Scribd
Upload
171 views5 pages

Chap7 Practice Key

The document provides the steps to decompose a relation R = (A, B, C, D, E) with functional dependencies Fc = {A → BC, CD → E, B → D, E → A} into BCNF. The relation is decomposed into two relations: R1 = (A, B, C, E) and R2 = (B, D). It is shown that this decomposition is lossless and dependency preserving by proving the relations are each in BCNF based on the given functional dependencies.

Uploaded by

Pallavi Kumari
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
171 views5 pages

Chap7 Practice Key

The document provides the steps to decompose a relation R = (A, B, C, D, E) with functional dependencies Fc = {A → BC, CD → E, B → D, E → A} into BCNF. The relation is decomposed into two relations: R1 = (A, B, C, E) and R2 = (B, D). It is shown that this decomposition is lossless and dependency preserving by proving the relations are each in BCNF based on the given functional dependencies.

Uploaded by

Pallavi Kumari
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Database Design – Practice Questions Solution

1. Given the relation schema R = (A, B, C, D, E) and the canonical cover of its set of
functional dependencies

Fc = { A → BC
CD → E
B→D
E→A }

Compute a lossless join decomposition in Boyce-Codd Normal Form for R. Show your
steps clearly to get full marks!

Using the algorithm to decompose a relation to BCNF from figure 7.6 in


text:

1. result = {(A, B, C, D, E)}; done = false;

2. Note that we are given the canonical cover Fc in the question. This means
that we can avoid computing the closure of F and just use Fc and
Armstrong's axioms to determine if a given functional dependency is in F+.

3. (A, B, C, D, E) is not in BCNF because B → D is not a trivial dependency and


it is not a superkey for (A, B, C, D, E):

A → BC given
A → B, A → C decomposition
B → D, so A → D given, transitive
A → CD union
CD → E, so A → E transitive
A → ABCDE union of above steps
E → A, so E → ABCDE given, transitive
CD → E, so CD → ABCDE transitive
B → D, so BC → CD augmentation
BC → ABCDE transitive

Since BC is a candidate key, B cannot be a superkey. As soon as we find


one functional dependency that does not meet the criteria for BCNF, the
schema is not in BCNF.
4. B → D holds on (A, B, C, D, E),
B → ABCDE is not in F+ (i.e. can't be computed using Armstrong's Axioms
from the canonical cover Fc) and
B ∩ D = empty set, so:

result = {(A, B, C, D, E) – (A, B, C, D, E)} ∪ {(A, B, C, D, E) – D} ∪ (B, D)


result = {empty set} ∪ (A, B, C, E) ∪ (B, D)
result = {(A, B, C, E), (B, D)}

5. We determine that (B, D) is in BCNF because the nontrivial functional


dependency B → D is given, so B is a superkey for schema (B, D).

6. We determine that (A, B, C, E) is in BCNF because for the nontrivial


functional dependencies given, A → BC and E → A, both A and E are
superkeys for the schema (A, B, C, E), since A → ABCDE and E → ABCDE
from step 3.

2. Suppose you are given the following functional dependencies:

fd1: name → address, gender


fd2: address → rank
fd3: rank, gender → salary

a) Give a primary key of the relation r(name, address, gender, rank, salary). Prove your
answer formally using Armstrong's Axioms.

name is the primary key. To prove this, first prove that name is a
superkey, then prove it is a candidate key and finally a primary key.

To prove that name is a superkey:


1. name → address decomposition on fd1
2. name → gender decomposition on fd1
3. name → rank transitivity on 1. and fd2
4. name, gender → salary 3. & pseudo-transitivity on fd2
5. name → name, gender augmentation of 2.
6. name → salary transitivity on 4. and 5.
7. name → name, address, union of 1., 2., 3., 6. & augmentation w.
rank, gender, salary name

From the given FD's, it is not possible to have an FD that only has 'name'
on its right-hand side, therefore, the attribute name must be part of any
superkey. As the definition of a candidate key is a minimal superkey and
{name} is a superkey, and the empty set is not a superkey, 'name' must be
a candidate key.

Since name is a candidate key and it is the only candidate key, it is also
the primary key.

b) (5 marks) Normalize the relation r(name, address, gender, rank, salary) to 3rd normal
form, ensuring that the resulting relations are dependency-preserving and lossless-join
decompositions. Specify the primary keys in the normalized relations by underlining
them.

Use the algorithm on page 230 of the text to work through the
decomposition. Note that the fd's given form a canonical cover Fc, so one
possible decomposition is as follows:

r1(name, address, gender)


r2(address, rank)
r3(rank, gender, salary)
( or r3'(address, gender, salary) is also possible)

Mention MUST be made about why the decomposition is a lossless join


and dependency preserving. Note the text mentions that the
decomposition results in a lossless join, dependency-preserving
decomposition, so this would be sufficient.
Otherwise, a decomposition is a lossless join if, for all relations r on
schema R that are legal under the given set of functional dependency
constraints,

r=Π R1 (r) ⋈ Π R2 (r) ⋈ Π R3 (r)

If we were required to prove this, note that the universal relation r is


first decomposed into two smaller relations r1 and r2. The relation r2 is
then further decomposed to r21 and r22. If we can show that r21 and r22 is
a lossless-join, we can recover the relation r2. Then if we can show that
r1 and r2 also form a lossless-join, then we can recover the universal
relation r and the entire decomposition is a lossless join.

To show that two relations r1 and r2 form a lossless join, we must show
one of the following:

r1 ∩ r2 → r1
r1 ∩ r2 → r2

In our case, the intersection of r1 and r2 is address, so we must


determine if either:
address → name, address, gender or
address → address, rank holds.

We can get the second FD by augmenting the given FD address → rank


with address, so this is a lossless-join and we can recover (name, address,
gender, rank).

The intersection of (name, address, gender, rank) and (rank, gender,


salary) is (rank, gender). We must determine if either
rank, gender → name, address, rank, gender or
rank, gender → rank, gender, salary holds.

We can get the second FD by augmenting the given FD


rank, gender → salary with rank and gender, so this decomposition is also
a lossless-join and therefore the entire decomposition is a lossless-join.
A decomposition is dependency preserving if F - F' = {∅}, or {F1 ⋃ F2 ⋃
F3}+ = F+. This can be proven by calculating the closures as indicated. An
easier way, based on page 223 of the text, is to show that all FD’s in Fc
can be tested in a single relation in the decomposition. So,
name → address, gender can be tested in r1
address, → rank can be tested in r2
rank, gender → salary can be tested in r3

You might also like