Introduction to Database Systems Functional Dependencies

Werner Nutt

Functional Dependencies
1. Meaning of FDs

1. 2. 3.

Meaning of FDs Keys and Superkeys Inferring FDs

Functional Dependencies
A FD is written X A or X Y Notation: X, Y, Z represent sets of attributes; A, B, C, represent single attributes X A (X determines A) is an assertion about a relation R: whenever two tuples of R agree on all the attributes of X, then they must also agree on the attribute A, or t1[X] = t2[X] implies t1[A] = t2[A] for all t1, t2 in R (analogously for X Y) Convention:We say X A holds in R Notation: No set braces in sets of attributes: just ABC, rather than {A,B,C }.
3

Example
Table Drinkers(name, addr, beersLiked, manf, favBeer)

name Janeway Janeway Spock

addr Voyager Voyager Enterprise

beersLiked Bud WickedAle Bud

manf A.B. Petes A.B.

favBeer WickedAle WickedAle Bud

Reasonable FDs to assert: 1. name addr 2. name favBeer 3. beersLiked manf

Example FDs

name Janeway Janeway Spock

addr Voyager Voyager Enterprise

beersLiked Bud WickedAle Bud

manf A.B. Petes A.B.

favBeer WickedAle WickedAle Bud

Because of name addr

Because of name favBeer

Because of beersLiked manf

FDs With Multiple Attributes

FDs with more than one attribute on the right dont increase expressivity but allow for convenient shorthands that combine FDs Example: name addr and name favBeer become name addr favBeer More than one attribute on the left may be essential. Example: bar beer price

Functional Dependencies
2. Keys and Superkeys

1. 2. 3.

Meaning of FDs Keys and Superkeys Inferring FDs

Keys of Relations
R relation, K a set of attributes of R K is a superkey for relation R if K functionally determines all of R K is a key for R if K is a superkey, but no proper subset of K is a superkey (that is, K is a minimal superkey)

Sometimes we call keys also candidate keys, to indicate they are candidates for choosing the primary key
8

Example
Drinkers(name, addr, beersLiked, manf, favBeer) We have name addr favBeer beersLiked manf Therefore, {name, beersLiked} determine all the other attributes Hence, {name, beersLiked} is a superkey
9

Example (cntd.)
Neither {name} nor {beersLiked} is a superkey: name manf doesnt hold beersLiked addr doesnt hold Hence: {name, beersLiked} is a key There are no other keys, but lots of superkeys: Any superset of {name, beersLiked} is a superkey

10

ER and Relational Keys

Keys in ER concern entities Keys in relations concern tuples Usually, one tuple corresponds to one entity, so the ideas are similar But in poor relational designs, one tuple may represent several entities, so ER keys and relational keys are different

11

Example Data
name Janeway Janeway Spock addr Voyager Voyager Enterprise beersLiked Bud WickedAle Bud manf A.B. Petes A.B. favBeer WickedAle WickedAle Bud

The relational key is {name, beersLiked} But in E/R, name is a key for Drinkers beersLiked is a key for Beers. The relation contains 2 tuples for Janeway entity 2 tuples for Bud entity.

12

Where Do Keys Come From?

1. Just assert one key K The only FDs are K A for all attributes A 2. Assert FDs and deduce the keys by systematic exploration ER model gives us FDs from entity-set keys and from many-one relationships Other FDs we may know from our domain knowledge
(no two courses take place in a room at the same time)
13

Functional Dependencies
3. Inferring FDs

1. 2. 3.

Meaning of FDs Keys and Superkeys Inferring FDs

14

Inferring FDs
We are given FDs X1 A1, X2 A2,, Xn An , and we want to know whether an FD YB must hold in any relation that satisfies the given FDs Example: If A B and B C hold, then surely A C holds

Important for design of good relation schemas

15

Inference Test
We are given a set of FDs and want to know whether YB follows from the given FDs. Test: We consider two tuples and assume they agree in all attributes of Y: Y 0000000. . . 0 00000?? . . .?

16

Inference Test (cntd.)

Apply the given FDs to infer that these tuples must also agree on certain other attributes If B is one of these attributes, then Y B is true Otherwise, the two tuples, with any forced equalities, form a two-tuple relation that satisfies the given FDs but does not satisfy Y B. This would show that Y B does not follow from the given FDs.
17

Inference Test: Example

Drinkers(name, addr, beersLiked, manf, favBeer) with 1. name addr 2. name favBeer 3. beersLiked manf Which of the FDs below follow from the given FDs? 4. addr favBeer 5. name beersLiked favBeer

18

Closure Test
An easier test is based on the concept of attribute closure Let R be a relation, F a set of FDs over R, Y a set of attributes of R. The closure of Y with respect to F, written Y+, consists of all attributes that are determined by Y, given F. Observation: Y B follows from F if and only if B Y+
19

The Closure Algorithm

Initialization: Y+ = Y Loop: Look for an FD X A in F such that X Y+ and A Y+ Add A to Y+ Until: there is no applicable FD left in F Output: Y+

This is a typical example of a fixpoint algorithm

20

Closure Algorithm: The Idea

X Y+

A new Y+

21

Example
Contracts(cno, supplno, projno, depno, partno, qty, value) Short: CSPrDPaQV A designer has found the following set of FDs: C is a key, i.e., C SPrDPaQV A project purchases each part using a single contract, PrPa C A department purchases at most one part from a supplier, SD Pa His colleague has come up with a slightly different set: A project purchases each part using a single contract, PrPa C A contract determines project, supplier and department, C PrSD SPrD is a key, SDPr CPaQV Are the findings of the second designer different from those of the first?
22

Finding All Implied FDs

We know how we can determine whether one FD follows from a set of FDs F = { X1 A1,, Xn An } Question: How can we find all such FDs? Motivation: To get a better schema, we normalize, i.e., we break one relation schema into two or more schemas.

23

Finding All Implied FDs: Example

Relation R with attributes ABCD, with FDs AB C, C D, D A. Decompose R into ABC, AD, (i.e., project onto ABC and AD) Question: What FDs hold in ABC ? Answer: Not only AB C, but also C A !

24

Why?
ABCD a1b1cd1 a2b2cd2 d1=d2 because CD a1=a2 because DA a1b1c a2b2c

comes from ABC

Thus, tuples in the projection with equal Cs have equal As: C A.

25

Projecting FDs
How can we find the FDs that hold on the projection of R? Basic Idea: 1. Start with the given FDs 2. Find all nontrivial FDs that follow from the given FDs (nontrivial = left and right sides disjoint) 3. Restrict to those FDs that involve only attributes of the projected schema
26

A Simple, Exponential Algorithm

1. For each set of attributes X, compute X+ 2. Add X A for all A X+ X 3. However, drop XY A whenever we discover X A
(Because XY A follows from X A in any projection)

4. Finally, return only FDs involving projected attributes

27

Optimizations
Suppose that Z is the set of all attributes of R. Then: + = Z+ = Z If X Y and X + = Z then Y + = Z

This can be used for optimizations!

28

Example
Relation ABC with FDs A B and B C Project onto AC A+ = ABC A B, A C
(Optimization: We do not need to compute AB+ or AC+ )

B+ = BC B C C+ = C nothing BC+ = BC nothing Resulting FDs: A B, A C, and B C. Projection onto AC: A C.

(The only FD that involves a subset of { A,C })
29

A Geometric View of FDs

We consider the set of all instances of a particular relation R That is, all finite sets of tuples that have the proper number of components. Each instance is a point in this space

30

Example: R(A,B)

{(1,2), (3,4)}

{} {(1,2), (3,4), (1,3)}

{(5,1)}

etc.

31

An FD is a Subset of Instances
For each FD X A, there is the subset of all instances that satisfy the FD Thus, we can identify an FD with a region in the space An FD is trivial if and only if it is represented by the entire space Example: A A.
32

Example: A B for R(A,B)

{(1,2), (3,4)}

AB
{} {(1,2), (3,4), (1,3)} {(5,1)}

33

Representing Sets of FDs

If each FD is a set of relation instances, then a collection of FDs corresponds to the intersection of those sets
(Intersection = all instances that satisfy all of the FDs)

Instances satisfying A B, B C, and CD A

AB BC CD A
34

Entailment of FDs
F = { X1 A1,,Xn An } set of FDs An FD Y B follows from F or is entailed by F if every instance that satisfies all FDs in F also satisfies Y B

This can be visualized: If Y B follows from the set F = { X1 A1,,Xn An }, then in the space of instances the region for Y B must include the intersection of the regions for the FDs Xi Ai . That is: Every instance satisfying all the Xi Ai surely satisfies Y B. But an instance could satisfy Y B, yet not be in this intersection.
35

Example

AB

AC BC

36

References
In preparing the lectures I have used several sources. The main ones are the following: Books: A First Course in Database Systems, by J. Ullman and J. Widom Fundamentals of Database Systems, by R. Elmasri and S. Navathe Slides: The slides of this chapter are based on slides prepared by Jeff Ullman for his introductory course on database systems at Stanford University
37