5 Specifying Aggregate Data:

Arrays
Arrays
COL 100 - Introduction to Computer
p
Science
II Semester 2015-2016
Department of Computer Science and Engineering
Indian Institute of Technology Delhi

Aggregate Data
Data Types seen so far:
int, char, float, double,...
These are Scalar data types

Aggregate data: collection of variables

of same type: ARRAYS
of different types: struct/class/union/...

Jan 2016

P. R. Panda, IIT Delhi

The Array Data Type

Representing
p
g Collection of
Data
ordered set of elements
all of same type
position has significance
(compare: set)
e.g., marks scored by 100
students

Jan 2016

P. R. Panda, IIT Delhi

10

-2

Declaring an Array
Declaration has 4 components
specifier (optional)
base type
declarator
name
declarator operator ( [ ], etc.)

initialiser (optional)

extern int Marks [4] = {5,10, 15, 20};

Specifier
Jan 2016

Base type

Declarator Initialiser

P. R. Panda, IIT Delhi

Array Access
Individual elements can
be accessed
READ or WRITTEN

A
Array access has
h ttwo
components
array
y name
index (starts at 0)

X = Y; and X = 0; not
allowed
ll
d ffor arrays iin
C/C++

Jan 2016

10

-2

X[2] = 5; // Array on LHS => WRITE

X[3] = X[3] + 1; // Array on RHS => READ

P. R. Panda, IIT Delhi

Storing an Array in Memory

Array size must be constant
int X[5]; // X[0] to X[4]
int Y[size]; // ?

We need to know how much

space it occupies
Staticallyy (at
( COMPILE-time))
...before execution

Jan 2016

P. R. Panda, IIT Delhi

10

-2

int X[5];

Storing an Array in Memory

Elements are stored in sequence
((continguous)
g
)
How do we implement A[i] = b;?
First compute address of A[i]
Need to know bytes occupied by As
elements
This is Hardwdare
Hardwdare-dependent
dependent
sizeof operator

Store value of b at this address

Jan 2016

P. R. Panda, IIT Delhi

int A[4];
Address Memory
(Byte)
0
1

100
?
?
?

A[0]
A[1]
A[2]
A[3]

Array Index Range

Index expression must be within
RANGE (0 to N-1 for A[N])
What if we have A[4] = 7 or A[-1]
A[ 1] = 7 ?
When is the error detected?

Safety vs.
vs Efficiency
Native types targeted at EFFICIENCY
If we need safety, we should build our
own types

Jan 2016

P. R. Panda, IIT Delhi

int A[4];
Address Memory
(Byte)
0
1

100
?
?
?

A[0]
A[1]
A[2]
A[3]

Initialisation

Local variables have garbage

(undefined) value until explicit
initialisation
Variables can be initialised when
they
y are declared
Variables in global scope must
be initialised with constant
expression
e
p ess o

int i = 0, j = 1;

equivalent
int ii, j;
i = 0;
j = 1;

initialisation done once, before

program begins

Jan 2016

P. R. Panda, IIT Delhi

int i;
cout << i;
prints garbage

int x = 0;
int y = x; /* error */
main ( ) {
int p = y;
int q = p + 1; /* OK */
}
9

Initialising Arrays

List of initialisers in {...}

Array size can be inferred from list size
Too few initialisers: rest are zero
Too many: error
int A [ ] = {10, 20, 30, 40};
equivalent to:
int A [4] = {10, 20, 30, 40};

int A [4] = {10, 20, 30, 40};

abbreviation for:
i t A [4];
[4]
int
A [0] = 10;
A [1] = 20;
A [2] = 30;
A [3] = 40;

int A [4] = {10, 20};

equivalent to:
int A [4] = {10, 20, 0, 0};

In general, INITIALISATION is NOT the

same as ASSIGNMENT (details later)
Jan 2016

int A [2] = {10, 20, 30}; /* error */

P. R. Panda, IIT Delhi

10

Strings in C++
string data type: array of
characters
Terminated by \0
Manipulation:

Jan 2016

Copying a string (s = t)
Comparing strings (s == t)
Extracting characters (s[1] == b)
Concatenating/Appending (s + t)
others...later

P. R. Panda, IIT Delhi

string s = abc;
abc ;

a
b
c
\0

11

Operations with Arrays

Sum of Array elements
0
1
2
3
4

Initialising an Array
for (i = 0; i < 5; i++)
A [i] = 0;

0
A

1
0

2
0

for (i = 0; i < 5; i++) {

t = A [i];
A [i] = A [4 - i];
A [4 - i] = t;
}
Jan 2016

sum = 0;
for (i = 0; i < 5; i++)
sum = sum + A [i];

5
sum = 25

What does this loop do?

A

P. R. Panda, IIT Delhi

12

Searching in an Array
Given array A[N] and value V, is V
present in the Array? 0 1 2 3
A

fo nd = false;
found
for (i = 0; i < N; i++)
if (A[i] == V) {
found = true; // V is present
break;
}
Jan 2016

H many comparisons?
i
?
How
best case
average case
worst case

P. R. Panda, IIT Delhi

13

Prime Number Algorithm

Avoiding
multiples of 2
and 3
Avoiding
duplicate loop
bodies

Jan 2016

x = 6; root_n = sqrt (n);

prime = true;
while (x < root_n) {
if ((n % ((x+1)) == 0))
{prime = false; break}
if (n % (x+5) == 0)
{prime = false; break}
x = x + 6;
}
cout << prime ? Prime
Prime : Not
Not Prime
Prime;;

P. R. Panda, IIT Delhi

14

Prime Number Algorithm

Avoiding duplicate loop bodies
x = 6; root_n = sqrt (n);
prime = true;
while
hile ((x < root_n)
root n) {
if (n % (x+1) == 0)
{prime = false; break}
if (n % (x+5) == 0)
{prime = false; break}
x = x + 6;
}
cout << prime ? Prime : Not Prime;
Jan 2016

P. R. Panda, IIT Delhi

x = ???;
root n = sqrt (n);
root_n
(n)
prime = true;
while (x < root_n) {
x = ???
if (n % x == 0)
{prime = false; break}
}

15

Prime Numbers:
Seive of Eratosthenes
Generalise earlier strategy
Avoid multiples of ALL smaller primes
Algorithm:
For 2 <= x <= n
...

Jan 2016

P. R. Panda, IIT Delhi

16

Algorithm

Array of [2..n]
Initialise:

All numbers UNCROSSED

x=2

WHILE ((x <= n))

Jan 2016

Proceed to next uncrossed number x. This is a

PRIME
CROSS all multiples of x
P. R. Panda, IIT Delhi

17

Problem with Algorithm

Too much space
Need array of size ~ n

Let us keep array of prime numbers

Jan 2016

P. R. Panda, IIT Delhi

18

Maintaining Array of Primes

Append to p when we find new prime
p [0] = 2;
p [1] = 3;
p [2] = 5;
limit = 2; plim_square = 5*5;
if (x >= plim_square)
plim square) {
limit = limit + 1;
plim_squa
p
square
e = p[
p[limit]
t] * p[limit];
p[ t];
}

Jan 2016

P. R. Panda, IIT Delhi

Array p of Primes
0 1 2 3 4
2

11

limit last_prime

19

Checking for Prime

Check if x is divisible by any number in p
j = 2;
prime = true;
while (prime && j < limit) {
if (x % p[j] == 0)
prime = false;
j = j + 1;
}

Jan 2016

P. R. Panda, IIT Delhi

Array p of Primes
0 1 2 3 4
2

11

limit last_prime

20

Too Many Divisions

MOD operator (%) is expensive
Same as division
Try reducing divisions

Consider a given p[j] = 31

0

9 10 11

11 13 17 19 23 29 31 37

Suppose we just checked x = 1643 (multiple of 31)

After
Aft thi
this, which
hi h multiple
lti l off 31 will
ill show
h
up
as a candidate x?
Jan 2016

P. R. Panda, IIT Delhi

21

Next Expected Multiple

No need to check any x between 1643 and
1643+62 for divisibility by 31
For every p[j], we can store the next relevant
multiple
Compare x with p[j]
if x is greater
greater, update the multiple
if x is smaller?
if x is equal?

Jan 2016

P. R. Panda, IIT Delhi

22

Anticipating the Next Multiple

New array multiple[]
multiple[j] contains
the next anticipated
multiple of p[j]
if not reached, then x is
nott divisible
di i ibl b
by p[j]
[j]

Jan 2016

j = 2;
prime = true;
while (prime && j < limit) {
while (multiple [j] < x)
multiple [j] += p[j] * 2;
if (x == multiple [j])
prime = false;
j = j + 1;
}

P. R. Panda, IIT Delhi

23

Overall Algorithm
Initialise
While (x < n)
get next x
if (x >= square of limit prime)
Update limit

loop through prime multiples

Update multiples
Prime
Pi
T
Test: compare with
i h current multiple
li l

if (x is Prime)
Append x to array
Jan 2016

P. R. Panda, IIT Delhi

24

Searching a Sorted Array

Assume we have a sorted array
values are in increasing (decreasing) order

Search for a given value V

A

Possible to search using less than 5

comparisons (worst case)

Jan 2016

P. R. Panda, IIT Delhi

25

Binary Search
mid
L

Divide
Divide and
Conquer strategy
at every stage, we
reduce the size of
the problem to half
the earlier stage

Strategy: Compare
with the middle
element of current
range, and
eliminate half of
the range

Jan 2016

// Algorithm Binary Search

int A[N], V, low, high; // Search for V in A[ ]
...
low = 0; high = N-1;
do {
// Calculate mid position k of remaining array
mid = (low + high)/2;
if (A[mid] == V) break; // found!
if (V > A [mid])
low = mid + 1;
else
high = mid 1;
} while (low <= high);
cout << (A[mid] == V? yes : no);

P. R. Panda, IIT Delhi

26

Sorting an Array
Rearranging array contents in
increasing or decreasing order
0
A

1
9

2
6

Sort in increasing order

A
Jan 2016

P. R. Panda, IIT Delhi

How do we sort?
27

Simple Sorting Algorithm

A[0]

A[i] A[i+1]

A[N-1]

for (i = 0; i < N; i++) {

k = position of min. element
between A [i] and A [N-1]
Swap A [i] and A [k]
}

Jan 2016

P. R. Panda, IIT Delhi

28

Simple Sorting Algorithm

A[0]

for (i = 0; i < N; i++) {

k=p
position of min. element
between A [i] and A [N-1]
Swap A [i] and A [k]
}

A[i] A[i+1]

A[N-1]

k = i;;
for (j = i+1; j < N; j++)
if (A[j] < A [k])
k = j;

t = A [ i ];
A [ i ] = A [ k ];
A [ k ] = t;
Jan 2016

P. R. Panda, IIT Delhi

29

Matrix Multiplication
Input
I
t - 44
4 4 matrices
ti
A and
dB
Output
p - Product matrix C = A B
A

C
1,j

i,1 i,2 i,3 i,4

2,j
3,j

i,j

4,j

Ci,j = Dot Product (Row i of A, Column j of B)

Jan 2016

P. R. Panda, IIT Delhi

30

Matrix Multiplication Formulation

A
1

4
1

2
3

i,j

j
k=1

k=2

k=3

k=4

Ci,j = Ai,k Bk,j

k=1

Jan 2016

P. R. Panda, IIT Delhi

31

Representing a 22-Dim Matrix

0

Declare 2-dimensional array:

y
float A [4] [4];
ARRAY of ARRAYS
Both dimensions start at 0 and
end at 3

Generalises to n dimensions
int B [3][4][6];
3-dim array.
y Can represent
p
3x4x6
matrix.

Jan 2016

P. R. Panda, IIT Delhi

0 0,0 0,1 0,2 0,3

1 1,0 1,1 1,2 1,3

2 2,0 2,1 2,2 2,3
, 3,1
, 3,2
, 3,3
,
3 3,0

Ai,j stored
t d in
i
array location A[i][j]
32

Dot Product Algorithm

Range 0-3 instead of 1-4

A
0

Specification

i,0 i,1 i,2 i,3

0,j

1,j

2,j

3,j

k=1

=
Jan 2016

k=2

ij
i,j

Ci,j
i j = Ai,k
i k Bk,j
kj
k=0

Implementation

j
k=0

n-1

c[i][j] = 0;
for (k = 0; k < 4; k++) {
c[i][j] += A[i][k] * B[k][j];
}

k=3

C
P. R. Panda, IIT Delhi

33

Matrix Multiplication Algorithm

A
0

i,0 i,1 i,2 i,3

0,j

1,j

2,j

3,j

j
k=0

k=1

k=2

k=3

n-1

i,j
ij

Ci,j = Ai,k Bk,j

main()
{
float a[4][4], b[4][4], c[4][4];
.../* read values into arrays */
for (i = 0; i < 4; i++)
for (j = 0; j < 4; j++) {
c[i][j] = 0;
for (k = 0; k < 4; k++) {
c[i][j] += A[i][k] * B[k][j];
}
}
}

k=0

C
Jan 2016

DotProduct
P. R. Panda, IIT Delhi

34

Storing MultiMulti-dimensional Arrays

ROW-MAJOR order

Memory

elements of a row are

contiguous
ti
iin memory
followed in C/C++

Alternatively,
COLUMN-MAJOR
order
elements of a column
are contiguous

0 0,0 0,1 0,2 0,3

1 1,0 1,1 1,2 1,3

2 2,0 2,1 2,2 2,3
3 3,0 3,1 3,2 3,3

Jan 2016

P. R. Panda, IIT Delhi

A [0][0]
A [0][1]
A [0][2]
A [0][3]
A [1][0]
A [1][1]
A [1][2]
A [1][3]
A [2][0]
A [2][1]
A [2][2]
A [2][3]
A [3][0]
A [3][1]
A [3][2]
35
A [3][3]

Accessing MultiMulti-dimensional
Arrays
y

Addr_A

y = A[i][j]
First compute address of
A[i][j]
Read data at this address

0 0,0 0,1 0,2 0,3

1 1,0 1,1 1,2 1,3

2 2,0 2,1 2,2 2,3
3 3,0 3,1 3,2 3,3

int A[N][N];
Address of A[i][j] =
Addr_A + sizeof(int) x (N x i + j)
Jan 2016

P. R. Panda, IIT Delhi

A [0][0]
A [0][1]
A [0][2]
A [0][3]
A [1][0]
A [1][1]
A [1][2]
A [1][3]
A [2][0]
A [2][1]
A [2][2]
A [2][3]
A [3][0]
A [3][1]
A [3][2]
36
A [3][3]

Can we improve the MM code?

main()
{
float a[4][4], b[4][4], c[4][4];
.../* read values into arrays */
for (i = 0; i < 4; ii++))
for (j = 0; j < 4; j++) {
c[i][j] = 0;
f (k = 00; k < 44; k++)
k ){
for
c[i][j] += A[i][k] * B[k][j];
}
}
}
Jan 2016

P. R. Panda, IIT Delhi

37

Improve the MM code

Is the += useful?
avoids repeated computation
off the
th address
dd
off c[i][j]
[i][j]
in general, some complex
analysis is necessary to
establish that two
expressions c[i][j] are
identical

Address of c[i][j] unchanged in

inner loop
Replace c[i][j] by scalar
avoids multiple address
computations and memory
accesses

Jan 2016

main()
{
float a[4][4], b[4][4], c[4][4];
.../* read values into arrays */
for (i = 0; i < 4; i++)
for (j = 0; j < 4; j++) {
c[i][j]
[ ][j] = 0;;
for (k = 0; k < 4; k++) {
c[i][j] += A[i][k] * B[k][j];
}
}
}

P. R. Panda, IIT Delhi

38

Better Version
main()
{
float a[4][4], b[4][4], c[4][4];
.../* read values into arrays */
for (i = 0; i < 4; ii++))
for (j = 0; j < 4; j++) {
c[i][j] = 0;
f (k = 00; k < 44; k++) {
for
c[i][j] += A[i][k] * B[k][j];
}
}
}
Jan 2016

main()
{
float a[4][4], b[4][4], c[4][4], t;
.../* read values into arrays */
for ((i = 0;; i < 4;; i++))
for (j = 0; j < 4; j++) {
t = 0;
f (k = 00; k < 4;
for
4 k++) {
t += A[i][k] * B[k][j];
}
c[i][j] = t;
}
}

P. R. Panda, IIT Delhi

39

Which is Better?

Functionally equivalent code may not be equally efficient!

for (i = 0; i < N; i++)
for (j = 0; j < N; j++)

for (i = 0; i < N; i++)

for (j = 0; j < N; j++)

A[i][j] = 10;

Jan 2016

A[j][i] = 10;

P. R. Panda, IIT Delhi

40