4 DB Relmod

Introduction to
Database Systems
The Relational Data Model
Werner Nutt
1
4. The Relational Data Model
4.1 Schemas
1. Schemas
2. Instances
3. Integrity Constraints
2
Different Schemas are Based
on Different Concepts
Conceptual
Conceptual Logical
Logical Physical
Physical
Schema
Schema Schema
Schema Schema
Schema
ER: Tables/Relations: File organisation:

• Entities, • column names/attributes • File types
• Relationships, • rows/tuples • Index structures
• Attributes
3
A Table
Table name
Column names
Product:
Name Price Category Manufacturer
gizmo $19.99 gadgets GizmoWorks
Power gizmo $29.99 gadgets GizmoWorks
SingleTouch $149.99 photography Canon
MultiTouch $203.99 household Hitachi
4
Rows
Review of Mathematical Concepts (1)
• Sets: S, T, S1, ..., Sn, T1, ..., Tn, { }
Cardinality of a set S denoted as |S|
• Cartesian Product of sets (also cross product):

S × T set of all pairs (s,t) where s ∈ S and t ∈ T
S1 × ... × Sn set of all n-tuples (s1,..., sn)
where si ∈ Si
• Relation R over S, T:
subset of S × T, written R ⊆ S × T
We write (s,t)∈R or, equivalently, sRt
5
• Relation R over S1, ..., Sn:
subset R ⊆ S1 × ... × Sn
The number n is the arity of R
(R is binary if n=2 and ternary if n=3)
• Function f from S to T, written f: S → T

associates to every element s∈S
exactly one element of T, denoted as f(s)
6
• Partial function f from S to T, written f: S ∼→ T
associates to some element s∈S
exactly one element of T, still denoted as f(s)
We write f(s) = ⊥ (read “bottom”)
if f does not associate any element of T to s
• A relation R over S1,...,Sn is total on Si

if for every si ∈ Si
there are sj ∈ Sj, j≠ i, such that (s1,..., sn) ∈ R
In other words:
every element of Si occurs in some tuple of R
7
• A relation R over S and T is functional in its
first argument if
sRt1 and sRt2 implies that t1= t2
for all s∈S, t1, t2 ∈T.
In other words, for every s∈S,
there is at most one t∈T related by R to s
• Analogously, a relation R over S1, ..., Sn can be functional

– in an argument i, or
– in a tuple of arguments, say (i,j,k)
8
How Many … ?
Consider sets
S, T with |S| = N and |T| = M
S1, ..., Sn with |Si| = Ni
• How many elements can a relation over S1, ..., Sn have?

At least? At most?
• How many relations over S, T are there?
How many over S1, ..., Sn?
• How many functions from S to T are there?
• How many partial functions from S to T are there?
9
How Many … ? (Cntd.)
• How many relations are there over S and T

that are functional in the first argument?
• How many relations are there over S and T

that are total on S?
10
Tables Look Like Relations …
A1 A2 A3 ... An
C Attributes {(a1, a2, a3, …, an),
a a1 a2 a3 an (b1, b2, a3, …, cn),
r
d b1 b2 a3 cn (a1, c3, b3, …, bn),
i Tuple
n a1 c3 b3 bn
...
a . (x1, v2, d3, …, wn)}
l .
i . Component
t
y x1 v2 d3 wn
Over which sets does
this relation range?
Arity
In Databases: Distinguish between

• Schema (structure ) and
• Instance (content) 11
Relation Schemas
A relation schema
R(A1:D1, ..., An:Dn)
consists of
• a name, say R
• a nonemtpy set of attributes, say A1,..., An
• a type or domain, say Di = dom(Ai) , for each attribute Ai
Example: Product (Prodname: Name,

Price: DollarPrice,
Category: Name,
Manufacturer: Name)
12
Types and Domains
Type: Class of atomic values, e.g.,
• integers, reals, strings
• integers between 15 and 80,
strings of (up to) 50 characters
Domain: Set of atomic values, that have a specific meaning

in an application, e.g.,
– Name, EmployeeAge
• Domains have a type, e.g.,
– EmployeeAge = Int[15,80] Domains allow for
• Domains may have default values an additional layer
of abstraction
13
4.2 Instances
1. Schemas
2. Instances
14
Relation Schema and Instance
• A tuple of values (v1, ..., vn ) satisfies
the relation schema R(A1:D1, ..., An:Dn) if vi ∈ Di (i=1,…n)
• An instance of R is a set of tuples that satisfy the schema of R

(i.e., a relation over D1, ...,Dn)
• Analogy with programming languages:

– schema = type
– instance = value
• Important distinction:
– Relation Schema = stable over long periods of time
– Relation Instance = changes constantly, as data is
inserted/updated/deleted
15
Example
Domain declaration:
Name=String(30), DollarPrice=Decimal (10,2),
Relation schema:
Product(Prodname: Name, Price: DollarPrice,
Category: Name, Manufacturer: Name)
Instance:
Prodname Price Category Manufacturer
gizmo 19.99 gadgets GizmoWorks

Power gizmo 29.99 gadgets GizmoWorks
SingleTouch 149.99 photography Canon
MultiTouch 203.99 household Hitachi
16
Database Schema and Instance
Database Schema
Set of relation schemas, e.g.,
Product (Productname, Price, Category, Manufacturer),
Vendor (Vendorname, Address, Phone), …
To keep things simple,
Database Instance we have dropped types/domains
Set of relation instances,
one for each relation in the schema
Important distinction:
– Database Schema = stable over long periods of time
– Database Instance = changes constantly
17
Updates
A database reflects the state of an aspect of the real world:
The world changes Î the database has to change
Updates to an instance:
1) adding a tuple
2) deleting a tuple
3) modifying an attribute of a tuple
What could be updates to a schema?
• Updates to the data happen very frequently.

• Updates to the schema: relatively rare, rather painful.
Why?
18
Null Values
Attribute values Three meanings of null values

• are atomic 1. not applicable
• have a known domain 2. not known
• can sometimes be “null” 3. absent (not recorded)
Student
studno name hons tutor year thesis title
s1 jones ca bush 2 null
s2 brown cis kahn 2 null
s3 smith null goble 2 null
s4 bloggs ca goble 1 null
s5 jones cs zobel 1 null
s6 peters ca kahn 3 “A CS Survey”
19
Order and Duplication
In tables:
• Order of attributes is fixed
• Order of rows is fixed (i.e., tables with different order
of rows are different)
• Duplicate rows matter
In mathematical relations:
• Order of tuples and duplicate tuples do not matter
• Order of attributes is still fixed
Question:
Can we model relations so that we get rid of
attribute order?
20
Reminder: Relations as Subsets
of Cartesian Products
• Tuple as elements of String x Int x String x String

E.g., t = (gizmo, 19.99, gadgets, GizmoWorks)
• Relation = subset of String x Int x String x String
• Order in the tuple is important !

– (gizmo, 19.99, gadgets, GizmoWorks)
– (gizmo, 19.99 , GizmoWorks, gadgets)
• No explicit attributes, hidden behind positions
21
Alternative Definition:
Relations as Sets of Functions
• Fix the set A of attributes, e.g.
A = {Name, Price, Category, Manufacturer}
• Fix D as the union of the attribute domains, e.g.,

D = dom(Name) ∪ dom(Price) ∪ dom(Category)
∪ dom(Manufacturer)
• A tuple is a function t: A → D
{Prodname
{Prodname gizmo,
gizmo,
• E.g. Price 19.99, This is the model
Price 19.99,
Category
Category gadgets,
gadgets, underlying SQL
Manufacturer
Manufacturer GizmoWorks}
GizmoWorks}
• Order in a tuple is not important,

22
attribute names are important!
Notation
Schema R(A1, ..., An), tuple t that satisfies the schema
Then:
• t[Ai] = value of t for attribute Ai
• t[Ai, Aj, Ak]
= subtuple of t, with values for Ai, Aj, Ak
Example: t = (gizmo, 19.99, gadgets, GizmoWorks)

– t[Price] = 19.99
– t[ProdName, Manufacturer] = (gizmo, GizmoWorks)
23
Two Definitions of Relations
• Positional tuples, without attribute names

• Tuples as mappings/functions of attributes
In theory and practice, both are used, e.g.,

– SQL: tuples as functions
– QBE (query by example): positional tuples
We will switch back and forth between the two…
24
Why Relations?
• Very simple model
• Often a good match for the way we think about our data
• Foundations in logic and set theory
• Abstract model that underlies SQL, the most important

language in DBMSs today
– But SQL uses “bags” while the abstract relational

model is set-oriented
25
4.3 Integrity Constraints
1. Schemas
2. Instances
26
Integrity Constraints
Ideal: DB instance reflects the real world
In real life: This is not always the case
Goal: Find out, when DB is out of sync
Observation:
Not all mathematically possible instances make sense
Idea:
• Formulate conditions that hold for all plausible instances
• Check whether the condition holds after an update
Such conditions are called integrity constraints!
27
Common Types of Integrity Constraints
• Functional Dependencies (FDs)
– “Employees in the same department have the same boss”
• Superkeys and keys (special case of FDs)

– “Employees with the same tax code are identical”
• Referential Integrity (also “foreign key constraints”)

– “Employees can only belong to a department that
is mentioned in the Department relation”
• Domain Constraints
– “No employee is younger than 15 or older than 80”
Integrity constraints (ICs) are part of the schema

We allow only instances that satisfy the ICs 28
Functional Dependencies (Example)
Emp (Name, taxCode, Dept, DeptHead)
A state of Emp that contains two tuples with

– the same Dept, but different DeptHead
– the same taxCode, but different Name, Dept, or DeptHead
is definitely out of sync.
We write the desired conditions symbolically as

– Dept → DeptHead
– taxCode → Name, Dept, DeptHead.
We read:
– “Dept functionally determines DeptHead”, or
– “Name, Dept, and DeptHead functionally depend on taxCode”
29
Functional Dependencies
DB relation R.
A functional dependency on R is an expression
A1,...,Am → B1, ...,Bn
where A1,...,Am and B1, ...,Bn are attributes of R.
An instance r of R satisfies the FD if for all tuples t1, t2 in R
t1[A1,...,Am] = t2[A1,...,Am]
implies
t1[B1, ...,Bn] = t2[B1, ...,Bn]
How many FDs are there on a given relation?

30
FDs: Example
Emp (EmpID, Name, Phone, Position)
with instance
EmpID Name Phone Position
E0045 Smith 1234 Clerk
E1847 Jones 9876 Salesrep
E1111 Smith 9876 Salesrep
E9999 Brown 1234 Lawyer
Which FDs does this instance satisfy?

• EmpID → Name, Phone, Position
• Position → Phone
• Phone → Position
31
General Approach for Checking FDs
To check A → B on an instance,
• erase all other columns
… A … B
X1 Y1
X2 Y2
… …
• check if the remaining relation is functional in A
Why is that correct?
32
FDs: Example (Cntd.)
Check Position → Phone !
EmpID Name Phone Position

E0045 Smith 1234 Clerk
E1847 Jones 9876 Salesrep
E1111 Smith 9876 Salesrep
E9999 Brown 1234 Lawyer
Is the white relation functional in Position?
33
FDs, Superkeys and Keys
Person (SSN, Name, DOB)
SSN → Name, DOB
Product (Name, Price, Manufacturer)

Name → Price, Manufacturer
Name → Name, Price, Manufacturer
Name, Price → Name, Price, Manufacturer
Book (Author, Title, Edition, Price)

Author, Title, Edition → Price
• A set of attributes of a relation is a superkey if

it functionally determines all the attributes of the relation
• A superkey is a candidate key if
none of its subsets is a superkey
34
Candidate keys are minimal superkeys
Keys: Definitions
• Superkey
– a set of attributes whose values together uniquely
identify a tuple in a relation
• Candidate Key
– a superkey for which no proper subset is a superkey:
a superkey that is minimal
– Can be more than one for a relation
• Primary Key
– a candidate key chosen to be the main key
– One for each relation,
indicated by underlining the key attributes
Student(studno,name,tutor,year) 35
Example: Multiple Keys
Student (Lastname, Firstname, MatriculationNo, Major)
Candidate key Candidate key

(2 attributes)
Superkey
Note: There are alternate candidate keys
• Candidate keys are

{Lastname, Firstname} and
{StudentID}
36
Foreign Keys
A set of attributes in a relation that exactly matches the
primary key in another relation
– the names of the attributes don’t have to be the same
but must be of the same domain
Student (studno, name, hons, tutor, tutorroom, year)
Staff (lecturer, roomno, appraiser, approom)
Notation:
FK1: Student (tutor, tutorroom) references Staff (lecturer, roomno)
FK2: Staff (appraiser, approom) references Staff (lecturer, roomno)
37
Satisfaction of Foreign Key Constraints
“FK: R(A) references S(B)”
To keep things
is satisfied by an instance of R and S if
simple, we assume
for every t1 in R there is a t2 in S such that that “lecturer” alone
t1[A] = t2[B], is the primary kay
provided t1[A] is not null of Staff
Student Staff
studno name hons tutor year lecturer roomno appraiser
s1 jones ca bush 2 kahn IT206 watson
s2 brown cis kahn 2 bush 2.26 capon
s3 smith cs goble 2 goble 2.82 capon
s4 bloggs ca goble 1 zobel 2.34 watson
s5 jones cs zobel 1 watson IT212 barringer
s6 peters ca kahn 3 woods IT204 barringer
capon A14 watson
lindsey 2.10 woods
barringer 2.125 null
Foreign key constraints are also called
38
“referential integrity constraints.”
Updates May Violate Constraints …
Updates are
Insertions, Deletions, Modifications

of tuples
Example DB with tables as before:
Student (studno, name, hons, tutor, year)

Staff (lecturer, roomno, appraiser)
The DB has key and foreign key constraints
Questions:
• What can go wrong?
• How should the DBMS react?
39
Insertions (1)
If the following tuple is inserted into Student,
what should happen? Why?
(s1, jones, cis, capon, 3)
Student Staff
capon A14 watson
lindsey 2.10 woods
40
Insertions (2)

(null, jones, cis, capon, 3)
Student Staff
capon A14 watson
lindsey 2.10 woods
41
Insertions (3)
(s7, jones, cis, null, 3)
Student Staff
capon A14 watson
lindsey 2.10 woods
42
Insertions (4)
(s7, jones, cis, calvanese, 3)
Student Staff
capon A14 watson
lindsey 2.10 woods
43
Deletions (1)
If the following tuple is deleted from Student,
is there a problem? And what should happen?
(s2, brown, cis, kahn, 2)
Student Staff
capon A14 watson
lindsey 2.10 woods
44
Deletions (2)
And if this one is deleted from Staff ?
(kahn, IT206, watson)
Student Staff
capon A14 watson
lindsey 2.10 woods
45
Modifications (1)
What if we change in Student
(s1, jones, ca, bush, 2)
to
(s1, jones, ca, watson, 2) ?
Student Staff
capon A14 watson
lindsey 2.10 woods
46
Modifications (2)
And what if we change in Student
(s2, brown, cis, kahn, 2)
to
(s1, jones, ca, bloggs, 2) ?
Student Staff
capon A14 watson
lindsey 2.10 woods
47
Modifications (3)
And what if we change in Staff
(lindsey, 2.10, woods)
to
(lindsay, 2.10, woods) ?
Student Staff
capon A14 watson
lindsey 2.10 woods
48
Modifications (4)
Now, let’s change in Staff
(goble, 2.82, capon)
to
(gobel, 2.82, capon) …
Student Staff
capon A14 watson
lindsey 2.10 woods
49
Summary: Reactions to Integrity Violations
If an update violates an IC, the DBMS can
• Reject the update
• Repair the violation by

– inserting null
– inserting a default value
– cascading a deletion
– cascading a modification
50
Summary
• The relational model is based on concepts from set theory
(and logic)
• It formalises the concept of a table
• Distinguish:
– relation schema: relation name, attributes,
domains/types
– relation instance: relation over domains of attributes
• Two formalisations of tuples
– positional tuples vs. tuples as functions on attributes
• Integrity constraints: Domain cs, FDs, Keys, FKs
• Updates may violate ICs
⎯ and the DMBS has to react
51

4 DB Relmod

Uploaded by

Copyright:

Available Formats

4 DB Relmod

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

4 DB Relmod

Uploaded by

Copyright:

Available Formats

Introduction to

The Relational Data Model

ER: Tables/Relations: File organisation:

gizmo $19.99 gadgets GizmoWorks

Power gizmo $29.99 gadgets GizmoWorks

SingleTouch $149.99 photography Canon

MultiTouch $203.99 household Hitachi

• Cartesian Product of sets (also cross product):

• Function f from S to T, written f: S → T

• A relation R over S1,...,Sn is total on Si

• Analogously, a relation R over S1, ..., Sn can be functional

• How many elements can a relation over S1, ..., Sn have?

• How many functions from S to T are there?

• How many partial functions from S to T are there?

• How many relations are there over S and T

• How many relations are there over S and T

In Databases: Distinguish between

Example: Product (Prodname: Name,

Domain: Set of atomic values, that have a specific meaning

• An instance of R is a set of tuples that satisfy the schema of R

• Analogy with programming languages:

gizmo 19.99 gadgets GizmoWorks

What could be updates to a schema?

• Updates to the data happen very frequently.

Attribute values Three meanings of null values

• Tuple as elements of String x Int x String x String

• Relation = subset of String x Int x String x String

• Order in the tuple is important !

• No explicit attributes, hidden behind positions

• Fix D as the union of the attribute domains, e.g.,

• Order in a tuple is not important,

Example: t = (gizmo, 19.99, gadgets, GizmoWorks)

• Positional tuples, without attribute names

In theory and practice, both are used, e.g.,

We will switch back and forth between the two…

• Very simple model

• Foundations in logic and set theory

• Abstract model that underlies SQL, the most important

– But SQL uses “bags” while the abstract relational

4.3 Integrity Constraints

Such conditions are called integrity constraints!

• Superkeys and keys (special case of FDs)

• Referential Integrity (also “foreign key constraints”)

Integrity constraints (ICs) are part of the schema

A state of Emp that contains two tuples with

We write the desired conditions symbolically as

An instance r of R satisfies the FD if for all tuples t1, t2 in R

How many FDs are there on a given relation?

Which FDs does this instance satisfy?

• check if the remaining relation is functional in A

Why is that correct?

Check Position → Phone !

EmpID Name Phone Position

Is the white relation functional in Position?

Product (Name, Price, Manufacturer)

Book (Author, Title, Edition, Price)

• A set of attributes of a relation is a superkey if

Student (Lastname, Firstname, MatriculationNo, Major)