mist. You name it!
Your Assignment
Make a conceptual data model based on the information in the text. Mark all the pieces
in the diagram that can be confirmed from the text.
"Erats have names that are unique. Erats can have argos.
Argos have names as well. The name of an argo must be
unique within the erat it belongs to. Erats mutually have
rondels. There are only a few different types of rondels. Erats
can have one or more ubins. A ubin always consists of one or
more argos of the erat, one or more rondels of the erat, or
combinations of the two."
.............................................................................................................................................
Practice 6—4: Synonym
.........................................................................................................................................
Practice 6—4: Synonym

A synonym is, according to a dictionary, “a word having the same meaning with
another (usually almost the same).”
Examples:
practice - exercise
order - command
entity - being
order - sequence
order - arrangement
command - demand
Your Assignment
Make a conceptual data model that could be the basis for a dictionary of synonyms.
.........................................................................................................................................
6-27
®
.........................................................................................................................................
Instructor Notes
Instructor Note
Topic Timings
66
Lecture 30 minutes
A dva nced M od eling T opics
Practice 30-60minutes
Total 60-90 minutes
Some of the patterns in this chapter have only

O verview
been touched in the practices.
• P atte rn s
• D ra w in g c o n ve n tio n s
• G en e ric m o d elin g
6 -2
Very common pattern.

P atte rn s: M aster–D etail
In object modeling there is a special name for
A
this relationship: an aggregation.
consists
part of
of Do not use parent/child as metaphor as this is a
B
false one.
• C h arac te ristic: c o n sists of
A n in stan c e o f B o n ly e x ists in th e co n tex t o f an A
• M eta p h o r: M a s te r – D etail
6 -3
Special case of Master-Detail.

Pattern: B as ket
A A X
Y
Z
Note not only different products in the basket,
consists
X B
but also different types of products. Here:
part of
of Y
groceries and a clock.
B Z
• C h arac te ristic:
ris tic:
co
onn tain
ta in e
err fo
forr v
vaariou
rio u s typ es
e s o f item
ite m s
• Ite
Itemm s m ay b e o f d iffere n t ty p
iffe ren pees
s
• M eta p
phh o r: S h o p p in g B a
assk
keett
6 -4
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Patterns: C lass ific ation
classifying
classified by
P
• Ch a ra cteristic : cla s sifie d b y, g ro u p ed b y

Q e xis ts in d ep e nd e n tly, m a y be re la te d
• M e ta p h o r: E M P LO Y E E – D E P A R T M E N T
6 -5
Many people think a hierarchy is equivalent to a

Patterns: H ierarchy
recursive 1:m relationship. This is not true. A
hierarchy has many internal rules (such as you
A
# Id cannot be your boss’ boss) which are not
represented by the recursive relationship.
• C h arac te ris tic: m a n ag e r o f / s u b o rd in a te o f Do not use parent/child as a metaphor as this is
• A d d itio n al co n s traints to g u ard hiera rc h ica l na tu re
• M eta ph o r: M o th er – C h ild
a false one.
6 -6
A Chain is a special case of a hierarchy and a

P atte rn s: C hain
preceded
special case of a network as well.
by B
B EA D
# Id
follow e d
C H A IN
by
A
B E AD
# Se qno
• C h arac te ris tic: p re ce d ed b y / fo llo w e d b y

• S e q u en c e is im p o rtan t
• M eta ph o r: E le p h an ts
6 -7
Very common structure.

P atte rns: N etw o rk
A
A A
• Ch a ra cteristic : p airs
E ve ry A c an b e c on ne cte d to ev ery A
(s o m etim es : to eve ry o th er A )
• M e ta p h o r: W eb D o c u m en t w ith H y p erlin k s
6 -8
Very powerful structure.

B ill of M aterial
product of
PRODUCT C O M P O S IT IO N
w ith * Q uan tity N ee ded
# C ode part in
in
PR
P R O DU
D UC T S C O M P O S ITIO N S
C ode N am e P rod_code P art_code Q uantity
914.53 AAAAAAAAA 854.01 604.18 1
914.54 AA 854.01 604.19 1
914.55 BBBBBBBBB 854.01 914.54 2
914.56 CCCCCCC 914.54 914.55 1
DDDDD 914.54 914.56 1
934.76 915.12 3
6 -1 0
.........................................................................................................................................
6-29
®
.........................................................................................................................................
The data in the table in the previous illustration

854.0
854.011
match the information given in this picture.
914.54 914.54
91 4.54
60 4.18
604.18 914.55
914.56
604.19
6 -1 1
This is a complex notion that you may run into

Sy m m e tric R elationsh ip
every now and then.
GROUP
Other examples:
G ro up_ id S
# Id 1 S1
cconsists
onsists of 2 1
2
S2
S3
• Alternative for
in
2 S4
S
3
3
S5
S6
• Synonym for (see practice)
6 -1 2
Most often the role model is used when

Patterns: R o les
subtyping does not work due to overlap in the
A P roles. In this example a president may have
Q been minister or party leader.
• Ch a ra cteristic : is / is 1 :m (o r 1 :1 ) relatio n s h ip s
• M e tap h o r: P ers o n – M an y H ats
(n o t n e ce ss arily co n c u rren t...)
6 -1 3
The PERSON-ROLE-ROLE TYPE model uses

R oles
P ER S O N R O LE
ROLE as a supertype for all possible roles. This
TY PE
R O LE
model is very useful if the variety of roles may
roles change during the life cycle of the system.
PERSON P R E S ID E N T
a ppointing
C O U N TR Y
Refer to the section on generic modeling for
ap pointed by
M IN IS TE R D E P A R TM E N T more on this.
P A R TY P A R TY
LE A D E R
6 -1 4
Fan trap is the subject of practice 3-8 Holiday

F an T rap
and possibly in individual solutions to other
B
A
A B
practices. This trap is always near.
C
AB
AC BC
• Ch a ra cteristic : rin g o f m :m related e n tities

• M e ta p h o r: A B C C om b in a tio n
6 -1 5
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
An example where the left and right model

Fan Trap R eso lve d
make sense:
A B C A B C A B C A: Translation Agency
AB BC
B: Language
ABC ABC ABC
C: Translator
A B fu nc tio n s a s B C fu n ctio n s as AB combines the available languages at an
lis t o f v alue s lis t o f va lu es
6 -1 6
agency. BC combines the available languages
per translator.
A data warehouse usually has the fact/
P atte rns: D ata W a rehou se
dimensions structure that is represented by the
July
B C
star or snowflake model.
A X D
F E
• C h arac te ristic: m u ltid im e n sio n a l, m an y , m a n y

d e ta il in sta n ce s
• M eta p h o r: star m o d e l
S ta rs m a y b e s tran g ely s h ap e d :
• S n ow flak e m o d el
6 -1 7
Good practice to do jointly as a whole group.

Prac tice: Patterns
• M o d e l o f m o ve s in a c he s s g a m e
• M o d e l o f ten d e rs (q u o tatio n s )
• M o d e l o f re cip e s
• M o d e l o f all p e o ple in vo lved in c o lle g e: s tu d e n ts ,
tea ch ers , pa re n ts , …
• R e nta ls in a v id eo s h o p
• M o d e l o f p h as es in a p ro c es s
6 -3 1
The main message should be:

D raw in g C onv entions
It does not matter what convention you use as
high volum es
long as you use one.
Note that the conventions apply mainly to the
high v olum es
diagrams you make for your own and your
team’s understanding.
N o t im p o rta nt w hich co n v en tio n y o u ch o o s e,
as lon g as y ou fo llo w o n e o f th e m
Diagrams for the individual business areas may
6 -1 8
use a different layout as these diagrams are

supposed to communicate with the subject
matter experts.
.........................................................................................................................................
6-31
®
.........................................................................................................................................
U se C o nven tions S ensibly
Bu t:
Re a da b ility firs t
6 -1 9
The right-hand model requires less careful

M ode l R e ada bility
reading and is therefore more accessible.
B A B
A
C
E C
D
F E
D
• T ak es s pa c e
• S u b je ct to tas te F
6 -2 0
G ene ric M odeling

M A N U FA C TU R E R M A N U F A C TU R E R A R TIC LE
* N am e * N am e TY P E
FILM A R TIC LE
* A sa TR IP O D o W eight
* H eight
LE N S o Focal D istance
CAM ERA o H eight
* Focal
BOD Y o A sa N um ber
D istance
* W eight o ...
6 -2 1
This model works when the number of

G ene ric M odeling
properties per article type is limited.
A R TIC LE TY P E
* D efinition P rop 1
o D efinition P ro p2
A R TIC LE
o P roperty1
In Oracle Designer, the User Defined Objects
o P roperty2
o D efinition P rop 3
o D efinition P ro p4
...
o P roperty3
o P roperty4
make use of this concept.
o P roperty5
o P roperty6
o P roperty7
M A N U FA C TU R E R
o P roperty8
* N am e
6 -2 2
Expensive, but very flexible. A single column

G en eric M odel
value in a normal table is now stored as row in
A R TIC LE T Y P E
the table based on ARTICLE PROPERTY
A R TIC LE P R O P E R TY
VALUE.
Searching for articles using some property
A R TIC LE P R O P E R TY V A LU E
o V alue
value is much more complex using this
structure.
6 -2 3
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
G en eric
h aving som e kind of

having
relationship w ith
TH IN G
having som e kin
kind d of
relationship w ith
6 -2 4
M ore G ene ric
TH IN G
A S S O C IA TIO N
6 -2 5
This used to be the internal model for Oracle

M ore G eneric Plu s
Designer. Views based on this model showed
the actual object types.
TH IN G
TY P E A S S O C IA TIO N
TY P E
TH IN G
A S S O C IA TIO N
6 -2 6
M ost G ene ric?
TH IN G TY P E A S S O C IA TIO N
TY P E
T H IN G
A S S O C IA TIO N
P R O P E R TY
TH IN G P R O P E R TY V A LU E
6 -2 7
Many real-life systems combine generic and

B es t of T w o W o rlds
down-to-earth structures.
Generic modeling is not better than down-to-
C U S TO M E R ‘generic’
A R TIC LE TY P E earth modeling, nor the other way around. Both
ORDER HEADER
A RT
R T IC LE P R O P E R TY
have advantages.
O R D E R ITE M
‘dow n to ea rth’ A R TIC LE P R O P E R TY V A LU

L UE
6 -2 8
.........................................................................................................................................
6-33
®
.........................................................................................................................................
Sum m ary
• P atte rn s
– S h o w s im ilarities
– In ve n t yo u r w h ee l o n ly o n c e
• G en e ric m o d els
– R e d u ce th e nu m b er o f e n tities d ra m atic ally
– A re m ore c o m p le x to im p le m en t
– A re v ery fle xible
– A re u s u ally th e b es t ch o ic e in u n s ta b le
s itu atio n s
6 -2 9

Prac tices
• P atte rn s Practice 3day 4day

• D a ta W are h o u se
• A rg o s an d E ra ts Patterns Yes Yes
• S yn o n y m
Data Warehouse Opt Yes
Argos and Erats Cha Cha
Synonyms Opt Cha
6 -3 0
“... coffee per head in Japan...” leads to an

•W hat is the sa les volum e in $ of c offee last m onth co m pared w ith the coffee sa les
attribute population per country. The other
volum e sam e m onth la st year?
•W hat is the sa les volum e in $ of c offee per head in Ja pan com pared w ith the
ave rage coffee sales v olum e in the M oon light countries around th e w orld?
pieces of information were probably already
•W hat is the grow th of the sales vo lum e in $ of coffee in S w eden com pared w ith the
gro w th of sales volum e of all products in the sa m e geographical are a? W hat is the
gro w th in local currenc y?
found in an earlier stage.
•W hat w as the total sales volum e in $ of coffee last m onth, com pared w ith the to tal
coffee sales volum e in the sam e m onth last yea r, for the shops that have been open
for at least 18 m onths?
•W hat is the grow th of the sales vo lum e in $ of nonfoo ds com pared to tha t of foods?
•W hat is the be st day of the w eek for total sales in the various countries? H ow is that
related to the average ? Is the best day of the w eek dependen t on the type of
loc ation?
•W hat products are m ost profitable per country? G lob ally?
•D oes the serv ice level (#em ploye es per 1000 item s s old) have influence o n sales?
6 -3 2
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
This is a very difficult practice and clearly

P ractice: A rg os and E rats
shows how much of your own interpretation
"E rats have nam es th at are u nique . E rats can have arg os.
Argos h ave nam es a s w ell. Th e n am e of an argo m ust be
you would usually use when you make a model.
uniq ue w ithin the erat it be longs to. E rats m utua lly have
ron dels. There a re o nly a few differen t typ es of ron dels. Erats
can h ave one or m ore ubin s. A ubin alw ays consists o f on e o r
Here interpretation is impossible as the central
m ore argos of th e erat, o ne o r m ore ron dels o f th e era t, o r
com b inatio ns of the tw o." concepts are nonsense words. When people
were asked to model the original text it seemed
much easier; many things that you cannot really
6 -3 3
tell from the text were assumed, and nobody
noticed this!
(For your eyes only: the text is a summary of
entity modeling, read entity for erat, attribute
for argo, relationship for rondel and uid for ubin
and immediately it falls into place!)
Synonym itself is not an entity, although most
Prac tice: Syn ony m
analysts will start with it.
p rac tice
o rd er
- e xe rc ise
- com m and
A synonym leads to a symmetric relationship: if
en tity
o rd er
- b e in g
- s eq u e n ce
two or more words share the same meaning
o rd er - a rra n g em en t
co m m a n d - de m a n d
then apparently a synonym exists.
6 -3 4
.........................................................................................................................................
6-35
®
.........................................................................................................................................
.............................................................................................................................................
.........................................................................................................................................

Name: Full Email:

Lesson number: 6

Detail comments
Content Page
Type Number
.........................................................................................................................................
6-37
®
.........................................................................................................................................
.............................................................................................................................................
Suggested Graphics
.........................................................................................................................................
Suggested Graphics

.........................................................................................................................................
6-39
®
.........................................................................................................................................

520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.............................................................................................................................................
7
.................................
Mapping the ER Model

.....................................................................................................................................................
Introduction
Lesson Aim
This lesson describes some principles of relational databases and presents the various
techniques that can be used to transform your Entity Relationship model into a
physical database design.
Schedule
About the slide

See Page 34
• Why use design modeling?
• Introduction to the components:
– Tables
– Columns
– Constraints
• Basic Mapping
• Complex mapping
7-2
Topic See Page

Introduction 2
Why Create a Database Design? 4
Transformation Process 6
Naming Convention 8
Basic Mapping 12
Relationship Mapping 14
Mapping of Subtypes 20
Summary 29
Practice 7—1: Mapping Supertype 30
Practice 7—2: Quality Check Subtype Implementation 31
.....................................................................................................................................................
Introduction
.....................................................................................................................................................
Topic See Page

Practice 7—3: Quality Check Arc Implementation 32
Practice 7—4: Mapping Primary Keys and Columns 33
Objectives
• Understand the need of a physical database design
• Know the concepts of the relational model
• Agree on the necessity of naming rules
• Perform a basic mapping
• Decide how to transform complex concepts
.....................................................................................................................................................
®
7-3
.....................................................................................................................................................
Why Create a Database Design?

The Entity Relationship model describes the data required for the business. This model
should be totally independent from any implementation considerations. This same ER
model could also be used as a basis for implementation of any type of DBMS or even
a file system.
About the slide
Why Create a Data Design Model? See Page 34
• Closer to the implementation solution

• Facilitates discussion
• Ideal model can be adapted to an RDBMS model
• Sound basis for physical database design
7-3
A New Starting Point An Entity Relationship model is a high-level representation

which cannot be implemented as is.
People creating these models may not be aware of physical and database constraints,
but they still have to provide a conceptually “workable” solution. This is why it is
important to have a validated and agreed ER model before going into the physical
database design.
Transforming the ER model, creates a “first-cut” database design. This first-cut design
is intended to serve as a new basis for defining the physical implementation of the
database.
This new model can easily be used for further discussions between designers,
developers, and database administrators.
.....................................................................................................................................................
Why Create a Database Design?
.....................................................................................................................................................
Presenting Tables
Tables are supported by integrity rules that protect the data and the structures of the
database. Integrity rules require each table to have a primary key and each foreign key
to be consistent with its corresponding primary key.
About the slide
Presenting Tables See Page 34
Table: EMPLOYEES
columns
Id Name Address Birth_date Dpt_id

126 PAGE 12, OXFORD ST 03-03-66 10
rows 349 PAPINI 53, HAYES AVE 10-08-77 20
785 GARRET 08-12-55 10
primary key unique key foreign key

column column column
EMPLOYEES (EPE)
pk * Id
Table diagram: EMPLOYEES uk1 * Name
o Address foreign
uk1 * Birth_date key
fk * Dpt_id
7-4
Tables A table is a very simple structure in which data is organized and stored.
Tables have columns and rows. Each column is used to store a specific type of value.
In the above example, the EMPLOYEES table is the structure used to store
employees’ information.
Rows Each row describes an occurrence of an employee. In the example, each row
describes in full all properties required by the system.
Columns Each column holds information of a specific type like Id, Name, Address,
Birth Date, and the Id of the department the employee is assigned to.
Primary keys The Id column is a primary key, that is, every employee has a unique
identification number in this table which distinguishes each individual row.
Unique keys Both columns Name and Birth_date are associated with a Unique key
constraint which means that the system does not allow two rows with the same name
and Birth_date. This restriction defines the limits of the system.
Foreign keys The foreign key column enables the use of the Dpt_id value to retrieve
the department properties for which a specific employee is working.
.....................................................................................................................................................
®
7-5
.....................................................................................................................................................
Transformation Process
Using transformation rules you create a new model based on the conceptual model.
About the slide
Transformation Process See Page 35
Conceptual Model
Relational Model
7-5
Conceptual Model The way you can describe requirements for the data business
requires using a semantically rich syntax through graphical representation. As you
have seen in previous chapters, many of the business rules can be described with
graphical elements such as subtypes, arcs, relationships (barred and nontransferable
ones). The only constraints in expressing business complexity that you have
encountered so far are the graphical limitations. We know that this model acts as a
generic one, because it is not related to any physical considerations. Therefore it can
be used for any type of database. Nevertheless, it may be that the DBMS type you
want to use (relational or others) does not support all of the semantic rules graphically
expressed in your ER model.
Relational Model. The Relational model is based on mathematical rules. This

means that when you try to fit all of the syntax from the ER model into the physical
database model, some of it may not have any correspondence in the relational model.
To preserve these specified rules, you have to keep track of them and find the correct
way to implement them.
.....................................................................................................................................................
Transformation Process
.....................................................................................................................................................
Terminology Mapping
About the slide
Terminology Mapping See Page 35
ANALYSIS DESIGN
ER Model Physical Design
Entity Table
Attribute Column
Primary UID Primary Key

Secondary UID Unique Key
Relationship Foreign Key
Business Constraints Check Constraints
7-6
Changing from one world to an other also means changing terminology.

Using a very simple basis:
• An entity leads to a table.
• An attribute becomes a column.
• A primary unique identifier produces a Primary key.
• A secondary unique identifier produces a Unique key.
• A relationship is transformed into a Foreign key and foreign key columns.
• Constraints are the rules with which the database must cope to be consistent. Some
of the business rules are translated into Check Constraints, other complex ones
require additional programming and can be implemented at client side or server
side or both.
This initial mapping of an ER model is limited to the design of tables, columns. and
constraints that can be declared. A declarative constraint is a business constraint that
can be ensured at the server level using database language statements only and
requires no coding.
.....................................................................................................................................................
®
7-7
.....................................................................................................................................................
Naming Convention
Before transforming the ER diagram you probably need to define a naming convention
so that people working on the project use the same standards and produce the same
model from the same source. Rules explained here are the ones used within Oracle.
Even though they are efficient, they are not the only ones that you can use and you or
your company may provide the company’s own standard as part of its method.
About the slide

General Naming Topics See Page 35
Decide on a convention for:

• Table names
• Special characters (%, *, #, -, space, …)
• Table short names
• Column names
• Primary and Unique Key Constraint names
• Foreign Key Constraint names
• Foreign Key Column names
7-7
Naming of Tables
The plural of the entity name is used as the corresponding table name. The idea is that
the Entity is the concept of an abstract thing—you can talk about EMPLOYEE,
CUSTOMER, and so on, so singular is a good naming rule, but a table is made up of
rows (the EMPLOYEES table, or CUSTOMERS table) where the plural is more
appropriate.
Naming of Columns
Column names are identical to the attribute names, with a few exceptions. Replace
special characters with an underscore character. In particular, remove the spaces from
attribute names, as SQL does not allow spaces in the names of relational elements.
Attribute Start Date converts to column Start_date; attribute Delivered Y/N transforms
to Delivered_y_n (or preferably Delivered_Ind). Often column names use more
abbreviations than attribute names.
.....................................................................................................................................................
Naming Convention
.....................................................................................................................................................
Short Names
A unique short name for every table is a very useful element of the name of foreign
key columns or foreign key constraints. A suggested way to make these short names is
based on the following rules:
• For entity names of more than one word, take the:
– First character of the first word.
– First character of the second word.
– Last character of the last word.
For example entity PRICED PRODUCT produces PPT as a short table name.
• For entity names of one word but more than one syllable, take the:
– First character of the first syllable.
– First character of the second syllable.
– Last character of the last syllable.
For example EMPLOYEE gives EPE as a short name.
• For entity names of one syllable, but more than one character, take the:
– First character.
– Second character.
– Last character.
For example FLIGHT gives FLT.
This short name construction rule does not guarantee uniqueness among short names
but experience has proved that duplicated names are relatively rare.
In case two short names happen to be the same, just add a number to the one that is
used less often giving, for example, CTR for the most frequently used one and then
CTR1 for the second one.
Naming of Foreign Key Constraints

The recommended rule for naming foreign key constraints is
<short name of the from table> _ < short name of the to table> _ < fk>.
For example, a foreign key between tables EMPLOYEES and DEPARTMENT results
in constraint name epe_dpt_fk.
Naming of Foreign Key Columns

Foreign key columns are prefixed with the short name of the table they refer to. This
leads to foreign key column names like dpt_no. Limiting the attribute name to 22
characters enables you to add two prefixes plus two underscores to the column name.
This may occur in the event of cascade barred relationships. This is discussed later in
the lesson.
.....................................................................................................................................................
®
7-9
.....................................................................................................................................................
Multiple Foreign Keys Between Two Tables

If there are two (or more) foreign keys between two tables then the foreign keys and
foreign key columns would be entitled to the same name. In this situation, add the
name of the relationship to both foreign key names. Do the same with the foreign key
columns. This way you will never mistake the one foreign key for the other.
For example, in the model of Electronic Mail entity LIST ITEM has two relationships
with ALIAS (one of them is at the subtype level). The naming would result in the two
foreign key names: lim_als_in and lim_als_referring_to. The foreign key columns
would be named Als_id_in and Als_id_referring_to.
Naming of Check Constraints

Check Constraints are named <table short name>_ck_<sequence_number>, such as
epe_ck_1, epe_ck_2 for the first and second check constraint on table EMPLOYEES.
Naming Restrictions with Oracle

Each RDBMS can have its own naming restrictions. You need to know if the
convention you decide to use is compatible with it.
Naming Restrictions with Oracle
• Table and column names:

– Must start with a letter
– May contain up to 30 alphanumeric characters
– Cannot contain space or special characters
• Table names must be unique within a schema.
• Column names must be unique within a table.
7-8
• You can use any alpha-numeric character for naming tables and columns as long as
the name:
– Starts with a letter.
– Is up to 30 characters long.
– Does not include special characters such as “!”.
.....................................................................................................................................................
Naming Convention
.....................................................................................................................................................
• Table names must be unique within the name space that is shared with views and
synonyms.
• Within the same table two columns cannot have the same name.
• Be aware also of the reserved programming language words that are not allowed to
be used for naming objects. Avoid names like
– Number
– Sequence
– Values
– Level
– Type
for naming tables or columns. Refer to the RDBMS reference books for these.
.....................................................................................................................................................
®
7-11
.....................................................................................................................................................
Basic Mapping
Entity Mapping
Before going into complex transformation we will look at the way to transform simple
entities.
Basic Mapping
1 - Entities
2 - Attributes
3 - Unique identifiers Table Name: EMPLOYEES
Short Name: EPE
Primary EMPLOYEE EMPLOYEES (EPE)

UID # Id pk * Id
* Name uk1 * Name
o Address o Address
uk1 * Birth_date
* Birth Date
Secondary
UID
7-11
1 Transform entities into tables using your own naming convention or the one
previously described.
In this example the entity EMPLOYEE produces a table name EMPLOYEES and
a short name EPE.
Use a box to represent tables on a diagram.
2 Each attribute creates a column in the table and the characteristics such as
mandatory or optional have to be kept for each column. Using the same notation
“*” or “o” facilitates recognition of these characteristics on a diagram.
3 All unique identifiers are transformed. A primary unique identifier is transformed
into a Primary key. The notation “pk” next to the column name indicates the
Primary key property. If more than one column is part of the primary key, use the
“pk” notation for each column.
Secondary unique identifiers, even if they do not appear on your ER diagram, need to
be implemented. To preserve this property, secondary UIDs are transformed as unique
keys. In the above example, the values for the combination of two columns must be
unique. They belong to the same unique key and each column has a uk1 notation to
indicate this. If, in future, another unique key might exist for that table, it would be
notated as uk2.
.....................................................................................................................................................
Basic Mapping
.....................................................................................................................................................
Rules for Relationships

About the slide
Rules for Relationships See Page 36
EMPLOYEE
# Id DEPARTMENT
* Name # Id
o Address * Name
* Birth Date
fk2 = epe_epe_fk
EMPLOYEES (EPE)
pk * Id
* Name DEPARTMENTS (DPT)
fk1 * Dpt_id pk * Id
fk2 o Epe_id fk1 = epe_dpt_fk uk * Name
7-12
Foreign Key Columns: A relationship creates one or more foreign key columns in
the table at the many side. Using previous naming rules, the name of this foreign key
column is Dpt_id for the relationship with Department and Epe_id for the recursive
relationship. This ensures that column names such as Id, coming from different tables,
still provide a unique column name in the table.
Depending on whether or not the relationship is required, the foreign key column is
mandatory or optional.
Foreign Key Constraints: The foreign key constraints between EMPLOYEES and
DEPARTMENTS is epe_dpt_fk. The recursive one between EMPLOYEES and
EMPLOYEES is called epe_epe_fk.
.....................................................................................................................................................
®
7-13
.....................................................................................................................................................
Relationship Mapping
Mapping of One-to-Many Relationships
As previously mentioned, some of the meaning that is expressed in an ERD cannot be
reproduced in the physical database design.
.
About the slide
Mapping 1:m Relationships See Page 37
XS
fk o Y_id
XS
fk * Y_id
7-13
A relationship in an ER Diagram expresses the rules that apply between two entities,
from two points of view. The notation used in the ERD is rich enough to tell, for
example, that the relationship is mandatory on both sides. The illustration shows that
the 1:m relationships that are mandatory at the one side are implemented in exactly the
same way as the ones that are optional at the one side. This means that part of the
content of the ER model is lost during transformation, due to the relational model
limitations. You need to keep track of these incomplete transformations; they must be
implemented using a mechanism other than a declarative constraint.
Mapping of Mandatory Relationship at the One Side

In case of the implementation of a relationship that is mandatory at the one side you
need to check two things.
• You cannot create any master record without at least one detail record.
• When deleting details you must be sure that you do not delete the last detail for a
master record, or alternatively, you must delete the master record together with its
last detail.
.....................................................................................................................................................
.....................................................................................................................................................
You can implement code to check this on the server side or on the client side. In an
Oracle environment this was usually done at the client side. Since Oracle 8, on the
server side Oracle offers implementation possibilities that were not available in
previous releases.
Optional Composed Foreign Keys

When a foreign key is made of two or more columns, and the foreign key is optional,
all foreign key columns must be defined as optional. Note that if you enter a value in
one of the foreign key columns, but not in the other one, Oracle will not fire the
foreign key constraint check.
You would need additional code to check that either all or none of the foreign key
columns have a value, but exclude the possibility of a partially-entered key.
Mapping of Nontransferable Relationships

About the slide
Mapping Barred and Nontransferable See Page 37
Relationships
X Y
# Id
* C1 # Id
* C2
XS (X) YS (Y)
pk * Id fk = y_x_fk pk * Id
* C1 pk, fk * X_id
* C2
7-14
This relationship property does not migrate to the physical database design because it
has no natural counterpart in an RDBMS, although you can code a solution at the
server side. In the example, you would create an update trigger at table YS that fails
when the foreign key column X_id is updated.
Mapping Barred Relationships

A barred relationship, like any other relationship, is mapped into a foreign key. The
foreign key column is also part of the primary key, and thus plays a double role.
.....................................................................................................................................................
®
7-15
.....................................................................................................................................................
Mapping of Cascade Barred Relationships

A Cascade Barred relationship may lead to long column names as the illustration
shows.
About the slide
Mapping Cascade Barred Relationships See Page 37
A B C D
# Id # Id # Id # Id
* C1 * C2 * C3 * C4
AS (A) BS (B) CS (C) DS (D)

pk * Id pk * Id pk * Id pk * Id
* C1 * C2 * C3 * C4
fk,pk * A_id fk,pk * B_id fk * C_id
fk,pk * B_a_id fk * C_b_id
fk = b_a_fk
fk * C_a_id
fk = c_b_fk
fk = d_c_fk
7-15
To avoid column names that could end up with more than 30 characters, the suggested
convention is never to use more than two table prefixes.
The usual choice for the foreign key column names is:
<nearest by table short name> _ <farthest table short name> _ <column name>
In the above example the foreign key column in DS that comes all the way from AS
via BS and CS is named C_a_id instead of C_b_a_id.
As the short names are usually three characters long, this rule explains why attribute
names should not have more than 22 characters.
.....................................................................................................................................................
.....................................................................................................................................................
Mapping of Many-to-Many Relationships

When transforming a many-to-many relationship, you create an intersection table.
About the slide
Mapping m:m Relationships See Page 37
X Y
# Id # Id
* C1 * C2
XS YS
pk * Id pk * Id
* C1 X_YS * C2
pk,fk1 * X_id
pk,fk2 * Y_id
fk1 = xy_x_fk fk2 = xy_y_fk
7-16
The intersection table contains all the combinations that exist between XS and YS.
• This table has no columns other than foreign key columns. These columns together
form the primary key.
• The rule for naming this table is short name of the first table (in alphabetical order)
and full name of the second one. This would give a many-to-many relationship
between tables EMPLOYEES and PROJECTS an intersection table named
EPE_PROJECTS.
• Whether the relationship was mandatory or not, the foreign key columns are
always mandatory.
Note this table is identical (maybe except for its name) to the table that would result
from an intersection entity that could replace the m:m relationship.
.....................................................................................................................................................
®
7-17
.....................................................................................................................................................
Mapping of One-to-One Relationships

About the slide
Mapping 1:1 Relationships See Page 38
X Y
# Id # Id
* C1 * C2
YS (Y)
XS (X)
fk = y_x_fk pk * Id
pk * Id * C2
* C1 fk,uk * X_id
Choose which side for FK for other cardinalities
7-17
When transforming a one-to-one relationship, you create a foreign key and a unique
key. All columns of this foreign key are also part of a unique key.
If the relationship is mandatory on one side, the foreign key is created at the
corresponding table. If the relationship is mandatory on both sides or optional on both
sides, you can choose on which table you want to create the foreign key. There is no
absolute rule for deciding on which side to implement it.
If the relationship is optional on both sides you may decide to implement the foreign
key in the table with fewer numbers of rows, as this would save space.
If the relationship is mandatory at both ends, we are facing the same RDBMS
limitation you saw earlier. Therefore, you need to write code to check the mandatory
one at the other side, just as you did to implement m:1 relationships that are mandatory
at the one end.
Alternative Implementations
A 1:1 relationship between two entities can be implemented by a single table. This is
probably the first implementation to consider. It would not need a foreign key
constraint.
A third possible implementation is to create an intersection table, as if the relationship
was of type m:m. The columns of each of the foreign keys of the intersection table
would be part of unique keys as well.
.....................................................................................................................................................
.....................................................................................................................................................
Mapping of Arcs
About the slide
Mapping Arcs See Page 38
Explicit implementation
USER
# Id
LIST ITEM * Name
ALIAS
# Id
fk1 = lim_x_fk
USERS (USR)
LIST_ITEMS (LIM)
fk2 = lim_usr_fk pk * Id
pk,fk1 * X_id * Name
fk2 o Usr_id
fk3 o Als_id ALIASES (ALS)
fk3 = lim_als_fk pk * Id
+ check constraint
7-18
The first solution illustrated above shows that there are as many foreign keys created
as there are relationships. Therefore a rule must be set to verify that if one of the
foreign keys is populated, the others must not be populated (which is the exclusivity
principle of the relationships in an arc) and that one foreign key value must always
exist (to implement the mandatory condition).
From a diagram point of view, all foreign keys must be optional, but additional code
will perform the logical control. One solution on the server side is to create a check
constraint at LIST_ITEMS as is:
CHECK ( usr_id IS NOT NULL
AND als_id IS NULL)
OR ( usr_id IS NULL
AND als_id IS NOT NULL).
This controls the exclusivity of mandatory relationships.
In case the relationships are optional, you need to add:
OR (usr_id IS NULL AND als_id IS NULL)
An other syntax that is often used:
DECODE (usr_id,NULL,0,1)
+ DECODE (als_id,NULL,0,1)=1;
(or =<1 for optional relationship).
You can also map arcs in a different way using the generic arc implementation. This is
a historical solution that you may encounter in old systems. You should not use it in
new systems. It is discussed in the lesson on Design Considerations.
.....................................................................................................................................................
®
7-19
.....................................................................................................................................................
Mapping of Subtypes
In mapping subtypes, you must make a choice between three different types of
implementations. All three are discussed in detail.
Mapping Subtypes
Variety of implementation choices
P K
# Id # Id
* Xxx
A • Supertype
Q
o Yyy # Id
• Subtype
R B
• Both Supertype
* Zzz # Id and Subtype (“Arc”)
L
# Id
7-19
Supertype Implementation
This choice produces one single table for the implementation of the entities P, Q, and
R. The supertype implementation is also called single (or one) table implementation.
Rules
1 Tables:
– Independent of the number of subtypes, only one single table is created.
2 Columns:
– The table gets a column for all attributes of the supertype, with the original
optionality.
– The table also gets a column for each attribute belonging to the subtype but the
columns are all switched to optional.
– Additionally, a mandatory column should be created to act as a discriminator
column to distinguish between the different subtypes of the entity. The value it
can take is from the set of all the subtype short names (DBE, DBU in the
example). This discriminator column is usually called <table_short_ name> _
type, in the example Dba_type.
3 Identifiers:
– Unique identifiers translate into primary and unique keys.
.....................................................................................................................................................
Mapping of Subtypes
.....................................................................................................................................................
– Unique identifiers at subtype level usually translate into a unique key or check
constraint only.
About the slide
See Page 38
Supertype Implementation
P K
# Id # Id
* Xxx
A PS (P)
Q
o Yyy # Id
pk * Id
* Xxx
R B o Yyy
* Zzz # Id o Zzz
fk1 * A_id
fk2 o B_id
• Mandatory
L discriminator * P_type
# Id column
• Additional
constraints
7-20
4 Relationships:
–Relationships at the supertype level transform as usual. Relationships at
subtype level are implemented as foreign keys, but the foreign key columns all
become optional.
5 Integrity constraints:
– For each particular subtype, all columns that come from mandatory attributes
must be checked to be NOT NULL.
– For each particular subtype, all columns that come from attributes or
relationships of other subtypes must be checked to be NULL.
Note: You may avoid the use of the discriminator column if you have one
mandatory attribute in each subtype. The check is done directly on these columns to
find out what type a specific row belongs to.
When to Consider Supertype Implementation

The single table implementation is a common and flexible implementation. It is the
one you are likely to consider first and is specially appropriate when:
• Most of the attributes are at the supertype level.
• Most of the relationships are at the supertype level.
• The various subtypes overlap in the required functionality.
.....................................................................................................................................................
®
7-21
.....................................................................................................................................................
• The access path to the data of the various types is the same.
• Business rules are globally the same for the subtypes.
• The number of instances per subtype does not differ too much, for example, one
type having more than, say, 1000 times the number of instances of the other.
• An instance of one subtype can become an instance of another, for example,
imagine an entity ORDER with subtypes OPEN ORDER and PROCESSED
ORDER, each subtype having its own properties. An OPEN ORDER may
eventually become a PROCESSED ORDER.
Additional Objects
Usually you would create a view for every subtype, showing only the columns that
belong to that particular subtype. The correct rows can be selected using a condition
based on the discriminator column. These views can be used for all data operations,
including inserts and updates. All applications can be based on the view, without loss
of performance.
The supertype table plus subtype views is an elegant and appropriate implementation
and should be considered as first choice.
Consequences for Tables Based on K and L

The foreign key in the table based on K is straightforward.
The foreign key of the table based on L is more complex. The supertype
implementation would mean that the foreign key refers to a valid P, not to the more
limited set of R’s. This must be checked with an additional constraint.
.....................................................................................................................................................
Subtype Implementation
.....................................................................................................................................................
This subtype table implementation (often loosely referred to as two-table
implementation) produces one table for each of the subtypes, assuming there are only
two subtypes, such as Q and R.
About the slide
Subtype Implementation See Page 39
P K QS (Q)
# Id # Id pk * Id q_a_fk
* Xxx * Xxx
A o Yyy
Q fk * A_id
o Yyy # Id
R B
* Zzz # Id RS (R)
fk1=r_a_fk
pk * Id
* Xxx
* Zzz fk2=r_b_fk
L
fk1 * A_id
# Id fk2 * B_id
7-21
Rules
1 Tables:
– One table per first level subtype.
2 Columns:
– Each table gets a column for all attributes of the supertype, with the original
optionality.
– Each table also gets a column for each attribute belonging to the subtype, also
with the original optionality.
3 Identifiers:
– The primary unique identifier at the supertype level creates a primary key for
each of the tables. Alternatively, if the subtypes had their own UID, this one
can be used as the basis for the primary key.
– Secondary identifiers of the supertype become unique keys within each table.
4 Relationships:
– All tables get a foreign key for a relationship at the supertype level with the
original optionality.
.....................................................................................................................................................
®
7-23
.....................................................................................................................................................
–For the relationships at the subtype levels, the foreign key is implemented in
the table it is mapped to. The original optionality is retained.
– No specific additional checks are required. Only when the Id values must be
unique across all subtypes would it need further attention.
When to Consider a Subtype Implementation

You can regard this implementation as a horizontal partitioning of the supertype. It
may be appropriate when:
• The resulting tables will reside in different databases (distribution). This may
occur when different business locations are only interested in a specific part of the
information.
• When the common access paths for the subtypes are different.
• Subtypes have almost nothing in common. This may occur when there are few
attributes at the supertype and many at the subtype levels. An example can be
found in the Electronic Mail model. Entity ADDRESS has two subtypes: MAIL
LIST and ALIAS. These subtypes only share the fact that they can be used as
addressee for a message, but their other properties are completely different.
• Most of the relationships are at the subtype level. This is the case especially if both
tables are to be implemented in different databases, and the foreign key integrity
constraint for the supertype may not be verified in all cases.
• Business functionality and business rules are quite different between subtypes.
• The way tables are used is different, for example, one table being queried while the
other one is being updated. A one-table solution could result in performance
problems.
• The number of instances of one subtype is very small compared to the other one.
Additional Objects
Usually you would create an additional view that represents the supertype showing all
columns of the supertype and various subtypes. The view select statement must use the
union operator. The view can be used for queries only, not for data manipulation.

The foreign key in the table based on L is straightforward and should refer to the table
based on R.
The foreign key of the table based on K is now more complex. This must be
implemented as two optional foreign keys, one to each of the tables based on Q and R.
An extra check is needed to make sure that both foreign keys do not have a value at the
same time; this is identical to an ordinary arc check.
.....................................................................................................................................................
.....................................................................................................................................................
Both Supertype and Subtype “Arc” Implementation

About the slide
Supertype and Subtype (Arc) Implementation See Page 39
PS (P)
P K
# Id pk * Id
# Id
* Xxx * Xxx fk3 =
fk1,uk1 o Q_id
A fk2,uk2 o R_id p_a_fk
Q
oYyy # Id fk3 * A_id
R B
* Zzz # Id fk1 = fk2 =
p_q_fk p_r_fk
QS (Q) RS (R)
L pk pk * Id
* Id r_b_fk
# Id o Yyy * Zzz
fk * B_id
7-22
This choice produces one table for every entity, linked to foreign keys in an exclusive
arc at the PS side. It is the implementation of the model as if the subtypes were
modeled as standalone entities with each one having an is subtype of / is supertype of
relationship to the supertype. These relationships are in an arc. Therefore this
implementation is also called Arc Implementation. See also the chapter on Constraints
for more details about subtypes compared to the arc.
Rules
1 Tables:
– As many tables are created as there are subtypes, and one for the supertype.
2 Columns:
– Each table gets a column for all attributes of the entity it is based on, with the
original optionality.
3 Identifiers:
– The primary UID at the supertype level creates a primary key for each of the
tables.
– All other unique identifiers transform to unique keys in their corresponding
tables.
4 Relationships:
– All tables get a foreign key for a relevant relationship at the entity level with
the original optionality.
.....................................................................................................................................................
®
7-25
.....................................................................................................................................................
– Two additional columns are created in the table based on the supertype. They
are foreign key columns referring to the tables that implement the subtypes.
The columns are clearly optional as the foreign keys are in an arc. The foreign
key columns are also part of the unique keys because, in fact, they implement a
mandatory one-to-one relationship.
– An additional check constraint is needed to implement the arc.
When to Consider a Both Supertype and Subtype Implementation

This solution performs a double partitioning. It is used relatively rarely, but could
be appropriate when:
– The resulting tables reside in different databases (distribution). This may occur
when different business locations are only interested in a specific part of the
information.
– Subtypes have almost nothing in common and each table represents
information that can be used independently. For example, when the PS table
gives all global information and both QS and RS give specific information, and
the combination of global and specific information is hardly ever needed.
– Business rules are quite different between all types.
– The way tables are used and accessed is different.
– Users from different business areas need to work with the same rows at the
same time, but with different parts of the rows, which could result in locking
problems and a performance issue.
Additional Objects
Although you would hardly use them, you could consider creating additional views
that represent the supertype and various subtypes in full.

Both foreign keys can be implemented straightforwardly without additional checks.
.....................................................................................................................................................
.....................................................................................................................................................
Storage Implication
The illustrations show the differences between the one, two and three table
implementations. In most database systems empty column values do take some bytes
of database space (although this sounds contradictory). In Oracle this is very low when
the empty columns are at the end of the table and when the data type is of variable size.
Supertype Implementation All rows for both types are in one table. Note the
empty space in the Q rows at the R columns and vice-versa.
About the slide
Storage Implication See Page 39
Supertype Implementatioin
discriminator column
cols cols cols
P
P Q R
Q
rows Q
R
rows R
Subtype Implementation In the two table implementation the “empty space” of

the one-table implementation is gone. This is a horizontal split of the table.
Storage Implication
cols cols cols cols

P R P Q
rows Q
rows R
.....................................................................................................................................................
®
7-27
.....................................................................................................................................................
Arc Implementation In this three table implementation the one table is sliced
vertically into a P-columns-only portion. The remaining part is horizontally split into
the Q and R columns and rows. An additional foreign key column at P, or a foreign key
column at Q and R each is needed to connect all the pieces together.
Storage Implication
Supertype and Subtype (Arc) Implementation
cols
P
fk cols fk cols
rows Q Q R
rows Q
rows R rows R
7-26
.....................................................................................................................................................
Summary
.....................................................................................................................................................
Summary
Summary
• Relational concepts
• Naming rules convention
• Basic mapping
• Complex mapping
7-27
Relational databases implement the relational theory they are based on.
A coherent naming rule can prevent many errors and frustrations and adds to the
understanding of the structure of the database schema.
You have seen how to map basic elements from an ER model such as entities and
relationships. These can be done in a very simple way. There are also complex
structures which require decisions on how to transform them. Some ER model
elements can only be implemented by coding check constraints or database triggers.
These are specific to Oracle and not part of the ISO standard for relational databases.
.....................................................................................................................................................
®
7-29
.....................................................................................................................................................
Practice 7—1: Mapping Supertype

Moonlight Coffees
Goal (See Page 40)
In this practice, you create a complex mapping and test your
understanding of the transformation process.
Scenario
Here is part of the Moonlight ER model showing the entity DEPARTMENT. One of
the analysts has decided to implement the DEPARTMENT entity and its subtypes as a
single table.
pp g p yp
DEPARTMENT report
# Id HQ COUNTRY
of
* Name * Address ORGANIZATION
* Head Count reporting # Tax Id Number
to
OTHER DEPARTMENT
Your Assignment
1 What would have been the rationale of this choice?
2 On the table diagram, name all the elements that must be created following this
supertype implementation. Use the naming convention as described in this lesson,
or use your own rules. Give proper names to the columns and foreign key
constraints and identify check constraints, if any.
DEPARTMENTS ( )
7-29
.....................................................................................................................................................
Practice 7—2: Quality Check Subtype Implementation
.....................................................................................................................................................
Practice 7—2: Quality Check Subtype Implementation

Moonlight Coffees
Goal (See Page 40)
In this practice you perform a quality check on table mappings
that were created by someone else who is supposed to use the
naming convention that is described in this lesson.
Scenario
Here is a part of the Moonlight ER model.
COUNTRY
# Code
with PRODUCT GROUP with
# Name in
with SHOP
with
in # No
* Name
PRODUCT * Address
for
GLOBAL of * City
PRICE LIST # Code LOCAL
# Start Date o Size # Name
* End Date
with with
in of
GLOBAL PRICE
* Amount
7-30
Your Assignment
Perform a quality check on the proposed subtype implementation of entity
PRODUCT.
lpt_shop_fk
GLOBAL_PRODUCTS (GPT) LOCAL_PRODUCTS (LPT)

pk * Code pk # Name
o Size fk * Shop_no
* Pgp_name fk * Pgp_name
.....................................................................................................................................................
®
7-31
.....................................................................................................................................................
Practice 7—3: Quality Check Arc Implementation

Moonlight Coffees
Goal (See Page 41)
The purpose of this practice is to do a quality check on table
mappings that were created by someone else who is supposed to
use the naming convention that is described in this lesson.
Scenario
This practice is based on the same ER diagram as the previous practice.
Practice: Quality Check

Arc Implementation
PRODUCTS (PDT) fk1=pdt_pgp_name

pk * Code fk2=pdt_gpt_code
fk1
* Pgp_name
fk2 * Gpt_code fk3=pdt_lpt_name
fk3 * Lpt_name
GLOBAL_PRODUCTS (GPT)
gpt_pgp_fk
pk * Code
o Size
LOCAL_PRODUCTS (LPT)
fk1=shp_lpt_fk
pk * Name
pk, fk1 o Shp_no fk2=pgp_lpt_fk
fk1 * Pgp_name
7-32
Your Assignment
Perform a quality check on the proposed supertype and subtype implementation of the
entity PRODUCT and its subtypes. Also, check the selected names.
.....................................................................................................................................................
Practice 7—4: Mapping Primary Keys and Columns
.....................................................................................................................................................
Practice 7—4: Mapping Primary Keys and Columns

Moonlight Coffees
Goal (See Page 41)
The purpose of this practice is to do a complex mapping of
primary keys and columns.
Scenario
This practice is based on the same model that was used in the previous practice.
Your Assignment
Identify the Primary key columns and names resulting from the transformation of the
GLOBAL PRICE entity. Give the short name also.
GLOBAL_PRICES ( )
.....................................................................................................................................................
®
7-33
.....................................................................................................................................................
Instructor Notes
Instructor Note
Topic Timings
77
Lecture 45 minutes
M apping the Entity M od el
Total 60-90 minutes
This lesson deals with a relational database

O verview
design. The approach would be the same
• W h y u se d es ig n m o de lin g ? independently of the RDBMS used. It applies
• Intro d u c tio n to th e c o m p o n en ts:
– T ab le s as well to DB2, Sybase, Informix, SQL sever,
– C o lum n s
– C o n s train ts or Oracle. It is technique dependent (relational)
• B a sic M ap ping
• C o m p lex m a pp in g but software independent
7 -2
A data design model can be derived from an ER

W hy C re ate a D ata D esign M ode l?
model (which is the flow down approach we
• C lo s er to th e im plem en tatio n s o lu tio n advise and follow in this course), but it could
• F a cilitates d is c us sio n
• Ide al m o d e l c an b e a d ap ted to a n R D B M S m o d e l also be created directly from an existing
• S o un d b as is fo r p h ys ica l d atab a se d es ig n
database.
7 -3
Students should be aware of Table structures.

Pres enting Tables
Ta ble: E M P LO YE E S
Theoretically every single table must have a
c olu m n s
primary key. Oracle allows tables to be created
Id N a me A dd r es s B ir t h_ d a te D p t_ i d
ro w s
12 6
34 9
P A GE 1 2, OX FO R D S T
P A PI N I 5 3, HA YE S AV E
0 3 -0 3 - 66
1 0 -0 8 - 77
10
20
without primary keys. Oracle also allows a
78 5 G A RR E T 0 8 -1 2 - 55 10
prim ary ke y
co lum n
un iqu e k ey
co lu m n
fo re ign key
co lum n
foreign key to reference a unique key instead of
EM P LO Y E ES (EP E )
pk * Id
a primary key.
Ta ble d ia gram : E M P LO Y E E S uk 1 * Name
uk 1 *
fk *
o A d dress
B irth_date
D p t_id
fo re ign
key This slide shows the way a table can be
7 -4 visualized and the way we represent tables in
this book in the table diagrams. The notation is
similar (but not identical) to the one used in
Oracle Designer.
.....................................................................................................................................................
Instructor Notes
.....................................................................................................................................................
The bits and pieces that go into the wastebasket

Trans form ation Proc ess
C onceptual M o del are the things that have no counterpart that can
be declared in the relational world. Most of
these pieces however, can be coded manually.
R elational M o del
7 -5
Note there are physical objects that do not have

Te rm ino lo gy M a pping
a ER counterpart. The most obvious one is a
A N A L Y S IS DE S IG N
view.
E R M odel P hysical D es ign
E ntity Table
A ttribute C olum n
P rim ary U ID P rim a ry K ey Primary key, unique key and foreign key are
S econdary U ID U niqu e K ey
R elationship Foreig n K ey constraints that can be declared, that is, they can
B usin ess C onstraints C heck C onstraints
be created without any coding in third-
7 -6
generation languages.
The naming convention described here and

G en era l N a m ing T opics
used in this book is the one used in Oracle’s
D ecide on a convention for: method: CDM.
• T able n am es
• S pe cia l ch aracte rs (% , *, #, -, sp ac e, … )
• T able sh ort nam es
• C o lum n na m es
• P rim a ry an d U n iqu e K ey C on straint n am es
• F oreig n K e y C o nstra int nam es
• F oreig n K e y C o lum n nam es
7 -7
N am ing R estrictions w ith O racle
• T a b le an d c o lu m n n am e s:
– M u s t start w ith a letter
– M a y c o n tain up to 30 alp h an u m e ric c ha ra cters
– C a n n o t co n tain sp a ce o r s p ec ia l c h ara cte rs
• T a b le n am e s m u st b e u n iq u e w ith in a s ch e m a .
• C o lu m n n am e s m u st b e u n iq u e w ith in a tab le.
7 -8
.....................................................................................................................................................
®
7-35
.....................................................................................................................................................
First map entities to tables.

B asic M apping for
E ntities
1 - En tities
T able N am e: E M P LO Y E ES
S ho rt N am e: EP E
E M P LO Y E E S (E P E )
E M P LO Y E E
7 -9
Next map attributes to columns.

B asic M apping for
A ttrib utes
1 - En tities
2 - A ttrib ute s
T able N am e: E M P LO Y E ES
E M P LO Y E E
# Id * Id
* N am e * N am e
o A d dress o A ddress
* B irth D ate * B irth_da te
7 -1 0
Create primary and unique keys.

B as ic M app ing
Then the foreign keys and all additional
1 - En tities
2 - A ttrib ute s
constraints can be defined.
3 - U niq ue ide ntifiers T able N am e: E M P LO Y E ES
Prim ary E M P LO Y E E E M PL O Y EE S (E P E )
U ID # Id pk * Id
* N am e uk 1 * N am e
o A d dress o A ddress
uk 1 * B irth_da te
* B irth D ate
Se cond ary
U ID
7 -1 1
A relationship is mapped into two things:

R ules fo r R elation ships
• A foreign key constraint
E M P LO Y E E
# Id
* N am e
o A ddress
D E P A R TM E N T
# Id
* N am e
• One or more foreign key columns
* B irth D ate
fk 2 = epe_ep e_fk
The representation of the tables and foreign
pk * Id
keys does not correspond to an official standard
* N am e
fk 1 *
fk 2 o
D pt_id
E pe_id
pk
fk 1 = epe_d pt_fk uk
D E P A R TM E N TS (D P T )
* Id
as there is no official convention. The
* N am e
convention used here is selected for educational
7 -1 2
reasons.
.....................................................................................................................................................
Instructor Notes
.....................................................................................................................................................
The mapping of a 1:m relationship that is

M app ing 1 :m R elationsh ip s
mandatory at the 1-end is initially identical to
XS
that of a 1:m relationship which is optional at
fk o Y _id the 1-end.
XS Very often the “mandatory 1” was not
fk * Y _id
implemented at all, or was implemented in the
applications.
7 -1 3
Since Oracle 8i there are possibilities for
implementing it at the server side. Solutions
involve triggers and need to make use of the
deferrable option for constraints.
The fact that a relationship is barred is mapped
M ap pin g B arred an d N on tra nsferable
R elatio nsh ips on a primary key.
X
# Id
Y Nontransferability cannot be declared.
* C1 # Id
* C2
The code on the server side must check that
X S (X ) Y S (Y )
foreign key columns are not updatable. You
pk *
*
Id
C1
fk = y_x_fk pk
pk, fk
*
*
*
Id
X _id
C2
could use a before update trigger for each row
to check this.
7 -1 4
Clearly, the table based on entity D has a

M app ing C asca de B a rre d R elatio nships
foreign key that consists of a lot of columns.
A B C D
# Id
* C1
# Id
* C2
# Id
* C3
# Id
* C4 Note the name of the foreign key columns in
table DS: C_id, C_b_id and C_a_id (instead of
A S (A ) B S (B ) C S (C ) D S (D )
C_b_a_id).
pk * Id pk * Id pk * Id pk * Id
* C1 * C2
fk,pk * A _id
* C3
fk ,pk * B _id
fk ,pk * B _a_id
fk
fk
* C4
* C _id
* C _b_id
We do not use a “barred foreign key” notation
fk = b_a_fk
fk = c_b_fk
fk = d_c_fk
fk * C _a_id
as this is very confusing for Oracle Designer
7 -1 5
users. There the bar across a foreign key has the
completely different meaning of restrict delete.
Whatever the optionalities of a m:m
M ap pin g m :m R elationsh ip s
relationship, it will be implemented as an
X Y
intersection or binary table. The intersection
# Id
* C1
# Id
* C2 table always has two mandatory foreign keys.
XS YS
pk * Id pk * Id
* C1 X _Y S * C2
pk,fk1 * X _ id
pk,fk2 * Y _ id
fk1 = xy_x_fk fk2 = xy_y_fk
7 -1 6
.....................................................................................................................................................
®
7-37
.....................................................................................................................................................
A 1:1 relationship is implemented as a 1:m

M app in g 1:1 R elation ships
relationship. The foreign key columns are also
X
# Id
Y
# Id
* C2
part of a unique key to “safeguard the 1”.
* C1
Note that the first implementation to consider is

X S (X )
fk = y_ x_fk
Y S (Y )
pk * Id
a single table implementation. In this case you
pk * Id
* C1
* C2
fk, uk * X _ id would consider table XS with mandatory
C ho ose w hich side fo r F K fo r other ca rd ina lities
columns Id, C1 and optional C2.
7 -1 7
Foreign keys in an arc consist of optional

M app in g A rcs
E xplicit im plem en tation
columns only. An additional check constraint
LIS T ITE M
USER
# Id
* N am e
must be created to safeguard the fact that at
A LIA S
# Id
least one of the foreign keys has a value.
fk1 = lim _x_fk
U S E R S (U S R )
LIS T_ITE M S (LIM )
fk2 = lim _usr_fk pk * Id
pk,fk 1 * X _id * N am e
fk 2 o U sr_id
fk 3 o A ls_ id A LIA S E S (A LS )
fk3 = lim _als_fk pk * Id
+ che ck c on strain t
7 -1 8
M app in g Sub typ es
Va rie ty of im p lem en tation ch oic es
P K
# Id # Id
* X xx
A • S upe rtype
Q
Y yy # Id
o • S ubtyp e
R B
• B oth S up ertyp e
* Zzz # Id an d S ub typ e (“A rc”)
L
# Id
7 -1 9
Supertype or “single table” implementation is a

Su pertype Im plem entatio n
straightforward implementation that ignores the
P
# Id
* X xx
K
# Id differences between the subtypes and
Q
o Y yy
A
# Id
P S (P )
pk Id
acknowledges what they have in common.
*
R
* Zzz
B
# Id
*
o
o
X xx
Y yy
Zzz
Mandatory attributes and relationships for the
fk 1 * A _id
L
• M a n d ato ry
d isc rim in ato r
fk 2 o
*
B _id
P _type
subtypes translate to optional columns and need
# Id
•
co lu m n
A d ditio n al
additional checks. Entities referring to a
co n s tra in ts
7 -2 0
subtype need an extra check to make sure the
reference is to a valid row.
.....................................................................................................................................................
Instructor Notes
.....................................................................................................................................................
Subtype or “two-table” implementation is also a

Sub type Im p lem e nta tion
straightforward implementation. This
Q S (Q )
P
# Id
* X xx
K
# Id pk *
*
Id
X xx
q_a_fk implementation mainly acknowledges the
Q
o Y yy
A
# Id
fk
o
*
Y yy
A _id differences. All attributes and relationships
R
* Zzz
B
# Id R S (R )
fk 1 = r_a_fk
translate into columns of the right optionality.
pk * Id
L
*
*
X xx
Zzz fk 2 = r_b_fk
An “incoming” relationship at supertype level
fk 1 * A _id
# Id fk 2 * B _id
needs extra attention as the single relationship
7 -2 1
must be implemented as two foreign keys with
an arc.
Both supertype and subtype implementation or
Sup erty pe and S ubtype (A rc) Im p lem e nta tion
P S (P )
“arc implementation” for short. Rather
P
# Id
* X xx
K
# Id
pk *
*
Id
X xx
fk 3 =
unknown but interesting implementation when
fk 1 ,uk 1 o Q _id
Q
o Y yy
A
# Id
fk 2 ,uk 2 o
fk 3 *
R _id
A _id
p_a_fk
the subtype details can often be used apart from
R
* Zzz
B
# Id fk 1 =
p_q_fk
fk 2 =
p_r_fk
the supertype information.
L
Q S (Q )
pk * Id
R S (R )
pk * Id r_b_fk
All attributes and relationships translate into
# Id o Y yy * Zzz
fk * B _id
columns of the right optionality. All
7 -2 2 “incoming” relationships map to appropriate
foreign keys. The major drawback is the fact
that you may need to join QS and RS with PS
too often.
An alternative implementation would be to
create foreign keys at the side of table QS and
RS both referring to PS. The arc would remain
in place at PS.
Supertype or “single table” implementation:
Sto ra ge Im plication
S upe rtyp e Im plem entatioin empty space in the R columns for Q rows and
discrim inato r colu m n
co ls cols cols
P
vice-versa. Additional type column.
P Q R
row s Q
Q Subtype or “two-table” implementation: no
R
empty space, no additional columns needed.
ro w s R
Arc implementation: no empty space,
additional foreign key columns.
7 -2 4
Sum m ary
• R e la tio n a l c o n ce p ts
• N a m in g ru les c on v e ntio n
• B a sic m a p p in g
• C o m p lex m a pp in g
7 -2 7
.....................................................................................................................................................
®
7-39
.....................................................................................................................................................

Practice
• M ap p in g S u p e rtyp e Practice 3day 4day

• Q ua lity C h ec k
S u bty p e Im p le m en ta tion Mapping Supertype Yes Yes
• Q ua lity C h ec k
•
S u pe rtyp e a n d S u b typ e (A rc) Im p le m en tatio n
M ap p in g P rim ary K e ys a n d C o lu m n s
Quality Check Subtype Impl. Yes Yes
Quality Check Arc Impl. Opt Yes
Synonyms Opt Cha
7 -2 8
Practice 7-1
Prac tice: M a pping Su pertype
repo rtin g to repo rt o f
If delegates want to use their own naming
D E PA R T M EN T
# Id
* N am e
HQ
* A d dress
repo rt
of COUNTRY
O R G A N IZ A T IO N
rep orting # Tax Id N um ber
convention it is fine, of course. The goal of this
* H ead C o un t
O T H ER D E PA R TM EN T
to
practice is to understand the mapping.
D E P A R TM E N TS ( )
7 -2 9
Practice 7-2
P ractice : Q uality C h eck
Sub type Im p lem e nta tion This practice should not take much time and
lpt_shop_ fk
can be done as a joint class activity.
G L O B A L _PR O D U C T S (G P T) LO C A L_PR O D U C TS (L PT )
pk * C od e pk # N am e
o Size fk Sh o p_no
*
* Pg p_n am e fk * Pg p _n am e
7 -3 1
.....................................................................................................................................................
Instructor Notes
.....................................................................................................................................................
Practice 7-3
P ractice : Q uality C h eck
A rc Im p le m e ntation This practice should not take much time and
PR OD U C T S (PD T )
pk * Code
fk 1 = pdt_pgp _nam e
fk 2 = pdt_gpt_ code
fk 1 P gp _nam e
*
fk 2 * G p t_co de fk 3 = pdt_lpt_n am e
fk 3 * L pt_na m e
G L OB A L_P R O D U C T S (G PT )
gpt_pgp_fk
pk * Code
o S ize
LO C A L_PR O D U C TS (L PT )
fk 1 = shp_lpt_fk
pk * N am e
pk, fk 1 o Sh p _n o fk 2 = pgp_lpt_ fk
fk 1 * Pg p _n am e
7 -3 2
Practice 7-4
Prac tice: M a pping Prim a ry K eys
and C olum ns This practice should not take much time and
G LO B A L_P R IC E S ( ) can be done as a joint class activity.
7 -3 3
.....................................................................................................................................................
®
7-41
.....................................................................................................................................................
.....................................................................................................................................................
.....................................................................................................................................................

Name: Full Email:

Lesson number: 7

Detail comments
Content Page
Type Number
.....................................................................................................................................................
®
7-43
.....................................................................................................................................................
.....................................................................................................................................................
Suggested Graphics
.....................................................................................................................................................
Suggested Graphics

.....................................................................................................................................................
®
7-45
.....................................................................................................................................................

520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.....................................................................................................................................................
8
.................................
Denormalized Data
.........................................................................................................................................
Introduction
Lesson aim
This lesson shows you the most common types of denormalization with examples.
Schedule
About the slide

See Page 31
• Denormalization
• Benefits
• Types of denormalization
8-2
Topic See Page

Why and When to Denormalize 4
Storing Derivable Values 6
Pre-Joining Tables 8
Hard-Coded Values 10
Keeping Details With Master 12
Repeating Single Detail with Master 14
Short-Circuit Keys 16
End Date Columns 18
Current Indicator Column 20
Hierarchy Level Indicator 22
Denormalization Summary 24
Practice 8—1: Name that Denormalization 25
Practice 8—3: Denormalize Price Lists 29
Practice 8—4: Global Naming 30
.............................................................................................................................................
Introduction
.........................................................................................................................................
Objectives

• Define denormalization and explain its benefits
• Differentiate and describe the different circumstances where denormalization is
appropriate
.........................................................................................................................................
®
8-3
.........................................................................................................................................
Why and When to Denormalize

Definition of Denormalization
Denormalization aids the process of systematically adding redundancy to the database
to improve performance after other possibilities, such as indexing, have failed. You
will read more on indexing in the lesson on Design Considerations.
Denormalization can improve certain types of data access dramatically, but there is no
success guaranteed and there is always a cost. The data model becomes less robust,
and it will always slow DML down. It complicates processing and introduces the
possibility of data integrity problems. It always requires additional programming to
maintain the denormalized data.
About the slide

Denormalization Overview See Page 31
Denormalization
• Starts with a “normalized” model
• Adds “redundancy” to the design
• Reduces the “integrity” of the design
• Application code added to compensate
8-3
Hints for Denormalizing

• Always create a conceptual data model that is completely normalized.
• Consider denormalization as the last option to boost performance.
• Never presume denormalization will be required.
• To meet performance objectives, denormalization should be done during the
database design.
• Once performance objectives have been met, do not implement any further
denormalization.
• Fully document all denormalization, stating what was done to the tables, what
application code was added to compensate for the denormalization, and the
reasons for and against doing it.
.............................................................................................................................................
Why and When to Denormalize
.........................................................................................................................................
Denormalization Techniques and Issues

In the next pages you see a number of denormalization techniques that are used
regularly. For every type of denormalization you see an indication of when it is
appropriate to use it and what the advantages and disadvantages are.
About the slide

Denormalization Techniques See Page 31
• Storing Derivable Values

• Pre-joining Tables
• Hard-Coded Values
• Keeping Details with Master
• Repeating Single Detail with Master
• Short-Circuit Keys
8-4
The following topics are covered:

• Storing Derivable Values
• Pre-joining Tables
and the most common specific examples:
• Derivable End Date Column
• Derivable Current Indicator column
• Hierarchy Level Indicator
.........................................................................................................................................
®
8-5
.........................................................................................................................................
Storing Derivable Values

When a calculation is frequently executed during queries, it can be worthwhile storing
the results of the calculation. If the calculation involves detail records, then store the
derived calculation in the master table. Make sure to write application code to re-
calculate the value, each time that DML is executed against the detail records.
In all situations of storing derivable values, make sure that the denormalized values
cannot be directly updated. They should always be recalculated by the system.
About the slide

Storing Derivable Values See Page 32
Before
A B
pk * Id pk,fk * A_id
* X pk * Sequence_No
* Quanity
Add a column to store derivable data in the

“referenced” end of the foreign key.
A
After pk * Id
* X
* Total_quantity
8-5
Appropriate:
• When the source values are in multiple records or tables
• When derivable values are frequently needed and when the source values are not
• When the source values are infrequently changed
Advantages:
• Source values do not need to be looked up every time the derivable value is
required
• The calculation does not need to be performed during a query or report
Disadvantages:
• DML against the source data will require recalculation or adjustment of the
derivable data
• Data duplication introduces the possibility of data inconsistencies
.............................................................................................................................................
Storing Derivable Values
.........................................................................................................................................
E-mail Example of Storing Derivable Values
EMail Example of Storing Derivable Values

Before
REC_MESSAGES (RME) MESSAGES (MSE)

USERS (USR)
pk,fk * Usr_Id pk * Id
pk * Id Subject
pk,fk * Mse_Id *
* Per_name * Text
Store derivable column in the ‘referenced’ end of the foreign key.
MESSAGES (MSE)
After pk * Id
* Subject
* Text
* Number_of_times_received
8-6
When a message is delivered to a recipient, the user only receives a pointer to that
message, which is recorded in RECEIVED_MESSAGES. The reason for this, of
course, is to prevent the mail system from storing a hundred copies of the same
message when one message is sent to a hundred recipients.
Then, when someone deletes a message from their account, only the entry in the
RECEIVED_MESSAGES table is removed. Only after all RECEIVED_MESSAGE
entries, for a specific message, have been deleted, the should the actual message be
deleted too.
We could consider adding a denormalized column to the MESSAGES table to keep
track of the total number of RECEIVED_MESSAGES that are still kept for a
particular message. Then each time users delete a row in RECEIVED_MESSAGES,
in other words, they delete a pointer to the message, the Number_of_times_received
column can be decremented. When the value of the denormalized column equals zero,
then we know the message can also be deleted from the MESSAGES table.
.........................................................................................................................................
®
8-7
.........................................................................................................................................
Pre-Joining Tables
You can pre-join tables by including a nonkey column in a table, when the actual value
of the primary key, and consequentially the foreign key, has no business meaning. By
including a nonkey column that has business meaning, you can avoid joining tables,
thus speeding up specific queries.
You must include application code that updates the denormalized column, each time
the “master” column value changes in the referenced record.
About the slide

Pre-Joining Tables See Page 32
Before
B
A
pk pk * Id
* Id
fk * A_id
* Col_a
Add the non_key column to the table with the foreign key.
B
After
pk * Id
fk * A_id
* A_col_a
8-7
Appropriate:
• When frequent queries against many tables are required
• When slightly stale data is acceptable
Advantages
• Time-consuming joins can be avoided
• Updates may be postponed when stale data is acceptable
Disadvantages
• Extra DML needed to update original nondenormalized column
• Extra column and possibly larger indices require more working space and disk
space
.............................................................................................................................................
Pre-Joining Tables
.........................................................................................................................................
EMail Example of Pre-Joining Tables

Before
FOLDERS (FDR) RECEIVED_MESSAGES (RME)
* pk,fk * Mse_id
pk Id
* pk,fk * Flr_id
Name
* Date_received
Create a table with all the frequently queried columns.
RECEIVED_MESSAGES (RME)
After pk,fk * Mse_id
pk,fk * Flr_id
* Date_received
* Fdr_Name
8-8
Example
Suppose users often need to query RECEIVED_MESSAGES, using the name of the
folder where the received message is filed. In this case it saves time when the name of
the folder is available in the RECEIVED_MESSAGES table.
Now, if a user needs to find all messages in a particular folder, only a query on
RECEIVED_MESSAGES is needed.
Clearly, the disadvantage is extra storage space for the extra column in a, potentially,
very large table.
.........................................................................................................................................
®
8-9
.........................................................................................................................................
Hard-Coded Values
If a reference table contains records that remain constant, then you can consider hard-
coding those values into the application code. This will mean that you will not need to
join tables to retrieve the list of reference values. This is a special type of
denormalization, when values are kept outside a table in the database. In the example,
you should consider creating a check constraint to the B table in the database that will
validate values against the allowable reference values. Note that a check constraint,
though it resides in the database, is still a form of hardcoding.
Whenever a new value of A is needed the constraint must be rewritten.
About the slide

Hard-Coded Values See Page 32
Before
B
A
pk Id pk * Id
*
Type fk * A_id
*
Remove the foreign key and hard code the allowable values and
validation in the application.
B
After pk Id
*
* A_Type
8-9
Appropriate
• When the set of allowable values can reasonably be considered to be static during
the life cycle of the system
• When the set of possible values is small, say, less than 30
Advantages
• Avoids implementing a look-up table
• Avoids joins to a look-up table
Disadvantages
• Changing look-up values requires recoding and retesting
.............................................................................................................................................
Hard-Coded Values
.........................................................................................................................................
Email Example of Hard-Coded Values

Before
USERS (USR)
BUSINESS_TYPES (BTE)
pk * Id
pk * Id fk * Bte_id
Name * Per_name
Hard code the allowable values and validation in the

application.
After USERS (USR)

pk * Id
* Business_type
* Per_name
8-10
Example
ElectronicMail would like to know some background information about their users,
such as the type of business they work in. Therefore EM have created a table to store
all the valid BUSINESS_TYPES they want to distinguish. The values in this table are
set up front and not likely to change.
This is a candidate for hard-coding the allowable values. You could consider placing a
check constraint on the column in the database. In addition to that, or instead of that,
you could build the check into the field validation for the screen application where
users can sign in to the EM service.
.........................................................................................................................................
®
8-11
.........................................................................................................................................
Keeping Details With Master

In a situation where the number of detail records per master is a fixed value (or has a
fixed maximum) and where usually all detail records are queried with the master, you
may consider adding the detail columns to the master table. This denormalization
works best when the number of records in the detail table are small. This way you will
reduce the number of joins during queries. An example is a planning system where
there is one record per person per day. This could be replaced by one record per person
per month, the table containing a column for each day of the month.
About the slide

Keeping Details with Master See Page 32
Before
B
A
pk,fk * A_id
pk * Id pk * Type
* Amount
Add the repeating detail columns to the master table.
After pk * Id
* Type1
* Amount_1
* Type2
* Amount_2
* Type3
* Amount_3
8-11
Appropriate
• When the number of detail records for all masters is fixed and static
• When the number of detail records multiplied by the number of columns of the
detail is small, say less than 30
Advantages
• No joins are required
• Saves space, as keys are not propagated
Disadvantages
• Increases complexity of data manipulation language (DML) and SELECTs across
detail values
• Checks for Amount column must be repeated for Amount1, Amount2 and so on
• Table name A might no longer match the actual content of the table
.............................................................................................................................................
Keeping Details With Master
.........................................................................................................................................
EMail Example Keeping Detail with Master

Before
STORAGE_QUOTAS (SQA)
USERS (USR)
pk,fk * Usr_Id
pk * Id pk * Storage_type
* Name * Allocated
* Available
Add the repeating detail columns to the master table.
USERS (USR)
After pk * Id
* Name
* Message_Quota_Allocated
* Message_Quota_Available
* File_Quota_Allocated
* File_Quota_Available
8-12
Example
Suppose each e-mail user is assigned two quotas—one for messages and one for files.
The amount of each quota is different, so both have to be tracked individually. The
quota does not change very frequently. To be relationally pure, we would create a two-
record STORAGE_TYPES table and a STORAGE_QUOTAS table with records for
each user, one for each quota type. Instead, we can create the following denormalized
columns in the USER table:
• Message_Quota_Allocated
• Message_Quota_Available
• File_Quota_Allocated
• File_Quota_Available
Note that the name of table USERS does not really match the data in the denormalized
table.
.........................................................................................................................................
®
8-13
.........................................................................................................................................
Repeating Single Detail with Master

Often when the storage of historical data is necessary, many queries require only the
most current record. You can add a new foreign key column to store this single detail
with its master. Make sure you add code to change the denormalized column any time
a new record is added to the history table.
About the slide

Repeating Current Detail with Master See Page 33
Before
B
A
pk,fk * A_Id
pk * Id pk * Start_date
* Price
Add a column to the master to store the most current details.
A
After
pk * Id
* Current_price
8-13
Appropriate
• When detail records per master have a property such that one record can be
considered “current” and others “historical”
• When queries frequently need this specific single detail, and only occasionally
need the other details
• When the Master often has only one single detail record
Advantages
• No join is required for queries that only need the specific single detail
Disadvantages
• Detail value must be repeated, with the possibility of data inconsistencies
Additional code must be written to maintain the duplicated single detail value at the
master record.
.............................................................................................................................................
Repeating Single Detail with Master
.........................................................................................................................................
EMail Example of Repeating Single Detail

with Master
Before
MESSAGES (MSE) ATTACHMENTS (ATT)
pk * Id pk * Id
* Subject pk,fk * Mse_id
* Text * Name
Add a column to the master to store the most current details.
MESSAGES (MSE)
After
pk * Id
* First_attachment_name
* Subject
* Text
8-14
Example
Any time a message is sent, it can be sent with attachments included. Messages can
have more than one attachment. Suppose in the majority of the messages that there is
no or only one attachment. To avoid a table join, you could store the attachment name
in the MESSAGES table. For those messages containing more than one attachment,
only the first attachment would be taken. The remaining attachments would be in the
ATTACHMENTS table.
.........................................................................................................................................
®
8-15
.........................................................................................................................................
Short-Circuit Keys
For database designs that contain three (or more) levels of master detail, and there is a
need to query the lowest and highest level records only, consider creating short-circuit
keys. These new foreign key definitions directly link the lowest level detail records to
higher level grandparent records. The result can produce fewer table joins when
queries execute.
About the slide

Short-Circuit Keys See Page 33
Before
B C
A
pk * Id pk * Id
pk * Id fk A_id fk * B_id
*
Create a new foreign key from the lowest detail to the

highest master.
A B C
After pk Id
pk * Id * pk * Id
fk * A_id fk * B_id
fk * A_id
8-15
Appropriate
• When queries frequently require values from a grandparent and grandchild, but not
from the parent
Advantages
• Queries join fewer tables together
Disadvantages
• Extra foreign keys are required
• Extra code is required to make sure that the value of the denormalized column
A_id is consistent with the value you would find after a join with table B.
.............................................................................................................................................
Short-Circuit Keys
.........................................................................................................................................
EMail Example of Short-Circuit Keys

Before
RECEIVED_
USERS (USR) FOLDERS (FDR) MESSAGES (RME)
pk * Id pk * Name pk * Id
* Name fk * Usr_id fk * Fdr_name
Create a new foreign key from the lowest detail to the highest
master.
After
RECEIVED_
FOLDERS (FDR) MESSAGES (RME)
USERS (USR)
pk * Name pk * Id
pk * Id fk * Usr_id fk * Fdr_name
uk * Name fk * Usr_name
8-16
Example
Suppose frequent queries are submitted that require data from the
RECEIVED_MESSAGES table and the USERS table, but not from the FOLDERS
table. To avoid having to join USERS and FOLDERS, the primary or a unique key of
the USERS table can been migrated to the RECEIVED_MESSAGES table, to provide
information about USERS and RECEIVED_MESSAGES with one less, or no, table
join.
.........................................................................................................................................
®
8-17
.........................................................................................................................................
End Date Columns

The most common denormalization decision is to store the end date for periods that
are consecutive; then the end date for a period can be derived from the start date of the
previous period.
If you do this, to find a detail record for a particular date you avoid the need to use a
complex subquery.
About the slide

End Date Column See Page 33
Before
B
A pk,fk * A_id
Add an end date column to speed up queries so that they can

use a between operator.
B
After pk,fk * A_Id
pk * Start_date
* End_date
8-17
Appropriate
• When queries are needed from tables with long lists or records that are historical
and you are interested in the most current record
Advantages
• Can use the between operator for date selection queries instead of potentially time-
consuming synchronized subquery
Disadvantages
• Extra code needed to populate the end date column with the value found in the
previous start date record
.............................................................................................................................................
End Date Columns
.........................................................................................................................................
Example of End Date Column

Before
PRICES (PCE)
PRODUCTS (PDT)
pk,fk * Pdt_id
pk * Id pk Start_date
*
* Name Price
*
Create an extra column derivable End_date column.
After PRICES (PCE)

pk,fk * Pdt_id
pk * Start_date
* Price
o End_date
8-18
Example
When a business wishes to track the price history of a product, they may use a PRICES
table that contains columns for the price and its start date and a foreign key to the
PRODUCTS table. To avoid using a subquery when looking for the price on a specific
date, you could consider adding an end date column. You should then write some
application code to update the end date each time a new price is inserted.
Compare:
...WHERE pdt_id = ...
AND start_date = ( SELECT max(start_date)
FROM prices
WHERE start_date <= sysdate
AND pdt_id = ...
)
and
AND sysdate between start_date and nvl(end_date, sysdate)
Note that the first table structure presupposes that products always have a price since
the first price start date of that product. This may very well be desirable but not always
the case in many business situations.
Note also that you would need code to make sure periods do not overlap.
.........................................................................................................................................
®
8-19
.........................................................................................................................................
Current Indicator Column

This type of denormalization can be used in similar situations to the end date column
technique. It can even be used in addition to an end date. It is a very common type of
denormalization.
Suppose most of the queries are to find the most current detail record. With this type of
requirement, you could consider adding a new column to the details table to represent
the currently active record.
You would need to add code to update that column each time you insert a new record.
About the slide

Current Indicator Column See Page 33
Before
A B
pk,fk * A_id
Add a column to represent the most current record in a

long list of records .
After B
pk,fk * A_Id
pk * Start_date
o Current_indicator
8-19
Appropriate
• When the situation requires retrieving the most current record from a long list
Advantages
• Less complicated queries or subqueries
Disadvantages
• Extra column and application code to maintain it
• The concept of “current” makes it impossible to make data adjustments ahead of
time
.............................................................................................................................................
Current Indicator Column
.........................................................................................................................................
Example of Current Indicator Column

Before
PRICES (PCE)
PRODUCT (PDT)
pk,fk * Pdt_id
pk * Id pk Start_date
*
* Name Price
*
Add a column to represent the most current record, in a long

list of records.
PRICES (PCE)
After pk,fk * Pdt_id
pk * Start_date
* Price
o Current_indicator
8-20
Example
In the first table structure, when the current price of a product is needed, you need to
query the PRICES table using:
AND start_date = ( SELECT max(start_date)
FROM prices
WHERE start_date <= sysdate
AND pdt_id = ...
)
The query in the second situation would simply be:

AND current_indicator = ’Y’
.........................................................................................................................................
®
8-21
.........................................................................................................................................
Hierarchy Level Indicator

Suppose there is a business limit to the number of levels a particular hierarchy may
contain. Or suppose in many situations you need to know records that have the same
level in a hierarchy. In both these situations, you will need to use a connect-by clause
to traverse the hierarchy. This type of clause can be costly on performance. You could
add a column to represent the level of a record in the hierarchy, and then just use that
value instead of the connect-by clause in SQL.
About the slide

Hierarchy Level Indicator See Page 33
Before
A
pk * Id
fk * A_id
Create a column to represent the hierarchy level of a record.
After A
pk * Id
fk * A_id
* Level_no
8-21
Appropriate
• When there are limits to the number of levels within a hierarchy, and you do not
want to use a connect-by search to see if the limit has been reached
• When you want to find records located at the same level in the hierarchy
• When the level value is often used for particular business reasons
Advantages
• No need to use the connect-by clause in query code
Disadvantages
• Each time a foreign key is updated, the level indicator needs to be recalculated,
and you may need to cascade the changes
.............................................................................................................................................
Hierarchy Level Indicator
.........................................................................................................................................
Example of Hierarchy Level Indicator

Before
FOLDERS (FDR)
pk * Id
fk * Fdr_id
* Name
Create a column to represent the hierarchy level of a record.
After FOLDERS (FDR)

pk * Id
fk * Fdr_id
* Name
* Level_no
8-22
Example
Imagine that because of storage limitations, a limit has been placed on the number of
nested folders. Each time a user wants to create a new instance of a folder within an
existing folder instance, code must decide if that limit has been reached. This can be a
slow process.
If you add a column to indicate at what nested level a FOLDER is, then when you
create a new folder in it, you can decide immediately if this is allowed. If it is, the level
of the new folder is simply one more than the level of the folder it resides in.
.........................................................................................................................................
®
8-23
.........................................................................................................................................
Denormalization Summary
Denormalization is a structured process and should not be done lightly. Every
denormalization step will require additional application code. Be confident you do
want to introduce this redundant data.
Denormalization Summary
Denormalization Techniques
• Storing Derivable Information
– End Date Column
– Current Indicator
– Hierarchy Level Indicator
• Pre-Joining Tables
• Keeping Detail with Master
8-23
.............................................................................................................................................
Practice 8—1: Name that Denormalization
.........................................................................................................................................
Practice 8—1: Name that Denormalization

Moonlight Coffees
Goal (See Page 34)
Learn to discriminate the type of denormalization depicted.
Your Assignment
For the following table diagrams, decide what type of
denormalization is used and explain why the diagram depicts the denormalization you
have listed.
Use one of:
• Storing derivable information
1
SHIFTS (SFT)
WEEKDAYS (WDY)
pk * No
pk * Code fk * Wdy_code
* Name Start_time
*
* End_time
* Wdy_name
PROD_GRPS (PGP) PRODUCTS (PDT) PROD_NAMES (PNE)
pk * Name pk * Code pk * Name

fk * Pgp_Name fk * Pdt_code
fk * Pgp_name
PRICE_LISTS (PLT)
COUNTRIES (CTY)
pk,fk * Cty_code
pk * Code pk * Start_date
* Name o End_date
* Current_price_ind
.........................................................................................................................................
®
8-25
.........................................................................................................................................
Practice 8—2: Triggers

Goal (See Page 35)
The purpose of this practice is to investigate which database triggers are needed to
handle a suggested denormalization.
Your Assignment
1 Indicate which triggers are needed and what they should do to handle the
denormalized column Order_total of ORDER_HEADERS.
ORDER_HEADERS (OHR) ORDER_ITEMS (OIM)

pk * Ohr_id
pk * Id
pk * Seqno
* Order_total * Item_total
Table Trg Type Column Needed? What should it do?

OHR Insert
Delete
Update Id
Order_total
OIM Insert
Delete
Update Ohr_id
Item_total
8-29
.............................................................................................................................................
Practice 8—2: Triggers
.........................................................................................................................................
denormalized column Lcn_address of EMPLOYEES.
LOCATIONS (LCN) EMPLOYEES (EPE)

pk * Id
pk * Id
fk * Lcn_id
* Name
* Address
* Lcn_address

LCN Insert
Delete
Update Address
other cols
EPE Insert
Delete
Update Lcn_id
Lcn_address
8-31
.........................................................................................................................................
®
8-27
.........................................................................................................................................
denormalized column Curr_price_ind of table PRICES.
PRODUCTS (PDT) PRICES (PCE)

pk * Pdt_id
* Name o End_date
* Curr_price_ind

PDT Insert
Delete
PCE Insert
Delete
Update Pdt_id
Start_date
End_date
Curr_price_Ind
8-33
.............................................................................................................................................
Practice 8—3: Denormalize Price Lists
.........................................................................................................................................
Practice 8—3: Denormalize Price Lists

Moonlight Coffees
Goal (See Page 36)
The aim of this practice is to decide on the type of
denormalization you could use, and what code is needed to ensure
database integrity.
Scenario
End users have started to complain about query performance. One of the areas where
this is particularly noticeable is when querying the price of a global product. Since
there is a large list of records in the GLOBAL_PRICES table, and it needs to be joined
with the PRICE_LISTS table, it is not surprising the queries can take a long time.
Optimizing the queries using other techniques have failed to result in acceptable
response times.Therefore the decision is to use some denormalization to correct this
problem.
The corporate office also has another concern. They would like to notify the local
shops of any new price list changes of global products, prior to their effective date.
They would like to enter the new price list information when it is decided, not when
the start date is reached. You need to add provision to alleviate this restriction.
Your Assignment
Describe what type of denormalization you would implement and what code you
would add to ensure the database does not lose any integrity. The next diagram shows
the current table schema. Consider both issues described above when deciding which
types of denormalization to implement.
PRICE_LISTS (PLT) GLOBAL_PRICES (GPE)
pk * Start_date pk,fk * Plt_start_date

pk,fk * Cty_code pk,fk * Plt_cty_code
* Amount
.........................................................................................................................................
®
8-29
.........................................................................................................................................
Practice 8—4: Global Naming

Moonlight Coffees
Goal (See Page 36)
To convert user requirements into denormalized table designs
Scenario
The corporate office has decided to formalize English as the
corporate language. Headquarters has asked the IS department to arrange for all global
products to store their names in English. On the other hand, countries must be able to
store their native language equivalent.
Your Assignment
Using the design below, denormalize the table design and describe the additional code
that will allow this requirement to be implemented.
LANGUAGES (LGE)
pk * Code
* Name
PRODUCT_NAMES (PNE)
PRODUCTS (PDT)
pk * Code pk,fk * Pdt_code
o Size pk,fk * Lge_code
* Name
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Instructor Notes
Instructor Note
Topic Timings
88
Lecture 45 minutes
D enorm alize d D ata
Total 60-90 minutes
Set the scene. Define denormalization and

O verview
explain the benefits. The majority of this lesson
• D e no rm aliza tio n covers examples of denormalization.
• B e ne fits
• T y p es o f d en o rm alizatio n
The presentation makes use of a generic table
scheme. The matching slide of the e-mail
system is only printed. You could consider
reversing this.
8 -2
Emphasize you need to start with normalized

D en orm alization O verview
data. As you denormalize the schema, you add
De n o rm a liza tio n
redundancy and will need to write code to
•
•
S tarts w ith a “n o rm a liz ed ” m od e l
A d d s “re d u n d an c y” to th e d es ig n
support the integrity of the database. The
•
•
Re d u ce s th e “in te g rity” of th e d es ig n
A p p lica tio n c od e a d d ed to c o m p en s ate
process of denormalization is ongoing, and you
should consider adding or removing it,
depending on the database performance.
8 -3
Just name the six types of denormalization, and

D en orm alization Tec hnique s
explain there may be more, but these are the
• S to rin g D eriva b le V a lu es only ones we will cover in this course.
• P re-jo inin g T ab les
• H a rd -Co d e d V a lu e s Remember, any time you add redundancy to the
• K e ep in g D e tails w ith M as te r
• R e pe a ting S in g le D e ta il w ith M as ter schema, you denormalize, and will have to
• S h ort-C irc u it K ey s
write (and maintain) code.
8 -4
.........................................................................................................................................
®
8-31
.........................................................................................................................................
In all cases of denormalized data, you need

Storing D eriva ble V alues
B e fore
code to prevent direct updates of the
A
pk * Id
B
pk,fk * A _id
denormalized values. These should always be
* X pk * S equence_ N o
* Q uanity
recalculated by the system.
A dd a colum n to store de rivable data in the
“refere nced” end of the foreign key.
A
A fter pk * Id
*
*
X
Total_quantity You will need to add code to B to make sure
that any time records in B are inserted, updated,
8 -5
or deleted, the total_quantity is recalculated for
every row in A that is referred to in an affected
row of B.
You will need to add code to the A table so that
Pre-Joining Ta ble s
B e fore
any time Col_a is changed all those
B
A
pk * Id pk
fk
* Id
corresponding denormalized values in the
* C ol_ a * A _id
A_col_a column of table B are changed too.
A d d the non_key colum n to the table w ith the foreign key.
B
A fter
pk * Id
fk * A _id
* A _col_a
8 -7
There must be code to check the values in the

H ard-C od ed Value s
A_type column against the allowable values,
B efore
A
pk Id
B
pk * Id
regardless of how records are entered into the
*
* Type fk * A _ id
database.
R em ove the fo reign key and hard code the allow able valu es and
validation in the application.
B
A fter pk Id
*
* A _Type
8 -9
This is considered to be denormalized although

K eeping D eta ils w ith M as te r
B e fore
there is no derivable information involved here.
B
A
pk * Id
pk,fk
pk
*
*
A _id
Type
Code to check a condition for Amount1 must be
*
A dd the repeating detail colum ns to th e m aster table.

A m ount
repeated for Amount2 and so on.
A
A fter pk * Id
* Type1
* A m ount_1
* Type2
* A m ount_2
* Type3
* A m ount_3
8 -1 1
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Code in B is required which fires when a record

R epe ating C urrent D etail w ith M aster
B e fore
in B is updated or a new record is inserted in B.
B
A
pk * Id
pk,fk
pk
*
*
A _Id
S tart_date
The code should find out if the new value has
* P rice
A dd a colum n to the m aster to store th e m ost current details.

consequences for a Current_price in A and
A
update column Current_price if so.
A fter
p k * Id
* C urrent_price
8 -1 3
You need to have code on insert and update of

Sh ort-C irc uit K ey s
B e fore
the B table so that when the value of A_id
A B
pk * Id
C
pk * Id
changes, the column A_id in table C is updated
pk * Id fk * B _id
fk * A _id
correspondingly.
C reate a new foreign key from the low est detail to the
highest m aster.
A B C
A fter pk Id
pk * Id * pk * Id
fk * A _id fk * B _id
fk * A _id
8 -1 5
Add code to fire on the insert of the B table to

End D ate C o lum n
B e fore
take the value of the new Start_date record and
B
A
pk Id
pk,fk
pk
*
*
A _id
S tart_date
populate the preceding End_date column.
*
A dd an e nd date colum n to spee d up queries so that they can

Note that the current B record for a record of A
use a betw een operator.
B
is not necessarily the record where B.End_date
A fter A _Id
pk,fk
pk
*
*
*
S tart_date
E nd_date
is NULL.
8 -1 7
You will need code to fire on insert or update of

C u rre nt Indicator C o lum n
B e fore
B, to check if updates of the Current_indicator
A B
pk,fk * A _id
column are necessary and if so, set the new
pk * Id pk * S tart_date
value.
A dd a co lum n to represent the m ost current record in a
long list o f records .
Note that usually two updates are needed: one
A fter B
pk,fk
pk
*
*
A _Id
S tart_d ate
to reset the record with Current _indicator = ‘Y’
o C urren t_indicator
(the old current) and one to set the indicator of
8 -1 9 the new current record.
The code should fire on insert and delete of the
H ie ra rchy Lev el Indicator
B e fore
A table, and when the A.A_id column is
A
pk * Id
updated. Level_no must be recalculated for the
fk * A _id
records of A which are affected.
C reate a colum n to represent the hierarchy level of a reco rd.
A fter A
pk * Id
fk * A _id
* Level_no
8 -2 1
.........................................................................................................................................
®
8-33
.........................................................................................................................................
D en orm alization Su m m a ry
D e no rm aliza tio n Te c h niq u es

• S to ring D eriva ble In fo rm atio n
– E nd D ate C o lum n
– C u rren t In dicator
– H ie ra rch y Leve l In dicato r
• P re-Jo in in g T ab les
• H a rd -C o d e d V a lu e s
• K e ep in g D e tail w ith M a ster
• R e pe ating S in g le D e ta il w ith M as ter
• S h ort-C irc u it K ey s
8 -2 3

Prac tices
• N a m e th a t D en o rm a liza tio n Practice 3day 4day

• T rigg ers
• D e no rm alize P rice L is ts Name that denormalization Yes Yes
• G lo b al N am ing
Triggers Yes Yes
Denormalize Price Lists opt opt
Global Naming opt cha
8 -2 4
Practice 8-1
Practice : N a m e T hat D en orm alization (1/3 )
1 This practice should not take much time and
S H IFTS (S FT)
W E E K D A Y S (W D Y )
pk * No
pk * C ode fk W dy_co de
*
* N am e S tart_tim e
*
* E nd_tim e
* W dy_n am e
8 -2 5
Practice 8-1
P R O D _G R P S (P G P ) P R O D U C TS (P D T) P R O D _N A M E S (P N E )
pk * N am e pk * C ode pk * N am e
fk * P gp_N a m e fk * P dt_code
fk * P gp_nam e
8 -2 6
.............................................................................................................................................
Instructor Notes
.........................................................................................................................................
Practice 8-1
P R IC E _LIS TS (P LT)
C O U N TR IE S (C TY )
pk,fk * C ty_code
pk * C od e pk * S tart_date
* N am e o E nd_date
* C urrent_price_in d
8 -2 7
Practice 8-2
Practice: Trigge rs (1/6)
1 It is important that the students are complete
O R D E R _H E A D E R S (O H R ) O R D E R _ITE M S (O IM )
pk * Id
pk
pk
* O hr_id
* S eqno
in the list of needed triggers and that they
* O rder_total * Item _total
are confident about that.
That takes away the fear of denormalizing
from some and the nonchalance from
others.
8 -2 8
Practice 8-2
2
LO C A TIO N S (LC N ) E M P LO Y E E S (E P E )
pk * Id
pk * Id
fk * Lcn_id
* N am e
* A d dress
* Lcn_address
8 -3 0
Practice 8-2
3
P R O D U C TS (P D T) P R IC E S (P C E )
pk * P dt_id
pk * Id pk * S tart_date
* Name o E nd_date
* C urr_price_ind
8 -3 2
.........................................................................................................................................
®
8-35
.........................................................................................................................................
The triggers alone are not enough. As time goes

by a new price may become current. The
T able T rg Ty pe C o lum n N ee ded ? W h at sho uld it do ? system needs a time-related mechanism to set
PDT In sert
D elete the indicator, not just a mechanism that depends
PCE In sert
D elete on DML operations.
U pd ate P d t_id
S tart_da te
E n d_d ate
C u rr_p rice_In d
8 -3 3
Practice 8-3
Prac tice: D en orm alize Price Lists
If there is time, discuss the type of code needed
P R IC E _ LIS TS (P LT)
pk * S tart_date
G LO B A L_P R IC E S (G P E )
pk,fk * P lt_start_date to set the denormalized values.
pk,fk * C ty_ code pk,fk * P lt_cty_code
* A m ount
S p ee d u p p erfo rm an c e fo r q u e rie s o n A m o u n t.
Ins e rt n ew pric e lists b efo re the ir effec tive d ate .
8 -3 4
Practice 8-4
P ractice : G lo bal N am in g
If there time allows, discuss the type of code
LA N G U A G E S (LG E )
pk * C ode
* N am e
needed to set the denormalized value.
P R O D U C T_N A M E S (P N E )
P R O D U C TS (P D T)
pk * C ode pk,fk * P dt_cod e
o S ize pk,fk * Lge_co de
* N am e
8 -3 5
.............................................................................................................................................
.........................................................................................................................................

Name: Full Email:

Lesson number: 8

Detail comments
Content Page
Type Number
.........................................................................................................................................
®
8-37
.........................................................................................................................................
.............................................................................................................................................
Suggested Graphics
.........................................................................................................................................
Suggested Graphics

.........................................................................................................................................
®
8-39
.........................................................................................................................................

520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.............................................................................................................................................
9
.................................
Database Design
Considerations
.....................................................................................................................................................
Introduction
Lesson Aim
This lesson illustrates some principles of the Oracle RDBMS and presents the various
techniques that can be used to refine the physical design.
Schedule
See page 36
Overview
About the slide
See page 36
• Oracle specific Design Considerations
• Data Integrity Issues
• Performance Considerations
• Storage Issues
9-2
Topic See Page

Introduction See page 2
Reconsidering the Database Design See page 4
Oracle Data Types See page 5
Most Commonly-Used Oracle Data Types See page 6
Column Sequence See page 7
Primary Keys and Unique Keys See page 8
Artificial Keys See page 11
Sequences See page 13
Indexes See page 16
Choosing Columns to Index See page 19
When Are Indexes Used? See page 21
.....................................................................................................................................................
Introduction
.....................................................................................................................................................
Topic See Page

Views See page 23
Use of Views See page 24
Old-Fashioned Design See page 25
Distributed Design See page 27
Benefits of Distributed Design See page 28
Oracle Database Structure See page 29
Summary See page 31
Practice 9—1: Data Types See page 32
Practice 9—2: Artificial Keys See page 34
Practice 9—3: Product Pictures See page 35
Objectives
• Describe which data types to use for columns
• Evaluate the quality of the Primary key
• Use artificial keys and sequences where appropriate
• Define rules for referential integrity
• Explain the use of indexes
• Discuss partitioning and views
• Recognize old-fashioned database techniques
• Explain the principle of distributed databases
• Describe the Oracle database model
.....................................................................................................................................................
®
9-3
.....................................................................................................................................................
Reconsidering the Database Design

Each RDBMS has its own internal mechanism. This lesson discusses the major
features provided by Oracle to get the best RDBMS performance.
About the slide
See page 36
Why Adapt Data Design?
• User Expectations
• Volumes
Adapted
• Hardware
Initial design Physical
• Network
Design
• O.S.
• Oracle specifics
9-3
You have to analyze a large number of parameters to obtain a correct adapted physical
design from the initial design. Note the “a correct”, not “the correct”. Like many
design issues, there is no absolute truth here.
The points noted here are the most important ones—there are others.
• The expected volume of tables, the hardware characteristics like CPU speed,
memory size, number of disks and corresponding space, the architecture—client/
server or three tier, the network bandwidth, speed, and the operating systems are
determinants.
• User requirements are an other big issue. Depending on the response time, the GUI
and the frequency of use of modules, they influence the objects that can be used in
Oracle to cope with user expectations.
• Depending on the version of Oracle you are using, some elements may or may not
exist.
.....................................................................................................................................................
Oracle Data Types
.....................................................................................................................................................
Oracle Data Types

About the slide
See page 36
Oracle Data Types
• Depending on:
– Domains
– Storage issue
– Performance
– Use
• Select a data type for columns:
– Character
– Number
– Date
– Large Objects
9-4
When you create a table or cluster, you must specify an internal data type for each of
its columns. These data types define a generic domain of values that each column can
contain.
• Some data types have a narrow focus, like number and date. Some data types are
general purpose data types, like the various character data types.
• Some data types allow for variable length, some do not.
Choosing a large fixed length for a column to store very few bytes for most of the
rows can result in a huge table size. This may affect performance as a row may
actually contain only a few bytes and yet be stored on multiple blocks, resulting in
a great number of I/O’s, and therefore decreasing performance.
• One cannot search against the Large Object Data Types; they cannot be used in a
where clause. They are only retrievable by searching against other columns.
.....................................................................................................................................................
®
9-5
.....................................................................................................................................................
Most Commonly-Used Oracle Data Types

• CHAR(size) These are fixed-length character data of length-sized bytes.
Maximum size is 2000 bytes.
Typical use: for official International Currency Codes which are a fixed three
characters in length such as USD, FFR.
• VARCHAR2(size) Variable-length character string having maximum length-sized
bytes. Maximum size is 4000, and minimum is 1. This is the most commonly-used
data type and you should use it if you are not sure which one to use. It replaces the
old Oracle version 6 CHAR data type.
Typical use: for storing individual ASCII text lines of unlimited length ASCII
texts on which you need to be able to search using a wildcard.
• NUMBER This data type is used for numerical values, with or without a decimal,
of virtually unlimited size. Use this data type for data on which calculation or
sorting should be possible. Avoid its use for numbers like a phone number, where
the value does not have any meaning.
Typical use: amount of money, quantities, generated unique key values.
• DATE Valid date range from January 1, 4712 BC to December 31, 4712 AD. A
date data type also contains time components. You should use it only when you
know the full date including day, month, and year. The time component is often set
to 00:00 (midnight) in normal use of dates.
Typical use: any date where the full date is known.
• LONG Character data of variable length up to 2 gigabytes. Obsolete since
Oracle8. Was used for ASCII text files where you do not need to search using the
wildcard or substring functionality. Use CLOB data type instead.
Typical use: for storing the source code of HTML pages.
• LONG RAW Raw binary data of variable length up to 2 gigabytes. Obsolete since
Oracle8. Was used for large object types where the database should not try to
interpret the data. Use BLOB data type instead.
Typical use: images or video clips.
• CLOB Character large object type. Replaces LONG. Major difference: a table can
have more than one CLOB column where there was only one LONG allowed.
Maximum size is 4 gigabytes.
Typical use: see LONG.
• BLOB Character large object type. Replaces LONG RAW. Major difference: a
table can have more than one BLOB column where there was only one
LONGRAW allowed. Maximum size is 4 gigabytes.
Typical use: see LONG RAW.
• BFILE Contains a locator to a large binary file stored outside the database to
enable byte stream I/O access to external LOBs residing on the database server.
Typical use: movies
.....................................................................................................................................................
Column Sequence
.....................................................................................................................................................
Column Sequence
The sequence of columns in a table is relevant, although any column sequence would
allow all table operations. The column sequence can influence, in particular, the
performance of data manipulation operations. It may also influence the size of a table.
About the slide
See page 37
Suggested Column Sequence
• Primary key columns

• Unique Key columns
• Foreign key columns
• Mandatory columns
• Optional columns
Large object columns always at the end
9-5
The suggested optimal column sequence is the following:

1 Primary key columns
2 Unique key columns
3 Foreign key columns
4 Remaining mandatory columns *
5 Remaining optional columns *
*
In cases where the table contains a LONG or LONG RAW column, even if it is a
mandatory column, make it the last column of the table.
The rationale is that null columns should be at the end of the table; columns that are
often used in search conditions should be up front. This is for both storage and
performance reasons.
.....................................................................................................................................................
®
9-7
.....................................................................................................................................................
Primary Keys and Unique Keys

About the slide
Primary Keys See page 37
CREATE TABLE countries

( code NUMBER(6) NOT NULL
, name VARCHAR2(25) NOT NULL
, currency NUMBER (10,2) NOT NULL
);
ALTER TABLE countries
ADD CONSTRAINT cty_pk PRIMARY KEY
(code);
Constraint and Index name
9-6
Primary Keys
They are a strong concept that is usually enforced for every table.
• They can be made up of one or more columns; each has to be mandatory.
• They are declarative as a constraint and can be named. When creating a primary
key constraint, Oracle automatically creates a unique index in association with it.
• A foreign key usually refers to the primary key of a table, but may also refer to a
unique key.
Tables that do not have a primary key should have a unique key.
Note: Although Oracle allows a primary key to be updated, relational theory strongly
advises against this.
Unique Keys
A unique key is a key that for some reason was not selected to be the primary key. The
reasons may have been:
• Allowed nulls. Nulls may be allowed in Unique keys columns.
• Updatable. Unique key values may change but still need to remain unique. For
example, the home phone number of an employee or the license plate for a car.
There may be more than one unique key for each table.
Note: A Unique index is the additional structure Oracle uses to check the uniqueness
of values for primary keys and unique keys. Creating a unique key results
automatically in the creation of a unique index.
.....................................................................................................................................................
Primary Keys and Unique Keys
.....................................................................................................................................................
How to Choose the Primary Key

Following analysis there is a choice of what you want to use for a primary key. It does
not have to be seen or known by the user—it can do its work completely in the
background.
About the slide
See page 37
Primary Keys
• Choosing the Right Key

– Simplicity
– Ease of use
– Performance
– Size
– Meaningless
– Stability
9-7
Desirable Properties for Primary Key
Simple: A primary key should be as simple as possible although Oracle8 allows it to

consist of up to 32 columns. Primary key columns can be of various data types. Note
that UIDs, as they arise from data analysis, are often composed, not simple. You need
to consider replacing such a primary key by a simple key.
Easy to Use: Primary keys are normally used in join statements, so a primary key
should be easy to use. Writing a SQL statement to create a join between two tables is
easier if two columns only, rather than a large number, are involved in the join
predicate.
Does Not Kill Performance: A join operation using a single key usually performs
much better than a join using four key columns.
Small Size: Large-sized primary keys lead to large-sized foreign keys referencing
them. In general, the referencing table contains far more rows than the referenced
table. An oversized primary key can lead to a multiple of unnecessary bytes.
.....................................................................................................................................................
®
9-9
.....................................................................................................................................................
Meaningless: You could, for example, choose to use the name of a country as a
primary key, but even recent history has shown that countries may change their names.
Opt for numeric values rather than character values, and if using numbers, avoid
numbers with any particular meaning.
Stable: You should try to avoid selecting a primary key that is likely to be updated.
Bear in mind that it is very rare for real world things to stay stable for ever.
.....................................................................................................................................................
Artificial Keys
.....................................................................................................................................................
Artificial Keys
An artificial key is a meaningless, usually numeric, value that is assigned to a record
which functions as the primary key for the table. Artificial keys provide an interesting
alternative to complex primary keys. Artificial keys are also called surrogate keys.
About the slide
See page 37
Artificial Keys
AS (A) BS (B) CS (C)

pk * Id pk * Id pk * Id
* C1 * C2 * C3
fk1 = d_a_fk fk2 = d_b_fk fk3 = d_c_fk

DS (D)
u
pk ,fk1 * A_id
u
pk ,fk2 * B_id
XS (X) u
pk ,fk3 * C_id
pk Id * C4
* pk * Id
fk1 * D_a_id fk = x_d_fk
fk1
fk ** D_b_id
D_id
fk1 * D_c_id
o C5
9-8
Advantages
Artificial keys have the following advantages over composed keys:
• The extra space that is needed for the artificial key column and index is less, often
far less, than the space you save for the foreign key columns of referring tables.
• Join conditions consist of a single equation.
• The joins perform better.
• Internal references, which are completely invisible to the user, can be managed.
The modeled UID can than be implemented as a unique key, and made updatable
without needing cascade updates.
• Because they are meaningless, it is difficult to memorize them. Users will not even
attempt this.
• Some people really like them.
Disadvantages
Disadvantages of artificial keys are:
• Because they are meaningless, they always require joins to collect the meaning of
the foreign key column.
.....................................................................................................................................................
®
9-11
.....................................................................................................................................................
• More space is required for the indexes, if you decide to create an additional unique
key that consists of the original primary key columns.
• Because they are meaningless, it is difficult to memorize them. Users always need
a list of values or other help for entering the foreign key values.
• Some people really hate them.
Deciding About Artificial Keys?
Before Design
Negative: It would corrupt your data model, as you would add elements that have
no business meaning.
Positive: There is a close mapping between the conceptual and technical model
that reduces the chances of misunderstanding.
After Design
Positive: It really is a design decision based on current performance
considerations.
Tools like Oracle Designer let you decide about artificial keys during the initial
mapping of the ER model. This is a nice compromise.
.....................................................................................................................................................
Sequences
.....................................................................................................................................................
Sequences
About the slide
See page 38
Sequences
225
224
223
CREATE SEQUENCE sequence_name

INCREMENT BY number
START WITH number
MINVALUE number
MAXVALUE number
CACHE number | NOCACHE
CYCLE | NOCYCLE;
9-9
Some Sequence Characteristics

• A sequence is a database object that can generate a serial list of unique numbers
for columns of database tables.
• A sequence provides the quickest way of generating unique numbers.
• Sequences simplify application programming by automatically generating unique
numerical values that can be used as artificial key values.
• A sequence may be used to generate sequence numbers for any number of tables.
Usually a separate sequence is created for each table with an artificial key,
although there is no special need for that.
• A sequence guarantees generation of unique ascending or descending numbers. A
sequence does not guarantee that all consecutive numbers are actually used.
.....................................................................................................................................................
®
9-13
.....................................................................................................................................................
Foreign Key
By definition, Foreign Keys must refer to primary key or unique key values. You
should consider what should happen if the primary key (or unique key) value changes.
About the slide
See page 38
Foreign Key Behavior
Delete Update
Restrict
Cascade
Default / Nullify
Supported by Oracle through declaration
9-10
Referential Integrity
There are two aspects to consider:
• The rules you want to implement to support business constraints
• The functionalities Oracle provides for these rules
Relational theory describes four possible kinds of behavior for a foreign key. For every
foreign key decide what kind of behavior you want it to have.
The behaviors describe what the foreign key should do when the value of the key it
refers to changes.
Restrict Delete
Restrict delete means that no deletes of a primary (or unique) key value are allowed
when referencing values exist. This is supported by Oracle. This is the most
commonly used foreign key behavior.
Restrict Update
Restrict update means that no updates of a primary (or unique) key value are allowed
when referencing values exist. This is supported by Oracle. Note that this behavior is
unnecessary in the case of artificial keys as these are probably never updated.
.....................................................................................................................................................
Sequences
.....................................................................................................................................................
Note that restrict update is not the same concept as nontransferability. Restrict update
prevents the update of a referenced primary key value. Nontransferability means that
the foreign key columns are not updatable.
Cascade Delete
Cascade delete means that deletion of a row causes all rows that reference that row
through a foreign key marked as “cascade” will be deleted automatically. Cascade
delete is an option that Oracle supports.
The complete delete operation will fail if, during the cascade, there is a record
somewhere that cannot be deleted. This may happen if the record to be deleted is
referred to through a restrict delete foreign key.
Cascade delete is a very powerful mechanism that should be used with care.
Cascade Update
Cascade update means that after a primary key value is updated, this change is
propagated to all the foreign key columns referencing it.
Cascade update and nontransferability often come together.
Default and Nullify

The default and the nullify option mean that on delete or update of the primary key
value, the related foreign key values will acquire a default value or will be set to
NULL.
These options can be implemented by creating an update database trigger on the table
referred to by the foreign key. Clearly, the nullify option is only valid if the foreign key
is optional.
Typical Use
Usually, many foreign keys are defined as restrict delete. This does not prevent the
referred record being deleted; it just forces the user to consciously remove or transfer
all referring rows.
Of course, when you use artificial keys you can set all foreign key update properties to
“restrict” as there will never be a good reason for updating an artificial key value.
.....................................................................................................................................................
®
9-15
.....................................................................................................................................................
Indexes
Indexes are database structures that are stored separately from the tables they depend
on. In a relational database you can query any column, independently of the existence
of an index on that column.
About the slide
See page 38
Indexes
• Performance
Name Phone
b
c
ALBERT 2655
d
ALFRED 3544 ef
gh
ij
ALICE 7593 kl
m
ALLISON 3456
no
pq
ALVIN 8642
rs
tu
ALPHONSO 2841 vw
xyz
• Uniqueness
9-11
Indexes are used for two reasons:

• To speed up queries
• To ensure uniqueness if required
Performance
Indexes are created to provide a fast method to retrieve values. However, indexes can
slow down performance on DML statements.
Oracle provides a wide range of index types. You must choose the type which is
suitable for its intended use.
Uniqueness
A unique index is an efficient structure to ensure that the values are not duplicated
within the set of columns included in the index. Unique indexes are automatically
created when you create a primary or unique key. The name of the index in that case is
the same as the name of the key constraint.
.....................................................................................................................................................
Indexes
.....................................................................................................................................................
Index Types
About the slide
See page 38
Choosing Indexes
B*tree
Bitmap
aba .1.2.5 X Y Z
abb .1.4.5
Reverse 0 1 0
aba .1.2.5 1 0 0
abb .1.3.5 bba .1.3.5 0 0 1
abc .1.1.5 cba .1.1.5 1 0 0
bba .1.4.5 ... 0 0 1
...
C1 C2 C1 C2
I.O.Table abc Y
aba X
abb Z aba X
abc Y abb Z
bba Z bba X
bbc X bbc Z
9-12
B*Tree
The classical structure of an index, if not explicitly specified otherwise, is the B*Tree
(also known as Tree balanced) index. It is specially designed for online transaction
processing systems. They have a proven efficiency and Oracle has offered them for
some time. They easily support insert, update, and delete.
Typical use: General purpose
Reverse Key
Based on that classical structure of the B*Tree, Oracle offers a reverse key index
which has most of the properties of the B*Tree but in which the bytes of each indexed
column are reversed.
Typical use: In an Oracle Parallel Server environment, where such an arrangement

can help avoid performance degradation in indexes
Bitmap
A bitmap index stores for each individual value of the indexed column, if a row
contains this value or not.
Typical use: Data warehouse environment. Bitmap indexes have a proven efficiency
in On Line Analytical Process systems when ad-hoc queries can be intensive and the
number of distinct values for the indexed column is not high.
.....................................................................................................................................................
®
9-17
.....................................................................................................................................................
Bitmap indexes require less space than a B*Tree index but they do not support inserts,
updates, and deletes as well as a B*Tree.
Index Organized Table

An index organized table is a table that contains rows that are stored in an ordered
way, using the B*Tree technique. It provides the speed that indexes provide and does
not require a separate index. The only restriction in its use is that you cannot create
additional indexes for this Index Organized table.
Typical use: Tables that are always accessed through exactly the same path, in
particular when storing large objects.
Concatenated Index
You can create an index that includes more than one column. These are called
concatenated indexes. The order in which you specify the columns has a strong impact
on the way Oracle can use the index. Set the column that is always in a Where clause
as the first column of the index. This is called the leading part of the index.
Function Based Index

Since Oracle8i it is possible to create an index based on a SQL function.
Typical use: Create an index on the first three characters of a name using the substr
function or the year component of a date using the to_char function.
.....................................................................................................................................................
Choosing Columns to Index
.....................................................................................................................................................
Choosing Columns to Index

About the slide
See page 39
Which Columns to Index?
• Primary key columns and Unique Key columns

(Up to Version 6)
• Foreign Key columns
• When significant better performance can be
observed in SELECT statements
! Avoid indexing:
• Small tables
• Columns frequently updated
9-13
Candidate Columns for Regular B*Tree Indexing

• Columns used in join conditions to improve performance on joins
• Columns that contain a wide range of values
• Columns that are often used in the Where clause of query
• Columns that are often used in an Order By clause of a query
Candidate Columns for Bitmap Indexing

• Columns that have few distinct values such as, for example, a column containing
indicator values (Y/N) or a column for gender
Columns Less Suitable for Indexing

• Columns that contain many NULL values where you usually search rows with the
NULL values
Columns that Cannot or Should Not be Indexed

• LONG and LONG RAW columns cannot be indexed
• Columns that are hardly ever used in Where / Order By clauses
• Small tables occupying only few data blocks
.....................................................................................................................................................
®
9-19
.....................................................................................................................................................
Temporary Indexes
• Indexes can be created and dropped for a particular incidental use. For example,
you can decide to create an index right before a report is run and then drop it
afterwards.
General Recommendations
• Limit the number of indexes per table. Although a table can have any number of
indexes this does not necessarily improve performance; the more indexes, the
more overhead is incurred when there are updates or deletes.
• As a rule of thumb, if there is any doubt, do not create the index. You can always
create it later.
• It is very likely that the initial set of indexes will have to change after some time,
because of changes of the characteristics of the system. Typically, the number of
different values in a column can initially be very low but increase during the life
cycle of a system. Initially, an index would not be of value but it would be later.
.....................................................................................................................................................
When Are Indexes Used?
.....................................................................................................................................................
When Are Indexes Used?

About the slide
See page 39
When Can Indexes be Used?
• When referenced in a Where clause or Order By

• When the Where clause does not include some
operators
• When the optimizer decides
• With hints in the SQL statement
9-14
You may have created an index to improve performance but without seeing any
benefits.
For Oracle to use them, indexed columns need to be referenced in the Where clause of
a SQL statement, or in the order by, while the Where clause must not include the
following:
• IS NULL
• IS NOT NULL
• !=
• LIKE
• When the column is affected by an operation or function (unless you use a
function-based index and the condition uses the same function)
For example, suppose column X contains many nulls and a few numeric, positive
values. Suppose queries often select all rows having a NOT NULL value. Finally,
suppose an index is created on X.
In this case, the condition WHERE X > 0 is preferable to WHERE X IS NOT NULL
because in the first situation Oracle would use an index on X and in the second Oracle
would not.
Yet, even if it was written in this way, it is the optimizer’s choice to decide whether to
use indexes or not. The decision is based on rules or on statistics.You can stimulate the
optimizer to use indexes using hints in your SQL statements.
.....................................................................................................................................................
®
9-21
.....................................................................................................................................................
Table Partitioning
Oracle provides an interesting feature to solve performance and administration
problems on tables with a large number of rows.
About the slide
See page 39
Partitioning Tables and Indexes
CUSTOMERS
Col1 Col2 Col3 Region
CUSTOMERS_R1
CUSTOMERS_R2
9-15
Partitioned Table
Since Oracle8, when creating a table, you can specify the criteria on which you want
to divide the table and make a horizontal partitioning. There are then as many
partitioned tables as there are distinct values in the column. Each partitioned table has
a specific name but access is made referring to the global name of the table. The
optimizer then decides which partition to access, depending on the value of the Where
clause.
The main issue of this feature is to manipulate considerably smaller pieces of data and
then improve the speed of SQL statements. Suppose you want to query on customers
located in a specific region, Oracle does not need to access all rows of the
CUSTOMERS table but can limit its search to the piece holding all customers of this
region only.
Logically, the table behaves as one object; physically, data is stored in different places.
Partitioned Index
Using the same idea, an index may be partitioned. It does not need to match with the
table partitioning. It may have different partitioning criteria and have a different
number of partitions to the table. This may be useful in the situation where the answer
to particular queries can always be found in the partitioned index.
.....................................................................................................................................................
Views
.....................................................................................................................................................
Views
A view is a window onto the database. It is defined by a SELECT statement which is
named and stored in the database. Therefore a view has no data of its own—it relays
information from underlying tables.
About the slide

See page 39
Views
T1 T2 T3 T4
V1 V2 V3 V4
• Restricting access
• Presentation of data
• Isolate applications from data structure
• Save complex queries
• Simplify user commands
9-16
Usages of Views
• Restricting access: The view mechanism is one of the possible ways to hide
columns and rows from the tables it is based on.
• Presenting data: A view can be used to present data in a more understandable way
to end-users. For example, a view can present calculated data built from
elementary information that is stored in tables.
• Isolating application from data structures: Applications may be based on views
rather than tables, where there is a high risk that the structure might change. If a
view is used, the application would need no maintenance providing the view
remains untouched, even though the underlying tables were modified.
• Saving complex queries and simplifying commands: Views can be used to hide the
complexity of the data structure, allowing users to create queries over multiple
tables without having to know how to join the tables together.
• Simplifying user commands.
.....................................................................................................................................................
®
9-23
.....................................................................................................................................................
Use of Views
Reasons for Views
• Advantages
– Dynamic views
– Present denormalized data from normalized
tables
– Simplify SQL statements
• Disadvantages
– May affect performances
– Restricted DML in some cases
9-17
Advantages
• You can use a view to present derived data to end users without having to store
them in the database. Typically, you would show completely denormalized, pre-
joined information in views that would allow end users to write simple SELECT
statements like SELECT * FROM ... WHERE ...
• Views can be made dynamic, for example, showing data that depend on which user
you are or what day it is.
For example, you could create a view that shows localized help messages.
According to the user name, the system can find the preferred language in a
PREFERENCES table and next return a message in this language. A single view
returns different values depending on the name of the user.
Another example type of view can be used to allow a user to access data between
8:00 am and 6:00 pm on weekdays only.
Disadvantages
• Views are always somewhat slower, which is due to the fact that the parse time is
slightly longer. Once a table and its columns are found, the query can be
immediately executed. Query criteria are linked with “and” to the criteria of the
view. This can affect the execution plan generated by the optimizer.
• Even if views behave almost like tables, there are still some restrictions when
using views for insert, update, and delete statements.
.....................................................................................................................................................
Old-Fashioned Design
.....................................................................................................................................................
Old-Fashioned Design
Going through existing systems, you may find some old-fashioned design techniques.
These techniques were used at the time the RDBMS features were not so advanced.
About the slide
See page 40
Old Fashioned Design
• Unique index
• Views with “Check option” clause
• Generic Arc implementation
9-18
Unique Index
Unique Indexes used to be created manually on the primary key columns because the
primary key constraint could not be declared up to Oracle7.
Check Option Views

In earlier versions of Oracle, it was not unusual to create a view “with a check option”.
These views, now obsolete, could be used to some extent to enforce data integrity and
referential integrity before Oracle7.
There is no functionality in a view with a check option that cannot be coded in a
database trigger. The declaration of integrity constraints and coding of database
triggers is now the preferred way to handle this.
.....................................................................................................................................................
®
9-25
.....................................................................................................................................................
Generic Arc Implementation

About the slide
See page 40
Generic Arc Implementation
X
A # Id
# Id * Name
* Name
Y
# Id
* Name
AS (A)
...
* Table_name (X or Y)
* Fk_id
9-19
The generic arc implementation is a fossil construction you may find in old systems.
In the implementation of the arc of entity A in the example, the three relationships in
the arc were merged into one generic foreign key column Fk_id. Added to table AS is
a NOT NULL column that keeps the information about which table the foreign key
value refers to. This used to be a popular technique because it could make use of a
NOT NULL constraint on Fk_id when the arc was mandatory.
This solution for implementing arcs should now be avoided for the following
limitations:
• Since Oracle7 the arc can now be implemented by simply declaring two foreign
keys and writing one check constraint.
• The joins may be very inefficient as, in many cases, you would need the time-
consuming union operator:
select A.Name, X.Name, ’X’ Type
from AS A, XS X
where ...
union
select A.Name, Y.Name, ’Y’
from AS A, YS Y
where ...
• Foreign key constraint for the foreign key column cannot be declared since it
cannot reference more than one primary key.
.....................................................................................................................................................
Distributed Design
.....................................................................................................................................................
Distributed Design
This is characterized as many physical databases, located at different nodes, but
appearing to be a single “logical database”.
About the slide
See page 40
Distributed Database
• Different physical databases appear as one logical

database
9-20
Characteristics
• Multiple physical databases
• One logical database view
• Possibly dissimilar processors
• Kernel runs wherever a part of the database exists
The multiple physical databases are not necessarily copies of each other or part of each
other.
You can decide on how to spread the individual table content across the different
databases on the different partitioning principles. You can decide for a vertical or
horizontal technique, or a combination of both.
.....................................................................................................................................................
®
9-27
.....................................................................................................................................................
Benefits of Distributed Design

About the slide
See page 40
Benefits of Distributed Databases
• Resilience
• Reduced line traffic
• Location transparency
• Local autonomy
• Easier growth path
but
• Increased, distributed, complexity
9-21
• Improved flexibility and resilience. Access to data is not dependent on only one
machine or link. If there is any failure then some data is still accessible on the local
nodes. A failing link can automatically be rerouted via alternative links.
• Improved response time by having the data close to the usual users of the data.
This may reduce the line traffic dramatically. For example, in the model of
ElectronicMail, it is very likely that each country will have its own database. This
database will store in its own messages table the messages that belong to the
people registered in that country.
• Location transparency allows the physical data to be moved without the need to
change applications or notify users.
• Local autonomy allows each of the physical databases:
– To be managed independently.
– To have definitions and access rights created and controlled locally.
• An easier growth path is achieved:
– More processes can be added to the network
– More databases can be included on a node.
– Software update is independent of physical structure.
Disadvantage
A major disadvantage of distributed design is the often very complex configuration:
with the data the complexity is also distributed. System maintenance is complicated.
.....................................................................................................................................................
Oracle Database Structure
.....................................................................................................................................................
Oracle Database Structure

About the slide
DATABASE See page 41
Database Structure
consists
of part of
TABLESPACE consists
of
resides container part of
in SEGMENT
of
OTHER TABLE INDEX
SEGMENT SEGMENT SEGMENT
sliced in sliced in consists
of
located in part of part of
residence
of TABLE OR INDEX PARTITION part of
DATA FILE EXTENT USED FREE
consists resides in
of part of residence of
DATA BLOCK
9-22
Tablespaces
The diagram shows the structure of a Oracle database.
An Oracle database consists of one or more tablespaces. Each tablespace can
hold a number of segments, and each segment must be wholly contained in
its tablespaces. The SYSTEM tablespace is created as part of the database
creation, and should be reserved for the Oracle Data Dictionary and related
tables only. You should not create application data structures in this
tablespace. You are advised to create separate tablespaces for different types
of segments.
Segments
A segment is the space occupied by a database object. There are three types
of segments: a table segment, an index segment or an other segment, that is
used for clusters. Only the other segments must be part of one tablespace.
Partitions
Usually, a segment is assigned to a single tablespace. However, with Oracle8
it is possible to spread a table or index segment into more than one
tablespace. This technique is called partitioning. A partition is the part of a
table segment (or index segment) that resides in one tablespace.
.....................................................................................................................................................
®
9-29
.....................................................................................................................................................
Extents
Each time more space is needed by a segment, a number of contiguous
blocks is allocated as an extent. There is no maximum limit on the number
of extents that can be allocated to a segment. It is usually preferable to avoid
an excessive number of small extents by ensuring that the segment has a
sufficiently large initial extent.
Data Files
Data files are the operating system files that physically contain the database data. Data
files consist of data blocks.
Data Blocks
A data block is the smallest amount of data Oracle reads in one read operation. A data
block always contains information from one extent only.
There is a distinction between the logical table, made up of rows with columns, and
the physical table, taking space that is made up of database blocks organized in extents
and located in data files.
.....................................................................................................................................................
Summary
.....................................................................................................................................................
Summary
Summary
• Data Types
• Primary, Foreign, and Artificial Keys
• Indexes
• Partitioning
• Views
• Distributed design
9-23
• Oracle provides a large choice of data types for the columns of the tables.
• Primary keys are needed for tables. Artificial keys can be a good solution to
implement complex primary keys.
• Indexes improve performance of queries and provide a mechanism for
guaranteeing unique values.
• Partitioning tables can also be a solution to performance problems.
• Views are a flexible, secure, and convenient object for users.
• Distributed Design is a complex technique. It allows data to be located closer to
the user.
.....................................................................................................................................................
®
9-31
.....................................................................................................................................................
Practice 9—1: Data Types

Moonlight Coffees
Goal (See page 41)
The purpose of this practice is to perform a quality check on
proposed data types.
Scenario
Use the model that illustrates Moonlight pricing.
CURRENCY of COUNTRY
# Code # Code Moonlight Pricing
with
in in with with
PRODUCT GROUP in
# Name with SHOP
from to with # No
EXCHANGE in * Name
RATE * Address
PRODUCT
# Month of * City
* Rate for
GLOBAL LOCAL
PRICELIST # Code # Name
# Start Date o Size
* End Date with
with with with of

LOCAL PRICE LANGUAGE
# Start Date # Code
in of
o End Date
GLOBAL PRICE of
of * Amount
* Amount in
PRODUCT NAME
* Name
9-25
.....................................................................................................................................................
Practice 9—1: Data Types
.....................................................................................................................................................
Your Assignment
1 Here you see table names and column names and the suggested data type. Do a
quality check on these. If you think it is appropriate, suggest an alternative.
Table Column Suggested Your Choice

Data Type Data Type
COUNTRIES Code Varchar2(2)
CURRENCIES Code Varchar2(3)
EXCHANGE_RATES Month Date
Rate Number(8,4)
PRICE_LISTS Start_date Date
End_date Date
PRODUCT_GROUPS Name Char(8)
PRODUCTS Code Char(10)
Size Number(4,2)
Pdt_type Number(1)
2 Suggest data types for the following columns. They are all based on previous
practices.
Table Column Your Choice Data Type

GLOBAL_PRICES Amount
LOCAL_PRICES Start_date
End_date
Amount
SHOPS Name
Address
City
3 What data type would you use for a column that contains times only?
.....................................................................................................................................................
®
9-33
.....................................................................................................................................................
Practice 9—2: Artificial Keys

Moonlight Coffees
Goal (See page 42)
You are coming to the end of your contract for Moonlight Coffees.
The job is almost finished!
Scenario
You need to make decisions on possible artificial keys for some of the Moonlight
tables. The model is the same as the one used in the previous practice.
Your Assignment
1 Indicate for each table if you see benefits of creating an artificial key and why.
COUNTRIES
GLOBAL_PRICES
PRICE_LISTS
2 For which tables (if any) based on the Moonlight model does it not make any sense
at all to create artificial keys?
.....................................................................................................................................................
Practice 9—3: Product Pictures
.....................................................................................................................................................
Practice 9—3: Product Pictures

Moonlight Coffees
Goal (See page 42)
The purpose of this practice is to modify a design to serve new

requirements.
Scenario
This is your last task for Moonlight coffees. Tomorrow you are free to forget all about
Moonlight and only drink coffee!
The decision has been made to make the first steps into the e-commerce market. One
objective is to allow customers to consult Moonlight’s website. This site should
provide product information. For each product at least two additional attributes have
been identified.
The first is the attribute Picture for images of the products. The second is an attribute
HTML Document that holds the product description that can be displayed with a
browser. Other attributes may follow.
Your Assignment
1 Decide what data type you would advise to be used for each column.
2 You have heard that an old Oracle version would not accept more than one long
type column per table. You are not sure if this is still a limitation. Advise about the
implementation.
.....................................................................................................................................................
®
9-35
.....................................................................................................................................................
Instructor Notes
Instructor Note
Topic Timings
99
Lecture 45-60 minutes
D ataba se D e sign C on sid eration s
Total 60-90 minutes
This chapter deals with Oracle in particular.

O verview
The suggestions made here are directly
• O ra cle sp ec ific D es ig n C o n s id eratio n s influenced by properties and behavior of the
• D ata In teg rity Is su e s
• P e rfo rm a nc e C o n sid e ra tio n s Oracle database. Some of the objects discussed
• S to ra ge Iss u es
here, such as sequences, do not exist in other
databases; the various database systems use
different ways to organize memory and the
9 -2
various optimizers do not behave in the same
way.
These parameters are generic ones. Many more
W hy A dapt D ata D e sign?
exist behind each one.
• U s er E x pe cta tion s
• V o lu m e s
A d a p ted
• H a rd w a re
Initial design P h ys ic al
• N e tw o rk
D e sig n
• O .S .
• O rac le sp e cific s
9 -3
Be aware of restrictions that may apply

O rac le D a ta T ypes
according to the programming language used
• D ep e n d in g o n :
– D o m a in s for coding applications. You may have
– S to ra g e is su e
– P erfo rm a n ce problems with large object type data types.
– Use
• S e le ct a d a ta typ e fo r c o lu m n s:
– C h arac te r
– Num ber
– D a te
– L a rg e O b je cts
9 -4
.....................................................................................................................................................
Instructor Notes
.....................................................................................................................................................
This sequence is suggested by Oracle CDM.

Sug gested C olum n Seq uenc e
• P rim a ry ke y co lu m n s
• U n iq u e K ey c o lu m n s
• F o re ig n k ey c o lu m n s
• M an d ato ry co lu m n s
• O ptio n a l c o lu m n s
L a rg e o bjec t co lum n s a lw ays a t th e e n d
9 -5
Up to release 6 of Oracle, users had to create

P rim ary K e ys
the unique indexes for Primary keys manually.
C R E A TE TA B LE countries
( code N U M B E R (6) N O T N U LL
, nam e V A R C H A R 2(25) N O T N U LL
, currency N U M B E R (10,2 ) N O T N U LL
);
A LTE R TA B LE countries
A D D C O N S TR A IN T cty_pk P R IM A R Y K E Y
(code );
C on strain t an d Inde x n am e
9 -6
Since Oracle8, 32 columns can be concatenated

P rim ary K e ys
in an index. Prior to Oracle8 the limit was 16
• C h oo sin g the R igh t K ey columns.
– S im plicity
–
–
E a se o f u se
P e rfo rm a nce
When you use a meaningful primary key, it is
– S ize
– M ean in gles s most likely that you may have to update it some
– S tability
day. If this happens once during the life cycle of
a system it may not be a problem, but if it
9 -7 happens far more frequently, you should
reconsider the choice.
There are a lot of advantages in choosing
A rtificia l K e ys
artificial keys, but it should be a well-
A S (A ) B S (B ) C S (C )
pk * Id
* C1
pk * Id
* C2
pk * Id
* C3
considered choice, not an automatism.
fk 1 = d_a_fk fk 2 = d_ b_fk fk 3 = d_c _fk
D S (D )
u
pk ,fk 1 * A _id
u
pk ,fk 2 * B _id
X S (X ) u
pk ,fk 3 * C _id
pk Id * C4
* pk * Id
fk1 * D _a_id fk = x_d_fk
fk1
fk ** D _b_id
_id
fk1 * D _c_id
o C5
9 -8
.....................................................................................................................................................
®
9-37
.....................................................................................................................................................
Sequences are specific to Oracle. If, for some

S equ ence s
reason, no gap between numbers is allowed
(other than the gap explicitly chosen by the
2 25
2 24 value of increment by), you need to use a
2 23
C R E A TE S E Q U E N C E se quence_nam e
IN C R E M E N T B Y nu m ber
different mechanism.
S TA R T W ITH nu m ber
M IN V A LU E nu m ber
M A X V A LU E nu m ber
C A C H E num ber | N O C A C H E
C Y C LE | N O C Y C LE ;
9 -9
This is relational theory.

Foreign K e y B eh avior
Nullify can be seen as a special case of Default.
D ele te U pdate
R e strict
Oracle supports all through coding, the marked
C a scade
ones can be declared.
D e fault / N ullify
Su pp orted b y O racle th ro u gh de claration
9 -1 0
Indexes are very a very powerful means to

Inde xes
• P erfo rm an ce
improve performance.
N am e
A LB ER T
P hone
26 55
b
c
Nevertheless, it is always a matter of balancing
d
A LF R E D 35 44 ef
gh
ij
between better performances on queries on one
A LIC E 75 93
side and poorer DML statements on the other.

kl
m
A LL IS O N 34 56
no
pq
A LV IN 86 42
rs
A LP H O N S O 28 41
tu
vw
xyz
xy z
Usually only few indexes are created when a
• U niqu en ess system goes live. During the life cycle of a
9 -1 1
system indexes are added if performance so
dictates.
The function-based index is not represented
C hoos in g Ind exes
B *tree
here (the reverse index can be seen as a special
B itm ap
case).
a ba .1 .2.5 X Y Z
a bb .1.4 .5
R everse 0 1 0
a ba .1 .2.5
a bb .1 .3 .5
a bc .1 .1.5
b ba .1 .4 .5
bb a .1.3 .5
c ba .1 .1.5
...
1
0
1
0
0
0
0
0
0
1
0
1
Note that delegates may assume that table data
...
are always stored in an ordered way.

C1 C2 C1 C2
I.O .Table ab c Y
aba X
abb Z ab a X
abc Y ab b Z
bb a Z bb a X
bb c X bb c Z
9 -1 2
.....................................................................................................................................................
Instructor Notes
.....................................................................................................................................................
General rule: start with a few indexes and add

W hich C olum ns to Ind ex?
new ones when necessary.
• P rim a ry ke y co lu m n s a n d U n iq u e K ey co lum n s
(U p to V e rs io n 6)
Some of the best performance improvements
•
•
F o re ig n K e y c o lu m n s
W h e n sig n ific a nt b e tte r p e rfo rm an c e ca n b e
have been accomplished by dropping excessive
o b s erv ed in S E L E C T s ta tem e n ts
indexes.
! A vo id in dexin g:
• S m a ll ta bles
• C o lum n s fre q u en tly u p d ated
9 -1 3
To force the use of indexes, you can use hints.

W he n C an Inde xes be U s ed?
For example, you can use the following syntax:
• W h e n re fe ren ced in a W h ere c lau s e or O rd e r B y SELECT /*+ INDEX (employees I_name) */
• W h e n th e W h ere c la u se d o es n o t in c lu d e so m e
o pe ra to rs
• W h e n th e o p tim izer d ec id es
Id
• W ith h in ts in th e S Q L sta te m en t
, name
, address
FROM employees
9 -1 4 WHERE name > ’L’;
Table partitions help to keep local tables small

Partitioning Tab le s a nd In dexe s and focussed on local needs.
C U S TO M E R S
C o l1 C ol2 C o l3 R eg io n
C U S TO M E R S _R 1
C ol1 C o l2 C ol3 R egio n
C U S TO M E R S _R 2
C ol1 C o l2 C ol3 R egio n
9 -1 5
Views do not have a counterpart in ER

View s
T1
T1 T2 T
T33 T4
modeling, although some may result from
supertype/subtype implementation.
V1 V2 V3 V4
• R es trictin g ac cess
• P resen ta tion o f d ata
• Iso la te a pp lica tion s fro m data s tru cture
• S av e co m p lex qu eries
• S im plify u ser co m m an ds
9 -1 6
.....................................................................................................................................................
®
9-39
.....................................................................................................................................................
R easo ns for V iew s
• A d va n tag e s
– D y n am ic vie w s
– P re se n t de n o rm a lize d d ata fro m n o rm alized
ta b les
– S im p lify S Q L s tatem en ts
• D is a dv an ta g es
– M a y a ffe ct pe rfo rm a n ce s
– R e stricted DM L in so m e c as es
9 -1 7
Some checks can be implemented in a view

O ld Fashion ed D es ig n
with check option. We advise that view should
• U n iq u e in d e x not be used but database triggers should be used
• V iew s w ith “C h e ck o p tio n ” c lau s e
• G en e ric A rc im p le m e nta tio n instead. The reason is primarily to keep all
checks in one place.
9 -1 8
Generic arcs should not be used in any new

G en eric A rc Im plem entatio n
system.
X
A # Id
# Id * Name
* N am e
Y
# Id
* Name
A S (A )
...
* Table_nam e (X o r Y )
* Fk_id
9 -1 9
Distributed database: one database design,

D istrib uted D a tab ase
implemented in several physical databases,
• D ifferen t p h ys ic al d atab a se s ap p e ar as o n e lo g ic al
d a ta b as e
acting as one logical database.
9 -2 0
Do not go into details about two-phase commit

B ene fits of D istributed D ata base s
and transaction mechanisms and so on. Only
• R e silien c e touch the surface.
• R e du ce d lin e traffic
• L o c atio n tra n sp a ren c y
• L o c al au to n o m y
• E as ie r g ro w th p a th
but
• Inc re as ed , d istribu te d , c o m p le xity
9 -2 1
.....................................................................................................................................................
Instructor Notes
.....................................................................................................................................................
Database Structure described in ER terms.

DATABASE
D atab ase Structure
consists
o f part of
TA B LE S P A C E consists
of
resides container part of
in SEGMENT
of
O TH E R TA B LE IN D E X
SEGMENT SEGMENT SEGMENT
sliced in sliced in consists
of
located in part of part of
residence
of TA B LE O R IN D E X P A R T ITIO N part of
D A T A FILE E X TE N T U S E D FR E E
consists resides in
of part of residence of
D A TA B LO C K
9 -2 2
Sum m ary
• D a ta T y pe s
• P rim a ry , F o re ig n , a n d A rtificial Ke y s
• Ind e x es
• P artitio n in g
• V iew s
• D is trib u ted d e sign
9 -2 3

Prac tices
• D a ta T y pe s Practice 3day 4day

• A rtific ial K ey s
• P ro d u ct P ictu re s Data Types Yes Yes
Artificial Keys Opt Cha
Product Pictures opt opt
9 -2 4
C U R R EN C Y
# C od e
of C O U N TR Y
# C o de M oon ligh t Pricing
Practice 9-1
w ith
in in w ith w ith
PRODUCT GROUP in
# N am e w ith SH O P
fro m to w ith # No
EX C H A N G E in * N am e
RATE * A d dress
PR O D U C T
# M o n th of * C ity
* R ate for
GLOBAL LO C A L
P R IC EL IST # C o de # N am e
# Start D ate o Size
* En d D ate w ith
w ith w ith w ith of

LO C A L P R IC E LA N G U A G E
# Start D ate # C od e
in of
o En d D ate of
G L OB A L P R IC E of * A m o un t
* A m o un t in
PRODUCT NAME
* N am e
9 -2 5
.....................................................................................................................................................
®
9-41
.....................................................................................................................................................
Practice 9-2
D a ta T ypes (1)
Table C o lum n S uggested Y our C hoice

This practice should not take much time and
C O U N TR IE S
C U R R E N C IE S
C o de
C o de
D ata Type
V archar2(2)
V archar2(3)
D ata Type
E X C H A N G E _ R A TE S M o nth D ate
R a te N um ber(8,4)
P R IC E _LIS TS S ta rt_date D ate
E n d_date D ate
P R O D U C T_G R O U P S Name C har(8)
P R O D U C TS C o de C har(10)
S iz e N um ber(4,2)
P d t_type N um ber(1)
9 -2 6
Practice 9-3
D a ta T ypes (2)
Table C olu m n Y our C hoice D ata Type

This practice should not take much time and
G LO B A L_P R IC E S
LO C A L_P R IC E S
A m o unt
S tart_date
E nd_ date
A m o unt
S HOPS N am e
A ddress
C ity
9 -2 7
.....................................................................................................................................................
.....................................................................................................................................................

Name: Full Email:

Lesson number: 9

Detail comments
Content Page
Type Number
Photocopy this page and fax to :Oracle Designer Education Products @ +(44) 118.924.5181
.....................................................................................................................................................
®
9-43
.....................................................................................................................................................
.....................................................................................................................................................
Suggested Graphics
.....................................................................................................................................................
Suggested Graphics

.....................................................................................................................................................
®
9-45
.....................................................................................................................................................

520 Oracle Parkway
Thames Valley Park
Reading - Berkshire
England
fold here
.....................................................................................................................................................
A
................................
Solutions
.....................................................................................................................................................
Introduction to Solutions
Before You Proceed
In this section you find solutions to the practices. A set of solutions—not the definitive
solutions. Solutions for practices are highly dependent on the business context.
Although some background is given in the practice text, a lot remains unsaid. You
have had to use your imagination and make your assumptions to fill in the blanks.
Even in the real-life modeling world of the analyst there is no such thing as the
conceptual model. You could even argue that there is no such thing as the real world.
There is no is. Everything is perceived by an observer and it is the perception only that
you can model.
Bear this in mind when you compare the given solutions with those you created. If
there are discrepancies, try to be aware of your assumptions and try to find those that
are at the basis of the presented solutions.
This text is not meant to be a disclaimer. The solutions presented here were created
with care, and are as close as possible to the likely direction a real-life model would
take.
Solution List
Practice See Page

Practice 1—1 Instance or Entity: Solution 4
Practice 1—2 Guest: Solution 5
Practice 1—3 Reading: Solution 6
Practice 1—4 Read and Comment: Solution 7
Practice 1—5 Hotel: Solution 8
Practice 1—6 Recipe: Solution 9
Practice 2—1 Books: Solution 11
Practice 2—2 Moonlight: Solution 12
Practice 2—3 Shops: Solution 13
Practice 2—4 Subtypes: Solution 14
Practice 2—5 Schedule: Solution 15
Practice 2—6 Address: Solution 16
Practice 3—1 Read the Relationship: Solution 18
Practice 3—2 Find a Context: Solution 19
Practice 3—3 Name the Intersection Entity: Solution 20
Practice 3—4 Receipt: Solution 21
Practice 3—5 Moonlight P&O: Solution 23
Practice 3—6 Price List: Solution 27
.....................................................................................................................................................
A-2 Data Modeling and Relational Database Design
Introduction to Solutions
.....................................................................................................................................................
Practice See Page

Practice 3—7 E-mail: Solution 28
Practice 3—8 Holiday: Solution 30
Practice 4—1 Identification Please: Solution 32
Practice 4—2 Identification: Solution 34
Practice 4—3 Moonlight UID: Solution 37
Practice 4—4 Tables: Solution 38
Practice 4—5 Constraints: Solution 39
Practice 5—1 Shift: Solution 40
Practice 5—2 Strawberry Wafer: Solution 41
Practice 5—3 Bundles: Solution 42
Practice 5—4 Product Structure: Solution 44
Practice 6—1 Patterns: Solution 45
Practice 6—2 Data Warehouse: Solution 46
Practice 6—3 Argos and Erats: Solution 47
Practice 6—4 Synonym: Solution 48
Practice 7—1 Mapping Supertype: Solution 49
Practice 7—2 Quality Check Subtype Implementation: 50
Solution
Practice 7—3 Quality Check Arc Implementation: Solution 51
Practice 7—4 Primary Keys and Columns: Solution 52
Practice 8—1 Name that Denormalization: Solution 53
Practice 8—2 Triggers: Solution 54
Practice 8—3 Denormalize Price Lists: Solution 56
Practice 8—4 Global Naming: Solution 58
Practice 9—1 Data Types: Solution 59
Practice 9—2 Artificial Keys: Solution 61
Practice 9—3 Product Pictures: Solution 62
.....................................................................................................................................................
Data Modeling and Relational Database Design A-3
.....................................................................................................................................................
Practice 1—1 Instance or Entity: Solution

List which of the following concepts you think is an Entity, Attribute, or Instance. If
you mark one as an entity, then give an example instance. If you mark one as an
attribute or instance, give an entity. For the last three rows, find a concept that fits.
A possible solution:
Concept E/A/I? Example Instance or Entity

PRESIDENT E Gandhi, Washington, Gorbachov
ELLA FITZGERALD I STAR, SINGER, PERSON
DOG E Snoopy
ANIMAL E Cat, Dog, ...
HEIGHT A PERSON, BUILDING, ...
TYPE OF TRANSPORT E CAR
Number of Wheels A CAR
My current car I CAR
A-1
In fact, all concepts can be both an entity and an instance of an entity. This depends on
the business perspective. Some businesses have strange perspectives...
Even Ella Fitzgerald could be an entity. Think in the context of a look-alike contest or
the context of an art collection where there may be several instances of an Ella
Fitzgerald.
.....................................................................................................................................................
Practice 1—2 Guest: Solution
.....................................................................................................................................................
Practice 1—2 Guest: Solution

Draw a line between the attribute and the entity or entities it describes.
A possible solution:
Address
Arrival Date
Family Name
GUEST Room Number
HOTEL Floor Number
ROOM Number of Beds
Number of Parking Lots
Price
TV set available?
A-2
Note that several entities seem to have the same attribute. In fact you should view this
as entities having different attributes that, by coincidence, have the same name.
The rule is that an attribute always belongs to one entity. Per entity the data type or
length of the attribute may be different.
Some of the presented answers are disputable. You could argue that a GUEST has an
Arrival Date when old arrival dates are not of interest and may be overwritten.
If you want to keep the number of beds per room, you do not need to keep it at the
level of the hotel, as you can derive the value based on the sum of all beds per room.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 1—3 Reading: Solution

Which text corresponds to the diagram?
EMPLOYEE assigned to DEPARTMENT

responsible for
A Each EMPLOYEE may be assigned to one or more DEPARTMENTS

Each DEPARTMENT must be responsible for one or more EMPLOYEES
B Each EMPLOYEE must be assigned to one or more DEPARTMENTS

Each DEPARTMENT may be responsible for one or more EMPLOYEES
C Each EMPLOYEE must be assigned to exactly one DEPARTMENT

Each DEPARTMENT may be responsible for exactly one EMPLOYEE
A-3
Option B is the correct reading.
.....................................................................................................................................................
Practice 1—4 Read and Comment: Solution
.....................................................................................................................................................
Practice 1—4 Read and Comment: Solution

1 Read each of the relationships in the model presented here.
2 Next, comment on the relationship you just read. Use your knowledge of people
and towns.
Practice: Read and Comment
PERSON born in TOWN

birthplace of
living in
home town of
visitor of
visited by
mayor of
governed by
(as mayor)
A-4
From top to bottom:

Each Person must be born in one or more Towns.
Each Town may be birthplace of exactly one Person.
Comment: both sides seem te be of the wrong degree.
Each Person must be living in one or more Towns.
Each Town may be home town of one or more Persons.
Comment: The first certainly has a wrong degree. The optionality seems wrong as
well.
Each Person may be visitor of one or more Towns.
Each Town must be visited by one or more Persons.
Comment: The optionality of the second seems very likely, but is probably wrong,
depending of the definition of Town.
Each Person may be mayor of exactly one Town.
Each Town may be governed by (as mayor) exactly one Person.
Comment: Both seem fine, except that if you need to keep historical information,
then both sides of the relationship must be “many”.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 1—5 Hotel: Solution

1 Comment on the relationships of the model presented here.
1 Possible comments:
• The relationship the lodging for/ in between entity STAY and HOTEL should
be replaced by a relationship between ROOM and HOTEL, such as located
in/location for.
• Every PERSON should have the possibility of more than one STAY; a STAY
should be for a single person only. When two persons sign in it should lead to
two instances of STAY.
• The relationship between PERSON and HOTEL seems to be redundant.
This is a possible model:
HOTEL
* Address
location for favored
located in owned known by
by by
ROOM
* Room Number
location for owner aware in favor
of of of
taking place in
STAY endured by PERSON

* Arrival Date enduring * Name
A-5
2 Make up two more possible relationships between PERSON and HOTEL that
might be of some use for the hotel business.
2 See the diagram for possible relationships between person and hotel.
.....................................................................................................................................................
Practice 1—6 Recipe: Solution
.....................................................................................................................................................
Practice 1—6 Recipe: Solution

1 Analyze the example page from Ralph’s famous Raving Recipes book and list as
many different types of information that you can find that seem important.
2 Group the various types of information into entities and attributes.
3 Name the relationships you discover and draw a diagram.
5DOSK·V5DYLQJ5HFLSHV
6RXSV $oRUGDDOHQWHMDQD
bread soup from Portugal
vegetarian
for 4 persons:
15 min
easy 1 onion
4 cloves of garlic
1 red pepper
1 liter of vegetable broth
4 tablespoons of olive oil
4 fresh eggs
1 handful of parsley or coriander
salt, pepper
9-12 slices of (old) bread
preparation Cut the onion into small pieces and fry together
with the garlic. Wash the red pepper, cut it in
half, remove the seeds and fry it for at least 15
page 127
A-6
Here is a list of types of information. There may be others and you may have
named them differently. An example from the text is given in italics.
Recipe (Açorda)
Original Name (Açorda alentejana)
Description (Bread Soup from Portugal)
Vegetarian Indicator (Vegetarian)
Number of Persons (4)
Ease of Preparation (Easy)
Preparation Time (15 min)
Preparation Description (“Cut the onion ...”)
Ingredient (Garlic, bread)
Quantity (4, for the cloves of garlic, 1 for the pepper)
Unit (liter, piece, handful)
Recipe Group (Soups)
.....................................................................................................................................................
.....................................................................................................................................................
Note that the ingredients are somehow sorted. There must be hidden intelligence
to do that. This might lead to:
Ingredient Type (vegetable, spice)
A possible model for the recipes:
BOOK
-Title a RECIPE must be classified in
containing exactly one RECIPE GROUP
published in A RECIPE GROUP may be
RECIPE classification for one or more
-Name RECIPES
-Description
-Vegetarian?
-Number of People classified
in RECIPE
-Preparation Time
GROUP
-Preparation Ease
-Preparation Text classification -Name
for
prepared
with
used
in
INGREDIENT
-Name a RECIPE must be prepared with
-Quantity one or more INGREDIENTS
-Unit An INGREDIENT may be used in
-Type
exactly one RECIPE
A-7
Other solutions may include entity AUTHOR, BOOK, BOOK TYPE, RECIPE,
RECIPE ITEM, INGREDIENT and INGREDIENT GROUP.
.....................................................................................................................................................
Practice 2—1 Books: Solution
.....................................................................................................................................................
Practice 2—1 Books: Solution

1 In this text the word book is used with several meanings. These meanings are
different entities in the context of a publishing company or a book reseller. Try to
distinguish the various entities, all referred to as book. Give more adequate names
for these entities and make up one or two attributes to distinguish them.
2 Create an ER model based on the text. Put the most general entity at the top of your
page and the most specific one at the bottom. Fit the others in between. Do not
worry about the relationship names.
Different meanings of the word Book from the text. Attributes are between
brackets. Attributes marked with an asterisk should preferably be modeled as a
relationship to another entity.
MANUSCRIPT (Title, Year of Completion, Author*)

EDITION (Sequence Number of Printing)
COPY (position on the shelf)
TRANSLATION IN LANGUAGE (Language*, Translator*)
TITLE (Authorized Text)
PRINTING (Font*, Year of Printing)
VERSION (Hardcover Indicator)
and maybe more. These may be related in this way:
MANUSCRIPT LANGUAGE
TRANSLATION
TITLE
EDITION
PRINTING
VERSION
COPY
A-8
A particular copy of a book is a copy of a version of a printing of an edition of a

title of a translation in a language of a manuscript.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 2—2 Moonlight: Solution

1 Make a list of about 15 different entities that you think are important for Moonlight
Coffees. Use your imagination and common sense and, of course, use what you
find in the summary that is printed below.
1 Possible entities sorted alphabetically:
COUNTRY PRODUCT
COUNTRY COMMUNITY PRODUCT GROUP
CURRENCY PRICE
DEPARTMENT SALE
DRINK SHOP
EMPLOYEE STOCK OPTION
EXCHANGE RATE STOCK PRICE
FOOD TICKER SYMBOL
JOB
LOCATION TYPE
PASTRY
A-9
2 Write a formal definition of the entity that represents:

2 Possible definition:
A shop is a place where Moonlight sells, did sell, or will sell products to
customers, a retail outlet. An example of a shop is the one in ... that was opened in
... .
An employee is a person that works, did work or will work for Moonlight in a
named job and is paid by Moonlight. An example of an employee is ... who works
as ... in ... .
.....................................................................................................................................................
Practice 2—3 Shops: Solution
.....................................................................................................................................................
Practice 2—3 Shops: Solution

Use the information from the list as a basis for an ER model. Pay special attention
to find all attributes.
SHOP in COUNTRY
* Number * Code
* Name of * Name
* Location
* City located
o Telephone in LOCATION TYPE
* Open Date * Description
of
or COUNTRY
* Code
* Name
SHOP
* Number of
* Name in
located located LOCATION
o Telephone in in TYPE
LOCATION
* Open Date * Name * Description
of of
A-10
Both solutions are fine. The second sees a location like JFK Airport as a possible
location for several shops. Maybe most locations have one shop only.
Note that in the second solution a location does not necessarily have a shop. This
allows for future locations. In this respect the second model is somewhat more
flexible.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 2—4 Subtypes: Solution

Find all incorrect subtyping in the illustration. Explain why you think the
subtyping is incorrect. Adjust the model to improve it.
All examples are cases of wrong subtyping.
The subtypes of DISABLED PERSON are not mutually exclusive: DEAF and
BLIND may overlap.
The subtypes of CAR ar not exhaustive: there are cars that are neither a station
wagon nor a sedan.
Entity BUILDING should have at least two subtypes.
If ROOM WITH BATH is a subtype of HOTEL this would mean that every
ROOM WITH BATH is a HOTEL.
DOMESTIC ANIMAL and MAMMAL are not mutually exclusive. In fact, DOG
is subtype of both, instead of the other way around.
Improvements
Solution: Subtypes
DISABLED CAR
PERSON
STATION WAGON
DEAF AND BLIND
SEDAN OTHER
DEAF, NOT BLIND CAR
BLIND, NOT DEAF
OTHER DISABLED
BUILDING HOUSE OTHER
HOTELROOM DOG
ir
ROOM WITH BATH DOMESTIC
r epa
nd
ANIMAL
yo
beMAMMAL
OTHER ROOM
A-11
.....................................................................................................................................................
Practice 2—5 Schedule: Solution
.....................................................................................................................................................
Practice 2—5 Schedule: Solution

Use the schedule that is used in one of the shops in Amsterdam as a basis for an
entity relationship model. The schedule shows, for example, that in the week of 12
to 18 October Annet B is scheduled for the first shift on Monday, Friday, and
Saturday.
SHOP
* Name
* Location
* City
with
for
SCHEDULE prepared
by EMPLOYEE
* Start Date
* Family Name
* End Date maker of
* First Name
with * Short Name
within
SCHEDULE for
ENTRY
* Weekday with
* Shift Number
A-12
An alternative solution models SCHEDULE ENTRY differently:
with
within
SCHEDULE for
ENTRY
with
* Monday Shift
* Tuesday Shift
* Wednesday Shift
* Thursday shift
* Friday Shift
* Saturday Shift
* Sunday Shift
The first model does allow for more than one shift per employee per day; the
second does not.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 2—6 Address: Solution

An entity, possibly PERSON (or ADDRESS) may have attributes that describe the
address as in the examples below.
1 How would you model the address information if the future system is required to
produce accurate international mailings?
PERSON
Address text line1
1 Address text line2 COUNTRY
Address text line3
Address text line4
PERSON
Street or “PO Box” Indicator ADDRESS
Street TYPE
2
House Number
City
Post Code
COUNTRY
Province or State
1 Solution 1 can easily cope with the various address formats. It simply
recognizes the fact that an address may consist of 4 lines of text at maximum.
This avoids the need to interpret a given address. Is 1045 ST 88 the house
numbered 1045 on Street 88 or is it the house numbered 88 at 1045 Street?
2 Solution 2 cuts the address into individual pieces and accounts for post boxes
as well. It assumes the interpretation is no problem. By adding entity
ADDRESS TYPE the sequence of display of the various address pieces can be
set.
2 Would your model from the previous practice also accept the addresses below?
3 Check if your model would be different if the system is also required to have
facilities to search addresses in the following categories. Make the necessary
changes, if any.
All addresses:
• In Kirkland
• With postal code 53111 in Bonn
• That are P.O. Boxes
• On:
– Oxford Road or
.....................................................................................................................................................
Practice 2—6 Address: Solution
.....................................................................................................................................................
– Oxford Rd or
– OXFORD ROAD or
– OXFORD RD
in Reading
3 The second model does allow most of the required queries, although the last
one may need some trials. If a system needs to allow quick and easy searches
on the address information, often a separate attribute is added that stores the
address information in uppercase using a particular set of rules to sort street
name elements and to abbreviate.
For example, a street name like “President A. B. Collins Road” would be
taken as COLLINS, RD PRES A B. Such an attribute is not derivable from
the address line unless the system has a built-in parser for this purpose.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 3—1 Read the Relationship: Solution

Read the diagrams aloud, from both perspectives. Make sentences that can be
understood and verified by people who know the business area, but do not know
how to read ER models.
Every ALU must be of

ALU of BRY
exactly one BRY
with Every BRY may be with
one or more ALUS
Every PUR may be bazooned in

PUR bazooned in YOK
one or more YOKS
Every YOK may be bazooned by bazooned by
one or more PURS
KLO bilought in Every KLO must be bilought in one or

HAR
more HARS
glazoed with Every HAR may be glazoed with
exactly one KLO
A-15
Even if the phrases sound as if they come from some strange Science Fiction
movie, make sure you are fluent in reading them. Only start judging what you
have just heard after reading the lines. In a real life context, you sometimes read
what you think is represented, rather than what is actually modeled. It is really
hard to switch off your interpretation machine.
.....................................................................................................................................................
Practice 3—2 Find a Context: Solution
.....................................................................................................................................................
Practice 3—2 Find a Context: Solution

Given the following ER diagrams, find a context that fits the model.
Here are just some examples of possible contexts.
COUNTRY
birthplace of PERSON
born in
location of TOWN
located in
PERSON PERSON
MALE
FEMALE mother of MALE FEMALE child
son of of
mother of
most popular
movie star in COUNTRY
PERSON
with most popular
born in movie star
birthplace of
A-16
.....................................................................................................................................................
.....................................................................................................................................................
Practice 3—3 Name the Intersection Entity: Solution

1 Resolve the following m:m relationships. Find an acceptable name for the
intersection entity.
2 Invent at least one attribute per intersection entity that could make sense in some
serious business context. Give it a clear name.
PRODUCT DEPARTMENT
STORE
in
with
of at
SALE
* Date
PERSON SAILBOAT
in INTERPRETER LANGUAGE
with
of at
with
CREW with
MEMBERSHIP of in
* Role FLUENCY
* Score
A-17
These are only possible solutions, based on just one possible context and one
assumption on desired functionality.
.....................................................................................................................................................
Practice 3—4 Receipt: Solution
.....................................................................................................................................................
Practice 3—4 Receipt: Solution

Use the information from the receipt and make a list of entities and attributes.
Served by: Dennis

Till: 3 Dec 8, 4:35 pm
-----------------------
CAPPUCC M 3.60
* 2 7.20
CREAM .75
* 2 1.50
APPLE PIE 3.50
BLACKB MUF 4.50
<SUB> 16.70
tax 12% 2.00
<TOTAL> 18.70
=======
CASH 20.00
RETURN 1.30
-----------------------
Hope to serve you again
@MOONLIGHT COFFEES
25 Phillis Rd, Atlanta
A-18
An assumption is that all, non-derivable, information that is displayed on the

receipt must be stored for at least some time.
COUNTRY
* Name
with
in
SHOP
* Name
o VAT%
with
in
TILL PERSON
* Number * First Name PRODUCT
with with * Short Name
at at * Price
SALES HEADER in
* Date & Time
* Amount Cash
referring
with
part of to
SALES ITEM
* Quantity
A-19
.....................................................................................................................................................
.....................................................................................................................................................
This model is based on several assumptions. One is that all entities may exist in
isolation: for example, there may be countries (recorded in the system) without
shops, shops without tills, and tills without sales. This explains all 1:m
relationships that are optional at the 1 side.
Sales tax or VAT percentage for these kinds of products is a fixed value at a local
shop level, such as in, for example, the USA, or at Country level, such as for
example in many European countries. In this model the choice is made to model it
at shop level.
Many countries know two VAT percentages: a low percentage for normal and a
high percentage for luxury goods. Moonlight may have to deal with both as
mineral water and coffee possibly fall in different categories.
(The day after writing this text the Dutch parliament accepted two VAT
percentages for one and the same product—drinking water. The initial amount is
seen as normal consumption, the excess as luxury.)
The Attribute Price of PRODUCT could be modeled as a separate entity as
products may have various prices over time.
.....................................................................................................................................................
Practice 3—5 Moonlight P&O: Solution
.....................................................................................................................................................

1 Create a entity relationship model based on the following personnel and
organization information:
1 Note the optional relationships from EMPLOYEE to DEPARTMENT and
SHOP. These result from “... country organizations or for a shop.”
All Moonlight Coffee employees work for a department such as

“Global Pricing” or “HQ”, or for a shop. All employees are at the
payroll of one of our country organizations. Jill, for example, works as
a shop manager in London; Werner is a financial administrator
working for Accounting and is located in Germany …
DEPARTMENT SHOP COUNTRY

* Name * Name ORGANIZATION
* City
with
with with
works works
for for at payroll of
EMPLOYEE
* Name
* Job
A-20
.....................................................................................................................................................
.....................................................................................................................................................
2 Extend or modify the diagram based on this information:

2
DEPARTMENT report of
* Name HQ
OTHER reporting to
DEPARTMENT COUNTRY
ORGANIZATION
… All shops belong with
to one country with with
belongs
organization (“the to
countries”). There is
SHOP
only one country
* Name
organization per
* City on
country. All countries
works with payroll
and departments
for works for of
report to HQ, except
HQ itself … EMPLOYEE
* Name
* Job
A-21
3 And again:
3
DEPARTMENT
HQ report of
* Name
OTHER reporting to
DEPARTMENT COUNTRY
ORGANIZATION
with with
… Employees can
work part time. with
Lynn has had an at belongs to
80% assignment payroll
of SHOP
for Product
* Name
Development since EMPLOYEE
* City
the 1st September. * Name
with
Before that she had with
a full-time position. to of to
ASSIGNMENT
* Job
* Start Date
* Part Time %
A-22
.....................................................................................................................................................
.....................................................................................................................................................
4 Change the model—if necessary and if possible—to allow for the following new
information.
a Jan takes shifts in two different shops in Prague.
4 a. Does not require changes in the model. People can be employed in various
places—the model allows that.
e To prevent conflicting responsibilities, employees are not allowed to work for
a department and for a shop at the same time.
b. See diagram.
b Last year Tess resigned in Brazil as a shop manager and moved to Toronto.
Recently she joined the shop at Toronto Airport.
c. See diagram.
c To reduce the number of direct reports, departments and country organizations
may also report to another department instead of Headquarters.
d. See diagram.
d The shops in Luxembourg report to Belgium.
e. Cannot be modeled with what you have learned so far.
4a: -
COUNTRY 4c DEPARTMENT
* Name HQ
report of
ORGANIZATION
4b
OTHER
with DEPARTMENT
EMPLOYEE
with
for for COUNTRY in COUNTRY
PAYROLL ORGANIZATION * Name
ENTRY of
* Start Date with of
4d located
belongs to in
SHOP
4e: - * No
* Name
A-23
5 Would your model be able to answer the next questions?

a Who is currently working for Operations?
b Who is currently working for Moonlight La Lune at the Mont Martre, France?
5 a, b. Given the previous models, these questions can be answered.
.....................................................................................................................................................
.....................................................................................................................................................
c Are there currently any employees working for Marketing in France?

c. Departments are not related to Countries—so there is no Marketing in
France. However, we can find who (if anyone) is working for the Marketing
Department who is also on the French payroll.
d What is the largest country in terms of number of employees? In terms of
managers? In terms of part-timers?
d. The question can be answered by counting the distinct current assignments
per country organization.
e When can we celebrate Lynn’s fifth year with the company? When can we do
the same with Tess’ fifth year with Moonlight?
e. The model does not (yet) contain an attribute End Date for ASSIGNMENT
and PAYROLL ENTRY. Given Tess’ career at Moonlight, there is a need for
attribute End Date as the periods are not necessarily contiguous.
f What country has the lowest number of resignations?
f. Can be answered.
Note:
This model is quite complex, as both departments and country organizations
are part of the organization. Departments do not have a location, unlike
country organizations. Departments are ‘virtual’ organizations. Think about
a department as a collection of tasks rather than a group of people. A
department can still exist even if there are no people working for it - the tasks
still remain. The location of the department may be defined by ‘the office of
that country organization which currently has the manager of the department
on its payroll.’
A country organization is not the same as a country. It is wise to distinguish

the two as countries are beyond the scope of Moonlight’s power, whereas
country organizations can be defined by Moonlight. A country organization
in Haiti may not (yet) exist, but the country does.
.....................................................................................................................................................
Practice 3—6 Price List: Solution
.....................................................................................................................................................
Practice 3—6 Price List: Solution

Make a ER model based on the pricelist from one of the Moonlight Coffee Stores.
COUNTRY
* Currency
with coffees
in teas
PRODUCT
SHOP foods
GROUP
* Name nonfoods
with supplements
* Address in
* City
PRODUCT
with
at * Short Name
o Size
PRICE LIST
* Start Date with
* End Date
with
in of
PRICE
* Amount
A-24
One of the questions that arises here is the following: is decaffeinated a

PRODUCT? In some respects it behaves differently from other products—you
cannot buy it separately, nor in combination with, for example, apple pie. On the
other hand it is something you can buy, and it has a price just like the other
products.
In this solution, decaffeinated is regarded as a product in a special product
group: supplements (maybe you have a suggestion for a better name...). Products
in this group need special attention, but otherwise can be treated as products.
Note also the entity PRODUCT GROUP. PRODUCT GROUP instances are
represented by the lines between the various products on the price list.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 3—7 E-mail: Solution

Take the given model as starting point. Add, delete, or change any entities,
attributes, and relationships so that you can facilitate the following functionality:
See the numbers in the illustration for the various assignments.
1 A user must be able to create nick names (aliases) for other users.
2 A folder may contain other folders.
3 A user must be able to forward a composition. A forward is a new message that is
automatically sent together with the forwarded message.
4 All folders and lists are owned by a user.
containing
USER owner of NICK
NAME within FOLDER
1 owned
referred
by 2
to
for
USER
4
forwarded owner owner
with of of
containing owned owned
COMPOSITION
3 by by
is within
forward FOLDER LIST
of
A-25
.....................................................................................................................................................
Practice 3—7 E-mail: Solution
.....................................................................................................................................................
Challenge:
5 A mail list may contain both users and other lists.
6 A mail list may contain external addresses, like “giovanni_papini@yahoo.com”.
7 A nickname may be an alias for an external address.
owner owner
USER USER
of of
5 owned 6 owned
by by
LIST LIST
owner is referred
is referred is referred owner
of to is
to to of
owned referred to
referring referring referring owned
by referring
to to to by
to
LIST ITEM
LIST ITEM
o External User
owner of NICK NAME

USER
owned by o For External User
7 referred
to
for
A-26
.....................................................................................................................................................
.....................................................................................................................................................
Practice 3—8 Holiday: Solution

Comment on the model given below that was based on the scenario text.
“Paul and I hiked in the USA. Eric and I hiked in France and
we rented a car in the USA last year”.
COUNTRY TRANSPORT
France Boots
USA Boots
USA Car
RT
C TRANSPORT
F r OU COUNTRY
O
US n NT
a
C ots SP
US A ce RY
Bo AN
A
TR
C
Er OM
Bo ar
s
COMPANION
ot
ON
i
Er c PA
Pa c i NI
NI
O
E c PA
ul N
Er OM
C
i
Pa ric
ul
A-27
When you join the information in COUNTRY-COMPANION table and

COMPANION-TRANSPORT table, this is the result:
COUNTRY COMPANION TRANSPORT

France Eric Boots
France Eric Car
USA Eric Boots
USA Eric Car
USA Paul Boots
This is obviously wrong. The model somehow generated more than was put into
it!
The reason? It is the combination of Paul, Hiking, USA that is important, just as
the combination of Eric, Hiking, France is important. It is the combination of the
three values that matter. See the model and table structure in the next illustration.
.....................................................................................................................................................
Practice 3—8 Holiday: Solution
.....................................................................................................................................................

France Eric Boots
USA Paul Car
HOLIDAY

France Eric Boots
USA Eric Car
USA Paul Boots
A-29
.....................................................................................................................................................
.....................................................................................................................................................
Practice 4—1 Identification Please: Solution

Describe how you would identify the following entities, making up any attributes
and relationships you consider appropriate.
A city
Name and Country is not always sufficient. In some countries you would also need
Province or State when there are more cities with the same name. Some countries
use a postal code that is unique per town or city.
Of course geographical identification, like Latitude 4 East, Longitude 53 North is
unique, but only practical in specific situations.
A Contact Person for a Customer

This is within our reach. Name, Job (or Telephone Number) and Customer is
probably enough. Or, maybe, just an Id.
A Train
If you think of the hardware, trains have a Id and often also a unique Name.
If you think of the 9:05 from Paris to Madrid, the identification consists of the
Time, the Start, and Destination. The railroad company uses a simple Code for
internal use.
A Road
In most countries the important roads have a Name. This name may be a number
or a combination of characters and numbers, like “Route 69” or “N356”, but this
is still a name. Unimportant roads may be unnamed. These can only be identified
by using geographical information, like both Ends. Ends may be given as a City
or as geographical coordinates.
A Financial Transaction
Use the combination of the From Account, the To Account, the Date and Time.
Note that you need Time because there may be several transactions with the same
From and To Account on the same Date.
An Academy Award
A person would receive an Oscar in a particular year because of a particular
reason (best supporting actress, best music score) that is related to a movie, or for
all of their activities in the movie world.
An Oscar is identified by the Person Receiving, the Year, the Movie (optional), and
the Reason.
.....................................................................................................................................................
Practice 4—1 Identification Please: Solution
.....................................................................................................................................................
A Painting
This is a difficult one. Although probably all famous paintings have a title or
name, not all paintings can be identified by Title and Painter and (optional) the
Year or Date. Paintings in art collections are often identified indirectly by an Id on
a sticker at the back of the painting, plus electronic photographs and descriptive
texts. The latter provide a possible check that stickers have not been switched.
Paintings are often not simply identified and also often simply not identified.
A T.V. show
This is also a tricky one from the perspective of a consumer. Often a show has a
Name and a Sequence Number (or Broadcast Date). Often Names are so common,
for example, News, that the Television Company is added. This, however, is not
enough, for example, when there is a rerun of the show by another company.
For the production company a T.V. show is probably just a project with an Id.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 4—2 Identification: Solution

Are the entities in the next diagrams identifiable?
A B C
1 # Xx * Yy # Zz
A B
C # Id
2 # Code
A
* Xx
3 B C with D
# Yy # Zz # Id
of
A-30
1 Because every A is identifiable by attribute Xx, every B and C are as well.

2 B is identifiable by Id. A is identifiable, only if B and C are. This leaves C.
Because of the arc, not every C will have the barred relationship with B. But
that leaves C’s unidentified.
3 D is identifiable by Id. Every C is identifiable by ZZ and D. Every B is
identifiable by Yy. As every B and every C is identifiable, so is every A.
.....................................................................................................................................................
Practice 4—2 Identification: Solution
.....................................................................................................................................................
Solution: Identification 1
4 5
P Q P
# Id # Name
PERSON
son of
MALE FEMALE
# Seqno mother of # Name
# Birth Date
6 partner in partner in
with husband with wife
MARRIAGE
# Start Date
A 31
4 Every Q is identifiable by Id. P’s that are related to an instance of Q are

identifiable by that instance, but not every P is related.
5 Conceptually there is no problem here. Every P is identified by its Name and
the reference to its “predecessor”. You could argue about how to deal with the
first P, but the question in return would be: who told you there is a first P?
More of a problem is the fact that the identification of an instance of the n th
generation takes n+1 names. This in particular makes this type of
identification highly impractical.
Note: the next model describes a context that may be different from the world you
are familiar with.
6 Entity FEMALE is identified by Name and Birth Date. This may not be true
for the entire population of the world, but may be true for a much smaller set
of people a business is interested in.
Entity MALE is identifiable as well: I am the only second son of my mother.
Every PERSON is identifiable because every person is female or male. Entity
MARRIAGE is identifiable by the combination of both MALE and FEMALE
that participates and the Date of the marriage.
7 Given the above model, answer the following questions.
7
a Can person A marry twice?
a Yes
b Can person A marry twice on the same day?
b Yes
.....................................................................................................................................................
.....................................................................................................................................................
c Can person A marry with person B twice?

c Yes
d Can person A marry with person B twice on the same day?
d No
e Can person A be married to person B and person C simultaneously?
e Yes
f Can person A be married to person A?
f No, because a marriage is always between a male and a female and person.
A cannot be both male and female as these are distinct subtypes of person.
.....................................................................................................................................................
Practice 4—3 Moonlight UID: Solution
.....................................................................................................................................................
Practice 4—3 Moonlight UID: Solution

Use what you know about Moonlight Coffees by now, and, most importantly, use
your imagination.
1 Given the model below, indicate UIDs for the various entities. Add whatever
attributes you consider appropriate. Country organizations have a unique “tax
registration number” in their countries.
2 Are there any arcs missing?
DEPARTMENT HQ
# Name report of
OTHER reporting to
COUNTRY
DEPARTMENT COUNTRY in # Code
with with ORGANIZATION
EMPLOYEE of
# Id with
with
located
with assigned to in
for for SHOP
of PAYROLL # No
ENTRY
# Start Date
with
to for to
ASSIGNMENT as JOB
# Start Date # Title
in
A-32
For transparency reasons, only the attributes that are needed for identification
are added in this model.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 4—4 Tables: Solution

Read the text on ISO Relational tables.
Do a quality check on the ER model based on the quoted text and what you know
about this subject. Also list constraints that are mentioned in the text but not
modeled.
Erroneous:
• Both relationships from FOREIGN KEY to TABLE should not be barred.
• A Table is not identified by Name only, but also by the owning USER. The
same is true for a KEY and a FOREIGN KEY.
• Every KEY, PRIMARY, UNIQUE and FOREIGN must be named uniquely
per USER.
These errors have been removed in the diagram below.
USER
owner of owner of owner of

in in in
from with KEY
FOREIGN KEY TABLE
# Name # Name # Name
with
to PRIMARY
of
referenced with
with in UNIQUE
of
with
with
for in
from for
ASSOCIATION in COLUMN
# Name in USAGE
# Seqno # Seqno
in * Data Type of
to o Not Null
A-33
Not shown:
• COLUMN in a KEY must be a COLUMN of the TABLE the KEY belongs to.
• COLUMNS in a FOREIGN KEY must refer to COLUMNS in a KEY of the
TABLE the FOREIGN KEY refers to.
These constraints cannot be shown and must be listed separately.
.....................................................................................................................................................
Practice 4—5 Constraints: Solution
.....................................................................................................................................................
Practice 4—5 Constraints: Solution

Change the diagrams to model the constraint given.
1
EMPLOYEE
# Name OTHER managed
CEO EMPLOYEE by
2 ALIAS
manager # Name
of USER owned LIST
# Name by
owner of NICKNAME
3
with
subfolder USER
within
FOLDER # Name
# Name owner
of
owned
by
A-34
1 Every EMPLOYEE must have a manager, except the Chief Executive Officer.
1 By creating a CEO subtype of EMPLOYEE the constraint is easily modeled.
You may argue the use of this CEO subtype if modeling the constraint is the
only reason for its existence.
2 A user may not use the same name for both NICKNAME and LIST name.
2 LIST and NICKNAME share the same namespace.This can be modeled with
a supertype. Note the repositioning of attribute Name.
3 A top level FOLDER must have a unique name per user; sub folders must have a
unique name within the folder where they are located.
3 Adding the arc and barred relationships is enough. Note that the recursive
relationship from FOLDER to FOLDER is made mandatory at the many end.
See also the remark at Practice 4- Identification 5: Solution about recursive
identification.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 5—1 Shift: Solution

List the various date/time elements you find in this Shift scheme and make a
conceptual data model.
CALENDAR DAY
SHOP # Date
* Name * Public Holiday Indicator
with of starting
on starting on
WEEKDAY
# Code SHIFT
* Name SCHEDULE
with with or on
on on
SHIFT
SHIFT SCHEDULE
# No * Start Date
* Start Time with
* End Time
A-35
In this solution, entity CALENDAR DAY is used. This might be sensible in the
context of a resource-planning environment where public holidays usually play a
role. Be aware of the fact that the dates of public holidays are completely
different around the world. In other terms, Public Holiday Indicator is not single
valued in a international context.
In a global context the issue of daylight saving time (which is not common and
certainly does not start at the same date around the world) and TIME ZONE
play a role.
.....................................................................................................................................................
Practice 5—2 Strawberry Wafer: Solution
.....................................................................................................................................................
Practice 5—2 Strawberry Wafer: Solution

Revisit your model and make changes, if necessary, given this extra information.
COUNTRY
# Code
with with
PRODUCT GROUP in
# Name with SHOP
with # No
in * Name
* Address
PRODUCT of * City
for GLOBAL LOCAL
PRICE LIST # Code # Name
# Start Date o Size
* End Date with
with with with of

LOCAL PRICE LANGUAGE
in of # Start Date # Code
o End Date
GLOBAL PRICE of
of * Amount
* Amount in
PRODUCT NAME
* Name
A-36
In this model several minor and major changes were made, compared to the
model that was presented earlier (Practice 3 - Price List: Solution).
The major change is the issue of the prices that are either determined on national
level or at shop level. This has lead to two price entities: GLOBAL and LOCAL
PRICE. Note the different attributes and relationships of these two.
The price list also shows the same product range as the earlier one, but now
products can be named in various languages.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 5—3 Bundles: Solution

1 Modify the product part of the model in such a way that the desired calculations
can be completed.
PRODUCT PRODUCT
1 GROUP 2 GROUP
BUNDLE PRODUCT PRODUCT

OTHER
BUNDLE
PRODUCT
with in
of referring with in
BUNDLE ITEM to of referring
* Quantity Needed BUNDLE ITEM * to
PRODUCT
Quantity Needed
GROUP
3
PRODUCT
ASSEMBLY ITEM
* Quantity Needed
A-37
1 The first model is probably what you came up with first. A bundle consists of
several products. You model this using the BUNDLE ITEM entity. But this
solution is not correct as a bundle can consist of another bundle as well, such
as SuperSweetTreat. Model 2 takes care of that. A BUNDLE ITEM can now
consist of one or more instances of BUNDLE and instances of OTHER
PRODUCT.
Note that model 2 assumes there is always a clear distinction between a
product and a bundle, because they are modeled as subtypes. Model 3 does
not assume that. A bundle is treated as any product here. You do not need to
worry if you should consider apple pie with whipped cream as a bundle or as
a single product.
.....................................................................................................................................................
Practice 5—3 Bundles: Solution
.....................................................................................................................................................
2 Change the model in such a way that it allows for:
DecafPunch (DP) = {Coffee (C) or Tea (T)} and {Blackberry Muffin (BM)}
Consider this as:

PRODUCT
AS1 = (C or T)
GROUP
DP = (AS1 and BM)
# Name
PRODUCTS
PRODUCT Code And_or Pg_name
# Code C ..
o And/Or Indicator T …
AS1 OR ….
DP AND …..
ASSEMBLY ITEM
ASSEMBLY ITEMS
* Quantity Needed
Prod_code Using_code Quantity
AS1 C 1
AS1 T 1
DP AS1 1
DP BM 1
A-38
2 This is a tricky one. You can regard a DecafPunch as a product group with
two products: DecafPunch Coffee and DecafPunch Tea, both with the same
price (because of the or). This point of view is already covered in the
PRODUCT GROUP / PRODUCT model.
In the solution presented, this aspect is ignored. This solution focuses on how
you can model the and or or type of constructions.
You see that DecafPunch can be regarded as a series of elementary groups that
are connected with and or or. An elementary group is seen as a group that
consists of either a single product or a product pair that is connected with an
and or or. Products can always be described this way (this follows from
mathematical logic).
See the table structure to check that the model works.
Note This model complexity may not make much sense in the context of
Moonlight. However, this type of model certainly does make sense for the bill
of material structures many production companies use for storing the
composition information of their products, with alternative compositions and
suchlike.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 5—4 Product Structure: Solution

1 Create a model for a product classification structure.
LEVEL1 fixed number

1 of levels with
LEVEL2
PRODUCT
CLASS
LEVEL3 # Name within
or generic:
LEVEL4
2 + coffees
+ teas
LEVEL5 + foods
product
+ nonfoods
+ supplements
PRODUCT + bundles
A-39
1 When the number of levels is known and fixed, the left model can be used.
Note that the model forces you to use all five levels for every product. To avoid
this you could create a relationship from PRODUCT to all LEVEL entities,
plus an arc across all.
The left model gives a lot of freedom. No limits to the minimum or maximum
number of levels. Freedom to change the classification during the lifetime of
the system.
2 (Optional) How would you treat the bundled products?
2 The problem with the bundles is that, strictly speaking, they cannot be
classified in class Drinks nor Foods as they can consist of both. Probably the
easiest way to handle this is to regard a bundle as an instance of PRODUCT
CLASS.
.....................................................................................................................................................
Practice 6—1 Patterns: Solution
.....................................................................................................................................................
Practice 6—1 Patterns: Solution

What pattern do you expect to find in the given contexts? If you do not see it, make
a quick sketch of the model. Use your imagination and common sense.
• Model of moves in a chess game Chain or Network

• Model of quotations M/D or Basket
• Model of recipes Bill of Material
• Model of all people involved in Roles
college: students, teachers,
parents, …
• Rentals in a video shop M/D or Basket
• Model of phases in a process Chain or Network
Moves in a Chess Game

A MOVE moves a chess piece from one FIELD to another. This is a typical
Network pattern. You can also see the MOVES as a Chain of ordered steps in the
chess game.
Quotations
A quotation makes an offer to deliver a number of products of a certain quantity.
This structure is similar to a ORDER and will have the Master-Detail or the
Basket pattern.
Recipes
A recipe usually describes how to make something: in order to create a ... you
need two ..., one ... and so on. This is typical Bill of Material pattern.
People Involved in a College

Pattern: Roles. A teacher can also be parent of a student, a parent can be an
former student and so on.
Rentals in Video Shop

A RENTAL is an event such as a sale or a quotation: Master-Detail or the Basket
pattern.
Phases in a Process
PROCESS and PHASE have a Master-Detail pattern. Be careful at this point. The
crucial thing about phases in a process is that the phases are usually ordered.
The phases may form a Chain if there is always one phase at most that precedes
and follows a particular phase.
If, on the other hand, the phases can be preceded and followed by any number of
phases (maybe even including itself) than the structure would be a Network.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 6—2 Data Warehouse: Solution

1 Check the Moonlight models you created so far. Do they cater for answering the
listed questions. If not, make the appropriate changes.
1 No formal solution.
YEAR
# No
with with
in
QUARTER
with assigned
in with
to
MONTH WEEK PRODUCT within
# No # No CLASS with
with with with
in in of GEOGRAPHY
DAY PRODUCT * Number of Inhabitants within
# No # Id * Number of Employees
with in with
of of of
SALES VOLUME
* Quantity
* Value in Local Currency
* Value in $
A-41
2 For a data warehouse data model, suggest the central “facts” entity.
2 The diagram answers more than the practice asked for. The central entity is
SALES VOLUME.
This is a typical snowflake model, rather than a star model. All three
dimensions, time, geography, and products, have their own internal structure.
Note that a week is not assigned to a month. Weeks are assigned to a year,
even though not all weeks fit exactly into a calendar year. Weeks have a fixed
length, like day and year, but behave differently as a week is not based on an
astronomical fact.
The product structure we defined earlier can be used here. The distinction
between local and global products has gone.
Geography is the generic word for an area (such as a part of country or a
country or a continent) that is of interest.
.....................................................................................................................................................
Practice 6—3 Argos and Erats: Solution
.....................................................................................................................................................
Practice 6—3 Argos and Erats: Solution

Make a conceptual data model based on the information in the text. Mark all the
pieces in the diagram that can be confirmed from the text.
ERAT
# Name
unknown
have have have have optionality
belongs unknown
of of to of
degree
RONDEL ARGO UBIN
* Type # Name
in in with
with with with
UBIN ITEM
Constraint not shown:

A ubin always consists of one or more argos of the erat, one or more...
A-42
Not all information you need is present in the given text. You may be surprised
how much actually is still left to check.
The only part you can be sure about is the UBIN - UBIN ITEM to model the
given “A ubin always consist of one or more ...”. This is a typical basket
construction.
Note: this text was created by using some find/replace commands in a normal
text. Ask your instructor about the original text.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 6—4 Synonym: Solution

Make a conceptual data model that could be the basis for a dictionary of
synonyms.
MEANING
# Id
* Description synonym
shared by
with
WORDS
WORD Word Mng_id
# Word Practice 1
Exercise 1
MEANINGS Order 2
Id Description Sequence 2
1 Action, actual doing ... Arrangement 2
2 Regular arrangement ... Order 3
3 Order, command ... Command 3
4 A vehicle with two wheels ... Demand 3 homonym
5 Bike 4
The solution is, probably, surprisingly and frustratingly easy.

Think of synonyms as words that share the same meaning. Do not confuse the
concept meaning with the concept word. Although you express a meaning using a
word or words, it is not the same as a word. Words are representations of
meaning.
Note that there is no entity SYNONYM. A synonym is something circumstantial:
if there is a MEANING shared by two or more words, then there is a synonym.
Conversely, if there is a WORD with two or more MEANINGS, there is a
homonym.
.....................................................................................................................................................
Practice 7—1 Mapping Supertype: Solution
.....................................................................................................................................................
Practice 7—1 Mapping Supertype: Solution

1 What would have been the rationale of this choice?
1 Possible considerations:
– The number of departments is probably not large, say 100 or less.
– There are only few specific attributes for the subtypes.
– One subtype, HQ may have only one instance.
– HQ seems to be modeled only to show the fact that all departments except
HQ must report to a department.
2 On the table diagram, name all the elements that must be created following this
supertype implementation. Use the naming convention as described in this lesson,
or use your own rules. Give proper names to the columns and foreign key
constraints and identify check constraints, if any.
2 See the diagram below.
dpt_dpt_fk
DEPARTMENTS (DPT)
pk * Id
* Name
* Dpt_type
* Headcount
o Address
fk o Dpt_id_reporting_to
Note the optional Address column. Note also the optional foreign key column and
its name. Following the rules, the name could have been Dpt_id. As many people
would assume that to be a primary key column name, the name is followed by the
name of the relationship that the foreign key represents.
Two constraints are needed:
• To guard the subtype:
( Dpt_type = ’HQ’
and Address is not null
and Dpt_id_reporting_to is null)
or
( Dpt_type <> ’HQ’
and Address is null
and Dpt_id_reporting_to is not null)
• To guard the hierarchical structure of the organization.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 7—2 Quality Check Subtype Implementation:

Solution
Perform a quality check on the proposed subtype implementation of entity
PRODUCT.
gpt_pgp_fk lpt_shp_fk
GLOBAL_PRODUCTS (GPT) LOCAL_PRODUCTS (LPT)

pk * Code pk * Name
o Size pk,fk * Shp_no
fk * Pgp_name fk * Pgp_name
Errors:
GLOBAL_PRODUCTS
• The foreign key column Pgp_name is not marked with fk.
LOCAL_PRODUCTS
• Column Shop_no should be named Shp_no
• Column Shp_no is also part of the primary key.
• The foreign key to table SHOPS should be named lpt_shp_fk
.....................................................................................................................................................
Practice 7—3 Quality Check Arc Implementation: Solution
.....................................................................................................................................................
Practice 7—3 Quality Check Arc Implementation: Solution

Perform a quality check on the proposed supertype and subtype implementation of
the entity PRODUCT and its subtypes. Also, check the selected names.
Errors:
PRODUCTS
• Gpt_code and Lpt_name are both part of a unique key uk1
• The last characters of the names of the foreign keys should all read _fk
• A fk column is missing as the primary key of LOCAL_PRODUCTS consists
of two columns. Column Lpt_shp_no must be added. This column is part of
fk3 and part of uk2.
GLOBAL_PRODUCTS
• This is an awkward one:
The foreign key line in the diagram must be removed as
GLOBAL_PRODUCTS does not have its own foreign key to
PRODUCT_GROUPS.
LOCAL_PRODUCTS
• Column Pgp_name must be removed. Shp_no is the only foreign key column
of this table.
PRODUCTS (PDT) fk1=pdt_pgp_fk

pk * Code fk2=pdt_gpt_fk
fk1 Pgp_name
*
fk2, uk1 o Gpt_code
fk3, uk2 o Lpt_name
fk3, uk2 o Lpt_shp_no
GLOBAL_PRODUCTS (GPT)
pk * Code
o Size
LOCAL_PRODUCTS (LPT)
lpt_shp_fk
fk3=pdt_lpt_fk pk * Name
pk, fk * Shp_no
A-46
.....................................................................................................................................................
.....................................................................................................................................................
Practice 7—4 Primary Keys and Columns: Solution

Identify the Primary key columns and names resulting from the transformation of
the GLOBAL PRICE entity. Give the short name also.
GLOBAL_PRICES (GPE) fk1= gpe_plt_fk

pk, fk1 * Plt_cty_code
pk, fk1 * Plt_start_date fk2= gpe_gpt_fk
pk, fk2 * Gpt_code
* Amount
Note that GLOBAL_PRICES has three columns that are part of its primary key.
All these columns are also part of foreign keys. The first two come from
PRICE_LISTS and have prefix Pct_. The third comes from
GLOBAL_PRODUCTS and has prefix Gpt_.
Note the double prefix in column name Pct_cty_code as this column refers to
Cty_code of PRICELISTS, which is a foreign key column itself.
.....................................................................................................................................................
Practice 8—1 Name that Denormalization: Solution
.....................................................................................................................................................
Practice 8—1 Name that Denormalization: Solution

For the following table diagrams, decide what type of denormalization is used and
explain why the diagram depicts the denormalization you have listed.
Use one of:
1
Type Prejoin tables
Why The Name column from the WEEKDAYS table has been added to the
SHIFTS table, as the Wdy_name column.
Type Short-Circuit Key
Why A new foreign key constraint and column has been added to the
PROD_NAMES table, to join to the PROD_GRPS table.
Type Invalid (could have been either End Date Column or Current Indicator)
Why When you add a column to a table to facilitate issues concerning date
retrieval or insert, add either an end-date column or current indicator column,
not both.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 8—2 Triggers: Solution

denormalized column Order_total of ORDER_HEADERS.
1 How to handle Order_total for ORDER_HEADERS
ORDER_HEADERS (OHR) ORDER_ITEMS (OIM)

pk * Ohr_id
pk * Id
pk * Seqno
* Order_total * Item_total

OHR Insert Y Order_total := 0
Delete N
Update Id N
Order_total Y prevent update
OIM Insert Y recalculate Order_total
Delete Y recalculate Order_total
Update Ohr_id Y recalculate Order_total
Item_total Y recalculate Order_total
denormalized column Lcn_address of EMPLOYEES.
2 How to handle Lcn_address for EMPLOYEES
LOCATIONS (LCN) EMPLOYEES (EPE)

pk * Id
pk * Id
fk * Lcn_id
* Name
* Address
* Lcn_address
.....................................................................................................................................................
Practice 8—2 Triggers: Solution
.....................................................................................................................................................

LCN Insert N
Delete N
Update Address Y Cascade to Employees
other cols Y If pk updated than
extended cascade
EPE Insert Y Set correct Lcn_address
Delete N
Update Lcn_id Y Set correct Lcn_address
Lcn_address Y Prevent update
denormalized column Curr_price_ind of table PRICES.
3 How to handle Curr_price_ind for PRICES
PRODUCTS (PDT) PRICES (PCE)

pk * Pdt_id
* Name o End_date
* Curr_price_ind

PDT Insert N
Delete N
PCE Insert Y Prevent overlap in price periods
Delete N
Update Pdt_id Y Set Curr_price_ind to NULL
Start_date Y Re-evaluate Curr_price_ind
End_date Y Re-evaluate Curr_price_ind
Curr_price_Ind Y Prevent update by user
Note that this solution is incomplete. There is no guarantee that the

Curr_price_ind is indeed set for a price with Start_date ≤ Sysdate ≤ End_date. To
achieve that you need to set a timer that checks, say once a day, if the End_date of
a price that is marked as current has expired.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 8—3 Denormalize Price Lists: Solution

Describe what type of denormalization you would implement and what code you
would add to ensure the database does not lose any integrity. The next diagram
shows the current table schema. Consider both issues described above when
deciding which types of denormalization to implement.
You could handle the design using two denormalization techniques. You could use
the Current Column Indicator technique for the performance issue, and the End
Date Column to allow managers to pre-enter price lists.
Slow Performance
You can increase performance on queries, by adding a denormalized column to
the GLOBAL_PRICES table. Since this table will store a large volume of records
of which only very few will be current, you could consider adding a
Current_indicator column to the table.
Some additional code is required to populate the column correctly.
The code we need to add should read the Start_date column in the PRICE_LISTS
table and set the current record indicator to reflect the current amount. This code
will populate the current record indicator, but when should it fire? Let us assume
the price of a product can only change once a day. With this scenario, we can
simply set a timer for twenty-four hours, and when it expires, fire the code to
update current record indicators, if there have been any changes in the price lists.
Pre-entering Price Lists

The issue of pre-entering price list information, before its effective date will
require another type of denormalization.
Implementing the corporate policy of not waiting to make changes to the Price
list requires a similar, and different, type of denormalization. We need to add an
End Date Column to the PRICE_LISTS table. If we add this new column, then we
can use it to show a price’s expiration date, before that date is reached. So, rather
than calculating the expiration date for a price to be equal to the start date of the
next sequential price, we can enter the start date, price, and end date for the price
when we insert the original record. In some cases the end date may not be known
at the time the record is inserted, so we will make that column optional.
The code we need to add should fire any time a price list is inserted or updated.
The code will read the start date of the current price list, then update the end date
of the prior price list record. There is a small cost during DML, but it is minimal.
The diagram below shows the table diagram after both denormalization
techniques have been implemented.
.....................................................................................................................................................
Practice 8—3 Denormalize Price Lists: Solution
.....................................................................................................................................................
PRICE_LISTS (PLT) GLOBAL_PRICES (GPE)
pk * Start_date pk,fk * Plt_start_date

pk,fk * Cty_code pk,fk * Plt_cty_code
* End_date * Amount
* Current_indicator
.....................................................................................................................................................
.....................................................................................................................................................
Practice 8—4 Global Naming: Solution

Using the design below, denormalize the table design and describe the additional
code that will allow this requirement to be implemented.
One solution to this requirement could be to store the Name for a PRODUCT
with English as language in a new denormalized column in the PRODUCTS
table. This Corporate_name column would then contain the corporate names for
every global product.
You would also need to add some application code to update the Corporate_name
column in the PRODUCTS table, if there are any inserts in the
PRODUCT_NAMES table and if there are changes made to the Name column
from the PRODUCT_NAMES table where the language is English.
The diagram below show the way the tables’ design could be structured.
LANGUAGES (LGE)
pk * Code
* Name
PRODUCTS (PDT) PRODUCT_NAMES (PNE)

pk * Code pk,fk * Pdt_code
o Size pk,fk * Lge_code
* Corporate_Name * Name
.....................................................................................................................................................
Practice 9—1 Data Types: Solution
.....................................................................................................................................................
Practice 9—1 Data Types: Solution

1 Here you see table names and column names and the suggested data type. Do a
quality check on these. If you think it is appropriate, suggest an alternative.
1
yp ( )
Table Column Suggested Your Choice

Data Type Data Type?
COUNTRIES Code Varchar2(2) Char(2)
CURRENCIES Code Varchar2(3) Char(3)
EXCHANGE_RATES Month Date Number(2)
Rate Number(8,4) Number(12,4)
PRICE_LISTS Start_date Date Date
End_date Date Date
PRODUCT_GROUPS Name Char(8) Varchar2(15)
PRODUCTS Code Char(10) Varchar2(15)
Size Number(4,2) Char(1)
Pdt_type Number(1) Char(3)
Fixed length values, like the codes for countries and currencies, can best be
stored using a Char data type. “Month” is not a date and can best be stored
using a number as this allows sorting. In general: do not be hesitant to allow
long character names and other values. You will regret it if you are a little too
economical!
2 Suggest data types for the following columns. They are all based on previous
practices.
2
yp ( )
Table Column Your Choice Data Type

GLOBAL_PRICES Amount Number(15,3)
LOCAL_PRICES Start_date Date
End_date Date
Amount Number(15,3)
SHOPS Name Varchar2(50)
Address Varchar2(50)
City Varchar2(50)
.....................................................................................................................................................
.....................................................................................................................................................
3 What data type would you use for a column that contains times only?
3 A time (like attribute Start_time and End_time in entity SHIFT) can be
implemented in several ways. One is to use a CHAR(5) to record times like
08:55. If calculations are needed, it is often implemented as two Number(2,0)
columns, one for the hours and the second one for the minutes. A third
possibility is to store it as a DATE column with a fixed date (like 01/01/01 but
perhaps on second thoughts, ...!).
Most countries around the world use the two-decimal format for money
amounts in the local currency. The other countries use no decimals. There is,
as far as is known, only one country that uses three decimal places.
.....................................................................................................................................................
Practice 9—2 Artificial Keys: Solution
.....................................................................................................................................................
Practice 9—2 Artificial Keys: Solution

1 Indicate for each table if you see benefits of creating an artificial key and why.
COUNTRIES
GLOBAL_PRICES
PRICE_LISTS
1
aCOUNTRIES have an three-character internationally-used code, which
can be used as a primary key. This code is like an abbreviation and is quite
easy to memorize. I am not sure what happens when a country renames
itself. This may result in a new code. Not the most convincing candidate
for an artificial key.
b GLOBAL_PRICES has no need for an artificial key as there are currently
no tables referring to it. If, however, there is a chance there will be
references in the future, an artificial key would make sense.
c PRICE_LISTS seems a good candidate for an artificial key, as the UID
consists of two components, one of which is a date.
2 For which tables (if any) based on the Moonlight model does it not make any sense
at all to create artificial keys?
2 An artificial key on tables EXCHANGE_RATES, GLOBAL_PRICES (see
above), LOCAL_PRICES, PRODUCT_NAMES would only be an interesting
alternative if the primary key for these tables was referenced in other (future)
tables.
.....................................................................................................................................................
.....................................................................................................................................................
Practice 9—3 Product Pictures: Solution

1 Decide what data type you would advise to be used for each column.
1 Which data type would you use for each column?
Picture LONG RAW
Html_doc LONG
2 You have heard that an old Oracle version would not accept more than one long
type column per table. You are not sure if this is still a limitation. Advise about the
implementation.
2 Advise about the implementation. See the table structure diagram. There can
be one table for multiple LONG columns for a product. There must be a
second table for the LONG RAW columns. This model is, in fact, a vertical
partitioning of a product record. It would also be a good structure when
tables allow more than one long type column. Separating the huge columns
from the small ones increases the speed of a read considerably.
PRODUCTS (PDT) TEXT_DOCUMENTS (TDT)
pk * Code pk,fk * Pdt_code varchar2(3)

……. pk * Info_type varchar2(20)
o Information CLOB
BINARY_DOCUMENTS (BDT)
pk,fk * Pdt_code varchar2(3)

pk * Info_type varchar2(20)
o Information BLOB
A-58
.....................................................................................................................................................
B
.................................
Normalization
Lesson B: Normalization
.........................................................................................................................................
Introduction
Lesson aim
This lesson describes the steps to normalizing table data to the third normal form for
those cases when there is no possibility to perform a full data analysis.
Overview
• Table Normalization
• Normal Forms of Tables
8-2
Topic See Page

Introduction 2
Normalization and its Benefits 3
First Normal Form 7
Second Normal Form 9
Third Normal Form 11
Normalization During Data Modeling 13
Summary 16
Objectives
• Define normalization and explain its benefits
• Place tables in Third Normal Form
.............................................................................................................................................
B-2
Normalization and its Benefits
.........................................................................................................................................

Why and When to Normalize Tables
Before we even talk about why you should normalize, first consider when you should
normalize. If you are developing an application system, and use the techniques of
entity relationship modeling, then you will not need to normalize. One of the
advantages of entity modeling is that the resulting table design is already normalized,
provided there are no obvious errors in the ER model.
The only time you will need to normalize the data is if there has been no time to build
an entity model and when a set of tables is already available. You can then employ the
normalization techniques following the initial database design as a last chance to
check for existing database integrity.
History of Normalization
Normalization is a technique established by the originator of the relational model, E.F.
Codd. The complete set of normalization techniques, include twelve rules that
databases need to follow in order to be described as truly normalized. It is a technique
that was created in support of relational theory, years before entity modeling was
developed. The entity modeling process has incorporated many of the normalization
techniques to produce a normalized entity relationship diagram.
Two terms that have their origins in the normalization technique are still widely in use.
One is normalized data, the other is denormalization.
Objective of Normalization
The major objective of normalization is to remove redundant data from an existing set
of tables or table definitions, thereby increasing the integrity of the database design
and to maximize flexibility of data storage. Removing redundant data helps to
eliminate update anomalies. The first three normal forms progress in a systematic
manner to achieve this objective.
There are many other normal forms in addition to the first three, and they deal with
more subtle anomalies. In general, the IT industry considers normalization to the Third
form an acceptable level to remove redundancy. With a few exceptions, higher
normalization levels are not widely used.
The major subject of normalization is tables, not entities.
.........................................................................................................................................
®
B-3
.........................................................................................................................................
Why Normalize?
• An Entity Model is not always available as a

starting point for design
• To reduce redundant data in existing design
• To increase integrity of data, and stability of
design
• To identify missing tables, columns and
constraints
Note: Third normal form is the generally-accepted

goal for a database design that eliminates
redundancy.
8-3
Normalization Compared to Normalized Data

Normalized data is data that contains no redundancies. This is important as data
redundancy may cause integrity problems. Normalization is the activity, the process,
that leads to a normalized data structure as does entity relationship modeling.
Benefits of Normalized Data

The major benefits of a correctly normalized database from an Information Systems
perspective include:
• Refinement of the strategy for constructing tables and selecting keys.
• Improved communication with the end-users’ application activities.
• Reduced problems associated with inserting and deleting data.
• Reduced enhancement and modification time associated with changing the data
structure.
• Improved information for decisions relating to the physical database design.
• Identification of potential problems that may have been overlooked during
analysis.
.............................................................................................................................................
B-4
.........................................................................................................................................
Recognize Unnormalized Data
USER USER MSE REC_ SRVR SERVER

_ID _NAME _ID DATE SUBJECT TEXT _ID _NAME
---- ----- ----- ----- --------------- ---------- ---- ------
2301 Smith 54101 05/07 Meeting Today There is.. 3786 IMAP05
2301 Smith 54098 07/12 Promotions I like to. 3786 IMAP05
2301 Smith 54445 10/06 Next Assignment Your next. 3786 IMAP05
5607 Jones 54101 05/07 Meeting Today There is.. 6001 IMAP08
5607 Jones 54512 06/07 Lunch? Can you... 6001 IMAP08
5607 Jones 54660 12/01 Jogging Today? Can you... 6001 IMAP08
7773 Walsh 54101 05/07 Meeting Today There is.. 9988 EMEA01
7773 Walsh 54554 03/17 Stock Quote The latest 9988 EMEA01
0022 Patel 54101 05/07 Meeting Today There is.. 2201 EMEA09
0022 Patel 54512 06/07 Lunch? Can we ... 2201 EMEA09
8-4
Unnormalized Data
Data that has not been “normalized” is considered to be “unnormalized” data or data in
zero-normal form. This data is not to be confused with data that is denormalized. If no
ERM was created at the start of a database design project, you are likely to have un-
normalized data, not denormalized data. If you want to add redundancy, for faster
performance or other reasons, you follow the rules defined during the process of
denormalization. But, to denormalize data you must start with normalized data. You
cannot denormalize an unnormalized design, just as you cannot de-ice your car, if
there is no ice on it.
.........................................................................................................................................
®
B-5
.........................................................................................................................................
Normalization
Normalization consists of a series of rules that must be applied to move from a
supposedly unnormalized set of data to a normalized structure. The process is
described in various steps which lead to a “higher” level of normalization. These
levels are called normal forms.
Normalization Rules
Normal Form Rule Description
First Normal Form The table must express a set of

(1NF) unordered, two-dimensional tables.
The table cannot contain repeating
groups.
Second Normal Form (2NF) The table must be in 1NF. Every

non-key column must be dependent
on all parts of the primary key.
Third Normal Form (3NF) The table must be in 2NF. No non-key

column may be functionally dependent
on another non-key column.
“Each non-primary key value MUST be dependent on the key,

the whole key, and nothing but the key.”
8-5
.............................................................................................................................................
B-6
First Normal Form
.........................................................................................................................................
First Normal Form

Definition of First Normal Form (1NF)
The table must express a set of unordered, two-dimensional tables structures. A table
is considered in the first normal form if it contains no repeating groups.
Steps to Remove Repeating Groups

1 Remove the repeating columns from the original unnormalized table.
2 Create a new table with the primary key of the base table and the repeating
columns.
3 Add another appropriate column to the primary key, which ensures uniqueness.
4 Create a foreign key in the new table to link back to the original unnormalized
table.
Converting to First Normal Form

---- ----- ----- ----- --------------- ------------
---- ------
2301 Smith 54101 05/07 Meeting Today There is....
3786 IMAP05
2301 Smith 54098 07/12 Promotions I like to...
3786 IMAP05
2301 Smith 54445 10/06 Next Assignment Your next...
3786 IMAP05
5607 Jones 54512 06/07 Lunch? Can you.....
6001 IMAP08
5607 Jones 54101 05/07 Meeting Today There is....
6001 IMAP08
5607 Jones 54660 12/01 Jogging Today? Can you.....
6001 IMAP08
7773 Walsh 54101 05/07 Meeting Today There is....
9988 EMEA01
7773 Walsh 54554 03/17 Stock Quote The latest..
9988 EMEA01
0022 Patel 54101 05/07 Meeting Today There is....
9988 EMEA01
0022 Patel 54512 06/07 Lunch? Can we .....
9988 EMEA01
1. Remove repeating group from the base table.

2. Create a new table with the PK of the base table and the reporting
group.
8-6
.........................................................................................................................................
®
B-7
.........................................................................................................................................
First Normal Form—Single Record

USERS (unnormal)
---- ----- ----- ----- ------------- ---------- ---- ------
2301 Smith 54101 05/07 Meeting Today There is.. 3786 IMAP05
5607 Jones 54512 06/07 Lunch? Can you... 6001 IMAP08
7773 Walsh 54101 05/07 Meeting Today There is.. 9988 EMEA01
0022 Patel 54101 05/07 Meeting Today There is.. 9988 EMEA01
USER USER SRVR SERVER

_ID _NAME _ID _NAME
USERS ---- ----- ---- ------
(normal) 2301 Smith 3786 IMAP05
5607 Jones 6001 IMAP08
7773 Walsh 9988 EMEA01
0022 Patel 9988 EMEA01
8-7
First create a second table to contain the repeating group columns. Then create a
primary key composed of the primary key from the unnormalized table and another
column that is unique. Finally create a foreign key to link back to the first table.
First Normal Form—Repeating Groups
RECEIVED_ USER MSE REC_

_ID _ID DATE SUBJECT TEXT
MESSAGES ---- ----- ----- --------------- ----------
(1NF) 2301 54101 05/07 Meeting Today There is..
2301 54098 07/12 Promotions I like to.
2301 54445 10/06 Next Assignment Your next.
5607 54101 05/07 Meeting Today There is..
5607 54512 06/07 Lunch? Can you...
5607 54660 12/01 Jogging Today? Can you...
7773 54101 05/07 Meeting Today There is..
USER USER 7773
SRVR 54554
SERVER03/17 Stock Quote The latest
_ID _NAME 0022
_ID 54101
_NAME05/07 Meeting Today There is..
---- ----- 0022
---- 54512
------06/07 Lunch? Can we ...
2301 Smith 3786 IMAP05
7773 Walsh 9988 EMEA01 USERS (normal)
8-8
.............................................................................................................................................
B-8
Second Normal Form
.........................................................................................................................................
Second Normal Form

Definition of Second Normal Form (2NF)
A table is in second normal form if the table is in the first normal form and every non-
primary key column is functionally dependent upon the entire primary key. No non-
primary key column can be functionally dependent on part of the primary key.
Depends on is defined as: a column B depends on column A this means that B must be
re-evaluated whenever A changes.
A table in the first normal form will be in second normal form if any one of the
following apply:
• The primary key is composed of only one column.
• No nonkeyed columns exist in the table.
• Every nonkeyed attribute is dependent on all of the columns contained in the
primary key.
Converting to Second Normal Form
1. Determine which non-key columns are not

dependent upon the table’s entire primary key.
2. Remove those columns from the base table.
3. Create a second table with those columns and the
columns from the PK that they are dependent
upon.
8-9
Steps to Remove Partial Dependencies

1 Determine which nonkey columns are dependent upon the table’s entire primary
key.
2 Remove those columns from the base table. Create a second table with those
nonkeyed columns and a copy of the columns from the primary key that they are
dependent upon.
3 Create a foreign key from the original base table to the new table, linking to the
new primary key.
.........................................................................................................................................
®
B-9
.........................................................................................................................................
Tables Already in Second Normal Form
USERS
USER USER SRVR SERVER

_ID _NAME _ID _NAME
---- ----- ---- ------
2301 Smith 3786 IMAP05
Is the USERS table already in 2NF?
8-10
Convert to Second Normal Form

RECEIVED_ USER MSE REC_
MESSAGES _ID _ID DATE SUBJECT TEXT
---- ----- ----- -------------- ---------
(1NF) 2301 54101 05/07 Meeting Today There is.
RECEIVED_ 2301 54098 07/12 Promotions I like to
2301 54445 10/06 Next Assignmen Your next
MESSAGES 5607 54101 05/07 Meeting Today There is.
(2NF) 5607 54512 06/07 Lunch? Can you..
USER MSE REC_5607 54660 12/01 Jogging Today? Can you..
_ID _ID DATE7773
7773
54101 05/07 Meeting Today There is.
54554 03/17 Stock Quote The lates
---- ----- -----
0022 54101 05/07 Meeting Today There is.
2301 54101 05/07
0022 54512 06/07 Lunch? Can we ..
2301 54098 07/12
2301 54445 10/06 MESSAGES MSE
5607 54101 05/07 (2NF) _ID SUBJECT TEXT
5607 54512 06/07 ----- ------------ ---------
5607 54660 12/01 54101 Meeting Toda There is.
7773 54101 05/07 54098 Promotions I like to
7773 54554 03/17 54445 Next Assignm Your next
0022 54101 05/07 54512 Lunch? Can you..
0022 54512 06/07 54660 Jogging Toda Can you..
54554 Stock Quote The lates
8-11
.............................................................................................................................................
B-10
Third Normal Form
.........................................................................................................................................
Third Normal Form

Definition of Third Normal Form (3NF)
A table is in third normal form if every nonkeyed column is directly dependent on the
primary key, and not dependent on another nonkeyed column. If the table is in second
normal form and all of the “transitive dependencies” are removed, then every non-
keyed column is said to be “dependent upon the key, the whole key, and nothing but
the key”.
Converting to Third Normal Form
Remove any columns that are dependent upon

another non-key column:
1. Determine which columns are dependent upon
another non-key column.
2. Remove those columns from the base table.
3. Create a second table with those columns and the
non-key columns that they are dependent upon.
8-12
Steps to Remove Transitive Dependencies

1 Determine which columns are dependent on another non-keyed column.
2 Remove those columns from the base table.
3 Create a second table with those columns and the non-key columns that they are
dependent upon.
4 Create a foreign key in the original table linking to the primary key of the new
table.
.........................................................................................................................................
®
B-11
.........................................................................................................................................
Tables Already in Third Normal Form
No non-key column can be functionally dependent

upon another non-key column.
RECEIVED_
MESSAGES MESSAGES
(2NF) (2NF)
USER MSE REC_ ID SUBJECT TEXT
_ID _ID DATE ----- -------------- ---------
---- ----- ----- 54101 Meeting Today There is.
2301 54101 05/07 54098 Promotions I like to
2301 54098 07/12 54445 Next Assignmen Your next
2301 54445 10/06 54512 Lunch? Can you..
5607 54101 05/07 54660 Jogging Today? Can you..
5607 54512 06/07 54554 Stock Quote The lates
5607 54660 12/01
7773 54101 05/07
7773 54554 03/17 Are these two tables in third
0022 54101 05/07
0022 54512 06/07 normal form? Why?
8-13
Converting to Third Normal Form
USERS
SRVR
ID NAME _ID
USERS ---- ----- ----
USER USER SRVR SERVER 2301 Smith 3786
_ID _NAME _ID _NAME 5607 Jones 6001
---- ----- ---- ------ 7773 Walsh 9988
2301 Smith 3786 IMAP05 0022 Patel 9988
0022 Patel 9988 EMEA01 MAIL_ ID NAME
SERVER ---- ------
3786 IMAP05
6001 IMAP08
9988 EMEA01
8-14
The theory of normalization goes further than the third normal form to cater for
several problematic constructions that may remain. Those normal forms are outside
the scope of this lesson.
.............................................................................................................................................
B-12
Normalization During Data Modeling
.........................................................................................................................................

If you have created a correct entity model, then the tables created during design will
conform to the rules of normalization. Each formal normalization rule has a
corresponding data model interpretation. The interpretations are as follows:
First Normal Form in Data Modeling

USER
# Name
* Person Name
* Message Receive Date
o Message Subject
o MessageText
First Normal Form Rule

A table must contain no repeating groups.
Corresponding Data Modeling Rule

All attributes must be single-valued.
RECEIVED
MESSAGE received by USER
# Receive Date # Name
o Subject receiver * Person Name
o Text of
8-15
First Normal Rule

A table must contain no repeating groups.
Corresponding Data Modeling Rule All attributes must be single-valued.

You can often recognize the misplaced attributes by the fact that there is the same
(entity) name in the attribute name, such as Message Subject and Message Text.
.........................................................................................................................................
®
B-13
.........................................................................................................................................
Second Normal Form in Data Modeling

Second Normal Form Rule
Every non-key column must be dependent upon all the parts of
the primary key.
An attribute must be dependent upon its entity’s entire unique
identifier.
RECEIVED including MESSAGE
MESSAGE # Id
# User Name included o Text
* Receive Date in
* Subject
RECEIVED including MESSAGE

MESSAGE # Id
# User Name included o Text
* Receive Date in * Subject
8-16
Second Normal Rule

Every nonkeyed column must be dependent upon all parts of the primary key.
Corresponding Data Modeling Rule An attribute must be dependent upon its

entity’s entire unique identifier.
.............................................................................................................................................
B-14
.........................................................................................................................................
Third Normal Form in Data Modeling

USER
# Name
* Person Name
* Password
* Server Id
* Server Name
Third Normal Form Rule
No non-key column can be functionally dependent upon
another non-key column.
No non-UID attribute can be dependent upon another non-
UID attribute.
USER MAIL SERVER
# Name assigned to
# Id
* Person Name
distribute * Name
* Password
mail to
8-17
Third Normal Rule

No nonkeyed column can be functionally dependent upon another nonkeyed column.
Corresponding Data Modeling Rule No non-UID attribute can be dependent upon

another non-UID attribute
.........................................................................................................................................
®
B-15
.........................................................................................................................................
Summary
Summary
1NF The table must express a set of unordered, two-

dimensional tables. The table cannot contain
repeating groups.
2NF The table must be in 1NF. Every non-key column must

be dependent on all parts of the primary key.
3NF The table must be in 2NF. No non-key column may be

functionally dependent on another non-key column.
An entity model transforms into normalized data design.
8-18
.............................................................................................................................................
B-16
Instructor Notes
.........................................................................................................................................
Instructor Notes
Instructor Note
Topic Timings
A
A
Lecture xx minutes
N orm alization
Practice xx minutes
Total xx minutes
Co pyrig ht  O racle Co rpora tion , 19 99. A ll rights reserve d.
Tell the students you will show them how to

O verview
remove redundancy in an existing set of tables.
• T a b le N orm aliza tio n
• N o rm al F o rm s o f T a b le s
8 -2
Things have changed over the past ten years or

W h y N o rm a lize?
so. More people create data models, and fewer
• A n E n tity M o d e l is n o t a lw a ys a va ila b le as a
startin g p o in t for d es ig n start outright at table design. An ERM will
• T o re d uc e red un d a n t d ata in e xis tin g d e sig n
• T o in c re ase in te g rity o f d a ta , an d s ta b ility o f translate to a normalized database design. So,
d es ig n
• T o id e n tify m iss in g tab le s, c olu m n s an d
co n s traints
why would you need to normalize a table
design? Typically an effort to take existing
N o te: Th ird n o rm a l fo rm is th e g en e ra lly -a cc e pte d
g o a l fo r a d ata b as e de s ig n th at e lim in ates
re d u n d an c y.
table structures and move them to new
8 -3
implementation technology, without revisiting
the business requirements, are candidates for
normalization. In this course we will only look
at normalization at the third level, but Codd
described twelve normal levels in his original
assertions.
Unnormalized data is not ‘de-normalized’.
R eco gnize U nn orm alize d D ata
Point out some of the repeating values.
U SE R U S ER MSE R EC _ SR V R S ER V ER
_I D _ N AM E _ID D AT E S UB J EC T TE XT _I D _ NA M E
- -- - - - -- - - -- - - - -- - - - -- - -- - -- - -- - -- -- -- - - -- -- -- - - - -- - --
2 30 1 S m it h 5 41 0 1 0 5/ 0 7 M ee t in g T o da y Th er e is .. 37 8 6 I MA P 05
2 30 1 S m it h 5 40 9 8 0 7/ 1 2 P ro m ot i on s I li k e t o. 37 8 6 I MA P 05
2 30 1 S m it h 5 44 4 5 1 0/ 0 6 N ex t A s si g nm e nt Yo ur n ex t. 37 8 6 I MA P 05
5 60 7 J o ne s 5 41 0 1 0 5/ 0 7 M ee t in g T o da y Th er e is .. 60 0 1 I MA P 08
5 60 7 J o ne s 5 45 1 2 0 6/ 0 7 L un c h? Ca n y o u. .. 60 0 1 I MA P 08
5 60 7 J o ne s 5 46 6 0 1 2/ 0 1 J og g in g T o da y ? Ca n y o u. .. 60 0 1 I MA P 08
7 77 3 W a ls h 5 41 0 1 0 5/ 0 7 M ee t in g T o da y Th er e is .. 99 8 8 E ME A 01
7 77 3 W a ls h 5 45 5 4 0 3/ 1 7 S to c k Q uo t e Th e l a te st 99 8 8 E ME A 01
0 02 2 P a te l 5 41 0 1 0 5/ 0 7 M ee t in g T o da y Th er e is .. 22 0 1 E ME A 09
0 02 2 P a te l 5 45 1 2 0 6/ 0 7 L un c h? Ca n w e . .. 22 0 1 E ME A 09
8 -4
.........................................................................................................................................
®
B-17
.........................................................................................................................................
Normalization at the table level produces the

N orm alizatio n R u les
N orm al Form R ule D escription same table structure as one created following a
First N orm al Form
(1N F)
The table m u st express a set of
unordered, tw o-dim ensional tables.
The table can not contain repeating
completed entity model.
groups.
S econd N orm al Form (2 N F) The table m u st be in 1N F. E very

non-key colu m n m ust be dependent
on all parts o f the prim ary key.
Third N orm al Form (3N F ) The table m u st be in 2N F. N o non-key

colum n m ay be functionally depende nt
on another no n-key colum n.
“ E ach non-prim ary key value M U S T be dep endent on the key,

the w hole key, and noth ing but the key.”
8 -5
Purple highlights the columns removed from

C onv erting to Firs t N orm al Form
U S ER US E R MSE R EC _ SR V R S ER V ER
the unnormalized table.
_ID _N A ME _ID D AT E SU B J EC T TE XT _ID _N A ME
- - -- -- - -- - -- - - - -- - - -- - - -- - -- -- - -- - -- -- - -- - -- ---
- -- - -- - --
2 3 01 Sm i th 5 41 0 1 0 5/ 0 7 Me e t in g T od a y Th er e i s .. .37
. 86 I MA P 05
2 3 01 Sm i th 5 40 9 8 0 7/ 1 2 Pr o m ot i on s I li k e t o. .37. 86 I MA P 05
2 3 01 Sm i th 5 44 4 5 1 0/ 0 6 Ne x t A s si gn m en t Yo ur ne x t. .37. 86 I MA P 05
5 6 07 Jo n es 5 45 1 2 0 6/ 0 7 Lu n c h? Ca n y ou . .. .60
. 01 I MA P 08
5 6 07 Jo n es 5 41 0 1 0 5/ 0 7 Me e t in g T od a y Th er e i s .. .60
. 01 I MA P 08
5 6 07 Jo n es 5 46 6 0 1 2/ 0 1 Jo g g in g T od a y? Ca n y ou . .. .60
. 01 I MA P 08
7 7 73 Wa l sh 5 41 0 1 0 5/ 0 7 Me e t in g T od a y Th er e i s .. .99
. 88 E ME A 01
7 7 73 Wa l sh 5 45 5 4 0 3/ 1 7 St o c k Q uo te Th e l at e st .99
. 88 E ME A 01
0 0 22 Pa t el 5 41 0 1 0 5/ 0 7 Me e t in g T od a y Th er e i s .. .99
. 88 E ME A 01
0 0 22 Pa t el 5 45 1 2 0 6/ 0 7 Lu n c h? Ca n w e . .. .99. 88 E ME A 01
1. R em o ve repeating group from the base table.

2. C reate a new table w ith the P K of the base table and the reporting
group .
8 -6
Just show them the new table structure. Point

Firs t N o rm al Form — S in gle R eco rd
U S E R S (unnorm al)
U SE R U S ER MSE RE C _ S R VR SE R VE R
out the primary key of the USERS table:
_ ID _ N AM E _ ID
- -- - - - -- - -- - --
2 30 1 S m it h 54 1 01
DA T E SU B JE CT TE X T _ ID _ N AM E
-- - -- -- - -- -- - - -- - - -- - -- - - -- - - - -- -- - -- -
05 / 07 Me e ti ng T od a y Th e re i s. . 3 7 86 IM A P0 5
USER_ID.
5 60 7 J o ne s 54 5 12 06 / 07 Lu n ch ? Ca n y o u .. . 6 0 01 IM A P0 8
7 77 3 W a ls h 54 1 01 05 / 07 Me e ti ng T od a y Th e re i s. . 9 9 88 EM E A0 1
0 02 2 P a te l 54 1 01 05 / 07 Me e ti ng T od a y Th e re i s. . 9 9 88 EM E A0 1
US ER US E R S RV R S E RV E R
_I D _N A M E _I D _ NA M E
USERS -- -- -- - - - - -- - - - -- - -
(norm al) 23 01 Sm i t h 3 78 6 I M AP 0 5
56 07 Jo n e s 6 00 1 I M AP 0 8
77 73 Wa l s h 9 98 8 E M EA 0 1
00 22 Pa t e l 9 98 8 E M EA 0 1
8 -7
Primary key of RECEIVED_MESSAGES is

First N orm al Fo rm — R e peating G ro ups
US E R MS E R EC _
User_id and Message_id. The foreign key
R E C E IV E D _
MESSAGES
(1N F)
_I D
-- - -
23 0 1
_I D D AT E
- - -- - - -- - -
5 4 10 1 0 5/ 0 7
S UB J EC T
- -- - -- - -- - -- -- -
M ee t in g T o da y
TE X T
-- - -- -- - --
Th e re i s ..
constraint that refers to USERS includes the
23 0 1 5 4 09 8 0 7/ 1 2 P ro m ot i on s I l ik e t o.
23 0 1
56 0 7
56 0 7
5 4 44 5 1 0/ 0 6
5 4 10 1 0 5/ 0 7
5 4 51 2 0 6/ 0 7
N ex t A s si g nm en t
M ee t in g T o da y
L un c h?
Yo u r ne x t.
Th e re i s ..
Ca n y ou . ..
User_id column.
56 0 7 5 4 66 0 1 2/ 0 1 J og g in g T o da y? Ca n y ou . ..
77 7 3 5 4 10 1 0 5/ 0 7 M ee t in g T o da y Th e re i s ..
U SE R U S ER 77
SR 7
V3R 5SE
4 55
R V4ER0 3/ 1 7 S to c k Q uo t e Th e l at e st
_ ID _ N AM E 00
_2I2D 5 4_10
N A1ME0 5/ 0 7 M ee t in g T o da y Th e re i s ..
- -- - - - -- - 00
-- 2
-2- 5--
4 51
- -2--0 6/ 0 7 L un c h? Ca n w e . ..
2 30 1 S m it h 37 8 6 IM A P 05
5 60 7 J o ne s 60 0 1 IM A P 08
7 77 3 W a ls h 99 8 8 EM E A 01 U S E R S (norm al)
0 02 2 P a te l 99 8 8 EM E A 01
8 -8
C o nvertin g to S econd N orm al F orm
1 . De te rm in e w h ich n o n-k ey co lu m n s are n o t

d ep e n d en t u p o n th e ta b le ’s en tire p rim ary k ey .
2 . Re m o ve th o se c o lu m n s fro m th e b ase ta ble.
3 . C re a te a se co n d ta b le w ith th o se c olu m n s an d th e
c olu m n s from th e P K th a t th ey are d e p en d e n t
upon.
8 -9
.............................................................................................................................................
B-18
Instructor Notes
.........................................................................................................................................
Moving to second normal form removes partial

T ables A lready in Sec ond N orm al Fo rm
dependencies. The mandatory requirement for
USER S any table to be considered for second normal
U S ER
_ID
- - --
US E R
_N A ME
-- - --
S R VR
_ ID
- - --
S E RV E R
_ NA M E
- - -- - -
form is a concatenated primary key. “You gotta
2 3 01 Sm i th 3 7 86 I M AP 0 5
5 6 07
7 7 73
0 0 22
Jo n es
Wa l sh
Pa t el
6 0 01
9 9 88
9 9 88
I M AP 0 8
E M EA 0 1
E M EA 0 1
be a girl to try out for the girl’s team.”
Is the U S E R S table already in 2N F?
8 -1 0
The primary key of Received_Messages is

C o nvert to S econd N orm al F orm
R E C E IV E D _
MESSAGES
US ER
_I D
MS E R E C _
_I D D A T E S U BJ E CT T E XT
User_Id and Mse_Id. Primary key of
-- -- -- -- - - - - -- - - -- - -- - -- - -- - - - -- -- - --
R E C E IV E D _
(1N F) 23 01 54 10 1 0 5 / 07 M e et i ng To d ay T h er e i s.
23 01 54 09 8 0 7 / 12 P r om o ti o ns I li ke to
23 01 54 44 5 1 0 / 06 N e xt As s ig n me n Y o ur n e xt
MESSAGES is Mse_Id. Only the column
MESSA GES
(2N F)
U S ER MS E
56 07 54 10 1 0 5 / 07 M e et i ng To d ay T h er e i s.
56 07 54 51 2 0 6 / 07 L u nc h ?
R E C_ 77
C a n yo u ..
56 07 54 66 0 1 2 / 01 J o gg i ng To d ay ? C a n yo u .. (Rec_Date) representing the date a particular
_ID _I D D A TE 77 73 54 10 1 0 5 / 07 M e et i ng To d ay T h er e i s.
- - -- - - -- -
2 3 01 5 4 10 1
2 3 01 5 4 09 8
- 73
- - -- 00
0 5 /0 00
7 22
54 55 4 0 3 / 17 S t oc k Q u ot e
54 10 1 0 5 / 07 M e et i
0 7 /1 2 22 54 51 2 0 6 / 07 L u nc h ?
ng To d ay
T h e la t es
T h er e i s.
C a n we ..
message was received by a specific user,
2 3 01 5 4 44 5 1 0 /0 6 MESSAGES
5 6 07 5 4 10 1
5 6 07 5 4 51 2
5 6 07 5 4 66 0
0 5 /0 7
0 6 /0 7
1 2 /0 1
(2N F)
M SE
_ ID SU B JE C T T E XT
-- - -- -- - -- - -- - -- - - - -- -- - --
54 1 01 Me e ti n g T od a T h er e i s.
belongs in RECEIVED_MESSAGES. The
7 7 73 5 4 10 1 0 5 /0 7
7 7 73 5 4 55 4
0 0 22 5 4 10 1
0 0 22 5 4 51 2
0 3 /1 7
0 5 /0 7
0 6 /0 7
54 0 98 Pr o mo t io n s
54 5 12 Lu n ch ?
I li ke to
54 4 45 Ne x t A ss i gn m Y o ur n e xt
C a n yo u ..
54 6 60 Jo g gi n g T od a C a n yo u ..
Mse_Id, in RECEIVED_MESSAGES, is the
8 -1 1
54 5 54 St o ck Qu o te T h e la t es
foreign key column in the fk constraint to the
MESSAGES table.
Moving to third normal form removes transitive
C onve rting to T hird N orm al F orm
dependencies. Make sure no non-keyed column
R e m o v e a n y co lu m n s th at a re d ep e n d en t u p o n
a n o the r n o n -ke y co lu m n :
is dependent on any other non-keyed column.
1 . D eterm in e w h ic h c olu m n s a re d ep e n d en t u p o n
a n o the r n o n -ke y co lu m n .
You must have a direct dependency on the
2 . R em o v e th o s e co lu m n s fro m th e b as e ta b le .
3 . C rea te a s ec o n d ta b le w ith th o se c o lu m n s an d th e
primary key, not an indirect one through
n o n -ke y co lu m ns th at th e y a re d ep e n d en t u p o n .
another non-keyed column
8 -1 2
The RECEIVED_MESSAGES table is in third

Tab les A lre ady in T hird N orm al Fo rm
normal form., due to ‘tricky’ circumstances. It
N o n o n -k ey c o lu m n c an b e fu n ctio n a lly d ep e n d en t
u po n an o th er n on -k ey c olu m n .
R E C E IV E D _
has only one non-keyed column, so there is no
ME SSAGES
(2N F)
U S ER MS E R E C_
MESSAGES
(2N F)
ID SU B JE C T TE X T
way it could be dependent on another non-
_ID _I D
- - -- - - -- -
2 3 01 5 4 10 1
2 3 01 5 4 09 8
D A TE
- - -- -
0 5 /0 7
0 7 /1 2
-- - - - -- - -- - -- - -- - --
54 1 0 1 Me e ti n g T od a y
54 0 9 8 Pr o mo t io n s
54 4 4 5 Ne x t A ss i gn m en
-- - -- - -- -
Th e re is .
I l ik e t o
Yo u r n ex t
keyed column. In the MESSAGES table you
2 3 01 5 4 44 5
5 6 07 5 4 10 1
5 6 07 5 4 51 2
5 6 07 5 4 66 0
1 0 /0 6
0 5 /0 7
0 6 /0 7
1 2 /0 1
54 5 1 2 Lu n ch ?
54 6 6 0 Jo g gi n g T od a y?
54 5 5 4 St o ck Qu o te
Ca n y o u. .
Ca n y o u. .
Th e l a te s
just want to ask “Is the Subject of the message,
7 7 73 5 4 10 1
7 7 73 5 4 55 4
0 0 22 5 4 10 1
0 0 22 5 4 51 2
0 5 /0 7
0 3 /1 7
0 5 /0 7
0 6 /0 7
A re th ese tw o tables in third
norm al form ? W hy?
dependent on the text, or is the text dependent
8 -1 3
on the subject, or are both dependent on the
Mse_Id?”. In this case, both are directly
dependent on the primary key (Mse_Id).
.........................................................................................................................................
®
B-19
.........................................................................................................................................
The USERS table does need a little work.

C o nverting to Th ird N orm al Fo rm
Column Server_Name is directly dependent on
USERS
USERS
ID NA M E
S R VR
_ ID
the Srvr_Id, not the User_Id. A new table is
-- - - -- - - - - - --
US E R
_I D
US E R
_N A M E
SR VR
_ ID
SE R V ER
_ N A ME
23 0 1
56 0 7
77 7 3
Sm i t h
Jo n e s
Wa l s h
3 7 86
6 0 01
9 9 88
created containing both the column violating
-- - - -- - - - -- -- -- - - --
23 0 1
56 0 7
77 7 3
Sm i t h
Jo n e s
Wa l s h
37 86
60 01
99 88
IM A P 05
IM A P 08
EM E A 01
00 2 2 Pa t e l 9 9 88
the dependency (Server_Name) and the column
M A IL_
00 2 2 Pa t e l 99 88 EM E A 01
SERVER
ID
- -- -
3 78 6
N A ME
- - -- - -
I M AP 0 5
it is dependent on (Srvr_id). The Srvr_Id is still
6 00 1 I M AP 0 8
9 98 8 E M EA 0 1
part of the USERS table, but only fills the role
8 -1 4
of foreign key column to the MAIL_SERVERS
table.
The next three slides highlight the connection
First N orm al F orm in D a ta M ode ling
U SER
of normalization to the entity modeling. Each of
# N am e
* P erson N am e
* M essage R eceive D ate
the three normal forms is shown with its
o M essage S ubject
First N orm al Form R ule

o M essageText
corresponding data model rule.
A table m ust c ontain no repeating grou ps.
C orresponding D ata M o deling R ule

A ll attributes m ust be single-value d.
R E C E IV E D
MESSAGE received by USER
# R eceive D ate # N am e
o S ubject receiver * P erson N am e
o Text of
8 -1 5
Second normal form - remove partial

Sec ond N orm al Fo rm in D ata M odeling
S eco nd N orm a l F orm Ru le
dependencies.
E very n on -ke y co lum n m u st be dep en den t upo n all the pa rts of
th e p rim ary k ey.
C orresp on din g D a ta M od elin g Ru le
A n attrib ute m ust be de pe nde nt up on its en tity ’s e ntire un iqu e
ide ntifier.
R E C E IV E D in clud ing MESSAGE
MESSAGE # Id
# U ser N am e inclu ded o Text
* R eceive D ate in
* S ubject
R E C E IV E D inclu din g MESSAGE

MESSAGE # Id
# U ser N am e inclu ded o Text
* R eceive D ate in * S u bject
8 -1 6
.............................................................................................................................................
B-20
Instructor Notes
.........................................................................................................................................
Third normal form - remove transitive

Th ird N o rm a l Form in D ata M o delin g
USER
dependencies
# N am e
* P erson N am e
* P assw ord
* S erver Id
* S erver N am e
T hird N o rm al Fo rm R u le
N o non -k ey colu m n ca n b e fun ctio na lly dep en den t u po n
an other n on -ke y co lum n .
C orresp on din g D ata M od elin g Ru le
N o non -U ID attrib ute c an b e d epe nd en t u po n a nothe r n on -
U ID attribu te .
USER M A IL S E R V E R
# N am e assig ned to
# Id
* P erson N am e * N am e
* P assw ord d istribu te
m ail to
8 -1 7
Sum m ary
1N F Th e table m ust exp ress a se t o f u no rd ered , tw o -

dim en sio na l ta bles . T he tab le cann o t co ntain
re pe atin g grou ps.
2N F Th e table m ust be in 1N F. E very n on -ke y co lum n m u st

be de pen de nt on all parts of th e prim ary k ey.
3N F Th e table m ust be in 2N F. N o n on -ke y co lu m n m ay b e

fu nc tion ally de pe nd ent on an other n on -ke y co lu m n .
A n en tity m od el tran sform s into n orm alize d d ata de sig n.
8 -1 8
.........................................................................................................................................
®
B-21
.........................................................................................................................................
.............................................................................................................................................
B-22
Index
.....................................................................................................................................................
Index BLOB 9-6

business function 1-23
business rules 4-2
A
arc 1-27, 4-12 C
both supertype and subtype implemen-
tation 7-25 cascade composed UID 4-7
exclusive 4-12 cascade delete 9-15
rules 4-14 cascade update 9-15
arc implementation 7-25 chain
generic 9-26 pattern 6-10
rules 7-25 CHAR 9-6
arc or subtypes 4-16 check
arcs conditional domain 4-20
incorrect 4-15 conditional relationship 4-20
mapping 7-19 front door 4-20
artificial key 9-11 range 4-20
attribute 1-13 state value transition 4-20
multiple state value triggered 4-20
UID 4-7 check constraint 4-13
redundancy 2-16 classification, pattern 6-7
single CLOB 9-6
UID 4-7 column
single valued 1-13 current indicator 8-20
attribute constraint 4-19 column sequence 9-7
attribute representation columns 7-5
mandatory 1-19 choosing for index 9-19
optional 1-19 end date 8-18
attributes 3-19 foreign key
naming 2-15 naming 7-9
recycling 6-20 composed
tracking 2-14 UID 4-7
volatile 1-14 concatenated index 9-18
attributes modeled as PROPERTY in- concept
stance 6-20 evolution 2-11
conceptual data modeling 1-8, 2-5, 3-26
conceptual model 7-6
B conceptual modeling 1-4
B*Tree conceptual models 1-28
index 9-17 conditional domain check 4-20
barred relationship 4-6 conditional nontransferability 5-9
basket, pattern 6-6 conditional relationship 4-20
BFILE 9-6 constraint
bill of material check 4-13
pattern 6-12 declarative 7-7
binary table 3-26 constraints 4-2
bitmap index 9-17 check
.....................................................................................................................................................
®
Index-1
Index
.....................................................................................................................................................
naming 7-10 data blocks 9-30

foreign key data files 9-30
naming 7-9 extents 9-30
hierarchy 6-9 Oracle 9-29
special 4-20 partitions 9-29
time-related 5-8 segments 9-29
convention tablespaces 9-29
naming 7-8 DATE 9-6
conventions date
sensible use 6-18 end 5-7
crowsfoot 3-7 start 5-7
current indicator column 8-20 date as Opposed to day 5-5
date or day 5-5
date time 5-6
D declarative constraint 7-7
data 2-4 default and nullify 9-15
normalized B-3 definition
unnormalized B-5 denormalization 8-4
data blocks 9-30 definition of an entity 1-10
data files 9-30 degree 3-7
data modeling delete
conceptual 2-5 cascade 9-15
physical 2-5 restrict 9-14
data type denormalization
BFILE 9-6 definition 8-4
BLOB 9-6 denormalization techniques
CHAR 9-6 derivable values 8-5
CLOB 9-6 hard-coded values 8-5
DATE 9-6 pre-joining tables 8-5
LONG 9-6 derivable 1-8
LONG RAW 9-6 derivable values
NUMBER 9-6 storing 8-6
VARCHAR2 9-6 design
data types distributed 9-27
Oracle 9-5 old fashioned 9-25
data warehouse 2-6 design strategy
pattern 6-16 client-server 7-12
data warehouse system data warehouse approach 7-8
design strategy 7-8 discriminator column 7-20
star model 7-10, 9-14 distributed design 9-27
database benefits 9-28
hierarchical 2-6 domain 4-19
network 2-6 conditional 4-20
object oriented 2-6 drawing conventions 6-17
relational 2-6
semantic 2-6
database structure
.....................................................................................................................................................
®
Index-2
Index
.....................................................................................................................................................
E optional composed 7-15

foreign keys 7-5
electronic mail 2-9
form
elements
first normal B-7
arc 1-27
second normal B-9
nontransferability 1-27
formal description of the entity 2-7
subtype 1-27
front door check 4-20
unique identifier 1-27
function
end date 5-7
business 1-23
end date columns 8-18
modeling 1-23
entities
function based index 9-18
event 2-20
functionality 1-23, 2-13
intangible 2-20
tangible 2-20
entity 3-25 G
formal description 2-7
generic arc implementation 9-26
inheritance 2-17
generic model 6-22
intersection 3-25
generic modeling 6-19
naming 2-7
generic models 6-20, 6-21
subtypes 2-17
graphical elements 1-17
supertype 2-17
entity DAY 5-6
entity definition H
evolution 2-11
entity life cycle 2-12 hard-coded values 8-10
entity relationship diagram 1-17 hidden relationships 4-18
entity relationship model 1-17 hierarchies
entity relationship modeling 1-7, 1-28 disputable 6-8
ER diagram false 6-8
soft box 1-18 hierarchy
ER model constraints 6-9
transform 7-4 pattern 6-8
evolution of a concept 2-11 hierarchy level indicator 8-22
exclusive arc 4-12 historical price 5-10
extents 9-30 homonyms 2-8
house building metaphor 1-5
F
I
fan trap
pattern 6-15 identification 4-4
first normal form B-7 in database 4-5
foreign key indirect 4-8
cascade delete 9-15 problems 4-4
cascade update 9-15 real world 4-5
columns 7-13 identifiers
constraints 7-13 information-bearing 4-11
default and nullify 9-15 incorrect arcs 4-15
incorrect UIDs 4-10
.....................................................................................................................................................
®
Index-3
Index
.....................................................................................................................................................
index entity 2-12

choosing columns 9-19 logging 5-4, 5-17
partitioned 9-22 logic
unique 9-8 referential 5-9
when used 9-21 LONG 9-6
index organized table 9-18 LONG RAW 9-6
index types 9-17
B*Tree 9-17
bitmap 9-17 M
concatenated index 9-18 mandatory 3-7, 3-10
function based index 9-18 many to many (m:m) 3-9
reverse key 9-17 mapping
tree balanced 9-17 basic 7-12
indexes 9-16 entity 7-12
indicator nontransferable relationships 7-15
hierarchy level 8-22 relationship 7-14
indirect identification 4-8 terminology 7-7
information 2-4 mapping arcs 7-19
types 1-24 mapping barred relationships 7-15
information-bearing identifiers 4-11 mapping many-to-many relationships 7-
inheritance 2-17 17
instances 1-10, 1-11 mapping one-to-one relationships 7-18
integrity mapping subtypes 7-20
referential 9-14 master
intersection entity 3-25 keeping details 8-12
repeating single detail 8-14
master detail, pattern 6-5
J
model
journalling 5-4, 5-17 conceptual 7-6
relational 7-6
modeling
K generic 6-19
keeping details with master 8-12 modeling time 5-4
key multiple attribute
artificial 9-11 UID 4-7
foreign 9-14
primary
N
desirable properties 9-9
keys name space 7-11
foreign 7-5 naming
primary 7-5, 9-8 attributes 2-15
short-circuit 8-16 check constraints 7-10
unique 7-5, 9-8 convention 7-8
entities 2-7
foreign key columns 7-9
L foreign key constraints 7-9
life cycle relationships 3-5
.....................................................................................................................................................
®
Index-4
Index
.....................................................................................................................................................
restrictions with Oracle 7-10 roles 6-14

tables 7-8 patterns 6-4
naming relationships 3-5 physical data modeling 2-5
negotiated prices 5-14 pre-joining tables 8-8
nested subtypes 2-19 price 5-10
network negotiated 5-14
pattern 6-11 price history 5-10
network structures 6-11 price list 5-10, 5-12
nontransferability 1-27, 3-8 priced product 5-10
conditional 5-9 primary key
normalization B-3 desirable properties 9-9
normalized data B-3 primary keys 7-5, 9-8
nouns 2-7 primary UID 4-9
NUMBER 9-6 primary unique identifier 3-18
product 5-10
properties
O primary key 9-9
old fashened design
generic arc implementation 9-26
R
unique index 9-25
old fashioned design range check 4-20
check option views 9-25 recursive relationship 3-4
OLTP system 7-6 recycling of attributes 6-20
one to many (1:m) 3-9 redundancy 2-16
one to one (1:1) 3-9 relationships 3-15
onstraint referential integrity 9-14
attribute 4-19 referential logic 5-9
optional 3-7 relational databases 2-6
optionality 3-6 relational model 7-6
Oracle data types 9-5 relationship
Oracle database structure 9-29 conditional 4-20
many to many 3-11
mapping 7-14
P master-detail 6-5
partitioned index 9-22 one to many 6-5
partitioning tables 9-22 recursive 3-4
partitions 9-29 relationship ends
pattern degree 3-7
basket 6-6 optionality 3-6
bill of material 6-12 relationship name 3-5
chain 6-10 relationship representation 1-20
classification 6-7 relationships 1-15, 3-19
data warehouse 6-16 barred
fan trap 6-15 mapping 7-15
hierarchy 6-8 hidden 4-18
master detail 6-5 mandatory 1-21
network 6-11 many to many 3-9
.....................................................................................................................................................
®
Index-5
Index
.....................................................................................................................................................
mapping 7-17 start date 5-7

many to one 3-9 state value transition check 4-20
mapping nontransferable 7-15 state value triggered check 4-20
mapping one to many 7-14 storage implication 7-27
one to many 3-9 arc implementation 7-28
one to one 3-13 subtype implementation 7-27
one-to-one supertype implementation 7-27
mapping 7-18 storing derivable values 8-6
resolving 3-25 subtype 1-27
resolving other 3-26 implementatioin
symmetric 6-13 rules 7-23
UID 4-8 implementation 7-23
relationsships rules 2-18
optional 1-21 subtypes 2-17
repeating single detail with master 8-14 mapping 7-20
representation 4-4 nested 2-19
reserved words. 2-15 subtypes or arcs 4-16
resolving other relationships 3-26 supertype 2-17
resolving relationships 3-25 implementatioin
restrict rules 7-20
delete 9-14 implementation 7-20
update 9-14 supertype and subtype implementation
reverse key index 9-17 arc 7-25
roles symmetric relationships 6-13
pattern 6-14 problem 6-13
rows 7-5 solution 6-13
rules synonyms 2-7
about arcs 4-14
business 4-2
subtype 2-18 T
transformation 7-6 table
binary 3-26
index organized 9-18
S
naming 7-8
second normal form B-9 tables 7-5
secondary UID 4-9 partitioning 9-22
segments 9-29 pre-joining 8-8
sequences 9-13 tablespaces 9-29
set theory 1-12 terminology mapping 7-7
sets. 1-12 three-tiered architecture 7-13
short-circuit keys 8-16 time
similar structure 6-4 modeling 5-4
single attribute time-related constraints 5-8
UID 4-7 tracking attributes 2-14
Snowflake model 6-16 transformation rules 7-6
soft box 1-18 transforming the ER model 7-4
special constraints 4-20 tree balanced index 9-17
.....................................................................................................................................................
®
Index-6
Index
.....................................................................................................................................................
types of information 1-24
U
UID
cascade composed 4-7
composed 4-7
multiple attribute 4-7
primary 4-9
relationships 4-8
secondary 4-9
single attribute 4-7
unique identifier 1-27, 4-6
primary 3-18
unique index 9-8
unique key 7-18
unique keys 7-5, 9-8
unnormalized data B-5
update
cascade 9-15
restrict 9-14
V
values 1-13
derivable
storing 8-6
hard-coded 8-10
VARCHAR2 9-6
views
usage 9-23
volatile attributes 1-14
W
words
reserved 2-15
.....................................................................................................................................................
®
Index-7
Index
.....................................................................................................................................................
.....................................................................................................................................................
Index-8

Data Modeling and Relational Database Design Instructor Guide PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Data Modeling and Relational Database Design Instructor Guide PDF

Transféré par

Droits d'auteur :

Formats disponibles

Data Modeling and Relational

Course Code 20000GC10

This documentation contains proprietary information of Oracle Corporation. It is

Restricted Rights Legend

Lesson 2: Entities and Attributes in Detail

Lesson 3: Relationships in Detail

Practice 3—5: Moonlight P&O 3-33

Lesson 5: Modeling Change

Lesson 6: Advanced Modeling Topics

Lesson 7: Mapping the ER Model

Lesson 8: Denormalized Data

Instructor Post Lesson Teaching Evaluation 8-37

Lesson 9: Database Design Considerations

Practice 2—1 Books: Solution A-11

Practice 9—2 Artificial Keys: Solution A-61

About the slide

Topic See Page

Topic See Page

Why Conceptual Modeling?

About the slide

• It describes exactly the information needs of the

What You Learn in This Course

• Which constraints and other database objects you require.

The House Building Metaphor

About the slide

What? as Opposed to How?

Entity Relationship Modeling

Entity Relationship Modeling

Has a robust syntax

held by the type o

read diagrams… REVIEW

...although they may

look rather complex

at first sight for

What is Involved in Modeling?

Goals of Entity Relationship Modeling

• Capture all required information

The goals of conceptual data modeling are to ensure that:

Entity Relationship modeling is independent of the hardware or software used for

About the slide

PERSON Mahatma Gandhi

Entities and Sets

• An entity represents a set of instances that are of

• Also represents something of significance to the

Entities Have Attributes

Values and Data Types

Attribute is Single Valued

About the slide

• Also represent something of significance to the

Entities usually have relationships. Here are some examples.

About the slide

EMPLOYEES have JOBS

PRODUCTS are classified by a PRODUCT TYPE

PEOPLE make TICKET RESERVATIONS

A relationship connects two entities. A relationship represents a significant

About the slide

Entity Relationship Models and Diagrams

About the slide

During design, entities usually lead to tables.

Mandatory attribute, that is, known and available

During design, attributes lead to columns.

About the slide

An employee has exactly one job