Design Theory for Relational

Databases

DESIGN THEORY FOR RELATIONAL DATABASES 1

 Objectives

Understand concepts of:

Functional Dependencies
Normalization
Decomposition
Multi-valued Dependencies

DESIGN THEORY FOR RELATIONAL DATABASES 2

 Contents

Functional Dependencies
Rules about FDs
Key & Super-Key
Normal forms

DESIGN THEORY FOR RELATIONAL DATABASES 3

 Functional dependency
 A functional dependency: constraint between
two sets of attributes in a relation
 A set of attributes X (include A1A2…An) in R
functionally determine another attribute Y (include
B1B2…Bm), also in R, (written X → Y) if and only if
each X value is associated with precisely one Y
value
 A functional dependency A1A2…An  B1B2…Bm
holds on relation R if two tuples of R agree on all of
the attributes A1, A2, …, An then they must also
agree on all of the attributes B1, B2, …, Bm

DESIGN THEORY FOR RELATIONAL DATABASES 4

 Functional dependency

Easy to see that: the following FD is true

◦ title,year  length, genre, studioName
Exercise: How about the FD
◦ title,year  starName
title, year  startName does not hold in Movies1 relation

DESIGN THEORY FOR RELATIONAL DATABASES 5

 Functional dependency
 Review key of relation, candidate keys(alternate
keys), primary key
 Super-key
 A set of attributes that contains a key is called a super-
key
 Every super-key satisfies the first condition of a key: it
functionally determines all other attributes of the
relation
 If K is a key, L is a super key, then: K  L
 A key is also a super key

DESIGN THEORY FOR RELATIONAL DATABASES 6

 Functional dependency
Armstrong’s Axioms
Fundamental Rules: Let X, Y, Z are sets of attributes
 Reflexivity:
If X is a subset of Y, then YX
 Augmentation
If XY, then XZYZ for any Z
 Transitivity
If XY and YZ, then XZ
Additional rules: Let X, Y, Z, W are sets of attributes
 Union/Combining: if XY AND XZ then XYZ
 Decomposition/Splitting: if XYZ, then XY and XZ
 Pseudotransitivity: If X  Y and WY  Z then WX Z

Trivial FDs: right side is a subset of left side

 Ex: FLD FD
DESIGN THEORY FOR RELATIONAL DATABASES 7
 Functional dependency
 A set of FD’s S follows from a set of FD’s T
if every relation instance that satisfies all the FD’s in T
also satisfies all the FD’s in S
Two sets of FD’s S and T are equivalent if and only if S
follows from T, and T follows S

DESIGN THEORY FOR RELATIONAL DATABASES 8

 The Closure of Attributes
 The closure of a set of attributes {A1, A2, …, An} under
FD’s in S (denoted {A1, A2, …, An}+) is the set of
attributes B such that every relation that satisfies all the
FD’s in set S also satisfies A1A2…An B
 That is, A1A2…An B follows from the FD’s of S
 A1, A2, …, An  {A1, A2, …, An}+, because A1A2…AnAi
is trivial
 The Closure of Attributes

Algorithm 3.7: Closure of a set of attributes

Input: A set of attributes {A1,A2,…,An} and a set of
FD’s S
Output: The closure {A1,A2,…,An}+
1. If necessary, split the FD’s of S, so each FD in S
have singleton right side
2. Let X be a set of attributes that will become the
closure. Initialize X to be {A1,A2,…,An}
3. Repeatedly search for some FD: B1B2…Bm  C,
such that B1, B2, …, Bm are in X, but C is not
a) If such C is found, add to X, and repeat the search
b) If such C is not found, no more attributes can be added to X
4. The set X is the correct value of {A1, A2, …, An}+
 The Closure of Attributes
R(A, B, C, D)
S={A→B, B→C, C→D, D→A}
Compute {A}+ ? {B}+ ?
What are some the keys of R?
 Closing Sets of Functional Dependencies
Suppose a set of FD’s S, any set of FD’s T equivalent
to S is said to be a basis for S.
Then we say T is a basis for S
Just work with only FD’s that have singleton right
sides
A minimal basis for FD’s S is a basis B that satisfies
three conditions:
All the FD’s in B have singleton right sides
If any FD is removed from B, the result is no longer
a basis
If for any FD in B we remove one or more attributes
from the left side, the result is no longer a basis
 Closing Sets of Functional Dependencies
Example
R(A,B,C)
S={A B, A  C, B  A, B  C, C  A, C 
B, AB  C, BC  A, AC  B, A  BC, B 
AC, C  AB}
R and its FD’s have several minimal basis
 {A  B, B  A, B  C, C  B}, or
 {A  B, B  C, C  A}
 What happens to …
… a set of FD’s S of R when we project R on
some attributes?
That is, suppose a relation R with set of FD’s S,
and R1=L(R). What FD’s hold in R1?
 Projecting Functional Dependencies
To find a functional dependencies of projection,
we
Follow from S, and
Involve only attributes of R1
 Projecting Functional Dependencies
Algorithm 3.12: Projecting a Set of FD’s
Input: R, R1=L(R), S a set of FD’s that hold in
R
Output: the set of FD’s that hold in R1
Method:
 T is the set of FD’s that hold in R1. Initially, T is empty
 For each set of attributes X of R1, compute X+. Add to
T all non-trivial FD’s X → A such that A is both in X+
and an attribute of R1
 Construct a minimal basis from T
 Projecting Functional Dependencies
Algorithm 3.12: Projecting a Set of FD’s (cont)
Compute a minimal basis from T
 If there is an FD F in T that follows from other FD’s in
T, then remove F from T
 Let Y  B is a FD in T , with at least two attributes in
Y, and let Z is Y with one of its attributes removed:
 If Z  B follows from the other FD’s in T (including Y  B), then
replace Y  B by Z  B
 Repeat the above steps in all possible ways until no
more changes to T can be made
 Projecting Functional Dependencies
Two notations
(1) Closing the empty set and the set of all
attributes cannot yield a nontrivial FD
(2) If we have already knew that the closure of
some set X is all attributes, then we cannot
discover any new FD’s by closing supersets of
X
Projecting Functional Dependencies
Example: Suppose R(A,B,C,D) has FD’s A→B, B→C,
and C→D. R1=A,C,D(R). Find the FD’s of R1?
◦ Compute the closure of the singleton set
◦ {A}+={A,B,C,D}, and B is not in R1, then new FD’s A→C,
A→D
◦ {C}+={C,D}, then new FD’s C→D
◦ {D}+={D}, no new FD’s
◦ Compute the closure of the doubleton set
◦ Since {A}+ include all attributes, no care any more for
supersets of {A}
◦ {C,D}+={C,D}, no new FD’s holds in R1
◦ Finally, there are three FD’s A→C, A→D, C→D
hold in R1
◦ A→D is transitive from A→C, and C→D
◦ So, minimal basis is {A→C, C→D}
 Anomalies introduction

Careless selection of a relational database schema

can lead to redundancy and related anomalies
So, in this session we shall tackle the problems of
relational database designing
Problems such as redundancy that occur when we try
to cram too much into a single relation are called
anomalies
 Anomalies
title year length genre studioName starName
Star Wars 1977 124 SciFi Fox Carrie Fisher
Star Wars 1977 124 SciFi Fox Mark Hamill
Star Wars 1977 124 SciFi Fox Harrison Ford
Gone With The Wind 1939 231 drama MGM Vivien Leigh
Wayne's World 1992 95 comedy Paramount Dana Carvey
Wayne's World 1992 95 comedy Paramount Mike Meyers

The principal kinds of anomalies that we encounter are:

Redundancy: information maybe repeated unnecessarily in several
tuples (exp: the length and genre)
Update Anomalies: We may change information in one tuple but
leave the same information unchanged in another (exp: if we found
that Star Wars is 125 minutes long, we may change the length in the
first tuple but not in the second and third tuples)
Deletion Anomalies: If a set of values becomes empty, we may lose
other information as a side effect (exp: if we delete “Fox” from the set
of studios, then we have no more studios for the movie “Star Wars”
 Decomposition

The accepted way to eliminate anomalies is the decomposition

of relations
Decomposition of a relation R involves splitting the attributes of
R to make the schemas of 2 new relations
Definition: Given a relation R(A1,..,An), we say R is
decomposed into S(B1,..,Bm) and T(C1,..,Ck) if:
+ {A1,..,An} = {B1,..,Bm} U {C1,..,Ck}
+ S = ∏B1,..Bm(R)
+ T = ∏C1,..,Ck(R)
 Example: Decomposition
title year length genre studioName starName
Star Wars 1977 124 SciFi Fox Carrie Fisher
Star Wars 1977 124 SciFi Fox Mark Hamill
Star Wars 1977 124 SciFi Fox Harrison Ford
Gone With The Wind 1939 231 drama MGM Vivien Leigh
Wayne's World 1992 95 comedy Paramount Dana Carvey
Wayne's World 1992 95 comedy Paramount Mike Meyers

title year length genre studioName title year starName

Star Wars 1977 124 SciFi Fox Star Wars 1977 Carrie Fisher
Gone With The Wind 1939 231 drama MGM Star Wars 1977 Mark Hamill
Wayne's World 1992 95 comedy Paramount Star Wars 1977 Harrison Ford
Gone With The1939
Wind Vivien Leigh
Wayne's World1992 Dana Carvey
Wayne's World1992 Mike Meyers
 Discuss

title year length genre studioName title year starName

Star Wars 1977 124 SciFi Fox Star Wars 1977 Carrie Fisher
Gone With The Wind 1939 231 drama MGM Star Wars 1977 Mark Hamill
Wayne's World 1992 95 comedy Paramount Star Wars 1977 Harrison Ford
Gone With The1939
Wind Vivien Leigh
Wayne's World1992 Dana Carvey
Wayne's World1992 Mike Meyers

The redundancy is eliminated (the length of each film appears only

once)
The risk of an update anomaly is gone (we only have to change the
length of Star Wars in one tuple)
The risk of a deletion anomaly is gone (if we delete all the stars for
Gone with the wind, that deletion makes the movie disappear from the
right but still be found in the left)
 Decomposition: The Good, Bad and Ugly

We observed that before we decompose a relation schema

into BCNF, it can exhibit anomalies; That’s the “Good”

However, decomposition can also have some bad:

◦ Maybe we can’t recovery the original information; OR
◦ After reconstruction, the FDs maybe not hold
 Example: Loss of information after
decomposition
R R1 R2
R3
A B C A B B C
A B C
1 2 2 3 1 2 3
1 2 3
4 2 2 5 1 2 5
4 2 5 4 2 3
4 2 5
Suppose we have R(A,B,C) but neither of the
FD’s B->A nor B->C holds.
R is decomposed into R1 and R2 as above
When we try to re-construct R by Natural Join of
R1 and R2, we have: R3 = R1 X R2 (but R3 <>
R1 => We lost information)
 Example: Dependency Loss
If we check the projected FD’s in the relations of
the decomposition, can we can be sure that when
we reconstruct the original relation from the
decomposition by joining, the result will satisfy
the original FD’s?
Normal Forms
 Dependencies: Definitions
Multivalued Attributes (or repeating groups): non-
key attributes or groups of non-key attributes the values
of which are not uniquely identified by (directly or
indirectly) (not functionally dependent on) the value of
the Primary Key (or its part).

Partial Dependency – when a non-key attribute is

determined by a part, but not the whole, of a COMPOSITE
primary key.

Transitive Dependency – when a non-key attribute

determines another non-key attribute.
 1NF
1NF A relation R is in first normal form (1NF) if and
only if all underlying domains contain atomic values
only
Take the following table.

StudentID is the primary key.

Is it 1NF?
No. There are repeating groups (subject, subjectcost, grade)

How can you make it 1NF?

Create new rows so each cell contains only
one value

But now look – is the studentID primary key

still valid?
No – the studentID no longer uniquely
identifies each row

You now need to declare studentID and subject

together to uniquely identify each row.

So the new key is StudentID and Subject.

So. We now have 1NF.

Is it 2NF?
 2NF
A relation R is in second normal form (2NF) if and only
if it is in 1NF and every non-key attribute is fully
dependent on the primary key

StudentName & Address are dependent on studentID

(which is part of the key)

But they are not dependent on Subject (the other part of the key)
And 2NF requires…

All non-key fields are dependent on the ENTIRE key (studentID +

subject)
 So it’s not 2NF

How can we fix it?

 Make new tables
Make a new table for each primary key field
Give each new table its own primary key
Move columns from the original table to the new
table that matches their primary key…
 Step 1

STUDENT TABLE (key = StudentID)

 Step 2

STUDENT TABLE (key = StudentID)

SUBJECTS TABLE (key = Subject)

 Step 3

STUDENT TABLE (key = StudentID)

SUBJECTS TABLE (key = Subject)

RESULTS TABLE (key = StudentID+Subject)

 Step 3

STUDENT TABLE (key = StudentID)

SUBJECTS TABLE (key = Subject)

RESULTS TABLE (key = StudentID+Subject)

 Step 4 - relationships

STUDENT TABLE (key = StudentID)

SUBJECTS TABLE (key = Subject)

RESULTS TABLE (key = StudentID+Subject)

 Step 4 - cardinality

STUDENT TABLE (key = StudentID)

1 Each student can only appear

ONCE in the student table SUBJECTS TABLE (key = Subject)

RESULTS TABLE (key = StudentID+Subject)

 Step 4 - cardinality

STUDENT TABLE (key = StudentID)

1
SUBJECTS TABLE (key = Subject)

Each subject can only appear

ONCE in the subjects table

RESULTS TABLE (key = StudentID+Subject)

 Step 4 - cardinality

STUDENT TABLE (key = StudentID)

1
SUBJECTS TABLE (key = Subject)

1
A subject can be listed MANY
times in the results table (for
different students)
8

RESULTS TABLE (key = StudentID+Subject)

 Step 4 - cardinality

STUDENT TABLE (key = StudentID)

1
SUBJECTS TABLE (key = Subject)

1
A student can be listed MANY
times in the results table (for
different subjects)
8

RESULTS TABLE (key = StudentID+Subject)

 A 2NF check

STUDENT TABLE (key = StudentID)

1
SUBJECTS TABLE (key = Subject)

SubjectCost is only
8

dependent on the
8

primary key,
Subject
RESULTS TABLE (key = StudentID+Subject)
A 2NF check

STUDENT TABLE (key = StudentID)

1
SUBJECTS TABLE (key = Subject)

1
8

Grade is only dependent

on the primary key
(studentID + subject)
RESULTS TABLE (key = StudentID+Subject)
A 2NF check

STUDENT TABLE (key = StudentID)

1 Name, Address are only

dependent on the SUBJECTS TABLE (key = Subject)
primary key 1
(StudentID)
8

RESULTS TABLE (key = StudentID+Subject)

STUDENT TABLE (key = StudentID)

1
So it is SUBJECTS TABLE (key = Subject)

2NF! 1
8

But is it
8

RESULTS TABLE (key = StudentID+Subject)

3NF?
 3NF

A relation R is in third normal form (3NF) if and

only if it is in 2NF and every non-key attribute is
non-transitively dependent on the primary key.
An attribute C is transitively dependent on attribute A if
there exists an attribute B such that: A->B and B->C

Note that 3NF is concerned with transitive

dependencies which do not involve candidate keys. A
relation with more than one candidate key will clearly
have transitive dependencies of the form: primary_key
-> other_candidate_key -> any_non-key_column
 A 3NF check

STUDENT TABLE (key = StudentID)

Oh oh… 1
SUBJECTS TABLE (key = Subject)

What?
8

RESULTS TABLE (key = StudentID+Subject)

 A 3NF check

STUDENT TABLE (key = StudentID)

1
HouseName is
SUBJECTS TABLE (key = Subject)
dependent on both
1
StudentID +
HouseColour
8

RESULTS TABLE (key = StudentID+Subject)

 A 3NF check

STUDENT TABLE (key = StudentID)

1
Or HouseColour is
SUBJECTS TABLE (key = Subject)
dependent on both
1
StudentID +
HouseName
8

RESULTS TABLE (key = StudentID+Subject)

 A 3NF check

STUDENT TABLE (key = StudentID)

1
But either way,
non-key fields are SUBJECTS TABLE (key = Subject)

dependent on MORE 1

THAN THE PRIMARY

KEY (studentID)
8

RESULTS TABLE (key = StudentID+Subject)

 A 3NF check

STUDENT TABLE (key = StudentID)

1
And 3NF says that
non-key fields must SUBJECTS TABLE (key = Subject)

depend on nothing 1

but the key

8

RESULTS TABLE (key = StudentID+Subject)

 A 3NF check

STUDENT TABLE (key = StudentID)

1
WHAT DO SUBJECTS TABLE (key = Subject)

WE DO? 1
8

RESULTS TABLE (key = StudentID+Subject)

Again, carve off the offending fields

SUBJECTS TABLE (key = Subject)

8

RESULTS TABLE (key = StudentID+Subject)

 A 3NF fix

SUBJECTS TABLE (key = Subject)

8

RESULTS TABLE (key = StudentID+Subject)

 A 3NF fix

8
1
1

SUBJECTS TABLE (key = Subject)

8

RESULTS TABLE (key = StudentID+Subject)

 A 3NF win!

8
1
1
8

1
SUBJECTS TABLE (key = Subject)
RESULTS TABLE (key =
StudentID+Subject)

Or…
 The Reveal

Before…

After… 1

8
1
1
8

SUBJECTS TABLE (key = Subject)

RESULTS TABLE (key = StudentID+Subject)

 3NF
No transitive dependencies

Table contains data from an embedded entity with non-key attributes.

TABLE SUB-TABLE

TABLE ??

SUB-TABLE

BCNF is the same, but the embedded table may involve key attributes.
 BCNF

A relation R is in BCNF if and only if: Whenever there is

a Non-Trivial FD
A1A2..An -> B1B2..Bm for R, it is the case that:
{A1,..,An} is a super-key for R

That is: the left side of every Non-Trivial FD must be a

super-key
 Example: BCNF or not

title year length genre studioName starName

Star Wars 1977 124 SciFi Fox Carrie Fisher
Star Wars 1977 124 SciFi Fox Mark Hamill
Star Wars 1977 124 SciFi Fox Harrison Ford
Gone With The Wind 1939 231 drama MGM Vivien Leigh
Wayne's World 1992 95 comedy Paramount Dana Carvey
Wayne's World 1992 95 comedy Paramount Mike Meyers

The above relation is not in BCNF because:

Consider the FD:
{title,year} -> {length, genre, studioName}
We know that {title, year} is not a super-key
 Example: BCNF or not

title year length genre studioName

Star Wars 1977 124 SciFi Fox
Gone With The Wind 1939 231 drama MGM
Wayne's World 1992 95 comedy Paramount

The above relation is in BCNF because:

{title,year} -> {length, genre, studioName}
And: neither title nor year by itself functionally
determines any of the other attributes
 Example: BCNF or not

title year starName

Star Wars 1977 Carrie Fisher
Star Wars 1977 Mark Hamill
Star Wars 1977 Harrison Ford
Gone With The1939
Wind Vivien Leigh
Wayne's World1992 Dana Carvey
Wayne's World1992 Mike Meyers

The above relation is in BCNF because: it has no

Non-Trivial FD
 Differences between BCNF and 3NF

3NF Boyce-Codd
a nontrivial functional a nontrivial functional dependency
dependency: X => A holds X => A holds in R, then:
in R, either a) X is a superkey of R
(a) X is a superkey of R, or
(b) (b) A is a prime attribute of
R.

Note:A functional dependency X => Y is a full functional

dependency if removal of any attribute A from X means that the
dependency does not hold any more; A partial functional
dependency is not a full functional dependency
 BCNF decomposition algorithm
(self studying)
Input: A relation R with a set of FD’s F
Output: A BCNF decomposition of R with lossless join
Method:
◦ At each step compute the key for the sub-relation R
◦ if not in BCNF, pick any FD X->Y which violates
◦ break the relation into 2 sub-relations
◦ R1(XY)
◦ R2(S - Y)
◦ this has a lossless join
◦ project FD's onto each sub-relation
◦ continue until no more offending FD's
 3NF decomposition algorithm
– self studying
Input: A relation R with a set of FD’s F
Output: A decomposition of R into a collection of
relations, all of which are in 3NF. This decomposition has
a lossless join and dependency-preservation.
Method:
◦ Find minimal basic for F, say G.
◦ ∀ X-A ∈ G, use XA as the schema of one relations in
the decomposition.
◦ If none of the sets of relations from Step 2 is a super
key for R, add another relation whose schema is a key
for R.
 Summary 1
Decompose a relation into BCNF is a solution for eliminating
anomalies
But BCNF can cause information loss and dependency loss
3NF is a relax solution of BCNF that keep loss-less join and
dependency-preservation properties
 Summary 2:
2NF 3NF Boyce-Codd
every nonprime a nontrivial functional a nontrivial functional
attribute A in R is not dependency: X => dependency X =>
partially A holds in R, A holds in R, then:
dependent on any key either a) X is a superkey of
of R (a) X is a superkey of R
R, or
(b) (b) A is a prime
attribute of R.

Note:A functional dependency X => Y is a full functional dependency if

removal of any attribute A from X means that the dependency does not
hold any more; A partial functional dependency is not a full functional
dependency